Topics relate to Fluent, CFX, Turbogrid and more

Calculation doesn’t converge : Residus curve oscillates for infinite

    • manelaero2020

      Hello evry one ;

      I am simulating a swirling flow inside a cyclonic filter.

      I have tried two turbulence models :

      1/ Standard K_Epsilon

      2/ K-epsilon -RNG Swirling flow ( this is the most appropriate for swirling flows)

      Methods are :

      3/ Simple for velocity pressure coupling ----For both turbulence models

      4/Upwind second order for turbulence , mass , momentum ---- For both turbulence models

      Calculation with the standard K epsilon model , has converged after 280 itérations

      But Calculation with the K epsilon RNG swirling flow model has not converged , the residus curve continued to oscillate ( instable) even aftre 6000 iterations ( a time of 1:30 h) . I have stopped calculation , because i have concluded that solution will not converge , i will just waste my time.

      For more details , please see figures attached.

      Since RNG Swirling flow is better than k epsilon to model cyclone inside filter , i prefer using it rather than standar k epsilon model.

      I ask you please :

      A/ Can you find an explanation about why the residus curve oscillates continually and solution has not converged even after more than 6000 iterations ?

      B/ Advice me to change something ? in the order to accelerate convergence of the model ?

      Remark : I have tried to reduce the value of residus from 0.001 to 0.01 but , also used Quick Scheme instead of simple but the problem is always here

      Thanks tou very much

      Kind regards

    • Federico2594
      Hi Aero2020,nDo you have evaluate your mesh?nI think probably that the mesh is not good to describe your phisic problem. Are you in Grid Indipendent solution?nDo you have evaluate your y+ for your turbolence model? Is the inflation layer correct? nDo you begin the calculation with first order to prepare the solution field and than change to second order?nWithout being presumptuous i think you have to see the theory before and then start the calculation.nRegards,nFedericon
    • Rob
      Forum Moderator
      Read up on precessing vortex cores: you've probably refined the mesh sufficiently to pick up the transients in the flow field. Also have a look at the model limitations for k-e and k-w models in highly swirling flows: hint, read up on the Reynolds Stress Model. nmakes good points; in this case y+ is less important as the complex flow is away from the wall: so you need to resolve the volume as well as the near wall. You also want to look at using a hex or poly mesh. n
Viewing 2 reply threads
  • You must be logged in to reply to this topic.