A Multivariate Adaptive Regression Splines model for determining horizontal wall deflection envelope for braced excavations in clays

摘要

The horizontal wall deflection of the braced excavations is influenced by many factors, including the excavation geometry, wall stiffness, soil parameters, struts spacing and so forth. In this study, three-dimensional (3D) finite element (FE) analyses were carried out to examine the corner restraining effects. A series of 3D FE analyses using the Hardening Soil (HS) model were carried out to investigate the effects of the strength of the clay, wall stiffness, excavation length, excavation depth and width, on the horizontal wall deflection envelope induced by braced-excavation in clays. Based on the results, a Multivariate Adaptive Regression Splines (MARS) model able to accurately learn the complicated implicit relationship between the maximum wall deflection envelope and these influential factors as well as the various interaction factors is developed. The developed MARS model is of good interpretability and enables the design engineer to estimate the shape of the wall deflection profiles. In addition, wall deflection profiles computed by this method compare favourably with a number of field and published records.