Metaheuristic Inspired Algorithm for Supervisory Fuzzy Controller for Plant wide

Abstract

Plantwide faces challenges to obtain controllers that are fit to correct errors from actual and desired values. This makes industries such as methanol production industry to have nonlinear process in which the process conditions such as temperature, pressure, and level are interacting each other and causing to reduce product quality, equipment damage, explosion, and flooding. Therefore, the goal of this research is to investigate tuning algorithm performance for multiple loop control system by developing metaheuristic tuning algorithm performance for fuzzy logic controller and by developing dynamic process models for plantwide application. However, plantwide process dynamic is simulated using UniSim software for controlling temperature of cooler, pressure and level of flash drum. Mathematical modelling is performed to derive transfer function of the respective process while taking the values of variables from the process simulated in UniSim. Fuzzy Logic Controller (FLC) is designed as there are three inputs and outputs as each one have gaussian type membership functions in which finally rules were defined. Since manually configuration control system is time consuming and abundant trial and errors to get the optimal parameters Particle Swarm Optimization (PSO) is implemented to use as metaheuristic algorithm to optimise control performance by finding particle best to update global best through successive iterations and swarm size. Thus, it is found that control performance increased as the PSO used to optimize control system for Proportional-Integral-Gain (PID) and FLC. The result got shows significant decrease of overshoot and undershot and responding faster with less rise time and settling time after using PSO to optimise conventional controller that combined with FLC. For example, pressure is controlled quickly to reach stability within 1 second after optimised FLC and PID as it has rise time of 2.8704 x 10^-11 sec for the PSO-PID and 3.3242 x 10^-12 sec for Optimised FLC.