A Review on The AC Servo Motor Control Systems

Abstract

AC Servomotors are widely used in the industries for the control of static and dynamic loads. Precise control of position, speed, and torque are the main issues with the AC Servomotor. AC Servomotors are highly demanded by the industries to have a precise response under dynamic load conditions. Many control techniques are commercially available for the control of AC Servomotor under static and dynamic load conditions. However, all of these control techniques have advantages and limitations. Many investigations are done on the control of AC Servomotor, but comprehensive surveys on the control of AC Servomotor were still limited. In this paper, most of such commercially available control techniques are investigated, discussed, and compared.

Full text

VOL. 19, NO. 2, 2020, 22-39
www.elektrika.utm.my
ISSN 0128-4428
22
A Review on the AC Servomotor Control Systems
Abdul Wali Abdul Ali 1*, Fatin Asmida Abdul Razak2 and Nasri Hayima2
Centre for Electric Energy and Automation (CEEA) Faculty of Engineering, Multimedia University, Cyberjaya, Malaysia
*Corresponding author: walikdr17@gmail.com, Tel: 601114320521
Abstract: AC Servomotors are widely used in the industries for the control of static and dynamic loads. Precise control of
position, speed, and torque are the main issues with the AC Servomotor. AC Servomotors are highly demanded by the
industries to have a precise response under dynamic load conditions. Many control techniques are commercially available for
the control of AC Servomotor under static and dynamic load conditions. However, all of these control techniques have
advantages and limitations. Many investigations are done on the control of AC Servomotor, but comprehensive surveys on
the control of AC Servomotor were still limited. In this paper, most of such commercially available control techniques are
investigated, discussed, and compared. It was found that all of the available control techniques have drawbacks, such as step
response issues, waveform oscillatory errors and fluctuations, instability of the system, switching losses, sensitivity to
parameter variations and external disturbances, and low dynamic responses. There is no control technique available which
could solve all the issues simultaneously.
Keywords: AC Servomotor, Control Stability, Dynamic Load, Static Load, Step Response, Control System
© 2020 Penerbit UTM Press. All rights reserved
Article History: received 31 March 2020; accepted 26 July 2020; published 29 August 2020.
1. INTRODUCTION
A servomotor is a motor employed for the control of
position or speed in the closed-loop control systems. The
functions of the servomotor are to turn over a wide range
of speed and also to perform the control position and
speed instructions given. DC and AC servomotors are
utilized in applications due to their machine structure in
general. When the condition is low power and variable
speed, the AC servomotors are the ones favored in control
systems due to its control capabilities [1–3]. Besides, the
applications of the AC servomotors can be found in
conveying technology, printing, wood processing, textile
industry, plastics industry, food and packaging industry,
packaging and filling plants, and machine tools. There are
two types of AC servomotors available which are a
squirrel cage asynchronous and a permanent magnet
synchronous. In the field of control of mechanical
linkages and robots, research works are mostly done only
on the DC motors. A literature review regarding the AC
servomotor motion control and tracking characteristics is
limited since the AC servomotor technology is
respectively new. AC servomotors applied in some
research articles are overviewed herein. Moreover, Lin et
al. [15] have conducted a study on simulation and
dynamic performances of electrical machines; the
transformer, the DC machine, the polyphase induction
machine, the polyphase synchronous machine, and the
single-phase induction machine with an electric machine
simulation program. Takahashi [24] presented an
environment to model and simulate mechatronic devices;
electrical motors (AC and DC), electronics, fluid power
and control, and mechanical systems. Also, Seki et al.
[10] have described a study on a high-performance servo
drive system and characteristics of a salient pole
permanent magnet motor. Wang et al. [32] analyzed the
performances of AC servo drives utilizing synchronous
and asynchronous motors. A mathematical model is given
with the control scheme and supported by experimental
results. Besides, a study done by Sravya [11] completed a
robust control of an AC induction servomotor for a
motion-control system. Plus, X. Li [43] introduced a
comparative study between two permanent magnet AC
machines by using numerical simulation and also
experimentally. The simulations were included for a two-
joint rigid robot directly driven by induction motors.
Furthermore, Zhang et al. [37] have worked on high
dynamic speed sensorless AC drives. Experimental
verification was achieved with an induction motor. The
on-line-mode parameter tuning was applied to eliminate
the steady-state error. Ollervides et al. [18] have dealt
with the problem of mechanical resonance in a system
comprising a permanent magnet synchronous servomotor
and a load with experimental verification. A study on the
performance of PD controlled servo systems was
conducted by F. Lin et al. [15]. A mathematical model
was presented with simulation results and the
experimental implementation was included. In a motion
control and implementation, X. Lx [13] performed
motion control of robots by induction motors to trace the
given directions by introducing a current controller. A
complete PM AC servomotor model was developed by P.
Puttaswamy et al. [21] too, which implemented a neural
network self-tuning PI control scheme. The authors
presented the experimental results in the study.
2. CONTROL METHODS
Many control methods are available for AC Servomotor.
Some of the control systems focus on the position control

Abdul Wali Abdul Ali et al. / ELEKTRIKA, 19(2), 2020, 22-39
23
only while some control systems focus on the speed or
torque control, respectively. However, some control
systems focus on the control of position, speed, and
torque of the AC Servomotor at the same time.
2.1 POSITION CONTROL
In this part of the paper, the control techniques for the
position control of the AC Servomotor are discussed.
These control methods are only used for the position
control of the motor.
M. Sazawa and K. Ohishi designed a control system
based on a fast continuous path tracking mechanism for
the flexible position control in 2009 [2]. In this control
strategy, the saturation of the torque and the coordinated-
motion were taken into account. For the starting and final
position, large accelerated or decelerated torques were
needed; thus, the PI controller had small loops of the
control acceleration to provide some output. This limited
output method, compensated for the corresponding output
and constructed the limiter of the speed controller. On the
other hand, a time of deceleration, which was the duration
from the commencing brake time to the finishing brake
time, was proposed, and within this time, the control
system was decelerated utilizing the highest torque. The
commencing time of the brake was known by the torque
of the load and the targeted position. The distance of the
movement was delighted as the error of the position until
the targeted position. This control technique had a good
step response, and it did not have oscillatory errors and
overshoot. Moreover, it also reduced the tracking error
that occurred because of the dynamic load. However, it
did not fully eliminate the error. Its tracking radius error
was 0.02 rad. The system complexity and high gains of
the controller may have consumed a lot of energy due to
the PI controller. The system settled at 1.041 seconds,
whereas under conventional fast continuous path tracking
control, the system settled at 1.046 seconds.
M. Vijayakarthick et al. proposed a new Modified
Repetitive Control Strategy to track a position control
performance of AC Servomotor in 2012 [3]. The
proposed control method was used to track the reference
signal and reject the load signal and was designed based
on the Principle of Internal Model. IMP states that perfect
tracking can be achieved, and a signal can be completely
rejected if the closed-loop model is stable. For the high-
frequency signal, the stability of the system could be
affected by the noise. This problem could be solved by
adding a low pass filter to the control loop. The
sensitivity function could be taken into consideration
because the stability had a direct relation with the
sensitivity; in addition, a rational factor was incorporated.
Moreover, this control method was not affected by
external disturbances or load variations. This repetitive
controller did not affect the stability of the system. It had
a good adaptive tracking ability, and in a practical
situation, it was very robust opposing the noises. Its
theories of control and knowledge of the system were not
required to be very difficult or complicated. Furthermore,
according to the experimental results, there was an
existence of the tracking error in the output waveform,
but this error was less than PD and conventional
repetitive control strategy. The error started reducing after
70 seconds, subsequently existed some delay in the
tracking response of the periodic reference trajectories.
The settling time of the system was around 100 seconds
under a 2% tolerance level.
In order to meet out the problem with Iterative
Learning Control (ILC), an Enhanced Iterative Learning
Control (EILC) method was proposed by S. Sathishbabu
and P. K. Bhaba in 2012 [4]. The ILC control scheme
features were learning filter and low-pass filter; the rate at
which the error signal was converged could be known
from the learning gains. This version was improved by
EILC, where the main idea in this version was that the
error signal was applied by the Filter before the control
signal for the next trial was computed. Practically, the
developed processes may have caused some unexpected
variations from the main or actual process. Subsequently,
few disturbances caused by the Filter could occur and
affect the closed-loop stability in the presence of high
frequencies. However, this issue could be reduced by
adding a Low Pass Filter in the loop of the control.
Furthermore, this control method also provided good
robustness opposing to the noises in the practical. It did
not need a complicated environmental model, theories of
control, and information of the system. Even though the
absolute error in the tracking error response of this
system was less than ILC, some errors still existed and
should have been further reduced. Its tracking error
response for the first five iterations was huge. The
absolute error at the second iteration was 1200 radians.
After the second iteration, the error started to reduce, and
finally, after eight iterations, it reached up to 80 radians
and stayed at the same value for the rest of the iterations.
This control system also was affected by the load
variation.
Although the reference waveforms were periodic in the
practical, the conventional controllers were not able to
track it properly with a good performance. To reduce this
problem, A.Ali and A. Alquhali suggested an AIMC
strategy [5]. This strategy states that the control scheme
can be developed perfectly if the controller is designed
based on the definite scheme of process Gp(s). This
indicated that the identifier needed to have complete
knowledge about the process. The feedback was only
needed if the knowledge about the process was
incomplete or inaccurate. By using mathematical
equations, the transfer function Gp(s) of the AC servo
system was identified. Plus, Bode plot techniques were
utilized to analyze the closed-loop transfer function of the
motor. In general, it was assured that IMC achievements
were better than the PID controller, specifically, in terms
of rising time, settling time, overshoot, and its greater
gain margin. Even though IMC was better than PID in
terms of performance, it still required further
improvement to have the ability to control the dynamic
load and remove the transient and steady-state errors
fully. The settling time of the AIMC controller was 8.13
seconds under a 2% tolerance level.
A research paper conducted by N. Wang and W. Lin in
2016 initiated a robust tracking control method [6]. This
method guaranteed the design of the controller so that the
AC Servomotor could have accurate tracking of the
position. The proposed strategy was implemented on the
base of vector control together besides the law of
mechanics of the Screw Ball and the development of a
recent dynamic model. The formulation of precise
position tracking control was done. Using Lyapunov

