Data-Driven-Design (D3) of multi-material systems: A novel framework and its application to viscoelastic metamaterials

While multimaterial additive manufacturing enables finely programmed heterogeneity, there remains no robust and objective-driven framework to assign materials across complex architectures under practical constraints. We introduce Data-Driven-Design (D³) as a robust computational framework for multi-material lattice design, optimized with respect to a prescribed performance objective. The framework relies on representing physical constraints, material data, and design objectives as sets in a phase space and formulating the material selection problem as a distance minimization problem among the encoded sets. We showcase the approach in multi-material design of viscoelastic lattices provided with measurements of complex moduli as a function of frequency with the design objective of maximizing dissipation. For our numerical experiments, we import dynamic viscoelasticity measurement for twenty five different materials from literature, and show that multi-material designs can match or outperform the dissipation obtained from homogeneous designs made of the most dissipative material among the data registry. In a finite lattice example, we show that D³ design yields a mechanical dissipation with 300% increase compared to best homogeneous design from a limited collection of materials. Beyond viscoelastic lattices, the D³ framework generalizes naturally to multi-physics and multi-objective metastructure design, offering a unified, data-driven approach to optimal material selection under complex constraints.

Duke Scholars

Author Hossein Salahshoor Civil and Environmental Engineering