Turbulence parameters and convergence
 

I have observed that turbulent parameters played a major role in the convergence of a simulation. Initially when I simulated with the default values of intensity ans viscosity ratio (5 and 10) the convergence is bad and when I switched over to use intensity and hydraulic diameter (= 4A/P) there is an improvement in the convergence.

Could some one please provide a reasoning to this behavior

Yes, turbulence makes a big difference to fluid flow. I am glad you figured that out :)

In simple 2-eqn turbulence models the turbulence is modelled to increase the local viscosity (and diffusivity for other equations). This means turbulence is assumed to slow a flow down and make it more viscous - this is in effect a reduction in Reynolds number and convergence is easier in low Reynolds number flows compared to high. (Only up to a point, CFX has great troubles converging for very small Re numbers <<1).

Hi Glenn, Thank you very much for the clarification so when a convergence difficulty is observed with a simulation how is it possible to sense at as an effect of improper or unphysical turbulence settings.

And if a good convergence is acheived by say reducing the viscosity ratio and increasing the intensity of turbulence then is it implied that such turbulence settings represent the real physics

The decision as to whether to choose a laminar or turbulent model only depends on the Reynolds number and whether the flow is turbulent or not. But the choice of which turbulence model to use for a turbulent flow is complex - all I will say is that the SST turbulence model is the best all-round model and use that one unless you have a very good reason to choose something else.

I did not say that! That is not correct.

? I have no idea how one implies the other. I think you need to read some background textbooks on turbulence modelling to understand what it is all about. "Turbulence modelling for CFD" by wilcox is good, and most introductory CFD textbooks include a section on turbulence modelling.