Probabilistic analysis of responses of cantilever wall-supported excavations in sands considering vertical spatial variability

The influence of vertical spatial variability of sands on the excavation-induced lateral wall deflection and bending moment of excavations supported by cantilever retaining walls is investigated in this paper. Herein, the random finite element method (RFEM) is adopted to explicitly study the effect of one-dimensional spatial variability of internal friction angle of sands on the predicted wall and ground responses. The RFEM analysis consists of three components: (1) finite element method for analyzing lateral wall deflection and bending moment, (2) random field theory implemented with Monte Carlo simulation (MCS), and (3) statistical interpretation of MCS results through confidence intervals. This study reveals the importance of random field modeling in coping with the spatial variability of sands in the problem of supported excavations: (1) neglecting spatial variability of soil property will cause an overestimation of the variation in the predicted wall deflection and bending moment; (2) the estimated probability of failure based on a well-established serviceability limit state may be overestimated or underestimated depending on the chosen limiting lateral wall deflection. This study further investigates the effect of the number of MCS on the confidence intervals of the predicted statistics of the maximum lateral wall deflection and the maximum bending moment. The results also demonstrate that the confidence interval analysis of the predicted statistics of the maximum lateral wall deflection and the maximum bending moment provides a rational tool for interpreting the statistical data from RFEM.