Systematic Review on Fault Diagnosis on Rolling-Element Bearing

To maintain machinery operations smoothly, Rolling-Element Bearings (REBs) are utilized so that the entire equipment’s safety is ensured. Sometimes, the safety of the machinery is still questioned because REBs are prone to faults. Even a small issue in REBs may break down the entire system unexpectedly. This may result in the rise of maintenance and downtime cost which significantly increases the operational cost of the machinery. Thus, it is important to recognize bearing faults in machinery and this leads to the improvement in the system’s reliability by avoiding unexpected accidents.

Different fault diagnosis methods have been introduced so far. This paper attempts to make a survey on different fault diagnosis works and their performance. This review paper provides a comprehensive coverage of various signal processing techniques such as EMD, EEMD, VMD-EMD, PEEWMD, FAEMD, APSFDM, CEEMD, SVD, GBMD, IENEMD-ATD, EEMD-MPE-BP, FPSEWT, TKEO, Adaptive spectral kurtosis technology, AMOMEDA, DAMM-SUM, APSFDM, EFD, SEAEFD, CEEMD-ATD, FDM, etc. Furthermore, Machine Learning algorithms such as KNN, ANN, ANN-DA, LSSVM, MSPC, BSE, DFAE, ANN-KNN, etc. have been examined. Meanwhile, Deep Learning algorithms like CNN, DNN, DCNN, Deep ResNet Structure, PCNN, Multi-task CNN, TL-SPF, CNN-ResNet Structure, PNN-FIFD, VSI-CNN, CNN-GAF, and etc., have also been discussed. Moreover, Reinforcement Learning algorithms such as Deep Reinforcement Learning, Neural Network with RL, Deep feature enhanced RL, DL-RL, PPO, ML-TRL, DRTCNN, RL-NAS, and DEPDRL have been analysed.

This survey analysed existing works on fault diagnosis on REB in three different perspectives. They are (1) Analysis on several REB fault diagnosis techniques, (2) Analysis on Empirical Mode Decomposition (EMD) based REB fault diagnosis techniques, and (3) Analysis on performance measures of existing REB fault diagnosis techniques.

Finally, this survey highlights the limitations and challenges in the existing REB fault diagnosis techniques. Researchers aiming to develop an efficient method for early and effective diagnosis, can find useful information and future directions in this survey.