Reclaimed Asphalt Test Specimen Preparation Assisted by Image Analysis

A Venlo H-350/225 X-ray CT system (shaw inspection systems) with the IMPS operating software was used for capturing the two-dimensional (2D) images at the interval of 1 mm along the specimen height with a resolution of approximately $0.083\mathrm{mm}/\mathrm{pixel}$ (Figs. 2 and 3). The imaging software, ImageJ, was used to threshold the air void particles from X-ray images and the volume fraction was calculated using Eq. (1). The result was then validated through comparison with the one measured using the buoyancy test method. The thresholded images of the air voids were then analyzed for specimen’s homogeneity assessment.

At first, the homogeneity of the air void distribution was assessed using cores and ring analysis. In this analysis, the images were virtually cut into cores of increasing radius (25 mm) from the center of the specimen to the specimen’s edge and the rings of increasing radius (25 mm) from the edge to the center of the specimen. The air void distribution within the cores and rings were then analyzed. Subsequently, another analysis was conducted to quantify the uniformity of the air void distribution across the depth of the specimen in vertical and radial directions [Eqs. (2) and (3)]. The latter applies the concept of uniformity index (UI) introduced by Masad et al. (2009). For UI vertical the polynomial function, $f(x)$ is a function of air void percentage, $x$ over the specimen height limits $a$ and $b$. The fourth-order polynomial of the function $f(x)$ is obtained from a best fitting of the plot of the measured air voids versus specimen height. For UI radial the polynomial function, $f(r)$ is calculated based on a function of air voids over the radius, $r$. A UI value close to zero indicates a uniform air void distribution.

Under this investigation, a set of procedures consisting of three methods of gyratory compaction have been conducted for obtaining a targeted test specimen of $100\times 100\mathrm{mm}$ with homogeneous air void distribution. On the assumptions that eventual difficulties in the compaction are proportional to the amount of RA, only the V-mix and RA30-mix have been considered. The methods are summarized as follows:

Details of the compaction procedures B and C are illustrated in Fig. 4. The specimens were then X rayed for internal structure images and laboratory tested for air void content [Buoyancy test BS EN 12697-6:2012 (BSI 2012)]. Table 2 summarizes the specimen properties as a result of undertaking the three compaction procedures.

From Fig. 5, both virgin and RA specimens prepared with procedure A have higher air void content at the top and bottom sections compared with the middle section indicating nonuniformity in the air void distribution throughout the specimens. This agrees with the higher values obtained for the vertical uniformity index compared with the radial one. In fact, final trimming (at least 15 mm) is needed at both ends of the specimens due to uneven surfaces caused by the preset angle (0.82°) during compaction. As a result, procedure A allows obtaining a homogenous specimen but with an $\mathrm{H}/\mathrm{D}$ ratio much lower than 1, which is not what was targeted. On the other hand, as presented in Fig. 6, the RA30 specimen, produced with $\mathrm{H}/\mathrm{D}$ ratio higher than 1, shows more air voids at the middle of the specimen compared with the other sections. This shows that the slender specimen contains nonhomogenous air void distribution. However, when the specimen was produced using procedure C, in which the core of $D=100\mathrm{mm}$ was cored from a specimen with the original size of $D=150\mathrm{mm}$, a homogenous air void distribution can be observed throughout the specimen as shown in Fig. 7. It should be noted that high air void content observed on the first few slices on the top and bottom sections is caused by the cutting and trimming works.

The second investigation of this study was to obtain a $100\times 100\mathrm{mm}$ specimen with homogenous air void distribution at the designed air void content. Using the compaction method proposed in procedure C, a preliminary investigation was carried out by compacting a number of $150\times 150\mathrm{mm}$ specimens at different densities for optimization (100, 95 and 90% from the original target density at 2.6% target air void content). Three replicates were prepared for each target density.

Table 3 summarizes the average values of resulting densities and air void contents of specimen C before and after the coring work. The idea was to obtain a relationship between the measured air void content of the cores ($100\times 100\mathrm{mm}$), and the target density to be set on the gyratory compactor for manufacturing the $150\times 150\mathrm{mm}$ specimens. The target density for compacting $150\times 150\mathrm{mm}$ specimen to obtain a core ($100\times 100\mathrm{mm}$) with homogenous air void content of 2.6% was then estimated from the linear relationships in Fig. 8 with the $R$-squared ($R2$) values close to 1.

From the plot, the estimated target densities for all mixture types are as follows:

In order to ensure repeatability of the test results, 24 replicates for each mixture type were then prepared by using the estimated density values. The cores of $100\times 100\mathrm{mm}$ were tested for air void content and X rayed for image analysis. This was to confirm that the specimens meet the target air void content (2.6%) and the air voids were homogeneously distributed through the $100\times 100\mathrm{mm}$ core. Table 4 and Fig. 9 give the results of the buoyancy test and an image analysis of the air void distribution. Both results prove that the cores of $100\times 100\mathrm{mm}$ (for all mixture types) with homogeneous air void distribution at the target air void content can be achieved using the method established in procedure C. Fig. 9 shows that the air void distributions in virgin and reclaimed asphalt specimens are uniformly distributed from top to bottom with average values differing around $+0.5\%$ from the desired one. The small variation in the average values indicates good repeatability within the replicates. This is also agreed by the uniformity index values calculated in both vertical and radial directions with the values close to zero (Table 4). In addition, the virgin mix shows slightly less homogeneous air void distribution throughout the specimen than the RA-mix. This could be due to the downsizing process of the RA recovered from the asphalt pavement. In fact, fractionation of the reclaimed asphalt pavement involves crushing and screening, which can result in a reduction of the original size and shape of the aggregate particles (become rounded and uniform), this consequently affects the aggregate interlock in filling up the volume of the specimen during compaction.