Construction and Building Materials 25 (2011) 2813–2821

Contents lists available at ScienceDirect

Construction and Building Materials

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

Fundamental mechanisms of bonding of glass fiber reinforced polymer

reinforcement to concrete
Wai How Soong a, J. Raghavan a,⇑, Sami H. Rizkalla b
a
Composite Materials and Structures Research Group, and Department of Mechanical and Manufacturing Engineering, University of Manitoba, Winnipeg, Canada MB R3T 5V6
b
Department of Civil, Construction, and Environmental Engineering, North Carolina State University, Raleigh, NC 27695, USA

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

Article history: Fundamental mechanisms of bonding between glass fiber reinforced polymer (GFRP) bar and concrete are
Received 7 September 2010 presented. Contributions from chemical bonding, bearing resistance, and frictional resistance to bond
Received in revised form 23 December 2010 were delineated by measuring the following: the load corresponding to complete debonding of the
Accepted 24 December 2010
bar, the load corresponding to onset of sliding and pullout of the bar along the entire embedment length,
Available online 21 January 2011
and the frictional load corresponding to frictional resistance to sliding. Research findings indicate that
while chemical bonding was the main contributor to the interfacial bond strength, the other two mech-
Keywords:
anisms contributed to the pullout strength of the bar. Correlation between the bar’s surface geometry and
Polymer composite
Reinforced concrete
the contributions from the three mechanisms are discussed.
Pullout Ó 2010 Elsevier Ltd. All rights reserved.
Interfacial strength
Bond strength
Fiber reinforced polymer

1. Introduction strength of the concrete, confinement pressure exerted by the con-

crete on the bar, and surface geometry of the bar.
Recently, polymer composite materials are used in civil engi- Al-Zahrani [1] and Benmokrane et al. [2] have observed a de-
neering applications in various forms including rebars for concrete crease in interfacial bond strength with increase in embedment
structures, sheets for flexural and shear strengthening, and sheets length, as shown in Table 1 for different glass fiber reinforced plas-
to wrap concrete columns and bridge piers to increase the confine- tic (GFRP) bars with axi-symmetric lugs [1] and helical lugs [2]. It
ment. Fiber Reinforced Polymer (FRP) bar is an excellent alternative has been well established that shear stress at the fiber–matrix
to steel bars due to their relatively higher corrosion resistance and interface, at any applied load during a single fiber direct pullout
high specific strength and modulus. Several successful applications test, is not constant. Al-Zharani et al. [3] and Benmokrane et al.
can be found in North America. While design codes are well estab- [4] have measured this shear stress distribution experimentally,
lished for the use of steel bars in concrete members, they are slowly for concrete reinforced with FRP bar. The interfacial bond strength
evolving for FRP bars. Development of these codes requires a good tabulated in Table 1 has been determined using average shear
understanding of bond between FRP bars and concrete and the rela- stress and it varies with embedment length, since the shear stress
tionship between the bond strength, and various material and test distribution and the average shear stress would change with
parameters. Past research has substantially advanced the knowl- embedment length. In addition, reported bond strength values cor-
edge in this area. Nevertheless, a substantial variation is observed respond to bar pullout rather than onset or completion of debond-
among the published data on interfacial bond strength, which sug- ing. This is also a reason for the variation observed in Table 1 and
gests that a comprehensive understanding is yet to evolve. will be discussed in subsequent sections.
Several variations of bar pullout and beam test have been used Test results indicate that increasing bar diameter caused de-
to study the bond characteristics of FRP bars. Various parameters, crease in the bond strength as given in Table 2 for smooth bars
whose effect on interfacial bond strength have been studied in [1] and bars with lugs [5]. While variation of shear stress distribu-
the past, are bar embedment length, bar diameter, compressive tion could be one reason, Tighiouart et al. [5] have indicated that
bleeding of water could be another reason since an increase in
bar diameter could increase the amount of water trapped near
⇑ Corresponding author. Address: Department of Mechanical and Manufacturing
the bar leading to higher amount of voids and lesser contact area.
Engineering, E2-327, EITC Building, University of Manitoba, 75, Chancellor Circle,
Winnipeg, Canada MB R3T 6A6. Tel.: +1 204 474 7430; fax: +1 204 275 7507.
Al-Zharani et al. [3] have varied the compressive strength of the
E-mail address: rags@cc.umanitoba.ca (J. Raghavan). concrete, reinforced with bars wrapped with lugs, from 31.4 MPa

0950-0618/$ - see front matter Ó 2010 Elsevier Ltd. All rights reserved.
doi:10.1016/j.conbuildmat.2010.12.054
2814 W.H. Soong et al. / Construction and Building Materials 25 (2011) 2813–2821

Table 1 Table 4
Published results on effect of embedment length on bond strength. Published results on effect of lugs on bond strength.

Reinforcement Bar Embedment Average Reference Reference Reinforcement Description Lug Lug Bond
diameter length bond width depth strength
(mm) (mm) strength (mm) (mm) (MPa)
(MPa)
Al-Zahrani Smooth GFRP Smooth – – 0.97
C-BAR™ (P1) 12.7 63.5 18.4 Benmokrane [1]
et al. [2] Machined Lug 3.8 1.3 39.7
12.7 127.0 14.5 GFRP
Machined Lug 8.9 1.3 23.2
H.B. Rebar 12.5 63.5 15.1
GFRP
12.5 127.0 12.7
Benmokrane C-BAR (P1) Lug 2.8 1.3 18.4
Machined 12.7 63.5 39.7 Al-Zahrani et al. [4]
GFRP [1]
12.7 127.0 29.7 Note: Embedment length = 63.5 mm; bar diameter = 12.7 mm.

