fit_single_crystal_parameters demo isn't converging

[Gbowker]

The demo runs successfully, producing all the correct outputs, but the predicted stress-strain curve never converges on the experimental curve no matter how many iterations:

All simulations:

Final simulation:

As you can see the simulated curves are not approaching the experimental data at all. Therefore the parameters are not fitted.

Please use the demo workflow from here and my notebook from here

Thanks,

Guy

[aplowman]

There is a bug in the way the damping is applied in the formable code here, it should be np.diag(np.diag(jac_prod)) instead of np.diag(jac_prod). I have re-run the demo workflow with that corrected and the first iteration is considerably better, as below. However, it then fails because the slip resistance goes immediately negative. So we need to add constraints to the optimisation, which I will try to do when I have some time.

[Gbowker]

How do you mean add restraints? So I can get started on this?

[aplowman]

Fitting parameter bounds; in this example, the slip resistances need to be positive. Yes, you are welcome to make a start on it!