b'TECHNICAL & RESEARCH7. Microscopy. One quadrant of each sample wasmembranes decreased with age, while the cut through the impact site to view the membraneexposed bitumen area generally increased with and cover board in cross-section with the use of amembrane age. There was a larger increase in microscope (Figures 9 and 10).exposed bitumen area between five- and ten-year-old membranes than there was between new and five-year-old membranes (Figure 11). Figure 11. Comparison of new, five, and ten-year-old membranes over gypsum cover board. Figure 9. Cross-section ofFigure 10. Cross-section For each membrane age, the simulated hail 10-year-old gypsum cover boardof new wood fiber cover sample. Note separation of facerboard sample. Note that noimpact resulted in minor granule loss that yielded from gypsum (yellow arrow)separations are visible.slightly more exposed bitumen. The exposed and radial crack throughoutbitumen area generally increased more upon thickness of gypsum (orangeimpact with gypsum cover board than it did with arrow). wood fiber cover board. Results of Impact TestingThe impacts did not fracture the membrane The following table describes the average resultsreinforcement in any of the samples.of the impact testing in terms of granule counts and exposed bitumen (expressed as a percentage of Cracking of the gypsum cover board was observed area) within the 3-inch diameter impact zone forat the impact site in all gypsum cover board samples each type of sample tested: that were tested. The impacts also caused cover board facer separation from the gypsum core at Table 1. Average manual granule counts and exposedthe impact sites in two thirds of the gypsum cover asphalt area pre- and post-impact board samples. The wood fiber cover board samples were visually unaffected by the impacts.Comparison between image analysis and manual granule counts revealed that image analysis accuracy is variable due to differences in granule shape, color, texture, and size as well as features like granule overlap or fracturing. For this reason, manual counts were found to be more accurate and ultimately what was relied upon for Conclusions the study. However, image analysis was found to be more useful in identifying The following conclusions were drawn from theexposed bitumen area due to the high results of the impact testing and analysis: color contrast between the granulesIn general, regardless of cover board typeand the bitumen, a property that image or impact, the quantity of granules on theanalysis is proficient in detecting.'