Machine learning for expediting next-generation of fire-retardant polymer composites

Abstract

Machine learning algorithms have emerged as an effective and popular decision-making tool for solving complicated engineering problems and challenges. Although introducing these algorithms can accelerate the optimization of fire retardants for polymeric materials by replacing traditional tedious and time-consuming trial-and-error methods, this tool remains at the elementary stage of designing fire retardants for polymeric materials, and thus to date there is a lack of insightful yet review on this topic. Herein, we review the most practical and accurate algorithms used to predict flame retardancy features, such as limiting oxygen index (LOI) and cone calorimetry results, of their polymeric materials. We highlight the merits of some current algorithms, including artificial neural network (ANN), Lasso, Ridge, ANN (L-ANN), and extreme gradient boosting (XGB). Finally, key challenges with existing algorithms for predicting next-generation fire retardants, followed by some proposed solution and future directions. This review will help expedite the development of optimized fire retardants accelerated by machine learning.