Scholarlink

<p>Internet of Things (IoT) concept transforms on the industry has assisted in capturing the significant amount of smart data to perform the intelligent organization of the industrial system. The bearing fault diagnosis is one of the application areas of Industrial IoT (IIoT)/I4.0 to provide the sustainable rotating machinery for Industry. The performance of bearing fault diagnosis is dependent on the features extracted from the vibration data for the training of the classifier. A significant feature space exists in the literature, but no dynamic approach has yet been presented to identify the most promising feature set to represent faults under varying operating conditions for fault diagnosis. Therefore, this paper proposes a new fault diagnosis framework based on clustering and ensemble technique. In the first phase of the framework, a novel feature selection routine is exploited to determine optimal feature subset. The feature ranking and clustering methods have collaboratively worked together to generate dynamic optimum feature subset, which is then presented to ensemble classifier for fault diagnosis. The second phase of the framework is an enhancement of ensemble classifier (EC) learning for achieving the decent performance of fault classification. The improvement in EC learning has accomplished by eliminating class imbalance data subset to construct a non-dominant solution and allow the EC set to grow automatically. The proposed framework is experimented with two vibration data sets and examined the significance of the suggested feature selection routine as well as EC learning routine in comparison with existing alternatives. The overall experimental result signifies that the proposed framework obtains the dynamic optimal feature subset and improves the EC learning to enhance fault classification accuracy.</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

Data-driven approach under varying operating conditions for bearing fault diagnosis

  • 期刊论文

  • 发表日期：

    2026年2月2日

查看全文
  • Sandeep S. Udmale,
  • Aneesh G. Nath,
  • Durgesh Singh,
  • Sanjay Kumar Singh

摘要

Internet of Things (IoT) concept transforms on the industry has assisted in capturing the significant amount of smart data to perform the intelligent organization of the industrial system. The bearing fault diagnosis is one of the application areas of Industrial IoT (IIoT)/I4.0 to provide the sustainable rotating machinery for Industry. The performance of bearing fault diagnosis is dependent on the features extracted from the vibration data for the training of the classifier. A significant feature space exists in the literature, but no dynamic approach has yet been presented to identify the most promising feature set to represent faults under varying operating conditions for fault diagnosis. Therefore, this paper proposes a new fault diagnosis framework based on clustering and ensemble technique. In the first phase of the framework, a novel feature selection routine is exploited to determine optimal feature subset. The feature ranking and clustering methods have collaboratively worked together to generate dynamic optimum feature subset, which is then presented to ensemble classifier for fault diagnosis. The second phase of the framework is an enhancement of ensemble classifier (EC) learning for achieving the decent performance of fault classification. The improvement in EC learning has accomplished by eliminating class imbalance data subset to construct a non-dominant solution and allow the EC set to grow automatically. The proposed framework is experimented with two vibration data sets and examined the significance of the suggested feature selection routine as well as EC learning routine in comparison with existing alternatives. The overall experimental result signifies that the proposed framework obtains the dynamic optimal feature subset and improves the EC learning to enhance fault classification accuracy.