Entropy-based fault detection approach for motor vibration signals under accelerated aging process

The purpose of this study is to analyze motor vibration signals due to the bearing fault, which is artificially generated by aging process. Vibration signal data recorded by the experimental setup has been conditioned by a high-pass filter (Butterworth type) to reach the regarding frequency components of the bearing failure. Spectral analysis has been applied to realize the degradation on the bearing and the power spectral density figures revealed that the magnitudes of frequency components between 1.5-4 kHz bandwidth increased after every aging cycle. Vibration signals were investigated statistically by examining four main statistical parameters: mean value, standard deviation, skewness and kurtosis. Evaluation of these parameters indicated that significant variance occurred on standard deviation. At this point Shannon entropy became an approach to analyze the variance on the standard deviation. The probability of the aging cycles has been defined as a function of standard deviation values for each aging cycle. Entropy definition, which is a function of probability, determines the uncertainty level on the data and it has been examined to identify the effect of the aging progress on the bearing by examining the transferred entropy amount between aging cycles.