Fault Identification in Bearing using MODWPT and Chi-Square

Bearing fault identification is the most critical aspect in the predictive maintenance of any rotating machine. The reliability and efficiency of an automated system highly depend on the production line and health of each machine. The bearing component, which erasures the smooth operation and is responsible for reducing friction needs special care in maintenance and early detection of bearing faults helps in reducing the maintenance cost. The proposed methodology explores the comparative study of Maximum Overlap Discrete Wavelet Transform (MODWT) and Maximum Overlap Discrete Wavelet Packet Transform (MODWPT) followed by the feature selection using the Chi-Square test. The results further explore the comparative analysis of various machine learning classifiers and in conclusion, the combined approach of the MODWPT with Supervised Vector Machine (SVM) outperformed other tested combinations for the Case Western Reserve University (CWRU) Dataset. The proposed methodology provides a computationally efficient and high-precision technique to detect the bearing fault contributing to the advancement of intelligent fault diagnosis systems.