An Analysis of Reclaimed Asphalt Pavement from a Single Source—Case Study: A Secondary Road in Romania

Abstract

The paper presents a comprehensive analysis on reclaimed asphalt pavement (RAP) milling material collected from a single source, namely from a secondary road in Romania, county road DJ109. The following characteristics are investigated: particle size, binder content, material variability and uniformity, and the clustering phenomena. Variability is demonstrated using the results of particle size gradation and binder content. The coefficient of uniformity and the coefficient of curvature demonstrate that the RAP used in this research is a well-graded material. However, the visual analyses conducted on RAP highlight the presence of RAP particle agglomeration and the need for further testing. The study presents three different experimental phases: (i) RAP-milling old asphalt pavement, RAP; (ii) RAP milling after binder extraction, RAPabe; and (iii) RAP after Los Angeles crushing, RAPac. After processing, the coarse part (C) had a great influence on the fine part (F), and F/C ratio increased, respectively, from 0.4 to 1.5 and 1.61. Material variability on the extended site, the difference between the design values and particle size, as well as the existing clustering process indicated that RAP material collected from secondary roads must be pre-processed prior to its storage and reuse in the recycling process. Considering that secondary roads represent 71% of the overall network of asphalt course roads in Romania, and around 24,000 km of roads are in need of at least extensive maintenance (wearing asphalt courses) or rehabilitation, RAP is a highly recyclable material. Therefore, this study provides advice and guidance for re-using RAP in new pavement mixtures.

1. Introduction

The quality of RAP in the production of hot-recycled asphalt in conventional hot-mix plants depends on several aspects: origin source, particle size, binder content, the variability and clustering phenomena. These aspects are analyzed individually in other scientific papers, while this study presents a comprehensive analysis of all above-mentioned aspects, which is a particularity of the applied methodology. RAP material performances are highlighted in previous researches/studies for both types of origin source, namely unique source or single project site [1,2,3,4,5], as well as multiple sources or project sites [6,7,8,9,10,11,12] collected using milling or other techniques or produced in laboratory after applying long-term aging techniques [13]. Thus, RAP obtained from unique or multiple sources presents significant inhomogeneity and is prepared with pre-processing techniques such as screening, sieving, crushing, storing separately according to its gradation stockpiles [14,15], or even mixing and storing in hybrid RAP stockpiles [16] suitable for further use in a new asphalt mixture. The aim of this research is to analyze and to characterize the properties of the RAP stockpiles collected by milling from a single source, a secondary Romanian road, county road DJ 109. This research is focused on a new category of roads, namely secondary arteries, different than the categories identified in the referenced research, such as highways and national and urban roads.

Secondary roads represent 71% of the overall network of asphalt course roads, which is 43,090 km [17]. Furthermore, 40% of modern Romanian roads have overdue service life [18], which represents a major source of waste that can be reused in new recycled mixtures [19,20] with great benefits for the environment [21,22,23]. This is similar to the overall situation in the European Community [24].

The novelty in this research is the case study that presents a specific area of research that is insufficiently analyzed and a type of road that has a big impact on the sustainability of road works considering the high quantity of RAP material available from secondary roads. This is a first attempt to study RAP characteristics from a secondary road located in Romania as an initial stage in the RAP-recycling process. The results of the material analysis provide information regarding RAP-milling material as a first study on the mineral skeleton.

The methodology developed and presented in this research is based on previous studies, but it provides an improvement as it analyses the properties of RAP in three different experimental phases: (i) before processing, RAP; (ii) after extraction, RAPabe; and (iii) after crushing, RAPac.

The grading properties were determined in all experimental phases. The results were used in comparison to the design values from present specifications referring to mixtures in the base layer (BAD 22.4) and subbase layers (AB 22.4 and AB 31.5) of road structures, according to AND 605/2016 [25]. RAP can be re-used in new pavement mixtures in a limited share and for certain gradation.

Aggregate gradation and particle distribution of the nine samples were statistically studied to obtain a generalization of the entire road sector. Considering the small set of data, t-test values were employed in determining the confidence interval. The results of the generalization are closer to the design values. However, other research states that no geographical or regionalization of the properties of the RAP stockpiles could be made [26]. RAP clustering is analyzed in this research mainly because it is a very important factor that affects the performances of design gradation of recycled asphalt mixtures [6,9,15,27,28,29,30]. Results of preceding research classify clusters according to the type of particle agglomeration structure into strong clusters, weak clusters, and old particle aggregate [29,31,32]. In this research, the clustering phenomenon was confirmed and quantified using two different fractions (≤4 mm and >4 mm), which were identified in all three different experimental phases. As a result of the two tests—after binder extraction and after crushing—the coarse part (C) (particle size > 4 mm) and the fine part (F) (particle size ≤ 4 mm) were analyzed against RAP using the F/C ratio. Previous research papers have identified clustering phenomena by applying a similar particle size limit and using conventional processes (RAP crushing) [8] or new methods (RAP microwave or oven heating) [33,34].

The paper is organized as follows: Section 2 presents the materials used in the research, with respect to the origin source, collection process, and sampling. Section 3 presents the test methods applied to the sample group in all experimental phases: (i) RAP-milling old asphalt pavement, (ii) RAP milling after binder extraction, and (iii) RAP after Los Angeles crushing. The following characteristics are investigated and presented in Section 4: particle size, binder content, material variability, and uniformity and the clustering phenomena. Finally, several conclusions are presented, and recommendations are highlighted.

2. Materials

The RAP used in the study was collected by milling the surface course as part of a major rehabilitation project of county road DJ109, with a total length of 11.19 km (km 20 + 786 ÷ km 31 + 976). According to the technical expertise carried out in 2016 [35], this is a flexible pavement consisting of 5 ÷ 10 cm surface course and 20 ÷ 30 cm sub-grade course. The thickness of the courses is not evenly distributed along the road. Figure 1 presents images of the old surface pavement. The pavement condition index (PCI) was evaluated at 12%, indicating a serious class. It correlates with an international roughness index (IRI) evaluated at 6.5 m/km. The following surface defects were observed: raveling, polished aggregates, and bleeding surface, which were covering 40% of the road. The following structural defects were observed: patching and potholes, longitudinal and transversal cracking, rutting, shoving, and corrugation, which were covering 25% of the road. The technical solution proposed for this road was milling of the surface course and replacement of the entire road structure.

The value of cumulative ESALs (Equivalent Single-Axle Road-115kN) obtained for this road sector was 200,000, indicating a T4 traffic category road.

The pavement sector is heterogeneous due to various pavement ages and the overdue life cycle, different damage states with multiple defects, and numerous maintenance interventions to improve pavement condition performed over time (filling cracks, patching, fog seals, slurry seals, bituminous surface treatments, and asphalt wearing course application). These characteristics of the sector have direct impact on the inhomogeneity of RAP. This is also demonstrated by the results of the binder content (See

Table 1. Binder content of RAP samples.

Binder Content	Sample 1	Sample 2	Sample 3	Sample 4	Sample 5	Sample 6	Sample 7	Sample 8	Sample 9
Binder content (%)	5.29	3.55	4.79	5.16	5.09	7.44	3.68	5.33	4.51

).

Binder content is considered an important index in the evaluation of RAP material performances, and the results are comparable to similar research [5,36]. More detailed research investigates binder content on different particle sizes and proves that fine aggregate micelles adsorb asphalt in a large amount, indicating the adsorption ratio values [6,37].

A cold-milling machine, Wirtgen W 2000, was used in the collection process to provide RAP. The moving speed of the milling machine during the collection process was 10 m/min, and the milling depth was 5 cm of the original upper layer. The material was deposited in 9 conical piles, each being obtained from a segment of approximately 1 km. The samples were collected from each stock using the sampling method specified in SR EN 932-1. This resulted in 9 boxes of approximatively 7 kg each, corresponding approximately to one sample per km.

3. Test Methods

The methodology used is presented in Figure 2. It improves the similar methodologies by investigating multiple characteristics of RAP in three different experimental phases. RAP material is characterized primarily with respect to aggregate gradation, which supports the comparison to the design values of the base/subbase asphaltic layers. The results are used to evaluate deviation from the design values in order to highlight the variability of the samples collected from a single source.

The study presents three different experimental phases conducted and presented individually and in comparison, namely: (i) RAP-milling old asphalt pavement, RAP milling processed, (ii) RAP milling after binder extraction, and (iii) RAP after Los Angeles crushing.

The sample group was collected from each stock according to SR EN 932-1.

The sieving experiment was conducted on RAP materials in all three different experimental phases, both before processing and after processing with bitumen-extraction test and fragmentation test (Los Angeles crushing) to highlight the percentages retained on the following sieves: 45 mm; 31.5 mm; 22.4 mm; 16 mm; 11.2 mm; 8 mm; 4 mm; 2 mm; 1 mm; 0.125 mm; 0.063 mm.

Clustering in RAP material represents the agglomeration of RAP particles observed after milling. These clusters contain different size aggregates, fine particles, and binder. Several studies classify the clusters into different types based on clustering particle size, binder content, and agglomeration mode. The main cluster types were identified in this study as well: weak RAP (within the black circle in Figure 3), strong RAP (within the grey circle in Figure 3), and RAP aggregate particle (within the white circle in Figure 3).

The Los Angeles abrasion test is used to simulate the breaking process of the aggregates in order to demonstrate the dislocation of the clusters. The crushing test was conducted at room temperature using a rotation number of 500r. RAP was fractioned in laboratory in gradations similar to the design values of virgin aggregate particles for conventional mixtures BAD 22.4, AB 22.4, and AB 31.5.

The bitumen-extraction test using the centrifugal separation method is performed to investigate the cluster phenomena in RAP materials as well as to determine the bitumen content.

The mean values of percent passing were determined for each sieve, and an estimate of the values of this parameter was calculated for the entire road sector using a confidence interval. T distribution was used considering the small data set and the fact that the variance for the entire road sector is unknown. The formula for the confidence interval for the mean is:

$$\overline{\mathrm{x}}\pm {\mathrm{t}}_{\mathrm{n}-1,\mathsf{\alpha }/2}\frac{\mathrm{s}}{\surd \mathrm{n}},$$

(1)

$$\overline{\mathrm{x}}=\frac{{{\displaystyle \sum }}_{\mathrm{i}=1}^{\mathrm{n}}{\mathrm{x}}_{\mathrm{i}}}{\mathrm{n}},$$

(2)

$$\mathrm{s}=\sqrt{\frac{{{\displaystyle \sum }}_{\mathrm{i}=1}^{\mathrm{n}}{\left({\mathrm{x}}_{\mathrm{i}}-\overline{\mathrm{x}}\right)}^2}{\mathrm{n}-1}},$$

(3)

where: $\overline{\mathrm{x}}$—Sample mean;

${\mathrm{x}}_{\mathrm{i}}$—Percent passing;

n—Sample size;

${\mathrm{t}}_{\mathrm{n}-1,\mathsf{\alpha }/2}$—Reliability factor;

n − 1—Degrees of freedom;

α—Level of significance;

α = 100%—Level of confidence;

s—Standard deviation for the sample.

When generating the confidence intervals, the value of t-statistic, for a degree of freedom equal to eight and a confidence level of 95%, was obtained from the T-table.

4. Results and Discussion

4.1. Properties of RAP Milling before Processing

The RAP aggregate fraction was characterized by using aggregate particle size distribution according to the Romanian Standard AND 605/2016. According to AND 605/2016, the sieve fraction is 0/0.063, 0.063/0.125, 0.125/1, 1/2, 2/4, 4/8, 8/11.2, 11.2/16, 16/22.4, 22.4/31.5, and 31.5/45 mm. The results are presented in contrast to the design gradation of the base asphaltic course BAD 22.4 (Figure 4a) and subbase asphaltic course AB 22.4 (Figure 4b) and AB 31.5 (Figure 4c), respectively.

RAP particle size deviations compared to the design values indicate the variability of the material (Figure 4d).

The mean, the standard deviation, and the coefficient of variance of the nine samples were determined for each sieve size. Except for the 4 mm and 8 mm sieves, which show a relative homogeneity based on the coefficient of variance value between 10% and 20%, the results for the other sieves show significant variability for the retained percent.

Variability of RAP properties results also in comparison to the minimum binder content according to Romanian standard AND 605: 4.2% for BAD 22.4 base layer, 4% for AB 22.4 and AB 31.5 sub-base layer, and 5.7% for BA16 wearing course. The values presented in

Table 2. RAP sieving matrix—retained percent (%).

Sieve (mm)	31.5	22.4	16	11.2	8	4	2	1	0.125	0.063	Avg
Sample 1	8.98	5.62	9.99	13.27	12.01	22.05	11.92	7.21	8.19	0.44	-
Sample 2	0.68	1.78	12.70	16.33	15.22	24.42	12.67	7.54	7.55	0.73	-
Sample 3	1.42	2.91	10.33	18.49	14.99	25.41	11.40	6.51	7.72	0.53	-
Sample 4	5.07	1.59	6.52	13.82	12.72	26.81	17.10	10.93	5.33	0.08	-
Sample 5	3.21	5.94	11.81	18.18	14.56	21.10	12.02	7.15	5.86	0.11	-
Sample 6	5.10	0.36	0.91	7.95	9.77	27.34	18.11	12.77	17.08	0.57	-
Sample 7	0.00	5.78	20.07	19.29	15.61	21.71	9.50	4.95	2.81	0.14	-
Sample 8	1.64	2.47	5.22	13.48	12.86	27.46	16.41	10.21	9.90	0.29	-
Sample 9	1.75	3.13	8.18	11.80	11.35	23.94	16.31	11.37	11.73	0.33	-
Mean	3.10	3.28	9.52	14.74	13.23	24.47	13.94	8.74	8.47	0.36	9.08
St dev	2.84	2.04	5.37	3.68	2.00	2.47	3.05	2.64	4.14	0.23	2.59
Cv (%)	92	62	56	25	15	10	22	30	49	63	42

show that the samples’ binder content varies from 3.55% to 7.44%.

Figure 5 highlights the greatest heterogeneity related to sieve size 16 mm and 0.125 mm.

The characterization of RAP aggregate uniformity uses the coefficient of uniformity (Cu) and the coefficient of curvature (Cc) to highlight grading characteristics. Cu is determined by the D60 to D10 ratio. D60 and D10 values represent the sieve opening size (mm) through which 60% and 10%, respectively, of the aggregate passes. Cc is determined by the ratio of squared value of D30 to D60 multiplied by D10. D60, D30, and D10 values represent the sieve opening size (mm) through which, respectively, 60%, 30%, and 10% of the aggregate passes. The results of Cu for all samples are higher than 5, which shows that RAP is a well-graded aggregate. This fact is also demonstrated by the results of Cc, which vary between 1 and 2. The values of the coefficients of uniformity and curvature presented in

Table 3. RAP uniformity results.

Uniformity	Sample 1	Sample 2	Sample 3	Sample 4	Sample 5	Sample 6	Sample 7	Sample 8	Sample 9
D60	10.63	9.41	9.74	7.96	11.01	5.67	12.48	7.37	7.37
D30	4.35	4.19	4.56	3.59	4.91	1.96	6.30	3.16	2.79
D10	1.14	1.18	1.22	1.42	1.55	0.60	2.41	0.98	0.84
Cu	9.3	8.0	8.0	5.6	7.1	9.4	5.2	7.5	8.8
Cc	2	2	2	1	1	1	1	1	1

are comparable to gradation characteristics in similar research [7]. Ref. [38] expanded the research of Cu in correlation to the characteristic particle size for two experimental stages before processing and after binder extraction.

Following the characterization of the RAP samples’ variability and uniformity, the variability of the entire road sector was obtained using statistical analyses of the small set of data. For this purpose, t-test values were employed to estimate the confidence interval for this statistical model (see

Table 4. Statistical values of percent passing (%) of the statistical model for the entire sector.

Sieve (mm)	31.5	22.4	16	11.2	8	4	2	1	0.125	0.063
Mean	96.90	93.60	84.08	69.36	56.12	31.67	17.73	8.99	0.52	0.17
Variance	8.13	12.94	43.82	89.42	122.42	84.49	41.58	18.09	0.11	0.02
St dev	2.85	3.60	6.62	9.46	11.06	9.19	6.45	4.25	0.33	0.13
Cv (%)	3	4	8	14	20	29	36	47	63	79
St error	0.95	1.20	2.21	3.15	3.69	3.06	2.15	1.42	0.11	0.04
T-stat 95%	2.31	2.31	2.31	2.31	2.31	2.31	2.31	2.31	2.31	2.31
Confidence interval (%)	94.70	90.83	78.98	62.07	47.60	24.59	12.77	5.71	0.27	0.06
	99.10	96.37	89.18	76.64	64.64	38.74	22.70	12.26	0.77	0.27

).

The results in the statistical model are further employed to obtain the gradation model for the entire sector.

The results of the gradation curves model presented in Figure 6 show that statistically generalized values of the estimated percent passing are closer to the limits of the design values for all analyzed road layers.

These results indicate that the variability generalized at the level of the entire sector is not as obvious as in the individual samples.

4.2. Properties of RAP Milling after Binder Extraction

A sieving analysis was conducted on RAPabe, the material obtained after binder extraction. The results are presented in contrast to the design gradation of the base asphaltic course BAD 22.4 (Figure 7a) and subbase asphaltic course AB 22.4 (Figure 7b) and AB 31.5 (Figure 7c), respectively. The results show that RAPabe does not have appropriate gradation compared to design value gradations of the virgin aggregate. Most of the values exceed the upper limit of design values.

Deviations of the particle size of RAPabe in comparison to the design values indicate the variability of the material (Figure 7d). The mineral skeleton analysis of RAPabe highlights the particle migration between the sieves, which is an indicator of the presence of cluster phenomenon, analyzed in Section 4.4.

4.3. Properties of RAP Milling after Crushing

A parallel test, Los Angeles crushing, was conducted. The sieving analysis was performed on RAPac. The results are presented in contrast to the design gradation of the base asphaltic course BAD 22.4 (Figure 8a) and subbase asphaltic course AB 22.4 (Figure 8b) and AB 31.5 (Figure 8c), respectively. The results show that RAPac does not have appropriate gradation compared to the gradation of the design values of the virgin aggregate. Most of the values exceed the upper limit of design values.

Deviations of the particle size of RAPac in comparison to design values indicate the variability of the material (Figure 8d). The mineral skeleton analysis of RAPac highlights the particle migration between the sieves, which is an indicator of the presence of cluster phenomenon, analyzed in Section 4.4.

4.4. Cluster Phenomenon

The result of processing tests, the bitumen-extraction test, and fragmentation test presented in Figure 7 and Figure 8 were further used to investigate the cluster phenomenon in the RAP material collected from the secondary road DJ 109. Clustering in RAP is an issue of instability of the aggregate skeleton in case RAP is proposed for further use in a new recycled material for an asphalt layer.

