-
Simulation Modelling Practice and Theory 18 (2010) 277–290
Contents lists available at ScienceDirect
Simulation Modelling Practice and Theory
journal homepage: www.elsevier .com/ locate/s impat
Hardware-In-the-Loop for on-line identification and controlof
three-phase squirrel cage induction motors
Ashraf Saleem a,*, Rateb Issa b, Tarek Tutunji a
a Mechatronics Engineering Department, Faculty of Engineering,
Philadelphia University, Jordanb Mechatronics Engineering
Department, Faculty of Engineering, Balqa’ Applied University,
Jordan
a r t i c l e i n f o a b s t r a c t
Article history:Received 8 June 2009Received in revised form 28
September2009Accepted 9 November 2009Available online 17 November
2009
Keywords:Identification and controlInduction
motorHardware-In-the-LoopARMA models
1569-190X/$ - see front matter � 2009 Elsevier
B.Vdoi:10.1016/j.simpat.2009.11.002
* Corresponding author. Address: P.O. Box 1, PhilE-mail address:
[email protected] (A.
This paper describes a strategy for identification and control
of three-phase squirrel cageinduction motors. The strategy in this
work is divided into 3 stages: on-line identification,off-line
controller design, and on-line control. First, the transfer
function is identified on-line. Next, the controller design is
performed in a pure simulation environment usingthe identified
transfer function. Finally, the designed controller is applied to
the real sys-tem.
Simulation and experimental results are presented to show the
validity of the proposedstrategy. Advantages of the proposed
strategy include high accuracy in the identified sys-tem,
simplicity, and low cost.
� 2009 Elsevier B.V. All rights reserved.
1. Introduction
Due to their simple structure, reliability of operation and
modest cost, the squirrel cage induction motors are the mostwidely
used electrical drive motors. The progress of semiconductor
technology in the last years made it possible for staticconverters
to be built at acceptable price, and therefore induction motors has
a future in variable speed drives. In inductionmotors, specific
demands are required at steady state and transient state operation.
These demands depend on load type andduty cycle of the motor. The
steady state condition is treated as a special case of the more
general solution [1].
The theory of the induction motor for dynamic conditions is
somewhat involved because of the rotating magnetic fields,the
spatial relationships of which depend on speed and load. On the
other hand, the equivalent circuit usually derived forsteady state
operation with sinusoidal voltages proves to be inadequate when
dealing with transients or when the motoris supplied from a static
converter.
The mathematical model to be used is tailored to the needs of
controlled drives. It incorporates most of the qualitativefeatures
of an actual motor but would not, be accurate enough for design
purposes [2]. Induction motors exhibit nonlineardynamic behavior
and therefore it is a challenge to establish an adequate
mathematical model for controller design pur-poses. Substantial
research in the past decades focused on the derivation of suitable
mathematical models in order to designappropriate controllers for
these motors [3–5].
Linear regression models have been generally used to approximate
the behavior of nonlinear systems. Some of thesemodels use the
black box concept to map input/output data patterns using adaptive
mathematical functions. Researchershave worked on identification of
induction motor parameters. Koubaa [6] and Castaldi et al. [7] used
a recursive predictionerror method and adaptive observers to
estimate the motor parameters, such as the rotor resistance, rotor
inductance, and
. All rights reserved.
adelphia University, 19392, Jordan. Tel.: +962 799979892; fax:
+962 64799037.Saleem).
http://dx.doi.org/10.1016/j.simpat.2009.11.002mailto:[email protected]://www.sciencedirect.com/science/journal/1569190Xhttp://www.elsevier.com/locate/simpat
-
Nomenclature
vds d-axis component of the stator voltage, Vvqs q-axis
component of the stator voltage, Vids d-axis component of the
stator current, Aidq q-axis component of the stator current, Ai0dr
d-axis component of the rotor current referred to the stator, Ai0qr
q-axis component of the rotor current referred to the stator, ALs
stator inductance, HL0r rotor inductance referred to the stator,
HR1 stator winding resistanceR02 rotor winding resistanceLm mutual
inductance between rotor and stator, HLss Ls + Lm, HL0rr L
0r þ Lm, H
xs stator electrical angular speed, rad/sxr rotor electrical
angular speed, rad/sxm 2P xr , rotor mechanical angular speed,
rad/sP number of polesTe electromagnetic torque, N mTm load torque,
N mJ equivalent moment of inertia, kg m2
278 A. Saleem et al. / Simulation Modelling Practice and Theory
18 (2010) 277–290
stator leakage inductance of a three-phase induction machine.
Their work used a complicated setup experiment based on aDSP tool.
Moreover, their aim was to estimate the motor parameters rather
than identify the transfer function.
