Iterations... what is this ?

[Anonymous]

Hello, 

i did a simulations with 1 iteration and with 1000 iterations, the results are 98% the same.

how(what exactly) does CFD calulates the model, where is the difference between low and high iterations(steps) ?

do i geht better results with more iterations ? or is just the number of nods of the mesh important for better results ?

Thank you.

[marwan_azzam]

Hello @Anonymous

To put it simply, in a CFD analysis, the solver does not directly calculate the final results (velocity, pressure, etc...) in one straight shot.  It calculates once, checks for convergence then calculates again.  Each cycle is called an iteration.  By default, the solver repeats this process till it converges or reaches the maximum number of iterations specified by the user (whichever comes first).  This happens no matter if you have a coarse or fine mesh.

As far as the mesh density, a rule of thumb is that the finer the mesh the more accurate the results are.  Of course refining the mesh beyond a certain point yields minimal of no gain in accuracy.  This is what we call mesh convergence, not to be confused with convergence described in the paragraph above.

Now as far as your model giving practically the same results with 1 or 1000 iterations I cannot comment on that without seeing the model.  It is unusual.

Marwan

[frederic.gaillard.7]

Hello 

I had the same question ! 

Here is what i know , but you should confirm my answer with the autodesk guys. 

• do i geht better results with more iterations ?
  • If you model is correctly set, Yes definitively !
    • In steady state mode you don't have to choose the number of iteration, the solution will stop by herself when the convergence assessement will meet the condition. To do that you have to enable  the toggle box ''intelligent control solution''  in solution control and choose the convergence requierement that you want to meet. 
    • In transcient mode i have exact the same question as you. not sure if i can provide you a good answer. 
• ''or is just the number of nods of the mesh important for better results ?''
• In fact it's a combinaison of both (mesh and number of iteration). You have mesh dependancy which cause a slight change in result when you increasing the number of element. If your mesh is too coarse, your simulation will not be that accurate as a dense et uniform meshing. 
  

You can check my post on the forum ''advection scheme'' here i join a word document where i simulate the same geometry with different convergence assessement criteria and different number element (number of mesh) 

Fred