- **FLUENT**
(*http://www.cfd-online.com/Forums/fluent/*)

- - **ON TURBULET MODEL FOR A NATURAL CONVECTION PROBLEM
**
(*http://www.cfd-online.com/Forums/fluent/29317-turbulet-model-natural-convection-problem.html*)

ON TURBULET MODEL FOR A NATURAL CONVECTION PROBLEM
Hallo! friends,
I am presently simulating a natural convection problem inside a vertical tube of 8cm*1500cm dimension. The Rayleigh number(~2*10exp9) predicts a turbulent convection inside. Who will advice me which turbulent model in Fluent is suitable to this problem. With advanced Thanks, Varghese |

Re: ON TURBULET MODEL FOR A NATURAL CONVECTION PRO
Why don't you select one by doing some comparisons of the results with experiments.
I did runs with ke, ke low-Re, rng and rns for one of my problems and found that rns is best, but it takes a lot of computing resources, so I'm using ke for most problems. It depends what's important in your problem, and the computing trade-offs which is best. You need to determine this for your own problem. Greg |

Re: ON TURBULET MODEL FOR A NATURAL CONVECTION PRO
Dear Greg,
Thank you very much for the response. I am looking for experimental data for the problem,but failed to find one so far. My aim is to study the height to radius ratio of a vertical cylinder with bottom heating(whose height is fixed at 1500cm),to find the threshold ratio when the convection predominates conduction.I donot know what shall be the norm in choosing one of the turbulent models. Personally, you said, you prefer rns model(is it the same as realizable k-e model,or?) thanks for any useful tip, Varghese |

Re: ON TURBULET MODEL FOR A NATURAL CONVECTION PRO
Well the rns model is the Reynolds Stress Model. I don't necessarily prefer it - it just gave slightly better results, than the others. But this comes at a cost.
For some problems it may not be better. I don't know. If you can't find measurements for your problem, look for a similar problem and model that, in order to give some indication of the effect of turbulence model on the results. This way you can choose/fix a turbulence model and then look at your problem. I would expect that the relative behaviour of the models would be similar, but this may not be the case once you start looking a minute details. Thus, this is where you'll need to define the scope of your study. Greg |

Re: ON TURBULET MODEL FOR A NATURAL CONVECTION PRO
Dear Greg,
Thank you for the suggestions! I have already implimented the k-e model for the problem and I can compare some features of interest with other models. For example I have an idea to compare the Nusselt number, which, hopefully predicts the predominance of convection in the system. Now, the Fluent gives an estimate of the surface Nusselt number in the wall fluxes category as a local non-dimensional coefficient of heat transfer. Does its definition in the wall fluxes category, forbids one to have a volume/mass averaging over this quantity? Can one legitimately interpret that when the mass-weighted average of the surface Nusselt number becomes greater than 1, the convection heat transfer predominates over the conduction heat transfer? Is the wall fluxes defined locally for each control volumes? In that case which kind of averaging will allow one to have a prediction for the total system? With greetings, Varghese |

Re: ON TURBULET MODEL FOR A NATURAL CONVECTION PRO
Yeah the Nusselt number is generally a good measure for lookin at a natural convection system.
Fluent's definition of Nu may not be the same as what you require for your problem. In one of my problems, another author used a different Nu definition and so the best way to compare was to write a udf to calculate the Nu like that author and not worry about Fluent's calculations. You might do something similar, or use custom field functions if the transformation is fairly straightforward.... Greg |

Re: ON TURBULET MODEL FOR A NATURAL CONVECTION PRO
Hallo! dear Greg,
Thank you for sharing your experiences. As you have mentioned, I realize that there exits many definitions for the Nusselt number in literature. My effort is, to understand how Fluent defines the Nusselt number. It says, Nu is evaluated locally and can be averaged suitably. Does it mean that Nu is evaluated for the interior zones also! If it is so, how it defines T-wall, which appears in the definition of Nu through h-eff, for each interior cell? Or if it evaluates Nu only for the boundary cells, which value does it adopt for the same? In any case, how to set T-ref and l-ref for each cell! Thanking you for the troubles you take to answer me, With Greetings, Varghese |

All times are GMT -4. The time now is 17:27. |