A new adaptive configuration of PID type fuzzy logic controller. (May 2015)

• Record Type: Journal Article
• Authors: Fereidouni, Alireza; Masoum, Mohammad A.S.; Moghbel, Moayed
• Is Part Of: ISA transactions. Volume 56(2015:May)
• Page Start: 222
• Page End: 240
• Publication Date: 2015-05
• Subjects: Fuzzy logic -- PID controller -- Scaling factors; Engineering instruments -- Periodicals; Engineering instruments; Periodicals; Electronic journals; 629.805
• DOI: 10.1016/j.isatra.2014.11.010
• Languages: English
• ISSNs: 0019-0578

Abstract

Abstract: In this paper, an adaptive configuration for PID type fuzzy logic controller (FLC) is proposed to improve the performances of both conventional PID (C-PID) controller and conventional PID type FLC (C-PID-FLC). The proposed configuration is called adaptive because its output scaling factors (SFs) are dynamically tuned while the controller is functioning. The initial values of SFs are calculated based on its well-tuned counterpart while the proceeding values are generated using a proposed stochastic hybrid bacterial foraging particle swarm optimization (h-BF-PSO) algorithm. The performance of the proposed configuration is evaluated through extensive simulations for different operating conditions (changes in reference, load disturbance and noise signals). The results reveal that the proposed scheme performs significantly better over the C-PID controller and the C-PID-FLC in terms of several performance indices (integral absolute error (IAE), integral-of-time-multiplied absolute error (ITAE) and integral-of-time-multiplied squared error (ITSE)), overshoot and settling time for plants with and without dead time. Highlights: An adaptive configuration for a PID type fuzzy logic controller has been introduced, implemented and verified. The proposed adaptive structure enhances the performance of the conventional PID controller and also conventional PID-FLC for plants with and without dead time. Two different approaches have been introduced for determining the initial scaling factors; first, the conventional method that defines them based on its well-tuned linear counterpart and second, a stochastic method that defines them based on a new h-BF-PSO algorithm.