Artificial neural network-assisted study on thermohydrodynamic behavior of tetrahybrid nanofluids in a porous stretching cylinder

Abstract

This study explores the flow dynamics and thermal characteristics of a tetrahybrid nanofluid over a stretching cylinder, considering the effects of a magnetic field and internal heat generation. Two distinct tetrahybrid nanofluids are examined for the comparative analysis of temperature, pressure, velocity distributions, skin friction, and heat transfer performance: one composed of Ag+SiO${}_2$＋TiO${}_2$＋Al${}_2$O${}_3$ suspended in kerosene oil, and the other consisting of Au+CuO＋Fe${}_3$O${}_4$+ Multi-Walled Carbon Nanotubes ($MWCNTs$) dispersed in water. The governing equations are solved numerically using the fourth-order Runge–Kutta method coupled with a shooting strategy and artificial neural network (ANN). Parametric studies revealed that the Au+ CuO＋Fe${}_3$O${}_4$+Multi-Walled Carbon Nanotubes ($MWCNTs$) nanofluid exhibited superior thermal performance, characterized by higher Nusselt numbers, while the Ag＋SiO${}_2$＋TiO${}_2$+Al${}_2$O${}_3$ nanofluid provided enhanced momentum transport and higher velocity profiles. Au+CuO＋Fe${}_3$O${}_4$+Multi-Walled Carbon Nanotubes ($MWCNTs$) shows stronger pressure resistance near the surface, while Ag＋SiO${}_2$＋TiO${}_2$＋Al${}_2$O${}_3$ yields greater skin friction due to higher effective viscosity. An artificial neural network (ANN) was trained using Bayesian regularization to accurately predict skin friction and Nusselt number values. The Au+CuO＋Fe${}_3$O${}_4$+Multi-Walled Carbon Nanotubes ($MWCNTs$) nanofluid is well-suited for high-efficiency thermal management systems, including electronic cooling, solar collectors, and magnetic heat delivery. In contrast, Ag+SiO${}_2$＋TiO${}_2$＋ Al${}_2$O${}_3$ nanofluid offers advantages in polymer extrusion, coating, and lubrication processes. The inclusion of porous media effects further broadens applicability to geothermal systems, packed-bed reactors, and smart heat exchangers. ANN predictions closely match the numerical results with correlation coefficients above 0.9999, demonstrating its reliability as a surrogate model.