CFD Online Discussion Forums

CFD Online Discussion Forums (
-   CFX (
-   -   Monte Carlo vs Discrete Transfer (

TobiasZ January 23, 2006 05:48

Monte Carlo vs Discrete Transfer
Dear all,

I have used the Discrete Transfer (DT) and Monte Carlo (MC) radiation models to predict the radiation heat exchange between two parallel walls with different temperatures and emissivities of 0.9.

For the MC model I obtain arbitrary non-physical spots of high rediative heat losses and gains on both walls(even for 1.000.000 histories). For the DT model I obtain good results with quite homogeneous wall radiative heat flux on the surface (using 8 rays, as default).

I've read in the user manual that for highly transparent media, such as for air, the MC model should be used to predict the radiative interaction between the model surfaces; and the DT model was ostensively used with limited success.

+ Does the DT model consider reflection at surfaces? + Is the symmetry plane treated like a mirror for radiation? + Can the DT model be used also for models with non-homogeneous boundaries for example for building simulations (and not only as everywhere suggested for furnances)? + Why my errors for the MC model?

Can anyone bring light into that problem?

Mike January 23, 2006 11:59

Re: Monte Carlo vs Discrete Transfer
In each case, do the radiation diagnostics written to the OUT file each iteration look acceptable? M

TobiasZ January 24, 2006 04:31

Re: Monte Carlo vs Discrete Transfer
Hi Mike and thank you for your reply,

For MC model avr values are: %SDSur 25; %SDVol 60; %Lost 0; %Imbal 0 For DT model avr values are: Its 6; VolChg 0; SurChg 6E-3; %Lost 0; %Imbal 1.3

Unfortunately I don't know much about these values, but I guess these values are not bad.

May it be possible, that my symmetry planes make a problem for the MC model?

TobiasZ January 25, 2006 04:32

Re: Monte Carlo vs Discrete Transfer
Hi Mike,

Yesterday I saw that Service Pack 1 is out now for CFX 10.0. This patch repairs the error of calculating the rays each step for the Discrete Transfer Model. However, this didn't affect my results.

May I ask you a few more questions about the DT model?

Does the solver calculate the ray tracks for each node on the boundaries or only at certain arbitrary points?

I guess the DT model calculates reflection, but as diffuse reflection and not as specular reflection?

Let's assume the case that the temperature boundary conditions are initially homogeneous and change over time further on, then the solver wouldn't create and store any tracks in the beginning of the simulation and would not be capable to comprise the radiation and reflections during the simulation. Is this correct?

I'd be grateful, if you have any idea.



opaque January 25, 2006 11:21

Re: Monte Carlo vs Discrete Transfer
Dear Tobias,

ANSYS CFX 10.0 SP1 fix for DTM you are referring to does not affect the results from previous calculations, it only restores the speed that 5.7.1 used to have..

Keep in mind that the radiation tracker (DTM, and MC) in ANSYS CFX does not work on the CFD mesh, but on a coarsened mesh (same mesh is the coarsening ratio is 1). The DT model traces from the centre of each radiation element face on the domain boundary. Those tracks (pure geometry) are stored in the radiation track file, and reused for later coefficient loops. This information is constant in time, unless the mesh is deforming and the tracker will recompute the traces again.

If your temperature/emissivity boundary conditions are a function of time, that should not affect the tracks in any way. Only the energy emitted, or reflected is affected. The tracks do not contain energy, only distances and location within the mesh..Summary: you are safe on this issue.

Good luck, Opaque..

TobiasZ January 25, 2006 12:11

Re: Monte Carlo vs Discrete Transfer
Opaque, thank you for your quick reply. This information helped much. I think I understand your point.



All times are GMT -4. The time now is 07:40.