عنوان مقاله [English]
Rivers accumulate huge amounts of floating debris including the trunk, branches and leaves during the floods, leading to increase the depth of local scour around bridge piers. A large number of the laboratorial and field studies have been performed to understand the mechanism of scouring phenomenon under floating debris. Over two past decades, different types of the artificial intelligence methods have been used to estimate the maximum scour depth around bridges piers. In this study, the Neuro-Fuzzy model based on group method data handling (NF-GMDH) was used to estimate the scour under effect of debris accumulations. The NF-GMDH network was developed using evolutional algorithms: genetic algorithm (GA), particle swarm optimization (PSO), and gravitational search algorithm (GSA). Parameters effective on the maximum scour depth included average velocity of upstream flow of the bridge pier, critical velocity of river bed sediments, depth of flow in section without debris, thickness of submerged debris, debris diameter, average particle size, pier diameter, and channel width. After training and experiencing each NF-GMDH models, the performances of each one was evaluated through statistical parameters. The results showed that the models proposed had better performance compared with emperical relationships. NF-GMDH-PSO (R=0.8413 and RMSE=0.37) and NF-GMDH-GA (R=0.8407 and RMSE=0.3640) had relatively similar performance. Finally, sensitivity analysis indicated that the ratio of pile diameter (D) to mean diameter of bed sediments (d50) has the most influence on determination of maximum scour depth.