CFD Online Discussion Forums (http://www.cfd-online.com/Forums/)
-   Main CFD Forum (http://www.cfd-online.com/Forums/main/)
-   -   CFD package for very-low-speed convective flows in large structures (http://www.cfd-online.com/Forums/main/1554-cfd-package-very-low-speed-convective-flows-large-structures.html)

 Matthew Kuperus Heun November 18, 1999 13:58

CFD package for very-low-speed convective flows in large structures

I must select a CFD package for use on the following problem:

Determine internal and external convective heat transfer coefficients for the following structures:

* 100-m diameter semi-spherical thin-membrane balloon

* 100-m cylinder balloons with fineness ratios of 1 through 10

* Natural-shaped balloons

The following boundary conditions are of interest:

* Constant flux

* Constant temperature

* Step-change in flux

It is expected that buoyancy-driven natural convection flows will dominate the internal and external flow fields at high altitudes (35 km). Can anyone suggest a tool that would be appropriate for this application?

Matt H.

 Mahesh Prakash November 18, 1999 22:03

Re: CFD package for very-low-speed convective flows in large structures

Dear Mathew,

As far as I know, for pure natural convection turbulent flows the k-epsilon turbulence model does not perform very well in being able to predict the heat transfer coefficient. A lot of work has been done and is being done in this area at Delft University. I have thoroughly studied their work for a square cavity with heated vertical walls and adiabatic horizontal walls for the turbulent case. The low Reynolds numnber model of Jones and Launder performs best in predicting the heat transfer coefficient (approximately 3% difference between experiment and numerical results). For other models the difference can be as large as 70%!!!

Recently I also looked at a case with the top and side walls at the same low temperature and the bottom wall at a higher temperature and found that there is a significant difference in the prediction of heat transfer coefficient with the std. k-epsilon model (approx. 50% overprediction). I have yet to try other models for this geometry.

Thus the turbulence model that you will be using is of paramount importance in natural convection flows especially if your aim is to be able to predict the heat tranfer coefficient. I would suggest you to use the low-Re model of Jones and Launder to start with and carry out some validation exercise.

Hanajalic's group at Delft is working with Reynolds stress models to study the turbulent Ra-Br problem. It would be a good idea to read the work that they are doing right now to see whether you could use their model.

 All times are GMT -4. The time now is 04:26.