Negative eigenvalues 2

[Senad83]

Hey,

I'm doing buckling analysis with 5 eigenvalues. One of the is negative, but my analysis still completes with out errors or warnings. Does anybody know how I can avoid the negative eigenvalues?

Senad

 

[corus]

Never heard of negative oens before. Are you they're just not almost zero, but a little negative? - just to be tehcnical.
Zero values will indicate the body is free to translate in a particular direction.

corus

 

[Senad83]

No, they are negative. I know of some reasons with rigid body motion, but the suprising thing is that the analysis completes.

Senad

 

[johnhors]

Buckling analysis can and will produce both positive and negative eigenvalues. They are both vaild "mathematical" solutions for the problem, just that the negative ones have no real relevance to the physical world.

 

[bassmanjax]

Negative eigenvalue just means that the load to cause the buckle is of the opposite sign to the applied load. So if I applied a reference load of 1 MPa and the first eigenvalue was -0.5, then the load required to trigger the buckling mode is (1 x -0.5) = -0.5 MPa.

You may well get spurious buckling modes with eigenvalue extraction, I know I've seen them. After all, the eigenvalues are just the mathematical solution to the problem, they may not necessarily be physically possible. It's up to you, the analyst, to determine what is significant and what isn't.

Having said that, it's probably worth double-checking your constraints and loads anyway.

Regards

Martin Stokes CEng MIMechE

 

[bassmanjax]

You all type wwaayyyyy faster than me ...;-)

Martin Stokes CEng MIMechE

 

[johnhors]

Bassmanjax,

Sorry for typing faster !

It's possible to set up a buckling analysis without an applied load, I've done this when writing beam analysis software using a "geometric stiffness" matrix (in place of the mass matrix for a natural freqency analysis) in conjunction with the standard stiffness matrix, and using the same eigen extraction routine as used in natural frequencies. This analysis produced probably as many negative eigenvalues as positive ones.

 

[Senad83]

Thanks a lot.

 

[dkeatley]

If you did the analysis correctly; to my understanding a negative eigenvalue means the structure buckles for that applied load. If I am wrong someone please correct me.

 

[TGS4]

dkeatley - you are wrong. See johnhors' comment above.

 

[DLTguy]

What type of elements are using ? Thin Shell Elements?

Regards,
DLT

 

[Senad83]

I'm using 8-noded shell elements.

 

[cad1amma]

hi!

I'm doing a buckling for non linear analysis for two differents shell structures; I have obtained negative values for both cases, and for all the eigen values.

I know this is possible, but how can I to know which structure is stronger?

I mean, the first eige value for one is -1,54 and for the other -2,34.

Thank very much in advance,
cad1

 

[bassmanjax]

If your reference load is the same in each case, then the structure with an eigenvalue of -2.34 will buckle at a higher load. The negative sign just means that the buckle load is in the opposite direction to the reference load.

Section 6.2.3 in the v6.7 Users Manual covers this quite comprehensively.



Martin Stokes CEng MIMechE

 

[bebme]

What if you have negative eigenvalues in a static contact analysis problem? I get this as one of the few warnings in my impact model and the problem does not finish because of too many increment attempts. Would this be a major problem for the solver finding a solution? Is there no way of getting rid of these warnings?

 

[bassmanjax]

If you're modelling an impact problem, you would be better advised to look at using Explicit rather than Standard.

Regards

Martin Stokes CEng MIMechE

 

[arbor]

hi bassmanjax,

I am interested in positive eigenvalues only in my analysis. However the analysis hasn't been able to give me the first positive eigenvalue after many negative values. I don't know how long it needs to run to find the positive value. Is there anyway to get positive eigenvalues only?

Thank you,

Arbor

 

[johnhors]

Arbor,

This doesn't sound right, I've never heard of this happening before! Could you attach a "small" input file that behaves this way?

 

[johnhors]

Is your applied load, putting the structure into tension rather than compression? If yes, try reversing it.

 

[arbor]

johnhors,

I am trying to model the buckling of a pressure vessel head knuckle under internal pressure (there will be tangential compressive stress in the knuckle due to internal pressure). However, the model kept giving me negative values for the shell and head buckling which would be due to external pressure. I haven't been able to obtain the first positive number due to the real load yet. So it would be nice for me if there is a way that ABAQUS only finds the positive value for the real applied load.

Arbor