Journal of Theoretical and Applied Information Technology

15th April 2022. Vol.100. No 7

© 2022 Little Lion Scientific

ISSN: 1992-8645 www.jatit.org E-ISSN: 1817-3195

COMPUTATIONAL AND INTELLIGENT OPTIMIZATION

TUNING METHOD FOR PID CONTROLLER
ROSALIA H. SUBRATA1, FERRIANTO GOZALI2, ENDANG DJUANA3
1,2,3
Electrical Engineering Department, Faculty of Industrial Technology, Universitas Trisakti, Jakarta,
Indonesia
E-mail: rosalia@trisakti.ac.id, ferrianto@trisakti.ac.id, 3edjuana@trisakti.ac.id
1 2

ABSTRACT

Proportional Integral Derivative (PID) control systems are quite common in the commercial, institutional,
industrial plants and many real-life applications due to their advantages over other kind of controllers. Proper
application and tuning of PID parameters can provide many benefits and increase process efficiency.
However, improper implementation, due to lack of understanding and poor tuning processes, are often be the
main cause behind many problems. Conventional PID tuning methods can’t meet the demands as the system
becomes more and more complex due to trial and error approach, therefore a more systematic and automated
approach is needed. This paper discusses the background and theory behind PID controller and bring a clear
understanding how to perform tuning process for PID controllers and presents computational and intelligent
optimization techniques such as Genetic Algorithms, Particle Swarm Optimization and Differential Evolution
used to make the control deviation of step response close to zero, have faster response and smaller or even
without overshoot. Through literature survey from existing publications, several issues related to the
implementation of the PID controller to enhance the performance of such as model plant development,
parameter setting, control characteristics expected and tuning processes are discussed and summarized. The
conclusion is that computational and intelligent optimization tuning method produce less overshoot, shorter
rise and settling time and less errors in system responses, therefore producing more optimal tuning and at the
end more optimal system performance.
Keywords: PID Parameter, Computational Method, Tuning Optimization.

1. INTRODUCTION We can say that process control deals with the

process of maintaining the output of a specific
After the discovery of PID control by Elmer process in industrial plant within a desired range to
Sperry in 1910 [1], and tuning method of PID maintain the quality of an end product. In most of
proposed by J. G. Ziegler – N. B. Nichols (Z-N) in industrial process, it is very important to maintain the
1942 [2], the PID control is growing rapidly. Now, it control of the performance in order to keep the
is common to find many PID controllers or modified operations running under certain predefined
PID controllers used as industrial controls. Many condition and to set more precise limits to maximize
tuning methods have been suggested in theory and the quality of the product and safety.
implemented on-site manually or automatically. However, usually it is difficult to achieve desired
However, most of industrial controllers are still performance for time-variant system. Therefore, it is
implemented based on and around classical PID needed that control parameters of the process are
algorithms [1]. adjusted automatically by the controller itself and
In the field of process control systems, plant’s become adaptive to the behavior of the process. In
mathematical models are very essential for steady state, control parameters are not needed to be
analytical design methods. But when the analytical adjusted in every step of the process but only when
design methods cannot be used, PID control proves control parameters become out of the range defined
useful. The basic PID controller or a modified PID and the performance is insufficient [4]. The
controller proved useful in providing satisfactory adaptivity of the controllers depend on how the
control, but it is not possible to provide optimal controllers utilize the performance evaluation of the
control for all conditions [3]. process and adjust the control design system.

2007
 Journal of Theoretical and Applied Information Technology
15th April 2022. Vol.100. No 7
© 2022 Little Lion Scientific

