Hi. I am trying to design a shoring system consisting of diaphragm walls with counterfort walls as stiffeners for a case at hand. I get instability problems and if i don't press escape I receive a dialog saying that the matrix is not positive definitive - critical load may have been exceeded. Could that be due to the location and direction of elastic supports. Or is it because I should be using a different solver. Any ideas or advice.
Structural Design Engineer
MZ & Partners Engineering Consultancy
When getting this message I used to check the full note of Analysis->Calcultion report->full note and make sure that all the load cases are convergent. By doing this, despite the matrix message, could we assume that the calculations, forces, displacements etc are correct? Iguess so. Please confirm.
No, as the analysis may converge for the different state of equilibrium of the structure. You need to determine where you are with the current load level first.
Try to run linear and non-linear buckling analysis (with no P-Delta) for the case or combination which reports this issue and check the value the critical coefficient. Try to reduce the load and check if the original load converges with no warnings displayed and then investigate the deformation of the model and compare it with the one you have for the original load.
How is that posible?
Artur is actually explaining this (correct me Artur I'm misunderstanding).
you have got a structure that is sensitive to the load case you have are studying in your non-linear buckling analysis. It tends to be unequilibrated easely -> critical factor in the linear buckling analysis is close to 1 or maybe <1.
so when performing a non linear analysis on this load case -> cannot converge.
1) check the value of critical coeffcient you have with linear analysis
2) If <1, before running a non linear analsys, reduce the value because you are sure that it will not converge.
3) If you tick p-delta analsys maybe even if the coeff is >1 but close to it, it will also crash because the 2nd order effect will have an effect on the critical coeff (reducing it actually). You can easely the this effect when running a buckling analysis on a column with a litlle lateral load : critical coeff will be smaller if you take into account 2nd order effect.
4) send your model if lost.
The question is, in this scensario, the message of the matrix non positive can be taken as an simple warning and not as an error?
If you will still have doubts please send me model.