Construction and Building Materials 63 (2014) 294–302

Contents lists available at ScienceDirect

Construction and Building Materials

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

Moisture susceptibility evaluation of asphalt mixes based on image

analysis
Soroosh Amelian a,⇑, Sayyed Mahdi Abtahi a, Sayyed Mahdi Hejazi b
a
Department of Civil Engineering, Isfahan University of Technology, Isfahan 84156-83111, Iran
b
Department of Textile Engineering, Isfahan University of Technology, Isfahan 84156-83111, Iran

h i g h l i g h t s

 By applying an image analysis technique, visual rating in boiling water test is converted to a more objective evaluation.
 The stripping results of boiling test show significant linear relationship with TSR and ESR values.
 There is not a significant correlation between RMS values and the boiling test results.

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

Article history: Moisture damage in hot mix asphalt (HMA) is one of the major concerns in durability of flexible pave-
Received 3 September 2013 ments. Numerous methods have been developed to evaluate moisture susceptibility of HMA in last
Received in revised form 27 February 2014 decades. Some of these methods are simpler and less costly (qualitative tests) and some of them are more
Accepted 1 April 2014
reliable.
In this research, a digital image analysis approach was utilized to convert boiling water test (ASTM
D3625) from visual rating to objective evaluation. Some laboratory tests were conducted on specimens
Keywords:
to compare the stripping percentages obtained from image analysis of the boiling water test and modified
Asphalt mixture
Moisture susceptibility
Lottman test (AASHTO T283) results. In AASHTO T283 test, in addition to Indirect tensile test, the
Image analysis dynamic modulus |E| test and the Marshall stability test were performed. Therefore, three criteria; ten-
Stripping sile strength ratio (TSR), |E| stiffness ratio (ESR) and Retained Marshall stability (RMS) were used to com-
Dynamic modulus pare the results of the two methods. The dynamic modulus test was conducted in indirect tension mode;
Indirect tensile strength and a linear viscoelastic solution was used for calculation of |E|.
Findings showed that the results of boiling water test have significant relationship with TSR and ESR.
Good correlation was found between three tests; however, the results of boiling test did not show signif-
icant relationship with RMS.
Ó 2014 Elsevier Ltd. All rights reserved.

1. Introduction can be classified into two major categories including those per-
formed on loose mixtures, such as the static immersion test and
Moisture susceptibility is an HMA mixture’s tendency toward the boil test, and those performed on compacted mixtures, such
stripping [1]. Stripping is fundamentally the loss of bond between as the immersion compression, indirect tensile strength, and mod-
the asphalt and aggregate. Moisture damage continues to be one of ulus tests [4]. Advantages of tests in loose mixtures, is that they are
the major causes of premature failure of HMA pavements. Moisture simpler and less costly to run than tests conducted on compacted
damage in HMA occurs due to a loss of adhesion and/or cohesion, specimens; and also they require simpler equipment and proce-
resulting in reduced strength or stiffness of the HMA and the devel- dures. But defining a pass/fail criterion is not an easy task for most
opment of various forms of pavement distress [2,3]. of these tests. For example, visual evaluation is used in the static
Numerous laboratory tests have been developed over the years immersion test to determine the degree of stripping below or
to identify the moisture sensitivity of HMA. The tests for identify- above 95%, a criterion that is not very repeatable between different
ing the moisture damage potential of an asphalt/aggregate mixture operators and different laboratories [5]. To eliminate the disadvan-
tage from this test, by using of computer tools, the visual judgment
⇑ Corresponding author. Tel.: +98 913 305 6179; fax: +98 311 391 2700. can be converted to a more objective evaluation. Merusi et al. used
E-mail address: [email protected] (S. Amelian). an image based algorithm to compare stripping resistance of

http://dx.doi.org/10.1016/j.conbuildmat.2014.04.012
0950-0618/Ó 2014 Elsevier Ltd. All rights reserved.
 S. Amelian et al. / Construction and Building Materials 63 (2014) 294–302 295

asphalt mixtures modified with different synthetic wax [6]. A sim- 0.075 0.3 2.36 4.75 12.5 19
ilar work has been done to evaluate the amount of HMA stripping 100 MDL
by Kim et al. [7]. Gradation
AASHTO T283 is one of the most commonly used procedures for Restricted Zone
Control Points
80
determining HMA moisture susceptibility [5]. In this method
Retained indirect tensile strength (ITS) is used as indicator of mois-
ture susceptibility. Nadkarni et al. [8] used the dynamic modulus

