Philosophies and fallacies in turbulence modeling

We present a set of positions, likely to be controversial, on turbulence modeling for the Reynolds-Averaged Navier Stokes (RANS) equations. The paper has three themes. First is what we call the fundamental paradox of turbulence modeling, between the local character of the Partial Differential Equations strongly favored by CFD methods and the nonlocal physical nature of turbulence. Second, we oppose two philosophies. The Systematic philosophy attempts to model the exact transport equations for the Reynolds stresses or possibly higher moments term by term, gradually relegating the Closure Problem to higher moments and invoking the Principle of Receding Influence (although rarely formulating it). In contrast, the Openly Empirical philosophy produces models which satisfy strict constraints such as Galilean invariance, but lack an explicit connection with terms in the exact turbulence equations. The prime example is the eddy-viscosity assumption. Third, we explain a series of what we perceive as fallacies, many of them widely held and by senior observers, in turbulence knowledge, leading to turbulence models. We divide them into hard fallacies for which a short mathematical argument demonstrates that a particular statement is wrong or meaningless, and soft fallacies for which approximate physical arguments can be opposed, but we contend that a clear debate is overdue and wishful thinking has been involved. Some fallacies appear to be intermediate. An example in the hard class is the supposed isotropy of the diagonal Reynolds stresses. Examples in the soft class are the need to match the decay rate of isotropic turbulence, and the value of realizability in a model. Our hope is to help the direct effort in this field away from simplistic and hopeless lines of work, and to foster debates. (C) 2015 Elsevier Ltd. All rights reserved.