Road Materials and Pavement Design, 2017

https://doi.org/10.1080/14680629.2017.1350598

Predicting thermal cracking of asphalt pavements from bitumen and mix

properties
Bagdat Teltayeva∗ and Boris Radovskiyb
a Kazakhstan Highway Research Institute, Almaty, Kazakhstan; b Radnat Consulting, Irvine, CA, USA

(Received 11 October 2016; accepted 24 June 2017 )

The objective of this study was to develop approximate relationships for prediction of
the rheological properties and tensile strength of asphalt concrete with conventional bitu-
men binder to address the low-temperature cracking analysis. The parameters of modified
Christensen–Anderson–Marasteanu model for relaxation modulus of bitumen were related
with its penetration index and softening point by approximate formulas based on Van Der
Poel’s nomograph for determining bitumen stiffness. Hot mix asphalt relaxation function has
been described by the Christensen–Bonaquist model. Hot mix asphalt tensile strength as a
function of loading time and temperature was related to the bitumen stiffness by equation
based on Heukelom’s data and the Molenaar-Li empirical formula. The strength of hot mix
asphalt has been also studied by performing tensile tests at constant strain rate and temper-
atures. The results were analysed considering the calculated stress increase with temperature
drop at various cooling rates. Critical temperature was estimated from comparison of thermal
stress and tensile strength of asphalt concrete at constant stress rate.
Keywords: asphalt pavement; thermal stress; low-temperature cracking; relaxation modulus;
tensile strength; critical temperature

1. Introduction
Low-temperature cracking is one of the major causes of failure of asphalt pavements in cold
weather climates, including much of the United States, Canada, Russia, Ukraine, Kazakhstan,
and other countries at extreme northern and southern latitudes. Thermal cracking has been asso-
ciated with the volumetric change in the asphalt concrete layer. When the temperature drops,
the pavement tends to contract its volume but the thermal strains are unrealised. That causes the
thermal stresses in pavement. They build up until they reach the strength of the material, leading
to the formation of cracks to relieve these stresses.
Monismith, Secor, and Secor (1965) developed a theoretical calculation method for the ther-
mally induced stress in asphalt pavement as in an infinite viscoelastic beam based on Humphreys
and Martin (1963) solution. Boltzmann’s superposition principle and hereditary integral form
of the constitutive equation for linear viscoelastic material was applied to relate time-dependent
stresses and strains. This method is currently used for the estimation of critical cracking tem-
perature by many researchers. Hills and Brien (1966) devised a simple pseudo-elastic method
to compute the thermal stresses in asphalt pavement. The predicted stress was dependent on
loading time equal to time interval of numerical integration over which the change of stress
was computed. Fromm and Phang (1971) eliminated the dependency on the time interval of

*Corresponding author. Email: bagdatbt@yahoo.com

© 2017 Informa UK Limited, trading as Taylor & Francis Group

2 B. Teltayev and B. Radovskiy

numerical integration by specifying loading times. However, the results of the stress compu-
tations were still directly influenced by the loading time specified. Christison, Murray, and
Anderson (1972) employed five different methods of stress computation, including the Moni-
smith’s and Hills-Brien’s methods, and concluded that a potential of low-temperature cracking
for a given pavement can be evaluated if the mix stiffness and strength characteristics are known
as a function of temperature and time of loading. Buttlar and Roque (1994) developed the indi-
rect tension test (IDT) to measure the creep compliance and tensile strength of asphalt mixtures
at low temperatures. The IDT testing of cores is very appropriate for the evaluation of existing
pavements. However, several differences between the axial tension and IDT tests raise ques-
tions about the interchangeability of the creep compliance and tensile strength values obtained
from the two test methods. Bouldin, Dongre, Rowe, Sharrock, and Anderson (2000) consid-
ered the thermally induced stress in asphalt binder as in viscoelastic bar and reported that the
midpoint of the binder’s glass transition is close to the pavement critical cracking temperature.
Based on the study of Bouldin et al. (2000), in AASHTO and ASTM specifications, MP1a-02
(AASHTO, 2002), PP 42-07 (AASHTO, 2007), and ASTM D 6816-02 (ASTM, 2002), asphalt
mixture thermal stresses were calculated from asphalt binder thermal stresses multiplied by an
empirical pavement constant equal to 18. This was done to obtain a critical cracking temperature
at which the mixture thermal stress curve and binder strength curve intersect (Moon, Marasteanu,
& Turos, 2013). First, the binder tensile strength at a constant strain rate of 3%/min cannot be
used as same as the mix tensile strength at different stress rates. Second, only one pavement
constant is not sufficient to convert the binder properties to the mix properties. This method
cannot reflect the difference in thermal stresses for mixtures prepared with the different aggre-
gate source and gradation and the asphalt mixture volumetric (i.e. air void, binder quantity, etc.).
For instance, Roy and Hesp (2001) for limited set of binders estimated the variation of pavement
“constant” between 3.4 and 16.7. Hu, Zhou, and Walubita (2009), Prieto-Muñoz, Yin, and Buttlar
(2013), and Dave, Buttlar, Leon, Behnia, and Paulino (2013) replaced the one-dimensional pave-
ment model with the two-dimensional stress analysis of low-temperature cracking implemented
within a viscoelastic finite element modelling framework. Farrar, Hajj, Planche, and Alavi (2013)
combined the thermal stress restrained and unrestrained specimen tests of asphalt mixture. The
measured restrained thermal stress development was very similar to the thermal stress build-up
calculated from the Boltzmann’s hereditary integral. Relaxation modulus of asphalt mixture was
determined by approximate conversion of complex modulus estimated from the model based on
the law of mixtures for composite materials and from the measured complex modulus of the
recovered asphalt binder (Farrar et al., 2013).
The low-temperature transverse cracking in asphalt pavements is the result of a combination
of two distress mechanisms: single-event thermal cracking and thermal fatigue. The single-
event thermal cracking as the significant contributor to transverse cracking of pavements was
considered in this study. To better understand the ways to reduce low-temperature cracking, a
mechanistic model is needed that includes the predicted rheological and fracture properties of
asphalt mixture.
The objective of our study was to develop the relationships for prediction of the viscoelastic
properties and tensile strength of asphalt concrete with conventional bitumen binder as a function
of time and temperature and apply them to the low-temperature cracking problem.

