CFD Online Discussion Forums

CFD Online Discussion Forums (http://www.cfd-online.com/Forums/)
-   Main CFD Forum (http://www.cfd-online.com/Forums/main/)
-   -   CFD Study of a Diffuser (http://www.cfd-online.com/Forums/main/70778-cfd-study-diffuser.html)

lemat1 December 5, 2009 13:47

CFD Study of a Diffuser
 
Hi all, i am currently conducting a comparative study of underbody performance if a bluff body with a diffuser. The model is fixed with a moving ground at 40m/s. i am assessing drag and downforce through changes in ramp angle, and yaw condition using a constant ride height. I am pretty sure that vortices should be shed from the body, and that these will affect the drag on the body yet i am unable to detect any in my results. I am using an implicit unsteady flow condition, K-E turb model and RANS solver. My Re number is 1.4x10^7, giving an estimated (from research) strouhal number of 0.28


my questions are;

1 - should i worry about vortices? im am comparing results against themselves so if i ignore this on all is that valid?(i.e. a constant effect) - i do not want exact downforce figures just which is most efficient (lift/drag) - could even switch to steady state?.-----i suspect this is wrong

2 - my time step is set 0.002 seconds which corresponds to 50 steps per vortex shed, at f=9.7Hz i.e. delta t = 1/(50*f) - is this enough/too fine?

3 - how many internal iterations per time step? in a steady state i see convergence at about 250 steps, so is this a good starting point?

4 - the model is 1.1m*0.7 m, so consequently the tunnel huge to maintain 5% blockage, is it possible i am not seeing vortices due to mesh density? im at nearly 1 million cells - using the surface wrapper and polyhedral mesh. refinement around the bluff body! - is this ok?

Thanks all

Regards

Mat

CapSizer December 5, 2009 17:53

1 million cells for a model like that sounds quite low. Also, you are flirting with flow separation in an adverse pressure gradient. This is not the natural habitat of the k-epsilon turbulence model.

lemat1 December 5, 2009 19:47

Quote:

Originally Posted by CapSizer (Post 238840)
1 million cells for a model like that sounds quite low. Also, you are flirting with flow separation in an adverse pressure gradient. This is not the natural habitat of the k-epsilon turbulence model.

which would you recommend? k-omega?,,,wrt to cell count i dont have access to a cluster, the uni one is broken, so i have to keep the cell count down!

CapSizer December 6, 2009 02:27

Well, you are somewhat resource-limited then, so probably a good idea to check just how far away from mesh insensitivity you are. As for turbulence models, first of all get hold of Wilcox's book! SST is probably not a bad place to start, but you need to take things step by step here. Going for a low Re-number method is probably advisable.

lemat1 December 6, 2009 07:17

ok, ive started by switching my turb model to k omega, and re solving the steady state cases, then look to incorporate the unsteady solver. any thoughts on the time step i proposed?

CapSizer December 6, 2009 14:24

It doesn't sound unrealistic, but it also helps to a) check your maximum and average CFL numbers, and b) to see how much your solution changes as you refine the time step.

lemat1 December 6, 2009 15:04

Quote:

Originally Posted by CapSizer (Post 238873)
It doesn't sound unrealistic, but it also helps to a) check your maximum and average CFL numbers, and b) to see how much your solution changes as you refine the time step.

Im not familiar with a CFL number, could you explain?

wrt to time step will the solution always be time step dependent since you will always pick up something extra(some tiny negligible disturbance?) with the smaller and small time step?

Thanks

Mat

CapSizer December 6, 2009 15:10

You need to do some reading about CFL numbers before you go any further. As for the second question, do it and see. It is a good way to learn.

lemat1 December 7, 2009 06:55

Quote:

Originally Posted by CapSizer (Post 238876)
You need to do some reading about CFL numbers before you go any further. As for the second question, do it and see. It is a good way to learn.

ok!, at least tell me what CFL stands for so i can look it up!

wrt to my original questions - for a comparison case can i just use steady state analysis? - no actual force numbers will be calculated!

Thanks

Mat

CapSizer December 7, 2009 08:45

1. Also look under Courant number. 2. Maybe, depends on whether or not there is flow separation.

Generally though, your university lecturer is being paid to answer these questions ....


All times are GMT -4. The time now is 13:11.