T-S Fuzzy Modeling and Application Based on Satisfactory Optimization

A T-S model fuzzy modeling method based on satisfying degree function is presented for a class of complex systems. Using the sampling data, the model parameters are initialized by fuzzy clustering and its premise parameters are rectified by learning off-line using back-propagation algorithm. Introducing the conception of character satisfying degree function to rectify online the forgetting factor of recursive least square method, the consequent parameters of the fuzzy rules are self-learning online by recursive least square method. Consequently, the precision and the identify speed of the T-S model are improved. Applying to locomotive brake control unit, the result shows the effectiveness of the proposed method.