Poincare Boundary Condition #782

YigitElma · 2023-12-03T02:55:37Z

Resolves #204
Resolves #773

This PR focuses on solving equilibrium by giving a Poincare section at zeta 0 surfaces instead of giving the last closed flux surface, LCFS.

This method gives exactly same results for high NFP >12, but different results for low NFP cases. Note: as far as my observation, the difference is caused by the NFP, but it might caused by something else that I am missing.

Tasks,

Add tests (make them run in reasonable time)
make sure API works with poincare (try to resolve inaccuracy issue) (kind of done)
make sure we don't use eq.surface anywhere expecting explicitly a toroidal surface
move utility functions for making poincare XS from an equilibrium to proper part (utils.py getters.py) Moved it to Equilibrium class i.e. eq.set_poincare_equilibrium() gives you separate equilibrium
Get poincare to work with BoundaryRSelfConsistency (and Z), and remove poincareR and poincareZ objectives
Add get_fixed_xsection_constraints() utility
Add new fixed and self consistency objectives for Poincare BC

Previous tasks that I am not sure about,

remove bdry_mode from Equilibrium and instead add new DOFs of the equilibrium Rp_lmn Zp_lmn Lp_lmn and xsection.
When creating an Equilibrium with a Poincare section, check that initial axis is consistent with the equilibrium (it should just be a circular axis at the poincare section axis position) (I think it is like that but I don't know what Dario meant exactly)
decide if LambdaGauge is needed or not with Poincare equilibria

github-actions · 2024-01-16T20:41:48Z

|             benchmark_name             |         dt(%)          |         dt(s)          |        t_new(s)        |        t_old(s)        | 
| -------------------------------------- | ---------------------- | ---------------------- | ---------------------- | ---------------------- |
 test_build_transform_fft_lowres         |     +1.07 +/- 1.43     | +5.58e-03 +/- 7.47e-03 |  5.29e-01 +/- 6.3e-03  |  5.24e-01 +/- 4.0e-03  |
-test_equilibrium_init_medres            |    +11.22 +/- 0.85     | +5.00e-01 +/- 3.79e-02 |  4.96e+00 +/- 2.1e-02  |  4.46e+00 +/- 3.2e-02  |
-test_equilibrium_init_highres           |    +26.54 +/- 3.82     | +1.55e+00 +/- 2.23e-01 |  7.39e+00 +/- 1.4e-01  |  5.84e+00 +/- 1.7e-01  |
 test_objective_compile_dshape_current   |     -2.26 +/- 5.35     | -9.53e-02 +/- 2.26e-01 |  4.12e+00 +/- 1.3e-01  |  4.22e+00 +/- 1.8e-01  |
-test_objective_compute_dshape_current   |    +34.64 +/- 1.84     | +1.80e-03 +/- 9.58e-05 |  7.00e-03 +/- 8.6e-05  |  5.20e-03 +/- 4.2e-05  |
 test_objective_jac_dshape_current       |     +4.53 +/- 7.41     | +1.91e-03 +/- 3.13e-03 |  4.42e-02 +/- 2.4e-03  |  4.23e-02 +/- 2.0e-03  |
 test_perturb_2                          |     +3.59 +/- 2.15     | +7.07e-01 +/- 4.23e-01 |  2.04e+01 +/- 2.8e-01  |  1.97e+01 +/- 3.2e-01  |
 test_proximal_freeb_jac                 |     +0.11 +/- 2.09     | +7.89e-03 +/- 1.51e-01 |  7.21e+00 +/- 9.4e-02  |  7.21e+00 +/- 1.2e-01  |
 test_solve_fixed_iter                   |     +1.20 +/- 2.51     | +3.88e-01 +/- 8.08e-01 |  3.26e+01 +/- 6.7e-01  |  3.22e+01 +/- 4.5e-01  |
 test_LinearConstraintProjection_build   |     +1.80 +/- 2.95     | +1.86e-01 +/- 3.04e-01 |  1.05e+01 +/- 1.3e-01  |  1.03e+01 +/- 2.7e-01  |
 test_build_transform_fft_midres         |     +4.42 +/- 6.26     | +2.82e-02 +/- 3.99e-02 |  6.65e-01 +/- 9.8e-03  |  6.37e-01 +/- 3.9e-02  |
 test_build_transform_fft_highres        |     +4.92 +/- 4.14     | +4.85e-02 +/- 4.08e-02 |  1.03e+00 +/- 3.6e-02  |  9.86e-01 +/- 1.9e-02  |
-test_equilibrium_init_lowres            |    +14.68 +/- 2.24     | +6.39e-01 +/- 9.77e-02 |  4.99e+00 +/- 4.0e-02  |  4.35e+00 +/- 8.9e-02  |
 test_objective_compile_atf              |     +1.65 +/- 3.73     | +1.40e-01 +/- 3.17e-01 |  8.64e+00 +/- 3.1e-01  |  8.50e+00 +/- 4.4e-02  |
-test_objective_compute_atf              |    +16.51 +/- 4.87     | +2.71e-03 +/- 7.99e-04 |  1.91e-02 +/- 6.0e-04  |  1.64e-02 +/- 5.3e-04  |
 test_objective_jac_atf                  |     +2.99 +/- 1.84     | +5.94e-02 +/- 3.65e-02 |  2.05e+00 +/- 2.2e-02  |  1.99e+00 +/- 2.9e-02  |
 test_perturb_1                          |     +5.23 +/- 4.14     | +7.94e-01 +/- 6.30e-01 |  1.60e+01 +/- 3.9e-01  |  1.52e+01 +/- 4.9e-01  |
 test_proximal_jac_atf                   |     +0.83 +/- 0.71     | +6.92e-02 +/- 5.91e-02 |  8.37e+00 +/- 5.3e-02  |  8.31e+00 +/- 2.7e-02  |
 test_proximal_freeb_compute             |     +3.47 +/- 1.91     | +6.51e-03 +/- 3.59e-03 |  1.94e-01 +/- 2.0e-03  |  1.88e-01 +/- 3.0e-03  |
 test_solve_fixed_iter_compiled          |     +2.99 +/- 2.80     | +6.21e-01 +/- 5.81e-01 |  2.14e+01 +/- 4.2e-01  |  2.08e+01 +/- 4.1e-01  |

desc/objectives/linear_objectives.py

desc/optimize/optimizer.py

desc/basis.py

f0uriest

I think the better API would be to introduce new degrees of freedom Rp, Zp etc like we have for the LCFS and axis, and then constrain those to be self consistent.

This would eliminate the boundary_mode thing, and allow more flexibility for setting up constraints.

Each equilibrium, in addition to having a .surface and .axis attribute would have a .poincare or .section property with the ZernikeRZToroidalSurface object. This would then have an assigned zeta value so we wouldn't need to pass around zeta to different functions.

YigitElma · 2024-02-01T04:52:25Z

I think the better API would be to introduce new degrees of freedom Rp, Zp etc like we have for the LCFS and axis, and then constrain those to be self consistent.

This would eliminate the boundary_mode thing, and allow more flexibility for setting up constraints.

Each equilibrium, in addition to having a .surface and .axis attribute would have a .poincare or .section property with the ZernikeRZToroidalSurface object. This would then have an assigned zeta value so we wouldn't need to pass around zeta to different functions.

