Self-Tuning Neuro-Fuzzy Controller by Genetic Algorithm with Application to a Coupled-Tank

Abstract

Abstract This paper describes a self-tuning Neuro-fuzzy Controller by Genetic Algorithm (NFCGA) applied to a coupled-tank fluid level control. The controller use a simplified fuzzy algorithm which based on the Radial Basis Function neural network (RBF). Genetic Algorithm (GA) is used to tune simultaneously all the parameters of controller from a random state. A plant model is used initially to tune the parameters of the neuro-fuzzy controller. A coupled-tank system for fluid level control is used as a test bed. Nonlinearity is inherent in the plant due to the pumps, valves and sensors. The NFCGA is compared with a manually tuned conventional fuzzy logic controller (CFLC) and a PID controller in terms of setpoint tracking, load disturbance rejection and changes in the plant dynamics. It is found that the NFCGA copes well over the complexities in the plant, and has several advantages over the other two controller.