Analytical and artificial neural network modeling of double-diffusive magnetized electroosmotic ellis nanofluid flow through a tapered non-uniform channel

This study investigates the nonlinear transport dynamics of an Ellis nanofluid in a tapered asymmetric peristaltic channel embedded in a Darcy porous medium under electroosmotic effects. The mathematical model incorporates double-diffusive transport, viscous dissipation, Brownian motion, thermophoresis, magnetohydrodynamics, and electrokinetic interactions. The governing set of coupled nonlinear partial differential equations is transformed into a system of ordinary differential equations using lubrication approximations and Debye-Hückel linearization, with the aid of appropriate dimensionless variables and solved using the Homotopy Perturbation Method (HPM). The dataset generated in HPM is further validated and used to predict the nonlinear flow characteristics, for which an Artificial Neural Network (ANN) is developed and trained. A tangent hyperbolic activation function and the Adam optimization algorithm are used in the ANN and the predictive performance is checked using mean squared error, regression analysis, and error histogram. The results show that the magnetic field reduces the fluid velocity but improves the temperature distribution while the electroosmotic forcing increases the fluid velocity and improves the fluid temperature distribution. Furthermore, Brownian motion makes the temperature rise but reduces the concentration of nanoparticles, and thermophoresis will increase both the temperature and the concentration of nanoparticles. The good agreement between the ANN predictions and the HPM solutions confirms the reliability of the proposed computational framework for analyzing complex electroosmotic Ellis nanofluid transport. The suggested model gives useful insights on the design and optimization of electroosmotic microfluidic devices, targeted drug delivery and lab-on-a-chip technologies using non-Newtonian nanofluids.