Passing %
60
test instead of ITS test in this field. The analysis indicated that there
was no statistical significant difference between the measured TSR
40
and ESR values for the same mixture. The correlation obtained
between the two ratios was fair to good. Because the dynamic
modulus test is nondestructive and is a major input into mechanis- 20
tic empirical pavement design guide, they recommended the ESR
to potentially replace TSR testing to assess field moisture damage
0
for asphalt mixtures [8]. Kim et al. [9] presented the results from
an analytical/experimental study on the dynamic modulus of Sieve Size mm
HMA tested in the indirect tension (IDT) mode using the theory Fig. 1. Gradation of aggregates used in this study.
of linear viscoelasticity. Comparison of results from the axial com-
pression and IDT test methods showed that the dynamic modulus
mastercurves and shift factors derived from the two methods are
in good agreement [9].
Table 1
Properties of asphalt binder.
2. Objective and scope of the research
Properties Standard Measured values Requirements
Density (g/cm3) ASTM D70 1.021 –
Objectives of this research are to:
Penetration at 25 °C (0.1 mm) ASTM D5 66 60–70
Softening point (°C) ASTM D36 49.5 49–56
 Convert the boiling water test from subjective evaluation to a Ductility at 25 °C (cm) ASTM D113 >100 >100
more objective estimation and more reliable test, so that it does Viscosity at 135 °C, cSt ASTM D2170 295.5 –
not depend on visual rating and operator’s judgment.
 Determine if there is a significant relationship between results
of image analysis of boiling water test and results of AASHTO
T283 Test.
Hydrated lime slurry was used as anti-stripping, which 2% hydrated lime by dry
weight of aggregates mixed into water (ratio of 1–3 by weight). The precoated
The scope of work included conducting both test methods on aggregate was then stockpiled for a marination period of 48–72 h to take place
specimens prepared from different types of aggregates. Digital chemical reactions on the surface of aggregates. Zycosoil was added in the molten
images taken from boiling water test were analyzed and compared asphalt with the amount of 0.5% by weight of asphalt binder and mixed thoroughly
[10].
with TSR, ESR and RMS obtained on dry and moisture conditioned
specimens after AASHTO T283.
3.3. Boiling water test

3. Material and methods The boiling water test is a simple test to assess the effect of water on the adhe-
sion between aggregate and asphalt binder. This test involves immersion of bitumi-
3.1. Materials nous-coated aggregate in boiling water for 10 min and the retained coated area is
evaluated by visual rating. According to ASTM D3625, 250 g of the loose mixture
Five types of aggregates with different mineralogy were used in two groups to samples with temperature of 85–95 °C were placed in approximately 800 mL dis-
prepare asphalt mixtures with different moisture sensitivity. The first group tilled boiling water for 10 min.
included limestone and slag–limestone aggregate (substituting the coarse fraction As mentioned before, the boiling test is qualitative which depends on judg-
of limestone aggregate with slag) that with high carbonate content has a stronger ment of the operator. In order to eliminate this disadvantage, a digital image anal-
bond with asphalt binder. The second group comprised of quartzite, granite and ysis approach was employed to convert the test from subjective rating to a more
andesite aggregates. These aggregates with high silica content have shown poten- objective evaluation. After cooling the boiling mixture to room temperature, the
tial for stripping. BOF slag was procured from Zob-Ahan steel manufacturing com- water was decanted and the wet mixture emptied onto an absorbent towel and
pany located in Isfahan city. The gradation of all aggregates is shown in Fig. 1. Two allowed to dry for an hour and then some digital color photographs were taken
types of anti-stripping additives were used; hydrated lime as filler and a nano- from the mixture. Two image processing software were employed; in the first
based material namely Zycosoil as liquid anti-stripping produced by Zydex Indus- step, the images in JPEG format were imported into Image-Pro Plus software in
tries India. It is a water soluble organosilane compound that has been designed order to segment, detect and separate the background and sparkles of the mix-
to reduce the moisture susceptibility of the compacted asphalt concrete. [10]. ture’s image. The modified images were saved in TIFF format to avoid decrease
Asphalt binder with 60/70 penetration grade was procured from Isfahan of quality and size and then transferred into Image Toll software. In this step,
petroleum refinery company and its basic properties are presented in Table 1. the color images were transformed to 8-bit grayscale images and thresholding
was used to create binary images. In digital images, each pixel has a luminance
value, regardless of its color. Luminance can also be described as brightness or
3.2. Mix design and specimens preparation intensity, which can be measured on a scale from black (zero intensity) to white
(full intensity). Most image file formats support a minimum of 8-bit grayscale,
The superpave mix design was utilized to determine the optimum binder con- which provides 256 levels of luminance per pixel and range from 0 (perfect black)
tent of HMA mixtures. The target air void was selected 4% for all mixtures. The opti- to 255 (perfect white). According to ASTM D3625, any thin, brownish, translucent
mum asphalt content was obtained 4.6% for limestone, 5% for slag–limestone, 4.9% areas should be considered fully coated. Therefore different threshold levels were
for quartzite, 4.4% for granite and 5.1% for andesite aggregate. All mixtures were applied to images and then a threshold value of 65 was selected to distinguish the
mixed in laboratory and then compacted on the IPC Servopack Gyratory Compactor. white and black pixels (stripping and bituminous-coated aggregate respectively).
The specimens were compacted into a 100 mm (4 in.) diameter gyratory mold and This threshold level seems to be a more realistic value to recognize the stripping
approximately a height of 65 mm. In TSR testing, limited additional specimens were areas of samples. However, applying different threshold levels indicated that the
compacted by a standard Marshall hammer to assess the effect of compaction threshold values between 62 and 68 did not significantly affect in stripping per-
method. centage results; but for a consistent comparison, it is important to use the same
296 S. Amelian et al. / Construction and Building Materials 63 (2014) 294–302

