Application in Servo System of Dynamical Neuro-fuzzy Network Controller

A novel dynamical neuro fuzzy network (DFNN) is proposed by adding a recurrent layer between the normalized layer and output layer of the forward neuro fuzzy network ANFIS. DFNN combines the advantages of fuzzy system, neural network and type Ⅲ controller. The structure of DFNN and a parameter regulating method which is based on the shrinking span membership functions and BP algorithm are proposed. The experiment results show that DFNN has a better response than the traditional PID+forward controller especially in the situation when the forward signal is difficult to obtain.