Abdul Wali Abdul Ali et al. / ELEKTRIKA, 19(2), 2020, 22-39
24
access, a convex optimization approach was used to
formulate the existence conditions for the controller. To
control the motion of the Motor, some more conditions
needed to be considered, such as non-linear effects of the
friction and unknown disturbances in the system. The
authors also stated that the Neural Network could also be
used to make the position control more accurate and to
approximate the non-linear friction effects. The adaptive
control method could be used to compensate for the
parameter variations. Besides, this proposed control
strategy had a good dynamic response but could lead to
transient errors in the step response. The system position
control waveform settling time was 12 seconds under a
2% tolerance level in uncertainty cases. Also, there were
some position tracking errors as the output signal never
tracked the input signal exactly at any point. Hence, the
system fault tolerance control and network framework
needed further improvement.
2.2 SPEED CONTROL
In this part of the paper, the available control techniques
for speed control of the AC Servomotor are discussed.
These control methods are only used for the speed control
of the motor.
A study conducted by F. F. Cheng and S. N. Yeh in
1993 suggested a novel fuzzy logic controller (FLC)
utilized to control the speed of the AC Servomotor [7].
The authors used a microprocessor to reduce the circuit
components for price reduction and the enhancement of
reliability. In order to control the motor torque
instantaneously, field orientation was indirectly endorsed,
including the current-regulated PWM and VSI. The
functions of membership of FLC were known by the
normalization of the input variables and the predefinition
of the linguistic codes. For the high and fast performance
AC servo system, a rules table was proposed based on the
following two concepts. Firstly, if the change in the error
of the angular velocity was zero, and the output was set to
a value, the controller would maintain the output.
Secondly, change in the output was conducted based on
the magnitude and the error in the angular velocity and
the change of error in the angular velocity signs, which
means whenever there was a departure among both of the
values that were existing, the output recovered to set. The
implementation of 49 rules was done by utilizing the
linguistic codes; the output variables were also composed
of linguistic codes. When there was equality between the
input and the output variables, the relation led to unity.
The relation was 0.5 degrees if near to each other and
zero degrees, for others. Furthermore, this control method
is a low cost, simplified hardware, and reliable. It has a
good dynamic response, and the switching loss was small,
minimized the chattering phenomenon, and the harmonics
in the current. However, it also had some drawbacks,
such as viscous friction existed in practical. The control
system contained reduced torque pulsations, step
response issues, complex calculations, and heavy
parameter variations.
Besides, E. Yolacan and M. Aydin conducted an
experiment on the vector-based speed control of
permanent magnet AC servomotor with FEA in 2012 [8].
According to the authors, in the Vector Control (Field
Oriented Control) of the AC machine, three-phase AC
stator current was transformed into a d-q axis rotating
reference frame. This method used the current and speed
variables of the motor, which were compared with the
reference signal. The PI regulators were used for reducing
the error to zero that was generated from the differences
between the measured and the reference signal. The
vector control technique was implemented based on the
phase and current magnitude. Plus, the three-phase
currents of the motor were measured and sent to the
controller. The controller would calculate the convenient
signals to generate the Pulse Width Modulation (PWM)
signals with respect to the phase current and the position
of the rotor. In this experiment, the Ds1104 dSpace
control board was used. Based on the results obtained
from the experiment, this control method has advantages
such as a good step response, no overshoot, a rise time of
1.5 seconds, and a settling time of 2.85 seconds. Also, the
dynamic torque response in this control system was very
low, and it was sensitive to parameter variations in the
control system.
In addition, I. Kadan et al. designed a low-speed
control of AC servomotors in no-load condition [9]. The
designed control method utilized adjustable torque ramp
option. During the mode of torque control, the torque
ramp option was specified to know the supplied control
voltage of the drive proportional to the specific amount of
torque. A proper control signal could be sent to the drive-
by assigning a proper value to the torque ramp option.
Since the control signal was properly measured, and the
prior amplitudes of the noises were known. Hence, this
helped to reduce the effects of the noise in the later
iterations. Moreover, the motor torque was regulated by
scaling down the least noise affected signal by the drive.
As the drives were strongly safeguarded, the noise could
not affect the internal signal of the drive. Dynamic
modeling was done for the rotating shaft of the motor.
Also, the motor speed could be well regulated at desired
values, and the effect of noise could be reduced to some
extent through sending a properly measured signal to the
drive. The control voltages had some fluctuations because
of a key attached to the shaft. The key weight became
noticeable because there was no load attached to the
motor shaft, and the motor speed was very low. Even
though the torque ramp parameter reduced the noise in
the system to some point, the motor speed still contained
some fluctuations and oscillations.
Another research paper was published by Y. Seki et
al., in 2015, suggesting another control strategy [10]. The
suggested control strategy improved the voltage
utilization on the constraint of the region of flux
weakening by the combination method of flux weakening
control and the Inverter modulation arrangement. When
saturation in the voltage appeared, the control of flux-
weakening operated utilizing the d-axis current and
terminated the saturation in the voltage. Saturation in the
voltage occurred when the voltage vector crossed the
output limitation. In order to keep this vector in the
inscribed circle, the desired d-axis referenced current was
known by the inverter control scheme. Furthermore, the
speed control arrangement of the Servo system was
considered by the phenomenon of the windup and the
control methodology of the conventional flux weakening.
The speed response tracked the reference speed and
determined the reference q-axis current. If the reference
q-axis current was restricted by the current limiter, the

Abdul Wali Abdul Ali et al. / ELEKTRIKA, 19(2), 2020, 22-39
25
feedback saturated q-axis current could correct the PI
controller. Since the Inverter output voltage had a limit,
therefore when the reference voltage crossed this limit,
the inverter would not be able to produce the referenced
output. To avoid instability, saturation in the voltage was
considered to correct the integral calculation of the
current controller. The d-axis current controller was not
employed by the anti-windup strategy of the control. This
was because the de-coupling control was disturbed by the
feedback of the voltage saturation. In short, this control
method brought the improvement in the Inverter voltage
utilization, the reduction of copper loss, and good
transient response. The copper loss was reduced
compared to the conventional differential signal control
method, but it still contained at the quantity of 12% under
25% load condition and 14% under 75% load condition.
The d-axis current used for the control of the flux
weakening was reduced from 0.99A to 0.71A. On the
other hand, the expending voltage restrictions caused the
voltage saturation to occur. The speed and current
responses contained oscillatory errors which may affect
the accuracy of the control system.
Speed control of electric drives using Active
Disturbance Rejection Control was initiated by Sravya K
et al. in 2016 [11]. In this control method, first of all, the
Servomotor was modeled using mathematical equations.
Phase voltages, flux linkages, currents, electromagnetic
torque, rotor angle, and speed were obtained from the dq0
equivalent circuit of the motor. In this control strategy,
three ADRC loops were designed; two were used for
current control, and another one was used for the purpose
of speed control. Furthermore, firstly, the input reference
signal, was tracked by the tracking differentiator.
Afterward, the Tracking Differentiator (TD) organized
the process of the transition and produced the derivative
value of the signal that was supplied at the input. The
deviations of the input could be smoothened in order to
reduce the overshoot at the output. The Extended State
Observer (ESO) was utilized to evaluate the system
disruption and uncertain model’s effects. Also, Non-
Linear State Error Feedback (NLSEF) was used to
produce the control rules for the system by combining the
output of ESO and TD, where ESO being a fast tracker,
compensated some disturbances. For that reason, NLSEF
and TD were replaced with proportional parameters to
meet the desired responses. In short, ADRC control could
compress the unknown disturbances internally or
externally more accurately than PI regulators. ADRC was
better than conventional PI regulators with respect to
rising time and settling time. Overall, the improvement in
the speed response and dynamic speed response of ADRC
based control of AC Servomotor were still needed.
Moreover, a study on the design of the Fuzzy Logic
Controller (FLC) for speed regulation of Permanent
Magnet Synchronous Motor dually driven system
considering the complexity of fuzzy designs was carried
out by B. Shikkewal and V. Nandanwar in 2012 [12]. The
alternate methodology in designing of the fuzzy system
was to tune by the Universe of Discourse method. This
involved modifying the rules and also the number of rules
in which a scaling factor was employed before each input
and after each output of the control structure. In many
cases, this scaling factor was taken by trials and error
methods. For the study purposes, two structures, namely
standard design, and the case design were considered.
The standard structure was designed with 49 rules
consisting of seven membership functions for each
variable (error and rate of change of error) designed by a
conventional procedure. Meanwhile, in the case design,
the structure consisted of Nine rules with Three
membership functions designed by tuning with the
Universe of Discourse method. Plus, the Mamdani Fuzzy
Inference type was used for a Fuzzy Inference System
scheme. Both scenarios were compared and tested for
different loads and wider ranges of speed. It was inferred
from the analysis that both scenarios resulted in a similar
performance, and thus, a number of rules could minimize
the FLC system’s complexity. Besides, the major
advantage of fuzzy logic control was the capability to
handle impreciseness, its high adaptability, and free from
mathematical modeling (because of its flexibility to
operate with linguistic schemes). However, the setback
with the design of the FLC scheme was that it required
some intuitive understanding of the process. The
simulation analysis also expressed that tuning by alternate
approach resulted in faster rise time. The scaling factor
played a prominent role in stability and oscillations. Thus,
the scaling factor had to be addressed with proper care to
avoid some fluctuations in the speed response.
Apart from improving the dynamic performance, in
order to simplify the tuning methodology, the
conventional ARDC scheme was decoupled, such that
disturbance rejection was independent of the reference
tracking. This control method is suggested by X. R. X. Lx
et al. in their research paper in 2015 [13]. According to
the authors, the analysis was carried out on the PMSM
drive for better regulation of speed. In the conventional
ADRC control structure, the proportional coefficient
parameter contained the coupled effect on tracking and
disturbance rejection. The underlying phenomenon
behind the proposed scheme was the estimation of total
disturbance rather than the estimation of target
disturbance alone. Plus, the velocity feedback contained
noise, which was generally eliminated by using Low Pass
Filter. However, the introduction of filter affected the
system closed-loop performances due to the lag
associated with it. Hence, a velocity estimator was used
to improve performance. The existing framework of ESO
was analyzed for different structural possibilities, and the
modified Linear Extended State Observer was proposed.
With the LESO scheme, tuning became easier by making
the disturbance attenuation effect separated from the
proportional coefficient parameter. This resulted in the
decoupled ADRC scheme. The proposed scheme was
analyzed, and the validation was carried out using a real-
time implementation of the control algorithm using d-
SPACE DS1103. The proposed scheme was analyzed for
different tracking and disturbance rejection performance.
According to the results obtained, it was inferred that in
the case of conventional ADRC, changes in the
proportional coefficient parameter affected the
disturbance rejection property to a great extent. In this
case, a perfect trade-off was maintained for optimal
tracking and disturbance rejection, whereas in the
decoupled ADRC scheme, the LESO bandwidth and
proportional coefficient were independent of each other,
and the system could have better performance with easy
tuning. However, it was also inferred that bandwidth of

Abdul Wali Abdul Ali et al. / ELEKTRIKA, 19(2), 2020, 22-39
26
ESO affected the tracking performance with the
introduction of ripples in the case of conventional ADRC,
and was not affected in the case of decoupled ADRC. In
the case of the decoupled ADRC, the speed and current
response of the system contained a lot of oscillations.
2.3 POSITION, SPEED AND TORQUE CONTROL
In this part of the paper, the control techniques used to
control the position, speed, and torque of the motor
simultaneously are discussed.
A control method used to control the position, speed,
and torque of the motor simultaneously was proposed by
Z. Chen et al., in 2001 [14]. In this method, multi-layer
momentum neural networks of back-propagation joined
with the PID controller were used to control the
Servomotor. The multi-layer neural network contained
Three layers, which were the input, hidden, and output
layers. Each signal flows over a weight known as
synaptic weight. The polarities of weights depended on
the acceleration of the signals. All the input signals were
accumulated and transferred to the output signal through
the summing node using a transfer function. The transfer
function could be either step type or threshold type. It
would transfer one if the input outpaces the threshold
value or else zero. The desired pattern could be achieved
by training this network. Plus, since the scattered
intelligence devoted by the weights, therefore this
network was able to learn. Moreover, this control strategy
had a fast dynamic performance and there was no
overshoot. However, using this control system, the rise
time of the speed response would be two seconds, which
would lead the system to be slow.
F. J. Lin et al. developed the Three-position
controllers, which were Variable Structure Controller
(VSC), Variable Structure Adaptive (VSA) controller,
and Variable Structure Direct Adaptive controller
(VSDA) [15]. These three controllers worked on the same
principle. The VSC and VSA were improved and
replaced by VSDA. This control system contained Two
main parts, which were the VSDA controller and the
Servomotor drive system. The conventional drive system
was supported by the VSDA controller, which provided
great advantages. VSDA contained the PD controller,
variable structure law, and direct adaptive law. In the
design of the VSDA controller, for the practical
operation, it was hard to achieve the uncertainty
constraints in advance. At the same time, a conservative
design control scheme caused the chattering issue more
severe. For that reason, the Direct Adaptive law was used
in this controller, designed to estimate the uncertainty
bounds. If there were no load disturbances and parameter
variations in the system, a very small positive value of
variable structure law would be enough to keep the
amplitude of chattering small and hold the system stable.
However, if external load disturbances and parameter
variations were present, departing from the sliding
surface, it would need the regular update of variable
structure law, which was created by the adaptive structure
to drive the trajectories of the system back towards the
sliding surface quickly. One of the advantages of this
control system is that it had good robustness to parameter
variations and external disturbances. Also, it did not
require knowledge of uncertainty bounds. This control
method also had good performance with reduced
chattering, but the chattering still existed and should have
been even reduced more. The load regulation response of
this control system was not fully improved as it was a
little bit sluggish.
Moreover, an Iterative Learning control method was
supported by the fact that the same recreating effects
would lead to the same error at each time or each run, the
control signal could be adapted by the recording of such
errors. In the next run, this control signal was assigned to
the process for the reason to reduce the error. So, the
error decreased with the increasing number of trials. This
control method introduced a new method to remove the
zero-crossing and the effect of the dead-time from the
PWM Inverter’s output waveform. For the creation of an
appropriate dead-time compensation voltage, a learning
algorithm and PI regulators were plugged in parallel for
the creation of dead-time compensation voltage. With the
help of the motor current and the angle of the rotor, the
actual currents were forced by the learning algorithm for
the tracking of the reference current. With a start in
learning, the error between the actual current and
reference current decreased with time. With further
progress in the process, d-q current approached to
reference current without the requirement of a proper
dead time compensating signal generation. Moreover,
after achieving an acceptable convergence, the learning
process was ended, and the learned compensating signals
were recorded. By making the learning process offline for
different values of the load current and the operating
frequency, the learning process became faster. These
signals were stored in memory. At the time of the online
operation, the operating frequency of the drive and the
currents of the load were identified with the proper
selection of these stored patterns. These were superposed
to the output of the PI regulator. To take the external
disturbances into account, Periodic online learning might
be carried [16]. In short, as the proposed learning control
and PI control operated together, for that reason, the
oscillations and fluctuations in the current waveform
were removed. However, removing these oscillating
errors took a few cycles that may have affected the
current response.
Another study conducted by S. Zorlu et al. in 2006
also suggested a new control technique using a custom-
designed Motion Control Card [17]. In this technique, the
implementation of control is done utilizing a Personal
Computer (PC) with a Motion Control Card. The torque
could also be controlled using the Motion Control Card
(MCC). In this application, a Low Pass Filter was used to
filter the Encoder signals. A PM DC generator was
connected at the output shaft of the Servomotor, and then
a dynamic load was connected at the output of the
generator to observe the nature of the Servomotor under
the different load conditions. Field-effect Sensors were
used in the power circuit of the Servomotor to measure
the separate voltages and current. Also, DC bus voltage
and motor currents were measured using voltage and the
current Sensors. The Inverter was driven by four 15V
isolated voltage sources, and the MCC was connected to
the PC using Industry Standard Architecture (ISA) bus.
Eight-bit port at the output of the Inverter signals, Analog
to Digital Converter (ADC), and Encoder chips were
utilized to measure the voltage and current. ADC was
used in order to have Zero phase shifts between the

