首页 > AI前沿 > Predicting Surface Displacement Fields of Laser Powder Bed Fusion Parts Using STL-Derived Graphs and Part-Level Descriptors

Predicting Surface Displacement Fields of Laser Powder Bed Fusion Parts Using STL-Derived Graphs and Part-Level Descriptors

arXiv机器学习 2026-11-20 00:07 11 阅读 查看原文
用户需要我将一段新闻(实际上是学术摘要)文本排版成语义化的HTML片段。我需要遵守规则: 1. 只输出HTML片段,不使用Markdown围栏,不加解释。 2. 每个自然段落使用 `

` 包裹,长文使用 `

`/`

` 划分小节。 3. 列表用 `
  • `,引用用 `
    `,代码用 `
    `,重要句子用 ``。
    4. 保留原文所有信息,不得增删改内容。
    5. 不要输出 ``/``/`` 等文档级标签。
    
    观察给定文本:这是一个关于激光粉末床熔融(LPBF)零件变形预测的学术摘要。它不是传统的新闻,但排版原则是相同的。
    
    分析原文结构:
    第一句:“Predicting part distortion before fabrication remains challenging in laser powder bed fusion (LPBF), while thermomechanical simulation is computationally expensive for repeated design evaluations.” —— 介绍研究背景与挑战。
    第二句:“This study presents a two-branch graph attention network (GAT) to predict three-dimensional surface displacement directly from STL-derived graphs.” —— 介绍提出的方法。
    第三句:“The first branch learns geometric relationships among connected vertices of a densified STL surface mesh, while the second incorporates 12 part-level descriptors of mass distribution, build-plate contact, overhang, size, and shape.” —— 介绍网络的两个分支。
    第四句:“Displacement labels from ANSYS Additive Print simulations are mapped onto STL vertices using Gaussian-kernel weighted averaging.” —— 介绍标签映射方法。
    第五句:“Evaluated on 117 parts using five-fold cross-validation split by part, the model achieves an R^2 of 0.946 and a mean absolute error of 0.0178 mm for displacement magnitude.” —— 评估结果。
    第六句:“Relative to a node-wise multilayer perceptron, GAT improves R^2 from 0.913 to 0.946; for build-direction displacement, R^2 increases from 0.903 to 0.939.” —— 与多层感知机的对比。
    第七句:“GAT also outperforms GraphSAGE (R^2 = 0.936) and a graph convolutional network (R^2 = 0.934) across all five folds.” —— 与其他图神经网络对比。
    第八句:“Ablation experiments show that removing mass-distribution descriptors reduces magnitude R^2 to 0.900, while removing build-plate contact and overhang descriptors reduces it to 0.920.” —— 消融实验结果。
    第九句:“These results demonstrate the value of combining local mesh connectivity with global geometric descriptors to predict LPBF distortion, enabling faster design iteration and earlier identification of distortion-prone geometries.” —— 总结意义。
    
    思考如何将其划分为小节:
    可以是:
    

    研究背景与方法

    包含前4句的段落。或者将它们拆分成多个 `

    `。 实际上,原文是一整段文字。我应当保留所有信息。规则提到“每个自然段落用

    包裹,长文用

    /

    划分小节”。 我们可以这样组织:

    研究背景与网络架构

    Predicting part distortion before fabrication remains challenging...

    This study presents a two-branch graph attention network (GAT)...

    The first branch learns geometric relationships...

    Displacement labels from ANSYS Additive Print simulations...

    实验评估与结果

    Evaluated on 117 parts using five-fold cross-validation...

    Relative to a node-wise multilayer perceptron...

    GAT also outperforms GraphSAGE (R^2 = 0.936) and a graph convolutional network (R^2 = 0.934) across all five folds.

    消融实验与结论

    Ablation experiments show that removing mass-distribution descriptors...

    These results demonstrate the value of combining local mesh connectivity...

    或者我也可以把第一部分直接作为一段,因为它本来就是一个自然段。原文是连续的文本,没有明确的段落分隔。但用 `

    ` 划分小节会更好。 检查是否需要使用列表、引用、代码或强调。原文是纯文本学术摘要,没有明显的列表、引用或代码。我可以用 `` 来强调重要句子。例如,主要结论可以用 `` 包裹。 重要句子: - “This study presents a two-branch graph attention network (GAT) to predict three-dimensional surface displacement directly from STL-derived graphs.”——这是核心研究内容。 - “These results demonstrate the value of combining local mesh connectivity with global geometric descriptors to predict LPBF distortion, enabling faster design iteration and earlier identification of distortion-prone geometries.”——这是总结意义。 我们可以把这些句子用 `` 包裹。 注意:原文有 “R^2”。在 HTML 中,也许应该保留为纯文本 “R^2”,因为规则说“不得增删改内容”。用户没有要求将上标转换为 ``,所以我们保留原样为 “R^2”。 最终的 HTML 结构:

    研究背景与方法

    Predicting part distortion before fabrication remains challenging in laser powder bed fusion (LPBF), while thermomechanical simulation is computationally expensive for repeated design evaluations. This study presents a two-branch graph attention network (GAT) to predict three-dimensional surface displacement directly from STL-derived graphs. The first branch learns geometric relationships among connected vertices of a densified STL surface mesh, while the second incorporates 12 part-level descriptors of mass distribution, build-plate contact, overhang, size, and shape. Displacement labels from ANSYS Additive Print simulations are mapped onto STL vertices using Gaussian-kernel weighted averaging.

    实验评估与对比分析

    Evaluated on 117 parts using five-fold cross-validation split by part, the model achieves an R^2 of 0.946 and a mean absolute error of 0.0178 mm for displacement magnitude. Relative to a node-wise multilayer perceptron, GAT improves R^2 from 0.913 to 0.946; for build-direction displacement, R^2 increases from 0.903 to 0.939. GAT also outperforms GraphSAGE (R^2 = 0.936) and a graph convolutional network (R^2 = 0.934) across all five folds.

    消融实验与研究结论

    Ablation experiments show that removing mass-distribution descriptors reduces magnitude R^2 to 0.900, while removing build-plate contact and overhang descriptors reduces it to 0.920. These results demonstrate the value of combining local mesh connectivity with global geometric descriptors to predict LPBF distortion, enabling faster design iteration and earlier identification of distortion-prone geometries.

    再次检查内容是否与原文完全一致。 原文: Predicting part distortion before fabrication remains challenging in laser powder bed fusion (LPBF), while thermomechanical simulation is computationally expensive for repeated design evaluations. This study presents a two-branch graph attention network (GAT) to predict three-dimensional surface displacement directly from STL-derived graphs. The first branch learns geometric relationships among connected vertices of a densified STL surface mesh, while the second incorporates 12 part-level descriptors of mass distribution, build-plate contact, overhang, size, and shape. Displacement labels from ANSYS Additive Print simulations are mapped onto STL vertices using Gaussian-kernel weighted averaging. Evaluated on 117 parts using five-fold cross-validation split by part, the model achieves an R^2 of 0.946 and a mean absolute error of 0.0178 mm for displacement magnitude. Relative to a node-wise multilayer perceptron, GAT improves R^2 from 0.913 to 0.94