CFD Online Discussion Forums

CFD Online Discussion Forums (http://www.cfd-online.com/Forums/)
-   FLUENT (http://www.cfd-online.com/Forums/fluent/)
-   -   Calculating time (http://www.cfd-online.com/Forums/fluent/47777-calculating-time.html)

thuy April 2, 2008 05:23

Calculating time
 
Hi all,

I have to simulate the premixed combustion in a furnace. The dimension of my model is 4m*3m*3m, and there are 12 burners. I created the geometry with Gambit and meshed it. The number of mesh is about 60 000 (type tetrahedral). I used "premixed combustion", the model "k-epsilon",the under-relaxation factor=0.5 for alls, and time stepping "fixed"=0.01s. Have you any idea about calculating time for this model? It took 2h (calculating time) for the result at 0.3s (real time). How can I reduce the calculating time?

Thank you in advance.

Thuy

Aris April 7, 2008 09:31

Re: Calculating time
 
Cannot you switch to steady? Are you using only 1cpu? If you have a two/four cores in your pc you can run fluent in parallel. From command prompt you should go at your folder (the folder with your case) and type: flunt -rx.x.xx 3ddp -t4 -rx.x.xx is the version of fluent you use. For example --r6.3.26 3ddp is for 3d double precision. -t4 means that you will use 4 cores (computing nodes). If you have dual core you should write -t2.

You should also split your mesh from fluent panel.

Good luck! Isn't your time step a little big? Have you checked the courant number?

thuy April 7, 2008 10:08

Re: Calculating time
 
Hi Aris,

Thanks for your reponse. I simulated my model with my laptop HP 5000 Processor 1.6 GHz RAM 512MB, so i can't do anything with the parallel processing.

I agree with you that the time step which i set is rather big. But If I choose the time step adaptive (t initial=0.0001s), the calculating time is longer. I checked the courant number,it varied in servaral zones from 1.83e-4 to 2.06.

And U talked about "split mesh", why and how I have to split them? If the number of Mesh increases, the calculating time is more important?

Thank you very much! Thuy


Aris Nikolopoulos April 8, 2008 04:07

Re: Calculating time
 
Spliting the mesh refer to running in parallel. So let's leave it out of our disgustion.

First courant number > 1 is a problem. In fact, with such CFL results' accuracy is doubtfull. You should have a time step 1-3 msec. In every cell you should have CFL<1

About computational cost. Your CPU and RAM are low. About CPU you cannot do many thing. Of course you can buy a new one. You should also check your RAM. If you see that RAM used when you run your case is above 512 MB, this is the MAIN reason that you have such computational cost. Your system in that case takes from your hard disk the nessecary space (swap memory) and this is a very slow procedure.


thuy April 8, 2008 13:04

Re: Calculating time
 
Thanks again for your reply. I checked my RAM, it is 485MB. It does not yet exceed the limit but as you say, it is possible to slow my calculate. I should buy a new CPU!

Just one last question: Can I devide the dimension of my model by 10? It could reduce the number of mesh, so the computational cost, but i'm not sure it still gets the same combustion phenomenal.

Sincerely Thuy

Aris Nikolopoulos April 10, 2008 05:23

Re: Calculating time
 
No you cannot! If you should also scale Reynolds e.t.c. and you will see that that the appropriate cell volume for your simulation will be decreased. Thus you will have the same mesh!


All times are GMT -4. The time now is 00:01.