ESV.Twelfth.Int.Conf.Volume.ONE.PART.TWO

diagnosis and preventive maintenance. Both means will 

gain increasing importance in future as they help to reduce 

down times. And it is hardly exaggerated to say that trucks 

which suffer from the gap in the electronic data processing 

Improving Safety in Heavy Trucks 

Thage Berggren, 

President and Chief Executive Officer, 

Volvo GM Heavy Truck Corporation, 

United States 

Glad to be here to represent Volvo at this panel discussion 

about heavy truck safety. 

Hope you realize you're in a country where safety is a 

mania second only to worshiping the sun. It's not safe to be 

lukewarm about safety in this country. 

So, you're in the right place at the right time-safety has 

never been a hotter topic in the heavy trucking industry. 

Electronic$ is also a hot topic, and I'm going to discuss 

how electronic devices may be able to lower the accident 

rate and help truck drivers do their job better. 

The following electronic systems are available or are in 

the process of development. 

AGS- Automatic gear shifting 

HUD-Head-up display 

DPS- Driver positioning $y$tem 

Driver seat with memory 

Steering wheel with memory 

Extemal minors with memory 

ALI- Axle loading indication 

MTS-Mobile telephone system 

Yellow pages 

Positioning 

Route guidance 

Navigation 

Personal alert 

RDS- Radio data and information system 

GPS- Satellite positioning system 

ELF- Electronic line following 

OTS- On-board traffic signs 

TWS-Truck work station 

Function programs 

Transport programs 

Communication programs 

Navigation, maps, etc. 

Pleasure Programs 

Technology NOT available commercially or LACKING 

technology: 

222 

ECC- Enhanced cruise control 

Keeps safe distance under varying speeds 

and conditions 

within the information chain fleet-vehicle-customer will be 

equipped with electronic infrastructure in future, so as to 

enable an unintemrpted flow of information parallel to 

goods transport. 

ADS- Automatic Driving System 

Convoys 

Merging, passing, exiting 

Avoiding collisions 

That's a rather long and impressive list. 

But the REAL question is what technologies will be the 

most useful, the most effective, AND the most appealing to 

customers? 

And what electronic equipment makes sense on a heavy 

truck? 

After their cost, there are three things to consider about 

electronic safety devices-reliability, reliability, and reliability. 

I think we all leamed about the importance of reliability 

in safety-related equipment from the antilock brake fiasco 

ofnearly l0 years ago. 

My first topic is electronic equipment to monitor the 

driver. 

Available devices and technology include: 

MSG-Maximum speed governor 

ET- Electronictachograph 

TRS- Trip recording systems 

This technology has varying degrees of merit and 

usefulness. 

In the end, however, no electronic or mechanical warning 

device is as effective as the waming system already built 

into the human body. 

I have my doubts about any alarm bell being able to 

change driver reactionTbehavior. 

Shouldn't take too much decision-making responsibility 

away from driver. 

If we could find an electronic device that kept a driver 

alert, should it warn the driver? Or the dispatcher? Or should 

it stop the truck? This is not an easy decision. 

Another hot topic today is for electronic devices to provide 

better feedback to the driver on vehicle dynamic performance*braking, 

skidding, roll, etc. 

Technology available includes: 

ABS-Lock-free brakes, tractor-to-trailer 

compatible 

ASC- Anti-spin traction control 

FIS- Fault indication systems 

DIS- Driver information sy$tem, including 

computer

The increase in safety when applying electronics is 

caused by the fact that the driver is relieved of control 

functions which can be expressed by explicit algorithms. By 

way of example this is demonstrated in figure 2. In the 

middle it is indicated that in the conventional vehicle the 

driver himself acts as a controller who compares the target 

and actual data and continuously derives from them the 

adapted positioning commands. With the application of 

electronic control the driver's part is restricted to act as a 

pilot as indicated in the lower paft who prescribes the target 

data and leaves the execution as well as the reaction to 

disturbing factors to the electronic control. It is immediately 

evident that thus a capacity reserve is built up on the part of 

the driver, which is either available for superior guiding 

functions or helps to improve reaction and alertness in unforeseeable 

situations. 

One objection against the application of electronics in 

vehicles is the supposed unreliability. Two tendencies are to 

be seen which allow the conclusion that electronic control 

will equal the high degree of reliability which the other 

vehicle components have reached already. Within the field 

of electronics as such the development offers increasing 

possibilities to provide additional routines for recognizing 

and correcting failures already in the design phase. Besides, 

electronic systems are increasingly becoming integrated 

components of the vehicle whose requirements concerning 

housing or $urounding conditions can be considered in the 

design phase of the vehicle already. 

For a representative electronic system the anti-lock braking 

system, sound statistical data as to reliability are available. 

This system has been under observation for the longest 

period of time as it was introduced for commercial vehicles 

in '83. If one examines the causes of the individual failures, 

the wiring hamess and the connection plugs are found to be 

the major trouble sources (figure 3). This statistic also includes 

one-shot troubles which take place under irreproducible 

conditions. From this it is evident that increasing 

the reliability means primarily increasing the soundness 

of the electro-mechanical connections. 

Irt the beginning it was pointed out that the purpose of 

electronics is to act as a controller in the place ofthe driver. 

As a precondition for this it is necessary to get a better 

insight in the human control mechanism. Practically it is 

hard to prove as the test conditions cannot be reproduced 

with sufficient accuracy and above all observations beyond 

the dynamic limit or in critical driving situations can only be 

performed to a limited degree. In this situation the Mercedes-Benz 

Driving Simulator in Berlin constitutes an efficient 

development tool (figure 4). Based on hardware which 

reaches the boundaries of the technical possibilities available 

today all sensual impressions----optical acoustical or 

mechanical-are simulated as realistically as possible so 

a$ to te$t the reaction of the driver under the desired 

conditions. 

Of the many interesting studies regarding vehicle dynamics 

or driver behaviour under stress conditions, some studies 

of the appropriate failure mode are mentioned in connection 

with the development of 4-wheel steering or fully active 

suspension control. The purpose was to determine which 

loss ofcontrol under certain critical driving conditions can 

be compensated for by the driver and which intrinsic safe 

performance is needed. 

A study which seems to be rather curious at a first glance 

should also be mentioned, namely to gain a deeper understanding 

of the drivers of tank trucks. The simulation of a 

real accident during which a car had a partially offset headon 

collision with a tank truck was intended to prove whether 

the driver of the tank truck would have had a chance to 

evade the car by a sufficiently quick steering manoeuvre 

without endangering himself by leaving the road. The influence 

of the moving liquid and the resultant shifting of the 

center of gravity was of major importance since it was 

assumed that the fuel tank and its chambers were filled to 

only two-thirds. 

Figure 5 shows the real accident situation above, in the 

middle and in the lower part two $cenes in the driving 

simulator which illustrate how the accident happened as 

seen by the truck driver. Figure 6 shows the steering performance 

of 3 characteristic drivers during the collision. Additionally 

the truck speed and the maximum steering angle 

speed are indicated. Driver I was the only one out of 20 who 

managed to avoid both colliding with the car and leaving the 

road. Driver 2 as a representative of the average succeeded 

only in avoiding the collision, but he came off the road. 

Finally driver 3 shows the pattem of a driver who maintains 

the highest speed. Although he generated the highest steering 

angle speed he did not succeed in staying on the road. 

This example shows very clearly that certain studies cannot 

be made at all without the driving simulator. An evalua' 

tion of statistical data-a method which is often used to get a 

deeper understanding of specific behaviour pattems

SLC- Air suspension leveling conrrol 

ASS- Active suspension systems 

Another area of interest are the electronic and conventional 

devices that show potential for improved direct and 

indirect driver field of view. 

Probably the most important device is larger windshields. 

Increasing glass area, changing to curved windshields and 

making corner posts slimmer have improved driver visibility 

greatly. 

But increasing the greenhouse area means decreasing the 

amount of metal in the cab. This makes protecting the driver 

more difficult-and leads to the conclusion that a steel cab 

is the best way to make the driver more secure. 

Improved aerodynamics also increase visibility by producing 

sloping hoods and rounded fenders. 

Another conventional solution is electric mirrors-mirrors 

adjustable by a flick of a switch that produce a wider 

field of vision for the driver. Devices that keep these minors 

clean and defogged also are impofiant to safety. 

Head-up display technology also improves the driver's 

field of view. With readings projecred on rhe windshield, he 

can monitor instruments and still keep his eyes on the road. 

Also important is television monitors, which are especially 

good for eliminating blind spots. 

A final item to consider is if a too-comfortable cab affect 

a driver's alertness. 

Unfortunately, this is not yet a problem. 

We've come a long way in improving cab design and 

environment but there's a long way to go. 

Trucks must b€ designed to provide a certain level of 

creature comforts for drivers. 

This is what I call the man-machine concept. 

Cabs must be designed so they are comfortable, convenient 

and logical. They must be designed to eliminate any 

chances of driver error that might lead to accidents. 

Increased cab volumes in the U.S.-a direct result of the 

1982 Surface Transportation Act--ended restrictions on the 

overall length of tractor-trailer combinations. 

These roomier cabs made driving easier, more convenient. 

Cabs should be designed to require a certain level of 

effort, a certain level of participation from the driver. 

There must be enough for a driver to do to keep him alert 

but not too much to do to distract him from driving safely. 

Too much automation could induce boredom. 

Sound frequency research have made progress in finding 

the "right" 

frequencies to keep drivers alen and reducing 

noise-induced fatigue. 

For example, low-frequency sounds puts drivers to sleep, 

while high-frequency sounds cause fatigue. So the effective 

cab design must find a happy sound frequency medium. 

Vibrations can cause discomfort and fatigue. 

The body tolerates vertical vibration better than horizontal 

vibration. Suspensions and seat designs can reduce vertical 

vibrations and increase comfort. 

But further advances in braking and acceleration have to 

be refined to reduce horizontal vibration. But research has 

determined that you shouldn't isolate the driver from the 

feel of the road too much. 

While much is possible to reduce accidents and make the 

operation of trucks safer, I'd like to close by making a few 

important points. 

L Accident prevention is the key safety goal of truck 

manufacturers. Volvo Truck's Accident Investigation Team 

has been in operation for 25 years studying how engineering 

and manufacturing improvements can help avoid accidents. 

2. Safety has little impact of cost lF safety features are 

designed into the truck rather than being added on. 

3. There are a number of electronic devices that could 

help prevent truck accidents-IF used right. We must establish 

our priorities in adopting safety features and devices in 

trucks. 

While we can't put a price tag on human life, there is a 

price tag on commercial vehicles. 

4. It is the markerplace that will ultimately determine 

WHICH safery technology it will accept and WHEN it will 

accept them. 

In the meanwhile, we in manufacturing will continue to 

protect drivers with current safety devices: 

Three-point safety belt 

Cab safety cage principle 

Steering wheel redesign 

Pedal area redesign 

Better dash layout 

Reinforced firewalls 

Better-protected brake hoses 

We believe in gradual, realistic advances in engineering 

and govemment regulation that results in significant advances 

in improving equipment and reducing accidents. 

223

Safety Situation of Heavy Thuck Occupants in TFaffic Accidents 

Written Only Papers 

Dietmar Otte, 

Traffic Accident Research Unit, 

Medical University Hannover (FRG) 

Hermann Appel, 

Institut of Vehicle Engineering, 

Technical University Berlin (FRG) 

Abstract 

Two hundred sixteen accidents of trucks with the permissible 

gross-weight of ovet 7 tons were analyzed- These 

accidents were documented by a scientific research team, 

directly at the scene of the accident. The injury situation of 

the truck passengers and the injury risk from these trucks for 

other traffic participants is described. A high safety standard 

for truck passengers is apparent, but only in collision 

with accident parties of lower weight categories. 

Introduction 

The injury risk for truck passengers increases in 

collisions with parties of higher weight categories. The 

correlation between AIS and mass factor is also analyeed- 

Feasible measures for the minimization of accident 

con$equences were demonstrated, especially for 

modification of the outside truck structures. 

Within the total picture of all traffic accidents, trucks are 

noted for tittle accident participation and also for less injury 

severity for truck passengers. Truck passengers represent 

only 2.3 percent of the number of all traffic victims (figure 

l). On an average only 0.17 percent of injured truck 

passengers can be assigned to each truck accident with 

personal and extensive material damage (figure 2)' In 

comparison with other traffic participants, no great injury 

risk is apparent for truck passengers. This evidently high 

safety standard of trucks seems to minimize the need for 

preventive measures like safety belts, safety steeringwheels 

and for modifications of energy-absorbing 

structures of the passenger compartment. The proportion of 

personal injuries resulting from accidents with trucks 

lllA ltrt n*, {8ttt9 huil dnd l.lllrtl Pflo|lt 

Floure 1. Dlstrlbutlon ol traftlc Fartlclpants of all Inlured 

peisona. $ource: D€Psrtment of Stgtlstlcs FRG. 

zu 

trrA rrr, 

.ash pdrtlslpd"tt tgffiffi*,4 

FIoure 2, Ouots of Inlured persons and accldentg each 

pa*rtlclpants. 

Source: Ddpartmint of Statisti€ FRG. 

Floure 3. Portion of accidents wlth psrsonal Inlury (100% = all 

ac*cldents of sach panlclpant).-Source: Departmenl ol 

Statlstlcs FHG. 

(figure 3) clearly shows the danger to which other traffic 

participants are exposed by trucks. Similar to car accidents 

with a proportion of 48.6 percent personal damage, truck 

accidents result in 42.8 percent of personal damage. This 

means that priority must be given for measures to ensure the 

passive safety of trucks. 

Objective 

ln order to determine the need for such measures and to 

evaluate the safety of todays trucks, an analysis of the 

accident event is necessary. A rough orientation of the 

accident situation, like the frequency of certain accident 

types and locations can be taken from statistical reports 

which are based on police records. These are collected at the 

Statistic Federal Office at Wiesbaden (West Germany) for 

the whole of the German Federal Republic (SIBA 1)' They 

provide an excellent general view of the federal and 

regional accident situation and help to recognize 

fundamental focal points. But an optimum of reduction of 

accident causes and accident consequences is only possible 

after cognition and evaluation of injury-causing facts, of 

typical kinematic movement procedures and a

comprehensive description of the individual injuries 

sustained. Such a detailed accident analysis of individual 

cases is available in Germany by the research project 

*'Investigations at the scene of the accident," for which a 

specially trained team documents accident characteristics. 

independent of police investigations. 

Working method and data recording 

In the greater vicinity of Hannover (West Germany), an 

interdisciplinary research team, consisting of doctors and 

engineers drives to the scene of an accident, immediately 

after the accident has taken place. They are informed of the 

accident by the central rescue headquarters. The accident 

site is approached in vehicles equipped with blue light and 

sirene. On arrival they immediately start collecting information 

about accident traces, vehicle deformations and impact 

points on persons. With a stereoscopic camera, true^toscale 

drawings are produced which make a reconstruction 

and an analysis of the accident and collision phases possible. 

Otte (2) and Dilling (3) give a detailed description of the 

procedure. Injuries are documented in detail (4), according 

to type, localization and severity degree AIS. 

From these investigations 216 accidents concerning 

trucks with the permissible gross weight of 7.5 metric tons 

were at our disposal for evaluation. Figure 4 shows the 

distribution of the collision parries, divided inro weight 

categories. The most frequently registered truck accidents 

are those with cars (40.3 percenr). Multiple collisions are 

with 14.4 frcrcenr of all collisions remarkably frequenr, 

especially for truck-trailers ( 19.8 percent) and tractor-trail- 

er units (13.6 percent). 

In this case more than two objects or 

traffic participants respectively are involved as collision 

partners in one accident. 

!ollltlon- 

pdrll|cr 

lolril 

t4-l 

vehlcle llpe 

lruck 

r,rl - rrl 

a{l 

Itucl 

> ltl 

frtt I 

lr!llar 

unll 

atr 

rml* 

lrctlrr 

unll 

tx, 

lolql lnl tl6 zl 6l t6 

ll 

!dt 

rueh 

pcde!lrlttn 

oblrcl 

muluplc 

to.l 

lr.t 

t6.t 

r.l 

t.t 

lr.l 

70,r tt.t tt.r tt,t 

lt.t 2',.6 lt.t tr.o 

f.f 7t,6 tr.r zo.E 

- 6.t t.0 r.t 

1., 9.1 

f .t tr,r l9.t lt,6 

Flgure4. Golllrlon-partners of trucks u 7,5t (10O96=ail vshtctcs 

sEch truck colloctlyc). 

Special features of the accident situation 

Truck accident$ yary according to the different types of 

trucks. Accidents with trucks of lower weight categories are 

more frequent inside towns, whereas accidents with heavy 

truck-trailer and semi-trailer units are more frequent outside 

town limits, especially on motorways. On federal highways 

accidents with truck-tractor trains were observed in 34.1 

percent and with truck trains to 47.7 percenr (figure 5), 

while approximately only l8 percent of the involved trucks 

on federal highways belonged to the weight cfltegory below 

13 tons. In comparison, the truck weight category between 

7.5 tons to 13.0 tons on highways represents with g.l percent 

only a very small proportion of accidents. In town 

districts, however, it represents fl proportion of 36. I percent 

of all accident vehicles. 

htghwcy country urbsn 

femltrqller 

unlt 

trcrller unlt 

truclr > l3t 

truclr 7.5t-l3t 

Flgure 5. Reglonal locallzatlon ol truck (ts 7,5 t)-accldcntacsnss 

(lfil% = cach area). 

lnrpdcl-drcd lmpscl-ares trucft 

rollklo,n-Fqilnil tro$t rldc tc€t olharr 

cgl 47.17" (r- ll 

Itfirl 

rldc 

ragl 

olh.rl 

lrucfr Z3.l% 

(!.{lt 

lroill 

rldr 

,aaz 

olhcri 

lwo-wheeler 

19,1"/, 

lrurl 

rlrlt 

latlt 

othrn 

(i:i. 

t7.o% 

?9.5% 

t.0t6 

2.t% 

to.9tt 

9.3% 

23,t% 

1.57. 

r3.9% 

l{.r% 

9.t7, 

!.+% 

lt.t% 

2.3t6 

la.o16 25.6% 

1.ru, 

L27, 

{t.7% 

l3,e% 

,,.87. 

t.3% 

I,tt6 

2.87. 

t.t7. 

l.l'6 

2.t% 

pedeslr.5,97 8I.87o 9.t% 9.lYo 

obfcct4.3,% t?.54/n 50.0% 12.5t6 

Flgure 6. lmpacl-areas on the truck ln rllailon to the lmpact 

arcac of lhe colllslon partnerr (1fit% = slt cotilelone dech 

partner).

The collision types of trucks are determined by the varying 

accident situations. 

The collision constellations of the truck are shown in 

figure 6. Trucks collide mainly frontal. This fact is verified 

by the graphical illustration of detailed impact points in 

figure 7. 

lront tolcl! n-94 (50,8%) 

lell eenlri rlghl 

?t.3 

ffiil 

Il.g r effil 

ilt:f,',rE$ rlffi 

Hm 

n.16 (35.6*) 

wldc-ldccd 

I8,2 sl.e 

tegEEEEEEEEEEEEEEEt 

tots$ n-27 ll+.97e, 

n-a (6.9t) 

ffi eottto ol eyeltsts/Pedestrtans 

Flgure 7. lmpact areas on the truck In detsll detcription. 

In frontal collisions, cars are most frequently hit laterally 

by trucks (29.5 percent). In comparison, many collisions of 

trucks against trucks occur with front and rear involvement 

(23.3 percent plus 25.6 percent). Two-wheel users frequently 

collide frontally with the side of trucks (41.7 percent). 

Pedestrians are hit frontally to a percentage of 81.{t. 

The side area of the truck shows the rear region, especially 

the rear axle region as a frequent impact point. The rear 

226 

region of the truck sides, especially the rear axle region, is a 

frequent impact point (see figure 7). 

Injury situation 

Injury severity 

The description of the injury severity is based on the 

MAIS (maximum AIS) which includes all severity degrees 

from 0 = uninjured to 6 = dead. It describes the maximum 

severity degree ofeach single injury. Ofthe 261 truck passengers 

documented in the case collective of accidents with 

personal damage mentioned here, 85.4 percent remained 

uninjured. In comparison, only 10.3 percent of uninjured 

were registered for the collision parties (figure 8). 38.5 

percent of the collision partners sustained slight injuries 

(MAIS l/2), but 23.1 percent received severe or most serious 

injuries (MAIS 5/6). With truck passengers injuries of 

severity degrees 5/6 were found in only 0.4 percent (n=l). 

This clearly demonstrates the injury risk for traffic participants 

in collision with trucks. External traffic participants 

like two-wheel users and pedestrians are especially exposed 

(figure 9). ln collision constellations with these traffic participants 

no truck passelgenr were injured, but no extemal 

traffic participants remained uninjured. Two-wheel users 

are especially at risk. 39.5 percent of them received most 

severe injuries (MAIS 5/6) and 42.1 percent of them were 

seriously injured (MAIS 3/4). Pedestrians also sustained in 

two-thirds of the cases serious injuries (MAIS exceeding 2). 

tnlury- lruclt 

MAIS 

dll occupqnli 

+tal 

0 ro 86.4% 

u2 

f, 

Eolllslon-pdrlner 

I 

occupdnl|/ 

lrilo-rf,hril-rtdcrr/ 

trrdlrlrlonl ,*- rro 

to.t* 

tt.rz 38.5% 

3t4 2.7% 24.2% 

5t6 0.4"/o ?,3,.le/c 

Flgure 8. Inlury-severlty-grades for occupsnts ol trucks (t 7, 5 

t) and theh colllelon-partners. 

MAIS 

lructr F cdr ruck Hlwo-wh€€lE 

trucl- rFt b.lld IrueI- lrtsl- 

Ddqlrldn 

loldl 

0 

n.l0a n-tl 

t5.{% ET 

n-{l 

E 

n.3l h.It 

@ 

h.ll 

u7 

3t4 

1.9% ffi,n G 

[ 

l".n'- f,..on 

",'* 

E 

laz,r% 

b E*. 

5t6 15,+% ffi''* 

1,..0* 

Figure 9. Injurity-severlty-grsd6s ot occupants of trucks (> 7,5 

t) and their colllslon-psrtner$.

Injured truck passengers were as a rule registered only in 

collisions with vehicles and objects (figure 9/10). No severe 

or most serious injuries (MAIS exceeding 2) could be observed 

for 104 truck occupants, in collisions with cars. Only 

L9 percent of the truck occupants sustained injuries, but 

only 15.4 percent of the belt-wearing car passengers remained 

uninjured and one third of them sustained serious 

injuries MAIS exceeding2 (32.7 percenr). Injuries to rruck 

passengers in single accidents of trucks were observed to 50 

percent, but exclusively of severity degrees l/2. In collisions 

between truck$, a significantly greater injury risk ex- 

ists for truck passengers, especially when the mass of the 

collision parties is quite similar. Figure l0 demonstrates the 

fact that in accidents of trucks with collision parties of lesser 

mass, for instance of the weight category less than 3.5 tons, 

passengers in trucks up to 7.5 tons remained to 84.2 percent 

uninjured. Here severity degrees MAIS 5/6 were not observed. 

In comparison passengers in trucks of weight category 

from 7.5 tons onward in collision with vehicles of the 

same weight category remained in only 52.6 percent uninjured, 

7.9 percent of them were seriously injured (MAIS 

3/4) and 2.6 percent were most seriously injured or killed 

(MArS 5/6). 

Influence of weight relation 

The statistic-mathematical proof about the connection 

between the danger to truck passengers by the mass relation 

of the collision parties is verified when applying the relation 

'Weight 

of truck from 7.5 tons onward to the weight of the 

collision partner' and the injury severity MAIS caused by 

the accident for this type (figure I l). This so-called mass 

factor was on the one hand determined for the passengers of 

the collision partner of the truck (figure I l-left graphic) 

and on the other hand for the passenger of the truck (figure 

I l-right graphic). With a mass relation of approximarely 

5, this means for the pa$sengers of the collision partner that 

they will probably remain uninjured in the accidenr. The 

injury probability in the arithmetic middle should be expected 

for a mass relation of approximately 3, but as shown 

in the run of the curve, a similar injury probability exists for 

all severity degrees. 

cCllrlil tsucl*lrucl rt,tl lruIFlrel

the pelvis/abdomen and thorax region for pedestrians' Ttvowheel 

users are very much at risk for fractures of the cervical 

vertebra (26.3 percent of two-wheel users). Noticeable 

are neck injuries, also with car passengers, who to 20'4 

percent sustained these injuries. In other traffic collisions' 

car passengers very rarely suffer neck injuries (Rether, 5). 

Passengers of trucks of the weight category less than 3'5 

tons also sustain substantial injuries to the whole body. 

Injury causes 

Passengers of trucks from 7.5 tons onward are mainly 

injured by parts of the vehicle front (figure l3), such as the 

windscreen (to 18.8 percent), the dashboard (36.8 percent), 

and the front foot-room ( I I .3 percent). Due to the fact that 

passengers of trucks weighing less than 7.5 tons are more 

often laterally impacted, the percentage of lateral interior 

structures as injury-causing parts is increasing (10.2 percent), 

while injuries caused by the dashboard and windscreen 

region are observed to a lesser degree. For lighter as 

well as for heavier trucks frequent injury causes are found 

outside the vehicle (for both collectives over l3 percent). 

Injuries sustained by truck passengers are mainly of a less 

serious kind. Approximately 85 percent are of severity degrees 

AIS I and 2. Severe injuries are quite rare in heavy 

trucks. In trucks weighing less than 7.5 tons injury severity 

degrees from AIS I upward are more frequent. This is 

especially the case in lateral collisions, by lateral interior 

parts as well as by parts outside the vehicle. 

ccura ol lnlury 

dsrhHH 

Itml lrg Fan 

lFlatld-lclarcl 

lnl.rld-ast 

thrflil lllllngr 

oulrld. vrhlqlr 

olhrr 

lolcl 

txl 

lr,l 

tfrt 

lruch r t.5l 

qll dEupsll 

lr.i 

6.J 

al 

t.t 

t(l 

e3 

ru urlrn >: 

10{I0 

9t.t t t 

tdo.0 

I00,0 

IOO.0 

t00,0 

tt9 1l.r 

t0.0 140 

tn ur le3 

Irftt < t,tl 

oll *EuFdrll 

lolql 

r- li! 

ttl 

rl.5 

tt.l 

,.2 

l0.t 

L6 

r,! 

It.1 

tt,6 

ru vrlrr >r 

F llcx + 

n,, 1l 

to.t rt l 

tt r u.i 

6d' 55.5 

too.! 

loo.o 

to.6 tt.a 

,l,t r|'l 

rll 

E-oGuFnlt 

9,0 

,1' 

4r 

tl.J 

t.0 

tl.t 

4.6 

tt6 

t$l-b.ilrd 

tu urlrn ' 

te6 u,a 

,Lt n.9 

t .5 tl.a 

dr.! trc 

teo t[o 

tt.0 lt.o 

tt,t 04., 

lo,t lr.t 

Floure 13. Cauees of lnlurlst for trafflc members In traffic accl' 

d6-nts wlth trucks z 7,5 I and Indlcatlon of the severltygrades. 

Car passengers are distinctly more severely injured, even 

with belt usage. 73.8 percent of the injuries are of severity 

degree AIS I and 2. Mainly regarded as injury causes are 

parts of the dashboard (to 26.7 percent), lateral compafiment 

parts (21.3 percent) and also by parts of the interior 

like steering wheel and safety belt (11.8 percent). Injuries 

caused by the belt are exclusively soft-part injuries, due to 

the close fitting belt strap holding the body. For belt-wearing 

car passengers 4.6 percent of all injuries are still inflicted 

by parts outside the vehicle. This may be due to extensive 

intrusions and deformations of the passenger 

compartment of cars colliding with trucks. Figure 14 shows 

228 

the injury situation of two-wheel users and pedestrians in 

collision with trucks from 7.5 tons upward and demonstrates 

that pedestrians as a rule sustain injuries by the truck 

side, which proves to be especially dangerous for this group 

of traffic participants, in view of protruding, often edEY 

structures. 29 percent (20.5 and 8.5 percent respectively) of 

two-wheel users sustain injuries by side structures of trucks. 

Injuries due to overrolling are also quite frequent, i'e.23.7 

percent of injuries inflicted on two-wheel users and even 

63.4 percent of injuries to pedestrians are induced by over* 

rolling. These injuries are often serious (for two-wheel 

users 38.2 percent exceeding AIS I, pedestrians 52.5 percent 

exceeding AIS l). 

:dus6 ol lfllury 

rollover 

toqd gurlEte 

lront €nd 

wlnd!dreen rcglon 

sldc 

regl 

rpcctllc lruch-pqrl5 

own lwo-wheeler 

own h6lmsl 

olhcts 

lolql 

n. lE6 

69,{ 

ls.{ 

t1o 

p6d6rlrldnr 

6,6 

ojo 

,., 

lnlurT-sevcrlll 

AIS r/2lArs >2 

- 100% - 

47.5 52.5 

88,0 lu.0 

'o: il:' 

s3.3 6&7 

100.0 

100.o 

lnlurles 

loldl 

n- 618; 

23.7 

22,.O 

'j 

e0.5 

$.9 

?,f 

8.5 

i,l 

0,6 

nturY-swerllI 

us vtltts >t 

F 

61.6 3E,t 

8t.7 le.3 

66.0 34,O 

6t.t 3r., 

6tr0 36,0 

i7.8 t2.2 

84,1 15,9 

87.6 l2.S 

100.0 

Fioure 14. Cauees of iniurles In trsfflc sccldent wlth trucks E 7' 

5 fand Indlcatlon of the- severlty-grades' 

Discussion and consequences 

216 trucks of a weight category from 7.5 tons upward 

were documented, reconstructed and analyzed by investigations 

at the scene of the accident. For the analysis impact 

points, vehicle deformations and mass proportions of the 

collision parties were examined and compared with the 

injury severity of truck passengers and collision pafiner. 

For passengers of trucks from 7.5 tons onward, a high 

standard of safety is evident. Only in collisions with vehicles 

of a heavier weight category the situation becomes 

dangerous for truck passengers. The significant influence of 

mass relations between collision parties on the injury probability 

and severity could be established within the framework 

of this study. All analyses did show that there is much 

more safety for truck pas$engers in heavy truck trains and 

truck-trailer units than in trucks of lighter weight categories. 

for instance the solo truck of less than 3.5 tons. This 

calls for meartures to achieve more safety for passengers, 

especially as far as the last mentioned truck group is concerned. 

As these vehicle groups are to a high percentage also 

involved in pedestrian collisions (see Otte, 6) and cause one 

third of accidents inside towns and closed areas, special 

measures are necessary for the safeguarding of extemal 

traffic participants. For the lighter trucks especially, modification 

concepts for the passenger comPartments like 

those installed in cars are necessary. The use of safety belts 

should also be obligatory for them.

A high injury risk is evident for car passengers, especially 

when the passenger compartment is part of the deformation 

area. Even for belt-wearing pas$engers a greater than average 

injury risk still remains. This may be due ro high vehicle 

front and also to the relatively small amount of energy 

absorption of the lateral car $tructures. Collisions between 

trucks and external traffic participants are regarded a$ e$pecially 

dangerous. While pedestrians are usually hit by the 

box-type shape of the truck and consequently sustain partly 

very serious injuries to the whole body, two-wheel users 

often receive very serious injuries by the side of trucks. 

There is a pronounced high risk for the momentous overroll 

injuries, mainly for pedestrians due to the high ground 

space. The, as a rule unprotected side of trucks makes an 

underdriving of the vehicle body for the two-wheeler possible. 

Edgy frame structures are an additional injury risk. 

The above analysis ofreal accidents emphasizes the need 

for modification measures for trucks. In view of the cognitions 

derived from our investigations, these should be as 

follows: 

For the protection of external traffic participants like 

pedestrians, cyclists and riders of motorized two wheelers: 

r I pulled down truck front to reduce the danger of 

underdriving and eliminate the danger of 

overrolling. 

r An energy-absorbing front area, for instance by 

special inserted impact regions on frequent impact 

points, for the reduction ofhead, thorax, and 

leg injuries for these traffic participants. 

r The bumper of trucks should have a thick rubber 

covering, if possible in foot-rest level of the twowheeler, 

in order to ensure more protection 

against tibia injuries at low impact speed. 

r Side and rear protection to eliminate underdriving 

of two-wheelers between the axles and truck 

body. Only such underdriving protections should 

be used that cover the side regions completely. 

Gaps and edgy conceptions (type of frame fixture) 

make an entangling with the two-wheelerpossible 

and the resulting injury severity could not b€ reduced, 

for instance in the case of a possible head 

impact with edgy frame structures. 

r The elimination of edgy vehicle parts which are 

especially injury causing with disposal and agricultural 

vehicles. 

For the protection ofother vehicles, especially cars: 

r Height adaption of energy-transmitting regions 

and implementation of energy-absorption regions 

on trucks, for the reduction of impact force by the 

truck also. This would lead to a compatible collision 

constellation. 

For the protection of truck pfls$engers 

themselves: 

r Establishing of a safety standard similar to that for 

cars. Basically, the same safety standard should 

be created for truck passengers as for cars; i.e. 

energy absorbing front structure, reinforcing of 

the passenger compartment, covering of the dashboard, 

fixing of safety steering wheels and first 

and foremost fixing and use of safety belts in 

automatic 3-point quality for all truck passengers. 

References 

(l) Statistisches Bundesamt Wiesbaden StraBenverkehrsunfiille 

1987. Fachserie 8. Verkehr Reihe 3.3 

Verlag Kohlhammer, Mainz (1988). 

(2) Otte, D. et al, Erhebungen am Unfallort, Unfall-und 

Sicherheitsforschung, Heft 37, Bundesanstalt fitr 

StraBenwesen, Bergisch Cladbach ( 1982). 

(3) Dilling, J.; Otte, D., Die Bedeutung drtlicher 

Unfallerhebungen, im Rahmen der Unfallfor$chung, 

Unfall.-und Sicherheitsforschung, StraBenverkehr, Heft 56, 

Bundesanstalt ftir StraBenwesen, 5948, 1986. 

(4) American Association forAutomotive Medicine, The 

Abbreviated Injury Scale-Revision 1985, Morton Grove, 

Illinois, USA (1985). 

(5) Rether, J.; Otte, D., Verletzungen der 

Halswirbelsiiule, beim Verkehrsunfall, Unfallheilkunde 87, 

52f530, 1984. 

(6) Otte, D., Collision Situations and Consequences of 

Injuries in Accidents of Heavy Trucks, Proc. OECD 

Sympos. 

"The Role of Heavy Freight Vehicles in Traffic 

Accidents", Mon$eal (Canada), 28.-30. April 1988. 

Improvements in the Fronto-Frontal Collision between Personal Vehicle (PV) and 

Heavy Duty Vehicle (HDV) 

Written Only Paper 

Pierre Soret, 

Renault Vehicules Industriels 

Abstract 

Safety is a major objective in the program of the 

experimental HDV Virages specifically to obtain 

improvements in the case of the fronto-frontal collision 

between a PV and a HDV. 

The architecture and the structure of the tractor have been 

designed in this aim. 

To evaluate the obtained performances: 

r Choiced a typical PV in the market. 

r Studied and defined criteria of dammage for the 

structure and for the occupants of the car. 

r Realized crashes of thi$ PV instrumented as usual 

with 2 dummies installed in it: first, againsr a 

$topped serial HDV as reference, and second, 

229

against a simulated front part of the experimental 

vehicle Virages. Crashes have been realized in 

different cases of speed for the car. 

r Compared the results of the two types of crash 

tests and evaluated the gains obtained by the 

experimental vehicle Virages' 


The programme of the experimental HDV Virages 

includes the improvement of active and passive security. 

Among the subjects of this improvement the reduction of 

the gravity of the fronto-frontal collision between an HDV 

and a PV is an important one, because this type of collision 

is the most frequent and the most severe in the collisions 

between these vehicles. More, this type of collision creates 

the greater number of injuries and deaths among the road 

users when a HDV is concemed. So the programme Virages 

includes tests to evaluate the improvement made about this 

subject. 

Conditions of the Tests 

The study of the circumstances of such accidents 

demonstrates that the collisions may concern either the total 

width or a variable part of the cars. In all the cases it is 

always the driver's side of the car which is concerned' A two 

third of the width collision is the most representatiYe case. 

ln many cases the paths of the vehicles are $lightly 

parallel. 

The speeds of the vehicles at the moment of the collision 

are unknown because, at present, there are no means to have 

a seriou$ evaluation of these. Therefore we choose to take 

the speeds as the main parameter of evaluation. 

In the study no element proved that the repartition of the 

speeds between the vehicles has an influence on the results 

of the collision. So it was possible to choose to have the 

HDV in a steady state situation. This simplified 

considerably the running of the tests and permitted only the 

use of just a front part of vehicle Virages set on a wall. 

?t3 

widlh 

of bh he 

car 

t-l\ 

-J 

Flgure 1. Tests condltlona 

230 

# 

-|-l 

l-l 

t-l 

#.+ 

mean$ of two or three tests, the speed which conduct to the 

limits of the chosen criteria. This is the speed limit for this 

arTangement of vehicles. So, different speed limits can be 

defined for the arrangement of different HVD with the same 

car. These speed limits characterize the behaviour of these 

HDV concerning the reference chosen car. 

In conclusion, the behaviour of Virages was evaluated by 

comparison with a representative one of the current HDV 

tractor in France: the R 310. 19 T from Renault Vdhicules 

lndustriels. 

Taking in account, first, that the analysis ofaccidents did 

not prove a particular PV was concerned, Second, that there 

was no barrier available for the purpose of such tests, we 

decided to choose a R 9 Renault which is a reuresentative 

medium car in France. 

Results of the tests 

The tests were conducted at three different speeds in the 

case of the collision against the R 310. They quantify the 

variation of the criteria versus the speed variation. It is 

possible to conclude that the speed limit for this case is 

approximately 59 kmTh (36,6 nph). As a matter of fact, at 

this speed, the HIC and D I criteria for the driver reach the 

maximum permitted levels (see figures 3 and 4). The other 

criteria and those for the front passenger are acceptable 

(figures 5 and 6). 

s0 57 s9 

Flgure 3. D 1 Crlterlon vsluos aftEr teEta 

50 57 

Flgure 4. HIG Grltcrlon valucs 

59 

60 

50 

20 

50 57 59 66 

FIgure 5. Thorar ecc6leretlon crltorlon valuss 

57 s9 

Flgure 6. Femurr ttraln crllerlon valuea 

At 66 km/h (41 mph) no limit level is reached in the 

collision ofthe car against Virages for all the criteria, but the 

gap is small in the case of the HIC of the driver. 

An improvement of the speed limit from 59 to 66 km/h or 

more has been made with Virages. 

Conclusions 

The height of the rigid parts of the actual HDV is the 

reason of the severity of the collision with PV. The working 

parts of the car run under the bumper and the frame of the 

HDV. They are only pressed against the ground till they 

Flgure 7. Posltlon ol the rlgld parts ol a PV end an actual HDV 

23l

each the front axle or the tyres. The bumper of the HDV 

may penetrate in the in-door space of the car (figure 7). 

The improvements made in the vehicle Virages are 

(figure 8): 

232 

r bumper and rigid parts of the frame at the $ame 

height as the height ofthe working parts ofthe car 

r reduced front over hang of the HDV = the tyres 

limit the deformation of the bumper 

r reduction ofthe roughness ofthe bumper and the 

rigid parrs of the HDV. 

In conclusion, such changes in the way to construct the 

front of a HDV allow a significant improvement in the speed 

limit of a collision between a HDV and a PV. 

J 

l 

Figure 8. Poeltlon of the rigld partE ol a PV and of Vlragea

Technical Session 2A 

Frontal Crash Protection 

Chairmanr Kennerly Diggs, United States 

Analysis of Frontal Crash Safety Performance of Passenger Cars, Light Tfucks 

and Vans and an Outline of Future Research Requirements 

James R. Hackney, William T. Hollowell, and 

Daniel S. Cohen 

United States Department of Transportation, 

National Hi ghway Traffic Safety Administration 

Abstract 

Since 1979, almost three hundred vehicle crash tests have 

been conducted in the New Car Assessment Program 

(NCAP). Parameters from the crash tests of the vehicles are 

presented. Analysis of these parameter$ relative to the frontal 

crash safety are performed. The studies are conducted 

with fleet weighted populations and include both passenger 

cars and light trucks and vans. Comparisons of vehicle 

performance are made which suppon the feasibility of potentially 

improved safety levels within the vehicle. 

Based on the potential for improved safety and on recent 

projections of fatalities and injuries by the National Highway 

Traffic Safety Adminisrration, an outline of possible 

future frontal crashwonhiness research is presented. 


Since 1979, the National Highway Traffic Safety 

Administration (NHTSA) has conducted crash tests of 233 

different makes and models of passenger cars (PCs) and 4l 

light trucks and vans (LTVs) in the New Car Assessment 

Program (NCAP). Each of these vehicles was crashed into a 

rigid barrier at a test speed of 35 mph. This is five mph faster 

than the prescribed speed for compliance with Federal 

Motor Vehicle Safety Standards (FMVSS). The NCAP is an 

experimental consumer information program which 

develops data on frontal crashes. In this program, a given 

vehicle is tested once at the nominal test speed of 35 mph 

with restrained and instrumented 50th percentile part 572 

dummies in the driver and right front seat pa$senger 

locations. The crash test$ are designed to indicate, for 

vehicles within the same weight class, the relative levels of 

occupant protection and vehicle safety in this crash 

condition. The data from these tests provide NHTSA with 

the most extensive set of frontal crashworthiness 

information ever assembled on production vehicles. 

At the Tenth International Technical Conference on 

Experimental Safety Vehicles, a review of the data from the 

NCAP vehicles which had been tested was presented (l).* 

This report updates the data contained in that review with 

the results from an additional 102 vehicles which have been 

tested through model year (MY) 1988. In addition to 

*Numbers in parcntheses dcsignate refercnces at end of paper 

presenting similar tables and charts as in the 1985 paper, this 

update provides trends and analyses which are based on 

fleet weighted observations and provides comparisons of 

PC to LTV parameters. The fleet of tested vehicles 

estimated to be on the roads in 1988 consists of over 37 

million PCs and almost 9 million LTVs. Comparison of the 

performance of the tested 198 I PC fleet to the tested 1988 

PC fleet shows a significant improvement in potential safety 

performance for restrained occupants in high speed frontal 

crashes. Comparisons of specific late model (MY 1986 to 

1988) vehicles within each weight class indicate the 

potential for significant safety improvements for the future 

vehicle fleet. 

The analyses ofthese frontal crash data and projections of 

fatalities and injuries are the basis for an outline of future 

frontal crashwonhiness research activ ities. These activities 

will study the projected remaining safety problem after the 

full implementation of Federal Motor Vehicle Safety 

Standard (FMVSS) 208. It is projected that even with the 

significant benefits of FMVSS 208 and mandatory use laws, 

more than 10,000 fatalities and 120,000 moderate to seyere 

injuries will still occur in frontal crashes each year in the 

United States. (Note: From the FMVSS 208 Final Rule 

notice in the Federal Register (2), mid-point estimates of 

benefits show a reduction of over 6.000 fatalities and over 

100,000 moderate to critical injuries per year when all 

pas$enger cars in the fleet meet FMVSS 208 requirements). 

TFends From the Anthropomorphic 

Dummy Responses 

In each of the NCAP tests, Part 572 anthropomorphic 

dummies are located in the driver and right front passenger 

seats of the vehicles. Standard in$trumentation includes: 

triaxial accelerometers in the head and chest. and load cells 

in the right and left femurs. The head injury criteria (HIC), 

three millisecond clip chest accelerations (Chest Gs) in gs, 

and femur loads are derived from the accelerometer and 

load cell outputs as specified in FMVSS 208.t Only the HIC 

and Chest Gs will be used in this section to examine the 

trends of the dummy responses. 

Some of the analyses in the following sections will 

separate the vehicles into two performance groups to allow 

easier comparisons and trend studies. Performance Group I 

consists of vehicles in which no dummy response exceeded 

a HIC of 1250 or a Chest C of 70. Performance Group 2 

consists of vehicles in which at least one dummy response 

exceeded these values. 

I In response to a petition, in I 986, the NHTSA revised rhe procedure for calculating 

HIC such that the maxirnum timt intcrval did not dxceed 36 milliseconds. Prior year 

HIC values in the NCAP tests have not been changed. Howevet, the effect of the 

change was examined and is insignificant relative to the analyses in this paper.

The analyses for the dummy responses and for many of 

the other parameters in the following sections of this paper 

are done forboth the unweighted and weighted NCAP data. 

The unweighted data are the data from actual vehicles 

which have been tested. The weighted data are these data 

which have been weighted by the estimated distribution of 

the tested vehicle in the fleet for a given year. This estimated 

distribution is derived from total vehicle registrations for 

1979 to 1986 models with vehicles identical to the test 

vehicles as retrieved from Polk's Annual National Vehicle 

Population Profiles. For 1987 and 1988 models the 

registrations are estimated from vehicle sales reported by 

Automotive News. To compute accumulated vehicle 

registrations through the model years survival rates shown 

in table I are used. These rates are derived from the Polk 

data. 

o I 

Table 1. Survlval ratss derived lrom Polk reglttration data. 

v€htcle A6E I 2 5 I 

Suwival Rste 0.998 0.984 o.972 o.958 0.932 0.895 0.84I 0.769 

Analysis of PC dummy responses 

Table 2 contains 

HIC and Chest Gs 

r 200 

L t000 

6 

o 

a 

t 

u 

I 

o 

,R 

the average and median values of the 

of the driver and passenger dummies 

Driver H lCs 

600 

t979 t9A0 t96t tg62 tt6J t9t. 19A5 1986 rgtt 

15 

{o 

JO 

t979 1980 

-** 

- greighted Avcrsgc 

t 

- 

l..l6wc;EhtEd Avcrog! 

- 

'Wcight.d Ucdiqn 

\ 

- -- - 

I 

Llnw.ightcd u"wrightcd Medidn Mtu 

-* 

_*ts 

Lofcrl Modrl Ycor 

Driver Chest Gs 

l-':"'el:lT.:1": 

- wiighttd AvGrofli 

Unwiight.d Av6roga 

- - WAightad M.diff 

- - Unwalghtad Mtdlod 

r9a2 1983 r9E4 1985 1966 1967 t9t8 

Ldfi3l Modrl Y.or 

Flgure 1. Comparlson of dummy resPon$os for NCAP Psssonger cars' 

234 

through the ten MYs for PCs. This table also gives the 

cumulative number of tested vehicles and the representative 

tested fleet. Figure I depicts these data in graphical form. 

These data indicate a significant trend of lower responses 

from 1979 to 1988 for unweighted and weighted data. 

Table 2. Averagee snd medians of occuPant HlCs and chest 

Gs-paaaengsr cars, 

o 

I 

r ?oo 

$ rooo 

a 

D 

I 

t 

o 

xodrl 

Yr*tr 

1919 

1979.80 

1979.81 

1979- 8? 

1979 - 83 

1979 - E4 

1979.85 

1979.86 

L979-81 

7979.88 

tud.t 

YsarE 

800 

a 

I 

o 

$ 4 5 

a 

6 

ca 

f+o 

E 

I 

Drlvsf 

HIC 

L210 1056 

1279 1088 

12?1 IOTE 

1166 976 

Irzs 960 

rr01 939 

1118 941 

1095 91I 

1084 897 

1065 894 

Drtvcr 

HIC 

Frrr.n8tr 

HIC 

Unerlahtrd 

1,21t tt63 

r29! 1179 

L241 1171 

1168 1033 

1132 1031 

1093 

I0l0 

to76 962 

1054 945 

1016 884 

977 836 

T+l8,hted 

PrrranSat 

HIC 

Av. ll|d Av€ hd 

197 9 1045 955 1121 1033 

1979.80 1071 965 1124 994 

1979.8 I IO40 939 1097 994 

t979-82 1016 939 1081 991 

I979 - 83 9E5 910 1049 9I9 

1979. E4 964 87t 996 865 

r979.85 941 649 966 859 

1979.86 926 845 936 829 

t979-87 917 826 903 799 

1979-88 903 E02 867 718 

Drtvrt 

Chart C 

56 59 

57 s4 

56 54 

53 51 

53 50 

s2 50 

31 50 

51 50 

51 50 

51 50 

Drtvsr 

Chest G 

53 54 

5L 52 

54 52 

53 50 

s2 50 

51 48 

50 48 

49 4E 

49 48 

49 47 

- Wciqht.d Avfog. 

UnFiightcC Avtrqg. 

- - w.ightd M.dlcn 

- - Unsr'ghttd u.dion 

600 

I 979 tgao tg8l 1962 1983 r98+ rg85 1986 1987 

Lotarl llodd Yasr 

F.r.cn8Gr 

freat C Nsbcr 

4A 45 

50 46 

49 45 

46 41 

45 43 

45 43 

44 42 

44 42 

44 47 

30 

51 

54 

EE 

t02 

133 

180 

?06 

233 

Parrin8rr 

ch€lt c NCAF 

FlG6t 

4t 45 

48 45 

4t 45 

47 45 

4L 42 

43 40 

42 40 

42 40 

41 40 

Passenger H lGs 

Paesenger Ch€st Gs 

- Wrightld Avlrqgc 

Unwcightrd Avarogc 

- - Wclghtrd Ucdion 

-. UnwliqhtGd Midiqn 

e197900 

52375 39 

7A219EO 

10407385 

12441 425 

17065150 

21709065 

27O6O6r3 

3r924r96 

37044106 

1982 19SJ t98{ 1965 t966 t967 1968 

Ldtrrl llodrl Y.dr

B6 

I 

H l o 

A 

f 

eg { ! 

I 

E 

I r o 

E 

t"" 

E 

6so 

a 

E 

E,o 

E 

a 

Figure 2. 

llaet. 

It.s l9ll rlll lsl itu ttlt t9ta 19t7 tel 

ilodrl ysr cf Vrhlcl. Ft.rt 

Safety performance trBnds of NCAP pas$enger car 

Figure 2 shows the percent of the weighted vehicles in 

Performance Group 2. This plot indicates that the portion of 

the tested vehicle fleet with the higher dummy responses 

hasdecreased from 5 I percent in 1979 to 28 percent in 1988. 

It may also be noted that 43Vo of the PCs mer FMVSS 208 

requirements even at the higher test speed. 

Comparison of the l98l fleet to the 1988 fleet 

From a comparison of two fleet years, l98l and 1988, the 

distributions of the Performance Croups can be $tudied in 

figure 3. The data in this figure show a significant shifr to 

lower level responses has occurred from the l98l to 1988 

fleet. Similar trends for the weighted and unweighted data 

are apparent. The l98l weighted passenger car fleet contains 

eight million vehicles and the 1988 fleet contains over 

g 

ro 

[.o 

E 

$ 

Ftrfcrmonct GrouF 

Flgure 3. Comparleon of lg81 vs 198E NGAP PG fleets. 

37 million vehicles. The l98l unweighted fleet contains 64 

vehicles and the 1988 unweighted fleet contains 233 

vehicles. 

The distributions of the weighted fleets by weight class 

(l) are shown in figure 4. The four weight classes aresmall 

(curb weight below 2,450 lb.), compact (curb weight 

2,451 to 2,950), intermediate (curb weight 2,951 to 

3,450 lb.), and large (curb weight above 3,451 lb.). The 

average weight of vehicles of the 8l fleet is 3278 lbs and the 

average weight of the 88 fleet is 3 189 lbs. (Noter The decision 

to compare the 198 I fleet to the 1988 fleet is based on 

the assumption that after three years of the NCAP that 

enough time had elapsed for the rnanufacturers to become 

responsive to the test data. Similar comparison can be made 

for any other fleet year). 

G lgal Pf, Fl!!t (Weiqhtsd) 

E:l rc88 FC Frcdi (wi,qhr!d) 

Flgure 4. Dlstrlbutlon ol 81 and 86 NCAP llaets (by vehlcle 

welght class). 

Weig hted Unweighted 

P.rlc.m€ncr Croup

PC dummy performance by weight class 

The real world safety risk is dependent on the weight of 

the vehicle (i.e. from the laws of physics, in a two vehicle 

frontal impact, the lighter vehicle will experience the larger 

velocity change). In the following paragraphs the NCAP 

vehicles are separated into the four weight classes as previously 

defined ( I ). The dummy responses and design parameters 

of each weight class will be examined. 

The following analyses are conducted only to compare 

the performance of different weight class vehicles in the 

single vehicle crash into a fixed rigid banier (which may 

also be viewed as a crash of two identical vehicles colliding 

head-on with each vehicle traveling at 35 mph). In this 

crash, each vehicle must dissipate its own kinetic energy. 

These analyses do not imply that performance in these tests 

are indicative of the risks across weight classes. It has been 

documented (4) that the occupants of smaller vehicles are at 

greater rink than those of larger vehicles due to the traffic 

mix. 

Table 3 contains the average and median values of the 

HIC and Chest Gs of the driver and passenger dummies for 

the four weight classes. This table also gives the number of 

tested vehicles and the size of the weighted fleet for each 

weight class. These data show minor differences across 

weight classes. 

Trble 3. Averagee and medlans of occupant HICs and chest 

Gs-pas8eng€r cars by welght claaa. 

Uc ltht 

CIar6 

sMll 

CqEpac t 

InteEd 

IrrEc 

g. tght 

Clilr 

gMIl 

Coqrct 

Int.md 

InrE. 

To further examine differences in performance between 

weight classes, figure 5 wa$ generated which compares the 

weight classes in the Performance Group 2. The information 

in this figure indicates better performance of the heavier 

vehicles for the unweighted data. However, when the data 

are weighted by the vehicle registrations, these performance 

differences are reduced. This indicates that especially in the 

small weight class that the poorer performing vehicles have 

the lower sales volume. From closer examination of the 

individual vehicle data, it also found that many of the poorer 

performing small cars were tested in the early years of the 

NCAP and that attrition is gradually eliminating these from 

the fleet. 

Analysis of LTV dummy responses 

LTVs were not included in the NCAP tesring unril MY 

1983. Since then, 4l LTVs have been tesred. Due to the 

236 

Drlvar 

HIC 

PsEeFfitar 

HIC 

tlffitght.d 

Drlvar 

ChEt c 

AS H.d Avs l,tsd Av. Med Awe lt d 

1097 974 

1072 845 

1008 880 

1037 954 

Drlvef 

HIC 

1090 959 

936 77r 

885 778 

887 808 

Paaaangar 

HIO 

tleltht€d 

49 47 

5t 5l 

5t 50 

Drlwr 

Chrrt C 

Psda6nB.r 

Ch.rt C llwbef 

Terted 

L5 4! 

L2 4I 

41 42 

43 4r 

Avc Hsd Av€ lt d Ave Hed Avc Hed 

882 

918 

905 

9L7 

708 

810 

784 

882 

921 821 

802 653 

802 752 

1094 1084 

49 47 

47 46 

51 50 

49 45 

EO 

77 

41 

23 

Prsrenger 

chc8t G TICAP 

FI..t 

42 40 

39 38 

4t 40 

46 47 

r2254210 

tt322t7Q 

79439I1 

3523265 

Fi 

N 

[ - 

6 - 

E 

? 

q 

(J 

o 

Conpact lntsrhEdioic ( o/ge 

Flgure 5, PCs in perlormance group 2 (by welght claae). 

lable l. Averagea and medians of occupsnt HlCs and chest 

Gs-llght trucks and vang 

LI1t 

Tlr. 

l{F/r 

PUc 

VAilt 

Drlvcr 

HIC 

PaaBGngsr 

HIC 

Unwctghted 

Drlver 

Cheat G 

Ave Hed AvG H+d Ave H€d Av. Kcd 

1023 950 

1201 rr47 

1565 1432 

1148 1134 

1065 754 

1108 II23 

51 49 

55 57 

59 55 

Totrl 1282 rl00 r067 56 54 47 

LTV 

IYp6 

uE\tr 

PUg 

vrJf 

Drlvar 

HIC 

l{rlghted 

PatsenSet 

HIC 

Drlvsr 

oheet c 

PssasntGr 

Ch.rt C Nwber 

Teeted 

45 48 

47 41 

49 41 

Ave l{6d Avc Hed Avs HGd Avc l{ed 

869 850 

1t0I 7L47 

7362 984 

1309 1321 

979 699 

1100 7798 

50 48 

53 50 

sparsity of data, data for individual MYs may not be signifi* 

cant. However, the 4l tested LTVs represent over 9 million 

vehicles in the 1988 fleet and more than 30Vo of IJIY registrations 

since MY 1983. Therefore. it is concluded that the 

total tested fleet data can provide significant data for examining 

the dummy responses. Table 4 gives the averages and 

medians of the LTVs. In comparing these values to the 1988 

PC fleet values in table 2, it is noted that the dummy respon$es 

in the LTV fleet are approximately 207o higher for 

both HICs and l07o higher for Chest Gs. 

Figure 6 shows the di$tribution of the PC fleet and the 

LTV fleet in the two performance groups. For the LTVs, 

567o of the unweighted fleet and 457o of the weighted fleet 

are in Performance Croup 2. For the PCs, these values are 

387o and 25Vo.The comparison in this figure of the performance 

group distributions points out the large difference in 

the test performance between LTVs and PCs. 

The LTVs can be examined in three categories: light duty 

trucks, vans, and multi-purpose vehicles. Table 4 also provides 

the dummy response values in these categories. From 

observing the average values, it is noted that the vans had 

the dummies with the highest HIC and Chest G responses of 

the three categories. 

These data indicate that the frontal crash test performance 

is not as good for LTVs as for PCs. However, the manufacturers 

have shown the capability of producing reasonably 

good performers in all the LTV categories. In fact, I I of the 

43 LTVs meer rhe requiremenrs of FMVSS 208 ar the higher 

crash speed. 

l0 

I4 

FsrEenget 

Ch6rt C NOAP 

Fleet 

44 43 

45 45 

50 ttj 

r46139? 

4869077 

2431951 

Totsl 1135 985 1067 1071 52 50 46 45 8162420

Pdormonor Qrggp 

Flgure 6. Comparlson of LTV and PC flests. 

Vehicle Design Parameters 

As in the previous paper (l), parameters of safety belt 

performance, steering assembly effects, and structural 

performance have been tabulated. In the following 

examination of these data, any changes which have 

occurred in trends from the first paper will be noted and 

comparisons to the LTV fleet will be made. 

Belt performance pflrameters 

Presently 33 States and the District of Columbia have 

mandatory use laws. These laws along with the emphasis on 

belt use by the manufacturers and safety groups have raised 

the usage rate observed in a NHTSA 19 city survey from 

less than ZOVo in the early 1980s to more than 457o in 1988. 

In addition, the requirements of FMVSS 208 to provide 

passive protection in PCs have led to the introduction of 

automatic belt systems in a large number of PC makes and 

models since 1985. Front seat occupants of PCs equipped 

with these automatic belt systems have a belt usage rate of 

about 857o. Under these conditions, the optimum design of 

the safety belt system is more important than ever before. 

Two primary functions of the belt system$ are to prevent 

or minimize occupant contact with interior surfaces and to 

maintain the occupant in the vehicle during a crash event. 

Belt system parameters which can affect these functions in a 

crash event include the time to onset load on the occupant, 

the rate of loading, and the peak load. The onset time is the 

time at which the torso belt load reaches 20O lb. The rate is 

the increase in load for each millisecond from 200 lb. to 

nominally 1,000 lb. A typical illustration of these parameters 

is shown in figure 7. These parameters have been calcu- 

P.rfcrEcdca Grilp 

Iated for all the NCAP tests. The average onset time for PCs 

is 35 ms and for LTVs 30 ms. The average rate is 47 lbs/ms 

for PCs and 58 lbs/ms for LTVs. 

2000 

@ 

1 750 

1 500 

o 

t 

tooo 

,l rro 

250 

o 

Ftsure 7. lttu'tratton "t b"ifiilH'0Tfr1","r". 

Steering assembly effects 

Even for the belt restrained occupant, head contact with 

interior vehicle components can lead to serious injuries. For 

the driver, the steering assembly is the most likely area of 

contact. In the NCAP 35 mph tests, very few of the dummies 

in the driver location were restrained sufficiently to avoid 

contact with some portion of the steering assembly. Sl%o (63 

of 205) of the driver dummies experienced head contacts 

into the steering assembly that placed the vehicle in Performance 

Group 2. For the small PC class,387a (30 of 79) were 

in Group 2, for the compacts 297o (19 of 65), for the intermediates, 

26Vo (l I of 42), and for the large PCs 167o (3 of 

l9). The LTVs show an even higher probability of severe 

head to steering assembly contacts with 46To (18 of 39 

vehicles) in Performance Croup 2.It is also noted that the 

237

average velocity change due to this contact is higher for the 

LTVs than for PCs (15.4 mph as compared to 13.3 mph). 

The severity of these steering assembly impacts is dependent 

on several vehicle performance parameters including: 

r The intrusion of the steering assembly into the 

companment 

r The area of the steering assembly which the head 

impacts and the force*deflection characteristics of 

the contact surface 

r The amount that the belt system reduces the ve- 

locity of the occupant before contact occurs 

r The effects of the structural performance that also 

can lead to differences in head contact velocities. 

In addition to these performances parameters, the driver 

location relative to vehicle interior components may contribute 

to possible head contacts. In figure 8, the distances 

from the driver to the vehicle interior are defined. These 

distances apply to the Part 572 dummy as seated according 

to the NCAP test procedure. Average values are given in 

table 5. Only minor differences are noted between the 

weight classes of the PCs. However, the head to header and 

Flgure 8. Drlver locstlon relstlvo to yehlcle lnterlor. 

Table 5. Average dlmenelonc for drlvgr locetlons. 

WetBht 

CIarr 

snsIl 

Conpact 

Intsnediate 

IrrEe 

AII PCe 

AlI LTIIg 

238 

HGad to 

Headet 

14 .50 

14. 14 

14.66 

l4. ll 

18 .04 

Drlvgr bcetfon (lncher) 

Heed to 

lJlndrh t6 Id 

19 .42 

I9. 36 

19,s8 

20.44 

19.53 

Ch6st to Knee to 

Dash 

L4.94 

14. 50 

14. 20 

l1 cc 

14. 56 

13-55 

6.12 

5.70 

6.29 

6.80 

6 .09 

6 .46 

head to windshield distances are significantly greater for 

LTVs when compared to PC$. 

From examination of the tested vehicles which have airbag 

supplementary re$traint systems, no indications of serious 

head to steering assembly impacts were found. As noted 

in the FMVSS 208 Final Rule (2), ". . . the most effective 

system is an airbag plus a lap and shoulder belt . . . Airbags 

with lap belts also provide better protection at higher $peeds 

than safety belts do, and they will provide better protection 

against debilitating injuries (e.g. brain and facial injuries) 

than safety belts". With many of the PC manufacturers 

providing the driver airbag as $tandard equipment in the 

near future. the data from the NCAP indicate that a reduction 

in head and facial injuries in frontal impacts for these 

vehicles may be expected. However for LTVs, unless manufacturers 

take action to improve the level of safety performance, 

the data indicate that head to steering assembly 

impacts for belt restrained occupants may be a continuing 

problem. 

Structural performance parameters 

The third vehicle attribute which has been previously 

examined from the NCAP data and is updated here is the 

performance of the frontal structure. During a crash, the 

frontal structure needs to provide a reasonable occupant 

compartment deceleration signature (crash pulse) and a collapse 

mode that does not allow excessive intrusion into the 

occupant space. 

The crash pulses of the NCAP vehicles are measured by 

accelerometers located in the occupant compartments. 

From these crash pulses, approximate force-crush curves 

are generated using the acceleration time data and the vehicle 

test weight. The parameters of interest from these forcecrush 

curves include the peak dynamic force, crush at this 

peak force, an estimated linear "stiffness", and the peak 

dynamic crush. The ayerage$ of these parameters and the 

test weights for PCs and LTVs are given in table 6. For PCs, 

nominally, an increasing peak force and increasing crush 

occur with increasing test weight. 

ForLTVs, significant differences are noted in all parameters 

when compared to PCs with much higher peak force and 

stiffness and lower crush. This comparison points out two 

problems: 

L In single vehicle frontal crashes, LTVs present a more 

severe environment to the occupant and a more difficult 

design problem for the safety engineer. 

2. In multi-vehicle crashes were LTVs collide with PCs, 

PCs will dissipate more of the kinetic energy, crush more, 

and have more compartment intrqsion than the LTVs. 

Table 6. Average vsluee of structural pardmeters. 

T. tSht 

chrt 

Sratt 

Coqact 

Intsil 

IrrBG 

AI1 

AII 

PCs 

LnI 

TcNt Wettht 

(Ibr) 

2J71 

3128 

3631 

4327 

3145 

3899 

Drtv€r bcEElon ( lnchil) 

P6!k Forc! 

(lbs) 

82094 

94073 

rl92e7 

12309? 

97477 

144113 

"Btlffncr4i 

(Ib'/tn) 

3795 

3848 

4509 

4259 

399 7 

t2397 

Dtsp g Psal 

Forc. (ln) 

25.8 

27.3 

29 .3 

25.2 

kxtnw Dttp 

(rn) 

27,t 

30,3 

30. I 

33, 2 

29.5 

23 .8

Recent TFends and Reseflrch Directions 

Recent trends 

The above paragraphs have presented and discussed 

some of the occupant response trends and parameters of all 

tested vehicles. It may be of interest to examine these trends 

relative to a select group of the more recent vehicles which 

have very low dummy responses. In MYs 86, 87, and 88, 33 

PCs had dummy response$ in which the HICs were less 90O 

and the Chest Gs were less than 55. These 33 PCs are 45Va of 

the tested PCs since MY 85. Of the PCs tested prior to MY 

86, only 287o had dummy responses in these lower levels. In 

comparison to the FMVSS 208 requirement$, 557o of the 

PCs in these three MYs met all requirements as compared to 

36Vo in the previous seven years. 

The 33 late model PCs which fell in the lower response 

levels are distributed evenly among the small, compact, and 

intermediate weight classes. The restraint systems in these 

vehicles include standard three point belts, two and three 

point automatic belts, and driver airbags. The manufacturers 

of these vehicles include both domestic and foreign. 

The performance of these vehicles strongly indicate the 

capability and the increasing trend of the manufacturers to 

incorporate design parameters in their vehicles which produce 

low dummy responses in the 35 mph NCAP test. 

Research directions 

Presently, FMVSS 208 requires a minimum level of protection 

for PC occupants in frontal crashes up to 30 mph. 

The lives and injuries which will be saved by FMVSS 208 

are very significant. (Note: From the FMVSS 208 Final 

Rule notice in the Federal Register (2), mid-point estimate$ 

of benefits show a reduction in all crash modes of over 6,000 

fatalities and over 100,m0 moderate to critical injuries per 

year when all passenger cars in the fleet meet FMVSS 208 

requirements). However, recent NHTSA projections of fatalities 

and injuries estimate that even after the fult application 

of the passive restraint requirements of FMVSS 208 to 

both PCs and LTVs, frontal impacts will account forapproximately 

10,000 fatalities and 120,0fi) AIS 2-5 injuries. A 

problem of this magnitude demands that we pursue research 

to address it. 

Safety problem 

FMVSS 208 has brought with it an increasing array of 

different types of automatic restraint sysrems in today's 

production vehicles. It is necessary to track the real world 

collision experience of these vehicles to better define new 

safety program directions and to refine existing program 

areas. 

As we examine the field performance of restraint $ystems, 

we will identify the safety problem at three crash 

severity ranges-low, moderate, and high speed impacts. 

Low speed impacts are defined as those below l5 mph (the 

nominal threshold for air bag deployment), moderate-level 

impacts are defined as between l5 and 30 mph (the range in 

which FMVSS 208 systems are expected to be most effective), 

and high speed impacts are defined as greater than 30 

mph. 

Real world crash data analysis will be conducted to determine 

the magnitude of the safety problem and the type of 

crash, vehicles involved, and human factors associated with 

the specific safety problems. These analyses will attempt to 

reach the following goals: 

l Setting priorities in terms of which types of crash 

events, types of vehicles, or occupant characteristics 

account for the highest percentage of serious injuries 

and fatalities. 

2. Evaluation of effectiveness of different safety 

concepts including air bag and automatic belt restraint 

systems in the crash environment. 

3. Injury mechanism identification including the 

body regions injured and the source of the injuries for 

specific safety areas. 

4. Identification of laboratory te$t conditions based 

on the crashes that account for the most serious 

injuries. 

The Fatal Accidenr Reporting System (FARS), NASS, 

State files, foreign files, and the special investigations will 

be used to conduct these analyses. 

Crash testing and sled testing 

While the above data analyses will provide useful guidance, 

it will be several years before sufficient accident data 

are collected to make definitive statements of the effectiveness 

of the restraint $ystems and the remaining fronml safety 

problem. Testing offers the most immediate means of determining 

answer$. Tests can provide early identification as to 

the possible limits of performance for a variety of restraint 

systems. 

A laboratory testing program could be designed to provide 

information in the following areas: 

I. Establishing baseline conditions for different 

types of automatic restraint systems under different 

crash conditions including different impact speeds, 

off-set conditions, and angle impacts. 

2. Determining upper bounds of the effectiveness of 

different types of automatic restraint systcm$ for different 

crash conditions. 

3. Evaluating the effect ofdifferent vehicle and occupant 

factors such as seating position, seating posture, 

seat adjustment, etc. on the performance of automatic 

restraint systems. 

Biomechanics research 

Projects in biomechanics will be directed toward improving 

and expanding injury criteria and human $urrogate capabilities. 

Present test devices could be modified to allow 

additional measurements for as$essing different injury 

modes and for extending injury measurement capabilities to 

other body regions. 

239

Component testing of restraint slstems and 

frontal interior surfaces 

As shown in the NCAP tests and as determined from data 

studies, occupant contact with the frontal interior surfaces is 

still possible even in vehicles equipped with automatic restraint 

systems. Testing in this area could establish the properties 

of restraint systems and frontal interior surfaces. 

The purpose of this data collection effort is to establish a 

baseline related to the material properties of these comPonents 

to aid in the identification of possible countermeasure$ 

and to provide input data for occupant simulation 

models. 

Vehicle structures models and occupant 

simulation models 

Vehicle structures models and occupant simulation models 

will be upgraded to allow parameter studies to be conducted 

to assist in the identification of the problems, in the 

development of test procedures, and in the development of 

potential countermeasures. 

Summary 

After ten years of testing in NCAP, significant trends 

toward improved dummy responses can be noted for the PC 

fleet. Average HIC and Chest G values have decreased by 

l0 to20To and PCs which are in Performance Group 2 (i.e. 

HIC greater than 1250 and/or Chest G greater than 70) in the 

weighted fleet have dropped from 5l7o to less than25Vo. 

437o of the tested PCs meet FMVSS 208 requirements in the 

NCAP test conditions. 

When comparing unweighted dummy respon$es across 

four PC weight classes, some degradation in safety 

performance can be noted from the large PC to the small PC- 

However, the weighted fleet indicates that the poorer 

performing small PCs are also the lower sales volume PCs. 

Within the weighted fleet, the level of safety performance of 

the small PCs may not be significantly less than the larger 

vehicles for the given test conditions. 

In examining the LTV dummy responses and comparing 

them to the PC performance, HICs and Chest Cs for the 

LTVs are approximately l0 to 207o higher than for the PCs. 

457o of the weighted LTV fleet are in Performance Group 2 

as compared toL|Vo of the weighted PC fleet. Even though 

LTVs have higher dummy responses than PCs, the 

manufacturers have demonstrated the capability to provide 

improved performance since 28Vo of the tested LTVs meet 

FMVSS 208 requirements in the NCAP test conditions. 

Vehicle design parameters associated with safety belts, 

$teering assemblies, and frontal structure are examined and 

comparisons between PC and LTV characteristics are made. 

For PCs, the average onset time for the occuPant to begin 

loading the belt is 35 ms and the average rate of loading is 47 

lbs/ms. For LTVg, these values are 30 ms and 58 lbs/ ms. 

The steering assembly effects indicate that on the aYerage 

more severe head contacts occur in LTVs than in PCs. The 

average velocity change of the head due to steering 

240 

assembly impact is 15.4 mph for LTVs and 13.3 for PCs. 

46Vo of the LTVs are placed in the higher response group 

due to head to steering assembly contacts. Only 3 l7o of the 

PCs are placed in the higher respon$e group due to these 

contacts. 

Much higher peak force and linear stiffness and 

significantly lower crush are noted in the frontal structures 

for LTVs than for PCs. The peak force is 50Vo greater for 

LTVs and the stiffness is three times that of the average PC. 

The average maximum crush is approximately 24 inches for 

LTVs and is almost 30 inches for the PCs. This comparison 

points out two problems: 

L In single vehicle frontal crashes, LTVs present a 

more severe environment to the occupant and a more 

difficult design problem for the safety engineer. 

2. In multi-vehicle crashes where LTVs collide with 

PCs, PCs will dissipate more of the kinetic energy, 

cru$h more, and have more compartment intrusion 

rhan the LTVs. 

In an examination of the late model PCs (MYs 86, 87, and 

88). it is found that 33 PCs (457o of the tested PCs since MY 

85) had dummy responses in the level I and 2 groups. The 

33 late model PCs are distributed evenly among the small, 

compact, and intermediate weight classes. All types of 

re$traint systems and domestic and foreign made vehicles 

are included. The performance of these vehicles strongly 

indicate the capability and the increasing trend of the 

manufacturers to incorporate design parameter$ in their 

vehicles which produce low dummy response$ in the 35 

mph NCAP test. 

Based on a recent NHTSA projection of fatalities and 

injuries and the existing performance of passenger vehicles 

as determined from the NCAR an outline of potential future 

research activities for enhanced frontal protection is 

presented. These activities include: 

l. A study to determine the existence and magnitude 

of any safety problems in frontal impacts after the 

implementation of FMVSS 208. 

2. A laboratory testing program which consists of 

full scale crash and sled testing. 

3. Continued projects in biomechanics to improve 

and expand injury criteria and human $urrogate 

capabilities. 

4. Continued projects in component testing to 

establish the properties of restraint systems and frontal 

interior surfaces. 

5. Application of structures models and occupant 

simulation models to extend the results of the 

laboratory test$ and to assist in the problem 

identification, te$t procedures and countermeasures 

development. 

References 

(1) Hackney, James, and Carlin Ellyson, "A Review of 

the Effects of Belt Systems, Steering Assemblies, and 

Structural Design on the Safety of Vehicles in the New Car

Assessment Program," Proceedings Tenth International 

Technical Conference on Experimental Safety Vehicles, pp 

38H13. 1985. 

(2) 49 CFR Part 571, Federal Motor Vehicle Safety 

Standard; Occupant Crash Protection, pp 28962-29010, 

Federal Register, vol.49, No. 138, July 17, 1984. 

(3) Hackney, James, and Vincent "The 

Quarles, New Car 

Assessment Program-Status and Effect," Proceedings 

Ninth International Technical Conference on Experimental 

Safety Vehicles, pp 809-824, 1982. 

(4) Cenelli, Ezio C., *'Risk of Fatal Injury in Vehicles of 

Different Size," NHTSA Report DOT HS 806-653, November 

1984. 

Inadequacy of 0" Barrier Test With Real World Frontal Accidents 

C. Thomas, S. Koltchakian, C. Thrriere, 

Laboratoire de Physiologie et de Biomdcanique, 

Associd h Peugeot S.A./Renault, 

France 

C. Got, A, Patel, 

Institut de Recherches Orthopddiques 

France 

Abstract 

The representativity of a global frontal impact must be 

assessed in relation to its concordance with the characteristics 

of frontal impacts in real world accidents in order to 

guarantee that the progress accomplished will be effectively 

translated on the road. 

The survey is based on the technical and medical investigation 

into 746 cars and 403 occupants restrained by 3,point 

seatbelts involved in severe frontal impacts with velocity 

change equal to or greater than 40 km/h. 

Precise descriptions are made of overlaps of front end 

with obstacle, deformation geometries, intrusion levels into 

the passenger compartment in terms of velocity change 

(delta-V) and mean acceleration (E m) of the car. 

The frequency and the severity of injuries to front occupants 

are analysed in terms of the relationship of the impact 

with 0o banier and 30o oblique barrier tests. 

The results show that the 0o barrier test does not reproduce 

mean acceleration values, intrusion levels or risks 

incurred by the majority of belted occupants. 

Deformations of vehicles. mean accelerations and conditions 

under which serious injuries occur with the largest 

number of belted occupants in real world frontal impact 

conditions, justify the choice of the 30" oblique barrier test 

as the most representative test. 


Frontal impact is, by far, the one causing the most deaths 

and serious injuries in real world accidents. 

That is why, as far back as the sixties, before extensive 

surveys were conducted and the principal results made 

known, it was decided in the United States that frontal 

impact against a wall perpendicular to the trajectory of the 

vehicle, herein called 0", would be the reference statutorv 

test. 

In the light of the first results of the Peugeor S.A./Renault 

accident survey, begun in 1970 (l)*, it did not take long to 

realize that at least 7IVo of frontal impacts were 

asymmetrical. 

Since the objective was to improve the secondary safety 

of cars by taking the reality of accidents into account, it 

consequently seemed logical to the French manufacturers to 

wonder about the representativity of the 0" barrier (2), 

propose a method for analysis of severity of frontal 

collisions (3), justify their preference for a test against an 

oblique wall forming an angle of 30" with the "0o barrier," 

namely I 20" with the longitudinal centre-line of the vehicle 

(4), provide accidentological justifications (5), present the 

results of the test whereby an experimental vehicle hits a 30" 

barrier at 70 km/h (6), confirm their choice in favour of the 

30o barrier (7) and even deplore not having been listened to 

(8). 

As far back as 1976, K. Friedman (9) substantiated the 

a$sertion that 9OVo of all frontal deformations were 

asymmetrical in the United States. At the same time, an 

English survey (10), based on an analysis of 143 impacted 

cars, concluded in the interest of an impact test at 50 km/h 

against a barrier with overlap of between 25 and SOVo, on 

account of the fact that 56Vo of the directions of impact are 

Iongitudinal to the nearest I0o and that 53 of the cars 

presented an overlap with the obstacle of 25 to SOVo. 

Recently, Daimler-Benz ( I I ) gave support to the interest of 

this test, taking into account a32Vo frequency of the "30 to 

507o" overlaps among the 822 vehicles of its make involved 

in frontal impacts and for which the front end overlap has 

been able to be quantified. Let us specify that this test is 

presented more as a tool for development of front end 

structure than as statutory alternative (12). 

It was therefore interesting to bring up to date and specify 

the principal data regarding relationships with experimental 

frontal tests, based on a sample of severe real world 

accidents. 

Method Used 

Our survey files on real world accidents were used as 

basis and were sorted in order to select cars equipped with 

3-point restraint seatbelts involved in severe frontal impacts 

with velocity change (delta-V) greater than or equal to 40 

24r

km/h. The estimation of the violence of the impact (delta-V 

E mean) used is that described in various publications (5, 

l3). 

The distributions of overlaps and deformation geometries 

are first of all described. These greatly contribute to the 

relationships with frontal tests such as 0o barrier and 30' 

barrier. The analysis is completed by examinations, in terms 

of delta-V of the levels of mean acceleration attained and 

the frequencies of critical intrusion into the pa$senger 

compartment. 

The risks of injury evaluated according to the AIS scale 

( l4) of belted front occupants only in this sample are studied 

in terms of the relationship with frontal tests. 

Distribution of deformations of cars subjected 

to severe frontal impact 

Since 1970, 3976 cars subjected to frontal impact have 

been analysed both from the technical and the medical 

viewpoints by the team making investigations into real 

world accidents. 

Without exclusion, this survey was conducted on all car 

accidents occurring in the same geographic zone to the West 

of Paris and causing injuries to at least one car occupant. 

This region, consisting of urban and rural zones, comprises 

different types of road networks (including motorways). 

Retrospective analysis sometimes shows that the most seri* 

ous accidents are over-represented in the sample. In particular, 

the mortality rate is double that observed at national 

scale. Let us specify that the $urvey did not cover only cars 

of French make. The share of cars of foreign make is very 

close to that observed at French national scale. 

Insofar as a truly representative frontal impact mu$t enable 

the protection offered by seatbelts to advance even 

further, it is important that the description of the accident is 

not biased by a large number of low-violence, and therefore 

not very severe, impacts for belted occupants. That is why, 

to be precise and pertinent, it is important that the subsample 

retained takes only cars into account that have been 

subjected to a delta-V at least equal to or greater than the 

threshold of appearance of the majority of serious and fatal 

lesions on belted front occupants. These occur in 85Vo of 

cases with delta-V greater than or equal to 40 krn/h, which 

already constitutes a high, but neces$ary violence threshold, 

taking into account the objective sought after. 

Besides this, the deformations must involve the front end 

over its full height. 

This amounts to eliminating 3230 cars from the basic 

sample that meet the following criteria, sorted in 

succession: 

. cars not equipped with 3-point restraint seatbelts-470 

cars hitting an overhanging obstacle (generally, 

commercial vehicle under-run)-3 I 9 

cars for which dissipated energy against deformable 

obstacle is not known-263 

*Numbers in parEnthcses designate rcferences at end of paper 

242 

cars for which opposing vehicle has not been examined-Z37 

cars involved in front-to-front impacts between 

cars "with glance-off," i.e. with complete stoppage 

of the vehicle at a distance of more than l0 

metres beyond the other car-140 

(Note: The equivalent speed against the barrieralso 

called equivalent energy speed (E.E.S)-is 

greater than or equal to 40 km/I for 8l cars. In 40 

of the 81 cases, the overlap with the opposing car 

was of the " l/4 track type"). 

Cars with delta-V estimated at less than 40 

km/h-1601 

The sample thus selected includes 746 cars whose deformation 

characteristics are quantified and the impact violences 

estimated. 

The overlap of the front end of the car with the obstacle is 

Flgurc 1. Clds$lflcatlon code lor frant damage locatlon8.

described by using a classification code (figure l) derived 

from the "Collision Deformation Classification" (SAE 

J224, March 1980). 

This improved code defines not only the overlap of the 

front end with the obstacle, but also takes into account the 

involvement of the side-member(s) in the energy dissipation 

process. This complementary data is necessary for understanding 

the behaviour of the structure undergoing frontal 

impact, the finality of which is above all preservation of 

the passenger compartment. To simplify the distribution of 

overlaps, the most frequent left-hand side deformations 

have been grouped together with right-hand side 

deformations. 

Distribution of types of overlap and obstacles encountered 

is given in table l. Distributed deformations, orthogonal 

or oblique, involving the whole of the front end of the 

cars, represent less than one third of the cases (213[7a6). On 

the other hand, cars subjected to partial overlap with the 

obstacle, involving half or more of the front end ( l/2 track 

and 2/3 track) are the most numerous. This group alone 

represents almost half (3511746) of the total sample, 

The majority of the obstacles encountered are cars (503 

cases), followed by stationary obstacles (199 cases) and 

commercial vehicles (44 cases without under-run, generally 

against commercial vehicle roadwheels). Let us note, inci- 

dentally, that more than half the cars of the sample studied 

were involved in a front-to-front collision with another car. 

The relationship with a type of test cannot only be determined 

on the sole basis of overlap of the vehicle with the 

obstacle. It is therefore necessary to complete the data by: 

deformation geometry, mean acceleration level, and to examine 

their consequences in terms of intrusion into the 

passenger companment. 

Distribution of types of overlap and deformation geometry 

are given in table 2. 

Damage is classified into four categories, according to 

whether deformation of the impacted vehicle is rectangular 

(i.e, perpendicular to the longitudinal centre-line ofthe ve- 

Tablc 1. Dletrlbutlon of cara sublected to delta-V > 40 km/h accordlng to type of overlap and type ol obstacle. 

f rmt-front 

Clrl front-rld. 

front-rur 

trrc,/brldr plrr 

(rnd 

Strtlomry 

obctaclc 

rrll 

othrrc 

col I lrlonr) 

Cmrclal vrhiclu 

Table 2. Dl$trlbutlon ol cars sublected lo delta-v > 40 km/h accordlng to type of overlap and type ol delormatlon. 

rrctrngulrr 

t tB' 

IYPEB obllq|lf 

16' to ,l{l' 

Af mkrr 

){t. 

oEF(nilTl(ta plnpolnt 

Dlrtrtbrltd 

(Do) 

Dlrtrlbutrd 

(or ) 

121 

35 

3 

6 

22 

5 

2/$ trrck 

(YeZo) 

2/3 track 

(YoZr 

) 

12E 

I 

3 

r0 

tg 

0 

IYPE OF OYENI.AP 

l/2 trrcx 

(Hz{) 

TYPEIF WERTAP 

1/2 treck 

(Yr Z,r ) 

1/3 trrok 

(LrRr) 

l/3 trtck 

(LrRr) 

l/4 trrck 

(toRo) 

l,/4 track 

(LoRc) 

c.ntrrl 

(Co) 

IOTAL 

130 l0 tl I 0 0 ta0 

E3 ta6 r11 t3 0 0 E43 

0 l5 16 30 u 0 r0a 

0 t? {6 3'o t8 IE la3 

TOTAT 213 t00 rE2 e2 l2 IE t8 

t03 65 

4l 

t4 0 

30 

I 

2 

3l 

il 

t 

37 

I 

t 

Critrrl 

(Cc) 

0 

0 

0 

TOTAT 

r6 T ta I 3 I 4tl 

TOTAL 213 r60 162 g2 12 It 7{0 

20 71 

1C 

I 

0 

434 

s3 

10 

| 2tl 

6i 

I 

243

hicle) or, on the contrary, oblique by approximately 30" on 

the one hand or more than 40" on the other. Pinpoint impacts 

against trees, poles or posts of all sizes are classified 

separately. 

It appears that 30o oblique deformations are most represented 

(343 cars), followed by rectangular deformation 

(166 cars), pinpoint deformation (133 cars) and finally 40" 

or more oblique deformation (104 cars). 

A first attempt at comparison with experimental tests (0' 

barrier, 30o barrier, impact$ against posts or others) can be 

made at this stage of the analysis. 

lmpacts related to the "0" barrier" test assume that the 

overlap is distributed over the full front end of the car and 

that the deformation is rectangular (to the nearest tlSo). 

Both these conditions are met by | 30 cars only. The mean 

acceleration levels must also be of the same magnitude as 

those recorded in the "0o barrier" tests for a given test 

speed. 

t5 

E 

I 

cl 

H 

{ 

Srs 

lil 

(t 

o { 

z 

{ 

lrJ 

I 

/ 

''/ 

' .--- I 

. / + + / l 

.: t-/t' ./ 

i i,4 +' r .," 

( i' ; ..i. ,,''. . / 

l. /ri.,..i.. . r,nt'' / 

i. / t....r .,/! ..-./ 

l' / .-. t.. ,./.'.. . J 

L: t,..' ,/ 

lr lilri:| .r. 

/ 

i. l: ,' ':/ 

l^ l: .'/ ''.^)t. 

/' 

lil::j"'' ./ 

$-'fi"" 

it!':t"" , 

l.:' ,/ 

./ 

\. " ,'/ . FEAL r*o AccroEHrB 

I / +EAHHIEHTeaTE 

V - - rso EAL ACCIDEFTTE 

-- 77 EAffiIEF rEat8 

.,J 

40 !o Eo 79 E0 EO 

TEET EFEED Of, VELOCITY CHANcE (TN Kn/h) 

Fioure 2. Dlstrlbution of the mean acceleratlon ol vehicles ac' 

co*rding to: thc veloclty change in real lmpscts rglated to 0'tesl' 

the tetl apeed for 0" barrler fests. 

The severity of these impacts in the "Velocity change 

(delta-V)/mean acceleration (a)" diagram is given in figure 

2. Comparison with the severity of the "0" barrier" tests is 

presented in the same figure by using the results of 77 

experimental tests on the most representative cars of the 

impacted vehicles. For the latter, the velocity taken into 

account is the experimental test speed, so a$ to remain 

homogenous with real world accidents, where the rebound 

speed is unknown. The result is that the cloud of points 

representing the "0o barrier" tests is off-set by approximately 

5 g towards the highest mean accelerations. Finally, 

only half(66/l 30) ofthe cars presenting a rectangular defor- 

244 

/ 

mation geometry distributed over the full front end give rise 

to mean acceleration levels comparable with those recorded 

in the "0o barrier" tests. Several reasons explain this result: 

. Only 33 of the 130 cars hit rigid, almost undeformable 

obstacles, such as walls or commercial 

vehicle roadwheels. 

r In 64 out of the 97 impacts against another car, the 

mean accelerations of vehicles in real world accidents 

are lower than in the 0" barrier test, as; 

r The hit areas ofthe opposing vehicle are less stiff 

(side doors-I7 cases----or rear end-3 cases). 

r The opposing vehicle hit front-on presents either 

oblique deformation-26 cases-or different architecture 

enabling imbrication of the most rigid 

members, without real crushing as in the 0'barrier 

test-18 cases. 

r Finally, the deformation geometry is rarely perfectly 

rectangular and symmetrical in real world 

accidents. The fact of assuming obliquity capable 

of attaining a maximum of 15" on the front end 

contributes towards increasing the maximum 

crush somewhat and consequently reduces the 

mean level of acceleration. 

Impacts related to the " 30o barrier " test are cars comprising 

an oblique deformation geometry of between 15o and 

40" and involving at least halfthe front end ofthe car. In this 

configuration, the deformation energy is dissipated by the 

centre and the side of the car. The most strained members of 

the front end unit are basically the side-member located on 

the side of the impact, body side and the power unit. 

If reference is made to table 2, it appears that oblique 

deformation geometries at 30o, with either distributed overlap 

(83 cases), 

"2/3 track" overlap (126 cases), or "UZ 

track" overlap (l I I cases), constitute a very homogeneous 

group both by types of deformation observed and by mean 

acceleration levels at given delta-V. Figure 3 gives the envelopes 

of the cases related to the " 30" barrier" test according 

to which the overlap is of the distributed type, 2/3 track type 

or even ll2 track type. A great similarity of acceleration 

levels with delta-V between 40 and 60 km/h is to be observed. 

Beyond that, for a few rare case$, acceleration levels 

are recorded that are greater than the mean connected with 

large overlap on the one hand or with impacts against obstacles 

that are relatively stiffer than the average front end units 

of the cars on the other hand. 

In orderto compare real world accidents related to the 30" 

barrier test and experimental " 30" barrier " tests, we applied 

the same method as that used in real world accidents for 

calculating the mean acceleration. The squared experimental 

test speed is divided by twice the maximum residual 

crush of the front end increased by 2OVo to take elastic 

recovery into account. Figure 4 gives the test speed and the 

mean acceleration of 8l cars tested against a 30o barrier. 

The scatter ofpoints representing the 320 cars related to 

the "30" barrier" test involved in real world accidents is 

delimited by dotted lines. The shaded area represents 48 

cars subjected to low accelerations in real world accidents.

|I 

E 

g 

tH 

F 

f; lrl 

lu 

o 

(J 

{ 

7 

{t 

ul 

I 

TOTAL OVEfI-IP (tl-Bgl 

+ +, --- E/g FHfi{T-EHD oVEH-AP (H-reE) 

r r. - t./c FHoflt-EhE ovEH-Ap N-rttl 

'f 

tt 

f .. il 

rr* 

tt+*++ +++" 

lhflrt .'t r +dit. r" 

i. ;o1*.. 'r. i ,.,, ' 

l.rlcqfr tr r.'r 

VELOCITY cHAftr (ln KrL/h, 

Flgure 3. Severlty ol lmpacts (veloclty chsng€, mean acceltration 

ol vehlcles) for 320 rsal frontal lmpacts related to 30'test. 

Apart from these cases, it emerges that, for the272 remaining 

cases, namely 85Vo of the sample of the 320 real cases 

related to the 30" barrier, reasonable adequacy exists between 

the mean acceleration levels recorded in real world 

accidents and in experimental "30" barrier" testing. 

In total, the relationship with the "30" barrier" test concerns 

320 cars, namely 42Vo of the 746 cars of the sample. 

This high percentage marks out the "30o barrier" test as the 

most representative test for restituting the deformations of 

cars subjected to a delta-V of more than 40 km[r. This result 

is due to the fact that the "30o barrier" test characterizes 

front-to-front collisions of highway reality parricularly 

well. 234 cars presenting deformations related to the 30o 

barrier test were involved in this type of configuration. This 

represents 73Eo (2341320) of the cars with deformations 

bordering on the "30o barrier" test and close on 607o 

(2341434) of the cars of the sample involved in frontto-front 

collisions (see table l). 

One single test, whatever it may be, cannot obviously 

cover the range of overlaps observed in real world 

accidents. 

Other remaining cases concern 39Vo (295t746) of all the 

cars of the sample. They are distributed as follows, by type 

of overlap (see table 2): 

t " 

ll4 track overlap" concem 72 cars, namely l0ola 

of the total sample analysed. The deformations 

involve the front roadwheel and support together 

with the entire side of the passenger compartment. 

The severity of these impacts in terms of velocity 

change and mean acceleration is generally lower 

than in the remainder of the accidents on account 

't 

I' 

E 

EH 

F 

{ 

ilI tE 

UJ 

.J 

(t 

{ 

z 

||r 

I 

t0 

+ so' ilffirEn rEBT 

- - HE L IOH-O ACCIDEHTE 0{-940) 

- to' IAffiIEF TEETA (H-el) 

so {Q EO SO 

TEET 3PEEO OR VELOCITY 

7Q gO AO 

CHAHBE (In Eil./h} 

Flgure 4. Dlstrlbutlon of lhe mean accelcrstlon ol vehlcles ac. 

cordlng to: the veloclty change ln real impactt r€lsted to 30' 

test, the teEt Bpeed tor 30. barrler lest8, 

of glance-off of the vehicle from the ob$tacle 

(above all in violent impacts) and the significance 

of residual crush. 

t "ll3 

track" overlaps are observed in 9? cases. 

The deformation geometries observed are quite 

different and could not be reproduced by one single 

test. The most frequently analysed are oblique 

and on account of this give rise to limited deformation 

of the end of the side-member. Only in the 

remaining cases (rectangular or pinpoint deformations), 

has the side-member been strained 

without bending and the amount of crushing is 

close to the maximum residual crush on the 

vehicle. 

t "l/2 

and 213 track" overlaps concern I 14 vehicles 

outside the cases related to the -'30" barrier" 

test. Here again, the deformation geometries involving 

roughly half the car are very dissimilar. 

Among these cases, the most numerous result 

from impacts against stationary obstacles such as 

large trees, bridge parapets or ditch sides. 

r Finally, centered poles concern only l8 cars. 

For the same delta-V, type of overlap and deformation 

geometry of the front end lead to quite different strains on 

the structure of the front end unit, which greatly influence 

the level of resistance of the passenger compartment. 

Frequencies of intrusion into the passenger compartment 

are analysed for the three main types of deformation encountered 

in real world accidents, which are: 

. impacts related to the "30o barrier" (320 cases), 

24s

off-set impacts (164 cases). The l/3 track and l/4 

track overlaps are covered here. 

r impacts related to the "0o barrier" (130 cases). 

During development of a vehicle, efforts are made to 

re$pect the integrity of the passenger compartment in order 

to preserve A DECELERATION SPACE for the re$trained 

occupant (and not survival space as is too often indicated)' 

Intrusion is quantified in the real world accidents of our 

sample by rearward displacement of the Iower windscreen 

header at A pillar level. As will subsequently be seen, since 

belted occupants are concerned, the risk is considerably 

aggravated when the reduction in deceleration space thus 

measured exceed$ 250 mm. Figure 5 gives the frequency of 

intrusions exceeding this value in classes of delta-V for the 

three types of most represented impacts. Critical intrusion 

frequencies differ distinctly from one type of impact to 

another. Thus, in the 51-60 km/h delta-V class, critical 

intrusions represent only 37o of the cases for impacts related 

to the 0o barrier against 29Va for those related to the 30" 

barrier and 60Vo in the case of the most off-set imPacts' It 

will be seen later that reductions in passenger compartment 

space constitute a significant limitation to the maximum 

efficiency of seatbelts. That is why it is important that a 

representative frontal impact used within a statutory frame- 

246 

eroportions (%) of cars 

with paeeenger 

intrueion 

compartment 

eb 

work verifies, among other things, the resistance of the 

passenger compartment. lt is undeniable that a test with 

overlap ofthe " l/3 track" type would present the advantage 

of concentrating efforts on thi$ single aspect provided that 

protection criteria are met. 

But solutions would risk being adapted to a low proportion 

ofcases, taking into consideration the $tatistic reality of 

accidents. In fact, out of 100 "critical" intrusions (rearward 

displacement of lower windscreen header > 250 mm), the 

largest number (38) occuned with impacts related to the 

"30" 

barrier" against 3l with 

"l/3 

or l/4 track" off-set 

impacts, 6 with impacts related to the 0" barrier and finally 

25 distributed among the various types of remaining 

deformations. 

Finally, the 30" banier test presents the following advantages 

from the sole viewpoint of accident analysis: 

r higher proportion of cars involved in severe impacts 

(3201746, namely 42Vo intotal), 

. realistic mean acceleration, restituted to 857o by 

the experimental test, 

. asymmetrical deformation of the car giving rise to 

a significant risk of critical intrusion with delta-V 

between 5l and 60 km/h, 

r larger number of cases of critical intrusion 

d"# 

","" 

q$ 

""-' 

-or": 

-v 

delta-V 

40*50 51-60 61-70 >70 

Figure S. Frequenclea oJ Intruslon (resnrard dlaplacoment ol the lowor windscreen hoader at A pillar level > 250 mm) by delta'V 

clCsses for thb throe doformatlon typeE' 

( in 

km/h)

Belted occupants involved in severe frontal 

impacts 

This second part aims to specify the risks incurred by 

belted front occupants in road accidents in terms of the 

different relationships with experimental frontal tests. 

A sample of 403 occupants taken from the first part of the 

survey was selected according to the following criteria; 

r front seat occupants without overload from rear 

pa$senger, 

. wearers of inertia reel seatbelts (or correctlv adjusted 

static seatbelts), 

r delta-V between 40 and 70 km/h. 

The distribution of the global gravity of lesions according 

to the place occupied in relation to the type of related impacts 

(table 3) reveals that: 

r Impacts bordering on the "30o barrier" test embrace 

the majority of recorded serious victims 

since they constitute 43Eo (39191) of the seriously 

injured (M.AIS 34-5) and half of those killed 

(10/20). 

r Impacts related to the "0" barrier" test represent 

23Vo of the seriously injured and concem 4 of the 

recorded 20 killed. 

r Gravity and mortality rates of occupant$ per types 

of related impact are similar, apart from off-set 

impacts with "l/3 or l/4 track" overlap, where 

they are less. This is due to the absence ofcases in 

the 6l-70 km/h delta-V class on account of the 

significant glance-off of cars in relation to the 

obstacle. 

Aggravation of risks connected with intrusion at equivalent 

delta-V appears clearly in the analysis of real world 

Table 3. Dlttrlbutlon of slobal sravlty ol leglons for 403 bcltsd 

occunant$ Involved In sEvere liontallmpacts (40 s delta-v < 70 

tm/h) accordlng to placeoccupled In relatlon to typssof relatBd 

lmpEcts. 

''t/1 

R. I rt.d 

lructt 

"rc." 

.t 

Ut-lricl- 

gfiD|rt 

ALIOdETHER 

r.AIS MIYERE FfrOT.T 

PA88ESER8 

0- t-2 

3-{-6 

xt t l.d 

TOTAL 

rr-2 

3-'a-6 

(t I l.d 

IOTAL 

(Fl-? 

3-4-6 

Kt I lnl 

T0rA[ 

G-t-2 

3*,Fs 

Xl I lill 

roTAl, 

0-r-2 

F,FE 

Xl I l..l 

TOIAT 

3,4 

r6 

I 

87 

26 

I 

6t 

6 

2 

23 

13 

I 

rec 

0e 

r5 

60 

121 

62 

'all 

273 

t3 

6 

3 

40 

t3 

2 

21 

6 

o 

t0 

a 

0 

t0 

et 6 

2Z 

s6 

33 

20 

r3l, 

IOTAL 

11 

2l 1 

tl 

te? 

3t 

10 

l?c 

t9 

l4 

2 

3t 

rl a 

c6 

60 

2t2 

el 

t0 'altJ 

accidents, while it does not pre$ent any major problem in the 

0" barrier experimental tests. 

Intrusion is quantified by the rearward displacement of 

the lower windscreen header closest to the occupant. This is 

qualified as "critical" when the displacement exceeds 250 

mm. 

A "critical" intrusion has been recorded for only l1Vo of 

the occupants (see table 4). All delta-V's intermingled, it 

concerns 6Vo of the slightly injured, 25Vo of the seriously 

injured and 807o of the killed. As from a delta-V of 5l-60 

km/h, the number of serious victims with critical intrusion 

( l3 seriously injured and 4 killed) is almost as great as when 

the intrusion is moderate or nil (19 seriously injured and I 

killed). 

Tsble 4. Dlatrlbutlon of 403 bslted occupsnts by claaaea of 

della-V ln relatlon to lhe Intruslon. 

Intruttd laval 

(rildrrd dlsle 

cnnt ol lffir 

Ylndrcr[n hrrdar) 

Lll or rddarrta 

(( 2S0 r) 

c/tttfil 

(re60:) 

oYrrrl I 

ravarlty 

r.AI8 o-r-e 

x.AI8 3-4-6 

Xt I lrd 

TOTAT 

t{.4I8 Ft-2 

l.AI8 3-{-6 

Kt I l.d lolAL 

o.lt&V (ln h,/h) 

rO-60 51-t0 cr-70 r0rAL 

20r 

2A 

t 

236 

f 

I 

0 1{ 

56 

r0 I 

76 

t3 4 26 

I 

ID 

? 

32 

z 

, 

t2 2t 

f--'l pasaenger comPartment 

Le8enq : U lntruBiond aso mm Nr?:i:::i::;Tgil:t."' 

.I1 L\\\\\I 

:t9. 

I NSSSSN 

40-50 

CIaE66a of 

."^ N\\\ 

I N\\\\ 

5r-60 

vclocity chenge 

( in kh/h) 

t?6 

ia 

a 

t7 

zl 

t0 

61-70 

Flgure 6. Gravlty rate by classee of veloclty change lor 403 

belled front occupanls accordlng to passenger companment 

lntruslon. 

Aggravation of the risk connected with intrusion at comparable 

delta-V is particularly obvious when the gravity 

rates (M.AIS l3limplied) presented in figure 6 are examined. 

In the 40-50 km/h and 5 140 km/h delta-V classes, the 

gravity rates in case of critical intrusion are respectively 

multiplied by 4. I and 2.8 compared with occupants undergoing 

nil or moderate intrusion. The 1z test shows that the 

differences observed are highly significant at the .05 threshold. 

In the 5l-60 km/h and 6l-70 km/h delta-V classes, the 

mortality rates (killed/implied) are multiplied by l0 when 

the occupants suffer "critical" 

intrusion, but the difference 

is not significant, on account of the low numbers involved. 

247

Comparison of risks with impacts related to 

tt0o barrier" and tt30o 

barrier" tests 

To be strict, comparison of the global gravities of lesions 

in the two samples should take into account the three physical 

parameters recorded at the time of real world accidents, 

which are: velocity change (delta-V); mean acceleration 

(fl); intrusion level (critical or not). 

Figure 7 gives this information for drivers according to 

whether the impact is related to the 0" barrier or the 30" 

barrier. On this side of 60 km/h. it is to be observed: that no 

death occurs in either sample; that for the seriously injured, 

the risk depends either on the mean acceleration level (0' 

banier), or because of the fact of critical intrusion (30o 

barrier). 

it 

c18 

g 

E 

cl 

H 

Irg 

E 

tg 

d IJ 

(, 

< 8 

z 

il 

I 

BlE 

C 

g 

E tTI 

rt 

G 

ltl 

IT 

t 

S e 

T 

H 

Floure 7. M.AIS of belted drlvers according to Yeloclty change! 

m6an acceleratlon of vehicle and intruslon for impact$ related 

to 0'test or 30'test. 

The lowest gravity rates are recorded in the absence of 

intrusion. They are.l0 (8/80) and.24 (10/41) respectively 

for drivers involved in impacts related to the 30" barrier and 

the 0" barrier. This difference is explained by a lower level 

of mean acceleration (approximately 4 g) to the benefit of 

drivers involved in "30" barrier" impacts. 

On the other hand, this advantage is completely wiped out 

in impacts related to the 30'barrier when the driver suffers 

critical intrusion. The gravity rate attains .36 (5/14). The 

difference with drivers related to the 30'barrier not suffer- 

248 

IIfACTS ffi-ATED TO O.TEST 

Fffi ELTEII OFIVEFS. 

. t AItr o-l-E 

+ HAIS 9-rl-6 

T KILLEP 

VELOCTTY CHAHEE (Tn K[/h) 

II#ACTS FELATEII TO 30. TEST 

FM EELTED OFIVERS, 

. XAIE o-t-e 

+ HAIE E-4-E 

x KILLED 

o lrlTr{ INTFUEIO}I 

YELOCITY Cl{Al*lEE (rn Krn./h) 

ing considerable intrusion is highly significant at the .05 

threshold (x2 = 6.61 ). 

With delta-V between 60 and 70 km/h, in spite of mean 

accelerations ofbetween l5 and 2l g, three out ofthe nine 

drivers involved in impact bordering on the 0" barrier did 

not suffer serious injury. The only killed was 42 years of age 

in a car subjected to considerable passenger compartment 

intrusion. Once again, for drivers whose impact is related to 

the 30" barrier, the balance is connected with presence or 

absence of critical intrusion. 

Out of l4 drivers, victims of critical intrusion between l0 

and 14 g mean acceleration, 8 were killed, 5 seriously injured 

(M.AIS 3-4-5) and only I was slightly injured 

(M.AIS l-2). In the absence of intrusion, the balance is 

more favourable since out of the l2 drivers concerned, the 

records show 0 killed, I seriously injured and 4 slightly 

injured. Since front passengers were involved (see figure 8), 

no case ofdirect intrusion took place on the occupant on this 

side of a delta-V of 60 km/h, either among cases related to 

the 0" barrier or among those related to the 30" barrier. The 

gravity rates of these front passengers are .16 (7 144) in cases 

bordering on the 30' barrier and higher, .23 (417) for cases 

assimilated to the 0" barrier. These rates are very similar to 

those observed for drivers under the same conditions (.10 

and .24 respectively). 

With delta-V between 60 and 70 km/h, high gravity is to 

6 

c l E 

6H 

Ers 

tr 

UT 

UJ 

.J 

o 

r t B 

z 

4 

lrl 

I 

6rE 

t 

g 

z 

Ers 

{ 

fi 

lrJ 

IJ 

S e 

z 

{ 

trl 

E 

IIfACTS FEIATED TO O. TEST 

FOH BELTED FHONT.PASSEI{GEtri. 

VELOCITY CHANGE (TN Kfi./N) 

IHFACTS HELATED TO S{T.TEST 

FOH BELTED FtrT{T-PASSE}EEFS. 

r l 

. HAIE o-l-e 

+ HAIE E-4-E 

|. KILLED 

O HTTH IHTruBION 

VELOCITY CHANBE (IN KNlh) 

Fioure L M.AIS ol belted front-passengers according to veloc' 

Itv-chanoe. mean accelerstion bf vehicle and Intrusl-on lor impicts 

related to 0'test or 30'test.

e observed (l killed and 4 seriously injured) for the 5 

pas$engers involved in impacts related to the 0o banier. This 

gravity appear$ less with the I I passengers involved in 

impacts related to the 30" barrier, who did not suffer intru* 

sion (l killed, 5 seriously injured and 5 slightly injured) on 

account of lower mean accelerations. 

Distribution of injuries by bodily area 

Non-minor injuries (AIS > 2 and AIS > 3) are given by 

bodily area (figure 9) for 261 ofthe 273 drivers and 126 of 

the 130 front passengers. 

fi otnrr, 

'.] 

':J 

DRIr/ERS (N- 261) 

| 

,*n""r" rcr.t;d to 30r test [l :lH::t 

LJ oo tcat 

tTHORAX I'PPER -ffih 

DOR$O ABITo|IEN PELVIS I4UEN 

LIXBS LTITBAR 

SPINE 

LD,qS 

FROI{I PA$$ENGEN$ 126 ) 

*%ru-ffi-k LIEA,D H8tr ?HORN( Dofi3o rmmx pnlvrs 

UPPEN 

LIXBS 

LU;BAN 

SPUC 

ITgER 

LIIGS 

Flguru 9. AIS l 2 snd AIS r 3 Inlury dlatrlbutlon$ for belted 

drlvera and F.F. pas8angers by bodfly ar6as accordlns to the 

slmllarlty to 0. or 30" bariler terts (frodtal lmpEcts lrom:10 to 70 

km/h of delta-v). 

The exact balance of lesions arising at the time of immediate 

death is not known, failing authorization to perform 

autopsies. This concerns l2 drivers suffering critical intrusion 

(of which 6 in impacts related to the 30o barrier) and 4 

passengers (2 in impacts related to the 0o barrier without 

intrusion and 2 in impacts related to the 30" barrier of which 

I with intrusion). Consequently, it must be bome in mind 

that the frequency and the gravity of the injuries to drivers 

suffering critical intrusion is under-estimated in what 

follows. 

Among drivers, the majority of the lesions of moderate 

severity or more (AIS > 2) involve the head followed by the 

limbs. On the other hand, regarding injuries AIS > 3, the 

classification is reversed. This is explained by the very high 

frequency of AIS 2 lesions to the head, such as brief loss of 

consciousness or simple fracture of the maxillo-facial area, 

while open fractures of the limbs with AIS 3 rating are 

relatively more numerous- 

For belted passengers, the abdomen followed by the thorax 

constitute the bodily areas most frequently attained by 

serious lesions. 

Globally, both with drivers and passengers, the share of 

injuries arising from impacts related to the 0" barrier is fairly 

stable (between 20 and 307o) from one bodily area to another, 

apart from a few rare exceptions. This finding in fact 

hides considerable differences in risk connected with the 

types of related impacts. 

Table 5 gives the AIS > 2 and AIS 2 3 injury rates by 

bodily area according to the place occupied in terms of the 

related impact according to whether or not there was critical 

intrusion into the pas$enger compartment. On account of 

the low numbers in certain categories of cases with high 

intrusion, 3 drivers in impacts related to the 0" barrier and 7 

passengers (2 in 30" barrierand 5 with otherclassifications) 

do not appear. 

Tsble 5. AIS > and d'AlS > 3 rates bv bodllv area of belted 

drlv€rs and lront Dasaanq€ra accordlno to tvoe of frontsl 

lmpact and Intrusldn (frontll Impact$ : c0-kmlh-i delta-V < 70 

km/h). 

r H€rO 

r IECX 

. TTERAX 

r (rFfEF I 

Ll|fs) 

. otnSo 

LUfAfi 

SFIItE 

r A8tldlEll 

r PEIVIS 

r L(IER ) 

LtnS) 

r HCIO 

. tEcx 

I rHnAI 

r UeeEr ) 

LIr.08) 

.|Xnflt ) 

Luita ) 

8FiltEt ) 

r AEllOa€l 

. PELYIS 

r L(ilER ) 

LIr.S8) 

A - AI8 f2 

atiIVERS 

,"t,o'ilT# Oth.il 

(n=s) 

(n=e3) 

| 

.50 l.re 

0 l.0r 

,t6 | .06 

'ei 

I' 

.0e .0, 

| 

.116 | .01 

ii 

11.00 

ll 0 

ll .04 

ll .04 

tl 

tl 

ll.0r 

tl 

ll:li ll .re 

tl 

| :ll 

( n:t{t) 

.22 

.02 

.03 

.t2 

,01 

,02 

,01 

,t6 

0 

.01 

0 

.03 

0 

.01 

o 

.05 

Crlt lc.l 

Intrurl m 

l.ll t0' 

(n:tl) 

. 'ai 

.0,1 

. t4 

.tt 

o 

. ta 

.m 

. tl 

.06 

. t'a 

.tl 

0 

.t4 

.11 

.71 

.tt 

0 

.06 

Ffio*r P|sSErlEtS 

tll or ld tntrutto 

|!ll 50' 

(n'61) 

,zz 

0 

.tt6 

,t4 

otharr 

( n'at ) 

,tt 

,0e 

,l5 

'1f 

The lessons to be drawn from this table are the following, 

per bodily area: 

t Head: The AIS > 2 lesion rate with drivers is 

comparable in real world accident related to the 0' 

barrier (.50) and in impacts with critical intrusion 

(.45 and .50). The rates of serious and fatal injuries 

(AIS > 3) are higher in the cases of critical 

o 

0 

.06 

.1t 

. tt 

0 

.tl 

0 

.06 

0 

.10 

.0'l 

.lo 

.0e 

0 

.04 

,06 

o 

.10 

.06 

.12 

.04 

.02 

.02 

0 

,02 

.0i 

0 

,02 

,02 

.02 

249

intrusion (.14 and .19) than in cases related to the 

0o barrier (.06r. 

t Thorax: The highest AIS > 2 and ) 3 lesion rates 

are observed for passengers involved in impacts 

related to the 0o barrier. 

. Upper /imbs; The highest risks are observed in 

cases of critical intrusion. 

t Abdomen.'The lesion rate is only lower with drivers 

involved in asymmetrical impact without 

intrusion. 

r Pelvis: The risk is only really present with drivers 

involved in impacts related to the 30" barrier with 

critical intrusion. 

t Lower limbs:The AIS > 2 or 3 risks are only very 

high among drivers suffering critical intrusion. 

In short, for all bodily areas (except the thorax), it appears 

that the risks are always higher when the driver is victim of 

critical intrusion into the passenger compartment. These 

risks exceed those observed in impacts related to the 0" 

barrier, which themselves remain higher than the gravity 

rates recorded in other impacts where there was no significant 

reduction of the pa$senger compartment space. 

Let us finally point out that 47Vo (68/146) of all the 

serious and fatal injuries (AIS E 3) are observed in impacts 

related to the 30o barrier against 24Vo in impacts related to 

the 0o barrier and 29Vo distributed among the various remaining 

types of impact. 

Discussion 

The 0" barrier test i$ widely used in both American and 

European standards. It is primarily supported by the 

NHTSA (15), which estimates that a 0o barrier test is 

preferable to a test against an oblique barrier insofar as the 

latter may constitute an obstacle to the development of 

inflatable air-bags. J. Hackney ( l6), taking the NCSS file as 

basis, adds that the risks connected with directions of 12 

o'clock impacts with deformation distributed over the 

entire front end are higher in real world frontal impact. Our 

survey shows that this is not exact in the case of belted 

occupant$ victims of a significant reduction in passenger 

compartment space. 

