[sasanghomi]

Dear friends,

These days, I am simulating some benchmarks about radiation. There are two questions in my mind;

1) The problem is that when some adiabatic walls are involved in modelings, the accuracy of results decreases tangibly.
It should be noted that I am applying DTM (Surface to Surface) in the simulations.

2) There exist some parameters relevant to advanced options for radiation problems (Thermal Radiation Control). I could not find any tutorial in which these parameters are discussed. Does anybody here have any experience about changing these parameters?

Any hint is appreciated.

Best Regards

[ghorrocks]

I do not know what CFX does if you define a wall to be adiabatic but then use a radiation model and make radiative heat transfer occur at the wall. You would have to experiment with that.

It is one of the great problems with CFX that they do not have any tutorial examples showing actual use of the software - covering things like working out mesh size or time step size, or adjusting a model like a radiation model to give accurate results. The tutorials just show how to get the model to run, they don't show how to get it to run accurately.

I do not know of any tutorials showing how to adjust the radiation model for accuracy. All I can suggest is to do a sensitivity analysis on all the likely looking parameters and find out which ones are important for your model. It is time consuming, but at least you then know for sure which parameters are important for your case.

[sasanghomi]

Thank you so much dear Glenn Horrocks.

[Opaque]

If you feel the problem is with the adiabatic walls, can you isolate a simpler test case that illustrates what the problem is?

The Discrete Transfer model discretization parameter is the Number of Rays. Which value to use is problem dependent, and there is no single answer for it.

How did you determine the accuracy is not good? What steps have you taken to study the accuracy of the model?

Some of those answers may help some in the forum to suggest the next step.

[sasanghomi]

Originally Posted by Opaque

If you feel the problem is with the adiabatic walls, can you isolate a simpler test case that illustrates what the problem is?

The Discrete Transfer model discretization parameter is the Number of Rays. Which value to use is problem dependent, and there is no single answer for it.

How did you determine the accuracy is not good? What steps have you taken to study the accuracy of the model?

Some of those answers may help some in the forum to suggest the next step.

In fact, I am simulating a benchmark which is solved analytically in Incopropera Fundamental Heat Transfer book. There is a cylinder whose top and bottom faces have constant temperatures and the lateral face is adiabatic. The problem is that Wall Radiative Heat Flux on the top and bottom faces have different values compared to the results of analytical solutions. (Incopropera Fundamental Heat Transfer book) (A discrepancy of 20%)

Best Regards

[ghorrocks]

Good to see you are doing some benchmark simulations. It is a very good idea to check you can get the benchmarks right before you tackle your own simulation.

The key tunable parameter in the Discrete Transfer radiation model is the number of rays parameter mentioned by Opaque. Have you done a sensitivity analysis on that parameter?

[Opaque]

Perfect!!!

Now careful. How is your setup in ANSYS CFX?

If you are taking an example case in Incropera&DeWitt from the pure thermal radiation chapter, you must keep in mind those problems are w/o conduction, no advection, and transparent media only. Therefore, you cannot solve the energy equation; otherwise, you are not solving the same problem. Perhaps, lowering the thermal conductivity as low as you can while the equation can still be solved robustly.

Once the above is under control, you still need to consider:
- How are you modeling your adiabatic walls? Black, or highly reflective?
For pure radiation problems, adiabatic walls are basically re-radiating walls
- What is the aspect ratio of the cylinder, say Height / Diameter?
The number of rays is a function of the aspect ratio of the cylinder. Let us say, your aspect ratio is 10 (long cylinder), some rays do not have "direct view" of the whole cylinder since, i.e. you need more rays so they can travel to the other side and contribute to the irradiation calculation.

Hope the above gives a better insight on the issues when doing numerical thermal radiation calculations.

[evcelica]

Originally Posted by Opaque

Therefore, you cannot solve the energy equation; otherwise, you are not solving the same problem. Perhaps, lowering the thermal conductivity as low as you can while the equation can still be solved robustly.

Or just turn off solving of the fluids and energy equations:
In CFX Pre: go to the top:

INSERT >> SOLVER >> Expert Parameter, go all the way to the right under the model overrides tab.

Set "solve fluids" and "solve energy" to "f" and "solve radiation" to "t"