Accurate torque-sensorless control approach for interior permanent-magnet

synchronous machine based on cascaded sliding mode observer

Kai-Hui Zhao1, Chang-Fan Zhang1, Jing He1, Xiang-Fei Li1, Jiang-Hua Feng2, Jian-Hua Liu3, Tao Li3
1
School of Electrical and Information Engineering, Hunan University of Technology, Zhuzhou 412007,
People’s Republic of China
2
CRRC Zhuzhou Institute Co., Ltd, Zhuzhou 412007, People’s Republic of China
3
School of Traffic Engineering, Hunan University of Technology, Zhuzhou 412007, People’s Republic of China
E-mail: [email protected]

Published in The Journal of Engineering; Received on 2nd May 2017; Accepted on 16th June 2017

Abstract: To improve the accuracy of torque control for vector control of interior permanent-magnet synchronous machine (IPMSM), this
study proposes a torque-sensorless control method based on cascaded sliding mode observer (SMO). First, the active flux model is discussed,
which converts the model of IPMSM into the equivalent model of surface-mounted permanent-magnet synchronous machine. Second, to
reduce chattering caused by system parameters variations and external disturbances, the cascaded observer is designed, which is composed
of a variable gain adaptive SMO and an active flux SMO. The variable gain adaptive SMO is designed to estimate the speed, rotor position
and stator resistance in the d–q reference frame. The active flux SMO is designed to estimate the active flux and torque in the α–β reference
frame. Global asymptotic stability of the observers is guaranteed by the Lyapunov stability analysis. Finally, simulations and experiments are
carried out to verify the effectiveness of the proposed control scheme.

1 Introduction disturbance, the ability to handle inaccurate mathematical model.

To reduce chattering phenomenon caused by parameter variations
Permanent-magnet synchronous machine (PMSM) has been widely
and external disturbances of PMSM, the cascaded observer is
used in industrial applications due to its good dynamic response,
designed, which is composed of a variable gain adaptive sliding
high power density, high torque-to-current ratio, and excellent
mode observer (SMO) and an active flux SMO. The variable gain
tracking precision [1]. The accurate torque control is the basic re-
adaptive SMO is designed to estimate the speed and stator resist-
quirement for torque control in the application of high-speed train
ance in the d–q reference frame. The estimated speed and stator
and electric vehicle. However, it is difficult to attain this target
resistance are then imputed to the active flux SMO. The ‘active
because of the variations of interior PMSM (IPMSM) parameters.
flux’ concept [24–29] provides alternatives to convert the
To achieve the high-precision torque control for vector control of
model of IPMSM into an equivalent model of SPMSM. The
PMSM, many methods are proposed to estimate load torque, such
active flux SMO is designed with sigmoid function to estimate
as the direct calculation method, the full-order and reduced-order
the active flux and calculate the torque in the α–β reference
observer, the model reference adaptive system, Kalman filters,
frame. Combining the torque command, the torque-sensorless
parameters estimating strategy and so on. A parameter estimation
control system of PMSM is constituted. Simulation and experimen-
method has been proposed to achieve accurate torque control for
tal results demonstrate the feasibility of the proposed control
the IPMSM by maximum torque per ampere (MTPA) control
method.
scheme in [2]. A load torque observer of Kalman filter has been pro-
The remaining of this paper is organised as follows: The system
posed to form forward compensation for surface-mounted PMSM
description is described in Section 2, and the unified PMSM
(SPMSM) in [3, 4]. A model reference adaptive identification
model is derived in the same section. A variable gain adaptive
method of torque closed-loop control has also been proposed in
SMO is designed in Section 3, and the SMO of active flux is
[5]. A load torque observer and a load torque derivatives observer
designed in Section 4. The simulation and experimental results
are designed based on the sliding-mode control method in [6].
are shown in Section 5. A brief conclusion is given in Section 6.
The accurate torque control method of IPMSM is proposed by para-
meters estimation in [7].
To achieve high-precision torque control for vector control 2 System description
of PMSM, it is necessary to obtain the speed or position by mech-
This section briefly demonstrates the mathematical model of
anical sensors. However, it performances tend to be poor. To over-
SPMSM and IPMSM. IPMSM model can be converted into an
come the drawbacks of the mechanical sensor, and achieve
equivalent SPMSM model by ‘active flux’. Then, the unified
sensorless operation, several algorithms have been suggested
PMSM model can be derived.
in the recent literature [8–16]. Sensorless control can achieve
the rotor position and speed exploiting the voltage and current.
The benefits of sensorless control are reduced the maintenance, 2.1 Mathematical model of SPMSM in the α–β reference frame
the complexity and the size of PMSM, and increased the reliability.
The stator voltage equations of the PMSM in the α–β reference
There are two kinds of sensorless controls: the signal injection
frame are as follows [30]:
[8–10] and the fundamental models of PMSM [10–16].
Considering the advantages of sliding-mode control and ‘active ⎧
flux’ concept, a torque-sensorless control method is proposed ⎪ dc
⎨ u a = Rs ia + a
for vector control of IPMSM in this paper. Sliding-mode control dt (1)
[17–23] is an attractive algorithm due to several features, such as ⎩ u = R i + dcb
⎪
b s b
robustness of system parameter uncertainties and external dt

