Abstract:Until relatively recently most mechanical metamaterial classes being studied have been composed of a single solid constituent phase and design has focused almost exclusively on structural geometry. Additional design dimensions can be introduced by accepting heterogeneity and varying materiality, i.e., allowing mechanical properties to vary across the metamaterial's unit cells or even from cell to cell in the metamaterial domain, creating composite metamaterials. This higher dimensionality significantly expands the effective property envelope, but the additional complexity also presents a significant hurdle. To overcome the design challenge, an automated design framework is proposed that leverages modern evolutionary computation techniques, combined with finite element analysis for fitness evaluation, to a discretized or voxelated design domain. However, this approach introduces stochastic and statistical aspects to the design process, which requires additional processing to successfully extract useful solutions. A case study is presented in which the proposed automated design framework is used to generate 2D structures that exhibit pentamode-like behavior. Pentamode metamaterials, which are best known for extreme bulk-to-shear modulus ratios (B/G), offer unique control over effective elastic properties and make for a particularly interesting test case. The evolutionary objective was defined as maximizing B/G over a voxelated square 2D domain. It was found that the evolutionary process converges to a solution relatively rapidly, generally in less that a hundred generations. B/G ratio values of 10,000 and more were obtained, largely exceeding those commonly found in the literature for experimental pentamode metamaterials. These generated designs feature reduced stress concentrations due to the elimination of point-like connections between lattice struts, which addresses a key practical limitation of diamond lattice pentamodes. It was observed that whatever the initial variety of elastic moduli values across the voxels of the design domain, as evolution progressed this variety collapsed to a much smaller number, most often a binary composite of very stiff voxels with a limited number of much softer voxels at key locations that acted as hinges.

Pentamode metamaterial design via generative modeling and deep learning

Acoustic metamaterial design framework using deep learning and generative modeling

Optimal Design of One-Dimensional Elastic Metamaterials Through Deep Convolutional Neural Network and Genetic Algorithm

Topological design of pentamode metamaterials with additive manufacturing

Design and optimization of three-dimensional composite multilayer cylindrical pentamode metamaterials for controlling low frequency acoustic waves

Composite hexagonal pentamode acoustic metamaterials with tailored properties

Data-driven design and optimization of ultra-tunable acoustic metamaterials

An automated design framework for composite mechanical metamaterials and its application to 2D pentamode materials

High Bulk Modulus Pentamodes: the Three-Dimensional Metal Water

The Ultra-low Frequency and Broad Band Gaps Characteristics of Multilayer Composite Cylindrical Three-dimensional pentamode metamaterials

Deep learning, deconvolutional neural network inverse design of strut-based lattice metamaterials

Hierarchical Combinatorial Design And Optimization Of Quasi-Periodic Metamaterial Structures

A transformable anisotropic 3D penta-mode metamaterial

Lattice metamaterials with controllable mechanical properties inspired by projection of four-dimensional hypercubes

Controlling auxeticity in curved-beam metamaterials via a deep generative model

Additively manufactured hybrid anisotropic pentamode metamaterials

A hybrid deep learning approach for the design of 2D low porosity auxetic metamaterials

Two-Dimensional Pentamode Metamaterials: Properties, Manufacturing, and Applications

Inverse Design of Mechanical Metamaterials Using a Point-Cloud-Based Deep Generative Model

Deep learning for Dirac dispersion engineering in sonic crystals

Deep learning-based design of ternary metamaterials for isolating full-mode waves