09.03.2014 Views

Sliding-Mode Observer with Resistances or Speed Adaptation for ...

Sliding-Mode Observer with Resistances or Speed Adaptation for ...

Sliding-Mode Observer with Resistances or Speed Adaptation for ...

SHOW MORE
SHOW LESS

Create successful ePaper yourself

Turn your PDF publications into a flip-book with our unique Google optimized e-Paper software.

<strong>Sliding</strong>-<strong>Mode</strong> <strong>Observer</strong> <strong>with</strong> <strong>Resistances</strong> <strong>or</strong> <strong>Speed</strong> <strong>Adaptation</strong><br />

f<strong>or</strong> Field-Oriented Induction Mot<strong>or</strong> Drives<br />

Ciro Picardi<br />

Francesco Scibilia<br />

Dept. of Electronics, Computer and System Science<br />

University of Calabria<br />

Dept. of Electronics, Computer and System Science<br />

University of Calabria<br />

Via P. Bucci, 42C<br />

Via P. Bucci, 42C<br />

87036 Rende (CS), ITALY 87036 Rende (CS), ITALY<br />

picardi@deis.unical.it<br />

francesco.scibilia@gmail.com<br />

Abstract – A sliding-mode observer w<strong>or</strong>king in parallel <strong>with</strong> a<br />

particular adaptive scheme is proposed f<strong>or</strong> field-<strong>or</strong>iented<br />

control of induction mot<strong>or</strong>. The observer detects the rot<strong>or</strong> flux<br />

components in the two-phase stationary reference frame. The<br />

adaptive scheme is able to on-line estimate either the mot<strong>or</strong><br />

resistive parameters <strong>or</strong> the rot<strong>or</strong> speed. In the f<strong>or</strong>mer case, the<br />

adaptive scheme allows to have continuously the exact values of<br />

the stat<strong>or</strong> and rot<strong>or</strong> resistances, parameters usually subjected to<br />

variations from their nominal values. In the second one, the<br />

scheme can be used to realize an induction mot<strong>or</strong> sens<strong>or</strong>less<br />

speed control. The analytical development of the sliding mode<br />

observer and the adaptive scheme is fully explained. Simulation<br />

results, based on a Matlab/Simulink/Real-Time-W<strong>or</strong>kshop<br />

model designed on purpose, are presented to verify the validity<br />

of the proposed adaptive observer structure.<br />

I. INTRODUCTION<br />

The induction mot<strong>or</strong> has found very wide industrial<br />

applications due to its well-known advantages as simple<br />

construction, reliability and low cost. The most popular high<br />

perf<strong>or</strong>mance induction mot<strong>or</strong> control method is that one<br />

known as Field-Oriented Control. It is based on a d-q<br />

reference frame rotating synchronously <strong>with</strong> the rot<strong>or</strong> flux<br />

vect<strong>or</strong>, which allows to achieve a decoupled control between<br />

the flux and the produced t<strong>or</strong>que, likewise to a separately<br />

excited DC mot<strong>or</strong> [1].<br />

In Direct Field-Oriented Control, both the instantaneous<br />

magnitude and position of the rot<strong>or</strong> flux vect<strong>or</strong> are supposed<br />

to be precisely known. However, as the rot<strong>or</strong> flux cannot be<br />

directly measured, eff<strong>or</strong>ts have been made to estimate the<br />

rot<strong>or</strong> flux using various kinds of observers, based on the<br />

measurements of the stat<strong>or</strong> currents, the stat<strong>or</strong> voltages and<br />

the mot<strong>or</strong> speed.<br />

An imp<strong>or</strong>tant problem is that the exact values of the mot<strong>or</strong><br />

parameters, from which the observer and some high<br />

perf<strong>or</strong>mance control systems depend, are different from<br />

nominal values and change <strong>with</strong> respect to the temperature<br />

and the operating conditions. Another question is the need of<br />

a speed sens<strong>or</strong> to provide the rot<strong>or</strong> speed measurement,<br />

necessary to regulation purpose as well as observer<br />

operation. The presence of this sens<strong>or</strong> increases the drive<br />

cost and can reduce the robustness of the overall system;<br />

m<strong>or</strong>eover, in some cases a speed sens<strong>or</strong> cannot be mounted,<br />

such as mot<strong>or</strong> drives in a hostile environment and very highspeed<br />

mot<strong>or</strong> drives.<br />

The extended Kalman filters f<strong>or</strong> simultaneous estimation<br />

of the rot<strong>or</strong> flux, the speed and some mot<strong>or</strong> parameters<br />

