Utilizing neural networks to illustrate the dynamics of viscous fluid flow over curved surface with homogeneous and heterogeneous reactions

Abstract

This study examines the influence of homogeneous-heterogeneous reactions and viscous dissipation on the magnetohydrodynamic (MHD) boundary layer flow over a curved stretching sheet, incorporating the effects of partial slip and a non-uniform heat source. Understanding these interactions is crucial for optimizing heat and mass transfer in industrial applications where precise thermal and solutal control are required. The governing partial differential equations are transformed into a system of coupled ordinary differential equations using similarity transformations and solved numerically via the Runge-Kutta method with a shooting technique. A comparative analysis with existing studies further validates the accuracy of the present findings, providing strength into flow control mechanisms and heat transfer enhancement strategies relevant to industrial thermal systems. Moreover, results indicate that increasing the magnetic field parameter increases the shear rate by 65.62 %, whereas thermal dissipation reduces the heat transfer rate by 13.63 %. Additionally, an Artificial Neural Network (ANN) model is employed to predict drag force, heat transfer, and mass transfer rates, achieving a validation accuracy exceeding 99 % with a mean squared error (MSE) of approximately 10⁻¹¹ and a regression coefficient (R) close to 1 for each case. Moreover, the inputs for predicting drag force are provided to the ANN by varying the values of curvature parameter $\left(K\left(1-5\right)\right)$, magnetic parameter $\left(M\left(1-3\right)\right)$ and slip parameter $\left({\lambda }_1\left(0.1-0.3\right)\right)$, the heat transfer rate controlled by appropriately adjusting the parameters $K\left(1-5\right),M\left(1-3\right)\text{,}$ heat source parameters $\left(A^{\ast }\left(0.5-1.5\right),B^{\ast }\left(0.1-0.4\right)\right)$ and Eckert number $\left(Ec\left(0.1-0.3\right)\right)$ and the solutal rate is determined by adjusting the parameters $K\left(1-5\right)\text{,}$ Schimdt number $\left(Sc\left(3-6\right)\right)\text{,}$ homogeneous and heterogeneous reaction parameters $\left(k_1\&k_2\left(0.1-0.25\right)\right)$. The ANN predictions show strong agreement with numerical results, confirming the reliability of the model for real-world thermal engineering applications.