abt wall functions and k-epsilon
 

Hi

I am trying to write a code for turbulent flows. I used the standard wall functions and solved a lot of problems, but I am finding it hard to solve the cases where there is abrupt contraction or expansion .e.g. in backward facing step or forward facing step problems, I am finding difficulty of convergence and with only very small underrelaxation factors for k & epsilon, I could get convergence.

Does it mean that I have to go to two layer or three layer approaches (e.g. Amano / chiang) for such cases rather than the standard wall functions.

Thanking you !

Re: abt wall functions and k-epsilon
 

The k-e model plus wall functions should converge relatively easily for this basic case. If you use simple equilibrium wall functions, the values of k and e are fixed at the near-wall point, and so that should not be a problem if you are computing the friction velocity correctly from the log law. The wall friction force should be linearised in the momentum equation to facilitate convergence. If you need heavy relaxation on k and e, it may suggest that you are using inappropriate linearisation practices for the source terms in these equations. Does your code converge if you employ a uniform turbulent viscosity plus wall functions?

Re: abt wall functions and k-epsilon
 

hi, Thanks for your kind cooperation.

Now the code is behaving much better.

I am using

if y+ > 11.63,

g = (tau_wall ^2)/(density*kappa*u_tau*y_p)

and

epsilon=u_tau^3/(kappa*y_p)

and

if y+ < 11.63 then,

g-density * epsilon = 0 ( source term of tke equation)

epsilon=2 * nu * tke / y_p ^2 (boundary condition for epsilon equation)!

is the above formulation ok ?

I have to reduce under-relaxation factors if i use a finer grid now but for a coarser grid now i am able to go take under relax fact of 0.2 which was only .05 earlier !

Re: abt wall functions and k-epsilon
 

For y+ > 11.63, k=u_tau^2/0.3 with simple equilibrium wall functions. In my experience no convergence problems accrue if the same relations are employed when y+ < 11.63, wherein u_tau is computed for laminar flow. For me a linear relaxation factor of 0.2 or smaller seems rather low for securing convergence. A value of 0.4 or 0.5 ought to be sustainable.

Procedure here:
 

Please, take a look at section 7.3.4, page 345 in the book of Poinsot and Veynante.

Re: Procedure here:
 

Hi RMG, Can you tell me some details ? I think we don't have any book of these authors

Poinsot and Veynante

in our library.

Thanking you.

You should buy it, it is good
 

You should buy it, it is good