Scholarlink

<p>This work presents a novel domain-adaptive, multi-channel residual network for accurate and robust bearing fault diagnosis capable of operating under diverse conditions. Additionally, the limitations of scarce data and variable operating conditions, pressing the need for more robust fault diagnosis methods, are addressed. To effectively manage such discrepancies, a maximum mean discrepancy (MMD) and network architecture incorporating a pre-trained deep residual network are used to facilitate effective knowledge transfer between source and target domains. Furthermore, efficient use of multi-channel data for specific diagnostic tasks is prioritized to optimize dynamic extraction of relevant features. Employing dynamic signal processing techniques, the proposed method extracts distinguishing features from multi-channel sensor data. Variational mode decomposition (VMD) is implemented to securely extract distinctive fault characteristics (e.g., specific fault types). Prior to this selective extraction, noise is removed, significantly improving accuracy and enhancing model stability. Dynamic signal decomposition is performed on augmented, multi-channel signals, and the variational mode decomposition (VMD) is applied to overcome noisy, distorted, and complex signals, significantly improving fault recognition. The proposed methodology effectively addresses the domain shift problems inherent in real-world scenarios, resulting in more consistent and robust solutions for bearing fault diagnosis. This work especially addresses limitations in the availability of sufficient data for specific fault types, demonstrating improved efficiency in identifying and classifying bearing faults. The proposed method combines domain adaptation techniques with an artificial intelligence (AI) approach to develop an automated fault diagnosis system.</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

Adaptive Multi-Sensor Residual Framework for Robust Gear Anomaly Detection in Aviation Systems

  • 期刊论文

  • 发表日期：

    2026年1月25日

查看全文
  • Anping Wan,
  • Fei Zhang,
  • Khalil AL-Bukhaiti,
  • Xiaomin Cheng,
  • Xiaosheng Ji

摘要

This work presents a novel domain-adaptive, multi-channel residual network for accurate and robust bearing fault diagnosis capable of operating under diverse conditions. Additionally, the limitations of scarce data and variable operating conditions, pressing the need for more robust fault diagnosis methods, are addressed. To effectively manage such discrepancies, a maximum mean discrepancy (MMD) and network architecture incorporating a pre-trained deep residual network are used to facilitate effective knowledge transfer between source and target domains. Furthermore, efficient use of multi-channel data for specific diagnostic tasks is prioritized to optimize dynamic extraction of relevant features. Employing dynamic signal processing techniques, the proposed method extracts distinguishing features from multi-channel sensor data. Variational mode decomposition (VMD) is implemented to securely extract distinctive fault characteristics (e.g., specific fault types). Prior to this selective extraction, noise is removed, significantly improving accuracy and enhancing model stability. Dynamic signal decomposition is performed on augmented, multi-channel signals, and the variational mode decomposition (VMD) is applied to overcome noisy, distorted, and complex signals, significantly improving fault recognition. The proposed methodology effectively addresses the domain shift problems inherent in real-world scenarios, resulting in more consistent and robust solutions for bearing fault diagnosis. This work especially addresses limitations in the availability of sufficient data for specific fault types, demonstrating improved efficiency in identifying and classifying bearing faults. The proposed method combines domain adaptation techniques with an artificial intelligence (AI) approach to develop an automated fault diagnosis system.