Big difference in temps between real and computed

evidence · January 5, 2014, 14:16

Hi guys,

i'm a new to CFD so i started with an experiment covering thermal management but i cant get that done properly in CFD tool (I'm using Autodesk CFD'14 software).

So i'm having a real selfmade device and it's exact 3D model. I think everybody know its a LEDs mounted on a alluminium heatsink: 15 Cree XB-Ds soldered on alluminum star-like heatsinks wich are mounted on a big alluminium heatsink. The temperature under the bottom-left "star" is 69 degrees in real.

So i export that model into the CFD software. Assign proper materials:

Alluminium to Big heatsink and 15 star-like heatsinks
Silicon to LED
Contact resistance of 6K/mm2*W to surfaces between big and star-like heatsinks, and LED between star-like heatsinks
Air to air, of course

As the boundary conditions i choose Pressure of 0 Pa on top and bottom surfaces of Air volume, and temperature of 25 C at the bottom of the same volume.

Also i choose a Total heat generation of 1,5W to each of the LED, because Cree XB-D has a 75% of heat generation of its power, and i run them at 2W each.

So when i push solve (gravity goes along Z and having 50 itterations) i get frustrating results of 265 C.

I also tried applying Thermal Compact model to LED (JB: 6.5C/W, JC: 1000C/W) but the temp raised up for 50% so i removed it.
So maybe anyone could what i'm doing wrong?

lovecraft22 · January 5, 2014, 16:03

Don't know anything about the software you are using but… did you make sure your case has converged?

evidence · January 5, 2014, 18:11

Thank for a reply.

No, i dont, but i do not know if its necessary, because it's having a temp of 250 C and i dont think that it would fall to 60-70 C. But i should try giving it some 300 itterations to make it converged. Will see

evidence · January 6, 2014, 06:15

Ran it to convergence and it gave me a result of 179 C which is not suitable yet

lovecraft22 · January 6, 2014, 16:07

Yes, but it is already 100deg lower than before. Are you sure this time it is fully converged?
It takes time for a simulation to converge, usually it is not a matter of short periods…

evidence · January 6, 2014, 17:18

Yeap, the programm says that it fully converged. Maybe there is not enough place for heat dissipation, I mean I need to make the air volume more larger?

beer · January 6, 2014, 18:46

Hi

Just a short thought of mine: It seems you have considered the cooling due to convection, have you also used the Boltzmann radiation? I can make a difference sometimes.

Regards

evidence · January 7, 2014, 08:07

Thank you,

have checked the Radiation to include and now i have the converged results of 85 degress, however i thought it might be less than 69 due to the programm has the "ideal" conditions for running the simulation

beer · January 11, 2014, 06:00

Well at least it`s getting better

Other possible errors: mesh resolution near the cooling body: if you can't resolve the (temperature) boundary layer with the mesh, the heat flow may be calculated incorrectly.

Another approach is to check for the turbulence. Especially at high temperature gradients the buoyancy causes a small turbulence which increases the effective heat conductivity. In trubulence modells this effect is already considered. So even if the main flow is not turbulent at all, local turbulence can occur and affect your heat flow. So think about including a simple k-epsilon modell maybe to see if it helps.

I would be interested if it helps, too. If it does, please let me know.

Cheers

evidence · January 11, 2014, 16:52

Thank you Beer, will try

But there's another problem - i have mistaken the models due to my stupidity. So the temparature for the above model is only 58 C in real life. So i gone digging

evidence · January 26, 2014, 06:18

Hi there,

so i went with ANSYS steady-state thermal and the results are very impressive.

My real life model had a temp near the heating element near 52 C with the ambient air of 19. And ANSYS calculated that my temp will have nearly 46 C (i have made a minimum set of settings, only including standard radiation-into-air 19 C with default settings and heat generation of 0,265 W/mm3 due to element has area of 3,9 mm3 and thermal dissipation of 1,4W).

I have made couple of other real-life experiments and put them into ANSYS and the results were nearly the same as the real ones, having the difference of 7-9% due to not-ideal real conditions.

One more - i have choosen 200 itterations but the temperature was the same as in the first step, so i've had a table of 200 rows showing me the same temp values - is it okay? Thank you

January 5, 2014, 14:16	Big difference in temps between real and computed	#1
evidence New Member Stan Join Date: Jan 2014 Posts: 7 Rep Power: 12	Hi guys, i'm a new to CFD so i started with an experiment covering thermal management but i cant get that done properly in CFD tool (I'm using Autodesk CFD'14 software). So i'm having a real selfmade device and it's exact 3D model. I think everybody know its a LEDs mounted on a alluminium heatsink: 15 Cree XB-Ds soldered on alluminum star-like heatsinks wich are mounted on a big alluminium heatsink. The temperature under the bottom-left "star" is 69 degrees in real. So i export that model into the CFD software. Assign proper materials: Alluminium to Big heatsink and 15 star-like heatsinks Silicon to LED Contact resistance of 6K/mm2*W to surfaces between big and star-like heatsinks, and LED between star-like heatsinks Air to air, of course As the boundary conditions i choose Pressure of 0 Pa on top and bottom surfaces of Air volume, and temperature of 25 C at the bottom of the same volume. Also i choose a Total heat generation of 1,5W to each of the LED, because Cree XB-D has a 75% of heat generation of its power, and i run them at 2W each. So when i push solve (gravity goes along Z and having 50 itterations) i get frustrating results of 265 C. I also tried applying Thermal Compact model to LED (JB: 6.5C/W, JC: 1000C/W) but the temp raised up for 50% so i removed it. So maybe anyone could what i'm doing wrong?

January 11, 2014, 06:00		#9
beer Member Join Date: Dec 2012 Posts: 92 Rep Power: 13	Well at least it`s getting better Other possible errors: mesh resolution near the cooling body: if you can't resolve the (temperature) boundary layer with the mesh, the heat flow may be calculated incorrectly. Another approach is to check for the turbulence. Especially at high temperature gradients the buoyancy causes a small turbulence which increases the effective heat conductivity. In trubulence modells this effect is already considered. So even if the main flow is not turbulent at all, local turbulence can occur and affect your heat flow. So think about including a simple k-epsilon modell maybe to see if it helps. I would be interested if it helps, too. If it does, please let me know. Cheers Last edited by beer; January 11, 2014 at 07:22.

Similar Threads
Thread	Thread Starter	Forum	Replies	Last Post
FLUENT received fatal signal (ACCESS_VIOLATION)	samy	FLUENT	0	November 10, 2007 13:09

January 5, 2014, 16:03		#2
lovecraft22 Senior Member lore Join Date: Mar 2010 Location: Italy Posts: 460 Rep Power: 18	Don't know anything about the software you are using but… did you make sure your case has converged?

January 5, 2014, 18:11		#3
evidence New Member Stan Join Date: Jan 2014 Posts: 7 Rep Power: 12	Thank for a reply. No, i dont, but i do not know if its necessary, because it's having a temp of 250 C and i dont think that it would fall to 60-70 C. But i should try giving it some 300 itterations to make it converged. Will see

January 6, 2014, 06:15		#4
evidence New Member Stan Join Date: Jan 2014 Posts: 7 Rep Power: 12	Ran it to convergence and it gave me a result of 179 C which is not suitable yet

January 6, 2014, 16:07		#5
lovecraft22 Senior Member lore Join Date: Mar 2010 Location: Italy Posts: 460 Rep Power: 18	Yes, but it is already 100deg lower than before. Are you sure this time it is fully converged? It takes time for a simulation to converge, usually it is not a matter of short periods…

January 6, 2014, 17:18		#6
evidence New Member Stan Join Date: Jan 2014 Posts: 7 Rep Power: 12	Yeap, the programm says that it fully converged. Maybe there is not enough place for heat dissipation, I mean I need to make the air volume more larger?

January 6, 2014, 18:46		#7
beer Member Join Date: Dec 2012 Posts: 92 Rep Power: 13	Hi Just a short thought of mine: It seems you have considered the cooling due to convection, have you also used the Boltzmann radiation? I can make a difference sometimes. Regards

January 7, 2014, 08:07		#8
evidence New Member Stan Join Date: Jan 2014 Posts: 7 Rep Power: 12	Thank you, have checked the Radiation to include and now i have the converged results of 85 degress, however i thought it might be less than 69 due to the programm has the "ideal" conditions for running the simulation

January 11, 2014, 16:52		#10
evidence New Member Stan Join Date: Jan 2014 Posts: 7 Rep Power: 12	Thank you Beer, will try But there's another problem - i have mistaken the models due to my stupidity. So the temparature for the above model is only 58 C in real life. So i gone digging

January 26, 2014, 06:18		#11
evidence New Member Stan Join Date: Jan 2014 Posts: 7 Rep Power: 12	Hi there, so i went with ANSYS steady-state thermal and the results are very impressive. My real life model had a temp near the heating element near 52 C with the ambient air of 19. And ANSYS calculated that my temp will have nearly 46 C (i have made a minimum set of settings, only including standard radiation-into-air 19 C with default settings and heat generation of 0,265 W/mm3 due to element has area of 3,9 mm3 and thermal dissipation of 1,4W). I have made couple of other real-life experiments and put them into ANSYS and the results were nearly the same as the real ones, having the difference of 7-9% due to not-ideal real conditions. One more - i have choosen 200 itterations but the temperature was the same as in the first step, so i've had a table of 200 rows showing me the same temp values - is it okay? Thank you