Application of artificial neural networks for predicting lateral and uplift capacity of buried rectangular box carrying pipelines

Abstract

Urban and offshore subterranean comprehensive pipelines are currently utilized frequently as infrastructure for fitting different engineering pipelines used for electricity, signal transmission, natural gas, petroleum, heating, water supply, and drainage systems. This study utilizes the finite element limit analysis (FELA) and artificial neural network (ANN) to evaluate the pull-out capacity factor of the buried rectangular box carrying the pipelines with internal inclined force in cohesive soil. In FELA, rigorous upper bound and lower bound solutions are performed to achieve the exact result. In this study, a dimensionless parametric analysis is carried out by considering the effect of five parameters viz. buried depth ratio, width-depth ratio, overburden factor, internal inclination angles, and interface factor on the pull-out capacity factor of a buried rectangular box, which can be called as a rectangular pipeline. Two significant samples of the investigated parameters are selected to evaluate the soil failure pattern (failure envelope) using the shear dissipation of soil failure for the buried rectangular pipeline subjected to inclined loading. Using the FELA results, the predictive equations are proposed using ANN, and then sensitivity analysis is performed to examine the sensitivity of each investigated essential parameter on the stability of the underground rectangular pipeline. The predictive ANN model for the buried rectangular pipe subjected to inclined loading is presented as design charts for practice and further investigation.