Chaos, Solitons and Fractals 19 (2004) 1217–1224

www.elsevier.com/locate/chaos

Nonlinear chaos control in a permanent magnet

reluctance machine
Ahmad M. Harb *

Department of Electrical Engineering, Jordan University of Science and Technology, P.O. Box 3030, Irbid 22110, Jordan
Accepted 19 June 2003

Abstract
The dynamics of a permanent magnet synchronous machine (PMSM) is analyzed. The study shows that under
certain conditions the PMSM is experiencing chaotic behavior. To control these unwanted chaotic oscillations, a
nonlinear controller based on the backstepping nonlinear control theory is designed. The objective of the designed
control is to stabilize the output chaotic trajectory by forcing it to the nearest constant solution in the basin of at-
traction. The result is compared with a nonlinear sliding mode controller. The designed controller that based on
backstepping nonlinear control was able to eliminate the chaotic oscillations. Also the study shows that the designed
controller is mush better than the sliding mode control.
Ó 2003 Elsevier Ltd. All rights reserved.

1. Introduction

The permanent magnet synchronous machines (PMSM) are widely used in the industrial applications. The avail-
ability of PM with considerable energy density led to the development of DC as well as synchronous machines with PM
filed excitation. The synchronous machine, with its conventional filed excitation in the rotor, is replaced by the PM
excitation. The large majority of commercially available PM synchronous motors are constructed with the permanent
magnets mounted inside the steel rotor core. This type of machine named interior permanent magnet (IPM) [1]. IPM
synchronous machine can offer a high-performance drive by utilizing the torques due the magnets and reluctance
variation. An IPM synchronous machine has problems at high speed and flux-weakening region due to limit on the
threshold voltage [2,3]. The applications of such machine are vary over a wide range, i.e. low-performance applications
to high-performance applications. As an examples of these application include, variable––speed pumps and blowers,
high-speed material handling, space and aircraft applications, and submersible vehicles.
The modern nonlinear theory such as bifurcation and chaos has been widely used to study the stability of a nonlinear
system, for more details, see [4–6]. It has been used to study the permanent magnet machines by many researchers.
Hemati and Leu [7] used chaos theory to study the dynamics of brushless DC motors. Hemati and Kwatny [8] and Park
et al. [9] used bifurcation theory to study PMSM. The studies show that these kinds of machines are experiencing
chaotic oscillations. These undesirable chaotic oscillations need to be eliminated.
So, the main objective of this paper is to design a nonlinear controller to eliminate the undesirable chaotic oscil-
lations. So far many researchers have introduced different kinds of controllers. Abed and Fu [10,11] and Harb et al. [12]
used a nonlinear state-feedback controller to change the subcritical Hopf bifurcation to be supercritical one and
eventually to eliminate the chaotic behavior. Krstic et al. [13], Zaher and Zohdy [14], Harb et al. [15] and Harb and
Ahmad [16] used the concept of the backstepping recursive nonlinear control theory to design a nonlinear controller in

*
Tel.: +962-2-720-1000x22504; fax: +962-2-709-5018.
E-mail address: [email protected] (A.M. Harb).

0960-0779/$ - see front matter Ó 2003 Elsevier Ltd. All rights reserved.
doi:10.1016/S0960-0779(03)00311-4
1218 A.M. Harb / Chaos, Solitons and Fractals 19 (2004) 1217–1224

order to eliminate the chaotic oscillation. Harb and Natsheh [17] used the sliding mode theory to control the chaotic
behavior of Lorenz system. Yassen [18] used feedback and nonfeedback methods, based on the Lyapunov direct
method and Routh-Hurwitz criteria, to suppress chaos to unstable equilibria or unstable periodic orbit. Laval and
MÕSirdi [19] used variable structure controller to control chaotic or hyper-chaotic oscillations of a nonlinear circuit. Li
and Tian [20] used finite time control techniques to solve the synchronization problem of two chaotic systems. Aston
[21] used small parameter perturbations to stabilize fixed point.
In this paper, to control the chaotic oscillations in the permanent magnet machines, a nonlinear controller based on
backstepping recursive theory is designed. Therefore, next section introduces the mathematical model of PMSM. In
Section 3, the controlled simulations, based on the designed nonlinear controller, is discussed. The method of sliding
mode is discussed in Section 4. Finally, conclusions are stated in Section 5.

