[medaouarwalid]

Hi. I am doing a numerical study on air jets. i want to compare 3 rans models: K epsilone , rsm and sst k omega. Should i do a mesh independent study for every model then comparing ?. Or i chose one model , i do the mesh study then with the suitable mesh i do simulation with the 2 other models and i compare wich model is better in my case

[CFDfan]

Originally Posted by medaouarwalid

Hi. I am doing a numerical study on air jets. i want to compare 3 rans models: K epsilone , rsm and sst k omega. Should i do a mesh independent study for every model then comparing ?. Or i chose one model , i do the mesh study then with the suitable mesh i do simulation with the 2 other models and i compare wich model is better in my case

My 2 cents.
I have never noticed that the model got re-meshed by the software after I changed the turbulence mode, so I would optimize the mesh with one of the modes and run the other two turbulence modes with the same mesh. My feeling is that it will give you more realistic comparative results as well because if you change both the mesh and the turbulence mode you'd never know what is causing the difference (if any).

[medaouarwalid]

Originally Posted by CFDfan

My 2 cents.
I have never noticed that the model got re-meshed by the software after I changed the turbulence mode, so I would optimize the mesh with one of the modes and run the other two turbulence modes with the same mesh. My feeling is that it will give you more realistic comparative results as well because if you change both the mesh and the turbulence mode you'd never know what is causing the difference (if any).

I thaught it would be better if i compare results of the models with their final mesh. For example, if independence of k epsilone occured with 3 milions éléments. It doesent mean that this number will lead to an independence using RSM, maybe with RSM i will need 5 milions. So i compare results of 3 milions K epsilone with 5 milions RSM. I hope you understand me

[FMDenaro]

My opinion is that each test would represent a different study and meaning of the comparison.
One could wonder about the best performance of several softwares and/or turbulent models on the same chosen grid.
As well as one could check the statistical solution for "vanishing" grid size (provided that the model allows that).


I don't think that a way of doing a comparison excludes the others.

[CFDfan]

Originally Posted by medaouarwalid

I thaught it would be better if i compare results of the models with their final mesh. For example, if independence of k epsilone occured with 3 milions éléments. It doesent mean that this number will lead to an independence using RSM, maybe with RSM i will need 5 milions. So i compare results of 3 milions K epsilone with 5 milions RSM. I hope you understand me

That makes sense too, but as FMDenaro noticed above, the test becomes three separate studies, which blurs the comparison targets. If the results from all three studies are the same, or with very small deviation, you could derive which method gives you the shortest solution time, which is good. If the results differ and you have the experimental results to compare with you could derive which turbulence model is the most accurate, which is very good too. If however you don't have experimental results to compare with and the CFD results from the three studies differ then what would be the conclusion? In any case I would appreciate if you a willing to share your conclusions when you complete the test

[winscombebob]

Originally Posted by medaouarwalid

Hi. I am doing a numerical study on air jets. i want to compare 3 rans models: K epsilone , rsm and sst k omega. Should i do a mesh independent study for every model then comparing ?. Or i chose one model , i do the mesh study then with the suitable mesh i do simulation with the 2 other models and i compare wich model is better in my omegle-online.com case

Do a mesh independent study for every model