Deep learning investigation of ternary hybrid nanofluid flow with nonlinear heat source-sink and Cattaneo-Christov heat flux model with Joule effect

Abstract

This study aims to develop a supervised learning artificial recurrent neural network algorithm supported by Levenberg-Marquardt (ARNN-LM) to analyze the impact of physical parameters, including Magnetic parameter Ma, factor pertaining thermal radiation Nr, factor pertaining thermal relaxation, Prandtl number Pr, Suction and injection parameter S, Temperature ratio parameter, heat source/sink parameter and Joule parameter J on the flow behavior of the fluid, and temperature distribution in steady nonlinear two dimensional ternary hybrid nanofluid flow with heat source-sink and Cattaneo-Christov heat flux model passing through a disk with Joule effect (THNF-CCHFMDJ). The investigation of the hybrid nanofluids including silver, manganese zinc ferrite, and copper nanocomposites on a spinning disc is appealing significant consideration due to its diverse applications. To simplify the analysis, nonlinear partial differential equations were effectively transformed into dimensionless systems by the application of similarity transformations. The numerical dataset of the proposed model has been constructed by varying various parameters for nine scenarios that are used in a Levenberg Marquardt-based intelligent computing method to build networks for approximating the numerical solutions of THNF-CCHFMDJ. The ARNN-LM methods is trained using 70% of the data, while 15% is allocated for testing and the remaining 15% for validation. It is observed that across a spinning disk, the fluid temperature increases while the fluid velocity decreases when the magnetic parameter values are decreased. At the higher levels, the temperature profile and radiative heat flux demonstrate substantial growth. The effectiveness and significance of designed ARNN-LM is demonstrated through regression (RG) index measurements, error histogram (EH), auto-correlation (AC) studies, auto-correlation analysis and convergence curves showing a minimal level of mean square error (MSE) for the comprehensive simulations of THNF-CCHFMD.