Safety control of façade system installation using hybrid artificial intelligence method

Abstract

Fall from height is often the main cause of injury and death in construction site. Although workers are aware of the dangers related to not wearing safety harnesses, many people forget or intentionally do not wear them when working at heights. Personal Protective Equipment (PPE) that complies with safety regulations is widely used to ensure worker safety. Training is considered effective in reducing risk-taking behaviors and improving the working practices of construction workers. However, direct supervision for occupational safety training still has limitations. This study utilized the new YOLOv8 algorithm "You only look once" (YOLO), which includes 5 variations of it including YOLOv8n, YOLOv8s, YOLOv8m, YOLOv8l, YOLOv8x, to ensure safety during the installation and construction of the facade system. A dataset consisting of 10043 images was searched and collected to establish a digital safety monitoring system through training and testing phases. The YOLOv8 algorithm has an average detection speed of up to 136 frames per second, meeting the requirement for real-time object detection. This study provides an optimal solution when the model will be stored on a cloud server and automatically notifies the manager.

Key words: Safety; artificial intelligence; facade system.