Enhancing Flame Retardancy in Polypropylene Composites: A Bayesian Optimization Approach

This study presents a novel application of Multi-Objective Bayesian Optimization (MOBO) to enhance the formulation of flame-retardant polypropylene (PP) composites. Our goal was to optimize the chemical composition of intumescent polypropylene (PP) formulations by maximizing the Limiting Oxygen Index (LOI) and minimizing filler content. Two different initialization procedures were studied: one using a Space Filling Design and the other based on literature-based formulations. The optimization model is based on a Gaussian Process (GP), employing the parallel acquisition function qNEHVI for suggesting experiments. Our research successfully navigates the complexities of material property optimization by proposing an optimal formulation within a constrained evaluation framework. This involved five iterations of three parallel evaluations, totaling a budget of 20 points. This work underscores MOBO’s potential as a transformative tool in advanced materials science, particularly for developing high-performance flame-retardant materials.