How we work
A camera, lighting and a fixture at the inspection point. An industrial computer runs the models locally, with no cloud dependency. A pass or fail signal goes to your PLC for rejection or alarm, and every part is logged with its image.
Four things at the inspection point
Step 1
Camera, lighting and fixture
The part is presented the same way every time. Fixed lighting and a fixture at the inspection point matter more than the camera: they are what make one part comparable to the next.
Step 2
An industrial computer at the line
The models run on a computer in your plant. Nothing is sent to a cloud service, so the station keeps working when the connection does not, and your part images stay with you.
Step 3
A pass or fail signal to your PLC
The station gives your PLC a signal it can act on, for rejection or for an alarm. How you use it is your decision, and it is wired to your existing control scheme.
Step 4
Every part logged with its image
Each inspected part is recorded with the image it was judged on. When a call is questioned, there is something to go back and look at.
A paid four week pilot
We would rather run a pilot on your parts than describe what we could do in principle. It runs for four weeks, at your line, and it ends in a measurement rather than a presentation.
Week 1
Scoping
- We agree which parts and which defects are in scope.
- You give us defect samples, including the ones you currently argue about.
- We write down the success criteria together, before any model exists.
Week 2
Optics and image capture
- Lighting trials at your line, on your parts.
- We fix the camera position and the fixture.
- Image capture begins once the view is stable.
Week 3
Model build
- We build the model on the images captured at your line.
- Where a dimension matters, measurement is added alongside the defect check.
Week 4
Shadow run
- The station runs beside your inspectors, judging the same parts they do.
- Disagreements are collected and reviewed rather than explained away.
What you receive
At the end you receive a report.
- Recall, measured on your own parts
- False reject rate, measured on your own parts
- Cycle time, measured on your own parts
A clear go or no-go recommendation, including when the answer is no.
The pilot fee is credited against a production system.
What plant managers ask us
What do you need from us to start?
Parts, including defective ones, and access to the line for the lighting trials. If you have images already, we will look at them, but we expect to capture our own at your line.
Does anything leave our plant?
The station runs the models on a computer in your plant and does not depend on a cloud service. During a pilot we need to work with sample images, and what we may keep is agreed in writing before we start.
Will it talk to our PLC?
Yes. The station gives a pass or fail signal to your PLC, wired to your existing control scheme, and you decide whether that drives a rejection gate or an alarm.
What happens when it flags a good part?
That is a false reject, and it has a cost, so the pilot measures it rather than ignoring it. You see the rate alongside recall and decide whether the trade is worth making.
What if the answer is that it will not work?
Then the report says so. A pilot that ends in a no-go has still told you something worth knowing, which is why we would rather run one than describe what we could do in principle.
Can you inspect a part that is not on your list?
Ask us. We work on drill bits, nuts and bolts today. Whether a different part is a fit depends on whether the defect is visible under fixed lighting at the resolution we can hold, and we will tell you plainly if it is not.