[ Convergence problem ] occurs when i change the material properties. [ N-S module ]

[leezw37]

Hello, everyone. I am a confused new user of MOOSE and need your help! Anything will be appreciated!

When i was using Navier Stokes Module, I got a convegence problem as soon as i changed the material properties of a test file.And here is the address: moose/modules/navier_stokes/test/tests/finite_element/ins/velocity_channel/velocity_inletBC_by_parts.i

All i have done to make this didn't converge is changeing "mu" from 1 to 1e-5 in the 88th line. I didn't get it.

(Why 1e-5?
Actually, I am doing a more complex thing with the propeties of air : rho = 1.225, mu = 1.7894e-05.
When i used real properties of air or water, instead of 1 like so many other test files, i got this problem. And this problem occurs in many other test files.)

MOOSE is no doubt a powerful tool, and i think there must be a way to deal with the complex matertial properties.Is there anyone can help me to find out the reason and solution? Love you.

========================================================================================
#moose/modules/navier_stokes/test/tests/finite_element/ins/velocity_channel/velocity_inletBC_by_parts.i
# This input file tests outflow boundary conditions for the incompressible NS equations.

[GlobalParams]
  gravity = '0 0 0'
  integrate_p_by_parts = true
[]

[Mesh]
  type = GeneratedMesh
  dim = 2
  xmin = 0
  xmax = 3.0
  ymin = 0
  ymax = 1.0
  nx = 30
  ny = 10
  elem_type = QUAD9
[]


[Variables]
  [./vel_x]
    order = SECOND
    family = LAGRANGE
  [../]
  [./vel_y]
    order = SECOND
    family = LAGRANGE
  [../]
  [./p]
    order = FIRST
    family = LAGRANGE
  [../]
[]

[Kernels]
  [./mass]
    type = INSMass
    variable = p
    u = vel_x
    v = vel_y
    pressure = p
  [../]
  [./x_momentum_space]
    type = INSMomentumLaplaceForm
    variable = vel_x
    u = vel_x
    v = vel_y
    pressure = p
    component = 0
  [../]
  [./y_momentum_space]
    type = INSMomentumLaplaceForm
    variable = vel_y
    u = vel_x
    v = vel_y
    pressure = p
    component = 1
  [../]
[]

[BCs]
  [./x_no_slip]
    type = DirichletBC
    variable = vel_x
    boundary = 'top bottom'
    value = 0.0
  [../]
  [./y_no_slip]
    type = DirichletBC
    variable = vel_y
    boundary = 'left top bottom'
    value = 0.0
  [../]
  [./x_inlet]
    type = FunctionDirichletBC
    variable = vel_x
    boundary = 'left'
    function = 'inlet_func'
  [../]
[]

[Materials]
  [./const]
    type = GenericConstantMaterial
    block = 0
    prop_names = 'rho mu'
    prop_values = '1  1e-5'  # mu was 1 in the original inputfile
  [../]
[]

[Preconditioning]
  [./SMP_PJFNK]
    type = SMP
    full = true
    solve_type = NEWTON
  [../]
[]

[Executioner]
  type = Steady
  petsc_options_iname = '-ksp_gmres_restart -pc_type -sub_pc_type -sub_pc_factor_levels'
  petsc_options_value = '300                bjacobi  ilu          4'
  line_search = none
  nl_rel_tol = 1e-12
  nl_max_its = 6
  l_tol = 1e-6
  l_max_its = 300
[]

[Outputs]
  [./out]
    type = Exodus
  [../]
[]

[Functions]
  [./inlet_func]
    type = ParsedFunction
    value = '-4 * (y - 0.5)^2 + 1'
  [../]
[]

[lindsayad]

For Re > 1, you should put supg = true in your GlobalParams block. Unless you add streamline-upwind Petrov-Galerkin stabilization, you will essentially have a central-differencing scheme for advection which does not capture the physics and is unstable.