30.06.2013 Views

Chapter 10 - AS Nida

Chapter 10 - AS Nida

Chapter 10 - AS Nida

SHOW MORE
SHOW LESS

You also want an ePaper? Increase the reach of your titles

YUMPU automatically turns print PDFs into web optimized ePapers that Google loves.

<strong>Chapter</strong> <strong>10</strong><br />

Specification Limit<br />

Process Capability, and Gage<br />

Measurement


Specification limit<br />

• Specification limit : conformance boundary<br />

specified formal characteristic.<br />

• Type of Specification<br />

– Two-sided : upper and lower limits<br />

– One-sided : either an upper or lower limit


Statistical tolerance limits<br />

• Statistical tolerance limits: limits of the<br />

interval for which it can be stated with a<br />

given level of confidence that it contains at<br />

least a specified proportion of the<br />

population.


Setting Up Statistical Tolerance<br />

Limit<br />

• Use control chart to measure of process<br />

capability to established requirement<br />

• From observations of a given quality<br />

characteristics from one or more samples.


Statistical tolerance limits<br />

• When process is in control, data can be used to<br />

determine statistical tolerance limit.<br />

• Assume that a process involves a quality<br />

characteristic that follows a normal distribution<br />

with mean µ, and standard deviation, σ. The upper<br />

and lower natural tolerance limits or statistical<br />

tolerance limits of the process are<br />

UNTL = µ + 3σ<br />

LNTL = µ -3σ


Statistical Tolerance Limits<br />

• Page 325<br />

• Obtain a sample of n observation<br />

• Check for normality assumption<br />

• Calculate statistics parameters<br />

• Select the desired confidence level and<br />

population proportion to be cover by limits<br />

• Determine k2 • Compute limits


Types of specification conflict<br />

• Type I : distribution too wide for double<br />

specification<br />

• Type II : Distribution not centered correctly<br />

for single specification limit<br />

• Type III : Specification limits too wide for<br />

acceptable product


Type I : distribution too wide for<br />

• Possible action<br />

double specification<br />

– Changing the process<br />

– Changing the specification<br />

– Setting up an inspection/sorting operation to<br />

find, remove, or repair the units of product<br />

falling outside either specification limit<br />

– Adjusting centering of distribution to strike a<br />

balance among relative costs


Type II : Distribution not centered<br />

correctly for single specification limit<br />

• Possible action<br />

– Attempting to center the process at a value far<br />

enough from specification limit<br />

– Changing the specification limit<br />

– Setting up inspection/sorting operation to find<br />

and remove or repair the units of product that<br />

fall out side the specification limit


Type III : Specification limits too<br />

wide for acceptable product<br />

• Possible action<br />

– Conducting ore formal program of process<br />

experimentation to determine what range of<br />

value cause nonconformities<br />

– Tightening the specification limits<br />

– Establishing shop or manufacturing limits on a<br />

temporary basis until final values are<br />

determined


Process capability analysis<br />

• Process capability refers to the uniformity of the<br />

process.<br />

• Variability in the process is a measure of the<br />

uniformity of output.<br />

• Two types of variability:<br />

– Natural or inherent variability (instantaneous)<br />

– Variability over time<br />

• Process capability analysis is an engineering study to<br />

estimate process capability.<br />

• In a product characterization study, the distribution of<br />

the quality characteristic is estimated.


