Multi-scale Predictive Simulation Methods for Turbulent Flow

dale I. Pullin：这项研究将在几个新的方向上发展我们对湍流预测和数值模拟的能力。目标是将传统的大涡模拟（LES）的现有范例扩展到从能量范围到Kolmogorov和Batchelor尺度的长度/时间尺度，用于一些湍流统计，同时也解决亚网格尺度建模中可怕的近壁问题。这项工作将把拉伸涡（SV）、亚网格尺度（SGS）模型扩展到多尺度LES (MSLES)，几乎涵盖了湍流活动长度和时间尺度的整个范围，用于一些输运统计。目的是使目前的类les方法实质上更接近理想的DNS（直接数值模拟）。该研究将开发算法，结合Navier-Stokes方程的拉伸涡解，作为LES的基础，以及湍流精细尺度的综合模型，以产生MSLES能力。这种耦合应该能够将几个湍流混合流的统计数据从分辨尺度延续到亚网格尺度。应用将包括几种典型的湍流，包括衰减湍流、平均密度梯度下的被动标量混合和变密度湍流混合。对于最后一类流，还将执行直接数值模拟（DNS）运行，以验证MSLES。虽然不期望真正的点和时间方向的DNS解析，但与传统的LES技术相比，MSLES方法可以在预测许多湍流统计数据，特别是湍流混合流方面提供实质性的改进。这项研究还将进一步开发、测试和应用一种亚网格尺度的方法，这种方法是专门为壁面湍流的近壁区域设计的，这是我们预测湍流模拟能力的主要障碍。两个新概念已经形成。第一种是一种特殊的近壁亚网格尺度元素，它基于局部内尺度与壁法向平均和壁平行滤波相结合的思想。第二种是扩展的涡旋，sgs通量的形式，以模拟近壁的主要动力学行为，纵向涡旋在其流向动量的壁法向运输。该方法将在位于对数律区域下限的虚拟墙处实现边界条件。粘性子层没有被分解。智力优势：这项研究将提供一个基于物理的分析框架，将传统的LES扩展到多尺度LES，并解决近壁、亚电网尺度的建模问题。在大雷诺数范围内，大规模流体动力学模拟范式对科学和工程的许多不同领域的应用产生了巨大的影响。更广泛的影响：这项研究将扩展复杂流体行为的多维数值模拟能力。MSLES将在数值模拟中提供一些类似DNS的保真度度量，但在计算资源方面的成本仅为DNS的一小部分。随着桌面计算能力的持续增长，我们预计这种类型的模拟将在十年内在工程应用程序级别上随时可用。

CBET- 0651754Dale I. Pullin This study will develop our capability for the predictive, numerical simulation of turbulent flows in several new directions. The goal is to both extend the present paradigm of conventional large-eddy simulation (LES) to span length/time scales from the energy-containing range to the Kolmogorov and Batchelor scales, for some turbulence statistics, and also to tackle the formidable near-wall problem in sub-grid-scale modeling. The work to be performed will extend the stretched-vortex (SV), subgrid-scale (SGS) model to multi-scale LES (MSLES) across almost the full-range of active length and time scales in turbulent flow, for some transport statistics. The aim is to move present LES-like methodology substantially closer to the ideal of DNS (direct numerical simulation). The study will develop algorithms that couple stretched-vortex solutions of the Navier-Stokes equations both as a basis for LES, and as a comprehensive model of turbulent fine scales, to produce an MSLES capability. This coupling should enable continuation of the statistics of several turbulent mixing flows from resolved to subgrid scales. Applications will comprise several canonical turbulent flows including decaying turbulence, passive-scalar mixing in the presence of a mean density gradient and variable-density turbulent mixing. For this last class of flows, direct-numerical simulation (DNS) runs will also be performed for validation of the MSLES. While not expecting true point- and time-wise DNS resolution, the MSLES methodology can provide a substantial improvement in the prediction of many turbulent statistics, particularly for turbulent mixing flows, over conventional LES techniques. The study will also further develop, test and apply a subgrid-scale approach designed specifically for the near-wall region of wall-bounded turbulent flows, a major roadblock to our capability for predictive simulation of turbulent flows. Two new concepts have been formulated. The first consists of a special near-wall, subgrid-scale element that is based on the idea of local inner scaling combined with wall-normal averaging and wall-parallel filtering. The second is an extension of the stretched-vortex, SGS-flux formalism, to model the principal dynamical behavior of near-wall, longitudinal vortices in their wall-normal transport of streamwise momentum. The method will implement boundary conditions at a virtual wall positioned at the lower limit of the log-law region. The viscous sublayer is not resolved. Intellectual merit: This study will provide a physically-based, analytical framework for the extension of conventional LES to both multiscale LES, and to the solution of the near-wall, subgrid-scale modeling problem. The paradigm of large-scale fluid-dynamical simulation has had an enormous impact on many diverse areas of science and engineering applications over a large range of Reynolds numbers.Broader impact: This research will extend capability to perform multi-dimensional, numerical simulations of complex fluid behavior. MSLES will provide some measure of DNS-like fidelity in numerical simulation but at a small fraction of the DNS cost in computational resources. With the continued explosion in the availability of desktop computing power, we expect this style of simulation to become readily available at the engineering application level within a decade.