Assigning a previous solution to part of the mesh

[rnpratoori]

Hello,

I have an exodus solution file which was solved on a domain of 4x1. I want to use this in another simulation where the domain is 4x3. The 4x1 solution occupies the bottom part of the 4x3 domain and the rest of the 4x2 domain has ConstantIC

I used the below snippet:

[Mesh]
  [main]
    type = GeneratedMeshGenerator
    dim = 2
    nx = 400
    ny = 300
    xmax = 4
    ymax = 3
  []
  [top_block]
    type = ParsedSubdomainMeshGenerator
    input = main
    block_id = 1
    combinatorial_geometry = 'y > 1'
  []
[]

[Variables]
    # polymer volume fraction
    [c1]
        order = FIRST
        family = LAGRANGE
    []
    # Chemical potential (nJ/mol)
    [w1]
        order = FIRST
        family = LAGRANGE
    []
    # polymer volume fraction
    [c2]
        order = FIRST
        family = LAGRANGE
    []
    # Chemical potential (nJ/mol)
    [w2]
        order = FIRST
        family = LAGRANGE
    []
[]

[ICs]
    [c1]
        type = SolutionIC
        from_variable = 'c'
        solution_uo = 2phase
        variable = c1
        block = 0
    []
    [c2]
        type = SolutionIC
        from_variable = 'c2'
        solution_uo = 2phase
        variable = c2
        block = 0
    []
    [top_c1]
        type = ConstantIC
        value = ${delta}
        variable = c1
        block = 1
    []
    [top_c2]
        type = ConstantIC
        value = ${delta}
        variable = c2
        block = 1
    []
[]

[UserObjects]
    [2phase]
      type = SolutionUserObject
      mesh = 'output/2phase_nd_2.e'
      system_variables = 'c c2'
      timestep = LATEST
    []
[]

This gives the error -

*** ERROR ***
/home/rnp/MOOSE/projects/panther/problems/3phase_nd.i:94.5:
The following error occurred in the UserObject '2phase' of type SolutionUserObject.

Failed to access the data for variable 'c' at point (x,y,z)=(    3.48,  1.00211,        0) in the '2phase' SolutionUserObjectBase

Why are these points being accessed when they are beyond the dimensions of the block?

[GiudGiud]

Hello

This has to do with how we compute the initial conditions using the generic projector algorithm. We basically loop over nodes, then edges and sides and volumes to make sure the variable is continuous, even though multiple ICs are at play here on the two blocks. Note that the nodes are shared between the blocks.

I am creating an issue and tentatively have a patch already #30826 (passes your case I think)

[rnpratoori]

@GiudGiud The PR works, but I noticed the shared nodes are not assigned any values - neither from the SolutionIC nor from the ConstantIC. Can this be fixed?

[GiudGiud]

I ll take a look.

[GiudGiud]

@rnpratoori made some changes to the PR. Does it work for you now?

[rnpratoori]

@GiudGiud this works perfectly. Thank you!

[mangerij]

@GiudGiud any status on this PR, will it be merged?

EDIT: nevermind, I did more reading. I understand the issue.

Do you think it is possible, instead of loading a variable that exists across a block restriction, to just partition that variable into each of its blocks and load them separately? Ttrying that now.

[GiudGiud]

You could try that but I think it hits the same issues on the boundaries near the blocks it is not defined on.

This PR is not very far but it's not quite ready. It's a complex problem to get it right for all FE families.
For lagrange it already works as is