Fault Diagnosis of Rolling Element Bearings on Low-Cost and Scalable IIoT Platform

Abstract

Across industries, rolling element bearings have been found to be among the largest contributors to machine downtime and the costliest to replace. Thus, a reduction in unexpected bearing failures will lead to an increase in machine availability, a reduction in maintenance and downtime costs, and an improvement in customer satisfaction. Over the past few decades, deep learning has been recognized as a useful tool for fault diagnosis of rolling element bearings. Although existing deep learning approaches can achieve satisfactory accuracy in bearing fault diagnosis, most of these approaches rely exclusively on data and do not incorporate physical knowledge into the learning and prediction processes. To address this challenge, this study develops a deep learningbased fault diagnosis approach by proposing a novel physics-based feature weighting (PFW) technique. The proposed PFW technique leverages the fault characteristic frequencies of a bearing to weigh the vibration features based on the amount of faultrelated information that they are expected to carry. The weight learning component of PFW comes from the ability to learn (or optimize) the parameters of the weighting functions while training a deep learning model. The performance of the developed approach with the proposed PFW technique is compared with that of traditional convolutional neural network.