Customizable metamaterial design for desired strain-dependent Poisson's ratio using constrained generative inverse design network

Inverse design of metamaterial structures with customized strain-dependent Poisson's ratio has significant potential across various applications. However, achieving precise control over these mechanical properties presents a challenge due to the complex relationship between geometry and mechanical performance. Here, we present a novel data-driven approach utilizing a constrained generative inverse design network (CGIDN) to address this challenge. The CGIDN uses backpropagation to efficiently navigate the design space and achieve target mechanical properties with high accuracy. Our method starts by generating a comprehensive dataset of Poisson's ratio-strain curves for various geometries incorporating cuts. These curves are then compressed using principal component analysis (PCA) to reduce dimensionality while preserving essential features. A deep neural network (DNN) is then trained to map input geometric parameters to these principal components, with the architecture optimized using grid search. The CGIDN facilitates the inverse design process by recommending geometric parameters for unit cell designs that match specified target Poisson's ratio-strain curves. We validated the effectiveness of our approach through Finite Element Analysis (FEA) and experimental verification. The FEA results for the designed unit cells showed high agreement with the target and predicted curves, demonstrating the accuracy of the CGIDN model. Further, tensile tests on specimens confirmed that the inverse-designed structures reproduced the desired mechanical behavior upon scale-up. Our method, which enables efficient and accurate design of metamaterials with tailored mechanical properties, holds promise for applications in wearable devices, soft robotics, and advanced sensor systems.