Auto-generation of fuzzy control rule base

A method using a self-organizing feature map network to generate a rule base for fuzzy control from measured data is presented. An unsupervised competitive learning algorithm is used to speed up the learning and convergence of training and to reduce the degree of noise disturbance. We study the consistency and completeness of the fuzzy control rules and we also design an interactive utility system for the fuzzy controller in order to collect sample data and to examine the proposed method by computer. The fuzzy controller formed by a self-generated rule base has been applied to a time-delay system in simulation experiments. The step response shows that the method of generating the rule base is correct.