You are on page 1of 19

Finite-Element Simulation of Instrumented Asphalt

Pavement Response under Moving Vehicular Load


Ogoubi Cyriaque Assogba 1; Zhiqi Sun 2; Yiqiu Tan 3; Lushinga Nonde 4; and Zheng Bin 5

Abstract: Understanding the structural behavior of layered asphalt pavements subjected to dynamic moving wheel loads is a crucial require-
ment for the future design of more-durable pavement structures. Therefore, to accurately evaluate the dynamic response of an instrumented
Downloaded from ascelibrary.org by Imperial College London on 01/04/20. Copyright ASCE. For personal use only; all rights reserved.

semi-rigid base asphalt pavement under full-scale moving heavy vehicular loading, this study established a full-scale three-dimensional
viscoelastic finite element model. Sinotruck HOWO A7 6 × 4 dump truck moving wheel loads and associated contact stresses effect on
the pavement structure were implemented in the implicit dynamic analysis. The dynamic properties of the asphalt concrete were incorporated
in Abaqus through a Prony–Dirichlet series. The rationality and validity of the developed model were successfully verified by field data
collected from a Rizhan–LanKao highway pavement test section in Shandong province, China. The pavement dynamic response computed at
different depths under the truck dual tires and at a different lateral position in the mid-depth of asphalt concrete middle course were analyzed.
Results indicated that both tensile and compressive three-directional dynamic strain, stress coexist in the pavement structure. In addition, an
alternating change was found between negative and positive dynamic shear stress and strain in the structure. Finally, this study revealed that
the peak value of the dynamic strain–stress in vertical, transverse, and longitudinal directions, as well as the shearing strain–stress response of
the pavement structure, are affected not only by the lateral distribution of the analysis points but also according to depth. DOI: 10.1061/
(ASCE)GM.1943-5622.0001616. © 2020 American Society of Civil Engineers.
Author keywords: Semi-rigid base asphalt pavement; Dynamic response; Viscoelastic material; Finite-element simulation; Field
measurement.

Introduction materials and the vehicle wheel loading are the most important part
of the pavement structural mechanical response analysis. The cur-
In China, more than 90% of the highway pavements are constructed rent mechanistic analysis and design method for asphalt pavement
with a semi-rigid road base (Sun et al. 2018). Semi-rigid base as- in China is based on the multilayer elastic theory and the static re-
phalt pavement is a complex multilayered system consisting of as- sponse analysis. In this approach, the pavement structure is modeled
phalt mixtures as a surface layer, cement-treated road base and as layered systems with linear elastic behavior and is assumed to be
subbase, cement-stabilized subgrade, and natural soil. Unfortu- subjected to a stationary vehicular loading. This includes the pave-
nately, in recent years, severe pavement distress and premature fail- ment dynamic response predictions or analysis; the asphalt concrete
ure have occurred in semi-rigid base asphalt pavement. The (AC) mixture is characterized as a linear elastic material without any
mechanical response analysis of a multilayered pavement system other energy dissipation source when computing the stress–strain
is the appropriate method to explain the pavement distresses and response induced by the moving traffic load on the asphalt pavement
its failure mechanism. The constitutive models of pavement system structure. In addition, the existing pavement inspection method can
explain only the degree of damage from the phenomenological point
1
Ph.D. Candidate, School of Transportation Science and Engineering,
of view, and cannot obtain the work information within the asphalt
Harbin Institute of Technology, Harbin 150090, People’s Republic of China pavement structure. It is unable to realize the timely perception of
(corresponding author). ORCID: https://orcid.org/0000-0002-1581-5187. pavement damage, the diagnosis of damage causes, or the mastery
Email: assocyr@hotmail.fr; assocyr@stu.hit.edu.cn of damage evolution law.
2
Ph.D. Candidate, School of Transportation Science and Engineering, Therefore, several types of research in recent years have focused
Harbin Institute of Technology, Harbin 150090, People’s Republic of on the analysis of flexible-pavement dynamic response subjected to
China. Email: 17S132062@hit.edu.cn moving vehicular load. To accurately predict the mechanical re-
3
Professor, School of Transportation Science and Engineering, State Key sponse of the pavement and identifying the causes of its distress
Laboratory of Urban Water Resource and Environment, Harbin Institute of
Technology, Harbin 150090, People’s Republic of China. Email: tanyiqiu@
and failure, some researchers have used various intelligent sensors
hit.edu.cn to monitor the internal information of asphalt pavement structure
4
Ph.D. Candidate, School of Transportation Science and Engineering, (Alqadi et al. 2008; Doyle et al. 2014; Chen et al. 2015; Li et al.
Harbin Institute of Technology, Harbin 150090, People’s Republic of 2015; Lu et al. 2016; Sha et al. 2016), whereas others have used
China. Email: nonde.lushinga@gmail.com numerical simulation methods, mainly finite-element (FE) methods
5
Graduate Student, School of Transportation Science and Engineering, (Howard and Warren 2009; Kim et al. 2009; Sun and Duan 2013;
Harbin Institute of Technology, Harbin 150090, People’s Republic of Assogba et al. 2019). These studies have demonstrated that the as-
China. Email: 14b932006@hit.edu.cn
sumptions made in the actual mechanistic analysis and design
Note. This manuscript was submitted on November 6, 2018; approved
on August 29, 2019; published online on January 3, 2020. Discussion per- method for asphalt concrete pavement are inconsistent with the dy-
iod open until June 3, 2020; separate discussions must be submitted namic properties of the asphalt concrete (viscoelastic behavior) and
for individual papers. This paper is part of the International Journal of the realistic loading conditions. Therefore, they cannot achieve ac-
Geomechanics, © ASCE, ISSN 1532-3641. curate pavement response evaluations and performance predictions.

© ASCE 04020006-1 Int. J. Geomech.

Int. J. Geomech., 2020, 20(3): 04020006


This study used the powerful commercial FE analysis software the viscosity of the surrounding environment, the energy dissipa-
Abaqus (version 2017) to develop a full-scale three-dimensional tion of the soil foundation, and so forth. Because of its complexity,
(3D) FE model to accurately quantify the critical stress–strain re- it is very difficult to accurately determine the damping matrix in FE
sponse of an instrumented asphalt pavement with a semi-rigid base analysis. The calculation methods of damping often use a stress-
under dynamic moving traffic loads. This analysis modeled the in- energy factor method, modal damping method, stiffness factor
strumented road test section of the Rizhan–LanKao G1511 high- method, and Rayleigh damping method. This FE analysis used
way pavement in Shandong province, China. the Rayleigh damping method, which assumes that the damping
The main objective of this study was to develop a 3D nonlinear, matrix is a linear combination of the mass matrix and stiffness-
dynamic FE model that integrates the viscoelastic behavior of the proportional damping. The corresponding equation is
asphalt concrete material to simulate the instrumented highway
pavement test section in order to accurately predict the dynamic C ¼ α½M þ β½K ð4Þ
responses of the semi-rigid base asphalt pavement structure accord-
ing to the analysis point distribution in term of depth and lateral where α = mass damping coefficient; and β = stiffness damping
Downloaded from ascelibrary.org by Imperial College London on 01/04/20. Copyright ASCE. For personal use only; all rights reserved.

position. coefficient.
Using the orthogonality condition, coefficients α and β can be
calculated by the vibration’s modes and the corresponding modal
Pavement Dynamic Analysis by Finite-Element damping ratio. Their distinctive equations are expressed in the fol-
Modeling lowing form:
2 3
Basic Equations and Concepts of Pavement Dynamics   wj −wi  
α wi wj 6 7 ξi
¼2 4 1 1 5 ð5Þ
The wheel loads of a moving vehicle, which are applied to the pave- β w2j
− w2i − ξj
ment system at various moving speeds along the road, are undeni- wj wj
ably dynamic. As a such, the contact stress and position of the wheel
loads are time-dependent and may be considered equal to the sum where wi and wk = vibration frequencies of order i and mode; the
of the static load and a continuously altering load. Various factors corresponding damping ratios ξ i and ξ k of order i and order k
such as pavement surface irregularity, vehicle speed, weight, and mode can be measured by the resonant column test or cyclic tri-
suspension system can affect the constantly changing dynamic load. axial test (Zhong et al. 2002). The damping ratio can be approxi-
Oscillatory movement of the vehicle due to the vehicle suspension mated as a constant within a certain frequency range by assuming
system causes the dynamic wheel loads to vary about their mean that the damping coefficients α and β have the same critical damp-
amplitudes when the vehicle is in motion. The dynamic character- ing ratio ξ, and therefore Eq. (5) can be simplified as follows
istic of the materials considered as isotropic elastic and linear (Bathe 1982):
according to the principle of virtual work applied to a domain Ω    
(Tautou et al. 2017) can be expressed as α 2ξ wi wj
¼ ð6Þ
→ → β wi þ wj 1

div σ þf ¼ ργ ð1Þ
→ The critical ξ damping rate for a given i order can be expressed
where σ = Cauchy stress tensor; f = denote the applied force; as a function of mass proportional damping α and stiffness-

ρ = density; and γ = acceleration vector. proportional damping β as follows:
Using the traditional Hooke’s law, the Cauchy stress factor
can be obtained α βw
ξ¼ þ ð7Þ
2w 2
σ ¼ λ∇U · I þ μð∇U þ ∇U Þ T
ð2Þ
The pavement structure’s smallest natural frequency and the
where U = displacement vector; I = identity matrix; and λ and highest loading frequency frequently are taken as critical frequen-
μ = Lamé’s coefficients. cies. The fraction of critical damping of the pavement structure ex-
According to the standard FE procedure (Bathe 1982; Lee et al. perimentally is between 0.02 and 0.05 (Zhong et al. 2002; Alqadi
1999), the governing equation of a nonconservative dynamic sys- et al. 2008). Therefore, the pavement materials, from the road base
tem including material damping can be represented by Eq. (3). This to natural soil were characterized as elastic with no other source of
mathematical statement can be solved using the implicit or the energy dissipation. The most extreme critical damping rate (5%)
explicit integration method in Abaqus. This study used the implicit was used for the layers.
direct integration method because it is more efficient for the level of
the frequencies observed in pavement structure simulations (Gungor
et al. 2016)
Pavement Structure Finite-Element Modeling
½Mfüg þ ½Cfu̇g þ ½Kfug ¼ FðtÞ ð3Þ
Pavement Model Geometry and Element Mesh
where ½M = mass matrix; ½C = damping matrix; ½K = stiffness
matrix; fu̇g = velocity vector; füg = acceleration related to nodes; Using Abaqus 2017, a full-scale 3D FE model was established
fug = displacement vector; and FðtÞ = external force vector related to simulate the Rizhan–LanKao G1511 highway pavement test
to structure dynamic system. section (mile stake K313 þ 600 − K314 þ 500) in Shandong prov-
ince, China. The real pavement structure system extends horizontally
and vertically into infinity. Modeling such a road system in a 3D FE
Material Damping
model will undoubtedly lead to computational or edge-effect errors.
The damping mechanism is one of the factors that seriously affects To considerably reduce such errors, the model domain should
the materials dynamic response. It is related to the structure itself, be larger. However, a larger model requires smaller elements, and

© ASCE 04020006-2 Int. J. Geomech.

