Hi,
I attached my mesh and .xml file, as I have a SIGFPE error after the first iteration and I don't really know why ...
I think I have a problem in my mesh, I launched 'check mesh', but I don't know where we can see if the mesh is bad...
Thanks for your help.
Regards
Gauthier
SIGFPE error after 1 iteration
Forum rules
Please read the forum usage recommendations before posting.
Please read the forum usage recommendations before posting.
-
gdambrin
SIGFPE error after 1 iteration
- Attachments
-
- NACA_0012.xml
- (6.6 KiB) Downloaded 237 times
-
- NACA_0012_2-med-tar.gz
- (5.22 MiB) Downloaded 228 times
-
James McNaughton
Re: SIGFPE error after 1 iteration
check_mesh says
Warning
=======
2 elements of type quad4 are mis-oriented
Warning
=======
25 elements of type hexa8 are mis-oriented
Although doesn't give locations of these. Viewing the PREPROCESSOR.CASE and having a zoom into the trailing edge you can see some issues though. There look to be some very skewed (possibly overlapping) cells at the trailing edge on the lower surface. Using structured near the profile is a good idea but the skewing of cells may lead to errors. Around the leading edge your mesh is good but at the trailing edge try something like in this tutorial: http://courses.cit.cornell.edu/fluent/airfoil/step2.htm
Hope that helps,
James
Warning
=======
2 elements of type quad4 are mis-oriented
Warning
=======
25 elements of type hexa8 are mis-oriented
Although doesn't give locations of these. Viewing the PREPROCESSOR.CASE and having a zoom into the trailing edge you can see some issues though. There look to be some very skewed (possibly overlapping) cells at the trailing edge on the lower surface. Using structured near the profile is a good idea but the skewing of cells may lead to errors. Around the leading edge your mesh is good but at the trailing edge try something like in this tutorial: http://courses.cit.cornell.edu/fluent/airfoil/step2.htm
Hope that helps,
James
-
Yvan Fournier
Re: SIGFPE error after 1 iteration
Hello,
I'll add that when cells have highly skewed faces, the orientation test may return false positives. Actually, when only a few cells seem mis-oriented, you can suspect this is the case. When cells are mis-oriented due to bad local numbering, you would expect all cells of a given type to be badly oriented (such as prisms in EnSight format written by some meshing tools who seem to have misread the format documentation), or a whole set of cells should be badly oriented (for example when part of the mesh is built using a symmetry but the meshing tool does not adjust the local numbering correctly).
We should improve the message for badly oriented cells to say "seem badly oriented", but in any case, when you have such cells, the mesh quality is locally bad.
A SIGFPE occurs when a floating-point exception occurred, usually either a division by zero (ie. bad initialization), or operations leading to values not representable by double-precision floating-point values, which typically occurs when huge values are encountered when a calculation starts diverging. So I would suspect divergence here.
I would either of suggest 2 tests: split warped faces (for example faces with a warping angle above 10 degrees). This may be activated either in the GUI or in usini1.f90. use relaxation of pressure increase.
Best regards,
Yvan
I'll add that when cells have highly skewed faces, the orientation test may return false positives. Actually, when only a few cells seem mis-oriented, you can suspect this is the case. When cells are mis-oriented due to bad local numbering, you would expect all cells of a given type to be badly oriented (such as prisms in EnSight format written by some meshing tools who seem to have misread the format documentation), or a whole set of cells should be badly oriented (for example when part of the mesh is built using a symmetry but the meshing tool does not adjust the local numbering correctly).
We should improve the message for badly oriented cells to say "seem badly oriented", but in any case, when you have such cells, the mesh quality is locally bad.
A SIGFPE occurs when a floating-point exception occurred, usually either a division by zero (ie. bad initialization), or operations leading to values not representable by double-precision floating-point values, which typically occurs when huge values are encountered when a calculation starts diverging. So I would suspect divergence here.
I would either of suggest 2 tests: split warped faces (for example faces with a warping angle above 10 degrees). This may be activated either in the GUI or in usini1.f90. use relaxation of pressure increase.
Best regards,
Yvan