This is an open access article published by the IET under the Creative Commons J Eng, 2017, Vol. 2017, Iss. 7, pp. 376–384
Attribution License (http://creativecommons.org/licenses/by/3.0/) doi: 10.1049/joe.2017.0160
where ca and cb are the α–β axis stator flux linkages, respectively, where cext represents the active flux, which is a sum of two terms –
and given by the following equations in SPMSM: the conventional permanent-magnet flux linkage and a flux linkage
due to saliency [25]. This active flux is the fictitious permanent-

ca = Ls ia + cra magnet flux linkage of an IPMSM.
(2)
cb = Ls ib + crb Equation (7) can now be expressed as

where cra and crb are the α–β axis rotor flux linkages, respectively, 3
Te = np cext iq (9)
and cra = cr cos ue , crb = cr sin ue . 2
Substituting (2) into (1), the equations of SPMSM in the α–β axis
reference frame can be expressed as follows: Substituting (8) into (5), (5) can be written as

⎧ 
dia Rs 1 ve cd = Lq id + (Ld − Lq )id + cr = Lq id + cext
⎪
⎪
⎨ dt = − L ia + L ua + L crb (10)
s s s
(3) cq = Lq iq
⎪
⎪ di R 1 v
⎩ b = − s ib + ub − e cra
dt Ls Ls Ls Active flux components in the d–q axis reference frame are as
follows:
In (1)–(3), ua , ub , ia and ib are the α–β axis stator voltages (V) and
⎧ 
currents (A), respectively. Ls is the inductance, Rs is the stator resist- ⎨ cext,d = cd − Lq id = cr + Ld − Lq id = cext
ance (V) and cr is the amplitude of the permanent-magnet flux (11)
linkage (Wb), ve is the electrical angular velocity (rad/s), ue is ⎩
cext,q = cq − Lq iq = 0
the electrical angular (rad).
Analysis of (11) indicates that the active flux is aligned along the
2.2 Mathematical model of IPMSM in the d–q reference frame d-axis as is the permanent-flux linkage. Hence, the active flux com-
The stator voltage equations of PMSM in the d–q reference frame ponents in the α–β axis reference frame are as follows. This is
are as follows [30]: shown in Fig. 1.
⎧ 
⎪ dc cext,a = cext cos ue
⎨ ud = Rs id + d − ve cq (12)
dt (4) cext,b = cext sin ue
⎩ u = R i + dcq + v c
⎪
q s q e d
dt
Using (12), converting (10) into the α–β reference frame, the
where cd and cq are the d–q axis stator flux linkages, respectively,
stator flux can be derived from the active flux [25]
and given by the following equations in SPMSM:

 ca = cext,a + Lq ia
cd = Ld id + cr (13)
(5) cb = cext,b + Lq ib
c q = L q iq

Substituting (5) into (4), considering cr can be regarded as time in-

2.4 Unified PMSM model
variant in a short-time period, so we get dcr /dt ≃ 0. The equations
for IPMSM in the d–q axis reference frame can be expressed as Substituting (13) into (1), the stator voltage equation of the IPMSM
follows: in the α–β reference frame can now be expressed as follows:
⎧ ⎧
⎪ did Rs Lq ud ⎪
⎪ di dc
⎪
⎪ dt = − L id + ve L iq + L
⎨ ⎨ ua = Rs ia + Lq a + ext,a
d d d dt dt (14)
(6) ⎪
⎪ dib dcext,b
⎪
⎪ diq R L uq c ⎩ ub = Rs ib + Lq +
⎪
⎩ = − s iq − v e d id + − v e r dt dt
dt Lq Lq Lq Lq
Since the time constant of the mechanical system is much larger
The torque equation of PMSM can be expressed as than that of the electrical system in the PMSM, the following

3
 
Te = np cr iq + Ld − Lq id iq (7)
2

In (4)–(7), ud , uq , id , iq , Ld and Lq represent the d–q axis stator

voltages (V), currents (A) and inductances (Wb), respectively,
(Ld , Lq ), Rs is the stator resistance (V), cr is the amplitude of
the magnet flux linkage (Wb), ve is the electrical angular velocity
(rad/s), ue is the electrical angular (rad) and np is the number of
pole pairs.

