Fatigue of Cold Recycled Cement-Treated Pavement Layers: Experimental and Modeling Study

Abstract

Fatigue is the main design criterion for cold recycled cement-treated mixtures (CRCTMs). However, the literature shows that the fatigue behavior of such mixtures is still not well known. For example, the effect of increasing reclaimed asphalt pavement (RAP) contents is yet a topic of discussion. This experimental and modeling study helps fill knowledge gaps on CRCTM fatigue behavior using long-term curing fatigue tests and three design methods currently being used in different countries. The objectives of this study were: (1) to characterize the mechanical and fatigue behavior of mixtures of RAP, aggregates and cement; (2) to evaluate the fatigue life of pavements with base and subbase layers of such mixtures using the novel Brazilian design method (MeDiNa); and (3) to compare the results with those obtained using the South African Pavement Engineering Manual (SAPEM) transfer functions and the American Association of State Highway and Transportation Officials AASHTOWare Pavement Mechanistic-Empirical Design (PMED) software. The mixtures were tested in the laboratory using flexural static and cyclic tests, and the required parameters to use the methods were obtained. Experimental results and modeling demonstrated a superior fatigue behavior of recycled layers with higher RAP contents. On the other side, layers with lower RAP contents abruptly lost stiffness in short periods, making thicker structures necessary. Therefore, using high RAP contents is not only a sustainable practice, but also a technical benefit. The equivalent single axle loads obtained using the SAPEM were higher than those obtained using MeDiNa, while the PMED ones were higher than both previous methods. Despite the inherent differences, this suggests that MeDiNa is more conservative. It also highlights the importance of calibration based on long-term pavement performance data.

1. Introduction

Pavement recycling is an alternative and sustainable rehabilitation technique used worldwide because of its environmental-friendly aspects [1]. There are two main kinds of pavement recycling: hot recycling, in which the materials need to be heated, and cold recycling, which consists of treating the existing pavement materials without heating processes. In cold recycling, asphalt additives, that is, emulsion [2] or foamed asphalt [3], and chemical additives are used alone or in combination to enhance the old pavement materials’ behavior. Although it has several advantages [4], there is still a lack of application of cold recycling in comparison to other kinds of rehabilitation treatments (e.g., mill and fill).

Cold recycling with Portland cement is a technique that can treat most distresses of old asphalt pavements causing less environmental impact than other techniques [5]. The technique is usually employed in situ with the help of a recycling machine (reclaimer), being called full-depth reclamation (FDR). However, it is also possible to produce the cold recycled cement-treated mixture (CRCTM) in a plant and then use it to rehabilitate old pavements or even to construct new ones.

The main difference between this kind of mixture and traditional cement-treated materials (produced with natural raw materials) is the presence of reclaimed asphalt pavement (RAP) in the former. As the RAP content in a CRCTM increases, its strength and stiffness generally decrease. This behavior is a consensus in the literature, as reported in a broad state of art review by Fedrigo, Núñez and Visser [1].

The number of studies on the fatigue of CRCTMs is limited [6,7,8,9,10,11,12,13,14], and some factors affecting it are still not completely understood. For instance, the effect of RAP, since previous research reports either better [6,12] or worse [10,11,14] fatigue behavior with increasing RAP contents. Moreover, most researchers used 28-day fatigue tests since it is a consensus that cement virtually reaches its maximum strength after such an age. However, curing time might affect the variability of fatigue results and consequently, their modeling, thus curing times longer than six months are advised [15,16]. Furthermore, advanced techniques have been used to understand important characteristics of different pavement materials (e.g., RAP asphalt-aggregate interaction, particle shape, loading rate) [17,18,19], including the fatigue behavior of CRCTMs [20]. Despite that, engineers still often use parameters and methods of traditional cement-treated materials to design CRCTM layers, which might increase the uncertainty of the whole project.

The mentioned facts show that research towards engineering practice regarding CRCTM fatigue is still needed. The present experimental and modeling study helps fill the mentioned knowledge gaps on CRCTM fatigue behavior using long-term curing fatigue tests and three design methods currently being used in different countries. The limitation of this study is that the modeling focuses mainly on the CRCTM layers’ behavior based on the CRCTMs’ laboratory characterization, even though data from previous research were used for the other materials.

This study had three main objectives:

• To characterize the mechanical and fatigue behavior of CRCTMs (RAP, aggregates and cement) in the laboratory;
• To evaluate the fatigue life of pavement layers (base and subbase) made of CRCTMs using the laboratory data and the novel Brazilian pavement design method (MeDiNa);
• To compare MeDiNa results with those obtained using two well-established design methods: the South African Pavement Engineering Manual (SAPEM) transfer functions and the American Association of State Highway and Transportation Officials AASHTOWare Pavement Mechanistic-Empirical Design (PMED) software.

2. Cold Recycled Cement-Treated Mixtures Laboratory Characterization

This section presents the materials and methods used in the experimental part of the work as well as the obtained results.

2.1. Materials and Methods

Four CRCTMs constituted by RAP, aggregates and Portland cement (C) were studied.

Table 1. Cold recycled cement-treated mixtures (CRCTMs) composition.

Mixture Code	Mixture Number	C (%)	RAP (%)	GCS (%)	VA (%)
4%C-50%RAP-50%GCS	1	4.0	50	50	-
2.5%C-50%RAP-50%GCS	2	2.5	50	50	-
2.5%C-30%RAP-70%GCS	3	2.5	30	70	-
2.5%C-35%RAP-35%GCS-30%VA	4	2.5	35	35	30

shows the composition of such mixtures, while Figure 1 shows their grain size distributions. Table 1 also presents the mixtures’ codes and numbers. Three mixtures had only recycled graded crushed stone (GCS—which was collected from the same existing pavement structure as the RAP) in their composition, while a fourth also had virgin aggregates (VA), which is a common practice to achieve a denser grain size distribution. The cement contents were chosen to achieve lightly (2.5%) and heavily (4.0%) bound materials and were computed based on the dry mass of RAP and aggregates.

Prismatic specimens (400 mm × 100 mm × 100 mm) were compacted using a hydraulic press to achieve the mixtures’ compaction parameters. The specimens were cured for 1 year in a wet chamber at a temperature of 23 °C ± 2 °C and relative humidity above 90%. It was decided to use a long curing time to reduce the inherent variability of fatigue tests by reducing the effect of the cement reactions. In France, a 1-year curing time is standard practice for fatigue tests for cement-treated materials [15], while in Australia, the recommendation is a minimum curing time of 6 months [16].

Flexural static and cyclic (fatigue) tests were performed using the four-point bending test configuration described by Castañeda López et al. [6]. Flexural static tests were performed following the National Cooperative Highway Research Program (NCHRP) [21] test method for cement-treated materials. As per the mentioned method, a monotonic load increase was applied at a constant rate of 690 kPa/min (2.3 kN/min). Equation (1) was used to compute the flexural tensile stress. The stress corresponding to the ultimate load is the flexural tensile strength (FTS). The flexural tensile strain was computed using Equation (2). Flexural tensile strain at break (ε_b) corresponds to 95% of the ultimate flexural load. Flexural elastic modulus (E_i) was determined from the stress-strain relationships (secant modulus corresponding to 40% of the flexural tensile strength). For each mixture, three specimens were tested and the results were analyzed using simple statistical analysis (average and standard deviation).

$${\mathsf{\sigma }}_{\mathrm{i}}=\frac{{\mathrm{P}}_{\mathrm{i}}·\mathrm{L}}{\mathrm{w}·{\mathrm{h}}^2}$$

(1)

$${\mathsf{\epsilon }}_{\mathrm{i}}=\frac{108·\mathrm{h}·{\mathsf{\delta }}_{\mathrm{i}}·{10}^6}{23·{\mathrm{L}}^2}$$

(2)

where σ_i (MPa) is the flexural stress corresponding to force P_i (N); ε_i (microstrain) is the flexural tensile strain corresponding to LVDTs average displacement δ_i (mm); L is the length between supporting rollers (300 mm), and; w and h are the average width and height of the specimen (mm), respectively.

The procedures used for flexural fatigue tests were based on Austroads’ experience [16]. Specimens were subjected to 5 Hz haversine cyclic loading without resting periods. The applied stress ratios (SR) varied from 30% to 95%. The tests were terminated after specimen failure or after a million loading cycles. For each mixture, nine specimens were tested, but some prematurely failed before data was not collected. Equation (3) presents the model used to analyze the fatigue behavior of the mixtures, which correlates the number of cycles to failure and the stress ratio. This model was chosen because it is used in the MeDiNa software. The fatigue model parameters for each CRCTM were obtained using regression.

$$\mathrm{N}={10}^{{\mathrm{k}}_1+{\mathrm{k}}_2·\mathrm{SR}}$$

(3)

where N is the number of cycles to failure; SR is the stress ratio, that is, the ratio between applied and ultimate stress (in decimal), and; k₁ and k₂ are regression coefficients.

2.2. Results

Figure 2 presents the average values of FTS and E_i obtained for the CRCTMs, as well as the standard deviation. The results follow some trends reported in the literature. For instance, the strength increases with increased cement content and reduces with increased RAP content. Likewise, the highest modulus value was observed for the mixture with the higher cement content, while the lowest was for the mixture with the higher RAP content. Some reasons causing the reduction trend with RAP addition might be: (1) RAP has ag-glomerations formed by fines and asphalt binder, which are weaker and deform more than natural aggregates; (2) residual asphalt binder in RAP reduces the surface area that could be coated with cement, inhibiting bonding points; and (3) residual asphalt binder creates rounded aggregate, reducing interlocking [1]. It is important to note that although the modulus values seem high in comparison to those generally reported in the literature (average of about 4000 MPa) [1], the specimens were cured for 1 year. Moreover, Austroads [16] reported modulus values as high as 15,000 MPa for cement-treated materials with similar cement contents as those used in the present study.

Table 2. CRCTMs characterization.

Mixture	1	2	3	4
Maximum dry density, MDD (g/cm3) 1	2.26	2.26	2.26	2.36
Optimum moisture content, OMC (%) 1	7.5	7.5	7.5	7.5
Final flexural elastic modulus, Ef (MPa) 2	688	322	475	590
Flexural tensile strain at break, εb (με) 3	362	332	307	187
k1 3	9	8.9	7.9	9.1
k2 3	−8.6	−6.1	−8.6	−10.3
R2 3	0.75	0.69	0.73	0.69

¹ Obtained using AASHTO Modified compaction test; ² Modulus value at the end of the stabilized layer life: E_f = 10%.E_i (the computed parameter used in the MeDiNa software); ³ Obtained using flexural tests.

presents compaction parameters, the ε_b, and the fatigue model parameters for each mixture. Figure 3 presents the CRCTMs fatigue curves. It is well established that Portland cement concrete has a fatigue endurance limit of 50% of its static flexural tensile strength (SR = 0.5). CRCTM 1 (2.5%C-50%RAP-50%GCS) showed the highest fatigue life for an SR of 0.5, while CRCTM 3 (2.5%C-30%RAP-70%GCS) showed the lowest. This indicates that a higher RAP content might increase the mixture fatigue life, which might be related to RAP’s flexibility, increasing the mixture’s capacity to withstand deformation. The increasing ε_b with the RAP content corroborates this. The obtained values of strain at break were higher than those reported in the literature for traditional cement-treated mixtures [16], which also indicates that CRCTMs are more flexible, contributing to the enhancement of their fatigue characteristics.

3. Cold Recycled Cement-Treated Pavement Layers Modeling

This section presents pavement modeling using MeDiNa, SAPEM transfer functions, and AASHTOware PMED.

3.1. Fatigue Life Evaluation Using MeDiNa

This section shows the methods used for running the MeDiNa analyses and the corresponding results.

3.1.1. Methods

The hypothetical pavement structure shown in

Table 3. Initially analyzed pavement structure.

Layer	Material	Thickness (mm)	Modulus (MPa)	Poisson’s Ratio
Wearing course	Asphalt concrete (Asphalt PEN 30/45 #12.5 mm Sepetiba) 1	150	9000	0.30
Base	Stabilized (studied CRCTMs) 2	200	Ei and Ef for each mixture	0.25
Subbase	Granular (GCS—Gneiss C5) 1	200	381	0.35
Subgrade	Silty soil (Brazilian MCT classification NS’) 1,3	-	189	0.45

¹ Used data were the default present in the MeDiNa software, which are based on laboratory tests on each material; ² For the CRCTMs, the moduli values and fatigue parameters presented in Figure 2 and Table 2 were used; ³ MCT denotes Miniature Compacted Tropical, and NS’ denotes Non-lateritic silty soil.

was initially used in the analyses (initial structure). For the base layer, the data obtained for the CRCTMs were used. The initial thickness was 200 mm and the Poisson’s ratio was assumed to be 0.25, following research on CRCTMs by Kleinert [22]. Since the focus of this study was the cold recycled cement-treated layers, MeDiNa default materials were used for the other layers. However, it is important to note that all MeDiNa data are based on previous research results obtained using the default materials.

The design traffic for 10 years was set as 3 × 10⁶ equivalent single axle loads (ESALs), representing heavy Brazilian traffic. The reliability level was set as 85%, the default for arterial roads. The friction coefficient between layers was considered null since it is a default characteristic of the MeDiNa software. It is highlighted that this leads to higher stress and strain values than the full-adhesion condition often used by other pavement analysis software. After analyzing the initial structure (Table 3), it was optimized for each CRCTM characteristic to withstand the number of ESALs (optimized structure). The structures were optimized by increasing or decreasing the layers’ thickness (wearing course, base and subbase) and changing the subbase material (granular for CRCTM). All other data remained the same as the initial structure.

3.1.2. Results

Table 4. MeDiNa results in terms of cracking and rutting for the initial structures.

Limit/Mixture	Cracking (%)	Rutting (mm)
MeDiNa limit	30.0	13.0
CRCTM 1	8.5	1.6
CRCTM 2	14	1.8
CRCTM 3	15.7	1.6
CRCTM 4	13.5	1.6

shows the results (in terms of cracking and rutting) obtained for the initial structure using each CRCTM as the base layer. The table also shows the MeDiNa limits for cracking and rutting, considering a reliability level of 85%. Cracking relates to the asphalt wearing course and rutting to the cumulative response of aggregate layers and subgrade. Figure 4 presents the cold recycled cement-treated base layer modulus degradation over time (ratio of each month modulus, E, to the initial modulus, E_i), following the MeDiNa sigmoidal model for cement-treated materials.

Figure 4 corroborates that CRCTMs with higher RAP content show better fatigue behavior than those with lower RAP content. However, for the same RAP content, the CRCTM with the higher cement content (CRCTM 1, 4%C-50%RAP-50%GCS) performed better in the pavement structure than that with the lower cement content (CRCTM 2, 2.5%C-50%RAP-50%GCS). This is explained by the flexibility properties that the RAP material gives to the cement-treated mixtures, increasing its ability to withstand deformation (ε_b increases). The effect of the cement content aligns with the report by Zhao et al. [14], that is, a better fatigue behavior with increasing cement content. This might be related to the increase of the initial stiffness of the layer since it takes longer to reduce to a terminal condition. By the end of the design period, the base layers with the higher RAP content (50%) kept approximately 30% of their initial stiffness. In comparison, the base layers with the lower RAP content suffered an abrupt modulus reduction, demonstrating that the pavement structure was unsuitable for the design traffic.

Table 5. Optimized pavement structures.

Layer	CRCTM 1 Structure	CRCTM 2 Structure	CRCTM 3 Structure	CRCTM 4 Structure
Wearing course	Asphalt concrete, 110 mm	Asphalt concrete, 125 mm	Asphalt concrete, 150 mm	Asphalt concrete, 150 mm
Base	CRCTM 1, 200 mm	CRCTM 2, 200 mm	CRCTM 3, 280 mm	CRCTM 4, 300 mm
Subbase	Granular, 200 mm	Granular, 200 mm	CRCTM 3, 200 mm	CRCTM 4, 210 mm
Subgrade	Silty soil	Silty soil	Silty soil	Silty soil

presents the optimized pavement structures. The table shows that it was necessary to change the material of the subbase layers of the structures using CRCTMs 3 and 4 (those with lower RAP contents); the CRCTMs replaced the granular material. On the other hand, it was possible to reduce the thickness of the asphalt-wearing course of the other two structures. The values of cracking and rutting obtained for the optimized structures are presented in

Table 6. MeDiNa results in terms of cracking and rutting for the optimized structures.

Limit/Mixture	Cracking (%)	Rutting (mm)
MeDiNa limit	30.0	13.0
CRCTM 1	28.8	2.3
CRCTM 2	29.2	2.2
CRCTM 3	16.1	0.7
CRCTM 4	9.8	0.7

.

Figure 5 shows the cold recycled cement-treated layers’ modulus degradation over time for the optimized structures. The layers with the higher RAP content (CRCTM 1 and 2) show a better fatigue behavior than those with lower RAP content (CRCTM 3 and 4) even after the optimization. However, now the behavior of the formers is more similar. By the end of the design period, the CRCTM 1 and 2 base layers kept approximately 20% of their initial stiffness. CRCTM 3 and 4 layers suffered a more pronounced initial stiffness loss. They kept about 10% of their initial stiffness after 10 years, even with the addition of a cold recycled cement-treated subbase layer in the structure.

It is worth noting that Castañeda López et al. [6] studied the laboratory fatigue behavior of mixtures made of RAP, aggregates and cement, reporting the superior behavior of a mixture with 4% cement and 50% RAP, similar to the present one. Furthermore, Fedrigo et al. [12] reported that the laboratory fatigue behavior of RAP, soil and cement mixtures improved with increasing RAP contents. Both studies corroborate the present data and analysis. They also state that the RAP content effect is more complex, depending on the cement content and the thickness of the layers. On the other hand, other authors state that high RAP contents reduce the fatigue life of CRCTMs [10,11,14] and that their behavior is intermediate between those of traditional cement-treated mixtures and asphalt mixtures [7]. Due to that, there is an argument that it is necessary to add virgin aggregates to CRCTMs to increase the surface area to be coated with cement and the interlocking [11,23,24]. However, none of the mentioned studies used long-term cured specimens and neither evaluated design methods currently used in different countries.

3.2. SAPEM Transfer Functions Analyses and Comparison with MeDiNa

This section presents the methods and results for the SAPEM transfer functions. The results are then compared with those of the MeDiNa optimized structures.

3.2.1. Methods

The South African mechanistic pavement design method, initially introduced by Theyse, De Beer and Rust [25], is recognized worldwide. One of its main characteristics is that its transfer functions are based on field experiments. In contrast, most design methods are based on laboratory results that must be calibrated using field data. This is also the main reason to use it and to collate its results with those obtained using MeDiNa (laboratory-based).

Considering that, the MeDiNa optimized structures were evaluated using the SAPEM transfer functions to obtain the ESALs that they could withstand. The transfer functions are not presented here since they are appropriately described in the manual [26]. However,

Table 7. Inputs for the SAPEM transfer functions.

Material	Failure Criterion	Critical Parameter	Observation
Asphalt concrete	Fatigue	Tensile strain at the layer bottom	Transfer function for 8000 MPa stiffness 1,2
Chemically stabilized	Fatigue	Tensile strain at the layer bottom	Transfer function for chemically stabilized materials 3
Granular	Shear	Principal stresses at the layer mid depth	Transfer function for granular material G3 (moderate moisture condition) 1
Subgrade	Permanent deformation	Vertical compressive strain	Transfer function for subgrade 1

¹ All data were based on the SAPEM manual default values for each considered material. ² There are no parameters for a 9000 MPa stiffness (used in MeDiNa analyses) in the SAPEM manual. ³ CRCTM ε_b from Table 2, which were obtained using flexural tests.

presents some important characteristics of the method and the inputs used. The critical stresses and strains were computed using the software of the MeDiNa package. Since SAPEM does not have the option for a reliability level of 85%, the analyses considered a 90% reliability level. Moreover, the granular phase of the chemically stabilized layer (post-severe cracking) was not considered.

It is highlighted that the SAPEM transfer functions for chemically stabilized materials are strain-based (computed strain to strain at break ratio), and different from the MeDina model (stress-based). Some studies indicate that strain-based models are preferable for such materials [6,27,28,29]. The flexural tensile strain at the break of the studied CRCTMs was presented in Table 2 (Section 2.2). Such values are higher than those typically observed in the literature for traditional cement-treated materials due to the presence of RAP in the mixtures, which makes them more flexible [6,12].

3.2.2. Results

Table 8. ESAL results for the SAPEM transfer functions.

