Machine Learning-Driven Optimization of Reentrant Chiral Auxetic Structures for Superior Energy Absorption

Abstract Auxetic materials, characterized by their unique negative Poisson's ratio, exhibit exceptional tensile and impact strength, outperforming traditional bulk materials. This makes them ideal for high-performance applications in aerospace, automotive, and biomedical industries, where energy absorption is vital. This study optimizes the topology of reentrant chiral auxetic (RCA) structures. An analytical model was developed to analyze energy absorption with specific unit-cell variables. The mechanical behavior of these metamaterials was parameterized based on the length and height of struts and internal angles between the cell struts. A Taguchi design (L27) was employed to evaluate the impact of six geometric factors on the energy absorbed per unit volume (EAV) and the specific energy absorption (SEA) of the RCA structure. Additionally, an analysis of variance was conducted to statistically assess the comparative significance and contribution percentage of each factor. The Taguchi results have shown that the height of the struts significantly affects both the EAV and SEA. Machine learning models, namely polynomial regression and support vector regression, were developed to enhance predictive capability and optimize within the Taguchi design space. These models captured the nonlinear relationships between geometric parameters and energy-absorption metrics, and were subsequently used to optimize RCA structures with close agreement to finite element analysis results.