You are on page 1of 12

23

CHAPTER 2

LITERATURE REVIEW OF THE CONTROLLERS FOR


THE CONICAL TANK LEVEL PROCESS

2.1 INTRODUCTION

Proportional Integral Derivative (PID) Controllers are widely used


in process industries for controlling the different process variables, due to
their simplicity, flexibility and efficiency. The tuning of the PID Controller is
very essential for the satisfactory operation of the process. In this thesis, the
application of the Conventional PI Controller, Genetic Algorithm tuned PI
Controller, Neuro tuned PI Controller, and Fuzzy Logic Controller for the
conical tank process is investigated. The controller performance is obtained,
in terms of ISE, ITSE, IAE and ITAE, and also through the time domain
specification criteria for servo, regulatory and servo-regulatory operations for
the conical tank process. Therefore, a literature review on the different
controllers and their performance is done, and reported in this chapter.

2.2 LITERATURE REVIEW

Proportional Integral Derivative (PID) Controllers have been


widely used in many process industries for several decades, due to their
simplicity, flexibility and efficiency. Zeigler and Nichols (1942) developed
the PID Controller tuning concept for open and closed loop operations. The
controller tuned parameters with Zeigler-Nichol’s method, perform well in
disturbance rejection, and poorly in tracking reference changes. Cohen and
24

Coon (1953) proposed the open loop tuning or process reaction curve method.
In the open loop tuning method, the connection between the controller and the
final control element is disconnected, and the step input is applied as an input
to the final control element and the process parameters are obtained from the
process reaction curve. The above mentioned method requires limited process
knowledge, but it offers lower damping and high sensitivity to the system.
Wills (1962) proposed the tuning maps concept for three mode controllers. In
the above mentioned paper, the overall approach for the calculation of
proportional gain, integral time and derivative time was discussed, and the
tuning concepts were highlighted. Miller et al (1967) proposed a comparison
of the controller tuning technique, in which the investigation of the process
reaction curve method is demonstrated and the process reaction curve
approximation is done; comparatively Miller et al (1967) is better than Cohen
and Coon (1953).

Parker (1972) introduced the concept of the design of Proportional


Integral Derivative (PID) Controllers by the optimal linear regulator theory.
In the above method, the optimal regulator was used to generate the
Proportional Integral Derivative (PID) Controller’s parameters, such as
proportional gain, integral time and derivative time. Hagglund et al (1985)
developed the auto tuning technique for the Proportional Integral Derivative
(PID) Controller, based on the dominant pole design. Astrom et al (1986)
introduced the concept of the expert systems in which the software programs
are supplemented by man machine interfaces. The design and application of
knowledge based expert systems for system control, improves the efficiency,
effectiveness and performance under uncertain and varying operating
conditions. Rivera et al (1986) introduced the design procedure of the Internal
Model Controller to establish PID rules with a well described approach. The
Internal Model Controller was developed (Morari and Zufiriou 1987), and it
25

overcomes the problem of disturbance rejection but its design calculation is


complicated for higher order processes.

Deshpande (1988) introduced optimization methods to obtain the


PID parameters by the optimization of the Integral Square Error (ISE),
Integral Absolute Error (IAE), Integral Time Square Error (ITSE) as the
performance indices. Corripio (1990) compared the performance of the error
criteria, in which the Minimization of Integral Absolute Error (IAE) generally
produces intermediate responses between fast responses, obtained from the
minimization of the Integral Square Error (ISE), and slow responses obtained
from the minimization of the Integral multiplied Time Absolute Error (ITAE).
Next, Hang et al (1991) developed the online auto-tuning of the Proportional
Integral Derivative (PID) Controller, based on the cross correlation technique,
described in the paper. In the above method, the Pseudo Random Binary
Signal (PRBS) is used as a reference signal to tune the auto tuner circuit.
Hang et al (1991) introduced refinement in the Zeigler Nichols tuning
formulae. Astrom et al (1992) introduced the concept towards intelligent
Proportional Integral Derivative (PID) Controllers. In the above paper, the
possibility of different intelligent controllers for the Proportional Integral
Derivative (PID) controllers, is also discussed.

Porter and Jones (1992) proposed the concept of the Genetic


Algorithm based technique for finding the tuning parameter of the digital PID
Controller. The above mentioned method is a simple, generic approach. A
classical method of optimization, namely, the gradient method by Varsek et al
(1993) is an often used method to find the optimal solution. The shortcoming
of the gradient descent method is its sensitivity to select the initial value. The
above mentioned method is well suited for the Proportional Integral
Derivative (PID) Controller, where the controller structure and parameters are
26

given. Zhuang et al (1993) introduced the concept of the automatic tuning of


the optimal Proportional Integral Derivative (PID) Controller.

