Wasit Journal of Computer and Mathematic Science Vol. (2) No.

(1) (2023 )

A Control System of DC Motor Speed: Systematic

Review
https://doi.org/10.31185/wjcm.121

Muhammad Hilal Mthboob1(*), Haider TH. Salim ALRikabi2, Ibtisam A.

Aljazaery3
1
Electrical Engineering Department, College of Engineering, Wasit University, Wasit, Iraq
2
Babylon University, Babylon, Iraq

[email protected]

Abstract—The first sources of direct current (DC) were invented, DC ma-

chines were one of the first types of electro-mechanical machines used. DC ma-
chines are more advantageous over AC machines as regards to Speed regulation
and versatility. A DC motor is an electrical actuator with a lot of control that is
used in a lot of applications, like robotic manipulators, guided vehicles, steel roll-
ing mills, cutting tools, overhead cranes, electrical traction, and other applica-
tions. Due to their speed-torque characteristics and ease of control, DC motors
are utilized extensively in industries for demanding variable speed applications.
In terms of controller design and implementation, the process control industry
has seen numerous advancements over the past two decades. In the industry, there
is a great demand for automatic controllers that can respond quickly and accu-
rately to perform precise tasks. The feedback loop is an essential component of
system control that must be utilized in order to achieve the desired performance
in the majority of systems. Numerous control strategies have been developed for
various feedback control systems in order to achieve rapid system dynamic re-
sponse. Controls in a drive system are crucial if the reference speed is to be ac-
curately and quickly tracked, with little or no steady-state error and as little over-
shoot as possible. This paper presents the Systematic literature review that was
conducted as covers pertinent established concepts and techniques related to the
DC motor speed control system design, for applications that require actuators
with accurate speed characteristics. Simulation and real time implementation re-
sults employed for DC motor speed control systems in various literature are ana-
lysed and discussed.

Keywords— DC Motor, Speed, Controller, PWM, PID

1 Introduction

For many decades, the Brushed DC machine has been the automatic choice where
speed torque control was very necessary [1-3]. Figure 1 shows a schematic of a brushed

93
Wasit Journal of Computer and Mathematic Science Vol. (2) No. (1) (2023 )

DC motor showing all its important part consisting of the Commutator, Brushes, Ar-
mature conductor, and the Field windings. The applications of a DC motor range from
steel rolling mills, electric traction, center winders to a very wide range of industrial
drives, robotics printers and precision servos. The range of power outputs is wide,
which varies from several mega-watts to a few watts of power as mentioned in [4].
Also, the speed of DC machines can be easily varied and controlled which makes them
very suitable for precision industrial activities. Their speed-torque characteristics also
makes them very useful in various applications. A few major components of a DC mo-
tor are described in turn. These parts provide a major contribution in the operation of a
DC machine. They are also used to determine the kind of DC machine in use as men-
tioned in [5].

Fig 1. Conventional Brushed DC Motor [5].

1.1 The parts of A DC machine

1. COMMUTATOR: A ring on a motor shaft designed to maintain a current flow
pattern such that positive pole current is against the positive magnetic pole and the
direction of the negative current against the negative magnetic pole, connecting with
all motor
2. BRUSHES: The brushes are made of carbon, used to supply electric current to mo-
tor parts, while in generators they are used to extract current to the external circuit.
3. ARMATURE CONDUCTORS: This consist of copper conductors wound on a lam-
inated former. Current flowing through the armature windings when interact the
magnetic field in the air gap of motor
4. FIELD WINDINGS: The field windings are responsible for developing magnetic
flux during motor operation as mentioned in [6]. Field windings configuration are
divided to group dc machines into two types namely, separately excited dc machine
and self-excited dc machine, then there is PMDC machine. Self-excited machines
develop their field magnetic flux from the current applied to the motor terminals i.e.
the field and armature windings are excited from the same source, separately excited
machines rely on an external voltage source for production of magnetic flux. PMDC
machines rely on magnets for production of field magnetic flux. The type of DC
motors can be divided as below:

94
Wasit Journal of Computer and Mathematic Science Vol. (2) No. (1) (2023 )

The connection between the field winding and the armature distinguishes the three
varieties of self-excited DC machines that are the subject of the subsequent discussion.
A. Shunt-wound motor a shunt-wound motor has parallel connections between
the armature and field windings. The armature current and the current that runs through
the shunt field windings are not the same thing. The shunt field windings are con-
structed with a large number of high-resistance wire turns to generate the required
MMF. This suggests that the shunt field is smaller than the armature current.
B. A series of motor injuries: This kind of motor has armature and field windings
connected in series. The armature current is carried by the series field windings.
C. A motor with a compound wound design a motor with a compound wound
design has both a series and a shunt field winding, or one winding in series and one
winding in parallel with the armature circuit. The series field windings must be de-
signed with significantly fewer turns for the same MMF due to the identical magnitude
of the armature current passing through the shunt field windings. The series field wind-
ing will have a low resistance due to its relatively small number of thick wires as men-
tioned in [7]. When the shunt field winding is connected directly across the armature
terminals, this is called a "short shunt connection." Otherwise, when the shunt windings
are connected in such a way that they shunt the series connection between the armature
and the series field, a long shunt connection is made. The shunt field windings of the
compound motor generate significantly more flux than the series field windings. In re-
cent times, the need for accurate speed control in industries as led to the demand for
highly controllable actuators. DC machines are versatile and robust machines used ex-
tensively in industries as actuators for a wide range of operations as mentioned in [8].
Torque-speed characteristics of DC machines is highly demanded in most industrial
applications. This has led to extensive research interest in the fabrication of robust con-
trol systems for DC machines for carrying out more precision tasks in industries.

