Deep-learning-based low-frequency bandgap prediction and elastic wave propagation properties of two-dimensional locally resonant metamaterials

Flexible design of bandgaps is a challenge in phononic crystals for vibration damping and noise reduction applications. In this study, based on novel 2D localized resonance (LR) metamaterials, the correlation between the spatial distribution of the materials and the bandgap properties is innovatively constructed, and the bandgap formation mechanism under the synergistic effect of microstructural parameters is revealed. An intelligent design framework integrating finite element method (FEM) and artificial neural networks (ANNs) is developed, which realizes the rapid generation of massive data sets through parametric modeling techniques, significantly shortens the traditional trial-and-error design cycle, and comparatively verifies the enhancement effect of different network architectures on the bandgap prediction accuracy. It is further found that the bandgap range can be synergistically regulated by the periodic array structure size and uni/biaxial compressive strain, which verifies that the low-frequency vibration suppression bandwidth can be effectively expanded by the combination of periodic array structure size optimization and dynamic compression control, and can be used as a strategy to realize the active design of the bandgap. This predictable bandgap design paradigm breaks through the limitations of the traditional trial-and-error method, and provides a new theoretical framework and technical path for the development of vibration- and noise-reducing metamaterials with adaptive adjustment capability, which shows important application value in the fields of intelligent vibration isolation equipment and precision instrument protection.