Construction and Building Materials 169 (2018) 306–314

Contents lists available at ScienceDirect

Construction and Building Materials

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

Cold recycling of lime-fly ash stabilized macadam mixtures as pavementbases and subbases

https://doi.org/10.1016/j.conbuildmat.2018.03.0300950-0618/� 2018 Elsevier Ltd. All rights reserved.

⇑ Corresponding author.E-mail address: liqiang2526@njfu.edu.cn (Q. Li).

Qiang Li a,⇑, Zhibing Wang b, Yuliang Li c, Jianlin Shang d

a School of Civil Engineering, Nanjing Forestry University, Nanjing 210037, ChinabNantong Nengda Construction Investment Co., Ltd, Nantong 226009, Chinac Jiangsu Transportation Institute Group, Nanjing 210000, Chinad Jiangsu Daorun Engineering Technology Co., Ltd, Nantong 226010, China

h i g h l i g h t s

� Pavement performance of recycled lime-fly ash stabilized macadam was evaluated.� Effects of cement contents, gradations, and RM properties were evaluated.� Relationships between RM properties and pavement performance were developed.

a r t i c l e i n f o

Article history:Received 31 October 2017Received in revised form 24 February 2018Accepted 2 March 2018

Keywords:Lime-fly ash stabilized macadamCement stabilized cold recyclingReclaimed materials propertiesGradation optimizationPavement performance

a b s t r a c t

The cold recycling technology has been increasingly used to rehabilitate semi-rigid pavement bases andsubbases in China due to the focus of the sustainable development. This paper presents the mechanicaland pavement performances of cold recycled lime-fly ash stabilized macadam (LFSM) mixtures usingPortland cement as the stabilizing material. The unconfined compressive strength, indirect tensilestrength, frost resistance, dry shrinkage (DS) resistance, and temperature shrinkage (TS) resistance of dif-ferent cold recycled mixtures (CRMs) with three cement contents, two reclaimed materials (RM) sources,and two aggregate gradations were measured, respectively. The locally recommended criteria for RMproperties were proposed based on performance correlations between RMs and CRMs. It is found thatthe gradation optimization by adding appropriate amount of the coarse virgin aggregate (VA) can greatlyimprove the mechanical strength and pavement performances of CRMs. CRMs with the clean RM do notshow significantly worse DS and TS resistance than cement stabilized macadam mixtures with 100% ofthe VA. The CRM strength is highly affected by the RM strength and angularity. Field investigations con-firm the feasibility and application potentials of the cold recycled LFSM layer as pavement bases andsubbases.

� 2018 Elsevier Ltd. All rights reserved.

1. Introduction

Semi-rigid (inorganic binder stabilized) materials have widelybeen used for pavement engineering in many countries for decadesdue to advantages of high strength, good loading distribution,reduced thickness requirement, improved workability, andincreased resistance to climatic effects [1–9]. Especially in China,the semi-rigid base asphalt or concrete pavement is the predomi-nant structure type for high-level highways and urban roads sofar. The lime-fly ash stabilized macadam (LFSM) mixture has beenselected as one of the main semi-rigid base or subbase materials

since the 1980s. It is composed of aggregates with a proper grada-tion, 18–25% lime-fly ash of the aggregate weight, and water at theoptimum content. The weight ratio of lime to fly ash generally var-ies from 1:2 to 1:4. The combination of coal fly ash and lime in thewet process generates the calcium silicate hydrate and calciumaluminate hydrate in the gel condition. Subsequently, thesehydrates crystallize to form interparticle bonds for cementingaggregates [10]. It supplies an opportunity to use industrial by-products and local materials. Therefore, LFSM is widely acceptedas a cost-effective paving material.

A lot of distresses, such as reflective cracking, raveling, and pot-holes, appear in LFSM bases and subbases under continuous effectsof vehicle loading and environmental conditions. These pavementdistresses greatly reduce the pavement service quality and driving

Q. Li et al. / Construction and Building Materials 169 (2018) 306–314 307

safety. In China, high-level highways and urban roads constructedearly have entered into a peak period of reconstruction and reha-bilitation. Over 2.2 million tons waste pavement materials are pro-duced every year through the reconstruction and rehabilitation ofexisting highways [11]. Moreover, a large number of highways andurban roads are faced with upgrading and alignment reconstruc-tion due to the continuous city development. With mounting con-cerns about the sustainable development, energy savings,greenhouse gases reduction, and conservation of natural resources,the cold recycling technology provides a technically sound, cost-effective, and environmentally friendly method to restore thepavement performance of LFSM bases and subbases [11,12].

Using stabilizing materials can improve the cold recycled mix-ture (CRM) performance. Stabilizing materials are mainly classifiedas asphalt binders (foamed asphalt and asphalt emulsion) andhydraulic binders (Portland cement, lime, and fly ash). By compar-ison of different stabilizing materials, the asphalt stabilized CRMhas a low strength at early-age. Therefore, it is easy to generate rut-ting, moisture damage, and fatigue cracking in the pavement struc-ture using the asphalt stabilized CRM layer [13]. Hydraulicstabilized materials can significantly improve the CRM durabilityand decrease the permanent deformation [4,14]. However, thecement content should be strictly controlled to avoid the shrinkagecracking. Cost, performance, and climatic conditions should beconsidered to choose the optimum stabilizing materials type[15]. It is possible to select a single binder, or a combination oftwo binders, or even a mix of several pre-dosed binders [4]. Port-land cement is the most widely used stabilizing material for CRMsin China by the balance of effectiveness and economy.