Abdul Wali Abdul Ali et al. / ELEKTRIKA, 19(2), 2020, 22-39
27
signals. The speed was measured using the HTCL 2016
chip, and an algorithm was built to realize the rotor’s
initial position. Based on the results obtained, the current
and voltage waveforms contained oscillations, which may
have led the system to be noisy. This control method had
a fast-dynamic response where its settling time was very
less, but an average error still existed in the speed
response of the system. The Motor at half load could not
achieve the targeted speed as fast as it could achieve at
full load.
Furthermore, an electronic feedback control system
based on the embedded Digital Signal Controller (DSC)
Microcontroller is also another control approach designed
by J. Ollervides et al. in 2011 [18]. The control method
consisted of a Switch power amplifier which fed Actuator
inputs that attributed a dual full H-bridge driver. The
driver IC was embedded with the DSC Microcontroller.
Two quantities of Hall effect current Sensors were used
to obtain the feedback current. Besides, the angular
velocity and position of the shaft of the motor were
measured by an incremental optical encoder. The Hybrid
Stepper Servomotor was used as an Actuator. The two-
phase hybrid Stepper Servomotor contained two electrical
components and one mechanical component. These
components were connected by the back-EMF
(Electromotive Force) and the transmission of torque.
This control method was aimed for position tracking and
torque control. Torque could be controlled if the
difference between actual current and the desired current
was zero. In order to keep this difference as Zero or to
make the stator current trajectories follow the desired
current trajectories, the input control voltage was
designed. The switch power amplifier was used to
reproduce the signal that was fed to the Motor, and it also
worked as an integrated circuit for current sensing.
Lastly, this electronic control system of the Servomotor
provided enormous benefits in portable drive system
applications. On the other side, because of the electronic
drive network, the cost and complexity of the control
system increased and also consist of many angular errors
and oscillatory errors in the shaft rotor position of the
motor.
Besides, L. Zhang et al. proposed another control
system based on the XC164CM microcontroller in 2012
[19]. In this control system, the Incremental Encoder was
used to achieve the angular position of the AC
Servomotor, since the speed was determined from the
position that changed over time. This value and the
feedback value were both sent to a speed-PID controller
for the adjustment. This speed-PID controller was used to
adjust the output (d, q-axis current), and then this output
value was sent to the current-PID adjuster. Three-phase
stator current was detected by ACS712 (the hall device).
Plus, the Clark and Park transformations were utilized to
convert the three-phase current to d-axis and q-axis
currents. This current was transferred to the current-PID
controller, and the resulted signal was transferred to the
Inverter. A Space Vector Pulse Width Modulation
(SVPWM) signal was generated from the CCU6 unit,
whose input was the two-phase current, which was the
conversion of the Inverter output using Park inverse
transformation. This SVPWM signal was then sent back
to the Inverter in order to generate a PWM signal to
control the Motor. The benefit of this control system is
that this control system had simplified hardware design
and good reliability. On the contrary side, the Inverter
output contained a lot of oscillatory errors, which may
have led the system to be noisy. Also, since the PID
controller was used, because of its high gains, it may have
consumed more energy and led the system to be noisy in
practical as proved by the results in Chapter Four.
Apart from that, a PLC controller used to control the
three-phase AC servomotor drive was initiated by M.
Sreejeth et al. in 2012 [20]. The authors explained that a
PWM signal from an AC drive was sent to the stator
windings of the motor. The Motor was controlled by the
PLC output, established on the mentioned ladder logic.
Thus, the Motor parameters were able to change online
and offline. However, there were some specific tasks that
needed to be done in the online mode and offline mode.
In online mode, the Motor parameters such as average
load, speed, and feedback pulses, etc. were recorded. In
offline mode, the Motor drive could be directly fed by the
parameters. Also, for offline control mode, the parameters
could be directly fed to the Servomotor drive. A data bus
named Modbus-RTU required field data from the Motor
and transmitted it to the PC for the process. The data bus
was also used to scatter the logical signals towards the
Servo drive obtained from the PLC. Besides, there was an
error between the reference speed and original speed from
which the reference torque was generated utilizing the
speed controller. During the constant torque operation,
the ratio of the reference torque and the torque constant of
the Motor was treated in order to calculate the reference
quadrature axis current. The phase currents of the stator
were achieved by employing the park’s inverse transform
with the help of the rotor position feedback and d-q axis
currents. The output was fed to the hysteresis-band-
controller after comparing the reference current with the
actual current using the PWM converter. The hysteresis-
band-controller output was used as a gate pulse for the
Inverter, and the variable frequency and voltage of the
Inverter were supplied to the stator windings of the motor
in order to achieve the commanded speed. Plus, with high
speed and high current, the Total Harmonic Distortion
(THD) level was lower, and Motor was in a good
performance. However, the harmonics in the line current
increased when the Motor was running at a lower speed.
Hence there were effects on the precision and
performance of the drive. At 25% load and 1500rpm
speed, the harmonics in the line current were 71.7%, and
at 3000rpm, this distortion reduced to 61.7%. The speed
response of the system was affected by fluctuations.
The servo motor also could be controlled by utilizing a
Multi-layer Neural Network, as discussed in the research
paper written by P. S. Puttaswamy and K. D. Dhruve in
2013 [21]. This method adopted the use of a neural
network for the adaptive direct torque control of AC
Servomotor. The Artificial Neural Network had higher
precise control than a conventional PID controller. This is
due to the fact that the proposed method could tune the
conventional PID controller parameters more accurately
with the neural network technology. To demonstrate the
speed control loops, a PWM inverter fed the motor. The
speed control loop that had a PID controller produced the
quadrature axis current. This quadrature current produced
the electromagnetic motor torque. This method was
known as DTC. The d-q frame was transformed into the

Abdul Wali Abdul Ali et al. / ELEKTRIKA, 19(2), 2020, 22-39
28
a-b frame, and the resulting currents were fed to the
motor. The advantage of this control method is it has a
good speed control performance for larger and smaller
variations in Motor parameters and the load conditions.
Also, the motor achieved the required speed after a small
overshoot and some delay. The performance of the
system was further improved for small parameter
variations rather than large parameter variations.
In addition, J. Yin et al. proposed another servo motor
controller, which utilizes a Fuzzy Adaptive PID based on
the DSP [22]. This control strategy was adopted using
Three control loops, which were the current loop, speed
loop, and position loop. This control strategy played a
role in improving the dynamic response of the system and
adjusting the faults at the slow speed. This control
strategy was designed based on the linear motor position
motion law, according to which a large error could be
quickly eliminated by increasing the weight of the error
control function. To eliminate the overshoot of the system
with an increase in error, the weight of the error-change
control action was increased. Also, the addition of a
special Analog to Digital (A/D) acquisition Chip helped
to improve the precision of the system, which could
recognize the feedback currents and the collection
conversion. The system used a Chebyshev type II filter
for digital filtering. In short, this control system had a
wide speed range and did not need the modeling of the
object that was going to be controlled. The positioning
accuracy was high, but the position control curve
contained small overshoot errors, and the rise time was
two seconds. Plus, the settling time was 2.5 seconds, and
the proposed control procedure led to the frequency of the
system to fluctuate.
Another control approach was proposed by K.
Matsuura et al. in 2014 [23]. According to the authors, in
the AC Servomotor, the phase currents were detected by
the Current Sensors. Three Current Sensors were used to
measure the three-phase current. Sometimes two Current
Sensors were also used (such as u-phase and w- phase).
However, the problem with this method was that these
sensors had characteristics variations such as gain and
offset variations. The current ripples were caused due to
this issue. To overcome this problem, a DC-link Current
Sensor was practiced. After the measurement of this
current, the construction of a three-phase current was
done utilizing a three-phase algorithm. For the three-
phase current to be not affected, the current sensor should
have had an offset, but if there was gain variation, the
current would be affected. Hence, it was a necessity to
recognize and compensate for the variations in the gain of
the sensors. Moreover, a reconstruction circuit was
obtained by using multiple RC series circuits. Six
samples and hold circuits were needed to achieve the
dynamics of the DC-link current for each Switch. Also,
the Inverter was driven using the SVPWM technique, so
that the convenience of achieving the dynamics in the
DC-link current prior and after the switching was
increased. Lastly, by using this method, there was no
occurrence of the current ripples. This is because only a
single sensor was utilized for the detection of the current,
and secondly, the current measurement gain deviation
was compensated. However, there was a presence of
electrical parameter variations in the practical work, and
some current variations in reconstructed current still
existed, which could affect the dynamic behavior of the
system.
T. Takahashi and I. Rectifier designed a controller
using a single chip motion control engine IC [24]. The
authors used the IRMCK201 IC, which was a single Chip
entire solution for the closed-loop torque and speed
control of the AC drives. This engine did not require
complex AC servo algorithm development, and with the
use of this IC, a complete control strategy could be
implemented with a minimum number of components and
design effort. Furthermore, this IC did not only contain
motion peripheral functions (such as PWM, current
sensing interface and Encoder counter circuit, etc.), but it
also contained complete algorithms for speed and field
orientation in the hardware form, named as Motion
Control Engine (MCE). Also, this control engine
contained control elements such as Proportional Integral,
Clark transformation, vector rotator, etc. This IC did not
require any coding or programming. Hence it could be
easily tuned and also could adapt to new motor
parameters easily. Besides, it contained memory registers
that could be scanned or written using a mating
Microprocessor RS232C by serially interfacing it with a
computer. The scanning and writing of the registers could
be done using a computer. For instance, if a specific value
for the switching frequency was chosen, then it could be
simply written to the specific register. The benefits of this
type of controller are that the IC computed very fast for
the closed-loop control algorithm, which led to a good
dynamic performance of the speed and torque of the
system. It was a single-chip solution for complete closed-
loop control. Plus, it could be easily tuned with different
specifications holding Motors, which helped to
implement a control algorithm easily. The torque control
loop had a good step response, yet it contained small
oscillations. Lastly, its voltage switching waveform and
motor current waveforms contained oscillations and were
not smooth, which may cause noise in the practical.
In harsh environments such as underwater, the sensor
of the Motor was one of the main issues. Therefore, the
Sensor-less Servo system was suggested and
implemented by B. Allotta et al. [25]. One of the critical
parts of the Sensor-less control algorithm was the
position-speed estimation. The Filter that had to be
implemented to estimate the rotor position in order to
perform an optimal commutation of the currents on
windings. The most common technique for the Sensor-
less control strategy of PMSM was established on the
observation of back-EMF as emf was proportional to the
rotor speed. This technique performed well when the
Motor spine at a speed over 10-20% of the nominal value.
As a result, the feed-forward start-up of the Motor was
required. Another technique was based on the injecting
signal that excited the machines at frequencies that were
different from the operating frequency of the machine and
with a negligible influence on the mechanical behavior of
the machine. However, this kind of method required
accurate current measurement and precise and reliable
current Sensors. For PMSM with star-connected
windings, it was also possible to perform Direct Flux
Control that was able to directly estimate the motor flux
linkages. Plus, an Improvement was done in the smart
back-EMF estimators that were the determination of the
rotor speed and position through back-EMF estimation.

Abdul Wali Abdul Ali et al. / ELEKTRIKA, 19(2), 2020, 22-39
29
The first part of the estimator generated the esteem of the
back-EMF and the second part estimated the rotor speed
and position through back-EMF. In short, the robustness
of the controller against a harsh environment is that the
real speed of the motor was lower than the reference
speed. This led to an average error between the estimated
speed and reference speed of the motor. Also, the current
and voltage waveforms contained a lot of oscillations.
Furthermore, C. H. Yan and J. H. Hui designed a full-
closed loop servo control system in 2015 [26]. The semi-
closed loop control system was widely used but for the
enhanced control precision, the full-closed loop method
was introduced. Gear Measuring Center used the
technology of the measurement for Four axes which were
X (tangential movement), Y (radial movement), Z (axial
movement) and C (spindle rotary movement) axes. These
axes were directed by the Servomotor using the numerical
control method in order to understand the linkages
between these axes. Besides, two ways communication
was constructed between a single board computer and a
motion controller. The computer sent motion command to
the controller based on the received motion state from the
controller. The position of individual axes in real-time
was measured using circular grating. In the semi-closed
loop control method, the encoder was utilized as a speed
closed-loop control as well as position closed-loop
control but some errors existed in the transmission chain.
To overcome this defect, a grating was used as position
closed-loop control, so the Encoder would only work as
speed closed-loop control. The speed closed-loop
contained Three internal closed-loops feedback control
systems (loop of the current, loop of the speed and loop
of the position). The current-loop was created internally
in the drive and it was used to improve the dynamic
response of the system. As a conclusion, this control
system had no distinct sensitivity to its components
fluctuations. The control method could improve the
overall performance of the Motor. Yet, the drawbacks of
this control method were it was more expensive and
complex, there consisted risk of instability, and it may
have created an oscillatory response and reduced the
overall gain of the system.
Besides, an automatic control loop tuning initiated by
S.-M. Yang and K.-W. Lin’s study in 2015 presented a
new scheme for the AC Servomotor drives parameters;
the determination and auto-tuning [27]. For the current
control loop, the determination of the electrical
parameters such as inductance and resistance was done.
On the other side, for the speed and position loop tuning,
the determination of mechanical parameters and torque
constant were done. According to the authors, the drive
contained two inner loops for the current control and two
outer loops for the speed and position control. The path in
the speed loop was realized with the speed command. The
PI regulators were utilized in the q and d axes current
controllers and there was a limitation of at least half of
the rated single-phase voltage. Also, the decoupling of the
cross-coupling and back-emf voltages were done utilizing
the predicted speed of the rotor and the electrical
parameters of the system. The cross-coupling voltages
and back-emf voltages were decoupled using the
estimated electrical parameters and rotor speed. To
prevent the error caused by the motion of the rotor, the
resistance of the stator should have been known. The
measurement of this resistance was done by employing a
pulse of d-axis voltage. Based on this pulse, the
measurement of the d-axis current at the steady-state was
performed and then from this measurement, the resistance
was calculated. The inductance was calculated by the
measurement of the peak current, which was done by
applying pulses of q and d axes voltages. Mechanical
Parameters, Feed-forward voltage and torque constant
were known by utilizing the theoretical current and speed
waveforms. In practice, the entire auto-tuning procedure
took roughly 1.4 seconds for accomplishment. The
advantages of this control system are this method led to a
good dynamic response, although the parameter
identification was not free of errors. Next, the transient
response and the frequency of the closed-loop showed
consistency with the tuning. However, some errors
existed in the measurement of auto-identified parameters
when compared with manual measurement, these errors
were within 10% and at the same time. The system
current waveforms contained many oscillatory errors and
these errors should have been reduced for further control
precision.
Next, Y. Sang et al. proposed a practical AC servo
motor controller based on the STM32 microcontroller in
their research paper in 2015 [28]. The proposed servo
drive had three control modes, which were the position,
speed, and torque. For speed control, the Microcontroller
produced two-way pulse signals and then the signal was
fed to the servo drive to control the Motor. The Encoder
produced the feedback signal. The rotation angle of the
Servomotor was controlled by the deviation signals that
were generated by the comparison of the target value and
feedback value. The position control accuracy was
depending on the number of pulses that encoder produced
per revolution. This particular Servomotor had two input
ways of speed command, analog input and register input.
Moreover, the Microcontroller realized the speed control
of Motor through digital-to-analog conversion. For
instance, a higher value of analog voltage led to a higher
value of speed. Finally, closed-loop control was known
by the feedback of speed loop Encoder. On the other
hand, the torque was controlled by the produced
instruction of the Chip through the digital-to-analog
conversion to be sent to the CN1 terminal of the drive. In
such a way, the Servo drive would rely on the internal
current loop to realize closed-loop control. The current
loop was used to determine the anti-interference ability
and response speed of the system. Lastly, the position
loop was the most important part of the stability of the
system. The actual position from the feedback loop was
compared with the target position set and then position
regulator produced speed commands. In short, this
control system had a precise static performance. It had a
simple hardware circuit, strong real-time performance,
low cost, fast processing speed, and reliable operation.
However, this control system had two drawbacks, firstly,
while adjusting the gain in manual mode, machine
rigidity and surroundings had an enormous influence on
the selection of Bandwidth. Secondly, the transmission
inertia affected the stability and dynamic response of the
Servomotor.
In the conventional AC drives the performance of the
current loop was affected by the saturation of the
magnetism. This affected performance subjected to the

Abdul Wali Abdul Ali et al. / ELEKTRIKA, 19(2), 2020, 22-39
30
precision of tracking and the optimized methods like field
weakening. An algorithm was used to control the current
vector trajectory for the production of perfect current that
produced torque within the limit of the supplied voltage.
This algorithm is designed by J. Bermingham et al. and
the principle was named field weakening [29]. According
to the authors, the optimized torque current was
calculated by utilizing the compensated PI regulator
based on the voltage vector control method. The levels of
the optimized torque under dynamic speed and system
parameters were achieved through voltage vector
trajectory commands generated by this algorithm. In this
scheme, the voltage commands were generated under
Four operation conditions, which laid within the speed
and current boundaries of the motor. The voltage vector
commands were implemented utilizing the voltage
control strategy. Also, these commands were used to
prevent the interaction or the sensitivity among the
current loop and voltage loop. The voltage loops played a
vital role in the regulation of the Motor voltage at the
defined boundaries. This occurred when the realization of
the current commands of the drive could not be done in
the field weakening region. The decoupling process
ensures that the d and q axes voltage control loops
controlled the d and q axes current control loops
respectively. In conclusion, this control system was
robust against the effects of large parameter variations
and its speed range and torque per set point of the speed
were high. Plus, the voltage-vector control maintained
continues Motor voltage and eliminated the risk of the
degradation in the performance in the field weakening
region. The principle of field weakening was
comprehensive in terms of the theories but practically, it
was difficult to achieve the objective correctly. Its voltage
and current waveforms were not very smooth which may
affect the speed and torque performances of the Motor.’
Apart from that, the performance improvement of the
torque ripple suppression by using the FLC approach for
PMSM was proposed by M. Gong et al. in 2015 [30]. In
this proposed design, the conventional Direct Torque
Control (DTC) scheme was replaced with the FLC
structures. The hysteresis controller of the DTC scheme
considered only the signs of flux and torque error rather
than their amplitudes. If the amplitude of the errors
exceeded the hysteresis boundary, then only vectors were
changed else they remained the same irrespective of how
large or small the error was. Besides, these vectors of
voltages given by Voltage Source Inverter (VSI) with
flux and torque error information were used to control the
Motor. The authors also employed discrete control
methodology, which was performed with respect to the
defined discrete instances. Thus, the torque ripples and
flux ripples were higher than hysteresis limits and the
conventional DTC resulted in large flux and torque
ripples. To overcome this issue, the conventional DTC
scheme was replaced with the FLC structure. The voltage
vector selection of optimal value was achieved by
inferring the ranges of torque errors and flux errors.
Another FLC was employed into the structure to
determine the action time of the vectors. This was
obtained by inferring the duty ratio, which was also
determined by the second FLC in the structure.
Furthermore, the proposed scheme was analyzed and
validated using four various parameters like speed,
torque, flux and current for both DTC & FLC schemes.
Lastly, this control system was elucidated from the
analysis that the proposed FLC methodology resulted in
smaller ripples at a steady-state than the conventional one
and also with quicker dynamic performance. With the
proposed method, the torque ripple was almost reduced
by 75%, the flux ripple was reduced by 50% and the
reduction of fluctuation and stator current. However, the
setback with design of FLC scheme was that it required
some intuitive understanding of the process. Even though
the fluctuations and oscillations in the torque and speed
response were reduced, there were some oscillations and
fluctuations, which needed to be further reduced.
Sliding Mode Control (SMC) method was considered
to be one of the successful non-linear control
methodologies owing to its robustness. However, in the
case of the non-linear system, some chattering
phenomenon was being observed during the process of
sliding at the end surfaces. To cope with the chattering
phenomenon, a composite technique was studied by H.
Wang et al. in 2016 [31]. The studied composite scheme
contained Continuous Terminal Sliding Mode Control
(CTSMC) with ESO. The CTSMC was employed to cope
with the chattering phenomenon. However, in the
presence of stronger disturbance in the PMSM system,
this would result in steady-state speed fluctuation. To
solve this issue, an ESO with adjustable gain was
introduced. Besides, this particular study was contributed
to the attenuation of the chattering phenomenon along
with the speed regulation and disturbance rejection of
PMSM. The CTSCM scheme was used as a feedback
regulation in order to stabilize the PMSM drive dynamics
in a finite time. ESO was employed as feed-forward
compensation to regulate the steady-state speed
fluctuation of PMSM drive in the presence of
disturbance. To validate the proposed methodology, the
scheme was implemented with DSP TMS320F2808 with
100MHz clock frequency. Based on the results obtained,
it can be seen that the CTSMC had shorter settling time
and the overshoot was almost minimal for both CTSMC
and SMC schemes. However, with the introduction of
sudden load, better disturbance rejection property was
observed in the composite mode only. Also, it was
inferred that the composite method produced better
tracking properties and faster convergence compared to
the conventional one. In order to improve the disturbance
rejection property of the CTSMC, switching gain was
adjusted by varying from smaller to a larger value. The
more the value of gain, the better the rejection. On the
other hand, speed fluctuation was more for larger gain
values. Hence, it was quite complex to achieve a trade-off
in rejection of disturbances and minimal speed fluctuation
with an adjusting gain. The same analysis was carried out
for the composite mode where for a smaller gain, it
resulted in good disturbance rejection and small steady-
state fluctuations. This was far less than the one by
CTSMC with the adjustable gain method. Lastly, it was
noted that the speed response of the system with
composite mode resulted in some dead time and resulted
in more rise time than the other two schemes. It was also
observed that the gain parameter had a coupled effect on
tracking and disturbance rejection characteristics.
Predictive controllers were the class of optimal control
methods that were usually employed if in a case to

Abdul Wali Abdul Ali et al. / ELEKTRIKA, 19(2), 2020, 22-39
31
determine the future behavior of the system. Model
Predictive Control (MPC) was such an optimal control
scheme and was quite popular in industries because of its
capability to forecast the behavior of the system in
addition to its control action according to the target
optimizing function. A simple and effective Predictive
Functional Control (PFC) methodology for PMSM drive
to improve its controller performance was proposed by S.
Wang et al. in 2015 [32]. According to the authors’
explanations, the PFC was majorly employed to
overcome the limitations of MPC. MPC involved a higher
computational burden since it required optimization at
every sampling instant. The PFC was essentially an
alternate version of MPC. This is because the PFC
retained most of the advantages like handling constraints
and online optimization with a low computational cost.
The reduction in computational complexity was achieved
through more intuitive design procedures. Besides, the
speed control of PMSM using a cascade structure was
carried out by this methodology. The methodology was a
two-level design procedure. In the first level, forecasting
of the current values was carried out which was followed
by an optimization procedure to optimize the level one
value using a suitable cost function. The system
performed well for a few well-defined conditions.
However, when subjected to strong disturbances, the
performance was not quite satisfactory. Hence, Improved
Predictive Functional Control (IPFC) was proposed to
compensate for unknown dynamics and stronger external
perturbations. The proposed IPFC was a composite
scheme that employed PFC with ESO. This was utilized
for effective controller performance. Feed-forward
compensation was provided for better disturbance
rejection properties. The entire control algorithm was
implemented using DSP TMS320F2808 with 100MHz as
clock frequency and the implementation of the control
algorithm was carried out using C-language
programming. One of the advantages of the PFC is it
resulted in minimal overshoot and small settling time
compared with PI. During the introduction/removal of
load, speed fluctuations were less in PFC. The real-time
experimental results also elucidated that PFC provides
better disturbance rejection and faster recovery time after
the load introduction/removal. To improve disturbance
rejection capabilities further, the system was assessed
with IPFC where a feed-forward compensation was
introduced for better disturbance rejection. This resulted
in the smallest settling time and very minimal overshoots
than PFC and PI schemes. In addition to that, during the
load introduction/removal, faster recovery and very
minimal fluctuations were observed. Lastly, in the real-
time analysis, PFC + ESO resulted in very good
disturbance rejection properties. However, the speed and
current waveform of the system contained a lot of
oscillations/ripples using the proposed methodology.
Furthermore, C. Dang et al also suggested analysis and
reducing methods of cogging torque on permanent
magnet AC servo motor in 2014 [33]. According to the
authors, the permanent magnet poles unavoidably
interacted with armature iron core and hence produced
cogging torque that caused vibration and noise, which
affected the operation performance and control accuracy
of the Motor. The computation techniques for the cogging
torque included energy method, the Maxwell tensor
method, and the Finite element method. The three
predeveloped methods were used to reduce the cogging
torque effectively. These methods implied the adjustment
of the width of the slots, the use of unequal thickness of
the permanent magnets and the use of unequal widths of
the permanent magnets. Based on these methods, the
finite element method was used to set; a variety of
programs were compared and an optimal solution was
proposed. Based on the experiments and analysis of this
investigation, it was found that the cogging torque was
affected by the solder bath placed on the stator. When the
stator yoke portion was uniform, it did not affect the
magnetic circuit of the stator. After notching the solder
bath periodically, the resistance of the circuit got affected.
This affected flux density, thereby affecting the air-gap
magnetic and this energy was converted into cogging
torque. Nevertheless, the cogging torque still was reduced
to some extent. It was assured that thickening the stator
yoke to reduce the saturation level could reduce the
impact of the solder bath and reduce the cogging torque
further but it would lead to the reduction of the power
density of the motor.
A methodology for reduction of torque ripple for
Brushless Direct Current (BLDC) Motor was proposed by
optimizing reference current utilizing the Integral
Variable Structural Control (IVSC) strategy. This
methodology was suggested by C. Xia et al. in their
research paper in 2014 [34]. The authors explained that
there were two types of commutation modes which were
low-speed commutations with and high-speed
commutations. The actual type of commutation was
determined with respect to the relation between back-
EMF (Electromotive force) and voltage of DC links. The
commutation controlled by double-phase switching for
current optimization of non-commutated windings was
used for low-speed commutation. However, if the same
was applied for high-speed commutation, the non-
commutated line would fail to trace the reference current
and would also introduce ripples in the torque response of
the system. In order to avoid this, three-phase switching
was explored in this work for high-speed commutation
instead of double phase switching. Furthermore, the
control structure contained two loops. The current
optimization was carried out in the inner loop and the
respective speed control was carried out in the outer loop
using the PI technique. This method involved the
estimation of the back-EMF using Luenberger full order
state observer. Also, the optimization was carried out in
accordance with the back-EMF waveform of both the
modes. This proposed scheme was validated in real-time
using the DSP TMS320F28335 for surface mounted
BLDC system. The stability of IVSC was further
analyzed with the Lyapunov candidate function. In short,
this method was employed with an IVSC strategy owing
to its advantages like robust disturbance rejection
capabilities and wider noise band suppression
competency. These features were fully utilized, which
resulted in the avoidance of a negative chattering
phenomenon. The reduction in the torque ripple and the
improvement in controller performance over a wider
range of load and speed were observed. Lastly, it was
noted that the use of three-phase commutation switching
action increased the commutation time, cost and also
decreased efficiency of the process. However, the

Abdul Wali Abdul Ali et al. / ELEKTRIKA, 19(2), 2020, 22-39
32
methodology played a dominant role in ripple reduction.
The performance comparison of PMSM for the control
of speed using Proportional Integral (PI) and FLC
schemes was carried out. This study was conducted by S.
P. Singh et al. in 2016 [35]. The FLC technique involved
the designing of input-output mapping of the controller’s
behavior which was generally expressed as fuzzy rules.
The input to the FLC was a crisp value that would be
fuzzified and this value was taken for fuzzy operations.
After that, mapping of degree of matching to fuzzy output
by suitable implication methodology was carried out.
Followed by, aggregation of output and defuzzification
mechanism was performed. The output of the FLC was
again a crisp value because of the choice of the Fuzzy
Inference System (FIS - Mamdani) used for the study. In
this study, error and the derivative of error of speed are
used as input to FLC. Besides, each input was defined
with Seven membership functions with a maximum of 49
rules that were designed for tuning the process. Mamdani
type inference mechanism was used as the FIS scheme
for the control of speed for the motor drive and the center
of gravity method was utilized for the defuzzification
process. As a conclusion, the PI scheme was easily
affected by non-linear dynamics and time-varying
parameters. To cope with that, an intelligent controller
using FLC was incorporated. The simulation analysis was
performed for various conditions like no-load, full load
and varying load scenarios. For no-load condition, the
overshoot of PI was 0.38% with settling time of 0.16
seconds, whereas, for full load condition, the overshoot
was 0.033% and settling time was 0.034 seconds. Under
varying load conditions, the overshoot was 0.6% for PI
and 0.053% for FLC with a settling time of 0.16 seconds
for PI and 0.04 seconds for FLC respectively. It was
inferred from the simulation analysis that FLC
outperforms PI scheme owing to its highly adaptable
nature. However, the major setback of designing the FLC
was that it required knowledge/experience of humans
regarding the particular system of interest for which FLC
design had to be carried out.
A novel methodology for Active Disturbance
Rejection Control (ADRC) for suppressing the overshoots
of the highly dynamic input signal was outlined. This
proposed scheme is designed by Tianrui Luan et al. in
2016 [36]. The proposed scheme was subjected to an
analysis of the PMSM system. ADRC structure
encompassed Three sub-components namely TD, ESO
and State Error Feedback (SEF) respectively. There were
many non-linear parameters associated with the ADRC
scheme and these parameters introduced difficulty in
tuning. To avoid this and also to obtain speed regulation,
a simplified approach was proposed. The proposed
approach primarily eliminated the TD component and
further linearization was carried out in ESO and SEF
components. Hence, this reduced the tuning complexity
and also resulted in better dynamic performance and
improved robustness to disturbances. Besides, the
proposed scheme was a two-step procedure where the
primary step was to tune tracing and the disturbance
estimation parameter as they were supposed to tune in
accordance with each other. The secondary step was to
find the optimal tuning parameter of the controller and
the settling time was chosen as an objective criterion. The
benefits of the proposed scheme are the ADRC possessed
smaller rise time and also faster recovery time whenever
the load was removed/added when compared with the PI.
Next. the rise time was around 0.066 seconds for ADRC
whereas, for the PI, it was 0.095 seconds during
implementation. The separation principle could be used in
ADRC for designing ESO and SEF. It was noted that
there was no authenticated mathematical evidence that
claimed the design procedure for ADRC parameters. The
used method gave only the generalized disturbance
estimated by ESO. Plus, the combined expression for
dynamic disturbances of both internal and external effects
was not used here. However, the torque and speed
response contained oscillations and fluctuations.
Apart from that, the experiments on Induction Motor
with the objective to develop a digital controller for
Induction Motor (IM) Drive was carried out by Y. Zhang
et al. in 2013 [37]. The improvement of Field Oriented
Control (FOC) for Induction Motor Drive (IMD) was
achieved by using the ADRC scheme. The analysis was
performed without any Speed Sensor. The
implementations of control algorithms were carried out in
Digital Signal Processors (DSP) and Field Programmable
Gate Array (FPGA) was used for some basic logical
manipulations. The concept behind this control strategy
was to identify the unknown disturbances using extended
observers and to compensate for it in real-time. The Two
control algorithms namely ADRC and PID were
implemented in DSP processors. The PI parameters were
tuned by trial and error method whereas the ADRC was
tuned with Linear Extended State Observer (LESO) and it
followed the conventional systematic procedure.
Moreover, the real-time implementations were carried out
with the TMS320F2812 DSP processor and EP2C5T144
FPGA processor. However, the major drawback of FOC
which required co-ordinate transformation and current
controllers was modified with the ADRC scheme. The
analysis showed that the ADRC outperformed the PID
scheme without overshoots and oscillations. Lastly,
owing to tuning by trial and error, the speed response of
the PI controller resulted in larger settling time, rise time
and overshoots whereas the ADRC resulted in faster
settling and very minimal overshoot. Though, this
approach of using DSP and FPGA introduced complexity
in design, space and also in cost aspect.
W. Bin et al. conducted a study on the application of
control techniques with respect to the region of operation
for achieving the full range speed in 2014 [38]. Based on
the authors' research, the constant torque and constant
power zone were taken for analysis. The Interior
Permanent Magnet Motor (IPMM) drive also was taken
for study. For the constant power range, in order to keep
the Motor with the unchanged power, the Field
Weakening (FW) control was applied. Meanwhile, for
regions with constant torque, Maximum Torque Per
Ampere (MTPA) was applied for full torque performance
requirements. When Motor was operating at a constant
torque region, the MTPA control scheme was utilized to
use the advantage of the torque reluctance of the Motor.
When producing constant torque with minimal current, by
using this control scheme, Motor efficiency could be
highly improved by decreasing the copper loss. On the
other hand, if the speed of the Motor was more than the
base speed, this would result in the increase of the back-
EMF. However, the higher Inverter voltage introduced

Abdul Wali Abdul Ali et al. / ELEKTRIKA, 19(2), 2020, 22-39
33
bounds on the Motor terminal voltage. Therefore, to
maintain the balance in voltage level during high-speed
rotation, the stator current was manipulated by increasing
the demagnetization current. In this scenario, the flux
weakening scheme should be used. Furthermore, the
important machine parameters based vector control
method was obtained through Finite Element Analysis
(FEA). For Motor speed below the base speed, the
analysis was carried out with two schemes namely control
and MTPA control. On the other hand, for Motor speed
greater than base speed, the analysis on smooth switching
from MTPA to FW was carried out. Based on the
analysis, the results showed that the improved dynamic
performance was obtained by the appropriate choice of
control strategy according to the particular operating
region. However, during switching, torque ripples were
evident from the study which may have required some
ripple suppression techniques to be incorporated with the
proposed methodology.
For Motor control, generally high precise and reliable
Sensors were required for accurate positioning. However,
owing to reasons like space constraint, cost, and system
complexity, these Sensors were not installed especially in
the case of serious industrial environments. Therefore,
sensor-less speed identification was another important
area gaining wider attention. Z. Ding et al. proposed
another speed control technique for permanent magnet
synchronous motor in 2014 [39]. The proposed
methodology suggested Sensor-less speed identification
and speed control using a Sliding mode approach. The
analysis was performed on the PMSM system. The
conventional PI scheme was replaced with SMC structure
to improve the performance mainly because of the
presence of non-linear characteristics associated with the
PMSM drive. The speed identification was carried out
with variable structure Model Reference Adaptive
Control (MRAC) scheme where SMC was integrated
with MRAC structure. Also, the MRAC scheme
contained an actual model and a reference model. They
were compared each time and according to the deviation
between them, corresponding control action was carried
out with a suitable adaptation mechanism. This scheme
also was incorporated with the SMC scheme. The control
law was constructed with a sigmoid function. Since the
traditional SMC utilized signum function for switching
resulted in a chattering phenomenon. Therefore, to
overcome this drawback, the signum function was
substituted with the sigmoid function. SMC based MRAC
was used for speed identification. Furthermore, the
proposed structure was validated for the reach-ability
condition using the Lyapunov analysis. The SMC
function was designed with the variable exponential rate
reaching law for quicker sliding and for the elimination of
chattering effect. For the proposed structure, Lyapunov
analysis was carried out for stability assurance. In short,
this proposed method was elucidated that the estimated
speed was tracked quickly and properly. It was inferred
that the SMC scheme resulted in faster response, lesser
speed overshoots and high robustness. However, the
precise performance largely depended on the reference
model used in the structure. Also, the system torque and
speed response contained oscillations and fluctuations
It was highly challenging for the system performance to
be robust when the system was subjected to unknown
disturbances and uncertainties in the parameters unless
the system was employed with more sophisticated control
algorithms. To cope with the time-varying parametric
uncertainties and perturbations, non-linear control
techniques were largely adopted to improve the
performance of the system. The SMC was one of the non-
linear robust strategy employed in many fields because of
its advantages like easy tuning and implementation for
highly non-linear systems. However, the system would be
associated with some chattering phenomenon because of
its discontinuous switching action. To compensate for
disturbances, an observer could be introduced and based
on this effect, switching gain had to be selected
appropriately to minimize the chattering. Apart from the
above-discussed method, the chattering phenomenon
along with time-varying uncertainties was also addressed
with other composite schemes. Therefore, Generalized
Proportional Integral (GPI) observer with SCM
(GPI+SCM) for better rejection of disturbances and
unknown uncertainties of the system were proposed by H.
Wang et al [40]. The proposed methodology was tested
for speed regulation of the PMSM drive. The procedure
involved the design of the GPI observer for disturbance
rejection and SMC for speed regulation. The GPI
observer for PMSM drive was designed as a function of
external perturbations, frictional loads and current
tracking error. The final gain parameter tuning was
obtained as the function of the single parameter by
directly equating to the characteristic polynomial of the
observer using a direct comparison method by assuming
that poles were in complex left half far away from the
imaginary axis. Besides, the speed control regulation was
achieved by combining SMC with GPI observer. It was
stated that convergence of the speed error to equilibrium
point asymptotically was assured if the switching gain
was larger than the gain of the disturbance error
estimation parameter. In addition, SMC did not require
rejecting disturbance with feed-forward compensation
using a disturbance estimator. As a conclusion, both the
schemes possess minimal overshoot with small settling
time. However, for the application of constant load
torque, it was inferred that speed recovery in the
composite scheme was faster than the SMC. The
composite method could estimate and reject ramp
disturbances. When the slowly varying signal was applied
with a sufficiently larger gain for SMC, it was difficult to
suppress the disturbance. However, the composite mode
rejected in a better way with much smaller gain value.
Lastly, during the recovery/rejection, some overshoots
and oscillations (in case of ramp signal) were observed in
the speed regulation performance, which may have
needed to be further addressed.
Besides that, the SMC also was widely adopted in
most of the highly non-linear electromechanical systems
owing to its highly remarkable features such as excellent
tracking properties and robustness for disturbance
rejection and time-varying parameters of the system.
However, some of the chattering phenomena were
observed because of the discontinuous switching whose
again could be reduced by adjusting switching gain
parameters. Even the highly sophisticated control
algorithms might have performance degradation if in case
a fault associated with any Actuator or Sensor
components of the system. To overcome this issue, many

Abdul Wali Abdul Ali et al. / ELEKTRIKA, 19(2), 2020, 22-39
34
fault diagnostic approaches were developed in recent
trends. The Fault Tolerant Control (FTC) approach using
the SMC strategy for PMSM Drives had been studied by
H. Mekki et al. in 2015 [41]. The fault associated with
stator asymmetries had been addressed in this research.
The healthy model of PMSM was applied with an integral
surface SMC strategy. The surface integral SMC for
PMSM drive was provided with Three surfaces, utilized
for speed, direct and quadrature current regulators. The
proposed scheme was analyzed for closed-loop stability
performance with Lyapunov stability studies. The faulty
model for eccentric fault diagnosis of PMSM drive was
developed. The objective was to make the system to
provide continued better performance, even if in case the
fault was being identified during the process such that
reconfiguration was not required during the process.
Furthermore, to improve the controller performance
under faulty conditions, the internal switching SMC
scheme was reinforced with the FTC approach. Lastly,
the Lyapunov stability analysis was carried out for the
FTC approach and also for continued stable performance
[41]. In short, this novel SMC provided good robust
tracking properties. However, in the presence of static
stator fault, the performance was degraded largely. The
FTC with SMC resulted in good reference tracking and
rejection of load disturbance even when the system was
subjected to eccentricity faults. It was inferred that the
system possessed faster dynamic performance and it was
highly robust under faulty conditions. However, systems
speed and current responses contained oscillations and
fluctuations under stator faults and needed to be further
improved.
The PFC accuracy was largely dependent on its
external perturbations. This had been addressed with a
composite scheme like PFC+ESO and the performance
was improved. However, apart from the external
disturbances, speed feedback quantization error also
largely influenced the performance of the system. A
methodology for better dynamic performance was
proposed with PFC by incorporating a Kalman filter
(KF). This methodology was investigated and suggested
by H. Liu and S. Li in 2012 [42]. By employing the
Kalman filter, a better estimation of states was made
possible by eliminating system and measurement noise.
KF was mainly used to obtain the information of load
torque, position of the rotor and the speed. Its ability to
operate in a noisy environment (both system’s and
measurement noise) and also the disturbance estimation
resulted in better performance. Besides, the PFC is the
modified version of MPC with minimal computational
complexity. It also contained some base function in terms
of control variable of interest and it is a predictive model.
Plus, it contained some amount of error correction
phenomenon between predicted and the actual model.
The receding optimization control methodology with the
objective to minimize the variance employed with
quadratic cost functions. This approach was added with
the KF for compensation of quantization errors in speed
feedback loop mainly because of rotor precision
limitations. The estimated and validated information were
used in the PFC control scheme for performance
enhancement of the system. The proposed scheme was
implemented in real-time using Infineon’s XMC4500,
120 MHz clock CPU frequency of ARM Cortes core.
Incremental Photo Encoders were used for rotor position
detection. It can be concluded that both schemes were
stable and could detect speed changes. However, there
was lag associated with detection by M/T method
(measuring both frequency and time speed method) and
also its tracking performance was not as accurate as
compared to the KF method. Secondly, observer
performance analysis was carried out with Disturbance
Observer-Based (DOB) method. For changes in load
torque, the DOB method introduced some overshoots and
it converged within 0.3 seconds. Meanwhile, KF was
being stable within 0.15 seconds and the overshoot was
comparatively much lesser than the DOB method.
Besides, it was observed that dynamic properties of
PFC+KF outperform PI with less rise time, settling time
and minimal overshoots. The steady-state ripples,
overshoots and settling time were more in PI, less in PFC
and relatively minimal in PFC+KF scheme. Compared to
PFC, PFC+KF possessed a small speed drop and faster
recovery time. This was made possible by KF, which
helped in creating a better estimation of information for
PFC and high-efficiency performance. However, the
system speed and current waveforms contained a lot of
oscillations and ripples under all of these implemented
control methods.
Furthermore, an adaptive control technique was widely
preferred in cases where the system was expected to
adapt to the changing process parameters/environmental
conditions experienced during the processes. According
to the research conducted by X. Li and S. Li, Speed loop
control for PMSM using adaptive control were initiated
[43]. The Model Reference Adaptive Control (MRAC)
scheme was employed for the system to accommodate the
parameter changes. MRAC scheme consisted of three
main components namely, reference model, adaptation
mechanism and the controller. The reference model was
generally chosen as to how the system was expected to
behave. Meanwhile, the adaptation mechanism was about
how the system had to accommodate to the changes and
at what rate the system had to be adapted like the
reference model and also to account for perturbations.
Thirdly, the controller was the basic control structure or
control law, which was usually employed in the system.
Every time, the deviation between the plant model and
the reference model was accounted for. Based on the
deviation, according to the adaptation mechanism,
corresponding control actions were provided each time.
Also, the control performance was quite improved and
the adaption mechanism was carried by using Lyapunov
stability analysis by the appropriate choice of candidate
function. However, to improve the disturbance rejection
capabilities, the system was incorporated with ESO to
compensate for uncertainties and unknown dynamics.
The stability analysis was carried for the composite
method. The proposed scheme was implemented using
DSP TMS320F2808 with a clock frequency of 100MHz
and C-programming language was used to implement the
control algorithm. In short, the MRAC resulted in better
performance than PI in a relative way in terms of fewer
overshoots and steady-state oscillations. MRAC had a
good dynamic load performance. It resulted in a smoother
response and very minimal overshoot. Even in the
presence of the load torque disturbances, MRAC
performed better than PI. However, for better rejection,

Abdul Wali Abdul Ali et al. / ELEKTRIKA, 19(2), 2020, 22-39
35
feed-forward compensation was employed by introducing
an observer. The composite structure resulted in better
properties than MRAC with relatively less settling time
and overshoots. Furthermore, the composite scheme
resulted in very good disturbance rejection properties
with very short recovery time. This composite MRAC
(MRAC+ESO) scheme possessed improved controller
performance with high robustness. It was also to be noted
that the composite method degraded the adaptation
capability of the system. This was mainly evident from
the steady-state ripple which was always greater than the
MRAC scheme.
Apart from the handling of parameter variation and the
unknown dynamics of the system by adaptive structures,
for enhancing the performance of the closed-loop, a
fractional adaptation mechanism was developed. The
MRAC using factional adaptive scheme was utilized in
the PMSM drive for velocity control. Two mechanisms of
adaptation were elaborated in this work. The mechanisms
were a gradient-based fractional order adaptation scheme
and Lyapunov based fractional order adaptation scheme.
It was inferred that for the gradient-based adaptation
approach that convergence of the system could be
increased by increasing the adaptation rate (𝛾). However,
In the Lyapunov adaptation mechanism, it was inferred
that larger the 𝛾, slower the adaptation rate. Larger 𝛼 was
better for the transient response and disturbance rejection
features. These results were completely contradictory to
the gradient-based adaptation method. Apart from that,
the presence of oscillations in the transient response also
was observed. To overcome this oscillatory behavior, the
normalized gradient approach had to be introduced. The
methodology for velocity control of the PMSM system
consisted of two loops. The inner loop controlled by
feedback linearization to obtain the reference voltage
information and the outer loop employed with Reference
Signal Tracking (RST) scheme for current information.
Also, the outer loop control parameters were tuned by a
fractional adaptive mechanism using the MRAC
technique. Moreover, Lyapunov analysis was carried out
for the design of the fractional-order MRAC scheme.
Besides, the stator reference current parameters were
controlled with the RST scheme. The parameters of RST
were adapted periodically using the gradient-based
approach or Lyapunov adaptation with MRAC structure.
The stator reference voltages were compensated with a
feedback linearization mechanism. It should be noted that
the reference model time constant had to be always
greater than the time constant of the inner loop to
preserve the system dynamics [44]. As a conclusion, to
demonstrate the robustness of the system, parameter
uncertainty was tested by varying inertia to +30% for
both the approaches. The responses did not introduce any
compromise in the performance. Better transient response
was achieved at lower in case of gradient approach
whereas, for the later, the system performed well with
large values of 𝛼. Secondly, the proposed scheme was
compared with the adaptive backstepping algorithm. It
was inferred that the proposed FO+MRAC scheme had a
higher tendency to cope with mechanical uncertainties
and to reject disturbances in an effective way than the
later. However, the appropriate choice of fractional order
derivative had to be provided with some guidelines to
improve the closed-loop performance.
Within real complex industrial situations, handling of
disturbances by non-linear systems was quite challenging.
Many disturbance rejection techniques were employed to
cope with slow varying signal or periodic disturbances. A
methodology to reject multiple disturbances using the
IMC principle had been identified. This methodology was
suggested in the study conducted by Y. Tan et al. in 2015
[45]. The authors proposed the composite control
technique which contained a disturbance model being
embedded into the disturbance observer for better
rejection. This composite model was called as Internal
Model Extended State Observer (IMESO), which was
employed to reject multiple disturbances. Despite many
observer schemes, the reason for the choice of ESO was
that the design of the observer by this method required
very limited information such as the order of the system,
the inputs, and the outputs. The study and the analysis
were carried out on the PMSM drive for speed loop under
multiple disturbance conditions. Both the simulation and
experimentation were carried out for the Three cases
namely, proportional feedback with ESO (P+ESO),
Proportional feedback with IMESO (P+IMESO) and PI
scheme. For the torque ripple information, the obtained
speed response of PI was analyzed with Fast Fourier
Transform (FFT) to characterize the corresponding
harmonics before the experiments. The experiment was
carried out for the reference speed of 1500rpm. The
proposed scheme was implemented in real-time using the
DSPTMS320F2808 processor with associated Power
modules. Based on the experimental results, it is
elucidated that the P+IMESO performs better with
multiple disturbance rejection than the other two
schemes. However, the equivalent performance was being
observed in the case of time-domain properties. All these
schemes possessed small settling time and very minimal
overshoot. Also, it was inferred that the proposed scheme
had a smaller settling time and little overshoot. The speed
ripples observed during the process were small for
P+IMESO, which was approximately 2.10 rpm, large for
P+ESO and larger for PI schemes during the disturbance
rejection of periodic ones. Lastly, the proposed scheme
eliminated slowly varying signal and sinusoidal signals
effectively. The results elucidated the fact that the
proposed methodology (P+IMESO) was capable to reject
multiple disturbances and the unknown dynamics
associated with the process. However, overshoots during
load torque addition/removal could be reduced for
performance enhancement.
Apart from that, the tuning of Integer Order PID
schemes was usually carried out by using either
frequency or time domain specifications. To facilitate the
tuning of the FO-PID controller, a graphical approach had
been discussed to achieve the robust behavior of the
system. FO-PID (Fractional Order PID) consists of Three
gains as same as the IO-PID (Integral Order PID). In
addition to that, Integral Order (𝜆) and derivative order
(𝜇). FO-PID became IO-PID if 𝜆 = 𝜇 = 1, even though,
FO-PID resulted in more flexibility and improved
performance, the tuning was quite complex. The gains of
FO-PID were determined by using the D-composition
method chosen from the left half of the complex plane,
and the degree of stability was defined with a parameter
𝜎. The graphical approach elucidated that the stabilizing
region was obtained by the D-decomposition technique

Abdul Wali Abdul Ali et al. / ELEKTRIKA, 19(2), 2020, 22-39
36
by analyzing Three regions of operation namely, Infinite
Root Boundary (IRB) lines, Complex Root Boundary
(CRB) curves and Real Root Boundary (RRB) curves.
The D-composition was obtained by dividing the interval
into Three zones, namely at 0, (0, +∞), and +∞
corresponding to each region. The procedure was mainly
carried out to reduce the complexity involved in
fractional order PID tuning. The validation of the
proposed scheme was performed in real-time for the
PMSM drive. The DSP56F8346 was the module used for
control in this study [46]. In short, both the schemes were
analyzed for Servo and the regulatory properties. It was
observed that tracking of step signal by FO-PID was
faster compared to IO-PID, and both the schemes resulted
in no overshoots. For regulatory conditions, load torque
had been changed abruptly. It was observed that FO-PID
had a faster recovery to maintain tracking with smaller
undershoots than IO-PID. Thus, the experimentation
elucidated the robust behavior of fractional order PID in
terms of the dynamic performance produced relatively
faster rise time and better disturbance rejection. Even
though the graphical tuning method was proposed,
assurance of optimal stability value had to be explored.
Besides that, the popularity of PI controllers in
industries was mainly because of its simplicity and
effectiveness for the reasonable performance of the
system. PMSM drives generally employ the FOC
algorithm, which basically involved some transformations
to bring the current parameter to the required form (from
AC to DC). However, this process was quite time-
consuming and used lots of memory. To overcome this,
Simplified Vector Control (SVC) had been proposed by
W. K. Wibowo and S. Jeong in 2013 [47]. The SVC
algorithm was obtained from the inverse of the coordinate
transformations used in FOC. Three PI controllers were
used in the PMSM control structure. Two corresponds to
current and one for speed. These PI were tuned with a
heuristic-based optimization algorithm known as the
Genetic Algorithm (GA) to obtain optimal tuning
parameters for each of the three controllers. Besides, the
GA was a local search algorithm originated by the
concept of the survival of the fittest. The algorithm had
three steps namely selection, crossover and mutation. The
potential candidates were selected for the process from
the initially generated solutions. They were taken for
reproduction either by using two parents (crossover) or by
a single parent (mutation). The Mean Square Error (MSE)
was used as an objective function and some constraints
with respect to the specifications in the time domain such
as overshoot, rise time and settling time were provided to
the system. After a suitable number of iterations, the
algorithm resulted in a solution satisfying both the
objective and the constraints imposed on it. Furthermore,
the proposed scheme was analyzed for implementation
using simulation. The analysis elucidated the fact that the
proposed scheme was feasible for environmental
conditions where no higher sophistication and precision
were required. The tracking of reference speed, response
tracking for varying speed and load rejection were all
analyzed. Also, it was elucidated that the performance of
the system satisfied the requirements stated. The
overshoot was 0.41%, settling time was 0.3 seconds and
the rise time was 0.37 seconds. The algorithm reflected
fairly precise control and could be easily realizable in
Microprocessors because of its handle-able sampling time
based on the Nyquist sampling criterion. Lastly, the
algorithm resulted in a sub-optimal solution since it was a
local search algorithm.
Last but not least, although the heuristic-based
approaches like GA and Particle Swarm Optimization
(PSO) tuned PI controllers gave a better performance than
conventional PI, owing to the converge of parameters in
sub-optimal level, the performance efficiency still
hindered. To overcome this, the algorithm, which resulted
in global optimal was taken for analysis. R. Kannan et al
chose a Bio-geographical based optimization (BBO) for
the tuning of parameters for PI controller [48]. BBO was
the algorithm that was inspired as the collection of
biological distribution of species from the geological
area. The objective was to minimize the difference
between actual and the reference speed of the PMSM
drive. Habitat Sustainability Index (HSI) was mostly used
for determining the optimal solution in BBO. The value
was high in favorable regions and low in the case of
unfavorable regions. Moreover, the migrations were also
possible into or out of the regions, namely, emigration or
immigration. So, HSI was improved largely by migration
phenomenon. After several trials, solution convergence
was attained. For the converged parameters, the controller
was tested for real-time performance. Apart from that, the
parameters obtained through Zeigler Nicolas (Zn), Real
Coded Genetic Algorithm (RGA), and BBO were tested
for different scenarios like no-load, full load, and
disturbance rejection conditions. The tuning of PI
parameters was carried out offline. The several
performance measures, namely Integral Absolute Error
(IAE), Integral Time Absolute Error (ITAE), Integral
Square Error (ISE), and Integral Time Square Error
(ITSE) had been evaluated for all the three schemes. It
was inferred that the BBO tuned controller reduced
steady-state error up to 75.2% and transient error up to
52%. However, for even better performance, swarm-
based optimization techniques could be employed.
2.4 AC SERVOMOTOR CONTROL TECHNIQUES
COMPARISON
All of the available control techniques for the AC
Servomotor were compared in terms of their advantages
and disadvantages, as shown in Table 2.1 below. The
precise control of speed, position, and torque were the
main issues with the AC Servomotor. Many control
techniques and algorithms such as neural networks, fuzzy
logic, iterative control, vector control, scalar control,
electronic power drives, repetitive control, PLC control,
internal model control, current differential signals control,
digital signal processing, FPGA, sliding mode control,
continuous path tracking, direct torque control, predictive
functional control, and active disturbance rejection
control were implemented for the solution of these issues.
However, all of the above-stated control techniques had
their advantages and limitations. Most of these control
techniques led to the step response issues, waveform
oscillatory errors and fluctuations, instability of the
system, switching losses, sensitivity to parameter
variations and external disturbances, and low dynamic
responses. Each control technique led to at least two of
the issues mentioned above. There was no control

Abdul Wali Abdul Ali et al. / ELEKTRIKA, 19(2), 2020, 22-39
37
technique implemented that could solve all of these issues
at the same time while offering a high precision control
Control
Control
Technique
Advantages
Disadvantages
Position
Continues path
tracking
[2]
Good step
response
Dynamic load
issues and more
energy
consumption
Repetitive
control
[3]
Robustness
against noises,
unaffected by
load variations
and more stable
Step response
issues, position
tracking error
Enhanced
Iterative
Learning [4]
Robustness
against noises
Position tracking
error and dynamic
load issues
Internal Model
[5]
Good
performance
and low
tracking error
Step response
issues and dynamic
load issues
Robust
tracking [6]
Good dynamic
load response
Step response
issues tracking
error, network
framework and
fault tolerance
issues
Speed
Fuzzy logic [7]
Good dynamic
load response,
small switching
losses and
minimize
chattering
Existence of
viscous friction and
step response
issues, rotor time
constant is
unknown, heavy
parameter
variations
Vector-based
speed control
[8]
Good step
response
Dynamic load
issues
Low-speed
control
[9]
Good control on
the regulation of
the motor speed
Fluctuation and
oscillatory errors in
the speed response
Current
differential
signals
[10]
Reduction of
copper losses,
improvement in
inverter voltage
utilization
Waveforms
Oscillatory errors
and voltage
saturation
Active
Disturbance
Rejection [11]
The ability of
unknown errors
compression
and good step
response
Improvement
needed in speed
response and
dynamic speed
response
Fuzzy Logic
[12]
Highly
adaptive, good
tracking
performance
and robustness
Fluctuations and
oscillatory errors
Decoupled
ADRC for
PMSM drives
[13]
Independency
of the system
parameters and
easy tuning of
the controller
Oscillations are
evident in both
speed and current
responses
Position,S
peed and
Torque
Neural
Networks [14]
Good dynamic
load response,
no overshoot
and oscillatory
errors
System response is
slow, high rise time
Variable
Structure
Direct
Adaptive [15]
Good dynamic
load response
Presence of
chattering and step
response issues
Iterative
learning [16]
Fluctuations in
the waveforms
are eliminated
The elimination of
fluctuations takes
few cycles that
may affect the
current response
Vector control
using custom-
designed
motion control
card [17]
Good dynamic
load response
Oscillatory errors
and fluctuations
XC164CM
Microcontrolle
r
[18]
Simplified
hardware design
and good
reliability
Presence of
oscillatory errors
PLC
[19]
Good
performance
and low THD
level at high
speed
Fluctuations in the
step response of the
system and high
THD level at low
speed
Multi-layer
Neural
Network [20]
Good dynamic
load response
Delay and
overshoot in the
speed response
Fuzzy
Adaptive PID
[21]
Good position
and speed
control
Fluctuations in the
system frequency
and the rise time of
the speed response
is high
Current control
using the DC-
link current
sensor
[22]
Good system
response
Electrical
parameters
variation in
practical
Single-chip
motion control
IC [23]
Fast response
and good
dynamic load
response, The
IC can be easily
tuned to be used
for different
motors
Oscillatory errors
in the step response
of the torque and
the voltage
switching
waveforms
Sensor-less
servo control
system
[24]
Robustness
against harsh
environment
such as
underwater
usage
Amplitude error
between the real
speed and
reference speed
and oscillatory
errors in the system
response
The full
closed-loop
control system
of Gear
Measuring
Centre
[25]
Good system
performance
and robustness
against the
variations of the
system
components
More expensive,
complex, risk of
instability and
oscillatory errors
Automatic
control loop
tuning [26]
Good dynamic
load response
Presence of
oscillatory errors
and error in the
parameter auto-
identification
STM32
Microcontrolle
r
[27]
Strong real-time
performance,
reliable
operation and
static precision
Dynamic load issue
Optimized
control of
servo-motor
drives in the
field
weakening
region
[28]
Robustness
against
parameters
variations
More complex, its
voltage and current
waveforms contain
oscillations, which
may affect the
performance of the
system

