The simplest view of the behaviour of a turbulent fluid is that it is just like that of a laminar one but with an increased viscosity, which however may vary from place to place.
This concept is usually attributed to Boussinesq (1877) and is called the "EFFECTIVE-VISCOSITY HYPOTHESIS".
One important implication is that turbulent shear stresses are proportional to velocity gradients.
The hypothesis has been known for many years not to be valid in all circumstances; but it is often so close to the truth that it is very widely used.
So central is it to turbulence modelling that models can be usefully classified according to their use or avoidance of it.
Turbulence models in PHOENICS can thus be classified as belonging to one or other of three groups, namely:-
Group-1 models can then be further classified in accordance with the methods which they adopt for CALCULATING THE MAGNITUDE of the effective viscosity.
Further distinction between models can be made by reference to their handling (or non-handling) of:
Sub-group 1.1, in which no differential equations are used
The practices for computing the effective viscosity (EV for short) are:
|prescribed EV||EV is given a uniform value|
|LVEL||EV is computed from the velocity, the laminar viscosity and the distance from nearby walls|
|Prandtl mixing-length||EV is computed from the velocity gradient length and a prescribed length scale|
|Van-Driest||as for Prandtl mixing length, but with low-Reynolds-number modification|
|Prandtl energy||EV is prescribed-length * SQRT(KE) where KE is energy of turbulence computed from a differential transport equation|
|TWO-LAYER KE-EP||as for KE-EP (see below), except that only the KE equation is solved near the wall, where the length scale is treated as known|
|k-epsilon (KE-EP)||EV is proportional to KE**2/EP, where KE and EP (dissipation rate of KE) are computed from differential transport equations|
|CHEN-KIM KE-EP||as for KE-EP, but with a "dual-time-scale concept" making the formulae depend upon the energy-production rate P|
|RNG-derived KE-EP||as for KE-EP, but with a "re-normalization- group concept" making the formulae depend upon the energy-production rate P|
|LAM-BREMHORST||as for KE-EP, but with low-Reynolds number extension requiring knowledge of the distance from the nearest wall.|
|Saffman-Spalding KE-VO||EV is proportional to KE/W**0.5, where KE and W (RMS vorticity fluctuations) are computed from differential transport equations|
|Kolmogorov-Wilcox KE-OMEGA||EV is proportional to KE/OMEGA, where KE and OMEGA ("turbulence frequency") are computed from differential transport equations|
|TWO-SCALE KE-EP||EV is computed in a similar manner to that of KE-EP model; but there are two turbulence- energy variables, KP and KT, and two dissipation-rate variables, EP and ET; so it is necessary to solve four differential transport equations.|
|REYNOLDS-stress||EV is not used. Instead, the shear stresses are themselves the dependent variables of differential transport equations, usually six in number.|
|Smagorinsky||EV is used only to resolve the small-scale subgrid-scale motion, the main transfers of momentum being computed by performing three-dimensional time-dependent solutions of the Navier-Stokes equations with the finest affordable space and time sub-divisions.|
|Two-fluid||EV is either not used at all, or is deduced from the local velocity differences between the two intermingling fluids which are used to describe the turbulent fluid mixture.|
|Multi-fluid||as for TWO-FLUID, except that, there being many fluids present, EV can be derived from their various velocities in a wider variety of ways.|