[selig5576]

I am currently working on a incompressible solver and I am at the stage where I want to implement a turbulence model. I am currently using a Chorin Projection method with Rhie-Chow interpolation. My solver (so far) is in 2D, so I was curious if there are any benchmarks I can test my 2D solver on, given I have a turbulence model.

In the case where I would implement the k-epsilon model and test it on a lid driven cavity how would I define the fluctuating component of the velocity, u'.

Since k is defined as

then would I need to initialize that and then solve the transport equations?

for (int i=1; i <= nx; i++)
{
  for (int j=1; j <= ny; j++)
 {
   k[i][j] = 0.5*(u_avg[i][j] + v_avg[i][j]);
 }
}

then I integrate k and e..is this the correct approach?

[valgrinda]

Since you use a RANS turbulence model, u' and v' are never directly available. They are included in the eddy-viscosity term in the Navier-Stokes equations. Typically, one would initialize k and epsilone based on empirical equations based on the flow intensities and the characteristic length scale L, e.g. the diameter of a pipe. See Versteeg's CFD book version 1, p. 72.


Sent from my iPhone using CFD Online Forum mobile app