Transportation Research Part D 94 (2021) 102811

Contents lists available at ScienceDirect

Transportation Research Part D

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

Impact of dynamic wheel load on roadway

infrastructure sustainability
Seunggu Kang a, *, Imad L. Al-Qadi b, Osman Erman Gungor b
a
Department of Civil Engineering, Merrimack College, 315 Turnpike St., North Andover, MA 01845, United States
b
Department of Civil and Environmental Engineering, University of Illinois at Urbana-Champaign, 205 North Mathews Ave., Urbana, IL 61801,
United States

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

Keywords: Dynamic wheel load (DWL) may result in additional pavement damage. This study examined the
Pareto frontier impact of DWL on pavement performance using mechanistic tractor-trailer and three-dimensional
Life cycle assessment (3-D) viscoelastic finite element (FE) pavement models. Based on the result of FE simulations, a
Life cycle cost analysis
correlation between the pavement responses and DWL was established. Using the mechanistic-
Rolling resistance
MEPDG
empirical pavement design guide approach, the pavement service life was predicted. A case
Road roughness study based on life cycle assessment (LCA) and life cycle costing (LCC) methodologies was per­
formed to study the DWL impact on greenhouse gas (GHG) emissions, energy consumption, and
cost. The study identified the probabilistic distribution of pavement service life due to DWL. The
effect of pavement service life on the environmental impact of each LCA stage was investigated.
Finally, optimal sustainable solutions considering both environmental and economic impacts
were introduced using a Pareto frontier analysis.

1. Introduction

Pavement surface smoothness is an important indicator of pavement performance and imparts a number of benefits, including
improved pavement service life, reduced fuel and vehicle operating cost (VOC) (FHWA, 2016). Road roughness excites vehicle systems,
resulting in repeated dynamic wheel loads on pavement that can cause premature localized pavement failure. While dynamic load
coefficient (DLC) is defined as the standard deviation of dynamic wheel load (DWL) divided by the mean of DWL (Eq. (5)), various
studies show a positive correlation between DLC and road roughness. The DLC of heavy trucks can range from 0.4 to 3.1 (Sweatman,
1983; Lieh and Qi, 1995; Guo et al., 2012; Van Cuong et al., 2019). DWL is also closely related to pavement critical responses. Various
researchers have determined the pavement critical responses due to dynamic wheel load using instrumentation (Huhtala et al., 1989;
Al-Qadi and Bhutta, 1999; Al-Qadi et al., 2004; Al-Qadi et al., 2011) and numerical simulations such as finite element analysis (Al-Qadi
et al., 2005; Oh et al., 2007; Al-Qadi and Elseifi, 2007; Wang and Al-Qadi, 2009; Hernandez et al., 2013; Xia et al., 2015; Al-Qadi et al.,
2018). Some studies employed empirical transfer functions or mechanistic-empirical pavement design guide (MEPDG) to predict
pavement performance based on critical pavement responses (Al-Qadi et al., 2018) or other input conditions such as pavement ma­
terial properties and structures (Schwartz et al., 2013).
The aforementioned studies address various important aspects of vehicle-pavement interaction (VPI) and the corresponding impact
on pavement performance. However, some limitations have been reported, including the following:

* Corresponding author.
E-mail addresses: kangs@merrimack.edu (S. Kang), alqadi@illinois.edu (I.L. Al-Qadi), oermangungor@gmail.com (O.E. Gungor).

https://doi.org/10.1016/j.trd.2021.102811

Available online 20 April 2021

1361-9209/© 2021 Elsevier Ltd. All rights reserved.
S. Kang et al. Transportation Research Part D 94 (2021) 102811

• Linear elastic material properties have been assumed.

• Simplified tire imprints and contact stresses have been used.
• The relationship between pavement performance and critical pavement responses have not been established.
• Static wheel loads have been used to predict pavement performance.

In addition, VPI is typically responsible for the majority of the environmental and economic impacts during the pavement life cycle.
Hence, it is important to accurately predict these impacts. The objective of this paper is to establish research framework to investigate
the connection between DWL and roadway infrastructure sustainability. To accomplish this objective and to overcome aforementioned
limitations of existing literature, the followings were considered in this paper:

• Generation of DWLs using a mechanistic tractor-trailer model excited by random pavement roughness.
• Using realistic material properties in finite element analysis (FEA) including viscoelastic asphalt concrete (AC) layers and stress-
dependent base modulus.
• Utilizing measured tire contact stresses, which are non-uniform and three-dimensional.
• Critical pavement responses, from the finite element (FE) simulations, were used in the MEPDG approach for probabilistic pave­
ment performance prediction.
• Optimal sustainable solutions were identified based on greenhouse gas (GHG) emissions, energy consumption, and cost quanti­
fication. Life cycle assessment (LCA) and life cycle costing (LCC) were used.

2. Methdology

The research framework illustrated in Fig. 1 begins with generating random pavement roughness based on zero-mean stationary
white noise that passes through transfer functions. The tractor-trailer model was excited with a simulated pavement profile at various

Fig. 1. Description of research framework.

2
S. Kang et al. Transportation Research Part D 94 (2021) 102811

roughness levels to compute truck DWL. As a result, the correlation between DWL and various roughness levels was determined. Using
a 3-D FE viscoelastic pavement model and various levels of DWL, pavement critical responses were computed. The correlation between
DWL magnitude and pavement responses was then determined. For pavement performance prediction, the MEPDG’s distress and IRI
prediction models were adopted.
Using the distribution of pavement service life resulted from DWL, LCA and LCC of pavement based on previously developed life
cycle impact assessment (LCIA) and cost databases were carried out. The following sections discuss the generation of road roughness
and DWL, pavement FE models, pavement performance prediction, and pavement LCA and LCC in detail.

2.1. Pavement roughness

Pavement roughness is often characterized by power spectral density (PSD), defined by the International Standardization Orga­
nization (ISO8608:1995) as follows,
( )− ν
Ω
̂
S ξξ (Ω) = S0 (1)
Ω0

where Ω is spatial frequency, S0 is degree of roughness, Ω0 is reference spatial frequency at 1 rad/m, and ν is waviness of pavement
profile. The PSD may be fitted with a transfer function, Hξw (ω), that filters white noise, w(t), to produce pavement roughness, ξ(t),
evaluated at x = Vt, where V is vehicle speed. Because w(t) is a zero-mean stationary white noise, its mean and correlation function are
E[w(t) ] = 0 and E[w(t)w(t + τ) ] = 2πS0 δ(τ), respectively. Formulating this system into a state-space representation, the following
expression may be realized as shown in Eq. (2).
ẋr (t) = Ar xr (t) + Br w(t)

yr (t) = Cr xr (t) (2)

where yr (t) is road roughness vector, [ξ(t)ξ(t) ]T , and xr (t) is state vector. Ar , Br , and Cr are state, input, and output matrices,
˙
respectively. A sample illustration of random road roughness generated based on Eq. (2) is shown in Fig. 2.

