Developing a Digital Twin-based Framework for Construction Fire Hazard Recognition Training

Abstract

Construction sites are dynamic environments Where fire hazards pose significant safety risks. Effective hazard recognition is the key step to prevent such hazards. While existing research has developed various training methods, limited studies focus on fire hazards in construction. Moreover, most approaches fail to adapt to evolving site conditions. To address these gaps, this study proposes a digital twin (DT)based framework for dynamic fire hazard recognition training, with current validation progress presented. This framework integrates 360° imagery, 3D Gaussian Splatting, and Immersive Virtual Reality (IVR) to create an adaptive and interactive training environment. A user-centered training approach is designed to enhance personalized learning by incorporating individual trainee profiles and situation awareness (SA) assessments. Additionally, a cloud-based data system enables long-term tracking and scenario updates based on historical hazard data and trainee performance. This modular conceptual framework provides a foundation and guideline for future research on dynamic fire hazard recognition training. Further work will focus on framework validation through pilot studies in real construction settings to assess training effectiveness and usability.