POWER ENGINEERING AND ELECTRICAL ENGINEERING VOLUME: 15 | NUMBER: 5 | 2017 | DECEMBER

Application of Sensorless Sliding Mode Observer

in Control of Induction Motor Drive

Chau Si Thien DONG 1 , Hau Huu VO 1 , Thinh Cong TRAN 1 ,

Pavel BRANDSTETTER 2 , Petr SIMONIK 2

1
Department of Automation Engineering, Faculty of Electrical and Electronics Engineering, Ton Duc Thang
University, 19 Nguyen Huu Tho street, Tan Phong ward, District 7, Ho Chi Minh City, Vietnam
2
Department of Electronics, Faculty of Electrical Engineering and Computer Science,
VSB–Technical University of Ostrava, 17. listopadu 15/2172, 708 33 Ostrava, Czech Republic

[email protected], [email protected], [email protected],

[email protected], [email protected]

DOI: 10.15598/aeee.v15i5.2626

Abstract. Induction motors are widely used in an in- control, PI controllers are used. PI controllers depend
dustry and it is necessary to improve control methods on parameters of the system and are sensitive to ex-
for induction motors to increase the efficiency of them. ternal noise. Sliding mode control was introduced in
In this paper, sliding mode controllers are proposed in- 1977, which overcomes disadvantages of PI control [1].
stead of traditional PI controllers in vector control of However, disadvantage of sliding mode control is chat-
induction motor drives. Moreover, rotor speed is es- tering phenomenon. In recent years, many researchers
timated by a sliding mode observer. In addition, the have been involved in solving problems such as fuzzy
robustness of control and observer algorithms are also sliding mode control [4], rotor resistance estimation [3],
proved by Lyapunov’s criterion. The experiments are chattering reducing [1], etc.
obtained in different speed changes of an induction mo-
tor drive. These experimental results confirm the dy-
namic properties of a sensorless sliding mode control During last years, sensorless control of AC machine
of an induction motor drive. has been developed. Sensorless means that the rotor
speed is estimated from an observer or an estimator
such as Extended Kalman Filter (EKF), Model Ref-
erence Adaptive System (MRAS) [5], Luenberger Ob-
Keywords server (LO), Sliding Mode Observer (SMO) [2] and [6],
etc. Sensorless control increases the reliability and de-
creases cost and complexity of the system. Among
Induction motor, Lyapunov, sliding mode con- them, MRAS and EKF are suitable for applications
trol, sliding mode observer, vector control. with a medium speed. In contrast, LO and SMO are
robust to noise and parameter variation [7].

1. Introduction In this paper, integral sliding mode control is pro-

posed in controlling of induction motor drives. More-
over, a sliding mode observer is used to estimate rotor
Induction motors are a type of AC electrical drives.
speed. Firstly, a mathematical description of the in-
They are widely used because they are robust, sturdy
duction motor is derived. Next, the sliding mode ob-
and require low maintenance. However, control of an
server, sliding mode controller and the proof of their
induction motor is very difficult because of variable fre-
robustness are in the second part. The structure of the
quency, complex dynamic and parameter variation, etc.
sensorless sliding mode control of the induction motor
[1].
is introduced in the third part. A laboratory stand is
Vector control is one the most popular methods for built and some experimental results are presented in
controlling induction motor drives because of efficiency the fourth part. Finally, some conclusions are summa-
in a wide speed range. However, in traditional vector rized.

c 2017 ADVANCES IN ELECTRICAL AND ELECTRONIC ENGINEERING 747

POWER ENGINEERING AND ELECTRICAL ENGINEERING VOLUME: 15 | NUMBER: 5 | 2017 | DECEMBER

2. Mathematical Model of 3.1. Sliding Mode Observer

Induction Motor The sliding mode observer can be given as:

The state representation of an induction motor is: x̂˙ = Â · ~x̂ + B · ~u + G · sign(S),