But what is desirable for any regulation is moreover 

absolutely essential for a system of inter-classification of 

vehicles, especially if its objective is comparative data on 

potential vehicle safety. In the United States, prediction of 

the safety level of tested cars is based on impact against 

orthogonal barrier within the framework of the New Car 

Assessment Program. The explanation of the divergences 

between this program, also called "Crashworthiness 

Rating," and real world accidents has been sought in terms 

ofrepresentativity ofthe test and consistency ofthe criteria 

measured on dummies with reference to the injuries 

observed in real world accidents (17, 18, 19,20). It appears 

that recourse to an asymmetrical test of the impact against 

"30" 

barrier" type and the addition of supplementary 

protection criteria concerning the quality of pelvic restraint 

250 

and facial protection are minimal improvements to be made. 

In Europe, at statutory scale, two initiatives $upporting 

the 30o barrier test were taken in the pa$t but were not 

concretized. As far back a 1974, the European Committee 

for Experimental Vehicles (21) declared itself in favour of a 

global asymmetrical test (30" barrier or 50dlo off-set banier). 

The other initiative is a draft regulation for global frontal 

impact with test against 30'barrier drawn up in 1983 under 

the care of the European Economic Commission. This 

document (22), adopted by a group of vehicle 

manufacturing experts, wanted to con$titute a statutory 

altemative to the often ancient regulations at present in 

force. 

The asymmetrical test against off-set banier with overlap 

in the region of 40Vo above all aims to verify that the 

behaviour of the front end structure of the car does not lead 

to critical intrusion without however resorting to 

rigidification of the front unit of the car to such an extent 

that the tolerances of the occupants risk being exceeded, 

especially in symmetrical impacts. Analysis of real world 

accidents clearly indicates that efforts must be made to 

attain this object. However, it is necessary to emphasize 

here that the largest number of critical intrusions is not 

observed in impacts of the " l/3 or l/4 track" type, but occur 

in cases where overlap with the obstacle is greater than or 

equal to half the front end of the car. 

The interest of a statutory te$t i$ to propose one single 

global procedure aiming to faithfully restitute the attendant 

risks. connected with intrusion on the one hand and 

accleration on the other hand, suffered by the majority of 

seriously injured and killed belted occupants. This means 

that the deformations of the vehicle must be programmed 

taking two complementary criteria into consideration: 

r integrity of the passenger compartment under 

asymmetrical impact conditions, 

r compliance with protection criteria measured on 

dummies. 

With more than 4OVo of serious victims involved in 

impacts related to the 30" barrier, it appears that this test 

constitutes the berrt compromise that can be made to 

advance the protection offered in frontal impact. 

Since the subject was voluntarily limited to the study of 

real world accidents. reference should be made to the work 

of G. Stcherbatcheff for a comparative survey of different 

frontal te$ts at experimental scale (23). 

Conclusions 

Analysis of a sample comprising 746 impacted vehicles 

ofall makes and 403 front occupants restrained by seatbelts, 

involved in velocity changes (delta-V's) equal to or greater 

than 40 km/h shows that: 

. Only lTVo of the cars present deformation related 

to the 0" barrier and mean accelerations 

undergone by these vehicles are less than those 

obtained with the 0o barrier test in half the cases.

The strict relationship to the 0o barrier test is 

therefore only observed for less than l07o of 

frontal real world impacts. 

r Killed and seriously injured belted occupants in 

impacts related to the 0" barrier represent less than 

a quarter of all the serious victims of the sample. 

r Risks of AIS E 3 lesions to drivers are lower in 

impacts related to the 0o barrier than in impacts 

with critical intrusion into the passenger 

compartment observed in asymmetrical impacts. 

r Majority of the cars involved in real world frontto-front 

road accidents present deformations and 

mean accelerations comparable with those 

observed against the 30o barrier. 

. Impacts related to the 30o barrier represent 427o of 

the cars studied, 43Vo of $eriously injured belted 

victims and half the killed belted victims. 

r With delta-V between 40 and 60 km/h, intrusion 

multiplies the risk of being seriously injured by 4. 

r Largest number of severe intrusions into the 

passenger compartment is recorded in impacts 

related to the 30" barrier. 

In short, it appears that the global frontal impact test 

against 30o barrier effectively constitutes the best 

compromise that can be made to translate risks of intrusion 

and the exceeding of tolerances observed in reality with the 

largest number of seriously injured belted victims within 

one same test. 

References 

(l) F. Hartemann, "Enqu€te midicale et technique sur les 

accidents de la route", Inginieurs de I'Automobile, 1972, 

n" 12 

(2) C. TarriEre, "De la rCalitd des blessures aux critEres 

pris en compte dans les tests de sdcurit0", Ingdnieurs de 

I'Automobile, 1972, n" 12. 

(3) P. Ventre, J. Provensal, "Proposition d'une mCthode 

d'analyse et de classification des sdvCritds de collisions en 

accidents riels", IRCOBI Conference, Amsterdam, June 

26.1973. 

(4) P. Ventre, "Proposal Methodology for Drawing-up 

Efficient Regulations", 4th International Congress on 

Automotive Safety, San Francisco, July l,t-16, 1975.; 

(5) C. Tarribre, A. Fayon, F. Hartemann, "The 

Contribution of Physical Analysis of Accidents Towards 

Interpretation of Severe Traffic Trauma", l9th Stapp Car 

Crash Conference, San Diego, November l7-19, 1975. 

(6) R.H. Arendt, "Crash Test Performance of Renault 

Basic Research Vehicle",6th ESV Conference, Washington 

DC, October 12-15, 1976. 

(7) J.A. Desbois, *'Peugeot Status Report", 6th ESV 

Conference, Washington DC, October l2-15, 1976. 

(8) P. Ventre, "Renault Status Report", 6th ESV 

Conference, Washington DC, October l2-15, 1976. 

(9) K. Friedman, "Phase II RSV Accident Analyses 

Techniques", 6th ESV Conference, Washington DC, 

ocrober l2-15. 1976. 

(10) B.S. Riley and C.P. Radley, "Traffic Accidents to 

Cars with Directions of Impact from the Front and the 

Side", 6th ESV Conference, Washington DC, October 12- 

15. t976. 

(ll) F. Zeidler, H.H. Schreier and R. Stadelmann, 

"Accident Research and Accident Reconstruction by the 

EES-Accident Reconstruction Method", SAE Congress, 

Detroit, Michigan, Feb. 25-March I, 1985. 

(12) L. Criisch, K.H. Baumann, H. Holtze and W. 

Schwede, "Safety Performance of Passenger Cars 

Designed to Accommodate Frontal Impacts with Partial 

Barrier Overlap", SAE Congress, Detroit, Michigan, Feb. 

Z7-March 3, 1989, SAE Paper 890 748. 

(13) Laboratory of Physiology and Biomechanics 

Associated with Peugeot S.A./Renault "Assessment of 

Crash Severity", Published at the Workshop on assessment 

of crash severity, Cothenburg, Sweden, September 1984. 

(14) American Association for Automotive Medicine, 

"Abbreviated Injury Scale-l980 revision". 

(15) NHTSA, "FMVSS 208", Docket74.l4, Notice 38, 

April 1985. 

(16) J. Hackney, "Comparison of 0o and Oblique Tests", 

l0th ESV Conference, Oxford, England, July l-4, 1985. 

(17) J.R. Stewart and E.A. Rodgman, "Comparisons of 

NCAP Crash Test Results with Driver Injury Rates in 

Highway Crashes", Final Report, October 1984, H.S.R.C.' 

University of North Carolina, Chapel Hill, N.C. 27514, 

U.S.A. 

(18) C. Thomas, F. Hartemann, J.Y. Fordt-Bruno, C. 

TarriEre. C. Chillon. C. Hufschmitt, C. Got and A. Patel, 

"Crashworthiness 

Rating System and Accident Data: 

Convergences and Divergences", SAE International 

Congress and Exposition, P14l, Advances in belt restraint 

systems, Detroit, Michigan, U.S.A., Feb. 27-March 2, 

1984. 

(19) J. Provensal, F. Hartemann and C. TarriEre, "Five 

years of New Car Assessment Program: balance and current 

conclusions 

", Proceedings of l0th Intemational Technical 

Conference on Experimental Safety, Oxford, England, July 

14. 1985. 

(20) C. TarriBre, "Comment amCliorer I'addquation 

entre la protection Cvalu€e dans les tests et la sdcuritC rdelle 

con$tatde sur la route", Proceedings of lrr'Eme Confdrence 

Canadienne Multi-disciplinaire sur la Sdcuritd RoutiEre, 

Montrdal, 26-28 mai 1985. 

(21) H. Taylor, "EEVC Status Report," 5th Intemational 

ESV Conference. London. June 1974. 

(22) Nations Unies-Commission Economique pour 

I'Europe, "Projet de rdglement sur le choc frontal global", 

TRANS/SCI1WP29/R237, Revision I du 24 ao0t 1983. 

(23) G. Stcherbatcheff, "Frontal Crash Tests and 

Compatibility", l2th ESV Conference, Gothenburg, 

Sweden. Mav 29-June 2. 1989. 

251

Reference Frontal Impact 

G. Stcherbatcheff, R. Dornez and D. Pouget, 

Renault 

Abstract 

Automotive structural design is influenced by the choice 

of a method for evaluating the protection level provided by a 

vehicle in frontal impact. This choice influences the vehicle's 

performance in real accidents. 

This paper compares different reference frontal collisions; 

offset car-to-car impact with 507o overlap on the left 

side; 30' angled barrier impact on the left side with or 

without anti-slip system; offset half-barrier impact with 

45Vo overlap on the left side; and impact against 0" angled 

barriers. 

I clc (rr lo rf Yitlr dftttl 2 

3 

f iltld !*tiff 

r{||| ifi t{t 

I. 

lN AC 

5 .rr: 0.fltlad brrft vitt clfrrl 6 ]AB 

FIgure 1. lmpact conflgulallons. 

252 

IJJJJJJJJJI 

l0r rdr{ rnr. 

lfx#d brri.r 

4rl xti rlit rltfrE 

ts ||d|{ lrrhr 

These different impact configurations are compared in 

terms of $tructure and test dummy injury criteria by main 

component analysis of the data. The configurations are 

evaluated from the viewpoint ofrepresentatives ofreal road 

conditions. The question is posed as to their influence on 

compatibility, especially in side collisions. 


Thirty-nine frontal collisions against various barriers and 

car-to-car were performed and analyzed. The vehicles 

selected are five recent models of the Renault range. The six 

configurations, tested at 56 km/h, are shown in figure l. 

The first configuration, impact between two identical 

vehicles, aims at simulating a category of collisions 

representing 60Vo of severe injuries in frontal impact. The 

impact against a 30o angled barrier corresponds to a planned 

European standard. The offset half-barrier is a barrier used 

by certain laboratories and car makers for the development 

of their vehicles. The impact against a 0" angled barrier 

corresponds to European Standard ECE l2 and U.S. 

Standards FMVSS 204 and 208, the statutory speed being 

48.3 km/h instead of 56 km/l which is the speed taken into 

account in this study. 

Methodology 

For each of the 39 tests. 22 factors were selected 

describing the test conditions, structural deformations and 

test dummy injury criteria. 

The 858 (39 X 22) data items are presented in reduced 

form in table l. Table 2 shows the mean values of these 

factors for each test configuration. 

Above all, however, this data was subjected to analysis 

with the SPAD software (Portable System for Data 

Analysis) (1).* 

In view of the number offactors adopted and the scatter of 

certain factors, 39 tests may seem a limited number for 

precise analysis. However, it will be noted that the software 

used provides indicators allowing evaluation of the analysis 

validity level and also shows factors poorly represented in 

the data set and for which no conclusions can be reduced. 

It is clearly advisable to increase the data set with time so 

as to verify that the conclusions drawn from the 39 tests are 

confirmed. 

Factors flnalyzed 

All the tests were performed at a speed of 56 km/h, which 

was a constant in the study. The slight fluctuations in test 

speed around 56 kmlh do not affect the study results, and 

this was verified beforehand. 

The first two factors listed below describe the test conditions, 

while the other 20 factors indicate the te$t results. 

N.B.: Factors l, 2 and 4 are qualitative (or illustrative) 

factors which do not contribute to calculation of the main 

axis, unlike the other factors which are quantitative. However, 

these qualitative factors, when projected onto the main 

*Numbers in parentheses designate rcfetences at end of paper.

Table 1. Test reaults (for each quantltatlve fsctor, the mean calculated on all tests ls equsl to 100). 

i 

I 

I 

t 

I 

I 

I 

I 

I 

a 

I 

t 

I 

I 

t 

t 

t 

I 

{ I 

{ 

J 

t 

I 

5 

I 

I 

I 

I 

I 

I 

;s. 

2 H 

i 

I 

I 

I 

s 

I 

I 

5 

I 

I 

I 

t 

I 

, 

I 

a 

I 

I 

I 

I 

{ 

I 

8 

t 

I 

I I 

t 

I 

I 

a 

{ 

r 

f 

a: 

fl 

v 

] 

r0i 

fl 

E 

t 

ll 

ti 

] 

tl 

lll 

rll 

T 

ll 

T 

tz 

?{ 

l0l 

tl 

1 

t 

!a 

E 

I 

at 

rol 

td 

tol 

It 

lt; 

] 

nl 

l+ 

llaad 

l. Ulnl lrr 

I 

Y 

Y 

i 

I 

T 

I 

Y 

I 

f 

f, 

I 

I 

t 

I 

Y 

I 

I 

Y 

t 

t 

I 

t 

I 

I 

I 

t 

Y 

I 

I 

I 

r 

t 

I 

t 

t 

rl 

il 

r01 

J' 

I 

Itl 

t 

rd 

I 

I 

7l 

tl 

tl 

E 

tlt 

lf, 7l 

T I 

lg 

I 

rta 

t? 

r0l 

IC 

T 

l: 

7l 

I 

td 

IT 

ul 

tfl T 

tl 

tot 

t0l 

Tfl 

lol 

n 

ll 

rfi T 

til 

I 

I 

I 

?t 

?l 

ltt 

lf, 

.rl 

I 

Ir( 

l0l 

lol 

n t 

T 

n 

F TT 

rtl 

lll 

ls 

l0t 

t 

rd,l 

Ill 

l0t 

rl 

lol 

la 

rl 

fl ll 

tl 

ll 

il 

f( 

I 

lu il 

t{ 

T 

la 

tl 

at 

DI 

lll 

r.l 

rl 

It 

ta 

lf, 

tl 

tl 

ru 

ll 

rI 

n 

rl 

tl 

ltl 

H 

aa 

lf 

ul 

tal 

! 

Itl 

It 

lll 

fi 

fl 

J 

t3 

{t 

lf 3 

fl 

n, 

3 

t, 

! 

I 

t: 

?l 

lt, 

ll 

t 

f 

t 

lr 

l0l 

ll: 

a 

tl 

al 

tl 

a fl 

frt 

Pralc 

'I 

|. 

ll 

fl 

It 

tl 

|| 

a 

fl 

IC 

fi 

{ 

ll{ tt 

lll 

Itl 

?l 

I 

tl 

lq t 

an 

tl 

rll 

lol 

5t 

ttI 

?l 

T 

t 

t! 

l?l 

ttl 

lrr 

lrI 

rll 

tt! 

llr 

t 

ill 3 

3 t 

t 

t 

t 

I 

E 

E 

t 

ll1 

lll 

l1 

r0l 

I 

lol 

l0r 

l0l 

li 

llr 

lc 

E 

l{l 

tl 

It 

ll 

l. 

U( 

ul 

ttl 

tn 

I 

tl ! 

ll 

? 

! 

||l 

It 

It 

tl 

ll{ 

l{t 

til 

lt 

It 

ltl 

I$ x 

tr 

CT 

t?l 

rfl 

Itr 

I 

II 

t{ 

ltl 

u 

rt( 

tl 

ul 

H llt 

lo{ 

al 

tl 

! 

l 

tl 

n 

L 

|8! 

l0l 

?l 

t 

l0l 

tl 

|. 

|{ 

x 

lrl 

lll 

tl 

t 

t 

T 

to{ 

s{ 

tl 

I 

tol 

77 

It{ 

fl 

tl 

lc t5 

lx 

ul 

tl 

ut 

ull 

t4 

qr 

t 

tt 

tl 

l{ 

u[ 

rfi 

rol 

t 

i 

l|1 

rtr 

llt 

TT 

lol 

f 

t 

T 

l0l 

lfi 

I 

fl 

rt 

lr 

IOt 

l9t 

tl 

4 

lsl 

lll 

lol 

l8{ 

tl 

ld 

tl 

r! 

tl 

T 

fl 

t! 

ttl. 

cFa! rt vl l0tt ill[ |lffirlf 

rdl 

Table 2. Test regultg: mean values of sach quantitatlve faclor for €sch impact conflguratlon (lor each quantltatlv€ fsctor, ths mean 

calculatsd on all te$ts ls equal to 100) 

tsrd 

Dt|tC 

tl 

t1 

l0i 

lr 

t{ 

rll 

Factor 

No. Labcl 

I 

r0{ 

l0l 

ut 

l|l 

ul 

ltl 

I 

I 

lfi 

lda 

lil 

tt 

|| 

lr t 

tli 

lll 

I 

r0r 

t 

lrl 

llI 

l0l 

fi 

ll 

3 

rll 

lq 

lri 

tf, 

t 

t 

il 

ll 

tl 

I 

lq 

tl 

t8l 

tlt 

rl| 

ru tt 

|r 

d 

ll? 

u0 

lot 

rol 

lll 

ll: 

fl 

l0i 

td 

rl( 

rll 

lll 

t0i 

t' 

11 

ttt 

T 

r0f 

tll 

T 

|{ 

tl 

tr 

tl 

ll 

tl 

I 

I 

It 

lr1 

ttl 

rT I?l 

ll 

frl 

IE 

tl 

rt 

I.t 

t 

tl! 

T 

5 

IG 

fl 

tt! 

?t 

|{ 

t tal 

It 

lfi 

ll 

$ 

ltl 

'l 

ttt 

EJ 

Ir I 

{li 

r|l 

ral 

I 

ril 

f, 

)t 

.'fl1 

-r|il 

-lll 

-r|i 

{ll 

-1rt/ 

Dllvtl ttltErclfl ll ?ll.x llAl.tltl ll aEcrl I Pttx. DltE ttl 

ul 

tl 

t6 

l0r 

tt 

rtr 

Irul 

tl 

H 

tl 

l0l 

9t 

Itt 

tEuH 

la 

?t 

tl 

ttl 

rq 

J 

lil 

ll: 

f 

I 

fi 

a1 

lral 

pflrc 

tl 

?l 

lfl 

ltt 

lri 

rai 

T 

r 

l0l 

l0t 

ft 

tta 

t[uH 

t1 

lfi 

t{ 

t0{ 

tl 

llrl 

t1 

tl 

tl 

le| 

fl 

tt: 

tlD GRt.i 

llJ l0l 

ll0 rq 

l0t l0: 

It tc 

u0 ll: 

tt .| 

Yharlt-rtZ 

Ilr Itl 

ltt ltt 

rgt tl 

lfi tl 

rfl l?3 

u 

-tDl 

rt1 

ttr 

rifi 

-tl 

a 

l0t 

?1 

l1 

l1 

tq 

rfi 

YI 

T 

I 

tl 

ll( 

t 

t rtt 

I 

t 

I 

* 

lfi 

td 

crtr r{D(L 

r{ 

ct 

il 

ll( 

lrl 

tr 

flrwrll 

Edl 

-ll 

.{A 

'Ut 

-|fl 

-s 

{ I 

.Il 

-trl 

al 

- ttl 

3rl 

-xl 

{ 

It 

lli 

-tl 

tttY Fr|| d.d Ltrata I I I 

fd3t condltloB 

N, A, Ref Test cffiflguFrtion 

30 A8 

caF-to-car wlth ofrrat 

30' esrcd buFi.r (rtsdrd 

brrricr) 

30 sH 3 30' uslGd bsri$ PIus rlde Ydl 

30 HS 4 

30' rr[IGd btf,Fl.r Yttlt 

stt-rllD ryrtf 

OFTS 5 0' utl.d buFirr rlth offrct 

oB 6 

fl.4. r -:l Vrhtcl. tyF. 

O' fiul.d bsrlrr (itsdrt{ 

b|frlcF) 

axes, give the analysis its full value. 

(a): Values resulting from film analysis (figure 2). 

(b); Definitions of mean decelerations until engine stoppage 

and until maximum vehicle crush (figure 3). 

tll 

fl 

11 

llt 

lli 

ril 

lc lr 

f 

tl 

fi 

rfl 

t tl 

f, 

tl 

?l 

lc 3 

tl 

|| 

tl 

lq 

I 

t: 

lq 

t 

lot 

lfl 

l{ a/ 

lal 

lll 

f, 

lr 

rt( 

rot 

rlj 

lll 

?t 

?l 

fi 

rl 

tl 

lci 

I 

tl 

'l 

t{ 

?1 

I 

rfi 

t! 

lll 

I 

lll 

lll t1 

riE 

l|| 

t 

tl 

lfl s 

I 

ll 

lr 

?1 

tl 

?l 

tlt 

ll,i 

l{i 

ltl 

t{l 

rtl 

ts 

lrl 

L 

T 

|r 

H 

lr 

ltl 

I 

I 

n 

tl 

J 

tl 

I 

la 

f 

t 

fl 

tl 

t! 

t 

I 

li 

ll 

lu I 

$ 

lol 

tq tl 

lol 

s 

l{ 

rI 

t?l 

l$ 

lri 

rll 

Iq 

t 

rrl 

Itl 

tu 

l|1 

Iti 

lll 

ril 

l7l 

It 

{l 

t{ 

tq 

tl 

3l 

ll 

7i 

tt 

{l 

l 

f 

lo{ 

l0{ 

lll 

ll 

tf 

ru 

r3 

lla t| 

x 

la 

tl 

tt 

t{ 

I 

tu 

lll 

lll 

t 

r} 

ul 

lrl 

ltt 

r0r 

llt 

{ 

T 

tot 

lll 

?l 

fl 

t 

a 

tl 

I 

l1 

?! 

rq 

l0r 

ltl 

t 

tI 

ui 

I 

ltr 

It 

f, 

lll 

tl 

a; 

?i 

Il 

ll 

tdr 

lI| 

11 

t1 

tli 

Itl 

tt 

(c); Passenger compartment integrity: maximum permanent 

intrusion of the windshield lower cross member and 

firewall (figure 4). 

Quantitative factorsi Main results 

Main axes and correlation circle 

A main component analysis is performed on all the data 

for the 39 tests. Each test is assigned an identical weight. 

Processing involves a transfer from a space of dimension 

n = l9 (number of quantitative factors) to a main plane l-2 

(annex I ) which provides an optimum representation of the 

data set in terms of inertia (figure 5). 

In the case under study, the l-2 plane found seems quite 

satisfactory and should be a good representation of real 

conditions, since these axes repre$ent 587o ofthe total inertia, 

whereas it is common, in data analysis, to represent only 

about 307o. Axis I by itself represents 43Vo of the inertia, 

and axis 2 represents | 5o/o. Tt,e other axes have an inertia of 

less than l07o (figure 6). 

253

Tut Frultr 

IlEL.r 

J DIIIC Drl%r ilutru EIC 

4 Y (or il) 

H6rrl lDact (or m hilil l|prct) 

6 thr ' rtErlns ulHl 

5 Dltl Y R [*IN th6ro 

6 DIU3 IR 3 || tldrir 

tlulM l.ft fff lorc6 

I DNF XulN ltlht fru fo*r 

9 9ET' lBNcnBGr ilulro HIC 

10 9fl t R rulu thoru 

DIF xut|u IGft fdr foE€ 

12 Pnr XulE rt8ht lffir foE6 

Vrhlclc 

13 t2 Tlil rt ehtch wlrlclc cruh l| rutrs (r) 

r6 

rB al 

r9 .2 

20 t 

ZL 

22 

d!d fitr 

'}I 

l?l 

I 

I 

I 

I 

drs. H*lffi fihlcle crush (a) 

t 

Lm8{tu.llnal ront of which 

rt riri t2 (r) 

Trffiwse w#nt of whiclc 

rt tts tl (i) 

thdtt vchlcli rotitton rt tlE t2 (r) 

pdH 

pdf 

.$P 

I li\ 

t 

I I 

I 

lbar \tahlcli dtrrlurtlffi utll 

m8tn rtoppqiE (b) 

lbs vahtcl. d.cilultlon fH cils{ir ttofprt 

utll u, shlcld cruh (rt ttr t2)(b) 

t s rrdrlclc drcditat{ott sttl 

trs € (b) 

lt*lrlt Dtrrfirfirt ttrtulm ol thr llndthlau 

Iffir cffir nrbar (c) 

llulE p.ffit lntruto$ of tI| 

frffiitl (c) 

I rrh 

r|| dtililir ruh{rr|rl0-ll 

\ l 

l l 

\l 

lr 

II 

I 

I 

I 

thJr 

llrrrltrJwt rtr{r|| 

(flh rnddrl 

Yirir 

Flgure 2. Vehlcle klnematlcs. Deflnltlon of maxlmum cruah, 

longitudinal and transv€rse movsments and rotation. 

254 

I 

rl 

I 

rl 

ItlntrUl 

rl=lild r2:,Wr2, r=y413gg 

dl dmrr-61 dmrr 

Figure 3. Vehicle klnematlcs: deflnltlon ol mean deceleratlon 

vsluee. 

Flgure 4. Integrity ol the vehlcle'$ passenger compartmGntr 

locatlon of permanent 

d€formation, 

The correlation circle of radius I makes possible the 

following: 

I 

I 

, 

, 

I 

'r rr*-----l I 

I 

I 

I 

I 

I 

I 

, 

I 

i-_*_____-i 

Dae. r drrr brdtallrl FE mil adl*Itlil * tb L'f r*r4.hLld 

ffi raLr 

ia.l+ r rds. bnaltrdli.l Ftlmnt a.f*Drbi C tlr llrnl 

(a) Viewing, in the l-2 plane, of factors poorly represented 

in that plane and therefore unusable on first 

analysis. Their image points are close to the centre of 

the circle. 

In the present case, 4 factors should be eliminated: 

driver and passenger femur forces. Their "bad" repre- 

sentation is probably due to excess scatter of femur 

forces and/or their small amplitude.

D0tr \ 

+ \ 

Ddt 

Ar'r 2 !l57rj 

Flgure 5, FeProsantatlon ol quantltatlve lactors and correlstlon clrcls In the mrln plane 1-2. 

Ft'r 

+ 

/ 

,/ 

| + + 

P|rrc 

Otx 1 

+-l 

/ stHt 

/ { 

/ , 

/ rl 

+ 

Anlt&37r1

l1 

s 

It 

lo 

It 

t0 

t 

0 

Flgure 6. Contrlbutlon of lhe main axes to totsl in€rtla. 

(b) Selection of the most significant factors, i'e., 

those closest to the circle of radius l, since they contribute 

extensively to the inertias of the main axes 

I and 2. Factor$ located within the circle of radius 0.7 

are not in the main plane and cannot be used on first 

analysis. 

(c) Revelation of (positively or negatively) correlated 

factor$ and independent factors' 

Positively conelated factors are grouped clo$e together: 

e.9., dmax, x, Y, t2. 

Some of these factors could be eliminated for subsequent 

analyses. 

Negatively correlated factors are diametrically opposite: 

e.g., a2, dmax. 

Independent factors are located on radii perpendicular to 

one another: thetz and dmax. 

Main correlations 

The main axis of the l-2 diagram, bearing 43Vo of the 

inertia, gives the main physical parameters of impacts performed 

at constant speed, namely, mean vehicle deceleration 

and a2 during the second phase of the impact, and 

negatively correlated factors such as crush and duration of 

impact. 

The secondary axis on the l-2 diagram, bearing l1Vo of 

the inenia, gives the rotation factor thetz, but especially, as 

shown in a subsequent study performed with additional 

256 

t0 ll tt tr ta tl tr lt tl It 

factors not included in the present analysis, the vehicles' 

weight. 

Insofar as concerns the vehicle occupants, very good 

correlation of driver head and thorax injury criteria (DHIC' 

DTH, DTH3) is observed with deceleration a2' The driver 

"is 

canied" by the axis of the impact. 

The passenger head and thorax criteria (PHIC, PTH)' 

although correlated, are Iocated on an axis at 45" relative to 

the axis of the impact. This results in sensitivity of these 

criteria not only to vehicle deceleration but also to parameters 

located on the secondary axis. One possible interpretation 

is the influence of vehicle rotation on passenger movements, 

allowing for the direction of Iateral restraint of the 

passenger shoulder-belt relative to the axis of the vehicle- 

In addition to these main results, the following observations 

can be made; 

r Mean deceleration a calculated one total crush is 

slightly offset relative to main axis I. This deviation 

is due to the position of al , still further offset, 

al being linked to the vehicle's $tructural 

$trength, but also, through its correlation with 

thetz, to the type of barrier causing rotation and 

involving a variable share of the vehicle's front 

structure, 

r Vehicle longitudinal movement x is naturally correlated 

to crush. but also to transverse move-

ment y. This latter relation, in accordance with the 

kinematics of impact in the case of 30" and 0" 

angled barriers, is more difficult to interpret for 

car-to-car impacts and impacts against offset 

half-barriers. 

r Passenger compartment integrity, pdw and pdf, is 

located on an axis at 45o relative to the crush. The 

intrusion/crash relation is natural. With respect to 

the intrusion component on axis No. 2, this is 

linked to the vehicle's architecture via the 

relation: 

Intrusion = Crush-Engine compartment clearance. 

To summarize, the correlation circle shows the following 

main directions: 

r Vehicle deceleration and deformation. driver 

lnJury cntena. 

Vehicle weight and rotation (figure 7). 

Figure 7. Directions ol the physical charactsrlstlcs of lmpact 

ln the maln plane 1-2. 

Qualitative factor; Type of impact 

conliguration 

The 6 impact configuration (qualitative factor) were superimposed 

on thequantitative factors in the main plane l-2. 

Representation of a qudlitative factor by its modalities 

(figure 8) 

Each modality (specific impact configuration) of a factor 

is characterized by its position relative to all the quantitative 

factors. By projecting it on the l-2 plane, its relation with 

the various quantitative factors located in that plane can be 

characterized. 

The diagram in Figure 9 is interpreted as follows: 

The Centre O corresponds to the mean of all tests for each 

quantitative and qualitative factor. 

A Point P which is the image of a quantitative factor 

(physical characteristics of the structure or test dummy) is 

located at a distance close to unity from the centre O, or, 

again, at a standard deviation of O in non-reduced form, the 

standard deviation being calculated on all values of the 

factor. 

Figure 8. $upcrposillon of qualitatiye factor$ on quantltatlye 

laclors In thd maln plane l-2. 

;ET =oar. 6F- 6' 

oFFS 

dP 

pdt 

FT-: *m ntn d fr.rdl hlffBln 

- 

h off$r trd.ilahntirt. 

ldl rffid{dtlhr{lilE$i 

h ll hil rdiirdlir, 

6t : rlniard drrblhfr aa tfufl 

htlrJa.r, 

. +*t* 

Flgure 9. Characterl?atlon of an lmpact conflguratlon accord- 

Ing to s qusntltatlve fsctor. 

The Characteristir:.r Point H of the mean of an impact 

configuration according to a quantitative factor (physical 

characteristics ofthe impact) can be obtained by projecting 

point M, which is the image of the mean of impact configuration 

results on the straight line OP. The relation OH/OP 

indicates, in non-reduced form, as a number of standard 

deviations relative to the mean of the factor for all 39 impacts, 

how the vehicle behaves depending on the factor in 

question. 

N.B.; This is strictly true only if point P is in the l-2 plane 

and therefore located on the correlation circle. 

Example: For the firewall intrusion factor and the OFFS 

impact configuration (offset half-barrier), the characteristics 

point is located at approximately 1.5 standard deviations 

from the general mean of all firewall intrusions, indicating 

the severity of this type of impact with respect to 

passenger compafi ment integrity. 

Comparison of impact conftguration 

The distribution of impact configurations in figure 8 calls 

for the following initial comments: 

r The impact against a 0' angled barrier appears 

greatly offset relative to the other impact configurations. 

It is characterized by high levels of deceleration 

in the initial and final phases of the 

251

impact, and by slight crushes and intrusions in the 

passenger companment. 

This configuration is located more than 5 standard deviations 

away from the general mean. As such, it is completely 

different from all the other configurations and is characterized 

by high driver and passenger injury criteria. 

r The images of the other impact configurations' 30 

AB, 30 SW 30 NS, OFFS and C/C, show similarity 

between the OFFS and C/C configurations 

on the one hand, and between 30 AB and 30 SW 

on the other hand. The 30 NS impact is located 

between these two pairs of impact configurations, 

but slightty offset along the axis of acceleration. 

. The configuration pair 30 AB and 30 SW is characterized 

in relation to the otherconfigurations by 

slight accelerations at the start of impact, strong 

rotation and relatively slight passenger compartment 

intrusions. The thorax criteria at 30 AB and 

30 SW are similar to those for the other configurations, 

except for OAB which is more severe' 

r The impact configuration pair OFFS and C/C is 

characterized in relation to the other configurations, 

except for OAB, by a higher mean acceleration 

at the start of impact, high crush levels and 

especially strong intrusions. The thorax criteria in 

OFFS and C/C are similar to the other configurations 

except for OAB, which is more severe. 

r Configuration 30 NS is distinguished from the 30 

AB, 30 SW OFFS and C/C configurations by 

slightly greater vehicle deceleration, especially at 

the end of impact, and by passenger compartment 

intrusions which are intermediate between 30 AB 

and 30 SW on the one hand, and between OFFS 

and C/C on the other hand' 

In this 3ONS configuration, the thorax criteria are similar 

to the other configurations except for OAB. 

Head/steering wheel impact 

Head/steering wheel impacts are a qualitative factor, as is 

the "impact configuration: factor. These two factors are 

dealt with in the same way in the l*2 plane. 

For the 39 tests, 32 head/steering wheel impacts are 

recorded. 

The "balance of head/steering wheel impact" image is 

located in the vicinity of the 30 AB impact (5 cases out of 7 

without head/steering wheel impact). The "head/steering 

wheel impact" image is offset towards the C/C configurations 

(8 head/steering wheel impacts for l0 impacts) and 

especially 30 SW 30 NS, OFFS and OAB (100d/o of 

head/steering wheel impacts). 

The location of head/steering wheel impacts in collisions 

"with 

head impact" differs depending on the type of impact: 

in a longitudinal vertical plane passing through the 

centre of the steering wheel in OAB impacts, or against the 

steering wheel spokes and rim in 30 SW 30 NS, OFFS and 

C/C impacts. 

258 

Frontal Impact in Real Road Conditions 

The six impact configurations of which the characteristics 

have been analyzed with respect to structure and injury 

criteria should be compared with real road conditions' The 

above data is based on the accidentological survey performed 

by the PSA-Renault Association (APR). 

Type of barrier (tahle 3). 

Accidentological data show that of the fatalities in car 

frontal impacts, 38Vo are due to impacts against the rigid 

fixed barriers, while 36Vo arc due to cat-to*car collisions. 

This breakdown is different for the population of severe 

injuries: 24o/o against fixed barriers and 597o car-to-car. 

Table 3. Frontsl impact: PsA-Fensult statistics concernlng 

pas$Gnger 

car occuPant8. 

f,itllcd 

gavirult lnjuEd 

F|rod 

obitacL 

3716 

24p 

Obrt aL 

Prrrirntar 

CE 

36,5 

69r5 

IrucL 

?5rg 

re,3 

Front structure deformation (Jigure I0). 

In severe frontal collisions, the APR accidentological 

survey shows that: 

ffs f!!1 

FT--] 

I r I l r l I 

r t 

t t t l 

--1-/+ 

22V. 

,*rtt**tt"; 

t | | t 

t t t t 

*; 

AV>10 krn/h 

*= 

t'"t"n--.,,, 

100 

100 

t 

l 

t 

l 

t 

t 

l 

l 

*; 

Flgure 10. Car clagsiflcatlon according to lront struclure de' 

foimatlon. PSA-Henault statistlcs concerning Passenger 

cars, 

r The 507o offset and 70Vo offset configurations 

represent 47Vo of accidented vehicles at velocity 

changes greater than 40 km/h; 

r There are more accidents involving at least 707o 

vehicle offset(28Vo) than accidents involving less 

IhaI 5lo/o offset (?ZVo): 

r Deformation of the vehicle front face is oblique, 

ranging between 15" and 40o in46Vo ofcases, and 

non-oblique (

Velocity change 

The breakdown ofsevere injuries and fatalities according 

to velocity change is as follows (APR survey): 

Delta-V 

50 km/h > 50Vo of severe injuries 

54 km/h > SOVI of severe injuries + fatalities 

65 km/h > 50Vo of fatalities 

Mean Deceleration 

44Vo of fatalities and 487o of severe injuries correspond to 

impacts at a mean deceleration ranging between 9 and l29: 

r Median of severe injuries: l0 g at a delta-V of 50 

km/h 

r Medianoffatalities: l3 gatadelta-Vof65kmih. 

Trajectory of car occupants 

The direction of movement of the occupant relative to the 

axis of the vehicle provides an indication concerning the 

direction of the forces exerted on the structure. 

In over 'lOVo 

of cases, the occupants trajectories are located 

at ll5' relative to the axis of the vehicle (table 4). 

Table 4. Dlrectlon ol vehlcle occupant tralectory rslatlve to vehlcle 

arls ln trontsl colllelons. 

Drlvar 

PeD8.r 

Illbd 

gtvrFlr inJuEd 

f,llhd 

Erv.Ely hJurrd 

-{E' te -16 

ll Eourr 

33t 

21r 

l4t 

20x 

H eadl steering wheel impacts 

-l!' tc +lt' 

l2 gouil 

7rr 

72x 

77x 

741 

+16' to +{6 

I Hour 

The frequency of head/steering wheel impacts in frontal 

collisions is an additional item of information helping to 

specify the direction of the occupant's trajectory" 

For a collision of violence ranging between 40 and 60 km/ 

h velocity change, the frequency of head/steering wheel 

impacts is 507o. 

Reference impact: discussion 

The selection ofa single reference configuration for frontal 

impact with dummies is an important factor with a view 

on the planned test standards. The barrier and the impact 

velocity must be defined. 

In the present study, the collision between a deformable 

moving barrier of given mass has intentionally not been 

taken into account. This configuration would result in 

excessive scatter of the speed variation between vehicles 

tested, depending on their weight. Collisions between heavy 

vehicles and a fixed barrier at high velocity change would 

not be represented in this type ofcollision, due to the excessively 

small velocity change in relation to a much lighter 

moving barrier. 

0r 

7l 

9t 

6r 

The C/C configuration will not be adopted, since it has 

the disadvantage of consuming two vehicles, which is economically 

prohibitive at the prototype te$t stage. 

The comparison of experimental results for the 30 AB, 30 

SW 30 NS, OFFS and OAB configurations with accidentological 

data shows that an impact against a 0'angled plane 

barrier (OAB) is not representative of real road conditions, 

for the following reasons (2): 

r Excessive mean vehicle deceleration at a given 

velocity change; 

r Unrepresentativeness of vehicle front structure 

deformations; 

r Low intrusion at a given velocity change; 

r Direction of forces on the vehicle oriented at 0o 

with a head/steering wheel impact frequency of 

1007o, which contradicts real road conditions 

where a frequency of 50Vo is observed. 

The other impact configurations (30 AB, 30 SW 30 NS 

and OFFS) are closer to real accidentological conditions. 

The experimental study shows that the 30 AB and 30 SW 

configurations are relatively similar to one another in terms 

of structure and dummy injury criteria. The 30 AB configuration 

seems better from the viewpoint of test implementation, 

the distance from the lateral barrier being hard to check 

very precisely in the 30 SW configuration. 

From a comparison of the 3 configurations 30 AB, 30 NS 

and OFFS with real accidentological conditions, it can be 

concluded that they have good representativeness in terms 

of front structural deformation, intrusion, and mean 

deceleration. 

Each of these configurations gives priority to one of the 

factors characterizing the impact. 

The OFFS impact configuration accentuates intrusion, 

front structural deformation being closer to a 507o offset 

than a 70Vo offset. Configurations 30 AB and 30 NS are 

closer to the structural deformations most commonly represented 

in real accidents. The mean direction of forces exerted 

on the structure in real accidents and resulting in a 50Vo 

frequency of head/steering wheel impacts is probably intermediate 

between the direction of forces exerted on vehicles 

in 30 AB and 30 NS impacts (head/steering wheel impacts: 

l\Vo and 1007o respectively). The deviations between the 

results achieved on thorax in the three configurations are 

within ll0Vo (3O NS) > OFFS > 30 AB). 

At the end of this study, it seems that the configuration 

most repre$entative of real road conditions is impact against 

a 30' angled barrier, the coefficient of friction at the interface 

between the barrier and the vehicle front face remaining 

to be specified. 

The speed of collision is an important item of information, 

defining the desired protection level. 

An impact speed of 56 km/h against a 30" angled barrier 

would cover over 507o of severe injuries and fatalities. 

259

The experimental study showed the difficulty of control- 

ling the frequency of head/steering wheel impacts. In the 

case of a head/steering wheel impact, only one point of the 

steering wheel is concerned, and this is inadequate to verify 

the steering wheel's qualities. This is why it seems advisable 

to supplement the impact with dummies by an additional 

test on the entire steering wheel, of the "component test" 

type, which remains to be defined. 

Reference impact and compatihility 

Numerous studies have emphasized the importance of 

inter-vehicle compatibility in frontal and side impacts. The 

notions of compatibility of weight, stiffness and architecture 

in frontal and side impacts have been illustrated (3), (4), 

(5). 

The choice of a single reference frontal impact, while it 

may represent a progress, could not solve the complex problems 

posed by stiffness and architecture compatibility, especially 

in side impacts. Further to the proposals made at the 

I lth ESV Conference (5), thorough studies should be continued 

along these lines. 

Conclusion 

Thirty-nine experimental frontal collisions were 

performed against various barriers to define a reference 

configuration established on the basis of a comparison with 

accidentological data. 

An analysis of the data by means of a specific software 

was performed for all the test data. 

This analysis showed a reference plane incorporating 

53Vo of inertias and allowing a study of the correlations 

between vehicle and dummy parameters' 

This study shows that one impact configuration, the 

orthogonal barrier, is very strongly offset relative to 

collisions against an angled barrier, offset half-barrier and 

car-to-car impacts, and also relative to real accidentological 

conditions. 

The collision against a 30" angled banier $eems most 

representative of real accidentological conditions, with the 

coefficient of friction between the banier and the vehicle 

front face remaining to be specified. 

A speed of 56 km/h could cover over SOVo of severe 

injuries and fatalities in frontal impacts. 

It seems advisable to supplement the global impact tests 

with dummies by an additional test on the entire steering 

wheel, of the "component test" type, which remains to be 

defined. 

The choice of a single reference configuration in frontal 

impact, while it represents a progressr could not by itself 

solve the complex problems posed by inter-vehicle 

architecture and stiffness compatibility, especially in side 

impacts. More detailed studies should be continued along 

these lines. 

260 

Main component analysis-all factors are quantitative 

and continuous: 

Xtj = t"tt result i for factor j 

All factors are centred and reduced: 

*ii=@ oxj 

Computation of main axis of inertia defining the l-2 

plane. The summation of the euclidean distances between 

all couples of tests is maximized: 

fector 

(xij-ri';1 " 

Analysis of the image of the factors in the l-2 plane: 

r Factors close to the correlation circle are in the l* 

2 plane 

r Factors inside the circle ofradius 0.7 are not in the 

plane and are not taken into account for the 

analysis 

r The correlation between two points is: 

Conelation (Xj,Xj') = Cos(oxj,oxj').llxj1l.I IXj'I I 

xj" 

jr19 

drr { j-r 

corr(IJTXJ')=1 

cott(XJ'NJ")=0 

Annu I

References 

(l) SPAD.N, 

"SystBme 

Portable pour I'Analyse des 

Donndes", L. Lebart, A. Morineau, T. Lambert-CISIA. 

(2) J. Provensal, "Five 

Years of New Car Assessment 

Program: Balance and Current Conclusions", lOth ESV 

Conference, Oxford, July 1985. 

(3) P. Ventre, "Homogeneous Safety Amid 

Experimental Investigation of Rear Seat Submarining 

Thomas F. Maclaughlin, Lisa K. Sullivan 

National Highway Traffic Safety Administration 

Christopher S. O'Connor, 

Transportation Research Center of Ohior 

Abstract 

An experimental investigation was conducted to 

determine the effects of certain seating and restraint 

parameter$ on the tendency for an adult rear $eat passenger 

to submarine 

(i.e., for the lap belt to ride over the pelvic iliac 

crest$ and penetrate the abdomen) in a 30 mph delta-v 

frontal collision. Fourparameters were investigated: 

type of 

restraint (lap belt only or three-point belt), seat cushion 

stiffness, seat cushion height, and lap belt angle (within the 

range from 20 to 75 degrees, as specified in FMVSS 210, 

"Seat Belt Assembly Anchorages"). The experiments were 

done on the HYGE sled, using a Hybrid III dummy with a 

"submarining pelvis" which contains three load cells 

mounted on each iliac crest to indicate lap belt location. The 

test matrix was a fractional factorial design, which enabled 

determination of the statistical significance of the 

parameters on submarining tendency and other occupant 

responses. Lap belt angle was found to be a highly 

significant parameter-the shallower the angle, the greater 

the submarining tendency. The tendency to submarine also 

appeared to be greater for three-point belted occupants than 

for lap-only belted occupants, although lap belt forces were 

much less for three-point belts. Results indicated that only 

one injury (AIS l) would have occurred out of six cases of 

submarining in three-point belts. 


United States FMVSS No. 210, "Seat Belt Assembly 

Anchorages", requires that the lap belt angle from the 

Seating Reference Point (SRP) to the anchorage fall within 

the range from 20o to 75o relative to the horizontal. 

Although front seat lap belt angles are typically close to 75o, 

rear seat installations in a number of cars, particularly small 

cars, have lap belt angles near 20". Accident data indicate 

Heterogeneous Car Population", 3rd ESV Conference, 

Washington, May 1972. 

(4) P. Ventre, "Proposal for Test Evaluation of 

Compatibility Between Very Different Passenger Cars", 

4th ESV Conference, Kyoto, March 1973. 

(5) G. Stcherbatcheff, "Compatibility 

in Side 

Collisions", I I th ESV Conference, Washington, May 1987. 

that, with the lap belt angle at or near 20o, there may be a 

possibility that rear seat occupants will slide under the lap 

belt (submarine), exposing themselve$ to abdominal 

injuries. Other parameters, such as seat cushion stiffness, 

the presence of a shoulder belt, etc., also may influence 

submarining tendency and other occupant re$ponses. 

We conducted a sled te$t program to determine (l) what 

parameters are significant in causing rear seat occupant 

submarining and (2) the effects of the more significant 

parameters on the occurrence of submarining and on other 

dummy responses. Prior to testing, we reviewed current 

literature to identify parameters which are believed to 

contribute to occupant submarining, and which are 

reproducible and repeatable 

in a controlled test situation. A 

listing (not necessarily all-inclusive) of parameters which 

appear to affect the tendency of the occupant to submarine is 

contained in table l. Based on this review, the following 

factors were chosen for investigation: (l) type of restraint, 

(2) seat cushion stiffness, (3) seat cushion height, and (4) lap 

belt angle. These four factors were incorporated into a 

generic HYCE sled buck, allowing forthe simulation of the 

rear occupant compartment of any vehicle. A series of 22 

HYCE sled tests was then performed. For more details than 

are contained in this paper, the reader should see the project 

final report (l).* 

Table 1. Psrametsr8 

lhat Itfect $ubmarlnlng 

A. Belt Parameters: 

l. Lap Only versus Lap/Shoulder Restraint 

2. Lap Belt Angle in Side View 

3. Lap Belt Angle in Top View 

4. Retractor Force 

5. Slack in the Lap Belt 

6. Slack in the Shoulder Harness 

7. Hysteresis in the Shoulder Hamess Webbing 

8. Location of the Buckle 

9. Latch Plate Design 

10. Shoulder Belt Anchor Location 

I l. Lap and Shoulder Belt Lengths 

rCunently with the Ford Motor Company *Numhcrs in parentheses designate refercnces trt end of papcr

B. Seat Parameters: 

L Angle of the Seat Cushion 

2. Stiffness of the Seat Cushion 

3. Friction Coefficient between Occupant and Cushion 

4. Angle of the Seat Back 

5. Seat Height 

C. Constraint Forces: 

I. Restraint on Knees by Front Seat Back 

2. Toe Board Restraint 

D. Irrput Forces and Accelerations: 

l. Time History of the Deceleration 

2. Deceleration Maenitude 

E. Occupant Parameters: 

1. Amount of Clothing on the Occupant 

2. Friction Coefficient between Occupant and Belt 

3. Initial Position of the Occupant 

4. Size of the Occupant 

5. "Relative Bigness" of the Occupant (Weight 

divided by Height cubed) 

6. Slope and Stiffness of the Abdomen 

7. Locations of the Centers of Mass of the Body 

8. Orientation of the Sartorius 

9. Joint Stiffnesses and Damping Coefficients 

10. Contraction of the Quadriceps Muscle 

Test matrix design 

The sled test matrix was designed to provide the most 

information possible about the $tatistical significance of the 

four factors in causing rear seat occupant submarining. The 

four factors and their levels were; 

Table 1. Parameler8 thst sffect $ubmarlnlng 

Factor l4wl 

tlrpr or restrrrn. 

}|:"llllrllt{.r, 

dedt cu'hlqn Etlffnes' 

ilfl I quntrftsd rn trs.ar 

ss,r hcishr ** I 

hfEh J 

S::l*tl:lf::"::. 

"serc Hclght D€cetuIn€tlonn 

lep b6lt af,ald 20' 

(off horlrof,tal) 34" 

:i: 

Three of the factors were tested at two levels; the fourfh 

factor, lap belt angle, had five levels. The matrix is shown in 

figure l, where conducted tests are designated by '*X's". 

One repeat test was conducted to provide a basis forestimating 

experimental error, which was desirable to statistically 

analyze results. 

This matrix is called a "half-factorial" or "half*replicate" 

because it calls for testing under only half the conditions 

present. The main features of factorial experimental 

designs are that (l) they allow the effects of independent 

variables on the dependent variable to be determined quan- 

262 

Flgure 1. Sled test mslrix, 

titatively (statistically) and (2) they allow determination of 

interaction effects among independent variables. (For example, 

knowing the "first order", or "two-way", interaction 

effect between lap belt angle and seat cushion stiffness 

provides an answer to the question: Is the effect oflap belt 

angle on submarining different for soft seat cushions than 

for stiff?). In half-factorial designs, such as ours, some of 

the main effects and two-way interactions are confounded 

with higher-order interactions. It is assumed, because it is 

usually true, that any effect seen is due to the lower-order 

interaction; i.e., that higher-order interactions are negligible 

compared with main and two-way interaction effects. 

Our primary interest was in the factor Belt Angle. Thus, 

we designed the sled test matrix to obtain statistical information 

about the four main effects and the two-way interactions 

involving the main effect Belt Angle (Belt Angle/Belt 

Tlpe, Belt Angle/Seat Height and Belt Angle/Cushion Stiffness). 

Information about all other interactions was not discernable 

due to the confounding present. 

For the statistical analyses (presented later in the paper), 

the null hypothesis was that the four main effects do not 

have an effect on the causation of occupant submarining or 

other responses of interest. The level of significance for 

rejecting that hypothesis was chosen tobe SVo (57o is generally 

used in analyses of this type). Thus, if the "p-value" 

from the analysis was 57o or less, the risk of rejecting the 

null hypothesis when it was in fact true is minimal and we 

are confident in saying that the factor in question does have 

an effect on submarining. In addition, we felt that levels of 

significance in the 57o to approximately lSEo range indicated 

"marginal" significance (i.e., we were unwilling to 

accept unconditionally the hypothesis that no effect existed, 

if the analysis indicated an 85Vo chance of an effect being 

present). 

HYGE sled buck 

We determined the values of several parameters in current 

vehicles Io enable fabrication of a "generic" HYGE 

sled test buck. Nine parameters were selected which described 

the geometry of the rear passenger compartment 

(see figure 2). On the basis of 1986 sales figures and vehicle 

availahility, we selected ten domestic and imported automobiles 

in different weight categories to determine average

L-AO 

L-3 

L-41 

L-48 

L-50 

L-5r 

H-31 

h 

d 

- - 

Flont Scrt BacL AnIl. 

Front of R.rr 3.rt t|ck to BeL of Front SorE Bmlr 

RGrtr s.at Brck AnEh 

E.rE occr+.nt Kilaa PlEt to Front S6at lrcl 

Frotrr Occr4tst H-Pt. to Lrr OGduSst H-Ft, 

f,e$ Oscr+mt H-Pt. to Hr.l pt. Dl|te. 

Rcar 0ccrry6t H-Ft. to H,6iI pt. Vtrtlc.I |tlrtea 

v.rtlcrl Dl|t.rr ftfi 16rr Occr.prntrr H-Pt. to fop of Ftmt 

3.rt tack 

tlrpth of Lrr 8.rt cMhton fril Front Edt. of cuhlon to 

occupantrr H-Pt. 

Flguro 2. Paaeenger reat compartm€nt g€ometry parameters. 

values for the nine parameters. 

The vehicles were as 

follows; 

Weight 

category 

I 

I 

2 

2 

2 

3 

4 

4 

5 

5 

vehicle 

1985 Ford Escort 

I984 Honda Accord 

1985 Pontiac Grand Am 

1985 Pontiac Sunbird 

1983 Toyota Camry 

1985 Chrysler New Yorker 

1985 Olds Ciera 

1986 Ford Taurus 

1983 Volvo GL 

1986 Buick Electra 

The average values for the 

below; 

Specification 

L40 

r_3 

L4l 

L48 

L-50 

L-51 

H-31 

h 

d 

1986 

sales po.sition: 

#6 domestic 

#3 imported 

#4 domestic 

#3 domestic 

#l import€d 

#9 domestic 

#l domestic 

#5 domestic 

#8 imponed 

#l I domestic 

nine parameters 

are listed 

average: 

25.5" 

26.76" 

25.5" 

2.23" 

3l.g l" 

35.94' 

t0.67" 

r7.97" 

15.69" 

The seat belt hardware consisted of the standard, automatic-locking 

type retractor generally used in the rear outboard 

seating positions. The original manufacturer's webbing 

was removed and replaced with a 5-bar webbing from 

one roll to minimize webbing differences. The same type of 

retractor was used for both the lap only and the three-point 

belts. The "D" ring chosen for the three-point belts was the 

locking type----over one half of the vehicles randomly surveyed 

had this type of "D" ring. Belt $ystems were replaced 

after each te$t to ensure integrity of each system. Load cells 

were used to record the inboard and outboard lap belt loads 

and the shoulder belt load, where applicable. 

The Escort front bucket seat was chosen to represent the 

average front $eat in overall height and seat back angle. 

Although not measured, its seat back stiffness appeared to 

be typical of most car seats; there was very little structure to 

resist rear occupant knee penetration. The Escort seat was 

replaced after each test to ensure an undamaged surface for 

the occupant to contact. The HYGE sled buck, in a pretest 

set-up, is shown in figure 3. 

Flgure 3. HYGE sled t€st sst-up. 

Seat cushion stiffness determination 

The stiffness values for the seat cushions were determined 

to identify the "softest" and "hardest" seats from 

the l0 vehicles selected above. Each rear seat cushion was 

tested in two places: (l) at the centerofthe occupant seated 

position and (2) on the forward edge of the cushion. Figure 4 

shows the forward position te$t set-up. The apparatus consisted 

of a hand-pumped hydraulic jack with a load cell and 

string potentiometer ailached. The load was applied at a rate 

of approximately l" per minute through a universal swivel 

joint to an 8" diameter plate. Force applied to the plate and 

Flgure 4. Cughlon force-deflectlon teet ssl-up (lorward 

poaltlon).

vertical displacement of the plate into the cushion were 

measured. 

Force versus deflection curyes for all l0 vehicles are 

shown in figures 5 and 6. The Toyota Camry cushion was 

one of the stiffest, and the Buick Electra cushion was one of 

the softest; therefore, these two were selected for use in the 

sled test matrix. 

j 

3 

o 

0 

o 

L 

FIgure 5. Sest stlffneea at center (in loadlng). 

0 

o 

L 

g,o 0,5 1.0 1.5 e.0 e. s 3.0 3. s 

0isplacemenL 

0isplacement 

tin) 

Flgure 6. Seat Btlffness at front (ln loadlng). 

Seat height determination 

q.0 

The vehicle rear seat height was measured as the vertical 

distance from the Seating Reference Point (SRP) to the 

occupant's resting heel position ("H*31" in figure 2). We 

obtained these values for a total of 3l vehicles (the l0 

vehicles selected above plus a random sample of 2l vehicles). 

They were non-uniformly distributed; only 6.574 of 

the sample were in the 5.5"-9.5" range and 93'5o/o were in 

the 9.5"-13.5". The median range values for these two 

groups were 7.5" and I1.5", respectively. The final height 

values chosen were 8'/ and 12". 

Crash pulse simulation 

The sled tests were conducted at a test velocity of 30 mph 

using a half-sine pulse. The pulse duration was approximately 

100 msec, resulting in peak sled accelerations of 

approximately 22 G's. This crash simulation pulse was 

based on similar pulses used for other occupant submarining 

studies performed by Adomeit (2, 3) and DeJeammes 

(4). 

264 

g,g 0. s t.s r,5 ?.s 

0isPlacement 

Vol vo rrt 

----+l- ci*. nt 

**El*trr nrt 

+cF! taat 

+G.H Ar mrt 

#Eccrt ceat 

+tr6ird mrt 

+tku YFkF *.t 

-----*-Fccrd |0r 

-- Tuua taai 

Hybrid III 

The anthropomorphic test device used for the sled series 

was the Hybrid III 50th percentile male dummy' A simple 

procedure was developed for seating the dummy such that 

its seated portture in the rear seat would appear more natural 

than was obtained by using the standard FMVSS 208 seating 

procedure, developed primarily for front seating. First, a 

male human subject (approximately 50th percentile) was 

instructed to sit in a "normal and comfortable" position in a 

typical automobile rear seat. The dummy was then placed 

next to the subject, and its posture adjusted to simulate that 

of the human. This was repeated with three other human 

subjects. The procedure which resulted was to seat the dummy 

initially according to FMVSS ?08, then move the pelvis 

forward enough to allow a space of about 2 inches between 

the rear seat back and the buttocks of the dummy. 

Instrumentation for the Hybrid III consisted of: 

r Triaxial accelerometer packages in the head, thorax 

and pelvis. 

r Femur load cells and knee shear transducers in the 

legs. 

r The "submarining" pelvis, consisting of three 

load bolts distributed along the anterior surfaces 

of the right and left ilium. 

. Denton six-axis upper neck load cell. 

r Abdominal air bladder insert. 

Filtering of the data consisted of SAE Class 1000 for the 

head accelerometer data, SAE Class 600 for neck forces and 

moment$, pelvic load bolts and femur loads and SAE Class 

180 for thorax and pelvic accelerometer data and knee 

shears. The abdominal bladder insert data and all of the seat 

belt loads were filtered using a SAE Class 60 filter- 

Prior to this project, an air pressure abdominal bladder 

insert for the Hybrid III dummy (5) was developed for 

measuring the depth and rate of steering wheel penetration 

into the abdominal region. We used the bladder insert during 

this project in an attempt to (l) determine if and when 

submarining occurred, and (2) mea$ure the amount of abdominal 

penetration caused by the belt when submarining 

occurs, hoping to determine the severity of the submarining' 

We found that, in its present form, the insert did not clearly 

indicate occurrence of submarining or penetration resulting 

from submarining, and would need more development work 

for this application. Therefore, it is not discussed further in 

this paper (see reference l). 

Sled test results 

A summary of the test conditions and whether or not 

submarining occurred are contained in table 2' Test #843 

wa$ to have been a three-point test. However, the retractor 

failed to lock during the test making the test 'tuspect for 

consideration in the matrix. Thus, the matrix contains 2l 

tests-2O of different conditions and I repeat test. 

Of these 2l tests, full submarining occurred in 9 tests, 

"one-sided" 

submarining in 2 tests and submarining did not 

occur in 10 tests. In general, we were able to detect very

Table 2. Sled tsst condlllona and results. 

I T6st I s.rr BGlr lEGrt t ltl S.rt I Bcrr l8ubr|rlnttgl f 16 of I 

tro. 

! ! 

Typ6 | Antlr lo$hrmlH.rthtl occut r is'rtmrrnrngj 

!--:::'!-:-------:-!-::'.....1-------l------l----...----t-----------i 

| 843 | * hp Onlyl 20 It t. I sdft I ttr | y.r I s7 u6c i 

r------t---------'-t---------t----...t..----t-----------i-----,-----i 

| 844 l11tr.6-polntl 20 D.r. I Sofr I Itr | Fr | 42 u6c i 

r------t---"'------t---------t---"---t------t----------.t,--,-------l 

| 845 lltr66-potntl 75 D.r, I Eofr I In I y.r I 12 ucc i 

t------t-----------t---------t-------t------t-----------t----,-,-,--l 

l#8461 lapontylTsltrt.l sofrlHtshl ffi | l| i 

l'-:-:-l---'-'-----l---------l-------l-----'l-----------t-----------l 

I 853 | Irp only | 20 Drg. | ltoft I HtEh I y6r i 58 -.. i 

! -':::-l:'------:-- 

! -::.-.---l ------- | ------ 1...-------- | ----------- i 

I E9l lThrao,potntl 20 Dit. I Hffrt I Hrth I yir j 43 .r.. i 

t------t-----------t---------t-------t------t--...---,--t-----------i 

| 89? llhr6a-polntl 75 D.E, I Hffd I HrCt I m i Xr i 

l--:::-!"-----:----!-::--'---l-------l------l------'----t-----------i 

lsgSllrponlylTsD.t, lHffdlInl m i nr i 

l'- :--l-----------1...------l-------l----.-l- i-----------i 

I f894 | lap only | 20 D6t. I thfd I Ifl | y$ | 4l u6c I 

I--:---l----------'l---------l---.---l'-----l-----------t- ---------l 

| 895 lThrrc-pottrrl47.5 ILg.l lterrt I Htth I t/2 yte i 75 raac i 

r------r-----------t---------t-------t,-----t---.-------t-----------l 

| 896 lThr66-potnrl47.5 Dr6.l Soft I In I yGr i 55 -.. i 

!--:::'!-"'---:-:--!::-:-.-'-l-------l------l-----'-----t-----------l 

| 897 | Irp only 147.5 Dct.l Hrrd I rtr | m I l{A i 

l--:::-l--------'.-l---------l-------l'-----l-----------l-----------l 

| 898 | tap odly 147.5 nrg.l soft I Hrth | rc i xr i 

l--::--l-----'-'---l---------l----'--l------l-----------l---,-------i 

| 899 lrtrer-poltrr|47.5 ltr8.l HrEd I Hrth | rc i Xe i 

l--:::-t-----------t---------t-------t,-,---t-----------t--- 

_____,_l 

| 904 | Irp only | 34 Deg. I soft I Hlgh I nd i rrr i 

t--:__-t-----------t---------t-------t*--,--t-----------t---_ _ __--l 

| 905 | hp only | 14 Drg. I Hrrd I rav I t., i tz rcac i 

l-':::-l-----------l---------l-------l----'-l-----------t--- -----,-l 

| 906 llhrae-potntl 34 D6g. I Soft I Ifl | yrr i aZ uac i 

! --:::- 

!:------':-- l':--'---- | -'-'--- l------l----:------ t ----------- 

porated three factors: ( I ) how high on the pelvis belt loading 

occurred, (2) whether submarining, when it did occur, was 

one-sided or two-sided and (3) the time of initial indication 

of submarining. 

The rating scheme consisted of six categories, which are 

listed below (although somewhat arbitrary, selection of the 

time intervals in categories 4-6 was based on the natural 

grouping that occurred in the experiments): 

L No submarining occurred; loads on pelvic lower 

load bolts only. 

2. No submarining occurred; loads on pelvic lower 

and middle load bolts ozly. 

3. One-sided submarining occurred. 

4. Full (two-sided) submarining occurred; initial 

submarining in 7(F-79 msec. 

5. Full (two-sided) submarining occuned; initial 

submarining in 50-59 msec. 

6. Full (two-sided) submarining occurred; initial 

submarining in 4G49 msec. 

In none of our tests did loading occur on the upper-most 

pelvic load cells without submarining occuning; nor did 

r 

| 907 lThrer-polnrl 34 Dag. I Herd I Htth I Lt/z ye, i SZ r*" submarining initiate in the time frame i 

of 6(F49 msec in any 

l-':---l-----'-----l---------l-.-----l------t----.------t-----------i 

| 9081 l,!ponlyl6lDrt,l Hrrdl rni of our tests. 

I lIA i 

!-':::'l-'----:'---l'::.-----l-------l------l-----------t-----------l Each of the 2l tests was rated according to the above 

| 9091 lrponlyl6lDrg.l BofrlHrEhl m i xe i 

l--::--l---'-'--'--l---------l-------l------t-'---------t---------- 

i scheme. The categorization results of the tendency for occu- 

| 910 lrtrcr-polntl 6I D.B, I H.rd I HrEh I no I NA i pant submarining are shown in the matrix format in figure 7. 

t------t-----------t----,----t-------t------t-----------i-------_-_i 

| 911 lrhrGo-potntl 61 D6t. I Soft I Itr | yrr j sn mec i 

t------t----'------t---------t-------t------t-----------t____----_-l 

* - fI *rs to hav6 b6rn r thrG6-polfiti ritrrctor frlldd to lock. 

# - lap B6lt brok. ht6 ln dvrnr (eftrr 80 ucc). 

clearly the occurrence of submarining from the films, the 

lap belt force-time response$ and the force-time responses 

of the load-measuring bolts on the pelvis (three located on 

each side). 

In addition to noting whether submarining occurred or 

not, the following five elements were analyzed: 

Tendency for rear seat occupant submarining. 

Occupant head forward (X) velocity. 

Occupant Head Injury Criterion (HIC). 

Occupant peak chest acceleration. 

Peak chest deflection for three-point belted 

occupants. 

Tendency 

for rear seat occupant suhmarining 

In addition to noting simply whether or not submarining 

occurred under certain conditions, we devised a rating 

method to indicate the "tendency" for submarining to occur. 

By "tendency", we mefln (l) if submarining did not 

occur, how close it came to occurring and (2) if it did occur, 

whether it was on both sides, and how early it was in the 

crash event. As a measure of the near-occurTence 

of submarining, 

we determined the location of the lap belt on the 

pelvic iliac crests, as indicated by the pelvic load cell bolts. 

From our literature review, time of occurrence. and whether 

one- or two-sided, were seen to be important indicators of 

submarining tendency. The rating method, therefore, incor- 

"4" 

't4 \ 

c.. 

\q,. I 

t. \\n 

| 

'**" 

'' -"., .1" l" 1.,, l"; l; 

I- l:l: l '.l ' I ' l ' l .. 

r \ | 'l* | I | | 

rlF rrs rrfi I 

I I *" l* 

i,i f""H,i,['r*.,"'* 

''. r . r. r6.rnrt nrFrd; t* F Flvi. 

I tu.i# rlhrl^rN d.v.r*. 

;l; 

;i il* 

1,,,,,,,,,1,;,i;,,,,1,,,,,,,,,),,,,,,,,,1,,,,,,,,,t,,,,,,,,,i,,,,,, 

l r * r t l 

|-" |;",. lo' ,1; .1; ;l; ;l; ; 

| | I : | r l r l r l l 

i ' ilii iil'ir$i iE llril f#Ei nririi i#iilifr ; il'fi it iii ii*, 

Flgure 7. Tandency 

lor aubmarlnlng. 

The categorization results were analyzed statistically to 

determine which of the four main factors and two-way interactions 

were significant. Figure 8 presents the four main 

effects of the test matrix. Each bar represents the average 

response for all tests conducted at the particular level (e.g., 

lap only) of the particular factor (e.g., Belt Type). Three of 

the factors have two levels each; one (Belt Angle) has five 

levels. If the difference between the average responses of 

different levels of a factor is "sufficiently large" compared 

with the response variances and the error estimate, then the 

statistical analysis will indicate that the main effect of that 

factor is "significant". Whether or not the main effect was 

265

found to be statistically significant is indicated beneath the 

effect, along with the p-value (as a percentage). Belt Angle, 

Belt Tlpe and Seat Height all had highly significant effects 

on the tendency to submarine. (Significance levels were less 

than l%). Cushion Stiffness had a marginal effect (l0.l1Vo 

significance level). No two-way interactions of main effects 

were significant (i.e., the effect a particular factor had on the 

submarining tendency was the same regardless of the level 

of other factors). 

o 

E r 

c 

E 

h 

t 8 

I 

c 

S e 

o 

F 

Figure 8. Tendency lor roar seat occuPant rubmarlnlng, 

The data indicating submarining tendency are presented 

in more detail in figures 9, 10, I I and 12. Each point in 

figures 9, l0 and I I represents an aYerage of two tests (or 

three, when the repeatability test result was included), chosen 

to illustrate particular main and interaction effects. 

In figure 9, lap-only and three-point belt results are plotted 

separately. It is clear that submarining tendency is differ- 

ent for the two restraint types (i.e., the Belt Tlpe main effect 

is significant). Also, lap belt angle strongly affects submarining 

tendency for both restraint types, indicating the absence 

of a Belt Angle/Belt Tlpe interaction effect. 

t 

llrr-PL lrlt 

+ 

s 

..s 

t. 

a 

Er 

$, 

E 

14 o||U lS 

"'{r.* 

td UF $dirdr 

lErnd. *ilrffi 

L+ Edt figh 

r o.tfr 

- ofif 

Orgts) 

Flgure L Tendency lor submarlnlng versu$ lap belt angle for 

belt type. 

In figure 10, results are shown separately forthe low and 

high seat heights. Submarining tendency differs for the two 

$eat heights, but is strongly influenced by lap belt angle for 

both heights (similar to Belt Type results in figure 9). 

Figure I I suggests similar trends for the factor Cushion 

Stiffness. However, separation of the two curves is less 

266 

Erlt Typr 

Angli 

Y$ - 0.721 Yrr - o.olli 

Hrlght Curhlon 

Vif - O.E9!i Msrllhil - 

I 0.77* 

f. 

!r 

g' 

I 

+t.E 

Lt tdt rlrb (D.F.) 

T*Hlt*ftliln- [|S 

l* r.d. filEH r o,flf 

Flgure 10. Tendency for submarlnlng veraua lap belt angle for 

$eat helght. 

distinct, indicating marginal difference between the two 

cushion stiffnesses in the submarining tendency versus lap 

belt angle relationship (as was indicated in the statistical 

analysis). 

g 

f,. 

i, 

i: 

I{ {7't tl 

|.oF ldt Atg|r (Drgu) 

qC$n trtrt lltliil.r r 

Htht ftr||cilFr 00ll 

toB Cflilc|r 

+ 

Hr{ Cfltit 

Flgure 11. Tendency for submarlnlng v€raus ldp bcll angle for 

cushlon Btlflnsss. 

Figure 12 contains the results for each individual test. An 

interesting observation, which puts some degree of uncertainty 

on trends discussed thus far, is that three of the curyes 

fall closely together, while the fourth is very different. One 

could conclude from this that the relationship between submarining 

tendency and lap belt angle is the same, regardless 

of type of restraint or seat cushion stiffness or seat height, 

except when the low seat is combined with the soft cushion. 

If this were true, it would indicate that the two-way interaction 

effect, Seat Height/Cushion Stiffness, is significant. 

This effect is confounded with the main effect of Belt Tlpe, 

and, as previously stated, when confounding occurs, we 

have assumed that the observed effect is due to the lower-order 

interaction (i.e., the higher-order interaction effect 

is negligible). It is possible that this assumption is incorrect; 

however, we feel that is highly unlikely, since both logic and 

past experience indicate that submarining is more likely to 

occur in a three-point belt than in a lap-only belt. The only 

way to know for sure, however, would be to conduct more 

tests in the matrix.

f: 

t 

R 

T, 

Lt Ertt AnCr (Dryr) 

flgg_re t!. Tendency for submarlning ysraus lap belt angle for 

bslt type/seat helghvcushlon rtlffness. 

There is another issue which raises some uncertainty; the 

effect of seat height on submarining tendency, as observed 

in the sled tests, may have questionable validity. The Hybrid 

III dummy has a fixed angle between the lower spine and the 

upper legs. Therefore, when fhe dummy is placed in the 

lowered seat, the pelvis rotates with the upper leg, exaggerating 

the pelvic angle and predisposing the dummy to 

submarining more than is likely for the human. (The reader 

should bear in mind that the feet are constrained against 

moving forward by the front seat). 

Occupant head forward (X) velocity 

The occupant's head forward velocity was derived by film 

analysis---digitizing the l" target at the head C.G. Iocation 

and obtaining both longitudinal and vertical displacements 

of the C.C. The longitudinal head trajectory was differentiated 

to obtain the corresponding velocity. 

Prior to testing, the occupant's head was located, on average, 

slightly over 20" behind the Escort front seat back. 

Therefore, 20" was considered the maximum distance the 

head could travel without impacting the front seat back. 

When the occupant was restrained by the three-poirrt belt, 

maximum head excursion was less than 20". When restrained 

by the lap belt only, his head travelled ar leasr 29". 

Thus, head impact was considered a possibility only for the 

lap-only belt system. 

!-#rr-t.st-dtl 

r..*hriB* 

I *E.l.li 

iu,FrHbldr 

Flgure 13. Head loward (X) veloclty (mph). 

In figure I 3, longitudinal head velocities are presenred in 

the matrix format. For the three-point belt, the velocities are 

maximum values, and range from l8 to 25 mph. For each lap 

belt only test, the longitudinal component of head velocity 

was determined at the point at which the head had travelled 

forward 20" lthe rearmost point at which head impact might 

occur). These are the values which appear in figure l3 for 

the lap belt only; they ranged from 26 ro over 33 mph, and, 

therefore, were considerably greater than maximum noncontact 

head velocities for the three-point belt restrained 

occupantrr, 

Flgure.l4- Fffect of lap brlt angle on head longltudlnal veloctty 

by restralnt. 

iliFifr'ffi* 

LS H firaL 

Y[ ' a,fll 

Flgure 15. Head Inlury crlt€rlon (HlC). 

ts lrli lflL 

h. at,ttt 

Since our main concem in analyzing head motion was 

to det€rmine potential contact velocities, we statistically 

analyzed head velocities separately for the lap-only and 

three-point belt restrained occupants. These results, shown 

in figure 14, indicate 

that Belt Angle has a small but significant 

affect on the head velocity of lap belt restrained occupants, 

at the point at which head impact may occur; but does 

not affect maximum head velocity of three-point belt 

restrained occupants. Also, this figure clearly shows that 

head velocities at 20" of excursion for the lap-only tests 

greatly exceed maximum head velocities for the three-point 

tests. 

267

Head injury criterion (HIC) 

The HIC was determined using the head C.C. resultant 

acceleration. Figure l5 contains the HIC values for the sled 

series. 

The statistical analysis indicated that none of the four 

main effects or interactions appear to have a strong effect on 

the occupant's HIC value, which is shown by the bar charts 

in figure 16. However, the difference between the average 

lap-only and the three-Point belt HIC's is substantial, and, at 

a 15.630/o level of significance, is considered marginally 

significant. HIC values, separated by Belt Tlpe and plotted 

against Belt Angle, are presented in figure I 7. These curves, 

and the individual HIC values shown in figure 15, show that 

HIC was far more erratic for occupants restrained by the lap 

belt only. Observations of the films indicated the reason; 

head contact occulTed in nine ofthe ten tests involving lap 

belt only, but did not occur in any of the eleven three-point 

belt tests. (Specifically, the head contacted either the front 

seat back or the dummy's knee in test 853, 893, 894, 897, 

898, 904, 905, 908 and 909. 

o 

; 

ta00 

| 600 

| 200 

e00 

Brlt TyFr Angh 

Mrrglnrl - l5.69tt il6 - 72.19% 

Hilght curhlon 

No - 35,51t N6. 65,43t 

Flgure 16. Head Infury crlt€lla for rear seal occuPants. 

s000 

20 

3n 

,.o rrn lt,t ,o.",".r 

Figure 17. Head injury criterlon (HlQ) lor rear seat occupsnts. 

Occupant peak chest acceleration 

The occupant'$ peak chest resultant accelerations are 

summarized in figure 18. 

The statistical results are shown in figure 19. Cushion 

was the only main effect which was shown to be significant 

in effecting the chest acceleration, with a p-value of 2-5To. 

268 

61 

The remaining main effects and interactions were nol 

sienificant. 

Brlt Typ. Anglr llrlght cu8hlon 

ilo-e0.13't ilo-34.70% Nd-60.74% Yrr-2.5o'{ 

Flgure 19. Peak chest accel€ration for rear Eest occuPants' 

Peak chest deflection 

for three-point belted 

occupants 

The peak chest deflections are summarized in figure 20 

for those tests where a three-point belt system was used. lt is 

interesting that the two tests where submarining did not 

occur resulted in the least amount of chest deflection. 

The statistical results are displayed in figure 2l ' Note that 

only I main effect is present in the statistical model-Belt 

Angf e. It is significant, with ap-value of l.llo/o. Seat Height 

.,F:r-1.0tt(Hrddl 

r.i*hhkd 

r . F l*rnr[ 

t2 - h.rl#l*hr[ 

Flgure 20. Peak chest deflection (lnches) lor three-point only,

and Cushion Stiffness are confounded, so the main effect 

due to either one cannot be discriminated without more 

testing. Interestingly, the confounded main effect is significant, 

with a p-value of l.3lVo, indicating that one of the 

factors, or a combination of the factors (low seat and soft 

cushion versus high seat and hard cushion) is significant. 

a 

E oE 

o 

o 

E 

,.illli',* 

Figure 21. Chsst d€flsctlon for lap/shoulder belted rear seat 

occupantg, 

Abdominal 

injury severity 

Leung, et al., (6) conducted l0 sled tests, in which human 

cadavers were restrained by three-point belts, to develop an 

abdominal injury criterion associated with submarining. 

Nominal sled velocity was 30 mph and peak decelerations 

ranged from 2 I to 30 g. Most of the cadavers were tested in a 

rear seat configuration. Upper and lower shoulder belt loads 

and inboard and outboard lap belt loads were recorded. 

Autopsies were performed. 

They observed a relationship between post-submarining 

peak force on the lap belt (average of both sides), normalized 

with respect to the mass of the part 572 dummy, and 

abdominal injury severity. This relationship, reproduced 

from Reference 6 and shown in figure 22, represents an 

approximate injury criterion which can be applied to our 

three-point belt test results. The criterion should not be 

applied to tests involving lap belt only, because the lap belt 

would be expected to penetrate a different region of the 

abdomen, causing injury to different organs, and, therefore, 

a different overall injury severity for a given force, than 

would occur in a three-point belt system. 

E 

T 

3 t 

i 

t 

I 

t 

kv.rnyerîffid5Fhr tJD 

tr .''sruH 

Flgure 22, Relatlonshlp between the standsrdlzed tenelon of 

the lap-beltr (after rubmarlnlng) and th€ abdomen AlS. 

Lap belt forces (average ofboth sides) from all ofour sled 

tests are shown in figure 23. For tests where submarining 

did not occur, values shown are maximum forces. For tests 

with submarining, values shown are the highest forces that 

were generated after submarining (full or half) occurred. No 

statistical analyses were performed on lap belt force because 

differences in injury consequences of high belt forces 

would be expected to greatly vary, depending on whether or 

not submarining occurs. As expected, maximum values 

(from tests with no submarining) were much higher for the 

lap-only belt (2ff)0 to 30fi) lbs or 8.9 km to 13.4 km) than 

for the three-point belt (approximately 14fi) lbs or 6.2 km). 

Post-submarining belt forces for the lap-only belts having 

the shallowest angles (20' and 34") were comparable in 

magnitude to peak forces that resulted in tests with no submarining. 

Maximum belt forces after full submarining in 

the three-point system ranged from 240 to 700 lbs ( l. I km ro 

3.1km). 

ilfrh;iH_"*' 

Flgure 23. Peak lap belt lorces (both sldes averaged) (tba). 

For the three-point belt tests, injury severities (AIS levels) 

were obtained from the post-submarining lap belt 

forces, by means of the relationship in figure 22. These are 

presented in figure 24,and indicate that, although full submarining 

occurred in six out of the eleven tests, only one 

injury would have occurred and it would have been of minor 

severity (AlS I ). (Although not specifically stated in reference 

6, it is assumed that the lap belt force/injury relationship 

of figure 22 applies only for full submarining). 

i,#;r'0 (|*rdEt 

I t*rnt{kd 

r , F blnt[ 

r/? - h.rrs nErnrfl 

Flgure E4. Inlury aeverlty (AlS) for three-polnt only.

Summary and Conclusions 

Based upon the results of the sled test series, the following 

summary and conclusions are made about the occurrence 

of submarining for rear seat occupants: 

l. The occurrence of submarining, in general, was 

very clearly detected by observing test films, forcetime 

responses of the load cells mounted in the submarining 

pelvis and force-time responses of lap and 

shoulder belt load cells. 

2. Of the four factors investigated in the test matrix, 

Lap Belt Angle has a statistically highly significant 

effect on the tendency for submarining to occur. Belt 

Tlpe (lap belt only versus three-point belt) and Seat 

Height appear also to have a significantz effect on, and 

Cushion Stiffness appears to marginally affect, submarining 

tendency. (While the effect of Lap Belt Angle is 

clear, a small degree of uncertainty exists with regard 

to the effects of the other factors, because of the confounding 

inherent in half-factorial designs). 

a. The shallower the angle (off horizontal), the 

greater the tendency for submarining to occur. A lap 

belt angle of approximately 45o appears to be a transition, 

below which the tendency for submarining to 

occur increases rapidly with decreasing angle. 

b. The tendency to submarine appears to be greater 

for three-point belted occupant$ than for lap belt 

only occupants. This appears, from the films, to be 

due to the shoulder belt ( I ) pulling up on the lap belt 

and (2) preventing forward jackknifing motion of 

the upper torso. 

c. It appears that submarining tends to occur more 

frequently when the height between the occupant's 

H-Pt. and heel resting position is low (8") than when 

that height is relatively high (l?"). The probable 

explanation for this is that with the lower heiSht, the 

occupant's pelvis is rotated rearward more than at 

the higher height. However, the effect of seat height 

on submarining tendency is questionable, due to 

lack of leg articulation in the Hybrid III dummy, and 

the resulting uncertainty regarding the dummy properly 

duplicating the human pelvis orientation for low 

seating heights. We feel the dummy's pelvis probably 

was rotated more in the lower seat than a human's 

would be, exaggerating the pelvic angle and 

predisposing the dummy to submarining more than 

is likely for the human. 

d. Although statistical significance is marginal, 

submarining tendency appears to be greater for the 

softer seat cushion. 

3. Forward head excursions and velocities were 

much greater for dummies in lap belts only than in 

three-point belts. Lap belt only restrained dummies 

had head velocities around 30 mph at a point represent- 

r Throughout rhe Summary and Conclusions, the tem "significant" or "signifi- 

caflce" is used to denote statistical signiflcance. 

270 

ing the rearmost location of a front seat back; they 

appeared to be slightly higher for steeper lap belt angles. 

Head excursionrt of three-point belt restrained 

dummies were less than that which would cause head 

contact; maximum head velocities were approximately 

2l-22mph. 

4. HIC value$ were marginally higher, and more 

variable due to some head/front seat back contacts and 

head/knee contacts, for dummies in lap-only belts than 

in three-point belts. HIC's averaged approximately 

1400 in the lap-only restraint, and 920 in three-point 

belts. 

5. Peak chest acceleration values were significantly 

affected only by seat cushion stiffness, averaging just 

under 60 g's for the hard cushion andjust under45 g's 

for the soft cushion. 

6. Peak chest deflections (three-point belt re$trained 

dummies only) were significantly affected by lap belt 

angle, ranging from about 2.1 inches for the $teepest 

angle to around 3 inches for the shallowest angles. The 

two tests where submarining did not occur resulted in 

the least amount of chest deflection. 

7. Maximum post-submarining lap belt forces of 

2839 and 2233 lbs occurred for the lap-only belts hav- 

ing the shallowest angles (20o and 34o, respectively). 

8. Maximum lap belt forces after full submarining in 

the three-point belt system ranged from 240 to 700 lbs. 

9. A useful, though approximate, injury criterion 

exists for evaluating severity of submarining for threepoint 

belt restrained occupants only. Injury severity 

can be inferred from peak lap belt forces that occur 

after full (two-sided) submarining (6). Although full 

submarining occurred in six out of eleven tests involving 

three-point belts, only one injury (AIS I ) was indi- 

cated. No criterion appears to exist for determining 

submarining severity when only a lap belt is worn; 

therefore, no conclusions can be made regarding sub- 

marining injury severity in lap-only belted occupants. 

Acknowledgments 

The views and findings in this paper are those of the 

authors and do not necessarily represent the policy of the 

NHTSA. 

The authors are grateful to Rodney Herriott, Doug Hayes, 

William Gwilliams, Claude Melton and Timothy Schock for 

their technical support in the fabrication ofthe test buck and 

processing of test data. 

A special thanks goes to Susan Weiser for the preparation 

of the manuscript.

References 

(l) Maclaughlin, T.F., Sullivan, L.K., and O'Connor, 

C.S., 

"Rear Seat Submarining Investigation," DOT HS 807 

347, National Technical Information Service, Springfield, 

VA 22161, May 1988. 

(2) Adomeit, D. and Heger, A., "Motion Sequence 

Criteria and Design Proposals for Restraint Devices in 

Order to Avoid Unfavorable Biomechanic Conditions and 

Submarining", SAE Paper 751 146. 

(3) Adomeit, D., "Seat Design-A Significant Factor for 

Safety Belt Effectiveness", SAE Paper 791004. 

(4) DeJeammes, M., et al, "Factors Influencing the 

Estimation of Submarining on the Dummy", SAE Paper 

8l1021. 

(5) Mooney, M.T. and Collins, J.A., "Abdominal 

Penetration Measurement Insert for the Hybrid III 

Dummy", SAE Paper No. 860653, SAE International 

Congress & Exposition, Detroit, Michigan, February 24- 

28. 1986. 

(6) Leung, Y.C., et al, "Submarining 

Injuries of Three- 

Point Belted Occupants in Frontal Collisions-Description, 

Mechanisms and Protection", SAE Paper 821158. 

An Experimental Study of Crashworthiness of Vehicle Front Structure 

Akihiro Miyajima, Noboru Takahashi, 

Gyoichi Hataya, 

Subaru Engineering Division, 

Fuji Heavy Industries Ltd. 

Abstract 

Frontal impact test data of vehicles which have various 

front structure configuration are analyzed, and relationship 

between crashworthiness and occupant injury level is 

discussed. 

Specific displacement, R (ratio of actual to free vehicle 

displacement) is introduced as an index for estimating the 

crashworthiness of front structure. Value of R at a certain 

time (here t = 40 msec) can represent the feature of 

crashworthiness in earlier period of impact and has close 

correlation with occupant injury level. To keep less R value, 

that is, to keep less vehicle displacement in earlier period of 

impact is effective to reduce injury severity. 

Presumed value of R is also obtained by computer 

simulation. [t could give an indication for predicting 

occupant protection performance with certain exactness in 

the first stage of vehicle body design. 


Vehicle crashworthiness is, needless to say, one of the 

most important factors which influence the occupant injury 

severity, as well as the performance of restraint systems. 

Many studies have been done theoretically and/or 

experimentally on this subject (Reference l, 2, 3, 4). 

Recently, large scale deformation analysis of vehicle 

structure by super-computer has come to be widely used to 

estimate crashworthiness, while occupant injury level is 

predicted by two- or three-dimensional victim behavior 

analysis (Reference 5, 6.). 

In order to provide appropriate crashworthiness at the 

early stage of vehicle development it is essential to e$tabli$h 

a simple and effective means to estimate the crashworthiness 

and to confirm its validity. From this point of 

view, as a practical approach, we analyzed impact test data 

of various vehicles, examined the relationship between 

occupant injury level and vehicle crashworthiness in earlier 

period of impact, and tried to introduce an index to be used 

to evaluate conveniently the crash performance of a vehicle 

front structure. We also compared this index with the 

calculated value by computer simulation. 

Vehicle Crashworthiness and Occupant 

Injury Level 

Vehicle deceleration curve 

Analysis of vehicle deceleration curve in a collision test 

is one of the mo$t frequently used means to eyaluate the 

crashworthiness. Vehicle deceleration is determined by 

impact speed, vehicle weight, stiffness of front structure 

members and configuration of rigid components, such as 

engine and drive train. 

Figure I shows certain vehicle deceleration curves in 30 

mph frontal impact tests on fixed barrier (measurement; on 

rear trunk floor). Specification of test vehicles is shown on 

Table l, where dimension a is the length of the front end 

(distance between the top of bumper and frontbulkhead), b 

is the distance from the top to powerplant, and c is the 

distance from the top to side frame (majorenergy-absorbing 

member). 

50 l0O raec 

rue 

Flgure 1. Vehlcle decelerallon ln 30 mph frontal lmpact test on 

flxed berrler. 

Basic diinensions and body structure of vehicle A and B 

are the same. but dimension b is not identical because of 

carrying different engine. Vehicle C has smaller body and 

powerplant compared with A and B. 

Difference of deceleration curve is derived from these 

differences of construction. Deceleration of vehicle A is 

27r

Table 1. Test vehlcle tpeclflcatlons 

bFct .Fd (hA) €.6 4E.r 49. r 

l.rrcrr v.rsht (E) r, a9' 1. Bl 

&lo4rtlo! ol !!rt htt 

hrlloltrlly o!tsrrt, 46t61 Horfuoltlllt oPo6.d. tutl 

r,3F r,2s 9r0 

JF 4m 

l@ !60 l?0 

115 725 415 

fr 1, 6l 

rd ril 6t 

t-Hs 

i{'!-{- 

H*-=+ V 

l i- ---1 

lower in -"r,,*, o*.,"0 ilt#;e that of vehicle c is 

higher. Vehicle B is the medium. Between these three cases, 

however, there are no big differences of the highest peak 

value of deceleration. 

Occupant injury level 

Occupant injury indices at the test of these three vehicles 

(mean value of several test cases) are shown in figure 2. We 

used passenger side data to eliminate the influence of secondary 

impact by steering post. From these data we can 

conclude that occupant injury level is higher when the vehicle 

has relatively lower deceleration in earlier period of 

impact (vehicle A), and on the other hand, we can expect 

lower injury level if it has higher deceleration in earlier 

period of impact (vehicle C). Vehicle B has the middle value 

of A and C. 

Flgure 2. Comparlaon of Inlury level of three lest vehlcles. 

Numerical expression of crashworthiness 

When discussing vehicle crash process, displacement is 

more convenient than deceleration to compare different 

vehicles. 

Figure 3 shows the vehicle displacement, S, to time, and 

Figure 4 shows its non-dimensional expression, R = S/Sc. 

Sc is the free displacement (viftual displacement) at 40 

msec, according to the impact speed. The time of 40 msec is 

adopted for convenience's sake. In the case of 30 mph to 35 

mph impact, value of free displacement, Sc = 0.53 m to 0.62 

272 

t 

m is close to the crash stroke, that is, effective energy 

absorbing deformation of compact/subcompact class vehicle. 

Accordingly, the vehicle displacement expressed as the 

ratio to this value is convenient to roughly estimate the 

crashworthiness. 

Flgure 3. Vehlcls displacement. 

vehtcle A. vo = \8.6 h/h 

lr0 

Celculatlon of n 

In this sense value of R may be called specific displacement, 

and less value of R to time means higher crushresistant 

characteristic in earlier period of impact. 

In the case of the three vehicles, as shown in figure 4, 

values of R are not so much diff'erent at 20 msec. At 30 msec, 

A and B has nearly the same, and C has less R value. At 40 

msec, where A(0.847) > B (0.795) > C (0.758), difference 

becomes distinct. 

& EEEC 

Flgure 4. Comparlaon of crashworthiness by non-dlm6nsional 

exPre99ron. 

Flgure 5. Inlluence ol drlve traln snd Impsct apeed 

(baaed on 

vehlcle A).

We also examined by test influence of different drive 

train configuration and different impact speed, in the case of 

the same body construction. The results show no significant 

difference ofR value in any case (figure 5). Consequently, 

with R value, we can quantitatively evaluate the basic characteristics 

of vehicle crashworthiness in relation to displacement 

and/or time. 

R value and occupant injury level 

In order to clarify the relationship between R value and 

occupant injury level, we examined test data of seventeen 

cases of 30 and 35 mph frontal crash of subcompact class 

passenger car, 

As shown in figure 6, a close correlation exists between 

the value ofR at 40 msec and occupant injury index, chest 3 

msec-G (correlation factor k = 0.9 l2). 

Figure 7 shows HIC versus R value at 40 msec for the 

same test data. Though it i$ not so much obvious as chest G, 

there is a tendency that HIC decreases as the value of R 

reduces (correlation factor k = 0.693). 

From the facts mentioned above, R value at 40 msec can 

be utilized as a practical index ofcrashworthiness for 30 to 

35 mph impact. Thus a bodystructure which has lower R 

value is preferable from the viewpoint of occupant 

protection. 

o 

d 

E 

E 

e 

E 

o 

Flgure S. Gorrelatlon botwo€n R at 40 mssc and chest 3 msec G. 

o k - 0.69, 

Flgure 7. Correlatlon between R at 40 maec and HlC. 

R value and ride-down effect 

Vehicfe l0nphl]5npt 

A o I 

B A A 

Ride-down effect generally increases with advance of the 

beginning of restraint of occupant with seat belt. We also 

examined the relationship between R value and ride-down 

effect. 

Figure 8 illustrates the vehicle deceleration and chest G 

of dummy versus vehicle displacement in frontal impact 

tests of the vehicle A, B, and C. The highest peak of chest G 

occurs immediately before the ultimate displacement. Rise 

up point of chest G (Sl), that is, beginning of occupant 

restraint is 300 mm for A. 230 mm for B and 195 mm for C 

re$pectively. Shaded area in figure I is expressed as 

E 

o 

r*t 

E -\ e ds . 

s1 

Vehiete A Drruy cheBt 

\ i 

: 

SI Se 500 m 

Vehlcle illsplBcment, g 

Flgure 8. Vehlcle and dummy chest deceleratlon vE. vshlcls 

dlaplacemant. 

k - -0,819 

rs2 '-\r. 

(-u. )' 

Flgure 9. Correletlon between R and rlde-down elfect. 

This value is proportional to the kinetic energy of dummy 

absorbed by deformation ofvehicle body in the period from 

the beginning of restraint (Sl) to the displacement at 40 

msec (S2), namely, the amount of ride-down effect. 

i 

273

Using the same test data shown in figures 6 and 7, relationship 

between E and R is obtained and $hown in figure 9, 

where E is corrected equivalent for 30 mph (48.3 km/h) 

impact. Correlation factor is k = -0.849, and less R value (t = 

40 msec) gives bigger ride-down effect. 

Computer Analysis of Crashworthiness 

As an index of vehicle crashworthiness, we calculated the 

value of R by mathematical simulation and compared with 

experimental data. 

The program employed forcomputation is DYCAST/GC 

Code (Reference 7, 8). 

Model 

The analyzed model is based on the vehicle B. For computation, 

engine hood, fenders and auxiliary Parts are neglected 

(figure l0). Components and members involved are 

substituted to finite elements as follows: 

r Sumpsl-non-linearspringelement 

. Side f14ns-nsn-linear spring element and beam 

. Crossmember-beam 

r Panel-shell 

r Powe{plant-rigidbody 

r Tire

Test data for the calculation model vehicle 

We also conducted impact test with the very vehicle used 

as the model for calculation. The results are shown in figure 

l2 and figure I 3 in comparison with calculation. Crash and 

displacement mode are fairly similar to that of calculation. 

R value at 40 msec obtained from the test data is 0.839, 

which has good coincidence with the calculated R value. 

From this experience, analysis with DYCAST simulation 

model seems to be useful for estimating vehicle 

crashworthiness. 

I 

Flgure 14. Vehlcle dl$placement In slmulatlon and cxperlm€nt. 

Conclusion 

(l) At a frontal impact test, relatively higher vehicle 

deceleration in earlier period of collision generally gives 

reduction of occupant injury level. 

(2) Specific displacement, the value of R = S/Sc (here at 

40 msec) can be used as an index for estimating vehicle 

crashworthiness. Less R value could provide lower 

occupant injury level. R value of a certain vehicle body is 

hardly affected by impact speed and drive train 

configuration. 

(3) Smaller R value generally means bigger ride-down 

effect. 

(4) R value could be calculated by computer simulation 

and utilized as an indication of vehicle structural design. 

We introduced an idea to estimate the vehicle 

crashworthiness using the value of R (specific 

displacement). It is, however, only one of the possible 

approach. The authors wish to conduct further experimental 

and analytical studies and to establish more effectual means 

to evaluate total crash performance of a vehicle. 

References 

Safer Steering Wheels to Reduce Face Bone Fractures 

Keith C. Clemo, 

Motor Industry Research Association, 

Nuneaton, Warwickshire, England 

Slade Penoyre, 

Transport and Road Research Laboratory, 

Crowthorne, Berkshire, England 

Martin J. White, 


Nuneaton, Warwickshire, England 

(l) G. Hataya, J. Takizawa, S. Kojima; Crash Analysis of 

Vehicle Structure Including Occupant. Bulletin of JSAE, 

No.8, 1975. 

(2) N. Aya, K. Takahashi, H. Nosho; A Method How to 

Estimate the Crashworthiness of Body Construction. 

Nissan Technical Review, No. I I, 1976. 

(3) M. Hashimoto, R. Nakahama; Influence of Crash 

Characteristics on Occupant Injuries. Journal of JSAE, No. 

4. Vol.4l. 1987. 

(4) M. Kondo, K. Takahashi; Relation between Vehicle 

Deceleration Curves and Occupant Responses in 

Collisions. Bulletin of JSAE. No. 35. 1987. 

(5) D.A. Vander Lugt, R.J. Chen, A.S. Deshpande; 

Passenger Car Frontal Barrier Simulation Using Nonlinear 

Finite Element Methods. SAE 871958. 

(6) J. Wismans, J.H.A. Hermans: MADYMO 3D 

Simulation of Hybrid-3 Dummy Sled Tests. SAE 880645. 

(7) R. Winter, J. Crouzet-Pascal, A.B. Pifko; Front Crash 

Analysis of Steel Frame Auto Using a Finite Element 

Computer Code. SAE 840728. 

(8) Grumman Corporation Research Center; DYCAST/ 

GC Nonlinear Structural Dynamic Finite Element 

Computer Code User's Manual, Ver. 1.4. 

Abstract 

The paper starts by reviewing accident data from UK and 

other countries which shows that restrained drivers frequently 

receive brain injuries and face bone fractures from 

striking their steering wheels in frontal crashes. These fracture$ 

are not usually fatal but are extremely unpleasant, 

requiring skilled surgery and long periods of immobilisation 

during recovery, and often resulting in permanent disfigurement. 

Possible counter-measures considered are seat 

275

elt pretensioners, air bags, and improved steering wheel 

designs, and the paper suggests that while pretensioners and 

air bags would be very effective their cost and complexity 

will prevent widespread use in high-volume cars for several 

years. In the meantime improved, less aggressive, steering 

wheel designs are seen as a good, cheap and quick way of 

reducing head and face injuries. A practicable safer wheel 

designed by TRRL and Sheller-Clifford was shown at the 

lOth ESV Conference at Oxford in 1985, together with a 

proposed impact test procedure using a flat disc of aluminum 

honeycomb. The honeycomb crush strength is chosen 

to be the same as the maximum pressure which the weakest 

important par"ts of the human face can withstand without 

bone fracture, so deformation of the honeycomb during 

impact indicates a wheel which is dangerously stiff' The 

present paper describes how this proposed test has since 

been developed, and gives test results for 15 designs of 

production wheels. It concludes that the test is easy to carry 

out and gives repeatable results, and suggests that use ofthis 

test in legislation would lead to steering wheel designs 

which would greatly reduce the risk of brain damage and 

face bone fracture. 


Seat belts have been shown to be very effective in 

reducing deaths and serious injuries to car occupants, with 

most studies sugge$ting that when worn they approximately 

halve the risk of death (1,2,3).* However, normal lap/ 

diagonal belts are not able to prevent a driver's face from 

hitting the steering wheel in a severe frontal crash, for 

reasons that are very obvious to anyone who tries a simple 

experiment-sit in the driver's seat with your belt fastened, 

jerk the belt to lock the inenia reel, then lean forward as far 

as the belt allows. Nearly all drivers find their faces or 

foreheads touch the wheel, even under the very low belt 

forces produced in the test. In a frontal crash, with the car 

decelerating at perhaps 30 g and a shoulder belt load of 

perhaps half a ton, the amounts of belt stretch and b-Jy 

movement are of course much greater, and the driver's head 

can meet the wheel at a high relative speed. Since most 

current production steering wheels are not designed with 

face impact or energy absorption in mind, it is hardly 

surprising that face injuries are both common and serious 

for restrained drivers. 

Accident Data on Face Injuries to 

Restrained Drivers 

The most comprehensive and recent source of accident 

data on restrained drivers appears to be the UK's continuing 

study organised by TRRL/DTp and being carried out by 

teams from the Institute for Consumer Ergonomics, 

Loughborough and the University of Birmingham, with 

financial suppoft from Rover, Jaguar, Ford, and Nissan (4). 

The most relevant data from this study on the driver face and 

head impact problem is contained in three recent 

publications by the TRRL and ICE teams (5,6,7). Taking 

+Numbers in parentheses designatc referencts at end of paper. 

276 

these in order of publication, the first (5) "examines the 

incidence, severity and causation of head injuries in car 

accidents in post seat belt legislation Britain". It uses data 

from 940 accidents and I 603 occupants, of whom 89 (6Vo) 

died, 401 (25o/o) werc seriously injured, 548 (347o) were 

slightly injured and 565 (357o) were uninjured. 936 (587a) 

were drivers, of whom 572 were restrained, 60 were 

unrestrained, and the restraint use of the remaining 304 was 

not known. 9 I of the drivers had head or face injuries from 

$triking their steering wheels, and nearly all of these were 

wearing seat belt$ (967o) and were in frontal impacts (95Vo). 

36 drivers received head (i.e. non-face) injuries from their 

wheels, 28 having AIS 2 brain damage (unconscious or 

amnesia) and ? having severe (AIS 3-6) brain damage. 84 

drivers had face injuries from steering wheel contact, 46 

being cut, bruises and abrasions (5 rated AIS 2),3l having 

single face bone fractures and 5 multiple fractures. The 

authors considered which parts of the steering wheels 

caused these injuries, and concluded that of I I AIS 3 or 

above injuries, 3 were caused by the hub and 8 by the rim or 

spokes. Taking all severities, the hub was repsonsible for 

34Vo and the rim or spokes for the rest. The author$ 

concluded that "the steering wheel has been shown to be a 

major cause of facial bone fractures and AIS 2 brain injury 

to drivers." 

The second relevant paper by the TRRl/Loughborough 

teams (6) was presented at I lth ESV Conference in 

Washington, May 1987. It starts by quoting Rutherford's 

findings (2) that compulsory wearing of seat belts was 

effective in reducing most injuries to drivers and front seat 

passengers, but that drivers' skull and face bone fractures 

increased by 8 and l07o respectively while the 

corresponding figures for front seat passengers fell by 7l 

and 46Vo. Since the main difference in a frontal crash 

situation between the driver and the front passenger is the 

presence of the steering wheel ahead of the driver, it seems 

reasonable to conclude that steering wheel impacts account 

for this large difference between the figures. The paper uses 

a more recent and Iarger data base than (5) with information 

from l6l8 vehicles and2720 occupants collected between 

January 1984 and June 1986. 213 of the vehicles were 

impacted frontally, and in these impacts with restrained 

front occupants it was found that the legs were the body 

region most often injured (63Vo of all injured occupants), 

followed by the head/face (52Vo). Excluding AIS I injuries 

leaves l08l of AIS 2 or more, of which the head/face 

accounts for 314 = 29Va and the legs for 230 = ZlVo. 

Considering the mechanism of injuries of AIS 2 or more to 

"vulnerable 

body regions", the steering wheel is the most 

important object struck, accounting for 34Vo of all injuries' 

while the next most frequent source is seat belt webbing at 

32Vo, and no other contact location exceeds 57o. When 

studying frontal collisions with passenger compartment 

intrusion, the steering wheel was found to be an even more 

important cause of injuries, accounting for 5l7o of all 

contacts while the next most common sources were "other 

vehicle" at lTVo and the A Pillar at 107o.

The third and most recent relevant paper published using 

this UK data base is (7). This pap€r says that 484 out of 2650 

re$trained drivers in frontal impacts sustained steering 

wheef head or face injuries (lBok),compared with2?l/2650 

= 87o sustaining torso injuries from their wheels. It was also 

found that the face injuries occurred at much lower crash 

speeds than the torso ones. The author of(7) attributes this 

difference to the fact that the $eat belt restrains the torso 

directly, so preventing torso/wheel contact except in very 

severe accidents, while the restraining load on the head can 

only be applied via the neck. The head is therefore much 

freer than the tor$o to move forward, and will receive less 

benefit from restraint use. This paper also points out a 

difficulty with using AIS to classify rhe seriousness of face 

injuries, because it does not take account ofthe potential for 

permanent disfigurement and consequent emotional and 

other problems. The author concludes that "the steering 

wheel is a frequent source of head and face injuries (to 

restrained drivers); the mo$t common non-minor head 

injury caused by steering wheel contacts is AIS 2 brain 

damage (unconsciousness); facial bone fractures are also 

common and more severe brain injuries do occur". 

While these recent papers provide strong evidence of the 

frequency and severity of head and face injuries from 

current steering wheel designs, the problem is by no means a 

new one. It was clearly identified by Gloyns et al in l98l 

(8), who concluded that both brain damage (concussion) 

and face bone fracture are very common in frontal impacts 

with restrained drivers. For example,2/3 of the drivers in 

the sample were injured only by making head or face 

contacts with their wheels, and of these 3l7a suffered a face 

bone fracture and 22Vo were concussed. The authors 

recommended improving steering wheel designs to increase 

their energy absorption, and suggested changes to 

legislation covering steering wheel testing. 

The problem of injuries produced by steering wheels is 

not of course unique to UK; researchers elsewhere have 

found very similar results when considering restrained 

drivers, and as belt wearing rates improve in other countries 

these injuries will become more important there too. For 

example (9) shows that in Finland in 1987 with front seat 

belt wearing rates of 82Vo in built-up areas and 92o/o on 

highways (in 1983/84), --steering wheel caused injuries 

have not shown a decreasing tendency during recent years 

", 

although injuries from other parts ofthe car had decreased. 

French workers had also identified the problem as long ago 

as l98l; (10) shows that 74 out of 405 belted drivers in 

frontal impacts had head to wheel impacts. To sum up, 

therefore, everyone who has studied this subject has found 

that head and face injuries from steering wheels are a 

serious problem for restrained drivers, and it is clear that a 

solution would be extremely worthwhile. 

Possible Solutions to Reduce Steering 

Wheel Injuries 

Since face bone fractures and brain injuries are being 

produced by head to wheel impacts, it is necessary either to 

prevent these contacts altogether, or to reduce their speed, 

or to design the wheel so it can absorb the head's kinetic 

energy without exceeding allowable human tolerance 

figures. Two ways of preventing head to wheel contacts, or 

ofreducing their speed, are to use seat belt pretensioners 

or 

steering wheel air bags. Either of these methods (or of 

course both together) should be very effective, panicularly 

if the pretensioner 

system not only tightens the belt but also 

moves the wheel forward as in the VW Audi Procon-ten 

design ( I I ). But pretensioners 

and air bags are complex and 

expensive, and it is not realistic to expect them to be used in 

popular cars soon, although they are gradually being 

introduced in quality cars and it is possible that when they 

prove to be effective there, public opinion will encourage 

wider use. However for high-volume cars the only 

financially viable approach to reducing drivers' head and 

face injuries seems to be the use of less aggressive steering 

wheels which are designed with energy absorption in mind. 

One such design was shown by TRRL and Sheller Clifford 

at the 1985 ESV Conference, figure I. This would be only 

very slightly more expensive than a standard hard wheel, 

and has been found to be highly acceptable to drivers in road 

trials. At least two car manufacturers (Volvo and GM) 

already make steering wheels using some of the same 

principles, and so the cost penalty is clearly not prohibitive. 

But other manufacturers have shown little enthusiasm for 

improving their wheels, and so it appears that it would be 

necessary to legislate for a test procedure to achieve wider 

use of safer designs. The aim of such a test procedure must 

be to limit the loads and decelerations of the driver's head 

from wheel impact to levels which will not cause injuries, 

and so a knowledge of human tolerance figures is an 

essential first step in designing the test procedure. 

,{4frrdr#v_ttrrrl 

2. fdr l|ltrl rpE*B. rdt rdl Fdrlid fid drhn d to buhlr *lfil 

3. I tilck mlt rim with itr -xlttrltr rfit|l rolntuEsttdlt pshlomd t lrr d 

r po&ibkeryfffi th.&lfittH. 

Flgure 1. TRFL/Sh€ll€r-Cllfford safety wheel

Human Tolerances For Head and Face 

Loadings 

Work on human tolerance to face loads goes back at least 

to the cadaver experiments of Rene Le Fort in Paris in 190 I , 

and his system ofclassifying the unpleasant fractures ofthe 

middle third of the face which occur when the tolerable 

loads are exceeded is still used today. Anyone who feels 

face bone fractures are not serious injuries (and who has a 

strong stomach) is advised to read some of the numerous 

medical papers on the problems of treating these fractures, 

eg. (12, 13). When trying to protect car occupants, two 

different kinds of possible injury must be considered; brain 

injury, caused by excessive deceleration, and face bone 

fractures, caused by excessive local loads. As with all 

questions of human tolerance, the available data is very 

limited for both these kinds of injury, but for brain damage 

the figure of 80 g for 3 ms has been accepted in legislation 

for many years (eg ECE Reg 21, FMVSS 201) and appears 

to be a reasonable criterion. For face bone fracture the 

problem is more difficult, as the skull is a very complex 

structure whose $trength varies over its different parts, and 

the risk offracture will depend on the applied force, the area 

over which this is applied, and the stiffness of the object 

which is loading the face. The best practical approach seems 

to be that taken by Nahum, (14, l5), who argued that 

pres$ure is the important criterion. His cadaver tests 

suggesred rhat a load of 200.-225 lb (0.9-l kN) when 

applied with an impactor of 1 sq. in. (645 sq mm) was 

unlikely to cause fracture of the zygomatic (cheek bone) or 

maxillary (top jaw) areas, Nahum also found that a much 

Iower load (50 lb, 0.22 kN) could break the nose, but the area 

of bone being loaded there is much less than I sq in; if this 

area is in fact about l" X l/4" then the allowable pressure is 

again 200 psi. 

Other workers, eg (16) have found similar levels of 

tolerable pressures for the zygomatic and maxillary regions, 

and although the subject does not appear to be well 

understood this pressure of approximately 200 psi seems a 

reasonable maximum to be allowed for steering wheel 

impacts. Cadaver testing cannot be carried out in UK, but 

we hope that countries where it can be done will in fact do it; 

such testing seems the only way of testing our hypothesis 

that an impacting device which crushes at 200 psi (1.4 

N/mmz) will not produce face bone fractures. 

Proposed Steering Wheel Test 

Procedure-1985 

At the 1985 ESV Conference in Oxford, TRRL presented 

a paper ( l7) proposing a steering wheel impact test based on 

existing "Interior Fittings" legislation, ECE Reg 2l or 

FMVSS 201. The test would use the same mass and speed 

( 15 lb, 6.8 kg and 15 mph, 24.1 kph) as those regulations but 

would replace their rigid metal spherical impactor with a 

disc of aluminium honeycomb whose crush strength 

corresponds to the human face tolerance, about 20O psi. Ttvo 

278 

pass/fail criteria were suggested, (l) that the impactor 

deceleration does not exceed 80 g for 3 ms (cumulative, so 

that a "ringing" accelerometer does not give a false pass), 

(2) that the permanent dent in the honeycomb is not deeper 

than 2 mm (measured only over the central 100 mm 

diameter area of the disc, so as to avoid edge effects-the 

disc's diameter is I 50 mm). In the I 985 work the aluminium 

honeycomb was used in its "as manufactured" condition, 

without pre-crushing, and a specification was chosen with 

the required initial crush strength in this condition. 

The 1985 paper Eave results for 7 standard production 

wheels and the TRRL/Sheller Clifford safety wheel. Only 

this safety wheel met the proposed deceleration and 

indentation depth criteria at all the points te$ted (hub, spoke/ 

rim junction, longest part of rim without a spoke, shortest 

part of rim without a spoke). The production wheels all 

passed on deceleration levels in the rim impacts but failed 

on indentation depth in those impacts, while in the hub 

impacts all failed on deceleration levels and only the Volvo 

passed on indentation depth. 

Development of Proposed Test Since 

1985 

Initial proposals for legislative test 

Following the publication of a proposed legislative test 

for steering wheels in the 1985 ESV Paper, a document was 

prepared for the European Community Group on Passive 

Safety (ERGA) for consideration for use in an EEC 

Directive to Member States. 

The document incorporated the test details proposed in 

the ESV Paper, with the addition of two tests to be 

performed on samples of the honeycomb material to verify 

its crush load performance. 

The document stipulated that each steering wheel should 

be impact tested at the four points chosen for the tests 

reported in the ESV paper, namely the hub, the join of the 

spoke and rim, the centre of the longest unsupported arc of 

rim, and also the shortest unsupported arc of rim. lt also 

allowed for a further impact point to be selected by the 

technical service responsible forthe tests where such apoint 

might be judged to offer a significantly greaterrisk of injury. 

Revised proposals incorporating pre-crushed 

honeycomb material 

In the discussion of this document by ERGA, it was 

suggested that there were two features of the honeycomb 

crush performance which made it unsuitable for the measurement 

of peak pressures. The first of these is the high 

initial peak which occurs at the beginning of crush on the 

honeycomb. A typical honeycomb crush load/deflection 

curve, as shown in figure 2, illustrates this. The initial peak 

is due to the buckling of the foil surface$ which make up the 

cells; once folding is established, the crush load falls and 

continues at an approximately constant level. It is clear that, 

if we choose our honeycomb material so that the height of

our initial peak matches our threshold pressure value, then 

continuing crush will occur at a significantly lower load, 

and therefore, penetration of the honeycomb represents a 

non-linear measure of severity. Also, and this was the second 

criticism of this material, the buckling behaviour of the 

foil, ie, its initial peak load, is significantly altered by imperfections 

in its shape, making a given sample very variable in 

this respect. The later working paper produced by TRRL to 

report initial trials of the honeycomb ( l8) confirms this. The 

paper reports initial peak values for a group of6 discs which 

vary between l. l8 N/mmzand 2. I I N/mmz, a spread of 56% 

of the mean. 

Wo(k doil to Inltirts 

hilur ol th6 horuybmb 

Flgure 2. Grush charactsrlstlcs of alumlnlum honeycomb 

As a result of these criticisms, a new approach to the 

preparation of the honeycomb material was adopted. Instead 

of using the honeycomb as it is normally supplied, it 

was decided to prepare each disc by pre-crushing it over the 

whole of its surface before test. The crush is continued up to 

a point where the crush load has passed the initial peak and 

fallen to the flat ponion of the curve beyond; this ensures 

that the buckling of the foil which causes rhe inirial peak is 

completed and the material is collapsing in a series of regular 

folds. Ifthe deflection is stopped at this point, the pattern 

of folding in the foil is preserved, and further crushing of the 

honeycomb will proceed as a continuation of the load/deflection 

curve already established. Thus, it is possible to 

predict the crushing load of each specimen by its performance 

in the pre-crush without the need for the tests mentioned 

in the previous section. Tests with pre-crushed honeycomb 

specimens proved that the behaviour of the 

honeycomb after interruption of the crush was exactly as 

predicted, the original curve being resumed as the crush is 

continued. The level ofthe flat part ofthe curve also proved 

to be much more repeatable than the initial peak for a given 

specimen. A sample of 6 pre-crushed discs gave crush loads 

varying between l.0l N/6rnz and 1.08 N/mm2, a spread of 

jVo compiled to 567o in initial peak loads of the previous 

group. 

In addition to the improved consistency of the precrushed 

specimen, the technique brings two practical advantages 

to benefit the test. The crushing gives the honeycomb 

disc a robust surface of folded foil which is much less 

susceptible to accidental damage than the exposed foil 

edges of the uncrushed material. The folded foil also lends 

itself to accurate measurement of the surface profile before 

and after impact, the tester being able to traverse the probe 

over the smooth surface, whereas measurements of the uncrushed 

foil edges required great care to avoid damage. 

One precaution which must be taken in pre-crushing is to 

lightly deform the upper surface foil edges before the crush 

load is applied. This ensures that the folding commences at 

the upper surface; without this, the honeycomb may fold at 

the lower surface or even at some point within the disc. The 

deformation may be achieved by a series of light hammer 

blows. 

Naturally, by pre-crushing the disc, the crushing load 

achieved in the impact test is considerably reduced. For the 

original material the mean steady crush load is 1.04 N/mmz 

which is considerably less than the previously quoted criteria 

of 1.4 N/mmz, representing the strength of the facial 

bone structure. A new material watr therefore selected for 

the tests, having the following specification: 

Manufacturer-{iba-Geigy (Aeroweb) 

Cell Size-{.25 in (6.35 mm) 

Foil Thickness--4.fi)Z in (0.051 mm) 

Material-Aluminium Alloy 5052 T (non 

perforated) 

Density-4.3 lb/frr (68.9 Kc/M3) 

Crush tests of 6 specimens of this material gave a mean 

crushing pressure of L7 I N/mmz with a spread from 1.70 

N/mmz to 1.72 N/mmz (l% spread). This is approximately 

20olo above the suggested human tolerance figure, but it was 

neces$ary to select a material which is a standard product to 

ensure availability, and it was felt that wheel manufacturers 

could reasonably object to a test material which is weaker 

than the tolerance figure. 

At the same time as changing the method of preparing the 

honeycomb discs, the opportunity was taken to give some 

attention to the way the profile of the disc surface and the 

subsequent penetration was measured. The TRRL working 

paper describes a technique for carrying out these mea$urements 

and a suitable apparatus. The honeycomb disc 

after impact is placed between two flat plates, the upper one 

having a circular hole 100 mm diameter, corresponding to 

the central zone over which measurements are made. A 3 kg 

weight is placed on the top plate, and 4 screws joining the 

plates tightened finger tight. Once clamped in this manner, 

the profile of the disc surface within the 100 mm diameter 

window is measured by taking readings from a dial test 

indicator mounted on a beam across the surface ofthe upper 

plate. The dial gauge has a 6 mm diameter probe with a flat 

end. To determine the penetration depth, five readings are 

taken of the highest point (undeformed surface) and five of 

the lowest point (bottom of dent). The difference between 

the mean values of these two quantities gives a measure of 

the penetration. 

A series of trials conducted by TRRL to investigate the 

repeatability of penetration depth measurements showed 

2t9

that estimates of depth by a single experimenter measuring a 

single disc (mean estimate of depth of crush 1.39 mm) gave 

a standard deviation of 0.016 mm, and that estimates by five 

different experimenters of the same disc gave a standard 

deviation of 0.034 mm. It seems clear that maximum penetration 

can be measured to sufficient accuracy to justify 

setting the permitted maximum at I mm, thus approaching 

the ideal criterion of zero penetration (since any permanent 

deformation of the disc shows that its yield strength was 

exceeded in the impact). 

U. K. test series to evaluate revised proposed 

legislative test 

In order to gain experience of the use of the revised 

proposed legislative test procedure under realistic conditions, 

TRRL awarded a contract to MIRA in March 1988 to 

carry out a series of tests on the steering wheels from eleven 

recent popular saloon cars, plus the TRRL/Sheller Clifford 

Safety Wheel. 

The tests were conducted to the procedure set out in the 

document ERGA S33 Rev I and incorporating all of the 

changes mentioned in the previous section. Two samples of 

each steering wheel were tested at each of the five impact 

positions, (that is, the four specified positions plus a further 

point chosen by MIRA to repre$ent the point judged to offer 

the worst hazard, not covered by the other four). 

The tests were conducted on the MIRA Head Impact 

Pendulum Rig, using a specially constructed pendulum having 

an effective mass of 6.8 kg and a centre of percussion 

coincident with the centre of the disc mounting face' This 

was achieved by designing the beam to extend below the 

centre of the impact face and above the pivot point by the 

correct amount. The pendulum incorporates two accelerometers, 

measuring the acceleration of the impact mass 

centre along a tangent to its arc of movement. The speed of 

the pendulum just before impact was measured optically. 

The impact velocity of the pendulum was determined by its 

position when released. 

The steering wheel was mounted on a rigid anvil for the 

test, using a steel stem to simulate the column inner shaft. 

The wheel's mounting on the anvil allowed a minimum of 

150 mm clear space behind the wheel. Care was taken to 

ensure that the profile ofthe end ofthe shaft and the securing 

nut matched that of the original components. 

The signals from the accelerometers and the photoelectric 

speed measuring device were amplified and passed 

to a digital data acquisition system, with a U.V. recorder 

providing a back up. The signals also passed through a low 

pass filter unit, having a re$ponse to SAE J 2l I Class 10fi)' 

The honeycomb material used for the discs was equivalent 

material to that specified in the TRRL Working Paper, 

but was obtained from a different manufacturer, namely 

Hexcel (UK) Ltd. The difference allowed a comparison of 

materials of nominally similar specification from different 

manufacturers. 

The honeycomb material was supplied in a continuous 

sheet 50 mm thick. Discs were cut from the sheet using a 

280 

circular cutter in the form of a cylinder of 150 mm inside 

diameter and 2 mm thickness, chamfered at one end to form 

a sharp edge at the inside. After cutting, the discs were 

inspected for irregularities in their cell structure and surface 

blemishes. 

The discs were then prepared for pre-crushing by lightly 

deforming the foil edges all over one surface using light 

hammer blows. and were then crushed to a thickness of 40 

mm, 

The crushing was conducted in a compression test machine, 

set to a crush speed of l2 mm/min. A loadcell was 

fitted between the fixture and the test machine crosshead to 

record the crush load and a potentiometer to record the 

deflection. The signals from these were amplified, passed 

through a low pass filter set to SAE J2l I Channel Class 60, 

and recorded on a digital recording apparatus. After crushing, 

the disc was again inspected and was then identified 

with a code number in marker ink on its crushed surface. 

The crush load and deflection of the specimen were plotted 

on a graph which was kept with the specimen for further 

examination. 

Results of U. K. test series-honeycomb 

performance 

A total of l4l discs were cut from the Hexcel mflterial. 

Three of these were rejected due to irregularities in theircell 

structure and the remainder were subjected to pre-crushing. 

The mean crush load of each of these in the last 2 mm of 

crush (8 to l0 mm deflection range) was noted. The mean of 

this value for all of the specimen$ was 30. 13 kN with a 

standard deviation of I.59 kN; this conesponds to pressure 

values of I.705 N/mm2 and 0.09 N/mm2 respectively. lt can 

be seen that this mean value is very close to the mean of 1.7 I 

N/mmz measured by TRRL in the Aeroweb sample, although 

it is again rather higher than the pre$$ure required to 

match the facial bone performance. This difference strongly 

favours the manufacturer, since a wheel which is upto24Eo 

higher in crush pressure than the original quoted threshold 

will still pass the test when using this material. 

In order to improve repeatability it was suggested by 

TRRL that all discs having a mean crush load outside the 

range 28 kN to 32 kN should be rejected. This led to I 2 discs 

being rejected which represents 8.57a ofour original stock. 

Results of U. K. test series*steering wheel 

performance 

The results obtained from the series oftests are given in 

table I . The pass/fail criteria proposed for the legislative test 

requires a penetration of the honeycomb not exceeding I 

mm and an impactor acceleration not exceeding 80 g for 

more than a 3 ms cumulative period. 

Reference to the table shows that, of the twelve types of 

wheel tested only one, the TRRl/Sheller-Clifford wheel, 

satisfied both requirements at all five positions. Of the other 

wheels, five failed at two positions, two at three positions, 

two at four positions and two at all five positions. It should 

be pointed out that, in several instances where failure oc-

Table 1. Results of UK Teat Serlea (11 PopulEr European Saloon Cars) 

Ychlcls Ht$ Spokr 

/Rrl 

Lrnctlon 

A 

B 

c 

D 

E 

F 

G 

H 

I 

J 

K 

ftIEEL T 2 

XIIEEL I 

z 

TTIEEL I 

2 

XI{EEL I 

2 

XTIEEL I 

2 

THEEL I 

2 

XI{EEL I 

z 

ITIEEL I 2 

T}IEEL I 2 

TI{EEL I 2 

XTIEEL I 

2 

TRRL SAFETY XTIEEL I 

rRRL SAFETY iHEEL Z 

L'4.J 

148.9 

lzt.5 

tz7.j 

Itt.8 

ut.r 

lzu.0 

125.8 

82.8 

84.0 

78. r 

78,7 

100.5 

l0?. r 

1t4.5 

lto. ' 

I20.9 

120.6 

126.6 

115.6 

t4t. t 

l?8.0 

7r.0 

72.9 

I rr CuilIstlvc Exccrdcncr (g) Hrxlrrr Honoycol$ Ptnrtrrtlon (m) 

4r,1 

tl.8 

40.5 

TB.' 

]6.8 

t5.0 

4{t,0 

19.0 

rt.8 

4t.8 

t1.5 

44,5 

48.5 

48.0 

4t. t 

49.I 

47.8 

47,2 

49.7 

49.5 

49.' 

tI.8 

28.0 

27.r 

Short 

Rfn 

tr.5 

4r.6 

t0.0 

t0. t 

22.O 

22.7 

?8.8 

29.8 

41.7 

t7.t 

40.t 

,7.7 

t4.5 

16.t 

2r., 

?4.0 

,7.6 

t9.r 

21.o 

28.0 

29.0 

t].0 

tt?.2 

hI.J 

Long 

Rtr 

22.' 

2r.6 

24.0 

zJ.5 

?4.6 

24.1 

10. t 

18.9 

2J,7 

2t.8 

2I.6 

t0.5 

50.0 

67.2 

69.8 

20.7 

20.8 

22.j 

?J.O 

2r.0 

2r.8 

2,.6 

ltt.2 

curred at more than one point, one of these points was a 

"worst 

case" position which may have been closely related 

to one of the other positions and not a distinct part of the 

wheel in its own right. Of the 34 instances of failure 

amongst the whole group, 8 were due to acceleration alone, 

l5 due to penetration alone and I I due to both acceleration 

and penetration. 

It can be seen that, although none of the commercially 

available wheels selected fully met the requirements, a large 

group were only marginally outside the test requirements 

and possibly require only minorchanges to be able to satisfy 

the propo$ed legislation. 

Considering the repeatability of the results, the mean 

difference between the cumulative 3 ms exceedence (acceleration 

level exceeded for a total of 3 ms) values for equivalent 

positions on two similar wheels is 3.1 g. The mean 

difference between maximum penetration values between 

equivalent positions on similar wheels is 0.32 mm. 

IorBt 

Ciaa 

t6.9 

It,9 

1r2.0 

rll.0 

rto.6 

r08.0 

Itr.8 

1t7.0 

85.1 

86. r 

78.0 

76.9 

94.8 

8r. t 

152.9 

Itl.5 

84.2 

81. r 

rlt.6 

It?.? 

88.5 

90,0 

25.2 

29.5 

HLb Spokc 

/Rlr 

Junctlor 

7.2 

7.0 

2.2 

1.05 

0 

0.1 

r,6 

2.t 

0 

0.1 

0 

0 

0.0t 

0.1 

2.5 

2.2 

0.t 

0.1 

2.0 

1.9 

2.6 

2.9 

0 

0 

0,6 

0.r 

1.8 

t.4 

r.tt 

2.7 

0.6 

0.7 

0 

0.4 

2.7 

2.' 

z,e 

2,5J 

2.6 

2.6 

0,65 

0.25 

0.8 

0.65 

2.77 

t.r5 

0.1 

0 

Short 

Rin 

0 

0,t 

0.15 

0.2 

0,2 

0.5t 

t,0 

0.6 

0.45 

0,9 

0.7t 

0.I5 

2.7, 

2.9' 

l.t 

2.0 

0.75 

1.00 

z.r, 

2.60 

2,r5 

r.8 

0 

0.2 

Long 

Rir 

0 

0.t 

2.7 

1.75 

0.t 

0.2 

o.7, 

0.4 

0.6 

0.4 

1.7 

0.8 

l.d 

l. tt 

2., 

l.t 

0.4 

0.4 

0.85 

0,80 

l.l5 

0.85 

o.z, 

0 

Further developments of legislative test 

llorrt 

Cuc 

0.4 

I.7 

2.6 

2.0 

0.4 

0 

2.6 

,.2 

0,2 

0,15 

0 

0 

0 

0 

l.r 

l.6t 

0.7t 

0.9t 

l.tt 

r.50 

,,2 

5.t5 

After the previously described series of tests had been 

conducted, increasing interest in the possibility ofnew legislation 

led to agreement within the ERGA Group that a 

programme of comparative trials should be carried out by 

several test houses throughout Europe. The participating 

bodies are: Netherlands (TNO), West Germany (BASI), 

United Kingdom (MIRA), France (UTAC) and Italy (FIAT). 

The proposed series of tests involves impacts using a 

modified form of the legislative test on three production 

steering wheels. These were selected for their general avail- 

ability, and were from the 1980 Ford Cortina Mk 4, 1984 

Ford Escort, and the 1987 GM/Vauxhall Cavalier. While all 

ofthese designs have since been superseded, they are typical 

examples of high-volume steering wheels of their re- 

spective dates. 

Each wheel is required to be tested at the stipulated im- 

0 

0

pact positions on the centre of the hub and the spoke/rim 

junction. Each facility will test each of the wheels 5 times at 

each of the positions. The test results are then to be compared 

to establish the repeatability ofresults between different 

test rigs at different facilities. In addition, the results are 

to be compared with results of tests involving the use of a 

rigid head form similar to that used in existing Interior 

Fittings legislation. The revised technique adopted for these 

tests incorporates four main changes designed to improve 

the repeatability and resolution of the results obtained. 

The first change is in the pre-crushing procedure for the 

honeycomb. One feature of the crush characteristics of the 

honeycomb material is the fluctuation in load which occurs 

over the nominally flat portion of the load deflection curve. 

This occurs as a consequence of the folding which takes 

place in the foil. It was felt that the variation in crush load 

between different discs could be considerably reduced ifthe 

crushing procedure were changed to allow the crush to be 

terminated at a given level of load in its fluctuation' rather 

than a fixed deflection as before. 

To achieve this, it wa$ proposed that the disc should be 

crushed to an appropriate thickness and then, without stopping 

the crush, the operator should wait until the risirrg load 

curve achieved the chosen threshold level. The crush would 

then be stopped immediately. Subsequent crushing of the 

disc during the wheel impact test would then continue at the 

threshold value. Trials of the new method proved that the 

technique was feasible, provided the crushing deformation 

rate could be reduced to approximately 4 mm/min. 

Tests conducted at the same time using round-edged 

probes penetrating limited areas of the disc surface led to the 

discovery that the "skin" formed on the surface of the disc 

by the folded Iayers of foil had a significant effect in the way 

the disc responded to limited areas of high pressure. In other 

words, the reinforced surface layer of deeply crushed discs 

tended to spread the effect ofa local high pressure area over 

a greater number of cells, reducing the disc's ability to 

detect such features. For this reason, it was decided that the 

honeycomb crush deflection should be reduced to a minimum. 

and therefore crush deflection limits of 2.5 mm to 3'5 

mm were adopted. This appears to be the minimum practical 

level of crush, considering the thickness lost in folding over 

the foil edges and the need to achieve a full peak to ensure 

crushing is occurring all over the top surface. 

The second change put forward was in the direction of the 

impact device in striking the wheel. It had long been recogni$ed 

that the previous stipulation (that the impact is directed 

perpendicular to the plane of the wheel) could lead to 

discs being deformed to a wedge shape when striking a hub 

surface which slopes in relation to the plane of the rim- The 

result of such a test would be a deformation occurring mainly 

outside the central 100 mm diameter zone (and also very 

difficult to measure). The procedure was therefore revised 

to allow for the impact to be re-aligned at up to 30 degrees 

from the original direction so that the face of the disc was 

parallel to the wheel sutface, or at a tangent to some "high 

spots" on the wheel. 

282 

The third change was in the technique of measuring the 

penetration. A new apparatus was designed by BASI in 

Germany similar to the measurement clamp described 

above but with a mask recessed into the top surface. This 

mask incorporates an array of holes which covets the central 

100 mm diameter measurement area. The mask may be 

aligned with the honeycomb to place the holes in line with 

the flat sides of the honeycomb cells. A probe is then lowered 

into the holes to measure the depth below the datum 

surface of the plate. The apparatus incorporates a base plate, 

onto which the disc is fastened by three serrated pins engaging 

in the cells of the honeycomb. The undeformed disc's 

surface is first measured, then the base plate and disc are 

transferred to the impact rig as a unit, then they are retumed 

to the measurement apparatu$ in the same alignment to be 

re-measured. 

International test series{hanges to impact 

rig requirements 

As well as the changes to the honeycomb preparation 

procedures described above, it was agreed to standardise the 

type of impact rig used for those tests, to remove one source 

ofvariability between countries. The test chosen is a vertical 

drop rig with the 6.8 kg impactor being guided in a 

straight line during the 24.1 km/h impact. This was preferred 

to a pendulum rig because pendulum design is complicated 

by the need to achieve the correct effective mass of 6.8 

kg while making the centre of percussion coincide with the 

impactor. Also the required mass distribution of the pendulum 

and its stiffness both in torsion and in bending should 

be specified to ensure that the honeycomb faceform cannot 

rotate significantly during impact and that nearly all the 

pendulum's kinetic energy is delivered rapidly to the wheel' 

Otherwise, as an extreme example, a pendulum consisting 

of a heavy chain carrying a light plate behind the honeycomb 

could be used to meet the energy/speedlmass requirements 

while allowing any wheel to pass! It would clearly 

have taken too long for all the test houses involved in this 

programme to agree on a pendulum design and to build new 

rigs, so a vertical drop rig was chosen. 

A free drop (unguided during impact) rig was considered 

but was rejected because slight errors in impact point on the 

wheel rim produce large differences in the free headform's 

motion as it rotates round the rim and spins off; if this 

happens the required energy of 152 J (from 6'8 kg moving 

atZ4.l fm/h) is not of course being absorbed by the wheel' 

A fully guided vertical drop rig therefore appears to be the 

best choice for this wheel impact test. 

Results of international test series-disc 

performance 

In addition to its role as one ofthe test houses participating 

in this study, MIRA was also called upon to prepare all 

of the honeycomb discs used in the te$ts. A further batch of 

Hexcel honeycomb material was ordered and a total of 210 

discs prepared. 

The new technique involving the crushing being abruptly

stopped at the threshold load proved relatively easy to carry 

out at the 4 mm/min crushing rate. However, variations 

between discs in the way the crush load fluctuated meant 

that in many cases it was not possible to achieve the threshold 

load of 3 I t 0.5 kN on a rising curve within the specified 

crush limits and 50 discs (247o of the total) were rejected 

because of this. Of the remaining specimens, however, the 

crush terminal load proved to be very consistent, with a 

standard deviarion of 0. l3 kN. 

Results of international test series-steering 

wheel performances 

Only the UK results are available at the time of writing, 

and these are summarised in tables 2 and3.In addition to the 

three production wheels, designated L, M and N in the 

tables, the UK programme also tested the TRRL/Sheller 

Clifford Safety Wheel. 

In these table$ the honeycomb indentation depths given 

are those measured using the TRRL method described on p7 

above, rather than using the German apparatus described on 

pl l. Trials with the Cerman apparatus on typical test discs 

showed that the mask with its array of holes was very timeconsuming 

to use, and in most cases no hole coincided with 

the lowest part of the dent. Measurements through the holes 

were therefore being taken from the sloping sides of the dent 

rather than from its flat bottom, and were neither as repeatable 

nor as accurate as those made by the much simpler and 

quicker TRRL method. 

Considering the honeycomb faceform tests first, the tables 

show good repeatability borh for indenrarion depth and 

for deceleration. On spoke/rim impacts, table 3, all four 

wheels passed on decelerarion (80 g/3ms; but only rhe 

TRRL/Sheller Clifford wheel passed on indentation ( I mm 

maximum allowable). Wheel L, with a mean dent depth of 

1.60 mm, was fairly close to passing, but the other two 

designs had dents of over twice the allowable depth. On the 

hub impacts, table 2, rhe TRRL/Sheller Clifford wheel 

passed easily on both dent depth and deceleration, and 

wheel L passed easily on dent depth and passed marginally 

on deceleration. The other two wheels failed very definitely 

on indentation depth ( I 9.6 and 5.6 mm) and wheel M clearly 

failed on 3ms deceleration. Wheel N showed extremely 

high peak deceleration levels but 3 out of the 5 samples 

passed on their 3 ms levels. This illustrates a well-known 

problem with using the 3 ms criterion-if the impactor hits a 

rigid non-yielding structure, a huge force will be developed 

on it but the whole speed change may be over in less than 3 

ms. This can give'a spurious pass result for a design which 

would be highly dangerous in a real head impact. To avoid 

the problem it has been suggested previously (19) that the 

principle of using a low-pass filter with a high cut-off frequency 

(eg CFC lfi)O) and a 3 ms window should be replaced 

by using a much lower cut-off frequency (eg CFC 

60) and a simple measure of peak deceleration. But this 

change has not been accepted for legislative use, and in 

practice the matter has not been important in Reg 2l testing 

because real car structures yield far enough to spread the 

deceleration pulse over longer than 3 ms. However, this is 

not true in the proposed steering wheel test, when the wheel 

alone is being tested on a rigid mounting, and a further 

criterion may be necessary to prevent manufacturers from 

arguing that a very hard wheel actually passed the letter of 

the test. Possibilities would be to use a lower CFC or to 

specify a maximum allowable peak figure of say 200 g as 

well as the 80 g/3 ms limit. 

The very deep dent in the honeycomb made by wheel M 

was cau$ed by the $teering column end, which projects 

about 20 mm from the wheel centre in this design and is 

covered only by a thin plastic trim which was punched 

through by the column end in each test. A major attraction of 

the proposed honeycomb test is that it would prevent this 

type of wheel mounting, which is now used by many manufacturers, 

from being fitted in future car designs. 

Tuming now to the suggested altemative of using a rigid 

spherical headform test, in spoke/rim impacts table 3 shows 

that all the 3 ms deceleration figures are well below 80 g. 

Since deceleration is the only measurement which can be 

made in this test, all the wheels would pass. This test clearly 

cannot detect the fact that the thin hard rims of wheels M and 

N would produce high local pressures which could break 

face bones, as shown by the honeycomb results in table 3. 

As mentioned above, (5) shows that spoke and rim impacts 

account for about 2/3 of all drivers' face injuries from their 

wheels, so the inability of a rigid headform test to detect 

dangerous spoke and rim designs makes it unsuitable for use 

in legislation. 

On the rigid headform hub impacts, table 2 shows rhat all 

four wheel designs greatly exceed the allowable 80 g. This 

was of course to be expected for designs M and N, as the 

headform simply hits the rigid end of the steering wheel 

mounting shaft, with no yielding structure. Testing of wheel 

N was stopped after two such impacts as the 500 g limit of 

the instrumentation was being exceeded. Wheel M was tested 

once at l/3 the standard speed (8.1 km/h instead of 24.1 

km/h), when it produced over 260 g. When used in cars 

these wheels would be mounted on steering columns designed 

to pass the Reg l2 ("Black Tufy") resr by yielding at 

below I l.l kN, which corresponds ro I I 100/6.8 X 9.81 : 

166 g, and so headform decelerations above this figure are 

not realistic. 

The TRRL/Sheller Clifford design and wheel L have rhe 

steering column end set well below the surface of the wheel, 

which is designed to yield to absorb energy. This feature 

worked well to produce low deceleration figures in the 

honeycomb tests, but in the rigid headform tests the spherical 

shape of the impactor allowed it to sink so far into the 

wheel padding that it hit the steering column end before it 

had stopped, whereas the flat honeycomb faceform had 

been decelerated more quickly immediately atter touching 

the wheel padding and so had stopped before reaching the 

column end. Since the flat disc is a more accurate representation 

of the shape of the human face than is the spherical 

headform (which was chosen to represent the round parts of 

the skull not the face), we feel that the high deceleration 

283

Table 2. Wheel Test Results. Flub Impacts Gulded Drop Flg Tests 

284 

l{heel 

TypeL 

I 

2 

1 

4 

5 

Hean 

Stsndard 

Deviation 

TyFeM 

I 

2 

1 

4 

5 

Mean 

Sbendard 

Devlation 

TypeN 

I 

z 

1 

4 

5 

Mean 

Standard 

Deviation 

TRRL/SheIler 

Clifford 

Safety Wheel I 

2 

, 

4 

, 

Mean 

Standerd 

Deviation 

DeFormable Honeycomb 

FeeeForm Tests 

Peek Decel ]ms DEcEI Max Dent 

I tl DePth' 

mm 

88.4 79.1 0.06 

87.4 78.0 0.05 

98.9 82.2 g.5Z 

99.7 77.5 0.06 

82.9 72.6 0,0J 

Bg.7 77.9 0.r4 

5.89 '.47 0,21 

241 ll9 L9.2 

2t8 rr} 2].0 

alt 99.7 r8,8 

22J Il4 18.2 

217 lZ2 18.7 

227 rI].5 19.6 

L2.6 8.56 1.94 

4rl 99.2 4.84 

68.4 6.04 

4lI 68.6 5.8] 

464 6]. r 5.91 

84.0 5.40 

429 76.7 5.61 

I0.0 r4.8 0.49 

74.J 70.4 0.0' 

7r.o 70.5 0.05 

7L.7 67.8 0.0I 

90.5 72,1 0.06 

76.8 7I.5 0.0] 

75.1 70.5 0.04 

1.47 1.65 0.014 

Rigid Spherical 

Impactor TEets 

PEak Jms 

Decel Decel 

s 9 

221 lr4 

254 rl0 

2r4 r04 

165 98.? 

L5? 92,6 

292 I0'.8 

42.' 8.7 

267 48.5 

( et 8.I 

kn/h) 

Not tesbedl 

See tExt 

511+ Il0 

526+ L57 

Not testedr 

Ses text 

I44 

19.l 

t88 102.6 

I87 98.6 

I79 

r99 

?L' 

96.9 

102.1 

106.' 

19' l0I.' 

r}.l ,.67

Table 3. Wheel Test Fesults. Spoke/Rlm Impactt Gulded Drop Flg Teats 

Wheel 

TypeL 

I 

2 

1 

4 

5 

Mean 

SbEndard 

Deviation 

Type 

H I 2 

1 

4 

5 

l'lean 

Standard 

Deviation 

TyPeN 

I 

z 

, 

4 

5 

Heen 

Sbanderd 

Devistion 

TRRL/Sheller 

Cll Fford 

Safety l{heel 

Mean 

StandErd 

Deviation 

I 

2 

j 

4 

5 

Deformable Honeycomb 

Faeeforn Tests 

Peak Decel ]ms Decel l,lex Dent 

I I Deptht 

mm 

106 6r.2 2.04 

I07 67.4 L77 

104 60,0 l.lg 

106 61.] L,52 

I09 65,0 I.50 

106.4 63.4 1,60 

l.B2 2.94 0. f 2 

101 70.0 2,9? 

I0l 66.7 ,.67 

106 6].0 2,5L 

104 60.0 2,27 

r05 51.7 2,4L 

lo].g 64.' ?.74 

L,gz 4,94 0.56 

lll 66.4 2,50 

105 58.7 2.88 

107 58.?. 1.6] 

108 J6.' 2.27 

r04 56.2 2.72 

1.07.4 59.2 2.40 

],5I 4.16 0.49 

52.2 Jt.g 0.0t 

4J. r t4.? 0,06 

52.J ,4,' 0.09 

44.1 t4.0 0.10 

48.5 11 .7 0.10 

48.0 ,4.4 0.07 

4.11 0.7J 0.0] 

Rigid Spherical 

Impector Tests 

PEek Jms 

Decel DecEl 

g 9 

97.1 52.7 

95.7 5r.5 

89.J 5t.0 

88.5 48.0 

94.2 45.6 

9t,0 50.? 

t,9l i,?l 

tt6 J6.5 

llg 5l.g 

II7 65.I 

120 65.1 

Il' 60.4 

117.0 61.8 

2.74 J.60 

r00 51.0 

lot 4J,2 

104 46.7 

107 '9.7 

r04 45.5 

roi,z 49,2 

2,77 6.51 

29,9 26.7 

,r,4 28.4 

t4.8 27.9 

t0.l 28,7 

29.9 26.' 

]1.6 ?1.6 

2.r2 1.05

levels found with the TRRL/Sheller Clifford design and 

wheel L indicate that the test method rather than the wheel 

designs is wrong. However if a manufacturer wishes to 

make wheels which pass this impact test with either a rigid 

spherical impactor or a flat honeycomb disc this could easily 

be done by using a greater depth of soft hub padding. 

To sum up these test results given in tables 2 and 3, it is 

clear that the rigid headform test would not be satisfactory 

because it cannot detect dangerous wheel designs which 

could produce face bone fractures. The honeycomb faceform 

test has been shown to have good repeatability and to 

be able to discriminate clearlv between different wheel 

designs. 

Steering Wheel Design Requirements 

to Pass the Faceform Test 

For the deceleration criterion part of this test and 

considering the wheel hub impact, the first e$sential is of 

course to use padding which provides a long enough 

distance for the impactor to stop before it hits the end of the 

steering column, without exceeding 80 g. Assuming the 

ideal case of constant deceleration at 80 g, the stopping 

distance required from velocity Y = 24.1km/h is given by 

5 = yz/2a = 28 mm, but in practice at least twice this figure 

should be used to allow for a more realistic deceleration 

curve. This necessary deceleration distance could of course 

be achieved by the wheel designer either by crushing foam 

plastic or by deforming metal, but the chosen method must 

not cause a local pres$ure which is high e;rough to deform 

the aluminium faceform in the test, ie. about 1.7 N/mm. The 

wheel designer must also make sure that if a large area of 

padding is used this padding is soft enough not to produce a 

total force on the whole faceform which gives a deceleration 

above 80 g, ie 80 X 6.8 X 9.81 = 5340 N. The faceform has a 

diameter of 150 mm and an area of 177fi) mm2, so if the 

padding covers the whole disc area then its crush strength 

must not exceed 534Oll77OO = 0.30 N/mmz. 

Considering rim impact, table 3 shows that deceleration 

is not critical for any of the wheels te$ted, because they all 

yield well below the force of5340 N needed to produce 80 g 

with 6.8 kg. But the problem now is to limit the pressure on 

the honeycomb to I .7 N/mmz. If a rim yield strength of 3330 

N is assumed (which would produce 50 g on the faceform) 

then this load must be distributed over at least an area of 

3330/l .7 : 1960 mmz. If the length of the "contact patch" 

between the rim and the faceform is approximately 150 mm 

(the faceform diameter) then the average width of this patch 

must be at least 1960/150 = l3 mm. Since many high quality 

steering wheels use rim section diameters of 25 or 30 mm 

this contact patch width requirement is quite reasonable, but 

it may require the designer to u$e a "chunky, soft feel" rim 

and to locate any reinforcing metal in this rim as far as 

possible away from the driver's face. On initial impact this 

will allow the wheel rim rather than the honeycomb to 

deform and spread the load over a large enough area to limit 

the contact pressure to 1.7 N/mmz. Similarly the wheel 

286 

spokes must be well padded and must be designed to yield at 

a suitable load when impacted. This will be easier to achieve 

with 3 or 4 spoke wheels but careful design should allow I 

or 2 spokes to be used if necessary for styling reasons; table 

3 shows that production wheel L, a 2 spoke design, came 

fairly close to passing the rim impact honeycomb test. 

This test may eliminate thin hard steering wheel rims, but 

may also make it difficult to provide a rim which is strong 

enough to be used by a large man to push his car out of a 

snowdrift, for example. However, this seems a small price 

to pay in return for a large reduction in face bone fractures. 

The design features described above (deep soft padding 

over the hub, thick soft rim with the reinforcing metal 

Iocated away from the driver, 3 or 4 spokes well padded and 

designed to yield when hit) have all been used in the TRRL/ 

Sheller Clifford safety wheel, figure l. Fifty of these wheels 

have been in road service since 1986 in Police, Army, and 

private Metros, with no reported problems and with users 

expressing a strong preference for these wheels over the 

standard ones. GM and Volvo wheels using the same design 

principles are already in widespread use, and while not quite 

meeting the proposed test requirements they would only 

need minor changes to do so. It is therefore clear that the 

honeycomb te$t can be passed by wheels of acceptable 

appearance and cost, and that provided the results from the 

other European countries confirm our findings on the 

repeatability and practicality ofthe test, there are no sound 

reasons to delay its introduction into legislation. 

Conclusions 

l. Recent accident data from UK and other countries have 

confirmed the results of earlier studies showing that in 

frontal crashes drivers wearing lap/diagonal seat belts often 

suffer brain damage and face bone fractures from hitting 

their steering wheels. 

2. Possible ways of reducing these injuries include the use 

of seat belt pretensioners and/or air bags, but these may be 

too expensive for widespread use in popular cars soon. A 

good cheap and quick way forward is therefore to improve 

the design of steering wheels to rnake them less likely to 

cause injuries when struck. 

3. Biomechanical data suggest that brain and face injuries 

are unlikely if the head deceleration during impact is kept 

below 80 g and the local pressure on the face is kept below 

1.4 N/mmz (200 psi). 

4. An energy absorbing steering wheel has been 

developed by TRRL and Sheller Clifford which can meet 

these requirement$ in an ECE Reg 2I|FMVSS 201 impact 

test (6.8 kg at 24.1 km/h, 15 lb at 15 mph). Road use has 

shown that this wheel is attractive to drivers and it would 

cost little, if any, more than standard non-energy-absorbing 

wheels. The paper has described the design requirements 

for this safety wheel. 

5. At the lOth ESV Conference in 1985 TRRLproposed a 

modification to the Reg 2I/FMVSS 201 Interior Fittings 

impact test by which an aluminium honeycomb faceform of

specified crush strength would be used instead of the 

standard rigid headform. This would en$ure that the 

allowable level of pressure on the driver's face is not 

exceeded in a steering wheel impact. The present paper has 

described subsequent development of this test to improve its 

repeatability. Results of tests on production steering wheels 

have been given which show that the current version of the 

test is highly repeatable, can discriminate between different 

designs of wheels, and is easily carried out by a competent 

test establishment. 

6. Results of similar tests on three designs of production 

wheels are now awaited from test houses in France. 

Holland, Germany and Italy for presentation to the EEC 

ERGA Working Group on Passive Safety. If these confirm 

the UK results given in this paper it is hoped that legislation 

can be agreed very $oon to require all steering wheels sold in 

Europe to meet this honeycomb faceform test. 

Acknowledgements 

The work described in this paper forms part of the 

programme of the Transport and Road Research Laboratory 

and the paper is published by permission of the Director. 

Crown Copyright. The views expressed in this paper are not 

necessarily those of the Department of Transport. Extracts 

from the text may be reproduced, except for commercial 

purposes, provided the source is acknowledged. 

References 

(l) Department of Transport, Compulsory Seat Belt 

Wearing, HMSO, London. October 1985. 

(2) Rutherford W.H. et al, The Medical Effects of Seat 

Belt Legislation in the United Kingdom, DHSS Research 

Report No. 13, HMSO, London, 1985. 

(3) Evans L. Occupant Protection Device Effectiveness in 

Preventing Fatalities, Proceedings of llth Intemational 

Conference on Experimental Safety Vehicles, Washington, 

USA, May 1987. 

(4) Mackay G.M. et al, The Methodology of In-Depth 

Studies of Car Crashes in Great Britain, SAE Paper 

850556. Detroit 1985. 

(5) Bradford M.A. et al, Head and Face Injuries to Car 

O c c upant s I n Ac c ide nt s-F i e I d D ata 1 98JJ6, Proceedings 

of IRCOBI Conference, September 1986. 

(6) Harms P.L. et al, Injuries to Restrained Car 

Occupants-What Are The Outstanding Prablems? 

Proceedings of I lth International Conference on 

Experimental Safety Vehicles, Washington, USA, May 

1987. 

(7) Thomas P.D. Head and Torso Injuries to Restrained 

Drivers From the Steering System, Proceedings of IRCOBI 

Conference, September 1987, Birmingham, UK. 

(8) Cloyns P.F. et al, Steering Wheel Induced Head and 

Faciul Injuries Amongst Drivers Restrained By Seat Belts, 

Proceedings of 6th IRCOBI Conference, September 1981. 

(9) Arajarvi E. A Retrospective Analysis of Chest Injuries 

in 280 Seat Beh Wearers, Accid. Anal. & Prev. Vol 20 No 4 

pp25l-9,1988. 

(10) Taniere C. et al, Field Facial Injuries and Study of 

Their Simulatian with Dummy, Proceedings of 25th Stapp 

Car Crash Conference. l98l. 

(l l) Sieffert U. Occupant Protection in Motor Vehicle 

Accidents, SAE Paper 870490. 

(12) Manson P.N. Structural Pillars of the Facial 

Skeleton: an Approach to the Management of Le Fort 

Fractures, Plast. Reconstr. Surgery, 1980 July 66(l) pp54- 

62. 

(13) Manson P.N. et al,Mrdlsce Fractures: Advantages of 

Immediate Extended Open Reduction and Bone Grafting, 

Plast. Reconstr. Surgery, 1985 76 ppl-10. 

(14) Nahum A.M. et al, Impact Tolerance of Skull and 

Face, Proceedings of the I 2th Stapp Car Crash Conference, 

Detroit, 1968, SAE Paper 680785. 

(15) Nahum A.M. The Prediction of Maxillofacial 

Trauma, Trans. Am. Acad. Ophth. Otol. 1977 54 pp932-3. 

(16) Hodgson V.R. Tolerance of the Facial Bones to 

Impact, Am. J. Anat. 1967 120 ppl 13-22. 

(17) Petty S.P.F. and M.A. Fenn, A Modified Steering 

Wheel to Reduce Facial Injuries and Associated Test 

Procedure, Proceedings of lOth Intemational Conference 

on Experimental Safety Vehicles, Oxford, UK, July 1985. 

(18) Robinson B.J. and S. Penoyre, Progress Towards a 

European Legislative Test for Steering Wheels, TRRL 

Working Paper WPIr'SD 90, Nov 1987. Issued as EEC 

Document ERGA Sll33. 

(19) Penoyre S. An Impact Test Programme Using Six 

Models of Lower-Medium-Sized Cars, Proceedings o{ lOth 

International Conference on Experimental Safety Vehicles, 

Oxford, UK, July 1985. 

The Inertial Flow Crash Sensor and its Application to Air Bag Deployment E 

Vittorio Castelli and David S. Breed 

Abstract 

The results of a theoretical investigation of a crush zone 

sensor based on the motion of a translating mass damped by 

inertial fluid flow through an orifice are reported. The 

characteristics of such a sensor are shown to be well suited 

to both frontal and side impacts. 


As a result of recent theoretical studies and field 

experience automotive engineers are directing increasing 

287

attention to the importance of crush zone sensing for triggering 

passive restraints such as air bags or seat belt tighteners 

(l), (?), (3), (4).* 

It now appears that side impact protection by means of 

inflatable restraints is also feasible. In this case, appropriate 

crash sensing must be in the crush zone although the desirable 

sensor characteristics may be somewhat different from 

the case of frontal impacts. 

Not much choice exists in the selection of crush zone 

$ensors. Ball-in-tube sensors. where the motion of a ball is 

damped by viscous gas flow through a small clearance, are 

the present industry standards. Crush detectors, which sense 

the progress of the crush zone in the vehicle body, are still 

under development. 

A new crash sensing device has been introduced which 

may possess characteristics more suitable and more easily 

tailored to all types of crush zone applications. It consists of 

an inertial mass, the travel of which is resisted by a damping 

force generated by inertial gas flow and thus quadratically 

related to the velocity. A spring provides the necessary bias 

force. 

This sensor has several favorable properties; it can be 

designed rather flat in the sensing direction and, thus, easier 

to mount; it can be designed to be much less sensitive to very 

short duration pulses, essential in sensing side impacts for 

some types of cars and an excellent property to provide 

immunity to hammer blows and other sharp impulses; it can 

be designed less sensitive to lateral vibrations than the ballin-tube 

sensor. 

Encouraged by such promise, an analytical and experimental 

study was undertaken to construct and validate a 

mathematical model for this device. The theoretical results 

are reported herein. 

Common strategies for crush zone crash 

sensing 

For orthogonal and angular frontal impacts, these authors 

have identified the existence of a crushing front which progresses 

through the vehicle in a manner characteristic of the 

vehicle itself, the type of crash, and the speed. Scaling 

techniques were introduced which help analyze the performance 

of a sensor in many different crashes based on data 

obtained in only a few tests. 

Proper frontal crash sensing has been performed by either 

crush zone sensing or pa$senger compartment (single point) 

sensing. These authors have recently concluded ( I ), (3) that 

the former strategy produces better results and has more 

general applicability. It requires the presence of at least one 

sensor within the crush zone for all crashes requiring restraint 

deployment. Crush sensing switches must signal the 

arrival of the crushing front. Other sensors must measure as 

closely as possible velocity change and trigger when its 

value equals or exceeds the injuring level (typically l0-12 

MPH). 

The sensing problem for side impacts can be stated in 

terms of the following principles (2): 

*Numbcrs in parenthcses designate rcferences at end of papcr. 

288 

(a) the time budget is very short since injuries to the occupant 

are the result of intrusion of the door structure. 

(b) the rapidity of the door structure intrusion causes passenger 

compartment sensing to trigger air bag deployment 

much too late to be effective. Crush zone sensing is necessary. 

The $ensors must be mounted on the door reinforcement. 

not on the skin. 

(c) at least one sensor must be in the early crush zone which 

means that, in many cases, up to three sensors per side may 

be necessary: just before the A-pillar, just after the B-pillar 

and at the door center. 

(d) in vehicles with stiff door structures, the sensors should 

measure velocity change threshold for short pulses (on the 

order of l0 milliseconds); the threshold speed should be 

equal to the injuring velocity change, i.e., l0-l? MPH.In 

cases of weaker side structures, the threshold velocity 

change for short pulses should be higher than injuring velocity 

change; possibly up to a factor of two. 

(e) marginal crarthes have a much increased time budget. 

(f1 sensing "L-crashes", where the door structure does not 

intrude toward the occupant, may not be as important due to 

the usually low level of the velocity changes involved. 

However, it could be accomplished by means of passenger 

compartment sensing. 

(g) the distance traveled by the occupant relative to his own 

vehicle before striking an injuring part of the interior is 

shorter than in frontal crashes. Therefore, the 5 inches minus 

30 millisecond criterion will have to be modified for 

these crashes (a further reason for its modification will be 

the requirement of much faster inflator action). 

Safing (arming) sensors for these applications could be 

crush sensing switches. 

In both frontal and side collisions, lateness in triggering is 

a very negative characteristic ofany sensor system because 

the out ofposition occupant can be injured by the deploying 

air bag. 

The inertial flow damped sensor 

The requirements outlined above suggest a sensor which 

responds to velocity change, with a response curve to acceleration 

pulses which can be tailored to several applications. 

For crush zone sensors, the response would be divided into 

three regions: the very short time duration, in which the 

threshold may be higher than the injuring velocity change; 

the intermediate pulse durations, in which the threshold is 

equal to the injuring velocity change; and the long pulses 

where the sensitivity decreases again. This response curve 

may look like the ones represented on figure l. The solid 

curve has the shape required for a side impact crash sensor 

for vehicles with weak door structures. This behavior may 

be deemed quite desirable for frontal crash sensing in order 

to provide immunity to sharp blows. The dotted curve is for 

side impact crash sensing in cars with very stiff side structure 

and displays the maximum sensitivity required of a 

frontal crush sensor. The ball-in-tube sensor also displays a 

slight decrease in sensitivity for short pulse duration. 

The fact that it may be desirable for the sensitivity to

20 

15 

10 

5 

0 

0 20 40 60 

Pulse Duration, msec 

80 

Side lmpact w.Soft Target 

Frontal lmpacts 

Side lmpact w.Hard Target 

Flgure 1. Crush Srnsor Charactsristic$ 

decrease for acceleration pulses of short duration suggests 

employing a physical effect which generates force at a higher 

rate than linearly with increasing velocity. Hence the idea 

of using a damping mechanism proponional to the square of 

the velocity. Fluid flow through an orifice has just such a 

behavior. 

A possible implementation of such a device is depicted in 

figure 2. An elastic diaphragm suppon$ a sensing mass and 

keeps it pressed against a top stop with an appropriate bias 

force. An acceleration pulse in the sensing direction causes 

the mass to move through the deflection of the diaphragm. 

This motion displaces the fluid filling the cavity (probably 

air) and forces it to flow through the orifice. The resistance 

to this flow causes a pressure difference between the two 

sides of the diaphragm which resists the motion of the 

sensing mass. As long as Mach number effects on the velocity 

can be neglected, this pressure difference is proportional 

to the area of the orifice and the second power of the fluid 

velocity. 

The orifice in the figure is shown of a circular shape. 

Other geometries are also usable, as they might facilitate the 

design of the diaphragm. 

Mathematical sensor model 

A very useful tool in adapting a sensor to a variety of real 

crash applications is a realistic mathematical model. Under 

Flgura 2. InFrtial Flou, Damped Sensor 

the stimulus of axial acceleration only, the behavior of this 

sensor is simulated by the following analysis. 

The nomenclature is 

A = sensed acceleration 

A = effective area on which pressure acts 

B = bias force 

C = discharge coefficient 

e = volume displaced by motion x 

f = force on the sensing mass 

G = sensor mass 

k = diaphragm spring constant 

m = gas maSS 

M = Mach number 

I = mass flow rate 

r = orifice radius 

R = gas constant 

t = time 

T = temperature 

y = gits velocity in the orifice

v = gas volume 

x = forward travel of the sensing mass 

7 = ratio of specific heats for the gas 

p = ga$ density 

Subscripts: 

d = downstream 

f = fill 

g=gas 

i = initial 

s = diaphragm spring 

u = upstream 

+ = forward 

*=aft 

For the sake of simplicity, the analysis is going to be 

presented for the case in which the ga$ velocity in the orifice 

does not achieve a Mach number of unity. Therefore, its 

accuracy range is limited to pressure ratios below the critical 

value (approximately 2: I for air). Furthermore, the analysis 

is going to consider perfect gases with constant specific 

heats. 

For a circular sharp edged orifice the discharge coefficient 

C is typically taken to be equal to 0.6. Its value may be 

quite different for other geometries. Thc gas constant for air 

ts 

^ 497?0 ft tb 

'( = t0 sfu;F- 

Thc rqurtion of motion of thr nnring mmr ir 

i. 

o T; = -frfr+ Go 

dt' 

and 

whcrt 

fr=B+Lx 

fr--(P* - P-l,t 

P+=I+RT, P* tP-trf 

F+=m+lV+, 9-=n*lV- 

V*t{*-r*, V_=l|r-+cr. 

Thc mrrt flow xrom thr orificr obryr thr 

follwingrquotiom 

d^, 

dr 

= +{l 

Pt Pit 

-i* = vi* Ei- = vlfr 

frr' 

?hr mmr flow rrtr through tht orificc 

givrnby 

,' - lzuP(P-llti 

9-nr'Crr(ffif 

rnd 

I'. 

"=f3)' 

'Pdl 

ar 

The mass flow rate through the orifice is 

given by 

and 

Q = n ,z c na( 

, = ( \ \ Y 

' Pt / 

2yP(P* l) 

RTu(y-l) 

y- I 

Starting from appropriate initial conditions, integration 

of the above equations in time with the time varying input of 

the body acceleration a, yields the motion of the sensing 

mass. When the travel reaches the contact spring, the spring 

force may change (note that the contact spring would typically 

start in contact with the sensing mass). When further 

travel causes the contact to be clo$ed. the sensor circuit is 

activated. The integration can continue and follow the motion 

further and through possible reversal, Particularly in 

cases where hard contact with the anvil is not achieved. an 

flccurate evaluation of the contact dwell can be made. 

For the sake of producing a useful model, integration of 

the above equation is best performed numerically using one 

of many available algorithms. The speed of integration 

using modern computers is so high that, for most applications, 

even the simple Euler method is adequate since the 

required rccuracy can be achieved by decreasing the time 

step siae. 

Refinements of the above-described sensor model can be 

achieved by including choking (mach number equal to unity) 

effects on the gas velocity in the orifice and by including 

the energy equation. The latter correction would account for 

the fact that the energy dissipation in the orifice flow cause$ 

an increase of the temperature (hence the pressure in the 

cavity). 

Simulation re$ults 

I 

)u 

A computer program representing the above equations 

was con$tructed and exercised under a variety of conditions. 

In order to exemplify the ability to tailor the properties of 

this sensor, this section displays families of responses to 

haversine acceleration pulses. 

Figure 3 shows the fire-no fire threshold for a sensor with 

the following dimensions: 

Sensingma$$EBgrams 

Bias force = 4.25 g's 

Bias stiffness = 3.64 lh/in 

Diameter = 2 inches 

Orifice radius = 0.045 inches 

Aft volume = 0.107 cu. in. 

Filling condition = I atmosphere 70 deg. F. 

The three curvcs indicate the effect of varying the distance 

which the sensing mass must travel before closing the

25 

20 

15 

t0 

5 

0 

------ 

AV, MPH 

0 20 40 60 80 


0.050" 

o,067 

" 

0.075" 

Flgure 3. Sensor Gherictsrl$tlcs. Effect of Travel 

contact. Longer travel desensitizes the sensor in a rather 

uniform manner at all pulse durations. 

The rather high slope of the curves with increasing pulse 

duration is due to the high spring rate used for the 

diaphragm. 

Figure 4 represents the effect on the response characteristics 

of the same sensor of variations in the aft volume. 

Smaller aft volumes cause a vacuum to be readily formed 

behind the diaphragm when its speed is high. Therefore, 

small aft volumes desensitize the sensor for short duration 

pulses. The effect is limited to short duration pulses; therefore, 

the abscissa axis has been expanded. 

Figure 5 indicates the effect on the sensor response of 

varying the orifice diameter. Smaller orifice diameters increase 

the flow impedance thus desensitizing the sensor. 

This effect is less pronounced for pulses of longer duration 

since, in this case, the sensor response is dominated by the 

bias. 

Figure 6 points out the effect on the sensor response of 

varying the bias force. A higher bias force increases the 

slope of the response curve for pulses of long duration. 

Therefore, the sensor becomes desensitized to gentle crash 

signals. 

The general behavior of the response curves carl be divided 

into major regions. The first region is characterized by an 

increased sensitivity to extremely short pulse$ (below 5 

AV, MPH 

0 10 20 30 40 


----** 0.471 cu.in. 

0.107 cu. in. 

0.050 cu. in. 

Flgure 4. $ensor Character|stlcs. Efioct of Aft Volume 

milliseconds) due to gas compressibility. This part of the Flgurr 5. Senror Characlerlrtlcr. Effect of Orlllce Dlsm€t€l 

25 

20 

r5 

10 

5 

0 

I 

r l 

AV, MPH 

\ ' t r r r . . . . " " : ; 

\- -::F 

0 20 40 60 80 

Pulse 

Duration, msec 

0.045" 

0.040" 

0.035" 

291

25 

20 

15 

10 

5 

0 

- * - * F - 

AV, MPH 

.., 

* t t 

a'*? 

20 40 60 80 


4.25 g's 

6.00 g's 

8.00 g's 

FIgure 6. Sensor Charsctsrl8tlcs. Efl€ci of Blss Force 

curves may not be realistic because the intervening structure 

would act as a mechanical filter and would prevent such a 

pulse to reach the sensor. The second region, from 5 to 15 

milliseconds displays the effect of the nonlinear character of 

the inertial flow damping: shorter pulses create a higher 

damping force. For pulses longer than l5 milliseconds the 

effect of the nonlinearity is mild and the sensor behaves as 

an acceleration integrator until the uniform positive slope 

characteristic of bias force and mass response takes over. 

The properties ofball-in-tube sensors are sensitive to the 

value of the gas viscosity which varies with temperature. 

Elaborate compensation $chemes must be employed in the 

design. The functioning of inertial flow damped sensors 

does not rely on viscosity. Thus the effect of temperature 

should be felt only when compressibility effects are present 

(pulses of very short duration). Figure 7 presents simulation 

results on the effect of ambient temperature on a sensor 

which had been filled with ambient air at 70 degrees Fahren* 

heit and sealed. Note that the exaggerated situations of an 

operating temperature 200 degrees F above and 104 degrees 

F below ambient was picked. As expected, the effect of 

ambient temperature in the range of pulse durations of interest 

is minimal. 

Conclusions 

The above-presented data give an idea ofwhat effect each 

of the essential design parameters have on the performance 

292 

25 r{ 

20 

15 

10 

I 

II 

I 

f 

I t 

JV, MPH 

20 

40 

PulseDuration, 

msec 

70 deg F 

270 deg F 

-34 deg F 

60 80 

Flgure 7. Ssneor Gharacterlstlcs. Effect of Operatlng 

Temperature 

of this type of sensor. The decreased sensitivity to short 

duration pulses makes it a particularly good candidate for 

mounting in lateral crush zonetr where the structural rigidity 

is low and in frontal crush zones where the danger of hammer 

blows exists. A flatter response curye can be obtained 

by proper choice of parameters. 

It can be shown that this sensor can also be used as a 

distance measuring device for families of similar acceleration 

pulses. However, this property is not applicable because 

this type of functioning requires rather large displacements 

of the sensor bodv. 

References 

(l) "Problems in Design and Engineering of Air Bag 

Systems", by David S. Breed & Vittorio Castelli, SAE 

Technical Paper 880724, March 1988. 

(2) "Trends in Sensing Side Impacts", by David S. Breed 

& Vittorio Castelli, SAE Technical Paper 890603, March 

89. 

(3) "Trends in Sensing Frontal Impacts", by David S. 

Breed & Vittorio Castelli. SAE Technical Paper 890750. 

March 89. 

(4) "New Sensor Developments Leading to Sensor System 

Simplication", by Roben W. Diller, SAE Technical 

Paper 

841218, May 1984.

Vehicle Tests Required For Air Bag System Design 

David J, Romeo, 

Autoliv North America, Inc. 

John B. Morris, 


Abstract 

At this time, a consen$us does not exist regarding the 

number and types of crash tests which need to be conducted 

to establish a basis for crash sensor threshold behavior. 

Experience obtained from the 1973-1976 General Motors 

air bag equipped cars ( l-3),* the Police Fleet Retrofit Driver 

Air Bag Program (4), and the Ford Tempo GSA Project 

(5), are reviewed and results from these projects, and from 

tests conducted by and for the automotive industry, are 

discussed. Such sensor threshold tests could be conducted 

as par"t of the design of an air bag system. 

Introduction and Summary 

Over the past twenty years, numerou$ methods have been 

used to design, develop, and evaluate air bag restraint 

$ystems. These methods have been reviewed, and this paper 

identifies five distinct categories or types of vehicle tests 

which should be performed. The tests are, in part, 

categorized by practical factors such as whether test 

dummies, test drivers or neither are required; whether a 

complete test facility is required, i.e., crash barrier, test 

dummies, high speed cameras, etc., and finally whether the 

entire system or only component$ are actually needed. 

The five types of vehicle tests are: 

l. High Speed Impact Te$t$ (test dummies, 

complete test facility, total destruction of vehicle). 

2. Low Speed Threshold Barrier Tests (no driver or 

dummy, moderate damage to vehicle). 

3. Fleet Testing (general popularion drivers, no 

damage to vehicle intended). 

4. Test Track Tests-Normal including rough road 

(test driver, little or no damage to vehicle). 

5. Severe Rough Road Exposures (with or without 

test driver, moderate damage to vehicle). 

In the following section, a general discussion of each of 

these categories is given. Specifics of proposed Severe 

Rough Road Tests are given. An attempt to evaluate the tests 

proposed in terms of real world air bag deployment 

experiences are presented. Finally, conclusions and 

recommendations are presented. 

Definition of Vehicle Test Types 

High speed impact tests 

These, of course, are where air bag deployment occurs. 

Although full systems tests are preferred, these tests can be 

+Numbers in parentheses designate refercnces at end of paper. 

conducted with or without a complete working sensordiagnostic 

assembly (time delayed barrier contact switches 

can be used to simulate sensor closure). Tests without cra$h 

sensors are sometimes conducted early in a development 

program to determine dummy performance given a 

specified sensing time, prior to the availability of 

production crash sensors. Equally clear, these are the 

expensive tests since a complete facility is required and the 

vehicle is destroyed. The baseline test here is the Federal 

Motor Vehicle Safety Standard (FMVSS) No. 208 test 

condition at 30 mph (48.3 kph) full frontal. 

The FMVSS No. 208 test, although obviously the most 

important with regard to safety standard compliance, 

appears to be less critical with regard to design ofthe air bag 

system, particularly the crash sensor. This is illustrated in 

table I which summarizes results for the 30 mph barrier 

crash for eight different vehicles, all with the same retrofit 

air bag system that was used in the Police Fleet 

Demonstration Program, reference 4. In general, excellent 

results were obtained in all of these vehicles despite the fact 

that they differ in size, weight, and engine drive train 


Table 1. Hesults ot pollce tleet drlver alr bag retrollt syslems ln 

varloue aulomoblles-3O mph, FiIVSS 208 craah teat. 

Ford LID 

Yeh lc l. 

D'odga Dlplmat 

Dodg€ Arl6s 

Chevrol6f Caprlc€ 

0pal Kadrtf 

Stcrl Ing 800 

YUgp GY 

Ford llustlng 

Dafe 

9/9/8i 

t2^t/6, 

5lt186 

1l?E/86 

4/ZlE7 

9/2tlS7 

tlilaE 

?lt6lE9 

_HIC- 

176 

9t9 

t76 

572 

d66 

505 

21' 

772 

Chgst ors 

S€n6'0r 

Cfosurs 

?0 msm 

This basic compliance test can be extended through the 

t30 degree barrier face, 35 mph vehicle assessment 

condition, and various dummy size, seat position, manual 

belt usage conditions as program budgets allow. The 

following table lists these in order of importance. 

The first three tests are Federal Motor Vehicle Safety 

Standard (FMVSS) No. 208 tests. The first test is the 

predominate test. It determines air bag performance when 

the safety belts are not used. The second test evaluates 

system performance in oblique crashes, a common 

occurence in real world accidents. The third test provides 

data on air bag performance when used with safety belts. 

The fourth te$t, the New Car Assessment Program (NCAP) 

test, provides performance data when the system is 

subjected to a more severe crash speed. The fifth test 

evaluates the adequacy of the crash sensor threshold setting 

/tE 

61 

,7 

57 

48 

5l 

5t 

48 

t6 

t6 

24 

t,f 

IE 

rt 

t4 

293

to react to this type of impact in time for the air bag to 

provide the protection desired. 

Speed 

30 mph 

30 mph 

30 mph 

35 mph 

30 mph 

Gondltlon 

full frontal 

+ 30 dBgree$ 

wmanual belts 

full trontal (NCAP) 

centered pole 

Bslt Use ln T6st 

No 

No 

Yes 

Yes 

No 

Additional high speed impact tests, which have been 

conducted (6), include: 

Speed 

30 mph 

E0 mph (closing) 

20 mph 

15 mph 

Condltlon 

30 deg, A-pillar 

car to car, half car offset 

underride 

frontal imDact 

The 60 mph closing speed offset car to car test is quite 

representative of many real world accidents. However, it is 

expensive and can be simulated with a single vehicle into a 

half barrier or pole offset to the driver side. This would be a 

worthwhile addition to a complete test matrix. 

Less can be said for the side impact, underride, and lower 

speed tests which all represent longitudinal velocity 

changes of a lower magnitude which in turn produce 

ambiguous results regarding sensor closure and dummy 

performance requirements. The results are considered 

ambiguous because at the relatively low longitudinal 

"delta-V's" 

experienced in these crashes the dummy injury 

indices are always low, regardless of whether or not the air 

bag deploys. Because of this the outcome always tends to be 

considered acceptable. ln other words, from the dummy 

results it is not possible to determine clearly a need for 

deployment and until a larger data base of real world 

accidents has been analyzed the need will continue to 

remain unclear. 

It is believed that the dummy results for tests which 

produce "barrierequivalent" velocity changes of l5 mph or 

less are oflittle value because ofthe lack ofbiofidelity at the 

deceleration levels experienced. Seat friction and, more 

importantly, muscular restraint of arms and legs are known 

to dominate occupant kinematics at these low speeds (7). 

The same reservations must be made with regard to 

computer simulation models which treat the occupant as a 

"free 

body." These results tend to generate sensor closure 

requirements inconsistent with real world experience. 

Reference 8, for example using the 5 mph, free body criteria 

concludes that air bag deployment is needed in situations 

such as running a car into deep water, a snow bank, or into a 

crash attenuation barrier. All of these events (crashes) 

produce vehicle decelerations in the range ofup to 6 or 8 g's 

and the model allows the dummy to strike the interior at a 

large "delta-V,o' indicating air bag deployment is 

warranted. Real world experience, on the other hand, 

294 

indicates that the air bag in fact is not needed in these 

situations. 

Consequently, if analysis of air bag component 

performance is to be conducted at these low speeds, it is 

suggested that incorporation of a preload into the dummy or 

computer model of the dummy be used. A preload of the 

order of4 g's is suggested as a starting point. 

Low speed threshold barrier tests 

These crash te$ts are conducted to determine the threshold 

of sensor closure in a frontal barrier crash. Based upon 

many factors, but primarily human injury tolerance, the 

threshold speed is normally chosen to be 1ill9 kph (9.9- 

I I.8 mph). By definition of threshold, 50 percent of the tests 

conducted at this condition will result in sensor closure (air 

bag deployment) and 50 percent will not. Therefore, it is of 

little value to conduct tests at this speed (unless an unreasonably 

large number of them are to be done). Rather, we 

mu$t test at speeds below and above where non-closure or 

closure must occur. Thus we define the terms; 

"Must 

not close with reasonable certainty" and 

"Must 

close with reasonable certainty." 

These conditions, based again on tolerance to injury, but 

also somewhat on system capability, are 4 kph (2.5 mph) 

above and below threshold velocity. 

Therefore, the following te{it protocol is established- 

For threshold of 16 kph, test at; 

t 12 kph with result that sensors must not close, and 

at 

r 20 kph with result that sensors must close. 

For threshold of 19 kph, test at; 

r l5 kph with result that sensors must not close, and 

at 

. 23 kph with result that $ensor must close. 

The exact threshold, i.e., between 16 and 19 kph, is selected 

based upon the particular project's analysis of human 

tolerance, car interior friendliness, system capability, indication 

of safety belt u$e, etc. 

Fleet tests 

These so called "fleet tests" should use as many vehicles 

(from one or two, up to several hundred) as possible. The 

complete air bag system is installed in the vehicle and the 

vehicle is driven oyer a wide range of normal conditions 

with mileage accumulation. The purpose of these test$ are 

many. Examples of design problems which have been experienced 

in the Police Fleet are listed below as illustrations. 

l. Appearance, aesthetics, consumer and service acceptance. 

The vehicles are made available to a wide range of 

people who look at, feel and experience the system. This is a 

"Show 

and Tell" exposure. Examples of concems to be 

evaluated are: 

r knee bar intrudes, is difficult to get in and out of 

car (was not a problem).

electrical connector coil broken due to improper 

steering wheel removal (was a big problem). 

r readiness indicator light stays on for sufficient 

duration to be seen and is located so that it provides 

the proper incentive when it stays lit (fault 

indicated) that service is sought. In the Police 

Fleet, the indicator light was, in some cases, ignored, 

and in some cases, rendered inoperative 

(broken). 

r appearance feedback-is air bag module attractive 

and does it obstruct view of in$truments. 

Steering wheel leather wrap was a big "plus" 

toward offsetting large air bag module 

appearance. 

2. Normal function of diagnostic and its exposure to 

normal vehicle electrical variants such as, low battery, electrical 

voltage surges, and/or interruptions during starting. 

Problems experienced 

include: 

. improper indication of fault due to electrical system 

transients. 

r diagnostic memory drained battery after several 

weeks of parking. 

. unnecessary fault readout information was given 

to the service mechanic. 

3. Mechanical response of components such as steering 

wheel, clock spring electrical connector, knee bar: 

. steering re$ponse affected by added mass ofsteering 

wheel air bag module. 

clock spring electrical connector drag and noise 

can be felt and heard 

interference with directional signals, horn. 

4. Environmental exposure to rain, cold, salt, etc. Problems 

have included: 

r moisture in diagnostic. 

r corrosion of sensors, wiring harness, connectors. 

5. Normal rough road, rough use exposure, parking lot 

speed bumps at low speed, i.e., less than 5 mph, impacts, 

door slams, hood slams, etc. No problems have been experienced 

to date from these exposures. 

Test track tests 

Test track tests are conducted for purposes similar to 

Fleet Tests except that the road (test track), driver (professional), 

and vehicle performance (speeds, cornering g's, 

etc.), are defined a priori. A second distinction is that expo- 

$ures more severe than normally encountered in the real 

world are studied. 

In some cases, many of the potential problems which 

eventually show up in the Fleet Tests are seen early enough 

to permit resolution before full scale production, as might 

be the case with Fleet Tests. 

Normally included in these evaluations, are a series of 

tests which include driving the vehicle over man made 

"Belgian 

Blocks", steps, bumps, pot holes and ramps. 

These are the so called "shake. rattle and roll" tests commonly 

used for durability testing. These tests have been 

used in earlier programs and found to be of little value in 

evaluating sensor closure. Put simply, they lack severity. In 

fact, it has been found that the inadequacy of these tests to 

produce potential crash sen$or closure situations has led to 

and forced the development of the "Severe Rough Road 

Exposures. 

" 

Severe rough road exposures 

These tests simulate vehicle exposure outside of the normal 

rough road durability type envelope often run on new 

vehicles. These tests all involve impacts which will damage 

the vehicle to some extent (scrapes, dents, tire and wheel 

failure) and cause deceleration pulses which may be sufficient 

for sensor closure. The extent of damage and severity 

of exposure to an occupant as a function of these impact 

parameters is not known at the beginning of the test. One 

can only state that the condition of sensor closure should be 

warranted. This can be thought of as shown in the following 

table. 

Vehicle 

damage 

None or slighl 

Minor 

Moderate 

Severe 

Inlury 

potcntial 

None 

None 

Minor 

Mod€rate 

Sensor 

closure 

No 

No 

Borderline 

Yes 

Equlval€nt 

barrler 

speed 

"delta-V" 

5 mph 

7 mph 

10 mph 

15 mph 

Four types of severe exposures have been proposed and 

will be discussed 

in the next section: 

2. Railroad Track 

3. Ditch 

4. Deer 

Specifics of Severe Rough Road 

Exposures and Deer Impacts 

Curb tests 

These tests involve driving the vehicle over a step or curb 

at increasing rates of speed and for increasing curb height to 

a point where sensor closure occurs. Its occurrence is 

compared to severity of exposure both with respect to the 

vehicle and the driver. The literature recommends curbs 

ranging from 100 mm to 200 mm in height and speeds from 

l0 to 50 kph. 

Some indication of the relationship of the recommended 

test curb height to wheel hub height is shown on the 

following table. Forreference, curb height can be expressed 

as a ratio ofvehicle ground to centerline ofhub distance. For 

13 inch to 15 inch wheels, this distance is approximately 

280 mm plus or minus l0 percent, and gives ratios of; 

295

Curb Height 

100 mm 

150 mm 

200 mm 

% Curb Height 

Axle Helght 

The following test protocol is recommended: 

h = 100 mm, v = 20,30,40,50 kph with driver 

h = 150 rlrl, v: 10, 20, 30, kph and up until tests are 

terminated due to sensor closure or severe vehicle 

damage 

h * 200 mm, may approach low speed barrier 

conditions 

Railroad track 

In this test, the vehicle is driven such that the undercarriage 

is snagged or impacted by a segment of railroad track' 

This simulates the single most common cause of unwanted 

airbag deployment in the early General Motors airbag fleet. 

Ceneral Motors recommends meeting a condition of a l'2 

inch (30 mm) interference at 20 mph without sensor closure. 

This recommendation may need further definition as to 

what part of the undercarriage to strike. The rail segment is 

approximately 300 mm in width. This can also be run at 

increasing speeds of 20, 30, 40 kph etc. until termination 

due to severity. 

Ditch 

Note that in the curb tests, vehicle loading occurs due to 

impact to the tires, wheels and front axle. In the railroad 

track test, which can also simulate a boulder contact, impact 

occurs to the vehicle undercarriage and in the low speed 

barrier tests, loading obviously occurs through the bumper. 

The ditch tests decelerate the vehicle through all three load 

paths, bumper, undercarriage and tires and, consequently 

are less quantifiable than the others. It is essentially an 

extension of a dirt road testing campaign and has its origin 

in the police fleet retrofit program where it was found to be 

ofconsiderable value. 

Showing Direction of Travel 

h . dround to center lind of axle WB . Wheel base 

Flgure 1. Concrete dltch. 

296 

.35 

.53 

.71 

The dimensions of the ditch are presented in figure I and 

may be modified to suit the particular vehicle. Tests can be 

initiated at 20 kph with a driver and continued as severity of 

results warrants, 

Deer impacts 

The Insurance Institute for Highway Safety, Reference 9, 

announced that in Pennsylvania the game commission reported 

that close to 40,000 deer were killed in 1988 in 

collisions with vehicles. Injuries were relatively rare, although 

considerable property damage frequently occurred. 

Reference 9 also reports that there are 20,000 such accidents 

per year in Michigan and when a car impacts a deer at high 

speeds, the car can crush locally and impart a l0 mph velocity 

change to the air bag crash $en$or while causing a relatively 

minor velocity change to the vehicle, resulting in an 

unwanted air bag deployment. On the other hand, the air bag 

may help prevent occupant injury in the event the deer 

carcass penetrates the windshield after the air bag deployed' 

General Motors, Reference 10, in recommending tests for 

evaluating air bag sensor performance in circumstances 

where deployment is not desired, proposes the following car 

to deer impact: 

"Tow 

the car at 50 mph into a horizontally oriented 

water-soaked boxer training bag simulating a deer. The 

bag is suspended at a height which aligns its center 

with the sensor. The water-soaked Everlast Model 

4543 training bag weighs 50 kg." 

Real world accident experience 

In this section real world air bag deployment experience 

is examined in an effort to establish the merit of the aforementioned 

vehicle tests. In general, the deployments should 

fall into the first, second or fifth categories listed under 

Introduction and Summary. That is "High Speed," "Low 

Speed," or Severe "Rough Road." ln a few case$, however, 

they fall under "Fleet" or "Test Track" tests. 

General Motors 1973-1976 ACRS vehicles 

References 4,5, and 6 provide data on General Motors air 

bag equipped vehicles produced from 1973 through 1976. 

Over I 1,000 air bag equipped vehicles were produced during 

this period. These data contain over 200 deployment 

accidents. As seen in the following table, accidents involving 

another car produced the most deployments in frontal 

impacts. Deployments produced by undercarriage contact 

are not included but are reported later. 

General Motols Flset Frontal lmpact Demonstration 

Object struck 

Car 

Pole 

Tree 

Truck and Van 

Other 

% of deploymenta N = 152 

64 

I 

I -l 

11

A distribution of all deployment accidents is shown on 

the following table. Undercarriage contact produced a significant 

number of the deployments. 

Crash mode 

Frontal 

Undercarriage 

$ide 

General llotor$ Flest 

The severity of the accidents resulting in the deployment 

of the air bag is shown in the following table. The velociry 

change, in most cases, was computed using the CRASH 

computerprogram. Reference I I reports that CRASH tends 

to underestimate velocity change at speeds under 30 mph, 

the speeds at which most crashes occur. The velocity 

changes shown in the following table were therefore corrected 

using an equation similar to that recommended in 

Reference ll. The velocity changes obtained from crash 

recorders installed in some of the vehicles were used as 

reported. 

Crash 

$EYerity 

Minor 

Moderate 

Severe 

Gcneral MotorE Fle€l 

Veloclty 

change 

Lessthan 13 

mph 

13-20 mph 

2G-50 mph 

Police fleet 198L1988 

% of deploymenls N = lg4 

79 

13 

I 

% of deploym8nt$ 

1.1= 

98 

26 

40 

34 

The following observations are made from accident repons 

of police vehicles in which retrofit driver air bag 

system$ were installed. Approximately 5fi) kirs were installed 

in the Ford LTD Crown Victoria, the Dodge Diplomat/Plymouth 

Gran Fury, and the Chevrolet Caprice. The 

number of deployment$ are quite small, 29 resulting from 

frontal impacts, neglecting undercarriage, for which we 

have data. As seen from the following table, a significant 

number of deer impacm resulted in air bag deployment. 

Obfoct struck 

Car 

Deer 

Pole 

Tree 

Other 

Pollce Fleat Frontal lmpact Dcploymsnts 

% of deployments N = 29 

55 

21 

10 

7 

7 

As with the C€neral Motors fleet, a significant number of 

deployments as a result ofundercaniage contact occurred in 

the police fleet. The impacts are summarized on the following 

table. 

|mpact 

Frontal 

Undercarriage 

Polics Fl€€t 

% ol deployments N = 34 

85 

15 

The severity ofthe accidents resulring in air bag deployment 

is shown in the following table. The number of velocity 

changes calculated is very small (N = 20). As with the 

General Motors data, the CRASH computed delta-V's have 

been corrected. 

Pollce Fleet 

Craeh severlty Voloclty change % ol deployments 

H=20 

Minor 

Moderate 

Severe 

Ford Tempo f985-1988 

Less than 13 mph 1 5 

13-20 mph 65 

More than 20 mph 13 

The Ceneral Services Administration purchased 5,O00 

1985 Ford Tempos equipped with driver side air bags. The 

following observations can be made from the experience 

with this fleet. As with other fleets, accidents involving 

another car produced the most deployments. The objects 

struck are summarized 

in the following table. 

Ford Tempo Fleet Frontal lmpsct Deployment8 

Oblect atruck 

Car 

Truck and Van 

Deer 

Tree 

Pole 

Other 

As with the other fleets, frontal impacts were the predominant 

crash mode. Undercarriage deployments were not as 

predominant in the Ford Tempo fleet. The types of impacts 

producing deployments are summarized in the following 

table. 

fmpsct 

Frontal 

Side 

Rollover 

Undercarriage 

Ford Tempo Fleet 

% of deployments N = 153 

63 

19 

5 

4 

3 

6 

% of deploymentc N = 165 

The crash severity of the deployment accidents are shown 

in the following table. The number of known delta-V's is 

rather small. As in the other cases. the CRASH calculation 

has been corrected. 

92 

5 

2 

1 

?97

Ford T6mpo Fleet 

Craeh eeverlty Veloclly change % ol deplolments 

Minor 

Moderate 

Severe 

Less lhan 13 mph 

13-20 mph 

More than 20 mph 

Air bag deployment situations versus 

proposed test matrix 

In previous sections of the paper tests were proposed to 

establish a basis for crash sensor threshold behavior. The 

accident data files were $earched to determine how frequently 

these tests simulated situations in which the air bag 

was deployed in on-the-road accidents. The following table 

summarizes air bag deployments in the three air bag fleets 

that relate to the various test conditions proposed. Although 

in the above examination of fleet deployment crash severity 

was classified as "minor." "moderate," or "severe," the 

following table divides these crash deployments into two 

categories, 

"High Speed" and 

"Low 

Speed," in keeping 

with the proposed test categories outlined in previous sections 

of the paper. The "Low Speed" category was arbitrarily 

set as those deployments below 15 mph and "High 

Speed" as those above l5 mph. Since the velocity change 

was not known in every case, the percentage obtained from 

those known was distributed over the entire file to determine 

the number of occurrences. 

Test 

High Speed 

Low Speed 

Severe Rough Rd 

Fleet 

No. of Occurrencee 

Follce TemPo 

The Fleet test numbers reflect non-crash deployments' 

Fleet tests provide invaluable information on the air bag 

system other than causes for undesired deployments. Information 

gained on maintenance problems and procedures, 

system durability and reliability, design problems, and sys* 

tem acceptability are not reflected in the table above. 

The section on Severe Rough Road Exposures outlined 

four tests to simulate exposures outside normal rough road 

driving. The following table summarizes the fleet experience 

relating to deployment accidents resulting from these 

types of exposures. 

T6st 

Ditch 

Curb 

Rail 

Deer 

Gl,l 

6 

7 

11 

3 

GM 

g9 

56 

27 

14 

No, of Occurrence$ 

Pollce Tempo 

3 

2 

6 

12 

17 

11 

e 

10 

76 

14 

108 

99 

I 

1 

i 

7 

The real world accidents that these tests simulate, in 

addition to railroad crossings, include impact with boulders, 

culverts, buried objects, irregular terrain, embankments, 

and rocks. as well as ditches and curbs. The deer test simulates 

accidents involving related low mass animals as well 

as deer. 

Conclusions 

I. At this time, there appears to be a lack of consensus as 

to definition of the vehicle tests required for air bag system 

development. 

2. A review oftests that are presently being used has led to 

a definition of five distinct categories of vehicle tests which 

could be conducted. Specifics ofthe te$ts in these categories 

are given. 

3. A category defined as "Severe Rough Road 

Exposures" includes tests of fundamental importance with 

regard to crash sensor requirements. 

4. A review of procedures in present use indicates that 

results of dummy data and simple computer modeling may 

be inappropriate for analysis of crash sensor requirements in 

the regime of relatively small "delta v" crash occurTences. 

5. Real world air bag deployment experiences with three 

groups or "fleets" of cars was analyzed and the relative 

merit or value of the te$t was established. 

6. Additional definition and supplementation of these 

tests are required. However, a move to establish a protocol 

in the vehicle test community may be warranted. 

This paper pre$ents the views of the authors, and not 

necessarily those of the National Highway Traffic Safety 

Administration. 

References 

( I ) DOT HS-802 30 I Executive and Tabular Summary of 

Air Bag Field Experience, Volume l, No. 1, April 1977 

(2) DOT HS-803 416 Executive and Tabular Summary of 

Air Bag Field Experience, Volume 2, No. I, May 1978 

(3) DOT HS*805 062 Executive and Tabular Summary of 

Air Bag Field Experience, Volume 3, No. l, August 1979 

(4) Romeo, D. J., and Morris, J. B., Driver Air Bag Fleet 

Demonstration Program, A 24 Month Progress Report, 

Tenth Intemational ESV Conference, DOT HS-806 916, 

Oxford, England July 1985 

(5) Backaitis, S, H., and Roberts, J. V., Occupant lnjury 

Pattern$ in Crashes With Air Bag Equipped Government 

Sponsored Cars, SAE 812216 

(6) Wilson, R. A., General Motors Corporation, Crash 

Testing The General Motors Air Cushion, Fifth 

International ESV Conference, London, England, June 

1974 

(7) Wagner, R., Audi A.G., A 30 MPH Front/Rear Crash 

With Human Test Persons, SAE 791030 October 1979, 

Ttventy Third Stapp Car Crash Conference 

(8) Breed, D., and Costelli, V., Problems in Design and 

Engineering of Air Bag Systems, SAE 880724 February 

1989 

(9) Insurance Institute For Highway Safety Status Report

Vol.24, No.2, Februarv 25, 1989 

(10) Letter, dated June 18, 1984, Ceneral Motors to Michael 

M. Finkelstein, NHTSA, forwarding a Description of 

General Motors Tests for Evaluating Air Bag 

The Development of an Advanced Airbag Concept 

Lennart Johansson, Jan Billig, 

Hugo Mellander, 

Volvo Car Corporation 

Bernd Werner, Peter Hora, 

Bayern Chemie GmbH 

Abstract 

Interaction between the head of the driver and the steering 

wheel may cause facial injuries, especially in high speed 

accidents. 

Air cushion technology provides the means of distributing 

and reducing the inertia forces acting on the head and the 

face. 

This paper describes the development of an airbag system, 

including the electrical sensor which is integrated in 

the steering wheel, and it focuses on the problems of positioning 

the sensor in the steering wheel. 

Information from necessary sensor testing, both on rough 

roads and in crash conditions, is presented. 

The improvements which are achieved, in terms of reduced 

violence to the face, have also been assessed experimentally 

by using a test dummy with a load sensing face. 


ln connection with the increased usage of seat belts in 

Europe, the number of serious injuries in collisions has 

decreased. 

% 

.lO 

so 

20 

10 

Uilattsd 

Flgure 1. Inlury rates for belted and unbelted front $eat 

occupant$ at dlflcrent statod sccldent Bp€eds. 

(Ref. 1) 

However, despite the belt, there is still the problem of 

skull and facial injuries sustained by the driver when his 

head hits the steering wheel in high speed accidents. Volvo's 

accident statistics show that lOVo of drivers in head-on 

collisions receive skull and facial injuries. See figure 2. 

Performance in Circumstances Where Deployment is not 

Desired 

(l l) Accuracy and Sensitivity of CRASH, DOT HS-806 

152, March 1982 

,-. Forehead 

Nos€ 

- Mouth/Teeth 

- Jew 

FIgurc 2. Faclal Inlurl€s ars mostly to the forehead, no$e, 

mouth/teeth snd lsw' 

Since the total pattern of injuries decreases with the usage 

of belts, the remaining facial injuries constitute a greater 

proportion of the injuries suffered than previously. 

Normally, the injuries are relatively slight, but this type of 

co$metic injury often creates psychological problems for 

the victims, and rehabilitation can, in some cases, take 

years. 

Our goal was to optimise an airbag system for the belted 

driver and, by using a high level ofintegration, to produce a 

cost-effective and compact system. A bag in the steering 

wheel would protect against skull and facial injuries in 

head-on collisions and increase the chances of survival in 

high speed collisions. Trimming a bag for a restrained 

occupant meant that the bag could be made smaller than the 

US-system (for unrestrained occupants). 

Eurobag U$.bag 

Flgure 3. Ths plctur€E show ths dlfference In sizB b€tw€€n a 

Egg lo*r a belted drlver (Eurobag) and a bag complylng wlth 

FMVSS zOE. 

The results after computer simulation and sled tests 

showed that a bag with a volume of approximately 35 litres 

(diameter 550 mm) without internal straps gave the best 

results. 

299

A small bag has advantages with regard to out-ofposition-exposure. 

omission of the strap$ meant that the 

bag could be made relatively simple in design. If a bag 

material with a coating which can be welded is used, a cheap 

method of manufacture can be developed whereby two discs 

(one of which has a pre-welded disc with the necessary 

attachment holes for the inflator) are welded together. (See 

figure 4). 

The advantages of this are, among other things, that the 

sensor can be fitted directly onto the inflator and, because of 

the fact that all system components are in the steering wheel, 

the system could be installed as an aftermarket item. 

Sensor development 

The Eurobag sensor, fitted in the steering wheel, is basically 

an electronic software controlled microprocessor 

system. The trigger algorithm follows 

ilffi 

"standard" accelera- 

Figure 4. Three parts welded together to a bsg. 

Since the inflator only needed to generate gas for a 35 litre 

bag, its dimensions could be reduced as shown in figure 5. 

Standrrd 

lnflator 

Ncw Inllator 

Fioure 5. A Eurobao Inflator (-157o smaller In dlamster and 

-2b* lower ln helght) compare'd with the conventlonal inflator. 

This gives not only a reduction in weight, but the new 

inflator also occupies less space in the steering wheel. 

System layout 

The whole system is located in the steering wheel, i.e. the 

cover, the bag, the inflator, the sensor, including diagnostic 

and reverse energy units, the connector coil and the warning 

lamp. (Figure 6) 

Figure 6. Eurobag concept. 

@F 

ffi 

WF*lg lrrltF 

tion signals within electronic airbag systems, i.e. above a 

certain acceleration threshold and when a certain velocity 

change has occurred, the power switch is triggered to ignite 

the squib. The threshold depends on the crash behaviour of 

the actual car. 

The development of the sensor was directed towards Volvo740[60. 

The 760 model also contains a tiltable steering 

wheel (tiltable +7'), and this must be considered in the 

development of the rlensor. (figure 7) 

Figure 7. The 760 steerlng wheelcan b€ tlltsd to thres posltione 

(+7o 

from the normal positlon). 

Flgure 8. Steering wheel wlth 5 accelerometers In th€ dlflsrent 

dlrectlons. 

Development of software 

The sensor development began with the accumulation of 

measuring data. With the aid of a measuring unit with five 

accelerometers fitted in the steering wheel, signals from a

number of different collisions were registered, to be analysed 

later. 

The accelerometers were placed in the X, Y and Z axes 

and f45o from X. (figure 8) 

Tests were carried out at different speeds from 4 mph to 

35 mph and were of types 0",30",90", 180o, pole and 

underride. (figure 9) 

oo ior lo. ttoo 

r t 

HHE\r 

r|1l 

HEEEE 

F'. uirf--'-- Ii 

ftT. _h E 

Flgure 9. Dlfferent types of colllalon Bltuallons, where measurements 

ol acceleratlon In the steerlng wheel were reglstered. 

Since the sensor should not trigger during different types 

of rough-road driving, measurements were carried out on 

that type of surface as a reference. Examples of "nontriggering 

situations" are curbs, road depressions and different 

types of bumps. (figure l0) 

Ost r|'m rrff, dffistr.r 

Gcdqrbad 

Flgure 10. Examples of rough-road sltuatlons. 

r Curb-stone testing was performed against a 100 

mm high curb at 35 mph and a 150 mm high curb 

at 30 mph. After such curb-stone driving, the car 

is damaged, but the driver unharmed. 

r Road depressions at speeds up to 55 mph and with 

maximum load in the car. 

r Driving on pot-holes and Belgian pavd with 

wheels both locked and rolling. 

r Test driving was also performed with unbalanced 

wheels at different speeds. 

The accumulation of signals was also supplemented by 

measurements made when hitting the steering wheel with 

different objects, e.g. a hammer and a fist. All tests were 

done with the steering wheel tilted at the three different 

angles. The measured acceleration levels from the different 

tests wsre used as input data to a computer simulation program, 

which simulates the sensor's signal processing. The 

calculated times were compared with the requirement according 

to the 125 mm criteria. 

The trigger time that the total $y$tem must meet wa$ 

determined by the conventional 125 mm requirement, i.e. 

the time required for an unbraked mass to move 125 mm in a 

crash. (figurE ll). This is to avoid interaction between a 

deploying bag and the occupant. 

Flgure 11. Unbraked msBB (the "worst case"" of an unrastralnad 

drlver) msy move, due lo Inertla, a marlmum of 125 mm 

before the bag le fully lnflated. 

Further analysis work has shown that when belts are used, 

a longer time delay can be accepted. But for the development 

of this sensor, the 125 mm requirement was adhered 

to, 

After a number of development stages, a signal processing 

algorithm was obtained which fulfilled the set triggering 

criteria, without inadvenent triggering during rough-road 

simulation. 

The algorithm is subdivided into four path$. Path I integrates 

the acceleration signal once to a delta velocity proportional 

voltage. Path 2 integrates the acceleration signal 

twice to a voltage that is proportional to the displacement of 

a free mass. Path 3 integrates the negative acceleration once 

to a velocity proportional signal and is used to recognize the 

direction of the crash. Path 4 compares the acceleration with 

a certain threshold and controls the connection ofthe signals 

evaluated in Path I to Path 3. In the logic block, the signals 

are connected together depending on certain thresholds and 

timing functions. 

In general it can be said that in a normal crash, the trigger 

signal for the squib is generated by Path l, if a certain 

threshold is exceeded. Path 2 to Path 4 are used to distinguish 

between a crash situation when a trigger signal should 

be generated and, for example, an acceleration signal generated 

in rough road conditions where no triggering is 

required. 

This algorithm was implemented in the hardware solution 

which was developed simultaneously. (figure l2) 

The programme also includes a diagnostic section which 

continuously controls the main component$ of the $y$tem. 

Also included is a pan of the progrflmme which calibrates 

the acceleration Fansducer during the final test in the fac- 

301

Figure 12. Schematlc blockdlagram showing the slgnsl conditloning 

parts of the sensor. 

tory via software and compensates the acceleration transducer 

sensitivity over variou$ temperatures. If a failure is 

detected a failure lamp lights up. 

Development of hardware 

With the help of simulation and practical tests it was 

found that an accelerometer in the direction of the steering 

shaft was sufficient to give acceleration signals which could 

be used. The hardware was built up around a microprocessor 

with the necesssary protective circuits and watchdog. 

The unit was fitted with a capacitor for a reserve energy 

supply in case the ordinary voltage feed is cut off in a 

collision. The voltage supply to the sen$or is fed through a 

contact reel of clockspring type since the slip rings for the 

horn were found unreliable. The tests performed with slip 

rings showed that in cold conditions condensation could 

freeze on the tracks and cause a voltage cut for up to l0 

minutes. Consequently the voltage feed via the slip rings is 

used for redundancy. 

The hardware concept, shown in the attached block diagram, 

is structured in three major parts: acceleration sensor 

with amplifier; microcontroller and memory; power switch 

and power supply. 

Acceleration sensor with amplifier 

Piezoelectric acceleration sensor with high shock resis* 

tance; amplifier as impedance converter and filter; diagnostic 

unit for acceleration sensor. 

Microcontroller and memory 

Triggering of power switch for squib ignition depending 

on implemented software controlled trigger algorithm; controlling 

of all diagnostic functions; software stored in the 

memory. 

Power switch and power supply unit 

Low impedance power switch for squib ignition and failure 

lamp control with integrated diagnostic functions; voltage 

regulator for conditioning of the car supply voltage; 

overvoltage protection and switch for reserve energy supply 

302 

on board. If faults occur in any of the components, the 

diagnostic unit gives a signal to a lamp located in the steering 

wheel cover. 

i 

ir'; r 

; l E f 

l L - t 

I 

Flgure 13. Elock dlagram showlng the hardware concept. 

The final prototypes were verified in a large number of 

sled tests and full-scale collision tests. The system fulfilled 

all criteria. Furthermore, a number of different rough-road 

tests were performed, and these proved positive, i.e. the 

sensor did not trigger. 

Crash test results 

Full-scale collisions have been performed from 0" and 

30", pole and under-ride at speeds of 30, 35 and 40 mph. The 

Eurobag system was found to function well as a complement 

to the seat belt in all the above situations. The sensor 

located in the steering wheel fulfils the specified requirements 

and functions with the steering wheel tilted in different 

positions. 

The tests showed that the Eurobag concept reduced Hic 

by approximately 25*30qo, see figure 14. 

Hlc 

800 

40 mph 3s mph 30 mph 

Sled t€st Barrler tsst Barrlor tcat 

Figure 14. Typlcal valuee from the evaluation tests. 

Another advantage with the Eurobag was that it reduced 

the forward movement of the head by approximately 20o/o. 

The forces on the torso belt were reduced by approximately 

l0o/o, see figure 15. 

Without a Eurobag there is a variation in measured Hic 

values due to different head impact location in the tests with 

I 

Sertb.tt ontY 

H zg% @ tu'o"t

KN 

I 

6 

I a.rbilt only 

$ euroleg 

Flgure 15. 1096 raductlon of torao belt forces wlth Eurobag. 

the steering wheel tilted at different angles. However, with 

the Eurobag concept, a much more consistent behaviour 

was observed. with lower Hic values. 

fiplcal readlngr from *led tetilng 

I Whh Eurohg 

tr w|thourEurob|g 

Flgure 16. Gomparison of faclal pr€saure wlth and wlthout Eurobag. 

The bar dlagram $howt the hlghoet achleved preesure ln 

any of the meaeurlng plEles wlth the samo designatlon. 

The Passive Restraint Approach 

Robert H. Munson, 

Automotive Safety Office, Ford Motor 

Company, United States 

Abstract 

This paper presents an overview of Ford's experience 

with "passive restraint" systems. It seeks to update the 

conference as to the available field experience with passive 

restraint systems being installed in Ford Motor Company 

vehicles. The present United States occupant crash 

protection standard requires a phase-in ofpassive restraints. 

Passive restraints can be automatic belts or air bags. An 

overview will be presented of Ford's passive belt and 

supplemental driver and passenger air bag experience, 

including: 

In order to confirm the reduction offacial injuries, a loadsensitive 

face dummy was used in the different collision 

situations. With the Eurobag concept, the facial pressure is 

reduced by an average of80Vo, (figure 16). 

Conclusions 

This project has shown that a small bag, optimized for 

belted drivers is an excellent complement to the seat belt. It 

protects the head against impact against the steering wheel 

and thereby reduces the risk of head and facial injuries in 

head-on collisions. 

The possibility to position the sensor in the steering 

wheel was successfully demonstrated. By using an 

electronic sensor, a sufficiently complex conditioning of the 

signal could be made to meet the non-triggering and 

triggering requirements and also to diagnose and indicate 

faults in the system. There is now the possibility to choose 

between a body-mounted crash sensor or a sensor located 

directly on the safety system. 

The advantages and disadvantages of these two 

approaches have to be further investigated and they will 

probably vary for different categories of cars. 

References 

(l) H Norin, G Karlsson, J Komer; "Seat Belt Usage in 

Sweden and its Injury Reducing Effect". SAE 840194. 

(2) S Koyabashi, K Honda, K Shitanoki: "R.eliability 

Considerations in the Design of an Airbag System". ESV 

1987. 

(3) C F Kirchoff et al; "Advanced Concepts for Driver 

Air Cushion Systems". ESV 1985. 

(4) D S Breed, V Castellir "Problems 

in Design and 

Engineering of Airbag Systems". SAE 880724. 

r The Decision Making Process-Supplemental 

Air Bag vs. Motorized Automatic Belts vs. 

Manual Automatic Belts 

r Air Bag and Passive Belt System Overview 

r The Effects of Mandatory Use Laws on Passive 

Restraint Decisions 

r Passive Restraint System Field Experience 

r The Continental Passenger Side Supplemental 

Air Bag 


This paper will focus on Ford's passive restraint 

experience since the lOth ESV conference, where another 

paper entitled "Supplemental Driver Airbag System-Ford 

Motor Company Tempo and Topaz Vehicles" was 

presented. 

303

Ford's passive restraint program consists of the air bag 

supplemental restraint system in combination with an active 

3-point safety belt on certain vehicles and motorized 

automatic shoulder belts with active lap belts for driver and 

right front passenger on other vehicles. It is estimated that 

through late April 1989, over 40,000 Model Year 1985 

through 1989 Ford Tempo and Mercury Topaz vehicles with 

the Supplemental Driver Air Bag Restraint system have 

been built and sold (see figure l). These vehicles have 

experienced more than three-quarters of a billion miles of 

travel. with an estimated 620 accidents that were severe 

enough to cau$e the air bag to deploy. Ford and NHTSA's 

monitoring programs have reports of 233 of these 

deployments. The results are summarized in figure 2. 

FLEETS 

a 

a 

FEOERAL 

OTHER 

HETAIL 

TOTAL 

r (As oF 4t20l8s) 

TEMPO I TOPPZ, SALES * 

l$85 t9E8 1987 1S88 1989 

s{xto 

2400 3700 

100 

1500 

4800 

5400 

3100 

8000 

2600 

3500 

TOTAL 

6500 

16,5{t0 

17,000 

40,000 

Figure 1. Sales of Tempo and Topaz equipped with the drlver 

air bag supplemental r€stlsint $y$tem. 

TEMPO / TOPAZ AIR BAG 

FLEET EXPERIENCE (AS OF 4-20-8e) 

ESTIMATED 

MILES DRIVEN (lulLLlONS) 

ACCIDENTS 

DEPLOYMENTS 

REPORTED 

DEPLOYMENTS 

BELT USE 

INJURIES: MODERATE 

sERrous 

FATALTTTES (DRTVER) 

Flgure 2. Tempoffopaz supplemental all bag fleet experlence. 

The 1989 model year Lincoln Continental is equipped 

with a standard driver and right front passenger Air Bag 

Supplemental Restraint System. For the 1990 Model Year, 

Ford plans to equip nine car lines with a driver-side Air Bag 

Supplemental Restraint System. 

The Decisionmaking Process 

800 

5200 

620 

23tt 

87% 

16 

3 

In July 1984, NHTSA amended FMVSS 208 to require 

restraints for both the driver and right front 

3 

passenger on U.S. pa$senger cars beginning, on a phasein 

basis, in 1987. Further, NHTSA required installation 

of passive restraints on a// U.S. passenger cars manufactured 

on or after September I, 1989 (i.e., the 1990 model 

year). The following summarizes a few of the many factors 

that Ford considered regarding passive restraint 

implementation. 

Ford began its evaluation program by prioritizing the 

anticipated practical safety benefits of various passive 

restraint alternatives on the basis of overall eff'ectiveness, 

anticipated usage and customer acceptance. The two-point 

motorized passive shoulder belt system was viewed very 

favorably, although more costly than non*motorized belt 

systems, based on anticipated usage and resultant 

effectiveness. Market research confirmed customer 

preference for motorized belts, but they were not considered 

practicable fbr larger cars with three front $eat positions. 

NHTSA, after extensive investigation, had concluded 

that air bags in combination with three-point active belts 

offered the highest effectiveness in terms of reducing both 

fatalities and injuries (see figure 3). However, this 

conclusion assumed that three-point active belt use was 100 

percent. Before an air bag supplemental restraint systembased 

program could be pursued feasibility and 

practicability had to be established and there had to be 

assurance of greatly increased three-point safety belt usage. 

Ford believed then as now that air bag practicability u,ould 

be established and the vast majority of the U.S. population 

would be covered by effective mandatory seat belt use laws. 

On that basis, Ford continued pursuit of an aggressive air 

bag implementation program that continues to the present 

time. 

IIS'URY LEVEL 

NHTSA ESTIMATED 

RESTRAIHT EFFECTIVENESS RANGES 

I/ 

E$TIIIATED RE$TRAINT EFFECTIVET{ES$ RAI{GES 

I-AP AI{D 

SHOULDER BELT 

AUTOilANC 

EELT 

AIR BAO 

(oNLn 

AIB BAS 

AI{D I-AP/ 

SIIOULOEf, 

EFLT 

FATAL 40-50* 35 - 5{lt6 e0-ttot 'ff - SSti 

ats 2-6 45- 55% 40 "55t6 It -t5tg 60-00t9 

-!Jf I{HIEA DOCI(ET 74-l4i NOTICE 80, *t'OCUFAHT CF EH FROTECTnili Fll{ L BULP. 

Flgure 3. NHTSA efiectiveness estimates. 

After thorough review ofthe technical issues and supply 

base requirements necessary for supplemental air bag 

programs, it was determined that programs calling for 

supplemental air bags on some vehicle lines and motorized 

belts on other vehicles would be the only practicable means 

of meeting the 1990 effective date. To allow for increased 

driver-side supplemental air bag availability, Ford 

petitioned NHTSA for extension of the 1.0 credit to allow 

for driver-side air bag and 3-point active belts after the I 989 

model year. 

In its petition to NHTSA, Ford stated that if the petition 

were granted, it would, in all likelihood;

lnstall supplemental driver air bags on a majority 

of its NAAO-designed cars. 

Install passenger-side air bags in some 1990 

model year cars as resolution of technical issues 

and supply base constraints allowed. 

These product plans have produced real and tangible 

effects. In the 1987 model year, Ford offered its first 

motorized two-point passive belt system as standard 

equipment on the Ford Escort and Mercury Lynx. In 1988, 

use of this system was extended to the Ford Tempo and 

Mercury Topaz, while continuing to offer a driver side 

supplemental air bag as an option. The all new 1989 Ford 

Thunderbird and Mercury Cougar models are equipped 

with the motorized belt system while the 1989 Continental 

is the first domestic U.S. car to offer both driver and right 

front passenger side supplemental air bags as standard 

equipment. 

For the 1990 model year, Ford will provide supplemental 

driver-side air bags on nine of its car lines; some of which, 

like Continental, will also offer standard supplemental 

passenger side air bags. We anticipate approximately one 

million 1990 Ford passenger cars will be sold with 

supplemental driver or driver/right front passenger air bags 

as standard equipment; we believe in the 1990 model year, 

Ford will produce and sell more cars equipped with standard 

supplemental air bags than any other manufacturer in the 

world (see figure 4). 

FORD f,OIOH COTPAI{Y 

FASSIVE FESTRAINT AVAILABILITY 

;gSry _4!!!L Afm|ltE Hwffi 

lttt ffi ruoEil wEr 

ilEffi ilrs-Halilolol r-ff 

r[r ffi ffiE[ wEr 

ltsffi ffiE[ WBI 

ffi MflSEIO' r-|w 

laaa ffi IffiE[ wat 

ffi IffiEN WEI 

r-tffi ffiE[ wET 

EFffi ffrH-Hllrulol r-ffi 

ffiEflil oi|Y[ffiAnrm l-ff 

itm il[i-filf|'l.fliilxrlmlllFmff tHtlf,l 

(EffiffiffiIfrtM 

-DEOI$EY*I41ffffi 

ffiilrus|.5rEsEreImrffi 

Fffiffi&1ffi 

Flgure 4. Ford passlve restralnt avallablllty. 

Air Bag and Passive Belt System 

Overview 

As shown in figure 5, the supplemental air bag system has 

four basic subsystems, the sensors, the diagnostic module, 

the electrical system and the air bag module or modules. A 

detailed description of these subsystems is included in 

attachment l. The Ford passive belt system, consisting of a 

motorized automatic shoulder belt, active lap belt and knee 

bolsters also is described in detail in attachment l. 

Mandatory Use Laws 

Unlike many other countries, the United States does not 

have a national law or regulation governing safety belt 

usage by the occupants of motor vehicles. It has been left to 

1 989 LINCOLN CONTINENTAL 

SUPPLEMENTAL AIR BAG RESTRAINT SYSTEM (SRS) 

Flgure 5. Supplemental alr bag $ystem. 

each of the fifty states to enact legislation on requiring 

safety belt u$age. Currently, thirty-three states plus the 

District of Columbia (see figure 6) have laws requiring 

safety belt usage at least for front seflt occupants. In 

addition, all fifty slates have enacted laws requiring the use 

ofchild restraints for infants and young children. Ford has 

been actively involved in supporting the passage of these 

laws and has encouraged safety belt usage through public 

support of safety belt education for our employees, our 

customers and the public at large. 

MANDATORY 

SEAT BELT USE LAWS 

Flgure 6. Mandatory use laws. 

Increased safety belt usage remains key to passive 

restraint system overall effectiveness. Prior to the 

enactment of the first state safety belt use laws in the mideighties, 

safety belt usage in the U.S. was only about 12 

percent. Today, it has reached 46 percent nationally and 

over 5l percent in $tates with mandatory use laws. Passive 

belts, either motorized or non-motorized, result in 

substantially higher usage rates than active belts, leading to 

greater net effectiveness in helping reduce fatalities and 

injuries on American highways. As active safety belt usage 

rates continue to increase, the combination of an active 

three point safety belt and supplemental air bags for the 

305

driver and right front passenger has the potential to provide 

eYen greater net effectiveness. 

Thus, it is very important that we in the safety community 

continue to strive for increased safety belt usage so the 

maximum benefit from the "passive" restraint systems now 

being installed in cars may be obtained. We have known for 

many years that getting vehicle occupants to buckle up is the 

single most important action for improving highway safety. 

It is clear that buckling up is still the most important action 

even in vehicles with supplemerrtal air bags or a 2-point 

motorized shoulder belt system. 

Air Bag and Passive Belt System Field 

Experience 

Ford vehicles with supplemental air bags have now 

experienced over three-quarters of a billion miles of travel. 

To date, I am pleased to report, the system has functioned as 

designed in every collision that has been reported to us. In 

addition, we are unaware of any air bag equipped car in 

service experiencing an inadvertent inflation of the air bag 

or bags. The vast majority of the mileage and reported 

collisions which caused the air bag to inflate have occurred 

in the Ford Tempo and Mercury Topaz vehicles, which have 

been in production since the 1985 Model Year. 

There are now also a small number of collisions which 

have resulted in the inflation of the driver and passenger 

supplemental air bag system in 1989 model year Lincoln 

Continentals. These data are summarized in figure 7. I 

would also like to emphasize two other points about the 

data. Active safety belt usage in the Ford Tempo and 

Mercury Topaz vehicles was reported as over 85 percent. 

This undoubtedly played a significant role in the 

effectiveness of the supplemental air bag. Many of the 

vehicles are owned by government and private fleets, which 

require or strongly encourage safety belt usage. Moreover, 

it is interesting to note that belted drivers have yet to suffer a 

non-fatal serious injury. All serious injuries reported were 

suffered by unbelted vehicle drivers. 

There have been two driver fatalities in these vehicles 

that resulted from head-on crashes with trucks (one belted). 

As can be seen from figures 8 and 9, neither vehicle had an 

intact passenger compartment following the crash and 

LINCOLN CONTINENTAL AIR BAG 

SALES (AS OF +20) 

DEPLOYMENTS (A$ OF 4.20) 

PASSENGERS 

INJURIES 

I'TODERATE 

SEHIOUS 

FATALITIES 

27,300 

37 

4 

DEPLOYMENT NONCE INFORMATION I9 SKETCTIY 

Flgure 7. Llncoln Contlnental alr bag supPlementsl restraint 

exp€rience, 

306 

neither incident was judged to be survivable by experts at 

the scene, regardless ofrestraint system design. In addition, 

our search of the FARS, orFatal Accident Repofiing Sy$tem 

data for the supplemental air bagequipped vehicles has 

provided information on two additional fatalities in these 

vehicles. The other two fatal accidents were side impact 

collisions, for which air bags are not designed to activate. 

Figure E. Tsmpa air bag fatal accldent No. 1. 

One resulted in a driver fatality and one resulted in a 

passenger fatality. Although the numbers are quite small, 

we currently have equal numbers of fatalities from frontal 

accidents and from non-frontal accidents. 

Flgure 9. Tempo air bag latal accident No.2. 

The data available from our passive restraint-equipped 

vehicles is too limited to allow us to evaluate system 

effectiveness quantitatively. However, detailed studies of 

both supplemental air bag and passive belt vehicles are 

ongoing. One key area of research underway is to compare 

safety belt usage rates in 1988 and 1989 model year 

Continentals. We are concerned that safety belt usage rates 

be maintained at their current high levels, or increased. Our 

studies of front seat occupants of new Ford and LincoltV 

Mercury vehicles including non-air bag 1988 Lincoln 

Continentals show safety belts usage of approximately sixty 

percent. We will be measuring usage rates in the air bagequipped 

1989 Continentals later this year in the same

locations to evaluate if supplemental air bags affect safety 

belt use (see figure l0). We are continuing to study usage 

rates and will study crash statistics to develop objective 

measures of our passive restraint system field performance. 

We look forward to reporting the results of these studies at a 

future date. 

BELT USAGE SURVEY PLANS 

. CONTINENTAL - wlTH / WITHOUT AIR BAGS, 

MODEL YEAR 1988 veruur 19Bg 

r BELT USE lH CURRENT CARS AND LIGHT TRUCK$ 

OUARTER 

I 

il 

ill 

IV 

TENTATIVE SCHEDULE 

vEHtcLE UI{ES 

TEMPO / TOPAZ, F-SERIES, BRONCO 

MUSTANG, PROBE, AEROSTAR 

CONTINEI{TAL, TAURUS / SABLE, 

ECONOLINE 

CROWN VIC / GRAND MARqU|s, TBIRD, 

RANGER / BROI{CO 

Flgure 10. Ford bslt uaage aurvey plana. 

Continental Passenger Side 

Supplemental Air Bag 

The 1989 Lincoln Continental has the first driver and 

right front passenger air bag supplemental restaint system 

introduced as standard equipment in the 1980's by a United 

States automaker. These vehicles have now been available 

to the public since last fall. Over 27,300 have been sold as of 

late April 1989. There have been 37 reported accidents of 

sufficient severity to have deployed the air bags. We are 

unaware ofany fatalities or serious injuries ofeither drivers 

or passengers in these vehicles. 

Summary 

We are clearly on the threshold of a new era in restraint 

systems. Ford will go from producing 70,000 driver or 

driver/right front passenger air bag-equipped vehicles in the 

I 989 model year to one million in the 1990 model year. The 

remainder of our passenger cars will have automatic 

shoulder belt restraint systems. The magnitude of these 

changes is likely to make the next few months a very 

challenging time for our assembly operations, and for the 

many involved supplier$. Our initial assessment of 

customer acceptance and observed usage rates in vehicles 

equipped with these systems are encouraging. Detailed 

studies are underway, and we look forward to seeing and 

reporting the results of passive restraint technology on the 

safety of automotive travel. 

Attachment L-Air Bag and Passive 

Belt System 0verview 

Sensors (Figure ll) 

The sensors are used to determine the onset of a crash 

severe enough to require the inflation of the air bag 

Flgure 1 1. Alr bag aensor. 

supplemental restraints and are of ball and tube design. 

They utilize a biasing magnet to hold the ball firmly at the 

aft end of the tube. The activation of the $ensor requires a 

deceleration of sufficient magnitude and duration to 

overcome the bias of the magnet and sustain the travel of the 

ball down the tube for several milliseconds, damped by the 

column of air in the tube. to close the contacts at the forward 

end ofthe tube. The ball and the contacts are gold plated for 

reliability. The sensor is designed to respond to impacts up 

to 30" left and right ofthe longitudinal axis ofthe vehicle. A 

typical system will have five sensors in the front of the 

vehicle. Three are "crash sensors" and two are "safing 

sensors". At least one crash $ensor and one safing sensor 

must close simultaneously to activate the air bag system and 

inflate the air bags. 

Diagnostic Module 

The air bag system diagnostic module continuously monitors 

the system readiness and verifies for the vehicle operator 

that it is ready to function whenever the vehicle is started. 

This is done through the readiness indicator lamp on the 

instrument panel, which illuminates for about six seconds 

each time the vehicle is started. Each inflator and sensor 

circuit is checked and, should a fault be found, the lamp will 

flash to alert the vehicle operator that service is required. 

The flashes are coded to indicate the location and nature of 

the service required, and the "fault codes" are provided in 

the service manual so the service technician knows which 

area of the system requires service. 

Electrical System (Figure 12) 

The electrical system con$i$t$ of a power circuit directly 

from the battery, and wiring to each crash sensor, to the 

diagnostic module, to the readiness indicator lamp, from the 

stan circuit, and to each air bag module. The driver side 

circuitry includes a "clock spring" mechanism to transmit 

the electrical signal to the $teering wheel-mounted driver 

module with a high degree of reliability. Current systems 

307

wtRtl{G 

HAEHESS 

BELT WAFI{ING LIGHT 

AHD CHII{E 

TEMPO/TOPAZ AIFBAG 

ELECTRICAL SYSTEM 

cFASH 8EI{$QXS 

"'lEBtil'" 

Flgure 12. Alr bag electrlcal system. 

also include wiring to a tone generator to provide indication 

of malfunction of the readiness indicator lamp circuit. 

Air Bag Modules (Figure 13) 

Flgure 13. Air bag moduleE. 

The air bag modules, either driver-side or passenger-side 

consist of the same basic components, although they are of 

different size and shape, and operate in slightly different 

manners. Each has a nylon "bag" although they are different 

in shape, size and venting. The driver-side is smaller, 

approximately two cubic feet. The nylon is neoprene coated 

and venting is by way of two or more vents on the steering 

wheel side of the bag. The bag also has internal $traps to 

help shape it into a flat cushion shape rather than a sphere. 

The passenger side "bag" is uncoated nylon, somewhat 

larger, without straps, and vented through the fabric porosity 

and through the aspiration ports between the inflator 

and the bag. Its shape is more spherical, but flattened somewhat 

against the windshield. 

Each module also has an inflator assembly, consisting of 

a steel canister containing the gas generant, a sodium azide 

compound which combusts and releases nitrogen ga$ to 

inflate the bag. The inflator is activated by an electrical 

signal which causes an igniter to initiate combustion of the 

gas generant. Filtering media is used to cool the gases and 

contain the combustion process within the inflator canister. 

