Classification and Regression Trees Approach for Predicting Current-Induced Scour Depth Under Pipelines

Abstract

Reliable prediction of scour depth is important in engineering analysis concerned with pipeline stability. The aim of this study is to develop an accurate formula for prediction of the current-induced scour depth under pipelines. Previous experimental data are collected and used as a database by which to study the effect of different parameters on the scour depth. Decision tree and nonlinear regression approaches are used to develop engineering design formulae for estimation of the current induced scour depth in both live bed and clear water conditions. It is demonstrated that the proposed formulas are more accurate than previous ones in predicting the scour depth in all conditions. Probabilistic formulas are also presented for different levels of risk, aimed at safe and economic design of submerged pipelines.