Modular mechanical metamaterials offer unique opportunities for programmable and reconfigurable functionality through simple geometric rearrangements. Inspired by the principle of modularity, we propose a new class of metamaterials composed of three standardized structural units, linear, yield-mimicking, and snap-through buckling elements, which can be assembled into two- and three-dimensional grids to manifest diverse nonlinear stress–strain responses. To accelerate the design and optimization process, we integrate deep learning (DL) into the metamaterial synthesis framework. A convolutional neural network-based predictor rapidly estimates the stress–strain curves of specific modular configurations, achieving high prediction accuracy. Furthermore, an inverse design framework based on a conditional variational autoencoder enables the automatic generation of modular configurations tailored to target stress–strain curves with high fidelity. This DL-driven framework facilitates the rapid, scalable, and reprogrammable design of nonlinear mechanical responses, bypassing the need for exhaustive simulations or manual tuning. This study establishes a universal, data-driven strategy for the inverse design of modular metamaterials, paving the way for intelligent, reconfigurable material systems in applications such as soft robotics and adaptive structures.

Created at: 2026-02-28 03:10:09 +0900 Updated at: 2026-10-07 05:46:11 +0900