You are on page 1of 14

Transportation Geotechnics 30 (2021) 100599

Contents lists available at ScienceDirect

Transportation Geotechnics
journal homepage: www.elsevier.com/locate/trgeo

Stress-strain response analysis of demolition wastes as aggregate base


course of pavements
Ehsan Yaghoubi a, *, Nithin Sudarsanan b, Arul Arulrajah c
a
College of Engineering and Science, Victoria University, Melbourne, Australia
b
Department of Civil, Construction, and Environmental Engineering, North Carolina State University, Raleigh, USA
c
Department of Civil and Construction Engineering, Swinburne University of Technology, Melbourne, Australia

A R T I C L E I N F O A B S T R A C T

Keywords: The sustainable construction approach by utilizing recycled aggregates has increasingly been the focus of
Demolition wastes highway construction industries and local road authorities in recent years. The efficient usage of recycled ag­
Resilient modulus behavior gregates for sustainable construction is owed to more than a decade of extensive experimental research aiming to
Pavement base
remove uncertainties in the properties and performance of recycled aggregates in transport infrastructures.
Recycled aggregates
Pavement response analysis
Nevertheless, a lack of knowledge exists about the stress-strain response of the pavement structures constructed
using recycled aggregates. The main goal of this study was to propose a response analysis approach to incor­
porate the repeated load triaxial (RLT) test results as a Level 1 MEPDG input parameter in a well-established
pavement analysis software named FlexPAVETM. In this approach, the aggregate base course was divided into
ten sublayers of equal thickness. Three different constitutive resilient modulus (Mr) models were utilized to
determine the Mr of each sublayer corresponding to the stress levels achieved at that sublayer. Modeling the
pavements in FlexPAVETM by assigning the corresponding Mr to the sublayers resulted in a more accurate stress-
strain response compared to the conventional linear elastic analysis approach. This approach provides a more
realistic and, accordingly, more accurate analysis of the behavior of unbound aggregates in pavements. The
secondary goal of this study was to investigate the stress-strain response of pavements with aggregate base
courses made of three types of recycled construction and demolition wastes, being recycled concrete aggregate
(RCA), crushed brick (CB), and waste excavation rock (WR). The resilient properties of the demolition wastes
were determined through repeated load triaxial (RLT) testing. A typical pavement profile consisting of asphalt
surface course, aggregate base course, and an A6 type of natural soil as subgrade was modeled using FlexPA­
VETM. The response analysis was undertaken following the proposed sub layered approach. Under the loading
conditions adopted in this research, with bulk stresses <185 kPa, CB and RCA exhibited a superior Mr response
compared to the conventional virgin materials, while WR showed an inferior behavior.

Introduction particular Recycled concrete aggregate (RCA), Crushed brick (CB) and
Waste excavation rock (WR) are considered as an attractive alternative
Construction activities are known as a significant source of CO2 to virgin aggregates in the highway construction industry, due to their
emission and associated climate impacts. Considerable growth of the resilient characteristics, durability and non-expansive nature [3,4]. RCA
construction activities has caused a substantial production of demolition is produced by crushing concrete structural elements extracted through
wastes in recent years. High levy and environmental issues associated demolition activities [5]. CB is a product of the demolition of structures
with landfilling of demolition wastes have made this traditional made of bricks, such as masonry buildings [6]. WR is a waste produced
approach of waste management undesirable [1]. Recycling and reusing as a result of digging or blasting in the ground for construction purposes,
of solid wastes as an alternative to virgin materials have drawn the such as digging trenches for water/sewer infrastructures [2].
attention of construction industries in order to lower such negative While the focus of the majority of research projects has been placed
impacts on the environment [2]. Recycled demolition aggregates, in on the laboratory and field experiments, analysis of the resilient

* Corresponding author at: College of Engineering and Science, Victoria University, Footscray, VIC 3010, Australia, PO Box 14428, Melbourne, Vic 8001, Australia.
E-mail addresses: ehsan.yaghoubi@vu.edu.au (E. Yaghoubi), nsudars@ncsu.edu (N. Sudarsanan), aarulrajah@swin.edu.au (A. Arulrajah).

https://doi.org/10.1016/j.trgeo.2021.100599
Received 11 August 2020; Received in revised form 2 June 2021; Accepted 3 June 2021
Available online 9 June 2021
2214-3912/© 2021 Elsevier Ltd. All rights reserved.
E. Yaghoubi et al. Transportation Geotechnics 30 (2021) 100599

