Entropy generation analysis on viscoelastic hybrid nanofluid flow between coaxial pipes with Darcy-Furchhiemer impact: an application of drilling fluid

This analysis aims to improve knowledge of the natural convection of a third-grade hybrid (SiO₂+TiO₂/KO) nanofluid flow in a porous medium through two concentric drilling pipes with entropy generation. A hybrid nanofluid was prepared by dispersing silicon dioxide (SiO₂) and titanium dioxide (TiO₂) nanoparticles in a base fluid of kerosene oil (KO). The current study also incorporates Darcy-Forchheimer and magnetic field effects. Effective heat management throughout the drilling process and improving drilling efficiency also depend on taking into consideration factors such as viscous dissipation, thermal radiation, and Joule heating. Also, an artificial neural network (ANN) is used to optimize the performance of the model. A system of nonlinear differential equations is created from suitable similarity variables. The fourth-order accuracy technique (Bvp4c) is used to generate numerical solutions. Graphical studies are used to show how different factors affect fluid flow, entropy generation, pressure, and heat transfer. The velocity profile falls because of the influence of nanoparticles volume fraction, Darcy-Forchheimer number, and the magnetic field parameter, which are crucial for controlling fluid flow during drilling operations. The nanoparticle volume fraction, Eckert number, heat source/sink, magnetic field, and radiation parameter all contribute to the increased temperature distribution, which is essential for preserving fluid stability and optimizing heat transfer. Entropy generation escalations with an upsurge in thermal radiation, Reynolds number, Hartman number, and Brinkman number. In addition, the skin friction reduces with an expansion of the Darcy-Forchheimer number while the Nusselt number declines with an upsurge in the nanoparticle volume fraction. The ANN is capable of accurately calculating system performance and executing complex data patterns. The overlapping of the results and lower absolute error for the model indicate that the optimizer was well built. Furthermore, the suggested solver’s competencies are provided by just a few of curves based on mean square error (Loss function), error histograms and regression.