Chaotic Analysis and Feature Extraction of Vibration Signals From Power Circuit Breakers

Abstract

Vibration signals generated during circuit breaker (CB) closing or opening operations contain rich information of operating mechanism (OM), transmission mechanism, and interrupter(s). Effective feature extraction method can mine the information as a criterion for diagnosis and maintenance. However, due to the very complicated mechanical system and extremely short operation time of CB, the vibration signal is highly nonlinear and non-stationary, which makes it very difficult to precisely extract effective features for machinery fault diagnosis. Chaotic nonlinear dynamic technique provides a new way to study the complex vibration signals. In this paper, phase space reconstruction is utilized to study the effect of faults on the chaotic attractor (the trajectories in multidimensional phase space) behavior. The power spectrum and Lyapunov exponent proof the existence of chaos in the CB's vibration signals. The invariant measures and ergodic quantities such as the largest Lyapunov exponent (LLE), correlation dimension (CD) and Kolmogorov entropy (KE) which can be estimated on the reconstructed attractor are presented as a set of new features for fault diagnosis. The estimation of these features was tested on the experimental data sets recorded from a 12-kV, 1250-A vacuum CB and a 252-kV, 4000-A SF6 CB, respectively.