Construction and Building Materials 196 (2019) 284–294

Contents lists available at ScienceDirect

Construction and Building Materials

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

Evaluation of the healing potential of asphalt mixtures based on a

modified semi-circular bending test
Jiwang Jiang a, Fujian Ni a,⇑,1, Fan Wu a, Husam Sadek b, Quan Lv c
a
School of Transportation, Southeast University, Nanjing, Jiangsu Province 210096, China
b
Department of Construction Management, College of Engineering, Louisiana State University, 3315 A Patrick F. Taylor Hall, Baton Rouge, LA 70803, USA
c
Key Laboratory of Road and Traffic Engineering of Ministry of Education, Tongji University, 304 Tongda Building 4800 Cao’an Road, Shanghai 201804, China

h i g h l i g h t s

An optimal rest period of 0.8s were determined.

Newly healing indexes were proposed.

Both binder properties and gradation affect the healing potential of asphalt mix.

The reliability of the newly proposed healing-based cyclic SCB test was validated.

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

Article history: Since the semi-circular bending (SCB) test has been widely used to evaluate the fatigue properties of
Received 22 March 2018 asphalt mixtures, it has been modified in this paper by adding a certain length of rest period in each cyclic
Received in revised form 11 October 2018 loading sequence to evaluate the healing potential of different asphalt mixtures. The digital image corre-
Accepted 29 October 2018
lation (DIC) method was utilized to characterize the tensile strain at the bottom of specimens during the
healing-based cyclic SCB test. The influence of notch depth of the specimen (0 mm, 15 mm and 25 mm)
and the length of the rest period (0 s, 0.4 s, 0.8 s and 1.2 s) on the fatigue and healing behaviour were
Keywords:
carefully evaluated. The fatigue life (Nf), total dissipated energy (Gsum), secant modulus (Ms), average
SCB test
Healing property
deformation growth rate (Dr) and average tensile strain (er) during the cyclic load with or without rest
Fatigue performance period were analysed. Test results show that the inserted notch reduces the fatigue life and deformation
Rest time of specimens, while can hardly change the load-sensitivity and secant modulus of the specimen. Most
Notch depth performance recovery happens within the rest period of 0.8 s. So it is better to conduct the healing-
based cyclic test with non-notched SCB specimen and 0.8 s rest period. Additionally, it is found that
asphalt mixtures with modified binder present both better fatigue and healing properties than asphalt
mixtures with neat asphalt, while the stone matrix asphalt mixture presents similar fatigue property
but much better healing potential than one dense graded mixture with the same binder.
Ó 2018 Published by Elsevier Ltd.

1. Introduction been considered in the design or practice [4]. The self-healing phe-
nomenon could be defined as any process by the material itself
Asphalt mixture is a self-healing material since the asphalt cause the improvement of a performance after an earlier damage
could heal itself [1]. Damage caused by the traffic loading or tem- on the performance of the material [5], which could be divided into
perature variations exists in the asphalt could be restored to a cer- two parts, one is autogenic self-healing process [6], another is
tain extent during the rest period, when the strength and stiffness autonomic self-healing process [7].
properties of the asphalt materials could be regained [2,3]. How- Many different tests have been used to characterize healing
ever, the self-healing behaviour of the asphalt mixture has not property of asphalt mixtures in varying scales. In general, they
can be divided into two categories with different ways to introduce
the rest period. One method lets the sample rest for a certain per-
⇑ Corresponding author. iod of time after being damaged to a certain level, which could be
E-mail address: [email protected] (F. Ni). referred to a storage healing test [8,9]. Another method is to set a
1
School of Transportation, Southeast University, Nanjing, Jiangsu Province 210096, short rest time in each load increment in the cyclic fatigue test,
China.

https://doi.org/10.1016/j.conbuildmat.2018.10.220
0950-0618/Ó 2018 Published by Elsevier Ltd.
 J. Jiang et al. / Construction and Building Materials 196 (2019) 284–294 285

which could be referred to an intermittent healing test [10]. As for Table 1

the intermittent healing test, the loading pattern matches the load- Aggregate gradation and mix design results.

ing in practice more closely [11]. In reality there is always a rest Sieve size (mm) Passing percent (%)
period between the loading of one axle to the next and of course SMA-13 AC-13
rest in between each truck passage. The intermittent healing test
16 100 100
is mainly cooperated with the fatigue test. Many fatigue tests, 13.2 95.5 95
including the four-point bend test [12], direct tension test [13], 9.5 62.5 76.5
push-pull test [14] and so on, have been added with different rest 4.75 27 53
times in each loading increment to evaluate the healing potential 2.36 20.5 37
1.18 19 26.5
of different asphalt mixtures. It could be found that the fatigue life
0.6 16 19
is extended significantly by adding the rest time at the same test 0.3 13 13.5
condition [15]. 0.15 12 10
However, the mechanism behind healing of asphalt mixtures is 0.075 10 6
Aggregate granite granite
still not fully understood up until now. As a result, a widely
Binder PG76-22 SBS PG76-22 SBS
accepted healing test, that can predict the healing behaviour of PG70-22 SBS
an asphalt mixture, is not available yet. The semi-circular bending 70# neat asphalt
(SCB) test has been widely used to evaluate the fatigue cracking Optimal binder content (%) 6.2 4.9
property of asphalt mixtures recently, since the simple test proce- Maximum density (g/cm3) 2.589 2.623

dure, more available test samples and reasonable stress pattern

[16,17]. Therefore, the SCB test was modified in this paper by add-
ing the rest period in each loading increment to evaluate the heal- (±0.5%). The cylindrical samples with diameter of 150 mm and
ing property of asphalt mixtures. The fatigue life, dissipated height of 180 mm were fabricated firstly. Then, both end of speci-
energy, stiffness and tensile strain from the SCB test are fully inves- mens was cut and three rings with a thickness of 50 mm were cut
tigated to better explain the mechanism behind healing of asphalt from the original sample by a masonry saw. After that, each ring
mixtures. was cut to two halves with or without a notch at the middle of
the bottom of the specimen based on different test conditions.
The air voids of each halve should be in the range of 3.5–4.5%
2. Objective and unqualified specimens were abandoned before testing.

