This study employs the NSGA-Ⅱ algorithm integrated with neural network models for the inverse design of a cosine beam honeycomb (CBH), aiming to achieve the coordinated control of zero Poisson's ratio (ZPR) characteristic and predetermined force-displacement curves. Structural sensitivity to various parameters is evaluated via univariate parameter analysis, followed by the utilization of Optimal Latin Hypercube Sampling (OLHS) to acquire uniformly distributed samples within the design space. 2500 corresponding models are generated using parametric modeling. Pythonbased scripts are employed to implement the automatic operation of finite element analysis, accomplishing the collection of Poisson's ratio curves and force-displacement curves for each metamaterial. A classifier is trained to rapidly determine the presence of ZPR characteristics in the structure, while a regressor is developed for the high precision fitting of the force-displacement curves. For multi-objective optimization in inverse design, the pre-trained neural network models are integrated with the NSGA-Ⅱ algorithm, with "satisfying ZPR characteristic" and "attaining the targeted force-displacement curve" as dual optimization objectives. Optimal structural parameters meeting the performance requirements are efficiently retrieved from the high-dimensional design space, thereby providing a systematic design methodology for the engineering application of zero Poisson's ratio metamaterials.
Wang et al. (Thu,) studied this question.