Hybrid GMDH-type neural network to predict fluid surface tension, shear stress, dynamic viscosity & sensitivity analysis based on empirical data of iron(II) oxide nanoparticles in light crude oil mixture

The effects of different parameters including nanoparticles mass fraction and temperature were investigated on rheological behavior and surface tension of iron(II) oxide/light crude oil nanofluid. Iron(II) oxide was dispersed in light crude oil by using ultrasonic processor. TEM images was provided in order to assess the size and morphology of iron(II) oxide nanoparticles. In addition, DLS analysis and Zeta potential test were performed on nanofluid for estimation of nanoparticles size distribution within the basefluid and stability of nanoparticles, respectively. The results of this study showed that for iron(II) oxide/light crude oil nanofluid the value of surface tension reach to its minimum value at the condition where nanoparticles mass fraction was chosen to be 2.0 wt% and temperature was set on 70 °C. These results showed that for iron(II) oxide/light crude oil nanofluid the rheological behavior of nanofluid is non-Newtonian at temperature of 40 °C and the suspension behave as a rheoplexy fluid in which by increasing the shear rate higher dynamic viscosity of nanofluid observed. Furthermore, nanofluid behaves as a Newtonian fluid for the temperature of higher than 55 °C. Finally, a comprehensive correlation was obtained for estimation of relative dynamic viscosity of nanofluid by hybrid group method of data handling (GMDH)-type neural network method. The correlation presented in this study shows that for the relative dynamic viscosity of iron(II) oxide/light crude oil as a function of nanoparticles mass fraction and temperature, the amount of the total deviation of calculated data from experimental values is less than 10%.