Beruflich Dokumente
Kultur Dokumente
Composite Structures
journal homepage: www.elsevier.com/locate/compstruct
Department of Civil and Environmental Engineering, KAIST, 291 Daehak-ro, Yuseong-gu, Daejeon 305-338, South Korea
Department of Nuclear Safety Research, Korea Institute of Nuclear Safety (KINS), 62 Gwahak-ro, Yuseong-gu, Daejeon 305-338, South Korea
a r t i c l e
i n f o
Article history:
Available online 28 August 2015
Keywords:
Strengthening/Retrofitting
Fiber-reinforced polymers (FPRs)
Bending beam test
Interfacial fracture energy (GF,int)
a b s t r a c t
This paper presents the experimental investigation of composite-mortar three-point bending beam test
setup, used to characterize the bond behavior between concrete/mortar and fiber-reinforced polymers
(FRPs). With this aim, a series of experimental studies have been conducted by considering different
FRP fabric types (carbon/glass and carbon/aramid), epoxy adhesives (E-1 and E-2), and notch depths
(50 and 33 mm). In addition, a fiber-reinforced mortar, with different fiber contents (0, 0.5, 1 and
1.5 wt.%) was also used to investigate the effects of short fibers on the interface behavior. From the
loaddisplacement curves in three-point bending beam tests, peak load (Ppeak), ultimate mid-span deflection (dult), and interfacial fracture energy (GF,int) of different bonded interfaces were evaluated. It is concluded from this study that this test setup is useful for the comparison of different bonded interfaces as
true interfacial failure was observed, but the interfacial fracture energy (GF,int) obtained from these tests
showed sensitivity to the notch depth. The incorporation of short steel fibers into mortar was not effective to improve the interfacial bond strength as not much fiber action was observed (near the bond line)
during testing.
2015 Elsevier Ltd. All rights reserved.
1. Introduction
FRPs are commonly applied on the external face of structural
elements with the use of an epoxy adhesive, and their strengthening performance have been reported by a number of researchers
[111]. The adhesive interface between FRPs and concrete plays
a key role in the structural performance of strengthened concrete
structures [12]. Among many possible failure modes (i.e., FRP rupture, concrete crushing, shear failure, interfacial plate-end debonding, and intermediate crack-induced interfacial debonding),
interfacial plate-end debonding is one of the most commonly
observed failure mode of FRP-strengthened concrete structures
[1315]. Therefore, an evaluation and accurate prediction of the
interfacial behavior of bonded interfaces has gained much attention by the research community [1620].
In order to understand the bond behavior of FRP-strengthened
RC structures, many experimental test methods (i.e., single or double shear-lap tests and three-point bending beam tests) and analytical bond strength models have been developed [17,2123]. A
number of researchers have adopted a lap-shear test (single or
double lap-shear tests) for developing bond strength models and
Corresponding author. Tel.: +82 42 350 3623; fax: +82 42 350 3610.
E-mail address: leeh@kaist.ac.kr (H.K. Lee).
http://dx.doi.org/10.1016/j.compstruct.2015.08.070
0263-8223/ 2015 Elsevier Ltd. All rights reserved.
354
sand was used as a main aggregate, while 30% of that (by weight)
was replaced with bottom ash to reduce the self-weight of mortar
and enhance sustainability. Four cases with different fractions of
steel fibers were studied (0, 0.5, 1.0 and 1.5 wt.%) to investigate
the effect of the steel fiber content on the interfacial bond
behavior. The length, diameter and tensile strength of the steel
fibers, used in this study, were 13 mm, 2 lm, and in the range of
27004000 MPa, respectively.
Three types of tests were performed to measure the basic material properties of plain and fiber-reinforced specimens (e.g., compressive strength, elastic modulus, Poissons ratio, tensile strength,
and fracture energy). For compression (ASTM C39 [32]; ASTM
C469 [33]) and splitting tensile tests (ASTM C 496 [34]), four cylindrical specimens, sized 100 200 mm (diameter height) were
tested for each series. In the designation of series, SF-F0, SF-F0.5,
SF-F1, and SF-F1.5 represents specimens with 0, 0.5, 1, and 1.5 wt.
