CFD Online Discussion Forums - adaptive time stepping and convergence

CFD Online Discussion Forums (https://www.cfd-online.com/Forums/)

- CFX (https://www.cfd-online.com/Forums/cfx/)

- - adaptive time stepping and convergence (https://www.cfd-online.com/Forums/cfx/147243-adaptive-time-stepping-convergence.html)

adaptive time stepping and convergence

Hi,

I was wondering if there is a way that you could set different residual convergence targets for different equations when using the adaptive time step. Is this possible? For example 1e-4 for continuity and 1e-5 for momentum etc. Thanks in advance.

I think there is a way
in solver control setting you can adjust convergence control criteria for continuity and momentum and turbulency separately.

No you can't do this. You can however have different schemes for advection and transient.

Quote:

Originally Posted by RicochetJ (Post 527943)

No you can't do this. You can however have different schemes for advection and transient.

Actually, you can. I asked Ansys support this questions a few years ago and their reply was that for transient models the option was not exposed in PRE, but could be changed using CCL. There will be an error message but it can be ignored.

This example shows how to set a different convergence critera for an additional variable.

Code:



FLOW: Flow Analysis 1

SOLVER CONTROL:

   Turbulence Numerics = First Order

   ADVECTION SCHEME:

     Option = High Resolution

   END

   CONVERGENCE CONTROL:

     Maximum Number of Coefficient Loops = 10

     Minimum Number of Coefficient Loops = 1

     Timescale Control = Coefficient Loops

   END

   CONVERGENCE CRITERIA:

     Residual Target = 1.E-6

     Residual Type = RMS

   END

   EQUATION CLASS: av

     ADVECTION SCHEME:

       Option = High Resolution

     END

     CONVERGENCE CRITERIA:

       Residual Target = 1.E-5

       Residual Type = RMS

     END

     TRANSIENT SCHEME:

       Option = Second Order Backward Euler

     END

   END

   TRANSIENT SCHEME:

     Option = Second Order Backward Euler

     TIMESTEP INITIALISATION:

       Option = Automatic

     END

   END

 END

END

Wow, that's good to know. I take it then you must replace the CCL file in command line prior to running the simulation.

Quote:

Originally Posted by RicochetJ (Post 528117)

Wow, that's good to know. I take it then you must replace the CCL file in command line prior to running the simulation.

Yes, do it either at the command line or right before you save the .def in PRE. Note that for additional variables there is just one set of options, so it is not possible to have different convergence criteria for different AVs.