I have written a creep model in UMAT. The formulations are used for small displacements. I have implemented a NEWTON-RAPHSON (NR) method in UMAT to update the stress. Here are my questions:
- The problem is that for each increment a few iterations are done. However in each iteration, total strain increment, dstran, becomes larger and the same time increment, dtime, is passed on to UMAT, whereby the NR method does not converge. I think dstran and dtime should decrease.
- Also the starting point for the iteration is the converged values of the last increment instead of last iteration which is different from what I have learned in my master studies.
- I am running the code with NLGEOM=OFF. I have not turned that on because my formulations are not for large displacements. What does switching NLGEOM on do? Could that be the cause of the previous problems?
- And the last
question, could the tolerance for an iteration method be less than machine precision?!! I think the answer is no.
Thank you in advance for your time.