A Turbing-Generator Unit Fault System Based on Wavelet Neural Networks

摘要

A noval method for the composite fault diagnosis based on wavelet neural network (WNN) is proposed. Wavelet transformation (WT) has the time-frequency location characteristic and multi-scale ability; artificial neural network (ANN) shows strong ability of self-adaption and self-learning and has wide applications in pattern recognition technique for complex non-linear system. The fault diagnosis model of turbine-generator unit is established and the method seriesly integrates the WT and the ANN. The paper also presented a new method of denoising by using the WT which solved two key problems: the determination of the threshold and the determination of the decomposition level. Fault character vector is extracted after denoising and sent to the ANNs to complete pattern recognition. With sufficient samples training, the type of fault can be obtained when signal representing fault is inputted to the trained ANNs. The diagnosis result approves to be accurate and comprehensive. The method can be generalized to other devices' fault diagnosis.