Recent efforts have been made to study the cold recycling tech-nology of semi-rigid materials. Mráz et al. [2] applied differenttypes of stabilizing materials for recycling the asphalt layer, gran-ular base layer, and cement stabilized macadam (CSM) layer. TheCRMwith asphalt emulsion and activated fly ash had a higher indi-rect tensile strength (ITS) at early-age, a similar stiffness modulus,and a similar moisture resistance compared with that with asphaltemulsion and cement. Du [3] investigated the effect of chemicaladditives on the performance of the asphalt emulsion recycledCSM mixture. Cementitious additives were recommended to beadded prior to asphalt emulsion since a higher mechanical strengthand resistance to moisture and rutting could be obtained. Ji et al.[16] reported that many factors had significant effects on thestrength of the cement stabilized CRM produced by the verticalvibration compaction method, such as the weight ratio of recycledcement base to recycled asphalt pavement, virgin aggregate (VA)content, cement content, and curing day. Cabrera et al. [17] foundthat adding the biomass bottom ash could improve the strength ofthe cement stabilized CRM and reduce the required cement con-tent. Shi et al. [18] developed an empirical relationship betweenthe ratio of recycled semi-rigid base materials to recycled asphaltlayer materials and the CRM base thickness. Yao [19] evaluatedthe surface andmechanical properties of recycled LFSM aggregates.The feasibility of using the cement stabilized CRM as pavementbases was validated. Zhang [20] designed a cement stabilizedCRM base composed of 80% recycled LFSM materials and 20% VAfor a practical project. The mixture 7-day unconfined compressivestrength (UCS) values at the cement content of 3.5% reached 4.6MPa and 5.2 MPa in the laboratory and field, respectively. It wassimilar to the mechanical strength of the CSM mixture with 100%of the VA. All these studies show the potential application prospectof CRM bases and subbases.

Although the cold recycling technology of semi-rigid materialshas been put into practical application in some regions of China,there are still some issues that should be solved urgently. In thecurrent Chinese specification JTG F41-2008 ‘‘Technical specifica-tions for highway asphalt pavement recycling” [21], although some

specific regulations in terms of the field investigation, raw materi-als selection, mix design, and construction are provided, only thestrength requirement is raised for the CRM base and subbase. How-ever, the other performance has not been considered, such as stiff-ness, moisture resistance, and shrinkage cracking resistance. Moreimportantly, there are no corresponding evaluation indicators andengineering properties criteria for reclaimed semi-rigid materials.The main objectives of this study are to evaluate the pavement per-formance of cement stabilized LFSM CRMs and to recommendengineering properties criteria of reclaimed LFSM materials forcold recycling. To accomplish it, the main engineering propertiesof waste LFSM materials from different sources were measured.After the mix optimization design, the mechanical performance,frost resistance, and shrinkage cracking resistance of the cold recy-cled LFSM mixture at three cement contents were measured in thelaboratory. The recommended control criteria of reclaimed materi-als (RM) properties and suitable layers for CRMs were proposedbased on performance correlations of the RM and CRM. The feasi-bility of the cold recycling technology was finally validated byengineering practices.

2. Experimental plan

2.1. Materials

2.1.1. Cement and VAThe composite Portland cement graded 42.5R was used as the stabilizing mate-

rial for cold recycling. The virgin limestone coarse and fine aggregates typicallyused in the local area were selected for the CRM gradation optimization. The mainengineering properties of the VA are shown in Table 1.

2.1.2. Reclaimed LFSM materialReclaimed LFSM materials were milled from distressed bases of two urban road

pavement sections located in Nantong of China. The rock type of recycled aggre-gates was limestone. Two sections had a similar original pavement structure con-sisting of a Portland cement concrete (PCC) surface of 24 cm in thickness over aLFSM base of 18 cm in thickness. After the removal of PCC slabs, the cold in-placerecycling (CIR) stabilized with cement was conducted on the LFSM base. Then, aCSM base of 18 or 20 cm in thickness and two asphalt concrete (AC) surface layersin a total of 10 or 12 cm (6 or 8 cm + 4 cm) in thickness were overlaid in turn. The 2cm design thickness variations of the CSM and AC layers between pavement sec-tions were due to variations in traffic levels and pavement elevation criteria. Themain engineering properties of the RM measured in the laboratory are also shownin Table 1. RM gradations are shown in Fig. 1.

It is found in Table 1 that RM aggregates have higher water absorption, crushingvalues, flat-elongated particle contents, and abrasion values compared with the VA.In other words, RM aggregates become worse in the compactibility, mechanicalstrength, angularity, and abrasion resistance after the long-term service and millingprocess. Deterioration of engineering properties is mainly due to the hardenedpaste of lime-fly ash hydration products coating on the surface of RM aggregates.It can be confirmed by observing the aggregate surface as shown in Fig. 2. Besides,the scanning electron microscope (SEM) test was also conducted on dry samples toobtain the microstructure surface morphology. It is clearly seen in Fig. 2 that manysmall lumps mainly containing SiO2, Al2O3, and Ca(OH)2 adhere to aggregates. Alarge number of minute air voids are formed on the surface, which is responsiblefor the higher water absorption. The hardened paste is weak and fragile. It decreasesthe strength, stiffness and abrasion resistance of RM aggregates. Moreover, manymicrocracks are formed inside the aggregate during the milling process. It alsomakes contributions to the poor strength and angularity. This may be solved afterrecycling because the added cement paste can fill voids and improve the interfacebonding. It is also observed in Table 1 that there are obvious differences of engi-neering properties between RM sources. The RM aggregate from Section 1 have ahigher strength (smaller crushing value) and a better angularity (lower flat-elongated particle content). However, the RM aggregate from Section 2 is cleaner(lower clay content) by comparison. Raw material quality, pavement damage sever-ity, milling methods, and storage environmental conditions are possible factors.

