Künstliches Gespür für Kunststoffe

Illustration Umweltpolitische Digitalagenda

In Germany there are still too few high-quality recycled materials from plastic waste, which are then reused in production. The ReCircE project not only aims to improve the waste sorting process using artificial intelligence, a Digital Product Passport is intended to create transparency for the entire chain of recyclable materials, thus making it easier to reuse granulate from complex products like electric kettles, car seats or toy dolls.

At 6am, 80 kilometres north-west of the city of Kaiserslautern in Germany, the first shift is starting at Papier-Mettler, Europe's largest producer of paper and plastic packaging. The factory gate rattles up as a lorry comes to a stop in front of it. The plastic waste being transported in the back of the lorry is about to get a new lease on life. Plastic films, bags and plastic packaging are collected in the region's supermarkets and department stores and brought to the recycling facility. Here, the plastic waste is crushed, washed and melted down to produce a fine granulate, which is a valuable raw material for new packaging.

Not enough recycling of plastic waste

700 kilometres away, in Berlin, Andreas Ciroth is pouring himself a cup of tea in his office. The founder of the company GreenDelta bought the electric kettle a few years ago. If the appliance breaks at some point, it cannot be entirely recycled. "The kettle is made of lots of different materials, and separating these materials for plastic recycling of electrical appliances is a very complex process," says Ciroth. It's just not worth it financially.

Most plastic waste in Germany is recovered for use as energy, for example, in waste incineration plants, but it is also sometimes shipped abroad. This is something environmental engineer Andreas Ciroth would like to change with artificial intelligence. And this is where the recycling experts from Rhineland-Palatine come into play. Papier-Mettler, Andreas Ciroth's company GreenDelta and other well-known partners from the private sector and scientific community joined forces under the ReCircE project to reorganise the entire sorting process in preparation for high-quality recycling. The consortium also includes the German Research Center for Artificial Intelligence, the Technical University of Darmstadt and the Fraunhofer-Gesellschaft for the Advancement of Applied Research. Together they applied to the BMU support programme as the Artificial Intelligence (AI) lighthouse project for the environment, climate, nature and resources.

Goal to use the same plastic compound up to 20 times

ReCircE aims to use artificial intelligence to remedy weaknesses in the sorting process. Near infrared spectroscopy has already been established as a sorting technology for packaging such as plastic films and bags. It detects the most common polymers and can sort them automatically. However, it is not yet possible to sort plastics made of complex products with highly diverse materials and possibly even harmful substances.

As a result, manufacturers of products like car parts, food packaging or textiles tend to use new plastic. Using granulates from recycled materials is generally not an option for them for quality and cost reasons. In theory, molecular chains of plastic could be used as many as 20 times, says Ciroth. However, for this to happen, the recycling process would have to be significantly improved. One breakthrough could be combining the established sensor-supported processes with the new possibilities offered by machine learning. Which is exactly where ReCircE is focussing its efforts.

Artificial intelligence will produce pure granulate in the future

It starts when waste is sorted on the conveyor belt. With an intelligent combination of infrared technology and 3D object design, sensor-supported machines could analyse the composition of the waste it is trained to detect and distinguish between, for example, commercial and household waste. The Fraunhofer Institute developed the prototype of a machine for ReCircE which can even be trained using artificial intelligence. "Over time, the machine will be able to recognise the origin of the waste and predict the likely composition more accurately," says Ciroth.

The waste is then better sorted by material class when it enters the melting process, where it can be further refined using AI. This occurs after the plastic has been melted into hard pellets. In the future, the aim is for AI to be able to detect the individual molecular chains in this state and treat the melted plastic in such a way that it breaks down into the individual fractions. The goal is not to have a mixed granulate, but to create four or five pure grades of a quality comparable to primary plastics. This will enable more plastics to be turned into a valuable raw material for industry.

Papier-Mettler will be the first company to test this new AI-supported process. Initially, the share of recycled materials in plastic bags and simple plastic films is to be increased from the current level of 80 percent to nearly 100 percent. The company will then start to use recycled granulate, also for industrial films, which Papier-Mettler currently makes with new granulate.

A Digital Product Passport aims to ensure transparency and repeated recycling

But the planned project intends to use AI even more extensively: the idea behind the Digital Product Passport, a kind of record of the product's life cycle, is make it easier for manufacturers and recyclers to deal with plastic properly. Until now, companies like Papier-Mettler have not known enough about the substances contained in certain products. And, when in doubt, they don't recycle them at all. Manufacturers, on the other hand, unintentionally use methods that exclude products from recycling, including adhesive that the recycler can no longer remove. A Digital Product Pass would provide this important information to everyone involved at the different stages of the material cycle: how materials were used, how they are combined and how the product needs to be disposed of.

GreenDelta has already developed an open source tool for this purpose that enables everyone from the manufacturer to the retailer and recycler to read and add information. Ciroth: "This transparency will let us create a material cycle that can be repeated numerous times."

However, in order for this tool to be of benefit, input from all stakeholders involved in a wide range of products is required, from for example electric kettles to dolls or smartphones. "It will be a major undertaking to convince a wide range of manufacturers and distributors to participate," says Ciroth. "We want to initiate this process and hope that it can become established throughout Europe over the next few years." A recycled electric kettle contains hardly any petroleum and generates 65 percent fewer CO2 emissions.

At some point, these efforts will lead to a plastic from the recycling process that can be used to make the electric kettle later used in GreenDelta's staff kitchen. Its production will not only minimise the use of fossil fuels, it will also result in 65 percent fewer CO2 emissions. Ciroth sees no alternative to plastics in the long run. No other material can be shaped and used as flexibly. "The only thing we need to get to grips with is achieving circular economy, otherwise plastic is an enormous environmental problem." But this is exactly what he and his colleagues at ReCircE are working on.

Further information