Research on Reducing Torque Ripple of DTC Fuzzy Logic-based
Gao Sheng-wei1.2, Wang You-Hua1, Cai Yan2, Zhang Chuang1
1. Province-Ministry Joint Key Laboratory of EF and
EAR. Hebei University of Technology
Tianjin, China
gaoshengwei1978@126.com
2. School of Electrical Engineering and Automation
Tianjin Polytechnic University
Tianjin, China
Abstract—The direct torque control (DTC) technique has been
widely used as a control strategy for motor drive system. High
torque ripple is a major problem when the motor is running at
low speed in DTC system. By analyzing the causes of torque
ripple, it is can be found that the change of the speed error
signal is a good indicator to determine the situation of motor
torque ripple. In this paper, a DTC strategy with a fuzzy logic
controller has been proposed. The inputs of the fuzzy
controller are the speed error signal and the change rate of
current signal. The output is incremental value of the torque
hysteresis amplitude. Dynamic adjustment of the torque
hysteresis amplitude is the ultimate goal. The simulation
results show that the DTC system with hysteresis amplitude
adjustment fuzzy logic controller can reduce the pulsation of
the motor torque and flux, and improve the performance of
low-speed operating conditions.
Keywords-fuzzy logic; torque hysteresis; amplitude
adjustable; direct torque control
I. INTRODUCTION
The direct torque control (DTC) has the advantage of fast
dynamic torque response and has good robustness to the
changes of rotor parameters etc. But t he torque ripple is an
inadequacy of the system. The pulsation becomes more
apparent especially in the low-speed operating conditions
[1][2]. Due to this fact, some solutions have been developed
in recent years.
One approach is to use a high switching frequency
inverter [3]. It can reduce the harmonic content of the stator
current, and reduce torque ripple. However, the loss of power
is increase and the efficiency of the inverter is reduced. Also
the cost of the inverter with high frequency switching
characteristics is high. Another method is to use two parallel
connected inverters generate multiple non-zero voltage
vector [4]. This method solved the torque ripple to some
extent, but the structure of the inverter is complicated.
Moreover the zero-phase sequence current needs to be
monitored in this connection. In addition, there are multi-
level inverter method [5] and space vector modulation
method [6]. The cost and complexity will increase in the
former method. The torque ripple and flux ripple can be
978-1-4244-5848-6/10/$26.00 ©2010 IEEE
631
reduced in the latter method. But it will change the switching
frequency.
A DTC method with fuzzy logic controller has been
presented in this paper which can dynamically adjust the
torque hysteresis amplitude. The experimental results show
that the program can effectively reduce the torque ripple and
improve the torque control performance in the low-speed
running.
II. ANALYSIS OF TORQUE RIPPLE
With regard to the principle of DTC in detail, these
literatures [7][8] can be concerned. In general DTC program,
the switching cycle can be expressed as t=1/f, where f is the
frequency of switch. During the switching cycle, the picked
voltage vector has been applied to induction motor always.
Then the change of torque is always along the same direction
too. In the case of small torque error, the actual torque can
reach the reference value in a small part-time of the all
switching cycle under the effect of the selected voltage
vector. In the remaining time, the torque will continue to
change along the original direction under the effect of the
voltage vector, so the torque will have a greater ripple. As
discussed before, torque ripple directly related to the
amplitude of the hysteresis boundary, and this pulse is
particularly prominent when motor is in low speed running.
When a semiconductor device with high switching
frequency is used, the width the hysteresis can be reduced
appropriate. However, due to the thermal limit of the power
devices, the width of hysteresis must be large enough, so that
the actual switching frequency of the inverter is below a
certain level. Because the hysteresis is used to deal with
unexpected worst-case, a fixed small hysteresis width also
make system performance inevitable degradation especially
in low-speed region.
In this paper, the fuzzy logic-based torque controller with
adjustable hysteresis amplitude was presented. The input
signals of the controller are the speed error and rate of
change of current. The output is the increment of the torque
hysteresis amplitude. According to motor running state, the
b
torque hysteresis amplitude T can be dynamically adjusted.
So that the flux and torque ripple were reduced. Thereby the
DTC system performance of the work was improved.
 III. WAY OF MINIMIZING THE TORQUE RIPPLE angular velocity in voltage model, and the rotor flux of
In the stationary stator coordinate system, the current model is related to the rotor rotation angular velocity
asynchronous motor electromagnetic torque can be generated in the αβ coordinates. Based on inverse transformation
by the following equation[9]: principles in feedback model, the model reference adaptive
3 controller about the rotor flux can be constructed by using
Te = n p (is βψ sα − isαψ s β ) (1)
2 of the voltage model and current model. The rotor speed ω r
According to motor model, stator flux vector is expressed as: can be identified through flux errors proportional-integral
and feedback control in MRAC. The identifier structure was
ψ sα = ∫ (u sα − isα Rs ) dt shown in Fig.1.
ψ sβ = ∫ (u sβ − is β Rs ) dt
*
In Fig. 1, the rotor flux reference value ψ r can be
acquired from the rotor flux-voltage reference model in the
ψ s = ψ s2α + ψ s2β αβ coordinate system. The calculation process is indicated
ψ n
Where s is the amplitude of stator flux, p is the number in (5). The rotor flux estimate value ψ r can be acquired
、 、
of motor pole pairs, Rs is the stator resistance, ψ sα ψ sβ from the rotor flux-current adjustable model in the αβ

