Construction and Building Materials 260 (2020) 119891

Contents lists available at ScienceDirect

Construction and Building Materials

journal homepage: www.elsevier .com/locate /conbui ldmat

100% recycled high-modulus asphalt concrete mixture design andvalidation using vehicle simulator

https://doi.org/10.1016/j.conbuildmat.2020.1198910950-0618/� 2020 Elsevier Ltd. All rights reserved.

⇑ Corresponding author.E-mail address: martins.zaumanis@empa.ch (M. Zaumanis).

M. Zaumanis ⇑, M. Arraigada, L.D. PoulikakosEmpa Swiss Federal Laboratories for Materials Science and Technology, Überlandstrasse 129, CH-8600 Dübendorf, Switzerland

h i g h l i g h t s

� RAP properties are well matched foruse in high-modulus asphaltmixtures.

� Laboratory performance of 100% RAPHMAC is close to reference mix.

� Traffic load simulator revealsinsufficient cracking resistance of100% RAP mixes.

� Crack propagation testing isrecommended for design of highcontent RAP mixtures.

� Higher binder content is required in100% RAP mixes compared toreference.

g r a p h i c a l a b s t r a c t

a r t i c l e i n f o

Article history:Received 12 November 2019Received in revised form 13 May 2020Accepted 7 June 2020

Keywords:AsphaltRoadHMACEMERecyclingMMLS3Traffic load simulatorDigital image correlation (DIC)FatiguePerformance

a b s t r a c t

High modulus asphalt concrete (HMAC) mixtures are designed for high rutting resistance, high modulusand excellent fatigue performance. This is achieved through the use of high content of hard bitumen, lowair void content, and use of performance-based testing for mixture design. Such approach is presumablywell-matched for application of Reclaimed Asphalt Pavement (RAP): the RAP binder is hard because ofaging, RAP mixtures inherently have low air voids and performance-based mix design is recommendedto increase the degree of reliability when using high RAP mixtures. Here we present a study focusedon designing and validating performance of HMAC from 100% RAP. Through multiple design iterationswe found that it was not possible to fully fulfill the fatigue, modulus and rutting requirements for eitherof the recycled HMAC mixture types (C1 or C2) although the performance of one of the 100% RAP mix-tures came close to the HMAC design requirements. Nevertheless, validating of the best-performing100% RAP HMAC slabs using vehicle load simulator demonstrated that the recycled HMAC is significantlyless resistant towards crack propagation compared to a conventional HMAC. This highlights the impor-tance of testing cracking resistance for high RAP mixtures.

� 2020 Elsevier Ltd. All rights reserved.

1. Introduction

A significant amount of research is focused on increasing theuse of reclaimed asphalt pavement (RAP) in production of hot

mix asphalt. As a result, RAP recycling rates have been steadilygrowing [1,2]. Nevertheless, in many locations RAP is still abun-dant and stockpiles keep accumulating. In Switzerland, for exam-ple, around 2.5 million tons of asphalt are milled annually.Around 30% (750,000 t) are unused and of the 70% that are recy-cled, a part is down-cycled for use in lower value applications[3]. The situation is similar elsewhere: as a result of the excess

2 M. Zaumanis et al. / Construction and Building Materials 260 (2020) 119891

RAP and distrust in high RAP hot asphalt mixtures, RAP is oftenused in construction of unbound base layers, for road shouldersand low traffic intensity roads thus diminishing its value [4,5]. Thisis a waste of resources. RAP is the most valuable when used in hotmix asphalt due to the ability to replace bitumen, which is themost expensive part of the mixture [6,7]. At the same time it is asustainable approach because use of RAP reduces the need fornon-renewable resources [8,9]. For these reasons, every effortshould be made to increase the proportion of reclaimed asphaltthat is re-used for production of new asphalt mixtures.

Further advancements in design of hot-mix asphalt are neces-sary to increase the mean RAP content in asphalt pavements. Thisconcerns all aspects of the road construction industry: milling tech-nologies, RAP management, production technologies, paving, mix-ture design, and quality control methods. The focus of this articleis on the potential of using RAP in an innovative way – for produc-tion of High Modulus Asphalt Concrete (HMAC) entirely from RAP.

HMAC, also known by the French abbreviation EME (Enrobés áModule Elevé), was developed in France with the objective toimprove the mechanical properties of asphalt concrete providinghigh modulus, good fatigue behavior and excellent rutting resis-tance. HMAC mixtures are primarily used for base and bindercourses and allow reducing pavement layer thickness or increasingpavement life span. Two types of HMAC are often distinguished: C1– with primary focus towards resistance to permanent deforma-tion and C2 – with high resistance to both permanent deformationand fatigue. These properties are achieved through use of high con-tent of hard (and often polymer-modified) binder, low air void con-tent and application of performance-based testing requirementsfor fatigue, modulus and rutting resistance.

Such approach seems well matched with the use principles ofRAP because:

1) RAP binder is aged thus naturally provide the required hard-grade binder for HMAC [10–13];

2) high RAP mixtures chronically demonstrate low air voids[14–17];

3) performance-based mixture design is recommended forhigh-RAP mixtures because of unknown binder blending,relatively little field performance experience and potentialfor cracking [4,13,18–20].

For these reasons, it is worth exploring if RAP could be used inHMAC.

RAP has already been used in HMAC mixtures with a perfor-mance similar to virgin mixtures. Miro et al. [14] concluded thatup to 30% RAP addition to HMAC did not deteriorate its propertiesboth for lab-produced and plant-produced mixtures. Addition of50% RAP, however, reduced fatigue performance. Similarly, studiesby Ma et al. [21,22] noted that up to 30% RAP can be incorporatedin HMAC. The authors did not encourage higher RAP contentsbecause of the deteriorating low temperature and fatigue crackingresistance. In a separate study Ma et al. [23] noted that addition ofsynthetic low-molecular weight polyolefin polymers can helpimprove the performance of HMAC mixtures with up to 60% RAP.Such conclusions are also supported by studies of conventionalasphalt concrete mixtures [24–26]. In fact, the increased risk ofcracking is the main reason for the reluctance of road agencies toallow higher RAP contents. Unfortunately, performance-basedcracking resistance tests are still not well developed [18]. For thatreason, a Mobile Model Load Simulator (MMLS3) was used in thisresearch to verify the cracking resistance of the designed mixtures.The vehicle load simulator applies a downscaled load to a pave-ment slab and multiple researchers have demonstrated the useful-ness of the simulator to evaluate fatigue resistance of asphaltpavement [27–29].

