Faults Detection Using Gaussian Mixture Models, Mel-Frequency Cepstral Coefficients and Kurtosis

Most machines failures can be associated with mechanical failures on bearing failures. This paper proposes a novel approach to detect and classify three types of common faults in rolling element bearings. The approach proposed here makes use Gaussian mixture model to classify, Mel-frequency cepstral coefficients (MFCC) and kurtosis are extracted from the bearing vibration signal and are used as features. A classification rate of 95% is obtained when using the MFCC features only while a classification rate improves to 99% when Kurtosis features are added to the MFCC..