A computational tool for particle-fluid flows

CBET-0754344 Prosperetti拟议活动的知识价值。 在自然界中经常出现带有悬浮颗粒的流动（例如，沙尘暴、沉积物迁移、生物流体）、技术（例如，污染物输送、流化床燃烧器、催化反应器、悬浮液）和制造（例如制药、食品加工、喷漆）。这些流动中的一些是层流（例如，用于诊断试剂盒应用的悬浮液、浆料、单分散聚合物微球），但是颗粒体积分数大。 更多情况下，它们是湍流的，湍流对颗粒的分布和分散具有强烈的影响，并且根据条件，颗粒本身可以深刻地改变相对于单相流的湍流特征。 人们一直致力于通过分析、计算和实验来理解这种所谓的流体与粒子之间的双向耦合的本质。数值模拟已被证明是特别有价值的，尽管这些系统的固有复杂性所带来的许多必要的简化。在斯托克斯流态下，存在用于稠密悬浮液的强大技术。湍流模拟，另一方面，已经进行了非常小的颗粒浓度（颗粒体积分数的顺序为10-4或更少）的条件下，与近似为质点的近似参数化的流体力的作用下移动的颗粒。粒子对流体的作用也大多近似于叠加点力。当颗粒尺寸小于所有的流体长度尺度（包括柯尔莫哥洛夫长度）时，该过程可能是合理的。 然而，即使在这种情况下，结果也有些不令人满意，因为它们依赖于参数化的力，而不是从第一原理计算获得的力。更重要的是，有很多非常重要的情况不能用这些方法来研究：悬浮在液体中的颗粒，而不是气体（例如沉积物，化学系统），非稀释系统（例如流化床），以及许多其他情况。 在最近NSF支持的工作中，提出者及其合作者开发了一种非常有效和精确的数值方法，用于模拟具有数千个悬浮颗粒的粘性不可压缩流体流动。 该程序充分考虑了颗粒的有限尺寸，不近似它们的形状，并且完全满足颗粒表面的无滑移条件。最大颗粒雷诺数取决于网格分辨率;在每个颗粒半径10个节点的非常易于管理的分辨率下，可以处理高达几十个的雷诺数。该方法的一个特别的优点是计算时间对悬浮在流中的颗粒数的弱依赖性。到目前为止，这种情况允许模拟具有多达约1000个悬浮颗粒的流动。 为了提高分辨率，能够模拟更多的粒子和更强烈的湍流，有必要开发新的计算策略：自适应网格，具有更好的可扩展性的算法，有效的预处理等。 这里提出的工作包括：（1）这个增强代码的开发，其验证和一些初步的模拟（例如，粒子落在静止的流体中，粒子落在衰减的均匀各向同性湍流中），以及（2）使用现代软件开发方法和工具，使代码易于被他人使用，使其通过一个维护的网站免费提供给研究界，该网站提供适当的文档、测试案例、教程、详细解释等。 如提案正文所述，含悬浮颗粒的流体流动是当代流体力学中最活跃的领域之一，相关论文每年被引用数百次，甚至在发表多年后也是如此。因此，要开发的代码满足了研究界的广泛需求，这些研究界已经准备好摆脱简单模型（点粒子，Stokes流等）。有限颗粒雷诺数，较大的体积分数和许多颗粒。这种计算工具的可用性将是模拟当代社会许多问题的核心实际情况的第一步，这些问题包括能源（例如石油工业，发电，污染），环境（例如海岸侵蚀，降水）和生命（例如花粉，谷物，药物）。

CBET-0754344ProsperettiIntellectual merit of the proposed activity. Flows with suspended particles arise frequently in Nature (e.g., dust storms, sediment transport, biological fluids), technology (e.g., pollutant transport, fluidized bed combustors, catalytic reactors, suspensions), and manufacturing (e.g. pharmaceutics, food processing, spray painting). Some of these flows are laminar (e.g. suspensions, slurries, monodisperse polymeric microspheres for diagnostic kit applications), but the particle volume fraction is large. More often they are turbulent and turbulence has a strong effect on the distribution and dispersion of the particles and, depending on conditions, the particles themselves can deeply change the turbulence character with respect to single-phase flow. Much effort has been devoted to the understanding of the nature of this so-called {\em two-way coupling} between fluid and particles by analysis, computation and experiment. Numerical simulation has proven particularly valuable in spite of many simplifications rendered necessary by the inherent complexity of these systems. Powerful techniques exist for dense suspensions in the Stokes flow regime. Turbulent flow simulations, on the other hand, have been conducted in conditions of exceedingly small particle concentration (particle volume fractions of the order of 10-4 or less) with the particles approximated as mass points moving under the action of approximately parameterized fluid forces. The action of the particles on the fluid has also mostly been approximated by superposing point forces. This procedure may be justified when the particle size is smaller than all the fluid length scales, including the Kolmogorov length. However, even in this case, the results are somewhat unsatisfactory as they rest on parameterized forces rather than forces obtained from first-principles calculations. More significantly, there are very many very important situations which cannot be studied by these means: particles suspended in a liquid, rather than a gas (e.g. sediments, chemical systems), non-dilute systems (e.g. fluidized beds), and many others. In recent NSF-supported work, a very efficient and accurate numerical method for the simulation of viscous incompressible fluid flows with thousands of suspended particles was developed by the proposer and his collaborators. The procedure fully accounts for the finite size of the particles, does not approximate their shape, and exactly satisfies the no-slip condition at the particle surface. The maximum particle Reynolds number depends on the grid resolution; with a very manageable resolution of 10 nodes per particle radius, Reynolds numbers up to a few tens can be handled. A particularly strong point of the method is the weak dependence of the computational time upon the number of particles suspended in the flow. This circumstance has permitted so far the simulation of flows with up to about 1000 suspended particles. In order to improve the resolution, be able to simulate a larger number of particles and more intense turbulence, it is necessary to develop new computational strategies: adaptive grids, algorithms with better scalability properties, efficient preconditioning etc. The work proposed here consists in: (1) The development of this enhanced code, its validation and some preliminary simulations (e.g., particles falling in an otherwise quiescent fluid, particles in decaying homogeneous isotropic turbulence), and (2) The use of modern software development methods and tools to permit the code to be readily used by others, making it freely available to the research community via a maintained web site with proper documentation, test cases, tutorials, detailed explanations etc.Broader impacts resulting from the proposed activity. As noted in the body of the proposal, the flow of fluids with suspended particles is one of the most active areas in contemporary fluid mechanics, with the relevant papers cited hundreds of times a year even many years after their publication. The code to be developed therefore meets a widely felt need in the research community which is ready to move away from simple models (point particles, Stokes flow, etc.) to finite particle Reynolds numbers, larger volume fractions and many particles. The availability of this computational tool will be a first step in the simulation of practical situations which lie at the heart of many problems of contemporary society such as energy (e.g.the oil industry, power generation, pollution), the environment (e.g. coastal erosion, precipitation) and life (e.g. pollen, grains, pharmaceuticals).