Rolling bearing fault diagnosis under practical compute constraints and long time series remains challenging in industrial settings. This study presented a compact same-source multi-view framework that combines a multi-scale gated Mamba sequence branch, a residual network with a convolutional block attention module (ResNet-CBAM) image branch, and lightweight low-rank cross-view fusion. Continuous wavelet transform (CWT) images were used as the default image view in the main experiments because they provide the most suitably maintained same-condition representation among the evaluated image views. Under a unified three-seed protocol, the proposed model reached 99.20% on Case Western Reserve University (CWRU) and 99.61% on Southeast University (SEU) in same-condition diagnosis, and attained the strongest cross-condition results within the internal comparison, with 27.39% on CWRU and 30.43% on SEU. The model used about $5.46\times10^5$ parameters and $5.64\times10^8$ multiply–accumulate operations (MACs). These results indicate a balanced trade-off between relative cross-condition generalization and computational cost, while also showing that broader operating-condition diversity, industrial validation, and deployment-oriented central processing unit (CPU) or edge profiling remain necessary.
Citation: Guangpeng Sun, Peiju Chang, Shaojuan Ma, Yucong Li. Compute-conscious multi-view neural framework with efficient state space models and reduced-order cross-attention for rolling bearing fault diagnosis[J]. Electronic Research Archive, 2026, 34(9): 6072-6100. doi: 10.3934/era.2026268
Rolling bearing fault diagnosis under practical compute constraints and long time series remains challenging in industrial settings. This study presented a compact same-source multi-view framework that combines a multi-scale gated Mamba sequence branch, a residual network with a convolutional block attention module (ResNet-CBAM) image branch, and lightweight low-rank cross-view fusion. Continuous wavelet transform (CWT) images were used as the default image view in the main experiments because they provide the most suitably maintained same-condition representation among the evaluated image views. Under a unified three-seed protocol, the proposed model reached 99.20% on Case Western Reserve University (CWRU) and 99.61% on Southeast University (SEU) in same-condition diagnosis, and attained the strongest cross-condition results within the internal comparison, with 27.39% on CWRU and 30.43% on SEU. The model used about $5.46\times10^5$ parameters and $5.64\times10^8$ multiply–accumulate operations (MACs). These results indicate a balanced trade-off between relative cross-condition generalization and computational cost, while also showing that broader operating-condition diversity, industrial validation, and deployment-oriented central processing unit (CPU) or edge profiling remain necessary.
| [1] |
P. Gupta, M. Pradhan, Fault detection analysis in rolling element bearing: A review, Mater. Today Proc., 4 (2017), 2085–2094. https://doi.org/10.1016/j.matpr.2017.02.054 doi: 10.1016/j.matpr.2017.02.054
|
| [2] |
M. Hakim, A. A. B. Omran, A. N. Ahmed, M. Al-Waily, A. Abdellatif, A systematic review of rolling bearing fault diagnoses based on deep learning and transfer learning: Taxonomy, overview, application, open challenges, weaknesses and recommendations, Ain Shams Eng. J., 14 (2023), 101945. https://doi.org/10.1016/j.asej.2022.101945 doi: 10.1016/j.asej.2022.101945
|
| [3] |
B. Peng, Y. Bi, B. Xue, M. Zhang, S. Wan, A survey on fault diagnosis of rolling bearings, Algorithms, 15 (2022), 347. https://doi.org/10.3390/a15100347 doi: 10.3390/a15100347
|
| [4] |
A. U. Rehman, W. Jiao, Y. Jiang, J. Wei, M. Sohaib, J. Sun, et al., Deep learning in industrial machinery: A critical review of bearing fault classification methods, Appl. Soft Comput., 171 (2025), 112785. https://doi.org/10.1016/j.asoc.2025.112785 doi: 10.1016/j.asoc.2025.112785
