2019 22nd International Conference on Electrical Machines and Systems (ICEMS)

Fuzzy Sliding Mode Control of Permanent Magnet

Synchronous Motor Based on the Integral Sliding
Mode Surface
Jia Longfei Huang Yuping Zheng Jigui
Beijing Research Institute of Precise Beijing Research Institute of Precise Beijing Research Institute of Precise
Mechatronics and Controls Mechatronics and Controls Mechatronics and Controls
Beijing, P.R. China Beijing, P.R. China Beijing, P.R. China
2277393124@qq.com huangyp@2008.sina.com zhengjigui@163.com

Chen Jing Tao Yunfei Li Pengfei

Beijing Research Institute of Precise Beijing Research Institute of Precise Beijing Research Institute of Precise
Mechatronics and Controls Mechatronics and Controls Mechatronics and Controls
Beijing, P.R. China Beijing, P.R. China Beijing, P.R. China
chenjing@bjtu.edu.cn tyf19881113@163.com lipengfei19871003@163.com

Abstract—In order to solve the parameter uncertainty, integral sliding mode controller, in which the fuzzy
reduce the number of fuzzy rules, and realize the high controller was used to adjust the coefficient before the
precision and steady state operation of PMSM under the saturation function of the fractional order sliding mode
uncertainty abrupt load torque, this paper proposes a fuzzy controller (FOSMC). The fuzzy controller has two inputs and
sliding mode control method for PMSM based on the integral one output, including 49(7×7) fuzzy rules. These schemes all
sliding mode surface. The fuzzy controller in this method is a need to determine the specific motor parameters, and there
one-dimensional fuzzy controller, which can realize high are more fuzzy rules in the fuzzy controller.
precision robust control of the motor with unknown
parameters. Simulations and experiments show that this In order to solve the parameter uncertainty, simplify the
method is superior to other methods in stability. fuzzy controller, reduce the number of fuzzy rules, and
realize the high precision and steady state operation of
Keywords—PMSM, fuzzy sliding mode control, integral PMSM under the uncertainty abrupt load torque, this paper
sliding mode surface, unknown parameters proposes a fuzzy sliding mode control method for PMSM
based on the integral sliding mode surface. The fuzzy
I. INTRODUCTION controller in this method is a one-dimensional fuzzy
Sliding mode variable structure control is a robust control controller, which establish five fuzzy rules, it can realize
algorithm which can overcome the problem of system high precision robust control of the motor with unknown
uncertainty, this algorithm is insensitive to system parameter parameters.
variation and load disturbance. Accordingly, the application
II. MATHEMATICAL MODEL OF PMSM
of sliding mode variable structure control in PMSM control
system has certain practical value, but the sliding mode Assuming that there are no damping windings on the
control has chattering property in the sliding mode stage [1- rotor, the reluctance of stator and rotor core is ignored,
5]. hysteresis loss and eddy current loss are not considered, the
three-phase windings are symmetrically distributed, and the
Fuzzy control and sliding mode control are combined excitation magnetic field and armature reactive magnetic
effectively as fuzzy sliding mode control (FSMC) to reduce field are sinusoidal in the air gap. In the conventional two-
chattering [6-10]. Petr [11] proposed an adaptive fuzzy axis d-q synchronous rotating reference frame, the
sliding mode control (AFSMC) method. In the control mathematical model of a typical PMSM can be described as
process, fuzzy self-learning mechanism is used to adjust the
sliding mode control parameters (extended feedback gain
and switching gain) so that the performance of the sliding R Lq 1
id = − s id + ωe iq + ud (1)
mode control is better than that of the fixed-parameter sliding Ld Ld Ld
mode control without chattering. Wai et al. [12] adopted an
adaptive fuzzy sliding-mode control system with an integral- ψf 1
L R
operation switching surface, which can control the position iq = −ωe d id − s iq − ωe + uq (2)
of an electrical servo drive. Lq Lq Lq Lq
If the integral sliding mode surface is used to design the
switching function, which can effectively weaken the high 3
Te = p ψ f iq + ( Ld − Lq )id iq  (3)
frequency chattering and realize the high precision fuzzy 2 
sliding mode control [13-15]. Mani et al. [16] proposed a
new adaptive fractional fuzzy integral sliding mode control where ud, uq, id, iq and Ld, Lq denote the d-axis and q-axis
(FFISMC) scheme for the stability of the permanent magnet voltages, currents and inductance of stator, respectively;
synchronous motor (PMSM) based wind energy conversion
system. Xia Chen et al. [17] designed a fractional order Rs is stator resistance;

978-1-7281-3398-0/19/$31.00 ©2019 IEEE

978-1-7281-3398-0/19/$31.00 ©2019 IEEE

 2019 22nd International Conference on Electrical Machines and Systems (ICEMS)

ωe is the rotor angular velocity; e (t ) = θ(t ) − θd (t )

 (9)
ψf denotes the magnetic flux linkage of the magnet of the
rotor; where e(t)、 e(t ) 、 e(t ) denote the angle error, speed error,
Te is the electromagnetic torque; acceleration error, and θ d (t ) = 1000t ， θd (t ) = 1000 ，
p is the pole pair number of the motor. θd (t ) = 0 , in this paper.
The motor studied is a surface mounted structure, Ld = Lq, The integral sliding surface is defined as
and id is controlled to zero, Then, the electromagnetic torque
can be simplified as t
s (t ) = θ(t ) −  θd (t ) − c1e(t ) − c2 e(t ) dt (10)
0

3
Te = pψ f iq (4) where c1 and c2 are non-zero normal numbers.
2
The derivative of (10) with respect to time
Considering the mechanical load, the dynamic equation
of the PMSM can be formulated as s(t ) = θ(t ) − θd (t ) + c1 e(t ) + c2 e (t ) (11)

Te − TL = J θ + Bθ (5) If the sliding mode control is in ideal state, then
s(t)= s(t ) =0, that is
where TL, J, θ and B denote the load torque, inertial value,
rotor mechanical angular displacement, and damping
e (t ) + c1e(t ) + c2 e(t ) = 0
 (12)
coefficient, respectively.
Therefore, the second-order dynamic model for the By determining c1 and c2, the tracking error e(t) and its
PMSM is re-expressed by using (4) and (5) as derivative approach zero.
The following analysis shows whether the system is
B 3 pψ f stable after adopting the integral sliding surface.
θ = − θ + iq + d (t )
J 2J
In the fuzzy controller described later, s(t) is used as
B
= − θ + k ⋅ u (t ) + d (t ) (6) input and iq (t ) equal to u(t) as the output. It is assumed that
J
iq (t ) =-s(t). Thus
= k ⋅ u (t ) + D (t )

B  e ( k + 1) = θ( k + 1) − θd ( k + 1) = θ( k + 1)


where D(t) includes TL, − θ and some other internal or
J B  3 pψ f
3 pψ f =− θ (k ) + iq ( k ) + d (t ) (13)
external disturbances, k = is constant. J 2J
2J B
= − (e( k ) + 1000) − k ⋅ s ( k ) + d (t )
III. DESIGN OF FISMC CONTROLLER J
The fuzzy controller with integral sliding mode surface as e ( k + 1) can be obtained by using (13), and then it can be

input is designed to replace the traditional PI controller in the deduced
speed loop. The design procedure is divided into two parts:
integral sliding mode surface selection and fuzzy controller
design. In this one-dimensional fuzzy controller, the integral  e( k + 1) = e( k ) + 
e ( k + 1) Δt
sliding mode surface s(t) is used as the input, and the control  e( k + 1) = e( k ) + e( k + 1) Δt

input u(t) is used as the output, which greatly reduce the  (14)
number of fuzzy rules. Differently from the traditional  s( k + 1) = e( k + 1) + c1e( k + 1) + c2 e( k + 1)
method, the controller output is obtained by anti-fuzzy  s ( k + 1) = s ( k ) + s( k + 1) Δt
transformation of barycenter method.
Using (13) and (14), d(t) is any number from minus 1,000
A. Selection of Integral Sliding Surface to plus 1,000, that is d(t)=2000*(rand(1)-0.5), Δt=0.0001.
The tracking error is defined as follows

e(t ) = θ (t ) − θ d (t ) (7)

where θd(t) is the motor angular displacement instruction.

