Scholarlink

<p>With the rapid advancement of aviation technology, ensuring the reliability of aircraft engines through accurate fault diagnosis remains a critical challenge. Current studies predominantly focus on component-level faults, while subsystem-level diagnostics—essential for comprehensive maintenance strategies—are under-explored. This paper proposes the Wavelet-CNNTrans Method, integrating continuous wavelet transform (CWT), convolutional neural networks (CNN), and Transformer architectures to address the dynamic, non-stationary flight data of turbine engines. Our objective is to enhance fault diagnosis accuracy by extracting time–frequency features from multivariate sensor data and leveraging deep learning for pattern recognition. Evaluated on NASA’s N-CMAPSS dataset, the method achieves 91.85% on accuracy in identifying subsystem-level faults under complex operating conditions, demonstrating its potential for improving engine safety and maintenance efficiency.</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

Wavelet-CNNTrans method for fault diagnosis of aircraft turbine engine subsystems utilizing flight data

  • 期刊论文

  • 发表日期：

    2026年6月25日

查看全文
  • Siqi Wu,
  • Zhiqiang Peng,
  • Quanbao Wang,
  • Kexin Li,
  • Yingli Xu

摘要

With the rapid advancement of aviation technology, ensuring the reliability of aircraft engines through accurate fault diagnosis remains a critical challenge. Current studies predominantly focus on component-level faults, while subsystem-level diagnostics—essential for comprehensive maintenance strategies—are under-explored. This paper proposes the Wavelet-CNNTrans Method, integrating continuous wavelet transform (CWT), convolutional neural networks (CNN), and Transformer architectures to address the dynamic, non-stationary flight data of turbine engines. Our objective is to enhance fault diagnosis accuracy by extracting time–frequency features from multivariate sensor data and leveraging deep learning for pattern recognition. Evaluated on NASA’s N-CMAPSS dataset, the method achieves 91.85% on accuracy in identifying subsystem-level faults under complex operating conditions, demonstrating its potential for improving engine safety and maintenance efficiency.