Integrating Stochastic Learning Approach to Analyze the Thermal Enhancement in Solar Aircraft Considering Nanofluid Flow through Cylindrical Surface

Abstract

In a two‐dimensional magnetic field, a Williamson‐nanofluid (WNF), an incompressible non‐Newtonian MHD fluid, flows past a stretched cylinder. It has several uses in a variety of domains, including material processing, heat transfer, and physiological fluids. The temperature‐dependent viscosity is modeled using the Reynolds viscosity formulation, with the cylinder surface subjected to a Newtonian heating boundary condition. Previous studies have already solved the problem using numerical methods (shooting method and Runge–Kutta method) to attain a solution. The governing physical model is formulated using the boundary‐layer approximation, which transforms the fundamental conservation laws into a system of nonlinear PDEs. Appropriate similarity variables are subsequently used to transform the dimensional PDEs into nonlinear ODEs. The resulting equations are solved numerically in Mathematica using the NDSolve algorithm. The flow, temperature, and concentration profiles are then calculated over a wide range of governing physical factors to examine their influence on the transport characteristics of the flow system. To further improve the predictive capability of the analysis, an AI‐based framework employing the NN‐LMBPA is implemented in MATLAB. The numerical solutions generated by the conventional solver are used to construct the training dataset for the NN. The performance of the proposed model is evaluated through function fitting, training‐state analysis, error histograms, regression analysis, and performance curves, providing a comprehensive assessment of its predictive accuracy and convergence characteristics. The results demonstrate that the governing flow parameters exert a significant influence on the flow, temperature, and concentration fields. In particular, the temperature distribution increases with increasing values of the Brownian‐motion parameter and the Nusselt number (Nt), whereas the velocity field is improved by increasing the melting parameter and Casson fluid parameter. Conversely, an increase in the thermal‐slip parameter suppresses the temperature distribution within the‐boundary‐layer. These findings highlight the value of the proposed AI‐assisted computational framework for accurately predicting the nonlinear transport behavior of complex nanofluid flow systems. As the Prandtl (Pr) number, Solutal slip parameter, and Nt grow, the concentration of flow drops. By comparing the results with previous research, the numerical code is validated and shown to be in good agreement.