Structural Engineering, Mechanics and Materials Seminar: Optimization Under Uncertainty for Design, Materials, and Large-Scale Computing

Seminar | November 4 | 12-1 p.m. | 502 Davis Hall

 James R. Stewart, Computational Sciences & Math Sandia National Laboratories

 Civil and Environmental Engineering (CEE)

This presentation highlights the many ways that optimization is used to support multiscale, multiphysics modeling and simulation. Examples include design and topology optimization, as well as PDE-constrained optimization and Bayesian inference for estimating material properties or input model parameters. In the context of additive manufacturing, the materials themselves become part of the design space, with control of the microstructure being extremely important yet highly uncertain. We show how anisotropy and variability in the microstructure affect design outcomes and performance. Additionally, one can introduce uncertainties into the design formulation through the introduction of risk measures. We then shift to very large-scale calibration of model parameters through a combination of PDE-constrained optimization and Bayesian inference. Results and work-in-progress are discussed for modeling Greenland and Antarctica ice sheets. The topics covered in this presentation provide a sampling of Sandia’s computational science capabilities, which are pushing the envelope for both design and large-scale, data-driven predictive simulation.

 Faculty, Students - Graduate, Students - Undergraduate

 Faculty, Students - Graduate, Students - Undergraduate

 sandwiches

 shaofan@berkeley.edu, 510-6425362

 Flyer

 FLyer