XANSYS Message: 22670 [Go back to message list]
[bookmark on del.icio.us]
No rating yet
Rate item:

Subject: Re: RBE3 constraint equation error
Author: Mark Flanagan
Date: 2001-04-12 14:24:00

Not having experience with this particular ANSYS element..but if it is based
on the NASTRAN RBE3... In Nastran, the RBE3 acts as a shortcut to a
multi-point constraint set. The degrees of freedom are tied directly (in a
weighted average sense), regardless of whether or not the DOF are actually
pointing in the same direction. So it is really important to have the
master and slave nodes using the same analysis cordinate system (read: nodal
coordinate system) unless you go through some baroque mathematical
contortions writing your constraint equations explicitly. NASTRAN also
recommends only using translational DOFs for slave nodes for their element,
which you appear to be satisfying, as you are using solids...

-Mark Flanagan
Senior Analyst
Dynamic Engineering, Inc.
757 873 1344 x-387

-----Original Message-----
Sent: Thursday, April 12, 2001 1:45 PM
To: xansys

Thanks Mark.

I am using the MASS21 element with 6 DOFs, so that part should be ok.

I have, however, rotated the master node, in order to be able to
apply my constraints correctly. Rotation about the chord axis, for
instance, which is not a global axis.

There is an option with the mass element, whether the element
coordinate system is parallel to the global coord. system initially,
or the nodal coordinate system.

What would the effect be from these factors, with regard to the RBE3?

If this is what causes the error, would it help if I rotated all my
slave nodes equal to the master node?

Now I think about it, I did not rotate my master node in my test
example, so I might have missed a parameter there.

Great! Something new to try out :)

Thanks again.

Uffe Eriksen,
Rambøll, Esbjerg,
Denmark

--- In xansys@y..., "Tate, Mark HS-SD" wrote:
> You should check that you are using a 3D mass element.
> Also make sure that nodal coordinates are not rotated.

> Mark C. Tate
> Principal Analysis Engineer
> Hamilton Sundstrand - A United Technologies Company
> mark.tate@h...

> -----Original Message-----
> From: udeansys@y... [mailto:udeansys@y...]
> Sent: Thursday, April 12, 2001 8:16 AM
> To: xansys@y...
> Subject: [xansys] Re: RBE3 constraint equation error

> Stan,

> Uhm... yes, I am applying standard DOF constraints (using the D
> command) to the master node, which is in the centre of the pipe
end.
> That's the whole point of what I am trying to do: place a simple
> support (as used in basic beam structures) on a 3D model of a
> pipe/beam. The RBE3 should then transfer rotations as well as
> translations to the nodes on the edge of the pipe, in the form of
> translations only, if you know what I mean.

> This is exactly what I did in my test case, which was a solid
> cylinder meshed with 20-node bricks. I selected all the nodes at
one
> end of the cylinder, including a centre node, which I had meshed
with
> a mass element (my master node).
> Then I used the command: RBE3,, ALL, ALL

> I did this in both ends of the cylinder. This should then enable me
> to place my DOF constraints on the master nodes only, so I can
create
> a pinned connection or whatever I want, right? The test worked
> beautifully with the PCG solver...

> But when I do the exact same thing on my real model, I get that
weird
> error, and ANSYS claims that my 'batch job has finished
successfully'
> and closes down completely (after an OK prompt). When I list the
CEs,
> they seem ok, in the sense that they 'convert' my rotational DOFs
to
> translations along the edge.

> If my model was ill-conditioned, wouldn't I 'just' get convergence
> problems? I don't even get to the real solution phase here...

> Thanks again :)

> Uffe Eriksen,
> Rambøll, Esbjerg,
> Denmark

> --- In xansys@y..., "Schwantes, Stan" wrote:
> > Uffe
> > Are you applying restraints to any master node? I get nervous
about
> > applying restraints to nodes involved in any type of constraint
> equation.
> > (Look at help on CE command). If some DOF's at the master node are
> > eliminated by the restraint, I'm not sure that ANSYS and the PCG
> solver
> > handle that correctly. Did you restrain a master node of RBE3 in
> your
> > successful small model test case?

> > In your non-midside node version the RBE3 worked correctly so
that
> may not
> > be the whole problem. It may be more of an ill-conditioned
matrix
> thing,
> > and the PCG solver does not handle ill-conditioned problems as
well
> as the
> > sparse solver does.

> > To avoid restraints on the master node, try a very short elastic
> beam tied
> > to the master node. The other end of the short beam is on the
pipe
> > centerline and has the restraints applied to it. This may give
you
> a better
> > conditioned system.
> > Hope this helps.

> > Stan Schwantes 952/931-5544
> > Alliant Techsystems, Inc
> > Defense Systems Group E-mail: stan_schwantes@a...

> > !!!!!!!!!!!!!!!!!!!!!!!!

> > -----Original Message-----
> > From: udeansys@y... [mailto:udeansys@y...]
> > Sent: Thursday, April 12, 2001 9:25 AM
> > To: xansys@y...
> > Subject: [xansys] Re: RBE3 constraint equation error

> > Thanks Stan!

> > 1. I am using ANSYS 5.6.2. We haven't had the opportunity to
> install
> > ANSYS 5.7 yet.

> > 2. I am doing a standard linear-elastic analysis, in order to
find
> > stress concentration factors in the joint, for later use in our
in-
> > house frame software.

> > 3. I will try the CECHECK command as soon as possible!

> > 4. You are right, I do not want that. We have specifically been
> asked
> > to simulate a simple support for the chord ends, i.e. just fixed
> for
> > translations in the centre of the pipe end, and with a 'roller'
in
> > one end.

> > 4.2. Actually, my RBE3 web at the brace end seems to work fine.
> It's
> > the RBE3's in the chord ends that are acting up, as far as I can
> tell.

> > I ran the analysis with the brace RBE3 only, and all the nodes in
> the
> > chord end fixed for translations, just as a test. That worked
fine.

> > But when I make RBE3's at the chord ends, the fatal error comes
> right
> > after 'preparing elements for solver'. Very weird...

> > 5. I would rather not have to use a beam web, but in the end I
> might
> > have to try it after all...

> > The funny thing is, I have done this type of analysis in all
shell
> > models heaps of times, with stiff endplates for all pipes. In
> theory,
> > I can't see why a shell endplate wouldn't work with solid element
> > pipes as well, but the results I get (with the PCG solver at
least)
> > are plain crazy! The whole thing rotating and expanding(!), even
> > though I have fixed it for axial rotations... I suppose it's a
> quirk
> > of the iterative PCG solver.

> > Thanks anyway, Stan, and keep'em coming! :)

> > Best regards,

> > Uffe Eriksen,
> > Rambøll, Esbjerg,
> > Denmark

Posts possibly associated with message #22670AuthorDateScore
22646RBE3 constraint equation errorUffe Dal Eriksen2001/04/12 
22651Re: RBE3 constraint equation errorStan Schwantes2001/04/12 
22653Re: RBE3 constraint equation errorUffe Dal Eriksen2001/04/12 
22655Re: RBE3 constraint equation errorStan Schwantes2001/04/12 
22662Re: RBE3 constraint equation errorUffe Dal Eriksen2001/04/12 
22666Re: RBE3 constraint equation errorMark Tate2001/04/120
22668Re: RBE3 constraint equation errorUffe Dal Eriksen2001/04/12 
22670Re: RBE3 constraint equation errorMark Flanagan2001/04/12 
22695Re: RBE3 constraint equation errorUffe Dal Eriksen2001/04/13