Applications

Check that the hardware is there before the part moves on

Press-in nuts, inserts and fasteners go missing, or the wrong ones get installed. A camera can see that the hardware looks different from a correct part. We want to be plain: VZSOR has not been validated for this yet, and a pilot would be the test.

What it flags
  • A press-in nut or insert that is absent
  • Hardware that looks different from the one in the good parts
  • A fastener that is missing from its hole
  • Hardware installed crooked or sitting proud of the surface
  • A part where the assembly step looks skipped

Where hardware goes wrong

Sheet metal and machined shops install hardware by hand or by press. A tired operator forgets one. A bin gets mixed. A press cycle fails without notice. The part moves on, and the miss surfaces at the customer. The hardware is small compared to the part, so a quick glance can miss it.

These misses are visual. The hardware is there or it is not, and it looks right or it does not. That makes them a candidate for a camera check. That makes them a candidate for a camera check, as long as the hardware faces the camera and can be seen in a fixed view. Parts where the hardware is hidden are a poor fit.

How the station would handle it

VZSOR is trained on photos of correctly assembled parts. A part with a missing nut or a different insert looks unlike the good set, so the station would flag it and show a heatmap at the spot. The operator then confirms whether hardware is truly missing.

A person looks at the flag and decides. Each result is saved with the photo, the part number, the lot or serial, and the operator, so you can show what was checked. This gives you a record of what was checked on each assembly.

Wrong versus missing

Missing hardware is a large visual change. Wrong hardware can be subtle. Two nuts of similar shape, differing in thread size or length, may look nearly identical from above. Spotting a wrong part takes more detail in the photo than spotting an empty hole, and the pilot would show whether the picture holds enough of it for your hardware.

VZSOR does not read thread size and does not measure anything. If the wrong part differs only in a way a camera cannot see, the station will not catch it. Keep your other controls for those cases. Similar parts that differ only in thread or length may pass.

Not yet validated

The validated work is one aluminum L-bracket and three defect types: dents, scratches and nicks. Assembly checks have not been tested. We are not claiming results here. Anyone who tells you a camera station handles every kind of assembly check before testing your parts is guessing, and so would we be.

A pilot would take one assembly, train on good ones, then run parts with hardware deliberately left out or swapped. The outcome would tell you whether this fits your parts. The outcome would tell you whether this fits your parts, with numbers from your own test set and not from ours. If it does not fit, you will know within the pilot window.

How a shop sets it up

  1. Choose one assembly and one view where the hardware is visible.
  2. Fix the part so the hardware sits in the same place in the frame each time.
  3. Photograph about 60 correct assemblies and run the wizard, about 30 minutes.
  4. Set aside parts with hardware missing or swapped for the pilot test set.

Limits, stated plainly

  • Not yet validated. No assembly or hardware tests have been run.
  • VZSOR does not identify hardware by size, grade or thread. Similar-looking wrong parts may pass.
  • Hardware on a face the camera cannot see is not checked.
  • Shiny hardware can make glare. Lighting needs to be tested on your part.

Questions

Does it work for press-in nuts today?
We do not know yet. VZSOR has been validated on dents, scratches and nicks on one bracket. Hardware presence is a use we expect to fit, and we would test it in a pilot rather than promise it.
Can it tell M4 from M5?
Not reliably, and we do not claim it. The station checks whether the part looks like your good parts. If two sizes look the same from the camera, it may pass both. Use a gauge or another control for that.
Do I need photos of parts with missing hardware?
Not to train it. Training uses good assemblies only. For a pilot test set, a few parts with hardware missing or swapped help us measure how well it flags them.
Does it count as a final check?
No. It flags suspect parts for a person. Your assembly sign-off stays with your people and your process. The station does not release parts to the next step by itself.

Try it on your parts.

$3,500 pilot on one part number, station on loan, 30–60 days. The fee is credited toward a station if you buy within 30 days.