A novel Hybrid Computational–Neural Framework to the Mechanisms of Blood Flow and Thermal Transport in Non-Planar Curved Arteries

: This study investigates physiological blood flow and heat transfer in a two-dimensional curved artery using the Carreau–Yasuda non-Newtonian model. Blood is considered an incompressible, laminar, and shear-thinning fluid. The finite element method (FEM) is employed to solve the coupled continuity, momentum, and energy equations and to obtain detailed distributions of velocity, pressure, temperature, streamlines, and isotherms. The effects of the Carreau–Yasuda model index (n), curvature-related parameter (λ), Reynolds number (Re), and Richardson number (Ri) on flow and thermal characteristics are examined. The results indicate that increasing the model index weakens the shear-thinning behavior, enhances axial velocity and pressure, and reduces convective heat transfer. Higher values of the torsion-related parameter intensify secondary flow structures and produce more complex streamline patterns with non-monotonic heat transfer behavior. Increasing the Reynolds number strengthens inertial effects, redistributes momentum toward transverse motion, and weakens thermal gradients due to reduced fluid residence time near the wall. Higher Richardson numbers enhance buoyancy-driven secondary motion and generate fluctuations in heat transfer because of the interaction between forced and natural convection. Furthermore, an artificial neural network (ANN) model is developed to provide rapid and accurate prediction of the hemodynamic and thermal responses. The ANN predictions show excellent agreement with the FEM results. The proposed hybrid FEM–ANN framework provides an efficient computational approach for analyzing non-Newtonian blood flow and heat transfer in curved arterial geometries, with potential applications in biomedical and cardiovascular engineering.