Construction and Building Materials 68 (2014) 572579

Contents lists available at ScienceDirect

Construction and Building Materials

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

Reliability-based assessment of deteriorating performance to asphalt

pavement under freezethaw cycles in cold regions
Wei Si 1, Biao Ma , Ning Li, Jun-ping Ren 2, Hai-nian Wang 3
Key Laboratory for Special Area Highway Engineering of Ministry of Education, Changan University, Xian, Shaanxi 710064, China

h i g h l i g h t s
 Compressive characteristics have been analyzed under freezethaw cycles.
 Reliability method was applied to analyze the pavement reliability functions with various uncertainties.
 Pavement capacity under freezethaw cycles was studied based on reliability method.

a r t i c l e

i n f o

Article history:
Received 25 October 2013
Received in revised form 8 June 2014
Accepted 3 July 2014
Available online 26 July 2014
Keywords:
Freezethaw cycles test
Resilient modulus
Reliability method
Monte Carlo Simulation
Fragility curves

a b s t r a c t
Accurate deterioration models play a critical role in designing and managing transportation infrastructure. Regular models just consider loading factor and its relative uncertainties. However, climate and
environment impacts are not considered or just taken as certain variables. Thermal cracks and moisture
distresses are principal distress forms in cold regions, where early damage is more signicant than general regions. In this paper, FreezeThaw (FT) cycle test was performed to investigate the impact of the
cold climate and moisture resistance on paving mixtures, and the compressive strength and resilient
modulus were studied. Then reliability method was applied to analyze the pavement reliability functions
with various uncertainties. The analytical results showed that the resilient modulus of asphalt concrete
mixture declined under FT cycles. Consequently, pavement structure capacity was reduced. The results
also illustrated that reliability method was capable of accommodating uncertainties in pavement parameters. The sensitivity analysis addressed that FT cycles had a signicant impact on estimating reliability,
especially with the large coefcient of variance. The larger coefcient of variance, the faster reliability
declined. Reliability analysis results indicate that decrease in variability of FT cycles can signicantly
increase the estimated reliability. As a result, FT cycles with uncertainty do harm to pavement loading
capacity, which should not be neglected in actual engineering. This paper also proposes some instructions
on simulating pavement performance models.
2014 Elsevier Ltd. All rights reserved.

1. Introduction
All civil infrastructures are faced with the problem of deterioration, and it has been recognized as a critical issue worldwide [1,2].
Pavement is a signicant part of transportation infrastructure,
which is exposed in atmospheric, and suffers the climatic,
environmental effect, and vehicular load directly [3]. Asphalt is a

Corresponding author. Tel.: +86 (029) 82334646.

E-mail addresses: siweichd@gmail.com, siwei26@utexas.edu (W. Si), mabiaochd@
163.com (B. Ma), 630589426@qq.com (N. Li), jpren1990@163.com (J.-p. Ren),
wanghainian@yahoo.com.cn (H.-n. Wang).
1
Tel.: +86 150 9401 1268.
2
Tel.: +86 151 9146 7341.
3
Tel.: +86 136 0916 1519.
http://dx.doi.org/10.1016/j.conbuildmat.2014.07.004
0950-0618/ 2014 Elsevier Ltd. All rights reserved.

thermoplastic material, and asphalt concrete (AC) mixtures present viscoelastic characteristics as well [3]. Numerous studies have
found out that distresses of pavement have important relationship
with climate and environment as well as loads [3,5]. Previous
researches showed that temperature and moisture are the most
impact factors to asphalt pavement performance [37]. Due to
the sensitivity to temperature and moisture, asphalt pavements
performance and serviceability are prone to deteriorate than other
infrastructures. Thermal cracks, moisture damage and other low
temperature distresses are the key issue for asphalt pavement in
cold regions [4].
Low-temperature cracking and thermal fatigue cracking are the
two forms of thermal cracking in asphalt pavement [3]. When
temperature falls rapidly or in continuous low temperature, temperature stress is formed in asphalt pavement. Low temperature

573

W. Si et al. / Construction and Building Materials 68 (2014) 572579

cracking occurs when the thermal tensile stress in asphalt pavement exceeds its tensile strength [3,4]. If the comprehensive stress
(temperature stress, load stress, etc.) is smaller than the tensile
strength, the interior micro-damage will accumulated rather than
forming cracks. After a large number of FreezeThaw (FT) cycles,
the comprehensive stress exceeds the ultimate tensile strength of
asphalt mixture, which leading to cracks and other apparent distresses [5,6]. Moreover, if the pavement is located in the moisture
environment, hydrodynamic pressure and vacuum constriction
will appear in the surface layer under repeatable vehicle loads
[57]. Water causes the disruption of the bond between the
asphalt and the aggregate at the asphalt-aggregate interface. This
premature failure of adhesion is commonly referred to as stripping
in asphalt concrete pavements [1]. FT cycles will result in stripping aggravation and voids increase on the surface of AC. Several
pavement distresses that can include stripping as the underlying
cause are rutting, cracking, raveling, ushing, and bleeding, furthermore which accelerates the degradation of pavement load
capacity.
During the FT cycles test, the damage of asphalt mixtures is
initiated by ice expansion load and accelerated by the interfacial
damage between asphalt and aggregate or fracture of asphalt mortar. The change of volume should be caused by the expansion of
water under low temperature in mixtures. The expansion will
enlarge the gap voids of the internal composition and reduce the
tensile strength of mixtures. Asphalt-aggregate bond is easily displaced from the aggregate by water under the impact of FT cycles,
which weakens the asphaltaggregate bond and increases the
stripping of aggregate.
Many researchers studied FT of various materials by using a
number of laboratory tests. Most of them were related to the
determination of the effects of FT on the properties of materials
such as strength, compressibility, porosity, pore size distribution,
and permeability. In most cases FT cycling of the samples is
limited to one cycle, or some xed cycles, it has not presented
the mixtures deterioration trend with FT cycles increasing. Lots
of researchers tried to explore the mechanism of AC performance
suffering temperature and moisture in cold regions. However,
due to the differences in experiment, model, and simulation
method, no standard evaluation criteria are built at yet.
In this paper, FT cycle test is proposed to analyze the inuence
of temperature and moisture on AC. During the FT cycles experiment, environment temperature changes from positive to negative
repeatedly, and the samples suffer repetitive thermal tress and
moisture impact. The deterioration of AC can be obtained through
FT cycle test. However, the test result just presents the deterioration trend of AC, which is not connected to the design equation or
pavement performance. Also, previous research neglected to
combine the materials experimental properties with pavement
design and performance evaluation. Consider with these drawbacks, reliability method is applied in this paper to analyze the
asphalt pavement AASHTO design equation based on the variation
of ACs compression properties under FT cycles. Reliability is an
important performance measure of pavement structural
(performance) condition and reliability-based procedures have
the capability of accommodating uncertainties in assessment.
More uncertainties will arise in pavement structure assessment
when AC is under FT cycles. Reliability models are probabilistic
models, which predict the failure probability of a given system.
The limit state function for exible pavements is typically dened
as a difference between the designed load applications. The supply
of a pavement is a criterion that can withstand certain loads before
failing, and the number of load applied is called demand correspondingly [8]. If there is a clear denition of a failure event and
the consequence of the failure, reliability models can be effectively
used to predict the performance and lifetime of pavements [1].

