Scholarlink

<p>Spindle thermal error is a primary factor affecting the machining accuracy of machine tools. However, the complexity of heat transfer poses significant challenges for the development of accurate thermal error models. This paper solves this problem by proposing a spindle thermal error method based on region-segmented thermal images. Thermal images are directly input into the model without relying on conventional temperature sensors. An image segmentation algorithm is then adopted to divide the thermal images into multiple regional units. In addition, positional encoding is introduced to preserve spatial contextual information, allowing to mitigate the issue of missing spatial correlation faced by conventional segmentation methods. The model removes background noise through a local attention mechanism and models cross-channel dependencies through a channel attention mechanism. Thus, it comprehensively uses infrared thermal image information. A thermal error prediction model is developed based on the spindle of a VMC450 machine tool. At an operating condition of 3000&#xa0;rpm, the model reaches a prediction accuracy of 99.28%. Afterwards, experiments at 2000&#xa0;rpm and 4000&#xa0;rpm are conducted, demonstrating that the proposed model exhibits prediction accuracies of 93.28% and 85.29%, respectively. This validates its high robustness and accuracy at different working conditions. The results obtained in this study provide a practical support for the promotion and application of the proposed model in thermal error compensation and accuracy optimization of various machine tools.</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

A Spindle Thermal Error Model Based on Region Segmented Thermal Images

  • 期刊论文

  • 发表日期：

    2026年3月16日

查看全文
  • Yu Zheng,
  • Yutao Zhu,
  • Guanting Pan,
  • Longtao Mu,
  • Yining Li,
  • Junhao Zhao,
  • Ying Wang

摘要

Spindle thermal error is a primary factor affecting the machining accuracy of machine tools. However, the complexity of heat transfer poses significant challenges for the development of accurate thermal error models. This paper solves this problem by proposing a spindle thermal error method based on region-segmented thermal images. Thermal images are directly input into the model without relying on conventional temperature sensors. An image segmentation algorithm is then adopted to divide the thermal images into multiple regional units. In addition, positional encoding is introduced to preserve spatial contextual information, allowing to mitigate the issue of missing spatial correlation faced by conventional segmentation methods. The model removes background noise through a local attention mechanism and models cross-channel dependencies through a channel attention mechanism. Thus, it comprehensively uses infrared thermal image information. A thermal error prediction model is developed based on the spindle of a VMC450 machine tool. At an operating condition of 3000 rpm, the model reaches a prediction accuracy of 99.28%. Afterwards, experiments at 2000 rpm and 4000 rpm are conducted, demonstrating that the proposed model exhibits prediction accuracies of 93.28% and 85.29%, respectively. This validates its high robustness and accuracy at different working conditions. The results obtained in this study provide a practical support for the promotion and application of the proposed model in thermal error compensation and accuracy optimization of various machine tools.