The residual is a measure of the local imbalance of each conservative control volume equation. It is the most important measure of convergence as it relates directly to whether the equations have been solved accurately. Additional information on how the residual is calculated is available. For details, see Linear Equation Solution in the CFX-Solver Theory Guide.
CFX uses normalized residuals to judge convergence. For details, see Residual Normalization Procedure in the CFX-Solver Theory Guide. By normalizing the residuals, you are presented with a relatively consistent means of judging convergence. The normalized residual is used to automatically stop the CFX-Solver run when the specified target level is reached.
In transient simulations, the same target residual is used to control the termination of the coefficient iterations, except that the residual now includes a contribution from the transient term. If the maximum number of coefficient iterations per time step has been set to a large value, such as 20, then when the coefficient iteration residuals are below the target residual, the coefficient iterations will stop. This method can be used to cause the CFX-Solver to run with many coefficient iterations early in the run (when residuals start relatively high) and fewer coefficient iterations later in the run.
To assess when convergence is reached, you can select a residual type, either MAX (maximum) or RMS (root mean square), and specify a target value.
Note that the residual level for turbulence transport equations
(for example, k, 
, Reynolds stress components, and
turbulent heat flux components) are not included by the solver in
deciding when convergence is reached.
Note also that for multiphase simulations, the convergence criterion for the volume fraction equations is a factor of 10 higher than the specified target residual.
Although the required convergence level depends on the model and on your requirements, the following guidelines regarding RMS residual levels may be helpful.
Values larger than 1e-4 may be sufficient to obtain a qualitative understanding of the flow field.
1e-4 is relatively loose convergence, but may be sufficient for many engineering applications. The default target RMS residual value is 1e-4.
1e-5 is good convergence, and usually sufficient for most engineering applications.
1e-6 or lower is very tight convergence, and occasionally required for geometrically sensitive problems. It is often not possible to achieve this level of convergence, particularly when using a single precision solver.
MAX residuals are typically 10 times larger than the RMS residual; The above guidelines for RMS residuals can also be applied to MAX residuals, with the targets increased by a factor of 10.
Sometimes, however, the MAX residuals are much larger (for example, a factor of 100) than the RMS residuals. In this situation, it is very likely that the region of high MAX residuals is isolated to a very small area of the flow, typically where some unstable flow situation exists (for example, a separation or re-attachment point, and so on). It may be the case that this small area of unstable flow / lack of tight convergence of the MAX residuals do not affect the overall prediction. To verify that the solution is acceptable, you should verify that the variation of relevant quantities (for example, lift, drag forces, efficiency of the device, and so on) is small.