Major uses of data from a process<br />

capability analysis<br />

1. Predicting how well the process will hold the tolerances.<br />

2. Assisting product developers/designers in selecting or<br />

modifying a process.<br />

3. Assisting in Establishing an interval between sampling<br />

for process monitoring.<br />

4. Specifying performance requirements for new<br />

equipment.<br />

5. Selecting between competing vendors.<br />

6. Planning the sequence of production processes when<br />

there is an interactive effect of processes on tolerances<br />

7. Reducing the variability in a manufacturing process.


Techniques used in process<br />

capability analysis<br />

1. Histograms or probability plots<br />

2. Control Charts<br />

3. Designed Experiments


Process Capability Analysis Using a<br />

Histogram or a Probability Plot<br />

Using a Histogram<br />

• The histogram along with the sample mean and<br />

sample standard deviation provides information<br />

about process capability.<br />

– The process capability can be estimated as<br />

x ±<br />

– The shape of the histogram can be determined (such<br />

as if it follows a normal distribution)<br />

– Histograms provide immediate, visual impression of<br />

process performance.<br />

3 s


Probability Plotting<br />

• Probability plotting is useful for<br />

– Determining the shape of the distribution<br />

– Determining the center of the distribution<br />

– Determining the spread of the distribution.<br />

σˆ<br />

=


Probability Plotting<br />

Cautions in the use of normal probability plots<br />

• If the data do not come from the assumed<br />

distribution, inferences about process capability<br />

drawn from the plot may be in error.<br />

• Probability plotting is not an objective procedure<br />

(two analysts may arrive at different<br />

conclusions).


Process Capability Ratios<br />

Use and Interpretation of Cp • Recall<br />

USL − LSL<br />

Cp =<br />

6σ<br />

where LSL and USL are the lower and upper<br />

specification limits, respectively.


Use and Interpretation of C p<br />

The estimate of C p is given by<br />

Ĉ p<br />

=<br />

USL − LSL<br />

6σˆ<br />

Where the estimate σˆ can be calculated using the sample<br />

standard deviation, S, or R<br />

/ d<br />

2


Use and Interpretation of C p<br />

One-Sided Specifications<br />

C<br />

C<br />

pu<br />

pl<br />

USL − µ<br />

=<br />

3σ<br />

µ − LSL<br />

=<br />

3σ<br />

These indices are used for upper specification and<br />

lower specification limits, respectively


Use and Interpretation of C p<br />

Assumptions<br />

The quantities presented here (C p, C pu, C lu) have some very<br />

critical assumptions:<br />

1. The quality characteristic has a normal distribution.<br />

2. The process is in statistical control<br />

3. In the case of two-sided specifications, the process mean<br />

is centered between the lower and upper specification<br />

limits.<br />

If any of these assumptions are violated, the resulting<br />

quantities may be in error.


Process Capability Ratio an<br />

Off-Center Process<br />

• Cpdoes not take into account where the<br />

process mean is located relative to the<br />

specifications.<br />

• A process capability ratio that does take<br />

into account centering is Cpk defined as<br />

Cpk = min(Cpu, Cpl)


Normality and the Process<br />

Capability Ratio<br />

• The normal distribution of the process<br />

output is an important assumption.<br />

• If the distribution is nonnormal, Luceno<br />

(1996) introduced the index, Cpc, defined<br />

as<br />

C pc<br />

=<br />

6<br />

USL<br />

π<br />

2<br />

E<br />

−<br />

LSL<br />

X<br />

−<br />

T


More About Process Centering<br />

• Cpk should not be used alone as an<br />

measure of process centering.<br />

• Cpk depends inversely on σ and becomes<br />

large as σ approaches zero. (That is, a large<br />

value of Cpk does not necessarily reveal anything<br />

about the location of the mean in the interval<br />

(LSL, USL)


