Title: The Simulation and Modeling of Turbulent Flows

Speaker: Krishnan Mahesh, University of Minnesota

Date/Time: Thursday, June 1, 2006, 9:00 – 10:00 am

Location: Building 980, Room 95 (Sandia NM)

Brief Abstract: Direct numerical (DNS) and large-eddy simulation (LES) have traditionally been restricted to fairly simple geometries. The first part of the talk will discuss our work on developing the capability to reliably perform such simulations in engineering configurations. Special attention has been paid to ensuring robustness and accuracy at high Reynolds numbers without numerical dissipation. The talk will briefly describe our incompressible algorithm (Mahesh et al. 2004, J. Comput. Phys., vol. 197) and present some representative solutions. We have recently extended this capability to compressible flows (Hou & Mahesh 2005, J. Comput. Phys., vol. 205) and will discuss the basic ideas.

We have recently developed an analytical theory (Park & Mahesh 2006, Submitted to J. Comput. Phys.) that predicts the effect of numerical error in turbulence simulations. The theory accounts independently for the effects of truncation error, aliasing error, and errors in the `subgrid' turbulence models. The theory is used to model the overall error in a second-order, energy-conserving, computational code. Good agreement with actual simulation results are obtained. A broader implication of the theory is its further development as a tool for uncertainty analysis of numerical solutions.

CSRI POC: S. Scott Collis, (505) 284-1123