(11)
ẋ = A · ~x + B · ~u, ŷ = C · ~x̂,
(1)
y = C · ~x, where x̂ is estimated state vector, ŷ is the estimated
T
where ~x = [iSα , iSβ , ψRα , ψRβ ] is the state vector, output vector, G is the gain matrix of the sliding mode
T
~u = [uSα , uSβ ] is the input vector and ~y is the output observer, S is the sliding mode surface. G and S are
vector. A, B and C are the state matrix, the input defined as:
 
matrix and the output matrix, respectively; and iSα − îSα
S = y − ŷ = , (12)

a11 a12 a13 a14
 iSβ − îSβ
a21 a22 a23 a24 
A= a31 a32 a33 a34  ,
 (2)  
g1 −g2
a41 a42 a43 a44 g2 g1 
G=
 . (13)
g3 −g4 
 
1 0 g4 g3
1 0 1
B= · , (3) The rotor speed is estimated online by a PI controller
σ · Ls 0 0
0 0 as following:
R
  ω̂R = KP ω · zω + KIω · zω dt, (14)
1 0 0 0
C= , (4)
0 1 0 0 where KP ω , KIω are proportional  and integral
 con-

stant of a PI controller, and zω = iSα − îSα · ψ̂Rβ −
L2m · RR + L2R · RS  
a11 = a22 = − , (5) + iSβ − îSβ · ψ̂Rα .
σ · LS · L2R

a12 = a21 = 0, (6) 3.2. Sliding Mode Control

Lm · RR In this section, the sliding mode control is proposed.

a13 = a24 = , (7) A sliding mode controller is designed in two steps:
σ · LS · L2R
choosing a sliding mode surface and choosing control
Lm · ωR signals.
a14 = −a23 = , (8)
σ · LS · LR Integral sliding mode surfaces are chosen as:
Lm · RR
a31 = a42 = , a32 = a41 = 0, (9) Sx = (i∗m − im ) dt + cx · (i∗m − im ) ,
R
LR (15)
RR
R ∗ ∗
a33 = a44 = − , a34 = −a43 = −ωR , (10) Sy = (ωm − ωm ) dt + cy · (ωm − ωm ) , (16)
LR ∗
where Sx and Sy are sliding surfaces, im is the reference
where iSα and iSβ are stator current components, magnetizing current amplitude, i is the actual magne-
m
uSα and uSβ are stator current components, tizing current amplitude, ω ∗ is the reference mechani-
ψRα and ψRβ are rotor current components in cal speed and ω = (2/p) · ωm is the actual mechanical
m R
the stator coordinate system [α, β] respectively; speed with p is the number of poles of induction motor.
RS and RR are stator and rotor resistances,
respectively; LS and LR are stator and ro- In Eq. (15) and Eq. (16), the cx and cy are chosen to
tor inductances; Lm is the mutual inductance; get the desired characteristics. The flux current com-
σ is the leakage constant and ωR is the rotor speed. ponent and torque current component are:
TR
i∗Sx = · (i∗m − im ) +
cx
3. Sensorless Sliding Mode (17)
TR
+im + · Kx · sign(Sx ),
Control of Induction Motor cx
J ∗
i∗Sy = · (ωm − ωm ) +
In this section, a sliding mode observer and an integral cy · KT
(18)
sliding mode controller are developed for controlling J
induction motor drive. + · Ky · sign(Sy ),
cy · KT

c 2017 ADVANCES IN ELECTRICAL AND ELECTRONIC ENGINEERING 748

POWER ENGINEERING AND ELECTRICAL ENGINEERING VOLUME: 15 | NUMBER: 5 | 2017 | DECEMBER

where TR = LR /RR is the rotor time constant, Similarly, the first derivative of Sy has the opposite
KT is the proportional constant between the motor sign with Sy and the condition for Ky is in Eq. (22).
torque and the torque current component, J is the mo-
Therefore, with the control signals as in Eq. (17)
ment of inertia.
and Eq. (18) and with the condition of Kx and Ky
Suppose that the rotor time constant and the load as in Eq. (21) and Eq. (22), the system is asymptotic
torque are changed but are bounded as: stable. However, to eliminate the chattering problem,
the saturation is used instead of signum function in
∆TR
≤ DTR , (19) Eq. (17) and Eq. (18).
TR

|TL | ≤ DTL . (20)

