Shuangzhang Tu

Integrated high performance computational tools for simulations of transport and diffusion of contaminants in urban areas

Rapid analysis of transport and diffusion of chemical and biological aerosols and contaminants in an urban environment is a critical part of any homeland security response team. High performance computing (HPC) is a valuable technique for such analysis. The time constraint needed to create fully developed complex 3D city terrain models to support such dispersion simulations requires a task of converting agency data to the format necessary on the simulation platform. Numerous data sets have been employed in the development of complex 3D city models. Such data include the use of multi-layer building morphology data, the use of geographic information system (GIS) based shapefiles and digital elevation models (DEM), and the use of remote sensing data such as Light Detection and Ranging (LIDAR). The constructed geometry models are used to generate large-scale computational domains on a platform that supports our HPC tools. These tools include fully automated unstructured mesh generation, parallel and scalable flow solvers based on stabilized finite element formulations and a remote client-server environment for large-scale flow visualization. The stabilized finite element formulations, which are based on the SUPG and PSPG techniques, are parallelized and vectorized on the Cray X1. The 3D validation problem involves transient simulation of flow past a building with a source point releasing traces. A 3D application problem is presented to demonstrate the capability of the integrated HPC tools.

Simulation of Hydrodynamic Cavitating Flows Using Stabilized Finite Element Method

This paper reports our progress on the simulation of cavitating ∞ows using the stabilized flnite element method. The solver is modifled from our incompressible free-surface ∞ow solver where the volume-of-∞uid (VOF) concept is applied. The liquid-vapor phase interface is treated as an internal interface captured by an interface function. The vapor bubble boundary can be located according to the value of the volume fraction of vapor. Source terms based on the bubble dynamics are added to the incompressible Navier-Stokes equations to model the physical phase change process (vaporization and condensation) due to cavitation. The discretization is based on the flnite element method stabilized using the Streamline-Upwinded/Petrov Galerkin (SUPG) method. The matrix-free GMRES solver together with block diagonal preconditioning are used to solve the large sparse linear system resulting from the flnite element discretization. The Message Passing Interface (MPI) functions are called to parallelize the code for large-scale applications on distributed memory computers. Preliminary results are presented to demonstrate the capability of our solver in simulating cavitating ∞ows.

Simulation of Contaminant Dispersion on the Cray X1: Verification and Implementation

Stabilized finite element formulation developed for simulation of dispersion of contaminants is implemented on the Cray X1. The stabilization is based on the SUPG and PSPG techniques. The governing equations are the incompressible Navier-Stokes equations coupled with the heat and mass transfer equations. The Boussinesq approximation in the momentum equation accounts for the density change due to thermal expansion. Fully implicit nonlinear systems of equations are solved iteratively using the matrix-free GMRES technique. The stabilized finite element formulation is parallelized and vectorized on the Cray X1. The three-dimensional validation problem involves transient simulation of flow past a building with source point releasing trances. Two-dimensional problems are simulated to compare the numerical results with analytical solutions.

Numerical Simulation of a Spinning Projectile Using Parallel and Vectorized Unstructured Flow Solver

The finite volume method (FVM) is the most widely used numerical method by computational fluid dynamics (CFD) researchers to solve the compressible Navier-Stokes equations. A successful FVM solver should be accurate, efficient and robust. High-order spatial discretization must be used for accuracy. Implicit time integration is usually adopted to obtain better efficiency, especially for high Reynolds number flows. For large-scale applications, the solver should be parallelized and even vectorized to be able to run on parallel and vector computer platforms.

High performance computing of compressible flows

This paper reports the performance of our parallel implicit finite volume solver for compressible flows. The Jacobian-free Generalized Minimal Residual method (GMRES) is used to solve the linear system resulting from the discretization. Furthermore, the matrix-free Lower-Upper Symmetric Gauss Seidel (LU-SGS) method is employed as a preconditioning technique to the GMRES solver. A new slope limiting procedure is designed to suppress the unphysical overshoots and undershoots of the numerical solution while not hampering the convergence of the steady-state simulation. The solver is also parallelized using mesh partitioning and message passing interface (MPI) functions. The Cray XI is used to measure the performance of the flow solver. A few 2D and 3D numerical examples are presented to demonstrate the performance of the present solver