Construction and Building Materials 247 (2020) 118616

Contents lists available at ScienceDirect

Construction and Building Materials

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

Characterization of the adhesive and cohesive moisture damage for

asphalt concrete
Mona Nobakht a,⇑, Derun Zhang b, Maryam S. Sakhaeifar a, Robert L. Lytton a
a
Zachry Department of Civil Engineering, Texas A&M University, 3135 TAMU, College Station, TX 77843-3135, USA
b
Texas A&M Engineering Experiment Station (TEES), Texas A&M University System, 7607 Eastmark Drive, College Station, TX 77840, USA

h i g h l i g h t s

An adhesive damage model is developed to predict the amount of moisture damage in the aggregate-asphalt interface.

A cohesive damage model is formulated to account for softening of the FAM due to the diffused water.

The models are developed based on intermolecular bond energy/force, and the mechanisms of cohesive and adhesive failures.

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

Article history: Durability of asphalt concrete is significantly influenced by moisture damage, resulting in early degrada-
Received 4 July 2019 tion of asphalt pavements. There are several tests and conditioning methods adopted by agencies and
Received in revised form 14 February 2020 researchers to identify the susceptibility of asphalt mixtures to the moisture damage. However, there
Accepted 27 February 2020
is a need to develop predictive models that account for the impacts of moisture on fundamental mech-
anistic properties of asphalt concrete. In this regard, the focus of this paper is to develop an adhesive and
a cohesive moisture damage model that enable the prediction of the amount of induced damages due to
Keywords:
the moisture diffusion in the asphalt concrete. The models are developed based on intermolecular bond
Adhesive and cohesive damage
Water vapor diffusion
energy/force, and the mechanisms of cohesive and adhesive failures. They assume water vapor diffusion
Moisture damage as the dominant moisture transport mode in asphalt pavements. The models are calibrated and validated
Asphalt concrete by conducting dynamic modulus test on Fine Asphalt Matrix mixtures as well as Bitumen Bond Strength
test on the adhesive bond between aggregate and asphalt binder. The results of validation process show
the competence of the proposed models in predicting the adhesive and cohesive moisture damage of
asphalt mixtures.
Ó 2020 Elsevier Ltd. All rights reserved.

1. Introduction Durability of AC is mainly affected by environmental factors

Asphalt Concrete (AC) is a composite material that consists of
three distinct phases; namely coarse aggregate, Fine Asphalt
Matrix (FAM) and air void. Coarse aggregate has the main contribu-
tion to compressive strength of the mixture, while FAM binds
coarse aggregate and supports tensile strength of the mixture.
The mechanical characteristics of AC depend on the properties of
FAM and aggregates as well as the strength of interfacial bond
between them. In other word, the degree to which a load is trans-
ferred from FAM to aggregate is controlled by the strength of the
interfacial bond.
⇑ Corresponding author.
E-mail addresses: mona.nobakht@tamu.edu (M. Nobakht), derunzhang@tamu.
edu (D. Zhang), msakhaeifar@tamu.edu (M.S. Sakhaeifar), r-lytton@civil.tamu.edu
(R.L. Lytton).
such as aging and moisture transport within AC pavements [1–
3]. The combined effect of these factors and traffic loading results
in the formation of various types of distresses such as pothole,
stripping, fatigue cracking, and so on, resulting in the reduction
of in-service life of pavements [4,5]. Moisture damage commences
when water – in the form of liquid or vapor – transports into the
AC pavements through three main mechanisms; namely perme-
ability, capillary rise and diffusion [6]. Water transport modes
are controlled by components of mixture such as air void sizes
and their connectivity, diffusivity of water molecules in the mix-
ture and asphalt binder, aggregate absorption capacity and so on.
It has been recognized that water vapor diffusion is constantly pre-
sent in the pavement and so it can be considered as one of the con-
tributors to the moisture damage. [7]. This vapor diffusion occurs
because of the Relative Humidity (RH) differential that exists
between atmosphere above the asphalt layer and the subgrade

https://doi.org/10.1016/j.conbuildmat.2020.118616
0950-0618/Ó 2020 Elsevier Ltd. All rights reserved.
2 M. Nobakht et al. / Construction and Building Materials 247 (2020) 118616
below the pavement structure. RH of the atmosphere changes with
location and weather. However, RH in the subgrade is always near
100%. Thus, it serves as a reservoir beneath the pavement, resulting
in a diffusive motion of water vapor toward the pavement. Several
research studies have reported the moisture damage in pavements
located in the region with desert/semiarid climate such as Arizona
[8,9]. This further implies the pivotal role of subsurface vapor dif-
fusion in formation of the moisture damage. Therefore, water in
the form of vapor is critical in impacting the properties and dura-
bility of asphalt mixtures. Along this line of research, Rueda et al.
experimentally evaluated the mechanical response of asphalt mix-
tures under different relative humidity environments and observed
that RH has a substantial influence on the linear viscoelastic prop-
erties of asphalt mixtures [10,11]. In another study, Caro et al.
studied the effects of mechanical and physical properties on the
response of asphalt mixtures subjected to water vapor diffusion
by using a finite element micromechanical model. They observed
that the diffusion coefficient of the asphalt mixture and aggregate
as well as the bond strength of the aggregate-matrix interface have
the most impact on the moisture susceptibility of mixtures [12].
The amount of water vapor diffused into the asphalt binder is
dependent on the water holding capacity of the binder, diffusivity
of water vapor into the asphalt binder, and RH. The two former fac-
tors are controlled by the thermodynamic and chemical character-
istic of binder, while the latter depends on the environmental
condition. Kassem et al. [13] measured the diffusivity of water
vapor in three mastics by measuring the rate of change in suction
with time using the thermocouple psychrometers. They found a
correlation between diffusion coefficients and moisture damage
observed in the field for asphalt pavements corresponding to the
mastics; a high diffusion coefficient was associated with a poor
moisture resistant pavement. In another study, Cheng et al. [14]
demonstrated that the mixtures composed of asphalt binders with
high water holding capacity are more susceptible to moisture
damage.
Moisture damage in asphalt concrete is defined as degradation
of mechanical properties due to the presence of moisture [6]. It is a
progressive deterioration which happens when water deteriorates
the adhesive bond between the aggregate surface and asphalt bin-
der, referred to as the adhesive damage, and the cohesive resistance
within the asphalt binder, known as the cohesive damage. Three
main interactions are involved in the adhesive and cohesive bond
[15]. The first interaction is related to electrostatic forces between
the ions in aggregate and binder, resulting from the Coulomb
forces. The second one corresponds to electrodynamic forces
between molecules, which results from Van der Waals intermolec-
ular bonding. Combination of first and second interactions,
referred to as Physico-chemical interaction, results in a thermody-
namic property of the material known as the surface free energy
[6]. The third interaction pertains to the covalent bond, resulting
from creation of a new compound due to the reaction of asphalt
binder with mineral aggregate. The energy that is required to form
an adhesive and cohesive bond are calculated from the surface free
energy of each material involved in the bond, that is, surface free
energy of aggregate and asphalt binder for the adhesive bond
and that of asphalt binder for the cohesive bond. When water
transports or diffuses into the asphalt mixture, water molecules
are absorbed by the aggregate and asphalt binder surface. This
results in a reduction in their surface free energy, and consequently
degradation in adhesive and cohesive bond strength.
With regard to the characterization of moisture damage, exten-
sive analytical and experimental efforts have been made to quan-
tify the moisture damage based on the thermodynamics and
mechanics. The quantification of moisture damage in most of the
studies in literature includes three steps: 1) designing a moisture
conditioning method, 2) defining a performance measure obtained
by conducting a performance test, and 3) calculating the damage
factor as the ratio of performance measured in dry and wet condi-
tions. AASHTO T283, Arambula et al., Copeland et al., Lytton et al.,
Song et al. [16–20] are several examples of the work conducted in
this area. For more information on the abovementioned approach,
refer to the review paper by Caro et al. [6]. Such a quantification
method provides a valuable tool that can be used in mix design
specifications to evaluate the moisture susceptibility of mixtures.
A threshold value can be assigned to the damage ratio to separate
moisture resistant mixtures from those susceptible to the mois-
ture. A recent survey shows 80% of State Department of Trans-
portation conduct some kind of testing related to moisture
damage resistance. However, there is a need for a performance pre-
diction model that can be used to characterize the moisture dam-
age in a performance-based framework [21].
The objective of this study is to develop an adhesive and a cohe-
sive moisture damage model that enable the quantification of the
detrimental effect of water vapor on adhesive and cohesive
strength of AC, respectively. The models are developed by provid-
ing links to the underlying causes of deterioration, and considering
the interactions among chemical and physical properties influ-
enced by moisture. The models assume water vapor diffusion,
which is the dominant moisture transport mode in asphalt
pavements.
This paper is structured in six sections. The current section pro-
vides a background about the moisture damage and a summary of
research studies conducted on this field, and also explains the
statement of need, research objectives and outline of the paper.
Section 2 and 3 focus on the formulation of the cohesive and adhe-
sive damage model based on two separate mechanisms involved in
the moisture damage. Test procedures and materials that are used
for calibration of the models are described in Section 4. The models
are then calibrated and validated in the following sections. Finally,
the last section concludes this study and presents some recom-
mendations for future work.
2. Adhesive damage model
Adhesive bond between aggregate and asphalt binder deterio-
rates as water enters the interface as shown in Fig. 1. Without
the presence of water, adhesive bond energy is characterized by
Eq. (1) [22], where DG indicates the difference between the final
and initial states of energy in an aggregate-asphalt bonding
system:
qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi qffiffiffiffiffiffiffiffiffiffiffi qffiffiffiffiffiffiffiffiffiffiffi
DGSA ¼ cS þ cA  cSA ¼ 2ð cLWA cS þ
LW
cþA cS þ cA cþS Þ ð1Þ
where,
DGSA = Adhesive bond energy;
Water film
Water vapor
FAM
Asphalt
Binder
Aggregate
Air Voids
Water Diffusion
Fig. 1. Adhesive bond deterioration due to water diffusion.
 M. Nobakht et al. / Construction and Building Materials 247 (2020) 118616 3

