Permanent Deformation Assessment of Cold Recycled Asphalt Mixtures

e objective of this study was to evaluate the behavior in terms of the permanent deformation of four cold recycled asphalt mixtures stabilized with asphalt emulsion and foamed asphalt. e analysis was conducted considering three ow number test procedures adapted for cold recycled mixtures: (i) ABNT NBR 16505 (method A); (ii) adaptation of the FN test for CRAM according to Kim and Lee’s method (method B); (iii) 3D-Move analysis software simulation of pavement structures to address the stress magnitude for the FN test (method C). e inuence of the curing process on the performance of the materials was also veried. e study used an extensive database of computed pavement responses from the four dierent asphalt pavement structures of the experimental section subjected to axial loading of a tandem axle using the 3D-Move pavement response analysis program. e FN results using the 3D-Move data (method C) demonstrated the good mechanical behaviour with respect to permanent deformation of the cold recycled mixtures subject to axial stress and temperature. e method proposed by Kim and Lee (method B) showed little ecacy for the analysis of the CRAMs tested in this work. e ABNT NBR 16505 method (method A), conventionally used for HMA, highlighted the inuence of temperature (60°C) on the CRAM mixtures. For all FN test conditions, the curing was a preponderant factor for the performance of cold recycled asphalt mixtures and should be carefully evaluated to simulate the early stage of the mixtures in the eld. It is also important to verify the inuence of humidity and temperature on the permanent deformation mechanism of cold recycled asphalt mixtures.


Introduction
Cold recycling is an increasing trend in rehabilitation, maintenance, and intervention of distressed asphalt pavements.
e main advantages of this technique are the economic and environmental bene ts allowing faster interventions and shorter tra c disruption [1,2]. In the recycling process, reclaimed asphalt pavement (RAP), usually generated from pavement milling [3], is incorporated to a new mixture through asphalt binder addition (stabilization). Virgin aggregates can also be incorporated to adjust the mixture's aggregate gradation [4][5][6].
Cold recycled asphalt mixture (CRAM) is usually stabilized with foamed asphalt or asphalt emulsion. Active llers may also be added (portland cement, hydrated lime, and others) to improve some mechanical properties, such as moisture resistance, and to increase early age resistance [5][6][7][8]. CRAM's mechanical behaviour is dependent on various factors, such as the curing process, reclaimed asphalt pavement (RAP) source and gradation, moisture content, type and content of asphalt stabilization agent (foam or emulsion), presence of active llers, and volumetric properties [2,9,10].
Considering the complex mechanical behaviour of CRAM, additional discussions are necessary on CRAM's mechanical behaviour in the laboratory. An appropriate laboratory characterization of the CRAM may increase projects' reliability, durability, and may reduce costs. erefore, this work aims to assess the permanent deformation of four di erent CRAMs (these mixtures have been used as the base course in an experimental test section), stabilized with asphalt emulsion and foamed asphalt, considering different flow number (FN) test procedures with respect to the loading condition and temperature, which were adapted for CRAM.
Rutting is considered by some researchers the main distress of asphalt pavements comprised by CRAM layers, including CRAM with the addition of portland cement or hydrated lime [5]. Rutting is affected by various factors: (i) asphalt content-higher stabilizing asphalt content leads to higher rutting probability [11,12]; (ii) active filler addition (such as portland cement and/or hydrated lime) accounts for an increase in permanent deformation resistance [13][14][15]; (iii) higher temperatures leads to higher permanent deformation susceptibility [16][17][18]; (iv) higher RAP content may increase rutting resistance [19]; (v) longer curing periods reduce the rutting potential [20]. In general, stiffer CRAMs are less susceptible to rutting.
Different permanent deformation test methods can be used in the laboratory to assess the behaviour of the recycled asphalt mixtures. For CRAMs with mechanical behaviour similar to granular materials, repeated load triaxial tests are used to obtain permanent deformation prediction models, with a frequency generally around 1 Hz to 5 Hz, and the number of cycles ranging from 10,000 to 1,000,000 [21][22][23].
When CRAM presents behaviors similar to a hot asphalt mixture (HMA), rutting tests are usually similar to those for HMA. One example is the flow number (FN) test, which is conducted without confinement, with load cycles limited to 10,000 cycles, and at 1 Hz [24,25]. Some studies show that temperature, cure, and stress state directly influence the resistance to permanent deformation of CRAMs [24,25]. e paper's main objective is to evaluate the permanent deformation of different cold recycled asphalt mixtures under different flow number test conditions with respect to stress and the temperature. e pavement structures of the test sections are presented in Figure 2. HMA stands for hot mixture asphalt, GAP means gap-graded asphalt mixture, and remaining infrastructure refers to the old pavement system that remained after pavement milling. After the original pavement milling, the remaining infrastructure was tested by means of a light weight deflectometer (LWD) and the mean modulus value of 118 MPa was found. Explanations on the base layer materials are given in the next section.

Materials.
Four different base layer materials were used to construct the test sections. e following nomenclature was adopted (Table 1): "AGG_xByF", where "AGG" string represents the main aggregate of the mixture, and "xB" and "yF" strings represent, respectively, the content and type of the asphalt binder and the content and type of the active filler. "B" is the type of asphalt binder (code E � emulsion, F � foam) and "F" is the type of filler (code C � cement, H � hydrated lime).
Two different RAPs (with different aggregate gradation) were used in the CRAM preparation and were obtained from the pavement milling of different heavy traffic highways in Brazil. All mixtures' gradations comply with Wirtgen's specification for CRAMs [6]. e aggregate gradations of the base layer materials are presented in Figure 3, along with Wirtgen's suggested limits for CRAM aggregate gradation. RAP_3E2C and RAP_3F2C mixtures have a nominal maximum aggregate size (NMAS) of 19 mm, and the RAP_2F1H has a NMAS of 25 mm. Base materials main characteristics are presented in Table 1.
From Table 1, significant total moisture content differences can be observed. is can be attributed to the presence of virgin aggregates, to the type of binder and to the aggregates' gradation. Virgin aggregates absorb more moisture than RAP. Also, foamed mixtures normally need more moisture to assure a proper binder distribution when compared to emulsion stabilized mixtures. e total moisture content determination was based on the higher specimens' dry density achieved after preliminary compaction and different moisture contents were assessed (5 different moisture content for each mixture).
Fine aggregate addition is important to ensure the best dispersion of the bitumen in CRAM. For the foamed asphalt mixtures, it occurs exclusively in the fine particles, thus increasing the cohesion of the compacted material. For this reason, some mixtures have also virgin aggregates in its composition.
To produce the foamed mixture RAP_3F2C, Wirtgen recycler model WLB 10 S was used. Uncompacted samples from the test section's base materials RAP_2F1H and GCS_2F1H were collected from the storage stockpiles during the test site construction and then compacted in the laboratory. RAP_3E2C was fully produced in the laboratory and was the only mixture that was also evaluated through the ignition test [26] to assess binder content. e average asphalt content found was 6.8% (1.8% emulsion residue + 5.0% RAP asphalt binder).

Compaction and Curing Process.
e test section samples were prepared by means of the vibratory compaction method, using a 1500 W vibratory hammer with a 5 kg surcharge. Samples with a 100 mm diameter and a final height of 150 mm were produced (in two layers) using the same compaction procedure reported by Meneses et al. [27]. e curing process consisted of maintaining the compacted specimens in the forced draft oven at 40°C for different curing times (0 or 28 days). e samples tested without curing (time zero) were sealed after compaction and tested at the compaction moisture content.

Flow Number.
e flow number (FN) test was used to evaluate CRAM's permanent deformation characteristics (primary, secondary, and tertiary zones) by subjecting the compacted specimens to a cyclic loading (1 Hz), without confinement. In the FN test, the total number of cycles and the duration of the test depend on the temperature and stress level [28]. e FN is then defined as the number of load cycles that takes the specimen to the beginning of the tertiary deformation zone, which corresponds to the minimum permanent axial deformation [29,30]. e deformation data and the number of cycles are adjusted with the Francken model, which was determined by Biligiri et al. [31].
In this study, three different test methods were conducted for a more comprehensive characterization of CRAM's permanent deformation: (i) ABNT NBR 16505 [32]; (ii) adaptation of the FN test for CRAM [24]; (iii) deviatoric stresses setup from 3D-Move analysis software simulation [33] of pavement structures.
is method aimed to evaluate the CRAMs in the same conditions that HMA mixtures applied in the wearing course are usually tested. Kim and Lee's [24] method (developed for cold recycled asphalt mixtures) consisted of 140 kPa axial stress cyclic loading at 40°C (conditions considered closer to the asphalt pavement base layer). e objective of this method is reaching the tertiary zone in a reasonable number of cycles that do not exceed 10,000 load cycles. Simulations of the pavement structures of

3D-Move Analysis
Simulation. e 3D-Move software is a continuum-based finite layer approach to estimate pavement responses. e software analysis and considers important factors, such as moving dynamic loads, tirepavement contact area, layer temperature, and viscoelastic properties of asphaltic materials [30,[34][35][36][37][38]. e four test sections ( Figure 2) were simulated using the 3D-Move analysis program. Poisson´s ratio (]) of the HMA, GAP, and base layers were considered 0.35, while the remaining infrastructure was 0.45. A single semi-axle load (dual tire) configuration was selected, with a circular contact area (radius � 0.107 m). e circular contact area is one of the most adopted by researchers and transport agencies [38]. e contact pressure was 560 kPa, and loads of 8, 10, and 14 tonnes were evaluated. Load speeds of 40, 50, 75, 80, 100, and 120 km/h were simulated. e remaining infrastructure of the test sections was tested by means of a light weight deflectometer (LWD) and the mean modulus value of 118 MPa was found.
Different temperature conditions were used to simulate the effect of the temperature dependence of the recycled materials on the pavement structure response. For the HMA sections, 50°C and 40°C were adopted for the surface layer and base layer, respectively; for the gap-graded sections, 55°C for the surface layer and 50°C for the base layer [39]. In addition to the information required for the simulation with the 3D-Move analysis software, mixtures and binders' viscoelastic properties were obtained through rheological tests.
For this work, only the vertical stress data from the simulations were considered, although displacements, strains, and horizontal stresses were also available from 3D-Move.
e 3D-Move simulations were employed only to verify an approximate stress state condition that the materials would be submitted in the field. Results of vertical stress for the base layers (at the alignment of the outer tire middle-where higher vertical stresses were found) at 40 km/ h are shown in Figure 4 (top, middle, and bottom layer).
For the RAP_3E2C, the maximum stress was 403.61 kPa at the top of the base layer. e stress level in the middle of the layer was approximately 60% lower than the maximum stress, and at the bottom, the stress level was 85% lower. erefore, the top of the base layer was considered the critical point for vertical stress. e trend observed for the other pavement structures was similar. e influence of the load speed on the base layer is shown in Figure 5. e load speed increase has caused a small stress value reduction. Considering the RAP_3E2C base layer, the stress level reduction was 5% when the load speed increases from 40 km/h to 120 km/h (from 403.61 kPa to 392.24 kPa). Again, the trend observed for the other pavement structures was similar. e vertical stress response was then evaluated in different structures (Figure 2) to analyse the influence of the applied load and speed on the mechanical response of the base materials. Figure 6 shows the vertical stress at the top of the base layer for the different CRAMs.
From Figure 6, it is observed that the vertical stress values were significantly higher for the segments with thinner surface layer (30 mm GAP -RAP_2F1H and GCS_2G1H). For both segments, there is a small increase in stress after the load increase, due to the thin asphalt layer and the higher temperature used in the simulations. At the ticker surface layer pavements (125 mm HMA-RAP_3E2C and RAP_3F2C), the load increase from 8 tf to 14 tf, which was approximately 15% the vertical stress (at 40 km/h). e stress values were lower when compared to the thin surface segments.

Flow Number.
e considered load configuration (for the flow number assessment) in 3D-Move was a dual wheel semi-axle, with 20 kN per tire. e contact pressure selected was 560 kPa, uniformly distributed on a 0.107 m radius circle. e loading speed was set to 40 km/h. e temperatures of 3D-Move simulations remained the same. e surface asphalt mixtures and the recycled bases were characterized by means of the dynamic modulus test (|E * | and δ), and the mixtures' asphalt binder was rheologically characterized by means of a dynamic shear rheometer (|G * | and δ). is data is mandatory in the 3D-Move software. e mixtures' |E * | master curves are presented on Figure 7.
FN test results for the three different load/temperature conditions are presented in Figure 8 and Table 2. Two replicates were tested for each condition (some samples presented issues during the test, especially the samples without curing, and were discarded). In some cases, the sample did not reach the FN, especially when the test was performed with a 140 kPa axial stress at 40°C, according to the method suggested by Kim and Lee [24] for CRAM stabilized with asphalt emulsion. e FN result of the HMA used in the wearing course of the segments with RAP_3E2C and RAP_3F2C is also presented in Table 2 for reference. e RAP_3E2C mixture behavior was the closest to the hot asphalt mixture behavior in terms of sensitivity to deformation. at behaviour may be related to the fact that this is the only mixture prepared with asphalt emulsion which did not include virgin aggregates (which has higher angularity and thus confers higher resistance to permanent deformation). However, the same did not happen for all CRAMs. Even for a high load cycle number (10,000 cycles), the foamed mixtures (RAP_3F2C, RAP_2F1H, and GCS_2F1H) did not reach the tertiary deformation zone for some procedures evaluated (especially after curing). It should be remarked that RAP_3E2C mixture. e method proposed by Kim and Lee [24] showed little efficacy for the analysis of the CRAMs tested in this work. e specimens have well withstood the stress and temperature proposed by the method.
All mixtures (without curing) have reached the FN under the 3D-Move test conditions after 28 days of curing, except for the GCS_2F1H mixture. In this method (3D-   Advances in Materials Science and Engineering  Advances in Materials Science and Engineering of the test for most of the mixtures without curing, although the stress level values were almost half of the stresses adopted from 3D-Move simulations. is indicates that 60°C is critical for the loading conditions of the evaluated mixtures and it raises the question if this is the best option for base materials. e mechanical response of the GCS_2F1H mixture was less influenced by the temperature variation, as most of its composition is virgin aggregates (lower binder content, if compared to the mixtures with RAP in its composition).
All methods highlighted the effect of curing on the permanent deformation behaviour of the tested materials.

