Professional Documents
Culture Documents
to
0.82 f
ci
to 0.70 f
ci
.
DURABILITY
MarchApril 2006 15
O
ver time, the initial prestressing force that is ap-
plied to a member decreases in magnitude. During
a period of approximately ve years, a prestressed
concrete member may lose as much as 25% of the initial pre-
stress force.
1,2
Elastic shortening, creep, shrinkage, and re-
laxation are the four primary contributors to prestress losses
in pretensioned beams. Of the four, elastic shortening is the
only loss that is not time dependent.
There are many methods used to estimate the effective
prestress force f
se
(prestress force existing after all losses).
The three most widely accepted methods are: (1) the Rened
Estimates of Time Dependent Losses method found in the
2004 American Association of Highway Transportation Of-
cials load-resistant factor design (AASHTO LRFD) Bridge
Design Specications, (2) the PCI Design Handbook method,
described by Zia et. al., and (3) lump sum estimates.
35
A new method is outlined in the recently published Na-
tional Cooperative Highway Research Program (NCHRP)
Report 496, the Detailed Method for Estimating Prestress
Losses.
6
The method in NCHRP Report 496 was developed
specically for girders cast with high-strength concrete.
In the research program reported herein, the prestress loss-
es of four I-shaped girders were measured and compared with
losses calculated using the following methods: the AASHTO
LRFD Bridge Design Specications, the PCI Design Hand-
book, and the recommendations from the NCHRP Report
496.
4
In addition, the beams and their respective prestressing
forces were designed to impose compressive stresses at re-
lease that exceeded the allowable release stresses found
in both AASHTO LRFD Bridge Design Specications
and the ACI Building Code Requirements for Structural
Concrete (ACI 318-05).
7
Currently, AASHTO LRFD and
ACI 318 limit the concrete compressive stresses after release
to 60% of the concretes release strength (0.60 f
ci
).
This research program examines whether compressive
stresses in excess of 0.60 f
ci
.
LITERATURE REVIEW
Prestress Losses
Several ongoing research projects have been conducted
recently to determine the accuracy of design equations used
for estimating prestress losses in high-performance concrete
(HPC) bridge girders. Roller et al. examined the prestress
losses in HPC bridge girders.
8
Along with prestress losses,
they also measured the creep and shrinkage of the concrete.
Five girders were cast, but prestress losses were reported for
only Girder BT3 and Girder BT5.
The two girders were subjected to two different curing
regimens. Girder BT3 was steam cured for 24 hours at 140 F
(60 C). Curing continued on Girder BT3 for an additional 10
hours until the forms were removed. Girder BT5 was cured
under a waterproof tarpaulin for 10 hours. The forms were
removed 12 hours after casting.
Prestress losses were measured using internal strain meters.
The research results showed that the AASHTO equations over-
estimated the losses by approximately 50% at 18 months for
Girder BT3. The researchers reported, however, that the steam
curing of Girder BT3 may have affected the prestress losses.
Compared with the early age losses of Girder BT5 (non-steam
cured), those of Girder BT3 were signicantly lower.
The results also showed that the AASHTO equations for
estimating creep and shrinkage may be overly conservative
for high strength concrete. The researchers recommended
further study of the creep and shrinkage behavior of HPC to
determine whether the AASHTO equations can be modied.
Roller et al. again examined the prestress losses, but this
research program examined the losses in a bridge built by the
Louisiana Department of Transportation and Development.
Construction of the bridge was completed in October 1999.
9
The prestress losses were reported for 12 months. The mea-
sured losses were approximately 35% less than the calculated
losses. These results were consistent with the Roller et al.
research.
8
Pessiki et al. examined the effective prestress force in
bridge girders that were 28 years old.
10
Load tests were per-
formed on the girders to determine the decompression load.
Visual observations, strain gauges, and displacement trans-
ducers were used in obtaining the decompression loads. The
average prestress loss for both girders was 18%, compared
with a loss of 33% predicted by the AASHTO equations.
Azizinamini et al.
also examined the available prestress in
an existing bridge girder, investigating the effective prestress
force in a 25-year-old girder.
11
The girder testing showed that
the prestress loss after 25 years of service was 20.7%, which
was less than the 25.7% predicted by the AASHTO equations.
Idriss measured the prestress losses of HPC bridge gird-
ers during construction and during service.
12
Deformation
sensors were placed in the girders to measure the prestress
losses. After ve months of service, the measured losses
were less than the losses predicted by AASHTO and the
Table 1. Target Release Stresses and Air Content of the Four Girders.
Girder
Targeted Allowable Compressive Stresses ( f
bot
/ f'
ci
)
Targeted Total
Air Content (%)
0.60 0.75 2 6
1 - X X -
2 - X - X
3 X - - X
4 X - X -
16 PCI JOURNAL 16 PCI JOURNAL
PCI Design Handbook. The AASHTO and PCI equations
predicted losses of 28% and 22.6%, respectively, which are
both greater than the measured losses of 11%. The reported
losses were measured after only ve months of service, how-
ever.
