The growth of computational resources in the past decades has expanded the application of Computational Fluid Dynamics (CFD) from the traditional fields of aerodynamics and hydrodynamics to a number of new areas. Examples range from the heat and fluid flows in nuclear reactor vessels and in data centers to the turbulence flows through wind turbine farms and coastal vegetation plants. However, in these new applications complex structures are often exist (e.g., rod bundles in reactor vessels and turbines in wind farms), which makes fully resolved, first-principle based CFD modeling prohibitively expensive. This obstacle seriously impairs the predictive capability of CFD models in these applications. On the other hand, a limited amount of measurement data is often available in the systems in the above-mentioned applications. In this work we propose a data-driven, physics-based approach to perform full field inversion on the effects of the complex structures on the flow. This is achieved by assimilating observation data and numerical model prediction in an iterative Ensemble Kalman method. Based on the inversion results, the velocity and turbulence of the flow field can be obtained. A major novelty of the present contribution is the non-parametric, full field inversion approach adopted, which is in contrast to the inference of coefficient in the ad hoc models often practiced in previous works. The merits of the proposed approach are demonstrated on the flow past a porous disk by using both synthetic data and real experimental measurements. The spatially varying drag forces of the porous disk on the flow are inferred. The proposed approach has the potential to be used in the monitoring of complex system in the above mentioned applications.
A turbulent eddy-viscosity surrogate modeling framework for Reynolds-Averaged Navier-Stokes simulations
用于不可压缩湍流稳态计算的雷诺平均Navier-Stokes（RANS）方程仍然是实际计算流体动力学（CFD）应用的主力。因此，速度或精度的提高有可能影响各种应用。在给定初始条件的情况下，我们引入了一个机器学习框架，用于RANS模拟稳态湍流涡粘性的代理建模。该建模策略被评估用于参数插值，同时数值地求解压力和速度方程以达到稳态，从而表示与机器学习混合的框架。我们实现{竞争}稳态的结果与显着减少的解决方案的时间相比，由Spalart-Allmaras一方程模型。这是因为所提出的方法允许相当大的松弛因子的稳态速度和压力求解器。我们的评估是针对具有相当大的网格各向异性和分离的后向台阶进行的，以代表实际的CFD应用。对于使用\textcolor{black}{either}改变入口速度条件或台阶高度的测试实验，我们看到求解时间减少约5倍。这些结果为快速探索参数空间提供了一个机会，这些参数空间在利用具有多个耦合偏微分方程的湍流闭合模型时被证明是令人望而却步的。\blfootnote{Code available publicly at \texttt{https://github.com/argonne-lcf/TensorFlowFoam}}。
Exploiting repeated matrix block structures for more efficient CFD on modern supercomputers
Heat and Mass Transfer Technological Centre, Technical University of Catalonia（热质传输技术中心，加泰罗尼亚理工大学）；Department of Information Engineering, University of Padova（帕多瓦大学信息工程系）；High Performance Computing and Visualiztion Team, SURF（SURF高性能计算与可视化团队）
Aerodynamic Drag and Heat Transfer Corrections for Dehydrated Pollen Particles: CFD-Based Modeling of Airborne Allergen Transport in Smart Urban Environments
CNRS, CORIA, UMR 6614, University of Rouen Normandy（法国国家科学研究中心CORIA联合研究单位6614，诺曼底鲁昂大学）；Université de Lille, Institut Mines-Télécom, Université d’Artois, Junia, ULR 4515-LGCgE, Laboratoire de Génie Civil et géo-Environnement（里尔大学、矿业电信联盟、阿尔图瓦大学、朱尼亚学院）；Lebanese International University (LIU)（黎巴嫩国际大学）；Gulf University for Science and Technology（海湾科学与技术大学）
A Posteriori Evaluation of a Physics-Constrained Neural Ordinary Differential Equations Approach Coupled with CFD Solver for Modeling Stiff Chemical Kinetics
Transportation and Power Systems Division, Argonne National Laboratory（阿贡国家实验室交通与动力系统分部）；Department of Mechanical Engineering, North Carolina State University（北卡罗来纳州立大学机械工程系）
DeepCFD: Efficient Steady-State Laminar Flow Approximation with Deep Convolutional Neural Networks
German Aerospace Center (DLR)（德国航空航天中心(DLR)）；German Research Center for AI (DFKI)（德国人工智能研究中心(DFKI)）；NUST School of Electrical Engineering and Computer Science（巴基斯坦国立科技大学）