(frequently only the rot<strong>or</strong> resistance, the parameter subjected<br />

to the widest variation) have been proposed as a potential<br />

solution to the above problems [2], [3]. Unf<strong>or</strong>tunately, this<br />

approach has some inherent disadvantages, such as the<br />

influence of noise and computation burden.<br />

In the last years, the sliding-mode observer has<br />

represented an attractive choice f<strong>or</strong> its being quite robust to<br />

disturbances, parameter deviations and system noise [4], [5].<br />

Adaptive flux sliding-mode observers, in which the mot<strong>or</strong><br />

speed is estimated by additional equations, have been<br />

designed trying to reduce the influence of parameters<br />

variations [6], [7], [8].<br />

This paper presents a sliding-mode observer w<strong>or</strong>king in<br />

parallel <strong>with</strong> a particular adaptive scheme. This scheme is<br />

able to on-line estimate either the mot<strong>or</strong> resistive parameters<br />

<strong>or</strong> the rot<strong>or</strong> speed. Thus, the adaptive sliding-mode observer<br />

allow to obtain a robust rot<strong>or</strong> flux estimation. M<strong>or</strong>eover, the<br />

configuration <strong>with</strong> the adaptive scheme f<strong>or</strong> the on-line<br />

estimation of stat<strong>or</strong> and rot<strong>or</strong> resistances values allows to<br />

implement high accurate speed controls, where there is the<br />

need to have constantly the right values of these parameters<br />

to preserve high level perf<strong>or</strong>mances. On the other hand, the<br />

configuration <strong>with</strong> the rot<strong>or</strong> speed estimation can be used f<strong>or</strong><br />

the implementation of a sens<strong>or</strong>less control.<br />

The paper explains the analytical development of the<br />

sliding-mode observer and the adaptive scheme.<br />

Finally, the validity of the proposed alg<strong>or</strong>ithms is verified<br />

by means of some simulation results, obtained using a<br />

Simulink model designed on purpose.<br />

II. FLUX OBSERVER-BASED FIELD-ORIENTED<br />

INDUCTION MOTOR DRIVES<br />

A. Induction Machine <strong>Mode</strong>l<br />

The induction mot<strong>or</strong> model, <strong>with</strong> the stat<strong>or</strong> currents and<br />

rot<strong>or</strong> fluxes defined as state variables, in the stationary α, β<br />

co<strong>or</strong>dinate system can be written as:<br />

⎡ p is<br />

⎤ ⎡ A11<br />

⎢ ⎥ = ⎢<br />

⎣ p φr<br />

⎦ ⎣ A21<br />

where p=d/dt<br />

R s + Rrp<br />

A11<br />

= − I<br />

σLs<br />

A12<br />

⎤ ⎡is<br />

⎤ ⎡B1<br />

⎤<br />

vs<br />

A<br />

⎥ ⎢ ⎥ + ⎢<br />

22 φr<br />

0<br />

⎥<br />

⎦ ⎣ ⎦ ⎣ ⎦<br />

A<br />

12<br />

1 ⎛ Rrp<br />

⎞<br />

= ⎜ I − ω J ⎟<br />

σL<br />

s ⎝ M ⎠<br />

A21 = Rrp<br />

I A − R rp<br />

22 = I + ω J<br />

M<br />

(1)


s =<br />

[ i i ] T<br />

i stat<strong>or</strong> current<br />

r<br />

sα<br />

LM<br />

Lr<br />

s β<br />

ϕ = φ , φ = [ φ φ ] T rot<strong>or</strong> flux<br />

s =<br />

r<br />

[ v v ] T<br />

sα<br />

s β<br />

r<br />

rα<br />

r β<br />

v stat<strong>or</strong> voltage<br />

R<br />

rp<br />

= R<br />

r<br />

⎛ L<br />

⎜<br />

⎝ L<br />

M<br />

r<br />

⎞<br />

⎟<br />

⎠<br />

2<br />

2<br />

M<br />

L<br />

M =<br />

L<br />

r<br />

σ = 1 −<br />

L 2<br />

M<br />

⎡1<br />

0⎤<br />

⎡0<br />

− 1⎤<br />

1<br />

I = ⎢ ⎥ J =<br />

⎣0<br />

1<br />

⎢ ⎥ B1<br />

= I<br />

⎦ ⎣1<br />

0 ⎦ σL s<br />

ω is the rot<strong>or</strong> electrical speed, subscripts α and β are used<br />

