extra tests ideas

[shimwell]

Test that point coordinates are generated within the ranges expected.

This is tricky for random samples as the results change

Allow for a wide tolerance to account for random distributions

Perhaps fixing the seed in openmc could help, openmc has a settings.seed value that defaults to 1

[shimwell]

testing the plotting routines ideas

check that plots result in file such as png files

check the axis ranges are correct

[LiamPattinson]

Hi Jonathan, for the first point, would I be right in thinking that you're interested in testing that TokamakSource is generating RZ coordinates correctly?

[shimwell]

That would be very nice, something that checks triangulation and elongation are correctly handled. What do you think @RemDelaporteMathurin ?

[RemDelaporteMathurin]

Hi @shimwell @LiamPattinson

So I think there are several things here and maybe we need seperated issues to make it clearer.

The TokamakSource class has a method called sample_sources() which creates a bunch of random (R, Z) coordinates based on the plasma geometry parameters (radii, elongation, triangularity, etc.). I think that's what @shimwell is refering to.

openmc-plasma-source/openmc_plasma_source/tokamak_source.py

Lines 198 to 215 in 3de9d21

	def sample_sources(self):
	"""Samples self.sample_size neutrons and creates attributes .densities
	(ion density), .temperatures (ion temperature), .strengths
	(neutron source density) and .RZ (coordinates)
	"""
	# create a sample of (a, alpha) coordinates
	a = np.random.random(self.sample_size) * self.minor_radius
	alpha = np.random.random(self.sample_size) * 2 * np.pi
	# compute densities, temperatures, neutron source densities and
	# convert coordinates
	self.densities = self.ion_density(a)
	self.temperatures = self.ion_temperature(a)
	self.neutron_source_density = neutron_source_density(
	self.densities, self.temperatures
	)
	self.strengths = self.neutron_source_density / sum(self.neutron_source_density)
	self.RZ = self.convert_a_alpha_to_R_Z(a, alpha)

One way to test this is to 1) test the method convert_a_alpha_to_R_Z() 2) for several sets of parameters (radii, elongation, etc.) test that the sampled coordinates are all in the shell produced by the plasma boundary equations :
R = major_radius + a * cos(alpha + triangularity * sin(alpha) ) + shafranov_shift
Z = elongation * a * sin(alpha)

openmc-plasma-source/openmc_plasma_source/tokamak_source.py

Line 178 in 3de9d21

def convert_a_alpha_to_R_Z(self, a, alpha):

[LiamPattinson]

Thanks @RemDelaporteMathurin and @shimwell, that's what I assumed the issue was about. I've managed to get such a test working, though with a few caveats:

• It currently only tests that the points generated are contained within a box with coordinates:
  • lower left (major_radius - minor_radius, -minor_radius * elongation)
  • upper right (major_radius + minor_radius, +minor_radius * elongation)
• It only confirms that the generated points are accurate for the tokamak parameters given in test_tokamak_source.py.

The first point is difficult to solve, as the plasma boundary is quite a complex shape to work with, but I might be able to constrain the test further with a bit more work. The second point should be quite easy to incorporate into my solution, as I'll just need to add a few more examples of tokamak geometries. Once I've done this I'll make a pull request.

[RemDelaporteMathurin]

Thanks for this. I guess that works but doesn't test the triangularity.

Perhaps we can focus on testing convert_a_alpha_to_R_Z. An "easy" way to do it is to create an inverse function convert_R_Z_to_a_alpha and check that convert_a_alpha_to_R_Z(convert_R_Z_to_a_alpha(R_test, Z_test)) == R_test, Z_test

So to create that inverse function, "simply" inverse the relations I gave above:

R = major_radius + a * cos(alpha + triangularity * sin(alpha) ) + shafranov_shift Z = elongation * a * sin(alpha)

Here's an idea of what it should look like:

[LiamPattinson]

I've tried looking at an inverse function, but I believe it can only ever be approximate as the a-alpha to RZ relations aren't bijective (at least according to "Tokamak D-T neutron source models for different plasma physics confinement modes", Fausser et al.). I believe I have an approximate solution working, but it'll take some further testing to make sure it's robust.

[RemDelaporteMathurin]

It isn't bijective as is but if you constraint a to be positive than each R, Z couple corresponds to one unique (a, alpha) point doesn't it?

[LiamPattinson]

I believe so, but the inversion may need to be performed numerically rather than analytically.

[RemDelaporteMathurin]

Ah maybe. Well even an approximate is better than what we have currently I suppose :-)
Thanks!