Fluids

Fluids

FSI – continuity stability solution

    • 1993saurav
      Subscriber

      I am doing a 3D FSI in a  rectangular pipe that has a small flap in the inside attached to the inner section of the pipe. The simulation is up and running with the usual setup but after a few timesteps, I get a very oscillating continuity chart and the thing doesn't converge. I have troubleshot it by reducing the flow courant number and explicit relaxation factors as shown but after a few successful time steps, the same thing happens with the solution which I have shown in the figure.


      I know for sure reducing the time step will stop that but I have also experienced that after maybe say a 100-200 converged timesteps, the same oscillation happens.what is the best way to get rid of it? how low can we go with URF?i have gone as low as 0.1 and nothings changed. Do we need to change the density and body forces URF as well? this is my current solution controls setting. Do I need to change CFL and ERF both or just one of them? I have read somewhere that only one needs to be hit and trialed with.


    • Rob
      Ansys Employee

      Drop the Courant Number & time step rather than the URF: you use the time step to control in a transient case. Also look at what is changing, and at what rate. 

    • 1993saurav
      Subscriber

      I have tried reducing the courant number to 10. The manual suggested that it should not be reduced below 10. Can I go below 10? Also while reducing the courant number, should I leave the URF as default? And when you say "look at what is changing and at what rate?", what do you mean by that?. I do monitor the surface integral pressure and the residual chart during the iterations. Is there something else I should be monitoring?

    • Rob
      Ansys Employee

      Courant Number can drop below 10, but will start to really slow the rate of convergence. Leave URF alone, but you may also need to reduce the actual time step. In terms of "what's changing" look at the contours of velocity etc. What is going on in the flow field, are all of the variables sensible, do you have any regions with high velocity/gradients? 

    • 1993saurav
      Subscriber

      yeah, I actually managed to complete the rectangular chamber simulation by adjusting the URF and other stuff and since it was a pretty simple simulation, I was able to get around some issues and get to the final solution.


      However, now I am working with a complex setup which is more like a varying diameter pipe, with flaps inside, acting as a valve. I have a velocity inlet and pressure outlet. The problem seems to be same i.e the oscillation residuals in continuity as I mentioned before. This time I cant bypass these issues. So, I did what you told me that is to reduce the courant number and keep the URF the same. I also reduced the time step to 1e-5 for courant number of 10. I noticed that the solution more like diverges when I go from courant number = 200 to 1. it's not oscillating around the convergence value .its diverging after a few iterations. I have included the pictures to explain this.the the second picture is for CFL=0.1 and time step=1e-5 the first one is for higher CFL =10 and time step 1e-5.


      I know there's a formula to calculate the time step based on CFL number but my initial velocity is 0.I have a transient velocity profile so I don't know if that is useful in the calculation. any insight into that as well would be helpful.

    • Rob
      Ansys Employee

      Drop the time step again and see how it behaves. Also look at the results at the time step before the solver fails. You're looking for skew cells (if mesh is deforming), narrow gaps, badly resolved flow gradients etc. 

Viewing 5 reply threads
  • You must be logged in to reply to this topic.