The main objectives of the work presented in this paper were as

follows: 4. Methodology

Evaluate the influence of notch depth and rest time on the fati- 4.1. Healing-based cyclic SCB test
gue property and healing potential of asphalt mixtures based on
the cyclic SCB test. As Fig. 1(a) shows, the setup for SCB test consists of two sup-

Propose a new analysis method to fully investigate the fatigue porting rollers at the bottom edge and a loading roller at the mid-
life, dissipated energy, stiffness, deformation and tensile strain dle point of the top edge. The span over the two supporters was
of the SCB sample during the cyclic loading according to load– 120 mm. In this study, stress-controlled mode was chosen for the
displacement curve and strain field obtained from Digital Image SCB fatigue test. So before the fatigue test, a SCB strength test
Correlation (DIC) methods. should be conducted firstly to obtain the peak load of SCB speci-

Analyse the healing properties of four asphalt mixtures with mens. The SCB strength test was conducted at 15 °C with a loading
two gradations and three binder types based on the healing- rate of 50 mm/min, and the peak load for each mixture is exhibited
based cyclic SCB test. in Table 2. Four replicates were prepared for the SCB strength test.
Then, three to five stress amplitudes were determined by the peak
load to be applied to the SCB fatigue test. The frequency of the
3. Materials and mix design
repeated compressive load is 10 Hz, which is approximately equiv-
alent to a vehicle speed of 50mph. In order to relief the healing pro-
A dense graded asphalt mixture (AC-13) and a stone matrix
cess, different lengths of rest period were added in the loading
asphalt mixture (SMA-13) were selected for this study. The nomi-
sequence. Fig. 1(b)–(e) show loading pattern with no rest time,
nal maximum aggregate size for both mixtures were 13 mm. The
0.4 s, 0.8 s and 1.2 rest period in each loading sequence,
coarse aggregates used in both mixtures were granite and the con-
respectively.
tent of the added cellulose fiber for SMA-13 mixtures was 0.3% by
A UTM-25 universal testing machine equipped with a tempera-
the weight of mix. Three asphalt binders with different PG grades
ture control chamber was used for the healing based cyclic SCB test
were utilized for the SMA-13 mixture, including a styrene–buta
at 15 °C. Prior to do the test, all specimens were placed in a ther-
diene–styrene (SBS) polymer-modified asphalt binder meeting
mostatic chamber for more than 4 h to reach the temperature of
the Superpave specification for PG 76-22, a SBS polymer-
15 °C, and four replicated specimens were prepared for each mix-
modified asphalt binder meeting the Superpave specification for
ture under each test condition. During the test, the vertical defor-
PG70-22 and a neat asphalt (70#) meeting the Superpave specifi-
mation of specimen was measured by Linear Variable Differential
cation for PG 58-28. As for the AC-13 mixture, only PG76-22 SBS
Transducers (LVDTs) which were installed in vertical direction.
modified binder was applied in the mix design. The base mixture
were first designed by the standard Marshall procedure and opti-
mal asphalt content for SMA-13 mixtures was determined to be 4.2. Analysis of the deformation curve
6.2% by the weight of mix while 4.9% for AC-13 mixtures. The
aggregate gradation and detailed volumetric parameters of both Fig. 2 shows one load-displacement curve of SMA-13 mixtures
mixtures are shown in Table 1. with PG76-22 binder under load level of 5kN with rest time of
All specimen tested in this work were produced by the super- 1.2 s. It could be found from Fig. 2(a) that the load-displacement
pave gyratory compactor (SGC) with a target air void level of 4% curve (blue line) of the healing-based cyclic SCB test could be
286 J. Jiang et al. / Construction and Building Materials 196 (2019) 284–294

5
(c)

Vertical load (kN)

4

0
0 0.5 1
Time /s
6

Vertical load /kN

4 (d)
3

1
(a) 0
0 0.9 1.8
Time /s
6 6
5
Vertical load (kN)

Vertical load /kN

(b) (e)
4 4

3 3

2 2

1 1

0 0
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 0 1.3 2.6
Time /s Time/s

Fig. 1. Set up of the healing-based cyclic SCB test: (a) loading equipment; (b) loading mode without rest time; (c) loading mode with 0.4 s rest time; (d) loading mode with
0.8 s rest time; (e) loading mode with 1.2 s rest time.

Table 2
Peak load of the SCB strength test of different mixtures.

*
SMA-PG76-0 mm means the SCB sample with notch depth of 0 mm, mix design of SMA-13 and PG76-22 SBS binder.

divided into three stages, and there are two inflection points dur- tion per cycle, it could be found there is a plateau between the first
ing all the loading sequences till the failure of the specimen. Before inflection and second inflection point, and the average value of the
the first inflection point, the vertical deformation increases sharply relative deformation growth per cycle between the first and second
since the densification of the specimen. The vertical deformation inflection point is termed as the average deformation growth rate
increases steady and slowly between the first and second inflection (Dr).
point, while after the second inflection point, the vertical deforma- As for each loading cycle, the loading frequency is 10 Hz while
tion grows faster and faster until the fatigue failure. So the number the deformation data acquisition frequency is 110 Hz, so it is pos-
of cycles at the second inflection point is termed as the fatigue life sible to plot the curve of vertical deformation versus cyclic load in
(Nf). The red line in Fig. 2(a) presents the relative vertical deforma- one cycle. As Fig. 2(a) shows, the green line shows the load-
 J. Jiang et al. / Construction and Building Materials 196 (2019) 284–294 287

6
Vertical deformation Vertical deformation per cycle
0.3 could be found that the secant modulus increases slightly when
the loading cycles increases and a plateau appears in the dissipate

Vertical deformation per cycle (mm)

Fatigue life (Nf)
energy curve versus loading cycles between the first and second
Vertical deformation (mm)

inflection point. So the average secant modulus (Ms) is defined as

4 0.2 the average value between first and second inflection point in
Cycle 2000
6
the curve of secant modulus versus loading cycles. The total dissi-
pated energy is defined as the accumulative dissipated energy
Cyclic load (KN)
4
Secant modulus
per cycle before the second inflection point.
2 2
Dissipate energy
0.1
per cycle
0
0 0.1 0.2 0.3 0.4 0.5
4.3. Incorporation of DIC methods
Vertical deformation (mm)

