[frossi]

Hi friends,

Can you please answer some of my questions of the CFX monitors? I am trying to understand how to determine convergence of a TRANSIENT solution more precisely.

1. In the first picture, you see the RMS plots. Is that the plot for the RMS of the whole solution, or the RMS change for each time step? How can I tell if the solution converged by looking at this plot?

2. The second picture shows the K-epsilon turbulence plot. Do I need to look at it for convergence?

3. I don't understand the plots in picture 3 (ANSYS field solver structural) and 4 (ANSYS interface loads structural). What do they represent? Do I need to look at those for convergence?


Thank you very much. Please let know

[ghorrocks]

1. The RMS plots are the RMS of the control volume residuals over the simulation at that iteration. If you have established a certain residual is required for convergence then you can use it as the convergence tolerance. In fact this is what is used to define convergence in the majority of simulations.

2. Yes, you need the residual for all equations to be acceptably low.

3. I will let other more experienced in FSI answer that - but this is described int he documentation so make sure you check that.

[frossi]

Thank you very much for the clarification. But I didn't understand when you say

The RMS plots are the RMS of the control volume residuals over the simulation at that iteration.

could you please explain in more detail please?

also, my convergence target is 1E-4, and the picture shows the plots well below that (around 1E-6). So does that mean that the solution already converged? why then continuing running the solver if the solution is already well below the tolerance?

Thank you again

[ghorrocks]

First of all - you need to understand how convergence works on a simple flow before you tackle a FSI simulation. Convergence on a FSI case is considerably more complex.

In a simple flow the residual of the linear solution is generated at each control volume as part of the solution procedure. Then either a RMS or MAX is used to generate a residual convergence number which is compared to the requested tolerance.

If your model is showing residuals way below the residuals tolerance this means the FSI loops are repeating the fluids solution, but the fluids solution is already solved to an accuracy beyond your tolerance. It is the FSI convergence which is doing this.

[frossi]

thank you, your clarification was extremely useful. One last question:

do we consider the solution to be converged only if the plot becomes a line (constant value), or a converged solution can also be an oscillating plot, which keeps a constant oscillation? (like in the picture).

[ghorrocks]

The image is from Fluent. I do not know how Fluent calculates its residuals so cannot say.

But for CFX, in most cases you are just trying to reach the residuals tolerance you define. So if it flat lines (either completely flat or oscillating) that means you have not achieved convergence. But read this FAQ for more information: http://www.cfd-online.com/Wiki/Ansys...gence_criteria