In this work, using additive manufacturing via fused filament fabrication 3D printing and finite element simulations conducted in ABAQUS software, the influences of structural parameters on the negative Poisson’s ratio of a newly designed auxetic structure were studied. The key parameters included the length of the arrow-head element (g), the length of the missing rib element (e), and the thickness of the structure (t). From linear to quadratic and polynomial models, the machine learning (ML) regression models have been used to investigate the structure-property relationship. The linear model was statistically significant, but with an R-squared value of only 70.88%, it shows that a large amount of variability in the data was not accounted for. The quadratic model slightly improved with an R-squared value of 77.17%. However, the polynomial model that depicts outstanding predictive power was the one having an R-squared value of 93.52%, which indicated high accuracy with high generalizability. Quantitatively, the support of findings from this study was evidenced by a very low root mean square error of 0.029, supporting evidence of a highly accurate polynomial ML algorithm. Optimization analysis revealed that the most negative Poisson’s ratios, less than −0.92, were reached at the highest magnitude of g and e. This research gave a methodological framework on how to predict and optimize the negative Poisson’s ratio of newly designed auxetic structures.