Reliability models distinguish only two conditions or states: a

functional (surviving) state and a failure state. Under FT cycles,
it directly impacts the states of reliability model.
The objectives of this research are to analyze the degradation of
resilient modulus of AC under FT cycles, and to use reliability
method to illustrate the impact of FT on AC structure capacity
behavior by the number of 18-Kip (80-KN) single axle load. The
scope of the paper is limited to the consideration of variability
associated with the response and utilization of pavements.

2. Experimental and reliability methods

2.1. Materials and FT cycle test
2.1.1. Materials used
The materials used in laboratory study were: Asphalt, aggregates, and ller;
which were used as virgin materials. Asphalt is SBR (Styrene Butadiene Rubber)
modied asphalt; aggregates and ller were obtained from lime-stone. Test results
of asphalt are presented in Table 1. Various engineering properties of asphalt,
aggregates and ller have met specications and standards through the experimental test.
AC used in this research with the 13 mm nominal maximum aggregate size AC13 mixture, as recommended by the specication of Ministry of Transportation of
China, which is listed in Table 2[9].
The optimum asphalt content (OAC) of the asphalt mixture was obtained by
applying Marshall Test: the cylindrical specimens made by the standard compaction hammer and cylindrical mold were 101.6 mm in diameter and 63.5 mm in
height [9]. On the basis of Marshall Test and considering the climate of cold regions
as well as trafc conditions, the nal OAC of AC-13 is 5.5% in this research.

2.1.2. Freezethaw cycle test

FT cycle test rst used in civil engineering is to evaluate the impact of FT
cycles on the performance of cement concrete [10,11] and soil [12,13]. As the application of exural pavement increases in cold regions, some researchers noticed that
asphalt pavement also suffered FT cycles inuence.
In terms of AC FT cycle test, there is no standard due to the differences on climate conditions. In AASHTO T283 testing procedure, the conditioned specimens
were vacuum saturated to 7080% saturation. Each vacuum saturated specimen
was tightly covered with plastic wrap and placed in a plastic bag with approximately 10 0.5 mL of water, and sealed. The plastic bags were placed in a freezer
at 18 3 C for 24 h. The specimens were removed from the freezer and placed
in a water bath at 60 1 C for 24 h with 25 mm of water above the specimens.
After 24 h in the 60 1 C water bath, the specimens were removed and placed in
a water bath at 25 0.5 C for 2 h 10 min to achieve room temperature [14]. In
one research, the samples were soaked in water for 12 h at room temperature, then
each was covered with plastic wrap, sealed in a plastic bag, and placed in an environmental chamber for 12 h at 17.8 C.The conditioned samples were exposed to
seven 24 h cycles [15]. One of the typical researches in China was rstly water conditioned by vacuum saturation for 15 min, and then subjected to 8 successive cycles
of freezing and thawing. Each cycle was consisted of freezing at 20 C for 8 h followed by soaking in water at 60 C for 4 h [5].
Considering present FT cycle test, a modied FT cycle test was proposed in
this paper. In the FT cycle test, specimen was placed in the plastic bag and
30 ml water was injected in the plastic bag, and sealed. Freeze temperature is
25 1 C and lasts12 h; then put the specimens into water-bath to be thawed at
25 1 C and lasts12 h as well. The freezer and water-bath are used to simulate
the FT cycles.
Table 1
Test results of asphalt parameters.
Asphalt state

Test item

Test
temperature

Result

Original
sample

Penetration/0.1 mm

15 C

51

Softening point (R&B) (C)

Ductility (cm)
Density (g cm3)
Flashing point (C)
Solubility (%)

25 C

5 C
25 C

123.1
47.6
>150
1.023
>260
99.6

15 C
25 C
5 C

0.2
65.3
58.1
>150

After aging

Mass Loss (%)

Vestigital penetration ratio (%)
Ductility (cm)

