Construction and Building Materials 254 (2020) 119234

Contents lists available at ScienceDirect

Construction and Building Materials

journal homepage: www.elsevier.com/locate/conbuildmat

Laboratory investigations on the mechanical properties of cement

treated RAP-natural aggregate blends used in base/subbase layers of
pavements
Muhammad Arshad 1
Department of Geological Engineering, University of Engineering & Technology, G.T Road, Lahore, Pakistan

h i g h l i g h t s

Cement dosage increase the drying shrinkage potential of the blended samples (BS).

1.5-4.5% cement dosage substantially improves the mechanical properties of the BS.

Cement dosage can considerably reduce the accumulative strains in the BS.

A regression model is presented for the estimation of compressive strength of the BS.

a r t i c l e i n f o a b s t r a c t

Article history: The use of reclaimed asphalt pavement (RAP) matrials in road construction, maintenance, and rehabili-
Received 29 September 2019 tation projects has a number of economic and environmental benefits. This study evaluates the perfor-
Received in revised form 7 April 2020 mance of cement-treated blends prepared by mixing the RAP content with fresh granular (G) material
Accepted 16 April 2020
in different percentages and for different curing periods. For this purpose a comprehensive laboratory
testing campaign was executed considering the resilient modulus tests, unconfined compressive strength
tests, long-term cyclic tri-axial load tests, creep loading tests, and drying shrinkage tests on the blended
Keywords:
samples. Analysis of the high-quality test result data demonstrates that applying even a small cement
Pavement structure
Reclaimed asphalt pavement
dosage, in the range of 1.5–4.5%, to the blended samples can considerably improve their resilience and
Material properties strength properties in addition to limiting the accumulative strains occurring under the applied cyclic
Empirical correlation and sustained creep loading. However, drying shrinkage of the blended sample may be an adverse effect
Laboratory experimentation due to the addition of cement content. Additionally, statistical analysis of experimental results shows
that resilience and strength properties of the cement-treated blends are significantly dependent on the
blend composition and curing period, in addition to cement dosage. A regression model is presented
for the estimation of compressive strength of the blended samples, which simultaneously takes into
account the blend composition, curing period, and cement dosage.
Ó 2020 Elsevier Ltd. All rights reserved.

1. Introduction RAP materials are generated each year globally but suffer from
poor practices and standards for the management of such gigantic
Reclaimed asphalt pavement (RAP) material is generated when quantities [3]. Reuse of this waste in asphalt mixtures for worn
the surface layer of the asphalt pavements are removed for rehabil- surfaces has perhaps become standard practice in many countries,
itation or resurfacing, or to obtain access to buried utilities such as especially in technologically advanced regions. This is because the
water and natural gas supplies, sewerage pipelines etc. Usually, use of RAP materials has demonstrated that it is a good alternative
this RAP material consists of good-quality, well-graded aggregates to using fresh granular aggregates due to the reduction in the cost
coated by an asphaltic layer, as reported by Ransinchung et al. [1] of the projects along with their environmentally friendly aspects
and documented by Sing et al. [2] among others. Millions of tons of [4–7]. Many researchers, including Thenoux et al. [8] and Liu
et al. [9], have reported that the recycling of asphalt pavement
may provide energy savings, economic payoffs, and increased con-
1
Formerly Graduate Student, Department of Civil Engineering, McMaster Univer- servation of natural resources. Despite such benefits, the main use
sity, Hamilton, Canada.
E-mail address: [email protected]
of RAP materials is limited to the surface course of the pavement

https://doi.org/10.1016/j.conbuildmat.2020.119234
0950-0618/Ó 2020 Elsevier Ltd. All rights reserved.
2 M. Arshad / Construction and Building Materials 254 (2020) 119234

structure and a noticeable surplus quantity of RAP is left unused
The long-term performance of a particular blended sample trea-
each year. During the last two decades, many studies have been ted with particular cement contents and subjected to 20,000
conducted with the goal of incorporating this excessive quantity load cycles after three different curing periods.
of RAP into the base/sub-base layers of flexible pavements.
However, the use of RAP materials in these layers is limited to 2. Material characterisation
20–30% by weight, with the rest being composed of fresh granular
materials, as reported by many researchers and practitioners [4– For this research, blends were prepared by mixing different per-
12]. Such limitations primarily occur since excessive inclusion of centages (by mass) of one type of RAP and one type of fresh gran-
RAP materials in the base/sub-base layer may prove to be detri- ular (G) sample with ordinary Portland cement (OPC). The natural
mental for the serviceable life of the road pavement, especially (fresh) granular sample, as well as the RAP materials, contained
when heavy traffic loading is anticipated. More recently, Arshad crushed limestone particles of angular to sub-angular shape. The
and Ahmed [10] and Arshad [11,12] have documented (based on flat and elongated particles in the samples/materials were not
experimental studies) that the addition of RAP contents to fresh more than 6%, as per ASTM D 4791. The specific gravity of the fresh
granular materials increases the resilient properties of the blends. granular and cleaned RAP samples varied from 2.60 to 2.65 as per
At the same time, such blended samples demonstrate significantly ASTM C 127.
higher accumulative deformation under repeated loading. This was Table 1 shows the matrix of the testing program designed for
observed with more clarity for the blended samples containing the comprehensive testing campaign including: resilient modulus
more than 50% RAP contents [1012]. tests, unconfined compressive strength tests, long-term cyclic tri-
Substitution of RAP with natural aggregates often reduces the axial load tests, creep loading tests, and drying shrinkage tests on
structural capacity of the base/subbase layers of flexible pavement, the blended samples.
and the use of 100% RAP material does not produce a high-quality Natural granular sample and RAP materials were air-dried in
base/subbase layer [13,14]. To improve the performance of base/ the laboratory at the ambient laboratory temperature (typically
subbase layers containing RAP, different types of stabilising agents 23 ± 2 °C). These samples were then thoroughly mixed using a
(additives) such as cement [15–20], fly ash [21,22], and slag [23] standard riffle box. A suitable quantity of each blend (0%, 25%,
have been studied. Cement stabilisation is a traditional technique 50% and 75% RAP) without the cement dosage were kept in impen-
generally used to stabilise low-quality pavement aggregates [23] etrable plastic bags until they were used for the different types of
and most researchers have recommended a 1–6% content of testing involved in this study.
cement by weight of blended aggregates [24–27]. What has gener- It should be noted that different procedures and/or criteria are
ally been found, so far, shows that cement-treated RAP mixtures adopted for the characterisation of RAP materials in different coun-
seem to have some improved mechanical behaviour under applied tries/region is quite varied; however, particle size distribution
loading. At the moment, however, the long-term behaviour under (gradation analysis) is the most commonly used among the differ-
creep loading, dry shrinkage, resilient performance and strength ent accepted standards. Nevertheless, a comprehensive portrayal
characteristics of these mixtures are still not clear owing to the of the RAP would involve a series of testing comprising, but not
limited number of studies performed on the subject. limited to, the evaluation of particle sizes along with the shape
The primary objective of this experimental study is to explore description of recovered aggregates from the RAP material, the
and evaluate the strength, resilience and long term performance determination of active binder amounts and its physical properties
under cyclic and creep loading of cement-treated blends prepared in the RAP, and the cleanliness and homogeneity of the RAP
by mixing RAP materials with fresh granular (G) samples in speci- [28,29].
fied proportions. More specifically, this study explores:
2.1. Particle size distribution

The effect of cement dosage and curing period on resilient mod-
ulus (MR), unconfined compressive strength (UCS), and indirect Fig. 1 shows the particle size distribution curves for the fresh
tensile strength (ITS) of the blended samples containing differ- granular and RAP material used for this research. Each curve corre-
ent percentages of RAP materials in the mixture; sponds to the average of three independent tests performed

The effect of cement dosage on drying shrinkage and sustained according to the AASHTO Designation T27-99 [30]. On observing
creep loading on the blended samples containing different per- the gradation curves it can be interpreted that the granular mate-
centages of RAP materials in the mixture; rial was coarser than the RAP sample.