2. Stiffness and relaxation modulus of bitumen

Stiffness modulus S(t) introduced by Van der Poel (1954) is the ratio of the constant applied
uniaxial stress σ0 to the resulting uniaxial strain ε(t) at time t. Based on the test results of 47
bitumens, Van der Poel developed a nomograph to determine the stiffness of asphalt cement as
 Road Materials and Pavement Design 3

a function of temperature T and loading time t (or frequency ω) if the softening temperature Ts
and the penetration index PI are known. To calculate binder stiffness, it is convenient to use the
BitProps program (free software) based on the scanned Van der Poel’s nomograph developed by
G. Rowe and M. Sharrock (Abatech, Inc.).
Several researchers proposed the empirical equations for stiffness modulus of bitumen depend-
ing on penetration and softening temperature based on Van der Poel’s data. Saal (1955) proposed
the relationship between the bitumen stiffness and its penetration for load duration t = 0.4 s and
PI = 0. Ullidz and Larsen (1984) proposed a simple equation for S(t) and limited its applica-
bility to the duration of loading 0.01 < t < 0.1, range of penetration index − 1 < PI < 1, and
temperature range 10°C < T < 70°C. Even in this narrow range of input parameters, the average
coefficient of variation of stiffness calculated using the proposed equation from Van der Poel’s
data is 40%. Shahin (1977) developed two regression equations based on Van der Poel’s nomo-
graph to relate stiffness modulus with time, temperature, penetration index, and softening point:
one of them for S < 1 MPa and another for S > 1 MPa. The range of their applicability in terms
of input parameters was not specified. Meanwhile, it is unknown what value of stiffness modulus
will be and which of these formulas should be used. Moreover, often for the same set of input
parameters, the first equation returns S < 1 MPa while the second gives S > 1 MPa and it is not
clear which of the results is correct. To our knowledge, the proposed formulas for the bitumen
stiffness based on Van der Poel’s data were mostly developed using purely regression techniques
without involving any rheological models.
In this study as a mathematical model for describing the stiffness of bitumen, Christensen and
Anderson (1992) expression was used:
  β −1/β
Eg t
S(t) = Eg 1 + , (1)
3η

where Eg is the uniaxial glassy modulus of binder (MPa), η is the steady-state viscosity (MPa·s),
and β is constant.
The instantaneous value for longitudinal modulus Eg was obtained by extrapolation of values
for stiffness modulus S(T, t) according to Van der Poel at low temperature T and small load
duration t for t → 0. With the purpose of extrapolation, we used the model developed to describe
the viscoelastic properties of amorphous glass forming polymers (Drozdov, 2001) based on the
theory of cooperative relaxation (Adam & Gibbs, 1965). For the asphalt with certain penetration
index, we selected the lowest temperature at which using the program BitProps the value of
stiffness S can be obtained for t = 0.00005 s. This lowest temperature varied from − 42° for
PI = + 2 to 9° for PI = − 3. The stiffness modulus S was obtained at that lowest temperature
for a set of eight loading times t = 0.00005, 0.0001, 0.0002, 0.0005, 0.001, 0.002, 0.005, and
0.01 s. Then using the Equations (18) and (37) of Drozdov (2001), the value of instantaneous
modulus was found by fitting the Van der Poel’s data for stiffness modulus at eight loading
times. It was concluded that for asphalts having the penetration index from − 2 to + 3, the
average value of instantaneous longitudinal modulus equals to E g = 2460 MPa with the standard
deviation of 7%.
After fitting Equation (1) to Van der Poel data, the following equations were obtained:
0.1794
β= , (2)
1 + 0.2084PI − 0.00524PI 2

η = aT Ahrr (T) · η(Tr ) (at T ≤ Ts − 10),

(3)
η = aT WLF (T) · η(Tr ) (at T > Ts − 10),
4 B. Teltayev and B. Radovskiy
    
12(20 − PI ) 0.2011
η(Tr ) = 0.00124 1 + 71 exp − · exp , (4)
5(10 + PI ) 0.11 + 0.0077PI
where Tr is reference temperature Tr = (Ts − 10) [°C], and aT (T) is time-temperature superpo-
sition function:
  
