Benchmarking Autoregressive Conditional Diffusion Models for Turbulent Flow Simulation

This archive contains spatio-temporal data from simulations of the Navier-Stokes equations: First, simulations of an incompressible wake flow at different Reynolds numbers simulated with PhiFlow. Second, a transonic cylinder flow simulated with SU2 at different Mach numbers. Finally, a curation of data from the Johns Hopkins Turbulence Database (JHTDB) is included, that features an isotropic turbulence flow simulated with a direct numerical simulation.
The isotropic turbulence data is made available under the Open Data Commons Attribution License (ODC-By) ( http://opendatacommons.org/licenses/by/). This means the data is open to use, but requires attribution to the original creators from the JHTDB (see https://turbulence.pha.jhu.edu/citing.aspx).
Furthermore, pretrained neural network model weights for flow prediction on each data set are provided, that can be used as described in more detail in our source code.

本归档文件包含基于纳维-斯托克斯方程（Navier-Stokes equations）的数值模拟所生成的时空数据集，具体涵盖三类内容：其一，采用PhiFlow模拟的不同雷诺数（Reynolds numbers）下的不可压缩尾迹流数值模拟结果；其二，采用SU2模拟的不同马赫数（Mach numbers）下的跨音速圆柱绕流数值模拟结果；其三，收录了来自约翰·霍普金斯湍流数据库（Johns Hopkins Turbulence Database, JHTDB）的经整理数据集，该数据集包含采用直接数值模拟（direct numerical simulation）得到的各向同性湍流流场数据。本各向同性湍流数据集采用开放数据Commons署名许可（Open Data Commons Attribution License, ODC-By）发布，许可详情可访问 http://opendatacommons.org/licenses/by/ 查询。该许可允许自由使用数据集，但需注明数据源自JHTDB的原始创作者，具体引用规范详见 https://turbulence.pha.jhu.edu/citing.aspx。此外，本归档还提供了针对各数据集的流场预测预训练神经网络模型权重，其详细使用方法可参见配套源代码中的说明。