Modulus tunability in hierarchical architectures: a machine learning-enabled approach

This paper presents a machine learning-enabled approach for achieving modulus tunability in hierarchical architectures, with a specific focus on the inverse design of modulus properties in both the x- and y-directions. By leveraging finite element analysis, a comprehensive dataset was generated, capturing the relationships between material allocation and directional modulus properties across isotropic and anisotropic configurations. A machine learning model was trained to predict the x- and y-directional modulus with remarkable accuracy (R-squared can reach 0.9997), enabling precise forward predictions of mechanical behavior. To achieve targeted modulus properties, an evolutionary algorithm was employed for inverse design, optimizing material allocations to match desired modulus magnitudes with high precision. This dual-direction modulus design methodology offers a powerful framework for tailoring hierarchical architectures to specific mechanical requirements, advancing applications in fields such as soft robotics, aerospace, and advanced manufacturing. The integration of machine learning and optimization highlights the potential for data-driven strategies in the design of customizable material systems.