March 13, 2026 10:57

Through the AutoROCK research project, the two German research institutes - SKZ and Fraunhofer IIS - aim to use artificial intelligence (AI) to automatically detect defects generated during profile extrusion, directly on the production line.

The goal is to identify quality deviations at an early stage, before they lead to scrap, material losses or costly rework, by dynamically adjusting process parameters.

The first step was to create a reliable database by specifically extruding profiles with characteristic defects. For the same purpose, several profile manufacturers supplied rejected products with typical defects, including cavities, foreign material inclusions, porosity and geometric errors.
The samples were analysed using state-of-the-art X-ray computed tomography systems and established non-destructive testing (NDT) methods.

The resulting dataset forms the basis for training an AI system able to automatically detect defects, classify them and assess their relevance. “Thanks to the variety of the collected data,” the researchers said, “the artificial intelligence will learn to identify even complex or overlapping defect structures.”

In parallel, a concept has been developed for a demonstrator (pictured) that will eventually enable inline measurements during extrusion. The requirements are demanding: it will have to ensure reliable measurements despite the continuous movement of the material, high extrusion speeds and limited installation space, while also meeting radiation protection requirements.
The device will also need to be optimised for integration into existing lines without disrupting the process. Practical tests on the demonstrator will take place during the year.

© Polimerica -  Reproduction prohibited, all rights reserved