2.3 Active flux model

Active flux has been defined in the literature [24–29] as

cext = cr + (Ld − Lq )id (8) Fig. 1 Active flux in coordinate frame

J Eng, 2017, Vol. 2017, Iss. 7, pp. 376–384 This is an open access article published by the IET under the Creative Commons
doi: 10.1049/joe.2017.0160 Attribution License (http://creativecommons.org/licenses/by/3.0/)
equation can be derived: where
⎧ ⎡ ⎤
⎪ dcext,a R̂s Lq ⎡ ⎤
⎪ = −ve cext,b v̂ e ⎥
⎨
dt ⎢ −L 0
T
⎢ Ld ⎥
⎪ dcext,b
(15) Aˆ = ⎢ d
⎥, D̂ = ⎣ −v̂ cr ⎦, x̂ = îd îq
⎪
⎩ ⎣ L R̂s ⎦ e
= ve cext,a −v̂ e d − Lq
dt Lq Lq

The equations for an IPMSM in the α–β reference frame can be ‘ˆ’ denotes 
the estimated values. v = K 1 esign(e), K 1 = k1 I,
expressed as follows: 
1 0
k1 . 0, I = , K 1 e is the sliding switch gain.
⎧ 0 1
⎪
⎪ dia Rs ua ve
⎨ dt = − L ia + L + L cext,b
⎪ Remark 2: The sliding switch gain K 1 e varies according to the
q q q
(16) stators current errors, which effectively reduces the chattering
⎪
⎪ dib R ub v
⎪
⎩ dt = − L ib + L − L cext,a
s e phenomenon.
q q q
Then, the error dynamics equation is obtained by subtracting (19)
from (17)
Remark 1: It is evident that (16) of the IPMSM by active flux is
equivalent to (3) of the SPMSM in the α–β axis reference frame. ė = Ae + Āx̂ + D̄ − v (20)
This results in a reduction in the dependence on the d-axis induct-
ance and thus eliminating the salient pole for IPMSM. The advan-
where
tages of this model include simple and generic equations with only
two required parameters Lq and Rs . ⎡ ⎤
The ‘active flux’ concept can turn all rotor salient-pole ac R̄s Lq
⎢ −L v̄e ⎥
machines into fictitious rotor non-salient-pole machines [24–29]. ⎢ d Ld ⎥
Ā = A − Â = ⎢ ⎥,
⎣ L R̄ ⎦
−v̄e d − s
Lq Lq
3 Sensorless control of IPMSM by a variable gain  
adaptive SMO cr T
D̄ = D − D̂ = 0 −v̄ e
Lq
In this section, a variable gain adaptive SMO for PMSM sensorless
control is designed to estimate the rotor speed and the stator resist-
ance in the d–q reference frame. Then, the estimated values are R̄s = Rs − R̂s , v̄e = ve − v̂e .
inputted to the active flux SMO. Based on the sliding-mode control theory, the sliding mode
manifold is defined upon the stator current errors
3.1 Design the variable gain adaptive SMO  T  T
s = s1 s2 = e = ed eq (21)
According to (6), the mathematical model for the IPMSM can be
expressed as follows:

ẋ = Ax + Bu + D
(17) 3.2 Lyapunov stability analysis, estimation of speed and position
y = Cx

Theorem 1: The stator current error system (20) converge to

where zero exponentially, ifthe sliding mode manifold is chosen as
⎡ (21), and k1 ≥ (Lve /2) − (Rs /Ld ) + h, h . 0, L = (Lq /Ld )−
Rs Lq ⎤ ⎡ 1 ⎤
(Ld /Lq ) . 0. Then, the stator resistance estimation can be derived
− ve 0
⎢ L Ld ⎥ ⎢ Ld ⎥ as [10]
A=⎢⎣
d ⎥, B=⎢ ⎥
Ld Rs ⎦ ⎣ 1⎦
−ve − 0  
Lq Lq Lq 1 1
   T R̂s = lR − ī î − ī î dt (22)
1 0 cr Ld d d Lq q q
C= , D = 0 −ve
0 1 Lq
Moreover, the electrical angular velocity can be derived as
 T  T
x = id iq , u = ud uq
 
Lq L c
v̂e = lv ī î − d ī î − īq r dt (23)
Define the stators current errors Ld d q Lq q d Lq

 T
T
e = ed eq = x − x̂ = īd īq (18)
Proof: The Lyapunov function is selected as
where īd = id − îd , īq = iq − îq .  2 
The proposed variable gain adaptive SMO can be designed as 1 T R̄s v̄2e
V= e e+ + (24)
follows in order to estimate both the speed and stator resistance: 2 lR lv

x̂˙ = Ax̂
ˆ + Bu + D̂ + v (19) where lR , lv are positive constants.

