Numerical and experimental study of natural convection in tilted parallelepipedic cavities for large Rayleigh numbers

Steady-state natural convection taking place in rectangular cavities filled with air is studied both experimental and numerically in the present work. The active walls, hot and cold, of the cavity are maintained isothermal at temperatures Th and Tc, respectively, and the other walls (channel) that close the cavity are adiabatic. Different angles of inclination α of the cavity from 0° to 360° are considered. This includes the analysis of several significant situations corresponding to inclinations of 0° (vertical active walls), 90° (hot wall down, Rayleigh–Bénard convection) and 270° (hot wall up, pure conductive mode). Two aspect ratios A = L/H = 0.75 and 1.5 are treated, being L the distance between the active walls and H its height. The numerical study is carried out by means of the finite volume method and provides the thermal and dynamic maps of the fluid for several geometrical configurations obtained while varying α, A and ΔT = Th − Tc. The range of the Rayleigh number, RaL, covered in our work extends from 10 to 108, allowing to treat numerous applications in engineering.The convective exchanges calculated by CFD are, for most of the treated cases, close to those obtained experimentally by means of a bench specifically designed for this purpose. The overall differences are relatively small, with an average discrepancy between calculations and measurements of about 6%, corresponding to the expected uncertainty of the model and measurements. Part of our results overlap with previous studies and the agreement with them is satisfactory. We provide new correlations of the type NuL–RaL which are useful for the sizing of structures based on this type of cavities.This numerical and experimental work comes to complete the numerous numerical studies found in the literature on this matter, often not accompanied by experimental measurements considering the difficulties to develop adequate assemblies, in particular for the broad range of RaL treated in this work. An estimation of errors has been performed in order to evaluate the accuracy of the experimental results.