Int. J. Geomech., 2020, 20(3): 04020006


Layer Pavement Structure Thickness and Material
Asphalt surface Course
appropriate and adequate boundary conditions. The boundary of a
35 mm SMA 13
FE model usually is artificial; the wave generated when a vehicle
Asphalt Middle Course 130 mm SBS-AC20
is in motion can propagate in the structure domain until it is re-
Asphalt Bottom Course 20 mm HE-AC5
flected in the artificial boundary and eventually pollutes the wave
Semi-rigid Base 180 mm Cement Treated
Bound material (CTB) propagation within the domain. Therefore, previous studies have
recommended the erection of nonreflective or absorbing bounda-
-
Semi-rigid Subbase 170 mm Cement Treated
Bound material (CTB)
ries (Hatzigeorgiou and Beskos 2010; Li et al. 2015; Chango
et al. 2019). Therefore, an infinite element was selected to eliminate
Cement-stabilized 180 mm Cement stabilized
Soil (CS)
the boundary effect of the pavement model, to absorb the wave en-
Subgrade
ergy and reduce the degree of freedom in the far field. Such ele-
ments are known to create a silent boundary without significant loss
Natural Soil Infinite In-place Soil of precision for dynamic analysis (Li et al. 2015). It is clear that
when the trucks were moving on the pavement structure, there was
Downloaded from ascelibrary.org by Imperial College London on 01/04/20. Copyright ASCE. For personal use only; all rights reserved.

no response recorded at infinity. In addition, and according to the


principle of Saint-Venant (Breuer and Roseman 1977), (1) It was
Fig. 1. Pavement structure sketch, layer thickness, and material. assumed that at the bottom of the pavement model there was no
displacement in the vertical and horizontal direction, thus the bot-
tom of the natural soil was constrained to zero along the x-, y-, and
z-directions (Taherkhani and Jalali 2018); and (2) the four lateral
therefore the computation will take a long time because it requires
sides of the pavement model were restrained from any perpen-
more resources to reach the required accuracy level.
dicular displacement to the side of the model [Fig. 2(b)]. Thus, the
The full-size FE model dimensions ðx; y; zÞ were carefully
x-axis displacements of two longitudinal sides of the FE model
chosen to be 25 × 20 × 70 m, based on the results from in situ
were constrained to be zero, and the y-axis displacement of two
measurement of the instrumented road test section and after at-
transverse sides of the FE model also was constrained to be zero
tempting to use various combinations of length, width, and depth
(Liao and Sargand 2011; Wang et al. 2011; Assogba et al. 2020).
of the model. In addition, existing 3D FE models in the literature
were considered as references (Beskou et al. 2016; Lu et al. 2016).
The thickness and the material of the layers of the pavement struc- Modeling of Moving Truck Loads
ture are illustrated in Fig. 1. The in-plane loading area of each dual
tire wheel path was about 0.508 × 60 m (width × length) with a The controlled-load truck multiaxle wheel loads were modeled as
fine mesh element to balance the cost of the calculation and accu- a transient local dynamic load. The user subroutine interface in
racy [Figs. 2(a and b)] ABAQUS was used to define the nonuniform distributed loads
In FE analysis, the consistency of the mesh size has a certain (DLOAD), which allows users to specify the variation of the dis-
influence on the result of the analysis. The mesh must be made in tributed load magnitude (F) as a function of time (TIME*), in the
such a way that it provides the most accurate results. Several re- specific coordinate (COORDS*) system with the domain element
searchers investigated the size of the mesh that is adequate for a number and applied load integration point number. The dynamic
dynamic model (Gungor et al. 2016; Taherkhani and Jalali 2018). load and its contact stress effect on the asphalt concrete pavement
Yoo and Al-Qadi (2008) suggested that in the loading area, 20 mm was incorporated into the implicit dynamic analysis using a devel-
can be adopted as the mesh length in the traffic direction and be- oped Fortran code (Hibbit, Karlsson, and Sorensen 2007).
tween 15 and 18 mm should be used in the lateral direction. In the
present study, the FE model meshing was performed in a way to Truck Wheel Load Configurations, Tire-Pavement
increase the accuracy of the model. Along the wheel path, a rela- Contact Area, and Associated Stress
tively fine mesh size was used because the stresses and the displace-
ment were high. A dense mesh size was adopted near the loading Wheel load configurations, tire–pavement contact area, and asso-
area, whereas a relatively coarse mesh size was used far away from ciated stress of the controlled-load truck used during the field tests
the loading area [Fig. 2(a)]. To avoid errors and warnings related to are required parameters for modeling the transient dynamic load
the mesh during the analysis, the depth of the 3D mesh in the stone of the vehicle. This allows accurately investigating the dynamic
mastic asphalt (SMA13) and high-elasticity and high-viscosity response of the instrumented semi-rigid base asphalt pavement sub-
(HE-AC5) layers was taken as half their respective thicknesses, jected to full-scale vehicular moving loading. Fig. 3 illustrates the
and that of the styrene-butadiene-styrene (SBS-AC20) layer was axle configuration of the Sinotruck HOWO A7 6 × 4 (China
one-third of its thickness. Similarly, the depth of the 3D mesh in National Heavy-Duty Truck Group, Shandong Province, China)
the base layer, the subbase, and the subgrade was taken equal to controlled-load truck, which included a single steering axle and
one-quarter of their respective thicknesses, and that of the natural a tandem driving axle.
soil was equal to one-eighth of its thickness. A continuum 3D The contact areas between the tire and pavement surface consid-
eight-node reduced integration element (C3D8R) was used in the ered in the FE model were generated from actual measurements
finite-element domain to enhance the convergence rate, whereas a (Table 1) from 11.00=22.5-11 and 11.00=R22.5-13 radial tires
continuum infinite 3D eight-node integration element (CIN3D8R) for the front axle and rear axle, respectively. The shape of each
was selected to define infinite boundaries on each side of the pave- tire footprint measured on the field was similar to an ellipse
ment model [Fig. 2(a)] (Tarefder and Ahmed 2014; Hernandez [Figs. 4(a and b)]. This was converted into a combination of two
et al. 2016; Reynaud et al. 2016). semicircles and a rectangle having a length L and a width 0.6L
[Figs. 4(a and b)] (Portland Cement Association 1966; Yoder and
Witczak 1975). The tire–pavement contact area Ac and the length
Boundary Conditions
were determined using (Hadi and Bodhinayake 2003; Huang 2003)
The erected boundary conditions had some effect on the accuracy
of the results from the FE analysis, so it was essential to adopt an Ac ¼ πð0.3LÞ2 þ ð0.4LÞð0.6LÞ ¼ 0.5227L2 ð8Þ

© ASCE 04020006-3 Int. J. Geomech.

Int. J. Geomech., 2020, 20(3): 04020006


rffiffiffiffiffiffiffiffiffiffiffiffiffiffi
Ac pffiffiffiffiffi In this FE model, the influence of each isolated wheel load
L¼ ¼ 1.3832 Ac ð9Þ
0.5227 under the truck axles was taken into account. However, only
the vertical contact pressure was considered to simulate the dy-
In this study, the tire–pavement contact area with each wheel load namic contact forces generated by the controlled-load truck in
was assumed to be rectangular. The real tire footprint was converted motion. This was due to the complexity of measuring the
into a rectangle with an equivalent area (Fig. 4). The length and the
three-directional stress state of the actual tire–pavement contact
width of the converted rectangle were 0.8712L and 0.6L, respec-
forces, and to the difficulty of expressing the contact forces in
tively. The tire–pavement contact area varied from one wheel tire
to another due to the load distribution (Table 1). For numerical mod- mathematical form to be implemented in the DLOAD subroutine.
eling purposes and to simplify the subroutine source code, the truck The vertical contact pressure of each tire was assumed to be uni-
tires were assumed to have the same tire–pavement contact surface. formly distributed over the contact area, and therefore easily was
The equivalent length and the width of each tire footprint was 28 determined by dividing the wheel load by the tire–pavement
and 20 cm. equivalent contact area.
Downloaded from ascelibrary.org by Imperial College London on 01/04/20. Copyright ASCE. For personal use only; all rights reserved.

U(z)=0
U(x)=0 UR(x)=UR(y)=0
UR(y)=UR(z)=0

20 m

U(x)=0
UR(y)=UR(z)=0

Displacement
U(x)
Direction

Rotation
UR(x)
Direction
U(x)=U(y)=U(z)=0
UR(x)=UR(y)=UR(y)=0

U(z)=0
UR(x)=UR(y)=0

(a)

Fine Mesh Size along Coarse Mesh Size far Dense Size Mesh
the wheel path away from loading area near loading area
5

Infinite Element
5

Infinite Element
Infinite Element
5

15
5

Finite Element

Infinite Element
5

5 60 5
(b)

Fig. 2. Finite-element mesh of the pavement structure: (a) 3D view of the selected domain (including finite and infinite elements) and boundary
conditions; and (b) in-plane geometry and mesh size.

