Method of fault diagnosis on multilayer BP wavelet networks and its applications

Stepped from the multi-level characteristic of system fault, this article puts forward a multilayer BP wavelet network fault-diagnosis model which is a mixer of level-distinguishing model on the basis of knowledge-classification and BP wavelet network, probes into the steps of this model, tests and verifies the correctness of these theories with practical examples. Systematic study of diagnosis knowledge avoids the combination-explosive problem of classification space, and dealing with fault from an overall viewpoint enables itself to solve problems quickly, and self-adapted adjustment of scale parameters of wavelet transform could improve convergence and discernment of the network. Then this method is suited for constituting of on-line fault-diagnosis.