ISSN: 1992-8645 www.jatit.org E-ISSN: 1817-3195

As PID controllers work based on combination of controller usually written in "parallel form" or in
suitable P, I and D parameters for each system in “ideal form” can be seen in eq. (1) and eq. (2) [1]:
order to produce optimal system performance, these = + + (1)
parameters (P, I and D) need to be tuned, a process
known as PID tuning. PID tuning means that the = 1+ + (2)
control loop has a specific goal which is achieved by
using the right P, I, and D parameters. PID tuning is
necessary in order to have the desired closed-loop
control. PID tuning can result in many benefits such
as the process can operate in a stable way, reduce Figure 1: PID Controller of a Plant
oscillations, reducing energy consumptions and
stability of the system can be achieved with a Output of the PID controller:
minimum amount of operator interventions.
= + + (3)
Nevertheless, conventional PID parameters tuning with , , positive.
methods have limitation in that they may produce
suboptimal performance, for which further = and = .
adjustment of PID parameters is needed. With the Eq (3) can be written as follows [3]:
advancement of computational methods, many = + + (4)
optimization techniques in engineering are used in
PID tuning to improve the capabilities of where = output controller; = error signal;
conventional PID parameter tuning methods. = proportional gain; = integral gain; =
Various meta-heuristic methods used in Artificial derivative gain; = integral time; = derivative
Intelligence such as Genetic Algorithm (GA), time.
Differential Evolution (DE), and Particle Swarm The Proportional also called P-controller
Optimization (PSO) have been used for determining will produce an output proportional to the error
the optimal PID tuning. These methods can produce signal . This controller always produces a
accurate PID parameters therefore producing system steady-state errors but it can provide stable
optimal system performance. operation. By improving the proportional constant
, the speed of the system response will improve.
In this paper a number of methods used for PID
The P-controller needs a manual reset if used
tuning are reviewed. We divided the tuning methods
independently because it never gets to a system
into two categories: Conventional Tuning Methods
steady state. However, if the proportional constant
and Computational and Intelligent Optimization
is too large, the system will take a long time to
Methods. A brief explanation of each method is
described then followed by examples of a case study reach steady-state condition. The Proportional
of the tunings that have been carried out previously, controller equation is formulated as [3]:
and presents the simulation results of the = (5)
implementation of PID tunings on the control system The Integral also called I-controller is needed to
process using these methods. Several examples of eliminate steady-state error and this controller can
case studies carry out comparative analysis, and integrate the error over a period of time so that the
explain the comparison results. error value becomes zero. The integral controller
equation is formulated as [3]:
2. PID CONTROLLER = (6)
Proportional Integral and Derivative or If a negative error occurs, the controller output will
abbreviated PID, and it is one kind of approach used decrease, meaning the response speed is reduced and
to control different process variables in industrial the system stability is disturbed. To increase the
applications. The PID controller block diagram of a response speed, the integral gain, is reduced.
plant is shown in Figure 1. These control elements The Derivatives also called D-controller will give a
can perform individually or collectively, for kick start at the output so as to improve system
example, a P-controller, a PI controller, or a PID response. The output produced by this controller
controller. Each control parameter can be adjusted depends on the rate of change of the error with
and controlled individually and each control element respect to time. The D-controller can anticipate
can be used for specific purposes depending on the
future error behavior unlike the I-controller. To
application. The transfer function of a standard PID
increase the speed of system response by increasing

2008
 Journal of Theoretical and Applied Information Technology
15th April 2022. Vol.100. No 7
© 2022 Little Lion Scientific

ISSN: 1992-8645 www.jatit.org E-ISSN: 1817-3195

the derivative gain . The derivative control desired, such as too large an overshoot in the step
equation is formulated as follows [3]: response, a fine tuning must be carried out until the
desired condition is reached [2]. In general, the
= (7) Ziegler Nichols tuning rule produces the value of the
The effect of each controller parameter P, I, D integral time, proportional gain and derivative time
independently on closed-loop performance is parameters based on the transient response
described in Table 1 while for open-loop characteristics of the plant as a starting point for fine
performance this table is useful as a first reference tuning [3].
for stability. However, , (or ) and (or )
3. PID TUNING METHOD
are mutually dependent in tuning for achieving
optimum performance [1].
As stated before, PID tuning relates to
Table 1: Effects of parameter P, I, and D processes of determining PID parameters based on
independently the method used to produce the engineering
specifications and desired closed-loop performance
characteristics. The best plant optimization can be
accomplished if the closed-loop control able to
control the process stability, reduce plant
oscillations, efficiency and minimize energy
consumption.
So finally, by combining these three parameters, the A number of conventional tuning methods
desired response for the system can be achieved. are done manually, difficult and take a lot of time. In
order to get the ideal PID parameters efficiently, it is
2.1 PID Controller Design necessary to tune the optimal parameters. There are
What needs to be done in designing a PID many methods of PID tuning. In this paper, PID
controller in a plant is first to derive the plant’s tuning methods are divided into two categories: 1).
mathematical model. In reality, it is not easy to Conventional Tuning Methods and 2).
derive the mathematical model of a complicated Computational and Intelligent Optimization
plant. In this situation, an experimental approach is Methods.
performed to improve the controller design. Then
select and apply various design techniques so that 3.1 Conventional Tuning Methods
the system achieves the desired criteria. To produce The steps to determine the PID controller
the desired transient and steady state characteristics, parameters with conventional methods, first make
needs to determine the appropriate control assumptions about the desired plant output, conduct
parameters. The process of setting the controller an analytical or graphical analysis, then determine
parameters in order to produce the desired system some features of the process. Tuning the controller
criteria is called controller tuning. parameters by this method in most cases does not
PID controllers are most often used by give perfect results and further tuning is required
practitioners in designing control systems because because it is based on the assumptions made.
the PID technique is quite mature and well Conventional PID tuning methods are briefly
understood. PID parameter tuning is very important discussed in the following sections.
to obtain the desired system characteristics. By 3.1.1 Ziegler-Nichols Method
tuning the PID parameters correctly, it can improve The Ziegler–Nichols tuning formulas were
slow system response, reduce large oscillations, developed empirically based on simulations of many
reduce steady state errors and improve system existing cases. The Ziegler–Nichols tuning method
stability. However, there is no single tuning strategy developed by John G. Ziegler and Nathaniel B.
that applies to all kinds of control systems, so no Nichols in 1942 is a heuristic tuning of a PID
single tuning method which able to tune all kinds of controller. To discover the parameters of the PID
loops optimally. controller, the following classical tuning methods
Ziegler and Nichols found a rule to adjust the are used: step response method and frequency
PID controller when the mathematical model cannot response method. There are two parameters in these
be known. The Ziegler-Nichols rule consists of a set methods process dynamics, namely the parameter
of values that will give stable system characteristics. which represent the process gain and the parameter
If the system provides conditions that are not as which represent how long the process is. These

