Two hundredths that stop a machine

First iteration: an inspection gauge in one day

The first response was as direct as possible: a mechanical go/no-go gauge integrated into one of the output chutes.

The idea is simple. The card slides down the chute and passes beneath a micrometer head set to the right height. If the card fits, it continues. If it is thicker, it is retained and the operator removes it.

It was designed, made and tested in a single day.

And the dimensional check worked. Good cards passed; thicker cards were retained.

But something emerged that the drawing had not warned about. As it slid, the card vibrated slightly and touched the head on its way through. Nothing showed on matt cards. On gloss-finished cards, a small mark could be left.

An inspection system that detects the defect correctly but damages a good part does not work for every product. And here, everything had to be sorted.

What a quick prototype teaches you

For me, this is the most useful lesson in the whole case.

You might think the first iteration was a failure. It was the opposite, for two reasons.

First: in one day, it validated the approach. It showed that separating by thickness solved the jamming problem, and that contact was the only remaining issue. Weeks spent designing the final solution on paper would have led to the same problem, only at the end, with everything already assembled.

Second, and less often mentioned: the first version kept working. It worked perfectly with matt cards. So, while the second version was being designed, the first continued sorting the matt cards. Production did not wait for the final solution, and the final solution did not have to be rushed.

A quick prototype does not have to be perfect. It has to teach you what you did not know and, if possible, buy you time to get the next step right.

Second iteration: no contact as the card passes

The second iteration took a couple of days. And it was not just mechanical work: design, changes to the PLC program, changes to the control screen, and changes to the control circuitry and wiring. It was integrated into the machine as another function, rather than as an add-on device.

The change in approach is easy to explain: the card no longer rubs against or strikes anything as it slides. It arrives, stops, is checked, and only then is a decision made.

To actuate the check, we used something the machine already had: the chip encoding station and its cylinder. Instead of adding a new actuator, the existing one was adapted to the mechanism.

The cycle works like this:

1. Position. The card travels down the chute and stops in the inspection area.
2. Check. The cylinder actuates the mechanism and the system checks the thickness.
3. Release or retain. A conforming card continues. A non-conforming card is retained, with an on-screen alert, until the operator removes it.

The result: the marks disappeared, including on gloss-finished cards.

Measuring the plastic without looking at it

There is a detail in the measurement method worth explaining.

Beneath the card, seated in a recess in the chute, is a steel insert that provides a reference. Measurement is performed by an inductive sensor, and an inductive sensor only detects metal. To the sensor, the plastic card does not exist.

So the sensor does not measure the card: it measures the distance to the steel beneath it. The card in between determines that distance. If it is thicker, the steel is a little further away, and the sensor detects it.

This has three advantages:

• It does not depend on the finish. Matt, glossy, printed or plain: the sensor does not care, because it is not looking at the card.
• The reference is always the same. A fixed steel insert, seated in its recess, always in the same place.
• Resolution to spare. An inductive sensor of this type has repeatability on the order of five microns. The difference to detect is twenty. There is margin to distinguish them reliably.

That last comparison should always be made before choosing a measurement method: the sensor's repeatability must be clearly smaller than the difference you want to detect. If they are close, you will have a machine that gets it right almost every time, which is another way of saying a machine you cannot rely on.

One chute out of four, and reversible

Two design decisions are not visible in the mechanism, but matter just as much.

Only one of the four chutes was modified. Inspection coexists with normal operation on the other three. There was no need to dedicate the whole machine to the issue: production continued while the batch was sorted.

The modification is reversible. The machine can be configured for normal production or thickness inspection.

The batch was sorted. And the machine gained something it did not have before: an inspection mode that can be activated when needed. If another material variation arrives tomorrow, nothing needs to be invented. Change the mode, and it is ready.

These decisions also make it easier to agree to modify a machine in production. A modification that requires everything to stop, or cannot be undone, is much harder to approve than one that affects a single chute and can be reversed.

The summary

If you have a mixed batch with a difference you cannot see, and need to sort it in the machine itself:

1. Validate the approach first, and quickly. A simple prototype in one day tells you whether separating by that dimension solves the problem.
2. Use the first version while you build the second, for everything it already handles well.
3. Check what inspection does to the good part. Detecting the defect is not enough if the inspection leaves a mark.
4. Avoid rubbing or impact as the part passes. Stop it, check it and decide before releasing it.
5. Use what the machine already has. An existing actuator is one you do not need to add.
6. Measure against a fixed reference, using a sensor whose repeatability is clearly smaller than the difference you want to detect.
7. Modify as little as possible, and make it reversible. One chute, not the machine. A mode, not a permanent change.

At 7Axis Industrial, we adapt existing machinery to integrate quality inspection: mechanics, PLC, control screen and wiring, without replacing the machine or stopping it more than necessary.

View the full case → Machine modification and improvement → Quality, inspection and poka-yoke →