Dispersion relation prediction and structure inverse design of elastic metamaterials via deep learning. (January 2022)

• Record Type: Journal Article
• Authors: Jiang, Weifeng; Zhu, Yangyang; Yin, Guofu; Lu, Houhong; Xie, Luofeng; Yin, Ming
• Is Part Of: Materials today physics. Volume 22(2022)
• Publication Date: 2022-01
• Subjects: Elastic metamaterials -- Dispersion relations -- Deep learning -- Inverse design; Materials science -- Periodicals; Physics -- Periodicals; Electronic journals; 530.41
• DOI: 10.1016/j.mtphys.2022.100616
• Languages: English
• ISSNs: 2542-5293

Abstract

Abstract: Well-designed metamaterial structures give rise to unprecedented properties that promise a diverse range of specific applications. Traditional methods typically rely on iterative searching in a vast design space aided by researchers' experience and optimization algorithms to obtain a structure with desired properties. Here, we establish a mapping between the structural topology and the dispersion relation of elastic metamaterials using deep learning approaches. Our results show that the proposed model enables accurate prediction of the dispersion relation for a given structure and the inverse design of near-optimal structures based on the target dispersion relation. Moreover, for the inverse design process, the input dispersion relations can be proactively tailored. Our deep-learning-based approaches have shown capability in accelerating the design and optimization process, paving the way to pursue new breakthroughs in metamaterials research. Graphical abstract: Image 1 Highlights: A mapping between the structural topology and dispersion relation of elastic metamaterials is established by the DL method. CNNs achieve accurate prediction of the dispersion relation for a given structural configuration. The cGANs enable the inverse design of near-optimal structures based on the target dispersion relations.