BLACKSBURG, VA -- A team of engineers and a mathematician from Virginia Tech and the University of Florida has won a $556,511 grant from the Air Force Office of Scientific Research to work on theory and algorithms for global/local design optimization. The design of large-scale systems such as automobiles, aircraft, and ships involve multiple disciplines such as structures, fluids, propulsion, economics, and manufacture. These systems require a decomposition of the global optimal-design problem into disciplinary sub-problems that ideally can be done concurrently and independently, said Layne Watson, professor of computer science and mathematics in the College of Arts and Sciences at Virginia Tech. "Typically the subsystems are coupled and cannot be optimized independently of each other without communicating data during the design process," Watson said. "As system complexity grows, this communication becomes a bottleneck, and the time required for high-fidelity subsystem simulations also grows. This project will develop mathematically rigorous decomposition theories so that large-scale design problems can be effectively solved on massively parallel supercomputers." Another aspect of the project is the development of cheap surrogate approximations to expensive simulations. This both reduces the subsystem evaluation time and improves the concurrency of the global design process. "Innovative and radical approaches to large-scale optimal design are being explored, both theoretically and empirically," Watson said. "Computational paradigms such as global/local optimization and cellular automata (models or approaches to computation), which are unproven but hold considerable promise for implementation on massively parallel computers, will be adapted to engineering design problems. "This project is interesting because it involves both mathematical theory and engineering experiments," Watson said. "Computer-science students are excited about working on a project where the things they design, like composite aircraft wing panels, are actually built and tested in the lab. Engineering students get to work with our Beowulf supercomputer in Torgersen Hall, something they normally wouldn't have access to."