nonlinear buckling analyse

[SKJoe]

Hi,

I do a set of nonlinear buckling analyses via APDL. I use shell181 and mpc 184. In some analyses the reason for termination was unconverged solution :

R E S T A R T I N F O R M A T I O N

REASON FOR TERMINATION. . . . . . . . . .UNCONVERGED SOLUTION

but for another ones :

R E S T A R T I N F O R M A T I O N

REASON FOR TERMINATION. . . . . . . . . .ERROR IN ELEMENT FORMULATION

...what does this mean ? Are results in this case valid (they seem to be valid) ? Did I set up something wrong ?

Regards,
Lubo

 

[cbrn]

Hi,
I don't know what you mean by "they seem to be valid", since normally both error messages are typical of a "wrong" solution.
Bear in mind that, when you are doing several sub-steps in sequence and you request results for every sub-step, the displayed solution is the last converged one at the last equilibrium iteration. It is identified by the fictitious sub-step number "999999". This may or not contain info about the solution failure.
"Unconverged solution" means that, at a certain point of the analysis, it was not possible to find an equilibrated force field to match the compatible displacement field. "Error in element formulation" comes when, for some reason, at least one element violates its mathematical formulation (e.g. the Gaussian nodes have come outside the domain defined by the "physical" nodes), because of what happened to the displacement field: that means that no compatible displacement field exists (or seems to exist).

I'm about sure that you already know that, for a non-linear buckling analysis, the first non-converged sub-step is the one for which the critical load has been exceeded, so this kind of termination is quite normal.

Hope this helps...

Regards

 

[SKJoe]

Hi,

Thank you for your reply cbrn ! I hoped that the answer will be as you wrote. I set these values for solution (the end of the previous load step is time = 6):

NCNV , 2
LNSRCH, 1
TIME , 46
AUTOTS, ON
NSUBST, 160,2000,160
NEQIT , 40

solution started with (46-6)/160 = 0.25 time increment, the last increment was 40/2000 = 0.02. In this increment 40 iteration was performed and then the solution was terminated. But I have just one more question. I realised that after the first bisection the next time increment was 0.1125 instead 0.25/2 = 0.1250. Why ? I thought that "bisection" means to divide by two. Is the reason relative to another used setings (LNSRCH, 1, AUTOTS, ON) ?

Regards,
Lubo

 

[cbrn]

Hi,
I don't remember how Line Search exactly works, but it has to do with secants, so depending on the slope / convexity of the curve, the bissection can lead to a timestep which is not half the previous one. This is the advantage of the Line Search wrt the "traditional" timestep division.

Regards