Advanced Algebraic Multigrid Applicationfor the Acceleration of Groundwater Simulations

The computationally most intensive process during a groundwater simulation is the solution of linear systems of equations. This can take more than 80 percent of the overall runtime. Hence, an efficient linear solver can save an enormous amount of runtime. However, it may strongly depend on the overall simulation environment which linear solver is the most efficient one for specific linear systems. In time-dependent processes, for instance, the type of the most suitable solver may even change over time. We present an advanced application of algebraic multigrid (AMG) which automatically adapts and optimizes its solving strategy, yielding a robust and very efficient process. This is achieved through two steering mechanisms. The first one controls the AMG method: it attempts to save computing time by reusing setup data from previous runs. The second one controls the switching between AMG and an alternative solver which is complementary to AMG. In groundwater simulation, a reasonable choice for this alternative solver is ILU-CG. The resulting solution process behaves like a “one-for-all” solver, at least from the user’s point of view.