Answers are grounded in this paper and cite their sources.
This approach demonstrates enhanced fault identification in noisy environments, indicating improved diagnostic accuracy.
To address the insufficient feature extraction, limited robustness, and low training efficiency of traditional fault diagnosis methods under strong noise, this paper proposes an intelligent bearing fault diagnosis approach that couples the Continuous Wavelet Transform (CWT) with a Spatial Pyramid Pooling and Dynamic-Tanh Swin Transformer (SPDST). First, CWT converts vibration signals into two-dimensional time-frequency scalograms. These images are then fed into a Swin Transformer equipped with a Dynamic-Tanh activation to capture fine-grained local features.Then, Spatial Pyramid Pooling (SPP) aggregates multi-scale feature maps to fuse global and local contextual information, thereby enhancing the representational capacity for fault identification. Experiments on the Case Western Reserve University (CWRU) bearing dataset show that the proposed CWT-SPDST method achieves a diagnostic accuracy of 99.55%; under high-noise conditions of -4 dB and -2 dB, the accuracy remains 91.07% and 93.30%, respectively. Ablation studies further confirm that the coupled architecture improves diagnostic accuracy and training stability while shortening training time.
No takes yet. Share an insight, caveat, or question.
Liu et al. (2026) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: