A new AI-driven testing system from University at Buffalo engineers can tell how much of a plastic product is truly recycled, offering a powerful tool for regulators and companies as global rules on plastic waste tighten.
The next time you see a water bottle or shopping bag boasting recycled content, a new technology under development at the University at Buffalo could one day verify whether that claim is real.
Engineers at UB have created a lab-based system that uses a mix of electrical and optical tests, plus artificial intelligence, to determine how much recycled plastic is in a product. In early trials on one of the world’s most common plastics, the method correctly identified recycled content levels more than 97% of the time.
The work, published in the journal Communications Engineering, aims to solve a major blind spot in the global push to cut plastic waste: there is currently no quick, reliable way to check how much recycled material is actually in a plastic item.
The team is targeting both speed and trust, according to corresponding author Amit Goyal, a SUNY Distinguished Professor and SUNY Empire Innovation Professor in UB’s Department of Chemical and Biological Engineering.
“Our goal is to create a quick and reliable tool that can be used to verify recycled material content, as well as enforce recycling regulations,” Goyal said in a news release.
Today, many companies voluntarily advertise recycled content, and governments in some places are starting to require it. But regulators and consumers largely have to take those labels on faith. Recycled plastic is melted, cleaned and remolded, ending up with a look and overall chemistry that closely resemble brand-new plastic.
At the microscopic level, though, recycling leaves fingerprints. Repeated processing can introduce tiny impurities and break some of the long polymer chains that give plastics their structure. Those subtle changes affect how the material behaves when exposed to electricity and light.
To pick up those signals, the UB team combined four different sensing techniques:
– Triboelectric testing, which measures how a plastic builds and holds static electricity when surfaces touch. Recycled plastics tend to hold a charge longer because of structural defects from prior processing.
– Dielectric and impedance spectroscopy, which applies an electric field to see how the plastic stores and loses energy. Recycled samples typically show lower energy storage and higher energy loss.
– Capacitance analysis, which tracks how quickly a plastic charges and discharges in a circuit. Shifts in timing can reveal changes in electrical properties caused by recycling.
– Mid-infrared spectroscopy, which shines specific wavelengths of light through the plastic to probe its chemical structure and detect fragmented polymer chains.
The researchers tested the approach on PET, or polyethylene terephthalate, the clear plastic used in soda and juice bottles, peanut butter jars and many food containers. They prepared samples with recycled content ranging from 0% to 50% and ran each sample through the four tests.
On their own, each technique offers only a partial view. To pull everything together, the team turned to machine learning, a form of artificial intelligence that can find patterns in large, complex data sets.
They trained a machine learning model on the combined test results so it could learn which patterns corresponded to specific percentages of recycled PET. Once trained, the system was more than 97% effective at determining the recycled content of the test samples.
Goyal called the project “an ideal example of combining cutting-edge innovation in science and engineering with AI for social good, and to potentially realize significant societal impact,” highlighting its potential to reshape how recycled plastics are monitored worldwide.
If such a system can be made practical outside the lab, it could give regulators a powerful enforcement tool, help honest manufacturers prove their claims and discourage greenwashing — the practice of exaggerating environmental benefits. It could also give recyclers and product designers better feedback on material quality, encouraging higher-value uses of recycled plastic instead of downcycling it into lower-grade products.
The UB team is already looking ahead to how this technology might be used in the field. Their next goal is to integrate the sensing methods and AI model into a single, portable device.
“By fabricating such a device, we hope to enable widespread, real-time monitoring of recycled plastics in commercial products,” Goyal added.
A handheld or benchtop tool could be used at factories, recycling facilities, ports and retail distribution centers to spot-check products. That kind of real-time verification may become increasingly important as more governments move from voluntary targets to binding rules on recycled content.
The timing is significant. States and countries around the world are considering or adopting regulations that require plastics to include a minimum share of recycled material. At the global level, the United Nations’ Intergovernmental Negotiating Committee is working on an international, legally binding agreement to end plastic pollution, which is expected to include stronger measures on plastic production and recycling.
If those rules are to be effective, officials will need ways to confirm that products actually meet the standards on paper. UB’s system is designed to fill that gap.
The team’s ultimate goal is not just better policing, but better plastics.
Goyal emphasized that the tool is meant to “improve the quality of plastic products and help reduce plastic waste, which will support a more circular economy where plastic pollution and its associated health and environmental risks are reduced.”
For now, the system has been demonstrated on PET in controlled experiments. Future research will likely focus on testing other types of plastics, refining the AI models and engineering a rugged, user-friendly device that can withstand real-world conditions.
If successful, the technology could help turn recycled-content labels from marketing claims into verifiable facts — and give students, consumers and policymakers a clearer picture of how far we have come in closing the loop on plastic waste, and how far we still need to go.
Source: University at Buffalo, SUNY