This is an open access article published by the IET under the Creative Commons J Eng, 2017, Vol. 2017, Iss. 7, pp. 376–384
Attribution License (http://creativecommons.org/licenses/by/3.0/) doi: 10.1049/joe.2017.0160
Assuming that changes in the stator resistance Rs and the electrical 4.1 Design the SMO of active flux
angular velocity ve occur slowly, it can be assumed that Rs and ve
According to (16), the mathematical model for the IPMSM can be
remain approximately constant within a given sampling interval.
shown as follows:
Differentiating (24) with respect to time and substituting (20) into
it, then the following equation is obtained: 
ẋ′ = A′ x + B′ u′ + Ed
(30)
R̂˙ R̄ v̄ v̂˙ y′ = Cx′
V̇ = e ė − s s − e e
T
lR lv
where
R̂˙ s R̄s v̄e v̂ ˙e
= eT Ae + eT Āx̂ + eT D̄ − eT v − −
lR lv  T  T  T
  (25) x′ = ia ib , u′ = ua ub , d = cext,a cext,b
1 1 R̂˙ s A′ = −(Rs /Lq )I, B′ = (1/Lq )I, E = −(ve /Lq )J
= e Ae − e v − R̄s
T T
ī î + ī î +
L d d d L q q q lR    
1 0 0 −1
  C = I, I = , J=
Lq L c v̂˙ 0 1 1 0
+ v̄e īd îq − d īq îd − īq r − e
Ld Lq Lq lv
The stators current errors are defined as
In order to drive the system to be convergent, i.e. let
 T
T
e′ = ea eb = x′ − x̂′ = īa īb (31)
eT Ae − eT v , 0 (26)
  where īa = ia − îa , īb = ib − îb .
1 1 R̂˙ Using the IPMSM model (30) in the α–β reference frame, the
R̄s īd îd + īq îq + s =0 (27)
Ld Lq lR SMO can be designed as follows:

 
Lq L c v̂˙ x̂˙ ′ = A′ x̂′ + B′ u′ + v′ (32)
v̄e īd îq − d īq îd − īq r − e =0 (28)
Ld Lq Lq lv  T
where x̂′ = îa îb , v′ = K 2 H(e′ ), K 2 = k2 I, k2 . 0. ‘ˆ’
denotes the estimated values. k2 is a sliding switch gain.
Since Ld , Lq , the condition (26) is described as follows: The switching function H(e′ ) is defined by the following sigmoid
  function [31]:
Rs 2 Rs 2 Lq Ld
e Ae − e v ≤ − ed − eq + ve
T T
− e e
Ld Ld Ld Lq d q 2
H(e′ ) = −1 (33)
1 + e−ae′
− k1 e2d sign(ed ) − k1 e2q sign(eq )
Rs    2 L 
v   2 where a is a positive constant that can be adjusted the slope of the
≤− e 2 − Rs 
e

 + e  
ed sigmoid function. The discontinuous sign function is replaced by a
Ld d Ld q 2 (29) continuous sigmoid function, which effectively reduces the chatter-
 
Lve  
 2
  2  2
  ing phenomenon.
+ eq  −k1 ed  −k1 eq  Hence, the error dynamics equation can be obtained by subtract-
2
    ing (32) from (30) as
Rs Lve 
=− − + k1 e2
Ld 2
ė′ = A′ e′ + Ed − v′ (34)
≤ − he 2

The sliding mode manifold is defined upon the stator current errors
 
where k1 ≥ (Lve /2) − (Rs /Ld ) + h, h . 0, L = (Lq /Ld )− s′ = e′ .
(Ld /Lq ) . 0.
Combining (27)–(29) will yield V̇ ≤ −he2 . According to the 4.2 Lyapunov stability analysis
Lyapunov stability theory, designed observer is stable. Therefore,
k1 , which is large enough, can ensure both the existence of
sliding motion and the asymptotical stability of sliding motion in Theorem 2: The stator current error system (34) converge to zero
′ ′
the global scope. exponentially, if the

 modemanifold
 sliding
  is chosen as s = e ,
From (27), the estimation of stator resistance may be derived as  
and k2 . max { − ċext,b /Lq , − ċext,a /Lq }.
(22). Also, from (28), the electrical angular velocity may be derived
as (23). Proof: The Lyapunov function is selected as
This completes the proof. □
1 T
V ′ = e′ e′ (35)
2
4 Design SMO of active flux for IPMSM
Then differentiating (35) with respect to time and substituting (34)
In this section, the SMO of active flux for IPMSM is designed to into it, then the following equation is obtained:
estimate the active flux in the α–β reference frame. Then, the esti-
mated torque could be derived based on the active flux and stator
V̇ ′ = e′ ė′ = e′ A′ e′ + e′ Ed − e′ v′
T T T T
current. (36)

J Eng, 2017, Vol. 2017, Iss. 7, pp. 376–384 This is an open access article published by the IET under the Creative Commons
doi: 10.1049/joe.2017.0160 Attribution License (http://creativecommons.org/licenses/by/3.0/)
where id , iq and voltages ud , uq . The rotor electrical angle ûe was
derived through the integration of the rotor electrical angular
Rs 2 Rs 2 velocity v̂e .
e′ A′ e′ = −
T
e − e (37)
Lq a Lq b (ii) Step 2: The active flux ĉext,a , ĉext,b is estimated using (42)
eb eb by the SMO of active flux exploiting the stator current ia ,
ea e
e′ Ed = ve cext,b − ve cext,a = − a ċext,a − ċext,b (38)
T
ib , voltage ua , ub and the estimated electrical angular velocity
Lq Lq Lq Lq v̂e , the estimated stator resistance R̂s .
(iii) Step 3: The estimated active flux amplitude ĉext was derived
using (43). The estimated current îq was derived through the
Substituting (37) and (38) into (36), then following equation is Park transformation of the estimated stator current îa , îb by
obtained: the active flux observer (32). The estimated torque could
Rs 2 e eb then be calculated by (44). The torque-sensorless closed-loop
V̇ ′ = − (e + e2b ) − a ċext,a − ċext,b − ea k2 H(ea ) − eb k2 H(eb ) control system is constituted.
Lq a Lq Lq
   
