Science Application: Simulations of Turbulent Flows with Strong Shocks and Density Variations

Principal Investigator:
Sanjiva Lele, Stanford University

Participating Institutions and Co-Investigators:
Stanford University: Sanjiva Lele, Parviz Moin
Lawrence Livermore National Laboratory: Andrew Cook, William Cabot, Bjorn Sjögreen
NASA Ames Research Center: H. C. Yee
University of California, Los Angeles: Xiaolin Zhong
 

Funding Partners:
U.S. Department of Energy - Office of Science, Advanced Scientific Computing Research, and the National Nuclear Security Agency.

Duration:
Five years (started in Summer 2006)

Objectives:
In the present SciDAC Science Application, we propose to develop a new capability based on high-order high-resolution schemes to simulate shock-turbulence interactions and multi-material mixing in planar and spherical geometrical configurations and study the turbulent RT and RM mixing. These problems are of relevance to ICF capsule implosion and supernovae explosions. The project also involves the development of novel sub-grid models for shock-turbulence interaction and multi-material mixing and its validation using the DNS data. The computational code development involves a sequence of benchmark problems to test the ability of the proposed computational algorithms and their implemented performance. Ensuring good scaling performance on massively-parallel computer architectures and developing systematic methods for quantifying uncertainty in nonlinear simulations of turbulent flows are other important parts of the proposed program.

We have identified specific algorithms which require careful and pragmatic assessment in selecting the specific algorithms whose synthesis will form the basis of the production code for the science application proposed. The first year of the SciDAC work will concentrate on algorithm assessments and refinements, and specific cross-comparisons to select the most promising algorithms. From year 2 onwards the combined team expertise will focus on the proposed science application using the most promising algorithms, subgrid modeling approaches and software components. While this science application will focus on the development of these capabilities in the context of turbulent RT and RM mixing, the algorithms, models and software developed under this program  will have a much wider applicability to diverse applications of interest to DoE and other government agencies and industry. It should also be possible adopt the new capabilities in other existing codes.

Science Application Partnership: Efficient Low Dissipative High Order Multiblock/Embedded Grid Solver for Shock/Turbulence Interactions

Principal Investigator:
H. C. Yee, NASA Ames Research Center

Participating Institutions and Co-Investigators:
NASA Ames Research Center: H. C. Yee
Lawrence Livermore National Laboratory: Bjorn Sjögreen
University of California, Los Angeles: Xiaolin Zhong

Funding Partners:
U.S. Department of Energy - Office of Science and Advanced Scientific Computing Research

Duration:
Three years (started in Fall 2007)

Objectives:
The present SciDAC Science Application grant is supported by a companion Science Application Partnership (SAP).

The goal of this SAP is to provide novel numerical methods and related tools to accurately simulate complex turbulent flows with strong shocks for complex geometries in an efficient, stable, accurate and reliable manner. These tools include improved algorithm development, efficient complex geometry handling, stable and accurate numerical treatment of physical and grid interface boundaries, local grid refinement, better uncertainty quantification of computed solutions, and an accurate 3-D Navier-Stokes/MHD (magnetohydrodynamics) solver that is applicable to a wide spectrum of flow types.

The proposed tools consist of an efficient low dissipative high order finite difference method with wavelet based flow sensors for the proper control of numerical dissipation for shock/turbulence/combustion interactions. Multiblock and embedded grids capabilities with chemical reaction terms, stable treatment of stiffness due to a wide range of temporal and spatial scales, hyperviscosity filter for material interface and for shock/turbulence interactions, and a new level set shock and interface front tracking method will be added to the single block code. These methods and the resulting solver promise to give a substantial gain in computational efficiency and reliability compared to present methods, in addition to being the kernel of the new simulation tools which are the aim and goal of the proposed subject science. With the considered multi-physics flows, it is anticipated that the multiblock framework capability will allow an optimum synthesis of these new algorithms in such a way that the most appropriate spatial and/or temporal discretizations can be tailored for each particular region of the flow. Effective use of software to realize the potential of the emerging petascale computers will be sought. The software will be documented, distributed, and maintained during the life span of this project.