Learning Forum

[Farah]

Hello folks! 

I'm running a two-way FSI simulation but don't know much about convergence. Here are my current settings and issues:

• Current Settings:

  • URF (Under-Relaxation Factor): 0.15 (from fluid to structural)
  • RMS Convergence Target: 1e-7

• Issues:

  • The graphs should both look sinusoidal, but the sinusoidal pattern in the fluid is more visible (larger) compared to the structural.
  • Changing RMS to 0.2 resulted in a "negative cell volume error."
  • Increasing the Young's modulus allows for a slightly higher URF, but the issue with the graphs persists.

Which factors should I change to make both graphs look alike (sinusoidal)?

[Federico]

Hello Farah, 

First, is this a steady-state anaylysis or a transient one? For the former, you should see steadily decreasing residuals, while for a transient analysis you would see a jump at a new Coupling Step, with steady decrease with each Coupling Iteration. Not sure what you mean with sinusoidal pattern here. 

These plots show the convergence of the transfered quantities, by computing the RMS change by for each new iteration. A solution is considered to have converged when is no longer changing with new iterations.

[Federico]

Second, Under relaxation factors will allow you to effectively tune the convergence rate of your solution. A low URF (near zero) will make convergence slower but make your solution more stable, and vice versa for URF closer to 1. 

There could be several reasons specific to your case which would explain why your solution sees negative volume cells with URF of 0.2, which is on the low side. Some of these reasons could be:

• use of incompressible fluid
• bad Dynamic Mesh settings
• time step size too large (if transient case)