Scholarlink

<p>Continuous linear segment paths are commonly employed in five-axis machining. Nevertheless, the discontinuities at segment corners induce abrupt tool velocity variations that generate detrimental vibrations and mechanical shocks. These issues emphasize the necessity of enhancing path continuity and smoothness. Traditional smoothing methods often suffer from limited computational efficiency, inconsistent smoothing performance, and inaccuracy in error control. To overcome these limitations, this study introduces a convolution-based path smoothing approach integrated with a smoothing error control model for five-axis machining tool paths. Initially, this approach decouples the tool trajectory in five-axis machining into its positional and orientational components, then constructed convolution function at segment corners to smooth the path, and finally performs synchronized smoothing to achieve G2-continuous trajectory optimization. Furthermore, a physics informed neural network model is constructed to dynamically regulate the convolution parameters, enabling real-time control over the smoothing error. Through conducting simulation and experimental validation, the effectiveness of the approach in reducing tracking errors and mitigating spindle vibrations has been demonstrated. Specifically, compared with conventional approaches, a significant reduction in spindle vibration amplitudes is observed, with decreases of 49.8%, 23.1%, and 18.3% in X, Y, and Z directions, respectively.</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

Convolution smoothing for five-axis machine path with physics informed neutral network-based smoothing error control

  • 期刊论文

  • 发表日期：

    2026年5月2日

查看全文
  • Kun Li,
  • Guanghui Zhou,
  • Zenghui Wang,
  • Zhijie Wei,
  • Qingyu Yang,
  • Hongwen Xing,
  • Wei Wang,
  • Chao Zhang

摘要

Continuous linear segment paths are commonly employed in five-axis machining. Nevertheless, the discontinuities at segment corners induce abrupt tool velocity variations that generate detrimental vibrations and mechanical shocks. These issues emphasize the necessity of enhancing path continuity and smoothness. Traditional smoothing methods often suffer from limited computational efficiency, inconsistent smoothing performance, and inaccuracy in error control. To overcome these limitations, this study introduces a convolution-based path smoothing approach integrated with a smoothing error control model for five-axis machining tool paths. Initially, this approach decouples the tool trajectory in five-axis machining into its positional and orientational components, then constructed convolution function at segment corners to smooth the path, and finally performs synchronized smoothing to achieve G2-continuous trajectory optimization. Furthermore, a physics informed neural network model is constructed to dynamically regulate the convolution parameters, enabling real-time control over the smoothing error. Through conducting simulation and experimental validation, the effectiveness of the approach in reducing tracking errors and mitigating spindle vibrations has been demonstrated. Specifically, compared with conventional approaches, a significant reduction in spindle vibration amplitudes is observed, with decreases of 49.8%, 23.1%, and 18.3% in X, Y, and Z directions, respectively.