Low-density polyamide 12 foams using Bayesian optimization and inverse design

Abstract

This study introduces a novel, comprehensive approach to optimizing and designing batch foaming of low-density polyamide 12 (PA-12) using advanced machine learning (ML) techniques. Bayesian optimization was used to minimize the foam density, which decreased from approximately 900 to 150 kg/m${}^3$ in a single new experiment. A PA-12 foam density of 50 kg/m${}^3$, the lowest achieved, was recorded. In addition, an inverse design approach was used to check the robustness of the model by identifying the specific processing parameters required to achieve the desired foam density. Finally, PA-12 foams with similar densities but different processing parameters were obtained using ML. The study highlights the effectiveness of integrating these ML methodologies in the development of lightweight, high-performance polymer foams, which is much more sustainable than traditional methods for achieving low-density foams.