Automated BIM-integrated cyber-physical system for real-time near-miss detection and proactive safety management on construction sites

Introduction : Automated near-miss detection in construction has remained technically constrained by two unresolved problems: the inability of conventional sensor systems to maintain localization accuracy as the physical propagation environment changes during active construction, and the absence of a scalable mechanism to extract hazard zone boundaries directly from building models without manual site surveying. Method : This study developed and field-validated a Cyber-Physical System that addresses both limitations by integrating Wi-Fi fingerprinting with BIM-based hazard zone identification on an active multi-story construction site. Worker positions were estimated in real time using an improved k-nearest neighbor algorithm operating on a radio map generated automatically from BIM geometry. A dedicated hardware calibration rig continuously recomputed the path-loss exponent to sustain fingerprinting accuracy as site conditions evolved. Results : When estimated worker coordinates intersected BIM-extracted unsafe zone boundaries, the system detected a near-miss and dispatched severity-classified SMS alerts based on worker PPE and training records. Field validation produced precision of 0.95 and recall of 0.96 under static conditions, and 0.93 precision and 0.95 recall under dynamic movement. Online calibration reduced false negatives by 68% over a three-day trial compared to a static radio map. Conclusions and Practical Applications : The system provides a replicable, infrastructure-light framework for automated near-miss monitoring and generates the continuous real-time data stream necessary for Digital Twin-enabled predictive safety management.