Joshua White | 18-ERD-027
The performance and safety of subsurface engineering projects depends on coupled fluid flow and geomechanical processes, and coupled multiphysics simulations are frequently used to design and manage these projects. Within these codes, the most expensive and least scalable component is almost always the linear solver used to address the implicit timestep advance. This bottleneck frequently limits the scale of computations that can be tackled.
In this work, we developed algebraic multigrid solvers that can tackle a wide variety of coupled physics in an efficient and scalable manner. The solvers have been implemented in an open-source solver package developed at Lawrence Livermore National Laboratory, known as HYPRE, and they are now available to the research and practice community. We also demonstrated the application of these methods by linking HYPRE to GEOSX, an open-source reservoir simulator developed at the Laboratory. We demonstrated scalable simulations of coupled fluid flow, transport, and geomechanics on problems orders-of-magnitude larger than currently possible.
This work has enabled significantly improved flow and geomechanics simulation capabilities for many applications of strategic interest to the Department of Energy (DOE), including geothermal energy systems, carbon storage projects, oil and gas reservoirs, and nuclear waste repositories. In addition, the solver methods enhance Livermore's core competency in high-performance computing, simulation, and data science, and provides a flexible capability that is useful for other multiphysics applications that address the Laboratory's national security and energy security missions.
Publications, Presentations, and Patents
Bui, Q. M., et al. 2020. "A Scalable Multigrid Reduction Framework for Multiphase Poromechanics of Heterogeneous Media." SIAM Journal on Scientific Computing 42, 2: B379-B396. LLNL-JRNL-772001
White, J. A., et al. 2019 "A two-stage preconditioner for multiphase poromechanics in reservoir simulation." Computer Methods in Applied Mechanics and Engineering 357: 112575. LLNL-JRNL-763612