Researchers at IMDEA Materials Institute have developed an artificial intelligence (AI)-based strategy to predict and assess the fire resistance of epoxy resins, one of the most widely used polymers in industry.
The new system, published in Polymer Degradation and Stability, enables the rapid and cost-effective identification of safer and more efficient materials, overcoming the lengthy and expensive trial-and-error cycles associated with traditional methods.
Epoxy resins are used in sectors such as construction, automotive, electronics, and aerospace. However, their inherent flammability poses a significant safety risk, limiting their use in critical applications.
To address this challenge, flame-retardant additives are incorporated into these materials. Among them, phosphorus-based flame retardants (P-FRs) are considered one of the most promising and safer alternatives, as they do not contain halogens.
“The traditional development of efficient flame retardants involves a design, synthesis, and laboratory testing process that is slow, costly, and highly sensitive to experimental conditions,” explains Dr. Qiong Tan, a Marie Skłodowska-Curie Actions Postdoctoral Researcher.
“To overcome this obstacle, we created a machine-learning model trained on data from 510 EP composite samples incorporating phosphorus-based flame retardants.”
“The model analyses the molecular structure of the flame retardants, the formulation of flame-retardant epoxy resins, and other variables to accurately predict two key indicators of fire resistance: the UL-94 vertical flammability rating and the Limiting Oxygen Index (LOI),” she adds.
The UL-94 classification is a test that evaluates the burning behaviour of polymeric materials under a standardised ignition source and classifies them based on their afterflame time, afterglow time, and flame dripping characteristics.
Meanwhile, the Limiting Oxygen Index (LOI) measures the minimum concentration of oxygen required in a gaseous mixture of oxygen and nitrogen that will support combustion of a material under specified test conditions.
“The combination of both indicators provides a much more comprehensive and reliable picture of the material’s real-world fire behaviour,” says Dr. Tan.
Beyond prediction, the main innovation of this work is a new evaluation framework that integrates AI results into a unified and easy-to-interpret classification system.
The framework categorises a material’s flame-retardant performance into four semantic levels (excellent, good, moderate, or poor), providing clear and direct guidance for engineers and materials designers. This methodology enables more informed and scientifically grounded decision-making, helping to prioritise polymers with the greatest potential.
The model has already been successfully validated using external case studies, demonstrating its robustness and practical applicability.
Through its High-Performance Polymers and Fire Retardants Research Group, led by Prof. De Yi Wang, IMDEA Materials plans to expand the database to include other types of polymers and flame retardants.
“Its direct applications could improve the safety of electronic components, electric vehicle batteries, aircraft interiors, and construction materials,” concludes Dr. Tan.
This work was supported by the European Union under the Marie Skłodowska-Curie Actions Postdoctoral Fellowships (HORIZON-MSCA-2024-PF-01), Project 101205162 – FireDesign.