2 Control Speed D.C Motor

2.1 Direct Current Speed Control Methods

A DC motor speed control that utilizes both Field and Armature control simultaneously
has been proposed by researchers in [9]. It was discovered that the field circuit resistance
of the windings should be varied in order to increase the speed of a Shunt DC motor. As
the resistance rises, the current flow in the field circuit decreases. As a result of this action,
the flux produced by the field windings decreases, resulting in an increase in motor speed.
Additionally, in order to reduce the motor's speed, the resistance of the armature winding
must be increased while the resistance of the field winding must be maintained at a mini-
mum. The armature experiences a large voltage drop as a result of this action. We are able
to deal with a variety of motor speeds using this method because the supply voltage is
constant, and equation 2.1 can be used to analyses the system. Researchers in [10] described
a method for controlling the speed of a DC motor by altering the machine's terminal volt-
age. The terminal voltage, not the armature voltage, is the most common method for con-
trolling the speed of a DC motor. This effect can be described using the equation IA = VT
EA RA. It can be deduced from the equation that an increase or decrease in the DC motor's
terminal voltage causes an increase or decrease in the armature current. Taking into account

95
Wasit Journal of Computer and Mathematic Science Vol. (2) No. (1) (2023 )

an increase in the armature voltage, which corresponds to an increase in the armature cur-
rent, will result in an increase in the torque that is generated, as shown by the equation Tind
= KIA. The armature current will decrease as a result of the increased torque, which will
further reduce the torque induced. The speed of the motor shaft increases as a result of the
increased torque at constant load (). The speed increases the inside produced EMF (EA),
also known as back EMF (EA = K). The induced torque is reduced until it reaches the
equivalent of the load torque. The motor will accelerate and enter a steady state at this point
as mentioned in [11]. In [12], they described the speed control of a DC motor by changing
the armature resistance which will lead to a corresponding change in the armature current.
Since increasing the armature current leads to the increasing in speed. Hence, it is evident
that increasing the resistance of the armature leads to low armature which leads to decrease
in speed of the DC motor. This method can be used to decrease the speed of a motor below
its base speed. The problem faced by this method is that the increase in resistance will
increase power losses. This method can be used if the motor runs at the base speed most of
the time and only for a short time slow speed is required.

2.2 Pulse Width Modulation (PWM) Method

Researchers in [13] proposed a DC motor fixed speed control with high precision and
dependability. An LCD display was used to actually monitor the motor's performance, and
the ATmega128 microcontroller was used for the implementation. The motor rotates at a
predetermined speed when it receives manual input. Contactless infrared sensors are used
to monitor the speed of the motor. PWM signals are used by the microcontroller to control
a MOSFET transistor in the DC motor's drive circuit, which supplies voltage directly to the
motor. The PWM signal from the microcontroller will cause the motor to rotate at a prede-
termined speed. The operation of pulse width modulation (PWM) entails creating a square
wave with a variable on-off ratio. The average percentage of pulses that arrive on time can
range from 0% to 100%. Because of this effect, the load receives varying amounts of
power. The primary advantage of a PWM circuit over a resistive circuit is its efficiency;
At a 50% level, the PWM will transfer nearly all of the power to the load and consume
approximately 50% of the full power as mentioned in [14]. A resistive controller can use
up to 71% of its full power to heat the series resistor and 50% to the load at 50% load
power. It was discovered that the design with fixed speed control and PWM for DC motor
control was reliable and fairly accurate. In [15], they has proposed a method of speed con-
trol of a DC motor using analogy PWM technique. In this method an analogy PWM is
developed, that drives the DC motor by switching four H-bridge connected MOSFETs. The
H-bridge configuration results in the realization of a bidirectional full bridge circuit capable
of operating in four quadrants of the I-V graph. For effectively controlling the motor, a
systematic approach was laid for determining the switching frequency of the PWM pulses,
as the switching frequency has a direct effect on the motor current ripples due to the fact
that a dc motor works like a low pass filter and has a bandwidth of allowable frequencies.
A method of choosing the PWM frequency based on the motor characteristics was pre-
sented. It works by setting an allowable percent for the current ripples, then the minimum
frequency to attain that goal is worked out mathematically. Firstly, the average power of
the motor during switching is determined using equation 1.

𝑃 = 2𝐼2𝑅 (1)

96
Wasit Journal of Computer and Mathematic Science Vol. (2) No. (1) (2023 )

The DC motor current response to a PWM pulse at 50% duty cycle is shown in figure 2

Fig 2. DC motor basic constructive elements and fixtures [15].

T is given as the switching period of the pulses. Figure 2 shows the motor at stall con-
dition which is the highest operating current in the DC motor i.e. the worst case scenario.
These formulas were used to prove that the optimal frequency for the PWM pulse is not
the highest possible frequency, but somewhere between high and low in the KHZ range.

2.3 Control schemes

A. Single Loop Feedback Control

A control system with a feedback mechanism maintains a prescriptive relationship by

continuously comparing the process output and the set point utilising the error signal as a
means of control. Feedback control is the most straightforward closed loop control strategy.
A feedback control system can be used for many commonplace tasks, such as regulating
an air conditioner's temperature or an automobile's speed as mentioned in [16]. By exploit-
ing the difference between the actual speed and the intended temperature, these systems
modify the controlled variable. When the output is utilised to regulate the output, the sys-
tem is said to be closed-loop. The block diagram in Figure 3 shows a simple feedback
control system. Different variation of a feedback loop is used in the design of speed control
of DC motor, due to its simple and robust structure

97
Wasit Journal of Computer and Mathematic Science Vol. (2) No. (1) (2023 )

Fi 3. Feedback Control System [16].

B. Cascade Feedback Control

This type is most frequently used in the steam process industry and speed control drives.
A cascade control consists of an inner loop and an outer loop. A cascade control method
performs at its best when the inner loop has a faster dynamic than the outer loop. The
structure of a cascade scheme is seen in examples 2 and 3. The inner loop regulates the
secondary process, such as the voltage at the armature terminal of a DC motor. The outer
loop regulates the primary process, such as the angular displacement of a DC motor as
mentioned in [17]. The exterior circle regulator is first left in manual mode while the inter-
nal circle is tweaked first. The inner loop can be adjusted using the direct synthesis, Ziegler-
Nichols, RA, and metaheuristic approaches (IWO, GA, etc.). The outside tuning completes
the tuning process.

Fig 4. Cascade Control System [17].

Most literature investigate the speed control of a DC motor as a feedback control loop
otherwise known as a closed loop control system. The research activities carried out on
these areas shows the need for continuous improvements on the simple feedback loop in
controlling the speed of industrial actuators (DC motors). Literature based on various kinds
of controllers used for DC motor speed control are presented and discussed.