2009
 Journal of Theoretical and Applied Information Technology
15th April 2022. Vol.100. No 7
© 2022 Little Lion Scientific

ISSN: 1992-8645 www.jatit.org E-ISSN: 1817-3195

parameters are expressed by simple formulas. The Nichols formulas for the PID controller parameters
Ziegler-Nichols tuning method is well implemented in terms of the and the [5]. = period of the
in systems with a small gain ratio and a smaller dead- dominant dynamics of the closed-loop system. This
time than the time constant. [5]. These methods are method is suitable only for closed loop PID
still widely used by controller manufacturers and controller systems. The use of this method aims for
process industry, both conventional methods and the system to achieve stability quickly. The
modified conventional methods. For some purposes, advantage of this Ziegler-Nichols method such as
the result is acceptable but not optimal for all and parameters are simpler to find accurately
applications. than the parameters and in the step response
In the step response method, parameters are method.
determined from the unit step response of the
process. To approximate the characteristics of the Table 3: The Ziegler-Nichols frequency response method
system, an open-loop test was carried out. First, PID controller parameters.
determine the point at which the slope of the step
response reaches its maximum, and then draw a
tangent to this point. The intersection of the tangent
and the coordinate axes gives the parameters and
(measured on the step response of the process), as
shown in Figure 2. The step response method can
only be used for stable processes. This method
provides PID parameters directly as function of The advantages of Ziegler-Nichols tuning method
and and the estimation of the period of the are its simplicity, intuitive, easy to use than other
closed-loop systems, as shown in Table 2 [5]. methods and requires only a few trials it is quite
simple to generate the appropriate controller
parameters while the disadvantages of this method
are the tuning result is sensitive to non-linearities
and approximations for the tuning parameters values
might not be entirely accurate for different systems.

3.1.2 Cohen Coon Method

The Cohen-Coon tuning method is derived
from the Ziegler-Nichols method, but in the process,
it uses more information from the system so as to
make the control performance much better. Cohen-
Coon tuning method is classified as an offline tuning
method. There are three parameters defined in this
Figure 2: Ziegler-Nichols step response method method, parameter is steady state gain, is time
Characterization delay, and is time constant.
Table 2: PID parameters for the Ziegler-Nichols step
Cohen-Coon tuning formulas as found in
response method Table 4 were obtained based on analytical and
numerical computational. The parameters = /
and & = / + are used in the table, =
apparent time constant, and & = relative dead time
[5]. The initial control parameters are measured
based on the response of the system with the time
constant and time delay as the output of a step
change at steady state. The main design criterion in
this method is rejection of load disturbances thus
The parameters in the frequency response method minimizes the steady state error [5].
named the ultimate gain ( ) and the ultimate period The advantages of Cohen-Coon tuning
( ). To determine the PID controller parameters, method are reducing errors in approximating the
first adjust the parameters until the control action in steady state gain and the time delay by using the
proportional = ∞ dan = 0, then improve the gain, good enough for systems with time delay, and
gain so the system starts to oscillate. The gain response to closed loop is faster than Ziegler-Nichols
obtained at this stage named and the period of the
oscillation is . Table 3 shows the simple Ziegler-

2010
 Journal of Theoretical and Applied Information Technology
15th April 2022. Vol.100. No 7
© 2022 Little Lion Scientific

ISSN: 1992-8645 www.jatit.org E-ISSN: 1817-3195

method. The disadvantage of Cohen-Coon tuning method for systems with high proportional gains and
method is that it cannot be used online. the rules providing unsatisfactory performance for
systems with long normalized dead time. Kappa-tau
Table 4: Cohen-Coon Method Controller tuning method is based on the parameters and &
Parameters. and designed for load disturbance response. The
method offers opportunity to distinguish between
setpoint and disturbance response. The advantages
of Kappa-tau tuning method are less oscillatory
response and optimal for disturbance rejection with
no overshoot while the disadvantages are it cannot
achieve closed-loop performance characteristics in
consequence of 0% overshoot.

