A small mark.
A better next cycle.

Krunon knows what it takes to make a consistent part. For more than 30 years, its team in Bunnik has brought together mould design, material knowledge and production experience to make plastic components in small and large series.

One of its most demanding lines works with a specialised elastomer. The material has a narrow process window: preparation, temperature and pressure all need careful control. Jesper de Groot and his team understand how to keep those variables in balance.

Even in a stable process, a faint surface mark can appear on roughly one in every 25 to 30 components. It is easy to miss, but enough for the part to fall short of the visual standard.

Finding and removing those parts by hand puts a repetitive task on a small team. And when the rejected part leaves the line, useful information about the process often leaves with it.

The brief was practical: recognise this particular defect, fit the existing line and give the equipment a signal it could act on.

An iPhone at the inspection point gives Limen a view of each component. The vision model distinguishes an acceptable part from one with the characteristic surface mark. A rejected part is labelled NOK.

That result travels over Modbus to Krunon’s existing control setup. The handling equipment separates the rejected component from the accepted product flow. Inspection becomes part of the line’s response, right when the defect appears.

Finding the right camera position took thought. Lighting, angle and the shape of the part all mattered. Once the position was settled, mounting the iPhone took two minutes.

Krunon’s handling equipment already used Modbus. Its own technician mapped Limen’s detection triggers to machine actions in the app. For this setup, no on-site integrator or rewritten PLC code was needed.

The first model was trained with Limen’s team, using images already captured on the line. The team labelled examples OK or NOK to teach the model the difference. By the end of the session, Jesper’s plan was simple: train the next model themselves.

As production continues, new examples give the team more material to review, label and use to refine the model. Their knowledge of the part stays at the centre of the process.

A surface mark is the visible result of a much larger process. Recording it makes the detection useful beyond sorting a single part.

Limen keeps the OK and NOK results, the images behind rejected parts and the rejection rate over time. Krunon can follow quality during production and compare what happened across runs.

An increase in rejections gives the team a reason to investigate the material, machine or process. With thresholds they define, they can sort a part, stop the machine or review the images before deciding what happens next. That decision can travel with them, instead of keeping them at the machine.

The image does not explain every cause on its own. It gives the people who know the line evidence to work with. The rejected part becomes a starting point for a better next cycle.

For Krunon, vision builds on the craftsmanship already in the team. Begin with a specific quality challenge, learn on real production and expand where it adds value.

Krunon and Limen plan to bring the same approach to other key lines: inspect continuously, respond to rejected parts and use the evidence to improve the process behind them.

One component inspected. One defect caught.
One more reason the next cycle can be better.