For the P-1, DTRM, S2S, and the DO radiation models, there are several parameters that control the radiation calculation. You can use the default solution parameters for most problems, or you can modify these parameters to control the convergence and accuracy of the solution. Iteration parameters that are unique for a particular radiation model are specified in the Radiation Model dialog box (for example, Energy Iterations per Radiation Iteration). Spatial Discretization (Discretization in the Theory Guide) and Under-Relaxation Factors (Under-Relaxation of Variables in the Theory Guide) are specified in the Solution Methods and Solution Controls task pages, respectively. The Convergence Criterion (Modifying Convergence Criteria) is set in the Residual Monitors dialog box.
There are no solution parameters to be set for the Rosseland model, since it impacts the solution only through the energy equation.
Important: If radiation is the only model being solved in ANSYS Fluent and all other equations are disabled, then the radiation transfer equation will automatically be solved for every iteration, regardless of what the Energy Iterations per Radiation Iteration field is set to in the user interface.
For the P-1 radiation model, you can control the convergence criterion and under-relaxation factor. You should also pay attention to the optical thickness, as described below.
The default convergence criterion for the P-1 model is 10-6, the same as that for the energy equation, since the two are closely linked. See Monitoring Residuals for details about convergence criteria. You can set the Convergence Criterion for p1 in the Residual Monitors dialog box.
Solution →
Monitors → Residuals
Edit...
The under-relaxation factor for the P-1 model is set with those for other variables, as described in Setting Under-Relaxation Factors. Note that since the equation for the radiation temperature (Equation 5–21 in the Theory Guide) is a relatively stable scalar transport equation, in most cases you can safely use large values of under-relaxation (0.9–1.0).
For optimal convergence with the P-1 model, the optical thickness
must be between 0.01 and 10 (preferably not larger than 5).
Smaller optical thicknesses are typical for very small enclosures
(characteristic size of the order of 1 cm), but for such problems you can
safely increase the absorption coefficient to a value for which
. Increasing the absorption coefficient will not change the
physics of the problem because the difference in the level of transparency of a
medium with optical thickness = 0.01 and one with optical thickness <0.01 is
indistinguishable within the accuracy level of the computation.
When the DTRM is active, ANSYS Fluent updates the radiation field during the calculation and computes the resulting energy sources and heat fluxes via the ray-tracing technique described in Ray Tracing in the Theory Guide. ANSYS Fluent provides several solution parameters in the expanded portion of the Radiation Model dialog box (Figure 12.28: The Radiation Model Dialog Box (DTRM)).
You can control the maximum number of iterations of the radiation calculation during each global iteration by changing the Maximum Number of Radiation Iterations. The default setting of 5 means that the radiosity will be updated up to 5 times. The actual number of iterations will be less if the residual convergence criterion is exceeded at any point during these iterations.
The Residual Convergence Criteria parameter (0.001 by default) determines when the radiation intensity update is converged. It is defined as the maximum normalized change in the surface intensity from one DTRM radiation iteration to the next (see Equation 12–9).
You can also control the frequency that the radiation field is updated as the continuous phase solution proceeds. The Energy Iterations per Radiation Iteration parameter is set to 10 by default. This means that the radiation calculation is performed once every 10 iterations of the solution process. Increasing the number can speed the calculation process, but may slow overall convergence.
For the S2S model, as for the DTRM, you can control the frequency with which the radiosity is updated as the continuous-phase solution proceeds. See the previous description of Energy Iterations per Radiation Iteration.
If you are using the pressure-based solver and you first solve the flow equations with the energy equation turned off, you should reduce the Energy Iterations per Radiation Iteration from 10 to 1 or 2. This will ensure the convergence of the radiosity. If the default value of 10 is kept in this case, it is possible that the flow and energy residuals may converge and the solution will terminate before the radiosity is converged. See Residual Reporting for the S2S Model for more information about residuals for the S2S model.
You can control the maximum number of iterations of the radiation calculation during each global iteration by changing the Maximum Number of Radiation Iterations. The default setting of 5 means that the radiosity will be updated up to 5 times. The actual number of iterations will be less if the residual convergence criterion is exceeded at any point during these iterations.
