Prediction Of Flow Through Rockfill Dams Using A Neuro-fuzzy Computing Technique

Abstract

Rockfill dams are economical and fast tools for flood detention and control purposes. Artificial intelligence approaches may provide user-friendly alternatives to very complex and time-consuming numerical methods such as finite volume and finite element for predicting flow through rockfill dam. Therefore, this paper examines the potential of coactive neuro-fuzzy inference system (CANFIS) for estimation of flow through trapezoidal and rectangular rockfill dams. The results showed that accurate flow predictions can be achieved with a CANFIS with the Takagi–Sugeno–Kang (TSK) fuzzy model and the Bell membership function for both trapezoidal and rectangular rockfill dams. Furthermore, Levenberg-Marquardt and Delta-Bar-Delta were the best algorithms for training the network in order to estimate flow through rectangular and trapezoidal rockfill dams, respectively. Overall, the results of this study suggest the possibility for using CANFIS for prediction of flow through rockfill dam.