3.1.5 Lambda Method

The Lambda tuning method was developed
3.1.3 Chien, Hrones and Reswick (CHR) Method for systems with a long dead time L and includes an
The Ziegler-Nichols step response method analytical tuning method of PID parameter. The time
was modified by Chien, Hrones and Reswick to constant of the closed-loop response is expressed by
make better damped closed-loop systems. The the λ parameter. The response is assumed to follow
Chien-Hrones-Reswick (CHR) autotuning method a first order with L and λ as a tuning parameter.
focuses on setpoint response and disturbance Lambda tuning is applicable only for PI controller,
response, the controller parameters formulas are parameters is given in Table 7 [5].
given in Table 5 and Table 6 [5]. This method
provides overshoot formulas for 0% and 20%. Table 7: PI controller parameters of Lambda Tuning
Determination of parameters a and L can be seen in Method.
sub chapter 3.1.1 Figure 2 of this article.

Table 5: Controller parameters of the CHR load

disturbance response method.

The advantages of Lambda tuning method are the

possibility of choosing how fast the response of the
controller, with a wide time delay and response with
no overshoot while the disadvantages are minimal
for disturbance rejection.

Parameters tuning formula for the 20% overshoot in 3.2 Computational and Intelligent Optimization
Table 5 are similar to the Ziegler-Nichols step Methods
response method as found in Table 2. However, for With the advancement of computational
the 0% overshoot, the gain and the derivative time methods, many optimization techniques in
are smaller and the integral time is larger. It can be engineering are used in PID tuning to improve the
said that the performance of proportional, integral capabilities of conventional PID parameter tuning
and derivative, are smaller [5]. methods. The main purpose is to make the control
deviation of step response close to zero, have faster
Table 6: Controller parameters of the CHR setpoint
response and smaller or even without overshoot.
response method.
Various meta-heuristic methods used in artificial
Intelligence such as Genetic Algorithm (GA),
Differential Evolution (DE), and Particle Swarm
Optimization (PSO) have been used for determining
the optimal PID tuning of various systems.
GA/DE/PSO is used to obtain PID parameters and
an output signal in the context of system
3.1.4 Kappa-tau (KT) Method requirements [6]. A general schematic of the system
Kappa-tau (KT) tuning method was can be seen in Figure 3.
designed to fill the gap of the Ziegler-Nichols

2011
 Journal of Theoretical and Applied Information Technology
15th April 2022. Vol.100. No 7
© 2022 Little Lion Scientific

ISSN: 1992-8645 www.jatit.org E-ISSN: 1817-3195

chromosome comprises of three PID gains

( , , with value bounds varied depend on the
objective functions used and written as binary
strings. The first and the most crucial step is to
encoding the problem into suitable GA
chromosomes. The more the chromosomes number,
Figure 3: General schematic of the system the better the chance to get the optimal results.
2. Initialization: Create generation of the initial
3.2.1 Genetic Algorithm (GA) population and fix in an explicit array using the
The Genetic Algorithm’s concept was first Ziegler-Nichols method [9] to ensure the system
found by J.H. Holland in 1970 and relies on natural stability and the convergence.
evolution principles [7]. A genetic algorithm (GA)
is a stochastic method for optimization problems 3. Fitness Value Calculation
based on a natural selection process that mimics There are two issues related to this step, the objective
biological evolution. Three basic operations or function and the fitness function. The objective
genetic operation in generic algorithm are 1) function is useful for ensure the validity of the fitness
Selection and Reproduction, 2) Crossover and 3) function in each chromosome. The function is set
Mutation. The application of these operations allows according to the system requirements to get the
the creation of new chromosomes which represents optimal value. Several objective functions can be
a solution of the problem and may be better than their
seen below:
parents.
The GA procedure starts by selecting the a. Integral Absolute Magnitude of the Error:
,
right number of chromosomes from the initial ()* = | | (8)
population where each chromosome represents an b. Integral of the Squared Error:
, .
adequate solution and is tested with a fitness (-* = (9)
function [7], [8]. The current population then
c. Integral of Time multiplied by Absolute Error:
undergoes an evolutionary process to produce a new ,
chromosome called a new generation. Chromosomes ( )* = | | (10)
in this new generation are better than the previous d. Integral of Time multiplied by the Squared Error:
, .
generation. This algorithm is repeated until it ( -* = (11)
produces an optimal solution [9], [10]. The process
Fitness function: The implemented objective
flowchart of GA is presented in Figure 4 [9], [11].
function is evaluated to reduce errors based on
system requirements. The result of this evaluation is
defined as a fitness function.
4. Selection: Selection is used to obtain
chromosomes which will be selected for
crossbreeding and mutation. The higher the fitness
value of the chromosome, the greater the chance of
being selected. When the process is complete all the
chromosomes are established.
5. Crossover and Mutation: Crossover is the GA
operator that collects two-bit strings as parent values
to form a new chromosome of the highest quality.
This operation is not performed on all existing
chromosomes but chromosomes were randomized to
perform crosses. If the crossover is not performed,
then the parent value will be passed down to the
offspring. A considerable population implicates a
smaller crossover rate. Mutation is the next GA
Figure 4: GA process flowchart. operator of that modifies the string codification
(changes a few strings at random) of the selected
PID parameter tuning process steps using GA: chromosome and replaces the genes lost from the
1. Generate Population: Generate the initial population due to the selection process that allows
population of a few chromosomes at random. Each

