Addressing Challenges in Modeling Subsurface Phenomena Using HPC, Computational Geometry and Applied Mathematics

Saumik Dana, Baylor College of Medicine

Abstract:  The problem of coupling of flow and geomechanics in subsurface has been a subject of intensive study in the realm of energy and environmental applications such as land subsidence and carbon sequestration among others. Land subsidence is a gradual settling or sudden sinking of the Earth's surface due to removal or displacement of subsurface earth materials. Carbon sequestration is the long-term removal, capture or sequestration of carbon dioxide from the atmosphere to slow or reverse atmospheric CO2 pollution and to mitigate or reverse global warming.

All of these problems have two characteristics: one, they are of multi-physics nature combining one or more of the effects, such as elasticity, flow, transport, chemical reactions etc. and, secondly, they are large scale problems marked by highly heterogeneous material properties. One of the main concerns with modeling is the disparity in length scales across physics for the coupled flow and mechanics models. We try to resolve this disparity by introducing computational geometry and HPC into a staggered solution algorithm. The algorithm solves the coupled problem by decoupling it and then solving the subproblems sequentially until convergence is achieved. The computational geometry aspect allows us to solve the decoupled problems on different grids. We then present some improvements to the algorithm for anisotropic cases. Some of the auxiliary method development will also be presented.

Also presented is a graph theory-based model order reduction for modeling flow and transport in subsurface discrete fracture networks. This work was done at Los Alamos National Lab.

Please use this link to attend the virtual seminar: