Neural networking-based approach for examining heat transfer and bioconvection in Non-Newtonian fluid with chemical reaction over a stretching sheet

This research investigates the impact of bioconvection and magnetohydrodynamics on Casson-Williamson fluid flow over a stretching surface while considering the effect of heat sources, thermal radiation, and chemical reactions. They are significant components in industrial processes and biomedical systems, such as targeted drug release and cancer therapies. The nonlinear governing partial differential equations (PDEs) are converted into ordinary differential equations (ODEs) employing similarity transformations and are numerically solved through a fourth-order procedure. Afterward, an Artificial Neural Network (ANN) with Levenberg-Marquardt training evaluates the flow pattern. The dataset is split into 70 % train, 15 % test, and 15 % validate to maximize model precision and generalizability. Mean Squared Error (MSE) is utilized to measure precision, whereas regression analysis (R ≈ 1) verifies strong prediction accuracy. Findings indicate that the momentum boundary layer reduces with increasing magnetic field and buoyancy ratio, whereas the Nusselt number increases with increased radiation parameters but decreases for growing heat source, Brownian motion, thermophoresis, and Eckert numbers. Skin friction also augments with greater magnetic, porosity, Rayleigh, and buoyancy parameters. These results contribute to the optimization of fluid flow in nanobiotechnology, chemical engineering, and heat transfer processes, underlining the potential of bioconvection and AI-based modeling for future development.