HomologyPostProcessing: false results for complicated geometry
Hi everyone,
I'm moving GetDP issue 164 here since it seems to be a problem with the HomologyPostProcessing
plugin instead of GetDP
.
We are trying to run a transient simulation of two insulated spiral pancake coils connected in series using an \vec{H}-\phi
formulation, see below for the geometry. Unfortunately, we could not reproduce the problem with easier geometries...
This model setup then has two cohomology basis functions. We follow Section 5 of the Gmsh cohomology paper to create a cohomology basis of 2 elements with evident physical interpretation:
Details on the creation of the modified cohomology basis
- The original cohomology basis are regions 8 and 9
- We modify it by calling the
HomologyPostProcessing
plugin to make it compatible with the loops 6 (terminal current) and 11 ("free" current circulating in the cylindrical connecting parts) - The resulting cuts are 12, 13 (and copies 3000001, 3000002), shown below
To check the cohomology basis, we then set up a very simple .pro file that only calls InitSolution
in the Resolution
and enforces a current of 1A in the first cut and 0A in the second (effectively "disabling" it). This should yield the following initial solution (log scale) with currents only in the conducting parts:
The above picture is for a spiral of 12 turns. When increasing the number of turns to 13, we have non-zero currents in the air domain:
- Using the cohomology basis created from the
HomologyComputation
(Regions 8 and 9 as commented in the.pro
) without the rearranging to make homology and cohomology compatible, again no currents in the non-conducting domain. Thus, it seems to be a problem with theHomologyPostProcessing
plugin - Removing the second cut region, i.e.
DOM_airHoleCut = Region[{ }];
, there are no currents in non-conducting domains. - Below 13 turns (exclusive), no air currents. Above 13 turns, air currents.
Please find below the simple .pro and two meshes for the two cases. We tested it with the official GetDP 3.5.0
release.
Thank you for the help!
Erik