Case study A · CalculiX-certified

The bracket that survives where the conventional part fails.

solver run 2026-07-24 · check these numbers against their artifacts

An additively-manufactured payload-interface bracket for a declared, generic 200 kg-payload rotorcraft load envelope — evolved from scratch, certified by the same solver and material as its conventional baseline.

The setup

The 200 kg payload figure is a public scale anchor. The envelope, load factors and torque were frozen in a preregistration before the first generation ran — they are PIE-declared generic values, not any manufacturer's. The claim could not be adapted to the result.

Load cases (all evaluated, worst case governs): +3.5 g vertical, −1 g vertical, ±1.5 g lateral / fore-aft, 500 N·m torsion about the mount axis
Safety-factor bar: 1.5 against the engine's AlSi10Mg yield strength (LPBF as-built material library entry)
Boundary conditions: bearing loads at the bolt bores, not clamped faces
Manufacturability: LPBF gates — local wall thickness, trapped voids, powder access, support fraction

The result

A side-by-side comparison of the conventional plate-and-gusset bracket and PIE bracket g00095. The first column names the property measured; the second and third columns give its value for each design.
	Conventional plate-and-gusset	PIE bracket g00095
Governing case	torsion	torsion
Governing safety factor	0.427 — fails	2.055 — passes
Worst other case	fore-aft, SF 1.64	vertical +3.5 g, SF 7.40
First structural mode	54.4 Hz	171.6 Hz
Mass	0.946 kg	1.165 kg
Printable (LPBF gates)	fails support gate	passes all gates

Both columns: CalculiX static + eigenfrequency on Gmsh tetrahedral meshes, identical load cases, identical AlSi10Mg material properties (same engine material-library entry). validated_sidecar

The honest trade, stated plainly: the evolved bracket is 23% heavier. This is not a lightweighting result. It is a survives-the-full-envelope result: the conventional design fails torsion outright at less than half the required safety factor, and the evolved part clears every declared case with margin. When a certified mass win exists, it will appear here with the same receipts. It does not exist today.

What was preregistered

Envelope, load cases, excitation spectrum, safety-factor bar, the fixed baseline, and the exact claim ceiling — frozen in bracket_demonstrator_prereg.json before iteration 1. The preregistration also lists what this campaign may never claim: flight or production readiness, a fatigue life, or a percent-lighter headline unless the candidate beats the baseline on every gate.

Conventional plate-and-gusset baseline bracket, rendered from its certification STL — The conventional plate-and-gusset baseline, rendered from its certification STL — the design that fails torsion at SF 0.427. The winner's mesh lives in its solver outputs; its render will appear here when the geometry export is regenerated.

the winner's certification record certification.json ✓ PASS

{
  "candidate_id": "bracket_g00095_21d1fa03afc24",
  "evidence_tier": "validated_sidecar_calculix_gmsh",
  "governing_case": "torsion",
  "governing_safety_factor": 2.0554,
  "load_cases": [
    { "case": "vertical_+3.5g", "safety_factor": 7.4004 },
    { "case": "vertical_-1g",   "safety_factor": 25.9015 },
    { "case": "lateral_1.5g",  "safety_factor": 24.8130 },
    { "case": "foreaft_1.5g",  "safety_factor": 10.7075 },
    { "case": "torsion",       "safety_factor": 2.0554 }
  ],
  "passes": true, "failures": []
}

ledger record e9ff00d8… · gmsh 31,487 nodes / 126,665 tets

Every number above traces to a CalculiX output inside a hashed, append-only ground-truth ledger record.

Back to evidence Run this on your component