Scholarlink

Purpose <p>Fault diagnosis in rotating machinery is critical for maintaining safe, reliable, and efficient operation in industrial applications. The classification of combined unbalance and shaft crack fault conditions remains challenging due to their interacting effects.</p> Methods <p>This study proposes a hybrid fault diagnosis framework integrating physics-based numerical modeling, experimental vibration measurements, and Artificial Neural Networks (ANNs). A multi-disk rotor-bearing system with unbalance and shaft crack faults is investigated through finite element modeling and experimental testing under various operating conditions. Vibration features are extracted in the frequency domain and used to train ANN-based classifiers.</p> Results <p>By combining prior knowledge of system dynamics with data-driven methods, the proposed approach allows classification of combined unbalance and crack fault conditions. Experimental results further revealed nonlinear interaction effects not fully captured by the numerical model. The proposed ANN-based hierarchical classification framework demonstrated promising classification capability under the examined fault conditions, with individual ANN classifiers achieving accuracies exceeding 83%.</p> Conclusions <p>The proposed methodology provides a computationally efficient fault classification framework with potential applicability in condition-monitoring systems, supporting the development of advanced fault diagnosis methodologies for rotating machinery.</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

Vibration-Based Multi-Fault Diagnosis of Rotating Systems using Artificial Neural Networks

  • 期刊论文

  • 发表日期：

    2026年6月22日

查看全文
  • Ioannis Tselios,
  • Georgios I. Giannopoulos,
  • Pantelis G. Nikolakopoulos

摘要

Purpose

Fault diagnosis in rotating machinery is critical for maintaining safe, reliable, and efficient operation in industrial applications. The classification of combined unbalance and shaft crack fault conditions remains challenging due to their interacting effects.

Methods

This study proposes a hybrid fault diagnosis framework integrating physics-based numerical modeling, experimental vibration measurements, and Artificial Neural Networks (ANNs). A multi-disk rotor-bearing system with unbalance and shaft crack faults is investigated through finite element modeling and experimental testing under various operating conditions. Vibration features are extracted in the frequency domain and used to train ANN-based classifiers.

Results

By combining prior knowledge of system dynamics with data-driven methods, the proposed approach allows classification of combined unbalance and crack fault conditions. Experimental results further revealed nonlinear interaction effects not fully captured by the numerical model. The proposed ANN-based hierarchical classification framework demonstrated promising classification capability under the examined fault conditions, with individual ANN classifiers achieving accuracies exceeding 83%.

Conclusions

The proposed methodology provides a computationally efficient fault classification framework with potential applicability in condition-monitoring systems, supporting the development of advanced fault diagnosis methodologies for rotating machinery.