Hi all,

The limiting of turbulent viscosity ratio is a thouroughly discussed issue in this forum. To summarize the discussion: 1. It is a problem that is observed at the beginning of the simulation without much effect on the results if the limitation vanishes as the simulation is stopped. 2. There seems to have some connection between the spatio-temporal discretization and the limitation of the turbulent viscosity ratio.

I have the following questions:

1. What is the exact reason for this limitation? Does it represents problems with the definition of turbulent viscosity ratio or its implementation in Fluent . Or is there some other reasons?

2. How far is the argument that the limitation of the ratio is innocent holds? Remember that the physically expected ratio remains in the order of two, while the limitation occurs at an order of 6.

Limiting the turbulent viscosity is a way to keep a "difficult case" from crashing due to convergence problems causing unphysically high turbulent viscosities. There is no guarantee that these unphysically high turbulent viscosities are not still present although you eventually manage to get under the "viscosity ratio limit".

Actually the fact that you reach the turbulent viscosity ratio limit when you run a simulation is a sign that you should be extra careful about checking your results and making sure that you don't have any critical regions where the turbulence fields have "blown up" and created honey (unphysically viscous regions) - this is a very common problem with for example the standard k-epsilon model.

My experience is that the use of a too strict turbulent viscosity ratio limit at the start of a simulation often can create problems later on - once the solution reaches this limit in a few cells this problem has a tendency to spread to nearby regions and eventually you end up with a turbulence field which "leans on the limit" in large parts of the domain. I would advice you to set the limit so large that it just only kills the critical initial transients and then quite quickly stops having any effect. If you don't manage to do this you should probably spend some time trying to improve your mesh in the regions where the problem occurs, change under-relaxation parameters for the turbulent fields or perhaps try to ramp up your inlet/outlet conditions to slowly obtain the conditions you want. YMMW. ;-)

Thank you Mr. Jonas Larson for your illuminating reflection over limiting eddy-viscosity ratio.

Let me point out some further observations I have with my present simulation. The problem is of a mild buoyant jet into a large quiesent waterbody.

1. As you have mentioned, initial transients has blown up the turbulent viscosity ratio, beyond the Fluent limits. I tried to refine the grid in the problem regions, only to realize it is not a affordable solution.

2. After a few more time steps, the peaks died out, but the domain still consists of "honey regions"( by the way it is a nice expression to describe the situation-"turbulent honey regions"! good..good!). The ratio assumes values of about 8000. You mention that this issue is bye-product of approximations in the standard k-epsilon model. Does it accompany the realisable version of k-epsilon model, which I have implemented for the current problem?

3. I also observe regions with extremely low values of the ratio, less than one. I expected the minimum value should be 1, because I imgined the ratio is between the effective viscosity calculated in the turbulent models and the molecular viscosity. Is there only the eddy viscosity at the numerator of the ratio? How does the turbulent models filter the eddy viscosity from the molecular one?

Greetings Varghese