f<strong>or</strong> α-axis and β-axis components. L M<br />

, L r<br />

, R r<br />

, L s<br />

and R s<br />

are<br />

mutual inductance, rot<strong>or</strong> inductance and resistance, and stat<strong>or</strong><br />

inductance and resistance, respectively.<br />

B. Direct Field-Oriented Induction Mot<strong>or</strong> Drive System<br />

Fig. 1 shows the block diagram of a direct field-<strong>or</strong>iented<br />

induction mot<strong>or</strong> drive.<br />

The three desired stat<strong>or</strong> currents f<strong>or</strong> the current-controlled<br />

voltage-source-inverter (CR-VSI) are calculated by the<br />

following transf<strong>or</strong>mation equations from the two-phase d-q<br />

rotating reference frame to the three-phase stationary a-b-c<br />

system:<br />

isad = − isqd<br />

sin θˆ + isdd<br />

cosθˆ<br />

1<br />

( isqd<br />

+ 3 isdd<br />

) − cosθˆ ( isdd<br />

− isqd<br />

)<br />

2<br />

1<br />

( i − 3 i ) − cosθˆ ( i 3 i )<br />

1<br />

isbd<br />

sinθˆ<br />

3<br />

2<br />

1<br />

i scd = sinθˆ<br />

sqd sdd<br />

sdd + sqd<br />

2<br />

2<br />

= (2)<br />

where i sqd is the t<strong>or</strong>que current component, i sdd is the field<br />

current component and θˆ is the instantaneous angle between<br />

d-axis and a-axis.<br />

Since d-axis has to be aligned <strong>with</strong> the rot<strong>or</strong> flux space<br />

vect<strong>or</strong> and α-axis of the α−β stationary reference frame is<br />

aligned <strong>with</strong> a-axis, θˆ can be obtained as follows:<br />

cos θˆ<br />

where<br />

ˆ<br />

/ ˆ<br />

= φrα<br />

φr<br />

φr<br />

β φr<br />

φ r α<br />

sinθˆ<br />

= ˆ / ˆ<br />

(3)<br />

ˆ and ˆ φ r β are the components of the estimate<br />

2<br />

vect<strong>or</strong> ˆφ r of the rot<strong>or</strong> flux and ˆ 2<br />

φ ˆ ˆ<br />

r = φ rα + φrβ<br />

.<br />

Taking into account the induction mot<strong>or</strong> model used in (1),<br />

the vect<strong>or</strong> ˆφ r can be obtained as:<br />

ˆ φ = ˆ (L /L )<br />

(4)<br />

r ϕ r<br />

r<br />

M<br />

Keeping in mind (4), in the following, the vect<strong>or</strong><br />

used referring as the rot<strong>or</strong> flux vect<strong>or</strong>.<br />

L<br />

r<br />

L<br />

s<br />

ˆϕ r will be<br />

Fig. 1 Block diagram of direct field-<strong>or</strong>iented induction mot<strong>or</strong> drive.<br />

III. CURRENT AND FLUX SLIDING-MODE<br />

OBSERVERS DESIGN<br />

Assuming that the speed is measured and the parameters<br />

are exactly known, the sliding-mode observer can be seen as<br />

composed by two parts w<strong>or</strong>king in parallel: a current slidingmode<br />

observer and a rot<strong>or</strong> flux sliding-mode observer.<br />

The current sliding-mode observer is defined by the<br />

following equation and it uses the measured stat<strong>or</strong> voltage<br />

and current vect<strong>or</strong>s as inputs:<br />

pˆi s = A11is<br />

+ A12<br />

ˆ ϕ r + B1v<br />

s + K sgn( ei<br />

)<br />

(5)<br />

where ˆι s is the estimation of i s ,<br />

e ˆ<br />

i is<br />

− is<br />

surface, and<br />

⎡k1<br />

0 ⎤<br />

K = ⎢ ⎥ is the current gain matrix<br />

⎣ 0 k2<br />

⎦<br />

ˆϕ r is the estimation of ϕ r ,<br />

= is the stat<strong>or</strong> current err<strong>or</strong>, chosen as sliding-mode<br />

