[litzj]

Yes, I know, thermal radiation and convection effect could analyzed.
(attempt itself is possible, and I use FLUENT 18 now.)

However, my question is that

Target furnace is about 1100 degree of Celsius and Input gas is about 7~800 degree while variation temperature of furnace heater is 50~80 degree at least.

Current target of temperature variation of furnace is less than +-5 degree of Celsius for specific position which is mechanically moved slowly.

I thought both improvement of the temperature variation and CFD correlation seem difficult. Is there any idea about this? (for better correlation and improvement itself)

From my literature survey, CFD analysis with very simple geometry furnace already shows discrepancy for more than 2~3%, about 20~30 degree at 1000 degree. Also, re-production of variation of heater temperature is not easy.

I hope some one recommend reference paper or method representing high temperature furnace with high accuracy in complex geometry

[piu58]

I think you cannot expect this degree of match. The main problem is that you cannot cover all the aspects of the reality in a CFD model.

I see two main reasons for doing CFD

1) You want to explore something. Therefore the simplest geometry and my be very simple boundary conditions are adequate. Beside this, all the physics should be reduced to the values you need to evaluate. You have to adapt your experiments to that simple case. In this case, the match between experiment and calculation should be very good.

2) You want to improve an existing facility. In this case the match doesn't need to reach the last few percents. It is far more important, that you see what happens if you change something. You need (more) to know the changes instead of the absolute values. It is mostly possible to calibrate your predicted result with the measuring values form the unchanged case.

[litzj]

Originally Posted by piu58

I think you cannot expect this degree of match. The main problem is that you cannot cover all the aspects of the reality in a CFD model.

I see two main reasons for doing CFD

1) You want to explore something. Therefore the simplest geometry and my be very simple boundary conditions are adequate. Beside this, all the physics should be reduced to the values you need to evaluate. You have to adapt your experiments to that simple case. In this case, the match between experiment and calculation should be very good.

2) You want to improve an existing facility. In this case the match doesn't need to reach the last few percents. It is far more important, that you see what happens if you change something. You need (more) to know the changes instead of the absolute values. It is mostly possible to calibrate your predicted result with the measuring values form the unchanged case.

Thanks to your comment. I appreciate it.

Unfortunately, correlation from using simple geometry is not available from my company; correlation via real facility is only option

And this is why I asked reference cases that existing CFD case or paper reproducing +-5 degree level variation in 1000 degree environment.

Requirement for that case is reducing +-5 degree variation in 1000 degree level furnace, and I think CFD reference model of mine should give enough fine confidence level for temperature variation.

[piu58]

You did not understand me fully:

I recommend that you take the measuring values to calibrate the results from your cfd results. Therefore ist is not necessary to use an oversimple geometry and physics.

[litzj]

Originally Posted by piu58

You did not understand me fully:

I recommend that you take the measuring values to calibrate the results from your cfd results. Therefore ist is not necessary to use an oversimple geometry and physics.

Oh I understood. Still, I could not convince that guys in the furnace lab give appropriate data, but it is worth to try it. Thanks again.