behavior of demolition wastes as an unbound structural layer of pave­ rigorous experimental research works on properties and the applica­
ments is still lacking. Modeling pavements made of recycled aggregates bility of recycled aggregates in geotechnical and pavement applications.
in response and performance analysis computer programs such as In addition, recycling industries have made significant improvements
FlexPAVETM by incorporating Level 1 MEPDG input and analyzing their and are now capable of producing high quality recycled aggregates with
behavior in comparison with that of conventional virgin aggregates is a consistent physical and strength properties [12]. However, a more
step forward to a sustainable design and construction approach. robust simulation of the unbound layers of pavements through well-
FlexPAVE™ [7] is a three-dimensional viscoelastic finite element established performance analysis software and utilizing Level 1
analysis program that can simulate the pavement response under real­ MEPDG data can further increase the industry’s confidence in using
istic moving vehicular loads with various axle configurations at various recycled aggregates in design and construction projects.
pavement temperatures. Several researchers have reported that the Poor quality and performance of unbound structural layers of
simulation predictions of FlexPAVE™ reasonably match with field ob­ pavements results in pavement distresses such as fatigue cracking and
servations. Although, it is worth mentioning that this program still re­ rutting [13]. Under the low-stress levels imposed on the unbound layers
quires improvements by adopting a more accurate laboratory-to-field of pavements through repeated traffic loads, permanent strains have
transfer function which is needed for obtaining more precise field per­ lower magnitudes compared to elastics strains. Therefore, an elastic
formance predictions [8–10]. The software is primarily equipped with response is considered suitable for unbound granular materials, and the
mechanistic models such as simplified viscoelastic continuum damage behavior of unbound pavement layers is determined by the elastic
(S-VECD) and linear elastic model. S-VECD model uses the fundamental modulus corresponding to the resilient behavior of the materials at that
properties of the asphalt concrete (AC) to predict the response behavior specific stress state [14]. Several researchers have compared the resil­
of AC surface layers. For the response analysis of base, sub-base, and ient properties of recycled demolition materials with those of virgin
subgrade layers, FlexPAVETM uses linear elastic models which can be aggregates [4,15–18]. While the majority of researchers have shown
improved using the approach proposed in the current study. The superior Mr properties for RCA compared to virgin aggregates, Lancieri
Mechanistic-Empirical Pavement Design Guide (MEPDG) [11] strongly et al. [16] indicated that higher Mr value was achieved in RCA compared
encourages Repeated Load Triaxial (RLT) testing of unbound materials to virgin aggregates under higher stress levels; however, under lower
for design purposes. The results of RLT tests can be used for the deter­ stress levels, RCA showed lower Mr values. Lancieri et al. [16] also re­
mination of constitutive Mr model parameters (k1, k2, and k3). These ported that over time, the Mr of RCA increased due to a self-cementing
model parameters are typically used as a direct input into the Pavement process. Arulrajah et al. [18] carried out an extensive experimental
Mechanistic Empirical (ME) software as the Level 1 MEPDG input. program and revealed that RCA and CB generally have higher Mr
Nevertheless, the current version of FlexPAVETM is not equipped with compared to typical quarried materials, while WR shows lower Mr under
such a feature. Currently, a multilayer elastic theory (MLET) analysis, the same stress levels. Saeed [19] investigated several mixtures of virgin
similar to Level 2 input in Pavement ME, is used in FlexPAVETM. In the aggregates with RCA and concluded that increasing the percentage of
MLET analysis a single value of the modulus of elasticity is provided as the RCA in the mix results in the reduction of Mr. Aside from favorable
an input parameter, which is a drawback that was intended to be mechanical behavior, the application of recycled aggregates can result
addressed in the current study. in less carbon footprint of construction activities [20]. Researchers such
The current study had two goals. The primary aim was to develop an as Cristelo et al. [15] and Arulrajah et al. [18] investigated the potential
approach for incorporating the RLT test results and to apply the environmental hazards of demolition and indicated that demolition
constitutive resilient modulus (Mr) models to the unbound aggregate wastes, in particular, RCA, CB, and WR are environmentally sound and
base course (ABC) in the pavement structural model in FlexPAVETM to can safely be used in pavements and highway embankments.
address a drawback of the software as explained in the previous para­ Investigating the stress distribution and associated compressive and
graph. The secondary aim was to utilize the proposed method for tensile strains due to traffic loading in a pavement structure made of
investigating the stress–strain response of three different types of recy­ recycled aggregates has not been focused on as much as experimental
cled demolition aggregates as an alternative to conventional virgin investigations. With a proven track record of mechanical properties, as
quarried materials used as ABC of flexible pavements. In this method, well as being environmentally friendly, pavement response analysis of
experimentally obtained results of the materials were used for devel­ the recycled aggregates layers through established computer programs
oping the pavement model in FlexPAVETM. The proposed approach in can provide an insight into the tensile strains and compressive strain
this research has not been previously used to improve the pavement generated in a pavement section, under moving load.
response analysis by the well-established FlexPAVETM program. The Numerous multilayer elastic-model based commercial software such
novelty of the research is the comparison of the pavement responses of as CHEVRON, BISAR, ELSYM5, KENLAYER, CIRCLY, and WELSEA were
the three CDW aggregates with conventional natural aggregate as the widely used by the pavement engineers to learn the pavement response.
ABC of a pavement system. The linear elastic analysis method assumes all the layers to behave
The outcomes of this research pave the way to a more pragmatic elastically, i.e., the layer that is stretched or deformed under a load
analysis and design of flexible pavements using RLT test results. Further, returns to its original state immediately once the loading is taken off.
the findings of the research expand the knowledge on the performance However, it is obvious that this is not the case in pavements, as other­
of recycled aggregates as the unbound base course of pavements and wise the rutting would not occur. Although the linear elastic analysis is
hence, promote the application of sustainable construction materials. considered as a reasonable and simple approach in specific cases, it
might not be sound to apply such an approach to various conditions such
Literature review as different environments, different loading conditions, and different
material types [21,22]. For instance, asphalt concrete requires some
Traditionally, in the road construction industry natural virgin ag­ time to regain its initial state after unloading, especially in a warm
gregates have been preferred by geotechnical and pavement engineers climate, due to its viscoelastic behavior. Also, while the unbound ag­
for several reasons. Primarily, the demolition material characteristics gregates of the base course have a stress-hardening behavior [23], fine-
are highly uncertain and dependent on the source; hence, using them grained soils of the subgrade have a stress-softening behavior [24].
demands more laboratory testing and quality control in comparison to The linear elastic approach is incapable of mimicking the above­
the virgin natural aggregates. Designers and contractors typically follow mentioned nonlinear behavior and therefore requires much more
a conservative approach and prefer using virgin materials with well- complicated materials models and testing to characterize the model
known performance and behavior due to decades of experience. In properties, for realistic - mechanistic numerical simulations [25].
recent years, these drawbacks have been increasingly tackled through Recently there has been increasing interest among pavement researchers

2
E. Yaghoubi et al. Transportation Geotechnics 30 (2021) 100599

in investigating the non-linear behavior of unbound and subgrade layer materials were obtained through an experimental program. The prop­
on pavement performance using numerical simulations [26–29]. The erties of the Belgrade aggregate base course (BABC) were extracted from
importance of a non-linear model for unbound and subgrade was Chow et al. [31] report. Particle size distributions (PSD) of the three
demonstrated by Wang and Al-Qadi [30], who utilized 3D finite-element recycled materials, BABC, as well as the upper and lower limits of
modeling analysis using ABAQUS and simulated a viscoelastic pavement Gradation C of ASTM-D1241 [32] for road base material, is presented in
response with a nonlinear anisotropic base support under moving Fig. 1. The PSD of recycled demolition materials fits within the upper
vehicular loading. They concluded that modulus distribution in the and lower limits of Gradation C except for slight deviations for RCA. The
granular base layer is affected not only by wheel load and pavement PSD of BABC does not completely fall within the ranges of Gradation C;
structure characteristics, such as asphalt layer stiffness and subgrade however, it falls within Gradation B limits of the ASTM-D1241 [32]
support, but also by temperature and vehicle speed because of the guideline. Fig. 1 also shows the maximum dry density (MDD) and op­
viscoelastic nature of the asphalt surface layer. Similarly, Kim et al. [28] timum moisture content (OMC) of the materials obtained following the
demonstrated the influence of incorporating non-linear models on the modified Proctor compaction method [33].
pavement response analysis of base and subbase layers and compared
the results against the linear elastic analysis. They found a significant Determination of resilient modulus properties
influence on the horizontal tensile strains that is a major source of fa­
tigue cracking. Laboratory testing of Mr is required for Level 1 MEPDG input. The
An objective of this study was to compare the stress–strain responses determination of the Mr is typically carried out through repeated load
achieved by incorporating three well established non-linear Mr models triaxial (RLT) testing. In this research, RLT testing was carried out
with those obtained through linear elastic analysis. Another objective following AG-PT/T053 [34] for RCA and WR and AASHTO-T307-99
was to demonstrate the necessity of incorporating the non-linear stress- [35] procedure for CB. Both procedures apply various combinations of
dependent geomaterial behavior into the numerical simulation to more applied dynamic vertical and static confining stresses. In the AG-PT/
accurately predict the pavement responses, which is essential in the T053 [34] method confining stresses are in a range of 20 to 150 kPa,
framework of a mechanistic-empirical pavement design approach. and the applied deviator stresses vary between 100 and 600 kPa. The
Furthermore, a comparative analysis of the predicted service life using range of confining stress and deviator stress in AASHTO-T307-99 [35] is
the linear elastic approach and the proposed non-linear approach was about 20 to 140 kPa and 20 to 275 kPa, respectively. RLT test results for
undertaken. In this regard, the experimentally obtained Mr properties of BABC were extracted from Chow et al. [31], who followed the AASHTO-
the typical virgin material and the demolition wastes were used for T307-99 [35] procedure.
modeling the ABC in a flexible pavement structure. In addition, the fa­ RLT specimens were prepared using a split mold with a 100 mm
tigue and rutting performance of the demolition wastes was evaluated diameter and 200 mm height. The recycled materials were compacted at
and compared against a type of virgin base course material. their OMC using the modified compaction energy [33] in eight layers.
Next, compacted specimens were carefully removed from the split mold,
Materials and methods quickly placed on the RLT cell pedestal to avoid loss of moisture and a
triaxial rubber membrane was placed over the sample using a vacuumed
Materials membrane expander. Next, the vacuum was released, the membrane
expander was removed and the membrane was sealed to the pedestal
Three types of selected recycled construction and demolition waste and the top loading cap with a set of O-rings following AASHTO-T307-
aggregates, being recycled concrete aggregate (RCA), crushed brick 99 [35]. For the preparation of the granular samples, as a measure to
(CB), and waste excavation rock (WR), were used in this research. A achieve the maximum consistency, two control sieves were used to split
North Carolina typical quarry material, namely Belgrade, suitable for the materials into three portions: less than2.36 mm, 2.36 less than 9.5
the aggregate base course was used as a control benchmark material. mm, and greater than 9.5 mm. These three portions were mixed in
In this research, the physical and stiffness properties of the recycled specific percentages, achieved from the PSD of the as-received materials

Fig. 1. PSD and compaction test results of recycled materials and BABC.