The derivative of (7) with respect to time

e(t ) = θ (t ) − θd (t ) (8)

The derivative of (8) with respect to time Fig. 1. Phase trajectory diagram formed by s(t) and s (t ) .
 2019 22nd International Conference on Electrical Machines and Systems (ICEMS)

Fig. 3 are quantitative factors to quantify the domains of s(t)

and u(t).
Establish the following five fuzzy rules in the fuzzy
system:
（1）If s is NB then u is PB;
（2）If s is NS then u is PS;
（3）If s is Z then u is Z;
（4）If s is PS then u is NS;
Fig. 2. Curve of e(t), e (t ) and 
e (t ) over time. （5）If s is PB then u is NB.
With the aid of MATLAB software, the phase trajectory
diagram formed by s(t), s(t ) is obtained as shown in Fig. 1,
and the variation curves of e(t), e(t ) , e(t ) are shown in
Fig.2.
Fig. 1 is the motion trajectory from the start point to the
end point of the phase trajectory diagram formed by s(t) and
s(t ) . It can be seen that under the action of uncertain
interference, the phase trajectory formed by the s(t) and s(t )
jitter around the end point (0,0) finally. If the interference is
Fig. 4. Membership function of fuzzy controller input.
zero, then s(t)= s(t ) =0. As can be seen from Fig. 2, when
t>0.05, both e(t ) and e(t ) are approximately equal to zero,
and when t>0.4, e(t) is approximately equal to zero. Because
of the interference, there is a certain jitter phenomenon in the
curve over time, and there is little jitter phenomenon in the
curve after integration, that is, the jitter is reduced by the
integral.
In conclusion, it can be seen from Fig. 1 and Fig. 2 that
the system is stable.
B. Fuzzy Controller Design Fig. 5. Membership function of fuzzy controller output.

The barycenter method is used for anti-fuzzification,

which takes the barycenter of the area enclosed by the
membership function curve and the abscissa as the final
output value of fuzzy reasoning.
IV. SIMULATIONS
The simulation model of fuzzy adaptive integral sliding
mode control system is constructed by MATLAB/Simulink,
Fig. 3. Fuzzy control block diagram at the speed loop. so the specific values of many parameters of PMSM are not
The traditional two-input fuzzy controller uses e(t), e(t ) considered. By simulating the difference between traditional
sliding mode control (SMC), fuzzy sliding mode control
or s(t), s(t ) as input, and this paper uses the switching (FSMC) and fuzzy sliding mode control based on integral
function s(t) as the input of the fuzzy controller, and u(t) as sliding mode surface (FISMC) under the action of irregular
the output, forming a fuzzy system of single input and single load torque TL are compared through simulation, in which
output. Constructing the fuzzy rule base on experience, thus TL=10t+4sin(100t)+5sign(t-0.2)-5sign(t-0.3) is shown in
greatly reduced the number of fuzzy rules. The fuzzy control Fig.6.
block diagram as shown in Fig. 3 is designed.
The fuzzy set is defined as follows:
s(t)={ NB NS Z PS PB }，u(t)={ NB NS Z PS PB }
where NB, NS, Z, PS and PB represent negative big,
negative small, zero, positive small and positive big
respectively. The input membership function of the fuzzy
system is shown in Fig. 4, and the output membership
function (of the fuzzy system) is shown in Fig. 5 The
continuity domain of s(t) and u(t) are all [-1,1]. k1 and k2 in

Fig. 6. A diagram of the load torque.

 2019 22nd International Conference on Electrical Machines and Systems (ICEMS)

A. Simulation Under Traditional Sliding Mode Control C. Simulation Under FISMC

In the traditional sliding mode control method, it is In the integral fuzzy sliding mode control method, the
assumed that the motor parameters are known. The sliding parameters of the motor are assumed to be unknown. The
surface and output are: concrete design of the controller in this method is shown in
the third part-"III. Design of FISMC controller".
s (t ) = ce(t ) + e(t ) (15) c1, c2, k1 and k2 are determined by experience, where k1
and k2 are quantitative factors to quantify the theory domains
 B  3 pψ f of s(t) and u(t), and Fig. 9 is obtained by simulation. In the
u (t ) =  −ce(t ) + θ(t )+θd (t )-η sgn( s )  / ( ) (16) chart, the velocity change curve is enlarged when the TL is
 J  2J
suddenly added (t=0.2s) and suddenly removed (t=0.3s).
where η>|D(t)|, which ensure the stability of the system.
Makes c=25，η=20/J.
The simulation obtains speed x2 which shown in Fig. 7
along with the time variation curve. In the chart, the velocity
change curve is enlarged when the TL suddenly added
(t=0.2s) and suddenly removed (t=0.3s).

Fig. 9. Speed simulation diagram under FISMC.

