PID Tuning Control Strategy using Cuckoo Search Algorithm for Pressure Plant

K.S.M. Jagindar Singh1, I. Elamvazuthi1*, Shaari, K.Z.K.2 Fernando V. Lima3

1 3
Department Electrical and Electronic Engineering Department of Chemical Engineering,
2
Department of Chemical Engineering West Virginia University,
Universiti Teknologi PETRONAS Morgantown, WV 26506, USA
32610, Bandar Seri Iskandar, Perak, Malaysia
1*
E-mail: [email protected]

Abstract—Plant process is a form of a process which requires and the system response was recorded. The research has shown
a robust controller to regulate the process. Therefore, tuning of the that the PSO-PID had improved rise time (0.297s) and settling
controller for the specific process is necessary and each controller time (4.03s) as compared to the rise time (0.494s) and settling
is process specific. This paper focuses on using the biomimetic time (11.1s) of conventional PID but a higher overshoot was
algorithm due to its’ competency to improve the system
observed. A liquid flow process was studied in [3]and the PSO-
performance in terms of rise time, settling time and overshoot to
regulate a closed loop pressure system. These biomimetic PID had provided better results than the IMC based controller.
algorithms excel in optimization problems. Algorithms that will be The process was modelled using First Order plus Dead Time
investigated are the Ziegler-Nichols (ZN) and Cuckoo Search (CS) (FOPDT) method from the Process Reaction Curve (PRC). The
in tuning the Proportional-Integral-Derivative (PID) controller improvement of the rise time and settling time is 55% and 66%
for a closed loop pressure plant system. respectively. Although the improvements are good, the
overshoot is approximately 25%. . Similar observation was also
Keywords— Pressure Plant; modelling; PID controller; First reported in [4] whereby the aim of the study is to design optimal
Order Plus Dead Time ; metaheuristic algorithms controller for room temperature. The mathematical model of the
I. INTRODUCTION air-conditioning system and room have been formulated for
simulation purposes. PSO-PID significantly reduced the
In literature, there has been various study conducted on the maximum overshoot, settling time and rise time for an air-
various types of process control loops. The type of control conditioning system. However, this has only been done in
which have been studied in literature are temperature, flow, simulation without real time application.
pressure and level.
The author in [5]had used genetic algorithm(GA) together with
Biomimetic control algorithm has been used to enhance the fuzzy PID and obtained good results. The aim is to save energy
conventional PID controller. Throughout the years, these in the production of lime. The GA is used to optimize the fuzzy
controllers have been tested in various applications by controller. The fuzzy is designed to control the air-fuel ratio.
researches. The nature of these algorithms which are universal Fitness function used for the GA is Integral of time absolute
make it possible to solve almost any optimization problem. In error (ITAE). The difference is that the GA is about obtaining
the case of the PID controller, it is aimed at reducing the the maximum so the ITAE function is slightly modified.
controller error which can be represented in terms of control However, there was no real time implementation as the
performance index (CPI)[1]. The algorithm will then work experiment was run on MATLAB. The water level control in
towards minimizing these CPI’s. Example of CPI’s that have [6] which uses GA has also proven to improve the system. It
been used are the Integral of Absolute Error (IAE), Integral of also has better response to disturbance and better dynamic
Time Square Error (ITSE), Integral of Time Absolute Error performances with smaller overshoot, less oscillation times and
(ITAE) and Integral of Square Error (ISE). These CPI’s are shorter adjustment time. The benefit of optimal control
actually objective functions which will be minimized by the presented in this study is that the controller automatically
algorithms. However, there is no restriction on CPI equations updates the new PID parameters once disturbance is detected,
as some researches have formulated their own objective thus the name optimal control. The proposed method in [7] uses
functions they aim to minimize. The possibilities of these form GA as well. The aim is to obtain high quality product while
of system tuning is endless. minimizing production of undesired products. The fitness
function included the reference temperature, reactor
The work reported in[2] is aimed at obtaining a fast and stable temperature, estimated reactor temperature and coolant
system for the level control of a three tank system. The system temperature. The proposed control scheme by the author has
was mathematically modeled and then tuned with the PSO significantly improved the conventional GA-PID controller.
algorithm with respect to an objective function. The PID was The proposed GA-PID has significantly reduced the overshoot
tuned offline and the values were updated into the controller produced due to the choice of the proper fitness function added
before the process. A unit step input was fed into the system