3
E. Yaghoubi et al. Transportation Geotechnics 30 (2021) 100599

as presented in Fig. 1. With this measure for consistency and compacting asphalt concrete (S9.5B) layer over an ABC with the thickness of 15.24
the samples using automatic compactors, and hence, a high degree of cm (6 in.) and an A6 subgrade. The S9.5B asphalt mixture is commonly
control in sample preparation, one specimen was prepared and tested for used in North Carolina as the surface course. This mixture is a hot mix
each material. A total of 60, 15, and 60 datasets were achieved through asphalt made of PG 58–28 asphalt and 40% of reclaimed asphalt pave­
RLT testing on RCA, CB, and WR, respectively. Also, 15 Mr datasets were ment. Stiffness properties of the control thin pavement model, including
available for BABC in Chow et al. [31] report. The model parameters for S9.5B, BABC, and A6 subgrade are presented in Table 1.
the three Mr predictive models were obtained through regression anal­ A simplified viscoelastic continuum damage model was employed in
ysis using these data sets. These models included bulk stress model the FlexPAVE™ simulations to characterize the asphalt concrete layer
proposed by Hicks and Monismith [36], deviatoric stress model pro­ with growing damage. However, as the current study focuses only on the
posed by Uzan [37] and modified universal (also known as the octahe­ pavement response, the relaxation modulus of asphalt concrete, repre­
dral) model recommended by AASHTO [38] as presented in Eqs. (1), (2), sented in the Prony series form was only required as the input data for
and (3), respectively. These models were developed for granular mate­ FlexPAVE™ simulations. The viscoelastic properties in terms of the
rials and the octahedral model is used in the MEPDG design procedure Prony series (Eq. (4)) were obtained by performing the dynamic
[11]. modulus (|E*|) tests on the mixture considered for the study following
AASHTO-T-378. The measured dynamic modulus at varying frequencies
θ
Mr = k1 .( )k2 (1) and temperature with the aid of time–temperature shift factors are fitted
pa
to the generalized Maxwell model to obtain the Prony coefficients. The
( )k 2 ( )k 3 Prony coefficients are presented in Table 2.
θ σd
Mr = k1 .pa . . (2) ∑
m
pa pa
E(t) = E∞ + Ei e− t/ρi
(4)
( )k 2 ( )k3 i=1
θ τoct
Mr = k1 .pa . . +1 (3) where, E∞ is the modulus of elasticity; t is time; Ei and ρi are the
pa pa
Prony fitting coefficients of the ith Maxwell element.
Where, k1, k2, and k3 are regression parameters, pa is the atmospheric For modeling in FlexPAVE™, a rectangular shape with an aspect
(normalizing) pressure, θ (=σ1 + σ 2 + σ3) is the bulk stress, σd is the ratio of (length/width) 11/7 was taken as the tire-pavement contact area
deviator stress and, and τoct is the octahedral shear stress. and a contact pressure of 827.4 kPa was assumed [39].
√̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅
(=13 (σ 1 − σ2 )2 + (σ 2 − σ 3 )2 + (σ3 − σ 1 )2 , where σ 1, σ2, and σ3 are the
principal stresses). The principal stresses of an element match the Car­ Determination of the aggregate base resilient moduli
tesian coordinates of the pavement sections when the wheel load is
exactly above the element in question. The principal stresses start to The Mr value at a specific depth location in an unbound layer de­
rotate (w.r.t in the Cartesian coordinates) as the wheel load moves away pends on the stress state that is induced by the wheel load and at-rest
from the element [14]. However, as the current stress state of interest is condition. Besides, the stiffness modulus of the constituent layers in a
on an element exactly underneath the wheel load, σ1 = σZ and σ2 = σ3 = pavement structure directly influences the stress distribution. Thereby
σh, where the Cartesian axis z is along the depth of the pavement, x is in an iterative method was adopted to calculate the Mr values [40–42]. The
transverse direction and y is in the traffic direction. recommended procedure involves the following steps that are adopted
in the MEPDG and the Pavement ME software [43].
Modeling in FlexPAVETM
1 Initial moduli are assumed for the unbound layers.
In the current study, stress–strain responses of the pavement struc­ 2 A stress state developed due to the wheel load is determined for mid-
ture were evaluated using the FlexPAVE™ program at 23 ◦ C pavement depth of each unbound layer (or sublayer) following Huang [41].
temperature under realistic loadings of moving vehicles. A typical North 3 The stresses obtained in Step 2 is used for the determination of the
Carolina thin pavement structure (Fig. 2), was analyzed under a 40 kN total stress state developed due to the wheel load and the overburden
single wheel load (single axle load of 80 kN) moving at a speed of 80 km/ pressure. Eq. (5) is used to compute the vertical component of the
hour. The pavement structure comprised of a 10.16 cm (4 in.) thick stress state (σz).
σ z = σ z(load) + σz(pave) (5)

where, σz(load) is the vertical pressure developed due to the wheel


load and σz(pave) is the at-rest vertical pressure from the overburden. σ z
(load) is computed using elastic layer theory and σ z(pave) is obtained using
Eq. (6).

n− 1
σ z(pave) = hn .γn + hi .γ i (6)
i=1

where, i is the layer above the nth pavement/soil layer for which the
Mr is being estimated (i = 1 is the surface course), hi is the thickness of
the ith layer or depth from the surface of the nth layer to the stress
calculation point of interest, and γ i is the unit weight of the ith layer.

Table 1
Thickness and stiffness properties of the materials in the control thin pavement.
Type of layer Name Thickness Poisson’s Ratio Modulus (MPa)

Asphalt Concrete S9.5B 10.16 (4 in.) 0.35 Table 2


Aggregate Base BABC 15.24 (6 in.) 0.4 102.0
Subgrade A6 Infinite 0.45 68.95
Fig. 2. Details of the thin pavement profile.

4
E. Yaghoubi et al. Transportation Geotechnics 30 (2021) 100599

Table 2 Although the procedure described above was relatively simple, its
Prony coefficients for the asphalt concrete. implementation was time-consuming due to the need to perform a
Ti (s) Ei (kPa) Ti (s) Ei (kPa) multilayer elastic theory (MLET) analysis run and the possibility of the
need for several iterations. The MLET analysis in this study was carried
2.00E + 11 2,380 2.00E-01 3,173,760
2.00E + 10 4,130 2.00E-02 3,761,120 out in the FlexPAVE™ program. With the aid of MATLAB code and Excel
2.00E + 09 7,330 2.00E-03 3,104,720 macros, the stress state at the mid-depth of each unbound layer (or
2.00E + 08 13,420 2.00E-04 2,497,320 sublayer) was calculated rapidly.
2.00E + 07 25,710 2.00E-05 1,851,650
2.00E + 06 52,450 2.00E-06 1,323,710
2.00E + 05 115,270 2.00E-07 917,490
Experimental results and analysis
2.00E + 04 270,180 2.00E-08 624,310
2.00E + 03 641,430 2.00E-09 419,320 Results of RLT testing are presented in Fig. 3 in a Mr-bulk stress plot.
2.00E + 02 1,401,350 2.00E-10 279,300 The power trend line in Fig. 3 shows the k1 and k2 parameters of the bulk
2.00E + 01 2,533,100 2.00E-11 185,000
stress model. Increased bulk stress as a result of increased confining
2.00E + 00 3,595,620 E∞ 60,490
stress and deviatoric stress, leads to a greater Mr. Higher confining stress
causes greater aggregate interlock, and greater deviator stress prompts
The estimation of the at-rest pressure coefficient, K0, was required for stress hardening, both of which result in lower strains and hence, greater
computing the contribution of the overburden stresses to the total lateral Mr [44,45]. In general, RCA and CB showed greater Mr and WR showed
stress. For cohesive soils, K0 is calculated using the Poisson’s ratio, μ, lower Mr compared to the typical virgin material used in this research.
according to Eq. (7). The value of μ used in the current research for the Higher values of Mr in RCA samples may be attributed to the fact that
subgrade was 0.45. RCA aggregates have rougher surfaces compared to CB and WR.
μ Yaghoubi et al. [23] observed micrographs from scanning electron mi­
K0 = (7) croscopy of RCA and CB and concluded that while almost the entire
1− μ
surface of RCA particles are rough, CB particles contain both rough and
For non-cohesive materials, K0 is calculated as a function of the angle smooth zones. In general, granular materials with rougher particle sur­
of internal friction, ϕ (in radians), and is obtained using Eq. (8). Internal faces are known to exhibit greater resilient modulus [14]. Fig. 3 also
friction angles used in this research for RCA, CB, WR and BABC were revealed that CB exhibited lower Mr values compared to RCA at bulk
49◦ , 48◦ , 51◦ and 42◦ , respectively. stresses greater than 185 kPa, while greater Mr values were observed at
K0 = 1 − sinϕ (8) bulk stresses lower than 185 kPa. This indicated the importance of
estimating the Mr of granular base material at different depths with
The total lateral stress, σ 3, used for the determination of the resilient various vertical and confining stress combinations. The Mr of granular
modulus at the depth of interest is obtained from Eq. (9). materials is highly stress-dependent [13].
In order to select the most accurate, Mr predictive model between
σ 2,3 = σh(load) + K0 σ z(pave) (9)
Equations 1 to 3, a regression analysis was carried out to achieve model
Where, σh(load) is the horizontal stress due to the wheel load on the parameters (k1, k2, and k3). In this regard, the root mean square devia­
pavement surface determined using elastic layer theory. tion (RMSD) for each data set was calculated following Azam et al. [46].
The %RMSD (Eq. (10)) is a measure for quantifying the differences be­
• The values of Mr are estimated using Equations 1 to 3 and vertical tween predicted and experimentally obtained values of Mr, in which
and lateral stresses computed in Step 3. lower RMSD means a more accurate prediction of data.
• The Mr obtained in Step 4 is compared with the Mr initially assumed √̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅
∑ [( )/
for each unbound layer. In case of significant difference, Steps 2 Mrp − Mrm Mrm ]2
%RMSD = 100 (10)
through 4 should be repeated as the 2nd iteration, and the iterations n
should continue until the assumed and computed Mr values
Where, Mrp is the predicted Mr, Mrm is the experimentally measured
converge. In this research, a difference of less than 5% between the
Mr, and n is number of available datasets, i.e., 60 for RCA and WR and
assumed and obtained values of Mr was considered acceptable.

Fig. 3. Resilient modulus vs bulk stress plots for RCA, CB, WR and BABC.

5
E. Yaghoubi et al. Transportation Geotechnics 30 (2021) 100599

15 for CB and BABC. with equal thickness of 15.24 mm. Interestingly, while RLT test results of
In this research, in order to obtain the model parameters with the RCA showed the highest values of Mr (Fig. 3), based on the stress levels
best fit to the experimental data, Microsoft Excel’s “solver” function was of the pavement modeled in FlexPAVETM, the applicable Mr value of CB
set to the objective of “minimum RMSD” by changing the values of k1 to was higher than that of RCA. This further illustrates the importance of
k3. The k parameters were achieved to predict Mr in MPa when input the significance of determination of the design Mr of structural pave­
parameters of pa, θ, σd and τoct were given in kPa. The goodness of fit of ment layers, based on the stress levels applicable to the pavement profile
each model was evaluated by calculating the coefficient of determina­ of interest.
tion (R2) as well as standard accuracy (Se/Sy) through comparison of the
predicted versus measure values of Mr. In the standard accuracy frac­ the proposed approach and effect of number of sublayers
tion, Se is the standard error of estimate and Sy is the standard deviation
[47]. The aggregate base course in the pavement structure tends to expe­
Values of k1, k2, and k3 together with %RMSD, Se/Sy and R2 corre­ rience various vertical and confining stresses at different depths.
sponding to each model and material are presented in Table 3. The Inputting different stress levels in the constitutive models presented in
evaluation of goodness of fit as presented in Table 3 was carried out Equations 1 to 3 results in various Mr values due to the stress-
following Witczak et al. [47] subjective criteria, in which Se/Sy ≤ 0.35 dependency of resilient modulus. The proposed approach of the cur­
and R2 ≥ 90 represent “Excellent”, 0.36 ≤ Se/Sy ≤ 0.55 and 0.70 ≤ R2 ≤ rent research to address the variation of Mr within the depth of ABC is to
0.89 represent “Good”, 0.56 ≤ Se/Sy ≤ 0.75 and 0.40 ≤ R2 ≤ 0.69 divide this layer into sublayers (SL). The applied stresses due to the
represent “Fair”, and 0.76 ≤ Se/Sy ≤ 0.90 and 0.20 ≤ R2 ≤ 0.39, traffic load and surcharge of top pavement layers can be estimated using
represent “Poor” fit. All three models showed an excellent fit with the the method explained in Section 3.4. In order to investigate the effect of
experimentally obtained Mr values; however, in general, the octahedral the sub-layering of ABC in the response analysis, two pavement profiles
model showed the greatest fit. Therefore, for the Mr analysis of the were modeled in FlexPAVETM. In one of the models, ABC was modeled as
unbound base layers in the following sections of this study, k parameters a single layer and in the other, ABC was divided into 10 sublayers. Both
obtained for the octahedral model were adopted. Fig. 4 demonstrates the models had the same asphalt concrete and subgrade types and thick­
predicted vs. measured Mr values for the materials of this research using nesses as shown in Fig. 2. The RLT test results of BABC were used, and
the octahedral model. Values of R2 being close to 1 and the concentra­ the Mr value corresponding to each sublayer was determined. The
tion of the data sets on the 1:1 line shows an “Excellent” goodness of fit. analysis was carried out under a 40 kN single wheel load.
In this research, in order to select the Mr corresponding to the stress Fig. 5a and 5b show the distribution of the vertical stress (σz) in the
levels of interest, a series of iterations were carried out using FlexPA­ depth of the pavement for non-sublayered ABC (NSL) and the ABC with
VETM outputs and the octahedral model with determined k parameters. 10 sublayers (10SL), respectively. The stress bulb diagrams of NSL and
In the first iteration, the average of Mr values obtained from RLT testing 10 SL notably resemble each other, indicating that the vertical stress
was defined in the ABC properties of FlexPAVETM as an initial Mr, and distribution in both models was similar. However, the strain bulb dia­
the response analysis was carried out. Next, the vertical and horizontal grams in Fig. 5c and 5d show different vertical strains (εz). Accumula­
stresses at several depths of ABC were extracted from the response tion of vertical strains results in rutting in pavements and was hence
analysis and input in the octahedral model to achieve the response essential to be investigated.
values of Mr. In the second iteration, the average response values of Mr In order to further investigate the effect of sub-layering, transverse,
obtained in the first iteration was taken as the input Mr value to carry and longitudinal strains were also extracted and presented in Fig. 6.
out another response analysis. The iterations were continued until the Transverse and longitudinal strains result in fatigue cracking in pave­
input and response Mr values matched. Following this approach, the ments and are hence crucial to be investigated. While both NSL and 10
values of Mr for RCA, CB, WR, and BABC were obtained to be 132, 160, SL models resulted in close longitudinal strains (Fig. 6c and 6d), the NSL
75, and 106 MPa, respectively. These values were the average of Mr model showed greater transverse strain, especially on the top of the ABC.
values for each material obtained at the mid-depth of 10 ABC sublayers The variation of vertical and transverse strains in NSL and ten models
under the same stress conditions indicates the importance of the sub-
Table 3 layering of ABC for more accurate response analysis. A higher number
Model parameters and the corresponding coefficient of determination for all of sublayers was expected to result in a more precise response analysis.
materials. On the other hand, too many sublayers made the modeling and analysis
Model Parameter RCA CB WR BABC time-consuming.
In order to determine the optimum number of sublayers, five pave­
Bulk Stress k1 132.67 139.92 58.994 88.154
Model k2 0.682 0.508 0.790 0.708
ment profiles were modeled in FlexPAVETM. All models had the same
R2 0.986 0.959 0.937 0.991 asphalt concrete and subgrade types and thicknesses, as shown in Fig. 2.
%RMSD 3.06 2.98 7.27 1.79 However, the ABC was divided into 5, 10, 15, and 30 sublayers with
Se/Sy 0.12 0.21 0.25 0.10 equal thicknesses to be compared with an ABC with no sublayers (NSL).
Goodness of Excellent Excellent Excellent Excellent
Vertical, transverse, and longitudinal stresses obtained from FlexPA­
fit
Deviatoric k1 1.237 1.609 0.547 0.962 VETM were input in the octahedral model to determine the Mr at the mid-
Stress Model k2 0.761 0.378 1.023 0.619 depth of each sublayer. Fig. 7 shows the variation of Mr in depth below
k3 − 0.079 0.131 − 0.262 0.091 the surface of the pavement for the five abovementioned models. The
R2 0.988 0.981 0.960 0.998 estimated Mr values for all sub-layered models match well.
%RMSD 2.94 2.60 6.80 1.04
Se/Sy 0.11 0.15 0.20 0.05
For investigating the effect of the number of sublayers, five pavement
Goodness of Excellent Excellent Excellent Excellent profiles were developed in FlexPAVETM with 5, 10, 15, 30, and no
fit sublayers of ABC. The vertical strain at the bottom of ABC and transverse
Octahedral k1 1.294 1.342 0.618 0.863 and longitudinal strains at the top of ABC were next extracted following
Model k2 0.764 0.387 0.996 0.640
the response analysis.
k3 − 0.136 0.394 − 0.432 0.202
R2 0.991 0.989 0.962 0.997 Fig. 8 shows the vertical strain vs. time for non-sub-layered (NSL)
%RMSD 2.74 1.64 6.23 1.40 and sub-layered models. As the number of sublayers increased, the
Se/Sy 0.10 0.11 0.20 0.06 percentage of difference in vertical strain compared to the model with
Goodness of Excellent Excellent Excellent Excellent fewer sublayers decreased. There was notably a relatively significant
fit
increase in εz from the NSL model to the 5SL model. There were also

6
E. Yaghoubi et al. Transportation Geotechnics 30 (2021) 100599

Fig. 4. Measured versus predicted Mr values using Octahedral model for (a) RCA, (b) CB, (c) WR, and (d) BABC.

Fig. 5. Distribution of vertical stress in the model with (a) no sublayers and (b) 10 sublayers, and vertical strain in the model with (c) no sublayers and (d)
10 sublayers.

relatively lower differences in εz between the sub-layered models. The Fig. 9 presents the transverse and longitudinal strains versus time for
maximum εz at ABC-bottom for NSL and 5SL models were 661 and 732, all models. Similar to the vertical strain response, transverse strain
respectively (10.8% difference). The difference between maximum εz for exhibit a relatively significant difference (2.6%) between NSL and 5SL
5SL and 10SL was 1.2%, and the difference between 10SL, 15SL, and models. The difference between 5 and 30 SL models were less than 0.1%.
30SL was less than 0.2%. The longitudinal responses did not show a significant difference

7
E. Yaghoubi et al. Transportation Geotechnics 30 (2021) 100599

Fig. 6. Distribution of horizontal strains: εx in the model with (a) no sublayers and (b) 10 sublayers, and εy in the model with (c) no sublayers and (d) 10 sublayers.

higher Mr compared to that of BABC, at the ABC-bottom, the Mr of BABC


was greater. Similar behavior was reported by Lancieri et al. [16] in
which the Mr of RCA was greater than a type of virgin aggregate under
higher stress levels and lower under smaller stress levels.

Comparison of the vertical strains

Using the Mr values corresponding to each sublayer of each material,


pavements were modeled according to Fig. 2 with ten sublayers of ABC
and 40 kN single wheel load. Three dimensional (3D) plots of Fig. 11
show the maximum vertical strains achieved at the top of the ABC for the
materials of this research. The values in brackets present the vertical
strains at the bottom of ABC. Based on the plots of Fig. 11, WR with the
Fig. 7. Variation of Mr in sub-layered models with depth. least Mr exhibited the greatest εz, and CB with the greatest Mr, showed
the least vertical strain. RCA and CB showed lower vertical strains
between non-sublayered and sub-layered models (less than 0.5%). compared to the typical virgin base material (BABC) and hence, less
Therefore, in the following sections of this study, pavements with ten potential for rutting. Therefore, they can be used for the construction of
sublayers of ABC were modeled in FlexPAVETM for the response analysis. the typical North Carolina thin pavements (Fig. 2) that are suitable for
low to medium trafficked roads. However, for using WR as the ABC,
Pavement response analysis and discussions either a thicker ABC is required to be designed and constructed, or WR
needs to be cement/lime treated.
In order to obtain the Mr corresponding to each of the ten sublayers An interesting point to note by observation of the vertical strain re­
of the recycled materials, first, a non-sublayered pavement system sponses at the top and bottom of the ABC was the difference in εz of ABC-
(Fig. 2) was modeled (LEA) and analyzed in FlexPAVETM. The Mr values Top and ABC-Bottom. In RCA, εz at the top of ABC was lower than the
obtained in the previous sections were used for each recycled material. bottom of ABC, while for the rest of the materials, it was the opposite.
Next, the stresses obtained from the analyses at evaluation depths cor­ Fig. 12 compares the vertical strains at the bottom and top of ABC for
responding to the mid-depth of each of the ten sublayers (following RCA and BABC. Typically, in the ABC, less strain is expected in the lower
Huang [41]) were input in the octahedral model (Eq. (3)), and the Mr layers due to lower stress levels. However, it should be noted that the
values were determined following the procedure explained in Section response Mr of the materials at the bottom layers is also lower, resulting
3.4. Fig. 10 shows the values of Mr versus depth below the pavement in greater strains. Based on Fig. 10, the rate of reduction of Mr by depth
surface for the materials used in this research. The determined Mr values was the highest for RCA among all materials. The relatively more
of sublayers were next input in the FlexPAVETM model for stress–strain reduced Mr at the bottom of ABC could be the reason for greater εz
response analysis. Interestingly, while at the ABC-Top, RCA exhibits a despite lower stress levels at the bottom. Such behavior was not
observed in the analysis outcomes of the pavement model with a single

8
E. Yaghoubi et al. Transportation Geotechnics 30 (2021) 100599

Fig. 8. Vertical strain versus time at ABC-Bottom.

Fig. 9. Plots of (a) transverse strain and (b) longitudinal strain versus time at ABC-Top.

layer of RCA as the ABC. This further marks the importance of dividing Comparison of the transverse and longitudinal strains
the ABC into sublayers for the stress–strain response analysis of
pavements. Fig. 13 shows the transverse and longitudinal strains at ABC-Top for
all the materials of this research. The positive values of strains indicate
the tensile nature of horizontal stresses at this depth of pavement.
Accumulation of the tensile strains at the top of ABC (below the asphalt

9
E. Yaghoubi et al. Transportation Geotechnics 30 (2021) 100599

vehicular loading at the bottom of the surface layer (εt) is termed as the
fatigue strain and the compressive strain on the top of the subgrade (εz)
is termed as rutting strains. These two strains govern the fatigue and
rutting life of the pavements, respectively. Equations 10 and 11 are
widely used for the determination of fatigue and rutting life of pave­
ments, respectively [41].

Nf = f1 × ε−t f2
× E− f3
(10)

Nr = f4 × ε−z f5 (11)

Where f1 to f5 are regression coefficients, E is the average elastic


modulus of the asphalt concrete layer, Nf is the fatigue life in terms of

Fig. 10. Variation of Mr with depth in the 10SL and NSL models for RCA, CB,
WR and BABC.

surface course) causes fatigue cracking in pavements. Both εx and εy


were the greatest for WR and the lowest for CB, due to the lowest and
highest stiffness properties, respectively. Also, εx and εy of RCA and CB
were lower than that of typical quarried materials marking these de­
molition aggregates suitable materials for ABC.

Comparison of the fatigue and rutting performance

The mechanistic-empirical pavement design makes use of the stress/


strain state of a pavement section to predict the pavement performance.
The primary objective in the pavement design is to determine sufficient
Fig. 12. Vertical strain versus time ABC-Top and ABC-Bottom for RCA
thickness of the structural layers to resist two commonly observed dis­
and BABC.
tresses, being fatigue and rutting. The tensile strain generated during

Fig. 11. Vertical strain at ABC-Top for (a) RCA, (b) CB, (c) WR and (d) BABC.

10
E. Yaghoubi et al. Transportation Geotechnics 30 (2021) 100599

Fig. 13. Plots of (a) transverse strain and (b) longitudinal strain versus time at ABC-Top for all materials.

standard axle load repetitions, and Nr is the rutting life in terms of purposes the fatigue life obtained from Equations 10 and 11 should be
standard axle load repetitions. calibrated with field data [41]. In the current research, however, the
Using the Prony coefficients presented in Table 2, the average elastic purpose is the comparison of the fatigue and rutting performance of the
modulus of the S9.5B for the vehicular speed of 80 km/h and the pavements with different base course types and analysis methods, and
pavement temperature of 23 ◦ C is 4,224 MPa (613 ksi). Table 4 presents hence values of Nf and Nr were directly utilized.
various values of regression coefficients proposed by different agencies For comparison of the fatigue performances, one of the base layers,
for Eq. (10), adopted from Huang [41]. It should be noted that regres­ being BABC with 10 sublayers (BABC-10SL) was taken as the reference
sion coefficients are obtained from laboratory fatigue tests under con­ and the percentages of difference between Nf of BABC-10SL and other
ditions that may be different from the field; thus, for the real-life design scenarios were calculated. Table 4 presents the percentage difference
between Nf values corresponding to RCA-10SL, CB-10SL, WR-10SL and
Non-Sublayered BABC (BABC-NSL) compared to BABC-10SL as the
Table 4
Fatigue life equation coefficients proposed by different agencies [41], and per­
reference scenario. An important point obtained from Table 4 is that
centage of difference in Nf compared to BABC-10SL as reference. using the conventional LEA method gives a fatigue life that is 8 to 14%
lower than the method proposed in this paper. Such a difference can
Agency/ Proposed Regression Difference compared to BABC-10SL
Organization Coefficients (%)
result in the misestimation of the design life of the pavements or
adequate thickness of the layers for the design life. Percentages pre­
f1 f2 f3 RCA- CB- WR- BABC-
sented in Table 4 also show that using RCA and CB results in a 38 to 74%
10SL 10SL 10SL NSL
(Linear and 67 to 141% increase in the fatigue life of the pavement, respectively.
Elastic) Using WR as the ABC of the pavement, however, results in a 32 to 48%
Asphalt 7.95E- 3.291 0.854 38 67 –32 − 8
reduction of the pavement fatigue life.
Institute 02 Table 5 presents various f4 and f5 values recommended by different
Shell 6.85E- 5.671 2.363 74 141 − 48 − 14 agencies adopted from Huang [41] for Eq. (11). Table 5 also presents the
Research 02 percentage of the difference between Nr values of RCA-10SL, CB-10SL,
Transport and 1.66E- 4.32 0 53 95 − 39 − 11
WR-10SL and BABC-NSL compared to BABC-10SL as the benchmark.
Road 10
Research Evidently, following the LEA approach, BABC shows a rutting life that is
Laboratory 57–67% greater than the approach proposed in this paper. Such a sig­
Belgian Road 4.92E- 4.76 0 59 109 − 42 − 12 nificant difference can result in under-designing the pavement layer
Research 14 thicknesses and thus premature pavement distresses. Interestingly, with
Center