© ASCE 04020006-4 Int. J. Geomech.

Int. J. Geomech., 2020, 20(3): 04020006


3825 1350

1830
2022

350
Downloaded from ascelibrary.org by Imperial College London on 01/04/20. Copyright ASCE. For personal use only; all rights reserved.

Front Drive Rear Drive


Steering axle
Tandem Axle Tandem Axle

Front Axle
Rear Axles

Fig. 3. Controlled-load truck configuration (unit: millimeter).

0.87 L
Table 1. Controlled truck axle weight and measured tire-contact area
Second- Third-
First-axle axle axle
Parameter wheel wheel wheel
Wheel configuration Single Dual
0.6L
Axle weight (kN) 71 187
Measured tire contact surface, Ac (cm2 ) 556.618 545.920 534.695
Length of actual area, L (cm) 32.632 32.317 31.983
Length of equivalent area, 0.87L (cm) 28.429 28.155 27.864
Diameter of semicircle or width of 19.579 19.390 19.190
equivalent area, 0.6L (cm) L
Average length of equivalent area, 28.149
0.87L (cm) Equivalent Contact Area
Average width of equivalent area, 19.386 Actual Contact Area
0.6L (cm)
(a)

Pavement Model Assumptions and Simplification 0.87 L

As in all research works using a finite-element model, some


assumptions were made in this study to establish a large-scale
3D viscoelastic multilayered system of the pavement structure.
The assumptions made to simplify the real road system and focus
0.6L
on the key factors were as follows:
1. Each pavement structure layer’s materials were assumed to be
homogeneous and isotropic. The materials of the road base, sub-
base, cement-stabilized subgrade, and natural soil were assumed
to respond linearly and elastically to the applied dynamic loads.
Asphalt layers were modeled as viscoelastic with a simple linear
relationship.
2. The interface between adjacent layers was assumed to be com-
pletely continuous, because the road test section was newly built.
3. When the vehicle was in motion, the stress and strain generated 0.6L
as well as the displacement of the pavement structure at an in-
finite distance from the vehicular loading was zero.
4. Tires and pavement were considered as uncoupled, and there-
fore were established as a one-solid model with known contact
L
stress (field measured contact stress). In addition, the surface of
the pavement model was a horizontal plane, and the slope and (b)
longitudinal unevenness of the road surface were neglected;
Fig. 4. Actual and equivalent contact areas: (a) single tire; and (b) dual
In addition, the pavement structure initial stress condition was
tires.
neglected.

© ASCE 04020006-5 Int. J. Geomech.

Int. J. Geomech., 2020, 20(3): 04020006


Pavement Material Characterization and Parameters 1.2Fr
E¼ ð10Þ
πD2 ε3
In FE analysis, the material characteristics are needed as input
parameters for pavement system modeling. Furthermore, accurate where E = elastic modulus (MPa); Fr = maximum applied load (N);
mechanistic analysis of the layered pavement system requires the D = sample diameter (mm); and ε3 = compressive strain of specimen
adoption of an ideal constitutive model for each layer (Ali et al. when loading reaches 0.3Fr .
2009; Li et al. 2017). Therefore, it is necessary to select an ideal Previous studies revealed that cement-treated bound material in
constitutive model that will be used to characterize the behavior and the field is susceptible to cracking. In addition, the modulus of a
the material properties of each layer. This section includes a brief cracked CTB is different from the modulus of the uncracked CTB
description of the materials of each layer of the road system as well measured in the laboratory (Freeme et al. 1987). Therefore, the
as the characterization of their relative properties. samples were tested as soon as the construction of the pavement
test section was completed to ensure that the CTB layer was not yet
cracked. Thus, the module tests were performed on the uncracked
Bonding Layers and Natural Soil Material Properties
Downloaded from ascelibrary.org by Imperial College London on 01/04/20. Copyright ASCE. For personal use only; all rights reserved.

CTB. Table 3 summarizes the calculated elastic modulus and other


Characterization material parameters.
The subgrade layer consisted of 1.5% cement-stabilized soil (CS),
whereas the natural soil consisted of soil materials in place. The
Asphalt Concrete Layers’ Material Property
top of the natural soil was improved by compaction without
Characterization
the addition of stabilizer (hydraulic binder, cement, and so forth).
The semi-rigid base and subbase layers consisted of 4% cement- It is clearly stated in the literature that the basic performance of the
treated bound materials (CTBs). The composition of the CTBs binder, the aggregate gradation, and the design properties of the
in the mixture is listed in Table 2. mixture have a great influence on the asphalt mixture mechanical
In this study, the materials from the road base to the natural soil properties. These mechanical properties directly influence the dy-
were assumed to be isotropic linear elastic. The cement-stabilized namic properties of the asphalt mixture. Therefore it is appropriate
subgrade soil and the natural soil Young’s moduli were obtained to briefly describe these.
from back-calculation after falling-weight deflectometer tests were The flexible layer of the instrumented road test section consisted
performed on the constructed road test section, and their values of three distinct asphalt mixtures. Stone mastic asphalt with stan-
are summarized in Table 3. The parameters of the CTBs were ob- dard asphalt binder was applied as the AC surface course. Styrene-
tained from laboratory-conducted uniaxial compression modulus butadiene-styrene–modified asphalt (SBS-AC20) was applied in
test. This, according to the Chinese test method J051-T0845 in the the AC middle course. The last asphalt layer was a mixture of high-
current JTG D50 (Chinese Standards 2017). The CTB samples were elasticity and high-viscosity and SBS-modified compound. The
core drilled from the field and cut to a size of 150 × 150 mm maximum nominal sizes of the aggregates constituting the SMA13,
(diameter × height). After being immersed in the water for 24 h, SBS-AC20, and HE-AC5 mixtures were, respectively, 19.5, 19, and
the samples were uniformly loaded with a universal testing machine 4.75 mm. The basic performance of each binder applied in each type
at the loading speed of 1 mm=min until failure. The applied load and of asphalt mixture is listed in Table 4, and the aggregate gradation of
the strain data were recorded to determine a reliable and accurate each mixture is illustrated in Fig. 5. The basic volumetric design
mechanical modulus of each simple by the following relationship: properties of each mixture are provided in Table 5. More details
were given by Tan et al. (2017) and Sun et al. (2018).
The literature clearly states that the behavior of the asphalt
concrete is not only time or frequency dependent but also temper-
Table 2. Composition of CTB in mix design ature dependent. In addition, asphalt concrete is characterized as
viscoelastic material whose stiffness depends on time, tempera-
Composition Weight ratio (%)
ture, and the frequency of the applied load (Li et al. 2015;
20–30 mm 28 Gungor et al. 2016). Asphalt layers from the instrumented pave-
10–20 mm 26 ment structure were assumed to be isotropic and to behave as a
5–10 mm 18
linear viscoelastic material. For isotropic viscoelasticity, there are
Sand 28
many constitutive models: the Kelvin–Voigt model, the Burgers

Table 3. Material parameters of pavement structure


Parameter
Damping ratio,
ζ ¼ 0.05
Unit weight Elastic modulus Poisson’s
No. Layer Material (kg=m3 ) (MPa) ratio α β
1 AC surface course SMA13 2,400 Viscoelastica 0.30 0.342 0.0069
2 AC middle course SBS-AC20 2,400 Viscoelastica 0.30 0.342 0.0069
3 AC bottom course HE-AC5 2,300 Viscoelastica 0.35 0.342 0.0069
4 Semi-rigid base CTB 2,100 14,000 0.25 0.342 0.0069
5 Semi-rigid subbase CTB 2,100 14,000 0.25 0.342 0.0069
6 Cement-stabilized subgrade CS 1,900 6,000 0.35 0.342 0.0069
7 Natural soil In-place soil 1,800 120 0.40 0.342 0.0069
Note: ζ = critical damping ratio; α = Rayleigh damping mass proportional; and β = Rayleigh damping stiffness proportional.
a
Elastic modulus for viscoelasticity (moduli time scale is instantaneous) (Table 6).

© ASCE 04020006-6 Int. J. Geomech.

Int. J. Geomech., 2020, 20(3): 04020006


Table 4. Basic performance of binder applied in each mixture Al-Qadi 2008). The dynamic complex modulus tests were con-
Asphalt mixture ducted as per ASTM-D3497 (ASTM 1995). Haversine loading
was applied to the specimen without impact and at loads varying
SMA13 SBS-AC20 HE-AC5
between 0 and 241 kPa for each load application for a minimum of
Asphalt cement type 30 s (not exceeding 45 s) at temperatures of −10°C, 4.4°C, 21.1°C,
Asphalt Compound modified 37.8°C, and 54.4°C at loading frequencies of 0.1, 0.5, 1, 5, 10, and
Performance binder with SBS Test method 25 Hz for each temperature. Figs. 6(a–d) depict the changes in the
complex modulus with the temperature for each loading fre-
Penetration, 25°C, 70 89 T0604-2011
100G, 5 S (0.1 mm) quency on each asphalt concrete mixture. These were used to pro-
Softening point (°C) 77 86 T0606-2011 duce a master curve and to capture the asphalt concrete material
Ductility, 5 cm=min, 42 62 T0605-2011 time, temperature, and loading frequency dependency. For any tar-
5°C (cm) geted temperature, such as the pavement in situ measured field tem-
Viscosity, 135°C (Pa · s) 1.048 1.622 T0625-2011 perature, the master curves can be shifted to obtain the shear
Downloaded from ascelibrary.org by Imperial College London on 01/04/20. Copyright ASCE. For personal use only; all rights reserved.

relaxation modulus and the bulk relaxation modulus of the asphalt


mixture.

Asphalt Concrete Material Time Dependency and


100 HE-AC5 Prony–Dirichlet Series
SMA-13
The time dependency of the asphalt mixture at each layer can be
Percent passing (%)

80 SBS-AC20
represented by a Prony–Dirichlet series in the form of shear and
60 bulk modulus (Hibbit, Karlsson, and Sorensen 2007)
 X
N 
40
GðtÞ ¼ G0 1 − Gi ð1 − eð−t=τ i Þ Þ ð12Þ
i¼1
20
 X
N 
0
0.075 0.15 0.3 0.6 1.18 2.36 4.75 9.5 13.2 16 19 26.5
KðtÞ ¼ K 0 1 − K i ð1 − eð−t=τ i Þ Þ ð13Þ
i¼1
Sieve size (mm)

Fig. 5. Aggregate gradation of each mixture. where N = number of Prony–Dirichlet series terms; G0 =
instantaneous shear modulus; K 0 = instantaneous bulk modulus;
and Gi , K i , and τ i = Prony–Dirichlet series coefficients.
A constant Poisson’s ratio (μ) of 0.30 was assumed for SMA13
and SBS-AC20, whereas 0.35 was selected for HE-AC5. The shear
Table 5. Volumetric design properties of each mixtures and bulk moduli of each asphalt layer were determined from the
Mixture VTM VFA VMA AC Pbe Gmm Gse Pba relaxation modulus EðtÞ using the following relationship:
SMA13 4.10 76.80 17.80 6.10 5.8 2.504 2.765 0.4 EðtÞ
SBS-AC20 4.17 69.90 13.67 3.92 3.6 2.794 3.007 0.3 GðtÞ ¼ ð14Þ
2ð1 þ μÞ
HE-AC5 1.84 91.71 21.68 8.73 8.1 2.538 2.960 0.7
Note: VTM = air void in compacted mix; VFA = void filled with asphalt EðtÞ
volume; VMA = void mineral aggregate; AC = asphalt cement contents; KðtÞ ¼ ð15Þ
3ð1 − 2μÞ
Pbe = effective asphalt binder; Gmm = maximum theoretical specific
gravity of mix; Gse = aggregate effective specific gravity; and Pba =
asphalt absorption (by aggregate). By normalizing the shear modulus by the instantaneous shear
modulus G0 , Eq. (15) can be simplified into a ratio and expressed
as follows:
GðtÞ
model, the generalized Maxwell model, and so forth. This paper gðtÞ ¼ ð16Þ
used the generalized Maxwell solid model [Eqs. (13) and (14)] to G0
characterize the viscoelastic behavior of the asphalt layers and to
simulate the dynamic behavior. The basic hereditary integral for- X
N
gðtÞ ¼ 1 − Gi ð1 − eð−t=τ i Þ Þ ð17Þ
mulation of an isotropic viscoelastic material can be expressed as i¼1
(Blab 2002; Hibbit, Karlsson, and Sorensen 2007; Hu et al. 2017)
Z Z where Gi = material constants; τ i = retardation time; and t =
t de t dðtr½εÞ reduced relaxation time.
σðtÞ ¼ 2Gðt − τ Þ dτ þ I Kðt − τ Þ dτ ð11Þ
0 dτ 0 dτ The Abaqus nonlinear curve-fitting utility was used to deter-
mine the Prony–Dirichlet series parameters. Then the RMS error
where GðtÞ and KðtÞ = shear and bulk relaxation function; e and (RMSE) function in Eq. (18) was used to accurately fit the mea-
tr½ε = deviatoric and volumetric parts of strain tensor; t = reduced ~
sured shear modulus gðtÞ to the calculated shear modulus gðtÞ
relaxation time; and I = tensor unit. (Hibbit, Karlsson, and Sorensen 2007)
Dynamic complex modulus tests were performed on each vffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
u N
mixture to obtain the relaxation modulus that was used to u1 X
determine the shear relaxation modulus and the bulk (volumetric) RMSE ¼ t ½g ðtÞ − g~ i ðtÞ2 ð18Þ
N i¼1 i
relaxation modulus (Alqadi and Elseifi 2006; Yoo and

© ASCE 04020006-7 Int. J. Geomech.

Int. J. Geomech., 2020, 20(3): 04020006


(a) (b)
Downloaded from ascelibrary.org by Imperial College London on 01/04/20. Copyright ASCE. For personal use only; all rights reserved.

(c)

Fig. 6. Dynamic modulus versus temperature for each loading frequency: (a) SMA13; (b) SBS-AC20; and (c) HE-AC5.

Asphalt Concrete Material Temperature Dependence and where α, β, γ, and δ = regression coefficients; t = time before con-
WLF Equation version under temperature test; and tr = time after conversion under
The properties and behavior of the asphalt concrete materials used in reference temperature.
the flexible layer of the semi-rigid pavement structure are highly According to Eqs. (12)–(21), the WLF equation constant and
influenced by the temperature. Therefore, in addition to the need Prony–Dirichlet series were calculated at the reference temperature
to take into account the time dependence of the asphalt material, of 10°C for each asphalt mixture (Table 6). Similarly, at any target
the temperature-dependent behavior also must be modeled using a temperature, the asphalt material parameters can be obtained by
seasonal time–temperature shift factor. This makes it possible to fitting the aforementioned equations. The asphalt reference temper-
convert the master curve from the test temperature to the reference ature used in the analysis was taken equal to the mean value re-
temperature, thus modifying the initial time (t) in a reduced time (tr) corded through the thickness of asphalt layers during the field
using Eqs. (20) and (21). Several relationships have been proposed to measurement.
model the time–temperature superposition of the asphalt concrete
(Pellinen et al. 2004; Chailleux et al. 2011). This paper used the
Williams–Landel–Ferry function (WLF) (Williams et al. 1955; Full-Scale Field Testing and Finite-Element Model
Hibbit, Karlsson, and Sorensen 2007) Validations
−C1 ðT − T ref Þ
logðαTÞ ¼ ð19Þ Pavement Section Instrumentation and Field Testing
C2 þ ðT − T ref Þ
The instrumented semi-rigid asphalt pavement modeled in this FE
where αT = time–temperature shift factor; C1 and C2 = regression analysis is a part of a project that focuses on the dynamic behavior
coefficients; T = analysis temperature; and T ref = reference of a typical pavement, designed to eliminate reflective cracks and
temperature. bottom-up fatigue cracking. Thus, a typical high-elasticity and
A sigmoidal function introduced in the AASHOTO pavement high-viscosity mixture, HE-AC5, was applied as a stress-absorbing
design guide (AASHTO 2004) was used to fit the relaxation modu- layer to resist fatigue cracking. Then SMA13 and SBS-AC20 were
lus data and obtain the regression coefficients. This function is used to enhance tire wear, permanent deformation resistance, and
described by the following expressions: durability.
The test field is located in Shandong province, China. The
α
logðEðtr ÞÞ ¼ δ þ ð20Þ Rizhan–LanKao test road is an important section of the national
1 þ expðβ þ γ logðtr ÞÞ highway network, G1511. Following a comprehensive instrumen-
tation layout [Figs. 7(a–d)], fiber Bragg grating (FBG) sensors with
t different packaging materials developed by Harbin University of
αT ¼ ð21Þ
tr Technology were adopted for the road test instrumentation during

© ASCE 04020006-8 Int. J. Geomech.

Int. J. Geomech., 2020, 20(3): 04020006


Table 6. Inputs parameters for defining viscoelastic behavior of each asphaltic layer at 10°C
SMA13 SBS-AC20 HE-AC5
τi Gi τi Gi τi Gi
No. Prony–Dirichlet series for generalized Maxwell model
−5 −1
1 1.947 × 10 1.593 × 10 1.003 × 10−4 1.519 × 10−1 2.921 × 10−5 2.019 × 10−1
2 7.661 × 10−4 1.681 × 10−1 3.440 × 10−3 1.326 × 10−1 9.227 × 10−4 2.113 × 10−1
3 3.968 × 10−2 2.089 × 10−1 1.651 × 10−1 1.884 × 10−1 3.689 × 10−2 1.775 × 10−1
4 1.004 × 100 2.022 × 10−1 2.472 × 100 2.603 × 10−1 1.248 × 100 1.797 × 10−1
5 2.267 × 101 1.776 × 10−1 4.009 × 101 1.789 × 10−1 6.323 × 101 1.499 × 10−1
6 1.279 × 103 8.091 × 102 2.857 × 103 6.813 × 10−2 3.399 × 103 7.165 × 10−2
Elastic modulus for viscoelasticity (moduli time scale is instantaneous)
10,000 15,000 6,500
Downloaded from ascelibrary.org by Imperial College London on 01/04/20. Copyright ASCE. For personal use only; all rights reserved.

Williams–Landel–Ferry equation constants


C1 23.74 27.061 25.44
C2 177.80 222.90 198.80
RMSE 0.11 0.02 0.07
Note: No. = number of Prony–Series; τ = relaxation time; and G = dimensionless relaxation modulus.

Median Strip

Overtaking Lane

Wheel path
Carriageway Strain Sensor zone
Traffic Direction

Wheel path

Emergency Lane K313+800 K314+000 K314+400

Longitudinal Strain sensors Vertical Strain sensors


Transversal Strain sensors Temperature and humidity sensors

(a)

0.32 0.32 0.50 0.32 0.50


Marking line between carriageway

Marking line between carriageway

Marking line between carriageway

34# 33# 22# 21# 10# 9# V3


and emergency lane

and emergency lane

and emergency lane


1.90

1.90

1.90

36# 35# 24# 23# V7 12# 11# V4


0.88

0.88

0.88

W2#

(b) (c) (d)

Fig. 7. Instrumentation layout of the experimental pavement structure and disposition details at mile stake K314 þ 400: (a) plan view of structural
instrumentation; (b) sensor layout details at the bottom of the SM13; (c) sensor layout details at the middepth of the SBS-AC20; and (d) sensor layout
details at the top of the HE-AC5.

© ASCE 04020006-9 Int. J. Geomech.

Int. J. Geomech., 2020, 20(3): 04020006


Downloaded from ascelibrary.org by Imperial College London on 01/04/20. Copyright ASCE. For personal use only; all rights reserved.

Fig. 8. FBG sensors used at Rizhan–LanKao test road section: (a) SFRP-FGB sensor; (b) LFRP-FGB sensor; and (c) TS-FGB sensor.

construction. To quantify the pavement dynamic response at three Field-Measured Dynamic Strain Response
different depths: 3.5 cm, on the bottom of the SMA13; 10.25 cm,
During the experiment, for each passage of the controlled-load truck
on the middepth of the SBS-AC20; and 18.50 cm, on the top of the over the instrument area, data were measured and recorded by the
HE-AC5. The road test pavement was instrumented on three dis- data acquisition system at a rate of 2,000 Hz. The recorded high-
tinctive sections located at miles K313 þ 800, K314 þ 000, and frequency signal provided a complete register of the instrument’s
K314 þ 400. A total of 36 short fiber-reinforced polymer (SFRP) response to each successive passage of the front and rear axles
FBG sensors [Fig. 8(a)] were installed in the wheel path to measure of the test truck. Figs. 10(a–c) show a typical three-directional strain
the pavement dynamic strain response in both transverse and signal obtained from the sensors integrated into the SBS-AC20 mid-
longitudinal directions, whereas 7 long fiber-reinforced polymer depth when the steering and driving axles of the test truck loaded at
(LFRP) FBG sensors [Fig. 8(b)] were installed to collect the ver- 258 kN passed over the instrument pavement at 20 km=h. For in-
tical dynamic strain. Four FBG temperature sensors (TS) [Fig. 8(c)] stance, the longitudinal dynamic strain in Fig. 10(c), clearly shows
were embedded outside the wheel path to monitor the in situ that the first peak strain values, at about 4.10 s, depict the passing of
temperature. All sensors were carefully embedded in the structure the truck front axle over the sensor. The following two peaks values,
during the construction period, and their positions were marked on at approximately 5.08 and 5.33 s, correspond to the dual tandem
the road surface to facilitate their identification during the field drive wheels of the truck passing over the sensor.
test. Figs. 9(a–d) display the installation of the various FBG sensors The original strain signals recorded as the test truck axles load
within the asphalt layer, the data monitoring system, and the passed over the instruments were polluted by noise [Figs. 10(a–c),
controlled-load test. unfiltered signals]. The presence of this unwanted signal noise can
A series of controlled wheel loading tests were carried out on the be related to the sensors themselves, to the surrounding materials,
test road section with the objective to investigate the dynamic re- and especially to the dynamic behavior of the pavement structure.
sponse of the pavement under a moving heavy vehicle load while As indicated in the literature, the structural dynamic response de-
exposed to environmental conditions. During the experiments, a pends on the ratio of the external loading frequency to the road
Sinotruck HOWO A7 6 × 4 dump truck with a single steering axle system natural frequency (Uddin 2003; Darestani et al. 2006). The
and tandem driving axles was used to conduct the load tests. Test range of natural frequency was found to be 6–14 Hz for flexible
truck loads were set at 149, 258, and 362 kN, with an axle calibra- pavements and 20–58 Hz for rigid pavements, whereas the truck
tion error range controlled at 1 kN. The detailed parameters of the loading frequency was about 6.5 Hz at 82 km=h and approximately
controlled-load truck used in this paper are listed in Tables 1 and 2. 4.6 Hz and 58 km=h (Gillespie 1993). Thus, the true amplitudes of
For each load configuration, the truck was driven directly over the original signal obtained from sensors integrated within the
the top of the sensors at targeted moving speeds of 20, 40, and pavement structure can fluctuate greatly. Therefore, it is important
60 km=h. Data collected from the FBG sensors were recorded to reduce the signal noise by removing these unwanted peaks
and analyzed for all the tests. before analyzing the response.
In this study, only the strain signal collected in Section K314 þ Signal processing tools such as the wavelet transform, smooth-
400 with the truck loaded at 258 kN and moving at 20 km=h sub- ing, and Fourier filter often are employed to improve the signal-to-
sequently were used to compare the dynamic response obtain from noise ratio (SNR) of the original signal. However, peak clipping is
the field test with that obtain by FE simulation in order to validate unavoidable when the original signals are processed by digital fil-
the developed full-scale 3D viscoelastic FE model. tering or smoothing. In this study, the FFT filtering tool available in

© ASCE 04020006-10 Int. J. Geomech.

Int. J. Geomech., 2020, 20(3): 04020006


Downloaded from ascelibrary.org by Imperial College London on 01/04/20. Copyright ASCE. For personal use only; all rights reserved.

Fig. 9. Installation of FBG strain sensors and field strain measurements in Shandong Expressway: (a) sensor positioning at the top of the HE-AC5;
(b) installation and mechanical paving; (c) controlled-load test in progress; and (d) data acquisition system.

Origin-Pro 2017 software was used to filtered the original strain field measured load duration was shorter than the calculated load
signals. The peak values of the filtered signal were less than those duration. The transverse dynamic strain taken at the bottom of the
of the unfiltered original signal [Figs. 10(a–c), filtered signals]. SMA13, in the middepth of the SBS-AC20, and at the bottom of
the HE-AC5 respectively indicated a difference of about 2%, 7%,
and 4% between the predicted and calculated maximum strains.
Comparative Analysis of FE Simulation Result and For vertical deformation [Figs. 11(c and d)], the response of the
Field-Measured Data measurement in the field was slower to recover and the loading time
The results of the 3D FE analysis and the field-recorded dy- was longer. In addition, the vertical deformation predicted at the
namic strains under the controlled-load truck at a moving speed middepth of the SBS-AC20 and at the bottom of the HE-AC5 had
of 20 km=h are depicted Figs. 11(a–f). The time histories of the an error range of less than 10%. Regarding the longitudinal strain
dynamic strain were measured at the same analysis point in the ac- [Figs. 11(e and f)], the peak values of the field-measured longitu-
tual pavement and FE Model. However, the time histories of the dinal strain were slightly lower than the predicted values, whereas
dynamic strain of the FE simulation in Figs. 11(a–f) were taken the strains in compression were relatively high. A difference of less
at Analysis points a, b and d in Fig. 12. In addition, the dynamic than 5% was observed between the maximum values of the pre-
strain histories of the field tests were those recorded by the sensors dicted longitudinal strain and those measured at the bottom of
installed at the bottom of the SMA13, the middepth of the SBS- the SMA13, in the middepth of the SBS-AC20, and at the bottom
AC20, and the top of the HE-AC (Fig. 7). of the HE-AC5.
The time histories of the dynamic strain predicted by the estab- An in-depth investigation of the comparative study of the result
lished FE model and the dynamic response recorded from the field of the FE simulation and measurements obtained from the field in-
testing were in a good agreement. Furthermore, the trend of change dicated certain differences. The reasons for the differences are
over time and the peak values were very close to each other, but 1. During the full-scale field testing, it was difficult to accurately
there was a certain difference in the residual strain and response control the speed of the controlled-load truck and the wheels’
time. load position, which ideally should be applied symmetrically in
Explicitly in the case of transverse strain [Figs. 11(a and b)], the wheel path and directly below the center of dual tires.
both predicted and measured responses had the same shape of 2. The dynamic responses recorded by the different FBG sen-
the curve (strain versus time). However, in terms of time history, the sors were influenced by the surrounding material constraints.

© ASCE 04020006-11 Int. J. Geomech.

Int. J. Geomech., 2020, 20(3): 04020006


Downloaded from ascelibrary.org by Imperial College London on 01/04/20. Copyright ASCE. For personal use only; all rights reserved.

Fig. 10. Unfiltered and filtered three-directional strain time histories recorded during the passing of the controlled-load truck loaded at 258 kN and
moving at 20 km=h: (a) transverse dynamic strain; (b) vertical dynamic strain; and (c) longitudinal dynamic strain.

The sensors were subjected to compressive stress not only at the between two distinct variables x and y can be obtained from the
approach of the controlled-load truck but also after the passage following expression: P P
of the truck. ðxi − x̄Þ ðyi − ȳÞ
rxy ¼ pP ffip P
ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi ð22Þ
3. The wheel load was assumed to be uniformly distributed over a ðxi − x̄Þ2 ðyi − ȳÞ2
rectangular contact area between each tire and pavement, which
differs from the real contact stresses and pressure induced by the P P
tires during field testing. Only accurate three-dimensional con- where x̄ ¼ ð1=nÞ Ni xi = mean of x; and ȳ ¼ ð1=nÞ Ni yi = mean
tact pressure and a nonuniform contact area can simulate with of y. The Pearson correlation coefficient (PCC) can take on any
very good accuracy the real conditions of stress distribution in value in the range ½−1; 1. There is no linear correlation between
the pavement structure. two variables when r ¼ 0. The correlation coefficient magnitude
All in all, the dynamic strains predicted by numerical simulation reveals the relationship strength, whereas its sign shows the direc-
at different depths within the pavement structure basically were tion of the relationship. As general a guideline for evaluating the
consistent with those measured in the field. Therefore, the estab- correlation strength (Cohen 1988), if 1 < jrj < 3, there is a small
lished 3D viscoelastic FE model can be used to accurately predict or weak correlation between variables; if 3 < jrj < 5, there is a
the structural behavior of the pavement subjected to full-scale truck medium or moderate correlation; and if 5 < jrj, this indicates a large
loading in motion. or strong correlation.
Tables 7–9 present the correlation matrix of the three directional
strains from field data and numerical simulations. The value of r
Correlations Analysis of FE Analysis Result and
was between 0.694 and 0.990. This proves that the measured and
Field-Measured Data
calculated dynamic responses at different depths had a positive or
As the comparative study between the simulation result and the field strong correlation. Thus, the responses had a statistically significant
data measurement showed, the established FE model can accurately relationship at the 0.01 level of significance for a two-tailed test.
predict the pavement dynamic response. Nevertheless, there was a Therefore, the margin of error between the simulation result and the
very slight difference between the curves’ shapes. To analyze the field data had a negligible influence on the accuracy of the predic-
influence of this margin of error on the accuracy of the prediction tion model. This clearly reinforces the previous statement that the
model, a correlation analysis of the FE simulation result and field predicted dynamic strains were in good agreement with the field
measured strain data was performed. The correlation coefficient rxy measurements. Consequently, the developed 3D FE model was

© ASCE 04020006-12 Int. J. Geomech.

Int. J. Geomech., 2020, 20(3): 04020006


Downloaded from ascelibrary.org by Imperial College London on 01/04/20. Copyright ASCE. For personal use only; all rights reserved.

(a) (b)

(c) (d)

(e) (f)

Fig. 11. Three-directional dynamic strains from FE simulation and field measurement during the passing of the test truck loaded at 258 kN and
moving at 20 km=h: (a and b) transverse dynamic strain; (c and d) vertical dynamic strain; and (e and f) longitudinal dynamic strain.

used to better understand the structural behavior of layered asphalt respectively taken at depths of 0.035, 0.100, 0.165, 0.185, 0.275,
pavements subjected to dynamic moving wheel loads. 0.450, and 0.535 m directly right below the center of the dual tires.
Points 1, 2, 3, 4, 5, 6, and 7 were selected respectively for lateral
positions 0, 0.050, 0.152, 0.255, 0.475, 0.675, and 0.915 m in the
Finite-Element Simulation Results Analysis and middle of the SBS-AC20 layer.
Discussion
Transverse Stress–Strain Analysis
Typical measurement points in the pavement structure were se-
lected along with the depth mainly under the center of the dual tires Figs. 13(a and b) and 14(a and b) respectively depict the time his-
but also at different lateral positions in the middepth of the SBS- tories of transverse strain and transverse stress at different depths
AC20 to ensure a better analysis of the dynamic response of the under the center of the dual tires and at different lateral positions in
asphalt pavement under a moving traffic load. The selected analy- the middepth of the asphalt concrete middle course before and after
sis points are shown in Fig. 12; Points a, b, c, d, e, f, and g were the crossing of the controlled-load truck.

© ASCE 04020006-13 Int. J. Geomech.

Int. J. Geomech., 2020, 20(3): 04020006


same patterns as the transverse stress with a slight delay. The maxi-
mum value of the transverse stress response was reached when the
load just reached the measurement point, whereas the maximum
value of the transverse dynamic strain response occurred 0.05 s
SMA13 a
b 1 2 3 5 6 7
after the load reached this point.
SBS-AC20 4
HE-AC5 c The analysis of the time histories of the transverse strain mea-
Base d sured at different lateral positions at the mid-depth of the SBS-
e
Subbase f AC20 illustrated the influence of the lateral position of the sensors
Subgrade g in the pavement dynamic response [Figs. 13(b) and 14(b)]. The
maximum value of the dynamic transverse strain and stress re-
Natural Soil
sponse occurred just below the center of the dual tires for the strains
and under the tires for the stresses. The dynamic transverse stress
and strain response generated as the load moved farther from the
Downloaded from ascelibrary.org by Imperial College London on 01/04/20. Copyright ASCE. For personal use only; all rights reserved.

measurement point decreased gradually. Therefore, the position of


the loading path and the embedded point of the sensors need to be
Fig. 12. Analysis points for dynamic response of asphalt pavement. controlled during the test process to ensure that the recorded dy-
namic response accurately reflects the field stress–strain state.

Table 7. Correlation matrix of field transversal strain and numerical Vertical Stress–Strain Analysis
simulations
Figs. 15(a and b) and 16(a and b) respectively display the time his-
Field measurement tories of vertical strain and vertical stress at different analysis points.
Numerical Significance
simulation SMA13 SBS-AC20 HE-AC5 (two-tailed) From the time histories of vertical stress, it can be seen that, before
and after the wheel load passes, a slight vertical compressive stress
SMA13 0.980a — — 0.000
SBS-AC20 — 0.972a — 0.000
occurs first, and then the tensile stress with a larger value appears
HE-AC5 — — 0.950a 0.000 just at the moment that the load is on the point of measurement. The
a
maximum value of the vertical stress occurred just under the wheel
Correlation is significant at 0.01 level (two-tailed). tire, and as the measurement points moved away from the tire, the
vertical stress decreased. The amplitude of the dynamic vertical
stress first increased with depth, reaching the maximum value at
Table 8. Correlation matrix of field vertical strain and numerical a depth of approximately 0.165 m, after which it decreased progres-
simulations sively from the peak value [Fig. 16(a)].
Field measurement The analysis of the vertical strain at different analysis points
Numerical Significance showed that as the measurement point moved away from the wheel
simulation SMA13 SBS-AC20 HE-AC5 (two-tailed)
load, the dynamic vertical strain gradually increased until reaching
SMA13 — — — 0.000 a certain maximum value at a lateral point of 0.695 m, after which
SBS-AC20 — 0.930a — 0.000 the strain started to decrease. From the pavement surface course to
HE-AC5 — — 0.694a 0.000 the base layer, the longitudinal strain responses were in compres-
a sion, and from the pavement base layer to subgrade, they were sub-
Correlation is significant at 0.01 level (two-tailed).
ject to tension. In the compressive zone, the vertical strain responses
decreased as the depth increased, whereas in the tensile zone they
Table 9. Correlation matrix of field longitudinal strain and numerical
increased as the depth increased.
simulations

Numerical
Field measurement
Significance Longitudinal Stress–Strain Analysis
simulation SMA13 SBS-AC20 HE-AC5 (two-tailed) Figs. 17(a and b) and 18(a and b) display respectively the time his-
SMA13 0.965a — — 0.000 tories of the longitudinal strain and longitudinal stress at different
SBS-AC20 — 0.990a — 0.000 depths in the center of the dual tires and at different lateral positions
HE-AC5 — — 0.961a 0.000 in the middepth of the AC middle course under the controlled-load
a
Correlation is significant at 0.01 level (two-tailed). truck.
At the analysis point at depth ¼ 0.100 m (the middepth of
the SBS AC20), as the loads approach the measurement points, the
Before and after the wheel load passed the measurement point, longitudinal dynamic stress response changed from compression to
the dynamic transverse stresses and dynamic transverse strain am- tension. Then it gradually increased to its maximum tensile stress
plitudes first increase, then decreased gradually after reaching the value when the wheel load was directly below the measurement
maximum value, and there was a certain residual strain at the end of point. The tensile stress decreased progressively and finally changed
the cycle (Figs. 13 and 14). The dynamic response analysis of trans- back to compression as the truck load moved farther from the meas-
verse stress recorded at different depths in the center of the dual uring points. Hence, as the truck approached and moved away from
tires [Fig. 14(a)] revealed that from depth ¼ 0.035 to 0.275 m, the measuring points, the longitudinal dynamic stresses response
the transverse strains were subjected to tension; they altered to was subjected to compression, whereas the stresses were tensile
compression after this depth. In the compression zone, as the depth under the wheel load.
increased, the amplitude of the transverse dynamic strain increased, The analysis of the influence of depth on the longitudinal strains
whereas it was attenuated in the tension zone as the depth in- and stresses showed that from the pavement surface to the road base
creased. The transverse strain response [Fig. 13(a)] followed the layer the dynamic longitudinal strain and stress responses were

© ASCE 04020006-14 Int. J. Geomech.

Int. J. Geomech., 2020, 20(3): 04020006


Downloaded from ascelibrary.org by Imperial College London on 01/04/20. Copyright ASCE. For personal use only; all rights reserved.

(a) (b)

Fig. 13. Time histories of transverse strain at different analysis points during the passing of the test truck loaded at 258 kN and moving at 20 km=h:
(a) points at different depths under the center of the dual tires; and (b) different lateral positions in the middepth of the AC middle course.

(a) (b)

Fig. 14. Time histories of transverse stress at different analysis points during the passing of the test truck loaded at 258 kN and moving at 20 km=h:
(a) points at different depths under the center of the dual tires; and (b) different lateral positions in the middepth of the AC middle course.

(a) (b)

Fig. 15. Time histories of vertical strain at different analysis points during the passing of the test truck loaded at 258 kN and moving at 20 km=h:
(a) points at different depths under the center of the dual tires; and (b) different lateral positions in the middepth of the AC middle course.

tensile under the wheel load, whereas they were subject to compres- behind the wheel load. When the stresses and strains under the
sion in front of and behind the wheel load [Figs. 17(a) and 18(a)]. wheel load were in a compression zone, the dynamic responses in-
From the base layer to the subgrade they were subject to compres- creased as depth increased, whereas when they were in a tension
sion under the wheel load, whereas they were tensile in front of and zone, the dynamic responses decreased as the depth increased.

© ASCE 04020006-15 Int. J. Geomech.

Int. J. Geomech., 2020, 20(3): 04020006


Downloaded from ascelibrary.org by Imperial College London on 01/04/20. Copyright ASCE. For personal use only; all rights reserved.

(a) (b)

Fig. 16. Time histories of vertical stress at different analysis points during the passing of the test truck loaded at 258 kN and moving at 20 km=h:
(a) points at different depths under the center of the dual tires; and (b) different lateral positions in the middepth of the AC middle course.

(a) (b)

Fig. 17. Time histories of longitudinal strain at different analysis points during the passing of the test truck loaded at 258 kN and moving at 20 km=h:
(a) points at different depths under the center of the dual tires; and (b) different lateral positions in the middepth of the AC middle course.

(a) (b)

Fig. 18. Time histories of longitudinal stress at different analysis points during the passing of the test truck loaded at 258 kN and moving at 20 km=h:
(a) points at different depths under the center of the dual tires; and (b) different lateral positions in the middepth of the AC middle course.

The analysis of the influence of lateral positions of measuring points was just below the contact surface of the wheel load (from
points on the longitudinal dynamic strain and stress revealed that lateral ¼ 0.050 to 0.255 m) [Figs. 17(b) and 18(b)]. The dynamic
the maximum values of the longitudinal strain and the stress re- longitudinal strain and stress responses decreased gradually as the
sponse were reached when the lateral position of the measurement measurement points moved farther from the wheel load.

© ASCE 04020006-16 Int. J. Geomech.

Int. J. Geomech., 2020, 20(3): 04020006


Downloaded from ascelibrary.org by Imperial College London on 01/04/20. Copyright ASCE. For personal use only; all rights reserved.

(a) (b)

Fig. 19. Time histories of shear strain at different analysis points during the passing of the test truck loaded at 258 kN and moving at 20 km=h:
(a) points at different depths under the center of the dual tires; and (b) different lateral positions in the middepth of the AC middle course.

(a) (b)

Fig. 20. Time histories of shear stress at different analysis points during the passing of the test truck loaded at 258 kN and moving at 20 km=h:
(a) points at different depths under the center of the dual tires; and (b) different lateral positions in the middepth of the AC middle course.

Shear Stress–Strain Analysis strain, after which they decreased gradually from the peak value. It
can be anticipated that the alternative change between positive and
The time histories of shear strain and shear stress measured at dif-
negative shear stress and strain under the complex action of the
ferent depths in the center of the dual tires and at different lateral
moving vehicular loads could be an important reason for the pre-
positions in the middepth of the asphalt concrete middle course
mature fatigue of pavement materials and the rutting of roads. Thus,
under the controlled-load truck are displayed in Figs. 19(a and b)
it is important to fully understand shear stress and shear strain in
and 20(a and b). Before and after the passage of each wheel axle,
semi-rigid types of flexible pavements in order to design more-
there was an alternating change between the negatives and the pos-
sustainable pavement.
itives shear stresses in the pavement structure. The absolute value
of the negative shear stresses was very slightly below that of the
positive shear stresses. The shear strain and shear stress dynamic Summary and Conclusions
responses recorded at different depths under the center of the dual
tires [Figs. 19(a) and 20(a)] and at different lateral positions in the This paper investigated the dynamic response of the semi-rigid type
middepth of the SBS-AC20 [Figs. 19(b) and 20(b)] followed the of flexible pavement structure under full-scale moving heavy
same pattern. The peak value of shear stress and shear strain at vehicular load. A full-scale 3D FE model of transient dynamic
the middepth of the asphalt surface middle course occurred directly analysis of the pavement was established to obtain critical strain
below the center of the tire, and as the wheel load moved away from and stress responses within the pavement structure. The developed
the measuring point, the amplitude of the dynamic responses pro- 3D FE model incorporates the viscoelastic behavior of the asphalt
gressively decreased. layers and the effect of the transient local dynamic wheel load. The
The amplitudes of the dynamic responses of the shear stress and critical dynamic responses of the pavement structure were analyzed
shear strain under the wheel load precisely at the center of the dual according to the depth under the center of the dual tires and the
tires increased initially with the depth, reaching the maximum value lateral distribution in the middepth of the SBS-AC20 layer. Pave-
at a depth of approximately 60 mm for the stress and 50 mm for the ment dynamic response measured at the Rizhan–LanKao G1511

© ASCE 04020006-17 Int. J. Geomech.

Int. J. Geomech., 2020, 20(3): 04020006


highway pavement test section was used to verify the rationality of may not entirely reflect the actual dynamic properties of the asphalt
the developed viscoelastic 3D FE model. Some of the preliminary concrete; (2) that the behavior of the nonasphaltic layer is linear
findings from this study are summarized as follows: and elastic in an isotropic system may not be representative of the
1. The pavement responses in terms of three-directional strain at stress-dependent state of these layers at some stage; (3) the tire–
the bottom of the SMA13, the middepth of the SBS-AC20, pavement contact stresses that were uniformly distributed on a
and the top of the HE-AC were calculated, and the results were rectangular equivalent area may not entirely be indicative of the
consistent with the full-scale field-measured responses. In addi- three-directional stress state and the nonuniform contact areas rep-
tion, the correlation analysis of the developed FE analysis result resenting the field real loading condition. The authors recommend
and the field-measured data showed that the influence of the the aforementioned concerns for further investigation in future
assumptions made in the development of this 3D viscoelastic studies.
FE model on the accuracy of the prediction of the model were
relatively insignificant. Data Availability Statement
2. Computed and measured dynamic responses indicated that be-
Downloaded from ascelibrary.org by Imperial College London on 01/04/20. Copyright ASCE. For personal use only; all rights reserved.

fore and after the wheel load passes to the measurement point, The data supporting the figures and tables in this paper, the numeri-
the transverse tensile strains and the vertical compressive strain cal model and subroutine code generated in this study, as well as
first increases, and then decreases gradually after reaching the other findings of this study are available from the corresponding
peak value Furthermore, there is a certain residual strain at the author by request.
end of the cycle. The longitudinal strain responses are tensile
under the wheel load, whereas they are subject to compression
in front of and behind the wheel load. Acknowledgments
3. The analysis of the three-directional dynamic response of the
pavement showed that the pavement asphaltic layers are sub- This work was funded by the Chinese Ministry of Science and
jected to tensile vertical stress and tensile transverse stress; the Technology under Grant No. 2014BAC07B00 and the Natural
semi-rigid base layer is subjected to tensile vertical stress and Science Foundation of China under Grant No. 51678207. The
compressive transverse stress; and the semi-rigid subbase, sub- equipment and software were supported by Harbin Institute of
grade, and natural soil are subjected to tensile vertical stress and Technology. The authors express their sincere gratitude to all the
compressive transverse stress. An alternating change between people involved in this research project. Finally, the authors thank
negative shear stress, strain, and positive shear stress occurred the reviewers for useful comments and the editors for improving the
in the pavement structure. manuscript.
4. The analyses of the pavement dynamic response according to
the depth demonstrated that with the increase of the depth, the
transverse strain in the asphalt layers increases, whereas it is References
gradually attenuated in other layers. The dynamic vertical stress
first increased with depth, reaching the maximum value at a AASHTO. 1993. AASHTO guide for design of pavement structures.
Washington, DC: AASHTO.
depth of approximately 0.165 m, after which it gradually de-
Ali, B., M. Sadek, and I. Shahrour. 2009. “Finite-element model for
creased from that maximum value. In the compression zone,
urban pavement rutting: Analysis of pavement rehabilitation methods.”
the vertical strain responses decrease as the depth increases, J. Transp. Eng. 135 (4): 235–239. https://doi.org/10.1061/(ASCE)0733
whereas in the tension zone it increases as the depth increases -947X(2009)135:4(235).
The longitudinal stresses and strains under wheel load increase Alqadi, I., H. Wang, P. Yoo, and S. Dessouky. 2008. “Dynamic analysis and
as depth increases in the compression zone and decrease as the in situ validation of perpetual pavement response to vehicular loading.”
depth increases in the tension zone. Transp. Res. Rec. J. Transp. Res. Board 2087 (2087): 29–39.
5. The analysis of the pavement dynamic response at different lat- Alqadi, I. L., and M. A. Elseifi. 2006. “Viscoelastic modeling and field
eral positions in the middepth of the SBS-AC20 showed that validation of flexible pavements.” J. Eng. Mech. 132 (2): 172–178.
as the measurement points move away from the wheel load, the Assogba, O. C., Y. Tan, Z. Sun, N. Lushinga, and Z. Bin. 2019. “Effect of
dynamic amplifications decreases. The maximum value of the vehicle speed and overload on dynamic response of semi-rigid base as-
phalt pavement.” Road Mater. Pavement Des. 1–31. https://doi.org/10
transverse strain occurs directly below the center of dual tires,
.1080/14680629.2019.1614970.
whereas the maximum value of stress occurs under the tires.
Assogba, O. C., Y. Tan, X. Zhou, C. Zhang, and J. N. Anato. 2020.
The vertical strain gradually increased until it reached its max- “Numerical investigation of the mechanical response of semi-rigid base
imum value at a lateral point of 0.695 m, after which the strain asphalt pavement under traffic load and nonlinear temperature gradient
began to decrease. The maximum value of the vertical stress oc- effect.” Constr. Build. Mater. 235 (Feb): 117406. https://doi.org/10
curs just under the tire. The longitudinal stress and the strain peak .1016/j.conbuildmat.2019.117406.
value occur below the wheel load contact area. The peak value of ASTM. 1995. Standard test method for dynamic modulus of asphalt mix-
shear stress and strain occurs directly below the center of the tire. tures. ASTM D3497. West Conshohocken, PA: ASTM.
Bathe, K.-J. 1982. Finite element procedures in engineering analysis.
Englewood Cliffs, NJ: Prentice-Hall.
Beskou, N. D., G. D. Hatzigeorgiou, and D. D. Theodorakopoulos.
Limitation and Further Research Direction
2016. “Dynamic inelastic analysis of 3-D flexible pavements under mov-
ing vehicles: A unified FEM treatment.” Soil Dyn. Earthquake Eng.
As with all research works, this study has some limitations. To
90 (Nov): 420–431. https://doi.org/10.1016/j.soildyn.2016.09.018.
achieve more-accurate pavement response assessment and perfor-
Blab, A. P. R. 2002. “Modeling measured 3D tire contact stresses in a vis-
mance predictions, it is recommended that future research consider coelastic FE pavement model.” Int. J. Geomech. 2 (3): 271–290. https://
a further improvement in pavement mechanical behavior analysis. doi.org/10.1061/(ASCE)1532-3641(2002)2:3(271).
Although this study formulated some assumptions in the actual 3D Breuer, S., and J. J. Roseman. 1977. “On Saint-Venant’s Principle in three-
viscoelastic FE model with reasonable accuracy: (1) that the asphalt dimensional nonlinear elasticity.” Arch. Ration. Mech. Anal. 63 (2):
concrete behaves as viscoelastic with a simple linear relationship 191–203. https://doi.org/10.1007/BF00280605.

© ASCE 04020006-18 Int. J. Geomech.

