Plate problem- simple reasoning required

[ateestructural]

This is a very simple problem using which I'm trying to validate my finite element code

I have two plates Plate 1 and Plate 2.

Plate 1

In this plate I connected elements ABEG and BCDF

Nodes E and F are co-incident

I connected E and F by rigid elements. I did NOT connect E and F for rotation X but for all other degrees of freedom with rigid elements

See attached pdf

Now, I have another plate

PLATE 2

In plate 2 I do NOT have co-incident nodes

I connected ABEG and BCDE. I released the rotation at E in plate ABEF

I use same supports and loads in both plate 1 and plate 2

Do plate 1 and plate 2 denote the same problems physically?

Because, I have coded in my finite element program releases for plates, and thus doing the validation

I get a difference ofby a factor of 10 in the displacement at hinged location in plates 1 and 2.

I have a feeling that the 2 problems are not physically same.

I do not the reason.

Can anyone express their point of view?

 

[JGard1985]

Hi does your FEA code have just 2 elements?

Jeff
Pipe Stress Analysis Engineer

http://www.xceed-eng.com

 

[rb1957]

why can't a model have two elements ? doing a patch test (to see how the elements work) you use a single element.

another day in paradise, or is paradise one day closer ?

 

[JGard1985]

It can, just wanted to make sure that the FEM mesh had enough integration points to capture the deflected shape and wasn't locked

Anyway IMO, based on statements made about boundary conditions and loads I think it is the same problem



Jeff
Pipe Stress Analysis Engineer

http://www.xceed-eng.com

 

[rb1957]

maybe the problem is within "my coded finite element program" ?

try running the two models in NASTRAN ??

another day in paradise, or is paradise one day closer ?

 

[whitwas]

Which direction is X?

If X is parallel to AB in the diagram, then, no, the two problems aren't the same. The one without the X-rotation connected allows the two plates to shear differently at that node. If you put a moment load about X on node F, then you should see the 2nd element having shear deformation while the first element doesn't. The shear deformation would appear as rotation about X without any corresponding displacement.

If X is normal to the plane and the elements have no drilling DOF, then they're equivalent. Although the nodes E and F could have different rotation angles caused by some other effect which doesn't matter because those DOFs aren't connected to the plate elements.

If X is parallel to AG then they're obviously different.

 

[ateestructural]

Yes, X is parallel to AB

You said

If X is parallel to AB in the diagram, then, no, the two problems aren't the same. The one without the X-rotation connected allows the two plates to shear differently at that node. If you put a moment load about X on node F, then you should see the 2nd element having shear deformation while the first element doesn't. The shear deformation would appear as rotation about X without any corresponding displacement.

Will it not be the same if I have no coincident nodes but a hinge (to release X rotation) at the location of the coincident node?

 

[ateestructural]

It can, just wanted to make sure that the FEM mesh had enough integration points to capture the deflected shape and wasn't locked

Anyway IMO, based on statements made about boundary conditions and loads I think it is the same problem

Could it be the mesh sensitivity is different for two problems?

 

[whitwas]

Sorry, I misunderstood the problem so please ignore my previous reply. Now plates 1 and 2 look like they should behave the same to me.

Have you already found that it works correctly for more intuitive cases like released Y-rotation on nodes B and E? If so, then perhaps the bug is that you've released the rotational DOF from bending but not from shearing? Does the element stiffness matrix have zeros in all of the node E row and column when it's released?

 

[ateestructural]

Have you already found that it works correctly for more intuitive cases like released Y-rotation on nodes B and E?

My Y rotation on nodes B and E was fixed. It otherwise reported instability error if ever I left the Y rotation free

then perhaps the bug is that you've released the rotational DOF from bending but not from shearing

I don't get that. I just released the X rotation and made the condensed stiffness matrix. Yes, it had 0's in the corresponding rows and columns

Does the element stiffness matrix have zeros in all of the node E row and column when it's released?

Yes, as I said above