The Hardware-In-the-Loop is an environment where virtual
components work in conjunction with real system’s compo-nents. It
is mainly employed to test a real control system on a virtual plant
in order to verify its performance before applyingit to the real
plant [8,9].
In this research, identification methods based on linear
regression methods are utilized on-line to develop transfer
func-tions of induction motors. These functions are then used to
design appropriate controllers under simulation environment.Once
verified, the controller is applied to the real AC motors. The
on-line identification and control are performed usingthe HIL
concept.
The 3-stage strategy proposed and implemented in this paper
contributes to the field of identification and control ofinduction
motors by clearly presenting organized steps for testing and
applying controllers. Other researchers conductedwork in the
parameter estimation, model identification, and controller design
of induction motors [3–7]. The implementa-tion of
Hardware-In-the-Loop (HIL) within the well defined strategy given
in this paper is an added value to the researchcommunity.
In previous works, system identification methods were developed
and applied to different engineering systems. Tutunjiet al. [10]
used impulse response data in a recursive gradient algorithm to
identify the transfer function of a DC motor andgyroscope.
Abedrabbo and Tutunji [11] presented identification model and
sensitivity analysis of hydrostatic transmissionsystem. Saleem et
al. [12] applied identification and control to a pneumatic servo
drives using mixed-reality environment.
This paper is divided as follows: Section 2 gives a brief
overview of induction motors model. Section 3 presents the
systemidentification and control and (Section 3.2) explains HIL
concepts used. (Section 3.3) demonstrates the proposed
strategy.Section 4 presents the experimental results and Section 5
concludes the paper.
2. Modeling and simulation of three-phase squirrel cage
induction motor
Many studies of the transient and steady state performance of
induction motors have used two axes (d–q) dynamic ma-chine model
for the solution of the motor performance equations [13,14], while
other studies have used a direct three-phasedynamic model that
seemed more convenient, due to the variables involved in such
modeling, in which they are the actualphysical quantities of the
motor [15]. Some authors have used dynamic model for small
perturbations and transfer function,or solutions for dynamic
behavior in complex symbolic form [16].
The steady state performances of the induction motors are
obtained using static model equations, derived from a dynamicmodel
by setting their derivatives to zero and solving the resulting
motor equations for the motor variables.
The state-space model of induction motor in standard form, with
respect to a synchronously rotating d–q coordinates[17,18], is as
follows:
x ¼ ids i0dr iqs i0qr xm
h iT— state vector
-
A. Saleem et al. / Simulation Modelling Practice and Theory 18
(2010) 277–290 279
Y ¼ ids i0dr iqs i0qr xm Te
h iT— output vector
u ¼ vds 0 vqs 0 Tm½ � — input vectorD ¼ ½0� — direct
transmission matrix
A ¼
xs þ P2 xmK2 P2 xmK1LmLss
R1Lss
K1 � R02
LmK2 0
R1Lss
K1 � R02
LmK2 �xs � P2 xmK2 � P2 xmK1
LmLss
0
� P2 xmK1LmL0rr
xs � P2 xmK1 �R1Lm
K2R02L0rr
K1 0
� R1Lm K2R02L0rr
K1 P2 xmK1LmL0rr
�xs þ P2 xmK1 0
� aLmi0qr
J 0aLmi0dr
J 0 0
266666666664
377777777775
— state matrix
B ¼
K1Lss
0 � K2Lm 0 00 K1Lss 0 �
K2Lm
0
� K2Lm 0K1L0rr
0 0
0 � K2Lm 0K1L0rr
0
0 0 0 0 � 1J
2666666664
3777777775� input matrix
C ¼
1 0 0 0 00 1 0 0 00 0 1 0 00 0 0 1 00 0 0 0 1
�aLmi0qr 0 aLmi0dr 0 0
2666666664
3777777775
— output matrix
a ¼ 34
P — constant
K1 ¼LssL
0rr
LssL0rr � L
2m
— constant
K2 ¼L2m
LssL0rr � L
2m
� constant
Referring to the described state-space model of induction motor,
a Simulink drive system block diagram was built in order tosimulate
the dynamic behavior as shown in Fig. 1.
The parameters of the induction motor may change during the
operation of the drive system, causing deviations betweenthe
corresponding signals of the model and the motor. The stator
resistance and rotor resistance change with temperature.This is a
relatively slow process as the thermal time constants are large.
Parameters changes produced by magnetic satura-tion affect the
stator reactance, rotor reactance and mutual reactance to some
extent [28].
The induction motor parameters were determined by testing the
motor under no-load and locked rotor conditions [35].The real
parameters of the induction motor were used in the simulated system
and are given in Table 1.
Fig. 2 shows the speed response comparison between the
mathematical model and the real system for step, trapezoidal,and
multi-trapezoidal inputs respectively. Results reveal that the
mathematical model was able to follow the real systemwith
negligible steady state error.
Fig. 1. Simulink model of the induction motor.
-
Table 1Induction motor data.
Parameter Value
Stator resistance R1 17.22 XStator reactance X1 19.2 XMutual
reactance Xm 252.6 XRotor resistance referred to the stator R02
18.69 X
Rotor reactance referred to the stator X02 19.2 XNominal voltage
V1n 220/380 VNominal speed nn 2870 rpmNominal output power P2N 300
WNominal current I1n 0.8 APower factor cos u1n 0.74Nominal
frequency f1n 50 HzNumber of poles P 2Moment of inertia J 0.02 kg
m2
280 A. Saleem et al. / Simulation Modelling Practice and Theory
18 (2010) 277–290
3. System identification and control through HIL
This section describes well known induction motors
identification and control techniques. Furthermore, it presents
theproposed strategy that based on HIL for the identification and
control of induction motors.
3.1. System identification
Mathematical models can be constructed using analytical
approach, such as physics laws, or using experimental ap-proach.
The later is usually used when critical information related to the
system is missing.
System identification is the field of approximating dynamic
system models from input/output patterns acquired throughphysical
experiments. This includes the experimental setup, data
acquisition, determination of an appropriate model, andthe design
of an algorithm for parameter convergence [9].
The system is driven by input ‘‘control” variables u(t), with
added noise v(t), to give an output y(t).The target is to establish
a mathematical model that mimics the original system and therefore
minimizes the error be-
tween the system and model outputs. System identification
incorporates the following steps:
1. Experiment design. This includes the choice of lab equipment
to be used such as computers, DAQ, and interface.2. Model structure
determination. The choice of the model can range from nonparametric
models, such as frequency analysis
and fuzzy, to parametric methods, such as difference equations
and neural networks.3. Experiment run. This is usually done by
exciting the system with an input signal (pulse, sinusoid, or
random) and measur-
ing the output signal over a specified time interval.4.
Algorithm choice and run. The algorithm used for convergence can
vary from simple one-shot least squares, recursive least
squares to advanced multi-structures such as back propagation.5.
Validation of results. The output of the identified model is
compared to the original system through different and ‘new’
input signals.
The identification is referred to on-line when steps 3 and 4 are
done concurrently and off-line when the results of step 3are
recorded and at a later stage step 4 is applied to the collected
data.
Researchers have applied several models and algorithms to
identify dynamic induction motor dynamics. Koubaa [6] useda linear
parameter estimation technique, based on recursive least squares,
to determine the rotor resistance, self inductanceof the rotor
winding, and the stator leakage inductance of a three-phase
induction. The model developed was based on stea-dy-state equations
of induction motor dynamics. Huang et al. [26] applied genetic
algorithms for parameter identification offield orientation control
induction motors. The model’s parameters were estimated using the
motor’s dynamic response to adirect on-line start. Burton et al.
[27] used neural network models and an algorithm based on random
search of the errorsurface gradient to identify and control
induction motor stator currents. A comparison to the
backpropagation algorithmwas also provided.
Holtz and Thimm [28] applied on-line identification technique
based on the evaluation of the dynamic response of theinduction
motor to a Pulse Width Modulation switch sequence to estimate the
rotor time constant and other machineparameters. Once the critical
parameters were estimated, a control method based on field oriented
control was applied toa vector-controlled induction motor drive.
Attaianese et al. [29] presented two on-line identifiers, based on
the model ref-erence adaptive control theory, for the rotor
parameters of the field-oriented induction motor drive. The
experimental re-sults used DSP system. Levi and Wang [30] proposed
a method for on-line mutual inductance identification in
vector-controlled induction machines.
-
0 5 10 15 20 25 30 35 40 45 500
0.5
1
1.5
2
2.5
3
3.5
4
Time
Spee
d (v
)
Math Model
Real Output
0 50 100 150-0.5
0
0.5
1
1.5
2
2.5
3
Real System
Math model
Spee
d (v
)
Time
0 100 200 300 400 500 6000
0.5
1
1.5
2
2.5
3
Time
Spee
d (v
)
Comparison of the mathematical model and the real system
responses to a multi trapezoidal input signal
Math Model
RealSystem
a b
c
Fig. 2. Mathematical model vs. real system responses to (a)
step, (b) trapezoidal and (c) Multi-trapezoidal inputs (speed scale
= 0.9 mv/rpm).
OutputInput
SystemModel
Delay
Delay
Delay
Delay
Fig. 3. ARMA model general structure.
