Session: 17-01-01: Research Posters

Paper Number: 148011

148011 - Neural Network Assisted Broadband Perfect Absorbing Meta-Layer Design for Wave and Vibration Attenuation 

Metamaterials and metasurfaces exhibit remarkable capabilities in controlling waves across optics, acoustics, and elastic waves. While recent attention on perfect absorber design underscores its significant ability to fully absorb incoming waves, challenges persist in achieving broadband functionality and accommodating large incident angles. Moreover, designing perfect absorbers for 1D or 2D structures often relies on intuition and bottom-up methods. This can limit system performance and add unnecessary weight to the overall system, resulting in a bulky design. To overcome those drawbacks, a top-down approach based on inverse design, such as topology optimization method, has been adopted to design broadband perfect absorbers. Meanwhile, machine learning methods, such as supervised learning, unsupervised learning, and deep learning, have been recently investigated in various fields through data-driven approaches for image or natural language processing. Neural network, as a subset of machine learning technique, inspired by the structure and function of the human brain, is particularly well-suited for capturing complex patterns and relationships in data. Implementing machine learning and neural networks in topology optimization has been shown to accelerate the design process and improve convergence rates. However, these methods have primarily been applied to static problems. In this study, we introduce a novel approach—employing neural network assisted topological optimization— in a dynamic problem to design a lightweight perfect absorbing meta-layer for flexural waves. The designed meta-layer is capable of handling large incident angles at broad bandwidths for perfect flexural wave absorption. In the design, we first implemented the neural network assisted topological optimization for the meta-layer on a 1D beam structure. The optimized structure, after removing the gray region, was validated using numerical simulation in COMSOL Multiphysics, and broadband absorption was achieved. Experiment was further carried out to characterize the absorption performance. The meta-layer was fabricated through advanced 3D printing techniques. The absorption performance of the fabricated sample was characterized using a laser doppler vibrometer in the experiment. The experimental results were compared to the numerical simulations, and the results matched well. Next, we extended the 1D design to 2D structures for broadband perfect absorption at large incident angles. The design was also validated through numerical simulation and experiment. As an application, we demonstrated broadband vibration attenuation using the designed meta-layer on both 1D and 2D structures. We compared its performance to traditional damping materials, highlighting its superior effectiveness in noise and vibration attenuation. Our proposed approach opens new avenues for the design of metamaterials and metasurfaces.

Presenting Author: Xiaopeng Li Toyota Research Institute of North America

Presenting Author Biography: Dr. Xiaopeng Li earned his Ph.D. in Mechanical and Aerospace Engineering and a master’s degree in Electrical and Computer Engineering in May 2020 from the University of Missouri. He joined Toyota Research Institute of North America as a research scientist in June 2020. His research expertise includes the development of active and passive acoustic and elastic metamaterials/metasurfaces and machine learning, with applications ranging from wave and vibration control to advanced engineering solutions.

Authors:

Xiaopeng Li Toyota Research Institute of North America