부산대학교 도서관

Recurrent neural network is a nonlinear dynamical system. This paper firstly introduces the principle of artificial neural network algorithm and the improvement of backpropagation, self-organizing competitive neural network and probabilistic neural network algorithm. At the same time, this paper applies it to fault diagnosis and prediction. Then this paper applies the improvement of back-propagation neural network algorithm, self-organized competitive neural network and probabilistic neural network algorithm to diagnose and predict faults. This paper innovatively introduces an improved backpropagation neural network algorithm with momentum factor to diagnose the actual data and compare it with the traditional one. Finally, this paper proves that the proposed method is effective through Simulink, Spice simulation and hardware circuit experiments.