The Polymer Society of Korea

Abstract

Recent advances in soft mechanical metatmaterials has highlighted the potential of 3D soft auxetic metastructures, however, accurate prediction of their dynamic properties remains challenging. In this work, two complementary approaches are presented to address this gap. First, a wave propagation-based theoretical–experimental method was developed to evaluate the dynamic properties of re-entrant honeycomb auxetic structures fabricated from viscoelastic polymers. Second, a vibrational physics-guided neural network (VPGNN) was proposed to directly estimate natural frequency and damping ratio from impulse responses by embedding analytical impulse response functions and physical metadata into the learning process. These strategy bridges experimental data, physical modeling and deep learning, enabling accurate, interpretable and scalable dynamic characterization of soft auxetic metastructures, thereby expanding their potential in engineering applications.