The pressure sensitivity of the blended samples under a range of The major gradation properties of the materials, including the
confining pressure, cyclic deviator stress, curing periods, and coefficient of uniformity (Cu) and coefficient of curvature (Cc), have
cement dosage, and; been summarised in Table 2. The gradation properties for the other

Table 1
Matrix of the testing program.
§
Fresh Number of Number of Number of Resilient Unconfined Indirect Dry shrinkage Creep load Long term cyclic
granular RAP dosage cement curing modulus compressive tensile test test triaxial loading
sample dosage periods test strength test strength test test
1 5 – – 15=5 – – – –
o @
1 5 3 3 153 153 153 *1  4  3  123 11
 3 = 45 3 = 45 3 = 45 3 = 36 1=6 32=6
Total 50 45 45 36 6 6

Note: Blended samples were prepared by mixing 0%, 25%, 50%, 75% and 100% (by mass) of RAP with the natural granular sample. Cement dosage of 1.5%, 3.0% and 4.5% were
made with reference to the weight of the total blend prepared.
§
including 0% RAP content.
* only 4 different percentages of RAP contents were considered.
o
only 2 different percentages of RAP contents and 1 curing period was considered.
@
only 1 type of blend along with 3 curing periods and 2 loading scenarios were considered.
 M. Arshad / Construction and Building Materials 254 (2020) 119234 3

cated the compaction characteristics of the materials. With the

fresh granular samples, the optimum moisture contents (wopt)
were measured at 6.25%, and the maximum dry unit weight was
observed to be 23.3 kN/m3. RAP measurements for the same crite-
ria were 7% and 21.5 kN/m3. The MPC tests were not performed for
the sample blends of fresh granular and RAP materials. It is
assumed that their wopt values would be in the range of 6–7.5%
with a dry weight of between 21.5 and 23.5 kN/m3.

3. Resilient modulus testing

The resilient modulus (MR) for pavement construction materials

is the ratio of the deviator stress to resilient strain as depicted in
Fig. 2. The MR value depends upon the loading sequence of the
specific test method applied, which is in turn dependent on the
gradation characteristics of the test specimen. For the study under
Fig. 1. Particle size distribution curves for fresh granular and RAP sample.
consideration, the test procedure was adopted as per AASHTO
T307-99 [32]. This also elaborates the sample preparation tech-
niques and the number of load repetitions for a particular load
Table 2 amplitude [12].
Particle size distribution characteristics of the fresh granular sample and RAP
materials.
3.1. Methods for sample preparation and the curing procedure
Material Fresh granular sample RAP material
D10 0.6 0.45 All of the test samples were compacted using an AASHTO T307-
D30 5 1.2
99 compliant vibration device [34]. A split-able mould with an
D50 15 3.5
D60 18 4.75 inside diameter of 152 mm (6 in.) and a height of 305 mm
Cu 30.06 10.56 (12 in.) was used for the preparation of test samples for the resili-
Cc 2.31 0.67 ent modulus tests and the others tests included in the testing cam-
%Fines 5 3 paign related to strength, creep and long term cyclic tri-axial
% Sand Size 25 57
loading tests. A split-able mould lined with a rubber membrane
%(4.75–9.5)mm 9 25
was placed on the base of the tri-axial cell and a vacuum was
applied to make a sealed contact between the membrane and the
specific blended samples were obtained according to the percent- inside surface of the mould.
age of granular materials and RAP making up that particular blend. The moisturised granular samples were compacted into 5 layers
in the mould using a vibratory hammer following the requirement
2.2. Physical properties of recovered bitumen binder of AASHTO T307-99 [32]. Before the start of the test, the sample
was wrapped with a second confining membrane to confirm that
To separate the aggregates from the binder agent (bitumen), the the accurate confining pressure could be held throughout the test-
RAP samples were put into the oven at a temperature of 140 °C. ing process.
These samples were washed with tap water to eliminate any impu- To estimate the level/amount of moisture required for compact-
rity from the samples. A rotational viscometer apparatus was used ing the MR test samples, the optimum moisture contents for the
to extract the binder contents by applying the solvent extraction samples (OMC or wopt), identified from the modified proctor test
technique. The recovered bitumen contents were found to be results were used. During the experiment itself, the compaction
approximately 4.3%. The softening point, as per ASTM D36-06, method and energy required for assembling the MR test samples
was 62 °C. Penetration at 25 °C, as per ASTM D5-06, was 52dmm, had to be changed from the parameters obtained from the modi-
and viscosity at 60 °C (as per ASTM D4402-06) was 23,500 P. fied proctor test, as some of the aggregates were unable to main-
tain their moisture content when subject to the vibrations
2.3. Physical and chemical properties of ordinary Portland cement applied to them during their preparation. This may result in a
(OPC) slight reduction of moisture content in the samples prepared for
the MR test. It is presumed, however, that this will have an insignif-
The Blaine values (fineness) of the used OPC was 338 m2/kg and icant effect on the final test results.
its specific gravity was 3.14. The mean particle size of the OPC was The cylindrically moulded samples were placed open-ended
found to be 4 lm. The mineralogical composition of the OPC (as per (from top and bottom) in a humidity room for the required curing
claimed by the manufacture) shows that it contains 59.6% C3S, period. The position of the cylindrical samples was also changed at
12.6% C2S, 7.86% C3A and 9.49% C4AF. The major chemical compo- least twice in a 24 h period in an attempt to avoid the concentra-
nent in the OPC was calcium oxide (CaO) being 63.7% followed by tion of moisture building up along one side of the sample. The tem-
silicon dioxide (19.7%) and the rest of the other components. Its 3, perature of the humidity room was maintained at 23 ± 2 °C along
7 and 28 days compressive strength was 25, 36 and 56 MPa, with relative humidity (RH) of greater than 90%. The specimens
respectively. were tested at ambient laboratory temperatures (typically
23 ± 2 °C) and RH of 55 ± 15%.
2.4. Modified proctor compaction test
3.2. Resilient modulus testing procedure
A Modified Proctor Compaction (MPC) test (AASHTO T180) [31]
was performed on all of the materials, with particles larger than The cylindrical test samples were subjected to cyclic axial stress
19 m being removed before testing. The results of the tests indi- at a particular confining stress within a conventional tri-axial cell.
4 M. Arshad / Construction and Building Materials 254 (2020) 119234

Fig. 3. A typical shape of applied repeated load cycles and the generated
deformation curve for the test specimen.

type of blend considered. The variation in MR value with bulk

stress can be approximated through power functions, demonstrat-
ing a very strong coefficient of determination (R2), typically in the
range of 0.97–0.99. On average, MR values were found to increase
with RAP contents at a particular loading combination and number
of load cycles.
For instance, the average MR value for fresh granular samples,
with the applied bulk stress range of 83–690 kPa, was found to
be 188 MPa. The relevant values for the blended samples contain-
ing 25%, 50% and 75% RAP contents were 211 MPa, 233 MPa, and
269 MPa. This surge in the MR values equates to an increase of
Fig. 2. A schematic view of elastic & plastic stain and definition of the resilient 12%, 24% and 43% respectively when equated to the MR values
modulus under repeated loading [33]. obtained for the fresh granular sample (188 MPa). These results
closely match findings presented by many researchers, including
For this research, the resilient modulus tests were carried out in
MacGregor et al. [35], Alam et al. [36], Bennert and Maher [34],
accordance with AASHTO T 307-99 [32]. In this standard, to simu-
and Kim and Labuz [37].
late actual traffic loading, a haversine-shaped loading waveform
must be used. The particular load-cycle of this waveform applies
the designated load for 0.1 s and the sample remains unloaded 3.3.2. Effect of cement dosage on MR values for a particular blended
for 0.9 s. sample and curing time
Fig. 3 depicts a typical repetitive load/stress pulse and the resul- Fig. 6 shows the effect of cement dosage on the MR values of a
tant residual deformation curve for a certain time period. AASHTO particular blended sample which was cured for a certain period.
T 307-99 [32] defines the loading sequence as consisting of a con- From this figure, it can be inferred that the addition of cement con-
ditioning stage of 750 load cycles, followed by 15 stages of 100 tent has a marked effect on the MR values of the blended sample
load cycles each. Each of the 15 stages has a different set of param- over the range of applied bulk stress values during the MR tests.
eters for the confining, maximum axial and cyclic deviator stress; For instance, Fig. 6(a) describes the effect of cement treatment on
this is shown in Table 3. a fresh granular sample which has been cured for one day before
To minimise the build-up of pore pressure, the drainage valve the MR testing. It can be concluded that the MR value increased
was kept open throughout the cyclical loading stages. There are with the addition of cement content. For the bulk stress at the
further procedural details for the tests including: sample installa- lower value, the increase in MR values is not so pronounced, how-
tion technique, electro-mechanics of the loading system (servo- ever, at the higher end the increase in MR values are clearly promi-
controlled electro-hydraulic material testing system (MTS) as nent. More specifically, at rb = 200 kPa the MR values for the blend
shown in Fig. 4), description of the load cells used and the linear treated with 0%, 1.5%, 3% and 4.5% of cement are 147 MPa,
variable differential transducers (LVDTs), and specifics details of 152 MPa, 160 MPa and 168 MPa, which in turn implies an increase
the data acquisition system. These details can be further investi- of 3.4%, 8.8% and 14.2% respectively over the control value obtained
gated in numerous publications [10,11]. The calculation of the resi- at 0% addition of cement content. The corresponding increase at
lient moduli was performed from the readings of the LVDT rb = 600 kPa rises to 15.3%, 20.5% and 27.6%.
(installed to measure the axial deformation) and the load cell for Fig. 6(b) portrays the effect of cement dosage in the range of 0–
the last five cycles of each loading combination. 4.5% on MR values for the blend prepared by mixing 75% fresh
granular (G) with 25% RAP contents. The prepared samples were
3.3. MR test results and discussion then cured for three days before the MR tests. As in the previous
case, with Fig. 6(a), the MR values not only increased with the bulk
stress value but with the addition of cement contents as well. For
3.3.1. Effect of RAP contents on MR values of the blended samples instance, at rb = 400 kPa, the MR value for the untreated blend
Fig. 5 explores the effect of RAP contents on the MR values of the (0% cement) is 280 kPa, however, this value rises to 320 kPa,
blended specimens prepared by adding 25%, 50% and 75% of RAP 380 kPa and 420 kPa when the blend was treated with 1.5%, 3.0%
(R) contents with natural granular sample (G). The control MR tests and 4.5% of cement content. These surges are equivalent to approx-
were conducted just after the preparation of the sample without imately 14.2%, 35.7% and 50.0% for the respective addition of
any curing time or cement dosage. This figure shows that the MR cement contents with reference to the values obtained for the 0%
value surges with bulk stress (rb = 2r3 + rt) irrespective of the cement content.
 M. Arshad / Construction and Building Materials 254 (2020) 119234 5

Table 3
The loading sequence for the resilient modulus test as per AASHTO T 307 protocol.

Seq. No. No. of load Appl. Confining stress Max. axial stress Cyclic deviator axial Contact stress Total axial stress Bilk stress
stress
r3 rmax rd 0.1rmax rt rb = 2r3 + rt
kPa kPa kPa kPa (kPa) (kPa)
0 750 103.4 103.4 93.1 10.3 206.8 414
1 100 20.7 20.7 18.6 2.1 41.4 83
2 100 20.7 41.4 37.3 4.1 62.1 104
3 100 20.7 62.1 55.9 6.2 82.8 124
4 100 34.5 34.5 31 3.5 69 138
5 100 34.5 68.9 62 6.9 103.4 172
6 100 34.5 103.4 93.1 10.3 137.9 207
7 100 68.9 68.9 62 6.9 137.8 276
8 100 68.9 137.9 124.1 13.8 206.8 345
9 100 68.9 206.8 186.1 20.7 275.7 414
10 100 103.4 68.9 62 6.9 172.3 379
11 100 103.4 103.4 93.1 10.3 206.8 414
12 100 103.4 206.8 186.1 20.7 310.2 517
13 100 137.9 103.4 93.1 10.3 241.3 517
14 100 137.9 137.9 124.1 13.8 275.8 552
15 100 137.9 275.8 248.2 27.6 413.7 690

A similar trend of increase in the MR values due to the addition

of cement content was observed for the other blends included in
this analysis, such as 50%G and 50% RAP, and 25%G and 75% RAP,
as shown in Fig. 6(c) and (d) respectively.

3.3.3. Effect of curing time on MR values for a particular blended

sample and cement dosage
Fig. 7 shows the effect of curing time on the MR values for the
particular blends treated with a specific percentage of cement con-
tents. From this figure, it can be observed that MR values may
increase substantially in line with the curing period that is applied
to the blends which contain a certain percentage of cement. Fig. 7
(a) depicts the behaviour of the 100%G sample treated with a 2%
cement content and cured for periods of one, three and seven days.
After one day of curing the average MR value over the considered
range of bulk stress, 84–690 kPa remained at approximately
207 MPa. For the samples tested after three and seven days of cur-
ing, the corresponding MR values reached up to 250 MPa and
325 MPa, i.e. an increase of 20% and 57% could be observed after
the respective curing periods of three and seven days. At the lower
Fig. 4. MTS loading system and other accessories used during the testing campaign. end of the bulk stress range, the MR value was 90 MPa for the sam-
Note: 1-Load frame; 2-Load cell; 3-LVDT; 4-Vibratory compaction device; ple tested after only one day of curing, which rose to 108 MPa and
5-Triaxial cell; 6-Test specimen; 7-Split mould; 8-Display unit (laptop computer); 140 MPa for the samples tested after three and seven days of cur-
9-Data acquisition system. ing. At the higher end of bulk stress, i.e. 690 kPa, the MR value
changed from 350 kPa to 421 kPa, and then 548 kPa, when MR test-
ing was conducted after one, three and seven days.
Fig. 7(b) describes the performance of 50%G and 50%RAP sam-
ples treated with a 4% cement content and cured for the periods
of one, three and seven days. The average increase in MR value over
the considered range of bulk stress 84–690 kPa was 30% and 90%
for the samples tested after the curing periods, compared to the
MR value obtained after one day of curing. At the lower end of
the bulk stress (84 kPa) the MR values were found to be
146 MPa, 190 MPa and 303 MPa after a curing period of one, three
and seven days. The relevant values were 418 MPa, 544 MPa and
790 MPa at the upper end of the bulk stress (690 kPa) applied dur-
ing the MR testing. A similar increasing trend in the MR values was
observed due to an increase in curing period for the blend contain-
ing 25%G and 75%RAP and stabilised at 6% of cement contents, as
shown in Fig. 7(c).

3.3.4. Statistical analysis of the MR values

Statistical analyses of the MR values obtained during the resili-
Fig. 5. Effect of RAP contents on the MR values of the blended samples. ent modulus tests were conducted considering:
6 M. Arshad / Construction and Building Materials 254 (2020) 119234

(a) (a)

(b)
(b)

(c)
(c)

Fig. 7. Effect of curing time on MR value for a particular blended sample and cement
dosage.

Variations in RAP content;

Variations in cement content;

Variations in the curing period.

The focus of the statistical analyses was to establish whether

the investigated test conditions produced a significant effect on
the MR value or not, although there was a noticeable change in
(d) the MR values. For this purpose, a t-test was performed on the data
set comprising of MR values obtained during the investigated test
conditions, as discussed earlier. A two-tail t-test was performed
using paired values of the two data sets at an alpha value of 0.05
Fig. 6. Effect of cement dosage on MR value for a particular blended sample and
curing time. (i.e. 5% level of significance for the two-tailed test, which is, in fact,
 M. Arshad / Construction and Building Materials 254 (2020) 119234 7

Table 4
Results of the t-tests based on MR values.

