FE Iterative Solver Reference
For the FE Magnetic Vector Potential solver, the Finite Element Solid Energy solver, and the Harmonic Balance Fe Magnetic Vector Potential solver Simcenter STAR-CCM+ offers an iterative solution approach based on the HYPRE library. The iterative solution approach is suitable for large model simulations, for which a direct solution can result in high memory and time requirements.
HYPRE provides some of the most commonly-used Krylov-based iterative methods to be used in conjunction with its scalable preconditioners. These include methods for nonsymmetric systems such as GMRES and methods for symmetric matrices such as Conjugate Gradient. For more information, see HYPRE.
Iterative Solver Properties
- Iterative
- Allows you to define the solver parameters for iterative solution by setting the following properties:
- Convergence
Tolerance—specifies the factor by which
the residuals of the linear system must be reduced before the solver
stops iterating. For example, with the default convergence tolerance
of 1.0E-6, the solver keeps iterating until
the residuals of the linear system drop by six orders of magnitude
(provided that the specified number of
Iterations is not reached before the
convergence tolerance is met).
For linear materials, you are advised to set the convergence tolerance to a low value. The default values are appropriate to most cases. For nonlinear materials, set the convergence tolerance to a higher value.
- Iterations—specifies the number of iterations for the iterative solver. The solver stops when it reaches the specified number of iterations, regardless of the specified convergence tolerance.
When the solver does not reach convergence within the specified number of iterations, Simcenter STAR-CCM+ prints a warning message to the output window. To visualize more information, set Verbosity to High. If the convergence rate is less than 1, you can increase the number of iterations. If the convergence rate is greater than 1, you are recommended to check the simulation setup and the mesh quality.
- Verbosity—sets the level of detail that is included in the output log. When set to High, Simcenter STAR-CCM+ prints a four-column table to the output window, showing the iteration number (column 1), the absolute residual norm (column 2), the convergence rate (column 3), and the residual norm relative to the initial value (column 4). Convergence is met when the convergence rate is less than 1.
- Convergence
Tolerance—specifies the factor by which
the residuals of the linear system must be reduced before the solver
stops iterating. For example, with the default convergence tolerance
of 1.0E-6, the solver keeps iterating until
the residuals of the linear system drop by six orders of magnitude
(provided that the specified number of
Iterations is not reached before the
convergence tolerance is met).
Flexible GMRES Solver Properties
Available only for the Harmonic Balance Fe Magnetic Vector Potential solver.
- Krylov Space Dimension
-
Sets the maximum size of orthogonal Krylov space vectors. Larger spaces enhance convergence and robustness, but increase the memory requirements necessary for the simulation to converge. The default value is 20.
Conjugate Gradient Solver Properties
Currently, this solver is available only for the finite element Magnetic Vector Potential solver, the Excitation Coil Lumped Parameter solver, and the Harmonic Balance Fe Magnetic Vector Potential solver (used as a preconditioner).
The Conjugate Gradient solver is a child node of the Iterative solver.
- Recompute Residual
- Specifies whether the solver recomputes the
residual from
. Set to one of the following values:
- 0: default value. The solver does not recompute the residual.
- n: where n is a positive number. The solver recomputes the residuals every n iterations.
- -1: the solver recomputes the residual after iterations, where is the number of iterations that you specify using the Iterations solver property.
Schwarz Solver Properties
- Direct—solves the lowest-order degrees of
freedom using the direct solver. See Sparse Direct Solver Controls.
This option is recommended for simulations with a coarse mesh. In general, the direct solver approach can require significant memory and time resources. The advantage of this approach is the direct solver factorizes the lowest-order degrees of freedom system once and then reuses the result in the following conjugate gradient iterations.
- Iterative—solves for the lowest-order degrees
of freedom using the iterative solver.
This option is recommended for simulations that have a large number of lowest-order degrees of freedom.
The Schwarz solver is a child node of the Iterative solver.