Rs  2 2
ċext,a ċext,b
=− e + eb − ea + k2 H(ea ) − eb + k2 H(eb )
Lq a Lq Lq
 
       
Rs      eb    
 ′ 2 e a     5 Simulations and experiments
≤− e − ċext,a  − ċext,b  − k2 ea  − k2 eb 
Lq Lq Lq To check the performance of the proposed scheme, simulations are
   
R  2   1     1   carried out on Matlab/Simulink, and experiments are carried out on
     
≤ − s e′  −ea  ċext,a  + k2 − eb  ċext,b  + k2 the RT-LAB platform. The block diagram of the IPMSM torque-
Lq Lq Lq
sensorless closed-loop control system is presented in Fig. 2. The
(39) MTPA [10] control scheme is carried out on an IPMSM.
   
The IPMSM parameters are listed in Table 1.
 
Chosen k2 . max { − ċext,b /Lq , − ċext,a /Lq }, then will yield The parameters of the variable gain adaptive SMO are set as
 2 follows: k1 = 300, lR = 0.1, lv = 50. The parameters of the SMO
V̇ ′ ≤ −(Rs /Lq )e′  . According to the Lyapunov stability theory, of active flux are set as k2 = 6000, a = 1. The same parameters
the designed observer is stable. Therefore, k2 , which is large are used both in simulation and experiments.
enough, can ensure both the existence of sliding motion and the
asymptotical stability of sliding motion in the global scope.
This completes the proof. □ 5.1 Simulation results
In order to investigate the dynamic performance of the proposed
torque-sensorless closed-loop control system, the initial rotor elec-
4.3 Estimate the active flux and torque trical angular velocity is set to 50 rad/s and subsequently increases
to 200 rad/s at 0.2 s; the initial stator resistance is set to 0.02 Ω and
Once the system reaches the sliding manifold, according to the
subsequently increases to 0.04 Ω at 0.4 s; the initial load torque is
sliding mode equivalent principle [32], the following equation is
set to 500 Nm and subsequently increases to 1000 Nm at 0.6 s.
obtained:
Figs. 3–6 demonstrate the performance of the proposed variable
s′ = ṡ′ = e′ = ė′ = 0 (40) gain adaptive SMO for IPMSM sensorless control. The estimated
d–q axis stator currents îd , îq are shown in Fig. 3. It is shown that
Substituting (40) into (34) yields the estimated q-axis stator current îq is increased when the torque
command is increased from 500 to 1000 Nm at 0.6 s. The actual
Ed = v′ (41) stator resistance Rs , the estimated stator resistance R̂s and the
error of stator resistance DRs are shown in Fig. 4. The actual elec-
The estimated active flux is derived as follows:
trical angular velocity ve , the estimated electrical angular velocity
 v̂e and the error Dve are given in Fig. 5. The actual electrical pos-
ĉ ext,b = Lq va /ve
(42) ition ue , the estimated electrical position ûe and the errors Due are
ĉ ext,a = −Lq vb /ve given in Fig. 6. Figs. 4–6 show that the increase of the electrical
angular velocity has an impact on the estimated stator resistance
The estimated active flux amplitude ĉext can then be estimated R̂s , the estimated electrical angular velocity v̂e and the estimated
 electrical position ûe at 0.2 s. However, they converge to the
ĉext = ĉext,
2
a + ĉext,b
2
(43) actual value quickly, which verifies the robustness of the proposed
approach.
The current îd , îq can be estimated using Park transformation of the Figs. 7–10 demonstrate the performance of IPMSM by the
estimated current îa , îb by the SMO of active flux (32). proposed active flux SMO. The estimated stator currents îa , îb are
Substituting the estimated amplitude of the active flux ĉext and shown in Fig. 7. The estimated ĉext,a , ĉext,b and the amplitude of
stator current îq into (9) yields the estimated torque active flux ĉext are shown in Fig. 8. The trace of the active flux is
3 shown in Fig. 9. The given torque Te∗ , the estimated torque T̂e
T̂e = np ĉext îq (44) and errors DTe are shown in Fig. 10.
2
It can be drawn from the simulation results:

Remark 3: Based on the above analysis, the procedure of the pro- (i) Before 0.2 s, the frequency of the stator current and active flux
posed scheme is summarised as follows: is low. After 0.2 s, the frequency of the stator current and
active flux increases when the electrical angular velocity
(i) Step 1: The electrical angular velocity v̂e is estimated using increases to 200 rad/s.
(23), and the stator resistance R̂s is estimated using (22) by (ii) Before 0.6 s, the stator current amplitude is small because the
the variable gain adaptive SMO exploiting the stator currents torque command is 500 Nm. After 0.6 s, the torque command

