Coupling FELA and computation intelligent in predicting behaviour of inclination strip anchor in anisotropic and heterogeneous clays

Abstract

This study introduces new numerical investigations of the uplift capacity of inclined strip anchors buried in anisotropic and heterogeneous clays. The analyses utilize Lower Bound (LB) Finite Element Limit Analysis techniques (FELA) under plane strain conditions, employing the Anisotropic Undrained Shear (AUS) failure criterion. The undrained shear strength of the clay increases proportionally with depth, and anisotropy is introduced by adjusting the ratios between the shear strengths obtained from three different tests. Four input parameters, namely, the embedment ratio (H/B), the inclination angle (α), the ratio of increasing undrained shear strength with depth (m = ρΒ/suTC0), and the factor of anisotropic parameters (re), are considered in analyzing the uplift coefficient (Fc), and a set of design charts is proposed. The effect of dimensionless parameters on the development of failure mechanisms is also investigated. Moreover, a novel method for assessing the ultimate uplift capacity of an inclined strip anchor in natural clay is given by integrating FELA with Multivariate Adaptive Regression Splines (MARS) and Extreme Gradient Boosting (XGBoost). Based on the optimal machine-learning models, this study also calculates the contribution of each input to the value of Fc via sensitivity analysis. The engineering computations for strip anchors embedded in anisotropic and non-homogenous clays can be improved with the results of these investigations and the proposed machine-learning models.