A novel method to identify and locate rotor-stator rub-rubbing fault

Abstract

Starts with the complexity of signals and a method to identify and locate a rubbing fault with dual complexity parameter is proposed. Firstly, signals are coarse-grained and its multiscale Lempel-Ziv complexity is calculated to measure the complexity of time sequence. Secondly, the complexity parameter of Hjorth parameters is taken advantage and the complexity of sequence is measured from the similarity of signal and sinusoidal signal. Two complexity parameters are utilized to extract fault information comprehensively. Thirdly, the dual complexity parameter of acceleration signals, which are collected by sensors from different positions and running states is calculated, and feature vector is constructed. Fourthly, an analysis is given to the consistency of dual complexity parameter of sensors in the same category and the difference between different categories. At last, the constructed dual complexity feature vector is combined with K-nearest neighbor classical algorithm to identify a rubbing fault and affected locations. For verifying the advantages of presented method, the presented method is compared with other conventional methods ground on the same data. The result indicates that the proposed method can extract the feature information of rotor-stator rubbing fault and affected position more precisely than other contrast methods. In the cross validation corresponding to two experiments in 2 days, the mean recognition accuracy of 10 successive tests can reach to 99.75%.