This is an open access article published by the IET under the Creative Commons J Eng, 2017, Vol. 2017, Iss. 7, pp. 376–384
Attribution License (http://creativecommons.org/licenses/by/3.0/) doi: 10.1049/joe.2017.0160
Fig. 2 Schematic diagram of IPMSM torque-sensorless closed-loop control system

Table 1 Parameters of IPMSM

Parameters Unit Values

rated voltage (UN ) V 1500

rated current (IN ) A 400
rated speed (ve ) rad/s 800
stator resistance (Rs ) V 0.02
q-axis inductance (Lq ) H 0.003572
d-axis inductance (Ld ) H 0.002
rotational inertia (J ) kg · m2 100
rotor PM flux (cr ) Wb 0.892
number of pole pairs (np ) pairs 4

Fig. 4 Actual, estimated stator resistances Rs , R̂s and the error DRs

Fig. 3 Estimated stator currents îd , îq

increases to 1000 Nm resulting in increase of the stator current

amplitude. Fig. 5 Actual, estimated electrical angular velocities ve , v̂e and the errors
(iii) When the motor starts or the speed changes, the stator current, Dv e
the active flux and the torque fluctuate, but they converge to
the actual value quickly.
To implement the proposed scheme, HILS experiments are
carried out on an OP5600 RT-Lab platform. The RT-Lab platform
5.2 Experiments results
is shown in Fig. 11, and the configuration is shown in Fig. 12. The
RT-LAB is a powerful, modular, distributed, real-time platform controller is a TMS320F2812 digital signal processor, which imple-
from OPAL-RT Technologies, Inc. It supports model-based ments high-performance control and computation. The PWM
design using rapid control prototyping and hardware-in-the-loop switching frequency is chosen as 5 kHz. The sampling period is
simulation (HILS) of complex dynamic systems [33]. chosen as 20 μs.

J Eng, 2017, Vol. 2017, Iss. 7, pp. 376–384 This is an open access article published by the IET under the Creative Commons
doi: 10.1049/joe.2017.0160 Attribution License (http://creativecommons.org/licenses/by/3.0/)
Fig. 6 Actual, estimated electrical positions ue , ûe and the errors Due
Fig. 9 Observed trace of active flux

Fig. 10 Given, estimated torques Te∗ , T̂e and the errors DTe
Fig. 7 Estimated stator currents îa , îb

Fig. 11 RT-Lab platform

Fig. 8 Estimated active fluxes ĉext,a , ĉext,b , and the amplitude ĉext

Figs. 13 and 14 demonstrate the performance of the IPMSM sen-

sorless control by the proposed variable gain adaptive SMO. Fig. 13
shows the waveform of the actual stator resistance Rs , the estimated Fig. 12 Configuration of the RT-LAB HILS system
stator resistance R̂s and the estimated stator currents îd , îq . Fig. 13a
shows the waveform when the electrical angular velocity increase
from 50 to 200 rad/s. It is evident that the estimated stator resistance
R̂s has fluctuation. Fig. 13b shows the waveform when the stator re- angular velocity and position can also rapidly track the actual
sistance increase from 0.02 to 0.04 Ω. It is evident that the estimated speed value.
stator resistance R̂s can rapidly track the actual stator resistance Rs . Figs. 15 and 16 demonstrate the effectiveness of the torque
Fig. 14 shows that the actual and estimated electrical angular veloci- control for IPMSM by proposed active flux SMO. Fig. 15 shows
ties ve , v̂e , respectively, the actual and estimated electrical posi- that the estimated active flux ĉext,a , ĉext,b , and the amplitude of
tions ue , ûe , respectively. It is evident that the estimated electrical active flux ĉext . Note that the actual active flux is unavailable and

This is an open access article published by the IET under the Creative Commons J Eng, 2017, Vol. 2017, Iss. 7, pp. 376–384
Attribution License (http://creativecommons.org/licenses/by/3.0/) doi: 10.1049/joe.2017.0160
the calculated active flux is determined by (42) and (43) using the velocity increase from 50 to 200 rad/s and torque is 500 Nm.
estimated electrical angular velocity v̂e . Fig. 16 shows the estimated Fig. 16b shows the waveform when the torque increase from 500
stator currents îa , îb and the given torque Te∗ , the estimated torque to 1000 Nm. It is evident that the estimated torque value T̂e can
T̂e . Fig. 16a shows the waveform when the electrical angular rapidly track the given torque value Te∗ .

Fig. 15 Experimental results. The estimated active fluxes ĉext,a , ĉext,b , and
the estimated amplitude of active flux ĉext

Fig. 13 Experimental results: The estimated stator currents îd , îq and the
actual, estimated stator resistances Rs , R̂s
a Electrical angular velocity increases from 50 to 200 rad/s
b Stator resistance increases from 0.02 to 0.04 Ω

Fig. 16 Experimental results: The estimated stator currents îa , îb and the
Fig. 14 Experimental results. The actual and estimated rotor electrical given, estimated torque Te∗ , T̂e
angular velocities v∗e v̂e and the actual and estimated rotor electrical posi- a Electrical angular velocity increases from 50 to 200 rad/s
tions ue , ûe b Torque increases from 500 to 1000 Nm

J Eng, 2017, Vol. 2017, Iss. 7, pp. 376–384 This is an open access article published by the IET under the Creative Commons
doi: 10.1049/joe.2017.0160 Attribution License (http://creativecommons.org/licenses/by/3.0/)
6 Conclusions [11] Yang S.C., Hsu Y.L.: ‘Full speed region sensorless drive of
permanent-magnet machine combining saliency-based and
This paper designs the cascaded SMO for torque-sensorless back-EMF-based drive’, IEEE Trans. Ind. Electron., 2017, 64, pp.
closed-loop control method of IPMSM. A variable gain adaptive 1092–1101
SMO is designed to achieve IPMSM sensorless control in a d–q [12] Saadaoui O., Khlaief A., Abassi M., ET AL.: ‘A sliding-mode observer
reference frame, and another active flux SMO is designed to esti- for high-performance sensorless control of PMSM with initial rotor
mate both the active flux and the torque in the α–β reference position detection’, Int. J. Control, 2017, 90, pp. 377–392
[13] Zhang X.G., Li Z.X.: ‘Sliding-mode observer-based mechanical par-
frame. The performance of torque-sensorless control for IPMSM
ameter estimation for permanent magnet synchronous motor’, IEEE
is achieved. The results of simulation and experimental validate Trans. Power Electron., 2016, 31, pp. 5732–5745
the feasibility and effectiveness of the proposed approach. In the [14] Song X., Fang J., Han B., ET AL.: ‘Adaptive compensation method
future, the cascaded non-singular fast terminal SMO will be for high-speed surface PMSM sensorless drives of EMF-based
designed for torque-sensorless closed-loop control of IPMSM, or position estimation error’, IEEE Trans. Power Electron., 2016, 31,
the chattering elimination and the fast convergence of non-singular pp. 1438–1449
fast terminal-sliding-mode problem of IPMSM system will be [15] Nguyen D., Dutta R., Rahman M.F., ET AL.: ‘Performance of a sensor-
considered. less controlled concentrated-wound interior permanent-magnet
synchronous machine at low and zero speed’, IEEE Trans. Ind.
Electron., 2016, 63, pp. 2016–2026
7 Acknowledgments [16] Comanescu M.: ‘Sensorless rotor position estimation of PMSM by
This work was supported by the Natural Science Foundation full-order and sliding mode EMF observers with speed estimate’,
Int. J. Veh. Auton. Syst., 2013, 11, pp. 157–189
of China under grant nos. 61473117 and 61503131, the Scientific
[17] He J., Zhang C.F., Mao S., ET AL.: ‘Demagnetization fault detection in
Research Fund of the Hunan Provincial Education Department permanent magnet synchronous motors based on sliding observer’,
under grant nos. 16A058 and 17B073, the Hunan Provincial J. Nonlinear Sci. Appl., 2016, 9, pp. 2039–2048
Natural Science Foundation of China under grant nos. [18] Zhao K.H., Chen T.F., Zhang C.F., ET AL.: ‘Online fault detection
2017JJ4031 and 2016JJ5012, the Hunan Province Education of permanent magnet demagnetization for IPMSMs by nonsingular
Department under grant no. 13CY018, the Teaching Reform fast terminal-sliding-mode observer’, Sensors, 2014, 14,
Fund of Hunan Province Education Department under grant no. pp. 23119–23136
[19] Huang G., Luo Y.P., Zhang C.F., ET AL.: ‘Current sensor fault diagno-
Hunan Education Notice [2016] no. 400, the Teaching Reform
sis based on a sliding mode observer for PMSM driven systems’,
of Degree and Postgraduate Education Fund of Hunan University Sensors, 2015, 15, pp. 11027–11049
of Technology under grant no. JG1604, the Key Laboratory for [20] Zhang C.F., Liao H.J., Li X.F., ET AL.: ‘Fault reconstruction based on
Electric Drive Control and Intelligent Equipment of Hunan sliding mode observer for current sensors of PMSM’, J. Sens., 2016,
Province Grant no. 2016TP1018, the Science and Technology 2016, pp. 1–9
