Multi-objective design of multi-material truss lattices utilizing graph neural networks

Area of Science:

• Materials Science and Engineering
• Computational Materials Design
• Additive Manufacturing

Background:

• Additive manufacturing (AM) enables architected materials with tailored properties.
• Multi-material AM expands design by combining distinct materials.
• Machine learning for lattice inverse design is limited to single-material systems.

Purpose of the Study:

• To develop a novel graph neural network (GNN) framework for the inverse design of multi-material truss lattices.
• To incorporate material properties as edge features in graph representations for GNNs.
• To enable fast and efficient inverse design for tunable thermal expansion and stiffness.

Main Methods:

• Utilized graph neural networks (GNNs) with material properties as edge features.
• Developed a graph representation for multi-material truss lattices.
• Validated the framework for single and multi-objective optimization tasks.

Main Results:

• Successfully designed multi-material lattices with tunable thermal expansion and stiffness.
• Demonstrated the framework's ability to explore a broad design space.
• Showcased GNNs' superior capacity in capturing multi-material structure-property relationships.

Conclusions:

• The proposed GNN framework offers a fast and efficient approach for multi-material lattice inverse design.
• This method significantly advances the design capabilities for architected materials in additive manufacturing.
• Continued GNN advancements will be crucial for realizing the full potential of multi-material truss lattices.