Hi, I am trying to solve a simple 1d visco-plasticity problem which is already solved using Abaqus here (chapter 2): Lele, S. P. (2008). On a class of strain gradient plasticity theories: formulation and numerical implementation (Doctoral dissertation, Massachusetts Institute of Technology). I’m …

Thanks Nate. It definitely helped. I have an exponent m, which is taken as 0.02. In the gradient it becomes negative (m - 1), which leads to division by zero, thus to nan. Just to check I changed it to 1.02, and the nan issue was resolved. But I have to solve for m=0.02. So, I gave different initi…

Consider Numerical continuation - Wikipedia to get to your desired value of m. To print the number of DoFs, if V is your function space: print(V.dofmap.index_map.size_global * V.dofmap.index_map_bs)

Thanks Nate. I’ll have a look at “Numerical continuation”. But I wonder how Abaqus solves it without resorting to such techniques. I think my mistakes lie somewhere else. You provided the functions for dofs in the function space; I was asking about my split functions ud and γp. I think problem lies…

Inserting into your code at the appropriate point: def count_dofs(V): print(V.dofmap.index_map.size_global * V.dofmap.index_map_bs) count_dofs(ME) count_dofs(ME.sub(0).collapse()[0]) count_dofs(ME.sub(1).collapse()[0]) Gives me the expected 4002 2001 2001

Getting residual "nan" from the start

variational formulation

nate October 2, 2023, 3:20pm 2

Perhaps Default absolute tolerance and relative tolerance - #4 by nate can help.

Topic		Replies	Views
Residual is nan General dolfinx	2	110	March 6, 2025
Newton solver converge initially, then diverge and blow up Legacy DOLFIN	8	641	January 15, 2024
Problem with residual variational formulation	3	786	July 30, 2020
Stop NonlinearVariationalSolver when reaching nan residual Legacy DOLFIN	0	83	June 5, 2025
NewtonSolver residuals go to zero, meaningless solution dolfinx	6	516	August 8, 2023

Getting residual "nan" from the start

Related topics