Digital Twin Enabled Collision Early Warning for Smart Site Safety System (SSSS)

• Published in: 2024 IEEE 20th International Conference on Automation Science and Engineering (CASE) pp. 3023 - 3028
• Main Authors: Zhao, Shuxuan, Yu, Chenglin, Qu, Xinye, Zhu, Zhengxu, Lei, JunJie, Zhong, Ray Y.
• Format: Conference Proceeding
• Language: English
• Published: IEEE 28.08.2024
• Subjects: Accidents; Activity recognition; Alarm systems; Collision warning; Construction; Construction industry; Deep learning; Detection algorithms; Digital twin; Digital twins; Monitoring; Safety; Synchronization
• ISSN: 2161-8089
• DOI: 10.1109/CASE59546.2024.10711385

Construction industry is one of the most dangerous industry sectors as its high accident rates and fatality rates. Being struck by working construction machines is one major reason for accidents in construction sites. In this research, a digital twin-enabled collision warning system is designed to avoid collision and improve safety in construction industry. The system is composed of physical entity layer, virtual entity layer, digital data layer, service layer and connection. The digital data layer collects and analyzes virtual information collected from the physical layer. The virtual entity layer is responsible for constructing digital models with the same constraints, behaviors, and rules as the physical models. The service layer can provide on-site services such as on-site detection, collision warning, and motion status monitoring for site managers and workers. Besides, deep learning-based algorithms are also designed to achieve the classification, location, and key points detection of construction machines. The experiment results show that the proposed deep learning detection algorithms can achieve more than 90% detection performance, and the proposed framework can provide precise on-site collision early warning.