[Logan Page]

Hi

I am masters engineering student and I'm currently trying to recreate the first part of results from a journal article: "Constructal multi-scale cylinders with natural convection" by T. Bello-Ochenda and A. Bejan, in FLUENT.

Simple put its a row of parallel, horizontal, cylinders (diameter Do) with a constant wall temperature (Tw) cooled by natural convection analyzed in 2D.

The current computation domain that i have setup for FLUENT is made up of:
1) a symmetry plane/edge on the left cutting through the middle of a cylinder
2) a symmetry plane/edge on the right cutting through the middle of the spacing (So) between 2 of the cylinders
3) a pressure inlet at the bottom edge (P = 0, T = 0)
4) a pressure outlet at the top edge (P = 0, T = 0)
5) and a cylinder wall (no slip, no penetration, Tw = 1)

The governing equations have been non-dimentionalised and the input parameters for a Rayleigh Number of 1000 for FLUENT are as follows:
- Do: 1
- Fluid density: 37.268 (boussinesq approximation)
- Fluid Cp: 0.72
- Fluid thermal conductivity: 1
- Fluid viscosity: 1
- Fluid thermal exp. coeff.: 1
- Operating pressure: 0
- Gravity: -1 (Y dir)
- Operating temp.: 0
- Specific operating density: 37.268

Using this model, FLUENT converges nicely for a given upstream (Hu) and downstream (Hd) length from the cylinder, but when i do a grid independence study i find that as i increase the downstream length (Hd) the heat transfer density rate continues to increase (ie there is a chimney effect created).

Does anyone have any suggestions / ideas on how to modify the boundary conditions to eliminate this effect?