2.2. Mix design

According to the previous recycling construction experience, cement contentsof 4.0%, 4.5%, and 5.0% were selected for CRMs with 100% of RM aggregates. It isobserved in Fig. 1 that two RM gradations are greatly and differently changed undereffects of traffic loading and milling machine according to the similar design grada-tions of two pavement sections during the previous construction period. The RMgradation is too coarse for Section 1. The passing percent values of 4.76 mm,

Table 1Main engineering properties of VA and RM.

Type Size (mm) Apparent density(g/cm3)

Water absorption (%) Clay content (%) Crushingvalue (%)

Flat-elongated particlecontent (%)

Abrasion value (%)

VA-coarse 19–31.5 2.735 0.6 0.5 18.5 11.0 16.4VA-fine �4.75 2.827 2.9 0.7RM-section 1 2.653 6.3 3.3 23.9 11.9 21.1RM-section 2 2.696 7.1 0.7 32.4 17.4 20.6

0 0.6 1.18 2.36 4.75 9.5 19 26.5 37.50.0750

10

20

30

40

50

60

70

80

90

100

Pass

ing

perc

ent (

%)

Sieve size (mm)

Upper limit Lower limitSection 1-RM Section 2-RMSection 1-CRM (Adding VA) Section 2-CRM (Adding VA)

Fig. 1. Aggregate gradations of RM and CRM.

SEM 4k timesCamera photo

Fig. 2. Surface and microstructure characteristics of RM aggregates.

308 Q. Li et al. / Construction and Building Materials 169 (2018) 306–314

2.36 mm, and 1.18 mm sieves are even very close to the lower limits specified bythe current Chinese specification JTG F41-2008 [21]. On the contrary, the RM grada-tion of Section 2 is a little fine. The passing percent value of each sieve is generallylarge. Therefore, both gradations were optimized by adding the VA based on themedian of recommended gradation limits by the specification [21]. Consideringthe construction technology and thickness requirements of the CRM layer, the VAcontent added should be controlled below 20%. Finally, 18% of the fine VA and19% of the coarse VA were added to RM aggregates from Sections 1 and 2, respec-tively. Only the cement content of 5.0% was selected for the optimized CRM grada-tions as shown in Fig. 1. A total of eight types of CRMs were used in this study.

2.3. Test methods

2.3.1. Compaction testThe compaction, UCS, ITS, and freeze–thaw (F-T) tests were conducted accord-

ing to the current Chinese specification JTG E51-2009 ‘‘Test methods of materialsstabilized with inorganic binders for highway engineering” [22]. The maximumdry density (MDD) and the optimum water content (OWC) were determined bythe heavy compaction test. Cement and water were added in turn to the dry RM.The loose CRM was poured into the cylindrical mold with the inner diameter of

152 mm and the height of 120 mm after twice mixing. The compaction processwas conducted in three layers and the compaction number of each layer was 98times. After several tests on five different water contents, the MDD and OWC at agiven cement content were calculated based on the correlation curve of the drydensity and water content.

2.3.2. UCS and ITS testThe UCS and ITS tests were performed on cylindrical CRM specimens with the

diameter of 150 mm and the height of 150 mm. Specimens were compacted atthe OWC and the compaction degree of 95%. Before testing they were firstly curedin plastic bags for six days at the temperature of 20 ± 2 �C and the relative humidityof 95%. Then, they were soaked in water for one day. The crosshead displacementrate applied on both tests was 1 mm/min.

2.3.3. F-T testThe F-T test was performed to evaluate the frost and moisture resistance of

CRMs. The specimen size, fabrication method, and curing conditions were the sameas those used for the UCS and ITS tests. However, the curing time was 28 days forthe F-T test. In a F-T cycle, specimens were frozen at the temperature of �18 �C

Q. Li et al. / Construction and Building Materials 169 (2018) 306–314 309

for 16 h, followed by thawing in water at the temperature of 20 �C for 8 h. The ratioof the UCS after five F-T cycles to the unconditioned UCS before the F-T effect wasdefined as the evaluation indicator for the F-T test.

2.3.4. Dry shrinkage (DS) testThe DS test equipment was composed of a static resistance strain surveying

instrument, a signal conversion box, 12X resistance foil strain gauges with thelength of 80 mm, shielded connection lines, and a computer based data acquisitionsystem. Parallel strain gauges were pasted on the specimen surface using the epoxyglue, as shown in Fig. 3. The DS test was conducted on CRM beam specimens withthe length of 400 mm, the width of 100 mm, and the height of 100 mm at the tem-perature of 20 �C and the relative humidity of around 50%. The air conditioner wasused to assist in controlling the room temperature. The strain data was automati-cally recorded until the water content of specimens basically reached a constantvalue (almost one month).

2.3.5. Temperature shrinkage (TS) testThe TS test was performed on the same beam specimen with pasted foil strain

gauges in a high-low temperature testing chamber as shown in Fig. 4. In the cham-ber, the testing temperature reduced from 60 �C to �10 �C with the rate of 1 �C/min.The temperature interval was 10 �C. The chamber maintained a constant tempera-ture for two hours after each cooling. The temperature and strain variations werealso automatically measured.

Fig. 4. High-low temperature testing chamber.

Table 2Compaction test results.

Section RM:VA Cement content (%) OWC (%) MDD (g/cm3)

1 100%:0% 4.0 12.0 1.8171 100%:0% 4.5 12.1 1.8201 100%:0% 5.0 12.3 1.8271 82%:18% (fine) 5.0 11.6 1.9172 100%:0% 4.0 11.7 1.8502 100%:0% 4.5 12.3 1.8532 100%:0% 5.0 12.6 1.8552 81%:19% (coarse) 5.0 11.2 1.937

3. Test results and analysis

3.1. Compaction properties

The dry density and moisture content are two important factorsthat significantly affect the performance of pavement materials [1].The MDD and OWC of CRMs for different cement contents, RMsources (pavement sections), and aggregate gradations are shownin Table 2. Both parameters linearly increase with the cement con-tent for each RM source. Adding a certain percent of the fine orcoarse VA can increase the MDD and decrease the OWC of theCRM. This is caused by the lower water absorption and the higherdensity of the VA surface compared with those of the RM aggregateas explained before. Although there are evident differences in theMDD between RM sources, the OWC values remain almost con-stant (11–12%) for both sections due to the higher water absorp-tion of RMs.

3.2. Mechanical strength

The average 7-day UCS and ITS of six replicates and error rangesfor different CRMs are shown in Fig. 5. The coefficient of variation(COV) values for all cases are less than 15%, indicating a goodreproducibility. Although the UCS increases with the cement con-tent, none of CRMs with 100% of the RM can reach the UCS of3.0 MPa regardless of RM sources. For Section 1, the CRM UCS at

Fig. 3. DS test s

the cement content of 4.0% is much lower than 2.5 MPa. It onlymeets the strength requirement of subbases according to the cur-rent Chinese specification JTG F41-2008 as shown in Table 3 [21].UCS values are just around 2.8 MPa even increasing the cementcontent to 4.5% and 5.0%. These strength values are only suitablefor bases of low-level pavements. For Section 2, UCS values at allcement contents are lower than 2.5 MPa and only available forsubbases. CRMs from Section 1 always show larger UCS values ata given cement content due to the lower crushing value and flat-elongated particle content (higher strength and better angularity)

pecimens.

Fig. 5. UCS and ITS tests results.

Table 3CRM strength criteria [21].

Layer 7-day UCS (MPa)

High-level pavement Low-level pavement

Base �3.0–5.0 �2.5–3.0Subbase �1.5–2.5 �1.5–2.0

310 Q. Li et al. / Construction and Building Materials 169 (2018) 306–314

of RM aggregates compared with those from Section 2. CRMs with-out the gradation optimization have much worse strength thanCSM mixtures with 100% of the VA at the same cement content.It confirms differences in mechanical properties between the RMand VA. CRMs from different sections show different UCS increasetrends with the cement content. This is caused by high samplevariations produced during the RM reclaiming and CRM fabricatingprocesses. The type of the VA (coarse or fine) added in the CRM sig-nificantly affects the UCS. There is almost no strength change afterthe gradation optimization for Section 1, indicating that adding thefine VA does not improve the CRM strength. Moreover, excessivefine aggregates may bring on great risks in the frost and shrinkageperformance. On the contrary, the UCS has a drastic increase by126% after adding 19% of the coarse VA for Section 2. The strengthof 5.14 MPa sufficiently meets the requirement for bases of high-level pavements. By comparison, the coarse aggregate shows amuch more significant effect on the CRM strength.

Effects of different factors on the ITS are similar to those on theUCS. The ITS increases with the cement content in the form of anonlinear power law function. For CRMs with 100% of the RM,the one from Section 1 always has a larger ITS than that from

Fig. 6. F-T tes

Section 2 at the same cement content. After the gradation opti-mization the CRM ITS from Section 2 has a great increase by 67%.However, that from Section 1 slightly decreases. It confirms againthe important role of the coarse VA in the CRM strength.

3.3. Frost resistance

The F-T effect can retard or accelerate the cementitious reactionoccurring in the CRM. The cementitious reaction slows down dur-ing freezing then resumes during thawing. F-T cycles causecementing bonds to be broken, which ultimately results in reduc-tions of the stiffness and strength of cement stabilized materials[23]. The average F-T test results of three replicates are shown inFig. 6. It is observed that strength losses after five F-T cycles areunder 20% for all cases, indicating the good frost resistance ofCRMs. The UCS ratio significantly increases with the cement con-tent for both sections because more cement particles are involvedin the cementitious reaction at a long time water condition. CRMsfrom both sections exhibit similar frost resistance at a givencement content. Adding the coarse VA greatly improves the frostresistance of CRMs since the coarser gradation has a lower densityand larger voids to provide higher hydraulic conductivity and frostresistance [24]. However, the CRM optimized by adding the fine VAis more susceptible to the F-T effect. The higher water adsorptionof the fine aggregate causes more severe strength deteriorationafter F-T cycles.

3.4. DS resistance

The macroscopic volume shrinkage occurs in CRM layers due tothe reduction of the internal water content. The shrinkage iscaused by the coupling effects of capillary tension, absorbed waterintermolecular force, and interlayer water evaporation [19]. Threereplicates were conducted for each DS and TS test and the averageresults were used for analysis. Variations of the average DS coeffi-cient with the water loss content (WLC) for different CRMs areshown in Fig. 7. The average DS coefficient increases firstly thendecreases with the WLC for each case. It remains stable in the latecuring period. The peak value appears in the first 2–6 days (1.5–3.4% of the WLC). It indicates that more attention should be paidto the early curing during the CRM layer construction process. Thisphenomenon can be explained through variations of the WLC andDS strain with the curing time. On the one hand, the water contentof CRMs constantly loses with time. The WLC greatly increases inthe early curing period. The accumulated WLC in the first six daysreaches approximate 50% of the total WLC because the free waterdistributed in connected pores inside the CRM specimen is quicklydrained into the air. However, the drainage of other types of water

t results.

