Initial conditions k and epsilon
 

Hello all!

I have been simulating the effects of the Atmospheric Boundary Layer (ABL) on a 3D high rise building with different Steady RANS turbulence models such as Standard kE, kOmega, kOmegaSST, realizable kE and RNG kE.

I have taken the Cd on the building and the Cp at 2/3H as 2 of my reference values in order to establish a comparation between my simulations with those 5 models and previous similar studies. However, the Cp obtained in all of my calculations is quite larger than those obtained by the mentioned previous studies. Apart from the mesh, which i have already checked and it is not the problem at all, i think that maybe I have used wrong values as the initial conditions of k, omega or epsilon.

My current values are:

turbulent kE: 0.7
turbulent epsilon: 2.0
turbulent omega: 3.0

My inlet U profile is based on a logaritmic law, and reachs 14.5 m/s at the roof. Uy and Uz are zero in the initial conditions.

By the way, I have read about a estimation based on the turbulent intensity. However, since I only use steady RANS models, I cannot estimate that intensity.


Are my initial values good? How should I change them if not? Do you know any way of estimate them?


Thank you all

Any advise will be absolutely helpful.

if you are using the log-law for velocity, then k and epsilon have to follow the formulations of the loglaw. Note that the loglaw is the balance between production of k and epsilon. in other words, uv du/dy = epsilon . Epsilon here has a y dependence which means you cannot assign it a fixed value. epsilon in the log-law ( local equilibrium) equals (friction velocity)^3/(kappa*y).
and k in the log law is (friction-velocity )^2/(sqrt C_mu).

Also, i would not use the standard k-epsilon for these type of flows because of the so-called stagnation point anomaly ( spurious production of k at stagnation region)

davidwilcox,

I will try it in my next simulations and, as soon as get any significant conclusion about it, I will let you know on this thread.

However, in some studies about simulating properly the ABL with CFD, the researchers have used uniform values for k and epsilon. Indeed, there is an inlet boundary condition in openFoam which, based on P.J Richards and R.P Hoxey (1993) logaritmic law, generates a non-uniform inlet U profile while using uniform profiles for both k and epsilon. And that boundary condition is the only one that exist on this programme to create a log law inlet profile, so I thought that it would be fairly correct.

That's precisely what made wonder about all this issue, I found different information from different researchs and, consequently I had no idea what to use. However, I agree with you about the local equilibrium, and I will create two non-uniform vertical profiles for k and epsilon and let's see how does it works.

As an aside, I have already read about the issue related to standard k-epsilon you have mentioned. That is one of the tasks of my project, how and why other turbulence models work more properly than standard kE in that especific point.

Thank you

Pelayo