Authorized licensed use limited to: Turkish Energy Nuclear and Mineral Research Agency. Downloaded on April 10,2025 at 09:15:20 UTC from IEEE Xplore. Restrictions apply.
 into the algorithm. This signifies the importance of the fitness species of the cuckoo lay their eggs in a communal nest. They
function in obtaining a good system response. also may remove the host bird eggs to increase the hatching
probability of their own eggs. This behavior is also observed
The work presented by author in [8] proposed the use of the Bat when the cuckoo eggs hatch. They usually hatch before the host
Algorithm to perform offline tuning of the servomotor for bird eggs hatch. Once the first cuckoo is hatched, it will blindly
position/shaft control. The motor model was derived and tuned propel the eggs out of the nest which will increase the cuckoo
in the MATLAB environment using Ziegler-Nichols, PSO and chick’s share of food which is provided by the host bird.
BA. Three case studies were conducted using three different
plant transfer functions. The Bat Algorithm outperformed the . Three rules are followed before implementing the cuckoo
PSO for the first transfer function in terms of rise time by 2.9% search as follow
and settling time by 30.5%. However, for the second transfer 1) Every cuckoo lays one egg one given time, and dump
function, the overshoot for Bat Algorithm is 13.7% higher than its egg in randomly chosen nest
the PSO. It still however improved the rise time and settling
2) The best nest with high quality of eggs will carry over
time by 15.97% and 92% respectively. to the next generations
Several other applications that involve other algorithms have 3) The number of available host nests is fixed, and the
also been reported in [9-16]This paper discusses the outcome of egg that is laid by the cuckoo is discovered by the host bird
the use of Cuckoo Search (CS) algorithm in tuning the with a probability between 0 and 1.
Proportional-Integral-Derivative (PID) controller for a closed
loop pressure plant system. The Levy flight is performed when a cuckoo is generating new
solutions in the algorithm. Various studies have shown that the
flight behavior of many insects and animals have demonstrated
II. ALGORITHMS the characteristics of Levy flight.

A. Ziegler Nichols x t x(t)  Levy() (1)

The ZN tuning method is a heuristic method which has a set of
rules to follow to tune the PID controller. Over the years, In most cases,  is equal to 1 which is the step size of the
various ZN tuning methods have been developed. The tuning
particle. x represents the solution. means the entry-wise
algorithm for this experiment is the Classical Tuning
product of the two vectors. The Levy flight provides a
Method[17].
random walk and the random step length is obtained from a
Levy distribution
The Integral (KI) and Derivate (KD) gains are set to zero at first.
Levy( ) ~ t  (1    3) (2)

The Proportional gain (KP ) is increased until the output of the control loop
has reached constant oscillation.
which is a continuous probability distribution for non-negative
random variables.

The ultimate gain (KU) and oscillation period (Tu) are then used to
determine the Proportional (KP), Integral (KI) and Derivative (KD) gains.

Figure 1. Ziegler Nichols Tuning Steps

Fig. 1 show the tuning steps for ZN. The tuning for the ZN has
three steps. The two values, KP and Tu are then used in Table 1
for obtaining the final parameter values for the PID controller.

Table 1. Ziegler Nichols Parameter Calculation

Control Mode KP TI TD
PID 0.60 KU TU /2 TU /8

Table 1 shows the calculation to obtain the parameter values for

the PID controller.

B. Cuckoo Search
The CS is a metaheuristic algorithm which is inspired by the
parasitic behavior of some cuckoo species[18][19]. Some

Authorized licensed use limited to: Turkish Energy Nuclear and Mineral Research Agency. Downloaded on April 10,2025 at 09:15:20 UTC from IEEE Xplore. Restrictions apply.
 This experiment consists of three tanks and two controllers. The
buffer tank VE210, buffer tank VE220 and cooling vessel
VE240. The buffer tanks are to supply air to the cooling vessel
VE240. The aim in this experiment is to control the pressure in
the cooling vessel VE240. The PID controller values for ZN-
PID and CS-PID were calculated in MATLAB and put into the
controller in the computer. The ZN-PID parameters were
optimized using the toolbox in MATLAB. The parameters for
the CS-PID were calculated by using the program in MATLAB
as well with initialization as in Table I.
IV. MODELLING
Modelling was implemented in this research to obtain the
characteristics of the

1.829e7.2 s
G plant ( s)  (3)
12.71s  1
Eq. 3 is the transfer function of the pressure plant obtained
through the PRC for a First Order Plus Dead Time (FOPDT)
model[12].

The controller was developed and tested in MATLAB. The ZN-

PID controller was designed using a MATLAB toolbox using
the step response tuning. The approach for the CS-PID was
different. Fig.4 shows the CS-PID design steps.
Figure 2. Flowchart of Cuckoo Search
Fig.2 shows the flowchart of the Cuckoo Search Algorithm. It
contains 4 actions and three decision making blocks.
III. EXPERIMENTAL SETUP
The control loop has be investigated in this paper is the PID
pressure control loop. The plant was modelled using the Algorithm Coding Computation parameter
update
Process Reaction Curve (PRC) obtained after providing the
system with a step input. The values that were obtained after
calculation were then inserted into the controller plate via
software. Figure 4. Controller Design Overview
The Cuckoo Search algorithm is coded in MATLAB from the
pseudocode in Fig.2. The algorithm is then run with respect to
the transfer function obtained from the PRC.

 ZN-PID
Matlab toolbox was used to obtain the PID parameter values for
the Ziegler-Nichols tuning. The controller transfer function is
represented in Eq.4.

0.079
G( s) ZN  PID  1.13   4.08s (4)
s

Figure 3. Setup for System Identification  CS-PID

Fig.3 shows the graphical user interface of the process plant. It The CS-PID controller was designed using the following
is following the Distributed Control System (DCS) architecture. parameters listed in Table 1. The parameter initialization is
The graph obtained here is used to obtain the transfer function required for the algorithm to run.
of the process.

Authorized licensed use limited to: Turkish Energy Nuclear and Mineral Research Agency. Downloaded on April 10,2025 at 09:15:20 UTC from IEEE Xplore. Restrictions apply.
 Table 2. Parameter initialization for CS ZN-PID has a faster rise time. The ZN-PID has a rise time of
Number of Nests 100 3.78 seconds faster as compared to the CS-PID. The CS-PID
Discovery Rate 0.25 settling time is 8.1 seconds faster than ZN-PID. The overshoot
Iterations 100 of the CS-PID is 28.8% lower than the ZN-PID. Therefore, the
CS-PID does provide a better step response as compared to ZN-
The controller equation for CS-PID is represented in Eq.5. PID.

0.0570 B. Experimental
G( s)CS  PID  0.9255   0.9940s (6)
s The PID controller values obtained in MATLAB simulation
was used in the controller that regulates the pressure in VE240
V. RESULTS AND DISCUSSION
cooling vessel. A step disturbance was applied to the system
A. Simulation and the graph was obtained through the software integrated with
The transfer function in Eq.3 was tuned using CS and ZN. The the plant.
plant was then given a step input and the corresponding step
response of the controllers have been illustrated in Fig.5.

Figure 5. Step response of ZN-PID and CS-PID Figure 6. ZN-PID Step Response

Fig.5 shows that the CS-PID has better step response as Fig. 6 shows the plot of the Manipulated Variable (MV) in red
compared to the ZN-PID in terms of settling time and which is the valve opening in percentage, the Process Variable
overshoot. Table 3 shows the parameter values for the (PV) in green which is the pressure and Set Point(SP) in blue
respective controllers. for the ZN-PID.

Table 3. Parameter values for PID controller

Controller KP KI KD
ZN-PID 1.13 0.079 4.08
CS-PID 0.9255 0.0570 0.9940

Table 4 contains the system performance of the controller. The

performance is seen in terms of rise time, settling time and
percentage overshoot.

Table 4. Summary of performance of controller

Controller Rise time(s) Settling Overshoot (%)
time(s)
ZN-PID 2.63 51.5 47.3
CS-PID 6.41 43.4 18.5

It can be observed that from the simulation, the CS-PID has Figure 7. CS-PID Step Response
provided better overall system performance. The CS-PID has
provided better settling time and less overshoot whereas the

Authorized licensed use limited to: Turkish Energy Nuclear and Mineral Research Agency. Downloaded on April 10,2025 at 09:15:20 UTC from IEEE Xplore. Restrictions apply.
 Fig. 7 show the step response for the CS-PID. Based on [6] J. B. -, D. L. -, and K. Z. -, “PID Parameters
observation, the rise time for the CS-PID and ZN-PID was 1.58 Optimization for Liquid Level Control System Based
minutes. However, the ZN-PID had an overshoot of 8.15% on Genetic Algorithm,” Int. J. Digit. Content Technol.
whereas the CS-PID had an overshoot of 2.5%. Furthermore, its Appl., vol. 6, no. 1, pp. 361–368, 2012.
the step response of the CS-PID has a lower steady state error
as compared to the ZN-PID. The oscillation of the ZN-PID was
[7] M. K. Tan, Y. K. Chin, H. J. Tham, and K. T. K. Teo,
greater which in return forced the MV to open and close very
“Genetic algorithm based PID optimization in batch
vigorously from 0% to 100% as compared to the CS-PID. This
will increase the wear and tear of the valve which is not good process control,” 2011 IEEE Int. Conf. Comput. Appl.
for the valve over a long time period. Ind. Electron., no. Iccaie, pp. 162–167, 2011.

