Engineering Highlights
Two real production-data findings from building this engine, and the fixes behind them — see CERTIFICATION_REPORT.md and PROJECT_STATUS.md in the source repos for the full write-ups.
REV-41 — Authorized-Overtime-Window PPT
§9.3 requires PPT for an OVERTIME shift to come from the authorized overtime window, not the standard shift window. The engine originally treated the whole authorized window as a subtracted deduction instead of PPT's basis — confirmed on this exact real shift, machine 15, 2026-06-16 (Holiday, with an authorized Overtime entry for machines 15 and 86).
PPT
Availability
OEE
Both real OVERTIME machine-shifts in the dataset (2026-06-16 Day)
| Machine | State | PPT | Downtime | Availability | Performance | Quality | OEE | Flags |
|---|---|---|---|---|---|---|---|---|
| MC015 | Overtime | 11h 10m | 0m | 100.0% | 81.0% | 100.0% | 81.0% | Overtime |
| MC086 | Overtime | 11h 10m | 0m | 100.0% | 73.1% | 100.0% | 73.1% | Overtime |
REV-42 — Consecutive-IDLE Detection
A machine left IDLE_NOT_SCHEDULED for more than 6 consecutive shifts, while comparable machines keep producing, gets an Information flag (F-IDLE). Machine 142's real June timeline has a genuine 14-shift streak — long enough to show the exact threshold boundary on real data.
2026-06-10 Night — 6th consecutive idle shift
streakLength: 6 · flagged: false
—
2026-06-11 Day — 7th consecutive idle shift
streakLength: 7 · flagged: true
"Beyond N shifts" means the streak must exceed the threshold — the 7th consecutive shift is the first one flagged, not the 6th. Real always-producing Circular peers (machines 1, 3, 9, 10) confirmed producing every one of these 14 shifts, satisfying the "comparable machines still producing" check.