UMAT/VUMAT for non-linear elasticity

[PWH]

Hello all,
I'm trying to write a non-linear elastic model UMAT. I used a linear elastic example as a template, since the only difference here is that the components of the elasticity matrix are strain-dependent rather than constant. They are easily obtained from the energy by

d2w/(du_ij)(du_kl).

In the linear elastic example, the Jacobian is just set equal to the elasticity matrix, but I suspect something more sophisticated needs to be done here?

I thought I'd try a VUMAT, as suggested elsewhere on this forum, to avoid the issue altogether- but at first glance the strain tensor is not even passed in to the VUMAT routine, so I'm not sure how to get those values.

Any suggestions on how to proceed? I've been looking for examples to work from, but what I find is often far more complicated than what I'm doing here.

[Jorgen]

You can probably estimate the Jacobian terms by the same equations as for the linear elastic case if you simply replace the two elastic constants with the actual strain-dependent values.

In a VUMAT you can either use the strain increments (which are provided by Abaqus) or you can calculate the strain from the deformation gradient (which is also provided).

- Jorgen

[PWH]

You can probably estimate the Jacobian terms by the same equations as for the linear elastic case if you simply replace the two elastic constants with the actual strain-dependent values.

That's exactly what I tried. I got solutions that depended on the increment size, and were either wrong or diverged. So I figured that approach was problematic.

The model can alternately be rearranged such that the moduli are stress (rather than strain) dependent, so I've done that with the VUMAT, but presently am trying to diagnose a floating point error that occurs apparently before any of the code executes :roll: