Application to Adaptive Backstepping for a

Permanent Magnet Synchronous Machine

Abderrahmane KECHICH, Benyounes MAZARI
Abstract
Permanent magnet synchronous machine (Shortened later on PMSM) has shown increasing popularity in
recent years for industrial drive application due to the recent development in magnetic materials, power
converters, and digital signal processors. In particular, PMSM drives are widely used in high performance drive
(HPD) applications. Fast and accurate speed response and quick recovery of speed from any disturbances are
essential. The control of a high performance permanent magnet synchronous motor drive for general industrial
application has received wide spread interest for researchers. In this work, a novel speed and position control
scheme for a (PMSM) is developed based on a nonlinear adaptive control scheme. The vector control scheme
is used to simplify control of the PMSM. System model equations are represented in the synchronously rotating
reference frame and provide the basis for the controller which is designed using the adaptive backstepping
nonlinear (Shortened later on ABNL) technique. Using Lyapunovs stability theory, it is also shown that the
control variables are asymptotically stable. The complete system model is developed and then simulated using
MATLAB/Simulink software. Performance of the proposed controller is investigated extensively at different
dynamic operating conditions such as sudden load change, command speed change, command position
change and parameter variation. The results show the global stability of the proposed controller and hence
found to be suitable for high performance industrial drive applications. The real time implementation of the
complete drive system is currently underway.
Keywords: synchronous machine, permanent magnet, adaptive, backstepping, controller, speed, position,
nonlinear, simulation

Introduction
Backstepping control is a relatively new
technique for the control of uncertain
nonlinear systems. The most appealing point
is the use of virtual control variables to make
the original high order system simple, thus
the final control outputs can be derived step
by step through suitable Lyapunov functions
ensuring global stability. An adaptive robust
nonlinear controller can be derived using this
control method in a straightforward manner.
Recently the newly developed adaptive back
stepping technique has been used in the
design of speed controllers for dc induction
motors and permanent magnet motors [3],
[8].This technique allows the designer to
incorporate most system non linear ties and
uncertainties the design of the controller. In
[2], [11], [6] the authors designed a nonlinear
controller that achieves rotor angular speed
and rotor flux amplitude tracking with
uncertainties in the rotor resistance and load
torque for an induction motor. Results show

Abderrahmane KECHICH, Eng.: Institut dlectrotechnique,

Bechar, Algrie; Benyounes MAZARI, Eng.: Institut
dElectrotechnique, Oran El Mnouer, Algrie.

that tracking objectives are achieved with

very little steady state error or overshoot.
Zhou and Wang [3] have developed a back
stepping based controller for a DC motor and
induction motor with uncertainties. First, the
authors derived a nonlinear model of a DC
motor with parameter uncertainties such as
motor inertia and load torque. Then the
control algorithm is derived from the model
equation. Hualin Tan [2] have also designed
an adaptive back stepping based controller
for position control of an induction motor.
Once again the simulated results showed
that tracking objectives were achieved. The
authors of [8] have successfully implemented
a back stepping based controller in real time
for a PMSM drive system. However, in
designing the controller the authors of did
not take into account all possible system
nonlinearities such as stator resistance
which varies with temperature. Stator
resistance can vary as much as with
temperature [5], [10].
II. Adaptive control of a PMSM
As stated before the PMSM, using the
adaptive backstepping technique, speed

ELECTROTEHNIC, ELECTRONIC, AUTOMATIC, 60 (2012), nr. 1

control is essential for a high performance

drive application, since a motor must have a
fast and accurate speed tracking response,
quick recovery of speed from any
disturbance and must be insensitive to
parameter variations [9]. Such applications
include rolling mills, machine tools, etc.
Position control of a motor is obtained

17

with two control loops- an outer position loop

and an inner speed loop. The position
controller generates the reference speed,
and the speed controller generates the
reference currents in the case of vector
control. A typical vector control scheme for
position or speed control of a PMSP is
shown in Figure.1.

Figure 1. Block diagram for adaptive backstepping based speed control of the proposed drive.

Typically conventional controllers such as

proportional integral (PI) or proportional
integral derivative (PID) have been used in
both loops. The main advantage to this is
simplicity. However, the range f operating
conditions is very limited, and the
performance of the controllers degrades
when uncertainties are introduced [1].
In order to maintain good control
performance an adaptive control scheme
must be used to compensate for
parameter variation [4]. Based on the motor
model, the d-q axis command voltage
components are generated from the
command d-q axis current components.
Using the inverse of Parks Transformation,
the command a-b-c phase voltages are
generated from the d-q axis components,
and then compared with high frequency
triangle waveforms to generate the PWM
logic signals for the inverter [12].
II. 1. Speed Control Design
The foundation of Backstepping is the
identification of a virtual control variable and
forcing it to become a stabilising function.

Thus, it generates a corresponding error

variable which can be stabilized by proper
input selection via Lyapunovs stability theory
[7]. This technique is very useful for online
estimation of parameters which cannot be
directly measured. Hence, this method is
suitable for speed control of a nonlinear
PMSM where parameters vary with magnetic
saturation such as inductance and stator
resistance which varies with temperature.
Moreover, the load torque unknown and
must be estimated as well for complete
nonlinear adaptive control.
The overall objective is to track the rotor
speed r . The tracking error is given by

e = r r

(1)

And the speed error dynamic is given by

.

e = =

1
j

3P

Bm r + TL 2 (m iq + (Ld Lq )id iq )

(2)

The stabilizing function is determined by

differentiating the Lyapunov function
1
V = e2
2

ELECTROTEHNIC, ELECTRONIC, AUTOMATIC, 60 (2012), nr. 1

18

~
Lq = L q Lq

to get
.

V =ee =

3P
e
(m iq + (Ld Lq )id iq )
B m r + TL
2
j

The d-q axis currents

i d and iq

(3)

are

identified as the virtual control variables to

stabilize motor speed. From (3) we choose
the following stabilising functions
(4)

i d = 0

(5)

Where k1 is a constant gain, id and iq are

the command currents. Substituting these
equations back into equation (3) the
Lyapunov function becomes
.

(6)

V = k1e 2

if k1 > 0 , then the function is negative

semi-definite which ensures asymptotic
stability.
If the d-q axis currents are identified as
the
virtual
control
variables,
the
corresponding error functions are defined as
eq = iq iq

(7)

~2 ~
~
~
1 2
L2 Lq TL2 R 2
e + ed2 + eq2 + d +
+
+
2
1 2 3 4

Like the speed error, these error functions

must also be reduced to zero. The current
error dynamics are (with i d = 0 )

e&d = i&d i&d =

Rid P r Lq iq v d
Ld

2(Bm k1 J )
[Te Bmr TL ] +
3Pm J

(9)

iq =

2
Bm r + TL k1 Je
3Pm

1
J

e& =

3P
~ 3P

TL + 2 m eq + 2 (Ld Lq )ed iq k1 Je

1 ~ ~&
1 ~ ~&
V&1 = ee& + ed e&d + eq e&q + Ld Ld + Lq Lq +
1
2

(14)

(15a)

1 ~ ~&
1 ~ ~&
TLTL + RR
3
4
3P
e ~ 3P

V&1 = TL + meq + (Ld Lq )ed iq k1Je +

2
2
J

Ri Pr Lqiq vd
+
ed d
Ld

2(B k J )
eq m 1 [Te Bmr TL ] +
3

PJ
m

(15b)

Riq + Pr Ld id + Pr m vq
+
eq

Lq

1 ~ ~& 1 ~ ~&
1 ~ ~& 1 ~ ~&
Ld Ld + Lq Lq + TLTL + RR
1
2
3
4

Lq

~
Ld = L d Ld

(13)

Now (12 can be differentiated to obtain:

(10)

vary witch magnetic saturation, stator

resistance R which varies with temperature,
and load torque which sometimes cannot be
measured directly. The corresponding error
variable are given by

Where TL is the estimated value of load

torque. Now from equation (7), (8) and (13)
the speed error dynamics is given by:

Riq + Pr Ld id + Pr m vq

The parameters that must be estimated

here are the inductances Ld and Lq which

(12)

Where 1 , 2 , 3 , 4 are constant gains

before differentiating, the speed error
dynamics (2) and q-axis command current
(4) must be modified to incorporate the
estimated load torque.

