negative buckling eigenvalue (Abaqus)

[cie-engineer]

Hello everyone !

I try to calculate buckle procedure for 2-girder model with intermediate stringers. As a result i obtained the negative eigenvalue (-3.8).
For taking into account the connection rotational stiffness between stringer and girder i used the "join" connector + spring elemts with N*m/rad stiffness. Is it correct ?
Same time the results from other software (Autodesk Robot) i have got completely different result.
What i have done wrong on abaqus ?

 

[FEA way]

You can find some tips regarding this in the documentation chapter Analysis —> Analysis Procedures —> Static Stress/Displacement Analysis —> Eigenvalue Buckling Prediction (paragraph "Understanding Negative Eigenvalues").

 

[cie-engineer]

Некоторые советы по этому поводу можно найти в главе документации Анализ —> Процедуры анализа —> Анализ статического напряжения/смещения —> Прогнозирование потери устойчивости по собственным значениям (пункт «Понимание отрицательных собственных значений»).

Thank you a lot ! It was very usefull to read that documentation part. Apart from that, on anower forum they answered me to try change coupling type from "Distributing" to "Kinematic", and it helped
But I haven't realized why the first one remains different, i still thinking about it:

 

[GVDR]

Distributing coupling is the equivalent of RBE3 in NASTRAN it is a connection made to transmit loads. The constraint forces the central node (your connection point) displacement to be the average of all the master points. You can thus have a relative displacement between connected points as long as its average at the connection point is null (think about ovalization mode of a cylinder). That may the reason for your negative eigenvalue depending on your setup.

Kinematic coupling is the equivalent of RBE2 in NASTRAN or other rigid elements, it is made to ensure displacement continuity. The constraint forces the all connected points to have the same displacement as the central connection node. This latter is here master. No relative displacement is possible for all connected nodes so you get a stiffer behavior, but no risk of an ill-posed problem as long as you do not generate over-constraints.

For your connections (welded ?), I doubt distributing coupling makes sense.