Application of ANN and FELA for Predicting Pullout Capacity of Caissons in Anisotropic Clays

New solutions for stability problems of offshore foundations

Abstract

This work aims to study the pullout capacity of suction caissons in anisotropic clays under axisymmetric conditions. There are three dimensionless parameters considered in this study as input parameters including the ratio of depth to width of a caisson, adhesion factor, and the ratio of undraind shear strength obtained from compressive and tensile triaxials. The results of the dimensionless pullout capacity of suction caissons are carried out by using finite element limit analysis software, namely OptumG2. The results show that the dimensionless pullout capacity of a suction caisson increases when a dimensionless input parameter increases. In addition, three dimensionless input parameters influence the failure mechanism of suction caissons, where the size of failure significantly depends on an increase or a decrease in a dimensionless input parameter. In addition, this paper presents an Artificial Neural Network (ANN)-based approach for predicting new solutions for stability problems of offshore foundations. The data was also analyzed using an artificial neural network (ANN) model which achieved an accuracy of R2 = 99.94 % in predicting the pullout capacity according to failure criteria.