11
E. Yaghoubi et al. Transportation Geotechnics 30 (2021) 100599

Table 5 non-hydrated cement. In another study, Arm [48] carried out RLT tests
Rutting life equation coefficients proposed by different agencies [41], and per­ on several identical CDW specimens containing crushed concrete ag­
centage of difference in Nr compared to BABC-10SL as reference. gregates and identical natural aggregates for comparison after various
Agency/ Proposed Difference compared to BABC-10SL (%) storing times, between 1 day to 730 days. She also carried out field
Organization Regression monitoring on pavement sections in which their ABC was constructed
Coefficients with RCA in comparison with those made of natural aggregates using
f4 f5 RCA- CB- WR- BABC-NSL FWD. Both laboratory and field results indicated a growth in stiffness
10SL 10SL 10SL (Linear modulus of CDW over time attributed to self-cementing properties due
Elastic)
to the presence of non-hydrated cement.
Asphalt Institute 1.36E- 4.477 − 11 119 − 45 67 In this study, RCA evidently contained mortar and CB contained up
09 to 30% by mass of crushed concrete particles and mortar. Based on the
Shell Research (85% 1.94E- 4 − 10 101 − 41 58
reliability) 07
abovementioned studies, an improvement in the stiffness moduli of ABC
Transport and Road 6.18E- 3.95 − 10 100 − 41 57 materials made of RCA and CB is expected in the long run. The most
Research 08 important factor for mitigating the environmental impacts on pavement
Laboratory (85% systems, next to the proper selection of binding materials for surface, is
reliability)
proper drainage [42]. The hydraulic conductivity of RCA and CB used in
Belgian Road 3.05E- 4.35 − 10 114 − 44 64
Research Center 09 this research was 6.9 × 10− 8 m/s and 1.16 × 10− 6 m/s, respectively,
which is comparable with that of typical natural aggregates, reported to
be greater than 1 × 10− 9 m/s [18]. Thus, with respect to permeability
respect to rutting life, among the three demolition wastes, only CB re­ and drainage capability, RCA and CB can behave as a natural crushed
sults in a rutting life that is greater than that achieved by using BABC as rock. WR is essentially a type of natural crushed rock, even though of a
the base course. lower quality compared to those used for the ABC of roads, and hence,
has a negligible sensitivity with moisture changes [42]. The perme­
Discussion on the long-term environmental impacts ability of WR used in this research was 2.45 × 10-6 m/s which is greater
than 1 × 10− 9 m/s (typical of natural aggregates [18]), and hence can
Following the fatigue and rutting performance analyses carried out provide sufficient drainage.
in the previous section, a discussion on the effect of environmental Sufficient drainage is known to significantly alleviate the long-term
factors on the long-term performance of construction and demolition impact of environmental factors such as precipitation [42], and RCA,
wastes (CDW) used as ABC was deemed beneficial. Environmental fac­ CB and WR have a permeability comparable with natural aggregates
tors such as moisture variations, freeze–thaw or temperature changes [18]. Thus, if not improved due to self-cementation, insignificant
have a greater influence on the subgrade or the surface layer compared different in the long-term performance compared to natural aggregates
to the ABC [42]. Hence, more attention has been paid to environmental is expected for CDW used in the ABC layer. Nevertheless, an investiga­
impacts on the subgrade and top-most structural layer of pavement tion on the effect of environmental factors on the long-term performance
systems. This is may be due to the misconception that the ABC layer is of the recycled materials used in this research is recommended for future
covered by an impermeable asphalt concrete layer and is placed over a studies.
permeable subgrade soil. In reality, asphalt surfaces are not 100%
impermeable and subgrade soils such as clay can have very low Conclusions
permeability, and hence considerable moisture variations can occur in
the ABC [42]. As a result, environmental effects, on the behavior and This study investigated the resilient response of RCA, CB, and WR as
long-term performance of the ABC layer should not be underestimated. the aggregate base course of a typical pavement profile under traffic
Such effects in unbound granular layers of pavements have traditionally loadings through an experimental program and using the FlexPAVETM
been considered by implementing adjustment factors (obtained through computer program. The following outcomes were achieved in this study:
comparison of the laboratory and field data) on the resilient moduli of The mean value of RLT test results suggested that RCA has the
ABC materials [41]. In AASHTO’s Mechanistic-Empirical Pavement greatest resilient modulus, followed by CB, BABC, and WR. The RLT
Design Method [11], the long-term effects of temperature changes and datasets for each material achieved under various stress levels were used
moisture variations in the performance of the pavement structure are to determine the model parameters of three well-known resilient
taken into account using an advanced modeling tool named the modulus predictive models. The octahedral model showed the best
Enhanced Integrated Climatic Model (EICM). goodness of fit and hence the most accurate predictions.
Studies on the long-term performance of CDW materials such as Under the stress levels applicable to the pavement structure and
those used in this research is very limited, possibly due to the physical loadings of this study (bulk stresses less than 185 kPa), CB exhibited the
properties of these materials being comparable to those of natural ag­ greatest resilient modulus despite RCA showing the greatest Mr values
gregates. A difference between CDW aggregates, such as RCA and CB, under stress levels of the RLT testing. This signifies the importance of
and natural aggregates is the presence of mortar and non-hydrated interpretation of RLT test results for the selection of the proper design
cement which may result in lower permeability of these materials resilient modulus for the applicable stresses within the ABC of each
compared to natural unbound aggregates [18]. However, the presence specific pavement structure.
of non-hydrated cement can have positive impacts as well. In a field Under the same loading conditions, in spite of the same distribution
study, Lancieri et al. [16] investigated the performance of two pavement of stresses, different response strains were observed through linear
structures in which their subbase was constructed with CDW. The per­ elastic analysis (LEA, pavement model with a single ABC layer) and the
formance evaluation was carried out 4 years and 8 years after the con­ pavement model in which the ABC was divided into ten sublayers. As an
struction using non-destructive techniques such as Falling Weight example, the vertical strain at the top of ABC demonstrated an
Deflectometer (FWD) and laser profilometer. Interestingly, Lancieri approximately 13% difference between the two models.
et al. [16] reported an improvement in the stiffness moduli of the CDW A scheme was proposed and followed for the incorporation of the
from the construction to the fourth and eighth years after the con­ octahedral model in FlexPAVETM response analysis. In this approach,
struction and concluded that the improvement was attributed to two instead of a single ABC with an assigned resilient modulus, the base
factors, being stress hardening over time due to traffic loads and residual course was divided into several sublayers, each of which was assigned a
self-cementing properties of CDW, which typically contain mortar and resilient modulus corresponding to the horizontal and vertical stresses of

