Ball Bearing Fault Diagnosis Using Mutual Information and Walsh–Hadamard Transform

Abstract

Bearing failure may result in the breakdown of machinery or possibly damage the human being operating the machinery. It is therefore necessary to diagnose bearing faults at early stage. Vibration signals are contaminated with noise. To minimize the effect of noise, Walsh–Hadamard transform is used for feature extraction. Statistical features from coefficients are calculated to form feature vector. Mutual information a type of feature ranking method is used to select most informative feature and subsequently to reduce size of feature vector. Coefficients from healthy and faulty bearings at different rotational speeds were calculated from signals. Statistical features are calculated from the acquired signals with different speeds. Two different classifiers like support vector machine and artificial neural network have been used for finding the accuracy. Result shows the methodology adopted is effective to diagnose various bearing faults.