cLW
A ; cA ; cA = Nonpolar, polar-acid and polar-base surface free
þ 
component of asphalt binder, respectively;
cLW
S ; cS ; cS = Nonpolar, polar-acid and polar-base surface free
þ 
component of aggregates, respectively;
cA ; cS ; cSA = Total surface free energy of asphalt binder, total sur-
face free energy of aggregate and interface energy between aggre-
gate and asphalt binder, respectively.
The surface free energy of asphalt binder and aggregate are typ-
ically estimated using the Wilhelmy Plate and Universal Sorption
test, respectively. When water vapor is present in the interface,
water molecules are absorbed by the aggregate. As a result, the
surface free energy of aggregate is reduced by the amount of
spreading pressure as shown in Eq. (2) [23]:
cSV ¼ cS  pe
where,
cSV = Surface energy of the aggregate in equilibrium with the
water vapor, ergs/cm2;
pe = Spreading pressure of aggregate due to adsorption of water
vapor, ergs/cm2.
Note that the subscript V in Eq. (2) denotes the presence of
water vapor at the aggregate-asphalt interface.
pe is defined as the reduction of surface energy of the aggregate
due to the absorption of water vapor, and can be calculated using
Gibbs adsorption model as illustrated in Eq. (3) [24]:
ZP0
RT n shown in Eq. (8). Note this equation will be proposed in this study
pe ¼ dP
mH2 O A p
0 aggregate-asphalt combinations tested.
where,
R = Universal gas constant, 8.314 J/Kmol;
T = Test temperature, K;
mH2 O = Molecular weight of water, 18.015 gr/mol;
A = Specific surface area of aggregate, m2/gr;
n = Amount of vapor absorbed on the aggregate, gr;
P0 = Saturated water pressure at the test temperature, Pa;
P = Water vapor pressure, Pa.
Changing the upper limit of integral to a specific water vapor P1,
Eq. (4) is proposed in this study to determine pe at the correspond-
ing relative humidity, RH ¼ pp1 :
0
A series of surface energy measurement tests were performed
in this study to further evaluate the relation shown in Eq. (6). Wil-
helmy Plate tests were conducted on two typical asphalt binders
denoted as unmodified #70 asphalt and #70 asphalt modified with
SBS modifier to estimate their surface free energy. Also, the Univer-
sal Sorption test was performed on four different aggregate types
including basalt, quartz sandstone, limestone and gravel with sili-
con dioxide (SiO2) content of 46.17, 90.55, 15.74, and 87.5%,
respectively, to measure their surface free energy as well as their
specific surface area. Eq. (6) was then used to calculate the adhe-
sive bond energy between aggregate and asphalt binder at various
RHs. The fracture work, W SVA , defined as the product of adhesive
bond energy DGSVA and specific surface area A, is subsequently
computed. The computed fracture work versus RH is shown in
ð2Þ Fig. 2 for both unmodified and modified binder. The results suggest
that the relationship between W SVA and RH can be characterized
using an exponential function to fit the experimental data:
W SVA ¼ DGSVA  A ¼ b1  a1 expðRHc1 Þ ð7Þ
where, a1, b1 and c1 are fitting constants specific for an aggregate-
asphalt combination. R2 for each asphalt-aggregate combination is
greater than 0.99, which validates the rationality of the suggested
model for characterizing the relationship between W SVA and RH.
Since the specific surface area is a constant for a given aggregate,
it can be inferred that the relation between adhesive bond energy
and RH can also be expressed through an exponential function, as
ð3Þ
to characterize the adhesive bond energy with respect to RH for
DGSVA ¼ b  aexpðRHc Þ ð8Þ
DGSA in Eq. (6) is a thermodynamic property that is affected by
the amount of water vapor accumulated in the interface. Thus, the
Adhesive Damage (AD) model for AC can be formulated by identi-
fying the underlying link between this thermodynamic property
and tensile adhesive bond strength. Tensile strength of an adhesive
bond can be determined from Eq. (9) developed by Lytton [28]:
(a) 900
800 Basalt
WSVA ×10^4 (ergs/g)

ZP1 Z 0
RHP
700
Quartz Sandstone
RT n RT n
peRH ¼ dP ¼ dP ð4Þ 600
Limestone
Siliceous River Gravel
mH2 O A P mH 2 O A P
0 0 500
400
The components of Eq. (4) are determined by conducting the
300
Universal Sorption Device (USD) test on aggregates at various val-
200
ues of RH. The adhesive bond energy in the presence of water vapor
100
can be estimated from Eq. (1) by replacing the surface free energy
0
of dry aggregate and asphalt with that of wet aggregate and binder, 0 20 40 60 80 100
which yields the following equation: RH (%)
DGSVA ¼ cSV þ cAV  cSA ð5Þ (b) 900
800 Basalt
where, DGSVA is defined as the adhesive bond energy of asphalt-
WSVA ×10^4 (ergs/g)

700 Quartz Sandstone

aggregate in the presence of water vapor. The spreading pressure Limestone
of asphalt binder can be negligible since binder is a low-energy 600 Siliceous River Gravel
material with respective to water [25–27]. Therefore, it is reason- 500
ably assumed in this study that cAV equalscA , implying that surface 400
free energy of asphalt binder is not reduced notably in the 300
presence of water vapor. Combining Eqs. (2) and (4) with (5), 200
the Eq. (6) is proposed in this study to estimate DGSVA at each 100
RH of interest: 0
0 20 40 60 80 100
Z 0
RHP RH (%)
RT n
DGSVA ¼ DGSA  dp ð6Þ
mA p Fig. 2. Fracture work between aggregate and a) unmodified binder and b) modified
0 binder.
4 M. Nobakht et al. / Construction and Building Materials 247 (2020) 118616

