关键词: 故障诊断, 神经网络, 数学模型, 旋转机械

Abstract: 　　The artificial neural network (ANN) is introduced into the condition monitoring and fault diagnosis (CMFD) of large rotating machinery Based on the theoretic analysis of ANN, an improved training method for back propagation (BP) algorithm is developed to accelerate the convergence and escape the local minimum. In order to solve the difficulties generated from the training of multi-output networks, the ANN topologic architecture is first decomposed, then recognized synthetically. This model has been successfully tested with a complex set of sampled fault data of rotating machinery.

Key words: Fault diagnosis, Mathematical model, Neural network, Rotating machinery