Forum Message

 

 

We have moved the forum to https://forum.ngsolve.org . This is an archived version of the topics until 05/05/23. All the topics were moved to the new forum and conversations can be continued there. This forum is just kept as legacy to not invalidate old links. If you want to continue a conversation just look for the topic in the new forum.

Notice

The forum is in read only mode.

DG and Newtons method linearization

More
6 years 3 months ago #320 by Matejczyk
Dear All,

First of all, congratulations to the developers of Netgen. I think it is a very good solver and what is more, the documentation is also properly developed which is not that usual.

I'm working on a solver for a system of 3 nonlinear equations. Following the tutorials and documentation, I wrote a Newton solver that solves the problem using a 1D mesh. As some of the parameters in my equations are discontinuous I got into troubles and wanted to implement a DG method to include the discontinuity of the parameters. I followed the DDG tutorial and added the jumps on the boundaries of the elements. In my case that would be points as I am working with 1D mesh.

Unfortunately, during the linearization of the form, I got an error

a.AssembleLinearization(u.vec)
RuntimeError: CalcLinearization for element-wise skeleton-VOL not implementedin AssembleLinearization


Is there anything I can do to overcome this problem or is it just something that is not implemented and won't change?
More
6 years 3 months ago #321 by cwinters
Hi,

could you please attach your python file.
Sounds like you used a combination of arguments in the SymbolicBFI which is not implemented.

Best regards
Christoph
More
6 years 3 months ago #322 by Matejczyk
Hi,

Here is the interesting part of the code. I cut it out of a bigger code to show where the problem is.

Best wishes,
B
Attachments:
More
6 years 3 months ago #323 by joachim
Hi B,

thank you for the flowers ...

it is as the error tells us, linearization for DG is not implemented:
RuntimeError: CalcLinearization for element-wise skeleton-VOL not implementedin AssembleLinearization

you can replace your DG formulation by an hybrid-DG (HDG), then it shall work.

It is not a principal problem with non-linear DG, but we mainly use HDG (rather than DG), and so nobody found time to implement that.

Best, Joachim
More
6 years 3 months ago #324 by Matejczyk
Hi Joachim,

Thanks for the answer.
I will implement it and see how it works.
Thanks,
Bart
More
6 years 3 months ago #328 by Matejczyk
Hi Joachim,

I was trying the HDG as you suggested but it looks like im still doing something very wrong. Now it crashes when I try to define the Facet space. Might it be a problem of 1D mesh?

The message I obtain when I try to create the Facet space is

setting all_dofs_together 0
Segmentation fault: 11
My space definition looks like
Y = H1(mesh, order=ord, dirichlet=[1,2])
F = FacetFESpace(mesh, order=ord)
and the second line is where the code collapses. Do you have an idea what might be the reason?

Best wishes and thanks again,
Bart
Time to create page: 0.131 seconds