A. Saleem et al. / Simulation Modelling Practice and Theory 18
(2010) 277–290 281
In this paper, the identification models used are based on
Auto-Regressive Moving-Average (ARMA) models. ARMAmodels are
difference equation models that map input–output data. The general
structure for an ARMA model is given inFig. 3.
The output is a linear difference equation of current and past
inputs and past outputs. This ARMA equation is given next
ŷk ¼Xnj¼1
ajyk�j þXmi¼0
biuk�i ð1Þ
where uk and yk are the inputs and outputs at discrete-time k,
and aj and bi are the ARMA parameters.
-
282 A. Saleem et al. / Simulation Modelling Practice and Theory
18 (2010) 277–290
The Z-Transform of the ARMA(m, n) model is calculated to yield
the transfer function of the model which is given by
Z yk �Xnj¼1
ajyk�j
( )¼ Z
Xmi¼0
biuk�i
( )ð2Þ
) HðzÞ ¼ YðzÞUðzÞ ¼
b0 þ b1z�1 þ � � � þ bmz�m1� a1z�1 � � � � � anz�n
ð3Þ
The main advantage of using ARMA models is the direct mapping of
those models into transfer functions. Once the trans-fer function
is established, they can replace the system in designing
controllers in a simulated environment.
The target is to minimize the Sum Square Error (SSE) between the
system and ARMA model outputs as given by
E ¼ 12
XKk¼1ðŷk � ykÞ
2 ð4Þ
In this paper, the recursive least square (RLS) algorithm is
used to minimize the SSE by approximating the optimum
ARMAparameters. Several methods such as steepest descent, gradient
search, Newton, and Levenberg–Marquardt can be usedwithin the RLS
to update the model parameters.
As previously indicated, identification techniques depend on two
main choices: model structure and algorithm. Theadvantage of the
ARMA structure is that it provides the transfer function parameters
which is essential to the proposed strat-egy. Other structures such
as frequency analysis or neural network can identify the system but
cannot provide the results ina transfer function format. As for the
algorithm, the RLS was used because it can have accurate results
with good convergenceproperties. These RLS features are well
established in the literature [9,34]. This feature is important in
this work as the sys-tem is identified in real time.
3.2. Induction motor control
The development of semiconductor power conversion technology has
led to a widespread use of electric drive with induc-tion motors,
and new control systems for these motors.
There are certain limitations in the use of a control method for
induction motors control system. It is difficult to identifycommon
approaches to the synthesis of induction motors control system. The
control of induction motors is complicatedsince it involves several
interacted variables. The most significant of which are as
follows:
(1) The electromagnetic torque of the induction motor is
determined by the product of the two resulting vectors of
theelectromagnetic parameters of the stator and the rotor and
therefore a function of four variables.
(2) There is a strong interaction magnetizing forces of the
stator and the rotor, the mutual state of which varies
contin-uously with the rotation of the rotor.
Fig. 4. Photograph of AC motor experiment set up.
-
A. Saleem et al. / Simulation Modelling Practice and Theory 18
(2010) 277–290 283
Induction motors, together with controlled power converters are
complex and nonlinear control systems [19]. Fullmathematical
description of such systems is quite cumbersome and hard to apply
to practical control systems. However,the design of induction
motors control systems is a proliferation of simple methods based
on the principle of cascade con-trol. The use of these techniques
allows reasonable opportunity to simplify the mathematical
description of the controlsystems.
In an adjustable AC drive system, for speed control, the
induction motor normally requires variable-voltage
variable-fre-quency power supply (voltage source inverter) [20,21],
or variable-current variable-frequency power supply (current
sourceinverter) [22,23].
The voltages and currents are described by complex
two-dimensional vectors. The mathematical treatment is different
inthe case of voltage control and current control of the induction
motor. Current control is normally preferred. However, thereare
some problems associated with current control [24] and are
summarized below:
(1) Since the stator equation is eliminated, the control system
is dependent on the rotor equation, and thus is criticallydependant
on the rotor inductive time constant. This time constant varies
with rotor resistance, which varies withrotor temperature, and it
is difficult to measure the rotor temperature. A more serious
problem occurs when the rotoriron goes into magnetic saturation and
the rotor time constant collapses. Then the whole control system
may collapse.
(2) There is a nonlinear relationship between current and
magnetic field strength. This nonlinearity is a source of
torqueripple.
FilterSignal
Generator
OutputPort
Input Port
ARMA Model
SIMULATION ENVIRONMENT
REAL SYSTEM
DAQ INTERFACE
TACHOMETER
AC MOTORINVERTER
PWM
RLS Algorithm
Fig. 5. Block diagram for on-line system identification based on
HIL.