More About Process Centering<br />

• An improved capability ratio to measure process<br />

centering is Cpm .<br />

USL − LSL<br />

Cpm =<br />

6τ<br />

where τ is the squre root of expected squared<br />

deviation from target: T =½(USL+LSL),<br />

τ<br />

2<br />

=<br />

[ ( ) ] 2 2<br />

2<br />

x − T = σ + ( µ T)<br />

E −


More About Process Centering<br />

• Cpm can be rewritten another way: (same as <strong>10</strong>.8)<br />

USL − LSL<br />

Cpm<br />

=<br />

2<br />

2<br />

6 σ + ( µ − T)<br />

where<br />

=<br />

C<br />

p<br />

1+<br />

ξ<br />

2<br />

− µ<br />

ξ =<br />

σ<br />

T


More About Process Centering<br />

• A logical estimate of C pm is:<br />

where<br />

Ĉ<br />

V<br />

pm<br />

=<br />

=<br />

Ĉ<br />

1+<br />

T − x<br />

S<br />

p<br />

V<br />

2


pk<br />

Confidence Intervals and Tests<br />

on Process Capability Ratios<br />

C pk<br />

• Ĉ pk is a point estimate for the true C pk , and<br />

subject to variability. An approximate <strong>10</strong>0(1-α)<br />

percent confidence interval on C pk is<br />

⎡<br />

1 1 ⎤<br />

⎡<br />

⎢1<br />

− Zα<br />

/ 2 + ⎥ ≤ C pk ≤ Ĉ pk ⎢1<br />

+ Zα<br />

⎢<br />

2(<br />

n −1)<br />

⎣<br />

9nĈ<br />

pk ⎥⎦<br />

⎢⎣<br />

/ 2<br />

1<br />

9nĈ<br />

pk<br />

+<br />

1<br />

2(<br />

n<br />

−<br />

1)


Confidence Intervals and Tests on<br />

Process Capability Ratios<br />

Example n = 20, Ĉ pk = 1.33. An approximate 95%<br />

confidence interval on C pk is<br />

⎡ 1 1 ⎤ ⎡ 1 1 ⎤<br />

1.33 ⎢1− 1.96 + ⎥≤Cpk ≤ 1.33 ⎢1+ 1.96 + ⎥<br />

⎣ 9(20)1.33 2(19) ⎦ ⎣ 9(20)1.33 2(19) ⎦<br />

• The result is a very wide confidence interval ranging<br />

from below unity (bad) up to 1.67 (good). Very little has<br />

really been learned about actual process capability<br />

(small sample, n = 20.)


Confidence Intervals and Tests<br />

on Process Capability Ratios<br />

Cpc • Ĉpc is a point estimate for the true Cpc , and<br />

subject to variability. An approximate <strong>10</strong>0(1-α)<br />

percent confidence interval on Cpc is<br />

where<br />

1+<br />

t<br />

Ĉ<br />

α<br />

, n −1<br />

2<br />

pc<br />

⎡ s<br />

⎢<br />

⎣ c<br />

c<br />

c<br />

n<br />

=<br />

⎤<br />

⎥<br />

⎦<br />

1<br />

n<br />

≤<br />

n<br />

C<br />

pc<br />

≤<br />

1−<br />

t<br />

∑ x − T<br />

i=<br />

1<br />

i<br />

Ĉ<br />

α<br />

, n −1<br />

2<br />

pc<br />

⎡ s<br />

⎢<br />

⎣ c<br />

c<br />

n<br />

⎤<br />

⎥<br />


Process Capability Analysis<br />

Using a Control Chart<br />

• If a process exhibits statistical control, then the<br />

process capability analysis can be conducted.<br />

• A process can exhibit statistical control, but may<br />

not be capable.<br />

• PCRs can be calculated using the process mean<br />

and process standard deviation estimates.


Gage and Measurement<br />

System Capability Studies<br />

Control Charts and Tabular Methods<br />

• Two portions of total variability:<br />

– product variability which is that variability<br />

that is inherent to the product itself<br />

– gage variability or measurement variability<br />

which is the variability due to measurement<br />

error<br />

σ<br />

= σ + σ<br />

2<br />

Total<br />

2<br />

product<br />

2<br />

gage


Control Charts and Tabular<br />

Methods<br />

X and R Charts<br />

• The variability seen on the Xchart<br />

can be<br />

interpreted as that due to the ability of the gage<br />

to distinguish between units of the product<br />

• The variability seen on the R chart can be<br />

interpreted as the variability due to operator.


