Skip to main content
QUICK REVIEW

[论文解读] GeoPT: Scaling Physics Simulation via Lifted Geometric Pre-Training

Haixu Wu, Wenhao Gao|arXiv (Cornell University)|Feb 23, 2026
Model Reduction and Neural NetworksPhysics and Astronomy被引用 0
一句话总结

GeoPT 通过使用带有合成速度场的提升动力学,从几何中预训练物理骨架,使在工业级流体和固体仿真中的数据效率高、微调更快。

ABSTRACT

Neural simulators promise efficient surrogates for physics simulation, but scaling them is bottlenecked by the prohibitive cost of generating high-fidelity training data. Pre-training on abundant off-the-shelf geometries offers a natural alternative, yet faces a fundamental gap: supervision on static geometry alone ignores dynamics and can lead to negative transfer on physics tasks. We present GeoPT, a unified pre-trained model for general physics simulation based on lifted geometric pre-training. The core idea is to augment geometry with synthetic dynamics, enabling dynamics-aware self-supervision without physics labels. Pre-trained on over one million samples, GeoPT consistently improves industrial-fidelity benchmarks spanning fluid mechanics for cars, aircraft, and ships, and solid mechanics in crash simulation, reducing labeled data requirements by 20-60% and accelerating convergence by 2$ imes$. These results show that lifting with synthetic dynamics bridges the geometry-physics gap, unlocking a scalable path for neural simulation and potentially beyond. Code is available at https://github.com/Physics-Scaling/GeoPT.

研究动机与目标

  • 通过降低对代价高昂的求解器生成数据的依赖,推动可扩展的神经仿真。
  • 引入动力学提升的几何预训练以编码几何-动力耦合。
  • 在具有合成动力学的大型几何数据集上进行预训练,以学习一个带动力学感知的先验。
  • 在多样的物理任务上对预训练模型进行微调,利用任务特定的动力输入。

提出的方法

  • 用合成速度场来扩充几何,创建动力学提升的输入。
  • 训练物理骨架(Transolver),在合成动力学下预测几何特征轨迹(提升预训练目标)。
  • 使用预训练损失,最小化预测轨迹与几何特征轨迹随时间的差异(L_pre^lifted)。
  • 通过用编码仿真设置S的任务特定速度场V_S替换随机动力学来微调,并最小化物理损失(L_fine)。
  • 将轨迹离散化为三步,并在公开几何存储库中的超过一百万个几何-动力样本上进行预训练。
Figure 1 : Neural aerodynamics simulation on DrivAerML (Ashton et al. , 2024 ) based on Transolver (Wu et al. , 2024 ) backbone. Geometry-only pre-training and conditioning refer to pre-training by predicting vector distance (Faugeras and Gomes, 2000 ) of given positions and utilizing geometry repre
Figure 1 : Neural aerodynamics simulation on DrivAerML (Ashton et al. , 2024 ) based on Transolver (Wu et al. , 2024 ) backbone. Geometry-only pre-training and conditioning refer to pre-training by predicting vector distance (Faugeras and Gomes, 2000 ) of given positions and utilizing geometry repre

实验结果

研究问题

  • RQ1动力学提升的几何预训练能否从未标注的几何体中学习到物理感知先验?
  • RQ2GeoPT 是否减少下游物理数据需求并在多种工业仿真中加速收敛?
  • RQ3GeoPT 与原生几何预训练或几何条件化基线在物理任务上的表现差异如何?
  • RQ4提升的预训练架构是否与不同的物理骨架和任务兼容?
  • RQ5该方法是否能扩展到流体和固体之外的领域,如辐射度/辐射回传?

主要发现

  • GeoPT 在多个工业基准上将所需的物理数据量减少了 20-60%。
  • GeoPT 在微调阶段的收敛速度提升最高可达 2×。
  • 在超过一百万个无求解器几何样本上的预训练,在空气动力学、水动力学和碰撞仿真等领域带来稳健的改进。
  • 几何提升的监督产生了与物理对齐的表示,优于仅几何预训练和几何条件化基线。
  • GeoPT 展现出随模型规模和预训练数据多样性扩展的可扩展性,并且对辐射度等其他物理领域具有泛化能力。
Figure 2 : GeoPT offers a way to scale up neural simulators with off-the-shelf geometries and enables fast fine-tuning for various physics.
Figure 2 : GeoPT offers a way to scale up neural simulators with off-the-shelf geometries and enables fast fine-tuning for various physics.

阅读本论文

在 Nubint 阅读器中打开原文,边读边就任意段落提问。

在 Nubint 中阅读本论文

本解读由 AI 根据论文原文生成,引用前请核对原文。