Using two-way ANOVA and hypothesis test in evaluating crumb rubber

modication (CRM) agitation effects on rheological properties of bitumen

Sassan Aaki

, Milad Memarzadeh
1
School of Civil Engineering, College of Engineering, University of Tehran, Tehran, Iran
a r t i c l e i n f o
Article history:
Received 22 May 2010
Received in revised form 31 October 2010
Accepted 13 November 2010
Keywords:
Bitumen
Crumb rubber modication
ANOVA
Hypothesis test
Low/high shear blending
Performance grade
a b s t r a c t
The effect of CRM aspects on rheological properties of asphalt binder such as improvement in the perfor-
mance grade (PG) for low, intermediate, and high service temperatures are evaluated and the binders
dynamic viscosity changes were studied in this research. Enhancement also resulted in the binders fati-
gue and rutting indices. At the same time, the effects of low/high shear blending conditions on the rhe-
ological properties were investigated. The effect of interactions between blending conditions and
modier content were studied by utilizing the analysis of variance (ANOVA) method and hypothesis tests.
2010 Elsevier Ltd. All rights reserved.
1. Introduction
1.1. Background
Use of crumb rubber in modication (CRM) of asphalt binders
causes improvement in performance of modied asphalt binders
both in low and high service temperatures [1]. CRM also improves
the performance aspects of the asphalt mixture such as thermal
susceptibility, elastic behavior, fatigue cracking resistance and
aging stability [1]. Researchers have found that the addition of
1018% crumb rubber can improve high and low temperature
properties of asphalt binders [25]. The digestion process of CRM
with binder is rather complicated and is dependent on the vari-
ables of CRM and binder (i.e., type, percentage, size, grade, etc.)
in addition to the mixing conditions (temperature, mixing time,
agitation, etc.) [6]. Lee et al. [7] investigated the performance prop-
erties of CRM binders due to CRM processing method and percent-
age through selected SuperPAVE rheological binder tests. The CRM
used in Lee et al. study was obtained from two sources. One source
used the ambient grinding method to process scrap passenger car
tires into crumb rubber and the other used the cryogenic grinding
process [7].
The effect of CRM on the properties of the asphalt binders were
evaluated in former and current studies [815]. In almost all of
these research discussions, conclusions are qualitative and based
on the raw test results and statistical methods are used scarcely
to investigate the results and compare the cases. Thodesen et al.
[16] utilized statistical methods in the development of an empiri-
cal model to determine G

/Sin d in crumb rubber modied binders.

They develop the empirical model depicting the changes in values
of G

/Sin d and failure temperatures. Besides, they use condence

intervals to evaluate these changes. Liu et al. [17] evaluated the
performance of different modied binders with different CRM con-
tents, particle size, and CRM type by using the analysis of variance,
ANOVA.
1.2. Research objective and scope
The main objective of this research is to evaluate the efciency
of crumb rubber content and low/high shear blending conditions
on the binder rheological properties. The comparison between
crumb rubber content and low/high shear blending capability on
the PG improvement, fatigue and rutting parameters improvement
was accomplished by using ANOVA method. Furthermore, the
hypothesis test was utilized to compare the efciency of the low
and high shear blending conditions and also to evaluate the mea-
sure of effectiveness of each condition for improving the PG of
the base binder.
0950-0618/$ - see front matter 2010 Elsevier Ltd. All rights reserved.
doi:10.1016/[Link].2010.11.028

Corresponding author. Tel.: +98 21 6111 2273; fax: +98 21 6640 3808.
E-mail addresses: aaki@[Link] (S. Aaki), [Link]@[Link]
(M. Memarzadeh).
1
Tel.: +98 917 1831597.
Construction and Building Materials 25 (2011) 20942106
Contents lists available at ScienceDirect
Construction and Building Materials
j our nal homepage: www. el sevi er . com/ l ocat e/ conbui l dmat
2. Materials and test methods
2.1. Base binder
A PG58-22 bitumen was selected as the base binder. The base binder was ob-
tained from one of the bitumen production plants in Iran [1]. The characteristics
of the base binder are presented in Table 1. The Superpave performance grading
(PG) testing protocol (ASTM-D6373) was used to evaluate the base and modied
bitumen samples [18] and the PG of binders was specied. High service tempera-
ture (HT) for a binder is determined as the temperature at which the G

/Sin d is
greater than 1 kPa for unaged binder and greater than 2.2 kPa for the rolling thin
lm oven aged condition (RTFO) (ASTM-D2872) [19]. Low service temperature
(LT) is 10 C less than the temperature at which the pressurized aging vessel
(PAV) (ASTM-D6521) [20] aged binder exhibits creep stiffness of less than
300 MPa and an m-value of greater than 0.3 at 60 s of loading (ASTM-D6648)
[21]. In addition to the performance requirements at high and low service temper-
ature; there is a limiting maximum stiffness at the intermediate service tempera-
ture (IT) to alleviate fatigue cracking, at which the binders G

