Skip to content
PRANINNOVATIONSProduction Specialists

Manufacturing · Vision · Tier-two aerospace supplier

Surface defect inspection, decided at the camera in 12 ms

Final inspection was manual and sampled, so surface defects were being found by the customer rather than on the floor. We surveyed the cell, captured production across three shifts, and commissioned an edge vision station that inspects every part and drives the existing reject gate.

12 ms

decision latency at the camera

99.4%

detection rate on hold-out set

100%

of parts inspected, up from sampled

The challenge

Parts left the cell faster than a cloud round trip would allow, and the plant handles controlled technology, so streaming imagery off site was not permitted. Manual inspection covered a sample, which meant the true escape rate was unknown rather than low.

What we did

  1. Two-day site survey covering optics, lighting, mounting and the existing reject mechanism
  2. Production footage captured across three shifts, including changeovers and night-shift lighting
  3. Annotation schema agreed directly with the quality team, so classes matched existing paperwork
  4. Model trained and optimised to hold line rate on the industrial edge unit being installed
  5. Integrated with the cell's PLC over OPC UA to drive the reject gate already in place
  6. Four-week supervised run in shadow mode before the system was permitted to act

Outcome

Every part is now inspected instead of a sample, the decision is made on the device at the cell, and no imagery leaves the plant network. The quality team owns the dataset and the annotation schema.

Start with a call, then a costed plan

Thirty minutes on the problem, the site and the constraints. If it looks like a fit, the next step is a four-week AI Pilot at a fixed price, with acceptance criteria signed before any code is written.