Research Information System
Expert Systems with Applications, vol.319, 2026 (SCI-Expanded, Scopus)
Digital Twin Project (DTP) success is key to improving the resilience of major water conservancy engineering. However, DTPs face complex risks. Existing research lacks a systematic risk system for DTPs. Current risk assessment and response models use one-time decision information, failing to update in a timely manner due to incomplete expert assessments. To address these gaps, an online learning-driven risk assessment and response decision model is constructed. First, a comprehensive risk indicator system of DTPs is constructed based on a literature review. Second, the online learning algorithm, Full Consistency Method (FUCOM) and coupling degree model are integrated to quantify and update risk loss considering coupling relation between risks. Third, an online learning-based risk response decision model is proposed to find optimal response measures. Finally, a numerical case is provided to validate the proposed model and the sensitivity and comparative analyses are conducted. The results indicate that the proposed model can effectively update risk assessment using real-world information and dynamically select optimal measures, significantly reducing expected risk losses. Besides, the proposed model exhibits particularly prominent robustness advantages in scenarios with high-intensity risk coupling networks and high exploration preferences. Finally, measure cancel cost impacts response decision results. This study can provide theoretical and practical guidance for risk governance of DTPs in major water conservancy engineering.