Influence of convection within and through perlite (a loose-fill insulant) on its apparent thermal conductivity from 20°C to 45°C

Abstract

Inclination of an enclosed, leaktight rectangular, 910mm × 316 mm, 10 mm thick layer of perlite, led to the maximum rate of heat transfer through it occurring with the layer at an angle of 75° to the horizontal. The apparent thermal conductivity of a horizontal layer of perlite, when heated from below, depends upon the velocity of any air draught blowing over the insulant's upper surface as well as upon the permeability of that upper surface. A horizontal air draught of 1·5 ms −1, at 100 mm above the top of a 100 mm thick horizontal layer of perlite (e.g. as could occur in lofts), resulted in an increase exceeding 20 per cent in the rate of heat transmission for a 5°C per centimetre temperature gradient through the insulant. Suppression of convection out of an otherwise open-topped 100 mm thick horizontal layer of perlite (above which no forced convection ensued) by means of the presence of building paper achieved a reduction of about 8 per cent in the rate of heat transfer through the insulant. The present investigation thus emphasises that such layers should be employed above insulant layers to inhibit convection losses, provided interstitial condensation will then not occur within the insulant layer. Leakage of air into the insulant layer from below (as occurs through ceilings beneath attic spaces) enhances buoyancy driven convection, resulting in increased rates of heat loss and hence higher apparent conductivities. Surprisingly, an increase in the thickness of the horizontal perlite layer with an open top, also leads to a significant rise in the mean apparent thermal conductivity of the insulant. Such conclusions cast doubts on the worthwhileness of many ‘economic thickness’ predictions made assuming a commercially supplied value (obtained under no internal convection conditions) for the insulant's apparent conductivity, and its invariance with the thickness of the insulant.