Testing scenario Comparison based on t-Critical t-Statistics Remarks (difference)

0 day curing & 0%OPC 100%G vs 50%G + 50%R 2.14 13.33 significant
100%G vs 25%G + 75%R 2.14 15.74 significant
100%G & 1 day curing 0%OPC vs 1.5%OPC 2.14 5.40 significant
0%OPC vs 3%OPC 2.14 6.10 significant
75%G + 25%R & 3 day curing 0%OPC vs 1.5%OPC 2.14 8.56 significant
0%OPC vs 3%OPC 2.14 10.54 significant
0%OPC vs 4.5%OPC 2.14 12.89 significant
50%G + 50%R & 7 day curing 0%OPC vs 1.5%OPC 2.14 10.94 significant
0%OPC vs 3%OPC 2.14 13.11 significant
0%OPC vs 4.5%OPC 2.14 15.38 significant
100%G & 1.5% OPC 1 day curing vs 3 day curing 2.14 8.22 significant
3 day curing vs 7 day curing 2.14 8.75 significant
1 day curing vs 7 day curing 2.14 8.88 significant
50%G + 50%R & 3% OPC 1 day curing vs 3 day curing 2.14 11.64 significant
3 day curing vs 7 day curing 2.14 13.72 significant
1 day curing vs 7 day curing 2.14 14.20 significant
25%G + 75%R & 4.5% OPC 1 day curing vs 3 day curing 2.14 16.58 significant
3 day curing vs 7 day curing 2.14 17.32 significant
1 day curing vs 7 day curing 2.14 17.05 significant

the probability of rejecting the null hypothesis when it is true). 15 different loading combinations applied sequentially as elabo-
This corresponds to a 95% confidence level which is a normally rated in AASHTO T307-99[32] for base/subbase materials. The
accepted analysis of this type [10,38]. The null hypothesis for the curves in the figure only reflect the history of residual deformation
t-test was assumed due to a significant difference between the in resilient modulus testing. The following observations can be
MR values obtained for the compared scenarios (i.e. t-statistic > drawn from analysing the data in Fig. 8:
t-critical), while the alternate hypothesis suggests that there was
insignificant difference between the MR values (i.e. t-statistic < 1. The addition of RAP content into the granular material produces
t-critical). Results of the t-tests are presented in Table 4. a noticeable increase in residual strain for the same stress his-
Table 4 demonstrates that t-statistic > t-critical for all the tory. Primarily, with an increase in the RAP content, the residual
scenarios/conditions presented in Table 4, which in turn implies deformation increases. For instance, the axial residual strain at
that there was a significant difference between the MR values the 900th load cycle is limited to 0.26% for the specimen that
obtained for scenario/condition considered. only contained a fresh granular sample, while at the same level
of cyclic load history the relevant value of axial residual strain
becomes 0.45%, 0.82% and 1.02% for the blends containing
3.3.5. Effect of curing time and cement dosage on residual axial strain 25%, 50% and 75% of RAP material. This element of increase in
during MR testing axial residual strain is undesirable even though the RAP mate-
Resilient modulus is one property amongst the other important rial has increased the MR value favourably;
properties a material requires for pavement design, as per the 2. For a particular value of confining pressure (cell pressure) r3,
Mechanistic-Empirical design approach. Based on the MR test the residual strain after 100 load repetitions increased with
results as discussed in Section 3.1, 3.2 and 3.3, it is obvious that the cyclic deviator stress amplitude. For example, for the load
the MR values of the blended samples containing fresh granular cycles 600–700, 700–800, and 800–900 (when rd/r3 = 0.90,
and RAP material increased in conjunction with an increase in 1.80 and 2.70), the incremental residual strain of the specimen
the percentage of RAP content. However, this increase in MR value containing 50% G and 50% R was 0.03%, 0.10%, and 0.47% respec-
should also be studied while considering the residual deformation tively. This situation, in turn, implies that the incremental resid-
occurring under the applied cyclic loading. This is due to the fact ual strain of the specimen has increased by a factor of fifteen
that pavement performance is directly linked to the accumulative
deformation trend of pavement materials. More precisely, the
accumulative residual deformation of pavement materials controls
rutting of pavement surface due to millions of load cycles during
the life span of the pavement structure. This situation clearly
demands an investigation or clarification on the effect of RAP on
the residual deformation of granular–RAP blends.
This section of the paper elaborates upon the effect of RAP con-
tents on accumulative deformation of different blends subjected to
resilient modulus testing. Although only 100 load cycles were
applied with specific stress states of confining pressure and the
cyclic deviator stress, as mentioned in Table 3, the effect of RAP
content, cement dosage, and curing period on residual deformation
can still be considered, at least qualitatively, representative.
Fig. 8 depicts the development of axial residual strain in fresh
granular (G) and it’s blending with 25%, 50% and 75% RAP material
during the resilient modulus testing. In this figure, the horizontal Fig. 8. The evolution of axial residual strain in fresh granular (G) and its blended
axis presents the total number of load repetition applied during samples during the resilient modulus testing.
8 M. Arshad / Construction and Building Materials 254 (2020) 119234

when the cyclic deviator stress amplitude was increased three- 1500 load cycles was 1.76%, which reduced to 1.23% and
fold. A roughly similar trend in residual strain evolution was 0.92% when the same blend was treated with 3% and 4.5% of
observed for the other blends mentioned in Fig. 8 under the cement content and tested after three days of curing (Fig. 9
given range of confining pressure and cyclic deviator stress (a)). For the pure granular sample (100% G) the corresponding
amplitude as specified in Table 3; value of residual strain is reduced to 0.24% and 0.19% with 3%
3. Apparently, the increase in MR value with a correlating increase and 4.5% of cement content respectively (having been tested
in residual strain is a contradictory development. However, one after three days of curing). It should be noted that the untreated
should realise that the MR value is the instantaneous slope of pure granular sample (100% G) yielded a value of 0.39% residual
the stress–strain curve at the end of a particular loading sce- strain after the same number of load repetitions under the same
nario comprising of 100 load cycles. From the author’s perspec- loading scenario;
tive, during the last few load cycles (from a series of 100 load 2. The increase in curing period may also stabilise the blend
cycles) the specimen achieves a relatively firm and stabilised noticeably. For the blend containing 25%G and 75%RAP and
state, and hence the instantaneous slope of the stress–strain treated with 3% cement content, the value of residual strain
curve is increased. Furthermore, RAP particles have a thin coat- dropped from 1.23% to 0.74% when the curing time was
ing of asphalt, which is a viscous material and the resilient increased from three to seven days. These figures, in turn, imply
modulus test being dynamic in nature, the viscous asphalt a reduction in residual strain value of 39% due to the additional
binder coating around the aggregate particles tended to provide four days curing. For the same blend when treated with 4.5%
higher resistance to deformation. cement content, the reduction in the residual strain value was
32% due to the additional four days of curing.
Fig. 9(a) and (b) reveal the effect of curing time and cement
dosage on the residual strain for the blends presented in Fig. 7.
The above remarks are typically applicable for all other combi-
Fig. 9(a) exclusively presents the evolution of residual strain when
nations of granular samples and RAP as shown in Figs. 8 and 9.
the relevant blends were treated with a cement dosage of 3% and
4.5%, and tested after a curing period of three days, whereas
Fig. 9(b) shows results for the curing period of seven days with 4. Unconfined compressive strength (UCS)
the same percentages of cement dosage
Careful consideration of the data presented in Figs. 8 and 9 UCS tests were performed on cement-treated cylindrical sam-
demonstrate that: ples prepared in triplicate and cured following the procedure
adopted for the preparation of specimens for resilient modulus
1. The dosage of cement may noticeably decrease the residual testing as per AASHTO T307-99 [32]. The duration of the curing
strain when compared to tests made without cement treat- period was kept to 3, 7 and 14 days before the USC testing. During
ment. For instance, for the untreated blend (Fig. 8) containing the UCS test, the loading rate was maintained at ~1.27 mm/min-
25%G and 75% RAP, the value of residual strain at the end of utes using the servo-controlled MTS. It should be noted that stable
specimens could not be prepared without cement dosage e.g., 100%
RAP material.
Fig. 10 shows the mean UCS values for each of the blended sam-
ples tested in triplicate along with error bars. In most of the cases,
the maximum and minimum UCS values remained within ±10%
Curing period: 3 days
(error bars) of the mean value. From this figure, it can be inferred
that the UCS values are not dependent only on the percentage of
the RAP contents in the blended samples but also on cement
dosage and the curing period before the UCS testing is carried
out. In general, UCS values were found to increase with cement
dosage, duration of the curing period, and the percentage of the
fresh granular material in the blended sample. For instance, after
three days of curing, the 100% RAP sample treated with 1.5% of
OPC had a UCS value of 125 kPa whereas the corresponding values
for blended samples prepared by mixing the 75% RAP with 25% G,

Curing period: 7 days

(days)
Fig. 9. The effect of curing time and cement dosage on the residual strain: (a) Three
days curing time and (b) Seven days curing time. Fig. 10. UCS values of the blended samples.
 M. Arshad / Construction and Building Materials 254 (2020) 119234 9

50% RAP with 50% G, 25% RAP with 75% G and 100% G were 567,
980, 1494 and 1660 kPa respectively. These values, in turn, corre-
spond to an increase of 353%, 684%, 1095% and 1228% when com-
pared to the reference value of 125 kPa. These trends of variation in
the UCS values are in agreement with finding of Taha et al. [15],
Yuan et al. [27] and Hoyos et al. [39] among others.
Similarly, after the curing of seven days and with an addition of
4.5% of cement, the increase in UCS values of the blended samples
may reach up to 596%, 766%, 1004% and 1135% for the blends con-
taining 25%, 50%, 75% and 100% of the fresh granular material,
when compared to the value obtained for the sample with 100%
RAP material and 4.5% cement dosage. A similar increasing trend
in UCS values due to the increase in fresh granular material was
observed when blended samples were cured for 14 days before
the UCS testing, irrespective of the cement dosage. The increasing
trend in UCS values of the RAP-aggregate blends with an increase
in cement content and curing period is in agreement with findings Fig. 12. Effect of curing period on UCS values of the blended samples.
reported by many researchers.
The reason for the increase in the UCS value with the fresh gran-
ular material in the blend may be due to an enhanced load transfer can be inferred that there is a quick gain in strength up to the cur-
occurring between the particles of the blend. More specifically the ing period of 7 days, and after that UCS value still increases but at a
presence of fresh aggregate in higher percentages causes greater slower rate. For example: with a blended sample containing 25%
levels of friction and interlocking actions between the aggregate RAP and 75% G material and treated with 4.5% cement, the UCS
particles with the reduction in asphalt coated viscous surfaces value after 3 and 7 days of curing is 2158 kPa and 4351 kPa respec-
within the blend. tively, i.e., a 100% increase in the UCS value due to the curing per-
iod of 4 days (from 3 days to 7 days), whereas after 14 days of
4.1. Effect of cement dosage on UCS value curing the UCS value reaches 5355 kPa. This increase of 1004 kPa
is only 46% when compared to the value obtained after 3 days of
Fig. 11 explicitly shows the change in the UCS values with curing, and it further reduces to 18% when compared to the value
cement dosage for a particular blend cured for a certain period. obtained after 7 days curing. A very similar trend in the increase in
For instance, a blended sample containing 50% G and 50% RAP the UCS values was observed for almost all the blends due to the
and cured for 14 days the UCS value increased from 2597 kPa to curing period.
3860 kPa when the cement contents were increased from 1.5% to The UCS value of 7 days cured and chemically stabilised base
4.5%, i.e. a 48% increase in the UCS value. In turn, this figure implies course materials should be in the range of 2070–2760 kPa as per
that the rate of gain of strength with cement content is 421 kPa per recommendations of Portland Cement Association (PCA) docu-
every 1% increase of cement. Similarly, for the 100% RAP and 100% mented by Thakur [13].
G material the corresponding increase in UCS value was found to
be ~56% and ~17%, respectively. Increasing the cement content of 4.3. Statistical analysis and modelling of the UCS values
a blend produces stronger cementitious bonding between the
aggregates in the blend/mixtures as reported by other researchers Table 5 shows the results of a two-tailed t-test in terms of com-
including Ma et al., [40] and Bessa et al. [41]. parison of t-statistic value with the t-critical value obtained at a 5%
Collectively, the effect of cement dosage found on UCS values of level of significance for the paired data. The data sets were com-
the blended samples is following the conclusions documented in prised of UCS values obtained for tested blends made with varying
the existing literature [20,42,43]. percentage of RAP/G material but at a particular value of OPC con-
tent and/or curing period. The null hypothesis for the t-test
4.2. Effect of curing period on UCS value assumed that there was a significant difference between the data
sets of UCS value (i.e. t-statistic > t-critical), while the alternate
Fig. 12 exclusively depicts the variation in UCS value with cur- hypothesis suggests that there was no significant difference
ing period and at a particular cement dosage. From this figure, it between the data sets of UCS value (i.e., t-statistic < t-critical).
The result of the t-test presented in Table 5 implies that UCS val-
ues are significantly dependent not only on the type of blend but
also on the percentage of OPC content in addition to the curing per-
iod (CP).
Fig. 13 shows that variations in the UCS values with the per-
centage of the granular material in the blend can be approximated
through a power function over the range of the cement dosage and
curing period. A careful analysis of the whole data sets suggests
that all of these three significantly interdependent variables can
be incorporated simultaneously into a single regression model pre-
sented as a power function for the estimation of UCS values:
h i
UCS ðkPaÞ ¼ s# 161ðXÞ0:47 ð1Þ

where s is the model parameter for considering the amount of OPC

in the blend and its value is 1, 1.4 and 1.6 for an OPC content of 1.5%,
Fig. 11. Effect of cement dosage on UCS values of the blended samples. 3.0% and 4.5% respectively; # is the model parameter for considering
10 M. Arshad / Construction and Building Materials 254 (2020) 119234

Table 5
Results of the t-tests based on UCS values.

Curing period Comparison based on OPC (%) t-critical t-statistic Remarks (difference)
3 days 1.5 vs 3.0 2.77 3.3 significant
1.5 vs 4.5 2.77 4.17 significant
3.0 vs 4.5 2.77 3.35 significant
7 days 1.5 vs 3.0 2.77 3.72 significant
1.5 vs 4.5 2.77 4.28 significant
3.0 vs 4.5 2.77 3.12 significant
14 days 1.5 vs 3.0 2.77 2.38 insignificant
1.5 vs 4.5 2.77 3.81 significant
3.0 vs 4.5 2.77 2.26 insignificant

OPC (%) Comparison based on curing period (days) t-critical t-statistic

1.5% 3 vs 7 2.77 3.38 significant
3 vs 14 2.77 3.00 significant
7 vs 14 2.77 2.46 insignificant
3.0% 3 vs 7 2.77 3.83 significant
3 vs 14 2.77 3.73 significant
7 vs 14 2.77 3.22 significant
4.5% 3 vs 7 2.77 3.83 significant
3 vs 14 2.77 3.65 significant
7 vs 14 2.77 3.13 significant

the curing period and its value is 1, 2.1 and 2.6 for the curing period
of 3, 7 and 14 days; X is the percentage of granular material in the
blend with the minimum value up to 0.5%.
Fig. 14 shows the distribution of the data point (determined and
estimated UCS values) about the unity line, along with the 95%
confidence interval (CI) and prediction interval (PI). This distribu-
tion of the data points about the unity line is found to be fairly dis-
tributed. The mathematical expressions for CI and PI are given in
Eqs. (2) and (3) below, respectively. CI defines the limits such that
there is a 95% probability that the true best-fit line for the popula-
tion falls within this interval. PI defines the limits such that 95% of
the estimated values to be found for a certain measured value are
within the interval range around the linear regression line:
vffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
!
u
u 1 ðX  X m Þ2
X ¼ y  ta SEEt þ ð2Þ
N Sxx

vffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
!ffi Fig. 13. Variation in the UCS values with granular material.
u
u 1 ðX  X Þ 2
X ¼ y  ta SEEt 1 þ þ
m
ð3Þ
N Sxx

where y is the trend line value at a particular pair of determined

and estimated values of UCS; t a is the t-statistic at a particular level
of probability (5% in this case) corresponding to N–2 degrees of
freedom; X m is the mean of measured values; X is a particular mea-
sured value; SEE is the standard error of estimate which is the stan-
dard deviation of predicted dependent variable values about the
estimated regression line; N is the number of paired observations
and Sxx
P
is the sum of the squared deviations of the predicted values
(Sxx = (xi- xm)2). Further detail on equation (2) and (3) can be found
in Ott and Longnecker [44].

5. Indirect tensile strength (ITS)

Indirect tensile strength is a primary parameter for the evalua-

tion of RAP materials that are to be used in the base layer [45,46].
ITS tests were performed as per ASTM D 6931 on the cylindrical
samples with a diameter of 150 mm and height 112 mm. The spec-
imens were cured following the procedure adopted for the prepa-
ration of specimens for resilient modulus test as per AASHTO T307- Fig. 14. Distribution of experimentally determined and estimated UCS values along
99. The duration of the curing period was kept to 3, 7 and 14 days the unity line.
 M. Arshad / Construction and Building Materials 254 (2020) 119234 11

before the ITS testing. The ITS value was then calculated using the 2. ITS values increased with the curing period; for the blend pre-
following equation: pared by mixing 50% RAP and 50% G and treated with 3%
cement content, the ITS value was found to be 160 kPa after
2P
ITS ¼ ð4Þ 3 days of curing and reached 345 and 677 kPa after 7 and
phD 14 days curing respectively. These figures, in turn, imply that
where ITS is in kPa; P (kN) is the maximum applied load indicated the rate of gain of ITS value is 47 kPa/day for the curing period
by the loading gauge of the testing machine; h(m) is the height of from 7 to 14 days. The corresponding figures for the 100% G
the tested sample and D(m) is the diameter of the tested sample. material and 100% RAP materials are 76 kPa/day and
Fig. 15 shows the ITS test results for the specimens prepared 2.56 kPa/day;
over a range of RAP/G contents and cement dosage, and cured for 3. ITS values were found to be significantly dependent on the per-
the period of 3, 7 and 14 days before testing. ITS tests were per- centage of the RAP content in the blend, irrespective of the cur-
formed in triplicate to ascertain the reliability of the ITS values. ing period and the cement dosage. For instance, the sample that
In most of the cases, the maximum and minimum ITS value for a contained 100% G material treated with 1.5% of cement and
particular sample remained within ±15% (error bars) of the mean tested after 7 days of curing period the ITS value was 426 kPa,
value. However, in general, the error bars reached up to 25% for which reduced to 238 kPa for the blend containing 50% RAP
the samples prepared at a cement dosage of 1.5% and/or cured and 50% G with the same cement dosage and curing period. This
for 3 days. The specific trend is variation in the ITS values, which value further reduces to 65 kPa for a sample containing 100%
was very similar to that observed during the UCS testing as RAP material.
depicted in Fig. 16:
In general, the ITS values remained within a range of 12–16% of
1. ITS values increased with an increase in the cement dosage, the UCS values for the different specimens included in the testing
which can be approximated through a linear function taking campaign.
cement dosage as an independent variable for all types of the
blends investigated in this research. The power function yielded 6. Drying shrinkage testing
a coefficient of determination (R2) in the range of ~0.87–0.98
(refer to Fig. 16), which is in fact square of the Pearson correla- In the existing literature, different methodologies have been
tion coefficient (R); adopted to investigate the drying shrinkage potential of cement-
treated geo-materials. The difference may be in terms of the spec-
imens preparation technique, size of the specimens (dimensions)
and/or shape (cylindrical or prismatic) of the specimens
investigated.
For the study presented in this research paper, cylindrical spec-
imens were prepared and cured following the same techniques as
adopted for the specimen used for the resilient modulus, or the
UCS tests, as explained earlier. However, to investigate the drying
shrinkage potential, samples were only cured for 24 h and then
dried in a controlled environment chamber with a relative humid-
ity of 55 ± 5% and an ambient temperature of 23° ± 2 °C. For each of
the blends, two specimens were prepared and the average values
of the drying shrinkage along with error bars were reported. The
cured samples were placed on the rigid platen and were wrapped
with a thin polythene sheet along the sides to prevent moisture
being lost from these curved surfaces. This measure was taken to
allow the maximum moisture loss from the top flat surface of
the specimen, which was then covered with Perspex squares. A
Fig. 15. ITS values of the blended samples. well-calibrated dial gauge having least-count of 0.001 mm was
mounted to the flat surface of the Perspex squares, as shown in
Fig. 17. This arrangement allowed the determination of the change
in the axial length of the specimen, which was then converted into
the percentage strain based on the initial length of the specimen
measured just before the start of the drying process. Fig. 18 shows
the shrinkage values (in percentage) observed after the respective
drying period of 3, 7 and 14 days.
From this figure, it can be inferred that the drying shrinkage
increased with the drying period and cement contents irrespective
of the type of blend investigated. The presence of RAP content, and
hence asphaltic coating, could not show a clear or significant effect
on the shrinkage potential evaluated. For instance, the blend pre-
pared with 100% G material and treated with 3% cement dosage
showed a drying shrinkage potential of ~0.025% at the end of 3 days
of drying, and this value reached to ~0.035% and ~0.05% at the end
of 7 and 14 days of drying. These figures imply that shrinkage
increased with curing time and higher rates were observed in the
early stages, as witnessed for the UCS testing. A very similar trend
was observed for the other blends prepared with RAP materials in
Fig. 16. Effect of cement dosage on the ITS values of the blended samples. the range of 25% to 75%, suggesting that the shrinkage behaviour of
12 M. Arshad / Construction and Building Materials 254 (2020) 119234

Fig. 19. Creep potential of the blended samples.

levels of cement dosage (0, 1.5 and 3.0%). The blended samples
Fig. 17. Drying shrinkage test in progress on cylindrical specimens wrapped with
were prepared and cured following the same procedure as adopted
polythene sheets.
for the resilient modulus or UCS testing (explained earlier). How-
ever, to investigate the creep potential, samples were cured for
24 h and then placed under the loading platen/chamber of the
MTS machine used for the resilient modulus testing in a controlled
environment of the relative humidity of 55 ± 5% and the ambient
temperature of 23° ± 2 °C. Also, one blend (75% G + 25% RAP with
1.5% OPC) out of six blends was cured for 3 days before the creep
testing. During the creep testing, each of the samples was sub-
jected to a cell pressure and deviator stress level of 68.9 kPa for
6000 min. The sustained loading and the response of the samples
were constantly monitored; this data was stored through the data
acquisition system as used during the resilient modulus testing.
Fig. 19 shows the effect of cement dosage, RAP content and the
curing period on the creep potential of the blended samples.
From this figure the following conclusions can be drawn:

1. RAP content in the blended sample substantially affected the

creep behaviour of the specimens under the same level of OPC
(days)
dosage. For instance, at the end of the creep testing period (after
Fig. 18. Drying shrinkage values of the blended samples. 000 min), the creep strain of the sample with 25% G + 75% RAP
and treated with 1.5% OPC had a cumulative creep strain of
the studied mixtures (with RAP addition) could be very similar to 0.07%. However, this value reduced to 0.03% when the blend
those observed for conventional cement-treated materials (with- contained 75% granular material along with 25% RAP. Similar
out the RAP addition). On average, the shrinkage potential observations have been reported by Cosentino et al. [47];
increased from 0.01% to 0.035% with an increase of cement dosage 2. Cement dosage caused the most significant impact on creep
from 1.5% to 4.5% at the end of 3 days of drying, i.e. an almost 250% strain. For example, the blended sample with 25% G + 75%
increase in drying shrinkage potential. RAP and treated with 0% OPC showed 0.127% creep strain after
This value decreases to 150% and then reduces to 100% at the end 6000 min of sustained loading. This value reduced to 0.07%
of 7 and 14 days of drying periods under the same dosage of cement (~45% reduction) and 0.055% (~56% reduction) when the same
content. The increase in drying shrinkage in conjunction with blend was treated with 1.5% and 3.0% OPC. Nevertheless, the
cement content is primarily due to the consumption of available impact of cement dosage was substantially milled for the blend
water in the hydration reaction of the cement. From the author’s per- containing 75% G + 25% RAP;
spective, an increase in cement contents for stabilisation of the geo- 3. Curing period before sustained loading also affected the creep
materials may lead to high moisture consumption and consequently behaviour noticeably. For instance, for the blended sample con-
high shrinkage, which may be followed by cracking of various inten- taining 75% G + 25% RAP cured for 24 h and treated with 1.5%
sities depending on the type of material stabilised. The use of low OPC, the creep strain value was 0.03% at the end of 6000 min
cement content may potentially contribute to the reduction of of sustained loading. This value reduced to 0.025% when the
potential for cracking although strength and stiffness requirements same blend was cured for 48 h before being subjected to the
of the blend may be compromised. same level of sustained loading.

