WorkRepresentative engagement02 / 12

Automated visual inspection of extruded aluminium profiles

A line-scan camera station with controlled lighting and an edge inference unit now grades surface defects at full line speed, replacing sampled manual inspection.

Sector
Manufacturing and industrial operations
Location
Germany
Duration
8 months
Services
Data science and AIElectronics and embedded systemsEdge intelligence and control
Customer-reported surface defects per monthClaims
-100102030M1M2M3M4M5M6M7M8M9M10M1 Claims: 19 claimsM2 Claims: 17 claimsM3 Claims: 21 claimsM4 Claims: 16 claimsM5 Claims: 14 claimsM6 Claims: 9 claimsM7 Claims: 6 claimsM8 Claims: 4 claimsM9 Claims: 3 claimsM10 Claims: 3 claims
claimsM10 · 3 claims

The challenge

Surface defects such as die lines, pick-up and blisters were found by manual inspection of samples at the saw. Defects reaching customers triggered claims and full-batch sorting.

Profiles varied in geometry, finish and reflectivity, and the line could not slow down for inspection.

What we did

We specified the capture station first: line-scan optics, dark-field and bright-field illumination, encoder-triggered acquisition and a mechanical arrangement that kept the profile in focus.

A defect detection and classification model was trained on images labelled jointly with quality staff, with an agreed definition of each defect class and its severity.

Inference runs on an industrial edge computer beside the line and reports to the quality system; uncertain regions are routed to a human review screen rather than auto-graded.

The outcome

Every metre of profile is inspected. Defect escapes fell sharply within two months, and the manual sample inspection was redeployed to root-cause work on the press.

A representative engagement. The figures are indicative of typical outcomes for this kind of work and are not drawn from a named client.

Work · 03Forecasting heat and electricity demand for a district utility

Contact

Tell us about the problem you need to solve

You may have a defined project, an underperforming process or a technical question that needs investigation. We can help assess the challenge, identify the information required and define an appropriate next step.