Active Learning‐Driven Inverse Design of Polyurethane Foams for EV Battery Applications

The rapid evolution of the electric vehicle (EV) industry demands advanced materials for battery protection, with polyurethane (PUR) foams emerging as a promising solution due to their thermal insulation, mechanical adaptability, and fire resistance properties. This study introduces an active learning‐driven inverse design (AL‐ID) framework, leveraging machine learning (ML) to systematically optimize PUR foam compositions exhibiting desired density and mechanical strength. AL was employed to iteratively refine the ML model by targeting high‐uncertainty regions, reducing experimental effort while improving predictive accuracy. Bayesian optimization (BO) further enhanced the search for optimal compositions by balancing exploration and exploitation. The framework demonstrated significant improvements in model performance, with Mean Absolute Error (MAE) and scores for density and mechanical strength predictions efficiently improving as the dataset grew. Besides successfully selecting 11 good material candidates out of 616,008 virtual compositions, the final ML models have shown small MAE values and good scores. This study underscores the potential of ML‐driven frameworks to accelerate material discovery.