12
E. Yaghoubi et al. Transportation Geotechnics 30 (2021) 100599

the mid-depth of the sublayer. Several models were developed and [3] Poon CS, Chan D. Feasible use of recycled concrete aggregates and crushed clay
brick as unbound road sub-base. Constr Build Mater 2006;20(8):578–85.
analyzed with a various number of sublayers from 5 to 30. A negligible
[4] Delongui L, et al. Construction and demolition waste parameters for rational
difference (less than0.2%) was observed in the response strains of pavement design. Constr Build Mater 2018;168:105–12.
models with 10 and a greater number of sublayers. Therefore, modeling [5] Gabr A, Cameron D. Properties of recycled concrete aggregate for unbound
and analysis of the typical virgin aggregates and demolition wastes were pavement construction. J Mater Civ Eng 2012;24(6):754–64.
[6] Yaghoubi E, Disfani MM, Arulrajah A, Kodikara J. Development of a void ratio-
continued by modeling the ABC with ten sublayers. moisture ratio-net stress framework for the prediction of the volumetric behavior
Under the same loading regime and subgrade conditions, RCA and of unsaturated granular materials. Soils Found 2019;59(2):443–57.
CB demonstrated less vertical and horizontal strains compared to the [7] Wang YD, Keshavarzi B, Kim YR. Fatigue performance prediction of asphalt
pavements with FlexPAVETM, the S-VECD model, and DR failure criterion. Transp
typical quarried material used in this research. Response analysis of WR, Res Rec 2018;2672(40):217–27.
on the other hand, showed the greatest vertical and horizontal strains. A [8] Cao W, Norouzi A, Kim YR. Application of viscoelastic continuum damage
thicker ABC or treatment using cement or lime could result in the approach to predict fatigue performance of Binzhou perpetual pavements. Journal
of Traffic and Transportation Engineering (English Edition) 2016;3(2):104–15.
comparable performance of WR to typical virgin aggregates. [9] Norouzi A, Kim D, Kim YR. Numerical evaluation of pavement design parameters
Comparison between the fatigue and rutting performances of the for the fatigue cracking and rutting performance of asphalt pavements. Mater
demolition wastes as the base course layer of pavements showed that CB Struct 2016;49(9):3619–34.
[10] Padula FR, Nicodemos S, Mendes JC, Willis R, Taylor A. Evaluation of fatigue
resulted in greater rutting life and fatigue life, RCA resulted in greater performance of high RAP-WMA mixtures. Int J Pavement Res Technol 2019;12(4):
fatigue life, but slightly lower rutting life, and WR resulted in lower 430–4.
rutting life and fatigue life compared to virgin aggregates. [11] AASHTO, Mechanistic-Empirical Pavement Design Guide: A Manual of Practice.
2008: Washington, DC.
Fatigue and rutting performance of the pavement system with the
[12] Cardoso R, Silva RV, de Brito J, Dhir R. Use of recycled aggregates from
same base course material (BABC) but using the conventional LEA construction and demolition waste in geotechnical applications: A literature
approach in comparison with the scheme proposed in the current study review. Waste Manage 2016;49:131–45.
showed a significant difference in the anticipated fatigue and rutting life [13] Cerni G, Corradini A, Pasquini E, Cardone F. Resilient behaviour of unbound
granular materials through repeated load triaxial test: influence of the conditioning
of the pavement. This emphasizes the importance of incorporating the stress. Road Materials and Pavement Design 2015;16(1):70–88.
resilient modulus models and a sublayered model in the pavement [14] Lekarp F, Isacsson U, Dawson A. State of the art. I: Resilient response of unbound
response analyses carried out by FlexPAVETM instead of using the con­ aggregates. J Transp Eng 2000;126(1):66–75.
[15] Cristelo N, Vieira CS, de Lurdes Lopes M. Geotechnical and geoenvironmental
ventional LEA approach. assessment of recycled construction and demolition waste for road embankments.
In this study, a discussion on the impact of the environmental factors Procedia Eng 2016;2016(143):51–8.
on the long-term performance of construction demolition waste was [16] Lancieri F, Marradi A, Mannucci S. C&D waste for road construction: long time
performance of roads constructed using recycled aggregate for unbound pavement
provided. A laboratory and field investigation on the effect of environ­ layers. WIT Trans Ecol Environ 2006;92.
mental factors on the long-term performance of the recycled materials [17] Leite, F.d.C., dos Santos Motta, R., Vasconcelos, K.L., and Bernucci, L., Laboratory
used in this research is recommended for future studies. evaluation of recycled construction and demolition waste for pavements.
Construction and building materials, 2011. 25(6): p. 2972-2979.
The proposed approach aims to provide a more realistic response and [18] Arulrajah A, Piratheepan J, Disfani MM, Bo MW. Geotechnical and
performance analysis of the flexible pavements using FlexPAVETM. The Geoenvironmental Properties of Recycled Construction and Demolition Materials
findings of this research can also help improve the knowledge of the in Pavement Subbase Applications. J Mater Civ Eng 2013;25(8):1077–88.
[19] Saeed, A., Performance-related tests of recycled aggregates for use in unbound
resilient behavior of recycled aggregates in pavement applications. This
pavement layers. 2008.
can encourage the pavement engineers, authorities, and practitioners to [20] Imteaz MA, Arulrajah A, Horpibulsuk S, Ahsan A. Environmental Suitability and
further consider recycled aggregates in the design and construction of Carbon Footprint Savings of Recycled Tyre Crumbs for Road Applications.
highways. International Journal of Environmental Research 2018;12(5):693–702.
[21] Ghadimi B, Nikraz H. A comparison of implementation of linear and nonlinear
constitutive models in numerical analysis of layered flexible pavement. Road
CRediT authorship contribution statement Materials and Pavement Design 2017;18(3):550–72.
[22] Gungor OE, Al-Qadi IL, Gamez A, Hernandez JA. Development of adjustment
factors for MEPDG pavement responses utilizing finite-element analysis. Journal of
Ehsan Yaghoubi: Conceptualization, Methodology, Formal anal­ Transportation Engineering, Part A: Systems 2017;143(7):04017022.
ysis, Investigation, Writing - original draft, Writing - review & editing, [23] Yaghoubi E, Disfani MM, Arulrajah A, Kodikara J. Impact of compaction method on
mechanical characteristics of unbound granular recycled materials. Road Materials
Visualization. Nithin Sudarsanan: Conceptualization, Methodology, and Pavement Design 2017:1–23.
Formal analysis, Writing - review & editing. Arul Arulrajah: Concep­ [24] Yang S-R, Huang W-H, Liao C-C. Correlation between resilient modulus and plastic
tualization, Supervision, Writing - review & editing, Resources. deformation for cohesive subgrade soil under repeated loading. Transp Res Rec
2008;2053(1):72–9.
[25] Huang H, et al. Advanced analytical tool for flexible pavement design and
evaluation. In: Airfield and Highway Pavements 2019: Design, Construction,
Declaration of Competing Interest Condition Evaluation, and Management of Pavements. VA: American Society of
Civil Engineers Reston; 2019. p. 61–71.
[26] Kim M, Lee JH. Study on nonlinear pavement responses of low volume roadways
The authors declare that they have no known competing financial subject to multiple wheel loads. Journal of Civil Engineering and Management
interests or personal relationships that could have appeared to influence 2011;17(1):45–54.
the work reported in this paper. [27] Kim M, Tutumluer E. Validation of a three-dimensional finite element model using
airfield pavement multiple wheel load responses. Road materials and pavement
design 2010;11(2):387–408.
Acknowledgments [28] Kim M, Tutumluer E, Kwon J. Nonlinear pavement foundation modeling for three-
dimensional finite-element analysis of flexible pavements. Int J Geomech 2009;9
(5):195–208.
The authors would like to thank Dr Richard Kim of the North Car­ [29] Ng K, Henrichs ZR, Ksaibati K, Wulff SS. Resilient modulus of subgrade materials
olina State University for providing access to FlexPAVETM and technical for mechanistic-empirical pavement design guide. Road Materials and Pavement
advice. Design 2018;19(7):1523–45.
[30] Wang H, Al-Qadi IL. Importance of nonlinear anisotropic modeling of granular base
for predicting maximum viscoelastic pavement responses under moving vehicular
References loading. J Eng Mech 2013;139(1):29–38.
[31] Chow, L.C., Mishra, D., and Tutumluer, E., Aggregate base course material testing
and rutting model development. 2014. p. 116.
[1] Saberian M, Li J. Investigation of the mechanical properties and carbonation of
[32] ASTM-D1241, Standard Specification for Materials for Soil-Aggregate Subbase,
construction and demolition materials together with rubber. J Cleaner Prod 2018;
Base, and Surface Courses. 2015, ASTM International: West Conshohocken, PA.
202:553–60.
[33] ASTM-D1557-12, Standard Test Methods for Laboratory Compaction
[2] Magnusson S, Lundberg K, Svedberg B, Knutsson S. Sustainable management of
Characteristics of Soil Using Modified Effort (56,000 ft-lbf/ft3 (2,700 kN-m/m3)).
excavated soil and rock in urban areas–a literature review. J Cleaner Prod 2015;93:
2012, ASTM International: West Conshohocken, PA.
18–25.

13
E. Yaghoubi et al. Transportation Geotechnics 30 (2021) 100599

[34] AG-PT/T053, Austroads repeated load triaxial test method: Determination of [42] Papagiannakis AT, Masad EA. Pavement design and materials. 1st ed. NJ, USA:
permanent deformation and resilient modulus characteristics of unbound granular John Wiley & Sons; 2017.
materials under drained conditions. 2000, Austroads, Sydney, Australia. [43] Khazanovich L, Celauro C, Chadbourn B, Zollars J, Dai S. Evaluation of subgrade
[35] AASHTO-T307-99, Standard Method of Test for Determining the Resilient Modulus resilient modulus predictive model for use in mechanistic–empirical pavement
of Soils and Aggregate Materials. 2007, American Association of State Highway design guide. Transp Res Rec 2006;1947(1):155–66.
and Transportation Officials, Washington, DC. [44] Nguyen B, Mohajerani A. Possible simplified method for the determination of the
[36] Hicks RG, Monismith CL. Factors influencing the resilient response of granular resilient modulus of unbound granular materials. Road Materials and Pavement
materials. Highway Res Rec 1971;345:15–31. Design 2016:1–18.
[37] Uzan J. Characterization of granular material. Transp Res Rec 1985;1022(1):52–9. [45] Puppala AJ, Hoyos LR, Potturi AK. Resilient Moduli Response of Moderately
[38] AASHTO, Guide for Design of New and Rehabilitated Pavement Structures. 2002, Cement-Treated Reclaimed Asphalt Pavement Aggregates. J Mater Civ Eng 2011;
National Cooperative Highway Research Program: American Association of State 23(7):990–8.
Highway and Transportation Officials, Washington, D.C. [46] Azam AM, Cameron DA, Rahman MM. Model for Prediction of Resilient Modulus
[39] De Beer M, Fisher C, Kannemeyer L. Towards the application of Stress-In-Motion Incorporating Matric Suction for Recycled Unbound Granular Materials. Can
(SIM). results in pavement design and infrastructure protection. 2004. Geotech J 2013;50(11):1143–58.
[40] Ullidtz P. Modelling flexible pavement response and performance. Denmark: Tech [47] Witczak M, Kaloush K, Pellinen T, El-Basyouny M, Von Quintus H, Report NCHRP,
Univ. of Denmark Polytekn; 1998. et al. TRB. Washington, D. C.: National Research Council; 2002.
[41] Huang, Y.H., Pavement analysis and design. 2004, Upper Saddle River, NJ: [48] Arm M. Self-cementing properties of crushed demolished concrete in unbound
PEARSON EDUCATION LIMITED. layers: results from triaxial tests and field tests. Waste Manage 2001;21(3):235–9.

14

You might also like