D. Comparison of the Three Methods

Through three methods of simulation, Fig. 7, Fig. 8 and
Fig. 9 are respectively obtained. By analyzing the speed
curves, the overshoot σ, rising time tr, jitter range in
stationary phase △sta, jitter range in sudden adding △add and
Fig. 7. Speed simulation diagram under traditional sliding mode control.
sudden removing △rem TL of the three methods are obtained.
Performance index values as shown in Table I, thereby sum
B. Simulation Under Fuzzy Sliding Mode Control up the advantages and disadvantages of the three methods.
In the fuzzy sliding mode control method without
integrals, the motor parameters of the motor are assumed to TABLE I
be unknown. The sliding surface as PERFORMANCE INDEX VALUES UNDER THREE METHODS
SMC FSMC FISMC
s (t ) = c1e(t ) + c2 e(t ) (17)
σ/% 49.31 1.91 0
s(t) as the input, and the output u(t) is obtained through tr/×10-2s 1.14 1.73 1.97
the fuzzy controller. c1, c2, k1 and k2 are determined by △sta/% 0.25 0.70 0.25
experience, and Fig. 8 is obtained by simulation. In the chart, △add/% 2.45 1.90 1.00
the velocity change curve is enlarged when the velocity just △rem/% 0.74 1.18 0.97
reached the expected velocity, sudden adding (t=0.2s) and
sudden withdrawing (t=0.3s) TL.
In SMC method, motor parameters are assumed to be
known, tr is short, the jitter range △sta is small in the
stationary stage, the jitter range △rem is small and the
response speed is fast when TL is suddenly withdrawn.
However, this method of the overshoot σ should be very
large, the jitter range △add is big when TL is added suddenly,
and the stability is poor.
In FSMC method, it is assumed that the parameters of the
motor are unknown. σ is small, but not zero. △sta is large,
there is large jitter in the stationary stage with poor stability.
The motor parameters are assumed to be unknown in the
FISMC method, although the tr under this method is long and
the response speed is slow. But σ of this method is zero, and
Fig. 8. Speed simulation diagram under FSMC. the △sta, △add and △rem is small, that is, no matter in what
stage, the system stability of this method perform well.
 2019 22nd International Conference on Electrical Machines and Systems (ICEMS)

V. EXPERIMENTS
Finally, a servo system including a permanent magnet
synchronous motor and a motor controller is developed. To
compare the simulation model, the system experiment is
carried out.
In order to verify the advantages and disadvantages of the
three algorithms, an experimental prototyping platform as
shown in Fig. 10 is developed.

Fig. 13. Speed curve under FISMC method.

As can be seen from these three figures, tr is shorter

under the SMC method, but σ is larger, the speed is unstable,
and jitter range is between ±57.6 rpm. tr is shorter under the
FSMC method, σ is smaller, but the speed is unstable, and
jitter range is between ±45.2 rpm. Under the FISMC method,
the σ is almost zero, the speed is stable, the jitter range is
between ±20.1 rpm, but the tr is longer.
The experimental results show that under the third
control method, the overshoot of speed change is small, the
Fig. 10. Experimental prototyping platform for PMSM drive system.
rising time is long and the speed jitter range is small. That is,
Under the same disturbance, the velocity analysis of under the condition of unknown motor parameters, the high-
PMSM is carried out by using the above three methods. precision speed control of the motor is not only realized, but
After setting the corresponding parameters of the three also weakens the vibration phenomenon of the motor. The
methods, the speed is given as 1000r/min, the speed signal is experimental results have high match with the simulation
collected, and then the experimental data curves shown in results, and the correctness of the simulation model is
Fig. 11, Fig. 12 and Fig. 13 are obtained. verified.
VI. CONCLUSIONS
The method of FISMC is designed by combining with
fuzzy control, the controller structure designed by this
method is simple and can accurately control the motor with
unknown parameters. Compared with SMC and FSMC, this
method has stronger stability to load disturbance and
interference, and has better control accuracy. However,
under the FISMC method, the response time of the system is
longer, the other two methods can be used when the motor
speed and the expected speed difference is large, and the
FISMC method can be used when the difference between the
motor speed and the expected speed is small, and the good
performance of fast motor response and good stability can be
used at the same time.
Fig. 11. Speed curve under SMC method.
REFERENCES
[1] T. Zwerger and P. Mercorelli, “Combining an internal SMC with an
external MTPA control loop for an interior PMSM,” 2018 23rd
International Conference on Methods & Models in Automation &
Robotics (MMAR), Miedzyzdroje, 2018, pp. 674-679.
[2] X. Sun, H. Yu and X. Liu, “Design and application of sliding mode
controller in PMSM position tracking control based on adaptive
backstepping,” 2018 Chinese Automation Congress (CAC), Xi'an,
China, 2018, pp. 3507-3511.
[3] F. MEHEDI, L. NEZLI and M. O. MAHMOUDI, “Speed control of
series-connected five-phase Two PMSM using sliding mode control,”
2018 International Conference on Electrical Sciences and
Technologies in Maghreb (CISTEM), Algiers, 2018, pp. 1-6.
[4] K. Suman and A. T. Mathew, “Speed control of permanent magnet
synchronous motor drive system using PI, PID, SMC and SMC plus
PID controller,” 2018 International Conference on Advances in
Computing, Communications and Informatics (ICACCI), Bangalore,
2018, pp. 543-549.
Fig. 12. Speed curve under FSMC method.
 2019 22nd International Conference on Electrical Machines and Systems (ICEMS)

