Iteration: convergence

    • raza

      Hello all, I'm fairly new to CFD and still trying to learn. I have performed a simulation for shell and tube heat exchanger and following is my iteration. Is the system converged? What are the criteria to prove that my results are accurate?

    • peteroznewman

      Hello Raza,

      I'm also learning CFD, so I welcome any expert comments. 

      My understanding is 10^-3 is a minimum convergence criteria on all the metrics plotted. However, if the curves are still showing a downward trend when they reach that value, then reduce the convergence criteria to 10^-6 and keep going.

      If the curves have gone flat long before reaching 10^-3, it probably means the mesh is not of sufficient quality or the elements are not small enough at the walls. Do you know what the Y+ value is for your mesh?  If you don't know what Y+ is, then you need to do some more study.

      You need experimental data to prove that your results are accurate, but you can show your model has converged by remeshing with smaller and smaller elements and plotting some important value from the result to show that value doesn't change when the elements get smaller.

    • rachels1001

      Hey Raza, one thing to also consider might be the under-relaxation factors you are running your simulation at. If your simulation residuals are starting to flat-line, you could try increasing them to get a more precise solution. 

      Additionally, while residuals are a really good indication of convergence, in some complex simulations you may never get sufficiently small residuals, so it's good to have extra monitors of known variables, such as heat flux, flowrates, etc. to get an idea of how your simulation is performing. These are found in the same post-processing tab as the residuals. 

    • pgl
      Ansys Employee

      Hi cevlikonja,

      Just wanted to say "+1 thumbs up" for helping our Student Community members with their CFD questions.

      It's good people like you that allow us to scale up this community into a valuable resource for our academic users.      


Viewing 3 reply threads
  • You must be logged in to reply to this topic.