CFD Online Discussion Forums (https://www.cfd-online.com/Forums/)
-   FLUENT (https://www.cfd-online.com/Forums/fluent/)
-   -   convergence problem(grids) (https://www.cfd-online.com/Forums/fluent/41016-convergence-problem-grids.html)

 Siddharth Khowala May 17, 2006 13:32

convergence problem(grids)

hi, When solving a 2D domain of (10*0.1) with a coarse grid(50,20) mesh convergence is obtained pretty quickly(residual was set as 1e-5 in both cases) but when solving the same problem with a fine grid of (150,150),it started giving reversed flows and started diverging.My question is how does grid affect the solution,would the solution obtained with a rough grid be accurate enough so that a fine grid would not be required.I also did the simulation using transient analysis + coarse grid with residuals set to 1e-3,still i am getting reversed flows.Plz help and shed some light in this matter. TIA

 Bagus Putra Muljadi May 18, 2006 02:47

Re: convergence problem(grids)

I think every correct solutions has to be grid independent. rough grid will lead to a poor monitoring on fluid properties in areas where the properties tend to change drastically in a short time.

I also think that reversed flows won't lead the iteration to divergence. Perhaps the reversed flows are already exist in the previous iteration but were mis-observed due to the rough grid.

 HS May 18, 2006 03:51

Re: convergence problem(grids)

I agree with Bagus. Backflow can occur at the beginning of iterations even if it does not exist in the final solution and is usually no problem. Though, if you suspect that it has anything to do with the solution diverging, make sure you have realistic backflow settings in your outlet boundary conditions.

An acceptable CFD solution should always be grid independent. That is, you should get the same solution on any fine enough mesh. If solving the problem on a finer mesh gives another (or no) solution, then you cannot trust your coarse mesh solution.

Also, you may want to think about how you constructed the finer mesh. Did you do a new mesh from scratch (that you know is OK) or did you use any of the adaption features of Fluent? In the latter case, make sure you check volume change and y+ etc, as your new mesh could be badly adapted and of low quality, which can have adverse effect on the chances of reaching convergence.

Hope any of this helps!

/Henrik

 Siddharth Khowala May 18, 2006 04:43

Re: convergence problem(grids)

hi, thanx for all the positive feedback for my question. Yes i had realized that my solution should be grid independent and i am doing the necessary to check it.The grid was constructed without using adaptive meshing and i guess that part is ok,but the backflow problem still persists .I am trying to model the roughness effects in laminar flows(i know it does not have much effect)and hence i guess to expect backflow would not be wrong in some situations.That fact is corroborated by my increase in velocity in the inflow which then dissipates these eddies caused due to roughness and solution converges well without any backflow.If i am wrong somewhere please feel free to point out the mistakes.Also any other idea would be welcome TIA

 All times are GMT -4. The time now is 18:53.