Traditional inspection and diagnosis methods for civil infrastructure operation and maintenance (CI O&M) rely heavily on human efforts. Such efforts are always affected by subjective judgment and human errors due to engineering knowledge and prior experiences of field engineers. On the other hand, recent development of AI-driven tools could achieve effective information acquisition but lacks interpretability and engineering credibility. How to integrate human knowledge with AI capacity for safe and effective CI O&M is thus necessary in this new era. This paper presents a human-in-the-loop digital twin (HITL-DT) framework that enables safety risk sensing, prediction and control for smart CI O&M. The proposed framework fuses human cognition (i.e., individual perception and team situation awareness), AI and engineering knowledge for 1) risk sensing and diagnosis based on spatiotemporal changes and 2) risk prediction and control for smart CI O&M. Qualitative analysis indicates that the HITL-DT approach produces more explainable, trustworthy, and actionable diagnostic outputs, which enhance the reliability and proactivity of CI O&M.
Sun et al. (Thu,) studied this question.