An adaptive hybrid computational framework for study of tornado dynamics

加拿大是世界上每年龙卷风数量第二多的国家。这种灾难性的天气可能造成巨大的财产损失，甚至夺去无数人的生命。虽然计算机技术的进步有望帮助龙卷风动力学的计算分析，但目前在这一研究领域发现了一些挑战。首先，当使用传统的数值方法求解基于Navier-Stokes方程的龙卷风模型时，常见的缺点在于该流动模型的宏观性质，导致难以捕捉可靠的龙卷风动力学研究所需的更精细的流动行为。其次，商业计算流体动力学（CFD）工具被发现无法获得令人满意的复杂龙卷风现象模拟，因为它们有限的界面常常阻碍用户以理想的方式调查龙卷风的细节。第三，并行处理被认为是处理大规模计算的必要条件，而这种技术在专门用于龙卷风动力学研究的可用内部代码中很少使用。为了克服这些主要挑战，本研究工作建立在基于气体动力学的晶格玻尔兹曼模型（LBM）上，因为它具有固有的并行性，可以用于算法开发，并且可以与其他先进的数值工具耦合。OpenLB是一个免费的开源LBM软件包，具有强大的并行执行能力，将作为基础代码，并将进行以下主要的新研究任务，旨在更经济、更可靠的龙卷风模拟：(1)发展晶格Boltzmann模型的有限体积（FV）离散化，从而消除LBM背景下的许多约束，从而使用非结构化网格和隐式求解方法，这在高湍流龙卷风风的数值研究中是必不可少的。(2)为避免龙卷风随时间演变时计算域外边界速度的更新，提出了一种基于浸入边界（IB）方法的龙卷风生成相互作用模型。(3)开发新的IB-FV-LBM自适应网格工具，该工具采用一种新的后验误差估计器，对应于三维环境下基于LES的湍流模型，并将其嵌入代码中，以提高仿真结果的可靠性。(4)建立一个相关的离散元模型，使空气-碎片相互作用能够更真实地研究龙卷风动力学。在OpenLB中嵌入上述特性后，生成的自适应混合CFD工具将能够执行大量的龙卷风构建交互测试。这将为更好地理解龙卷风动力学和提高建筑设计的抗风能力提供有价值的见解和指导。通过该项目培训HQP也将使加拿大在未来的风力工程创新中受益。

Canada ranks as the second country in the world with the most tornadoes per year. This disastrous weather may inflict colossal property damage and even take numerous human lives. While computer technology advancement promises to aid computational analysis of tornado dynamics, several challenges are currently found in this research area. First, when using conventional numerical methods to solve the Navier-Stokes equations based tornado model, the common drawback lies in the macroscopic nature of this flow model, leading to difficulties in capturing finer flow behaviors necessary for a reliable tornado dynamics study. Second, commercial computational fluid dynamics (CFD) tools were found unable to attain satisfactory simulations of complex tornado phenomena, because their limited interface often prevents the user from investigating tornado details in a desirable manner. Third, parallel processing is considered compulsory in dealing with large-scale computations while this technology has been little employed in available in-house code specifically used for the tornado dynamics study. To overcome these major challenges, this research work is built upon the gas kinetics based lattice Boltzmann model (LBM) owing to its inherent parallelizability for algorithmic development and noticeable extensibility towards its coupling with other advanced numerical tools. OpenLB, a free open-source LBM software package with strong parallel execution capabilities, will serve as the basis code, and the following major new research tasks aimed at economical and more reliable tornado simulations will be conducted: (1) Develop a finite volume (FV) discretization of the lattice Boltzmann model, so that many constraints in the LBM context can be removed, leading to the use of unstructured mesh and implicit solution methods, which are essentially demanded when a highly turbulent tornadic wind is numerically investigated. (2) Develop an innovative immersed boundary (IB) approach based tornado-building interaction model in order to avoid updating the velocity at the outer boundary of computational domain when the tornado evolves with time. (3) Develop new IB-FV-LBM adaptive meshing tools powered by a novel a posteriori error estimator corresponding to the LES based turbulence model in the 3-D context, and embed it in the code to improve the reliability of simulation results. (4) Develop a pertinent discrete element model to enable the air-debris interaction for a more realistic study of tornado dynamics. With the above features embedded in OpenLB, the resulting adaptive hybrid CFD tools will be able to perform a large series of tornado-building interaction tests. Valuable insight and guidance will be gained for better understanding tornado dynamics and improving the wind-resistant capabilities in building design. The training of HQP through this program will also benefit Canada with future wind engineering innovations.