Segregated

This feature can be used as an attribute for the Time and Stationary features. The approach taken is nonlinear Uzawa iterations in which user-defined groups of variables are solved for separately (a segregated step) while other variables are held fixed. The segregated steps for the segregated solver is handled by subattributes of the sort of SegregatedStep, LumpedStep, and LowerLimit.

The Segregated attribute supports the following properties;

Table 6-57: Valid Segregated Property/value Pairs
Property	Value	Default	Description
maxsegiter	positive integer	100, 25 (Time)	Maximum number of segregated iterations.
ntolfact	positive scalar	1	Tolerance factor.
plot	on \| off	off	Plot while solving.
plotgroup	String	default	Plot group to use for plot while solving.
probesel	all \| none \| manual	all	The probes to compute.
probes	array of strings		Probes to use when probesel=manual.
ratelimit	positive scalar	1	Limit on nonlinear convergence rate.
segaaccdim	positive integer	10	Dimension of Anderson iteration space when segstabacc = segaacc.
segiter	positive integer	1	Fixed number of segregated iterations.
segreserrfact	positive scalar	100	Residual factor for segtermonres=auto.
segstabacc	none \| segclfcmp \| segaacc	none	Stabilization and acceleration: None, pseudo time stepping (for stationary solvers), or Anderson acceleration.
segterm	iter \| tol \| itertol	tol	Segregated solver termination technique.
segtermonres	off \| on \| auto	auto	Termination criterion: Solution, residual, solution or residual.
subcfltol	positive scalar	0.1	Target error estimate for pseudo time stepping.
subinitcfl	positive scalar	5.0	Initial CFL number for pseudo time stepping.
subkdpid	positive scalar	0.05	PID regulator-Derivative for pseudo time stepping.
subkipid	positive scalar	0.05	PID regulator-Integrative for pseudo time stepping.
subkppid	positive scalar	0.65	PID regulator-Proportional for pseudo time stepping.
useratelimit	on \| off	off, on (time)	Use limit on nonlinear convergence rate.

Termination of the segregated solver is controlled by the property segterm. The default setting is tol, in which case the segregated iterations are terminated when, for each group, the estimated error is below the corresponding tolerance set by the main tolerance for the parent solver multiplied with the nonlinear tolerance factor ntolfact. However, a maximum number of allowed segregated iterations is chosen through the property maxsegiter; if the maximum is reached, the iterations are terminated and an error message is displayed. Termination after a fixed number of segregated iterations is achieved by instead choosing iter. The number of segregated iterations is controlled by the property segiter. The third available option for segterm is itertol, which is a combination of the other two options; the segregated iterations are terminated when one of the two convergence criteria of tol and iter is met. The property maxsegiter is only supported when tol is used for termination. For both the settings iter and itertol, the number of iterations is controlled by the property segiter.

The nonlinear solver uses an adaptive tolerance for termination of iterative linear system solvers. This adaptive tolerance is based on the maximum of ntol and itol. During the nonlinear iterations, it can, however, be larger or smaller than this number. The segregated solver uses the same tolerance as the linear solver when constant damping is used. However, when automatically adjusted damping is used, the adaptive tolerance of the nonlinear solver is used. The parametric solver uses the same tolerance as the corresponding stationary solver.

The property segstabacc enables or disables pseudo time stepping (for stationary problems) or Anderson acceleration. When enabled the pseudo time stepping is controlled by the scalar-valued regulator parameters subcfltol, subinitcfl, subkdpid, subkipid, and subkppid. For the Anderson acceleration, the parameter segaaccdim specifies the dimension of the Anderson iteration space.

The property segtermonres controls the termination criterion for stationary problems when segterm is not iter. When segtermonres=off the estimated error is solution-based, with segtermonres=on it is based on a relative residual and for segtermonres=auto the estimated error is the minimum of the solution and residual based errors. For segtermonres=auto the property segreserrfact is a scalar factor multiplying the relative residual error.