Grease Contamination Detection in the Rolling Element Bearing Using Deep Learning Technique

Abstract

Vibration Analysis is one of the most effective methods used for the condition monitoring of rolling element bearings. The early failure of bearing is mainly due to the presence of solid particles in the grease lubricants. The condition of lubrication in the bearing is an essential parameter to meet the various demanding conditions of the system. This paper aims to analyze the effect of lubricant contamination by solid particles on the dynamic behavior of rolling bearing and to classify them using a support vector machine (SVM) and deep learning algorithm. Experimental tests have been performed with 50 and 100 mg of sand dust particles added to the ball bearings to contaminate the grease lubricant at full load conditions. Vibration signals were analyzed in terms of RMS, kurtosis, skewness, and peak to peak for fault type classification using SVM. In deep learning, the raw vibration signals are converted into a spectrogram image and fed to convolution neural networks (CNN) for fault classification. The results indicate that both SVM and deep learning techniques are effective for fault classification under the influence of lubricant contamination.