Some questions about calculating time steps

[HPL001]

Hello moose community,
I have a problem. The adaptive time interval of the compute step suddenly increases at 5000. This is my compute setting.

[Executioner]
  type = Transient
  # scheme = BDF2
  # solve_type = PJFNK
  # petsc_options_iname = '-pc_type -ksp_gmres_restart -sub_ksp_type
  #                       -sub_pc_type -pc_asm_overla'
  # petsc_options_value = 'asm      31       preonly
  #                        ilu      2'
  solve_type = 'NEWTON'
  l_tol = 1e-04
  l_max_its = 30
  nl_max_its = 15
  nl_rel_tol = 1e-03
  nl_abs_tol = 1e-03
  start_time = 0.0
  end_time = 20000.0
  # dt = 0.00125

  [TimeStepper]
    type = IterationAdaptiveDT
    cutback_factor = 0.9
    dt = 0.001
    growth_factor = 1.2
    optimal_iterations = 15
  []

  # [Adaptivity]
  #   interval = 2
  #   refine_fraction = 0.2
  #   coarsen_fraction = 0.3
  #   max_h_level = 4
  # []
[]

and then, Here are the time steps I monitored.

[HPL001]

By the way, how should I output my initial condition? Below is my output, but it only outputs the data after the first time step, not the initial situation.

[HPL001]

[Outputs]
  exodus = true
  print_linear_residuals = true
  execute_on = timestep_end
  checkpoint = true
  solution_history = true
  # [out]
  #   type = VTK
  #   file_base = Sintering_output
  #   enable = true
  #   execute_on = timestep_begin
  # []
  # [out]
  #   type = Tecplot
  #   ascii_append = true
  # []
[]

[GiudGiud]

Change execute_on = timestep_end to "initial timestep_end"

[HPL001]

In addition, I have another question to learn. I used the phase field method and these kernels in the simulation of particle sintering. If I want to see changes in particle shape, should I use the variable bnds or c?

[Variables]
  [c] # 定义颗粒浓度变量（守恒变量）
    order = FIRST
    family = LAGRANGE
  []

  [w] # 定义总自由能的函数值变量
    order = FIRST
    family = LAGRANGE
  []

  [PolycrystalVariables]
  []
[]

[Kernels]
  [c_res]
    type = SplitCHParsed
    f_name = F
    kappa_name = kappa_c
    variable = c
    coupled_variables = 'eta0 eta1'
    w = w
  []

  [w_res]
    type = SplitCHWRes
    variable = w
    mob_name = M
  []

  [time]
    type = CoupledTimeDerivative
    v = c
    variable = w
  []
[]

[AuxVariables]
  [bnds]
  []
[]

[AuxKernels]
  [bnds]
    type = BndsCalcAux
    v = 'eta0 eta1'
    variable = bnds
    var_name_base = eta
    op_num = 2
  []
[]

[GiudGiud]

The time stepping behavior is expected. That s what the IterativeDtTimestepper does

If you want a different time stepping behavior, pick another one in the Executioner/TimeStepper block

[GiudGiud]

Should be gradual. I ll check this out.

I d still change the TimeStepper if you don't want it to change

[HPL001]

Thank you for your help in checking. I wonder if this phenomenon is normal. Also, is there any way I can limit his maximum time step?

[GiudGiud]

Yes you can set dtmax and you can set multiple timesteppers in the TimeSteppers block (notice the s) and the smallest step there by any of the steppers limits the step

[HPL001]

I set the maximum time step as you suggested, but something seems to be wrong

[Executioner]
  type = Transient
  # scheme = BDF2
  # solve_type = PJFNK
  # petsc_options_iname = '-pc_type -ksp_gmres_restart -sub_ksp_type
  #                       -sub_pc_type -pc_asm_overla'
  # petsc_options_value = 'asm      31       preonly
  #                        ilu      2'
  solve_type = 'NEWTON'
  l_tol = 1e-04
  l_max_its = 30
  nl_max_its = 15
  nl_rel_tol = 1e-03
  nl_abs_tol = 1e-03
  start_time = 0.0
  end_time = 20000.0
  # dt = 0.00125

  [TimeStepper]
    type = IterationAdaptiveDT
    cutback_factor = 0.8
    dt = 0.00001
    growth_factor = 1.5
    optimal_iterations = 10
    iteration_window = 5
    dtmax = 1.0
  []

  # [Adaptivity]
  #   interval = 2
  #   refine_fraction = 0.2
  #   coarsen_fraction = 0.3
  #   max_h_level = 4
  # []
[]

error：

*** ERROR ***
/mnt/d/projects/moose/modules/combined/sintering_datas/02_04_06/peacock_run_exe_tmp_MySintering02_04.i:485.5: unused parameter 'Executioner/TimeStepper/dtmax'

There seems to be no maximum time step in adaptive step size, and I looked in the manual and couldn't find it.

[GiudGiud]

you'll have to use two timesteppers to do this then.