Layer	CRCTM 1 Structure	CRCTM 2 Structure	CRCTM 3 Structure	CRCTM 4 Structure
Wearing course	2.7 × 107	1.0 × 107	1.3 × 108	2.1 × 108
Base	8.6 × 106	7.4 × 106	1.7 × 107	2.1 × 107
Subbase	8.0 × 1020	2.1 × 1017	1.1 × 107	1.1 × 107
Subgrade	1.3 × 1019	1.9 × 1018	2.9 × 1020	5.1 × 1021

presents the results obtained using the SAPEM transfer functions in terms of ESALs for each pavement layer of the optimized structure (Table 5, Section 3.1.2). The lower obtained ESAL values always correspond to the cold recycled cement-treated layers (base or subbase). However, these values were always at least twice the design traffic considered in the MeDiNa analyses (3 × 10⁶). This suggests a more conservative design by MeDiNa. Nevertheless, the South African transfer functions were developed for traditional chemically stabilized materials (natural raw materials) and were not calibrated for recycled materials (RAP presence).

In these analyses, the layers with lower RAP content showed the higher ESALs, probably due to the fatigue model being more appropriate for them (the less the RAP, the more similar to a traditional chemically stabilized material). This behavior also agrees with the arguments by Castañeda López et al. [6] and Fedrigo et al. [12] regarding the influence of the thickness of the pavement layers since CRCTM 3 and CRCTM 4 structures are more robust.

The high ESAL values for the granular subbase and subgrade are explained by the fact that such transfer functions were developed based on field results of pavement structures with wearing courses much thinner than those of the optimized structures. It is important to note that the analyses were run with null friction between pavement layers (MeDiNa default). Although, considering full friction between layers would lead to lower stress and strain values, consequently, increasing the ESALs the structures would withstand.

3.3. AASHTOWare PMED Software Analyses and Comparison with MeDiNa

This section presents the methods for analyzing the MeDiNa optimized structures using the AASHTOWare PMED software.

3.3.1. Methods

AASHTOWare PMED is internationally recognized as one of the most advanced pavement design methods. However, although research related to the method has been recently conducted [21,30,31], the chemically stabilized layers’ analysis still needs to be enhanced, especially the fatigue model calibration. The main differences between the PMED and the methods above are that it considers the climate and computes the individual damage caused by the different types of vehicles’ axles and loads.

The used transfer functions are not presented in this paper since they are detailed in the AASHTO manual [32], but

Table 9. AASHTOWare PMED used transfer functions characteristics.

Layer	Failure Criterion	Critical Parameter
Asphalt wearing course	Fatigue cracking	Tensile strain at the layer bottom and reflective cracks (stabilized base)
Chemically stabilized	Fatigue cracking	Tensile stress at the layer bottom

shows their main characteristics. Since the CRCTM layers are the focus, only fatigue was considered in the analysis. The climate of Porto Alegre city (Southern Brazil) was used to run the analyses. The reliability level was set as 90%. It is highlighted that the reliability is not considered for chemically stabilized layers, only for the asphalt-wearing course cracking, for which the software sets a default failure limit of 25% (MeDiNa uses 30%). The design traffic was set as heavy (Long-Term Pavement Performance, LTPP, Program default data available in PMED software), and the maximum allowable annual average daily truck traffic (AADTT) was found, then the ESAL was computed for each failure criterion.

Table 10. Inputs for the AASHTOWare PMED software.

Material	Input Parameters
Asphalt concrete, AC	MeDiNa default material with advanced characterization 1
Chemically stabilized	Studied CRCTMs (Figure 2) 2
Granular	A-1-a PMED default 3
Subgrade soil	A-2-5 PMED default 3

¹ Advanced characterization data by Barros et al. [33,34,35]; ² For the CRCTMs, data were those presented in Figure 2 and Table 2. However, in PMED, minimum input values for FTS and E_i are mandatory and were used (1.0 MPa and 1035 MPa, respectively); ³ Default data present in PMED for each material were used.

presents the inputs used in the analyses. The soil and aggregates were chosen from the PMED default list to meet the characteristics of those used in the previous analyses (AASHTO soil classification system). The CRCTMs and asphalt concrete characteristics were based on experimental data. A more detailed asphalt concrete characterization than that present in MeDiNa is needed in PMED analyses (e.g., dynamic modulus), so data from research by Barros et al. on the same asphalt concrete were used [33,34,35].

3.3.2. Results

Table 11. ESAL results for the AASHTOWare PMED software.

Analyzed Criterion	CRCTM 1 Structure	CRCTM 2 Structure	CRCTM 3 Structure	CRCTM 4 Structure
Asphalt concrete fatigue	1.9 × 108	2.3 × 108	4.8 × 108	3.9 × 108
Stabilized layers fatigue	2.0 × 108	2.5 × 108	5.1 × 108	4.1 × 108

presents the results obtained using the AASHTOWare PMED software regarding ESALs for the analyzed criterion (Table 5, Section 3.1.2). The CRCTM layers’ ESAL values obtained using PMED are all higher than the one used in MeDiNa analysis (3 × 10⁶), even though the former considers a lower limit for failure due to asphalt wearing course cracking (25% and 30%, respectively). These values are also higher than those obtained using SAPEM. Furthermore, they are slightly higher than those of the asphalt-wearing course. Although there are inherent differences from one method to another, there are four possible reasons: (a) there is no field calibration for the chemically stabilized layers’ models, (b) the reliability is not applied for such materials, (c) the method considers reflection cracking on the asphalt concrete due to a cracked stabilized layer, and (d) the last and probably the main reason is that PMED assumes full friction between all layers (lower stresses and strains compared to the previous analysis).

The higher ESAL values for the CRCTM layers with lower RAP content might also be related to the assumption of full friction between layers. This made the base and subbase layers in CRCTM structures 3 and 4 work together as a single layer, increasing its overall thickness and reducing the stress that reaches its bottom.

4. Conclusions

This paper presented an experimental (long-term curing flexural tests) and modeling (three different design methods) study on the fatigue of cold recycled cement-treated pavement layers, similar to those obtained using full-depth reclamation. Based on the analysis, the following conclusions related to each main objective can be drawn:

• The mechanical and fatigue behaviors of cold recycled cement-treated mixtures (CRCTMs) were characterized in the laboratory using long-term curing flexural tests. The results provided the necessary inputs to use the novel Brazilian pavement design method (MeDiNa) and two well-established pavement design methods: the South African Pavement Engineering Manual (SAPEM) transfer functions and the AASHTOWare Pavement Mechanistic-Empirical Design (PMED) software. The laboratory results also confirmed some literature-reported trends, such as the increase of strength and stiffness with increasing cement and decreasing RAP contents. Furthermore, they also showed that mixtures with higher RAP contents have better fatigue behavior, possibly due to RAP’s flexibility.
• The modeling using MeDiNa showed that cold recycled cement-treated layers with higher cement (4%) and RAP (50%) contents perform better. Layers with high RAP content kept 30% of their initial stiffness by the end of the design period. Then, reducing the asphalt-wearing course thickness of structures with such mixtures was possible. On the other hand, layers with low RAP content abruptly lost stiffness at the beginning of the design period, being necessary to increase the layers’ thickness and replace the granular subbase with a stabilized subbase. Since fatigue is the main failure mode of cold recycled cement-treated mixtures, it can be concluded that using high RAP contents is not only a sustainable practice, but also a technical benefit.
• Using the SAPEM field-based transfer functions, it was possible to analyze the pavement structures optimized using MeDiNa. The former resulted in equivalent standard axle load (ESAL) repetitions of at least twice that considered the MeDiNa analyses. ESALs obtained using the AASHTOWare Pavement Mechanistic-Empirical Design software were higher than both the other methods. Despite the inherent differences from one method to another, this suggests that MeDiNa is more conservative when designing a pavement structure. It also highlights the importance of calibrating laboratory models using long-term pavement performance monitoring data.

Further research to understand the fatigue of CRCTM should focus on the development/calibration of mechanistic-empirical structural design methods based on field data. Even though it is harder than in the laboratory, varying cement and RAP contents in test sections could help to understand their effect. Moreover, using accelerated pavement testing would help obtain reliable results in shorter periods.

It is worth noting that this study has certain limitations. The cold recycled cement-treated mixture characteristics reported here are valid only for the studied mixtures. The laboratory fatigue models were not calibrated to field conditions; hence, using such data in other analyses should be carried out with caution. The modeling focused mainly on the CRCTM layers’ behavior based on the CRCTMs’ laboratory characterization.