[8] K. Singh, P. Vasant, I. Elamvazuthi, and R. Kannan,

“PID Tuning of Servo Motor Using Bat Algorithm,”
Procedia Comput. Sci., vol. 60, pp. 1798–1808, 2015.
VI. CONCLUSION
[9] N. Hambali, A. Masngut, A. A. Ishak, and Z. Janin,
The simulation and plant experiment of the CS-PID and ZN- “Process Controllability for Flow Control System Using
PID have been successfully carried out. It can be observed that
Ziegler-Nichols ( ZN ), Cohen-Coon ( CC ) and Chien-
the CS-PID has a better overall system response than the ZN-
Hrones-Reswick ( CHR ) Tuning Methods,” no.
PID. Therefore, the CS algorithm could be used for PID tuning
algorithm for pressure plant. November, pp. 25–27, 2014.

[10] V. Balaji, M. Balaji, M. Chandrasekaran, and M. K. A.

ACKNOWLEDGMENT Ahamed, “Optimization of PID Control for High Speed
The authors would like to acknowledge the support of Universiti Line Tracking Robots .,” Procedia - Procedia Comput.
Teknologi PETRONAS and MyBrain15 which is the initiative Sci., vol. 76, no. Iris, pp. 147–154, 2015.
by the government of Malaysia in funding this project.
REFERENCES [11] I. Elamvazuthi, O. M. Aymen, Y. Salih, and H. Tawfeig,
“An Intelligent Control of Blood Pressure System
[1] J. Esch, S. X. Ding, N. Weinhold, and M. Schultalbers, using PID and Neural Network,” pp. 1049–1053, 2013.
“Control Performance Index Minimal Tuning of Set-
Point Weighted PID-Controllers for LTI Plants Based [12] H. M. Fayek, I. Elamvazuthi, N. Perumal, and B.
on Convex Optimisation,” pp. 1439–1444, 2013. Venkatesh, “A controller based on Optimal Type-2
Fuzzy Logic : Systematic design , optimization and
[2] K. V. Lakshmi and P. Srinivas, “Optimal tuning of PID real-time implementation,” ISA Trans., vol. 53, no. 5,
controller using Particle Swarm Optimization,” in pp. 1583–1591, 2014.
2015 International Conference on Electrical,
Electronics, Signals, Communication and Optimization [13] T. G. P. Vasant and I. E. K. Z. K. Shaari, “Game-
(EESCO), 2015, pp. 1–5. theoretic differential evolution for multiobjective
optimization of green sand mould system,” Soft
[3] “PSO based PID Controller Design for a Liquid Flow Comput., 2015.
Process,” vol. 4, no. 6, pp. 4252–4256, 2014.
[14] I. Elamvazuthi, P. Vasant, and T. Ganesan, “Handbook
[4] Z. Jun and Z. Kanyu, “A Particle Swarm Optimization of Optimization: From Classical to Modern Approach,”
Approach for Optimal Design of PID Controller for I. Zelinka, V. Snášel, and A. Abraham, Eds. Berlin,
Temperature Control in HVAC,” 2011 Third Int. Conf. Heidelberg: Springer Berlin Heidelberg, 2013, pp.
Meas. Technol. Mechatronics Autom., vol. c, no. 2, pp. 1025–1046.
230–233, 2011.
[15] P. N. Q. Nhon, I. Elamvazuthi, H. M. Fayek, S.
[5] L. I. Shaoming, Y. Peng, Y. Fan, and Z. Guoli, “On Lime Parasuraman, and M. K. A. A. Khan, “Intelligent
Kiln Temperature System Based on GA Fuzzy PID Control of Rehabilitation Robot: Auto Tuning {PID}
Algorithm and Optimal Control,” pp. 4545–4550, Controller with Interval Type 2 Fuzzy for {DC}
2014. Servomotor,” Procedia Comput. Sci., vol. 42, pp. 183–
190, 2014.

Authorized licensed use limited to: Turkish Energy Nuclear and Mineral Research Agency. Downloaded on April 10,2025 at 09:15:20 UTC from IEEE Xplore. Restrictions apply.
 [16] W. A. Soomro, I. Elamvazuthi, M. K. A. A. Khan, S. [18] X. S. Yang and S. Deb, “Cuckoo search via Lévy
Muralidharan, and M. Amudha, “PID Controller flights,” 2009 World Congr. Nat. Biol. Inspired
Optimization using Artificial Fish Swarm Algorithm,” Comput. NABIC 2009 - Proc., pp. 210–214, 2009.
vol. 2, no. 10, pp. 90–96, 2013.
[19] X.-S. Yang and S. Deb, “Engineering optimisation by
[17] X. S. Yang and S. Deb, “Cuckoo search via Lévy cuckoo search,” Int. J. Math. Model. Numer. Optim.,
flights,” 2009 World Congr. Nat. Biol. Inspired vol. 1, no. 4, pp. 330–343, 2010.
Comput. NABIC 2009 - Proc., pp. 210–214, 2009.

Authorized licensed use limited to: Turkish Energy Nuclear and Mineral Research Agency. Downloaded on April 10,2025 at 09:15:20 UTC from IEEE Xplore. Restrictions apply.