Sin d in the PAV

aged condition does not exceed 5000 kPa. The dynamic shear modulus, G

, and
the phase angle, d, of binders were determined by a dynamic shear rheometer
(DSR) using parallel plate geometry at 10 rad/s (1.59 Hz) (ASTM D7175) [22].
According to the standard test method test specimens were 25 mm in diameter
by 1 mm thick for high temperature (HT) tests (4682 C) and 8 mm in diameter
by 2 mm thick for intermediate temperature (IT) tests (440 C).
The dynamic viscosity tests are accomplished with the rotational viscometer
that is the rotating spindle-type viscometer (ASTM-D4402) [23]. Dynamic viscosity
determined at 95, 115, 135, 155, 165 and 175 C for each sample. In each temper-
ature the viscosity measured at the shear strain rate of 1.568 1= sec by the spin-
dle SC27, which has the diameter of 11.76 mm. The dynamic viscosity test is
accomplished on the unaged original binder, RTFO bitumen and PAV bitumen sam-
ples to examine the effect of CRM and also evaluate the modication resistance
against aging process [1]. Researches show that the PAV aging induced damages
to the bitumenpolymer network structure causes changes in the components of
the bitumen and degradation of modied binders. The function of the modier is
thus signicantly impaired. Eventually the rheological and mechanical properties
of the modied binder become similar to the base bitumen [1,2427].
2.2. Bitumen modication
The base bitumen, PG 58-22, was blended with various amounts of crumb rub-
ber, 1016%. Crumb rubber passing through a No. 50 (0.3 mm) sieve was used for
this purpose. The amounts of crumb rubber added to the bitumen were 10%, 14%
and 16% of the weight of the base bitumen (PG 58-22). It was added when the base
bitumen reached 170 C and blended for 240 min at 5500 rpm using a high shear
homogenizer mixer. In this process mixing temperature and blending speed were
selected so that a sufcient vortex was formed in the mixing chamber during the
entire process [1]. The low shear blended samples were mixed by a low shear rotary
blender at 600 rpm at 170 C.
3. Test results evaluation
3.1. Fatigue and rutting parameters evaluation
The enhancements resulted due to modication in G

Sin d,
which is known as the fatigue parameter, at 19 C and G

/Sin d,
the rutting parameter, at 64 C in comparison with the base binder
are shown in Figs. 1 and 2, respectively. As shown in Fig. 1, the
G

Sin d at 19 C has a minimum value in 14% crumb rubber content

and then increases as the result of increasing the content of crumb
rubber (CR) to 16% for both low and high shear blended samples. It
can be deducted from Fig. 2 that the rutting parameter, G

/Sin d, in-
creases as the result of increasing the crumb rubber content in case
of both low and high shear blended samples and the maximum
enhancement has occurred at 16% CR content. Figs. 35 show the
G

Sin d versus temperature for different CR contents in low/high

shear blended samples. It is resulted that the amount of G

Sin d
for low shear blended samples is less than high shear blended sam-
ples over the test temperature range. This indicated that the low
shear blending conditions make more enhancements on the IT per-
formance than the high shear blending conditions, which is an
unexpected result and can be studied further. Figs. 6 and 7 show
Table 1
The characteristics of the base asphalt binder.
Penetration grade 6070
Penetration (0.1 mm) at 25 C ASTM D5 70
Softening point (C) ASTM D36 48
Saybolt feurol viscosity (s) at 135 C ASTM E102 144.6
Ductility (cm) at 25 C ASTM D113 >100
Flash point (C) ASTM D92 276
Burning point (C) ASTM D92 362
Fraas break point (C) DIN EN 12593 15
Trichloroethylin solubility ASTM D2042 99.95
Penetration after RTFO 38
Ductility after RTFO (cm) at 25 C >100
Penetration index (PI) 0.9
0.0
0.5
1.0
1.5
2.0
2.5
0.09 0.1 0.11 0.12 0.13 0.14 0.15 0.16 0.17
G
*
.
S
i
n
(
d
e
l
t
a
)

(
M
o
d
i
f
i
e
d
)

/

G
*
.
S
i
n
(
d
e
l
t
a
)

(
B
a
s
e
)

Crumb rubber conetent
G*. Sin(delta) proportion at 19 C
High Shear
Low Shear
Fig. 1. G

Sin d versus crumb rubber content at 19 C.