% of steel fibers, respectively. The fracture energy (GF) of mortar
specimens was measured by the work-of-fracture method, proposed by Hillerborg [35] where GF is the energy required to create,
propagate and fully break a unit area of crack in the concrete [36].
Three-point bending beam tests were conducted on notched beam
specimens to determine GF. The dimensions of the specimens were
chosen according to the recommendations of RILEM Committee 50
[37] and RILEM committee 162 [38] for plain and fiber-reinforced
mortar specimens, respectively. The geometry and test setup of
the three-point notched beam tests is shown in Fig. 1. Details of
the calculation of the fracture energy can be found in RILEM recommendations [37] and Petersson [39].
0
All the specimens exhibited compressive strength (f c ) of above
50 MPa. While steel fiber-reinforced specimens (SF-F0.5, SF-F1, and
SF-F1.5) showed marginally higher compressive strength, but the
elastic modulus was relatively lower in the comparison with that
of the control specimens. The Poissons ratio of all the series was
found to be around 0.2, typical value for concrete and mortar.
During the test, the failure of control specimens was relatively brittle compared with the fiber-reinforced specimens due to a fiber
bridging effect. Averaged results of compression tests are detailed
in Table 1.
Table 1
Compressive strength, elastic modulus, and Poissons ratio of plain and fiberreinforced geopolymer mortar.
2. Material properties
2.1. Geopolymer mortar
Two types of binders, coal fly ash (FA) and granulated blast furnace slag (BFS) were used. Both the binders were mixed with a
weight ratio of 1:1 to prepare the geopolymer binder. Two types
of fine aggregates were used (river sand and bottom ash). River
Specimen
Compressive
strength (MPa)
Elastic modulus
(GPa)
Poissons ratio
SF-F0 (Control)
SF-F0.5
SF-F1.0
SF-F1.5
52.99
54.80
56.87
56.75
21.40
20.34
20.06
20.00
0.211
0.223
0.24
0.2
355
The splitting tensile strength (ft) and fracture energy (GF) of the
geopolymer mortar specimens were significantly enhanced by
fiber reinforcement. In splitting tensile tests, the crack widths of
fiber-reinforced specimens were relatively narrower compared
with the control specimens, and the specimens remained intact
due to the fiber bridging effect as shown in Fig. 2. Fiberreinforced specimens SF-F0.5, SF-F1, and SF-F1.5 showed an
increase in the tensile strength by 12%, 13%, and 26% that of the
control series (SF-F0), respectively. The splitting tensile test results
are summarized in Table 2. In notched beam tests, fiber-reinforced
specimens showed a longer post peak behavior attributed to fiberbridging and fiber pull-out phases, while plain mortar specimens
showed a sudden drop in the loading value after reaching the peak
load. Typical loaddisplacement curves of plain and fiberreinforced geopolymer mortar specimens are shown in Fig. 3. For
plain specimens, the load dropped to zero at about 1.3 mm
mid-span deflection while fiber-reinforced specimens exhibited
an ultimate mid-span deflection of about 79 mm. Only 0.5 wt.%
Tensile strength
(MPa)
Increase
ratio (%)
Fracture energy,
GF (N/m)
Increase
ratio (%)
SF-F0 (Control)
SF-F0.5
SF-F1
SF-F1.5
2.22
2.49
2.512
2.783
12.3
13.26
25.46
70.8
237.55
417.66
508.75
236
490
619
Table 3
Tensile test results of epoxy adhesives.
Properties
E-1
E-2
36.1
1.7
3.4
14
6
25.6
2.57
1.98
14
11
2500
Load (N)
2000
1500
Control
Fiber-reinforced
specimen
1000
500
0
0
Deflection (mm)
Fig. 3. Typical loaddeflection curves of plain and fiber-reinforced specimens.
