A systematic literature review of deep learning for vibration-based fault diagnosis of critical rotating machinery: Limitations and challenges

Over the last decade, thousands of papers on machine-learning for diagnosing faults in rotating machinery through vibration signals have been published. Specifically, deep learning, coupled with domain adaptation, has been replacing traditional physical and signal-processing techniques. This study systematically reviews the literature on deep learning for fault diagnosis in rotating machinery, focusing on real-world cases. The review points out current limitations in systems with several examples of labeled or unlabeled faulty signals. The study concludes by suggesting directions in which deep learning can be successfully implemented, contributing to the enhancement of current diagnostic capabilities.