At the same time, it has to be recognized that high RAP contentdoes not automatically lead to reduced cracking resistance. Ade-quate counter measures have to be taken – such as use of rejuve-nators, softer binder grades, performance-based mixture designmethods, etc. Success of such measures is supported by multiplestudies where low temperature cracking for high RAP mixtureshas demonstrated similar performance to conventional designs[5,10,30,31]. Likewise, improved fatigue performance of high RAPmixtures has been often reported [26,32–34]. One explanation isthat this performance depends on the test method and mode ofload application. At low strains (or low stress) which is relevantfor thick pavements a stiff mixture is likely to have a longer fatiguelife, while at high strains (or high stress) which simulates thinnerpavements a more flexible HMA will have a longer fatigue life. Thishighlights the importance of considering the pavement structure inconjunction with mix design. RAP may be beneficial to increasingpavement life if it is used in the right layer of a purposefullydesigned pavement structure.

1.1. Objective

The objective of the study is to investigate the potential todesign HMAC mixtures from 100% reclaimed asphalt pavementand validate the results using a vehicle load simulator.

2. Materials and methods

2.1. Materials

RAP milled from an undefined location in Switzerland was usedfor the research. It was screened in a RAP processing facility to frac-tions of 0/11 mm and 11/22 mm. The binder penetration of the0/11 mm fraction was 22 � 0.1 mm while the 11/22 mm fractionhad a binder penetration of 28 � 0.1 mm. The RAP binder contentand RAP aggregate grading curved (white curves) are summarizedin Table 1.

The 100% RAP mixtures were composed of reclaimed asphaltwith maximum grain size of 22 mm (black curve), but as it canbe seen in Table 1, 90% of the 11/22 fraction particles pass through16mm sieve. Thus, if only RAP aggregates were to be used, the onlyoption was to design a mixture with maximum aggregate size of16 mm. HMAC 16 mixtures are not specified in Switzerland sothere was no mixture available to use as reference. Instead, anHMAC 22 mixture designed according to the Swiss specificationswas supplied by an asphalt plant and used as a reference. In othercountries HMAC 16 mixtures are routinely used and have the sameperformance-based requirements as coarser mixtures [35]. It wastherefore decided to apply the Swiss performance-based require-ments of HMAC 22 for the 100% RAP HMAC 16 mixtures designedin this study.

The reference HMAC 22 mixture was used in a parallel project[36] to determine the thresholds for fatigue and modulus for inclu-sion in HMAC specifications in Switzerland. The required valuesused in the study are based on the recommendations from thatproject.

Virgin bitumen was used in the mixtures with the goal of ensur-ing the required performance-based properties of the HMAC mix-tures. Two types of virgin bitumen were used:

– Penetration grade 10/20 bitumen having penetration of 18 � 0.1 mm.

– Natural bitumen from Croatia having penetration below 1 � 0.1 mm and softening point of 115 �C. This bitumen has a highasphaltene content of >50% and was used to increase the stiff-ness of the mixture when necessary.

Table 1RAP aggregate grading curves (white curves).

RAP fraction Binder content,% Passing through sieve, %

0.063 0.125 0.250 0.500 1.0 2.0 4.0 5.6 8.0 11.2 16.0 22.4

0/5.6mm 5.6 12.8 18 26 33 41 65 84 100 100 100 100 1005.6/11.2mm 3.8 7.8 11 15 19 24 31 41 57 95 100 100 1000/11.2mm 5.3 13.8 19 26 35 45 6 82 92 99 100 100 10011.2/22mm 3.2 4.0 7 10 13 17 23 32 36 44 62 90 100

M. Zaumanis et al. / Construction and Building Materials 260 (2020) 119891 3

2.2. Methods

The experimental program is outlined in Fig. 1. HMAC C1 and C2were prepared for both reference and 100% RAP mixtures. The dif-ferent materials and their proportion that were used to prepare the100% RAP mixtures will be presented in detail along with the dis-cussion. Conventional tests were performed on all the mixtures butthe emphasis of the study was on ensuring correspondence toperformance-based design criteria. Finally, validation of the opti-mum design using Model Mobile Load Simulator was intended tocompare 100% RAP and the reference mixture.

2.2.1. Constituent material testsBinder was extracted from RAP according to EN 12697-1 using

toluene and recovered by rotary evaporator according to proceduredescribed in EN 12697-3. Gradation of the mixtures was deter-mined according to EN 12697-2 after extracting the binder (whitecurve).

Flow coefficient of fine portion (0.063–2 mm) of the aggregateswas tested according to EN 933-6. The test is performed by mea-suring the time that is necessary for a specified volume of thematerial to flow through a standardized funnel. The less angularthe aggregates, the faster they flow.

Penetration of bitumen was determined at 25 �C according toEN 1246. The penetration of the blend between RAP and virgin bin-

Fig. 1. Experimental program.

ders within the HMAC mixtures was estimated using Eq. (1). Thepenetration of the blend of RAP and virgin binders with the naturalbitumen was estimated using an equation provided by the produc-ers of the natural bitumen (Eq. (2)).

logPblend ¼ AlogPRAP þ BlogPvirg100

ð1Þ

Pblend – penetration of the blend of RAP and virgin binder,0.1 � mmPRAP – penetration of RAP binder, 0.1 � mm.Pvirg. – penetration of virgin binder, 0.1 � mm.A – percentage of RAP binder, %.B – percentage of virgin binder, %.

P ¼ Pblend � expð�0:0433 � CÞ ð2Þ

P – penetration of final binder in mixture, 0.1 � mm.Pblend – penetration of the blend of RAP and virgin binder,0.1 � mm.C – percentage of natural binder, %.

2.2.2. Mixture productionLaboratory mixtures were prepared using a heated batch mixer

at 175 �C. The constituent materials were heated at this tempera-ture in an oven for 3 h before mixing. This temperature wasselected according to the recommendations from Swiss specifica-tions for target binder grade (10/20). During mixing, the differentfractions of RAP were first mixed together for 1 min, followed byaddition of the virgin bitumen and additional 4 min of mixing.Because of RAP agglomerations and the relatively wide fractionsthat are normally used, high-RAP mixtures are prone to inhomoge-neous particle size distribution. To improve uniformity, the RAPobtained from the asphalt plant was split into subsamples ofapproximately 10 kg using a riffle box. These subsamples werethen used for laboratory mixing.

2.2.3. Sample preparationMarshall samples were prepared at 175 �C and compacted with

50 blows to each side according to EN 12697-34 after re-heating ofthe asphalt mixtures. To minimize aging, the samples were heatedin a microwave oven for 10 min, followed by a conventional ovenfor 50 min. The prepared Marshall samples were used to determinethe void characteristics for each mixture according to EN 12697-8.The determined bulk density fromMarshall samples was then usedas the target to calculate the necessary mixture mass for slab sam-ples. The slabs with dimensions of 18 cm � 50 cm � 10 cm wereprepared using a French roller compactor. To prepare the stiffnessand fatigue test samples, four 100 mm samples were cored fromeach slab. They were subsequently cut and polished to 40 mmheight. To avoid edge effect and maximize homogeneity, all facesof the samples were cut.

2.2.4. Stiffness and fatigueStiffness (according to EN 12697-26) followed by fatigue test

(EN 12697-24) on the same specimens was performed using cyclicindirect tensile test on cylindrical shaped specimens (IT-CY) of

4 M. Zaumanis et al. / Construction and Building Materials 260 (2020) 119891

150 mm in diameter and 60 mm in height. Stiffness tests were per-formed by applying a sinusoidal load at frequency of 10 Hz and10 �C, as defined for type testing by EN 13108-20. The load levelwas chosen to induce horizontal strains in the specimen in therange between 0.05 and 0.10‰. Although the European standardrequires four replicates for stiffness, in this case, only three speci-mens for each mixture were tested.

Fatigue test was performed at 10 �C by applying a sinusoidalrepeated loading at 10 Hz frequency on cylindrical samples (CIT-CY). The conventional failure criterion (Nf/50) of 50% loss of initialstiffness modulus (at 100 cycles) was used. Strain levels were cho-sen to induce failure of the specimens at three distinct levels (~105,~105.5, ~106). The standard requires three replicates at each strainlevel, however this was not always followed. If after testing of sev-eral first samples of a particular mixture, it was deemed that theresults would not satisfy the criteria, the subsequent samples werenot tested. Testing at different conditions allows calculatinganother conventional failure criterion – the strain at 1 millioncycles (denoted e6). This is calculated according to Eq. (3).

LogðNf Þ ¼ aþ 1b� LogðeÞ ð3Þ

where e is the amplitude of the tensile strain repeatedly applied, Nf

is the number of load applications to failure. The values a and b areregression constants, determined from plotting a regression linebetween fatigue failure criteria LogNf/50 and applied strain ampli-tude Logei. The value a is the ordinate of the regression and 1/b isthe slope. The fatigue and modulus requirements using CIT-CY testare currently not specified in Switzerland but a recent study pro-posed the requirements used in this article [36].

Fig. 2. MMLS3 [37].

Fig. 3. MMLS3 test

2.2.5. RuttingRutting resistance of the asphalt mixes was evaluated using

French Rutting Tester (FRT) at 30,000 cycles according to EN12697-22. The rutting resistance requirements are set for samplesprepared using a pneumatic wheel. However, a steel wheel com-pactor was used in this research. This method allows ensuringmorehomogeneous density throughout the slab which is importantwhen cores are drilled for stiffness and fatigue tests. Since it wasdesired to use the same sample preparation procedure for both rut-ting and fatigue tests, the steel wheel was preferred for both tests.

A comparison of rutting test results on slabs prepared usingsteel wheel and pneumatic wheel was made for an AC8 type mix-ture having the same porosity of around 4%. At 10,000 cycles theresults demonstrated 5.9% proportional rut depth for steel wheelsamples and 3.5% rut depth for the pneumatic wheel samples indi-cating that sample preparation with the standard method couldresult in a somewhat lower rutting than reported in this paper.

2.2.6. Mobile model load simulator (MMLS3)In order to upscale and validate the results obtained on labora-

tory samples, an MMLS3 test was performed at 20 �C. The MMLS3(illustrated in Fig. 2) is a downscaled accelerated pavement loadingdevice used for testing of pavement distresses under the load ofrepetitive rolling tires. The slabs used for the tests were 1.6 m longand 0.6 m wide with a thickness of 8 cm and they were made fromlaboratory-mixed loose material which was short-term aged for 4 hat 150 �C in a forced-draft oven. The virgin reference mixture wasnot aged because it was sampled from an asphalt plant where shortterm aging already had occurred. More information on the testdevice and slab preparation can be found in Zaumanis et al. [37].

Since cracking is the major concern for high content RAP mix-tures, the MMLS3 was used to determine the mechanical resistanceof slab specimens under rolling tire loading regime against fatiguecrack formation and propagation. To do this, the short edges of theslabs were laid onto steel profiles (supports) to induce bendingunder load. Between the steel profiles and below the slab, a thinrubber mat was placed to model a soft elastic foundation. Theslab- mat system simulates the condition of a flexible pavement,in which the asphalt layers have a continuous support of the sub-grade and bottom up cracking of asphalt pavement layers arisefrom repeated tensile strains due to traffic loading. A thermocouplewas installed inside the slab through a small hole drilled at oneside. Considering that mechanical friction produced by the rollingtires increases the temperature in the slab [38], air temperaturewas adjusted so that the slab was always at 20 ± 1 �C. To initiatea crack, a 35 mm deep transverse notch was cut at the center ofthe slab from the bottom side. The notch allows to control the loca-tion at which the crack starts to propagate through the slab. Thedetection and monitoring of the crack progression was made bytwo means (Fig. 3):

ing setup [37].

M. Zaumanis et al. / Construction and Building Materials 260 (2020) 119891 5

1. Digital image correlation (DIC) device: this non-contact opticaltechnique measures the deformation of a body under loading bytracking and correlating the displacements of random specklepatterns applied to the surface of the specimen. The movementis calculated from digital images obtained from 2 cameras posi-tioned in front of the area of interest; in this case, around thenotch of the slabs (DIC area in Fig. 3).

2. Linear variable differential transducer sensors (LVDTs): becauseof initiation and progression of micro and visible cracks, thestiffness of the slabs decreases. This leads to an increase inbending deformation under load. The vertical deflection wasmeasured through the use of LVDTs, as illustrated in Fig. 3. Con-sidering that the temperature and the loading speed are main-tained constant, any change in the amplitude of the verticaldeflection can be attributed to a change in the slabs’ stiffnessdue to cracking.

3. Results and discussion

Two types of mixtures were designed – HMAC C1 and HMAC C2.The major difference between the two is the higher fatigue resis-tance and higher modulus requirements for the HMAC C2 mixture.An iterative approach was used for designing the mixtures witheach subsequent design denoted by a new letter (A, B, C, D). The

Fig. 4. Gradation of HMAC C1 mixtures.

