[B_R_Khan]

Hi!

I have come across spectral distribution (2 1) as an input asked from user in almost all radiation models including solar load. I don't know what does it mean in the radiation models of openfoam..Is it related to number of bands? or is it power divided by wavelength? I'm lost. Please help!


Thanks

[JulioPieri]

I'm not an expert, but from the source code solarLoad.C:

{
qprimaryBf[patchID][localFaceI] +=
    (qPrim & n[localFaceI])
    * spectralDistribution_[bandI]
    * absorptivity_[patchID][bandI]()[localFaceI];
}

Which gives power (q) for an specific band (bandI).

Also, from other sources I get that it is a way to weight the flux.

Lastly, from fvDOM.H:

spectralDistribution is the energy spectral distribution of the collimated
     external beam.

Hope this helps

[B_R_Khan]

Originally Posted by JulioPieri

I'm not an expert, but from the source code solarLoad.C:

{
qprimaryBf[patchID][localFaceI] +=
    (qPrim & n[localFaceI])
    * spectralDistribution_[bandI]
    * absorptivity_[patchID][bandI]()[localFaceI];
}

Which gives power (q) for an specific band (bandI).

Also, from other sources I get that it is a way to weight the flux.

Lastly, from fvDOM.H:

spectralDistribution is the energy spectral distribution of the collimated
     external beam.

Hope this helps

Hi! Thanks for your reply

I have found the formula for calculating spectral distribution (screenshot attached).
So from the code, it looks like there are two bands ; band 1 has value of 2 W/m2.nm and band 2 has value of 1 W/m2.nm.
Do you know how bands are defined in radiation? like which wavelength range corresponds to which band?

Thanks

[JulioPieri]

As from the formula you posted, the spectral step is infinitesimal (del_lambda).

When you approximate it to a finite interval, you consider delta_lambda for you "integration" over the wavelength. So you calulate it for each interval o wavelength you want to consider.

Again, I'm no expert, but you might want to discretize your band (i.e. visible light spectrum) into, say, 10 steps. So you delta_lambda would be WholeBand/10. Each of this step will contribute differently to the flux

[B_R_Khan]

Originally Posted by JulioPieri

As from the formula you posted, the spectral step is infinitesimal (del_lambda).

When you approximate it to a finite interval, you consider delta_lambda for you "integration" over the wavelength. So you calulate it for each interval o wavelength you want to consider.

Again, I'm no expert, but you might want to discretize your band (i.e. visible light spectrum) into, say, 10 steps. So you delta_lambda would be WholeBand/10. Each of this step will contribute differently to the flux

Do you know which entry will correspond to which wavelength interval?
I have made some observations:
When I increase spectral distribution from (2 1) to (5 2) to see it’s effect on results, qr becomes 840 W/m2. Changed it to (10 20) and got value of 1200 W/m2. Again changed it to it to (20 10) and got value of 880 W/m2.
I don't know the explanation for my findings. Can you explain this weird trend?

[JulioPieri]

I'm not sure, but reading the solarLoad.C file, it seems it is weighting over bands

spectralDistribution_ =
        spectralDistribution_/sum(spectralDistribution_);

Two guesses here:
1- the first number is the number os bands and the second a weight factor.
2- the array is a distribution over the band in equal steps. For instance, you could write something like (1 2 3 2 1) so 10% would be on the first 1/6th, 20% on the second, etc. Try putting more numbers in the array to check this

[B_R_Khan]

Originally Posted by JulioPieri

I'm not sure, but reading the solarLoad.C file, it seems it is weighting over bands

spectralDistribution_ =
        spectralDistribution_/sum(spectralDistribution_);

I changed the spectral distribution from two entries to 5 entries as you mentioned but I got the following errors:

[2] --> FOAM FATAL ERROR: (openfoam-2106)
[2] Radiation bands : 5
Bands on patch : 0 is 2

I guess this means that the entries in spectral distribution are weights for each band

[JulioPieri]

Thanks! I was expecting to be the whole spectrum actually... Because it's weird that we limit the inputs in just two scalars. I'd expect more versatility

Let's see, isn't there any other file where you specify the number of bands?

Edit: try exploring the boundaryRad file. Take a look at the error:

// Model number of bands
         label nBands = radiation.nBands();
  
         const IOdictionary radiationDict(boundaryIO);
  
         forAll(mesh.boundary(), patchi)
         {
             const polyPatch& pp = mesh.boundaryMesh()[patchi];
  
             if (radiationDict.isDict(pp.name()))
             {
                 const dictionary& dict = radiationDict.subDict(pp.name());
  
                 radBoundaryPropertiesPtrList_[patchi].reset
                 (
                     boundaryRadiationPropertiesPatch::New(dict, pp)
                 );
  
                 if (nBands != radBoundaryPropertiesPtrList_[patchi]->nBands())
                 {
                     FatalErrorInFunction
                         << "Radiation bands : " <<  nBands << nl
                         << "Bands on patch : " << patchi << " is "
                         << radBoundaryPropertiesPtrList_[patchi]->nBands()
                         << abort(FatalError);
                 }
             }

So the "nBands" must be equal to "radBoundaryPropertiesPtrList" which is a subDict inside a radiation properties file.

Is your mesh heavy? Send me you case file (preferebly in zip format) and I'll have a look

[B_R_Khan]

Originally Posted by JulioPieri

Thanks! I was expecting to be the whole spectrum actually... Because it's weird that we limit the inputs in just two scalars. I'd expect more versatility

Let's see, isn't there any other file where you specify the number of bands?

