A Comprehensive Investigation of Fault Signatures and Spectrum Analysis of Vibration Signals in Distributed Bearing Faults

Abstract

In industry, distributed bearing faults are prevalent and stem from factors like inadequate lubrication, contamination, and electrical erosion, to name a few. These faults exhibit complex and unpredictable vibration patterns in contrast to localized ones including single and multi-point defects. The literature extensively discusses localized faults, particularly single-point defects. However, a thorough examination of distributed bearing faults is essential for gaining insight into the intricate nature of real bearing faults. This analysis is pivotal for understanding the various factors that impact vibration behavior and accurately interpreting them. To achieve this, the prevailing distributed bearing fault patterns are synthetically generated in a controlled laboratory environment. Subsequently, a comprehensive dataset encompassing lubrication, erosion, contamination, and flaking issues is proposed. For this purpose, two induction motor setups (3 and 10hp) are developed and operated at 10 different speeds and 5 varying load levels to capture vibration signals using a triaxial accelerometer. The results highlight the key characteristics of distributed faults introduced by random defects spread across bearing surfaces. These are absent in localized faults like single and multi-point faults. It is also shown that the unique harmonic frequencies are exclusively present in distributed bearing faults which helps to distinguish faults and boost the diagnosis algorithm performances compared to models trained solely by localized fault data.