Hybrid LES-RANS-Coupling and Wall Modeling for Complex Flows with Separation

目前法德联合研究项目的目标是发展迫切需要的技术，以便将大涡模拟(LES)及其导数应用于具有实际意义的高雷诺数湍流流动。两个主要的发展战略将被考虑，这两个战略密切相关，相辅相成。这些模型一方面是适当的墙模型与适当的(自适应)次网格尺度模型相结合。其目标是在近壁区域之间架起桥梁，从而实现高重复性模拟，而无需使用极其精细且耗费CPU时间的网格来解决墙的直接邻近问题。另一方面，将开发混合的LEAN-RAN方法，试图将模拟(理想情况下不是领域)分成RANS部分和LES部分。通常，RAN对于需要合理的CPU时间和内存的附加边界层是足够的，其中也可以应用LES，但需要极大的资源。相反，RAN在具有大规模分离和/或具有大尺度涡旋结构的流动中常常失效。在这里，Les无疑是更好的选择。基本概念是结合这两种方法的优点，至少为一类特殊的流产生一个最优解。真正的非分区混合技术是优选的，因为它避免了RAN和LES区域的预定义，从而导致或多或少自动地选择合适的模拟技术的方法。尽管组合LEAN-RANS方法的想法并不新鲜，但仍有许多悬而未决的问题需要回答。这尤其包括对LES和RAN之间适当的耦合技术、自适应控制机制和适当的SGS/RANS模型的需求。由于近似壁面边界条件部分依赖于RANS模型，而经典壁面函数是近壁区RANS模型的极限情况，因此，本提议的两个主题，即壁面模拟和LES-RANS耦合，是强相互关联的。为了研究上述悬而未决的问题，在第一步中，不可避免地要考虑一些几何上简单的试验情况，包括压力诱导流动分离和随后的再附着，例如试验构型T2，即周期性山丘上的流动。当然，最终目标是具有实际意义的流动配置，包括高Re情况。因为所有合作伙伴都已经研究了大攻角翼型周围的流动，所以这样的配置(测试用例T6)将作为第二个更复杂的测试用例。作为一个长期目标，将研究实际感兴趣的几何三维构型。一种选择是围绕简化的汽车模型(Ahmed车身，测试用例T4)流动。特别是对于混合方法，所选择的流动形态应该包括RAN可以合理解决的区域，以及必然需要LES而不是RAN的大尺度流动分离等复杂流动特征。这些类型的流对墙建模技术也是极具挑战性的，它允许基于相同的测试用例直接比较和评估这两种方法。LIMSI Orsay和LSTM Erlangen的CFD团队在复杂流动的LES和DNS的开发和应用方面拥有长期的丰富经验。在Ecole Centrale de Nantes，数值小组在复杂几何的RAN计算方面拥有多年的经验。由于这些主要专业领域的互补性，预计合作将特别富有成果，从而产生强大的协同效应。

The objektive of the present joint French-German research project is the development of urgently required techniques in order to apply large-eddy simulation (LES) and its derivatives to high Reynolds number turbulent flows of practical interest. Two main development strategies will be considered which are closely related and belong together. These are on the one hand appropriate wall models combined with suitable (auto-adaptive) subgrid scale models. The goal is to bridge the near-wall region and therefore allow high Re simulations without resolving the direct vicinity of the wall with extremely fine and CPU-time consuming grids. On the other hand, hybrid LES-RANS approaches will be developed, which try to split up the simulation (ideally not the domain) into a RANS part and an LES part. Typically, RANS is adequate for attached boundary layers requiring reasonable CPU-time and memory, where LES can also be applied but demands extremely large resources. Contrarily, RANS often fails in flows with massive separation and/or with large-scale vortical structures. Here, LES is without doubt the better choice. The basic concept is to combine the advantages of both methods yielding an optimal solution at least for a special class of flows. A real non-zonal hybrid technique is preferred since it avoids the predefinition of RANS and LES regions leading to an approach where the suitable simulation technique is chosen more or less atitomatically. Although the idea of combined LES-RANS methods is not new, a variety of open questions has to be answered. This includes in particular the demand for appropriate coupling techniques between LES and RANS, adaptive control mechanisms, and proper SGS/RANS models. Since approximate wall boundary conditions partially rely on RANS modeling and the classical wall function is a limiting case of a RANS model in the near-wall region, both topics of the present proposal, i.e., wall modeling and LES-RANS coupling, are strongly interconnected. In order to investigate the above mentioned open questions, in the first step it is definitely inevitable to consider some geometrically simple test cases including pressure-induced flow separation and subsequent reattachment, e.g. test configuration T2, the flow over periodic hills. The final objectives are, of course, flow configurations of practical interest including high-Re cases. Because all partners already studied the flow around airfoils at high angles of attack, such a configuration (test case T6) will serve as a second, more complex test case. As a long-term goal a geometrically three-dimensional configuration of practical interest will be investigated. One option is the flow around a simplified car model (Ahmed body, test case T4). Especially for the hybrid methods, the flow configurations chosen should include regions which can be reasonably resolved by RANS and additionally complex flow features such as large-scale flow separation which definitely require LES instead of RANS. These kinds of flows are also extremely challenging for the wall modeling techniques which allows a direct comparison and evaluation of both approaches based on the same test cases. The CFD groups at LIMSI Orsay and LSTM Erlangen have long lasting experience in the development and application of LES and DNS for complex flows. At Ecole Centrale de Nantes, the numerics group has many years of experience in RANS computations for complex geometries. Owing to these complementary fields of main expertise, the cooperation is expected to be particularly fruitful leading to strong synergy effects.