EmmiAI/Emmi-Wing

该存储库包含在NeurIPS 2025的物理科学机器学习研讨会上提出的数据集，用于评估神经代理模型在高度湍流跨音速条件下的性能。数据集包含248个案例，每个案例包含表面和体积的位置及场值（如压力、速度、涡度等），以及计算阻力和升力系数所需的信息。数据集遵循工业标准，使用OpenFOAM进行模拟和网格生成，并采用Spalart-Allmaras湍流模型。此外，数据集还提供了设计参数（包括马赫数和雷诺数）以及可能存在数值伪影的案例列表。

This repository contains the dataset proposed in the paper presented at the Workshop on ML for the Physical Sciences at NeurIPS 2025, used for evaluating neural surrogate models in highly-turbulent transonic regimes. The dataset includes 248 cases, each containing surface and volume positions and field values (e.g., pressure, velocity, vorticity), along with information needed to compute drag and lift coefficients. The dataset adheres to industrial standards, using OpenFOAM for simulations and mesh generation, and employs the Spalart-Allmaras turbulence model. Additionally, it provides design parameters (including Mach and Reynolds numbers) and a list of cases that may contain numerical artifacts.