Abstract
•Mixed convection in a 3D partitioned cavity with a rotating cylinder was studied.•Considerable changes in 2D local Nu are obtained with variation of Ri and ω.•Adding CNT particles increased the Nu by about 82% and 90% for the highest φ.•Location of cylinder and conductive partition parameters affect the heat transfer.•Interpolation with POD was used for forecasting of average Nu.
In this study, convective heat transfer features of CNT-water nanofluid in a three dimensional cubic enclosure which is separated by a conductive partition and having an inner rotating adiabatic circular cylinder were examined. Numerical simulations were performed with Galerkin weighted residual finite element method. Various values of Richardson number of the left cavity (between 0.01 and 100), angular rotational speed of the inner cylinder (between −34724 and 34724), vertical (between 0.25 and 0.75) and horizontal (between 0.1 and 0.3) locations of the cylinder, thickness (between 0.01 and 0.2) and thermal conductivity ratio (between 0.1 and 100) of the conductive partition and nanoparticle volume fraction (between 0 and 0.03) on the fluid flow and heat transfer were analyzed. It was observed that the average heat transfer enhances for higher values of Richardson number, angular rotational speed of the cylinder (generally in both directions), solid particle volume fraction of nanofluid and thermal conductivity of the partition. The rate of enhancement are in the range of 11–14% when configurations with the lowest and highest values of Richardson number are considered whereas 18% of enhancement is attained at the highest rotational speed of the cylinder as compared to motionless cylinder configuration. The enhancements in the average Nusselt numbers with the inclusion of the CNT-nanoparticle at the highest solid volume fraction are in the range of 82% and 90% as compared to water. The average heat transfer coefficient is estimated by using POD and interpolation among POD modes and modal coefficients which is found to be computationally inexpensive and the method produces accurate results as compared to high fidelity three dimensional intensive computational fluid dynamics simulations.