Wang and Kwok (1994) proposed the optimal design of the


Proportional Integral Derivative (PID) Controller, based on the Genetic
Algorithm. Ho et al (1995) introduced the performance and gain, and phase
margin with a well known Proportional Integral Derivative (PID) Controller
tuning formula. Dangprasert et al (1995) employed the Genetic Algorithm for
on-line parameter identification and controller tuning in the frequency control
of a power system. Zibo and Naghdy (1995) proposed the concept of the
application of the Genetic Algorithm for system identification. The limitations
and counter measures of the Proportional Integral Derivative (PID) Controller
for process control applications was introduced by Lee et al (1996). Rad and
Lo (1977) introduced the concept of a self tuning PID Controller, using the
Newton-Raphson method. Astrom et al (1998) proposed the design of
Proportional Integral (PI) Controllers, based on the non-convex optimization
technique. Ho et al (1998) developed the concept of optimal gain and phase
margin, for the calculation of the parameters of the Proportional Integral
Derivative (PID) Controller. Wang et al (1999) proposed the Proportional
Integral Derivative (PID) Controller tuning with improved performance for
the dynamic plant. Ya-Gang Wang and Hui-He Shao (2000) introduced a
simplified optimum method for the Proportional Integral (PI) Controller,
which is applicable to different complicated processes. The above mentioned
method produces high performance, and is widely used for linear self
regulating processes.

Astrom and Hagglund (2001) proposed the future of the


Proportional Integral Derivative (PID) Controller tuning concept, which is
applicable to the different complicated processes. Krohling and Rey (2001)
developed the design of the optimal disturbance rejection of the Proportional
27

Integral Derivative (PID) Controller, using the Genetic Algorithm. Liu and
Daley (2001) proposed the optimal tuning concept for the Proportional
Integral Derivative (PID) Controller, which is applicable to different
industrial system’s applications. Ari Ingimundarson and Tore Hagglund
(2002) compared the performance of the Proportional Integral Derivative
(PID) Controller and that of the dead time compensating controller, based on
the Integral Average Error (IAE) optimization technique. Ming Ge et al
(2002) introduced the robustness of the Proportional Integral Derivative (PID)
Controller in the LMI (Linear Matrix Inequality) technique. Its applicability
in different dynamic plants is also discussed in the paper. Next, Kao-Shing
and Hwang et al (2003) proposed the reinforcement learning algorithm to the
adaptive control scheme, which is applicable to different nonlinear dynamic
processes. Skogestad (2003) developed simple analytical rules for model
reduction, as well as the Proportional Integral Derivative (PID) Controller
rules for process control applications. Wang and Spronck (2003) proposed an
overall view of the Genetic Algorithm applied to different control engineering
problems. Lin and Chou (2003) proposed the Induction motor servo drive,
using the slide mode controller with the Genetic Algorithm. Teng et al (2003)
implemented the Genetic Algorithm (GA) which is applicable to the on-line
tuning parameter of the Proportional Integral Derivative (PID) controller, in
the liquid level system for the cylindrical tank.

Next, Astrom and Hagglund (2004) proposed the concept of


revisiting the Zeigler-Nichol’s step response method for the Proportional
Integral (PI) Controller. In the above paper, the frequency response method is
limited and the new tuning rule is introduced. Kazantzis et al (2005) proposed
the optimal tuning control technique for nolinear processes. Toscano (2005)
proposed a simple Proportional Integral (PI) and Proportional Integral
Derivative (PID) Controller, with a numerical optimization technique, which
is applicable to different processes. Jie-Sheng et al (2006) introduced the
28

optimal design of the Proportional Integral (PI) and Proportional Integral


Derivative (PID) Controller for the non-minimum phase system. Aytekin
Bagis (2007) proposed the Modified Genetic Algorithm to find the
Proportional Integral Derivative (PID) Controller’s optimal parameters, for
improved specifications for the dynamic plant. Recently, Valarmathi et al
(2007) developed the evolutionary algorithms like the Genetic Algorithm,
which is used to obtain the optimal parameters of the Proportional Integral
Derivative (PID) Controller in the chemical pH process. Kim et al (2008)
proposed the auto tuning Proportional Integral Derivative (PID) Controller,
based on the improved Genetic Algorithm for the reverse osmosis plant. Luo
et al (2010) developed the tuning of the Proportional Integral Derivative
(PID) Controller parameters in the hydraulic servo control system, based on
the evaluation of the Genetic Algorithm. Arya et al (2011) proposed the
evaluation of the Genetic Algorithm optimization technique for the tuning of
the Proportional Integral Derivative (PID) Controller. Mohd Sazli Saad et al
(2012) developed the Genetic Algorithm tuned Proportional Integral
Derivative (PID) Controller tuning, by a differential equation approach.

