Files
aimsight-labeling/auto/analyses/long-labels/WILDSTRUBEL/analysis.md
T
maximeandClaude Opus 5.5 ddb17c2189 auto: labels requalified under the 5-noise rules (only what stands out from the noise): 32 analyses refined, the 6 paused periods verified, the long labels requalified with the fixes the period analysts found; 1153 -> 961 labels
- 32 analyses: each label kept, rewritten or removed by its analyst (section Requalification of the report; mm magnitudes where the targets have them)
- SNCF01 x3, WILDSTRUBEL 2026-09-01 -> 2026-10-06 x3: verified from their existing labels under the new rules
- long labels (no analysis): auto/analyses/long-labels/<STATION>/; NORDEND-169 and -195 run to 07/11, so NORDEND-283 and -284 (their continuation) are removed
- reading-rules.md, stations.md: the citations of removed or rewritten labels point to their successor, or to [expert]

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
2026-11-17 23:58:00 +01:00

10 lines
4.0 KiB
Markdown
Raw Blame History

This file contains ambiguous Unicode characters
This file contains Unicode characters that might be confused with other characters. If you think that this is intentional, you can safely ignore this warning. Use the Escape button to reveal them.
# Requalification of the long labels: WILDSTRUBEL
- **WILDSTRUBEL-46** (Target issue, reference R3, 07/09 00:00 → 01/10 00:00): **kept** as is. R3 is rejected at 445 of 1088 measurements (41 %), almost all by day: 05–16 UTC, 42–43/44 on 07–09/09, 32–37/44 on 22–29/09 (23/29 on the restart day 22/09, 25/42 on 30/09: "a little less often in the last days of September" holds); at night 0–5 per night. Every measured day of the range, its own rejections (the other references over the same range: PA_8 121, R2 67, R1 31, PB3_2 16, PA1_2 15). A reference rejected every day under the ordinary daylight regime is a Target issue even without effect on the targets (`families.md`, "Target issue or episode of another family"; catalogue: "A target rejected every night or every day").
- **WILDSTRUBEL-69** (No data, 09/09 21:45 → 22/09 08:45): **kept** as is. No measurement of any sensor between these two instants (1195 missing steps; last measurement 09/09 21:45, first 22/09 08:45); no image (09/09 21:45 → 22/09 08:39) and no metric (09/09 22:00 → 22/09 07:00). Battery 9 % and BusVoltage 11.5 V at 09/09 21:00 (other nights 12.9–13.0 V at 25–60 %), then 100 % and 14.3 V at 22/09 08:00: the energy signature stated holds.
- **WILDSTRUBEL-70** (Displacement, 09/09 21:45 → 22/09 09:00): **kept** as is. `ranges.py --k 5 --min-measurements 6 --outside 2026-09-09T21:45Z/2026-10-01T00:00Z` (level and noise of the days before the outage, 01–09/09): from the first measurement after the outage, PB1_2 corr_y −11.3 noises (−0.19 px, about −1.1 mm) for 170 measurements, 22/09 09:00 → 24/09 04:45; PB_5 corr_x −6 to −7 noises (about −1.0 mm) 22/09 09:00 → 24/09; PB2_2 corr_y −7.9 noises (about −0.8 mm). Night levels (`days.py`), night 09/09 → night 22/09: PB1_2 corr_y −13.9 noises (−1.3 mm), PB2_2 corr_y −6.4 (−0.6 mm), PB_5 corr_x −5.0 (−0.7 mm); PA_9 corr_y +2.7, PA2_2, PA3_2, PB4 corr_x −1.6 to −2.3 noises (under the threshold on their own, named in the text as smaller steps, not wrong). At least one target beyond 5 noises: the step holds; its bounds follow the rule of a jump hidden in an outage.
- **WILDSTRUBEL-130** (Target issue, reference R3, 01/10 00:00 → 01/11 00:00): **kept** as is. 325 of 432 measurements rejected (75 %): by day 29–40 of 31–43 per day from 01 to 09/10 (31/31 on 07/10, 29/31 on 09/10), the only daytime measurement of 15/10 rejected; at night 0–2 per night (1 of 38 on 01/10). The other references over the range: PA1_2 90, PA_8 86, R2 40, R1 25, PB3_2 13. Its own rejections every measured day: Target issue kept. No measurement after 15/10 08:00 except one on 31/10 20:15 (night, accepted), as the text says.
- **WILDSTRUBEL-199** (No data, 15/10 08:00 → 31/10 20:15): **corrected** (description), review removed. Bounds hold: last measurement 15/10 08:00, next 31/10 20:15, no measurement in between. Metrics: two isolated hours on 25/10 (01:00, 03:00), then back on 30/10 18:00; BatteryPercent 78 % at 30/10 18:00, 80–84 % all night, 94 % at 31/10 05:00, 100 % from 31/10 06:00 to 15:00 (BusVoltage 13.1–13.2 V then 14.0–14.2 V). Correction applied (found by the period analyst): "the battery full" at the 30/10 return becomes "well charged but not full; full only on 31/10 in the morning"; images on 31/10 04:00–13:00 hourly, first displacement measurement 31/10 20:15 ("until that evening" added).
Sources read: `tools/GUIDE.md`, `tools/DATA.md`, `labeling/families.md`, `labeling/labeling-rules.md`, `labeling/reading-rules.md`, `labeling/stations.md`, `input.jsonl`, the NORDEND long-labels outputs (format). Commands: `scripts/r3_rates.py` (rejections per reference and per day/night, through `tools/scripts/data.py`), `inventory.py` (missing steps, image gaps), `metrics.py --names BusVoltage,BatteryPercent(,BatteryState) --metrics-only` on 08–23/09 and 24/10–01/11, `days.py` night levels 05–26/09 (`work/days_70.txt`), `ranges.py --k 5 --outside` 01–25/09 with the mm columns (`work/ranges_70.txt`). Labels written by `scripts/write_labels.py`; `validate.py`: 0 errors, 0 overlaps.