Entropy generation prediction for nanofluid flow using artificial neural networks: a comparative study of training algorithms

Abstract

The present study deals with the artificial neural network approach for entropy generation optimization of a micropolar tetra-hybrid nanofluid passing through a curved artery. In this work, the fluid is exposed to an inclined magnetic field, along with homogeneous and heterogeneous chemical reactions, viscous dissipation, and Joule heating combined with an external heat source. The governing partial differential equations are transformed into a set of ordinary differential equations by applying a group of self-similar transformations. These resulting ODEs are then solved using a shooting technique, ensuring an accurate resolution of the complex boundary conditions. The impacts of physical factors are examined for axial velocity, micro-rotational velocity, temperature, and concentration profiles with entropy generation optimization. A detailed comparison with previously published results confirms a strong agreement, validating the current approach. It is observed that the micro-rotational velocity increases by increasing the radius of curvature parameter, whereas a reverse trend is noted for the micropolar parameter. The Nusselt number is enhanced by an increase in the radiation parameter, while it is reduced by a higher radius of curvature parameter and Brinkman number. This study may help in selecting appropriate nanofluids for sterilization processes, hyperthermia applications, drug delivery systems, bioimaging techniques, maintenance of orthopedic implants, and micro-drug and hormone delivery devices. Furthermore, the integration of artificial intelligence methods enables real-time monitoring and prediction, thereby improving precision and effectiveness in medical and industrial applications involving nanofluids.