Research Overview


High Performance Computing
Use of nonstandard hardware accelerators to increase the performance of parallel supercomputers by at least an order of magnitude.
CFD with Graphics Processors
SmithWaterman (GEMS, JCP)
Scientific Computing with Graphics Processors
Tilera Pro64 (64 core processor)
Supported by DOE/Oak Ridge (Steve Poole) 

Numerical Methods
Design of 'mimetic' numerical methods that capture the physics of partial differential equations well.
Discrete Calculus Methods
Unstructured Staggered Mesh Methods
Secondary Conservation
Fractional Step Methods
Supported by a Stanford subcontract from DOE (ASCI)


Turbulence Modeling
Development of equation systems which mimic NavierStokes equations but which are computationally tractable on a PC.
OrientedEddy Collision Model
Universal RANS/LES (k/eps, RST)
Dissipation Tensor
Decay Rate
Transition
Turbine Blades
Turbulent Potential Model
Supported by:
NSF (Fluid Dynamics)
ONR (Pat Purtell/ Ron Joslin)
AFOSR (Tom Beutner)


Superhydrophobic Drag Reduction
Analysis of the fundamental physical mechanisms underlying superhydrophobic drag reduction.
High Reynolds Number Simulations
Turbulent Simulations
Laminar Theory
Experiments (2000)
Data Archive.


Turbulence Simulation
Direct Numerical Simulation of canonical turbulent flows which use physically realistic initial conditions and which use parallel staggeredmesh numerical methods that conserve mass, momentum, kinetic energy and vorticity/circulation, and
Rotating Decay
ReturntoIsotropy
Plane Strain
Isotropic Decay
Superhydrophobic Surfaces
ShearFree Turbulent Boundary Layers
Data Archive.
Supported by NSF (Fluid Dynamics)


FreeSurface Simulation
Calculation methods for the simulation of two phase flows with very large density discontinuities. High resolution of the freesurface interface with moving and adaptive staggered mesh methods.
Methods
Droplets
Supported by NSF exploratory grant (SGER)


Interactive Computational Fluid Dynamics
RealTime Direct Numerical Solution of Turbulent fluid flow. Allows the user to push the fluid with the mouse and obtain an intuitive feel for turbulence. An application for hardware acceleration (below).
PollenSeed (3D) {just email if this need a dll}
Stir Crazy (2D)
Source Code


Local Supercomputing
Parallel Machines built and maintained by the Theoretical and Computational Fluid Dynamics Laboratory.
Orion (8 GPUs, 1924 cores, pictured)
Cyclops (608 cores)
von Karman
(500 cores, decomissioned)
Supported by:
ONR (Pat Purtel/Candice Wark)
UMass College of Engineering (Mike Malone)
Department of Energy
(Steve Poole)


pict2.JPG 
