Abstract: Based on the concise analysis of the influence of support parameters to stability and deformation of surrounding rock mass in soft rock tunnel, Bayesian network and corresponding uncertainty reasoning principle has been introduced to develop an expert system for selection of support parameters of soft rock tunnel with large span. By combining prior knowledge of domain experts with work-site data recorder, the paper gets the posterior probability density of most nodes. The expert system is applied to compare several schemes about the support parameters of one tunnel. The field practices prove that the expert system has good applicability. The Bayesian network can reason the incomplete input information by using the prior knowledge of domain experts and work-site data recorder. The Bayesian network based expert system can not only be applied to making comparison and choice of the support parameters of soft rock tunnel, but also be applied to diagnosis of the collapse accident of soft rock tunnel.

Key words: Bayesian network, tunnel, surrounding rock instability, risk early-warning, expert system

摘要： 为了对隧道围岩失稳风险做出准确判断，预防围岩失稳灾害的发生，将贝叶斯网络方法应用于隧道围岩失稳的风险预警。在系统分析隧道支护参数影响围岩稳定性的基础上，通过引入基于贝叶斯网的知识表达和相应的不确定性推理原理，构造了隧道围岩失稳风险预警专家系统；并通过工程案例验证了隧道围岩失稳风险预警专家系统好的适用性。贝叶斯网专家系统可以充分利用专家的先验知识和已有案例提供的或概率分布，可以使推理在输入数据不完备的基础上进行，能够有效地实现围岩失稳风险预警。

关键词: 贝叶斯网络, 隧道, 围岩失稳, 风险预警, 专家系统