Combination of Wavelet Packet Band Energy Analysis and AR model Bi-spectrum in Application on Circuit Breaker Condition Detection

In this paper, the field vibration signal of circuit breaker (CB) with heavy noises disturbing is firstly de-noised with WP (wavelet packet) method by which the feature band of the signal is got. With singular value decomposition (SVD), the parameters and the orders of AR model are got in the modified Yule-Walker equation of the feature band signals. The bi-spectrum analysis is then used to done with the determined parameters, which provides the bi-spectrum characteristics for on-line fault diagnosis of CB. The results analysis of the field data collected from different period proves the validity of this method which can be used as a means on on-line fault diagnosis of CB.