Vibration based diagnostics for bearing faults and gear wear

Abstract

This thesis investigates vibration based machine condition monitoring and consists of two parts: bearing fault diagnosis and gear wear monitoring. In the first part, a signal processing method to diagnose localised bearing faults in the presence of periodic impulsive noise has been developed and tested on signals measured from two rigs. The method developed for bearing diagnosis aims at selecting suitable frequency bands that exclude those impulsive noises and retain the feature of impulsiveness of bearing fault signals. This is realised by using a parameter considering the energy and energy distribution of envelope signals in the frequency domain, and its efficiency is demonstrated by comparison with the Fast Kurtogram and Protrugram. The second part presents a newly developed vibration indicator for the monitoring of gear wear. Even though gear vibrations are closely related with tooth wear, a vibration based wear monitoring method has not been well established. In this work, an averaged logarithmic ratio (ALR) is calculated from time synchronous averaged gear signals to evaluate the effects of tooth wear on gear transmissions. The indicator ALR, calculated with a fixed reference, can be used as a wear severity index, while an alternative version of the indicator, mALR, is calculated with a moving reference and can be used to show the evolution of gear transmission features. In addition, these indicators can also be used as a general parameter for gearbox condition monitoring. Based on the proposed vibration indicators, a framework is established to integrate vibration and wear particle analysis for gear condition monitoring. In the framework, the proposed vibration indicators are first used for fault detection and classification. The faults are then categorised into two groups, structure- and wear-related faults, which dictates how the analysis methods are applied and ensures that wear particle analysis is only used when strictly required. The feasibility and efficiency of this framework are demonstrated by application to three laboratory gear tests, with promising results.