The Fuzzy Logic set theory concept similar to mathematics, was


first proposed by Lotfi A. Zadeh (1965) of the University of California.
Again, Lotfi A. Zadeh (1973) elaborated the concept of Linguistic variables
which are related to the fuzzy set theory. Mamdani and Assilian (1975)
developed the application of the Fuzzy Algorithm to approximate reasoning,
using Linguistic variables for controlling the steam generator for dynamic
plant control. Mamdani (1977) introduced the application of Fuzzy Logic to
approximate reasoning, using the Linguistic variable synthesis in the design
of controllers in industrial applications. Fukami et al (1980) proposed some
consideration of the fuzzy conditional inference for different fuzzy logic
applications. Tong (1980) obtained the evaluation of the fuzzy model derived
from the experimental data. Takagi and Sugeno (1985) developed the concept
29

of the fuzzy identification of the system, and also applied it for modeling and
control. Sugeno and Kang (1988) introduced the concept of the structure
identification of the fuzzy model. De Silva and Mac Farlane (1989)
introduced knowledge based control, for application to robotics. Li and Lan
(1989) developed the fuzzy logic algorithm which is applicable to servo
system control applications. Lee (1990) developed models of Fuzzy Logic
Controllers for applications in control systems. Chui et al (1991) developed
fuzzy logic for the control of roll and moment for the flexible wind of the
aircraft. Layne and Passino (1993) developed the fuzzy model reference
learning control, for cargo ship steering control. Garcia-Benitez et al (1993)
developed the rule based supervisory control of a two-link flexible
manipulator. Ying (1993) introduced the simplest fuzzy controllers, using the
different inference method for different nonlinear PI Controllers, with
variable gain adjustment. Kwong et al (1996) proposed the concept of fuzzy
learning control for different applications. Cheng et al (1996) developed the
Fuzzy Logic Controller application for the inverted pendulum.

Chao and Teng (2000) introduced the Proportional Derivative


(PD) Controller for the Self tuning Fuzzy Logic Controller, in which the
system performance is studied without the steady state error. Yi and Yuhazaki
(2000) developed the fuzzy logic controller for inverted pendulum control
applications. Jin Zhao and Bose (2002) developed the evaluation of the
membership functions for the fuzzy logic controller. Chia-Feng Juang (2003)
developed an automatic building approach to the special Takagi-Sugeno fuzzy
controller, for unknown modelling and stable control applications. Cheng-
Awang et al (2005) developed the concept of the Fuzzy based Proportional
Integral Derivative (PID) Controller for cement plant industries. Jian Lin et al
(2006) developed the neuro-fuzzy controller for temperature control
applications, in which compensatory operations and wavelet neural networks
are used. Ahmed et al (2007) developed online neutralization based on the
30

fuzzy logic controller. Lee and Gonzalez (2008) developed the fuzzy logic
versus PID Controller, for the position control of a muscle-like actuated arm
application. Rahul Malhotra et al (2011) developed the concept of boiler flow
control with the Proportional Integral Derivative(PID) Controller with the
Fuzzy Logic Controller. Li et al (2012) developed the concept of the
application of the fuzzy logic controller for maintaining the liquid level.

Mc Culloach and Pitts (1943) developed the first neural network in


which logical ideas of mathematics with bioscience were introduced. Frank
and Rosenblatt (1958) proposed the Perceptron model for single layer and
multilayer feedforward networks, in which the network is able to obtain the
weight only on a linearly separable task. Kohonen (1982) developed the Self
Organization Map (SOM) architecture of a competitive learning approach, in
which the network is a simplified model of the feature for the localized region
mapping of the brain, and published this concept in a book in the year 1989.
Rumel hart, Hinton and Williams (1985) introduced the concept of the
learning rule, which is applicable to the Feed forward Back Propagation
Neural Network. But it has the difficulty of a slow rate of convergence and
the local minima problem. Funahasi (1989) introduced an approximate
relation ship of continuous mapping by Neural Networks. Hornik et al (1989)
proposed that multilayer feed forward networks are universal approximators.
Narendra and Parthasarathy (1990) proposed the concept of process
identification and control of the dynamic system, using Neural Networks. Wu
and Hogg (1992) developed the Neural network regulator for turbo
generators. Linsker (2003) introduced the concept of the self organization
approach in a Perceptual network. Kim (2006) developed a Neural network
based tuning of the Proportional Integral Derivative (PID) Controller, for the
level control of a steam generator. Bhuvaneswari et al (2008) proposed the
Neural network based Model reference Adaptive Control for the conical tank
31

level process. Suresh et al (2009) proposed the design of a plant estimator,


using Neural networks for the control of the conical tank process, in which the
experiments are conducted for servo and regulatory operations. Wang et al
(2011) proposed a Neural Network based self tuning PI Controller, for the
precision motion of a PMAC motor.

