Accurate, geometry-free, computational fluid dynamics

计算流体动力学（CFD）是工程和物理学中的一种关键数值方法，其应用范围从汽车和机翼周围的风流建模到通过喷墨打印机喷嘴的墨水流动建模。其全球市值在2018年达到16亿美元，预计到2024年将增长到31亿美元。许多类的问题需要在复杂的时间演变的几何形状，或在有固体，液体和气体相的混合的情况下，流体流动建模。在这些情况下，依赖于自适应笛卡尔网格或静态扭曲网格的标准现成技术可能变得低效。在这个项目中，我们将进一步开发一种令人兴奋的新的无几何CFD方法，与迄今为止使用的方法不同，它也是准确的。这开辟了新的问题，这两个“几何自由”的模拟是有利的，但准确性也是至关重要的。这包括对移动障碍物、齿轮、旋转叶片和类似物周围的流体流动以及沿着渠道和大坝周围的水流进行建模。我们将采用STFC资助的IAA项目-无网格有限体积（MFV）中开发的一种新的精确、无几何、CFD技术，并将其应用于开源、优化的工程CFD代码DualSPHysics中。这是一个GPU加速的代码，已经包括了许多有用的模块，用于建模复杂的固体，移动，边界重要的广泛的CFD应用。我们还将在此代码中实现一种新的“粒子分裂”技术，使我们能够有效地捕获小尺度的边界特征，如粗糙表面和孔洞。DualSPHysics的开源软件许可证将使我们能够维护我们的MFV IP。具体来说，我们的改进可以用于闭源商业应用，使我们能够将我们的新技术带到TRL 8。

Computational Fluid Dynamics (CFD) is a key numerical method in engineering and physics, with applications ranging from modelling wind flow around cars and wings to modelling the flow of ink through an inkjet printer nozzle. Its global market value reached $1.6 billion in 2018 and is expected to grow to $3.1 billion by 2024. Many classes of problem require fluid flow to be modelled in complex time-evolving geometries, or in situations where there are mixes of solid, liquid and gas phases. In these situations, standard off-the-shelf techniques that rely on adaptive Cartesian grids, or static distorted meshes can become inefficient. In this project, we will further develop an exciting new geometry-free CFD method that, unlike previous methods used to date, is also accurate. This opens-up new classes of problem for which both "geometry-free" simulations are advantageous, but for which accuracy is also paramount. This includes modelling fluid flow around moving barriers, gears, rotating blades and similar, and water flow along channels and around dams. We will take a new accurate, geometry-free, CFD technique developed in an STFC funded IAA project - Meshless Finite Volume (MFV) - and implement it in an open-source, optimised, engineering CFD code, DualSPHysics. This is a GPU accelerated code that already includes many useful modules for modelling complex solid, moving, boundaries important for a wide range of CFD applications. We will also implement a new 'particle splitting' technique in this code, allowing us to efficiently capture small-scale boundary features like rough surfaces and holes. The open-source software licence for DualSPHysics will allow us to maintain our MFV IP. Specifically, our improvements can be used for closed-source commercial applications, allowing us to take our new technology to TRL 8.