TNO uses machine learning to cut polymer development time from years to months
The Netherlands Organisation for Applied Scientific Research (TNO) presented PolyScout at the Rethinking Materials 2026 summit in London this week, a project that combines machine learning with polymer science to accelerate development of bio-based and biodegradable materials for packaging.
The typical polymer development cycle takes 10 to 15 years. PolyScout compresses that timeline by years through machine learning models trained on validated experimental data rather than all available data indiscriminately.
"AI modeling is based on real, validated data from literature, experimentation, and collaborations with partners in the value chain," said Pieter Imhof, senior business developer at TNO and commercial lead at PolyScout. "This is different from the standard approach in which all available data is taken into account - garbage in equals garbage out."
Designing for regulations and real-world constraints
PolyScout works backward from market requirements rather than forward from material properties. The team translates regulatory demands, processing constraints, safety standards, and end-of-life considerations into material specifications.
One project involved designing a single-material soup pouch that could withstand high-temperature filling, prevent leakage, and meet strict safety standards. TNO collaborated with material suppliers, soup producers, and the brand owner to deliver a solution that met all constraints simultaneously.
Another application focused on sustainable packaging for medical devices, where regulatory compliance and performance requirements are equally stringent.
Faster validation and scaling
TNO can produce designed materials in-house, validate polymers through testing, and scale production to required volumes. This end-to-end capability eliminates handoffs between research and commercialization that typically add years to development.
The machine learning approach also explores a larger solution space than traditional methods, increasing the probability of finding viable materials within a shorter timeframe.
Imhof noted that TNO is also implementing a "Safe and Sustainable by Design" approach at the summit, which includes advanced bonding and debonding technologies. These allow researchers to build composites and laminates with desired properties while ensuring the materials can be returned to basic components at end-of-life-for example, using magnetic fields to trigger layer separation.
Learn more about AI for Science & Research and how machine learning applies to scientific discovery.
Your membership also unlocks:
