关键词: 技术诊断, 神经网络, 特征提取, 振动, 转子

Abstract: 　　In the process of computer-aided surveillance for large rotating machinery by means of vibrational signals, because of the inherent deficiencies of expert system and the operating characteristics of machine set, it is difficult to acquire a satisfied result by means of knowledge-based systems. Artificial neural networks exhibit attributes of associate memory, pattern matching, generalization and learning from examples. Considered as a kind of information processing method, it is promising and has a great potential used in the field of machinery fault diagnosis and surveillance. Field vibrational fault samples have been used to train the neural network. The features of several kinds of conventional faults have been discussed in detail, the fault feature extraction system is designed. In order to improve the generalization ability and decrease the on-line learning time, a two stage hybrid network has been proposed. The holospectrum and Wigner distribution have been used to verify the effectiveness of neural network.

Key words: Artificial neural network, Feature Extraction, Rotor Technical diagnosis, Vibration