Hi,
I am trying to model plain strain biaxial compression test for dense sand with ABAQUS Standard, I have implemented a critical state soil model in forms of UMAT. Well, as expected for dense sand, encounter numerical difficulties and convergence issues when I run the analysis.
I have attached the input file for the reference. Biaxial soil sample is 0.25mm wide and 0.5mm in height. It is pin supported in left bottom node and roller supported in all other bottom nodes. I have specified initial stresses (isotropic) in the initial step and confining pressure in the second step. In the final step I applied a 0.2 displacement to top nodes. At the same time, I have made four elements in lower right corner weaker than other elements to trigger localisation. I tried with both CPE4 and CPE8 elements. It seems linear elements are better for convergence.
I would appreciate if anyone can give a suggestion to me on how to increase the convergence (at least to see the initial softening). I tried automatic stabilisation, damping, solution controls, adaptive meshing etc. But nothing gave me encouraging results.
Is Riks method or dynamic implicit good for this kind of problems?
Should I make NLGEOM on or not? How it affect the results?
Are there any regularisation available in ABAQUS?
I appreciate any advice.
Thanks in advance,
Hansini
NLGEOM=on is probably required.
Dymamic, analysis=quasi-static may help with the convergence.
Dont use riks method.
Find the yield strain or displacement
Then in initial step define BC of fixed edge and in step1 define displacement for forced edge equal to yield strain.
So now use general static method for step1
Whit this by 100 to 200 inc,u could find ultimate comp strength
Thank you both. I apologise for the late reply.
When I try NLGEOM on with my UMAT, it does not even converge a single step, but with out NLGEOM it goes at least several increments.
I get warnings of negative eigen values. I can not decide whether the error is in UMAT or in input file. I am using a softening material model.
If at all it fails to converge, you may have to loosen the convergence criteria. Numerically, it is always difficult to get smooth convergence when you are using higher Phi values. Such problems doesnot appear in C-soil.
But, loosening the convergence criteria or by introducing automatic stabilization factor types should be used by referring to the manual and by checking that these alterations have not much of influence on the final results.
- Badges
- Science topic