308 

Finally, each module has a cover which protects the bag, 

but is designed to open when inflation is required. The 

driver modulelt use a molded design which tears at molded 

in seams as inflation commences. The passenger side may 

use a similar design, or a series of small interlocking plastic 

tabs to $ecure the two halves ofthe cover together, depending 

on the geometry of the cover. 

Other System Components (Figure 14) 

WAHNING 

This restraint module cannot be 

repalred. Use Ford publirhed diagnostic 

lnetructlons to determine if the unit is 

delective. lf defective, replace and dispoee 

of the entire unit ae directed in inetructionr. 

Und€r no circum$tancee Ehould diagnosis be \ 

performed using electrically powered test equipment 

or problng devlces. Tamperlng or mlshandling 

can reeult in personal injury. For spsciel 

handllnE handllng lnetructlons, refer to the Ford AlrbaE Alrbag 

Shop Manual. eogB.o4ogoso-AA 

CONTAIN$ $ODIUM AZIDE 

AND POTASSIUM NITRATE CON. 

TENTS ARE POISONOUS AND EX. 

TFIEMELY FLAMMABLE CONTACT WITH 

ACID. WATEH, OB HEAVY METALS MAY PRQ. 

DUCE HAHMFUL AND IRRITATING GASES OR 

EXPLOSIVE COM. 

POUNDS DO NOT 

DAI{GEF 

DISMANTLE. IN. 

CINEHATE, OR 

BRING INTO 

CONTACT WITH 

ETECTFICITY OR 

STOFE AT ]EM. 

PERATUHES 

EXCEEOI{G 2co"F. 

FIRST AID: lF 

CONTFNTS AHE 

OF CHILDREH 

SWALLOWED, IN. 

OUCE VOMITING - FOR EYE CQNTACT, FLUSH 

EYES WITH WATER FOR 15 MINUTES - IF 

GASES FROM ACID OR WATER CONTACT 

ARE INHALED, SEEK FFIESH AIFI*IN 

EVERY CASE GET PROMPT 

MEOICAL ATTENTIOT. 

THIS VEHICLE IS EOUIPPED WITH A SUPPLE- 

MENTAL DRIVER AIRBAG SYSTEM. TAMPER. 

ING WITH OR DISCONNECTING THE AIRBAG 

SYSTEM WIRING COULD DEPLOY THE BAG 

OR RENDER THE SYSTEM INOPERATIVE. 

WHICH MAY RESULT IN HUMAN INJURY. 

Flgure 14. Air bag system labellng. 

The supplemental air bag restraint system is designed to 

supplement the conventional lap/shoulder safety belt by 

providing additional protection for the face, chest and head 

in a frontal crash that is equal to or more severe than an 

impact at ?8 mph into a parked car of similar size, or a fixed 

object at l4 mph. Thus, there are many types of collisions 

where it is not designed to activate, and probably will not. 

Examples are side, rear and rollover types of accidents. The

Flgure 15. Automatlc shoulder belt passlve restralnt system. 

safety belt must be wom to help protect in these other types 

of accidents. We also believe that the safety belt works with 

the air bag in frontal collisions to help position and restrain 

the occupant for maximum benefit from the combination of 

belt and bag. However, in the event someone does not buckle 

up, knee bolster surfaces are provided to help control 

submarining and provide maximum benefit from the supplemental 

air bag. 

An additional area that should be rnentioned is labeling. 

Each vehicle equipped with the air bag supplemenral restraint 

system is labeled, from the vin plate "air bag" nomenclature, 

to the individual warning labels on the modules 

themselves. These labels are designed to aid the people who 

will use, service and, eventually, recycle the vehicle in 

accomplishing their contact with the vehicle without unreasonable 

risk ofinjury or accident. 

Passive Belt System (Figure 15) 

The Ford motorized automatic belt system consists of a 

motorized automatic shoulder belt with an inboard mounted 

Crash Simulation Methods for Vehicle Development at Nissan 

Tatsuya Futamata, Hiroyuki Okuyama, 

Nobuhiko Takahashi, 

Nissan Motor Co., Ltd. 

Abstract 

For vehicle frontal crash simulation, Nissan has been 

using a relatively simple lumped mass-spring simulation 

combined with an in-house frame crash program CRAFT. 

However, it has been recognized that this procedure does 

emergency locking retractor. The motors which position the 

belt are activated by the position of the door switch and the 

ignition switch. The belt will move forward to allow ingress 

and egress whenever the adjacent door is opened. The belt 

will move back to the rear, or restrained, position whenever 

the adjacent door is closed and the ignition is on. However, 

in the event of a moderate to severe accident, the inertia 

safety shutoff switch which disables the electric fuel injection 

system pump also shuts off the power to the safety belt 

system motors. The shoulder belts are designed to work 

with the active lap belt to provide maximum protection. In 

the event the lap belt is not fa$tened, a knee bolster surface is 

provided to help control submarining. Two approaches have 

been used to provide for emergency release of the automatic 

shoulder belt. Some vehicles have an emergency release 

lever mounted on the console adjacent to the retractor. This 

lever allows the belt to spool out eyen if the retractor remains 

locked after an accident. Other vehicles have a buckle 

for emergency release located at the outboard end on the 

moving carriage. 

not always have capability to simulate frontal crash responses 

for vehicle structural design changes directly. Thus, 

as an alternative to this conventional method, a commercial 

nonlinear dynamic finite element program PAM-CRASH 

have introduced and researched to simulate vehicle crash 

characteristics using that program run on a super computer. 

A step-by-step approach was taken to simulate a test 

vehicle crash response staning with its front side rail crash 

simulation, and finally, the simulated deceleration of the 

309

test vehicle was successfully conelated with its actual crash 

test result. 

The major issues to be solved in our further study will be 

model simplification and application of this simulation 

method to various types of test vehicles. 


The techniques used in carrying out large deformation 

vehicle crash analysis have recently been improved by the 

use of super*computers and commercial nonlinear finite 

element analysis programs which have been suitably 

developed specifically for these computers. 

In parallel with these improvements, a number of papers 

have appeared mainly conceming the modeling methods 

and the crash characteristics obtained in vehicle high-speed 

frontal crash simulations. 

There has been no report of good vehicle deceleration 

correlation between crash tests and finite element 

simulations by which the dummy injury criteria can be 

accurately determined. The principal reasons for this seem 

to be (l) computational cost (10 hours'to 20 hours'CPU 

time even with a super-computer), (2) software problems, 

and (3) the lack of a reasonable modeling technique. 

Computational cost may be significantly reduced in the 

near future by improving hardware capabilities in 

conjunction with optimizing software algorithms. Software 

problems include the lack of appropriate finite element 

types for modeling some vehicle components, as some 

papers mentioned. 

Modeling techniques become an issue when viewed from 

the fact that some paPers show that the simulated vehicle 

deceleration curves do not simulate the tendencies seen in 

test results especially with regard to the latter part of the 

deceleration's time history. 

To overcome these deficiencies, we have introduced a 

commercial nonlinear dynamic finite element analysis 

program called PAM-CRASH. The program simulates 

vehicle crash analysis and has been successful in 

reproducing the results of the test vehicles' deceleration 

characteristics. 

This paper, accordingly, discusses the modeling 

techniques acquired in our research which is necessary for 

simulating vehicle frontal crash characteristics. 

Conventional Method 

Nissan has thus far been usiltg "spring-mass simulation" 

combined with an in-house program we call CRAFT as the 

method for simulating vehicle component crash 

characteristics. Figure I shows its simulation concept. 

In CRAFT, a frame like a front side rail is modeled as a 

structure consisting of rigid beams and plastic hinges for 

connecting those beams. The program simulates its 

deformation pattern and force-deformation characteristics. 

The force deceleration characteristic$ are used for the 

corresponding spring element in the spring-mass model and 

the vehicle characteristics are determined by spring-mass 

310 

* 

tln 

,H 

Y\ 

HSa. I 

_ff 

Flgure 1. Sprlng-mass mBthod in conlunctlon wlth craft. 

simulation. This method has been recognized as being very 

useful because it requires a very little time to prepare the 

models and run the simulation calculations. 

Several issues have remained unresolved, however. One 

of them is that CRAFT does not have the capability to 

simulate the crash characteristics of frames having 

longitudinal shapes which are too straight, that is, no plastic 

hinges can be predetermined. This is especially true for 

those frames which tend to fold like an accordion. Another 

problem is that spring-mass simulation cannot deal directly 

with the structural design changes intended. 

Against this background, we introduced the commercial 

nonlinear dynamic finite element program PAM-CRASH 

for three purposes: to simulate every type of frame crash 

deformation pattem, to deal with partial or slight structural 

design changes, and to simulate vehicle deceleration 

characteristics with sufficient accuracy. 

Simulation of Vehicle Frontal Crash 

Deceleration Characteristics 

As noted above, PAM-CRASH is being employed to 

simulate the deceleration characteristics of vehicle 

passenger car components during crash tests. We take a 

step-by-step approach using partial models to finally 

correlate all vehicle model characteristics with the test 

results. Each step or phase is outlined below. 

Phase l. Frame model 

Figure 2 shows a typical vehicle deceleration curve at a 

crash speed of around 50 km/h (Prototype A). Figure 3 

presents typical characteristics during the first half of the 

crash. The dotted curve indicates the result of the simulation 

with a spring-mass model consisting of a lumped mass and 

nonlinear springs, as shown in figure l. This simulated 

result correlates fairly well with the crash test data result 

which, in turn, implies that the spring-mass model can effectively 

reproduce the crash phenomena as far as the first 

half of the crash is concemed. 

Figure 4 shows a time history of the energy absorbed by 

each spring element in this model. From figures 3 and 4, it 

can be recognized that the first peak in figure I is derived

g 

z 

v ECE 

5 

U 

t 

Flgure 2. Tlplcal dec€taraflon curve. 

0 z 

q 

E(E 

H 

LIJ 

O 

uJ 

o 

TIME (ms) 

_ 

.--..TEST 

MASS SPRING 

SIMUI.ATION 

Flgure 3. Comparlson of deceleratlon curv€ between mass 

Epr|ng slmulatlon 

snd crash test. 

mostly from the deformation force of the front portion of the 

front side rails. This means that the first step in simulating 

vehicle deceleration is to obtain correlation for the front 

side rail characteristics. 

The front portion of the front side rail was then modeled 

by thin plate (shell) finite elements, and a dynamic crash 

simulation was carried out using PAM-CRASH. In this 

case, modeling of the panel assembly was abbreviated as a 

thickness increase although the test piece was spotwelded. 

This modeling concept is based on the fact that our fundamental 

study indicated that the simulation results were not 

appreciably affected by different approaches to modeling 

panel assemblies (figure 5), 

Figures 6, 7, and 8 compare the results of the simulation 

L€g€nd 

E oTHEFS 

N RAD|AToR 

m HoooLEoGE 

E 

CTR MBH 

W re stDE RA|L 

Flgure 4. Engrgy absorbed by vehlcle components (springmaas 

slmulatlon). 

o 

I 

nME (MSECI 

Flgure 5. Effect of spot wsld modo|tng. 

rEST .f r-|l 

tL 

EMut.^Ttor [) 

C0iarr|ol{ ISO€ fr. 

srMutlTKil{ filt 

M.€tE Dfi|fiEss 

lfll 

M 

and the crash test ofthe front side rail. These figures clearly 

indicate that both the deformed shape and the force-defor, 

mation characteristics derived during finite element simulation 

correlate well with the test results. 

Phase 2. Engine compartment panel model 

The result of the spring-mass simulation (figure 4r time 

history of energy absorbed by each spring element) also 

indicates that the passenger compartment absorbs only 

small amounts of energy during phases I and II of the whole 

collision process. This implies that the passenger compartment 

moves almost as a rigid body with only a very small 

plastic deformation. Accordingly, we attempted to reproduce 

the results ofportions I and II ofthe deceleration curve 

3ll

Flgure 6. Sllnulatlon deformed shape of front slde rEll. 

Flgure 7. Teet deformed shape ol llont slde rall' 

UJ 

o cc 

o 

L 

Flgure 8. Force-tlms characterl8tlcs of lront alde rail. 

of the test results using finite element modeling of the engine 

compartment (the front portion of the body) and the 

simulation. 

As a preparatory stage, only the body panels ofthe engine 

compartment were modeled. The simulation result was 

312 

t 

tl 

ll 

tl 

1l 

li 

-T-- 

,'r\ 

'w^. F- \ 

*,._t't-.-. 

I l l l 

10 

- stMutr\TtoN 

.---TEST 

2b so 40 50 

TIME (msec) 

compared with the test result obtained under the same condition 

to confirm the appropriateness of the body panel 

modeling. 

The engine compaftment body panels of the te$t vehicle 

were assembled for use as the test piece. All panels of the 

test structure were spot-welded to each other. On the other 

hand, the panel connections made by panel flanges in the 

finite element analysis model were modeled as a thickness 

increase as in the case of the frame model (figure 9)' 

Flgure 9. Engine comPailment panel model' 

Figure 10. Slmulation delormed shape of €nglne compartment 

pa-nel etructure at 30 ms. 

Flgure 11. Delormedshape 

of engine comPailment Pan€l 

$tructure at 30 m3,

Figures 10, ll and 12 compare the simulation and test 

results in this case. These figures indicate that the finite 

element analysis model conelate$ well with the panel crash 

characteristics, and confirms the appropriateness of the 

body panel modeling. The calculation took I.5 hours of 

CPU time when PAM-CRASH Ver. 10.0 was run on the 

CRAY-XMP/I2. 

0 

z 

I {CE 

5IIJ 

o 

UJ 

o 

I 

I 

20 

TIME (ms) 

SIMUI-ATION 

.--.-* TEST 

i..i 

i i:.i 

Flgurc 12. _Dsceleratlon-tlme chsrscterlstlc of englne compart. 

m6nt panel. 

Phase 3. Addition of power train components 

For the final $tep, the large components inside the engine 

compafiment, such as the engine and the transmission, were 

modeled by shell elements to simulate the first half of the 

test vehicle crash deceleration characteristics. Those component 

models were connected with the engine compartment 

model as mentioned above. 

Lumped masses were attached to some of the component 

model nodes so that the total mass, the center of gravity, and 

the inertia moment of each component could be adjusted to 

the same values as those of the test vehicle. The engine 

mounting brackets were also modeled by thin shell elements 

and the rubber bushes were neglected. The suspension components 

were represented as lumped masses and added to 

the corresponding nodal points which support them. 

The crash characteristics of all components except for the 

body panels between the barrier and the front wall of the 

engine were represented by shell modeling of some of the 

major components and by the stiffness of the front portion of 

the engine model (figure l3). 

Although the body panel structure model was reformed to 

extend backward slightly, the floor panels of the passenger 

compartment were modeled only at the front portion for the 

following reason. On test vehicle A during the frontal crash, 

L' 

30 

Flgure 13. Englne compartmant slmulation model. 

the force from the power train components to the body 

panels is transmitted mainly through the closed sectional 

beams around the lower area of the dash panel. Therefore, 

it 

Flgure 1tl. Slmulatlon deformed shape. 

g 

z 

at- 

{ 

E 

IJJ 

tu 

(J 

uJ 

o 

Flgure 15, Comparlson of slmulailon and crash te$t dec€leratlon 

curves. 

3r3

can be assumed that the dependence of force transmission 

on the shear stress field of the floor panel is small enough to 

neglect the effect of modeling the rear portion of the floor 

panel. 

The rear half of the vehicle body was represented as a 

moving barrier attached to the front body, with the motion 

ofthe rear end nodes on the front body all being constrained 

except for the vehicle longitudinal direction. 

Figures l4 and l5 show the result of the simulation using 

this model in comparison with the vehicle test results. The 

figures clearly indicate that the simulation correlates closely 

with the first half of the test deceleration time history 

except for the shape of the second peak curve as shown in 

figure 15. This calculation took 5 hours of CPU time using 

PAM-CRASH Ver. 10.0 run on a CRAY-XMP/I2. 

Full Vehicle Simulation 

As outlined previously, considerable testing was 

involved to confirm the applicability of the finite element 

modeling method for simulating the crash characteristics of 

a vehicle front end. 

Here, the entire vehicle was modeled mainly by shell 

elements to simulate the vehicle frontal crash 

characteristics until crash deformation process was 

complete. The reason for this is that the pitching of the 

vehicle at the last half of the deceleration time history 

cannot be simulated by a model which has the rear half of 

the body acting as a rigid moving barrier to which a lumped 

mass is attached. On the other hand, some studies have been 

employed to determine the reason the Phase 3 model cannot 

simulate the second peak of the deceleration curve. 

Actual vehicle tests were carried out to analyze the 

phenomenon responsible for making this peak' It was found 

that the time of contact between the power train and body 

panels was distributed around the peak of the deceleration 

curve as shown in figure 16. 

Thus, the meshing and the material characteristics of the 

model around these contact area$ were reviewed, and the 

Figure 16. Test vshlcl€ dccelsration. 

314 

z 

Fd 

TE 

u 

t! 

r 

TIME (ms) 

shell modeling procedure was partially modified and 

lumped masses were added. 

The actual vehicle crash test around on initial velocity of 

50 km/h also demonstrated that the tires did not affect the 

deceleration characteristics significantly as far as this test 

vehicle and conditions were concemed. The doors were 

modeled by beam elements which have the correlated 

Y-direction crash characteristics of the door panels. The 

hood and the fender panels were excluded in both the 

simulation and test. Figure l7 shows the whole vehicle 

model that was finally devised. 

FIGUHE 17. FUII VEHICI"E $IMULATION MQDEL 

Figures l8 and l9 indicate the simulated results obtained 

using this whole vehicle model. These figures clearly show 

that both the deformed shape and the deceleration 

'-r- .-- - r- I 

*_.:1 ,,'- 

--..-..'.,'..1 

Figure 18. Compsrlson of simulation and crash test 

d€lormatlons.

g 

z 

I { 

EE 

3 

u, 

(J 

u 

o 

TIME (m$l 

Flgure 19. Comparlson of slmulstlon and crash test 

decelerstlon curves. 

characteristics of the passenger compartment from the 

simulation result correlate well with the test result. 

This calculation took 20 hours of CPU time using PAM- 

CRASH Ver. 10.2 run on a CRAY-XMP/I2. 

Conclusion 

This paper has presented a study in which frontal crash 

characteristics, especially the deceleration of the passenger 

compartment, were simulated using an explicit dynamic 

finite element code. Furthermore, the vehicle 

characteristics obtained with the finite element model were 

successfully correlated with actual test results using a stepby*step 

procedure, beginning with only a small portion of 

the whole vehicle body. 

Our future studies will focus on four principal objectives. 

One will be to reduce the 20 hours of CPU time needed for 

whole vehicle simulation. Simplification of the modeling 

involved will be a factor in this improvement. 

A second objective will be to find a way to make use of 

paftial vehicle models introduced as a means of developing 

the final whole vehicle mode. It will be possible in this sense 

to use the engine compartment model to simulate and 

reform the early portion of the deceleration curve. This will 

thus be a very useful first step in studying ways to improve 

Simulation of Vehicle Crashworthiness and lts Application 

Koji Kurimoto, Koji Taga, 

Hiroyuki Matsumoto, Yutaka Tsukiji, 

Mazda Motor Corporation 

Abstract 

Vehicle crashworthiness simulation has become increasingly 

important in recent years from the standpoint of determining 

feasible $tructure within a limited period. This paper 

describes simulation of a full scale passenger car with an 

ride-down efficiency by reforming the deceleration 

characteristics of passenger compartments. 

Another aim will be to use component models to ascertain 

force-deformation characteristics. Spring-mass simulation 

can be carried out using these force-deformation curyes as 

input data. This hybrid method will be most effective when 

used at the early conceptual design stage. 

Finally, the development of an efficient modeling 

methodology for vehicles having various types of power 

train layouts and body structure$ will be a key objective. 

References 

( I ) H.G. Hock, A. Poth and W Schrespfer, 

"lnfluence of 

Modeling Undercarriage and Power Train Componentrr on 

the Quality of Numerical Crash Simulation", The Second 

International Conference on Supercomputing Applications 

in the Automotive Industry, October 1988. 

(2) B. Hazet and M. Guiraud, "Some Applications of 

Crash Simulation in Car Manufacturing Industry", The 

Second International Conference on Supercomputing 

Applications in the Automotive Industry, October 1988. 

(3) J. Hillmann and W. Rabethge, "Computer Aided 

Development of a Vehicle Front-end Structure to Improve 

Safety for Car Occupants in Frontal Collision", ATZ, 

November 1988. 

(4) D.A. Vanderlugt, R.J. Chen, and A.S. Deshpande, 

"Passenger 

Car Frontal Barrier Simulation Using 

Nonlinear Finite Element Methode", SAE paper No. 

87 I 958. 

(5) P.D. Bis and J.F. Chedmail, "Automotive 

Crashworthiness Performance on a Supercomputer", SAE 

paper No. 870565. 

(6) T. Sakurai et al., "Crash Analyses on Passenger 

Cars", Journal of the Society of Automotive Engineers of 

Japan, Vol.42, No. 10, October 1988. 

(7) K. Ishii and I. Yamanaka, "Influence of Vehicle 

Deceleration Curve on Dummy Injury Criteria", SAE paper 

No. 880612. 

(8) K. Matsushita and S. Morita, "Relationship Between 

Vehicle Front-end Stiffness and Dummy lnjury During 

Collisions", The Eleventh International ESV Conference. 

May 1987. 

explicit finite element method in case of a frontal crash. 

Discussions on how to validate a full vehicle simulation 

model with respect to vehicle deformation and rotation of 

passenger compartment in a vertical plane are presented. 

For the former, it is shown that deliberated component models 

such as crossmember, engine supports, and sliding interfaces 

between suspension towers and dash upper, etc., result 

in a good correlation between experiment and simulation. 

For the latter, main components such as hinge pillars, doors, 

3t5

tires are characterized in such a way that good correlations 

between dynamic component tests and simulation are 

achieved. As a result, it is found that a simulation by a full 

vehicle model agrees quite well with the experimental results 

in relation to the rotational behavior of vehicle. 

Analytical results to represent the crash characteristics by 

rneans of intemal energies and transmitted forces among 

various components are also presented. This approach will 

be useful to examine and to improve the crashworthiness. 

Preface 

At high speed frontal collisions, a rotation of passenger 

compartment in a vertical plane, commonly referred to as 

"nose 

dive", as well as the longitudinal vehicle 

deformation often play important roles.(l) This rotation 

may give rise to substantial increase of crash space, or it 

may degrade restraint capability of a seatbelt system, due to 

an additional forward movement of the seatbelt anchorage 

points. This paper describes the application of a large scale 

computer model to simulate the vehicle dynamics in the 

frontal barrier crash. The vehicle is assumed to be a typical 

passenger car repre$ented by about 18,000 thin rthell 

elements governed by a three dimensional Lagrangian 

explicit finite element code, called PAM-CRASH.(2X3) 

The object of this study is to validate a full vehicle 

simulation model with respect to vehicle deformation and 

rotation. In addition, some discussions on the feasibility of 

treating crash process in relation to intemal energies and 

transmitted forces are presented. 

Mathematical Modeling 

Mathematical representation of an ordinary passengercar 

of front engine and front wheel drive is formed on a threedimensional 

space, assuming it crashes into rigid banier' 

A total of 18,000 thin shell elements are used for the 

bodywork and engine, to allow an accurate representation 

of their external geometries and of the internal impacts that 

occur during a crash. In addition, bar elements are used to 

approximate the suspension unit by providing equivalent 

horizontal and vertical stiffnesses to these components. Bar 

elements are also used to represent longitudinal door 

stiffness. Nodal constraint is used to approximate the engine 

supports. Bumper is excluded to reduce cpu-time. Instead, 

equivalent dimension of rigid body to residual of deformation 

of bumper is added as a substitute of the residual and 

longitudinal deformation of vehicle is expressed in such a 

way that difference of length between the bumper and its 

residual is added to calculated deformation of the 

bodywork. 

A mesh of approximately l0 X l5 mm elements are used 

at the front half of the engine compartment to capture the 

impact buckling modes. This then is made progressively 

coarse toward the rear ofthe car. Static strain-stress curves 

of steel, linearly interpolated by 8 points, are used to 

characterize the steel. The windshield is included in the 

model because of its structural impoftance, and represented 

316 

by a mesh of 100 X 100 mm elements with the material 

characteristics which is assumed to be equivalent to thin 

steel plate of I mm thickness. 

Time integration by finite differences yields the solution 

of the problem consisting out of acceleration, velocity and 

displacement time histories at each material particle or node 

of the struature. This time integration is performed using 

explicit method. 

Simulation Fidelities 

A full scale frontal barrier crash test at 35 mph was 

performed, obtaining both barrier force, which is defined as 

force measured from load cells set on the banier, versus 

vehicle deformation, and rotation and vertical movement of 

passenger companment versus time. 

Taking the observation from test results and the outcome 

of several componentwise simulation into account, a 

computer simulation is executed on the full scale vehicle 

simulation model, shown in figure l. 

Figure 1. Full scale model. 

The comparison of tested and calculated results with 

respect to the barrier force and the rotation and vertical 

movement of passenger compartment is shown in figures 2, 

3. and 4. At this moment, the results of simulation do not 

compare to the test results. 

li fr[1a1, orr 

lrsl, 

Flgure 2. Comparlson ol deformatlon charscterlstlce (orlglnal 

condition). 

Vehicle deformation 

Firstly, it is observed, 

in figure 2, a shift of 120 mm from 

the origin on the lateral axes is made for the calculation, 

because of the reason stated before. Secondly, three phe-

(ilsrrLr) 

ilend 

Flgure 3. Comparison of passenger compartment rotatlon 

(original condltlon). 

__11:l__!J:ht rr"ltl-l_lj. j 

Flgure 4, Compariron of pa88enger compartment yertical 

movement (at slde rail lront, orlglnsl condltlon). 

Large neeh SD4l I ncEh 

Flgure 5. Eft6cr by modlflcetlon ol crossmember. 

nomena which differ between test and simulation are observed 

as follows. 

(i) For phenomenon l, firsr peak force by simulation is 

recognized a$ a consequence ofan impact of the equivalent 

dimension of rigid body to residual of bumper deformation, 

to the rigid banier, which is neglected. Next, the value of the 

second peak force by simulation is higher than that by test. 

Judging from the difference of the deformation of crossmember 

between test and simulation as shown in figure 5, 

the cause is considered that mesh used to represent crossmember 

is too coarse for that to buckle at the point between 

its front-end and the engine support. Consequently, changing 

the mesh size for crossmember from 100 X l0O mm to 

30 X 30 mm, difference between test and simulation with 

respect to the deformation of the crossmember and barrier 

force are improved. 

(ii) For phenomenon 2, the deformation of the vehicle by 

simulation is smaller than that by test in figure 2, when the 

peak force is produced at the time when the engine contacts 

to the rigid barrier. Judging from the difference of the behavior 

of the engine between test and simulation as shown in 

Sol id nount Sol t nount 

Flgure 6. Effect by modificatlon of englne mounl. 

figure 6, the cause is considered that the engine supports 

modeled by using nodal constraint are too $tiff for the engine 

to move parallel to the ground a$ test. Consequently, 

changing the model for the engine supports to bar elements, 

which are characterized by using static stress-strain curves 

of the engine supports by test, good agreementri between test 

and simulation with re$pect to the behavior of the engine 

and the barrier force are achieved. 

; t , ; l 

- - i 

- l 

; i 

t' ,|'.1 

' f i 

!. L\' 

1 ! 

i ,., 1 

l 

I 

I '\\ ii 

Flgure 7. Comparlson ol deformed characterlstlc (modlfled 

condltion). 

(iii) For phenomenon 3, at this region, the barrier force by 

simulation, which is attained after the engine impacts barrier, 

is lower than that by test. Deliberately adding sliding 

interfaces behind the engine between suspension tower and 

dash upper, etc., good agreements between test and simulation 

with respect to the barrier force are achieved. 

Figure 7 shows the result of comparison of test with 

simulation with all of the above-mentioned measures at a 

crossmember, engine supports, and sliding interfaces. It is 

shown on this figure that these considerations result in good 

correlation between test and simulation with respect to barrier 

force. 

Vehicle rotation 

llnflj 

inr 

\ 

In figure 3 and 4, it is observed that the rotation and 

venical movement of passenger compartment differ between 

test and simulation. Comparing deformation process 

of each component during the crash between test and simulation, 

observing high speed movie films of the test and 

animated pictures of the simulation, it is recognized that 

deformation of the tires, of the doors and of the hinge pillar 

of passenger compartment are much different between test 

and simulation. Accordingly, validity of material characteristics 

or method of modeling for each of those components 

are examined further as follows. 

(i) The tire, so far, has been represented by thin steel 

317

I 

I i 

l^l- 

l;f 

1,,)f, 

I 

itrl 

lct 

lc) 

lIa- 

f,... 

q) 

L. 

(s 

dla 

00 

80 

[]i 0 

40 

40 B0 120 l[i0 200 

Dr:1'orrrra,Iion(mm) i 

Flgure 8. Deformatlon charsctsristic ol tlre 

plates of 2 mm thickness. A crash test of the tire into the 

rigid barrier is performed at 4.7 mph, which is selected as 

the speed for the tire to deform on the full scale vehicle crash 

test, obtaining barrier force versus tire deformation. By 

executing the simulation of the tire about deformation characteristics 

with various thickness of thin steel plates, it is 

found that using thin steel plates of 0.1 mm thickness results 

in good agreement between test and simulation with respect 

to barrier force as shown in figure 8. 

t-- 

I 

II 

I 

l^ 

lEa 

i-x 

lv 

I 

ior 

l(J 

i L 

l o 

it* 

| ," 

l*, 

t - 

I t".., 

I r.., 

I rtt 

lca 

I I 

II 

I 

40 

2,0 

ltl 

I 

?0 40 

Dr:lora,1,ion(mn) 

Figure 9. Deformatlon characterlstlcs of dool 

!i imrrla,Li 

'l'Er 

s 1, 

(ii) The door, hitherto, has been represented by three bar 

elements. Here again a crash test of the door into the rigid 

barrier is performed at 8.9 mph which is selected to obtain 

comparable deformation to compare the results with those 

by the full scale crash test. It is found that deliberated model 

of the door using shell elements results in good agreement 

between test and simulation with respect to barrier force as 

shown in figure 9. 

(iii) The hinge pillar, so far, has been represented by a 

mesh of 100 X 100 mm. A crash test of the hinge pillar into 

the rigid barrier is performed at 13.8 mph, which is selected 

by the same reaEon stated just above. It is found that using a 

mesh size of 30 X 30 mm results in good agreement between 

test and simulation with respect to barrier force as shown in 

figure 10. 

;{t 

*}( 

"tJ 

(t{ 

Ct 

J 

t- 

.9 

Btj 

[:i 0 

/t n 

'l 

\/ 

r-. ') fl 

at{ 

4t5 

Sirirrla,l;ion 

'l'c: 

sl t 

l) cr l' o r rna, L i o n ( rnni) 

Figure 10. Deformstlon characterlstlce of hlnge pillar' 

l0 

2 

0 50 100 

T iilc( ilsec) 

(4) l,cft llund slde 

T i ile( ilsec) 

(b) Rirht Haild Side 

Flgure 11. Comparlaon of passenger compartm€nt lotation 

(modified condltlon). 

Tine (rscc) 

0 E0 100 0 

(a) Left llrhd Sidtt 

g 

+ t00 

f, 200 

Tile (rsec) 

(b) Risht Hand Side 

Fioure 12. Compsrison of pas3enggr compartment vertlcsl 

mdvement (at slile rail front,'modlflet condiiion)'

Figure I I and 12 show the results of comparison of test 

with simulation with all the above-mentioned measures at 

tires, doors and hinge pillars. It is shown on these figures 

that these considerations results in good agreement between 

test and simulation with respect to the rotation and the 

vertical movement of passenger compartment. 

Decomposition of Transmitted Force 

and Internal Energy 

In order to understand well of crash phenomena in the 

process ofvehicle crash, it is necessary to obtain data such 

as force, deformation, etc., at various components of 

vehicle, as well as at vehicle as a whole. However, generally 

these data obtained from the direct measurement are limited 

because of large deformation, interference between each 

component, high deceleration during the impact, etc. 

Therefore, it is thought quite useful and also necessary to 

obtain the supplemental information by taking advantage of 

the simulation of mathematical model as described in this 

pflper. 

As typical examples, figure 14 and figure 15 show the 

transmitted force and internal energy among various 

components obtained from the simulation described above. 

These figures show thflt the force inflicted on the front end 

of the front rail is conveyed to respective components 

toward the passenger compaftment and that the energy is 

absorbed at these points. The location of these points are 

shown in figure 13. 

Flgure 13. Sch€matlcs of anglne cornpartment. 

Figure 14 indicates that, in this case, the engine seems to 

have a small effect on the transmitted force on the front rail 

at the rear part of engine support, because the value of 

transmitted force at front end of front rail nearly correspond 

to that at rear part of engine support. In addition, transmitted 

force of front rail seems to be divided equally at rear part of 

suspension tower into the rear end offront rail and the upper 

part of the engine compartment, namely, wheel apron. 

Figure 15 indicates that the percentage ofthe absorption 

of the initial kinematic energy possessed by the vehicle is 

about 307o for the front rails and about 25Vo for the 

cro$smember. Thus, the paths or amount of, transmitted 

- 

(u 

L 

o 

t! 

800 

600 

400 

200 

-f1snl end of f ront rail 

"-" Rear side of 

eng i ne support 

---.- Rear end of f ront rail 

..*.--Rear end of uheel apron 

0l 

0 l0 20 30 40 50 60 

T i me( msec ) 

Flgure 14. Decomporltlon of tren$mltted force. 

l5 

Eto 

>r 

bo 

L 

o) 

u D 

Flgurc 15. tl,ecomposlllon ol Internal energy. 

force and internal energy are easily obtained, which 

generally difficult to measure. 

Summary and Discussions 

ln it ia I K inemat ic 

Energy 

0 20 40 60 B0 100 

Time(ers) 

Front rail 

Crossnenbe 

Eng i ne 

t/hee I apro 

It is demonstrated that the vehicle model by finite 

element method is effectively and conveniently used. 

In order to achieve good correlation between the mathematical 

model and actual full scale frontal barrier crash test. 

it is found necessary to deliberately make model of 

components such as a crossmember, engine supports, etc., 

and to carefully characterize main components in such a 

way that good correlation between dynamic component test 

and simulation are achieved. 

Among this area to be examined further is a curtailment 

of the number of shell elements for reducing cpu-time. An 

approach to this area may lie in a simplification of the model 

based on closer examination of actual phenomena, by 

eliminating ineffective components. 

Although there is room for improvement, the computer 

simulation method presented here is considered acceptable 

for practical applications, for example, to various parameter 

studies. 

319

Acknowledgment 

The authors would like to thank the related people of 

Mazda Motor Corporation, who gave kind suggestion in 

preparing this paper. 

References 

(l) K. Kurimoto, et al., "An Analytical Management of 

Frontal Crash Impact Response", Proc. of lOth ESV conf., 

pp.452{459, 1985. 

(2) Paul Du Bois, et al., "Automotive Crashworthiness 

Performance on a Supercomputer", SAE paper 870565, 

1987. 

(3) Donald A. Valder Lugt, et al., "Passenger 

Car Frontal 

Barrier Simulation Using Nonlinear Finite Element 

Methods", SAE paper 87195t1, 1987. 

Occupant Simulator Preprocessors: Parameter Studies Made Easy 

Edwin M. Sieveka and Walter D. Pikey, 

University of Virginia Department of 

Mechanical and Aerospace Engineering, 

William T. Hollowell, 

U.S. Dept. of Transportation National Highway 

Traffic Safety Administration 

Abstract 

The use of computer simulated parameter studies 

provides a cost-effective means for examining a wide range 

ofaccident scenarios and vehicle designs. To facilitate these 

studies the National Highway Traffic Safety Administration 

is developing preprocessor programs for several occupantenvironment 

crash simulators. Given an existing input file, 

these preprocessors automate the task of generating the 

many, perhaps thousands, of additional files that are used in 

large parameter studies. This paper describes two such 

programs, PADPREP and MVMAPREP. PADPREP creates 

new input files for the PADS simulator which is used to 

study driver/steering-assembly interactions. MVMAPREP 

supports the MVMA-ZD program, which is used for 

passenger and pedestrian simulations. In addition to 

describing program use, demonstration parameter studies 

generated by each preprocessor are included as examples. 


A high priority in automotive safety research has always 

been to optimize occupant compartment design so that the 

largest margin of safety is provided over the widest possible 

range of accident conditions. The variables which need to be 

considered include seat design, restraint design, dashboard 

design, steering assembly design, collision speed, and 

compartment integdty (resistance to intrusion). Adequately 

assessing the effects of these many parameters can require 

thousands of test "crashes." Performing such a parameter 

study with actual vehicles is prohibitively expensive and 

time consuming; only a few subsystem and full scale crash 

te$ts can reasonably be performed on each production 

vehicle, and these are usually done at the impact speed 

required by government te$t specifications. Thus, the goal 

of optimizing vehicles for occupant safety, or of evaluating 

320 

the full performance range of existing vehicles, cannot 

begin to be realized without the aid of computer simulated 

crashes. 

In its frontal protection research program, the National 

Highway Traffic Safety Administration (NHTSA) has 

chosen to focus on the MVMA-ZD and PADS occupant 

simulators (1, 2, 3).+ MVMA-ZD is used for passenger 

simulations; PADS was designed to simulate primarily the 

driver and steering assembly interaction. Using only the 

simulators, parameter studies are difficult because they 

require manually editing hundreds of input files, which is 

tedious and time consuming. To expedite creating the 

numerous input files required in parameter studies, the 

NHTSA has sponsored projects at the University of Virginia 

to write preprocessor programs for both PADS and 

MVMA-2D. They are referred to as PADPREP and 

MVMAPREP. Starting from a single input file, these 

programs can generate sets of up to 1000 new input files for 

their respective simulator programs. 

This paper present$ sample parameter studies which 

demonstrate the use of the PADS and MVMA-ZD 

simulators in conjunction with their newly created 

preprocessors, PADPREP and MVMAPREP. 

Overview of the PADS Software 

System 

Pads 

The PADS occupant simulator was developed for the 

NHTSA by MGA Research Corp. and has been updated by 

the University of Virginia (4, 5). It can model both drivers 

and passengers but it is designed primarily to model 

driver/steering-assembly interactions. Six modes of 

steering assembly deformation are provided, including 

radial and tangential wheel-rim bending, whole-wheel 

crushing, wheel rotation, column rotation, and column 

stroke. Inertial resistance of the wheel/column mass is 

included, as well as frictional resistance at the $hear capsule 

and the column bushings. The capability of restraining the 

occupant with a 3-point belt system is also included. 

*Numbers in puentheses designate references at end of papcr.

The occupant's body is modeled with four jointed 

segments representing the head, torso, thigh, and lower leg. 

Five contact sensing circles, shown by dashed circles in 

figure l, detect impacts with vehicle surfaces other than the 

steering assembly. The surfaces recognized are the roof 

header, windshield, upper dash, middle dash, lower dash, 

seat cushion, and seat back. In driver simulations, torsocircle 

impact with the dashboards is assumed not to occur 

due to the presence of the steering wheel, but the head circle 

is used for roof and windshield contact and the knee circle 

for lower dashboard contact. 

Wheel contacts with the occupant are presently 

determined from the Beometry represented by the heavy 

lines in figure l. The torso region is translated forward, 

relative to the torso line-segment, by the radius of the upper 

torso circle; its thickness remains constant as the occupant 

rotates. It is divided into shoulder and abdomen regions. 

The head region is also rectangular but is based on the radius 

of the head circle. It, too is unaffected by rotation of the 

body or head. Future plans call for the head circle to be used 

for head/wheel contact, but this requires additional 

revisions to the contact algorithm. 

Flgure 1. PADS contsct zonee. 

Kinplot 

The PADS KINPLOT program produces plots of occupant, 

steering assembly, and contact surface position at 

specified time intervals; figure 2 shows an example of such 

a plot. Occupant position plots are essential when working 

with simulator programs as a qualitative check on the fidelity 

of the calculations. If one ignores position plots and 

concentrates too much on numerical results such as the 

Head Injury Criteria (HIC), the Chest Severity Index (CSI), 

or the peak head and chest accelerations, it is possible to 

obtain plausible quantitative answers which actually come 

from qualitatively unrealistic occupant kinematics. 

PADPREP 

The PADPREP preprocessor program accepts as input a 

standard PADS input file as well as another file called the 

experiment design file (EDF). The EDF file contains 

information for the PADS program and also a list of up to 

eight variables which are to be modified by PADPREP. The 

file is in template form with u$er prompts built in. This 

approach reduces the amount of interactive input required 

by PADPREP while still giving the user clear input 

directions; it also provides a useful reminder of the 

parameter study's contents. A sample EDF template is 

shown in figure 3. The fitrst five entries tell PADPREP about 

the type of input file being used and the type of output 

ffi tut af,tnl$ ffiffi flPtr 

I-EffiIft z-Tfl-AffiIE 

I 

--- ffirM Drarg ttd 

---- lM3 ru-lM'il 

(MIIEHnIBIfCfi|EII 

ArrCE Td ffDI O' lEHilD ffiPU? DISTEDI 

IrDlttrru z-MilfiD 3-lllUlffi VEGA ffiI a.xffi 

I 

arc ffi ffi tr!ftffi rcl DlDt ffir Ot ffi MIEI 

r) fftffi yw xril lldtt tmY 

b) IX?E-ilil LI'TIrc IT E IIT 

ot f,ItrFH Dlil-rrut 

d) fiEIff DEF'ITS 

.) cuffitlE rul-PHElaor tM 

tt CUMTIE S4il rr& ffiFr 

o.H.r mrs l|IEil 

(F f,fi aEPrsI rdat f ,g ttTt cffi il ap&!a] 

t0t0r0 

ffif, Til ffDI OI trCUPffiI 

l-DRrWf, Z-FEAEHI 

I 

aEH rd ff?l Ot tEI EAmIFT 

0 

o-ffiilgmlED l-N ffit :.tfllo fitt 

ErE rE xlmEt or vffiilg s Sctu (H. il 8)r 

j.N r ffiO 

ETE VNINC DIfA II IE 'IE4 IH trI N& LIEI 

VNl:IIr 

{vr!'rr[ xrEr Eaf il rE sgtttrtot 

Til{ 

cdfrect 

COHIE 

ATMIft IIffiM ruI COIxlnIIfi 

VEM VEE OF trEA trc 

0,5 

0.d3 

0.5 

0.03 

Flgure 3. PADPFEP erporlment deslgn flle. 

0 0 

0 t 

ad 

ibor 

32,1

desired. The next entry is the number of variables in the 

parameter study. At the end of the file is a table containing a 

list of variable names, with starling values, increment 

values, and the number of cases (increments) to be 

computed for each variable. Two additional entries are also 

included. The correlation flag permits a variable to be 

correlated with the one immediately preceding it if this flag 

is non-zero; a chain ofcorrelated variables can be built up in 

this manner. Finally, the .rcale mode field determines 

whether scaling (scale mode = 0) or direct input (scale mode 

= I ) is used. Either is permitted except for vector quantities; 

in which case, the new value is treated as a scale factor only. 

PCFG 

The PADS Control File Generator (PCFG), which was 

written at the DOT Transportation Systems Center (TSC), 

expedites the creation of PADS input data sets. It is designed 

to operate on a special PADS database that is maintained by 

the NHTSA. The user specifies the type of vehicle, the type 

of steering wheel, the collision acceleration pulse, occupant 

size and age, and other data, such as whether the occupant is 

belted or not. The PCFG can also be used to alter most of the 

PADS input variables and can do parameter ranging to create 

multi-file "run sets" for use in parameter studies. It 

differs from PADPREP in that new input files are always 

created by accessing original information from the PCFG 

database. PADPRER on the other hand, operates directly on 

an existing PADS data file. The two programs can operate 

effectively in tandem, with the PCFG used to create a baseline 

file which can then be modified easily with PADPREP. 

This combination is especially useful if the baseline file is to 

contain numerous changes to the database values since the 

PCFG can be used to "edit" the database information before 

creating an input file. Once this baseline file has been 

created, however, it is more efficient to perform parameter 

scaling with PADPREP since no further reference to the 

database is necessary. 

PADS Simulator Demonstration 

The examples which follow are a qualitative 

demonstration of the capabilities of PADS and PADPREP 

and their potential usefulness in occupant compartment 

parameter studies. 

Basic capabilities 

The set of plots in figure 4 demonstrates the various 

modes of steering assembly motion. The interior dimensions 

approximate those of a subcompact passenger car and 

the collision speed is 20 mph. In figures 4a4c the steering 

column position is such that the contact forces generated by 

the occupant are primarily parallel to the axis of the column 

and hence there is little column rotation but considerable 

column stroke and wheel crush. Head injury in this case 

would be due only to windshield contact. In figures 4d-4f, 

the column has been raised higher relative to the occupant. 

The first position is closer to the actual wheel position, but 

322 

the second situation could occur during an accident ifdistortion 

of the car's front end displaces the steering assembly 

from its normal location. In this case. the forces perpendicular 

to the column axis are much larger and significant column 

rotation occurs. More important, the column strokes 

much less, Because of the design of collapsible steering 

columns, if the perpendicular force is too large it creates 

excessive friction between the column's moving parts and 

the stroking motion is inhibited. Thus, a column which 

strokes properly and mitigates injury in its normal design 

configuration may perform poorly if its alignment relative 

to the occupant is altered during a crash. This is a good 

example of a case in which a simulator program can prove 

very useful. Since it is not practical to build actual cars with 

steering assemblies in different locations, the simulatorprogram 

can be used to estimate the range of displacements 

relative to the occupant over which the column's crush 

characteristics can provide significant injury reduction. 

Standerd C-olufll Rafued C;olumn 

time = .080 sec 

timc =.120ecc 

timc = .180 sec 

Flgure 4. Subcompact-20 MPH colllelon. 

Parameter study 

As a simple demonstration of a parameter study, PAD- 

PREP was used to generate nine PADS input files from the 

subcompact file that was used for figures 4a4c. The variable 

studied was the steering column stroke resistance and it 

was allowed to vary from 0.5 to 2.5 times its baseline value 

in increments of 0.25. Figure 5 shows the entries in the EDF 

template needed to generate this study. All of the PADS

simulations produced plausible occupant behavior. The injury 

measure numbers are of the appropriate magnitude but 

for now should be considered relative to each other only. 

The HIC and CSI results are plotted in figure 6. In general, 

the curves display the anticipated trend of increasing injury 

level as column stiffness increases. Note that both curves 

show a decrease in injury for columns softer than the baseline 

value. This could indicate that a softer column would 

provide for greater occupant safetyn but other factors such as 

injury levels at higher collision speeds would have to be 

considered as well. 

lEl rfl ffilt 6t vNlMt S ru (ll.t. Ot t)t 

t 

ffi vW ilTl Itr Tf,E tIru M Wg NBES LIXIr 

(V[M TIEA roAf IT TE ilAIIFIEDI 

V|W ATNTIfr IXffiT roEI' CffiIIfr 'ffi 

w vEu vE4 0t illt w Ht 

ffiE 0.5 0.I5 I 0 

Flgure 5. EDF rntrleE for column stlftnoss study. 

C.olumt Stilfnct 

Flgure 6. HIC and CSI reaponae surfacee. 

The peak acceleration levels for the herd and chest are 

plotted in figure 7. The chest results show only moderate 

sensitivity to column stiffness for values below 2 times the 

baseline, however, at higher stiffnesses the curve rises 

sharply before leveling off again. In the case of head acceleration. 

the results are even more dramatic. with the curve 

being completely flat through 2 times the baseline and then 

jumping suddenly by about 30 G's for the last two points. 

Other than to say that a softer column seems slightly bener, 

there is no clear indication from these results of what the 

optimum design should be. They do, however, illustrate 

another important piece of information that can come out of 

a parameter study. 

While one generally approaches a parameter study with 

the intention of finding an optimum design, it can also be 

important to identify design regions that are particularly 

bad. In this case, based on peak head and chest acceleration, 

the conclusion would be that column stiffness should be 

restricted to values less than about 1.75 times the baseline 

value. These results also illustrate how a study ofdifferent 

injury measures can be useful. The values of HIC and CSI 

Grlumn Stlffncq 

Flgurc 7. Peak acceleratlon r€sponse surfaces. 

rise steadily with column stiffness over most of the design 

space, but there is nothing dramatic or unexpected about the 

curves. The sudden jumps in the acceleration curves, however, 

clearly alert the investigator that a transition to a different 

column-performance regime has occurred. 

MVMA-2D Software Overview 

MVMA-zD 

The MVMA-2D occupant simulator has been developed 

at the University of Michigan Transportation Research 

Institute with support from the Motor Vehicle Manufacturers 

Association (6). The occupant's body is modeled by 

eight lumped masses with fourteen degrees of freedom. The 

body's "surface" can be represented by user defined 

ellipses. As in PADS, the vehicle's interior is represented by 

a series of line segments, but in MVMA-2D the number of 

segments can be chosen by the user. Contacts can occur 

Flgure 8. MViIA-2D/TRIX output. 

323

etween body ellipses and the vehicle line segments, and 

between the body ellipses themselves. The program also 

contains an airbag model, simple and complex belt models, 

and a steering assembly model. The latest documentation 

states, however, that the steering assembly model is not 

fully operational. An example of MVMA*ZD passenger 

and vehicle geometry is shown in figure L 

TRIX 

The TRIX program (shon for Tektronix Stick figure plotting 

program) is the MVMA-2D equivalent of the PADS 

KINPLOT program. It produces occupant position plots 

such as the one in figure L As with KINPLOT, examination 

of TRIX output is essential in order to verify that the occupant's 

motion is realistic and that occupant/vehicle contacts 

are properly accounted for. 

MVMAPREP 

The MVMAPREP program is a preprocessor designed to 

create multiple MVMA-ZD data files for a parameter study 

which uses an existing MVMA-2D data file as a starting 

point. As with the PADPREP program, the modifications to 

be made to the baseline file are specified in a template-style 

EDF file. An example template is shown in figure 9. Each 

line of user input must begin with a ">" marker in column 

one and all entries must be placed within the fields delimited 

by the dashed lines below each entry header. All numerical 

data items can be scaled or directly replaced by a user 

defined quantity. Character data items can be modified in 

the sense that ellipse contacts can be changed, added, or 

deleted, as well as material propefties associated with ellipses 

and vehicle surfaces. Justification of numerical data 

CN@ ETIPEE CffircT8I 

----- lHHffi DIAIOX tIU 

------ M-lD ffi-ffilgaoR 

ffil lxBM ffi IHI rxDlilsD rrff6 ---------------- 

-- oH lrf,lE Eclfrlffi srTf, t 

" 

> 

ITIPAE ES M c@rtr til cilT&t 

ttcf, 

sDorutr 

ssD 

aEN 

ffi 

ffit r? 

ffitD 

CErre! ETIPEE OR ffiIil ATERIE FNOruf,TY E'IGXI|EMI 

> @ F 

lttrtBt oE 

nlclof, xrE 

SIIRIE 

CErrcE trlERICd FMEB IY ffiIfd dR DIruCT ItrFM: 

Ill xox-PNtEtr 4sDl v&l$lEar 

tlltl 

ID* 

&IPAE OR 

ruolox trrE 

ffi 

ftc, trltD 

t f 

f , m E 

VEE 

t07 xrDaftT 2 3 r.1 0 

1 0 I I 6 1 0 0 . 1 

l2I euErnr I t 40,0 I 

(21 PIIIETER alml v*ISl88 (N. oF 8)t 

LItrE 

rDf 

ETI'8E ff 

MIil 

tIFEr r 

LIFSTAT 

f,rE 

- { i 

- 4 3 

Flgure L Experlment deslgn fil6. 

324 

0,5 

X 

('cC. Frrd ATffitlrc IfCUEm I Or C a vlslM 

| | vsur v&E lflcR. r f, NrE 

0,5 a 0 0 EtoftE 

a r 0 twtdcr 

within associated fields is not necessary. However, since 

MVMA-2D treats blanks within character data fields as 

significant, the user must insure that ellipse, region, and 

material names in the EDF file occupy the same position 

within the field that they occupy in the baseline file. 

The first two categories in the EDF file, ellipse contacts 

and material property assignment, represent one-time 

changes. MVMAPREP reads the baseline file into an internal 

workfile and makes all the above changes before generating 

the parameter study files. 

The first sub-category under numerical parameter 

changes is also used for one-time changes, in this case for 

numerical data which is not part of a parameter study. A line 

ID number is required for all enfries. The ellipse or region 

name is needed for those cartes for which there are multiple 

lines with the same line ID number. The occurrence number 

entry (OCC.#) allows a specific data point within a force/ 

deformation table to be altered or allows the entire table to 

be scaled. Once a match is found for the line ID number and 

the name field, MVMAPREP moves down (OCC.#-1) additional 

lines, if the occurrence number is positive, before 

making a change. If it is negative, MVMAPREP changes 

ABS(OCC.#) lines including the original match. The new 

value field contains scale factors or new values for direct 

input. Finally, the scale mode field (SM) determines whether 

scaling (SM = 0) or direct input (SM = I ) is used. Either is 

permitted except when the occulrence number is negative, 

in which case the new value is treated as a scale factor only. 

The last entry category in the EDF file is for numerical 

parameter study variables. Up to eight variables, producing 

up to lff)O different case combinations, can be processed in 

a single study. Most of the entry fields are the same as for the 

non-parameter study category, but a few have been added. 

The new value field has been replaced by a starting value 

field plus an increment field. The increment number field 

specifies how many levels of each variable are to be calculated. 

The correlation factor field (CF) permits variables to 

be correlated with one another in the same manner as described 

for the correlation flag in PADPREP. Finally, the 

last addition is the variable name field which allows the user 

to give name$ to the variables in the parameter study. 

MVMAPREP/MVMA-ZD Parameter 

Studies 

Following sections describe two related parameter 

studies performed with MVMA-2D with the aid of the 

MVMAPREP preprocessor. The variable chosen for these 

studies was the stiffness of the torso belt portion of a threepoint 

belt restraint system. The baseline data set was 

derived from a NHTSA simulation of a mid-sized passenger 

car sled test at 30 mph. 

The motivation for studying the torso belt stiffness came 

from the observation that the restraint belts permitted very 

little forward motion of the occupant's torso despite the 

relative severity ofthe crash (see figure l0). This suggested 

the possibility that belts with greater stretch (or spoolout)

would reduce injury measures, such as HIC and peak 

acceleration, by providing the occupant with a longer "ridedown" 

time. 

Flgure 10. Msxlmum forward occupant posltion wlth origlnal 

belt stlffnssE. 

Case one 

The force/strain data in the original data file was linear 

beyond a strain of 6Vo and reached a value of 9450 lbs. at a 

strain of lfi)7o. For the purpose of the parameter study, the 

following baseline data was chosen: 

Strain -0.00 0.25 0.50 0.75 1.00 

Force -0.00 2500 5m0 7500 l0(n0 

To create the study, the entries shown in figure I I were 

entered into a blank EDF template. The changes in the nonparameter 

study category ensure that the instrument panel 

area is sufficiently stiff for contact with the head to be 

obvious. Any change in belt design which permits head/ 

dash contact is considered unacceptable in this study and 

such contact should not be masked by an overly soft dashboard. 

The parameter study itself is described by the last 

line in the EDF file which specifies that the force/strain data 

table for the torso belt is to be scaled in ten l07o increments, 

beginning at l07o of the baseline value. The resulting family 

of force/strain curves. one curve for each of ten new 

MVMA-2D input files, is shown in figure 12. 

CBrEr ffid Tffi IT ffiIffi il DIffi Iffit 

III H-tM ml vull$rul 

LIil ETffi * trC. 'IE H 

rDI HIf,rM I . YM 

> 407 TItGlt -a I I0.0 0 

I ao7 EEI! -a r 3.0 b 

tzl rM tmt vw (ru. or lrr 

ttn Erflt il ffi, trH EErc rrffi | d t I 

IDI rEIilff T I vEil vEE IfcI. Ix IrE 

> tol affiT -3 t o.LI 10 0 0 tEffi 

Flgure 1 1 . EDF entrles lor belt stlfine$$ *tudy-Case 1. 

I 

I 

F 

o 

R 

c 

E 

Flgure 12. Bell stlflneas.4sse 1. 

l.f 6.4 a.a 

STRAIN 

The parameter study response surface$ for the HIC, peak 

head acceleration, and peak chest acceleration injury measures 

are plotted in figure 13. The most dramatic feature 

occur$ at the belt stiffness scale factor of l0vo. At this 

stiffness the occupant makes hard contact with the dashboard 

as shown in figure 14. For all other stiffnesses, even 

down to 20Vo of maximum. no dashboard contact occurs. 

however, there is also no dramatic improvement in the injury 

measures. The best case is for a scale factor of 0.4, for 

which the improvements in HIC, head g's, and chest g's are 

9Va, llo/a, and lTVa, respectively. Still, one would probably 

TBE. E 

I .r"." 

N 

J 

U 

R 

rla.. 

Y 

M 

E 

A 

s 

U 

R 

E 

rfttt. I 

tQl6. a 

AEEE. E 

768E. E 

65Ca,5 

t€ea,a 

{EEE. E 

tE6€,6 

eE6C. a 

1464. € 

6,6 

tf .5 

lo.c 

3E.E 

s.6 

t.l 

E.e 

c. a E.a 

Flgure 13. Inlury reeulta-Caee 1. 

5,1 

a,t 

BELT STIFFNESS SCALE FACIUR 

t.t 

32s

conclude that for belts with linearly increasing stiffness, 

there is little to be gained by simply reducing the stiffness of 

the original design. 

Flgure 14. Occupant-dashboard contact tor belt stlffness factor 

ot 0.1. 

Case two 

There are, however, other ways in which the belt performance 

can be changed. It is well known from shock isolation 

theory that having an energy absorber with piecewise 

constant force/deformation properties is the best way to 

minimize peak accelerations. Furthermore, the first constant-force 

plateau should be reached as quickly as possible. 

In the case of seat belts, some approximation of this effect 

might be achieved via the spoolout mechanism. 

To test a simplified version of such a force/strain curve 

for seat belts, the parameter study EDF template was modified 

as shown in figure 15. Before scaling, the force and 

strain data table was changed so that a constant force plateau 

extends from a strain of l2.5%a to 62.5Vo. Since it was 

observed in Case I that the maximum force stayed below 

25fi) lbs, this value was used for the plateau level of the 

stiffest curve (scale factor of l). The resulting family of new 

belt stiffness curves is shown in figure 16. 

The injury measure curves for this parameter study, 

shown in figure 17, are indeed much more interesting. Solid 

contact with the dashboard again occurs for a scale factor of 

cwd Iruw tM tt ruIU6 fi DIIH llml 

(llH-Pffirffiffir 

Ltn 

rDl 

tEDttrt 0t 

ffiIf, ffi 

ocd, tttD id t 

f f f f i r 

> aor rrrrsT -a 3 10,0 0 

> tot Effitt -a I t.0 0 

' ,01 rffi 2 2 0.lll I 

' 706 8ffi! I 5 2100. r 

> 70r rM a I 0,625 I 

> 70t dffiE! a I 2f,O0. I 

IzlDr$GHffiffi(H.qr't 

im &IrEil ffi.tttrttMIBffiad cavM 

rDl wrilIE | | vEE vEE lf,f,.tx IM 

- --*;--- 

0,I 0 0 tl[$ff 

Flgure 15. EDF entde$ for b€lt $tlffness study4ase 2. 

326 

F' 

U 

t9666 , E 

96eE ' a 

EE6€'6 

?666.e 

6CaO.e 

R 

t668.€ 

c 

E 

+EEE. E 

3EEE. E 

8685.6 

1666.6 

6.8 

6,6 6,a e.{ 6,3 E,a 1.S 

STRAIN 

Figure 16. Belt stillness.4aae 2. 

0.1 and a grazing contact also occurs for the 0.2 case. Injury 

measures for the three stiffest belts are similar to those 

obtained in Case I. This time, however, the curves show 

clear minima in the HIC and head acceleration values at a 

scale factor of 0.4. The corresponding chest acceleration 

value is also the next to lowest. The improvements in the 

three injury measures relative to the original linear baseline 

curve are 54Vo, 4OVa, and22Vo, respectively, a very substantial 

change. The occupant's maximum forward position 

when using this belt stiffness is shown in figure 18. 

While the results of this study have not determined the 

optimum design characteristics for torso belts, they have 

I 

N 

J 

U 

R 

Y 

M 

E 

A 

s 

U 

R 

E 

6.t 8.3 

BELT STIFFNESS SCAI.E FACTOR 

Flgure 17. lnlury results-Gase 2. 

E.J s.?

Tirnc(mmc): 120.fi 

Flgure 18. Maxlmum forward occupsnt po8ltlon wlth belt stlffness 

factor of 0.4. 

demonstrated the power of even a simple, qualitative occupant 

simulator parameter study to shed new light on safety 

system design. The simulation study has shown that occupant 

injury levels show remarkably little variation over a 

wide range of belt stiffnesses when using belts with linear 

stiffness characteristics, and that significant improvements 

may be possible through proper tailoring of the belt's force 

versus strain behavior. 

Simulator validation 

An often overlooked application of preprocessor generated 

parameter studies is in the area of simulator program 

and model validation. The usual approach to occupant simulator 

validation is to compare simulated occupant responses 

to measurements taken from a sled test or a fullvehicle 

crash test, of which only a limited number are available. 

Due to this small sample space, initial validation efforts 

do not always uncover program limitations that can 

cause problems when other designs and crash environments 

are simulated. Sometimes input data can be chosen during 

the initial validation effort that gives agreement with test 

results while masking a flaw in the program. Since extensive 

parameter studies place special demands on simulation 

programs by requiring them to run reliably over a wide 

range of input data, such flaws will often surface during a 

study which pushes the limit$ of the original model. 

Even when a program has been properly validated, it is 

still possible to construct a model with data that irr too tightly 

bound to a particular test. For example, a table of force 

deformation data which gives good agreement in a 25 mph 

simulation may be too limited in range for an impact at 35 

mph; leading to incorrect results. For such reasons, it is a 

good idea to subject a baseline model to some preliminary 

"test" 

parameter studies in order to make refinements be- 

fore using the model extensively. 

References 

( l) Cohen, D., L. Stucki, C. Ragland, 

"Development of 

Analytical Procedures to Characterize the Vehicle 

Environment in Frontal Impact Accidents", SAE paper no. 

850251,1985. 

(2) Stucki, L., D. Cohen, and C. Ragland, 

"Evaluation of 

Frontal Occupant Protection Using the Passenger/Driver 

Simulation Model", Proceedings of the Tenth 

International Technical Conference on Experimental 

Safety Vehicles, Oxford, England, pp.459-4?9, May 1985. 

(3) Cohen, D., and L. Simeone, "The Safety Problem for 

Passengers in Frontal Impacts-Analysis of Accident, 

Laboratory, and Model Simulation Data", Proceedings of 

the Eleventh International Technical Conference on 

Experimental Safety Vehicles, Washingron, D.C., lgB7, to 

be published. 

(4) McGrath, M. and D. Segal, "The Developmenr and 

Use of the PADS (Passenger/Driver Simulation) Computer 

Program", Final Report, National Highway Traffic Safety 

Administration (NHTSA), Conrract No. DTNH2Z-EZ-R- 

07017, March 1984. 

(5) Sieveka, E.M. and W.D. Pilkey, "Implemenration 

of 

the PADS Program in the Safety Systems Optimization 

Model", Final Report, National Highway Traffic Safety 

Administrarion (NHTSA) Conrracr No. DTRS-s7-95- 

00103, 1988. 

(6) Bowman, 8., R. Bennett, and D. Robbins, MVMA 

Ttvo-Dimensional Crash Victim Simulation, Version 4, 

Vols. l-3, University of Michigan Highway Safety Research 

Institute, Final Repofi No. UM-HSRI-79-5*1,2,3, 

June 1979. 

on the Safety Body Structure of Finite Element Method Analysis 

Toshiaki Sakuria, Toshihiro Aoki, 

Passenger Car Engineering Center 

Mitsubishi Motors Corporation 

Abstract 

In developing the safety body structure of a small vehicle, 

it is necessary to consider both the body structure system 

and restraint system. In this paper, body structure system is 

discussed without changing the current characteristics of 

restraint system. In order to determine the crashworthiness 

requirements of body structure, the body is divided into two 

subsystems viz. engine compartment and passenger 

compartment. Deceleration Ge and Gc are introduced for 

subsystems, while for total body structure, transfer function 

Dc is introduced. 

Thus, there exists the optimal crashworthiness balance 

between the engine compartment and passenger 

compaflment.

On the other hand, it is necessary to devise an analytical 

method to verify the requirements. Various analyses are 

studied. As a result, it is seen that FEM (Finite Element 

Method) is useful to predict the crashworthiness. It is also 

$een thflt the passenger comPartment protection predicted 

from FEM is verified by testing the prototype vehicle' 


Computer simulation makes it possible to evaluate the 

safety related characteristics of a vehicle. ln previous 

studies of crash analysis of a vehicle mass-spring 

simulation, static buckling analysis and experimental 

formulas have been carried out (l)+. However these 

methods could not respond to design of engineer for safety 

vehicle in an intial design stage of development- 

The objective of this paper in noticing body structure 

system without changing the current restraint system 

discusses the following items: 

(l) To investigate the experimental data variance at 

the fixed barrier frontal impact test' 

(2) To compare data with finite element code which 

is capable of treating non-linear materials and 

experimental results. 

(3) To illustrate crashworthiness of body structure 

by defining factors Ce, Gc and Dc as dependent 

variance. 

(4) To confirm that there exists the optimal 

crashworthiness balance between the engine 

compartment and passenger compartment on the basis 

of from FEM and verify it by testing the prototype 

vehicle. 

Primary Experiment 

Experimental Dsta Variance 

It is necessary to confirm the reliability of experimental 

data by evaluating calculation results and experimental 

data. A few impact tests have been carried out' Table I 

shows the coefficient of variation with the test results of 

initial maximum deceleration of a vehicle to the fixed 

Table L Cosfilclent of varl8tlon. 

u--H*" 

i : Mcan Valuc 6 ; $t6ndrrd Deviation 

barrier in Iow and high velocity. Table 2 shows the 

coefficient of variation with the respect of static and 

dynamic compression test of straight and curved frame. 

Table 3 illustrates the results of tests executed to study the 

coefficient of variation of occupant injury from sixteen 

+ Numbers in parentheses designdtc rcferences at end of ptper. 

328 

Coefficicnt of Variation 

X,=oli 

5mph fixed vanier 0.05 

30mph fixed vanier 0.07 

Tsble 2. Coefficlent of varlstlon. 

Coefficient of Variation 

0 0.05 0.1 

Straight 

Frame 

Curved 

Frame 

i-iTFI 

tFl : Sratic l-l ; Dynamic 

Table 3. Coetflclent ot varlatlon of occupant inlury. 

Occupant Injury x o x 

HIC 

CHEST g 

FEMER LOAD 

lb 

D:Driver Seat P:Passenger Seat 

vehicles under the fixed banier frontal impact tests. As 

shown in table 3, the coefficient of variation is considered to 

be about 0.07. In general the coefficient of variation is 

relatively small in case of tests carried out with less 

uncertain factor such as compression test with simple 

figure; on the other hand, it is somewhat larger in the tests 

with some degree of uncertainty involved such as complex 

injury evaluation tests. 

Comparisons between FEM and test re$ults 

Test vehicle 

Using a production vehicle as a basis, it is possible to 

check the FEM analysis performance and compare test results 

against model results. Test vehicle is a body in white 

equipped with suspension to move on in testing' Table 4 

describes the test conditions. 

Table 4. Test condltlons. 

D 685 140 0.20 

P 70t 8l 0.12 

D 4l 3.2 0.08 

P 33 2.r 0.06 

D 440 218 0.50 

P 579 247 0.43 

Vehicle Type 4 doors small sedan 

Vchicle Weight 470 kg 

Impact Velocity 30 mph (48km/h) 

Impact Form Fixed Varrier

FEM analy$is and model 

It is necessary to include the following items for analyzing 

crashworthiness 

of vehicle: 

( l) geometric non-linearity, 

(2) material non-linearity, 

(3) elasto-plasric rhin plate/shell, 

(4) non-linear spring element, 

(5) progressive rivet failure with structural collapse, 

(6) a slide line algorithm which prevents penetration 

of internal structure surface. 

Irt case of transient response problems such as crash, it is 

al$o necessary to solve the motion equation with time int€gration. 

Explicit integration is preferable for solving crash 

simulation problems compared implicit integration, Furthermore 

the dependence of strain rate of material is to be 

considered. 

One program is selected of a few crash simulation program 

codes vectorized for supercomputer (2). 

Next, FEM model in question is explained as follows. 

The body panel such as frame, outer panel, floor and roof are 

modeled with thin shell element to capture the impact buckling 

modes. In addition, bar element are used to approximate 

the suspension unit and wheel by providing equivalent 

horizontal and vertical stiffness for these components. A 

part of flange of panel connected with spot welding is modeled 

with two plate thickness. The weight of tires is anributed 

to the suspension model with lumped mass. Rigid walls 

may be defined which provide extemal impact surfaces, 

while a slide line algorithm prevent$ penetration of internal 

$tructure surfaces that may collide during the crash simulation. 

Slide line interface conditions are available for frames. 

In deciding mesh element of FEM analysis, they must be 

decided according to purposes, time and accuracy of calculation, 

and restraint conditions in the program itself. Especially 

in the explicit method, the solution advances through 

the impact duration using a small time step ba$ed on the 

velocity of sound (or wave propagation time) across the 

smallest mesh element. Thi$ ensures that the physical phenomena 

including stress wave effects are completely followed 

qnd that solution is stable. The velocitv of sound of 

steel Co is Co = [f/p. es is Co = 5 X 106 mm/sec, rhere fore 

the size of timestep is less than I (F seconds. In view of the 

above the optimum size of mesh element is 20 mm X 20 

mm. The number of mesh elements is about 9000 because 

the finer mesh contributes to deformed area. while the 

coarser mesh does to the undeformed zone such as rear body 

structure. This is why mesh becomes more coarse toward 

the rear of the vehicle. 

Comparisons between calculation and test 

results 

The correlation of the simulation against experimental 

results is a key point of the calculation. Specific items under 

consideration include: 

FI 

vl 

q' 

tr 

tr 

fit 

q.) 

q,J 

(J 

0) 

; 

q 

ô 

z?, 

tr4.- 

q) 

(J 

o 

IJ. 

L} 

(q 

r 

tr 

q 

0.0 

0.0 

calculation 

time (second) 

calculation 

time (second) 

0.3t) 

o'30 

Flgure L l{odal decclsrstlon versus fimo curyo and lmpact 

Ioad Y€rEUS 

tlme. 

. to verify the accuracy of the simulation with regard 

to the impact force and deceleration pulse. 

. to study collision modes of each structural member, 

critical member being front side member and 

passenger floor. 

The numerical and experimental deceleration and impact 

force illustrated for a 30 mph impact into a fixed barrier in 

firgure [. The important first 0.30 seconds are calculated and 

are shown in the figure. The first peak is a little delayed in 

the analytical model, but as shown in the figure, the result of 

calculation shows reasonable correlation with the experimental 

ones. Figure 2 shows a FEM model and deformation 

mode of calculation. A good correlation between the calcu- 

3?9

lated and experimental results is shown as in figure 2. 

Thus the FEM has been proved to be one of the valuable 

tools in predicting the crashwot"thiness of body structure 

with the aid of high speed computer. 

Flgure 2. FEM model snd doformatlon mode. 

Evaluation of Crashworthiness and 

Verification by Testing the Prototype 

Vehicle 

Evaluation of crashworthiness 

It is imponant for a safety vehicle how crashworthiness 

and occupant protection characteristics are developed and 

incorporated in vehicle designs. There are targets for 

occupant protection for example HIC in FMVSS No. 208. 

However, targets for body structure to construct to a safety 

vehicle cannot be found in previous papers. In order to 

determine the crashworthiness requirements of a new car 

design, two dependent variations are offered: one is 

deceleration Gc and Ge observed on the side-sill around 

peripheral front seat, and other decelerance Dc. 

Body deceleration Ge and Gc 

Figure 3 shows a definition of deceleration Gc and Ge. 

Here Ge is defined as average crash pulse in initial deformation 

and Gc is defined as average one in laterdeformation of 

vehicle. If deformation is assumed that occurs progressively, 

it is found that Ge is a crash pulse of engine 

compartment of vehicle and Gc is a one of the rear frame 

around the peripheral dash board and passenger 

compartment. 

Furthermore, a merit of this figure in presenting a deceleration 

deformation curve as shown in figure 3 is to 

estimate the distribution of energy absorption. 

Total crash energy E is denoted as follows: 

E=GeXt+GcXr (l) 

330 

cl 

0) 

tu 

o 

{) 

A 

Flgure 3. Deflnltlon of Ge and Gc. 

Thus Ge and Gc is related to crashworthiness, progression 

ofcrash and occupant injury as mentioned later' In 

addition, the fact that body deformation occurs progressively 

can be proved by a simple mass-spring simulation. 

The simulation results are shown in figure 4. From 

figure, a condition is found as follows: 

Kr/K2

x 

o 

g 

A 

SOhDh 

l - t 

.jtttt -_xlJ.ph,-\ 

\ \\ 

\..- 

r_-r 

-- \q-a- I 

- 2.8nph 

----- 

o- 

Body Cofigunrim 

Flguro 5. ,Dsceleratlon at varlous Impact velocltles and body 

conllguratlon. 

Verification by testing the prototype vehicle 

Various analyses are studied by considering Gc, Ge and 

Dc from FEM in order to obtain a high quality of crashworthiness. 

Figure 6 shows the target verified by testing the 

prototype vehicle. In checking the occupant injury twodimensional 

simulation is carried out (4). Figure 7 shows 

the results of calculation. As the figure 7 shows, occupant 

injury is unchanged. Figure I illustrates the sketch of the 

prototype body structure. Countermeasured components 

are the strengthened side member, the additional perimeter 

frames around peripheral wheel housing, the strengthened 

floor and side-sill, and the formed double floors. Total 

weight of this vehicle is not increased by adjusting the 

stringer onto the floor panel. 

Test Results 

Static crash te$t 

In order to confirrm a deceleration Gc and Ce of the 

prototype vehicle a static cra$h test is carried out as shown 

in figure 9. Figure 9 (a) shows rhe outline of rhe test 

procedure for determining the frontal buckling load and 

figure 9 (b) shows the outline of the test procedure for rear 

Ll 

o t.0 

.E 

d 

E 

tr 0.9 

0.8 

currcnl model 

Flgure 7. CElculatlon resulta of occupant Inlury. 

I 

5 

E 

{) 

t) 

o 

d 

Body Dcformation 

Flgure 6. Target verllled by te3tlng the prototypg vehlcle. 

Double Floor, 

Strengthened Side Member 

.Perimcicr Frame 

, .Strengthened Side-sill 

Flgure 8. Sketch ofth€ prototype body structure. 

Ldad Ccll 

/ 

t *,1--1 

AU -/ll)_ry/ I AV, 

U - U 

Flgure 9. Test method. 

- r-o 

Body Deformation Body Deformation 

0 

U 

0.9

part buckling load of body structure. In case of (b), the 

suspension part is incorporated so as to be similar to an 

actual vehicle. Figures l0 and I I show the test results. As 

shown in those figures, static buckling load has a tendency 

to increase. 

* 

b4 

G 

.g 

6 

I 

Flgurc 10. 

memDer. 

^f 

E{ 

.g 

d 

6 

100 200 

Deformation of Side Member (mm) 

Statlc compregelon test r€sults of thc lront $ldE 

50 100 150 

Deformation of Rear Side Member (mm) 

(Passenger Compartment) 

Flgure 11. Statlc compresclon t6$t resultE of the rear slde 

member. 

35mph barrier impact test 

Figure 12 shows the relationship between deceleration 

and deformation of vehicle in the 35 mph barrier impact test' 

Figure 13 illustrates the experimental results of deceleration 

Dc. As the figure shows, Ge and Gc increase, and Dc also 

increases. 

J9 

E 

o 

E 

0 

0 

o 

d 

Body Deformation (mm) 

Flgure 12. Relationshlp b€twocn doceleratlon and deformatlon 

In 35mph. 

332 

Model 

Flgure 13. D€celeratlon of the experimentsl results. 

ln addition, vibration phenomenon considered as one of 

the vehicle functions is investigated before the crash test. 

Figure 14 shows the relationship between response and 

frequency. In comparison with current body structure, its 

frequency increases. 

bo 

!4 

g 

Flgure 14. Relatlon betwsen lcsponse and frequency. 

From the above mentioned results, it is found that factors 

Ge. Gc and Dc come to be variable to estimate crashworthiness 

of body structure. It is also found that FEM analysis is a 

very useful tool to predict the crashwofihiness of body 

$tructure in an early design stage. 

Conclusion 

Deceleration 

Dc 

Cunent Model I 

Prototype Vehicle l.l 

(J 

- 

,F 0'l 

,4) 

6 

F 

0.01 

By introducing Ge, Gc and Dc for estimating the 

crashworthiness of body structure, the optimal design rule 

for safety vehicle which has been ambiguous is discussed. 

By using the FEM program of vectorized supercomputers 

on a commercial scale, complex and varied impact 

phenomena are clarified. 

(1) Variation of estimating the crashworthiness exists. 

(2) FEM analysis is very useful to predict the crashworthiness 

of body structure in impact. 

(3) Safety body designed from FEM analysis and testing 

a prototype vehicle is confirmed. 

The authors intend to continue their research into safety 

vehicle in the future.

References 

(l) T. Sakurai, M. Takagi, Nonlinear Structure Analysis 

on Vehicle Development-Present Status and Future 

Trends, J. JSAE (in Japanese), Vol. 43, No. l, 1 989. 

Identification of Safety Problems for Restrained Passengers in Frontal Impacts 

Daniel S. Cohen, 

Office of Crashworthiness Research, 

National Highway Traffic Safety 

Administration, 

U.S. Department of Transportation 

Lawrence F. Simeone, 

Transportation Systems Center, 

Research and Special Programs Administration, 


Donald J. Crane, 

MGA Research Corporation 

Abstract 

The objective ofthe frontal crashworthiness research and 

development effort at the National Highway Traffic Safery 

Administration (NHTSA) is to assess the safety problems 

associated with occupants of passenger cars involved in 

frontal impacts and to identify potential remedies for the 

problem. The focus of this paper is the identification of the 

safety problem for restrained right, front seat passengers. 

NHTSA's National Accident Sampling System files and 

individual state files contained in the Crashworthiness State 

Database were utilized to identify the crash environment, 

vehicle factors, and injury mechanisms associated with 

restrained pa$sengers. The New Car Assessment Program 

provided data on over 200 full scale crash tests containing a 

restrained right, front seat dummy. Analyses utilizing the 

accident and crash test data are presented in this paper. 

A laboratory test program has been conducted to 

characterize head to instrument panel contacts which 

appears to be a major injury mechanism. The material 

properties and geometric characteristics of the upper 

instrument panel have been collected and analyzed for 

several of the production vehicles tested as part of the New 

Car Assessment Program. Dynamic and static forcedeflection 

data was collected utilizing a head component 

impactor. Results of thi$ laboratory prografi are presented. 

The MVMA 2-D computer model is being utilized to 

simulate the associated New Car Assessment Program crash 

tests to assist in the identification of mitigation concepts. 

The paper discusses the input data requirements, the 

validation process, and results of initial simulations. 

(2) USER's Manual 

"PAM-CRASH". ESI. 

(3) T. Sakurai, Y. Kamada, Structural Joint Stiffness of 

Automotive Body, SAE paper 880550, 1988. 

(4) H. Katoh, R. Nakahama, A Study on the Ride-Down 

Evaluation. The Ninth ESV. 198?. 

Introduction-Passenger Safety 

Assessment 

The objective ofthe frontal crashworthiness research and 

development effort at the National Highway Traffic Safety 

Administration (NHTSA) is to assess the safety problems 

associated with occupants of passenger cars involved in 

frontal impacts and to identify potential remedies for the 

problem. The focus of this paper is the identification of the 

safety problem for re$trained right, front seat passengers. 

Insight into the safety problem for the restrained right, 

front seat passenger can be gained by examining available 

field accident data and crash test data. TWo accident data 

files were utilized within this study-the National Accident 

Sampling System (NASS) and the Crashworthiness Srare 

Database (CWSD). NASS (l)* files contain a nationally 

representative sample of crashes occurring within the 

United States. The data contained in the files are collected 

by teams located throughout the country. Detailed 

information on the vehicle, occupant, and crash factors are 

obtained from on-site inspections, vehicle examination, 

occupant interviews, and police and medical records. The 

CWSD (2) has been developed by NHTSA to support 

research programs aimed at improving vehicle 

crashworthiness. The database contains information 

extracted from several State files and the data contained in 

the files is primarily obtained from police investigations. It 

is being used to supplement the information contained in 

NASS. 

In addition to field accident analyses, the crash tests 

performed as part of the New Car Assessment Program 

(NCAP) have been examined to help in the definition of the 

safety problem. NHTSA has conducted more than 250 crash 

tests under this program since 1979. Vehicles in this 

program are subjected to a full frontal impact against a flat, 

stationary barrier at 35 miles per hour. This is 5 mph faster 

than the prescribed speed for compliance with Federal 

motor vehicle safety standards. The crash tests are designed 

to indicate, for vehicles within the same size class. the 

relative levels of occupant protection and vehicle safety 

performance. Each vehicle contains a restrained driver and 

passenger side dummy. Data is collected on structural and 

dummy responses during the crash and stored at NHTSA in 

a computerized data base (3). The analyses related to the 

*Numbers in parenlhcses dcrignatc rrfercnce$ ar rnd of paprr.

field accident data and the crash tests are presented in the 

next section titled "Problem Determination". 

Based on these analyses, it was decided to focus 

particular attention on head injuries caused by contact with 

the instrument panel. The material proper"ties and geometric 

characteristics of the upper instrument panel have been 

collected and analyzed for several of the production 

vehicles tested as part of the New Car Assessment Program. 

Dynamic and static force-deflection data were collected 

utilizing a head component impactor. Injury criteria data 

obtained from these component tests were compared to the 

full scale NCAP crash tests. This work is discussed in the 

section titled "Laboratory Testing for Head to Instrument 

Panel Impacts". 

The MVMA ?D Crash Victim Simulator is being utilized in 

this effort to help characterize the baseline conditions of a 

restrained front seat passenger. In future work this model 

will be utilized to assist in postulating and analytically 

evaluating mitigation concepts as to their effect on harm 

reduction. Input data issues and requirements related to the 

occupant characterization, instrument panel material 

properties, restraint system, and geometry are discussed. 

lnitial validation efforts and sensitivity analyses are 

presented. This work is contained in the section titled 

"Analytical 

Characterization of Baseline Vehicles". 

Problem Determination 

The following sections present analyses conducted using 

accident data and crash test data to better define the safety 

problem for restrained right, front seat passengers in frontal 

impacts. 

Field accident dflta 

Both NASS and CWSD accident data were utilized in 

performing the analyses presented in this section. The 

NASS files utilized included the calendar year 1979 to 1987 

individual files. Comparisons are made by the distribution 

of both "harm" and injuries. NASS utilizes a national expansion 

factor to make the NASS files representative of the 

population of accidents which occur in the United States. 

These factors are available for the 1979-86 NASS files, but 

not for the 1987 NASS file. Because of this difference and 

the limited sample size available for restrained passengers----€specially 

at the higher injury severity levels, 

both weighted and unweighted data are presented. The 

counts ofinjuries that are used in the comparisons presented 

in this section are based on a count ofall injuries received by 

the passenger. ln the NASS file, up to six individual injuries 

for each injured occupant may be coded into the 

computerized file, and it is all these injuries that are included 

in these analyses. 

Additional comparisons are made utilizing the concept of 

"harm" 

as originally defined by Malliaris, et al. (4,5). To 

obtain "harm", a harm weighting factor that is associated 

with each of the AIS injury severity levels is utilized. The 

harm factors are applied to all the injuries received by the 

passenger. The harm weight factors that were utilized are 

334 

based on the information included in reference 5. 

INJURY AIS 6 HARM 265 

LEVEL AIS 5 WEIGHTING 228 

AIS 4 FACTORS 62.8 

Ars 3 12.6 

AIS 2 3.8 

AIS I ,45 

Other weighting concepts that attempt to account for the 

consequences ofdifferent types and severities of injuries are 

being evaluated by NHTSA (6). 

The CWSD files were also utilized. The analyses were 

limited to the use of the Pennsylvania 1983-86 state files. 

The distributions are based on the police recorded body 

region which received the most severe injury. Counts are 

based on injuries that were considered either incapacitating 

or fatal. 

Figure I presents the distribution of injuries and harm to 

restrained passengers by body region injuried based on the 

O 

z 

U a n 

-) 

Lv 

u 

L 

HEAD FACE 

1 979-86 

WEIGHIED 

1979-87 

UNWEIGHTED 

1 979-86 

HARM 

NECK CHES 

BODY REGIONS 

Flgure 1. Restralned passengers In frontal lmpacts. Dlstrlbutlon 

of body reglona lnlured. NAS$, AI$ 2+ Inlurles. 

HEAD 

CWSD*PA, 

FACE NECK TRUNK 

BODY REGIONS 

LOW EX 

Figure 2. Reetralned passengers in frontal impact$. Distribution 

of body regions iniured. Crashworthiness State 

databasFPennsylvsnls f lles. Incapacltatlng and fatal Inlurles.

NASS files. Three comparisons are presented including 

distributions based on 1979-86 AIS 2+ injury weighted 

counts, 1979-87 AtS 2+ unweighted counts, and 1979-86 

harm weighted data. Head injuries account for l5 to l9Vo of 

the AIS 2+ total injuries for both the weighted and unweighted 

injury counts. For the harm weighted distribution, 

the head accounts for 25Vo of the total harm. Figure ? presents 

body region distributions based on the CWSD files. 

The head body region accounts for 25Vo of the injuries 

classified as incapacitating or fatal. 

NASS contains information as to the injury source for the 

body regions injured. Figure 3 presents the distribution of 

injury sources for head injuries. Because of the small sample 

size, only the analysis using unweighted data for the 

1979*87 file is presented. There is information on injury 

source for 2l injuries at the AIS 2+ level. Nine of the head 

injuries were attributed to contacting the windshield, five to 

the instrument panel, and 3 to the roof rails/pillars. 

(J30 

z 

:) 

o 

H20 

LL 

WDSHLD INST PNL PILLARS NON CT 

INJURY 

SOURCE 

(SAMPLE SIZE:21 INJURIES) 

OTHER 

'1979-87 NASS 

UNWTIGHTED 

Flgure 3. Hsstralned pss8angsrs ln frontal lmpacte. Inlury 

aources for head lmpacls. 197F87 NASS. Unwelghted. AI$ ll 

rnlur|€s. 

Crash test data 

The following analyses are based on the NCAP crash 

tests that are contained in NHTSA's vehicle crash test 

database. Each of the vehicles tested in this program was 

crashed into a rigid barrierat a test speed of 35 mph. This is 5 

mph faster than the prescribed speed for compliance with 

Federal motor vehicle safety standards. The crash tests are 

designed to indicate, for vehicles within the same size class, 

the relative levels ofoccupant protection and vehicle safety 

performance. 

Specific attention was directed at the association of the 

HIC injury criteria values with contact source. The following 

figures and tables are based on the passenger car tests 

and the restrained passenger dummy head injury criteria 

(HIC) and chest injury criteria (chest acceleration). As can 

be noted in table l, the average HIC value is approximately 

double (l174 vs. 602) for those dummies that contact the 

instrument panel compared to those receiving no contact. 

The chest acceleration levels for contact cases are also higher. 

The "other" contact category primarily consist of head 

to knee contacts. It can also be noted that contact with the 

instrument panel occurs in almost one-half of the crash 

tests. 

Table 1. Comparison ol restrained pe$$6nger lnlury crlterla by 

head contact bources. NCAP teet rdsulte. Faseehg6r cara only. 

Contact Source Sample Size l{IC Chest 

Acce l eratlon 

Instrument Panel ll9 rtf4 46 

None 76 602 3g 

Other 22 ll25 43 

Tsble 2. ComDarlaon of rsrtrElnsd Daassnser lllC levels dlstrF 

butlon by hedd contact source. NCAP test-r€sulta. Pasaenger 

csrs onty. 

HIC Level s 

0 to 500 

501 to 1000 

l00l to 1500 

l50l + 

Total 

Contact Sources 

Inst Pnl l{one Other 

3r. (4) 3614 (27) 9r( (?) 

38% (45) 6l* (46) 45x (10) 

38% (45) 4% (3) tSx (3) 

2r% (zs't 0y, (0) 3lx (7) 

100% (lre) 100% (76) t00%.(e2) 

Table 2 presents the distribution of different levels of 

HIC. The table shows thtt 59?o of the dummies that contacted 

the instrument panel received HIC readings equal to 

or greater than a HIC value of lfi)O, whereas for those 

dummies that did not contact any interior surface that number 

is only 47o. 

Figure 4 presents another illustration ofthis contact problem. 

The figure presents the HIC and chest acceleration 

readings for each dummy by contact source. The symbol 

"1" 

indicates instrument panel contact, 

"N" 

indicates no 

contact, and "O" indicates other contact. As can be seen 

from the figure, almost all the crashes that contain no passenger 

contact with the interior surface result in HIC readings 

lower than l0(X). 

Tables 3 and 4 present similar information for the light 

trucks tested as part of the NCAP program. The HIC and 

contact source trends are the same as those for passenger 

cars. 

Laboratory Testing for Head to 

Instrument Panel Impacts 

Based on the results ofthe field accident and the crash test 

analyses, it was decided to focus specific attention on the 

safety problem of head to instrument panel impacts. Head to 

windshield contacts appeared as a problem in the accident 

data but not in the full scale crash tests. This discrepancy is 

continuing to be investigated as part of the problem 

identification effort. The analytical simulation work that is 

presented in a subsequent section presents initial results 

related to this question. 

335

t{l cF 

300u 

o + II 

PUIT at HICFTC6F SYrtEd. lS VALiE OF CuillAct 

o l II 

II I 

l l 

I I 

I 

I r l 

I T TII 

t o t l 

0r I I ll 

I O I 

o l 

I 0 t r 

It I I 

t oll t I 

I I NI II II I 

I l r l 

I tt I o I I 

t{ | I I I tt I I I 

ir o r r I I I It I l{ 

III I IIII O 

I I IIIT II N I{ 

II I t{ I 

t{ tfitt{It{ Nil{t H H N 

l{ ril r rJ$fio H 

}{I tf.ll O ttf{t I l{ 

il t{ l{tl{l{ l{ t t{ rfiril o 

HH ]\t{ I{T ilI 

t { i l H 

- --+- --+-+*{+r*+-----------+- +--+r-----+-..--- ------+----r ------+- ------*-*_+*- --* -_+-+-++_++ 

o lo eo go /to 5(' a{} To Ht ett too 

Flgure 4. Plot of HIC v$. chGst acceleratlon for head contacl8 wlth the lnstrument psnel (l), no contact (N), and other contact (O). 

Table 3. Comparlaon of reatralned psssengsr Iniury critcria by 

head contact sourcec. NCAP t€st results. Llght trucks and vans 

only. 

Contact Source Sample Size HIC Chest 

Accel eratl on 

None 9 671 44 

0ther ll lt89 50 

Table 4. Comparlson of reetrained pageenger HIC levels 

dbtrlbutlon by hesd contsct tourc€. NCAP tcst results. Light 

trucks and vais only. 

HIC Level s 

Contact Source$ 

0 to 500 

501 to 1000 

l00l to 1500 

l50l + 

Tota l 

Instrument Panel e0 

l25r 48 

Inst Pnl None Other 

104 (a) 22s (?') 0* (0) 

20* (4) 67* (61 ?7% (3) 

45% (e) ll% (r) 45% (5) 

?5% (5) o% (0) 

loo# (20) 100% (e) 

38ff (3) 

r00* (lr) 

A laboratory testing program was initiated, which had as 

its primary objective the simulation of passenger head 

injury as a result of impacts with the upper dash panel of 

several NCAP passenger cars and light trucks. The 

336 

simulation was accomplished by the use of static and 

dynamic component level head te$ts. The data collected 

serves as first, a baseline data base for characterizing the 

material properties/geometry of instrument panels, and 

second, input data sets for analytical simulation models. 

Summary of laboratory tasks 

MGA Research Corporation, under contract with the 

Transportation Systems Center (TSC), United States Department 

of Transportation, performed the laboratory testing 

program involving the component level head tests of the 

passenger side upper instrument panel (7). MCA has developed 

both the equipment and procedures to conduct static or 

dynamic component level tests of a vehicle interior (8). The 

investigation included three groups of vehicles as listed in 

Table 5. The damaged NCAP vehicles were the vehicles 

used to perform the actual NCAP tests and were acquired 

under the contract with TSC. The group I vehicles matched 

the damaged vehicles but contained undamaged interior 

components. The group 2 vehicles were other undamaged 

vehicles that matched other vehicles tested within the 

NCAP program. They were purchased as vehicle front ends 

or cowl sections and contained the same options as the 

original NCAP vehicle. The light trucks were vehicles used 

by MGA to perform other interior component level tests for 

TSC.

The initial task in the project wa$ to perform analysis on 

the NCAP crash films to determine the passenger head 

impact velocity, impact angle and contact point. This was 

then followed by a test matrix that included three dynamic 

test types and one static test type. The test types were designated 

as dynamic generic, NCAP simulation, FMVSS 201 

and a static generic. The dynamic generic was a 25 mph 

impact at a fixed 50 degree impact angle measured from 

horizontal, using a 6 inch diameter l5 pound hemisphere 

shaped headform. This impact angle was determined from 

the average of the nine impact angles found in the film 

analysis. The analog to the dynamic generic was the static 

generic, performed at a quasi-static rate of 2 inches/minute. 

Both test types were performed on the group l, group 2, and 

light trucks. 

The NCAP simulation, performed on the damaged 

NCAP, group I and group 2 vehicles, made use of a Parr 572 

anthropomorphic head ballasted to l5 pounds. The impact 

velocity and angle were determined by the film analysis 

results. The FMVSS 201 test, a l5 mph impact using a 6 inch 

diameter hemisphere shaped headform ballasted to l5 

pounds, was performed on the 1985 Dodge Colt, 1985 Ford 

Tempo, 1986 Hyundai Excel and the five light trucks. 

Test procedures 

and data acquisition 

This section will review the test procedures used to perform 

the film analysis and component level head tests along 

with the data acquisition. The purpose of the film analysis 

was to determine the passenger head impact velocity, impact 

angle and contact location, and was performed on each 

of the nine group I and group 2 vehicles listed in table 5. 

Using a film digitizer and computer, the passenger head 

target motion was digitized, based on time, with respect to 

the vehicle. By knowing the film speed, the head target x 

Table 5. Vehlcles tested. 

hged 

lmP 

t/ehlclee 

gtup I 

ard 

ef,p-2 

Vdrlclg 

r.t#t 

ItrrclrE 

1986 tddc ErbJElf 

1986 ltyrdai EGI 

1986 lgrzu l{.tar|q 

1986 l{emrry Sable 

1986 Saab 90OO 

1986 tulck e*ry 

1985 Cbevmlert, ftrfrit 

1985 Dodgr SIt 

1985 rud Teq'o 

1986 Hytndat bdl 

1986 Isrzu l{taik 

1986 ilnz(b B-2(X)O Ptclop 

1986 letury Sable 

1986 Saab 90OO 

1985 CtEvrrl€t G-10 Van 

f986 Dodge Caratrarl 

1985 IbLd F-150 

1986 IloLd RanEer 

1985 NLssan Ptdq'p 

and z motion (where x is positive forward and z is positive 

upward on the vehicle) versus time was recorded. 

The head displacement was determined from the time of 

vehicle contact with the barrier up to the time of head 

contact with the upper dash. The head x motion versus time 

and head z motion versus time were then manipulated to 

determine the head impact angle and velocity. The velocity 

was determined by ploning the resultant motion versus time 

and finding the slope of that line in a region 20 msec prior to 

head contact. The impact angle was determined by crossplotting 

the head x motion with the z motion and calculating 

the downward angle with respect to the horizontal in a 

region just prior to contact. A summary of the film analysis 

results are listed in table 6 and a sketch of the impact velocity 

and location are shown in figure 5. 

Table 6. Summary of the fllm analysle r€Bults. 

Vehicte r'T]m+ Vcfefty (rFh) q)art Atqf€ (.leg) 

1986 tulc* €firy 17 

1985 CtffiIet SFrjrt 23 

1985 hdge SIt t5 

rgSE lbtd Tryo ZF.1 

1986 n$rdat EeI 28 

1986 IsrElr IJ.hrlr 22 

1986 rhrda Picld+] 26 

1986 l{emry Sablc L7.4 

1986 S|rb 9000 23 

IMPACT \ELOCITY 

IMPACT ANGLE 

INSTRUMENT 

PANEL 

Flgure 5. Pareenger head lmpact angle and veloclty deflnltlon. 

Dynamic and static interior components tests were performed 

using MGA's dynamic/static load frame on which a 

vehicle cowl section is fixtured as shown in figure 6. The 

vehicle cowl section is that portion of the vehicle cut forward 

of the shock towers and through the front seat floorpan, 

and fixtured using a combination of welding and bolting. 

This methodology has been found to be a very rigid 

system simulating the conditions found in the full vehicle. 

Each vehicle cowl section was purchased with the same 

options as the NCAP vehicle and included all components 

behind the instrument panel. An undamaged instrumenl 

panel was replaced after each test and installed identical to 

the original condition. The A-frame supports either a hydraulic 

cylinder for static testing or an impactor for dynamic 

testing. The cylinder or impactor can be oriented in any 

direction, through lateral motion, vertical motion and rotation 

in two directions. 

The static testing was performed using a hydraulic cylinder 

mounted to the A-frame as shown in figure 7. Attached 

36 

49 

44 

49 

49 

70 

66 

4E 

54 

337

Flgure 6. MGA statlc/dynamlc load frame. 

Flgure 7. Slatlc gsnerlc test Bet-up. 

to the end of the cylinder was a 6 inch diameter hemisphere 

shaped 201 headform with a load cell mounted behind it to 

record the force. A displacement potentiometer attached to 

the cylinder was used to measure the instrument panel deflection. 

The cylinder operation and data acquisition was 

fully computercontrolled using the method of displacement 

feedback. Force/deflection pairs were recorded every 0.040 

inches as the cylinder was extended and retracted at a rate of 

2 inches/minute. An example static generic force/deflection 

result is shown figure 8. The contact point for the test was 

338 

F 

o 

R 

c 

B 

( 

P 

o 

u 

ll 

D 

s ) 

ETATIC O8I{BRIC PAEEE}IGIBR EIDB UPPER Iil8TRU}IENT PAIIBL 

O fORCB/DEFI,ECTIOII 

o.oo l.oo 8.oo g.oo 4,oo E,o0 €.o0 ?.oo 

DEFIECTION (TNCHE$) 

Flgure 8. Statlc generlc force/deflectlon plot. 

determined by the film analysis or in the case of the light 

trucks was at the junction point of the upper and middle 

instrument panel. After the test was completed, the data was 

saved on a diskette, loaded to MGA's Micro VAX and converted 

to the NHTSA component data base format. All the 

data was then sent to NHTSA and is available on their VAX. 

As mentioned earlier, three types of dynamic te$t$ were 

performed, the dynamic generic, NCAP simulation, and 

FMVSS 201. A similar test procedure was used for each test 

type and only differed by the impact velocity, impact angle, 

contact location, and impact form. Figure 9 shows the dynamic 

test set-up. Strain gage type accelerometers were 

mounted in each headform to record the acceleration in the 

impacting direction. A displacement potentiometer attached 

to the back of the headform was used to measure the 

instrument panel deflection. A contact switch placed on the 

instrument panel recorded the time of impact. 

Flgure 9. Passenger slde dynamic head tBet set.up. 

The impactor is essentially a guided piston, with both the 

impacting sequence and data recording controlled through a 

computer. Figure l0 shows the back of the impactor and the 

computer, system controller, pneumatics cart and signal

Flgure 10. Dynamlc ta$tlng data acqulsltlon. 

conditioning used to execute the test. The impactor is fired 

by pressurizing nitrogen in an accumulator to a pressure 

corresponding to the desired velocity based on the impacting 

mass. Once pressurized, a safety pin is released and a 

fire button is pressed, discharging the nitrogen to the back of 

the piston. 

A 

c 

E 

L 

E 

R 

A 

I 

o 

N 

( 

o 

s ) 

F 

o 

R 

c 

E 

( 

F 

o 

u 

N 

D 

I 

) 

ro, o 

o. 

-lo. o 

-ao. o 

-s0.0 

-+0. o 

-80. o 

r 

I 

"tt- 

VERSUS TIHE 

-00. o 

o,0000 0.o8oo 0, rooo Q, r800 0.aooo o.86(10 

rIMB (FEGSHDS) 

Flgure 11. Dynamlc lert Ecceleratlon versus tlms. 

rooo 

800 

soo 

400 

soo 

o 

o,ooo o.Boo r,ooo I,Eoo s.o00 s,€oo 8.ooo 

DI8PI.ACBHEflT ( IilCEE8 ) 

Flgure 12. Dynsmlc force/deflectlon curya. 

Once the piston has stopped accelerating, the nitrogen is 

vented to the atmosphere, allowing the piston to travel ar a 

constant velocity. Immediately after venting, the data recording 

was triggered and velocity measured by a wand 

attached to the piston, passing through a light trap, The 

acceleration, displacement and contact were all recorded on 

a time basis according to the SAE J2l I standards. An example 

of an acceleration time curve is shown in figure I l, and 

through file manipulations can be reduced to the dynamic 

force/deflection curve as shown in figure 12. As with the 

static data, all the dynamic data was converted to the 

NHTSA component data base format and installed on the 

NHTSA data base. 

Results 

This section will discuss the results from the film analysis, 

dynamic tests, static tests and their relationship to the 

NCAP results. Table 7 provides a summary of the passenger 

side HIC based on the head resultant acceleration for each of 

the nine NCAP vehicles as determined by a 36 msec HIC 

program. These are the results used to correlate with the 

component level test results. 

Tsble 7. Summary ol NGAP passenger HtC. 

Tested Vehicle 

L985 Ford Telrpo 

1-986 llyundai B{ceI 

1986 Isuzu l+Iark 

l-985 l{azda Pidfl,p 

L986 l,Ienerry Sable 

1986 Saab 9OO0 

Hrc (t2-il) 

L449 

2662 

999 

1647 

597 

L443 

(24) 

(13) 

(36) 

(35) 

(35) 

(5) 

A comparison was made between the head impact velocity 

as determined through the film analysis and the NCAP 

resultant passenger HIC. A correlation plot comparing the 

two is shown in figure 13. The correlation coefficient, r, 

I 

o_ 

(J 

O 

L,J 

F 

{ 

L 

E 

O 

UJ 

r 

JO 

20 

10 

1000 2000 3000 

NCAP RESULTANT HIC 

Fasultint Hlc veraua 

fub""ii"]:. head lmpacl veloclty corrsla' 

339

determined for this plot was 0.86. These results indicate that 

as the passenger head velocity increases, the likelihood of 

an increase in head injury will result. Therefore, one approach 

to reducing the occupant head injury is to reduce the 

head velocity. This can be accomplished by modification of 

structural characteristics (crash pulse and intrusion), interior 

geometry, and restraint effectiveness. However, as discussed 

in the problem determination section to this paper, 

prevention of all contacts is not feasible and countermeasures 

must also focus on the properties of the instrument 

panel itself in mitigating contact forces. 

Two types of loading responses resulted from the static 

generic test type as shown by the overplot in figure 14. In 

one case, the trend shows a gradual rise in force to a peak 

load, which is characteristic of the softer instrument panels. 

This type of response was found in instrument panels that 

had approximately 0.54.75 inches of cushion covered by a 

vinyl type of material. The other trend shows a response that 

is much more steep and is characteristic of the stiffer instrument 

panels, constructed of hard plastic with little or no 

cushion material. The advantage of a softer instrument panel 

was reflected in the results obtained from the Mercury 

Sable. The combination of a soft, thick cushioned instrument 

panel and an impact velocity of l7 .4 mph resulted in 

the lowest NCAP HIC of the group at 597 over 36 msec. 

F 

0 

F 

E 

STATIC GENERIC FORCE/DEFLECTION COMPARISON 

O 86 ISUZU I-MABK 

x 8€ lilERCttRY SABLE 

0.00 l.oo a.o0 9.00 4.oo 5.oo 

DEFI,BTION 

(INCHES) 

Flgure 14. $tatlc generlc force/deflectlon reapons€a. 

Although the static tests can discriminate among the vari* 

ous types of instrument panel stiffnesses, it lacks the dynamic 

response, which is critical in defining the instrument 

panel impact properties and as input into the occupant sim* 

ulation models. An overplot comparison of a static generic 

and dynamic generic force/deflection from the same vehicle 

is shown in figure 15. The dynamic generic test show$ a 

large inertial spike occurring within the first two inches of 

deflection, and has a steeper initial slope as compared to the 

static test. 

Two types of responses from the dynamic tests are shown 

in the overplot of figure 16. In one case, the trend was a 

gradual rise to a peak force while the other shows an inertial 

340 

?00 

eo0 

I Eoo 

[ 

fl 

D 

s 

) 

+oo 

soo 

eoo 

roo 

t.] 

{\, d 

j 

-# fi 

f 

F 

0 

n 

c E 

( 

o 

U N 

D 

s ) 

Flgure 

F 

o 

c 

E 

( 

p 

o 

U 

N 

s ) 

o 

IS86 SAAB 9OOO _ FORCE/DEFLECTION COHPARISON 

DYNAI.TIC GENERIC 

STATIC CENEBIC 

I 600 

leoo 

soo 

600 

soo 

o 

o.00 r.oo e.oo 3.oo 4.o0 6.00 6.00 7.00 

DEFLECTION (INCHES) 

15. $tatic and dynamlc force/deflectlon comperlson. 

DYNAMIC GENERIC FORCE/DEFLECTION 

1986 $AAB 9000 

1S85 FORD TEMPO 

8000 

4600 

4000 

I EOO 

I OOO 

600 

0 

o. oo I .00 a.00 3. oo 4. oo 3. oo 6. oo 

Frgure 18. Dynamic ;;"#;-l 

I 

l \ ,nhn *- \r- 

N {'\/ {* 

spike within the first few inches of deflection. The trend 

most often seen was that with the inertial spike. 

Three methods were used to correlate the dynamic generic 

test results to the NCAP passenger HIC results. These 

included comparing the peak force/deflection, loading 

slope, and HIC to the NCAP resultant HIC. In no case did 

the component level HIC's match the NCAP HIC's. This is 

because a complex three dimensional event was being modeled 

using a one dimensional test device. Also, there is some 

level ofacceleration on the head due to the vehicle deceleration, 

which is not accounted for in the component test. 

The force/deflection peak is the peak load that occurs 

within the first two inches of instrument panel deflection, 

also called the inenial spike. This peak load and the initial 

loading slope or stiffness were determined from the dynamic 

generic force/deflection plot. If there was no distinguishable 

peak load, as in the case of a gradual rising load, the 

force was taken at one inch of deflection. The correlation 

coefficient for comparison of the peak force and the NCAP 

resultant HIC was found to be 0.58 and the plot is shown in 

figure 17. The loading slope was calculated from the initial 

rise of the force/deflection plot and was found to have a 

L+d{ 

(tfl-.- { 

{ *tr 

/ 

t l /-w 

--j{ 

HJffi 

5hJ 

n 

f t

a 

m I 

LI 

tr 

LL 

}{ 

Ld 

o- 

2000 

1 000 

r:0.58 

0 1000 2000 3000 

NCAP RESULTANT 

HIC 

Flgure 17. Dynsmlc gcnerlc peak force yorsus NGAP HIC correlatlon 

plot. 

Z. 

\ (n 

(D 

J 

t! 

o_ 

oJ 

n 

(J 

1 

O 

o J 

4000 

J000 

2000 

1 000 

0 1000 2000 J000 4000 


Flgure 18. Pynanlc Aenerlc loadlng slope verua NCAP HIC correlatlon 

plot. 

O 

I 

o 

LL 

0{ 

L.J 

Z 

|Jl 

= 

7 

O 

750 

500 

-tr.n 

LJ\J 

Fbure r9. 

pfol. 

r:o.57 

r=0.60 

0 1 000 2000 3000 


Dynamlc generlc HIG varsus NCAP HIC corrolatlon 

correlation coefficient of 0.60 and is shown in figure 18. The 

correlation based on the component level HIC was found to 

be 0.57 and is shown in the plot in figure 19. As seen from 

these plots, all of the methods for correlating the component 

level impact results with the NCAP passenger resultant HIC 

are quite similar. 

Comparison of the NCAP simulation tests performed on 

the damaged NCAP vehicles showed a much bettercorrelation 

with the NCAP resultant HIC. A correlation coefficient 

of 0.96 resulted from these tests and is shown in the plot in 

figure ?0. Although the results correlate well, information 

about the crashed vehicle must be known, such as the passenger 

head impact angle and velocity determined from the 

film analysis. This presents a problem when trying to predict 

the passenger injury criteria for a vehicle that has not 

been crashed. 

q I 

z { 

F J 

f 

a 

LrJ 

tr 

o_ 

{ 

O 

z 

3000 

2000 

1 000 

o 5oo 1 ooo 1 soo 

NCAP SIMULATION 

DAMAGTD VEHICLE HIC 

Flgure 20. Demegod ilCAP HIG versua NCAP raaultant HIC corrslEtlon 

plot. 

= 

z 

tr 

{ J 

l 

= 

a 

o- 

O 

z 

750 

500 

250 

r- 0.6 0 

0 1000 2000 3000 


Flgurr 21. NCAP slmulatlon HIC verrur ilCAP resultant HIC 

correlatlon plot. 

341

The other NCAP simulation testing was performed on the 

group I and group 2 vehicles. Again, there was no exact 

match of the component level HIC's compared to the full 

scale NCAP crash passenger HIC's. The correlation coefficient 

between these two $ets of data was 0.60 and has a 

correlation plot as shown in figure 21 . 

As for the FMVSS 201 results, each of the vehicles tested 

passed the criteria. The FMVSS 201 simulates a head impact 

utilizing a 15 pound head form and impact velocity of 

15 mph. The criteria $tate$ that the maximum acceleration 

level, for a 3 msec period, must not exceed 80 g's. 

As mentioned before, an exact simulation of the injury 

sustained by a passenger in a full scale crash was not possible 

in the laboratory testing program. This is because of the 

constraints placed upon a laboratory test procedure on a 

component level. These constraints include limiting the 

head motion to one direction, neglecting the affect of the 

vehicle deceleration on the HIC and not accounting for the 

belt loading affect on the passenger kinematics. Possible 

areas of an improved test procedure would be to incorporate 

the dummy neck response in a linear impact device. This 

could also be accomplished using a pendulum type device, 

along with simulating the belt loading. However, as will be 

discussed in the next section, component level test results 

are a very useful input to the occupant simulation models, 

which are used to model and predict full scale crash tests. 

Analytical Characterization of Baseline 

Vehicles 

The purpose of this aspect of the Restrained Passenger in 

Frontal Impact$ study was to develop computer simulations 

of selected frontal impact events. The cases are used as a 

tool to study the analytical aspects ofthe events in order to 

identify injury mechanisms and to develop mitigation 

concepts for passenger injury. The MVMA 2D Crash 

Victim Simulation model was selected for this study. 

The MVMA 2D model is a complex nine-mass, tensegment, 

spring-mass system which models total body 

occupant response in two-dimensions (9). It computes step 

by step occupant kinematic details, body part external 

forces and absorbed energies, injury criteria, and vehicle 

response. Frontal events from the National Highway Traffic 

Safety's (NHTSA) New Car Assessment Program (NCAP) 

were chosen a$ cases to be simulated. The two initial cases 

discussed here involve events in which the passenger's head 

struck the instrument panel. Both vehicles employed 

3-point belt restraint systems. 

Modeling approach 

Accelerometer and load cell data from the individual 

NCAP tests and laboratory component data were combined 

to con$truct each of the NCAP cases studied. Accelerometer 

data obtained from the passenger compartment strucrure 

was used to provide the deceleration time-history for the 

simulated passenger compartment and dimensional data 

from undamaged vehicles provided the initial passenger 

342 

clearances and angular orientation of the vehicle interior 

impact surfaces. Laboratory component data, discussed 

above, provided a source for specifying the force-deflection 

properties of the impacted surfaces. Restraint system geometric 

data was also obtained from the laboratory effort. 

Occupant response during the NCAP frontal event was 

obtained from accelerometer and load cell test data and 

from an analysis of the NCAP te$t film. Electronic test data 

provided head and chest deceleration and femur load profiles 

and the film data provided head trajectory and head 

velocity information- The case models were constructed 

and the occupant response output was compared to the corresponding 

NCAP occupant response data. 

The MVMA ?D 3-point "advanced belt system" module 

requires a considerable amount of detailed restraint system 

data not normally reported in NCAP tests. These include 

belt system geometry-the location of anchor points and 

"D" 

rings, occupant attachment points, and the effective 

belt lengths for each segment of the 3-point belt system. In 

additon, belt spool-out and inertia reel characteristics as 

well as the friction characteristics of belt slippage through 

the D-rings and belt slippage over the occupant's torso is 

required. In cases where data was unavailable, nominal 

values were inserted and adjustment was made to belt elongation 

characteristics to obtain the required occupant-restraint 

system response. 

Baseline case deYelopment 

Two cases initially developed for this study were a 1985 

Dodge Colt (10) and a 1985 Ford Tempo (11). Both cases 

were similar in that both vehicles employed 3-point "activen' 

belt systems and involved right front $eat passenger 

head contact with the upper portion of the instrument panel, 

The Colt NCAP test resulted in a passenger Head Injury 

Criteria (HIC) of 738 and the Tempo test resulted in a HIC of 

t436. 

It was noted in the Colt NCAP test that the occupant's 

head and upper torso twisted somewhat when contacting the 

upper instrument panel. This was probably due to early right 

knee contact which occurred about 25 milliseconds before 

the left knee contact. The result was a significant head 

acceleration component in the lateral direction (y-direction) 

when the head hit the panel. The MVMA model is a two' 

dimensional model (x-y) and the panel stiffness input data 

obtained in the laboratory effort was also coplanar (x-z). 

Thus for the Colt case, model occupant head response was 

compared to the resultant of the x and z NCAP accelerometer 

traces rather than the triaxial resultant. The Tempo 

NCAP test on the other hand, had a more coplanar occupant 

head impact. (The x-z resultant HIC of the Tempo NCAP 

test was 997o of the triaxial resultant HIC, while the Colt x-z 

resultant HIC was only about EOVI of the triaxial.) Chest 

response for both vehicle tests was mainly coplanar (x-z). 

Both NCAP test reports indicated knee contact with the 

lower portion of the instrument panel by the right front seat 

pa$senger followed by head contact with the upper portion 

of the in$trument panel. In both cases, film analysis indi-

cated intrusion of the instrument panel into the passenger 

compartment. Since head contact in both cases was partial 

(i.e. only the "forehead" of the passenger dummy made 

contact), it is reasonable to assume that without intrusion 

contact may have been avoided. Results of the comparison 

between the MVMA 2D simulations and the NCAP crash 

tests are shown in tables I and 9 and in figures 22 thru 27. 

The two MVMA 2D simulations match the NCAP resulrs 

fairly well. Femur results forthe Colt simulation show some 

disparity due to the uneven contact times of the knees during 

thE NCAP tCSt. 

I 

Tabre E. MVMA 2D slmulsrlon of NCAP tesr #791, (1985 Cotr). 

Hrq .,,, 

rl-t2 .. 

3[r hird 9.r ......,,,, 

3!E chcat g'r ,........ 

frnur fprco (lb.)..... 

hrrd v.loclty (rph) ... 

601.1 (r-r) 

s?-ll0 

11. r 

tr .9 

l{60 ( rur) 

25 (fIlrl 

525.5 

70-1 35 

7l .0 

3l .5 

leSl 

Table 9. MVMA 2D eimulatlon of NCAP t6$t #E42, (1985 Tempo). 

a 

tnjury clltrrlon I|CAF ?91 t$il tD Colt 

tnjury crttirlon NcA?_tat tlvt|A_2D trrpd 

r.l 

I.r 

tl.l 

t.l 

Erc ..,, ....... l{!5.7! lt13.3r 

tl-t? ,,...,.,. 7{-98 75-16 

3rr hrrd 9'r . . .... . , ,, 110.7 115.3 

t[ chirt 9" ......... 15.6 5t.l 

frlur lorci (lbrl ,,,.. ll27 (rurl l00l 

h.rd vrloctty (rph) ,.. 26 lftlr) t8 

r.t 

||.r 

r.t 

r.t 

il.1 

t.l 

l.l 

Crlt itt ll ltrrl*tfi r Grlt rct trt l?Dl 

H h.,lt.* *..lH,tD tFp X.itt 

t.r 

.t'l 

t.r 

r'ltr.lil'ltt.a-'trr.l 

rlllrnr$ 

Flgure 22. The lollowlno I$ a comparlaon ol MVMA 2D stmutatlons 

wlth NCAP test #791, 1985 Dbdge Colt. 

e3 

I 

t 

tl 

? 

tl 

f.t 

r.l 

Lr 

E.f 

r.l 

t;.1 

ta.l 

t.l 

t.l 

I.r 

fI.r 

I.r 

il,t 

-fI.l 

il.r 

t.r 

tf .t 

tt.i 

r5.l 

tt,t 

5.1 

l.l 

l.l t.l r.l 

r.l rD.l 

dlll*rrl 

t.l 

l.r 

;tt 

't/ 

r ! 

.t 

Orr+ h.rllrrt t.trr.fl6 

Colt llt$ E0 Srftlotro.r t| Colt mP ltl, l?ll 

l|ed ttt*lt1 fril Fllt tr.ttflr 

tt.t r.l lt.t rt.l ll.l 

rlllrr*r* 

Flgure 23. Comparlson of NGAP test lllm analyala wlth MVMA 

2D almulatlon-head veloclty. 

.E 

lrr htr trrrl rt Er.r 

t'i 

'l 

u 

+ 

? 

ii 

i\ 

I 

I 

i 

. a 

r -ra { 

U i '. 

tt 

343

||.r 

lr.r I.r 

t.l 

r.l 

Fl# 

Figure 24. Comparlson of NCAP te8t tllm analysls and MVMA 

2D almulation-head trslectory. 

I 

a 

ft.l 

t t.l 

ttt.l 

fr,l 

r.t 

r.l 

{|.r 

|r.l 

l,f 

t-l 

.|.l 

I.r 

cctt illn lD llrrldlm rlr Cclt lFll Trt l'lDl 

fH tr.JtcttA 

l.r 

t't 

t.r 

trF tn D lllldlrr rr TQf Fl trrt Flt 

lB hrltrt tt.lrrtril 

t.r -'- ||.r t.l - ill.r - . nl-l -- 

l.l l-l I.l r'E.r rl.l 

rrllrrr* 

Flgure 25. The followlng le a comparison of MVMA 2D slmuletlons 

wlth NCAP test # 842, 1985 Ford Tempol 

I 

I 

ll.r 

344 

ff.1 

{r.l 

r,r 

I.r 

E.r 

||.r 

t.a 

tFp 

G|nf, hrrltfrf F..lrr3lfr 

t't 

tfi lD 

. t 

F' 

*., 

r l l 

ilrt tD 

r'r 

rr,ll'.'l*.a-'t-.a 

illllr-rs 

L 

I 

I 

I.r 

tllql 

I.r 

r.f 

t.t 

-il.1 

il.. 

tl.r 

t{.t 

lr,l 

rt.l 

It.l 

t,l 

a.t 

l.l 

Lr l|.l I.r rt.l rt.l 

l.l t.l il.1 fr.l rt.l 

r|lll*-f 

tqo fffi tD llrhtlrr rt tr4o fEf Trrt lltE 

lld ttlcltg frn Fllr trl1lrr 

t.t --tt.t 

{l.t 

s.l 

r.t 

rr.l 

rlt.t 

l{l.l 

tll.l _ 

fl.l 

arl llaacdraa 

Flgure 26. Comparlson ol NCAP test fllm analyais with MVMA 

2D slmulatlon-head veloclty. 

r.l 

t.l 

|rro fllil lD lrirl.tlf,r rr TF fGil Trrt [tl€ 

lhd tr.Jfstfirl 

r.l 

?rr hcr lrtrl * hrr 

r'r 

t.r 

FltFls 

tsF ---- 

Figure 27. Comparlson of NCAP test fllm analysis wlth MVMA 

2D slmulation-head trslectory. 

f.l 

I.r

Parametric study 

To identify the critical factors leading to passenger 

injury 

during the frontal event, a parametric study was developed. 

Both events were nominally similar. At the start of the 

frontal impact, both dummies slide forward from theiroriginal 

seating position. Evidence indicates initial contact with 

the knees into the lower portion of the instrument panel 

followed by a pitching forward of the occupant's upper 

torso and impact of the forehead into the upperportion of the 

instrument panel. Several critical parameters were 

identified: 

r Restraint System Dynamic Compliance 

. Upper Instrument Panel Stiffness 

. Pflssenger Seat Position 

r Torso Belt Slack 

Parametric test results are shown in figures 28 thru 31. 

Restraint system dynamic compliance was identified as 

compliance of the restraint system during the event which 

included a combination of belt stretch, belt spool-out, and 

belt hardware distortion. Results indicate that a stiff belt 

system which minimizes occupant forward movement improves 

the passenger's injury response. The more compliant 

a re$traint system is, the more likely the occupant will strike 

the instrument panel. The severity of the impact is also a 

consequence of restraint compliance as the belt system can 

slow the forward movement of the occupant and minimize 

the impact. A sufficiently stiff restraint system can prevent 

head impact totally. As restraint sy$tem stiffness was continually 

increased however, an increasing trend in occupant 

injury was observed as the stiffer belt system absorbed less 

energy, transferring more of the crash energy to the 

occupant. 

When head impact with the instrument panel does occur, 

the stiffness characteristics ofthe panel has a direct effect on 

resulting head injury. In both cases it appeared that the 

restraint system, although allowing head contact, did mitigate 

the impact by slowing the occupant's forward velocity. 

Thus head contact was generally partially mitigated and the 

stiffness of the first two to four inches of the upper panel 

played an important role in injury response. Response was 

almost linear, the stiffer the panel the harder the head impact. 

HIC and head angular accelerations responses followed 

suit. The softer panels mitigated head injury but 

allowed deeper penetration of the head into the panel which 

increased the likelihood of impacting a panel structural 

member of localized "hard spot." 

Moving the passenger to a more rearward seating position 

helps to mitigate injury by allowing the restraint system 

more room to prevent passenger head impact. As noted 

above, head impact in both cases was partial, usually only 

the forehead of the occupant strikes the instrument panel. 

Thus, any rearward seating position from 2 to 4 inches 

helped to minimize head impact. Conversely, preventing 

panel intrusion can also accomplish the same ends. Moving 

the occupant forward tended to increase injury criteria in the 

cases studied since the closer proximity of the instrument 

panel allowed for a fuller head to panel contact. The Colt 

case showed a decrease in passenger head injury however, 

when in the extreme forward seating position. Thorax contact 

was initiated in this case and helped to dissipate the 

occupant's forward momentum. The Tempo case on the 

other hand, saw a significant increase in head injury for the 

extreme forward seating position (4-inches forward of the 

mid position). In the Tempo case the occupant struck the 

windshield initially, rebounded and then hit the upper instrument 

panel. 

As part of the parametric study, torso belt slack was 

introduced into the baseline cases. Negative slack was also 

introduced which had the effect of simulating initial belr 

tension in the system. Generally, the introduction of belt 

slack increased restraint system compliance and increased 

passenger head injury response. Initial belt ten$ion tended 

to improve passenger injury response somewhat. It was 

noted in the case of the Tempo, that with greater than 1.5 

inches of slack, head impact was squarely on the top of the 

instrument panel rather than a combination of top and middle 

panel impact. This reduced the head angular acceleration 

and neck injury response since impact with the almost 

horizontal middle panel had tended to force the head sharply 

rearward. The magnitude of the resulting head impact with 

the top panel however, sharply increased. 

Discussion of results 

Head injury was the major response variable in this preliminary 

analysis of restrained front seat passengers in frontal 

crashes. Throughout the parametric study, chest and 

femur injury response remained moderate and except for 

some extreme parametric variations, varied littte. Neck injury 

generally showed only a moderate response to parametric 

variations. (Differences in neck response between the 

Colt and Tempo cases however, were significant.) Restraint 

system operation was a key variable in the passenger's 

injury response. A sufficiently stiff restraint system can 

prevent head contact during the event. However, too stiff a 

restraint system can increase chest injury and head and neck 

response as more of the crash energy is transfered to the 

occupant. When the restraint system was too compliant, 

head contact with the instrument panel occurred and in 

some extreme cases, the head contacted the windshield. The 

extent of the head injury incurred by instrument panel impact 

is dependent on head contact velocity and panel stiffness. 

Head contact velocity is a function of crash pulse 

severity, passenger clearance, and the extent to which the 

restraint system slows the occupant prior to contact. In the 

two cases studied here, the disparity in head injury criteria 

between the two vehicles was largely a result of restrain 

system dynamic performance and crash pulse severity. 

The restraint system in the Colt case slowed the passenger 

forward travel sufficiently to mitigate head contact injury, 

while in the Tempo case, torso belt compliance allowed the 

occupant to impact the instrument panel at a significantly 

highercontact velocity (28 mph for the Tempo head impact 

and 20 mph for the Colt). 

Preliminary results indicate that restraint system perfor- 

345

tl 

I 

C 

I 

I a 

t I 

? 

I a3 

I 

I ? 

a 

lS Or.|| Cclt rra tG lfrl tFo ilrt tD ll3|ltrflr ftr 0rtr Cclt rr.l lttr FFt T.+o ilrt ND Srrrhttn 

ITil.I 

IS r|hill.rrl|rt llrt r CAtlrE 

rtll.l 

fllC v ||t f l||t'|frit Fil.t 8l,tttf5. 

m.r 

il.1 

rfl.t 

H|r.I 

tn.l 

rtt.l 

tH.l 

llr.r 

ttr,l 

t.l 

il.r 

n.l 

r.r 

t.l 

ttt.l 

r||!.1 

l.n.l 

iltr.r 

tH.l 

n,l 

t.l 

rf,.l 

H.r 

f.r 

r.lr 

l5.firtr 

tn-l 

*..|r*rf,r r| lrttrl|* trt- Cflfr*f 

ilH.l 

t,l 

n.l 

H.l 

t,l 

il.1 

t.r 

.I.l 

t.l 

l.tt 

. Tqo. 

'crtt 

l.I r,I r.rl 

t.l| l.I r.ll r.t 

llrtls t t .ld ttfr Gtlr|r*. 

|ld |rr*f rr frltruld lflrit Ct+llrt 

t.fl 

l'.' 

l... 

l'r 

t.E 

hl*lrr lEt .l|rt l|rl^r C;r|rr. 

F.r 

Lf 

lJ 

rJ I't 

l.r l'r 

r.r r.t 

Il*|n hr.rrd l*rhllc. 

Flgure 28. The followlng are restralnt syst€m compllance 

teals, iIVMA 2D craah vlctlm almulatlone: 

346 

t't 

t.t 

tl 

I 

C 

I 

I 

I 

t I 

a 

t IF 

t 

I G 

I 

I ? 

a 

rtl.l 

tH.i 

H.r 

tr.t 

aI.a 

lr.r 

lll.t 

t.l 

rn.l 

il.1 

t.l 

l.{ l.l t.t t,f 

l.t 1.3 r.r r.r r.l 

b|rttr tlrrl l[lfftr. 

l.l --- l.r t.l - l.t r.l 

l.t l.l l.l l.{ l.l 

|rlr0lr| trrrt ttllllrr 

H ftrlr ttrh*t; r lfrfr tltr*r* frrl ltlllrnr 

tn.l 

rrn.l 

ffi-l 

n.r 

t,r 

frt,f 

t.l 

I.l 

rn,r 

il.t 

t,r 

Itd I+*'f rlr $pr lnlrr:il' hrl ltlffil$ 

. t 

i - / 

.t 

hft 

,i 

JFr 

i. . . r{ 

il.t 

IT l'{ r'r l'l r'l t'l 

l.r r.t r'l t.. t.t 

blArr hl |lraf|nr 

Flgure 29. The followlng are upp€r lnatrument panel atlffneet 

tE8ts, MVMA 2D craeh vlctlm simulatlone: 

blt

L 

I I 

IID Drt||r Colt fJ lE Fr{ tqlc lllt lD lllhltrr l* Dfrr Gllt rt lG lr.r l* llt !D lrrrl*,rn 

tI.l 

lllt |,. lbrqr *cttr1 lc.ttlf,r 

t.l 

lllt rr Trro frll 3l*] 

fr.l 

I.l 

n,l 

In.t 

rn.l 

trn,l 

tLl 

ln.t 

I.l 

I.l 

rI.l 

In.l 

tfl.l 

rn.l 

It.f 

ilI.l 

t.l 

rI.r 

H.r 

I.l 

rt.r 

I 

I n.r 

3 n.r 

f 

I 

a 

t.r 

3 

t 

t 

I 

tr.r 

I.r 

.Ll 

{.1 

{,r l,l 

-l.l l.l 

Ld bltlr| (trr.tal, 

t.l 

lt=ln 

lh- IfGt ||;rrrar t.ttrrf frtttolr 

.Crrtl.*f. 

. a 

I 

{,t 

-l.l t.l 

t.t 

t.l 

l* hltlr| (|}r dt, lr*ln 

lH t;rlr *..lrr*rrl n rhrrir fr*|if htttrr 

trfi,l 

r l 

. a 

r a 

'brrl 

lF 

r-r 

t.l 

{t r.r 

l'l 

rr 

'rl 

H hrltrr tftr r|at, l* 

Flg_urg qq. The followlng ano pqrsongor roatlng poslilon te$ts, 

iIVMA 2D crsah vlcllm almulatlone: 

I 

Y 

f.l 

tl 

I 

C 

L 

I 

I 

tI.l 

I.l 

rt.l 

rI.l 

trD.l 

rtl.r 

t.l 

I.t 

Ll 

il.1 

It.l 

rt.l 

It.l 

ril.l 

ft.l 

t 

r I.l 

a 

rrt.l 

fr.f 

t.t 

I D.l 

a 

t 

r.l 

I 

. t.l 

I r.t 

t ? 

a I.l 

In.l 

r.l 

Crlt, 

r.l 

l.i t.t 

l|lt ll-r, |'*lr. 

lH t|!ff rr trrc l|lt tl-* 

. t. 

rqo j 

' ! 

t 

i 

-i 

Grlt 

r.a 

l.r t.t 

hll ll*lr, tr*ln 

hl l|rr; IGatH3th E rr* ErI ..m 

,ts'- t 

ih. 

r .i '. 

t 

. .1. 

t i r 

1 / r 

t. r-.1 

v ,y" 

-- r 

c.tL.f 

-l.l 

{.r 

l.l l.l 

l.r 

lrn lld, 

t.! 

frln 

t.l 

1.3 

Flgure 31. The followlng ar6 torso bett $teck tests, ilVilA 2d 

crash vlctlm slmulatlon$:

mance is a complex event: restraint system compliance, 

crash pulse severity, passenger clearances, instrument panel 

stiffness, and panel intrusion are all key elements affecting 

occupant injury. The two NCAP cases show that in addition 

to these apparent event variables, performance of key elements 

of the restraint system is also an important consideration 

in assessing the potential for mitigating passenger injury. 

Future work will focus on analyzing additional frontal 

crash events including light duty trucks, and on identifying 

and analyzing critical restraint system elements. 

Summary 

The objective of the frontal crashworthiness research 

program at the National Highway Traffic Safety Administration 

is to assess the safety problems associated with 

occupants of vehicles involved in frontal impacts and to 

identify potential remedies for the problem. The focus of the 

analyses presented in this paper is the identification of the 

safety problem for restrained, right, front, seat passengers. 

Analyses of accident data and crash test data indicate the 

pas$engers can still make contact with interior vehicle 

surfaces even why they are restrained. Head contact with the 

upper part of the instrument panel appears to be a serious 

safety problem, and research efforts were initiated through 

both laboratory and analytical studies to better characterize 

the injury mechanism. Initial results of the studies are 

presented in this paper. 

The laboratory effofis included the collection of both 

dynamic and static force-deflection characteristics of the 

upper instrument panels for a wide range of passenger cars 

and light trucks that had been tested as part of NHTSA's 

New Car Assessment Program. Test data was collected 

utilizing a component level head test which consists of a 

hemisphere shaped headform mounted on an impactor 

which is essentially a guided piston. The component level 

laboratory results were compared to the NCAP results and a 

fair degree of conelation was obtained. Exact simulation of 

the injury sustained by a passenger in a full scale crash test is 

not po$sible utilizing the present test devices and 

procedures. Constraints of the present system include 

limiting the head motion to one direction, neglecting the 

affect of the vehicle deceleration and belt loading on head 

accelerations, and a fixed effective mass. Future research 

will examine alternative component level test procedures to 

better simulate full scale crash tests. 

The analytical effort utilized the MVMA 2D occupant 

simulation model. Initial simulations focused on the 

validation of two NCAP tests, and the results of the 

validation and parametric studies are presented. The 

laboratory study ptovided input data on the material 

properties of the instrument panels and geometry of the 

vehicle interiors. The parametric study indicates the effect 

that crash pulse, intnrsion, seat position, restrain system 

compliance, and instrument panel stiffness have on head 

injury criteria. The extent ofthe head injury criteria incurred 

348 

by instrument panel contact is dependent on head contact 

velocity and instrument panel stiffness. Head contact 

velocity is a function of crash pulse severity, passenger 

clearance, and the extent to which the restraint sy$tem slows 

the occupant prior to contact. While the first priority in 

developing mitigation concepts should be aimed at 

eliminating or reducing head contact velocity, it will not be 

possible to eliminate all head contacts. For those situations, 

the instrument panel and other interior components need to 

be modified to mitigate head contact forces. 

References 

(1) National Accident Sampling System, 1986. A Repon 

on Traffic Crashes in the United States, National Highway 

Traffic Safety Administration, July, 1988. 

(2) Crashworthiness State Database, Reference Manual. 

To be published, National Highway Traffic Safety 


(3) NHTSA Data Tap Reference Guide, Volume I: 

Vehicle Crash Tests, National Highway Traffic Safety 

Administration, August 1985. 

(4) Malliari, Hitchcock, and Hedlund, "A Search for 

Priorities in Crash Protection," SAE International Congress 

& Exposition, SAE 820242, Detroit, Michigan, Feb. 1982. 

(5) Malliaris, Hitchcock, and Hansen, 

"Harm Causation 

and Ranking in Car Crashes," SAE lnternational Congress 

& Exposition, SAE 850090, Detroit, Michigan, Feb. 1985. 

(6) Luchter, Faigin, Cohen, and Lombardo, "Status of 

Costs of Injury Research in the United States," to be 

presented at the Twelfth International Technical Conference 

on Experimental Safety Vehicles, National Highway Traffic 

Safety Administration, June, 1989. 

(7) Crane, D., "Restrained Passenger Head Testing," 

Volume l, MGA Research Corporation, under contract to 

the National Highway Traffic Safety Administration and the 

Transportation Systems Center DTRS-57-87-C-0046, 

MGA Draft Report No. G88R{2, February 1989. 

(8) Segal, D., Kamholz, L., and Griffith, D., "Vehicle 

Component Characterization" Volumes I and 2, MGA 

Research Corporation under contract to the National 

Highway Traffic Safety Administration and the 

Transportation Systems Center, DOT HS 807 054 and DOT 

HS 807 055, January 1987. 

(9) Bowman, 8.M., Bennett, R.O., Robbins, D.H., 

MVMA Two-Dimensional Crash Victim. Simulation 

Version 5, Vols I, Il, lII. Transportation Research Institute, 

University of Michigan, Ann Arbor, MI 48109. Rpt. No. 

UMTRI-85-?4-?, October, 1986. 

(10) Carlson, L.E. and Leonard, P., New Car Assessment 

Program, Frontal Barrier Impact Test, 1985 Ford Tempo GL 

Z-Door. Mobility Systems and Equipment Company, Los 

Angeles, CA 90301 (Prepared for U.S. Dept. of 

Transportation, National Highway Traffic Safety 

Administration, Office of Market Incentives, Rpt. No. 

MSE-85-No. 4), July, 1985. 

(ll) Levan, W.E. and Alianello, D.A., New Car

Assessment Program, Frontal Barrier Impact Test, 1985 

Dodge Colt 4-Door Hatchback. Calspan Corporation, 

Buffalo, NY 14225 (Prepared for U.S. Dept. of 

Computational Crash Analysis at the Saah Car Division 

Larsgunnar Nilsson, 

Saab Car Division 

Saab Scania AB 

Abstract 

In the development of Saab cars, passenger safety is of 

the greatest concem. To verify the safety level achieved a 

large number of destructive crash tests are carried out. Each 

of the$e te$tri is time-consuming and costly. It is of interest to 

Saab, as well as the whole community of cflr manufacturers, 

to reduce the number of crash tests, and yet to achieve the 

target level of passenger safety. Computational mechanics 

has turned into a powerful tool which can complement crash 

tests and, furthermore, reduce the number of tests needed. 

Recently, Saab has developed a finite element model of 

the Saab 9000 T | 6 CS car. The calculated results show good 

agreement with available test data and also reveal crash 

phenomena which are hard to observe in the standard destructive 

physical tests. 

Computational analysis of crash problems is the new 

method of developing cars with higher passenger safety. 

Once developed and validated, the finite element crash 

model can be utilized for design studies or parametric studies 

and needs only a very short time to produce useful 

results. Thus, within a given time a larger number of design 

altematives can be evaluated and a final design with higher 

passenger safety can be obtained. 


Safety has always been of great concern in the 

development of Saab cars. During the I 950s to I 970s safety 

developments mainly concerned active safety. The success 

of Saab in rally competitions during that time was partly a 

result of high active safety. Passive safety has since then 

been given lncreased attention to meet customers' needs 

and government regulations. 

The engineering considerations in regard to 

crashworthiness are extensive, Many experiments are 

carried out as design studies or verification tests on both 

prototypes and cars taken from the production line. Both 

complete car structures and components are tested. 

Although possible, development of structural 

crashworthiness by testing alone is a much too costly and 

time-consuming process. 

It is of great interest to Saab to shorten the development 

phase and yet achieve the target level of passenger safety. 

During recent years, computer analysis of nonlinear 

contact-impact problems has developed remarkably. Today, 

Transportation, National Highway Traffic Safety 

Administration, Office of Market lncentives, Rpt. No. 

CAL-8-5-No. ?), February, 1985. 

it is possible to preform analysis of complex structures, 

which just a couple of years ago was considered to be 

science fiction. The rapid evolution in this field is a 

consequence of developments of more theoretically sound 

and efficient algorithms for the numerical solution of 

nonlinear structural dynamics, as well as the rapid growth in 

speed and availability of supercomputers. Computer 

analysis of crashworthiness is on the threshold of emerging 

as a tool that can substantially reduce the cost and time 

required for the development and certification of a new car 

design. 

Numerical Analysis of Vehicle 

Crashworthiness 

Review 

Computers have been used in the analysis of 

crashworthiness since the 1960s. In the first generation of 

analysis the vehicle was modelled as a lumped system 

consisting of mass and spring elements. The main problem 

with these models was to assign appropriate stiffness to the 

spring elements. These data must be found from 

experiments on structural components similar to those 

which are to be analysed. The resulting system constitutes a 

set o[ coupled ordinary differential equations, which must 

be integrated in the time domain. Because of the difficulties 

mentioned and the limited capacity of the mainframe 

computers available at that time (comparable with that of 

personal computers today), the number of mass and spring 

elements wari very limited. In general, the results from 

physical tests on the same structure or its components must 

be available in order to calibrate the model. As a 

consequence, the applicability of the lumped system was 

limited to analysis of structure$ very similar to the one 

tested. Thus, the lumped models could not be used to make 

predictions regarding new car designs without the existence 

of a physical prototype. A typical lumped mass system used 

at Saab in the 1970s is shown in figure I. 

The finite element method 

The finite element method (FEM) was developed during 

the 1950s and the 1960s mainly for the analysis of static 

problems. The first attempts to use FEM in the analysis of 

crashworthiness were made at the beginning of the 1970s. In 

these analyses, beam and rod elements were used along with 

discrete mass and spring elements. Although reflecting the 

real structure to a larger extent than the lumped models, the 

properties of the structural elements were still mainly found 

349

Flgure 1 . Mass'sprlng model of s Ssab 9fi1 vehlcle (1979). 

from physical tests. Thus, the criticism of the lumped massspring 

models is also valid for the first generation of FEM 

crash analysis models. A widely used computer program of 

the first generation type is KRASH (Wittlin and Gamon 

( I )*). 

With FEM first-principle analysis became possible, i.e. 

the structural properties are derived from the basic continuum 

mechanical relations. Thus, it became possible to make 

crashworthiness predictions of a vehicle structure simply by 

knowing the relevant material properties, the geometry, and 

the boundary and initial conditions. 

The equations of motion resulting from the FE-discretised 

system must be integrated in time, starting from 

known initial conditions. This is carried out as a step-bystep 

integration, either with an implicit or an explicit 

method. 

When an implicit method is used, a non-linear equation 

system must be iteratively solved in each time-step, which 

makes it a very time-consuming step. In crash problems the 

loads vary extremely rapidly. In order not to truncate these 

loads, very short time-steps must be used. Thus, implicit 

methods demand very large computer resources to integrate 

the response time of interest. Also, due to the severe nonlinearities 

involved, the iterative equation solver often fails 

to converge. Consequently, implicit methods have not been 

successful in the analysis of crash problems, although many 

ofthe general purpose FE code developers still do not seem 

to have given up hope. 

The generally used explicit method is the central difference 

method. Unlike the implicit methods, it does not require 

the assembly of system matrices and the solution of 

each time-step is trivial. However, the central difference 

method is only conditionally stable, i.e. the time-step must 

be less than a critical value proportional to the shortest time 

it takes a sound wave to travel between two adjacent nodes. 

It follows from the previous discussion that the time-step 

limitation is not too severe in the analysis of crash problems. 

* Numbers in parentheses designate references at end of paper. 

350 

Today, all the popular codes in crash analysis u$e the central 

difference method. 

Car body structures mainly consist of connected sheet 

metal paft$. To make an accurate FE model of the car body, 

shell elements are essential. Also solid, beam and rod elements 

are required in a complete vehicle structural model. 

The first FE-analysis, which used shell elements and 

explicit time integration, of a vehicle collision was carried 

out in 1975 by Welch, Bruce and Belytschko (?). It was also 

the first attempt to use FEM as a first-principle analysis of 

vehicle crashworthiness. At that time, the computers were 

not powerful enough to perform the analysis successfully. 

During a crash, the vehicle structure is severely deformed 

and the structural parts may be exposed to multiple contactimpacts. 

The first FE program, with an explicit time integration, 

that seriously addressed the contact-impact problem 

and that took full advantage of the supercomputer (Cray) 

was DYNA3D, developed by Dr. John O. Hallquist (3), (4) 

The first version of DYNA3D became available in 1976. 

DYNA3D has since developed remarkably, and is today 

probably the most well-known and widely used FE code in 

crash analysis. Most of the other commercial crash analysis 

codes, e.g. MSC/DYNA, PAMCRASH, and RADIOSS, 

have their origin in DYNA3D. 

DYNA3D originally lacked shell and beam elements. 

During that time, vehicle components such as a front beam 

were modelled by the use of solid elements. To account for 

bending effects in the metal sheets, at least two elements 

were required across the thickness. As previously mentioned, 

this resulted in a very small time-step, due to the 

short distance between nodal points in the thickness 

direction. 

Shell elements became available in PAMCRASH 1984 

and in DYNA3D 1985, see hallquist et al (5). The first shell 

element implemented in DYNA3D was the Hughes-Liu 

element (Hughes and Liu (6)) which is derived from a continuum 

formulation and allows very large deformations. 

Later, also the Belytschko-Tsay element (Belytschko and 

Tsay (7)) was implemented in DYNA3D, see Hallquist and 

Benson (8). The Belytschko-Tsay element is based on the 

Mindlin plate theory and thus only allows moderate shell 

deformations. On the other hand, the Belytschko-Tsay element 

requires fewer operations and makes the analysis 

faster. 

The above-mentioned developments resulted in a tre* 

mendous increase in computer analysis of vehicle crashwofihiness: 

The first complete vehicle (Citroen BX) frontal 

30 mph crash analysis was carried out 1985/1986 at Peugot 

S.A., see Chedmail et al (9), and since then most car manu* 

facturers have performed similar analyses. 

Saab started to use DYNA3D in the analysis of front 

beams, and front frame structures in 1984. DYNA3D has 

since then been used extensively, see Nilsson and Larsson 

( l0). Saab has also actively supported new developments in 

DYNA3D through Dr. Hallquist and through research programmes 

carried out by the author at Linkdping Institute of 

Technology, e.g. Zhong (l l) and Oldenburg (12).

Examples 

At Saab, computational crash analysis is considered to be 

an engineering activity of high priority and its importance is 

rapidly increasing. Since 1985, Saab has continuously 

aimed at developing qualified in-house resources, i.e. 

personnel, methodology, software and hardware, for 

advanced crash simulations. 

The following crash simulations have been carried out at 

the Saab Car Division using DYNA3D. All simulations, 

except the full frontal car crash, have been run in-house 

using a Cray 1A supercomputer. The full frontal car crash 

simulation was run on a Cray X-MP/28 supercomputer at 

SINTEF, Trondheim. A new Cray X-MP/48 supercomputer 

has recently been installed at Saab, and all current crash 

simulations are run on that computer. 

The following examples have been chosen to give a brief 

overview ofthe crash analysis carried out at Saab during the 

last couple of years. 

Structural Parts and Components 

Front beam 

The boxed beam is often used as a front structure member. 

It has been analysed extensively at Saab as a standard test 

example to debug new methods or new computer code 

versions and facilities. The finite element model of one half 

of the beam is shown in figure 2. Only one quarter of the 

beam is analysed with symmetric boundary conditions 

taken into account. At the right end of the beam, a rigid mass 

is attached. A total of 1280 4-node shell elements are used. 

The beam and the rigid mass have an initial velocity of 35 

mph leftwards along the length axis of the beam. The 

motion of the beam is constrained by a rigid wall, which is 

hit bv the left end of the beam at time zero. 

Flgure 2. Inltlal conllguratlon of ths lront baam. Ons qusrter of 

the b€am la analyaad wlth symmstrlc boundary condltlons. 

Figure 3 illustrates the deformation of the beam at 

different time states. It is observed that the folding 

mechanism is described in details by the model. This has 

been made possible by the single-surface-contact 

algorithms in DYNA3D, i.e. all contacts are automatically 

notified and analysed. Obviously, this facility is of the 

outmo$t necessity when a full car structure is considered. 

Figure 4 illustrates the typical time graph of the normal 

force on the rigid wall. Each drop in acceleration is caused 

by the formation of a new fold. 

Flgure 3. Ilelormatlonr of the fronl beam after 0, 4, 8, 1 2, 1 6 snd 

20 m$. 

a 

J 

J 

(I 

= 

LJ 

z 

o 

F 

(n 

z 

o 

IJ 

U 

at 

o 

L 

J 

(I E 

tr 

o z 

6, 

3,8 

3,s 

t,(xE 

{.508 

4,ffi 

r,Sft 

3.TIE 

E, !fi 

t.ffi 

t.5{E 

i.fff 

o 

H H H E 

d i d d 

ilhlu . o. (xxt(E*{b 

Hrtffi ' o.ctt!E.o. TII'lE 

Flgure 4. 

wall. 

H I 

Normal torce bstween tho lront beem and the rlgld 

Typically, 2 hours of Cray lA CPU time is required to 

obtain 20 ms of response time. 

Front frame structure 

The finite element model of aconceptual frontal structure 

is shown in figure 5. A total of 3500 shell elements were 

used to model the symmetrical part of the structure. The 

inefiia proper"tie$ of the rear part of the car body are represented 

by a discrete mass attached with rigid links to the rear 

parts ofthe front. Initially all nodal points have a velocity of 

35 mph in the forward direction, and the bumper hits a rigid 

wall extending perpendicular to the direction of motion. 

Figure 6 shows the deformed structure at selected time 

states. From these figures, together with a video animation 

produced from the detailed solution sequence, it is possible 

to acquire an understanding of the mechanical behaviour of 

the structure during impact. 

351

FIoure 5. Inltlal conllourstlon of the front lrame struclurs. Ons 

ha-lf ol the lrsme 15 analy*d wlth symmelrlc boundary 

condltlons. 

A total of 7 hours CPU time on the Cray 1A was required 

to obtain 40 ms of response time. 

Steering column 

In figure 7 the geometry and various deformed configurations 

of a Saab 9fi) steering*column system are shown. The 

model consists of 2l00shell elements. A rigid wall (mass 30 

kg) hits the steering-wheel with an initial velocity of 9 mph 

in the direction of the steering-column. 

The FE-model of the steering-column irtcludes details of 

the energy-absorbing sliding mechanism between the column 

segments as well as contact interfaces between all 

parts. 

Flguls 7. Inltlal conflguratlon of th€ lt€€rlng whssl'column 

syst€m. 

Figure 8 shows the deformed geometry at various time 

states. Details of the deformation process can be revealed by 

making the exterior parts invisible, as shown in figure 9. 

The typical force-time relation of the rigid wall is shown in 

figure 10. 

The steering-column model has been used for the purpose 

of design studies. 

A 20 ms response analysis requires approximately l0 

CPU hours on the Cray lA. 

352 

Flgure 6. Deformatlons of the front lrame slructure sfter 0, 10, 

20 and 30 ms.

t=5ms 

t=10ms 

t=15ms 

Flgure L Deformatlona of the steerlng wheeFcolumn syslem 

after 5, l0 and 15 mg. 

Flgure 9. Datalls ol thr strsrlng column sy$tem. Erterlor vlew 

after 0 and 15 mr. Sllcad-up vlcw aftrr 0, 10 and 15 mr. 

353

t 

f 

tr 

3 

U 

z 

o 

F 

o 

7 

o 

U 

tr 

J 

c 

E 

tts 

z 

c i ^ i i { i d i , ; i i i r N N 

T IFIE 

Flsurt 10. Normal lorco on thc rlgld wall (dummy) |mplnglng 

the ateerlng wheel-column syttom. 

Complete Car Structure 

Model preparation 

In a frontal car crash, extensive folding (figure 3), and 

bending collapses take place in the front structure, contactimpacts 

occurbetween engine/gearbox and radiatorffan and 

between enginelgearbox and firewall etc, and many 

structural members are highly deformed and distorted. It is 

an advanced task to develop a FE-model which, within 

acceptable tolerances, gives the required results' The 

modelling of the frontal crash in the case of the Saab 9000 

T16 CS (figure I I ) was an extensive iob, taking almost one 

year. This complete vehicle model has, however, since then 

also been used for many other purposes' e.g. side impact 

analysis, load distribution analysis, idle shake vibration 

analysis etc. 

Owing to time considerations, it was decided to model 

only one half of the body, assuming symmetry along the 

body centre line (i.e. the x-axis, y = 0). The engine/gearbox, 

the AC and some other systems were, however, fully 

described. From similar crash test$, we know that the 

wheels/tyres take part only to a minor extent in the 

deformation process. Thus, they have been omitted from the 

analysis model. Because of these arrangements, the results 

obtained were valid for up to about 80 ms response time. 

The FE-model (figure 12) consists of 7500 shell 

elements. In addition, beam elements are used as door 

substitutes. The total weight and weight distribution of the 

FE-model were chosen to correspond to the real vehicle. 

Initially, the vehicle has 35 mph velocity forward (xdirection) 

and hits the rigid wall, which has its extension in 

the yz-plane, at time zero. 

354 

Fioure 11. lnltlal conflouration$ ol the SEEb 9000 vehlcle FE' 

m6dei snd detalled vle* of the engine/gearbox/Ac sy8tem. 

Analysis 

Because of the primary memory restrictions of the Cray 

1A (l MW), the fuII frontal car crash analysis was run on a 

Cray X-MP/28 at SINTEF, Trondheim. To make this possible. 

we first installed our versions of DYNA3D and TAU- 

RUS (postprocessor of DYNA3D, see Hallquist and Brown 

(13) there. In total, four days were needed to produce a tape 

with the results database. The total CPU time needed for one 

run covering up to 80 ms respon$e time was 22 hours. 

A video animation of the full set of deformed states, with 

or without overlaid stress, energy or strain functions, was 

also produced by using MOVIE.BYU (Christiansen (14) 

on-line on the Cray at SINTEF. 

Results 

From the analysis, a very large results database was obtained. 

The following selected results are mainly chosen to 

illustrate the versatility, validity and accuracy ofthis type of 

study.

t=60 

FIgure 11. Soqucnco ol tho dolormed Saab 90(Xl vehlcls FE-modcl Et O, 20, 40 snd 60 ms after lmpect. 

Figure 12 shows a set of deformed states. When compared 

with high-speed films captured in similar tests, a very 

close agreement in the global deformations is observed. 

Figure 13 gives a close-up view of the engine compartment 

before impact and at one deformed state. The deformations 

obtained in the analysis are very close to those found in 

a real crash. 

355

I 40fls 

E rp{trirn.4ntc 

Flours 13. Glo*e-up vlew ol thc Saab 9000 vehlcle front struc' 

tuie and the correbpondlng FE-model' Undcformsd and de' 

lormed (40 me) configuratlons. 

Figure 14 gives close-up views of the front beam structure 

before impact and at one deformed state. The close 

agreement of the analysis results to the test is again 

illustrated. 

Figure 15 shows the measured and calculated acceleration 

signal (x-direction) from an accelerometer mounted on 

top of the sill. The measured signal has been integrated 

twice to find the velocity and the displacement' Furthermore, 

it has been filtered (60 Hz) before it was plotted 

together with the velocity, the displacement and the corresponding 

data from the computer analysis. As noted, a very 

close agreement of the analysis results with the tests has 

been obtained. 

The video animation of the analysis results gives a very 

detailed understanding of the mechanics taking place during 

the deformation process. On the deformed geometries also 

fringes of stress, strain or intemal energy intensities are 

shown as further help in understanding the proces$. 

With some exceptions, very good agreement has been 

achieved between the analysis results and the tests. After 

$ome minor corrections. this full vehicle model can be used 

with confidence in future analysis. The possibilities of conducting 

parametric studies or design studies are extensive. 

356 

ll 

/l 

,'t 

Flgure 14. Slde vlew of the Saab 9000 vehlcle front structure 

snd ths correspondlng FE'model. Undeformed and delormed 

(70 me) conflgurstlon$. 

Finally, this study shows that prediction analysis is possible 

today. The input data for this analysis are simply material 

parameters, geometry data, and boundary and initial 

conditions. Thus, it is a true first-principle analysis. 

Conclusions 

Experience at Saab in using advanced computer analysis 

of crash problems has increased quickly since 1984. This 

has been made possible by the breakthrough in 

computational crash mechanics (new theories, algorithms 

and codings in DYNA3D), fast access to supercomputer$, 

and the continuous and consistent build-up of in-house 

personnel resources and know-how, 

Computational mechanics is the new methods of 

developing cars with higher passenger safety. Once 

developed and validated, the finite element crash model can 

be utilized for design studies or parametric studies, and 

needs only a very short time to produce useful results. Thus, 

within a given time a larger number of design alternatives 

can be evaluated and a final design with higher quality and 

passenger safety can be obtained.

flgqrg 15. Eottom vlew ol tlq $aeb 9fi10 vehlcte FE-modet. 

Undoformed and delormed (20 ms) conflguratlons. 


The close cooperation of Dr John O. Hallquist in various 

aspects of DYNA3D is greatly appreciated. 

Mats Larsson and Roger Malkusson in the Stress 

Analysis Group at the Saab Car Division have assisted in 

most of the computer runs. 

References 

(l) Wittlin, G. and Gamon, M.A.: Experimental Program 

for the Development of Improved Helicopter Structural 

Crashworthiness Analytical and Design Techniques, 

USAAMRDL Technical Reporr, 72-72, 197 3. 

(2) Welch, R.E., Bruce, R.W. and Belytschko, T.: Finite 

Element Analysis of Automotive Structures under Crash 

Loadings, Report DOT-HS-105-3-697, IIT Research 

Institute, Chicago 1975. 

(3) Hallquist, J.O.: Preliminary User's Manuals for 

DYNA3D and DYNAP, University of California, Lawrence 

Livermore National Laboratory, Report UCID*17268, 

Livermore 1976. 

v 

(m,/=) 

16 

L?, .B 

I 

4 

0 

S 

(*) 

.4 

,2 

Experiment 

t (ms) 

100 

Flgure 16. Experlmenlsl snd cslculated accsl€ratlons, 

velocltles and dlrplaccments of the Saab 9000 sltl (clos6 t6 

B-po$t). 

(4) Hallquist, J.O.: DYNA3D, User's Manual, University 

of Califomia, Lawrence Livermore National Laboratory, 

Report UCID-19592, Rev 4, Livermore 1988. 

(5) Benson, D. Hallquist, J.O., Igarashi, M., Shimomaki, 

K., and Mizuno, M.: The Application of DYNA3D in Large 

Scale Crashworthiness Calculations, University of 

California, Lawrence Livermore National Laboratory, 

Report UCRI-94028, Livermore 1986. 

(6) Hughes, T.J.R. and Liu, W.K.: Nonlinear Finite 

Element Analysis of Shells: Part I. Three-Dimensional 

Shells, J. Comp. Meths. Appl" Mechs. Eng,, 27, 1981. 

(7) Belytschko, T. and Tsay, C.S.: Explicit Algorithms for 

Nonlinear Dynamics of Shells, in ASME AMD-48, 198I. 

(8) Hallquist, J.O. and Benson, D.: DYNASD, a 

Computer Code for Crashworthiness Engineering, ASKA 

user's conference, Baden-Baden 1986. 

(9) Chedmail, J.F., Du Bois, P., Pickett, A.K., Haug, E., 

Dagba, B., and Winkelm-itler, G.: Numerical Techniques, 

Experimental Validation and Industrial Applications of 

Structural Impact and Crashworthiness Analysis with 

Supercomputers for the Automotive Industries, in 

Supercomputer Applications in Automotive Research and 

Engineering Development (Ed. Marino, C.), Computational 

Mechanics Publications, Southampton 1986. 

(10) Nilsson, L. and Larsson, M.: Saab Crash programe 

Route to Customer Safety, The Saab-Scania Griffin, 

Linkiiping 1987. 

(ll) Zhong, Z.: On Contact-lmpact Problems, Thesis 

178, Department of Mechanical Engineering, Linkiiping 

University, Linktiping I 988. 

(12) Oldenburg, M.: Finite Element Analysis of Thin- 

Walled Structures Subjected to lmpact Loadings, 

357

Thesis 1988:69D, Department of Mechanical Engineering, 

LuleA University of Technology, Lulefr 1989. 

(13) Brown, B.E. and Hallquist, J.O.: TAURUS-An 

Interactive Post-Processor for the Analysis Codes NIKE3D, 

DYNA3D, TACO3D, and GEMINI, University of 

California, Lawrence Livermore National Laboratory, 

Report UCID-19392, Rev. l, Livermore 1984. 

(14) Christiansen, H.: MOVIE.BYU-a General Purpose 

Computer Graphics System, Civil Engineering Department, 

Brigham Young University, Salt Lake Cily 1984. 

Intrusion Effects on Steering Assembly Performance in Frontal Crash Testing 


Carl Raglando 

National Highway Traffic Safety Administration, 

U.S. Department of Transportf,tion 

Gayle Klemer, 

Research and Special Programs Administration, 


Abstract 

Dynamic intrusion of the steering assembly into the occupant 

compartment of a car has a significant effect on the 

injuries sustained by the driver. In an effort to quantify the 

injuries caused by this component, two subcompact vehicle 

models were tested at 25 and 35 miles per hour in full frontal 

and half offset barrier impacts. The two models were chosen 

from those tested under the New Car Assessment Program 

(35 mile per hour barrier tests) because they have very 

similar crash pulses but dissimilar steering assembly intrusion. 

This paper compares the differences between the two 

models in physical characteristics, $teering assembly design, 

occupant injury data and occupant kinematics in a 

program of eight tests using production vehicles and a 50th 

percentile unrestrained male Hybrid III driverdummy. Data 

from electronic tran$ducers and high speed film analyses 

are presented. Conclusions are drawn about the design of 

the steering assembly for mitigation of harm to the driver. 


The objective of Steering Assembly/Frontal 

Structure test program wart to as$ess the crash 

characteristic$ of the steering assembly and its effect on 

unrestrained drivers in frontal impacts. New Car 

Assessment Program (NCAP) 35 mph frontal barrier test 

data were used to $elect two vehicles which have similar 

crash pulses, but dissimilar steering column performance in 

terms of dynamic intrusion. The '87 Hyundai Excel and the 

'87 

Toyota Celica were selected on this basis. The crash 

pulses of both cars were similar, but the horizontal 

component of steering column dynamic intrusion was 

greater in the Hyundai than in the Toyota. 

The purpose of these tests was to determine the steering 

column behavior within a range of crash conditions using an 

unre$trained Hybrid III driver dummy. Additionally, 

mechanisms of intrusion were investigated to gain insights 

358 

into means of lowering the potential for occupant injuries 

caused by intrusion of the steering assembly into the 

occupant compartment. A test program of eight standard 

production '87 Hyundai Excel and '87 Toyota Celica 

vehicles was conducted. The matrix consisted of 35 mph 

and 25 mph full barrier and half barrier tests, as outlined in 

table 1. 

Table 1. Summary of test condltlons. 

Tart , HoCal GFAth H6dt 

2 

3 

E 

; 

I 

a 

Hyunoar Excr r 

Toyot! cGlrca 

Toyotr Crllcr 

Hycfrddl 

Eic+l 

Tdyota CGI I cr 

Hyundrl Excrl 

Yotett Crltct 

hyundrl Excrl 

Vehicle Characteristics 

To completely understand the steering assembly behavior 

of the vehicles, their crash characteristics are examined and 

compared. The information given in table 2 describes the 

physical characteristics of the vehicle and the test 

conditions for each test. The first column gives the test 

number in chronological order. The second column 

describes the vehicle make tested and the crash mode: FF for 

full frontal and HO for half offset frontal. The wheelbase is 

in the third column, with the fourth column giving the 

distance between the centerline of the front axle and the 

front bumper which is intended to approximate the available 

crush of the front structure. These two characteristics are 

given only once for each make since all cars of the same 

make were identical. The test weight, given in column four, 

includes the vehicle. vehicle instrumentation and the 

Table 2. Vehlcle chsracterlgtlcs. 

e5 nph 

55 mph 

36 mph 

23 nph 

23 mph 

35 nFh 

35 nih 

Ful I fFontrr 

Ful I lFqnttl 

Ful I lFont.l 

Ful I fFontgl 

Hal a ollict 

HrI I ollret 

Hsrt offrrt 

Hal t dllset 

l|odel/ tfheel Fr,Axle- Tcst Test 

l{o, Crash }lode be3e BunDer It, SDeed 

I nchcs I nches I bs ftbh 

I 

z 

3 

4 

5 

6 

7 

I 

Hyundal /FF 

Toyota/ FF 

Toyotil/FF 

Hyundr'l /FF 

Toyotr/H0 

Hyundel /H0 

Toyota/H0 

Hyundel /H0 

93,7 

99. I 

33. 3 

38. r 

e,660 

2 ,700 

?,750 

?,650 

2,780 

2.650 

?,760 

2.6?0 

24,7 

25.0 

ss,4 

34.8 

24,6 

24.5 

34,7 

34,7

dummy and its instrumentation. The test speed was required 

to remain within i0.5 mph of the nominal 25 and 35 mph. 

Table 3 contains the data describing the performance of 

the frontal structure of the vehicle used in each tesf. This 

table is arranged so that one can readily compare Hyundai/ 

Toyota pairs at each speed in each crash mode. The 

following abbreviations are used to identify the tests: 25 or 

35 for the nominal test speed in mph, FF for full frontal 

crash mode, HO for half offset frontal crash mode, H for 

Hyundai and T for Toyota. Maximum passenger 

compafiment acceleration in g's, maximum frontal dynamic 

crush, peak load cell barrier force, and the time (in 

milliseconds) at which each occurred are given. The 

passenger compartment acceleration was recorded by an 

accelerometer mounted On the rear seat $upport structure. 

The acceleration pulse was filtered through a Bufierworth 

double pass filter to -3db at 30 HZ curoff frequency, with a 

15 HZ corner and a 75 HZ stop frequency. The crush was 

then computed by double integration of the acceleration 

curve. The load cell barrier (LCB) force is the sum of the 

force histories of each of the 28 load cells used on the 

banier, 20 cells in the case of the half offset tests. The force 

summation was filtered through a Bufterworth double pass 

filter to -3db at 100 HZ cutoff frequency, with a 5O HZ 

corner and a 250 HZ stop frequency. This frequency wa$ 

chosen for the load cells due to their inherently lower 

frequency content (as compared to accelerometers in the 

vehicle) and due to the filtering effect of summing the 

channels of data (tending to cancel out random noise). 

Given last is the crush at the time peak LCB force occurred. 

Table 4 shows variables calculated from the data in table 3. 

Crush and the peak force from table 3 were used to compute 

frontal stiffness, the first variable in table 4- Frontal 

Table 3. Msasured crash date. 

FF35 H 

It Fr 

EF33 f 

Dt ht 

EOti H 

O+ mr 

EO?s T 

It mt 

HO'3 H 

ot mi 

t|oSE T 

3! mr 

unr * i 

f+;EE,1n-+TTFEEllgr, 

Tabh 4. Calculsted crash datB. 

FFzf H 

FFE3 T 

FFf3 H 

FFI3 T 

of23 H 

OFZE T 

OFI: H 

6F3E T 

* Ati 

EllIEg'TI l+*tll'g-1s66'-. 

14. t{ 

r6re6 

ae-1a 

zo-ro 

2o.27 

16.04 

2L-11 

stiffness, given in thousands of pounds per inch of crush, is 

followed by the maximum crush as a percentage of the total 

frontal distance available forcrush. The last column in table 

4 is the maximum energy absorbed by the frontal structure 

calculated by integrating the force displacement curae. 

In addition to the data presented in tables 3 and 4, plots of 

the passsenger compartment acceleration (crash pulse), 

velocity and displacement over time are helpful in 

comparing the two vehicles. Figure I compares the 25 and 

35 mph full frontal pulses of the Hyundai and Toyota, while 

figures 2 and 3 compare velocity and displacement, 

respectively, obtained by integrating the crash pulse. 

Figures 4, 5 and 6 make the same comparisons for the half 

offset crash mode. From these curves we can see that these 

two vehicles have reasonably similar crash pulses, with the 

largest variation occurring in the 35 mph tests. The effect of 

these variances will be further discussed 

in the section on 

Occupant Kinematics. 

It is important also to look at the vehicles' energy 

absorption capabilities. Figure 7 shows the comparison of 

Hyundai and Toyota force-deflection curves for 25 and 35 

mph full frontal tests. Figure 8 shows the integrals of these 

force-deflections. It is clear at 25 mph that the frontal 

structure of the Toyota absorbs more energy than does the 

Hyundai in this crash mode. These 25 mph curves also 

indicate that the Toyota has the potential to absorb more 

energy at higher crush. Energy absorption is the same up to 

approximately 3 inches of crush. After that point the Toyota 

absorbs more energy for every increment of increased 

crush. 

Figures 9 and l0 help to explain the above differences. 

Figure 9 is a pretest under,body view of the engine 

compartflent of a Celica; figure l0 is the same view of an 

Excel. Notice the T-shaped assembly (front suspension 

crossmember and center engine support member) in the 

Toyota. The Toyota lower control arms also provide added 

structural rigidity, at least from the wheels aft, due to their 

trailing longitudinal orientation and their attachment to the 

floorpan, This construction indicates that the Toyota may be 

structurally stronger in the latterpart of the crush. However, 

at 35 mph, the peak LCB force is essentially the same forthe 

two cars, but the peak force occurs slightly earlier in the 

Hyundai at approximately 16 inches of crush. This peak in 

the Hyundai did not occur at the lower speed, even though 

the crush exceeded 16 inches. This indicates that engine 

impulse or some other factor, rather than structural strength, 

influenced the Hyundai peak force. The reasons will be 

investigated in later discussions comparing crash speeds 

and crash modes. 

Comparison of structural performance in the half offset 

crash mode shows less difference tretween the two makes. 

Figure I I contains plots of the force-deflection curves 

resulting from the 25 and 35 mph half offset batrier tests. 

Figure 12 shows the corresponding energy absorption 

curve$. Again the Toyota frontal structure absorbs more 

energy at 25 mph than does the Hyundai, but the difference 

is less pronounced. In the 35 mph case the energy absorption 

359

tt rr* Ftll lm;nl H1r'ld,ti Ercel w. tqon C.UH lf rpt Frlll lrclittl Hyr'l,&ll E*d rlt. folo|r Ctl|r'r. 

0 25 SO t} 100 

Ilmr (mr) 

Figure 1. Crash pulse lor full lrontal tests. 

2t l;h PilI ltratd Hyr:drl Errrll ot Toyot: HE tt r;t Ftll ftot|rt Hystdtl Excrl pr. ttlotr Ctltcr' 

eg 

i 

t 

t'r (c) 

Flgure 2. Veloclty curves for full frontal tests' 

360 

- l{yurdd Ercd 

- lopto Cr$co 

- H|.Era 

||l rdCt 

+a-a.----l--tt'lr' 

r ro t t a t t rr i f'drsfrf.rrfrf,.f 

a 

it 

c 

o - 

oL 

! t( l - 

{t 

3 t 

I 

- llWndoi Excal 

- Toyuto Crrlico 

a 

a 

a/ 

IT 

r.-r----g: 

o ro f t 4 tP f rt f +*d..t.f*f.f.frf,rf 

tr 

(m)

-n 

u 

rt rFt ?:ll l'|f,td ''lllr.&j tsd rr. Tq'li. c'Glfu tll| ft* Fsll lto:td llystt/,tl Err:ll rr. ?qo]. C.zillc. 

{ 

-FFt hrr 

-- 

- 

c& 

1r.r.._._.t 

.t 

T- 

I 

l5 ;pt Htll clftct Hyr,tthl Ercrl or. Tqctn Ctl/rct Jll npl Hdl cflfi HyarJnl Ercrtr ot. toyolt Crrlkr. 

6 -to 

o 

! a!f -20 

g 

{ 

-J0 

r f t t a a a rr a ilt'rfrildrtrtfril | + + f a ] a r| t tf.|.rtdrl|rlrf1* 

rm (al 

- Hygn{oi Erccl 

r loyoto Crl'co 

_,10 

0 23 50 lJ r00 

Tlmr (nr) 

Figure 4. Crash pulsea lor hall off8et tests. 

- Xyg6Ooi Ercrl 

- tof.sto Cafco 

il 

I 

I 

I 

I t

2E rll I{r.V aUlE, IlytstuI E,rcel o* Toyrclt Cdlu tt r;r Hrll Oooootl.4 Hytr'sd.tl E-ctl sr' liottot. c.lle.. 

ttrr (ml 

Flgure 5. Velocity surves lor half offsst te$ts. 

- 

* a rr a adr{rtrt$19afrf 

|tr. (+) 

ttl nplr I|l.If Ofilr Hyr*ttltt Ercrl lr, Ilqn|n Ccllgr l$ erpl llrlf Ofittl Hltathl Excel ot, fqol. e3rlfl 

lh. lrbl 

Flgure 6. Dlsplacement curvss for hall oflsst tssts. 

362 

r rl t t a I t rr a Jdrsrt+lrr.t.tri 

t rr + t e f t f f {rlr$rfri+l.tta*+ 

o + il t {r |t, f f ft'rfrfrt+F1*rtrrn 

ttm (tl

t 

! ctocg 

J 

I 

o 

o 

l- 

E5 nph lull froatrl llyrndrl lrcrl rr. Toyotr Grllcr 3$ mgh full lrontel ltnlbdrt lrcrl rr. totot| crllcr 

rr0 

r00 

90 

80 

70 

60 

50 

4 

JO 

20 

t0 

o 

70000 

60m0 

30000 

$ 

! 

I {0000 

}r 

PJm00 

o 

c 

lr, 

E Htu,illol ErCd 

r loyoto Cdlr6 

ii 

a, 

a- 

a I 

, I 

,tA 

. l 

\; 

tt 

I 

I 

I 

I 

t 

to to Jo 

Olrphcrrnrnl (ln) 

Flgure 7. Forco vs. cruth plotr for full lrontal tssts, 

- flgrl 

I crlco 

t 

0. 

o 

...../ 

,'/ 

t 

0 

l t 

, 

t 

a 

a 

rt 

a 

E ctoo.g 

J 

a 

u 

o 

il0 

t00 

90 

EO 

fo 

s 

50 

rO 

JO 

20 

to 

o 

:/ 

I 

] 

ir) 

- f,1s61 

7 i 

t, 

l 

t l 

I 

] 

tr 

i'. 

l r , t 

t 

I 

I 

f0 20 JO 

Dlpbcrmrnt 

(ln) 

r Cetico t ttf 

0 

t0 15 20 25 

Displocrmrnf (in) 

- lltsndoi Erccl 

r Isyota Cclcc 

25 mph Full Ftontel Exccl oe, Cclica 35 mph Full Frontal ErceI oa, Ccllcs 

Olrplommrnl (in) 

Flgure 8. Energy ab8orptlon In full frontEl tests. 

..-P 

'rdf 

.,df 

fle 

;d 

g 

.i.dP 

-d 

c

Figure 9. Toyota lront underbody. 

Figure 1 0. Hyundal front undGrbody. 

is essentially equal for the two vehicles until about 23 inches 

ofcrush has occurred. From that point on the Toyota is again 

dissipating more energy than the Hyundai. Also note that 

the Hyundai uses 98.2% of its available frontal length at 

maximum crush at this speed. lt is no coincidence that the 

greatest intrusion is also observed during this test: 4.5 

inches rearward and 3.6 inches upward. (See table 7 

discussed in section IV). 

Figures l3 and 14 show the crash pulse comparison, by 

crash mode, for the Hyundai and Toyota, respectively. Both 

the low speed Hyundai test and the high speed Toyota tests 

had higher crash pulse peaks in the half offset tests. To 

explain this the corresponding set of force-deflection 

curves, figures I 5 and I 6, and energy/displacement curves, 

figures 17 and 18, are examined. The first conspicuous 

difference is that both vehicles are less stiff in the half offset 

crash, as might be expected, but this stiffness difference is 

364 

more pronounced at 25 mph than at 35, and more 

pronounced at 25 in the Toyota than in the Hyundai. The 

second salient feature of the curves is that the peak load cell 

force values generally correlate with peak acceleration (also 

see table 4). One can also see from figures I 5 and I 6 that the 

pairs of force-crush curves fbr the 35 mph Toyota and the 25 

mph Hyundai have very similar shapes as would be 

expected. However the 25 mph Toyota curves and the 35 

mph Hyundai curves do not show this similarity. This 

exception in both cases is that the peak force and 

acceleration in the offset tests were unexpectedly high 

values and occurred late in the event. This shows that 

"bottoming-out" or stiffening occurred when the crush 

exceeded a certain limit available in the front structure, 

causing a high but late crash pulse peak (consistent with the 

force peaks). Since the high speed Hyundai tests did not 

display this phenomenon, it is probable that the energy 

absorbed in both high speed tests as well as the offset 25 

mph test was of sufficient energy to bottom out the 

structure. Referring to table 4 shows that in the low speed 

tests this bottoming out phenomenon occurred somewhere 

between 55*68Va of the reported available crush. ln the high 

speed tests this occurred between 7V98Vo.In the low speed 

tests and the high speed full frontal test, the Toyota showed 

no signs of bottoming out. However in the high speed offset 

test, the structure apparently did bottom out. This indicates 

that the Toyota bottoms outatT0-80Vo of the available crush 

as repofted in table 4. This can perhaps be explained by two 

unique design features of the Toyota. One of these is the 

longitudinal engine support member which also serves to 

absorb frontal crush energy in a frontal impact. In the offset 

impact used in this testing this member was not fully 

engaged, thus softening the structure. The other difference 

is that the Toyota has trailing lower A-arms which add 

considerable stiffness to the structure at about 23 inches of 

collapse distance. Since the Toyota does not collapse this far 

in the half offset 25 mph test and the longitudinal member is 

not resisting collapse, stiffening is not seen at the lower 

speed. These diff'erences further indicate that the Toyota has 

a more effective collapsible structure. 

Comparisons similar to the ones made for crash mode are 

made by crash speed in figures 19 through 24. By 

comparison of force-crush curves at 25 and 35 mph the 

extent ofvelocity rate dependence can be examined. From 

the previous discussion it was discovered that the Hyundai 

structure bottomed-out in three of the four tests, yet the 

Toyota bottomed-out in only the 35 mph offset. Since 

structural forces are unpredictable when the structure is 

fully collapsed, the Toyota full-frontal tests more reliably 

explain the effect ofvelocity on crash pulse and force-crush. 

It can be seen from comparing the Toyota full frontal tests in 

figure 20, that the crash pulses are very similar in shape' the 

peak in the 35 mph test is higher, and the duration longer 

than in the 25 mph test. By looking at figure 22, it can be 

seen that the force-crush curves are also very similar, until 

the load drops off in the 25 mph test.

8i rlpb lhll olht Hyundri lrcel rr. Toyol.r Cdtcr lS nph llell olf*t Hyrodri Ercrl vr. Toyotr Ccller 

Ir0 

too 

90 

eo 

t 

Eto 

t 

o 

50 

:t 

Jto 

t 

g.o 

o 

t- 

J{r 

20 

l0 

0 

r llWlrdoi Ererl 

- tqsra Cdica 

,_rj 

, 

a, 

a 

t 

I 

I 

I 

I 

I 

I 

I 

0frphcrmrnl (ln) 

Flgure 11. Crash pulses lor Hyundal Excel full frontsl vs, hslf offact teete. 

10 

Dllplocrmrni (ln) 

25 mph HeA Offtct Exccl oe, C,clica 35 mph HaA Qffsct Ercel os. Celica 

50000 

t 

4 | 40ffi 

h 

gJ000o 

o 

g 

trJ 

?0000 

- f41l 

t C.Ico 

t 

,fr 

, 

ll 

0 

t, 

f t 

) 

o 

f7 

0 5 t 0 t 5 2 0 

Dhpfocrrrrnf (ln) 

Flgure 18. Energy abaorptlon In half off$Gt tc$ts. 

, 

2J 

|r 

il0 

t00 

90 

EO 

Erc 

t 

o 

o. 60 

o 

.= 

Sso 

E.o 

o 

lr 

'df 

Jo 

20 

t0 

0 

'dP 

.rdf 

fr- 

:'-e 

P 

,i -dFP 

d 

c 0 

- l+nrndoi frGd 

- IoFtq Cetico 

o 

a, 

r EICCI 

r Crlho 

to 

t 

, r 

f + t 

l r 

y r 

Â. I 

, I \ l 

l r 

, 

,, 

l r 

tr, 

\./i I 

I 

I 

I 

I 

I 

I , 

t0 t5 20 ?s 

Displocrmrnt (in) 

o 

,? 

J5 

365

i 

o 

C 

o 

I 

t I 

tl 

(t 

{ 

HFn&t Erorl l, *fh Fr/ll fmll.ttl ct' Hr,/f ofrrlt Hytntst Brrrll l8 lilh F:ll ftoilrd 

n Fsll lrutlol 

- lloll ollrrt 

r{ir.r- 

I. 

, ra 

fr.t^ -r 

\ 

I ] 

tt 

\. r t' 

\ r I ; 

\L /; 

5rt 

- Fdl lrsrtC 

r lHf oltrt 

ir t's ro is lm | 25 ls, I ts 25 r5 t00 

Iimr (rnr) 

llmr (mr) 

,*ta 

Figurel3.Grashpu|sestorHyundalExcellu|||rontalvs.haliolfsettests. 

6 

i) 

c o 

o 

o 

o 

rt 

g 

t0 

-t0 

-J) 

Toyott Ccllcc 2tl ;lph Frll ftonttl w' HrJf offr,l To1ot. Ctllct 35 ;plr Ftlt ftonttl 

timr (mr) 

Flgure 14. Crash pulaes for TOyOta Gelica tull frontal v3. half offset te$ts. 

366 

- Full frontol 

r Hdf oflrrt 

It 

' l 

l 

t 

t: 

t 

(, 

E 

-9* 

E I 

t r + 

(, 

{ 

I 

it 

C 

t t 

, l 

a u 

rit' Htf ollttl 

.ff. 

ot Htlf olltct 

ti 100 

i ts io 7s loo l?5 l5o l?5 25 rt0 r73 

l0 

0 

o -t0 

o 

L 

t 

t 

o - 

o 

{ 

r Full lrontsl 

- Holf offt.t 

I 

Iimr (mr) 

t75

t 

! ctooo 

J 

t 

|, 

o 

lr 

Hyuodrl kcrl 2$ rnph Full lrontrl vr. llrll olhrt Hyundrl Ercrl 35 mph hrll frontrl rr. Hrll otfrct 

tt0 

t00 

90 

8{t 

70 

60 

50 

{O 

JO 

20 

I Fuill fruntol 

- tlcll oftrrt 

!t 

tr 

i 

t 

I 

I 

I 

a 

I 

a TI 

a 

I 

I 

I 

I 

a 

Ohplecrmrnl (ln) 

il0 

r00 

90 

^8o 

I 

E c 7 0 

I 

o 

s" 60 

350 

o 

g + 0 

o 

E J o 

Flgure 1 5. Force ve. crush plots for Hyundal Excel full frontrl vs. half ofl$et terts. 

t 

! EI0oo 

J 

I 

u 

o 

Toyolr Cellcr 86 mph lull frootd vr. Hrll olfrct 

il0 

90 

80 

to 

60 

20 

t0 

0 

- Full lrq.rtql 

r Holl ott3rt 

0lrphornrnf (ln) 

Figure 16. Forcc ys- crush plota for Toyota cellca full lrontal vs. h8lf ofl'tt tssts. 

t 

v Etotro 

I 

I 

u 

o 

20 

t0 

o 

ro Jo 

Dhplocrmrnt (ln) 

ToyotrCcllcr 

35 mph Full frontrl vr. Hdl offret 

it 

,tt 

o 

o 

a 

p frrll frcnlol 

r llolf offtcl 

[ ; 

o 

a 

I 

t l 

t l 

/ t 

, 

a 

I 

a 

tt 

t 

I 

m 

Obplocrmrnl (ln) 

t 

- Full t|Untdl 

- itoll ctttrt 

IL 

t I 

I 

I 

I 

a 

I 

a 

at t I 

I 

I 

I 

I 

I 

a 

.0 

\

7 

romo 

6mo 

?0000 

tmo 

0 0 5 t o t 5 2 0 2 5 

Olgbcrmrnf (in) 

Flgure 1 7. Energy absorptlon by craeh mode lor the Hyundal Excel. 

368 

25 mph FuIl Frcntil os. Half Off*t 35 mph F14II Frcntal ot. Half olfwt 

I 

a 

lmmo 

fo 15 20 25 

Dirplocrmrnl (ln) 

25 mph Full Frontal os, HaA Off*t 35 mph FUII Frontal oe. HtU Oflset 

70000 

60ffi 

i0000 

I 

e 

I 10000 

t 

9J0000 

t 

C 

lr, 

2m00 

s Full Froatol 

tr Holl Oflrt 

t' 

rrf 

t 

t 

a 

t 

, 

0 

0 

, 

t 

0 

, 

o 

, 

o 

r0000 /rr 

t 

, 

0 

0 i 1 0 1 5 2 0 2 5 

ltirpfommrnf (ln) 

Flgure 18. Energy absorptlon by craeh mode for thc Toyota Gellca. 

.d 

.df 

t'r-f 

Re 

:p tr 

;+e 

"dfP 

"dtr 

..*p 

'*d 

,r-P 

F.e 

:#r 

o 

ioP 

.sP 

,d 

- Futt Fronlol 

- ndl Offr.t 

{," 

- Full Frontol 

r rhtl O{lut 

{,* 

, 

O 

f o 

T ' 

7 a 

t 

o 

, 

, 

I 

t 

, 

I 

I 

, 

r0 15 20 z5 

Dirplommrnt (ln) 

i, 

o 

o 

,o 

rqt 

o 

o 

a 

';t 

J5

Hyrndrl Ercd Frll frortrl tS r7t ur. 15 ;pt Hyr;drl E.rcd ttr,$ olftril 25 sph ur. t.f rrplr 

- ZJ rriptr 

rt J5 ffiph 

0 25 50 73 r00 t25 t5o l7S 

Iimr (mr) 

Figure 1L Crarh pulaea lor Hyundal ExcBl 25 mph vs. gE mph t6st$. 

.r 

C| 

C o 

g 

.l 

It, 

o 

{ 

t5 tm 

fimr (mr) 

Toyott Cclicr Fill ftoaul 28 lllph ot, 35 lllllh Tuyotn Ccllct Helf offtc| 28 nyh or. 35 ;p[ 

r ?S r?rph 

r J$ mpn 

It 

lr a 

f 

li 

t; 

I 

l 

, 

TT 

I. 

E 

o -r0 

7t t00 r75 25 t5 t00 

finr (mr) 

Flgure 20. Crash pulses for Toyota Cellca 25 mph vs. 35 mph t$ts. 

tirnc (mr) 

t 

o 

o 

t 

I 

() 

u - 

{ 

- 25 trph 

I JJ nptr 

i\ 

I 

I 

I 

t t 

/l t 

/; 

I 

a 

I 

, 

I 

r25 

aallrt 

t25 t50 r15 

\

7 

llyundri frcrl lrll frenlrl Ei lpb rr, $S ntph Hyrndri Exccl llrll olfrct Ei rnph vr. 35 mph 

'ilo 

Dlrplocrmrnl (h) 

Flgurc 21. Forcc vc. crush plot$ lor Hyundal Excel 25 mph Yt. 35 mph le8t$. 

t0 J0 

Dlrplocrmrnl (ln) 

Toyotr Cclicr Full frontrl 25 mph w' S$ mph Toptr Cclicr Hrlf olbct ?5 mph vr. 35 mph 

Olrplocrmrnf (ln) 

Flgurc 22. Forcc vr. crush plots tor Toyot! Cellca 25 mPh vs. 35 mph tt8ts. 

370 

r00 

t0 

0 

r lt 6jrft 

r ,Lmdr 

I 

90 

a 

t 

t0 

f 

t 

l 

Ero 

I 

I 

o 

I 

s60 

350 o .'n 

, 

E.o 

o 

rr ir 

J{) 

it 20 

il0 

t00 

90 

t0 

a 

lro 

f 

o 

F60 

*io 

t 

g { 0 

o 

lr 

Jo 

20 

t0 

0 

- 25 flUh 

r Ji rilfh 

I 

I 

I 

I t 

rll 

a l 

I 

t ; 

t ; 

t : 

\ \ ; 

l1 

t l 

t l 

trt 

I 

I 

I 

I 

I 

I 

I 

f 

I 

I 

a 

I 

a 

I 

E CIoco 

= J 

a 

It 

o 

b 

t 

r0 

00 

90 

80 

r0 

€o 

ll0 

t00 

90 

80 

E n 

J 

o o 6 0 

g 

gio 

! o 

d 

l* Jo 

20 

* 25 mplr 

r Jg lrtph 

- !$ mph 

r J5 mptt 

t 

It 

l r 

1 1 - 

I t l r 

f t t 

! n 

,'Fl \ 

20 

t 

I 

I I r 

'aatt 

a 

; r 

I t t 

t l 

t r 

: i 

c ; t 

l,^4 

0lrplomrnrnf (ln) 

i 

I 

I 

I 

I 

I 

a 

, 

l r 

l i 

t ; l_ _ __ _^.

.d 

{'"* 

.'Ff 

.-*f 

;--" 

I 

Jd 

E*cd FUII Frontd 15 lrph u.1, ZE nph Exccl Hr,$ Offtct 35 mph u,t. 25 mph 

"Sf 

-dtt 

0 

H tr$ rvrfh 

r 25 fieft 

r 

,rir, 

0 5 r 0 r $ 2 0 e 5 J 0 I J r0 t3 20 25 

Olrplommral (ln) 

Dlrplocrmrnt (ln) 

Flgure 23. Energy Abmrptlon by lmpact yrloclty for the Hyundal Ercel. 

rd 

rtdf 

Cclice Fu/l Erontd tE mph ot, 25 mph EcIIca Half Offrct 35 mph ot, 25 mph 

t-df 

He 

;"df 

e 

Jd 

.df 

-df 

- Jg 6pl 

- 23 ttplr 

tt 

0 5 t o r 5 ? o 2 t t J 0 J 5 r0 15 20 25 

Dlrplocrmrnl (la) 

Flgurc 84. Encrgy rbrorptlon by lmprct veloclty for thr Toyotr Gellcr. 

r--P 

.tof 

tt-f 

f,"-* 

:df 

g 

;d 

"dtr 

"df 

.df 

..dP 

'"-f 

f,,e 

:*P 

E 

;d 

.dtr 

"*F 

- JS mfrlr 

-t 2i nplr 

- J$ ;6p1 

r ?5 nrpfr 

7o 

a 

o 

rl 

I 

Olrplmrnrnt (ln) 

J5

Having discussed and compared the frontal structure, we 

will go on to explore the differences in occupant injury 

between the Hyundai Excel and the Toyota Celica in these 

eight tests. 

Occupant Injury Data 

An unrestrained 50th percentile male Hybrid III dummy 

was used in the driver position in all eight tests. To facilitate 

better camera coverage of the driver dummy's interaction 

with the steering assembly, no dummy was used in the right 

front seat position. Table 5 $ummarizes the occupant injury 

data recorded by electronic instrumentation and the 

laceration index (LI). Analysis of data from test films is 

presented in Section IV. 

Table 5. Occupant lnlury data. 

rest H,c EEll' i8?tl'li8f,i [FEl EEg: Lr 

FFES H 

ot- 

FFAS T 

ot 

FF35 H 

ot 

FF3 6oT 

HOES H 

9t 

HO25 T 

ot 

HO35 H 

ot 

HO35or 

r tl dnd te rFG dlven for I 

unltE: Loadr lF DouEdt 

DlcDIrccfrint ln lnchar 

Tlnb (t) ln nlltlrecgnds 

Lrccrdtlon IndGx (LI, lt unltliil 

The Head Injury Criterion (HIC) was calculated from the 

resultant of the three central accelerometers in a nineaccelerometer 

array. The maximum chest g's is the 

maximum acceleration of the torso which has duration 

greater than three milliseconds. Maximum femur load, 

shown in table 4, is the maximum axial compressive forcu in 

pounds measured by the femur load cell. The maximum 

abdominal displacement is calculated from data measured 

by a pressure transducer in the Hybrid III. The conversion 

formula between pressure in pounds per square inch (x) and 

displacement in inches (D) for the abdominal insert used in 

this dummy is: 

D = (0.00047288x5 - 0.02616x4 + 0.53471x3 

- 5.0362x2 + 26.981x + 5.31)125.4. 

Positive chest displacement is recorded by a 

potentiometer when a plate at the front of the chest is 

displaced toward the spine. The maximum displacement is 

shown in table 4 in inches. 

The laceration index is a comparative measure of 

laceration and abrasion of the face and scalp of the 

occupant. It is based on the number, length and depth ofcuts 

made, during the crash, in two layers of chamois fitted over 

the face and the top of the head of the dummy. An LI of I 

indicates that there were abra$ions to the outer chamois, but 

no penetrating lacerations. 

312 

1451 

BE: tE 

,111, 

66.25 

I 424 

74. r8 

a9-9? 

ll?2"' 

65 .47 

9lo 

?[i16" 

469 

t3.47 

toz, oo 

IB55 

t3:i6 

18llo 

92.77 

54.5 

49.1 

78.2 

lo5. a 

44. O 

45. O 

6+. O 

69. O 

r67 r 

r 039 

e5a3 

e3 t6 

l5l I 

103r 

2920 

1550 

l20a 

1463 

1676 

1409 

l20t 

I 65i 

I 796 

z.za 

90. o 

Eolt 

1 .69 

s3.77 

b16' 

1.93 Er?8. at.7e 

ioSlaz 

1.74 

97.72 

2.45 

79.27 

2.05 

95-25 

t.7l 

79 .95 

&r?tu 

ti?B' 

I .45 

99.07 

r.12 

65.97 

1 .17 

7J.tZ, 

1.00 

8, 29 

1. OO 

7.78 

r.oo 

4. 59 

l. oo 

3. 90 

The parts of the body most affected by the steering 

assembly are the head and chest. We begin the discussion of 

occupant data with those indicating head injury. From table 

5 one can see that, in general, the driver of the Toyota 

experiences a considerably lower HIC than the driver of the 

Hyundai. Also note that the laceration index for the Toyota 

driver is always quite high, while the LI for the Hyundai 

driver is never greater than I (abrasions only). This would 

seem to indicate that head injury to the Toyota driver results 

primarily from contact with the windshield while the 

Hyundai driver seldom hits the windshield; his injuries are 

caused by hitting something harder, the instrument panel, or 

steering wheel hub. The one exception to this is the HIC of 

1586 for the Toyota driver in the 35 mph full frontal test' 

This is borne out in the film analysis and the exception 

explained in the next section. 

Two observations predominate when looking at the 

maximum chest accelerations reported in table 5. With such 

similar crash pulses we could expect that since the Hyundai 

driver had a higher chest acceleration that the Toyota driver 

at 25 mph, the same would be true at 35 mph. However, 

quite the oppo$ite was recorded. In fact, the increase in three 

millisecond clip chest acceleration for the Toyota driver 

from the ?5 mph test to the 35 mph test is extremely out of 

line with the increase for the Hyundai driver. This points to a 

difference in the function of some part of the Toyota interior 

at the higher speed. The second oddity is that while the chest 

acceleration decreases in the half offset mode in both cars at 

25 mph, at 35 mph the chest g's for the Hyundai driver are 

greater than in the full frontal mode but the Toyota driver's 

chest acceleration is greatly reduced from the full frontal 

case. To adequately explain these seemingly incongruous 

data we must rely on analysis of high speed films of the tests 

to detail the kinematics of the dummy and vehicle interior 

surfaces and supplement what is known from electronic 

instrumentation. 

Occupant Kinematics 

Film motion analysis was conducted to determine 

kinematics of the dummy and its interaction with the 

steering assembly, instrument panel, windshield and other 

interior components. Figures 25 through 32 show the 

relative positions of the dummy and steering wheel/interior 

for the drivers at important times throughout the eYent. A 

summary of the dummy kinematics is presented in table 6. 

Relative velocities between the dummy and struck 

surface at contact time (known as "contact velocity") were 

evaluated as one of the parameters affecting occupant 

injury. Contact velocity was determined by adding the 

Table 6. Summary of drlver dummy klnematlcs. 

chcrt/iub 

H.rd/rln 

Hc.d/I. P. 

6'l 

8a 

Evint Tlffi - ill ll lracondr 

-H FFti-fl FF25-T FF35-T HO25-[ Ho3r-f, rcUr-T XOrt-T 

58 

50 

,1 

64 

69 

al 

at 

5l 

6l 

t1 

77 

73 

89 

67 

undctchlnad 

58 

57 

la 

f4 

73 

t3 

66 

55

ftdcrrn sontrott rlt 0 EBr? rr 

r l r e c S l 

Flgure 25. Dummy klnematlc plote of llyundal 2$ mph lull frontal. 

chln oantrotr rln rt t3.6 rr 

chmt cmtrctr hub rt E{.9 nr l,lorr controtr inrtrunrnt prrrl rt !t.G rr 

ec 

gl 

373

*aolrt cqrtrotr rir rt El.E n 

Elrtn ccrt*tr tcp of *rrl rt E8,t r 

I r r i l i l l { l 

Flgure 26. Dummy klnematlc plota ol Hyundal 35 mph full frontal. 

374 

Ehrrt cqrtrtr lrrb rt E.l.E rr 

Hrn ntrtr lrrtrurrrt prml rt ?4.1 rr

Sdorn contetr nln rt 68,e rl 

chln cont.rotr rh rt GrrB r. 

/q 

r r r t t S r a c 

FlgurcZ7. Dummy lrlnomstlc plots of Toyots ZS mph full frontal. 

cll|rt srtrctr lrJb rt f,l.E rr 

Ehir hltr upprr IPz h.rd cFrekd tt 6r.i

c 

RFdiln ca*,str rlr rt El.E m 

Figure 28. Dummy klnematic plots ot Toyota 35 mph full frontal' 

376 

Ohrrt/chln srrtrot rlr rt Gc.t rr 

ee !t 

H,rck controtr rln rt ??.a ilr

Hilr| srtrtr rl&fird art|ctr r/r rt n.l il. chln eentrotr rlr rt ?t.t rr 

Clrt cantmtr lrrE rt tt.l r mr contttr lnrtnnat Frtl rt Cg.O il 

r l r l l a r 

Flgure 29. Dummy klnematlc plot$ ot Hyundal 25 mph half offrst.

fSdoffi c*rtrcEr rir rt tlf.l rr 

Chrrt eoatrctr hub rt F?.t rr 

Flgure 30. Dummy klnematic plots of Hyundal 35 mPh half oflsot. 

378 

Chin oontrctr rir rt EG,E rr 

)hrd cmtrotr lnrtnrrnt prrrl rt ?4.1 rr

t 

filtln cflrtstr rir rt CC. l rr Ehin eontrstr rh rt ?t.i rr 

chrrt eqrtrtr lr.S rt ?tt.t rr l|rd srt*tr lnrtnnnt Frrrl, rt Bi!.9 rrr 

Flgure 31. Dummy klnematlc plots ot Toyote 25 mph hall oftsct. 

379

Sdsm lartrotr rh rt EB.? rr Hrrd Edrtrctr rln rt EE.l r. 

b) 

-{-d { 

jv 

*/," 

0hrrt ccrtrctr hrb rt ll.t rr 

./,, 

Flgure 32. Dummy klnematlc plots of Toyota 3$ mph half otfeet. 

380 

lhrd rtrlku rlndrtrirld rt ll.3 rr

velocity, at the time of chest contact, of an unrestrained one 

mass object (relative to the compartment) to the film 

determined velocity of the column. Figures 33 through 36 

are the contact velocity curves derived from the crash pulses 

of the eight crash tests and table 7 summarizes the steering 

column film analysis results. The resulting total contact 

velocities are shown in table 8. This method of determining 

contact velocity wa$ chosen since film analysis distortion 

effor was unavoidable with a wide angle lens used on the onboard 

camera. Additionally, the dummy's shoulder or chest 

target was not always clearly visible in the views of the offboard 

camera$, precluding their use for all tests. 

Comparisons are also shown in table 8 between contact 

velocities obtained using this method and contact velocities 

from film analysis only for the full frontal tests. There is 

generally very good agreement. Another use of the contact 

velocity derived from the crash pulse is to quantify the crash 

pulse differences for the specific contact times observed for 

the drivers in this test series. 

The first set of tests analyzed are the full frontal te$ts at 

the 25 mph crash speed. Figures 25 and ?7 show the dummy 

kinematics in these tests. For the Hyundai full frontal test at 

25 mph, the che$t to hub contact velocity, from table 8, is 

24.5 mph (including 2.4 mph rearward velocity of the 

steering assembly) at 65 milliseconds. There was 2.0 inches 

of rearward dynamic intrusion before the dummy contacted 

the wheel and the column/wheel began absorbing energy by 

stroking. ln the Toyota, contact velocity was 25.8 mph at 64 

milliseconds. There was a very small horizontal velocity 

component of intrusion of 0.2 mph and a correspondingly 

small horizontal intrusion component of 0.2 inches. Contact 

velocity fpr the Toyota compares reasonably well, though 

slightly higher, with that calculated for the Hyundai, despite 

Hytrl.dtl Erccl 25 tsph Ftll ftanltl 

f EJO 

lr 

E 

!ro 

|t 

t Goro 

fl -.rlfl lilr iliH dr lhl h 

1$ 

1$ 

f rP + 0 ro f, frdrrorf 

Gantoc,t tlmr (rnrl 

? a 

g E 

s ; 

C E 

s 3 

€ E 

f t , 

E € 

F O 

Flgure 33. Contact veloclty snd tr8vsl cuiyaa lor ?5 mph lull frontal teslE' 

t0 

the intrusion difference, since the intrusion in the Hyundai 

caused earlier contact than would have been expected 

otherwise. 

In the Toyota the three millisecond clipped chest 

acceleration was 49.1 g's, compared to 54.5 g's in the 

Hyundai at 25 mph. This shows that one should not 

conclude from the Hyundai and Toyota 25 mph tests that 

lower contact velocity caused by intrusion is desirable for 

improved occupant re$pon$e. Though the response 

difference is not large, it is especially significant in light of 

the fact that the Hyundai column stroked more than the 

Toyota's, see summary of post-test column measureflents 

in tabl€ 9. Therefore, steering assembly intrusion may 

adversely affect dummy response during dummy ride down 

and after the initial impulse is imparted to the dummy 

through the wheel. 

The next set oftests analyzed are the full frontal tests at 35 

mph. Figures 26 and 28 show the dummy kinematics in 

these tests. For the Hyundai test, the contact velocity is 30.7 

mph (including 3.6 mph horizontal velocity of the steering 

column and 3.0 inches dynamic intrusion). Chest contact 

occurred in the Hyundai crash test at 58 milliseconds. 

Contact velocity for the Toyota was 30.1 mph at 6l 

milliseconds. There was a very small horizontal velocity 

component of intrusion of 1.3 mph for the Toyota. These 

velocities, like the 25 mph test also compare reasonably 

well with one another. Part of the reason for the similarity in 

total contact velocity, from table 8, despite much higher 

intrusion velocity in the Hyundai is that contact occurred 

slightly earlier in the Hyundai due to the intrusion, as 

compared to the Toyota. Therefore, contact velocity was 

unchanged in this case by intrusion, but the column may 

have still been accelerating toward the dummy afterconlact. 

Toyota Cclica 25 mph Full fiontal 

rhd a|.iad llil ilr.lrt rl; lflrlE ltr 

o 1$ 1o f 1p + f to f, frfr.orf 

Gonloct ltmr (mr) 

c 

a0 ii 

I |,Eo 

E 

a 

2 0 i 

381

Hynndoi Exccl 35 mph FulI ftontnl Tulota Cclica 35 mph Full ftontal 

E oE,lo 

:= 

! !zo 

r, 

o 

C 

o 

oto 

acHl flril$f llftf miltfl rlllt 4fltd lia 

i / 

t / 

V I 

JO 

20 

? f 

g E 

a * 

9 F 

! 8 

# ! 

EE 

(J 

s rs +o f P f p 0 

10 f f .d+c.+s 

I0 

10 1o f ls + S ro f frdrr0rf 

Gonlocl llmr (mr) 

Flgure 34. Contact veloclty and travel cury€a for 35 mph full frontal tests. 

Gonlocl ttmr (mr) 

Hyundel Excal 25 mph Helf offtcl Tuyote Cclicn 25 mph Half offrct 

E o' 

Ero 

tr 

ii 

g 

t20 

(r 

o 

C 

o(Jro 

0 

$ r$ r0 f P + S r0 f Srd.,r0.1s I0 

rs 10 f S f $'t0 f eplqp.rogo 

Contoct tlmr (mr) 

Flgure 35. Contact velocity f,nd travol curve$ lor 25 mph half offset teste. 

382 

ro 

s3 

! r 

! E 

f;E 

f o 

E € 

F o 

o 

st!ill railoal tn rnartaa r|l -dHrl h 

rtfl Glild lfr frtfl rllr rHail hr cltd rd.rlfif llta [tt- rl] rlodrd hr 

| , / 

Conlocl llmc (ma) 

VI 

60 

50 

+0 

JO 

20 

r0 

0 

60 

50 

+0 

JO 

20 

t0 

5 

a() 

E 

o 

t 

E 

a ' 

o 

F 

C 

o 

t, 

c ol' 

E 

a t 

0

I cE 

= t|o 

t 

u 

E Co 

(J 

Hyandd Excel tE nph Half offttt Toyote Ccllca 35 mph Helf olfsct 

JO 

20 

t0 

ilH *rhd lh fifta dti rH|d h 

oroffPfC$f,drdtrorf 

Gonl'oci lfmr (mr) 

{0 

JO 

20 

t0 

0 

s 3" 

E + 

€-E 

€ ieo 

a Llr 

E € 

F o 

0ro 

Flgure 36. Contact veloclty and travsl currra for 35 mph half offeet tests. 

60 

lCtil ailhfl tna tiadt.. rll a|.hd tr 

0 1o 1o f tp f 1p 10 f f .d+o*f 

Gonlocf ttmr (rnr) 

50 

C 

.o+ 

I 

r, 

C 

1 6 g 

E 

- 

2 0 8 

t0 

0 

383

Table 7. Steering column Intrusion measurements from fllm 

analysls. 

t'1E3. ' 

FF'E 

HO2E 

boS6 

.l,r#-e; .:ssl * 

z -o 

,2 

3-' 

65 

o.a 

70 

li5 

l.a 

,2 

t., 

63 

8;' 

le6 

to' 

6.6 

4A 

1.5 

loro 

ts' 

le" 

-9t'* 

E.2l 

36 -3 

(ln) 

iHF 

(ln) 

'- 

(roh) 

vcon 

(ilDh) 

60 I+' Er' 

167 8ot lr3 

9f 8tt" 

9re 

l.l 

6l 

2.1 

65 

o.6 

66 

Ir rtForrnc 

Tsble 8. Che$t contact and velocity summary' 

T?' 

frt" 

9ta$ 

o.f 

65 

:- t8fillTllSf,l"s, not utid to c@putc totrl 

Table 9. Summary of steerlng column pre' and post'tsst 

measuremsnt$. 

TEAT 

FfiB:T t8 tt lo 

FFIE-T 

FFIE-r i8 t8 l:l 

tsoaE-H 

xo35-H N8 t3 it 

HOEE-T 

totE-T t8 

Etl.EiE r ' "r 

21 

z9 9 t:ts 

As previously pointed out, one should not use this to 

conclude that it is better to contact the wheel early to reduce 

the contact velocity because the effects of the wheel motion 

can still affect the dummy after initial contact. 

In looking at dummy response for the 35 mph Toyota test' 

parameters other than intrusion and contact velocity appear 

to dominate. The chest acceleration in the Toyota was 

unexpectedly high at 105.8 g's. Not only was this much 

higher than the 78.2 g's in the Hyundai test, but it was also 

disproportionately higher than the 49.1 g value seen in the 

lower speed Toyota test. Examination of the post-test 

measurements reveals that the Toyota steering assembly 

$rroked 2.4" at ?5 mph but only stroked 1.9" at 35 mph and 

the Toyota column stroked less than the Hyundai at both 

speeds, see table 9. This difference in stroke between the 

two Toyota tests is percentage-wise higher than that for the 

two Hyundai tests, which measurcd 2.7 inches and 2.2 

inches for the 25 and 35 mph tests, respectively. The posttest 

measurement of the steering column angle also shows a 

significant difference in the Toyota and Hyundai column 

behavior. The column angle for the Toyota did not change at 

all at 25 mph, but changed 14 degrees toward the vertical at 

35 mph. This large angle change creates the additional 

problem ofreducing the horizontal component ofthe stroke, 

making the column less effective in absorbing the 

horizontal component of the occupant's motion. 

Additionally, as the column angle changes the lower rim of 

the wheel comes in contact with the dummy's chest or 

abdomen at an increased velocity and force. Since no 

significant increase in abdominal displacement was noted, 

see table 5, it can be deduced that the relatively stiffToyota 

384 

IEST ilO, FF25H FFI6H FFZsI FF35I f,dzsH HO35H HOZ6T HO35T 

ch$t Contrct 

Tlna (hr) 

chart TFavcl 

Dl rtrnca (l nt 

[!ftEtt'$il . t.onr 

FllI lhtly6lt 

Contrct Vil. (Dph) 

illP"li!l'i't"or,r 

Totrl Contact 

Vrloc{ty (mphl 

65 5A 

lo. I lt 

2e. I E7 -L 

22.4 2.1.2 

2.4 3.5 

e{.8 to.7 

64 6l 

lr,9 12. r 

2s.6 2A.A 

26,6 29 

0.e 1,3 

e5.8 30, r 

77 sA 

t0,6 6, f 

z2,u rE.9 

-0.5*r 5,f 

22.9 21.L 

?4 66 

r t.4 10,8 

e3.0 27.1 

0.2 0.6 

?i.2 27.9 

three-spoke wheel significantly loaded the chest in this test' 

It is also important to note in this test that there is very little, 

0.8 inches, vertical intrusion in the Toyota at 35 mph' The 

large change in the angle of the Toyota column appears to be 

due to a non-axial load applied at the base of the steering 

column by the steering rack, which was forced rearward by 

the engine and structural supports intruding the floorpan' 

Additional film analysis was used to detemline that the total 

dummy motion (including stroke of the dummy's chest) was 

only 9.2 inches at 35 mph compared to ll.8 inches at 25 

mph. It was concluded that the large change in the column 

angle precluded axial stroke and did not allow for energy 

absorption in the non-axial loading direction. Also the front 

end of the column shaft was rigidly fixed or possibly moved 

by an intruding steering rack. This phenomenon explains 

the excessively high chest acceleration. 

Behavior of the Hyundai column was also investigated' 

The stroke of the column in the 35 mph test was 2.? inches 

and the stroke in the 25 mph test was 2.7 inches. This 

indicates that the column did not perform as well at 35 mph' 

The cause of this reduced column stroke appears to be 

inversely related to the angle change of the steering column 

during the test. The post-test measurements of column angle 

change along with the stroke distance of the column is 

summarized in table 9. The finding that stroke distance and 

angle change are inversely related is quite logical since 

dummy inertial loads are predominantly horizontal and any 

angle change of the column away from the horizontal will 

naturally reduce the applied force component along the axis 

of the column. Additionally, increasing the angle of the 

column with respect to the horizontal creates a larger 

component of force perpendicular to the column axis, 

resulting in proportionally higher f'riction forces along the 

column axis which further resist column stroke. 

Another aspect of the intrusion, which had a significant 

effect on head acceleration, was the vertical component of 

intrusion in the Hyundai (table 7).In the 35 mph impact, the 

instrument panel moved approximately 4.5 inches upward 

at 74 milliseconds. At 25 mph, the instrument panel moved 

1.7 inches upward at 82 milliseconds. These motions were 

significant to note because the dummy's nose squarely 

struck the instrument panel at these times. Had the intrusion 

not occurred, the dummy's head may have struck the 

windshield, receiving a lower HIC (due to the lower 

breaking force of the windshield) and probably lacerated the 

face chamois. In the 25 mph impact the HIC measured l45l 

compared to 1424 in the 35 mph impact. The proximity of 

these two measurements, in spite of different crash speeds, 

can be accounted for by the fact that the instrument panel in 

the higher speed test broke up before the dummy's head 

struck it, thus limiting the concussive force to the head. 

This vertical intrusion may have also affected the dummy 

chest response. When the wheel and column are thrust 

upward, the forces between the dummy and the wheel are 

transmitted in a more axial direction along the column. 

Furthermore, if the column is loaded at the bottom of the 

shaft from the intruding firewall, engine, steering rack, etc.,

then these intruding forces may be absorbed from the 

column stroke rather than being tran$mitted directly to the 

driver. It should be noted that this phenomenon is beneflrcial 

for absorbing intrusion forces only if the column is not fully 

stroked. 

Occupant kinematics were next evaluated for the offset 

tests by film analysis and by using the crash pulse to obtain 

the contact velocities. The first comparison is between the 

Hyundai and Toyota in the 25 mph test. Film analysis of the 

Hyundai (see figures 29 and 30) shows that the dummy's 

first torso contact with the wheel is the abdomen to the lower 

rim at approximately 67 milliseconds. Next, the chest 

contacts the hub at 77 milliseconds. Analysis from the 

overhead camera shows that no significant rotation (less 

than I.5 degrees counterclockwise) of the occupant 

compartment and very little lateral motion (about 2.0 inches 

to the right) have occurred by this time. The effect of the 

lateral motion is significant because the dummy hits the 

wheel a similar distance to the left of the wheel's center axis. 

This causes the chest to load the column non-axially, which 

causer; the stroking mechanism to perform poorly and the 

rim to act as a principal energy absorber for the thorax. This 

motion was observed to be similar for all half barrier tests 

conducted in this series as well as for previously conducted 

car-to-car offset tests. 

From the Hyundai crash pulse, the contact velocity was 

calculated to be 22.9 mph (see table 8) between the chest 

and hub. There was no additional intrusion velocity in this 

test and apparently the column had already started to stroke 

since the column velocity was measured to be -0,5 mph 

(away from the occupant). The chest 3 millisecond clipped 

acceleration was 44 g's. This acceleration is lower than the 

25 mph full frontal Hyundai acceleration, because of the 

non-axial loading of the hub. Therefore, the softer wheel 

provides a better stroking force at 25 mph than the column. 

In the Toyota crash test, the contact velocity was 

calculated to be 23 mph at the measured contact time of 74 

milliseconds. A very small intrusion velocity of 0,2 mph 

was recorded from film analysis, making the total contact 

velocity 23.2 mph. The three millisecond clipped chest 

acceleration is 45 g's. This figure is consistent with that 

obtained in the Hyundai test. It also appears to be in 

agreement with the dummy acceleration of 49.1 g's 

measured in the Toyota full ftontal test which, though 

slightly higher, can be explained by a slightly higher contact 

velocity of 25.8 mph. It appears that even though this 

vehicle also translated to the right like the Hyundai, its 

movement was slightly less at 74 milliseconds, measuring 

only 1.3 inches by overhead film analysis. This combined 

with a wider diameter steering wheel hub in the Toyota, see 

photographs of wheels in figure 37, allowed the column 

rather than the wheel rim to effectively restrain the 

dummy's torso. Comparison of the dummy's abdominal 

deflection showed remarkable similarity; L75" for the 

Hyundai and l.l4'for the Toyota, see table 5. 

Next, the 35 mph half offset tests were compared for chest 

response. Dummy kinematic plots for these te$t$ are shown 

Hwndd Sl€.rlng Wh€€l 

Toyob $tr.r|ng Who.l 

Flgure 37. Photographs of tteerlng wheels removed from both 

test v€hlcle8, 

in figures 30 and 32. The Hyundai driver dummy had an 84 g 

chest acceleration clipped value while the Toyota dummy 

had only a 69 g's, 3 millisecond clipped acceleration value. 

Examination of the contact velocity shows that the Toyota 

driver contacted the steering wheel at 27.9 mph (including 

.6 mph intrusion velocity), while the Hyundai driver 

contacted the wheel hub at only 24.1 mph (including 5.2 

mph intrusion velocity). The cause of the large difference in 

contact velocity can only partially be explained by crash 

pulse differences. Examination of contact velocity curves, 

figure 36, shows that contact velocity difference would only 

be 2 mph or less at the same contact time or spacing. Since 

the spacing was decreased by the intrusion in the Hyundai, 

contact occurred l6 milliseconds sooner. This fact accounts 

for the greatest difference in contact velocity, causing the 

Hyundai driver to make contact before the relative speed 

between the car and the dummy was allowed to increase to 

its maximum potential. However, since these relative 

motions are still occurring between the dummy and the car 

after the initial contact, forces and resulting accelerations 

are still being imparted to the dummy. Therefore, it must be 

concluded that the only explanation for the dummy 

response difference is intrusion while the dummy is in 

contact with the wheel. Further evidence to support this 

conclusion is the post-test measurements of steering column 

stroke which show that the Hvundai stroked more than the 

385

Toyota (1.9" versus 0.9" from table 9), apparently due to the 

intrusion rather than steering column performance 

differences. 

Head response comparisons are next discussed by car 

type since HIC appears to be a strong function of object 

struck and car design as well as crash severity. First the 

Hyundai was examined to explain the reason for relatively 

high HICs and the reason the head did not break the 

windshield in most cases. By film analysis it was 

determined that three out of four tests with the Hyundai had 

high HICs (142,t-1855) due to contact with the instrument 

panel lip. One exception was the case of the 25 mph offset 

test in which the HIC was only 910. The difference in this 

test was that no significant vertical intrusion of the 

instrument panel, only 0.1 inches, occurred. This allowed 

the dummy to engage the relatively soft windshield before 

hitting the less forgiving instrument panel. It is also 

interesting to note that, this is the only Hyundai windshield 

which cracked as a result of dummy head contact. In the 

other tests the vertical intrusion ranged from 1.4" to 3.7". 

Next the HICs of the Toyota driver dummies were 

investigated. The value from these tests ranged from 469 to 

1586. These values were consistent with crash severity, 

since the two low speed tests had a 664 and a 469 HIC value 

for the full frontal and the offset frontal, respectively. The 

corresponding values for the high speed tests were 1586 and 

1375, respectively. It was also noted that both dummies in 

the full frontal tests had higher HICs than those of dummies 

in matching offset tests. This is readily explainable since the 

full frontal crash pulses were of shorter duration, see figure 

14, and the dummy kinematics in the offset tests were such 

that the dummy struck the windshield/instrument panel 

away from the longitudinal centerline of the dummy's 

original position. This resulted in the dummy's head 

striking the instrument panel off-center where the distance 

from the instrument panel lip to the windshield is smaller 

allowing the dummy's head to strike the windshield before 

hitting the less forgiving instrument panel. 

The reason for the higher HICs in the higher speed Toyota 

te$ts was not just crash severity, but as previously 

mentioned, was more related to the object struck. At the 

lower energy levels, apparently the windshield was 

sufficient to attenuate the motion of the dummy. At the 

higher energy level of the 35 mph tests this was not the case. 

In this test the dummy's head slid down the relatively steep 

raked windshield and the chin struck the stiffer instrument 

panel. 

Steering Assemblies 

This section compares the steering assemblies for the two 

cars, including their mounting, intrusion protection and 

energy absorption design concepts. Each of the steering 

assemblies were unique in some way, each having some 

characteristics that would produce good occupant responses. 

No tear down was performed on the column$, so 

that specific energy absorbing mechanisms are unknown. 

For the high speed tests, the columns, mounting brackets, 

386 

etc. were removed from the vehicle after the post-test measurements 

were taken and were photographed. 

The first column investigated is the Toyota Celicacolumn 

shown in figure 38. These photographs show the wheel and 

column detached and a Toyota wheel/column from another 

Toyota which is attached to the mounting bracket. The design 

of the bracket appears to be very good for decoupling 

the intrusion for the column. It is designed to span the width 

of the car from A-post to A-post, such that the firewall may 

intrude some distance before contacting the column mount. 

Also, the lower column mount is relatively stiff so that when 

intrusion exceeds this value, which was not determinant due 

to complex geometry and uncertainties, it further resists 

intruding forces. The attachment of the column to the mount 

is fairly conventional with the bottom part of the column 

rigidly attached and a shear capsule mount at the upper end 

of the colurnn. The telescoping energy absorbing device is 

located between the two mounts. The steering rack is 

mounted directly to the firewall. One concern with this 

design is that the steering column mounting bracket has 

relatively low torsional stiffness, allowing large angle 

changes in the column when firewall intrusion forces the 

lower end of the column rearward. Another potential concern 

is that there is about a 5 inch portion of rigid column 

below the bottom mount and a shaft connected to the column 

by a U-joint. If this shaft were $horter or the coupling 

St66rlng Column Henrov6d from T€$t Vohlcl€ 

$t6€dng Golufin Atlach6d to Moudtlng BreckEta 

Flgure 38. Toyota steerlng and mounting hardwale removed 

from teet vchlcle.

were flexible, there would probably be less chance of in- 

truding the lower end of this column. 

The othercolumn looked at is the Hyundai Excel column 

which is quite different in design. Figure 39 shows one view 

of the Hyundai steering column and wheel isolated and 

another view with the mount bracket. Upon close examina- 

tion it can be seen that both upper and lower mounts are 

shear type mounts with the shaft portion of the shear capsule 

attached to the column. It is also readily observed that the 

telescoping energy absorbing mechanism is at the boilom 

end of the column. The practical difference between this 

design and the Toyota's is that the column will absorb some 

energy, albeit a very small amount, proportionally, of the 

engine intruding as well as the dummy's forward energy. 

Thercfore we have two very differcnt design concepts. The 

Toyota resists intruding forces and the Hyundai attenuates 

the intrusion by limiting the forces that can be transmitted to 

the dummy. The other flspect of the Hyundai design is its 

mounting. This system of mounting fixes the mounting 

bracket directly to the upper firewall. The only other addi- 

St€erlng Column F€mov€d from Teft V€hlclc 

$rrarlng Column wtth ltoundftg Brfttrtt 

Flgura 39. Hyundal rteerlng End mountlng htrdwara rtmoved 

from tart yehlcle. 

tion to this mount is an angled support bar going from the 

A-pillar to the mounting bracket. It can only be speculated 

that the function of this bar is to prevent side intrusion since 

it does not appear to serve any function for frontal impact 

protection. 

The potential problem with the Hyundai design is that 

attenuation of intrusion is accomplished only by the one 

method just discussed. By mounting the column bracket 

directly to the firewall, motion of the column occurs even at 

low crash sevedty as $hown by this test series. Another 

concem with this type of design is that when the available 

stroke of the column is used, about 3 inches in the Hyundai, 

there is a direct link between the intruding component and 

the occupant. Another disadvantage in this steering system 

type is that the inertial weight of the moving part or srroking 

pofrion of the steering column could potentially be very 

high. However this was probably nonhe cflse as the weight 

of the Hyundai steering column and wheel was only ll.2 

pounds (4.06 pound wheel and 7.14 pound column) compared 

to the Toyota assembly which weighed 15.75 pounds 

(4.25 pound wheel and I1.5 pound column). 

The designs of the two wheels were also compared. The 

photographs 

in figure 37 show the Hyundai Z-spoke design 

and the Toyota 3-spoke design. Judging from occupant response 

and wheel deformation, the wheel of the Hyundai 

was apparently softer than that of the Toyota. The softer rim 

was advantageous for abdominal protection but was, in 

general, not useful for chest protection. One exception was 

the 35 mph Toyota full frontal test in which a sofr lower rim 

would have lihely reduced chest acceleration. 

Summary 

These tests provide a better understanding of the behavior 

of wheels and columns when subjected to a limited number 

of real-world type crashes in which intrusion and offset 

Ioading play a major role. It was observed that, as it affects 

occupant response, the difference in structural response 

between crash mode$ wa$ minor compared to the 

differences in intrusion and loading direction of the steering 

assembly. It was also determined from this test series that 

interior design, shape, material properties and intrusion of a 

specific car all work together, synergistically, 

to determine 

the occupant protection potential. Furthermore these 

parameters are dependent on crash pulse since it governs the 

contact times and point of impact. Contact time$ are very 

important since the relative velocity between the occupant 

and interior component changes as a function of time and 

the interior components may be moving due to intrusion 

thus changing the velocity or the component struck. 

387

Influence of Corrosion on the Passive Safety of Private Cars 


Wolfgang Sievert, Ernst A. Pullwitt' 

Bundesanstalt Fiir Strassenwesen (BASI) 

(Federal Highway Research Institute) 

Dieter Wobben, Helmut Pfisterer' 

Rhein i sch-We stfiili scher Ti.iv Es sen (RWTUV ) 

Abstract 

The objective of this study was to determine significant 

differences in the degree of initial damage resulting from 

corrosion and to assess the relevant influence of corrosion 

using vehicle-occupant loading criteria and test-vehicle 

damage and loading. 

The BASI and RWTUV Essen conducted a joint $tudy of 

three different vehicle types to determine the pattems of 

impact behaviour resulting from corrosion. An older vehicle 

with corrosion damage and a younger vehicle with as 

little damage as possible were examined and tested in the 

impact test. 

The frontal collision type was selected for the study. The 

vehicles with their different levels of upkeep revealed significant 

differences in both the corrosion level and in the 

results of the impact test. 

A comparison of the results obtained from the impact 

tests revealed that initial corrosion damage had resulted in 

an unacceptable reduction in occupant safety levels in a 

number of aspects. One aspect which is particularly worrying 

is the fact that many weak spots contributing to a reduction 

in occupant protection were discovered for those vehicles 

which had only suffered low levels of corrosion and 

which would therefore have been expected to react a$ new 

vehicles. 

One aspect of even graver consequence than the corrosion-related 

failure of vehicle components was the failure of 

the seat belts in two tests. 


The passive safety of passenger cars which is aimed to 

protect vehicle occupants in the event of an accident is 

demonstrated by a series of tests conducted when obtaining 

approval for a new model. However, new vehicles make up 

only a small proportion of total vehicles on the road. The 

service life of a passenger car is currently over 10 years 

(Zl.l%a of the passenger cars on the road in the Federal 

Republic of Cermany in 1987 were more than l0 years old), 

while the average age for 1987 was 6.17 years. Age and 

performance can be assumed to bring about changes in 

vehicles which lead to their modified behaviour in 

accidents. Failure mechanisms and corrosion damage were 

observed in impact tests conducted at the BASI in another 

context and gave rise to this current project. 

During the course of its studies, the RWTUV determined 

a number of points with severe corrosion damage and 

388 

therefore considered it necessary to conduct an appropriate 

study. A newly developed test unit was used for this 

purpose. 

The two institutes selected the following types of 

vehicles, based on their experience: VW Golf I, type l7; 

Ford Fiesta: Daimler Benz W 123. 

For each of the vehicle types mentioned, one vehicle 

approx. l0 years old showing average corrosion damage for 

a vehicle of this age and a 6-year-old vehicle with as little 

corrosion damage as possible were examined and then 

subjected to impact testing. 

Objective of the Study 

The studies conducted were performed so as to allow 

significant differences resulting from differing degrees of 

initial corrosion damage on the basis of vehicle'occupant 

loading criteria and test-vehicle damage and loads. 

For this purpose, a corrosion test unit was first used to 

measure the corrosion damage of the vehicles employed in 

the study. 

The occupant loading criteria used for the assessment 

comply with ECE draft regulation 237 (Economic 

Commission of Europe) and take into account the headload 

(limit value HIC < 1000), the thoracic acceleration (limit 

value a.** E 60g) and the forces acting in the thighs (limit 

value F*"* < l0 kN). The pelvic acceleration (limit value 

ar*, { 80g) also served as an assessment variable in addition 

to the criteria laid down in this draft regulation. Relevant 

criteria for vehicle loading were drawn from ECE 

regulations Rl2, Rl4, R2l, R33 and R.237. 

Following evaluation of the corrosion results, the results 

of the impact test$ and the differences between vehicles 

with heavy corrosion and those with low corrosion levels, a 

test was to be performed to determine whether suitable 

measures needed to be introduced to improve the long-term 

effect ofpassive vehicle safety. 

Corrosion Study 

Selection of vehicles and test points 

Six passenger cars were selected for the project. Two 

vehicles were selected in each case from the same model 

(smaller modifications were possible) of manufacturers 

VW FORD and DAIMLER BENZ and formed a study pair. 

One vehicle of each pair showed only low levels of 

corrosion, if at all, and will be referred to below as having 

"low 

initial damage" (slightly conoded vehicle). The other 

vehicle, on the other hand, already showed clear signs of 

corrosion damage and will therefore be referred to as 

demonstrating "high initial damage" (heavily corroded 

vehicle). 

At least 50 test points were marked on each of the six 

vehicles. These points corresponded to the vehicle-specific 

weakpoints and to the body structures important for the 

introduction of forces into the vehicle in the event of a

frontal impact. A number of additional areas revealing signs 

of corrosion during the vehicle examinations were also 

included in the test. 

Studies conducted 

In order to obtain objective data on the extent of the 

corrosion, non-destructive testing was performed with a 

new test unit developed by RWTUV. This test instrument is 

a compact manual unit which functions on the inductive test 

principle. It makes use of the fact that the various conditions 

ofthe sheet steel ofthe car, e.g. intact steel sheet, corrosion, 

filler, thick coatings ofsurface-protection paint, have different 

electromagnetic properties. The test signal recorded 

with the probe is evaluated in a microcomputer on the basis 

of this "material effect", is allocated to a status class and is 

presented on a monitor in the form of a colour display. 

The following five colour allocations were used: 

I green; No or only very slight corrosion which is 

insufficient to impair ritability. The thickness of 

the paintwork or underseal protection is less than 

3mm. 

r green/red: Low to extensive corrosion. 

r red: Heavy to very heavy corrosion. Standard 

sheet can be easily destroyed by tools when this 

indication applies. 

r greenL/orange: Steel sheet intact, with an underseal 

coating of 3 to 6 mm. 

