摘要：This paper presents a neuro-fuzzy network where allits parameters can be tuned simultaneously usingGenetic Algorithms. The approach combines themerits of fuzzy logic theory, neural networks andgenetic algorithms. The proposed neuro-fuzzynetwork does not require a priori knowledge aboutthe system and eliminates the need for complicateddesign steps like manual tuning of input-outputmembership functions, and selection of fuzzy rulebase. Although, only conventional geneticalgorithms have been used, convergence results arevery encouraging. A well known numerical examplederived from literature is used to evaluate andcompare the performance of the network with othermodelling approaches. The network is furtherim plemented as controller for two simulated thermalprocesses and their performances are compared withother existing controllers. Simulation results showthat the proposed neuro-fuzzy controller whose allparameters have been tuned simultaneously usingGAs, offers advantages over existing controllers andhas improved performance
