Documente Academic
Documente Profesional
Documente Cultură
Abstract: This paper presents an investigation correlating tensile properties of glass fiber–reinforced polymer (GFRP) bars with the results
Downloaded from ascelibrary.org by Suleyman Demirel Universitesi on 03/07/19. Copyright ASCE. For personal use only; all rights reserved.
obtained from flexural testing. Composite bars, in contrast to conventional steel, when used in structures exposed to aggressive environments
can significantly increase the durability and lifetime. However, the use of GFRP bars in concrete structures is still limited due to the variability
of mechanical properties and bar quality among different manufacturers. Rational and fast testing methods for these reinforcing bars are
needed. A flexure test is relatively easy and potentially could be used to determine the tensile strength of bars, which is a primary mechanical
property that must be determined. This research investigates a possible correlation between the tensile strength and the modulus of rupture of
GFRP bars. A Weibull weakest link model is employed. The effect of bimodularity of GFRP material in tension and compression also is
investigated. DOI: 10.1061/(ASCE)MT.1943-5533.0002575. © 2018 American Society of Civil Engineers.
Author keywords: Glass fiber–reinforced polymer (GFRP); Reinforcement; Composite; Quality control; Civil engineering; Tensile
strength; Modulus of rupture.
Experimental Study
follows (Table 2). Most fiber-reinforced polymer bars contain some
The specimens for this study were provided by two major GFRP kind of surface finishing, which is not effective in carrying internal
bar manufacturers, Company I and Company II. All GFRP bar forces. For straight bars from Company I, with a sand coating, the
specimens were straight and were manufactured using the straight cross-sectional area was determined by diameter measurements ex-
bar (SB) manufacturing (pultrusion) process (Fig. 1). Two different cluding the sand coating. For straight bars from Company II, which
nominal bar diameters were used: #4 and #5 bars from Company I; had grooves cut into the bars, the diameters were measured inside
and 12 and 16 mm bars from Company II. The material character- the grooves. Cross-sectional areas were determined from these
istics of all bar specimens as specified by the suppliers are listed in diameters assuming circular bars.
Table 1. The bars from Company I were sand-coated straight bars Three different tests were performed. A tensile test (CSA 2012)
(SB-I), and the bars from Company II were ribbed straight bars determined the tensile strength in direct tension and the tensile
(SB-II). modulus. The specimen was attached to the testing machine by
Both manufacturers used vinyl ester resin. Detailed information an anchorage system (Fig. 2). The specimens (Fig. 3) included
about the resins is unknown due to the proprietary nature of the the free length (40 times the bar diameter) and twice the anchorage
products. However, for the particular work presented in the paper length listed in Table 3. Five samples for each bar type and each
the details of the resin are not important for validation of the pro- diameter were tested.
posed testing procedure. Flexure tests (ASTM 2014) determined the rupture modulus,
i.e., the tensile strength of the outer fibers in bending [Eq. (1)].
In the method, the specimen is tested as a simple beam in
Testing Procedures three-point bending [Fig. 4(b)]. The sample is cut longitudinally,
obtaining a less than semicircular cross section. Because the com-
The testing procedures and methodologies used in this research pressive strength of the GFRP bar is lower than their tensile capac-
were developed according to CSA S807-10 and CSA S806-12, sup- ity, a less than semicircular cross-section assures the tensile failure
plemented by information included in ACI standard 440.3R-04 of the specimen. Although the top (compressive) portion of the
(ACI 2004) and the ASTM protocols. The number of tests per- sample can be partially damaged by the cutting of the fibers, this
formed in this research program was specified based on guidelines should not have a signigicant influence on the overall test results
provided by CSA S807-10 (CSA 2010) standard. A minimum of (modulus of rupture). Each specimen was visually inspected to
five tests were done on identical samples. This was enough to en- confirm rupture of tensile fibres at failure.
sure the validity of the results. The research included investigations Sample dimensions depend on the bar diameter, and are re-
of short-term tensile, flexural, and compression properties of ported in Table 4. The shape of the semicircular cross section in
GFRP bars. Fig. 4(a) was created by water-jet cutting along the bar length,
Because the work compared test methods, it was considered im- and resulted in only minor variations in dimensions; the thickness
portant to obtain cross-sectional areas that are effective in carrying of all samples was measured before testing. Five samples for each
stresses. The geometrical properties of the bars were determined as bar type and each diameter were tested.
The tensile strength in bending can be calculated using the
standard formula for linear elastic beams
PLðh − cÞ
σb ¼ ð1Þ
4I
Fig. 3. Tensile test specimen M16 bar total length 1,400 mm: (a) top view; and (b) side view.
where σb = tensile stress in outer fibers at midspan (MPa) due to samples were attached to the testing machine by an anchorage de-
bending; P = load at midspan (N); L = span (mm); I = moment of vice, which applied a uniform compression stress (Fig. 5). The total
inertia (mm4 ); and c = distance from centroid to extreme ten- length of the specimen was 335 mm for bars from Company I and
sile fiber. 325 mm for bars from Company II. The load-displacement and
A compression test (ASTM 2003) determined the modulus of strain (measured by strain gauge) at the middle of the bar length
elasticity in compression. Similar to the tensile test, compression were recorded during the test. Five samples of only one type
(SB) and of one size (12 mm and #4) of each company’s specimens
were tested.
Table 3. Tensile test specimen dimensions
Effective Free length Anchorage Total Tests Results
Bar type diameter (mm) (mm) length (mm) length (mm)
The details of the tensile testing done for this research and issues
SB (1–5) 12=14 480=560 380 1,250=1,320
related to it were described by Arczewska et al. (2016). Results
SB (6–10) 16=18 640=720 380 1,400=1,480
from tensile testing of bars that failed by visible bar rupture were
Fig. 4. (a) Specimen cross section; and (b) flexure test device.
available for bars from Company II only. The anchorage length testing was based on ASTM D3410/D3410M-03. The test was per-
used for specimens (40 times the bar diameter) was insufficient formed for straight bars, #4 bars from Company I and 12-mm bars
for the type of bar coating used by Company I, and the test resulted from Company II. The results are shown in Table 7. Based on pre-
in slipping of the anchorage, resulting in lower recorded strength vious research (Arczewska et al. 2016), compression stiffness does
than the actual tensile strength. Therefore, the values specified by not differ for different bar diameters and therefore the same moduli
the manufacturer (Company I) were used for further investigation of elasticity were used for #5 bars from Company I and for 16-mm
in this paper (Table 1). Test results included the tensile strength of a bars for Company II.
bar, calculated as a maximum force recorded at failure divided by
the effective cross-section area, and the modulus of elasticity
(Table 5). Material Dual Modulus
The results from the flexure test were the tensile strength of the
external fibers, i.e., the modulus of rupture (Table 6) calculated FRP materials exhibit linear behavior but different stiffnesses (elas-
from Eq. (1). All the tested specimens failed by the desired tensile tic moduli) and strengths in tension and compression (Jones 1978).
rupture of the external fibers (Fig. 6). The results in Table 7 are The difference between tensile and compression properties varies
consistent with standard ASTM D4476/D4476M-14, which means depending on the type of FRP composites: for GFRP it is relatively
that the strains were calculated assuming the same stiffness moduli small, but for carbon FRP the tensile modulus is 2–5 times higher
in tension and compression. than the compression modulus. According to Jones (1977, 1978),
The needed information from the compression test, for this par- this difference in moduli is the result of fiber:matrix ratio in which
ticular investigation, was the modulus of elasticity in compression. fibers either tend to be bonded with the matrix, and hence stiffen the
Previous research (Arczewska et al. 2016) revealed issues associ- material when the tensile stress is applied, or fibers buckle when
ated with different types of compressive testing. For this paper, compression stress is applied, which can cause material weakening.
σc σt
Downloaded from ascelibrary.org by Suleyman Demirel Universitesi on 03/07/19. Copyright ASCE. For personal use only; all rights reserved.
The elastic theory of materials with dual elastic moduli was ¼ ð5Þ
studied by Jones (1977, 1978), Medri (1982), and Ambartsumyan Ec c Et ðh − cÞ
(1982). Researchers proposed a bilinear stress-strain relationship,
which simplified the actual material behavior. In fact, a nonlinear where σt = maximum tensile stress (rupture modulus); σc = maxi-
transition exists between the tension and compression linear por- mum compression stress; Et = tensile modulus; Ec = compression
tions of the stress-strain curve. This simplification can cause prob- modulus; c = height of compression stress block; (h − c) height of
lems for stresses close to zero, but it is justified for most practical tension stress block; εt = maximum tensile strain; and εc =
cases in which the analysis is about failure stresses, which is the maximum compression strains.
case in this research. The stiffness moduli in compression, Ec , and tension, Et , are
Despite the fact that GFRP certainly can be classified as a given in Tables 1, 5, and 7. For Company I, Ec ¼ 45 GPa and
dual-modulus material, and during bending both states of stress Et ¼ 65.6 GPa (#4) and 62.6 GPa (#5), and for Company II,
(tensile and compression) are present in the material, the ASTM Ec ¼ 53 GPa and Et ¼ 62 GPa (M12) and 63 GPa (M16).
D4476/D4476M-14 standard does not account for this in calcula- Considering the known failure moments M and the cross-
tions of bending capacity. This can lead to underestimation of the sectional areas, the calculations of the moduli of rupture are listed
rupture modulus. The dual-modulus relationship proposed by Jones in Table 8. The table lists results from a modified test, in which the
(1977, 1978) and Ambartsumyan (1982) was adopted for flexural modulus of rupture was calculated taking into account the bimod-
strength determination in this work. ular characteristic of the composite material.
Comparing these results with values obtained directly from the
flexure test (Table 6) shows that the modulus of rupture is higher
Three-Point Bending Test: Modified Approach when using the modified procedure than that based on the ASTM
D4476/D4476M-14 standard. The position of the neutral axis is
The modified approach refers to the calculations of rupture modu- sensitive to the difference between moduli and is shifted toward
lus using different moduli in tension and in compression. To deter- the section with a higher modulus of elasticity. Therefore, due to
mine the influence of GFRP dual modulus on bending properties, change of the stress-block geometry (position of neutral axis) and
the bilinear stress-strain curvature was introduced to the three basic the fact that the system needs to be in equilibrium [Eqs. (1) and (2)],
equations: equilibrium of forces [Eq. (2)], equilibrium of moments the calculated tensile strength of the outer fibers increased com-
[Eq. (3)], and linearity of strains [Eq. (4)] (Beer 2012). Distribution pared with results from the ASTM D4476/D4476M-14 standard.
of stresses and strains along the specimen cross section is shown in In the case of GFRP bars, the difference between the modified
Fig. 7. Furthermore, Eq. (5) follows from Eq. (4) to include the two and unmodified modulus of rupture was small and did not exceed
moduli in order to apply these equations for determination of strain 10%. However, this difference depends on the difference between
and stress states at failure moduli. The influence of different ratios of tensile to compression
Z Z moduli on the modulus of rupture is presented in Fig. 8. The curve
shows the analysis performed for one semicircular cross section.
σdA þ σdA ¼ 0 ð2Þ
At Ac
where
Fig. 8. Effect of modular ratio on calculated modulus of rupture Z
σ − σu m
(tensile strength in bending). β¼ dV
σo
Downloaded from ascelibrary.org by Suleyman Demirel Universitesi on 03/07/19. Copyright ASCE. For personal use only; all rights reserved.
Z m
σb − σu m σ 1 Effective Volume for Semicircular Cross Section
dV b ¼ V b b ð15Þ
Vb σ o σ o 2ðm þ 1Þ2
In a three-point bending test, tensile stress changes not only along
the sample height but also along the sample length. (Fig. 10). Thus,
which, after combining with Eq. (14), yields to determine the effective volume of a sample subjected to the
three-point bending, integration needs to be done over the portion
m m
σt σb 1 of the specimen that is stressed in tension.
Vt ¼ Vb ð16Þ The Weibull weakest link model is founded on the primary
σo σo 2ðm þ 1Þ2
assumption that in order for failure to occur, the largest flow not
only needs to be present in the sample volume, but it also needs
The ratio of stress in direct tension to tensile stress due to bend- to be located in the portion of the sample that experiences the larg-
ing can be expressed as est stress. This volume of the sample that is under the largest stress
is called an effective volume, and is described as an integral over
1
σb V t 2ðm þ 1Þ2 m the sample volume stressed in tension. The applied stress in
¼ ð17Þ Weibull theory is ðσÞ ¼ ½ðσ − σu Þ=σo m . Thus the effective volume
σt Vb
is equal to V Eb ¼ ∫ Vb ½ðσb − σu Þ=σo m dV b . Because σu is specified
as a stress under which the failure will not occur, it is assumed to
Simplifying Eq. (17) obtains be 0. The σo is specified as a failure stress, which for a single
sample can be represented by the maximum stress that sample will
1
σb V Et m experience, σmax . A general equation for effective volume in this
¼ ð18Þ case is
σt V Eb
Z m Z m
σb − σu σb
where V Et ¼ V t and V Eb ¼ ½V b =2ðm þ 1Þ2 are effective volumes V Eb ¼ dV b ¼ dV b ð20Þ
in direct tension and bending, respectively. Vb σo Vb σmax
In the simple case of direct uniform tension, the effective vol-
ume V Et is assumed to be the specimen volume V (V Et ¼ V t ¼ V). where σu ¼ 0; and σo ¼ σmax , where σmax = maximum tensile
In other loading cases, such as bending, the effective volume is stress due to three-point bending.
smaller than the total volume of the sample or the sample volume The tensile stress varies linearly with distance x (Fig. 11), thus
stressed in tension (V Eb < V b < V). the stress change along the sample length can be written as
The relation between V and V Eb can be expressed as (Weil and
Daniel 1964) x 2x
σb ¼ σ max ¼ σ ð21Þ
L
2
L max
V Eb ¼ kV ð19Þ
where x ∈ ½0; ðL=2Þ due to the symmetry of the bending moment
where k = loading factor; and V = total volume. diagram.
For a rectangular cross-section specimen under three-point Assuming that the tensile stress varies linearly with the height
bending, k ¼ ½1=2ðm þ 1Þ2 and the effective volume V Eb ¼ of the cross section, the stress distribution function can be deter-
½V=2ðm þ 1Þ2 . The concept is illustrated in Fig. 9. mined as
Fig. 10. Bending moment diagram and stress distribution in sample due to bending.
(a)
Weibull Modulus
The Weibull modulus, also known as a shape parameter (dimen-
sionless), describes the distribution of flaws in a material. The
higher the Weibull modulus, the more uniform is material that is
Downloaded from ascelibrary.org by Suleyman Demirel Universitesi on 03/07/19. Copyright ASCE. For personal use only; all rights reserved.
Table 9. Data set for unmodified results for #4 bar for Company I
Sample number Strength (MPa) x ¼ lnðstrengthÞ Pf ¼ ði − 0.5Þ=n Y ¼ lnfln½1=ð1 − PfÞg
1 1,569.6 7.358576089 0.1 −2.250367327
2 1,623.13 7.392111663 0.3 −1.030930433
3 1,645.98 7.40609123 0.5 −0.366512921
4 1,651.23 7.409275744 0.7 0.185626759
5 1,700.37 7.438601153 0.9 0.834032445
Table 10. Data set for modified results for #4 bar for Company I
Sample number Strength (MPa) x ¼ lnðstrengthÞ Pf ¼ ði − 0.5Þ=n Y ¼ lnfln½1=ð1 − PfÞg
1 1,741.6 7.46255951 0.1 −2.250367327
2 1,800.84 7.496008502 0.3 −1.030930433
3 1,826.62 7.510222544 0.5 −0.366512921
4 1,832.4 7.513381862 0.7 0.185626759
5 1,886.61 7.542536847 0.9 0.834032445
On the trend line in Fig. 11, the approximate number 300 de- Correlation between Tensile Strength due to
rives from the transformation of the three parameter Weibull Bending and Direct Tension Strength
strength distribution model Pf ¼ 1 − exp ½−ðσ − σu Þ=σ0 m to the
linear function lnfln½1=ð1 − PfÞg ¼ m½lnðσ − σu Þ − ln σ0 using Using information obtained from the previous section (Weibull
Downloaded from ascelibrary.org by Suleyman Demirel Universitesi on 03/07/19. Copyright ASCE. For personal use only; all rights reserved.
