Multiscale Simulations of Multiphase Systems

伍斯特理工学院多相流的直接数值模拟在过去几年中取得了巨大的进展，其中流动的所有连续尺度都得到了完全分辨。计算机能力的提高和新的算法现在使人们有可能对数百个粒子(液滴、气泡和固体)的非稳定运动进行足够长的时间跟踪，从而可以计算出流体混合物的有意义的平均值。然而，尽管有大量的信息和理解，域名系统正在提供相对复杂的流动，但实际系统仍然提供了挑战，仍然限制了可以模拟的系统的范围，即使我们将我们的研究限制在连续统理论很好地描述的系统上。正如人们可能预料的那样，问题是规模问题。从模拟开始，我们可以称之为“主导小尺度”的完全解决，我们经常发现，多相流也可以产生比主导流动尺度小得多的特征，包括非常薄的薄膜、细丝和液滴。通常，这些“特征”之间存在明显的尺度分离，通常局部演化的惯性效应相对较小，孤立地，这些特征通常由分析模型很好地描述。本文提出了一种将模型等小尺度特征融入多相流数值模拟的方法。这一发展将包括理论框架和数值方法的实施。这里要解决的另一个问题是如何将多相流的解析结果用于此类流动的工程模型，尽管这是一个更具探索性的问题。在用所谓的双流体模型模拟分散流方面已经取得了相当大的进展，但对于每一相既包括连续部分又包括分散部分的流动来说，所取得的进展要少得多。总体而言，多相流平均特性的模拟远远落后于均匀湍流的模拟，但多相流数值模拟的最新进展使得推进这类模拟变得迫切。这些方法将在一个通常表现出相当复杂的物理系统的背景下开发，这具有巨大的实际重要性。液体的雾化存在于许多应用中，虽然喷雾的计算模型已经达到了相当高的复杂程度，但这些模型都依赖于初始雾化的非常粗略的近似。由于初始液滴尺寸分布通常对液滴随后的演变至关重要，因此对雾化的缺乏了解是喷雾计算的主要瓶颈。拟议活动的智力优势在于提高了对多相流进行数值模拟的能力，并利用这些能力大大提高了对雾化的理解。这里提出的这项研究的更广泛的影响是双重的。首先，它将有助于重新定义多相流研究的技术水平，通过演示如何使用计算来检查到目前为止已经尝试过的更复杂的问题。其次，它还将帮助培训研究生在多相流中使用域名系统。

ABSTRACT - 0522581Worcester Polytechnic InstituteDirect Numerical Simulations (DNS) of multiphase flows, where all continuum scales of the flow are fully resolved, have progressed enormously in the last few years. Increase in computer power and new algorithms now make it possible to follow the unsteady motion of several hundred particles (drops, bubbles and solids) for long enough times so that meaningful averages for the fluid mixture can be calculated. However, in spite of the enormous information and understanding that DNS are providing for relatively complex flows, real systems provide challenges that still limit the range of systems that can be simulated, even when we limit our studies to systems well described by continuum theories. The problem is, as one might expect, one of scale. Starting with simulations where what we might call the ''dominant small-scales'' are fully resolved, it is frequently found that multiphase flows also can generate features much smaller than the dominant flow scales, consisting of very thin films, filaments, and drops. Frequently there is a clear separation of scales between these ''features,'' usually inertia effects are relatively small for the local evolution, and in isolation these features are often well described by analytical models. It is proposed here to develop an approach to incorporate such small-scale features as models into numerical simulations of multiphase flows. The development will consist of both a theoretical framework and an implementation into a numerical method. The other problem that will be addressed here, although at a more exploratory level, is how to use the results of DNS of multiphase flows for engineering models of such flows. Considerable progress has been made in the modeling of disperse flows using the so-called two-fluid model, but much less has been achieved for flows with where each phase consists of both a continuous and a disperse part. In general, modeling of the averaged properties of multiphase flows lags far behind what has been achieved homogeneous turbulent flows, but the recent progress in DNS of multiphase flows makes it urgent to advance such modeling. These approaches will be developed in the context of a physical system that usually exhibits considerable complexity, and that is of immense practical importance. Atomization of liquids is found in a multitude of applications and while computational modeling of sprays has reached a fairly high degree of sophistication, such models all rely on very crude approximations for the initial atomization. As the initial droplet size distribution is generally critical for the subsequent evolution of the drops, the lack of understanding of the atomization is a major bottleneck for computations of sprays.The intellectual merit of the proposed activity lies in the advancement of the capabilities to conduct numerical simulations of multiphase flows and the use of these capabilities to provide greatly improved understanding of atomization. The broader impact of the study proposed here are two-fold. First of all, it will help redefine the state of the art in multiphase flow studies by demonstrating how computations can be used to examine much more complex problems that have been attempted until now. Secondly, it will also help train graduate students in the use of DNS for multiphase flows.