2.2. Dynamic wheel loading

Pavement roughness excites driving vehicles, which produces DWL. DWLs vary in magnitude and may lead to premature localized
pavement failure. The type of wheel load is divided into static and dynamic loads. Static wheel load refers to the stationary vehicle
weight. The additional load comes from the dynamic vehicle-pavement interactions (VPI) caused by road irregularities. In this study,
the sum of the static and additional loads is referred to as the DWL. The heavy truck wheel load is a primary cause of load-related
pavement distresses, such as fatigue cracking and permanent deformation. This study therefore adopts the tractor-trailer model dis­
cussed in Kang et al.’s (2019) research to generate truck DWL (see Fig. 3). The tractor-trailer model represents an FHWA Class 9 truck
with five axles consisting of one steer axle and two tandem axles.
The magnitude of DWL depends on various parameters, such as the mechanical properties of suspensions and tires, vehicle mass,
and road roughness. The tractor-trailer model adopts vehicle mechanical properties from Law et al.’s (2002) study. Thus, the only
parameter to control is road roughness. This study employs the pavement roughness model discussed in Kang et al.’s (2019) work,
which translates white noise (or w(t)) into a random pavement profile, ξ(t). The DWL, FDWL (t), is expressed in Eq. (3).
FDWL = Fa + Fs (3)

Fig. 2. Generated random pavement roughness.

3
S. Kang et al. Transportation Research Part D 94 (2021) 102811

Fig. 3. A five-axle tractor-trailer model (after Kang et al. (2019)).

Here Fa and Fs are the additional wheel load, from the dynamic VPI, and the static wheel load, respectively. FDWL is calculated based on
the equation of motion. For the tractor-trailer model, Fa is expressed in Eq. (4).
[ ]
˙
Fa (t) = kt [xt (t) − ξ(t) ] + ct ẋt (t) − ξ(t) (4)

In Eq. (4), kt is the tire stiffness coefficient vector. xt (t) is the axle mass response vector. ξ(t) is the pavement roughness input vector,
and ct is the tire-damping coefficient vector. The single dot above xt (t) and ξ(t) in Eq. (4) refers to a first-order time derivative. For each
combination of pavement roughness (1.0 to 3.0 m/km) and axle (1 through 5), the statistics of the resulting DWL are analyzed. The
dynamic load coefficient (DLC) is introduced for the statistical characterization of DWL as follows:
σ
DLC = (5)
μ

here σ and μ represent the standard deviation and mean of DWL, respectively. The mean of half-axle forces for axles 1 through 5 are 29;
42; 42; 32; and 32 kN; respectively. It was found that there is a virtually linear positive correlation between DLC and IRI, which is in
agreement with other studies (Gillespie et al., 1993; Misaghi et al., 2021). For any arbitrary IRI, DLC at vehicle speed of 105 km/h can
be computed using the regression equation shown in Eq. (6).
DLCi (N) = ai ∙IRI (6)
In Eq. (6), i stands for the i axle in the tractor-trailer model as illustrated in Fig. 3. The regression coefficients needed to predict the
th

DLC of the ith axle are based on IRI are as follows: a1 =0.419, a2 =0.359, a3 =0.342, a4 =0.284, anda5 = 0.238.

2.3. 3-D finite element pavement models

The FEM is needed to account for more realistic conditions to model the tire-pavement interaction. Realistic conditions include
accurate tire contact area; 3-D non-uniform tire-contact stresses; pavement structure; and material properties. In this section, FEM is
used to model 3-D FE viscoelastic flexible pavements. Provided that a primary interest of this study is to quantify DWL’s effect on
pavement response and performance, the following section discusses the resulting pavement responses caused by tire-contact load
applied on 3-D FE pavement models.

2.3.1. Viscoelastic pavement materials

Asphalt concrete (AC) comprises various materials that include granular aggregates, asphalt binder, and mineral fillers. AC is a
viscoelastic material, and its modulus changes with temperature and time of loading. Dynamic modulus is a fundamental material
property describing the AC under different temperatures and loading frequencies. The MEPDG employs a sigmoidal function to fit the
dynamic modulus, as shown in Eq. (7) (Hernandez et al., 2016).
α
log(|E* | ) = δ + (7)
eβ− γlog(ξ)

In the above equation, |E* | represents the dynamic modulus, and ξ is a reduced angular frequency (Hz). δ is the minimum modulus
value. α is span of modulus values, and β and γ are shape parameters. The MEPDG also uses a second-order polynomial for the
time–temperature shift factor, as indicated in Eq. (8).

log(aT ) = α1 T 2 + α2 T + α3 (8)

here aT is a time–temperature shift factor for AC. T represents the temperature of interest, and α1 , α2 , and α3 are regression coefficients

4
S. Kang et al. Transportation Research Part D 94 (2021) 102811

(see Table 1).

This study considers two AC pavement structures — both thin and thick — and each section is comprised of three AC layers,
including the wearing surface and intermediate and base layers on top of granular aggregate and subgrade layers (see Fig. 4). The AC
sigmoidal function for dynamic modulus was converted into a Prony series for AC material characterization to be used in ABAQUS. For
the thin pavement, the thicknesses of the wearing course (WS), intermediate layer (IL), base layer (BL), and granular-aggregate layer
are 25 mm, 37.5 mm, 62.5 mm, and 150 mm, respectively. For the thick pavement, the thicknesses of WS, IL, BL, and the granular-
aggregate layer are 52.5 mm, 72.5 mm, 175 mm, and 300 mm, respectively. In this study, the thin and thick AC pavements are referred
to as AC125-B150 and AC300-B300, respectively (see Fig. 4). Here, AC indicates asphalt concrete, B indicates granular aggregate base,
and the numbers indicate the thickness in mm. Pavement FE models are also illustrated in Fig. 4.

