Data-driven inverse design of multifunctional bicontinuous multiscale structures.

Area of Science:

• Materials Science
• Computational Modeling
• Additive Manufacturing

Background:

• Bicontinuous multiscale structures possess unique interconnected networks offering superior performance.
• Designing these structures is challenging due to a lack of analytical models and computational complexity in optimization.
• Existing methods often require post-processing to ensure structural integrity and connectivity.

Purpose of the Study:

• To present novel 3D Large-range, Boundary-identical, Bicontinuous, and Open-cell Microstructure (L-BOM) datasets.
• To enable fast, data-driven inverse design of multifunctional bicontinuous structures.
• To overcome limitations in current design methodologies for complex microstructures.

Main Methods:

• Development of generative artificial intelligence models.
• Application of active learning techniques for dataset generation.
• Creation of datasets with identical boundaries and broad property coverage.

Main Results:

• Generated L-BOM datasets featuring bicontinuous open-cell architectures with guaranteed connectivity.
• Demonstrated broad property coverage for extensive performance exploration.
• Successfully utilized datasets for inverse design without requiring post-processing for connectivity.

Conclusions:

• The L-BOM datasets significantly advance the design capabilities for bicontinuous multiscale structures.
• These datasets facilitate exploration of a large property space for multifunctional materials.
• Potential applications include the design of bone implants, chair components, and materials with tunable elasticity and permeability.