Machine learning-based analysis of heat and mass transfer in casson hybrid nanofluid flow over a variable porous riga plate

Abstract

This study uses an artificial neural network (ANN) architecture in combination with the Levenberg-Marquardt backpropagation (LMBP) method to investigate the impact of heat radiation on the hybrid nanofluid flow through a variable porous medium beneath a stretchy Riga plate. The purpose of this work is driven by the increasing demand for accurately modeling complex nonlinear thermal systems involving electromagnetic control, porous media, and nanofluids, which are not sufficiently covered in the existing literature. Single-walled carbon nanotubes (SWCNTs) and multiwalled carbon nanotubes (MWCNTs) are suspended in propylene glycol, which is regarded as the base fluid. The Riga plate is a plate composed of alternating electrodes as well as permanent magnets to form electromagnetic forces, which act significantly on the flow and the heat transfer properties. The governing nonlinear partial differential equations (PDEs) are transformed into ordinary differential equations (ODEs) via similarity transformations in order to provide datasets for ANN training. The finite difference technique (FDM) is then used to numerically solve these ODEs. Then the successive over relaxation method, executed through custom MATLAB code, is used to obtain the solutions of these equations. The novelty of the work is in the integrated evaluation of hybrid nanofluid flow over a Riga plate in a variable permeable media with radiative heat transfer, as well as the use of an ANN-LMBP predictive model, which is trained on high-fidelity numerical data. The numerical data show that the thermal radiation increases the temperature distribution by about 21% in the case of SWCNT-based nanofluid and 18% in the case of MWCNT-based nanofluid. The porosity parameter decreases the velocity of fluid by approximately 12%, whereas the modified Hartmann number increases the fluid velocity by 15–17%. Additionally, SWCNTs exhibit 8–10% higher thermal conductivity compared to MWCNTs. The ANN-LMBP model demonstrates excellent predictive performance with a coefficient of determination R2 > 0.998 and a mean squared error on the order of 10−4. These results confirm that the proposed ANN-based framework is a fast, accurate, and computationally efficient tool for modeling nonlinear thermal fluid systems, with significant applications in thermal management, cooling technologies, energy systems, and industrial fluid flow processes. Clinical trial number not applicable.