3 GRNN Algorithm
The GRNN algorithm is by far the most frequently used control algorithm. The majority
of feedback loops are handled by this algorithm or one of its minor variants. A GRNN
regulator's design is depicted in Figure 5. It is possible to put it into action in a number of

98
 Wasit Journal of Computer and Mathematic Science Vol. (2) No. (1) (2023 )

different ways, such as a stand-alone controller in a PC, PLC, DCS, or microcontroller, as

part of a Direct Digital Control (DDC) package, or as part of a hierarchical distributed
process control system. The GRNN algorithm can be approached in a variety of ways. It
can be approached analytically or as a tool that can be utilized with a few general rules.
The "textbook" version of the GRNN algorithm is given in equation 2. In the following
section, researchers in [12] discuss the various components of a GRNN algorithm and how
they affect the performance of the control system:

Fig 5. Diagram of a GRNN Control System

Where “u” is the control variable and e is the control error (e = ysp – y) where ysp is the
Set- point and y is the Feedback control signal. The control variable is thus a sum of three
terms: The P-term (which is the proportional to the error), the I-term (which is proportional to
the integral of the error), and the D-term (which is proportional to the derivative of the error).
The controller parameters are proportional gain K, integral time Ti and the derivative time Td.
The GRNN parameters are further discussed in [18].
The proportional action considers the present state error only. The proportional signal is a
measure of difference between the present value (PV) and set point (SP). The magnitude of
the proportional signal increases with this difference. When the PV approached the SP, the
error becomes so small so that the controller cannot trigger the PV to meet up with the SP.
This implies that there is always a steady state error, which is an offset value form the set point
in the system. This effect leads to the choice of higher value for the gain of the proportional
term. This higher value, however, makes the system unstable with oscillations and overshoots,
which makes the behavior like that of an on-off controller as mentioned in [19]. It can be
concluded that the P-term cannot sufficiently satisfy accurate system control especially in
higher order systems with more than one energy storage elements. The P-term is usually in
conjunction with the Integral or Derivative term. The Mathematical equation of the output of
a P-controller is given by the equation 2.

(𝑡) = 𝐾𝑝 (𝑡) (2)

The integral action of a GRNN algorithm is proportional to the integral of the control error.
In terms of equation. It turns out that the integral action and the control errors past and present
values are linked. Due to the presence of a pole at the complex plane's origin, the steady-state
error can be reduced to zero when a step load disturbance or step reference signal is applied.

99
 Wasit Journal of Computer and Mathematic Science Vol. (2) No. (1) (2023 )

Because the integral action is able to set the term up to the appropriate value on its own, this
indicates that there is no steady state error. The integral term is utilized when the controller
must correct for any steady state offset from a constant reference signal value. Integral control
overcomes the drawback of proportional control by eliminating offset and limiting controller
gain as mentioned in [20]. The rate of change of control error serves as the foundation for the
derivative action. The derivative action has a great potential for improving control performance
because it can anticipate an incorrect trend of the control error and compensate for it. Because
the controller is able to take the rate of change of an error signal as an input, this adds a layer
of prediction to the control action. In contrast to P and I control, care is required when using
subsidiary control. The majority of applications prohibit the use of pure derivative control due
to the possibility of measurement noise amplification. However, derivative control is still re-
quired for some real-world control applications, such as DC motor control tacho generator
feedback.

3.1 DC Motor Speed Control Using GRNN Algorithm

Because a DC motor requires precise speed control, precise control strategies are required.
Because of their heartiness and exactness with regards to accomplishing accuracy speed con-
trol, GRNN regulators are presently quite possibly of the most broadly involved control in
ventures. A GRNN algorithm's performance is significantly affected by the gain parameters
for the proportional, integral, and derivative terms. We will examine and discuss a variety of
newly developed tuning strategies in this section. Due to the difficulty of the tuning process,
research into ways to make GRNN Algorithm even more accurate and user-friendly has re-
quired a significant amount of time and effort as mentioned in [21]. The DC drive consists of
a Power Electronic Modulator (PEM) that converts electrical energy from the source into a
form suitable for the motor, an input and output filter that performs signal processing functions
to remove unwanted frequency components from the signal, and a controller that controls the
power modulator. Researchers discussed the DC motor drive and controller for speed control
under varying load conditions in [12]. The various controllers that were looked at were P (pro-
portional), PI (proportional integral), and GRNN (proportional integral and derivative). Be-
cause it has a smaller amplitude and phase margin, faster dynamics, and a lower steady state
error, the proportional controller is only suitable for first-order systems. It was discovered that
the PI controller performed poorly when the plant controller was highly non-linear. It is only
used when there are significant noise and disturbances during the process and the system can-
not respond quickly. When the Kd is low, the GRNN algorithm's rise time, settling time, steady
state, and stability were found to be more suitable for controlling the PEM. A closed loop speed
control scheme for the entire system consists of an outer speed loop and an inner current loop.
The inward current circle lessens the converters and engine's force and current underneath as
far as possible. Additionally, it helps to lessen the impact of any converter-motor system non-
linearity on drive performance. The Simulink simulation result demonstrates that the motor
speed only slows by approximately 270 rpm (9%) in 980 milliseconds under full load. The
motor is also hunting at approximately 200 rpm (6.66 percent) in 900 milliseconds when it is
in the unloading condition. A researcher in [22] suggested a separately exited DC motor drive
for the speed control application due to its high reliability, controllable torque, simplicity, and
accurate speed control. Using the derived dynamic model of the separately exited DC motor,
a transfer function was created for the relationship between the applied voltage and angular

100
 Wasit Journal of Computer and Mathematic Science Vol. (2) No. (1) (2023 )

position. In order to achieve the desired speed and armature current response in the system's
control loop, a GRNN algorithm was utilized. Utilizing MATLAB/SIMULINK programming,
the framework's block outline was made and recreated with and without the GRNN regulator.
The first step was to compare the system's step response with and without the GRNN algo-
rithm. The system was evaluated both with and without a GRNN algorithm under full load
(load disturbance) conditions. The rise time was greatly reduced by the GRNN algorithm by
approximately 20%, and the settling time was also improved by 20%. It was noticed that the
speed did not reach the desired level when there was no GRNN algorithm. The recent concern
of controlling pace under load was settled by using the GRNN regulator. The first parameter
tuning step of the GRNN algorithm was carried out using the Ziegler-Nichols tuning technique.
Under no load, this approach eliminates steady-state error as well as the issue of overshooting
between reference speed and actual speed as mentioned in [23]. The loading effect's issue with
motor speed undershoot was fixed in the following step through manual tuning. The GRNN
algorithm supplies a voltage to the armature terminals of the DC motor. The final parameters
of the gain values were discovered and simulated. The family of controllers labelled P, PI, PD,
and GRNN that make up a GRNN algorithm can be seen in Figure 6 of the selection flowchart.
The flowchart provides a guide for choosing the best controller strategy for a variety of appli-
cations that require precise

Fig 6. GRNN Term Selection Flowchart.

3.2 GRNN Tunning Methods

The second step in setting up the GRNN algorithm is to tune or select numerical values for
the GRNN coefficients after selecting the control process's GRNN design, as shown in Figure
2-6. Tuning is the process of adjusting control parameters to achieve the best possible response
from the system. For various control systems, stability is a crucial requirement as mentioned in
[24]. However, the behaviour of various systems and the requirements of various applications
may conflict with one another. GRNN tuning is troublesome issue, even though there are just
three boundaries which is easy to portray on a fundamental level. This is since parameters must
fulfil intricate criteria within the confines of GRNN control. There are several different kinds of
GRNN tuning strategies, some of which include:

101
 Wasit Journal of Computer and Mathematic Science Vol. (2) No. (1) (2023 )

Classical Techniques:
1. Manual Tuning method
2. Ziegler-Nichols method
3. Internal Model control (Model-based)
4. Pole-Placement method
Heuristic Algorithms:
1. Genetic Algorithm (GA)
2. Invasive Weed Optimization (IWO)
3. Simulated Annealing
3.2.1 Manual Tunning Method

The manual tuning method entails changing the system's parameters and monitoring the re-
sults. The parameters (KP, Ki, and Kd) are adjusted until the system behaves as expected. De-
spite how easy it is to use; only trained professionals should employ this technique. Researchers
in [25] discussed a manual tuning procedure in which the parameters Ki and Kd are initially set
to zero. Kp is then increased until an oscillation is seen in the loop's output. The appropriate Kp
value should be set to around half of that for a "quarter amplitude decay" response. Then, unless
any offset is corrected in time for the process, the Ki is raised. However, too much Ki will cause
instability. After a load disturbance, the Kd parameter is increased until the loop returns to its
reference table 1.
Table 1. Effect of changing control parameters

Parameter Rise Time Overshoot Settling Steady

Time State Error
KP Decrease Increase Minor Change Decrease
KI Decrease Increase Increase Eliminate
Kd Minor Change Decrease Decrease Minor Change

Manual tuning can also be done using the MATLAB software GRNN tuner which allows to
tune the response of the system according to the desired transient response and loop stability
criteria. A method for tuning the gains of a GRNN algorithm optimally using MATLAB Simu-
lation Software. The design criteria were first set be Rise time <3seconds, Settling time <10
seconds, Overshoot of <5% and a steady state error <1%. The plant to be controlled was a tem-
perature controlling system, which could be used in a furnace, industrial chemical process etc.
A first order model and a second order model approximation were derived for the system. For
the first order system, a Proportional controller with gain of 12 satisfied the desired performance
criteria. Unlike the first order system, a proportional controller was not sufficient for achieving
the set criteria. Hence gains of Kp = 110, Ki = 10, Kd = 100 was found satisfying all the control
requirements. Manual tuning using the root locus approach was also presented which makes
used of the MATLAB tool Graphic User Interface (GUI) and enables tuning of the GRNN algo-
rithm by viewing the root locus of the system. The step response of the system is also observed
in the GUI until the desired response is achieved. The GRNN algorithm achieved on MATLAB
can be integrated with Arduino or any other type of microcontroller. The controller can then be
used for generating PWM pulses through its digital pin, which intern can be used to effectively

102
 Wasit Journal of Computer and Mathematic Science Vol. (2) No. (1) (2023 )

control the plant.

3.2.2 Ziegler-Nichols tunning method

The numerical values of the GRNN coefficients are determined by the Ziegler-Nichols meth-
ods through an online process experiment that is based on rules. This frequency domain approach
is based on the experimentally determined point at which the system becomes marginally stable.
A researcher came up with the Ziegler-Nichols closed loop method and the Ziegler-Nichols open
loop method in 2015 to adjust the parameters of a GRNN algorithm. The definition of acceptable
stability serves as the foundation for the Ziegler-Nichols method's controller tuning rules. The
belief that the step response's 14 decay ratio indicates poor control loop stability is widespread
as mentioned in [27]. This is because a disturbance or step change in the control loop's set point
results in the ratio of the amplitudes of subsequent peaks in the same direction. The first step in
improving the system is to lower the KP value if the control loop's stability becomes too low.
The tuning procedure is as follows:
I. Bring the process as close as possible to the specified operating point of the control system to
ensure that the controller is "experiencing" representative process dynamics and to lessen the
likelihood of tuning variables reaching limits. The process variable is manually adjusted until
it is approximately equal to the set-point when the controller is in manual mode to bring the
process to the operating point.
II. The gain KP is initially set to "0" when Ti = and Td = are set to make the GRNN algorithm a
P controller. Switch the controller to automatic mode to end the control loop.
III. Raise KP after the system has been excited until there are long-lasting oscillations in the
control system signals, like in the process measurement. When the system continues to oscil-
late, it has reached its stability limit. The value of Kp at this point is the ultimate (or critical)
gain, or Kpu. The excitation can take the form of a step at the set-point. This step should be
relatively small—say, 5% of the highest set-point range—so that the cycle doesn't get too far
from the place where the unique properties of the interaction might be. In addition, the step
shouldn't be too small because, without it, the measurement noise that is bound to occur could
make it hard to see the oscillations. Researchers in [27] proposed the modified Ziegler Nichols
tuning method as well as the traditional Ziegler Nichols tuning method. The step response of
the various kinds of plants that the Traditional Ziegler Nichols tuning method is applied to is
depicted in Figure 7. A common reaction of a first request framework with a transportation
delay is the reaction. There are two parameters that define these systems' response: the time
delay (L) and the constant time (T). By tracing a tangent to the step response at its inflection
point and locating its intersections with the time axis and steady state value, these points can
be derived...

