Journal of Safety and Environment

Addressing the conflict between subjective and objective weights in the risk assessment of underground heating pipe galleries， as well as the lack of dynamics in traditional models， this study proposes a dynamic assessment method that integrates game theory-based combined weighting with residual-driven feedback correction. This integrated framework explicitly acknowledges the evolving nature of infrastructure risks by establishing a closed-loop， self-optimizing evaluation system. First， initial subjective and objective weight sets are constructed using three methods: the G1 method (Order Relationship Analysis Method) determines subjective weights based on expert insights to identify implicit risks such as weld defects; the improved entropy weight method captures data discreteness; and the CRITIC (Criteria Importance Through Intercriteria Correlation) method integrates data variability and inter-indicator correlation to form complementary objective weights， mitigating the limitations of single-method approaches. Subsequently， game theory combined weighting achieves collaborative optimization by minimizing the Euclidean distance between the combined weight vector and the three initial sets， resulting in the Nash equilibrium between subjective experience and objective data. To enhance dynamics， a residual-driven multi-stage feedback correction mechanism is designed: each phase calculates the relative residual between predicted and actual risk levels， which dynamically adjusts the subjective weight coefficient via an exponential decay function and updates cloud model parameters (E_x， E_n， H_e). This continuous adaptation mechanism ensures that the assessment model remains synchronized with the real-time degradation process and environmental variations of the pipe gallery system throughout its operational lifecycle. Using a case study of a 12.3 km DN1200 heating pipe gallery in Hebei Province， results indicate that Section 2 in the old urban silty clay layer is the highest-risk priority， with a risk score of 52.5. Operation and maintenance indicators exhibit the highest sensitivity， contributing 35% to total risk. The subjective weight coefficient decays from 0.5 to 0.001， thereby reducing reliance on expert experience. The comprehensive score achieved by the proposed method (0.975 9) outperforms the G1 method (0.874 3)， improved entropy weight method (0.703 1)， and CRITIC method (0.892 4) by 12.3%， 8.7%， and 10.1%， respectively. This method effectively identifies high-risk sections and dominant factors， providing scientific support for full-life-cycle safety management of underground heating pipe galleries.