Application of improved neuro-fuzzy GMDH to predict scour depth at sluice gates

Abstract

An improved neuro-fuzzy based group method of data handling using the particle swarm optimization (NF-GMDH-PSO) is developed as an adaptive learning network to predict the localized scour downstream of a sluice gate with an apron. The input characteristic parameters affecting the scour depth are the sediment size and its gradation, apron length, sluice gate opening, and the flow conditions upstream and downstream of the sluice gate. Six non-dimensional parameters were yielded to define a functional relationship between the input and output variables. The training and testing of the NF-GMDH network are performed using published scour data from the literature. The efficiency of the training stages for the NF-GMDH-PSO is investigated. The testing results for the NF-GMDH network are compared with the traditional approaches based on regression method. A sensitivity analysis is carried out to assign the most significant parameter for the scour prediction. The results showed that the NF-GMDH-PSO network produced lower error in scour prediction than all other models.