A Robust Deep Learning System for Motor Bearing Fault Detection: Leveraging Multiple Learning Strategies and a Novel Double Loss Function

Abstract

Motor bearing fault detection (MBFD) is vital for ensuring the reliability and efficiency of industrial machinery. Identifying faults early can prevent system breakdowns, reduce maintenance costs, and minimize downtime. This paper presents an advanced MBFD system using deep learning, integrating multiple training approaches: supervised, semi-supervised, and unsupervised learning to improve fault classification accuracy. A novel double-loss function further enhances the model’s performance by refining feature extraction from vibration signals. Our approach is rigorously tested on well-known datasets: the American Society for Mechanical Failure Prevention Technology (MFPT), Case Western Reserve University Bearing Data Center (CWRU), and Paderborn University's Condition Monitoring of Bearing Damage in Electromechanical Drive Systems (PU). Results indicate that the proposed deep learning method outperforms traditional machine learning models, achieving high accuracy across all datasets. These findings underline the potential for applying deep learning in MBFD, providing a robust solution for predictive maintenance in industrial settings and supporting proactive management of machinery health.