356
manufactured by MSC Korea Co., Ltd [41]. During fabrication, carbon fibers were introduced along the principal axis with other
fibers (glass or aramid) being in the lateral direction (at 90). FRP
fabric specimens were tested in accordance with ASTM D 3039
[42]. Rectangular specimens (250 mm length 25 mm width)
were prepared following the recommended dimensions by ASTM
D 3039 [42]. In the designations of specimens, C, G, A, and E represents carbon fibers, glass fibers, aramid fibers, and epoxy adhesive,
respectively. The following number (1 or 2) represents type of
epoxy adhesive (E-1 or E-2). An automatic coating device was used
to accurately control the thickness of the adhesive, and adhesive
was applied from both the sides of FRP fabric. On average, a uniform thickness of 0.33 mm was achieved for all the composite
sheets while thickness of fabric itself was around 0.27 mm. After
curing for 7 days at an ambient temperature, the composite sheets
were cut into rectangular strips with a width of 25 mm and length
of 250 mm. Universal testing machine (INSTRON 5583), with a load
cell capacity of 50 kN, was used for tensile tests. Tests were
conducted under displacement control mode at a crosshead speed
of 2 mm/min as recommended in the ASTM standard [42]. An
extensometer, with a gage length of 50 mm, was used to record
the longitudinal displacement in gage length section. Specimens
were tested for each fabric type with both the epoxies (i.e., E-1
and E-2) until 5 specimens with failure in the gauge length section
were obtained.
It was observed that the tensile strength of C/G fabric composites was relatively more dependent on the epoxy type in comparison with that of C/A fabric composites. The average values of
tensile strength, ultimate strain capacity, and elastic modulus of
composite specimens with different epoxy adhesives are presented
in Table 4. Specimen C/G-E2 showed tensile strength of around 15%
higher than that of C/G-E1, while the tensile strength of specimen
C/A-E2 and C/A-E1 was almost similar (only 4% difference). Since
the elastic modulus of adhesive E-2 was higher than that of adhesive E-1 (Table 3), that of respective composite specimen C/G-E2
was similarly higher than that of C/G-E1. The elastic modulus of
specimen C/G-E2 was 28% higher than the corresponding specimen
C/G-E1 while specimen C/A-E2 showed only 5% higher elastic modulus than that of the specimen C/A-E1. The ultimate strain capacity
of specimens C/G-E1, C/G-E2, C/A-E1, and C/A-E2 were 1.35%,
Table 4
Tensile test results of hybrid FRP composites.
Specimen
Tensile
strength
(MPa)
Ultimate
strain (%)
Elastic
modulus
(GPa)
Curing
(days)
No. of
specimens
C/G-E1
C/G-E2
C/A-E1
C/A-E2
302.4
346.86
394.76
379.82
1.35
1.23
1.31
1.34
23.80
30.40
24.78
26.11
14
14
14
14
8
6
6
7
357
Specimen
Fabric type
Epoxy type
No. of specimens
1
2
3
4
5
6
7
8
9
10
C/G-E1-F0-50
C/A-E1-F0-50
C/G-E1-F0.5-50
C/A-E1-F0.5-50
C/G-E2-F0.5-0
C/A-E2-F0.5-50
C/G-E1-F1-50
C/A-E1-F1-50
C/G-E1-F1.5-50
C/A-E1-F1.5-50
Carbon/Glass
Carbon/Aramid
Carbon/Glass
Carbon/Aramid
Carbon/Glass
Carbon/Aramid
Carbon/Glass
Carbon/Aramid
Carbon/Glass
Carbon/Aramid
1
1
1
1
2
2
1
1
1
1
0
0
0.5
0.5
0.5
0.5
1
1
1.5
1.5
50
2
2
2
2
2
2
2
2
2
2
11
12
13
14
15
16
17
18
19
20
C/G-E1-F0-33
C/A-E1-F0-33
C/G-E1-F0.5-33
C/A-E1-F0.5-33
C/G-E2-F0.5-33
C/A-E2-F0.5-33
C/G-E1-F1-33
C/A-E1-F1-33
C/G-E1-F1.5-33
C/A-E1-F1.5-33
Carbon/Glass
Carbon/Aramid
Carbon/Glass
Carbon/Aramid
Carbon/Glass
Carbon/Aramid
Carbon/Glass
Carbon/Aramid
Carbon/Glass
Carbon/Aramid
1
1
1
1
2
2
1
1
1
1
0
0
0.5
0.5
0.5
0.5
1
1
1.5
1.5
33
2
2
2
2
2
2
2
2
2
2
two different depths (50 mm and 33 mm) were casted for two different series. Following the surface preparation, FRP fabric was
applied on the beam without scotch tape. After application of
FRP fabric, a thin layer of epoxy adhesive was applied on the top
of FRP and the second beam was kept on the top of the first beam.
