AeroJEPA: Learning Semantic Latent Representations for Scalable 3D Aerodynamic Field Modeling
作者: Francisco Giral, Abhijeet Vishwasrao, Andrea Arroyo Ramo, Mahmoud Golestanian, Federica Tonti, Adrian Lozano-Duran, Steven L. Brunton, Sergio Hoyas, Hector Gomez, Soledad Le Clainche, Ricardo Vinuesa
分类: cs.LG
发布日期: 2026-05-07
💡 一句话要点
提出AeroJEPA架构,通过联合嵌入预测实现可扩展的3D空气动力学场建模与语义表征学习。
🎯 匹配领域: 支柱二:RL算法与架构 (RL & Architecture) 支柱七:动作重定向 (Motion Retargeting)
关键词: 空气动力学 联合嵌入预测架构 隐式神经表示 代理模型 语义表征学习 计算流体力学
📋 核心要点
- 现有空气动力学代理模型在处理大规模3D流场时扩展性较差,且难以生成具备语义分析价值的潜在表征。
- AeroJEPA采用联合嵌入预测架构,通过解耦潜在空间预测与流场分辨率,实现高效的流场建模与语义特征提取。
- 实验证明该方法在复杂高保真流场及跨音速机翼设计中表现优异,支持概念向量运算及约束下的设计空间优化。
📝 摘要(中文)
空气动力学代理模型常用于替代高保真计算流体力学(CFD)评估,但在处理大规模3D流场时面临扩展性差及潜在空间缺乏语义分析能力的问题。本文提出了AeroJEPA,一种用于流场建模的联合嵌入预测架构(JEPA)。该方法不直接从几何形状预测完整流场,而是通过几何与工况的上下文潜在表示来预测流场的潜在目标表示,并可选地通过连续隐式解码器重建流场。这种设计将潜在预测与场分辨率解耦,同时促进了潜在空间的语义组织。在HiLiftAeroML和SuperWing数据集上的实验表明,AeroJEPA在作为连续代理模型时具有竞争力,能自然扩展至高分辨率输出,并学习到包含几何与空气动力学信息的语义表征,支持插值、线性探测及设计优化。
🔬 方法详解
问题定义:现有代理模型在处理高分辨率3D空气动力学数据时计算开销巨大,且模型学习到的隐空间往往是“黑盒”,缺乏对流场物理特征的语义理解,难以直接用于工程设计优化。
核心思路:引入JEPA(联合嵌入预测架构)范式,将预测任务从像素/网格空间转移到潜在空间。通过学习几何与工况的映射,强制模型在潜在空间中捕捉流场的本质语义,从而实现对复杂流场的解耦建模。
技术框架:系统包含上下文编码器(处理几何与工况)、目标编码器(处理流场状态)以及预测器。预测器基于上下文潜在表示预测目标潜在表示,最后通过连续隐式解码器(Implicit Decoder)将潜在向量映射回空间坐标下的物理场。
关键创新:核心创新在于将流场预测任务解耦为“潜在空间预测”与“场重建”。这种架构不仅降低了对高分辨率网格的直接依赖,还通过自监督学习机制使得潜在空间具备了线性可分性和语义可解释性。
关键设计:采用连续隐式解码器(如MLP架构)实现空间分辨率无关的重建;利用对比学习或预测性损失函数优化潜在空间,使其支持概念向量算术(Concept-vector arithmetic)和约束优化任务。
🖼️ 关键图片
📊 实验亮点
AeroJEPA在HiLiftAeroML和SuperWing数据集上展现了卓越性能。实验显示,该模型不仅能准确重建大规模边界层流场,其学习到的潜在空间还支持线性探测(Linear Probing)和概念向量运算,证明了其在设计优化任务中的有效性。相比传统方法,它在保持高精度的同时,实现了对复杂几何流场的高效泛化与可控生成。
🎯 应用场景
该研究主要应用于航空航天领域的飞行器气动外形设计与优化。通过提供高效、可解释的代理模型,AeroJEPA能够显著加速高保真CFD仿真流程,支持工程师在设计阶段进行快速的参数化探索、概念插值及基于语义的约束优化,从而缩短研发周期并提升设计质量。
📄 摘要(原文)
Aerodynamic surrogate models are increasingly used to replace repeated high-fidelity CFD evaluations in many-query design settings, but current approaches still face two important limitations: they often scale poorly to the very large fields arising in realistic 3D aerodynamics, and they rarely produce latent representations that are directly useful for analysis and design. We introduce AeroJEPA, a Joint-Embedding Predictive Architecture for aerodynamic field modeling that addresses both issues. Rather than predicting the full flow field directly from geometry, AeroJEPA predicts a target latent representation of the flow from a context latent representation of the geometry and operating conditions, and optionally reconstructs the field through a continuous implicit decoder. This formulation decouples latent prediction from field resolution while encouraging the latent space to organize semantically. We evaluate AeroJEPA on two complementary datasets: HiLiftAeroML, which stresses the method in a high-fidelity regime with extremely large boundary-layer fields, and SuperWing, which tests large-scale generalization and latent-space optimization over a broad family of transonic wings. Across these benchmarks, AeroJEPA is competitive as a continuous surrogate for aerodynamic fields, scales naturally to high-resolution outputs, and learns context and predicted latents that encode geometry and aerodynamic quantities not used directly as supervision. We further show that the resulting latent space supports controlled interpolation, linear probing, concept-vector arithmetic, and a constrained design latent-optimization experiment. These results suggest that predictive latent learning is a promising direction for scalable and design-meaningful aerodynamic surrogate modeling.