This is not clear in the doc of fluent because you cant use standard Reynolds averaging for variable density flows without introducing new correlation terms and thus you need new closure models.

On the other hand, nowhere in the doc, the Favre averaging method is exposed, so we should think that this method is not used so ...

"How does fluent takes into account the variable density flows ?"

If we assume that the standrard Reynolds formulation is used, then results may be seen as "favre averaged". This means mean(U)=mean(rho U)/mean(rho) witch is in fact tilde(U).

The problem is that if mean(U)=tilde(U) in the core of jet for example, it is completely different in the shar layers. Moreover, the two component velocity correlations are totally false because as favre averaged they include the density fluctuations !!!!

could someone give me some informations on the way FLuent simulate variable density flows. In my opinion, variable density is only taken into account through the turbulent viscosity witch involves the dynamic viscosity (nu=mu/hro). I hope i'm wrong, but if someone know, i'd be glad to have some more info.

Favre averaging only becomes important in the hyper-sonic regime. In subsonic and mildly supersonic flow you can safely neglect effects like pressure diffusion, pressure dilatation, pressure work and turbulent transport/molecular diffusion of turbulent energy.

A simple explantion as to why this can be neglected is that for most flows the turbulent fluctuations are not strong enough to affect the pressure & density fields significantly.

Besides, Fluent does not have any turbulence models where these effects are modeled - they are just neglected - and this is okay for virtually all subsonic and supersonic flows.

I went through most of these derivations quite thouroughly in the introductory part of my PhD thesis - you can download it from here (in gziped postscript format).

in addition to the numerical data i have some experimental acquisition of velocity and density (in an air-co2 jet). When you use favre or reynolds representation you dont have the same results, especially in the mixing layers where density fluctuations are high.

So my question would be : does the solution given by fluent can be considered as a fravre averaged result or as a simple Reyonlds averaged one ?

[spy_hzdr]

Dear Jimmy, did you get any solid answer on this question during the past 15 years?

[habo]

Same issue here, in 2021.

Quote from Anany, 2010, p. 15, Numerical Modelling of Combustion Processes at Elevated Pressures:

On the other hand, chemically reactive flows are characterised by strong heat release accompanying the chemical reactions and, hence, by temperature-coupled density variations. Applying the time-averaging procedure to reacting flows introduces extra source terms to the transport equations due to these density variations. In order to avoid the density-related source terms, mass-averaging (Favre-averaging) techniques instead of time-averaging ones are implemented to overcome such inconvenience.

I am working with chemical reacting flows in ANSYS Fluent therefore I assume that Favre-average NS equations are used per default. BUT nothing stats that, and so I am not sure myself.