A framework for comparative evaluation and data-driven constitutive modelling of cellular materials under dynamic loading

Cellular materials are widely used in impact-resistant structures due to their lightweight characteristics and excellent energy absorption capability. However, selecting suitable cellular materials for dynamic loading environments and accurately modeling their mechanical responses remain challenging, particularly under high-temperature and high-strain-rate conditions. In this study, a framework for comparative evaluation and constitutive modeling of cellular materials is proposed to support material selection and engineering modeling for impact-resistant applications. Compression experiments were conducted on three representative cellular materials, namely aluminum honeycomb (AH), aluminum foam (FA), and rigid polyurethane foam (RPUF), under varying strain rates and temperatures to systematically compare their mechanical responses and energy absorption characteristics. Microstructural observations using scanning electron microscopy (SEM) were performed to interpret the deformation and failure mechanisms. The results indicate that FA exhibits relatively stable energy absorption efficiency under elevated temperatures and high strain rates, while AH shows higher specific energy absorption at room temperature but greater sensitivity to extreme conditions. RPUF presents lower energy absorption capacity and stronger temperature dependence. Based on the comparative evaluation results, FA was identified as a suitable candidate for dynamic energy absorption applications. A dataset consisting of 305 stress–strain curves was established and a data-driven rate-dependent constitutive model (DD-RDCM) was subsequently developed for FA using an optimized neural network architecture combined with a particle swarm optimization–genetic algorithm (PSO–GA) strategy. Finally, drop-weight impact experiments on sandwich panels were conducted to evaluate the engineering applicability of the proposed approach. The DD-RDCM model successfully predicts the dynamic response of FA sandwich structures with prediction errors within 15% and improves prediction accuracy compared with the classical Sherwood–Frost model. The proposed framework provides a practical approach for the comparative evaluation, material selection, and constitutive modeling of cellular materials in impact-resistant structural design.