Construction and Building Materials 25 (2011) 1026–1036

Contents lists available at ScienceDirect

Construction and Building Materials

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

Analytical prediction of transfer length in prestressed self-consolidating

concrete girders using pull-out test results
Andrew Pozolo, Bassem Andrawes *
Department of Civil and Environmental Engineering, University of Illinois at Urbana-Champaign, Urbana, IL 61801, United States

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

Article history: While self-consolidating concrete (SCC) is comparable to conventional concrete (CC) in terms of strength,
Received 11 November 2009 the comparability of SCC’s bond to steel is less well-defined. A keen understanding of SCC’s bond strength
Received in revised form 1 June 2010 is essential to advance SCC within the prestressed concrete industry. This study presents an analytical
Accepted 19 June 2010
method for predicting the transfer length of steel strands in prestressed girders using pull-out test
Available online 23 July 2010
results. The experimental data from a series of 56 pull-out tests is utilized to derive bond stress–slip rela-
tionships for 12.7 mm steel strands embedded in SCC and CC. Modification factors are used to correlate
Keywords:
pullout bond stresses to transfer bond stresses in prestressed members, and the modified relationships
Self-consolidating concrete
Pull-out test
are integrated in three-dimensional finite element models to predict transfer lengths in prestressed
Transfer length SCC girders. The analytical predictions correlate well with experimental results and transfer length
Finite element analysis requirements of current US design codes.
Ó 2010 Elsevier Ltd. All rights reserved.

1. Introduction
In the early 1980’s, the number of skilled workers in Japan’s
construction industry had fallen to a level which spurred concerns
over the quality of the country’s concrete infrastructure [1]. To im-
prove concrete durability without the need for skilled labor,
researchers developed a high performance concrete which would
compact into formwork via its own weight. Today, self-consolidat-
ing concrete (SCC) has emerged as a viable alternative to conven-
tional concrete (CC) in many structural applications, particularly
those which require dense reinforcement or complex geometry.
Its unique workability and propensity to reduce construction time
and cost have made SCC an intriguing material to the international
design community.
Inconclusive research on SCC behavior in prestressed members
has thus far limited the technology’s impact on the United States’
prestressed concrete industry. A keen understanding of SCC’s bond
strength, including its impact on transfer length in prestressed
members, is essential to safely incorporate SCC in modern applica-
tions. To foster this understanding, several American universities
and State Departments of Transportation (DOTs) have recently
sponsored projects to analyze the bond characteristics and transfer
length of prestressing strands in SCC girders. Select data from these
studies is presented herein, though a comprehensive summary
may be found in a synthesis review executed by Andrawes et al.
in 2009 [2]. As evidenced in the review, results are somewhat
* Corresponding author. Tel.: +1 217 244 4178; fax: +1 217 265 8039.
E-mail address: andrawes@illinois.edu (B. Andrawes).
inconclusive. Studies by Ruiz et al. [3], Larson et al. [4], Hamilton
and Labonte [5], and Trent [6] showed transfer lengths of strands
in SCC to meet code provisions stipulated by both the American
Concrete Institute (ACI) [7] and the American Association of State
Highway Transportation Officials (AASHTO) [8]. Naito et al. [9]
noted that strands in SCC met code requirements, though they
did not satisfy minimum pullout loads suggested by Logan [10].
In contrast, studies by Zia et al. [11] and Girgis and Tuan [12]
showed inadequate transfer lengths in SCC when compared to code
provisions. Haq [13] and several of the aforementioned researchers
observed via pull-out tests lower bond strength in SCC than in CC.
Variations in mix constituents, strand types, and specimen types
throughout the studies provide no constant by which to compare
results. Furthermore, the large-scale nature of each study does
not encourage iterative testing to eliminate inconsistencies. Thus,
to augment previous research and explore the application of SCC
in highway bridges, the U.S. Illinois DOT (IDOT) has sponsored its
own study comprising, in part, the aforesaid synthesis review
and the contents of this paper.
The paper at hand first examines results from eight large-scale
projects that have published studies on SCC bond behavior per-
taining to transfer length in prestressed members. The governing
parameters in the studies are identified. The paper then utilizes
experimental data from eight sets of pull-out tests conducted at
the University of Illinois at Urbana-Champaign (UIUC) to derive
bond stress–slip relationships for 12.7 mm steel strand embedded
in SCC and CC. Modification factors are applied and the relation-
ships are incorporated in finite element (FE) analyses to predict
the transfer lengths of strands within a prestressed T-beam

0950-0618/$ - see front matter Ó 2010 Elsevier Ltd. All rights reserved.
doi:10.1016/j.conbuildmat.2010.06.076
 A. Pozolo, B. Andrawes / Construction and Building Materials 25 (2011) 1026–1036 1027

experimentally tested by Haq [13] and a box girder tested at
UIUC. The predicted transfer lengths are compared to experimen-
tal results and current U.S. design standards.
2. Previous research synopsis
Eight large-scale projects within the United States have pub-
lished studies on SCC bond behavior as it pertains to transfer
length within prestressed members. Developing universal bond
criteria from these studies is difficult given their range of test vari-
ables. Select data has been compiled in this article to provide brev-
ity in contrasting previous research. Table 1 lists the constituents
of the studies’ SCC mixes and of an IDOT-approved SCC. The table
shows high variability in the amount of air-entraining agent
(AEA) and viscosity-modifying admixture (VMA) used in the SCCs.
Of the 14 mixes, nine contained an AEA ranging from 70 to 861 mL/
m3 and half incorporated a VMA ranging from 482 to 5268 mL/m3.
The amount of VMA is particularly significant because it may ad-
versely affect bond [12]. Several mixes included supplementary
cementitious materials; both Hamilton and Labonte [5] and Girgis
and Tuan [12] tested mixes containing fly ash, while Naito et al. [9]
tested SCC containing blast furnace slag. External literature has
documented the effect of these additives on plastic mix properties,
though their impact on bond in hardened concrete remains unclear
[14,15].
In the United States, each state is responsible for developing its
own SCC mix guidelines, introducing myriad variables into SCC
bond data. Using results from earlier studies to predict bond ade-
quacy of a proposed SCC mix may be unsound when the studies’
mixes do not comply with standards applicable to the proposed
SCC. When compared to IDOT Bureau of Materials & Physical Re-
search provisions for precast SCC, nine mixes in Table 1 exceed
the maximum cement factor, eight exceed the limit for fine aggre-
gate proportions, and three have water/cement ratios outside the
current allowable range [16]. Noncompliance with IDOT standards
prevents these mixes from serving as a basis for bond criteria in
new Illinois SCC.
Table 2 contains, where available, the specimen types, concrete
strengths, bond strengths, and transfer lengths reported in the
eight aforementioned studies. Four studies entailed modified
Moustafa pull-out tests to qualify the bond characteristics of
strand in SCC as satisfactory when compared to the same strand
in CC [17]. Columns six and seven of Table 2 list the absolute and
normalized bond strengths as explicitly reported in the literature
or as derived from available data. When bond strength was calcu-
lated using pull-out test data, the pullout load was assumed to be
uniformly distributed along the embedded strands. The eighth col-
umn in Table 2 contains the experimental transfer lengths (Lt,test)
from each study, and the final column compares them to AASHTO
requirements (Lt,code), or a value 60 times the strand diameter [8].
In three studies, the ratio Lt,test/Lt,code for at least one specimen ser-
ies exceeded unity, indicating insufficient bond between strand
and SCC. Physical test variables not listed in Table 2 may have also
impacted test results and were not constant throughout the stud-
ies. For example, each study tested a different specimen type con-
taining a unique number of strands, not all of which were bottom
reinforcement; transfer lengths are known to vary between top
and bottom strands [4].
After reviewing the previous studies, it could be concluded that
although determining transfer length experimentally may be the
most rigorous method for assessing bond strength of strands in
SCC, the large number of material and geometric variables involved
would make such a method impractical in terms of time and cost.
Hence, a simple yet accurate analytical approach is needed to pre-
dict transfer length. The following sections propose a systematic
way to predict transfer lengths in prestressed members using

Table 1
SCC mixes from eight US studies and typical illinois mix.

Reference [4] [6] [13]

Larson et al. Trent Haq
Material Units SCC S1CCM S1CCM2 SCC1 SCC2A SCC2B SCC3
Cement kg/m3 444a 444a 441a 444a 415 415 415
Coarse agg. kg/m3 806 963 978 876 818 818 850
Fine agg. kg/m3 889 794 775 964 845 845 755
Fly ash kg/m3 – – – – – – –
Fine agg./total agg. 0.52b 0.45 0.44 0.52b 0.51b 0.51b 0.47
Water L/m3 134 168 168 153 163 163 183
W/C ratio 0.30c 0.38 0.38 0.35 0.39 0.39 0.44
AEA mL/m3 193 – – 508 203 474 861
HRWR mL/m3 2708 – – 3751 3950 3257 4162
VMA mL/m3 – – – – 1893 482 4162
Set retardant mL/m3 – – – 20,307 – 15,858 12,636
Reference: [5] [9] [3] [12] [11]
Hamilton et al. Naito et al. Ruiz et al. Girgis & Tuan Zia et al. Illinois
Material Units SCC SCC SCC Mix #1 Mix #2 SCC SCC
Cement kg/m3 446a 503a,d 563a 474a 374 480a 391
Coarse agg. kg/m3 774 978 800 760 777 788 917
Fine agg. kg/m3 838 760 873 840 859 770 857
Fly ash kg/m3 100 – – 89 59 – –
Fine agg./total agg. 0.52b 0.44 0.52b 0.53b 0.53b 0.49 0.48
Water L/m3 153 161 168 173 173 203 149
W/C ratio 0.34 0.32 0.30c 0.37 0.46c 0.42 0.38
AEA mL/m3 70 77 – – – 94 735
HRWR mL/m3 2491 5268 4042 542 542 3133 3094
VMA mL/m3 – 619 735 387 387 – –
Set retardant mL/m3 534 – – 193 193 1253 –
a
Exceeds maximum cement factor of 418 kg/m3.
b
Water/cement ratio falls outside range of 0.32–0.44.
c
Fine aggregate proportion exceeds maximum of 50% of total aggregate by weight.
d
Mix also contains 25 kg/m3 of slag.
1028 A. Pozolo, B. Andrawes / Construction and Building Materials 25 (2011) 1026–1036

Table 2
Pertinent data from eight US studies (see Table 1 for reference numbers).
pffiffiffiffi p pffiffiffiffi p
Specimen type Strength (MPa) Peak load (kN) Peak load fc0 (m N) Bond strength, U (MPa) U fc0 ( MPa) Lt,test (m) Lt,test Lt,code