data obtained from the three-point bending test. The number modulus), the effective volumes of the semicircular cross-sectional
300 is correlated with the σ0 characteristic strength (the scale samples were calculated from Eq. (24). Results are given in
parameter) and can be described as the stress at which 63.2% of Tables 12–15 for unmodified and modified methodologies for pre-
specimens will fail, and thus it can be written as 300 ¼ m ln σ0 . diction the modulus of rupture. Tables 12–15 also report the
If m ¼ 40 (as in this case), then σ0 ¼ 1808 MPa. This number tensile stress ratios (tensile stress of the outer fibers due to bending
is not directly used in the correlation between the tensile and flex- versus tensile stress from direct tension) calculated from Eq. (18).
ure strength, but it characterizes the distribution of results obtained Experimental tensile stress ratios are also presented for comparison.
from the three-point bending test. Direct tension strengths for bars from Company I were taken from
To interpret the importance of the bimodular effect (in flexure the manufacturer’s data (as explained previously). Direct tension
analysis) on the correlation between tensile and flexure tests, both test results for bars from Company II and bending test results
results (from the modified and unmodified methods) are presented for both companies were measured in this research.
subsequently. Because of the method’s repeatability, just one exam- For a better understanding of the results, a relative error
ple of the data table and the corresponding Weibull graph are shown (percentage value of the difference between values) for the results
subsequently. obtained from correlation and results obtained directly from tests is
The Weibull modulus m is independent of the methodology presented in Table 16 for bars from Company I and in Table 17 for
(constant or bimoduli analysis) used for the determination of bars from Company II.
Table 12. Tensile-flexure stress correlation (using unmodified data from flexure test) for Company I
V Et in σb , not Effective volume in bending, Tensile stress ratio calculated Tensile stress ratio (σb =σt )
Bar size Value σt tension (mm2 ) modified V Eb , for m ¼ 40 (mm2 ) σb =σt ¼ ðV Et =V Eb Þ1=m directly from tests
#4 Mean 1,312 86,161.6 1,638.06 0.536 1.349 1.25
SD — — 42.49 0.026 0.002 —
COV — — 0.03 0.048 0.001 —
#5 Mean 1,184 183,124.8 1,510.03 1.338 1.389 1.28
SD — — 42.22 0.077 0.002 —
COV — — 0.03 0.057 0.002 —
Table 13. Tensile-flexure stress correlation (using modified data from flexure test) for Company I
V Et in σb , not Effective volume in bending, Tensile stress ratio calculated Tensile stress ratio
Bar size Value σt tension (mm2 ) modified V Eb , for m ¼ 40 (mm2 ) σb =σt ¼ ðV Et =V Eb Þ1=m (σb =σt ) directly from tests
#4 Mean 1,312 86,161.6 1,817.61 0.476 1.353 1.39
SD — — 47.16 0.021 0.001 —
COV — — 0.03 0.043 0.001 —
#5 Mean 1,184 183,124.8 1,653.05 1.212 1.393 1.40
SD — — 46.28 0.072 0.002 —
COV — — 0.03 0.059 0.002 —
Table 14. Tensile–flexure stress correlation (using unmodified data from flexure test) for Company II
V Et in σb , not Effective volume in bending, Tensile stress ratio calculated Tensile stress ratio
Bar size Value σt tension (mm2 ) modified V Eb , for m ¼ 40 (mm2 ) σb =σt ¼ ðV Et =V Eb Þ1=m (σb =σt ) directly from tests
M12 Mean 1,222.98 54,259.20 1,927.02 1.356 1.657 1.58
SD 93.24 91.44 0.226 0.013
COV 0.08 0.05 0.166 0.008
M16 Mean 1,269.98 128,614.40 1,802.89 2.172 1.581 1.42
SD 50.48 76.32 0.039 0.001
COV 0.04 0.04 0.018 0.001
Table 16. Relative error for tensile–flexure strength correlation for Company I
Unmodified ASTM 447 Unmodified ASTM 447 Modified method Modified method
tensile stress ratio tensile stress ratio Error tensile stress ratio tensile stress ratio
Bar size calculated σb =σt from tests σb =σt (%) calculated σb =σt from tests σb =σt Error (%)
#4 1.35 1.25 8 1.35 1.39 2.8
#5 1.39 1.28 8.6 1.39 1.40 0.7
Table 17. Relative error for tensile–flexure strength correlation for Company II
Unmodified ASTM 447 Unmodified ASTM 447 Modified method Modified method
tensile stress ratio tensile stress ratio Error tensile stress ratio tensile stress ratio
Bar size calculated σb =σt from tests σb =σt (%) calculated σb =σt from tests σb =σt Error (%)
M12 1.66 1.57 5.7 1.66 1.64 1
M16 1.58 1.45 8.9 1.58 1.51 4
The error in both cases for the modified and unmodified • Direct tensile testing must remain a primary test for establishing
methodologies was relatively small (did not exceed 10%) compared tensile strength, stiffness, and the stress-strain relationship of
with results directly obtained from tests. This suggests that corre- GFRP bars. However, a faster, simpler, and cheaper flexure test
lation between the tensile test and the flexural test does exist for is shown to have the potential to be an efficient method for
GFRP bars, and the methodology presented in this paper has a determining tensile strength and stiffness of GFRP bars via the
potential for practical application in testing to determine tensile proposed correlation in this research.
strength.
Acknowledgments
Conclusions
The authors acknowledge the financial support from the Ministry of
The tensile–flexure strength correlation was investigated in this Transportation of Ontario (MTO), Ontario Centers of Excellence
study by implementation of the Weibull weakest link flaws (OCE), and Schoeck Canada. Materials were supplied by Schoeck
distribution model. To minimize the correlation error, the material Canada and Pultral.
bimodularity was introduced in the three-point bending test calcu-
lations, and the obtained results were compared with standardized
methodology assuming the same moduli of elasticity in tension and
References
compression (ASTM 2014). Based on the obtained results, the
following conclusions can be made: ACI (American Concrete Institute). 2004. Guide test methodology for fiber:
• The Weibull weakest link model proves to be a proper metho- Reinforced polymer (FRPs) for reinforcing or strengthening concrete
dology for obtaining a correlation between the tensile strength structures. ACI 440.3R–04. Farmington Hills, MI: ACI.
from direct tension and modulus of rupture from flexure. The Ambartsumyan, S. A. 1982. Elasticity theory of different modulus. Edited
presented procedure recognizes the difference between both by W. Ruifeng and Z. Yunzhen, 11–32. Beijing: China Railway Press.
tensile and flexure testing protocols (different load application Arczewska, P., M. Polak, and A. Penlidis. 2016. “Determination of proper-
and stress distribution along specimens and flaw distributions in ties and quality of glass fiber reinforced polymer composite reinforcing
the material) and allows predicting with satisfactory accuracy bars.” Int. J. Mater. Eng. Technol. 15 (4): 225–261. https://doi.org/10
.17654/MT015040225.
the tensile capacity of the GFRP bar using only the three-point
ASTM. 2003. Standard test method for compressive properties of polymer
bending test.
matrix composite materials with unsupported gage section by shear
• Based on the obtained results, it can be concluded that dual loading. ASTM D3410/D3410M. West Conshohocken, PA: ASTM.
moduli of the material do not have a significant influence on ASTM. 2014. Standard test method for flexural properties of fiber
the final correlation results for GFRP bars with similar stiffness reinforced pultruded plastic rods. ASTM D4476/D4476M. West
in tension and compression. However, for products with large Conshohocken, PA: ASTM.
differences between tensile and compression moduli, the bimod- Balendran, R. V., T. M. Ranat, T. Maqaood, and W. C. Tang. 2002.
ular characteristic of the material should be considered. “Application of FRP bars as reinforcement in civil engineering
Jones, R. M. 1977. “Stress-strain relations for materials with different Tripathi, V. K. 2003. “Standardization of test methods for property evalu-
moduli in tension and compression.” AIAA J. 15 (1): 16–23. https://doi ation of FRP bars.” Ph.D. thesis, Dept. of Civil Engineering, West
.org/10.2514/3.7297. Virginia Univ.
Jones, R. M. 1978. “Mechanics of composite materials with different Weil, N. A., and I. M. Daniel. 1964. “Analysis of fracture probability in
moduli in tension and compression.” Final Scientific Rep. Research nonuniformly stressed brittle materials.” J. Am. Ceram. Soc. 47 (6):
Sponsored under AFOSR Grant Number 73-7532 Air Force Office 268–274. https://doi.org/10.1111/j.1151-2916.1964.tb14413.x.