2.3.2. Unbound materials

Although nonlinearity may significantly affect the pavement responses for thin AC pavements, its effect may become insignificant
for thick AC pavements where tire loads are mostly distributed in the thick AC layer and stress levels in underlying layers are relatively
small (Hernandez et al., 2016; Gungor et al., 2016). Given that the AC300-B300 pavement structure has a sufficient AC thickness, the
granular-aggregate layer and subgrade are modeled as linear elastic. For the AC125-B150 pavement structure, however, the granular-
aggregate layer is modeled as nonlinear anisotropic material, and the subgrade is modeled as linear elastic material due to a smaller AC
thickness. MEPDG’s resilient modulus model for nonlinear and anisotropic granular materials is shown in Eq. (9) (ARA, 2004a).
( )k 2 ( )k3
θ τoct
Mrv = k1 pa +1
pa pa
( )k 5 ( )k6
θ τoct
Mrh = k4 pa +1
pa pa
( )k8 ( )k 9
θ τoct
Mrs = k7 pa +1 (9)
pa pa

here Mrv ,Mrh ,and Mrs are resilient moduli in the vertical, horizontal, and shear directions, respectively. θ represents bulk stress, while
τoct is octahedral shear stress. pa is a normalizing stress, while k1 through k9 represent regression coefficients summarized in Table 2.

2.3.3. Finite element pavement model

ABAQUS scripts for 3-D FE AC pavement models were developed by Al-Qadi and his co-workers over the past two decades at Illinois
Center for Transportation (Al-Qadi and Elseifi, 2007; Yoo and Al-Qadi, 2007; Wang and Al-Qadi, 2009; Al-Qadi and Wang, 2012;
Gungor et al., 2016; Gungor et al., 2017; Hernandez and Al-Qadi, 2017; Al-Qadi et al., 2018). The scripts were modified to simulate the
pavement structures shown in Fig. 4. The length and width of the AC125-B150 pavement are 5220 mm and 4578 mm, respectively,
with wheel load magnitude of 26 kN. The length and width of the AC300-B300 pavement are 8320 mm and 7678 mm, respectively,
with the wheel load magnitude of 26 kN. The length of both pavement models increases slightly with the magnitude of the wheel load
because the tire contact area increases with the wheel load magnitude. The dimensions of both pavement FE models were optimized to
achieve reasonable computational time and accuracy through mesh sensitivity analysis. The pavement FE models were refined to
achieve sufficient proximity in pavement responses to the result obtained from linear elastic analysis software (e.g., WinJULEA). In
addition, a reduction of elements for computational efficiency was achieved with a smooth transition to the infinite boundary of FE
pavement models through iterations. The resulting pavement FE models are depicted in Fig. 4 (c) and (d), and it is notable that the size
of AC300-B300 FE model is substantially larger than that of AC125-B150 FE model. In this study, 30 cases captured the effect of the
wheel load and vehicle speed. For each loading case, the tire-contact area was calculated based on measured tire footprints.
The temperature profiles for both pavement structures were derived from the analysis of a two-layer system comprising the AC and
granular layers (Wang et al., 2009; Hernandez et al., 2016). The average pavement-surface temperature in this study was assumed to
be 21 ◦ C (70◦ F). Fig. 5 illustrates the resulting temperature profiles for both pavement structures.
Vehicle speed is related to loading frequency which affects the AC dynamic modulus. The typical highway truck speed limit in
Illinois is 105 km/h (65 mph). Thus, this study adopts 105 km/h as the truck speed. The correlation between wheel load and pavement
responses (e.g., longitudinal tensile strain at the bottom of AC and vertical compressive strain on top of subgrade) is graphically

Table 1
Asphalt concrete dynamic modulus coefficients (after Hernandez et al.’s (2016)).
Pavement Layer δ α β γ α1 α2 α3
AC300-B300 Wearing Surface (WS) 1.9654 2.5711 − 1.5622 0.4982 0.0002 − 0.1053 6.3104
Intermediate Layer (IL) 1.7370 2.8245 − 1.2149 0.4656 0.0002 − 0.1053 6.3104
Base Layer (BL) 1.6067 2.8820 − 1.2840 0.4812 0.0004 − 0.1533 8.6010

AC125-B150 Wearing Surface (WS) 1.0967 3.5621 − 1.0530 0.3572 0.0002 − 0.1169 6.9827
Intermediate Layer (IL) 0.9694 3.6050 − 1.0560 0.3825 0.0001 − 0.0962 6.1900
Base Layer (BL) 0.3505 4.3576 − 0.8800 0.2829 0.0009 − 0.2285 11.6836

5
S. Kang et al. Transportation Research Part D 94 (2021) 102811

Fig. 4. (a) Thin (AC125-B15) and (b) thick (AC300-B300) pavement cross sections, and (c) thin and (d) thick pavement FE models.

Table 2
Regression coefficients for resilient modulus in vertical, horizontal, and shear directions (after Hernandez et al.’s (2016)).
Direction Vertical Horizontal Shear

Coefficient k1 k2 k3 k4 k5 k6 k7 k8 k9

Quantity 869.6 0.9785 − 0.5673 596.6 1.1419 − 1.3464 389.1 0.9083 − 0.2409

Fig. 5. AC temperature profiles for the AC125-B150 and AC300-B300.

6
S. Kang et al. Transportation Research Part D 94 (2021) 102811

illustrated in Fig. 6 for the AC300-B300 and AC125-B150 pavement structures, respectively. It is important to note that the correlation
presented in Fig. 6 is used to predict pavement critical responses corresponding to an arbitrary wheel load. Positive and negative signs
indicate tensile and compressive strains, respectively in Fig. 6. Notably, a near-linear correlation exists between the wheel load and
pavement-critical responses despite the viscoelastic nature of the AC materials. This is because the FE simulations adopt the same
loading frequency (vehicle speed) and temperature profiles. Herein, only the magnitude of the wheel load changes. In order to provide
validation, these critical responses were then compared with the results from WinJULEA, a multi-linear elastic theory (MLET)
pavement analysis software adopted by MEPDG. In most cases, the FEM results are similar to or slightly greater than that of the
WinJULEA results due to AC viscoelastic and dynamic analysis, as illustrated in Fig. 6.

2.4. Pavement performance prediction

This study adopts the MEPDG’s approach to predict pavement performance over its service life. In MEPDG, AC is implicitly
considered linear viscoelastic material through using complex modulus computation based on the frequency of loading. The load
frequency is determined based on a procedure that is a function of pavement geometry, temperature, tire radius, and axle load. In
addition, the AC layer is divided into several sublayers having different modulus values. Base and subgrade materials are assumed to
exhibit linear elastic behavior in MEPDG. To capture the temperature and moisture gradient within pavement, MEPDG has a sub-
layering procedure that further divides pavement layers into multiple thinner layers (i.e., sublayer). In the mechanistic part of
MEDPG, the pavement responses were computed given pavement geometry and tire loading. For response computation, this study uses
3-D FE models developed in ABAQUS. Later, these computed responses are linked to the pavement damage through transfer functions
(i.e., empirical part of MEPDG). The transfer function for IRI prediction is provided in Eq. (10).
IRI = IRI0 + K1 (FCt ) + K2 (RDt ) + K3 (TC) + K4 (SF) (10)

here, IRI0 is an initial IRI value. K1 through K4 are calibration factors equal to 40, 0.4, 0.008, and 0.015, respectively. FCt is the total
fatigue cracking obtained from summing top-down and bottom-up fatigue cracking. RDt represents the total rutting depth, while TC is
the length of thermal cracking and SF is a site factor (as shown in Eq. (11)). As seen, IRI is a function of different types of pavement

Fig. 6. Correlation between wheel load and longitudinal tensile strain at the boom of AC for (a) AC125-B150 and (b) AC300-B300, and correlation
between wheel load and vertical compressive strain at the top of subgrade for (c) AC125-B150 and (d) AC300-B300.

7
S. Kang et al. Transportation Research Part D 94 (2021) 102811

distresses, namely rutting, bottom-up and top-down fatigue cracking, and thermal cracking. The progression of each pavement distress
over time has its own procedure (ARA, 2004b; Gungor and Al-Qadi, 2020). This study uses the TC results from the MEPDG software for
the sake of the completeness of the model shown in Eq. (10).
[ ( )] [ ( )]
SF = Year1.5 ln (Precip + 1) + (FI + 1)p02 + ln (Precip + 1) + (PI + 1)p200 (11)

In Eq. (11), Year represents the pavement age, measured by years. The Precip is an annual average precipitation, which is measured
in inches. FI is an annual average freezing index, which is measured in Fahrenheit. As for p02 , it is the percentage passing the 0.02 mm
sieve, while p200 is the percentage passing the 200 sieve (0.075 mm).
The service lives of the AC300-B300 and AC125-B150 sections can be determined if any one of the thresholds for the IRI, RD, FC,
and TC is met. In this study, the thresholds for IRI, RD, FC, and TC are 2.7 m/km, 12.5 mm, 7.5 percent lane area, and 189.4 m/km,
respectively. The resulting service life is discussed in detail in the following section.

2.5. Pavement life cycle assessment and life cycle costing

DWL caused by road roughness may result in additional pavement damage. Given the randomness of pavement roughness, DWL
may result in a probabilistic distribution of a pavement service life. The primary goal of this section is to quantify DWL’s impact on the
GHG emissions and energy consumption during a pavement’s design life. Pavement life cycle stages considered in this study include
the material, construction, use, and maintenance stages (discussed in the following sub-sections). A case study is presented to examine
the impact of DWL on the LCA and LCC results based on the two highway pavement structures, which are illustrated in Fig. 4. The case
study follows the guidelines by the International Standardization Organization (ISO14044:2006) and the FHWA (Van Dam et al.,
2015). Environmental Protection Agency (EPA)’s Tools for the Reduction and Assessment of Chemical and Other Environmental
Impacts (TRACI) 2.1 was employed for impact characterizations in this study (Bare, 2012). The study also adopted a discount rate of
3% to account for the change in the monetary value with respect to time.

2.5.1. Material and construction stages

The material and construction stages include the production of raw pavement materials, the operation of AC mixing plants, the
transportation of raw materials — from the origin to plants — and AC — from plants to the construction site, mobilization, and the
operation of construction equipment. This study adopts the previously developed regional life cycle impact assessment (LCIA) and cost
databases and models which were based on regionally collected primary data from material producers and contractors in Illinois, and
commercial life cycle inventory (LCI) database – US Ecoinvent (Earthshift, 2013). The details of database and model development are
discussed elsewhere (Kang et al., 2014; Yang, 2014; Ozer et al., 2017). This study assumes that rehabilitation takes place at nighttime;
hence, the impact of traffic delay caused by work zones is not considered.

2.5.2. Maintenance stage

After the initial construction of pavement, regular maintenance should be applied to maintain the adequate level of pavement
serviceability. In this study, pavement service life refers to the number of years that a pavement remains in service before any
rehabilitation treatment. For simplicity, it is assumed that there is only one type of rehabilitation, a 2.5-in AC mill and inlay, at the end
of a pavement’s service life. As discussed in Section 2.4, the pavement’s service life is determined by the thresholds of IRI, RD, FC, and
TC. In other words, pavement rehabilitation is triggered when any one of these thresholds is met. The design life of pavement is
assumed to be 45 years for both AC300-B300 and AC125-B150 pavement structures (IDOT, 2013).

2.5.3. Use stage

Some studies examined rolling resistance (RR) due to deflection in addition to RR caused by roughness (Giustozzi et al., 2018; Xu
et al. 2019; Kim et al., 2019) but this study only considers roughness-induced RR in the use stage of LCA. The IRI progression of the
AC300-B300 and AC125-B150 pavement sections is obtained from Eq. (10). The excess fuel consumption due to pavement roughness is
captured using the half-car and truck models (Kim et al., 2017; Kang et al., 2019). This study adopted the roughness-induced vehicle

Table 3
Background information for pavement structures and traffic.
Pavement Structure AC125-B150 AC300-B300

Wearing surface in mm (in) 25 (1.0) 52.5 (2.1)

Intermediate layer in mm (in) 37.5 (1.5) 72.5 (2.9)
AC Base layer in mm (in) 62.5 (2.5) 175 (6.9)
Granular aggregate in mm (in) 150 (5.9) 300 (11.8)
Granular aggregate stiffness in kPa/mm (psi/in) 113 (415) 113 (415)
Subgrade soil stiffness in kPa/mm (psi/in) 19 (70) 19 (70)
Annual average daily truck traffic (AADTT) 4500 14,000
Annual average daily traffic (AADT) 23,900 59,600
Annual traffic growth (%) 1.0 1.0
Pavement length in km (mi) 1.0 (0.62) 1.0 (0.62)

8
S. Kang et al. Transportation Research Part D 94 (2021) 102811

operating cost model by Chatti and Zaabar (2012) and the fuel cost was assumed to be $0.79 per liter ($3.00 per gal). Table 3 presents
background information on pavement structures and traffic.
The AC300-B300 mimics the actual pavement structure of the Chicago area’s I-90 highway corridor (Jane Addams Memorial
Tollway). As of 2017, the annual average daily truck traffic (AADT) of this corridor is 59,600, and the annual average daily truck traffic
(AADTT) is 14,000 (IDOT, 2019). It is assumed that the AC125-B150 section is exposed to less AADTT as compared to AC300-B300, as
shown in Table 3. In addition to the assumption of the AADT, other assumptions made on truck traffic include the following:

• All axles are equipped with dual-tire assembly (DTA).

• DWL from each truck axle is applied on pavement separately.
• All trucks are Class 9.
• Truncated normal distribution is applied to the DWL to eliminate the effect of a negative wheel load.

The application of truncated normal distribution resulted in the removal of unrealistically long pavement service lives as seen in
Fig. 7. The pavement length is assumed to be 1.0 km. Based on the sensitivity analysis, a total of 7000 different DWLs were found to be
optimal to represent the adequate normality of the DWL while minimizing the computational cost. It can be assumed that there are
7000 one-km-long pavement sections, and that each section is exposed to the repetition of a particular DWL from the distribution to
fail. The performance of 7000 pavement sections is then analyzed to derive the service life distribution.
The service lives resulting from the applied DWL on the pavement are illustrated in Fig. 7. After fitting with lognormal distribution,
the mean and standard deviation of the service life distribution were calculated as presented in Table 4.
The impact of roughness-induced RR can be captured based on the maintenance schedule and IRI progression. The following
equation can be used to estimate the excess fuel consumption (EFC) caused by road roughness based on the above-mentioned vehicle
dynamics models for the passenger vehicle (PV) and truck (Kim et al. 2016; Kim et al. 2017; Kang et al., 2019).

EFCPV (MJ/km) = a1 IRI 2t + a2 IRIt + a3 (for PV)

EFCTruck (MJ/km) = b1 IRI 2t + b2 IRIt + b3 (for trucks) (12)

EFCPV and EFCTruck indicate the excess fuel consumption due to roughness for PV and trucks, respectively, at 105 km/h, which is
represented in the unit of MJ per km. a1 through a3 and b1 through b3 are regression coefficients presented in Table 5. IRIt is the value
of IRI in m/km at an arbitrary time, t.

3. Results and discussion

Once the service life of pavement due to DWL is determined, the environmental and economic impacts from material, construction,
maintenance, and use stages can be calculated using the aforementioned LCIA, cost databases, and the presented models. A case study
is performed for the two pavement sections, AC125-B150 and AC300-B300 (see Fig. 4). Based on the background information provided
in Table 3, the life cycle GHG emissions, energy consumption, and cost are evaluated.

3.1. Probabilistic LCA results

The pavement condition and rehabilitation frequency vary with the service life. This study presented probabilistic LCA results

Fig. 7. Distribution of service life for the AC125-B150 (left) and AC300-B300 (right).

9
S. Kang et al. Transportation Research Part D 94 (2021) 102811

Table 4
Mean and Standard Deviation of Pavement Service Lives.
Mean (Year) Standard Deviation (Year)

AC125-B150 10.2 2.9

AC300-B300 11.9 3.3

Table 5
Regression coefficients for PV and truck EFC caused by IRI.
Vehicle Type PV Truck

Coefficient Symbol a1 a2 a3 R2 b1 b2 b3 R2

Regression Coefficients IRI (m/km) 1.67 × 10− 2

2.02 × 10− 3
− 1.79 × 10− 2 0.99 2.73 × 10− 2
1.41 × 10− 1
− 1.66 × 10− 1 0.99

because of the variation in pavement service life due to DWL. As illustrated in Fig. 8, the total GHG emissions and energy consumption
follow lognormal distribution. The mean and standard deviation of GHG emissions and energy consumption per function unit for both
pavement sections are summarized in Table 6. The functional unit is one-lane-km in this study.
As indicated in Table 6 and Fig. 8, the mean values of GHG emissions and energy consumption for AC300-B300 are nearly two folds
that of AC125-B150. AC300-B300 section is significantly thicker than the AC125-B150 section; hence, a greater amount of materials
and fuels is consumed during construction and rehabilitation. The standard deviations of GHG emissions and energy consumption for
AC300-B300 are three folds greater than AC125-B150. The use stage has contributed significantly to that. The impact from the use
stage is significantly greater for AC300-B300, as its traffic volume, both PV and trucks, is more than twice that of the AC125-B150
section. The extreme high values of the GHG emissions and energy consumption are also attributed to the roughness-induced RR in

Fig. 8. Probabilistic life-cycle GHG emissions for (a) AC125-B150 and (b) AC300-B300, and energy consumption for (c) AC125-B150 and (d)
AC300-B300.

10
S. Kang et al. Transportation Research Part D 94 (2021) 102811

Table 6
Mean and standard deviation of the total greenhouse gas (GHG) emissions and energy consumption.
GHG Emissions (kg CO2 e) Energy Consumption (MJ)

Mean Standard Deviation Mean Standard Deviation

AC125-B150 1.37 × 106 9.40 × 104 2.11 × 107 1.47 × 106

AC300-B300 2.85 × 106 3.35 × 105 4.16 × 107 4.91 × 106

the use stage. The DWL can cause additional pavement damage, which triggers pavement rehabilitation treatments. The frequent
rehabilitation treatments keep the overall pavement roughness level low. This significantly reduces the roughness-induced RR in the
use stage.

3.2. Impact of pavement service life on life cycle stages

The results presented in Table 6 and Fig. 8 depict the total GHG emissions and energy consumption; but the breakdown of envi­
ronmental impacts among different life cycle stages is unclear. As indicated earlier, frequent rehabilitation consumes more materials
and fuels while maintaining the good general pavement condition. This results in the reduction of the use stage impacts. Depending on
DWL magnitude, the pavement service life varies significantly. It is well documented that a pavement section prematurely fails if
exposed to repeated extremely heavy wheel loads that exceeds the design. On the other hand, a pavement section may experience a
slower distress rate than expected if exposed to reduced wheel loads than the design ones. Therefore, sensitivity analyses were

Fig. 9. Percentage of the GHG emissions’ contribution to life cycle stages at different service lives for AC125-B300 (top) and AC300-B300 (bottom).

11
S. Kang et al. Transportation Research Part D 94 (2021) 102811

conducted to assess the contribution of each life cycle stage.

As illustrated in Fig. 9, the use stage contribution is generally greater for AC300-B300 than AC125-B150. For example, if the
pavement service life is 4 years, the contribution from the use stage is 36.1% and 54.8% for AC125-B150 and AC300-B300, respec­
tively. When the pavement service life of 12 years, the use stage contribution increases to 77.6% and 86.0% for AC125-B150 and
AC300-B300, respectively. Thus, regardless of the pavement structure, the use stage contribution increases with pavement service life.
At a low pavement service life — due to an extremely high DWL — frequent rehabilitation treatments are applied, and the overall
roughness condition is kept low. This results in high contribution in the maintenance stage; but low contribution in the use stage.
Conversely, at the tail end of a long pavement service life, less frequent rehabilitation and high levels of pavement roughness can
significantly increase the impacts of the use stage.

3.3. Economic and environmental impact optimization

As indicated earlier, a shorter pavement service life may result in a lower use stage environmental impact. Due to the dominant
contribution of the use stage as illustrated in Fig. 9, this may, in turn, lead to a reduction in the overall life cycle environmental impact.
However, the increased agency cost due to frequent rehabilitation may lead to an increase in the life cycle cost. A sustainable solution
should consider both environmental and economic impacts. For a set of non-dominated solutions, a Pareto frontier analysis may be
conducted (Wan, 1975; Censor, 1977; Ziyadi et al., 2018). In Fig. 10, the horizontal and vertical axes are the total cost and GHG
emissions, respectively. The blue crosses indicate potential scenarios corresponding to different pavement service lives. The orange
line represents Pareto front and the scenarios in the vicinity of Pareto front may be considered as possible optimal solutions. Based on
the application of an elitist non-dominated sorting genetic algorithm (Deb et al., 2000), a total of six non-dominated solutions were
identified as optimal solutions for each pavement section. Instead of comparing all six optimal solutions, two optimal solutions that
best balance between economic and environmental impacts were chosen for comparison as illustrated in Fig. 10. The first optimal
solution (OS1 ) emphasizes the economic impact, while the second (OS2 ) emphasizes the environmental impact. Pavement service lives,
GHG emissions, and life cycle costs for these two optimal solutions are compared in Table 7.
In Table 7, OS1 is taken as a baseline and the corresponding percent changes in service life, GHG emissions, and cost of the OS2 are
presented for comparison. If a transportation agency employs a periodic rehabilitation every 13 years (in lieu of 19 years), it may
increase the life cycle cost by 1.41% but reduce GHG emissions by 12.1% for AC125-B150. For AC300-B300, maintaining a pavement
service life of 7 years (in lieu of 13 years) may increase life cycle cost by 0.72% but reduce GHG emissions significantly by 20.2%. The
results suggest that a systematic pavement design and maintenance planning may benefit transportation agencies and optimize the
pavement life cycle economic and environmental impacts.

4. Summary and conclusions

Current pavement performance prediction models resulted from vehicle-pavement interaction often oversimplify pavement ma­
terial properties and tire contact stresses, lack correlation to critical pavement responses, overlook dynamic wheel loads, or neglect the
connection with roadway infrastructure sustainability.
This study examined the impact of dynamic wheel load (DWL) on pavement performance and the environment. Based on DWL
obtained from the mechanistic tractor-trailer model, non-uniform three-dimensional (3-D) tire contact stresses were applied on 3-D
viscoelastic finite element (FE) pavement models. A total of 30 FE simulations were performed to estimate the pavement’s critical
responses to various magnitudes of wheel load for two flexible pavement sections: AC125-B150 and AC300-B300. Based on the results
of FE simulations, a correlation between the pavement responses and the DWL was established. Using the mechanistic-empirical
pavement design guide (MEPDG) approach, a distribution of the pavement’s service life was obtained based on pavement damage
accumulation and the International Roughness Index (IRI) progression. In addition, this study presented a life cycle assessment (LCA)
and life cycle costing (LCC) case study to examine the DWL impact on the environment and economy in three dimensions: global
warming potential (GWP) (quantified by greenhouse gas [GHG] emissions), energy consumption, and cost. The following highlights
the main findings of the study:

• Based on the finite element analysis (FEA), the relationship between wheel load and pavement critical responses is nearly linear.
• DWL caused by road roughness generated from a zero-mean stationary white noise exhibits normal distribution.
• The service life of pavement is determined based on the corresponding thresholds of any of the following: IRI, rut depth (RD),
fatigue cracking (FC), and thermal cracking (TC).
• The pavement service life and LCA results may be demonstrated with lognormal distribution.
• Pareto frontier analysis may be used to identify an optimum solution considering economic and environmental impacts.

The case study considered the material, construction, maintenance, and use stages with a functional unit of one-lane-km. Based on
the maintenance schedule, determined by the pavement’s service life resulting from the DWL, probabilistic LCA results were identified.
A higher variation was observed for thick pavement section due to the use stage impact from the roughness-induced rolling resistance
(RR). The LCA results suggest that having high DWLs may cause an earlier pavement to fail and trigger pavement rehabilitation.
Frequent maintenance/ rehabilitation, however, may keep the pavement roughness level low, which would significantly reduce the
use stage environmental impact. However, the increased agency cost for frequent rehabilitation may lead to an increase in the life cycle
cost, but a reduction in GHG emissions and energy consumption. This implies that a systematic pavement design and maintenance

12
S. Kang et al. Transportation Research Part D 94 (2021) 102811

Fig. 10. Pareto frontier analysis for optimal policies for (a) AC125-B150 and (b) AC300-B300 sections.

Table 7
Pavement Service Lives and Impacts of Optimal Solutions.
AC125-B150 AC300-B300

Optimal 1 or OS1 (baseline) Optimal 2 orOS2 Optimal 1 orOS1 (baseline) Optimal 2 orOS2

Service Life (Year) 19 13 13 7

GHG Emissions (1000 kg CO2e) 1590 1390 3070 2450
Present Value ($1000) 7080 7180 27,900 28,100

Percent Change between OS1 and OS2 (%)

Service Life – − 31.6 – − 46.2
GHG Emissions – − 12.6 – − 20.2
Present Value – +1.41 – +0.72

planning may assist transportation agencies to optimize the pavement economic and environmental impacts during its life cycle.
For a future study, the following topics may be further researched:

• Impact of nonlinearity, resulting from pavement layer bonding conditions and granular materials, on critical responses and service
lives
• Effect of seasonal temperature variations on material properties and critical responses
• Effect of truck rolling motion resulted from different left and right wheel paths and crosswind using a 3-D truck model

Acknowledgments

The authors are representatives of Illinois Center for Transportation (ICT). Valuable inputs from Hasan Ozer, associate professor at
Arizona State University; Billie F. Spencer, professor at the University of Illinois at Urbana-Champaign; and Robin E. Kim, assistant
professor at Hanyang University (Seoul, South Korea) are greatly appreciated. The contents of this paper reflect the view of the authors,
who are responsible for the facts and accuracy of the data presented herein. The contents do not necessarily reflect the official view or
policies of ICT. This paper does not constitute a standard, specification, or regulations. This research did not receive any specific grant
from funding agencies in the public, commercial, or not-for-profit sectors.

Author contribution statement

All authors contributed to the paper equally.

References

Al-Qadi, I.L., Bhutta, S.A., 1999. In situ measurements of secondary road flexible pavement response to vehicular loading. Transp. Res. Rec. 2 (1652), 206–216.
Al-Qadi, I.L., Dessouky, S., Tutumluer, E., Kwon, J., 2011. Geogrid mechanism in low-volume flexible pavements: accelerated testing of full-scale heavily
instrumented pavement sections. Int. J. Pavement Eng. 12 (2), 121–135. https://doi.org/10.1080/10298436.2010.535534.
Al-Qadi, I.L., Elseifi, M.A., 2007. New generation of wide-base tires: Impact on trucking operations, environment, and pavements. https://doi.org/10.3141/2008-13.
Al-Qadi, I.L., Hassan, M.M., Elseifi, M.A., 2005. Field and theoretical evaluation of thermal fatigue cracking in flexible pavements. https://doi.org/10.3141/1919-10.

13
S. Kang et al. Transportation Research Part D 94 (2021) 102811

Al-Qadi, I.L., Hernandez, J.A., Gamez, A., Ziyadi, M., Gungor, O.E., Kang, S., 2018. Impact of wide-base tires on pavements: a national study. https://doi.org/10.
1177/0361198118757969.
Al-Qadi, I.L., Loulizi, A., Elseifi, M.A., Lahouar, S., 2004. The Virginia Smart Road: the impact of pavement instrumentation on understanding pavement performance.
J. Assoc. Asphalt Paving Technol. 73 (2004), 427–466.
Al-Qadi, I.L., Wang, H., 2012. Impact of wide-base tires on pavements: results from instrumentation measurements and modeling analysis. Transport. Res. Rec.: J.
Transport. Res. Board 2304, 169–176.
ARA, Inc., 2004a. Guide for Mechanistic-Empirical Design. Chapter 2. Material Characterization. Champaign, IL. Retrieved from http://onlinepubs.trb.org/
onlinepubs/archive/mepdg/Part2_Chapter2_Materials.pdf.
ARA, Inc., 2004b. Guide for Mechanistic-Empirical Design. Chapter 3. Design of New and Reconstructed Flexible Pavements. Champaign, IL. Retrieved from http://
onlinepubs.trb.org/onlinepubs/archive/mepdg/Part3_Chapter3_FlexibleDesign.pdf.
Bare, J., 2012. Tool for the Reduction and Assessment of Chemical and Other Environmental Impacts (TRACI, Version 2.1). [Software]. United States Environmental
Protection Agency, Cincinnati, OH.
Censor, Y., 1977. Pareto optimality in multiobjective problems. Appl. Math. Optim. 4, 41–59.
Chatti, K., Zaabar, I., 2012. Estimating the Effects of Pavement Condition on Vehicle Operating Cost (Report 720). National Cooperative Highway Research Program.
Washington, D.C.: National Academy of Sciences.
Deb, K., Agrawal, S., Pratap, A., Meyarivan, T., 2000. A Fast Elitist Non-Dominated Sorting Genetic Algorithm for Multi-objective Optimization: NSGA-II. Indian
Institute of Technology Kanpur, Kanpur, India.
Earthshift, 2013. US-Ecoinvent Database. Version 2.2. [Database]. Swiss Center for Life-cycle Inventories, St-Gallen, Switzerland.
Federal Highway Administration, 2016. Measuring and Specifying Pavement Smoothness (FHWA-HIF-16-032). Pavement Design and Analysis Team. Washington, D.
C. Retrieved from https://www.fhwa.dot.gov/pavement/pubs/hif16032.pdf (accessed Jun. 9, 2019).
Gillespie, T.D., Karamihas, S.M., Sayers, M.W., Nasim, M.A., Hansen, W., Ehsan, N., 1993. Effects of heavy-vehicle characteristics on pavement response and
performance. NCHRP Report 353. Transportation Research Board of the National Academies, Washington, D.C.
Giustozzi, F., Ponzoni, F., Louhghalam, A., Kirchain, R., Gregory, J., 2018. Sensitivity analysis of a deflection-induced pavement-vehicle interaction model. Road
Mater. Pavement Des. 20 (8), 1880–1898. https://doi.org/10.1080/14680629.2018.1479288.
Gungor, O.E., Al-Qadi, I.L., Gamez, A., Hernandez, J.A., 2016. In-situ validation of three-dimensional pavement finite element models. In: In the Roles of Accelerated
Pavement Testing in Pavement Sustainability. Springer, pp. 145–159.
Gungor, O.E., Al-Qadi, I.L., Gamez, A., Hernandez, J., 2017. Development of adjustment factors for mepdg pavement responses utilizing finite-element analysis.
J. Transport. Eng., Part A: Syst. 143 (7), 04017022.
Gungor, O.E., Al-Qadi, I.L., 2020. Wander 2D: A flexible pavement design framework for autonomous and connected trucks. Int. J. Pavement Eng. https://doi.org/
10.1080/10298436.2020.1735636.
Guo, G., Ding, W., Zhang, C., 2012. Analysis of stochastic dynamic load acting on rough road by heavy-duty traffic. In: Proceedings of the 12th COTA International
Conference of Transportation Professionals. Beijing, China, pp. 3194–3205.
Hernandez, J., Al-Qadi, I.L., 2017. Semicoupled modeling of interaction between deformable tires and pavements. J. Transport. Eng.—Part A: Syst. 143 (4),
04016015.
Hernandez, J.A., Al-Qadi, I.L., De Beer, M., 2013. Impact of tire loading and tire pressure on measured 3D contact stresses. In: Paper presented at the Airfield and
Highway Pavement Conference: Sustainable and Efficient Pavements — Proceedings of the 2013 Airfield and Highway Pavement Conference, pp. 551–560.
https://doi.org/10.1061/9780784413005.044.
Hernandez, J.A., Gamez, A., Al-Qadi, I.L., 2016. Effect of wide-base tires on nationwide flexible pavement systems: numerical modeling. Transp. Res. Rec. 2590,
