Scholarlink

<p>Chatter has a significant impact on the machining process, severely affecting part quality and production efficiency, making its detection crucial. Recent advances in deep learning have led to remarkable achievements in diverse application areas, creating new possibilities for effective chatter detection. In addressing the issue of discrete-edge end mill chatter detection, this work introduces IB-CNN, a hybrid deep CNN in which the Inception-L module is combined with MDAM to strengthen feature representation. The Inception-L module enables automatic extraction of features across different scales in milling signals, enhancing feature representation capability, while the MDAM module combines multi-head attention and deformable convolution, facilitating adaptive emphasis on chatter relevant regions within the model. This enhancement strengthens the model’s ability to capture chatter relevant characteristics. Discrete-edge end milling experiments conducted under diverse operating conditions are used to evaluate the proposed approach through analyses of classification performance and generalization behavior. Experimental findings show that the proposed approach delivers over 98% average accuracy, with its robustness confirmed by comparisons against established methods.</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

The milling chatter detection method for discrete-edge end mill based on attention mechanism

  • 期刊论文

  • 发表日期：

    2026年2月16日

查看全文
  • Wei Yang,
  • Minli Zheng,
  • Wei Zhang,
  • Ming Song,
  • Baojuan Dong

摘要

Chatter has a significant impact on the machining process, severely affecting part quality and production efficiency, making its detection crucial. Recent advances in deep learning have led to remarkable achievements in diverse application areas, creating new possibilities for effective chatter detection. In addressing the issue of discrete-edge end mill chatter detection, this work introduces IB-CNN, a hybrid deep CNN in which the Inception-L module is combined with MDAM to strengthen feature representation. The Inception-L module enables automatic extraction of features across different scales in milling signals, enhancing feature representation capability, while the MDAM module combines multi-head attention and deformable convolution, facilitating adaptive emphasis on chatter relevant regions within the model. This enhancement strengthens the model’s ability to capture chatter relevant characteristics. Discrete-edge end milling experiments conducted under diverse operating conditions are used to evaluate the proposed approach through analyses of classification performance and generalization behavior. Experimental findings show that the proposed approach delivers over 98% average accuracy, with its robustness confirmed by comparisons against established methods.