Machine Learning for Magnetohydrodynamic Nanofluid Flow

In this article, the two-dimensional magnetohydrodynamic (MHD) flow of a hyperbolic tangent fluid over a stretched surface containing nanoparticles is examined. Using appropriate scaling, the controlling nonlinear partial differential equations are converted into ordinary differential equations, which are then resolved by the shooting technique. We analyze how velocity, temperature, and concentration are affected by the following important parameters: Brownian motion parameter (Nb), Lewis number (Le), Prandtl number (Pr), Weissenberg number (We), Hartmann number (M), and thermophoresis parameter (Nt). The findings indicate that while Brownian motion and thermophoresis increase temperature, increases in Weissenberg and Hartmann numbers slow down fluid motion. As Nb, Pr, Le, and Nt grow, the local Sherwood number rises, but the Nusselt number falls. Additionally, an artificial neural network (ANN) is used to train the solution, and its accuracy, training dynamics, and computing efficiency are compared with the conventional BVP4C solver. The work emphasizes how adaptable ANNs may be as a substitute for conventional numerical techniques in boundary value analysis.