Our FR5 faceplate inspection is still a plan, but the example folder already labels travelling bright streaks as scratches. Small presentation changes move them. I need defensible physical references and image conditions before approving threshold work, otherwise the next person inherits our guesses as training material.
Put the uncertain labels on hold and ask whoever owns appearance acceptance to judge the identified physical plates. A moving bright line is an observation, not yet a defect classification.
Can each image still be linked to its actual plate? If that link is missing, a later physical judgement cannot automatically repair the old annotation.
And keep the original images with the unverified labels marked as such. You will want to compare what the camera showed with the later physical judgement, not overwrite the history until only a clean reference folder remains.
The plates are retained and individually marked. I've linked the images and asked quality to review them. One extra image has no plate identity, so it stays outside the reference set rather than borrowing a likely match.
Make the hold travel with copied images too. The original folder can be beautifully managed while a slide deck keeps teaching people that every bright streak is a scratch.
Quality's first review is back. Two plates with strong streaks meet the appearance requirement; another has a real scratch at the rim. I've corrected the linked labels and flagged the copied teaching slides for the same update.
Keep that rim location in the capture comparison. Reducing the reflections is only useful if the actual required defect remains observable under the intended presentation conditions.
What exactly changes when you say presentation changes: the plate's seating, the holder position or the lighting? Record those separately before testing an explanation.
Removal and reseating change the streak position with the camera and lamp untouched. I haven't measured holder repeatability yet. We are starting there, keeping deliberate lighting changes for a separate comparison.
Include repeated placements of the same plate as well as different plates. Otherwise part variation and seating repeatability get mixed into one result.
Maintenance owns the holder record; quality retains the plates and physical decisions. I am keeping the capture settings and image-to-plate links with the comparison. No threshold selected.
Reserve examples for later evaluation that are not used to choose those settings or thresholds. These first plates can teach you where the problems are without becoming the entire evidence for the final inspection.
Holder inspection found a seating stop fitted to an older drawing revision. Maintenance has documented the as-found position and is reviewing restoration to the specified arrangement. I am not attributing every streak to that finding.
Good distinction. A configuration mismatch is now identified; repeat presentation and defect visibility still need checking after any approved correction.
Stop restored under the reviewed procedure. Repeated placements of the identified plates are more consistent, but the rim scratch still becomes weak in one view. That candidate view is not sufficient on its own.
It stays in the set. Thanks Robin, the temptation was to move straight to the clearer centre scratch. Vision and quality are reviewing the missing rim visibility before any threshold decision.