Dear all,

I am working on Creep modeling. I started to review some established approach using user defined *Creep subroutine, with really good results. But now I am moving on tensor damaging models, and I am working on a *UMAT subroutine to model multi axial creep. This is my first experience with UMAT, so I decided to write a Norton law in multi axial form, just to practice with the code.

To test the code I rely on some benchmarking example, like axi-symmetric uniaxial specimen, notched bar and for 3D case a simply supported beam. I am modelling the creep behaviour with two different step, one elastic( static) and a second one with creep equations(Static). The Jacobian is elastic for all the procedure.

The problem that I am facing is the following one, after few increment ( which correspond to 10 20 hours ) I see strange changing in the stress field, S Mises for example. In the simplest case, the uniaxial specimen, the stress have to be always constant, instead after 17 increment it start to raise to no sense values. The code appear correct, also because it is really simple, Im trying to implement some error check on creep strain calculation, but such strategy is not giving good results.

Does anyone had any experience in this, or simple has any suggestion?