Innovative Research Team in Higher Educational Institutions of [21] Rubio J.J.: ‘Hybrid controller with observer for the estimation and
Hunan Province. rejection of disturbances’, ISA Trans., 2016, 65, pp. 445–455
[22] Rubio J. de Jesus: ‘Sliding mode control of robotic arms with dead
zone’, IET Control Theory Appl., 2017, 11, pp. 1214–1221
8 References
[23] Aguilar-Ibanez C.: ‘Stabilization of the PVTOL aircraft based on a
sliding mode and a saturation function’, Int. J. Robust Nonlinear
[1] Jarzebowicz L., Karwowski K., Kulesza W.J.: ‘Sensorless algorithm Control, 2017, 27, pp. 843–859
for sustaining controllability of IPMSM drive in electric vehicle after [24] Boldea I., Paicu M.C., Andreescu G.D., ET AL.: ‘‘Active Flux’
resolver fault’, Control Eng. Pract., 2017, 58, pp. 117–126 DTFC-SVM sensorless control of IPMSM’, IEEE Trans. Energy
[2] Preindl M., Bolognani S.: ‘Optimal state reference computation with Convers., 2009, 24, pp. 314–322
constrained MTPA criterion for PM motor drives’, IEEE Trans. [25] Foo G., Rahman M.F.: ‘Sensorless vector control of interior perman-
Power Electron., 2015, 30, pp. 4524–4535 ent magnet synchronous motor drives at very low speed without
[3] Zheng Z.D., Li Y.D., Xiao X., ET AL.: ‘Load torque observer of per- signal injection’, IET Electr. Power Appl., 2010, 4, pp. 131–139
manent magnet synchronous motor’, Trans. Chin. Electrotech. [26] Liu J.B., Nondahl T.A., Schmidt P.B., ET AL.: ‘Rotor position estima-
Soc., 2010, 25, pp. 30–36 tion for synchronous machines based on equivalent EMF’, IEEE
[4] Zheng Z.D., Li Y.D., Fadel M., ET AL.: ‘A rotor speed and load torque Trans. Ind. Appl., 2011, 47, pp. 1310–1318
observer for PMSM based on extended Kalman filter’. IEEE Int.
[27] Ancuti M.C., Tutelea L., Andreescu G.D., ET AL.: ‘Practical
Conf. Industrial Technology, Bombay, India, 2006, pp. 1057–1062
wide-speed-range sensorless control system for permanent magnet re-
[5] Zhu Y., Wu S.B., Wu Z.H., ET AL.: ‘Precise torque control method of
luctance synchronous motor drives via active flux model’, Electr.
IPMSM in vehicle’, Trans. Chin. Soc. Agric. Mach., 2014, 45,
Power Compon. Syst., 2014, 42, pp. 91–102
pp. 8–13
[28] Zhao Y., Zhang Z., Qiao W., ET AL.: ‘An extended flux model-based
[6] Zhang C.F., He J., Jia L., ET AL.: ‘Virtual line-shafting control for per-
manent magnet synchronous motor systems using sliding-mode rotor position estimator for sensorless control of salient-pole
observer’, IET Control Theory Appl., 2015, 9, pp. 456–464 permanent-magnet synchronous machines’, IEEE Trans. Power
[7] Huang W.Q., Zhang Y.T., Zhang X.C., ET AL.: ‘Accurate torque Electron., 2015, 30, pp. 4412–4422
control of interior permanent magnet synchronous machine’, IEEE [29] Kalluf F.J.H., Isfanuti A.S., Tutelea L.N., ET AL.: ‘1-kW 2000–4500 r/
Trans. Energy Convers., 2014, 29, pp. 29–37 min ferrite PMSM drive: comprehensive characterization and
[8] Kim S.I., Im J.H., Song E.Y., ET AL.: ‘A new rotor position estimation two sensorless control options’, IEEE Trans. Ind. Appl., 2016, 52,
method of IPMSM using all-pass filter on high-frequency rotating pp. 3980–3989
voltage signal injection’, IEEE Trans. Ind. Electron., 2016, 63, [30] Haitham A.R., Atif I., Jaroslaw G.: ‘High performance control of AC
pp. 6499–6509 drives with Matlab/Simulink models’ (Wiley, 2012)
[9] Wang G.L., Yang L., Zhang G.Q., ET AL.: ‘Comparative investigation [31] Ren J.J., Liu Y.C., Wang N., ET AL.: ‘Sensorless control of ship pro-
of pseudorandom high-frequency signal injection schemes for sensor- pulsion interior permanent magnet synchronous motor based on
less IPMSM drives’, IEEE Trans. Power Electron., 2017, 32, a new sliding mode observer’, ISA Trans., 2015, 54, pp. 15–26
pp. 2123–2132 [32] Spurgeon S.K.: ‘Sliding mode observers: a survey’, Int. J. Syst. Sci.,
[10] Foo G., Rahman M.F.: ‘Sensorless sliding-mode MTPA control of 2008, 39, pp. 751–764
an IPM synchronous motor drive using a sliding-mode observer [33] Krishna M., Daya F.J.L.: ‘Adaptive speed observer with disturbance
and HF signal injection’, IEEE Trans. Ind. Electron., 2010, 57, torque compensation for sensorless induction motor drives using
pp. 1270–1278 RT-Lab’, Turk. J. Elec. Eng. Comp. Sci., 2016, 24, pp. 3792–3806

This is an open access article published by the IET under the Creative Commons J Eng, 2017, Vol. 2017, Iss. 7, pp. 376–384
Attribution License (http://creativecommons.org/licenses/by/3.0/) doi: 10.1049/joe.2017.0160