The rot<strong>or</strong> flux sliding-mode observer, using the measured<br />

stat<strong>or</strong> current vect<strong>or</strong> as input, is defined by the equation:<br />

p ˆ ϕr<br />

= A 21 is<br />

+ A22<br />

ˆ ϕr<br />

+ Η sgn( ei<br />

)<br />

(6)<br />

⎡h11<br />

h12<br />

⎤<br />

where Η = ⎢ ⎥ is the rot<strong>or</strong> flux gain matrix<br />

⎣h21<br />

h22<br />

⎦<br />

Defining the rot<strong>or</strong> flux err<strong>or</strong> as e ϕ = ˆ ϕr<br />

−ϕr<br />

, we obtain<br />

the following simple err<strong>or</strong> equations:<br />

p e i = A12 eϕ<br />

+ K sgn( ei<br />

)<br />

(7)<br />

p eϕ = A22 eϕ<br />

+ Η sgn( ei<br />

)<br />

(8)<br />

The design of the current gain matrix K can be derived by<br />

using the Lyapunov’s stability the<strong>or</strong>em. We choose the<br />

candidate Lyapunov function as follows:<br />

V1 = ( e T<br />

i e i ) / 2<br />

(9)<br />

This function is positive definite, satisfying the first<br />

Lyapunov stability condition. Taking into account (7), we<br />

can write the time derivative of V 1<br />

as follows:


T<br />

T<br />

pV1 = e i (p ei<br />

) = ei<br />

[ f + K sgn( ei<br />

)]<br />

(10)<br />

<strong>with</strong> f = [ f 1 f ]<br />

T<br />

2 = A12eϕ<br />

From the condition of (10) being definite negative, we<br />

obtain the following conditions f<strong>or</strong> the gain matrix K:<br />

k < , k 0 and k 1 > f1<br />

, k2<br />

> f 2<br />

(11)<br />

1 0 2 <<br />

Acc<strong>or</strong>ding to the equivalent control method, in sliding<br />

mode the system constituted by the two observers behaves as<br />

if the term sgn( e i ) is replaced by its equivalent value<br />

x = sgn( e )]<br />

, calculated by (7) assuming e p e = 0 :<br />

[ i<br />

eq<br />

i = i<br />

−1<br />

x = −K<br />

A12e ϕ<br />

(12)<br />

Substituting (12) into (8) and taking into account that<br />

A = − σL , we obtain:<br />

22 s A 12<br />

p eϕ = −Qe ϕ <strong>with</strong> Q [ I HK<br />

−1 = σ Ls<br />

+ ] A 12 (13)<br />

Then, we can make e ϕ → 0 by choosing<br />

Q = q I <strong>with</strong> q > 0<br />

(14)<br />

that leads to the following relation to determine H :<br />

−1<br />

H = [ QA12 − σLs<br />

I]<br />

K<br />

IV. CURRENT AND FLUX SLIDING-MODE<br />

OBSERVERS WITH ADAPTIVE SCHEME<br />

(15)<br />

In <strong>or</strong>der to take into account the parameter variations and<br />

the absence of the speed sens<strong>or</strong>, the auth<strong>or</strong>s proposes an<br />

adaptive scheme w<strong>or</strong>king in parallel <strong>with</strong> the sliding-mode<br />

observers described in the previous section (Fig. 2).<br />

The equations of the two observers become:<br />

pi ˆ ˆ ˆ<br />

s = A11is<br />

+ A12<br />

ˆ ϕ r + B1v<br />

s + K sgn( ei<br />

)<br />

(16)<br />

p ˆ ϕ ˆ ˆ<br />

r = A 21 is<br />

+ A22<br />

ˆ ϕ r + Η sgn( ei<br />

)<br />

(17)<br />

where ˆι s , ˆϕ r , e i , K , H are already defined and<br />

A ˆ ij = Αij<br />

+ ∆Aij<br />

(i, j = 1,2) <strong>with</strong> ∆ A ij denoting the estimation<br />

of the parameter and speed variations from their real values.<br />

In particular, we consider that this adaptive scheme is able to<br />

operate in two different ways. In the f<strong>or</strong>mer, it gives the<br />

estimated values of the resistive parameters Rˆ s and Rˆ rp ;<br />

then, it can be seen as on-line identification scheme allowing<br />

the realization of high perf<strong>or</strong>mance drive systems, f<strong>or</strong> which<br />

the actual values of stat<strong>or</strong> and rot<strong>or</strong> resistances are required.<br />