u sβ 、u 、i 、i
sα sβ sα
are corresponding stator flux, stator
coordinate system. The calculation process is indicated in
(6). Current model is placed on feedback channel. The
voltage, and stator current value in the αβ coordinate rotation angular velocity can be identified by the closed-
system. loop control method for rotor flux .
2
The dynamic compensation torque is defined as follow: Lr Lm
∫
∗
ψ rα = [ (u sα − Rs isα )dt − ( Ls − )isα ]

r
dω Lm Lr
M =J = Te − Tload (2)
dt Lr
2
Lm
∫
∗
Where Te is the motor torque, J is the transient inertia of ψ rβ = [ (u s β − Rs isβ )dt − ( Ls − ) is β ]
Lm Lr
the system, Tload is the torque load. The speed error signal
∗ 2 2
 r − ω . From (1)and (2), we can ψr = (ψ rα ) + (ψ r β ) (5)
can be expressed as e = ω ref
1
get (3): ψ rα = ( Lm isα − ωr Trψ r β )
d (ωref + e) Tr p + 1
J = Te − Tload (3)
dt 1
A constant speed reference signal can be expressed ψ r β = ( Lm isβ − ωr Trψ rα )
Tr p + 1
using the following expression:
ψ r =
2 2
d ω ref (ψ rα ) + (ψ r β ) (6)
=0
dt
If the load is constant, then it can be seen that the rate u sα
of change of speed error is closely related with the motor u sβ ψr
∗
torque from the (3). Equation 3 can be reduced to:
de Te − Tload
= (4) r
ω
dt J
The conclusion can be drawn from the (4) that the rate of isα
change in the speed error signal is indeed a good indicator to ψ
r
is β
assess the motor torque ripple.
IV. SPEED IDENTIFY
Figure 1. MRAC speed identifier
For the speed sensorless DTC system, speed of
observation is an important part. Based on the rotor flux
error between the voltage model and the current model, the
Parameter of speed can be identified by using the Model V. REDUCE TORQUE RIPPLE USING FUZZY CONTROLLER
reference adaptive controller (MRAC) structure and the
closed-loop feedback control theory. In this article, the amplitude of torque hysteresis is not
It can be got from the mathematical equations of induction fixed. But according to the level of torque ripple, the
motor that the rotor flux has nothing to do with the rotor magnitude of torque hysteresis can be adjusted by the fuzzy

632
controller. The two inputs of the fuzzy controller are the
speed error e1 and the stator current error e2 . They are 
ω ∆bT ∆Te
r
decided by the following equations. Is
 ( k ) − ω ( k − 1)
e (k ) = ω (7) Figure 2. Torque hysteresis controller
1

e2 ( k ) = I s ( k ) − I s ( k − 1) (8)

Where the stator current I s can be represented as

Quantity
2 2
Is = I as + I β s
The increment of the torque hysteresis amplitude can be
output by the fuzzy controller. Then the required torque
hysteresis amplitude can be got by formula 9. Finally, the
Figure 3. Input and output membership function
torque ripple ∆Te can be suppressed.
bT ( k ) = bT ( k + 1) + ∆bT ( k ) (9) TABLE I. Fuzzy rules of Torque hysteresis Controller
The specific way of working was shown in Fig. 2
e2
If setting the amplitude bT too small, the overshoot may ∆bT
NH NM NS ZE PS PM PH
touch the upper boundary. Reverse voltage vector may be
selected. Corresponding, the reverse voltage vector will NH NH NH NH NH NM ZE ZE
rapidly reduce the torque. The torque ripple will be
NM NH NH NH NH NM ZE ZE
stubbornly high too.
Based on experience and simulation, the fuzzy controller NS NM NM NS ZS ZE PS PM
is designed. For different motor speed and current values, a
e1 ZE NM NM NS ZE PS PM PM
appropriate torque hysteresis boundary can be selected in
order to reduce torque and flux pulsation. The specific PS NS NS ZE PS PS PM PM
modus operandi of the fuzzy controller can be carried out in
PM ZE ZE PM PH PH PH PH
accordance with Table 1. The membership functions of the
、
input e1 ( k ) e2 ( k ) , and output ∆bT ( k ) were all chosen in PH ZE ZE PM PH PH PH PH

the form of isosceles triangle, as shown in Fig. 3.
The semantic rules can be expressed by the following
examples:
If e1 is NH or NM, and e2 is NH, then ∆bT is NH: this
PH: positive high; PM: positive median; PS: positive small; ZE: zero;
NH: negative high; NM: negative median; NS: negative small;
TABLE II. Induction motor rated parameter