Polystyrene boards and steel clamps were used to hold the beams
straight and aligned during the curing period.
For bending beam test specimens, the same test setup was
employed as described in Section 2 for the fracture energy tests
with exception of different loading span and beam length. Fig. 4
schematically illustrates the test set-up used for the three-point
loading bending beam test. A 5 kN load cell was used for these tests
with a moving head speed of 0.2 mm/min as recommended by
RILEM committee 162 [38]. Mid-point displacement (d) was
358
Table 6
Summary of the averaged bending beam test results.
Exp no.
Specimen
Ppeak (N)
dult (mm)
GF,int (N/m)
1
2
3
4
5
C/G-E1-F0-50
C/G-E1-F0.5-50
C/G-E2-F0.5-50
C/G-E1-F1-50
C/G-E1-F1.5-50
1353.8
1250.78
1142.89
1343.97
1402.81
1.18
0.98
0.87
0.89
0.75
106.7
102.3
97.34
99.36
94.04
6
7
8
9
10
C/A-E1-F0-50
C/A-E1-F0.5-50
C/A-E2-F0.5-50
C/A-E1-F1-50
C/A-E1-F1.5-50
1383.22
1231.15
1290
1206.63
0.92
0.78
0.76
0.75
97.87
85.9
86.30
85.92
11
12
13
14
15
C/G-E1-F0-33
C/G-E1-F0.5-33
C/G-E2-F0.5-33
C/G-E1-F1-33
C/G-E1-F1.5-33
1447
1319.45
1128.2
1196.8
1422.46
1.135
1.16
1.048
1.2
0.89
94.25
98.50
61.24
74.58
62.72
16
17
18
19
20
C/A-E1-F0-33
C/A-E1-F0.5-33
C/A-E2-F0.5-33
C/A-E1-F1-33
C/A-E1-F1.5-33
1152.69
1290.03
1314.5
1417.53
1265.49
1.05
1.19
1.045
1.006
0.83
68.31
76.05
64.70
70.28
57.65
(c) 1% fibers
(e) 50 mm notch
(f) 33 mm notch
359
360
361
362
363
5. Statistical analysis
5.1. Analysis of variance (ANOVA)
In this part of the study, an analysis of variance (ANOVA) of
the data was performed for the response variables, Ppeak, dult and
GF,int, and the influence of different experimental parameters
(i.e., FRP fabric type, epoxy adhesive, steel fiber content, and
notch depth) on the results was investigated. The analysis was
carried out for a confidence level of 95% using Minitab software
[48]. The ANOVA (general linear model) results for Ppeak, dult,
and GF,int are presented in Tables 79, respectively. The last
columns of the tables represent the percentage contributions (P)
of each factor to the total variation i.e., the degree of influence
on the response variables.
