Initial conditions for use with the Jacobian debugger

[Sina-av]

Hi all!

I want to check the Jacobian of a Moose app that I am writing with the help of the Jacobian debugger file, analyze_jacobian.py. For the simulations, I am using the following variables:

[Variables]
  [disp_x] []
  [disp_y] []
  [disp_z] []
  [NLvar] []
[]

Where disp_x, disp_y, and disp_z are the displacement degrees of freedom, and NLvar is just another variable governed by a PDE related to plastic deformation of the body.

Based on the webpage for the Jacobian debugger, it is recommended to use random initial conditions (ICs) to ensure that the values and gradients of the variables are non-zero.

In line with the use of ICs for the Jacobian debugger, I have some questions:

1. To clarify, is the Jacobian debugger only used for the IC and does not advance in time? Is there a way to run the Jacobian debugger for future time steps of a simulation?

2. If we can only run the Jacobian debugger for ICs of a system, I was wondering if there is a way to ensure plasticity in the model. In my Jacobian, I have derivatives of plastic deformation with respect to total deformation, and I want to check that as well. So, I was wondering if there is a way to ensure that, based on the given ICs, I will have plasticity in the system.

Thank you so much.

[lindsayad]

I myself do not use the analyze_jacobian.py script. I typically reduce my mesh size to be quite small, and then I run my simulation as normal, but with the PETSc command line options -snes_test_jacobian -snes_test_jacobian_view (these options are at the core of the analyze_jacobian.py script). With these options you can run your simulation for arbitrarily many time steps and so you will certainly test things like non-zero gradients.

I usually pair this with DOFMap output to help ascertain what variable's Jacobian entries are bad

[Sina-av]

Great, thank you so much. This exactly did what I wanted.