CFD Online Discussion Forums (http://www.cfd-online.com/Forums/)
-   ANSYS Meshing & Geometry (http://www.cfd-online.com/Forums/ansys-meshing/)
-   -   [ANSYS Meshing] Lift & Drag Meshing Techniques (http://www.cfd-online.com/Forums/ansys-meshing/100319-lift-drag-meshing-techniques.html)

 tylerplowright April 23, 2012 23:59

Lift & Drag Meshing Techniques

Hey,
Having a few problems trying to model a wind tunnel experiment we were required to do. In the experiment we put a small cube in a wind tunnel to see what the lift and drag forces were on it. We have results of around 0.3N of force in drag and no lift (a face was square to the wind). Only problem is when we model this in Ansys CFX we can't get anywhere near that result. We are modelling it as if it were in the tunnel so the distance between the object and the walls/outlet aren't ideal. Inlet speed is 20m/s, Outlet average static pressure is 0Pa (relative to atmospheric).

In nearly of the settings we tried, we haven't been able to get the drag force down below about 0.7N. If it was around about 0.4, I would be happy and would actually potentially be able to get some kind of meaningful result out of it but being over 100% out is a bit too much. We did only eyeball the angle so I'm not looking to model it exactly, but something sort of close would be nice.

As such are there any techniques with meshing that someone would recommend for best capturing the lift and drag of an object? Inflation? Sphere of Influence? Any tips or recommendations would be great.
Thanks
Tyler

 Far April 24, 2012 04:36

First of all, this problem is more related to solver than meshing.

First I discuss the meshing part:

For drag prediction you need to ensure three things:
1. Y+ should be less than 1.
2. Boundary layer should be resolved properly with at least 30-40 nodes
3. Wake region meshing requirements.

 Far April 24, 2012 07:14

From solver point of, you should go for LES or atleast DES/DDES. (may SAS do the trick, I am not sure)

 Ananthakrishnan April 24, 2012 07:33

Hi,
i have a few points to highlight..

1) drag and lift convergence depends upon your the set of equations you have selected..
for eg..only euler equations with or without energy equations, or higher order equations involving viscosity, turbulence, reynolds stresses, etc..These affect your final value as you move closer to reality.
No matter what, the numerical simulations have always resulted in a slightly higher value compared to experiments. This is because it is very very difficult to simulate real life conditions with approximations.
2) Apart from what far has told some other approximation for the solver are
a)using ideal gas instead of constant density
b)check your mach number..you might also want to use compressible flow instead of incompressible..
c) Dont use laminar flow models.
d) try fluent as well..

 tylerplowright April 25, 2012 04:14

Quote:
 Originally Posted by Far (Post 356617) First of all, this problem is more related to solver than meshing. First I discuss the meshing part: For drag prediction you need to ensure three things: 1. Y+ should be less than 1. 2. Boundary layer should be resolved properly with at least 30-40 nodes 3. Wake region meshing requirements.
Perfect, that has helped a lot. Made a few more things make sense too. What do you mean by wake region meshing however? I'm running a fine mesh on the entire object (seeing as it is quite small and the wind tunnel isn't large either), with all the defaults and a small amount of inflation.

Quote:
 Originally Posted by Far (Post 356645) From solver point of, you should go for LES or atleast DES/DDES. (may SAS do the trick, I am not sure)
How do you change the settings to LES, DES/DDES? I can find things like K-epsilon, SST etc but there isn't anything that looks like LES, DES/DDES.

Quote:
 Originally Posted by Ananthakrishnan (Post 356650) Hi, i have a few points to highlight.. 1) drag and lift convergence depends upon your the set of equations you have selected.. for eg..only euler equations with or without energy equations, or higher order equations involving viscosity, turbulence, reynolds stresses, etc..These affect your final value as you move closer to reality. No matter what, the numerical simulations have always resulted in a slightly higher value compared to experiments. This is because it is very very difficult to simulate real life conditions with approximations. 2) Apart from what far has told some other approximation for the solver are a)using ideal gas instead of constant density b)check your mach number..you might also want to use compressible flow instead of incompressible.. c) Dont use laminar flow models. d) try fluent as well..
I want to try and get it as close as I can without going totally overboard. Within 25 to 30% would be fine however due to the fact that it wasn't an exact setup (only setup according to eye). With that in mind however and your advice, I'm running ideal gas and using the SST model. I am going to play around with the other models/solvers to see if any help as well as try fluent. What do you mean by mach number? And is compressible flow really necessary seeing as it is at very low speeds (20m/s)?

Also, when I have inflation turned on I can't get the momentum and mass equations to converge. Specifically the RMS V Moment and the RMS W Moment, which both sit around 3e-03. The Mass and the RMS U moment both eventually go to below 1e-04. Is this likely to be causing odd results and if so any ideas on how to get around it?

Thanks
Tyler

 Ananthakrishnan April 27, 2012 14:43

1.Mach number--Yeah the speed is very less..I just wanted you to make sure that the mach number on the surface of the airfoil does not cross 0.6.
2.Convergence--From a first glance, i think you have suppressed some kind of variable (approximation) through the selection of model which might be getting generated during actual simulation.. (Huge temperature variations might also cause this problem. Try switching off the energy equation to confirm this hypothesis)..It must be affecting the final results but i am not expert enough to guess the weight age of effect..
All the best!!!

 tylerp4cfd May 2, 2012 23:53

Quote:
 Originally Posted by Far (Post 356617) 2. Boundary layer should be resolved properly with at least 30-40 nodes .
Once a simulation is run how can I find out how many nodes there are in the boundary layer?

 All times are GMT -4. The time now is 20:41.