Executive Summary : | Direct numerical simulation (DNS) and Reynolds-Averaged Navier-Stokes (RANS) are the two extreme methods for computational fluid dynamic (CFD) predictions of turbulent flow fields. DNS resolves all scales of motion, requires no turbulence closure models, and is considered as the most accurate method. However, resolving all scales (largest to smallest) makes DNS computationally very expensive, and thus unaffordable for many CFD users. On the other extreme, RANS tends to resolve only the largest scale of motion while modeling all other scales. RANS is computationally inexpensive andis widely affordable, but its over-reliance on turbulence closure models makes is susceptible to inaccuracies. In between DNS and RANS, there are "intermediate" methods. Such methods, also called the scale-resolving simulations (SRS), resolve a range of motion which is intermediate between RANS and DNS, require intermediate levels of modeling, and require intermediate quantum of computational resources. Large eddy simulation (LES), detached eddy simulation (DES) and Partially-averaged Navier-Stokes (PANS) methodology are some examples of such intermediate methods. The focus of the current proposal is the PANS method.
In recent years the PANS method has been shown to have superior performance than RANS in a variety of flow fields. However, a vast scope of improvement still exists, especially in massively separated flow fields past bluff bodies. Experiments show that the unresolved stress in the separated regions of a flow field is significantly misaligned with the resolved strain-rate. Like RANS, PANS too relies on turbulence closure models. To date, most turbulence models used in PANS paradigm use the so-called linear eddy viscosity models (LEVM), which assume the unresolved stress to be proportional to the resolved strain-rate tensor. This assumption constraints the turbulent stress tensor to be invariably aligned with the resolved strain-rate tensor. In highly separated flow fields, this behaviour is not consistent with experimental observations and is a reason behind inaccuracy of CFD predictions. Accordingly, the overarching objective of this project is to augment the PANS methodology with a non-linear eddy viscosity closure model (NLEVM). |