Scholarlink

<p>The multi-stage stamping of aluminum alloys generates densely distributed and highly overlapping acoustic emission (AE) activity; thus, isolating and characterizing burst-type signals using conventional AE parameters is difficult. The scarcity of defective samples in industrial production further limits the applicability of supervised crack-detection approaches, which require balanced datasets to establish decision boundaries. By analyzing AE signals collected directly from an automotive stamping line, this study confirms that crack formation in AA6014-T4 panels consistently produces high-energy burst transients through the deformation process. By leveraging this physical signature, a baseline-only anomaly detection framework was developed using lightweight and physically interpretable AE features. Field validation on a stamping line demonstrated that the proposed method remains robust under real-world industrial noise and process variability. Detection accuracies of 98.21% and 92.86% were achieved at two sensor locations, with the sensor closest to the stamping region attaining a 100% recall rate, ensuring that no cracks were missed. These findings verify that burst-type AE characteristics provide a reliable basis for real-time crack detection in multi-stage aluminum stamping, highlighting the practical value of baseline-driven monitoring strategies for industrial manufacturing.</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

Crack detection in multi-stage stamping using acoustic emission based on baseline feature distributions

  • 期刊论文

  • 发表日期：

    2026年1月10日

查看全文
  • Zhichao Sun,
  • Penghua Zhang,
  • Luis Paulo M. Lima,
  • Lijing Chen,
  • Yeseul Kong,
  • Gyuhae Park

摘要

The multi-stage stamping of aluminum alloys generates densely distributed and highly overlapping acoustic emission (AE) activity; thus, isolating and characterizing burst-type signals using conventional AE parameters is difficult. The scarcity of defective samples in industrial production further limits the applicability of supervised crack-detection approaches, which require balanced datasets to establish decision boundaries. By analyzing AE signals collected directly from an automotive stamping line, this study confirms that crack formation in AA6014-T4 panels consistently produces high-energy burst transients through the deformation process. By leveraging this physical signature, a baseline-only anomaly detection framework was developed using lightweight and physically interpretable AE features. Field validation on a stamping line demonstrated that the proposed method remains robust under real-world industrial noise and process variability. Detection accuracies of 98.21% and 92.86% were achieved at two sensor locations, with the sensor closest to the stamping region attaining a 100% recall rate, ensuring that no cracks were missed. These findings verify that burst-type AE characteristics provide a reliable basis for real-time crack detection in multi-stage aluminum stamping, highlighting the practical value of baseline-driven monitoring strategies for industrial manufacturing.