-
284 A. Saleem et al. / Simulation Modelling Practice and Theory
18 (2010) 277–290
A simple physical law proves that the induction motor should be
voltage controlled instead of current controlled [25].Voltage
control automatically eliminates the effect of the nonlinear iron
magnetization curve. Only a small correction termhas to be
modulated according to the stator voltage drop caused by the
nonlinear magnetization current. This correctionterm is most
important at low speed operation of the motor.
3.3. Proposed strategy based on Hardware-In-the-Loop
The HIL is an environment where software components within a
simulation program are worked in conjunction withreal system
components implementing an integrated system. This setting gives
the capability of monitoring the system’sbehavior by observing its
response using virtual monitoring components [31,32]. Usually, HIL
is used for testing
0 0.5 1 1.5 2 2.5 3
0
0.2
0.4
0.6
Time (secs)
Actual Output (Red Line) vs. The Predicted Predicted Model
output (Blue Line)
0 0.5 1 1.5 2 2.5 3
-0.06
-0.04
-0.02
0
0.02
0.04
Time (secs)
Error In Predicted Model
Spee
d (v
)er
ror
0 0.5 1 1.5 2 2.5 3
0
0.2
0.4
Time (secs)
Actual Output (Red Line) vs. The Predicted Predicted Model
output (Blue Line)
0 0.5 1 1.5 2 2.5 3-0.02
0
0.02
0.04
Time (secs)
Error In PredictedModel
Spee
d (v
)er
ror
a
b
Fig. 6. Impulse response of real system compared to the
identified model output for (a) 6th order model and (b) 22nd order
model.
-
0 20 40 60 80 100 120 140 160 180 2000
0.5
1
1.5
2
2.5
3
3.5
Time
Spe
ed (
v)
6th order model
Real System
Fig. 7. Step response of the real system vs. the identified 6th
order model.
0 50 100 150 200 250 300-0.5
0
0.5
1
1.5
2
2.5
3
Time
Spee
d (v
)
6th order
Math model
Real System
0 100 200 300 400 500 6000
0.5
1
1.5
2
2.5
3
Time
Spee
d (v
)
6th order
Math model
Real System
a b
Fig. 8. Comparison of real system’s, mathematical model’s, and
identified model’s responses to (a) trapezoidal input and (b)
multi-trapezoidal input.
A. Saleem et al. / Simulation Modelling Practice and Theory 18
(2010) 277–290 285
hardware controllers on software models. However, in this
research, software components are used to identify
physicalsystems.
The strategy adopted in this research was divided into three
successive stages: on-line identification, off-line
controllerdesign, and on-line control. First, the induction motor
system was connected to the computer through DAQ where an
RLSalgorithm was used to identify the transfer function on-line.
Next, the controller design was performed in a pure
simulationenvironment using the identified transfer function from
the first step. Once the controller design was completed, it was
ap-plied to the motor to perform on-line control. All stages were
applied in MATLAB/Simulink environment.
The advantages of the proposed method include the following:
accuracy in the identification and system control, optimiz-ing time
resources and minimizing the cost as a result of on-line
identification and off-line controller design, increased
flex-ibility of the controller, and user friendly.
The second step of the proposed strategy, which is the off-line
controller design, offers the following improvements
overtraditional design strategies:
1. The induction motor to be controlled will not be used during
the experimentation of the controller design and parametertuning
and therefore the down time of the induction motor will be
minimized. This might be a crucial time saving issuewhen the motor
is used production line. Equally important, damage to the motor due
to inappropriate parameter valuesis avoided.
2. The off-line tuning gives much more flexibility to the
designer in order to design and test many controllers in a
simulationenvironment. This will result in obtaining an optimized
controller in minimum time.
-
286 A. Saleem et al. / Simulation Modelling Practice and Theory
18 (2010) 277–290
4. Results and analysis
The 3-stage proposed strategy was implemented and tested in the
Laboratory. This section provides the details of thestrategy
implementation.
The experiment setup used a P4, 3 GHz desktop computer with a
National Instruments Data Acquisition (DAQ) card6036E that has a
sampling rate of 200 KS/s. The DAQ card had 16 analog inputs, 2
analog outputs, and input voltage rangeof ±10 V. The induction
motor system was composed of a motor, pulse width modulator,
inverter, and tachometer. Theinduction motor is a three-phase
squirrel gage with the following ratings: nominal power 300 W,
nominal speed2870 rpm, nominal torque 1 N m, nominal line voltage
380 V, nominal frequency 50 Hz. The output from the tachometerwas
0.9 mv/rpm. The experimental test rig photograph is depicted in
Fig. 4.