Table 2100% RAP HMAC C1 design and test results.

Parameter C1-A

RAP 0/5.6, % –RAP 5.6/11, % –RAP 0/11, % 32.8RAP 11/22, % 66.5Filler, % –Virgin bitumen, % 0.7Natural bitumen, % 0.07Total binder content (RAP + virgin), % 4.70Estimated binder penetration, 0.1 � mm 16Richness modulus 2.87Richness modulus @ 85% binder activation 2.39Air voids, % 2.2Modulus, MPa @ 10 �C, 10 Hz 25,151Fatigue strain at 106 cycles e6 @ 10�&10 Hz, lm/m 13.9Proportional rut depth @30,000 cycles, % –

* Requirement based on the recommendation of a research project [36].** Sample preparation differed from the standard method, likely resulting a somewhat

respective iterations of each mix type (C1 and C2) were designedin parallel.

3.1. Design of 100% RAP HMAC C1 mixture

Three distinct designs of 100% RAP HMAC C1 mixtures weredeveloped and tested (denoted C1-A; C1-B, C1-C). Grading curvesof the mixtures are illustrated in Fig. 4 and the mix compositionalong with the test results and requirements according to require-ments in Switzerland (SN 640 431-1c-NA) are summarized inTable 2. The fatigue results are visually illustrated in Fig. 5 and aconventional HMAC 22 C1 mixture (C1-Ref.) is added to providea reference of a mixture that passes the requirements.

C1-A mixture was designed using two fractions of RAP(0/11 mm and 11/22 mm) delivered by the asphalt producer. Theadded virgin binder content was chosen to satisfy the require-ments for Richness modulus (calculated according to Eq. (4), Rich-ness modulus is conceptually similar to calculation of binder filmthickness and results in minimum binder content for the specificgradation). Two binder types were selected – 10/20 penetrationclass bitumen and natural bitumen – at a proportion to provideresultant penetration close to the average requirement of 150.1 � mm. The performance-based test results for C1-A mixturedesign demonstrated very high modulus and a very low fatigue.

C1-B C1-C Required

13.3 10.5 –20.5 21.0 –– – –64.8 66.3 –2.6 – –1.75 2.1 –– – –5.14 5.58 �4.6021 21 15–253.11 3.57 �2.702.70 2.87 –2.0 2.0 3.0–6.022,646 20,850 �19,000*46.7 45.8 �50*– 6.8** �5.0

higher rut depth.

Fig. 5. Fatigue diagram showing failure criteria Nf/50 versus applied strain ampli-tude e for HMAC C1 mixtures.

6 M. Zaumanis et al. / Construction and Building Materials 260 (2020) 119891

It also had lower than required void content. Because of theseunsatisfactory results, rutting resistance was not tested for theC1-A mixture.

MR ¼ BGK

a �ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi0:25� 100�að Þþ2:3� a�bð Þþ12� b�cð Þþ150�c

1005q ð4Þ

Fig. 6. Gradation of HMAC C2 mixtures.

MR – Richness modulus.BGK – Binder content, %.a ¼ 2:65=qa, where qa is density of aggregates, Mg�m�3.a – mass of aggregates passing through 4.0 mm sieve, %.b – mass of aggregates passing through 0.25 mm sieve, %.c – mass of aggregates passing through 0.063 mm sieve, %.

C1-B mixture was designed to increase the void content,increase fatigue resistance and reduce modulus. It was done byre-sieving the 0/11 mm RAP fraction into 0/5.6 mm and5.6/11 mm fractions. This allowed developing a gradation closerto the lower requirement of the grading envelope in an attemptto increase air voids. At the same time, higher binder contentwas required to increase fatigue resistance and reduce modulus.It has been long discussed that part of the RAP binder film is notactivated and acts as part of the rock [33,39,40]. To account for thiseffect, in the C1-B and consecutive mixtures, we arbitrarilyassumed that the ‘‘active” portion of the binder is 85%. This thenrequired to increase the binder content to 5.14% in order to ensurethe Richness modulus of �2.7. Filler was added to increase the sur-face area of the aggregates. As can be seen in Table 2 these changesin mix design resulted in a significant increase in fatigue resistanceand a reduction of modulus. The cause for fatigue improvementcan be depicted from Fig. 5 where C1-B demonstrates significantlyhigher strain amplitude compared to the C1-A mixture, thus shift-ing the fatigue line to the right. However, the fatigue resistance isstill somewhat lower than the reference and the modulus is signif-icantly higher than the required value.

C1-C mixture was designed to further increase fatigue resis-tance and reduce modulus. To do this, filler was removed and bin-der content was increased to 5.58%. The proportions of othermaterials remained constant. These changes resulted in a slightdecrease in modulus and almost unchanged fatigue resistance.The fatigue resistance was about 10% lower than the requirementand approximately 20% lower than that of the reference mixture,indicating a shorter life cycle of the pavement. Nevertheless, itwas considered that it is impossible to further improve the perfor-mance using the available materials. Further fatigue improvementscould possibly be achieved by introduction of new additives. TheC1-C mixture was tested for rutting resistance where it demon-strated 6.8% proportional rut depth. This is higher than therequired �5.0%. As discussed in the Methods section, the samplepreparation differed from the requirement and it is possible thatsample preparation with a standard method would result is asomewhat lower rutting. This mix had air voids that were 1% lowerthan the requirement. However, since the main goal of the studywas to verify the performance-based properties of the mixtures,it was decided to use this mixture design for performance valida-tion of using MMLS3.

3.2. Design of 100% RAP HMAC C2 mixture

Four different 100% RAP HMAC C2 mixtures were designed (de-noted C2-A; C2-B; C2-C; C2-D) and their gradations are illustratedin Fig. 6. To ensure a distinct difference compared to the HMAC C1mixtures, only two RAP fractions, 0/11 and 11/22 mm, were used.Because of this, the gradation was finer compared to the C1 mix-ture but still corresponded to the 16 mmmaximum aggregate size.The mixture design proportions along with the test results are

summarized in Table 3. Fatigue test results are illustrated inFig. 7 and a conventional HMAC 22 C1 mixture (C2-Ref.) is addedto provide a reference of a mixture that passes the requirements.

C2-A mixture was designed by adding virgin 10/20 penetrationclass bitumen and natural bitumen in amount and proportions toensure the required richness modulus of 3.3 and penetration closeto the minimum requirement of 10 0.1 �mm. This resulted in highmodulus and unsatisfactory fatigue. The air void content, althoughlower than that of HMAC C1, was within the required range.

C2-B mixture was designed slightly coarser to increase air voidsand the content of natural bitumen was reduced to increase binderpenetration. As with the HMAC C1 mixtures, binder content wasincreased to reach the required Richness modulus assuming thatonly 85% RAP binder is activated. The results indicate slightincrease in fatigue resistance and no reduction in stiffness.

C2-C mixture was designed similar to C2-B with two differ-ences: (1) no natural bitumen was added to make the mixturesofter and (2) crumb rubber was added at 10% of virgin bindermass. The crumb rubber (particle size of <600 mm) was firstblended with the hot virgin bitumen and then added to the mixwith the intention of increasing the fatigue resistance. As shownin Fig. 7, this did not happen and instead the slope of the regressionbecame steeper reducing the strain at 106 cycles (e6). The modulusreduced slightly compared to C2-B mixture.

C2-D mixture was designed by removing the filler, and increas-ing the binder content to 6.5%. This resulted in reduction of airvoids and modulus, and increase in fatigue resistance. It was con-sidered that it is not possible to design a mixture with better fati-gue resistance using the existing RAP without adding any virginaggregates or new additives so the rutting resistance was tested.The FRT results demonstrate unacceptably high proportional rutdepth at 18.3% compared to permitted 7.5%. The reason for this islikely two-fold. First, the binder content for this iteration is veryhigh, reducing the interaction between the aggregates. Second,the RAP aggregates are not angular enough, and the frictionbetween the particles is reduced resulting in an increase of the per-manent deformations per wheel pass. A flow coefficient test per-formed on extracted RAP 0.063–2 mm fraction particle sizeaggregates produced a result of 29.5. Such result corresponds tothe lowest category in EN 13043 standard and can be consideredinsufficient for high rut resistance requirements, like HMAC mix-tures. A visual inspection of the RAP >2 mm fraction confirmed thatit also has rounded shape which could further contribute to weakrutting resistance.

Since both the fatigue and rutting test results were not satisfied,it was concluded that design of this mixture is not successful and it

Table 3100% RAP HMAC C2 design and test results.

Parameter C2-A C2-B C2-C C2-D Required

RAP 0/11, % 31.8 25.6 25.6 26.1 –RAP 11/22, % 65.2 70.0 70.0 71.1 –Filler, % 1.3 2.0 2.0 – –Virgin bitumen, % 1.45 2.28 2.40 2.80 –Natural bitumen, % 0.25 0.12 – – –Crumb rubber, % from virgin binder – – 0.10 – –Total binder content (RAP + virgin), % 5.51 6.04 6.04 6.5 �5.30Estimated binder penetration, 0.1 � mm 11 17 22 22 10–20Richness modulus 3.30 3.62 3.62 4.04 �3.3Richness modulus @ 85% binder activation 2.96 3.30 3.30 3.70 –Air voids, % 1.2 1.6 2.1 1.3 1.0–4.0Modulus, MPa @ 10 �C, 10 Hz 22,420 20,218 21,077 14,751 �19,000*Fatigue strain at 106 cycles e6 @10 �C&10 Hz, lm/m 33.9 42.2 37.6 56.7 �60*Proportional rut depth @30,000 cycles, % – – – 18.3** �7.5

* Requirement based on the recommendation of a research project [36].** Sample preparation differed from the standard, likely resulting in a somewhat higher rut depth.

Fig. 7. Fatigue diagram showing failure criteria Nf/50 versus applied strain ampli-tude e for C2 mixtures.

M. Zaumanis et al. / Construction and Building Materials 260 (2020) 119891 7

was not tested using MMLS3 traffic load simulator. A re-design ofthe mixture gradation and addition of portion of angular virginaggregates would be necessary to improve the performance.

3.3. Validation of C1-C mixture using MMLS3 vehicle simulator

The 100% RAP HMAC C1-C mixture design was selected for val-idation using MMLS3 traffic load simulator. Even though this mix-ture did not pass all the requirements for a conventional HMACmixture, it demonstrated the best performance from all the 100%RAP HMAC C1 mix types in the performance-tests. The perfor-mance of any of the 100% RAP HMAC C2 mixtures was too poorusing the performance-based tests to warrant testing using theresource-intensive MMLS3 test.

MMLS3 tests were intended to demonstrate the fatigue perfor-mance of the mixtures. The fatigue limit was defined as the num-ber of MMLS3 load applications when the bottom-up crack reachesthe specimen’s surface, after growing from the notch upwardsthrough the thickness of the slab. In some cases, the cracks mightbe visible to a naked eye thanks to a large differential movementbetween the crack edges produced by an MMLS3 tire passing.However, it was evident that the stiffness of the 80 mm thick slabsand the strong interlocking between the aggregates on both sidesof the crack produced a relatively small deflection under load.Therefore, it was difficult to see crack initiation with an unaidedeye. Consequently, only after post-processing of the result itbecame evident when the cracks developed.

Fig. 8 shows the results of the DIC analysis on the three testedslabs of 100% RAP HMAC C1-C mixture. The color maps illustratethe relative displacements in horizontal direction of a regionaround the notch when the MMLS3 tire is passing over the centerof the slab. These displacements are relative to the unloaded state,when no tire is touching the surface of the specimen. In the situa-tion where there is bending without cracks, the color map wouldpresent no discontinuity, i.e. there would be a smooth transitionbetween different tonalities. As seen in Fig. 8(a), this is the casefor the slab produced with the reference HMAC 22 C1 mixture evenafter 320,000 loadings. On the contrary, as it can be observed in theFig. 8(b) and (c) corresponding to slabs with 100% RAP HMAC C1-Cmixture, a discontinuity in the colors suggests the presence of acrack after just 1,000 load cycles. This means the slab is crackedalready very early during the test.

On the right side of Fig. 8, the horizontal displacements of twolines A and B are plotted to show the crack width produced at thespecific number of passes. Both 100% RAP slabs present a gap in theline with horizontal displacements at 1000 cycles. The formation ofa crack at the beginning of the loading means that the slabs werenot actually tested under a fatigue regime. Rather, they crackedsoon after starting of loading, thus showing a brittle behavior ofthe material. On the other hand, the horizontal displacements ofthe HMAC 22 C1 reference mixture present a continuous line evenafter several days of loading (320,000 cycles) indicating no fatiguedamage.

