Heat Radiation

Numerical methods and mathematical models of Elmer
kevinarden
Posts: 2237
Joined: 25 Jan 2019, 01:28
Antispam: Yes

Heat Radiation

Post by kevinarden »

I did not create this file Elmer did, and then it tried to read past the end of it creating an EOF error
system.zip
(117.78 KiB) Downloaded 218 times
Is the error in the case.sif?

MAIN: =============================================================
MAIN:
MAIN:
MAIN: -------------------------------------
MAIN: Reading Model: case.sif
LoadInputFile: Scanning input file: case.sif
LoadInputFile: Loading input file: case.sif
Loading user function library: [HeatSolve]...[HeatSolver_Init0]
LoadMesh: Base mesh name: ./.
LoadMesh: Scaling coordinates: 2.540E-02 2.540E-02 2.540E-02
LoadMesh: Elapsed REAL time: 0.0270 (s)
MAIN: -------------------------------------
AddVtuOutputSolverHack: Adding ResultOutputSolver to write VTU output in file: case
Loading user function library: [HeatSolve]...[HeatSolver_Init]
Loading user function library: [HeatSolve]...[HeatSolver_bulk]
Loading user function library: [HeatSolve]...[HeatSolver]
RadiationFactors: ----------------------------------------------------
RadiationFactors: Computing radiation factors for heat transfer
RadiationFactors: ----------------------------------------------------
RadiationFactors: Total number of Radiation Surfaces 6262 out of 6262
At line 565 of file /build/elmerfem-csc-eg-6VhCCM/elmerfem-csc-eg-8.4-0ppa0/fem/src/RadiationFactors.F90 (unit = 10, file = '././ViewFactors.dat')
Fortran runtime error: End of file

Error termination. Backtrace:
#0 0x7f5a9e7aa2da in ???
#1 0x7f5a9e7aaec5 in ???
#2 0x7f5a9e7ab68d in ???
#3 0x7f5a9e921a33 in ???
#4 0x7f5a9e91a9c4 in ???
#5 0x7f5a9e91c0f9 in ???
#6 0x7f5a9ef1d0d9 in radiationfactors_
at /build/elmerfem-csc-eg-6VhCCM/elmerfem-csc-eg-8.4-0ppa0/fem/src/RadiationFactors.F90:565
#7 0x7f5a9edbfaa8 in __mainutils_MOD_addequationbasics
at /build/elmerfem-csc-eg-6VhCCM/elmerfem-csc-eg-8.4-0ppa0/fem/src/MainUtils.F90:1441
#8 0x7f5a9efaff51 in addsolvers
at /build/elmerfem-csc-eg-6VhCCM/elmerfem-csc-eg-8.4-0ppa0/fem/src/ElmerSolver.F90:1080
#9 0x7f5a9efaff51 in elmersolver_
at /build/elmerfem-csc-eg-6VhCCM/elmerfem-csc-eg-8.4-0ppa0/fem/src/ElmerSolver.F90:427
#10 0x557f34e7317b in solver
at /build/elmerfem-csc-eg-6VhCCM/elmerfem-csc-eg-8.4-0ppa0/fem/src/Solver.F90:69
#11 0x557f34e72ece in main
at /build/elmerfem-csc-eg-6VhCCM/elmerfem-csc-eg-8.4-0ppa0/fem/src/Solver.F90:34
raback
Site Admin
Posts: 4812
Joined: 22 Aug 2009, 11:57
Antispam: Yes
Location: Espoo, Finland
Contact:

Re: Heat Radiation

Post by raback »

Hi Kevin,

There is already problem in computing the view factors. You can try to compute them separately by calling the binary "ViewFactors" as you would "ElmerSolver". This creates the ViewFactors.dat in the mesh directory. This basically is pure geometric information that tells the fraction of space angle each boundary element see each other.

The computation will fail typically for cases where the normal points to the wrong direction. This is bit like the FSI case: when we have an interface boundary the direction of the normal is ambiguous. There are two ways to give it: either specify "Radiation Target Body", or my recommendation, specify the emissivity in the material section (instead of the BC section). Then the code will automatically detect the body which is radiating and assume the outer normal to point out of it.

