Abstract

This paper addresses the application of rotor speed signal for the detection and diagnosis of ball bearing faults in rotating electrical machines. Many existing techniques for bearing fault diagnosis (BFD) rely on vibration signals or current signals. However, vibration- or current-based BFD techniques suffer from various challenges that must be addressed. As an alternative, this paper takes the initial step of investigating the efficiency of rotor speed monitoring for BFD. The bearing failure modes are reviewed and their effects on the rotor speed signal are described. Based on this analysis, a novel BFD technique, the rotor speed-based BFD (RSB-BFD) method under variable speed and constant load conditions, is proposed to provide a benefit in terms of cost and simplicity. The proposed RSB-BFD method exploits the absolute value-based principal component analysis (PCA), which improves the performance of classical PCA by using the absolute value of weights and the sum square error. The performance and effectiveness of the RSB-BFD method is demonstrated using an experimental setup with a set of realistic bearing faults in the outer race, inner race, and balls.

Introduction

FAULT diagnosis techniques are becoming more important as more engineering processes are automated, while the manpower needed to operate and supervise processes is reduced. Because rotating electrical machines (REMs) are at the heart of most engineering processes and are designed for tighter margins, there is a growing need for fault diagnosis for the sake of reliability. Different faults may occur in an REM, which can be classified as stator faults, rotor faults, static and dynamic eccentricities, and bearing faults [1]. Based on an IEEE motor reliability study for large motors above 200 hp [2]