4. Sensorless Sliding Mode
The system is asymptotic stable if and only if Kx Control Structure of
and Ky satisfy:
Induction Motor Drive
Kx > DTR · |i∗m − im | , (21)
cy The control structure of Induction Motor (IM) drive is
Ky > DTL · . (22) shown as in Fig. 1. This figure is the basic scheme of
J
vector control of an induction motor drive but PI con-
Proof: trollers in speed control loop and flux control loop are
replaced by Sliding Mode Controllers (SMC). A slid-
The Lyapunov function is chosen as:
ing mode observer is used to estimate rotor speed and
1 a current model is used to estimate the rotor flux com-
V = · Sx2 + Sy2 .


(23)
2 ponents. The control of induction motor is described
step by step as follows:
The Eq. (23) is positive definite function. The first
derivative of Lyapunov function is: • Step 1: iSa , iSb are measured by current sensors.

V̇ = Sx · Ṡx + Sy · Ṡy . (24) • Step 2: iSα , iSβ are calculated from iSa , iSb .
• Step 3: The stator angle γ̂ and rotor flux compo-
The Eq. (24) is negative definite function if: nents are estimated by a current model.

Ṡx = −Kx · sign(Sx ), (25) • Step 4: Calculate iSx , iSy from iSα , iSβ and γ̂.

Ṡy = −Ky · sign(Sy ), (26) • Step 5: Calculate i∗Sx and i∗Sy by sliding mode
controllers as in Eq. (17) and Eq. (18).
where sign is the signum function:
 • Step 6: Calculate the reference stator voltage com-
 1, x > 0, ponents u∗Sx and u∗Sy by PI controllers.
sign(x) = −1, x = 0, (27)

1, x < 0, • Step 7: u∗Sα and u∗Sβ are calculated from u∗Sx and
u∗Sy with the γ̂ from step 3.
We have: • Step 8: Calculate u∗Sa , u∗Sb and u∗Sc and feed to
1 PWM to control induction motor.
Ṡx = (i∗m − im ) − cx · · (iSx − im ) . (28)
TR • Step 9: The uSα , uSβ , iSα and iSβ are input of
sliding mode observer, which estimates the rotor
Replace iSx in Eq. (17) into Eq. (28), Eq. (28) be- speed. The stator voltage components uSα and
comes: uSβ depends on the u∗Sα and u∗Sβ and the used
∆TR
Ṡx = (i∗ − im ) − method of Pulse Width Modulation (PWM), sine
TR + ∆TR m (29) PWM or space vector PWM.
1
+Kx · · sign (Sx ) .
TR + ∆TR

The first derivative of Sx has the opposite sign with 5. Experimental Results
Sx so condition for Kx is derived as in Eq. (21). Next,
the first derivative of Sy is: A laboratory stand with an induction motor drive is
1 designed for the experimental verification of the cho-
Ṡy = cy · · TL − Ky · sign (Sy ) . (30) sen algorithms. The experimental platform is shown
J

c 2017 ADVANCES IN ELECTRICAL AND ELECTRONIC ENGINEERING 749

POWER ENGINEERING AND ELECTRICAL ENGINEERING VOLUME: 15 | NUMBER: 5 | 2017 | DECEMBER

imref - im DC link
ˆ m +
*
iSx +
*
uSx
SMC PI
iSx
- + uS*
Rotating 6
iˆm , ˆ m , iSx
* *
, iSy Decoupling to PWM Inverter
Stationary uS* 
*
m* +
*
iSy +
+ uSy
SMC +
PI
- -
ˆ m iSy ˆ
uS
uS  KM
SMO

ψˆ SR K ˆi S ˆ m Current
m
 Model
ˆ
iSx ˆ iSa
Stationary
iS
iSy iS  T3/2 iSb
to Rotating

IM

Fig. 1: Sensorless sliding mode control structure of induction motor drive.