0 0
During the loading of the healing-based cyclic SCB test, the dig-
0 900 1800 2700 3600 4500 5400 ital image correlation (DIC) system is utilized simultaneous to
Number of cycles measure the full-field strain of the front surface of specimens.
(a) Definition of fatigue life (Nf) and average deformation growth rate (Dr) Before the implementation of DIC technique, a series of surface
treatment on specimens should be employed. As exhibited in
Secant modulus per cycle dissipated energy per cycle
16 2 Fig. 3(b), the surface is painted uniformly matt white using a spray
matt paint and random black dots are speckled onto the dried
Secant modulus per cycle (106 N/m)

Dissipated energy per cycle (J)

white surface of specimens using a spray matt black paint. This
12 1.5 random speckle pattern on the surface could be used to capture
the full-field displacement. As Fig. 3(a) shows, a high-speed indus-
Average secant modulus (Ms)
trial digital camera is placed and regulated carefully to obtain the
8 1 images during the test with a frequency of 10 frames per second.
Two white lights amounted on both sides of the specimen are
Total dissipated energy (Gsum) applied to improve the illumination effects. An image process soft-
4 0.5
ware is used for analysis of the displacements/strain field of the
specimen. The Newton-Raphson method [18] is used as a sub-
pixel registration algorithm in the displacement measurement. A
0 0
0 900 1800 2700 3600 4500 5400 numerical differentiation process is used to obtain the strain con-
Number of cycles sidering the noise contained in the computed displacement.
(b) Definition of average secant modulus (Ms) and total dissipated energy (Gsum) The main cracking zone for the SCB fatigue test is determined as
±2.5 cm in the middle of the specimens as previous study shows
Fig. 2. Schematic map for the definition of fatigue parameters from the load- [19], since most tensile strain happens in the 5 cm range. Accord-
displacement curve from the cyclic loading test.
ing to the DIC methods, the tensile strain from the gauge length
of 5 cm in the middle of specimen could be obtained as Fig. 4
shows. The blue line is the accumulative tensile strain and the
deformation curve at cycle 2000. The area of the curve is deter- red line is the tensile strain growth per cycle. The average tensile
mined to be the dissipated energy per cycle and the secant modu- strain growth rate (er) is defined as the average value of the tensile
lus at this cycle was calculated by Eq. (1). strain growth at each cycle between the first and second inflection
point.
F maxi  F mini
M si ¼ ð1Þ
Dmaxi  Dmini
4.4. Definition of healing index
where: Msi = the secant modulus of cycle i; Fmaxi = the highest load
in cycle i; Fmini = the lowest load in cycle i; Dmaxi = the highest The self-healing phenomenon could be defined as any process
deformation in cycle i; Dmini = the lowest deformation in cycle i. by the material itself cause the improvement of a performance
As Fig. 2 (b) shows, the blue line presents the secant modulus after an earlier damage on the performance of the material. There-
changes with the increasing loading cycles and the red line exhibits fore, as Eqs. (2)–(6) shows, according to the fatigue parameters
the dissipated energy evolves with the increase of loading cycles. It defined as above, the healing indexes are defined as the increase

Tensile strain

Gauge length of 5cm

(a) (b)
Fig. 3. Utilization of the DIC methods: (a) set up of the DIC system; (b) strain contours of the surface of the SCB specimen.
288 J. Jiang et al. / Construction and Building Materials 196 (2019) 284–294

0.08
Tensile strain Tensile strain growth rate
4.E-04 where: HNf, HGsum, HMs, HDr, Her are the healing indexes of fatigue
life, total dissipated energy, stiffness, deformation and tensile
strain, respectively; N’f and Nf are the fatigue life of asphalt mixtures
Accumulated tensile strain

0.06 under load sequence with and without rest periods, respectively;

Tensile strain per cycle

G’sum and Gsum are the total dissipated energy of asphalt mixtures
under load sequence with and without rest periods, respectively;
0.04 2.E-04 M’s and Ms are the average secant modulus of asphalt mixtures
under load sequence with and without rest periods, respectively;
D’r and Dr are the average deformation growth rate of asphalt mix-
Average tensile strain growth rate ( r)
0.02 tures under load sequence with and without rest periods, respec-
tively; e’r and er are the average tensile strain growth rate of
asphalt mixtures under load sequence with and without rest peri-
0 0.E+00 ods, respectively.
0 900 1800 2700 3600 4500 5400
Number of cycles
5. Discuss of results
Fig. 4. Definition of the average tensile strain growth rate (er) from the DIC method.
5.1. Influence of notch depth
or decrease scale (or percentage recovery) of the fatigue
parameters. For most researches, a notch is mechanically inserted into the
N0f  Nf SCB halves and this notch creates a stress concentration that forms
HNf ¼ ð2Þ at the notch tip [20]. Unfortunately, the notched SCB sample could
Nf
hardly describe the full damage process and structure degradation
of asphalt materials since the primary artificial distress applied on
G0sum  Gsum
HGsum ¼ ð3Þ the sample, which may reduce the potential of healing possibility.
Gsum
To evaluate the influence of notch depth on the fatigue properties
of the asphalt mixture, the SCB samples of SMA-13 mixture using
M0s  Ms
HMs ¼ ð4Þ PG76-22 SBS binder are prepared by inserting different notch
Ms depths of 0 mm, 15 mm and 25 mm. As Table 3 shows, five stress
levels are selected to conduct the cyclic SCB fatigue test with no
D0r  Dr
HDr ¼ ð5Þ rest time. Four replicates are prepared for each test condition.
Dr Table 4 shows the test results of the fatigue test of SMA mix-
tures with varying notch depths under different stress levels. It
e0r  er
Her ¼ ð6Þ could be found that the coefficient variable of different fatigue
er parameters of sample with no notch presents lower than samples

Table 3
Stress levels of the SCB cyclic fatigue test with different notch depths.

Sample ID Peak load (kN) Load level and stress level

SMA-PG76-0 mm-0 s 15.05 Load (kN) 2.00 3.00 4.00 5.00 6.00
Stress level 0.13 0.20 0.27 0.33 0.40
SMA-PG76-15 mm-0 s 9.30 Load (kN) 1.00 1.50 2.00 2.50 3.00
Stress level 0.11 0.16 0.22 0.27 0.32
SMA-PG76-25 mm-0 s 7.50 Load (kN) 0.95 1.20 1.45 1.70 1.95
Stress level 0.13 0.16 0.19 0.23 0.26

Table 4
Fatigue parameters of the SMA mixture with different notch depths under varying stress levels.

Sample ID Stress level Nf Gsum (J) Ms (106N/m) Dr (mm) er

Mean COV1 Mean COV Mean COV Mean COV Mean COV
SMA-PG76-0 mm-0 s 0.13 175,250 10.4% 5282.6 12.7% 9.1 2.7% 3.1E-05 18.0% 2.2E-07 11.2%
0.2 22,935 17.5% 1255.4 25.3% 11.5 8.0% 1.8E-04 21.6% 2.6E-06 19.1%
0.27 6750 8.3% 727.5 8.7% 12.0 3.3% 4.2E-04 6.6% 3.9E-06 5.6%
0.33 2550 10.7% 512.1 9.1% 12.1 9.7% 1.9E-03 7.9% 1.2E-05 10.2%
0.4 1082 18.8% 264.8 16.8% 13.4 6.4% 3.1E-03 18.9% 2.3E-05 10.3%
SMA-PG76-15 mm-0 s 0.11 246,000 18.6% 1969.9 15.4% 8.5 6.3% 2.4E-05 28.8% 4.8E-07 25.6%
0.16 30,507 30.3% 563.1 30.6% 9.1 10.8% 9.9E-05 30.9% 1.6E-06 27.2%
0.22 11,000 25.7% 301.8 24.0% 10.2 13.4% 4.3E-04 24.1% 3.2E-06 18.7%
0.27 4109 24.7% 194.6 22.0% 10.5 17.0% 9.5E-04 19.9% 1.4E-05 18.3%
0.32 1679 25.5% 123.3 20.3% 10.7 3.1% 1.7E-03 32.7% 2.8E-05 26.2%
SMA-PG76-25 mm-0 s 0.13 37,785 31.0% 348.2 28.3% 8.7 3.7% 6.2E-05 29.4% 2.2E-06 21.9%
0.16 17,500 27.6% 267.7 25.1% 8.7 9.9% 1.7E-04 14.0% 3.4E-06 16.5%
0.19 12,455 17.6% 248.5 16.4% 10.2 13.0% 2.1E-04 35.8% 3.0E-06 24.1%
0.23 3071 24.4% 87.9 20.5% 10.8 9.0% 6.9E-04 18.6% 1.0E-05 19.2%
0.26 1477 31.2% 71.7 24.3% 10.9 6.8% 1.4E-03 20.0% 4.0E-05 27.5%

1. COV means coefficient of variation.

 J. Jiang et al. / Construction and Building Materials 196 (2019) 284–294 289

with 15 mm or 25 mm notch. Since it is hard to cut notches with at higher stress level. Similarly, as figure (c) and (d) show, the fati-
identical angle, length and width for each specimen, which may gue equations obtained by Dr and er of specimens with different
cause significant systematic error on the test results. Additionally, notch depths are parallel to each other. All these indicate that
the material property (aggregate or asphalt mastic) at the top point the pre-defined notch reduces Nf, Gsum, Dr and er of the specimen
of the notch could affect the results. during the fatigue loading at the same stress level, but the sensitiv-
As Fig. 4 exhibits, fatigue life (Nf), total dissipated energy (Gsum), ity of the specimens to stress levels is not changed with or without
average deformation growth rate (Dr) and tensile strain (er) could the notch. Fig. 5(e) demonstrates the relationship between Ms and
fit well a linear function with stress levels at the log-log coordinate stress level of the specimens with different notch depths. The aver-
with R2 more than 0.9. The fatigue equations deduced by different age secant modulus (Ms) fits a linear relation with the stress levels
fatigue parameter are shown as Eqs. (7)–(10). Fig. 5 (a) and (b) and data from specimens with different notch depths fit with one
show that specimens with no notch present longest fatigue life identical equation, which indicates that the inserted notch has
(Nf) and largest total dissipated energy (Gsum), followed by speci- negligible influence on the secant modulus during the cyclic load-
mens with 15 mm notch and 25 mm notch. The slope of the fatigue ing sequence.
equations obtained from specimens with different notch depths
logðNf Þ ¼ a1  logðrÞ þ b1 ð7Þ
are similar, which indicates that Nf or Gsum of specimens with
notch at lower stress level may equal to specimen without notch logðGsum Þ ¼ a2  logðrÞ þ b2 ð8Þ

SMA-PG76-0mm-0s
SMA-PG76-0mm-0s SMA-PG76-15mm-0s
SMA-PG76-15mm-0s SMA-PG76-0mm-0s SMA-PG76-15mm-0s
SMA-PG76-15mm-0s
SMA-PG76-25mm-0s
SMA-PG76-25mm-0s SMA-PG76-25mm-0s 4.0
5.5 log (Gsum) = -2.5339log ( ) + 1.4261
log (Nf) = -4.5031og ( ) + 1.2431
R² = 0.9994 R² = 0.9828
3.5
log (Gsum) = -2.5088log ( ) + 0.8371
Log (Nf)

log (Nf) = -4.66031og ( ) + 0.5344 R² = 0.988

Log(Gsum)
4.5
R² = 0.9574 3.0

2.5
3.5

log (Nf) = -4.50651og ( ) + 1.0186 2.0

R² = 0.992 log (Gsum) = -2.4244log ( ) + 0.4744
R² = 0.8853
2.5 1.5
-1.2 -0.9 -0.6 -0.3 0.0 -1.2 -0.9 -0.6 -0.3 0.0
Log( ) log ( )
(a) log (Nf) versus Log ( ) (b) log (Gsum) versus Log ( )
Log ( ) Log ( )
-1.2 -0.9 -0.6 -0.3 0.0 -1.2 -0.9 -0.6 -0.3 0.0
-2 -4.0
log (Dr) = 4.287log ( ) - 0.4273
R² = 0.9661 log ( r) = 3.8462log ( ) - 2.4166
R² = 0.7995
log (Dr) = 4.0482log ( ) - 0.7418
Log (Dr)

R² = 0.9982 -3 log ( r) = 3.7842log ( ) - 2.7744 -5.0

Log ( r)
R² = 0.9618

-4 -6.0
log (Dr) = 4.1692log ( ) - 0.8425
R² = 0.9825 log ( r) = 4.0147log ( ) - 3.0068
R² = 0.9676
SMA-PG76-0mm-0s
SMA-PG76-0mm-0s SMA-PG76-15mm-0s
SMA-PG76-15mm-0s -5 SMA-PG76-0mm-0s SMA-PG76-15mm-0s
SMA-PG76-15mm-0s -7.0
SMA-PG76-25mm-0s
SMA-PG76-25mm-0s SMA-PG76-25mm-0s
SMA-PG76-25mm-0s

