Local buckling pressure prediction of constrained pipes by FEM and machine learning techniques

The external hydrostatic pressure exerted on the buried pipe can induce local buckling. A hybrid approach integrating finite element model (FEM) and machine learning algorithms is proposed to estimate the critical buckling pressure of confined pipes. A finite element model is established, and its reliability is validated against analytical, numerical and experimental methods. A comprehensive parametric discussion on the critical buckling pressure is accomplished, covering gap size, thickness-to-radius ratio, friction coefficient and strength of the medium (soils and/or rocks). The study uses eight machine learning algorithms to evaluate the predictive performance of each model based on evaluation metrics and the Friedman test system. The results indicate that the tree model represented by the XGBOOST model performs the best in terms of prediction accuracy.