Design exploration of additively manufactured chiral auxetic structure using explainable machine learning

Abstract

A design exploration of hexachiral structures using explainable machine learning (ML) is performed in this work. The hexachiral structures are fabricated using resin via vat photopolymerization (VPP). The ML model is used to build the function that explains the association between Poisson’s ratio and the hexachiral design parameters. The data set for ML model construction is first collected by using the Halton sequence and simulated using the finite element method (FEM). To validate the data set, the results obtained from the FEM simulation are compared with those obtained from the compression test. The Gaussian Process Regression (GPR) models for Poisson’s ratio and porosity are constructed to extract important design insight. A Global Sensitivity Analysis (GSA) and Shapley Additive Explanations (SHAP) are used to analyze the sensitivity of the porosity and Poisson’s ratio to the hexachiral design parameters. GSA result shows that the strut’s thickness is the most decisive parameter that affects the Poisson’s ratio. The application of SHAP also reveals that the relationship between the strut thickness and Poisson’s ratio is nonlinear. Finally, the minimum Poisson’s ratio value is achieved by design with minimum strut thickness, minimum node radius, and maximum strut length.