(c) log (Dr) versus Log ( ) (d) log ( r) versus Log ( )

SMA-PG76-0mm-0s SMA-PG76-15mm-0s
16 SMA-PG76-25mm-0s

12

Ms = 14.775 + 7.0507
Ms

8
R² = 0.7819

0
0.1 0.15 0.2 0.25 0.3 0.35 0.4 0.45
Stress level ( )
(d) Ms versus stress level
Fig. 5. Relationship between different fatigue parameters versus stress levels from specimens with different notched depths.
290 J. Jiang et al. / Construction and Building Materials 196 (2019) 284–294

logðDr Þ ¼ a3  logðrÞ þ b3 ð9Þ 1.2 s are less than 10%, so the addition of this 0.4 s has little effi-
ciency on the performance recovery. Considering that the fatigue
logðer Þ ¼ a4  logðrÞ þ b4 ð10Þ test is time-consuming, rest time of 0.8 s is decided to be the opti-
mal rest time for the evaluation of healing potential of asphalt mix-
where: r = stress level; a1, a2, a3, a4, b1, b2, b3, b4 = fitting ture based on the cyclic SCB fatigue test.
coefficients.
5.3. Comparison between the cyclic SCB test and healing-based SCB
5.2. Influence of rest times test

To evaluate the influence of rest time on the fatigue and healing 5.3.1. Fatigue properties of different mixtures with or without rest
properties of the asphalt mixtures. The non-notched SCB samples period
of SMA-13 mixtures using PG76-22 SBS binder were prepared. Four Two material factors were considered in this paper. Different
rest times were selected to be applied in each loading increment, asphalt mixtures with three asphalt binders and two gradations
including 0, 0.4, 0.8 and 1.2 s. Three loading levels, including 4, 5, were selected to conduct the healing-based cyclic SCB test with
and 6kN, corresponding to stress ratios of 0.27, 0.33 and 0.4, rest period of 0.8 s and cyclic SCB test with no rest period. Fig. 8
respectively, were selected for each rest period to build the fatigue shows the fatigue equations of these mixtures fitted by different
equations. Four replicates were prepared for each test condition. fatigue parameters. As shown in Fig. 8(a), the fatigue life (Nf)
increases significantly from test without rest time to test with rest
5.2.1. Analysis of the fatigue parameters periods of 0.8 s except the SMA-13 mixture with 70# neat binder.
As Table 5 exhibits, the coefficient available (COV) of the test The Nf of SMA-13 mixture and AC-13 mixture with PG76-22 SBS
results with different rest times is less than 25%. Fig. 6 shows that binder are similar when followed by SMA-13 mixture with PG
fatigue life (Nf), total dissipated energy (Gsum), average deforma- 70-22 SBS binder and 70# neat binder. Analogously, except the
tion growth rate (Dr) and tensile strain (er) could fit well a linear SMA-13 mixture with the neat asphalt, the total dissipated energy
function with stress level at the log-log coordinate with R2 more presents a clear growth by adding the rest period. The Gsum of both
than 0.9 for most cases. Fig. 6(a) and (b) illustrate that Nf and Gsum SMA-13 mixture with PG76-22 and PG70-22 SBS binder and AC-13
increase significantly when the rest period increases from 0 to mixture with PG76-22 binder are similar and all much larger than
1.2 s. Fig. 5(c) and (d) shows that Dr and er decrease dramatically the SMA-13 mixture with 70# neat asphalt. Fig. 8(c) and (d) illus-
with the increasing rest period. As can be seen in Fig. 6(e), the trate that the average deformation growth rate (Dr) and average
secant modulus (Ms) changes slightly when the rest time is applied tensile strain growth rate (er) are significantly reduced by the addi-
in the loading pattern. tion of rest periods for all four mixtures and the SMA-13 mixture
with 70# neat asphalt presents the largest growth rate of both
5.2.2. Analysis of the healing index deformation and tensile strain. As can be seen in Fig. 8(e), both
Based on the analysis of the fatigue equations, it is found that SMA-13 mixture and AC-13 mixture with PG76-22 SBS binder have
addition of rest time could increase the fatigue life and total dissi- similar secant modulus (Ms) and larger than SMA-13 mixture with
pated energy, reduce the average deformation growth rate and PG70-22 SBS binder and 70# neat asphalt. The addition of rest time
average tensile strain rate, but have little effect on the secant mod- increases the Ms slightly for each mixture.
ulus. Therefore, the healing property from the addition of rest time
could be quantified by the difference of fatigue parameters, includ- 5.3.2. Analysis of the healing potential of different mixtures
ing Nf, Gsum, Dr, and er. Since the healing indexes calculated from Table 7 exhibits the healing indexes of four asphalt mixtures
data at different stress levels are similar, Table 6 presents the aver- calculated by fatigue life (Nf), total dissipated energy (Gsum), aver-
age value of the healing indexes of the SMA mixtures at different age deformation growth rate (Dr) and average tensile strain rate
stress levels. It could be observed that HNf and HGsum are higher (er). the healing index of HDr shows negligible difference between
when the rest time increases while HDr and Her are lower with different mixtures, which may be caused by the localization defor-
the increasing rest time. Fig. 7(a) shows that the amplitude of vari- mation near the top roller. So the healing index calculated by aver-
ation of the healing indexes from 0 to 0.8 s are much larger than age deformation growth rate (HDr) is not reliable. As Fig. 9 shows,
from 0.8 to 1.2 s. Fig. 7(b) illustrates the relative percentage of binder property has significant influence on the healing property of
recovery in each rest time interval. It could be found that the asphalt mixture. The limited data shows that HNf of the SMA-13
amplitude of variation of all four healing indexes from 0.8 s to mixture with PG76-22 SBS modified binder is more than 10 times

Table 5
Fatigue parameters of the SMA mixture with different rest times under varying stress levels.

