Hello everyone,

I am using Fluent to try to simulate the evolution of a turbulent boundary layer on a flat plate of 30 cm of length with a freestream velocity of 5 m/s (2D-Steady state).

The mesh is such as : Dx=1mm and Dy=y+ at the wall

I've made a theoritical turbulent boundary profile for the velocity inlet condition. This profil follows the regular laws, as :

--> U+=y+ [close to the wall]

--> U+=1/0.41*ln(y+)+5 [At the log region]

--> The deficient law of Coles farther from the wall

I ran some simulations using k-eps model with a wall function until the residuals are small enough (< 10e-8). When I compare the numerical results at several x locations with the log law, it doesn't match at all.

Has someone have done something similar ?

Thank you !

Seb

I've done a bit with the log-law.

Remember, Coles' function does not satisfy boundary conditions at the outer edge of the BL. Does that affect your CFD model?

Also, Coles' log-law constants are very dodgy. He obtained the values for kappa and B_0 using massaged velocity profile data. Sadly, his values are often still treated as universal constants by noobs.

Actually I thought that wall functions in the RANS models were made to make the flow to fit the sub-layer law and log-law with the "universal" constants.

At the entrance, my theoritical profile is such

U+ = 1/0.41 * ln(y+) + 5 [for 10 < y+ < 60]

When I get profiles at some location after the simulations, they all fit in a profile such as :

U+ = 1/0.41 * ln(y+) + 0.3 [for 10 < y+ < 30]

I was wondering if this was coming from the fact that I am simulating a Low Reynolds boundary layer (Rtheta ~ 600). I made a simulation with a higher velocity (and so a different boundary layer at the entrance) at U=100m/s instead of 5m/s and I found the same results...

After trying different mesh, different profiles at the velocity inlet and different models I found that the best model to simulate the evolution of a turbulent boundary layer is K-EPS --> Realizable --> Enhanced wall treatment.

Using this, the profile of velocity fits the log law fairly well;

Besides, the first cell doesn't have to be too small (> y+)