Scaling of the k and epsilon variables in the turbulent channel

we are trying to develop a fluid flow PINN model for a custom channel flow geometry. We are basing our case on the turbulent channel example. We are confused by the approach to calculate the k and epsilon and their scaled variables k_star and ep_star.
k: Min(log(1 + exp(Symbol(“k_star”))) + 1e-4, 20)
Epsilon: Min(log(1 + exp(Symbol(“ep_star”))) + 1e-4, 180)
Does anyone know why they are used and if there is any general guideline that describe these scalings?