Test results of Benmokrane et al. [2] and Al-Zahrani [1] are com-
pared in Table 5 to illustrate the influence of loading rate and
Table 2 inconsistency in the definition of contact surface area used in the
Published results on effect of bar diameter on bond strength. calculation of interfacial bond strength. Since the dimensions and
Reinforcement Nominal bar Embedment Average bond Reference the properties of the reinforcing bar and the concrete were same,
diameter length (mm) strength the difference in the maximum pullout load is believed to be due
(mm) (MPa) to the difference in the loading rates used by the two groups of
Type A 12.7 127 10.6 Al-Zahrani researchers. In addition, the interfacial bond strength reported by
et al. [3] Al-Zaharani [1] is higher than that reported by Benmokrane et al.
15.9 127 7.3
[2] despite the lower value for the measured pullout load and same
19.1 127 6.6
25.4 127 6.4 bar dimensions. This apparent discrepancy is believed to be due to
the difference in the definition and calculation of the contact sur-
Type B 12.7 127 12.3
15.9 127 10.8 face area. While Benmokrane et al. [2] have calculated the contact
19.1 127 – surface area using the average diameter of the reinforcing bar, Al-
25.4 127 7.4 Zahrani [1] has calculated the contact area using the actual diam-
Smooth GFRP 6.35 63.5 1.37 Al-Zahrani eter of the reinforcing bar and the dimensions of the lugs.
[1] The above discussion suggests that the interfacial bond strength
12.7 63.5 0.97 could be very much dependent on test parameters and surface
geometry of the bar, whereas the interfacial bond strength should
be a unique value independent of test parameters and surface
geometry of the bar.
to 66.1 MPa and have observed no change in the interfacial bond
strength. This is to be expected since the failure was interfacial. An-
other work of Al-Zahrani [1] has shown that the induced lateral 2. Fundamental behavior during pullout tests
force could be influenced by the mismatch in Coefficient of Ther-
mal Expansion (CTE) between the concrete and the bar as observed A brief discussion on the pullout behavior of a FRP bar from con-
in Table 3 for concrete reinforced with a smooth bar. A test temper- crete is included here to assist in the comprehension of the scatter
ature lower than the cure temperature resulted in negligible bond in published interfacial bond strength and to provide the basis for
strength probably due to lack of lateral pressure caused by the con- the research approach used in this study.
traction of the bar. However, a test temperature higher than the This development length for civil engineers (lc), which is also
cure temperature resulted in a bond strength higher than the ref- known as critical length among composite community, can be pre-
erence case, for which the test and cure temperatures were same. dicted using Eq. (1) based on assumption of constant shear stress
This is probably due to increase in lateral pressure due to expan- distribution along the embedded fiber length.
sion of the bar. These results are similar to the results of Leung
and Geng [6], who have shown that the interfacial bond for con- lc ¼ rf d=4su ð1Þ
crete reinforced with steel bars increases with lateral pressure.
where rf is the fiber strength, d is the fiber diameter and su is the
Researchers have also varied the surface geometry of the FRP
fiber–matrix interfacial bond strength.
bars to enhance the resistance to sliding and pullout strength as
Load applied to the concrete–FRP bar interface, during direct
tabulated in Table 4. Al-Zahrani [1] has observed decrease of the
pullout test of a concrete specimen reinforced with a FRP bar of
bond strength by increasing the lug width from 3.8 mm to
embedment length (l) 6 lc, is shown schematically in Fig. 1 as a
8.9 mm. The latter also changed the failure mode from interfacial
function of slip between the bar and the concrete. Fig. 1a repre-
failure to crushing failure of the concrete between the lugs.
sents the possible load–slip curves for a concrete specimen rein-
forced with a smooth FRP bar. Curve A represents the case when
Table 3 debonding of the smooth FRP bar occurs at a maximum load (F 0d ).
Published results on effect of lateral pressure on bond strength [1].
The applied load may drop either to zero or to a finite frictional
Curing temperature, Tc Testing temperature, Tt Interfacial bond strength force value, F 0f . Pullout behavior of the FRP bar may also follow
(°C) (°C) (MPa) curves B or C since the interfacial stress varies along the embed-
60 20 <0.04 ment length. Curve C represents possible case for a stable debond-
20 20 <0.04 ing of the bar, which starts at F 0i and progresses to completion along
20 60 1.79
the entire embedment length at F 0j . Curve B represents possible
20 20 0.831
partial debonding. In this case, the remaining bonded portion of
 W.H. Soong et al. / Construction and Building Materials 25 (2011) 2813–2821 2815

Table 5
A comparison of published results to illustrate the influence of test conditions and contact surface area definition on the interfacial bond strength.

Reference Reinforcement Nominal lug dimension Nominal bar Loading/ Max. pullout Interfacial bond
depth  width (mm) diameter (mm) displacement rate load (kN) strength (MPa)
Benmokrane C-BAR™ (P1) commercial 1.3  2.8 12.7 22 kN/min 46.0 18.4
et al. [2] available GFRP bar
Al-Zahrani Machined GFRP bar 1.3  3.8 12.7 0.125 mm/min 24.2 39.7
et al. [3]

bond strength. The maximum pullout load, Fp, can be predicted

(a) Fp (b) using the following equation
A’
B’ G’ Fp ¼ Fd þ Fb þ Ff ð2Þ
C’ D’ where Fd, Fb, and Ff are debond load, bearing load, and frictional load
Load