Sample ID Stress level Nf Gsum(J) Ms (106N/m) Dr (mm) er

Mean COV Mean COV Mean COV Mean COV Mean COV
SMA-PG76-0 s 0.27 6750 8.3% 727.5 8.7% 12.0 3.3% 4.2E-04 6.6% 3.9E-06 5.6%
0.33 2550 10.7% 512.1 9.1% 12.1 9.7% 1.9E-03 7.9% 1.2E-05 10.2%
0.4 1082 18.8% 264.8 16.8% 13.4 6.4% 3.1E-03 18.9% 2.3E-05 10.3%
SMA-PG76-0.4 s 0.27 8975 19.8% 2037.3 4.7% 11.9 9.8% 4.2E-04 10.0% 2.5E-06 11.2%
0.33 3753 1.0% 902.2 0.1% 12.2 2.4% 1.1E-03 15.5% 5.6E-06 12.3%
0.4 1840 23.3% 744.9 15.2% 13.5 5.9% 1.8E-03 19.7% 8.0E-06 20.1%
SMA-PG76-0.8 s 0.27 12,520 19.7% 2162.3 21.3% 12.8 9.0% 3.7E-04 1.1% 2.3E-06 2.4%
0.33 5568 18.5% 1362.3 16.8% 12.8 12.0% 8.1E-04 16.4% 3.6E-06 13.6%
0.4 2043 21.1% 741.0 25.4% 13.5 12.9% 1.7E-03 13.1% 5.9E-06 10.4%
SMA-PG76-1.2 s 0.27 13,840 15.7% 1949.0 24.1% 12.6 10.9% 2.3E-04 1.7% 1.7E-06 4.2%
0.33 5591 21.9% 1367.7 16.9% 13.1 6.3% 8.7E-04 23.7% 3.5E-06 17.3%
0.4 2293 17.7% 861.6 17.2% 13.9 2.5% 1.7E-03 10.5% 4.9E-06 9.7%
 J. Jiang et al. / Construction and Building Materials 196 (2019) 284–294 291

SMA-PG76-0mm-0s SMA-PG76-0mm-0.4s
SMA-PG76-0mm-0.4s SMA-PG76-0mm-0s SMA-PG76-0mm-0.4s
SMA-PG76-0mm-0.4s
4.0
SMA-PG76-0mm-0.8s SMA-PG76-0mm-1.2s
SMA-PG76-0mm-1.2s SMA-PG76-0mm-0.8s SMA-PG76-0mm-1.2s
SMA-PG76-0mm-1.2s
4.5
log (Nf) = -4.5741og ( ) + 1.5419 log (Gsum) = -2.0747log ( ) + 2.1189
R² = 0.9999 R² = 0.9922 3.5
log (Nf) = -4.6081og ( ) + 1.4937 4.0 log (Gsum) = -2.7216log ( ) + 1.7993

Log(Gsum)
R² = 0.9946 R² = 0.9917
log (Nf) = -4.0341og ( ) + 1.6502

Log (Nf)
R² = 0.9979 3.0
3.5
log (Gsum) = -2.5707log ( ) + 1.8046
R² = 0.8943
3.0 2.5
log (Gsum) = -2.5339log ( ) + 1.4261
log (Nf) = -4.5031og ( ) + 1.2431
R² = 0.9828
R² = 0.9994
2.5 2.0
-0.6 -0.5 -0.4 -0.3 -0.6 -0.5 -0.4 -0.3
Log ( ) log ( )
(a) log (Nf) versus Log ( ) (b) log (Gsum) versus Log ( )
Log ( ) Log ( )
-0.6 -0.3 -0.6 -0.5 -0.4 -0.3
-2 -4.0
log (Dr) = 4.1692log ( ) - 0.8425 log ( r) = 4.0147log ( ) - 3.0068
R² = 0.9825 R² = 0.9676

-3 -5.0

Log ( r)
log (Dr) = 3.6892log ( ) - 1.2494 Log (Dr)
R² = 0.9799 log ( r) = 2.9969log ( ) - 3.8713
log (Dr) = 3.9663log ( ) - 1.1799 R² = 0.9543
R² = 1 -4 log ( r) = 2.4459log ( ) - 4.2584 -6.0
log (Dr) = 5.1348log ( ) - 0.6727 R² = 0.9996
log ( r) = 2.726log ( ) - 4.1964
R² = 0.9713
R² = 0.9605
SMA-PG76-0mm-0s
SMA-PG76-0mm-0s SMA-PG76-0mm-0.4s
SMA-PG76-0mm-0.4s -5 SMA-PG76-0mm-0s
SMA-PG76-0mm-0s SMA-PG76-0mm-0.4s
SMA-PG76-0mm-0.4s -7.0
SMA-PG76-0mm-0.8s
SMA-PG76-0mm-0.8s SMA-PG76-0mm-1.2s
SMA-PG76-0mm-1.2s SMA-PG76-0mm-0.8s
SMA-PG76-0mm-0.8s SMA-PG76-0mm-1.2s
SMA-PG76-0mm-1.2s
(c) log (Dr) versus Log ( ) (d) log ( r) versus Log ( )
SMA-PG76-0mm-0s SMA-PG76-0mm-0.4s
SMA-PG76-0mm-0.8s SMA-PG76-0mm-1.2s
16

12

Ms = 13.327log ( ) + 8.2832
Ms

8
R² = 0.77

0
0.1 0.15 0.2 0.25 0.3 0.35 0.4 0.45
Stress level
(d) Ms versus stress level
Fig. 6. Relationship between different fatigue parameters versus stress levels with different rest times.

Table 6 (62.2% versus 35.7%). Furthermore, the gradation type also

Healing indexes of the SMA mixture with different rest times compared with no rest has significant influence on the healing index. Compared with
time. AC-13 mixture with PG76-22 modified binder, the HNf of SMA-13
Sample ID HNf HGsum HDr Her mixture with the same binder increases from 52.4% to 97.6%, HGsum
increases from 128.6% to 181% and Her decrease from 48% to
SMA-PG76-0.4 s Mean 50.1% 145.8% 28.5% 51.8%
COV 27.5% 21.4% 28.9% 18.7% 62.2%. Even-through the fatigue properties of SMA-13 mixture
SMA-PG76-0.8 s Mean 97.6% 181.0% 48.0% 62.2%
and AC-13 mixture with PG76-22 SBS modified binder are similar,
COV 18.5% 8.6% 20.8% 28.8% the healing potential of SMA-13 mixture is much better than AC-13
SMA-PG76-1.2 s Mean 112.1% 186.8% 49.8% 64.9%
mixture. Therefore, not only asphalt property but also the grada-
COV 6.3% 17.9% 12.7% 16.6% tion of asphalt mixture could influence the healing potential of
asphalt mixture significantly.

