4.5.3.1.4 Controlling the Numerical Solver:

PROMIS-NT allows to choose between two different numerical equation solver algorithms. One of them is a direct Gauß solver. The other one, an iterative solver, uses the bi-conjugate gradients stabilized algorithm (BICG-Stab) [63]. The PromisNTSetupModel parameter directSolver allows to choose between them.

The iterative solver needs an additional convergence criterion. This is controlled by the Parameter linearAccuracy. Permutation of the system matrix may lead to better convergence of the iterative solver because of an improved matrix structure. This option is handled with the Parameter sortNodes. The preconditioning matrices are calculated by an incomplete LU factorization of the equation matrix. The maximum fill-in for this matrices compared to the equation matrix is determined by maxFillin.

The user may choose between various damping schemes described in very detail in [15] by utilizing the Parameters dampScheme, delta, minDamp, and maxDamp.

The Parameters printLinear, iterationInfo, and updateInfo are responsible to control the ``verbosity'' of the transient equation solving and may be used to obtain additional information in case the user defined diffusion models cause convergence problems. All PromisNTSetupModel parameters controlling the numerical equation solving are summarized in Table 4.8.

Table 4.8: PROMIS-NT numerical equation solver setup

Name	Type	Description
directSolver	MdlBool	Use direct solver rather than iterative one. Default: false
linearAccuracy	double	The relative accuracy of the linear solver. Default: 1e-5
sortNodes	MdlBool	Sort matrix before solving. This can improve the convergence even for the iterative solver. Default: true
maxFillin	int	The maximum fill-in for the preconditioner. Default: 50
finalNorm	double	At this residual norm the iteration stops. Default: 1e-2
maxSolveTime	double	The maximum CPU time in seconds for one solving run of the linear solver. Default: 20.0
iterationLimit	int	Final number of iterations. Default: 30
dampScheme	MdlString	The damping scheme. One of "none", "test", "full", "bank", "curvature", "minimos", "local", "global", "potential". Default: "bank"
delta	double	The damp parameter for the damping scheme. Default: 0.1
minDamp	double	The minimum damp factor for the nonlinear iterations. Default: 0.008
maxDamp	double	The maximum damp factor for the nonlinear iterations. Default: 0.99
printLinear	MdlBool	Print the whole linear system to the file mat.dat if an error occurs in the solver. Default: false
iterationInfo	MdlBool	Print information about the linear solver progress and various quantity norms at each iteration. Default: true
updateInfo	MdlBool	Print information about the quantity updates. Default: false