¹ Department of Instrumentation and control Engineering,
Kalasalingam University, Krishnankoil, Tamilnadu, India
saravana.control@gmail.com; mpraja83@gmail.com; seshucontrol@gmail.com
² Department of Electronics and Communication Engineering,
PSR Engineering College, Sivakasi, India
krvalarmathi@yahoo.co.in

Abstract: Implementation of State Transition Algorithm (STA) for the tuning of multivariable Proportional Integral Derivative (PID) controller is reported. Two input and two output binary distillation column plant model by Wood and Berry is considered as a benchmark. Simulations are performed for three cases such as multivariable PID controller with decoupler, without decoupler and multivariable PI controller without decoupler. Integral of Time weighted Absolute Error (ITAE) and Integral of Absolute Error (IAE) are chosen as objective functions for the first case while Integral of Absolute Error alone is chosen for remaining couple of cases. Simulations are carried out for 20 independent trials by STA and STA with SBX crossover. Comparison of fitness value and time domain parameters for three cases against PBPSO is reported. Statistical performance measures of STA and STA-SBX algorithms are presented and simulation result implies that STA is a potential algorithm which provides better fitness for all the three cases and has global search capability. SBX crossover is enhanced by the consistency of the algorithm.

Keywords:  Multivariable PID Controller Tuning, State Transition Algorithm, Distillation column, MIMO process.

>>Full text>>
CITE THIS PAPER AS:
G. SARAVANAKUMAR, K. VALARMATHI, M. PALLIKONDA RAJASEKARAN, Seshadhri SRINIVASAN, M. WILLJUICE IRUTHAYARAJAN, Valentina E. BALAS, Tuning of Multivariable DecentralizedPID Controller Using State Transition Algorithm, Studies in Informatics and Control, ISSN 1220-1766, vol. 24 (4), pp. 367-378, 2015. https://doi.org/10.24846/v24i4y201501

Introduction

Proportional Integral Derivative controller is one of the simplest and most commonly used ones in various industries for control applications. Despite significant advancements in control technology, over 80% of industrial control loops are incorporated with PID controller. Though it is widely accepted, it should be properly tuned to meet desired behavior. Extensive work of Ziegler and Nichols [1] is the breakthrough in tuning methodology and Cohen Coon, Lambda tuning, and Chen Hrown Reswick (CHR) methods are a few of the other tuning methods reported in the literature [2]. Existing tuning methods are classified [3] based on nature and usage as analytical methods, Heuristic method, Frequency response method, Optimization method and Adaptive tuning methods. Among those, optimization method is widely utilized around the globe as it is conceptually simple and widely accepted one for tuning PID controller [2]. In this method, controller parameters are adjusted based on the chosen objective function chiefly integral performance measures. A classical optimization technique namely gradient method is often used to find optimal values. The shortcoming of gradient descent methods is sensitivity to the selection of initial values and their tendency to lock into a local extreme point [4]. Evolutionary Computation techniques are proposed to tune the PID controller by taking all non-linearity and additional process characteristics into account [5], [6]. Genetic Algorithm (GA) has the capability to solve nonlinear and complex optimization problems [7]. Porter and Jones proposed a GA-based simple and generic method of tuning digital PID controller [8].

Numerous work is available in the literature related to Distillation column control strategy [9-14]. Very recently, various optimization techniques such as Covariance Matrix Adaptation Evolution Strategy (CMAES), Particle Swarm Optimization (PSO), Differential Evolution (DE), Tribes Algorithm (TA), Ant Colony Optimization (ACO), Tabu Search Algorithm (TSA) and different Binary Particle Swarm Optimization are used to tune the PID control parameters [15-19]. Modified firefly algorithm for the tuning ofmultivariable PID controller is implemented for distillation column [20].

Iruthayarajan and Baskar compared the performance of various Real Coded Genetic Algorithm (RGA), Differential Evolution (DE), Modified Particle Swarm Optimization (MPSO) and Covariance Matrix Adaptation Evolution Strategy (CMAES) in the multivariable distillation column [21]. Discrete Binary PSO (DBPSO) algorithm, Probability based discrete binary PSO (PBPSO) and Modified Discrete Binary PSO (MDBPSO) are proposed to tune multivariable PID controller for the distillation column and compared the results with the RGA, MPSO and CMAES [22].

Amongst, PBPSO provided the best optimal value for all the three cases such as multivariable PID controller with decoupler, without decoupler and multivariable PI controller without decoupler. Recently proposed State transition Algorithm (STA) is applied to different benchmark problems which has the ability to reach a global optimal solution and shown that it has good convergence property when compared with a Real coded genetic algorithm (RCGA), Complementary Learning Particle Swarm optimization (CLPSO) and Strategy adaptation Differential Evolution (SaDE) [23]. Herein, Implementation of State Transition Algorithm (STA) and STA-SBX Algorithm to tune multivariable PID controller for the distillation column is proposed and compared with already reported best results.