In the latter way, the adaptive scheme provides the<br />

estimation ωˆ of the rot<strong>or</strong> speed and allows the realization of<br />

a sens<strong>or</strong>less speed control.<br />

Fig. 2 Configuration of current and flux sliding-mode observers<br />

w<strong>or</strong>king in parallel <strong>with</strong> the adaptive scheme<br />

Theref<strong>or</strong>e, in the case considered in this paper, the matrices<br />

are specified as:<br />

Âij<br />

Rˆ<br />

s + Rˆ<br />

⎛<br />

⎞<br />

rp<br />

Aˆ<br />

11 = − I<br />

⎜<br />

Rˆ<br />

Aˆ<br />

1 rp ⎟<br />

12 = I − ωˆ J<br />

σL<br />

⎜<br />

⎟<br />

s<br />

σLs<br />

M<br />

⎝<br />

⎠<br />

A ˆ Rˆ<br />

21 = Rˆ<br />

rp I<br />

Aˆ<br />

rp<br />

22 = − I + ωˆ J (18)<br />

M<br />

where Rˆ<br />

s = Rs<br />

+ ∆ Rs<br />

, Rˆ<br />

rp = Rrp<br />

+ ∆ Rrp<br />

and ˆ ω = ω + ∆ω<br />

.<br />

A. Adaptive Scheme f<strong>or</strong> Resistive Parameters Identification<br />

If a rot<strong>or</strong> speed measure is available ( ∆ ω = 0 ), from (1),<br />

(16) and (18) we obtain the following current err<strong>or</strong> equation:<br />

pe i = ∆ A11is<br />

+ A12eϕ<br />

+ ∆A12<br />

ˆ ϕr<br />

+ K sgn( ei<br />

) (19)<br />

where<br />

∆ Rs<br />

+ ∆ Rrp<br />

1 ⎛ ∆ Rrp<br />

⎞<br />

∆A11<br />

= −<br />

I ∆A<br />

12 = ⎜ I − ω J ⎟<br />

σL<br />

s<br />

σLs<br />

⎝ M ⎠<br />

The adaptive scheme can be derived by using the<br />

Lyapunov’s stability the<strong>or</strong>em. We choose the following<br />

positive definite Lyapunov function:<br />

T<br />

V2 = ( i ei<br />

) / 2<br />

2<br />

Rs<br />

/ 2ks<br />

2<br />

Rrp<br />

/ 2krp<br />

e + ∆ + ∆<br />

(20)<br />

where k s<br />

and k rp<br />

are positive constants.<br />

The time derivative of V 2<br />

becomes:<br />

T<br />

pV2 = ei<br />

(p ei<br />

) + ∆ Rs(p Rˆ<br />

s /ks<br />

) + ∆ Rrp(pRˆ<br />

rp /krp<br />

) (21)<br />

Taking into account (19), we obtain<br />

⎡ Rˆ<br />

s 1 T<br />

⎤<br />

pV2<br />

= pV1<br />

+ ∆ Rs<br />

⎢ p − ei<br />

is<br />

⎥ +<br />

⎢⎣<br />

ks<br />

σLs<br />

⎥⎦<br />

(22)<br />

⎡ Rˆ<br />

rp 1 ˆ ⎤<br />

T ϕ<br />

+ R ⎢<br />

r<br />

∆ rp p − ei<br />

( is<br />

− ) ⎥<br />

⎢ k σL M<br />

⎣ rp s<br />

⎥⎦<br />

where the function pV 1 is given by (10).<br />

Let the second term null and the third term null in (22), we<br />

can find the following adaptive scheme f<strong>or</strong> the resistive


parameters identification:<br />

Rˆ<br />

−1<br />

ks<br />

T<br />

= p ( ei<br />

is<br />

)<br />

(23)<br />

σL<br />

s<br />

s<br />

1 krp<br />

T ˆ ϕr<br />

Rˆ<br />

rp = p<br />

− [ i ( is<br />

− )]<br />

σLs<br />

M<br />

e (24)<br />

By means of (23) and (24) the time derivative pV 2 is equal<br />

to pV 1 and thus it is negative definite choosing the conditions<br />

given in (11). In other w<strong>or</strong>ds, the matrix K f<strong>or</strong> the adaptive<br />

current sliding-mode observer defined by (16) can be<br />

designed on the basis of conditions (11).<br />

Under the assumption that the sliding mode is attained and<br />

the resistive parameters identification is completed, the<br />

design strategy of the rot<strong>or</strong> flux gain matrix H is the same<br />

