AI-Based Predictive Maintenance Techniques for Bearings: Emerging Research Trends and Engineering Solutions

Bearings are essential components across a wide range of mechanical systems, facilitating smooth motion with minimal energy dissipation. Therefore, their maintenance and health monitoring are critical for ensuring system reliability and operational safety. With the increasing integration of advanced digital technologies into industrial processes, condition monitoring systems have become indispensable for the early detection of bearing wear and damage, allowing interconnected machines to gather vast amounts of data and convert it into actionable insights. While extensive datasets offer significant opportunities for analysis, they are often subject to noise, biases, and inconsistencies that complicate the development of accurate physical models, particularly in complex, dynamic systems. Machine learning has emerged as a powerful tool for extracting meaningful insights from large-scale datasets, enabling automated detection, classification, and prediction of bearing faults without the need for explicit programming, thereby reducing the necessity for human intervention. This thesis provides a thorough review of peer-reviewed scientific literature alongside currently available engineering solutions in the field of intelligent bearing fault diagnosis. In this context, it critically examines existing technologies in relation to the established body of knowledge, identifying and analysing any gaps that may exist between these domains. By investigating potential factors contributing to such disparities, this work offers informed projections for future developments in intelligent bearing fault diagnosis.