574

W. Si et al. / Construction and Building Materials 68 (2014) 572579

Table 2
Gradation composition of aggregates mixture.
16
100

13.2
95

9.5
76.5

4.75
53

2.1.3. Compression test and calculation formulas

The resilient modulus laboratory test was carried out by Universal Material Tester System (UMTS) which was performed by the uniaxial compression test. The
UMTS is a computer controlled system which can operate automatically. Loads
and deection of specimen was measured with computer by linear variable differential transducers (LVDTs). UMTS equipped environment chamber with an accuracy
of 0.1 C, and the loading speed could be chosen as experiment needed.
The sample is a cylindrical shape with 100 2 mm in diameter and height,
which was shaped using static pressure method. 15 0.5 C and 2 mm/min were
adopted as test temperature and load speed in this research. After loading test
under FT cycles, the performance of asphalt mixture was evaluated by compressive strength and resilient modulus, which can be calculated as follow [9]:

Rc

4P

pd2

where Rc is the compressive strength (MPa), P is the ultimate loading at failure (N), d
is the diameter of specimen (mm).
After compressive strength test, the ultimate loading was divided into ten average levels, which were used to determine the loadings in resilient modulus test.
0.1P, 0.2P, 0.3P, . . ., 0.7P were selected as the loadings.

qi

4P i

pd

Pi is the ith loading level (N), qi is the corresponding pressure under Pi (MPa).
Function (2) was used to calculate the actual pressure qi under the different
loading levels. By plotting scatter points in the coordinates, where pressure qi is
taken as the vertical ordinate and resilient deformation DLi as horizontal ordinate,
the fth loading level (0.5P) with pressure q5 and DL5 were selected to calculate the
resilient modulus. Formula is shown as

E0

q5  h
DL5

E0 is the resilient modulus (MPa), h is the sample height (mm). Mean value of four
specimens under each FT cycles is applied to analyze the compressive characteristics in this paper.
2.2. Reliability theory
One of the most traditional structural reliability methods is the StressStrength
(Demand-Supply) inference method. This method compares a random variable that
denes the level of supply and another random variable that species the applied
loads or demand; a failure occurs when the level of demand exceeds that of supply
[1,16,17]. A graphical representation of the demand-supply inference method is
presented in Fig. 1. As can be observed from Fig. 1, the non-compliance (failure)
occurs in the inference region of supply and demand distributions, where the
non-compliance region is proportional to the non-compliance probability [16].
The resulting factor of safety is called the reliability index (b), which increases
in value as overlap area increase (the supply over the demand). The fundamental
considerations in the structural reliability theory are: (1) mathematical formulation
of the limit state function; (2) characterization of the basic random variables; and
(3) evaluation of the multidimensional probability integral [1]. Reliability function
can be uncouples into the Supply and Demand functions, which are typically two
separate areas of study [17]. More specically, the structural reliability model is formulated in terms of n basic random variables X = [x1, ..., xn]T, and a limit state function G(x) which is the function that separates the failure and non-failure domains, S
and D are assumed supply and demand respectively, then Eq. (4) represents the
limit state function.

Gx S  D

With a dened limit state function and the structural failure expressed as an
event fGx 6 0g; mathematically, the probability of failure can be expressed as
an n-dimensional probability integral.

pf probGx 6 0

f xdX

Gx60

where f(x) is joint probability density function of the basic random variables X. Then
reliability can be expressed as R = 1  pf.
The results of the limit state function can be used to calculate the probability of
failure (non-compliance) and the b value. Ang and Tang described a process that can
be used to derive the expected value and variance of a design parameter [18]. This
process can also be used to derive the measure of safety, ES-D, given in Eq. (6).

2.36
37

1.18
26.5

0.6
19

0.3
13.5

0.15
10

0.075
6

ESD ES  ED

where E(S) and E(D) are the expected values for S and D. Since
can be shown that b is dened as noted in Eq. (7).

rSD

q
r2S r2D , it

ES  ED
b q
r2S r2D

The probability of non-compliance is an estimate chance that an engineering

system fails to perform its stated purpose under anticipated operational conditions.
Finally the probability of non-compliance can be expressed as:

pf probGx 6 0 Ub

where U denotes the standard normal distribution function. The failure occurs in the
non-compliance of strength and stress distributions, where the failure region is
proportional to the failure probability.
2.3. AASHTO design equation
The AASHTO design method is one of the most widely used methods for
designing exible pavement structures. This design procedure is based on the
results from the accelerated pavement testing, known as the AASHO Road Test
[19]. Without considering the term that corrects for the overall variance, the
strength of a pavement is dened by the AASHTO design equation as follows:

log W 18 9:36 logSN 1  0:20

logDPSI=4:2  1:5
0:4 1094=SN 1

5:19

2:32 log M r  8:07

where W18 is allowable number of the equivalent 18-kip (80-kN) single-axle loads
(ESAL) to cause the reduction of the serviceability level by the amount of DPSI; SN
is structural number; and Mr is effective resilient modulus of roadbed soil.
SN has been described as an abstract number representing the structural
strength of a pavement, which can be calculated as follow [19].

SN a1 D1 a2 D2 m2 a3 D3 m3 :

10

In which a1, a2, and a3 are layer coefcients for the surface, base, and sub-base,
respectively; and the corresponding thickness are D1, D2, and D3 respectively; and
m2, m3 are the drainage coefcient for the base course, and sub-base course
respectively.

3. Results and discussion

3.1. Impact of FT cycles
In order to illustrate FT cycles impact on mixture compression
characteristics, AC-13 with OAC 5.5% was taken for test. The FT
test method is extremely laborious and time consuming, 14 FT
cycles was selected to illustrate the impact of FT cycles on asphalt

Demand (Stress)
Probability Density Function

Sieve size (mm)

Passing rate (%)

Supply (Strength)

Non-compliance
(Failure region)
System Variable
Fig. 1. The probability of non-compliance.

