The research presented in this commentary and the associated publications portfolio addresses the challenges faced by the health management of rolling element bearings (REBs) in industrial rotating machines under multi-source interference scenarios. It provides new and robust condition monitoring and fault diagnosis approaches for improving the safety, efficiency and health management level of rotating machinery systems by analysing and mining REB-related health information in measured vibration signals. Under the main theme, the research is further divided into two sub-themes: condition monitoring of REBs and fault diagnosis of REBs. They respectively deal with the issues of health status monitoring and incipient fault diagnosis of REBs during their entire life cycle under complex operating scenarios. For effective condition monitoring of REBs, this research systematically investigates and develops health indicator-based methods through sparse statistical theories. Significant innovations and contributions include two new generalized mathematical frameworks for sparse statistical indicators (including sparsity measures), four new and robust families of sparsity measures (including the power functionbased Gini index I, II and III and the fully generalized Gini index) and the method for calculating condition monitoring baselines of machine health indicators by theoretically establishing the probability distributions under normal machine conditions. The developed methodology enables the creation of new bearing health indicators that are robust to multi-source interference noises. The proposed sparsity measures can efficiently quantify periodic transient fault features and characterize the degradation state of REBs. The established health indicator baseline provides an important reference for abnormal state detection. These theories and methods have high potential to markedly advance the condition monitoring level of REBs. For accurate fault diagnosis of REBs, this research develops new fault feature enhancement methods and spectrum methods through non-stationary signal processing theories. Important innovations and contributions include two theoretical frameworks for generalized statistical indicator-guided blind deconvolution methods, new blind deconvolution methods guided by robust sparsity measures for enhancing REB fault features, and generalized envelope spectrum, product envelope spectrum and weighted combined envelope spectrum for accurate diagnosis of REB faults. These provide new theories and methods for REB defect diagnosis under complicated interferences. Application in train axle-box bearings showcases that the developed approaches can efficiently extract bearing fault features submerged in noisy vibration measurements and diagnose various bearing defects. These innovative methods have significant potential to enhance the fault diagnosis capability and accuracy of REBs under complex operating conditions.
| Date of Award | 21 Nov 2025 |
|---|
| Original language | English |
|---|
| Sponsors | China Scholarship Council |
|---|
| Supervisor | Fengshou Gu (Main Supervisor) & Adam Bevan (Co-Supervisor) |
|---|