0

50

100

150

200

0 2 4 6 8

Aver

age

DS

coef

ficie

nt (μ

ε/%

)

WLC (%)

Cement content-4.0%Cement content-4.5%Cement content-5.0%Cement content-5.0%(with VA)

(a) Section 1

0

50

100

150

200

0 2 4 6 8

Aver

age

DS

coef

ficie

nt ( μ

ε/%

)

WLC (%)

Cement content-4.0%Cement content-4.5%Cement content-5.0%Cement content-5.0%(with VA)

(b) Section 2

Fig. 7. Variation of the average DS coefficient with the WLC.

Q. Li et al. / Construction and Building Materials 169 (2018) 306–314 311

(capillary water, absorbed water, bound water, and interlayerwater) inside the specimen is very slow. It causes the water lossspeed to gradually slowing down. On the other hand, the DS strainalso increases with the curing time. The increase rate graduallydecreases until it reaches a steady value.

Effects of cement contents, aggregate gradations, and RMsources on the total average DS coefficient of CRMs during thewhole testing process are shown in Fig. 8. The DS coefficient signif-icantly increases with the cement content for each section.Although as the cement content increases the WLC slightlyincreases, the shrinkage effect caused from the above physicalactions becomes larger and larger than the deformation constraint

0

20

40

60

80

100

120

140

3.5 4.0 4.5 5.0 5.5

Aver

age

DS

coef

ficie

nt (μ

ε/%

)

Cement content (%)

Section 1 Section 2Section 1 (Adding VA) Section 2 (Adding VA)CSM [25]

Fig. 8. Average DS coefficients.

due to the strength improvement. It results in a significant increaseof the DS strain. The OWC and WLC of CRMs from both sections aresimilar at a given cement content. However, the CRM from Sec-tion 1 with a higher strength and a coarser aggregate gradationproduces a much larger DS strain. It is mainly attributable to thehigher clay content of the RM compared with that from Section 2.Therefore, the CRM from Section 1 shows a 40–50% larger DS coef-ficient than that from Section 2. After the gradation optimization,CRMs from both sections have smaller WLC and DS strains dueto adding the clean VA with the lower water absorption and highershrinkage resistance. However, the VA type has a significant effectof the DS coefficient. Adding the fine VA causes a 10.3% increase inthe total average DS coefficient for Section 1. On the contrary, itshows a 9.3% decrease after adding the coarse VA for Section 2. Itindicates that adding appropriate amount of the coarse VA to formthe skeleton structure is also beneficial for improving the DS resis-tance of CRMs. It is worth mentioning that CRMs from Section 2 donot exhibit the significantly worse DS performance than the CSMmixtures with 100% of the VA by comparison with the previousstudy [25], as shown in Fig. 8. Contrarily, CRMs with the uncleanRM from Section 1 show the lower DS resistance.

3.5. TS resistance

Solid-liquid–gas phases in the CRM show different thermalexpansions and contractions along with the variation of the envi-ronmental temperature. Because the CRM is a poor conductor ofheat, there are large temperature differences between inside andoutside of the mixture at the initial cement hydration process. Itwill generate the thermal stress and deformation in the CRM. Insevere cases, cracks will appear [19]. Variations of the TS coeffi-cient with the testing temperature for different CRMs are shownin Fig. 9. Effects of cement contents, aggregate gradations, andRM sources on the average TS coefficient of CRMs are shown inFig. 10.

The TS coefficient corresponding to the median of the tempera-ture interval increases with the temperature. This is because thatthere are larger gaps between aggregate particles at higher tem-peratures. As the temperature decreases, aggregates are prone tobe contracted and closed. Moreover, the TS test conducted on dryspecimens excludes the opposite expansion effect caused by waterfreezing below 0 �C. The air effect is also negligible since it ismainly distributed in connected pores. Effects of cement contentsand aggregate gradations on the TS resistance are analogous tothose on the DS resistance. For both sections, the CRM with thehigher cement content has the larger TS coefficient at a given tem-perature because more cement hydration products with the highshrinkage effect are formed. Fine aggregates and fillers generallyproduce larger thermal strain than coarse aggregates. Therefore,the average TS coefficient increases by 9.9% after adding the fineVA for Section 1. However, it decreases by 21.1% after adding thecoarse VA for Section 2. Unlike the DS test results, the TS resistanceof the CRM from Section 1 is slightly better than that from Section 2at each cement content. It indicates that the higher clay content ofthe RM for Section 1 may not bring about the negative effect in thedry condition. The higher strength and coarser aggregate gradationplay more important roles in improving the TS performance forSection 1. Compared with the previous study [25], the CRM showsa little worse TS resistance than the CSM mixture with 100% of theVA, as shown in Fig. 10.

3.6. Performance correlations between RM and CRM

Qualitative relationships between RM properties and CRM per-formance were obtained from the above analysis. To quantitativelyestablish correlations, more test results from additional four

0

5

10

15

20

25

30

-10 0 10 20 30 40 50 60

TS

coef

ficie

nt (μ

ε/)

Temperature ( )

Cement content-4.0%Cement content-4.5%Cement content-5.0%Cement content-5.0% (Adding VA)

(a) Section 1

0

5

10

15

20

25

30

-10 0 10 20 30 40 50 60

TS

coef

ficie

nt (μ

ε/)

Temperature ( )

Cement content-4.0%Cement content-4.5%Cement content-5.0%Cement content-5.0% (Adding VA)

(b) Section 2

Fig. 9. Variation of TS coefficient with the temperature.

0

3

6

9

12

15

3.5 4.0 4.5 5.0 5.5

Aver

age

TS

coef

ficie

nt (μ

ε/)

Cement content (%)

Section 1 Section 2Section 1 (Adding VA) Section 2 (Adding VA)CSM [25]

Fig. 10. Average TS coefficients.

Table 4Main engineering properties of RM and CRM from additional sections.

Section RM

Waterabsorption (%)

Clay content(%)

Crushingvalue (%)

Flat-elongated particlecontent (%)

3 5.9 1.3 21.5 8.54 7.4 0.9 25.4 115 6.9 1.8 24.5 7.66 7.0 1.5 19.8 10.5

312 Q. Li et al. / Construction and Building Materials 169 (2018) 306–314

pavement sections of urban roads located in the same area werealso involved for statistical analysis. For these sections with similarengineering background and rehabilitation plans as Sections 1 and2, the main engineering properties of the RM and UCS values of theCRM without the gradation optimization were measured, as listedin Table 4. Only the strength correlation rather than other perfor-mance ones was taken into account in this study.

The analysis of variance (ANOVA) at a 95% confidence level wasperformed on datasets from the total six sections to examine thesignificance of different RM properties in affecting the CRMstrength. The Pearson product moment correlation coefficient R2

and scatter plots were also used for correlation confirmation. Thesum of squares of deviations SS, degree of freedom DF, mean squareerror MS, F-test value F-value, probability P-value, and the criticalvalue of the F-test F-critical are provided in Table 5. It is seen fromthe statistical analysis that the RM strength and angularity showsignificant effects on the CRM UCS at each cement content. Statis-tical F-values are much larger than F-critical values for the crush-ing value and flat-elongated particle content. Fairly good linearcorrelations exist between the crushing value or flat-elongatedparticle content and CRM strength. By comparison, the crushingvalue has a more significant effect and a better correlation thanthe flat-elongated particle content. Although larger statistical F-values are also obtained for the water absorption and abrasionvalue, no obvious correlations between the water absorption orabrasion value and CRM strength are confirmed after examiningthe Pearson correlation coefficients and scatter plots. Therefore,effects of the water absorption, clay content, and abrasion valueof the RM on the CRM strength are insignificant.

A nomogram shown in Fig. 11 was drawn based on the multiplelinear regression method to facilitate predicting the CRM strengthfrom RM properties. The crushing value and flat-elongated particlecontent were selected as the critical RM parameters. If they wereknown, the CRM 7-day UCS at a given cement content could bepreliminarily determined from the figure. According to the CRMstrength requirement from the current Chinese specification JTGF41-2008 [21] and the frequently used range of the cement con-tent, the locally recommended criteria for the crushing value andflat-elongated particle content of the RM without the gradationoptimization were backcalculated and shown in Table 6. Becauseit was developed on a limited amount of test data obtained fromsimilar construction projects, a further study is needed to validateand extend the findings in a wide range of milling methods, RMsources, and CRM design plans.

4. Field investigation

LFSM layers of Sections 1–6 were recycled as pavement sub-bases using CRM design plans determined from laboratory tests.In these construction projects, the optimum cement content(4.2% or 4.5%) was used for the RMwithout the gradation optimiza-tion. Firstly, the cement was paved by hand or mechanically on theLFSM layer of the old pavement after cleaning the garbage and

CRM UCS (MPa)

Abrasionvalue (%)

Cement content4.0%

Cement content4.5%

Cement content5.0%

23.2 2.49 3.02 3.3122.8 2.18 2.97 3.2017.7 2.38 3.11 3.3918.9 2.58 3.30 3.67

Table 5ANOVA test results.

Parameter Source SS DF MS F-value P-value F-critical R2

UCS (4.0%) Water absorption 63.6 5 61.6 306.2 7.9 � 10�9 5.0 0.04Clay content 6.1 5 1.3 2.7 0.1 5.0 0.01Crushing value 1593.2 5 1498.1 157.6 1.9 � 10�7 5.0 0.73Flat-elongated particle content 299.4 5 238.3 39.6 9.0 � 10�5 5.0 0.66Abrasion value 1047.5 5 1024.7 435.5 1.4 � 10�9 5.0 0.03

UCS (4.5%) Water absorption 48.8 5 45.8 173.2 1.2 � 10�7 5.0 0.02Clay content 10.3 5 4.9 9.0 1.3 � 10�2 5.0 0.07Crushing value 1511.4 5 1415.7 147.9 2.6 � 10�7 5.0 0.86Flat-elongated particle content 267.0 5 206.2 33.9 1.7 � 10�4 5.0 0.80Abrasion value 980.7 5 956.6 396.0 2.3 � 10�9 5.0 0.03

UCS (5.0%) Water absorption 42.8 5 40.0 144.9 2.8 � 10�7 5.0 0.01Clay content 12.6 5 7.0 12.7 5.2 � 10�3 5.0 0.01Crushing value 1478.7 5 1382.9 144.3 2.9 � 10�7 5.0 0.81Flat-elongated particle content 254.7 5 193.8 31.8 2.2 � 10�4 5.0 0.74Abrasion value 953.9 5 929.6 382.9 2.7 � 10�9 5.0 0.05

0

1

2

3

4

5

6

8

13

18

23

28

33

38U

CS

(MPa

)

Cru

shin

g va

lue

(%)

4.0

4.5

5.0

Flat-elongated particle content (%)6 8 10 12 14 16 18

Cement content (%)

Fig. 11. UCS Nomogram of CRM.

