Artificial neural network modeling of tangent hyperbolic ternary fluid flow with radiation effects

This study examines free convection boundary layer flow and heat transfer of a non-Newtonian tangent hyperbolic ternary nanofluid ( A g − C u − G O /kerosene) from a cone, considering the effects of thermal radiation and magnetic forces. Rosseland’s algebraic diffusion flux model is used to account for radiative effects, valid for optically thick nanofluids. The Keller Box method, a second-order implicit finite-difference technique, is employed to numerically solve the transformed conservation equations with realistic boundary conditions. The validity of the numerical code is verified by comparing its results with those from existing literature. Velocity and temperature profiles within the boundary layer are analyzed for the effects of the Weissenberg number, rheological power law index, surface temperature exponent, magnetic, and radiation parameters. The relative performance of ternary, binary, and unitary kerosene nanofluids is computed. The findings contribute to a deeper understanding of the interactions between fluid dynamics, thermal gradients, and radiation effects in non-Newtonian fluids, with implications for enhancing industrial processes like metal casting, chemical reactions, and renewable energy systems. Moreover, a physics-guided artificial neural network trained using the Levenberg–Marquardt (LM) algorithm is employed to predict the skin friction coefficient and heat transfer characteristics of the flow. The network is trained using numerical solutions of the governing equations, ensuring consistency with the underlying physical behavior. The ANN architecture is selected based on performance indicators such as rapid convergence, small gradient and Mu value and low prediction error on the Nusselt number. The close agreement between numerical and ANN results confirms the reliability and effectiveness of the LM-based physics-consistent neural network for heat transfer prediction.