Files
aimsight-labeling/auto/analyses/SNCF01_2026-10-04T0000_2026-10-14T0645/report.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

8.6 KiB

SNCF01: check of the labels of 4 to 14 October

Your question: check the 30 labels already placed between 4 October 00:00 and 14 October 06:45, under the new rule that only what stands out from the noise is labelled. For each label: keep, correct, remove, or add.

Verification

Counts:

  • 14 kept
  • 14 corrected
  • 2 removed
  • 0 added

Result: 28 labels.

Changes:

  • SNCF01-8: bounds tightened to 08:45→16:30 (was 08:15→17:15). The text now describes that day's targets. Reason: the agitation goes beyond the noise only between 09:00 and 16:15.
  • SNCF01-11: end moved from 16:45 to 15:15. Reason: the last movement beyond the noise is at 15:00; the 16:00 train has its own label.
  • SNCF01-13: removed. Reason: at 06/10 06:45 the low targets are only missing points. The targets that remain move 3 noises at most.
  • SNCF01-14: bounds changed to 07:00→13:15 (was 07:15→14:30). Reason: the agitation stops at 13:00.
  • SNCF01-15: end moved from midnight to 07/10 03:00, and the text rewritten. Reasons: a cycle never ends at midnight; the camera reaches its night position about 03:00; the old text had a copy error.
  • SNCF01-16: end moved from 21:00 to 08/10 04:15, and the text rewritten. Reasons: P5 is still off its level until 01:00, and the camera keeps coming back until 04:15. The day was weak and overcast; the old text was copied from a sunny day.
  • SNCF01-17: text corrected. Reason: BG.3, F4.5 and F4.6 do move a little beyond their noise at the 13:00 train. The old text said they "barely move".
  • SNCF01-24: removed. Reason: on 9 October (overcast) one single value reaches 5.1 noises. The rest is a missing point and the start of the thermal cycle.
  • SNCF01-27: bounds changed to 07:00→13:00 (was 07:15→14:15). Reason: the agitation ends at 12:45.
  • SNCF01-28: bounds changed to 10:00→15:30 (was 08:15→16:00), and the text now says weak and overcast. Reason: only a few values just beyond the noise.
  • SNCF01-29: end moved from midnight to 12/10 02:45, and the text rewritten. Reasons: same as -15; the camera stops coming back at 02:45; the old text had a copy error.
  • SNCF01-30: text corrected. Reason: all P targets are rejected at 14:15; only BG.2, BG.3 and F4.5 jump.
  • SNCF01-31: text corrected. Reason: CV.4 is not rejected; it jumps by about 1.3 mm for one measurement. The old text said "nothing else moves".
  • SNCF01-32: bounds changed to 07:30→14:45 (was 07:00→16:15). Reason: the agitation runs from 07:45 to 14:30.
  • SNCF01-34: text corrected. Reason: same as -30.
  • SNCF01-36: bounds changed to 07:45→14:00 (was 08:30→14:30). Reason: the agitation runs from 08:00 to 13:45.

Spanning labels:

  • None is wrong inside the range.
  • SNCF01-5 (F4.3) says about half of its measurements are rejected. Here it is about a quarter. The defect holds, but the text overstates it for this period.

In short

Most labels hold. The two removed labels were below the noise threshold. The corrections are tighter bounds and texts that had been copied from another day.

The curves cannot be trusted:

  • by day on sunny days, when the low targets shake;
  • at the six train passages, where the targets that stay visible jump by up to about 7 mm for one measurement.

Every day the camera warms and moves, and the compensation leaves up to about 1.4 mm on P7. No real movement of the wall is seen in these ten days, apart from the slow month-long drift already labelled.

What happened

The camera's daily cycle (every day). In the sun the camera turns a little and comes back at night. The compensation does not erase all of it:

  • P7 drifts by about 1 to 1.4 mm on sunny days (4, 5, 6 and 12 October), and less on overcast days.
  • P1 moves the other way.
  • On 12 October the curves do not come back during the night; they come back on the 14th.

There is no temperature sensor on this station, so the link to the heat is likely, not confirmed. The structure itself is not moving.

Daytime shaking of the low targets (sunny days). From morning to early afternoon, the corrected curves of the low targets jump from one measurement to the next and come back at once. These are mostly CV.4, CV.2 and F4.2, up to about 2 mm. Their recognition error rises at the same time. Some P targets jump together for one measurement. At night it stops. The shaking appears only on days with sun (shadows on the wall in the images). The cause is likely hot air shimmering above the tracks. It is weak on 11 and 13 October and does not reach the threshold on 7 and 9 October.

Trains (6 passages).

