MUSIG convergence difficulties

I am performing a steady state simulation of air water bubbly flow in a vertical pipe. I have a converged solution with a monodispersed set up.

I am now trying to do a polydispersed MUSIG model. The residuals of the size group equations are just oscillating around RMS 1E-2. All other residuals are well converged. I am monitoring the volume fraction at the outlet. It is oscillating rapidly with a very small amplitude, however the mean value has been constant for quite a while and not moved far from the monodispersed result.

I started with the auto timestep, I have tried increasing and decreasing it. I have tried 1 and 2 velocity groups with a number of different bubble sizes

No matter what I have tried the size fraction residuals will not decrease.


  • DrAmineDrAmine GermanyForum Coordinator

    I won't pay much attention to the residuals of the bin equations. They have been always a bit odd. Consider standard residuals as well as domain imbalances to judge convergence numerically as well as monitor points to judge it from physical point. Aslo check running with coupled volume fraction. Better to use a constant physical time for the false transient approach.

  • Thank you for the advice. I have been solving with the coupled volume fraction but I will switch to a constant timestep.

    The imbalances of some of the size fraction equations are very large but they are being normalized by very small values, so I am ignoring those. The mass and momentum residuals are well converged. However, my results are not matching the published result I am trying to reproduce.

    I do not think my results are converged because the Volume fraction of my second velocity group is nearly zero.

  • DrAmineDrAmine GermanyForum Coordinator

    Reduce first of all any kind of numerical errors and then start quantifying the models and kernels you are using (model errors) and reduce any kind of user errors ( wrong input, wrong BC,..)

Sign In or Register to comment.