In comparison, the calculated values are nal losses for
the entire service life of the beams. Naturally, as the girders
age, they will continue to lose prestress force. The total losses
considering the fullness of service life can be expected to dif-
fer from losses measured at only ve months of age.
The research demonstrates that the AASHTO equations
consistently overestimate the total prestress losses; instead of
arguing for changes to the prediction methods, however, most
researchers agree that more testing needs to be done before
the prediction equations are modied. The NCHRP Report
496 is an outgrowth from efforts of the Transportation Re-
search Board and afliate members to further document and
support alterations to the loss equations.
Allowable Compressive Stress at Release
The AASHTO LRFD Bridge Design Specications, the
ACI 318 Building Code, and the PCI Design Handbook limit
the concrete release stress in compression to 60% of the com-
pressive strength at release (0.60 f
ci
).
3,4,7
The reported pur-
pose of the limit is to control creep deformation and damage
due to micro-cracking.
13
Practitioners suggest, however, that
the limit for compressive stresses is articial and the beam
performance is not reduced by exceeding the compressive
stress limit at release.
In many design cases, strict adherence to the rule effec-
tively increases the need for harping strands or, alternatively,
increases the number and length of debonded strands in the
end regions of girders. In this manner, the current limit for
compressive release stresses effectively increases production
time and creates additional need for steam curing.
14
Unfortunately, little data exist to support increasing the al-
lowable compressive stress at release. Despite the dearth of
published supporting data, many precast, prestressed con-
crete manufacturers, as part of their standard practice, release
strands as long as resulting compressive stresses stay within
75% of the concrete strength. In support of this practice, the
PCI Standard Design Practice reports that no problems have
been reported by allowing compression as high as 0.75 f
ci
.
15
Pang investigated the effects of large compressive stresses
on the hardened properties of concrete.
16
He conducted creep
tests on concrete cylinders loaded at one day of age to 60%,
70%,
and 80% of the concretes one-day breaking strengths.
Cylinders loaded to 70% of breaking strength demonstrated no
adverse effects from the sustained loading at early ages.
One of the cylinders that was loaded to 80% of breaking
strength failed under sustained loading, possibly indicating
that 0.80 f
ci
.
Noppakunwijai et al. conducted an experimental research
program that examined the effects of high release stresses
on precast, prestressed concrete girders.
17
Two girders with
compressive stresses at release of 0.79 f
ci
and 0.84 f
ci
were
cast. The prestress losses of the girders were measured. They
concluded that higher release stresses did not have a negative
impact on the test specimen.
Despite these recent data, the amount of data on the release
stresses of prestressed concrete is very limited, particularly
experimental data on bridge girders. Therefore, the justica-
tion for increasing the allowable compressive stress limits at
release primarily has been based on common practices in the
precast/prestressed concrete industry.
EXPERIMENTAL PROGRAM
Scope
Four prestressed concrete girders were cast with targeted
release stresses of 0.60 f
ci
or 0.75 f
ci
;
inside at Fears Structural Engineering Laboratory
(FSEL);
and outside at FSEL.
**
At each location, the girders were simply supported with
cribbing and bracing located 12 in. (305 mm) from each end.
Typical support cribbing and bracing can be viewed in Fig. 2,
where the girders are stored outside FSEL.
The reported losses are prestress losses corresponding to
the center of gravity (cg) for the prestressing strands. Strain
readings are taken at DEMEC targets located on the sides of
each girders bottom bulb and at the DEMEC targets located
on the very top of each girder. The strain at the cg is then
determined from the strain readings taken from the DEMEC
target points.
The reported losses are then computed from the cg strains
by multiplying the strain by the strands elastic modulus.
Table 6 lists losses at the ends (average of the two ends) and
at the center of the girder.
Table 6 also lists the ratio of maximum compressive stress
at release to compressive strength at release. For Girder 4, the
compressive stress at the extreme bottom ber is computed to
be 4950 psi (34.1 MPa), or 56.9% of the concretes one-day
strength (that is, 8700 psi [60 MPa]).
Likewise, for Girder 1, the computed compressive stress
at the bottom ber is 5650 psi (40 MPa), or 64.9% of the
concretes one-day strength. For Girder 3, the computed
compressive stress at the bottom ber is 4250 psi (29.3 MPa),
or 69.3% of the concretes one-day strength (that is, 6130 psi
[42.3 MPa]). Finally, the highest ratio of stress to strength
occurred on Girder 2, where the maximum compressive stress
at release was 82.1% of the concretes compressive strength.
In engineering practice, the concrete is presumed to be lin-
Table 4. Fresh Concrete Properties for Girders 1 Through 4.
Fresh Concrete Properties Girders 1 and 4 Girders 2 and 3
Fresh concrete temperature (F) 68 60
Slump (in.) 10.0 9.75
Air content (%) 2.3 6.2
Unit weight (lb/ft
3
) 151.1 146.9
Air temperature (F) 36 35
Note: C = (
5
/
9
)(F -32); 1 in. = 25.4 mm; 1 lb/ft
3
= 16.02 kg/m
3
.
