Self-running Fault Diagnosis Method for Rolling Element Bearing

Abstract

While the machine is running, damaged components of the bearing trigger vibrations in the structure of the machine when it contacts other surfaces. These components appear at specific frequencies dictated by the geometry of the bearing and its rotation frequency. An autonomous fault detection method is therefore needed to improve the performance and the reliability of the mechanical system. This article aims to present an autonomous bearing fault detection process. This process takes into account the slip phenomenon by calculating a normalized indicator related to the existence of a bearing fault in a narrow band centered at the theoretical frequency. The latter is calculated from the geometry of the bearing, after preprocessing steps in order to equalize the baseline spectrum and to set an appropriate statistical threshold. An application on real data from the IMS database will be held at the end in order to detect and classify mechanical faults.