Setting the Solver Parameters

Set the solver parameters. The under-relaxation factors are important and are common to all solvers. The other settings are typically used for troubleshooting.

The default relaxation factors are chosen to give reliable convergence for common multiphase flow cases. If a multiphase study requires extra under-relaxation for convergence, you can improve performance by using the Explicit Under Relaxation factor for the additional amount. One strategy is to retain the Implicit Under-Relaxation Factor at its default value, or up to 0.8, and reduce the Explicit Under-Relaxation Factor. The product of the two factors then gives the overall relaxation required.

You are best to apply Explicit Under-Relaxation to both Velocity and Volume Fraction Solvers. Where the S-Gamma model is active, use the same relaxation scheme for the S-Gamma solver as for the Volume Fraction solver.

Steady-state flows that are specified with inlet boundary conditions benefit most from Explicit Under-Relaxation. In these cases, you can save up to 30% in the number of iterations that are required for convergence.

In addition to specifying numerical values, you can use expressions to set under-relaxation factors. This possibility allows you to adjust or ramp the under-relaxation factors over the course of the simulation by using simulation operations. It is possible to use global parameters to specify an input expression for any of the under-relaxation factors. See Global Parameters.

The steps in this procedure are intended to follow on from Step 8 in Modeling Eulerian Multiphase Flow.

To set the solver parameters:

  1. Edit the Solvers node and set the appropriate under-relaxation factors.
    Solver Node Description
    Segregated EMP Flow > Phase Coupled Velocity

    See Segregated EMP Flow Solver.

    Segregated EMP Flow > Pressure

    See Segregated EMP Flow Solver.

    Segregated Energy

    See Segregated Energy Solver for Multiphase.

    Volume Fraction

    See Volume Fraction Solver for Multiphase.

For steady state cases, the default values may be insufficient for convergence, for example due to strong force or source terms. For such cases, you can reduce the relaxation factors.

  1. For each solver in turn, starting with the velocity or volume fraction solver, reduce the relaxation factors towards 0.1 as required.

    You can use the response of the residuals as a guide to the effectiveness of each adjustment. For relaxation factors below 0.1, convergence can be slow. In these cases, it can be more effective to approach steady state using time-stepping, especially if a physical timescale can be resolved for the term that is preventing convergence or causing unsteadiness. Reducing the time-step size is the first thing that you can try when attempting to resolve difficulties in unsteady simulations.

    Alternatively, if the rate of convergence is too slow, you can improve performance by transferring some of the relaxation from the Implicit Under-Relaxation Factor (see Eqn. (920)) to the Explicit Under-Relaxation Factor. An Explicit Under-Relaxation Factor is available for the Phase Coupled Velocity, Volume Fraction, and S-Gamma solvers.

Some cases may require you to decrease the linear solver convergence tolerance. This setting is a property of the AMG Linear Solver component of each solver.

  1. Select the AMG Linear Solver component of the appropriate solver and set the Convergence Tolerance value.

    The default value is 0.1, but you can reduce this value to between 1.0E-3 and 1.0E-4 in most cases. You can investigate the justification for, and the effect of, this adjustment by changing the Verbosity setting from None to Low and then examining the diagnostics output from each solver over a single iteration.

Return to Modeling Eulerian Multiphase Flow and continue with Step 9.