I got the error in Viewfactor calculation
Inconsistent number of parents found 4134 (vs. 4164)
I noticed this is a test on nsurf. But it is not completely clear what is checked.
The error appears after changing the mesh density in gmsh.
With another mesh the error does not appear and everything works.
So I am wondering wher I should look for curing this problem.
The full output of this section was:
ViewFactors: Computing view factors for radiation body 1
ViewFactors: Number of surfaces participating in radiation: 4164
ViewFactors: Computing viewfactors...
Combining original boundary elements for shading
The combined set includes 838 line segments (vs. 4164)
Inconsistent number of parents found 4134 (vs. 4164)
ViewFactors: View factors computed in time (s): 39.82 39.95
ViewFactors:
ViewFactors: Viewfactors before manipulation:
ViewFactors: Minimum row sum: 9.895E-01
ViewFactors: Maximum row sum: 1.000E+00
ViewFactors: Number of rowsums below 0.5 is: 14 (out of 4164)
ViewFactors: Normalizaing Factors...
ViewFactors: Normalization iteration 1: 1.073E-08
ViewFactors: Normalization iteration 2: 2.169E+02
ViewFactors: Normalization iteration 3: 3.921E+17
ViewFactors: Normalization iteration 4: 6.504E+21
ViewFactors: Normalization iteration 5: 4.368E+50
ViewFactors: Normalization iteration 6: 1.400E+58
ViewFactors: Normalization iteration 7: 2.137E+70
ViewFactors: Normalization iteration 8: 4.173E+72
ViewFactors: Normalization iteration 9: 4.341E+84
ViewFactors: Normalization iteration10: 1.390E+89
ViewFactors: Normalization iteration11: 1.108+150
ViewFactors: Normalization iteration12: 5.165+159
ViewFactors
-
kevinarden
- Posts: 2310
- Joined: 25 Jan 2019, 01:28
- Antispam: Yes
Re: ViewFactors
Sounds like a meshing issue. Debugging would require posting of your files, geo, mesh, sif etc.
-
raback
- Site Admin
- Posts: 4827
- Joined: 22 Aug 2009, 11:57
- Antispam: Yes
- Location: Espoo, Finland
- Contact:
Re: ViewFactors
Hi
When computing 2D view factors one can combine the shading elements if they lie on the same line. First the code finds the superelements and then checks that each original element can be assigned to them. For some reason in this case not.
It can be that the code has failed somehow. This is just a consistency check. On the other hand it has been there untouched for many years. It would be interesting to see what kind of mesh caused this.
As a background the view factor computation is annoying. You must compute N*N binary integrals. When there is shading this is further multiplied by N. Hence if we can reduce N to M (<<N) the cost of shading computation is greatly reduced.
In 3D octree structure is used which is an alternative strategy to take down the time. Basically the strategies could be even combined.
-Peter
When computing 2D view factors one can combine the shading elements if they lie on the same line. First the code finds the superelements and then checks that each original element can be assigned to them. For some reason in this case not.
It can be that the code has failed somehow. This is just a consistency check. On the other hand it has been there untouched for many years. It would be interesting to see what kind of mesh caused this.
As a background the view factor computation is annoying. You must compute N*N binary integrals. When there is shading this is further multiplied by N. Hence if we can reduce N to M (<<N) the cost of shading computation is greatly reduced.
In 3D octree structure is used which is an alternative strategy to take down the time. Basically the strategies could be even combined.
-Peter
Re: ViewFactors
Hi
I identified the main problem.
It is the crystal/gas surface in a Czochralski growth furnace.
Going from 0.5 mm to 0.2 mm grid distance at boundary gives the problem.
When using
Viewfactor Combine Elements = Logical False
Normalization converges.
The crystal has the shape of the experiment thus it is not just cylindrical.
For the moment this will solve the problem, but may be one should look deeper what is going wrong.
I identified the main problem.
It is the crystal/gas surface in a Czochralski growth furnace.
Going from 0.5 mm to 0.2 mm grid distance at boundary gives the problem.
When using
Viewfactor Combine Elements = Logical False
Normalization converges.
The crystal has the shape of the experiment thus it is not just cylindrical.
For the moment this will solve the problem, but may be one should look deeper what is going wrong.