Scholarlink

<p>Gearbox defects are a leading cause of failure in mechanical transmission systems, resulting in significant operational disruptions and financial losses. This study introduces a novel diagnostic tool combining complemented empirical mode decomposition with adaptive noise (CEEMDAN) and Bayesian convolutional neural network (BCNN) to accurately identify compound faults. CEEMDAN enhances signal extraction and reduces residual noise, outperforming other empirical methods like EMD, EEMD, CEEMD, HVD, and VMD. The BCNN classifier, trained on mixed faults, quantifies uncertainty, enabling more reliable predictions. This approach achieves a superior classification accuracy of 99.12 % on the test dataset, proving its effectiveness in fault differentiation and enhancing diagnostic reliability.</p>

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

Bayesian convolution neural network for mechanical transmission fault diagnosis with empirical signal processing

  • 期刊论文

  • 发表日期：

    2025年5月31日

查看全文
  • Kamar Zaiem,
  • Hichem Bouras,
  • Mohamed Faouzi Rachedi

摘要

Gearbox defects are a leading cause of failure in mechanical transmission systems, resulting in significant operational disruptions and financial losses. This study introduces a novel diagnostic tool combining complemented empirical mode decomposition with adaptive noise (CEEMDAN) and Bayesian convolutional neural network (BCNN) to accurately identify compound faults. CEEMDAN enhances signal extraction and reduces residual noise, outperforming other empirical methods like EMD, EEMD, CEEMD, HVD, and VMD. The BCNN classifier, trained on mixed faults, quantifies uncertainty, enabling more reliable predictions. This approach achieves a superior classification accuracy of 99.12 % on the test dataset, proving its effectiveness in fault differentiation and enhancing diagnostic reliability.