CFD discrete adjoint benchmarks

This data set contains meshes for three discrete adjoint benchmarks along with ADflow configuration scripts to reproduce the results in the following paper: Gaetan K.W. Kenway, Charles A. Mader, Ping He, and Joaquim R. R. A. Martins. Effective adjoint approaches in Computational Fluid Dynamics, Progress in Aerospace Sciences, 2019. In this paper we used two CFD solvers that have discrete adjoints: ADflow (see link below to access code) and OpenFOAM (with an internal discrete adjoint). If you are interested in running your adjoint solver for comparison, we recommend to use the provided meshes (either in structured or unstructured format) and the same flow conditions and CPU cores as described in the paper. If you would like to generate your own meshes, we suggest using a similar cell number for each benchmark. When comparing the performance, please report the runtime (in Taubench work unit), peak memory usage (resident set size in GB), number of CPU cores and mesh cells, as well as a description of your adjoint solver and the computing platform (including the Tau reference runtime). The cases are based on the following benchmarks developed by the AIAA Aerodynamic Design Optimization Discussion Group (ADODG): • ADODG3: A low-speed rectangular wing. • ADODG4: The Common Research Model (CRM) wing. • ADODG5: The CRM wing-body-tail configuration. The filename suffixes are: *_Multiblock_Structured_Mesh.cgns: Multiblock structured mesh files for ADflow. *_Overset_Structured_Mesh.cgns: Multiblock overset mesh files for ADflow. *_Singleblock_Unstructured_Mesh.cgns: These meshes are identical to *_Multiblock_Structured_Mesh.cgns except that they are in single block unstructured mesh format. The far field, symmetry, and wall boundary names are FARFIELD, SYMMETRY, and WING, respectively. *_FFD.xyz: Free-form deformation (FFD) files for defining the shape design variables. *_ADflow_Run_Script.py: Python scripts for reproducing the ADflow adjoint results in the paper. See detailed instructions in the scripts. Code_Version.txt: The required codes to produce the results, including links and versions.本数据集包含三个离散伴随基准测试的网格文件，以及可复现下述论文结果的ADflow配置脚本：Gaetan K.W. Kenway、Charles A. Mader、Ping He与Joaquim R. R. A. Martins发表于《航空进展（Progress in Aerospace Sciences）》2019年刊的论文《计算流体动力学中的有效伴随方法（Effective adjoint approaches in Computational Fluid Dynamics）》。 本文中我们使用了两款搭载离散伴随模块的计算流体动力学（Computational Fluid Dynamics, CFD）求解器：ADflow（访问代码请见下方链接）以及内置离散伴随模块的OpenFOAM。若您希望使用自研伴随求解器开展对比测试，建议采用本数据集提供的网格（支持结构化或非结构化格式），并遵循论文中规定的相同流场条件与CPU核心数。若您计划自行生成网格，建议为每个基准测试设置相近的网格单元数量。 在进行性能对比时，请上报以下信息：运行时长（以Taubench工作单元为单位）、峰值内存使用率（以GB计量的驻留集大小（resident set size））、CPU核心数与网格单元数，同时需附上您所用伴随求解器与计算平台的详细说明（包括Tau参考运行时长）。 本数据集包含的三个基准测试均由美国航空航天学会（AIAA）气动设计优化讨论组（Aerodynamic Design Optimization Discussion Group, ADODG）开发，具体如下： • ADODG3：低速矩形机翼 • ADODG4：通用研究模型（Common Research Model, CRM）机翼 • ADODG5：CRM翼身尾组合构型 文件名后缀说明如下： - *_Multiblock_Structured_Mesh.cgns：适用于ADflow的多块结构化网格文件 - *_Overset_Structured_Mesh.cgns：适用于ADflow的多块重叠结构化网格文件 - *_Singleblock_Unstructured_Mesh.cgns：该类网格与*_Multiblock_Structured_Mesh.cgns内容完全一致，仅采用单块非结构化网格格式 - 远场、对称面与壁面的边界名称分别为FARFIELD、SYMMETRY与WING - *_FFD.xyz：用于定义形状设计变量的自由形式变形（Free-form deformation, FFD）文件 - *_ADflow_Run_Script.py：用于复现论文中ADflow伴随求解结果的Python脚本，脚本内附有详细操作指南 - Code_Version.txt：包含复现结果所需的代码信息，包括代码链接与版本号。