575

W. Si et al. / Construction and Building Materials 68 (2014) 572579

2.2

Resilient modulus, MPa

Compressive Strength, MPa

800

2.0

1.8

1.6

760

720
680

640

600
0

10

12

14

16

10

12

Freeze-Thaw cycles, cycle

Freeze-Thaw cycles, cycle

(a) Compressive strength

(b) Resilient modulus

14

16

Fig. 2. Impact of FT cycles on compressive characteristics.

mixtures. The tendency of compressive strength and resilient modulus after14 FT cycles are shown in Fig. 2.
Fig. 2 presents compressive strength and resilient modulus of
asphalt mixture, which tend to decline with the increasing of FT
cycles. In initial FT cycles, compression characteristics drop
quickly; after 8 FT cycles, the decline of compression characteristics become gentle; nally, as FT cycles increase, the degradation
tendency gradually turn to stable. After 14 FT cycles, the
compressive strength reduces nearly 0.33 MPa comparing with
unconditioned (has not experienced with FT cycles, i.e. FT cycle
equals 0) performance, the attenuation of which reaches to 15.6%.
The resilient modulus reduces nearly 147 MPa (attenuates 18.7%).
In Fig. 2, there are some abnormal points, which decrease
sharply during a certain FT cycles and then increase during the
next FT cycle. In terms of this phenomenon, the uneven aggregate
distribution takes main responsibility. Experimental error may also
have some inuence as well.
It is observed that the curves of compressive strength and
resilient modulus are nonlinear, and the variation trends can be
simulated by exponential form. Therefore, exponential model is
applied to simulate the variation of compressive characteristics
as the FT cycles changes. The used formula list as below [20,21]

y a b  ecx e

e  N0; d2

11

where y is the explained variable, here denotes the compressive

strength and resilient modulus (MPa); a is the parameter that represents the initial condition; b is the parameter that represents the
rate at which quality deteriorates with FT cycles; c is the parameter that represents the curvature of the function; x is the FT cycles;
and e is the error term. The mean of the random variable e is
assumed to be zero; the variance is homogeneous and uncorrelated
with the dependent variables.
Parameters are estimated using a least-square method. Due to
the random variable, e follows normal distribution. Chi-square
value is used to evaluate the quality of tting parameters [20,21].

After simulation, the tting results of exponential model are

listed in Table 3.Results show that exponential model well reects
the variation of compressive strength and resilient modulus; the
R2adj reached 0.95 and 0.88, respectively. ANOVA result indicates
that the test data and regression model have excellent relation at
the 0.01 signicant level from statistic view.
Mixture internal pores and air voids increase under the repetitive FT cycles. When AC is in moisture environment, water will ll
the air voids and turn into ice in freezing cycles, causing the volume expansion. This generates expansion force and results in the
micro-damage to mixture [5,15,22]. During the thaw cycles, with
the ice melting, more water will ll air voids and internal pores.
The bonding between asphalt membrane and aggregates decreases
under the water erosions [5,15,22].Therefore, after several FT
cycles, the bond force decreases dramatically, which leads to the
mixture loosens and fails.
Previous researches show that asphalt, aggregate composition,
and the same asphalt mixed with different aggregates (limestone
or granite) play an important role in the FT cycles impact on
asphalt mixtures. While this research just focuses on the deterioration trend of mixtures compressive characteristics with FT cycles
increasing, and the physico-chemical principles during FT cycles
is not considered. Hence SBR modied asphalt with AC-13 gradation is taken as example, which is applied to study the pavement
design and performance evaluation based on the uncertainty under
FT cycles.
3.2. Impact of asphalt-aggregate ratio
The resilient modulus with different asphalt-aggregate ratio
4.5%, 5.0%, 5.5%, 6.0%, and 6.5% under FT cycles is compared,
and the test results are shown in Fig. 3.
As shown in Fig. 3, the resilient modulus with different
asphalt-aggregate ratio decreases as FT cycles increase. With
asphalt-aggregate ratio increasing, the resilient modulus increases
at the beginning and then turns to decline, which presents

Table 3
Fitting results of exponential model.
a

Compressive strength
Resilient modulus

Statistics

Value

Error

Value

Error

Value

Error

Reduced chi-sqr

R2adj

1.45
625.33

0.38
23.80

0.67
151.92

0.37
25.58

0.06
0.21

0.05
0.11

8.99E04
366.67

0.95
0.88

576

W. Si et al. / Construction and Building Materials 68 (2014) 572579

700

Resilient Modulus, MPa

In which a1 is the AC layer coefcient; E is the resilient modulus of

asphalt mixture, which are suffered with FT cycles in this research.
The relationship between AC coefcients and resilient moduli is
generally based on layered elastic theory.
The AC coefcient a1(FT) and the SN after FT cycles can be
obtained by applying Eqs. (12) and (10). Due to the uncertainty
of resilient modulus under FT cycles, the SNFT maintains the
same uncertainty degree as resilient modulus.

4.5%
5.0%
5.5%
6.0%
6.5%

800

600
500
400
300

3.4. Developing reliability function for AASHTO design equation

200
100
0
0 2 4 6 8 10 12

14

0 2 4 6 8 10 12

14

0 2 4 6 8 10 12

14

0 2 4 6 8 10 12

14

0 2 4 6 8 10 12

14

Freeze-Thaw cycles, cycle

Fig. 3. Impact of asphalt-aggregate ratio on resilient modulus.

unsymmetrical bell shape. The peak value appears in the OAC;

asphalt-aggregate ratio at 5.0% has higher resilient modulus than
that asphalt-aggregate ratio at 6.0%. After 14 FT cycles, resilient
modulus of 5.0% and 5.5% are obviously greater than others. Furthermore, the attenuation of resilient modulus declines when
asphalt-aggregate ratio over 5.5%.
From Fig. 3 it can be seen that after 4th FT cycles, the resilient
modulus with an asphalt-aggregate-ratio of 5.0% declines sharply,
which is evidently smaller when asphalt mixtures suffered with
6th FT cycles compared with it suffered with higher FT cycles.
For this abnormal phenomenon, the uneven aggregate distribution
may take main responsibility, experimental error may also have
some inuence as well. If the resilient modulus with an asphaltaggregate ratio of 5.0% at 6th FT cycles is larger than it at 8th
FT cycle, the resilient modulus also presents decline trend.
Mixture resilient modulus is related to mixture porosity, aggregate distribution and internal structure defects. The internal damage will be enlarged after several FT cycles and this result in the
deterioration of mixture performance. Therefore, appropriate
asphalt-aggregate ratio limits the degradation.
In order to get the uncertainty of resilient modulus, ShapiroWilk method is used to test the distribution of resilient modulus
after FT cycle test. The results of ShapiroWilk test show that
the data was draw from a normally distributed population at the
0.05 level. The coefcient of variation (CV) of resilient modulus
under FT cycles is assumed to10%in this paper.
3.3. Calculation of SN under FT cycles
The strength and modulus of AC decline as FT cycles increases,
which results in reduction of pavements structural capacity.
Former researches show that SN is able to reect the structural
capacity of the pavement structure. As shown in Eq. (10), layer
coefcients, thickness, and drainage coefcient codetermine the
SN of a pavement. This paper concentrated on the impacts of FT
on AC layer coefcient.
The coefcients are indicators of the relative ability of a material to function as a structural component in the asphalt pavement.
According to the AASHTO 1986 Guide, the resilient modulus of a
material is the primary means to determine the layer coefcient
of the material. Layer coefcients for layers made of different
materials can be determined using empirical equations derived
from eld experiments. The following AC coefcient equation is
the commonly used relationship [23]

a1 0:4  log


E
0:44 0:20 < a1 < 0:44
3000 Mpa

12

The reliability function in pavement engineering applications,

the Supply analysis deals with the material properties and structural design characteristics contributing to the pavement strength;
whereas, the Demand analysis concerns trafc prediction [24].
The demand equation can be dened as the accumulated number of the equivalent 18-kip (80-kN) single-axle loads during the
elapsed time period. Asphalt Institute and AASHTO pavement
design guide recommend the use of a growth factor that considers
trafc over the entire design period, with the total trafc growth
factor expressed as [3,19]

TGF

1 rt  1
r

13

where TGF is total growth factor; r is yearly rate of growth; and t is

design period. With a modied yearly trafc in ESALS for the initial
year (ESAL0) and the total growth factor TGF, the demand or the
accumulated ESALS for an arbitrary time period t can be obtained as

Nt ESAL0  TGF ESAL0 

1 rt  1
r

14

Given that the demand is dened by Eq. (14) and the supply by
Eq. (9), the limit state function can be expressed as

GSNFT ; DPSI; M r ; ESAL0 ; r; t log W 18  log Nt

15

Base on the limit state function, the non-compliance (failure

domain) can be dened as fGSNFT ; DPSI; Mr ; ESAL; r; t 6 0g, and
the probability integral as follow:

Ft

f SNFT ; DPSI; Mr ; ESAL0 ; r; t

GSN FT ;DPSI;Mr ;ESAL0 ;r;t60

 dSNFT dDPSIdMr dESAL0 dr

16

Calculation of the integral dened in Eq. (16) is a tough

problem. Various approximation techniques such as Monte Carlo
Simulation (MCS), the rst order reliability method (FORM), and
the second order reliability method (SORM) are commonly used
[1,17,24,25]. FORM is an analytical method based on linear approximation of a nonlinear limit state at the design point. The main
advantage of the FORM is its analytical traceability and satisfactory
level of accuracy even for extremely small probabilities, while the
shortcomings are lack of accuracy for highly nonlinear limit state
function and the difculty in its iteration-based process of searching for the design point [1]. FORM shortcomings can be addressed
by using the second-order reliability method (SORM), and other
high-order of moments. Zhang and Damnjanovic used method of
moments to analyze pavement reliability function with the limit
state function of the design equations from AASHTO [1]. MCS is
the computational algorithms that rely on repeated random sampling to obtain numerical results. MCS is preferred over the FORM,
whose accuracy is increasing as sample size increases; meanwhile,
MCS is applicable to a wider range of problems. With the computer
technology and programming development, the limitations of MCS
about computationally intensive and difcult implemented in

577

W. Si et al. / Construction and Building Materials 68 (2014) 572579

mathematical programming are not the critical issue presently.

Considering this, MCS has been applied to analyze the limit state
function of reliability in this research.
The approximation given by Eqs. (8) and (16) can be improved
by subsequently employing a MCS. In MCS, the probability of
failure is approximated as [25]

pf 

N
^
1X
IGX j 6 0
N j1

Consequently, the Supply of reliability models has been changed under FT cycles; and it is different from present research that
Supply is always hypothesized as a constant value, while the
Demand changes with time (or loads). In this reliability model,
the Supply changes as well as Demand; Supply is also presented
by fragility curves with uncertainty, which made the reliability
model more complicated and practical to pavement deterioration
in cold regions.

17
3.5. Reliability analysis
^

where N is the total number of simulations conducted, X j represents

the jth randomly simulated vector of basic variables, and I[] is the
indicator function, having the value 1 if [] is true and the value 0
if [] is false.
As SNF-T deals with uncertainty, the Supply of pavement structure varies in the reliability model. The Supply curve becomes fragility curves instead of the constant curve. With the variation of
uncertainty, fragility curves of Supply can be divided into different
performance levels obtained by evaluating the limit state function
conditioning on Demand. Once the fragility of pavement is established, the designer has a exibility to specify any trafc utilization
model for predicting pavement reliability, and take reasonable preventive maintenance action as well.
Pavement parameters, except the resilient modulus of AC, used
in this paper were recommendation by Huang [3]. The thickness of
AC layer is 203 mm (8 in); the base layer is 178 mm (7 in); and the
sub-base layer is 292 mm (11.5 in). The resilient modulus of base
and subbase are 207 MPa (30,000 psi) and 76 MPa (11,000 psi),
respectively. The mean value of the effective resilient modulus of
roadbed soil is 39.3 MPa (5,700 psi), and DPSI is 2. The drainage
coefcients for the base course and sub-base course have the same
value 1.2. The layer coefcients for the base and the sub-base are
0.14 and 0.08, respectively. The fragility curves of log W18 under
FT cycles are shown in Fig. 4.
Fig. 4 shows that log W18 decreases as FT cycles increase.
Result indicates that even though the loading and other articial
impacts have been neglected, FT cycle still has a signicant inuence on the Supply. The degradation trend of Supply is severe in
initial FT cycles, and then becomes gentle with FT cycles
increase. The uncertainty of Supply contributes to the increase of
the non-compliance area as Supply reduces. Apparently, for a
new asphalt pavement, even though there are no vehicle loading
or other articial activities, deterioration also appears under the
climate and environment effects [3,5]. It makes sense that pavement in atrocious weather regions has more critical immature
damages than general regions.

In order to analyze the impact of FT cycles on pavement structure capacity, this paper gives the predicted number of ESAL in
design period as 5 million, and the design life is 50 years. The basic
random variables in the limit state function are assumed to be
independent and normally distributed. There are four random variables, which are modied yearly ESAL, resilient subgrade modulus,
yearly ESAL growth rate and SN under FT cycles. The mean of
modied yearly ESAL, resilient subgrade modulus, and yearly ESAL
growth rate is 109,261, 5700 and 0.05, respectively; and SN
changes with FT cycles increase. The CV of the four basic random
variables is 10%, 15%, 10%, and 10%, respectively [3]. Furthermore, a
sensitive analysis is conducted to illustrate the SN variation and
uncertainty under FT cycles, which intend to explain the impact
and effect of uncertainty in pavement structure capacity. The CV
s of 5%, 10%, 15%, 20%, and 30% on SN are applied. Due to the large
sample size used for the MCS, the MCS estimates should be close to
the actual level. For each simulation, the failure probability was
estimated with a MCS of 1 million samples.
Fig. 5 illustrates the variation of estimated reliability of different
CV s under FT cycles at 10th year. For different CV s, all the estimated reliabilities present decline trend with FT cycles increase.
Furthermore, with the increasing of CV, the decline trend turns to
be magnied. At the 5% CV, the estimated reliability is 99.5% at
the unconditioned case, and after 14 FT cycles it is 94.8%; the estimated reliability reduces 4.7% compared with unconditioned case,
which accounts for 4.73%. While at the 20% CV, the estimated reliability is 81.5% at the unconditioned case, and after 14 FT cycles it
turns to 72.1%. At the 15% CV, when it suffered with 14 FT cycles
the estimated reliability reduces 10.3%, the reduction accounts for
11.7% compared with unconditional.
Fig. 5 also presents that for the pavement structure at the same
CV, the estimated reliability declines with FT cycles increase. For
the unconditional case, the estimated reliability is 99.5%, 95%,
87.7%, and 81.5% at the 5% CV, 10% CV, 15% CV, and 20% CV
respectively. While when it experienced 8 FT cycles, the
estimated reliability is 95.8% and 74% at the 5% CV and 20% CV

Demand (Stress)

Fragility
curves

0 6.5
2
4
6
8
10
12
14
16

6.6

6.7

CV=5%
CV=15%

Degradation trend
of supply
Uncertainty

Supply (Strength)

Reliability

F-T cycles

Probability Density Function

1.0
log W18

CV=10%
CV=20%

0.8

Non-compliance
(Failure region)
0.0

System Variable
Fig. 4. Fragility curves of Supply under FT cycles.

0 2 4 6 8 10 12

14

0 2 4 6 8 10 12

14

0 2 4 6 8 10 12

14

Freeze-Thaw cycles, cycle

Fig. 5. Impact of FT cycles on reliability.

0 2 4 6 8 10 12

14

578

W. Si et al. / Construction and Building Materials 68 (2014) 572579

1.0
0 cycle
0 cycle
0 cycle
0 cycle
4 cycles
8 cycles
14 cycles

Reliability

0.8

0.6

CV=5%
CV=10%
CV=15%
CV=20%
CV=10%
CV=10%
CV=10%

0.4

0.2

0.0
0

10

15

20

25

30

35

40

45

50

Time, year
Fig. 6. Sensitivity of reliability function of FT cycles and standard deviation on
resilient modulus.

respectively. And when it suffered with 14 FT cycles, the estimated reliability is 94.8% and 72.1% at the 5% CV and 20% CV
respectively. Comparing the estimated reliability at the 5% CV
and 15% CV, the difference at the unconditioned is 11.8%, and it
becomes 17.4% after 14 FT cycles. Results indicate that FT cycles
have prominent impact on estimated reliability, especially for
higher CV. In general, the larger CV, the more signicant impact it
will be.
Fig. 6 illustrates the change in estimated reliability when different of FT cycles and CV s of the environment parameter changes.
After construction pavement structural reliability is the maximum;
then, deterioration initiates as a result of the vehicle loading and
the environmental conditions. It can be seen from Fig. 6 that all
the estimated reliabilities present decline trend with time increasing, and the erce phase is focused on 1030 years. During the
whole construction period (50 years), comparing the estimated
reliability of unconditioned case (0 cycles, CV = 10%), the reliability
with the same CV (CV = 10%) shows decline trend with FT cycles
increases, and these decline curves present almost parallel with
time increasing.
For the unconditioned case (CV = 10%), after 17 years operation,
the reliability of pavement structure capacity turns to below 80%,
after 20 years it turns to below 70%, and after 25 years it just has
the half reliability compared with the initial state. For the same
CV (CV = 10%) at the 10th year, the reliability is 95.2%, 91.5%,
87.8%, and 85.8% when it suffered with 0, 4, 8, and 14 FT cycles
respectively; at the 15th year, the reliability is 84.1% and 65.7%

