Multiphysics Modelling of Tri-Hybrid Nanofluid Transport over a Quadratically Radiative Riga Surface with Electroosmotic Effects and Pollutant Transport: Numerical Simulation, AI-Based Surrogate Prediction, and Industrial Applications

In this study, an integrated finite-difference and deep neural network analysis of the simultaneous multiphysics of coupled momentum, heat, nanoparticle, and pollutant transport in a Sutterby tri-hybrid nanofluid over a quadratically radiative Riga surface is developed. The Fe3O4, Al2O3, and TiO2 nanoparticles are considered, and a mixture of propylene glycol and water is used, along with electroosmotic forcing, electromagnetic actuation, nonlinear thermal radiation, viscous dissipation, Brownian motion, thermophoresis, and pollutant transport. The key novelty is the ability to formally connect these strongly coupled nonlinear transport processes with a deep neural network (DNN) surrogate model to allow for the rapid and accurate prediction of complex thermo-fluid and mass-transfer responses. The governing equations are then transformed into nonlinear equations and solved numerically using a finite difference method (FDM) coded in FORTRAN to produce high fidelity numerical data sets that are used to train a feed-forward back-propagation deep neural network (DNN) implemented in Python with PyTorch. The optimized DNN successfully predicted the velocity, temperature, field of nanoparticles, and pollutant concentration, with the RMSE less than (3.2 × 10−2) and (R2 > 0.987), which indicates that the optimized DNN possesses high predictive accuracy and reliability. The fluid motion is improved by the increased electroosmotic and electromagnetic parameters due to increased electrokinetic and Lorentz forces, and the increased Brinkman number increases the fluid viscous dissipation heating and raises the temperature field. The enhanced Brownian motion facilitates the migration of nanoparticles and greatly affects coupled heat and mass transport while pollutant-related parameters control concentration distributions. The proposed DNN framework has the potential of being an efficient surrogate for fast simulations, parametric evaluation, and optimization for nonlinear multiphysics systems, and has direct applications to thermal management systems in industry, microfluidic processing, electrochemical technologies, electromagnetic flow-control devices, industrial wastewater treatment systems, and pollutant removal systems.