|
| [5] |
W. Zhang, C. Li, G. Peng, Y. Chen, Z. Zhang, A deep convolutional neural network with new training methods for bearing fault diagnosis under noisy environment and different working load, Mech. Syst. Signal Process., 100 (2018), 439–453. https://doi.org/10.1016/j.ymssp.2017.06.022 doi: 10.1016/j.ymssp.2017.06.022
|
| [6] |
L. Wen, X. Li, L. Gao, Y. Zhang, A new convolutional neural network-based data-driven fault diagnosis method, IEEE Trans. Ind. Electron., 65 (2018), 5990–5998. https://doi.org/10.1109/TIE.2017.2774777 doi: 10.1109/TIE.2017.2774777
|
| [7] |
X. Dai, K. Yi, F. Wang, C. Cai, W. Tang, Bearing fault diagnosis based on POA-VMD with GADF-Swin transformer transfer learning network, Measurement, 238 (2024), 115328. https://doi.org/10.1016/j.measurement.2024.115328 doi: 10.1016/j.measurement.2024.115328
|
| [8] |
P. Bao, W. Yi, Y. Zhu, Y. Shen, B. X. Chai, STHFD: Spatial–temporal hypergraph-based model for aero-engine bearing fault diagnosis, Aerospace, 12 (2025), 612. https://doi.org/10.3390/aerospace12070612 doi: 10.3390/aerospace12070612
|
| [9] |
Y. Li, K. H. Bwar, R. Chai, K. M. Tse, B. X. Chai, Adaptive multi-view hypergraph learning for cross-condition bearing fault diagnosis, Mach. Learn. Knowl. Extr., 7 (2025), 147. https://doi.org/10.3390/make7040147 doi: 10.3390/make7040147
|
| [10] |
P. Kumar, I. Raouf, J. Song, H. S. Kim, Multi-size wide kernel convolutional neural network for bearing fault diagnosis, Adv. Eng. Software, 198 (2024), 103799. https://doi.org/10.1016/j.advengsoft.2024.103799 doi: 10.1016/j.advengsoft.2024.103799
|
| [11] | A. Vaswani, N. Shazeer, N. Parmar, J. Uszkoreit, L. Jones, A. N. Gomez, et al., Attention is all you need, in Advances in Neural Information Processing Systems, 30 (2017). |
| [12] | A. Gu, K. Goel, C. Re, Efficiently modeling long sequences with structured state spaces, in International Conference on Learning Representations, (2022), 1–27. |
| [13] | A. Gu, T. Dao, Mamba: Linear-time sequence modeling with selective state spaces, in First Conference on Language Modeling, (2024), 1–32. |
| [14] |
P. Wang, Y. Song, X. Wang, Q. Xiang, MD-BiMamba: An aero-engine inter-shaft bearing fault diagnosis method based on Mamba with modal decomposition and bidirectional features fusion strategy, Measurement, 242 (2025), 115870. https://doi.org/10.1016/j.measurement.2024.115870 doi: 10.1016/j.measurement.2024.115870
|
| [15] | S. Woo, J. Park, J. Y. Lee, I. S. Kweon, CBAM: Convolutional block attention module, in Proceedings of the European Conference on Computer Vision (ECCV), (2018), 3–19. https://doi.org/10.1007/978-3-030-01234-2_1 |
| [16] | E. J. Hu, Y. Shen, P. Wallis, Z. Allen-Zhu, Y. Li, S. Wang, et al., Lora: Low-rank adaptation of large language models, in International Conference on Learning Representations, 1 (2022), 3. |
| [17] | J. Arevalo, T. Solorio, M. Montes-y-Gómez, F. A. González, Gated multimodal units for information fusion, preprint, arXiv: 1702.01992. |
| [18] | Case Western Reserve University Bearing Data Center, Welcome to the Case Western Reserve University Bearing Data Center Website, 2026. Available from: https://engineering.case.edu/bearingdatacenter/welcome. |
| [19] |
S. Shao, S. M. McAleer, R. Yan, P. Baldi, Highly accurate machine fault diagnosis using deep transfer learning, IEEE Trans. Ind. Inf., 15 (2019), 2446–2455. https://doi.org/10.1109/TII.2018.2864759 doi: 10.1109/TII.2018.2864759
|
Figures(17) / Tables(13)
Guangpeng Sun, Peiju Chang, Shaojuan Ma, Yucong Li. Compute-conscious multi-view neural framework with efficient state space models and reduced-order cross-attention for rolling bearing fault diagnosis[J]. Electronic Research Archive, 2026, 34(9): 6072-6100. doi: 10.3934/era.2026268
DownLoad: