Previous and current variable values in pre-SMO calculation

[NateAM]

Check these boxes if you have followed the posting rules.

• Q&A General is the most appropriate section for my question
• I have consulted the posting Guidelines on the Discussions front page
• I have searched the Discussions forum and my question has not been asked before
• I have searched the MOOSE website and the documentation does not answer my question
• I have formatted my post following the posting guidelines (screenshots as a last resort, triple back quotes around pasted text)

Question

I'm using the automatic scaling calculation for a simulation using a custom kernel which requires both the previous and current value of the field variables and their gradients. When the pre-SMO calculation runs I've noticed that that the previous and current field variable values are the same. Is this intended behavior or does it seem like there is a bug somewhere?

The executioner part of my input deck is

[Executioner]
  type = Transient
  solve_type = NEWTON
  petsc_options_iname = "-pc_type -pc_factor_mat_solver_package"
  petsc_options_value = "lu       superlu_dist                 "
  line_search = none
  automatic_scaling = true
  nl_rel_tol = 1e-5
  nl_abs_tol = 1e-7
  nl_max_its = 50
  start_time = 0.0
  end_time = 1.0
  dtmin = 1e-12
  dtmax= 0.1
  [TimeStepper]
    type = IterationAdaptiveDT
    optimal_iterations = 4
    iteration_window = 3
    linear_iteration_ratio = 1000
    growth_factor=1.2
    cutback_factor=0.5
    dt = 0.1
  []
[]

I'm not completely sure that the identical values are happening during the pre-SMO calculation because my kernel fails to converge (due to the two deformations being the same) and it throws an exception that MOOSE catches. I recognize that the kernel-specific output won't make much sense, but you can see that the variables u and phi, and their gradients are identical for the current and previous timestep. All of the messages from my kernel are between the "="s.

Time Step 1, time = 0.1, dt = 0.1
Pre-SMO residual: 0.318904
Currently Executing
    Computing Pre-SMO Residual
      Finished Computing Residual                                                        [  9.47 s] [  174 MB]
    Finished Computing Pre-SMO Residual                                                  [  9.47 s] [  174 MB]

Performing automatic scaling calculation

    Computing Automatic Scaling
      Computing Jacobian                                                                 [  7.42 s] [  174 MB]
    Finished Computing Automatic Scaling                                                 [  7.42 s] [  174 MB]
    Computing Residual.                                                                  [ 12.47 s] [  174 MB]
 0 Nonlinear |R| = 1.696577e-01
    Computing Jacobian...                                                                [ 20.69 s] [  202 MB]
      0 Linear |R| = 1.696577e-01
      1 Linear |R| = 1.391660e-16
    Finished Computing Residual                                                          [  9.52 s] [  555 MB]
 1 Nonlinear |R| = 3.266552e-03
    Computing Jacobian..                                                                 [ 19.01 s] [  557 MB]
      0 Linear |R| = 3.266552e-03
      1 Linear |R| = 2.854340e-18
    Finished Computing Residual                                                          [  9.45 s] [  557 MB]
 2 Nonlinear |R| = 1.907495e-05
    Computing Jacobian..                                                                 [ 19.00 s] [  558 MB]
      0 Linear |R| = 1.907495e-05
      1 Linear |R| = 2.836699e-20
    Finished Computing Residual                                                          [  9.45 s] [  557 MB]
 3 Nonlinear |R| = 8.579929e-08
 Solve Converged!
  Finished Solving                                                                       [126.32 s] [  408 MB]

Outlier Variable Residual Norms:
  disp_z: 8.389971e-08

Postprocessor Values:
+----------------+----------------+
| time           | bot_react_z    |
+----------------+----------------+
|   0.000000e+00 |   0.000000e+00 |
|   1.000000e-01 |   5.104526e+02 |
+----------------+----------------+


Time Step 2, time = 0.2, dt = 0.1

    Computing Pre-SMO Residual
      Computing Residual.                                                                [ 10.11 s] [  408 MB]
Pre-SMO residual: 0.322034
    Finished Computing Pre-SMO Residual                                                  [ 10.12 s] [  408 MB]