@f0uriest I made some changes to the API. First of all, I created a new class called PoincareSurface (which is an inherited class of ZernikeRzToroidalSection) to be able to store lambda basis and coefficients in the section. That class needs more cleaning but my main aim is to decide which type of objectives to use by checking the eq.surface type. This can be either FourierRZToroidalSurface or PoincareSurface. I changed the previous eq.bdry_mode stuff to that. Also, I am using the zeta of the surface now, so no need to pass zeta value to optimizer and objectives anymore.

ddudt · 2024-02-07T20:39:09Z

Let's get rid of support for the Poincare BC from a text input file in this PR

desc/objectives/getters.py

…r other tests, create new input file

…optimizable

YigitElma · 2025-02-06T07:14:18Z

I changed the hacky way with another approach where I declare some parameters non-optimizable. To do so, I added a new function inside desc.optimizable called make_nonoptimizable and a setter for cls.optimizable_params. If this works, it should solve the speed issue.

CHANGELOG.md

YigitElma · 2025-02-06T07:41:06Z

docs/notebooks/tutorials/input.HELIOTRON_desc_no_continuation

@@ -3,34 +3,33 @@
 # on 11/20/2023 at 14:54:04.
 # For details on the various options see https://desc-docs.readthedocs.io/en/stable/input.html

-# global parameters 


Apart from bdry_mode = lcfs line, rest is due to black formatter

…izable by default, if FixSection added then they become optimizable

dpanici · 2025-02-06T20:11:46Z

desc/continuation.py

@@ -693,6 +695,8 @@ def solve_continuation(  # noqa: C901
        NotImplementedError,
        "Continuation method with anisotropic pressure is not currently supported",
    )
+    # TODO: add warning for cases with Poincare BC. This needs a better


is this done above?

f0uriest · 2024-12-18T19:36:44Z

desc/objectives/normalization.py

@@ -12,6 +12,11 @@ def compute_scaling_factors(thing):

    scales = {}

+    # change this to also work with Poincare XS where bdry is  ZernikePolynomial


I think you can get rid of this comment

f0uriest · 2024-12-18T19:38:05Z

desc/perturbations.py

@@ -67,6 +70,20 @@ def get_deltas(things1, things2):  # noqa: C901
            if not jnp.allclose(a2.Z_n, a1.Z_n):
                deltas["Za_n"] = a2.Z_n - a1.Z_n

+    if "surface_poincare" in things1:


we should call this section or xsection or whatever name we use elsewhere (pick one and use it consistently)

f0uriest · 2024-12-18T19:46:07Z

tests/test_perturbations.py

+    eq_poin.change_resolution(eq.L, eq.M, eq.N)
+    eq_poin.axis = eq_poin.get_axis()
+    eq_poin.surface = eq_poin.get_surface_at(rho=1)


why are these necessary? Shouldn't these already be set when you create the equilibrium?

f0uriest · 2024-12-18T19:56:43Z

desc/equilibrium/equilibrium.py

+
+        constraints = maybe_add_self_consistency(self, constraints)
+        # TODO: make this more general and probably put it somewhere else
+        if not is_any_instance(constraints, SectionRSelfConsistency):


if we end up keeping this, it should be in Optimizer.optimize not here, otherwise it won't get called in some cases

f0uriest · 2024-12-18T19:56:51Z

desc/equilibrium/equilibrium.py

+
+        constraints = maybe_add_self_consistency(self, constraints)
+        # TODO: make this more general and probably put it somewhere else
+        if not is_any_instance(constraints, SectionRSelfConsistency):


f0uriest · 2024-12-18T19:57:28Z

desc/equilibrium/equilibrium.py

+            warnif(
+                zeta != 0,
+                UserWarning,
+                "The lambda value of the returned surface will not be correct "


