I'm preparing the FR10 inspection lesson. Bright lines move between pictures of the same brushed plate, yet the folder calls all of them scratches. Nobody has linked those lines to a physical inspection.
the FR10 trim-plate examples have acquired scratch labels without inspection
CalebBrown0928 · 2025年10月13日 16:16 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
21 条回复
Keep the original pictures, but have quality identify and judge the actual plate before using those defect labels; a moving highlight is useful evidence about the image setup, not a ready-made reject example.
13分Keep undecided separate from accepted and rejected. Otherwise somebody will force a label just to finish the folder sorting.
7分We had a reference plate that picked up a new scratch during training. Everyone kept calling it the good one because that was written on its box. Whoever keeps your samples needs to check their condition, not only their number.
20分Beth Brooks, how often did you check yours after that? Before every lesson, or when somebody noticed a change?
22分Caleb, do those different views use the same holder position?
9分Jonas, the temporary holder allows a slight tilt. Quality has the identified plate now, and Ella Barnes, its label stays undecided until they inspect it.
-1分The holder review should establish repeatability across the intended part variation. Even a stable image of this single plate would not establish a suitable production presentation.
21分Do the files record the camera settings, or just the image names?
14分Beth's example is why a physical reference set needs an owner and handling rules, not merely a nice spreadsheet; otherwise the lesson can quietly change while the slide deck stays identical.
19分回复 ChrisBailey
And keep the full images linked to any close-ups. We nearly argued about the wrong edge because every crop filled the screen. A recognisable hole or corner makes the defect location much easier to explain.
4分Don't annotate over the defect. Separate marked copy, original retained.
13分回复 BethBell0672
Beth Bell, would you show the whole plate beside the close-up in training, so the loader can find the same place?
15分Ella Archer, before use after that incident, with inspection owning the replacement decision. Nora, yes, we also stopped passing the reference around loose. People meant well and handled it like an ordinary sample.
12分回复 ChrisBailey
Quality accepted the plate and retained a separate faint reject. Chris, settings were not saved for all old views. Henry, tooling is checking the holder range. New captures will record the setup.
19分Don't discard the failed old views just because their settings are incomplete. Mark the gaps and keep them out of tuning; they still explain why you restarted the comparison.
23分Elliot, yes. Whole plate plus close-up. Easier than guessing which shiny corner the picture means.
13分Thanks, Beth Brooks. I was thinking only about keeping the sample number attached, not checking whether the sample itself had changed.
22分Caleb, did the holder check explain the tilt, or is it still waiting behind the image work?
22分Tooling found the rim can catch one temporary support. Revised support is proposed, not fitted. The known reject stays in the repeat-image set; no threshold decision yet.
20分Check that revised support against the drawing variation, not only the two retained plates.
17分讨论已关闭
此讨论已连续六个月没有活动，现已关闭新回复。 最后活动： .