Control Charts and Tabular<br />

Methods<br />

Precision to Tolerance (P/T) Ratio<br />

• An estimate of the standard deviation for<br />

measurement error is<br />

σˆ<br />

=<br />

R<br />

gage<br />

d2<br />

• The P/T ratio is<br />

6σˆ<br />

gage<br />

P/<br />

T =<br />

USL−<br />

LSL


Control Charts and Tabular<br />

Methods<br />

• Total variability can be estimated using<br />

the sample variance. An estimate of<br />

product variability can be found using<br />

σ<br />

S<br />

2<br />

Total<br />

2<br />

σˆ<br />

=<br />

= σˆ<br />

2<br />

product<br />

σ<br />

2<br />

product<br />

= S<br />

2<br />

product<br />

2<br />

+ σˆ<br />

−σˆ<br />

+ σ<br />

2<br />

gage<br />

2<br />

gage<br />

2<br />

gage


Control Charts and Tabular<br />

Methods<br />

Percentage of Product Characteristic Variability<br />

• A statistic for process variability that does not<br />

depend on the specifications limits is the<br />

percentage of product characteristic variability:<br />

σˆ<br />

σˆ<br />

gage ×<br />

product<br />

<strong>10</strong>0


Control Charts and Tabular<br />

Methods<br />

Gage R&R Studies<br />

• Gage repeatability and reproducibility (R&R)<br />

studies involve breaking the total gage<br />

variability into two portions:<br />

– repeatability which is the basic inherent<br />

precision of the gage<br />

– reproducibility is the variability due to<br />

different operators using the gage.


Control Charts and Tabular<br />

Methods<br />

Gage R&R Studies<br />

• Gage variability can be broken down as<br />

σ<br />

2<br />

measurement<br />

error<br />

=<br />

σ<br />

2<br />

gage<br />

2<br />

reproducibility<br />

2<br />

repeatability<br />

• More than one operator (or different conditions)<br />

would be needed to conduct the gage R&R<br />

study.<br />

=<br />

σ<br />

+<br />

σ


Control Charts and Tabular<br />

Methods<br />

Statistics for Gage R&R Studies (The Tabular<br />

Method)<br />

• Say there are p operators in the study<br />

• The standard deviation due to repeatability can be found<br />

as<br />

R<br />

σˆ<br />

repeatabil ity =<br />

d<br />

where<br />

R =<br />

R<br />

1<br />

+ R<br />

2<br />

and d 2 is based on the # of observations per part per<br />

operator.<br />

p<br />

2<br />

+ +<br />

R<br />

p


Control Charts and Tabular<br />

Methods<br />

Statistics for Gage R&R Studies (the Tabular<br />

Method)<br />

• The standard deviation for reproducibility is given as<br />

where<br />

σˆ<br />

=<br />

R<br />

x<br />

x<br />

reproducibility<br />

x<br />

max<br />

min<br />

=<br />

x<br />

max<br />

− x<br />

d 2 is based on the number of operators, p<br />

1<br />

R<br />

d<br />

min<br />

1<br />

x<br />

2<br />

= max( x , x<br />

= min( x , x<br />

2<br />

2<br />

, …x<br />

, …x<br />

p<br />

p<br />

)<br />

)


Methods Based on Analysis<br />

of Variance<br />

• The analysis of variance can be extended to<br />

analyze the data from an experiment and to<br />

estimate the appropriate components of gage<br />

variability.<br />

• For illustration, assume there are a parts and b<br />

operators, each operator measures every part n<br />

times.


Methods Based on Analysis<br />

of Variance<br />

• The measurements, y ijk , could be represented by<br />

the model<br />

⎧ i = 1,<br />

2,...<br />

a<br />

⎪<br />

yijk = µ + τi<br />

+ β j + ( τβ)<br />

ij + εijk<br />

⎨ j = 1,<br />

2,...,<br />

b<br />

⎪<br />

⎩k<br />

= 1,<br />

2,...,<br />

n<br />

where i = part, j = operator, k = measurement.


Methods Based on Analysis<br />

of Variance<br />

• The variance of any observation can be given by<br />

2 2 2 2<br />

στ , σβ,<br />

στβ,<br />

σ<br />

2 2 2 2<br />

ijk ) y ( V σ + σ + σ + σ = τ β τβ<br />

are the variance components.


Methods Based on Analysis<br />

of Variance<br />

• Estimating the variance components can be<br />

accomplished using the following formulas<br />

σˆ<br />

σˆ<br />

σˆ<br />

σˆ<br />

2<br />

2<br />

τβ<br />

2<br />

β<br />

2<br />

τ<br />

=<br />

=<br />

=<br />

=<br />

MS<br />

MS<br />

MS<br />

MS<br />

E<br />

B<br />

A<br />

− MS<br />

n<br />

− MS<br />

an<br />

− MS<br />

bn<br />

AB<br />

AB<br />

AB<br />

E

Hooray! Your file is uploaded and ready to be published.

Saved successfully!

Ooh no, something went wrong!