Application of ANN and FELA for Predicting Bearing Capacity of Shell Foundations on Sand

In this study, the artificial neural network (ANN) model has been employed to create a data-driven prediction for the bearing capacity factor of shell foundations under both axisymmetric and plane strain conditions. To develop the set of training data in ANN, the lower and upper bound finite element limit analysis (FELA) has been carried out to obtain the numerical results of the bearing capacity factor of shell foundations. The N γ values were obtained from FELA and used as a training data set. The influences of all dimensionless variables such as the shape of the shell and the soil parameters on the bearing capacity factors and the failure mechanisms were investigated and discussed in detail. The sensitivity analysis of these dimensionless variables was also examined. To propose a design equation for predicting the values of N γ for both strip and axisymmetric shell foundations with rough and smooth interfaces, the ANN model was developed. The finding in this paper can be used by practicing engineers for the design of shell foundations under both axisymmetric and plane strain conditions.