A.C. Loos
The objective of this investigation to develop a comprehensive 3-D process simulation model for complex composite and to optimize the Resin Transfer Modeling (RTM) fabrication of aircraft stiffened structures. This effort will directly impact the process cycle development in the NASA Advanced Composite Technology (ACT) program. A joint research program between NASA Langley, Virginia Polytechnic Institute and State University, The College of William and Mary, Northrop Corp. and Douglas Aircraft Company is now underway to develop a science-based understanding of the RTM process. The use of this program will result in new low-cost manufacturing methods for damage tolerant structures.
The RTM simulation code includes numerical modules for the prediction of the resin free surface movement in the porous textile preform, a module for predicting the convective and conductive heat transfer in the polymer resin during the process, a of the porous preform, and a module to predict the compaction behavior of the textile preform and tooling assembly during the processing cycle. The main section of the 3-D simulation code consists of a set of five highly-coupled non-linear finite element solvers which produce the resin pressure, resin temperature, fiber/tooling temperature, and deformation fields for the problem.
Presently the simulation code uses a series of sparse solvers provided with the Cray Y-MP at NASA Langley. These include several from the SITRSOL numerical library from Cray Research Inc. In order to reduce the memory requirements involved in the comprehensive simulation, assembly of the global stiffness matrices is accomplished using the sparse column format. Efforts are underway which will allow the RTM code to be ported to a symmetric multiprocessing architecture system. The implementation on such a system will allow for the current simulations to be run on a midsize workstation platform. Scalability will allow for the user to summit RTM simulations of varying size to the META center computers and if more memory or CPU cycles are needed then additional resources will be automatically allocated from the server machines across the network. As the problem sizes grow the use of 2, 4, or 8 way memory interleaving will also be necessary. Access to scalable system will demonstrate a more general purpose computational paradigm to a broader audience.
With scalable systems it would be possible to simulate and visualize results in "realtime" in a CAVE environment for some of the more complex 3D structures. Visualization of 3D simulation results would be developed on desktop workstations first and converted to a CAVE environment when appropriate.