A Geometry-Based Parallel, Adaptive Finite Element Analysis Framework

72 A Geometry-Based Parallel, Adaptive Finite Element Analysis Framework--Simmetrix, Inc., 1223 Peoples Avenue, , Troy, NY 12180-3511; (518) 276-8045 Mr. Mark W. Bealll, Principal Investigator Mr. Mark W. Bealll, Business Official DOE Grant No. DE-FG02-97ER82476 Amount: $72,769 Two technologies are needed for the reliable simulation of physically and geometrically complex engineering and scientific problems. Parallel computing is needed to effectively deal with the large problem sizes, while adaptive methods are required to ensure the reliability of the calculation. Current efforts to join these two technologies suffer because there are disparate views of the computational processes which make their integration difficult. This project will combine a suite of geometry-based procedures with a new geometry-based, object-oriented, adaptive analysis framework to produce a product that can perform parallel adaptive finite element analysis in a real world environment. The Phase I project will develop the procedures needed for the analysis framework to work in a parallel computing environment. An investigation of the efficiency of the parallel adaptive analysis framework in a variety of parallel environments will be carried out. Emphasis will be placed on the application of the procedures on cost effective workstation networks. The Phase II project will concentrate on extending the Phase I system to address more complex non-linear problems and on improving the performance and robustness of the code to make it a viable commercial product. This will include building one or more full scale demonstration analysis products using the framework to help attract Phase III funding. Commercial Applications and Other Benefits as described by the awardee: The framework developed in this project would be used to build parallel, adaptive finite element analysis codes that run efficiently on machines ranging from single workstations, to clusters of workstations, to massively parallel supercomputers. The immediate benefit would be that these codes would be commercially available to allow users to design engineered products that are more efficient and less expensive. The object-oriented design of the framework allows new analysis capabilities to be quickly implemented and brought to market to help satisfy the ever growing need for simulation._