Machine learning-driven analysis of heat transfer and entropy generation in blood nanofluid flow for biomedical applications

Efficient thermal management is essential in blood flow systems for advancing biomedical applications such as thermal therapies, vascular implants, and lab-on-a-chip diagnostics. This study numerically investigates the forced convection of a Williamson fluid over three distinct geometrical configurations: a flat plate, a wedge, and a stagnation point, utilizing machine learning. The investigation incorporates the effects of a magnetic field, enabling a detailed assessment of magnetohydrodynamic influences on the fluid’s convective transport behavior. Additionally, the study accounts for nonlinear thermal radiation and the presence of gyrotactic microorganisms. Unlike prior research that considers each geometry independently, our approach integrates multi-geometry numerical simulations with predictive modeling, providing a comprehensive understanding of geometry-dependent transport phenomena. The governing nonlinear PDEs, including momentum, energy, concentration, and gyrotactic microorganism dynamics, are converted via similarity transformations and solved using the Runge–Kutta–Fehlberg (RKF) method. A multiple linear regression (MLR) on 600 simulation datasets then facilitates rapid prediction of the modified Nusselt number. The model incorporates magnetohydrodynamic effects, nonlinear radiation, viscous dissipation, and bioconvection, ensuring that the simulations reflect physiologically relevant conditions. Entropy generation analysis is performed to evaluate thermodynamic irreversibility, offering deeper insights into flow efficiency in biomedical contexts. The results reveal that the temperature ratio is the dominant parameter controlling heat transfer across all geometries, contributing 40.8%, 35.51%, and 32.13% for flat plate, wedge, and stagnation flows, respectively, while the Weissenberg number has the most significant impact near stagnation points. The Weissenberg number significantly affects the stagnation point geometry (15.50%), compared to lower impacts in the plate (5.35%) and wedge (11.68%) geometries. The study’s novelty stems from its simultaneous assessment of multiple boundary-layer geometries, integration of entropy analysis with machine learning predictions, and identification of geometry-specific factors that govern heat transfer in blood nanofluids.