when it suffered with 0 and 14 FT cycles respectively; while at

the 20th year, the reliability becomes 68.7% and 43.9% when it
suffered with 0 and 14 FT cycles respectively. Result indicates
that the estimated reliability is sensitive as FT cycles increase,
the difference between unconditioned and experienced 14 FT
cycles cases (CV = 10%), the reliability is 9.4% at 10th year, and it
becomes 24.8% at 20th year.
Fig. 6 also presents that for the unconditioned cases, the estimated reliability presents decline trend with the increasing of CV
in the rst 25 years. However, the variation of estimated reliability
shows inverse trend after 25 years; the larger CV, the higher reliability. Consequently, the Year 25 is the inection point for the
same FT cycles at different CV s. One reason may be used to
account for this phenomenon, the larger CV corresponding the
higher uncertainty; and the uncertainty will continual accumulating with time increasing. We know that deterioration and ageing
happens to materials and structures with time increasing, trafc
composites and loads also changes with time increasing, so it is
impossible to precisely predict the pavement performance for
several decades based on the present information. The larger
uncertainty results in the estimated reliabilitys variation range
expand, therefore when time over 25 years the larger CV corresponds higher reliability like Fig. 6 presents. When time increases
to a certain degree, the reliability range will expand with the
increase of uncertainty. Results indicate that a decrease in variability of FT cycles can signicantly improve the estimated reliability,
which also state that with uncertainties increase (larger CV), the
engineer will be faced with more puzzles and risks.
The recommended critical values of reliability level to urban
and rural interstate freeway are 85% and 80% respectively [19].
For the unconditioned case, the degradation of reliability present
different under the changing of CV, when the reliability declined
to the critical values, the corresponding service time as shown in
Fig. 7(a). For the xed CV (CV = 10%), the service time of the critical
values of reliability is changing with FT cycles increasing, which
as shown in Fig. 7(b).
Fig. 7(a) shows that when the estimated reliability declined to
the critical values, the service time is different with the variation
of CV s. For the critical values of 85% and 80%, at 5% CV the service
time is 22 and 23 years respectively; while at 30% CV the service
time is 7 and 10 years respectively. Fig. 7(b) presents that when
the estimated reliability reached to the critical values, initial FT
cycles has an important impact on its service time. With FT cycles
increases, the service time of 85% reliability level is 18, 16, and
14 years for 0, 4, and 8 FT cycles respectively; and the service time
of 80% reliability level is 20, 17, and 15 years for 0, 4, and 10 FT
cycles respectively. Results indicate that CV and FT cycles both

85% Reliability level

80% Reliability level

85% Reliability level

80% Reliability level

20

15

Time, year

Time, year

20

10

15

5
10
0
5

10

15

20

30

10

15

20

Coefficient of variation, %

(a) Unconditioned with CV changing

30

0 1 2 4 6 8 10 12 14

0 1 2 4 6 8 10 12 14

Freeze-Thaw cycles, cycle

(b) Fixed CV with F-T cycles changing

Fig. 7. Time variation under the critical reliability values.

W. Si et al. / Construction and Building Materials 68 (2014) 572579

have a signicant impact on service time, uncertainties and FT

cycles should be valued in the actual cold road engineering.
When asphalt pavement subject to FT impacts, it suffers accelerated damage leading to a reduced pavement life. As a result, the
asphalt pavement reduces in performance and increases in maintenance and rehabilitation costs. The time of failure is given by the
time at which an unacceptable level of service is reached by the
pavement structure or the delayed of repair work (maintenance
or rehabilitation) would increase the cost to an unacceptably high
level. Therefore, the suitable maintenance and rehabilitation
should be taken before the estimated reliability reached to the critical value. The pavement capacity and performance will be recovered to an acceptable level; it will be more economical and
efcient. Furthermore the reliability gures present that the
decline trends of estimated reliability are not constant, there are
some points corresponding the higher decline ratio. Based on these
variations, the state highway agencies can propose the corresponding maintenance plans, which can be used to guide the highway
construction and management.
In fact, due to the change of FT cycles, other mechanical properties of AC changes as well as resilient modulus. The penetration
of AC increases compared with that of unconditioned, since water
erosion makes the inner pores and air voids expanded. This makes
stripping aggravation, and accelerates the deterioration of structure capacity as well. Meanwhile, swelling and frost heave have
important impact on the layer coefcients and drainage coefcients of pavement structure, which also results in the loading
capacity decline. This paper just considered the inuence of FT
cycles on the AC layer coefcient, which simplies and weakens
the impact of FT cycles on actual pavement. In actual engineering,
the impact of FT cycles is more critical, therefore FT cycles as
well as other climate, environment and geological factors should
not be neglected in pavement performance assessment.
4. Findings and conclusions
This work was conducted to apply reliability method to explore
the effect of FT cycles with uncertainties on pavement structure
capacity. Main conclusions drawn from this study are summarized
as:
The compressive strength and resilient modulus of asphalt mixture decline when they suffered with FT cycles. Compressive
characteristics keep declining as FT cycles increases, which
declines rapidly in the initial FT cycles and then becomes gentle
after 8 cycles. Exponential model is used to simulate the variation
of compressive characteristics with the FT cycles changing; the
tting and ANOVA results indicate that the test data and regression
model have excellent relation.
Asphalt-aggregate ratio has an obvious inuence on the resilient modulus. With asphalt-aggregate ratio increasing, the resilient
modulus presents unsymmetrical bell shape; and the peak value
appears in the OAC (5.5%).
The estimated reliability presents decline trend as FT cycles
increase, and it turns to be magnied with the increasing of CV.
The sensitivity analysis is conducted to evaluate the effect of variation in FT cycles impact on SN with the reliability measures; calculation results indicate that the FT cycles variation has an
evident impact on the reliability, meaning that sometimes it is
more cost-effective to assess the impact of FT cycles than to
design a pavement with a higher level of capacity. FT cycles and
CV both have a signicant impact on pavement service time,
especially for higher reliability level demand highway; the suitable
maintenance and rehabilitation should be taken before the