3(30 + PI ) 1 1
aT Ahrr (T) = exp 11720 · − , (5)
5(10 + PI ) (T + 273) (Ts + 263)
 
2.303(T − Ts + 10)
aT WLF (T) = exp − . (6)
(0.11 + 0.0077PI )(114.5 + T − Ts )
For 1910 points in the range of PI from − 3 to + 2, (T − Ts ) from − 45°C to 10°C, and t from
10−4 to 104 s, the standard deviation of stiffness calculated using Equation (1) from the Van der
Poel’s data was 14.6% and the coefficient of determination R2 = 0.978.
By definition, stiffness modulus S(t) is the inverse of uniaxial creep compliance. A related
function is the shear creep compliance:
   1/β
1 Gg t β
J (t) = 1+ , (7)
Gg η

where β is given by Equation (2) and Gg = the glassy modulus in shear; Gg ≈ Eg /3 = 820 MPa
if the material is considered isotropic with Poisson’s ratio ν = 0.5, although it should be noted
that ν is time dependent (Di Benedetto, Delaporte, & Sauzeat, 2007).
There are different methods for converting the shear creep compliance J (t) to relaxation modu-
lus G(t). In this study, the modified Hopkins–Hamming algorithm (Tschoegl, 1989) was selected,
and relaxation modulus G(t) was approximated by Christensen, Anderson and Marasteanu
(CAM) model (Marasteanu & Anderson, 1999):
  b −k/b
Gg t
G(t) = Gg 1 + , (8)
η

where b and k are constants. It can be shown that the parameter k is a function of b from the
following equation (Ferry, 1980):
 ∞
G(t) dt = η, (9)
0

which can be treated as a definition of zero-shear viscosity η. After substituting Equations (8) to
(9) and integration, it is easy to find that exactly k = 1 + b. This leads to the following formula
for relaxation modulus at shear:
   −(1+1/b)
Gg t b
G(t) = Gg 1 + , (10)
η

where
 −1
1 ln(π )
b= + −2 . (11)
β ln(2)
The uniaxial relaxation modulus of binder can be found as Eb (t) ≈ 3G(t), where its shear
modulus G(t) is given by Equation (10).
 Road Materials and Pavement Design 5

Complex modulus of bitumen in shear was determined from creep compliance Equation (7) at
a frequency equal to the inverse of the loading time using the Schwarzl and Struik (1968) method
in the following form:
  β −1/β
Gg Gg
Gd (ω) = 1+ , δ(ω) = m(ω)π/2, (12)
(1 + m(ω)) ηω

where Gd (ω) is the norm of complex modulus, δ is phase angle, ω is frequency, (x) is the
gamma function, and
(Gg /ηω)β
m(ω) = . (13)
1 + (Gg /ηω)β
For example, comparison of experimental results (Christensen & Anderson, 1992) at
T = 25°C versus calculated from Equation (12) values of Gd (ω) and δ(ω) is shown in Figure 1
for SHRP core bitumen AAB-1 with pen 25 = 98 dmm, Ts = 47.8°C, PI = 0 (Mortazavi &
Moulthrop, 1993). Tests were performed on dynamic shear rheometer with a range in frequency
from 0.1 to 100 rad/s at temperatures − 35, − 25, − 15, − 5, 5, 15, 25, 35, 45, and 60°C. The
data at all temperatures were then shifted with respect to time to construct the master curve at
reference temperature 25°C (Christensen & Anderson, 1992). Notice that the agreement between
the calculated and experimental results for the norm of complex modulus and the phase angle is
good.
The relaxation spectrum is a useful, fundamental way of characterising the time-dependent
behaviour of asphalt binders (Christensen, Anderson, & Rowe, 2016). According to Alfrey’s
rule (Ferry, 1980), relaxation spectrum H (τ ) of binder is obtainable in first approximation as a
negative slope of the relaxation modulus. Differentiating Equation (10) with respect to ln t leads
to
    −(2+1/b)
Gg τ b Gg τ b
H (τ ) = Gg (1 + b) 1+ . (14)
η η

Figure 1. An example of comparisons between the model predictions and experimental data for complex
modulus and phase angle as a function of frequency: points – measured (Christensen & Anderson, 1992),
lines – calculated from Equation (12).
6 B. Teltayev and B. Radovskiy

Figure 2. The relaxation spectrum for bitumen AAB-1 at 25°C: dashed line – from experimental data
(Reproduced with permission from Christensen, 1992); solid line – calculated from Equation (14).

Relaxation spectrum calculated in Christensen (1992, Figure 3.4) for the same SHRP core
bitumen AAB-1 based on measurements of complex modulus is presented in Figure 2 together
with calculated from Equation (14). The agreement between them is good. From Equation (14),
it is easy to obtain the relaxation time corresponding to the maximum density of spectrum
 1/b
η b
τmax = , (15)
Gg 1 + b
and its maximum density
 
b + 1 (2b+1)/b
Hmax = Gg b . (16)
2b + 1
An acceptable estimate for the maximum density of relaxation spectrum of bitumen is Hmax =
Gg b/3. It follows that maximum density of spectrum is proportional to the parameter b of CAM
model and is independent on temperature of bitumen.
One can see that the important rheological properties of bitumen such as stiffness, relax-
ation modulus, complex modulus, and relaxation spectrum can be estimated from the pro-
posed equations using the simplest standard properties of bitumen: penetration and softening
temperature.

3. Stiffness and tensile strength of asphalt concrete

To analyse the temperature-induced stresses, the uniaxial relaxation modulus of asphalt concrete
Emix (t) is needed. Christensen and Bonaquist (2015) recently improved their model:
Emix (t) = Pc [Eagg · (1 − VMA) + Eb (t) · VFA · VMA], (17)

Pc = 0.006 + 0.994[1 + exp(−(0.663 + 0.586 · ln(VFA · Eb (t)/3) − 12.9VMA

− 0.17 · ln(εs · 106 ))]−1 , (18)
 Road Materials and Pavement Design 7

where Eb (t) is relaxation modulus of binder (MPa), Eagg is elastic modulus of aggregate (MPa),
VMA are voids in mineral aggregate (volume fraction), VFA are voids filled with asphalt (vol-
ume fraction), εs = 0.0001 is the standard target strain, and Pc is the contact factor introduced in
Christensen and Bonaquist (2015) and defined by Equation (18). The same model can be used
for the stiffness modulus of asphalt concrete Smix (t) if Eb (t) is replaced by binder stiffness S(t)
defined according to Equation (1).
To estimate a critical temperature, the tensile strength of asphalt concrete as a function of
temperature is needed. W. Heukelom has presented compelling evidence that the tensile strength
of mix is related to the properties of bitumen contained herein (Heukelom, 1966). He gave some
data of tensile strength measurements of eight dense-graded mixes carried out at a variety of
temperatures and speeds. Heukelom has showed that the relative tensile strength, that is, the
tensile strength divided by its maximum value, can be represented by one curve for all mixes
tested as a function of the stiffness of bitumen recovered (Heukelom, 1966, Figure 22):

Tensile Strength
R(S) = Of Mix. (19)
Maximum Tensile Strength

The Heukelom curve can be approximated by equation

0.774 + 0.039r + 0.141r4.547

R= , (20)
1 + 0.026r3.608 · exp(1.245r)

where r = log(Eg /S), Eg is the uniaxial glassy modulus of binder (MPa), S is the binder stiffness.
Molenaar and Li (2014) proposed an empirical equation to estimate the maximal tensile
strength of asphalt concrete Ph as a function of mixture stiffness and the volumetric composition.
Six dense-graded mixes and one porous mixture with a bitumen of 10/20 to 70/100 penetra-
tion and aggregate of maximum grain size from 4 to 32 mm were tested at the Delft University
of Technology in a temperature range of 5–35°C at constant tensile strain rate from 0.0001 to
0.04/s. Regression analysis was performed to predict the maximum tensile strength of mixtures
depending on mixture stiffness Smix at 20°C and asphalt-void ratio (VFA). Molenaar and Li
(2014) developed the following regression equation:

Ph = 0.505 · Smix
0.308
· VFA0.849 . (21)

The combination of Equation (20) based on the Heukelom curve for relative strength of mix with
empirical Equation (21) for the maximal strength of mix (Molenaar & Li, 2014) leads to the
following expression for tensile strength f as a function of temperature T and the loading time t:

0.774 + 0.039r + 0.141r4.547

f = 0.505 · Smix
0.308
· VFA0.849 · . (22)
1 + 0.026r3.608 · exp(1.245r)

where S is the binder stiffness (MPa) defined in Equation (1) as a function of time and tempera-
ture, Smix is the mix stiffness at T = 20°C and at fixed loading time of 0.06 s, and f is the tensile
strength under rectangular pulse loading of duration t.
Equation (22) relates the tensile strength of asphalt concrete f with temperature T and time to
failure tf in terms of the binder creep stiffness S(tf ), as Heukelom (1966) suggested. At constant
strain rate Vε , the tensile strength fε in the left-hand side of Equation (19) can be expressed as a
product of time to failure, the strain rate, and the secant modulus at failure Hmix (tf ). The secant
modulus Hmix (t) is related to the relaxation modulus Emix (t) by the following equation (Smith,
8 B. Teltayev and B. Radovskiy

1976):

1 t
Hmix (t) = Emix (t − τ ) dτ . (23)
t 0
Thus using Equation (22), one can find the time to failure tf numerically as a root of the following
equation:
0.774 + 0.039r + 0.141r4.547
tf Vε Hmix (tf ) = 0.505 · Smix
0.308
· VFA0.849 · , (24)
1 + 0.026r3.608 · exp(1.245r)
where r = log(Eg /S(tf )).
Then, the tensile strength of asphalt concrete fε at constant strain rate Vε can be calculated as
fε = tf Vε Hmix (tf ). (25)
The calculated tensile strength at constant strain rate was compared with test results of direct
tensile measurements at Braunschweig Technical University (Stock & Arand, 1993). The main
features of testing machine developed at the University of Braunschweig are a very stiff frame,
strain transducers for the measurement of the length of the specimen, and a step motor to apply
strain to a specimen at a pre-determined rate. The samples were prismatic with a cross section
of 40 mm by 40 mm. The specimen length was 160 mm. Seven asphalt concrete mixtures with
the same content of different binders of 4.7% by weight of total mix were testedat constant strain
rate Vε = 1 · 10−4 /s (1 mm/min) and at temperatures 20, 5, − 10, and − 25°C. The test results
for mix with unmodified bitumen B1 having the penetration index PI = − 0.692 and softening
temperature Ts = 49°C for the mix volumetric properties VMA = 0.15 and VFA = 0.733 are
shown in Figure 3 together with tensile strength curve calculated from Equation (25). The cal-
culated maximal strength and the general shape of curve agree with measured strength although
the curve looks shifted along the temperature axis by approximately 4–5°C. Possible reasons of
shifting will be discussed later.
In the Kazakhstan Highway Research Institute, the uniaxial tensile tests were performed on
testing system TRAVIS (InfraTest GmbH), which is a version of device developed at the Uni-
versity of Braunschweig (Figure 4).This testing system includes a compact test frame integrated