220/380 V, 50 Hz
3 phase supply
are done in this subroutine. The load torque is from
an electrodynamometer. At the speed of 100 rpm, the
Encoder
maximum load torque is just 1.5 Nm.
C1 Induction
C2 Motor Tab. 1: Parameters of the induction motor.

Parameter Value
PWM ADC QEP Rated power 2.2 kW
Rated speed 1420 rpm
TMS 320F28335 Rated voltage 230/400 V
DSP
Rated current 8.43 A/4.85 A
Rated torque 14.8 Nm
Fig. 2: Laboratory stand of control of the induction motor. Number of pole pairs 2
Stator resistance 3.44 Ω
Stator inductance 0.1546 H
Rotor resistance 1.7178 Ω
in Fig. 2. In the laboratory stand, a Siemens induc- Rotor time constant 0.09 s
tion motor with the catalog number 1LA7106-4AA10 Moment of inertia 0.005 kg·m2
is used. Parameters of the induction motor are given
in Tab. 1. At the end of the induction motor, an In our laboratory, the data are stored in the memory
incremental sensor is mounted and its resolution is of DSP and then transferred to the MATLAB environ-
3000 pulses per revolution. The induction motor is con- ment to draw responses. The sampling time is chosen
nected to an indirect frequency converter with the volt- as 10 ms. Because of the limitation of memory of the
age DC-link, which was designed at the Department of DSP, the response time is 4 s and the response of speed
Electrical Power Engineering, VSB–Technical Univer- is divided into two cases with and without load. With-
sity of Ostrava. The control of the voltage inverter is out load, the reference speed is 100 rpm in the forward
performed by the sinusoidal pulse width modulation rotation and 60 rpm in the reverse rotation. With load,
with control frequency 10 kHz and control amplitude the reference speed is chosen as 100 rpm and the load
10 V. Therefore, the transfer constant KM is 1/2. torque of 1.5 Nm is added at the time of 1.5 s.
The control system is a Digital Signal Processing Figure 3, Fig. 4, Fig. 5, Fig. 6 and Fig. 7 show the
(DSP) of Texas Instrument eZdspTM F28335 because time response of important parameters of the induc-
it is highly integrated and is a high performance so- tion motor drive, which were obtained by measurement
lution for a demanding control application. The most from the laboratory stand in case of no load. When
important part of program is ADC interrupt subrou- having load, the time response of rotor speed, torque
tine, which is done for every 50 µs. Setting reference and magnetizing current component and flux current
speed, measuring current components, calculating cur- component are in Fig. 8, Fig. 9 and Fig. 10, respec-
rent and flux components as well as control signals, etc. tively.

c 2017 ADVANCES IN ELECTRICAL AND ELECTRONIC ENGINEERING 750

POWER ENGINEERING AND ELECTRICAL ENGINEERING VOLUME: 15 | NUMBER: 5 | 2017 | DECEMBER

reference speed 10
100 estimated speed isx
actual speed isy

50 5
Speed (rpm)

Current (A)
0 0

-50 -5

0 0.5 1 1.5 2 2.5 3 3.5 4 0 0.5 1 1.5 2 2.5 3 3.5 4

Time (s) Time (s)

Fig. 3: Reference speed (blue), real speed (black), estimated Fig. 7: Magnetizing current component and flux current com-
speed (red). ponent.

15
120
10 100

5 80
Torque (Nm)

Speed (rpm)
0 60

-5 40

-10
20 reference speed
estimated speed
0 actual speed
-15
0 0.5 1 1.5 2 2.5 3 3.5 4 0 0.5 1 1.5 2 2.5 3 3.5 4
Time (s)
Time (s)

Fig. 4: Induction motor torque.