than the same mixture with 70# neat asphalt. HGsum increases 5.3.3. Comparison of the sensitivity of different healing indexes
from 23.9% to 181.0% when the binder of SMA-13 mixture is chan- Four parameters were used to evaluate the healing properties of
ged from the neat asphalt to the PG76-22 modified binder. Addi- the asphalt mixtures considering two factors, binder types and gra-
tionally, the recovery percentage of the Her of the PG76-22 dations. To verify the confidence of the test results, the statistical
modified binder is nearly two time of the Her of neat binder significance of the changes in the four healing indexes, depending
292 J. Jiang et al. / Construction and Building Materials 196 (2019) 284–294

HNf
HNf HGsum
HGsum HDr
HDr Hr r
H 150% HNf HGsum HDr Hr
200%

Percentage of revovery
150%
Pecentage of recovery

80%
100%

50%
10%
0%
0 0.4 0.8 1.2
-50%
-60%
-100% 0-0.4s 0.4-0.8s 0.8-1.2s
Rest time (s)
(a) (b)
Fig. 7. Healing indexes under varying rest times: (a) total percentage of recovery compared with rest time of 0; (b) Relative percentage of recovery in each interval.

SMA-PG76-0mm-0s
SMA-PG76-0mm-0s SMA-PG76-0mm-0.8s
SMA-PG76-0mm-0.8s SMA-PG76-0mm-0s
SMA-PG76-0mm-0s SMA-PG76-0mm-0.8s
SMA-PG76-0mm-0.8s
AC-PG76-0mm-0s
AC-PG76-0mm-0s AC-PG76-0mm-0.8s
AC-PG76-0mm-0.8s AC-PG76-0mm-0s
AC-PG76-0mm-0s AC-PG76-0mm-0.8s
AC-PG76-0mm-0.8s
SMA-70#-0mm-0s SMA-70#-0mm-0.8s 4.5 SMA-70#-0mm-0s
SMA-70#-0mm-0s SMA-70#-0mm-0.8s
SMA-70#-0mm-0.8s 4.0
SMA-70#-0mm-0s SMA-70#-0mm-0.8s
SMA-PG70-0mm-0s
SMA-PG70-0mm-0s SMA-PG70-0mm-0.8s
SMA-PG70-0mm-0.8s SMA-PG70-0mm-0s
SMA-PG70-0mm-0s SMA-PG70-0mm-0.8s
SMA-PG70-0mm-0.8s

4.0 3.5

Log(Gsum)
Log (Nf)

3.5 3.0

3.0 2.5

2.5 2.0
-0.8 -0.7 -0.6 -0.5 -0.4 -0.3 -0.8 -0.7 -0.6 -0.5 -0.4 -0.3
Log ( ) log ( )
(a) (b)
Log ( ) Log ( )
-0.9 -0.6 -0.3 -0.8 -0.7 -0.6 -0.5 -0.4 -0.3
-2 -4

Log ( r)
Log (Dr)

-3 -5

SMA-PG76-0mm-0s
SMA-PG76-0mm-0s SMA-PG76-0mm-0.8s
SMA-PG76-0mm-0.8s SMA-PG76-0mm-0s SMA-PG76-0mm-0.8s
SMA-PG76-0mm-0s SMA-PG76-0mm-0.8s
AC-PG76-0mm-0s
AC-PG76-0mm-0s AC-PG76-0mm-0.8s
AC-PG76-0mm-0.8s AC-PG76-0mm-0s
AC-PG76-0mm-0s AC-PG76-0mm-0.8s
AC-PG76-0mm-0.8s
-4 -6
SMA-70#-0mm-0s
SMA-70#-0mm-0s SMA-70#-0mm-0.8s
SMA-70#-0mm-0.8s SMA-70#-0mm-0s
SMA-70#-0mm-0s SMA-70#-0mm-0.8s
SMA-70#-0mm-0.8s
SMA-PG70-0mm-0s
SMA-PG70-0mm-0s SMA-PG70-0mm-0.8s
SMA-PG70-0mm-0.8s SMA-PG70-0mm-0s
SMA-PG70-0mm-0s SMA-PG70-0mm-0.8s
SMA-PG70-0mm-0.8s
(c) (d)
SMA-PG76-0mm-0s SMA-PG76-0mm-0.8s
AC-PG76-0mm-0s AC-PG76-0mm-0.8s
16 SMA-70#-0mm-0s SMA-70#-0mm-0.8s
SMA-PG70-0mm-0s SMA-PG70-0mm-0.8s

12
Ms

4
0.15 0.25 0.35 0.45
Stress level
(e)
Fig. 8. Relationship between different fatigue parameters versus stress levels of different asphalt mixture with or without rest time.
 J. Jiang et al. / Construction and Building Materials 196 (2019) 284–294 293

Table 7 (1) The predefined notch would reduce fatigue life (Nf), total
Healing indexes of the different asphalt mixtures. dissipated energy (Gsum), average deformation growth rate
Sample ID HNf HGsum HDr Her (Dr) and average tensile strain rate (er) of the specimens dur-
SMA-70#-0.8 s Mean 8.5% 23.9% 45.2% 35.7% ing the fatigue loading at the same stress level, but the load-
COV 25.7% 18.2% 27.0% 26.2% sensitivity of the specimens and the secant modulus(Ms) are
SMA-PG70-0.8 s Mean 74.3% 130.1% 48.1% 44.6% not changed with the addition of a notch.
COV 6.8% 13.8% 26.4% 5.0% (2) Longer rest period in each loading sequence means higher Nf
SMA-PG76-0.8 s Mean 97.6% 181.0% 48.0% 62.2% and Gsum while lower Dr and er and similar Ms, so the healing
COV 18.5% 8.6% 20.8% 28.8% process has little relationship with the average secant mod-
AC-PG76-0.8 s Mean 52.4% 128.6% 50.1% 48.0% ulus (Ms).
COV 25.0% 17.4% 27.9% 3.8% (3) The amplitude of variation of the healing indexes from 0 to
0.8 s are much larger than from 0.8 to 1.2 s, indicating that
most performance recovery happens in the rest time of
0.8 s. Since the healing based fatigue test is very time-
250% SMA-70#-0.8s SMA-PG70-0.8s consuming, 0.8 s was set to be the optimal rest period for
SMA-PG76-0.8s AC-PG76-0.8s the healing evaluation test.
200% (4) Healing indexes from fatigue life, total dissipated energy and
Percentage of recovery

tensile strain (HNf, HGsum and Her) could examine the healing
150%
property of different asphalt mixtures more reliable than
100% healing indexes from the average deformation growth rate
(HDr).
50% (5) Asphalt mixtures with modified binder present both better
fatigue property and healing potentials. While, the stone
0%
matrix asphalt mixture (SMA-13) with modified binder pre-
-50% sents similar fatigue property but much better healing
potentials when compared with the dense graded mixture
HNf HGsum HDr Hr
-100% (AC-13) with the same modified binder.
Healing index
Fig. 9. The healing indexes of different asphalt mixtures.
Conflict of interest

None.
Table 8
Sensitivity analysis of four healing indexes using one-way ANOVA.