A MooseException was raised during FEProblemBase::computeResidualTags

==========================
Convergence not achieved in material model. Requesting timestep cutback.
0 : Failure in relaxed solve
INPUT PARAMETERS FOLLOW:
time:
 2.000000e-01, 1.000000e-01,

fparams:
 2.000000e+00, 1.000000e+04, 1.000000e-08, 2.000000e+00, 3.192203e+00, -3.167845e+01, 2.000000e+00, 1.000000e+04, 1.000000e-08, 2.000000e+00, 0.000000e+00, 0.000000e+00, 2.000000e+00, 0.000000e+00, 0.000000e+00, 2.000000e+00, 0.000000e+00, 0.000000e+00, 2.000000e+00, 0.000000e+00, 0.000000e+00, 2.000000e+00, 0.000000e+00, 0.000000e+00, 2.000000e+00, 0.000000e+00, 0.000000e+00, 2.000000e+00, 6.964416e+02, 1.267138e+02, 5.000000e+00, -1.867498e+01, -3.781732e+01, 1.517765e+01, -2.407120e+01, -5.861821e+00, 1.100000e+01, 0.000000e+00, 0.000000e+00, 0.000000e+00, 0.000000e+00, 0.000000e+00, 0.000000e+00, 7.925235e+02, 0.000000e+00, 0.000000e+00, 0.000000e+00, 0.000000e+00, 2.000000e+00, -3.781732e+01, -5.861821e+00, 5.000000e-01, 5.000000e-01, 5.000000e-01, 1.000000e-09, 1.000000e-09,

current_grad_u:
 1.055797e-03, 5.136526e-04, 1.843982e-03,
 4.002290e-04, 1.684113e-03, 2.357700e-03,
 -2.437182e-05, -9.028907e-05, -7.940004e-02,

current_phi:
 -1.956531e-04, 3.049702e-06, 1.765037e-05, 2.840806e-06, -1.943198e-04, 2.230603e-05, -1.359193e-05, -1.715515e-05, -3.393720e-04,

current_grad_phi:
 -1.945539e-05, -2.592158e-05, -5.862817e-05,
 -3.320356e-06, -3.467268e-06, -2.339044e-07,
 -2.038262e-05, -4.046447e-06, 1.301341e-05,
 -3.317225e-06, -1.923737e-06, -2.083438e-07,
 -1.566101e-05, -3.290013e-05, -5.876160e-05,
 -3.335579e-06, -2.500360e-05, 1.655791e-05,
 1.558924e-05, 2.839010e-06, -7.544714e-06,
 2.378279e-06, 1.869235e-05, -9.588049e-06,
 -1.327543e-05, -2.236166e-05, -8.576912e-05,

previous_grad_u:
 1.055797e-03, 5.136526e-04, 1.843982e-03,
 4.002290e-04, 1.684113e-03, 2.357700e-03,
 -2.437182e-05, -9.028907e-05, -2.470542e-02,

previous_phi:
 -1.956531e-04, 3.049702e-06, 1.765037e-05, 2.840806e-06, -1.943198e-04, 2.230603e-05, -1.359193e-05, -1.715515e-05, -3.393720e-04,

previous_grad_phi:
 -1.945539e-05, -2.592158e-05, -5.862817e-05,
 -3.320356e-06, -3.467268e-06, -2.339044e-07,
 -2.038262e-05, -4.046447e-06, 1.301341e-05,
 -3.317225e-06, -1.923737e-06, -2.083438e-07,
 -1.566101e-05, -3.290013e-05, -5.876160e-05,
 -3.335579e-06, -2.500360e-05, 1.655791e-05,
 1.558924e-05, 2.839010e-06, -7.544714e-06,
 2.378279e-06, 1.869235e-05, -9.588049e-06,
 -1.327543e-05, -2.236166e-05, -8.576912e-05,

SDVS:
 0.000000e+00, 0.000000e+00, 0.000000e+00, 0.000000e+00, 0.000000e+00, 0.000000e+00, 0.000000e+00, 0.000000e+00, 0.000000e+00, 0.000000e+00, 0.000000e+00, 0.000000e+00, 0.000000e+00, 0.000000e+00, 0.000000e+00, 0.000000e+00, 0.000000e+00, 0.000000e+00, 0.000000e+00, 0.000000e+00, 0.000000e+00, 0.000000e+00, 0.000000e+00, 0.000000e+00, 0.000000e+00, 0.000000e+00, 0.000000e+00, 0.000000e+00, 0.000000e+00, 0.000000e+00, 0.000000e+00, 0.000000e+00, 0.000000e+00, 0.000000e+00, 0.000000e+00, 0.000000e+00, 0.000000e+00, 0.000000e+00, 0.000000e+00, 0.000000e+00, 0.000000e+00, 0.000000e+00, 0.000000e+00, 0.000000e+00, 0.000000e+00, 0.000000e+00, 0.000000e+00, 0.000000e+00, 0.000000e+00, 0.000000e+00, 0.000000e+00, 0.000000e+00, 0.000000e+00, 0.000000e+00, 0.000000e+00,

current_ADD_DOF:

current_ADD_grad_DOF:

previous_ADD_DOF:

previous_ADD_grad_DOF:


To recover, the solution will fail and then be re-attempted with a reduced time step.
==========================

  Nonlinear solve did not converge due to DIVERGED_FUNCTION_DOMAIN iterations 0
 Solve Did NOT Converge!
  Finished Solving                                                                       [ 13.82 s] [  408 MB]
Aborting as solve did not converge

[GiudGiud]

Hello

What is SDVs?
And ADD?

In any case, you should not be using the pre-SMO residual?
use_pre_SMO_residual should be set to False by default.
I think you may be using a legacy application that still uses those. See this announcement for the deprecation of using the pre-SMO residual
https://mooseframework.inl.gov/newsletter/2024/2024_04.html

Otherwise, how do you retrieve the current and previous value in the code?

[GiudGiud]

I collect the previous and current values of the degrees of freedom by declaring in myKernel.h:

ok this looks correct.

Maybe it's just because the pre-SMO calculation is depreciated that there is this problem at all?

It's not deprecated, it's just that the old values of anything is not generally used on the first time step, because they dont generally exist. I would expect them to return 0, but turns out they return the current value. This is probably a safer value for most problems, just not for yours

[NateAM]

It's complicated. Do you really need the old value? Is this for a restart simulation?
Could you just use a different expression for the material on the first time step?

Okay, it's mostly just for plotting purposes so maybe I don't. I don't explicitly ask for Pre-SMO anywhere in my input deck so this could be a function of my main.C not being updated to the new standard. I'll try and update and see what happens.

It's not deprecated, it's just that the old values of anything is not generally used on the first time step, because they dont generally exist. I would expect them to return 0, but turns out they return the current value. This is probably a safer value for most problems, just not for yours

That makes sense to me. What is confusing to me is that at every time step the pre-SMO residual is called and if my kernel can't converge locally in the pre-SMO residual I find that the current and previous values are the same even when it isn't the first time step. Is that expected?

[GiudGiud]

I dont think so.
If you could make a minimal non-working example I can take a look

[NateAM]

@GiudGiud

Sorry it's taken me a while to respond. I updated to the most recent version of MOOSE and I'm not able to reproduce the error so maybe this was something that was fixed. I'm still getting the warnings:

LEGACY MODES ENABLED:
 This application uses the legacy material output option: material properties are output only on TIMESTEP_END, not INITIAL. To remove this message, set 'use_legacy_material_output' to false in this application. If there are gold output files that contain material property output for which output occurs on INITIAL, then these will generate diffs due to zero values being stored, and these tests should be re-golded.

LEGACY MODES ENABLED:
 This application uses the legacy initial residual evaluation behavior. The legacy behavior performs an often times redundant residual evaluation before the solution modifying objects are executed prior to the initial (0th nonlinear iteration) residual evaluation. The new behavior skips that redundant residual evaluation unless the parameter Executioner/use_pre_smo_residual is set to true. To remove this message and enable the new behavior, set the parameter 'use_legacy_initial_residual_evaluation_behavior' to false in *App.C. Some tests that rely on the side effects of the legacy behavior may fail/diff and should be re-golded.

Though I don't think I have enabled the pre-SMO explicitly. My executioner has the form:

[Executioner]
  type = Transient
  solve_type = NEWTON
  petsc_options_iname = "-pc_type -pc_factor_mat_solver_package"
  petsc_options_value = "lu       superlu_dist                 "
  line_search = none
  automatic_scaling = true
  nl_rel_tol = 1e-5
  nl_abs_tol = 1e-7
  nl_max_its = 10
  start_time = 0.0
  end_time = 1.0
  dtmin = 1e-12
  dtmax= 0.1
  [TimeStepper]
    type = IterationAdaptiveDT
    optimal_iterations = 5
    iteration_window = 1
    linear_iteration_ratio = 1000
    growth_factor=1.2
    cutback_factor=0.5
    dt = 0.1
  []
[]

So does it have to do with using automatic_scaling?

[GiudGiud]

Automatic scaling does trigger an earlier Jacobian evaluation . So a similar effect as the legacy preSMO behavior