According to the ANOVA for Ppeak (Table 7), the steel fiber content has a statistical and physical significance on Ppeak with a percentage contribution of 21.76%. The effect of the other remaining
parameters seemed insignificant as the percentage contributions
were relatively low. From the analysis in Table 7, it is apparent that
the sum of the percentage contributions of all the parameters was
less than that of the associated error of 74%. These results are compatible with the experimental findings for Ppeak. In the experimental results, Ppeak did not show a particular tendency against
individual experimental parameters. That is, for different FRP
fabrics, specimens with 50 mm notch depth showed similar Ppeak
while specimens with 33 mm notch depth exhibited higher Ppeak
for C/G fabric specimens; similarly C/G fabric specimens showed
difference in Ppeak for different epoxy adhesives but values of Ppeak
for C/A fabric specimens were similar for different adhesive types
(see Sections 4.2 and 4.3).
Table 8 shows the ANOVA results for dult against different
parameters. From the analysis, it is evident that the notch depth
and fiber content were the most significant parameters with a percentage contribution of 34.04% and 34.91%, respectively. The percentage contribution was greater than that of the associated
error which was about 17% for dult. On the other hand, epoxy type
(P = 4.16%) has the least statistical and physical significance on the
response variable dult, followed by the fabric type having a percentage contribution of 9.45%.
The ANOVA results for the response variable GF,int are shown in
Table 9. The analysis revealed that the notch depth had the highest
influence on the interfacial fracture energy of FRP-to-mortar
bonded joints with a percentage contribution of 52.11%. The next
significant parameter was the fiber content (P = 13.45%) followed
by the fabric type with the percentage contribution of 12.47%
whereas epoxy type was the least significant parameter with a
percentage contribution of 7.85%. It is worth noticing that the
error associated with the ANOVA analysis for the GF,int was
approximately 14%. These findings further justify the dependence
of GF,int on the initial depth of notch in three-point bending beam
tests as notch depth was found to be the most significant
parameter.
5.2. Signal-to-noise (S/N) ratio
The concept of signal-to-noise (S/N) ratio, well known in the
communication industry, was first introduced by Taguchi in the
field of quality engineering for the evaluation of measurement systems as well as for the function of products and process [49]. Based
on the response characteristics of interest, several S/N ratios are
available i.e., larger/higher is better (HB), nominal is better (NB),
or lower is better (LB) [49].
For response variables Ppeak, dult, and GF,int, the HB was
employed since a higher value is desired for these response
variables. For HB, loss function Lij can be expressed by Eq. (1)
364
365
F-ratio
Percentage contribution
Fabric type
Epoxy type
Fiber content
Notch depth
Error
Total
1
1
3
1
12
18
1484
5822
38,324
167
130,313
176,111
1484.4
5822.3
12774.6
167.3
10859.4
0.14
0.54
1.18
0.02
0.84
3.31
21.76
0.1
73.99
100
Table 8
ANOVA for dult.
Source
F-ratio
Percentage contribution
Fabric type
Epoxy type
Fiber content
Notch depth
Error
Total
1
1
3
1
12
18
0.04254
0.01872
0.15713
0.15321
0.07853
0.45013
0.042544
0.018721
0.052376
0.153209
0.006544
6.5
2.86
8
23.41
9.45
4.16
34.91
34.04
17.45
100
Source
F-ratio
Percentage contribution
Fabric type
Epoxy type
Fiber content
Notch depth
Error
Total
1
1
3
1
12
18
561.3
353.4
607
2345.7
634.2
4501.6
561.31
353.38
202.34
2345.69
52.85
10.62
6.69
3.83
44.38
12.47
7.85
13.45
52.11
14.09
100
Table 9
ANOVA for GF,int.
