Reliability Analysis of Foundation Pit Directly Based on Field Information

There are two problems frequently encountered in the risk analysis of foundation pit through the probability based reliability method. One is the large amount of calculation and the other is insufficient statistic field information. To solve these problems, uniform experiment design (UD), non-parametric regression (NPR) and Bootstrap simulation (BS) techniques are introduced based on the concept of response surface (RS). UD and NPR are used to construct more efficient RS, based on which, the failure probability is calculated through field information and BS. Combined with a practical project, the probability distributions calculated by BS based on field observation datum, and those calculated by Monte Carlo simulation (MCS) based on assumed probability distribution are compared. The results show that the probability distributions obtained through BS and MCS are the same in most cases, but BS is more suitable for the case of small sample datum. The RS methods based on UD and NPR can provide higher accuracy with less computation while containing more uncertain parameters.