Comments on: A new low-Reynolds-number k-e model for turbulent flow over smooth and rough surfaces
Submitted by: Uriel Goldberg 
Year of issue: 1996
Paper starts on page: 255
Date submitted: Fri Aug 16  5:24:13 1996

While proposing a turbulence model for rough surfaces is undoubtedly a noble cause, there are three items in this paper which I find necessary to criticize. In order of severity, these are: (1) The formulation of f_mu: it is not f_mu which is responsible to cause the eddy-viscosity to vanish at a smooth solid surface; k^2 is taking care of of that. In fact, it is doing so overenthusiastically, since it forces a y^4 asymptotic behavior for nu_t as y --> 0. Since the correct behavior is y^3, it follows that f_mu ~ 1/y is the correct asymptotic behavior, certainly not y^2! (Eqs. 7 & 10.) (2) For over 10 years now DNS studies by people like Kim, Moin, Spalart, Mansour, and others, have consistently shown that epsilon attains its maximum value right at the wall. Surely, this cannot be ignored and must be pointed out in the discussion about the epsilon profile predicted by the model. It is not enough anymore to show the Laufer (1954) and Coles (1978) data and ignore the DNS results. (3) If one must use explicit wall distance in the near-wall formulation of a turbulence model, at least an effort must be made to avoid using y+ because of its singular behavior at separation and reattachment locations (since tau_w vanishes there). This drives f_mu to zero, creating laminar spots in the middle of a turbulent flow! Instead, one could use the Reynolds numbers k^(1/2) y/nu or (nu epsilon)^(1/4) y/nu to replace y+. This would render the model immune to separation and reattachment.
by Sisi Chen