Files

Antoine af195c3a75 docs: add handoff document for trajectory optimization setup

2026-01-29 16:46:55 +00:00

5.4 KiB

Raw Blame History

Handoff: Set Up Trajectory-Based Optimization for M1 Mirror

From: Mario (Clawdbot)
To: Claude (Windows/LAC)
Date: 2026-01-29
Context: M1 GigaBIT mirror optimization using new Zernike Trajectory Method

Summary

A new Zernike Trajectory extractor has been implemented and validated. It provides mode-specific optimization metrics (coma, astigmatism, trefoil, etc.) integrated across the full 20°-60° operating range instead of discrete weighted sums.

Your task: Set up an Atomizer optimization study using TPE optimizer with total_filtered_rms_nm as the primary objective, logging mode-specific metrics.

What's Already Done

✅ Extractor implemented: optimization_engine/extractors/extract_zernike_trajectory.py
✅ Validated on M1 model: R² = 1.0 (physics model confirmed)
✅ Documentation: docs/physics/ZERNIKE_TRAJECTORY_METHOD.md
✅ Example config: docs/examples/trajectory_optimization_config.yaml

Pull latest from Gitea:

git pull origin main

What You Need to Set Up

1. Create New Study: `SAT3_Trajectory`

Location: studies/SAT3_Trajectory/ (or similar)

2. FEA Configuration

The FEA must solve 6 subcases with these elevation angles:

90° (manufacturing reference)
20° (measurement/polishing reference)
30° (new - for trajectory)
40° (primary operational)
50° (new - for trajectory)
60° (secondary operational)

Subcase labels in NX must be the angle numbers (e.g., LABEL = 20, LABEL = 40, etc.)

3. Extractor Configuration

Use the trajectory extractor:

extractors:
  - id: ext_trajectory
    name: "Zernike Trajectory"
    type: zernike_trajectory
    config:
      reference_angle: 20.0
      # Subcases auto-detected from OP2 labels

4. Objectives

Primary objective (what optimizer minimizes):

objectives:
  - id: obj_total
    name: "Total Integrated RMS"
    source:
      extractor_id: ext_trajectory
      output_name: total_filtered_rms_nm
    direction: minimize
    weight: 1.0

Secondary objectives (logged but not optimized, weight=0):

  - id: obj_coma
    name: "Coma RMS"
    source:
      extractor_id: ext_trajectory
      output_name: coma_rms_nm
    direction: minimize
    weight: 0.0

  - id: obj_astig
    name: "Astigmatism RMS"
    source:
      extractor_id: ext_trajectory
      output_name: astigmatism_rms_nm
    direction: minimize
    weight: 0.0

  - id: obj_trefoil
    name: "Trefoil RMS"
    source:
      extractor_id: ext_trajectory
      output_name: trefoil_rms_nm
    direction: minimize
    weight: 0.0

  - id: obj_spherical
    name: "Spherical RMS"
    source:
      extractor_id: ext_trajectory
      output_name: spherical_rms_nm
    direction: minimize
    weight: 0.0

5. Optimizer Configuration

Use TPE (recommended for fresh start):

optimizer:
  type: tpe
  config:
    n_trials: 100
    n_startup_trials: 15

6. Constraints (Optional but Recommended)

constraints:
  # Sanity check: linear model should fit well
  - id: con_r2
    name: "Trajectory fit quality"
    source:
      extractor_id: ext_trajectory
      output_name: linear_fit_r2
    type: soft
    operator: ">="
    threshold: 0.95

7. Design Variables

Use Antoine's existing wiffle tree and geometry parameters. Key ones likely include:

Wiffle tree radial positions (R1, R2, R3)
Wiffle tree angular offsets
Rib thickness
Back pocket geometry
Support pad locations

Ask Antoine for the current parameter names and ranges.

Validation Checklist

Before running optimization, verify:

FEA solves all 6 subcases (90, 20, 30, 40, 50, 60)
OP2 contains displacement results for all subcases
Subcase labels in OP2 are angle numbers (check with extractor)
Extractor returns valid results (run standalone test first)
R² ≈ 1.0 (confirms physics model holds)

Test command:

from optimization_engine.extractors.extract_zernike_trajectory import extract_zernike_trajectory

result = extract_zernike_trajectory('path/to/solution.op2')
print(f"R² = {result['linear_fit_r2']}")
print(f"Total RMS = {result['total_filtered_rms_nm']} nm")
print(f"Coma RMS = {result['coma_rms_nm']} nm")

Expected Outputs

After running the extractor, you should see:

total_filtered_rms_nm: ~4-10 nm (integrated across angles)
coma_rms_nm: ~3-10 nm
astigmatism_rms_nm: ~3-10 nm
linear_fit_r2: ~1.0

During optimization, track:

How total_filtered_rms_nm decreases
Which modes (coma, astig, etc.) improve most
Whether R² stays high (if it drops, something is wrong)

Key Files to Reference

File	Purpose
`docs/physics/ZERNIKE_TRAJECTORY_METHOD.md`	Full method documentation
`docs/examples/trajectory_optimization_config.yaml`	Example config
`optimization_engine/extractors/extract_zernike_trajectory.py`	Extractor code
`2-Projects/P04-GigaBIT-M1/Technical-Analysis/Tensor_Project/TECH-SPEC-Zernike-Trajectory-Optimization.md`	Physics derivation + validation results

Questions?

If anything is unclear, the tech spec in PKM has the full physics derivation and validation results. The extractor auto-detects angles from OP2 subcase labels, so as long as the FEA is set up correctly, it should "just work."

Good luck! 🚀

Handoff document by Mario (Clawdbot), 2026-01-29

5.4 KiB Raw Blame History