motors: name anchors after the machine, not their position in the file

`Geometry2`/`Computation2` are named for being the second machine defined, not for
the machine they describe. Nothing in the name says 5-phase or 40-slot, and nothing
distinguishes it from the other 5-phase variants people run on branches.

One module per anchor, named `<class>_<phases>f_<slots>s`:

  motors/motor1.py -> motors/synrm_3f_36s.py
  motors/motor2.py -> motors/synrm_5f_40s.py

Each exports plain `Geometry` and `Computation`, so the module name carries the
identity and the class names stay uniform. The numbering existed only because both
machines shared a namespace; per-anchor modules remove the need for it.

  from motors.synrm_5f_40s import Computation, Geometry

The name carries only what the anchor fixes for good -- class, phases, slots. Turns
are deliberately left out: `Nc` is routinely swept, so a name containing it goes
stale. Same reasoning for currents and bus voltage.

Also restores the anchor definition to CONTRIBUTING (it survives on the
`contributing` branch but was dropped in #23, leaving "anchor" used three times and
never defined, and the sentence "add an **motors**" mid-substitution), and fixes that
sentence.

No behaviour change: the two anchor modules were diffed against the `motor1.py` /
`motor2.py` they replace, with the class renames normalised away, and are identical
apart from a one-line docstring -- including the fixes from #28 and #29, which landed
in `motor2.py` while this branch was open and had to be carried across by hand.

Callers must update their import path; there is no compatibility shim, so a stale
import fails loudly rather than silently resolving to an aliased class.

Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>

Author: smidl

CONTRIBUTING.md

@@ -17,10 +17,26 @@ If you find yourself editing `machine_design/`on your own branch, stop — that'
 | `machine_design/optimization/` | design agnostic optimizers and geometry generators |
 | `machine_design/parallel_calculation/` | parallel runners for sweeping many solves at once |
 | `experiments/<name>/` | your run configs, the profile/objective you're studying, analysis notebooks, result artifacts, write-ups |
-| `motors/` | concrete machine definitions (`motor1.py`, `motor2.py`), each subclassing the base `Geometry`/`Computation` classes |
+| `motors/` | one module per **anchor** (a named machine), each subclassing the base `Geometry`/`Computation` classes |
 | `tests/` | the test suite |
 
-Add a new machine by subclassing an existing `motors/motor*.py` file instead of editing an existing class in place, so old results stay reproducible.
+### Anchors — machines are named, not numbered
+
+An **anchor** is a named machine definition: everything needed to instantiate and solve one physical machine — class, slots, phases, winding (turns, `Rs`), end-winding `Lew`, **rotor radii**, geometry generator + parameter bounds, FEA template.
+
+The name carries only what the anchor fixes for good — machine class, phases, slot count. Values that are routinely swept (turns, currents, bus voltage) stay out of it, so a name never goes stale.
+
+One anchor per module in `motors/`, named `<class>_<phases>f_<slots>s`:
+
+| module | machine |
+|---|---|
+| `motors/synrm_3f_36s.py` | 3-phase SynRM, 36 stator slots |
+| `motors/synrm_5f_40s.py` | 5-phase SynRM, 40 stator slots, dq1+dq3 excitation |
+
+Each exports `Geometry` and `Computation`, so the module name carries the identity and the class names stay uniform — `from motors.synrm_5f_40s import Geometry, Computation` says which machine you are solving, where `Geometry2` did not.
+
+- **Derive, don't fork.** To study a variant (add magnets, change turns), subclass an existing anchor and override only what differs — do **not** edit an anchor's class in place, so old results stay reproducible. A genuinely new machine is a new anchor module.
+- **An anchor must declare all of the above**, rotor radii included. They are the radial window the barrier generators draw into; an anchor that does not set them cannot have a rotor built for it.
 
 ## Pull requests
 
@@ -60,7 +76,7 @@ All compute goes through the SLURM queue on `bayes` — see [`bayes-queue/README
 
 **I want to optimize an existing machine with an existing method** → in `experiments/<name>/`, write a config that picks an anchor + a method + your objective/constraints, and an sbatch to run it. Touch nothing shared.
 
-**I want to study a new machine** → add an **motors** (PR to `main`); if it's a variant, derive it from an existing anchor. Then use it from your experiment.
+**I want to study a new machine** → add an **anchor** (PR to `main`); if it's a variant, derive it from an existing anchor. Then use it from your experiment.
 
 **I want a new optimization method** → add it to `machine_design/optimization/` against the interface (PR to `main`). Now anyone can use it on any anchor.
 

motors/motor1.py motors/synrm_3f_36s.pymotors/motor1.py renamed to motors/synrm_3f_36s.py

@@ -1,3 +1,5 @@
+"""Anchor `synrm_3f_36s`: 3-phase synchronous reluctance machine, 36 stator slots."""
+
 from ansys.aedt.core import Maxwell2d
 from ansys.aedt.core.modeler.modeler_2d import Modeler2D
 

motors/motor2.py motors/synrm_5f_40s.pymotors/motor2.py renamed to motors/synrm_5f_40s.py

@@ -1,12 +1,18 @@
+"""Anchor `synrm_5f_40s`: 5-phase synchronous reluctance machine, 40 stator slots.
+
+Derives from `synrm_3f_36s`, overriding only the slotting and winding it needs for five
+phases. Excitation is dq1 + dq3.
+"""
+
 import numpy as np
 from ansys.aedt.core import Maxwell2d
 
 from machine_design.designs.computation import ComputationBase
 
-from .motor1 import Geometry
+from .synrm_3f_36s import Geometry as BaseGeometry
 
 
-class Geometry2(Geometry):
+class Geometry(BaseGeometry):
     def set_geom_params(self):
         super().set_geom_params()
         self.geom_params["SlotNumber"] = "40"
@@ -22,7 +28,7 @@ def set_winds_params(self):
         self.wind_params["Nc"] = "113"
 
 
-class Computation2(ComputationBase):
+class Computation(ComputationBase):
     def set_oper_params(self):
         f = 50  # [Hz]
         RotSpeed = 60 * f / self.geometry.PolePairs  # [rpm]

notebooks/run.py

@@ -15,7 +15,7 @@
     plot_barriers,
     save_params,
 )
-from motors.motor1 import Computation, Geometry
+from motors.synrm_3f_36s import Computation, Geometry
 
 config = load_config()
 aedt_version = config["aedt_version"]

notebooks/run_optimization.py

@@ -23,7 +23,7 @@
     objective,
     objective_transform,
 )
-from motors.motor1 import Computation, Geometry
+from motors.synrm_3f_36s import Computation, Geometry
 
 torch.set_default_dtype(torch.float64)
 

tests/test_ansys_regression.py

@@ -20,7 +20,7 @@
 from machine_design.config import load_config
 from machine_design.designs.design import Design as LiveDesign
 from machine_design.optimization.generators import HacklGenerator_OneLambda
-from motors.motor1 import Computation, Geometry
+from motors.synrm_3f_36s import Computation, Geometry
 
 pytestmark = pytest.mark.ansys
 

tests/test_design.py tests/test_synrm_3f_36s.pytests/test_design.py renamed to tests/test_synrm_3f_36s.py

@@ -2,7 +2,7 @@
 
 import pytest
 
-from motors.motor1 import Computation, Geometry
+from motors.synrm_3f_36s import Computation, Geometry
 
 
 @pytest.fixture

tests/test_design2.py tests/test_synrm_5f_40s.pytests/test_design2.py renamed to tests/test_synrm_5f_40s.py

@@ -1,22 +1,22 @@
 import pytest
 
-from motors.motor1 import Geometry
-from motors.motor2 import Computation2, Geometry2
+from motors.synrm_3f_36s import Geometry as BaseGeometry
+from motors.synrm_5f_40s import Computation, Geometry
 
 
 @pytest.fixture
 def base_geometry():
-    return Geometry()
+    return BaseGeometry()
 
 
 @pytest.fixture
 def geometry2():
-    return Geometry2()
+    return Geometry()
 
 
 @pytest.fixture
 def computation2(geometry2):
-    return Computation2(geometry2)
+    return Computation(geometry2)
 
 
 def test_geom_params_overrides_slot_number_only(base_geometry, geometry2):