How does one judge the convergence of an AcuSolve CFD solution?

acupro

The concept of convergence is subjective - different users will give different responses.  Some general comments related to AcuSolve:

1. Residual Ratio - an overall measure of how well the solution matches the equations being solved
2. Solution Ratio - an overall measure of how much the solution changes from time step to time step (for steady-state), or within the time step (for transient)

In general we would want both of these to be quite low.  'Low' is subject to interpretation, and the default convergence tolerance of 0.001 is usually good for AcuSolve simulations.  (This 1.e-3 value is similar to 1e-5 or 1-6 for some other solvers, as with AcuSolve the normalization factor is updated every time step.)  By default, the Residual Ratio would need to be below the specified tolerance (0.001 default) for pressure, velocity, temperature, species.  For other quantities like turbulence this could be a factor of 10 higher (0.01 default).  Also, by default, the Solution Ratio could be a factor of 10 higher than the residual ratio values.  You can review the settings in the CONVERGENCE_CHECK_PARAMETERS command which will be included in the <problem>.ss.inc file generated by the AUTO_SOLUTION_STRATEGY command.   This file shows the default stopping criteria, based on the convergence tolerance.  A value of 'standard' means at or below the convergence tolerance value, 'looser_by_10' means one order of magnitude higher, etc. This information will also be in the <problem>.<run>.echo file.

You would also want to track the results of interest and see how well they are reaching a 'constant' value for steady-state.  If the default convergence has been met, but the solution of interest is still changing more than is acceptable to you, then maybe reduce the convergence tolerance and allow it to converge more.  (This is something of a convergence sensitivity study.)  By the same token, if the results of interest have come to what you consider a steady or stable solution, but the default convergence has not been reached, you may still decide the solution is acceptable.  This is where 'engineering judgement' is useful.

Convergence itself does not necessarily mean you have an accurate solution.  You would still want to perform a mesh sensitivity study to see how the results of interest change as the mesh is refined (volume size, surface size, boundary layer parameters, etc.).  You could also perform other sensitivity studies (effect of small changes in boundary conditions, initial conditions, turbulence model choice, etc.)

Bottom line - 'Convergence' is more of a convention or practice rather than a hard definition.  The Altair CFD tools like AcuSolve provide good methods to help judge convergence.