2012
 Journal of Theoretical and Applied Information Technology
15th April 2022. Vol.100. No 7
© 2022 Little Lion Scientific

ISSN: 1992-8645 www.jatit.org E-ISSN: 1817-3195

the reappearance of genes that did not appear in the

initial population.

3.2.2 Particle Swarm Optimization (PSO)

The concept of Particle Swarm
Optimization (PSO) technique was introduced by
James Kennedy and Russell Eberhart in 1995 [12].
The PSO algorithm mimics the behavior of a herd
insects or birds to take a chance and gain advantage
while moving in certain space in finding food or
avoiding predators. Each particle with its own two
characteristics, namely its position and velocity, able
to remember the best position ever traversed in the
space. The interaction of particles in the PSO method
will produce the best particles in order to obtain the
optimal PID parameter specifications [13],[14].
There are 3 stages in the PSO algorithm:
The first stage is to generate particle’s positions and
velocities of the initial population randomly and they
move through the space at random speeds. The
second stage is to update speeds of the particles for
the next iteration, particles are able to remember Figure 5: General Particle Swarm Optimization process
every interaction, following their previous best flowchart
position and equivalent fitness. Velocity update is
based on the value of particle fitness which is not 3.2.3 Differential Evolution (DE)
only determine which particle has the best global Differential Evolution (DE), proposed by
value in current herd, but also determine the best Storn and Price [16] is a variant of computational
position of each particle over time. And finally, the algorithms used in optimization problems [17]. This
third stage is position update. A particle with the best algorithm is similar to GA and PSO, but Differential
fitness and unique for the entire space solution, is Evolution has some advantages such as easy and
called best global value of the swarm [12], [14], [15]. simple to use, the ability to locate the accurate global
The cost function applied to achieve the optimal PID optimum irrespective of the initial parameter values,
coefficient is ITAE as found in equation (10), it has rapid convergence and utilizes few control
because this function is useful for reducing errors parameters.
(positive and negative), and giving small overshoot Differential Evolution is a population-
percentages [12]. based heuristic search algorithm in which the
The process flowchart of Particle Swarm population of D-dimensional individuals is used to
Optimization is presented in Figure 5. optimize a problem. Each individual in the
population is a candidate solution to the problem and
is coded as a vector. The parameters of the
algorithms are the population in the generation G,
the dimension of the problem D and the population
size NP. DE is composed of four steps, which are
initialization, mutation, crossover and selection as
seen in Figure 6.

2013
 Journal of Theoretical and Applied Information Technology
15th April 2022. Vol.100. No 7
© 2022 Little Lion Scientific

ISSN: 1992-8645 www.jatit.org E-ISSN: 1817-3195

DE/current-to-pbest/1 and each strategy affects

population diversity differently which provides a
different rate of search convergency [22]. The
mutation factor F defines the search step of the
optimization. According to Storn and Price, F range
is in [0, 2] [16]. Generally, smaller values of F are
the better when the population is closer to the global
best value. In contrast, larger values of F are
preferred when the population is far from the global
best value [23].
Crossover: The Crossover Operation is performed to
increase the diversity of mutation vectors from
previous operation. In this operation, the crossover
procedure generates a trial vector between the
interaction of the mutation vector and the target
vector. There are some option popular crossover
strategies, such as classic binary crossover strategy,
exponential crossover strategy [24,25], eigenvector-
Figure 6: Flow graph of Differential Evolution based crossover strategy [26]. According to Storn
algorithm. and Price, the crossover rate CR range in [0,1]
reflects the probability in which the trial vector
Initialization: Three main parameters of the inherits mutation vector’s values [27] and has a
algorithm, population size (NP), mutation factor (F) direct influence on the diversity of the population
and crossover rate (CR) are defined in the [23].
initialization step [18]. The setting of these
Selection: The last process of the iteration is the
parameters influences the algorithm’s performance,
selection process. After the crossover operation, the
and the optimal setting depends on the problem to be
fitness value of the trial vector is calculated.
faced [19]. In general, these parameters are set after
Afterward, the trial vector and the target vector
the objective function is designed [20]. The DE
compete, and if trial vector has a better fitness value
optimization starts with a randomly generated
than target vector, then trial vector will replace target
population of NP individuals characterized by D
vector in the next generation.
parameters according to a uniform distribution
within the search space. Each individual represents a
candidate solution to the optimization problem.
4. RESULTS AND DISCUSSION
After initialization, the DE algorithm performs
mutation, crossover and finally selection operations
The implementation of many optimization
iteratively until the user-defined criteria are reached.
techniques to improve the capabilities of
In this iterative process, new individuals are
conventional PID parameter tuning methods such as
generated and evolved over generations. In this
GA, PSO and DE can be found in various
work, all generated individuals for the search space
applications. Through literature survey from existing
are feasible solutions since the problems are non-
publications, several issues related to the
constrained.
implementation of the controller is described as
Mutation: In each G-generation, the selected follows.
mutation strategy is used to create a mutation vector.
The popular mutation strategy is DE/rand/1. In this GA based PID controller tuning has been applied to
strategy, two vectors (individuals) are randomly Continuous Stirred Tank Reactor process with a
chosen, the sum of the weighted difference of two combination of ISE, IAE, and ITAE as the objective
individuals is multiplied by a mutation factor F, and function. The simulation results have concluded that
the result is added to a third random vector. The the optimal PID parameters obtained using GA have
difference vector automatically adapts to the scale of achieved satisfactory set point tracking and
the optimized function, and that is the key success disturbance rejection in the process [28].
factor of the DE algorithm [21]. There are also some GA based PID Controller tuning is implemented and
popular mutation strategies such as DE/rand/2, compared with the classical Ziegler-Nichols tuning
DE/best/1, DE/best/2, DE/current-to-best/1, method on a position control system of DC motor

