A neural network approach to estimating deflection of diaphragm walls caused by excavation in clays

Abstract

An artificial neural network (ANN)-based approach for predicting deflection of diaphragm walls caused by braced excavation in soft to medium clays is presented in this study. Five input variables, including excavation depth, system stiffness, excavation width, shear strength normalized with vertical effective stress, and Young’s modulus normalized with vertical effective stress, are adopted as inputs to the ANN. The database for training and testing the ANN is generated from hypothetical cases using finite element method. The performance of the developed ANN reveals that the influence of each input variable on the wall deflection is consistent with the excavation behaviors generally observed in the field. The validation using 12 excavation case histories collected in this study shows that the wall deflection caused by braced excavation can be accurately predicted by the developed ANN.