Fig. 8: Reference speed (blue), real speed (black), estimated
speed (red).
10
is alfa
is beta
8
5 load torque
motor torque
Current (A)

6
Torque (Nm)

0 4

2
-5
0
0 0.5 1 1.5 2 2.5 3 3.5 4
Time (s) -2

0 0.5 1 1.5 2 2.5 3 3.5 4

Fig. 5: Stator current vector components iSα and iSβ . Time (s)

Fig. 9: Induction motor torque (red) and reference load torque

10 (blue).
is alfa est
is beta est
5
10
Current (A)

isx
isy
0
5
Current (A)

-5
0

0 0.5 1 1.5 2 2.5 3 3.5 4

Time (s)
-5

Fig. 6: Estimated stator current vector components îSα and

0 0.5 1 1.5 2 2.5 3 3.5 4
îSβ . Time (s)

Fig. 10: Magnetizing current component and flux current com-

ponent.
The response of speed is good, the percentage of
overshoot is 9 %, the time response is 0.7 s and the
error between the reference speed and the actual speed
is just about 1 rpm. The speed decreases to 98.5 rpm turns to the value corresponding to the reference speed
at the load. After the control process, the speed re- (see Fig. 8).

c 2017 ADVANCES IN ELECTRICAL AND ELECTRONIC ENGINEERING 751

POWER ENGINEERING AND ELECTRICAL ENGINEERING VOLUME: 15 | NUMBER: 5 | 2017 | DECEMBER

6. Conclusion IEEE, 2013, pp. 43–49. ISBN 978-1-4799-0275-0.

DOI: 10.1109/ICoSC.2013.6750833.
In this paper, the sensorless sliding mode control with
[4] HSU, K., H. H. CHIANG, G. H. HUANG and
sliding mode observer is presented in a low speed range.
T. T. LEE. Enhanced fuzzy sliding mode con-
The theoretical assumptions are verified through simu-
trol to motion controller of linear induction motor
lations in MATLAB-Simulink environment and experi-
drives. In: 2014 IEEE International Conference on
ments on laboratory stand. The estimated rotor speed
System Science and Engineering (ICSSE). Shang-
can be used in the speed control of the induction motor
hai: IEEE, 2014, pp. 268–273. ISBN 978-1-4799-
drive. The absolute error between the reference speed
4367-8. DOI: 10.1109/ICSSE.2014.6887947.
and the actual speed in sensorless sliding mode control
of the induction motor is just 1 rpm without load and [5] YANG, S., X. LI, D. DING, X. ZHANG
1.5 rpm with load at the reference speed of 100 rpm and Z. XIE. Speed sensorless control of in-
(percentage of error is just 1 % to 1.5 %). When com- duction motor based on sliding-mode observer
paring with PI controllers, sliding mode controllers give and MRAS. In: 2016 IEEE 8th International
better results, the relative size of overshoot is smaller, Power Electronics and Motion Control Con-
settling error is smaller as well and it is more robust ference (IPEMC-ECCE Asia). Hefei: IEEE,
to external noise. When having load torque, in sliding 2016, pp. 1889–1893. ISBN 978-1-5090-1210-7.
mode control, the decreasing of rotor speed is smaller DOI: 10.1109/IPEMC.2016.7512583.
than in PI control.
[6] MAAMOURI, R., M. TRABELSI, M. BOUSSAK
and F. M’SAHLI. Sliding mode observer sensor-
Acknowledgment less vector controlled induction motor drive with
anti-windup PI speed controller. In: 2016 XXII
International Conference on Electrical Machines
This paper was supported by the projects: Centre (ICEM). Lausanne: IEEE, 2016, pp. 1187–1193.
for Intelligent Drives and Advance Machine control ISBN 978-1-5090-2538-1. DOI: 10.1109/ICEL-
(CIDAM) project, Reg. No. TE02000103 funded by MACH.2016.7732675.
the Technology Agency of the Czech Republic and