Int. J. Geomech., 2020, 20(3): 04020006


Chailleux, E., G. Ramond, C. Such, and C. de La Roche. 2011. “A under dynamic FWD loading.” Constr. Build. Mater. 141 (Jun): 23–35.
mathematical-based master-curve construction method applied to complex https://doi.org/10.1016/j.conbuildmat.2017.02.096.
modulus of bituminous materials.” Supplement, Road Mater. Pavement Li, S., Z. Guo, and Y. Yang. 2015. “Dynamic viscoelastic response of an
Des. 7 (S1): 75–92. https://doi.org/10.1080/14680629.2006.9690059. instrumented asphalt pavement under various axles with non-uniform
Chango, I. V. L., M. Yan, X. Ling, T. Liang, and O. C. Assogba. 2019. stress distribution.” Road Mater. Pavement Des. 17 (2): 446–465.
“Dynamic response analysis of geogrid reinforced embankment sup- https://doi.org/10.1080/14680629.2015.1080178.
ported by CFG pile structure during a high-speed train operation.” Lat. Liao, J., and S. Sargand. 2011. “Viscoelastic FE modeling and verification of
Am. J. Solids Struct. 16 (7). https://doi.org/10.1590/1679-78255710. a U.S. 30 perpetual pavement test section.” Road Mater. Pavement Des.
Chen, X., J. Zhang, and X. Wang. 2015. “Full-scale field testing on a high- 11 (4): 993–1008. https://doi.org/10.1080/14680629.2010.9690316.
way composite pavement dynamic responses.” Transp. Geotech. 4 (Sep): Lu, Z., Z. Hu, H. L. Yao, and J. Liu. 2018. “Field evaluation and analysis of
13–27. https://doi.org/10.1016/j.trgeo.2015.05.002. road subgrade dynamic responses under heavy duty vehicle.” Int. J.
Chinese Standards. 2017. Specifications for design of highway asphalt Pavement Eng. 19 (12): 1077–1086.
pavement. JTG D50-2017. Beijing, China: Ministry of Transport of Pellinen, T. K., M. W. Witczak, and R. F. Bonaquist. 2004. “Asphalt mix
the People’s Republic of China. master curve construction using sigmoidal fitting function with non-linear
Cohen, J. 1988. “Statistical power analyses for the behavioral sciences.”
Downloaded from ascelibrary.org by Imperial College London on 01/04/20. Copyright ASCE. For personal use only; all rights reserved.

least squares optimization.” In Recent advances in materials characteri-


Technometrics 31 (4): 499–500. zation and modeling of pavement systems, 83–101. Reston, VA: ASCE.
Darestani, M. Y., D. P. Thambiratnam, A. Nataatmadja, and D. Baweja. Portland Cement Association. 1966. Thickness design for concrete pave-
2006. “Dynamic response of concrete pavements under vehicular ments. Skokie, IL: Portland Cement Association.
loads.” In Vol. 92 of Proc., IABSE Symposium Report, International Reynaud, P., S. B. Nasr, F. Allou, T. Chaise, D. Nelias, and C. Petit. 2016.
Association for Bridge and Structural Engineering, edited by IABSE “3D modelling of tyre-pavement contact pressure.” Rev. Française Génie
(International Association for Bridge and Structural Engineering), Civil 21 (6): 712–729. https://doi.org/10.1080/19648189.2016.1150894.
59–66. Budapest, Hungary: International Association for Bridge and Sha, A., J. Wang, L. Hu, and X. Zou. 2016. Analysis of dynamic response of
Structural Engineering. asphalt pavement in heavy vehicle simulator tests. Cham, Switzerland:
Doyle, J. D., I. L. Howard, C. A. Gartrell, G. L. Anderton, J. K. Newman, and Springer.
E. S. Berney. 2014. “Full-scale instrumented testing and three-dimensional Sun, L., and Y. Duan. 2013. “Dynamic response of top-down cracked asphalt
modeling of airfield matting systems.” Int. J. Geomech. 14 (2): 161–170. concrete pavement under a half-sinusoidal impact load.” Acta Mech.
https://doi.org/10.1061/(ASCE)GM.1943-5622.0000272. 224 (8): 1865–1877. https://doi.org/10.1007/s00707-013-0849-7.
Freeme, C. R., M. De Beer and A. W. Viljoen. 1987. “The behaviour and
Sun, Z., Y. Xu, Y. Tan, L. Zhang, H. Xu, and A. Meng. 2018. “Investigation
mechanistic design of asphalt pavements.” In Proc., 6th Int. Conf., Struc-
of sand mixture interlayer reducing the thermal constraint strain in as-
tural Design of Asphalt Pavements. Ann Arbor, MI: Univ. of Michigan.
phalt concrete overlay.” Constr. Build. Mater. 171 (May): 357–366.
Gillespie, T. D. 1993. Effects of heavy-vehicle characteristics on pavement
https://doi.org/10.1016/j.conbuildmat.2018.03.095.
response and performance. Washington, DC: Transportation Research
Taherkhani, H., and M. Jalali. 2018. “Viscoelastic analysis of geogrid-
Board.
reinforced asphaltic pavement under different tire configurations.” Int.
Gungor, O. E., I. L. Al-Qadi, A. Gamez, and J. A. Hernandez. 2016. “In-
J. Geomech. 18 (7): 04018060. https://doi.org/10.1061/(ASCE)GM
situ validation of three-dimensional pavement finite element models.”
.1943-5622.0001183.
In The roles of accelerated pavement testing in pavement sustainability,
Tan, Y., Z. Sun, X. Gong, H. Xu, L. Zhang, and Y. Bi. 2017. “Design
edited by J. Aguiar-Moya, A. Vargas-Nordcbeck, F. Leiva-Villacorta,
parameter of low-temperature performance for asphalt mixtures in cold
and L. Loría-Salazar. Cham, Switzerland: Springer.
regions.” Constr. Build. Mater. 155 (30): 1179–1187. https://doi.org/10
Hadi, M. N., and B. Bodhinayake. 2003. “Non-linear finite element analy-
.1016/j.conbuildmat.2017.09.094.
sis of flexible pavements.” Adv. Eng. Software 34 (11–12): 657–662.
https://doi.org/10.1016/S0965-9978(03)00109-1. Tarefder, R. A., and M. U. Ahmed. 2014. “Modeling of the FWD deflection
Hatzigeorgiou, G. D., and D. E. Beskos. 2010. “Soil–structure interaction ef- basin to evaluate airport pavements.” Int. J. Geomech. 14 (2): 205–213.
fects on seismic inelastic analysis of 3-D tunnels.” Soil Dyn. Earthquake https://doi.org/10.1061/(ASCE)GM.1943-5622.0000305.
Eng. 30 (9): 851–861. https://doi.org/10.1016/j.soildyn.2010.03.010. Tautou, R., B. Picoux, and C. Petit. 2017. “Temperature influence in a dy-
Hernandez, J. A., A. Gamez, and I. L. Al-Qadi. 2016. “Effect of wide-base namic viscoelastic modeling of a pavement structure.” J. Transp. Eng.
tires on nationwide flexible pavement systems: Numerical modeling.” Part B Pavements 143 (3): 04017012. https://doi.org/10.1061/JPEODX
Transp. Res. Rec. J. Transp. Res. Board 2590 (1): 104–112. https://doi .0000013.
.org/10.3141/2590-12. Uddin, W. 2003. “Effects of FWD load-time history on dynamic re-
Hibbit, Karlsson, and Sorensen. 2007. Abaqus/Standard analysis user’s sponse analysis of asphalt pavement.” In Proc., Pavement Performance
manual. Providence, RI: Hibbit, Karlsson, Sorensen. Data Analysis Forum in Conjunction with MAIREPAV03. Guimaraes,
Howard, I. L., and K. A. Warren. 2009. “Finite-element modeling of Portugal: TRB Data Analysis Working Group.
instrumented flexible pavements under stationary transient loading.” Wang, G., R. Roque, and D. Morian. 2011. “Evaluation of near-surface
J. Transp. Eng. 135 (2): 53–61. https://doi.org/10.1061/(ASCE)0733 stress states in asphalt concrete pavement three-dimensional tire–
-947X(2009)135:2(53). pavement contact model.” Transp. Res. Rec. J. Transp. Res. Board
Hu, X., A. N. M. Faruk, J. Zhang, M. I. Souliman, and L. F. Walubita. 2017. 2227 (1): 119–128. https://doi.org/10.3141/2227-13.
“Effects of tire inclination (turning traffic) and dynamic loading on the Williams, M. L., R. F. Landel, and J. D. Ferry. 1955. “The temperature
pavement stress–strain responses using 3-D finite element modeling.” dependence of relaxation mechanisms in amorphous polymers and
Int. J. Pavement Res. Technol. 10 (4): 304–314. other glass-forming liquids.” J. Am. Chem. Soc. 77 (14): 3701–3707.
Huang, Y. H. 2003. Pavement analysis and design: United States edition. https://doi.org/10.1021/ja01619a008.
Englewood Cliffs, NJ: Prentice-Hall. Yoder, E. J., and M. W. Witczak. 1975. Principles of pavement design.
Kim, M., E. Tutumluer, and J. Kwon. 2009. “Nonlinear pavement founda- New York: Wiley.
tion modeling for three-dimensional finite-element analysis of flexible Yoo, P. J. and I. L. Al-Qadi. 2008. “The truth and myth of fatigue cracking
pavements.” Int. J. Geomech. 9 (5): 195–208. https://doi.org/10.1061 potential in hot-mix asphalt: Numerical analysis and validation.” In
/(ASCE)1532-3641(2009)9:5(195). Vol. 77 of Proc., Asphalt Paving Technology: Association of Asphalt
Lee, I.-W., D.-O. Kim, and G.-H. Jung. 1999. “Natural frequency and mode Paving Technologists-Proceedings of the Technical Sessions, 549–590.
shape sensitivities of damped systems: Part I, distinct natural frequen- Lino Lakes, MN: Association of Asphalt Paving Technologist.
cies.” J. Sound Vib. 223 (3): 399–412. https://doi.org/10.1006/jsvi.1998 Zhong, X. G., X. Zeng, and J. G. Rose. 2002. “Shear modulus and damping
.2129. ratio of rubber-modified asphalt mixes and unsaturated subgrade soils.”
Li, M., H. Wang, G. Xu, and P. Xie. 2017. “Finite element modeling and J. Mater. Civ. Eng. 14 (6): 496–502. https://doi.org/10.1061/(ASCE)
parametric analysis of viscoelastic and nonlinear pavement responses 0899-1561(2002)14:6(496).

© ASCE 04020006-19 Int. J. Geomech.

Int. J. Geomech., 2020, 20(3): 04020006

You might also like