Scholarlink

Purpose <p>Chatter is a common self-excited vibration phenomenon in milling processes, significantly impacting part quality and production efficiency. This study aims to develop a high-precision chatter identification method to address the limitations of single-sensor systems in capturing complex vibration patterns and to enhance machining stability.</p> Methods <p>A multi-sensor fusion chatter identification method based on deep convolutional neural networks (AlexNet-MF network) is proposed. Continuous wavelet transform (CWT) is employed to convert 1D vibration signals into 2D time-frequency images as input. Multi-sensor information is integrated through data-level, feature-level, and decision-level fusion strategies, and the model's generalizability is validated via milling experiments with different cutting tools.</p> Results <p>Experimental results demonstrate that the proposed method achieves an average chatter identification accuracy of over 95%, significantly outperforming comparative methods. The multi-sensor fusion strategy effectively improves the capture capability for complex vibration patterns, while the time-frequency image input enhances the robustness of feature learning.</p> Conclusion <p>The AlexNet-MF-based multisensor fusion method enables efficient identification of milling chatter. Its high accuracy and generalizability provide a reliable solution for vibration monitoring in industrial applications.</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

Milling Chatter Identification Based on Deep Convolutional Neural Networks and Multi-Sensor Fusion

  • 期刊论文

  • 发表日期：

    2025年6月3日

查看全文
  • Minli Zheng,
  • Wei Yang,
  • Yupeng Si,
  • Tong Zheng

摘要

Purpose

Chatter is a common self-excited vibration phenomenon in milling processes, significantly impacting part quality and production efficiency. This study aims to develop a high-precision chatter identification method to address the limitations of single-sensor systems in capturing complex vibration patterns and to enhance machining stability.

Methods

A multi-sensor fusion chatter identification method based on deep convolutional neural networks (AlexNet-MF network) is proposed. Continuous wavelet transform (CWT) is employed to convert 1D vibration signals into 2D time-frequency images as input. Multi-sensor information is integrated through data-level, feature-level, and decision-level fusion strategies, and the model's generalizability is validated via milling experiments with different cutting tools.

Results

Experimental results demonstrate that the proposed method achieves an average chatter identification accuracy of over 95%, significantly outperforming comparative methods. The multi-sensor fusion strategy effectively improves the capture capability for complex vibration patterns, while the time-frequency image input enhances the robustness of feature learning.

Conclusion

The AlexNet-MF-based multisensor fusion method enables efficient identification of milling chatter. Its high accuracy and generalizability provide a reliable solution for vibration monitoring in industrial applications.