Using nonlocal interface problem allows for 7x speedup in large-scale simulations

Multimaterial problems exist in mission applications such as mechanics and subsurface transport. To capture effects arising from long-range forces at the microscale and mesoscale that aren’t accounted for by classical partial differential equations, the MAThematical foundations for Nonlocal Interface Problems (MATNIP) project team developed a mathematically rigorous interface theory employing nonlocal models that use integral as opposed to differential operators. By applying their theory to domain decomposition methods, the team was able to achieve speedups of up to 7x in solve in solve time in large-scale simulations. The issue of parameter selection for nonlocal models from experimental data was successfully tackled using both adjoint optimization and ML approaches.

The Sandia team collaborated with Los Alamos and Oak Ridge National Laboratories, as well as Florida State University, Sandia Alliance partner University of Texas at Austin, and Lehigh University. The project team had fifteen publications, presented talks, and a former team member, Marta D’Elia, was nominated for the SIAM Wilkinson Price in Scientific Computing and Numerical Analysis for contributions to nonlocal modeling.

Sandia researchers linked to work

• Christian Glusa
• Mamikon Gulian
• Pavel Bochev
• Mario Martinez

Sponsored by

Publications

May 9, 2023