Computational Fluid Dynamics (A1)
Project PI: Graham Carey
PROJECT DESCRIPTION
Computational fluid dynamics (CFD) involves the computer simulation of fluid flow and transport processes in nature as well as in engineering analysis and design. It is now a recognized discipline and the applications are multidisciplinary in nature ranging from weather prediction to flow and heat transfer in heat exchangers and manufacturing processes. A finite element simulator for solution of the Navier Stokes equations and coupled transport processes is being developed under a NASA Earth and Space Sciences Grand Challenge project. The research applications include manufacturing processes such as those involved in semiconductor crystal growth and silicon oxidation where nonlinear coupling and free surface effects are important. n the proposed research, solution of these flow problems using finite element methods and domain decomposition will be explored. This will involve migration of Intel high end products to CFD applications. Next, parallel algorithm activities will be expanded to include tightly coupled clusters of Intel systems and carry out related performance studies on larger scale applications and high resolution simulations. The results from the flow applications studies and parallel algorithm work will be described in reports and technical papers.
TECHNICAL CHALLENGES
One aspect of the research plan is to demonstrate this migration for the high end Intel products. This issue will be addressed by employing both loosely coupled and tightly coupled Intel systems with concurrency of simulations (parallelism) across both physics and flow or transport subdomain partitions. The research will investigate and demonstrate the performance of both forms of configurations as part of the research effort. The flow applications will include simulation of surface tension driven flow and chemically reacting flows. Another important issue proposed is to address visualization of flow and transport results. This is a key difficulty especially for 3D flows and high resolution unstructured grids. Time dependent flow applications can easily generate large data sets and visualization can become the main bottleneck to an effective simulation. Study of both hierarchic and parallel visualization strategies will be employed for addressing this issue.
IMPACT
The proposed research will have a high impact because the CFD Lab is working on advanced applications in the target CFD area. Furthermore, our work will provide a testbed that will complement related activities using Intel clusters at the national laboratories (such as the BEOWULF and LOKI projects). Finally, the work will help migrate high performance Intel products into the graduate research environment.
EQUIPMENT
Cluster configurations (loose and tight clusters) are proposed with supporting networking hardware for fast communications between computers. The equipment will be installed in the CFD Laboratory. Year 1 - 16 Uniprocessors (no monitors); 4 Dual Processors with high graphics; 1 Dual Processor; 6 Monitors; Year 2 - 2 Dual Processors with high graphics; 3 Dual Processors; 5 monitors; Year 3 - 3 Uniprocessors; 1 Dual Processor; 3 Monitors
RESOURCES
The CFD Lab will provide laboratory space and research faculty, post-doctoral fellow and graduate student support for the application studies.
BENEFITS TO INTEL
This project will provide exposure of high-end Intel products for CFD. Our graduate students go on to work in major corporations (e.g. aerospace, the automobile industry, etc.) as well as small software development and consulting companies. The research will provide technical and performance data of interest to Intel.