18.3.4. Convergence Control Parameters

Relaxation Parameters

model coefficient relaxation

Type

Real

Default Value

1.0

Description

Sometimes poor convergence is observed with the turbulence model, characterized by bouncy and residual graphs. Changing this under-relaxation parameter can help, but do not use values lower than 0.1.

outer loop relaxations default

Type

Real

Default Value

0.75

Description

Parameter outer loop relaxations default, which defaults to 0.75, is designed to provide a default value for multiple relaxation parameters that operate at the outer loop level. Currently, parameter outer loop relaxations default provides a default value for only one relaxation parameter, relax mass, and only for steady-state cases.

Bouncy convergence, even with a small timestep, can sometimes be reduced by lowering this value to 0.5 or less. A value below 0.25 is not recommended because that would drastically slow down convergence.

relax mass

Type

Real

Default Value

0.75

Description

Changing the value of this parameter can aid convergence for problems that show residual oscillations in separation and reattachment regions. Because it relaxes the mass flow calculation procedure, it should not be used for transient problems (or should be set to 1).

For transient cases, parameter relax mass defaults to 1.

For steady-state cases, parameter relax mass defaults to the value of parameter outer loop relaxations default.

wallscale relaxation factor

Type

Real

Default Value

0.75

Description

This is a relaxation factor for the wallscale equation, which is solved for all and SST turbulence models.

The wall scale equation is a pure diffusion equation and is therefore very sensitive to non-orthogonal meshes (such as a mesh that is highly skewed at a corner). If convergence problems for a wallscale equation are observed, try reducing the value of the wallscale relaxation factor.

Convergence and Runtime Control

allow cbck for initcon

Type

Logical

Default Value

f

Description

If set to 't', expressions involving callbacks (such as integrated values) can be used to set initial conditions. By default it is not allowed because allowing callbacks introduces the risk of hanging in parallel.

auto turnoff solve eq flag

Type

Logical

Default Value

f

Description

This flag determines whether to stop solving individual equations when residuals have fallen below their convergence criteria. With this flag set to true then, for example, the solution for each of the U, V, W, and P-Mass equations is stopped when the corresponding equation residual falls below its target by more than 10%. The run is completed when convergence has been achieved and when the Additional Variable equations have been solved satisfactorily. Note that turbulence equations are ignored by the solver when determining the level of convergence, and so are not affected by this flag.

This is only functional for steady-state models, and when the model contains one or more Additional Variables. It is automatically set to 't' if the Auto Timescale feature is invoked.

check isolated regions

Type

Logical

Default Value

t

Description

For serial runs, the solver checks if any fluid domain contains volumetric regions that are isolated pockets. This check cannot be performed for parallel solver runs.

ignore solve flag on restart

Type

Logical

Default Value

f

Description

The solution of the wallscale equation stops after it has converged, even if the other equations have not yet converged. If you perform a restart and reduce the convergence criterion, it will not start solving again because the flag indicating it has converged is still set. To start solving the wallscale equation again on a restart, set this parameter to 't'.

imbalance normalisation type

Type

Integer

Default Value

2

Description

Specifies how imbalances are normalized:

0 - no normalization

1 - normalize by maximum contribution in domain

2 - normalize by maximum contribution in all connected domains

require coefloop convergence

Type

Logical

Default Value

f

Description

If set to 't' then the solver will stop in a transient run if convergence is not achieved within the coefficient loop

Memory Control

topology estimate factor

Type

Real

Default Value

1.0

Description

Note:  Do not use this parameter unless informed to in a run-time error message.

Adjusts an internal factor used to help when the solver is estimating the total memory required to store the control volume equation matrix. Sometimes the internal factor is not sufficient and the solver will stop indicating that the "topology estimate factor" parameter must be increased. If this is the case, supply a number for this parameter that is larger than 1; for example, 1.2 will increase the internal memory estimate by 20%, which is often sufficient.

topology estimate factor zif

Type

Real

Default Value

1.0

Description

Note:  Do not use this parameter unless informed to in a run-time error message.

Adjusts an internal factor used to help when the solver is estimating the total memory required to store the matrix coefficients at GGI domain interfaces. Sometimes the internal factor is not sufficient and the solver will stop indicating that the "topology estimate factor" parameter must be increased. If this is the case, supply a number for this parameter that is larger than 1; for example, 1.2 will increase the internal memory estimate by 20%, which is often sufficient.

High Speed Models

max continuity loops

Type

Integer

Default Value

1

Description

Sets the maximum number of continuity loops to perform within a timestep. The continuity loop iterates on the density*velocity nonlinearity in the continuity equation. The default value of 1 is usually appropriate. For high-speed supersonic flows (Mach numbers above 2), increasing the value to 2 may help convergence.