No abstract

No abstract

Protection of submarine pipelines against scouring processes is considered one of the most important issues in the design of submarine structures. Most reported experimental investigations of seabed pipeline scour were conducted to acquire relationships with precise estimations to predict the local scour depth below a pipeline. However, several empirical equations based on experimental datasets do not have a sufficient capacity of validation to evaluate the local scour depth due to the complicated mechanism of scouring phenomena. Thus, data-driven approaches based on gene-expression programming (GEP), evolutionary polynomial regression (EPR) and model tree (MT) have been used to develop explicit equations to estimate local scour depth below pipelines due to currents. The experimental datasets utilised in this study are grouped into two sections: clear-water and live-bed conditions. The effective parameters on the local scour include sediment size, pipeline diameter, initial gap between the pipeline and the bed sediment and the approaching flow characteristics. The training and performance testing of the proposed models are conducted using non-dimensional input–output variables. The efficiency of the training stages for GEP, MT and EPR is studied. The performances of the testing stages are compared with traditional approaches.

No abstract