Scholarlink

To correctly diagnose multiple faults in rotating elements particularly bearing, in this paper, the authors have proposed, wavelet-based denoising, empirical mode decomposition (EMD), and unsupervised machine learning-based method principal component analysis (PCA) and independent component analysis (ICA) for fault diagnosis. Bearing having faults in outer race, inner race, and angular misalignment induced artificially and assembled to experimental portable setup at the laboratory. Acquisition of the vibration data is done using FFT analyser which is SPM made LEONOVA DIAMOND® and data are processed using CONDMASTER RUBY® and MATLAB® software. Firstly, the signal is denoised using wavelet transform. After denoising, EMD is applied for denoised signal decomposition into different intrinsic mode functions (IMFs). After EMD, principal component analysis (PCA) is applied on IMFs followed by independent component analysis (ICA) to get independent components, which are then analysed to extract fault related characteristic features. This paper proposes Wavelet-EMD-PCA-ICA-based technique for fault diagnosis and shows convincing results particularly for small bearings and at lower rpm, especially when multiple faults are present, where fault signatures are difficult to identify in a normal FFT spectrum.

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

Integrating Machine Learning with Empirical Mode Decomposition for Multiple Fault Diagnosis in Rotating Machinery

  • 会议论文

  • 发表日期：

    2026年

查看全文
  • Sunil Kumar,
  • Swapna Singh,
  • Nitesh P. Yelve

摘要

To correctly diagnose multiple faults in rotating elements particularly bearing, in this paper, the authors have proposed, wavelet-based denoising, empirical mode decomposition (EMD), and unsupervised machine learning-based method principal component analysis (PCA) and independent component analysis (ICA) for fault diagnosis. Bearing having faults in outer race, inner race, and angular misalignment induced artificially and assembled to experimental portable setup at the laboratory. Acquisition of the vibration data is done using FFT analyser which is SPM made LEONOVA DIAMOND® and data are processed using CONDMASTER RUBY® and MATLAB® software. Firstly, the signal is denoised using wavelet transform. After denoising, EMD is applied for denoised signal decomposition into different intrinsic mode functions (IMFs). After EMD, principal component analysis (PCA) is applied on IMFs followed by independent component analysis (ICA) to get independent components, which are then analysed to extract fault related characteristic features. This paper proposes Wavelet-EMD-PCA-ICA-based technique for fault diagnosis and shows convincing results particularly for small bearings and at lower rpm, especially when multiple faults are present, where fault signatures are difficult to identify in a normal FFT spectrum.