Plastic is a much more complex material than you may think. Because it comes in many different types with varied compositions and characteristics, plastic is difficult to separate and recycle at the end of its life. Although, new technology developed by scientists in Denmark may soon change that.
The team behind the study developed a new camera technology that can tell the difference between as many as 12 different types of plastics (PE, PP, PET, PS, PVC, PVDF, POM, PEEK, ABS, PMMA, PC, and PA12) — which make up the vast majority of household plastic types.
As reported by TechExplore, the new technology allows for the separation of plastics based on a purer chemical composition than is possible today, paving the way for better ways to recycle plastics at designated facilities.
Combining machine learning with new camera technology
“With this technology, we can now see the difference between all types of consumer plastics and several high-performance plastics,” says Associate Professor Mogens Hinge, who is leading the project at Aarhus University. “We can even see the difference between plastics that consist of the same chemical building blocks, but which are structured slightly differently.”
According to Professor Hinge, the novel technology combines a hyperspectral camera in the infrared area with machine learning to identify the different types of plastic directly on the conveyor belt.
What technology is currently used to separate plastics?
Currently, recycling facilities use near-infrared technology (NIR) or density tests (determining whether the plastic sinks or floats) to separate certain types of plastics, such as PE, PP, and PET.
These methods, however, are less accurate than the new technology and aren’t as good at distinguishing plastics with a high chemical purity in their composition. The breakthrough separation technology thus opens new opportunities to increase the recycling rate of discarded plastic and reduce pollution.
Source study: Vibrational Spectroscopy — Plastic classification via in-line hyperspectral camera analysis and unsupervised machine learning