described in the previous section, that led to relation (15).<br />

B. Adaptive Scheme f<strong>or</strong> Rot<strong>or</strong> <strong>Speed</strong> Estimation<br />

In <strong>or</strong>der to derive an adaptive scheme f<strong>or</strong> rot<strong>or</strong> speed<br />

estimation we define the following definite positive<br />

Lyapunov function:<br />

T 2<br />

W = ( e ϕ eϕ<br />

) / 2 + ∆ω<br />

/ 2kω<br />

(25)<br />

where k ω is a positive constant.<br />

The time derivative of W is expressed as<br />

T<br />

pW = e ϕ (p eϕ<br />

) + ∆ω(p<br />

ˆ ω /kω<br />

)<br />

(26)<br />

By considering only the rot<strong>or</strong> speed as variable, we obtain:<br />

∆ A11 = ∆A21<br />

= 0, ∆A22<br />

= − σLs<br />

and ∆A12<br />

= ∆ωJ.<br />

So, the<br />

current and flux err<strong>or</strong> equations can be written as:<br />

pe i = A12 eϕ<br />

−(<br />

∆ω /σ Ls<br />

) J ˆ ϕr<br />

+ K sgn( ei<br />

) (27)<br />

pe ϕ = − σLs<br />

A12 eϕ<br />

+ ∆ωJ<br />

ˆ ϕr<br />

+ H sgn( ei<br />

)<br />

(28)<br />

In acc<strong>or</strong>ding <strong>with</strong> the equivalent control method, if the<br />

current traject<strong>or</strong>ies reach the sliding manifold, i.e. p e i = 0 ,<br />

from (27) and (28) we obtain:<br />

definite and the following equation f<strong>or</strong> the speed estimation<br />

is obtained:<br />

1 qk<br />

ˆ − ω T T<br />

ω = p [ x K J ˆ ϕr<br />

]<br />

(32)<br />

dσ L<br />

s<br />

V. SIMULATION RESULTS<br />

In this section, the perf<strong>or</strong>mance of the proposed observer is<br />

presented via simulation results. Fig. 3 shows the Simulink<br />

block diagram designed to simulate the drive system.<br />

As shown in Fig. 1, the field-<strong>or</strong>iented controller is based on<br />

a CR-VSI structure: the output of the speed regulat<strong>or</strong><br />

represents the q-axis desired current i sqd<br />

, while the field<br />

weakening block gives the d-axis desired current i sdd;<br />

a (d,q)<br />

to (a,b,c) transf<strong>or</strong>mation provides the components i sad<br />

, i sbd<br />

and<br />

i scd<br />

necessary f<strong>or</strong> the current regulat<strong>or</strong>s. These are fixed<br />

frequency hysteresis regulat<strong>or</strong>s and give the inputs to the<br />

VSI. Standard prop<strong>or</strong>tional plus integral controllers (<strong>with</strong><br />

anti-windup system) are used f<strong>or</strong> speed and flux regulat<strong>or</strong>s.<br />

The adaptive sliding-mode observer provides the rot<strong>or</strong> flux<br />

position θˆ , necessary f<strong>or</strong> the field <strong>or</strong>ientation, and the flux<br />

magnitude ˆφ r , used also to close the flux control loop as<br />

described in relations (2) and (3).<br />

The adaptive scheme can operate either as resistive<br />

parameter identification <strong>or</strong> as rot<strong>or</strong> speed estimation.<br />

The whole control system is discretized <strong>with</strong> a 10 µ s time<br />

step. In <strong>or</strong>der to simulate a microcontroller device<br />

environment the “Adaptive SM <strong>Observer</strong>” uses a 20 µ s<br />

sample time, the “<strong>Speed</strong> Controller” uses a 60 µ s sample<br />

time and the “F.O.C” uses a 20 µ s sample time.<br />

To further on investigate the implementation feasibility,<br />

two standalone real-time applications f<strong>or</strong> a 32-bit Generic<br />

