GA-Based Optimization for Multivariable Level Control System: A Case Study of Multi-Tank System

Abstract

This paper presents a systematic way to determine the trade-off optimized controller tunings using computation optimization technique for both servo and regulatory controls of the Multi-Tank System, as one of the applications under the multivariable loop principle. The paper describes an improved way to obtain the best Proportional-Integral (PI) controller tunings in reducing the dependency on engineering knowledge, practical experiences and complex mathematical calculations. Relative Gain Array (RGA) calculation justified the degree of relation and the best pairing for both interacted control loops. Genetic Algorithm (GA), as one of the most prestigious techniques, was used to analyze the best controller tunings based on factor parameters of iterations, populations and mutation rates to the applied First Order plus Dead Time (FOPDT) models in the multivariable loop. Amid simulation analysis, GA analysis’s reliability was justified by comparing its performance with the Particle Swarm Optimization (PSO) analysis. The research outcome was visualized by generating the process responses from the LOOPPRO’s multi-tank function, whereby the GA tunings’ responses were compared with the conventional tuning methods. In conclusion, the result exhibits that the GA optimization analysis has successfully demonstrated the most satisfactory performance for both servo and regulatory controls.