threshold value for all tests. After applying the specific threshold level, the number 3.4. Modified Lottman test
of black and white pixels was counted and finally the amount of sample surface
and stripping areas of mixtures were determined. According to AASHTO T283, six specimens with 100 mm in diameter and
The sparkle of bitumen is the main concern in image processing of asphalt mix- approximately 65 mm in height were prepared for each type of aggregate. The
tures. The light affords fresh bitumen to shine and this feature causes significant specimens were compacted to obtain 7.0 ± 0.5% air void in each sample. After
error in results of image processing, because when the images are transformed to removal from the mold, the specimens were stored at room temperature for 24 h
grayscale, the software identifies these sparkle areas as white pixels. To solve this and then the bulck specific gravity and the air voids were determined. The speci-
problem, two approaches can be followed. The first involves providing special light mens were divided into two subsets; one is the dry subset and the other is moisture
conditions during photography capture. This work may need some special tools conditioned. The dry subset was wrapped in leak proof plastic and then placed in
such as special cameras or lighting apparatus. The second approach is the elimina- 25 °C (77°F) for 2 h before testing. The other subset was conditioned as described
tion of the light reflection of particles utilizing a software to process the images. In below:
this study, it was decided to use tools and provide a condition as simple as possible
to make the method available in laboratory without the need of a special apparatus.  The samples were saturated in vacuum container for approximately 10 min to a
Therefore, a common digital camera (Samsung GT-I9100) was employed to capture degree of saturation between 70% and 80%.
the images of the samples with resolution of 3264  2448. The images were taken  After saturation, the samples were wrapped in plastic bags containing
with sufficient and indirect light to prevent creation of excessive sparkles, and then 10 ± 0.5 mL of water. Consequently, they were placed in a freezer at a temper-
the rest of sparkles were eliminated in software. The indirect light was provided by ature of 18 °C (0°F) for 16 h.
applying strong light behind opaque glass. In such condition, the creation of light  The samples were placed in a water bath at 60 °C (140°F) for 24 h.
reflections was significantly reduced, while there was still sufficient light for proper  The specimens were placed in water bath at 25 °C (77°F) for 2 h and then tested.
segmentation. An example of sparkle elimination is shown in Fig. 2. It is recom-
mended to use a light color as background of samples in order to avoid light reflec- For moisture susceptibility evaluation of mixtures, the indirect tensile test, the
tion. A light green towel was used as background. With this color, background dynamic modulus test and the Marshall stability test were conducted on the
detection can be done adequately. specimens.