Embedded Process<strong>or</strong> has been developed from the Simulink<br />

drive system model using Real-Time-W<strong>or</strong>kshop (RTW).<br />

The mot<strong>or</strong> parameter nominal values are the following:<br />

PW = 370 W P = 2 poles<br />

V = 230 Volts I s = 1. 7 Amps<br />

s<br />

R s = 24.6 Ohm R = 16. 1Ohm<br />

L = 39.8 mH L = 26. 1mH L = 1. 46 mH<br />

s<br />

r<br />

r<br />

M<br />

1<br />

eϕ = A12 − [( ∆ω /σ Ls<br />

) J ˆ ϕr<br />

− Kx]<br />

(29)<br />

−1<br />

pe<br />

ϕ = [ σLs<br />

K + H]<br />

x = QA12<br />

Kx<br />

(30)<br />

where x = sgn( e )]<br />

and Q is the matrix given in (14).<br />

[ i<br />

−1<br />

T<br />

12 ]<br />

eq<br />

Since [ A = A /d<br />

12<br />

<strong>with</strong> d = det [ A 12 ] > 0 and<br />

Q = q I <strong>with</strong> q > 0 , the time derivative of W becomes:<br />

q T T<br />

q T T pωˆ<br />

pW = − x K Kx + ∆ ω[<br />

− x K J ˆ ϕr<br />

+ ] (31)<br />

d<br />

dσ L<br />

k<br />

s<br />

ω<br />

The first term in (31) is negative definite; thus, imposing the<br />

second term null, the whole function pW is negative<br />

Fig. 3 High level block diagram of the drive system considered


A. <strong>Sliding</strong>- <strong>Mode</strong> <strong>Observer</strong> <strong>with</strong>out Adaptive Scheme<br />

The sliding mode observer is used to calculate the rot<strong>or</strong><br />

flux components in <strong>or</strong>der to implement the field <strong>or</strong>iented<br />

control. The mot<strong>or</strong> operates under nominal t<strong>or</strong>que load and it<br />

is driven to a desired rot<strong>or</strong> speed of 200 rpm.<br />

Fig. 4 shows the real and the estimated values of a<br />

component of rot<strong>or</strong> flux when the parameters are assumed<br />

exactly known. Fig. 5 shows the real and the estimated<br />

values of the same component of rot<strong>or</strong> flux, when we<br />

suppose the stat<strong>or</strong> and the rot<strong>or</strong> resistances are affected by<br />

+5% and +10% variations f<strong>or</strong>m their nominal values<br />

respectively. It can be seen that the estimated flux is accurate<br />

and robust <strong>with</strong> respect to the resistive parameters variations.<br />

Fig. 6 Stat<strong>or</strong> resistance identification, <strong>with</strong> 5% variation<br />

from nominal value<br />

Fig. 4 Real and estimated<br />

φ r α<br />

Fig. 7 Rot<strong>or</strong> resistance identification <strong>with</strong> 10% variation<br />

from nominal value<br />

Fig. 5 Real and estimated<br />

φ r α under parameters variation<br />

Fig. 8 Identification of a unknown stat<strong>or</strong> resistance<br />

B. <strong>Sliding</strong> <strong>Mode</strong> <strong>Observer</strong> <strong>with</strong> Adaptive Scheme<br />

f<strong>or</strong> Resistive Parameters Identification<br />

The adaptive scheme, w<strong>or</strong>king in parallel <strong>with</strong> the sliding<br />

mode observer, allows to obtain the c<strong>or</strong>rect value of resistive<br />

parameters, which can change from the nominal ones<br />

acc<strong>or</strong>ding to mot<strong>or</strong> temperature and operating condition.<br />

Fig. 6 and 7 show the results of the identification adaptive<br />

scheme when a step parameter variation of the two resistive<br />

parameters is considered. Fig. 8 and 9 show how the<br />

proposed adaptive scheme is able to estimate the c<strong>or</strong>rect<br />

values of the stat<strong>or</strong> and rot<strong>or</strong> resistances also when they are<br />

supposed completely unknown.<br />

Fig. 9 Identification of a unknown rot<strong>or</strong> resistance


C. <strong>Sliding</strong> <strong>Mode</strong> <strong>Observer</strong> <strong>with</strong> Adaptive Scheme<br />

f<strong>or</strong> Rot<strong>or</strong> <strong>Speed</strong> Estimation<br />

This adaptive scheme allows to eliminate the speed sens<strong>or</strong><br />

using an estimation of the rot<strong>or</strong> speed in place of the<br />

measured one.<br />

Fig. 10 and 11 show the transient behaviour of the system<br />

in two particular operating conditions: speed reversion and<br />

mot<strong>or</strong> braking, when the mot<strong>or</strong> is under nominal load t<strong>or</strong>que.<br />