Table 6Locally recommended criteria for RM engineering properties.

Layer Crushing value (%)

High-level pavement Low-level pavemen

Base �12 �20Subbase �28 �28

Table 7Field test results.

SectionCement content (%)

Deflectometer (0.01 mm) Before recycling AverageRepresentative

After recycling AverageRepresentative

Design

Back-calculated modulus (MPa) Before recycling AverageRepresentative

After recycling AverageRepresentative

UCS (MPa) AverageRepresentativeDesign

Q. Li et al. / Construction and Building Materials 169 (2018) 306–314 313

water. The paving uniformity was guaranteed by drawing squareson the LFSM layer surface. Then, the recycling equipment com-posed of a CIR machine and a water wagon drove at the speed of4–8 m/min to complete a series of recycling works, includingmilling, mixing, and paving to a fixed depth. The loose CRM wascompacted and shaped by drum vibratory road rollers and a motorgrader. Finally, the CRM layer was cured under the wet and cov-ered condition for strength growth.

Surface deflections of the LFSM layer were respectively mea-sured before and after (seven days curing) the CIR constructionby the falling weight deflectometer (FWD) at an interval of 50 m.The program MODULUS 6.0 was used to back-calculate the in-situ modulus of the CRM layer from FWD data. Cylindrical samplesof the full LFSM layer were also cored from the field for visualobservation and UCS measurement. The field test results are listedin Table 7. The representative value shown in the table is the con-fidence lower limit of the arithmetic mean of a given indicator andcalculated by the following equations.

R ¼ RA � ZaS ð1Þ

L ¼ LA þ ZaS ð2Þ

Flat-elongated particle content (%)

t High-level pavement Low-level pavement

�10 �14�18 �18

1 2 3 4 5 64.5 4.5 4.5 4.5 4.5 4.2

41.1 31.7 41.9 62.2 77.7 38.066.8 67.0 99.6 92.3 197.1 70.925.3 32.3 21.8 25.6 25.3 28.242 44.2 29.3 38.9 42.6 40.2<65.0 <61.5 <63.1 <67.3 <63.1 <63.1

1237 1630 1147 765 576 1356723 630 439 511 240 5462484 1470 3157 2272 2697 18431365 1272 1902 1599 1315 1480

2.9 2.5 3.0 3.0 3.0 2.82.7 2.3 2.7 2.7 2.8 2.5�2.0 �1.5 �2.0 �1.5 �2.0 �1.5

314 Q. Li et al. / Construction and Building Materials 169 (2018) 306–314

E ¼ EA � ZaS ð3Þwhere R is the representative value of the UCS, RA is the mean valueof the UCS, Za is the reliability coefficient (1.645 at a reliability of95%), S is the standard deviation, L is the representative value ofthe deflectometer, LA is the mean value of the deflectometer, E isthe representative value of the back-calculated modulus, EA is themean value of the back-calculated modulus.

It is seen that all CRM subbases have great improvements in theoverall strength and bearing capacity. Representative values of sur-face deflections decrease by 34–78% and representative values ofthe back-calculated modulus of CRM layers increase by 0.9–4.5times for different sections after the CIR construction. The surfacedeflections completely meet design requirements. Moreover, intactfield core samples can be drilled after a good curing. The UCS rep-resentative values are 35–80% larger than the design criteria. Engi-neering practices confirm that the cold recycling of the LFSM layeras high-level pavement subbases is feasible and applicable. It evenshows great application potentials as pavement bases so long asthe aggregate gradation of the CRM is effectively optimized by add-ing the appropriate VA.

5. Conclusions

In this paper an experimental evaluation of the pavement per-formance for cold recycled LFSM mixtures is presented. Engineer-ing properties criteria of reclaimed LFSM materials for coldrecycling are recommended. Some important observations andconclusions are as follows:

(1) According to the strength criteria, all CRMs without the gra-dation optimization can only be used as bases of high-levelpavements or subbases regardless of RM sources and cementcontents.

(2) All CRMs have good frost resistance after F-T cycles. TheCRM frost resistance significantly increases with the cementcontent.

(3) CRMs show the maximumDS coefficient in the first 2–6 daysof the curing period, indicating the significance of the earlycuring condition. CRMs with the clean RM exhibit similarDS performance to the CSM mixtures with 100% of the VA.The higher cement content and RM clay content are negativefor the DS resistance.

(4) The TS coefficient of CRMs increases with the temperature.CRMs show slightly worse TS resistance than CSM mixtureswith 100% of the VA. The lower cement content and higherRM strength are beneficial for improving the TSperformance.

(5) Adding appropriate amount of the coarse VA greatlyimproves the mechanical strength, frost resistance, DS resis-tance, and TS resistance of CRMs. However, adding the fineVA makes CRMs more susceptible to the F-T damage andshrinkage cracking.

(6) The RM strength and angularity have significant effects onthe CRM strength. A nomogram is developed by the multiplelinear regression to predict the CRM strength from thecrushing value and flat-elongated particle content of theRM. The locally recommended criteria for RM propertiesare preliminarily proposed.

