Optimisation of neural fuzzy control system - having multi-stage learning process in which three level neural network is used together with fuzzy logic rules

摘要

The neural network based fuzzy logic controller has a three-level structure. Located between two of the levels (I,II) are minimal networks for learning a set of three membership functions. The inputs (1,2) in the first level connect with pairs of elements (3,4 and 5,6), and with further elements (7,8,9 and 10,11,12). Between the second and third levels (II,III) are the inference elements of the neuro-fuzzy system. AND functions are provided by nine neurons (13-21). Outputs are generated by elements (22,24) that provide a 'defuzzyfying' action. The process operates to continuously 'learn' until an optimum condition is achieved. USE/ADVANTAGE - Simplifies utilisation of expert system.