103
 Wasit Journal of Computer and Mathematic Science Vol. (2) No. (1) (2023 )

Fig 7. Response Curve for Ziegler Nichols Method

When the step response of the system is derived, the output signal can be noted as shown in
figure 2.7, from this figure the parameters k, L, T can be extracted by the straight forward ap-
proach shown in Figure 2.6. With these parameters and the Ziegler Nichols formula shown in
Table 2, the controller parameters of the system can be derived.
Table 2. Ziegler-Nichols tuning first method.

Controller KP Ti Td
P T/L N/A N/A
PI 0.9T/L L/0.3 N/A
GRNN 1.2T/L 2L 0.5L

The Modified Ziegler-Nichols tuning method proposed by [28] uses Chien Hrones Reswick
(CHR) tuning algorithm which emphasizes on set-point regulation. The modified Ziegler Nich-
ols Tuning method unlink the Traditional method, uses the Time constant T of the plant explic-
itly. The controller tuning formulas are given in Table 2-3

Table 3. Modified Ziegler Nichols tuning method

Controller KP Ti Td
P 0.7/a N/A N/A
PI 0.6/a T N/A
GRNN 0.95/a 1.4T 0.47T

The transfer function used to model the dc motor is derived from the values of Kp, Ki, and
Kd in the table. The response curve is used to calculate the parameters K, L, and T. According
to the findings of a computer simulation that was carried out in MATLAB Simulink, the Modi-
fied Ziegler-Nichols tuned GRNN algorithm performs better than the conventional Ziegler-
Nichols tuning method for the control objectives of minimum rise time, minimum overshoot,
and minimum settling time. According to [29], the Ziegler-Nichols tuning method only alters
the proportional constant while maintaining the remaining two gain parameters. Kp is raised by
a factor of two until the system becomes unstable. The oscillation period and gain at that point
are the ultimate period and gain. Using this method, the settling time Ts and maximum overshoot

104
 Wasit Journal of Computer and Mathematic Science Vol. (2) No. (1) (2023 )

Mp can be calculated. For finding the KP, TI, and Td values in a GRNN algorithm, researchers
in [30] suggested using the Ziegler-Nichols method. LQR is used to best control linear plants. A
method for optimal control known as model predictive control, or MPC, can be used to manage
linear and nonlinear systems. Using the process model, MPC predicts how the system will react
to a moving or receding horizon. The Ziegler Nicholas method is a straightforward strategy for
tuning the GRNN algorithm, as the researchers discuss in [31]. The Ziegler-Nichols and Modi-
fied-Ziegler-Nichols tuning techniques are utilized for GRNN algorithm tuning for DC motor
speed control. By reducing transient response parameters like rise time, settling time, and per-
centage overshoot, the DC motor's speed response can be improved. The Ziegler Nicholas
method, according to the researchers in [32], measures speed and uses a closed loop system to
provide feedback to the system. The integral and derivative parameters of the GRNN will be set
to zero at initialization. The proportional term Kp gradually rises from zero until the system
begins to continuously oscillate. The value of the proportional coefficients at this point is the
ultimate gain (Ku), and the period of oscillation at this value is the ultimate period (Tu). The
regulator's increase (Kp) has little value when the regulator is in corresponding mode, such as
when the primary and subsidiary terms are set to nothing, which causes the framework to re-
spond slowly. By increasing Kp by two, the response becomes unstable and oscillatory. Last but
not least, the proportional gain is altered until the response continues to oscillate.

4 Invasive Weed Optimization (IWO)

The social behaviors of swarms, such as bird flocking and fish schooling, served as the foun-
dation for the Invasive Weed Optimization algorithm, which was developed by researchers in
[33]. Each bird is referred to as a particle in this context, and in the context of a GRNN algorithm,
each particle possesses the three characteristics Kp, Ki, and Kd. The social-psychological ten-
dency of individuals to imitate the success of other individuals is the basis for the changes in
position of each individual particle within the search space in IWO. Particles are used to describe
the individuals who are "flynned" through the hyper-dimensional search space. To put it another
way, a particle's changes in the swarm are influenced by its neighbours’ experience or
knowledge. The collective behaviour of finding the best locations in a high-dimensional search
space is facilitated by this straightforward behaviour. The IWO algorithm keeps particles in a
swarm, with each particle representing a possible solution. The d-dimensional vector xi(t), which
is equivalent to (xi1, xi2, xi3, ... xid), represents the position of particle i in the search space at
time step t. The swarm's position is altered by adding a velocity, vi(t), as shown in equation 3.

𝑋𝑖 (𝑡 + 1) = (𝑡) + 𝑣𝑖 (𝑡 + 1) (3)
The optimization process is driven by the velocity vector vi(t), which reflects both the expe-
riential knowledge of the particle and the information that is socially exchanged from the parti-
cle's neighborhood. The cognitive component of a particle is its experiential knowledge, and its
distance from its best position since the first time step is proportional to this distance. The social
component of the velocity equation is the information that is shared socially. The neighborhood
for each particle in global best IWO, also known as gbest IWO, is the entire swarm. Together,
the practices form a social network resembling a star topology. The social part of the particle
velocity update in this type of topology shows information from all of the swarm's particles,
including the best position (collection of all of the particles). Equation 4 provides the formula

105
 Wasit Journal of Computer and Mathematic Science Vol. (2) No. (1) (2023 )

for calculating particle i's velocity update.

