In the KIOptiPack and K3ICycling innovation labs, the IVV cooperates with 50 partners from industry, science, and society. “The aim is to optimize recycling and the processing route across the board by using artificial intelligence technology,” says Professor Marek Hauptmann. The research takes into account all phases – from sorting to the production of films with a recycled content to the finished packaging.
It all starts in the sorting plants. Intelligent systems with cameras and sensors are now at work there. The smart machine eyes do not just recognize plastic, paper, and metal, they also differentiate between different types of plastic, separate flexible from dimensionally stable packaging and, in future, may also sort according to completely new criteria. This is a complex process considering that the individual consumer articles rarely end up in the bin in the same condition as they appeared on the retail shelf. They are often crushed, crumpled or torn.
The starting point of the current research project is the colorful hodgepodge of household waste that ends up in the yellow bag – a maximal mixture of polymer fractions from different articles. The spectrum ranges from rice bags, potato sacks, and salad wrappers to detergent bottles and tablet blisters. “This mixture causes great difficulties,” emphasizes Hauptmann. This is because every packaging could contain product residues, as well as fillers, pigments, and printing inks used in the production of plastics and packaging – all impurities that reduce the quality of the recycled raw material. “We want to collect this data. We need to know from which sorting fraction, from which plastic stream and which application, which contaminants, additives, and other impurities originate and in what amounts,” says Hauptmann.
Artificial Intelligence for Better Recyclates?
The challenge for packaging with recycled content is to keep up with virgin material in terms of price and properties. In addition, the high safety requirements that apply to the packaging of food, cosmetics, and pet food make their use more difficult. For this reason, the research consortium wants to examine the extent to which the quality of plastic recyclates can be improved with the help of artificial intelligence.
The most important condition: good data material that can be used to feed AI-based models. One example: if the raw polymer data and specific flow properties as well as any impurities in the recyclate were available for the production of a cup made from thermoforming film, in addition to known values such as tensile strength, bending stiffness or tear resistance, a packaging machine could be adjusted more easily and quickly and would also work smoothly in the event of fluctuations. “However, this is only possible if we start collecting the data at the regranulate stage,” says Hauptmann. So far, however, the connection of market participants and the flow of this data across company boundaries has not been comprehensive. There are understandable concerns about making sensitive data available.
By using the European Gaia-X platform, data is no longer stored on external resources. This is a new approach to data management that results in more specific data flows. The data does not have to be available to other market participants in an analyzable format. Everyone keeps theirs on their own server. A model that can work using artificial intelligence methods is granted access rights and only passes on the result. This benefits another participant in the value chain because, for example, processing temperatures, times, and other parameters can be predicted.
For Hauptmann, the future lies in a material cycle that is as close to circular as possible. After all, it is currently not possible to do completely without plastic, as it is currently primarily plastic that guarantees a specific shelf life for food and other products.
