Application of multi-scale fuzzy entropy for roller bearing fault detection and fault classification based on VPMCD

Roller bearing is an integral component in various types of rotating machinery. Bearing fault detection is very important to prevent failure, increase safety, reduce production idle time and decrease maintenance cost. In this paper, Multi-scale Fuzzy Entropy(MFE) is used for fault detection of roller bearing and Variable predictive model-based class discrimination (VPMCD) is used as multi-fault classifier. Fuzzy entropy is calculated for complexity measure of time series constructed from motor vibration signal. Usually, as vibration signals tend to be non-linear, fuzzy entropy calculated for single scale may not contain all the fault information. Hence it is essential to calculate entropy for multiple scales. As a multi-fault classifier VPMCD has been used to classify bearing faults. Fault features created using MFE are used as an input for VPMCD classifier. VPMCD is applied here for roller bearing fault classification. The effect of motor rotational speed on the MFE values is investigated. Experimental analysis is conducted to evaluate performance of this method. The results of this experiment indicate that MFE and VPMCD together can achieve good accuracy and reliability in bearing fault detection and classification.