SVM broken bar detection based on analysis of phase current

This paper presents a method for broken bar detection based on motor current signature analysis of one phase induction motor current. The broken bar classifier is implemented by support vector machine due to its good robustness and generalization performances. The proposed method is based on frequency and time domain analysis of one phase steady state and transient current signal, which is only physical value to be measured. Broken bar related features are extracted using Fourier transformation, Welch's PSD estimation and statistical parameters: variance, skewness and kurtosis. It is shown that proposed method successfully detects faulty condition in various operational situations and sub-nominal load conditions. It is a reliable online method, which has high robustness to load variations and operating conditions. The proposed method is sufficiently accurate and fast which gives him capability for real-time online applications in industrial drives.