2. Mathematical model of PMSM

After Hemati and Kwatny [8], we used their derived model of PMSM. The electrical dynamics of PM under in-
vestigation is described by


1
i_q ¼ ð
Riq
nLd xid
nkt x þ vq Þ ð1Þ
Lq


1
i_d ¼ ð
Rid þ nLq xiq þ vd Þ ð2Þ
Ld
J x_ ¼ T ðI; hÞ
Tl ðtÞ ð3Þ

and the electromagnetic torque T ðiq ; id Þ is given by

T ðiq ; id Þ ¼ nðkt iq þ ðLd
Lq Þiq id Þ ð4Þ
pffiffiffiffiffiffiffiffiffiffiffi
where kt ¼ ð3=2Þke , T ðI; hÞ is the torque generated by the motor, and Tl ðtÞ represents the external torque applied to
the motor shaft, including viscous and friction, given by Tl ¼ bx þ TL , where b is the viscous friction coefficient and TL
account for any additional loads imposed on the system. For more details on modeling of PMSM may be found in [3].
After some transformation, done by Hemati and Kwatny [8], the system equations of PMSM machine take the fol-
lowing form:
x_ 1 ¼ v~q
x1
x2 x3 þ qx3 ð5Þ
x_ 2 ¼ v~d
bx2 þ x1 x3 ð6Þ

x_ 3 ¼ rðx1
x3 Þ
TeL ð7Þ

Under certain condition and transformation we found that the system under investigation is equivalent to well known
Lorenz system [8]. The Lorenz system is given by
x_ ¼ rðy
xÞ
y_ ¼ qx
y
xz ð8Þ
z_ ¼
bz þ xy

In Eqs. (5)–(7), for b ¼ 1, and v~d ¼ v~q ¼ TeL ¼ 0, which represents the unforced system, comparing these equations
with Eq. (8), one can see a great degree of similarity between the two sets of equations. So, the system under inves-
tigation is equivalent to Lorenz system and may possess similar dynamic characteristic of Lorenz system

3. Backstepping nonlinear theory

First, the uncontrolled simulations are displayed in this section. After Hemati and Kwatny [8], we use the following
parameters L ¼ Ld ¼ Lq ¼ 1:99 mH, R ¼ 0:9 X, kt ¼ 0:049 N m/A, J ¼ 5  10
6 Kg m2 , and b ¼ 2:3  10
2 N/rad/s. As
shown in Figs. 1 and 2, under certain condition, the study found that the PMSM is experiencing chaotic behavior. These
chaotic oscillations need to be removed or controlled. So, as you will see next section, the concept of nonlinear
backstepping control theory is used in order to design a control signal to eliminate the chaotic oscillations in PMSM.
 A.M. Harb / Chaos, Solitons and Fractals 19 (2004) 1217–1224 1219

20

15

10

X1
0

-5

-10

-15

-20
0 5 10 15 20 25 30 35 40 45 50
Time

Fig. 1. Time history (x1 –time) of the uncontrolled system.

30

20

10
X2

-10

-20

-30
-20 -15 -10 -5 0 5 10 15 20
X1

Fig. 2. State-plane (x2 –x1 ) of the uncontrolled system.

3.1. Control design and simulation

As been mentioned in the previous section that this kind of machine (PMSM) is experiencing chaotic behavior. So, in
this section the main objective is to design a nonlinear controller to eliminate these undesirable chaotic oscillations. To
this end, we introduced one control signal to Eq. (8), and let x1 ¼ x, x2 ¼ y, and x3 ¼ z, one can rewrite the system as
follows: For more details to understand the methodology of backstepping control theory, see Appendix A.
x_ 1 ¼ rðx2
x1 Þ
x_ 2 ¼ qx1
x2
x1 x3 þ u ð9Þ
x_ 3 ¼
bx3 þ x1 x2

Now, let the error signals to be as follows:

e1 ¼ x1
x1d ð10Þ
e2 ¼ x2 þ c1 e1 ð11Þ
e3 ¼ x3 þ c2 e1 þ c3 e2 ð12Þ
1220 A.M. Harb / Chaos, Solitons and Fractals 19 (2004) 1217–1224

For x1d ¼ 0, take the time derivative for Eqs. (10)–(12), one can obtain:
e_ 1 ¼ rðe2
c1 e1
e1 Þ ð13Þ
e_ 2 ¼ qe1
ðe2
c1 e1 Þ
e1 ðe3
c2 e1
c3 e2 Þ þ c1 e_ 1 þ u ð14Þ
e_ 3 ¼
bðe3
c2 e1
c3 e2 Þ þ e2 ðe2
c1 e1 Þ þ c2 e_ 1 þ c3 e_ 2 ð15Þ

Let:
1X 3
V ¼ ki e2i ; ki > 0 ð16Þ
2 i¼1

X
3
V_ ¼ ki ei e_ i ¼ k1 e1 e_ 1 þ k2 e2 e_ 2 þ k3 e3 e_ 3 ð17Þ
i¼1

Substitute Eqs. (10)–(15) into Eq. (17), and:

1. Choose ci and ki such that the terms in V_ are equal to zero,

i ¼ 1; 2; 3

2. For c1 ¼ c2 ¼ c3 ¼ 0, k1 ¼ k3 ¼ 0, and k2 ¼ k4 ¼ 1, choose u such that:

u ¼
e2
qe1 þ e1 e3 ð18Þ

Figs. 3 and 4 shows the controlled simulation results. Comparing Figs. 1–3, it is so clear that the system has recovered
from its chaotic behavior and exhibits a stable performance. This result was achieved using only one control signal. This
control signal is totally depend on the state-variables of the system. To show how the designed controller is so effective
in controlling the chaotic oscillation, the control signal is applied to the systems after 25 s as shown in Fig. 5.

4. Sliding mode theory

A control signal u is added to Eq. (8) and the new systems can be rewrite as [17]
x_ 1 ¼ rðx2
x1 Þ
x_ 2 ¼ qx1
x2
x1 x3 ð19Þ
x_ 3 ¼
bx3 þ x1 x2 þ u

0.7

0.6

0.5

0.4
X1

0.3

0.2

0.1

-0.1
0 5 10 15 20 25 30 35 40 45 50
Time

Fig. 3. Time history (x1 –time) of the controlled system.

 A.M. Harb / Chaos, Solitons and Fractals 19 (2004) 1217–1224 1221

0.9

0.8

0.7

0.6

X2
0.5

0.4

0.3

0.2

0.1

0
-0.1 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7
X1

Fig. 4. State-plane (x2 –x1 ) of the controlled system.

20

15

10

5
X1

-5

-10

-15

-20
0 5 10 15 20 25 30 35 40 45 50
Time

Fig. 5. Time history of the controlled system (with backstepping control after 25 s).

The two separate phases of the sliding mode design method are:
Phase 1
Sliding-surface design: an appropriate sliding surface must be selected to yield desirable performance.
S ¼ aðx1
x1ref Þ þ ðx2
x2ref Þ þ ðx3
x3ref Þ ð20Þ

where a is the tuning parameter

Phase 2
Sliding-reachability condition design
s_ ¼
Kc s
Kd signðsÞ ð21Þ

where Kc and Kd are positive constant design parameters

4.1. Construction of the control law

By differentiating the sliding surface and equating with the reachability condition and solving out for u:
u ¼
Kc S
Kd signðSÞ
arx2 þ arx1 þ a_x1ref
qx1 þ x2 þ x1 x3 þ x_ 2ref
bx3
x1 x2 þ x_ 3ref ð22Þ
1222 A.M. Harb / Chaos, Solitons and Fractals 19 (2004) 1217–1224

20

15

10

x1
-5

-10

-15

-20
0 5 10 15 20 25 30 35 40 45 50
time

Fig. 6. Time history of the controlled system (with sliding-mode control after 25 s).

80

60

40

20

0
x1

-20

-40

-60

-80
-20 -15 -10 -5 0 5 10 15 20
x2

Fig. 7. State-plane of the controlled chaotic behavior with sliding mode controller.

4.2. Simulation results

Fig. 6 shows the time history of the controlled system with sliding mode control after 25 s. Notice that the trajectory
in phase-plane, as shown in Fig. 7, will spend some time ‘‘winding around’’ one equilibrium point, before jumping to the
other side and winding around the other equilibrium point for a while. This process goes on forever, with the trajectory
never crossing itself (in three-space), while when the sliding mode is attained the trajectory will goes to the steady state
and the system is stable.

5. Conclusions

The application of the PMSM in the industries became well known and very important. It is vary over a wide range,
i.e. low-performance applications to high-performance applications. PMSM can offer a high-performance drive by
utilizing the torques due the magnets and reluctance variation. Unfortunately, these kinds of machines have problems at
high speed. The study shows that the machine is experiencing a chaotic oscillation. A nonlinear controller based on the
backstepping nonlinear control theory is designed. The designed backstepping control signal is compared with sliding
mode control results. The study showed that the designed controller was so effective to eliminate the chaotic oscillation.
 A.M. Harb / Chaos, Solitons and Fractals 19 (2004) 1217–1224 1223

Appendix A. Methodology of backstepping

To explain the methodology of the backstepping controller, let me start from a special case that has been taken from
[22]. It is the integrator backstepping as shown in Fig. 8(a).
Let
g_ ¼ f ðgÞ þ gðgÞn ðA:1Þ

n_ ¼ u ðA:2Þ

where g and n are the state variables and u is the control signal. If the component of Eq. (A.1) can be stabilized by a
smooth state feedback control law n ¼ /ðgÞ, with /ð0Þ ¼ 0, then, the origin of g_ ¼ f ðgÞ þ gðgÞ/ðgÞ, is asymptotically
stable. Suppose that we know a smooth and positive definite Lyapunov function V ðgÞ that satisfies the inequality
oV
½f ðgÞ þ gðgÞ/ðgÞ 6
W ðgÞ ðA:3Þ
og
where W ðgÞ is positive definite. Adding and subtracting gðgÞ/ðgÞ on the right hand side of Eq. (A.1), one obtain the
equivalent representation shown in Fig. 8(b) and (c) and mathematically as follows:
g_ ¼ ½f ðgÞ þ gðgÞ/ðgÞ þ gðgÞ½n
/ðgÞ ðA:4Þ

n_ ¼ u ðA:5Þ

or
g_ ¼ ½f ðgÞ þ gðgÞ/ðgÞ þ gðgÞz ðA:6Þ
z_ ¼ v ðA:7Þ

where z ¼ n
/ðgÞ and v ¼ u
/_ . Eqs. (A.4) and (A.5) are similar to Eqs. (A.1) and (A.2), except that now the first
component has an asymptotically stable origin when the input is zero. This feature will be exploited in the design of v to
stabilize the overall system. Using the following Lyapunov function as
1
Vc ðg; nÞ ¼ V ðgÞ þ z2 ðA:8Þ
2

η
u
⌠ g(η ) + ⌠
⌡ ⌡

ƒ(.)
(a)

u ξ η
⌠ + g(η) + ⌠
⌡ ⌡

-φ(η) ƒ(.)+g(.)φ(.)
(b)

Ζ η
u
+ ⌠ g(η) + ⌠
⌡ ⌡

-φ` ƒ(.)
(c)

Fig. 8. (a) The block diagram of the system, (b) introducing uðgÞ, (c) backstepping
uðgÞ through the integrator.
1224 A.M. Harb / Chaos, Solitons and Fractals 19 (2004) 1217–1224

oV oV oV
V_c ¼ ½f ðgÞ þ gðgÞ/ðgÞ þ gðgÞz þ zv 6
W ðgÞ þ gðgÞz þ zv ðA:9Þ
og og og
Choosing v ¼
ðoV =ogÞgðgÞ
kz, where k > 0, one obtain, V_c 6
W ðgÞ
kz2 , which shows that the origin (g ¼ 0,
z ¼ 0) is asymptotically stable. Since /ð0Þ ¼ 0, so, the origin (g ¼ 0, n ¼ 0) is also asymptotically stable. Substituting for
v, z, and /_ , one obtain the state feedback control law
o/ oV
u¼ ½f ðgÞ þ gðgÞn
gðgÞ
k½n
/ðgÞ ðA:10Þ
og og
If all the assumptions hold globally and V ðgÞ is radially unbounded, we can conclude that the origin is globally
asymptotically stable.

References

[1] Krishnan R. Electric motor drives; modeling. analysis, and control. New Jersey: Prentice Hall; 2001.
[2] Rahman MF, Zhong L. Control strategies for buried magnet motors for extended low and high speed ranges. In: ICEMA 96
Conference Proceedings, September, 1996. p. 500–4.
[3] Jahns TM, Klinman GB, Neumann TW. Interior permanent magnet synchronous motors for adjustable-speed drives. IEEE Trans
Ind Appl 1989;22(4):738–47.
[4] Abed EH, Alexander JC, Wang H, Hamdan AH, Lee HC. Dynamic bifurcation in a power system model exhibiting voltage
collapse. In: Proceedings of the 1992 IEEE International Symposium on Circuit and Systems, San Diego, CA, 1992.
[5] Rajagopalan C, Sauer PW, Pai MA. Analysis of voltage control system exhibiting Hopf bifurcation. In: IEEE Proceedings of 28th
Conference on Decision and Control, Tampa, FL, 1989.
[6] Harb AM, Nayfeh AH, Chin C, Mili L. On the effects of machine saturation on subsynchronous oscillations in power systems.
Electric Mach Power Syst 1999;28(11).
[7] Hemati N, Leu MC. A complete model characterization of brushless DC motors. IEEE Trans Ind Appl 1992;28(1):172–80.
[8] Hemati N, Kwatny H. Bifurcation of equilibria and chaos in permanent-magnet machines. In: Proceeding of the 32nd Conference
on Decision and Control, San Antonio, Texas, December, 1993. p. 475–9.
[9] Li Z, Park J, Joo Y, Zhang B, Chen G. Bifurcations and chaos in a permanent-magnet synchronous motor. IEEE Trans Circ Syst
2002;49(3).
[10] Abed EH, Fu JH. Local feedback stabilization and bifurcation control, I. Hopf bifurcation. Syst Control Lett 1986;7:11–7.
[11] Abed EH, Fu JH. Local feedback stabilization and bifurcation control, II. Stationary bifurcation. Syst Control Lett 1987;8:467–
73.
[12] Harb A, Nayfeh A, Mili L. Bifurcation control for mitigating subsynchronous oscillations in power systems. In: 14th PSCC,
Sevilla, Spain, 24–28 June, 2002.
[13] Krstic M, Kanellakopoulus I, Kokotovic P. Nonlinear and adaptive control design. John Wiley & sons Inc.; 1995.
[14] Zaher A, Zohdy M. Robust control of biped robots. In: Proceedings of ACC, Chicago IL, USA, June 2000. p. 1473–8.
[15] Harb A, Zaher A, Zohdy M. Nonlinear recursive chaos control. In: Proceedings of the American Control Conference (ACC),
Anchorage, Alaska, May 8–10, 2002.
[16] Harb A, Ahmad W. Control of chaotic oscillators using a nonlinear recursive backstepping controller. In: IASTED Conference on
Applied Simulations and Modeling, Crete, Greece, June 2002. p. 451–3.
[17] Harb A, Natsheh A. Sliding-mode control of a chaotic system, Int J Nonlinear Dyn, submitted for publication.
[18] Yassen MT. Chaos control of Chen chaotic dynamical system. Chaos, Solitons & Fractals 2003;15(2):271–83.
[19] Laval L, MÕSirdi NK. Controlling chaotic and hyperchaotic systems via energy regulation. Chaos, Solitons & Fractals
2003;15(2):285–95.
[20] Li S, Tian YP. Finite time synchronization of chaotic systems. Chaos, Solitons & Fractals 2003;15(2):303–10.
[21] Aston PJ. Controlling chaos in systems with O(2) symmetry. Chaos, Solitons & Fractals 1998;9(8):1289–96.
[22] Khalil H. Nonlinear systems. 3rd ed. Prentice Hall; 2002.