The RAP particle agglomeration phenomenon is indicated by the presence of particle sizes greater than the ones specified by the standard in RAP-milling samples, such as particle size greater than 22.4 mm for BAD22,4 and AB22,4, in share of 3%.

The clustering phenomenon in RAP was highlighted in this study using two processes: binder extraction and Los Angeles crushing. For all three practical stages, the separation in two different fractions (fine particles with size smaller than 4 mm and aggregate coarse with particle size greater than 4 mm) is presented in

Table 5. F/C Ratio matrix.

	Fraction (mm)	(i) before Processing (%)	(ii) after Binder Extraction (%)	(iii) after Crushing (%)
Course part (C)	>4	68	40	38
Fine part (F)	≤4	32	60	62
F/C Ratio		0.46	1.5	1.61

.

The particle size of 4 mm and above is used to highlight the disaggregation process. The results show that the clustering phenomena are not eliminated by simple sieving separation. Both processing methods have a great impact on the gradation of the milling material. The results obtained in the before-processing phase are comparable to [39], which reported 59% course aggregates, 31% fine aggregates, and 5% mineral filler. The percentage of particles with diameters smaller than 4 mm increases from 32% in RAP milling before processing to 60% in RAP milling after binder extraction and respectively, to 62% in RAP milling after crushing. The F/C ratio is higher than 1, indicating the clustering phenomena.

In all three experimental phases, the particle migration between the sieves is also highlighted in Figure 9.

The presence of particle agglomeration in RAP may affect design gradation. RAP processing results presented in Figure 10 demonstrate disaggregation.

5. Conclusions

This paper is a comprehensive study on RAP-milling material collected in a single project from a secondary road in Romania, county road DJ 109. The novelty in this research is the case study that presents a specific area of research that is insufficiently analyzed and a type of road that has a big impact on the sustainability of road works. The results of this analysis could not be presented in a detailed comparison to the results of similar studies due to several differences regarding the collection process type, the multisource samples, and the different category of roads, such as highways, national, and urban roads.

The data from this study demonstrate the great variability of the mineral skeleton of the analyzed sample group.

RAP variability may negatively affect the recycled asphalt mixtures when it is re-used in new mixtures. According to the findings in this paper, the properties of RAP stockpiles collected from this single jobsite vary greatly. Therefore, reclaimed asphalt pavement stockpiles collected from secondary roads need pre-processing prior to storage and re-use.

The coefficient of uniformity and the coefficient of curvature demonstrate that the RAP used in this research is a well-graded material. However, the visual analyses conducted on RAP highlighted the presence of RAP particle agglomeration and the need of further testing. The binder content varied due to the heterogeneity of the old surface pavement in the analyzed sector.

The cluster phenomena were identified in the RAP material using two processing tests: the bitumen-extraction test and the fragmentation test. The ratio of fine particle against aggregate course is higher than 1 in both phases (ii) after extraction and (iii) after crushing, while it is lower than 1 for the RAP material (i) before processing. The results indicated an important particle migration between sieves for the two processes, namely from the course part (sieve size above 4 mm) to the fine part (sieve size under 4 mm).

In Romania, the recycling of asphalt pavement materials in a plant is conducted according to the specification DD 509-2003 [40], which allows a substitution rate of maximum 30% of RAP exclusively in base layers and subbase layers. For this reason, RAP gradation, white gradation (RAPabe), and crushed RAP (RAPac) were analyzed in contrast to the design grading of base layer BAD 22.4 and subbase layers AB 22.4 and AB 31.5. The results indicated a significant material variability. Binder content of the samples was also used to demonstrate the variability of the RAP material.

The results of this research can be an asset for the road administration and road contractors to enhance their knowledge regarding RAP characteristics.

Further research on RAP material collected in the rehabilitation processes conducted on secondary roads in Romania will be considered using multiple sources or project sites based on material availability as a real opportunity of sustainable road works, considering that around 24,000 km of roads are in need of at least extensive maintenance (wearing asphalt courses) or rehabilitation.