Intelligent computational analysis of magnetohydrodynamic double-diffusive mixed convection in nanofluids

Abstract

The objective of the current computational study is to examine the influence of different magnetic field inclination angles, buoyancy ratios and rotational speeds of two cylinders on mixed convective heat and mass transfer as well as entropy formation in a lid-driven trapezoidal enclosure. As governing fluids, several water-based nanofluids and hybrid nanofluids with set solid volume fractions (5%) are utilized. In this study, the effects of SWCNT-water, Cu-water, and Al2O3-water nanofluids are explored, as well as the effects of three different types of SWCNT-Cu-Al2O3 -water hybrid nanofluids, each composed of varying proportions of SWCNT, Cu, and Al2O3 nanoparticles in water. The governing Navier-Stokes, thermal energy, and mass conservation equations are solved using the Galerkin weighted residual finite element method via numerical simulation to obtain the average Nusselt number, average Sherwood number, average temperature, and Bejan number as well as entropy generations output parameters inside the enclosure for various parameter values. The results are also presented in forms of streamlines, isotherm and isoconcentration contours for various parameters to get a better insight of heat and mass transfer with fluid flows inside the trapezoidal enclosure. In most cases under magnetohydrodynamic conditions, maximum heat and mass transfer along with maximum average total entropy generation occurs if the magnetic field is applied at an inclined angle with respect to the horizontal axis with maximum rotational speeds of the cylinders for each type of fluids in this current framework. Specially Cu and SWCNT nanoparticles show better results. Using the simulation data, an innovative artificial neural network model for accurate prediction is then developed. For training and validation data, it predicts convective heat and mass transfer with 96.81% accuracy and average dimensionless temperature and Bejan number with 98.74% accuracy. FEM and ANN are used to determine the optimal values for each of these input parameters, and a comparative study is conducted between FEM and ANN to determine the optimal output parameter values. Using a hybrid nanofluid (Cu-Al2O3-water), the performance of the ANN model constructed is tested for novel scenarios. For test data, it predicts convective heat and mass transfer with 97.03% accuracy and average dimensionless temperature and Bejan number with 99.17% accuracy. So, the new ANN model proposed accurately predicts the results for each type of governing fluids in the current framework.