Radiation can bring changes in polymers that enable their use in many industrial and healthcare applications. A new IAEA coordinated research project will explore how machine learning can improve prediction of structural changes in polymers caused by ionizing radiation. (Graphic: IAEA)
The IAEA is inviting research organizations to join a new project that will use machine learning to better predict structural changes in polymers caused by ionizing radiation.
Radiation-induced effects in polymers play a critical role across diverse industrial and healthcare applications, from nuclear power plant cable insulation to medical equipment sterilization.
Polymers are large molecules made up of repeating building units called monomers. They form long structures like chains or branches and can be both natural (such as proteins, cellulose and starch) or synthetic (such as nylon, paper, plastics, rubber). When polymers are subjected to radiation, their structure changes. This polymer modification process triggers various effects including oxidation, cross-linking and chain scission, leading to significant alterations in their chemical, physical and mechanical properties, which can be put to good use.
Running Blind on Radiation-Polymer Interaction
The challenge for engineers and materials scientists has been predictability. While it is well established that radiation causes fundamental changes in materials, a precise understanding of radiation-polymer interactions remains limited. Devising effective polymer design, modification and fabrication strategies for each new application requires expensive, time-consuming experimental testing. This slows down innovation, increases costs and can ultimately hinder long-term sustainability.
Today, machine learning (ML) can predict many things in complex systems, from weather patterns to consumer behaviour. However, the development of ML-based predictive tools for polymer changes under radiation has been hindered by the lack of comprehensive and reliable data catalogues. Key information that could drive progress is scattered across decades of academic papers, locked in proprietary industrial databases, or missing entirely. There is no unified, high-fidelity catalogue of validated data that an ML algorithm can use to learn the deep patterns in polymer changes under radiation.
CRP Objectives
To address this challenge, a comprehensive and validated database will be the first step for data-driven modelling of radiation effects in polymers. The new IAEA Coordinated Research Project (CRP), titled Data-driven Prediction of Structural Changes in Polymers Induced by Radiation , aims to address this gap. The five-year CRP will create a validated database of polymer-radiation interactions through a systematic review of existing literature and targeted experimental work to fill data gaps.
The ultimate goal is to enable the development of a robust database for ML predictive models that can simulate radiation-induced polymer behaviour under various conditions.
The CRP's methodology focuses on three pillars:
- Structured database: Designing the data structure by curating and validating decades of scattered literature data into a single, standardized source.
- Targeted experiments: Conducting experiments to fill data gaps where previous literature is missing or contradictory.
- ML for predictive models: Developing and training predictive models on radiation effects on polymers.
Running from 2026 until 2031, this CRP aims to:
- Identify the initial set of polymers and parameters of interest and assign different selected known polymers to participating research teams.
- Collect and validate existing data for corresponding polymers.
- Fill in data gaps and expand datasets on polymer-radiation interactions.
- Build a validated database and develop predictive models.
How to join the CRP:
