为了分析地铁隧道健康监测中各个指标之间的相互演化作用关系，提出了一种对监测数据进行分析的数据挖掘思路，将傅里叶幅值灵敏度检验扩展法（EFAST）应用于各项指标的全局敏感性分析中。依据现行隧道安全评价规范，建立反映隧道安全状态的指标体系；通过基于粒子群优化的最小二乘法支持向量机（PSO-LSSVM），模拟构建了各安全指标函数拟合模型，用以描述隧道各安全指标之间相互作用关系；在此基础上，采用EFAST对该模型的指标进行了全局敏感性分析。结果发现，对隧道安全性影响较大的敏感因素有管片接缝张开宽度、管片接缝接触应力和隧道沉降值；另外，还分析了不同的目标参数、不同的参数取值范围、不同的输入参数的分布形式对参数敏感度值的影响，从不同的角度揭示了盾构地铁隧道中参数的相互作用机理的演化规律。

A new index called global sensitivity index by global sensitivity analysis was proposed to evaluate the relationship between tunnel structure health monitoring data and tunnel health status. Global sensitivity analysis could quantitatively and systematically evaluate the parameter sensitivities and their variations under various conditions, resulting in an improved understanding of relationships between monitoring data and tunnel health status, and suggesting potential approaches to the improvement of tunnel health evaluation model. Moreover, a case study of global sensitivity analysis was presented to evaluate the monitoring data on Wuhan Yangtze River Tunnel. The achievement of this global sensitivity analysis leads to the parameter sensitivities had been quantified, and the sensitive and insensitive parameters were distinguished. The sensitive parameters could then be identified as major factors for monitoring or administrative maintenance. The variations in parameter sensitivity index associated with various target functions, parameter ranges and distribution were examined. The variations of parameter sensitivities were observed with various conditions, suggesting that sensitive and insensitive parameters maybe inversed under different conditions.

国家自然科学基金（51378235、71571078）；湖北省自然科学基金（2014CFA117）