AI 기반 산업 기계 끼임 사고 위험도 평가 시스템

초록

This study analyzes the structure and applicability of SafetyLens AI, an AI-based real-time monitoring system designed to prevent entrapment and crush accidents in manufacturing domain environments. Manufacturing accounts for 23.3% of all fatal industrial accidents, of which 12.7% are attributed to entrapment incidents. This highligh the severity of the issue and structural limitations of conventional single-camera and static sensor based – safety systems. The proposed system employs multiple 2D cameras and the YOLOv8 algorithm to detect workers’ hands and calculate the overlap ratio (OR) with designated risk zones, thereby enabling real-time risk assessment. When the OR exceeds a predefined threshold, warning alarms are activated and the equipment is automatically halted to prevent accidents. The calculation method for OR and specification of the risk evaluation criteria have further scope for improvement.

키워드