(7) Field investigations confirm the feasibility of the cold recy-cled LFSM layer as pavement bases and subbases.

Acknowledgements

The authors would like to acknowledge the financial supportfrom Science and Technology Project of Jiangsu ConstructionSystem (2017ZD146), Qing Lan Project, and Science ResearchProject by Nantong Nengda Construction Investment Co., Ltd(ND201600002).

References

[1] C.T. Gnanendran, D.K. Paul, Fatigue characterization of lightly cementitiouslystabilized granular base materials using flexural testing, J. Mater. Civ. Eng. 28(9) (2016) 04016086.

[2] V. Mráz, J. Valentin, J. Suda, L. Kopecky, Experimental assessment of fly-ashstabilized and recycled mixes, J. Test. Eval. 43 (2) (2015) 264–278.

[3] S. Du, Influence of chemical additives on mixing procedures and performanceproperties of asphalt emulsion recycled mixture with reclaimed cement-stabilized macadam, Constr. Build. Mater. 118 (2016) 146–154.

[4] J. Abdo, J.P. Serfass, P. Pellevoisin, Pavement cold in-place recycling withhydraulic binders: the state of the art in France, Road Mater. Pavement 14 (3)(2013) 638–665.

[5] F.P. Majarrez, Semi-rigid pavement performance and construction techniquesfor semiarid areas, Road Mater. Pavement 14 (3) (2013) 615–637.

[6] A. Mohammadinia, A. Arulrajah, J. Sanjayan, M.M. Disfani, M.W. Bo, S.Darmawan, Stabilization of demolition materials for pavement base/subbaseapplications using fly ash and slag geopolymers: laboratory investigation, J.Mater. Civ. Eng. 28 (7) (2016) 04016033.

[7] R. Lundstrom, R. Karlsson, L.G. Wiman, Influence of pavement materials onfield performance: evaluation of rutting on flexible, semi-rigid and rigid testsections after 7 years of service, Road Mater. Pavement 10 (4) (2009) 689–713.

[8] F. Netterberg, M. de Beer, Weak interlayers in flexible and semi-flexible roadpavements: part 1, J. S. Afr. Inst. Civ. Eng. 54 (1) (2012) 31–42.

[9] D.X. Xuan, A.A.A. Molenaar, L.J.M. Houben, Shrinkage cracking of cementtreated demolition waste as a road base, Mater. Struct. 49 (1–2) (2016) 631–640.

[10] R.B. Saldanha, N.C. Consoli, Accelerated mix design of lime stabilized materials,J. Mater. Civ. Eng. 28 (3) (2016) 06015012.

[11] X. Li, X. Yin, B. Ma, J. Huang, J. Li, Cement-fly ash stabilization of cold in-placerecycled (CIR) asphalt pavement mixtures for road bases or subbases, J. WuhanUniv. Technol. 28 (2) (2013) 298–302.

[12] A. Nataatmadja, Some characteristics of foamed bitumen mixes, Transp. Res.Rec. 1767 (2001) 120–125.

[13] T. Quick, W.S. Guthrie, Early-age structural properties of base material treatedwith asphalt emulsion, Transp. Res. Rec. 2253 (2011) 30–40.

[14] A. Kavussi, A. Modarres, Laboratory fatigue models for recycled mixes withbitumen emulsion and cement, Constr. Build. Mater. 24 (10) (2010) 1920–1927.

[15] A. Modarres, P. Ayar, Comparing the mechanical properties of cold recycledmixture containing coal waste additive and ordinary Portland cement, Int. J.Pavement Eng. 17 (3) (2016) 211–224.

[16] X. Ji, Y. Jiang, Y. Liu, Evaluation of the mechanical behaviors of cement-stabilized cold recycled mixtures produced by vertical vibration compactionmethod, Mater. Struct. 49 (6) (2016) 2257–2270.

[17] M. Cabrera, F. Agrela, J. Ayuso, A.P. Galvin, J. Rosales, Feasible use of biomassbottom ash in the manufacture of cement treated recycled materials, Mater.Struct. 49 (8) (2016) 3227–3238.

[18] F. Shi, Y. Chen, D. Liu, X. Li, Adaptability of cement stabilized full-depth in situregeneration base, J. Univ. Shanghai Sci. Technol. 39 (2) (2017) 188–193.

[19] K. Yao, Study on lime-fly ash macadam semi-rigid base reused in rigid baseMaster thesis, Shandong University, Jinan, 2009.

[20] Y. Zhang, Research on pavement regeneration technology on Yongyu line inJiangbei district of Ningbo Master thesis, Zhejiang University, Hangzhou, 2014.

[21] Research Institute of Highway, Technical Specifications for Highway AsphaltPavement Recycling (JTG F41–2008), China Communications Press, Beijing,2008.

[22] Research Institute of Highway, Test Methods of Materials Stabilized withInorganic Binders for Highway Engineering (JTG E51–2009), ChinaCommunications Press, Beijing, 2009.

[23] M.G. Rosa, B. Cetin, T.B. Edil, C.H. Benson, Freeze-thaw performance of fly ash-stabilized materials and recycled pavement materials, J. Mater. Civ. Eng. 29 (6)(2017) 04017015.

[24] Y. Xiao, E. Tutumluer, Y. Qian, J.A. Siekmeier, Gradation effects influencingmechanical properties of aggregate base granular subbase materials inMinnesota, Transp. Res. Rec. 2267 (2012) 14–26.

[25] Z. Li, Research on the performance of cement stabilized aggregate mixture andindexes correlation Master thesis, Southeast University, Nanjing, 2007.