Optimized physics-informed neural networks for deciphering of external source pollutants in a swirling flow induced by a constant torsional motion

The current study inspects the influence of thermophoresis and Brownian motion on fluid stream through a stretchy rotating cylinder with torsional motion. The consequence of heat source/sink and pollutant concentration on the liquid motion is also considered. The present study's partial differential equations (PDEs) are converted into a system of non-linear ordinary differential equations (ODEs) using similarity transformations. The obtained equations are then solved numerically using the Runge Kutta Fehlberg fourth-fifth order (RKF-45) scheme. The flow, heat and mass transport attributes are assessed using the Physics-informed neural network (PINN). This model is optimized by a hybrid genetic algorithm and particle swarm optimization to address the flow, heat and mass transport attributes via neural networks. The effect of several parameters on the thermal, concentration, and velocity profiles is depicted with graphical illustrations. The velocity, concentration, and thermal profile decline as the Reynolds number rises. The thermal profile upsurges with the rise in heat source/sink, Brownian motion, and thermophoretic parameters. As the Biot number intensifies, the temperature upsurges. For the heat source/sink parameter, heat transport rate is enhanced by approximately about 10.2 %. The improvement in the values of the Brownian motion parameter declines the concentration profile, and an opposite trend can be seen with the thermophoretic parameter. The concentration profile increases as the pollutant external source parameter intensifies.