Deep learning-based Adam optimization for magnetohydrodynamics radiative thin film flow of ternary hybrid nanofluid with oscillatory boundary conditions

Abstract

This work investigates the new and complete characteristics of radiation, magnetic field, and heat source/sink in the unstable thin-film flow of ternary and hybrid nanofluids over the stretching surface with oscillatory boundary conditions. of radiation, magnetic field, and heat source/sink in the unstable thin-film flow of ternary and hybrid nanofluids over the stretching surface with oscillatory boundary conditions. The flow field is mathematically formulated and solved numerically using BVP5C and deep neural networks with MATLAB software; considering industrial applications, Ethylene glycol (EG) is taken as base fluid, and the nanoparticles utilised in this study include Aluminium oxide $\left({\mathrm{Al}}_2{O}_3\right)$, carbon nanotubes with one or more walls (SWCNTs, MWCNTs). Further, the model is trained by adapting the deep neural network (DNN) technique. Graphical simulations are prepared for Case 1: $\mathrm{EG}+\text{SWCNT}+{\mathrm{Al}}_2{O}_3$ and Case 2: $\mathrm{EG}+\text{SWCNT}+\text{MWCNT}+{\mathrm{Al}}_2{O}_3$. To analyse the significance of unsteadiness, Prandtl, Eckert number, radiation, magnetic, film thickness, source/sink parameter on velocity, temperature and Nusselt number. The research showcases that heat transfer is high in $\mathrm{EG}+\text{SWCNT}+\text{MWCNT}+{\mathrm{Al}}_2{O}_3$ compared with $\mathrm{EG}+\text{SWCNT}+{\mathrm{Al}}_2{O}_3$ hybrid nanofluid. Increasing the layer thickness and unsteadiness parameters lowers temperature and velocity. Applied DNN model shown to be extremely useful for prediction and estimation. Obtained results are helpful in the formulation of advanced products and processes.