(8)

e d = i d i d

e&q = i&q i&q =

To reduce these estimation errors to zero,

another Lyapunov function is defined as
V1 =

2
( B m r + TL k 1 j e )
i =
3Pm

(11)

~
TL = TL TL
~
R = R R

Now

the

input

voltage

v d and v q are

chosen make equation (15) negative semi

definite to ensure global stability.
vd = R id P r Lq iq + k 2 ed Ld +

3P
Ld Lq iq e
2J

2Lq (Bm k1J ) 3P

miq + Ld Lq idiq Bmr TL + +

3PmJ 2

3
P
Riq + Pr Ldid + Pr m + k3eq Lq + meLq
2J

vq =

(16)

(17)

ELECTROTEHNIC, ELECTRONIC, AUTOMATIC, 60 (2012), nr. 1

Now substitute equation (16) and (17) into

(15) to get
e ~ 3P
3P
(Ld Lq )ed iq k1 Je +
m eq +
V&1 = TL +
J
2
2

ed
P
L
3
d
R R id P r L q Lq iq k 2 ed Ld
L d L q iq e +
Ld
2J

L q (Bm . k1 J )

Ld L d Lq + L q id iq

Jm

eq 2 L q (Bm k1 J )
+ P L L i + +

T
T
L
L
r
d
d
d

3 PJ m
L q

P
3

eL q m

R R i q k 3 e q Lq
2J

~ ~&
~ ~&
~
~
~
~
&
&
Ld Ld Lq Lq TLTL R R
+
+
+
11
2
3
4

Simplifying (18) we get

V&1 = k 1e k e k e +
2
2 d

designing this controller is the same as the

previous procedure for speed control.
However, in this case the model equation is
simplified by grouping parameters into one
variable for convenience; the model is
represented as follows.

i&q = 1 + avq

(25)

i&d = 2 + bv d

(26)

Te = K T i q + K S i d i q

(27)

& r = Ai&q + Bid iq + C + Dr

(28)

(18)

19

2
3 q

where

&
~ e 2 e (B k1 J ) L d
+
TL + q m
+
J
3
3 PJ m

&
~ 3 Pee d iq e q P r id e q (B m k1 J ) id i q L d
+
Ld

Lq
Jm
2J
1

&
~ 3 Pee d i q e d P r i d e q (B m k 1 J )i d iq L q
+
+
+
L q
+
2
Lq
J m
2J

&
e q iq
R
~ e i
R d d
+
Ld
Lq
4

a=
(19)

1
3P
1
, b=
, KT =
m ,
Lq
2
Ld

3P
(Ld Lq ) , A = K T
2
J
K
T
B
B= S , C= L , D= m
J
J
J

KS =

Step 1: Define the position tracking error

as

From (19) we can see the update laws for

parameter estimation are

e1 = = & r

e 2eq (Bm k1 J )
&

TL = 3 +
3PJm

Where is the desired reference

trajectory of the rotor angle, the position
error dynamic is then

3Peed iq eq Pr id eq (Bm k1 J ) id iq
&
L d = 1
+
+
2J
Lq
Jm

(20)

(29)

e&1 = & & (30)

(21)

The stabilising function is determined by

differentiating the Lypunov function V1 =

3Peed iq ed Pr id eq (Bm k1 J )id iq

&

+
+
Lq = 2

2J
Lq
Jm

e i
ei
&
R = 4 d d q q
L
Lq
d

(22)

Therefore, the following expression is

obtained:

V& = k1e 2 k 2 ed2 k3eq2

to get

V&1 = e1e&1 = e1 = e1 & r

(23)

(24)

