Technical information

3-21
3.4.2 Identification of Dataset for Structural Investigation
In general terms, this research aims at devising a method for applying APT results to
field conditions. As put forward in Chapter 1: (i) the scope is limited to the case of
duplicate pavement systems; (ii) the work plan consists of calibrating a mechanistic
model to APT conditions and extending it using laboratory data; and (iii) the extended
model is to be validated using NCAT results. Accordingly, it is argued that Section n1
dataset, collected in the APT between July and August 2004 (see Table 3.4.1), is best
suited for carrying out structural investigation and achieving the main study objective.
First and foremost, there is maximum similarity between this section and its replicate at
NCAT, especially in the initial part of the experiment when both pavements were in
their pristine state. This similarity was severely ‘damaged’ during the second round of
n1 testing that took place between August 2005 and April 2006 (see Table 3.4.3),
mainly because of the ‘artificial’ subgrade weakening and also because n1 became a
rehabilitated structure: the surface HMA was replaced after incurring about 90,000 load
passes and then the structure was allowed to rest/heal without traffic for almost a year
before more loads were applied. Second, in this dataset APT passes were applied
without wander, which means that the exact carriage position relative to the embedded
gauges is known and available. This information is critical for model calibration.
Finally, the construction operations are well documented and supplemented by
laboratory tests.
This state of affairs is not the case for Section n2. First, the original surface
HMA lift was replaced but the construction data and properties of the new mix are not
available. Second, loading of n2 between September 2004 and February 2005 included
wheel wander for which the exact carriage position was recorded only in the loading
direction but not laterally. When the exact position of the loading is unknown, the
approach followed herein fails because the APT model cannot be calibrated using
inverse analysis. Finally, the main difference between sections n1 and n2 (or
equivalently between N1 and N2) is the binder type. However, the master curves for
n2/N2 mixes 3 and 4 were very similar to the n1/N1 master curves for mixes 1 and 2
(respectively). Hence, the resilient response of the two sections should also be very
similar. In this connection, see also the FWD results in Figure 3.3.8. Moreover, as will