Session: 30-06 Fluid Dynamics

Paper Number: 125326

125326 - Evaluation of Subfilter Model Performance for LES of Supercritical Fluids 

The collective need for a more sustainable energy future has led to the rapid development of high-pressure ground-based power systems. These systems offer higher thermal efficiencies than their traditional low-pressure counterparts and can be built with carbon capture technologies embedded within them. Such systems provide an important alternative that complements the goal of carbon free energy production and a completely renewable energy production framework.

The development of these systems is naturally a significant focal point for the research community. The synergistic application of numerical and experimental analysis is crucial to minimizing the time-to-deployment of optimal configurations. Fortunately, powerful computing architectures are becoming increasingly accessible to engineering development teams. This permits the application of higher fidelity simulation methods than has been historically possible. In particular, the large-eddy simulation (LES) framework now offers a realistic alternative to the cheaper Reynolds-averaged Naiver-Stokes (RANS) method for design. The LES method, however, still applies an averaging operation to the governing equations, and consequently suffers from its version of the classical closure problem. Turbulence models must be developed and employed to circumvent this issue.

The last few decades have seen significant focus placed on the development of subfilter- or subgrid-scale models. These models can typically be reliably employed for LES of inert, incompressible turbulent flows. However, the extension of these models into more complex multiphysics regimes introduces uncertainties that must be quantified. Deficiencies associated with the well-known Smagorinsky closure in the high-pressure fluid regime have been studied extensively. These analyses have largely been done in an a priori sense whereby the flow field resulting from a direct numerical simulation (DNS) is filtered to obtain quantities of interest in compressible LES. The DNS calculation does not involve any turbulence models and so these data serve as a reference point from which model performance can be assessed. Model forms are then applied to the fields using reconstructed filtered quantities and conclusions regarding model performance may be drawn. However, from a fundamental viewpoint, an instantaneous realization of a field governed by the filtered system of governing equations must be different from the corresponding realization for the exact system. Hence, a priori analysis, while important, can only provide a partial picture of overall model performance.

To provide additional insights into LES of supercritical fluids, the current research applies an a posteriori analysis to quantify the performance of a suite of models associated with the dynamic Smagorinsky closure in the high-pressure, supercritical fluid regime. Datasets obtained include three LES calculations at distinct resolutions spanning from a wall-resolved LES through to the coarse grid limit. A fourth dataset is generated using DNS, which is used as a reference point for the analysis. The data is generated from a canonical spatially-evolving mixing layer. Full 3D mean and instantaneous data is available for all four datasets. The analysis performed provides a comprehensive picture regarding model performance in this regime.

Presenting Author: Dhruv Purushotham Georgia Institute of Technology

Presenting Author Biography: Dhruv Purushotham is a PhD candidate at the School of Aerospace Engineering at the Georgia Institute of Technology. His research focuses on supercritical fluids, large-eddy simulation, direct numerical simulation, high-performance computing and turbulence modeling.

Authors:

Dhruv Purushotham Georgia Institute of Technology
Joseph Oefelein Georgia Institute of Technology