Deep Reinforcement Learning-Based Early Fault Warning for Rolling Bearings in Weak Data Scenarios

Mingyu Yue

doi:10.54097/szffsx45

Authors

Mingyu Yue School of Mechanical Engineering, University of Leeds, Woodhouse Lane, Leeds, UK

DOI:

https://doi.org/10.54097/szffsx45

Keywords:

Deep Reinforcement Learning, Fault Diagnosis, Rolling Bearing, Weak Data.

Abstract

Rolling bearings are vital components in rotating machinery, and early detection of bearing faults is essential to avoid unplanned downtime, costly repairs, and safety hazards. In industrial environments, early fault signatures are often weak, noisy, and rare-creating a “weak-data” scenario marked by severe class imbalance, high annotation cost, and distribution shifts that undermine conventional deep learning methods. Deep reinforcement learning (DRL) offers a promising alternative by framing fault warning as a Markov decision process and leveraging sequential interaction and reward signals to learn from limited and imbalanced data. This paper presents a systematic the paper of DRL-based approaches for early bearing-fault warning, categorizing and analyzing key enabling techniques: reward shaping and balanced experience replay, data augmentation and generative synthesis, integration of deep feature extractors (CNNs/LSTMs), meta-reinforcement learning for few-shot adaptation, transfer and domain adaptation, exploration strategies, and partially observable modelling. The paper further discusses practical challenges—including reward design, sample efficiency, training stability, interpretability, and computational constraints for edge deployment—and outlines promising directions such as generative-model augmentation, attention and graph-based architecture, multi-sensor fusion, and lightweight distributed DRL. This paper aims to guide the development of data-efficient, robust DRL solutions for industrial predictive maintenance.

Downloads

Download data is not yet available.

References

[1] Çakır, E., Dumond, P. (2025) A comparative analysis of reinforcement learning and conventional deep learning approaches for bearing fault diagnosis. arXiv preprint arXiv: 2506.19929.

[2] Wang, S., Li, J., Xu, X., Wu, R., Qiu, Y., Chen, X., Qiao, Z. (2025) Optimization of a coupled neuron model based on deep reinforcement learning and application of the model in bearing fault diagnosis. Sensors, 25 (12), 3654.

[3] Pham, M. T., Kim, J. M., Kim, C. H. (2020) Deep learning-based bearing fault diagnosis method for embedded systems. Sensors, 20 (23), 6886.

[4] Wang, R., Jiang, H., Zhu, K., Wang, Y., Liu, C. (2022) A deep feature enhanced reinforcement learning method for rolling bearing fault diagnosis. Advanced Engineering Informatics, 54, 101750.

[5] Li, Y., Wang, Y., Zhao, X., Chen, Z. (2024) A deep reinforcement learning-based intelligent fault diagnosis framework for rolling bearings under imbalanced datasets. Control Engineering Practice, 145, 105845.

[6] Ding, J., Li, X., Gudivada, V. N. (2017) Augmentation and evaluation of training data for deep learning. In Proceedings of the IEEE International Conference on Big Data, pp. 2603 – 2611. IEEE.

[7] Lin, L., Wang, B., Qi, J., Wang, D., Huang, N. (2019) Bearing fault diagnosis considering the effect of imbalance training sample. Entropy, 21 (4), 386.

[8] Wang, S., Wang, D., Kong, D., Wang, J., Li, W., Zhou, S. (2020) Few-shot rolling bearing fault diagnosis with metric-based meta learning. Sensors, 20 (22), 6437.

[9] Hoang, D. T., Kang, H. J. (2019) A survey on deep learning-based bearing fault diagnosis. Neurocomputing, 335, 327 – 335.

[10] Kang, Y., Chen, G., Wang, H., Pan, W., Wei, X. (2023) A new dual-input deep anomaly detection method for early faults warning of rolling bearings. Sensors, 23 (18), 8013.

[11] Shaalan, A., Mefteh, W., Frihida, A. M. (2024) Review on deep learning classifiers for faults diagnosis of rotating industrial machinery. Service Oriented Computing and Applications, 18 (4), 361 – 379.

[12] Wang, Z., Xu, Z., Cai, C., Wang, X., Xu, J., Shi, K., Zhong, X., Liao, Z., Li, Q. (2024) Rolling bearing fault diagnosis method using time-frequency information integration and multi-scale TransFusion network. Knowledge-Based Systems, 284, 111344.

[13] Wang, J., Zhao, Y., Wang, W., Wu, Z. (2024) Improving bearing fault diagnosis method based on the fusion of time–frequency diagram and a novel vision transformer. Journal of Supercomputing, 81 (1).

[14] Zhao, S., Xing, Y., Chen, X., Ou, Y., Zhang, W., Lin, B. (2020) A power metering pipeline fault warning method based on deep learning. In IOP Conference Series: Materials Science and Engineering, 740 (1), 012111. IOP Publishing.

[15] Zhang, A., Li, S., Cui, Y., Yang, W., Dong, R., Hu, J. (2019) Limited data rolling bearing fault diagnosis with few-shot learning. IEEE Access, 7, 110895 – 110904.

[16] Zhu, W., Yu, C., Zhang, Q. (2022) Causal deep reinforcement learning using observational data. arXiv preprint arXiv: 2211.15355.

[17] Li, A., Luo, Y., He, Y., Cheng, Z., Wang, T. (2021) Fault diagnosis method of motor bearing based on deep transfer learning. In Proceedings of the IEEE China International Youth Conference on Electrical Engineering (CIYCEE), pp. 1 – 5. IEEE.

[18] Ren, H., Liu, S., Qiu, B., Guo, H., Zhao, D. (2024) A novel intelligent fault diagnosis method for bearings based on a multi-head self-attention convolutional neural network. Artificial Intelligence for Engineering Design, Analysis and Manufacturing, 38, e19.

[19] Chen, G., Qi, J., Gao, Y., Zhu, X., Dong, Z., Sun, Y. (2024) DGTRL: Deep graph transfer reinforcement learning method based on fusion of knowledge and data. Information Sciences, 658, 120019.

[20] Cao, H., Feng, F., Huo, J., Yang, S., Fang, M., Yang, T., Gao, Y. (2025) Model-based offline reinforcement learning with adversarial data augmentation. arXiv preprint arXiv: 2503.20285.

[21] Jiang, F., Liang, Q., Wu, Z., Kuang, Y., Zhang, S., Liang, J. (2024) A zero-cost unsupervised transfer method based on non-vibration signals fusion for ball screw fault diagnosis. Knowledge-Based Systems, 288, 111475.