Bayesian Machine Learning in Metamaterial Design: Fragile Becomes Supercompressible.

Area of Science:

• Materials Science
• Computational Materials Design
• Mechanical Engineering

Background:

• Traditional materials design relies on trial-and-error, limiting exploration of novel material properties.
• Future materials require adaptability, multi-functionality, and tunability beyond current capabilities.
• Computational approaches are essential for efficient exploration of complex material solution spaces.

Purpose of the Study:

• To develop a data-driven computational framework for designing novel metamaterials.
• To adapt metamaterial concepts for diverse properties, base materials, scales, and manufacturing.
• To demonstrate the fabrication and characterization of new supercompressible metamaterials.

Main Methods:

• Utilized a Bayesian machine learning approach for guided metamaterial design.
• Employed a computational, data-driven strategy to explore the material solution space.
• Fabricated and tested metamaterial designs at both macro and micro length scales.

Main Results:

• Successfully transformed brittle polymers into lightweight, recoverable, supercompressible metamaterials.
• Macroscale design achieved >94% strain and ~0.1 kPa recoverable strength.
• Microscale design achieved ~80% strain and >100 kPa recoverable strength.

Conclusions:

• A data-driven computational approach enables the design of advanced metamaterials with tailored properties.
• The developed framework facilitates the creation of adaptive, multipurpose, and tunable materials.
• The open-source code supports future research in metamaterial design and analysis.