Madubala et al (2004) proposed the real time control of the conical


tank process liquid maintenance, in which the Genetic Algorithm is used to
tune the membership function of the Fuzzy Logic Controller, and compared
its performance with that of the Proportional Integral (PI) Controller for the
nonlinear conical tank process. Anandanatarajan et al (2005) developed the
design of a Smith Predictor PI Controller. This was then compared with the
Conventional PI Controller, for the first order nonlinear process with dead
time of the conical tank process. Bhaba et al (2007) developed the real time
implementation of the Wiener Model Proportional Integral (PI) Controller for
the non linear conical tank liquid level process, in which the Integral Square
Error (ISE) criterion is obtained for the set point and load changes.
Bhuvaneswari et al (2008) developed the Neuro based Model Reference
Adaptive Controller for a nonlinear conical tank process, in which the time
optimal control for set point changes and adaptive parameter variation are
done, using a neural network. Further, Nithya et al (2008) developed the
design of a soft computing based controller for the non-linear conical tank
process, and the performance was evaluated based on criteria, such as the
Integral Square Error (ISE) and Integral Absolute Error (IAE).

Bhaba et al (2009) developed the real time implementation of a


new Coefficient of Discharge Method the Proportional Integral (CDM-PI)
Control scheme for a conical tank liquid level maintenance. The servo
tracking operation for different level changes for error criteria such as,
32

Integral Square Error (ISE), Integral Absolute Error (IAE) as well as time
domain specification criteria is analyzed in this paper. Next, Ram Kumar et
al (2010) proposed the real time application of the Ant Colony optimizing
algorithm, and implemented it for the conical tank process, in which the
performance of the Internal Model Controller (IMC) is compared with that of
the ant optimization colony algorithm through the performance error criteria.
Suresh et al (2010) proposed the design of a Fuzzy Estimator to assist the
fault recovery in a conical tank process in which the servo response is
compared with the regulatory operation. Anand et al (2011) introduced a
simplified tuned Adaptive Proportional Integral (PI) Controller for the conical
tank process, in which the performance of the Adaptive Proportional Integral
(PI) Controller is compared with that of the Conventional PI Controller.
Fareeza et al (2012) introduced the design of a plant estimator model, using
the Fuzzy Logic Controller (FLC) for the conical tank process, in which the
servo operation performance is compared with regulatory operation. Ganesh
Ram and Abraham Lincoln (2012) proposed the Fuzzy Adaptive PI Controller
for the nonlinear conical tank process, in which the Fuzzy Adaptive PI
Controller’s performance is compared with that of the Conventional PI
Controller. Arivalahan et al (2012) developed a Genetic Algorithm tuned PI
controller in which the optimal values of the PI Controller parameters are
obtained. In addition, the Neuro tuned PI Controller and Fuzzy Logic
Controller are designed for the conical tank level process. The developed
controllers are evaluated in terms of the performance indices, namely, ISE,
ITSE, IAE, ITAE, and time domain specifications, such as Peak over shoot
(Mp), Settling Time (Ts) and steady state error (ess) for servo, regulatory and
servo-regulatory operations. Further, the modelling aspect of the conical tank
level process is included, and the stability analysis of the system is evaluated
through the pole- zero plot and Nyquist stability criteria.
33

2.3 CONTROL OF THE CONICAL TANK WITH


COMPUTATIONAL INTELLIGENCE TECHNIQUES

The control of the conical tank is a challenging task, due to its


nonlinearity and constant change in the cross section. Therefore, in this thesis,
computational intelligence techniques, using the Conventional PI Controller,
Genetic Algorithm tuned PI Controller, Neuro tuned PI Controller and Fuzzy
Logic Controller are investigated for the conical tank process. The
Limitations of the Conventional PI Controller are:

Not preferable for nonlinear processes.

Settling time is more.

Not preferable under parameter variations.

Robustness is low.

The control action of the conical tank can be done by Servo,


Regulatory and Servo-Regulatory operations for different types of controllers.
In the servo operation, the set point is varied, and the process with the load
variable is kept constant. In the regulatory operation, the set point is kept
constant, and the process with the load variable is changed. In the servo-
regulatory operation, the set point and process with the load variable are also
variable. The comparison of the controller performance is obtained, in terms
of the ISE, ITAE, IAE, ITSE, and through Time domain specifications, and
also through the graphical approach.
34

2.4 SUMMARY

This chapter presented the literature review of the conical tank


process, and the different types of controllers used for the conical tank
process have been reviewed. The limitations of the Conventional PI
Controller for the conical tank process have been outlined. The suitability of
the computational by intelligent techniques like, the Genetic Algorithm tuned
PI Controller, Neuro tuned PI Controller and Fuzzy Logic Controller for these
tasks, has been identified. The performances of these controllers are
investigated in the coming chapters.