Van Driest Model
From ThermalFluidsPedia
External Turbulent Flow/Heat Transfer 
While the above twolayer model works well in the viscous sublayer and the fully turbulent region, a discrepancy between the twolayer model and experimental results in the transition region (near y^{ + } = 10.8) is apparent. The reason for this discrepancy is that the velocities in these two regions were assumed to be dominated by different mechanisms: molecular activities in the viscous sublayer and eddy level activities in the fully turbulent region. The contribution of the eddy diffusion in the viscous sublayer was completely neglected. In fact, relatively large elements of fluid in the viscous sublayer can lift off the wall and immediately be replaced by the other fluid from the fully turbulent region. This phenomenon is referred to as “bursting” (Kays, et al., 2005). Therefore, the eddy diffusivity in the viscous sublayer is not really zero and the only point that it becomes zero is at the wall where y^{ + } = 0.
To account for the effect of eddy diffusivity in the viscous sublayer, the Van Driest hypothesis states that the eddy diffusivity decreases as one nears the wall (Van Driest, 1956), instead of two regions where it is “on” or “off.” Therefore, we can use the Prandtl’s mixing length model in the entire turbulent boundary layer by introducing a damping function, i.e.,

where A is an effective sublayer thickness that must be determined empirically. Equation (4.431) can also be expressed in the following dimensionless form:

where A^{ + } = Au_{τ} / ν is the dimensionless sublayer thickness. The value of A+ must be determined in such a way that the calculated value of u+ outside of the sublayer matches the result obtained by the law of the wall, eq. (4.427). It is found that A^{ + } = 25 can produce the best match. Figure 4.36 shows comparison of the results obtained by the Van Driest model, the two layer model, and experiments. It can be seen that the results produced by the Van Driest model agreed very well with the experimental results in the entire turbulent boundary layer. It should be pointed out that A^{ + } = 25 can only produce good results for the cases without blowing or suction and without pressure gradient. When adverse pressure gradient () and wall velocity are present, the following correlation can be used:

where

are respectively the blowing velocity and the pressure gradient in the wall coordinate. The empirical constants in eq. (4.433) are a = 7.1, b = 4.25, and c = 10.0. If p^{ + } > 0, b = 2.9 and c = 0.