If k1 , k 2 , k 3 > 0 then it is proved that

equation (24) guarantees asymptotic stability
in the complete system.
II. 2. Position control
For position control, we regard the rotor
speed and d-q axis currents as the virtual
control variables [10]. The procedure for

1 2
e1
2

(31)

We now choose the first stabilising

function as

r = k1e1 + &

(32)

Equation (32) indicates the desired

velocity for position tracking. The next step is
to design a speed controller so that the rotor
speed will follow (32).
Step 2: Now we define the speed tracking
error as

e2 = r r = k1e1 + & r

(33)

From equation (33), the position error

dynamic can be written as

ELECTROTEHNIC, ELECTRONIC, AUTOMATIC, 60 (2012), nr. 1

20

e&1 = k1e1 + e2

(34)

The speed error dynamic is defined as

& r
&r
e&2 =

(35)

e&2 = k12e1 + k1e2 + && Aiq Bid iq C Dr

Now define a new Laypunov function as

V2 =

1 2 1 2
e1 + e 2
2
2

(36)

Differentiate to get

V&2 = e1e&1 + e2 e&2

[( )

V&2 =k1e12 +e2 1k12 e1 +k1e2 +&& Aiq Bidiq CDr

(37)

Since id and iq were identified as the

virtual control variables, we define the
reference currents as

iq =

1
1 k12 e1 + (k1 + k2 )e2 + && C Dr
A

[(

(38)

i d = 0 (39)
Substituting (38) and (39)
equation (37) would yield

back into

V&2 = k1e12 k 2 e22

(40)

Where k1 , k 2 > 0 are design constants,

thus the virtual control is asymptotically
stable. Since the parameters C and D are
unknown we must use their estimated
values C and D .
Thus
equation
(38)
becomes

1
iq = 1 k12 e1 + (k1 + k2 )e2 + && C D r
A

[( )

(41)

Step 3: The goal now is to make id and iq

follow the reference trajectory id and iq . The

e& 3 = iq i&q
~
~
A1 av (45)
e&3 = 4 + A5 +C6 + D6r + Be4iq 6 + D
q

And

e&4 = i&d = 2 bv d

e3 = i iq

(42)

e4 = i d i d

(43)

Using equation (42) and (43) the speed

error dynamics can be represented by

e&2 = e1 k2e2 + Ae3 + Be4iq +C + D

r
~
~
Where C = C C and D = D D are

(46)

Where 4 , 5 and 6 are known signals

defined in appendix.
Step 4: The final Lyapunov function
includes the current errors and parameter
estimation errors.

~ ~
1 2 2 2 2 C 2 D2
V3 = e1 + e2 + e3 e4 + +
2
n1 n2

(47)

Where n1 and n2 are adaptive gains. Now

differentiate and substitute all error dynamic
equations to get.

~& ~&
CC DD
V&3 = e1e&1 + e2e&2 + e3e&3 + e4e&4 +
+
n1
n2

(48)
~
~
V&3 = e1( k1e1 + e2 ) + e2 e1 k2e2 + Ae3 + Be4iq + C + Dr +
~
~
e3 4 + A5 + C6 + Dr 6 avq +
~& ~&
CC DD
+
e4 ( 2 bvd ) +
n1 n2

V&3 = k1e12 k2e22 k3e32 k4e42 + Ae2e3 +

&
~
C
C e2 + 6e3 + +

n1

&

D
~
D e2r + 6e3r + + e3 (k3e3 + 4 avq ) +

n2

e k e + B i e + + D A1 bv
4

4 4

(49)

The d-q axis reference voltages are

chosen to be

final current error signals are defined as

vq =

1
(k3e3 + 4 )
a

vd =

1
k4e4 + B iq e2 + 6 + D A1 2
b

(50)

(51)

Where k 3 , k 4 > 0 are designed constants,

the update laws are defined as:

(44)

the
parameter estimation errors. Now we define
the current error dynamics as

&
C = n2 (e2 + 6 e3 )

(52)

&
D = n3 (e2r + 6e3r )

(53)

ELECTROTEHNIC, ELECTRONIC, AUTOMATIC, 60 (2012), nr. 1

Substituting equation (50) and (53) into

equation (49) would yield

V&3 = k1e12 k2e22 k3e32 k4e42 + Ae2e3 < 0

(54)

For sufficiently large k 2 and k 3 , it is shown

that the complete system is asymptotically
stable.
III. Simulation of the complete PMSM
Drive System
III. 1. Speed control
To verify the effectiveness of the
proposed adaptive scheme, digital simulation
has been performed using Matlab/simulink.

21

The overall objective in a high

performance drive application is to force the
output speed of the motor to follow a desired
speed trajectory with no error.
Block
diagram
for
adaptative
backstepping based speed control of the
proposed drive is as shown in Figure 1.
III.2. Simulation results and discussion
Figure 4. (a) shows the speed response
of the proposed adaptive backstepping
based nonlinear (ABNL) controller and drive
system from a step change in reference
speed from 10 [rad/sec] to 100 [rad/sec] at
full load 1 [Nm].

Figure 4. Simulated responses of the proposed ABNL controller and drive at full load with step change in command
speed.

Next the results are shown in Figure 4 (b)

for a sudden change in torque from 1[Nm] to
3.85 [Nm] at rated speed for the proposed
controller. An up side down peak value
(-2.5 [Nm]) is applied on time 1.5 [sec].
Figures 4 (c) and 4 (d) show the currents
ids and ias which both converge to zero. The
fluctuations in case of current

ias

are

important according to ids due to the

undulant.
The actual speed converge with the

reference speed in a very short time ( ~ 0.10

[sec] ) with minimal overshoot and no stead
state error. The speed changes the sense of
rotation -100 [rad/sec] when a step changes
in load from 10 [Nm] to - 3 [Nm]. One, also,
remarks the actual speed converges with the
reference speed in a very short time.
III. 2.1. Position control
Block diagram for position control of the
proposed drive is shown in Figure 2.

22

ELECTROTEHNIC, ELECTRONIC, AUTOMATIC, 60 (2012), nr. 1

Figure 2. Block diagram for position control

The system is stable in both rotation

senses.
Figure 5. shows comparison between

speed response of the proposed controlled

ABNL and controlled PI.

Figure 5. Comparison between speed responses of the proposed controlled ABNL and controlled PI.

III. 2.2. Simulation results and discussion

In order to verify the effectiveness of the
proposed adaptive backstepping scheme for
position control of the PMSM, digital simulation
have been performed using Matlab / Simulink
software. The overall control bock diagram is

shown in Figure 2 and a detailed block diagram

of the adaptive backstepping controller is shown
in Figure 3. Now the results are shown for square
wave trajectory. Figure 6. (a) shows the rotor
position for a step change in reference position,
20 [rad] to 20 [rad].

Figure 6. Simulated responses of the proposed ABNL controller in Command position.

ELECTROTEHNIC, ELECTRONIC, AUTOMATIC, 60 (2012), nr. 1

The position converges to the reference

position (positive and negative) in a short
time with no overshoot and no steady state
error.
Figure 6(b) shows at 0.5 seconds, the
load torque that was changed from 0 [Nm] to
2 [Nm]. There is no noticeable change in the
position results, likewise in negative case.

23

Figure 6(c) shows current ias . Big fluctuation

appears during the variations of torques ( 0.5
[sec)] and 1.5 [sec]. Figure 6(d) shows the
current iqs that perfectly follows the torque
pace.
For comparison between the adaptive
backstepping order and PI control. The
results are shown in Figure 7.

Figure 7. Comparison between responses position of the proposed controlled ABNL and controlled PI.

