ITR/AP: Multiscale Models for Microstructure Simulation and Process Design
 |
 |
 |
 |
 |
This project brings together a team comprising engineers, materials scientists, computer scientists and mathematicians to develop advanced techniques for simulating the evolution and response of materials microstructures. Target applications include dendritic solidificationin the cryopreservation of biological materials, shock-driven phase transitions in shape memory alloys and coupled continuum-atomistic models. From the perspective of computational science and information technology, these diverse applications pose a number of common challenges. They depend on coupled, heterogeneous physical models that involve multiple length and time scales. This implies a level of computational complexity that requires massively parallel and adaptive solution methods, novel coupling strategies and novel numerical methods. Several applications involve complex moving interfaces that might exhibit changes in topological connectivity.
An interdisciplinary team in the Center for Process Simulation & Design is meeting these challenges through a program of information technology research. Domain-specific abstraction frameworks for parallel applications development and run-time load-balancing strategies for heterogeneous parallel applications have been successfully deployed. We are developing a new class of spacetime discontinuous Galerkin finite element methods that offer O(N) complexity and a rich structure for parallel implementation. However, the realization of these benefits requires new technologies for mesh generation and visualization. New techniques for generating causal spacetime grids that adapt simultaneously in space and time and new visualization methods that extract data from unstructured spacetime grids meet these requirements. Our visualization technology exploits the programmable capabilities of today’s heavily parallel and deeply pipelined graphics hardware to produce per-pixel-accurate renderings of the solution data. In combination, these technologies enable significant advances in computational materials research.