579

estimated reliability reached to the critical value. Results indicate

that a decrease in variability of FT cycles can signicantly increase
the estimated reliability; and uncertainties and FT cycles should
be valued in the actual cold road engineering.
Acknowledgments
The author Wei Si Gratefully acknowledges the China
Scholarship Council (CSC) for the nancial support to study at
The University of Texas at Austin. This research was supported
by the National Science & Technology Support Program during
the 12th Five-year Plan Period (No. 2014BAG05B04), the Construction and Science Project of the Ministry of Transport of China (No.
2013 318 490 010), the Doctor Postgraduate Technical Project of
Changan University (No. 2014G5210002). The authors thank Dr.
Zhanmin Zhang, Zhe Han, Kan Long for their help.
References
[1] Zhang Z, Damnjanovi I. Applying method of moments to model reliability of
pavements infrastructure. J Trans Eng 2006;132(5):41624.
[2] Hong F, Prozzi JA. Estimation of pavement performance deterioration using
Bayesian approach. J Inf Syst 2006;12(2):7786.
[3] Huang YH. Pavement analysis and design. 2nd ed. New Jersey: Prentice-Hall;
2004.
[4] Si W, Ma B, Xiao N, Gesang ZR. Analysis on compression characteristics of
asphalt mixture under freezethaw cycles in cold plateau regions. J Highway
Trans Res Dev 2013;30(4):448.
[5] Feng D, Yi J, Wang D, Chen L. Impact of salt and freezethaw cycles on
performance of asphalt mixtures in coastal frozen region of China. Cold Reg Sci
Technol 2010;62(1):3441.
[6] Kettil P, Engstrm G, Wiberg NE. Coupled hydro-mechanical wave propagation
in road structures. Comput Struct 2005;83(21):171929.
[7] Hbeda P. Water sensitivity of asphalt pavements. Swedish National Road and
Transport Research Institute (VTI) VTI note 35, 1998.
[8] Alsherri A, George KP. Reliability model for pavement performance. J Trans Eng
1988;114(3):294306.
[9] JTG E20-2011. Standard test material of bitumen and bituminous mixtures for
highway engineering. Beijing; China Ministry of Transport: 2011. [in Chinese].
[10] Fagerlund G. Scaling, absorption and dilation of cement mortars exposed to
freezing and thawing in NaCl solution. In: Marchand J, Pigeon M, Setzer MJ,
editors. Freezethaw durability of concrete. London: E&FN Spon; 1992.
[11] Jacobsen S, Sellevold EJ. Frost/salt scaling and ice formation of concrete: effect
of curing temperature and silica fume on normal and high strength concrete.
RILEM PROCEEDINGS 30. FreezeThaw Durability of Concrete, 1997. p. 9306.
[12] Talamucci F. Freezing processes in porous media: formation of ice lenses,
swelling of the soil. Math Comput Model 2003;37:595602.
[13] Padilla F, Villeneuve JP, Stein J. Simulation and analysis of frost heaving in
subsoils and granular lls of roads. Cold Reg Sci Technol 1997;25(2):8999.
[14] AASHTO. Resistance of compacted asphalt mixtures to moisture-induced
damage. AASHTO T 283, Washington, DC, 1991.
[15] Goh S, Akin M, You Z, Shi X. Effect of deicing solutions on the tensile strength
of micro- or nano-modied asphalt mixture. Constr Build Mater
2011;5(1):195200.
[16] Sundararajan C. Probabilistic structural mechanics handbook. New
York: Chapman & Hall; 1995.
[17] Ditlevsen O, Madsen HO. Structural reliability methods. Chichester: Wiley;
1996.
[18] Ang AH-S, Tang WH. Probability concepts in engineering: emphasis on
applications to civil and environmental engineering. 2nd ed. New
York: Wiley; 2007.
[19] AASHTO. AASHTO guide for design of pavement structures. AASHTO,
Washington, DC, 1993.
[20] Washington SP, Karlaftis MG, Mannering FL. Statistical and economic methods
for transportation data analysis. Chapman & Hall; 2011.
[21] Greene W. Econometric analysis. 5th ed. Upper Saddle River, NJ: Prentice-Hall;
2002.
[22] Merbouh Mhammed. Effect of thermal cycling on the creep-recovery
behaviour of road bitumen. Energy Procedia 2012;18:110614.
[23] Bahia Hussain U, Bosscher PJ, Christensen J, Hu Yu. Layer Coefcients for New
and Reprocessed Asphaltic Mixes. No. WI/SPR-04-00. Wisconsin Dept. of
Transportation, Division of Transportation Infrastructure Development,
Bureau of Highway Construction, Technology Advancement Unit, 2000.
[24] Deshpande Vighnesh P, Damnjanovic Ivan D, Gardoni Paolo. Modeling
pavement fragility. J Trans Eng 2009;136(6):5926.
[25] Melchers RE. Structural reliability analysis and prediction. 2nd ed. West
Sussex, England: John Wiley & Sons; 1999.