Machine Learning Prediction of Artificial Cilia-Driven Ethylene Glycol Nanofluid in a Porous Channel with Activation Energy Effect

Artificial-cilia-driven transport is an important phenomenon in microfluidic systems, biomedical devices, porous transport media, and thermal-management applications, where controlled momentum, heat, and mass transfer are required. Accurate prediction of such coupled transport processes becomes challenging when magnetic effects, porous resistance, variable thermophysical properties, and chemical reactions are considered simultaneously. In the present study, we focus on the numerical and machine-learning prediction of artificial-cilia-driven ethylene glycol–silver (Ag–EG) nanofluid flow through a curved porous channel. The mathematical model incorporates Darcy–Brinkman–Forchheimer resistance, magnetic effects, viscous dissipation, heat generation, variable viscosity, thermal conductivity and diffusivity, and Arrhenius activation-energy effects. The governing nonlinear momentum, energy, and concentration equations are formulated in curvilinear coordinates under low-Reynolds-number conditions and solved using an implicit finite-difference method. A feed-forward artificial neural network is developed using 5000 numerical samples, which are divided into 70% training, 15% validation, and 15% testing subsets. The Levenberg–Marquardt algorithm is employed for network training, while MSE, RMSE, and (R2) are used to evaluate prediction accuracy. The results show that increasing the Darcy number enhances fluid velocity because of greater porous-medium permeability, whereas increasing the magnetic and Forchheimer parameters reduces the flow through Lorentz-force and nonlinear porous resistance, respectively. Increasing the thermal-conductivity parameter from 0.0 to 0.6 decreases the maximum temperature by about 15%, while increasing the activation-energy parameter from 0.1 to 1.0 increases the average species concentration by approximately 18–22%. The ANN predictions show close agreement with the numerical results. The developed numerical–machine-learning framework provides an efficient approach for predicting coupled transport behavior in cilia-driven porous-channel systems.