Hello:
I am having problem with sphere. More specifically, GiD does not seem to
mesh things correctly, especially when hex elements are used.
I am attaching a model of 1/8th of a sphere with hex meshing. First, you
will notice that the structured meshing produces elements with inverted
curvatures. Second, if you consider the nodes generated on the sphere
surface, you will see that there is a non-negligible error on the distance
from the center.
The center of the sphere is (0,0,0) and the Radius is 10. This mesh
(quadratic) was generated under 7.2, but I have made attempts with GiD
betas and I don't seem to be getting better results. Node coordinates are
exported out of GiD in d.dddddE+ee format (standard).
I am attaching a text file with the computation of the radius for the
surface nodes. If I haven't made an error on this calculation, I find that
the maximum error on the radius is 0.031865351 !
The problem is that my solver needs to fit the spherical surface. The
fitting algorithm fails persistently due to these meshing errors.
Is there a way to improve this situation?
Thanks.
Philippe
-------------- next part --------------
Computation of the error on the distance of a surface node to the center (0,0,0).
Left column = Node number (from GiD)
Right column = Distance - Radius (Radius = 10)
Node error
1 0
2 5.788E-08
4 1.74244E-06
5 1.1235E-06
9 4.87018E-06
12 4.03409E-06
15 1.56118E-06
16 1.56118E-06
23 8.2583E-07
25 2.94669E-06
28 3.39598E-06
29 5.03592E-06
39 5.3721E-07
44 2.66444E-06
45 2.66444E-06
53 3.13826E-06
62 3.46338E-06
65 3.20855E-07
66 9.9175E-08
73 5.92291E-06
92 5.41672E-06
97 3.80948E-06
98 5.16338E-06
99 3.49688E-06
114 8.5421E-07
122 1.2709E-06
123 7.3383E-07
134 2.8643E-07
153 1.57817E-06
160 3.7346E-06
161 3.3221E-06
179 3.36324E-06
189 5.38601E-06
231 3.75334E-06
236 2.62218E-06
238 4.5307E-06
246 4.19896E-06
264 1.33633E-06
266 1.11222E-06
278 2.23777E-06
281 5.96825E-07
314 2.0765E-08
321 3.55625E-07
331 1.23115E-06
333 4.40753E-06
384 7.48525E-07
417 3.53521E-06
418 1.91414E-06
423 6.9049E-07
459 3.97658E-06
460 3.09449E-06
467 2.64394E-06
468 2.75539E-06
481 1.20073E-06
482 5.64413E-06
505 4.445E-07
536 0.000174737
539 4.12893E-06
540 3.95549E-06
566 5.95177E-06
570 2.10538E-06
588 2.30921E-06
621 0.000582394
622 0.008610147
640 7.95795E-06
641 1.73993E-06
650 0.00904243
669 3.56722E-06
703 0.001059687
705 3.93985E-06
706 0.01565972
747 3.32943E-06
793 2.50179E-06
794 5.33616E-06
812 4.5307E-06
819 0.001488804
820 0.021020989
833 0.018067566
835 6.02995E-06
849 0.014385084
884 5.31205E-06
898 0.008617308
909 1.04294E-06
926 0.024817165
927 0.001778326
984 1.47958E-06
988 0.01565483
1017 2.69756E-06
1018 2.76163E-06
1037 0.028602539
1039 0.027032994
1042 0.001881271
1079 0.016129001
1082 0.021021231
1129 4.22075E-06
1135 9.4019E-07
1147 0.027696764
1150 0.001778566
1175 0.024817949
1177 0.00903606
1260 0.031865351
1278 0.026825865
1279 0.001484689
1280 3.49688E-06
1281 1.54724E-06
1283 1.38411E-06
1297 0.027030601
1298 1.61754E-06
1311 1.1434E-07
1327 0.014311209
1331 2.35895E-07
1337 4.86081E-06
1351 0.000178523
1371 0.01438985
1377 3.39163E-06
1395 0.027697995
1404 0.024440948
1405 0.001055571
1419 5.6781E-07
1427 0.000582264
1471 3.23906E-06
1495 0.028075697
1496 0.001059931
1499 0.026821605
1523 0.000586403
1524 0.020547052
1534 1.75087E-06
1549 0.016134075
1564 0.001490698
1566 2.76163E-06
1569 0.008930413
1589 0.024442761
1590 3.97832E-06
1637 0.000171669
1638 0.001774412
1640 0.015176076
1679 0.009014889
1680 3.18174E-06
1724 0.020547743
1727 0.001879506
1729 0.017409898
1732 0.014311022
1754 2.74752E-06
1755 2.66444E-06
1759 0.008288539
1761 5.23911E-06
1762 5.1619E-07
1798 0.001773849
1802 2.29866E-06
1804 0.014330536
1810 0.015177936
1813 2.29537E-06
1834 3.27974E-06
1835 0.001486359
1836 2.50179E-06
1841 4.87204E-08
1851 0.008584039
1852 0.008941069
1854 0.008290026
1858 0.01606557
1862 0.001056042
1863 3.3992E-06
1868 2.81039E-06
1885 0.015571095
1892 1.98777E-06
1900 0.020880363
1908 0.000588517
1909 1.16502E-06
1916 1.8474E-07
1918 0.024628868
1919 0.01425007
1927 0.001014271
1928 0.000175813
1929 1.11806E-06
1931 0.026799615
1970 0.027459306
1973 5.19076E-06
1974 1.73993E-06
1978 0.008912509
1980 0.009385359
1983 0.026603845
1985 1.56118E-06
1986 1.56118E-06
1988 1.09186E-06
1991 2.44963E-06
1992 4.46278E-06
1993 0.017885566
1998 3.36864E-06
2001 0.024243155
2002 0.008276151
2004 1.53575E-06
2010 1.51906E-06
2011 0.004850785
2012 0.028230642
2013 0.020397895
2016 1.4268E-06
2018 5.61345E-07
2020 0.003185721
2022 1.38006E-06
2039 0.015089694
2041 0.00956225
2042 2.45924E-06
2043 0.031417191
2048 2.30591E-06
2051 0.008255785
2052 0.009464118
2055 3.37838E-06
2056 2.0639E-07
2059 0.02770827
2060 4.12893E-06
2061 1.16502E-06
2063 0.008529538
2064 0.008676645
2066 0.015399793
2067 0.020611327
2069 0.017235339
2070 0.003575688
2071 0.024268851
2103 0.026392171
2104 0.027016315
2105 2.81039E-06
2106 0.026179821
2107 1.1235E-06
2108 1.74244E-06
2109 4.16994E-06
2110 0.023885526
2111 0.008122877
2112 4.12014E-06
2113 0.010109646
2114 0.020135342
2115 5.85446E-06
2116 0.007934663
2117 0.01492474
2118 3.37824E-06
2119 0.003932679
2120 0.008203084
2121 4.84109E-06
2122 0.000492789
2123 1.20073E-06
2124 4.87204E-08
2125 0.008738978
2142 0.013704195
2143 0.015278995
2144 9.77035E-07
2145 0.013619947
2146 0.007157673
2147 0.008629767
2148 0.0069493
2149 1.3023E-07
2150 0.001878256
2151 0.00148703
2152 0.001487627
2153 0.000580251
2154 0.000573847
2155 4.0409E-06
2156 3.67818E-06
2157 4.74285E-06
2158 4.87018E-06
2159 3.49278E-06
2160 3.49688E-06
2161 4.85182E-06
2162 0
2163 7.76265E-07
2164 3.97658E-06
2165 3.66747E-06
2166 2.94669E-06
2167 3.54065E-07
2168 9.9175E-08
2169 2.74752E-06
2170 3.68825E-06
2171 8.2999E-07
2172 0
2173 2.62218E-06
2174 5.16338E-06
2175 1.56118E-06
2176 1.1235E-06
2177 1.23115E-06
2178 5.788E-08
2179 1.57817E-06
2180 2.66444E-06
-------------- next part --------------
A non-text attachment was scrubbed...
Name: SphereMeshWithHex.zip
Type: application/zip
Size: 454330 bytes
Desc: not available
Url : http://listas.cimne.upc.edu/pipermail/gidlist/attachments/20031105/6051c6af/attachment.zip
[GiDlist] Problem with spheres
Moderator: GiD Team
-
Enrique Escolano
[GiDlist] Problem with spheres
Hello Philippe
The problem is not the mesher, in your sample, some geometric surfaces have not exactly spheric shape.
The two big surfaces of each sphere layer are spheric, but the three more small are only an approximation,
probably are constructed from your boundary lines, not by rotation. Then the surface constructed from the
boundary is only an approximation to the desired spheric shape.
To see this, use Geometry-Edit-Edit Nurb Surface..., to see the Nurbs control polygon points
This is a image, selecting an approximate surface, It' s a untrimmed surface.
The neighbourg big surface instead is really spheric, it's a 1/8 of sphere, trimmed with your boundary lines.
Can delete the bad surfaces (first deleting volumes), and create new spheric surfaces, using as base shape, the shape of the 1/8 of sphere of the big surface, and trimming with the appropiate curves.
Geometry-Create-NurbSurface-Trimmed
select the big surface, as base shape, and select the four boundary lines to trim it.
Then obtain a new trimmed surface (in blue) with this control polygon :
Repeat this operation for the other approximate surfaces (in green), and re-build the volumes (can use AutomaticStruct).
Note: Must reassign your volume mesh information to obtain structured hexahedras.
I attach also the modified GiD model.
Regards
Enrique Escolano
----- Original Message -----
From: "Philippe Bouchilloux" pb at magsoft-flux.com
To: "GiDList" gidlist at gatxan.cimne.upc.es
Sent: Wednesday, November 05, 2003 4:21 PM
Subject: [GiDlist] Problem with spheres
Hello:
I am having problem with sphere. More specifically, GiD does not seem to
mesh things correctly, especially when hex elements are used.
I am attaching a model of 1/8th of a sphere with hex meshing. First, you
will notice that the structured meshing produces elements with inverted
curvatures. Second, if you consider the nodes generated on the sphere
surface, you will see that there is a non-negligible error on the distance
from the center.
The center of the sphere is (0,0,0) and the Radius is 10. This mesh
(quadratic) was generated under 7.2, but I have made attempts with GiD
betas and I don't seem to be getting better results. Node coordinates are
exported out of GiD in d.dddddE+ee format (standard).
I am attaching a text file with the computation of the radius for the
surface nodes. If I haven't made an error on this calculation, I find that
the maximum error on the radius is 0.031865351 !
The problem is that my solver needs to fit the spherical surface. The
fitting algorithm fails persistently due to these meshing errors.
Is there a way to improve this situation?
Thanks.
Philippe
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://listas.cimne.upc.edu/pipermail/gidlist/attachments/20031107/55b5f2cb/attachment.htm
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: image/png
Size: 46716 bytes
Desc: not available
Url : http://listas.cimne.upc.edu/pipermail/gidlist/attachments/20031107/55b5f2cb/attachment.png
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: image/png
Size: 64331 bytes
Desc: not available
Url : http://listas.cimne.upc.edu/pipermail/gidlist/attachments/20031107/55b5f2cb/attachment-0001.png
-------------- next part --------------
A non-text attachment was scrubbed...
Name: SphereMeshWithHex_Modified.gid.zip
Type: application/x-zip-compressed
Size: 342844 bytes
Desc: not available
Url : http://listas.cimne.upc.edu/pipermail/gidlist/attachments/20031107/55b5f2cb/attachment.bin
The problem is not the mesher, in your sample, some geometric surfaces have not exactly spheric shape.
The two big surfaces of each sphere layer are spheric, but the three more small are only an approximation,
probably are constructed from your boundary lines, not by rotation. Then the surface constructed from the
boundary is only an approximation to the desired spheric shape.
To see this, use Geometry-Edit-Edit Nurb Surface..., to see the Nurbs control polygon points
This is a image, selecting an approximate surface, It' s a untrimmed surface.
The neighbourg big surface instead is really spheric, it's a 1/8 of sphere, trimmed with your boundary lines.
Can delete the bad surfaces (first deleting volumes), and create new spheric surfaces, using as base shape, the shape of the 1/8 of sphere of the big surface, and trimming with the appropiate curves.
Geometry-Create-NurbSurface-Trimmed
select the big surface, as base shape, and select the four boundary lines to trim it.
Then obtain a new trimmed surface (in blue) with this control polygon :
Repeat this operation for the other approximate surfaces (in green), and re-build the volumes (can use AutomaticStruct).
Note: Must reassign your volume mesh information to obtain structured hexahedras.
I attach also the modified GiD model.
Regards
Enrique Escolano
----- Original Message -----
From: "Philippe Bouchilloux" pb at magsoft-flux.com
To: "GiDList" gidlist at gatxan.cimne.upc.es
Sent: Wednesday, November 05, 2003 4:21 PM
Subject: [GiDlist] Problem with spheres
Hello:
I am having problem with sphere. More specifically, GiD does not seem to
mesh things correctly, especially when hex elements are used.
I am attaching a model of 1/8th of a sphere with hex meshing. First, you
will notice that the structured meshing produces elements with inverted
curvatures. Second, if you consider the nodes generated on the sphere
surface, you will see that there is a non-negligible error on the distance
from the center.
The center of the sphere is (0,0,0) and the Radius is 10. This mesh
(quadratic) was generated under 7.2, but I have made attempts with GiD
betas and I don't seem to be getting better results. Node coordinates are
exported out of GiD in d.dddddE+ee format (standard).
I am attaching a text file with the computation of the radius for the
surface nodes. If I haven't made an error on this calculation, I find that
the maximum error on the radius is 0.031865351 !
The problem is that my solver needs to fit the spherical surface. The
fitting algorithm fails persistently due to these meshing errors.
Is there a way to improve this situation?
Thanks.
Philippe
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://listas.cimne.upc.edu/pipermail/gidlist/attachments/20031107/55b5f2cb/attachment.htm
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: image/png
Size: 46716 bytes
Desc: not available
Url : http://listas.cimne.upc.edu/pipermail/gidlist/attachments/20031107/55b5f2cb/attachment.png
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: image/png
Size: 64331 bytes
Desc: not available
Url : http://listas.cimne.upc.edu/pipermail/gidlist/attachments/20031107/55b5f2cb/attachment-0001.png
-------------- next part --------------
A non-text attachment was scrubbed...
Name: SphereMeshWithHex_Modified.gid.zip
Type: application/x-zip-compressed
Size: 342844 bytes
Desc: not available
Url : http://listas.cimne.upc.edu/pipermail/gidlist/attachments/20031107/55b5f2cb/attachment.bin
-
Philippe Bouchilloux
[GiDlist] Problem with spheres
Hello Enrique:
Thanks for your comments. I really thought that my surfaces were initially
trimmed from a sphere. Maybe I messed up. Anyway, thanks for pointing out
the Edit Nurb Surface command, which I was not using to verify these
surfaces. It is really handy.
Actually, wouldn't it be nice if we could impose a radius of curvature of a
Nurb surface, without necessarily having to go through the trimming
exercise? Maybe something to consider in your wish list?
So I will redo the sphere and check again.
However, your email does not explain why the mesh is deformed, even in the
spherical regions:
27f944.jpg
Any idea how to control this type of things and make sure that the
structured mesh is more uniform?
Thanks again for all the help.
Philippe
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://listas.cimne.upc.edu/pipermail/gidlist/attachments/20031107/76095375/attachment.htm
-------------- next part --------------
A non-text attachment was scrubbed...
Name: 27f944.jpg
Type: image/jpeg
Size: 126661 bytes
Desc: not available
Url : http://listas.cimne.upc.edu/pipermail/gidlist/attachments/20031107/76095375/attachment.jpg
Thanks for your comments. I really thought that my surfaces were initially
trimmed from a sphere. Maybe I messed up. Anyway, thanks for pointing out
the Edit Nurb Surface command, which I was not using to verify these
surfaces. It is really handy.
Actually, wouldn't it be nice if we could impose a radius of curvature of a
Nurb surface, without necessarily having to go through the trimming
exercise? Maybe something to consider in your wish list?
So I will redo the sphere and check again.
However, your email does not explain why the mesh is deformed, even in the
spherical regions:
27f944.jpg
Any idea how to control this type of things and make sure that the
structured mesh is more uniform?
Thanks again for all the help.
Philippe
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://listas.cimne.upc.edu/pipermail/gidlist/attachments/20031107/76095375/attachment.htm
-------------- next part --------------
A non-text attachment was scrubbed...
Name: 27f944.jpg
Type: image/jpeg
Size: 126661 bytes
Desc: not available
Url : http://listas.cimne.upc.edu/pipermail/gidlist/attachments/20031107/76095375/attachment.jpg
