Fault detection and classification of motor bearings under multiple operating conditions

Abstract

The article presents a framework for fault detection and classification to monitor the condition of motor bearings under multiple operating conditions. The condition monitoring of motor bearings is crucial for failure prevention, as bearings are prone to failure in challenging working environments. Intelligent fault diagnosis methods driven by deep learning and model-based approaches have been widely adopted to address these concerns. However, accurately diagnosing bearing faults across varying conditions and identifying multiple fault types remains challenging. The article proposes a multitask fault detection and classification approach for health monitoring using the HUST motor bearings dataset. The evaluation using HUST motor bearing datasets demonstrates robust performance across diverse operating conditions and in the presence of multiple faults. The HUST dataset is valuable for bearing fault diagnosis due to its diverse operating conditions and inclusion of multiple fault types, offering a realistic representation of fault scenarios derived from real bearing experiments. This methodology enhances the safety and reliability of mechanical equipment, with adaptability to various rotating scenarios. • Multitask learning approach for fault detection and classification for HUST motor bearings. • Designed a sensor-based diagnostic approach that preserve control loop integrity. • Incorporated domain knowledge to create adaptable diagnostic schemes for diverse conditions. • Validated using the diverse HUST bearing dataset, reflecting realistic fault scenarios.

Topics & Concepts