where

For the range of Rayleigh numbers given in Table 7, this equation gives Nusselt numbers in the range of 4.75 to 7.45.

The qualitative agreement of the local Nusselt numbers of the present study with is good. Discrepancies are possible in the average Nusselt number since these authors use thermocouples to determine the local Nusslet number. Resulting in loss of resolution, especially near the peak values. It is clear that the flush heater correlations substantially underpredict the average Nusselt number. As discussed by Park and Bergles [105], this discrepancy is due to the inapplicability of large plate correlations for short segments, especially near the leading edge. Conduction losses to the supporting plate are significant for small surfaces and lead to a higher measured Nusselt number. The cuboid model also underpredicts the Nussselt number because it is primarily designed for a fin assembly and not an isolated copper block. The Nusselt number for a block located in an array is smaller than for a single block since the fluid approaching the block is preheated and the thermal boundary-layer over the vertical face is thick. For the isolated heated surface considered in the present work, the boundary- layer thickness (Figure 4.38) is zero at the leading edge and at all Rayleigh numbers except the lowest value.

Closure

Heat transfer in natural convection from an isolated protruding heater is found to be larger than that computed from correlations for a flush heater, a large cuboid, and data for an array of blocks. Interferometric study shoes the local Nusselt number over the vertical face of the copper block to be quite large. This factor along with conduction to the supporting plate in the direction provides the reasons for this difference.