hi guys i am working in a automobile firm. the value of coefficient of drag that i get through CFX always comes more than the experimental drag values calculated in wind tunnel testing. we have the experimental data. can't figure out why it always comes more. i am using SST turbulence model. inlet parameters and the geometry of the vehicle is exactly the same as in wind tunnel. the cad geometry of the vehicle is in 1:1 scale.

Where does one start...there are hundreds of reasons why this could be the case. It's probably worth looking at a few of the old posts on this forum for some tips. What mesh have you used, tet or hex? Have you set the correct near wall node location for the SST model (Y+<1.0). Have you got enough resolution in your mesh to capture the pressure drag correctly. Is your test data really as good as you think it is? Have you got the same Reynolds number in your CFD model? Are you solving using steady state solver when really it is a transient problem. Have you made sure all the physical properties have been set correctly? These are the easy things to check, there are many more things which you might need to look at.

Hi,

As James says, there are a lot of things which need to be right before you get an accurate simulation. There are some useful papers on the CFX-Community website where they model external flows on vehicles. I recommend you have a look at them to see how other people have done it.

Regards, Glenn Horrocks

I worked in the same area so I can give this suggestions:

-Check the y+. -Hybrid (Tetra+Prisms) mesh is not so good, so try to find the Mesh Indipendence. -Increase the Domain dimensions.

Stex

thanks guys for the suggestions. i will be looking at it. but i need suggestions in one more point? is there any mathematical reason that while doing analysis through CFD we will always get more numerical value of drag as compared to experimental one. when i say mathematical i mean the way CFD works out. is it due to discretization techniques or something else.

Hi,

Some turbulence models have known deficiencies, such as k-epsilon in the separated region behind bluff bodies. It over-estimates the turbulence generated, this can change the separation points and consequently change the predicted drag. Remember that CFD is a numerical solution to a bunch of mathematical models which roughly describe fluid flow. A perfectly converged solution is no good if the underlying mathematical model is not appropriate.

Regards, Glenn Horrocks