Machine learning based evaluation of thermal signature and slip flow dynamics in a lubricated vertical duct exposed to solar energy-induced heating

The aim of this paper is to present a thorough investigation into the impact of an external magnetic field on the flow and thermal properties of nanofluid within a vertically oriented square duct exposed to sunlight-induced heating. The duct incorporates two rectangular fins attached to one set of adjacent walls, receiving uniform sunlight intensity. In contrast, the other two walls feature a thin lubricant film which causes the fluid slip rather than adhesion. Assuming hydrodynamically and thermally developed laminar flow, the study employs a specific thermal boundary condition, enforcing a constant heat flux per unit axial length and a uniform peripheral temperature at the sun-exposed duct walls. Considering the balance between computational efficiency and solution accuracy, the finite volume method is employed to solve the governing partial differential equations. It has been observed that the magnetic parameter induces magnetic damping, resulting in a reduction in velocity distribution and qualitative alterations in flow patterns within the duct due to the heightened impact of the applied magnetic field. The presence of taller fins acts as significant obstacles which divert the flow and cause additional frictional losses, resulting in an overall reduction in flow speed. Elevated Rayleigh numbers amplify natural convection currents, producing a more pronounced change in velocity distribution while qualitatively transforming the thermal signature within the duct. Further, the slip parameter increases the overall velocity distribution while exerting a minimal impact on the thermal distribution, due to a reduction in resistance along the walls with lubrication, facilitating smoother fluid motion and leading to a more widespread velocity profile throughout the duct. Finally, to bring a contemporary perspective to the research methodology, modern techniques such as machine learning were integrated. The widely adopted Neural Fitting app was employed to enhance the precision of the analysis.