Wiring Vision into Andon and Interlocks: Signals, Reset, Bypass

A vision system that only records raises no objections, because it changes nothing about how the line already runs. The moment its verdict drives an andon lamp or a machine interlock, the conversation gets serious, because it stops the line. This step — from visible to intercepting — is the point of the whole exercise and also the most likely to generate opposition. Three things have to be settled before wiring begins: what form the signal takes, whether a dry contact, a network message or a hard wire; who may reset an alarm and under what conditions; and how the system exits safely if it fails. Projects that go to interlock without settling these are, in practice, bypassed quite quickly.

Prompt or interception, decided per item

Prompt and interception make very different demands. A prompt only has to deliver information to a dashboard or a station display, tolerates latency and the occasional lost message, and can travel over a network. An interception has to drive an actuator or prevent the next action, which means considering the consequence of a false actuation — one spurious line stop costs production directly, so the demands on reliability, response time and fail-safe behaviour are far higher. Within one station, different check items often suit different forms. Sequence errors such as a missing part or a skipped step usually justify interception, while a small positional deviation may be better as a prompt. Assigning each check item to prompt or intercept is the first step in interlock design, and it belongs to process and production jointly, not to the technical side alone.

Reset: authority, condition and record

How reset works decides whether the interlock means anything. If the button sits within the operator's reach and one press resumes the line, then on a tight takt it becomes muscle memory within days and the interlock exists in name only. A better arrangement grades reset authority by check item: minor deviations may be cleared by the operator, with the system counting how often; sequence or safety-related errors need a team leader or process engineer, and the reset is recorded with who cleared it and when. The reset condition needs defining too — is it "the workpiece has been removed" or "the step has been redone"? Those two require different detection. Making reset a recorded action rather than a click is what turns an interlock into something that actually holds.

Bypass prevention: make the honest path the easy one

Any mechanism that affects takt will be worked around: a wire pulled, the camera covered, an interlock parameter disabled. Technical lockdown is rarely airtight and it puts the shopfloor in opposition. The more effective idea is to make the intended path easier than the workaround. False alarms low enough that the system is not a burden. A reset that takes two presses and no waiting. A prompt that tells the operator what is wrong and what to do, rather than announcing NG. When handling an alarm takes clearly less time than bypassing it — and than the consequences of having bypassed it — the motive disappears on its own. This is why reducing false alarms matters more at the interlock stage than at the logging stage; the available levers are in cutting false alarms.

The record belongs in the interlock chain too

An interlock action with no record cannot be reviewed afterwards. The record should contain at least the trigger time, the check item, the verdict, the image, the time the signal was issued, the time of reset, and who reset it. With those, two questions become answerable: was this stop a false alarm or a real problem, and how long did recovery take. These are the same fields described in write-back field design, with interlock timestamps added. Joining the interlock record to production data is the only way to answer, months later, how many real problems the interlock actually stopped — which is the question that decides whether it stays.