PREDICTION OF SOME PHYSICAL PROPERTIES OF NANOFLUIDS INCLUDING VARIOUS METAL OXIDES USING ARTIFICIAL NEURAL NETWORK

In this study, physical properties of nanofluids such as viscosity and thermal conductivity are investigated. Although there are many experimental and theoretical correlations and models available to predict these parameters, predictions are highly conflicting because the involved mechanisms are not fully understood. For instance, the predicted values of dynamic viscosity of gamma-Al2O3/water nanofluids (at a volume fraction of 0.001) by different models range between 0.0005623 and 0.0403 kg/m.s and thermal conductivity ranges within 0.866 and 1.551 W/m center dot K under the same conditions. In this work, 184 experimental data of thermophysical properties of various metal oxide nanoparticle including Al2O3, SiO2, TiO2, Fe2O3, MgO, and CuO at a temperature range of 30 to 65 degrees C, diameter range of 10 to 50 nm, and volume fraction between 0.004% to 2% were collected from various publications. In this work, artificial neural network (ANN) has been implemented. The parameters of thermal conductivity and dynamic viscosity have been correlated to nanofluid concentration, temperature, particle size and molecular weight. It has been found that the topography of (4, 4, 4) from the ANN provides about 2% and 1% cross-validating and testing error respectively. Additionally, a sensitivity analysis was performed and the effect of individual parameters was analyzed.