With Monte-Carlo-Simulation H-Energy of Fluid is not converging
 

Hi, I have the following question and really hope you can help me:

I have one Fluid Domain with one inlet with a constant mass flow at constant temperature and one opening with constant pressure. In the Fluid Domain there is the geometry of a temperature sensor which contains some Solid Domains. The purpose is to simulate the temperature profile of the sensor after a sudden change of the inlet temperature. All boundaries of the system are adiabatic. Without radiation the simulation works fine.

Now I want to add radiation, in fact Monte-Carlo-Radiation. I only use it in the Fluid Domain and I have the CCL-script at the End for defining multiband spectral data. To get an inicial guess I calculate a stationary case at first. In this case CFX starts working and all residuals fall down to 1e-06 except of the H-Energy of the fluid, as the screenshots show. What is the reason for that? Can anybody tell me a solution?

I use 1e+06 Histories and participating media model. The H-Energy is also not converging with 1e+04 an 1e+07 histories or with a grey spectral model and even if I calculate 1000 iterations...

http://lherzog.de/1.JPG


http://lherzog.de/2.JPG

http://lherzog.de/3.JPG

LIBRARY:
CEL:
EXPRESSIONS:
#
# Standard
#
InitialTemperature = 282[K]
InputTemperature = 1153[K]
SpecHeatCap Inconel600 = 562.11625[J/(kg*K)]
SpecHeatCap MgO = 1194.0885[J/(kg*K)]
SpecHeatCap Nicrosil = 545.32775[J/(kg*K)]
SpecHeatCap Nisil = 545.32775[J/(kg*K)]
ThermalConductivity Inconel600 = 18.46825[W/(m*K)]
ThermalConductivity MgO = 23.38217753[W/(m*K)]
ThermalConductivity Nicrosil = 21.89189375[W/(m*K)]
ThermalConductivity Nisil = 30.818[W/(m*K)]

#
# wavenumber limits
#
Kayser01 = 75000 [1/m]
Kayser02 = 115000 [1/m]
Kayser03 = 220000 [1/m]
Kayser04 = 240000 [1/m]
Kayser05 = 330000 [1/m]
Kayser06 = 405000 [1/m]
Kayser07 = 480000 [1/m]
Kayser08 = 580000 [1/m]
Kayser09 = 670000 [1/m]
Kayser10 = 755000 [1/m]
KayserHigh = 2571200 [1/m]
KayserLow = 1 [1/m]
#

#
# Dimensionless Wavenumbers
#

k01 = Kayser01 * 1[m]
k02 = Kayser02 * 1[m]
k03 = Kayser03 * 1[m]
k04 = Kayser04 * 1[m]
k05 = Kayser05 * 1[m]
k06 = Kayser06 * 1[m]
k07 = Kayser07 * 1[m]
k08 = Kayser08 * 1[m]
k09 = Kayser09 * 1[m]
k10 = Kayser10 * 1[m]
khigh = KayserHigh * 1[m]
klow = KayserLow * 1[m]
kdim = wavenumber*1[s]
#
# absorption coefficients [1/m]
#

absorp01 = 7.474906482
absorp02 = 0.231927953
absorp03 = 1.05844541
absorp04 = 33.46560635
absorp05 = 0.031275922
absorp06 = 1.323535546
absorp07 = 0.016650979
absorp08 = 0.097641775
absorp09 = 0.003470572
absorp10 = 0.084420621
absorp11 = 0.000247321

#
# Step Function per Interval
#
int01 = (step(k01-kdim)-step(klow-kdim))
int02 = (step(k02-kdim)-step(k01-kdim))
int03 = (step(k03-kdim)-step(k02-kdim))
int04 = (step(k04-kdim)-step(k03-kdim))
int05 = (step(k05-kdim)-step(k04-kdim))
int06 = (step(k06-kdim)-step(k05-kdim))
int07 = (step(k07-kdim)-step(k06-kdim))
int08 = (step(k08-kdim)-step(k07-kdim))
int09 = (step(k09-kdim)-step(k08-kdim))
int10 = (step(k10-kdim)-step(k09-kdim))
int11 = (step(khigh-kdim)-step(k10-kdim))

#
# absorption
#
absorpIR = 1*[1/m]*(absorp01*int01+absorp02*int02+absorp03*int03+abs orp04*int04+absorp05*int05+absor0p6*int06+absorp07 *int07+absorp08*int08+absorp09*int09+absorp10*int1 0+absorp11*int11)
END
END
END

Since every other equation seem to have converged, have you looked at other energy related quantities such energy flows, or radiation flows at boundaries of interest ?

The Monte Carlo radiation solution uses a brand new radiation field every iteration, and it usually shows as high level residual in the energy equation. Any particular reason you are using MC instead of Discrete Transfer for a fluid only simulation. The MC is more time consuming than DT with incremental benefit unless you get into the limitations of DT (not valid for semi-transparent solids and possibly ray effects due to low ray count). I would run the DT model as well and compare the solution to see if there is something to worry about.

Thanks for your quick response!

I use Monte Carlo because I've read an Ansys Radiation Modeling Guide and the disadvantages of DT sounded really bad. You could summarize it with "if you have time and compute power use Monte Carlo" ;-)
You're right, radiation is only enabled in a Fluid Domain. I have some Solid Domains in there, but they're all opaque.

In the following graphic attached I've plotted some energy flow at a Fluid Solid Interface where the energy flow of the two sides of the interface should be the same, right? But they really are not...
I'm particulary experienced in fluid simulation, but totally new to radiation modelling. Are these high energy residuals normal for Monte Carlo? How can I trust the solution?

Tomorrow, I will try DT, but I would be glad if I or somebody could solve the problem with Monte Carlo...

http://lherzog.de/4.JPG

Now there are some questions left:

If I select that MC-radiation is only solved every n iterations, how are the other iterations treated? Is there no radiation or is the radiation just halted, the same way they're halted if I use model overrides? Then I could save some time while I let the model converge without radiation and then turn on radiation for only some iterations, right?

Presently I give a try to DT, but it's very slow, much slower than Monte Carlo...