Upgrades to static EM simulations. #1347
Conversation
|
I would recommend leaving the number of wave numbers in the perpendicular horizontal component as |
|
Also, if you don’t mind, please split the two suggestion’s up into two separate PR’s. |
| ) | ||
| else: | ||
| raise NotImplementedError( | ||
| "Only isotropic and linear isotropic conductivities implemented." |
There was a problem hiding this comment.
This formulation should actually be able to handle all anisotropic conductivities (as we don’t need to invert the matrix). Let me double check what the tensor reduces to.
There was a problem hiding this comment.
This whole thing started because the current formulation does not naturally handle any form of anisotropy. E.g. the sig_1 and sig_2 for linear anisotropy are used to form the mass matrix within the div-grad term. And the sig_3 is used for the term associated with the wave domain. Although we could likely add full tensor properties, I don't anticipate anyone using it in practice.
There was a problem hiding this comment.
Oh I just meant in the future it can support more than just axial anisotropy, compared to the cell centered one that would only support axial.
Then perhaps we need to consider a third option. When we generate meshes for the 2.5D problem, and when we ask for things like cell center locations, the coordinates are x (along-line) and y (elevation). This is also the case when defining 2.5D DC/IP surveys. So 'nky' is not the number along the y-direction. Could we have a property like 'n_wave_coefficients'? |
|
@jcapriot what do you think about defining new properties like 'MeSigma2D' and 'MnSigma2D' as opposed to overriding 'MeSigma' and 'MnSigma'? If we add the new properties, I will need to ensure they are cleared upon a model update. |
We create meshes with x1 and x2 dimensions that correspond to x and z. The only properties of meshes that specifically refer to dimensions are on a Also, making it the way you’ve suggested would mean that for 2.5D the anisotropy tensor would be ordered: along-line, elevation, cross-line, and for 3D it would be along-line, cross-line, elevation. So simply for consistencies sake we should continue referring to the cross line dimension as y. |
Leave the properties named the same, in fact I would just call the |
Codecov ReportAttention: Patch coverage is
Additional details and impacted files@@ Coverage Diff @@
## main #1347 +/- ##
==========================================
+ Coverage 82.33% 82.36% +0.03%
==========================================
Files 165 165
Lines 25249 25385 +136
==========================================
+ Hits 20788 20908 +120
- Misses 4461 4477 +16 ☔ View full report in Codecov by Sentry. |
This pull request is to add functionality that was deemed useful after the AgroGeo24 workshop. In this pull request, we aim to add linear anisotropy to 2.5D DC/IP simulation classes. This includes the forward problem and derivatives that can be used to develop and inversion framework.