[MARS-UD-100] HTC with two materials

[rwendner]

for problems with two diffusion materials we face a number of issues while using the current version of the tetbased HTC model.

Jan Vorel temporarily fixed some - see latest to push to gitlab. However, a more consistent fix is still required, also addressing some of the following inconsistencies

(1) the creation of a DiffusionNodeList by reading from file is not supported - this should be changed, using the same syntax as for the NodeList command.
( LengthUnits mm
Particles
ReadFile CylNodeL.mrs )

(2) the keyword "ImportNodeList" for the DiffusionNodeList is inconsistent with the analog keyword for the DiffusionTetList (here simply TetList is used)

(3) The DiffusionTetList automatically creates also a DiffusionNodeList. The label for this DiffusionNodeList can not be set - this should be added.

(4) The definition of the DiffusionProblem allows the call of multiple DiffusionElementLists but only of one DiffusionNodeList. In order to solve a problem with two adjacent domains with different materials a common mesh needs to be used.
Consequently, a common diffusionNodeList needs to be created manually and used for the creation of both diffusion meshes (DiffusionTetLists). Without the quick fix as pushed this was not possible.

The attached files contain an example the works, but only with the fix as pushed to gitlab.

[zhouxinwei]

I see two ways to do it:

i) we start with one tet mesh, then divide to two sub domains with different materials.

ii) we start with two tet meshes (with shared nodes on the interface), combine the nodes into a single one.

In both cases, we need to have only one diffusion node list, which is fed into the solver.

Which case fits into your workflow better? The logic for the input file and the implementation (bullet points 1 and 4) will be different depending on the choice.

(2) I don't understand the question, please elaborate.

(3) it is available:

Code: Select all

UserDefinedList ttLDiff DiffusionTetList {
  TetList ttL
  DiffusionMaterial matDiff1
  EditNodeList {
    Rename abc // here we reset the name
  }
}


[rwendner]

at the moment we follow strategy (i). However, for more complex problems maybe (ii) is a better workflow.
Can we implement both ?

@(2) I simply suggest to make the keywords of the different functions equivalent.
If we use just "TetList" in the DiffusionTetList then we should also just use "NodeList" in the DiffusionNodeList instead of "ImportNodeList" etc.

@(3) OK, thanks. Since in this step we create from a NodeList a DiffusionNodeList I feel that the command "EditNodeList" is misleading. I would prefer having a command like "DiffusionNodeListLabel" which makes it clear that a new list is created of type "DiffusionNodeList"

@(4) please check Jan's quick fix the allow work flow (i)

[zhouxinwei]

@(2) They are doing different things. TetList is the base geometry mesh for the diffusion tet list, that's not importing. In the diffusion framework, the ImportNodeList is used to combine different DIFFUSION node lists to one, as described in the development note. Its purpose is to help in a situation, e.g. prescribe BC on two diffusion node lists, it's easier to combine then operate on them together.

If we want to combine two node lists into one, try comparing:

Code: Select all

UserDefinedList comb DiffusionNodeList {
  ImportNodeList ndL1
  ImportNodeList ndL2
}


Code: Select all

UserDefinedList comb DiffusionNodeList {
  NodeList ndL1
  NodeList ndL2
}


@(3) Every element list has a node list. In the diffusion element list, the node list is also type of diffusion. There shouldn't be confusion.

@(4) The fix may work for the case you were doing. But the ImportNodeList is not supposed to import a FE node list, see @(2). Right now, the diffusion framework begins with a tet mesh.

There is work to be done to make the framework ready for sublists with different materials. I will keep you updated.

[zhouxinwei]

Logic to model multiple tet element lists with different materials is implemented, please see development notes in MARS-UD-100 repository in GitLab.