I orange: Steel sheet intact, with an underseal coating 

thicker than 6 mm er with corrosion holes 

repaired using filler (e.9. polyester). 

The test area was examined by positioning the probe of the 

test instrument at points in a l0 x l0 mm grid; the colour 

indication was transferred to a corresponding printed form. 

Results on vehicles with no or very slight 

corrosion 

These three newer vehicles revealed no visible signs of 

corrosion during the thorough visual inspection. The nondestructive 

test produced "green" indications almost exclusively. 

One exception was the DB W 123 where "red" 

indications indicated considerable corrosion in the area of 

the belt mounting point on the left-hand inside sill (despite 

no external sign of corrosion). This was also confirmed by 

mechanical means. 

In order to attain the conditions of a non-corroded vehicle 

for the pulpose of the comparative studies, a section was 

welded onto the vehicle before the crash test in order to 

make the mechanical stabilitv similar to that of a new 

vehicle. 

Results with heavily corroded vehicles 

The older vehicles revealed clear signs of corrosion over 

various areas during the visual inspection. These were examined 

and documented in greater detail during the subsequent 

non-destructive test. 

In order to allow a quick overview of the corrosion situation 

for each vehicle, a simplified classification into three 

categories was performed after the data had been recorded; 

r No or only very slight rusting (e.g. rust film). 

r Surface corrosion, also penetrating the sheet at 

individual points. 

. Very heavy corrosion, partially with cracks and 

holes. 

These findings were transferred to sketches of the relevant 

vehicle sections. 

Results on the Golf, 1977 model 

The following visible signs of corrosion were discovered: 

. Rusting in the engine chamber, A-pillar rusted 

through in places. 

r Slight rusting on one door. 

r Driver and front-passenger access areas rusted 

through in places. 

r Right-hand belt securing point rusted through. 

r Very heavy corrosion, including holes, in the passenger 

compartment at the seam between the floor 

panel and the inside sill. 

r Areas of corrosion on both sills. 

. Very heavy corrosion with underbody rusted 

through over a very large area behind the lefthand 

transverse link securing point. 

Results on the Fiesta, 1978 model 

The following visible signs of corrosion were discovered; 

. Engine chamber, left-hand side: heavy surface 

corrosion. 

r Slight rusting of the front-passenger door. 

r Relatively heavy surface corrosion in the area of 

the door sill. 

r Surface corrosion on both seat mounts, 

. Very heavy corrosion in the area of the door sill. 

. Very heavy corrosion on the underfloornext to the 

sill; very heavy corrosion at the jacking point, 

including holes in pans. 

Results on the Daimler Benz 200, 1976 model 

The following visible signs of corrosion were discovered: 

. Slight corrosion in the engine chamber. 

. Heavy corrosion on the driver's door. 

. Severe corrosion on the front-passenger's door, 

r Heavy corrosion in the area of the belt securing 

point. 

r Extensive surface corrosion of the seat mount in 

the passenger compartment. 

r Heavy surface corrosion with holes in the wheel 

house. 

r Slight corrosion of the sill in the floor area. 

389

Impaqt Tests 

The collision type selected for the test was a 0" frontal 

collision with an impact velocity of 50 km/h, figure 1. This 

collision type accounts for over 607o of accidents involving 

passenger cars. Due to its frequency and the high risk of 

injury with frontal impacts, this type of impact has more test 

specifications and criteria than any other and is a vital aspect 

of the type approval procedure (e.g. behaviour of steering 

system). 

Test configurfltion 

The vehicles were each equipped with a Hybrid ll50Vo 

male dummy in the front seats. Dummy positioning and 

recording of the test data were performed primarily in accordance 

with ECE draft regulation R.237. An exception 

was made with the DB W123, where the seat position selected 

was not the rearTnost position, but instead corresponded 

to the dummy's body size. Further test parameters 

were also taken from this draft or from regulation R33. The 

impact forces were measured with a dynamometric measuring 

wall in order to determine the changes in the front 

rigidity of the test vehicles. Figure I shows the individual 

elements of this measuring wall. 

Results of the impact tests 

The test results are $hown below so that, for each vehicle 

type examined, beginning with the VWColf, it is possible to 

draw a comparison between the two vehicles of the vehicle 

pair (vehicle with slight initial damage and vehicle with 

heavy initial damage). The results were also compared with 

criteria from the above-mentioned ECE regulations. The 

load on the vehicles is described first, followed by the load 

on the durnmies. 

Tests with the VW Golf TVpe 17 

The accelerations and forces measured on the test vehicles 

are much lower on the heavily conoded vehicle. The 

high acceleration measured at the transmission tunnel can 

be attributed to the interaction of the sensor with parts of the 

passenger compartment. The deformations measured on the 

vehicle with heavy initial damage are much higherthan with 

the slightly corroded vehicle. Table I summarises the vehicle 

load values. 

The test requirements which are demanded in the various 

regulations and guidelines and which were taken into account 

accordingly in these tests related to: 

390 

t Displacement of the steering.tyJtern.-This test 

must normally be performed without dummies' 

The information provided by the measurements in 

the tests discussed here are therefore not fully 

comparable with the type approval tests per' 

formed on new vehicle models. In tests performed 

with the vw Golf, the steering column displacement 

in the heavily conoded Golf is larger than in 

the slightly corroded vehicle. 

Table 1. Vehlcle load vrluG$ In test$ wlth the VW Golf, Type 17. 

Dttuatat 

rccEllFrTloll at 

truB[LBELon tunnGl 

6r* tSllttBl 

ai" [sllttu] 

riiee [s]/ttEl 

IEl 

aarr tgl ,/ t ltrEl 

gill 

li* lsl ir[#] 

H.avlly 

Cotrodcd 

llT Golf 

dilTllil*".rffil 

Tcst vi rluEa lor 

I sltqhtly 

I corroddd 

(KoR r) I v* corf (RoR 2) 

ll. 9 

36 / 47aa 

L7 / 72fri 

t7 ,/ 37ia 

2L,9 / L3 

ltHIiHl,llH 

zz,$ / 23 

--* 

2L,2 / 12 

ro.3 / 1r4 

36.9 / 11 

8L,9 / t2 

47,7 / 45 

8.8 / 25 

s{5 

657 

r Failure of the DerluElng 86flEor 

39 / 34ua 

20 / 38fra 

{l ,/ 35rs 

16. 7 

72/ 35[6 

36 / 37\B 

33 / 50Da 

ea.a / L{ 

30.s / 48 

--t 

26.2 / L4 

9,4 / L6 

{o. 5 ,/ lx 

eo.6 / 5L 

58.7 / 2A 

s.s / 27 

636 

554 

The dynamic displacement could not be measured in its 

entirety since the marking points required for the measurement 

did not remain in camera on the film recording due 

to the fact that the front deformation was considerably higher 

than expected. The steering column displacement was 

203 mm even for the slightly corroded Golf and was therefore 

considerably above the criterion limit value (127 mm) 

for new vehicles. 

t Change in the size of the passenger compartment.---This 

requirement also has to be determined 

without dummies in a type approval test. 

The test values are nevertheless also interesting 

for tests with dummies, eyen if the presence of a 

dummy may prevent excessive change. The criterion 

limit value was not complied with for the 

right-hand seat of the heavily corroded VW Golf. 

t Door opening behaviour.--:lha two Golfs did not 

differ considerably in this point. The doors ofboth 

vehicles could be opened without the use of tools, 

although only with difficulty and not to their full 

opening angles. The changes in the door opening 

dimensions (measured longitudinally and diagonally) 

lay over the range 5-66 mm. They did not 

reveal any clear relationship with the level of initial 

damage, the highest value being measured on 

the vehicle with low corrosion level. 

t Fuel leakage /osses.-These were not discovered 

in any project test and are therefore not dealt with 

here in any greater detail. 

The occupant loads measured on the dummies indicate a 

relationship between initial damage and dummy loading 

which would tend to demonstrate that higher levels of initial 

damage result in higher levels of deformation, that the$e

esulted in lower vehicle decelerations and, consequently, 

lower loads on the occupants. However, this relationship 

only applied when no steering-wheel or knee impacts occurred. 

Force measurements in the belts revealed e.g. such a 

tendency. The dummy values are reproduced in tables 2 and 

In both tests, the dummy in the driver's seat suffered a 

head impact against the steering wheel and the criterion 

limit value was exceeded, this being particularly true in the 

case of the heavily corroded VW Golf due to the greater 

displacement of the steering column and the change in the 

seat position. 

Thoracic loads corresponded to the various passenger 

compartment decelerations and were found to be higher for 

the slightly corroded vehicle. This fact is also reflected in 

the high€r belt forces. 

Pelvic loading was also higher for the slightly corroded 

vehicle. No significant forces were introduced into the pelvis 

by a knee impact. 

In addition to these loads, which could be measured by 

means ofphysical values, the occupants ofthe heavily corroded 

vehicle were found to be subject to additional risks. 

The positions of the belts and the behaviour of the dummies 

in the two front seats would also suggest '*submarining". 

The faulty behaviourof the seats and the restraining systems 

resulted from the considerable change in the position ofthe 

seats due to the inadequate rigidity of the floor panel. It 

should be mentioned here that the underbody of the two 

vehicles examined was not of identical design, the younger 

Golf being fitted with a standard reinforcement in the seat 

area in the form of a wedge-like section (series 

modification). 

The load situation for the passenger dummies is virtually 

the same as for the driver. Head loading in the Golf tests 

with the more severely corroded vehicle produced low values, 

however, despite the light impact against the dashboard. 

No contact occurred in the comparative test, although 

the positional change of the head resulting from the 

high passenger compartment acceleration occurred so 

quickly that accelerations of more than 40g were measured 

for a period of 55 ms. This resulted in the head-load limit 

value being exceeded slightly by a value of HIC = 1023. A 

knee impact was determined for the passenger dummy in the 

heavily corroded Golf, although the loads remained uncritical 

here, too, see the summary shown in table 3. 

Tests with the Ford Fiesta 

These tests were characterised by a seat belt failing in 

each impact. The test values are shown in table 4. 

The test values show the same tendency as the tests involving 

the Colf-the slightly corroded vehicle was more 

rigid. 

t Displacement of the steering calnnn.-Up to the 

dummy impact, the steering-column displacement 

in the tests differed considerably; the displacement 

of 260 mm for the heavily corroded 

vehicle was well above the limit value ( l?7 mm); 

Table 2. Vsluea ol the dummy load (drlver) ln the VW Golf, Tlpe 

17. 

TeBt Point 

(Prrt of Bddy) 

Heavily 

TrBt ValuGE fof 

I stightty 

Corrod.d I corro*d 

192t127 

1177 

' 

\tr/51 

162 

Linit V8IGE 

of 

Lold 

I I W Golf ((01 l) | vu Gotf (Km ?)l crltffir I 

xErp 

emr/rrfiE tgJ 

Hrc (Htc 36*) 

ITIORAX 

amx/r3dB lgl 

slfi 

SEAT BELT 

tlmx lklll/t tnEl 

F2mx tku/t Iffil 

tlmx tklll/t trEl 

F4mx tkfll/t I|EI 

FEtV I S 

rfitsx/83m [g] 

FEruR 

F trux tkII 

8.2 

5,4 

5.2 

7-2 

65 

6t 

65 

71 

52t51 

0.u 

0.1. 

1000 il000) 

66e3me 

fl0001 

80s3ffi 

r0 k[ /8 kxSs 

10 kx /E kx3c 

r HlC56=(HeEd 

Injury Crltlfiffi) cstcutatiol.r lnteryail. = 16ffi 

r* sl ($everity trdcx) todd Griteriofl, is m lohglr erFtoyed 

' 

in current reguldtiffis 

t.r6,* 

Table 3. Values of lhe dummy load (paeaenger) In the VW Golf. 

Tct Point 

(Part df Body) 

IIEID 

emx/tm tgl 

Hrc (Hrc 36*) 

THMIT 

rEr/rlffi I9I 

slfi 

SEAI BETI 

rro tru/t trll 

hrx rku/r rffiIl 

FImx [kxl/t tftBIl 

ro** rkrt/t Imll 

PErVt $ 

rmr/rlc tCI 

FENN 

Flm (ku 

Frmx IkXI 

Ifit v.lEr in 

HGrvtty I Silghtty 

corfodid 

I Corro*d 

w Gotf (xm r) | Yv colt (xfl a) 

16167 

371 (et8) 

,ot?9 

212 

4,5 

0.8 

e.1 

t.5 

74t44 

6l 

60 

6E 

97 

7.0 

1.6 

94151 

roel (6el) 

51t49 

196 

8,2 I 63 

J-?, t 59 

1-?t& 

6.8 I 7a 

49t47 

1.7 

0.6 

Lirlt Yrlrt 

of 

Lo.d 

crltfflr 

1000 (1000) 

e0gf* 

t1000t 

80t* 

10 kI /8kxrG 

10 kr /8krr# 

HIC lt6 r (llcd Injury Cf lt$im) crlculatlon intcrvEl . I6n6 

r* SI (SGvrrity ln(|6xt lold critrrlon. lr m longrr 64|to)Gd 

In currerit rcf,ulrtiffi 

?44t92 

taiz * 

44t11 

305 

4,0t68 

1.0t& 

2,Tt& 

4,4 I 92 

35/55 

0.9 

l.l 

the comparative vehicle had a displacement of 

I l0 mm and thus remained within the limit value. 

. Door opening behaviour.-The door opening 

behaviour was critical for both vehicles. Tools 

391

Table 4. Test values of the vehicle load lor ths Ford Flests. 

VEHICI.E 

ISAD 

PAR,AI.IETER 

lcclljlRlTlof, at 

tranBfrlBdlon tunnrl 

!r* tqll t ldsl 

ai" tsllttffil 

6ii." tgllttEll 

azx tsllttnsl 

tsl 

ICCII.InIIIOI at slll 

!r*r tql ,/ t tusl 

";;; lsllttb81 

DEtOnXtTrof, rof,clE 

iiiii iHi ii iil:i 

T6Bt Val 

HeavLIy 

CoErodEd 

FIEATTA (I(OR 3) 

uea for 

stlghtly 

corroddd 

FIESTA 

(I(OR 4) 

16.0 / 

iimiiniiiini 

ililTlill r*l 

t3.2 / 49 

L3.A / 77 

37.9 / Ic 

15, ? 

33,4 / 6r 

3O.4 / 7,6 

2L.7 / 83 

L4,7 / 2r 

I 

46.9 / 24 

3O,7 / z2 

r7.9 / LL 

L3.7 / 67 

5O,5 / 16 

LA[.a / 23 

81.6 / 19 

9,7 / LO 

655 

591 

44.5 / 2E 

24,4 / 76 

45,3 / za 

17.7 

36,L / za 

4L.6 / 28 

35.3 / 29 

23,7 / r2 

16.3 / 9 

54.9 / 23 

25,7 / Zr 

z6.e / 23 

f3,3 ,/ 10 

49.2 / L6 

L7a.7 / 22 

ag,t / LA 

Lz,O / 33 

560 

50{ 

had to be used to open both doors of the heavily 

corroded vehicle and the driver's door of the newer 

vehicle. The function of the tail door was not 

impaired in these tests, however, so that the criterion 

stipulated by ECE-R33 is thus satisfied in 

full. 

. Chang,e in the size of the passenger compart' 

rnent.-This was critical for the heavily corroded 

vehicle with regard to two aspects-the distances 

from the front seats to the dashboard and the distance 

from the right-hand seat to the front face of 

the footwell. The considerable reduction in the 

fiee distances between the test points-the distance 

from the seat to the dashboard was approx' 

150 mm-resulted both from the seats being displaced 

and from a deformation in the face wall. 

The load on the dummy in the driver's seat (measured 

with regard to the load criteria) was not critical in the test 

with the heavily corroded Ford Fiesta, see table 5. 

The failure of a seat belt in each impact means that the 

tests can only be evaluated to a limited extent, since the 

various dummy values cannot be compared with each other. 

The assessment of the effect of the corrosion damage therefore 

has to rely more heavily on the comparison of the 

results obtained for the load criteria of the dummy held 

correctly by the belt. 

The faults in the seat belts were as foll ows: Opening of the 

seat helt buckle Jttr the passenger dummv in the test with the 

heavily t:ttrroded vehit:le and tearing away of the lap helt 

from the seat fitting o/ the driver's danrn-y in the test with the 

slightly conttded vehicle. No signs of unusually heavy wear 

to this belt were determined prior to the test' The belt tore 

only after the belt-restraint fieams were tom open. 

392 

Table 5. Test values ol the dummy load (driver) In the Ford 

Flests. 

I+st Polht 

(Port of Body) 

lEts 

mx/a3m lgl 

Htc ililc 36r) 

I!!8A[ 

mx/sJc 

sl *r 

EEAI-!E.tI, 

Flmx IkII/t 

FZmx tkill/t 

F3mx tkll/t 

F4mx IkII/t 

e$yls 

ffix/dlm lgl 

ft!u8 

Itfiox tkill 

Frux IkII 

TffiI 

TIIBT 

TEI 

tBI 

N;tvl ly 

corrodfd 

FIESTA ((M 

9ZlE1 

6E9 (664) 

66t55 

4tl 

4,1 | 71 

o-9 I 76 

7,-9 I 58 

4.t I t6 

76t71 

2.0 

2-1 

valEB for 

I srishtty 

I corrodcd 

l) | FrEsrA fl(oR 4) 

146/1 0l+ 

Tv (572r+ 

4,E 

?,0 

5.1 

/'.E 

48147+ 

397 + 

49+ 

6gl 

5l+ 

51+ 

98191 + 

1-T + 

5.9 + 

Llillt VrtEB 

of 

tod 

crl tcr i r 

r000 (1000) 

609!ffi 

tl000t 

aohm 

10 kx /Eklrlffi 

t0 kx /8kx3G 

r HIC 56. (Hced Injufy Criteriffi) crtculttlon Int.rvrL | 36m 

*r SI (sivirlty lndex) told crit$im, is no lo.igor rytopd 

In currfilt regutEtiffi 

+ tsp bett tear at thc hcight of th! Eert flttlhe 

The driver dummy in the heavily corroded Fiesta is decelerated 

well by the belt, his thorax hitting the steering 

wheel as the latter is displaced into the Passenger compartment. 

This is followed by a slight head impact in which the 

lower cranium facial strikes the upper section of the steering-wheel 

rim. 

With the slightly conoded Fiesta, the kinematics of the 

driver dummy were uniform for the first 50 ms, after which 

the dummy, without additional restraint by the torn belt, was 

thrust onto the steering wheel and suffered an impact involving 

the abdomen, the thorax and, somewhat later, the 

head. Despite the belt tear, the head loads remained small 

and were comparable with those experienced in the heavily 

corroded vehicle, though this was not the case with the 

pelvic loads. 

The passenger dummy values allow even less comparison 

since the dummy in the heavily corroded vehicle cannot be 

evaluated due to the fact that it was nst restrained by the 

belt. With the exception of the very high head load resulting 

from a double impact against the windscreen frame, loading 

tended to be low, however, see table 6, since the dummy was 

still decelerated by the belt for approx. 60 ms. 

Tests with the DB W 123 

As already mentioned before, the belt mounting points in 

the door sill of the slightly corroded vehicle had to be 

reinforced by steel sections since an unusually high level of

Tablc 6. Paaaenger dummy lofd$ In thr Ford Flesta. 

T6t Folnt 

(Piit of Body) 

rH/t_ tgl 

Eln 

SEAT BEIT 

Fl* 

Feaa* 

F!-* 

Fl* 

EEWIE 

r hl(/'t; tcl 

[Eta 

I F16* Ikn I 

I Fffi tkxr 

! 

eo9leol + 

I5E4 (1t57) + 

Tilt Valul| for 

llrtrity ltttghllt 

Corfod.d I ctr?o.|rd 

FlEttA (rn 3) | FrErrA ((fl t) 

6015E + 

4tt + 

.v# 

6E 

55 

57 

6tt 

tlrlt Vrl'Jii 

of 

Lord 

crl tGrir 

lnEAp | | | | 

Lx/tr 

Isl 

||lc (|ilc 16r) 

$S8AI 

tHI/t t*l 

(kxl/t IFI 

(txl/t tBl 

(ku/t tBI 

t.9 

r,t 

:|E 

2.4 

59+ 

60+ 

60+ 

59+ 

41/Il + 

e,E + 

5.6 + 

6il6i2 

9et (Et4) 

t7t',' 

5$ 

1.9 

?,( 

5.3 

5.6 

2,4 

2-6 

1000 (t000) 

6ott0001 

8o"J* 

r0 tI /EtrI; 

l0 kr /Ekrlr 

r fllC 36 r (l|nrd lnjury Critcrlffi) c.lcut.tim intcrvrt . !6rr 

** 9l (Ssrity ldir) tord criteim, tr ho ln|f,r Wloyed 

+ Brtt hEklc oF.ild ilflie th6 t;Et 

corrosion (in relation to the overall condition) was ascertained 

in this area and this vehicle was intended to represent 

a vehicle with little or no corrosion. 

For the same reason, the front belt systems were completely 

renewed in order to avoid similar faults as in the tests 

with the Ford Fiesta (original DB spare pans). 

Vehicle loads also corresponded to the above-mentioned 

expectations in these tests. Thble 7 shows the vehicle load 

values. The deformation behaviour of the vehicles was very 

similar although the deformation with the heavily corroded 

vehicle was some 70 mm greater. 

t Steering systemdisplacement---:lhesteeringsystem 

displacement in the$e tests was very low and 

was only 16 mm for the heavily corroded vehicle. 

As with the comparable values in the tests with the 

other vehicle types, this value was also measured 

with dummy occupants. 

. Door opening behaviour.--The door opening behaviour 

in both tests was virtually faultless, although 

the opening mechanism of the slightly corroded 

vehicle was difficult to operate. 

t Passenger compartment size.-The minimum 

test point dimensions were complied with in both 

tests (uncritical in both tests). 

The behaviour of the belt mounts. refer to what mentioned 

before, wa$ not as it should have been in the case of 

the heavily corroded vehicle since the areas around the 

securing nuts were torn out of the sill. The left-hand belt 

Table 7. vohlcl€ load vElus8 In the tests wlth the DB w 123. 

Y.IfICItt 

INAD 

PARAI{RTER 

tilnrrlfilon tunn l 

l* [Ei ir[#i 

rt3js tftlltlltl 

I 

tcl 

r?x tql ,/ t [!rJ 

lCciLlnrTIOI ft 

!IIl 

:i* [Ei lilH] 

L4,L / 2L 

intffii 

L7,2 

ffi 

/ L4 

L94,2 / 34 

Ln,3 / 15 

77,9 / 2A 

L7.9 / r1 

79.6 / 13 

L54.7 / 5A 

7L.2 / 28 

r3.9 ,/ 68 

Dttonllrror 

{rr* ,1y.. tml 

699 

q.* peffi. tml 

6et 

T€rt Vrlues for 

tlravlly I sttgrrtry 

corrodtd I CoEroaled 

DB n 1e3 (XOR 6) | DB tf 123 (XOR S) 

la. e 

29.9 / 4L 

3O.O / rO 

3L.5 / t? 

tF,f, / 4L 

18, 6 ,/ 61 

47.O / 4L 

16. I 

38.7 / 4l 

42.2 / 50 

17,6 / 50 

27.7 / 26 

34.O / 33 

r8tl.9 / 3{ 

L?.Q / L6 

26.7 / 35 

zz.a / 46 

94,5 / L7 

LA4,L / 5r 

7L.6 / 3L 

r8.e / 55 

622 

551 

mounting point was thereby displaced 8 cm to the front. lf 

the belt forces measured during the impact are taken as a 

basis, the tensile force with which the securing nuts were 

tom out of the left-hand mounting point was 4.5 kN. The test 

force for the individual mounting points as stipulated in 

ECE-Rl4 is 6.8 kN for the sill point. 

Thefront seatr were pressed out oftheir original position 

towards the doors due to the manner in which the belt forces 

were introduced into the seats and the transmission tunnel 

section. This was accompanied by a positional change in the 

seats in longitudinal direction of around 30 mm. This seat 

behaviour probably resulted in the passengers suffering 

head impacts against the dashboard in both cases. A further 

result of this positional change was that the passenger dummy 

was clearly swung out of the shoulder belt. This was not 

so evident with the driver dummies due to their impact with 

the steering wheel. 

The loads on the driver dummies do not allow complete 

comparison with each other due to a measuring defect on the 

dummy in the slightly corroded vehicle. The loads on the 

driver dummies are shown in table 8. Whereas the dummy in 

the slightly corroded vehicle suffers only abrief impact with 

its forehead against the hub of the steering wheel at the end 

of the forward displacement, the driver in the heavily corroded 

vehicle hits the cover of the hub with his full face due 

to the larger forward displacement resulting from the belt 

coming away from the mounting point. The limit value of 

the HIC is clearly exceeded. The remaining dummy loads in 

the heavily corroded vehicle are low. The load on the thorax 

in the newer vehicle is considerably higher than the limit 

value, and the pelvic load is also in the region of the limit 

value. The form of belt mounting and the forward displacement 

of the seat probably laid behind the passenger dum- 

393

Table 8. Vslues mea$ured lor the drlver dummy loads in the DB 

w 123. 

Tcat Point 

(PErt of Body) 

IIEAD 

amex/Eror' tcl 

Hlc (Hlc 16r) 

THORAX 

amEx./E3ffi [gl 

sl ** 

SSAI BEI.T 

Ftmx lkill/t tftEll 

FziEx tkxl/t tffill 

F36sx tkl{l/t tmtl 

f4mx tk||l/t trrBll 

PELVIS 

ailsx/'rfl* tsl 

FEIfJR 

tlm6x tkrl 

Frm'x [k|l 

Tcet Vetuoe for Itlrlt vetr,reg 

HeEvily lstlghtly lof 

corroded lcoffodcd lLosd 

DE u 1a3 (KoR 6)l DB H l?5 (Kon 5)l Critcria 

'161 

/ 154 

1598( 1599) 

57t52 

576 

7.tt& 

1.1 I 69 

t.E | 69 

5,1 I 76 

58t56 

2.7 

2.7 

1. 

8nt73 

116 

9.1 t8 

0,6 | 19 

5,1 / 67 

5.EtE4 

100/80 

Lq 

7,1 

't000 (1000) 

60BI* 

tl000l 

E0S3iG 

10 k[ /8krbffi 

r0 kil /8krLffi 

* HIc 56= (Hc!d lnjury critrriffi) c8tculdtlffi lnterval = 3&m 

** sl (severity lrdEx) toEd criterim, is m lmg€r optoyed 

+ cEbte of the heEd scceterstlon Ednaor fl+tured aftcr E7m 

mies suffering head impact on the dashboard and to the limit 

value being exceeded. (table 9). 

The pelvic load of the dummy in the newer vehicle lay 

close to the limit value. 

Table 9. values mea8ured for the passcngcr dummy load In the 

DB W 123. 

TeBt Polht 

(Pirt of Body) 

HEAI) 

rmx/|'tm tll 

Htc (Htc 116*) 

TttoR^x 

rmx/r5|E tcl 

stfi 

BEAI EELT 

Ftmx fldl/t IEI 

Fhrx tklll/t tffiI 

flmx tklll/t Iml 

F{,n6x [krl/t ttr8l 

PELYIS 

ribx/hr* tgl 

FEI.TJR 

Ftmx tkxl 

ftilx [ku 

394 

I 

I 

I IIII 

Hsvl ly 

corrodrd 

DE U|a (rff 6) 

?s7m 

lo!8 * 

37ttl 

319 

r,a 

1.9 

?.7 

5,6 

Idt VrlH f6r 

tl 

7t 

6E 

73 

41tW 

1.5 

1,1 

sl lsht ty 

corrodrd 

Ds uul (ril 5) 

1i|4.t90 

1&9' 

58t77 

It5 

E.Zl& 

4.9 | 65 

4.6 I 6? 

5.5t& 

56t54 

0.4 

0 

Llrl t vrlH 

of 

Lord 

Crl t.rlr 

1000 (1m0) 

60TJ* 

10001 

E0TJ* 

l0 kx /8lx!E 

l0 kx /Elxts 

r illC !6 . (HeEd 

lnjury Crltrrlo|r) crlcutrtlorr Intrrvot ' 56m 

fi sI (scvefity lrdrx) tord critffim. ir m longer rrytoyrd 

* 

t< 36m 

The risk of occupants striking sharp edges in the tests 

with the DB W I ?3 is limited to the area of the dashboard or 

centre console, 

Relationship between the results of the corrosion test and 

the impact behaviour. 

The following sections compare the results obtained for 

non-destructive corrosion testing with the damage and results 

determined in the impact tests. 

VW Golf 

The 6 points marked inthe engine chamber all indicated 

corrosion damage in the areas of the welded and flanged 

seams. The damage suffered in the impact consisted of 

considerable deformation to the front section of the vehicle, 

with a number of components (fender and door plate) showing 

signs of fracturing. 

In the area of thewindscreen both A-pillars were found to 

have rust blistering and were rusted through at points. The 

impact test caused the A-pillars to $tart to tear at these points 

and resulted in a high level of deformation (A-pillar bent in 

roof section). The high level of corrosion damage in the area 

of the A-pillar at the level of the windscreen frame and the 

unsatisfactory securing of the door plates resulted in the 

upper doar secrian being bent out of the door plane and thus 

served as a source ofdanger to the occupants ofthe vehicle. 

It should nevertheless be remembered that the side windows 

had been wound down to allow better filming and that this 

may have influenced the strength of the door frame. 

The fact that both sills (inside and outside) showed heavy 

signs of corrosion and were rusted through in parts and that 

the floor panel had partially come away meant that the si/l 

was bent in the area of the right-hand door. 

In the area of the right-hand belt mounting point the 

corrosion test revealed high Ievels of surface and throughrusting 

on the inside sill. However, the belt mounting points 

had been strengthened by attaching measuring equipment, 

thus preventing the points being torn out in the test. The 

effect of fitting the test equipment was avoided in subsequent 

tests; it is therefore not possible to evaluate these 

areas in the Golf tests. 

Heavy corrosion was ascertained at several points in the 

areas linking the floor panel to the inside sill. This caused 

the underbody to crumple in the impact test, which in turn 

resulted in changes in the position of the/ront seats. 

The damage determined in the passenger compartment of 

the heavily damaged Golf by means of corrosion test equipment 

caused sections of the floor panel to come away from 

the inside sill in the impact. The sharp, rusty edges of the 

sheeting projecting into the passenqer compartment thrs 

represented a considerable source of danger for the vehicle 

occupants. Sharp edges or dangerous areas of unevenness in 

the area which parts of the body may strike are unacceptable 

according to the criteria of the ECE regulations. Figure I 

shows one example of such sources of danger after impact in 

the heavily corroded Golf. 

Ford Fiesta 

The differences described in the level of corro$ion 

resulted in differences in the deformation depth and differ-

Atsa ol tho ddvsfs 8oel. 

Ar€a of lhs front-passor1gofs 

S€st, 

Flgure 1., Defonlatlon of th€ lloor pan€l 

In tests wlth the heavlly 

corrodod VW Golf 

ent behavio;rr of the individual parts, e.g. in the areas of the 

lower windscreen frame including the A-pillars, the connection 

between the fenders and the side walls of the engine 

chamber with the parts functioning as side members. The 

areas of interest in the underbody area are the transitions 

from the A-pillar to the door sills and the sections of the 

underbody connected to the transverse links. 

The deformation and damage suffered by the underbody, 

the right-hand windscreen frame and the dashboard in the 

passenEer compartment of the heavily corroded Ford Fiesta 

represented sources of injury for the occupants. 

The floor panel in the area of the two sills was found to be 

particularly severely damaged even during the corrosion 

examination stage. 

Daimler Benz W 123 

The corrosion damage ascertained on supporting elements 

in the front section of the vehicle produced only 

slightly greater deformation in the case of the heavily corroded 

DB W 123. The transirion point from rhe A-pillar ro 

the door sill revealed high levels of corrosion damage, including 

rusting-through of the parts, even though the areas 

examined showed no external signs of rusting, as just 

mentioned before. Despite the high levels of corrosion damage 

measured at these points, the slightly corroded and 

heavily corroded vehicles behaved in more or less the same 

way during the impact; the slightly greater deformation 

depth can be seen from the distance between the wheels and 

the wheel house. 

The corrosion measurements revealed particularly interesting 

results for the belt mounting points of the inside sill 

for both the slightly corroded and heavily corroded vehicles 

of this vehicle type. This corrosion resulted in one of the belt 

mounting points being torn out during impact. 

Evaluation of the results 

The vehicles classified as "slightly corroded" or "heavily 

corroded" showed very significant differences in both 

the corrosion examination and impact te$ts. 

The corrosion damage was determined successfully for 

the purposes of the project. For the vehicles categorised as 

"slightly corroded", it was possible to demonstrate 

that the 

absence of corrosion predicted by the measurement did 

actually exist. This was demonstrated by the more "re- 

silient" behaviour in the crash and by the absence of rust on 

the fractured metal parts revealed after the crash. Furthermore, 

the results obtained with the corrosion test unit were 

found to be very good since they were able to reveal the 

important and unusually heavily corroded points in the 

Daimler Benz's inside sill. 

In many cases, the corrosion damage determined on the 

"heavily corroded" vehicles could be attributed directly to 

failure of vehicle components. 

The comparison of re sults obtained from the impact tests 

revealed that, in many points, the level of protection 

afforded to the occupant$ was reduced by an unacceptable 

degree by the corrosion damage. The limit values demanded 

by the various regulations governing the behaviour ofvehicle 

components were clearly exceeded in a number of cases. 

Particularly alarming is the fact that the slightly corroded 

vehicles, which were expected to behave similar to new 

vehicles, were found to exhibit a number of weak points in 

the field of occupant prorecrion (see rable l0). The following 

failure mechanisms were found to be the main causes of 

failure for the heavily corroded vehicles: 

The displacement of the steering rolumn into the passenger 

compartment in the case of the Golf and Fiesta was 

found to be considerably higher rhan the legal limit of 127 

mm. 

The criteria for cftanges in the size of the passen7er compartmeil 

were complied with apart from one exception. 

This exception related to one passenger compartment dimension 

in the te$t with the VW Golf. 

The door opening behaviour was only critical in the test 

with the Ford Fiesta. The corrosion damage aggravates what 

is obviously a design weakness of this vehicle type; the door 

opening behaviour was also less than perfect in the comparative 

test. 

The hehaviour of parts of the passenger comparlment 

with respect to the danger of injury they represent for the 

395

Table 10. Overvi6w ol the malor te8t results. 

vehicle 

typB 

wl GOI,F 

Type 17 

FORD FIESTA 

DB w l?3 

V€hicla Danag€ 

Heavily Corrod;d I Sliqhtly corrod€d 

- crltlcal dlsplaceucnt 

of stddring Eyst€D 

- DeteffilnatLon of paa- 

Eenger conpartfient 

Eiz€ rfter inpact, 

1 valuG failGal thc 

critsrion 

- ilrc exceeded for 

drlver dumy 

- Probabl€ rr8ubEariningn 

of driver dumy 

- Fallure 6f belt buckl€ 

right-hand aeEt 

- Critical di8plac€nsnt 

of at€Gring Ey6teD 

- crltlcal d6or openlng 

behaviout 

- Dat€mimtion of paa- 

EEngcr colpartnant 

sLz€ aftBr irpact, 3 

valu€a faif€d th€ 

criteria 

- Belt nount toEn out of 

driver seat 

- Hrc dxcdrdBd f6r bdth 

driv€r and ffont-seat 

paaEenger 

- critlcal diaplacenent 

of stecrlng Byrtch 

- Hrc exceeded fdr 

drlver and front-aeat 

paF6qng€r 

- Iap bdlt of driv€r 

duMy torn 

- PGlvic load crit€rion 

excecdGd 

- crltlcal door opening 

bchavLour 

Heavy corroEion at 

belt nountlng polnt in 

EIIIt lnproved f,or the 

tcst 

HIC exceeded for 

drlver duMy 

critlcal pelvic load 

for driver duffiy 

occupants was critical in all vehicle types. A direct risk of 

injury through corroded body parts was particularly high in 

the VW Golf test. While such a risk was also present in tests 

involving the Ford Fiesta, this risk was lower. The risks 

determined with the Daimler Benz were less grave and 

related to a number of the control elements and the 

dashboard. 

The dummy load limit values, the criteria which most 

clearly illustrate the risk of injury to vehicle occupants, 

were exceeded, particularly as regards the head load. The 

occupant loads in the DB W 123 were unexpectedly high 

when measured against the positive results obtained for the 

vehicle-specific characteristics. 

The assessrzent of the procedure for determining corro' 

sion and performing tests must take into account the following 

points for an envisaged continuation of the project: 

'Ihe 

determination of corrosion was primarily schematic 

in form and was only in part related to vehicle type and 

corrosion damage. This meant a very high number of test 

points. It would appear advisable for the future to perform 

only a few initial measurements with the corrosion test unit 

at the relevant areas of the vehicle type concemed and to 

perform exhaustive measurements of only the most pronounced 

areas. Such relevant areas are e.g. the belt mounting 

points, the seat mounts, the sills and similar frame 

components. 

'fhe 

impact teJtJ were chiefly conducted with the seat 

belts and seats already fitted in the vehicles at the time of 

purchase. 

Irrespective of their existing corrosion damage, the failure 

of these component$ can result in high dummy loads, as 

396 

was also demonstrated in the tests involving the Ford Fiesta. 

Future tests must therefore bear this in mind and exchange 

the old seat belt system$ for new ones. The used systems 

must be examined and evaluated in a separate test (component 

test) in addition to the impact test involving the vehicle. 

Due to the importance of how the vehicle seats behave in 

frontal impacts, these should also be examined thoroughly 

prior to the test and, ifnecessary, exchanged and examined 

separately. 

As shown by the tests with the VW Golf in which the 

manner used to fit the acceleration sensor may have prevented 

the belt mounting point being tom out of its fitting 

location, it is particularly important to ensure that any 

measuring instruments fitted are taken into account when 

evaluating the results. 

The consequences of an accident of comparable severity 

as the crash tests conducted here would result in injuries to 

occupant$ of "heavily corroded" vehicles ranging from 

very severe to fatal. The tests outlined here allow this conclusion 

to be drawn, although the low number of tests conducted 

to date, the small range of models covered and the 

little experience gained in preparing tests, means that it has 

not yet been possible to determine corrosion-related damage-and 

consequent reductions in safety-which cover 

several vehicle types and which can be precisely quantified. 

The failure of the seat belts in two tests is of equal if not 

greater importance than the conosion-related failure of vehicle 

components. 

The results of the "slightly corroded" vehicles in the 

crash test were also unsatisfactory in a number of points' 

Although the german regulation for car approval (STVZO) 

does not stipulate any dummy load values, global tests using 

dummies are essential for assessing the complex interaction 

of the various vehicle components in a frontal impact. 

Summary and Conclusions 

The examination of three different vehicle types (VW 

Golf. Ford Fiesta and DB W 123) to determine their 

corrosion-related behaviour in impacts was performed 

jointly between the BASI and the RWTUV Essen. The 

RWTUV Essen was responsible for selecting the vehicles 

and determining the degree of initial damage resulting from 

corrosion. A corrosion test instrument newly developed by 

the RWTUV Essen was used for determining the level of 

corrosion. The BASt conducted the impact te$t$ and test 

evaluation. 

Significant differences were ascertained between the 

"slightly 

corroded" and 

"heavily 

conoded" vehicles. The 

differences related to test values for and damage to the test 

vehicles, and to test values for the dummies. 

The most grave events for the occupant$ resulting from 

corrosion-related initial damage were as follows: 

r The belt mount was tom out. 

r The behaviour of the $teering system. 

r The behaviour ofthe vehicle seats in conjunction 

with the vehicle's underbodv or seat mount$.

The behaviour of body parts or parts of the 

controls with regard to the risk of injury they 

represent for the occupants. 

Even the tests with the '-slightly corroded" vehicles 

yielded results which were also unsati$factory when 

measured against the regulations used for assessment 

purposes. 

The following conclusions can be drawn about the faulty 

behaviour ofolder vehicles and, in particular, those vehicles 

which are not yet so old: 

The resistance of vehicles to corrosion-related and userelated 

damage must be enhanced, particularly in the light 

of the problems highlighted here relating to the insufficient 

stability of the floor assembly, the significant displacement 

of the steering system into the passenger compartment, and 

The Breed All-Mechanical Driver Air Bag Evaluation 

Wrinen Only Paper 

Jerome M. Kossar, 


Abstract 

Results of a program evaluating a potentially lower cost 

driver air bag system are presented. The nonelectrical, air 

bag retrofit system integrates the complete air bag, crash 

sensing, and inflation functions within a single module 

mounted over the steering wheel hub. The Nationat 

Highway Traffic Safety Administration contracted Breed 

Corporation to develop police car retrofit driver protection 

systems. Predictive analytics compared favorably with 

results measured in tests. Test dummy injury measures 

indicate good protection for the range of driver sizes, both 

with and without supplementary belt restraints. 

Environmental testing indicated that long term elevated 

temperature exposure be limited to temperatures no higher 

than 90oC (194"F) in order to insure high operational 

reliability. 

This paper presents the views of the author, and not 

necessarily those of the National Highway Traffic 

Administration (NHTSA). The numbers in parentheses are 

references, listed at the end ofthe paper. 


The all mechanical Breed Corporation retrofit driver air 

bag system was the subject of a National Highway Traffic 

Safety Administration program. The program concerned a 

nonelectrical air bag system with its crash sensor and 

initiating subsystem as an integral part of the inflator, which 

in turn is housed within the air bag module mounted over the 

hub of a steering wheel. The potential economies 

anticipated from the elimination of multiple remote crash 

sensors, their wiring requirements, and their electronic 

system on*board diagnostic and defect warning apparatus 

were the incentives for NHTSA interest. It is possible that 

the strength of the belt mounts and belt systems themselves. 

The inadequate safety of those vehicles with either no or 

only slight corrosion was demonstrated in a global test. 

The corrosion-related failure mechanisms of the vehicles 

in the frontal impact tests examined in this paper suggest 

that the risk to the occupants of a vehicle involved in a 

lateral collision (where the stability of the vehicle structure 

exerts a direct influence on the occupants) increases at a 

much higher rflte as a result of corrosion than is the case in a 

frontal impact. 

The high number of, in part, dangerous failure events 

resulting from a low number of impact tests and the 

possibilities for reducing such failures in future with a low 

level of effort demands a fuller examination of the problems 

relating to older vehicles which have suffered priordamage. 

the currently labor intensive production of air bag systems 

of this type could be highly automated, and if this 

automation were accomplished, the installed system costs 

could be approximately one third less than that of the 

multiple, electrically signalling, crash sensor systems 

prevalent today. Such automation of production would, 

however, be costly to develop and is not a product of this 

low production volume program. 

The simplicity of mechanical design and the presence of 

two independent sensing and initiating mechanisms within 

the Breed sensor/initiator added to the anticipated system 

reliability. A schematic of one half of the Breed sensor/ 

initiator showing the intemal mechanism of one of the two 

identical and independent means of crash sensing and 

inflation intitiation is seen in figure I. At program inception 

there were reservations concerning the technical feasibility 

of a timely distinction between desirable and undesirable 

deployment situations based on decelerations of the 

steering wheel. The first efforts of the program were, 

therefore, directed at establishing such feasibility. 

After establishing technical feasibility, the next efforts 

involved specific car selection and further development and 

evaluation of the air bag system. The last program efforts 

involved work towards the qualification of the inflator/ 

sensor/initiator assembly. Environmental conditioning and 

testing, which are representative of industry qualification 

practice, have indicated need for design improvement to 

ensure operational reliability after long time exposures to 

the highest temperature employed in some manufacturer's 

system qualification. 

Establishing Feasibility 

The unique air bag mechanical sensor/initiator to be 

employed and its isolated location within the pyrotechnic 

gas generator (inflator) mounted to the steering wheel have 

produced concems as to the ability of such a system to 

397

f 

ILfriil"ll:"ttr 

thi iinrln! r$ n b.ll lt h.ld ltr tlt rotltlil rlffi !y tlf, sFlil 

A dGGlFrtlon ot Fr thm 1.7 6ri 13 ru$lrud to oY6ffi tl$ 3prlfi{ lillil lffi bl$ 

Ind cMc th. rffrlit btll td m toffiFd. 

(hci In r cnih xlth cntr thrtr | a.7 6 dac.lailtlon, tho fttl6 of tl* brll 13 d|laa 

n rlffid by $r lIfi tlrosgh th. b.ll dlfrtn clsrrffi In th cyllrrhf tdH. 

lf r ilr dHeleHtcs lo|l{ ffiOh rbovc 4.7 E. tlE fo$d EYHI cl ttf b.ll rtll 

iluBG thc s0rlm lildcd ldcr io tffi tls'F rhrft srlflclatly tE din t|| tlrl]l| pln. 

Flgure 1. Sensor/lnitlator. 

distinguish between desirable deployment and nondeployment 

events. If deployment was required, could initiation be 

produced early enough in the crash event to provide timely 

air bag inflation for driver protection? The criteria initially 

employed in this program to assess the timeliness of inflator 

initiation were: (a) initial air bag loading on the driver must 

occur before the time the driver moves forward more than 

five inches relative to the compartment interior; and (b) 30 

milliseconds must be allocated to the time between inflator 

initiation and accomplishment of air bag inflation to the 

extent required to impose initial restraining loads on the 

driver. 

As a service to this evaluation, steering wheel hub acceleration 

were recorded during head-on aligned crashes between 

a deformable moving barrier and four different 1983 

model yearcars: The Honda Accord; Renault Fuego, Dodge 

Omni; and Chevrolet Celebrity. These hub acceleration 

crash histories were used as inputs by the Breed Corporation 

to their computerized math model which predicts deployment 

initiation time of their mechanical sensor. Each prediction 

is based on the fixed physical design parameters of 

their sensor and the specific crash pulse under consideration. 

The theoretical results of the Breed modeling are 

shown in table l. As this table illustrates, the predicted 

initiations of driver air bags in these compact car test 

crashes appear close to the timing criteria requirements but 

they are generally several milliseconds later than suggested 

by the criteria. This infers that some adjustments of the 

sensor design parameters would be advisable to provide 

improved protection for compact cars. 

In tests conducted at the Ohio Transportation Research 

Center (TRC) in East Liberty, Ohio, rigid barrier frontal 

crash tests of a Ford LTD with an installed Breed steering 

wheel mounted sensor (reference l)* produced no sensor 

+Numbcrs in parentheses designate references at end of papcr 

398 

EAEEO AIF BAC 

cBAElt SEFOn/ntTtAIOi 

HffiE 

Table 1. Predlcted alr bag Inltlstlon time. 

CGIffi ELOTITI PtEDICIED IXITIAIIil TIflIIED IXITIATIfl ' 

wfftclE (l'lFllt (sffis) (EECofOtt 

:Iil'r : :.:i: :.:i: 

rA! c[Ev. ErERrry 61 0.0et 0,0e0 

i ESE il EilrlEGXt tlr^f ocdFtllt/Atft 8 6 ct{Tmr cffi3 tY Tft IIE ffiilr ffE foHftb 

triggering in 5.1 and 7.1 MPH frontal crashes while a 9.1 

MPH crash did cause the sensor to trigger. Preliminary 

rough road driving tests were also performed. From these 

initial barrier crashes and driving tests it was concluded that 

the Breed sensor/initiator provided adequate resistance to 

undesirable deployment. 

The last feasibility evaluation test series of frontal 

crashes employed three Ford LTD's which had been retrofitted 

with developmental air bag systems. Prior to the conduction 

of these crash tests, NHTSA was supplied with a proprietary 

math model of the mechanical sensor by the Breed 

Corporation which allowed the user to predict initiation 

time based on the crash pulse experienced. Two crash tests 

were conducted on each of the three cars, the first at a speed 

below that required for air bag deployment and the second 

under crash conditions for which deployment was desirable 

for driver protection. Based on actual measured acceleration, 

NHTSA, using the Breed sensor math model, was able 

to compare predicted and actual sensor performance. 

The initial test of each of the three Ford LTD's was a low 

speed frontal banier crash. The type and severity of the 

second crash test of each car was varied. The first car was 

subjected to 30.0 mph frontal fixed rigid barrier (FRB) 

crash; the second a 12.2 mph FRB crash, and the third a 29.9 

mph frontal centerline impact against a l2 inch diameter 

rigid pole. The bumper was stiffened by the addition of 38 

pounds of steel prior to the pole test. Table 2 presents the 

injury measurements obtained from unbelted Part 572 50th 

percentile male dummies seated in the driverposition. Table 

3 allows comparison of actual sensor initiations with those 

predicted from recorded crash pulse measurements. As table 

2 illustrates, all the injury measures from dummy 

instrumentation satisfy NHTSA requirements and table 3 

demonstrates timely and predictable initiation of the air bag 

deployment. 

The design of the developmental air bags used in this 

feasibility series was varied. The first Ford employed an 

airbag made from two 28 inch diameter flat disks of coated 

nylon sewn together at their circumferences and incorporating 

two 1.2 inch diameter vents. The second air bag was 

smaller, fabricated from 26.5 inch diameter flat fabric and 

containing four 1.0 inch diameter vents. In a successful 

attempt to insure that the air bag expanded over the lower 

quadrant of the steering wheel before the dummy forward 

motion blocked this deployment path, 11 inch long tethers 

were added which attached internallv to the front and rear

Table 2. Dummy datE lrom Ford LTD feaalblllty teete. 

TES DIHIPTIdI 5{,.0 pfl frflr t 

ilotD Mt*lct 

HIG 2'F 

eril4 xlc FENb 6a - lll E 

flE3r ffi. 0 (tt8t 

HtrT *rGt. lbd( 

t x. tdtt Lm 

lltr 

ll4r 

a!G 

tt6 

t4t7 Lr. r ll || 

tctu,taF 

le.e |til rffl 

rr0r0 $lttEf 

r rEtD lrrtilt tcfErEtlilil H lnr[li iil clldurril 

r G||Err ErTrErF rccEtEt ilil 5t ry tnt m cr|'qt'Ailor 

E 

ut.9 l+r rfrE 

llr 0lr. FotE 

t6? 

t5-l{tE }t.tttE 

AOG 4IGfr 

t7 ans 

lott Lr. r ?l B tsta Lr. d ?tt F 

l0l5 ll. t el E r$e u. a t7 fr 

Table 3. Breed Inltlator math modcl performance. 

il ils tESt 

ffltlil *fi[ IttlltTrfl 

r?9 LE | 4.5 Ft ff 

iHlll llCrD wtEr 

riltsa4.tffi 

Fffi[ IIEID 'IITIIT 

17? LrD D rz.r rff t4.5 E 

fHr[ ilGID lllllll 

,ilLEra.7S ff 

tHf[ llolb slilh 

il 

hli EEL ETFffii 

teft. ruilfl [EtETtil il ltltIlilfl 

fitrEilA 

Elltilfi *tEt p ftrE F ittE 

'.PILH F IITE 

r?r9 

tE | 10.0 pff A E (ltl! SfEtll0 ffiL rIE t a6 S E E 

fHI[ ttElD $mlH t.Pltu ilE I et s 

rstba8.9H 

Iffil ilIET FflI 

!1.1 B 

3fthlF rot t ft:t 5 trrt 

t-Pttu s rttf 

3ftilI$#L flEIlaB SE 

t-Ptln ilttsE 

rlutr6 541 5 iilc m filt 

l.'lLW E fnl 

IIEETIIGH4L IIE9BH SE 

t-FIrd ftE a ll B 

5ll: (*t) DEFr[t EHtnTS *lt#t tfi H*o 

surfaces of the air bag. These tethers sewn on the centered 

3.5 inch radius, arrested the air bag during its rearward 

deployment causing a more rapid radial expansion of the air 

bag into the space between the advancing dummy and the 

lower portion of the steering wheel. The tethered bag was 

provided with a single vent of L75 inch diameter. 

Study of the high framing rate film exposed in each of the 

feasibility series deployment crashes revealed deficiencies 

in performance not evident from the dummy instrumentation 

measurements. During the 30 mph FRB crash it was 

observed that at the time of maximum chest loading the 

plane of the steering wheel was almost perpendicular to the 

dummy chest. The steering wheel rim, thus radially loaded 

by the dummy, clearly constituted a significant and unsafe 

concentrated load input to the dummy rib cage. The steering 

wheel plane rotation was produced as a consequence of the 

steering column having rotated to a near vertical position by 

the time peak chest loading occuned. Thus, the normal air 

bag benefits of well distributed loading on the occupant 

were not achieved during the time of maximum driver restraint 

loading primarily due to the large forward and upward 

rotation of the steering column. 

Post crash inspection revealed buckling of the main 

bracket suppofr of the column resulting from an excessive 

forward pitching moment being reacted from the steering 

column into the firewall. This buckling allowed the forward 

rotation of the column observed in test. The forward pitching 

moment in the steering column is increased beyond that 

of a non-air-bagged column as a consequence of the effective 

moment arm of the steering column being extended by 

the inflated air bag. The column pitching moment is created 

by the driver's non-axial loading ofthe aft end ofthe steering 

column which produces increased bending of the column 

when coupled with the longer moment arm created by 

the interfacing air bag. The observed column rotation produced 

by air bag addition and the consequent threat to the 

driver highlights the importance of assuring adequate column 

stability and column mounting strength when driver air 

bags are introduced. 

The overall results obtained in the feasibility evaluation 

encouraged NHTSA to proceed with development of a retrofit 

driver air bag system that would be suitable for installations 

in police fleets. 

Car Selection 

Since the feasibility evaluation had uncovered steering 

column instability under the loads of the air bag restrained 

driver and no simple retrofit to attain stability was 

recognized, it was necessary to select another police car for 

use in this program. For this purpose a Chevrolet Impala and 

Dodge Diplomat were equipped with experimental Breed 

driver air bag systems and 30 mph frontal crash tests were 

conducted. The air bag for the Impala had a flat diameter of 

28 inches, employing six l3 inch long internal tethering 

straps arrayed on a centered 7 inch diameter circle with two 

1.2 inch diameter vent holes. The only difference in the air 

bag used in the Dodge Diplomat was that the vent holes 

were reduced to l.l inch diameters. Injury measured from 

both tests satisfied criteria as is evident from the values 

shown below. 

TEST CAR 

1983 Chevrolet Inpala 

sHESrc 

FEMTTR r-oAp$ (LBs) 

289 44 1850 1000 

1983 Dotlge D1plonat 46I 49 2120 t5 30 

Under loads ofthe knee bolster the inboard portion ofthe 

lmpala instrument panel collapsed. This explains the low 

right femur load. Post test inspection of the Diplomat 

revealed that the higher left femur load resulted from 

bearing of the deformed knee bolster upon a steering 

column mounting bolt. Either of these problems could be 

avoided through design modification. It was decided to 

proceed with retrofit air bag development for installation in 

the Impala. 

399

Sled test program 

A sled test program to demonstrate the frontal impact 

performance of the Breed all mechanical driver's side 

retrofit air bag system was conducted in the Transportation 

Research Center of Ohio. The Hyge sled facility was used 

with a reinforced Chevrolet Impala (GM "8" Body) sled 

buck. Various conditions of speed, dummy siee, dummy 

type, and air bag/seat belt combinations were tested. For 

comparative purposes, these tests also included drivers with 

no restraints as well as drivers employing only seat belts. 

Both straight frontal and angle fixed rigid barrier impacts 

were simulated with vehicle speeds of 30, 35, and 40 mph. 

Dummies employed were the 5th percentile female, 50th 

percentile male (part 572),95th percentile male, and Hybrid 

III (50th percentile male). 

The base vehicle for the buck was a 1979 Chevrolet 

Impala, four door sedan, equipped with a split back bench 

type front seat. So as to provide an unbiased comparative 

evaluation, the steering column, steering wheel, instrument 

panel, instrument panel suppofring structure, and, when 

required, windshield were replaced for each test. 

It had been determined that the steering system of the 

Impala provided excessive axial play between the steering 

wheel and steering column which could produce delay or 

distortion of the car crash deceleration pulse reaching the 

sensor. A retrofit modification within the aft end of the 

steering column was therefore used to reduce axial 

clearance. It had also been determined that the steering 

column resistance to axial stroking was less than ideal for 

application of the air bag system. To compensate, 

supplementary stroking resistance was provided by the 

application of two u-bolt type clamps to the smaller 

diameter tube of the two tube telescoping steering column. 

The added stroking resistance was produced by frictional 

loads between the smaller diameter column tube and the 

attached u-bolts a$ the u-bolts were forced by the 

telescoping process to slide down the smaller diameter 

column tube during column $troking. 

The airbag was a tethered circular planform bag 

fabricated by sewing two neoprene coated nylon cloth disks 

together at their periphery so as to yield a 28 inch inside 

diameter when flat. Tethering consisted of six l2 inch long 

$trap$ on a 7 inch diameter circle concentrically sewn to the 

air bag front and back inner surfaces. Air bag venting was 

provided by two 1.2 inch diameter holes through the 

forward air bag surface. The decorative and protective air 

bag cover used was standard for use on the Ford 1985 

Tempo/Topaz air bag equipped cars. The inflators were 90 

gram units containing the Breed mechanical sensor/initiator 

supplied by Morton Thiokol Inc. The last thirteen tests 

employed the final design sensor/initiator which was 

automatically armed upon insertion of the air bag module 

into the steering wheel cavity. The earlier air bag tests 

required manual arming through insertion of a screw prior 

to air bag module attachment to the steering wheel. 

Calibration, function and performance were identical in 

both sensor/initiator models used. 

400 

The steering wheels used were the standard for use in the 

Ford 1985 Tempo/Topaz air bag equipped cars. It measures 

approximately I inch deeper than the standard Impala 

wheel, and therefor locates the wheel rim surface plane 

about I inch further aft in the occupant compartment. 

The knee bolsters used employed a steel tubing frame 

around the periphery ofits base which straddled the steering 

column. The inboard and outboard lowest ends of the frame 

were bolted through the horizontal lower instrument panel 

support of the Impala to retain the knee restraint in place. 

Twenty gauge sheet steel spanned the peripheral tube frame 

and served a$ the mounting surface for the energy 

management crushable foam as well as the bearing surface 

for knee load transfer into the Impala lower instrument 

panel. The crushable energy management material was a 3 

inch thick composite consisting of two lrlz inch thick foams 

of different materials bonded together. The top material (at 

the knee contact) wa$ two pound per cubic foot polystyrene 

foam and the bottom material was 6 pound density closed 

cell polyethylene, The upper layer which was slotted on its 

knee contact surface served the dual function of energy 

absorption and capture of the knee for controlled 

penetration. A vinyl cover was $ewn over the foam 

composite. A light weight brace was bolted between the 

inboard point of bolster attachment and the accelerator 

bracketry on the fire wall. This brace provided an inboard 

load path for knee load transfer from the instrument panel so 

as to avoid bending failure of the lower instrument panel 

support as had occurred in the first crash test of the Impala. 

To compensate for the extra I inch depth of the air bag 

steering wheel as compared to the original stock Impala 

wheel, the seat po$itioning in these sled tests were a$ 

follows: 

5th percentile female...first seat rail notch aft of 

forward 

50th percentile male...first seat rail notch aft of 

midpoint 

95th percentile male...farthest rearward seat rail notch 

Most importantly, it should he noted that each air bag 

inflation was initiated only by activation of the Breed 

sensor| initiaror. The Breed all mechanical sensor/initiator 

responded only to the acceleration ofthe steering wheel on 

the sled buck. exactlv as it would in a real car crash. 

Sled Test Program Results 

Some general trends observed from the sled test are: 

l. In each ofthe sled tests, when the retrofit air bag 

system was added to the three point belt (lap plus 

shoulder belts), large reductions of the Head Injury 

Criteria (HIC) and substantial reductions in measured 

loads in the lap belt and shoulder belt occurred while 

satisfying all FMVSS 208 dummy injury measures. 

2. The all mechanical retrofit air bag system alone, 

when not supplemented by belts, satisfied all FMVSS 

dummy injury measurement criteria in the sled tests

when applied to the 5th percentile female 

anthropomorphic te$t device up to at least 30 mph, for 

the 95th percentile male anthropomorphic te$t device 

up to at least 35 mph, and for the 50th percentile male 

dummies up to nearly 40 mph. 

However, it must be emphasized that sled tests may 

provide optimistic evaluations of restraint $y$tem 

performance as compared to real car crashes resulting from 

the failure of sled testing to produce the passenger 

compartment intrusion so frequently found in real crashes. 

Such compartment intrusion can foreshorten the occupants 

forward travel within the compartment and can also cause 

higher impact speeds between the occupant and the interior 

intruding surfaces. Either of these intrusion effects would 

increase loads on the occupant. 

It must also be emphasized that an airbag system by itself 

does not provide significant protection for side impacts, 

rollovers, and many ejection situations. To provide 

protection in these other serious conditions, the air bag 

svstem must be combined with some form of seat belt. Also. 

Tabl6 4. $led tests ol lmpala retroflt EyEtrms Psrt 572 50th percentlle msle dummler. 

SITI^ÎE AIR BI8 

CICST 

DRIER IESTNAIIT I'|88 

cn f,| $cE lHTlAIlil lltc 

t E. CUP 

AIR BtrC $Y$TEI flLY 

AIR BIB SYSTEII ilLY 

AIR NTG $Y$TEI + ITF EELT 

AtR B C 8YS. + l.|P & flfl.fl.DEt BETTS 

AIN BAG SY$. + IIP T $|f,'LDET BELIE 

LIP & SIfiIfDER IELTS fltY 

LTP EELT ilLY 

TOTATLY UNESTRAIH 

AIR I|lE SY5TEI flLY 

AIR BIE SYSTEII + LAP ELTS 

ATR IIG $YS. + ITF & EflJI.OEN BELTS 

TTP T SIHf,DER BELT$ ilLY 

TOTALLY IJTRESTRAIED 

AIN BAB SYSTET (TLY 

AIN BAE EY$TEI + LTP EELT 

LTP & TIflI-D€R 8ELI8 ilLY 

- 

fr 

tro Fll 

5{l Ftl 

]N IHI 

g) rfrl 

:n l?rl 

:10 prl 

l{l rprl 

gl Pfl 

36 Ftl 

5E ptl 

T' FII 

:[t prl 

55 rPtl 

40 f?H 

40 t?H 

40 pr 

zt ts. 

24 E. 

22 rG. 

58 t8.r 

e4 E. 

2a 18. 

?5 B. 

24 F. 

23 13. 

25 [s. 

since the kinematic and biomechanic fidelities of test 

dummies are imperfect, protection levels indicated in 

testing for the range of sizes in the human population should 

not be interpreted as precise. Additionally, the performance 

observed for the various restraint conditions tested in this 

Impala sled test buck may not be representative of other car$ 

with other belt systems or when other air bag sy$tems are 

employed. The purpose of the sled test program was to 

provide basis for evaluation of the potential benefits which 

might be appreciated by police officers according to the 

range of their sizes and their restraint usage practices if their 

vehicles were retrofitted with the Breed air bag. 

Table 4 shows sled test data from all Part 572, Subpart B, 

50th percentile male dummies tested with the final design of 

the Breed all mechanical driver retrofit air bag system. 

From this table it is clear that the crash sensor function 

worked consistently well with the exception of the one 

sensor which used an obsolete and improper component in 

its assembly. Generally, when the proprietary computerized 

math model of the sensor mechanism was used to establish 

the anticipated air bag initiation time based on measured 

t44 

210 

t6{t 

Sllr 

?91 

85t 

7v 

661 

440 

Ir9 

415 

t$8 

1?a7 

Itc 

l7E 

tft 

6lcr 

39G 

39G 

7B 

9G 

IEFI ffi.R 

(Fd.58) 

tgfl 

ttSt 

741 

714r 

#7 

t7l 

w 

2506 

{cl 1214 

ilc 

5t{ 

TIGIIT FTI 

(Fil.ffi) 

LATE ITITIATIfl REf,I.TED FROI USE Of II ffiil.ETE OII?IIEIT IT TIG A$$E5|.Y oF Tl: SEXSm C USltC BllSltC lr rEcll ilsll 

STEEIITG GOIJII TETIOFIT CLT|pTIG MTFIED TO Pf,OYIffi TM Fq.M ADOITIilAL STNtrIXC NESI$TATCE ABIilE MilAl RETNOFIT LEYET 

1E 

tilg 

tzn 

1n 

tlE 

fi9 

ll87 

tot 

tztt 

lTZ 

t7a 

r6tt6 

1695 1799 

788 il0 

{Et $ 1n 

401

steering wheel hub acceleration records, the predicted and 

actual measured deployment initiation times were in 

agreement within 2 to 3 milliseconds. 

Table 4 also provides driver test data allowing 

comparison between the protection offered by the retrofit 

air bag system when used alone, when used in combination 

with the Impala three point belt (lap plus shoulder belt), and 

when used as a supplement to a lap belt. This table 

additionally provides data from tests of the 50th percentile 

male driver restrained only by his lap and shoulder belts, by 

a lap belt only, and without any restraint. Since tests were 

conducted simulating frontal barrier crash speeds of 30 

mph, 35 mph, and 40 mph, the table demonstrates the 

influence of these crash speeds. 

In the tests simulating 30 mph crashes it is seen that the 

excessive chest accelerations experienced with the lap belt 

only (72G) and totally unrestrained driver (90G) are 

reduced to 39G by the three point belt alone but the three 

point belt also produced an undesired increase in HIC value. 

When the retrofit air bag system is introduced, either alone, 

or in combination with belts. the chest acceleration is seen 

to be at the same or a lower level than the three point belt 

produced when used alone, but each of the air bag 

combinations are seen to reduce HIC to one third or less of 

the value produced by the three point belt. 

Looking at the 35 mph barrier test simulations in table 4, 

without use of the air bag, FMVSS 208 head injury criteria 

is exceeded when the dummy is unrestrained or when 

restraint is provided only by the lmpala three point belt. The 

use of the air bag system alone or supplemented by lap or 

three point belts reduces the HIC to low levels with all other 

injury measures satisfying FMVSS 208 injury criteria in 

these 35 mph frontal barrier crash simulations. 

The 40 mph testing $hown in Table 4 demonstrates that 

the retrofit air bag system when used alone is almost a 

Table 5. Sled tasts ol Inpala retroflt systems Hybrld lll SOlh psrccntlle mEle dumml6$. 

DRII'ER RESTRAII{T I,SED 

AIR BAG SY$TEI.I OI{LY 

AIR EAG SYStFr,l OILY 

AIR BAC SYSTEI{ + LAP BELT 

AtR 8A6 SYS. + LAP & SHSJIDER BELT9 

LAP & $I{ULDER BEIT$ qItY 

TOTAILY UIIRESTRAIIED 

AIR 8AG SYSTEII + LAP BELT 

AIR BAS SY$, + I.AP & S|n'|JER BETTS 

402 

strutArED AIR BAIi 

CRASI{ SPEED IIIITIATIOII 

f0 lftl 

30 lpH 

I0 rffi 

30 t?H 

30 r?H 

30 rfH 

55 lfl{ 

35 rf,H 

22 ris. 

2l rs. 

eu rc, 

1{0 DATA 

22 l4s. 

20 xs, 

satisfactory restraint for the 50th percentile male at this 

relatively high speed. When the air bag $ystem wa$ 

supplemented by a lap belt, the dummy injury measures 

satisfied FMVSS 208 designations for the head, chest, and 

femurs. 

Sled tests conducted using the Hybrid III 50th percentile 

male dummy (Pat 572, Subparr E) provide an interesting 

illustration of the influence of a more compliant and 

biofidelic chest on the chest acceleration experienced by the 

dummy. Table 5 shows test results using the Hybrid III 

dummy in the driver position. In general injury measures 

from the Part 572, Subpart B, dummy (the only 50th 

percentile male dummy referenced for use in our standards 

before the advent of the Hybrid III) are seen to be in 

reasonably good agreement with those of the Hybrid III 

except for the chest accelerations which are lower in the 

Hybrid III tests. The most extreme difference occurred in 

the cases of the unrestrained occupant. The Part 572, 

Subpart B, dummy registered a 90G chest acceleration 

while the Hybrid III experienced 48G in the 30 mph tests. It 

should also be noted that all measures shown on table 5 for 

the air bag system, both alone and supplemented by belts, 

satisfy the injury criteria, including chest deflection 

limitations, even at 35 mph. 

Tests using the 95th percentile male dummy are shown in 

table 6. These data demonstrate dummy injury measures 

satisfying FMVSS 208 requirements at 30 mph and 35 mph 

when the air bag system is the only restraint employed for 

the large driver. It also illustrates satisfaction of FMVSS 

208 criteria at 35 mph when the air bag system is 

supplemented by lap or lap plus shoulder belts. For this 

heavier driver in the 35 mph test$ it can be noted that the 

supplement of the three point Impala belts with the air bag 

system reduced chest acceleration to 47G from the level of 

5 I G that was experienced when only the air bag system was 

oolFREsslsl CHEST 

Ittc cflEsr (tlcHEs) 3 lts. GLIP 

104 

263 

122 

183 

793 

586 

?ffi 

4ttc 

t.3 

e.0 

1.3 

1.5 

1.9 

?.8 

1.4 

1.9 

5es 

34G 

25c 

37c 

56c 

48G 

386 

5ZG 

LEFT FENN 

(Pil10$) 

856 

1144 

IIO DATA 

419 

e86 

2530 

644 

922 

RtBllT FEI'uR 

(PffOS) 

1275 

t 195 

IIO DATA 

249 

?63 

14t6 

602 

426

Tabl6 6. Sled teets of lmpala rstroflt systsma 95th percentlle male dummlee. 

DTIYER REETIAITT I'8EO 

ltneG$YgtEtfltt 

AIR BAB EYSTEI ilLY 

AIN ltB gfETH + LTP BELT 

AtR BAG 8YS. + lJlF I 8llll,oEr ELT$ 

LAF & E|lOJltrEN BELTS ilLY T 

LAP & SIfrJLDER BELTS ilLY 

ÎR BAG SY$TEI,I flLY 

ATR BAG SY$TEil + LAP BELT 

AIR BIG SVS. + IJP & SI(UI,DER BELTS 

stn uTED Att 8Ac 

GMSII SPEED IIITIATIfl IIIC 

t{l FH 

55 Fll 

l5 rfrl 

55 Frl 

15 FX 

S Ftl 

25 x$. 

21 rs. 

?4 r$. 

24 n8. 

employed. Use of the three point belt without air bag is 

shown to ptoduce an excessive HIC level of 1590. 

Results of sled tests with the 5th percentile female 

dummy are shown in table 7. These results illustrate 

satisfaction of FMVSS 208 in the 30mph simulated crash of 

the air bag system when used alone or in combination with 

the three point belt or lap belt as supplernentary restraint. It 

is of interest to note that the 5th percentile female dummy 

injury measures, if compared to FMVSS 208 requirements, 

indicate inadequate protection from the stock three point 

belt system when the belts are employed as her only 

restraint. 

Table I shows sled test results obtained when the retrofit 

air bag system is installed on a tilt wheel. The 50th 

percentile male dummy was used to investigate the 

con$equences of a tilt wheel adjusted to its maximum 

forward tilt position. From these results it is seen that 

adequate protection was provided under conditions of 

maximum forward wheel tilt, however, it is of intere$t to 

fiz 

tc8 

$1 

Et7 

1590 

HrEtt 

I t8. cttP 

tEc 

5tG 

T9G 

47c 

4E6r 

49c 

LEFT fEI.I 

(Pf,TDS} 

r SlflLDER EEII Bffi AY TllE TIIG YIC ETEERIIC HlEEl. tfifn lll CDTTTCTED TllE H.nfi AEDfltIt ETEERIIG Ootltfr 8TRfiE 

STARTE AT TIIC OF ICPA.FFER STEERIIG IIIEET NII flTIET 

Tabl6 7. Sled teste of lmp8la ratroflt ryrtema Sth percentlle female dummles. 

sI;J(ATED AIE BAG 

ORI\GR RESTRAIXT I'|SED 

cn sH EPEED nililATlolt HrC 

l0 ffH 

l0 tfH 

30 pH 

e5 HS. 

E xs. 

24 rs. 

LAP ^P SIHf,.DET BETIS OILY IO I"H 1635r 

1476 

t544 

519 

5lr4 

trg|r 

t94 

IIGIIT FET.N 

(Pim$) 

rtr57 

1772 

258 

{5r 

2E 

:t2s 

compare these results with those shown in table 4 under the 

same test speed conditions but with a stock steering wheel 

rim plane angle. The comparison is as follows: 

It is seen that the tilted wheel cases produced higher HIC 

and chest accelerations. It is believed that the air bag 

deployment which is directed on a higher trajectory as a 

consequence of the steering wheel tilt plane causes a 

reduced air bag interaction with the chest while increasing 

head loading. A similar influence is seen in comparing the 

chest accelerations in the ca$e$ of the tilted wheel when the 

air bag system is used as sole restraint and when it is 

supplemented by a lap belt. The lap belt supplement raised 

the chest acceleration when the wheel was tilted forward. 

whereas, in all cases tested with the steering wheel at a 

standard angle the lap belt supplement consistently lowered 

chest acceleration. With the lower deployment angle 

provided by the standard wheel angle, the air bag efficiently 

restrains the upper torso and head so that added lower torso 

and upper Ieg kinetic energy absorprion provided rhrough 

28 

585 

* FONE.IGTD II?ACTED I.PFER RIT OF STEERITG IJIIEEL ATD THEI TIIE HIE 

57c 

LEFT F$TN 

(PqJps) 

J41 

N5 

re3 

&5 

RTGIIT FEIIJR 

(FOJIDS) 

725 

81 

97 

UE

Table 8. Speclal sled teste wlth 50th percentllG Part 572 nele dummles. 

SI I,I,JLATED 

DRII/ER RESTRAI}IT USED CRASH SPEED 

AIR BAG SYSTEII OiILY 50 I,IPH 

AIR B^G SYSTEI,I + LAP BELT 30 IIPH 

DRIVER RESTRAI}IT USED 

AIR BAG SYSTEI'I OILY 

HIC 

CHEST ACCELERATION 

Hrc 160 

CHEST ACCELERATION 3OG 

SIl.IJLATED 

CRASH SPEED 

30 l'lPH 

TILT TIHEEL STEERING COLU}IN TESTS 

SLEO TESTS COTIDUCTEO I'ITH }IA)(IIL{I,''I FORTIARD TILT OF UPPEN RIH 

AIR BAG 

I}IITIATIO{ 

zA tls. 

?? irs. w6 

AilGtED BARRIER SLED TESTSIruLATION 

AIR BAG 

IIIITIATIOI 

26 irs. 

ua[ruuu-Brxlu.r 

restraining forces of the lap belt reduce the energy 

absorption required from the air bag and thereby reduce 

head and chest loading. With the tilted wheel, however, the 

occupant jack-knife phenomena produced by a lap belt 

changes upper torso angle into an already tilted air bag 

inducing an early air bag penetration at the steering wheel 

lower rim. Once the dummy chest is bottomed against the 

lower rim the radial stiffnes$ of the steering wheel rim plane 

induces higher chest loading than exists without the 

supplementary lap belt. 

One test was conducted with the crash pulse of a 

perpendicular frontal barrier crash but with the sled bu"k 

rotated l5o to the sled motion axis conservatively 

simulating a 30o angled barrier impact on the front left 

corner (driver side) ofthe car. The results ofthat test are also 

presented in table L Driver dumrny injury measures 

satisfying FMVSS 208 were recorded in this test of the air 

bag system without belt supplements. As can be seen in the 

table, air bag initiation occurred at 26 milliseconds during 

this te$t where the sled axis was 15" off the axis of the Breed 

sensor/initiator. This initiation time which is a few 

milliseconds later than that which the aligned crash sled 

simulations produced, reflects the directional sensitivity of 

the sensor. 

The influence of the retrofit air bag system on a driver 

restrained by a three point belt is of great interest. Table 9 

shows the results of matched test set$. Each set consists of 

two tests in which the sarnE dummy size was employed in 

tests at the same speed. In the first test of each set only a 

three point belt (lap plus shoulder belts) was used to restrain 

the driver, while in the second, both the three point belt 

404 

NOR.I.IAL RII'I ANGLE 

144 & 210 

35G & 37G 

316 

43G 

296 

54G 

HIC 

516 

HIC 

CHEST 

5 ilS. CLIP 

43c 

54G 

CHEST 

5 lrs. CLIP 

37G 

LEFT FEiUR 

(PflJilDS) 

1425 

LEFT FEI.I,JR 

(PflJNDS) 

1285 

RIGHT FEiIUR 

(POUNDS) 

15n 

398 

RIGHT FEI{UR 

(FCIJNDS) 

1 132 

system and retrofit air bag system were simultaneously 

employed. The two most important functions of the air bag 

system as supplement to the three point belts are seen to be 

the consistent large reductions in the threat ofhead injuries, 

as evidenced by the head injury criteria (HIC) values 

obtained, and the reduction of loads in the lap and shoulder 

belts. The large reductions of HIC provided by the air bag 

system supplement to the three point belts is clearly an 

advantage. The torso belt load reductions produced with the 

air bag system are also of particular interest since some of 

the early research with belt sy$tems indicated strong 

correlation between numbers of rib fractures produced and 

measured torso belt load (reference 2). As is shown in this 

table the reduced torso belt loading provided by the air bag 

system supplement is substantial, averaging over 407o in 

these tests. 

Impala Car Crash Tests of Final 

System 

A series of five Impala crashes were conducted at the 

Transportation Research Center of Ohio utilizing the final 

design configuration of the Breed retrofit driver allmechanical 

air bag system. The tests consisted of two 

frontal fixed rigid barrier crash tests at 30 mph, a centered 

frontal pole crash at 30 mph, a 30 mph right front corner 

impact into a 30" fixed rigid barrier, and a 35 mph frontal 

fixed rigid barrier crash. Except for the car used in the first 

test, during test preparation the vehicles were subjected 

only to the Impala retrofit operations developed for 

installation of the Breed all mechanical retrofit driver air 

bag system. The test dummy in each of these crash tests was 

an unbelted 50th percentile male Part 572, Subpart B, 

placed in the driver position. 

The test car for the first demonstration crash of the final 

design air bag system was a l98l model year Chevrolet 

Impala. After purchase of this used car, the tilt wheel

Table 9. SIed tests of lmPsla retrollt tystomr alr bag system Influonco on performance ol thraa polnt belt. 

DRIVER RESTRAI}IT USEO 

SIIIULATED 

CRASH SPEED 

LAP & SHCIJLDER 

BELTS OILY 30 IiIPH 

AIR BAG SYS. + LAP & SHf,JLDER BELTS 30 IIPH 


BELTS O{LY 35 IIPH 

AIR BAG sYS. + LAP & SHqTLDER 

BELTS j5 ;.;pH 

LAP & SHflJLDER 

BELTS OILY 

AIR BAG sYs. + LAP & SHflTLDER 

BELTS 

LAP & SHflJLDER 

BELTS OILY 

AIR 8AG sYs. + LAP & SHf,TLDER BELTS 

50 HPH 

50 tlPH 

35 tiPH 

35 l,lPH 


BELTS O{LY 30 I,IPH 

AIR BAG SYS. + LAP & SHOJLDER BELTS 30 I,IPH 

steering column, which was its original equipment, was 

removed and replaced with a non-tilt wheel steering 

column. Normal air bag retrofit operations were then 

followed during the subsequent installation of the air bag 

system. The frontal barrier test was conducted at a measured 

impact speed of 30.0 mph. Posr rest srudy of the high 

framing rate camera film revealed anomalies in the function 

of the energy absorbing steering column which had been 

installed as a replacement. The column end closest to the 

driver was clearly shown in the test frlm during the crash 

eyent. The film revealed that the steering column dropped 

downward towards rhe dummy's lap during the initial air 

bag deployment sequence, prior to subsrantial loading. This 

observed motion indicates that the steering column broke 

free of the shear capsule much earlier and at lower load than 

had been observed in any earlier or subsequent test. 

Furthermore, the interval of column stroke extended over 

only approximately 25 ms., whereas, a 40 ms. stroke time is 

CHEST LAP BELT SHCIJLDER BELT 

HIC 3 l'ts. CLIP (PCITNDS) (FCuilDS) 

5OTH PERCEIITILE IIATE DI,I,I.IY . PART 572 SIJBPART B 

855 

29,| 

I 158 

415 

39G 

39c 

51c 

54G 

re61 

6?8 

SOTH PERCEIITILE IIALE DUIiI,IY . HYBRID III 

793 

185 

95TH PERCEIITILE IIALE rull|Y 

r590 

817 

56c 

38G 

49G 

47c 

2456 , 

1452 

1431 249t3 

.|129 

t15Z 

lB09 7l,4/ 

978 1416 

2272 

1500 

'TH PERCEIITTLE 

FEMLE DUITIIIY 

1635 57c 5U+ 

585 5tG 476 

2240 

1826 

1920 

961 

the characteristic timc observed in other 30 mph frontal 

tests. It is strongly suspected that the steering column 

replacement performed prior to the air bag system retrofit 

introduced uncharacteristic column stroking response 

during which significantly less driver kinetic energy was 

absorbed within the column stroking mechanism. The 

dummy injury measures were therefore suspected to be non 

representative of the retrofit air bag system and a repeat of 

the test was deemed necessary. 

The results of this first test are shown in table 10. Also 

shown on this and subsequent Hbles ofcrash test results are 

the measured times of inflator initiation and the predicted 

initiation time based on the Breed proprietary mechanical 

sensor math model. To obtain prediction from the Breed 

model the steering wheel hub acceleration time history 

measured during the crash test was used as the input 

excitation to the model. Dr. David Breed had permanently 

encoded the design pflrameter$ of the Breed sensor used in 

405

the NHTSA sponsored work before the math model 

computerized program was supplied to the Govemment 

technical manager. The proprietary parameter values in the 

executable only program include the weight of the sensing 

mass sphere, the spring con$tant of the biil$ spring, the 

initial bias spring force and corresponding bias acceleration 

on the sensing mass, the travel required to trigger the firing 

pin, the diameter and clearance of the sphere within the 

interior bore of the cylinder within which the sphere must 

move the prescribed amount to produce triggering, and the 

various geometrical locations and dimensions of the "D" 

shaft and levers within the sensor/initiator. Based on the 

input acceleration history the model establishes the 

Reynolds number of the air flow passing thru the annular 

type orifice between the sensing sphere and cylinder wall in 

order to establish the presence of laminar or turbulent flow 

and the air velocity for determination of air damping 

retarding the sphere's motion. The math model output 

includes the sensor/initiator firing pin release time which it 

calculates based on the characteristics of the crash pulse 

which is input. Acceleration measurements shown for the 

chest are maximum 3 millisecond duration values (3 ms. 

clip). 

A second frontal rigid barrier crash test was conducted at 

an impact speed of 30.2 mph. The test car was a 1983 

Chevrolet Impala. As in the first and all subsequent car 

crash te$ts, no belt restraints were employed on the 50th 

percentile male dummy which was in the driver position. 

The Breed air bag $ystem performance was the same as had 

been demonstrated in the sled tests. In this crash test the data 

from the fore-aft acceleration of the head center of gravity 

was not recovered beyond 9l millisecond$ after crash onset. 

This prevented an accurate determination of the head injury 

l98t ll?ArA 

lqtfE il.fALA 

l98B il?ALA 

I9f,E IIPALA 

* 

FRilIAL FIXED RIGID SANRIER 50.? I?II 

CEIITERTIIIE FOTE E9.E I?R 

1I{I DECREE FRilTAI BARRIER 30.? }fII 

FROTTAT fIXED RIEIb FAIRIER !5.0 I"H 

criteria (HIC), but an estimate based on similarity of head 

acceleration in sled and crash measurements prior to 

channel loss in the crash test was made and is shown on table 

I l. Table I I also contains the test data from the center line 

pole crash, the 30" angled fixed rigid barrier with the initial 

barrier contact on the front right comer of the bumper, and 

the 35 mph frontal fixed rigid barrier crash. The data shown 

illustrates that head, chest and femurs are provided good 

protection in the 30 mph crashes into the front and angular 

rigid barriers as well as into the rigid pole impacting on the 

car centerline. The 35 mph frontal barrier crash dummy 

injury measures indicate satisfactory head and femur 

protection with a marginal protection of the chest. The car 

preparation for the pole impact did not include and 

additional bumper stiffening such had been added to the 

Ford LTD in the earlier evaluation of mechanical sensor 

feasibility. 

It should be noted that the inflator initiation times 

measured in these varied crashes demonstrate the Breed 

sensor characteristic of quicker response foi the more 

violent crash pulses experienced in the 30 and 35 mph 

frontal rigid barrier crashes with later responses in the 

slower onset pulses of the pole crash and the angular barrier. 

Inflator initiation times of 62 ms. in the pole crash and 32 

ms. in the 30o angled barrier crash are significantly later 

than occurred in the frontal fixed rigid barrier tests. Despite 

the later air bag deployments in these slower onset crash 

pulse tests, it is clear from the dummy measures that a driver 

would be provided good protection. This is a natural 

consequence of the sensor/initiator responding only to the 

acceleration present in the occupant compartment which, 

not by chance, besides producing the triggering of the 

inflator also produces the unrestrained driver's motion 

Tsble 10. Flrst demonatrrtlon crash t6$t ol Breed r€trofit drlver alr bsg Eystem with suspect ateerlng column tunctlon' 

CAR TESTED CNA$[ COIIDITIOI{ CSASI{ SPEED |TC CHE$T L. FEN'R R. FE]IJI II{FI^TM ITITIATIOTI IMEI PREDICTIO{ 

19Bl II?ALA FROI{TAI FIED RrGIp BARRTER :m.0 HPll xl|* 59 G 13:t0 tB, 1m0 tB. 31 ts. :t0 lr$, 

r lllc cstculEtim cor.rtd rut b6 [l8d€ dJc to 1068 of vcrtlcal head exis rceclcrutlon det8 in t68t' 

Table 11. Reeults of tlnsl demon$tratlon crs8h t€Ets of Breed retloflt drlvsr air beg $ystem. 

C,AN TE$TS CRABI{ OETDITIilI CRASII SPEED IIIC CIIESI t. fEilN R. FEIi'R ItFtAt* trllTl^ttofl lfDEt PtEDtcTIill 

t3l 

E 

58it 

ilt6 

4eG 

21 c 

1??0 rB. 

1220 LB. 

7:t0 LB. 

1470 LB. 

1450 LB, 

t(60 tB. 

6{t G 1750 LB. 1260 LB, 

20 ils. 

25 rS. 23 lts. 

62 l{S. 6t tG. 

Foru-rft hcrd *crtlretlq'r rccord toft rfter 9l [8, th|,E pre,vEhtlng pr€ci8e ilc dotcminrtlorr. E6sid ori einiterity to 

hrd eccclrrrtlotB In pilor stGd tcstirE, Gxtr+otrtloh of thG drts pr8t 91 |B, prorddct r Hlc of sbot t ioo. 

32 t+$. l{0T cALcutATEo 

20 HS,

within the compartment prior to air bag deployment. 

Therefore, in theory and as demonstrated in these resrsr an 

air bag deployment which is timely for driver protection can 

be produced based on the accelerations present in the 

passenger compartment. The car crash testing conducted in 

this program was insufficient to confirm with great certainty 

that the unique sen$or response characteristics designed 

into this retrofit driver air bag module are indeed 

appropriate to a$sure timely deployment and resulting 

adequate driver protection under a// real world crash 

configurations for which air bag protection may be 

desirable. 

Off Road Car Testing 

A series of rough road tests were performed at the 

Transportation Research Center of Ohio on a l9g3 

Chevrolet Impala to determine under what marginal 

conditions a Breed air bag module would deploy. The repon 

of the te$ts conducted was written by the engineer who 

drove the test car, Mr. Robert C. Esser (reference 3). The test 

driving surfaces are well described in his report and include 

cobblestone road, staggered bumps, single bump, potholes, 

and a small and large earthen ditch. Progressively severe 

conditions were imposed on the car by conducting the 

passages at increasing speeds. Finally, deployment was 

indicated in a 20 mph impact of the front bumper into the 

vertical end of the large ditch which consisted mainly of 

hard packed clay. In addition to the severe bumper mount 

distortion created in the final sequence of large ditch tests 

other damage was also produced. A significant sheet metal 

crease wa$ produced on the left front fender over the wheel 

well and the right front fender was pushed t/z inch rearward. 

Subsequent inspection revealed a cracked radiator fan 

shroud and wrinkling of the left inner fender. 

The test driver who was wearing a wide web four point 

belt restraint offers the following opinion in his written 

report; --The air bag actuation threshold at any higher level 

could result in unacceptable discomfort or injury to the 

driver even with use of the production three point seat belt 

$ystem." This opinion, although admittedly subjective, was 

provided by an experienced engineer in the area of crash 

research and testing who is the only person known to have 

experienced threshold deployment conditions with the 

Breed air bag module installed. Since no acceleration 

measurements were made during this test series and movie 

coverage was minimal, the change of car velocity in the 

deployment event can only be estimated, and that estimate is 

approximately l0 mph with an upper limit value of l5 mph. 

Environmental Qualification Testing 

An environmental exposure and test series which was 

typical of qualification testing for air bag gas generators 

employed by car manufacturers was conducted on Breed/ 

Thiokol inflator modules which housed the all mechanical 

pyrotechnic sensor/initiators in addition to the normal gas 

generator constituents (reference 4). After reme- 

dial design implementations to correct deficiencies which 

were discovered early thru preliminary exploratory environmental 

testing, and a great deal of difficulties controlling 

shaker test equipment outputs to insure that desired test 

exposure extremes were not exceeded, the qualification 

program was completed. The program demonstrated the 

Breed sensor/initiator calibrations to be unaffected by the 

extensive expo$ures to vibration, thermal and mechanical 

shock, humidity, and temperature 

extremes which were 

used in the qualification test series. The gas generating 

characteristics 

of the inflator were also shown to be satisfac- 

tory after these environmental qualification exposures (reference 

5). There was, however, a determination from a 

special testing program conducted by the Indian Head Naval 

Ordnance Station, that the originally specified maximum 

temperature for long term exposure jeopardized sub- 

$equent reliability of inflator function. It was established 

that 105"C (22loF) desensirized rhe stab primers which 

must function as the first element in the pyrotechnic chain of 

inflator ignition. 90"C ( 194"F) was found ro be a safe limit 

for the cumulative 88 hour exposure employed in qualification 

testing which would not degrade system reliability 

(reference 6). 

Conclusions 

The Breed retrofit driver protection air bag system has 

been evaluated in this program. The system has been refined 

and perfected during the course of this work. Some significant 

improvements in design and function were achieved as 

a consequence. The initial concept and design of an all 

mechanical sensor/initiator integrally housed within the 

body of an air bag inflator was shown to be feasible. It was 

demonstrated that timely air bag deployments could be 

achieved without the use of multiple crash sensors located 

in the front end of the car and without expense and complexity 

of reliance on electronics fortimely crash severity recognition 

and air bag deployment intitiation. 

This program has demonstrated that: 

( I ) Good frontal crash protection is provided to the 

driver under all of the crash conditions tested and at 

crash speeds 

in excess to those which the Impala belt 

system can provide adequate protection. 

(2) The air bag sysrem performs well with belr sys- 

tems and provides enhancement 

to the protection offered 

by the belts atone. 

(3) The retrofit system provides good accomodation 

and protection ro the wide range of potential driver 

sizes. 

(4) The sensor function which distinguishes need for 

air bag deployment is immune to the effects of environmental 

exposures, 

(5) When the inflator is tested in the configuration of 

a car installation, the inflator design provides adequate 

protection from moisture to its pyrotechnic contents. 

(6) The primers within the Breed sensor/initiator 

407

suffer a loss in reliability when exposed to 

temperatures above l94oF (90'C) for a period of 88 

hours, 

References 

(l) Stultz, J., (Transportation Research Center of Ohio) 

"Bumper Tests", Report Number 840608' July 1984' 

Vehicle Research and Test Center, East Liberty, Ohio. 

(2) Eppinger, R.H., (National Highway Traffic Safety 

Administration) 

"Prediction of Thoracic Injury Using 

Measurable Experimental Parameters '" Report On The 

Sixth I nternational Technical C onference on Experimental 

Safety Vehicles (October 1976) DOT HS 80? 50l- 

(3) Esser, R.C., (National Highway Traffic Safety 

Administation) 

"Breed Airbag Program Rough Road 

Driving Tests," Report No. RD 38'H, September 1985, 

Vehicle Research and Test Center, East Liberty, Ohio. 

(4) Schneiter, F.E., (Morton Thiokol Inc.) "Test Plan For 

Environmental Simulation And Evaluation Testing of 

Inflator with Breed Sensor," Doc. No. UFR-1208(C)' 

Morton Thiokol Inc., Utah Division' 

(5) Schneiter, F.8., (Morton Thiokol Inc') "Final 

Report-special Test Program-Morton Thiokol Inflator 

with Breed Mechanical Firing Device" NHTSA Program 

No. DTNH22-87-P-01432, Morton Thiokol Inc', Utah 

Division. 

(6) Michienzi, M., (Indian Head Naval Ordnance Station) 

"special Test Report For The Breed Corp. Stab Detonator," 

Report Number STR86006, April 1987, Ordnance Devices 

Department, Naval Ordnance Station, Indian Head, MD. 

Design of a Crash Attenuation System for a Lightweight Commuter Car 


L. Glen Watson, P. Barry Hertz, and 

Dean L. Beaudry, 

University of Saskatchewan 

Abstract 

Nexus I is a single Passenger, three-wheeled, highly fuel 

efficient, lightweight vehicle which was developed with 

funding by the Canadian Department of Transport, at the 

Univeriity of Saskatchewan (1, 2,3)*' Unlike most safetyoriented 

vehicles Nexus was designed with high fuel 

transport efficiency (4) as the primary objective- Nexus 

was, in addition, developed to comply with the intent of the 

Canadian Motor Vehicle Safety Standards (CMVSS) for 

passenger cars. This design orientation has led to a very 

unique vehicle that tries to combine two attributes that are 

widely regarded as incompatible: safety and fuel 

conservation. This paper deals primarily with the design of 

an unusual feature of Nexus, the nose mounted Crash 

Attenuation System (CAS). This crash attenuation system is 

designed to protect the frame of the car and hopefully the 

passenger from any damage in a frontal collision. 

Design Considerations 

In the Canadian automobile crash environment the 

largest number of cars are subject to frontal collisions (5). 

This suggests that there should be a very large payoff in 

protecting against impacts from the front. Since only one 

prototype of Nexus was being built and because the 

designers wished to comply with the CMVSS frontal 

collision standards with regard to passenger protection, it 

was decided to add an energy absorbing device to the nose 

I Nexus is defined to he a connection in a linked Sroup or series. 

*Numbers in parentheses dcsignate reference$ at end of papcr. 

408 

of the vehicle. In the event of a frontal crash this energy 

absorbing device would bring the car and the passenger to a 

controlled stop from at least 48 kmlh, without decelerating 

the frame of the vehicle in excess of 30 "g's". 

This deceleration level was chosen since a finite element 

analysis of the frame of the vehicle had indicated that the 

frame would not yield under a 30 "g" deceleration- The 

aerodynamic profile of the vehicle, structural considerations 

and anthropometric constraints combined to limit the 

energy adsorption device to a length of 0.71 meters. Further, 

if one assumed that the device had the shape of a right 

parallelepiped it would have to fit in a space 0.25m by 

0.35m by 0.7Im. The mass of Nexus and the driver was 

estimated to be 410 kg and the CMVSS required that the car 

should be tested in a 48 km/h barrier collision. A final 

constraint on the design was the proposal by Rauser (6) that 

the ideal form of acceleration pulse is an initial peak 

followed by a square wave as shown in figure l - 

Deflection 

Flgure 1. A stepped square wsve responae.

Possible 

Crash Attenuation 

Mechanisms 

The design problem thus became one of finding some 

way to absorb (48 km/h)z X al0kg/2 or 472X l0e Newton 

meters of energy in rhe CAS 0.25m by 0.35m by 0.7Im or 

0.062 cubic meters in volume. As Nexus was being 

designed to be a lightweight vehicle, the enargy absorbing 

system was also required to be Iight in weight. During the 

design, three types of systems were considered in depth. 

The candidate systems were: 

l. a yielding -'diamond" linkage (7), 

2. a rigid polyurerhane foam "nose" (B), and 

3. an aluminum honeycomb 

..nose" (9). 

Simulations of the linkage indicared rhat it could easily 

absorb the required energy. However, there were suspicions 

thar the Iinkage might fail by buckling wirh liille or no 

energy absorption. In addition, the required attachment$ to 

the frame would have necessitated some changes to the 

"rigid" 

bulkhead at the front of the frame. 

The proposed foam nose was to be built from layers of a 

locally manufactured rigid polyurethane insulation. The 

foam nose CAS appeared to be an attractive choice since it 

would result in a low mass, economical, energy absorbing 

device. Many of the advantages of the foam nose were 

shared by the aluminum honeycomb design; however, the 

design team initially felt that aluminum honeycomb would 

be unduly expensive. Eady in the construction phase ofthe 

vehicle u decision to use the rigid polyurethane foam as a 

CAS was made. 

The first problem encountered in the design of the rigid 

foam CAS was to determine its material properties. The 

foam stiffness was found to be rate sensitive and temperature 

sensitive; further, it was found to have two disparate failure 

modes in compression (8). 

Under normal temperature conditions the marerial failed 

by a plastic buckling of the cell walls. For temperatures 

below the glass transition temperature the material failed 

through a brittle failure of the cell walls. Figure 2 illustrates 

the temperature dependence of the material, It was 

determined that in spite of the temperature and velocity 

dependencies it would srill be possible to build a CAS from 

the rigid polyurethane foam. 

The first barrier impact test$ were done on quarter scale 

models and these indicated the foam system was feasible, It 

was noted however that the force displacement curve did 

not have the ideal shape shown in figure L Instead, the force 

irtcreased very gradually. This was due to the fact that the 

front of the CA$ had a much smallerarea than the back since 

it was built to fit into the aerodynamic shell of Nexus. 

The next step was the full scale testing of the CAS. This 

testing was done at the University of Saskatchewan 

pendulum testing facility ( I | ). The full scale resr was only a 

partial success. The energy absorbed was as large as was 

specified in the initial design, unfonunately the car was 

heavier than was anticipated when the CAS was designed. 

a rro 

J 

; 

rf uo 

vl 

g l,l 

1t 

T 

F 200 

160 

Stndnrufe,i (/rn) 

fleure.2. 

Deelgn curygq relettng straln_ rar€ and temperature 

dependence of one rlgld, cloeedcell polyurethane. 

After the CAS had absorbed the design energy, it effectively 

bottomed out and the acceleration rate exceeded 30 g's. In 

addition, the CAS exhibited a large amount of ..spalling 

off " of material which was not yet loaded and hence had not 

absorbed a significant amount of energy. 

The next stage of the CAS development was to build 

quarter scale models of aluminum honeycomb. To avoid the 

problems associated with the tapering of the CAS, the 

aluminum honeycomb CAS was made in the form of a right 

parallelepiped. The one quarter scale models performed 

flawlessly. The scaled energy was absorbed and the CAS 

was deformed uniformly. Finally, full scale models of the 

CAS were built and tested; unfortunately, the tests were 

only a partial success. The quarter scale models had been 

made from either a single sheet of honeycomb or from two 

sheets bonded together. Due to cost constraints it was 

decided to construct the full scale CAS from four layers of 

honeycomb with layers of fiberglass and polyester resin 

between them. This arrangement proved to be barely stable 

and in some of our impact test$ one of the interior layers of 

honeycomb was extruded laterally almost intact and it 

absorbed almost no energy (figure 3). 

The remainder of the CAS did not have enough material 

to absorb all of the energy and once again the CAS bottomed 

out undcr the load. However, in other experiments the Crash 

Attenuation Systems which remained stable under testing 

Flgure 3. Fallure of an unstsble GAS.

easily absorbed all of the required energy' Figure 4 shows 

the acceleration deflection curve for an aluminum honeycomb 

CAS; as can be seen the CAS performed very well' 

The dip in the midpoint of the curve was probably 

associated with some lateral motion of one of the 

honeycomb layers. It is believed that the instability could be 

removed by using a single thicker layer of honeycomb. 

Drcdil{fmrutrrllmr 

lc tull slt dynamh trtr of r,hrminurn ltFnc:|Gsrlb 

ena rdid p6fu-rttfrhiit-Fim qrrrgf abmrtiry tilnrPff moddr 

g 

Floure 4. Deceleratlon veraua Time tor full scale te$ts of rlgld 

poTyurethane fosm and alumlnum honeycomb modele. 

Conclusions 

In this paper we have introduced the idea of a nosemounted 

crash energy absorber which would limit the 

deceleration to the frame of a vehicle in the event of a barrier 

collision. The work that we have presented demonstrates 

that such a safety feature can be added with very little 

increase in the mass of the vehicle. In addition to mechani 

cal linkage systems, rigid polyurethane foams and 

aluminum honeycombs show good promise of providing 

adequate crash protection. 

References 

(l) Eichendorf, R., Gerwing, D., Hertz, P.8., "Design of 

the Nexus Vehicle", Int. J. of Vehicle Design (to appear). 

Passenger Car Crash Worthiness in Moose/Car Collisions 

Written Only Papers 

P. Ltivsund, G. Nilson, M. Y. Svensson, 

Chalmers University of Technology, Sweden 

J. G. Terins, 

Volvo Car Corporation, Sweden 

Abstract 

This paper describes the development and verification of 

a moose-dummy for testing the crash worthiness of cars in 

moose-car accidents. 

410 

(2) Beaudry, D.L., Hertz, P.B', and Watson, L'G. 

"Construction of the Nexus Vehicle", Int. J. of Vehicle 

Design (to appear). 

(3) McEachern, R.A., Watson, L'G', Hertz' P.B" 

"Predicting the Strength of Welded Aluminum Structures", 

Transactions of the Society of Automotive Engineers 1988 

Paper No. 880902. 

(4) Hertz, P.8., "Technical Aspects of Transport Energy 

Efficiency" Department of Mechanical Engineering' 

University of Saskatchewan, July 1982. 

(5) Transport Canada Report TP948CR7705, 

"Impact 

Configurations and Severities in Canadian Passenger 

Vehicle Collisions". 

(6) Rausern M., "Energy absorbtion of passenger car 

body structures made of steel and aluminum", Int' J' of 

Vehicle Design. Special Issue on Vehicle Safety, 1986 pp. 

l r3-128. 

(7) Gerwing, D.H., Hertz, P.B., "Phase l-Report 

Design of 'Nexus' Vehicle Transport Canada" Contract No' 

osv83-00060. 

(8) McEachern, R.A., Sargent, C.M., and Watson, L.G', 

"Observations on the Material Properties of Low Density 

Rigid Polyurethane Foam" submitted to Materials Science 

and Engineering. 

(9) McEachern, R.A., Watson, L.C., and Hertz, P.8., 

"Frontal 

Crashworthiness Modelling of a Lightweight 

Commuter Car" Presented at the Winter Annual Meeting of 

the American Society of Mechanical Engineering, Boston, 

Mass.. December 13*18, 1987. Published in Trends in 

Vehicle Design Research 1987, ASME DE- Vol. l l pp. 53s9. 

(10) Beaudry, D.L., Watson, L.G. and Hertz, P.8., "The 

Application of Finite Element Modeling in the design of a 

Crash Attenuation system for a Lightweight Commuter 

Car" Proceedings of the 4th International ANSYS 

Conference, May l-5, 1989, Pittsburgh, Pennsylvania. 

(ll) Ken, D., Fischer, T., Werle, S., Hertz, P.B. "Full 

Scale Crash Testing Facility for Lightweight Vehicles" Int. 

J. of Vehicle Design (to aPPear)' 

The dummy is based on and verified against the results of 

a staged collision in which a passenger car impacts a moose 

cadaver. The cadaver test is also described in the report. 

Moose-car accidents contribute to about 2 percent of the 

death casualties in the Nordic road traffic and the type is 

even more represented among slighter injuries. Since 

several attempts to reduce the frequency of moose-car 

accidents have proved to have minor effect improvements 

of the crash worthiness of passenger cars in this kind of 

impact might be desirable.


The Scandinavian moose, sometimes also called elk. is a 

huge member of the deer family. A new-born moose weighs 

about l0 kg and a full grown male might, in rare cases, 

weigh up to 1000 kg but wirh an average of between 300 and 

400 kg. 

In other words this wild animal is almost as heavy as a 

horse and of similar size but has longer and slimmer legs. 

This means that when an ordinary passenger 

car impacts a 

moose, only a negligible part of the animal's mass will be 

struck by the strong, frontal part of the car. Most of the 

animal's mass instead will be impacted by the windshietd 

and roof construction. Car-moose collisions occur in the 

countryside, on roads with speed limits in the range of 7()- 

I l0 km/h. Requirements on unobstructed field of view and 

modern styling trends have made many cars inadequate to 

withstand such an impact. 

The animals are well hidden in the vegetation and do 

often enter the road very swiftly and without warning. 

Cautiousness and driving experience thus offer poor 

protection against this type of accident. Passenger car- 

moose collisions contribute to about ZVo of the death 

casualties in the Nordic road traffic (l), (2).* 

Different attempts to prevent the accidents have been 

undertaken. The most important are: 

r Trees and bushes along the roadsides of the major 

roads have been cut down to give clear sight a few 

meters on each side of the road. 

Road signs have been put up in order to wam the 

drivers at places where the animals are known to 

cross frequently. 

r Wildlife fences have been mounted along such 

parts of the major roads where moose most 

frequently cross. 

The first two methods $eem to have slight effect. The 

third method has provcd to be efficient, but is rather 

expensive. The animals change habits and find new places 

to cros$ the roads. 

It has become obvious that improvement of the crash 

worthiness of passenger cars in this type of accident would 

be a better way to decrease the numbers of killed and injured 

road users. This has been observed by the Swedish 

authorities and in lg82 a working group was established 

with the official research institutions and the car 

manufacturers. 

A moose-dummy was assembled and crashtests were 

conducted (3). Although obtaining valuable dara from rhese 

tests the group decided in 1985 that an improved moosedummy 

was needed for the future work. This paper 

describes the development of the latter dummy. 

Materials and Methods 

This project can be divided into two sections: 

(l) A staged collision with a passenger 

car impacting a 

*Numtrrs in prrenthe$cs dcsignate references at cnd of paper 

moose cadaver in order to get insight into the mechanisms of 

moose-car collisions. 

(2) Design of a moose-dummy and validation by a staged 

collision similar to (l) with rhe cadaver replaced by the 

dummy. 

General test conditions 

The tests were made with an equivalent test set up. The 

mid plane of the car coincided with the center of mass of the 

moose-cadaver/dummy (figure l). The speed of the test cars 

were 79 and 76 ft6/h respectively. Each collision was high 

speed filmed from both sides of the vehicle and from above. 

W 

Flgure 1. TBst conflguratlon. 

Test cars 

The cars were instrumented with accelerometers (Endevco 

2262) and force sensor$ (Load-indicator, AB20). The 

accelerometer$ were placed on the roof at the b-pillars and 

on the floor at the side members. The roof-rails were sawn 

up just behind the b-pillars and force sensors were installed. 

The cars were equipped with high speed film cameras. The 

two vehicles had cameras mounted on top of the boot, covering 

the wind screen area from behind. In the cadaver resr two 

floor-mounted cameras covered the wind-screen and frontal 

half of the roof construction from below. 

In the dummy test the deformation of the roof structure 

and the penetration of the dummy into the passengers companment 

was covered by one floor-mounted camera which 

covered four specially designed gauge rods. Three rods 

were made of PVC-pipe and were fixed to the upper windshield 

frame, one in the middle and the other two in line with 

the front seat occupants positions. All three rods pointing 

straight backwards along the inside of the roof. The fourth 

rod was made of a thin aluminum pipe and was designed to 

measure the penetration of the dummy through the windshield 

opening. [t was mounted horizontally in a rig which 

allowed the rod to glide with friction. It was placed in the 

midplane of the car inside the passengers compartment at a 

height corresponding to the middle of the wind screen opening. 

At the frontal end of the rod was fixed an aluminum 

plate 150 X 200 mm. The fronral end had its initial position 

300 mm behind the upper wind screen frame. 

The vehicles were equipped with an automatic braking 

system which was pneumatically powered and electromechanically 

triggered. The device was connected to the 

original hydraulic braking system of the car. 

Moose cadaver 

The moose cadaver was of a four-year-old male, weighing 

260 kg. The center of mass of the moose body was at the 

7:th rib, 1.35 m above the ground. The cadaver hung in four 

4ll

steel wires that were cut off, with specially designed cutting 

devices (Norabel AB), at first contact between the car and 

the moose. The cutting device was made up by a small 

housing through which the wire was passing. Inside the 

housing, a wedge, driven by explosive agent, cut the wire 

when an electrical detonator was triggered. The trig-to-cut 

time did not exceed 2 ms. 

Moose dummy 

The design and construction of the moose dummy was 

based mainly on the experiences from the cadaver test' In 

that test the moose-body responded roughly like a water 

filled sack, that is no evident rigidity due to e.8. the skeletal 

structure could be observed, to the impact. 

The dummy thus mainly consisted of water. The water 

was divided into impermeable compaftments, twenty high 

pressure hoses (Heliflex, I1299) with an inner perimeter of 

320 mm and an outer of 340 mm. The hoses were filled with 

water to 5.7 dm3/m which gave a cross sectional area corresponding 

to a rhomboid with diagonals with the rutio 2ll. 

The hoses were tightened at the ends by welding and all 

air inside was removed. To avoid ripple inside the hoses 

they also contained five interlayers of a polypropylene fiber 

mat (ENC-TEX AB, Y065) (figure 3). The mat consisted of 

two sheetings separated by thin, transverse, synthetic fibers 

which gave a high flow resistance. 

Flgure 3. Gross-Eoctlon of one hoae placed lnelde lts Poclrct In 

thE mstrir. Lengtht In [mml. 

The hoses were mounted together in a material matrix. 

The matrix consisted of a number of sheets of a knitted 

fabric (ENG-TEX AB, Yl55) sewn together with a 6 mm 

wide flat-lock seam (figure 2). The cross section of the 

dummy thus got a rhomboidal pattern. Each rhomboid 

"pocket" had a perimeter of 390 mm. Each hose was covered 

with a polyethene film to allow easy glide inside its 

pocket, which is essential in a bending motion of the 

dummy. 

Flgure 2. s) Crosg-sectlon and b) alde vlew, of the mooss dummV. 

tengttrb in [mm]. Ths dummy was hung wllh lta lower edge 

1m above the ground. 

Results 

Cadaver test 

With this test setup the forelegs were hit by the car front, 

while the hind legs were passed. When the front hit the 

forelegs the hooves bent up under the bumper thu$ fixing the 

legs at this position during the first 50 ms after first contact' 

This forced the moose body downwards, towards the 

bonnet. Observations from the high speed films indicated 

Flgure 4. Drawingt from the hlgh'apeed tilm of the cgdsvcr test 

sh-owlng the lntru$ion ot ths moose'body lnto th6 pass€ng_€rs' 

compar:tment. The numbers show the tlmB past from flrtt 

oontact betwssn the vehlcle and the cadaver In [m$1.

that not much of the cadaver's mass was rotated in this 

initial sequence even though lhe supedicial parts, the skin in 

particular, were considerably twisted (figure 4). 

At approximately 80 ms after first contact the pelvis was 

hit by the right frontal part of the roof. This occurred at a 

speed of ?6 km/h and induced a force peak of about 20 kN in 

the right force sensor and caused an acceleration peak of 

about 30 g at the b-pillar (figures 5 and 6). 

The force then rapidly decreased at the right side and 

slowly increased at the left side. The force pulse on the left 

side was lower but with a longer duration. The maximum 

value was about l4 kN (figure 6). The impact had a duration 

of about 80 ms during which the car velocity dropped from 

78.9 to 65.9 kmih, in complete accordance with the law of 

impulse. The pulses in the floor accelerorneters never 

exceeded I 0 g and had an average value of about 4 g (figure 

5). 

Irl 

40 

30 

20 

10 

0 

0 lO t0 !20 lco 100 1.0 IErl 0 {0 80 lto 160 200 lto lm.l 

Flgure 5, s-1. Accelerometer signalr from ths cadsv€r t€st 

(a,b,c) and from the dummy tegt (ii,e,l). Flgures a) and d) shsw 

thedght-b-plllar slqnals, !l and c) show thb left-b:illlar ilgnals 

End c) and f) show lloor slgnals. 

The moose body proved to be very deformable and 

flexible. It bent around the a-pillars and the wind-screen 

frame and penetrated deep into the compartment. Maximum 

penetration occurred after 150 ms, at which tirne the moose 

body was as far inside as a few tenths of mm:s in fronr of the 

b-pillars (figure 4). 

The wind-screen dropped into the car at an early stage, 

thereby shielding the two floor-mounted cameras. Thus the 

penetration could only be observed from one side view and 

the boot-mounted camera. 

The residual deformations of the car roof was up to 180 

mm in the x-direction and 130 mm in the negative 

z-direction (figure 7). 

tkill 

10 

t5 

rz 

16 

t2 

3 

{ 

0 

a 

{ 

0 

0 {0 60 12C 16A 20A [mEI 

kNl 

Figure 6, a-f. Force ssnaor slgna|e from the cadaver (a,b,c) and 

the dummy (d,e,f). a) and d) rlght alde algnals, b) and e) left slde 

8lgnal3, c) and f) superposition ol the signsls from both sldes. 

Dummy test 

The dummy showed great similarities with the cadaver in 

the interaction with the roof construction. The penetration 

into the passenger's compartment was about as deep as in 

the cadaver test. The acceleration and force pulses also 

showed good correlation (figures 5 and 6). 

The impact speed was 76 km/h, corresponding to the 

windshield impact at the previous test. The residual deformations 

of the car roof was up to 400 mm in the x-direction 

and 180 mm in the negative z-direction (figure 7). 

Discussion 

Cadaver test 

Statistics show that moose in the first month$ of their 

second year of life are in majority amongst the animals 

involved in car-moose accidents. In this age they usually 

have a body mass in the range of 18f280 kg. We therefore 

decided to use a cadaver in this weight range. The cadaver 

used was a starved four-year-old male weighing 260 kg. A 

normal weight for this animal would have been about 350 

kg. 

The cadaver was dissected immediately after the test. The 

strength of the rib bones was then tested and found 

20 

16 

t2 

s 

0 

t6 

6 

2A 

z4 

20 

r6 

t2 

6 

4 

0 

0 

0o 110 l60 2oo 

t20 t60 eoo 

413

Flgure 7. Residual windshleld lrame deformations: a) top vl€w 

b) lrontal view 

considerably higher than of ribs of a one-year-old animal. 

The evidently flexible and ductile behavior of the body, 

shown in the high-speed films, indicated that the moose 

body was loading the roof construction with a pressure that 

was rather evenly distributed over the contact surfaces. The 

final deformations of the car gave the same indications. It 

was believed that the skeleton plays a minor role in this kind 

of impact. 

The center of mass of the cadaver used in this test is about 

100 mm higher than what is typical for a one-year-old 

animal. It was however concluded that the cadaver was an 

acceptable model for a one-year-old moose. 

Due to rotation of the moose, the spine "rested" against 

the upper wind-screen frame during the crash. It is difficult 

to tell whether the spine in another configuration would 

behave as a load carrying structure or not, which would give 

implications to the dummy development. 

The speed of impact was chosen in the range of 70-l l0 

kmTh since this is the interval where the injurious accidents 

appear. 80 kmlh was considered a reasonably tough speed 

which was expected to give moderate deformations and 

thereby to enable meaningful force measurements. 

Accidental data show that impacts where the center of mass 

of the animal coincides with the mid plane of the car are the 

414 

most serious (2). That is why this configuration was chosen 

in the two tests. 

The Volvo 244 was chosen as test aar since it was the most 

common car in the Scandinavian countries at that time. A 

large reference material was available in Volvo Car 

Corporation's accidental data, with a great number of well 

documented moose-car collisions. 

In the Volvo accident files the severity is noted by the 

Vehicle Deformation Index (VDI) scale (SAE J 224 B, 

Collision Deformation Classification). 

The car in the cadaver test got a VDI of 60, which places 

the crash among the severe in the most frequent group (VDI 

55-60) in Volvo's accidental data for this kind of impact. 

This accords quite well to previous discussions. The most 

severe accidents with moose have a VDI in the region 75- 

80. Accidents with VDI > 80 are very rare. Volvo experts 

found the deformations very characteristic for this type of 

accident (l). 

The moose was equipped with three pairs of 

accelerometers (Kyowa AS200A). That is three devices on 

the hit $ide and three at the reciprocal spots on the sther side. 

The accelerometers were meant to give baseline data for 

estimating the compression of the moose body. However, 

the rotation of the outer parts of the moose took the sensor$ 

out of direction at an early stage and this is why the available 

information is of minor relevance. 

Dummy development 

The cadaver test indicated that the moose could be described 

mainly as a water-filled sack in this kind of impact' 

Thus, a strong, soft and non-elastic container with a viscous 

content of a density close to I000 kg/m3 was expected to be 

an appropriate dummy solution. Roughly, the container 

should have the shape of a moose body. 

As mentioned the crash-configuration with the center of 

gravity of the moose hitting the car's mid plane is considered 

the worst one. The moose itself is rather asymmetric in 

all dimensions around its center of gravity, but to increase 

repeatability and to simplify moose-dummy construction 

the dummy was made symmetric around its transversal 

plane through the centre of gravity. 

In all dimensions the mass distribution was arranged to 

simulate the configuration of the cadaver (from centre of 

gravity to head end in the asymmetric case) in the critical 

part of the collision, 80-160 ms after first contact when the 

head and the hind legs are moving at the sides of the car in a 

horizontal plane. 

To accomplish this we used hoses of different lengths 

(figure 2). 

The hoses were only filled to half their maximum volume 

in order not to hinder the bending. Since the sheets in the 

material matrix are made of a knitted fabric and thus are 

tensile, the matrix must have a pre-tension in the vertical 

direction. Pre-tension decreases the deformation of the matrix 

when the dummy is hung up. The cross sectional area of 

each pocket in the matrix will increase and the length of the 

matrix decrease duc to gravity when the dummy is hung

up. Each pocket was given a perimeter of 320 mm in the 

empty unloaded matrix and had a perimeter of 390 mm 

when the dummy had been put together and hung up. The 

length of each pocket in the empty matrix must be lSOVo of 

the hose length in order to get the same length as rhe hose 

when the dummy is hung up. 

Dummy test 

This test showed good agreement with the cadaver test. 

Both the signals from the accelerometers and the deformation 

of the car showed such good agreement that we considered 

the dummy to be a valuable and inexpensive test tool 

for this kind of impact. The dummy test was a more severe 

impact to the car due to the fact that the mass of the dummy 

was equal to the mass of the cadaver with forelegs. The mass 

of the forelegs of the cadaver was accelerated by the car 

front before the body impacted the wind-screen area and did 

not contribute to the wind-screen impact but did reduce the 

car speed prior to wind-screen impact- The speed at windscreen 

impact was set 3 km/h lower in the dummy test, to 

give the same impact speed at the wind-screen area in both 

tests. The mass to be accelerated by the wind-screen/roof 

construction wa$ however still higher in the dummy case. In 

this respect the 260 kg-dummy was reciprocal to a 310 kgmoose. 

The contact with the forelegs in the cadaver test 

pulled the cadaver slightly downwards" Thus the lower edge 

of the wind-screen frame and the top of the dashboard took a 

greater part of the impact energy in the cadaver test. This 

also adds to the fact that the dummy impact was more severe 

to the roof construction. 

The car in the dummy test got a VDI of 75, which places it 

among the moderate in the most severe group. 

Acceleration measurements were left out in the dummy 

test since the design of the dummy did not allow the same 

method for mounting the accelerometers and since no reference 

information from the cadaver test was available. 

General considerations 

The injury-inducing mechanisms in moose-car impacts 

can be divided into two major groups. 

( I ) Direct contact between passenger's head and moose 

body. Prior to this study the generally accepted theory was 

that contact occurred between the head and the upper 

wind-screen frame due to the fact that the frame is pushed 

backwards during the crash. This study however shows that 

the moose body is much deeper inside the companment than 

the frame through all the critical part of the crash. The 

occupants head will thus hit the moose body rather than the 

car structure. A deeper analysis of Volvo's accidental data 

tends to confirm this conclusion. The medical reports on 

head-injured pas$enger$ in all cases show diffuse, blunt 

trauma. 

The deformed wind-screen frame is however a great risk 

factor in the so-called secondary impacts. That is when the 

car after impacting the moose collides with something else, 

e.g. a tree or another car. Secondary impact occurs in approximately 

20Vo of the moose-car accidents. 

Volvo's accidental data also showed that the use of seat 

belts significantly decreases the risk for severe injuries in 

this kind of impact. All accidents in all Volvo models where 

injuries were reported over a period of five years were 

studied. We found 396 injured passengers, all from the front 

seat. Of these ,372had used their seat belt. ln this group, we 

found one killed (AIS 6) and two critically injured (AIS 5). 

In the left 24 cases where the seat belt was not used or the 

usage was hard to determine, four fatally and two critically 

injured were found. 

Even with restrictions and reservations to this brief study 

it is likely that significant differences will remain. 

(2) Glass pieces from the wind-screen give face and arms 

injuries ( I ), (2). This is a very common injury mechanism in 

moose-car collisions but the injuries are in general moderate. 

Severe cases can occur e.g. when glass lacerates the 

eyes, but these are very rare. 

Reduction of the amount of glass pieces emerging from 

the wind-screen would reduce the number of injuries of this 

kind. 

References 

(l) Lind, B. (1981) "Viltolyckor, sammanstiillning av 

olycksmaterial", Report from Volvo Car Corporation, 

Criteborg (in Swedish). 

(2) Thorson, J. (ed.) (1985) "Moose Collisions and 

Injuries to Car Occupants", Reporl from Departments of 

Environmental Medicine, Surgery and Forensic Medicine, 

University of Umei and Occupational Health and Safety for 

State Employees, Umefr (in Swedish). 

(3) Turbell, T. (1984) "Simulated Moose-Collisions, a 

methodology study", VTl-meddelande nr 402, Swedish 

Road and Traffic Research Institute, Linkiiping (in 

Swedish). 

Acknowledgement 

We would like to thank professor Bertil Aldman, head of 

the Department of Injury Prevention at Chalmers University 

of Technology, for valuable comment and advices to our 

work. 

We would also like to thank veterinarian Bengt-Ole 

Rtiken forsharing with us his deep knowledge of the mooseanatomy 

and for help with rigging, dissection and postcrash 

analysis ofthe cadaver. Thanks to Swedish Road and 

Traffic Research 

Institute for practical help and to Swedish 

Transport Research Board for financial $upport.

Investigation, Evaluation, and Development of AdvancedConcepts in Three-Point 

Belt Comfort Enhancement Devices 


David J. Biss, 

David James, Ltd. 

Abstract 

An investigation was conducted into the $ources of slack 

caused by comfort and convenience devices, which can 

affect the safety performance of the basic three-point belt 

design. Limited field observations were made to correlate 

$ources of slack identified from the literature and personal 

experience to the wearing configurations of shoulder straps 

on the typical American highway. From this background, a 

research program was carried out to investigate with $led 

tests in the laboratory the dynamic restraint performance of 

three-point belts with previously observed and documented 

amounts of slack. The tension relief (eliminator) "windowshade" 

device (TRD) was chosen as the item for laboratory 

scrutiny because it is the single most frequent cause of 

slack in American belts and because it produces a pattern of 

slack which mimics that from many other sources. The 

results of these sled tests define the expected envelope of 

performance of a typical TRD. From these results, design 

improvements to three-point belt comfort and convenience 

devices are suggested which will improve, in an immediate 

and cost effective manner, the safety performance and reliability 

of belts containing these devices. 


The Three-Point Safety Belt 

It is particularly appropriate to be discussing the safety 

performance of the three-point belt system here in Giiteborg 

thirty years after Mr. Nils Bohlin, a native son of Sweden 

and a long time resident of Gdteborg, received a Swedish 

patent (l)* on this device. The foresight ofthe inventor is 

best demonstrated by noting that the basic configuration 

illustrated in his patents ( 1,2,3) of a single strap, three point 

belt system anchored at the B-pillar, tunnel and sill areas 

(figure l) is used in the majority of the worlds' automobiles 

today. Retractors have been added to the basic design, and 

certain derivative designs have appeared, some with 

multiple straps, dual spool retractors and other features, but 

the safety performance expectations for these later 

additions, changes and refinements, are still judged against 

the basic requirements for safety belt performance Bohlin 

laid down over three decades ago. The most basic of these 

performance requirements was that a properly designed 

three-point belt must restrain a person in a "physiologically 

favorable manner" (2) by interacting appropriately with the 

human anatomy. 

The subject of the present study encompasse$ the safety 

performance implications of adding "comfort and 

*Numbert in parenthescs designate rfcrences at end of paper 

416 

Deut5cfe Patente 

EnNnB 6 

. PricdttrhJifrNeil0lg87 

NILS TVAR BOHTIN 

Gotcbary (Schrvcdcn) 

Sicherheitsgxt ltir Fahrzeuge 

p4cnticd vom 21. AugEt lt59 ft 

t)rFDrdnunlF-5i(h.rhcit*un wrcin v*ntli.hfi Schri$ in dcr EnRirllunr dcr 

trnckhrhwor'rhrungcn lir Emfrfrhmge- Dic bir d$in bckrnnten Sirhcrhcirfiunc 

ih'm cincn shidlicltn. nrh untcn rcrichtctcn Druc! rufd* Rnrtlnr rur 

unrl ltxrtrusn cimn *t*ndichcnTtil dcrSprnnunp im Brurtlorb rufdhltwtohh*. 

&'hli;r trfin'Jlng bcrillt cimn Sichcrb.irl$.r, dcr rcwohl rlcn Obtrldrp+ 

Ejf-Q$h .lqn LJtrrErkilru in phyriolqi*h gnnsti3cr Vci* fcthilt, Drzu dicm lc cin 

tkrlcntxxhlg a bcidcn Scitcn d6 Sitrc! ud eid rurmmcnhin6enrJc Sclrling+ 

Ertildfi ilt Brutr und Hofrg!fi. 

Flgure 1. Bohlln's patent lllustration lor the threo-point $af6ty 

bslt. 

convenience" enhancement devices to Bohlin's basic 

design of the three-point belt. 

Safety Belt Comfort and Convenience. 

The history of the three-point safety belt is inseparably 

interwoven with the development and addition of devices to 

enhance the comfort and convenience of the wearer (4,5). 

The concept behind certain devices such as the emergency 

locking retractor (ELR) are so fundamental to the acceptability 

of the three-point belt to the average user that there 

has been little debate concerning the appropriateness of 

introducing reliable ELR designs. 

Other devices such as comfort clips and tension relief 

devices (TRD) devices-more generically known as windowshade 

devices) have generated much more controver$y 

because they introduce pre-impact wearing slack that is 

obvious to the trained eye might well enhance belt user 

injuries. 

The discussions and arguments overthe years concerning 

the safety implications of introducing these and other comfort 

and convenience devices have generally centered 

around a lack of reliability to lock when required, or the

propensity to introduce slack, either in the preimpact, or the 

crash environment (4,5,6,7). Fundamental in these argument$, 

implied or explicit, is some underlying defirnition or 

concept of the "comfort and convenience" gained by the 

addition of these devices against which the downside risks 

can be compared. 

Probably the most comprehensive attempt to formalize a 

definition of comfort and convenience and to quantify attributes 

thereof, was in the NHTSA Notice of Proposed 

Rulemaking, Notice 17, of December, I979 (8). This action 

elicited numerous comments which can be summed up as 

follows; 

. , . the relationship of comfort and convenience to 

wearing rates has not been established and there is 

evidence to the contrary (9). 

Comfort and convenience goal.r.-Although this attempt 

at quantifying and regulating the comfort and convenience 

of safety belts by the NHTSA met with response$ typical of 

that above ( 10, I I , I 2), clearly there are some basic attributes 

safety belt designers have been striving for over the years. 

The more progressive features seem to have emanated 

mainly from the application of fundamentally sound and 

ethical design practices by applying research, evolution, 

and simply user feedback. Some characteristics of the main 

comfort and convenience concepts and devices are; 

r Neat, safe and clean stowage; convenient to grasp, 

extract, buckle and unbuckle 

r Limited or adjustable impingement on sensitive 

areas: 

Face 

Neck 

Female breasts 

r Permit occupant non-crash/non-braking movements 

inside the vehicle 

t Limit webbing retraction forces (European design 

philosophy) 

t Cancel webbing retraction forces (US design 

philosophy) 

There are distinctly different perceptions in Europe and 

the United States concerning the relationship between the 

fundamental safety requirements of safety belts, and the 

effect comfort and convenience features might have on this 

safety performance. The most enlightened approaches in 

Europe stress an integrated approach with attention to the 

fundamentals of safety belt performance and comfort such 

as anchorage location, and reliable and smooth retractor 

function, including attention to webbing tension, component 

friction, etc. (4,13,14). This European philosophy can 

be summarized by the concept that a large measure of 

"comfort" 

is derived from the security of knowing rhat the 

safety belt system provided will protect to the best extent 

possible 

in a crash (15). 

The American approach has been to add extemal devices 

such as windowshades, guide loops and comfort clips to 

the belt to achieve some definition of comfort 

(5,7,9,10,11,12,16).In the past, the more frequent and varied 

styling changes in U.S. cars $omewhat precluded the 

European approach. This American design philosophy predetermined 

that there would be numerous forced tradeoffs 

between safety belt safety performance and comfort 

(17,18,19). 

Also, comparing pre and post mandatory belt use law 

(MUL) rates in both regions (20), the u$age rate in European 

vehicles (operatedboth in Europe and the Unired States) has 

been consistently and significantly higher than in American 

automobile$. The differing European and American philosophical 

approaches to three-point belt design and safety 

performance evaluation cannot be traced to fundamentally 

different medical or biomechanical considerations. By all 

measures, the European integrated safety approach has been 

more successful because, the safety belt designs which have 

resulted are not significantly compromised over the original 

Bohlin concept, and yet, are more widely accepted by the 

vast majority of a population spectrum not fundamentally 

different than that found in the United $tates. 

Comfort and Convenience 

Enhancement Devices 

Discussion 

E me r g e ncy I o c k i n g 7's I t'ss 1p t'.-The emergency locking 

retractor (ELR) was the first major supplemental device 

added to the basic three-point belt. Its development was 

underway in Sweden in 1962 and regulations for approval 

were adopted there in 1967. The ELR became standard 

equipment on Volvo cars in 1968 (4). It is interesting that the 

first Swedish regulations on ELRs required that they be 

sensitive to both webbing and chassis accelerations, and this 

requirement has carried through to the present day. Soon 

after ELRs were introduced into significant numbers of cars 

in Sweden, adjustments were made to locking sensitivities 

to satisfy user complaints concerning lock-up during 

fastening of the belt. As other countries introduced ELRs in 

the coming years they had the benefit of the prior Swedish 

experience as well as generating a learning curve of their 

own (19). 

The typical American designed ELR is vehicle 

acceleration sensitive (VSR) only, FMVSS 2(X) requires a 

locking sensitivity of 0.7 G before the webbing extends I 

inch. The typical European retractor is sensitive to both 

vehicle and webbing accelerations (mandatory in Sweden) 

and is required by EEC 771541to lock at 0.45 G before 2 

inches of strap movement. The Federal requirements are 

that the retractors lock when tilted over l5 deg. EEC77 /541 

requires that locking take place when the ELR is tilted l2 

deg. or more. 

Emergency tensioning retractor.-The next major 

practical advancement in ELR technology was when 

Daimler-Benz, in conjunction with REPA (21,22), 

developed the emergency ten$ioning retractor (ETR). The 

first generation design would, upon receiving a signal from 

a crash sensor, pretension the continuous loop belt webbing 

4t7

to 3fi)'-400 lb. Later design objectives were primarily to 

eliminate slack by rewinding up to l0 cm of webbing when 

slack was pre$ent. The ETR retained all the comfort 

advantages of the ELR with the added safety performance 

that the pretensioning during frontal collisions not only 

eliminated any slack from the belt, but tied the occupant 

tightly to the vehicle for maximum ridedown benefits. A 

number of patents (24,25,26) for pretensioners had been 

granted in the 1970's but Daimler was the first to implement 

the technology into a significant number of its Mercedes 

cars beginning in 1980. The Daimler ETR, however, 

initiates only upon a frontal impact within the sensitivity 

zone of the air bag crash sensor. Patents by Takata (23) and 

VW (24) pointed out accurately that intermittent locking of 

ELRs during rollovers and other multiple impact crash 

modes permitted gross amounts of spool-out in some 

instances, and their patents put forth pretensioning and 

positive locking as countermeasure for these and other crash 

modes. 

Passive belts.-Some passive belt designs have major 

convenience features incorporated into their designs. The 

shoulder strap of the Toyota/Takada motorized anchor 

system is extremely easy to enter-so easy in fact that even 

a seasoned belt wearer at times forgets to use the manual lap 

belt. Additional periodic chime wamings and easier access 

to the stowed and donned lap belt buckle should be 

considered for inclusion into this system so that lap belt 

usage is encouraged to the extent possible. 

Semi-passive helt features,-An example of a semipassive 

belt feature, as the concept is used here, is the 

Mercedes motorieed extension arm that hands the occupant 

the belt buckle from difficult stowage positions in coupes, 

etc. 

The comfort and convenience implications of other 

recent passive and semi-passive belt-designs will be 

discussed below in connection with slack they can be 

expected to introduce. 

Dual spool ELR retract4l's.-Qys1 the years various 

designs of dual spool ELR retractors have been used, 

usually in a configuration of one retractor for the lap belt 

and one for the shoulder belt. On some models. Volvo uses a 

dual spool, dual sensitivity retractor in a single strap design 

to minimize webbing on the reels while maintaining 

acceptable comfort levels. If in these multiple retractor 

designs, the retractors function smoothly and lock reliably, 

the safety performance of these systems should be 

equivalent to that of the basic three-point continuous loop 

three-point system. 

A system in which separate lap and shoulder belts are tied 

together at a latch plate is becoming more common in 

American passenger vehicles. This is particularly true for 

certain "passive" belt designs in which the two ELR 

retractors are mounted in the door. In the American dual 

spool retractor design, however, vehicle-sensitive-only 

retractor$ of U.S. Federal specification are being used in the 

lap belt. Should there be no lock-up or delayed lock-up in 

one of these $y$tem$, particularly during a rollover, the 

418 

occupant can face increased excursions and possible 

ejection from the belt. 

Automatic lotking retactors.-Automatic locking 

retractor (ALR) devices certainly earn a mention in the 

history of devices added to belts to increase comfort 

because their intended function is to stow the belt neatly and 

make retrieval more convenient. ALRs have been used 

primarily with lap-strap-only seat belts-not typically with 

belts incorporating shoulder harnesses. If these devices 

function properly and the positive locking feature locks 

reliably, the dynamic performance of the strap will closely 

emulate that of a strap (albeit with higher elongation 

properties) without a retractor. 

Comfort c/lps.-Comfort clips to prevent the webbing 

from rewinding fully against an occupant's shoulder/chest 

were introduced almost simultaneously with the 

introduction of ELR retractors which placed the shoulder 

strap under constant tension against the occupant's chest. In 

Europe, comfort clips were phased out as the retractor 

designs were refined, while in the United States, the basic 

concept of the comfort clip evolved into the windowshade 

device. In conjunction with the installation of shoulder belt 

retractors, comfort clips were installed as standard 

equipment in many American vehicles before the 

windowshade device came into widespread use. The 

fundamental problem with comfort clips was that the lay 

usercan, and routinely does, set an arbitrary amount ofslack 

in the belt without having any real biomechanical 

appreciation or guidance a$ to the resulting hazards 

(r7,r9,27). 

Routing guides, loops, hooks, etc.-Before 1974, the 

major U.S. manufacturers uniformly used a hard mounted, 

no retractor shoulder belt either in a four-point 

configuration or in a Type 2a detachable shoulder strap. 

After 1974, three-point belts appeared in the U.S. with the 

ELR and other structural anchors positioned to 

accommodate the more divergent styling contours of the 

American car. Many times supplementary routing loops, 

hooks, etc, were used to guide the shoulder belt strap over 

the occupant's shoulder and away from the face/neck area. 

Many of these guide loops were attached to adjustable head 

restraints and were designed in conjunction with an ELR 

located on the $ide roof rail behind the occupant. 

Figure 2 is a photograph of an unprompted driver on the 

road which shows the triangulation slack effects such 

routing loops and hooks routinely introduce into a belt 

system. These hooks and loops are usually made of plastic 

and, upon restraint crash loading, particularly when they are 

the element forming the apex of the triangle, they will break 

and the real slack in the belt will manifest itself as the strap 

snaps into direct tension. Although guide loops can be of 

value in certain applications, such as coupes and 

convertibles, their use should be closely monitored to 

ensure that unnecessary and excessive slack is not a 

consequence of this use. 

Te nsi on re lief' (w indow s hade ) device s.-Tension relief 

devices (TRD), or more appropriately called tension

Flgure 2. Trlangulatlon rlack a$socletod wlthgulde loopf. 

elimination devices (19), are generically called 

"windowshade" 

devices in the U.S. because this device is 

designed to be set and reset at different lengths of webbing 

slack analogous to the functioning of a roll-down 

windowshade. The original design concept for these 

devices was pioneered by General Motors (12) and was 

introduced shortly after the 1974 Federal requirements for 

the mandatory installation of rhree-point belts. The TRD 

was promoted as a feature which would make three-pofnt 

belts more acceptable to the wearers and increase belt use. 

The Federal Rulemaking history on rhis device has been 

confusing and contradictory. These devices were first 

addressed in Federal Rulemaking in 1976 (16) with the 

admonition that "the tendency for such retractors to permit 

introduction ofexcessive slack is an argument against their 

use." Nonetheless TRDs were subsequently permitted with 

the requirement that their proper use be explained in the 

associated vehicle owner's manual. 

When further comfort and convenience inspired Federal 

Rulemaking was initiated in late 1979, the (U.S.) Moror 

Vehicle Manufacturers Association stated in their response. 

"Experience 

shows such features as emergency locking 

retractors, tension-relieving devices, webbing guides and 

improvements in anchorage locations have not resulted in 

significant increased belt usage" (10). The Ceneral Motors' 

response to this Docket stated, ". . .we (CM) pioneered the 

tension-relieving device because we believe that persons 

disposed to use belts might be discouraged by constant belt 

pressure. . . ." But the information that is available is 

contrary to the NHTSA contention that the lack of comfort 

and convenience in today's belt systems is the cause of low 

use. . . (l?). 

The controversy surrounding the introduction of the 

windowshade was evident even inside General Motors. A 

1976 management evaluation (28) of the windowshade 

TRD reported, "On numerous occasions during the driving 

Frocess I glanced down to find the shoulder belt with 

excessive slack. . . . In all, I find the mechanism tricky and 

don't have confidence in what it's going to do next." 

In the belt supplier industry, the view concerning the 

widespread introduction of the TRD was ambivalently 

expressed in 1977 by Henderson (19): 

"Although 

the author believes the tension 

eliminator device represents a considerable step 

forward in safety belt comfort when used intelligently, 

he is uncertain of the wisdom of risking the excess 

slack which reduces the safe deceleration space for the 

occupant within the vehicle." 

This author goes onto explain the availability of a "belt 

tension reducer" which provides two levels of retraction 

force and ". . . avoids the danger of excess slack and no 

action by the user or complicated automatic devices are 

necessary to ensure colTect stowage" (19). 

The TRD device does make the shoulder harness more 

'*comfortable," 

in the sense that it reduces the webbing 

tension forces on the shoulder to zero. This type of 

"comfon," 

however, is the same type of comfort achieved 

if the belt is not on at all. In Europe another dimension is 

included in the concept of "comfort" which emphasizes the 

feeling of security obtained by having a shoulder strap snug 

against the body ready to protect in a collision (15). The 

windowshade device in its present design configuration 

then appeals to tho$e who are not predisposed to wear the 

belt in the first place. It is questionable whether the 

generally good performance of the basic three-point belt 

should be compromised for all users, in order to make a 

special appeal to recalcitrants. 

Although a number of TRD designs were refined over the 

years in attempts to minimize the possibilities for excess 

slack, many were not. Currently there are still designs on the 

U.S. market which will routinely introduce l8 inches of 

slack, or indeed any amount of available slack, during the 

same type of occupant movements the ELR was originally 

designed to permit. 

Tension reduction ( lowering) devices.-Tl,Ds are similar 

to TRDs because, within a design envelope of belt 

movement a few inches forward of the normally seated 

occupant, the tension forces are reduced by a dual level 

rewind spring device that can be reset to maximum 

retraction force with a windowshade release type 

movement. The main difference with the TRD 

windowshade, however, is that the TLD does not eliminate 

the tension; rather, it reduces the retraction tension force in 

this comfort zone to a level which will still keep the strap 

snug on the shoulder. Henderson (19), in his 1977 report on 

devices available to enhance comfort and convenienge, 

described in some detail the functioning of rhe TLD. Certain 

high-line Toyota models currently incorporate this TLD as a 

comfort extra. Although this device avoids the excessive 

slack problems of the TRD, the reduced tension can become 

a problem if the belt twist$ or encounters excessive friction 

in the hardware guides. These potential problems should be 

amenable to correction by careful, coordinated design of the 

components. 

Adjustable shoulder helt anchor point.-Some 

manufacturers have chosen to address the belt fit problem 

4r9

through the use of an adjustable, load carrying shoulder belt 

anchor point. This feature is standard on some Mercedes 

and Saab models, and has been an available option on 

Volvo. The attractiveness ofthis feature is that it can solve 

the belt fit problem without any degradation in optimum 

belt performance. However, it requires a motivated user to 

change the anchor position and a knowledgeable user to set 

the anchor in a biomechanically advantageous position. For 

most u$ers. and for most belts without windowshades, a 

comfortable anchor location with the belt on the shoulder 

will ensure near optimum belt safety performance as well. 

Even with these caveats, if this device i$ set at the opposite 

extreme indicated by occupant size, the effectiveness ofthe 

shoulder belt restraint is not grossly compromised. The 

adjustable shoulder anchor device is most useful for regular 

occupants of the same vehicle riding in the same seating 

positions, particularly for children. 

The Safety Implications of Comfort 

Devices 

The safety implications of the comfort enhancement 

devices discussed here are most frequently attributed to the 

possibility these devices will inadvertently or deliberately 

introduce slack into the belt system in the process of trying 

to achieve some sort of "comfort." An in-depth discussion 

of the safety effects of this slack requires consideration of 

how such slack originates and of what its effects are? 

Safety Belt Slack: A Definition and History 

The terms "belt slack," "occupant ride-down," and "restraint 

system efficiency" have been used frequently over 

the years in the literature, but typical use refers only to 

general concepts with little precision or definition. Slack as 

it is used here falls into two categories. First, pre-impact 

slack will mean a loose belt which is prone to as$ume a 

dangerous position on the body before the impact. Second, 

dynamic impact slack arises as the occupant moves forward 

inside the vehicle with low or no restraint forces due to preimpact 

slack, delayed retractor lock-up, or restraint system 

compliance. 

One early school of thought for maximizing belt restraint 

system performance was to simply tie the occupant as tightly 

as possible to the vehicle interior in an attempt to achieve 

the $ame chest accelerations as the vehicle's crash pulse. 

In 1968 Bertil Aldman and Ame Asberg of Gdteborg 

performed some of the earliest reported rigorous research 

(29) into the issue of slack effects on three-point belt restraint 

performance. They modeled simplified three-point 

belt systems with high and low strap elongation properties 

restraining a simplified single mass occupant, the responses 

of which were comparable to chest responses for an actual 

vehicle occupant. The modeled systems were then subjected 

to input crash pulses from actual vehicles of various 

onset rates, amplitudes and durations. 

Their research and a review of previous work (30) did 

show that for an acceleration test environment with a low 

4ZO 

rate of onset, and for tight restraint straps that resulted in 

negligible occupant displacement relative to the vehicle, the 

occupant's chest respon$es were approximately the same 

level as that for the vehicle. However, for actual car crash 

environments and practical automotive restraint characteristics 

they found: 

"The 

resulting peak acceleration ofthe occupant is 

always higher than this average deceleration ofthe car. 

. . . impact amplification always exists when (realistic 

automotive) belts are used. . . . The amplification results 

from the general shape ofthe input (crash) pulse, 

the stiffness of the restraint, and any slack present in 

the system. The introduction of any appreciable slack 

is disastrous to the protective capacity of stiff systems 

by producing high peak accelerations with a high rate 

of onset. . . . The safety obtainable in any car is wasted 

if it is equipped with a stiff restraint system that can be 

used with a slack. Therefore. it seems advisable in the 

future to take into consideration the deformation characteristics 

of the car and the restraint together. " 

Aldman's and Asberg's research proved to be insightful 

because, even though they did not directly mention the slack 

effects of ELR retractors on three-point belt performance, 

their results were directly applicable to the issue of the I 968 

introduction of ELRs into production cars in Sweden. 

ln 1977 Morris analyzed three-point belted occupant response 

data from full scale frontal barrier tests of 1976 

model year automobiles to determine the effects of slack on 

restraint system performance (3 l). Morris developed a useful 

definition of ridedown by overlaying the shoulder belt 

forces onto the acceleration crash pulse (figure 3) and calculated 

the ridedown from the overlap of these curves. His 

approach took into account the influences of ELR characteristics 

such as lock-up time, amount of film-spool, etc., as 

well as webbing elongations. His results showed a strong 

correlation between ridedown duration and peak shoulder 

! 

4 

c 

t 

o 

E 

Illustratlons of polnts used to detennlne 

duratlons of llde down. 

Flgure 3. A dellnltlon of rlde down (from Morrlg, reference 31).

elt loads. None of the vehicle crashes Morris analyzed had 

appreciable pre-impact slack set in the windowshade, so 

any slack time he documented was from ELR and webbing 

compliance influences. 

In 1980 Reichert and Bowden (32) of Departmenr of 

Transport Canada published the results of sled tests conducted 

to a$sess the effect of slack caused by windowshade 

comfort devices on the safety performance of three-point 

belt systems. They compared three-point belts with no retractors, 

fl retractor in the shoulder belt, and a retractor in the 

lap belt each with 2.5, 5 10, 15, and 25 cm of slack. Their 

results showedl 

. . . (The no retractor) belts produced an almost 

linear increase in slack time, head excursion, peak head 

acceleration, and HIC with increasing slack. 

. . . (The retractor at the shoulder harness anchor) 

belts showed much higher values of head excursion 

and HIC at low slack (settings) and (a) lesser dependence 

on slack. Peak acceleration ofthe chest was less 

than with type A belts at large slack(s). 

. . . (The retractor at the lap anchor) belts performed 

similarly to the Type A (shoulder rerracror) belts. 

Retractors modify the effect of slack on occupant 

dynamics and appear to reduce the prolection against 

head contact provided by seat belts. 

Aldman's and Ashberg's research again proved, in retrospect, 

insightful because, their results were later to be confirmed 

by Monis, and Reichert and Bowden, among others. 

From this research (29,31,32) and others (7, 13,24,33, 

34, 35, 36, 37, 38, 39), the theoretical limits on three-point 

belt restraint system performance in a particular vehicle for 

the driver or passenger can be determined by analyzing the 

vehicle stopping distance; the vehicle deceleration crash 

pulse the principle direction offorce; the elastic and energy 

absorbing properties of the restraint system; and, the distance 

inside the vehicle available for occupant translation 

and articulation. Applying these general principles ro a specific 

vehicle environment will, in conjuncrion with judicious 

design decisions, lead to optimizing the three-point 

belt performance in that vehicle. This process will only 

succeed, however, ifthe full envelope ofpossible effects on 

belt performance of slack caused by devices such as ELRs, 

TRDs, routing loops, moving anchor points, etc. are quantified 

and considered. 

Slack Caused by Comfort and Convenience 

Devices 

Slack caused by comfort and convenience devices arises 

from two sources, pre-impact slack in which a loose belt is 

prone to assume a dangerous position on the body before 

impact; and, dynamic impact slack in which the occupant 

move$ forward inside the vehicle with low or no restraint 

forces due to pre-impact slack, delayed retractor lock-up, or 

restraint system compliance. 

Slack caused by ELRs.--The primary failure mode of 

concern for the ELR is that it may not lock or might lock late 

during an impact. Although one might be able ro express the 

probability of this happening as a small fraction, rhe consequences 

are nonetheless catastrophic when it does happen, 

indicating that special attention during an effects analysis is 

warranted. The function of the ELR is to permit wearers to 

move their upper torsos around freely inside the vehicleup 

to the limits of webbing available-in the non-braking, 

non-crash situation. This free movement capability changes 

from a desirable attribute in the non-crash situation to a 

failure mode if exhibited in the crash condition. 

The typical American designed ELR is vehicle acceleration 

sensitive only with the locking mechanism being a 

suspended, Iead-weighted pendulum lifting a locking bar 

into the path oflocking teeth on the retractor spool ends. The 

typical European retractor is sensitive to both vehicle and 

webbing accelerations. FMVSS 209 requires a vehicle acceleration 

locking sensitivity of 0.7 C before webbing extends 

I inch, while EEC77l54l requires European retractors 

to lock at 0.45 C before 2 inches of strap movement. 

Federal regulations also require that the retractors lock 

when tilted over 15 deg. EEC 77 l54l requires that locking 

take place when the ELR is titled 12 deg. or more. The 

American Federal ELR then is less sensitive to these two 

standard prescribed locking modes than is the EEC ELR and 

there has been some discussion that this lower sensitivity 

can be expected to affect the outcomes for accidents on the 

highway (4,6,15,23,24), particularly for multiple impacts 

and rollovers. 

Horizantal plane impacts.-The ideal conditions to promote 

reliable vehicle sensitive ELR pendulum locking are 

impacts confined principally to a horizontal plane in which 

the vehicles do not roll appreciably and which generate a 

single delta-V crash pulse which maintains a relatively constant 

absolute level without radical fluctuations. Such conditions 

are generated in the typical frontal barrier crashes 

but discussions have occuned concerning whether these 

laboratory conditions and the FMVSS 209 type test apparatus 

adequately evaluate ELR's in foreseeable accidents on 

the highway. Bohlin in l98l (4) stated: 

"There 

are no regulations today on the testing of the 

vehicle sensitive locking function which probably reflects 

the reality. . . . It is therefore proposed that the 

regulations regarding the emergency locking features 

of the retractor should be seriously reconsidered." 

In 1975 Mackay (6) analyzed failure modes for ELR belts 

and observed: 

"The 

third failure mode applies to inertia reels 

where complete failure to lock or locking so late as not 

to provide adequate occupant protection, has been recorded. 

Such cases are characterized by occupant trajectories 

and contacts similar to those observed for 

unrestrained occupants. . . . Subsequent examinations 

of such reels have not shown a cau$e, and we conclude 

that either preimpact braking or some dynamic characteristic 

of the locking mechanism is responsible." 

421

In the early 1970's NHTSA found that emergency locking 

retractors installed in some automobiles were not locking 

reliably when subjected to the FMVSS 209 acceleration 

test (40,41,42). In the case of one American style retractor 

subsequent investigation showed that initial tip-to-tip contact 

of the ratchet teeth with the lock pawl teeth could cause, 

in some circumstances, the retractor to delay locking or not 

to lock at all. In response to this investigation the manufacturer 

made a number of changes in the locking mechanism 

which apparently rectified the immediate problem. At the 

conclusion of this investigation (41), however, questions 

sunounding the actual reliability of these and similar ELRs 

to lock when required in real accidents were not completely 

resolved. The retractor$ subject to this NHTSA investigation 

were the same basic design as those used in most 

American cars up to the present. 

Rollover and multiple impacts.-*-{here have been instances 

investigated by this author where occupants ofvehicles 

involved in rollover accidents were found with threepoint 

ELR belts still around them but with excessive 

amounts of slack up to and including the full unwound 

amount. There have been other similar notations during the 

years of experience with the ELR concerning the possibility 

that the simple undamped locking mechanisms in some 

ELRs might not lock reliably in rollovers or multiple impacts. 

A number of patents address these vehicle dynamics 

considerations including one by Takata-Kojyo (?3) which 

stated: 

While these (typical ELR) systems possess many 

advantages, they are frequently unreliable since the 

reel braking is either of too short a duration, or is 

maintained for an indefinite period and they otherwise 

leave much to be desired. . . . Another object of the 

pressnt invention is to provide an improved inertia 

responsive safety belt locking system in which an adequate 

braking interval is assured independently of the 

sequence of events following the braking actuation. 

A patent by Volkswagen (24) also described the real 

world crash environment in which ELR retractors are called 

on to reliably lock. 

Currently known automatic (ELR) belt winders . . . 

release . . . the safety belts after tightening (locking) 

thus defeating the desired restraining action. . . . a 

stopping or locking device is provided for maintrining 

the increased safety belt tension, once applied, in order 

to prevent the vehicle occupants from falling free in the 

direction of travel in the event of continued deceleration. 

Such a stopping device is especially important in 

accidents in which the vehicle undergoes several successive 

impacts . . . or ovefiums. 

These patents as well as indications from field accident 

investigations, indicate that certain oscillatory dynamics 

are present in multiple crash modes which are strongly 

suspect in causing intermittent locking, or failure to lock, of 

ELRs (4,6). A major difficulty surrounding public discus- 

422 

sion of these possible failure modes is that if such a failure 

occur$, the pattern of evidence looks the same as if the 

occupant was unrestrained. Realistic multiple impact dynamic 

testing related to real world accidents should be undertaken 

with the specific intention of finding the intermittent 

and unlocking failure modes of typical ELR designs. 

Once these failure modes have been identified, discussions 

can commence onhow to evaluate different ELR designs for 

this type of vulnerability, and, the search for improved 

designs will be further stimulated. 

Film spool s/acft.-A major finding of the NHTSA 35 

mph New Car Assessment Program begun in 1979 was that 

the film spool effect

years and had been phased in for a number ofyears before 

that. Thus, over 100 million automobiles presently on the 

American highways are equipped with this device resulting 

in significant safety implications to the driving public. 

The typical belt tension eliminator (reducer) "windowshade" 

device can be readily added to almost any ELR 

spool shaft. The TRD introduces slack of two types into the 

three-point belt: pre-impact slack, which precludes the belt 

from automatically and completely retracting to fit the 

wearer; and dynamic slack which can sacrifice crash phase 

ridedown and increase occupant body segment excursions 

inside the vehicle. 

Concerning pre-impact, wearing slack-there have been 

analogies drawn between the intended slack a windowshade 

device introduces into a torso belt and the recommended 

slack (one fist's width) in the pre ELR torso belts. This 

analogy is misleading from a number of standpoints. 

Major changes in operational performance of three-point 

belrs occurred wirh the addirion of rhe ELR (32). The most 

major of these was that the conventional ELR introduced a 

certain amount of unavoidable dynamic slack into the system 

from, at minimum, film-spool and the elongation of the 

associated extra webbing. Even though this dynamic ELR 

slack has been minimized in many recent designs, it still is 

the ca'te that any slack introduced by the windowshade 

device is cumulative to normal ELR associated slack. 

In addition to slack effects which permit additional forward 

motion of the occupant with little or no restraint 

forces, there are slack effects which permit the shoulder 

strap to assume, pre-impact, a dangerous configuration on 

the body. This occurs, for example, when the strap lies off 

the corner of the shoulder. This study has demonstrated that 

this configuration can occur with as little as 1.3 inches of 

pre-impact slack and this study further confirms dangerous 

abdominal loading as an expected consequence in both 

frontal and frontal oblique impacts. 

Further contradictions to the assertion that the present 

designs of the TRD will introduce only a prescribed amount 

of slack, are observations as well as personal experience 

showing the windowshade can, and often does, introduce 

gross amounts of slack (4 to 24 inches) during normal driving 

movements ( 17, 18, 19,27). These are movements of the 

type the ELR was originally designed ro permir, but the 

intent of the basic non-TRD ELR design is that the belt 

webbing must unwind and rewind to follow the wearer at all 

times. 

Typically how the TRD causes excess slack can best be 

explained using the schematic graph shown in figure 4 in 

which belt length unwound from the reel is plotted against 

time periods of arbitrary duration. The first time interval 

represents the occupant pulling out extra belt webbing to get 

it around himself for buckling and then the belt rewinding 

snug back onto the shoulder. The second interval includes 

the limited movements of the shoulder/torso with the belt 

snug. The third event is the occupant leaning forward within 

the windowshade cam setting distance, then leaning back 

with the windowshade still set. This prescribed amount of 

l 

f 

5 

! 

) 

I 

o 

.J 

I 

5tsdlrdr. u'Eiltotron r". r."rô pqst --f-- 

rne 

Figure 4. A human factors analyria of the operatlon of the tenslon 

reliel "windowshade" devlce. 

slack has been in the range of2 to 5 inches for various TRD 

designs, but in recent years is typically around 2 inches. The 

next interval depicts the wearer's motion to reset the windowshade 

to the snug configuration. 

The next two intervals show what typically happens when 

a driver, for instance, leans forward to look for cross traffic 

or to manipulate the radio controls. If the shoulder of the 

wearer oscillates slightly while the belt is in rhe extended 

position, the entire sequence mimics the motion of the occu- 

pant putting on the sear belt in the first place. The TRD is 

capable of setting all the way out to the end of the spooled 

webbing, and being a simple mechanical cam device, it 

cannot discriminate many times between the motions of 

putting the belt on and just leaning forward with a slight 

rocking motion. Of course, one can stress that TRD users 

should closely monitor their belts for no more than the 

prescribed amount of slack, but this does not answer the 

question of what the hazards are during the periods the belt 

strap is away from the body in its TRD set, or process of 

being reset, condition. 

An interesting insight comes to light when one reviews 

the history of the windowshade TRD in conjunction with an 

analysis of the operating characteristics of the combined 

TRD-ELR system. Figure 5 shows webbing tension forces 

ploned against forward extension of the belt for two makes 

of American cars equipped with the TRD and for a European 

car without the TRD. The starting point was with the 

strap snug on the occupant's shoulder. Two competing 

makes of American cars exhibited essentially the same webbing 

tension versus webbing extension characteristic and 

these forces are about 50d/o higher than those exhibited by 

the European ELR. Also, in the critical area where the belt is 

snug against the normally seated wearer, the American tension-extension 

characteristic rises much faster than that for 

the European ELR. 

The present TRD design then has been used in somewhat 

of a self-fulfilling prophecy role-thar of eliminating retraction 

fbrces which are in the first place higher than for 

belt systems which do not contain the device. This dichotomy 

is understandable when considering one of the main 

design requirements of the TRD. The tension elimination 

condition must be immediately cancelled when the door is 

- 

.

n 

c 

'{J 

- 

U 

c 

30 

c E 0 

ut 

0J 

0, 

tr ia 

E 

-a 0, 

> 

r 6 

J 

c 

rI 

c 

ol 4 

c 

_a 

0 4 S 

Shoulder Forword 

Movennent (tn,) 

r l l 

0 1 0 e 0 3 0 4 0 

Shoutder Forword Movenrent (crr,) 

Floure 5. Webblnq tengion versua lonrard extenslon for TRD 

an? non-TRD thre-e-point belta. 

unlatched because, in turn, there must be an immediate and 

strong spool torque applied to overcome the inertia from a 

stationary belt, and thus avoid having a non-retracting belt 

fall out the door to be damaged or to trip an exiting occupant. 

A door-operated plunger is used with TRDs to cancel 

the windowshade setting and permit this retraction when the 

door is opened. 

Although many concerns have been raised over the years 

about the use ofa device to deliberately introduce slack into 

the three-point belt ( 19,28,32); it was not until 1987 that the 

first comprehensive study on how belts equipped with the 

TRD were being worn on the highway was carried out and 

published by the Insurance Institute for Highway Safety 

(?7). Observations for this IIHS study were taken of suburban 

Maryland drivers and they showed that for restrained 

drivers ofTRD equipped cars, 27 percent had I to 2 inches 

of slack in their belts while 8 percent had 3 inches or more. 

As a direct comparison, foreign cars without the TRD were 

monitored and only 5 percent of the belted drivers in those 

had I to 2 inches of slack while none were observed to have 

over 3 inches of slack. 

This author has performed a brief audit of the IIHS study, 

again for suburban Maryland drivers, and has found it to be 

a representative sample of the slack visible through the side 

window from a distance. A difference in procedures with 

the IIHS study was followed in this audit in that drivers were 

approached and asked to have the position of their threepoint 

belt photographed from next to the car where the 

photographer could get a picture of the entire shoulder strap 

and parts of the lap strap. If these drivers moved their belts 

in any way before the photo, they were not included in the 

study. 

The photographic results of this study showed that in 

many instances where the IIHS probably would have quan- 

424 

LrEanql 

A 

x 

o 

{ 

1985 Chevy Cavqlltr 

1986' l{.rcrebs lsE 

1985 t'lercury Horqus 

,rt 

l 

tified 0 to 2 inches ofslack, from an obserrration point close 

to, and looking down into, the car; there was observed an 

additional I to I inches of slack lying in the occupant's lap 

next to the latch plate (see figure 6). The findings from the 

IIHS study are valuable considering the observation viewpoint 

they had; however, the amount and frequency of slack 

they reported is probably understated because they could 

not $ee the latch plate area where a significant portion of 

slack webbing slack typically accumulates. 

It was also noted in the present observations that about 5 

percent of the occupants were using the windowshade device 

as a facilitator for wearing the shoulder harness under 

the arm without the normal discomfort caused by retractor 

tension (see figure 7). This photographic based study is 

continuing and additional results will be published 

elsewhere. 

The TRD can inadvertently introduce slack from occupant 

dynamics during pre-impact vehicle dynamics such as 

braking maneuvers and running over rough ground. During 

such occupant jostling, the TRD can set itself resulting in a 

slack belt at impact. It is appropriate to also add here that in a 

number of the sled tests reported in the next section, the 

windowshade slack did not reset to zero in the belt even after 

a 30 mph impact. Thus, the implications for the present 

design of the TRD is that it reduces the protection of the 

3-point safety belt in multiple impacts in yet another way' 

The EEC regulations 77l54l give a good overview of the 

concems which should be considered in conjunction with 

TRDs. 

3.2.2.1. the straps are not liable to assume a dangerous 


3.2.2.2. that the danger of a correctly positioned 

belt slipping from the shoulder of a wearer 

as a result of his/her forward movement is 

reduced to a minimum. 

Again the findings from the sled test portion of this study 

confirm the significant probability of the belt straps slipping 

from the proper areas of the body to dangerous areas because 

of the dynamic influences of slack during frontal or 

frontal oblique impacts (e.g. shoulder to the abdomen). 

The implications on belt performance of TRDs as they are 

presently designed sometimes lead to biearre results. Observations 

by the author, which are exceptionally difficult to 

record via photography, are that with the TRD set, and the 

near window open, excess slack ends up in a wind blown 

loop behind the occupant. The shoulder strap is snug on the 

user but the wearer many times does not realize there is a big 

Ioop of excess belt strap behind him, and apparently from 

the number of times this has been observed, does not realize 

the hazard of the situation. 

During the history of the TRD windowshade in American 

vehicles there has been until recently a notable absence of 

publicly released data from U.S. researchers both on the 

effects on belt wearing patterns on the highway, and on the 

biomechanical implications for impact protection. There

Flgure 6a 

Flgure 6s 

Fl$urla 6a41. Ilplcal slack In TRD-wlndowshade balts. 

has been no comparable study to that of the 1980 Reichert 

Canadian study (32) in the United States and this has impeded 

both public debate and Rulemaking on this subject. In 

198 I , Culver and Viano (38), reported results for slack belt 

tests in far side oblique impacts but did not report the effects 

of shoulder strap aMominal loading. In 1982 NHTSA conducted 

four sled tests (43) to demonstrare the potential effects 

of slack caused by the TRD windowshade on threepoint 

belt performance. 

Flgure 6d 

Flgure 6b Flgure 6e 

Flgure 6l 

Although the presently reported sled tests certainly do not 

represent a full qualification test program a particular TRD 

device; this test program and associated analysis is presented 

here as a $uggested outline of the types of tests which 

should be run to assess the effects of the types and amount$ 

of slack caused by comfort devices as typically found with 

users on the highway. Because the TRD is the most frequent 

source of this user slack (27), it was chosen as the device for 

an indepth sled test research program to as$ess the affects of 

425

Flgure 7d 

Flgure 7c 

Flgure 7l 

Figures 7a-71. $lsck shoulder belts under arm in THD equlpped vehicles. 

slack on three-point belt performance. Many of the findings 

are applicable to other comfort enhancement devices as 

well. 

Slack effects,from 

"passive" belts.-Some passive shoulder 

belts can be considered convenience enhancement devices 

while others will have to be judged on their individual 

merits. Figures 8a-8c show one design for a three-point 

"passive" 

belt which uses dual spool retractors mounted in 

426 

Figure 7e 

the door, and even though this system contains a normal 

buckle, theoretically at least, one can get into and out ofthis 

car without unbuckling the belt. This design has qualified 

for NHTSA approval, but surveys have shown (44) that the 

usage rate is lower for this "passive" system than for the 

manual three-point belt system it replaced in the same mod- 

el car.

Flgure 8a 

Flgure 8b 

Flgure Ec 

Flgure 8. Slack In thre€-polnt,.passlve" 

belt daelgne. 

There has been a misconception that the TRD device 

would be phased out with the introduction of passive belts. 

The "passive" systems shown in figures 8a-8c are in the 

cars which make up a significant portion of the U.S. market 

share and they do contain the TRD windowshade device. 

When looking at figures 8b-8c it is difficult to determine 

why a TRD is included in these systems because, when the 

belt is taut, the strap is two inches forward of the average 

size occupant's shoulder and at least one inch forward of the 

sterDum. 

From the field observations mentioned earlier. for this 

design of "passive" belt in a four-door sedan, the only way 

for the belt strap to make contact with the shoulder is by 

introducing slack via the TRD. This is, however, not a safety 

feature because this type of strap positioning is unpredictable 

and simply adds more dynamic slack into the system. 

Slack efficts of other "comfort" devices.-A number of 

references to slack caused by other comfort and convenience 

enhancement devices have been included in this 

paper such as routing loops breaking and so on. A number of 

the tests discussed below contained slack, albeit from the 

TRD, which would be typical of these other situations and 

the results should be applicable accordingly. 

A Sled Test Program to Evaluate the 

Effects of Slack-as Introduced bv the 

Windowshade Device 

A laboratory sled test program was undertaken to study 

the effects of belt slack on the $afety performance of a 

typical American-made single retractor, continuous loop, 

three-point seat belt system (VSR/ELR equipped with a 

TRD) using the Hybrid III dummy as rhe resr device. The 

TRD feature was chosen for attention because it is the 

leading cause of slack in seat belts in the U.S. (27). Many 

observations and findings of this study are, however, 

independent of the TRD-the TRD just being the cause of 

the $lack which can occur by other means. For instance, the 

findings from the oblique impact tests apply to slack related 

issues such as jammed webbing, guide loop, triangulation, 

anchor point location, etc. 

General Test Protocol 

Test Environment.-Table I summarizes the test protocol 

and the results for the slack belt tests. Twenty-eight sled 

tests were run in a recent model Ford LTD/Crown Victoria 

front seat sled buck as representing 

the average full-sized 

American car interior layout. For economy of testing and 

repeatability, all occupants were tested in the left front seat 

position. The steering system was installed or removed to 

create a driver or mirrored passenger 

environment. 

H'f i;Jil' o",n;"':lio':*"0_:i::*"01,""1". 

*, ffi 

SH/SLT IIF 

to 

I 

I 

ll 

rB 

r9 

r4 

26 

z1 

m 

b 

26 

22 

t6 

l2 

r{ 

!q.l 9 

-1.0 0 468 tlj 30. I 

)q I o o.o o 525 {oq ,7.7 

19 | d 0.5 o 675 58? j6.o 

n / o Lq 0 6{4 486 38.4 

to / 0 r.6 d ?:0 605 4o.o 

30 / o 2.3 0 02r 730 44.d 

10 / O 3.o 0 1026 88o 4r.9 

Jo / o 5.o 0 910 69E 46,6 

p / 9 t.3 o rd23 986 43.9 

N / 6 q.o 0 !359 1359 rB.i 

30 / o 18,0 0 715 126 37.6 

10 / o r8,o o i98 ?98 {2-6 

10 / 38 

-r.o 

10 / 18 o,o 

30 / 38 0.0 

l0 / 3E r,l 

30 / 1E 2,5 

30 / 36 3,0 

3d / 16 3,0 

10 / 3E 1,D 

36 / 18 3.0 

3d / l8 2.3 

15'/ o o.o 

J5 / O 3.o 

10 / O o.o 

JO / O 3.0 

fi/n -1.0 

10 / o l.o 

422 ts 4,2 

352 228 31,6 

616 {1t 38,2 

379 l0l 29.9 

587 485 37,6 

t96 ?83 3o.t 

F4? 4r4 42.8 

{rJ 325 37,1 

256 16? 34.7 

W 351 J7.9 

ll92 1161 {9-r 

149r r37o ,2.1 

q52 852 32,7 

b97 Ei7 l?,8 

rlr0 1rl0 17,4 

1165 1165 t9-0 

15fr 

l407 

1302 

r4d3 

rl97 

rcrS 

1{2? 

l6r6 

14to 

1292 

ETE 

16fl 

2059 

u36 

18?6 

14C9 

1"95 

1836 

r50l 

r.26 

1,?t 

1 -Zt 

1,19 

l,66 

1.ff 

l.0l 

o.B 

o-s 

- Tssts ss Jo +h B,B,V,, Acruar hlh V. 3t.5 rph. 

- T€8tr EE 15 lph LE,v,, Acrud hlr. v. 37,5 !ph, 

- In t$ts l0 s 11. rhr hlr es 6ff s. shmld;r, 

- q/d dsnotss qu6ttlongtr chsst d6r6tkitm drrr tn Tssrs t? d t8 

. n/t denot€8 ihsF rs no rFmsducsr lof thlr r.rr. 

L19 P SS 

r,54 Piss 

1.70 PA53 

r.45 Prss/m ffi 

1.38 FtSs 

0.?0 F^gs/of s|M 

r,75 FS 

3.2t q,/d' Ptgs/ffi w 

z,E? q/d PAf4/|ff m 

r,16 P^$s 

1.{{ PrES 

1.37 Pr.$s 

FS 

PASS 

PS 

PS 

PH 

Pffi 

Pls 

PAS/oF SLffi 

PASS 

PAS 

PS/OF ruR 

PH 

1,30 Dn 

Ll[ DR 

1.66 Dn 

1,37 DR/RPqT ' 13

The three-point belts used were the stock LTD belts obtained 

through the dealer. These belts were single spool, 

vehicle sensitive/emergency locking retractor (VSR/ELR), 

three-point continuous loop systems with a cinch lock latch 

plate such that when the belt was on, webbing could move 

from the lap strap into the shoulder strap, but as designed, 

not the other way. These belts were equipped with a tension 

relief '-windowshade" device (TRD). 

The sled pin chosen for the Transportation Research Center 

of Ohio (TRC) HYGE sled provided a reasonable analog 

to the 30 and 35 mph barrier crash pulses for this vehicle 

yielding 32.5 and 37.5 delta-V's respectively for these two 

test conditions. 

Data reutrded.-Table I shows the data collected. In 

addition to the data shown, three 1000 frame-per-second 

cine cameras were stationed for front and side orthogonal 

views. The normal components of head and chest accelerations, 

along with chest deflections, were recorded on the 

Hybrid III which was calibrated immediately prior to the 

test series. Belt loads and neck forces were also recorded 

while femur loads were not. 

Specific test protocol.-These sled tests were conducted 

in two primary phases: With a head-on, 0 deg. orientation, 

and, with the sled buck rotated 38 deg. to the left. This sled 

rotation simulated far side oblique impacts with occupant 

PDOF angles to the inside of the car for both the normally 

oriented driver and the mirrored passenger. Thirty-eight 

degrees was chosen because this is approximately the angle 

at which the shoulder belt traverses across the car from the 

B-pillar D-ring to the latch plate. If the belt works properly, 

one might expect the shoulder strap to provide effective 

restraint of the upper torso in a direction approximating the 

belt loop orientation in the car. Another rationale for choosing 

the 38 deg. angle was that given practical resource 

limitations, this one sled angle is a reasonable simulation of 

impact orientation angles on the highway up to approximately 

twice the sled angle, or 76 degrees. 

For the 0 deg. frontal sled impacts, repeat tests were 

performed and data trends plotted to permit identification of 

outlier points beyond the predictable or explainable envelope 

of conditional relationships. Results for these tests 

confirmed that the test conditions had been well controlled, 

and that the Hybrid III dummy had performed in a repeatable 

and predictable manner. Reference 45 describes the 

specific details of the Hybrid III performance in these tests. 

ForTests I through I I and 18, slack in the passenger shoulder 

belt was gradually increased from -1.0 in. to 18.0 in. 

Tests I 2, I 3 and I 5 were of the driver at 30 mph with 0.0 and 

3.0 in. of slack. Test l4 was of the driver at 35 mph with -l.0 

in. of slack, while Tests l6 and I 7 were of the passenger at 

35 mph with 0.0 and 3.0 in, of slack, respectively. 

The three independent variables evaluated during the 

oblique tests were shoulder belt slack (defined as the extra 

webbing through the shoulder belt D-ring at the beginning 

of the test), the position of the belt with respect to the arm/ 

shoulder joint, and lap belt slack etrtablished by measuring 

between the center lower abdomen and a corresponding 

428 

point on the taut belt. Upon detailed analysis of the results, 

the effect of 0 to 3.0 in. of lap belt slack was not strongly 

significant in the type belt tested using an ATD. 

Frontal0 Degree Sled Tests 

Data analysis and results.-For the 0 deg. tests, the outcome 

variables most sensitive to the slack independent variable 

were head speed and displacement relative to the interior. 

Figure 9 shows the head excursion envelopes for 

passengerTests I through ll,16,lTand lS,whilefigure l0 

shows the head excursion envelope for the driver occupant 

at 30 and 35 mph, Tests l2 through 15. Figure I I shows head 

speeds relative to the interior versus time and displacement 

for Tests I and 9. Figures l2 through l8 show individual 

cine frames for selected 0 deg. frontal tests showing key 

kinematic times during each event such as maximum head 

velocity, impact head speed, and head maximum forward 

displacement. 

I'lote Heod Stortinq Poliltioni Hove Been Adlusted ior Cqmer0 Pordllor 

Figure 9. Head excurslon trslectories lor passenger te8t8. 

Noter Heod Stortinq Positions Hove Been Adjusteci for Cdmero Porollox 

Flgure 10. Head ercursion traiectorles for driver tssts. 

Chest results.-The shoulder belt strap forces (along 

with transfer forces from the head/neck when contact ocr 

curred) were the restraint forces on the chest during these 

tests. The present condition of -1.0 in. of slack in Test I was 

achieved by tightening and locking the belt at 40-50 lb 

preload. From Table l, this modest preload did not increase

0, ,^ 

L'] IU 

E 

a 

'11 

o . 

tu 

E + 

OJ 

T 

l4 

e 

Flgure 12a 

-C o_ 

H 

_r: 

o 

0l 

_ 1 0 

o 

U 

T 

5 

TEST *01 

30 mph,/0'Possenge. 

- 1.0 in Elqck 

Its I $09 

30 nph/O'Possengsr 

8.0 in sldak 

5 1 0 1 5 a 0 e 5 

Heod Excursion (in) 

l l l r l r 

0 1 0 e 0 3 0 4 0 5 0 6 0 7 0 8 0 

Heod Excursion (cn) 

Flgure 11. Felative hesd speeds versus hsad excurslons for 

-1 .0 in. slack (teat #1) and 8.0 in. stack (te8t #g). 

Flgure 12b 

Flgure.I2. .Test #l-pasarngel wlth -1.0 In. slack. Speed: 30 

mph; dlrectioni 0 dogr€€8. 

Figure 13a 

Floure 13b 

Flfiure..l3..Test+Z-psea€nger wlth 0.0 in. stack. Speed: i0 

mph; directioni 0 degr€e8. 

Flgure 14a 

Flgure 14b 

Flgure..14. .Test #O8-paslsnger wlth 5.3 ln. slack. Speed: 30 

mphi direction: 0 d6grcas. 

429

.-l 

Figure 15b 

Flgure 15. Test #09-pase€nger with 8.0 in. slack. Speed: 30 

mph; direction: 0 degrees, 

Figure 16a 

Flgure 16b 

Flgure 16. Test #l2drlver wlth 0.0 ln slack. Speed: 30 mph; 

dlrectlon: 0 degrees. 

430 

Flgure 17b 

Flgure 17. Test #1S-driver with 3.0 In. slack. Speedr 30 mph; 

dlrectlon:0 degreea. 

Flgure 18a 

Fioure 18b 

Fi[ure 18. Test #14+]lver with *1.0 in. slack. Speed: 35 mph; 

direction: 0 degrees.

chest deflections while it reduced the chest accelerations 

from 37.7 to 30.1 G's. The shoulder strap load increased 

from 1407 to 1504 pounds with this amount of preload. For 

belt slacks up to 3 inches, chest accelerations were directly 

proportional to slack while belt loads and chest deflections 

were generally inversely proportional to slack. For slack 

conditions greater than 5 inches, the effects on chest responses 

of head/neck impacts to the instrument panel began 

to dominate. 

The highest chest deflection for the 0 deg. test series was 

I.86 in. for the 35 mph driver with a moderately pretightened 

belt. The highest chest acceleration of 52.3 G 

resulted for the 35 mph passenger with 3.0 in. belt slack. For 

the 30 mph passenger tests, there were no chest accelerations 

or deflections approaching the corridor limits established 

for this ATD (less than 60 G-3 ms clip and 2.0 in. 

chest deflection). These results are most probably attributable 

to the combined factors of vehicle crash pulse, retractor 

spool-off, and webbing stretch properties having to some 

extent been tuned to the dynamic loading characteristics of 

the Hybrid III chest response corridors. This is also consistent 

to some extent with the findings of Reichert (32) discussed 

above. 

Effects on chest responses of the belt initially off the 

shoulder against the basic condition of the belt strap on the 

shoulder were as$es$ed in two frontal tests. For a slack of 5 

in. with the strap off the shoulder (Test l0), the belt loaded 

into the right lower abdomen below the rib cage, and across 

the lower left rib cage structure. This concentrated loading 

on the Hybrid III chest structure, with reduced restraint 

contribution from the shoulder, caused both chest accelerations 

and deflections to increase. 

With l8 in. slack and the belt strap initially off the shoulder 

(Test I I ), the chest accelerations were below those for 

the l8 in. on-shoulder slack condition (Test l8), while chest 

deflections were the same. Belt loads were recorded in Test 

I I but not Test 18, so a direct comparison is not possible. 

The belt loads for Test I I were somewhat below peak onshoulder 

values because of influences from the head/neck 

impacting the instrument panel. 

Head results"-Head (center of) speeds relative to the 

interior of the compartment ver$us time and versus head 

excursion were plotted for each of the eighteen 0 deg. frontal 

tests in which varying amounts of torso belt slack was 

set. These head kinematic plots have been published recently 

elsewhere (45) and the remaining results from this study 

will be summarized here. Figure I I shows an overplot of the 

head speed versus excursion data for the passenger occupants 

in Tests I and 9. The head speed versus time and head 

speed versus forward excursion plots were con$tructed for 

each frontal test by detailed kinematic analyses of the high 

speed cine films. Cross checks indicated that the data was 

consistent and the Hybrid III's performance was repeatable. 

The head speed increased uniformly and predictably as the 

amount of slack in the shoulder belt is increased. Predictable 

exceptions occurred in Tests l0 and I I where the belt 

strap started off the shoulder and those results will be dis- 

cussed in more detail below. The kinematics plots of head 

motions of Figure 9 and l0 have been adjusted for parallax 

so that these plots look like the cine film frames and the 

time-based event$ match. The basic trajectory data were not 

corrected for parallax and thus the derivative results of 

relative head speeds and excursions inside the compartment 

are somewhat understated, but uniformly so. 

Results from this sled test program show the peak head 

velocities and excursions within the compartment are sensitive 

and repeatable outcome variables when plotted against 

the control variable of shoulder belt slack. In this series of 

tests, and indeed in any similar belt restraint tests conducted 

in a realistic compartment interior, head response variables, 

such as accelerations and the derived Head Injury Criteria 

(HIC), are strongly influenced by the energy absorption 

characteristics of what surface the head impacts and the 

relative orientation during that impact. 

These considerations were explored to some extent by 

TarriEre in 1982 (46), when he quantified the component 

forces acting on a three-point belted driver's head impacting 

a vehicle's steering wheel. TarriEre's argument, however, 

that the HIC limits can be safely increased to 1500 for belted 

drivers, gives insufficient attention to the fact that for this 

occupant, many impacts of the steering system with the 

head occur into the delicate facial area. The relatively weak 

facial bones are easily fractured causing serious cosmetic 

injuries, and these structures can also be driven rearward 

into the brain area. TarriBre's other argument, that no head 

injury is expected to occur if there is no head contact, is 

somewhat confirmed by the presently reported 30 mph 0 

deg. frontal results. When no impact occurred, the HIC36 

was below 1000. In the 35 mph 0 deg. tests, however, the 

HIC36 was over 1000 even for the no contact situation (Test 

l6). 

A 1986 study by Grdsch, et al. (47) confirms these concems 

regarding the insensitivity of the HIC to predict facial 

injuries and develops a Facial lnjury Criteria (FIC) as a 

basis for evaluating injuries to belted occupants. The results 

from his analysis indicate when comparing different energy 

absorbing steering wheel treatments, including air bags, 

some will exhibit higher HICs, but lower FICs. These authors 

$tate, with considerable substantiation, that in such 

instances, the FIC should be the controlling factor. 

These concepts can be related to the results of the present 

study by reviewing plots of maximum head velocity against 

the head forward excursion for the -1.0 (Test I ) and the 8.0 

(Test 9) in slack conditions. The envelope of head kinematics 

traced out for these conditions show the vulnerability of 

the head to impact injury is related not only to the head 

velocity (kinetic energy) at any given point, but is also 

related to how far forward in the compartment (how close to 

interior surfaces) the head is positioned when going that 

speed. For Test l, with the moderately pretightened shoulder 

belt, the maximum relative head speed was 17.4 mph 

which occurred after the center ofthe head had traveled l2 

inches-not near to striking any interior surfaces or objects. 

431

For Test 9, with 8 inches of slack in the shoulder belt, the 

maximum head speed was 29.4 mph which occurred after 

the center of the head had traveled 22 inches. This maximum 

head speed point was just one inch short of the 23 inch 

mark where the head impacted the instrument panel at a 

speed of 27.9 mph. 

The present approach to analyzing the effect on head 

re$ponse$ of increasing belt slack uses the rationale that the 

relationship of higher relative head speed and greater forward 

excursion inside the compartment is a good predictor 

ofthe potential for serious head and facial injuries. Substantiation 

for such an approach is given in terms of the kinetic 

energy dissipated in stopping the head from various relative 

interior velocities. For instance, the kinetic energy of the 

head (using Grdsch's effective mass of 16 pounds) at a 

relative head speed of 17.4 mph (which is the head speed for 

the best performing belt test of this series) is 162 ft-lb, 

whereas the kinetic energy for the head with 8 in. of slack in 

the belt and a head speed of 29.4 mph is 462 ft-lb, a factor of 

2.85 increase. As a comparison, the kinetic energy asso- 

l6 

)to 

x 

E 

o 

E.- 

>rc 

E 

U 

5ro 

.E 

E 

I 

u70 

u 

-60 

0 

r 

30 mph,zO'Possenger 

I 

I 

o/ 

ô /o 

o / 

I 

5 1 0 

ShoutdeF Bett Stock (h,) 

l0 a0 30 

Shoul.der Bett Stock (cm,) 

-"-* 

Flgure 19. Maxlmum h6ad velocltle$ veraus ehoulder belt 

elack. 

n 

o 

Totat Herd TFavet 

ForraFd Axls Travet 

I 

j 

I 

T 

/ 

{ 

J- 

,t 

)6 

ry 

IFtr 

5 t 0 

Bett Stlck (hches) 

-10 0 r0 20 30 40 50 

Shoulden Bett Slqck (cm,) 

Flgure 20. llsxlmum head cxcurslona vsrsut ehoulder belt 

slack. 

432 

T 

t 

E - - 

6 

)a: 

E 

I 

f 

.E 

t 

tu 

I 

U 

30 

9as 

F 

b 

20 

--E 

ciated with the maximum relative head speed of 34.2 mph in 

these te$ts, is 625 ft-lb, a factor of 3.86 increase. 

Figure l9 summarizes the maximum relative head-tointerior 

velocity versus shoulder belt slack for the 30 mph 

frontal pa$senger tests in which the strap was on the shoulder. 

Figure 20 summarizes the maximum head excursion 

inside the compartment versus shoulder belt slack for these 

same conditions. For ten of the l8 frontal 0 deg. sled tests in 

this program, the test conditions were closely controlled, 

and the results for these ten tests can be directly compared in 

Figures 19 and 20. A perceptible break or knee in these 

curves occurs between five and eight inches slack, reflecting 

the onset of head impacts with the instrument panel for 

slack setting over 5 inches. In future testing it would be 

interesting to investigate possible influence of the Hybrid 

III head/neck/torso dynamics on these findings. 

To consider the effect of the torso strap being off the 

shoulder before impact, Figures 2l and 22 werc prepared 

which include Tests ll and 18. A simple normalization 

proce$$ was used by plotting the actual head speed and 

displacements for these off-shoulder tests, and then considering 

what additional apparent slacks might be introduced 

to have the off-shoulder data points fall on the on-shoulder 

l6 

0 

ir. 

= 

9'" 

t 

f, 

5ro 

Should€r Bclt $tqck (h,) 

| | | _.. _l .__-___ _._- 

-lt 0 l0 e0 30 40 

Shoutder Belt S(ack (cF) 

Flgure ?1. Msrlmum head vslocitles versus shoulder belt 

slack wlth ofl-shoulder condltions included. 

c70 

' 6 0 

t 

s 

t 

F30 

= 

6 

>ei 

E 

! a 

E 

30 

te5 

Note, The Bett off-Shoutdrr Increases the Effectlve StaEk Byl 

- Test 10, Fron 5 h, to / h = e h 

- feEt U, Fron 18 h. to 24 h, = 6 h, 

30 nph 0'Fdssenoer 

I 

/ 

.v 

I 

fr 

lo 

F-dI 

/ 

5 1 0 1 5 

ShoeLder Bett St0ck (h,) 

q-- 

---F 

. 1 0 0 t 0 d 3 0 4 0 5 0 6 0 7 0 

ShoqldfF Bett $tock kE,) 

Flgure 22. Msxlmum hesd excureions vsrsus ehouldsr belt 

slack with ofl-shoulder condltions lncluded.

curves. The results of this analysis indicate$ that with the 

belt off the shoulder, a slack of 5 inches looks kinematically 

like a slack of 7 inches. With l8 inches of slack. the belt off 

the shoulder looks kinematically like a slack of 24 inches. 

The 5 to 7 inch slack analogy for the on and off-shoulder 

conditions is reasonable to describe this effect quantitatively. 

However, at the l8 inch slack setting the head 

impacts to the instrument panel were so heavy that this kind 

of comparison is presented here as descriptive only. It is 

interesting to compare the responses for the off-shoulder 

conditions in Figure 2l and 22 with those in table I . For the 

5 inch off-shoulder condition, the chest acceleration, shoulder 

belt load, and chest deflection was the highest for any of 

the 30 mph frontal 0 deg. passenger tests, whereas the head 

respon$eri were not the highest in this series. 

Figure 23 compares the peak head speed versus shoulder 

belt slack for the previously discussed 30 mph frontal 0 deg. 

pa$$enger tests with these same variables for the 30 mph 

frontal 0 deg. driver tests and the 35 mph frontal 0 deg. 

driver/passenger tests. This last comparison can be made 

because the peak head speed occurs before impact. Even so, 

when comparing the 30 mph driver with the 30 mph passenger, 

the driver peak head velocities are slightly and consistently 

higher. This effect has not as yet been totally explained 

but it is probably due in part to the positioning of the 

hands and arms holding the steering wheel. 

dt+ 

u 

t u 12 

o ' - 

oj 

-b 

o 

0r .^ 

I T U 

E 

f, 

X 

E ^ 6 

_c 

r 

1i 

H.= 

tu 

d 

0J 

I 

Ea0 

f 

T 

d 

E 

3C Eph/o' + 

Drtver 

/ 

'A 

6 

0 5 

Shoutder Bett Stock (inches) 

0 1 0 

Shoutder Bett SIock (cm,,r 

Figure ?3. Itlgxlmum head veloclty veraua shoulder belt alack 

for trontal drlvers and passengers 

at 30 and 35 mph. 

Shoulder belt slack in the range of0 to 5 inches is found 

most frequently in American cars because the automobile 

and belt manufacturers designed the U.S. type windowshade 

tension relief devices to introduce slack amounts 

in this range. These devices will however, introduce any 

amount of slack up to the total amount of webbing available 

under certain circumstances. The results of these tests, as 

presented in figure 23, show that the effects ofintroducing 5 

inches of slack into a shoulder belt is approximately equiva- 

/o 

' 30 Aph/o' 

ParrtngaF 

lent to raising the barrier equivalent impact severity from 30 

to 35 mph, a severe result when kinetic energies are considered. 

At a given slack, going from 30 to 35 mph B.E,V. 

raises the maximum head speed about 6.5 mph. 

Figure 24 relates both HIC and HIC36 to the shoulder belt 

slack for Tests I through 9 and I 8, all of the pa$senger tests 

at 30 mph. The HICs rise predictably for slacks of - I .0 to 8 

in., but then fall significantly from 8 to l8 in. More data 

points would be desirable in this range to fully explain this 

drop at the higher slack levels, but in the absence of such 

additional data points, the following observations can be 

made here. First, the top and brow of the instrument panel in 

the late model Ford LTD is constructed of a plastic substructure 

which is a reasonably good energy absorber if impacted 

in an advantageous direction. With the l8 in. of slack in 

these tests the Hybrid III's head hit more of the flat upper 

yielding surface of this instrument panel, and in two tests 

yielded HIC's between 700 and 800. 

Leeendr 

O Btss Possenger HIC 

tr Blss Passenoea HIC36 

# Esser HIC 

P Roneo HIC 

O Blss 3-poht Ilnrver HIC 

1 6 

Shoutder 

E l0 le l,t 16 l8 

Bett Stock (lnches) 

t0 a0 30 r0 

Shoutder Bett Stqck (cft.) 

Figure 24. Comparlsons of HIC vs, Shoulder Belt Slack trom 

Vsrlous Research Programr. 

Figure 24 also compares the HIC values from this study 

with those previously reported by Esser and Romeo (43,48) 

of three-point belt sled tests performed by TRC in Ford 

sedan sled bucks with the $ame Ford belts. For the Esser 

passenger tests at 30 mph, the HIC values match the present 

data closely up to 2 inches. He then has no data to report 

until a forth test at 163/+ inches slack. In his tests, the instrument 

panel was constructed of a sheet metal substruature 

which probably explains the large difference in the HIC 

values for his I 63/+ in. slack test and the results for the I 8 in 

slack tests reported here. These results again demonstrate 

the significance of not only of the head speed at impact, but 

433

Test fl2, Driver: J0 mph 

Test #15, Driver 

Projection of Test #1 PoEs6nger Results to q 30 mph, -1.0 in slock Teet for Driv€r 

Figure 25a-25c: Head Tralectorles lor 0 deg. 30 mph Drlver Wlth 

-i-o, o.o and 3.0 inches Slack. 

the energy absorbing characteristics of the surface 

impacted. 

Romeo conducted a 30 mph three-point belted driver sled 

te$t with a Hybrid II in the same body buck used for this 

study and his test produced a HIC of 865, essentially the 

same a$ obtained in this study. In his study, Romeo also 

tested a lap belted only driver with the same design of 

steering wheel and column used in this study and recorded a 

HIC of 2584, considerably more than the peaks recorded in 

either the present tests, or in Esser's tests. 

Figures 25a and 25b compare the head kinematics for the 

driver occupant with 0.0 and 3.0 inches of slack for the 30 

434 

mph test condition. As the results discussed above showed, 

the steering wheel is well within the head swing trajectory 

of the driver even with zero slack. However, the location of 

the dummy's face/head impacts into the steering wheel 

changes with the amount of slack as does the impact speed. 

The steering wheel used in this test series would be classified 

by Glyons (49) as a non-energy absorbing type in which 

the plastic molding is primarily for cosmetic purposes. Although 

the highest HICs for the driver tests at 30 mph was 

1165, l2o/o over the 1000 limit FMVSS 208 uses to predict 

brain injury, analyses similar to Grdsch (47) and Gloyns 

(49) would most probably predict serious facial injuries for 

all the driver tests at 30 mph with slacks of zero or more. 

The results of the present study have been structured in 

such a way as to permit interpolation of the results to related 

test conditions which were not actually run on the sled. For 

instance, figure 25c shows the predicted results of a driver 

test at 30 mph 0 deg. frontal with a pretightened belt. 

Whereas the head impact speeds for the zero and 3.0 inch 

slack conditions were in the 20 mph range, pretightening the 

belt to between 40 and 50 pounds drops the head impact 

speed below 8 mph. This again emphasizes the usefulness of 

the head speed envelope ("how far-how fast") analysis 

developed in this study. 

The influences of the force-deflection characteristics of 

the various surfaces impacted by the head during the presently 

reported tests have been considered, but the mean'r to 

rigorously control these impact conditions were not immediately 

at hand. It was found during these tests that the initial 

head impact point for both the driver and passenger test$ 

could be repeated predictably. It was, however, more difficult 

to control the mutual head-to-surface force-deflection 

characteristics of these impacts which determine variations 

in head acceleration, and thus HIC levels (and with extended 

capability, facial injury indices). The presently reported 

results show that this may be possible using the Hybrid III 

dummy and with considerably more resources than were 

expended here-for instance purchasing new steering columnrl 

and wheels for each test and instrumenting the faces 

per Grrisch. Short of this, using the maximum relative head 

velocities as a predictor of head/face injury potential is an 

intermediate, but useful tool in studying the effects of shoulder 

belt slack. 

Oblique $led Test Results 

The 38 deg. orientation of the sled body buck for Tests I 9 

through 28 was approximately the angle at which the shoulder 

harness defines a plane cutting across the car. If the strap 

stays on the rib cage structure during oblique impacts in this 

range of severity, one would expect that the belt should 

provide effective restraint. 

Figures 26a through 26e show two high speed movie 

views of the kinematics of the Hybrid III in the 30 mph, 38 

deg. oblique sled test in which the belt was pretightened to 

40*50 lb (approximately-1.0 in. slack). This pretightening 

pocketed the belt into the dummy's clothing and skin, and as 

the impact progressed, it became obvious that this pre-

pocketing contributed to the belt effectively staying on the 

shoulder and chest structure (clearly across the stemum) 

throughout the test. The frame in this figure shows the 

Hybrid III at its maximum forward excursion (before rebound) 

to the front and right. 

Figure 26a 

Figure 26b 

Figure 26c 

Flgure 26d 

Flgure 26e 

Flgures 26a-26e: Test #1g-Passeng€r wlth -1.0 ln. slack. 

Sp-eed: 30 mph; obllque 38 degreee. 

Figures 27a through 27e show two high speed movie 

views of the kinematics of the Hybrid III for these same test 

conditions but with the belt slack set at 0 in. In this condition, 

when the belt is initially positioned on the shoulder, it 

seats or pockets itself into the occupant's clothing and flesh 

early in the impact sequence and the strap stays on the 

shoulder/chest structure-well up on the sternumthroughout 

the impact. 

The effect of slack on Hybrid III responses in frontal sled 

tests has been discussed above and is a good starting point to 

discuss the effects of slack in the oblique sled tests reported 

here. In these tests, added to the forward motions are the 

lateral motions caused by the 38 deg. sled orientation. When 

the torso strap is off the shoulder, 1.3 to 3.0 inches of slack is 

enough to permit the belt to hang low on the chest/sternum 

area pre-test. During the impact the belt changes position 

coming off the right rib cage and is barely on the left rib cage 

structure, if at all. This amount of slack also allows the belt 

to hang over the corner of the shoulder, but in the Hybrid III, 

that is still within the range of the notch in the shoulderjoint 

flesh at that point. As the impact progre.ese$ the belt catches 

in the shoulder notch, but this is still not enough to keep the 

belt from assuming a dangerous biomechanical position 

across the lower thorax and abdomen during loading. 

43s

Flgure 27a 

Figure 27b 

Figure 27c 

Figure 27d 

Floure 27e 

Fllurea 27a-27e: TsEt #26-pa$sanger with 0.0 in. slack. 

Speed: 30 mph; oblique 3E degrees. 

With 3 inches of slack in the belt, but initially positioned 

on the shoulder, the belt strap initially pocketed itself into 

the occupant's clothing and chest flesh at the beginning of 

the impact sequence. As the sequence progressed, the shoulder 

strap stayed on the rib cage long enough to effectively 

restrain the forward and lateral motions of the thorax without 

the belt slipping into the abdominal area. Similar results 

were obtained in Tests 20 and 22 where the belt was initially 

on the shoulder but with 2.5 and 2.3 inches of slack, respectively. 

Again the belt stayed on the rib cage long enough to 

arrest the forward and lateral motions without the strap 

slipping off into the abdominal area. In both these tests the 

shoulder belt strap passed over the notch in the shoulder 

flesh of the Hybrid III without snagging, as frequently happened 

in terrts where the belt was initially off the shoulder. 

Figures 28a through 28g show that with as little as l.3 

inches of slack, the belt can be positioned off the shoulder, 

permitting the occupant's rib cage/shoulder structure to 

come out of the belt early and cause severe abdominal 

loading. The initial position of the shoulder strap is near the 

top of the shoulder flesh notch of the Hybrid III. During this

Flgure 28d 

Flgure 28f 

Flgure 289 

Figure 2Bc 

Fllures 28a-289: Test #zFpas$enger with 1.3 in. slack. $peed: 30 mph; oblique 38 degrcee. Belt Inltlally poaltloned otl-ahoulder. 

437

impact the belt partially snags in this notch, but despite this, 

the dummy still twisted out of the belt as the impact progressed 

and the strap loaded heavily into the lower chest/ 

abdomen area. Again, this test emphasizes the importance 

of the initial position of the shoulder strap on an occupant 

just before an impact. 

On this point, it is interesting to draw the reader's attention 

back to one of the significant claims in the original 

Bohlin patent for the three-point belt (1,2,3). The threepoint 

belt, when positioned on the body properly, should 

apply restraint forces to the body in a "physiologically 

correct" manner to engage the strong skeletal structures of 

the body. Clearly, the results of these sled tests demonstrate 

non-compliance with this fundamental principle when the 

belt is positioned off the shoulder or even on the corner of 

the shoulder before the impact. 

The initial position of the belt at the beginning of the 

impact is a strong determinant of where the belt will physiologically 

load the anatomy during the impact. As little as 

1.3 inches of slack allows the belt to drop 6 to 8 inches 

vertically on the chest/sternums, and this initial orientation 

seems to be a dominant factor in the configuration of the belt 

loading during impact. In this case, again, loading dangerously 

into the lower chest/abdominal area. 

A sled test was conducted in which the occupant was 

leaned forward with the belt strap on the shoulder/thorax 

with 3 inches of webbing unwound from the retractor (Test 

?8). In this test the occupant's forward and lateral motions 

were effectively restrained while the strap stayed on the rib 

cage shoulder structure. Thus, even with the occupant leaning 

forward to take advantage of the freedom of movement 

provided by the ELR, and with the shoulder strap snug and 

positioned properly, the slack effects are minimal if the 

occupant does not strike anything. The findings of this 

study, that where the belt starts out on the anatomy is critical 

to safety belt performance, was confirmed again in this test. 

There is no analog in the human for a notch in the flesh at 

the shoulder. It is doubtful that the human shoulder would 

act like the Hybrid III shoulder in impacts such as these 

tests, catching the belt during oblique tests. As such, it is 

probable that the belt would end up even lower on the 

human or cadaver than on the Hybrid III as photographed 

here. Accordingly, oblique occupant retention tests on 

three-point belts which have a tendency to introduce slack 

during use should be re$ted not only with the Hybrid III, but 

with surrogates in which the shoulder simulates the human 

both in the pre-test and test conditions. Altematively, an 

improved Hybrid III shoulder should be produced so that 

the range of biofidelity of this ATD can be extended to 

frontal oblique impacts. 

These results strongly suggest that public information 

campaigns are needed to better educate belt users about the 

biomechanical implications both of wearing shoulder belts 

slack and of not having them positioned on the clavicle area 

before any potential collision. 

438 

An Assessment of the Safety Effects of 

Slack on the Safety Performance of the 

Three-Point Belt 

Summary of sled test findings,-The results of this sled 

test program lead to an overview of the safety effects on 

three-point safety belt performance of the TRD 

windowshade which can be summarized as follows: 

r Biomechanically, where the shoulder strap is 

initially positioned on a person before an impact 

has just as significant an effect on crash 

perforrnance as does the amount of slack. 

r For the three-point belt tested, the most sensitive 

outcome variables were the head speeds and 

excursions inside the compartment versus 

shoulder belt slack. 

r Modest levels of pretightening, 40-50 pounds, 

significantly reduce first the probability, and 

second the severity oflhe occupant's head striking 

the interior; and, also significantly reduces the 

likelihood that the occupant's torso will come out 

of the belt in oblique impacts. 

r Forthe passengeroccupant in the three-point belt/ 

interior system tested, there was only a slight 

degradation of occupant injury measures with one 

inch of slack and the belt on the shoulder as 

compared to the zero slack setting. However, 

r As little as 1.3 inches of slack permits the belt to 

lie over the corner of the shoulder resulting in 

poor belt placement and a dangerous 

configuration of belt loading during the impact. 

r Thus, for the passenger occupant, there is a 

narrow range of slacks, one inch or less, which 

might be permitted for comfort reasons under 

certain circumstances. 

r For driver occupant$, the steering wheel is in the 

head swing zone for all slacks greater than zero. 

The speed at which the head impacts the steering 

wheel changes modestly from 19.7 to ?2.3 mph 

when the slack is increased from 0 to 3.0 inches 

but as the slack is increased, the face is more likely 

to be directly impacted versus the forehead/skull. 

With modest pretightening of the shoulder belt 

4G-50lb, the head impact speed drops to 7.8 mph 

and the head contact point moves even further 

away from the face. 

r The driver obtains all the same benefits of modest 

pretightening ofthe belt as the passenger, plus the 

benefit that the severity of his head impact with 

the steering wheel significantly reduced. 

r These results indicate clearly the immediate need 

for either air bags or energy absorbing steering 

wheel treatments for drivers of all three-point belt 

equipped cars to protect the face during impacts as 

many of the works referenced herein have pointed 

out before.

The Emergency 

Locking Retractor 

Reconsidered 

ELRts as crash sensors 

Over the past twenty years intense efforts have gone into 

the analysis of air bag crash sensors. Such efforts have 

routinely included many non-standard tests to determine the 

sensitivity of the sensor to initiate reliably in a wide variety 

of crash modes. At least in the public arena there has been no 

reponed comparably intense analysis designed to determine 

the limits of reliable operation and failure modes of the 

emergency locking retractor. As mentioned above there has 

been some discussion (4,6,23,24) about the possibilities for 

the ELR to not lock in certain circum$tances. but when these 

types of situations have reached the public discussion stage 

(41,42), the resolution involved rhe rarionale that possible 

failure to lock or delayed lock would only occur in a few 

isolated instances. Such a rationale has neverbeen accepted 

in the development and evaluation of air bag crash sensors. 

Based even on the limited public record, it seems 

warranted to initiate programs to evaluate the reliability of 

various ELR designs to lock reliably. The first area to 

consider is the rollover mode where there are multiple 

impacts from many different directions on the vehicle 

chassis, where the occupant motion is probably not in phase 

with these vehicle dynamics, and where the ELR 

requirement to maintain its locked position-that the 

occupant maintain constant tension on the webbing-is 

likely to be violated a number of times during the impact. A 

dynamic version of the FMVSS 301 rollover test in which a 

balanced body buck with a three-point belted occupant is 

rotated at a speed of, for instance, one revolution per second 

for a sustained period of time would be a reasonable staning 

point. Funher investigative steps should include preimpact 

shake dynamics to simulate rough ground or partial fall 

conditions just prior to impact. 

Design Concepts Growing Out of This 

Study 

Rethinking the Tension Relief Device 

r The present design of the TRD is unsatisfactory 

because, even within its prescribed operation 

envelope, it introduces slack which can be 

hazardous in oblique impacts, and it routinely 

introduces even more than the recommended 

amount. The root of the problem isthat the present 

TRD is a negative option devirc as shown in 

Figure 29a, and gives the wearer the above 

described amounts of slack whether or not he 

wants it, knows what to do with it. or knows the 

biomechanical implications of it. There are better 

alternatives as discussed below. 

r Although the European approach to safety belt 

performance effectively precludes the TRD 

I 

E 

J 

o 

l 

.E 

s 

! 

l 

I 

o 

I 

E 

l 

o 

l 

E 

t 

! 

) 

o 

T 

o 

J 

O 

E 

tr 

Posrtrve actron, e.S" Ou5hing button, 

setE on€ the controlted slock, bst 

only Ffoh ofrgindl bett-on posilon. 

AAy for{o/d notton beyond 

E In. rf5ati bstt to Enug, 

TRn connot be ret *ben leq^hg 

lo.wq.d. fhere ls oniy one TRD 

setthg per belt *eo.Lhg. 

B 

---f-- 

l 

I 

; 

6 

j 

t : l 

t t 

Figure$ 29s-29b: Operatlonal comparison of present TRD 

wlndowshada with controlled elack zone device.- 

device, $ome type of slack eliminator with a 

strictly controlled limit to the amount of slack 

possible might be indicated for obese persons and 

other "special conditions." However, based on 

the European experience with comfort designs 

and wearing rates, and the industry'$ own 

assessment of the impact of the TRD on usage 

rates, it appears unnecessary to expose all users to 

the downside risks associated wilh the pre$ent 

operational characteristics of these devices. Thus, 

the windowshade should be sold only as an option 

for special cases and only in a highly modified 

form. 

An all new concept is shown in Figure 29b. 

Comparing with the present negative option 

device in Figure 29a, the new concept is a positive 

option device that will only introduce a strictly 

controlled amount of slack (one inch or less as 

indicated by the results of this study) and only sets 

one "comfort zone" per belt application. Such a 

device would appropriately be called a 

"Controlled 

Tension Zone Device (CTZ)." This 

concept eliminates the grossly excessive slack 

that the TRD can introduce, and restores the best 

functional attributes of the ELR, that of keeping 

the belt strap near the che$t. One concept is to 

have a positive action "push button" or some 

other actuator available to the occupant to only 

439

introduce the controlled zone slack ifthe occupant 

actually feels enough discomfort to positively 

take action. Otherwise the ELR works as normal. 

Another version of the concept is that the forward 

motion of the shoulder, much like the present 

TRD, would set the one time limited slack into the 

CTZ, but if there was any forward motion at all 

beyond the stop, the device would reset the 

webbing to snug. 

Pretighteners vs. Pretensioners vs, 

Positive Locking Devices-a Proposed 

Definition 

The results of this study confirmed the many benefits 

reported elsewhere ofpretightening a belt in the early stages 

of a crash event. Although it would at first seem to be a 

solution to the TRD slack problem to install such 

pretensioners in conjunction with the TRD, special 

considerations are in order. Pretensioners (ala Daimler- 

Benz) are not indicated for use with TRD belts because the 

relatively high pretensioning force (300-400 lbs) could 

snap the belt tight from a slack position creating a dangerous 

cabling action. An alternative to the high level pretensioners 

is one suggested by this study, that of a pretightener capable 

of modest tension in the 4f50 lb range. 

When considering these more modest levels of 

pretightening many new possibilities open up for improving 

the safety performance and reliability of typical three-point 

belt designs. This modest pretightening should accomplish 

the same positive locking task discussed above (in which 

damping was used) to prevent intermittent locking and 

unlocking of the retractor spool. With the pretightening 

forces low, pre-impact anticipatory initiation can be 

considered in which the belts would tighten and positively 

lock during emergency braking, tripping actions, rolling 

motions, etc. Because the pretightening forces are low, and 

thus the energy requirements are low, the units can be 

designed to be easily resetable by the user in either the 

preimpact, or impact initiation scenario. The initiation 

sensing mechanism of the pretightener can take consider* 

able advantage from the extensive pool of analysis 

techniques developed for air bag crash sensing and other 

shock type sensing tasks. 

In short, an easily resetable pretightener/positive locking 

device can be triggered which would: 

r Eliminate the slack from tension eliminators and 

other causes. In the process of eliminating this 

slack, the easily resettable pretightener will 

position the torso strap properly on the shoulder 

which was shown in this study to be critical to the 

safety performance of the three-point belt' 

r Positively lock (damp) the ELR retractor and 

prevent unlocking and reel-out during multiple 

impacts and rollovers. 

r A continually renewable energy source would 

recharge the pretightening device while the 

vehicle is in normal use, examples: 

Normal occupant movements of putting on and 

wearing the belt could store energy in springs for 

the pretightener function. 

Electrical energy from the vehicle could be used 

at low rates to recharge the pretightener energy 

source. 

r Alternatively, for higher set crash levels of 

initiation, but yet lower than the damage point of 

the retractor hardware, easily serviced replaceable 

energy sources are a possibilitY. 

r The attractiveness of these concepts are that they 

can be incorporated in an incremental fashion and 

each one will improve the performance of the 

present TRD belt system. A coordinated program 

to incorporate these measures as a system will, of 

course, lead to the optimal pay-off. 

Conclusion 

The present design of the TRD windowshade is 

hazardous in a number of ways and urgently needs 

rethinking in order to conform to the original safety goals of 

the three-point belt patent. To the extent that tension 

elimination devices will continue to be sold to the general 

public, major redesigns are necessary to meet these goals. 

Some concepts are presented here, based on considerable 

first-hand analysis experience, which would go a long way 

towards meeting these goals. Additionally, the types of 

concepts, designs and devices which will improve the 

tension elimination/reduction devices will, in most 

circumstances, greatly improve the performance and 

reliability of other three-point belt hardware components 

such as the emergency locking retractor. 

References 

(l) Sverige Patent227 568, "Siikerbetssele 

vid fordon," 

Uppfinnare; N.I. Bohlin, AB Volvo, G0tenborg, Patenttid 

Frfln Den 29 Augusti 1958. 

'(2) "Safety 

United States Patent 3,043,625, Belt," Nils 

Ivar Bohlin, Gdteborg, Sweden, assignor to Aktiebolaget 

Volvo, Gciteborg, Sweden, filed August 17, 1959, Serial No' 

834,258. 

(3) German Patent No. 1101987, (Award presented as 

Most Important Patent No. 8) "Safety Belt for Vehicles," 

Nils Ivar Bohlin, Giiteborg, Sweden, patented from August 

?4, 1959 on. 

(4) Bohlin, N.1., "Refinements of Restraint System 

Design-A Primary Contribution to Seat Belt Effectiveness 

in Sweden," Proceedings, International Symposium on 

Occupant Restraint, American Association for Automotive 

Medicine, Toronto, Ontario, Canada, June l-3, 198 I. 

(5) Johannessen, H. George, "Historical Perspective on 

Seat Belt Restraint Systems," SAE Paper 840392, 

International Congress & Exposition, Detroit, Michigan, 

February 27-March 2, 1984, pp.217-225.

(6) MacKay, G.M., P.F. Gloyns, H.R.M. Hayes and D.K. 

Criffiths, "Europcan Vehicle Safety Standards and Their 

Effectiveness," University of Birmingham, U.K., Proceedings, 

4th Intemational Congress on Automotive Safety, 

NHTSA, Washington, D.C., July, 1975, pp. 431454. 

(7) Bohlin, N.L, "Srudies of Three-Poinr Restrainr 

Hamess Systems in Full Scale Barrier and Sled Runs," AB 

Volvo, Giiteborg, Sweden, 8th Stapp Car Crash and Field 

Demonstration Conference, Wayne State University, 

Detroit, Michigan, October, 2l-2?, 196/.. 

(8) Federal Register, Vol. 44, No. 251, Monday, 

December 31, 1979,49 CFR Part 571, (Docket No.7zl-14: 

Notice l7), "Federal Motor Vehicle Safety Standards: 

Improvement of Seat Belt Assemblies." 

(9) Ford Motor Company, Reply to DocketT4-14; Notice 

17, NPRM, Letter to Joan Claybrook, Adminisrraror, 

NHTSA, April I, 1980. 

(10) Motor Vehicle Manufacturers Associarion, Reply to 

Docket 74-14, NPRM, Letter ro Joan Claybrook, 

Administrator, NHTSA, April l, 1980. 

(ll) Chrysler Corporation, Reply ro Docket 74-14, 

NPRM, Letter to Joan Claybrook, Administrator, NHTSA, 

March 28. 1980. 

(12) General Motors Corporarion, Reply to Docket 74- 

14, NPRM, Letter to Joan Claybrook, Administrator, 

NHTSA, April l, 1980. 

(13) Danner, M., K. Langwieder, and Th. Hummel, 

"The 

Effect of Restraint Systems and Possibilities of Furure 

Improvements Derived from Real-Life Accidents," SAE 

Paper 840394, Society of Auromotive Engineers, 

Warrendale, PA, 1984. 

(14) Kendall, D.L., M. Fowkes, I. Gazeley, and C.M. 

Haslegrave, "Comfort 

and Convenience of Safety Belts in 

Everyday Use," Report No. l98l/2, The Moror Indusrry 

Research Association, Warwickshire, England. 

(15) Personal Comrnunicarion with N.I. Bohlin. 

(16) Federal Register, Vol. 41, No. 243, Thursday, 

December 16, 1976,49 CFR Part 571, (Docket No. 7zt-14: 

Notice 7), "Federal Motor Vehicle Safety Standards." 

(17) Moffan, Charles, A., Edward A. Moffatt. and Ted R. 

Weiman, 

"Diagnosis of Seat Belt Usage in Accidents," 

SAE Paper 840396, Warrendale, PA 15096. 

(18) Bedard, P., "Slack Thinking in Detroit," Car and 

Driver, Vol.32, No. 12, June 1987, pp. 190-191. 

(19) Henderson, C., "Comfort and Convenience 

Improvements to Increase Safety Bett Utilization," SAE 

Paper 770187, Internfltional Automotive Engineering 

Congress and Exposition, Detroit, Michigan, February 28- 

March 4. 1977. 

(20) Grimm, A.C., "International 

Restraint Use Laws." 

University of Michigan, UMTRI, Research Review, Vol. 

18, No.4, Jan.-Feb. 1988. 

(21) Mitzkus, J.E. and H. Eyrainer, 

--Three-Poinr 

Belr 

Improvements for Increased Occupant Protection," SAE 

Paper 840395, Society of Automotive Engineers, 

Warrendale. PA. 1984. 

(22) Grtisch, L., H. Kaiser, and W. Schmid, 

"Mathematical 

Movement and Load Simulation for 

Persons Involved in an Automobile Accident," SAE Paper 

871 109, Governmentflndustry Meeting and Exposition, 

Washington, DC, May l8-21, 1987. 

(23) United States Parent 3,825,205, 

"Moror 

Vehicle 

Safety Devices," T. Takada, Takata Kojyo Co. Ltd., filed 

June 8. 1972. 

(24) United States Patent 3,87l,47O, 

"Safety Belt 

Tensioning Device," Schwanz et al., Volkswagenwerk 

Aktiengesellschafr, fi led October 21, 1972. 

(25) United States Patent 3,871,683, 

"Vehicle 

Safety Belt 

Tightener," S. Otani, Nissan Motor Company, Ltd., filed 

January 23, 1974. 

(26) United States Patent 3,929,2O3, 

"Mounting 

Device 

for Shoulder Harness Type Safety Belt," Y. Nagazumi, 

Nissan Motor Company, Ltd., filed January 17, 1974. 

(27) Ciccone, Michael A. and JoAnn K. Wells, 

"Improper 

Shoulder Belt Use by Maryland Drivers," 

Insurance Institute for Highway Safety, Washington, DC, 

June 1987. 

(28) Emshwiller, J. R., "Seat-Belt 

Slack: Comfort Device 

in U.S. Car Raises Safety Concern," The Wall Street 

Journal, July 31, 1987. 

(29) Aldman, Bertil and Arne Asberg, "Impact 

Amplification in European Compacts," Laboratory for 

Traffic Medicine, Swedish Council on Road Safety 

Research, SAE Paper 680790, Society of Auromotive 

Engineers, Warrendale, PA, 1968. 

(30) Stapp, J.P., "Human Exposure ro Linear 

Deceleration," Parts I arrd 2, AF Technical Report 5915, 

I 949-19s l. 

(31) Monis, J.8., "Seat Belt Performance in 30 MpH 

Barrier Impacts," U.S. Department of Transportation, 

NHTSA, Report No. DOT HS-802 480, April 27, t977, 

National Technical Information Service, Accession No. 

PB-269 962. 

(32) Reichert, J.K. and T.J. Bowden, "A Study to 

Determine the Quantitative Effects of Seat Belt Slack," 

DCIEM Report No. 8fR-64, Prepared for Deparrment of 

Transport Canada, Defence and Civil Institute of 

Environmental Medicine, Downsview, Ontario, November 

1980. 

(33) Horsch, J.D., "Occupant 

Dynamics as a Function of 

Impact Angle and Belt Restraint," General Motors 

Research Lab., SAE Paper 801310, Society of Auromorive 


(34) Backaitis, S.H., L. Delarm, and D.H. Robbins, 

"Occupant 

Kinematics in Motor Vehicle Crashes," SAE 

Paper 820247, Society of Automotive Engineers, 

Wanendale, PA. 1982. 

(35) Rattenbury, $.J., P.F. Golyns, H.R.M. Hayes, D.K. 

Griffiths, "The Biomechanical Limits of Seat Belt 

Protection," Proceedings, American Association for 

Automotive Medicine, Louisville, KY, October 34, 1919. 

(36) Shanks, J.E. and A.L. Thompson, 

"Injury 

Mechanisms to Fully Restrained Occupants," McGill 

MI

University, Montreal, SAE Paper 791003, Society of Automotive 


(37) Niedener, P., F. Waltz, and U. Zollinger, "Adverse 

Effects of Seat Belts and Causes of Belt Failures in Severe 

Car Accidents in Switzerland During 1976," Twenty-First 

Stapp Car Crash Conference, SAE Paper 770916, Society of 

Automotive Engineers, Warrendale, PA, 197'7. 

(38) Culvea Clyde C., and David C. Viano, "Influences 

of Lateral Restraint on Occupant Interaction With a Shoulder 

Belt or Preinflated Air Bag in Oblique Impacts*," SAE 

Paper 810370, Intemational Congress and Exposition, Detroit, 

Michigan, February, 1981. 

(39) Dance, M. and B. Enserink, 

"Safety Performance 

Evaluation of Seat Belt Retractor$, SAE" Paper 790680, 

Passenger Car Meeting, Dearbom, MI, June I l-15, 1979. 

(40) 49 CFR, Ch. V (10-l-86 Edition), National Highway 

Safety Traffic Administration, DOT. 

(41) CIR 1419, 1436, U.S. Department of Transportation, 

NHTSA, t976*r977. 

(42) CIR 1424, U.S. Department of Transportation, 

NHTSA, 1970-75. 

(43) Esser, R.C., "Restraint Model Validation Tests 

(Phase l)," Report No. SRL-51, U.S. DOT/NHTSA, Vehicle 

Research and Test Center, February 1983. 

(44) "NHTSA To Check Showrooms To See If Belts Are 

Attached," IIHS Status Report, Vol. 24, No. 4, April 2?' 

1989. 

Investigation of Fire in Cars 


Eva Johansson, 

Saab Car Division 

Abstract 

With the purpose to test how an engine compartment fire 

develops toxic gases and fire in the passenger compartment 

five full-scale tests have been performed by Saab Car Division 

together with the National Testing Institute of Sweden' 

This is a description of the test method, performed tests 

and the conclusions from the tests. 

The firewall was videorecorded during the tests to see 

where, when and how the fire passed though the wall into 

the passenger compartment. 

A heat sensing camera and thermocouples attached to the 

bulkhead were used to measure the differences in temperature 

during the tests. 

To achieve data about the smoke CO and CO2 concentrations 

were sampled during the test. 

The conclusion is that this method is useful for testing the 

fire spread through the bulkhead and the development of 

toxic gases in the passenger compartment. 

The materials that are used to seal around the passages for 

electrical wiring, flexible tubings etc. are extremely important 

for the fire spread in cars. 

442 

(45) Biss, D.J., "Safety Performance Evaluation of Slack 

Effects in Three-Point Belts Using the Hybrid III Dummy in 

Frontal and Frontal Oblique Sled Tests," Presented at the 

l6th Annual Intemational Workshop on Human Subjects 

for Biomechanical Research, Atlanta, Georgia, October 16, 

1988. 

(46) TaniEre, C., D. Lestrelin, G' Walfisch, and A. Fayon, 

"The 

Proper Use of HIC Under Different Typical Collision 

Environments, 

" 

Ninth Intemational Technical Conference 

on Experimental Safety Vehicles, U.S. Department of 

Transportation, NHTSA, Kyoto, Japan, November l-4, 

1982, pp. 321-336. 

(47) Grtisch, L, et al., "New Measurement Methods to 

Assess the Improved Protection of Airbag Systems," Daimler-Benz 

AG, 30th Annual Proceedings, American Association 

for Automotive Medicine, October 6-8, 1986, Montreal, 

Quehec. 

(48) Romeo, D., and R.C. Esser, 

"Airbag Demonstration 

Program (Volume I)," Report No. RD 583-2, U.S' DOT/ 

NHTSA, Vehicle Research and Test Center, January 1984. 

(49) Gloyns, P.F., S.J. Rattenbury, A.Z. Rivlin, et al.' 

"steering 

Wheel Induced Head and Facial lnjuries 

Amongst Drivers Restrained by Seat Belts," Accident Research 

Unit, University of Birmingham, U.K., VIth, International 

IRCOBI Conference on The Biomechanics of Impacts, 

Salon de Provence, France, September 8,9,10, l98l' 


Fire is something that man has leamt to use for different 

purposes. But we are well aware of the risks if the fire gets 

out of control. Statistics about deaths caused by fire in 

Sweden during 1986 shows a total number of 105 deaths of 

these fires most of them were caused by fires in houses and 

only four of them were fires in cars. Although fire in cars is 

not so common it is very important to investigate this 

question to avoid unnecessary risks. 

To start a fire you need three things heat, oxygen and a 

material that can bum. 

There are many hot areas in a car, exhaust systems with 

catalytic converter, engine parts etc. Flammable materials 

and flammable liquids, fuel and different oils are also 

available. All around there is oxygen in the air, so all the 

necessary ingredients to get a fire started are available in a 

car, primarily in the engine comPartment. 

A fire consists of three different phases flammability 

phase, burning phase and the last phase when all the 

flammable material is burnt and the fire ceases (see figure 

l). 

The important thing in the flammability phase is to 

increase the time before ignition. This could be done in two 

ways, either you can choose different materials that are 

difficult to ignite or you can add different chemicals to

Flgure 1. The dltferent phases 

In a flre, 

d.veloped fire 

plastics but unfortunately most of these chemicals also 

increase the toxicity of the smoke. 

In the second phase, the burning phase the temperature i$ 

so high that all organic materials bum. Then the important 

thing is to prevent or at least delay the time for fire spread 

into the passenger compartment. For that a car body is 

needed, especially bulkhead and floor, that is designed to do 

so. 

The test method used as described below is aimed for 

studying the second phase in a car fire. 

Test method 

The tests were conducted at the Swedish National Tbsting 

Institute in Boras. In a big hall ( l8 X 34 m) called fire church 

it is possible to perform full scale fire tests indoors. The 

reason for this is that parameters such as wind and 

temperature can be eliminated. Before testing the vehicle 

was prepared and instrumented as follows. 

Preparation of the Yehicle 

l. The car body was cut off just behind the B-pillars. 

2. The front ofthe car body was supported at the rear end 

to get an horizontal floor. 

3. A board of 20 mm thick plastic glass was mounted at 

the back to allow sampling of CO and CO, concentrations in 

a closed passenger compartment. 

4. The front seats, most of the instrument panel and the 

carpets were removed for full vision of the passages in the 

bulkhead. 

5. The engine compartment was left in running order with 

all equipment in$talled, including liquids. The engine was 

not running before or during the test. 

Trm 

6. The length of the fuelpipes wa$ increased and the pipes 

were bent up along the side structure and filled with fuel to 

get some pressure in the system. 

Instrumentation 

l. Thermocouples were attached close to all the different 

pa$sages 

through the bulkhead and in the motor compartment. 

In each test about twenty thermocouples were used. 

2. Concentration of CO and CO, in the smoke was sampled 

inside the vehicle where the drivers head would have 

been located. 

Other test equipment 

L Videocamera to record the bulkhead area during the 

test. 

2. Heat sensing camera as a complement to the thermocouples 

to see the temperature distribution of the bulkhead. 

3. Cameras to take pictures of special events. 

4. A digital timer. 

Description of the test 

r Four small cubes (dimension 50 X 50 X 50 mm) 

filled with 200 ml heptan (in each cube) was placed just 

in front of the bulkhead in the same height as the floor, 

two on the left hand side and two on the right hand side 

(see Fig. 3). The reason to ignite there is to test the 

worst case. If the fire start$ in the front this would 

increase the time for the fire to enter the pa$senger 

compartment. 

r The cubes with heptan were ignited and the hood 

was closed. 

r During the test the bulkhead was recorded with a 

videocamera, a heatsensing lR-camera,* temperatures 

were measured and the firesmoke sampled. 

r When the passenger compartment was filled with 

smoke the plastic glass at the rear was removed. 

r The test continued until the fire entered the passenger 

compartment. Then the fire was extinguished. 

*IR = the camera works with infrared technology. 

Flgure 2. The car aftor preparailon. Flgure 3. lgnltlon at polnt 1 and 2.

Performed tests 

Altogether five tests were perfbrmed. Three of them with 

Saab 9000 and two with Saab 900. The following is a 

description of the differences in the tests and the results' 

Test No. l-Saab 9000 

As a special investigation the fuelpipes (normally metal) 

were changed to plastic pipes in this test. The reason for this 

was to see whether the fire followed the pipes backwards 

under the floor to the fuel tank. The ventilation fan was not 

running. 

The temperatures close to all passengers and in the engine 

compartment was measured. 

Results: 

The temperatures on the bulkhead was about 20G-300"C 

when the seals around the passages staned to melt away' 

The concentrations of CO and CO2 increased slowly according 

to figure 4 and 5. 

slEffia 

I(OLOXIDI(OISENTRATIff{ 

zsn 

TPPIT] 

- e s e { B 

rilfoflil 

Flgure 4. CO concentratlon. 

5 AB Cmg 8790t7 

r(otDlo)irH(or.cEt{TFAufti tPHrl 

69ffi 

- 2 8 2 ' 

Flgure 5. CO" concantratlon. 

444 

l? 

t4 

ta t{ 

tht|PEnlflf, cc) 

rEtfER nn (c) 

ffi.e 

$ArE ffiV Bli 

6.s l0,o 

TID (lfil{} 

EAAB @EDII EIE 

rrD fl{s{) 

$lr| ffif,tfV Bl8 

TtD fl{D{) 

rE.e 

rE.e

Er?ERAnn (c) tAtt ffi+f,f,Y EtB 

TErfEtAnn (G) lAll ffifff EIE 

s.s 6.e le .8 

TD (l0l) 

FIgure 6. Temperaturcr mcsrursd st the bulkhead. 

Special observations after ignition (minutes;seconds); 

8:fiF-Smoke started to enter the passenger compartment. 

In this test the ventilation fan was not running. 

A question rose, would there have been 

more smoke earlier in the passenger compartment 

if the ventilation fan had been running? 

8:45-Seals around passages started to melt. Flames in 

the engine compartment were visible through the 

bulkhead. 

l0:05-Fire entered the passenger compartment. 

l2:20-The test was stopped and the frre extinguished. 

The fuelpipes in the engine companment were bumt to 

the point where they were bent in under the floor. 

Test No.2*--Saab 900 

The ventilation fan was not running. The temperatures 

close to all passages and in the engine compartment were 

measured. 

Test results: 

The temperatures on the bulkhead ranged from 50"C up to 

325" C. The lowest temperature was measured in the area 

where the bulkhead is made of two metal sheets attached to 

each other. The air between the metal sheets works as an 

insulation. 

