[Pierre1]

Typically there are two convergence criteria

1. Residuals drop to enough low value
2. Output parameters (the parameters we care about in the simulations results) don’t change anymore with more simulation iterations

It seems that we can only set up the 1st criteria in CFX and Fluent. Regarding the second one, we need to either monitor by eyes or setup to run absolutely enough iterations number to make sure the simulations satisfy it. Both of these two ways are costly.

Anyone has a better way to set up the second criteria?

[ghorrocks]

In CFX V15 you have more options to declare convergence than those you list. I don't have CFX V15 at hand so I cannot check this, but I remember you can declare convergence on:

* Residuals
* Imbalances
* CPU time
* A user specified criteria, usually that a user specified value has converged to within a tolerance

That selection will cover most people's requirements.

[Pierre1]

Originally Posted by ghorrocks

In CFX V15 you have more options to declare convergence than those you list. I don't have CFX V15 at hand so I cannot check this, but I remember you can declare convergence on:

* Residuals
* Imbalances
* CPU time
* A user specified criteria, usually that a user specified value has converged to within a tolerance

That selection will cover most people's requirements.

Fortunately I use Ansys 15.0. But I didn’t that many functions as you said. Attached figure shows all the options that I have regarding the convergence criteria. It does work work with Residuals, conservation, CPU time but nothing else.

[ghorrocks]

Residuals - that is controlled by "Residual Type" and "Residual Target"
Imbalances - that is controlled by "Conservation Target"
CPU time - that is controlled by "Elapsed Time"
user specified criteria - that is controlled by "interrupt control"

So they are all there. My memory was correct