Quality

Control charts of the reject rate and the defect mix behind it, over the last 24 hours. When a chart goes out of control the usual cause is the line, not the model — a new tool, a different batch of material, a lamp that has aged. Check the line first.

Parts inspected

800

2 stations reporting

First-pass yield

76.9%

Held

9.8%

Rejected

13.4%

Overrides

0.0%

Above 20% for a shift is a signal to review the labels and the threshold

Reject rate, fleet-wide

A p-chart: the control limits move with the subgroup size, so a quiet night shift and a busy day shift are judged on the same terms

In control

Defect mix

What the fleet is finding, most common first

Why parts did not pass

Every non-passing decision carries a coded reason. These are the ones that fired

Code	Reason	Kind	Parts
201	Known defect detected with high confidence.	defect	107
204	Classifier and anomaly model disagree.	defect	23
202	Result is not confident enough to decide automatically.	defect	15
205	Mark found but smaller than the size limit for its class.	defect	13
103	Reflection covers part of the inspected area.	evidence	12
301	Part does not look like anything the model has seen.	novelty	11
109	Camera did not deliver a frame in time.	evidence	4

By station

One chart per station: a fleet-wide excursion and one station drifting look identical in aggregate

ST-01in control

ST-02in control