GiD forum

Hello:

When I use the Geometry | Create | Object | Circle function in GiD, I
obtain an almost closed circle defined by a line and a single point. A
surface is also automatically generated.

I found out that if I remove the surface and split the line in slices (just
like a pizza pie), then the radius of each slice changes. The variations
are actually sufficient to produce spurious results in my FE solvers. So,
obviously, that is a problem.

I found out that if I create my circle using 2 arcs (2 lines and 2 points),
then I can slice these arcs and the numerical errors on the radius become
very small. The same thing for circles generated by rotation extrusion.

Also note that the disc used as the base of the Cylinder object is actually
constituted of two arcs of circle, and does not present the strange
behavior of the single-point-circle.

I want to ask the following:

- Do you agree with my remarks, i.e. the Circle object in GiD is defined in
such a way that it is not a very good representation of a true circle? So
this object should be used carefully if numerical analysis is to be
performed. Maybe you can give me some recommendations about the usage of
that circle?

- Would it be possible to have a new object, a "true circle" that would be
defined with two arcs, just like the base of a cylinder object?

- I believe GiD uses single precision for the definition of objects. Is
this true? Do you see that as a limitation?

Thanks.

Philippe

Hello,

When you create a circle with that function, GiD creates a rational
NURBS line. Its expression, for the unit circle is:

Control points:
1 0.000000,1.000000,0.000000 W=1.000000
2 -1.000000,1.000000,0.000000 W=0.707107
3 -1.000000,0.000000,0.000000 W=1.000000
4 -1.000000,-1.000000,0.000000 W=0.707107
5 0.000000,-1.000000,0.000000 W=1.000000
6 1.000000,-1.000000,0.000000 W=0.707107
7 1.000000,0.000000,0.000000 W=1.000000
8 1.000000,1.000000,0.000000 W=0.707107
9 0.000000,1.000000,0.000000 W=1.000000

Knots:
1 0.000000
2 0.000000
3 0.000000
4 0.250000
5 0.250000
6 0.500000
7 0.500000
8 0.750000
9 0.750000
10 1.000000
11 1.000000
12 1.000000

Following the NURBS theory, this should represent an exact circle.
So,
the mathematical representation of this line should by exactly a circle.
If
you divide this circle into parts, every arc should be converted into
another
NURBS line that should represent exactly that arc. If you have found
that the
representation is not accurate, maybe you have discovered a bug in the
division
code. Please send us the exact example to make it easier to discover the
problem.

GiD internal operations in the preprocessing part are made in double
precision.

About the mesh simmetry in a simmetric geometry, as unstructured mesh
is an
heuristic algorithm, you can never guarantee that you will obtain the
same mesh
although the geometry is simmetric. You cannot even guarantee that you
will obtain
the same mesh on two runs of the mesher in the same geometry. But good
analysis
codes do not depend on the mesh, do they?

Regards,

--
Compass Ing. y Sistemas Dr. Ramon Ribo
http://www.compassis.com ramsan at compassis.com
c/ Manuel Girona, 61 bajos tel. +34 93 204 10 82
08034 Barcelona, Spain fax. +34 93 204 19 09

-----Mensaje original-----
De: gidlist-admin at gatxan.cimne.upc.es
[mailto:gidlist-admin at gatxan.cimne.upc.es] En nombre de Philippe
Bouchilloux
Enviado el: jueves, 19 de junio de 2003 21:47
Para: GiDList
Asunto: [GiDlist] Circles


Hello:

When I use the Geometry | Create | Object | Circle function in GiD, I
obtain an almost closed circle defined by a line and a single point. A
surface is also automatically generated.

I found out that if I remove the surface and split the line in slices
(just
like a pizza pie), then the radius of each slice changes. The
variations
are actually sufficient to produce spurious results in my FE solvers.
So,
obviously, that is a problem.

I found out that if I create my circle using 2 arcs (2 lines and 2
points),
then I can slice these arcs and the numerical errors on the radius
become
very small. The same thing for circles generated by rotation extrusion.

Also note that the disc used as the base of the Cylinder object is
actually
constituted of two arcs of circle, and does not present the strange
behavior of the single-point-circle.

I want to ask the following:

- Do you agree with my remarks, i.e. the Circle object in GiD is defined
in
such a way that it is not a very good representation of a true circle?
So
this object should be used carefully if numerical analysis is to be
performed. Maybe you can give me some recommendations about the usage
of
that circle?

- Would it be possible to have a new object, a "true circle" that would
be
defined with two arcs, just like the base of a cylinder object?

- I believe GiD uses single precision for the definition of objects. Is

this true? Do you see that as a limitation?

Thanks.

Philippe


_______________________________________________
GiDlist mailing list
GiDlist at gid.cimne.upc.es
http://gid.cimne.upc.es/mailman/listinfo/gidlist

Hello Philippe.

As explain Ramon, it exists in GiD two mathematical expresions for a circle:
The classical circle (with a center 2D, a radius and a transformation matrix
to locate in 3D position), and the NURBS expression.

The first expression set the radius explicit, and for the second expression
the radius is implicit, and can have a great numerical error.

The numerical error can increase if your model have a size too big, too
small, or is translated far of the origin.

Can you send us the sample (what is the location, radius, number of
divisions, etc)?

To increase the precision can try to scale and/or translate your entities.
You can also create an arc with the classical expression using
geometry-create-arc instead to create with the circle object or by
rotation.

In other hand, the cylinder uses also NURB expression. The precission
problem must also appear in this case!

Enrique

----- Original Message -----
From: "Philippe Bouchilloux" pb at magsoft-flux.com
To: "GiDList" gidlist at gatxan.cimne.upc.es
Sent: Thursday, June 19, 2003 9:46 PM
Subject: [GiDlist] Circles


Hello:

When I use the Geometry | Create | Object | Circle function in GiD, I
obtain an almost closed circle defined by a line and a single point. A
surface is also automatically generated.

I found out that if I remove the surface and split the line in slices
(just
like a pizza pie), then the radius of each slice changes. The variations
are actually sufficient to produce spurious results in my FE solvers. So,
obviously, that is a problem.

I found out that if I create my circle using 2 arcs (2 lines and 2
points),
then I can slice these arcs and the numerical errors on the radius become
very small. The same thing for circles generated by rotation extrusion.

Also note that the disc used as the base of the Cylinder object is
actually
constituted of two arcs of circle, and does not present the strange
behavior of the single-point-circle.

I want to ask the following:

- Do you agree with my remarks, i.e. the Circle object in GiD is defined
in
such a way that it is not a very good representation of a true circle? So
this object should be used carefully if numerical analysis is to be
performed. Maybe you can give me some recommendations about the usage of
that circle?

- Would it be possible to have a new object, a "true circle" that would be
defined with two arcs, just like the base of a cylinder object?

- I believe GiD uses single precision for the definition of objects. Is
this true? Do you see that as a limitation?

Thanks.

Philippe


_______________________________________________
GiDlist mailing list
GiDlist at gid.cimne.upc.es
http://gid.cimne.upc.es/mailman/listinfo/gidlist