The DIC results were confirmed by the LVDT measurements.The analysis of data collected from LVDTs is based on the calcula-tion of the deflection amplitudes from the raw data vs. the accu-mulated MMLS3 cycles, as shown in Fig. 9. It can be expectedthat a deformation of a bending plate on an elastic foundationshould produce a steady growth of the deflection in the center ofthe span due to accumulation of micro cracks. The micro crackswill eventually start producing a dramatic reduction of stiffness,revealed as an inflection of the curve. This is observed as anincrease of the deflection amplitudes until the slab is completelycracked and acting as two separate plates. This behavior was notobserved in the 100% RAP nor in the reference mixture.

The reference HMAC 22 C1 slab showed no sign of cracking evenafter >320,000 load applications, which corresponds to 43 h ofuninterrupted MMLS3 operation. Since the machine cannot rununattended, the loading was stopped during the night while airtemperature was maintained at 20 �C. These breaks in testingappear as discontinuities in the deflection amplitude vs. load cyclescurve in Fig. 9 and are probably due to small temperature differ-ences inside the slab and due to crack healing, typically producedin asphalt materials during resting periods [41]. At the same time

Fig. 8. Pavement response due to repeated loading of MMLS3 for a) reference mixture, b) and c) 100% RAP HMAC C1-C mixture.

Fig. 9. Raw deflection data of the plates (left) and calculated deflection amplitude vs. accumulated number of MMLS3 load cycles (right).

8 M. Zaumanis et al. / Construction and Building Materials 260 (2020) 119891

M. Zaumanis et al. / Construction and Building Materials 260 (2020) 119891 9

it is evident that the curve does not produce the sudden increase indeflection amplitude which would characterize a crack.

As evident from analysis of the DIC results, both 100% RAPHMAC C1-C slabs cracked suddenly after the first few load applica-tions. The test was not discontinued at this point, but rather thedata was accumulated up to around 60,000 cycles. Although theslabs were completely cracked, the interlocking of aggregates inthe crack allowed a load transfer between both parts of the slab.Just like with the reference mixture, no inflection point is observed.In fact, without the DIC results, an observer of the LVDT resultsmight erroneously conclude that there is no crack. This signifiesthe importance of using DIC when performing MMLS3 tests.

It can be seen in the Fig. 9 that the RAP slabs exhibit a higherdeflection amplitude compared to the reference slab, but this islikely not related to the stiffness of the mixture. The stiffness of100% RAP and the reference mixture are actually relatively similaraccording to mixture design results (Table 2). The higher deflectionamplitude of 100% RAP slabs can be explained by the fact that theslab is cracked shortly after the beginning of the loading.

The MMLS3 tests were intended as a validation of fatigue resis-tance of the mixture. Unfortunately, in none of the cases fatigueendurance was actually measured. In the case of 100% RAP mix-tures, the slabs broke shortly after start of loading while for the vir-gin mixture the load was not high enough to induce fatiguedamage. It is therefore not possible to make any conclusionsregarding fatigue behavior with this set up. Unfortunately, repeat-ing the test at different test conditions was not possible due totime constraints.

While the objective to evaluate the fatigue resistance was notachieved, an important conclusion can be made from the MMLS3test results. The results demonstrated that with the same loadingconditions for the RAP and reference mixtures, the RAP mixturewas much more brittle compared to the reference mixture. Thisis despite the fact that the performance-based test results (mostimportantly modulus and fatigue) were actually relatively similarfor the 100% RAP and reference mixture. A likely explanation forthe differences in the MMLS3 test results is then the difference inresistance to crack propagation. Note that the notch in the centerof slab is an artificial crack, intended to initiate crack propagationwhereas fatigue tests that were performed as part of mixturedesign only consider the formation of microcracks. They do notcharacterize the second stage of material failure – crack propaga-tion. When loading was applied to this artificial crack usingMMLS3, the crack in 100% RAP mixtures propagated much fasterthan that in the reference mixture. This indicates that for high-RAP HMAC mixtures measuring the crack propagation might bean important mix design consideration. Because of the small sam-ple size in this study, this conclusion needs to be further addressedin another research.

In an actual pavement, the two parameters – crack initiation andcrack propagation are related. For a stiffer mixture, like the RAPmixture used in this research, the development of micro cracksmay be delayed because of ability of the material to absorb strainenergy before they start to have tensile failures [32,34]. Once initialmicro cracks have formed networks, resistance to crack propaga-tion plays an important role. In our study, the fatigue test character-izes crack initiation while the MMLS3 reflects crack propagationbecause the slabs had an artificial notch. Based on the test results,we can see that the performance in each of these tests for the100% RAP was very different as compared to the reference mixture.

4. Conclusions and recommendations

In theory, the design principle of HMAC is a good match forusing high RAP content in mixtures. This is because of the require-

ment to test the mixture using performance-based tests instead ofrelying mostly on volumetric properties like it is done for asphaltconcrete (AC) type mixtures. The aged, hard binder of reclaimedasphalt also often matches requirements for hard binder grade inhigh modulus asphalt concrete mixtures (HMAC). This study pre-sented the design of HMAC mixture made from 100% reclaimedasphalt. The following conclusions can be drawn from the study:

1) It was not possible to design a 100% RAP HMAC C1 or C2mixture to pass the performance-based test requirements.This is likely due to the presence of aggregates having lowaggregate angularity in the RAP (either because of low angu-larity materials used in the original mixture design or due todegrading during milling and processing). This indicatesthat, although the binder may fit HMAC mixtures, imple-mentation of management procedure of the RAP is necessaryto also control the properties of the aggregates.

2) The Model Mobile Load Simulator (MMLS3) results did notallow making conclusions regarding fatigue resistance ofeither the RAP or the virgin mixtures because the selectedloading conditions did not induce fatigue failure. However,the results did allow to conclude that at the conditionsimposed by the test, the RAP mixture was much more brittlethan the virgin mixture. Such results were not expectedbased on the laboratory mix design results, demonstratingthe importance of upscaling the laboratory results beforeallowing such pavements into practice.

3) The failure of the existing mix design procedure to capturethe dramatically insufficient resistance to crack propagationof HMAC mixtures, suggests that performance-based mixdesign methods might need to be re-evaluated when usinghigh RAP mixtures. One parameter that could aid character-ization of high RAP mixtures is testing of susceptibility tocrack propagation as part of mix design.

