Define Controls Using the AMG-CG Linear Solver
-
From the Linear solver options, select
AMG-CG.
The app calculates the pre-conditioner matrix using the algebraic multi-grid conjugate gradient (AMG-CG) method. The AMG-CG linear solver is recommended for most simulations.
- Enter a value for the Iteration limit, which is the maximum number of iterations for the linear solver.
- Enter a value for the Convergence check frequency, which is the interval (in linear iterations) to check linear convergence.
-
Enter a value for the Linear convergence limit, which is the
point at which the iterative solver stops.
The iterative solver stops when the relative residual norm of the system of equations and the relative correction of the solution norm fall below this convergence limit.
-
Select one of the following patterns for filtering errors from the AMG
cycle type options:
Option Description V Filters errors using a V-cycle pattern. W Filters errors using a W-cycle pattern. The cycle type pattern filters the error from the linear solver to accelerate the convergence.
-
Select one of the following techniques for removing high-frequency errors from the
Residual smoother options:
Option Description ILU Removes high-frequency errors using the incomplete lower-upper factorization smoother. SSOR Removes high-frequency errors using the symmetric successive over-relaxation smoother. - Enter a value for the Maximum size of coarsest level.
- Enter a value for the Maximum number of levels, which is the maximum number of coarse grid problems.
- Enter a value for the Number of pre-smoothing steps, which is the number of pre-relaxations for the fixed multi-grid cycles.
- Enter a value for the Number of post-smoothing steps, which is the number of post-relaxations for the fixed multi-grid cycles.