Now, in your case there seemed to be just two materials, both of them metals. Perhaps you're looking at all for "diffuse gray" radiation but "idealized" instead. For the diffuse the setup should be such that the boundaries should see itself. If you think of an barrel, then external radiation boundaries can only be "idelized" whereas the internal BCs may be "diffuse gray" because the geometry makes the temperatures interdependent.

The radiating boundary may be partially open "Radiation Boundary Open = Logical True". Then the user may specify the fraction of omitted space angle to be idealized. Without this keyword it is assumed that the view factors always sum up to 2PI gradians (half space) and this information is used to normalize the view factors to be more accurate.

-Peter
kevinarden
Posts: 2237
Joined: 25 Jan 2019, 01:28
Antispam: Yes

Re: Heat Radiation

Post by kevinarden »

Makes sense, thanks. Helpful, i will try again.

Kevin.
kevinarden
Posts: 2237
Joined: 25 Jan 2019, 01:28
Antispam: Yes

Re: Heat Radiation

Post by kevinarden »

I simplified the model to 2D. With and without the internal air modeled. With the internal air modeled I can see the heat transfer to the air. The out box temperature above the heat source box reaches 413.4. Without the air in the box modeled, the answer is the same with or without radiation boundaries on and it is 521.6. If I specify radiation boundary target the solver ends with an error calculating view factors.
air.png
(95.76 KiB) Not downloaded yet
no_air.png
(30.84 KiB) Not downloaded yet
2dmeshnoair.zip
(201.03 KiB) Downloaded 200 times
raback
Site Admin
Posts: 4812
Joined: 22 Aug 2009, 11:57
Antispam: Yes
Location: Espoo, Finland
Contact:

Re: Heat Radiation

Post by raback »

Hi

Works well for me. There is no check that the ViewFactors.dat is of right size so remove that. Also the ViewFactor computation may require that the "ELMERSOLVER_STARTINFO" points to the correct file. ElmerSolver uses just a system command to call the viewfactor computation.

-Peter
Attachments
radiation.sif
(3.48 KiB) Downloaded 241 times
kevinarden
Posts: 2237
Joined: 25 Jan 2019, 01:28
Antispam: Yes

Re: Heat Radiation

Post by kevinarden »

It does work well, thanks.

Kevin
kevinarden
Posts: 2237
Joined: 25 Jan 2019, 01:28
Antispam: Yes

Re: Heat Radiation

Post by kevinarden »

Did not work in 3D.
3dmesh.zip
(206.28 KiB) Downloaded 217 times
ViewFactors: Number of nodes in mesh: 484574
ViewFactors: Computing view factors for radiation body 1
ViewFactors: Number of surfaces participating in radiation: 83264
ERROR:: Viewfactors: Memory allocation error. Aborting
STOP 1
raback
Site Admin
Posts: 4812
Joined: 22 Aug 2009, 11:57
Antispam: Yes
Location: Espoo, Finland
Contact:

Re: Heat Radiation

Post by raback »

Hi

Note that you need a dense matrix for the view factors with size 83264 x 83264 i.e. about 6.93e9 entries. Even if you had this much memory in your system the code will fail because max value of 32 bit integer is 2.14e9. Hence the quadratic scaling makes the radiation model of Elmer poorly scalable. This is inherent to methods that require computation of view factors.

You could also do this explicitely using Monte-Carlo methods, for example. Hoever, this will have a toll on the convergence and accuracy as well. However, it could scale better with problem size.

Personally I find the "discrete ordinates" method interesting as it would be deterministic, offer scaling and be more easily applicable in parallel as well.

Unfortunately neither of the alternative methods have been implemented.

-Peter
kevinarden
Posts: 2237
Joined: 25 Jan 2019, 01:28
Antispam: Yes

Re: Heat Radiation

Post by kevinarden »

Ok, good to know, thanks for the info.
kevinarden
Posts: 2237
Joined: 25 Jan 2019, 01:28
Antispam: Yes

Re: Heat Radiation

Post by kevinarden »

On this case the windows version, not updated in a while, actually experience cooling (lower temps on the boundary compared to initial) but on the updated Linux version it did not run.
3dmesh.zip
(206.28 KiB) Downloaded 194 times
Post Reply