Project Reg. No. SP2017/104 funded by the Student [7] ZHANG, Y., Z. ZHAO, T. LU, L. YUAN, W. XU
Grant Competition of VSB–Technical University of Os- and J. ZHU. A comparative study of Luenberger
trava. observer, sliding mode observer and extended
Kalman filter for sensorless vector control of in-
duction motor drives. In: 2009 IEEE Energy Con-
References version Congress and Exposition. San Jose: IEEE,
2009, pp. 2466–2473. ISBN 978-1-4244-2893-9.
[1] OLIVEIRA, C. M. R., M. L. AGUIAR, J. B. DOI: 10.1109/ECCE.2009.5316508.
A. MONTEIRO and W. C. PEREIRA. Vec-
tor Control of Induction Motor using a Slid-
ing Mode Controller with Chattering Reduc- About Authors
tion. In: IEEE 13th Brazilian Power Electron-
ics Conference and 1st Southern Power Elec- Chau Si Thien DONG was born in Ho Chi Minh
tronics Conference (COBEP/SPEC). Fortaleza: City, Vietnam. She received her M.Sc. in Auto-
IEEE, 2015, pp. 1–6. ISBN 978-1-4799-8779-5. matic Control from Ho Chi Minh City University of
DOI: 10.1109/COBEP.2015.7420071. Technology in 2003. Her research interests include
[2] KIM, J., J. KO, J. LEE and Y. LEE. Rotor Flux nonlinear control, adaptive control, robust control,
and Rotor Resistance Estimation Using Extended neural network and electrical drives.
Luenberger-Sliding Mode Observer (ELSMO) for
Three Phase Induction Motor Control. Canadian Hau Huu VO was born in Binh Thuan, Vietnam. He
Journal of Electrical and Computer Engineering. received his M.Sc. degree in Automation Engineering
2017, vol. 40, iss. 3, pp. 181–188. ISSN 0840-8688. from Ho Chi Minh City University of Technology in
DOI: 10.1109/CJECE.2017.2682259. 2009. His research interests include robotics, control
theory and modern control methods of electrical drives.
[3] TALHAOUI, H., A. MENACER and
R. KECHIDA. Rotor resistance estimation using Thinh Cong TRAN was born in Da Nang
EKF for the rotor fault diagnosis in sliding mode City, Vietnam. He received his M.Sc. degree in
control induction motor. In: 3rd International Electrical and Electronic Engineering from Ho Chi
Conference on Systems and Control. Algiers: Minh City University of Technology in 1998. His

c 2017 ADVANCES IN ELECTRICAL AND ELECTRONIC ENGINEERING 752

POWER ENGINEERING AND ELECTRICAL ENGINEERING VOLUME: 15 | NUMBER: 5 | 2017 | DECEMBER

research interests include microcontroller systems and soft computing methods in the control of electrical
modern control methods of electrical drives. drives.

Pavel BRANDSTETTER was born in Os- Petr SIMONIK was born in Ostrava, Czech
trava, Czech Republic. He received his M.Sc. and Republic. He received his M.Sc. and Ph.D. from
Ph.D. from Brno University in 1979 and 1987, respec- VSB–Technical University of Ostrava in 2002 and
tively. He is now full professor and dean of Faculty 2006, respectively. He is now vice-dean of Faculty
of Electrical Engineering and Computer Science at of Electrical Engineering and Computer Science at
VSB–Technical University of Ostrava. His research VSB–Technical University of Ostrava. His research
interests include AC controlled drives, sensorless interests include automotive electronic systems, power
control and applications of observers, estimator and active filters, and power semiconductor systems.

c 2017 ADVANCES IN ELECTRICAL AND ELECTRONIC ENGINEERING 753