deflection of plate on elastic support (zigzag?)

rhoka

hi All,

i have interested in feature of elastic support, took some comparisons for small problems.
MECWAY look convergences with mesh refinement to reach deflection targets were's about -0.92in maximum except in result of formation in a whole mesh.

i just wondering why MECWAY get unexpected result, is this right? when deflection formations are zigzag (upright/downright) at each node mesh.

another question is, how can i apply translation spring at each node of mesh? should i do it manually? this required to represent sliding resistant were i'ts about 10% or higher compare to perpendicular springs.

attached input files and screenshot.

thanx u

Victor

These zig-zag displacements are characteristic of hourglassing

http://www.quora.com/What-is-hourglass-effect-in-finite-element-analysis

It can happen with quad4 shells in Mecway if they only have point constraints or loads.

In this case, you can avoid it by applying the force to some element faces instead of a single node.

rhoka

from your links: hourglassing is another problem come from reduced integration, i'm being learn for these. previously i know little about shearlocking, and at the past i daily use SAP were element are linear and formulation based on incompatible were these phenomenon didn't shown.

is MECWAY have brief documentation about element formulation and recomendation, i did not found in manuals.

when problem in shearlocking, many recomendation using quadratic element. MECWAY provide these element. attached screenshot after i rerun analysis using quadratic,

btw, why in MECWAY for 3D shell an element linear tri3 can not combine with quad4?

Victor

Not much documentation on the formulations sorry, other than p48-49 of the manual. They're pretty much the same as you'll find in many textbooks. The shells are flat plate elements combined with membrane elements. Quad4 and quad8 use reduced integration for shear stiffness to avoid shear locking.

There's no tri3 thick shell simply because it wasn't implemented. There used to be a tri3 thin shell but it turned out to have some consistent error so I removed that long ago.

rhoka

thanks for answering, glads you're using standard element formula were available in many textbook so i can learn from them.

no linear tri3 is not a big problems since quadratic (tri6 and quad8) element are available.

the problems only in nodes, try to simple case (attached)
4.618nodes linear element vs 13.928 nodes quadratic element, both has the same number of element. it's about four times larger in number of nodes.

but, this can be tweaking in meshing to creates half of number element. luckily quadratic mesh can catch curved geometry.

VMH

Rhoka, what software are u using to create that shell/surface model like your attached pic. Ive been looking for one similar to FreeCAD that we can modify geometry after creating them (parametric modeling or the one similar to Ansys or Abaqus). Is it expensive? Thanks

rhoka

btw, i'm using FreeCAD as well and export as *.stp

works with surfaces modeling not solid.

it's quiet easy using boolean operation

VMH

Thanks for the info. I knew about FreeCAD but couldn't edit the geometry after creating the surfaces for geometry like you one attached (changing ODs or in-plane or out-plane angles).

rhoka

i didn't try parametric modeling, but from the documents: yes, it should be possible using scripting

VMH

I haven't tried using scripting. Thanks

VMH

rhoka, I figured out how to model shells/faces in FreeCAD that are fully editable later without using scripting for parametric study/design changes. Example attached (trunnion).

VMH

Attachments.

VMH

Attachments.

rhoka

hi VMH,

first i didn't know if specific CAD discussion is appropriate here, since MECWAY only linked with STEP/IGES were it's standard format and it can be produce by many CAD programs.

parametric programming in CAD means: all object dependencies will update consistently if one part of object changed. an example are in pipe connection, if main tube diameter are change then the secondary tube were connected will search for surface intersection automatically.

so, in the beginning i talk about parametric modeling for surface complex object is only possible trough scripting in FreeCAD.

VMH, i seek a solution for this case and found a simple way even this is not really parametric. i work in solid modeling modules, split a component and save in each *.stp format. then you can combine each part (assembling) with MECWAY capabilities and meshing surface only, removing unnecessary surface later.

attached a screenshot in CAD modeling were it has two parts, one for main object and another are for hole intersection filler.

about removing unnecesary surface, i post to reply in latest your video thread.

VMH

I know what parametric modeling is and it can be done for shell/surface model (I attached the file already. If you update one dimension, everything that is associated with it will automatically update).

I feel it is applicable to Mecway forum because some may be interested in the capability in modeling all shell/surface model to be imported into Mecway for complex geometry and the able to change them as needed.

I'm just letting you know I figured out a way to model shell/surface model that can be fully edited as we discussed a few days ago. Disregard if not interested.

J_Marc

Hi,
This is in regards to the original plate over elastic foundation post. I have attached a 2m square plate with 1m square centre surface load of 80te. I was just curious about the problem above. Everything was good "von mises stress" etc had a symmetry to the contour until I viewed the "moment per length" contours. Is this caused due to the nature of the meshing or hour-glassing talked about?

Victor

Hello J_Marc

The asymmetry is because the elements' local coordinates (U,V,W) aren't all aligned the same way and the moment per length is shown in the element coordinate system. You can see their orientations by turning on the Show element axes button.

To align U,V with Z,X:
1) Select all the elements
2) Right click on one of them and go to "Element properties"
3) Set Orientation to "U axis"
4) Enter 0,0,1 for X,Y,Z. This makes it parallel to the global Z axis.
5) Press OK

For isotropic materials, this won't make any difference to the solution except to change the coordinate system used to display moments and anything else using element coordinates.

J_Marc

Thanks