When collecting old electronic devices, we regularly conduct spot checks. A truck filled with discarded devices is taken to an interim station, such as Les Hautes Ardennes in Vielsalm. There, the devices are placed one by one on a weighing table or platform, and six different cameras take a photo. All this information is fed into a database in the cloud, and then a computer algorithm predicts what kind of product it is.
“At first, the algorithm could only distinguish between product groups: large household appliances, gardening equipment, IT equipment, etc.”, explains Eric Dewaet, CEO of Recupel. “This meant that we knew the main categories of what was being collected. Based on that information, we calculated the contribution that the manufacturers had to pay for the transport and recycling of electronic devices."
“Since 2020, the algorithm can recognise the exact device, such as a hedge trimmer, a weed burner, or a blender. As a result we have a more detailed picture of the material being collected. This information is important for recyclers, because if they know what kinds of valuable materials they can recover and resell from devices, and how much effort is needed for this recovery process, they can make a better estimate of the cost of recycling.”
Recognising reusable devices
The AI system promises to deliver more benefits in the future. “Among other things, we're looking at whether the algorithm could predict whether a device contains a battery”, says Dewaet. “That's important for a recycler to know, as lithium-ion batteries are very flammable if you recycle them incorrectly.”
“The system may also be able to learn to recognise reusable devices by the age of the model and any visible damage. In due course, we'd like to use the algorithm at the locations where the electronic device is being recycled, in doing so our trucks would no longer have to make an intermediate stop.”