7. Creep potential of the blended sample under sustained In the existing literature, different numerical ranges of creep
loading strain or potential have been reported; however, the trends of
‘creep strain vs time’ lines have an almost similar trend. This is
Creep potential of two different types of blends (75% G + 25% primarily because different researchers have used different types/
RAP and 25% G + 75% RAP) was investigated at three different combinations of the blended samples and stabilising/cement
 M. Arshad / Construction and Building Materials 254 (2020) 119234 13

material in addition to loading conditions and ambient deviator stress of 124 kPa after 1 day of curing, the rate of increase
environments. of strain becomes 7.1E-5% per cycle for the load cycles in the range
From the author’s perspective, asphalt material has viscoelastic of 2000–20,000. This value reduces to 2.55 E-6% per cycle for the
properties and the presence of asphalt material leads to reorienting samples cured for 7 days. On the other hand, this figure remained
of the granualr particles under long-term static loading. Reposi- limited to 1.7 E-5% and 6.5 E-6% per load cycle applied at a cyclic
tioning of particles in the long-term causes the higher creep values deviator stress level of 31 kPa on the samples cured for 1 and 7 days
of the blended samples containing higher percentages of RAP con- respectively.
tent when compared to those prepared with a lesser amount of
RAP material in the blend. Similar observations have been reported
9. Summary and conclusions
by other researchers [15,27].
The author offers the following points as a conclusion to this
8. Long-term performance of blended samples subjected to research paper:
cyclic loading
1. Mechanical properties of the cement-treated blended sam-
To explore the curing effects on long term performance under ples were found to be significantly dependent on the con-
cyclic loading, six repeated load tri-axial tests were performed on tent/combination of the blend, dosage of the cement
blended samples containing 50% G and 50% RAP and treated with treatment and the curing period;
3% cement. Three tests were conducted under a cyclic deviator 2. The addition of RAP materials with natural granular samples
stress value of rd = 31 kPa and a confining stress value of yields higher resilient moduli and lower UCS and ITS values
r3 = 34.5 kPa, while for the other three tests the corresponding val- than those obtained for natural granular samples when the
ues of stress conditions were 124 kPa and 137.9 kPa as shown in loading conditions were the same.
Fig. 20. In both of the loading scenarios, 20,000 load cycles were 3. The addition of ordinary Portland cement, even in small
applied using a haversine stress wave, as in the case of the MR test- quantities (1.5–4.5%), to the fresh granular and blended
ing. From this figure, it can be inferred that the accumulative strain samples, may increase their resilient modulus, UCS and ITS
value is not only dependent on the loading condition, but it is also values by a considerable extent. Similarly, curing of the
highly dependent on the curing period. For instance, the accumu- cement-treated specimens yields a noticeable increase in
lative strain after 10,000 load cycles under the cyclic deviator these strength parameters.
stress of 124 kPa is 2.43% for the sample which has a curing period 4. For all the tested samples, the resilient modulus was typi-
of 1 day, while this value reduces to 1.37% and 1.06% for the sam- cally found to be directly proportional to the bulk stresses
ples tested at the end of 3 and 7 days of curing. applied to them. The mutual relation between the resilient
Similarly, for the samples subjected to cyclic deviator stress of modulus and the bulk stress of a specific material can be rea-
31 kPa, the accumulative strain value after 10,000 load cycles sonably determined using a power-law function having a
remained limited to 0.60%, 0.37% and 0.28% for 1, 3 and 7 days of coefficient of correlation in the range of 0.96 to 0.99;
curing. These values suggest that the accumulative strain value 5. Similarly, the variation in UCS values of the blended samples
of the samples may reduce in the range of approximately 50–60% with the RAP contents and with the curing period can be
due to the curing period of 7 day, when compared to those sub- approximated using a power-law function having a coefficient
jected to 1 day of curing. A similar trend of reduction in the accu- of correlation in the range of 0.96 to 0.99. Likewise, a correla-
mulative strain values due to curing period was observed for tion between the UCS and ITS values with cement content/
loading cycles in the range of 2000–20,000. Furthermore, it is obvi- dosage can be established using linear regression model;
ous from the figures that practically 60% of the total residual strain 6. The creep strain of the blended samples (with RAP contents)
occurred through the first 2000 load cycles out of the 20,000 total under sustained loading may be many folds of the values
load cycles applied for all the cases investigated. obtained for fresh granular samples. This situation was
Nevertheless, it is interesting to note that there is a tendency for found more pronounced in the case of samples prepared at
the elastic shakedown to be less noticeable for samples subjected 0% cement dosage;
to a higher level of cyclic deviator stress and shorter curing period 7. The yield of residual strain is not only dependent on the type
when compared to those subjected to lower levels of cyclic devia- and composition of the blend, the dosage of cement treatment,
tor stress and a longer curing period. For instance, for the cyclic and the duration of the curing period before cyclic loading, but
it is also dependent on the ratio of cyclic deviator stress to the
confining stress(rrdc Þ during the cyclic triaxial testing. More
specifically, a higher value of rrdc produces more residual strain;
8. The performance of blended samples in the domain of residual
strain when subjected to thousands of load cycles (20,000 for
this study) is highly pressure-sensitive. Their tendency towards
elastic shakedown increases in line with increases in the curing
periods, and decreases in line with increases in the amplitude
to the application of the cyclic deviator stress value.
9. The addition of ordinary Portland cement (1.5–4.5%), to the
blended samples, may prove to be a reliable stabilising agent
against the residual and creep strain. However, drying
shrinkage, which may lead to cracking and other instabilities
due to the moisture consumption during the hydration of
the cement paste can be a major concern for cement-
stabilised base layers of the flexible pavement. These factors
must be addressed to avoid undesirable performances of
Fig. 20. Long term performance of blended samples subjected to repeated loading. these base layers;
14 M. Arshad / Construction and Building Materials 254 (2020) 119234