gij 10logLij
Specimen
Ppeak (N)
dult (mm)
GF,int (N/m)
1
2
3
4
5
C/G-E1-F0-50
C/G-E1-F0.5-50
C/G-E2-F0.5-50
C/G-E1-F1-50
C/G-E1-F1.5-50
62.63
61.94
61.16
62.57
62.94
1.44
-0.18
-1.21
-1.01
-2.50
40.56
40.20
39.77
39.94
39.47
6
7
8
9
C/A-E1-F0-50
C/A-E1-F0.5-50
C/A-E2-F0.5-50
C/A-E1-F1-50
62.82
61.81
62.21
61.63
-0.72
-2.16
-2.38
-2.50
39.81
38.68
38.72
38.68
11
12
13
14
15
C/G-E1-F0-33
C/G-E1-F0.5-33
C/G-E2-F0.5-33
C/G-E1-F1-33
C/G-E1-F1.5-33
63.21
62.41
61.05
61.56
63.06
1.10
1.29
0.41
1.58
-1.01
39.49
39.87
35.74
37.45
35.95
16
17
18
19
20
C/A-E1-F0-33
C/A-E1-F0.5-33
C/A-E2-F0.5-33
C/A-E1-F1-33
C/A-E1-F1.5-33
61.23
62.21
62.38
63.03
62.05
0.42
1.51
0.38
0.05
-1.62
36.69
37.62
36.22
36.94
35.22
n
1X
1
n k1 y2ij
Lij
Figs. 1113. From the main effect plots for Ppeak (Fig. 11), it is evident that fiber content is the most influential parameter attributed
to its sharp graphical slope, followed by epoxy adhesive. A much
lower slope of FRP fabric and notch depth shows that the effect
of these parameters is not significant.
62.8
62.6
62.4
62.2
62
61.8
61.6
C/G C/A
E-1 E-2
N33 N50
[4950], while the S/N ratio gij for the ith performance characteristic in the jth experiment is given by Eq. (2) [49].
1
0.5
0
-0.5
-1
-1.5
-2
C/G C/A
E-1 E-2
N33 N50
366
40
39.5
For prediction of the response variables, mathematical relationships between the experimental parameters and performance
measures were established. A multiple linear regression analysis
was conducted for this purpose using Minitab software [48]. The
generic multiple linear regression model is given by Eq. (3)
39
38.5
38
37.5
Y bo b1 X 1 b2 X 2 bN X N
37
36.5
C/G C/A
E-1 E-2
N33 N50
The main effect plots of S/N ratio for dult are shown in Fig. 12. It
can be inferred that the notch depth and fiber content had almost
an identical level of influence on the dult as the graphical slopes
were approximately same (consistent with the ANOVA results in
Section 5.1), followed by the FRP fabric and epoxy type. dult
decreased with the increase in fiber content and notch depth.
The maximum dult can be obtained for the FRP fabric C/G with
epoxy adhesive E-1, applied to the mortar specimens with
50 mm notch depth and no steel fibers.
Similarly, Fig. 13 shows that the notch depth had the steepest
slope followed by the steel fiber content, FRP fabric type, and epoxy
adhesive type. The GF,int can be maximized by keeping the FRP
fabric, epoxy adhesive, and fiber content at level 1 with notch
depth at level 2.
where FA denotes the FRP fabric type; E is the epoxy type; FI is the
fiber content; and N represents the notch depth. The models were
verified by analysis of variance (ANOVA) and experimental data.
The ANOVA results of Ppeak, dult, and GF,int for the linear regression
models are given in Table 11. The significance (P) value for Ppeak
was 0.601 which is higher than that of a-value (0.05), while that
for dult and GF,int was relatively low (0.00). This indicates that the
Table 11
ANOVA for Ppeak, dult, and GF,int for multiple linear regression models.
Response variable
Source
Ppeak
Regression
Residual error
Total
4
14
18
29565.56
146545.1
176110.6
7391.4
10467.51
dult
Regression
Residual error
Total
4
14
18
0.35
0.1
0.45
GF,int
Regression
Residual error
Total
4
14
18
0.706
0.088
0.007
3792.48
709.116
4501.6
F-ratio
948.12
50.65
Significance (P)
0.601
12.22
0.000
18.72
0.000
Table 12
Experimental and predicted results for Ppeak, dult, and GF,int.
No.
Ppeak (N)
Exp.
Pred.
1
2
3
4
5
1353.8
1250.78
1142.89
1343.97
1402.81
1307.16
1314.09
1224.09
1321.02
1327.95
6
7
8
9
1383.22
1231.15
1290
1206.63
11
12
13
14
15
16
17
18
19
20
dult (mm)
Exp.
Pred.
3.45
4.82
6.63
1.71
5.34
1.18
0.98
0.87
0.89
0.75
1.05
0.95
0.88
0.86
0.77
1292.85
1299.78
1209.78
1306.71
6.53
5.28
6.22
7.66
0.92
0.78
0.76
0.75
1447
1319.45
1128.2
1196.8
1422.46
1310.73
1317.66
1227.66
1324.59
1331.52
9.42
0.14
8.10
9.65
6.39
1152.69
1290.03
1314.5
1417.53
1265.49
1296.42
1303.35
1213.35
1310.28
1317.21
11.09
1.02
7.69
7.57
3.93
GF,int (N/m)
Exp.
Pred.
11.44
2.86
0.57
3.48
2.09
106.7
102.3
97.34
99.36
94.04
110.46
104.16
93.78
97.86
91.56
3.40
1.78
3.66
1.51
2.64
0.93
0.84
0.76
0.75
1.50
7.25
0.52
0.27
97.87
85.9
86.30
85.92
98.61
92.31
81.93
86.01
0.75
6.95
5.06
0.11
1.135
1.16
1.048
1.2
0.89
1.25
1.16
1.08
1.06
0.97
9.13
0.34
2.87
11.42
8.25
94.25
98.50
61.24
74.58
62.72
89.72
83.42
73.04
77.12
70.82
4.81
15.31
16.15
3.29
11.43
1.05
1.19
1.045
1.006
0.83
1.14
1.05
0.97
0.95
0.86
7.73
12.18
7.37
5.37
3.38
68.31
76.05
64.70
70.28
57.65
77.87
71.57
61.19
65.27
58.97
12.28
5.89
5.42
7.13
2.24
models for dult and GF,int are highly significant statistically at 95%
confidence level but not for Ppeak. Since no parameter was found
to be significant as evident by ANOVA for Ppeak (see Section 5.1),
the value of P being high seems logical for the regression model.
From experimental data presented in Table 12, the multiple linear regression models predicted Ppeak, dult, and GF,int with an average
error of 5.93%, 5.16%, and 5.78%, respectively. The model predictions showed a high degree of congruity with the experimental
results.
6. Conclusions
This paper has presented an experimental study on the interfacial bond behavior of FRP-to-mortar bonded joints. A detailed parametric study was conducted to investigate the validation of threepoint bending beam test setup for the prediction of true interfacial
bond behavior. Effects of different FRP fabrics (C/G and C/A), epoxy
adhesives (E-1 and E-2), and different notch depths (50 and
33 mm) were studied during these experiments. In addition, effect
of steel fibers on the interfacial bond strength was also investigated by incorporating three different percentages of short steel
fibers (0, 0.5, 1 and 1.5 wt.%). Ultimately, the experimental results
were statistically analyzed (ANOVA and S/N ratio) to investigate
the influence of key experimental parameters. Moreover, mathematical models (multiple linear regression) were proposed to estimate the relationship between factors and performance measures.
The following is the summarized conclusions of this study.
Results of epoxy coupon specimens highlight the importance of
determining the properties of each epoxy adhesive for more
precise prediction of the interfacial bond behavior. Test conditions should be similar to the field condition.
Tests of FRP composite specimens revealed that the FRP composites exhibit different stressstrain behavior with different
epoxy adhesives. Their tensile strength, elastic modulus and
ultimate strain capacities vary with the epoxy adhesive and
thus the properties of each strengthening system should be
carefully determined before application.
Fiber-reinforcement in concrete can enhance the material properties of concrete itself and fibers can efficiently reduce the
crack propagation within concrete [29,30]. However, the results
of bending beam tests revealed that the fiber reinforcement was
ineffective to improve the interfacial bond strength. The interfacial fracture energy (GF,int) of plain and fiber-reinforced specimens was almost identical. However, for higher fiber contents
(1% and 1.5%), the GF,int reduced by 1350% that of the corresponding plain geopolymer mortar specimens. One reason for
this reduction in the GF,int can be due to the weak fibermatrix
interface as a large amount of fly ash was contained in geopolymeric binder. This phenomenon was observed in a study conducted by Zhang et al. [47].
