Application of Machine Learning to Investigation of Heat and Mass Transfer Over a Cylinder Surrounded by Porous Media—The Radial Basic Function Network

This paper investigates heat and mass transport around a cylinder featuring non-isothermal homogenous and heterogeneous chemical reactions in a surrounding porous medium. The system is subject to an impinging flow, while local thermal non-equilibrium, non-linear thermal radiation within the porous region, and the temperature dependency of the reaction rates are considered. Further, non-equilibrium thermodynamics, including Soret and Dufour effects are taken into account. The governing equations are numerically solved using a finite-difference method after reducing them to a system of non-linear ordinary differential equations. Since the current problem contains a large number of parameters with complex interconnections, low-cost models such as those based on artificial intelligence are desirable for the conduction of extensive parametric studies. Therefore, the simulations are used to train an artificial neural network. Comparing various algorithms of the artificial neural network, the radial basic function network is selected. The results show that variations in radiative heat transfer as well as those in Soret and Dufour effects can significantly change the heat and mass transfer responses. Within the investigated parametric range, it is found that the diffusion mechanism is dominantly responsible for heat and mass transfer. Importantly, it is noted that the developed predictor algorithm offers a considerable saving of the computational burden.