2014
 Journal of Theoretical and Applied Information Technology
15th April 2022. Vol.100. No 7
© 2022 Little Lion Scientific

ISSN: 1992-8645 www.jatit.org E-ISSN: 1817-3195

order 3. The response produced by the PID controller provides a more fluctuating fitness value than DE.
with the GA method has a rise time and settling time Both of these methods succeeded in finding the
much faster than produced by the classical method optimal gain parameter. Overall, the ability and
and the error produced by the GA is much smaller effectiveness of the DE algorithm to achieve the best
than the error produced by the classical method [29]. PID gain configuration in maintaining the electrode
position is better than the PSO algorithm [34].
PSO-based Tuning of a PID Controller is
implemented for a High-Performance Drilling PID Controller using DE Method is implemented to
Machine and is compared with the classical Ziegler- determine the rotational speed of a DC motor with a
Nichols tuning method. The simulation results high stable output. The simulation results display the
conclude that tuning PID parameters based on the rise time, settling time, and overshoot values of the
PSO algorithm provides less overshoot, faster to DC motor and can be concluded that the PID tuning
reach settling time, rise time; peak time are faster; using the DE method produces an optimal and
and reduced errors [30]. effective system [35].
PID controller tuning using PSO method was also
5. CONCLUSION
implemented on four different systems, 1) Linear
hydraulic system to control the position and velocity
of fluid flow in hydraulic cylinders driven by 1. Even though conventional PID tuning have been
hydraulic pumps. 2) Linear electrical control with widely used in industrial fields because of their
disturbance (D.C motor). 3) Linear thermal control simple principles, easy implementation and wide
system with delay. 4) Non-Linear Permanent applications, most of the system exhibits a large
Magnet Synchronous Motor. The simulation results percent of overshoot.
showed that the PID controller tuned by PSO 2. Computational and Intelligent PID tuning such as
method, provides accurately the desired closed loop Genetic Algorithm, Particle Swarm Optimization
dynamics (overshoot, rise time, settling time, and and Differential Evolution which are based on
steady state error) in a very short time for a small Artificial Intelligence-based computational
number of iterations. Overall, the PSO method could methods, can produce optimal tuning of PID
be considered as an effective and reliable auto tuning parameters, therefore producing optimal system
process for PID controllers [31].
performance.
A comparative study of the performance of PID 3. These computational and intelligent PID tuning
controllers tuned with Genetic Algorithm, Particle methods exhibits less percent of overshoot, faster
Swarm Optimization and Hybrid Genetic Algorithm rise and settling time and less errors than
Particle Swarm Optimization applied to the conventional methods.
Automatic Voltage Regulator system. The result 4. For real-valued function optimization, due to the
showed that Genetic Algorithm has a lower fitness
encoding of floating-point parameters into bit
cost than the Particle Swarm Optimization method,
strings, Genetic Algorithms are less suitable but on
Particle Swarm Optimization method is faster than
the Genetic Algorithm, while the Hybrid Genetic the other hand, evolution strategies of Differential
Algorithm Particle Swarm Optimization method is Evolution are well suited for real-parameter
more stable and effective [32]. optimization.

A comparative study of the performance of PID REFERENCES

controllers tuned by GA and PSO for DC motor
position controllers concluded that GA method [1] Kiam Heong Ang, Gregory Chong, and Yun Li ,
provides a slow response for the system while PSO “PID Control System Analysis, Design, and
method provides a better system response for the Technology”, IEEE Transactions On Control
system with a lower percentage of overshoot. PSO Systems Technology, Vol. 13, No. 4, July 2005,
based PID controller provides an ideal performance pp. 559-576.
for position control of DC motor related to the
[2] J. G. Ziegler and N. B. Nichols, “Optimum
system specifications [33].
settings for automatic controllers,” Trans.
A study compares DE algorithm and PSO algorithm ASME, Vol. 64, 1942, pp. 759–768.
for optimizing PID parameter tuning for Electrical [3] Katsuhiko Ogata, “Modern Control Engineer-
Discharge Machining Systems. With IAE fitness ing”, Prentice Hall, 5th Edition, New Jersey
function, DE algorithm gives a lower fitness 2010.
function value than the PSO algorithm. While PSO

2015
 Journal of Theoretical and Applied Information Technology
15th April 2022. Vol.100. No 7
© 2022 Little Lion Scientific

ISSN: 1992-8645 www.jatit.org E-ISSN: 1817-3195

[4] Takuya Kinoshita, Toru Yamamoto, “Design [14] James Kennedy and Russell Eberhart, “Particle
of a Smart Adaptive Control System”, 9th swarm optimization,” Proceedings of
International Symposium on Advanced Control International Conference on Neural Networks
of Chemical Processes, IFAC Papers Online, (ICNN) 1995, Australia, Vol. 4, IEEE Xplore,
Vol. 48, No. 8, 2015, pp. 1309-1314. 2002, pp.1942-1948.
[5] Karl Johan Astrom, Tore Hagglund, “PID [15] M. M. Reza Alavi Milani, T. Cavdar, and V. F.
Controllers: Theory, Design and Tuning”, Aghjehkand, “Particle swarm optimization -
Instrument Society of America (ISA), 2nd based determination of ziegler-nichols
Edition, 1995. parameters for pid controller of brushless dc
[6] Kachitvichyanukul, “Comparison of Three motors,” 2012 International Symposium on
Evolutionary Algorithms: GA, PSO and DE, Innovations in Intelligent Systems and
Industrial Engineering & Management System, Applications, IEEE Xplore, 2012, pp. 1–5.
Vol. 11, No. 3, September 2012, pp. 215–223. [16] Kenneth V. Price, Rainer M. Storn, Jouni A.
[7] J. H. Holland, “Adaptation in natural and artificial Lampinen, “Differential Evolution: A Practical
systems: an introductory analysis with Approach to Global Optimization”, Springer-
applications to biology, control, and artificial Verlag Berlin Heidelberg, Germany, 2005.
intelligence”, MIT press, 1992. [17] Ruijun Dong, “Differential evolution versus
[8] E. Purwanto, E. Wahjono, I. Ferdiansyah, D. S. particle swarm optimization for PID controller
Yanaratri, L. P. S. Raharja, R. P. Eviningsih, design”, Proceedings Fifth International
and G. Basuki, “Implementation of genetic Conference on Natural Computation, ICNC’09,
algorithm for induction motor speed control IEEE Xplore, 2009, pp. 236–240.
based on vector control method”, 2019 [18] Lyuyang Tong, Minggang Dong, ChaoJing, “An
International Seminar on Research of improved multi-population ensemble
Information Technology and Intelligent Systems differential evolution”, Neurocomputing, 2018,
(ISRITI), IEEE Xplore, 2020, pp. 244–247. Elsevier, pp. 130–147.
[9] E. Flores-Moran, W. Yanez-Pazmino, and J. [19] A.W. Mohamed, A.K. Mohamed, “Adaptive
Barzola-Monteses, “Genetic Algorithm and guided differential evolution algorithm with
fuzzy self-tuning pid for dc motor position novel mutation for numerical optimization”,
controllers,” 19th International Carpathian International Journal of Machine Learning and
Control Conference (ICCC), IEEE, 2018, pp. Cybernetics, Springer, 2019, pp. 253–277.
162–168. [20] Tanabe, R.; Fukunaga, A.S. “Improving the
[10] D. K. Meena and S. Chahar, “Speed control of search performance of SHADE using linear
dc servo motor using genetic algorithm”, 2017 population size reduction”, Proceedings of the
International Conference on Information, 2014 IEEE Congress on Evolutionary
Communication, Instrumentation and Control Computation (CEC), Beijing, China, July 2014;
(ICICIC), IEEE, 2017, pp. 1–7. pp. 1658–1665.
[11] D. A. R. Wati and R. Hidayat, “Genetic [21] Vladimir Stanovov, Shakhnaz Akhmedova ,
algorithm-based pid parameters optimization Eugene Semenkin “Selective Pressure Strategy
for air heater temperature control,” in differential evolution: Exploitation
International Conference on Robotics, improvement in solving global optimization
Biomimetics, Intelligent Computational problems”, Swarm and Evolutionary
Systems, IEEE Xplore, 2013, pp. 30–34. Computation 2019, Vol. 50, Nov. 2019, pp. 1-
[12] R. Eberhart and J. Kennedy, “A new optimizer 41.
using particle swarm theory”, Proceedings of [22] Janez Brest, Mirjam Sepesy Maučec, Borko
the 1995 Sixth International Symposium on Bošković, “Single Objective Real-Parameter
Micro Machine and Human Science, IEEE Optimization: Algorithm jSO”, 2017 IEEE
Xplore, 2002, pp. 39–43. Congress on Evolutionary Computation (CEC),
[13] M. E. F. Moran and N. A. P. Viera, IEEE Xplore, Spain, 2017, pp. 1311-1318.
“Comparative study for dc motor position [23] Wu Deng, Junjie Xu, Yingjie Song, Huimin
controllers”, IEEE Second Ecuador Technical Zhao, “Differential evolution algorithm with
Chapters Meeting (ETCM), IEEE, 2017, pp. 1– wavelet basis function and optimal mutation
6. strategy for complex optimization problem”,
Applied Soft Computing Journal, Vol. 100,
2020, Elsevier, pp. 1-16 .

2016
 Journal of Theoretical and Applied Information Technology
15th April 2022. Vol.100. No 7
© 2022 Little Lion Scientific

ISSN: 1992-8645 www.jatit.org E-ISSN: 1817-3195

[24] Shi-Zheng Zhao, Ponnuthurai Nagaratnam [33] Eduardo Flores-Moran, Wendy Yanez-Pazmi-
Suganthan, “ Empirical investigations into the no, Luis Espın-Pazmino, Ivette Carrera-
exponential crossover of differential Manosalvas, and Julio Barzola-Monteses,
evolutions”, Swarm Evolutionary Computation, “Particle Swarm Optimization and Genetic
Vol. 9, 2013, Elsevier, pp. 27-36. Algorithm PID for DC motor position
[25] Xin Qiu, Kay Chen Tan, Jian-Xin Xu, controllers”, IEEE ANDESCON, Oct 2020,
“Multiple Exponential Recombination for IEEE Xplore, 6 pages.
Differential Evolution”, IEEE Transactions on [34] Trias Andromeda, Azli Yahya, Syahrullail
Cybernetics, Vol. 47, No. 4, April 2017, pp. Samion, Ameruddin Baharom and Nor Liyana
995–1006. Hashim , “Differential Evolution for
[26] Shu-Mei Guo, Chin-Chang Yang, “Enhancing Optimization of PID Gain in Electrical
Differential Evolution Utilizing Eigenvector- Discharge Machining Control System”, ICETI
Based Crossover Operator”, IEEE Transactions 2012, Transactions of the Canadian Society for
on Evolutionary Computing, Vol. 19, 2015, pp. Mechanical Engineering, Vol. 37, No. 3, 2013,
31–49. pp. 293-301.
[27] R. Storn, “On the usage of differential evolution [35] Siti Fatimah Anggraini, Alfian Ma’arif , Riky
for function optimization”, Proceedings of the Dwi Puriyanto, “PID Controller on DC Motor
North American Fuzzy Information Processing, and Tuning Using the Differential Evolution
Berkeley, CA, USA, 19–22 June 1996, pp. 519– Method”, TELKA, Vol.6, No.2, November
523. 2020, pp. 147-159.
[28] A. Jayachitra and R. Vinodha, “Genetic
Algorithm Based PID Controller Tuning
Approach for Continuous Stirred Tank
Reactor”, Hindawi Publishing Corporation
Advances in Artificial Intelligence, 2014, pp. 1-
8.
[29] Neenu Thomas, P. Poongodi, “Position Control
of DC Motor Using Genetic Algorithm Based
PID Controller,” Proceedings of the World
Congress on Engineering, London, U.K., Vol. 2
, July 2009, pp. 1618-1622.
[30] S. M. G. Kumar, J. Deepak and R. K. Anoop,
“PSO based Tuning of a PID Controller for a
High Performance Drilling Machine”,
International Journal of Computer
Applications, Vol. 1, No. 19, 2010, pp. 12-18.
[31] M.H.T. Omar, W.M. Ali, M.Z. Mostafa, “Auto
Tuning of PID Controller Using Swarm
Intelligence”, International Review of
Automatic Control, Vol. 4, No. 3, Praise Worthy
Prize S.r.l., 2011, pp. 319-327.
[32] Faouzi Aboura, “Tuning PID Controller Using
Hybrid Genetic Algorithm Particle Swarm
Optimization Method for AVR System”, 2019
International Aegean Conference on Electrical
Machines and Power Electronics (ACEMP) &
2019 International Conference on Optimization
of Electrical and Electronic Equipment
(OPTIM), Istambul Turkey, IEEE Xplore, 2020,
pp. 570 – 574.

2017