4) In order to improve fatigue performance, 100% recycled mix-tures required higher binder content than normally found inHMAC mixtures. This is likely because of not fully activatedRAP binder. However, in this study, even increasing of bindercontent did not allow fulfilling the fatigue requirements andtherefore use of additives should be considered in mixturescontaining high content of RAP.

5) Sieving of RAP into three fractions as opposed to using twofractions allowed more flexibility to design the mixtureresulting in a better performance.

6) Use of linear variable differential transducer sensors did notallow detecting propagation of crack in the MMLS3 slabs.The use of digital image correlation was the preferred option.

In general, the results of this work demonstrate the difficulty indesigning 100% RAP HMAC mixtures. However, it has to be consid-ered that HMACmixtures are used for high-performance pavementson highways but the RAP used in this study was from an unknownorigin with low quality aggregates. An appropriate management ofthe RAP might enable using the material in HMAC in very high con-tent. Another potential approach is to use high content of RAP inHMAC mixtures designed for intermediate and low traffic intensityroads as a replacement to conventional asphalt concrete mixtures toreduce pavement layer thickness, increase pavement life expec-tancy, and reduce costs. Further research on pavement design, andcost effectiveness is encouraged to verify this.

CRediT authorship contribution statement

M. Zaumanis: Conceptualization, Methodology, Software, Vali-dation, Formal analysis, Investigation, Data curation, Writing -

10 M. Zaumanis et al. / Construction and Building Materials 260 (2020) 119891

original draft, Visualization, Supervision, Funding acquisition. M.Arraigada: Software, Validation, Investigation, Data curation, Writ-ing - review & editing, Visualization. L.D. Poulikakos: Conceptual-ization, Methodology, Validation, Writing - review & editing,Supervision, Project administration, Funding acquisition.

Declaration of Competing Interest

The authors declare that they have no known competing finan-cial interests or personal relationships that could have appearedto influence the work reported in this paper.

Acknowledgements

This contribution is part of a Swiss national project titled ‘‘Sus-tainable Fully Recycled Asphalt Concrete” funded by the Swiss Fed-eral Office For The Environment grant No. UTF489.19.14/IDM2006.2423.487. The authors are grateful to F. Piemontese forassisting with the MMLS3 measurements and the H. Kienast, C.Meierhofer, R. Takacs, and S. Küntzel from Empa Concrete andAsphalt laboratory for performing tests.

References

[1] EAPA, Asphalt in figures 2016, 2018. http://www.eapa.org/userfiles/2/AsphaltinFigures/2016/AIF_2016.pdf.

[2] NAPA, Asphalt pavement industry survey on recycled materials and warm-mixasphalt usage 2016, Lanham, MD, 2017, http://www.asphaltpavement.org/PDFs/IS138/IS138-2016_RAP-RAS-WMA_Survey_Final.pdf.

[3] S. Dünner, VSS 2005/454 Forschungspaket Recycling von Ausbauasphalt inHeissmischgut: EP3: Stofffluss- und Nachhaltigkeitsbeurteilung, 2013.

[4] M. Zaumanis, R.B. Mallick, Review of very high-content reclaimed asphalt usein plant-produced pavements: state of the art, Int. J. Pavement Eng. 16 (2015)39–55, https://doi.org/10.1080/10298436.2014.893331.

[5] R. West, J.R. Willis, M. Marasteanu, Improved Mix Design, Evaluation, andMaterials Management Practices for Hot Mix Asphalt with High ReclaimedAsphalt Pavement Content, National Center for Asphalt Technology, Auburn,AL, 2013.

[6] A. Copeland, Reclaimed asphalt pavement in asphalt mixtures: state of thepractice, McLean, VA, 2011.

[7] M. Zaumanis, R.B. Mallick, R. Frank, 100% Hot mix asphalt recycling: challengesand benefits, Transp. Res. Procedia (2016) 3493–3502, https://doi.org/10.1016/j.trpro.2016.05.315.

[8] R.B. Mallick, M.J. Radzicki, M. Zaumanis, R. Frank, Use of system dynamics forproper conservation and recycling of aggregates for sustainable roadconstruction, Resour. Conserv. Recycl. 86 (2014), https://doi.org/10.1016/j.resconrec.2014.02.006.

[9] Q. Aurangzeb, I.L. Al-Qadi, H. Ozer, R. Yang, Hybrid life cycle assessment forasphalt mixtures with high RAP content, Resour. Conserv. Recycl. 83 (2014)77–86, https://doi.org/10.1016/J.RESCONREC.2013.12.004.

[10] M.C. Cavalli, M. Zaumanis, E. Mazza, M.N. Partl, L.D. Poulikakos, Effect of ageingon the mechanical and chemical properties of binder from RAP treated withbio-based rejuvenators, Compos. Part B Eng. 141 (2018) 174–181, https://doi.org/10.1016/j.compositesb.2017.12.060.

[11] M. Zaumanis, J. Oga, V. Haritonovs, How to reduce reclaimed asphaltvariability: A full-scale study, Constr. Build. Mater. 188 (2018) 546–554,https://doi.org/10.1016/j.conbuildmat.2018.08.137.

[12] T. Koudelka, P. Coufalik, J. Fiedler, I. Coufalikova, M. Varaus, F. Yin, Rheologicalevaluation of asphalt blends at multiple rejuvenation and aging cycles, RoadMater. Pavement Des. 20 (2019) S3–S18, https://doi.org/10.1080/14680629.2019.1588150.

[13] S. Im, P. Karki, F. Zhou, Development of new mix design method for asphaltmixtures containing RAP and rejuvenators, Constr. Build. Mater. 115 (2016)727–734, https://doi.org/10.1016/J.CONBUILDMAT.2016.04.081.

[14] R. Miró, G. Valdés, A. Martínez, P. Segura, C. Rodríguez, Evaluation of highmodulus mixture behaviour with high reclaimed asphalt pavement (RAP)percentages for sustainable road construction, Constr. Build. Mater. 25 (2011)3854–3862, https://doi.org/10.1016/j.conbuildmat.2011.04.006.

[15] W. Mogawer, A. Austerman, L. Mohammad, M.E. Kutay, Evaluation of highRAP-WMA asphalt rubber mixtures, Road Mater. Pavement Des. 14 (2013)129–147, https://doi.org/10.1080/14680629.2013.812846.

[16] R.S. McDaniel, A. Shah, G.A. Huber, A. Copeland, Effects of reclaimed asphaltpavement content and virgin binder grade on properties of plant produced

mixtures, Road Mater. Pavement Des. 13 (2012) 161–182, https://doi.org/10.1080/14680629.2012.657066.

[17] R. West, A. Kvasnak, N. Tran, B. Powell, P. Turner, Testing of moderate and highreclaimed asphalt pavement content mixes, Transp. Res. Rec. J. Transp. Res.Board. 2126 (2009) 100–108, https://doi.org/10.3141/2126-12.

[18] M. Zaumanis, L.D. Poulikakos, M.N. Partl, Performance-based design of asphaltmixtures and review of key parameters, Mater. Des. 141 (2018) 185–201,https://doi.org/10.1016/j.matdes.2017.12.035.

[19] M. Zaumanis, M.C. Cavalli, L.D. Poulikakos, How not to design 100% recycledasphalt mixture using performance-based tests, Road Mater. Pavement Des.(2019), https://doi.org/10.1080/14680629.2018.1561381.

[20] M. Zaumanis, A. Valters, Comparison of two low-temperature cracking testsfor use in performance-based asphalt mixture design, Int. J. Pavement Eng.(2019), https://doi.org/10.1080/10298436.2018.1549323.

[21] T. Ma, H. Wang, X. Huang, Z. Wang, F. Xiao, Laboratory performancecharacteristics of high modulus asphalt mixture with high-content RAP,Constr. Build. Mater. 101 (2015) 975–982, https://doi.org/10.1016/j.conbuildmat.2015.10.160.

[22] T. Ma, Y.L. Zhao, X.M. Huang, Y. Zhang, Using RAP material in high modulusasphalt mixture, J. Test. Eval. 44 (2016) 781–787, https://doi.org/10.1520/jte20150194.

[23] T. Ma, X. Ding, D. Zhang, X. Huang, J. Chen, Experimental study of recycledasphalt concrete modified by high-modulus agent, Constr. Build. Mater. 128(2016) 128–135, https://doi.org/10.1016/j.conbuildmat.2016.10.078.

[24] K.K. Tam, P. Joseph, D.F. Lynch, Five-year experience of low-temperatureperformance of recycled hot mix, Transp. Res. Rec. (1992) 56–65.

[25] R. West, J. Michael, R.E. Turochu, S. Maghsoodloo, Comparison of virgin andrecycled asphalt pavements using long-term pavement performance SPS-5data, in: Transp. Res. Board 90th Annu. Meet., Transportation Research Board,Washington, D.C., 2011.

[26] I.L. Al-Qadi, Q. Aurangzeb, S.H. Carpenter, W.J. Pine, T. J., Impact of high RAPcontent on structural and performance properties of asphalt mixtures, IllinoisDOT, Springfield, IL, 2012.

[27] S. Lee, Y. Seo, Y.R. Kim, Experimental validation of laboratory performancemodels using the third scale accelerated pavement testing, KSCE J. Civ. Eng. 10(2006) 9–14, https://doi.org/10.1007/bf02829299.

[28] H. Kim, K. Sokolov, L.D. Poulikakos, M.N. Partl, Fatigue evaluation of porousasphalt composites with carbon fiber reinforcement polymer grids, Transp.Res. Rec. (2009) 108–117, https://doi.org/10.3141/2116-14.

[29] Y. Huang, L. Wang, H. Xiong, Evaluation of pavement response andperformance under different scales of APT facilities, Road Mater. PavementDes. 18 (2017) 159–169, https://doi.org/10.1080/14680629.2017.1329871.

[30] A. Shah, R. McDaniel, G. Huber, V. Gallivan, Investigation of properties of plant-produced reclaimed asphalt pavement mixtures, Transp. Res. Rec. J. Transp.Res. Board. 2014 (1998) 103–111.

[31] M. Zaumanis, R.B. Mallick, R. Frank, Evaluation of different recycling agents forrestoring aged asphalt binder and performance of 100 % recycled asphalt,Mater. Struct. Constr. 48 (2015), https://doi.org/10.1617/s11527-014-0332-5.

[32] J.S. Daniel, N. Gibson, S. Tarbox, A. Copeland, A. Andiescu, Effect of long-termaging on RAP mixtures: laboratory evaluation of plant produced mixtures,Asph. Paving Technol. 82 (2013) 327–365.

[33] T. Bennert, R. Dongree, Backcalculation method to determine effective asphaltbinder properties of recycled asphalt pavement mixtures, Transp. Res. Rec. J.Transp. Res. Board. (2010) 75–84, https://doi.org/10.3141/2179-09.

[34] B. Huang, Z. Zhang, W. Kinger, Fatigue crack characteristics of HMA mixturescontaining RAP, in: 5th Int. RILEM Conf. Crack. Pavements, RILEM, Limoges,France, 2004 pp. 631–638.

[35] W. Bankowski, L. Wiman, B. Kalman, M. Tusar, I. Pap, D. Pangarowa, J.Hendrikson, A. Strineka, Sustainable pavements for European new memberstates (SPENS), 2009.

[36] L.D. Poulikakos, Standardization requirements for the Swiss annex to EN forcomplex modulus and resistance to fatigue, Project Nr. VSS 2014/502, FB 1660,(2019) 74.

[37] M. Zaumanis, M. Arraigada, S.A. Wyss, K. Zeyer, M.C. Cavalli, L.D. Poulikakos,Performance-based design of 100% recycled hot-mix asphalt and validationusing traffic load simulator, J. Clean. Prod. 237 (2019) 117679, https://doi.org/10.1016/j.jclepro.2019.117679.

[38] S. Bhattacharjee, R.B. Mallick, Effect of temperature on fatigue performance ofhot mix asphalt tested under model mobile load simulator, Int. J. PavementEng. 13 (2012) 166–180, https://doi.org/10.1080/10298436.2011.653565.

[39] M. Zaumanis, L. Boesiger, B. Kunz, M.C. Cavalli, L. Poulikakos, Determiningoptimum rejuvenator addition location in asphalt production plant, Constr.Build. Mater. 198 (2019) 368–378, https://doi.org/10.1016/j.conbuildmat.2018.11.239.

[40] M. Zaumanis, R.B. Mallick, Finite element modeling of rejuvenator diffusion inRAP binder film – simulation of plant mixing process, Multi-Scale Model.Charact. Infrastruct. Mater. 8 (2013) 407–419, https://doi.org/10.1007/978-94-007-6878-9_30.

[41] P. Ayar, F. Moreno-Navarro, M.C. Rubio-Gámez, The healing capability ofasphalt pavements: a state of the art review, J. Clean. Prod. 113 (2016) 28–40,https://doi.org/10.1016/j.jclepro.2015.12.034.