(a)

The sparkle
of bitumen

(b) (c)

Fig. 2. (a) Captured image, (b) Thresholding without sparkles elimination and (c) Thresholding with sparkle elimination.

Fig. 3. (a) IDT and (b) Dynamic modulus, tests equipment.

 S. Amelian et al. / Construction and Building Materials 63 (2014) 294–302 297

Table 2 TSR ¼ ðS2 =S1 Þ  100

Results of boiling water test using image analysis technique.
where S1 is the average tensile strength of the dry subset, while S2 is the average
Mix type No additive Hydrated lime Zycosoil tensile strength of the conditioned subset.

Average SD Average SD Average SD

3.4.2. Dynamic modulus |E| test
Surface stripping (%) The |E| test was conducted on conditioned and unconditioned specimens to
Quartzite 40.3 3.37 3.5 0.21 1.4 1.25 determine the |E| stiffness ratio (ESR) of samples. In this study, the E test was con-
Granite 15.8 0.98 1.43 0.10 1.37 0.07 ducted in indirect tension mode (Fig. 3b) and an analytical solution using the theory
Andesite 86.5 7.20 1.39 0.11 0.82 0.06 of linear viscoelasticity, as presented by Kim et al. employed as follows [9]:
Limestone 1.6 0.05 – – – –
Slag–limestone 1.3 0.03 – – – – P 0 b1 c 2  b 2 c 1
jE j ¼ 2
pad c2 V 0  b2 U0
Note: SD: Standard deviation.

b1 U 0  c 1 V 0
m¼
b2 U 0 þ c2 V 0
3.4.1. Indirect tensile strength test where P0 is the amplitude of sinusoidal load used in the dynamic modulus test. a is
In the modified Lottman test, a conventional method for moisture sensitivity the loading strip width, d is the thickness of the specimen, U0 and V0 are the constant
investigation of asphalt mixtures is the indirect tensile (IDT) strength test amplitudes of horizontal and vertical displacements respectively and m is the Pois-
(Fig. 3a). A load was applied along the diameter of the specimens at a constant rate son’s ratio. By assuming the value of Poisson’s ratio the measurement of vertical
of 50 mm (2 in) per minute and the maximum load was recorded. The ratio of aver- deformations are not required [11]. Since a value of 0.35 for Poisson’s ratio has
age tensile strength of the conditioned and dry specimens or tensile strength ratio been found to be reasonable for asphalt mixture at 25 °C (77°F) [12]. Therefore, in
(TSR) was computed by the following equation: this research, a Poisson’s ratio of 0.35 was assumed and only the horizontal

Mixture Granite HMA Quartzite+HL HMA

Original images

Background
and sparkles
detection

Determination of
the sample area

Determination of
the stripping area

Stripping% 15.8 3.5

Fig. 4. Image analysis steps.

298 S. Amelian et al. / Construction and Building Materials 63 (2014) 294–302

Table 3
Results of indirect tensile test.

Mix Condition Average air voids (%) Saturation (%) Average tensile strength (kPa) Average TSR (%)
Limestone Conditioned 6.7 79.0 1581.3 83
Unconditioned 6.6 – 1906.5
Slag–limestone Conditioned 7.0 80.2 1721.8 93
Unconditioned 7.2 – 1845.8
Granite Conditioned 7.1 75.5 1238.8 70
Unconditioned 7.3 – 1767.3
Granite + HL Conditioned 7.2 72.5 1826.5 105
Unconditioned 6.9 – 1741.8
Granite + ZS Conditioned 6.6 79.3 1539.2 92
Unconditioned 6.8 – 1675.6
Quartzite Conditioned 6.8 73.4 751.7 41
Unconditioned 6.7 – 1837.6
Quartz + HL Conditioned 6.7 70.7 1599.9 89
Unconditioned 6.9 – 1796.2
Quartz + ZS Conditioned 7.0 78.1 1311.2 76
Unconditioned 7.2 – 1723.2
Andesite Conditioned 6.9 72.8 250.1 18
Unconditioned 6.9 – 1384.9
Andesite + HL Conditioned 7.1 71.1 1194.0 89
Unconditioned 7.3 – 1339.4
Andesite + ZS Conditioned 7.0 72.6 1021.8 78
Unconditioned 6.8 – 1307.7

displacements were measured during the tests. The applied load level was damage. And granite mixture with average stripping of 15.8%
adjusted so that the horizontal strain in the specimens was in the range of shows moderate but not sufficient resistance to stripping.
100–130 microstrains. After preconditioning of specimens to obtain uniform
Both anti-stripping additives, including hydrated lime slurry
deformation, the test was continued with sinusoidal loading at frequency of 1.0 Hz
and the average amplitude of the load and the displacements over the last five
and Zycosoil, substantially improved the stripping resistance of
loading cycles were measured for calculation of |E|. HMA mixtures. The stripping percentage for all mixtures treated
The coefficients, b1, b2, c1, and c2 were calculated for specimen diameter and by anti-stripping agents was less than 3.5. In comparison between
gauge length of 100 mm with the following values: all mixtures, the andesite HMA with stripping percentage of 86.5
b1 ¼ 0:0711; b2 ¼ 0:0013; c1 ¼ 0:0054; c2 ¼ 0:0199 showed the maximum stripping potential; on the other hand, the
andesite HMA treated by Zycosoil with stripping percentage of
The |E| stiffness ratio was found with the following equation [8]:
0.82% indicated the minimum stripping potential.
ESR ¼ ðE2 =E1 Þ  100 Fig. 4 shows the image analysis steps for two types of aggre-
gates with different stripping percentages. As mentioned before,
where E1 is the average dynamic modulus of the dry subset, while E2 is the average
dynamic modulus of the conditioned subset. the images should be taken with sufficient and indirect light. Some
aggregates need extra lighting in comparison with conventional
3.4.3. Marshall stability test
aggregates due to their darker colors. In this study, for andesite
In the Marshall-Immersion test, Marshall stability is used as a strength param- HMA mixtures, images were captured by using flash light in order
eter rather than compressive strength in ASTM D1075 and AASHTO T165 to evalu- to identify stripping areas correctly.
ate moisture sensitivity (5). The moisture conditioning of specimens is identical in
both methods except for the freeze cycle as performed in AASHTO T283. In the pres-
ent study, the specimens were conditioned according to AASHTO T283 (like the |E| 4.2. Indirect tensile test
and the IDT tests) and the Retained Marshall stability (RMS) was then found with
the following equation [13]: The average indirect tensile strength of the mixtures is given in
RMS ¼ ðMS2 =MS1 Þ  100 Table 3. It can be seen that among the original mixtures without
using anti-stripping additives, only the limestone and slag–lime-
where MS1 is the average Marshall stability of the dry subset, while MS2 is the aver- stone HMA mixtures had TSR values above 75%. Also all the HMA
age Marshall stability of the conditioned subset.
mixtures prepared with anti-stripping additives had TSR more
than 75%. In addition, all the hydrated lime-treated aggregate
4. Results and discussion had TSR values of more than 85%. The minimum TSR is related to
andesite HMA (18%) while the maximum TSR belongs to granite
4.1. Boiling water test HMA modified by hydrated lime (105%).
In AASHTO T283, the results of visually estimated moisture
Thirty-three samples were tested in boiling water in order to damage in fractured specimens, is needed to be reported. After
provide three replicates for each type of mixture. Table 2 shows the indirect tensile test was conducted, some images were taken
the average results and standard deviation of the test. As described from fractured specimens and the amount of moisture damage
in the experimental procedures section, the specific threshold level determined by image analysis technique with the same procedure
was applied to distinguish the stripping and non-stripping areas of described in boiling water test. In Fig. 5 the average stripping per-
loose asphalt mixtures. The stripping percentage of each sample centages of fractured specimens are correlated with the average
was determined by dividing the amount of stripped surface by stripping percentages of boiling water test. With R2 = 0.905, there
the amount of total surface of sample. With regard to Table 2, in is a close relationship between two stripping criteria.
mixtures with no additive, limestone and slag–limestone mixtures For limestone and slag–limestone aggregates, the IDT test was
with stripping of 1.6% and 1.3% show appropriate resistance to repeated by using a Marshall hammer instead of gyratory compac-
stripping; conversely, quartzite and andesite mixtures with tor to compact the specimens. The TSR obtained from two proce-
stripping of 40.3% and 86.5% show excessive potential to moisture dures did not show a significant difference, but the ITSs from
 S. Amelian et al. / Construction and Building Materials 63 (2014) 294–302 299

ESR also belongs to granite HMA samples prepared with Zycosoil

2
additive with the value of 143%.
R = 0.905

4.4. Marshall stability test

Stripping in IDT Test %

Table 5 illustrates the average Marshall stability for dry and

conditioned specimens. The maximum moisture sensitivity is
related to the andesite aggregates with RMS of 67%. In comparison
with IDT and |E| tests, it can be seen that the RMS is more than the
TSR and ESR for all HMA mixtures prepared without using anti-
stripping additives. However in anti-stripping treated samples,
there is not a specific pattern.

4.5. Correlation analysis between boiling test and AASHTO T283 test

In order to find the relationship between the boiling water test

Stripping in Boiling Test % and the modified Lottman test, the results of boiling test and
AASHTO T283, including TSR, ESR and RMS were correlated. As it
Fig. 5. Correlation of stripping percentages in boil test and IDT test obtained from
image analysis.
can be seen from experimental results, the percentages of bitumi-
nous-coated aggregates in boiling water test are always equal or
less than 100%; while in the modified Lottman test, in some sam-
three specimens of each subset were closer and more consistent in ples, the ratio of strength or stiffness for conditioned and dry spec-
the gyratory compacted mixes than the Marshall compacted mixes. imens is more than 100%. The moisture susceptibility tests are
employed to assess the moisture destruction and to determine
the amount of decrease in strength and/or stiffness of HMA; and
the purpose of them is not to assess the improvement in properties
4.3. Dynamic modulus |E| test of HMA in presence of water. Therefore, it seems reasonable to
limit the ratios of strength and stiffness which are more than
The average dynamic modulus of specimens is shown in Table 4. 100–100%.
In original mixtures without anti-stripping additives, the quartzite Figs. 6–8 show the correlation between stripping percentages of
and andesite samples with ESR value of 41% and 14% respectively, each mixture obtained from image analysis after the boiling test
were extremely sensitive to moisture destruction; however, gran- and TSR, ESR and RMS criteria. In each figure, the stripping percent-
ite mixture with an ESR value of 77% showed relatively good per- ages are correlated once with the original ratios obtained from
formance in dynamic modulus test compared to indirect tensile AASHTO T283 test, and another time with ratios limited to 100%.
test with TSR value of 70%. In all HMA mixtures with anti-stripping As it can be seen in all cases, the ratios limited to 100% have a bet-
additives, the ESR values were more than 80%. It can be observed ter relationship with the stripping percentages than the original
that the ESR in all samples treated with Zycosoil, is more than ratios. Fig. 6 indicates that the coefficient of determination for ori-
the same samples in TSR testing. Like the boiling water and the ginal TSR ratios (6a) and TSR ratios limited to 100% (6b) is 0.87 and
indirect tensile tests, the andesite HMA showed the maximum 0.89 respectively. Correlation of ESR ratios is illustrated in Fig. 7.
moisture susceptibility with an ESR value of 14%. The maximum The coefficient of determination for original ratios is 0.70 (7a);

Table 4
Results of dynamic modulus test.

Mix Condition Average air voids (%) Saturation (%) Average dynamic modulus (MPa) Average ESR (%)
Limestone Conditioned 6.8 72.0 1287.8 88
Unconditioned 6.8 – 1463.4
Slag–limestone Conditioned 7.0 74.3 2000.7 91
Unconditioned 7.2 – 2198.6
Granite Conditioned 6.7 73.7 2067.7 77
Unconditioned 6.8 – 2700.2
Granite + HL Conditioned 7.2 77.3 3180.4 127
Unconditioned 7.3 – 2503.8
Granite + ZS Conditioned 6.8 72.8 3170.6 143
Unconditioned 6.7 – 2217.2
Quartzite Conditioned 7.3 76.4 1378.7 41
Unconditioned 7.1 – 3333.4
Quartz + HL Conditioned 6.6 78.1 2892.8 87
Unconditioned 6.7 – 3325.1
Quartz + ZS Conditioned 6.5 79.6 2805.6 87
Unconditioned 6.8 – 3228.7
Andesite Conditioned 6.9 77.5 319.0 14
Unconditioned 7.2 – 2343.3
Andesite + HL Conditioned 7.2 74.2 1567.5 84
Unconditioned 7.5 – 1876.8
Andesite + ZS Conditioned 6.5 74.8 1502.7 108
Unconditioned 6.8 – 1387.6
300 S. Amelian et al. / Construction and Building Materials 63 (2014) 294–302

Table 5
Results of Marshall stability test.

Mix Condition Average air voids (%) Saturation (%) Average Marshall stability (KN) Average RMS (%)
Limestone Conditioned 6.8 77.0 10.35 120
Unconditioned 7.0 – 8.63
Slag–limestone Conditioned 7.3 75.8 11.25 115
Unconditioned 7.1 – 9.78
Granite Conditioned 7.0 74.7 8.46 93
Unconditioned 6.9 – 9.10
Granite + HL Conditioned 7.2 73.2 10.06 115
Unconditioned 7.4 – 8.75
Granite + ZS Conditioned 6.5 74.9 7.28 85
Unconditioned 6.8 – 8.54
Quartzite Conditioned 6.6 72.8 6.92 73
Unconditioned 6.7 – 9.47
Quartz + HL Conditioned 6.6 79.1 8.25 92
Unconditioned 6.5 – 8.96
Quartz + ZS Conditioned 7.1 78.2 7.81 89
Unconditioned 6.9 – 8.78
Andesite Conditioned 6.4 71.9 5.86 67
Unconditioned 6.6 – 8.76
Andesite + HL Conditioned 6.9 77.3 7.17 85
Unconditioned 7.2 – 8.43
Andesite + ZS Conditioned 6.9 72.6 6.78 81
Unconditioned 6.6 – 8.37

(a) (b)
Y= -0.99X+89.1 Y= -1.02X+91.2
R2 = 0.87 R2 = 0.89
Stripping (%)

Stripping (%)

TSR (%) TSR (%)

Fig. 6. The correlation between tensile strength ratios and stripping percentages in boiling test: (a) Original TSR ratios and (b) TSR ratios limited to 100%.

(a) (b)
Y= -0.63X+68.2 Y= -0.95X+89.3
R2 = 0.70 R2 = 0.93
Stripping (%)

Stripping (%)

ESR (%) ESR (%)


Fig. 7. The correlation between |E | stiffness ratios and stripping percentages in boiling test: (a) Original ESR ratios and (b) ESR ratios limited to 100%.
 S. Amelian et al. / Construction and Building Materials 63 (2014) 294–302 301

(a) (b)
Y= -0.94X+101 Y= -1.86X+177.5
R2 = 0.38 R2 = 0.58

Stripping (%)

Stripping (%)
RMS (%) RMS (%)

Fig. 8. The correlation between Retained Marshall stability and stripping percentages in boiling test: (a) Original RMS ratios and (b) RMS ratios limited to 100%.

however, this value has increased to 0.93 for ESR ratios limited to  Two anti-stripping additives used in this study including
100% (7b). hydrated lime and a nano-based material (Zycosoil) were effec-
Moreover, in the original ratios, the TSR values with R2 = 0.87 tive and improved the moisture susceptibility of HMA mixtures.
have the best correlation and in the ratios limited to 100%, the  There was a close relationship between stripping percentages
ESR values with R2 = 0.93 have the best correlation with the strip- obtained from boiling water test and stripping percentages of
ping percentages obtained from image analysis of the boiling test. fractured specimens after the indirect tensile test.
The correlation analysis shows that with R2 = 0.89 and R2 = 0.93,  Unlike TSR and ESR criteria, it was observed that there was not a
the boiling water test results have significant linear relationship significant correlation between the results of Retained Marshall
with the TSR and ESR values respectively. stability (RMS) and the boiling water test.
Unlike these two criteria, it was observed that with R2 = 0.38 for
original ratios and R2 = 0.58 for ratios limited to 100% (Fig. 8), there
is not a significant correlation between RMS criterion and boiling
Acknowledgments
water results. This phenomenon could confirm the fact that the
Marshall test is not sensitive enough to measure the effect of mois-
The Authors would like to thank the Isfahan Technical & Soil
ture destruction.
mechanic Laboratory and the Isfahan Highway Administration for
providing laboratories and equipments needed to complete this
5. Summary and conclusions
study. The Authors gratefully acknowledge Dr. Ahmad Goli for
his useful guidance.
In this study, a digital image analysis approach was applied to
find whether it is practical to convert the visual and subjective
evaluation of stripping in boiling water test to a more objective References
evaluation and to convert the boiling method to more reliable test.
For this purpose, the boiling water test was conducted on HMA [1] Construction of hot mix asphalt pavements. Manual series no. 22, 2nd ed.
Asphalt Institute; 1998.
mixtures prepared with five different types of aggregate and two [2] Hicks RG. NCHRP synthesis of highway practice 175: moisture damage in
anti-stripping additives; and then image analysis was utilized to asphalt concrete. Washington, DC: Transportation Research Board, National
distinguish stripping and bituminous-coated aggregate. Further- Research Council; 1991.
[3] Solaimanian M, Kennedy TW, Elmore WE. Long-term evaluation of stripping
more, the indirect tensile test, the dynamic modulus |E| test and and moisture damage in asphalt pavements treated with lime and
the Marshall stability test were conducted on conditioned and antistripping agents. Report CTR 0-1286-1F. Texas Department of
unconditioned specimens according to AASHTO T283 to determine Transportation, Center for Transportation Research, University of Texas at
Austin; 1993.
the level of relationship between results of image analysis of boil- [4] Solaimanian M, Bonaquist RF, Tandon V. NCHRP report 589: improved
ing test and results of the modified Lottman test. conditioning and testing procedures for HMA moisture
Statistical analysis indicated close relationship between the susceptibility. Washington, DC: Transportation Research Board; 2007.
[5] Solaimanian M, Harvey J, Tahmoressi M, Tandon V. Test methods to predict
results of boiling water test obtained from image analysis tech-
moisture sensitivity of hot-mix asphalt pavements. Moisture sensitivity of
nique, with TSR and ESR values limited to 100%. With R2 value of asphalt pavements: a national seminar, San Diego, California; 2003.
0.89 and 0.93 for TSR and ESR, the modified boiling test by image [6] Merusi F, Caruso A, Roncella R, Giuliani F. Moisture susceptibility and stripping
analysis technique could be considered as an effective method to resistance of asphalt mixtures modified with different synthetic waxes. In:
Transportation Research Record: Journal of the Transportation Research Board,
predict moisture destruction of HMA mixtures in a more objective No. 2180, Transportation Research Board of the National Academies,
manner in comparison with original boiling test. Based on the lab- Washington, DC; 2010. p. 110–20.
oratory experiments and analysis, the following conclusions can [7] Kim YR, Pinto I, Park SW. Experimental evaluation of anti-stripping additives in
bituminous mixtures through multiple scale laboratory test results. Construct
also be summarized: Build Mater 2012;29:386–93.
[8] Nadkarni AA, Kaloush KE, Zeiada WA. Using the dynamic modulus test for
 Image analysis technique was found an efficient method to con- moisture susceptibility evaluation of asphalt mixtures. In: Transportation
Research Record: Journal of the Transportation Research Board, No. 2127,
vert estimation of stripping in boiling water test from subjective Transportation Research Board of the National Academies, Washington, DC;
rating to more objective evaluation. 2009. p. 29–35.
302 S. Amelian et al. / Construction and Building Materials 63 (2014) 294–302

[9] Kim YR, Seo Y, King M, Momen M. Dynamic modulus testing of [11] Katicha SW. Development of laboratory to field shift factors for hot-mix
asphalt concrete in indirect tension mode. In: Transportation Research asphalt resilient modulus. Dissertation 2003. Virginia Polytechnic Institute
Record: Journal of the Transportation Research Board, No. 1891, and State University; 2003.
Transportation Research Board of the National Academies, Washington, DC; [12] Standard test method for indirect tension test for resilient modulus of
2004. p. 163–73. bituminous mixtures. ASTM Standard D 4123-82, ASTM International, West
[10] Kim J, Moore JR. Laboratory evaluation of ZycoSoil as an Conshohocken; 1982.
anti-stripping agent on superpave mixtures. National Center for [13] Test method for resistance to plastic flow of bituminous mixtures using
Asphalt Technology, Auburn University; 2009 [submitted for Marshall apparatus. ASTM Standard D 1559-89, ASTM International, West
publication]. Conshohocken; 1989.