Prediction of Flame Retardant Coagent Characteristics Using Deep Learning

This technical report discusses the development of a deep learning model aimed at predicting the characteristics of flame retardant coagents for polyester resins. The primary objective is to identify alternatives to antimony trioxide, a commonly used flame-retardant aid, due to environmental and geopolitical concerns associated with its use. The report details the methodology employed, which includes understanding the flame retardancy mechanism and extracting critical parameters for candidate materials. The authors constructed a model to predict bromination energy and boiling points of potential flame retardants, emphasizing the importance of these properties in ensuring effective flame retardancy. The findings indicate that the identified candidate materials exhibit flame retardancy comparable to antimony trioxide while maintaining essential physical properties. The report concludes by highlighting the significance of materials informatics in accelerating the discovery of new flame retardant materials, which could lead to safer and more sustainable options in polyester resin applications.