next up previous contents
Next: Implementation of the positive Up: Numerical aspects of two-equation Previous: Space discretization of the

2D implementation of low-Reynolds-number modeling

 

When implementing the Lam-Bremhorst model, a serious aspect to consider is the near-wall behaviour of the damping functions. For example, consider the damping function

 

Approaching the wall, desired asymptotic behaviour depends upon accurate values of this function. Using Taylor series expansions, this damping function can be written as

Furthermore, the limiting form of should be asymptotically consistent with the near-wall behaviour of time-averaged properties. For example, it is well known that the asymptotic variation of turbulent energy and dissipation rate near the wall are

Hence, we have

Substituting in (6.42) gives

Because the order of the leading term of is larger than that of , it sometimes happen, numerically, that is relatively large even when . Therefore, if and , should be set to its limit value. In order to avoid very small (or negative) values of k in regions in which becomes very small, the implementation of the damping function is slightly modified: should not be smaller than 0.01.

The calculation of the normal distance from a grid point in the inner region to the wall, i.e. Y, will now be described. We consider the following situation as depicted in Figure 6.1. The distance of a cell-center point P from a boundary

  
Figure 6.1: Calculation of normal distance between node P and boundary point B

surface can be found as the scalar product of a vector connecting a boundary point B and P and the unit normal vector :

The index numbers of the cell containing P and the near-wall volume including B are identical. The co-ordinates of B and P are obtained from the co-ordinates of cell vertices by linear interpolations, hence

The unit normal vector is computed as follows:

In the presence of several walls - the usual case - is, in some sense, a weighted average of distances to all points on the solid boundary seen by the point at which Y is to be computed. There is no consensus, however, on how Y should be computed in general. A simple approach, adopted in complicated geometrical domains, is to take Y as the distance to the nearest wall.

In the case of k- model, standard boundary conditions must be employed at a solid wall, i.e. for the momentum equations noslip conditions are imposed on the boundary, whereas for turbulent energy k a homogeneous Dirichlet condition holds. However, in order to avoid non-positive values of k, may be taken as boundary condition on the wall. In order to impose condition (6.22), the equation for is solved up to the second grid point closest to the wall, applying the condition

at the first grid point P above the surface. Here, is given by (6.23), is the nondimensional distance from the wall and is the friction velocity with the wall shear stress.


next up previous contents
Next: Implementation of the positive Up: Numerical aspects of two-equation Previous: Space discretization of the

Tatiana Tijanova
Wed Mar 26 10:36:42 MET 1997