Simultaneous bearing fault diagnosis and severity detection using a LAMSTAR network-based approach

Bearings are one of the most important components in many industrial machines. Effective bearing fault diagnosis and severity detection are critical for keeping the machines operate normally and safe. In this study, the problem of simultaneous bearing fault diagnosis and severity detection with deep learning is addressed. Existing solutions developed using deep learning rely on fault feature extraction using complicated signal processing techniques. They perform bearing fault diagnosis and severity detection separately and normally require extensive supervised fine tuning. This study presents an effective deep learning-based solution using a large memory storage and retrieval (LAMSTAR) neural network. The developed approach can automatically extract self-learned fault features and perform bearing fault diagnosis and severity detection simultaneously. The structure of the LAMSTAR network is determined by optimally selecting the sliding box size of the input time-frequency matrix. The effectiveness of the proposed approach is validated using data collected from rolling element bearing tests.