A robust method for bridge safety risk assessment using improved multi-state fuzzy Bayesian network

This paper proposes a robust method utilizing Multi-state Fuzzy Bayesian Network (MFBN) to evaluate bridge safety risks, offering a foundation for risk control. Initially, a bridge collapse fault tree and a directed acyclic graph are constructed to analyze causal relationships. Subsequently, bridge nodes are categorized into three states, enhancing traditional binary states. Expert judgment ability and subjective reliability levels are considered to ensure survey data reliability, using confidence indices to establish multi-state fuzzy conditional probability tables. Lastly, the improved similarity aggregation method incorporates age as a factor to aggregate expert opinions, determining safety risk probabilities of root nodes. This approach assesses bridge safety risk levels with prior knowledge and evidence, identifying critical nodes contributing to heightened risks, aiding in the formulation of risk management strategies. Application to two urban bridges demonstrates the method's effectiveness and robustness, positioning it as a valuable decision-making tool for bridge safety risk management.