S. Aaki, M. Memarzadeh/ Construction and Building Materials 25 (2011) 20942106 2095
the G

Sin d versus temperature at different CR contents for low and

high shear blended samples, respectively.
For this reason the most improvement in fatigue parameter oc-
curs at 14% CR content for both low and high shear blended
samples.
Figs. 810 show G

/Sin d, rutting parameter, versus temperature

for different CR contents in low/high shear blended samples. The
G

/Sin d of high shear blended samples are less than of low shear
blended samples over the test temperature range. Based on the
qualitative evaluation of the test results, it can be concluded that
the low shear blending conditions have more effect on the HT per-
formance improvement than the high shear blending conditions,
i.e. low shear blended samples show better HT performance than
the high shear blended samples with equal CR contents. The statis-
tical analysis and quantitative evaluation of the test results is also
discussed in the following sections of this paper. Figs. 11 and 12
show G

/Sin d, rutting parameter, versus temperature at different

CR contents for low and high shear blended samples, respectively.
3.2. PG improvement
The LT improvements of modied binders for both low and high
shear blended samples are shown in Fig. 13. The LT improvement is
dened as extension of the lower bond of the performance grade
0.0
1.0
2.0
3.0
4.0
5.0
6.0
7.0
8.0
9.0
10.0
0.09 0.1 0.11 0.12 0.13 0.14 0.15 0.16 0.17
G
*
/
S
i
n
(
d
e
l
t
a
)

(
M
o
d
i
f
i
e
d
)

/

G
*
/
S
i
n
(
d
e
l
t
a
)

(
B
a
s
e
)


Crumb rubber content
G*/Sin (delta) proportion at 64 C
High Shear
Low Shear
Fig. 2. G

/Sin d versus crumb rubber content at 64 C.

0.00
1.00
2.00
3.00
4.00
5.00
6.00
7.00
8.00
9.00
10 12 14 16 18 20 22 24 26
G
*
.
S
i
n
(
d
e
l
t
a
)


(
M
P
a
)
Temperature (C)
10 % Crumb rubber
High Shear
Low Shear
Fig. 3. G

Sin d versus temperature at 10% CRM binder.

2096 S. Aaki, M. Memarzadeh/ Construction and Building Materials 25 (2011) 20942106
(PG) of base binder resulted by modifying, and calculated as shown
below:
LT improvement LT
Base binder
LT
Modified binder
It can be seen that the binder modied with 14% and 16% CR
makes the most improvement in LT in high and low shear blended
samples, respectively. Another interesting result is that the LT
improvement in low shear blended samples is lower than the high
shear blended ones, so high shear blended samples show better LT
performance than the low shear blended samples with equal CR
contents.
Fig. 14 shows the IT improvement of modied binders for both
low and high shear blended samples. The IT improvement is de-
ned as the amount of lowering the intermediate service temper-
ature of the binder after modication and calculated as shown
below:
IT improvement IT
Base binder
IT
Modified binder
It is obvious that the binder modied with 14% CR has highest
improvement in IT for both low and high shear blended samples.
In this case the IT improvement for low shear blended samples is
higher than the high shear blended ones.
The HT improvement of modied binders for both low and high
shear blended samples is shown in Fig. 15. The HT Improvement is
dened as the amount of rising the high service temperature after
modication and calculated as shown below:
HT improvement HT
Modified binder
HT
Base binder
0
1
2
3
4
5
6
10 12 14 16 18 20 22 24
G
*
.
S
i
n
(
d
e
l
t
a
)


(
M
P
a
)
Temperature (C)
14% Crumb rubber
High Shear
Low Shear
Fig. 4. G

Sin d versus temperature at 14% CRM binder.

0
1
2
3
4
5
6
7
10 12 14 16 18 20 22 24 26
G
*
.
S
i
n
(
d
l
e
t
a
)


(
M
P
a
)
Temperature (C)
16% Crumb rubber
High Shear
Low Shear
Fig. 5. G

Sin d versus temperature at 16% CRM binder.

S. Aaki, M. Memarzadeh/ Construction and Building Materials 25 (2011) 20942106 2097
It is apparent that the binder modied with 14% and 16% CR has
the highest enhancement in HT for high and low shear blended
samples, respectively. In this case, the low shear blended samples
leading to more improvement in HT than the high shear blended
samples.
3.3. Dynamic viscosity
Fig. 16 shows the dynamic viscosity of CRM binders at 135 C. It
is clear that the dynamic viscosities of modied binders increase as
a result of increasing the CR contents. It is clear from Fig. 16 that
the dynamic viscosity of the modied binders for low shear
blended samples is more than the high shear blended samples.
4. Statistical analysis, method
4.1. Analysis of variance (ANOVA)
The analysis of variance is used to evaluate the efcacy of
crumb rubber content and the low/high shear blending conditions
on the binder properties and to determine which has the superior
effect on performance measurement of base binder.
4.1.1. The ANOVA method
The test has two discrete independent variables, X and Y. X can
select n levels which are X
1
; X
2
; . . . ; X
n
; while Y can select m levels
which are Y
1
; Y
2
; . . . ; Y
m
; respectively. If the test is done according
0.5
5.0
0 5 10 15 20 25 30
G
*
.
S
i
n
(
d
e
l
t
a
)


(
M
P
a
)
Temperature (C)
Low shear blending
10% CR
14% CR
16% CR
Fig. 6. G

Sin d versus temperature for low shear blended samples.

1
10
10 12 14 16 18 20 22 24
G
*
.
S
i
n
(
d
e
l
t
a
)

(
M
P
a
)
Temperature (C)
High shear blending
10% CR
14% CR
16% CR
Fig. 7. G

Sin d versus temperature for high shear blended samples.

2098 S. Aaki, M. Memarzadeh/ Construction and Building Materials 25 (2011) 20942106
to each set of parameter levels (X
i
, Y
j
), and the performance test re-
sult is z
ij
, therefore all the z
ij
s are independently related to each
other and follow the normal distribution.
The basic method of two-way ANOVA is as follows [28]:
Mean calculation
z
1
mn
X
n
i1
X
m
j1
z
ij

z
i

1
m
X
m
j1
z
ij

z
j

1
n
X
n
i1
z
ij

z
1
n
X
n
i1

z
i

1
m
X
m
j1

z
j
4:1
where z represents the arithmetic mean value of the measurements.
Sum of squares of deviations calculation
The sum of squares of deviation is
TSS
X
n
i1
X
m
j1
z
ij

z
2
SSX SSY SSE 4:2
0.0
1.0
2.0
3.0
4.0
5.0
6.0
7.0
8.0
9.0
10.0
55 60 65 70 75 80
G
*
/
S
i
n
(
d
e
l
t
a
)


(
k
P
a
)
Temperature (C)
10% Crumb rubber
High Shear
Low Shear
Fig. 8. G

/Sin d versus temperature at 10% CRM binder.

0.0
1.0
2.0
3.0
4.0
5.0
6.0
60 65 70 75 80 85
G
*
/
S
i
n
(
d
e
l
t
a
)


(
k
P
a
)
Temperature (C)
14% Crumb rubber
High Shear
Low Shear
Fig. 9. G

/Sin d versus temperature at 14% CRM binder.

S. Aaki, M. Memarzadeh/ Construction and Building Materials 25 (2011) 20942106 2099
where SSX represents the sum of squares of deviations caused by
factor X, SSY represents the sum of squares of deviations caused
by factor Y and SSE represents the sum of squares of error.
Let,
R
i

X
m
j1
z
ij
; C
j

X
n
i1
z
ij
; G
i

X
m
j1
z
2
ij
; G
i

X
n
i1
z
2
ij
R C
X
n
i1
X
m
j1
z
ij

X
n
i1
R
i

X
m
j1
C
j
4:3
G
X
n
i1
X
m
j1
z
2
ij

X
n
i1
G
i

X
m
j1
G
j
4:4
The sum of squares of deviations can be calculated as:
TSS G
R
2
mn
G
C
2
mn
SSX
1
m
X
n
i1
R
2
i

R
2
mn
SSY
1
n
X
m
j1
C
2
j

C
2
mn
SSE TSS SSX SSY 4:5
0.0
1.0
2.0
3.0
4.0
5.0
6.0
7.0
8.0
9.0
60 65 70 75 80 85
G
*
/
S
i
n
(
d
e
l
t
a
)


(
k
P
a
)
Temperature (C)
16% Crumb rubber
High Shear
Low Shear
Fig. 10. G

/Sin d versus temperature at 16% CRM binder.

0.1
1
10
100
46 51 56 61 66 71 76 81 86
G
*
/
S
i
n
(
d
e
l
t
a
)


(
k
P
a
)
Temperature (C)
Low shear blending
10% CR
14% CR
16% CR
Fig. 11. G

/Sin d versus temperature for low shear blended samples.

2100 S. Aaki, M. Memarzadeh/ Construction and Building Materials 25 (2011) 20942106
Efciency test
The mean value of sum of squares of deviations should be cal-
culated, as the rst step
MSX
SSX
n 1
MSY
SSY
m1
MSE
SSE
n 1m1
4:6
By then the efciency test should be followed by below
equations:
FX
MSX
MSE
FY
MSY
MSE
4:7
where the ratio, F(X), follows F distribution with (n 1, (n 1)
(m 1)) degree of freedom (dof). For a given signicant level, a, if
FX > Fa n 1; n 1m 1; factor X has a signicant effect
on test results, otherwise, has no signicant effect. This method
can easily apply to factor Y.
4.2. Hypothesis tests
In this paper the hypothesis test is used to compare the ef-
ciency of low and high shear blending conditions and determine
which one has the superior effect on improving the performance
0.5
5
55 60 65 70 75 80 85
G
*
/
S
i
n
(
d
e
l
t
a
)


(
k
P
a
)
Temperature (C)
High shear blending
10% CR
14% CR
16% CR
Fig. 12. G

/Sin d versus temperature for high shear blended samples.

0
1
2
3
4
5
6
7
8
0.09 0.1 0.11 0.12 0.13 0.14 0.15 0.16 0.17
L
T

i
m
p
r
o
v
e
m
e
n
t
Crumb rubber content
LT improvement
High Shear
Low Shear
Fig. 13. The LT improvement of the binder with increasing in crumb rubber content.
S. Aaki, M. Memarzadeh/ Construction and Building Materials 25 (2011) 20942106 2101
of asphalt binder in low, intermediate and high service
temperatures.
4.2.1. The mean hypothesis test method
Due to test conditions, the hypothesis test of two populations
with unknown variances and equal number of samples is used in
this paper. The steps of this method are [29]:
1. H
0
: l
1
l
2
0
2. H
1
: l
1
l
2
> 0 or H
1
: l
1
l
2
< 0
3. T
f ;n
1
n
2
2


X
1

X
2
sp

1
n
1

1
n
2
p
where H
0
is the null hypothesis, H
1
is the alternative hypothesis,
T is the test statistics, X
1
and X
2
are the mean value measure-
ments of two population samples, n
1
and n
2
are the numbers
of each sample that has the equal quantity, and s
2
p
is the com-
bined variance of the two samples which can be calculated as:
s
2
p

n
1
1s
2
1
n
2
1s
2
2
n
1
n
2
2
where s
2
1
and s
2
1
are the variances of two samples.
4. T has the t-distribution with n 1 degree of freedom, which a
is the signicant level.
5. The acceptable area is dened as:
T < t
1a
0
1
2
3
4
5
6
7
8
9
0.09 0.1 0.11 0.12 0.13 0.14 0.15 0.16 0.17
I
T

i
m
p
r
o
v
e
m
e
n
t
Crumb rubber content
IT improvement
High Shear
Low Shear
Fig. 14. The IT improvement of the binder with increasing in crumb rubber content.
0
5
10
15
20
25
0.09 0.1 0.11 0.12 0.13 0.14 0.15 0.16 0.17
H
T

i
m
p
r
o
v
e
m
e
n
t
Crumb rubber content
HT improvement
High Shear
Low Shear
Fig. 15. The HT improvement of the binder with increasing in crumb rubber content.
2102 S. Aaki, M. Memarzadeh/ Construction and Building Materials 25 (2011) 20942106
If the above equation is true, the null hypothesis can be
rejected.
5. Discussion
5.1. The result of ANOVA
The two-way ANOVA was used to evaluate and compare the
efciency of the crumb rubber content and the blending
conditions.
5.1.1. LT improvement
Table 2 shows the results of ANOVA used for evaluating the ef-
ciency of CR content and low/high shear blending conditions on
the PG of the binder at LT, low service temperature. Variable X rep-
resents the CR contents and Y represents the low/high shear blend-
ing conditions. As illustrated in Table 2, the ANOVA results that
neither the crumb rubber content nor the blending conditions have
exhibit a meaningful improvement on the performance of the
modied binders at low service temperature, when using a 0.95
signicant level; but with 0.8 signicant level, it can be concluded
that the blending conditions has more effect on the improvement
of the modied binders performance at low service temperature.
5.1.2. IT improvement
The results of ANOVA used for evaluating the efciency of
crumb CR content and low/high shear blending conditions on the
performance of the modied binders at IT, intermediate service
temperature are shown in Table 3. It can be inferred fromtable that
with 0.99 signicant levels, the blending conditions are more sig-
nicant than the CR content in improvement of modied binders
IT. In this case the CR content does not have meaningful effect on
the improvement of PG. With 0.95 of signicant level both of the
0
500
1000
1500
2000
2500
3000
3500
0.09 0.1 0.11 0.12 0.13 0.14 0.15 0.16 0.17
V
i
s
c
o
s
i
t
y

(
M
o
d
i
f
i
e
d
)

-
V
i
s
c
o
s
i
t
y

(
B
a
s
e
)
Crumb rubber content
Dynamic viscosity at 135 C
High Shear
Low Shear
Fig. 16. The changes of dynamic viscosity of the binder with increasing in crumb rubber content.
Table 2
ANOVA method for PG improvement in LT, low service temperature.
Crumb
rubber
content
Low
shear
(Y
1
)
High
shear
(Y
2
)
R
i R
2
i
G
i
X
1
10% 4.3 5.1 9.4 88.36 44.5
X
2
14% 3.8 7.3 11.1 123.21 67.73
X
3
16% 4.9 5.5 10.4 108.16 54.26
C
j
13 17.9 R = C 30.9
C
2
j
169 320.41
P
C
2
j
489.41 G
G
j
56.94 109.55
P
R
2
i
319.73 166.49
TSS 7.355
SSX 0.73
SSY 4.002
SSE 2.623
MSX 0.365
MSY 4.002
MSE 0.874
FX 0.417 F
0.95
19 F
0.8
4
FY 4.576 F
0.95
18.5 F
0.8
3.56
Table 3
ANOVA method for PG improvement in IT, intermediate service temperature.
Crumb
rubber
content
Low
shear
(Y
1
)
High
shear
(Y
2
)
R
i R
2
i
G
i
X
1
10% 4.7 2.7 7.4 54.76 29.38
X
2
14% 7.7 5.1 12.8 163.84 85.3
X
3
16% 6.9 4 10.9 118.81 63.61
C
j
19.3 11.8 R = C 31.1
C
2
j
372.49 139.24
P
C
2
j
511.73 G
G
j
128.99 49.3
P
R
2
i
337.41 178.29
TSS 17.088
SSX 7.503
SSY 9.375
SSE 0.21
MSX 3.752
MSY 9.375
MSE 0.07
FX 53.595 F
0.95
19 F
0.99
99
FY 133.929 F
0.95
18.5 F
0.99
98.5
S. Aaki, M. Memarzadeh/ Construction and Building Materials 25 (2011) 20942106 2103
CR content and the blending conditions have meaningful effect on
the IT improvement. It can be notied that in case of 0.95 blending
conditions is still has more signicant effects.
5.1.3. HT improvement
Table 4 shows the results of ANOVA used for evaluating the ef-
ciency CR content and low/high shear blending conditions on the
performance of the modied binders at HT, high service tempera-
ture. As demonstrated in Table 4, with 0.95 signicant levels, none
of the CR content and the blending conditions can improve the HT,
signicantly; but with the signicant level of 0.9, both of them
have similar signicant effects on improving the HT.
5.1.4. Dynamic viscosity
The results of ANOVA used for evaluating the efciency of the
CR content and blending conditions on the viscosity increase of
the modied binders at 135 C are shown in Table 5. It can be con-
cluded from the table that with 0.9 signicance levels, the CR con-
tent has the more effect on increasing the dynamic viscosity of the
modied binders than the blending conditions. In this case, alter-
nating the blending conditions makes no signicant changes on
the dynamic viscosity of the modied binders.
5.2. The result of hypothesis test
Tables 68 show the results of hypothesis test that have been
utilized to evaluate the improvement efciency of performance
Table 5
ANOVA method for viscosity improvement at 135 C.
Crumb rubber content Low shear (Y
1
) High shear (Y
2
) R
i R
2
i
G
i
X
1
10% 888.417 697 1585.417 2,513,546 1,275,093
X
2
14% 1586.027 1407 2993.028 8,958,215 4,495,133
X
3
16% 2906.361 1853 4759.361 22,651,518 11,880,544
C
j
5380.806 3957 R = C 9337.806
C
2
j
28,953,068 15,657,849
P
C
2
j
44,610,917 G
G
j
11,751,703 5,899,067
P
R
2
i
34,123,279 17,650,770
TSS 3,118,335
SSX 2,529,204
SSY 337,870.4
SSE 251,260.1
MSX 1,264,602
MSY 337,870.4
MSE 83,753.36
FX 15.099 F
0.9
9
FY 4.034 F
0.9
8.53
Table 4
ANOVA method for PG improvement in HT, high service temperature.
Crumb
rubber
content
Low
shear
(Y
1
)
High
shear
(Y
2
)
R
i R
2
i
G
i
X
1
10% 13.6 10.3 23.9 571.21 291.05
X
2
14% 16.9 15.6 32.5 1056.25 528.97
X
3
16% 20.3 14.8 35.1 1232.01 631.13
C
j
50.8 40.7 R=C 91.5
C
2
j
2580.64 1656.49
P
C
2
j
4237.13 G
G
j
882.66 568.49
P
R
2
i
2859.47 1451.15
TSS 55.775
SSX 34.36
SSY 17.002
SSE 4.413
MSX 17.18
MSY 17.002
MSE 1.471
FX 11.678 F
0.95
19 F
0.9
9
FY 11.557 F
0.95
18.5 F
0.9
8.53
Table 6
The mean hypothesis test results of LT improvement.
Crumb rubber
content
Low shear High shear
10% 4.3 5.1
14% 3.8 7.3
16% 4.9 5.5
Mean 4.33 5.96
Variance 0.30 1.37
s
2
p
T t
0.99,4
t
0.95,4
0.84 2.18 3.75 2.13
Table 7
The mean hypothesis test results of IT improvement.
Crumb rubber
content
Low shear High shear
10% 4.7 2.7
14% 7.7 5.1
16% 6.9 4
Mean 6.43 3.93
Variance 2.41 1.44
s
2
p
T t
0.99,4
t
0.95,4
1.93 2.20 3.75 2.13
Table 8
The mean hypothesis test results of HT improvement.
Crumb rubber
content
Low shear High shear
10% 13.6 10.3
14% 16.9 15.6
16% 20.3 14.8
Mean 16.93 13.57
Variance 11.22 8.16
s
2
p
T t
0.99,4
t
0.95,4
9.69 1.32 3.75 2.13
2104 S. Aaki, M. Memarzadeh/ Construction and Building Materials 25 (2011) 20942106
grade (PG) regarding different blending conditions. In these
hypothesis tests, the null hypothesis, H
0
, is l
1
= l
2
, where l
1
represents the mean value of the PG improvement for low shear
blending population and l
2
represents the mean value of the PG
improvement for high shear blending population and the alterna-
tive hypothesis, H
1
, is l
1
< l
2
or l
1
> l
2
. As calculated in Table 6,
the region for rejection of test with 0.99 signicant levels is
(3.747, 1) and the value of T in this test is not in the region
for rejection then the null hypothesis can be accepted, which
means the l
1
= l
2
, the low shear and high shear blending modi-
ed samples have the same LT properties. The region for rejection
with 0.95 signicant levels is (2.132, 1) and the value of T in this
case is in the region for rejection then the null hypothesis can be
rejected, l
1
< l
2
, in this case the high shear blended samples
cause the higher improvement in LT, which is also shown in
Fig. 13).
Table 7 shows the results of hypothesis test that is utilized for
analyzing the IT improvement of the binder. It is clear from the
Table 7 that with 0.99 signicant levels, the T value is not in the
region for rejection then the null hypothesis can be accepted.
The null hypothesis can be rejected with 0.95 signicant levels,
l
1
< l
2
, which means that the low shear blended samples are more
signicant in improving the IT of the binder that is shown in
Fig. 14.
The results of hypothesis test that is utilized for analyzing the
HT improvement of the binder are shown in Table 8. It can be inter-
preted from the table that the null hypotheses are accepted with
either 0.99 or 0.95 signicant levels, which means that the low
shear blended samples and high shear blended samples has the
same effect on the HT improvement. This result is also shown in
Fig. 15 with a little diversity between the graphs of the low and
high shear blended samples. In this case the alternative hypothesis
was l
1
< l
2
.
6. Conclusion
In this study, two statistical methods were used to evaluate the
performance grade of the asphalt binder with various CR contents
with different blending conditions. The qualitative and quantita-
tive evaluation of test results is:
The evaluations of test results show that high shear blending
has more effect on improvement of LT than the lowshear blend-
ing while low shear blending has more effects on the IT and HT
improvements than high shear blending.
Modied binders made with low shear blending have better
performance in the intermediate service temperatures and are
more resistant to fatigue cracking than those modied by the
high shear blending.
The results of ANOVA show altering the blending condition has
more effect on the LT, IT and HT of the modied binders than
changing the CR contents. This means that the blending condi-
tions has superior effect on the performance of modied bind-
ers that CR contents in LT, IT and HT.
ANOVA shows that the dynamic viscosity of the modied bind-
ers is less sensitive to blending conditions than to the CR con-
tents, which means alternating the blending conditions makes
no signicant changes on the dynamic viscosity of the modied
binders.
The results of hypothesis test shows that the high shear blend-
ing is more signicant on the LT improvement of the modied
binders than low shear blending, while in the case of IT and
HT improvement of modied binders, the low shear blending
condition has a greater effect than the high shear blending
condition.
Acknowledgments
The authors would like to express their sincere appreciation to
Professor Nader Tabatabaee supervisor of the SuperPAVE labora-
tory of Civil Engineering Department, Sharif University of Technol-
ogy, and also wish to acknowledge Eng. Zia Alavi and Eng. Zahra
Kamali for their valuable efforts on conducting the laboratory tests.
References
[1] Aaki Sassan, Tabatabaee Nader. Proposals for modication of Iranian bitumen
to meet the climatic requirements of Iran. J Construct Build Mater
2009;23:214150.
[2] Potgieter CJ. Bitumen rubber asphalt, design and construction procedures in
South Africa. In: Proceedings of the 8th conference on asphalt pavements for
Southern Africa (CAPSA04), Sun City, South Africa; 2004.
[3] Balige M. Crumb rubber bituminous mixes from waste tires in Argentina.
Asphalt Mag 2004:245.
[4] Sebaaly PE, Epps J. Effect of crumb rubber particle size and content on the low
temperature rheological properties of binders. In: Proceedings of the
transportation research board annual meeting, Washington DC; 2002.
[5] Navarro FJ, Partal P, Martinez-Boza F, Gallegos C. Thermo-rheological
behaviour and storage stability of ground tire rubber-modied bitumens. J
Fuel 2004;83(1415):20419.
[6] Shen Junan, Amirkhanian Serji, Xiao Feipeng, Tang Boming. Inuence of surface
area and size of crumb rubber on high temperature properties of crumb rubber
modied binders. J Sci Direct Construct Build Mater 2009;23:30410.
[7] Lee Soon-Jae, Akisetty Chandra K, Amirkhanian Serji N. The effect of crumb
rubber modier (CRM) on the performance properties of rubberized binders in
HMA pavements. J Sci Direct Construct Build Mater 2008;22:136876.
[8] Thodesen Carl, Shatanawi Khaldoun, Amirkhanian Serji. Effect of crumb rubber
characteristic on crumb rubber modied (CRM) binder viscosity. J Construct
Build Mater 2009;23:295303.
[9] Abdelrahman Magdy A, Carpenter Samuel H. Mechanism of interaction of
asphalt cement with crumb rubber modier. TRR 1999;1661:10613.
[10] Bahia Hussain U, Davies Robert. Effect of crumb rubber modiers (CRM) on
performance-related properties of asphalt binders. AAPT 1994;63:41438.
[11] Buncher Mark S, Brown ER. Use of crumb rubber modied in hot asphalt. In:
Proceeding of the 23rd air transport conference; 1994. p. 5363.
[12] Shen J, Amirkhanian S. The inuence of crumb rubber modier (CRM)
microstructure on the high temperature properties of CRM binder. Int J
Pavement Eng 2005;6(4):26571.
[13] Kim S, Loh SW, Zhai H, Bahia H. Advanced characterization of crumb rubber-
modied asphalts, using protocols developed for complex binder. Transport
Res Rec: J Transport Res Broad 2001;1767:1524.
[14] Zanzotto L, Kennepohl G. Development of rubber and asphalt binders by
depolymerization and devulcanization of scrap tires in asphalt. Transport Res
Rec: J Transport Res Broad 1996;1530:519.
[15] McGennis RB. Evaluation of physical properties of ne crumb rubber modied
asphalt binders. Transport Res Rec: J Transport Res Broad 1995;1488:6371.
[16] Thodesen Carl, Shatanawi Khaldoun, Amirkhanian Serji, Bridges William.
Development of an empirical model for determining G

/Sin d in crumb rubber

modied binders. J Construct Build Mater 2009;23:19227.
[17] Liu Shutang, Cao Weidong, Fang Jianguo, Shang Shujie. Variance analysis and
performance evaluation of crumb rubber modied (CRM) asphalt. J Sci Direct
Construct Build Mater 2009;23:27018.
[18] ASTM-D6373. Standard specication for performance graded asphalt binder.
American Society for Testing and Materials; 1999.
[19] ASTM-D2872. Standard test method for effect of heat and air on a moving lm
of asphalt (rolling thin-lm oven test). American Society for Testing and
Materials; 1997.
[20] ASTM-D6521. Standard practice for accelerated aging of asphalt binder using a
pressurized aging vessel (PAV). American Society for Testing and Materials;
2003.
[21] ASTM-D6648. Standard test method for determining the exural creep
stiffness of asphalt binder using the bending beam rheometer (BBR).
American Society for Testing and Materials; 2001.
[22] ASTM D7175. Standard test method for determining the rheological properties
of asphalt binder using a dynamic shear rheometer (DSR). American Society for
Testing and Materials; 2005.
[23] ASTM-D4402. Standard test method for viscosity determination of asphalt at
elevated temperatures using a rotational viscometer. American Society for
Testing and Materials; 2002.
[24] Wu SP, Pang L, Mo LT, Chen YC, Zhu GJ. Inuence of aging on the evolution of
structure, morphology and rheology of base and SBS modied bitumen.
Construct Build Mater 2008;23(2):100510.
[25] Cortizo MS, Larsen DO, Bianchetto H, Alessandrini JL. Effect of the thermal
degradation of SBS copolymers during the aging of modied asphalts. Polym
Degrad Stabil 2004;86(2):27582.
[26] Ruan Y, Davison RR, Glover CJ. The effect of long-term oxidation on the
rheological properties of polymer modied asphalts. Fuel 2003;82(14):
176373.
S. Aaki, M. Memarzadeh/ Construction and Building Materials 25 (2011) 20942106 2105
[27] Ouyang C, Wang S, Zhang Y, Zhang Y. Improving the aging resistance of
styrenebutadienestyrene tri-block copolymer modied asphalt by addition
of antioxidants. Polym Degrad Stabil 2006;91(4):795804.
[28] Kleinbaum David G, Kupper Lawrence L. Applied regression analysis and other
multivariate methods. Duxbury Press, A Division of Wadsworth Publishing
Company, Inc.; 1978.
[29] Ang Alfredo H-S, Tang Wilson H. Probability concepts in engineering. 2nd
ed. John Wiley & Sons Inc.; 2007.
Sassan Aaki received his BS in Civil Engineering in 1998 from Sharif University
of Technology in Tehran, Iran and continued his studies in highway engineering.
He received his MS degree from SUT in 2000 with a thesis on comprehensive
pavement management system development and its implementation in a branch
of the Tehran freeway network. He was studying rheological asphalt binder
properties and worked on Iranian asphalt binder modication during his PhD
studies, including work on projects on pavement evaluation and pavement
rehabilitation design. He played an instrumental role in setting up the Superpave

asphalt laboratory of the Department of Civil Engineering at SUT. He received his

PhD degree from SUT in 2009. And till now he is working in School of Civil
engineering, College of Engineering at the University of Tehran as Assistant
Professor.
2106 S. Aaki, M. Memarzadeh/ Construction and Building Materials 25 (2011) 20942106