Neurofuzzy-Based GMDH-PSO to Predict Maximum Scour Depth at Equilibrium at Culvert Outlets

Abstract

In this study, the neurofuzzy-based group method of data handling (NF-GMDH) as an adaptive learning network was utilized to predict the maximum scour depth at equilibrium downstream of culvert outlet structures. The NF-GMDH network was developed using particle swarm optimization (PSO). Effective variables on the maximum scour depth at equilibrium included those of sediment size downstream of culvert outlets, the geometry of culvert outlets, and the flow characteristics upstream and downstream of the culvert. Training and testing performances of the NF-GMDH-PSO network were carried out using nondimensional data sets that were collected from the literature. The testing results of the NF-GMDH-PSO model were compared with the gene-expression programming (GEP) and traditional equations. The NF-GMDH-PSO network produced a lower error of maximum scour depth at equilibrium prediction than those obtained using the other models. Also, the most effective parameter on the maximum scour depth at equilibrium was determined using a sensitivity analysis approach.