" #0:25
8 m  1 fere with the intermolecular bonding as opposed to what occurs in
rt ¼ ð Þ  ð c Þ  EA
ðEA DGSA Þ0:5 ð9Þ polymers. Therefore, Young modulus of the wet aggregate is
3t A 1þ ES approximately the same as that of the dry aggregate. The above-
Where, mentioned explanation implies the following expressions:
t = Binder thickness, m;
Ewet
S ’ Edry
S
m = Number of cracks;
A = Cross section area of the binder, m2; Edry dry
S  EA ð13Þ
 dry dry
Ewet E þE
c = Mean crack size, m; S
 Ewet
S A
’1
dry
ES þEwet
EA ; ES = Relaxation modulus and Young modulus of asphalt and S A

aggregate, respectively, Pa; Applying Eqs. (7) and (13) and into (12) yields:
DGSA = Adhesive bond energy between asphalt and aggregate, " #
0:5 0:5
ergs/m2. Ewet b  aexpðRHc Þ
dAD ¼ 1  ð Adry Þ  ð Þ ð14Þ
Eq. (9) is developed based on micromechanical and fracture EA ba
analyses of adhesive damage in the asphalt-aggregate system. Lyt-
ton [28] related the adhesive bond strength to thermodynamic and Using the formula of ideal gas low, RH can be written as the
microstructural properties of the materials involved in the bond. function of water vapor concentration as demonstrated in Eq. (15):
The equation was derived using the energy balance concept, which
P nRT RTM RT
states that the apparent strain energy within the bulk specimen RH ¼ ¼ ¼ ¼ Cm ð15Þ
P0 P0 V P0 mH2 O V P0 mH2 O
(i.e., cracked and intact sections combined) equals the sum of the
true strain energy within the intact materials and surface energy where, Cm is the water vapor concentration at the aggregate-asphalt
absorbed within the crack surface. When adhesive failure happens interface, and its value is controlled by the diffusivity of water vapor
in the aggregate-asphalt system subjected to a tensile force, part of in AC and diffusion time. The final form of AD model developed in
the stored strain energy is released due to the newly created this paper is illustrated in Eq. (16).
cracked surface. On the other hand, the surface energy increases 2 3
c 0:5
given the creation of new free surfaces (i.e., cracks). The equation wet 0:5 b  aexpðfP0 mRTH O C m g Þ
6 EA 7
was derived through establishing the energy balance for all kinds dAD ¼ 1  4ð dry Þ ð 2
Þ 5 ð16Þ
EA ba
of energies involved in the bond and also through considering
the viscoelastic and thermodynamic properties of the aggregate-
asphalt system.
The Adhesive Damage (AD) model is formulated in this study as 3. Cohesive damage model
follows to account for the deterioration effect of water vapor in the
adhesive bond: Asphalt is a thermoplastic material that is composed of polar
and non-polar hydrocarbon chains, aromatic ring and condensed
rwet
dAD ¼ 1  t
ð10Þ ring molecules. The bonding force holding together the molecules
rdry
t is the van der Waals force [31], which is recognized as a weak sec-
ondary bonding force. The high molecular weight and long chains
where, rwet
t denotes the tensile strength of the adhesive bond
cause an entanglement of asphalt molecules, as shown schemati-
between binder and aggregate in the presence of water vapor.
cally in Fig. 3. The lack of strong intermolecular bonds eases the
Replacing dry properties with wet properties rwet
t can be obtained
mobility of chains. Thus, the entangled mobile chains result in vis-
as:
coelastic properties of the asphalt binders and FAMs.
2 30:25 The Lennard-Jones (LJ) potential is used in this study to model
m wet 1 7  wet AF 0:5
6 8 the interaction between asphalt binder molecules in wet and dry
rwet
t ¼ 4ð Þ  c  Ewet
5 EA DGSVA ð11Þ
3t A 1þ A condition. The LJ potential describes the potential energy of inter-
Ewet
S action between two non-bonding atoms or molecules based on
 their distance of separation. That is, the potential energy between
Lytton [28] showed that mA c value is constant for a certain mate-
rial. This value represents the amount of cracks developed in the
material when tensile stress reaches the maximum value in the

stress–strain curve. Thus, mA c value can be assumed to be identical
in dry and wet condition given that the same aggregate-asphalt rm
materials are involved in the bond in both dry and wet condition.
Substituting Eqs. (9) and (11) and in Eq. (10), and crossing out the
similar terms from numerator and denominator give the following
(a)
expression for AD model:
" 0:25 0:5
#
Ewet Edry dry
S þ EA Ewet DGSVA 0:5
dAD ¼ 1  ð S
 wet Þ ð A
Þ ð Þ ð12Þ
Edry
S
ES þ Ewet
A Edry
A
DGSA

Young modulus of aggregates is significantly larger than that of rw

asphalt binder even at low temperature and high loading fre-
quency [29,30]. Also, the effect of absorbed water on changing
Young modulus of aggregates is negligible. Aggregates are catego-
rized as the refractory ceramic material, which are often porous (b)
and able to absorb water through connected pores. Given the
strong interatomic bonding between atoms and molecules in cera- Fig. 3. Schematic illustration of high molecular chain of asphalt in (a) dry and (b)
mic materials, the absorbed water vapor molecules could not inter- wet conditions.
 M. Nobakht et al. / Construction and Building Materials 247 (2020) 118616 5

2
molecules varies with their intermolecular distance. The potential d V dF 156A 42B
equation, shown in Eq. (17) and illustrated schematically in Fig. 4, E/ ¼ ¼ 14  8 ð23Þ
dr 2 dr r r
accounts for the difference between attractive forces (i.e., dipole–
When water vapor transports into the asphalt, small molecules
dipole, dipole-induced dipole, and London interactions) and repul-
of H2O diffuse between the molecular chains and occupy positions
sive forces:
 12    among them. Therefore, the molecules are pushed away and the
g g6 rm 12 rm 6 distance between them increases. The separation of chains results
V LJ ¼ 4e ð Þ  ð Þ ¼ e ð Þ  2ð Þ ð17Þ
r r r r in an increase in r value in Eqs. (22) and (23). Consequently, the
intermolecular bonding force and modulus of elasticity diminish
g  0:89rm ð18Þ and the asphalt becomes softer. The schematic picture of this
mechanism is presented in Fig. 3(b). This change in the molecular
Where,
structure of asphalt is the main cause of cohesive failure occurring
V LJ = Lennard-Jones Potential;
in FAM given the moisture diffusion. The Cohesive Damage (CD)
e = Depth of the potential well;
model is formulated to account for the deterioration effect of mois-
g = The finite distance at which the potential is zero;
ture on modulus of FAM as follows:
rm = Distance at which the potential reaches its minimum;
r = Distance between the molecules. Ewet P  ðE0 wet Þ þ ð1  PÞE0 dry
The r12 term is the repulsive potential describing the repulsion dCD ¼ 1  ¼1 ð24Þ
Edry E0 dry
between molecules due to the overlap of electron orbits, and r6 is
related to the attraction between molecules. A simplified formula- Where,
tion for LJ potential can be shown as follows: Ewet = Overall modulus of wet sample
Edry = Overall modulus of dry sample
A B
V LJ ¼  ð19Þ E0wet = Average modulus in wet state based on molecular
r 12 r6
separation;
Where, E0dry = Average modulus in dry state based on molecular
A ¼ 4eg12 ð20Þ separation;
P = Percentage of molecules affected by moisture.
B ¼ 4eg6 ð21Þ E0wet and E0dry are defined based on Eq. (23), assuming that the ini-
tial average molecular distance r m is changed to r w after water
The first derivative of the LJ potential with respect to the dis-
molecules diffuse within chains:
tance gives an expression for the intermolecular force [32]:
dV 12A 6B 156A 42B
F¼ ¼ 13 þ 7 ð22Þ E0dry ¼  8 ð25Þ
dr r r r 14
m rm
Also, the modulus of elasticity is a measure of resistance to
detachment of the adjacent molecules, and is proportional to slope 156A 42B
E0wet ¼  8 ð26Þ
of the force-distance curve at r m [32]: r 14
w rw
Substituting Eqs. (25) and (26) into Eq. (24) gives the following
Equation for CD model:
 
rm 8 1 þ 3:7ðA=BÞ  r 6
dCD ¼ P 1  ð Þ  ð w
Þ ð27Þ
rw 1 þ 3:7ðA=BÞ  r m
6
Intermolecular Potential (V)