Larson et al.
SCC1 Top strand beam 51.8 – – – – 0.53 0.70
SCC1 Bot. strand beam 48.1 – – – – 0.71 0.93
SCC1 T-beam 54.9 – – – – 0.69 0.90
CC Pullout 38.6 176 28.3 9.6 1.54 – –
Trent
S1CCM 152 mm  152 mm – – – – – 0.71 0.83
S1CCM2 152 mm  152 mm – – – – – 0.66 0.77
CC 152 mm  152 mm – – – – – 0.84 0.98
Haq
SCC1 T-beam/pullout 53.0 74 10.1 3.0 0.42 0.76 0.99
SCC2A T-beam/pullout 53.0 116 15.9 4.8 0.65 0.69 0.90
SCC2B T-beam/pullout 46.2 86 12.6 3.5 0.52 0.76 0.99
SCC3 T-beam/pullout 46.2 89 13.1 3.7 0.54 0.76 1.00
CC T-beam/pullout 38.2 135 21.9 5.6 0.90 0.5 0.66
Hamilton et al.
SCC AASHTO Type II 58.6 – – – – 0.36 0.47
CC AASHTO Type II 58.6 – – – – 0.35 0.46
Naito et al.
Logan CC Pullout 27.6 140 26.7 7.3 1.39 – –
SCC Bulb Tee – – – – – 0.4 0.52
HESC Bulb Tee – – – – – 0.4 0.53
Ruiz et al.
SCC 165 mm  305 mm 75.8 – – – – 0.51 0.56
Girgis & Tuan
Mix 1 (SCC) NU1100 I-beam 53.0 193 26.5 – – 0.91 1.00
Mix 2 (SCC) NU900 I-beam 55.2 292 39.4 9.0 1.21 1.07 1.17
Mix 3 (CC) NU1350 I-beam 48.1 214 30.8 – – 0.51 0.56
Zia et al.
SCC AASHTO Type III 38.3 – – – – 0.82 1.08
SCC AASHTO Type III 37.6 – – – – 1.27 1.67
CC AASHTO Type III 32.4 – – – – 1.12 1.47

pull-out test data and finite element analysis. The proposed meth- Table 3
od may be particularly useful when large-scale testing is not Batch constituents for CC and SCC mixes.

