Fluids

Fluids

Residuals and Convergence

    • Noelly
      Subscriber

      Hello, 


      I know there are many posts about residuals and convergence, sorry to bring up another one...


      I am using a k-omega model and this is what my residual plot looks like:



      I am not very experienced with Fluent, I just finished the Cornell series "A hands-on introduction to engineering simulations" and I guess I still need more practice. But would you mind helping me with the following:


      1. From this discussion (https://forum.ansys.com/forums/topic/how-to-get-system-convergence/?order=all#comment-7bb6d7e2-172d-49de-82e8-aaa90073b115) I learned that there are scaled and non scaled residuals, and that I can monitor other variables to check convergence. Why is it the case that it is more difficult for the continuity residuals to achieve the target convergence in some cases? 


      2. I am confused as to how the solver treats the values we input as convergence criteria. What I mean is does the solver tries to reach convergence "equally" or is there an existing "competition": if for example I choose 1e-5 for continuity and 1e-3 for all the other, will the solver behave differently from a case where all the values would be 1e-3? 


      3. I observed similar trends in this thread as well (https://forum.ansys.com/forums/topic/convergence-of-steady-state-turbulent-airflow/?order=all#comment-f8a999b3-4598-456c-9b6b-aa9d00b5fad6)and found out that there are Fluent lectures available. Can anyone help me accessing them please? I could not find them on the website. Or is it part of the subscription? If so would you have resources that I could consult specifically about the topic of convergence?


      (Oh and I checked, the values for my average and maximum skewness are 0.23 and 0.87 respectively)


      Thanks!

    • Rob
      Ansys Employee

      Simply residuals are a function of the error in the model. The higher the residual the less good the solution is. Note, we can have fairly high residuals due to local effects but overall have a decent solution.  Read up on the mass flux reports and variable monitors too. 


      Continuity is related to the mass imbalance in the model, but is also (as you can see) the hardest residual to converge: plot contours of Residual - Mass Imbalance and compare that with other variables.  Meshing is more that just cell quality, you need to consider resolution and growth rate. A solution will hit convergence (numerically) when it hits all of the residual check values. Reaching 1e-5 will more accurate numerically than 1e-3  BUT that doesn't mean the solution is correct as you may have made an error in one of the boundary settings. 


      Look in Help in the solver. From about R19.x there are some tutorials available in the top level section.  Course lectures are available via the ANSYS Learning Hub which needs a subscription. There are other materials, but I don't have links (they're not ours!). 


       


       

    • Noelly
      Subscriber

      Thanks!

Viewing 2 reply threads
  • You must be logged in to reply to this topic.