Finite element and neural network analysis of magnetohydrodynamics induced natural convection of Ag–MgO hybrid nanofluid in a porous wavy trapezoidal cavity
Umer Farooq, Yasir Ul Umair Bin Turabi, Shahzad Munir
Abstract
This study investigates the thermal behavior prediction in porous wavy trapezoidal cavities utilizing magnetohydrodynamics (MHD) induced hybrid nanofluids through a combination of neural networks (NNs) and finite element-based analysis. The novel cavity configuration filled with (Ag–MgO–H2O) hybrid nanofluids with the effect of an inclined MHD is subjected to isothermal heating from the wavy lower wall and circular cylinder and cooled from the left and right walls, while the upper wall is insulated. The finite element method is employed to perform simulations and numerically calculate the desired solutions of the governing model. A comprehensive range of parameters, including the Hartmann number Ha, undulations N, Rayleigh number Ra, nanoparticles volume concentration ϕ, amplitude of wave a, Darcy number Da and inclination γ, is explored to understand their influence on flow patterns, heat transmission, Nuavg, and total entropy formation. In addition, the study implements an artificial NN model to forecast the cavity's overall thermal behavior in a variety of circumstances. The present outcomes revealed that the rate of heat transfer increases with the increasing numbers of the undulations N, Da, and γ, while it reduces by growing Ha. The total entropy generation is directly correlated with N and Da, and it is inversely correlated with Ha. The Nuavg for Ha=0 and ϕ=0.04 is increased 1.4 times more that of Ha=50 and ϕ=0.01. The training state figure, after 648 epochs, demonstrated the optimization progress and convergence with gradient =7.5697 × 106 and Mu = 1 × 10−7.