A Statistical Moment Based Framework for Probabilistic Assessment of Tunnels with Correlated Input Variables

The accurate prediction of tunnel convergence is essential for ensuring tunnel stability, but it is challenging due to uncertain geotechnical properties. Probabilistic assessment offers a wide range of potential model responses, but complete knowledge of the density function of input parameters is mandatory. This study introduces a novel statistical moment-based framework for assessing the tunnel stability, which is capable to incorporate the correlation structure of the random variables. The proposed framework is applied to both analytical and numerical tunnel models. The accuracy of the proposed method has been demonstrated through a comprehensive comparison of the failure probability estimated through direct Monte Carlo Simulations. The point estimation-based version of the proposed method that is available in the literature underestimates the probability of failure for higher coefficient of variation of the random variables. The computation efficiency is also compared and discussed. The proposed method is much generalized whereas the point estimation-based version of the proposed method can be adopted for random variables with low coefficient of variations because of its exceptional computational efficiency.