Machine Learning-Assisted Sensitivity Evaluation for Heat Transfer Performance of Inclined Heat Transfer Systems

Abstract

Many technical systems rely on efficient temperature regulation. These systems include thermal shields for aircraft, metal forming, microelectronics, solar energy collectors, nuclear cooling units, polymer extrusion, and nuclear power plants. It is common for these technologies to function in environments with very hot and uneven surfaces, which calls for fluids with exceptional performance. To address this challenge, the heat transfer and nonlinear flow characteristics of a hybrid nanofluid, which includes water-dispersed copper (Cu) nanoparticles and aluminium oxide (Al2O 3 ) nanoparticles are investigated over an inclined stretching/shrinking sheet. The combined impacts of heat source and thermal radiation are considered in the study. The physical model has highly coupled, nonlinear governing equations due to the incorporation of nonlinear surface motion, inclination-induced gravity effects, and hybrid nanoparticle interactions. We transform them into a system of MATLAB-solvable nonlinear ordinary differential equations by utilizing similarity variables. There is also strong agreement when results from certain limiting circumstances are compared to previously published data. Several physical parameters control the problem, and the impact of these parameters on different flow distributions is studied extensively using both tabular and graphical methods. This study analyzed nonlinear hybrid nanofluid flow over an inclined stretching/shrinking surface influenced by magnetic forces, radiation, suction, and heat generation, with machine learning and Sensitivity analysis applied for predictive modeling. Furthermore, a global sensitivity analysis based on Sobol indices is performed to quantify the relative importance of governing parameters and their interaction effects on the skin-friction coefficient and local Nusselt number. The results identify the dominant parameters controlling momentum and heat transfer characteristics. Multiple Linear Regression (MLR) served as an efficient surrogate, achieving high accuracy (R 2 = 0.98 for C f , R 2 = 0.92 for N ux ) in estimating flow and thermal characteristics.