Innovative Computational Strategies for Large Scale and Coupled Problems

The scope of this research covers multiscale and strongly-coupled multiphysics problems. The three major activities are:

  • Hierarchical adaptive modeling strategies for simulating response phenomena occurring at disparate spatial and time scales, using reasonable computer resources. The strategies use multiple mathematical models in different regions of the structure to take advantage of efficiencies gained by matching the model to the expected response in each region. Adaptivity in the strategy minimizes reliance on a priori assumptions about the response. An object-oriented interactive environment is being developed that not only allows the use of different numerical algorithms in different parts of the structure, but also allows the algorithms to be changed dynamically during computation.
  • Strategies for the effective use of diverse high performance computing platforms. The platforms considered include high performance computers such as clusters of processors, large-scale parallel computers, terascale and grid computing platforms. General guidelines are produced for designing future large-scale simulation programs that will run efficiently on distributed heterogeneous computing platforms or on hardware tailored for specific applications.
  • Effective hybrid strategies, including numerical/analytical and numerical/neuro-computing methods.