Ph.D. Proposal

Dhruv Purushotham 

(Advisor: Prof. Oefelein)

"Characterization of Subgrid Closure Performance for Large-Eddy Simulation of 
Supercritical Turbulent Flows" 

Wednesday, March 8th 

2:00 p.m. 

Montgomery Knight Room 325 

Abstract 

High-pressure flow physics has historically been important to propulsion systems. The last few years, however, have seen the development of high-pressure ground power systems which promise to operate at higher efficiencies than their traditional counterparts. The increased interest in high-pressure/supercritical fluid physics in recent decades has led to much experimental and computational research in this area. 

Three techniques exist for the study of turbulent flows through computation. These are the Direct Numerical Simulation (DNS) technique, the Large-Eddy Simulation (LES) technique, and the Reynolds-Averaged Navier Stokes (RANS) approximation. With the increasing availability of powerful computing architectures, LES represents an attractive option for the community due to a good balance of accuracy and cost. The methodology is mature, yet relies upon closures and assumptions derived from ideal gas behaviors to operate. The framework hence naturally suffers when applied to high-pressure flows with significant real gas effects and the related nonlinearities. As a result, the research community has placed significant attention on the improvement of the LES method so it can be justifiably applied at these exotic conditions. 

The proposed work makes use of both the DNS and LES methodologies to systematically evaluate the performance of a baseline set of commonly used LES closures. The momentum closure in LES calculations is achieved using models associated with the mixed-dynamic Smagorinsky model, while energy closure is achieved using a gradient diffusion model. A suite of four simulations is proposed to be carried out, including one DNS. The three LES calculations include a wall-resolved simulation, and two additional calculations with successive levels of spatio-temporal coarsening. All simulations are proposed to occur in a canonical spatial mixing layer configuration. 
The DNS data set will help drive a physics-based evaluation of current model formulation. This data set will also provide detailed, full-field information which can be used to quantitatively assess the LES closure performance on a cell-to-cell basis, which will help delineate regions of good closure performance from those of poor performance. In this way, the proposed work may identify a causal link between supercritical flow physics and LES model performance. Outcomes of the work involve novel insight regarding the physics of supercritical fluid mixing, a rigorous assessment of model performance and a characterization of model performance as a function of grid resolution. 

Committee

•    Prof. Joseph C. Oefelein – School of Aerospace Engineering (advisor)
•    Prof. Adam M. Steinberg – School of Aerospace Engineering
•    Prof. Jerry M. Seitzman – School of Aerospace Engineering
•    Prof. Devesh Ranjan – School of Mechanical Engineering
•    Dr. Ramanan Sankaran – Oak Ridge National Laboratory