Deep learning-based topology design of periodic barrier for full-mode waves. (3rd January 2022)

• Record Type: Journal Article
• Authors: Liu, Chen-Xu; Yu, Gui-Lan
• Is Part Of: Construction & building materials. Volume 314:Part B(2022)
• Publication Date: 2022-01-03
• Subjects: Periodic wave barrier -- Deep learning -- Design -- Bandgap -- Vibration isolation; Building materials -- Periodicals; 624.18
• DOI: 10.1016/j.conbuildmat.2021.125579
• Languages: English
• ISSNs: 0950-0618

Abstract

Highlights: A novel deep learning model is proposed. The topology design of periodic barrier for full-mode waves is realized. The proposed model is suitable for most sites with different soil parameters. The "one-to-many" design is provided to meet some specified requirements. A periodic barrier tailored for a practical example isolates target vibrations. Abstract: A deep learning model is proposed to solve the design problem of periodic wave barrier with consideration of full-mode waves including in-plane mixed mode and out-of-plane shear mode, and the effect of site conditions on design is taken into account. The proposed model is composed of a variational autoencoder (VAE) and an autoencoder (AE) with two pretrained decoders. It can perform designs for different situations and give multiple structures for the same target within a very short time. Large number of targets in the testing set are considered, and the designed results highly meet the expectations. A periodic wave barrier is designed by the approach for a practical example, and the vibrations in main frequency range are attenuated greatly. The deep learning method makes the design of periodic wave barrier smart, efficient, accurate, and universal.