A Novel Fuzzy CMAC Network and Its Application

A novel fuzzy CMAC network is proposed,using brannew changeable reception field in place of constant reception field.Meanwhile the net introduces a channelopening layer of the rules to replace usual hash layer to avoid the collision problem.The membership function of fuzzy net adopts the triangle function.The parameter training adopts the improved evolutionary programming algorithm with the proposed recursivechanging rate.The weights of the memory region are trained by the whole leastsquares methods.It is concluded that in the soft sensing of the transformation rate of acrylonitrile,the proposed fuzzy CMAC network has an excellent performance in the generalization and learning speed.