Add forward model for prisms #86
Conversation
|
The test functions should include a few that check if the forward model has symmetry properties. |
|
@leouieda I'm wondering if the |
|
I have some doubts about the |
We agreed to change the vertical coordinate so it points upwards. See #89 |
The current |
Remove then old minus sign. Because g_z is the downward component, we want it pointing downwards.
Add test that compares the analytical solution of the g_z generated by an infinite slab with the one generated by a large enough prism. Also tests convergence: checks if the result gets closer to the analytical solution if the horizontal dimensions of the prisms are increased.
santisoler
force-pushed
the
prism_gravity
branch
from
158f916
to
9bb491d
Compare
It introduces some fixes to the equations that define the horizontal components of the gravity acceleration.
Remove the set of points located around the center of the prism: g_z must be zero on those points and testing their values is complicated.
Modify the safe_atan2 function in order to follow the Fukushima2019 definition. Add reference to Fukushima2019 on doc/references.rst. Update test function for the new safe_atan2 function.
Add set of points located on the upward == 0 plane. The values of g_z on those points must be all equal to zero.
|
@leouieda Besides the problem regarding the convergence test for very large prisms, I think this is looking really good, specially after implementing the |
|
One more thing I want to discuss is the fact that we are not allowing users to compute gravitational fields on computation points that falls inside the prism. Nevertheless, every paper claims that it can be safely done without modifying the analytical solution. Should we allow users to do so,? How do you think we should test it if so? |
Remove the check that used to raise an error if the computation point fell inside the prism. Add test function for g_z symmetry if computation points fall inside the prism, and include computation points inside the prism on the test function for potential symmetry.
Co-Authored-By: Leonardo Uieda <[email protected]>
Co-Authored-By: Leonardo Uieda <[email protected]>
Co-Authored-By: Leonardo Uieda <[email protected]>
Co-Authored-By: Leonardo Uieda <[email protected]>
Co-Authored-By: Leonardo Uieda <[email protected]>
Define a single set of computation points for the two examples. Modify the size and location of the single prism. Improve how the gz results are printed in order to avoid printing a numpy array. Create a title for the two prisms example.
Redefine coordinates from a stacked numpy array to a tuple in order to avoid unnecessary memory allocation. Fix Numba jitted function that expected a numpy array.
Change order of lower and higher boundaries when checking if an error is raised when invalid prism boundaries are passed. Remove unneeded lines.
Make use the already existing function to compute the gravity field of an infinite slab instead of rewriting it.
|
Thanks for the review @leouieda and sorry for all the mistakes! Feel free to take a look at this when you have a spare time! |
Add size of density array and number of prisms on the error message when they mismatch in order to make clearer what's failing. The improvement of such error messages was triggered after #98.
|
Merging when CIs finish building against master |
|
Just waiting for #105 to fix the linting errors |
|
Thanks @leouieda !
Haha, it's nice to see how we improved our way of doing things! |
Add forward model for prisms in Cartesian coordinates following
Nagy et al. (2000) and Nagy et al. (2002). Compute the potential and the
downward component of the gravitational acceleration generated by prisms
on computation points. Perform non-jitted checks for valid prisms
boundaries. Define a safe_atan2 function following Fukushima (2019) in
order to guarantee that the generated field satisfies Poisson's
equation. Use a safe_log function to allow the computation of
gravitational fields on singular points of the analytical solution.
Add test functions for symmetry and result comparison with an infinite
slab (Bouger plate).
Fixes #49
Reminders
make formatandmake checkto make sure the code follows the style guide.doc/api/index.rst.