104–112. https://doi.org/10.3141/2590-12.
Huhtala, M., Pihlajamaki, J., Pienimaki, M., 1989. Effect of tires and tire pressures on road pavements. Transp. Res. Rec. 1227, 107–114.
Illinois Department of Transportation, 2013. Chapter Fifty-four Pavement Design. Bureau of Design and Environment Manual. Accessed Jan. 26, 2019.
Illinois Department of Transportation, 2019. Getting around Illinois. Retrieved from https://www.gettingaroundillinois.com/default.htm (accessed June 27, 2019).
International Organization for Standardization, 1995. Mechanical vibration – Road surface profiles – Reporting of measured data. ISO 8608:1995(E). Geneva,
Switzerland.
International Organization for Standardization, 2006. Environmental management life cycle assessment requirements and guidelines. ISO 14044:2006(E). Geneva,
Switzerland.
Kang, S., Ozer, H., Al-Qadi, I.L., Spencer, B.F., 2019. Stochastic analysis of rolling resistance energy dissipation for a tractor-trailer model. https://doi.org/10.1177/
0361198119840344.
Kang, S., Yang, R., Ozer, H., Al-Qadi, I.L., 2014. Life-cycle greenhouse gases and energy consumption for material and construction phases of pavement with traffic
delay. Transp. Res. Rec. 2428, 27–34.
Kim, R.E., Kang, S., Spencer, B.F., Al-Qadi, I.L., Ozer, H., 2016. New stochastic approach of vehicle energy dissipation on nondeformable rough pavements. J. Eng.
Mech. 143 (4) https://doi.org/10.1061/(ASCE)EM.1943-7889.0001192.
Kim, R.E., Kang, S., Spencer, B.F., Ozer, H., Al-Qadi, I.L., 2017. Stochastic analysis of energy dissipation of a half-car model on nondeformable rough pavement.
J. Transport. Eng. Part B: Pavements 143 (4). https://doi.org/10.1061/JPEODX.0000014.
Kim, R.E., Kang, S., Spencer, B.F., Al-Qadi, I.L., Ozer, H., 2019. Impact of pavement roughness and deflection on fuel consumption using energy dissipation. J. Eng.
Mech. 145 (10) https://doi.org/10.1061/(ASCE)EM.1943-7889.0001653.
Lieh, J., Qi, W., 1995. Simulation of dynamic truck loading on pavements using measured road roughness. Transp. Res. Rec. 1501, 13–21.
Misaghi, S., Tirado, C., Nazarian, S., Carrasco, C., 2021. Impact of pavement roughness and suspension systems on vehicle dynamic loads on flexible pavements.
Transport. Eng. 3 https://doi.org/10.1016/j.treng.2021.100045.
Oh, J., Fernando, E.G., Lytton, R.L., 2007. Evaluation of damage potential for pavements due to overweight truck traffic. J. Transp. Eng. 133 (5), 308–317. https://
doi.org/10.1061/(ASCE)0733-947X(2007)133:5(308).
Ozer, H., Yang, R., Al-Qadi, I.L., 2017. Quantifying sustainable strategies for the construction of highway pavements in Illinois. Transport Res. D-TR E 51, 1–13.
https://doi.org/10.1016/j.trd.2016.12.005.
Schwartz, C.W., Li, R., Ceylan, H., Kim, S., Gopalakrishnan, K., 2013. Global sensitivity analysis of mechanistic-empirical performance predictions for flexible
pavements. Transp. Res. Rec. 2368, 12–23. https://doi.org/10.3141/2368-02.
Sweatman, P.F., 1983. A Study of Dynamic Wheel Forces in Axle Group Suspensions of Heavy Vehicles, Australian Road Research Board, Special Report No. 27.
Van Cuong, B., Van Quynh, L., Long, L.X., 2019. Influence of heavy truck operating condition on dynamic load coefficient. https://doi.org/10.1007/978-3-030-
04792-4_49.
Van Dam, T., Harvey, J.T., Muench, S.T., Smith, K.D., Snyder, M.B., Al-Qadi, I.L., Ozer, H., Meijer, J., Ram, P.V., Roesler, J.R., Kendall, A., 2015. Towards Sustainable
Pavement Systems: A Reference Document. FHWA-HIF-15-002. Federal Highway Administration, Washington, DC.
Wan, Y.-H., 1975. On local Pareto optima. J. Math. Econ. 2 (1), 32–42.
Wang, H., Al-Qadi, I.L., 2009. Combined effect of moving wheel loading and three-dimensional contact stresses on perpetual pavement responses. https://doi.org/10.
3141/2095-06.
Wang, D., Roesler, J., Guo, D.Z., 2009. Analytical approach to predicting temperature fields in multilayered pavement systems. J. Eng. Mech. 135 (4), 334–344.
Xia, R., Li, J., He, J., Shi, D., 2015. Effect analysis of vehicle system parameters on dynamic response of pavement. Math. Probl. Eng. https://doi.org/10.1155/2015/
561478.
Xu, X., Akbarian, M., Gregory, J., Kirchain, R., 2019. Role of the use phase and pavement-vehicle interaction in comparative pavement life cycle assessment as a
function of context. J. Cleaner Prod. 230, 1156–1164. https://doi.org/10.1016/j.jclepro.2019.05.009.

14
S. Kang et al. Transportation Research Part D 94 (2021) 102811

Yang, R., 2014. Development of a Pavement Life-cycle Assessment Tool Utilizing Regional Data and Introducing an Asphalt Binder Model. MS Thesis. University of
Illinois at Urbana-Champaign, Urbana, IL.
Yoo, P.J., Al-Qadi, I.L., 2007. Effect of transient dynamic loading on flexible pavements. Transp. Res. Rec. 1990 (1), 129–140.
Ziyadi, M., Ozer, H., Kang, S., Al-Qadi, I.L., 2018. Vehicle energy consumption and an environmental impact calculation model for the transportation infrastructure
systems. J. Cleaner Prod. 174, 424–436. https://doi.org/10.1016/j.jclepro.2017.10.292.

15