our recipe check trusts whichever insert I say I fitted

AmyBaker0467 · 5 Apr 2025, 02:05 UTC

Closed
AM
AmyBaker0467
I can select the right recipe with the wrong cover nest installed on our proposed FR5 station, and the insert also fits backwards. Better labels feel like only half an answer.

15 replies

NI
NinaChen1197
Replying to AmyBaker0467

Labels help someone choose, but they don't tell the control system what is there. Can your tool designer make the wrong orientation physically impossible before you start adding clever checks?

8 points
AM
AmyBaker0467
Replying to NinaChen1197

The designer can add an asymmetric locating feature. That would stop reversal, but the other product-family insert would still fit the base correctly.

7 points
NI
NinaChen1197
Replying to AmyBaker0467

Then that's two different problems. Key the orientation, and get controls to review how installed insert identity is checked against the selected recipe. Don't ask one feature to do both jobs if it can't.

9 points
AM
AmyBaker0467
Replying to NinaChen1197

We had been calling both 'wrong nest', which probably explains the argument. I'm splitting them in the review, with seating as another condition to check.

9 points
NI
NinaChen1197
Replying to AmyBaker0467

Yes, a recognised insert can still be sitting on a chip. We learned that with a perfectly correct label on a fixture that wasn't down on its supports.

3 points
MI
MinaBennett0772
Replying to AmyBaker0467

What does the current 'nest present' input actually detect? If it's just metal somewhere near a sensor, don't rename it 'seated' and hope nobody notices the difference.

-5 points
NI
NinaChen1197
Replying to MinaBennett0772

Fair question. I assumed there wasn't a sensor yet, but you haven't said either way.

9 points
AN
AnikaAdams0160
Replying to AmyBaker0467

Also check how stores names the two inserts, because ours shared a stock description long after the drawings stopped being interchangeable, which made the wrong pick rather easy.

-1 points
NI
NinaChen1197
Replying to AnikaAdams0160

Keep the human-readable identity as well as whatever controls proposes. Maintenance still needs to identify the thing in their hand when it isn't installed.

4 points
MA
MayaChen1171
Replying to NinaChen1197

What should happen if the identity can't be read? Put that in the requirements now. A failed read mustn't quietly mean 'use the last recipe'.

22 points
AM
AmyBaker0467
Replying to MayaChen1171

There is no sensor yet, just a proposed present signal. Controls is reviewing identity, seating and unreadable states separately; the last recipe won't count as confirmation of the installed insert.

20 points
MA
MayaChen1171
Replying to AmyBaker0467

And who can authorise the changeover after a mismatch? The operator needs an approved way forward, not a box they can tick until the warning goes away.

9 points
AM
AmyBaker0467
Replying to MayaChen1171

Our process owner will define that with the operators. We haven't chosen the identification hardware, and I won't pretend the revised wording is a tested system.

23 points
AM
AmyBaker0467
Replying to AmyBaker0467

Tooling has proposed the asymmetric feature, stores has distinct insert descriptions, and controls has the failure cases. Next is a combined design review, still before automated use.

23 points
NI
NinaChen1197
Replying to AmyBaker0467

Include a wrong-family insert fitted the right way in the checks. That's the one your new key deliberately won't stop, so it's a good test of whether the identity check really does its job.

15 points

Discussion closed

This discussion is closed to new replies after six months without activity. Last activity: .