Collapse load for an in-plane loaded panel 2

[tuss]

Due to symmetry I have modelled only a quarter of a sandwich panel, loaded in plane in one direction. To apply the load I have used a prescribed displacement. When I then want to find the corresponding load, I have summarised the reaction forces in every node on the loaded edge. However, because I have only modelled a quarter of the panel, to achieve the buckling load for the entire panel, I should double the achieved load, shouldn't I?

 

[rb1957]

if you modelled the unloaded edge as an axis of symmetry, then the total load applied to the panel (to get the results you have) is double the model applied load.

but why would your arbitary(?) applied load be the buckling load of the panel ?

 

[SWComposites]

tuss - the minimum buckling mode may not be symmetric wrt the centerlines of the panel - think of a buckling mode that has 2 half waves in the length or width of the panel - in this case the mode shape displacement is not symmetric. Therefore, you must model the entire plate if you are doing a buckling analysis.

 

[rb1957]

SWComposites

wouldn't that be the 2nd buckling mode, and so less critical than the 1st (assuming that the middle of the panel can deflect) ?

i'm more curious that tuss thinks he has found the buckling load in the first place ??

 

[gwolf]

This is how you do it:

a) Unless you are talking high speed loads, the first, half wavelength, mode is the only one two consider.

b) Use one of the combined mode shape/buckling proceedures in your FE code to calculate the load for you then look at the text output file to see what it calculates.

c) Run a modal analysis, deform your original mesh using (probably, but not definitely) the first mode and scale the max deformation to be the max skin waviness. Then apply an overly large load in a NONLINEAR static analysis and see what load fraction the analysis bombs out at - this will be the buckling load.

d) Manually put a small dimple in your mesh, again within surface waviness limits, and do a nonlinear analysis as in c), the load will be very similar to that in c)

Regards,

gwolf.

 

[SWComposites]

a) the first buckling mode has only one halfwave length for a ~square plate. For a plate with a 2:1 aspect ratio the first mode has 2 half waves in the long direction, etc.

b) I'm assuming that tuss ran an eigenvalue buckling analysis; in this case you take the eigenvalue times the applied loads to get the theoretical buckling load.

 

[tuss]

Thank you for your answers!
However, I accidently wrote that I wanted the buckling load for the panel which was not correct, I just wanted the applied load corresponding to the prescribed displacement. The buckling load (for the first eigenmode) I achieved from a buckling analysis by multiplying the load with the eigenvalue. The load I'm now interested in is in the nonlinear postbuckling analysis. Then I just got confused concerning if I should double the load to get the corresponding load for the entire panel, which I intuitively thought.

 

[rb1957]

tuss,

you've modelled a 1/4 panel and applied an edge displacement, and you're trying to analyze the full-size panel for post-buckling behaviour ... DON'T.

the (presumably shear) buckling behaviour of the 1/4 panel will be very different to the behaviour of the full panel.

in the linear range, i think the load on the full size panel is 1x the 1/4 panel ... 2x from the width is obvious, but i think you have a 0.5x from the length, consider strain in the length direction, the full size panel will have 1/2 the strain as the 1/4 panel ('cause it'll have twice the length).

for the full size panel you better off modelling the entire panel

good luck