Comparing Test Results for Coarse Aggregate
Pairwise comparisons among different levels of material and
test method were conducted using Tukey’s tests. A summary of
Tukey’s test results for coarse aggregate materials is presented in
Table 3-12. The “difference of means” values shown in Table 3-12
indicate that the Phunque method yielded higher Gsa, Gsb,
and Gssd results and lower water absorption values than the
AASHTO T 85 method. Based on a significance level of 0.05,
the differences in the test results between the Phunque flask and
AASHTO T 85 methods were statistically significant for each of
the coarse aggregate materials tested in this study, as shown in
the “significant” column of Table 3-12. These differences can be
seen in Figure 3-7, which presents graphical comparisons of the
average Gsa, Gsb, Gssd, and water absorption results for the five
coarse aggregate materials tested in Experiment
2.
The differences in Gsb and Gssd values for the two methods
were considered significant from a practical point of view. Part
of the differences can be attributed to how the two methods
account for absorption. In essence, the Phunque method
does not take into account the water absorption that takes
place from the time when the first aggregate particles enter
the water in the flask until the initial reading is made of the
water level in the neck of the flask. This may be as much as
30 seconds. Obviously, water is absorbed into the aggregate
particles during this time. From Figure 3-7(d), it can be
seen that the materials with the greatest differences in water
absorption were the recycled concrete and the blast furnace
slag. The aggregate particles for these materials have the largest
pore sizes and therefore are capable of absorbing water
more quickly than the other materials, resulting in lower
water absorption values. To correct the problem, timing for
the initial water level reading should be varied based on the
absorption of each material. However, it is difficult to shorten
the 30-second period for the initial reading because it would
take that much time to introduce all of the sample into the
flask, clean the neck, and take the reading. Further analysis of
the test results using a correlation between the water level and
time at which the water level reading is taken is conducted in
Experiment 5 to estimate the water level prior to 30 seconds.