Figure 3. Measured tensile strength values at constant strain rate (Stock & Arand, 1993) and those
calculated from Equations (24) and (25).
 Road Materials and Pavement Design 9

Figure 4. Machine used for the mechanical testing in the Kazakhstan Highway Research Institute.

in the temperature chamber. The load is applied to the specimen via heavy-duty screw jack and
stepping motor. An electronic load cell is directly attached to the spindle. The equipment is con-
trolled by a PC connected to the motor and the transducers. PC is used for measurement data
logging, control of the test procedure, and the temperature test chamber.
Dense-graded asphalt mixture was prepared with the use of granite aggregate fractions of
5–10 mm (20%), 10–15 mm (13%), and 15–20 mm (10%) from the Novo-Alekseevsk rock pit
(Almaty region); sand of fraction 0–5 mm (50%) from the plant Asphaltconcrete-1 (Almaty city)
and activated filler (7%) from the Kordai rock pit (Zhambyl region). The air-blown bitumen was
produced by the Pavlodar petrochemical plant from the crude oil of Western Siberia (Russia).
After short-time ageing, the bitumen penetration was 70 dmm (at 25°C), softening temperature
Ts = 48°C, and PI = − 0.91.A mixture designated as K2 was prepared with 4.8% of bitumen
by weight of aggregates. Mixture was compacted with the Cooper roller compactor (model CRT-
RC2S) to the average void content of 3.6%. Rectangular specimens (50 × 50 × 160 mm3 ) were
then sawed from these slabs and glued with the epoxy resin to the mounts. Specimens were
tested with nominal deformation rate of 1 mm/min (Vε = 1 · 10−4 /s) . Tests were performed at
temperatures 20, 10, 0, − 10, − 20, and − 30°. The strength values were obtained from five
tensile tests for all temperatures (Figure 5). The average coefficient of variation was of 15.6%.
Figure 5 presents the comparison of the measured tensile strength of asphalt concrete samples
and the strength of asphalt concrete predicted form Equations (24) and (25) for Vε = 1 · 10−4 /s
and the mix volumetric properties VMA = 0.144, and VFA = 0.75 (dashed curve).The mod-
ulus of aggregate of bulk specific gravity Gsb = 2.760 was estimated using the correlation
recommended in Christensen and Bonaquist (2015) as Eagg = 7650G1.59 sb = 36, 000MPa.
In Figure 5 as before in Figure 3, the calculated maximal strength and the shape of the predicted
curve (dashed red) agree with measured tensile strength but the predicted curve is constantly
shifted along the temperature axis to higher temperatures even more than in Figure 3. This shift
can be caused by various reasons. One of them might be the time-temperature superposition.
Obviously, the binder, regardless of the aggregates skeleton, drives the temperature dependency
of a bituminous material (Di Benedetto et al., 2011). In this paper, the time-temperature superpo-
sition function for the binder aT (T) in Equations (5) and (6) was determined from Van der Poel’s
data. Meanwhile, Anderson et al. (1990) found that Van der Poel’s nomograph underpredicted
the bitumen stiffness at long loading times and low temperatures. Although the time-temperature
sensitivity of bitumen at low temperatures is an important issue and deserves further research
10 B. Teltayev and B. Radovskiy

Figure 5. Measured and calculated tensile strength values: dashed curve – calculated from Equations (24)
and (25) at constant strain rate Vε = 1 · 10−4 /s, dotted curve – calculated from Equation (22) for average
time to failure tf = 40 s.

efforts, this reason does not explain the shift at temperatures T > 0°C on Figure 5 (the dashed
curve).
Another possible reason of shift is the effect of machine compliance on specimen strain rates.
To facilitate data acquisition, the system TRAVIS was designed to extend the testing time at
cold temperatures. As a result, because of machine compliance, the on-specimen strain rate was
not constant, given that the cross-head displacement rate 1 mm/min remains constant throughout
the test, due to the high stiffness of the material at low temperatures. Actual time to failure was
between 40 and 50 s at temperatures 20, 10, 0, − 10–20°C and around 20–25 s at T = − 30°C.
To check the validity of this reason, we calculated the tensile strength of the same asphalt mix
K2 at loading time 40 s from Equation (22) for PI = − 0.91, Ts = 48°C, VMA = 0.144, and
VFA = 0.75 (dotted curve in Figure 5). Shift of calculated curve to cold temperatures closer to
measured strengths confirms the significant effect of machine compliance on test results. If a true
constant-rate-of-deformation test is to be performed on asphalt concrete, feedback to loading
unite must be from on-specimen LVDTs and not from the cross-head displacement as was the
case in this experiment. This constant-strain loading requirement is difficult to achieve. More-
over, although used by some researchers (Bouldin et al., 2000; Christison et al., 1972; Stock
& Arand, 1993), the advantages of constant strain-rate test are doubtful of its value for low-
temperature cracking problem because the longitudinal strain in asphalt pavement induced by
cooling is unrealised, that is, equals to zero up to appearance of the transverse crack. For the
single-event low-temperature cracking analysis, the stress-controlled strength test looks more
important than strain-controlled test.
At constant stress rate Vσ , time to failure equals the tensile strength fσ over the stress rate
tf = fσ /Vσ . In that case, the binder stiffness in the right-hand side of Equation (22) depends on
 Road Materials and Pavement Design 11

Figure 6. Tensile strength of asphalt concrete as a function of temperature calculated from Equation (26)
at different stress rates.

tensile strength and the stress rate. The tensile strength fσ at constant stress rate Vσ can be found
numerically from Equation (26) as a function of T and Vσ :

0.774 + 0.039r + 0.141r4.547

fσ = 0.505 · Smix
0.308
· VFA0.849 · , (26)
1 + 0.026r3.608 · exp(1.245r)

where r = log(Eg /S(tf = fσ /Vσ )).

Figure 6 depicts the tensile strength calculated from Equation (26) for the same as before
asphalt mix K2 at different constant stress rates.
To illustrate, the mix strength at T = − 2°C and Vσ = 0.05MPa/s equals to 3.1 MPa
(Figure 6). In that case, time to failure is 62 s; stiffness of binder at T = 20°C and t = 0.06 s
according to Equation (1) is S = 6.02 MPa; stiffness of mix at T = 20°C and t = 0.06 s
from Equation (17) equals to Smix = 6040MPa. Parameter r is equal to r = log(2460/S(T =
−2◦ C, t = 62 s)) = 2.611, and the right-hand side of Equation (26) returns the tensile strength
of 3.1 MPa. A 10-fold increase in stress rate shifts the strength vs. temperature curve to higher
temperatures as much as around 7°C (Figure 6).

4. Thermal stresses and critical temperature

Predicted relaxation modulus Emix and the stress rate dependent tensile strength fσ were used
to estimate the critical single event cracking temperature. In this study, the critical cracking
temperature (Tcr ) is defined as the temperature at which the asphalt concrete tensile strength at
constant stress rate crosses the thermal stress curve.
Low-temperature-induced stresses in asphalt pavement were calculated as in an infinite
viscoelastic beam resting on frictionless rigid foundation:
 t  
d
σ (t) = − αmix (T(τ ))Emix (ξ(t) − ξ(τ )) T(τ ) dτ . (27)
0 dτ
12 B. Teltayev and B. Radovskiy

where t is the present time (s), τ is the passed time (s), T(τ ) is the temperature variation with time
(°C), αmix is the asphalt mixture coefficient of thermal t contraction (/°C), Emix is the relaxation
modulus (MPa) and ξ(t) is the reduced time: ξ(t) = 0 dτ/aT (T(τ )).
Coefficient of thermal contraction was assumed constant αmix = 2.5 · 10−5 /◦ C. The proper-
ties of binder and mix were taken as before: Ts = 48°C, PI = − 0.91, E agg = 36,000 MPa,
VMA = 0.144, VFA = 0.75. The relaxation modulus of asphalt concrete was determined from
Christensen–Bonaquist model – Equation (17).
A sinusoidal variation between the maximum and minimum temperatures during a day was
assumed in the first example (Figure 7).The temperature drops from 5°C to − 35°C and returns
to 5°C over 24-hour period. Thermal stress build-up was calculated according to Equation (27).
A tensile stress rate was estimated as Vσ (t) = dσ (t)/dt and it varies from zero to 3.8·10−4 MPa/s
with the average of about 2·10−4 MPa/s. Tensile strength of asphalt concrete fσ at Vσ = 2 ×
10−4 MPa/s according to Equation (26) as a function of T is shown on the same graph. The
critical cracking temperature at which the thermal stress curve and asphalt concrete strength
curve intersect is Tcr = − 33°C.
In the second example, the pavement temperature drops from 5°C to − 35°C at different con-
stant cooling rates. As expected, the stress accumulation rate was found dependent on cooling
rate (Figure 8). Cooling rate increase greatly increases the thermal stress rate. At the cooling
rate of 2°C/h that is similar to the rates experienced by real pavements, the loading rate at the
temperature close to − 30°C is around Vσ (t) = 2 · 10−4 MPa/s (Figure 8). Figure 9 is a plot of
stress according to Equation (27) at cooling rate of 2°C/h and of asphalt concrete strength ver-
sus temperature from Equation (26) at Vσ (t) = 2 · 10−4 MPa/s. The critical temperature is about
Tcr = − 34.5°C.
Low-temperature cracking of asphalt pavements is a widespread and costly problem in cold
regions. Ability to predict the temperature-induced stress in asphalt pavement and asphalt con-
crete strength starting from the properties of binder and mix should be very helpful for designing
mixes that are more crack resistant.

