Abstract:The trade-offs between different mechanical properties of materials pose fundamental challenges in engineering material design, such as balancing stiffness versus toughness, weight versus energy-absorbing capacity, and among the various elastic coefficients. Although gradient-based topology optimization approaches have been effective in finding specific designs and properties, they are not efficient tools for surveying the vast design space of metamaterials, and thus unable to reveal the attainable bound of interdependent material properties. Other common methods, such as parametric design or data-driven approaches, are limited by either the lack of diversity in geometry or the difficulty to extrapolate from known data, respectively. In this work, we formulate the simultaneous exploration of multiple competing material properties as a multi-objective optimization (MOO) problem and employ a neuroevolution algorithm to efficiently solve it. The Compositional Pattern-Producing Networks (CPPNs) is used as the generative model for unit cell designs, which provide very compact yet lossless encoding of geometry. A modified Neuroevolution of Augmenting Topologies (NEAT) algorithm is employed to evolve the CPPNs such that they create metamaterial designs on the Pareto front of the MOO problem, revealing empirical bounds of different combinations of elastic properties. Looking ahead, our method serves as a universal framework for the computational discovery of diverse metamaterials across a range of fields, including robotics, biomedicine, thermal engineering, and photonics.

What problem does this paper attempt to address?

The problem that this paper attempts to solve is: in material design, the trade - offs between different mechanical properties pose fundamental engineering challenges. Specifically, how to balance the stiffness and toughness of materials, the relationship between weight and energy - absorption capacity, and various elastic coefficients. Although gradient - based topology optimization methods are very effective in finding specific designs and properties, they are less efficient in exploring the vast design space of complex metamaterials and cannot reveal the achievable boundaries of interdependent material properties. In addition, parametric design and data - driven methods also have limitations. The former lacks geometric diversity, and the latter is difficult to extrapolate from known data. To solve these problems, this study proposes a new framework that formalizes the simultaneous exploration of multiple competing material properties as a multi - objective optimization (MOO) problem and uses a neuroevolutionary algorithm to solve this problem efficiently. Specifically, the researchers adopted Compositional Pattern - Producing Networks (CPPNs) as a generative model for unit cell design. By modifying the Neuroevolution of Augmenting Topologies (NEAT) algorithm, they evolved CPPNs so that they could create metamaterial designs on the Pareto front of MOO problems, thereby revealing the empirical boundaries of different elastic property combinations. The following are the key formulas of this method: 1. **CPPN output**: \[ h_k=\phi_k\left(\sum_{l \in \text{Pre}(k)} W_{lk}\cdot h_l + b_k\right) \] \[ \xi_j(\mathbf{X}_r)=h_{\text{out}} \] 2. **Normalized strength value**: \[ \bar{\xi}_{jr}=\frac{\xi_{jr}-\xi_{\min}}{\xi_{\max}-\xi_{\min}} \] 3. **Classification label**: \[ \text{Label}_j(x_r,y_r)= \begin{cases} 1, & \text{if }\xi_{jr}\geq T\\ - 1, & \text{if }\xi_{jr}<T \end{cases} \] 4. **Boundary point calculation**: \[ \mathbf{X}_b=\rho\cdot\mathbf{X}_p+(1 - \rho)\cdot\mathbf{X}_q,\quad\text{where }\rho=\frac{|\xi_{jp}-T|}{|\xi_{jp}-\xi_{jq}|} \] 5. **Fitness function**: \[ \mathbf{f}(D_j)=(E_j,\nu_j) \] 6. **Pareto front definition**: \[ P_F =\{\mathbf{f}(D_j)\mid D_j\text{ is retained by the reference - vector - guided selection strategy}\} \] 7. **Neuroevolution update formula**: \[ \theta_j^{t + 1}=\mathcal{G}(\theta_j^t,\mathcal{M},\mathcal{C}) \] Through this method, researchers can generate diverse metamaterial structures automatically, without gradients, without data, and in a highly scalable manner, and build a comprehensive metamaterial database.

Populating cellular metamaterials on the extrema of attainable elasticity through neuroevolution

Data-Driven Design of Thermal-Mechanical Multifunctional Metamaterials

Evolutionary Discrete Multi-Material Topology Optimization Using CNN-Based Simulation Without Labeled Training Data

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

Systematic topology optimization of elastic metamaterials for broadband bandgaps and customized mechanical properties

Hierarchical Combinatorial Design And Optimization Of Quasi-Periodic Metamaterial Structures

Large Permeability Metamaterial Design Optimization for Low Frequency Applications

Neural Metamaterial Networks for Nonlinear Material Design

Evolutionary topology optimization for mechanical metamaterials with auxetic property

Inverse Design of Mechanical Metamaterials with Target Nonlinear Response via a Neural Accelerated Evolution Strategy

Parameter Space Exploration of Cellular Mechanical Metamaterials Using Genetic Algorithms

An Efficient Robust Optimization Methodology for Metamaterial Unit Design

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

Computational Design of Flexible Planar Microstructures

Data-driven topology optimization of spinodoid metamaterials with seamlessly tunable anisotropy

Topology optimization of irregular multiscale structures with tunable responses using a virtual growth rule

Controllable and scalable gradient-driven optimization design for two-dimensional metamaterials based on Deep learning

A Deep Learning Approach for Reverse Design of Gradient Mechanical Metamaterials

Design of multifunctional metamaterials using optimization

Hierarchical combinatorial design and optimization of non-periodic metamaterial structures

Prospecting for Pluripotency in Metamaterial Design