Adaptive Neurofuzzy Network Based PI Controllers with Multi-objective Functions

Abstract

As the performance of PI controllers can deteriorate rapidly for highly nonlinear systems, nonlinear PI controllers are developed. An approach to design these controllers is to switch between several linear PI controllers using fuzzy logic based on the Takagi-Sugeno model. Following this approach, nonlinear PI controllers are derived in this paper using B-spline neurofuzzy networks. Design guidelines and on-line training of the proposed controller are devised, and the performance is illustrated by a simulated two-tank water level control rig. Comparison with conventional PI controllers is also made.