AI-based sorting: the smart way to recycle plastics

Image: Piles of plastic waste sorted by colour; Copyright: STEINERT — STEINERT's AI-based technologies are making the sorting of plastic waste at RE Plano even more reliable and efficient. Copyright: STEINERT

22.11.2023

Updated on 24.02.2025: We have revised this article to include the latest developments and provide the most up-to-date information.

Artificial intelligence and plastics recycling – two concepts that have little in common at first glance, but which form an impressive symbiosis in practice. STEINERT and RE Plano are impressive proof of this. Thanks to AI, they have succeeded in overcoming the challenges of sorting polyethylene and polypropylene, paving the way for innovative recycling solutions.

Image: Michael Köster; Copyright: STEINERT — Michael Köster. Copyright: STEINERT

How AI Enhances Plastic Sorting Efficiency

In an interview with K-MAG, Michael Köster talks about the collaboration, the innovative sorting technology and the extent to which AI-controlled processes improve the quality and efficiency of recycling plastic waste.

Mr Köster, how did the partnership between STEINERT and RE Plano come about and what goals are you pursuing with the implementation of AI-based sorting technologies?

Michael Köster: In order to enable closed-loop applications for PP and PE plastics, such as those that exist for PET, RE Plano was looking for a partner that could sort these materials by colour and material in a correspondingly good quality. This is possible with the sensor fusion of colour and NIR cameras in STEINERT UniSort sorting systems and is even more reliable and efficient thanks to AI-based sorting programs.

What are the challenges involved in sorting plastic fractions such as polyethylene and polypropylene and how does AI-based technology help to overcome them?

Köster: There are foreign substances in the material flow that cannot be sorted using conventional technology. One example of this is silicone cartridges made from HDPE, which can contain silicone residue and therefore contaminate the PE sorting product. STEINERT sorting systems are able to sort out these materials using AI-based sorting programmes. The same applies to single and multi-layer packaging. In addition, the reliability of sorting white and natural-coloured materials can be improved.

Exhibitors & products around digitalisation

Intelligent Object.Identifier – The Future of AI Sorting

STEINERT's Intelligent Object.Identifier technology combines various state-of-the-art technologies. Can you tell us more about this?

Köster: In our UniSort sorting systems, we use an NIR and a colour camera that look at the same point on the conveyor belt. We call this sensor fusion. As the information from both sensors overlaps so perfectly (as the material on the belt cannot move between sensor 1 and sensor 2), this information is the best teaching material for an AI to generate our sorting programmes from. We also use a hyperspectral imaging NIR camera. This has many times the spatial and spectral resolution within the near-infrared spectrum compared to conventional point-to-point scanners. This additional information also helps to generate particularly reliable AI sorting programmes.

Image: Sorting plant; Copyright: STEINERT — The UniSort sorting systems from STEINERT provide optimum training material for an AI software. Copyright: STEINERT

Success Story: AI Sorting at RE Plano

What advantages does it offer over conventional sorting methods?

Köster: New sorting applications are made possible and existing ones can be executed with greater reliability.

Can you share some specific results or successes that have been achieved since the implementation of STEINERT technology at RE Plano?

Köster: The RE Plano plant in Bochum did not exist before, which is why no improvements can be named in comparison. However, purities of over 97% are produced there, which is the first time a purity similar to that of PET cycles has been achieved. and that the recycling of PP and PE is improved compared to the industry.

The Future of AI in Plastic Recycling

What advice would you give to other companies in the plastics recycling industry that are interested in implementing AI-based sorting technologies? Are there any particular challenges or stumbling blocks they should be aware of?

Köster: Nothing that can't be solved together. In order to train the AI, the materials have to be scanned in advance in our test centre. This is a more complex process than with conventional materials. However, we have already done this for many products, such as the silicone cartridges, so that these sorting programmes can be used immediately.

What motivates you personally most about working in the field of plastics recycling?

Köster: It is important to find a balance between reducing plastic on the one hand and the sustainable use of recyclable plastics on the other. The growing demand for high-quality plastics such as those used in food packaging or the automotive industry – to name just a few – is also increasing the demands on plastics recycling. New and exciting tasks await us, because we have a great responsibility for the future of our planet and our children!

Want to learn more about AI-powered plastic solutions and the use of artificial intelligence? Read our latest insights on K-MAG!

Elena Blume (Editorial team K-Mag)