A Multistate Bayesian Network-Based Approach for Risk Analysis of Tunnel Collapse

Collapse is a typical disaster during tunnel construction and may cause tremendous loss; therefore, risk evaluation can help minimize such damage by taking preventive measures. This paper proposes a method to analyze the risk of tunnel collapse based on multistate fuzzy Bayesian network (MFBN). The method screens the risk factors responsible for tunnel collapse by means of the fault tree analysis and establishes the corresponding Bayesian network model. Meanwhile, the triangular fuzzy number is utilized to describe the possibility of node failure. The fuzzy failure probability under each failure state is acquired through expertise. When integrating experts’ opinions, an improved similarity aggregation method is proposed; this method comprehensively gauges their judgment ability and subjective recognition degree and alleviates the influence of excessive differences in opinions. Furthermore, a multistate fuzzy conditional probability table is established by combining probability interval division and expert knowledge to describe the intensity–dependence relationship among nodes. After defuzzification, the collapse probability and critical risk factors can be determined through MFBN-based inference. In addition, the method allows for the dynamic analysis of risks in tunnel construction. Applying this method to Yanglin Tunnel, the results demonstrate the feasibility and application potential of this method, and it can provide important supporting information for risk prevention and control during construction.