Edit: try exploring the boundaryRad file. Take a look at the error:

// Model number of bands
         label nBands = radiation.nBands();
  
         const IOdictionary radiationDict(boundaryIO);
  
         forAll(mesh.boundary(), patchi)
         {
             const polyPatch& pp = mesh.boundaryMesh()[patchi];
  
             if (radiationDict.isDict(pp.name()))
             {
                 const dictionary& dict = radiationDict.subDict(pp.name());
  
                 radBoundaryPropertiesPtrList_[patchi].reset
                 (
                     boundaryRadiationPropertiesPatch::New(dict, pp)
                 );
  
                 if (nBands != radBoundaryPropertiesPtrList_[patchi]->nBands())
                 {
                     FatalErrorInFunction
                         << "Radiation bands : " <<  nBands << nl
                         << "Bands on patch : " << patchi << " is "
                         << radBoundaryPropertiesPtrList_[patchi]->nBands()
                         << abort(FatalError);
                 }
             }

So the "nBands" must be equal to "radBoundaryPropertiesPtrList" which is a subDict inside a radiation properties file.

Is your mesh heavy? Send me you case file (preferebly in zip format) and I'll have a look

I think the number of bands can be increased by changing the number of entries in absorption emission coefficients in boundary radiation properties file as well as spectral distribution elements in radiation properties file.

I did not change number of entries in absopritvity and emissivity arrays in radiation boundaries before and hence got the error:

panel
{
    type        opaqueReflective;

    // Fraction of the reflected is diffussive
    fd          0; // 0: all specular 1: all diffusive

    wallAbsorptionEmissionModel
    {
        type            multiBandAbsorption;
        absorptivity    (0.3 0.7);
        emissivity      (0.3 0.7);
    };
}

I have not used any other radiation model like viewfactor or fvDOM along with solarLoad as simpleSolarCar tutorial too doesn't include any other radiation model. Viewfactor model gives option of nbands in its usage. Solar load does not ask for any nbands input though

I will try it with 5 entries of spectral distribution & 5 absorptivity/ emissitivity entries and let you know about the results. But the question still stands that which band interval ( UV, visible etc?)will correspond to which value of spectral power.

Sorry I can't send the case files right now without asking my supervisor Let me ask him and I will get back to you. I used simpleSolarCar tutorial from openfoam.com as the base files.

Thanks!!
Bushra

[JulioPieri]

Well, that I don't know for sure. I'd expect to be just infrared, since we're interested in heat transfer, but I just read that Fluent uses a parameter to weight between visible and infrared. Try digging into the OF code, or looking at Fluent's manual! Let me know if you figure out.

[B_R_Khan]

Originally Posted by JulioPieri

Well, that I don't know for sure. I'd expect to be just infrared, since we're interested in heat transfer, but I just read that Fluent uses a parameter to weight between visible and infrared. Try digging into the OF code, or looking at Fluent's manual! Let me know if you figure out.

So I changed the absorptivity and emissivity of panel in radiation boundary condition to (0.3 0.7 0.3 0.3 0.3) and spectral distribution to (2 1 3 2 1) and the simulation gave no error! The maximum irradiance on panel surface decreased from 880 w/m2 with (2 1) spectral distribution to 550 W/m2 with the current five band spectral distribution.

I have skimmed through Fluent manual and yes you are right, it considers two bands for diffuse light i-e visible and infrared. If openfoam is accepting 5 bands in solar Load then maybe it is dividing the infrared, visible and UV range into 5 equal wavelength intervals with each interval having it's own flux value defined in spectral distribution.

I will look more into it. Thanks!

[Yann]

Hi Guys,

I am far from a radiation expert, but I tried to play around the radiation models in OpenFOAM and bumped into pretty much the same questions than you do.

From what I have seen in the code, I think there not any notion of wavelength anywhere because it is not relevant for the code. You can split your spectrum in N bands, give weights to each band thanks to the spectral distribution parameter, and adjust your material properties giving the emissivity and absorptivity on each band.

The code doesn't need to know whether it is UV, visible or whatever, as long as it has the information to compute the proper spectral distribution and the proper wall behavior.

This is up to the user to know which band he is defining, and correctly adjust the spectral distribution and wall emissivity/absorptivity to be physically correct.

Does that make sense to you?

Cheers,
Yann

[B_R_Khan]

Originally Posted by Yann

Hi Guys,

I am far from a radiation expert, but I tried to play around the radiation models in OpenFOAM and bumped into pretty much the same questions than you do.

From what I have seen in the code, I think there not any notion of wavelength anywhere because it is not relevant for the code. You can split your spectrum in N bands, give weights to each band thanks to the spectral distribution parameter, and adjust your material properties giving the emissivity and absorptivity on each band.

The code doesn't need to know whether it is UV, visible or whatever, as long as it has the information to compute the proper spectral distribution and the proper wall behavior.

This is up to the user to know which band he is defining, and correctly adjust the spectral distribution and wall emissivity/absorptivity to be physically correct.

Does that make sense to you?

Cheers,
Yann

yes you are right, I have come across the same conclusion after playing around different radiation models. So far, I have seen option of defining wavelength range in wideBandAbsoprtionEmission model only but there is no tutorial that uses this model and I had no success in using this model correctly