The Residual Convergence Criteria (0.001 by default) determines when the radiosity update is converged. It is defined as the maximum normalized change in the radiosity from one S2S radiation iteration to the next (see Equation 12–10).
For the discrete ordinates model, as for the DTRM, you can control the frequency that the surface intensity is updated as the continuous phase solution proceeds through the Energy Iterations per Radiation Iteration setting. See the description of DTRM Solution Parameters.
For most problems, the default under-relaxation of 1.0 for the DO equations is
adequate. For problems with large optical thicknesses (), you may experience slow convergence or solution oscillation.
For such cases, under-relaxing the energy and DO equations is useful.
Under-relaxation factors between 0.9 and 1.0 are recommended for both
equations.
For the MC model, as for the DTRM, you can control the frequency that the surface intensity is updated as the continuous phase solution proceeds through the Energy Iterations per Radiation Iteration setting. See the description of DTRM Solution Parameters.
Once the radiation problem has been set up, you can proceed as usual with the calculation. Note that while the P-1 and DO models will solve additional transport equations and report residuals, the DTRM and the Rosseland and S2S models will not (since they impact the solution only through the energy equation). Residuals for the DTRM and S2S model radiation iterations are reported by ANSYS Fluent every time such an iteration is performed, as described below.
When running the S2S model with the parallel solver, note that coupled walls are encapsulated, which may cause problems with the partitioning of the mesh. See Troubleshooting for recommendations on correcting such problems.
The residual for radiation as calculated by the P-1 model is updated after each iteration and reported with the residuals for all other variables. ANSYS Fluent reports the normalized P-1 radiation residual as defined in Monitoring Residuals for the other transport equations.
After each DO iteration, the DO model reports a composite normalized residual for all the DO transport equations. The definition of the residuals is similar to that for the other transport equations (see Monitoring Residuals).
ANSYS Fluent does not include a DTRM residual in its usual residual report that is issued after each iteration. The effect of radiation on the solution can be gathered, instead, via its impact on the energy field and the energy residual. However, each time a DTRM iteration is performed, ANSYS Fluent will print out the normalized radiation error for each DTRM radiation iteration. The normalized radiation error is defined as
(12–9) |
where the error is the maximum change in the intensity (
) at the current radiation iteration, normalized by the
maximum surface emissive power, and
is the total number of radiating surfaces. Note that the
default radiation convergence criterion, as noted in DTRM Solution Parameters, defines the radiation calculation
to be converged when
decreases to
or less.
ANSYS Fluent does not include an S2S residual in its usual residual report that is issued after each iteration. The effect of radiation on the solution can be gathered, instead, via its impact on the energy field and the energy residual. However, each time an S2S iteration is performed, ANSYS Fluent will print out the normalized radiation error for each S2S radiation iteration. The normalized radiation error is defined as
(12–10) |
where the error is the maximum change in the radiosity (
) at the current radiation iteration, normalized by the
maximum surface emissive power, and
is the total number of radiating surface clusters. Note
that the default radiation convergence criterion, as noted in DTRM Solution Parameters, defines the radiation calculation
to be converged when
decreases to
or less.
Sometimes, you may want to set up your ANSYS Fluent model with the radiation model active and then disable the radiation calculation during the initial calculation phase. For the P-1 and DO models, you can turn off the radiation calculation temporarily by deselecting P1 or Discrete Ordinates in the Equations dialog box, which is accessed by right-clicking Solution Controls in the tree (under Solution) and clicking Equations... in the menu that opens. For the DTRM and the S2S model, there is no radiation calculation to disable in the Equations dialog box. You can instead set a very large number for Energy Iterations per Radiation Iteration in the Iteration Parameters group box of the Radiation Model dialog box.
If you turn off the radiation calculation, ANSYS Fluent will skip the update of the radiation field during subsequent iterations, but will leave in place the influence of the current radiation field on energy sources due to absorption, wall heat fluxes, and so on. Turning the radiation calculation off in this way can therefore be used to initiate your modeling work with the radiation model inactive and/or to focus the computational effort on the other equations if the radiation model is relatively well converged.