The output rotor position does not

converge to the command position and there
is a large overshoot and long settling time for
square wave trajectory. One can see in the
zoom between time 0.1 [sec] and 1.1 [sec]
that the performance to the adaptive
backstepping order is very good vis--vis the
classic order PI.
An adaptive backstepping is based on
nonlinear control technique for a PMSM has
been developed for both position and speed
control.
III. Conclusion
The PMSM can be used for variable
speed high performance electric motor
drives. This performance depends on the
types of controllers used. It was found that
adaptive backstepping based controllers
gave very good speed performance with
PMSM. The motor model equations provided
the basis for the design. Stability was proven
using Lyapunovs stability theory. The
simulation results model for the complete
drive system shown the performance of both
controllers was investigated at different
operating conditions such as sudden change
in command speed, load and parameter
variation. It was found from the results that

the proposed controller is robust and could

be a potential candidate for high
performance industrial drive applications.
Both adaptive controllers were also
compared with conventional fixed gain PI
controllers and it was found that the
performance of both ABNL based controllers
is superior to the PI controllers.
References
[1] Pillay P. and Krishnan R., Control
Characteristics and Speed Controller Design
for
High
Performance
Permanent
Synchronous Motor Drives, in IEEE
Transactions on Power Electronics, Vol. 34,
N 2, Apr., 1990, pp. 151-159.
[2] Tang H. and Chang J., Adartive Position
Control of Induction Motor Systems under
Mechanical
Uncerttainties,
in
IEEE
International Conf. on Power Electronics and
Drive Systems, Vol. 2, 1999, pp. 597-602.
[3] Zhou J., Wang Y. and Zhou R., Adaptive
Backstepping control of separately excited Dc
motor with uncertainties, in International
conf. on Power System Tech., Vol. 1, 2000,
pp. 91-96.
[4] Petrovic V., Ortega R., Stankovic A.M. and
Tadmor G., Design and Implantation of an
Adaptive Controller for torque Ripple
Minimization in PMSM motors, in IEEE

24

[5]

[6]

[7]

[8]

ELECTROTEHNIC, ELECTRONIC, AUTOMATIC, 60 (2012), nr. 1

Transactions on Power Electronics, Vol. 15,

N5, Sep., 2000, pp. 871-880.
Haque M.E. and Rahman M.F., Influence of
Stator Resistance Variation on Direct Torque
Controlled Interior Permanent Magnet
Synchronous Motor Drive Performance and
Its Compensation, in Conf. Rec. IEEE IAS
Annual Meeting, Vol. 4, 2001, pp. 2563-2569.
Huang C.I., Chen K.L., Lee H.T. and Fu L.C.,
Non Linear Adaptive Backstepping Motion
Control of Linear Induction Motor, in Conf.
Rec. American Control Conference, Vol. 4,
2002, pp. 3099-310496.
Shyu K.K., Tseng Y. and Yang D., A Newly
Robust Controller Design for the Position
Control of Permanent Magnet Synchronous
Motor, in IEEE Transaction on Industrial
Electronics, Vol. 49, N 3, 2002, pp. 558-565.
Vilathgamuwa M., Rahman M.A., Tseng K.
and Uddin M.N., Non linear Control of
interior Permanent Magnet Synchronous
Motor, in IEEE Transsaction on Industry

Application, Vol. 39, N 2, Mar./Apr., 2003,

pp. 408-415.
[9] Chen D.F., Liu T.H. and Hung C.K., Non
Linear Adaptive Backstepping controller
Design for a Matrix-converter Based PMSM
Control System, in Conf. Rec. IEEE Annual
Meeting, Vol. 1, 2003, pp. 673-678.
[10]Uddin M. and Lau J., Adaptive backstepping
based non linear control of an IPMSM Drive,
in Power Electronic Society Conference,
Aachen Germany, (2004), pp. 3451-3457.
[11]Lau J. and Uddin M., Non Linear Adaptive
backstepping based
control of
IPMSM
Drive, in International Machines and Drive
Conference, San Antonio TX, USA, (2005),
pp. 1689-1694.
[12]Benabou A., George S. and Clenet S.,
Permanent Magnets Modelling for Dynamic
Application, in Journal of Magnetism and
Magnetic Materials, Vol.320, N 6, March,
2008, pp. 830-835.