Test Section Monitoring.
e rutting of the experimental sections was monitored for 45 months (RAP_3E2C and RAP_3F2C) and 14 months (RAP_2F1H and GCS_2F1H). e field measurements were performed on the wheel path (inner and outer) and at every 20 meters of the sections. e mean results and their standard error are presented in Figures 9 and 10.
Besides the inherent high variability of some results (attributed to the manual measurement method), clear trends are observed. e rutting values of RAP_2F1H and GCS_2F1H sections were much higher than the RAP_3E2C and RAP_3F2C sections, mainly due to the pavement configuration (RAP_3E2C and RAP_3F2C structures have a thicker HMA layer), but may also be related to the construction process quality, as 2 mm rutting values were observed right after the sections' construction. erefore, it can be considered that RAP_3E2C and RAP_3F2C CRAMs have performed satisfactorily (only around 1 mm rut), even after almost four years of heavy traffic loads. RAP_2F1H and GCS_2F1H CRAMs have shown a worse performance, but once again, this may be related to the high stress that the CRAM layer was submitted due to the thin gap-graded asphalt mixture layer (30 mm). All test sections presented stabilized results after around 10 months of construction, attributed to the curing of the base materials. e field permanent deformation values are sufficient to validate the FN test results. e fact that some mixtures did not reach the flow number during FN tests can be related to a high permanent deformation resistance of the mixtures, since good rutting performance was observed in the field.
It is noted that RAP_3E2C and RAP_3F2C mixtures performed similarly in the field, from the rutting monitoring perspective. But these mixtures presented different behaviour under FN tests. is fact can be attributed to the differences in FN and field stress state and strain levels. It should also be remarked that in the field, the CRAM layers are under the surface course layer, which is responsible for absorbing traffic loads and better distributing them to the lower layers.

Conclusions
e present study evaluated the permanent deformation of CRAMs using three variations of the flow number method: (i) ABNT NBR 16505 [32]-204 kPa and 60°C (method A); (ii) the method proposed by Kim and Lee [24]-140 kPa and 40°C (method B); (iii) using the results of 3D-Move software (to define the deviatoric stresses), through the analysis of an extensive database obtained from the four asphalt pavement structures analyzed, submitted to dynamic loads of traffic at various speeds-400/440/550 kPa and 50°C (method C). Based on the results presented, the following conclusions can be made: [20,42].
(i) 3D-Move software simulations proved to be a good analysis tool and aided the laboratory test's stresses definition. e software efficiency was corroborated by the data obtained from the test sections survey (ii) From the simulations, it was possible to observe the influence of the load speed and load magnitude, which was reflected on the FN test results (iii) e curing was a preponderant factor for all studied mixtures, providing a greater resistance to permanent deformation at all conditions selected for the FN test. e FN test's critical temperature needs to be better investigated to better correlate with the field data (iv) e RAP_3E2C mixture was more influenced by the temperature and load magnitude in the laboratory tests. is may have happened due to the highest asphalt binder content among the mixtures tested and the form of binder dispersion (emulsion) in the mixture. In addition, the 98% RAP composition may have contributed to a greater susceptibility to permanent deformation at the laboratory (v) e stress condition applied to the material plays a major role in the permanent deformation behaviour. e thicker wearing course layer on the top of the CRAM layer will protect it from major rutting issues (vi) Field monitoring of the test sections has shown a trend that corroborates the FN test results: a good permanent deformation resistance performance was observed after months of monitoring and stabilization of the plastic deformation probably attributed to the curing process of the CRAM materials.
Data Availability e data will be available from the corresponding author upon request by email (kamilla.vasconcelos@usp.br).

Conflicts of Interest
e authors declare that they have no conflicts of interest.