Buoyancy-driven convection and entropy generation analysis of Ostwald–de Waele nano-suspension in a trapezoidal porous domain with V-shaped baffles using multi-layer perceptron learning algorithm

Nanofluids have the potential to completely transform the solar-thermal sector due to its capability for solar-thermal absorption. In view of state-of-the-art developments, a thorough grasp of the design requirements, manufacturing methods, application domains, and technical difficulties of these innovative solar ideas is desperately needed. The present study aims at the enhancement of the efficiency of the solar photovoltaic modules subject to free convective flow of nanofluid suspension in a trapezium-shaped porous domain with V-shaped cold baffles placed on the sides of an inserted inner cylinder. Thermal transfer and irreversibility optimization are carried out. The porous domain is saturated with a water-based nanofluid and fraction of CuO nano-sized solid particles and acting as a non-Newtonian shear-thickening fluid. The efficacy of various arrangements of V-shaped baffles, along with other effective parameters, such as Flow index parameter (n = 1.2, 1.5, and 1.8), Darcy number (Da = 10^–3, 10^–2, and 10^–1), volume fraction of CuO solid nanoparticles in water (ϕ = 0.01, 0.03, and 0.05), Rayleigh number (Ra = 10⁵ and 10⁶) on the hydrodynamic, thermal, and entropy generating features, is evaluated. A computational fluid dynamic-based finite-element method (FEM) is implemented for numerical exploration of governing equations that model the physical problem and one of the supervised machine learning algorithms named multi-layer perceptron (MLP) would be exerted for estimation of the values of velocities, stream function, and temperature. The findings emphasized the importance of including cold baffles on the inserted cylinder to improve the rate of heat transfer in this configuration. Furthermore, the results indicate that the value of mean square error for validation would be 3.4907e−4 which proves that established neural network is perfectly capable of estimating the amounts of U, V, Ψ, and T.