feasible. Mix constituents Units CC SCC

Type III cement kg/m3 397 392
Coarse aggregate kg/m3 1096 952
3. Experimental tests Fine aggregate kg/m3 699 853
Air-entraining agent mL/m3 1276 542
3.1. Pull-out test description High-range water reducer mL/m3 1741 3133
Water L/m3 132 109
The US Precast/Prestressed Concrete Institute (PCI) recommends the Moustafa W/C ratio – 0.33 0.28
pull-out test to qualify bond characteristics of strands embedded in concrete as sat- Coarse/fine aggregate ratio – 1.57 1.12
isfactory [18]. Moustafa first performed pull-out tests on lifting loop strands in Fine/total aggregate ratio – 0.39 0.47
1974, considering 9.5 mm, 11.1 mm, and 12.7 mm diameter strands with embed-
ment lengths between 305 and 762 mm [17]. Each strand was loaded by hydraulic
jacks, and the relative displacement between concrete and strand was monitored
Table 4
throughout loading until failure occurred. Since no bond quality standard exists
Experimental plastic SCC and CC test results and typical SCC test values.
for prestressing strands, Moustafa’s simple method was adopted with slight modi-
fication in subsequent studies conducted by Rose and Russell [19] and Logan [10] to SCC mix properties Units SCC, Exp. SCC, Typ.
study bond performance of strands intended for pretensioned applications. For con-
Slump flow mm 560 560–760
sistency, the latter studies utilized 12.7 mm diameter strands embedded 457 mm in
J-ring value mm 50.8 <50
large block specimens.
L-box value % 75 >75
The PCI [18] directly refers to Logan [10] for a detailed outline of the Moustafa
VSI – 0–1 <2
method, which served as the basis for this study’s experimental procedure. The
pull-out test is attractive given its simplicity and relatively low cost. In this study,
a total of 56 pull-out tests were performed on 12.7 mm seven-wire low-relaxation Air entrainment % 5.7
strands embedded in SCC and CC blocks. Concrete compressive strengths and the CC mix properties Units CC, Exp.
strands’ force-slip responses were recorded at curing ages of 1, 3, 7, and 28 days;
Slump mm 177.8
this data was utilized to derive bond stress–slip relationships for SCC and CC.
Air entrainment % 6.3
Four pullout block specimens and thirty-two 152 mm  305 mm cylinder spec-
imens were cast simultaneously. Half of the specimens used SCC and the other half
used CC. The batch constituents for the SCC and CC are given in Table 3. Both mixes
used Type III cement, a coarse aggregate with maximum 12.7 mm nominal diame-
ter, and a natural sand fine aggregate. Use of a high-range water reducer (HRWR) ability, moderate filling ability, and little but acceptable segregation of the mix
and an air-entraining agent (AEA) ensured proper workability for each mix; no vis- [20]. Table 4 also shows fresh CC properties and the entrained air for both concretes.
cosity-modifying admixture (VMA) was used. Standard slump flow, J-ring, L-box, Each pullout block was 610 mm  610 mm  1473 mm and contained fourteen
and visual stability index tests were conducted on the fresh SCC. The results are 12.7 mm diameter, 1860 MPa tensile strength strands with 457 mm embedment.
compared to common industry standards in Table 4 and show moderate passing All block dimensions including strand spacing, longitudinal reinforcement, and
 A. Pozolo, B. Andrawes / Construction and Building Materials 25 (2011) 1026–1036
clear cover are shown in Fig. 1. Concrete was cast outdoors with temperatures
above the IDOT minimum of 3.8 °C. Space heaters were placed around the speci-
mens and the casting bed was covered to ensure adequate overnight curing
conditions.
The servo-controlled assembly in Fig. 2 was utilized to apply load to strands in
the pull-out tests. A linear variable differential transducer (LVDT) attached to the
cylinder monitored displacement of an aluminum plate secured to the top of the
piston. Strands were loaded at a constant displacement-controlled rate of
10.16 mm/min, resulting in loading rates below the maximum 88.9 kN/min as set
forth by Logan [10]. Load was applied continuously until strands were completely
pulled out or fractured. Five criteria were obtained for each pull-out test: (1) First
slip load, (2) Peak pullout load, (3) Displacement at first slip, (4) Displacement at
peak load, and (5) Depiction of failure. The former four criteria were taken from
the force-slip response for each strand, while the latter criterion was made by visual
observation.
3.2. Pull-out test results
Fourteen pull-out tests and eight uniaxial compression tests were conducted
each day; half were on SCC specimens and half were on CC specimens. Fig. 3 pre-
sents the force-slip response of seven strands embedded in the SCC block tested
3 days after concrete placement. Typical of all the results regardless of concrete
age or type, the figure shows force-slip responses characterized by an initial linear
region followed by a subsequent nonlinear region. The linear behavior was ob-
served while strand remained fully bonded to concrete and the steel deformed elas-
tically. Localized bond failure then brought about a significant reduction in pullout
stiffness at the point of first slip, resulting in nonlinear behavior. Progressive bond
failure caused a gradual reduction in pullout stiffness until reaching peak pullout
capacity, after which load resistance declined and ultimate bond failure occurred.
When comparing bond performance in different concretes, normalization tech-
niques should be considered. Numerous studies on bond behavior have shown a
correlation between bond strength and ðfc0 Þ1=2 , where fc0 is concrete strength ob-
tained from cylinder compression tests [21–22]. ACI provisions, moreover, state
that development lengths of reinforcing bars are inversely proportional to ðfc0 Þ1=2 ,
which implies a linear relationship between bond strength and ðfc0 Þ1=2 [7]. However,
methods of normalization vary among studies depending upon the concrete
strength range, confinement within specimens, and country in which testing occurs.
Previous research and most European design codes consider bond strength to vary
proportionally with ðfc0 Þ1=3 [23]. Other studies have found a more precise correlation
between bond strength and ðfc0 Þ1=4 [24]. Based on similarities to the eight studies
discussed in Section 2, the current study assumes a linear relationship between
bond strength and ðfc0 Þ1=2 . This article presents both absolute and normalized exper-
imental results.
152 305 152 152 229
51 mm
Foam Cube
#10 Stirrups
(U.S. #3)
#13 Longitudinal Bars
(U.S. #4)
51 mm Clear Cover
(all sides)
Fig. 1. Dimensions (mm), strand placement, and reinforcement of pullout block.
1029
The compressive strengths of SCC and CC at each testing age are shown in Fig. 4.
Both concretes achieved adequate strength for initial tests and strengthened over
time, reaching 35 MPa after 28 days. To better assess the bond behavior of SCC
and CC using a common datum, the effect of normalization on first slip and peak
pullout loads is highlighted between Figs. 5 and 6. Fig. 5 shows the average absolute
first slip and peak pullout loads at all ages with an assumed error of plus or minus
one standard deviation. Pullout loads increased over time, ranging in SCC from 161
to 171 kN and in CC from pffiffiffi135
ffi to 161 kN. Fig. 6 shows first slip and peak pullout
loads normalized using fc0 . Normalized first slip loads differed between SCC and
CC by an average of 10% for all tests. Only the 1-day tests showed first slip normal-
ized loads in SCC higher than those in CC. Normalized pullout loads differed be-
tween SCC and CC by as much as 25% at 1 day and as little as 1% at 3 days. Only
the 7-days tests produced lower normalized pullout loads in SCC than in CC, an
anomaly explained by the high SCC compressive strength recorded at that age.
3.3. Pullout bond stress–slip relationships
Idealized responses were developed for each set of seven pull-out tests by cal-
culating the average load at four critical locations corresponding to the points of
first slip, peak pullout, ultimate failure, and midpoint between first slip and peak
pullout. The load at each point was correlated to bond stress by assuming the entire
embedded strand uniformly resisted pullout load. The formation of an idealized
pullout bond stress–slip relationship is illustrated in Fig. 7 for 3 days SCC. Slip
was calibrated using steel constitutive properties to eliminate the effect of strand
elongation.
Fig. 8 displays the idealized relationships for both concrete types at all curing
ages. Until the first slip occurred, SCC and CC relationships were extremely compa-
rable. The peak normalized bond stress of SCC was 25% greater than that of CC 1 day
after concrete placement. After 7 days the converse held true; peak normalized
bond stress of CC was approximately 25% greater than that of SCC. The differences
between the concretes’ peak bond stresses at 3 and 28 days were negligible. From
the figure’s stress–slip relationships and comparable results from analogous re-
search, it was concluded that the strands in this study displayed sufficient bond
to SCC; therefore, they were utilized in the project’s subsequent phase in which a
full-scale girder was cast for transfer length tests.
3.4. Transfer length test description
Following the pull-out tests, an 8.53-m long prestressed hollow box girder was
cast for transfer length testing. The girder was cast with the same SCC mix as in the
pullout blocks and had cross-sectional properties shown in Fig. 9. The girder’s 16
uncoated 12.7 mm, 1860 MPa strands were prestressed to 1396 MPa. Strands were
229 229 229 229 229 152
457 559 457
1030 A. Pozolo, B. Andrawes / Construction and Building Materials 25 (2011) 1026–1036

Steel Strand 127

Prestressing Chuck 95
Protective Steel Plates 26

Load Cell 152

Aluminum Plate & 51
Steel Adapter Piece

LVDT

457

Hollow Core Cylinder

(159 mm outer diameter)

Concrete Block

Fig. 2. Pull-out test assembly (dimensions in mm).

Force-Slip Responses for 3-Day SCC 3.5. Transfer length results

200
The girder’s transfer length at each day was calculated using the 95% Average
Maximum Strain Method (95% AMS) [2]. The method takes the transfer length of
160 a prestressed beam as the distance from the beam’s end to the point at which sur-
face strain first reaches 95% of the average plateau strain value; the plateau is rea-
Force (kN)

120 sonably determined by inspection and indicates no further increase in concrete

Peak Pullout Points strain toward the center of the span. An example of the 95% AMS method is shown
in Fig. 11. The 28-days experimental transfer length for the box girder was found to
80 be 532 mm, a value below both the ACI (50db = 635 mm) and AASHTO
First Slip Points
(60db = 762 mm) limits [7–8]. The girder’s 28-days compressive strength was
40 50 MPa.

0 4. Finite element analysis

0 20 40 60 80
Slip (mm)
The pullout bond stress–slip relationships developed in Section
Fig. 3. Force-slip response of seven strands in SCC 3 days after concrete placement. 3.3 were utilized to define parameters for finite element analyses
of large-scale prestressed members. A pullout model was first
developed to test the method in which bond-slip was physically
40 incorporated in the FE analysis. The same methodology was then
35 applied to predict transfer lengths in a prestressed T-beam and
Cylinder Strength (MPa)

the box girder described in Section 3.4.

30
25 4.1. FE pullout model description
20
The finite element program ANSYS was employed to examine a
15 representative portion of a pullout block. The pullout model, shown
10 CC in Fig. 12a, was primarily used to validate the analytical method
SCC chosen to incorporate bond-slip between steel and concrete. Preli-
5
minary analysis showed a 254 mm  254 mm  610 mm concrete
0 prism surrounding one steel strand could accurately capture the re-
1 3 7 28
sponse observed experimentally. The prism was modeled with a
Age (days) single 12.7 mm diameter steel strand embedded 457 mm along
Fig. 4. Average compression cylinder strengths of SCC and CC at various ages. its center axis. The strand extended to a point located 610 mm
above the concrete surface, where a displacement was applied in
ramped increments of 2.54 mm to simulate load transfer via the
prestressing chuck. Nodal constraints allowed the steel strand to
flame-cut in sequential order after concrete had cured for 24 h, meeting release
strength of 34.5 MPa. move only along the line of pullout action and restricted the con-
Surface target points spaced 50 mm apart were attached to both sides of the gir- crete prism from movement at its upper surface.
der along a path at the depth of the strands’ center of mass, as shown in Fig. 10. A Concrete was modeled using SOLID65 brick elements capable of
mechanical gauge was used to measure the distance between targets immediately cracking under tension, crushing under compression, and plastic
before and after releasing the strands, and at 3, 7, 14, and 28 days of concrete cur-
ing. Strains between targets at all ages were calculated with respect to the initial
deformation [25]. The elements had three translational degrees
reading taken prior to prestress release. Three concrete cylinders were tested at of freedom at each of their eight nodes. Concrete properties were
each age to determine the girder’s compressive strength. defined in ANSYS according to the Willam and Warnke failure
 A. Pozolo, B. Andrawes / Construction and Building Materials 25 (2011) 1026–1036 1031

a 200 b 200
150 150

P (kN)
P (kN)
100 100
CC CC
50 SCC 50 SCC

0 0
1 3 7 28 1 3 7 28
Age (days) Age (days)

Fig. 5. Average absolute first slip (a) and peak pullout (b) loads.

a 1.60 b 1.60
1.40 1.40
P′ = P/√f′ c (m√kN)

P′ = P/√f′ c (m√kN)
1.20 1.20
1.00 1.00
0.80 0.80
0.60 0.60
0.40 CC 0.40 CC
SCC 0.20
SCC
0.20
0.00 0.00
1 3 7 28 1 3 7 28
Age (days) Age (days)
Fig. 6. Average normalized first slip (a) and peak pullout (b) loads.

2.5 SCC Bond Stress-Slip Relationships

2.5
U ′ = U/ √ f ′c (√MPa)

2
U′ = U/ √f′c (√ MPa)

2.0

1.5 1.5

1.0
1
1 Day 3 Days
Experimental Relationships 0.5
0.5 7 Days 28 Days
Idealized Relationship
0.0
0 20 40 60 80
0 Slip (mm)
0 10 20 30 40 50 60 70 80
Slip (mm)
CC Bond Stress-Slip Relationships
Fig. 7. Idealized and experimental bond stress–slip for strands in 3 days SCC. 2.5
U ′ = U/ √ f ′c (√MPa)

2.0
model [26]. The uniaxial cracking stress was taken as the modulus
of rupture, and the uniaxial crushing capability was deemed irrel- 1.5
evant and was, therefore, disabled to avoid convergence problems.
The concrete stress–strain behavior adopted in the analysis was 1.0
a multi-linear curve obtained according to the Todeschini stress– 1 Day 3 Days
strain model, which is defined using the following: 0.5
7 Days 28 Days
 
2fc0 e 0.0
eo 0 20 40 60 80
fc ¼  2 ð1Þ
1þ e Slip (mm)
eo
Fig. 8. Normalized bond stress–slip relationships for SCC (top) and CC (bottom) at
2f 0 four ages.
eo ¼ c ð2Þ
Ec
where f is the stress at any strain e, and eo is the strain at peak com-
f pressive strengthp fc0ffiffiffi[27]. The concrete modulus of elasticity Ec was
Ec ¼ ð3Þ ffi
e taken as 57,000 fc0 . After conducting a preliminary mesh refine-
1032 A. Pozolo, B. Andrawes / Construction and Building Materials 25 (2011) 1026–1036

152 610 108

140 152
152

16 8 121
178
356
686
32 19

356
51, typ. 102

140 51
38

127 51 203 3 spa. at 51 203 51 127

Fig. 9. Geometry and strand locations for hollow box girder (dimensions in mm).

Fig. 10. Surface target points along hollow box girder used for transfer length tests.

ment study, it was decided to use a concrete prism comprising 2400

Transfer Length (L t) using 95% AMS SOLID65 elements.
0.0020 Steel strand was modeled as LINK8 spar elements which re-
95% AMS sisted uniaxial tension–compression forces and did not consider
0.0018
0.0016 bending. The elements had three translational degrees of freedom
at each of their two nodes. Steel was assigned a multi-linear isotro-
Strain (mm/mm)

0.0014
pic stress–strain curve for 1860 MPa strand, a 98.7 mm2 cross-sec-
0.0012
tional area.
0.0010
0.0008
4.2. Bond-slip model
0.0006
0.0004
Lt To model bond-slip behavior at the concrete–strand interface,
0.0002
three COMBIN39 spring elements connected each pair of coinci-
0.0000
0 200 400 600 800 1000 1200 dent concrete and steel nodes, as shown in Fig. 12b. Two of the
springs were relatively rigid and acted only to prevent coincident
Distance from Beam End (mm)
nodes from slipping relative to each other in the horizontal plane.
Fig. 11. Transfer length data for hollow box SCC girder at 28 days. The third spring acted only in the direction of pullout and was
 A. Pozolo, B. Andrawes / Construction and Building Materials 25 (2011) 1026–1036
(a) Prism geometry
Fig. 12. Pullout prism (a) and rheological representation (b) of concrete-strand interface with gray steel node and black concrete node.
2.5
Normalized Bond
2.0
Stress (√MPa)
1.5
1.0
0.5
0.0
0 20 40
Slip (mm)
Fig. 13. Ideal and analytical bond stress–slip
pffiffiffiffi relationships for SCC and CC 3 days
after concrete placement, normalized by fc0 .
characterized by nonlinear force–deflection inputs defined by the
idealized experimental pullout force-slip responses. The model
contained 57 COMBIN39 elements. As seen in Fig. 13, the global
FE model output correlated well with experimental data and
showed that nonlinear springs were capable of capturing bond-
slip.
5. Transfer length estimation
5.1. Correlating pullout bond stress to transfer bond stress
Modification factors were derived to correlate pullout bond
stresses to transfer bond stresses before experimental stress–slip
relationships could be utilized in FE analysis to predict transfer
lengths. Average bond stress–slip relationships derived from pull-
out data underestimate the stress–slip of strands in prestressed
members for several reasons. First, for simplification, the average
bond stresses are determined by assuming uniform bond stress
distribution along the entire embedded strand; however, the actual
bond stress varies with the depth of embedment as shown in
1033
COMBIN39 - UX COMBIN39 - UY COMBIN39 - UZ
(b) Concrete-strand interface
Pullout Load
Initial Load
Idealized Curves Intermediate Load
CC Model Output
SCC Model Output
60 80
Ultimate Load
Fig. 14. Variation of bond stress during pull-out test.
Fig. 14. The figure illustrates the location of peak stress shifting
along the strand at increasing load increments. Similar progression
patterns have been measured in steel reinforcing bars in both con-
centric [28] and eccentric [29] pullout specimens. As such, the
length of strand resisting pullout load is much shorter than the full
embedment length, especially at small loads. Second, the physical
conditions in pullout blocks differ from those in transfer zones of
prestressed members. As noted in Sozen et al. [30], strands in pull-
out blocks undergo radial contraction while being pulled, yet
strands in transfer zones expand when released from tension; this
phenomenon, known as wedging action, exerts pressure on the
surrounding concrete and increases the frictional component of
bond. Finally, stirrup reinforcement in prestressed beams actively
confines concrete in transfer zones, which has been shown to sig-
nificantly improve the bond capacity of strands [31]; concrete in
pull-out tests is typically unconfined.
1034 A. Pozolo, B. Andrawes / Construction and Building Materials 25 (2011) 1026–1036

The impacts of non-uniform stress distribution, prestressing,
and confinement on bond are excluded in deriving average pullout
bond stress–slip relationships. For this study, to correlate pullout
bond stress to transfer bond stress, modification factors were de-
rived to account for non-uniform stress distribution and wedging
action; no increase in bond capacity was made to account for con-
finement. Data from Sozen and coauthors [30] was utilized to cal-
culate a ratio between maximum bond stress and average bond
stress (Umax/Uavg = 2.1) in a pullout specimen with 457-mm
embedment. Then, data from Cousins et al. [32] was utilized to cal-
culate a ratio between average bond stresses of prestressed and
non-prestressed pullout specimens (Up/Unp = 1.91) containing un-
coated 12.7-mm diameter strands. A combined modification factor
(MF) of 4.0 was obtained by multiplying the two aforementioned
ratios, and was applied directly to the stress–slip relationships pre-
sented previously in Fig. 8. The impact of this modification is seen
in Fig. 15, which contrasts an unmodified and modified relation-
ship. In the figure, the circled region shows where stress–slip
behavior is most critical in determining transfer length; the range
of slip values in this region are reasonable bounds for typical con-
crete-strand slip in prestressed members.
5.2. FE modeling technique verification
To validate the methodology and assumptions presented thus
far, analysis was conducted on a FE model of a beam tested exper-
imentally by Haq [13]. The T-beam was 11.6 meters long with
cross-sectional properties as shown in Fig. 16. SOLID65 elements
were used to model the concrete, the material properties for which
were derived as per Section 4.1 using the 55.4 MPa 28-days
8 ments retained the same material properties as in the pullout
7 model but now included an initial strain of 0.00521 mm/mm to
U′ = U/√f′c (√MPa)
strength reported in the study for mix SCC2B (see Table 1). LINK8
elements modeled the beam’s two prestressing strands, which in-
cluded an initial strain of 0.00579 mm/mm to impart an estimated
prestress of 1118 MPa. All concrete and steel elements in the beam
were 152.4 mm long. As in the pullout model, three COMBIN39
elements were affixed between each pair of coincident concrete
and steel nodes to represent bond-slip along the beam’s longitudi-
nal axis. This method assumed that camber was small enough to
neglect the vertical component of prestressing force after beam
deformation. The normalized average force-slip response from
pull-out tests using mix SCC2B, multiplied by the modification fac-
tor (MF = 4.0), was utilized to derive force–deflection parameters
for the COMBIN39 springs. The average analytical end-slip, or the
relative movement between the prestressing strand and the con-
crete surface at a beam’s end, was found to be 3.24 mm. When
compared to Haq’s average experimental end-slip of 3.31 mm,
the analysis yielded a relative error of 2%. Therefore, the assump-
tions regarding modification factor (MF) and the overall FE model-
ing technique were deemed acceptable.
5.3. FE Box girder model
The hollow box girder described in Section 3.4 was modeled
along one axis of symmetry to reduce the computational effort re-
quired in FE analysis, as shown in Fig. 17. All points along the axis
of symmetry were restrained to prevent out-of-plane deformation.
Concrete material properties were derived as per Section 4.1 using
the 28-days compressive strength of 50 MPa. To account for the
stirrups and longitudinal reinforcement in the experimental girder,
appropriate steel volume ratios for smeared reinforcement were
assigned to SOLID65 elements in the FE model. LINK8 steel ele-

6 impart a 1032 MPa stress within the strands (after prestress

losses). Most elements in the girder were 152.4 mm long, though
5
Idealized Transfer Bond Stress elements at both ends were refined to more accurately capture
4 Idealized Pullout Bond Stress stresses and strains within transfer zones. The model contained
3 26,532 SOLID65 elements and 858 LINK8 elements. Three COM-
2 BIN39 elements were affixed between each pair of coincident con-
1 crete and steel nodes to represent bond-slip along the girder’s
longitudinal axis. The normalized experimental bond stress–slip
0
0 20 40 60 80 relationships for 28-days SCC, multiplied by the modification fac-
Slip (mm) tor MF = 4.0, were used to calculate the force–deflection parame-
ters for the COMBIN39 elements.
Fig. 15. Unmodified (pullout) and modified (transfer) bond stress–slip relationship Transfer lengths for the FE girder were obtained using the end-
for 3-days SCC; circled region is most critical in determining transfer length. slip method in which the relative slip D between the prestressing

114 153 114

76

#13 Bars
(U.S. #4)

305 Stirrups placed 305 mm

on center (U.S. #2)

51 12.7 mm
Prestressing Strands
51 51 51
Fig. 16. Geometry and reinforcement for T-beam analysis (dimensions in mm).
 A. Pozolo, B. Andrawes / Construction and Building Materials 25 (2011) 1026–1036
Fig. 17. FE girder model discretization with strands shown as circles.
Analytical and Experimental Transfer Lengths
1000
Transfer Length (mm)
900
800 AASHTO
700 ACI
600
500
400
300
200
100
0
T-Beam
Fig. 18. Analytical and experimental transfer length values for T-beam and box
beam.
strand and the concrete surface at a beam’s end is directly corre-
lated to transfer length Lt as follows:
aEps D
Lt ¼
fsi
where a depends on the stress distribution along the strand, Eps is
the strand’s modulus of elasticity, and fsi is the strand’s initial stress
[2]. The end-slip for each case was taken as the average of the end-
slip values obtained for all strands within the girder. The parameter
a in Eq. (4) is typically assigned a value between 2 (constant stress
distribution) and 3 (linear stress distribution); this analysis as-
sumed a constant stress distribution along each strand.
Calculated using the average end-slip values for all strands in
the FE girder, the analytical transfer length was 519 mm, or 2.4%
less than the experimental value of 532 mm. Fig. 18 compares
the girder’s analytical and experimental transfer lengths to current
ACI (50db = 635 mm) and AASHTO (60db = 762 mm) limits [7–8].
The figure also compares analytical and experimental transfer
lengths for the T-beam, where both lengths were calculated using
Eq. (4). As illustrated, the transfer lengths predicted using FE anal-
ysis and pull-out test data correlated well with experimental val-
ues and code-specified limits. It is expected that the Illinois SCC
utilized in the pullout experiments, the SCC girder, and the FE anal-
ysis will perform adequately in prestressed members.
6. Conclusions
This paper first examined results from eight experimental large-
scale published studies on bond behavior and transfer length in
1035
Refined Mesh for 914.4 mm
from Each End
SCC. The paper then presented the results of 56 pull-out tests in
Experimental the form of bond stress–slip relationships for 12.7 mm seven-wire
Analytical
steel strand embedded in SCC and CC. When incorporated in FE
beam models, modified stress–slip relationships reasonably pre-
dicted the transfer lengths of strands in two prestressed beams.
The study afforded the following observations:
(1) Pull-out test results at various ages showed strand
performance in SCC to be comparable with strand perfor-
mance in CC. Normalized pullout loads differed between
Box Beam the two concrete types by as little as 1% after 3 days of
curing.
(2) Average pullout bond stress–slip relationships obtained
using experimental pullout data establish lower bounds for
bond strength because they do not account for non-uniform
stress distribution along strands or radial expansion due to
prestressing. A modification factor of 4.0 was derived in this
study to correlate pullout bond stress–slip relationships to
transfer bond stress–slip relationships.
(3) Using the modified bond stress–slip relationships, transfer
ð4Þ
lengths predicted via FE analysis for a SCC T-beam and hol-
low box beam were 952 mm and 519 mm, respectively, with
errors of 6.7% and 2.4%, respectively.
(4) Both analytical (519 mm) and experimental (532 mm) trans-
fer lengths for the SCC box girder were below the ACI and
AASHTO limits. The SCC mix tested in this study is therefore
expected to perform adequately in prestressed members.
(5) While the study utilized SCC adhering to IDOT standards, the
analytical procedure relating pullout stress–slip relation-
ships to transfer stress–slip relationships may be useful in
analysis of any pretensioned specimen. In-depth research
may be warranted to refine the stress–slip modification fac-
tor and validate the FE analysis.
References
[1] Okamuru H, Ouchi M. Self-compacting concrete. J Adv Concr Technol
2003;1(1):5–15.
[2] Andrawes B, Shin M, Pozolo, A. Transfer and development length of
prestressing tendons in full-scale AASHTO prestressed concrete girders using
self-consolidating concrete. Illinois center for transportation report FHWA-
ICT-09-038, Illinois Department of Transportation, Springfield, IL; 2009.
[3] Ruiz E. et al. Estimating transfer lengths of precast/prestressed beams cast with
self-consolidating concrete. Paper no. 07-2276, transportation research board
86th annual meeting, Washington, DC; 2007.
[4] Larson K, Peterman R, Esmaeily A. Evaluating the time-dependent
deformations and bond characteristics of a self consolidating concrete mix
1036 A. Pozolo, B. Andrawes / Construction and Building Materials 25 (2011) 1026–1036
and the implication for pretensioned bridge applications. Report no. FHWA-
KS-01-1, Kansas Department of Transportation, Topeka, KS; 2007.
[5] Hamilton H, Labonte T. Self-consolidating concrete (SCC) structural
investigation. Report no. BD545 RPWO#21, Florida Department of
Transportation, Gainesville, FL; 2005.
[6] Trent J. Transfer length, development length, flexural strength, and prestress
losses evaluation in pretensioned self-consolidating concrete members. M.S.
Thesis, Virginia Polytechnic Institute and State University, Blacksburg, VA;
2007.
[7] ACI 318-08. Building code requirements for structural concrete
and commentary. American Concrete Institute, Farmington Hills, MI;
2008.
[8] AASHTO LRFD Bridge Design Specifications. 3rd ed. American Association of
State Highway and Transportation Officials, Washington, DC; 2004.
[9] Naito C. et al. Comparative performance of high early strength and self
consolidating concrete for use in precast bridge beam construction. ATLSS
report no. 05-03, Advanced technology for large structural systems,
Bethlehem, PA; 2005.
[10] Logan D. Acceptance criteria for bond quality of strand for pretensioned
prestressed concrete applications. PCI J 1997;42(2):52–90.
[11] Zia P, Nunez R, Mata L. Implementation of self-consolidating concrete for
prestressed concrete girders. FHWA/NC/2006-30 final report, North Carolina
department of transportation, Raleigh, NC; 2005.
[12] Girgis A, Tuan C. Bond Strength of self-consolidating concrete for prestressed
concrete applications. Report no. SPR-PL-1 (037) P528 & EACNH67114,
Nebraska Department of Roads, Omaha; NE; 2005.
[13] Haq M. Effect of self consolidating concrete (SCC) mix proportioning on
transfer and development length of prestressing strands. M.S. Thesis, Michigan
State University, East Lansing, MI; 2005.
[14] Lachemi M et al. Self-consolidating concrete incorporating new viscosity
modifying admixtures. Cem Concr Res 2004;34(6):917–26.
[15] Khayat K, Guizani Z. Use of viscosity-modifying admixture to enhance stability
of fluid concrete. ACI Mater J No. 94-M40 1997:332–40.
[16] Self-consolidating concrete for precast products (BDE). Bureau of materials
and physical research, Illinois Department of Transportation, Springfield, IL;
2007.
[17] Moustafa S. Pullout strength of strand and lifting loops. Technical bulletin 74-
B5, Concrete Technology Associates, Tacoma, WA; 1974.
[18] Interim guidelines for the use of self-consolidating concrete in precast/
prestressed concrete member plants. Precast/Prestressed Concrete Institute,
Chicago, IL; 2003.
[19] Russell B. Rose D. Investigation of standardized tests to measure the
bond performance of prestressing strand. Research report, fears
structural engineering laboratory, University of Oklahoma, Norman, OK; July
1996.
[20] Test Methods for Self-consolidating concrete. Technical Bulletin TB-1506C, W.
R. Grace & Co.-Conn., Cambridge, MA; 2006.
[21] Chan Y, Chen Y, Liu Y. Development of bond strength of reinforcement steel in
self-consolidating concrete. ACI Struct J 2003;100(4).
[22] Esfahani M, Lachemi M, Kianoush M. Top-bar effect of steel bars in self-
consolidating concrete (SCC). Cem Concr Compos 2008;30(1).
[23] Mitchell D, Marzouk H. Bond characteristics of high-strength lightweight
concrete. ACI Struct J 2007;104(1).
[24] Darwin D et al. Development Length Criteria for Conventional and high
relative rib area reinforcing bars. ACI Struct J 1996;98(5).
[25] ANSYS structural analysis guide. Release 11.0, ANSYS, Inc., Canonsburg, PA;
2007.
[26] Willam K, Warnke E. Constitutive model for the triaxial behaviour of concrete.
Seminar on concrete structures subjected to triaxial stresses. In: International
association of bridge and structural engineering conference, Bergamo, Italy;
1974.
[27] Todeschini C, Albert B, Kesler C. Behavior of concrete columns reinforced with
high strength steels. Proc J Am Concr Inst 1964;61(6):701–16.
[28] Feldman L, Bartlett F. Bond stresses along plain steel reinforcing bars in pullout
specimens. ACI Struct J 2007;104(6).
[29] Perry E, Thompson J. Bond stress distribution on reinforcing steel in beams and
pullout specimens. Proc J Am Concr Inst 1966;63(8).
[30] Anderson G, Rider H, Sozen M. Bond characteristics of prestressing strands.
structural research series 283, University of Illinois at Urbana-Champaign,
Urbana, IL; 1964.
[31] Stocker, M. and Sozen, M. Investigation of prestressed reinforced concrete for
highway bridges, Part V: bond characteristics of prestressing strand.
Engineering experiment station bulletin 503, University of Illinois at Urbana-
Champaign, Urbana, IL; 1970.
[32] Cousins T, Badeaux M, Moustafa S. Proposed test for determining bond
characteristics of prestressing strands. PCI J 1992;33(1):66–73.