When Eqs. (20) and (21) are substituted in Eq. (27), the CD
model can be rewritten as:
η "  8  14 #
rm rm
dCD ¼ P 1 þ 1:17  2:17 ð28Þ
rw rw
ε r
The next step in development of the CD model involves identi-
rm fying the functions that properly describe the changes of P and rw
as a function of water content. The more amount of water is
absorbed, the more molecules are affected and the more separation
happens within the molecular chains. Therefore, P and r w functions
should be increasing with respect to the water content. The rate of
water diffusion in FAM is fast at the beginning of process, and grad-
ually decreases with time until it becomes close to zero. As diffu-
Intermolecular Force

r P and rw functions should have asymptotes so as to illustrate the
Fig. 4. Schematic illustration of Lennard-Jones potential and intermolecular force.
sion is the main reason for changes inPand r w , the same trend as
that observed for diffusion rate should be seen for Pand r w . This
further implies that these functions should be concave. In addition,
equilibrium state that happens at the end of diffusion. Three func-
tion types satisfying the abovementioned criteria include logarith-
mic, negative exponential and monomial with the exponent
between 0 and 1. All three functions are examined in the study
and fitted to the cohesive damage measured in the lab in order
to identify the best matching functions. The cohesive damage is
6 M. Nobakht et al. / Construction and Building Materials 247 (2020) 118616

measured in the lab by following the procedure explained in the

next sections. Eqs. (29) and (30) show the functions that gave
the best fit, and so are selected in this study to model Pand r w :
P ¼ 1  expða0 wÞ ð29Þ

rm rm rm 1
¼ ¼  ¼ ð30Þ
r w r m þ d0  wc0 rm 1 þ d0 wc0 0
1 þ b wc0
rm

Where,
w = Gravimetric water content, %;
0
a0 ; b ; c0 = Fitting parameters.
The ultimate function for CD model is obtained by substituting
Eqs. (29) and (30) and into Eq. (28) as follows:
" #
1:17 2:17
dCD ¼ ½1  expða0 wÞ  1 þ 0
8  0
14 ð31Þ
1 þ b wc0 1 þ b wc0
For a viscoelastic material such as FAM, the effect of absorbed
water in degradation is affected by temperature and loading fre-
quency. In general, increasing the temperature and decreasing
the loading frequency result in an aggravation of cohesive deterio-
ration. Cohesive damage occurs when water deteriorates the cohe-
sive bond within the FAM molecules. When temperature is raised,
the molecular motion of water molecules increases, and they gain
more energy to force apart macromolecules of FAM, resulting in a
increase in the intermolecular distance. According to the LJ poten-
tial formula, this results in more reduction in the secondary inter-
molecular bonding forces and energies, consequently more
reduction in the cohesive bond. On the other hand, when loading
time increases (which is equivalent to decreasing the frequency),
in each loading cycle water molecules have more time to push
apart macromolecules of the FAM and exacerbate the chain separa-
tion, resulting in more cohesive damage. This implies that fitting
0 Depending on the asphalt thickness, testing temperature, rheol-
parameters involved in representing rw in Eq. (30) (i.e., b and c0 )
are functions of loading frequency and temperature. This conclu-
sion is further examined in Section 5.3. In order to calibrate CD
model, a set of laboratory testing is conducted on FAM samples,
which is explained in the next section.
4. Materials fabrication and test procedure
4.1. Bituminous bond strength test
Parameters of the AD model are determined by conducting
Bituminous Bond Strength (BBS) test in accordance with AASHTO
TP-91 [33] on moisture conditioned and dry specimens. As men-
tioned in Section 2, the adhesive damage model is formulated
based on the adhesive strength between aggregate and binder in
wet and dry condition, where tensile strength values appear in
both nominator and denominator of the model (i.e., Eq. (10)).
According to this formulation, some variables involved in the rt
formula (i.e., Eq. (9)) (e.g., thickness) are crossed out from nomina-
tor and denominator. Therefore, it can be assumed that the impacts
of variables such as binder thickness and loading rate in changing
the adhesive damage are negligible given that they affect both rwet t
and rdry
t in a similar way. In other words, while adhesive tensile
strength is dependent on variables such as binder thickness, the
adhesive damage can be assumed to be invariant to these variables.
Therefore, the BBS test can be used to determine the tensile bond
strength between aggregate and asphalt binder and validate the
accuracy of the AD model in predicting the adhesive damage due
to moisture diffusion.
As shown in Fig. 5, the BBS device is composed of a pressure
hose, pneumatic adhesion device, piston, reaction plate and metal
pull-out stubs. A pull-off force is applied on the test specimen by
Fig. 5. BBS test components.
the metal stubs. When applied stress exceeds the adhesive
strength between asphalt and aggregate or cohesive strength
within the binder, the failure occurs. The Pull-off Tensile Strength
(POTS) is determined from Eq. (32):
ðBP  Ag Þ  C
POTS ¼ ð32Þ
Aps
Where,
C = Piston constant;
BP = Burst pressure (kPa);
Ag = Contact area of piston with reaction plate (mm2);
Aps = Area of the pull-stub (mm2).
ogy of the asphalt, rate of loading and water presence in the inter-
face, type of failure can be either adhesive, cohesive or
combination of both. It has been recognized that the presence of
moisture at the interface is one of the main causes for adhesive
failure [34,35]. Therefore, it can be assumed that moisture condi-
tioned specimens tested in this study experience the adhesive fail-
ure regardless of the amount of other abovementioned variables.
The AD model developed in the previous section is aimed to char-
acterize the bond strength with respect to the concentration of
water vapor at the interface. Thus, a test set-up was designed such
that the moisture concentration gradient at the interface between
aggregate and binder change with time. The test set-up is shown in
Fig. 6. In order to properly simulate water vapor transport as it
happens within the mixture, the substrate surface was prepared
by cutting thin slices of 2 cm from a S3 PG 64-22 compacted
asphalt mixture and asphalt binder was adhered to the area of
the AC slice occupied by large aggregate as shown in Fig. 6.
The S3 PG 64-22 mixture is a 19 mm Superpave mixture com-
posed of 85% limestone aggregate (30% #67 Rock, 10% 1/200 Chips,
45% Sand) and 15% fine RAP, and its binder content is 4.4% by
weight. Table 1 illustrates the volumetric properties of the sub-
strate. After cutting, substrates slices were submerged in distilled
water at 25 °C to remove any surface residue from the cutting pro-
cess. They were then kept at the room temperature in front of a fan
for two days to remove the absorbed water. Then, acetone was
applied on the aggregate surface exposed to air, and metal stubs
to remove dust. The metal stub has a diameter of 12.7 mm, and a
0.8 mm thickness surrounding edge. Four cuts exist along the stub
edge to allow the flow of excess asphalt binder out as the stub is
pressed on the aggregate surface.
PG 64-22 binder was used as an adhesive material to attach
metal stub to the aggregate surface (Fig. 7). The substrate slice,
 M. Nobakht et al. / Construction and Building Materials 247 (2020) 118616 7

Cellophane
HMA

Pull-off stub
Asphalt Binder
X
Aggregate
Substrate Slice

Water Diffusion

Fig. 6. BBS test sample set-up.

Table 1 moisture conditioning levels, which were defined based on the

Volumetric Properties of S3 PG 64-22 AC Slice. varying time exposed to moisture. Specimens were covered with
Gradation cellophane to prevent water vapor from evaporating from the sur-
Sieve Size Percent Passing (%)
face. This provides the required boundary condition for solving a
one-dimensional diffusion equation. The diffusion equation was
19 mm 100
12.5 mm 90
used to estimate the water vapor concentration at the interface
9.5 mm 79 as a function of time.
4.75 mm 57 In total 33 specimens were prepared; 15 test specimens were
2.36 mm 41 placed in the moisture room at 100% RH and 35 °C temperature
1.18 mm 28
to prepare the wet specimens. 15 conditioning times of 1, 3, 12,
0.60 mm 19
0.30 mm 12 24, 32, 48, 72, 96, 120, 144, 216, 312, 384, 432, 500 and 700 h were
0.15 mm 7 considered for conditioning specimens. When conditioning time
0.075 mm 5.1 was over, specimens were taken out and cellophane was removed
Volumetric from the top surface and surrounding edges. The BBS test was then
NMAS1 19 mm conducted on wet samples at the loading rate of 0.82 MPa/s (120
RAP2 (%) 15
psi/s) to measure the rTwet data that are needed to calculatedAD ,
Air Void (%) 7.1
VMA3 (%) 15.9 according to Eq. (9).
VFA4 (%) 55 Furthermore, 15 specimens were put in a chamber at 35 °C to
Binder Content (%) 4.4 prepare the dry specimens; the results of BBS test on dry speci-
Dimensions mens provide the rTdry data that are needed to calculatedAD . Dry
Diameter 150 mm specimens were subjected to the same conditioning time as those
Thickness 20 mm
considered for conditioning wet specimens. When conditioning
1
Nominal Maximum Aggregate Size. time was completed, dry specimen was removed from the chamber
2

3
Reclaimed Asphalt Pavement.
Voids in Mineral Aggregates.
4
Voids Filled with Asphalt.
stubs and asphalt binder were placed in the oven at 65, 85 and
164 °C, respectively, for at least 30 min. Asphalt binder was then
placed immediately on the surface of stubs, and the stubs contain-
ing asphalt were pressed firmly against the aggregate surface with-
out applying the torsion until no excess asphalt binder flows. The
excess asphalt was then scrubbed from the aggregate surface to
ensure that adhesive measurements solely reflect the aggregate-
asphalt bond strength. Four stubs were pressed on each slice,
and each slice was considered as a test specimen containing 4 sam-
ples for bond strength measurements. As shown in Fig. 7, the top
surface and surrounding edges of substrate were covered firmly
with cellophane. Test specimens were then subjected to predefined
and tested. For both wet and dry specimens, four rT measurements
were recorded for each conditioning time since four stubs were
pressed on each slice. Also, one specimen was tested as a control
specimen, and two were prepared to determine the diffusion coef-
ficient and maximum water uptake of the test specimens. The sat-
urated surface dry (SSD) weight of these two specimens was
measured at various time intervals for one month to calculate
the diffusion coefficient. Diffusion coefficient was then employed
in a close form solution to estimate the water vapor concentration
at the interface of aggregate and asphalt binder at different times.
4.2. Dynamic modulus test
The parameters of the CD model are determined by conducting
the dynamic modulus test on FAM specimens. FAM specimens
were composed of a PG 64-22 asphalt binder and limestone

Fig. 7. Sample preparation.

8 M. Nobakht et al. / Construction and Building Materials 247 (2020) 118616

aggregates. They were prepared by using fine portion (i.e. aggre- coelastic region. The results of dynamic modulus test on wet spec-
gate portions passing No. 16 sieve) of a full mixture of the S3 PG imens provide the Ewet data that are needed to calculate dCD in Eq.
64-22 mixture. That is, the gradation of FAM follows the gradation (23). Dry samples that were placed in the chamber at 35 °C were
of fine portion in S3 PG 64-22 mixture. The binder content of FAM tested after 2, 4, 12, 20, 32, 40, 60 and 90 days of conditioning time.
was determined by following the approach developed by Sousa Two replicates were prepared for each conditioning time, and |E*|
et al. [36] (2013). His approach is based on separating the fine por- testing was conducted in the same approach as that adopted for
tion of a mixture from the coarse portion, and burning that in the moisture conditioned specimens. The results of dynamic modulus
ignition oven at 427 °C. The measured asphalt content for FAM test on dry specimens provide the Edry data that are needed to cal-
specimens, referred to as S3 PG 64-22 FAM, was measured as culate dCD in Eq. (23).
8.5% by weight of the mixture. Loose FAM mixtures were
conditioned in the oven at 135 °C for two hours to simulate
short-term aging, and then compacted by Superpave Gyratory 5. Results and discussion
Compactor (SGC) in 150 mm diameter and 120 mm height. The
required weight of specimens was calculated such that their air 5.1. Diffusion model
void contents were less than 2%. The final cylindrical specimens
of 38 mm in diameter and 110 mm in height were cored and cut, The AD model is the function of water vapor concentration.
and their volumetric properties were measured following the Therefore, first step toward calibration of the model is to deter-
corresponding AASHTO standards. mine the water vapor concentration at the interface as the function
The number of 34 S3 PG 64-22 FAM specimens was fabricated, of time. Moisture diffusion within the asphalt mixture when placed
and their dry weight was recorded. 18 specimens were placed in in the moisture room is assumed to follow Fick’s second law:
the moisture room at 100% RH and 35 °C temperature while 16
@C @2C
specimens were put in a chamber at 35 °C. 8 conditioning times ¼D 2 ð33Þ
@t @x
of 2, 4, 12, 20, 32, 40, 60 and 90 days, were defined for conditioning
specimens that were placed in the moisture room. Two replicates Where,
were prepared for each conditioning time. When conditioning time C = Water vapor concentration, gr/mm3;
was over, specimens were taken out from the moisture room and t = Time, s;
prepared for dynamic modulus, |E*|, testing. Brass buttons were D = Diffusion coefficient, mm2/s;
glued to the surface of test specimens using a button gluing jig x = Position, mm.
as shown in Fig. 8. Three LVDTs were mounted on the buttons The diffusion process for the test configuration is the case of an
and dynamic modulus test was conducted in the strain-control absorption by a membrane where the initial concentration inside
mode by using AMPT device. In order to keep the moisture content the specimen is zero, and bottom surface is kept at a constant con-
constant during the test and provide the testing environment close centration. The solution for the absorption in the form of a trigono-
to the environment condition, conditioned specimens were tested metrical series is presented as follows [37]:
at 25, 10, 5, 1, 0.1 Hz frequencies and only one temperature of
35 °C. The objective was to ensure that water vapor concentration C ¼ C0
" #
resulting from the conditioning in 100% RH is not disturbed during 4X 1
ð1Þn n o ð2n þ 1Þpx
 1 exp Dð2n þ 1Þ2 p2 t=4L2 cos
the test. Since the test chamber for |E*| testing does not provide RH, p n¼0 2n þ 1 2L
when conditioned specimens are placed into the chamber, concen-
tration gradient between specimens and surrounding causes des- ð34Þ
orption of water from the tested specimens. In order to minimize
where, D is the diffusivity of the membrane. The amount of
this desorption, the test was conducted at the conditioning tem-
absorbed water entered the membrane at time t is determined by
perature (i.e., 35 °C herein). Therefore, specimens could be tested
integrating C values over the entire volume [37]:
immediately after withdrawing from the moisture room and so
" #
water vapor concentration can be assumed to be approximately X
1
8 n o
constant during testing. Peak-to-peak strain level was set to be Mt ¼ M1 1  exp Dð2n þ 1Þ p t=4L
2 2 2
ð35Þ
within 45 and 60 micro-strain to ensure that no damage occurs n¼0 ð2n þ 1Þ2 p2
during the loading, and that FAM samples remain in the linear vis-
where, M1 denotes the maximum water uptake after infinite time.

Fig. 8. (a) Brass buttons glued to FAM specimen; (b) dynamic modulus set-up.
 M. Nobakht et al. / Construction and Building Materials 247 (2020) 118616
As stated in Section 4.1, two samples were prepared and placed
in the moisture room to estimate the diffusion coefficient and max-
imum water uptake of the test specimens. The weight of samples
was recorded every other day for 30 days to estimate the amount
of water uptake. Eq. (35) is then fitted to the measured mass of
absorbed water to backcalculate the diffusion coefficient and max-
imum water mass diffused in thin specimens. Then, Eq. (34) is
employed to estimate C at X = 0, which is denoted by Cm in Eq
(16). Figs. 9 and 10, exhibit the change of Mt and Cm with diffusion
time, respectively.
5.2. Calibration of AD model
Results of the BBS test on dry and moisture conditioned
specimens are presented in Fig. 11. It is observed that as condi-
tioning time increases, the reduction in POTS increases as well,
which is obviously caused by the increased amount of accumu-
lated water vapor concentration at the interface with time.
Fig. 12 shows the images of failure on the surface recorded at
various conditioning times. The first image on the left indicates
the failure surface for the dry specimen; as moving to the right,
conditioning time increases and so a greater amount of water
vapor are diffused into the substrate, resulting in the growth
of moisture damage.
The ratio of the POTS for moisture conditioned specimens to
POTS for dry specimens was determined, and then subtracted from
one; the remainder indicates the measured adhesive damage
induced at the interface. Adhesive damage is plotted versus condi-
tioning time in Fig. 13. Each data point indicates the average adhe-
sive damage introduced into four tested bonds. The Cm
corresponding to each conditioning time is extracted from
Fig. 10, and then adhesive damage is plotted vs Cm in Fig. 14. The
AD model is then fitted to the measured damage to find the fit-
ting parameters described in Eq. (16). Cm can be converted to RH
using Equation to plot the damage against RH as presented in
Fig. 15.
For the specific material tested in this study, a, b and c are found
to be 97.57 ergs/cm2, 304.85 ergs/cm2 and 0.4, respectively. These
parameters account for the specific adhesive bond energy between
the aggregate and the asphalt binder tested. For the test configura-
tion selected in this study, one month of conditioning at 100% RH
and 35 °C is needed to change RH at the interface from 0 to 1. When
RH at interface reaches 1, diffusion process gradually stops given
that concentration gradient along the specimen—which is the driv-
ing force for the diffusion—approaches zero. The time frame over
which bond deterioration happens depends on specimen dimen-
0.6
0.5
Water Uptake, %
0.4
0.3
0.2
0.1
Measured Data
Model Fitting
0.0
0.0E+00 5.0E+05 1.0E+06 1.5E+06 2.0E+06 2.5E+06 3.0E+06
Time, Seconds
Fig. 9. Moisture uptake versus time for S3 64-22 HMA slice.
9
Moisture Concentration (Cm) , gr/mm3
4.5E-08
4.0E-08
3.5E-08
3.0E-08
2.5E-08
2.0E-08
1.5E-08
1.0E-08
5.0E-09 Moisture Concentration at Interface
0.0E+00
0.0E+00 1.0E+06 2.0E+06 3.0E+06 4.0E+06 5.0E+06 6.0E+06
Time, Seconds
Fig. 10. Water vapor concentration at the interface.
sions, aggregate and asphalt source, moisture concentration
available in the environment and temperature. Considering the
actual dimensions of the pavement in the field, the maximum
damage generating in the bond occurs at much longer time than
the calculated value here.
Ewet
The A
dry term in the AD model is obtained by conducting the
EA
regression analysis; It is observed that this ratio decreases almost
linearly with RH from 1 to near 0.88. Such a trend is expected given
that water vapor molecules diffuse into the asphalt binder after
they reach the interface. However, since the diffusivity of asphalt
binder is smaller than that of asphalt mixture [38], and since the
time frame examined in this study is short, the induced damage
in asphalt binder is not significant. To simplify the AD model, the
averaged ratio, which is 0.96, is used in the model. Eqs. (36) and
(37) present the final form of AD model for the tested materials.
As shown in Fig. 13, the AD model with R2 = 0.98 is robust in pre-
dicting the adhesive damage:
0:4 0:5
dAD ¼ 1  ð1:42  0:42eð0:025C m Þ Þ ð36Þ
0:4 0:5
dAD ¼ 1  ð1:42  0:42eðRHÞ Þ ð37Þ
One of the main objectives of this study was to formulate the
AD model based on the fundamental of physics, and illustrate the
calibration process. Considering all the inputs required for calibrat-
ing the model, and moisture conditioning configuration needed to
estimate the variable of the model, the BBS test was selected in this
study for the calibration purposes. As mentioned earlier in
Section 4, substrates and asphalt binder were heated at 65 and
164 °C, respectively, and by following the ASSHTO TP-91 standard,
BBS test samples with strong asphalt-aggregate bond were built.
However, the temperature at which the samples were fabricated
might not exactly mimic typical mixing temperatures and conse-
quently the created aggregate-asphalt bond might not exactly rep-
resent the bond developed during mixing asphalt mixtures.
However, this does not cause a major issue regarding the AD model
as long as the bonding is strong enough. The AD model is defined
based on the ratio of wet to dry adhesive tensile strength, and
the extent of damage can be considered independent to factors
influencing both wet and dry bond strength equally. It has been
recognized that surface tension of asphalt binders decreases lin-
early with temperature [39]. As the adhesive bond energies pertain
directly to the surface tension of asphalt binder, it can be also
assumed that the adhesive bond energy declines linearly with tem-
perature. Therefore, to simplify the model, it can be assumed that
the wet and dry tensile strength change linearly with the fabrica-
tion temperature and almost with a same rate. In this way, the
10 M. Nobakht et al. / Construction and Building Materials 247 (2020) 118616

2
Aged
1.8 Wet-Aged
1.6

POTS, MPa 1.4

1.2

0.8

0.6

0.4

0.2

0
0 1 3 12 24 48 72 96 120 144 216 312 384 432 500 700
Conditioning Time, hours

Fig. 11. Measured POTS for dry and wet specimens.

Fig. 12. Growth of moisture damage with conditioning time, increasing from left to right as 0, 96, 216 and 432 h.

0.7 0.70
R2=0.98
0.6 0.60
Adhesive Damage

0.5 0.50
Adhesive Damage

0.40
0.4
0.30
0.3
0.20
0.2
Measured Data
0.10
Model Fitting
0.1
Measured Data 0.00
0.0 0.0E+00 1.0E-08 2.0E-08 3.0E-08 4.0E-08
0 100 200 300 400 500 600 700 800 Moisture Concentration (Cm) , gr/mm3
Time, Hours
Fig. 14. Adhesive damage vs Cm.
Fig. 13. Adhesive damage versus time.

ficients of the model. The AD model can be extended to account for

impact of temperature at which samples were fabricated gets can-
celed out given that it changes both numerator and denominator of
the model almost in a same manner.
Also, it should be noted that the abovementioned temperatures
are fabrication temperatures, at which samples were produced,
and they are different than conditioning and loading temperature,
which are 35 °C for samples tested in this study. This temperature
was selected given that it can reasonably represent the pavement
temperature for a long period of time during a year, and as pre-
sented earlier in this section, the AD model was calibrated only
at this temperature. Unlike fabrication temperature, the condition-
ing and loading temperature can impact the AD model since they
affect the variable of the model (i.e., Cm or RH) as well as the coef-
these variations using the linearity assumption mentioned earlier
and the fact that the impact of RH on decreasing the bond energy
is stronger at high temperature. The extension of the AD model is
beyond the scope of this paper.
5.3. Calibration of CD model
As mentioned previously in Section 4.2, wet specimens were
conditioned under 100% RH and 35 °C for varying predefined time
length. Although not very severe, conditioning specimens at 35 °C
increases the rate of asphalt binder aging to some extent. In order
to factor out the impact of aging and only account for the impact of
moisture damage, dry specimens were placed in the chamber at
 M. Nobakht et al. / Construction and Building Materials 247 (2020) 118616 11

0.70 As shown in Fig. 16, the CD model with R2 = 98 is robust in predict-

ing the cohesive damage when loading temperature is the same as
0.60 conditioning temperature.
Dependency of the cohesive damage on frequency is captured
0.50 by parameters bʹ and cʹ while aʹ accounts for the percentage of
Adhesive Damage

molecules affected by the amount of absorbed water. Fig. 17 shows

0.40 the relationship between these parameters and loading frequency
for S3 PG 64-22 FAM specimens tested here. It is observed that a
0.30 power law and logarithmic function are capable of appropriately
modelling bʹ and cʹ, respectively, as a function of loading frequency.
0.20 In general, fitting parameters aʹ, bʹ and cʹ account for changes hap-
Measured Data pening in the intermolecular bonding force of the FAM subjected to
0.10 a certain moisture conditioning. Three functions that describe P, bʹ
Model Fitting
and cʹ are presented in the following equations:
0.00
0.0 0.2 0.4 0.6 0.8 1.0 P ¼ 1  expð0:25wÞ ð39Þ
RH
0
b ¼ 0:0205fr0:181 ð40Þ
Fig. 15. Adhesive Damage vs RH.

c0 ¼ 0:13logðfrÞ þ 0:18 ð41Þ

35 °C for the same conditioning time as those for wet specimens.
Then, dynamic modulus values of wet and dry samples subjected Substituting Eqs. (39)–(41), and for aʹ, bʹ and cʹ in Eq. (31) gives
to each conditioning time were used in the left part of Eq. (23) to the final form of CD model as follows:
calculate the corresponding moisture damage ratio for each mois- "
ture conditioning time. In other words, the ratio of measured |E*| 1:17
dCD ¼ ½1  expð0:25wÞ  1 þ 8
values of moisture conditioned specimens to measured |E*| values ð1 þ 0:0205fr0:181 w0:13logfrþ0:18 Þ
of dry specimens is subtracted from one. The remainder indicates #
2:17
the cohesive damage induced in each conditioned specimen.  14
ð42Þ
The gravimetric water content corresponding to each condition- ð1 þ 0:0205fr0:181 w0:13logfrþ0:18 Þ
ing time is calculated as:

mw 5.4. Extension of CD model

w¼  100 ð38Þ
m
As stated in the preceding section, for moisture conditioned
Where,
FAM specimens, |E*| testing is conducted at several frequencies
mw = Mass of absorbed water in FAM specimen after condition-
but just one temperature of 35 °C to ensure that the water content
ing time of t, gram;
remains constant during testing. Based on the concept of time–
m = Mass of dry FAM specimen, gram.
temperature superposition for a viscoelastic material such as
The measured damage ratio is then plotted versus gravimetric
water content in Fig. 16. Each data point represents the average
cohesive damage introduced into two replicates. The CD model is (a) 0.035
then fitted to the measured damage ratios to find the fitting
parameters of aʹ, bʹ and cʹ illustrated in Eq. (31). Evidently, the 0.030
cohesive damage aggravates with an increase in the water content. 0.025
As mentioned in Section 4.2, in order to keep the water content
constant during testing, |E*| was conducted only at 35 °C but at dif- 0.020 y = 0.0205x-0.181
b′

ferent loading frequencies. It is observed that the cohesive damage

R² = 0.98
0.015
changes with the frequency of loading. In other words, at lower fre-
0.010
quency, the effect of diffused water in reduction of |E*| is greater
than that at higher frequency. Such a trend is expected because 0.005
at low frequency, water molecules have more time to push apart 0.000
macromolecules of the FAM and exacerbate the chain separation. 0 5 10 15 20 25 30
Frequency, Hz
0.6 (b) 0.40
R2=98%
0.35
dCD =1-|E*|aged-wet/|E*|aged-dry