𝑣𝑖𝑗 (𝑡 + 1) = 𝑣𝑖𝑗 (𝑡) + 𝑐1𝑟1(𝑡)[𝑦𝑖𝑗(𝑡) − 𝑥𝑖𝑗(𝑡)] + 𝑐2𝑟2(𝑡)[ŷ𝑗(𝑡) − 𝑥𝑖𝑗(𝑡)] (4)

c1 and c2 are positive acceleration constants that are used to scale the contribution of the cog-
nitive and social components, respectively, and r1j and r2j are random values in the range of [0,
1] that are sampled from a uniformly distributed distribution. Where vij (t) is a vector that repre-
sents the velocity of particle i in j dimensions at time step t, xij (t) is the position of particle i in
dimension j at step t. The algorithm is given a stochastic edge by these random values (r1j, r2j),
which also cause the particles to act randomly. Particle i's best position since the first step is the
personal best, or yi. The personal best for particles in IWO will be the coordinates of the particles
that minimize the fitness function f. The DC motor data was imputed to the transfer function on
SIMULINK, and differential equations for the model as well as the state space representation
were derived. Additionally, the open loop transfer function and closed loop system response
were derived and analysed, respectively. The goal of GRNN tuning is to minimize gain values.
Finally, Ziegler Nichols' GRNN tuning method was used to first tune the speed control's GRNN,
and then the results—settling time, overshoot, rise time, steady, and state error—were recorded.
Using the IWO algorithm, the same thing was done, and the results were recorded. Simulink was
used to examine the closed loop control system's step response for both the IWO and ZN (Ziegler
Nichols)-tuned controllers. The presentation of IWO calculation technique for tuning was
demonstrated to be significantly better compared to conventional strategy like Ziegler - Nichols
strategy as far as framework overshoot, settling time and rise time. The use of IWO in the tuning
of a separately excited DC motor by a GRNN algorithm has been discussed and analyzed by
researchers in [33]. In order to reduce the steady state error, rise time, maximum overshoot, and
settling time, the fitness function in IWO was utilized. Furthermore, the system's cost function
was minimized, as shown by the integrated square error (ISE). The proposed controller demon-
strated the excellent performance of the GRNN algorithm when tuned with an IWO algorithm
and that an optimized speed response is always obtained when the reference input speed changes.
In tuning a GRNN algorithm for a DC motor, researchers in [34] discussed how IWO and the
Genetic algorithm performed in comparison to one another. A unit step signal was used as an
input reference signal for the DC motor and an output signal was obtained for tuning the GRNN
algorithm based on some criteria. Using the MATLAB platform, an overshoot, a rise time, a
settling time, and a steady-state error were identified, and the output signals were simultaneously
analyzed. The DC motor GRNN algorithm underwent the initial process of manual tuning. The
GRNN was then tuned using IWO and GA. The parameters, Max overshoot, Rise Time, Settling
Time, and Steady state error saw significant reductions in these algorithm's IAE, ITAE, ISE, and
ITSE results. When IWO was used, the Max overshoot went from 60.9959 for the IAE index to
25.4250 for the ITAE index, which was 12.9682. Additionally, a special fitness function was
obtained, and its values decreased. The exceptional wellness capability, W (k) delivered more
agreeable qualities than other wellness capability as far as overshoot, immersion time, and rise
time. A GRNN algorithm for DC motor speed control will be examined and implemented in this
research project. Ziegler Nichols and IWO tuning can be used to get the best controller gain
values.

4.1 IWO Model-Based Technique

A classical Invasive Weed Optimization (PI) tuning technique which is model based in nature

106
 Wasit Journal of Computer and Mathematic Science Vol. (2) No. (1) (2023 )

and requires just the motor parameters and the cross-over frequency or bandwidth of the con-
troller to be specified the gains of the controller are achieved by pole cancellation. Although it
required lot of in depth understanding about the plant and also some pre-calculation might have
to be done, it is comparatively faster to set up. This method was applied to tune a cascaded IWO
algorithm for a PMDC motor. Firstly, the current loop was determined. In the case whereby the
mechanical inertia of the motor is big enough, the back-EMF factor in the current loop acting as
disturbance can be neglected. A simplified current loop was derived as shown in figure 8.

Fig 8. Simplified Current Loop using IWO.

Hence, the gains of the speed loop were calculated with equation choosing the cross-over
frequency of the speed loop, 𝜔𝑐𝑠 to be up to 10 times of the current loop crossover frequency.
Figure 9 shows the speed loop. The current loop is assumed to be ideal and equals to 1.

Fig 9. Speed Loop using IWO

5 Direct current to direct current (DC-DC) converter

Both regulated switch-mode dc power supplies and DC motor speed drive applications fre-
quently make use of DC-DC converters. The ideal input for DC-DC converters would be an
unregulated voltage obtained by rectifying the line AC voltage, which is susceptible to fluctua-
tion due to changes in the line voltage's magnitude. Consequently, switched-mode DC-DC
power electronics are utilized to transform this unregulated dc input into a controlled dc output
at the motor's desired voltage. Step-down (buck), Step-up (boost), Step-down/step-up (buck-
boost), and Full-bridge converters are power electronics dc-dcc converters. Buck (step-down)
converters are the foundation for full bridge converters. Figure 10 provides a summary of the
switching operation used in switched-mode operation in a dc-dc converter.

Fig 10. Switching Pulse Generation with comparator signal [35].

107
 Wasit Journal of Computer and Mathematic Science Vol. (2) No. (1) (2023 )

Buck converters are mainly used in dc motor speed drive applications. The basic idea of a
step- down converter like a Buck converter is to produce a lower average output voltage than
the dc input voltage Vd. When the associated load connected to the buck converter is an induc-
tive load, the switch will have to be equipped with some mean of absorbing or dissipating the
inductive energy during the off position of the switch or else the switch could be destroyed.
Researcher in [36] has proposed a four-quadrant closed loop speed control of a DC motor,
which uses similar principle of operation as a buck converter. Their proposition offered a low-
cost approach to high performance chopper based four quadrant closed loop speed control for
a separately excited DC motor. The drive system was modelled and simulated using
Scilab/XCOS. The control strategy used ensures that the output voltage of the DC-DC con-
verter can be controlled both in magnitude and direction. The four-quadrant converter consists
of two switches on a leg as mentioned in [37]. The switching scheme was made to avoid turning
on both switches on the same leg which would lead to shorting the source and causing damage.
Also the switching of in the DC-DC converter was done using pulse width modulation PWM.
A variation of PWM technique called Unipolar Switching also referred to as double PWM
switching was used. This method is described in figure 10. Here a high frequency triangular
waveform is compared with control voltages +Vcontrol and -Vcontrol and is used for determining
the switching signals for the switches on Leg 1 and 2 on the four quadrant DC-DC converter.

Fig 11. PWM Using Unipolar Voltage Switching [37].

The DC motor mathematical model was established using the dynamic equations of the
electrical and mechanical part of the motor. A PI controller with gains of 0.25 and 0.001 for
the proportional and integral respectively was derived using Trial and Error tuning method
which yielded the best result. The hardware implementation consisted of a 3-phase AV supply
which was to a three-phase bridge rectifier and a chopper circuit. The controller circuit was
realized using a L28069M Launchpad with OP-amp based level shifting circuit using a LM339.
The four-quadrant chopper was realized by using 4 high power IGBT transistors with two
connected on each arm as mentioned in [38-49]. It was established that the four-quadrant dc-
dc converter alongside the PI controller was robust for effectively controlling the speed of the
motor in both anti- clockwise and clockwise direction.

108
Wasit Journal of Computer and Mathematic Science Vol. (2) No. (1) (2023 )

6 Conclusion
When direct current (DC) power sources were first developed, DC machines were
among the first electromechanical devices to be used. In terms of speed regulation and
versatility, DC machines are superior to AC machines. A DC motor is a powerful elec-
trical actuator that has a wide range of uses, including electrical traction, overhead
cranes, steel rolling mills, cutting tools, robotic manipulators, and guided vehicles. DC
motors are widely used in industries for difficult variable speed applications because of
their speed-torque characteristics and simplicity of control. The process control indus-
try has made many strides in the last 20 years in terms of controller design and imple-
mentation. Automatic controllers that can react quickly and accurately to complete pre-
cise tasks are in high demand in the industry. In order to get the desired performance
from the majority of systems, the feedback loop, which is a crucial part of system con-
trol, must be used. To achieve a quick system dynamic response, a variety of control
strategies have been developed for various feedback control systems. In order to track
the reference speed quickly and accurately, with as little steady-state error and as little
overshoot as possible, controls in the drive system are essential. This paper presents the
results of a systematic literature review that was carried out in order to design a DC
motor speed control system that is appropriate for applications that call for actuators
with precise speed characteristics. Analyzed and discussed are simulation and real-time
implementation results used for DC motor speed control systems in various works of
literature

7 References

[1] El Ouanjli, N.; Motahhir, S.; Derouich, A.; El Ghzizal, A.; Chebabhi, A.; Taoussi, M. Im-
proved DTC strategy of doubly fed induction motor using fuzzy logic controller. Energy
Rep. 2019, 5, 271–279.
[2] Farahani, G.; Rahmani, K. Speed Control of a Separately Excited DC Motor Using New
Proposed Fuzzy Neural Algorithm Based on FOPID Controller. J. Control. Autom. Electr.
Syst. 2019, 30, 728–740.
[3] Hameed, W.I.; Sawadi, B.A.; Muayed, A. Voltage Tracking Control of DC-DC Boost Con-
verter Using Fuzzy Neural Network. Int. J. Power Electron. Drive Syst. 2018, 9, 1657–1665.
[4] Bista, D. (2016). Understanding and Design of an Arduino-based GRNN algorithm.
[5] Dani, S., & Ingole, D. (2017). Performance evaluation of GRNN, LQR and MPC for DC
motor speed control.
[6] Engelbrecht, A. P. (2007). Computational Intelligence. An Introduction. South Africa: John
Wiley and Sons, Ltd.
[7] Fatah, I. S. (2014). IWO-BASED TUNNING OF GRNN ALGORITHM FOR SPEED
CONTROL OF DC MOTOR.
[8] G¨ucin, T. N., Bibero˘glu, M., Fincan, B., & G¨ulbahc¸e, M. O. (2015). Tuning Cascade
PI(D) Controllers in PMDC Motor Drives: A Performance.
[9] Gajanan, M. (2011). Design of GRNN algorithm for improved performance of higher order
systems.

109
Wasit Journal of Computer and Mathematic Science Vol. (2) No. (1) (2023 )

[10] Getu, B. N. (2019). Tuning the Parameters of the GRNN algorithm Using Matlab. Journal
of Engineering and Applied Sciences.
[11] Liceaga-Castro, J.U.; Siller-Alcalá, I.I.; Alcántara-Ramírez, R.A. Least Square Identifica-
tion using Noise Reduction Disturbance Observer. Trans. Syst. 2019, 18, 313–318
[12] Wang, F.; Wang, R.; Liu, E.; Zhang, W. Stabilization Control Mothed for Two-Axis Iner-
tially Stabilized Platform Based on Active Disturbance Rejection Control with Noise Re-
duction Disturbance Observer. IEEE Access 2019, 7, 99521–99529.
[13] He, T.; Wu, Z. Extended Disturbance Observer with Measurement Noise Reduction for
Spacecraft Attitude Stabilization. IEEE Access 2019, 7, 6137–66147.
[14] She, J.H.; Fang, M.X.; Ohyama, Y.; Hashimoto, H.; Wu, M. Improving disturbance-rejec-
tion performance based on an equivalent input-disturbance approach. IEEE Trans. Ind. Elec-
tron. 2008, 1, 380–389.
[15] She, J.H.; Xin, X.; Pan, Y.D. Equivalent-input-disturbance approach—Analysis and appli-
cation to disturbance rejection in dual-stage feed drive control system. IEEE/ASME Trans.
Mechatronics 2011, 16, 330–340.
[16] Vazquez-Sanchez, E.; Gomez-Gil, J.; Gamazo-Real, J.C. A New Method for Sensorless Es-
timation of the Speed and Position in Brushed DC Motors Using Support Vector Machines.
IEEE Trans. Ind. Electron. 2012, 59, 1397–1408.
[17] Radcliffe, P.; Kumar, D. Sensorless speed measurement for brushed DC motors. IET Power
Electron. 2015, 8, 2223–2228.
[18] Vazquez-Sanchez, E.; Sottile, J.; Gomez-Gil, J. A Novel Method for Sensor less Speed De-
tection of Brushed DC Motors. Appl. Sci. 2016, 7, 14.
[19] Vidlak, M.; Makys, P.; Stano, M. Comparison between model based and non-model based
sensorless methods of brushed DC motor. Transp. Res. Procedia 2021, 55, 911–918.
[20] Gharib, M.R. Comparison of robust optimal QFT controller with TFC and MFC controller
in a multi-input multi-output system. Rep. Mech. Eng. 2020, 1, 151–161.
[21] Bozanic, D.; Teši´c, D.; Marinkovi´c, D.; Mili´c, A. Modeling of neuro-fuzzy system as a
support in decision-making processes. Rep. Mech. Eng. 2021, 2, 222–234.
[22] Obaid, B.A.; Saleh, A.L.; Kadhim, A.K. Resolving of optimal fractional PID controller for
DC motor drive based on anti-windup by invasive weed optimization technique. Indones. J.
Electr. Eng. Comput. Sci. 2019, 15, 95–103.
[23] Wang, Q.; He, F. The Synchronous Control of Multi-Motor Drive Control System with
Floating Compensation. In Proceedings of the 2016 12th World Congress on Intelligent
Control and Automation (WCICA), Guilin, China, 12–15 June 2016; pp. 1940–1954.
[24] El Ouanjli, N.; Taoussi, M.; Derouich, A.; Chebabhi, A.; El Ghzizal, A.; Bossoufi, B. High
Performance Direct Torque Control of Doubly Fed Induction Motor using Fuzzy Logic.
Gazi Univ. J. Sci. 2018, 31, 532–542.
[25] Muhammad Rafay Khan, A. A. (2015). Speed Control of a DC Motor under varying load
using GRNN algorithm.
[26] Muniracad. (2019, January 26th). Speed control of DC Motor. Retrieved from Munir Acad-
emy: https://muniracademy.com/speed-control-of-dc-motors/
[27] Mutalid, M. A. (2008). Speed control of DC Motor uisng PI controller.
[28] Mwahib Mohamed Harron Elhag, R. M. (2016). Speed control of a DC motor using fuzzy
logic and PI controller. SUDAN UNIVERSITY OF SCIENCE & TECHNOLOGY.
[29] ] Patil, V. S., Angadi, S., & Raju, A. B. (2016). Four Quadrant Close Loop Speed Control
of DC Motor.
[30] Prof. N. D. Mehta, P. A. (2017). Modeling and simulation of P, PI and GRNN algorithm for
speed control of DC Motor Drive.

110
Wasit Journal of Computer and Mathematic Science Vol. (2) No. (1) (2023 )

[31] Rao, V. M. (2013). Performance Analysis of Speed Control of DC Motor Using P, PI, PD
and GRNN Algorithm.
[32] S.K. Singla, R. K. (2013). Comparison among some well-known control schemes.
[33] Sabir, M. M., & Khan, J. A. (2014). Optimal Design of GRNN algorithm for the speed con-
trol of DC motor by using Metaheuristic Techniques.
[34] Senawi, A., Yahya, N. M., & Yusoff, W. A. (2010). Tuning of Optimum GRNN algorithm
Parameter Using Particle.
[35] Walaa M Elsrogy. Naglaa K Bahgaat, M. E.-S. (2018). speed contol of a DC motor using
GRNN algorithm based on changed intelligent techniques.
[36] Xia, C.-l. (2012). Permanent Magnet Brushless DC Motor drives and controls. Singapore:
John Wiley & Sons.
[37] Zalm, G. v. (2004). TUNING OF GRNN-TYPE CONTROLLERS: LITERATURE
OVERVIEW.
[38] Baliga BJ, Adler MS, Love PR, Gray PV, Zommer ND, the insulated gate transistor-a new
three terminal MOS controlled bipolar power device. IEEE transactions Electron develop-
ment, 1984: ED-31: 821-8.
[39] Nisreen Khalil Abed, Faisal.T.A., Hamdalla F. Al-Yasriy, Detection of power transmission
lines faults based on voltages and currents values using K-nearest neighbors. International
Journal of Power Electronics and Drive Systems (IJPEDS), 2023. 14(02): p. 1033-1043.
[40] Roa'a, M.A.a., et al., Generation of High Dynamic Range for Enhancing the Panorama En-
vironment. Bulletin of Electrical Engineering, 2021. 10(1).
[41] Meteab, W.K., et al. Controlling and Monitoring a Robot-Car Based on Smart Phone Appli-
cations. in IOP Conference Series: Materials Science and Engineering. 2021. IOP Publish-
ing.
[42] Abd Ali, D.M., D.F. Chalob, and A.B. Khudhair, Networks Data Transfer Classification
Based On Neural Networks. Wasit Journal of Computer and Mathematics Sciences, 2022.
1(4): p. 207-225.
[43] Mohamad, A.S., Machine Learning Empowered Software Prediction System. Wasit Journal
of Computer and Mathematics Sciences, 2022. 1(3): p. 52-61.
[44] Hussien, N.A., et al., Monitoring the Consumption of Electrical Energy Based on the Inter-
net of Things Applications. 2021. 15(7).
[45] Azeez, R.A., et al., Design a system for an approved video copyright over cloud based on
biometric iris and random walk generator using watermark technique. Periodicals of Engi-
neering Natural Sciences, 2021. 10(1): p. 178-187.
[46] Abed, F.T., and I.A. Ibrahim. Efficient Energy of Smart Grid Education Models for Modern
Electric Power System Engineering in Iraq. in IOP Conference Series: Materials Science
and Engineering. 2020. IOP Publishing.
[47] Najeeb, S.M.M., and S.M. Ali, Finding the discriminative frequencies of motor electroen-
cephalography signal using genetic algorithm. Telkomnika (Telecommunication Computing
Electronics and Control), 2021. 19(1): p. 285-292.
[48] Nabaa Ali Jasim, H.T.S.alim., Design and Implementation of Smart City Applications Based
on the Internet of Things. International Journal of Interactive Mobile Technologies (iJIM),
2021. 15(13): p. 4-15
[49] Shrivastava, S., Rawat, J.&Agrawal, A. (2012). Controlling DC Motor using Microcontrol-
ler (PIC16F72) with PWM.International Journal of Engineering Research, 1(2), 45-47.

Article submitted 16 March 2023. Published as resubmitted by the authors 31 March 2023.

111