ORAFOL halved lens inspection time using Zebra AI machine vision

Some of ORAFOL Fresnel Optics' Fresnel plastic lenses are as small as 4mm across. Operators used to find defects by holding every lens up to the light under specific lighting conditions. An AI machine vision system from Zebra Technologies has since cut inspection times in half. Zebra says the bespoke inspection systems built on its software have now handled more than 1.5 million optical lenses.

The halving figure comes from a vendor case study, which Zebra announced on October 26, 2026 in a paid Business Wire release, so treat it as a reported result. The more useful detail is where the custom work went: the complex microstructures required the development of a bespoke optical turret. For a quality or operations lead scoping vision inspection on small, finely structured parts, that split is a useful planning reference.

Checking lenses against the light

ORAFOL Fresnel Optics is a German-based division of the ORAFOL Group. It supplies more than 600 different components to customers around the world in the machine vision, sensor, lighting, display and solar markets. Every item has to meet exact quality standards before it leaves the facility, and the Fresnel lenses were the hard case. Some are as small as 4mm across with complex microstructures, which makes inspection difficult.

The product range adds to the problem. ORAFOL's own product pages show Fresnel lenses from a few millimetres to several hundred millimetres in diameter. Some also have a diffuser microstructure embossed on the flat back, so a single part can be patterned on both faces. That variety could help explain why an off-the-shelf inspection setup was hard to find.

The old method was manual. Operators held each lens to the light under set lighting conditions and looked for defects. Zebra's account says the work was repetitive and tiring and took as many as 15 operators to support production lines. ORAFOL looked for a supplier that could automate the job and couldn't find one able to handle the complex Fresnel lenses, so the system had to be built to order.

The automation joined a quality system that was already mature. ORAFOL's Optic Solutions division says on its website that it has held DIN ISO 9001 certification since 1996, that it uses a mix of standard measurement systems and customer-specific setups, and that every product is inspected either manually or automatically.

The same page says quality assurance covers the whole manufacturing process and is continuously monitored and optimised. It lists structure inspection, geometry inspection and automatic optical inspection as separate checks, and illustrates the last with high-volume inspection of small polymer Fresnel lenses. That suggests the Zebra system took over one defined inspection job inside an existing regime rather than replacing the regime itself.

What the system changed on the line

The first system runs on Zebra Aurora Vision Studio with Aurora Deep Learning, which Zebra markets as a no-code studio. Besides cutting inspection time in half, Zebra says it enabled 24/7 workflows and end-to-end component traceability. It also freed frontline workers to focus on more strategic tasks.

Rico Traxdorf, who works in production and technology at ORAFOL Fresnel Optics, said consistency was the biggest gain. The team writes a specification into the system, and any scratch outside it is rejected. That turns accept-or-reject from one person's call at the end of a long shift into a written tolerance applied the same way to every part.

The reported results go beyond speed. Zebra lists round-the-clock operation as a separate result from the time cut, so the halving figure alone may understate the change in available inspection hours. For a supplier whose customers build machine vision and sensor products, a per-component inspection record could be as useful in a customer quality review as the faster cycle time.

Standard camera, custom turret

Zebra's account also gives the hardware side. According to Zebra, the solution saves ORAFOL both time and money. A standard camera was sufficient, because Aurora software works with any camera. The special engineering went into a bespoke optical turret that was developed for the complex microstructures. That suggests the custom work sat in the fixture rather than in the camera.

India Automation Hub reports that standard cameras alone weren't enough and that ORAFOL paired them with cameras built to reproduce the structure of a Fresnel lens. By its account, that let the system pick up defects that conventional inspection had missed, and it credits ORAFOL's gains to fitting the cameras to the part as well as to the software. The two accounts differ on the hardware.

They agree on the bigger point, though. Whether it's described as a turret or as purpose-built cameras, the part-specific optics is where the engineering effort went. A buyer's planning question is therefore less about which camera to buy and more about who designs the way the part is presented and imaged.

A second line, built with an integrator

Rising demand led ORAFOL Fresnel Optics to build a second system, more advanced than the first, for high-volume inspection of optical windows. It worked with ID Engineering, a Zebra Industrial Automation System Integrator, and used Aurora Vision Studio with deep learning again. Quality engineer Alexandra Frenzel said a problem no supplier could originally solve has become a core part of production for the company's high-volume products. She added that the system plays a crucial role in delivering the consistently high, flawless quality ORAFOL's customers expect.

The second build changed the product and kept the software. That suggests the first project gave ORAFOL a reusable base it could apply to a different part. Jason Harvey, Zebra's VP of AI and automation for EMEA, described the deployment as an example of deep-learning vision and partner expertise working together on a difficult problem.

At ORAFOL, the practical version of Harvey's line is fairly specific: a written tolerance makes the reject decision at the inspection station, and each component carries a traceable record of it.

There is a counterpoint, and it comes from the same facts. No-code describes how the software is configured, not how much engineering the project takes. ORAFOL still needed a custom turret for the first system and a specialist integrator for the second. The case also reports the time saving as a ratio to manual inspection, without absolute cycle times, so plants can't benchmark throughput per lens from it directly.

What the optical-window line will show

The figure worth watching is the one ORAFOL hasn't reported yet: results from the optical-window system. If a second, higher-volume product family gets gains close to the first on the same software, the pattern would hold up: standard software and cameras, with the custom work concentrated in how the part is held and lit. If the windows needed a very different hardware build, the lesson for other precision-parts makers would be narrower. Each new part shape would come with its own fixture project.