The objective of this article is to investigate the mechanism of unsteady heat and mass transfer (HMT) in a magnetohydrodynamics Carreau nanofluid flow (MHD-CNFF). To accomplish this, we utilized a novel artificial intelligence-based (AI) method, deep learning predictive technique combined with Levenberg-Marquardt scheme (DLPT-LMS). The similarity transformation is applied to convert the partial differential equations (PDEs) into a set of non-linear ordinary differential equations (ODEs). We numerically solved these resulting ODEs using the Adam numerical approach in sophisticated “Mathematica” software. The influence of Brownian motion parameter, thermophoresis parameters, Prandtl number, Lewis number, and mass transfer parameters on the temperature, velocity, and concentration profile is analyzed using a synthetic dataset. This evaluation is conducted across different variation of MHD-CNFF by leveraging DLPT-LMS. The absolute error (AE) across various iteration of MHD-CNFF demonstrated a progressive drop, indicating the accuracy of DLPT-LMS. The outcomes of this investigation revealed that the AE ranged from 10−3 to 10−8 which shows a minimum error between actual and predicted value. Additionally, the MSE confirmed the accuracy of proposed DLPT-LMS in predicting the behavior of velocity, temperature, and concentration profiles, with lower values showing better predictions. The model accuracy was reinforced by performance results, including regression analysis and error histograms. Error values gradually decreased during training, validation, and testing phases, exhibiting a robust convergence and indicating the highly reliability of AI-based algorithm for investigating intricate fluid dynamics in MHD-CNFF systems.
Read full abstract