Comparison of Modeling Soil Parameters Using Random Variables and Random Fields in Reliability Analysis of Tunnel Face

Abstract

Uncertainty is a notable feature of the geotechnical field. Reliability analysis methods provide a more rational solution than deterministic ones in the presence of the spatial variability of soil properties. Two types of reliability analysis methods were employed in the face stability analysis of pressurized tunnels in this study: the random variable method (RVM), which ignores spatial autocorrelation of soil parameters, and the random finite-difference method (RFDM), which considers spatial autocorrelation. In the RVM, the first-order second-moment (FOSM) method was adopted to calculate the reliability index with less computational effort. Then, the computational efficiency and accuracy of the two types of reliability analysis methods were systematically compared through a typical tunnel problem. Further sensitivity studies using the RFDM were performed to explore the effects of the scale of the fluctuation and autocorrelation functions. The authors found the probability of failure predicted by the RVM to be significantly overestimated because the spatial structure of soil parameters were ignored. Finally, using equivalent soil parameters instead of realistic ones, the corresponding results predicted by the equivalent FOSM method were found to be consistent with those of the RFDM, but they required far less computational effort. Thus, the equivalent FOSM method can be used to efficiently predict the stability of the tunnel face in variable soils, and it provides a practical tool for designers.