Not converge about DIVERGED_NANORINF and DIVERGED_LINEAR_SOLVE problem.

[lipxisi]

Hello

In the case of 06.i of porous_flow I modified geometry, fluid properties and permeability to suit my model. However, the solver does not converge, sometimes there will be DIVERGED_NANORINF and DIVERGED_LINEAR_SOLVE related problems, like

[Time Step 3, time = 150, dt = 50
 0 Nonlinear |R| = 1.449103e-08
 Linear solve did not converge due to DIVERGED_NANORINF iterations 0
Nonlinear solve did not converge due to DIVERGED_LINEAR_SOLVE iterations 0
 Solve Did NOT Converge!
Aborting as solve did not converge]

sometimes there will be directly skip linear convergence problems.
[Time step 5,time =18000,dt =3600
0 Nonlinear lRl=4.939397e-15 
Solve converged!]

I don't know what is the main reason for this, and I should adjust those areas. Though I have looked at https://mooseframework.inl.gov/moose/application_usage/failed_solves.html, but there is no good solution.

inputfile

[Mesh]
  [cube]
    type = GeneratedMeshGenerator
    dim = 2
    xmax = 9.5
    ymax = 9.5
    nx = 100
    ny = 100
  []
  #[rename]
  #  type = RenameBoundaryGenerator
  #  input = cube
  #  old_boundary = '0 1 2 3'
  #  new_boundary = 'out_side out_side out_side out_side'
  #[]
[]

[GlobalParams]
  PorousFlowDictator = dictator
[]

[Variables]
  [porepressure]
    initial_condition = 0
  []
  [tracer_concentration]
    initial_condition = 0
    #scaling = 1e5
  []
[]

[ICs]
  [tracer_concentration]
    type = ConstantIC
    value = 1
    variable = tracer_concentration
    boundary = top
  []
  [poutside]
    type = ConstantIC
    variable = porepressure
    boundary = top
    value = 5e5
  []
[]

[BCs]
  [tracer_concentration]
    type = DirichletBC
    value = 1
    variable = tracer_concentration
    boundary = top
  []
  [1tracer_concentration]
    type = DirichletBC
    value = 0
    variable = tracer_concentration
    boundary = bottom
  []
  [poutside]
    type = DirichletBC
    variable = porepressure
    boundary = top
    value = 5e5
  []
  [1poutside]
    type = DirichletBC
    variable = porepressure
    boundary = bottom
    value = 0
  []
[]

[PorousFlowFullySaturated]
  porepressure = porepressure
  coupling_type = Hydro
  gravity = '0 0 0'
  fp = the_simple_fluid
  mass_fraction_vars = tracer_concentration
  stabilization = KT # Note to reader: 06_KT.i uses KT stabilization - compare the results
  flux_limiter_type = superbee
[]

[FluidProperties]
  [the_simple_fluid]
    type = SimpleFluidProperties
    bulk_modulus = 13E9
    viscosity = 9E-2
    density0 = 2787.0
  []
[]

[Materials]
  [porosity]
    type = PorousFlowPorosity
    porosity_zero = 0.15
  []
  [permeability_caps]
    type = PorousFlowPermeabilityConst
    block = '0'
    permeability = '1E-19 0 0   0 1E-19 0   0 0 1E-19'
  []
[]

[Preconditioning]
  active = basic
  [basic]
    type = SMP
    full = true
    petsc_options = '-ksp_diagonal_scale -ksp_diagonal_scale_fix'
    petsc_options_iname = '-pc_type -sub_pc_type -sub_pc_factor_shift_type -pc_asm_overlap'
    petsc_options_value = ' asm      lu           NONZERO                   2'
  []
  [preferred_but_might_not_be_installed]
    type = SMP
    full = true
    petsc_options_iname = '-pc_type -pc_factor_mat_solver_package'
    petsc_options_value = ' lu       mumps'
  []
[]

[Functions]
  [dt_controller]
    type = PiecewiseConstant
    x = '0    3600   18000  43200'
    y = '50   500  1800  43200'
  []
[]

[Executioner]
  type = Transient
  solve_type = Newton
  end_time = 3.6e5

  [TimeStepper]
    type = FunctionDT
    function = dt_controller
  []
  nl_abs_tol = 1E-10
  l_abs_tol = 1e-15
[]

[Outputs]
  exodus = true
  print_linear_residuals = true
[]

Thank you for checking!

[GiudGiud]

Hello

I dont usually say this but your convergence criteria are too tight

  nl_abs_tol = 1E-10
  l_abs_tol = 1e-15

you can leave the first one if it works, but the second one is too close to numerical precision. In fact since the residual is checked by the nonlinear solver, you dont need to converge the linear solve anywhere near as tight. 1e-3 is often enough

Guillaume