Fault Diagnosis of Journal Bearings Based on Artificial Neural Networks and Measurements of Bearing Performance Characteristics

摘要

Two of the most common defects in rotating systems are abnormal wear of the bearing bushing and bearing misalignment. The present paper introduces a new fault diagnosis model that uses artificial neural networks (ANN) in order to identify the increase of wear depth and/or the increment of the misalignment angle. Reynolds equation is solved by FEM and provides data about bearing wear and misalignment. The proposed model uses eccentricity, altitude angle and minimum film thickness and feeds with their values an ANN that is trained in order to provide reliable identification of the variation of each defect. The accuracy of the proposed model is demonstrated - for several misalignment angles, worn depths and L/D ratios - for a worn/misaligned rotor bearing and its applicability as a real-time condition monitoring system is discussed.