Source Factors Df Sum of Square F value Pr > F References

HNf Binder 2 1.2797 53.805 0.0001473*
[1] M. Castro, J.A. Sánchez, Fatigue and healing of asphalt mixtures: discriminate
Gradation 1 0.3062 10.989 0.029516* analysis of fatigue curves, J. Transp. Eng. 132 (2) (2006) 168–174.
HGsum Binder 2 3.8575 98.882 2.553E-05* [2] P. Karki. An integrated approach to measure and model fatigue damage and
Gradation 1 0.4121 11.047 0.0292747* healing in asphalt composites, 2014.
[3] J.S. Daniel, Y.R. Kim, Laboratory evaluation of fatigue damage and healing of
HDr Binder 2 0.0162 0.213 0.8137296 asphalt mixtures, J. Mater. Civ. Eng. 13 (6) (2001) 434–440.
Gradation 1 0.0222 0.611 0.4780248 [4] J. Qiu, M.F.C. Van de Ven, S. Wu, et al., Investigating the self healing capability
of bituminous binders, Road Materials and Pavement Design 10 (sup1) (2009)
Her Binder 2 0.1648 14.192 0.0053132*
81–94.
Gradation 1 0.0302 69.945 0.0011177*
[5] H. Mihashi, T. Nishiwaki, Development of engineered self-healing and self-
repairing concrete-state-of-the-art report, J. Adv. Concr. Technol. 10 (5) (2012)
170–184.
on the binder types and gradations, was examined by one-way [6] Q. Lv, W. Huang, X. Zhu, et al., On the investigation of self-healing behavior of
bitumen and its influencing factors, Mater. Des. 117 (2017) 7–17.
analysis of variance (ANOVA), as Table 8 shows. The p-value are [7] S.M. Abtahi, M. Sheikhzadeh, S.M. Hejazi, Fiber-reinforced asphalt-concrete–a
presented on the basis of an overall significant level fixed at 0.05. review, Constr. Build. Mater. 24 (6) (2010) 871–877.
It is noted that both binder type and gradation have significant [8] A. González, J. Norambuena-Contreras, L. Storey, et al., Self-healing properties
of recycled asphalt mixtures containing metal waste: An approach through
effect on HNf, HGsum and Her. However, the healing index from
microwave radiation heating, J. Environ. Manage. 214 (2018) 242–251.
the average deformation growth rate (HDr) can hardly been influ- [9] Q. Lv, W. Huang, F. Xiao, Laboratory evaluation of self-healing properties of
enced by the materials factors, which indicates that the healing various modified asphalt, Constr. Build. Mater. 136 (2017) 192–201.
index of HDr is unreliable to evaluate the healing property between [10] P. Ayar, F. Moreno-Navarro, M.C. Rubio-Gámez, The healing capability of
asphalt pavements: a state of the art review, J. Cleaner Prod. 113 (2016) 28–40.
different asphalt mixtures based on this healing-based cyclic SCB [11] J. Jiang, F. Ni, L. Gao, et al., Developing an optional multiple repeated load test
test. On the other hand, all the healing indexes from fatigue life, to evaluate permanent deformation of asphalt mixtures based on axle load
total dissipated energy and tensile strain (HNf, HGsum and Her) could spectrum, Constr. Build. Mater. 122 (2016) 254–263.
[12] A.K. Khiavi, R. Rasouli, Laboratory evaluation of loading frequency effects on
present the difference of the healing properties of different asphalt HMA self-healing, Constr. Build. Mater. 162 (2018) 1–8.
mixtures very well. [13] J.F. Su, P. Yang, Y.Y. Wang, et al., Investigation of the self-healing behaviors of
microcapsules/bitumen composites by a repetitive direct tension test,
Materials 9 (7) (2016) 600.
6. Conclusions [14] F. Canestrari, A. Virgili, A. Graziani, et al., Modeling and assessment of self-
healing and thixotropy properties for modified binders, Int. J. Fatigue 70
(2015) 351–360.
In this paper, the fatigue properties and healing potential of dif- [15] R. Li, F. Xiao, S. Amirkhanian, et al., Developments of nano materials and
ferent asphalt mixtures were evaluated by a healing-based cyclic technologies on asphalt materials–a review, Constr. Build. Mater. 143 (2017)
SCB test. The influence of notch depths and rest periods on fatigue 633–648.
[16] J. Jiang, F. Ni, Q. Dong, et al., Research on the fatigue equation of asphalt
and healing performance were also carefully investigated. The mixtures based on actual stress ratio using semi-circular bending test, Constr.
findings and conclusions are as follows: Build. Mater. 158 (2018) 996–1002.
294 J. Jiang et al. / Construction and Building Materials 196 (2019) 284–294

[17] G. Saha, K.P. Biligiri, Novel procedural pragmatics of dynamic Semi-Circular strength degradation using digital image correlation methods, Measurement
Bending test for fatigue evaluation of asphalt mixtures, Road Mater. Pavement (2018).
Des. (2017) 1–8. [20] M.R. Ayatollahi, E. Mahdavi, M.J. Alborzi, et al., Stress intensity factors of semi-
[18] Z. Wang, S. Wang, Z. Wang, An analysis on computational load of DIC based on circular bend specimens with straight-through and chevron notches, Rock
Newton-Raphson scheme, Opt. Lasers Eng. 52 (2014) 61–65. Mech. Rock Eng. 49 (4) (2016) 1161–1172.
[19] J. Jiang, F. Ni, Q. Dong, et al., Fatigue damage model of stone matrix asphalt
with polymer modified binder based on tensile strain evolution and residual