4d1e0a7f8e
feat(iace): GT-Bremse coverage — 59 expert measures + 7 hazard patterns
...
Systematic gap analysis of the Bremse ground-truth file (60 entries,
100 unique expert measures) revealed only ~5% library coverage. This
commit closes the documented gaps with concrete, norm-anchored
mitigations.
Library additions (M481-M539, 59 entries):
- M481-M482 Low-voltage isolation (>= 2,0 / 2x1,0 / 1,0 MOhm +
IP2X/IPXXB per EN 60204-1 Ziff. 6.2/8.2.3) — primary
trigger of this work
- M483-M485 Pneumatic safety (component pressure rating, hose
retention, depressurization per EN ISO 4414)
- M486-M490 Robot-cell access (tool-secured fence, dual-channel
door monitor, intentional restart, anti-trap inside
opening, HMI sight line per ISO 10218-2)
- M491-M493 Teach mode (key/password mode selector, safe reduced
speed <= 250 mm/s, hold-to-run with 3-stage enabler
per ISO 10218-1)
- M494-M500 Geometry constants (Safe Limited Position, reach-over
250 mm @ 2250 mm fence, conveyor opening >= 850 mm,
25 mm finger gap, band speed <= 100 mm/s per
EN ISO 13857 / EN 619)
- M501-M507 Enclosure load rating, gripper fail-safe, centring
gripper stop on door, MWF nozzle integration, floor
load capacity per DIN 1055-3
- M508-M517 Electrical cabling + PE protection (environment-rated,
drag chain, strain relief, 10 mm² Cu PE, dual PE,
monitoring, continuity check, class-II equipment,
SELV/PELV per EN 60204-1)
- M518-M522 RCD, cable cross-section, overcurrent in each active
conductor, IP22 water ingress, lockable main switch
- M523-M539 Teach-locked door, WZM door interlock, dual-channel
door switch, machining-doors-closed for aerosol
retention, post-NOTHALT release, >25 kg lifting aid
(DGUV 208-016), 95-120 cm control height, ergonomic
conveyor height, SDS/PSA reference, BA instructions
for depressurization/clamp release/max weight/pinch
warning/slip warning/dead-state cleaning
New hazard patterns (HP1710-HP1717):
floor overload, gripper failure throw, compressed-air injury in
machining cell, manual handling load + awkward posture, MWF skin
contact, live-cabinet cleaning short, pneumatic stored-energy.
Existing patterns rewired to the new measures: HP1600, HP1602-1606,
HP1610-1612, HP1620-1622, HP1630/1631/1633, HP1640/1641, HP1660/1661,
HP1675, HP1685, HP1688, HP1689, HP1698-1704.
Tooling:
- scripts/gt_measure_gap_analysis.py: 4-signal fuzzy matcher
(Jaccard, token recall, substring containment, norm-reference
overlap). Outputs markdown + JSON.
- gt_coverage_test.go: 23 expert-validated (GT-Nr, pattern, measure)
triples + a norm-reference presence test for every new expert
measure (no generic 'do X safely' entries allowed).
- .gitea/workflows/ci.yaml: new iace-gt-coverage job enforces
MIN_COVERAGE_PCT (70%) on Strong+Weak GT coverage; never lower
without explicit decision.
Coverage shift: 5% Strong -> 30% Strong, 0% -> 72% Strong+Weak.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com >
2026-05-16 13:08:52 +02:00
bf9d8a5ed3
fix(iace): resolve M-ID collisions for electrical/pressure patterns
...
6 supplementary measures (M410-M420) were silently overwritten by
metalworking duplicates in measureByID lookups, so robot-cell electrical
patterns resolved to chip-extraction/cleaning fallbacks instead of
equipotential bonding, creepage, EMC, or hose-burst protection. Rename
supplementary IDs to M475-M480 and rewire 13 affected pattern references
in robot_cell + robot_cell_ext.
HP1640 (direct contact with live parts, GT 2.2): priority 98->99, drop
RequiredEnergyTags gate so it fires in robot cells without an electrical
tag, expand mitigations to 5 concrete TRBS 2131 / IEC 60204-1 / EN 61140
measures (basic protection, double insulation, earthing, insulation
monitoring, equipotential bonding) — was previously losing to HP1688
even though HP1688 describes a different scenario.
HP1688 (touch voltage from potential differences): priority 98->96 so it
no longer outranks HP1640 for the direct-contact case; mitigations
expanded from M410-only to 4 concrete electrical measures.
Add regression tests pinning HP1640 contact-protection resolution and
M475 = Potentialausgleich. Existing TestGetProtectiveMeasureLibrary_-
UniqueIDs now actually enforces uniqueness (previously masked by
last-wins map override).
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com >
2026-05-16 10:12:55 +02:00
c075ecb721
feat(iace): Phase 2 — RAG-validierte Massnahmen-Erweiterung 241→428, Evidenztypen 25→50
...
- 62 Kap.-Verweise durch Themen-Deskriptoren ersetzt (0 verbleibend)
- measures_library_trbs.go: +85 Massnahmen (M217-M301) aus TRBS 1111/1201/2111/2121/2131/2141/2152
- measures_library_osha.go: +70 Massnahmen (M302-M371) aus OSHA Machine Guarding/LOTO/Electrical/Robots/Noise/Ergonomics/Pressure Vessels
- measures_library_trgs.go: +11 Massnahmen (M372-M382) aus TRGS 600/500/401/402/509/727/555
- measures_library_supplementary.go: +21 Massnahmen (M383-M403) aus RAG-Gap-Analyse gegen 6.141 extrahierte Obligations
- Brandschutz (8): TRGS 509/510/511/741/751 — Brandkonzept, Loeschanlagen, Brandmeldung, Fluchtweg
- Strahlung/Laser (5): OSHA TM Ch.6 — Laserklasse, Laserschutz, LSB, Absaugung
- TRBS 1115 Cybersecurity MSR (3): Cyber-GBU fuer Safety-SPS, Pruefung, Aenderungsmanagement
- TRBS 1112 Instandhaltung (3): GBU Instandhaltung, kontrollierte Handsteuerung, Fremdfirmenkoordination
- ASR (2): Sicherheitsbeleuchtung, Quetschschutz kraftbetaetigte Tueren
- tag_resolver.go: +25 Evidenztypen (E26-E50) — Materialzertifikat, EMV, Druckpruefung, Laser, ATEX, SIL/PL-Validierung, SBOM
Methodik: Systematische Obligation Extraction aus 152 Dokumenten (TRBS/TRGS/ASR/OSHA)
in Qdrant bp_compliance_ce (83.222 Chunks), Gap-Analyse gegen bestehende Bibliothek,
eigene Formulierungen (keine Normtext-Reproduktion).
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com >
2026-05-10 07:07:34 +02:00
Benjamin Admin