An intelligent fault diagnosis method of high voltage circuit breaker based on improved EMD energy entropy and multi-class support vector machine

Targeting the characteristics of machinery vibration signals of high voltage circuit breaker (CB), a new method based on improved empirical mode decomposition (EMD) energy entropy and multi-class support vector machine (MSVM) to diagnose fault for high voltage CB is proposed. In the fault diagnosis for the high voltage CB, the feature extraction based on improved EMD energy entropy is detailedly analyzed. A new multi-layered classification of SVM named 'one against others' algorithm approach is proposed and applied to machinery fault diagnosis of high voltage CB. The extracted features are applied to MSVM for estimating fault type. Compared with back-propagation network (BPN), the test results of MSVM demonstrate that the applying of improved EMD energy entropy to vibration signals is superior to that based on wavelet packet analysis (WPT) and hence estimating fault type on machinery condition of high voltage CB accurately and quickly.