Simplified dynamical neuro-fuzzy network controller and application

The authors previously presented a dynamical neuro-fuzzy network (DFNN). Based on the forward neuro-fuzzy network ANFIS, it combines the advantages of fuzzy system, neural network and PID algorithm by adding a recurrent layer between the normalized layer and output layer of ANFIS. This paper simplified the structure of DFNN and a simplified dynamical fuzzy neural network (SDFNN) is proposed. SDFNN adopts only one recurrent neuron in recurrent layer by connection to the outputs of fuzzy zero rules, so a kind of PID controller is realized by SDFNN. The simulation results show SDFNN has the quick response as ANFIS does but possess the higher performance accuracy than ANFIS