CFD Online Discussion Forums

CFD Online Discussion Forums (
-   Main CFD Forum (
-   -   *HELP* Convergence problem!! Master dissertation deadline in 2 weeks!!! (

Jennifer Cheung August 24, 2011 04:14

*HELP* Convergence problem!! Master dissertation deadline in 2 weeks!!!

Im a student and Im new to CFD. I got this research project due in two weeks time. Im looking in to partitioning on flow rates with natural ventilation. However convergence (I think) problem occur. I tried refining the mesh and doing all kinds of random stuff (removing furniture, changing probe positions, etc) but it wont help. Residuals are really large and the flow paths werent the way I expected it to be - coming out of one of my light well openings instead of going in.
I use quadratic loss coefficient of 2.69 at my openings.
The model has 62x54x52 cells and the simulations run for 5000 iterations.
Anyone knows what to do?

Many thanks!!!!!!


Jennifer Cheung August 24, 2011 04:17

1 Attachment(s)
the monitor plots x

Eezyville August 25, 2011 20:23

Hey Jennifer,

You aren't giving us enough information to work with. But I think you should recheck your boundary conditions. You said that your light well opening is pushing air out instead of pulling it in. Check the back pressure and make sure its low enough to suck air into the well. You may also consider playing with the time step. You may need to lower it if you increased the number of cells.

Also you didn't mention what code you are using but you should look and see if you can use a Riemman Solver to add dissipation to the solution. If it converges from the solution with dissipation added turn off dissipation and run from the converged solution and hope that the dissipation free solution converges.

Good luck

Jennifer Cheung August 26, 2011 03:56

Hi thanks for your suggestions!! i tried to provide more information as i can. i was trying to upload my model but i didnt know how to :p.
the software im using is PHOENICS. i ve tried using false-time-steps, but whatever value i tried (tried 0.25, 0.1, 0.05, etc), the results went mad e.g. air velocity 3000m/s :confused: and i didnt understand why.
im very new to CFD (only half a year) so im not sure if there is an option for Riemman Solver in PHOENICS. but i will try to find out!
i also notice that after 100 iterations the flow path was correct and the flow rates/velocity/temperature are reasonable, but the errors are not small enough. however if i continue to run it for more iterations the flow paths changed and the errors become bigger and bigger. if you know whats wrong that will be a great news for me!!
ps: just tried apply pressure -0.01pa at my lightwell opening and it didnt work :(.

many thanks!!!

Jennifer Cheung August 26, 2011 09:25

some information:

im looking into the effect of internal partitioning on air flow rates. my model is a one storey connecting to a light well at one end and a stack at the other end. im investigating 3 cases. a partition with small openings is added in each case. the partitions are in series. all partitions are set as adiabatic with 0 thickness.
the first 2 cases are fine. problems described above occurred when i added the third partition.

mettler August 26, 2011 10:03

I agree with Eezyville, I would try adding more cells - A lot more cells. You should do a grid check anyway. Still, I would add a zero to each of your cell numbers and try that.

Jennifer Cheung August 26, 2011 12:58

thanks for your advise mettler!! i go try it now and will let you what ive got! :)

All times are GMT -4. The time now is 01:09.