High-efficient and reversible intelligent design for perforated auxetic metamaterials with peanut-shaped pores

Abstract

Among various types of auxetic metamaterials, the perforated materials with peanut-shaped pores exhibit numerous advantages such as simple fabrication, high load-bearing capability, low stress-concentration level and flexibly tunable mechanical properties, and thus they have received much attention recently. However, one challenging is to make a high-efficient and reversible design of such metamaterials to meet diverse auxetic requirements, without the need to model them through conventional physics- or rule-based methods in time-consuming and case-by-case manner. In this study, a data-driven countermeasure is introduced by coupling back-propagation neural network (BPNN) and genetic algorithm (GA). Firstly, a dataset including microstructure-property pairs is prepared to train BPNN to determine the hidden logic mapping relationship from microstructural parameters to Poisson ratio. Then, GA is employed to optimize the mapping relationship to find the corresponding optimal solutions of microstructural parameters meeting the target Poisson’s ratio. The efficiency and accuracy of specific optimal designs is verified by the tensile experiment and finite element simulation. Subsequently, more optimal solutions corresponding to positive, zero or negative Poisson’s ratios are achieved under constrained/unconstrained conditions to accelerate the design of auxetic metamaterials by this interdisciplinary tool in which the auxetic characteristics and artificial intelligence are interconnected mutually.