一种基于RBF神经网络的智能故障诊断方法

Intelligent Fault Diagnosis Based on RBF Neural Network

根据神经网络独特的容错、联想、推测、自适应、自学习等优点,针对BP网络在故障诊断应用中收敛速度慢等不足,研究了基于RBF神经网络的智能故障诊断方法。该诊断方法只需要足够的具有代表性的故障样本用以训练神经网络,然后将归一化的故障信息输入给训练好的神经网络,根据其输出结果就可以判断发生的故障类型。利用该诊断方法,对发动机转子系统故障诊断进行了仿真,仿真结果表明,基于RBF神经网络的智能故障诊断方法效果良好。

Because of the flaw of the slow convergence rate of the BP neural network, intelligent fault diagnosis is studied based on RBF neural network with its unique advantages of fault containing, association, guessing, self- adaptation mad self-study. This method of fault diagnosis only needs enough representative fault swatches to train the neural network. As the unitary detecting information is inpuf into the neural network, its output can figure out what kind of fault there is. With this fault diagnosis method, the engine rotor fault diagnosis system was simulated. It was showed in the simulation results that the intelligent fault diagnosis based on RBF neural network is effective.