Efficient fault diagnosis of ball bearing using ReliefF and Random Forest classifier

Abstract

The present study focuses on identifying various faults present in ball bearing from the measured vibration signal. Features such as kurtosis, skewness, mean, and root mean square, and complexity measure such as Shannon Entropy are calculated from time domain and Discrete Wavelet Transform. To select the best wavelet function, Maximum Energy to Shannon Entropy ratio criterion is used. Information Gain and ReliefF ranking methods are used to assess the quality of features and features are ranked based on the weight gain obtained from the methods used. Support Vector Machine and Random Forest classifier are selected to identify bearing faults and comparison is made to diagnose faults on the ranked feature set. Experiments are conducted on Case Western Reserve University bearing data sets. Results show that ranking method is useful for identifying best feature set and to improve classification accuracy simultaneously. Cross-validation efficiency of 98.38% is obtained when ReliefF is used with Random Forest.