Figure 7. Thermal stress and estimated fracture temperature for sinusoidal temperature variation during
a day.
 Road Materials and Pavement Design 13

Figure 8. Thermal stress rate at different cooling rates.

Figure 9. Thermal stress and estimated fracture temperature at constant-rate cooling.

5. Conclusions
The following conclusions can be drawn based on the analysis performed in this study:

• The important rheological properties of bitumen, such as the creep compliance (Equation
(7)), relaxation modulus (Equation (10)), complex modulus (Equation (12)), and relaxation
14 B. Teltayev and B. Radovskiy

spectrum (Equation (14)), can be related with its simplest properties such as penetration
and softening temperature based on experimental data and rheological models (CA model
in this study) using the interrelationships of linear viscoelasticity rather than applying the
purely regression techniques. It would be of interest to enlist a more comprehensive rhe-
ological model such as the 2S2P1D model (Di Benedetto, Olard, Sauzeat, & Delaporte,
2004), for modelling the viscoelastic properties of both bitumen and mixture.
• The direct tensile strength of asphalt mixture as a function of temperature can be esti-
mated depending on properties of binder and mix volumetric composition from empirical
equations (22), (25), and (26). Evaluation of asphalt concrete strength needs a more
comprehensive mechanically based approach to address a ductile-to-brittle transition
at low temperatures, perhaps based on new theory of materials failure (Christensen,
2013).
• The feasibility of constant strain-rate test of asphalt mixture is doubtful of its value for low-
temperature cracking problem because the longitudinal strain in asphalt pavement induced
by cooling is unrealised up to appearance of the transverse crack. Moreover, the constant-
strain loading requirement at various temperatures is difficult to achieve due to the testing
machine compliance. For low-temperature cracking analysis, the strength test at constant
rate of stress loading looks more valuable.
• The proposed method to determine the critical single-event cracking temperature as the
temperature at which the thermal stress curve intersect the constant-rate strength curve
leads to reasonable results.

Additional research is needed to verify the proposed method to calculate the thermal stress
build-up from a single cooling event, to estimate the critical temperature and to extend the
application of the proposed method.

Disclosure statement
No potential conflict of interest was reported by the authors.

Funding
This work was supported by Road Committee of Ministry for Investments and Development of the Republic
of Kazakhstan (Contract No. 36 dated 21.07.2016).

References
AASHTO. (2002). Standard specification for performance graded (PG) asphalt binder. AASHTO Provi-
sional Standards, MP1a-02, Washington, DC.
AASHTO. (2007). Determination of low-temperature performance grade (PG) of asphalt binders. AASHTO
Provisional Standards, PP 42-07. Washington, DC.
Adam, G., & Gibbs, J. H. (1965). On the temperature dependence of cooperative relaxation properties in
glass-forming liquids. The Journal of Chemical Physics, 43, 139–146.
Anderson, D., Christensen, D., Dongre, R., Sharma, M., Runt, J., & Jordhal, P. (1990). Asphalt behaviour at
low service temperatures. Federal Highway Administration-RD-88-078, U.S. Department Transporta-
tion, Washington, D.C., 337p.
ASTM. (2002). Standard practice for determining low-temperature performance grade (PG) of asphalt
binders D 6816-02. Annual book of ASTM standards, road and paving materials: Vehicle-pavement
systems, 04-03. West Conshohocken, Pennsylvania.
Bouldin, M., Dongre, R., Rowe, G., Sharrock, M. J., & Anderson, D. (2000). Predicting thermal cracking
of pavements from binder properties-theoretical basis and field validation. Journal of the Association
of Asphalt Paving Technologists, 69, 455–488.
 Road Materials and Pavement Design 15