still need to do this

YigitElma self-assigned this Dec 3, 2023

PlasmaControl deleted a comment from github-actions bot Dec 19, 2023

PlasmaControl deleted a comment from github-actions bot Jan 16, 2024

dpanici requested review from f0uriest, ddudt and dpanici January 17, 2024 20:33

YigitElma marked this pull request as ready for review January 17, 2024 22:45

f0uriest reviewed Jan 26, 2024

View reviewed changes

desc/objectives/linear_objectives.py Outdated Show resolved Hide resolved

f0uriest reviewed Jan 26, 2024

View reviewed changes

desc/optimize/optimizer.py Outdated Show resolved Hide resolved

f0uriest reviewed Jan 26, 2024

View reviewed changes

desc/basis.py Outdated Show resolved Hide resolved

f0uriest requested changes Jan 26, 2024

View reviewed changes

YigitElma marked this pull request as draft January 31, 2024 20:14

dpanici removed request for ddudt and dpanici January 31, 2024 20:39

PlasmaControl deleted a comment from review-notebook-app bot Feb 1, 2024

github-actions bot reviewed Feb 8, 2024

View reviewed changes

desc/objectives/getters.py Outdated Show resolved Hide resolved

desc/objectives/getters.py Outdated Show resolved Hide resolved

desc/objectives/getters.py Outdated Show resolved Hide resolved

github-actions bot reviewed Feb 8, 2024

View reviewed changes

desc/objectives/getters.py Outdated Show resolved Hide resolved

YigitElma and others added 3 commits February 13, 2024 23:59

Merge branch 'master' into dp/poincare

c60852c

fix node pattern error test, SOLOVEV poincare shouldn't give error fo…

8f92b7a

…r other tests, create new input file

change Poincare solution method for tests to my version

fd82c12

YigitElma and others added 5 commits January 30, 2025 18:02

Merge branch 'master' into dp/poincare

995369a

add make_nonoptimizable function

e45365e

update how I manage the self consistency stuff regarding the make_non…

b1f9596

…optimizable

maybe remove noqa

816a409

Merge branch 'master' into dp/poincare

f261a0d

fix the arg not in list issue

7a18a95

YigitElma commented Feb 6, 2025

View reviewed changes

CHANGELOG.md Outdated Show resolved Hide resolved

YigitElma commented Feb 6, 2025

View reviewed changes

YigitElma and others added 5 commits February 6, 2025 13:27

also add make_optimizable function, make Poincare variables not optim…

d227d09

…izable by default, if FixSection added then they become optimizable

Merge branch 'master' into dp/poincare

6a8c81d

Update CHANGELOG.md

72863ec

remove xsection updater from perturb

b713bbe

Merge remote-tracking branch 'origin/master' into dp/poincare

1549d44

YigitElma requested review from a team, rahulgaur104, f0uriest, ddudt, dpanici, sinaatalay and unalmis and removed request for rahulgaur104, ddudt, unalmis, dpanici, f0uriest and a team February 6, 2025 20:06

update changelog

94d909d

dpanici reviewed Feb 6, 2025

View reviewed changes

f0uriest reviewed Feb 10, 2025

View reviewed changes

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Poincare Boundary Condition #782

Poincare Boundary Condition #782

YigitElma commented Dec 3, 2023 •

edited

Loading

github-actions bot commented Jan 16, 2024 •

edited

Loading

f0uriest left a comment

YigitElma commented Feb 1, 2024

ddudt commented Feb 7, 2024

YigitElma commented Feb 6, 2025

YigitElma Feb 6, 2025

dpanici Feb 6, 2025

f0uriest Dec 18, 2024

f0uriest Dec 18, 2024

f0uriest Dec 18, 2024

f0uriest Dec 18, 2024

f0uriest Dec 18, 2024

f0uriest Dec 18, 2024

		@@ -12,6 +12,11 @@ def compute_scaling_factors(thing):

		scales = {}

		# change this to also work with Poincare XS where bdry is ZernikePolynomial

Poincare Boundary Condition #782

Are you sure you want to change the base?

Poincare Boundary Condition #782

Conversation

YigitElma commented Dec 3, 2023 • edited Loading

github-actions bot commented Jan 16, 2024 • edited Loading

f0uriest left a comment

Choose a reason for hiding this comment

YigitElma commented Feb 1, 2024

ddudt commented Feb 7, 2024

YigitElma commented Feb 6, 2025

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

YigitElma commented Dec 3, 2023 •

edited

Loading

github-actions bot commented Jan 16, 2024 •

edited

Loading