Performance Evaluation of an On-Line Self-Learning Fuzzy Logic Controller Applied to Non-Linear Processes

Abstract

The generation of rule-bases in conventional fuzzy logic controllers can be a difficult and time consuming problem for implementation by process operators thus affecting their wider applicability. A Self-Learning Fuzzy Logic Control (SLFLC) offers a possible solution. A performance study is therefore presented to evaluate the performance of a proposed SLFLC by analysing its transient performance for a variety of on-line tests and examining its ability to generate a consistent set of rules, based on a predetermined criteria. The results presented show that even with a limited knowledge of the process, the self-learning procedure is able to develop a suitable set of rules and produce a satisfactory process performance with some degree of robustness and repeatability when applied to a non-linear single-input single-output (SISO) or multi-input multi-output (MIMO) laboratory liquid-level processes.