Scholarlink

Bearing faults are a primary cause of machinery failures, underscoring the need for accurate and robust diagnostic methods, particularly under noisy conditions. This paper introduces the SCSSA-VMD-CNN-Transformer (SVCT) framework, which integrates the Sine Cosine Sparrow Search Algorithm (SCSSA)-optimized Variational Mode Decomposition (VMD) with a CNN-Transformer architecture for enhanced bearing fault diagnosis. SCSSA adaptively optimizes VMD parameters to decompose noisy vibration signals into intrinsic mode functions (IMFs), facilitating effective feature extraction and classification by the CNN-Transformer model. Experiments conducted on the Case Western Reserve University (CWRU) bearing dataset, across four noise levels (0, 0.1, 0.2, 0.3), demonstrate that SVCT improves the average classification accuracy by 2.48% and reduces the mean squared error by 41.73% compared to a baseline CNN-Transformer. Notably, the highest accuracy gain of 3.3% is observed at the highest noise level. Confusion matrix analyses further confirm SVCT’s robustness, maintaining high classification rates even in the presence of severe noise. SVCT exhibits a smaller accuracy degradation (12.2%) compared to the baseline (13.5%) as noise levels increase, alongside significant reductions in RMSE (25.3%) and MAE (38.13%), as well as an improved R2 score (2.46% relative gain). These findings underscore the potential of SVCT for reliable fault diagnosis in noisy industrial environments, contributing to advancements in predictive maintenance technologies.

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Optimizing VMD Parameters Using SCSSA with CNN-Transformer for Fault Diagnosis

  • 会议论文

  • 发表日期：

    2026年

查看全文
  • Jiajing Zhou,
  • Yuanjun Guo,
  • Dianan Liu,
  • Zhile Yang

摘要

Bearing faults are a primary cause of machinery failures, underscoring the need for accurate and robust diagnostic methods, particularly under noisy conditions. This paper introduces the SCSSA-VMD-CNN-Transformer (SVCT) framework, which integrates the Sine Cosine Sparrow Search Algorithm (SCSSA)-optimized Variational Mode Decomposition (VMD) with a CNN-Transformer architecture for enhanced bearing fault diagnosis. SCSSA adaptively optimizes VMD parameters to decompose noisy vibration signals into intrinsic mode functions (IMFs), facilitating effective feature extraction and classification by the CNN-Transformer model. Experiments conducted on the Case Western Reserve University (CWRU) bearing dataset, across four noise levels (0, 0.1, 0.2, 0.3), demonstrate that SVCT improves the average classification accuracy by 2.48% and reduces the mean squared error by 41.73% compared to a baseline CNN-Transformer. Notably, the highest accuracy gain of 3.3% is observed at the highest noise level. Confusion matrix analyses further confirm SVCT’s robustness, maintaining high classification rates even in the presence of severe noise. SVCT exhibits a smaller accuracy degradation (12.2%) compared to the baseline (13.5%) as noise levels increase, alongside significant reductions in RMSE (25.3%) and MAE (38.13%), as well as an improved R2 score (2.46% relative gain). These findings underscore the potential of SVCT for reliable fault diagnosis in noisy industrial environments, contributing to advancements in predictive maintenance technologies.