CFD Online Discussion Forums

CFD Online Discussion Forums (http://www.cfd-online.com/Forums/)
-   FLUENT (http://www.cfd-online.com/Forums/fluent/)
-   -   Discuss about computation time of FLUENT (http://www.cfd-online.com/Forums/fluent/45814-discuss-about-computation-time-fluent.html)

Kelly August 22, 2007 11:01

Discuss about computation time of FLUENT
 
Hello,

I am simulating flow using FLUENT in a small scale rectangualer pond (only 2m*4.6m*2m).The computations are really so demanding and time consuming.1 second is computed in about 2-3 hours. I wonder is it applicable to use FLUENT in a long time simualtion such as a storm which may last for several hours or even a day. If so it will last for at least half a year to finish computation.

My Problem:

- 3D unsteady simulation, - VOF (air-water), - k-epsilon, - PISO, - pressure: Body Force Weighted, - default under-relaxation factors - mesh size=0.1 m.

Can you present here some of your experiences? Thanks,

Kelly

Dusan August 22, 2007 11:20

Re: Discuss about computation time of FLUENT
 
Hello Kelly,

I have the same problem but it is in 2D and I want to shake it, with pond. And your Courant number is OK?

Kelly August 22, 2007 15:11

Re: Discuss about computation time of FLUENT
 
Thanks,Dusan. My Courant number is 0.25.

AAA August 22, 2007 16:04

Re: Discuss about computation time of FLUENT
 
Hi

May I ask: 1)how many cells there are in the domain, 2)what is the time step size and number of time steps, 3)what computaional power do you have?

Regards

AAA

Kelly August 22, 2007 16:26

Re: Discuss about computation time of FLUENT
 
AAA,thanks for your attention! The problem have 97293 cells; time step size=0.001s (I tried bigger ones but failed to get convergence),number of time steps=5000.I am not sure about the computational setting of this lab computer but it should have at least 512mb RAM.

Kelly

AAA August 23, 2007 09:17

Re: Discuss about computation time of FLUENT
 
Hi

The total duration of your run is 5 seconds with 1000 steps per second. This is comutationally demanding. Have you obtained any satisfactory results yet? If yes, have it run over night. Otherwise, have less number of time steps, say 1000 (1 second scene) to see what happens. PISO gives me trouble sometimes too. Try using another scheme (SIMPLIC) and see if the calculations speed up. As for pressure discretization, I know that body force weighted is to be used here, but try using 2nd order. finally, consider lowering under-relaxation factors.

Good luck

AAA

Kelly August 23, 2007 12:32

Re: Discuss about computation time of FLUENT
 
Hi AAA,

I could get satisfactory results even though it is computationally demanding. Still I am thinking how to get over this long lasting computaion time. I tried as you suggested but failed: firstly there is a warning "turbulent viscosity limited to viscosity ratio****" and then ended with divergence after several iterations. I am trying to change UR factors to aviod this now. Hopefully it will work.

Thanks for your suggestion.

Kelly

AAA August 24, 2007 20:43

Re: Discuss about computation time of FLUENT
 
Hi Kelly

Let's take care of the warning "turbulent viscosity limited to viscosity ratio****" which is not physical. This problem is mainly due to one of the following:

1)Poor mesh quality(i.e.,skewness > 0.85 for Quad/Hex, or skewness > 0.9 for Tri/Tetra elements). {what values do you have?}

2)Use of improper turbulent boudary conditions.

3)Not supplying good initial values for turbulent quantities.

I hope this helps.

Regards

AAA



All times are GMT -4. The time now is 02:58.