Buttlar, W. G., & Roque, R. (1994). Development and evaluation of the strategic highway research pro-
gram measurement and analysis system for indirect tensile testing at low temperatures. Transportation
Research Record, 1454, 163–171.
Christensen, D. W. (1992). Mathematical modelling of the linear viscoelastic behaviour of asphalt cements
(thesis). The Pennsylvania State University.
Christensen, R. M. (2013). The theory of materials failure. Oxford: Oxford University Press.
Christensen, D. W., & Anderson, D. A. (1992). Interpretation of dynamic mechanical test data for paving
grade asphalt cements. Journal of the Association of Asphalt Paving Technologists, 61, 67–116.
Christensen, D. W., Anderson, D. A., & Rowe, G. W. (2016). Relaxation spectra of asphalt binders and the
Christensen-Anderson rheological model. Journal of the Association of Asphalt Paving Technologists,
Meeting Pre-prints, 86, 546–577.
Christensen, D. W., & Bonaquist, R. (2015). Improved Hirsch Model for estimating the modulus of hot mix
asphalt. Journal of the Association of Asphalt Paving Technologists, 84, 527–557.
Christison, J. T., Murray, D. W., & Anderson, K. O. (1972). Stress prediction and low temperature frac-
ture susceptibility of asphaltic concrete pavements. Journal of the Association of Asphalt Paving
Technologists, 41, 494–523.
Dave, E., Buttlar, W. G., Leon, S. E., Behnia, B., & Paulino, G. H. (2013). IlliTC – low-temperature cracking
model for asphalt pavements. Road Materials and Pavement Design, 14(S2), 57–78.
Di Benedetto, H., Delaporte, B., & Sauzeat, C. (2007). Three-dimensional linear behavior of bituminous
materials: Experiments and modeling. International Journal of Geomechanics, 7, 149–157.
Di Benedetto, H., Olard, F., Sauzeat, C., & Delaporte, B. (2004). Linear viscoelastic behavior of bituminous
materials: From binders to mixes. Road Materials and Pavement Design, 5, 163–202.
Di Benedetto, H., Sauzeat, C., Bilodeau, K., Buannic, M., Mangiafico, S., Nguyen, Q. T., . . . Van Rompu,
J. (2011). General overview of the time-temperature superposition principle validity for materials
containing bituminous binder. International Journal of Roads and Airports, 1, 35–52.
Drozdov, A. D. (2001). A model for the viscoelastic and viscoplastic responses of glassy polymers.
International Journal of Solids and Structures, 38, 8285–8304.
Farrar, M. J., Hajj, E. Y., Planche, J. P., & Alavi, M. Z. (2013). A method to estimate the thermal stress
build-up in an asphalt mixture from a single-cooling event. Road Materials and Pavement Design,
14(1), 201–211.
Ferry, J. D. (1980). Viscoelastic properties of polymers (3rd ed.). New York, NY: Wiley.
Fromm, H. J., & Phang, W. A. (1971). Temperature susceptibility control in asphalt cement specifications.
Highway Research Record, 350, 30–35.
Heukelom, W. (1966). Observations on the rheology and fracture of Bitumens and asphalt mixes. Journal
of the Association of Asphalt Paving Technologists, 35, 358–399.
Hills, J. F., & Brien, D. (1966). The fracture of Bitumens and asphalt mixes by temperature induced stresses.
Prepared discussion. Journal of the Association of Asphalt Paving Technologists, 35, 292–309.
Hu, S., Zhou, F., & Walubita, L. F. (2009). Development of a viscoelastic finite element tool for asphalt
pavement low temperature cracking analysis. Road Materials and Pavement Design, 10(4), 833–
858.
Humphreys, J. S., & Martin, C. J. (1963). Determination of transient thermal stresses in a slab with
temperature dependent viscoelastic properties. Transactions of the Society of Rheology, 7, 155–
170.
Marasteanu, M. O., & Anderson, D. A. (1999). Improved model for bitumen rheological characteriza-
tion. Paper presented at the Eurobitumen workshop on performance related properties for Bituminous
Binders, Luxembourg.
Molenaar, A. A. A., & Li, N. (2014). Prediction of compressive and tensile strength of asphalt concrete.
International Journal of Pavement Research and Technology, 7, 324–331.
Monismith, C. L., Secor, G. A., & Secor, K. E. (1965). Temperature induced stresses and deformations in
asphalt concrete. Journal of the Association of Asphalt Paving Technologists, 34, 248–285.
Moon, K. H., Marasteanu, M. O., & Turos, M. (2013). Comparison of thermal stresses calculated from
asphalt binder and asphalt mixture creep tests. Journal of Materials in Civil Engineering, 25(8), 1059–
1067.
Mortazavi, M., & Moulthrop, J. S. (1993). The SHRP Materials Reference Library. SHRP-A-646.
Washington, DC: The Transportation Research Board.
Prieto-Muñoz, P. A., Yin, H. M., & Buttlar, W. G. (2013). Two-dimensional stress analysis of low-
temperature cracking in asphalt overlay/substrate systems. Journal of Materials in Civil Engineering,
25(9), 1228–1238.
16 B. Teltayev and B. Radovskiy

Roy, S. D., & Hesp, S. A. M. (2001). Fracture energy and critical crack tip opening displacement: Fracture
mechanics-based failure criteria for low-temperature grading of asphalt binders. Paper presented at the
proceedings of Canadian Technical Asphalt Association, Toronto, Ontario (pp. 185–212).
Saal, R. N. J. (1955). Mechanical testing of asphaltic bitumen. Paper presented at the Fourth World
Petroleum Congress, Rome, Section VI/A, Paper 3, 1–17.
Schwarzl, F. R. L., & Struik, C. E. (1968). Analysis of relaxation measurements. Advances in Molecular
Relaxation Processes, 1, 201–255.
Shahin, M. Y. (1977). Design system for minimizing asphalt concrete thermal cracking. Paper presented at
the proceedings of fourth international conference on the structural design of asphalt pavements, Ann
Arbor, University of Michigan (pp. 920–932).
Smith, T. L. (1976). Linear viscoelastic response to a deformation at constant rate: Derivation of physical
properties of a densely cross linked elastomer. Transactions of the Society of Rheology, 20(1), 103–117.
Stock, A. F., & Arand, W. (1993). Low temperature cracking in polymer modified binder. Journal of the
Association of Asphalt Paving Technologists, 62, 23–53.
Tschoegl, N. W. (1989). The phenomenological theory of linear viscoelastic behaviour. Heidelberg:
Springer-Verlag.
Ullidz, P., & Larsen, B. K. (1984). Mathematical model for predicting pavement performance. Transporta-
tion Research Record 949, TRB, 45-54.
Van der Poel, C. (1954). A general system describing the visco-elastic properties of bitumens and its relation
to routine test data. Journal of Applied Chemistry, 4, 221–236.