In this setup, one input and one output signals were used. The
output was generated from the computer and sent from theDAQ via D/A
channel to the motor system. While the input signal was read from
the tachometer via the A/D channel. Bothsignals were used in a
MATLAB/Simulink environment where the software algorithm was
implemented. Real time windowstarget (rtwt) toolbox within MATLAB
was utilized for real time system interface.
4.1. Stage one: system identification
Fig. 5 shows the block diagram for the adopted on-line system
identification stage. One way to implement the systemidentification
is to use an ideal impulse as the excitation signal. However, for
implementation purposes a pulse is used in-stead. The period of the
pulse should be short enough to represent an ideal impulse but also
long enough to activate themotor to its desired settling speed.
Several tests and experiments were conducted to choose the pulse
width value to be0.15 s. A pulse with 5 V amplitude was applied
from the computer to the induction motor system. Measured
tachometeroutput was sampled at 0.05 s and combined with the input
signal to formulate patterns. These patterns were used in aRLS
algorithm to establish the estimated system transfer function via
ARMA models. The sampling rate was selected in orderto obtain
appropriate number of samples (e.g. 300 samples) to be used for the
RLS algorithm.
The choice of ARMA model order affects convergence and therefore
several orders were simulated in order to choose themost
appropriate. Two of those model orders, 6th and 22nd, are presented
in Fig. 6 where motor response vs. identified
Table 2statistical data for the real, the identified model, and
the mathematical model responses.
Input signal Statistical criteria
System Min Max Mean Standard deviation
Trapezoidal Real �0.0473 2.884 1.446 1.235Identified Model
�0.01737 2.858 1.447 1.282Mathematical Model 0 2.856 1.529
1.139
Multi-trapezoidal Real �0.05157 2.889 1.891 1.126Identified
Model �0.01737 2.858 1.891 1.163Mathematical Model 0 2.856 1.916
1.083
Fig. 9. Identified model controlled response off-line with PID
controller (P = 3.68, I = 0.02, D = 0).
-
A. Saleem et al. / Simulation Modelling Practice and Theory 18
(2010) 277–290 287
model responses to the pulse input are provided. Also, the
errors between the motor and model responses are provided inthe
same figure. Both order models gave a good approximation of the
original motor response. Although the 22nd ordermodel resulted in a
slightly smaller error, the 6th order model was used throughout
this paper because lower order modelsconverge faster.
The identified model was further tested with a step input. The
model response was compared to the real system as shownin Fig. 7.
It is observed that the identified model had a very close behavior
to the real system in terms of rise time, overshoot,and settling
error.
Furthermore, trapezoidal and multi-trapezoidal inputs were
applied to the identified model, real system, and mathemat-ical
model. Results are presented in Fig. 8. Again, results showed that
the 6th order identified model used does very well inapproximating
the real system behavior. In fact, the identified model did better
than the mathematical model. These resultsare presented in Table 2.
Comparison among the three responses showed the mean and standard
deviations of the identifiedmodel is almost equal to the real
system. Therefore, the model was considered appropriate to be
utilized in control systemdesign via HIL.
Speed Profile
OutputPort
Input Port
SIMULATION ENVIRONMENT
REAL SYSTEM
DAQ INTERFACE
TACHOMETER
AC MOTORINVERTER
PWM
Controller
-+
Fig. 10. Block diagram for on-line control via HIL.
-
Fig. 11. Simulink block diagram for on-line control.
Fig. 12. Real system speed response with on-line PID controller
(P = 3.68, I = 0.02, D = 0).
288 A. Saleem et al. / Simulation Modelling Practice and Theory
18 (2010) 277–290
4.2. Stage two: simulation control
The 6th order model obtained from the identification stage was
used to replace the real system within Simulink in orderto design
the suitable controller. At this stage, the designer can utilize
any control method that is appropriate on the iden-tified model and
tests the results in a simulated environment. The scope of this
paper is concerned with the strategy de-scribed in Section 3 rather
than selecting among different control methodologies.
In this paper, a controller based on
Proportional–Integral–Derivative (PID) [33] was applied due to its
simplicity and widepractice in the industry. A trapezoidal speed
profile was given as the desired response. The PID tuning was
applied usingZiegler–Nichols tuning method. The final PID
parameters were set to P = 3.68, I = 0.02, and D = 0. The result is
presentedin Fig. 9 which shows that the controller was successful
in forcing the model to follow the desired speed.
4.3. Stage three: on-line control
Once the controller was optimized on the identified model in a
simulated environment, the identified model was thenreplaced by the
real system for the final controller test. In this stage, the
designed controller was applied to the real systemthrough Real time
windows target within Simulink according to the experimental setup
explained in Section 4. Fig. 10 showsthe block diagram for the
on-line control. Note that this setup differs from the system
identification setup, Fig. 5, in that itapplies a controller block
instead of the ARMA/RLS block.
