Data-driven prediction of thermo-solutal transport in sodium alginate-based Casson radiative magnetohydrodynamics ternary nanofluid squeeze flow

This study focuses on unsteady squeezing flow between parallel plates for a sodium-alginate-based Casson ternary nanofluid containing graphene, graphene oxide, and silver nanoparticles in the presence of a transverse magnetic field, thermal radiation, and a homogeneous chemical reaction. Using a similarity transformation, the governing mass, momentum, energy, and species equations reduce to a coupled boundary value problem on the similarity coordinate. A collocation solver in Python produces a numerical dataset across the relevant dimensionless parameters. We then train a compact artificial neural network that maps the parameters and the similarity coordinate to the dimensionless velocity, temperature, and concentration fields. The surrogate reproduces the numerical solutions with near unity alignment on an unseen test set. Bar comparisons and identity-line scatter plots show excellent agreement, error histograms are centered at zero with small variance, and bootstrap ensembles deliver narrow predictive intervals. Parameter sweeps confirm that the network preserves the expected physics: a larger Casson parameter reduces the flow intensity, a larger Schmidt number thins the concentration layer, and a larger Eckert number elevates the temperature due to viscous dissipation. The results show that increasing Ec from 0.4 to 0.5 causes an increase of 22.13% in the initial temperature profile. Also, increasing Sc from 1.5 to 2.5 causes a decrease of 4.69% in the concentration profile. Gradient-based sensitivity identifies nanoparticle volume fractions and the Casson parameter as the dominant inputs within the explored design space. The approach provides a fast and accurate predictive tool for radiative reactive magnetohydrodynamic squeezing of Casson ternary nanofluids and enables rapid parametric exploration without repeated boundary value solves.