CFD Online Discussion Forums (https://www.cfd-online.com/Forums/)
-   CFX (https://www.cfd-online.com/Forums/cfx/)
-   -   CFX thermo-fluid 2D simulation problems (https://www.cfd-online.com/Forums/cfx/93136-cfx-thermo-fluid-2d-simulation-problems.html)

 maryliz October 5, 2011 23:39

CFX thermo-fluid 2D simulation problems

Hi,
I've been trying to run a simulation using CFX, but i'm having a few problems. My problem envolves cooling two integrated circuit boards coupled with 5 heat source place at the lower board (http://imageshack.us/photo/my-images...eometrial.jpg/). I'm considering laminar flow and using a 2D Symmetry approach.
The mesh (http://imageshack.us/photo/my-images/20/malhab.jpg/) was generated using the sweep command with quad elements and edge sizing in all heat source faces and fluid outer faces.
I divided the problem in 3 domains (heat source, IC boards and fluid). The fluid is air, set as ideal gas, which enters the channel at 293.15K and 1.5 m/s. At the channel outlet, it was set the average static pressure as 0 Pa and press profile blend equas 0.05 (ansys default).
The heat source generate 5W and this energy is dissipated only at the top face of the heat source, the other faces are considered adiabatic. The boards in its whole length are also adiabatic.
The problem is that, the heat sink should be gradually cooled by the fluid as the simulation goes by. However, the heat source temperature remains around 3000K at the end of simulation.

I wonder if anyone has any ideia about what am I doing wrong. Or even give me some tips to improve the problem solution, such as: how improve mesh quality, or set some different boundary conditions. I posted the picture of my mesh and the residues I obtained in one of my simulation.
:)

 sberez October 6, 2011 05:20

If you can upload an image of your results (Temperature contour) and also more details of the boundary conditions and domain properties will be helpful.

 ghorrocks October 6, 2011 07:12

Are the interfaces between the domains properly set up? Does this simulation need radiation?

 maryliz October 6, 2011 22:07

Boundary Conditions of the chip:
- solid domain;
- domain motion: stationary;
- heat transfer-thermal energy;
- no conditions initial
- source energy: total source 5W;

Boundary Conditions of the fluid:
- material: air ideal gas;
- reference pressure 1 atm;
- buoyancy: gravity x e z = 0 and gravity y= -g;
- buoyancy reference density= 1.204 kg/m^3
- heat transfer thermal enegy and turbelence laminar;
- initialization conditions: U=1.5m/s, automatic with value 293.15K and static pressure is automatic;
- inlet: subsonic, 1.5m/s and total temperature = 293.15K;
- outlet: relative pressure 0 Pa and pressure profile blend 0.05 and substrato is adiabatic
I noticed that the simulationh results has been showing some unexpected behavior, the chips heat transfer behavior also varies . That is, the chips next to the outlet becomes more cooled than the first one, when the correct is the opposite situation (of course).
(http://imageshack.us/photo/my-images/80/100xn.jpg/)

 ghorrocks October 6, 2011 23:00

I would check your interfaces. If the fluid/solid interface is not present then the solid will heat up and have no mechanism to loose the heat. This will cause the temperature to go very high - which is just what you are seeing.

 sberez October 7, 2011 03:16

i think that there are other few points you should check:
I guess that 5W if for the whole heat sink (at 3D), because it's 2D divide it by the depth of the component.
I'm not sure that laminar flow is right for this simulation.

 maryliz November 1, 2011 00:26

I need to find the entropy generation in a channel with heaters in the CFX. Because the program does not calculate the entropy generation, I've decided to make a new approach: I've collected the speed and temperature gradient data generated by CFX, calculated the entropy generation at each point and then I took the entropy generation mean value. This approximation is valid? What are the other ways to find the entropy generation? How can I introduce the entropy generation PDEs in CFX? I know that I can do this by creating a FORTRAN routine, but I haven't found where do I write this routine in CFX.

I know this subject has already been discussed in this forum, but I looked for this and haven't found any answer about it.