Dec 6, 2024
9:00am - 9:15am
Hynes, Level 2, Room 203
Jeonghoon Park1,Jaebum Noh1,Jehyeon Shin1,Grace Gu2,Junsuk Rho1
Pohang University of Science and Technology1,University of California, Berkeley2
Jeonghoon Park1,Jaebum Noh1,Jehyeon Shin1,Grace Gu2,Junsuk Rho1
Pohang University of Science and Technology1,University of California, Berkeley2
Recent advancements in 3D printing have facilitated the fabrication of complex 3D mechanical structures, including metamaterials with unique properties such as negative effective bulk modulus and band gaps. While chiral mechanical metamaterials have demonstrated complete band gaps to block elastic waves of all polarizations, previous research often overlooked structural stability. Designing stable metamaterials requires simultaneous consideration of both dynamic properties (e.g., bandgaps) and static properties (e.g., stress distribution), presenting challenges for traditional heuristic-based design approaches.<br/>We propose a novel artificial intelligence-driven method for the inverse design of 3D chiral mechanical metamaterials with desired static and dynamic characteristics. Our approach utilizes a generative neural network pipeline comprising a conditional GAN, an encoder, and a regressor module. This framework allows for the simultaneous optimization of dynamic and static properties, resulting in mechanical metamaterial designs that enhance stability while preserving desired physical characteristics. The conditional GAN generates design parameters based on input band gap frequencies, while the encoder transforms these parameters into embeddings, facilitating easier training. The regressor module addresses the non-uniqueness issue in inverse design. Through numerical simulations, we confirm that the produced structures exhibit the intended bandgap frequencies and low stress distributions. Experiments on the structures fabricated through 3D printing excellently corroborate the numerical analysis results.<br/>Our research demonstrates the potential of AI-driven inverse design in developing complex 3D chiral mechanical metamaterials with optimized static and dynamic properties, overcoming limitations of conventional design methods. This approach opens new possibilities for creating 3D-printed mechanical metamaterials with tailored properties for various applications in vibration control, acoustics, and structural engineering.