10. The presence of RAP content, and hence asphaltic coating, [18] A. Grilli, E. Bocci, A. Graziani, Influence of reclaimed asphalt content on the
mechanical behaviour of cement-treated mixtures, Road Mater. Pavement Des.
could not show a clear or significant effect on the shrinkage
14 (3) (2013) 666–678.
potential evaluated for the blends considered in this study. [19] A.J. Puppala, L.R. Hoyos, A.K. Potturi, Resilient moduli response of moderately
cement-treated reclaimed asphalt pavement aggregates, J. Mater. Civ. Eng. 23
(7) (2011) 990–998.
[20] A.R. Ghanizadeh, M. Rahrovan, K.B. Bafghi, The effect of cement and reclaimed
Declaration of Competing Interest asphalt pavement on the mechanical properties of stabilized base via full-
depth reclamation, Constr. Build. Mater. 161 (2018) 165–174.
[21] A. Mohammadinia, A. Arulrajah, S. Horpibulsuk, A. Chinkulkijniwat, Effect of
The authors declare that they have no known competing finan- fly ash on properties of crushed brick and reclaimed asphalt in pavement
cial interests or personal relationships that could have appeared base/subbase applications. J. Hazard. Mater. 321(2017):547–556.
to influence the work reported in this paper. [22] M. Hoy, S. Horpibulsuk, A. Arulrajah, Strength development of recycled asphalt
pavement–fly ash geopolymer as a road construction material, Constr. Build.
Mater. 117 (2016) 209–219.
Acknowledgements [23] M. Hoy, S. Horpibulsuk, A. Arulrajah, A. Mohajerani, Strength and
microstructural study of recycled asphalt pavement: Slag geopolymer as a
pavement base material. J. Mater. Civ. Eng. 30(8) (2018):04018177.
The author gratefully acknowledges the research training pro- [24] W.S. Guthrie, A.V. Brown, D.L. Eggett, Cement stabilization of aggregate base
vided by the McMaster University, Hamilton, Canada. The author material blended with reclaimed asphalt pavement, Transp. Res. Rec. 2026 (1)
(2007) 47–53.
also acknowledges the valuable comments and suggestions of the [25] A.J. Puppala, A. Pedarla, B. Chittoori, et al., Long-term durability studies on
anonymous reviewers for the improvement of this article. The chemically treated reclaimed asphalt pavement material as a base layer for
author also acknowledges the unprecedented cooperation and sup- pavements, Transp. Res. Rec. 2657 (1) (2017) 1–9.
[26] M. Isola, G. Betti, A. Marradi, G. Tebaldi, Evaluation of cement treated mixtures
port of the laboratory staff in the execution of the experimental
with high percentage of reclaimed asphalt pavement, Constr. Build. Mater. 48
work presented in this research. (2013) 238–247.
[27] D. Yuan, S. Nazarian, L.R. Hoyos, A.J. Puppala, Evaluation and mix design of
cement-treated base materials with high content of reclaimed asphalt
References pavement, Transp. Res. Rec. 2212 (1) (2011) 110–119.
[28] G. Tebaldi, E.V. Dave, P. Marsac, et al., Synthesis of standards and procedures
[1] G. Ransinchung, S. Singh, S. Abraham, Feasibility of reclaimed asphalt for specimen preparation and in-field evaluation of cold-recycled asphalt
pavement in rigid pavement construction. In: Engineering Challenges for mixtures, Road Mater. Pavement Des. 15 (2) (2014) 272–299.
Sustainable Future: Proceedings of the 3rd International Conference on Civil, [29] G. Tebaldi, E. Dave, P. Marsac, et al., Classification of recycled asphalt pavement
Offshore and Environmental Engineering (ICCOEE 2016, Malaysia, 15-17 Aug (RAP) material. In: Proceedings of the 2nd international Symposium on
2016). (2016) of Conference: CRC Press. Asphalt Pavements and Environment. (2012). Fortaleza, Brazil.
[2] S. Singh, K. Monu, R.N.G. Ransinchung, Laboratory investigation of rap for [30] W.H. Chesner, R.J. Collins, M.H. MacKay, J. Emery, User Guidelines for Waste
various layers of flexible and concrete pavement, Int. J. Pavement Eng. (2019) and by-Product Materials in Pavement Construction, Recycled Materials
1–14. Resource Center, 2002.
[3] S. Yu, S. Shen, C. Zhang, et al. Evaluation of the blending effectiveness of [31] 180 AASHTO T, Modified method of test for the moisture-density relations of
reclaimed asphalt pavement binder. J. Mater. Civil Eng.. 29(12) soils, American Assoc. State High. Trans.Offic. (2010).
(2017):04017230. [32] AASHTO T307-99 Standard method of test for determining the resilient
[4] N. Tran, Z. Xie, G. Julian, et al. Effect of a recycling agent on the performance of modulus of soils and aggregate materials. American Association of State
high-rap and high-ras mixtures: Field and lab experiments. J. Mater. Civil Eng. Highway and Transportation Officials, Washington. 99(2012).
29(1) (2016):04016178. [33] Y.H. Huang, Pavement Analysis and Design, Prentice Hall, Englewood Cliffs
[5] X. Shi, A. Mukhopadhyay, D. Zollinger, Sustainability assessment for portland New Jersey, 1993.
cement concrete pavement containing reclaimed asphalt pavement [34] T. Bennert, A. Maher, The Development of a Performance Specification for
aggregates, J. Cleaner Prod. 192 (2018) 569–581. Granular Base and Subbase Material, Dept. of Transportation, New Jersey,
[6] X. Shi, D.G. Zollinger, A.K. Mukhopadhyay, Punchout study for continuously 2005.
reinforced concrete pavement containing reclaimed asphalt pavement using [35] MacGregor J., Highter W. DeGroot D. Structural numbers for reclaimed asphalt
pavement me models, Int. J. Pavement Eng. (2018) 1–14. pavement base and subbase course mixes. Transp. Res. Rec.: J. Transp. Res.
[7] M. Kumari, G. Ransinchung, S. Singh, A laboratory investigation on dense Board (1687) (1999):22–28.
bituminous macadam containing different fractions of coarse and fine rap, [36] T.B. Alam, M. Abdelrahman, S.A. Schram, Laboratory characterisation of
Constr. Build. Mater. 191 (2018) 655–666. recycled asphalt pavement as a base layer, Int. J. Pavement Eng. 11 (2)
[8] G. Thenoux, A. Gonzalez, R. Dowling, Energy consumption comparison for (2010) 123–131.
different asphalt pavements rehabilitation techniques used in Chile, Resour. [37] W. Kim, J. Labuz, Resilient modulus and strength of base course with recycled
Conserv. Recycl. 49 (4) (2007) 325–339. bituminous material. (2007).
[9] S. Liu, A. Shukla, T. Nandra, Technological, environmental and economic [38] R.L. Ott, M.T. Longnecker, An introduction to statistical methods and data
aspects of asphalt recycling for road construction, Renewable Sustainable analysis. (2015): Nelson Education.
Energy Rev. 75 (2017) 879–893. [39] L.R. Hoyos, A.J. Puppala, C.A. Ordonez, Characterization of cement-fiber-treated
[10] M. Arshad, M.F. Ahmed, Potential use of reclaimed asphalt pavement and reclaimed asphalt pavement aggregates: Preliminary investigation, J. Mater.
recycled concrete aggregate in base/subbase layers of flexible pavements, Civil Eng. 23 (7) (2011) 977–989.
Constr. Build. Mater. 151 (2017) 83–97. [40] T. Ma, H. Wang, Y. Zhao, X. Huang, Laboratory investigation on residual
[11] M. Arshad, Correlation between resilient modulus (mr) and constrained strength of reclaimed asphalt mixture for cold mix recycling, Int. J. Pavement
modulus (mc) values of granular materials, Constr. Build. Mater. 159 (2018) Res. Technol. 8 (1) (2015).
440–450. [41] I.S. Bessa, A.L. Aranha, K.L. Vasconcelos, et al., Laboratory and field evaluation
[12] M. Arshad, Experimental evaluation of constitutive models for the estimation of recycled unbound layers with cement for use in asphalt pavement
of resilient modulus values and an assessment of pressure sensitivity of rehabilitation, Mater. Struct. 49 (7) (2016) 2669–2680.
granular blends containing reclaimed asphalt pavement materials, Constr. [42] W.S. Guthrie, D. Cooley, D.L. Eggett, Effects of reclaimed asphalt pavement on
Build. Mater. 180 (2018) 491–502. mechanical properties of base materials, Transp. Res. Rec. 2005 (1) (2007) 44–
[13] J.K. Thakur, J. Han, R.L. Parsons Factors influencing deformations of geocell- 52.
reinforced recycled asphalt pavement bases under cyclic loading. J. Mater. Civil [43] A.V. Brown, Cement stabilization of aggregate base materials blended with
Eng. 29(3) (2016):04016240. reclaimed asphalt pavement. (2006).
[14] Q. Dong, B. Huang, Laboratory evaluation on resilient modulus and rate [44] R. Ott, M. Longnecker An introduction to statistical methods and data analysis:
dependencies of rap used as unbound base material, J. Mater. Civ. Eng. 26 (2) Brooks. Cole, Cengage Learning. 1273(2010).
(2014) 379–383. [45] A. Grilli, M. Bocci, A.M. Tarantino, Experimental investigation on fibre-
[15] R. Taha, A. Al-Harthy, K. Al-Shamsi, M. Al-Zubeidi, Cement stabilization of reinforced cement-treated materials using reclaimed asphalt, Constr. Build.
reclaimed asphalt pavement aggregate for road bases and subbases, J. Mater. Mater. 38 (2013) 491–496.
Civ. Eng. 14 (3) (2002) 239–245. [46] T. Ma, H. Wang, X. Huang, et al., Laboratory performance characteristics of high
[16] J. LaHucik, S. Schmidt, E. Tutumluer, J. Roesler, Cement-treated bases modulus asphalt mixture with high-content rap, Constr. Build. Mater. 101
containing reclaimed asphalt pavement, quarry by-products, and fibers, (2015) 975–982.
Transp. Res. Rec. 2580 (1) (2016) 10–17. [47] Cosentino P., Kalajian E., Shieh C.-S., et al., Developing specifications for using
[17] S.E. Euch Khay, S.E. Euch Ben Said, A. Loulizi, J. Neji, Laboratory investigation of recycled asphalt pavement as base, subbase or general fill materials, phase ii.
cement-treated reclaimed asphalt pavement material. J. Mater. Civil Eng. 27(6) (2003).
(2014):04014192.