Finally, Fig. 12 refers to a real-time simulation done using<br />

the standalone application developed from the Simulink<br />

model of this scheme.<br />

VI. CONCLUSION<br />

In this paper a particular adaptive scheme w<strong>or</strong>king in<br />

parallel <strong>with</strong> a sliding-mode observer has been developed.<br />

The Lyapunov’s stability the<strong>or</strong>em has been used in the<br />

development, granting the stability of the adaptive slidingmode<br />

observer.<br />

This adaptive scheme cannot estimate simultaneously the<br />

resistive parameters and the speed, because the estimation is<br />

done using only the stat<strong>or</strong> electrical quantities and so the<br />

err<strong>or</strong> signal suitable to be used as “engine” f<strong>or</strong> the estimation<br />

mechanism, is not enough. Thus, the proposed configuration<br />

can be used either when the speed is measured but the exact<br />

values of stat<strong>or</strong> and rot<strong>or</strong> resistances are required as in high<br />

perf<strong>or</strong>mance drives <strong>or</strong> to realize a sens<strong>or</strong>less control where<br />

the characteristics of robustness and parameter insensitivity<br />

of the sliding-mode approach are considered adequate.<br />

The paper has shown some simulation results, based on the<br />

development of a Matlab/Simulink/Real-Time-W<strong>or</strong>kshop<br />

model, which have proved the validity of the proposed<br />

scheme in different significant operating conditions.<br />

V. REFERENCES<br />

Fig. 10 Reference and estimated speeds<br />

Fig. 11 Reference and estimated speeds<br />

[l] P. Vas, Vect<strong>or</strong> Control of AC Machines, New-Y<strong>or</strong>k;<br />

Oxf<strong>or</strong>d University Press, 1990, p.120.<br />

[2] M. La Cava, C. Picardi and F. Ranieri, “Application of<br />

the extended Kalman filter to parameter and state<br />

sstimation of induction mot<strong>or</strong>s, ”, International Journal<br />

of <strong>Mode</strong>lling & Simulation, vol. 9, no.3, 1989, pp.85-89.<br />

[3] R. Kim, S. K. Sul, and M. H. Park, “<strong>Speed</strong> sens<strong>or</strong>less<br />

vect<strong>or</strong> control of induction mot<strong>or</strong> using extended<br />

Kalman filter,” IEEE Trans. Ind. Applic., vol. 30,<br />

Sept/Oct 1994, pp.1225-1233.<br />

[4] V. I. Utkin, “<strong>Sliding</strong> <strong>Mode</strong> Control Design Principles<br />

and Application to Electric Drives”, IEEE Trans. Ind.<br />

Electronics, vol. 40, no. 1, Feb. 1993, pp. 23-26<br />

[5] J. Y. Hung and W. Gao and J. C. Hung, “Variable<br />

Structure Control: A survey”, IEEE Trans. Ind.<br />

Electronics, vol. 40, no. 1, Feb. 1993, pp. 2-22.<br />

[6] H. Kubota and K. Matsuse, “<strong>Speed</strong> Sens<strong>or</strong>less Field-<br />

Oriented Control of Induction Mot<strong>or</strong> <strong>with</strong> Rot<strong>or</strong><br />

Resistance <strong>Adaptation</strong>”, IEEE Trans. Ind. Applic., vol.<br />

30, no. 5, Sept/Oct 1994, pp. 1219-1224.<br />

[7] M. Tursini, R. Petrella and F. Parasiliti, “Adaptive<br />

<strong>Sliding</strong>-<strong>Mode</strong> <strong>Observer</strong> f<strong>or</strong> Sens<strong>or</strong>less Control of<br />

Induction Mot<strong>or</strong>s”, IEEE Trans. Ind. Applic., vol. 36,<br />

no. 5, Sept/Oct 2000, pp. 1380-1387.<br />

[8] J. Li, L. Xu and Z. Zhang, “An Adaptive <strong>Sliding</strong> <strong>Mode</strong><br />

<strong>Observer</strong> f<strong>or</strong> Induction Mot<strong>or</strong> Sens<strong>or</strong>less <strong>Speed</strong><br />

Control”, Ind. Appl. Conf., 39 th<br />

Annual Meeting<br />

Conference, Oct. 2004, pp.1329-1334.<br />

Fig. 12 Reference and estimated speed in real-time simulation

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

Saved successfully!

Ooh no, something went wrong!