Abdul Wali Abdul Ali et al. / ELEKTRIKA, 19(2), 2020, 22-39
38
Direct Torque
Control based
FLC [29]
Reduction in
torque ripple
and flux ripple,
good dynamic
response
Oscillatory errors
and fluctuations in
the waveforms
Continuous
Terminal
Sliding Mode
with Extended
State Observer
[30]
Small steady-
state response
and disturbance
rejection
Speed response
leads to dead time
and high-rise time,
effects of gain
parameters on the
system
performance
Predictive
Functional
Control with
Kalman Filter
[31]
Good dynamic
performance
and fast
recovery time
Oscillatory error
and ripples in the
system waveforms
3. CONCLUSION
A general review on the control of the AC Servomotor
was done; this review was taken from the research
articles and the research papers that were published for
the control of the AC Servomotor. It covered all of the
commercially available control methods for the AC
Servomotor, which include control of speed, torque and
position of the AC Servomotor. Matlab Simulink, C
programming, and Programmable Logic Controller (PLC)
programming were used in the implementation of these
control approaches. The advantages and drawbacks of
these control strategies were discussed and compared in
this review. This review was analyzed and related to the
industrial application to investigate the most demanded
solutions of the problems by the industry. From the
analysis, it was obtained that it is a necessity in the
industry to precisely control the position, speed and
torque of the AC Servomotor when handling a dynamic
load; however, such a precise control method has not
been designed yet.
REFERENCES
[1] R. M. Phelan, Automatic Control Systems, Cornell
University Press, 1977.”
[2] M. Sazawa and K. Ohishi, “Fast continuous path
tracking control considering high precision and
torque saturation,” in 2009 35th Annu. Conf. of
IEEE Ind. Electron., 2009, pp. 3089–3094.
[3] M. Vijayakarthick et al., “Position Tracking
Performance of AC Servo Motor Based on New
Modified Repetitive Control Strategy,” vol. 10, no.
January, pp. 119–128, 2012.
[4] S. Sathishbabu and P. K. Bhaba, “Tracking Position
Control of AC Servo Motor Using Enhanced
Iterative Learning Control Strategy,” vol. 3, no. 6,
pp. 26–33, 2012.
[5] A. Abdul Ali and A. Alquhali, "Improved Internal
Model Control Technique for Position Control of
AC Servo Motors", ELEKTRIKA- Journal of
Electrical Engineering, vol. 19, no. 1, pp. 33-40,
2020. https://doi.org/10.11113/elektrika.v19n1.179
[6] N. Wang and W. Lin, “Robust tracking control of
AC servo system including a ball screw,”
Neurocomputing, vol. 179, pp. 110–117, 2016.
[7] F. F. Cheng and S. N. Yeh, “Application of fuzzy
logic in the speed control of AC servo system and
an intelligent inverter,” IEEE Trans. Energy
Convers., vol. 8, no. 2, pp. 312–318, Jun. 1993.
[8] E. Yolacan and M. Aydin, “Vector based speed
control of permanent magnet AC servomotor with
FEA and experimental verification,” in 2012 9th
France-Japan & 7th Europe-Asia Congr. on
Mechatronics (MECATRONICS) / 13th Int’l
Workshop on Research and Educ. in Mechatronics
(REM), 2012, pp. 324–328.
[9] I. Kardan et al., “Low speed control of AC servo
motors in no-load condition,” 2014 2nd RSI/ISM
Int. Conf. Robot. Mechatronics, ICRoM 2014, pp.
245–249, 2014.
[10] Y. Seki et al., “Quick and stable speed control of
SPMSM based on current differential signal and
extension of DC-link voltage utilization in flux-
weakening region,” Proc. - 2015 IEEE Int. Conf.
Mechatronics, ICM 2015, vol. 1, pp. 709–714,
2015.
[11] Sravya K et al., “Speed control of electric drives
using Active Disturbance Rejection Control,” 2016
Bienn. Int. Conf. Power Energy Syst. Towar.
Sustain. Energy, pp. 1–6, 2016.
[12] B. Shikkewal and V. Nandanwar, “Fuzzy Logic
Controller for PMSM,” Int. J. Electr. Electron.
Eng., vol. 1, no. 3, pp. 73–78, 2012.
[13] X. R. X. Lx et al., “A Decoupled Active
Disturbance Rejection Controller for PMSM Speed-
regulation System with Position Feedback,” 2015.
[14] Z. Chen et al. , “Intelligent control of alternative
current permanent magnet servomotor using neural
network,” ICEMS 2001 - Proc. 5th Int. Conf. Elect.
Mach. Syst., vol. 2, pp. 743–746, 2001.
[15] F. J. Lin et al., “Variable structure adaptive control
for PM synchronous servo motor drive,” IEE Proc.
- Elect. Power Appl., vol. 146, no. 2, pp. 173–185,
1999.
[16] I. C. Regulation, “A New Iterative Learning Control
Method for,” pp. 1460–1465.
[17] S. Zorlu et al., “Vector Control Of An AC
Brushless Servomotor Using A Custom-Designed
Motion Control Card,” in 2006 IEEE Int. Symp. on
Ind. Electron., 2006, no. 7, pp. 2516–2521.
[18] J. Ollervides et al., “Feedback Electronic Power
Drive for a Brushless AC Servomotor,” in 2011
IEEE Electron., Robotics and Automat. Mech.
Conf., 2011, pp. 264–269.
[19] L. Zhang et al., “Design of AC servo motor control
system based on XC164CM microcontroller,” Proc.
2012 Int. Conf. Ind. Control Electron. Eng. ICICEE
2012, pp. 1277–1279, 2012.
[20] M. Sreejeth et al., “Monitoring, control and power
quality issues of PLC controlled three-phase AC
servomotor drive,” 2012 IEEE 5th Power India
Conf. PICONF 2012, 2012.
[21] P. S. Puttaswamy and K. D. Dhruve, “Multi-layer
Neural Network for Servo Motor Control,” vol. 71,
no. 14, pp. 32–37, 2013.
[22] J. Yin et al., “Fuzzy Adaptive PID Controlling of
Servo Motor System Based on DSP,” Fifth Int.
Conf. Intell. Human-Mach. Syst. Cybern., pp. 26–
28, 2013.
[23] K. Matsuura et al., “Robust current control
insensitive to gain deviation and offset of inverter

Abdul Wali Abdul Ali et al. / ELEKTRIKA, 19(2), 2020, 22-39
39
DC-link current sensor for SPMSM,” in 2014 Int.
Power Electron. Conf. (IPEC-Hiroshima 2014 -
ECCE ASIA), 2014, vol. 5, pp. 1516–1521.
[24] T. Takahashi and I. Rectifier, “High Performance
AC Drive by Single Chip Motion Control Engine
IC,” Interface, pp. 1–8.
[25] B. Allotta et al., “Development of sensorless PM
servo-system for harsh environments,” in 2015
AEIT Int. Annu. Conf. (AEIT), 2015, pp. 1–6.
[26] C. H. Yan and J. H. Hui, “Full-Closed Loop Servo
Control System of Gear Measuring Centre,” 2015
3rd Int. Conf. Comput. Comput. Sci., pp. 37–40,
2015.
[27] S.-M. Yang and K.-W. Lin, “Automatic Control
Loop Tuning for Permanent Magnet AC Servo
Motor Drives,” IEEE Trans. Ind. Electron., vol. 46,
no. c, pp. 1–1, 2015.
[28] Y. Sang et al., “The practical control technology
design for AC servo motor based on STM32 micro-
controller,” in 2015 IEEE Advanced Inform. Tech.,
Electron. and Automat. Control Conf. (IAEAC),
2015, pp. 1–5.
[29] J. Bermingham et al., “Optimized control of high-
performance servo-motor drives in the field-
weakening region,” Conf. Proc. - IEEE Appl. Power
Electron. Conf. Expo. - APEC, vol. 2016–May, pp.
2794–2800, 2016.
[30] M. Gong et al., Bio-Inspired Computing -- Theories
and Applicat., vol. 562. Berlin, Heidelberg:
Springer Berlin Heidelberg, 2015.
[31] H. Wang et al., “Continuous terminal sliding mode
control with extended state observer for PMSM
speed regulation system,” Trans. Inst. Meas.
Control, pp. 1–10, 2016.
[32] S. Wang et al., “A High Performance Permanent
Magnet Synchronous Motor Servo System Using
Predictive Functional Control and Kalman Filter,”
J. Power Electron., vol. 15, no. 6, pp. 1547–1558,
Nov. 2015.
[33] C. Dang et al., “Analysis and reducing methods of
cogging torque on permanent magnet AC servo
motor,” in 2014 17th Int. Conf. on Elect. Mach. and
Syst. (ICEMS), 2014, pp. 2136–2140.
[34] C. Xia et al., “Torque Ripple Reduction in
Brushless DC Drives Based on Reference Current
Optimization Using Integral Variable Structure
Control,” Ieee Trans. Ind. Electron., vol. 61, no. 2,
pp. 738–752, 2014.
[35] S. P. Singh et al., “Performance Comparison of
PMSM Drive using PI and Fuzzy Logic based
controllers,” pp. 563–569, 2016.
[36] Tianrui Luan et al., “A novel Active Disturbance
Rejection Control speed controller for PMSM
drive,” in 2016 IEEE 8th Int. Power Electron. and
Motion Control Conf. (IPEMC-ECCE Asia), 2016,
no. 2, pp. 116–120.
[37] Y. Zhang et al., “An Active Disturbance Rejection
Control of Induction Motor Using DSP + FPGA,”
pp. 4047–4052, 2013.
[38] W. Bin et al., “A New Speed Control Strategy for
Interior Permanent Magnet Motor at Full Speed
Range,” pp. 533–538, 2014.
[39] Z. Ding et al., “Speed identification and control for
permanent magnet synchronous motor via sliding
mode approach,” Syst. Sci. Control Eng., vol. 2, no.
1, pp. 161–167, Dec. 2014.
[40] H. Wang et al., “Generalized Proportional Integral
Observer Based Sliding Mode Control Method for
PMSM Speed Regulation System,” pp. 3482–3487,
2015.
[41] H. Mekki, “Fault Tolerant Control of PMSM Drive
Using Sliding Mode Strategy,” pp. 576–581, 2015.
[42] H. Liu and S. Li, “Speed Control for PMSM Servo
System Using Predictive Functional Control and
Extended State Observer,” IEEE Trans. Ind.
Electron., vol. 59, no. 2, pp. 1171–1183, Feb. 2012.
[43] X. Li and S. Li, “Speed Control for a PMSM Servo
System Using Model Reference Adaptive Control
and an Extended State Observer,” vol. 14, no. 3, pp.
549–563, 2014.
[44] F. Authors, “COMPEL : The International Journal
for Computation and Mathematics in Electrical and
Electronic Engineering Article information :,” 2015.
[45] Y. Yan et al., “Disturbance Rejection Control
Method Based on Composite Disturbance Observer
for Permanent Magnet Synchronous Motor,” no.
61473080, pp. 3137–3142, 2015.
[46] Y. Wang et al., “Magnet Synchronous Motor
Control System,” pp. 5065–5070, 2016.
[47] W. K. Wibowo and S. Jeong, “Genetic algorithm
tuned PI controller on PMSM simplified vector
control,” J. Cent. South Univ., vol. 20, no. 11, pp.
3042–3048, Nov. 2013.
[48] R. Kannan et al., “Selection of PI Controller Tuning
Parameters for Speed Control of PMSM using
Biogeography based Optimization Algorithm,” no.
ii, pp. 0–5, 2016.