By Sophie Lindqvist

Automotive paint shops now run surface inspection systems that detect defects as small as 40 to 50 microns, roughly half the width of a human hair. Wolfgang Berggold, business line manager at Isra Vision, calls that level of resolution very high for series production.

The benchmark matters because BMW's Regensburg plant in southeast Germany was the first car factory in the world to deploy automated optical inspection (AOI) for series production in March 2023.

How does AOI work in the paint shop?

AOI targets the small specks and bumps that Rudolf Hummelsberger, vice president of painted body at BMW, says appear on the topcoat (the final paint layer). These marks usually come from fine dust, and robots sand them out.

Standard machine vision fails on polished, reflective body panels. The workaround is deflectometry, a method in which monitor screens project striped patterns onto a body and cameras read the reflections. Software then maps the reflections to locate uneven areas.

The system stores each defect's position, size, and shape in a digital catalogue, and the algorithm compares new data against that catalogue in real time. Hummelsberger says the approach ensures that no defect, however small, escapes detection. The continuous digital record also feeds statistics that flag process anomalies before they become scrap.

What can AOI still not do?

The current system does not cover panel edges or the final millimetres next to doors and other joints. Skilled humans still inspect those zones manually, and a laser projector marks the areas the system has not yet reviewed, so operators focus only on the uninspected regions.

Software at BMW's Munich plant compares live data from dust particle sensors inside the paint booths with AOI output. During dry, hot spells, the algorithm can call for filter changes in the air handling system before defects spike.

Where is the technology heading?

BMW frames its roadmap cautiously. Hummelsberger told AMS: "We consider the potential for using AI in our production systems to be very high." Pilot installations are running, but he declined to detail them.

Future use cases include paint robots that reprogram themselves and closed-loop quality control that adjusts process parameters without operator input. Hummelsberger also pointed to new possibilities for finishing area digitalisation through supplier partnerships, though he offered no timeline.

What challenges does customisation pose?

Individual paint finishes, including multi-tone topcoats, and new substrates from paint suppliers can confuse an AI system trained on a narrower material set. "The use of different substrates or individualisation of the car body finish, such as multi-colour, poses challenges for the technology and must be technically evaluated on a case-by-case basis during implementation," Hummelsberger said.

Kiran Mohan, vision system engineer at Shelton Vision, which supplies inspection machines for textiles, car headliners, airbags, non-woven surfaces and composites, said the same problem applies across materials. He expects AI to expand into more inspection areas as camera, lighting and processing power improve.

What to watch next

• Whether BMW's pilot installations move into series production, and on what timeline.
• How AOI suppliers handle new substrates and multi-tone finishes as carmaker order books shift toward greater individualisation.
• The first paint shop to deploy self-reprogramming robots, a capability Hummelsberger flagged as a near-term target.