Optimization of Scaled Parameters and Setting Minimum Rule Base for a Fuzzy Controller in a Lab-Scale pH Process

Abstract

Experimental and simulation studies were conducted to design a multiregional fuzzy logic controller (FLC) for a lab-scale pH process system. Scaled coefficients of the controller were optimized offline to obtain the best controller performance in terms of cost function and function evaluation number using two widely used global optimization methods; namely, Genetic Algorithm (GA) and Particle Swarm Optimization (PSO). Various tuning parameters of these optimization algorithms were investigated in detail to compare their performance on this subject. The cost functions obtained via two algorithms were very close to each other, but the function evaluation number was almost 4-fold due to complicated computation of the GA compared to the PSO. On the other side, working with the PSO was much easier owing to fewer adjustable parameters. In the second part of the study, the rule base of the controller was minimized using two different new methods. In this context, the number of rules was reduced by 43% and 56% using the fuzzy matching set and the personal initiative methods, respectively. Rule base reduction increased the performance and the effectiveness of the controller, an important issue for real-time applications.

Systematic Procedure for Optimal Controller Implementation Using Metaheuristic Algorithms

“…A very large number of papers describe control systems that use artificial intelligence techniques for all the aspects that involve nonlinearities, or imprecise, incomplete, and uncertain knowledge. We recall here only a few topics; Model's Parameters Determination - Altinten (2007), System Identification - Makas and Yumusak, (2016), Optimization of Control Parameters - Kesarkar and Selvaganesan (2015); Aras et al (2011); Jabri et al (2011); Zhang et al (2018); Optimal Sensors' Placing - Bruant et al (2011); and Singh and Hahn (2006), Optimal Control Problems - Bououden et al (2015); Naghizadeh, et al (2016); Hu, X.B. et al (2004); Faber et al (2005); Qian et al (2012); and Wong et al (2016).…”