Neural network and Taguchi design optimization of solvent fraction effects on heat dissipation in Bodewadt flow

Abstract

The work examines thermo-hydrodynamic features of Boger fluid over a stretchable disk, underlining its significance in heat transfer processes across energy and cooling systems. Bodewadt flow governs the fluid dynamics, ensuring a consistent angular momentum at distant points, regulated by centrifugal forces and radial pressure gradients. This research explores three-dimensional incompressible Bodewadt flow of Boger fluid on a stretchable disk, considering the role of endothermic/exothermic reaction and Arrhenius activation energy. The mathematical framework employs similarity transformations to convert a set of partial differential equations into ordinary differential equations. By employing the Runge-Kutta shooting technique, the study ensures an accurate and computationally efficient numerical solution. In addition, Taguchi-based statistical design enables efficient parameter selection while ensuring robust assessment of heat transfer behavior. Advancing further, artificial neural networks with the Levenberg–Marquardt backpropagation algorithm are employed to improve prediction accuracy. Additionally, the proposed fluid model shows optimal accuracy, confirmed by a mean square error of $4.6658\times {10}^{-7}$. Findings reveal that Brownian motion significantly governs heat transfer, contributing 91.48 %, while the maximum heat transfer rate recorded reaches 0.949473.