Up to 2,400 electric motor housings a day now pass through fully automated AI surface inspection at an automotive supplier plant in the Czech Republic, and the supplier says the system has cut annual complaints by two thirds and driven unrecognised surface defects to zero.

The installation, designed by Vivaldi Digital Solutions GmbH, runs up to 70 USB3 uEye+ XC autofocus cameras from IDS Imaging Development Systems across several inspection cells. An Intel-based edge server computes the AI analyses. The components are aluminium housings for electric motors bound for German OEMs.

The tolerances are tight. "There are strict requirements at the plant in the Czech Republic. According to the defect catalogue, defects from 0.2 millimetres must be reliably detected. The test time per component must not exceed 36 seconds," said Uwe Siegwart, Managing Director at Vivaldi. Every part is tested, 100 per cent automatically.

How the System Works

"The IDS cameras integrated into the test cells capture high-resolution images of the surface of aluminium components from German OEMs, in this case an electric motor," Siegwart said. Images are pre-processed for exposure, white balance and colour correction before the AI classifies defects — scratches, pores, dirt, cold runs, chatter marks.

The architecture is deliberately modular:

• Up to eight cameras connect per USB hub, capturing several inspection areas simultaneously.
• A maximum of two hubs feed each of Vivaldi's self-developed "cam controllers", which bundle and process image signals.
• Multiple cam controllers network into the edge server, which runs the AI analyses and talks directly to the system controller that triggers and synchronises the cameras.

Aluminium is the hard part. Shiny surfaces reflect light, and those reflections can obscure or distort defects. The AI has to separate normal material structure from genuine flaws across material and environmental variations — the difference between a scratch and a specular highlight.

Why the uEye XC?

Camera selection came down to geometry. "The decisive factor was that the autofocus camera can capture several surfaces at different distances from the camera," said Eike Humpert of Vivaldi's Business Development. That flexibility matters when different component variants share a line. A macro attachment lens shortens the minimum object distance when a variant demands closer inspection.

Sensor technology does the rest. "The BSI pixel technology enables the sensor to deliver consistently precise and reliable results, even in changing light conditions," said Jürgen Hejna, Product Owner 2D cameras at IDS. BSI (back-side illumination) puts the wiring behind the photosensitive layer, improving light capture and reducing pixel noise. Hejna also credits the 24x digital zoom, auto white balance and precise colour correction for the system's reliability.

What Does This Change for Traceability?

Traceability is the real prize here, not just defect detection. "Image processing is a central component for traceability, which is the actual technological centrepiece of the solution, and IDS is the ideal partner for this point," Siegwart said. Digitised quality data, captured in real time and linked to other production parameters, reduces rejects and rework — errors minimised, costs down, processes optimised, in his words.

For a German supply base under sustained cost and quality pressure, that linkage is the argument. Fluctuating detection rates, weak batch-level traceability and shifting plant conditions have historically meant downtime, complaints and rework. A camera network that resolves defects down to 0.2 mm, inside a 36-second takt (cycle time per component), with serial-level data attached, addresses all three at once.

What to Watch Next

Vivaldi expects the economics to keep improving. Cheaper, higher-resolution cameras and optimised tooling will push real-time analysis into the hands of end users rather than vision specialists, Humpert said, with demand concentrated on improved resolution, more compact designs and price-optimised cameras.

The credible next milestones: whether Vivaldi ports the SensorBox beyond this single Czech site, whether the two-thirds complaint reduction holds across other component geometries, and whether sub-0.2 mm defect catalogues arrive as camera resolution climbs. Supplier claims of zero missed defects should also be tracked against warranty and return data over a full model year — the plant-level numbers look strong, but one site is one site.