The interfacial fracture energy (GF,int) exhibited dependence on
the initial depth of notch. With the increase of notch depth from
33 to 50 mm, GF,int also increased. Plain mortar specimens exhibited about 13% and 43% increase in GF,int as notch depth increased
from 33 to 50 mm for C/G and C/A fabric specimens, respectively.
The bending beam test setup is useful to study the effect of different parameters such as different FRP composites and epoxy
adhesive types. This test setup can be used for a comparison
of different strengthening systems. The interfacial fracture
energy (GF,int), however, was found to be sensitive to the depth
of notch. For consideration of GF,int as a true material property, it
should be independent of the notch depth. Therefore, a more
detailed experimental verification is recommended before consideration of GF,int as a true interface property.
367
368
[36] Hillerborg A, Modeer M, Petersson PE. Analysis of crack formation and crack
growth in concrete by means of mechanics and finite element. Cem Concr Res
1976;6(6):77382.
[37] RILEM 50-FMC Draft Recommendation. Determination of the fracture energy
of mortar and concrete by means of three-point bend tests on notched beams.
Mater Struct 1985;18(4):28790.
[38] Rilem TC. 162-TDF. Test and design methods for steel fiber reinforced
concrete: bending test (final recommendations). Mater Struct 2002;35:
57982.
[39] Petersson PE. Crack growth and development of fracture zone in plain concrete
and similar materials. Division of building materials. Lund Institute of
Technology, Report No. TVBM-1006; 1981. p. 174.
[40] ASTM D638. Standard test method for tensile properties of plastics. ASTM
International. West Conshohocken (PA, USA); 2010.
[41] MSC Korea Co., Ltd. <http://www.mschkorea.com> [last accessed: 01.05.15].
[42] ASTM D3039/3039M. Standard test method for tensile properties of polymer
matrix composite materials. ASTM International. West Conshohocken (PA,
USA); 2008.
[43] Wu ZS, Matsuzaki T, Tanabe K. Interface crack propagation in FRPstrengthened concrete structures. In: Proceedings of the 3rd international
symposium on non-metallic (FRP) reinforcement for concrete structures
(FRPRCS-3), Sapporo, 1416 October 1997. p. 31926.
[44] Wu ZS, Matsuzaki T, Tanabe K. Experimental study of fracture mechanism of
FRP-strengthened concrete beams. In: JCI symposium on FRP reinforced
concrete structures, JCI 1998. p. 11926.
[45] Meier U. Post strengthening by continuous fiber laminates in Europe. In:
Proceedings of the 3rd international symposium on non-metallic (FRP)
reinforcement for concrete structures (FRPRCS-3), Sapporo, 1416 October
1997. p. 4158.
[46] Agarwal BD, Broutman LJ. Analysis and performance of fiber composites, 2nd
ed. John Wiley & Sons Inc.; 1999 [p. 449].
[47] Zhang Y, Sun W, Zongjin L, Zhou X. Geopolymer extruded composites with
incorporated fly ash and polyvinyl alcohol short fiber. ACI Mater J 2009;106
(1):310.
[48] Minitab software (trial version) <http://www.minitab.com/en-us/>.
[49] Taguchi G, Chowdhury S, Wu Y. Taguchis quality engineering handbook.
Berlin: John Wiley & Sons; 2005. p. 1662.
[50] Gauri SK, Chakraborty S. A study on the performance of some multi-response
optimization methods for WEDM processes. Int J Adv Manuf Technol
2010;49:15566.
[51] Tosun N, Cogun C, Tosun G. A study on kerf and material removal rate in wire
electrical discharge machining based on Taguchi method. J Mater Process
Technol 2004;152:31622.