A Bottom-Up Design Framework for Multifunctional Lattice Metamaterials.

Area of Science:

• Materials Science and Engineering
• Artificial Intelligence in Design
• Metamaterials

Background:

• Increasing demand for lightweight, multifunctional lattice metamaterials.
• Limitations of traditional inverse design methods (e.g., topology optimization) in exploring design space.
• Need for advanced methods to achieve complex structural designs for enhanced properties.

Purpose of the Study:

• To introduce a novel generative AI framework for inverse design of lattice metamaterials.
• To enable the design of shell lattice structures with optimized energy absorption and broadband sound absorption.
• To overcome limitations of traditional methods by enhancing design freedom and structural complexity.

Main Methods:

• A generative AI framework combining 3D Gaussian voxel generation and deep learning.
• Hybrid architecture: 3D convolutional neural network (CNN) and conditional deep convolutional generative adversarial network (cDCGAN) for energy absorption prediction and generation.
• Genetic algorithm for tuning heterogeneous geometries for broadband sound absorption.

Main Results:

• Designed shell lattice structures demonstrate superior multifunctionality.
• Achieved 40%-200% greater energy absorption compared to conventional shell lattices.
• Exhibited high average sound absorption coefficient (∼0.7) across a broad bandwidth (1000-5800 Hz).

Conclusions:

• The proposed generative AI framework overcomes drawbacks of existing inverse design approaches.
• Voxel-level model generation informed by physical insights enhances design capabilities.
• Demonstrated effectiveness for creating advanced multifunctional lattice metamaterials.