首页 > AI前沿 > Fuse-PINN: Combining Finite Element Methods and Physics-Informed Neural Networks to enable massively parallel simulations of non-linear PDEs

Fuse-PINN: Combining Finite Element Methods and Physics-Informed Neural Networks to enable massively parallel simulations of non-linear PDEs

arXiv机器学习 2026-11-14 00:55 12 阅读 查看原文
用户需要将一段英文新闻正文排版成语义化的HTML片段。 规则: 1. 只输出HTML片段本身,不要Markdown围栏、不要解释。 2. 每个自然段落用`

`包裹,长文用`

`/`

`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。
    
    分析正文内容:
    正文是一段关于物理信息神经网络(PINNs)和Fuse-PINN框架的学术新闻或摘要。
    原文:
    "Research in the domain of Physics-Informed Neural Networks applied to PDE modelling has advanced significantly in recent years, providing a wide ecosystem of novel methods. However, one frequent limitation is the inability of PINNs to scale out on High-Performance Computing architectures to address more ambitious challenges, particularly non-linear physics models with turbulent dynamics in large spatial domains. Typical studies restrain demonstrations to toy models with simple physics equations, few variables, and rarely any complex non-linear dynamics evolved over long timescales. Part of the AI community has even adopted the idea that "PINNs don't scale". This work proves otherwise. We present a mathematically robust formulation of Finite-Element-Method for Neural Networks (FEM-NN) implemented into a comprehensive, modular framework: Fuse-PINN. This formulation provides the basis for a sclable parallelisation of Physics-Informed Neural Networks on GPU-based HPC systems. Furthermore, an Energy Natural Gradient Descent optimiser is implemented to allow high-precision convergence. Fuse-PINN is demonstrated on physics models relevant to Magnetic Confinement Fusion, including non-linear convective models with fine-scale dynamics. Near-perfect scaling is achieved on multi-GPU runs with increasing problem size."
    
    虽然用户说"长文用

    /

    划分小节",但这篇文本其实比较短,可以分成几个逻辑部分,或者直接用一个或多个`

    `和`

    `。 为了符合"长文用

    /

    划分小节"的要求,我可以划分几个小节: 小节1:背景与挑战 (Background and Challenges) 小节2:Fuse-PINN框架与优化 (Fuse-PINN Framework and Optimization) 小节3:实验与结果 (Experiments and Results) 不过,规则4要求“不得增删改内容”。如果加了小标题,算不算“增改内容”? 用户说“保留原文所有信息,不得增删改内容”。如果添加小标题,就是增加了原文没有的文字。因此,最好不添加额外的小标题文字,或者仅使用原文中的某些句子作为小标题?但原文没有明显的小标题。 或者,整个文本作为一个长段落,或者分成几个自然段落,用`

    `来表示大标题(如果没有大标题,可以不用,或者用原文的第一句话?)。 等等,规则2说“每个自然段落用`

    `包裹,长文用`

    `/`

    `划分小节”。如果原文本身没有小节标题,强行加小节标题会违反“不得增删改内容”。所以最安全的做法是:不添加任何原文没有的标题文字。如果一定要用`

    `,可以把原文的某些核心句提取出来做标题?但这也算修改。 更好的理解是:如果原文有结构,就划分;如果没有,就全部用`

    `。或者,将整个文本视为一个段落,用`

    `包裹。 但为了体现排版工程师的能力,我可以把原文按逻辑断句,分成几个`

    `。 原文段落: 1. Research in the domain of Physics-Informed Neural Networks applied to PDE modelling has advanced significantly in recent years, providing a wide ecosystem of novel methods. However, one frequent limitation is the inability of PINNs to scale out on High-Performance Computing architectures to address more ambitious challenges, particularly non-linear physics models with turbulent dynamics in large spatial domains. 2. Typical studies restrain demonstrations to toy models with simple physics equations, few variables, and rarely any complex non-linear dynamics evolved over long timescales. Part of the AI community has even adopted the idea that "PINNs don't scale". This work proves otherwise. 3. We present a mathematically robust formulation of Finite-Element-Method for Neural Networks (FEM-NN) implemented into a comprehensive, modular framework: Fuse-PINN. This formulation provides the basis for a sclable parallelisation of Physics-Informed Neural Networks on GPU-based HPC systems. Furthermore, an Energy Natural Gradient Descent optimiser is implemented to allow high-precision convergence. 4. Fuse-PINN is demonstrated on physics models relevant to Magnetic Confinement Fusion, including non-linear convective models with fine-scale dynamics. Near-perfect scaling is achieved on multi-GPU runs with increasing problem size. 这样分成4个`

    `,不增加任何额外文字,完全保留原文。 关于``,可以把重要的名词或结论加粗,比如 "PINNs don't scale", "Fuse-PINN", "Energy Natural Gradient Descent optimiser" 等。规则说“重要句子可用``”。 关于引用,"PINNs don't scale" 可以用 `

    ` 或者 ``,规则说“引用用`
    `”,所以:
    "PINNs don't scale"
    检查原文是否有代码,没有。 检查是否有列表,没有。 排版设计:

    Research in the domain of Physics-Informed Neural Networks applied to PDE modelling has advanced significantly in recent years, providing a wide ecosystem of novel methods. However, one frequent limitation is the inability of PINNs to scale out on High-Performance Computing architectures to address more ambitious challenges, particularly non-linear physics models with turbulent dynamics in large spatial domains.

    Typical studies restrain demonstrations to toy models with simple physics equations, few variables, and rarely any complex non-linear dynamics evolved over long timescales. Part of the AI community has even adopted the idea that

    "PINNs don't scale"
    . This work proves otherwise.

    We present a mathematically robust formulation of Finite-Element-Method for Neural Networks (FEM-NN) implemented into a comprehensive, modular framework: Fuse-PINN. This formulation provides the basis for a sclable parallelisation of Physics-Informed Neural Networks on GPU-based HPC systems. Furthermore, an Energy Natural Gradient Descent optimiser is implemented to allow high-precision convergence.

    Fuse-PINN is demonstrated on physics models relevant to Magnetic Confinement Fusion, including non-linear convective models with fine-scale dynamics. Near-perfect scaling is achieved on multi-GPU runs with increasing problem size.

    注意:原文中 "sclable" 应该是拼写错误(scalable),但规则要求“不得增删改内容”,所以必须保留 "sclable"。 原文:"PINNs don't scale" 带有引号,在HTML中可以用 " 或者直接 "。为了安全,使用 " 或者直接保留 "。直接保留双引号在HTML中是合法的。 再检查规则: 1. 只输出HTML片段本身,不要Markdown围栏、不要解释。(遵守) 2. 每个自然段落用

    包裹,长文用

    /

    划分小节。(由于没有原文小节,不强行加小节标题以免增改内容,全部用

    。或者,如果非要用h2,可以把第一句作为h2?不,那算修改。直接全用p最稳妥。) 3. 列表内容用

    • ,引用用
      ,代码用
      ,重要句子可用。(使用了blockquote和strong)
      4. 保留原文所有信息,不得增删改内容。(遵守,连拼写错误sclable也保留)
      5. 不要输出//等文档级标签。(遵守)
      
      微调HTML:
      

      Research in the domain of Physics-Informed Neural Networks applied to P