Scholarlink

<p>Structural health monitoring (SHM) of spur gears is essential for ensuring the reliability and longevity of mechanical systems in industrial applications. This study presents a framework for spur gear fault diagnosis that integrates vibration and acoustic emission (AE) data using signal processing and supervised machine learning. Five gear fault conditions were created experimentally to evaluate diagnostic performance. The vibration and AE data were acquired using an adxl335 accelerometer and a max9814 microphone, recorded through an NI PCIe-6361 data acquisition card. The signals were analyzed in the time and frequency domains to extract fault signatures and assess the effectiveness of each sensor. Then machine learning models were trained on vibration data, acoustic data, and their fused combination. The framework demonstrated that models can be trained effectively on both homogeneous single-sensor data and heterogeneous fused data. This work underscores the potential of multi-sensor SHM systems with data fusion for predictive maintenance, providing insights into sensor selection, fusion strategies, and their role in optimizing fault diagnosis for rotating machinery. The findings support Industry 4.0 by enabling more accurate and efficient monitoring of critical mechanical components.</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

Multi-sensors Structural Health Monitoring of Spur Gears: A Comparative Analysis and Data Fusion Approach with Supervised Machine Learning

  • 期刊论文

  • 发表日期：

    2025年10月18日

查看全文
  • Muhammad Talha Asif,
  • Shahzad Ahmad,
  • Jia Ge,
  • Muhammad Arslan Ashraf

摘要

Structural health monitoring (SHM) of spur gears is essential for ensuring the reliability and longevity of mechanical systems in industrial applications. This study presents a framework for spur gear fault diagnosis that integrates vibration and acoustic emission (AE) data using signal processing and supervised machine learning. Five gear fault conditions were created experimentally to evaluate diagnostic performance. The vibration and AE data were acquired using an adxl335 accelerometer and a max9814 microphone, recorded through an NI PCIe-6361 data acquisition card. The signals were analyzed in the time and frequency domains to extract fault signatures and assess the effectiveness of each sensor. Then machine learning models were trained on vibration data, acoustic data, and their fused combination. The framework demonstrated that models can be trained effectively on both homogeneous single-sensor data and heterogeneous fused data. This work underscores the potential of multi-sensor SHM systems with data fusion for predictive maintenance, providing insights into sensor selection, fusion strategies, and their role in optimizing fault diagnosis for rotating machinery. The findings support Industry 4.0 by enabling more accurate and efficient monitoring of critical mechanical components.