CFD and machine learning approach based-predictive modeling of scouring below submarine pipeline under wave and current condition

Abstract

The major free spans of submarine pipelines, caused by the combined action of waves and currents can pose significant risks to pipeline operational safety. In this study, a numerical model based on the Reynolds-averaged Navier–Stokes (RANS) equation and sediment transport principles was established to investigate local scouring around a pipeline under wave and current conditions. The influence of the pipeline diameter D, flow incident angle a, current velocity Uc and KC number on the scour depth was analyzed. Based on a total of 145 sets of numerical simulation cases, a dataset was established, and all data were normalized and divided into a training set (80%) and a test set (20%). A decision-tree regression model was used to train the dataset, and a machine-learning prediction model was constructed. In the testing stage, the model had R2 = 0.94, RMSE = 0.043 for the scour depth S prediction; R2 = 0.92, RMSE = 0.096 for the dimensionless scour depth S/D prediction; and R2 = 0.97, RMSE = 0.267 for the scour hole width L prediction. The machine-learning prediction model exhibited good performance and computational efficiency, and can be used for rapid prediction of submarine pipeline scour characteristics.