Group method of data handling to predict scour depth around vertical piles under regular waves

AbstractThis paper presents a new application of the Group Method Of Data Handling (GMDH), to predict pile scour depth exposed to waves. The GMDH network was developed using the Levenberg–Marquardt (LM) method in the training stage for scour prediction. Scour depth due to regular waves was modeled as a function of five dimensionless parameters, including pile Reynolds number, grain Reynolds number, sediment number, Keulegan–Carpenter number, and shields parameter. The testing results of the GMDH-LM were compared with those obtained using the Adaptive Neuro-Fuzzy Inference System (ANFIS), Radial Basis Function-Neural Network (RBF-NN), and empirical equations. In particular, the GMDH-LM provided the most accurate prediction of scour depth compared to other models. Also, the Keulegan–Carpenter number has been determined as the most effective parameter on scour depth through a sensitivity analysis. The GMDH-LM was utilized successfully to investigate the influence of the pile cross section and Keulegan–Carpenter number on scour depth