Hi,

I'm having trouble producing turbulence in a buoyant flow induced by a heated cylinder. According to theory and articles the flow should indeed be turbulent but never is, no matter how high the Rayleigh number.

As of now I've run both steady and unsteady simulations on different grids with varying density. I have taken in consideration many of the tips given in this forum for my simulations but to no avail.

I've read about false time stepping where the under-relaxation factors are set for each cell, supposedly this should be good for a buoyant flow. Is there such a function in Fluent?

It is turbulent at Ra 10^9. You will only see a plume rising above the cylinder. For turbulent flow you will see a later separation.

I suggest that you model the half the domain and use symmetry because its easier to get a converged steady state solution. If you model full doamin, during your solution you will see the sway in the plume and that makes the problem transient. By modelling half of the domain u can reach steady state solution.

Best regards Haroon

Hi,

I have also noticed the sway in the plume as you describe it and I have done my simulations with symmetry for the most part. My Ra number is 10^11 so the flow is most likely turbulent, however this is not visible in the Nusselt x-y plot around the cylinder. The Nusselt number should show a sharp increase in the transition region and the decrease to 0 when the flow separates from the cylinder face.

Have you experienced this effect when simulating a turbulent buoyant flow?

Hi,

I think you need to plot Nusselt number against the angle, and the separation angle is over 120 for the tubulent case as far as remember. In fluent you will have to plot the cylinder boundary wall against the Nu. Yes I did get reasonable results for cylinder.

The Nu may not be completely zero where the separation occurs. There is recirculation in the wake region due to which the Nu is always a value slightly greater than 0. As compared to the rest of the boundary this value will be very low so you can say theoritically it is zero.

Hope this helps.

Hi,

thank you for your response. I am plotting the nusselt number in a correct way and the plots and consistent with the lamniar case, only showing a very small "hump" at 60 degrees and then decreases steadily to 140-150 degrees where it drops as a result of separation.

My residuals are very oscillative, maybe the solution is not yet converged and could benefit from lower relaxation? I already have low relaxation on pressure, momentum and body force, (0.2 - 0.5).

I am using density as a function of temperature for the water (incompressible-ideal-gas), my temperature difference between pipe and water is approx. 15K.

hi The residuals will oscillate because of the plume sway so you cannot stop that.

I thinks bousinesq density model works ideal for bouyancy driven flow. The thermal expansion coeffienceint fro water at 15K is 0.0004 approximately.