H. Safikhani - Department of Mechanical Engineering, Amirkabir University of Technology, Tehran, I.R. Iran
A. Abbassi - Department of Mechanical Engineering, Amirkabir University of Technology, Tehran, I.R. Iran
S. Ghanami - Department of Mechanical Engineering, University of Sistan & Baluchestan, Zahedan, I.R. Iran

In the present study, Computational Fluid Dynamics (CFD) techniques and Artificial Neural Networks (ANN) are used to predict the pressure drop value (Δp ) of Al۲O۳-water nanofluid in flat tubes. Δp  is predicted taking into account five input variables: tube flattening (H), inlet volumetric flow rate (Qi  ), wall heat flux (qnw  ), nanoparticle volume fraction (Φ) and nanoparticle diameter (dp ). The required output data for training the ANN are taken from the results of numerical simulations. The numerical simulations of nanofluid are performed using two phase mixture model by FORTRAN programming language. The flow regime and the wall boundary conditions are assumed to be laminar and constant heat flux respectively. The ANN results are compared with the numerical simulated one and excellent agreement is observed. To view the accuracy of ANN model, statistical measures R۲ , RMSE and MAPE are used and it is seen that the ANN model has high accuracy in predicting the (Δp ) values.  

ANN, GMDH, Mixture model, Nanofluid, Pressure drop