I am preparing an FR5 faceplate inspection station and our example images are being labelled before we have reliable references. Bright streaks move with small presentation changes, yet people call them scratches. What should I establish about the physical pieces and capture conditions before commissioning and threshold work?
A small faceplate movement changes which streak we call a scratch
IngridCasey · 2026年7月28日 08:57 UTC
主帖中文翻译暂不可用，以下显示英文原文；回复保留原文。 查看英文原文
12 条回复
回复 IngridCasey
Can you still identify and retrieve the physical faceplates behind those images? Ask the acceptance owner to examine those pieces under an agreed viewing method. I would not use the image labels to decide their physical condition.
21分回复 MinaChan1120
Yes, the original faceplates are retained with image identities. I can ask quality to examine them; no independent decisions have been recorded for these examples yet.
7分回复 IngridCasey
Don't swing from all scratches to all harmless reflections. A moving bright patch can hide a real mark as well as impersonate one.
11分回复 IngridCasey
Keep a decision per physical faceplate, including any undecided region, and link each capture to that identity. A folder called good is a poor database once someone starts copying crops into presentations.
5分The holder configuration belongs with the capture as well. If maintenance later replaces or adjusts it, the next person needs to know what presentation the image comparison actually established, not just which exposure produced a clean picture.
14分Jamie, record removal and reseating in the comparison too. Repeated frames of an untouched plate can look consistent while never testing the presentation difference Ingrid described.
9分Quality has examined the retained pieces and marked one disputed streak region as acceptable surface. Another piece has a confirmed scratch elsewhere. We have physical decisions now, but no repeatability result for the imaging arrangement.
14分回复 IngridCasey
Keep that confirmed scratch in the next comparison. It stops the nicest-looking view winning by hiding the defect you actually need.
7分回复 IngridCasey
And retain the misleading old captures as evidence of the presentation problem, separate from the tuning set. Their history remains useful even after the labels attached to them are corrected.
21分I've linked quality's decisions to the pieces and kept the old captures outside the tuning references. The proposed comparison records holder, lighting and reseating; I will not describe that plan as a validated inspection method.
5分回复 IngridCasey
Who will store those reference faceplates between comparisons? A labelled image survives a meeting more easily than a small part left on somebody's bench, and future support needs both.
16分Also agree when a reference needs re-examination. I don't know your handling conditions, but the original physical judgement should not be assumed unchanged after damage or a replacement sample.
13分