Natural convection immersion cooling of an array of vertically oriented heated protrusions in an enclosure filled with a dielectric liquid: effects of enclosure width, Prandtl number and component orientation

Abstract

The natural convection heat transfer characteristics of a 3X3 array of vertical oriented heated protrusions, immersed in a dielectric liquid, were investigated. Aluminum blocks, 24mm x 8mm x 6mm, were used to simulate 20 dual in-line packages. Surface temperature measurements of the components were made by embedding copper-constantan thermocouples below the surface of each component face. A constant heat flux was provided to each component using an Inconel foil heating element. Power supplied to each component varied from0.115W to 2.90W. The aluminum blocks were mounted on a plexiglass enclosure with inner dimensions:I = 203.2mm II= 152.0mm W = 82.6mm, and a wall thickness of 24.5mm. The upper boundary was maintained at 10(o)C, while all other exterior surfaces were insulated. The chamber width, measured from the surface of the circuit board to the opposite, inner wall of the enclosure, was varied from 42mm to 7mm by inserting plexiglass spacers into the enclosure. Two dielectric liquids, FC-75 and FC-43, were used as working fluids. Non-dimensional data from this study was combined with the data obtained by Aytar (1991) for a horizontal component orientation, to develop an empirical correlation which predicts the Nusselt number as a function of Rayleigh number, Prandtl number, component orientation and chamber width. This correlation was found to be accurate to within 11% of the original curve fit data. Heat transfer in FC-75 was found to occur mainly by convection arising from buoyancy forces, regardless of chamber width. Heat transfer in FC-43 was found to occur mainly by molecular diffusion for chamber widths of 11mm or greater, and by convection at a chamber width of 7mm. The maximum uncertainty in the Nusselt and Rayleigh numbers was 2.5%, based on a zeroth order uncertainty analysis, and occurred at the lowest power level, where the maximum uncertainty in the temperature measurement resided.