Machine learning-enabled rational design of organic flame retardants for enhanced fire safety of epoxy resin composites

This study proposed an approach utilizing machine learning (ML) to accelerate the design of organic flame retardants (FRs) for epoxy resins (EPs), avoiding the limitations of traditional trial-and-error methods. For the first time, ML models have been established and considered for five pivotal parameters: limiting oxygen index (LOI), peak heat release rate (PHRR), total heat release (THR), time to ignition (TTI), and vertical combustion test (UL 94) level. These models were employed to consider and assess the significance and relevance of FRs structure and addition amount to the essential flame retardancy of EPs. The ML models showed excellent performance, with the coefficient of determination scores around 0.8 for the test set. Utilizing key structural insights gleaned from these ML models, a FR referred to as BDOPO was employed here to experimentally verify the changes in the properties of EP composites loaded with different amounts of BDOPO (EP/BDOPO), and the results showed that, except for the TTI, the ML models could accurately predict all the other properties of EP/BDOPO. The study also elucidated the flame retardancy mechanism of BDOPO in EP. This approach provides an effective method for designing organic FRs for high-performance EP.