Ff
F d’ Fd respectively. Fd, Fb, and Ff represent contributions to pullout load
A G E’ from chemical bonding, bearing resistance, and frictional resistance
B
Ff ’ D respectively.
C
F i’
E Fi The interfacial bond strength determined using the debond load
would be a unique value independent of the surface geometry of
0 0 the bar and the test conditions. However, the interfacial bond
Slip strength determined using the pullout load would not be a unique
value because the pullout load is dependent on these three mech-
Fig. 1. A schematic of load versus slip behavior during single fiber direct pullout
anisms, two of which are dependent on the surface geometry of the
test.
bar and the test conditions. Unfortunately, most of the previous
studies have used the pullout load to determine the interfacial
bond strength. Since the parameters that influence the bearing
the bar debonds in an unstable manner at the peak load and the and frictional loads were varied arbitrarily, the reported values
load drops suddenly to F 0f . Therefore, for cases B and C partial slid- for the interfacial bond strength also vary by a wide margin, from
ing and pullout of the bar in the debonded region, and progressive researcher to researcher. In addition, owing to lack of delineation
debonding of the bar in the bonded region would occur simulta- of contributions from the three mechanisms, previous researchers
neously during loading. Since frictional force supports the sliding were unable to correlate the measured pullout load to test param-
and pullout, the actual load–slip behavior is dependent on the fric- eters and surface condition of the bar.
tional resistance, the interfacial bond strength, and the length of Hence, the objectives of this research are to (a) delineate the
bonded and debonded regions at any given load. contributions from the three mechanisms (chemical bonding, bear-
The interfacial bond strength calculated using F 0d , for case A, or ing resistance, frictional resistance) to the pullout load, (b) corre-
F 0i , for cases B and C, would be a measure of the chemical bonding late these contributions to bar characteristics such as surface
between the FRP bar and the concrete. Owing to weak chemical roughness, lug pitch, and number of lugs per unit embedment
bonding between concrete and the FRP bar, researchers have mod- length, and (c) study the effect of loading rate on pullout load.
ified the FRP bar surface, using lugs and sand particles, to enhance Additionally, interpretation of the published results is also compli-
its resistance to sliding and pullout. Fig. 1b represents the load–slip cated due to the variation in the definition of the contact surface
plot for debonding and pullout of a bar with lugs or sand particles. area from one researcher to another and by the assumption of con-
Due to shear stress distribution along the embedment length, dur- stant shear stress distribution along the embedment length. The
ing a pullout test, debonding of the FRP bar would start at Fi and former issue is addressed in this study through explicit definition
would complete at Fd, for all the three cases in Fig. 1b. Curve A0 rep- of the contact surface area. The latter issue is deferred to a future
resents the case when the applied load continues to increase even study though the impact of this could be observed in this study too.
after the completion of debonding at Fd. This increase is attributed Since the completion of this study, there has been a number of
to the bearing resistance, caused by mechanical interlocking of the published experimental research studies [7–11] focusing on the
surfaces of the concrete and the bar, and frictional resistance, pullout behavior of FRP bars from concrete. However, none of them
caused by the bar’s surface roughness. When the lugs or the sand have delineated the contributions to pullout load from various
particles along the entire embedment length are sheared at the mechanisms, as done in this study.
maximum load Fp, the bearing resistance due to them is eliminated Finally, even though sliding and pullout of the bar after com-
and the load drops suddenly to the frictional force, Ff. Alternatively, plete debonding is not focused in this study, it is worth mentioning
shearing of lugs and sand particles can be progressive as observed to understand the measured load–slip curve. If the bar is of infinite
in this study. Curve C0 represents the case when the shearing of length and the interface is not altered during pullout, the bar will
lugs or sand particles is complete before reaching the maximum pullout at constant F 0f or Ff and curves D or D0 would be recorded
load. Curve B0 represents the case when the shearing is partial respectively. If the bar is of finite length, load required to overcome
and hence, a load drop is registered at the maximum load due to the friction will reduce with fiber pullout due to reduction in the
the sudden shearing of the remaining intact lugs or sand particles. embedment length (l) and curves E or E0 would be recorded. If con-
Between Fi and Fd, partial sliding and pullout of the bar occurs stant frictional resistance is assumed, then curves E and E0 would
along the debonded length. Assuming that the contribution from have a constant slope. If frictional resistance is reduced due to
bearing and frictional resistances to Fd, during this partial sliding abrasion between the bar and the concrete, curves E and E0 would
and pullout, is negligible either Fi or Fd can be used in determining have non-linear slopes as shown in Fig. 1b. This is termed as slip
the interfacial bond strength. However, most of the researchers weakening. In many cases, a periodic increase and drop in the load
have used the load at which sliding and pullout of the bar occur is observed during sliding and pullout as shown in curve E0 . This is
along the entire embedment length to determine the interfacial due to one or more of the following: (a) resistance from the lugs in
2816 W.H. Soong et al. / Construction and Building Materials 25 (2011) 2813–2821

the free end of the bar while it moves against the concrete during were allowed to cure for 28 days. The specimen identification codes are tabulated
in Table 6. The compressive strength of the concrete was determined as per ASTM
testing of a specimen type used in this study; all lugs in the free
C39-86 to be 49.77 ± 2.26 MPa.
end of the composite reinforcing bar were removed in this study
to avoid this; (b) resistance caused by the debris of sheared lugs
3.3. Direct pullout test
or sand particles; (c) resistance due to mechanical interlocking of
the surfaces of the bar and the concrete due to surface roughness. The direct pullout test was carried out in stroke-control mode using a MTS
Alternatively, if frictional resistance is increased due to debris from hydraulic test frame with a maximum loading capacity of 1000 kN. The test setup
interaction between the bar and the concrete during pullout, is shown in Fig. 4. During loading the concrete cylinder was pressed against a steel
curves G or G0 would be observed. Such a behavior is often termed plate, bolted to the ground using two 2.25 in. diameter bolts. Since the test speci-
men surface was not flat, a thin layer of plaster was applied on the test specimen
as slip strengthening. surface to ensure uniform contact with the steel plate. Two LVDT’s were attached
rigidly to the specimens using specially designed fixture as shown in Fig. 4. Relative
3. Experimental details slip between the bar and the concrete was measured using these LVDTs at both
ends of the bar. The LVDT data was acquired using a data acquisition system and
3.1. Material stored in a computer. All tests were done at a displacement rate of 1.3 mm/min.
Two additional loading rates of 0.26 mm/min and 6.5 mm/min were used to study
Four types of GFRP bars used in this study are shown in Fig. 2. The diameters of the influence of loading rate on pullout load. While most of the specimens were un-
smooth (S) and sand-coated (SC) bars were 12.7 mm and 13.6 mm respectively. loaded after recording the minimum load to which the applied load dropped be-
Their average tensile strength and modulus, as per manufacturer’s data sheet, were yond Fp, few specimens were loaded beyond this point to record the pullout
683 MPa and 40 GPa respectively. The bars with lugs (RL), which were basically re- behavior. A minimum of three specimens was tested to obtain each data point.
sin-impregnated fiber strands, had shape and dimensions as shown in Fig. 3. They All tested specimens were dissected to confirm the mode of failure. The speci-
had two longitudinal lugs positioned opposite to each other. The helical lugs were mens were cut using a water-cooled 14 in. diameter diamond tipped blade to a dis-
wound on the bar between the longitudinal lugs such that it started at one longitu- tance of about 1 in. from the bar. Subsequently all the specimens were pried open
dinal lug and ended at the other. The number of lugs in the as-received RL bars was using a chisel and a hammer to reveal the interface. In order to study the progres-
12 per 88.9 mm, i.e. a pitch of 4.4 mm. Their average tensile strength and elastic sion of debond, lug failure, and shearing of sand particles during loading, RLC, SLC,
modulus, as per manufacturer’s data sheet, were 770 MPa, and 42 GPa respectively. TLC, SCC, MC, and SC specimens were loaded to pre-selected load levels below Fp
These as-received RL bars were machined in a lathe to remove the lugs. The pitch of and then unloaded. Subsequently the specimens were dissected to examine the
lugs was altered by removing alternate lugs or two consecutive lugs to result in bars interface.
with six (SL) and three (TL) lugs per 88.9 mm respectively. The pitch of the lugs in SL While the nominal bar diameter, measured using a digital caliper, was used in
and TL bars was 11.95 mm and 26.9 mm respectively. Complete removal of lugs re- calculating the contact surface area for the S, M and SC specimens, Eq. (3) was used
sulted in machined (M) bars. While S bars were used as the reference, SC and M bars to calculate the contact surface area for the RL, SL and TL type specimens.
were used to study the influence surface roughness. RL, SL and TL bars were used to
A ¼ Abar þ ðN  Ahlug Þ þ Allug  ðN  Aov erlap Þ ð3Þ
study the influence of lugs and the pitch of lugs. The air-entrained concrete used in
this study was a proper mixture of water, Portland cement, sand and gravel in the
where Abar is the total surface area without lug = dl, Ahlug is contact surface area for
ratio 1:2.1:4.31:6, as suggested by Portland Cement Association [12].
one helical lug = 2  (p  (D2  d2)/(4  sin a) + (pDa/sin a)), N is number of helical
lug, Allug is contact surface area for longitudinal lugs = 2(2h + w)p, and Aoverlap is the
3.2. Test specimen overlap area between bar and lug = pda/sin a.
Values for D, d, a, h, w, P, a are given in Fig. 3. It was assumed in this study that
The pullout specimen was a concrete cylinder with the FRP bar embedded at its the area over which debonding occured was the same area over which subsequent
center, as shown in Fig. 4. Diameter and height of these specimens were 152.4 mm frictional sliding occured. Hence, the same contact surface area was used in the
and 304.8 mm respectively. Bonding between the bar and the concrete was pre- determination of both the interfacial bond strength and the frictional stress. How-
vented at both the ends of the cylinder using PVC tubes, to eliminate any end effect. ever, this assumption has to be examined in future studies since progressive shear-
The length of this tube at the load end was maintained at 152.4 mm. The length of ing of lugs and sand particles was observed in this study.
this tube at the free end was varied to accommodate bars with various embedment While the interfacial bond strength was determined using the debond load, no
lengths. Preliminary experiments were carried out using RL and M specimens, with attempt was made to determine the pullout strength since it would not be a unique
various embedment lengths given in Table 6, to determine the embedment length value that can be used in any design.
that would yield interfacial failure during subsequent tests. Based on these tests, an
embedment length of 88.9 mm was chosen for subsequent tests. A steel tube was
bonded to the load end of the bar using a room temperature curing epoxy and used 4. Results and discussion
as a gripping system for the GFRP bars.
The specimens were cast using a plastic mold. The bars were held in position at The experimental program undertaken in this study was de-
the center of the mold using a specially designed wooden fixture. The PVC tubes
were bonded to the bars using clay, which was subsequently removed after curing.
signed to measure the debond load (Fd), the pullout load (Fp), and
Concrete was mixed in a concrete mixture, poured into the mold, and packed using the frictional load (Ff). Representative load–slip curves for SC,
a hand-held ram. The molds were covered with a plastic sheet and the specimens MC, SCC, and SLC specimens are shown in Fig. 5. Elastic deforma-
tion of the bar has been deducted from the LVDT measurements
to obtain the relative slip between the bar and concrete. Load ver-
sus slip relationship for the free end is nearly a mirror image of that
for the load end. The load at which the free-end slip started to in-
crease was defined as Fd since the free end of the bar could start to
slip only after the debonding had progressed completely from the
load end to the free end. For all specimens, the load-end slip
started to increase immediately after the start of application of
load while the free-end slip was zero until Fd. Since the free-end
slip started to increase beyond Fd, debonding was believed to be
complete at Fd. All specimens loaded to a load slightly higher than
Fd were unloaded and dissected to confirm the mode of failure.
Clean bar surface without any adhering concrete particles con-
firmed that the failure mode was interfacial [13]. This confirmed
that debonding was complete at Fd.
The maximum measured load was defined as Fp and Ff was de-
fined as the value to which the applied load dropped beyond Fp.
Subsequently, the bearing load was determined using Eq. (2) and
Fig. 2. Four types of polymer composite reinforcing bars used in the present study. experimentally measured values of Fd, Ff and Fp. Thus, the contribu-
 W.H. Soong et al. / Construction and Building Materials 25 (2011) 2813–2821 2817

D =13.94 mm

Helical Lug d = 12 mm

b =0.95 mm
p = 4.4 mm
a = 2.8 mm
w = 2.8 mm
α

α ≈ 76°
Longitudinal Lug h = 1.23 mm

Fig. 3. Shape and dimensions of lugs.

less than Fd. This relative motion would be resisted by the lugs
MTS Actuator and the sand particles bonded to the surface of the bar and hence
the observed increase in load between Fd and Fp is due to frictional
and bearing resistance. Beyond Fp, the applied load suddenly
Load Cell dropped to a minimum value, Ff, in all specimens except SC. Unlike
the schematic shown in Fig. 1 this load drop was accompanied
Grip with simultaneous pullout of the bar. Beyond Ff (Fp for SC), a
stick–slip type of pullout behavior was noticed for all specimens.
This is believed to be due to surface roughness and, in case of bars
with lugs or sand particles, to be due to sheared lugs or sand par-
LVDT
ticles caught between the sliding bar and the concrete. It can be ob-
( Top) served in Fig. 5 that the maximum and minimum loads for MC and
Metal SLC specimens, during the stick and slip behavior, decrease with
plate
increase in slip. This indicates a change in surface characteristics
of the bar and the concrete, and hence, the frictional condition.
Fixed Debonder (6”) An examination of the dissected RLC and SCC specimens, loaded
to Fp, revealed complete shearing of lugs and sand particles as
12” Embeddment Length shown in Figs. 6 and 7 respectively. Since lugs and sand particles
Variable Length are the source of bearing resistance, it can be concluded that the
Debonder bearing resistance did not contribute to the load recorded beyond
LVDT Fp. If there were no frictional resistance, the load would drop to
( Bottom ) zero beyond Fp. Hence, the recorded Ff is a measure of contribution
6” diameter
of frictional resistance to measured Fp. Since SC samples did not
Specimen
have any surface features that would cause mechanical interlock-
ing, bearing resistance was assumed to be zero and hence, the in-
Ground crease in load between Fd and Fp was taken to be due to frictional
resistance to slip.
Fig. 4. A schematic of specimen dimensions and test setup used in the present The measured pullout load (Fp) is plotted in Fig. 8 as a sum of
study.
the debond load (Fd), the frictional load (Ff), and the bearing load
(Fb) for various types of bars. The pullout load increased from SC
to SCC and this is attributed to increase in surface roughness from
tions from chemical bonding, bearing resistance, and frictional SC to SCC. It also increased from TLC to RLC and this is attributed to
resistance to the measured pullout load were delineated in this increase in the number of lugs from TLC to RLC. It is interesting to
study. note that the pullout load for SCC specimens is comparable to that
While the relative movement between the bar and the concrete for RLC specimens.
is possible along the entire embedment length at loads greater Normally, relative sliding of two surfaces at a frictional interface
than Fd, this is possible only along the debonded length at loads would not occur until the interfacial shear force exceeds the fric-

Table 6
Test specimen identification codes.

Bar type Bar diameter (mm) Pitch of lugs (mm) Bar embedment length in pullout specimen (mm) and test specimen codes
127.00 95.25 88.90 63.50 31.75
RL 13.94 4.40 RLA RLB RLC RLD RLE
SL 11.95 – – SLC – –
TL 26.90 – – TLC – –
SC 13.6 – – – SCC – –
M 12.0 – MA MB MC MD ME
S 12.5 – – – SC – –
2818 W.H. Soong et al. / Construction and Building Materials 25 (2011) 2813–2821

70 60

60 Free End SCC Load End

50
F
50 b
40

Load (kN)
Load (kN)

40
30 F
30 F F F
f
d p f
20
20
SLC
10 F
d
10
MC
SC
0 0
RLC SLC TLC SCC MC SC
-30 -20 -10 0 10 20 30
Slip (mm) Fig. 8. Measured pullout load plotted as a sum of measured debond load, measured
frictional load, and calculated bearing load for various specimens.
Fig. 5. Representative load–slip curve for SC, MC. SCC, and SLC specimens.

ble for a frictional interface without any chemical bonding and

mechanical interlocking, is still an unresolved issue.
Sliding of the bar with respect to concrete indicates the transi-
tion from static to dynamic slip. The load corresponding to this
transition may be lower (if this occurs in the partially debonded re-
gion) or above (if this occurs only after complete debonding along
the entire embedment length) Fd. For all specimens, the slope of
the load–slip curves above Fd was significantly lower than the
slope below Fd. This suggests that the slip beyond Fd is dynamic slip
and that the measured Ff corresponds to dynamic friction. Any con-
tribution from static friction may be included in Fd. The Fd and Fp
values measured for SC specimens were 3.59 ± 0.84 kN and
4.26 ± 0.24 kN respectively. The difference of 0.67 kN between
these two average values is taken to be Ff since the surface of the
bar was smooth without any surface features for mechanical inter-
locking. Since this is within the scatter for Fd and Fp, it appears like
Fstatic = Fdynamic and ldynamic = lstatic. Further confirmation of this
using varying confinement pressure is required. It should be noted
Fig. 6. Complete shearing of all lugs along the embedment length of a RLC specimen
loaded to Fp. that ldynamic would vary for different bars.

4.1. Contribution from chemical bonding

Since debonding is complete at Fd, it would be expected to be a

measure of chemical bonding and the interfacial bond strength (sd)
determined using this would be expected to be same for all speci-
mens. However, it varied from a minimum of 1.01 MPa for SC spec-
imen to a maximum of 5.68 MPa for SCC specimen as shown in
Fig. 9. Possible reasons for this variation are:

(a) Underestimation of contact surface area, especially in SCC

and MC samples, since calculation using nominal bar diam-
eter would have excluded the increase in surface area due to
surface roughness.
(b) Contribution from the bearing resistance, especially in SCC
and LC samples, since measured Fd corresponds to debond
completion rather than debond onset. The increase in Fd
with increase in the number of lugs in LC (i.e. RLC/TLC/SLC)
Fig. 7. Complete shearing of sand particles along the embedment length of a SCC specimens (Fig. 8) appears to support this reasoning.
specimen loaded to Fp. (c) Contribution from static friction as discussed above. How-
ever, contribution from the dynamic friction to Fd is believed
tional resistance for sliding. This is referred to as static friction and to be minimal because the magnitude of load-end slip
the maximum force (Fstatic) for initiation of the sliding is related to recorded at Fd is very low. This is supported by the observa-
normal force (N) acting on the interface through static friction tion of clean bar surface without any scoring marks during
coefficient, lstatic(=Fstatic/N). The force (Fdynamic = Ff) required to post-mortem examination of the specimens loaded to Fd.
maintain the sliding of the interfaces may be equal to or less than
(Fstatic). In case of the latter, ldynamic < lstatic, which has been noted Nevertheless, the value of 1.01 MPa obtained for SC specimens
by numerous published literature. Whether this would be applica- would be the maximum limit to the interfacial strength since the
 W.H. Soong et al. / Construction and Building Materials 25 (2011) 2813–2821 2819

6 pitch of lugs. Contribution from frictional resistance (hence ldy-

namic) for LC specimens was higher than that for the SC specimens

5 but lower than that of MC specimens. This is believed to be due

Interfacial Strength (MPa)

to (a) longitudinal lugs with relatively higher surface roughness,

4
(b) rough bar surface in locations where lugs were removed to alter
the pitch, and (c) the resistance offered by the sheared lugs caught
between the bar and concrete.
3
Since complete shearing of sand particles from the bar’s surface
is observed in Fig. 7, it can be concluded that Ff recorded for SCC
2 specimens is not representative of the frictional resistance encoun-
tered during loading to Fp. In addition, examination of dissected
1 SCC specimens loaded to load levels between Fd and Fp revealed
progressive shearing of the sand particles with increase in load
0 [13]. This suggests that the frictional stress changed continually
RLC SLC TLC SCC MC SC during loading and the values in Fig. 10 represent the state of fric-
Fig. 9. Interfacial bond strength for various specimens.
tion after the load drop at Fp. Future studies should focus on char-
acterizing this change in frictional stress during loading. Since the
change in friction stress during sliding and pullout was not charac-
bearing resistance is zero and the frictional resistance is negligible. terized, error in the values of Ff and Fb is unknown. Similar conclu-
This value is comparable to the published values. Since lateral sion is valid for LC specimens since progressive shearing of lugs
pressure was not determined, its contribution to Fd is unknown. was observed with increase in load.

4.3. Contribution from bearing resistance

4.2. Contribution from frictional resistance
The bearing resistance is negligible in SC and MC specimens due
Frictional load varied with type of bar as shown in Fig. 8. Lateral
to lack of geometrical features on the bar surface that can cause
pressure was same for all specimens. In order to eliminate the ef-
mechanical interlocking of the bar with the concrete. However,
fect of difference in contact area, these frictional loads are normal-
the bearing resistance is substantial in SCC and LC specimens. It
ized with contact surface area and plotted in Fig. 10 as frictional
is interesting to note (Fig. 8) that the bearing resistance due to
stresses (sf) for various specimens. Comments made in Section 4.1
small sand particles in SCC specimens is comparable to the bearing
regarding the error in the calculation of the contact surface area for
resistance due to the lugs in SLC and TLC samples.
MC and SCC specimens are also applicable here.
The bearing resistance increased from SLC to RLC due to de-
It can be inferred from Fig. 10 that the frictional stress is negli-
crease in the pitch of lugs (i.e. increase in the number of helical
gible for SC specimens. This is to be expected since the surface of
lugs). In order to quantify the influence of the pitch and the num-
bar in SC specimens was smooth. However, the MC specimens
ber of helical lugs on the bearing resistance, a number of LC spec-
had relatively rougher bar surface after removal of lugs in a lathe.
imens were unloaded after loading to specific load levels between
This resulted in a higher ldynamic and frictional stress in MC
Fd and Fp, and were dissected and examined to determine the num-
specimens when compared to SC specimens. SCC specimens with
ber of sheared lugs. It can be inferred from Fig. 6 that all the lugs in
sand-coated bars recorded the maximum frictional stress and this
the LC specimens sheared at Fp. However, at loads less than Fp few-
is believed to be due to higher ldynamic for the sand–concrete
er helical lugs were sheared as shown in Fig. 11 for a RLC specimen
interface when compared to lower ldynamic for the relatively softer
loaded to 0.85Fp. At this load, only two lugs near the load end were
polymer composite–concrete interface. Since the SCC specimens
sheared. The sheared lugs appear milky-white while the intact lugs
had higher contact area due to higher surface roughness, when
have the same color as the bar. It is also interesting to note that the
compared to SC and MC specimens, any error in the contact area
very first lug at the load end was not sheared. This was noticed in
used in determining sf could also be another contributing factor
all specimens with bars with lugs. Experimental studies [1,4] have
to the observed difference. The frictional stress values for LC spec-
shown that the maximum interfacial shear stress occurs just below
imens were same. This is to be expected since ldynamic and the sur-
face roughness in these specimens would not change with the