0.5
0.30 y = 0.13Log(x)+0.18
0.4
0.25
0.3 0.20
c′

0.2 0.15
Measured at Fr=0.1 Measured at Fr=0.5 0.10
Measured at Fr=1 Measured at Fr=5
0.1 Measured at Fr=25 CD Model at Fr=0.1
CD Model at Fr=0.5 CD Model at Fr=1 0.05
CD Model at Fr=5 CD Model, Fr=25
0 0.00
0.0 2.0 4.0 6.0 8.0 10.0 12.0 0 5 10 15 20 25 30
Water Content (w), % Frequency, Hz

Fig. 16. Cohesive damage at different frequencies vs. water content. Fig. 17. Fitting parameters (a) bʹ and (b) cʹ versus frequency.
12 M. Nobakht et al. / Construction and Building Materials 247 (2020) 118616
FAM, the effect on the material property caused by the frequency
changes is similar to that caused by temperature changes. With
respect to the effect of diffused water vapor on intermolecular
bonding, increasing the temperature causes the same effect on
the reduction of intermolecular bond and cohesive damage as
decreasing the frequency. Therefore, the CD model can be extended
to account for the variation in both loading temperature and fre-
quency. It should be noted that the incompressibility of water
vapor within asphalt mixtures is not taken into account here for
simplification. Accordingly, a time–temperature shift factor is used
to combine the effect of temperature and frequency.
Referring to Eqs. (24) and (31), the |E*| for the moisture condi-
tioned FAM can be formulated as follows:
jE jwet ¼ jE jdry  ð1  dCD Þ
1:17
¼ jE jdry  f1  ½1  expða0 wÞ  ½1 þ
ð1 þ b wc Þ
2:17
 0 14
g
ð1 þ b wc0 Þ
As explained in Section 5.3 and shown in Eqs. (40) and (41), bʹ
and cʹ are formulated as the functions of the loading frequency.
Since the CD model is applied on the |E*| of dry samples, the shift
factor function of the dry specimens can be used to combine the
effect of the temperature and frequency into a single factor,
referred to as the reduced frequency:
f R ¼ adry
T f
Where,
f = Frequency;
adry
T = Shift factor of the dry FAM specimen;
Substituting Eq. (44) into (42) yields the final form of CD model
for the tested FAMs as follows:
"
1:17
dCD ¼ ½1  expð0:25wÞ  1 þ
0:181
ð1 þ ð0:0205f R
#
2:17
 14
0:181
ð1 þ ð0:0205f R Þwð0:13Logf R þ0:18Þ Þ
6. Validation of the AD and CD models
After determining all of the coefficients in the AD and CD model,
these models can be used to predict the adhesive damage induced
in the interface between aggregate and asphalt binder, and cohe-
sive damage occurred in the FAM due to moisture diffusion. It
should be noted that parameters of the adhesive and cohesive
damage models are specific to each material, and so they change
when a different mixture is used. Given that the parameters were
driven based on the principles of physics and mechanics, they
account for material characteristics involved in the adhesive and
cohesive damage. In sections 5-2, and 5-3, the adhesive and cohe-
sive models were calibrated for S3 PG 64-22 HMA and FAM speci-
mens and the parameters of the models were determined for these
materials. In addition to these parameters, the adhesive and cohe-
sive models involve two variables as Cm and w, which are func-
tions of conditioning period, RH, and water vapor diffusivity of
materials. In this section, performance of the models was verified
for new test sets from the same material as those used for the cal-
ibration. The new test sets were subjected to different conditioning
periods to develop different set of variables, and then the models
were verified by comparing the measured damage with the pre-
dicted damage.
An example of the comparison between the predicted and mea-
sured adhesive bond strength, and predicted and measured
dynamic modulus of a moisture conditioned FAM specimen are
shown in Figs. 18 and 19, respectively. Three test specimens were
prepared and conditioned in the moisture room for 250, 350 and
600 h following the approach detailed in Section 4. The bond
strength was then measured and compared with the values pre-
dicted using CD model. The comparison illustrated in Fig. 18 shows
that the validation results are acceptable considering the variation
that exists in the bond strength. Furthermore, two FAM specimens
were fabricated and conditioned in the moisture room for 50 and
70 days. The dynamic modulus test was then conducted on the
conditioned specimens at 0.1, 1, 5, 10, 25 frequencies and 35 °C
temperature. Fig. 19 shows the comparison between the measured
values with |E*| values calculated using CD model, which implies
the merit of the proposed model.
0 8
7. Conclusion and future work
ð43Þ
An adhesive and a cohesive moisture damage model were
developed in this study. The adhesive damage model accounts
for deterioration of the adhesive bond strength between aggregate
and asphalt binder as water vapor reaches the interface. The model
is formulated by linking a thermodynamic property (i.e., adhesive
bond energy) to a mechanical property (i.e., tensile strength of
the interfacial bond) of asphalt mixture. Adhesive bond energy is
characterized based on the water vapor concentration at the inter-
face and RH. A diffusion model is employed to estimate the water
ð44Þ
vapor concentration at the interfacial bond. Parameters of the
adhesive model were determined by conducting the Bitumen Bond
Strength test on dry and moisture conditioned test samples. It was
observed that as conditioning time increases, reduction in the ten-
sile bond strength increases as a result of increased amount of
accumulated water vapor concentration at the interface. The vali-
dation results show that the adhesive damage model performs well
in modelling the effect of water vapor concentration on the bond
8 strength deterioration.
Þwð0:13Logf R þ0:18Þ Þ The cohesive damage model proposed in this study accounts for
softening of the FAM due to the diffused water vapor. The model
ð45Þ was formulated using the Lennard-Jones potential, and relating
its components to the diffused water content. Calibration of the
model was completed by conducting the dynamic modulus test
on FAM specimens conditioned at 100% RH and 35 °C. It was
observed that the cohesive damage changes with the frequency
of loading; at lower frequencies, the effect of diffused water in
reduction of |E*| is greater than that at higher frequencies. A good-
ness of fit of R2 = 98% for the cohesive damage model suggests the
capability of this model in quantification of the impact of diffused
water vapor on the stiffness degradation of FAM. The cohesive
1.4
Measured Strength
Adhesive Bond Strength (MPa)
1.2 Predicted Strength
1
0.8
0.6
0.4
0.2
0
250 350 600
Moisture Conditioning Time (hours)
Fig. 18. Comparison of predicted and measured adhesive bond strength.
 M. Nobakht et al. / Construction and Building Materials 247 (2020) 118616
1400
1200
Predicted |E*| (MPa)
1000
800
600
400
200
0 and diffusion coefficient in hot-mix asphalt and their relationships to moisture
0 200 400 600 800 1000 1200
Measured |E*| (MPa)
Fig. 19. Comparison of predicted and measured |E*| of the FAM specimen.
damage was extended based on the concept of time–temperature
superposition to predict the cohesive moisture damage at various
frequencies and temperatures.
AC is a composite material consisting of aggregates, FAM and air
void. The mechanical characteristic of AC depends not only on the
properties of FAM and aggregates but also on the interfacial bond
strength between the aggregates and asphalt. The AD and CD mod-
els proposed in this study can be integrated into a single moisture
damage model to account for the detrimental effect of moisture on
the mechanical properties of asphalt pavements. Such a moisture
damage model can be incorporated in pavement design programs
