Predictive modelling of flow dynamics in micropolar hybrid nanofluids subjected to magnetic dipole influence using nonlinear autoregressive artificial neural networks with exogenous input

Abstract

In the present research article, we explore the intelligence-based numerical computation of a nonlinear autoregressive artificial neural network with exogenous input using the Levenberg-Marquardt algorithm (ANN-LMA) trained to analyze the unsteady two-dimensional micro-polar flow of an incompressible hybrid nanofluid (Al 2 O 3−Fe 3 O 4/Ethylene Glycol) under the impact of the magnetic field introduced by the magnetic dipole on a nonlinear stretched sheet. In the first step of the research, partial differential equations are converted into nonlinear ordinary differential equations using appropriate transmission and then solved numerically via the Successive Over Relaxation method after applying the finite difference method. The impact of numerous emerging parameters on the solutions is displayed graphically, and the physical significance is discussed. The results show that the concentration profile displays a dwindling trend for the Brownian motion parameter, and the opposite trend is witnessed for the thermophoretic parameter. Moreover, for increasing values of the dimensionless distance parameter and ferromagnetic interaction parameter, the temperature and velocity profiles exhibit the opposite tendency. In the second step of the study, we validated the Successive Over Relaxation outcomes with an ANN-LMA. For this purpose, the suggested ANN-LMA is trained for six different scenarios of the reference dataset points of the given flow model. The analysis shows that Successive Over Relaxation and LM-SNNs yield similar results, with a decreasing trend for the Brownian motion parameter and an opposite trend for the thermophoretic parameter.