In this study, forced convection of nanofluid flow in various channel geometries with a hydraulic diameter of 16 mm and length of 1.5 m under laminar flow condition has been investigated numerically. Constant heat flux of 6 kW/m2 has been applied on to the surfaces of the channels. Fe3O4/water nanofluid has been used in the analyses to enhance the convective heat transfer of the base fluid. Analyses have been performed for Reynolds numbers between 500≤Re≤2000, and for volume concentrations of nanoparticles between 1% and 5% in cylindrical, square, rectangle, and triangle cross-sectioned channel geometries. The finite volume discretization method has been used to solve the governing equations. The effects of some parameters; Reynolds number, nanoparticle volume fractions, channel geometries on the average Nusselt number, Darcy friction factor and entropy generation have been investigated in detail. The results indicate that nanofluid offers further convective heat transfer enhancement according to base fluid and cylindrical cross-sectioned channel gives the best heat transfer performance among other cross-sectioned channel geometries. Using water as a working fluid, cylindrical cross-sectioned channel geometry gives the highest heat transfer rate among other channel geometries, whereas triangle one gives the lowest. Cylindrical cross-sectioned channel geometry offers up to 77.6% enhancement compared to triangle cross-sectioned channel geometry for the same hydraulic diameter and same heat flux. However, triangle cross-sectioned channel geometry has highest convective heat transfer increment ratio (4.12%) for changing working fluid as water to nanofluid. Also, some new Nu correlations based on the channel geometries and nanoparticle volume fractions were proposed in the present study.

In this study, the experimental results using mono (Fe3O4/water and Cu/water) and hybrid (Fe3O4-Cu/water) type nanofluid with nanoparticle volume concentrations of (0≤φ≤0.02) under laminar flow conditions (994≤Re≤2337) were compared with the results obtained by ANN. While the Reynolds number (Re), hydraulic diameter (Dh), thermal conductivity (k) of working fluid, and volume concentration of the nanoparticles (φ) were selected as input layers, the Nusselt number (Nu) were considered as output layers. The %75 of the findings obtained from experiments were used to train Artificial Neural Network (ANN). The estimated data by ANN is in perfect agreement with the experimental data. The success of ANN was deter-mined by comparing it with SVM, Dec Tree, and their variations. Mean square error (MSE), root mean square error (RMSE), R-sq (R2), and mean absolute error (MEA) were considered in evaluating the results obtained. According to findings, MAE 0.00088274, MSE 1.4106e-06, RMSE 0.0011877 and R2 1.00 were measured. These findings show that the use of ANN is a feasible way to predict the convective heat transfer performance of hybrid nanofluid under a magnetic field (MF).