and performance-related specifications to help with designing
long-lasting pavements and prevent premature failures. Develop-
ment and validation of the moisture damage model will be studied
as a continuation of this study in future.
CRediT authorship contribution statement
Mona Nobakht: Conceptualization, Data curation, Formal anal-
ysis, investigation, Methodology, Supervision, Validation, Visual-
ization, Resources, Writing-original draft, Writing-review &
editing. Derun Zhang: Data curation, Formal Analysis, Methodol-
ogy. Maryam Sakhaeifar: Methodology, resources, Writing-
review & editing. Robert Lytton: Conceptualization, Methodology,
Supervision.
Declaration of Competing Interest
The authors declare that they have no known competing finan-
cial interests or personal relationships that could have appeared
to influence the work reported in this paper.
References
[1] M. Ling, X. Luo, F. Gu, R.L. Lytton, Time-temperature-aging-depth shift
functions for dynamic modulus master curves of asphalt mixtures, Constr.
Build. Mater. 157 (2017) 943–951.
[2] M. Nobakht, M.S. Sakhaeifar, Dynamic modulus and phase angle prediction of
laboratory aged asphalt mixtures, Constr. Build. Mater. 190 (2018) 740–751,
https://doi.org/10.1016/j.conbuildmat.2018.09.160.
[3] D. Zhang, B. Birgisson, X. Luo, I. Onifade, A new short-term aging model for
asphalt binders based on rheological activation energy, Mater. Struct. (2019),
https://doi.org/10.1617/s11527-019-1364-7.
[4] M. Nobakht, M.S. Sakhaeifar, D.E. Newcomb, Selection of structural overlays
using asphalt mixture performance, J. Mater. Civ. Eng. 29 (2017), https://doi.
org/10.1061/(ASCE)MT.1943-5533.0002070 04017209.
[5] M. Ling, X. Luo, Y. Chen, S. Hu, R.L. Lytton, A calibrated mechanics-based model
for top-down cracking of asphalt pavements, Constr. Build. Mater. 208 (2019)
102–112.
[6] S. Caro, E. Masad, A. Bhasin, D.N. Little, Moisture susceptibility of asphalt
mixtures, Part 1: mechanisms, Int. J. Pavement Eng. 9 (2008) 81–98.
[7] R. Luo, T. Huang, D. Zhang, R.L. Lytton, Water vapor diffusion in asphalt
mixtures under different relative humidity differentials, Constr. Build. Mater.
136 (2018) 126–138.
13
[8] Hicks, R., Santucci, L. & Aschenbrener, T. 2003. Moisture sensitivity of asphalt
pavements: a national seminar. San Diego. California, 2-21.
[9] Tong, Y. 2013. Fatigue resistance of asphalt mixtures affected by water vapor
movement.
[10] E.J. Rueda, S. Caro, B. Caicedo, Influence of relative humidity and saturation
degree in the mechanical properties of Hot Mix Asphalt (HMA) materials,
Constr. Build. Mater. 153 (2017) 807–815.
[11] E.J. Rueda, S. Caro, B. Caicedo, Mechanical response of asphalt mixtures under
partial saturation conditions, Road Mater. Pavement Des. 20 (6) (2019) 1291–
1305.
[12] S. Caro, E. Masad, A. Bhasin, D. Little, Micromechanical modeling of the
influence of material properties on moisture-induced damage in asphalt
mixtures, Constr. Build. Mater. 24 (7) (2010) 1184–1192.
[13] E. Kassem, E. Masad, R. Bulut, R.L. Lytton, Measurements of moisture suction
1400 damage, Transp. Res. Rec. 1970 (2006) 45–54.
[14] D. Cheng, D.N. Little, R.L. Lytton, J.C. Holste, Use of surface free energy
properties of the asphalt-aggregate system to predict moisture damage
potential (with discussion), J Assoc. Asphalt Paving Technol. (2002) 71.
[15] A.W. Hefer, D.N. Little, R.L. Lytton, A synthesis of theories and mechanisms of
bitumen-aggregate adhesion including recent advances in quantifying the
effects of water, J. Assoc. Asphalt Paving Technol. 74 (2005) 139–196.
[16] AASHTO T283. 2007. Standard method of test for resistance of compacted
asphalt mixtures to moisture-induced damage. AASHTO Provisional
Standards: Washington, DC, USA.
[17] E. Arambula, E. Masad, A.E. Martin, Moisture susceptibility of asphalt mixtures
with known field performance: Evaluated with dynamic analysis and crack
growth model, Transp. Res. Rec. 2001 (2007) 20–28.
[18] A.R. Copeland, J. Youtcheff, A. Shenoy, Moisture sensitivity of modified asphalt
binders: factors influencing bond strength, Transp. Res. Rec. 1998 (2007) 18–28.
[19] Lytton, R. L., Masad, E. A., Zollinger, C., Bulut, R. & Little, D. N. 2005.
Measurements of surface energy and its relationship to moisture damage.
[20] I. Song, D.N. Little, E.A. Masad, R.L. Lytton, Comprehensive evaluation of
damage in asphalt mastics using X-ray CT, continuum mechanics, and
micromechanics (with discussion), J. Assoc. Asphalt Paving Technol. (2005) 74.
[21] McCarthy, L. M., J. Callans, R. Quigley, and S. V. Scott III. 2016. Performance
Specifications for Asphalt Mixtures. Synthesis No. 492. National Cooperative
Highway Research Program. Washington, D.C.
[22] R. Luo, D. Zhang, Z. Zeng, R.L. Lytton, Effect of surface tension on the
measurement of surface energy components of asphalt binders using the
Wilhelmy Plate Method, Constr. Build. Mater. 98 (2015) 900–909.
[23] D. Zhang, R. Luo, Development of a method to determine surface energy
components of mineral fillers, Constr. Build. Mater. 146 (2017) 370–380.
[24] D. Zhang, R. Luo, Modeling of adsorption isotherms of probe vapors on
aggregates for accurate determination of aggregate surface energy
components, Constr. Build. Mater. 134 (2017) 374–387.
[25] R.J. Good, Spreading pressure and contact angle, J. Colloid Interface Sci. 52
(1975) 308–313.
[26] D. Zhang, R. Luo, Z. Zeng, Characterization of surface free energy of mineral
filler by spreading pressure approach, Constr. Build. Mater. 218 (2019) 126–
134.
[27] D. Zhang, R. Luo, Modifying the BET model for accurately determining specific
surface area and surface energy components of aggregates, Constr. Build.
Mater. 175 (2018) 653–663.
[28] Lytton, R. L. 2015. Micromechanics of Civil Engineering Materials, Class Notes.
Zachry Department of Civil Engineering, T. A. M. U. (ed.). College Station, TX.
[29] F. Zhou, F. Lydon, B. Barr, Effect of coarse aggregate on elastic modulus and
compressive strength of high performance concrete, Cem. Concr. Res. 25
(1995) 177–186.
[30] R.G. Allen, D.N. Little, A. Bhasin, R.L. Lytton, Identification of the composite
relaxation modulus of asphalt binder using AFM nanoindentation, J. Mater. Civ.
Eng. 25 (2012) 530–539.
[31] LIU, C, Van der Waals force and asphalt concrete strength and cracking, J. Eng.
Mech. 131 (2005) 161–166.
[32] W.D. Callister, D.G. Rethwisch, Materials Science and Engineering, John Wiley
& Sons NY, 2011.
[33] AASHTO TP-91. 2015. Standard Method of Test for Determining Asphalt Binder
Bond Strength by Means of the Binder Bond Strength (BBS) Test. AASHTO
Provisional Standards: Washington, DC, USA.
[34] K. Kanitpong, H. Bahia, Relating adhesion and cohesion of asphalts to the effect
of moisture on laboratory performance of asphalt mixtures, Transp. Res. Rec.
1901 (2005) 33–43.
[35] R. Moraes, R. Velasquez, H.U. Bahia, Measuring the effect of moisture on
asphalt–aggregate bond with the bitumen bond strength test, Transp. Res. Rec.
2209 (2011) 70–81.
[36] P. Sousa, E. Kassem, E. Masad, D. Little, New design method of fine aggregates
mixtures and automated method for analysis of dynamic mechanical
characterization data, Constr. Build. Mater. 41 (2013) 216–223.
[37] CRANK, J. 1979. The mathematics of diffusion, Oxford university press.
[38] E. Arambula, S. Caro, E. Masad, Experimental measurement and numerical
simulation of water vapor diffusion through asphalt pavement materials, J.
Mater. Civ. Eng. 22 (2009) 588–598.
[39] Pauli, A. T. 2014. Chemomechanics of damage accumulation and damage-
recovery healing in bituminous asphalt binders. Doctoral dissertation, Sieca
Repro, Delft, Netherlands.