对长距离输水隧洞盾构施工可能的风险路径进行预测分析，对保障隧洞工程安全施工具有重要作用。基于 大量盾构施工风险事故统计分析，提取隧洞工程盾构施工风险致因指标，利用解释结构模型对盾构施工风险因素 进行因果层次关系分析和网络拓扑图构建；建立基于贝叶斯网络的盾构施工风险事件预测模型，运用贝叶斯网络 的反向诊断推理技术计算风险事件发生的最大可能路径，确定导致盾构施工风险事件的关键因素；将所得的风险 预测模型应用到西霞院灌区工程穿沁隧洞盾构施工项目中。结果表明，模型运算得到的风险事故路径与工程现 阶段实际施工情况一致，验证模型的可靠性与适用性，并根据风险事故路径预测结果为穿沁隧洞后续施工的风险 隐患排查治理提供指导意见。

In order to solve the problems of uneven distribution of water resources, mismatched water and soil resources, and serious shortage of water resources, China has launched many long-distance water diversion projects during recent years. With the implementation of these projects, a large number of headrace tunnels has also been constructed, which is continuously developing and improving the construction methods, and now the shield method has gradually become the main construction method of tunnels. As the shield tunneling is a complex system process influenced by a variety of risk factors, construction risk accidents occur from time to time, resulting in delays of the construction period as well as loss of life and property. Therefore, predicting and analyzing the possible risk path of the shield tunneling of long-distance water conveyance tunnel plays an important role in ensuring the safe construction of the tunnel project.At present, the risk analysis of tunnel shield construction schedule have several limitations. First, the correlation between the risk factors of shield tunneling of headrace tunnel is not considered; Secondly, the existing research results are mostly based on subway and traffic tunnel, and the differences in construction technology and risk management of long-distance headrace tunnel of hydraulic engineering are not considered. This paper takes the shield tunneling of long-distance water conveyance tunnel in the water diversion project of hydraulic engineering as the research object. Statistical analysis are made on a large number of collected data of water diversion tunnel construction cases, and the risk causing indicators and possible risk events are extracted of tunnel engineering shield tunneling. By using interpretative structural model, the causal hierarchy analysis and network topology diagram construction are conducted. Then a prediction model of shield construction risk events based on Bayesian network is established, and the reliability of the model is preliminarily verified through Bayesian network learning.Finally, the risk event prediction model obtained has been applied to the shield construction project of Xixiayuan Chuanqin Tunnel. The reverse diagnosis reasoning technology of Bayesian network is used to calculate the maximum possible path of risk events and determine the key factors leading to risk events in shield construction. The results show that the risk accident path calculated by the model is consistent with the actual construction situation at the current stage of the project. The reliability and applicability of the model are verified, and according to the prediction results of the risk accident path, the guidance is provided for the troubleshooting and governance of the risk hidden dangers in the subsequent construction of Chuanqin Tunnel.Conclusion: (1) The risk path prediction model of tunnel shield construction is established. Through the interpretation of the structural model, the correlation between various risk events that affect the tunnel shield construction progress is determined, and then the Bayesian network model is established, forming an effective method to predict the development trend of the tunnel shield construction risk events through the tunnel shield construction risk factors. (2) The reliability and applicability of the model are verified. Through the application of the shield construction project of Chuanqin Tunnel Project, the predicted result of risk causal chain is consistent with the actual construction situation, and the model can provide real-time decision support for managers. (3) This model can be directly applied to shield tunneling of headrace tunnels with similar construction conditions; for different situations, such as special geological conditions and different shield machines, some parameters of the model need to be appropriately changed, but the model and calculation process established in this paper are applicable.