Synapse
⌘+K
Synapse
PulseExploreClubsResearchersJournals
Instagram
HomeClubsExplore
March 21, 2026Transactions of the Canadian Society for Mechanical Engineering

A Bearing Fault Diagnosis Method Based on Continuous Wavelet Transform and an SPDSwin Transformer

View Full Paper
Ask AI
Bookmark
Share

Authors

TLTaisu LiuXJXiaoyi JingPLPeitong Liu

Discussion

Loading...

Member takes

Overview

This approach demonstrates enhanced fault identification in noisy environments, indicating improved diagnostic accuracy.

Key Points

  • The aim is to develop a robust method for diagnosing bearing faults by overcoming limitations of traditional techniques.
  • Utilized Continuous Wavelet Transform (CWT) to convert vibration signals into time-frequency scalograms.
  • Implemented a Swin Transformer with Dynamic-Tanh activation for fine-grained local feature extraction.
  • Employed Spatial Pyramid Pooling (SPP) to merge multi-scale feature maps for enhanced context representation.
  • Conducted experiments on the Case Western Reserve University (CWRU) bearing dataset.
  • Achieved a diagnostic accuracy of 99.55% with the CWT-SPDST method.
  • Maintained accuracies of 91.07% and 93.30% under high-noise conditions of -4 dB and -2 dB, respectively.
  • Ablation studies confirmed improvements in diagnostic accuracy and training stability.

Cite This Study

Liu et al. (2026) studied this question.

synapsesocial.com/papers/69be36bf6e48c4981c675f28https://doi.org/10.1139/tcsme-2025-0202
View Full Paper
Ask AI
Bookmark
Share