XANSYS Message: 22596 [Go back to message list]
[bookmark on del.icio.us]
No rating yet
Rate item:

Subject: Re: STRUC - Coupling Solid and shells? Or PCG solver....
Author: Mark Rodamaker
Date: 2001-04-10 12:58:00

Todd,
Higher order elements do not have physical stiffness at their corner
nodes. Sometimes the stiffness is actually negative depending on whether the
element is hex or tet shaped. Read in the following commands to see what I
mean. The total reaction in the Z direction is 1 but the distribution is a
little nonintuitive.

/PREP7
et,1,95
mp,ex,1,1e7
BLOCK,,1,,1,,1,
/view,,1,2,3
/vup,,z
/pnum,node,1
/type,,0
esiz,1
vmes,1
eplo
nsel,s,loc,x,0
d,all,ux
nsel,s,loc,y,0
d,all,uy
nsel,s,loc,z,0
d,all,uz
alls
nsel,s,loc,z,1
sf,all,pres,1
alls
fini
/solu
solve
fini
/post1
set,1
prrs

Mark Rodamaker
MCR Associates
Sunnyvale,California
voice: 408-736-1636

----- Original Message -----
To:
Sent: Tuesday, April 10, 2001 9:27 AM

> >Remember to attach beams only at
> >midside nodes if the solids are higher order (95's or 92's).

> Mark, I have never heard this advice before. Can you elaborate on this?

> Thanks Mark.

> Todd Meidinger
> Bobcat/IR
> Design Engineer/Analyst

> "Mark
> Rodamaker" To:
xansys
> > m> Subject: Re:
[xansys] STRUC - Coupling
> 04/10/01 11:35 Solid and
shells? Or PCG solver....
> AM

> Adding endplates is fine but applyting your loads to 1 node of the shell
> endplate is questionable. It might be better to put a node in the center
of
> the tube and connect beams or pipes to this node. The cluster of beams
> attaching at a node is structurally stable. Remember to attach beams only
> at
> midside nodes if the solids are higher order (95's or 92's). Make sure you
> have enough constraints that the system is constrained against free body
> translations and rotations. For well conditioned problems, PCG is the
> fastest solver available and will give identical answers to the other
> solvers but it will get there faster. However, if the problem is poorly
> conditioned , it will have a hard time. A problem can be poorly
conditioned
> if it is poorly constrained. It appears you have this problem by your
> description. The other cause of poor conditioning is materials with very
> different moduli ( you have only 1 material ) or really severe aspect
> ratios
> ( I doubt yours are very poor). So, it is quite likely you have not
> constrained your model quite enough.

> Mark Rodamaker
> MCR Associates
> Sunnyvale,California
> voice: 408-736-1636

> ----- Original Message -----
> From:
> To:
> Sent: Tuesday, April 10, 2001 4:13 AM
> Subject: [xansys] STRUC - Coupling Solid and shells? Or PCG solver....

> > Hi there,

> > I am about to do a linear static analysis of a tubular joint (i.e. a
> > pipe with a smaller pipe connected to it at an angle). It is a part
> > of an offshore platform.

> > I have previously done a lot of these with all shell elements. But
> > this model needs to be modelled in solid brick elements.

> > My problem is with the boundary conditions of the chord (the
> > bigger, 'through' pipe), and the loading of the brace (the smaller
> > pipe). I need to apply these at the ends of the pipes, in the centre
> > line, where there is normally a hole. For instance, I need to
> > simulate the chord being simply supported, so I need to 'release' the
> > rotational DOFs, which can't be done by simply fixing all the nodes
> > on the pipe circumference (in the end).

> > Usually, when doing a similar shell model, I use endplates with
> > increased stiffness, and I simply apply my loads and BC's to the
> > centre node of each endplate.

> > But since solid brick elements only have the 3 translational DOFs at
> > each node, this can't be done (I need to apply a moment to the free
> > end of the brace, and I need to fix the model for rotation about the
> > chord axis).

> > So I figured that I could still use a shell endplate, which should
> > allow me to apply rotations to the centre node, which would then
> > be 'converted' to translations on the entire solid element edge of
> > the relevant pipe.

> > Is this assumption correct, and have others used this method?

> > My problem is, I get some really weird results.... When I load the
> > brace axially, for instance, the whole model seems to 'expand'
> > (radially to the chord) and rotate some 45 degrees, even though I
> > have fixed the chord ends for axial rotations! Moment loading on the
> > brace won't even converge (still a linear analysis).

> > I am forced to use the PCG solver, as the frontal and sparse solvers
> > produce TRI files that exceed the 8Gb limit and closes the process.
> > I have no experience with this solver, so I am not sure if my problem
> > lies with the PCG solver or if there is something wrong with my basic
> > assumptions regarding the use of a stiff shell endplate, as explained
> > above.

> > What I am basically looking for here is:

> > A confirmation (or not) about the validity of using shell endplates
> > to connect and transfer forces/moments to the edges of a solid
> > element pipe.

> > Any advice available on the PCG solver. The description for the PCG
> > solver fits my model very nicely (huge, linear solid element model
> > etc.)

> > But is there anything I should be careful about with this solver? I
> > am using the default options right now...

> > I hope any of this makes sense :)

> > Any advice will be much appreciated!

> > Sincerely,

> > Uffe Eriksen,
> > Denmark

> > ude@r...

Posts possibly associated with message #22596AuthorDateScore
22571STRUC - Coupling Solid and shells? Or PCG solver....Uffe Dal Eriksen2001/04/10 
22591Re: STRUC - Coupling Solid and shells? Or PCG solver....Mark Rodamaker2001/04/10 
22593Re: STRUC - Coupling Solid and shells? Or PCG solver....Todd Meidinger2001/04/10 
22594Re: STRUC - Coupling Solid and shells? Or PCG solver....Rahul Hiwale2001/04/10 
22596Re: STRUC - Coupling Solid and shells? Or PCG solver....Mark Rodamaker2001/04/10 
22597Re: STRUC - Coupling Solid and shells? Or PCG solver....Uffe Dal Eriksen2001/04/10 
22704Re: STRUC - Coupling Solid and shells? Or PCG solver....Ing. Luk s Vas¡Šek LukasVa@2001/04/13