Fig. 11 shows the Simulink/software program used to run the
on-line controller experiment. Note here that the PID con-troller
is based in the PC software. The ‘‘Real System” in Fig. 11 is a
Simulink block that is interfaced with the DAQ. The DAQin turn
acquires the I/O data from the real system, induction motor, and
forwards it to this block.
-
Fig. 13. Real system speed response to a multi-trapezoidal input
with on-line PID controller (P = 3.68, I = 0.02, D = 0).
A. Saleem et al. / Simulation Modelling Practice and Theory 18
(2010) 277–290 289
For the final step, speed profiles that are similar to the
profiles used in stage two are applied. The speed response of
thereal system is shown in Fig. 12 for the trapezoidal and Fig. 13
for the multi-trapezoidal reference profiles. The real
system,induction motor, was able to follow the desired reference
signals with minimum error. This means that the designed
con-troller in stage two performed well on the real system. It is
worth noting that further fine tuning can be applied at this
stage.
The effects of parameters, such as temperature, on the induction
motor model are established in the literature [28]. Theidentified
model used in this paper establishes a reference model which is
used for off-line controller initial design. The thirdstep in the
proposed strategy, the on-line control, will compensate for the
parameter variation through on-line fine tuning.Therefore the
performance of the identified model is adequate for our purpose
within our strategy.
5. Conclusions
In this paper, a strategy to identify and control an induction
motor is provided. The proposed strategy is composed ofthree
stages. In the first stage, ARMA models were used to identify the
motor transfer function. In the second stage, the iden-tified model
was used in a simulation environment in order to design the
controller off-line. In the last stage, the virtual con-troller was
applied to the induction motor via HIL environment.
In the identification stage, an impulse was applied to the motor
and the speed was measured. Patterns of measured in-put–output
signals were used in a software environment to identify the
transfer function by optimizing its parametersthrough an RLS
algorithm. In the control stages, the identified model was adopted
within a simulation environment to testand optimize appropriate
controllers. These controllers were then applied to the real
system.
Experimental results using PC/DAQ and MATLAB/Simulink were
carried out. Results showed that the proposed strategywas able to
identify and control the motor behavior for different speed
profiles.
References
[1] E.Y.Y. Ho, P.C. Sen, Digital simulation of PWM induction
motor drives for transient and steady-state performance, IEEE
Trans. Ind. Electron. (1986) 66–77.
[2] W. Leonhard, Control of Electrical Drives, Springer-Verlag,
Berlin, Heidelberg, 2001.[3] Bartomiej Beliczynski, Lech Grzesiak,
Induction motor speed estimation: neural versus phenomenological
model approach, Neurocomputing 43 (1–4)
(2002) 17–36.[4] H. Madadi Kojabadi, Simulation and experimental
studies of model reference adaptive system for sensorless induction
motor drive, Simulat. Model.:
Pract. Theory 13 (6) (2005) 451–464.[5] B. Bose, Adjustable
speed AC drive – a methodology status review, Proc. IEEE 70 (2)
(1982) 116–135.[6] Y. Koubaa, Recursive identification of induction
motors, Simulat. Model.: Pract. Theory 12 (2004) 363–381.[7] P.
Castaldi, W. Geri, M. Montanari, A. Tilli, A new adaptive approach
for on-line parameter and state estimation of induction motors,
Control Eng. Pract.
13 (2005) 81–94.[8] S. Ertem, Y. Baghzouz, A fast recursive
solution for induction motor transients, IEEE Trans. Ind. Appl.
IA-24 (1988) 138–144.[9] T. Soderstom, P. Stoica, System
Identification, Prentice Hall, 1989.
[10] T. Tutunji, M. Molhem, E. Turki, Mechatronics systems
identification using an impulse response recursive algorithm,
Simul. Model.: Pract. Theory 15(2007) 970–988.
[11] S. Abedrabbo, T. Tutunji, Identification and analysis of
hydrostatic transmission system, Int. J. Adv. Manuf. Technol.,
doi:10.1007/S00170-007-0966-3.2007.
[12] A. Saleem, S. Abedrabbo, T. Tutunji, On-line identification
and control of pneumatic servo-drives via mixed reality
environment, Int. J. Adv. Manuf.Technol. (2008).
[13] J.H. Chang, B.K. Kim, Minimum-time minimum-loss speed
control of induction motors under field-oriented control, IEEE
Trans. Ind. Electron. IE-44(1997) 809–815.
[14] G. Singth, V. Pant, Analysis of a multiphase induction
machine under fault condition in a phase redundant AC drive system,
Electr. Mach. Power Syst. 28(2000) 577–590.
