November 13, 2018

 

computer model image
Bayesian calibration and validation of atomistic (left) and coarse-grained (right) models will play a key role in developing AEOLUS methods to learn from data and optimally design experimental campaigns. Credit: J.T. Oden, K. Farrell, D. Faghihi.

A joint university—U.S. Department of Energy (DOE) laboratories team of researchers led by The University of Texas at Austin Institute for Computational Engineering and Sciences (ICES) Professors Omar Ghattas and Karen Willcox has been awarded a four-year, $10 million grant by the DOE's Advanced Scientific Computing Research program to create the "AEOLUS" center for applied mathematics research in experimental design, optimal control, and learning, with application to advanced manufacturing and materials.

ICES professors George Biros, Robert Moser and J. Tinsley Oden, also an ASE/EM professor, are co-principal investigators on the AEOLUS center. Other institutions involved include Brookhaven National Lab, MIT, Oak Ridge National Lab, and Texas A&M University.

The AEOLUS center aims to create new unified mathematical, computational, and statistical approaches that exploit the interplay between large-scale simulations and experiments. AEOLUS will target (1) learning predictive models from complex data via Bayesian inference and optimization, and (2) optimizing experiments, processes, and designs using the resulting uncertain models.

"To fully realize the power of scientific simulation as a basis for scientific discovery, technological innovation, and rational decision-making, it is now imperative to move beyond simulation to tackle the outer loop of optimization for learning from data, experimental design, and control with complex uncertain models," said Omar Ghattas, a professor in the Departments of Geological Sciences and Mechanical Engineering and AEOLUS co-director.

"Given data, how do we infer complex models with associated uncertainty? How do we design new experiments to reduce the uncertainties in these models? And finally how do we optimally design and control systems governed by these models? Carrying out these tasks for the large, complex models that characterize leading edge DOE applications has been intractable historically."

"Our target area of advanced materials and manufacturing is critically important to the Department of Energy's mission to ensure the nation's security and prosperity, and is a rich source of challenging problems in inference, experimental design, and optimal control," said ICES director Karen Willcox, a professor in the Department of Aerospace Engineering and Engineering Mechanics, and AEOLUS co-director.

"Our testbed problems in additive manufacturing and materials self-assembly require multifaceted and integrated advances in applied mathematics. We are looking forward to a strong partnership with Brookhaven and Oak Ridge National Laboratories to tackle these challenging problems."