Early Classifications of Bearing Faults Using Hidden Markov Models, Gaussian Mixture Models, Mel-frequency Cepstral Coefficients and Fractals

Most rotating-machines failures are often linked to bearing failures. Proper condition monitoring on bearings is therefore essential to reduce the duration of machine down-times. This paper introduces feature extraction methodologies that can facilitate early detection of bearing faults. The time-domain vibration signals of a rotating machine with normal and defective bearings are processed for feature extraction. Both linear and non-linear features are extracted using Multi-Scale Fractal Dimension (MFD), Mel frequency Cepstral Coefficients and kurtosis. The extracted features are then used to classify faults using Gaussian Mixture Models (GMM) and hidden Markov Models (HMM). Results demonstrate that HMM outperforms GMM in classification of bearing faults. However, the disadvantage of HMM is that it is computationally expensive to train compared to GMM.