had been removed). The material was rolled with
tween measurements). This required an adjust-
a steel roller to increase the adhesion of the mem-
ment of the procedure to minimize that weight
brane to the plate. The other technique used heat
loss/gain. We were able to tare out those losses
to seal the modified bitumen membrane to the
and gains.
plate. Again the sample was centered on the mask
Water losses and desiccant gains are to be ex-
in the mounting jig, the sample back was heated
pected when the cup is opened for weighing.
with a propane torch until the asphalt started to
Losses and gains can occur through the closed
melt, and then the steel plate was pressed down
seals of the apparatus. We suspect that our seals
onto the sample.
leaked as, in hindsight, we used an inappropriate
Once the bottom of a membrane was adhered
material (nylon) for the threaded rods (shown at
to the mounting plate, the top of the sample was
the corners of the cup assembly in Figure 15). Be-
coated with wax around its perimeter to a width
cause of this choice, we could not apply as much
equal to that of the mask. The mask was then
force to clamp the assembly together as should
heated for a several seconds and applied to the
have been used. Thus the moisture could leak in
wax on the sample. Finally the edge of the sample
or out through the machining grooves at the bot-
and mask was sealed with wax applied on in mul-
tom of the O-ring slot, along the surface of the
tiple layers. The sample application procedure re-
mating cup, or through scratches along the sur-
quires practice to become familiar with all the
face of the sample plate ring.
possible difficulties that can occur during the pro-
cedure.
Test results
Each membrane sample was used for two per-
Table 6 and Appendix E show the modified-cup
meance tests. The first test was the modified-cup
test results. The y-axis range on both sides of all
test with the desiccant above the sample and the
six graphs in Appendix E is 5 grains (since 1 gram
water below as shown in Figure 14. The second
= 15.432 grains, the y-axis range is only about a
water-cup test used only the water cup fastened
third of a gram). There are two measurements of
to the sample plate. Three membranes with a 0.06-
water vapor transmission and permeance for each
in.-diam. hole punched through them with a
membrane in the modified-cup test method. One
heated steel rod were also retested using the wa-
is the weight gain by the desiccant; the other is
ter-cup method.
the weight loss from the water. If there were no
leaks and the membrane absorbed no moisture
then the weight gain would equal the weight loss.
Test chamber and controls
An insulated walk-in box was used as the test
We suspect that our apparatus leaked or that our
chamber. A proportionally controlled electric
ability to tare out losses or gains was not adequate.
heater maintained the temperature in the cham-
Consequently, we used the smaller of the weight
ber. That temperature was kept at 84F plus or
gain of the desiccant or the weight loss of the wa-
minus about a degree. An open pan of water main-
ter to calculate the weight change per hour for the
tained the relative humidity of the chamber at
material, since that smaller value has the lower
about 16%, plus or minus about 2%.
extraneous loss or gain included. For low per-
meance materials, ASTM E96 specifies that the
reported permeance is the linear regression slope
Measurement problems
At the beginning of the tests, we discovered that
of the weight change vs. time divided by the
there was a significant variation in successive
sample area and the average pressure difference.
weights of individual desiccant and water cups
The third column of data in Table 6 shows the
due to static electricity created by handling of the
weight change per hour used to calculate the per-
cups. We finally solved this problem by isolating
meance (linear regression slope of the lines in
the sample far enough from the test scale with a
App. E). Dividing the weight change per hour
piece of wood, so that the attraction between the
by the sample area yields the water vapor trans-
scale's case and the cups due to the static electric-
mission (WVT). Dividing the WVT by the pres-
ity was minimized.
sure difference across the sample yields the per-
When we began making test weights in the
meance. Table 7 and Appendix F show the more
modified test procedure, we discovered that there
conventional water-cup test results. The y-axis
was a significant amount of weight change while
range of all six graphs in Appendix F is 5 grains.
we were making and recording the weights (about
Again the slope method or "numerical analysis"
the same magnitude as the weight changes be-
was used to determine the weight change per hour
18
Previous Page
