s baghalian - Department of Civil Engineering, Razi University, Kermanshah, Iran
h Bonakdari
f Nazari - Department of Mechanical Engineering, Young Researchers Club
m Fazli

In this paper the flow field in a 90o bend is analyzed numerically and also predicted using Artificial Neural Networks (ANN). In this study, first of all, for solving governing equations of flow, a computational fluid dynamic (ANSYS-CFX 12) code is used. Then, by presenting a 3D numerical modelin 2 phases, variation of velocity field in different sections of a bend is studied. In addition results ofnumerical model are validated by experimental data. After that for prediction of velocity field in variable sections of bend, an ANN is trained based on Back-Error Propagation (BEP) technique. The inputs of the ANN are depth, radius and angle of different points of bend, and target outputs are corresponding velocities. Results of ANN test show that they have good agreement between actual and predicted data

Numerical study, Bend, Velocity field, Artificial Neural Network, Prediction