Scientists at TNO, the Netherlands Organisation for Applied Scientific Research, have develop an artificial intelligence (AI) model that speeds up the development of biodegradable polymers.
Called polyScout, the machine-learning programme can predict the chemical structure of polymers.
“If you need a material with certain properties, you enter those properties into the model, and then the model produces a chemical structure for a suitable polymer,” scientist Milad Golkaram explained.
Golkaram’s team started developing the technology in 2022. The predictive model follows a number of steps to produce polymer chemical structures that researchers can then test in the lab. First, it searches for data in existing literature about existing polymer structures. It then creates new data through experiments, obtaining data-driven structure-property functions which return polymer structures when researchers input desired properties. This allows it to uncover new, safe, and sustainable polymers.
“Back when I was studying, we had to go into the lab to develop new polymers,” Golkaram said. “It was a matter of trial and error, a process that can take years. With machine learning, our model learns more and more about the properties of polymers. It can recognise correlations faster than people can and draw conclusions from them. This means you actually get the desired result almost immediately.”
polyScout is already in use in the JTF-project, where TNO is partnering with Senbis, a producer of biodegradable plastics and fibres focused on addressing microplastics pollution. TNO is assisting Senbis by using polyScout to help develop a biodegradable polyester to be used as textile fibre.
“Some are sceptical about the use of AI, or think it’s too complicated. They are always positively surprised when we show them what we can do with polyScout. As a result, we now have several companies who want to work with us, especially in the Netherlands. Getting a foothold in the rest of Europe is our next challenge,” Golkaram said.