  • Trains are seen on the images on 5 October 16:00, 7 October 13:00 and 16:00, and 12 October 16:00.
  • The passages on 7 October 19:45 and 11 October 14:15 are deduced from the same signature.
  • On these trains, most targets and two of the three references are hidden. The targets left visible (BG.2, BG.3, F4.5, sometimes some P) jump by 5 to 7 mm for one measurement, then come back. Confirmed.
  • On 12 October 06:45, a smaller passage makes CV.4 jump by about 1.3 mm for one measurement. Likely (no image).

Corrected curves in x of P7, P1 and CV.4, in mm

P7 (blue): the daily thermal hump. CV.4 (green): the daytime shaking. P1 (orange): the drop at the 7 October 16:00 train.

Train passing on 7 October at 13:00

The blurred train crosses the lower part of the wall, in front of the targets.

Missing measurements (4 nights). On 9, 10, 13 and 14 October, one measurement is missing for all sensors, around 01:30-02:00. On the 9th and 14th FLORIDA misses the same instant. The cause is outside the station and not established.

What to do

Nothing on site for these events. The six targets that are badly seen over the whole month are already labelled for action (BG.1, F4.1, CV.7, F4.4, F4.3, CV.6).

Not established

  • The daily cycle and the shaking are tied to heat by the images (sun) and their shape only. A temperature sensor on the station would confirm it.
  • On 7 and 9 October, a single value per day reaches the threshold; I left them as noise. If you want such isolated values labelled, say so.

Labels

Start (UTC) End (UTC) Family Targets In a few words
04/10 08:45 04/10 16:30 Visibility CV.4, F4.2, BG.2, BG.3, P daytime shaking, sunny
04/10 09:00 05/10 06:30 Thermal displacement P7, P6, P5, P1 camera daily cycle
05/10 06:30 06/10 07:15 Thermal displacement P7, P6, P5, P1 camera daily cycle
05/10 07:00 05/10 15:15 Visibility CV.2, CV.4, F4.2, P1-P3 daytime shaking, sunny
05/10 15:45 05/10 16:15 Obstruction BG.2, BG.3, F4.5, P4 train, seen
06/10 07:00 06/10 13:15 Visibility CV.4, F4.2, CV.2 daytime shaking, sunny
06/10 07:15 07/10 03:00 Thermal displacement P7, P6, P5, P1 camera daily cycle
07/10 07:30 08/10 04:15 Thermal displacement P5, P7, P6, P1 weak cycle, overcast
07/10 12:45 07/10 13:15 Obstruction BG.3, F4.5, F4.6 train, seen
07/10 15:45 07/10 16:15 Obstruction BG.2, BG.3, F4.5, P1 train, seen
07/10 19:30 07/10 20:00 Obstruction BG.2, BG.3, F4.5, P2-P7 passage at night, deduced
08/10 06:45 08/10 16:30 Visibility CV.2, CV.4, F4.2, P6, P7 daytime shaking, sunny
08/10 08:15 09/10 04:30 Thermal displacement P7, P6, P5, P1 weak cycle, night overshoot
09/10 01:45 09/10 02:15 No data all one measurement missing, with FLORIDA
09/10 07:00 10/10 06:00 Thermal displacement P5, P7, P6, P1, P2 cycle without evening return
10/10 01:45 10/10 02:15 No data all one measurement missing
10/10 07:00 11/10 08:30 Thermal displacement P7, P6, P5 camera daily cycle
10/10 07:00 10/10 13:00 Visibility CV.4, CV.9, CV.2, F4.2, P2, P3 daytime shaking, sunny
11/10 08:30 12/10 02:45 Thermal displacement P7, P6, P5, P1 camera daily cycle, overcast
11/10 10:00 11/10 15:30 Visibility CV.4, F4.2, CV.2, P1, P2 weak shaking, overcast
11/10 14:00 11/10 14:30 Obstruction BG.2, BG.3, F4.5 passage, deduced
12/10 06:30 12/10 07:00 Obstruction CV.4 partial passage, deduced
12/10 07:30 12/10 14:45 Visibility CV.4, F4.2, CV.2, BG.3, P daytime shaking, sunny
12/10 08:00 14/10 04:00 Thermal displacement P7, P6, P5, P1, P4 two-day cycle, night without return
12/10 15:45 12/10 16:15 Obstruction BG.2, BG.3, F4.5 train, seen
13/10 01:45 13/10 02:15 No data all one measurement missing
13/10 07:45 13/10 14:00 Visibility CV.4, F4.2, F4.6, P1, P2 daytime shaking
14/10 01:15 14/10 01:45 No data all one measurement missing, with FLORIDA