Conditions on Newton-Raphson iterations

Question

Conditions on Newton-Raphson iterations

DorianDepriester opened this issue 4 years ago · 3 comments

Hello there,
It appears that the number of non-linear iterations for each increment is only driven by its maximum value (parameter Maximum non linear iterations)

Indeed, changing the values for Absolute nonLinear solver tolerance or Relative nonLinear solver tolerance seems to have no effect: the number of iterations remains 4 (if Maximum non linear iterations = 4).

As far as I understand, the computed increment must be updated as soon as one of the aforementioned criteria is reached. Am I misinterpreting something?

Answer 1 · 2021-01-22T16:05:03.000Z

Hi @DorianDepriester ,

In the current version of the code, we only practice the constant number of nonlinear iterations and do not use "Absolute nonLinear solver tolerance" or "Relative nonLinear solver tolerance" to change the number of iterations. For our applications so far, since we compute the precise algorithmic tangent modulus, the code converge using Maximum non linear iterations=4. Using higher number is unnecessary and typically doesn't help the convergence. In the elastic part, we might be able to use less number, however, they only include a small portion of computation.

Thanks
Reza

Answer 2 · 2021-01-22T16:15:44.000Z

Also, if you have any application you want to model in PRISMS-Plasticity, you can contact me at yaghoobi@umich.edu to discuss that specific application in more details.

Thanks
Reza

Answer 3 · 2021-01-23T10:42:26.000Z

Ok, thank you Reza for your detailed answer!
Currently, my application is generic, for I try to set up a simple workflow for performing CPFEM from real microstructures imaged by EBSD. It is based on a software of mine, named MTEX2Gmsh.
If you are interested, check out the procedure I have proposed.

I will contact you by email if I have further questions/issues.

Best regards,
Dorian