Application of Finite element method and Machine Learning in analyzing the stability of slope supported by one row of vertical piles

Abstract

This study proposes a new approach for analyzing the stability of a slope supported by one row of vertical piles, using finite element method (FEM) combined with machine learning algorithm. Initially in the proposed approach, the FEM model was built with commercial software Plaxis, and the factor of safety (FS) was obtained by shear strength reduction method. The values of FS were calculated in different cases of pile locations, pile spacing, cohesion and internal friction angle of soil. The results from FEM models were subsequently used to train the Artificial neural network (ANN). The ANN model with explicit formulas to obtain FS helps to speed up the calculation process and facilitate the optimization in slope design. Besides, the machine learning model evaluates the importance of the input features (pile location, pile spacing, cohesion and internal friction angle of soil, slope angle) to the FS of the slope reinforced by one row of vertical piles. Results show that soil shear strength and slope angle have greater impact on FS than pile locations and spacing do.