Designing aperiodic metamaterials using mechanical neural networks.

Area of Science:

• Metamaterials Science
• Mechanical Engineering
• Artificial Intelligence

Background:

• Designing aperiodic metamaterials for multiple behaviors is challenging due to complex interdependencies.
• Existing methods struggle to account for all real-world factors in metamaterial design.

Purpose of the Study:

• To develop a novel design approach for aperiodic metamaterials using mechanical neural networks (MNNs).
• To enable the accurate achievement of desired multiple behaviors in metamaterials by considering all relevant factors.
• To demonstrate the MNN approach by designing two distinct shape-morphing aperiodic metamaterials.

Main Methods:

• Utilized mechanical neural networks (MNNs) to determine optimal arrangements of passive beam designs within aperiodic metamaterials.
• MNNs learned by tuning the stiffness of mechatronically controlled active beams to match passive beam states.
• Iterative physical learning process continued until desired material behaviors were achieved.

Main Results:

• Successfully designed two aperiodic metamaterials with distinct shape-morphing capabilities using the MNN approach.
• Demonstrated the MNN's ability to navigate complex design spaces and achieve specific functional behaviors.
• Validated the effectiveness of the MNN in considering all real-world factors for metamaterial design.

Conclusions:

• Mechanical neural networks offer a powerful and efficient method for designing complex aperiodic metamaterials.
• The MNN approach overcomes limitations of traditional methods by integrating physical learning and comprehensive factor consideration.
• This work paves the way for advanced metamaterials with tailored multi-functional shape-morphing properties.