[15] S.N. Ghani, Digital computer simulation of three-phase
induction machine dynamics – a generalized approach, IEEE Trans.
Ind. Appl. IA-20 (1988) 106–113.
http://dx.doi.org/10.1007/S00170-007-0966-3.2007http://dx.doi.org/10.1007/S00170-007-0966-3.2007
-
290 A. Saleem et al. / Simulation Modelling Practice and Theory
18 (2010) 277–290
[16] B.K. Bose, Scalar decoupled control of induction motor,
IEEE Trans. Ind. Appl. IA-20 (1983) 216–225.[17] O.I. Okogo, MATLAB
simulation of induction machine with saturable leakage and
magnetizing inductances, Pacific J. Sci. Technol. 5 (1) (2003)
5–15.[18] Hussein Sarhan, Rateb Issa, Modeling, simulation and test
of inverter induction motor drive system with improved performance,
J. Eng. Sci., Asuit Univ.
33 (5) (2005).[19] G.A. Capolino, Non-linear Modeling and
Control of Some Classes of Induction Machine, Springer, Berlin,
1989.[20] K. Shyu, Hsin Jang Shick, A new switching surface sliding
mode speed control for induction drive system, IEEE Trans. Power
Electron. 11 (4) (1996)
660–667.[21] S. Wade, M. Dunnigan, B. Williams, A new method of
rotor resistance estimation for vector – controlled induction
machines, IEEE Trans. Ind. Appl. 44
(2) (1997) 247–257.[22] T. Ito, T. Yamaguchi, R. Ueda, T.
Mochizuki, S. Tokata, Analysis of field orientation control of
current source inverter drive induction motor systems, IEEE
Trans. Ind. Appl. 1A-19 (2) (1983) 206–210.[23] M. Salo, H.
Tuusa, A vector-controlled current-source PWM rectifier with a
novel current damping method, IEEE Trans. Power Electron. 15 (3)
(2000)
464–470.[24] P. Hussels, M. Mehne, Synthesis of cycloconverter
and current-source-inverter for high power high performance drives,
Electr. Eng. 74 (1) (1990).[25] Ragnar Jonsson, Improved Control
Method of an AC Induction Motor and Device Therefor, European
Patent Office, 1988.[26] K.S. Huang, Q.H. Wu, D.R. Turner,
Effective identification of induction motor parameters based on
fewer measurements, IEEE Trans. Energy Convers. 17
(1) (2002) 55.[27] Bruce Burton, Farrukh Kamran, Ronald G.
Harley, Thomas G. Habetler, Martin A. Brooke, Ravi Poddar,
Identification and control of induction motor
stator currents using fast on-line random training of a neural
network, IEEE Trans. Ind. Appl. 33 (3) (1997) 697–704.[28] Joachim
Holtz, Thomas Thimm, Identification of the machine parameters in a
vector-controlled induction motor drive, IEEE Trans. Ind. Appl. 27
(6)
(1991) 1111.[29] Ciro Attaianese, Alfonso Damiano, Gianluca
Gatto, Ignazio Marongiu, Aldo Perfetto, Induction motor drive
parameters identification, IEEE Trans. Power
Electron. 13 (6) (1998) 1112.[30] Emil Levi, Mingyu Wang, Online
identification of the mutual inductance for vector controlled
induction motor drives, IEEE Trans. Energy Convers. 18
(2) (2003) 299.[31] X. Wu, H. Figueroa, A. Monti, Testing of
digital controllers using real-time hardware in the loop
simulation, in: 2004 IEEE 35th Annual Power
Electronics Specialists Conference, 2004, PESC 04, vol. 5, 20–25
June 2004, pp. 3622–3627.[32] Lu Bin, Wu Xin, H. Figueroa, A.
Monti, A low-cost real-time hardware-in-the-loop testing approach
of power electronics controls, IEEE Trans. Ind.
Electron. 54 (2) (2007) 919–931.[33] K.H. Ang, G. Chong, Yun Li,
PID control system analysis, design, and technology, IEEE Trans.
Control Syst. Technol. 13 (4) (2005) 559–576.[34] G. Franklin, J.
Powell, M. Workman, Digital Control of Dynamic Systems, third ed.,
Addison-Wesley, 1998.[35] A. Fitzgerald, Charles Kingsley, Stephen
Umans, Electric Machinery, McGraw Hill Book Company, 1983.
Hardware-In-the-Loop for on-line identification and control of
three-phase squirrel cage induction motorsIntroductionModeling and
simulation of three-phase squirrel cage induction motorSystem
identification and control through HILSystem
identificationInduction motor controlProposed strategy based on
Hardware-In-the-Loop
Results and analysisStage one: system identificationStage two:
simulation controlStage three: on-line control
ConclusionsReferences