Preconditioning Newton-Krylor Methods for Variably Saturated Flow
- LLNL
In this paper, we compare the effectiveness of three preconditioning strategies in simulations of variably saturated flow. Using Richards' equation as our model, we solve the nonlinear system using a Newton-Krylov method. Since Krylov solvers can stagnate, resulting in slow convergence, we investigate different strategies of preconditioning the Jacobian system. Our work uses a multigrid method to solve the preconditioning systems, with three different approximations to the Jacobian matrix. One approximation lags the nonlinearities, the second results from discarding selected off-diagonal contributions, and the third matrix considered is the full Jacobian. Results indicate that although the Jacobian is more accurate, its usage as a preconditioning matrix should be limited, as it requires much more storage than the simpler approximations. Also, simply lagging the nonlinearities gives a preconditioning matrix that is almost as effective as the full Jacobian but much easier to compute.
- Research Organization:
- Lawrence Livermore National Lab., CA (US)
- Sponsoring Organization:
- USDOE Office of Defense Programs (DP) (US)
- DOE Contract Number:
- W-7405-ENG-48
- OSTI ID:
- 791027
- Report Number(s):
- UCRL-JC-137011
- Country of Publication:
- United States
- Language:
- English
Similar Records
A Parallel, Fully Coupled, Fully Implicit Solution to Reactive Transport in Porous Media Using the Preconditioned Jacobian-Free Newton-Krylov Method
Nonlinearly preconditioned Krylov subspace methods for discrete Newton algorithms