{"id":1118,"date":"2013-09-26T17:02:44","date_gmt":"2013-09-26T06:02:44","guid":{"rendered":"https:\/\/www.computationalfluiddynamics.com.au\/?p=1118"},"modified":"2013-09-26T17:02:44","modified_gmt":"2013-09-26T06:02:44","slug":"cfd-turbulence-part5-scale-resolving-simulations_srs","status":"publish","type":"post","link":"https:\/\/www.leapaust.com.au\/blog\/cfd\/cfd-turbulence-part5-scale-resolving-simulations_srs\/","title":{"rendered":"Turbulence Part 5 – Overview of Scale-Resolving Simulations (SRS)"},"content":{"rendered":"
<\/div>

In our recent posts on turbulence, we have approached the topic from an industrial perspective where most turbulence modelling is still conducted using RANS models (typically two-equation models which make use of the Boussinessq hypothesis, i.e. the turbulent viscosity). \u00a0While RANS models are indeed the workhorse for industrial CFD simulations, in an increasing number of cases companies and researchers are recognising the need to resolve a greater spectrum of turbulence (e.g. large-scale separation, strongly swirling flows, acoustics, etc.).\u00a0 In these cases, the information provided by RANS or unsteady RANS (URANS)\u00a0is limited and we need to consider scale resolving simulation (SRS) techniques.<\/p>\n

\"Q-criterion_RANS-SRS\"<\/a>
Difference in isosurface of Q-criterion between RANS (left) and SRS (right) of crossflow over a cylinder<\/figcaption><\/figure>\n

While rarely used a decade ago outside of academic research, SRS is now increasingly being embraced for many industrial CFD applications across Australia and New Zealand.\u00a0 This trend is driven by a combination of continued hardware improvements, improved parallel scalability of modern ANSYS CFD solvers and a number of new turbulence modelling strategies which have made SRS more practical to implement on complex industrial geometries. \u00a0The main drawback of SRS has been that it requires a suitably fine meshing resolution as well as suitably small timestep in order to obtain accurate results (within a practical timeframe). In the past, these two factors had combined to push SRS out of reach for most industrial CFD users (beyond the lucky few who had access to supercomputing hardware!).<\/p>\n

\"mesh-SRS\"<\/a>
Mesh resolution for a scale-resolved simulation (SRS) of flow over a car mirror<\/figcaption><\/figure>\n

When faced with the necessary compromises, industrial CFD users had typically chosen a URANS approach which provides a nonphysical single-mode transient behaviour that is dominated by the turbulent or RANS length scale. In contrast, a SRS strategy aims to resolve all (or a portion) of the turbulent spectrum and considers all turbulent length scales to the grid or dissipative limit. \u00a0In ANSYS CFD, the most detailed approach that is available is Large-Eddy Simulation (LES)<\/span>. \u00a0A LES approach essentially resolves all large turbulent structures (which have the bulk influence on the flow-field) up to the available grid limit and dissipates the energy from Kolgomorov scales via the introduction of a sub-grid scale model. \u00a0The use of LES in a practical timeframe is not feasible for wall-bounded flows at high Reynolds numbers, due to the requirement for excessive grid spacing in the boundary layer when using LES. \u00a0For this reason, LES has previously been regarded as most applicable to free shear flows.\u00a0 However, recent development of a Wall-Modelled LES (WMLES) approach has meant that less stringent meshing (and, consequently, temporal resolution) requirements can be used in the boundary layer which means that WMLES is applicable to a wider range of industrial problems.<\/p>\n

\"les-backward-facing-step\"<\/a>
Large eddy simulation (LES) of the flow over a backward-facing step to capture separated profile<\/figcaption><\/figure>\n

To rectify this issue relating to suitable grid resolution in the boundary layer, ANSYS CFD has also developed a number of hybrid SRS models which limit the resolution of turbulent structures to all regions which fall outside the boundary layer. This strategy is optimal for wall-bounded applications, since RANS models are utilized locally to resolve the boundary layer whilst the turbulent scales are properly resolved in the unsteady regions.<\/p>\n

 <\/p>\n

The choice of most suitable SRS modelling strategy will depend upon your application.<\/p>\n