[5] V. Repecho, D. Biel, and A. Arias, “Fixed switching period discrete- Control and Optimization of Dynamical Systems, vol. 47, no. 1, pp.
time sliding mode current control of a PMSM,” IEEE Transactions on 540-545, 2014.
Industrial Electronics, vol. 65, no. 3, pp. 2039–2048, 2018. [12] R.J. Wai, C.M. Lin, C.F. Hsu, “Adaptive fuzzy sliding-mode control
[6] S. Hussain and M. A. Bazaz, “Fuzzy integrated sliding mode for electrical servo drive,” Fuzzy Sets and Systems, vol.143, no.2, pp
controller for vector controlled PMSM,” 2014 6th IEEE Power India 295-310, 2004.
International Conference (PIICON), Delhi, 2014, pp. 1-6. [13] B. Jiang, H. R. Karimi, Y. Kao, and C. Gao, “A novel robust fuzzy
[7] Q. Nguyen Khanh, N. D. That, Q. Nguyen Hong and Q. P. Ha, integral sliding mode control for nonlinear semi-Markovian jump T–S
“FPGA-based fuzzy sliding mode control for sensorless PMSM fuzzy systems,” IEEE Transactions on Fuzzy Systems, 2018.
drive,” 2012 IEEE International Conference on Automation Science [14] Y. Pan, C. Yang, L. Pan, and H. Yu, “Integral sliding mode control:
and Engineering (CASE), Seoul, 2012, pp. 172-177. Performance, modification and improvement,” IEEE Transactions on
[8] K. M. A. Prasad and U. Nair, “An intelligent fuzzy sliding mode Industrial Informatics, vol. PP, no. 99, pp. 1–1, 2017.
controller for a BLDC motor,” 2017 International Conference on [15] S. Park, T. Yoon, G. Kwak and H. Ahn, “T-S fuzzy control of
Innovative Mechanisms for Industry Applications (ICIMIA), IPMSM using backstepping integral sliding mode,” 2015 15th
Bangalore, 2017, pp. 274-278. International Conference on Control, Automation and Systems
[9] S. Abderazak and N. Farid, “Comparative study between sliding (ICCAS), Busan, 2015, pp. 1113-1118.
mode controller and fuzzy sliding mode controller in a speed control [16] P. Mani, R. Rajan, L. Shanmugam and Y. Joo, “Adaptive fractional
for doubly fed induction motor,” 2016 4th International Conference fuzzy integral sliding mode control for PMSM model,” in IEEE
on Control Engineering & Information Technology (CEIT), Transactions on Fuzzy Systems.
Hammamet, 2016, pp. 1-6.
[17] X. Chen, S. Li, Y. Cui, Y. Xu and M. Xu, “Research on permanent
[10] S. Wang, C. Lin, C. Tseng, J. Chou and B. Syu, “Design of a fuzzy magnet synchronous motor based on fractional order fuzzy sliding
sliding-mode controller for induction motor vector control systems,” mode control,” 2017 Chinese Automation Congress (CAC), Jinan,
2016 International Automatic Control Conference (CACS), Taichung, 2017, pp. 4123-4127.
2016, pp. 206-211.
[11] P. Husek, “Adaptive fuzzy sliding mode control for uncertain
nonlinear systems,” Third International Conference on Advances in