This week in SimPEG…
SEOGI KANG has been building on Dom Fournier’s work implementing different norms, including sparse and sharp norms, and using Iteratively Reweighted Least Squares to update the model when sparse norms are considered. Dom’s work focussed on their application in potential fields inversions and Seogi is now applying them for DC resistivity in a groundwater example.
Dom Fournier is taking another perspective on his research with lp-norms in the regularization. He has done a review of the IRLS directive for sparsity+blocky models and taken a first stab at exploring the range of possible solutions with a
throw enough mud at the wall, some of it will stick
type strategy. Hopefully the outcome will be a “most likely” model with an associated uncertainty estimation.
Lindsey Heagy has been looking at the role of magnetic permeability in frequency domain electromagnetic experiments over steel cased wells. This is an example of an inductive-source response of a steel cased well, inspired by (Augustin, 1989). We have a large loop on the surface and are looking at the magnetic flux density (b) in a conductive borehole and a conductive, permeable borehole. There is a lot going on here, but a couple things to note: (1) magnetic permeability boosts the magnetic flux in the borehole walls (and boosts the maximum secondary magnetic flux you can achieve), (2) it reduces the frequency at which we start to observe EM induction effects.
Releases this week
Discretize 0.1.15 has been released. This implements an
aveCCV2F on the
TensorMesh so that cell centered vectors can be averaged to faces. Thanks to SEOGI KANG for the implementation and Craig Miller for the issue that prompted this development. The latest release is on pypi and can be obtained by opening a command prompt and running
pip install --upgrade discretize
In the meeting
We discussed an initial implementation of a
Simulation class within SimPEG to replace the
Problem and the
prob.pair(survey) interaction on SimPEG currently required for running a forward simulation. You can see the initial development on the Simulation pull request.
See you next week!