Experimental and numerical study of inverse natural convection-conduction problem in a fully partitioned cavity

The inverse three-dimensional (3D) computational fluid dynamics method combined with the least squares method and overdetermined temperature measurements on the partition is used to study the natural convection-conduction conjugate heat transfer problem in a fully partitioned cavity. An unknown heat transfer rate, an appropriate flow model, heat transfer coefficients on all solid surfaces, and heat transfer characteristics are predicted. The conditions for verifying the accuracy of the obtained estimates are as follows. The root mean square error (RMSE) obtained by an appropriate flow model needs to be smaller than that obtained by other flow models. The obtained Nusselt number (Nu_h) on the hot wall must also line up with current correlations. The results show that the zero-equation model and the RNG k–ε model are suitable for problems with a horizontal partition and a vertical partition, respectively. This implies that choosing a suitable flow model may be influenced by the partition site. The partition-induced obstruction and buoyancy-driven flow may interact to create a complicated flow field and temperature distribution. The RMSE value and the maximum error of Nu_h obtained by the appropriate flow model are about 0.45 and 4% for the horizontal partition and about 0.8 and 14.5% for the vertical partition, respectively. The obtained estimate of Nu_h is closer to the proposed correlation. The Q values for horizontal and vertical partitions in this study are predicted to be 4.05, 5.06, and 4.64 W, respectively. The heat loss can be as high as 43 and 33% for horizontal and vertical partitions, respectively. Few researchers have used this inverse CFD method to predict the present problem. A literature review found that these findings have not yet appeared in the open literature. Thus, this study is novel and original.