[rxgrch]

Hi foamers:

what I am doing is to implement a high order scheme for the FVM's face interplation, but now I face a problem with the lduMatrix, as we know, the matrix values just store for the face shared by two cells, and indexed by face, which defined by fvMeshLduAddressing. to implement my high order face interplation, the value not shared by two adjacent cells also had to be stored in the lduMtrix, which result in the modification of fvMeshLduAddressing. however, I found this modify of fvMeshLduAddressing may trigger a chain reaction in the code, which I had to modify the existing low order schemes, a heavy task. So can anyone help me with this problem, and point out if there are something wrong with my implement, or some advices? Thank you very much!

[santiagomarquezd]

Hi, I would suggest to read this material:

https://docs.google.com/file/d/1Y0cX...it?usp=sharing

http://www.sciencedirect.com/science...17931002003307

in order to check if is it possible to write your scheme in the FOAM's TVD framework. In addition take into account that schemes with double upwind cell information cannot be implemented in non-structured meshes in general since this cell cannot be always univocally determined.

Regards.

[rxgrch]

Originally Posted by santiagomarquezd

Hi, I would suggest to read this material:

https://docs.google.com/file/d/1Y0cX...it?usp=sharing

http://www.sciencedirect.com/science...17931002003307

in order to check if is it possible to write your scheme in the FOAM's TVD framework. In addition take into account that schemes with double upwind cell information cannot be implemented in non-structured meshes in general since this cell cannot be always univocally determined.

Regards.

Thanks for your reply, and all my work is now under the structured mesh, your document and paper seemed to be of great value for my work. it seemed that the four TVD schemes you mentioned in the paper still used just two cells C and D in Fig 2(b), the cell U actually isn't used, right? Did all the four sheme had been implement in the current OpenFOAM version? Or just your in-house code? Thank you!

[santiagomarquezd]

Hi, even if you are using structured meshes it is possible to remove the dependency of the U cell (take for example the QUICK limiter) and write the method in terms of C and P (or P and N in FOAM nomenclature), using (16) in the work of Darwish. This work explains the way TVD is implemented in FOAM. I think you would have to try this way first before changing the lduAddressing.

Regards.