Atomizer/projects/hydrotech-beam/playbooks/DOE.md

Playbook — DOE (Phase 1 Landscape)

Scope

Design of Experiments for Hydrotech Beam: baseline + LHS samples to map design space.

Standard flow

1. Prepare clean state (archive previous results if needed)
2. Run DOE with explicit backend
3. Validate outputs
4. Evaluate gate criteria
5. Decide next phase (rerun bounds/constraints or move to TPE)

Commands

Real engineering DOE

python .\run_doe.py --backend nxopen --model-dir "<PATH_TO_NX_MODELS>" --clean --study-name hydrotech_beam_doe_phase1_real

Debug DOE only (synthetic)

python .\run_doe.py --backend stub --clean --study-name hydrotech_beam_doe_phase1_stub

Interpretation checklist

• geo_infeasible high: geometry bounds/checks may be too broad or too strict
• fully_feasible = 0: constraints may be too strict for current design space
• Very low runtime and suspiciously smooth outputs: likely stub execution

Decision branch

• If gate passed: proceed to TPE seeded by feasible points
• If gate failed:
  • adjust constraints and/or DV bounds
  • rerun clean DOE

Documentation requirement

After each DOE run, update:

• run record
• decision and rationale
• next action owner