situation corresponds to the torque error has a large Parameters Rating

overshoot. To reduce the torque error, the value ∆bT should Rated Power 1.7kw
be significant reduced. Rated speed 1440rpm
Pole pairs 2
VI. SIMULATION RESULTS AND ANALYSIS
In this paper, the simulation about the DTC system
Stator resistance 2.25Ω
containing the fuzzy controller was executed by using Rotor resistance 2.5Ω
MATLAB software, and the simulation results has been Stator inductance 0.666h
compared with the general DTC system. The simulation Rotor inductance 0.666h
model works in low-speed (10 rd/s) status. Load torque is
15N/m. Other parameters were shown in table 2. The Mutual inductance 0.666h
simulation results were shown in Fig. 4 and Fig. 5. Inertia 0.02Kg ⋅ m2
Since the fuzzy controller can provide suitable amplitude
Sampling time 3E-6
according to system operation station, the torque ripple was
significantly reduced as shown in Fig. 4. And the pulsation
of the flux track was substantially reduced as shown in Fig. 5
when using fuzzy controller in DTC system.
It was drawn from the Fig. 4 and Fig. 5 that the DTC
system with the fuzzy controller is superior to general DTC
system in steady-state performance.

633
 Torque
(N/m) REFERENCES
[1] Lascu C, Boldea I, Blaabjerg F, “Variable-Structure Direct Torque
Control - A Class of Fast and Robust Controllers for Induction
Machine Drives,” IEEE Trans. J. Ind. Electron., Vol. 51, no.4, pp. 38-
45, Feb. 2004.
[2] Miroslaw Wlas, “Artificial Neural Network Based Sensorless
Nonlinear Control of Induction Motor,” IEEE Trans. J. Energy
Time (s) Time (s) Convers., vol. 20, no. 3, pp. 520-527, Dec. 2005.
[3] N Benaidja, “Identification of Asynchronous Machine Parameters by
Evolutionary Techniques,” Electric Power Components and Systems,
(a) (b) vol. 34, no. 12, pp. 1359 - 1376. Jan. 2006.
Figure .4 The waveform of torque ripple [4] Mustapha Messaoudl, “A Robust Sensorless Direct Torque Control of
(a). The waveform of torque ripple in general DTC system Induction Motor Based on MRAS and Extended Kalman Filter,”
(b). The waveform of torque ripple in the DTC system with Leonardo Journal of Sciences, Vol. 12, pp. 35-56. Jan. 2008.
the fuzzy controller [5] R.krishnan and A.S.bharadwaj, “A review fo parameter sensitivity
and adaptation in indirect vector controlled induction motor drive
systems,” IEEE Trans. J. Power Electron, vol. 6, no. 4, pp. 623-635,
Feb. 2007.
ψ sβ [6] M.Hinkkanen and K.Luomi, “Modified integrator for voltage model
flux estimation of induction motor,” IEEE Trans. J. Ind. Elevtaon.,
vol. 50, no. 4, pp. 818-820, Aug. 2003.
(Wb) [7] N Benaidja and N K henfer, “Identification of A synchronous
Machine Parameters by Evolutionary Techniques,” Electric Power
Components and Systems, vol. 34, no. 12, pp. 1359-1376, 2006.
ψ sα [8] Baburaj Karanayi, and Muhammed Fazlur Rahman, “Stator and rotor
ψ sα
resistance observers for induction motor drive using fuzzy logic and
(Wb) (Wb) Artificial neural networks,” IEEE Trans. J. Energy Convers., vol. 20,
(a) (b) no. 4, pp. 771-780, Jan. 2005.
Figure .5 The trajectory of stator flux [9] R.Portillo, M.M.Prats, J.I.Leon, J.A.Sanchez, and L.G.Franquelo,
“Modeling strantegy for back-to-back three level converters applied
(a). The trajectory of stator flux in general DTC system
to high power win trrbined,” IEEE Trans. J. Ind. Elevtaon., vol. 53,
(b). The trajectory of stator flux in the DTC system with
no. 5, pp. 1483-1491, Oct. 2006.
the fuzzy controller

VII. CONCLUSION
DTC strategies are simple and efficient control strategies
to the motor drive system. However some drawbacks
associated to the high torque ripple are present in this type of
control techniques. In this paper, a novel method of reducing
torque ripple has been put forward by analyzing the reasons
of torque ripple. It is the main character of the method that
the fuzzy logic controller combines with the DTC. The speed
error signal and the rate of change of current signal are the
input of the fuzzy logic controller. The torque hysteresis
amplitude increment is the output of the fuzzy logic
controller. The controller can dynamically adjust the torque
hysteresis amplitude. The simulation results show that the
method can effectively improve the shortcoming that the
ripple of torque in the conventional DTC system. The
method provides an effective scheme for reducing motor
torque ripple and noise in the low-speed operating
conditions.
ACKNOWLEDGMENT

（
The authors would like to thank Tianjin Science and
Technology Development for Higher Education Fund
）
20071119 for supporting this project.

634