5
Frictional Stress (MPa)

0
RLC SLC TLC SCC MC SC
Fig. 11. Partial shearing of lugs along the embedment length of a RLC specimen
Fig. 10. Frictional stress for various specimens. loaded to 0.85Fp.
2820 W.H. Soong et al. / Construction and Building Materials 25 (2011) 2813–2821

80 the error band for the slope (2.59 kN) obtained using experimental
Load = 2.59* (number of lugs) +28.04 results plotted in Fig. 12. However, no failure of concrete was ob-
70
served between the lugs in the LC samples tested in this study. Fur-
60 ther investigation on the stress distribution along the embedment
length and near the lugs is required to resolve this.
50 Finally, the extrapolated value for Fp, for the case of no lug, is
Load (kN)

28 kN which is comparable to the value of 23.56 ± 3.21 kN ob-

40
tained for MC specimens. The average of Fd for RLC, SLC and TLC
30 is 18.22 ± 3.1, which is indicated as average debond load in
Fig. 12. The difference between Fp and Fd, for the case of no lug,
20 MC Specimen
should be due to the contribution from the frictional resistance,
Average Debond Load = 18.22 kN and the bearing resistance due to surface roughness. However,
10
the latter can be neglected based on the results for MC specimens
0 in Fig. 8. Hence, the difference of about 10 ± 3.1 kN should be due
0 2 4 6 8 10 12
to frictional contribution and it is comparable to frictional values
Number of Lugs plotted in Figs. 8 and 9 for LC specimens, a minimum of 6.3 kN
for TLC to a maximum of 11 kN for RLC. Thus, the data in Fig. 12
Fig. 12. Progression of lug failure in LC specimens with increase in applied load.
independently support the research approach used in this study.
The bearing resistance also increased from SC to SCC due to in-
the concrete surface. Hence, it is to be expected that the first lug creased surface roughness. When the sand particles were com-
near the concrete surface would not shear. pletely sheared from the bar surface at Fp, as shown in Fig. 7, the
The applied load versus the number of helical lugs sheared at nominal diameter of the bar decreased from 13.6 mm to
that load is shown in Fig. 12, for all specimens with varying num- 12.25 mm. Hence, the maximum bearing resistance in SCC speci-
ber of lugs and pitch. Since the number of sheared lugs was deter- mens is limited by the shear strength of the bond between the sand
mined by dissecting the specimens unloaded from a load level, the particles and the bar. Assuming that the entire bar surface is cov-
load values may or may not correspond to the actual load values at ered with sand particles, bond strength is estimated to be
which the lugs sheared. Hence, the trend suggested by the data in 3.95 MPa (13,520N/(3.14  12.25 mm  88.9 mm). Thus, if the con-
Fig. 12 is emphasized here rather than the absolute values. A linear tact surface area between the bar and the sand particles is known,
increase in the applied load with increase in the number of sheared the bearing resistance due to the sand particles may be predicted.
lugs confirms that the increase in load between Fd and Fp is mainly
dependent on the bearing resistance offered by each lug and inde-
4.4. Effect of loading rate
pendent of the pitch of lugs. The average shearing load per lug, gi-
ven by the slope of the linear-fit line in Fig. 12, is 2.59 kN. Using
The pullout load for SCC specimens increased with increase in
this value, the average shear strength (i.e. shear strength of the
loading rate as shown in Fig. 13. However, such a clear trend was
bond between the lug and the bar) of each lug is determined
not observed in RLD specimens. This is believed to be due to higher
(2.59 kN/pda) to be 24.55 MPa. This value is similar to that deter-
contribution from frictional resistance in SCC specimens when
mined by Al-Zahrani [1]. Hence, the bearing resistance can be
compared to RLD specimens. Further investigation is required to
determined using the following equation
clearly understand the relationship between the loading rate, and
F b ¼ sl Al N l ð4Þ the frictional resistance and the bearing resistance. Nevertheless
these preliminary results highlight the need to control the loading
where sl is the shear strength of a helical lug, Al is the contact sur-
rate while evaluating the interfacial bond strength using the pull-
face area of a lug (pda), and Nl is the number of lugs per embedment
out test.
length of the bar, which depends on the pitch. Theoretically, it is
In summary, the approach used in this study has clearly delin-
possible to increase the bearing resistance and Fp by increasing Al
eated the contributions from various mechanisms to the interfacial
and/or Nl provided the concrete between the lugs can withstand
bond strength and the pullout load (i.e. the resistance to bar pull-
the compressive and shear stresses introduced in it by the applied
out). The results of this study have also demonstrated the depen-
load. This is supported by the results of Al-Zahrani et al. [3], who
observed a change in failure mode from shearing of lug to fracture
of concrete between the lugs when the width of the lug was in- 70
creased. Hence, the maximum bearing stress contribution from
one lug can be determined from Eq. (5a), and either (5b), (5c) 60 SCC
depending on the lowest Fmax.
50
Pullout Load (kN)

RLD
F b per helical lug ¼ F max in the concrete ð5aÞ
40
where
2 30
F max ¼ rc Ac = sin a for compressive failure ð5bÞ

20
F max ¼ sc Ac =ðsin a cos aÞ for shear failure ð5cÞ

rc is the compressive strength of the concrete, sc is the shear 10

strength of the concrete, and Ac is the area of the concrete between
the lugs (p(D2  d2)/4). For the lug shape and dimensions shown in 0
0.26 1.3 6.5
Fig. 3, Fmax would be minimum for compressive failure even if the Loading Rate (mm/ min)
shear strength is assumed to be one-half of the concrete compres-
sive strength of 49.77 MPa. The calculated Fmax of 2.09 kN is within Fig. 13. Influence of loading rate on pullout load.
 W.H. Soong et al. / Construction and Building Materials 25 (2011) 2813–2821 2821

dence of these contributions on parameters such as bar surface Acknowledgements

geometry, test conditions, and concrete strength. Since these
parameters varied from researcher to researcher in the past, the re- The authors would like to acknowledge the financial support for
ported strength corresponding to the pullout load also varied this project from NSERC (Natural Sciences and Engineering Re-
widely. Hence, instead of using this strength corresponding to pull- search Council of Canada), the University of Manitoba, and material
out load for design as suggested in the reviewed literature, the ap- and testing support from ISIS (Canadian Network of Centers of
proach used in this study is suggested. Eq. (2) can be modified, for Excellence for Intelligent Sensing for Innovative Structures). In
bars with lugs, as addition, the authors would like to acknowledge the technical
F p ¼ sd Ad þ sl Al Nl þ ldynamic N ð6Þ assistance of Mr. Moray McVey of the University of Manitoba.

where Ad is the contact surface area between the concrete and the References
bar. Other parameters have been defined before. The maximum lim-
it for Fp is the force required to fracture the bar and this can be cal- [1] Al-Zahrani MM. Bond behavior of fiber reinforced plastic reinforcements with
concrete. PhD thesis, University Park, Pennsylvania, Pennsylvania State
culated using the tensile strength and the cross sectional area of the University, USA; 1995.
bar. The areas, Ad, and Al are functions of bar embedment length and [2] Benmokrane B, Maceachern M, That TT, Zhang B. Evaluation of bond
bar diameter. If the lug dimensions and shape, bar diameter, shear characteristics of GFRP rebars embedded in polymer and normal cement
concrete. Technical report no. 7, University of Sherbrooke; 1999.
strengths, N, and ldynamic in Eq. (6) are known, the critical embed- [3] Al-Zharani M, Mesfer M, Al-Dulaijian SU, Nanni A, Bakis CE, Boothby TE.
ment length (i.e. development length) that would yield maximum Evaluation of bond using FRP bars with axisymmetric deformations. Constr
Fp can be predicted using Eq. (6). Alternatively, the dimension and Build Mater 1999;13(6):299–309.
[4] Benmokrane B, Tighiouart B, Chaallal O. Bond strength and load distributions
shape of lugs wrapped on a bar, which would result in maximum of composite GFRP reinforcing bars in concrete. ACI Mater J 1996(May–
pullout load, can be designed for a chosen embedment length. June):246–53.
[5] Tighiouart B, Benmokrane B, Gao D. Investigation on the bond of fiber
reinforced polymer (FRP) rebars in concrete. In: Saadatmanesh H, Eshani MR,
5. Conclusions editors. Proceedings of the second international conference on composites in
infrastructure, Tucson, Arizona, USA; 1998. p. 102–112.
Debonding load, corresponding to debond completion, is much [6] Leung CKY, Geng Y. Effect of lateral stress on the debonding and pullout of steel
fibers in a cementitious matrix. In: Buyukozturk O, Wecharatana M, editors.
lower than the pullout load, corresponding to onset of sliding and Proceedings of the 1993 ACI spring convention, ACI Special Publication, SP-
pullout of the bar along the entire embedment length. While the 156; 1995. p. 153–72.
former is mainly a measure of chemical bonding, the latter is a [7] Zenon Achillides, Kypros Pilakoutas. Bond behavior of fiber reinforced polymer
bars under direct pullout conditions. J Compos Constr 2004(March/
sum of contributions from chemical bonding, frictional resistance, April):173–81.
and bearing resistance. The maximum value for the interface bond [8] Okelo Roman, Yuan Robert L. Bond strength of fiber reinforced polymer rebars
strength for the concrete–bar specimen used in this study is in normal strength concrete. J Compos Constr 2005(May/June):203–13.
[9] Tastani SP, Pantazopoulou SJ. Bond of GFRP bars in concrete: experimental
1.01 MPa. The resistance from lugs to bar pullout is comparable study and analytical interpretation. J Compos Constr 2006(September/
to that from sand particles bonded to the bar. Bearing resistance October):381–91.
due to lugs is a function of shear strength of the bar-lug interface, [10] Baena Marta, Torres Lluis, Turon Albert, Barris Cristina. Experimental study of
bond behavior between concrete and FRP bars using a pullout test. Composites
lug dimensions, pitch of the lug, and number of lugs, and is limited
Part B 2009;40:784–97.
by the concrete’s compressive or shear strength. Bearing resistance [11] Esfandeh Masoud, Sabet Ali R, Rezadoust Amir M, Alavi Mohammed B. Bond
due to sand particles is a function of surface roughness amplitude performance of FRP rebars with various surface deformations in reinforced
and is limited by the shear strength of the bond between the sand concrete. Polym Compos 2009:576–82.
[12] Design and Control of Concrete Mixtures. 11th ed. IL, USA: Portland Cement
particles and the bar. Frictional resistance is a function of surface Association.
roughness and it may vary during loading due to progressive [13] Wai How Soong. Bonding between the concrete and fiber reinforced plastic
shearing of lugs or sand particles. Loading rate influences the pull- (FRP) rods. MSc thesis, Winnipeg, Manitoba, University of Manitoba, Canada;
2001.
out load.