OUTER PLANET 46 PETplanet Insider Vol. 25 No. 11/24 www.petpla.net OUTER PLANET AI based sorting technology based on an article by Johannes Laier The German Würzburg-based company WeSort.AI GmbH develops advanced sorting technology using artificial intelligence (AI), enabling waste to be sorted with much greater granularity. This allows new material streams and purity levels to be achieved. Granular waste sorting involves two consecutive steps. First, waste objects are precisely captured using advanced sensor technology, and then algorithmically analysed by AI. This includes, for example, the detailed recognition of object type, material type, colour, manufacturer, and distinctions between food and non-food packaging through novel AI object recognition algorithms. Essentially, the AI not only identifies the material but also recognises the waste object itself. The AI from WeSort.AI has already been trained with several million images of various waste objects, enabling it to cover a wide range of knowledge. In the second step, the recognised waste must be sorted very precisely. For this, a compressed air nozzle bar is controlled by intelligent control software, blowing the waste into different sorting chambers. WeSort.AI’s technology covers both steps. The innovation lies primarily in the AI used to evaluate the sensor data and create a truly intelligent material stream. Current state of technology: The most advanced sorting technology currently available is primarily nearinfrared (NIR) sorters. In this process, waste on the conveyor belt is detected using NIR sensors, and the material composition is identified on a pixelby-pixel basis. NIR sensors typically penetrate a few nanometers to 1mm into the surface, and thus are only able to identify the material at the surface. This often leads to errors, especially with labelled packaging, such as paper around a yogurt cup or a PE sleeve around a PET bottle. In such cases, the yogurt cup may mistakenly be assigned to the paper fraction instead of the PS fraction, degrading the quality of the recycled material. AI technology, on the other hand, is capable of recognising packaging and other waste as a whole. Similar to the human eye, AI not only gathers material information in the form of pixel areas but also understands that it is dealing with a yogurt cup made up of materials (64% PS, 22% paper, 14% aluminium) from manufacturer “XY,” for example, in white and green colours, with a weight of 22g, classified as “food” quality, and a material value of €0.01. This new “knowledge layer” added to the material stream enables smarter detection and, in the next step, sorting. Thanks to AI, the yogurt cup can be recognised as such and correctly assigned to the PS fraction. With the technology, new fractions could be created based on food-grade quality, packaging geometry, or brand, alongside sorting by material types or new individual objects. Object-based recognition allows for both material classification and the identification of specific objects. For example, cheese packaging and yogurt cups can be sorted into the food fraction, while detergent bottles fall into the non-food fraction. Plastics originally intended for food packaging can be reused without being downcycled, reducing the need for new production. Another advantage lies in identifying disruptive and hazardous materials, allowing for their targeted removal from the process. AI material flow analysis and quality assurance Beyond the mentioned applications, this AI technology is claimed to offer a variety of other potential applications to enhance the efficiency of recycling plants. It allows for the rapid and precise collection and analysis of material flow data. Trends are immediately identified, enabling prompt responses to any issues that arise. www.wesort.ai F.l.t.r.: The brothers Johannes and Nathanael Laier founded WeSort.AI in 2021. This year, the company has been awarded this year’s prestigious Deutscher Gründerpreis [Founder’s Award] in the Start-up category.
RkJQdWJsaXNoZXIy NTY0MjI=