Comparison of flow pattern in a 60° sharp bend by using FLUENT software and artificial neural network , support Vector Machine Methods

Abstract

This paper presents an experimental and numerical study of the flow patterns in a strongly-curved 60o open channel bend. Corresponding numerical model is based on the Fluent software and ANN,SVM methods. The use of artificial intelligence methods and Support vector Machine in different hydraulic sciences has become conventional in recent years. In this study the turbulence model is used to simulate turbulent flow parameters and Compared of flow pattern in a 60° sharp bend by Using FLUENT software and ANN, SVM Methods. The results show that , enjoying low error values, the FLUENT model has an acceptable level of consistency with the available experimental results. ANN model can predict velocity pattern fairy accurately. The error values of FLUENT and ANN models are smaller in the outer wall (contraction zones) in comparison with the inner wall (separation zone). It could therefore be said that the error value is greater in high- velocity areas (erosion- prone areas) than in low- velocity areas (sedimentation- prone areas). Careful examination of these models will make it clear that both FLUENT and SVM model underestimate and the ANN model overestimates. The error value is very small in the cross sections after the bend in all three models.