Smoke started to enter the passenger compartment earlier 

in this test. The concentration of CO and COr were higher in 

this test according to figure 7 and 8. 

(The toxicity of the smoke is a difficult area. The time to 

get hurt by it depends on different parameters for instance, 

how fast you breathe, the size of the person, an adult can 

manage more, etc.) 

Special observations after ignition (minutes:seconds): 

Z;fi)'--Smoke started to enter the passenger compartment. 

6:(XI--Some seals around passages melted and we could 

see flames in the engine compartment through 

the bulkhead. 

8: l4-Flames entered the passenger compartment. 

l0:5fThe test was stopped and the fire extinguished. 

ff13*iffi*"r*t*^rrm tP'nl 


SAAB ggg 

KOLDIOXTH(O}ICEHTRATT(}{ EPFIIl 

- 2 9 2 . 1 

Flgurc 8. COz concentratlon. 

11 

ll 

n ,l 

r 

A 

t MJ 

tg 

I 

Il 

V 

\ 

t2 

t,t 

445

IEI?ERANN C$T iAAD F+NOY fi! 

IEIfENANN CG) 

TD flrDll 

r8.0 

8 rt gHnfi' &t 

e.o 6.8 

r0.8 tE.t 

TD 00{} 

Test No.3-Saab 90fi| 

In this test the fire was ignited in a different way. The 

Saab 9000 is constructed with a "box" in front of the bulkhead 

(see figure l0). In this test only one cube filled with 

100 ml heptan was used and placed in the center and at the 

bottom of "the box " (see figure I 0). The purpose was to test 

the innerwall of the "the box" 

The ventilation fan was not running. The temperatures 

close to all passages and in the engine compartment was 

measures as in test no; l. 

Test results: 

The fire spread into the engine compartment earlier than 

into the passenger compartment. The fire spread into the 

passenger compartment 2-3 minutes faster than in test no: I 

(Saab 9000, ventilation fan not running). Temperatures and 

concentrations of CO and COr during the test according to 

figure ll and 12. 

Special observations after ignition (minutes;seconds); 

5;00-The fire entered the engine companment. 

446 

TETfERÎIR CC) $Arc ffif,olr Et 

gn. 

8.0 6.e 18.8 

TID CIIDO 

Flgure 9. Temperatures at lhe bulkhesd. 

Flgure 10. "Box" ln the lront of the bulkhead Saab 9000. lgnl. 

tlon at polnt l. 

7:26-The fire entered the passenger compartment. 

9:00-The temperatures on the bulkhead are 100- 

150"c. 

l0;45-The te$t was stopped and the fire extinguished. 

Test No.4-Saab 9000 

This test was exactly the same as test no: I except for one

I(OLOXIDXOHCEHTFATIfl{ TPHT] sAlE s99 - 83 rEnPE*rnn rcf,|o cl stl! sc 

eSss 

eE6S 

I E6S 

IBBP' 

sg9 

4 3 6 

nMEUtN) 

Flgure 1 1. CO concentrstlon. 

XOLOIOXIDKONCENTRATTON 

IPPH] SAAB SS0S B3 

I essg 

I BEE6 

a66B 

6EEE 

4gog 

2006 

s 

Figure 1?. CO? concentrEtlon. 

IE|fENITLR tCNÔ CJ 6A t tfll - lfiov lit 

56s 

eEc 

250 

t5t 

tee 

so 

e 

't*l-,-1 

ffi 

'i l\ 

,/,-* 

I 

A 

*/1"J ! 

te t5 

TtD EnD{r 

t5 

rID [HIN] 

ftl 

h t 

D 1 

s 

e6a 

fl 

l5r 

ls 

6a 

e 

TE|fERÎtft TGfiÔ CJ 

rHe 

ffi 

leG 

,|e6 

fl 

ist 

e 

6ae 

{le 

43S 

c 

sAA! Sa - 811 

6 

SA/IE ffi 

TEIfERAITJfi TEF^D C] 5lA6 ffie 

lss 

t 

- llt 

re 

./i 

A ,fq 

+r 

! i 

/ r"i 

r.' | | r14 

I 

) \ - 

Flgure 13. Temprratured 

at the bulkhead. 

I 

/l/\ 

I u 

T 

--/^ i!::--'- 

i 

t5 

rID EITII{] 

rrp tnitxl 

TrD l|uf{l 

t6 

h l 

lra 

rra 

h t 

to ll 

tc l? 

Tr ll 

ft la 

lu 15 

fr 16 

tc l{ 

Tr 19

KOLOXIPXOTIGEHTRATI't{ CPHIT SAAB ST6 - PROV 84 

e500 

3ffi9 

r60s 

tsso 

540 

o 


KOLDIOXIDXOHCEHIRATT(I{ TPFI{] 

eE66 

ts€6 

6 

Flgure 15. CO2 concenlratlon. 

TE}fEFAT(h EGEAD CJ SllE ffiO - Ffiry !,1 

o6d 

EE' 

,165 

t66 

zeo 

tee 

5 

448 

sAlB gggE - PROV 84 

TID (HINI 

h l 

lrl 

es 

sAlS afre - PHw F. 

rr, rE ao 

s^l! s€66 ' PFOV 84 

salB 9666 - PFOV F4 

rID [HIt{' 

s E t6 15 ea 

Flgure 16. Temperatures at the bulkhead. 

rID ETIIH] 

Ta5 

-,-.- .To I 

Tr9 

te l0 

To ll 

Tp l7 

fs la 

re 2d

parameter. The ventilation fan was running. The purpose of 

this was to investigate whether the concentration of CO and 

CO, increased and if fire spread eadier into the passenger 

compartment. The temperatures close to all passages and in 

the engine companment was measured as in test no: I. 

Flgure 17. GO concsntrstlon. 

XOLDIOXIDI(OHCEHIRATIOT.I 

TPPTI] 

| 2S€g 

Flgure 18. CO? concentrstlon. 

rE|fecAtr,* t6FAo cl 

5EE 

sAts5 Ffilry !5 

SAAA 9gS - PRov Bs 

saJlB s66 

ta lE 

PROV 85 

t5 

ITD T'dIH) 

TID TI{IN] 

TID IItINl 

trl 

tE 

?+2 

toJ 

SÂB DO6 - tsLov Bs 

SllBs - PRovBS 

ILFI,tP^TUR TBftrD c,I SAAB 9{TS _ PROY BS 

5AS 

u5s 

asq 

lsB 

t60 

sg 

g 

6 e te 16 

sAlB sB - PROV ES 

Flgure 19. TemBeratures at tha bufkherd. 

TID THIH] 

t2 t5 

IID IHIHI 

r5 

TIO THTN] 

-'-- 

lo 6 

-.+ Tc t 

- -.- t{ | 

-- Ts I 

1c I0 

tr ll 

Iu t5

Test results: 

Compared to test no: I (Saab 90(X), ventilation fan not 

runrring) a small increase of CO and a large increase of the 

CO2 concentrations were measured according to figure 14 

and 15. 

The smoke entered the passenger compartment a little 

earlier than in test no:1. 

When comparing the temperatures from this test with 

those from test no: l, a delay for about 3 minutes was found. 

This delay was found also in the time for fire entering the 

passenger compartment. The shape of the temperature 

curves was almost identical with the ones from test no:l 

except for the delay. 

Main observations after ignition (minutes:seconds): 

l0:05-The wiring to the ventilation fan was short-circuited, 

the fan stopped. 

l3:25-Some seals around passages has melted away. 

The flames in the engine compartment were 

visible. 

I 3 :55-Fire entered the passenger compartment. 

l5:45-The test wa$ stopped and the fire extinguished. 

Test No.S-Saab 900 

This test was exactly the same as test no:2 except for one 

parameter. The ventilation fan was running. The purpose of 

this was to investigate whether the concentration of CO and 

CO2 increased and if fire spread earlier into the passenger 

compaftment. The temperatures close to all passages and in 

the engine compartment was measured as in test no:2. 

Test results: 

The temperatures measured in this test were almost identical 

with those from test no:2. Also the time when the fire 

entered the passenger compartment was almost exactly the 

same (minutes;seconds test no:1 time 8:14 and test no:5 

time 8:00). 

The concentrations of CO and CO2 did not increase compared 

with test no:2 (Saab 900 ventilation fan not running). 

Main observations after ignition (minutes:seconds): 

5;30-The wiring to the ventilation fan was short-circuited, 

the fan stopped. 

6:45-Some seals around passages had melted away. 

7:Z5-Flames were visible through the bulkhead. 

8:00-Fire entered the passenger compartment. 

l0:19-The test was stopped and the fire extinguished. 

Conclusions 

The repeatability of the test method is good. 

The test flethod is useful for testing firespread through 

the bulkhead and development of CO and CO2 

concentrations in the Passenger compartment. 

450 

HOTOFRUHSTEHPTRAIURENrffiaD Cl sAE SH6 

I oos 

6 6 l o 1 6 m 

TlD thll{l 

Flgure 20. Temperaturea In the englne compartment Sssb 

9000. 

e I s I te t6 

TIO IIIIH] 

Flgure 21. Temperatures In the engine compartment Saab 900. 

If and when the fire spreads into the passenger compartment 

the concentration of CO increases rapidly. The 

concentration is highest when the passenger compartment is 

completely closed. 

If the ventilation system continues to run when a fire is 

beginning in the engine compartment the amount of smoke 

in the passenger compartment increases. 

The bulkheads fire resistance could be increased by 

design. Important for the bulkhead fire resistance are the 

following design features: 

r Temperature insulation of the bulkhead, for 

instance two sheets of metal with an insulating material 

in between. 

r Seals used around passages for wires, flexible 

tubes etc. Use seals that are difficult to ignite. 

. Protect passage$ and holes from flames and heat. 

Locate them as low as possible, use glass fibre or 

metal-foil to reflect the heat. 

r Avoid large holes. 

. Keep the number of holes as low as possible.

The Effect of Restraint Design and Seat Position on the Crash Tlajectory of the 

Hybrid III Dummy 


D G C Bacon. 


Crash Protection Centre 

Ahstract 

In a series of HyGe sled tests on the Hybrid III dummy, 

the effects of seat belt and seat geometry on dummy trajectory 

were examined. A three dimensional spatial measurement 

system was used to evaluate the dummy behaviour 

alongside the standard instrumentation results. 

The first part of the test series compared dummy size and 

then went on to examine the effect of the influence of belt 

parameters. These included belt length, stiffness and effects 

of weblocks and pre-tensioners. An under$tanding was 

gained of the effects of these parameter changes and the 

degree to which a car with poor occupant protection performance 

in a crash may be improved without re-developing 

the vehicle structure. The second part assessed dummy behaviour 

when the seat was out of the standard position. 

These tests involved changing seat back angle, cushion 

angle and cushion friction. The results show the effect on 

injury levels of changes from the settings used in compliance 

testing. The results in this paper provide guidance to 

improved occupant protection through changes in restraint 

system design. 


The aim of this Industry and Government sponsored 

project, coordinated by MIRA, was to develop a reliable 

method of determining the trajectory of a dummy in a crash 

situation and thus to provide vehicle manufacturers with a 

means of correlating dummy movement with variations in 

seat and belt parameters. Changes in crashworthiness 

performance could then be assessed more accurately by the 

techniques developed in this study. 

Initial studies of repeatability in measuring trajectory 

showed that variations were less than lOmm in the dummy 

movement up to a position where impact with a steering 

wheel would occur. Test conditions could be sufficiently 

controlled to enable the small effects of some parameter 

changes to be determined. 

The project was considered in four overlapping phases: 

A Literature Search.-Relevant papers to assist the 

test and analysis prograrnme. 

Development of a Spatial Measurement System.- 

This phase covered the adaptation of a proprietary but 

relatively undeveloped $patial tracking system to 

perform the task of recording the dummy trajectory in 

three dimensions. This included the writing of 

computer programs for analysing and presenting these 

trajectories. 

Test Series.-A programme of dynamic sled tests 

performed on the MIRA HyGe facility. Restraint 

$yrrtem parameters were varied and the effect 

determined on dummy trajectory and injury criteria. A 

total of about 50 tests were planned covering a matrix 

of test configurations. 

Analysis.-Instrumentation data from the dummies 

were analysed by the appropriate routines to generate 

the injury criteria. The trajectories were analysed for 

peak excursions, velocities and other characteristic 

dimensions. 

Changes in restraint sy$tem parameters were evaluated 

against the generated data and conclusions drawn about 

their effectiveness. The topics of Hybrid III repeatability, 

Hybird II/III comparison, seat belt anchorage locations and 

dummy submarining have been covered in another paper 

(l).* This paper presents part of the project work 

investigating the effect of dummy size, seat belt type and 

seat configuration on the crash trajectories. 

Development of a Spatial Measurement 

System 

One of the initial phases of the project wa$ to ascertain 

which of the available measuring systems would be suitable 

for data collection in a HyGe test environment. Three 

systems were evaluated at MIRA under test conditions and 

the selected system is described here. 

VICON is a three-dimensional measurement system 

using photogrammetric techniques, interfaced to a PDP- 

I 1fl3 computer system for data capture and presentation. A 

cine film digitisation $y$tem can also be interfaced with the 

analysis software. VICON records the po$ition of retroreflecting 

markers in the camera's field of view and the 

markers are illuminated by lights clo$e to the camera lens 

axis (figure l). 

rNumbers in puentheses designtrte rcfercnces f,t end of paper 

Flgure 1. Traloctory mof,sursmsnt $ystem. 

451

The standard system uses high stability, 625 line, 2:l 

interlaced TV cameras to produce 50 fields of data per 

second. Each camera in the system produces a wide band 

( l2 MHz) video signal, which contains over a quarter of a 

million pixels in each TV frame. The interface extracts the 

relevant co-ordinate information from the signals from each 

camera and passes it to the system computer. 

For the applications which require a higher sampling rate 

than 50 frames/second, high speed cameras are used. These 

cameras offer sampling rates of up to 200 Hz, and use a 

rotating shutter to freeze the movement. 

Any bright points in the picture are seen as shoft pulses in 

the video signal, typically of lfi) nanosecond duration' A 

Marker Detector module connected to each camera picks 

out these pulses and converts them to accurate, digital 

timing signals. A Co-ordinate Generator transforms the 

timing pulses into crystal controlled counts of the position 

of each point resolving the horizontal position to l;1Ofi) and 

the vertical position to l;3fi) at 60 Hz. For 2fi) Hz frame 

rates the horizontal resolution is l:400. 

These data are transferred via FIFO buffers onto 

computer disc together with any other data that may have 

been recorded at the same time. The resulting disc file is 

limited only by disc capacity and may run to several million 

co-ordinates. 

Once the data capture is complete, the software begins to 

process this vast amount of data. First, a data reduction 

algorithm calculates the centroid of each marker to provide 

a file of co-ordinates representing the position of each 

marker in each TV frame. Next, the data is sorted 

interactively, labelling every marker in every frame for all 

the cameras. Sorting algorithms aid this process, tracking 

the markers through the data field under the operator's 

control. This produces a two-dimensional trajectory file. 

Finally, a three-dimensional mathematical model of the 

motion is created from the two-dimensional trajectory file 

using the principles of close range photogrammetry and 

camera parameter calibration. 

Structured programming techniques have been used to 

provide a modular system of over 20 separate programs in 

an integrated set which runs under the host operating system 

with multi-user capability. 

The applications software package processes the threedimensional 

data available to give a wide range of graphical 

output. 

The following table gives the main specification$ of 

VICON $patial measurement sy$tem. 

Test Equipment and Procedure 

HyGe Test 

A rig was constructed to mount on the HyGe sled which 

was equipped with a right hand production seat. The seat 

itself had the runners removed and was bolted to a rigid 

bracket. Several brackets were manufactured to suit the 

different seat types and configurations required for the test 

programme. 

452 

Table 1. VICON Speclflcstlon 

Accuracy SD ; Zmm with 2m field of vlew 

Preclelon 5D ( Zmm with ?n fleld of view 

Sample Hate t00 Hz 

Number of Cameras 2 - 7 

Number of Matkers t-t0 

SD = Stenderd Deviation 

The acceleration pulse (figure ?) wari chosen to simulate a 

48 km/h (13.4 m/s, 30 mile/h) car impact. This provided 

representative test conditions which would have relevance 

to a range of vehicles. This pulse gave a sustained 

acceleration level of20-249 and delta-V in the region of52 

km/h (14.4 mls.32 mile/h). 

Flgure 2. llyGe eled pulse. 

Anthropometric Dummies 

Unless otherwise specified the Hybrid III (reference 2) 

50th percentile male dummy was used. The 5th percentile 

female dummy and the 95th percentile male dummy were 

used in two tests. 

Before each test series, the dummies were fully calibrated 

to part 572 ofthe Code ofFederal Regulations in accordance 

with Motor Vehicle Safety Standard 208. For each test the 

dummy was positioned as specified in Section l0 of this 

regulation. The position of the dummy was measured prior 

to the test. The dummies were also inspected after each run 

and the joint torques checked. 

Restraint Systems 

Anrhorag,es.-The initial anchorage positions used were 

based on the manufacturers' drawings of anchorage location 

relative to H-point. Variations were made from these

positions to determine an optimum restraint position. All 

seat belt anchorage positions were measured relative to the 

dummy H-point position as determined using a reference 

manikin. The standard seat belt position was as fitted in a 

production vehicle with the seat at mid position. 

Seat belts.-The standard production belts used for the 

tests were of lZVo elongation webbing and 3m in length 

(unless otherwise stated). The length of belt stored on the 

reel before each test was in the region of 625 mm. 

Seats 

Three types of seat were used for the tests. The two 

prinicpal models (Type A and Type B) were producrion 

seats supplied by vehicle manufacturers. The third was a 

leather covered version of Type A seat. This was used to 

investigate the effects of seat friction. 

For the standard configuration the seat$ were installed 

with the standard cushion angle as fitted in production and 

the back angle set to 25 degrees unless otherwise specified. 

Some tests were performed using one seat with a range of 

back and cusion angles. New seats were used for the majority 

of the tests. 

Photography and Spatial Measurement 

High reflectivity markers were fixed to the dummy in the 

positions shown in figure 3. When illuminated by a high 

inten$ity light source, on the same optical axis as the viewing 

camera, these markers show up as bright spots. The 

markers themselves were in the shape of spheres, hemispheres 

or discs which had been covered with retro-reflective 

tape. The positions of the viewing cameras were similar 

to those shown in the sketch of the VICON system in figure 

l. 

z 

,*J 

Flgure 3. VICON marksr po8ltlons and axls convrnllon8. 

Three l6mm high speed cameras running at 1000 pps 

were also used to provide a visual record of the tests. 

To assess the contact velocities and contact angles ofboth 

the head hitting the steering wheel and the knee hitting a 

knee bolster pad, the planes of a fictitious steering wheel 

and bolster pad were added to the trajectory plots. These 

planes were defined using standard EEC data measured 

from acutal production vehicles. 

Test Instrumentation 

For all the tests the dummy was instrumented with triaxial 

accelerometers in the head, chest and pelvis. The Hybrid III 

dummy also contained a chest displacement transducer and 

instrumented neck. Load cells were fined to the lap and 

diagonal belts and a wire potentiometer was used to measure 

the amount of seat belt reel-out. 

For all the test$ the instrumentation results were analysed 

ro the requirements of FMVSS 208. 

The accelerometer outputs from the head were analysed 

to produce the Head Injury Criterion (HtC). Even though 

the dummy head did not strike any structure in these rests 

and HIC without impact is not widely accepted, the HIC was 

used to rank the severity ofthe head accelerations. 

Co-ordinate Sign Convention 

For all tests the coordinate system used was as follows: 

X-longitudinal, positive rearwards 

Y-lateral, positive to dummy right hand side 

Z-vertical, positive upwards 

These coordinates are funher described in figure 3. 

Test Programme 

The test programme was divided into several sessions 

where the tests were grouped conveniently according to the 

parameters 

being studied and also to provide a logicat progression 

to the project. In outline, the subject matter for this 

paper was as folJows; 

r Comparison of 5th percentile female and 95th 

percentile male dummies 

r Effect of seat belt webbing stiffness 

r Effect of seat belt length 

. Pre-tensioners and Weblocker investigation 

r Effect of seat back angle 

r Effect of seat cushion angle 

r Effect of seat cusion friction 

Table 2 summari$es the relevant tests. To aid comprehension 

of the programme which involves several variables, 

some sub-matrices of the table are given in the discussion of 

the results. 

Results 

Where possible graphical data were prepared foreach test 

as follows: 

. Top of Head, forehead, chest, pelvis, knee 

Trajectory (X-Z plane) 

X Displacement and velocity/time 

Z Displacement and velocity/tirne 

Resultant velocity/X displacement 

Change in angle/time (X-Z plane) 

r Shoulder forward rwist/time (X-y plane) 

In addition summary data from the instrumentation results 

and trajectory analysis were provided as follows: 

r Head 

HIC 36 ms.250 ms 

Resultant acceleration (3 ms) 

Forehead peak resultant velocity 

453

Table 2. Testa carrled out In parametrlc study. 

Rm No T6Et spBciflcettm Comonts 

l6 

2l 

22 

2' 

24 

2' 

26 

27 

28 

z? 

l0 

il 

12 

11 

I4 

,5 

t6 

t7 

]8 

19 

40 

4l 

42 

41 

44 

46 

47 

,2 

5J 

54 

56at BA CA lclt 

B 

B 

B 

B 

B 

A 

A 

5U 

A 

su 

5U 

LF 

5U 

su 

SU 

SU 

A 

5U 

5U 

LF 

5U 

A 

LF 

A 

B 

B 

B 

I 

B 

5td 

std 

std 

5td 

std 

std 

100 

100 

400 

400 

std 

5td 

srd 

5td 

5td 

5td 

std 

400 

400 

400 

400 

400 

400 

400 

25F 

5td 

std 

std 

5td 

std 

std l2c 

std l7s 

5td 8* 

std 5r 

std Ltl 

std t2$ 

std uc 

std tz$ 

std tzn 

5td t?x 

5td FA 

5td l2c 

5td l?f, 

std RA 

std FF 

14.50 LzZ 

14.5c l2x 

14.50 r2r 

14.50 r2x 

-5.50 lzx 

-5,50 l?s 

-5.50 FA 

-5.50 FA 

-5.50 r2s 

-5.50 r2x 

std r2s 

std r2s 

std l2r 

Sttl l?r 

std l2c 

B -TypaBueat 

A - Typc A saet ln dtendrrd ffidltlon 

stj - Typs A aart, anti-aubfEflnlng pen 

remvrd 

LF - Lddthdr ecat (low frlction) Type A 

RA - Lorsr anchorEgrs mvad hlghar end to 

tha fedf 

FA - Ldwer enchorEgca mvod lowar end fofrerd 

FF - Foot r66t mv6d foEHard 

454 

Bad6lln6 t66t 

EffBct of H6bblng stlffruBs 

Effact of wabblnq atlffmaa 

Effact of bclt langth 

Effgct of balt lanqth 

Standard Typa A aaata 

Slgnlflcent Bubrarinlnq 

Stghtftcant Buboarlnlng 

5th oarcantlb famlp 

95th pilccntlls mls 

{ablockar uscd 

HEblock6r u$d 

Hl16 pr6-trnalonrr 

BA - Beck Angle 

CA - Cu8hlon Angl6 

tH - lzf wcbblng 50ftn 

longar than atandard 

stl - lzs xebblnq 500m 

slbrtsr than standard 

Forehead resultant'impact' velocity at steering 

wheel plane 

Max. change in head angle 

Head angle at 'impact' 

Max. angular velocity 

Max. forward displacement 

Chest 

Resultant acceleration (3 ms) 

Deflection (Hybrid III only) 

Max. Forward displacement 

Max. Vertical displacement 

Max. Change in angle (X-Z plane) 

Shoulder forward twist (X-Y plane) 

Pelvis 



Max. Change in angle 

Knee 



'Impact' 

velocity with knee bolster plane 

Seat Belt 

Lap load 

Diagonal load 

Reel-out 

Dummy Size 

Tests were made with a 5th percentile female (Run 46) 

and 95th percentile male (Run 47) to compare their performances 

with the 50th percentile Hybrid III (Run 16). The 

restraint systems and seats for the three tests were the same. 

The comparative trajectories for the head, chest, pelvis and 

knees are shown in figure 4. Summarized te$t data for the 

tests are given in table 3. 

Flgure 4. Dummy slze comparlson. Head, chest and pelvla 

trsl€ctorles. 

Table 3. Dummy motlon comparlson r€sults- 

Rh 

No 

I6 

46 

47 

- stlmRD sth ftEt[tnu 

I XXEE ilOI AVAITABLT I 

)my Typrl5izr HIC 

( I6'ns) 

HIII 50th 

prre;ntib nilc 

5th psrc6ntl16 

fenele 

95th prrcrntllB 

m16 

194 

I094 

1t2 

tl6dd 

xcurrion 

(mm) 

650 

622 

829 

H6Ed 

Psak 

/6loclty 

12.9 

12.5 

Il,6 

lhad Inpaet 

Veloclty 

(n/a) 

ll.9 

ib ContEct 

8.9 

Chset 

AccsIsrEtlffi 

(9 ovor !m) 

16.7 

44,9 

J2.O 

The instrumentation results for the 5th percentile female 

show high HIC and chest acceleration values. The acceleration 

of the dummy is expected to be higher than a 50th 

percentile due to the lower mass if subjected to the same 

forces. As the diagonal belt runs high across the chest, 

shoulder rotation and head forward excursion were low. It 

can be seen from figure 4, the 5th percentile dummy does 

not make contact with the steering wheel plane when sitting 

in the standard position. In practice a 5th percentile driver 

would be expected to adjust the seat, bringing the occupant 

closer to the vehicle interior. In this case the head could 

contact the steering wheel at any velocity up to the peak of 

12.5 m/s. 

The 95th percentile male dummy produced a greater moment 

about the diagonal belt and was less well restrained. 

There was considerable rotation ofthe upper body out ofthe 

belt with a head forward excursion of some 830mm. This 

was high compared with 650mm for the 50th percentile 

Hybrid III and 620 mm for the 5th percentile female Hybrid 

II. By the same reasoning that the 5th percentile HIC and 

chest acceleration are high, the 95th percentile results are 

low by comparison. 

As with the sth, the 95th percentile occupant would be 

likely to adjust the driving position, changing the head 

impact position. The velocity at impact would be expected

to be higher than the 8.9 m/s recorded but lower rhan rhe 

I I.6 m/s maximum head resultant velocitv. 

Belt Type 

Tests were conducted to compare the standard belt system 

(Run l6) with a range of belt options; 

| . l7 Vo Webbing stiffness, standard is l27o (Run 2 I ) 

2. 8% Webbing stiffness (Run 22) 

3. Belt Sfi)mm longer than standard (Run 24) 

4. Belt 50omm shorter than standard (Run 23) 

5. Weblock sy$tem$ (Runs 52,53) 

6. Pre-tensioner system (Run 54) 

Webbing Stiffness 

Trajectory results for the 3 tests with different webbing 

stiffnesses are shown in figure 5. 

- SIAilDAR0 IELI 

---- t?* ut6Bt[G 

-.-.- E* wEEBttG 

(tHEst I PELvrs 

ilOT AVAIIAELE ' 

5,,9t#ij,ryJi"Tlns 

Bllffness comparlson. Head, chest and pet- 

In the case of the lTEo belt there was an increase in the 

head forward excursion. The 87o belt showed a slight decrease 

in excursion from standard but this could be within 

normal scatter. The trajectory graphs show the increase in 

head and chest forward movement due to the l77o belt. The 

head impact velocity into the steering wheel plane increased 

for both the 8% and lTVo belts due to the changed impact 

location. The llVo belt also allowed more movement of the 

chest. The belt stiffness appeared to have had most effect on 

chest acceleration (see table 4 for test data). The instrumentation 

results showed an increase, in chest 3ms exceedence. 

for the stiffer belt and a corresponding decrease for the 

softer one. Both HIC values however were greater. 

Belt Length 

Trajectory results for the three tests with different webbing 

lengths are shown in figure 6. 

Changing the length of the seat belt had a much greater 

effect on the dummy performance than the belt stiffness. 

There was a clear increase in head excursion from 650mm to 

70omm for the longer belt (Run 24) and a decrease to 600 

for the shorter (Run 23). The corresponding head impacr 

velocities were also respectively increased and reduced (see 

table 4 for test data). The chest forward rotation in the case 

of the long belt was some 15 degrees higher than with the 

shorter one. 

Fig.ure 6.. Webblng length comparleon. Head, chest and petyt$ 

lralectorles. 

Seat belt reel-out as recorded by the wire potentiometer 

was down to 25 mm for the short belt and up to 90mm for the 

long belt. The standard belt reel-out was normally in the 

region of 40-50mm. 

Weblock and Pre-tensioner Systems 

The effect of the weblock (Runs 5?,,53) was similar ro 

that of the shortened belt. Reel-out was of course reduced 

and head excursion brought down (figure 7). The impact 

velocity was significantly reduced from the standard configuration 

(table 4). 

0.t 

- STArmno fftI 

---- vtl Ldrr 

- - - PntrtBtotEn 

3tttfir6 Yxttl 

FtltE 

aff 

-04 -0 4 -0.{ -01 

tttt't 

Flgure 7. Weblock and pro-trnalonor comparleon. Head, chest 

and pelvla tralectorlee .- 

Table 4. B€lt type results. 

nI 

Xo l.lt lyp6 HIT 

(l6s) 

l6 

2l 

22 

24 

52 

9l 

54 

Stond.rd 

l7t 

8t 

-5oth 

+50lh 

l|block 

I.bIock 

P16trnrLmr 

t9l 

429 

58r 

441 

546 

,26 

?BJ 

272 

H6od HrEd :hsEt lhr.t Ch$t B.lt 

I+rct 

Yrloctty (n) 0 sv6f ItrE) in) (dss ) ttut 

(r/c) 

(n) 

tt.9 

12,J 

12,l 

lI.7 

rr.] 

I0,2 

9.4 

10.0 

650 

677 

640 

600 

toJ 

,20 

t00 

izJ 

The pre-tensioner system used for the tests was of the 

type normally activated by a length of cable attached to a 

collapsing part of the vehicle structure. A bracket was 

36,7 

r5, ] 

19,5 

Jr,6 

16,6 

t,9 

tt, r 

t5.8 

250 

t{l0 

7J2 

24I 

264 

214 

2fi 

n5 

44 

40 

J4 

t7 

{4 

42 

4' 

49 

52 

25 

90 

ft 

29 

d0 

455

mounted to the sled to retain the cable outer and the cable 

itself was attached to the sled rails with a shear bolt. This 

bolt may have sheared slightly early but the general performance 

of the pre-tensioner system was good (Run 54). The 

seat belt reel-out was not down as far as in the weblock tests 

but the head excursion and impact velocity were both down. 

Both these $ystems were seen to have a beneficial effect on 

rhe HIC. 

Seat Configuration 

Cushion and Seat Back Angles 

All seats were of standard production design, with cloth 

covers and incorporating an anti-submarining pan. The 

design cushion angle was 4.5o up at the front and the 

standard back angle was taken to be 25o reclined. The 

anchorage locations were standardised for most of the test 

runs, 

Data from tests in table 5 have been summarized in Table 

6. With a constant cushion angle of 4.5o, dummy trajectories 

for the three back angles are compared in figure 8' The 

effects of the three cushion angles against constant back 

angle is shown in figure 9. 

Tsble 5. Test matrlx lor cushlon angle and sest back angle 

lnvestlgatlon. 

Back Angle 

Cuehion 

Anglo 

100 250 40o 

-5 .50 44 42*t 41++ 

4. 50 26 25 2S 

14.50 55 ,7 

* with lower anctroragoa for*ard and louer 

++ illth leather s6et (1or friction) 

Table 6. Seat conflguratlon resulte. 

Rm 

t\t5 

25 

26 

28 

tl 

Run 

Ntmbers 

The base configuration for this series was the Type A seat 

set with a 25 degree back angle and 4.5 degree cushion angle 

(Run 25). This produced slightly different results from the 

Type B system, notably a higher belt reel-out and head 

excursion. The head velocity was, however slightly lower' 

456 

HIC 

]60t ) 

15 ]70 

t7 

40 

42 

4t 

44 

Il5 

27z 

420 

h2 

599 

47i 

475 

494 

468 

lbrd tbid 

Imcct 

/sloclty 

(n/r) 

!xcurtlon 

(n) 

lt.9 

lr.0 

I2,2 

It,8 

ll.2 

lr.6 

t2,2 

12.6 

ll.9 

lr.0 

685 

610 

750 

7J2 

660 

795 

748 

77} 

7t7 

660 

Ch6t 

Frd 

dtBpl 

(n) 

z6s 

290 

T(R 

276 

268 

,7e 

,21 

t9t 

t04 

212 

Chirt 

Dffi 

dlBpl 

(-) 

I6l 

14I 

196 

16r 

t4r 

L52 

188 

268 

205 

170 

Chert 

Rot. 

(dlc) 

44 

J7 

57 

47 

60 

47 

l8 

57 

42 

P6lYlr 

Fild 

dtrpl 

(u) 

168 

189 

t85 

202 

148 

166 

22J 

,66 

20r 

198 

PcIvld 

Dom 

dlspl 

(m) 

48 

E' 

lI4 

49 

4l 

2E 

4t 

148 

54 

58 

P6lYlr 

Rot, 

(d"9) 

t0 

I 

t4 

22 

I5 

ll 

A 

r6 

r5 

Flgure E. Back angle comparlson' H€ad, chest, pelvls end kne6 

trai6ctorles. 

Flgure L Cuahlon angle comparison. head, chett' pelvls and 

knee trslsctories. 

Changing the seat back angle to l0 degrees reduced both 

forward excursion and head impact velocity (Run 26). 

Although head overall excursion from rest is low, the head 

actually moves slightly further forward than standard due to 

the more forward statting position. The belt, however, pulls 

the dummy up more sharply than in the standard position' 

The fact that less belt is left on the reel in this position could 

have a contributing factor to this. The overall change in 

angle of the body segments during the test was reduced, as 

would be expected, although the head angular velocity 

showed little change. Similar effects were observed in 

reverse when the seat back angle was increased to 40 

degrees (Run 28). 

With the standard back angle and a cushion angle of 14'5 

degrees there was little change in the results (Run 35). A 40 

degree back angle (Run 37) with this cushion angle 

produced a similar effect to that of the standard back angle' 

Excursion and impact velocity both increased. 

It appears that, with a good lap belt, the cushion angle has 

little effect on the dummy trajectory. 

Seat Friction 

All seats were fitted with the standard anti-subrnarining 

pan. The lower friction was achieved with a leather seat 

covering. 

A limited number of runs were carried out with the low 

friction seat and comparisons were made with cloth covered 

seats. Comparative head, chest, pelvis and knee trajectories 

are given in figure 10. Table 7 provides test information on 

the friction series of tests.

0.t 

0.r 

c! 

- mrr iar, fr|il|lo ciiitn|ril 

---- tEllHtr stl. Stmm 

r*l6ulATr0l 

- -,- ltfiffin ttal. S' 

6tct. -5 5. fl,rtffi 

,*na*tr.\ 

slttfrtr vEtt 

?L^xl 

# 

ms 

HEAO 

-r'+ -l I -t a -0. -0.i -0 { -t t o0 6 t 0.a. t,. 

Flg.urr 10. Low frlcllon eeat tssts. Hced, cheat, pelvls and knee 

tralsctorles. 

Table 7. Test matrlr tor rsst lrlctlon Invosilgf,tlon. 

Cuchlon 

Angle 

* 

++ 

Beck Anglo 

l0c 250 400 

-9.50 40r, 45* 

4.50 'I*, 2t+* 

14.50 

loathor soit - frlctlon coefficlent 0.45 

cloth seat - frictlon coafficlont 0.7 

Run 

Nulbere 

In the standard position (Run 3l) there was a general 

increase in forward movement of the dummy in the region 

of 30mm at the pelvis and 50mm at the head. Impact velocities 

however were similar. When this seat was tested in the 

extreme situation of a 40 degree back angle and a -5.5 

degree cushion angle (Runs 40,43) there was lirtle change 

in the dummy overall movement save the expected increase 

in the chest linear and angular displacement. 

Conclusions 

Of the three spatial measurement systems evaluated in 

the early stages of the project, the VICON system was 

selected as providing automatic data capture capacity to the 

specification of 30 markers at 200 Hz sample rate using 3 

video cameras. This equipment was u$ed successfully to 

gather dummy trajectory data on the 50 tests planned on this 

project. 

The tests with the varying dummy size and seat positions 

emphasise that wherea$ compliance testing is performed 

under one set of conditions there is a wide range of others 

that need consideralion during the design and development 

stages. For example, the tests have shown that a good lap 

belt fit allows a wide range of cushion angles to be used 

safely. 

The seat belt tests showed that belt stiffness changes do 

not have as much effect as changes in the belt length. A 

shorter belt shows less reel-out and restrains the occupant 

bener. Weblocks and pre-tensioners 

also achieve this. 


The Motor Industry Research Association would like to 

thank the following for their contributions and assistance in 

the project, and their agreement to publish. 

UK Department of Trade and Industry 

ASE (UK) Ltd 

Rover Croup Ltd 

Britax Safety Systems Ltd 

Ford Motor Company Ltd 

Jaguar Cars plc 

Rolls Royce Motor Cars Ltd 

Transport and Road Research Laboratory 

Oxford Metrics Ltd 

References 

(l). Freeman, C.M. and Bacon, D.G.C. "The 

3-Dimensional Trajectories of Dummy Car Occupants 

Restrained by Seat Belts in Crash Simulations." 32nd Stapp 

Car Crash Conference, May 1988. 

(2) -'Hybrid III-A Biomechanically Based Crash Test 

Dummy". J. K. Foster, J. O. Kortge and M. J. Wolanin. 2lst 

Stapp Car Crash Conference, 1977. SAE No. 770938.

Technical Session 2B 

Accident Investi gation 

and Data Analysis 

Chairman: S. Christopher Wilson, Canada 

In-Depth Study of Motor Vehicle Accidents in Japan 

Masahiro ltoh, Koshiro Ono, 

Japan Automobile Research Institute, Inc. 

Yasuhiko lrie, 

Traffic Safety and Nuisance Research Institute 

Abstract 

The Ministry of Transport (MOT) is carrying out a motor 

vehicle accident investigation every year for the accurate 

determination of the actual status of motor vehicle accidents, 

clarification of the relationship between motor vehicle 

construction/equipment and human injuries, reinforcement 

of motor vehicle safety standards (Safety Regulations 

for Road Vehicles) and the promotion of technical evaluation 

of individual items of existing standards. Each investigation 

consists of case studies aimed at the in-depth study of 

specific details of accidents, conditions of accidents occurred 

in investigation area$, and statistic survey to find the 

prop€r position of each accident subject to investigation. 

Roughly l0O cases of accidents are studied and 4,500 cases 

are subject to the statistic survey for the duration of four 

months in every year. Investigation data are processed into 

computer files to allow a proper analysis, according to the 

purpose of study at any time as called for. 


The MOT of Japan has been carrying out the 

investigation and analysis of motor vehicle accidents every 

year since 1973, in line with the recommendation submined 

by the Council for Transport Technics in 1972, which 

pointed out that **it is necessary to clarify the correlation 

between traffic accidents and motor vehicle construction/ 

equipment (interior parts and components), and take 

effective $afety measures" to upgrade motor vehicle safety. 

The second recommendation was submitted by the Council 

in 198 I, which stated that the system and contents of the 

investigation should be strengthened. ln response to the 

recommendation, the methodology of investigation and 

analysis employed in past has improved since 1986, and 

areas, duration and number of accidents subject to 

investigation have been expanded, along with further 

upgrading and reinforcement of motor vehicle safety 

standards. Moreover, the investigation/analysis system has 

been improved to obtain useful data for the international 

harmonization of the standards. 

The Traffic Safety and Nuisance Research Institute of the 

MOT and Japan Automobile Research Institute, Inc. (JARI) 

are engaged in such investigations. An appropriate study 

theme, such as the effectiveness of fastening seat belts, the 

effectiveness of steering with energy absorbing system, the 

actual condition of side collision, the occurrence of vehicle 

fire and the compatibility of large trucks, etc. is selected 

each year, and approximfltely 100 cases of accidents are 

studied in detail for four months every year. Case study/ 

analysis on correlation among the striking (impact) speed, 

vehicle deformation, injured region of occupants, and 

damageable objects of vehicles is done for each accident. 

All traffic accidents involving human fatalities/casualties 

that have occurred during the investigation period in areas 

concerned are also studied. The MOT reflects findings of 

the analysis to upgrade and reinforce motor vehicle safety 

standards (to equip the driver's seat belt with ELR device, 

to use HPR glass for motor vehicle windshields, 

reinforcement of fuel leakage preventive requirements, 

etc.), as well as the confirmation of each item of standards 

and the intemational harmonization of such standards. 

The present status of motor vehicle accident 

investigations and analysis carried out since 1973 will be 

described in this paper, and analytical results on 

relationships between types ofaccidents and vehicle speeds 

immediately before collision, occupant injuries and impact 

speeds, comparison of severity of occupant injuries in terms 

of EBS (Equivalent Barrier Speed) between occupant$ use 

seat belts and those not use them, the degree of sideward 

intrusion of the vehicle on the side collision and the severity 

of occupant injury and so on will be also reported. 

Motor Vehicle Accident Investigation 

Methodology 

Accident investigation are being carried out by JARI as a 

contract work entrusted by MOT. The "Motor Vehicle 

Accident Analysis Committee" consisting of experts in 

motor vehicles, medicine, traffic engineering, govemment 

administration agencies concerned, etc. is formed within 

JARI for deliberations on the methodology of investigation 

and analysis, etc. and for the preparation of annual reports 

based on findings of the investigation and analysis. 

The accident analysis section of Traffic Safety and 

Nuisance Research Insitute of the MOT and the accident 

analysis team of the JARI are engaged in the investigation 

under the joint cooperation of the National Police Agency, 

the Ministry of Construction, the Medical Association of 

Japan and so on. 

The duration of investigation is four months per year, 

459

followed by five months of the analysis and preparation of a 

rePon. 

Objectives of investigation and analysis 

Indepth investigation for the determination of the actual 

status of accidents in order to clarify the relationship between 

motor vehicle construction/equipment and the severity 

of human injury, contrilute to the upgrading and reinforcement 

of motor vehicle safety standards, and the 

verification of effect of the standards. 

Accidents Subject to Investigation 

Case studies 

Investigation are done on vehicle to vehicle accidents and 

single vehicle accidents involving mainly passenger cars 

and trucks (hereinafter refened to four or more wheeled 

vehicles unless otherwise specified) with relatively large 

damages of vehicles of accidents in which the correlation 

between the vehicle construction/equipment and injuries 

appear to be significant, and those deemed effective for the 

objectives of investigation. 

Statistic surveys 

Statistic surveys are carried out for all traffic accidents 

involving injuries occurred during the four month period of 

investigation in a specific area (Ibaraiki) in order to determine 

the statistic position of each accident subject to 

investigation. 

Areas and number of accidents investigated 

Seventy or more accidents are investigated out of all 

accidents occurred in the specific area. For particular accidents 

such as those involving of injured occupants use seat 

belts and vehicle fires. 30 or more accidents were selected 

out of accidents occurred in areas around the specific area 

ffi*'.**" 

trlr.. rlrlr.r 

tlll 

AEcla.rr i.r.rlitrrirr 

+ 

{ 

ff' 

(D lrr6iratori lo to thr Flie hcrdqurrkE 0r Frlin 

ti! dlracM rid condition 0[ ercb rccidcni tt rrE 

of t.lctbom. 

O Ttc Flicc isdqsrtcr+ inlorr ol Ellftt rpf.otrirti 

rtrrdcrr to h iircrtitrtcd by tba inrEulalqr L!f,cD 

lilBtrutis tm lcs to ihc Flrcc ttrtlor, tbr accidtrt 

sE tE rrFrr EhoE ctc. 

Flgure 1. Outllna of csso study method. 

460 

(Tokyo, Chiba, and Tochigi) and other areas in Japan (mainly 

Hokkaido, Sizuoka, Aichi, Hyogo, Okayama and 

Hiroshima). 

Method and System of Investigation 

Case studies 

Outline of the method of the case studies and the system 

of investigation are as shown in figure I . Investigators of the 

expert level are carrying out investigations according to the 

following procedure. 

(l) Investigation in the Specific Area (Ibaraiki) 

Investigators go to the police headquarters of the 

prefecture or confirm the occurrance and condition of each 

accident by means of telephone, and select appropriate 

accidents to be investigated. For each accident thus 

selected, the investigation team goes to the police station 

and finds out the nece$sary information, and carries out the 

investigation on traffic environments on the accident scene, 

together with the investigation on vehicle involved in the 

accident ifthe vehicles are still on the scene ofthe accident. 

If the vehicles are already moved out of the scene, the team 

finds the place where the vehicles were moved, then goe$ to 

the repair shop, etc. to check on the vehicles. Afterwards, 

the team goes to the hospital, etc. for the investigation of 

regions, details and severity of injuries of person(s) 

involved in the accident. 

(2) Investigation in Other Areas 

lnvestigators obtain accident information from the Road 

Traffic Information Center, etc. and select proper objects of 

investigation, or phone into the prefectural police 

headquarters for the injury of occurrence of accidents or 

confirmation of details of each accident so as to select 

proper objects of investigation. Then the inve$tigation team 

carries out the necessary investigation according to the 

same procedure as that of the specific area. 

Statistic survey 

ln order to determine conditions of accidents occured in 

the specific area during the investigation period and to identify 

parameter for the selected cases to be studied, investigators 

go to the prefectural police headquarters and check on 

the number of vehicles involved, the number of persons 

involved per type of accident (fatality, $evere injury and 

minor injury) and so on for all traffic accidents which involve 

casualties or fatalitie$. 

Data collection forms 

As the purpose of the motor vehicle investigation and 

analysis is the clarification of the relationship between the 

motor vehicle construction/equipment and the severity of 

human injury, items to be investigated were so set that 

conditions of vehicle construction/equipment, and the correlation 

between parts of components that contacted and 

occupants, and injured regions of occupants could be clearly 

determined. Details of investigation were al$o set that

traffic environments on occurTence ofaccident can be determined. 

Figure 2 shows major items of investigation in systematic 

manner classified by "human factors", "motor vehicles" 

and "traffic environments". All investigation items 

are converted into computer codes, and details entered in 

the data collection forms are stored in computer files after 

the completion of investigation and data processing, which 

are used in subsequent analysis. 

@t 

g 1l 

rffiil 

t4l 

(fudrit Ertscrir, rErul 

affif 0l *cuMrrr I 

Flgure 2. Detalls and malor Item$ of Invertigation. 

Outline of Accidents Subject to the 

Investigation 

Number of accidents and number of persons 

involved in accidents classified by fiscal year 

and types of accidents 

Table I shows the number of accidents classified by fiscal 

years. The number ofaccidents subject to case studies so far 

Table 1. Outllne ol numbers ot dlfferent types of accldentg 

lnYestlgated. 

ft- ts f T u 

H[1. S f,u( b 69. il ut@& 157{15 S 

lfftt?.8 fle{. e flffis m{19.n il66 

tilflg lD lt0l s l0{ts ll0[# ril0trs 

tfl0t ls fot s s67. fl il0?.6 ilsca ll 

Fdr*|.. l0 *t0o.1.. d rr !'crf l. 111illfi I{ LE llo0.p lza0zE 

57(tS t?{ls 

ffi0m rff(Iffi lmi lm tfi( Im lott( t@ 

il5 t I lm lul 

l0J tr9 

il9 s il $l 

il s 

ti 

I t t3 

lro ts ln I& lsl 

Drrr.rr | ourr Esnrlr lsl u fll lBl 

ldiltrril *rorctcli/H r'&rr r rtcl'rri lfi l9 il0 

ml s fll E? WI 

k ef FrHr lrlld lhtrl'ird) u{ {6 6loth sof, s E7@. S il5tiD 

lld{3afi l*{t.0 tsH.s 155(S. m lfit(5{. m 

Ht{t Is Det0 lflts ls ilt2t ffi Irr5ffifl 

mt(lm s(tE B(IE Hi tffi I{{{ lffi 

amounts to 1,015, consisting of 157 cases of vehicle to 

vehicle head-on collisions, 264 cases of side collisions, 168 

cases of rear-end collisions, 168 cases of rear-end 

collisions, 245 cases of single vehicle accidents, 124 cases 

of four-wheeler to two-wheeler or bicycle accidents and 57 

cases of vehicle to pedestrian accidents. 

Vehicles involved in the accidents consist of 1,04 1 

passenger cars, 139 vans, 514 trucks and l3 other 

vehicles-1,707 vehicles in total. The number of occupants 

in the vehicle$ amounts to 2.834. Other included in case 

studies are 80 motorcycle/mopeds and 44 bicycles, 

involving 210 riders/cyclists. As for the classification of 

accidents involving occupants, there are 265 fatalities, 

1,664 casualties and l,l l5 no-injury case$. 

Number of vehicles classified by principal 

direction of impact force and general area 

of damage (forms of impacts) 

The number of vehicles and ratio of the principal direction 

of impact force, generfll area of damage and type of 

accident for vehicle to vehicle collisions and single vehicle 

collisions subject to case studies are shown in figure 3. 

= Rerr-end colltsion 

: 

ffi srnrle vthicle sccideit 

lrnr-iidcd 

| ?7( l.9t) 

0thrrr qrnrrr | ztg (15 

I ztg (li, ir) ll) 

Abrcrt I 19( l.3t) 

Toul I l$l( l00f) 

x0tc 

0ilrrr: l0f. 0lt. 0t8, crc 

lbr.nt: Firc hrr!lr ntt 

rictr r{iiiil. lrtl 

irtq rrtrr. !tE, 

Lrrvc 0ut unltrori, 

Flgure 3. Numbers ol vehlcles and rEtloe claaalfled by forma of 

vehlcle lmpact8 and type of accfdenla. 

The classification is done by using three digits ofcodes of 

CDC (Collision Deformation Classification according to 

SAE J224)-namely, frontal collisions are classified into 

I lF, l2F and 0lF, side collisions 02R.03R.04R.08L. 09L 

and l0L, rear-end collisions 058, 068 and 078. Other colli- 

461

sions such as l0F.01R.048. etc. are classified as those with 

different directions of impact force, and vehicle fires during 

running without involving collision are classified as 

"absent". 

The number of vehicle subject to analysis is I ,43 I , which 

consist of782 vehicles involved in front collisions (54.67o), 

187 vehicles involved in side collisions (l3.\Vo), 197 vehicles 

involved in rear-end collisions (l3.8%o),219 vehicles 

involved in collisions of other direction of impact force 

(l5.3Vo),27 vehicles involved in collisions from both sides 

(l.9Vo) and l9 vehicles without collisions (l.3Vo). 

Number of occupants and severity of injuries 

classified by principal direction of impact 

force 

Number of occupants and ratios classified by the princi 

pal direction of impact force, general area of damage and the 

severity of injury are shown in figure 4. The greatest number 

( 1,280) of occupants classified by the prinicpal direction of 

impact force is observed in frontal collisions, followed by 

345 occupants in side collisions and 287 occupants in rearend 

collisions. 

Figure 4. Numbers of occupants and ratlos classlfled by lorms 

ol vehlcle Impacts and severlties of iniuties, 

Ratios of injury severity classified by the principal direction 

of impact force shows that rear-end collisions account 

for the highest rute of 9l.7Vo for minor injury and no-injury 

cases. Side collisions for the highest rate of 35.6Vo for severe 

injury and fatality cases. 

462 

InjurY sGYerll, 

fl r. t-rts o 

ffi r.t-rrs o.s*z'o 

El - r.,r-rts z. s*s. o 

ffi r.r-rrs o,o-.r.o 

Itt' 0 0.5 

-2,t 

I 

[6 l{l 

Irr.rr, l0r, 0lr. tlr, .r., 

Atr.*rj lilr lilrlr rrl 

rlir rrrrira. lrll 

t t 

-i.0 

t0 

tr 

6.0 

*t.! 

tr 

tr 

5it I. tl) 

15t ilt. en 

t0t L tD 

il5 t010 ttt ttr 2t5il t00n 

Position of accidents subject to case studies 

Statistic $urveys have been also carried out since 1986 for 

all traffic accidents occurred in the specific area during the 

investigation period, for the clear determination of parameter 

for the selected of accidents subject to case studies. 

The total number of accidents covered by the statistic 

surveys in three years up to 1988 (total duration: l2 months) 

is 12,647. 

Vehicle to vehicle collision and single vehicle accidents, 

which are main targets of investigation, amount to 5,929 of 

which 198 accidents have been subject to ca$e studies. 

Comparison of severity of injury between statistic surveys 

and case studies shown following results. Statistic 

surveys yielded ll9 fatalities (Z.OVo),795 severe injuries 

(l3.4Vo) and 5,015 minor injuries (84.67o). According to 

case studies, on the other hand, fatalities are 80 cases 

(4O.4Vo), severe injuries 90 cases (45.SVo) and minor injuries 

28 cases (14.l7o), with higher ratios of fatalities and 

severe injuries than the statistic surveys. 

Comparison of vehicle running speeds immediately before 

collisions between statistic surveys and case studies 

indicates that case studies show higher speeds in terms of 

cumulative incidence rates, as shown in figure 5. It is also 

found from the classification by type of accidents that the 

speeds are generally higher for case studies as shown in 

figures 6 to 9. According to the comparison of the speeds by 

the 50o/o line of cumulative incidence rates, single vehicle 

accidents show the highest speed range. Likewise, it is also 

found by looking at the relationship between the severity of 

occupant injury and the speed immediately before collision 

that the speeds of case studies are generally higher for 

severe and minor injuries than the statistic surveys, though 

the $ame tendency is observed between them for fatalities, 

as most of them covered by the statistic surveys are also 

included in the case studies as shown in figures l0 to 12. 

s 

2 500 

?000 

ffiffiffiH\*^ 

200 roo il 

1 500 

I 000 

500 

4 150 5 

l 

E tn 

I 

z 

VAsraLisric 

+ survey 

! crsc 

+ 3tu(ty 

i0 lrt 50 6d ti irt q0 I00 I l0 120 130 lao 

Speed (lu/h) 

Floure 5. Numbers of vehlclee and cumulativg incidence rate$ 

cldssified by vehicle speeds lmmedlately before colllglon. 

q 

G 

0 

0 

o 

5 0 E E 

o 

l 

d 

I 

0 o

400 

o 

i r00 

E 0l 

E 2oo 

E 

0 

i 

E too 

Head-on colllslon 

@*ristic 

+ SUtrct 

I cr* 

4F stud, 

t0 a0 50 60 70 E0 90 lo0 ilo 120 ljo lto 

Speed (ltr/h) 

I' 

0 0 ; 

0 

d 

h 

o 

g 

o 

E 

5 0 t 

o 

d 

t 

0 " 

Flgurc 6. Numb€ra ol vehlclea and cumulative incidencc ratas 

claaslfled by types of accldente and vehlcle speeda lmmedlately 

before colllslon. 

n 

500 n H n 

*de nffi ffiHn n 

colrrtroh 

?oo roo F 

1 000 

.9 rso 

o 

* 100 

E r o 

Flgure 7. Numbere ot yehlcle$ and cumulatlve lncldence rates 

classlfled by type$ of accldsnts and vehlcle apeeda lmmedlately 

before colllsion. 

o 

E 

I 

E 

2 500 

2000 

1 500 

1 000 

500 

150 

100 

strucl 

Ych ic lc 

lu !0 50 6[ 70 80 90 100 ll0 120 l]D 140 

H A 

Speed (krlh) 

3tl rklnt 

rchiclc 

Flgurc 8. Numbert ol vehlclee and cumulatlve Incidcnco ratos 

clagalfled by types ot accldcnts and vahlcle apeeda lmmediat6ly 

before colllslon. 

6 

o 

fu :r 

Relr-€nd colltsloo 

o o F 

@**iatic 

+ $rttt 

E *tc 

+ rtrdt 

0 l0 20 t0 a0 t0 60 70 8o 90 100 rrd lto rI0 lr0 

SDe6d (lclh) 

50 

d 

o 

o 

-E 

o 

n 

5 

I 

0 a 

o 

6 

p 

I 

a 

q A 

q 

0 6 

r A 

o 

I 

Slngle vehlcle lccldent 

@s[uslrc 

+ Sufvey 

rl lu ?0 l0 .0 50 60 ?r) E0 90 I00 lt0 120 ll0 ll0 

Speed (!u/h) 

loo F 

o 

! 

0 

0 

5 0 E 

E 

o 

G 

I 

0 u 

Flgure 9. Numbers of vehlcles and cumulatlve lncldence rstes 

clasalfled by types of accldents and vehlcle speeds lmme. 

dlately before colllslon. 

E c o 

6 

o 

! 

3 3 0 

I r o 

I 

E to 

Fetal I ty 

0 l0 20 lo a0 t0 60 70 E0 90 l0o ll0 120 130 140 

Speed (kr,/h) 

roo t 

Flgure 10. Numbers of occupEnls and cumulatlve Incldence 

rates claseiflsd by E€vsrltl€a of Inlurlee and vehlcle epeeds 

lmmedlatoly belore colllslon. 

E 

d 

r 

5 

g 

o 

I 

?00 

150 

100 

50 

severe lnJury 

o l0 ?0 30 a0 50 60 r0 80 90 100 tto t?0 t30 lao 

speed (kr/h) 

Flgure 11. Numberg of occupants and cumulatlvs lncidence 

rates cle8slfied by E€varltl€a of Inlurlea and vehlcle apeedr 

lmm6dlat6ly b€fors colllElon. 

It is found from foregoing data that fatalities and severe 

injuries are the majority of accidents subject to case studies 

and that the speeds immediately before collision are also 

higher. 

0 

€ 

o 

o 

qr1 E U 

o 

F 

I 

! 

0 u 

100 g 

o 

0 

G 

0 

o 

c 

e 

5 0 t 

E 

o 

! 

a 

I 

J 

o '

E rro 

! 

c 

E 

j roo 

1 1 0 

- 

na 

NHH 

illnor lnjury ^ 100 r 

@stattsltc 

+ surv€y 

! crsc 

+ ttudy 

,l lrl :c J0 40 50 60 70 80 90 100 ll0 I20 130 110 

Speed (krlh) 

Floure 12. Numbsrs of occupanls and cumulatlve Incidence 

rafes claesltled by g€verltlc+i ol iniuriee and vehlcle rpeeds 

lmmedlately belore colllslon. 

Analysis of Accidents Involving 

Passenger Cars 

Of all vehicles involved in accidents, passenger cars 

(limited to bonnet-type cars) with the highest vehicle ratio 

and relatively small differences in vehicle construction and 

weigh were analyzed in terms of the principal direction of 

impact force, general area of damage, relationship between 

the impact speed (equivalent barrier speed) and the severity 

of injury, etc. 

Principal direction of impact force and 

number of vehicles classified by equivalent 

barrier $peed 

The distribution of speeds of vehicles subject to case 

studies is studied according to the classification by the principal 

direction of impact force. It is found from figure 13 

that speeds are slightly higher for side collisions than those 

of frontal collisions in terms of equivalent banier speeds 

(values obtain by calculating energy absorption rated from 

o 

E 

I t- 

Floure 13. Numberr of yrhlclss and cumulatlve Incldcnce 

rsfcs clrttlfled by prlnclpal dlrectlong ol lmpact lorce and 

equlvElsnt bsrrlar sprcdS. 

4M 

10 

I 

8 

o1 

ol 

o-j 

70t 

o 

L 

I 

ol 

10 

o 

ffiFtort 

! siac 

ffilar 

+ trort 

+ Stdc 

+F lcr r 

10 20 30 40 50 60 70 80 90 100 

Equlvalent berrler epeed (krlh) 

E 

o 

d 

0 

U 

o 

E 

100 ; 

E 

e 

! 

I 

a 

o 

6 

0 a 

E 

o 

q 

I 

0 u 

amount$ of crushes of vehicles and converting the values 

into barrier impact test speeds), and the speeds of rear-end 

collisions are lower than others. The speeds on the 507o line 

of cumulative incidence rates is 25km/h for side collisions. 

2lkm/h for frontal collisions and lOkm/h for rear-end 

collisions. 

Severities of occupant injuries classified by 

principal direction of impact force and 

equivalent barrier speed$ 

Injuries ofoccupants involved in accidents are compared 

according to classifications of principal direction of impact 

force and equivalent barrier speeds. Severities of occupant 

injuries are classified by J-AIS (compared with AIS: J-AIS 

0.5 and 1.0 are equivalent to AIS l; likewise, J-AIS 1.5 and 

2.0 = AIS 2; J-AIS 2.5 and 3.0 = AIS 3; J-AIS 6.0 through 

9.0 are for the classification of fatalities by hours, which are 

equivalent to AIS 6), and the severity of the maximum 

injury of each occupant is expressed as "M.J-AIS". 

Cumulative incidence rates of the maximum severity of 

occupant injuries are classified into fatalities (M.J-AIS 6.0 

through 9.0), severe injuries (M.J-AIS 2.5 through 5.0), 

minor injuries (M.J-AIS 0.5 through 2.0) and no-injury 

(M.J-Ar S 0). 

The data compared by the principal direction of impact 

force are shown in figures 14 to 16. Although this compari- 

d 

e 

o 

6 

I 

O 

90 

80 

70 

60 

50 

40 

30 

?0 

l0 

0 

I{urbcr 

of 

occultnta 

+il.J-AIS 0 204 

* il.J-AIS 0,5-2. 0 383 

+il.J-AIS 2.5-5. 0 114 

+il.J-AIS 6,0-9. 0 5 9 

0 lil 20 l0 {0 50 60 ?0 80 90 100 

Equlvalent barrler epeed (krlh) 

Flgure 14. Cumulative incldence rat€l of occuPant Inlurles involved 

In lrontsl colll$lons classlfled by equlvalent barrler 

sp€eds and aeverltles of lnluries. 

fl 

6 

o 

d 

s 

E 

o 

E 

g 

E 

tr 

G 

)I 

3 

90 

80 

70 

60 

50 

40 

JO 

?o 

10 

0 

+t.J-Ars 0 

l||bcf 

of 

ffiGUDtnrl 

60 

+t,J-AIS 0,5-2. 0 r0r 

+il.J*AIS 2,5-S. 0 6 3 

+ ti, J-Ats 6,0-9. 0 { 2 

0 l0 e0 50 .0 50 60 t0 80 90 100 

Equlvrlent barrler epeed (kr/h) 

Figure 15. Cumulatlve Incldencs rates of occupant lnlurlea lnvolved 

In elds collaslons classlfled byequivaltntberrlsr speedg 

and severlllea of lnlurler.

e 

6 

o 

€ 

I 

a 

u 

rQo 

90 

EO 

70 

60 

50 

40 

JO 

20 

10 

0 

l[bcr 

of 

trcuFrltE 

+ll.J-AI$ 0 5l 

* ll,J-AI9 0.$-2.0 lrg 

+ il,J-AIS ?,5-5,0 3 

+ll,J-AIS 6,0-9,0 6 

l0 20 l0 .0 50 60 70 E0 90 I00 

Equlval6nt barrter Epeed (kr/h) 

Flgure 16. Cumulatlve Incldence ratgs ol occuofnt Inlurle$ Involved 

In rear.end colllslons classllled by sqLlvaledt berrler 

spasda and aavsrltlea of Inlurler. 

son shows similar tendencies of injury between frontal and 

side collisions, injuries occur from relatively low impact 

speeds for rear-end collisions but speeds to have caused 

fatalities are relatively high. It should be noted, however, 

that this tendency is not clearly determined, as the number 

of occupants involved in fatal accidents is small. 

Comparison of severities of occupant injuries 

between those using sert helts and those 

not using them 

The mandatory use of seat belt was reinforced since November 

1986 in Japan, on the whole, the rate of belt use (it 

wzs 92Vo for drivers involved in vehicle to vehicle and 

single vehicle accidents according to the national statistics 

of 1987) has been improved to have contributed to the 

reduction of injuries in accidents. Using rates in case studies 

are, however, generally quite low, and it was as low as 237o 

for drivers involved in accidents subject to investigation in 

1988. This low using rate is presumably due to rhe facr that 

cases with reduced injuries owing to seat b€lt use normally 

would not be included in case $tudies, a$ severe injuries and 

fatalitie$ are main objects of case studies, which resulted in 

the seemingly low using rate. Effects of seat belts on the 

reduction of injuries are observed here, where front seat 

occupants are considered as the object of analysis. The 

o 

H ' 

- 

I 

t 

0 

B 3 

d 

5 2 

= . 

E 

0 

""" 

o o o 0 6 

"Affi 

d866"; 

OeOOc o o o a 

0 l0 20 30 .0 50 60 ?0 80 90 l0o 

Equlvalent barrler speed (krlh) 

Be I ted 

I 

A 5 

A ro 

Flgurs 17. Soverltlee ol occupant Inlurles Involved In lrontll 

colllslons clsrslflrd by cqulvdlcnt bairlsr apoodr. 

number ofoccupants subject to the analysis is 78 for those 

who use seat belts (3 or more point belts) and 505 for those 

who do not use seat belts. 

Figure l7 shows the relationship between the severity of 

the maximum injury and the equivalent barrier speed. Effects 

of seat belt use were not recognized in this figure. 

Thus, occupants with negative factors to have reduced the 

effect of seat belts are excluded from those who use seat 

belts. Namely, occupants of vehicles collided by trucks with 

large amounts of intrusions (30cm or more) into the vehicle 

compartments, causing underride, or where rear seat occupants 

not use seat belts pushed front seat occupants through 

seat backs, etc. are excluded for the purpose ofcomparison. 

Table 2 shows the breakdown of occupants who are 

deemed to have such negative factors. 

Table 2. Number of occupantr lor whlch thc sfiect of teat bolt 

u$o on tho reductlon ol Inlury |3 prerumably reduced. 

[ause of prcsufiably rcduccd 0r i ver Co-dr i ver Totr I 

Largc ailount of intrusron (30cr or nore) l 3 l7 

Underr rde col I rsron I I 

llfong use 2 

lnflucnce of unbeltcd rerr se!t occuprnt 2 4 

Torn I L O o 32 

According to this comparison, the effect of seat belts can 

be recognized, as the injury severity is lower for those using 

seat belts than those not using seat belts, as shown in figure 

18. 

ra 

- 

3 

I 

t) 

l r 

€ 

1 2 

J 

E 

0 

a "..' 

f i l a 

0 l0 20 30 r0 t0 60 ?0 80 90 100 

Equlvdlent bsrrler Bpeed (krlh) 

Unbe 

o 

c.} 

\ , 

Ited 

I 

l0 

20 

Flgure 18. Ssvorltlor of occupanl Inlurlcr Involvsd In lrontal 

collltlonr clamllled by equlvalent balrler rpesdr. 

A comparison of cumulative incidence rates is also made 

between occupants using seat belts and those not using 

them, according to the classification by severities of injuries. 

Figure l9 shows cumulative incidence rates of occupants 

using seat belts, while figure 20 shows the rates of 

those not using seat belts. It is found from these figures that 

equivalent barrier speeds tend to be higher for those using 

seat b€lts in each category of injury severity. 

50 

455

a 

0 

0 

0 

a 

o 

6 

I 

O 

ool 

rol 

uoJ 

701 

uol 

50-l 

4 0'i 

,'ol 

20 1 

':l 

Belted 

liurber 

OI 

occupants 

+!f.J-AIS 0 22 

+ll.J-AIS 0,5-2,0 28 

++lq,J-AIS ?.5-5.0 16 

+lt.J-Ars 6.0-9,0 t2 

! I j ;n li 40 50 b0 70 irl ttr 100 

Equlvalent barrler speed (krlh) 

Flgure 19. Cumulatlve lncldence rstes of occupant Injurlet Involvsd 

In frontal colllslons clas$lflcd by equlvalenl barrler 

speeds and severllles of Inlurler. 

This difference is particularly significant for cases of no- 

injury (M.J-AIS 0) and minor injuries (M.J-AIS 0.5-2.0), 

which may be also considered as an effect on the injury 

reduction by use of seat belts. 

As have been compared so far, the effect of seat belt use 

on the injury reduction is recognized, but occupant injuries 

caused by the improper use of seat belts or injuries of front 

seat occupants resulted from the pushing by rear seat occupants 

not using seat belts are also found recently. Activity to 

prompt on promote the proper use of seat belts and the 

development of a more comfortable seat belt system are 

considered necessary, since such problems are likely to 

increase along with the increasing use of seat belts. 

+t 

E : area of slde slII lower Plane 

to the slndow frane lorer Plane 

ril 

-fit i 

l l r l 

lli 

J( 

il: 

M : 

* 

o^ 

$ro 

Ito 

Hoo 

0 

E s o 

5 + o 

g J o 

8 2 0 

i 1 0 

6 o 

0 1 0 2 0 1 0 4 0 5 0 6 0 7 0 8 0 

Equlvelent berrler speed (krlh) 

Unbelted 

l{u[ber 

of 

0ccuPants 

* rrr.J-AIS 0 135 

+ ll|.J-AIS 0,5-2.0 25r 

-* Irl,J-Ars 2.5-5.0 71 

+ ltl.J-AIS 6.0-9.0 42 

Flgure 20. Cumulative incidenc€ rates ot occupant lniurles ln. 

volved In frontal colllalona claeelfled by €qulvalent barrier 

eposde and severltles ol injuriet. 

Side collision accidents 

The number of vehicle collided sideward by other vehicles 

or objects (CDC codes; 01, 0?,03, 04, 05R; 07, 08,09, 

10, lll-) and subjected to investigation classified by the 

principal direction of impact force and general area of damage 

is 192 (right-side collisions: 92 vehicles; left-side collisions: 

100 vehicles) as shown in figure 21. As for the principal 

direction of impact force, direction of 02 o'clock and l0 

o'clock from the forward direction are most frequently 

found, including directions of 03 o'clock, while collisions 

from the backward direction are few. By looking at intruded 

vehicles classified as E or A according to the CDC (E: area 

of side sill lowerplane to the window frame lowerplane; A: 

area of side sill lower plane to the roof top), intrusions into 

.!3tl 

presence 

of lntruslon 

related vehicle 

A : area of side slll lower plane 

to the roof top 

ffiH 

Flgure 21. Numbers of vchlcles classlfled by prlnclpal dhectlons of lmpsct force, general arGas of damage$, and absenc€/presence 

of lntruslon Into passengsr compartment. 

466

their passenger compafrments are found in 103 vehicles out 

of 149 vehicles classified as E, and 42 vehicles ofout of43 

vehicles classified as A. As forthe general area of damage, P 

(middle), F (front) and Y (front + middle) account for the 

majority, with great amounts and rates of intrusions for P 

and Y in particular. The most frequently found principal 

direction of impact force into the passenger compartment is 

09 o'clock. 

Relationship between equivalent berrier 

speed and maximum intrusion into passenger 

compflrtment 

The relationship between the equivalent barrier speed 

and the maximum intrusion into passenger compartment in 

vehicle to vehicle and single vehicle accidents is shown in 

f,tgure 22. Vehicles collided from direction of 0t,02,03,09, 

l0 or I I o'clock, with the general area of damage in P, D 

(entire side plane), Y, Z (middle + back) and the range of E, 

A or H (bumper upper plane to the roof top) are covered 

here. It is found from this figure that some correlation 

exists between the equivalent barrier speed and the maximum 

intrusion into passenger compartment$, but significant 

variations are also found from single accident vehicles. 

Therefore, it was decided to limit objects of the analysis 

to vehicle to vehicle collisions, and the relationship mentioned 

above was analyzed again. 

The relationship between the equivalent barrier speed 

and the maximum intrusion into passenger compartments 

according to the classification by the general area of damage 

was thus re-evaluated. Namely, general areas of damages 

are classified into R D (E, A, H) and Y, Z (E, A, H), and 

principal directions of impact force are classified into three 

categories of 03,09 o'clocks, 0l,l I o'clocks and 02,10 

o'clocks. Figure 23 shows vehicles with the general area of 

damage in the zone of P, D while figure 24 shows vehicles 

with the general area of damage in the Y,Z zone, with each 

of them classified by the principal direction of impact force. 

; 

150 

It0 

130 

r?0 

..9 tto 

g 100 

9 s o 

E ro 

* 

Eo 

I 

5 

. - 

JU 

! , 0 

f 3 0 

?0 

l0 

0 

DuatE arcr: P. tt. Y, z (8, A, ll) 

Dlrectlon: 01. 02. 03, 09, 10, ll 

O c 

o o o A o 

" g8 4 

:lir:a 

iUE o - 

o o 

o 

0 l0 20 30 .0 50 60 t0 80 90 100 

Equlvalent barrler epeed (k|/h) 

vchlclc to vGhlclt 

coII l3 lon 

o l 

o l 

o 5 

Slntlr vchlclc 

ecldGnt 

a l 

4 ? 

A 3 

Flgure 2e. texlmum Intrurlonr Into pessenger compartmGntr 

In $ldt colllslons clscalflcd by squlvalent barrler spcrds (vehlcle 

to vehlclo colllslons and rlngle vehlcle accldents). 

150 

l a0 

t30 

.9 rro 

d 100 

q 90 

i 8 0 

! s o 

I 

r l0 

4 3 0 

?0 

l0 

0 

150 

la0 

| 30 

'* 1?O 

I ' - - 

9 rro 

E 100 

3 g o 

5 so 

h 

E ? 0 

- 60 

!l so 

! r o 

s 3 0 

t0 

DilidE arer: P, D (E. A. Hl 

Dlrectlon:03,09 

0 l0 20 30 10 50 60 10 80 90 100 

Equlvalent barrler epeed (krlh) 

DU|l $c6: I, Z (E, A, [) 

Dlr*tlon:03,09 

01, lI 

0e. l0 

0 l0 20 t0 r0 50 60 ?0 80 90 100 

Equlvalent blrrler apeed (kr,/h) 

- 0s, 09 

o l 

d 2 

o 5 

Flgure 24. Maxlmum Intruglong Into paraengcr compartmsnta 

In slde colllsions classlfled by oqulvalcnt barrler apeeda (vehlcle 

to vehlcle colllrlons). 

Comparison of both figures shows that the intrusion into the 

P, D zone tends to be slightly greater than that of the Y, Z 

zone, and the maximum intrusion into the passenger compartment 

is slightly greater where the principal direction of 

impact force is in the range of 03,09 o'clocks, while it is 

smaller where it is within the range of 0l,I I o'clocks. This 

tendency is more significant in the Y, Z zone. As described 

so far, some correlation is found between the equivalent 

barrier speed and the maximum intrusion into the passenger 

compartment by the classification of principal direction of 

impact force as mentioned above. 

Injuries of vehicle occupants 

- 03. 09 

o l 

o 2 

Flgure 23. Maxlmum Intruslons Into psrasng€r compErtmenls 

ln elde colllelons classllled by equlvalcnt barrlrr spaadr (vehlcle 

to vehlcle colllalona). 

Table 3 and table 4 show numbers of injuries as classified 

by damageable parts and components against front seat 

occupants and regions of injuries. As regards damageable 

parts and components against the occupants, those at both 

sides of passenger compartments account for the highest 

rate of 42.8Vo in vehicle to vehicle collisions, while they 

account for 37.SVo in single vehicle accidents, and objects 

out$ide vehicles are direct causes ofdamages in other cases. 

As for the classification by regions of injuries, head, neck 

and chest account forhigherrates in vehicle to vehicle colli- 

467

sions, while head accounts for the highest rate in single 

vehicle accidents. Damageable parts or components caused 

relatively severe injuries are door panels, A-pillars and vehicle 

interior deformations and exterior parts/components 

of the other vehicle in vehicle to vehicle collisions, while 

A-pillars, roofside rails and objects outside vehicle in single 

vehicle accidents. 

Relationships among the severity of the maximum injury 

of occupants, equivalent barrier speed and the maximum 

intrusion into the passenger compartment were studied 

next, by classifying general area of damages into P, D, Y and 

Z (E, A, H) and principal direction of impact force into two 

categories of 01, 02, 03 o'clocks and 09, 10, 1 I o'clocks, 

with front seat occupants as the objects of analysis. Figure 

25 and 26 show the relationship between the equivalent 

barrier speed and the severity ofoccupant injury classified 

by the direction of impact and the location of sitting. Although 

a clear tendency cannot be found due to great variations 

among individual vehicles, it is found that severity of 

injuries in struck side occupants are generally severe than 

those of the opposite side occupants, and for the occupants 

of left seats in particular where the principal direction of 

impact force is with the zone of 09, 10, ll o'clocks' The 

occuffence of injuries are found for both right and left seat 

occupants where the equivalent barrier speed is l0km/h or 

higher and fatalities are found where the equivalent barrier 

speed is 35km/h or higher for both right and left seat occupants, 

where the principal direction of impact force is in the 

zone of 01, 02, 03 o'clocks. In case where the principal 

direction of impact force is in the zone of 09, 10, I I 

o'clocks, injuries are found for both left and right seat 

occupants where the equivalent barrier speed is l5kmft or 

higher and fatalities are found where the equivalent barrier 

speed is 25km/h or higher for left seat occupants and 45km/ 

h or higher for right seat occupants. Likewise, the relationship 

between the maximum intrusion into the Passenger 

compartment and the severity of occupant injury has been 

$tudied. It is found as shown in figure 27 and 28 that variations 

among individual vehicle are even greater than the 

above. The occurrence of injuries are found where the 

amount of intrusion is Ocm or more for left seat occupants, 

and 5cm or more for right seat occuPants, while the occurrence 

of fatalities are found where the intrusion is 35cm or 

more for both right and left seat occupants, where the principal 

direction of impact force is in the zone of 01, 02, 03 

o'clocks. In case where the principal direction of impact 

force is in the zone 09, 10, I I o'clocks, injuries are found 

from the Ocm intrusion for both left and right seat occupants, 

while the fatalities are found from the intrusion of 20cm for 

Table 3. Numbsrs of Inlurlet ol occupants Involved In slde colllslons classifled by damageable parts, reglons and aeverltles of 

Inlurlee (vehlcle to vehlcle colllrlona). 

t. 

lj 

hrrlr ldrr.rurill 

I lrrsr 

tli 

I t,rl lrt 

{.r'..0r orlil '.1'.1 

crrrrErl r.tror Elrr. I'rdr. rM*i 1 rDo.r Jr.br 

1[ttllrr1010t0 

i 

ol'." 0 r.5 t0 t! I { (q 5 0 r 0 | 0 e 0.1'll,l r.o r.: ro rt ro.'jj 

I 

I t 

! ? ! rrr0r0,.jie.ir I r 0 r r ro !.0 | oro t 1! rr ro r.0.1'll,l ? 0 | 0.1".0. i | 0 il r[:l 

t ? t t l ? i l 

I l ' 

lfr flFt7 ! r r r rFral lm a ! e i l r r f r 

r{[ti xr*rrt u-rr]r I E 

Tgble 4. Numbers ot hlurles of occupants Involved ln elde colllslons claaallled by damageable parts, reglont snd seYsritie$ of 

inlurles (slngle vehlcle accldents). 

- I rrnd8hrcld 

E I insrurelr gutl 

"" 

Slrli:",l,'iT'' 

" "lrmf 

I *rdt-rrrl 

! | ourirda ottftr 

= | rqo EurrrcG 

468 

t,0 l.tr 10 z5 1.0 {.0 5.0 ? 0 9 0 4'lt 

1 1 | 

l t z 

t l 

t r i 

l l l 3 6 

2 l l l 

2 2 l 1 6 

t 1 1 5 

t l 

l t 

l r 

2 a 6 2 9 l l ? l 

ctrYrcrl rrlr0n 

?lll|lI'en* trrbrl tn* tr.br 

*.1f0-,j ,,, t0 r.0 t0 s.0 I o s.o,11l,l r.o e0,lll,l I.s Lo to r.o,l'l!,r 5 J.0 r, 0t;i; nru!lr rlnitolr^r(r 

l t l r S 

l l l S 

I t 2 2 t t 5 l l t 2 r 2 1 

2 2l r r 2

left seat occupants and from the 4thm intrusion for right 

seat occupants. 

h 

s 

I 

T n 

- - 

i r 

o 

l'2 

E r 

0 

DuatE $cr: P. It, Y, Z (E, A, H) 

Dtrcctlon:0L 02,03 

0 t0 ?0 30 a0 50 60 ?0 80 90 100 

Equtvalent berrler epeed (krlh) 

Rltfit lrotrt 

trauFfit 

o l 

o 2 

U N 

Oro 

Lcft frort 

ftcuDmt 

a l 

A 2 

A 5 

Figure ?5. Saverltlee of Inlurlor of occupsnt8 Involved In slde 

colllslon$ clesslflrd by cgulvalsnt barrlsr apeede (vehlcle to 

vehlcle colllslon*). 

0 

I 

6 

H ? 

I 

? r 

- 

E l 

0 

B s 

6 

a l 

) 

0 

Dutf, rr€r: P, D, Y, Z (E. A' ll) 

Ittrectlon:09,10.1f 

A 

a . A a 

o g q 

o € o O 

A o . o O s 

a O o 

l0 t0 t0 a0 50 50 ?0 80 90 100 

Equlvelent birrler epeed (kr,/h) 

Rldrt lront 

occupmt 

o l 

6 2 

o 5 

Olo 

trlt front 

occuP6ht 

a l 

A ? 

A 5 

Figure 26. $avsrltles of Inlurlea of occupanla Involved In slde 

collislons classlflad by equlvalent barrler apeeds (vehlcle to 

Yehlc16 colllslona). 

I 

t 

3 1 

I 

? 5 

E 

5 

h 

o 

B r 

6 

1 l 

I 

E 

0 

Itr.ar rrcr: P, lt. Y, Z (8. A, lll 

Itlrmtls: Ol,01,0t 

A O 

0 l0 t0 30 {0 50 60 ?0 N0 90l0it0ttf,tf,tot50 

llarllul lntruslon (cr) 

llttt lront 

ftcrDer 

o l 

o t 

o s 

Olo 

L.ft front 

*euFmt 

A l 

a l 

A 5 

Flgure 27. Severltles ot Inlurlcr of occupants Involved In slde 

colllslons classllled by merlmum Intruslons Into pssssnger 

compartm€nta (vehlcle to vehlcle colllslons). 

6 

- 6 

-- i 

I 

t 

0 

B 3 

s 

5 ? 

= . 

E 

0 

Durte rrci:'P. D. I, Z (E, A" H) 

Dlrectlon:09. 

r0.11 

0 l0 20 30 r0 50 60 ?0 80 s0l0ht0r2hti.0l50 

llarlrur lntru8lon (cr) 

Rlttt front 

oecuptnt 

p l 

o z 

Oto 

Latt front 

dcuDfit 

a l 

A 2 

Flgure 28. Seyerltlrs of Inlurler of occupanl8 Involved In slde 

colllslons classlfled by marlmum Intruslons Into psssongsl 

compartments (yehlcle to vehlcle colll$lon$). 

Summary 

Accident case studies being canied out at present and 

stati$tic surveys for the clear determination of parameter for 

the selected accidents have been described in this report, 

together with the outline of accidents investigation so far 

and some examples of analysis on seat belts and side 

collisions. 

The following can be pointed out by summarizing findings 

of this report. 

(l) Rates of fatalities and severe injuries, and vehicle 

speeds immediately before collision are higher in accidents 

covered by the case studies than those ofother accidents in 

general. 

(2) The distribution of number of vehicles classified by 

the principal direction of impact force and the equivalent 

barrier speed has been studied for passenger cars. It is found 

that equivalent banier speeds are slightly higher in side 

collisions than those in frontal collisions, while the speeds 

are lower in rear-end collisions. Likewise, the distribution 

of severities of occupant injuries has been studied. It is 

found that the tendency of injuries is similar between frontal 

and side collisions. while it is different for rear-end 

collisions. 

(3) The effect of rhe seat belt use on the reduction of 

injury has been compared between passenger car occupants 

use seat belts and those not use them who were involved in 

frontal collisions, according to the classification by the 

equivalent barrier speeds. As a result, the effect ofseat belt 

use on the injury reduction is recognized on comparison 

taking account of passenger companment intrusion rate, 

underride collisions into trucks and the effect on rear seat 

occupants without use seat belts. 

(4) Conelation is found between the equivalent barrier 

speed and the maximum intrusion into the passenger 

compartment in vehicle to vehicle side collisions by 

limiting the zone of principal direction of impact force and 

the general area of damage. For single vehicle accidents, 

469

however, variations among individual vehicles are great and 

clear correlation cannot be determined. This is presumable 

due to the difference in object to which each vehicle collided. 

For occupant injuries, those of oacupants in struck 

vehicle tend to become severer in general, but the tendency 

is not clear when it is considered in relation to the equivalent 

barrier speed or the maximum intrusion into the passenger 

compartment. 

Postscript 

Objectives, methodology, outline and some of analytical 

results of motor vehicle traffic accident investigations 

carried out so far under the Ministry of Transport of Japan 

have been reported. We intend to carry out more studies/ 

investigations and analysis of data in order to reflect the 

findings to motor vehicle safety standards. 

References 

(l) The Councils for Transport Technics, 

"Technical 

Measures for Safety of Motor Vehicles-First Program for 

Future Motor Vehicle Safety Standards", September, 1972 

(Japanese). 

(2) The Councils for Transport Technics, 

"Technical 

Measures for Safety of Motor Vehicles-Second Program 

Plan for Future Motor Vehicle Safety Standards", October, 

1980 (Japanese). 

(3) Land Transport Engineering Department, Regional 

Transport Bureau, Ministry of Transport, "Motor Vehicle 

Accident Investigation and Analysis Reports", 1973 to 

1988 (Japanese). 

(4) Koshiro Ono and Takashi Sato, "Review of MOT 

Vehicle Accident Investigation in Past Seven Years", 

Proceedings Ninth Intemational Technical Conference on 

Experimental Safety Vehicles, Kyoto, Japan, October 1982. 

(5) SAE Recommended Practice J2?4a, "Collision 

Deformation Classification 

", 

(6) P.L. Harms, M. Renouf, P.D. Thomas and M. 

Bradford, "Injuries to Restrained Car Occupants; What are 

the Outstanding Problems?", Proceedings Eleventh 

International Technical Conference on Experimental Safety 

Vehicles, Washington, D.C. United States, May 1987. 

(7) Dietmar Otto, Norbert Siidkamp and Hermann Appel' 

"Residual 

Injuries to Restrained Car Occupants in Frontand 

Rear-Seat Positions", Proceedings Eleventh 

Intemational Technical Conference on Experimental Safety 

Vehicles, Washington, D.C. United States, May 1987. 

(8) David C. Viano, "Limits and Challenges of Crash 

Protection", Accident Analysis and Prevention, Vol. 20, 

No.6. December 1988. 

The Crash Avoidance Rollover Study: A Database for the Investigation of Single 

Vehicle Rollover Crashes 

E,A. Harwin, Lloyd Emerp 

National Highway Traffic Safety 



Abstract 

In an effort to more fully understand the contributions of 

the vehicular, environmental and driver characteristics to 

the frequency of rollover crashes, a dedicated database with 

a broad spectrum of variables, has been developed. This 

database, denoted by the acronym CARS (Crash Avoidance 

Rollover Study;, includes data from approximately three 

thousand single vehicle rollover crashes. Vehicle types 

encompassed by this study include passenger cars, light 

trucks, utility vehictes and vans. The contents of the CARS 

database were collected over a one and a half year period by 

trained investigators from the Maryland State Police and 

supplemented by information derived from analyses by 

NHTSA staff. This database contains all levels of reported 

accident and injury severity. 


The National Highway Traffic Safety Administration 

(NHTSA) is involved in continuing research into the causes 

470 

and the effects of rollover crashes. The goal of this research 

is to answer the following questions: 

r What vehicle factors influence rollover 

propensity? 

r What environmental factors affect rollover 

frequency? 

r What driver factors are causal to rollover 

involvement? 

r What changes in vehicle design, highway design, 

and driver behavisr will reduce the risks 

associated with rollovers? 

This research ha$ taken several approaches including 

accident data analysis, development of sophisticated 

vehicle handling and rollover computer simulations and 

full-scale experimental testing. Analysis of a validated 

rollover database is considered essential to support these 

efforts, as well as aiding in the develoPment of future 

rollover countermeasures. 

Past analyses of rollover crash data, revealed the areas of 

information where existing files do not meet the needs of 

ongoing rollover research. Thus NHTSA decided to 

construct a database which could yield a maximum amount 

of information on topics such as accident scene 

characteristics, the driver condition and the precrash 

vehicle stability and handling behavior.

Existing Data Bases 

Three major sources of accident data, developed by 

NHTSA, are available to support accident analysis research. 

They are the Fatal Accident Reporting Sysrem (FARS), rhe 

National Accidenr Sampling Sy$rem (NASS) and the Crash 

Avoidance Research Datafile (CARDfile) (l).* 

FARS is an excellent source of infonnation pertaining to 

the consequences 

ofaccidents when fatalities are involved. 

FARS variables and collection criteria allow tletailed 

evaluation of topics related to vehicle crashworthiness 

characteristics. However, since FARS deals solely with the 

accidents in which a death occurs, accidents with less severe 

injury levels are not included in the database. Also, as 

Edwards (?) indicates, the lack of success in crash 

avoidance problem identification, using crashworthiness 

oriented databa$es, stems from the fundamental differences 

in the approach to accident harm mitigation. That is, the 

accident data collected by crashworthiness researchers is 

more closely related to understanding injury causes, in 

order to lessen their severity, than to accident prevention. 

NASS has an extensive variety of data variables in its 

files. Not only do the files have information pertaining to 

the consequences of rollover crashes, but the automated 

NASS files may be supplemented by examining the hard 

copy files for additional information concerning precrash 

conditions (3). These files have been used by researchers as 

the benchmark of national representativeness for police 

reported accidents. However, as NASS is a sample, it lacks 

the numbers of observations deemed appropriate for 

analyses on the vehicle make/model level. 

CARDfile is a database constructed from state police 

accident reports from five states, and was designed to 

improve the precrash information available at the vehicle 

make/model level. The file variables were selected to 

include indicators of precrash stability and control. The 

large size and representativeness of the CARDfile data files 

allows statistical analyses to be performed at the vehicle 

make/model level. However, to maximize the NHTSA 

capability to analyze single vehicle rollover accidents with 

respect to precrash handling, stability, roadway 

characteristics, and driver condilions, it was determined 

that an enhanced CARDfile style database, with a broader 

range of variables, was necessary. 

CARS Database 

Design Criteria 

NHTSA research has shown that over ninety percent of 

vehicle rollover accidents do not involve another vehicle. 

This finding, along with past analyses of single vehicle 

rollover accidents, rruggests that the medium best suited for 

rollover data analysis would be a detailed study of single 

vehicle rollover crashes. If complete, this data could 

provide an excellent opportunity to test hypotheses relating 

to the role of the vehicle, the accident scene, and the driver 

in these crashes. It was determined that a single vehicle 

rollover accident database, with emphasis on issues relating 

rNumbcrs in parenthescs designale references at cnd of paper. 

to crash avoidance, would be designed using police repoft$ 

and supplemented 

by the following $ources; 

. A trained accident reconstructionist would 

investigate and verify the original state accident 

repon. 

r The accident reconstructionist would interview 

the original accident investigating officer for 

verification of certain data and for additional 

information. 

r The reconstructionist would interview the driver 

to derive additional accident data characteristics 

such as precrash handling or road condition. 

r The vehicle registration record would be obtained 

for the rollover accident vehicle from the state 

registration file. 

r The complete driving record of the rollover 

vehicle driver would be obtained to aid in 

assessing driver related factors. 

r NHTSA analysts would provide $tabiliry and 

vehicle damage analysis results. 

The resulting CARS files were constructed from every 

available single vehicle rollover accident occurring in the 

state of Maryland from August 1987 until December of 

1988. Its files consist of approximately three thousand 

accidents for the vehicle classes of passenger car, light 

truck, rrtility vehicle, and van. Since these rollovers are of 

sufficient number, and were obtained over a continuous 

period of eighteen monrhs, CARS has rhe following 

beneficial characteristics: 

Accidents resulting in all levels of injury severity 

were obtained. 

Sufficient numbers of accidents, representing a 

current vehicle make/model population, were 

obtained. This allows for analyses to be 

performed based on specific make and models. 

A broad spectrum of accident variables were 

obtained which permit the reconstruction of a 

vehicle's probable precrash movements. 

The numbers oferrors introduced by handling and 

coding mistakes were minimized by the 

redundant collection of data and by performing 

multiple levels of data verification. 

Since CARS represent$ a continuous time period 

and a defined geographical area, the commonly 

used data normalizing metrics for exposure, such 

as vehicle registration, can be easily employed. 

Data Collection Methodology 

The Crash Avoidance Rollover Study, CARS, was 

designed to meet the above data criteria to provide an 

enhanced crash avoidance rollover database. past and 

present efforts of database collection, such as NASS, rely on 

the utilization of numerous teams of accident 

reconstructionists. It is costly, in time and available 

resources, to train sufficient numbers of teams necessarv for 

471

a project the size of CARS. A more cost effective solution 

was to mobilize the trained accident reconstructionists of 

the Maryland State Police. This elite group of officers, 

trained in accident investigation techniques, were already 

geographically placed throughout the state. This special 

team of accident reconstructioni$ts were given specific 

collection instructions and data acquisition forms. When a 

rollover, which met the above criteria, was reported, a 

member of the team was dispatched to collect the required 

data. 

Data source redundancy and data cross checking 

capability were designed into the CARS collection 

methodology. For example, the Maryland Motor Vehicle 

Accident Report gave the location and description of the 

accident site. This was validated and upgraded by on-site 

measurements and photographs of the accident scene. 

Data from multiple sources were also collected to 

facilitate reconstruction of accident precrash and causal 

factors (table l). The data sources for these elusive factors 

include the following: 

r Maryland State Motor Vehicle Accident Report 

r Driver interview by accident reconstructionist 

r Original investigating officer interview taken by 

the accident reconstructionist 

r Additional physical evidence was collected at the 

accident scene and from the crash vehicle by the 

accident reconstructionist 

Table 1. Dsta sources. 

FOLICE ACCIDENT REFOFT 

VEHCLE FEFONT 

VErcLE 

This information was then used for accident evaluation 

by NHTSA engineers. For example, the precrash behavior 

of the vehicle, before rolling, was estimated. A 

determination was made as to whether the accident vehicle 

was skidding, what type of skid occurred (such as spinning 

or sideslipping) and an assessment was made as to the 

probable cause of that skid (such as braking or steering). 

One engineer performed this evaluation to minimize 

inconsistencies. 

Data was also derived from photographs of the CARS 

vehicles. Structural damage to accident vehicles was 

evaluated by using a modified version of the SAE 

Recommended Practice (J224). This practice provides a 

basis for the classification of a vehicle's deformation level 

caused by an accident. ln the CARS database, vehicle body 

damage was coded using this SAE zero to nine scale to 

represent the extent of vehicle body deformation. Zero is 

coded for no body damage and nine is coded when body 

472 

BEOBTRATION 

ENCINEERINC EVALUATION 

INTERVIEWS:OFIVER ANO OFFCEA 

OREINAL IHVESiIdAIdO OFFEEF 

RECONSTFUCTIOffIST 

SIÎE AEOI$TFATPN FILEg 

AECONATFUCTIONET 

BTATE LECEHSINO FILES 

tust cM# oaTA AHAIYST$ 

flffig4 ENCINFFF 

RECONTBWTFNI8T 

GSNERAL AACEENT/ORIVER INFOAKATIfr 

VFHICTE MEASUREMENIg AfiD FHOTOS 

MAXE/MOOEL VEAIFEATION 

TIhE MEAEUREMEflTS AilO MODEL INFOFMATION 

MEASUNEMENTF/HOTOE 

O 

ORIV€R TICKET FECOFO 

VEHICLE OEFORMAIION EVALUATPN 

FHECAAS STABILITY ABEE9EEilT 

DHNEA, fiOÔWAY, VErcLE OSOITEH 

intrusion is complete. This coding allowed for as detailed of 

an analysis as the photographic evidence could provide. 

One team of NHTSA crashworthiness analysts performed 

this coding to maximize consistency. 

Data Quality Control 

Data quality control procedures were instituted at every 

level of the database construction process. 

The Maryland State Police selected to investigate CARS 

crashes were those who had been trained and certified by the 

State of Maryland as accident reconstructionists. These 

officers were chosen for their proven abilities and interest in 

accident reconstruction. This group was then briefed by the 

Maryland State Police project leader and NHTSA engineers 

regarding the objectives of the vehicle rollover project and 

the proper use of the data forms. All vehicle measurements, 

accident scene measurements, and accident reconstruction 

procedures were thoroughly discussed with this group. 

The next level of data quality control was intrinsic to the 

CARS data gathering design. As mentioned above, multiple 

sources of key, yet difficult to acquire, information were 

used. Data indicating such accident descriptors as driver 

physical condition, vehicle speed, and vehicle skidding 

were subject to this redundant data gathering method' The 

third level of data quality control was implemented by the 

NHTSA receiving engineer. Each accident data form was 

checked for missing or improperly entered data, and for 

completeness and appropriateness with respect to the CARS 

collection criteria. Forexample, if an incoming accident file 

lacked a driver record it would be withheld until the missing 

material could be recovered from the Maryland State Police 

computer files. 

Quality control efforts were rigorously implemented at 

the data entry level, Using separate sessions, the raw data 

was double entered. The two resulting data files were then 

subjected to a computer data comparison program where 

each data inconsistency was documented. The appropriate 

accident file was then pulled for each flagged observation, 

and, wherever possible, corrected. 

A fifth level of quality control was applied to assure a 

dependable vehicle make and model coding. The vehicle 

identification number, CARDfile output code, vehicle type 

code, and vehicle photographs were double checked for 

consistency. The verified vehicle code was then processed 

by a CARS program which was designed to separate truck, 

vans, and utility vehicles into distinct make/model groups. 

For example, the make/model category for a Ford pickup 

was broken down into Fl50's, F250's, etc. 

The most ambitious step of the data verification 

procedure used a series of computer programs, written 

expressly for CARS. These programs checked all variables 

against an acceptable, predetermined value range. When 

inconsistencie$ were reported, records would be pulled, and 

the values verified or corrected.

CARS Datahase Organization 

CARS is composed of five data files and two user files. 

The data files are entitled Accident, Vehicle, Tire. Driver. 

and Passenger. The two user files are the Vehicle Parameters 

and the Variable Library, The above files are in SAS 

(Statistical Analysis System) format and are installed on a 

VAX I l/780 mini-computer. Their configurarion, as 

outlined in table 2, allows forrapid examination of variables 

taken from one file or, for more complicated analyses, file 

merging. Examples of the use of these files, individually 

and collectively, may be found in the Analysis section of 

this paper. 

Table 2. Crash avoldance rollover study detabase. 

Accident file 

+ 

lr"*tro I 

The Accident file contains over sixty variables which 

describe such accident related factors as the roadway characteristics, 

environment. roadside attributes and the sequence 

of crash events (table 3). The variable, -tripping,' 

which estimates the rollover initiating event, proves to be 

especially beneficial in demonstrating the effects of roadway 

and roadside features on the risk of single vehicle 

rollover accidents. The variables which describe the land 

use at the crash site may be of particular research interest. 

Table 3. Accldent flle contents. 

TIME OF CFASH 

OESCRIPTOF TYPE 

SEQUENCE OF EVENTS 

$EVEBITY MEASURES 

ROADWAY DESCFIPTOHS 

HOADSIDE DE$CRIPTOHS 

WEATHEB/LIGHT 

PRECRASH MOVEMENT INDICATOFS 

LOCALE 

OTHEB 

NUMBEF OF VAHIABLES 

2 

7 

2 

17 

o 

6 

€ 

3 

t1 

Past NHTSA analyses have shown that rural environments 

are significantly related to the incidence ofrollovercrashes 

(4). 

Vehicle file 

The Vehicle file contains over sixty variables which allow 

investigation of vehicle related factors in rollover 

crashes (table 4). This file was generated using an extensive 

make/model coding and verification system (see the section 

on Data Quality Control) so that accident analyses may be 

performed on distinct make/model groups. Vehicle level 

precrash stability (the incidence of skidding, for example) 

and crash outcome (the amount of vehicle deformation, 

number of quarter tums rolled, etc.) variables allows the 

researcher to obtain insight into the entire crash accident 

scenario. This file also contains parameter measurements of 

the accident-involved vehicle. 

Table 4. Vehlcle flle contents. 

Tire file 

MAKE/MODEL IDENTIFICATION 

VEHICLF CONDITION 

MEASUFEO PARAMETER$ 

PHECFASH MOVEMENTS 

CRASH MOVEMENTS 

VEHICLE MODIFICATIONS 

VEHICLE LOADING 

MILEAGE 

OTHEF 

The Tire file contains twenty-five variables which describe 

the tire tread depth, inflation pressure, and rolling 

radius of the crash vehicles' tire (table 5). Tread depth 

provides data for assessing a tire's influence on the dynamic 

performance on wet surfaces. Tire inflation pressure is necessary 

for the analysis of limit performance of tires in cornering 

and braking maneuvers. Such tire data is also essential 

to rollover simulation validation. 

Driver file 

The Driver file is composed of more than forty variables 

related to the driver demographics, driver condition at the 

time of the crash, driving history, and any possible accidenr 

avoidance errors or attempts made by the driver (table 6). 

For example, of primary importance to the understanding 

of a driver's behavior at the time of the crash. is the use of 

alcohol. As shown by Terhune (5), in a study of injured 

4't3

Table 5. Tlre flle contenta. 

DEscRrproR TYPE I Huueen oF VARIABLE$ 

MANUFACTUHER I 4 

MoDEL NAME | 4 

OEM SIZE RATING I 4 

TFEAD DEPTH I 4 

GAUGE PRESSUHE I 4 

RoLLTNG FADrus | 4 

orHER | 1 

Table 6. Drlver flle contenla. 

DISCHIPTOF TYPE NUMBER OF VAHIABLE$ 

DFIVER CONDITION 

DRIVEF HISTOBY 

DFIVEF AVOIDANCE ATTEMPT 

DRIVEF ERFOF 

IN.JUFY SEVERITY 

EJECTION 

BESTBAINT USE 

DRIVER DEMOGRAPHIC$ 

OTHEF 

drivers, alcohol use is consistently under reported. To mitigate 

this bias, in our database, data from multiple sources of 

use indicators were included in CARS. Despite these efforts, 

the thirty percent alcohol impairment rate found in 

this file must still be considered more of a lower bound 

estimate for the actual rate of alcohol involvement in rollover 

accidents. 

Passenger files 

The Passenger file, with its ten variables, was established 

to aid in analyses directed at rollover outcome harm mitiSation. 

These variables rePort on injury serverity, age/sex 

demographics, and restraint use (table 7). 

474 

17 

q 

2 

11 

1 

2 

1 

3 

1 

Table 7. Passenger flle contsnts. 

DESCRIFTOFT TYPE I NUMBER OF VAHIABLES 

PASSENGEF INJUFY SEVEFITY I 1 

EJECTION 

FESTRAINT USE 

DEMOGRAPHICS 

OTHER 

Vehicle parameter file 

This file contains twenty-nine variables of measured, 

calculated, or identifying vehicle characteristics. The measurements 

were performed by the Vehicle Research and Test 

Center, NHTSA, using an apparatus called the Inertial Parameter 

Measurement Device (7). The file parameters include 

such items as center-of-gravity data and moment of 

inertia measurements. These parameters allow the testing of 

statistical hypotheses relating the risk ofrollover to vehicle 

geometric configuration (table 8). Currently, these measurements 

exist for approximately one hundred make and 

models by model year. 

Tsble 8. vshlcle parEm€t€r flle contents. 

DESCRIPTOB TYPE I NUMBEF OF VARIABLES 

MODEL TDENTIFICATION I 15 

MEASURED DIMENSION$ I 5 

GENTEF oF GRAVITY POSITION$ I 2 

INEFTIA MEASUHEMENTS I 3 

DERTVED RATIOS I 2 

oTHER I 2 

Library 

Attached to the CARS data files is a user format library 

file containing the definitions of the variable codes. This file 

provides the user with a formatted and therefore easily read 

output from CARS. 

Single vehicle accident database supplement 

The above CARS files are designed to be supplemented 

with a database of all single vehicle crashes. This database 

will be gathered from the Maryland Automated Accident 

files, for the same time period as CARS. It will have a 

structure similar to that of CARS. However, as this informa- 

2 

,l 

2 

3

tion will be taken from a single source, the variable roster 

will be more limited. This database may be used as a merric 

of single vehicle accident exposure. 

Analysis-Sample Results 

The CARS database files may be examined using 

traditional SAS techniques for descriptive and inferential 

analyses. Frequency distributions of variables have been 

explored in a continuing quality control effort. CARS has 

also been subjected to several preliminary exarninations to 

assess its capability to support re$earch. The following 

analysis samples demonstrate the versatility of this database 

to provide insight into the problems of rollover accidents. 

Accident file 

The relative frequency of tripping is of primary importance 

to the issue of rollover. Tripping, by way of a working 

definition, is a rollover initiating event where a force, other 

than roadway friction, is generated by a mechanical interaction 

of the vehicle with its environment. Untripped rollover 

is defined as a rollover which results solely from the interaction 

of the vehicle tires with a hard surface. Table 9 contains 

the distribution of these estimated sources of tripping. It is 

of interest to note that the untripped rollover is a relatively 

rare event (less than ten percent of the database). 

Tabfe L Trfpplng mechanlrma. 

UNTFIPPED 

MECHANISM 

PAVEMENT EDGE 

DITCH 

SOIL,/FLAT 

GUAFDFAIL,/BARRIER 

EMBANKMENT/SLOPE DOWN 

OTHER,/UNKNOWN 

PERCENT OF DATABASE 

9.8 

7.4 

3.7 

23.3 

24.O 

7.O 

14.4 

10.4 

Figure I shows the distribution of the above tripping 

mechanisms as a function of the variable .land use.' This 

variable is a description ofthe general character ofthe land 

in the vicinity of the crash site and takes on the values of 

'urban' or 'rural.' The two most notable differences 

Trlpplng mechanlsms In rollover Eccldents: by land ute. 

FJERCT N I 

30 

7i 

!llr9** 

t.o'"' sn'j' qo'L/t\'A],,qate[s.,,qopt,,,*auoilir 

uiiet'oAftiin"nut'o' ;\Hte 

i u' 

IRIPF'INTJ Mt.L I IANI.]M 

Flgure 1. Maryland GAFS Flle data. 

E 

UFBAN 

N nLrrr 

between these distributions are found for'roadside curb,' 

cited 3.5 times more often in urban settings, and 'ditch,' 

cited 2.2 times more often in rural areas. 

The CARS variables which describe features adjacent to 

the lanes of travel show a proportionately higher ratio of 

curbs in urban environments (3.6 times greater than that 

found for rural). These variables also show that a larger 

proportion of ditches are found along rural roadways (2.5 

times more often than in the urban environments), figure Z. 

PF RL]t. NT 

60 

50 

40 

/gFFertR 

oul+oer* 

Roadalde 

fsalures present ln rollover craehes. 

c'ss**o$.ute=r"oet.Ioptoo$$ t'F\ "$-** 

67.4 

20.3 

8.0 

J.E 

o.7 

URAAN u 

R(JRI N 

Flgure 2. Featurer adlacsnt lo lanee of trayel. Up to two cltscl 

P€r oD8ervatlon. taryland CAFS Flle data. 

Vehicle 

file 

Rollover analyses are frequently performed on data broken 

out by vehicle type, (forexample, passengercars, pickup 

trucks, etc.). The distribution of these designations is 

reported in table 10. 

Table 4. Vehlcle type dlstrlbutlon. 

VEHICLE TYPE PEBCENT OF DATABASE 

PASSENGEF CARS 

PICKUF TRUCKS 

UTILITY VEHICLES 

VANS 

OTHER 

The frequency distribution of the reconstructed variable, 

'Precrash 

Vehicle Dynamic Behavior,' which resulted from 

the stability analysis performed by NHTSA sraff, (rable I I ) 

reveals that the greatest proponion (sixty-eight percent) of 

precrash stability conditions involves some type of skidding. 

Thus, Table I I shows that five percent of the single 

vehicle rollovercases were initiated by a vehicle displaying 

a precrash front wheel lateral skid (understeer) type ofvehi- 

cle skidding, fifty percent displayed a precrash rearwheel 

lateral skid (oversteer) 

type ofvehicle skidding, and thideen 

percent displayed a four wheel lateral skid (neutral steer) 

type of vehicle skidding. 

475

Table 11. Precrash vehlcle dynsmlc behavior (stabllity 

condltlon). 

STABILITY CONDITION 

NO $KID INDICATED (STABLE) 

FRONT WHEEL/LATERAL (UNDEBSTEEF) 

FIEAR WHEEL LATEBAL (OVEHSTEEH) 

FOUB WHEEL LATEHAL (NEUTRAL STEEF) 

UNKNOWN 

Table 12 shows a breakdown of the above sixty-eight 

percent of skidding accidents, evaluated as to instability 

cause or instability precipitator. A noteworthy finding in 

table l2 is the discovery that fifty percent of the skidding 

type of rollover accidents were caused by the lateral force 

generated by going around a curve in the road too fast. 

Twenty-four percent of the rollover accidents were caused 

by the driver putting in a severe steering input while on a 

straight road such as passing another vehicle, The eighteen 

percent precrash skidding attributed to snow and ice includes 

accidents occurring on both curves and straight 

roads. 

Table 12. Cauee ol dynamlc ln$tablllty (up to two clted per 

case). 

CAUSE OF INSTABILITY PEFCENT OF DATABASE 

LATERAL CUBVE FORCE 

ACCELEFATION INPUT 

STEERING INPUT 

BBAKING INPUT 

STEEFING AND ACCELERATION 

STEEBING AND BRAKING 

OTHER (SNOW, ICE, ETC) 

UNKNOWN 

The distribution of vehicle body deformation damage 

areas, table 13, shows that the most frequently damaged 

portions of a vehicle involved in a rollover are the vehicle 

top and sides. What is less expected is the large (forty-eight 

percent) proportion of vehicles which have measurable 

frontal damage. 

Table 13" Vehlcle arca of deformatlon. 

52 

1 

24 

1 

18 

10 

+ AFEA OF DAMAGE PERCENT OF DATABASE 

TOP DAMAGE 

FRONT DAMAGE 

-REAB DAMA-GE 

DAMAGE 

_LEFT 

RIGHT OAMAGE 

+ MTJLTIPLE AREAS OF DAMAGE MAY BE CODED FOR EACH ROLLOVEF 

476 

48 

26 

Vehicle parameter file 

By accessing variables from the Driver, Vehicle, and Vehicle 

Parameter files, the relationships between a selection 

of vehicle geometric parameters, quarter turns rolled, mean 

driver injury severity, and estimated speed were explored. 

This was performed using nineteen make/models which had 

measurements available and were represented by a minimum 

of ten observations in the database. All values represent 

make/model based means. The estimate for driver injury 

is derived from the police reported, ordinal values of'0,' 

representing no injury, to '4,' indicating a fatality. The 

estimate for speed, is based on data from the police report, 

the investigator's report, the driver's estimate, or the speed 

limit. The variable for the quarter turns rolled estimates the 

number of quarter turns the vehicle underwent. A matrix of 

Pearson correlation coefficients and their associated P values, 

probabilities that there is no association, are reported in 

table 14. 

Table 14. Pearson correlatlon coeff lclentg. 

INJUFY r,004 

0.000 

0.578 

._ 0,0,1_.._ 

$PEED 

0.578 

0.387 

0.10 1 

1,000 "0.089 

0.0 10 0,000 

--O.r-S$ 

0.74 

FOLL 

0.367 

1.000 

WEIGHT 

10 r. 

_il 

-0,2 l8 

0.370 

o.720 0.000 

0.t0t -0.3?g 

dg19_ 0.1 75 

6FF 

t__ _ 

FOLL INEFTIA 

0.403 

0,087 

0,074 

0.785 

-0.32S 

d. r70 

0,397 

0.10 1 

0.367 

0,r22 

0.0c 

0.73 

INJUHYi MEAN DFIVEB INJUFY 

BOLL: MEAN OUAHTEF TOFNS FOLLED 

?14 

38d 

-.0t I I 

0.65? 

-0.416 

0.o77 

-0,219 

0.3€9 

0.t01 

0.E80 

0.403 

0.067 

- is;r 

0.101 

'0.3t5 0.2 14 

dtlL'__ 0.380 

r,000 -0.€08 

0.000 0.00E 

0.608 

0 006 

0.81 I 

1.0001 

0.81t 

0.0001 

t.000 

0.000 

-0.390 

oi_": 

-0.543 

0 0087 

0.074 

0,765 

0.387 

0,!?? 

r:; i 

0.€51 

0.81 I 

o.0n0r 

-0 3s0 

0.09s 

1.000 

0,000 

0.63r 

0.004 

SFEED: MEAN EETIMATEO 6PEEO 

WEIGHT: MEAN CUFB WEIGHT 

FOLL INEFTIA 

-0.3?s 

0.170 

0.083 

.o:ltt 

-0.416 

o.o77 

0.8 t? 

0.0001 

-0.583 

0.00s 

0.9s1 

0.004 

1.000 

0.000 

llll illii^*ii.'iL:"'i:'"T"?i,1""1'*Il1fl i:''ffitl'UlT!?i,'" L,NEAF a'soctA'o' 

19 PASS€NGEF CAFS, FIOKUF THUCX$, UTIIITY VEHICLES AND VANS 

MINIMUM OF 1O OBSEFVATIONS PEF MAFE/MODEL 

Using a five percent criterion that the probability of no 

association is significant, the top line of the table indicates 

that as the mean injury increases so does the average estimate 

for speed. That is, if there is more initial energy imparted 

to the rollover speed, there is an increased risk of injury. 

Several of the measured vehicle Parameters, including 

weight, wheelbase, half track to center*of-gravity height, 

and roll moment of inenia are also correlated according to 

the five percent standard. For example, the mean roll moment 

of inertia has a negative and significant association 

with the mean half track width to center-of-gravity height 

ratio (7). 

Conclusions 

Based on the following attributes, CARS is a versatile 

and cost effective medium in which researchers may gain 

insight into single vehicle rollover accidents: 

r The CARS database structure, of five data files 

and two user files, allows for ease of access within 

one file or file aggregation. 

The detailed information within CARS allows for 

the testing ofa broad range ofhypotheses relating

ESV.Twelfth.Int.Conf.Volume.ONE.PART.TWO

Create successful ePaper yourself

Delete template?

Save as template?