Table 5. Hardened Concrete Properties of Girders 1 through 4.
Age of Test Girders 1 and 4 Girders 2 and 3
Average Compressive Strength (psi)*
1 day (at release) 8700 6130
14 days 10,190 7110
28 days 11,060 8390
56 days 12,440 9200
180 days 14,460 10,850
360 days 15,610 11,460
Average Modulus of Elasticity (ksi)*
1 day (at release) 5600 4700
14 days 5800 4900
28 days 6000 5500
56 days 6300 5400
180 days 6800 5500
360 days 6900 6000
* Reported values are the average of three tests.
Note: 1 psi = 0.006895 MPa; 1 ksi = 6.895 MPa.
20 PCI JOURNAL 20 PCI JOURNAL
ear and elastic for stress calculations. In other words, stresses
are proportional to strains and the relationship is dened by
the elastic modulus. Accordingly, the normal engineering
practice to compute the bottom ber precompression stress is
given by the equation:
f
bot
= F
si
1
A
tr
+
e
tr
S
b
( )
tr
where F
si
is the prestress force immediately prior to release
and the cross-section properties are all transformed.
If gross properties are used, then the prestress force used
in the equation must be the prestress force immediately after
prestress release to obtain reasonably accurate estimates for
concrete stress. By using the transformed cross-section prop-
erties, the exact computation of concrete stresses (and elas-
tic shortening losses) can be obtained without iteration.
The equation shown, however, assumes Hookes Lawa
linear relationship between stress and strain. In actual gird-
ers, however, with compressive stresses approaching and ex-
ceeding 60% of the compressive strength of the concrete, the
stress versus strain relationships are decidedly nonlinear.
To account for that nonlinearity, an effective modulus of
elasticity (E
eff
) was determined and employed in the calcula-
tion of transformed section properties. The effective modulus
was obtained from the parabolic stress versus strain relation-
ship widely known as Hognestads model and given by the
expression
18
:
f
c
f
c
'
= 2
r
c
r
o
[
\
|
)
j
~
r
c
r
o
[
\
|
)
j
2
|
|
|
|
|
|
|
|
The effective modulus was obtained from the slope of a
secant drawn from the origin to a point on the stress versus
strain parabola with the corresponding concrete stress re-
quired to provide equilibrium. The effective moduli range
from 500 ksi to 900 ksi (3.4 GPa to 6.2 GPa) less than the
measured modulus.
It should be noted that a lesser concrete modulus will re-
sult in a proportionately greater elastic shortening loss. The
effective modulus (E
eff
) was used to compute stresses in the
concrete after release and the elastic shortening losses.
The losses reported in Table 6 do not include losses due to
steel relaxation. Because relaxation occurs without an accom-
panying change in strain, relaxation cannot be measured with
DEMEC points. Relaxation losses were estimated using the
current AASTHO LRFD equations and are incorporated in the
total loss calculations that are discussed later in the paper.
DISCUSSION OF TEST RESULTS
Fresh Concrete Properties
One goal of the research was to develop HPC that could
easily be used in the local precast concrete industry. The
sponsor of the project, the Oklahoma Department of Trans-
port, was also interested in our ability to produce air-en-
trained HPC suitable for fabrication of bridge girders.
The slumps of the two mixtures were 9.75 in. and 10 in.
(246 mm and 254 mm), indicating that both mixtures had suf-
cient workability to produce the girders. As can be seen in Fig.
5 and 6, the girders had a smooth nish and there were no hon-
eycombed areas.
Girders 1 and 4 had a targeted air content of 2%, and Girders 2
and 3 had a targeted air content of 6%. One mixture was batched
for Girders 1 and 4, and another mixture was batched for Gird-
ers 2 and 3. The measured total air contents for both mixtures
were within 0.50% of the targeted air contents indicating our
ability to produce concrete within specied air content ranges.
Table 6. Measured Prestress Losses.
Average Measured Prestress Losses (ksi)
*
Beam
Age
(Days)
Temperature
(F)
Girder 1 Girder 2 Girder 3 Girder 4
f
bot
/ f'
c
= 64.9% f
bot
/ f'
c
= 82.1% f
bot
/ f'
c
= 69.3% f
bot
/ f'
c
= 56.9%
Ends
Center Ends
Center Ends
Center Ends
Center
1 (at
release)
36
and
0.75 f
ci
.
Similarly, Girder 4 was designed to have allowable com-
pressive stress of 0.60 f
ci
.
The 700 psi (4.8 MPa) increase in strength beyond targeted
Table 7. Total Prestress Losses.
Girders Location
Total Prestress Losses (ksi)
Measured
(1 year)
*
AASHTO
Report 496
1
Ends 0.72 0.81 1.07
Center 0.68 0.77 1.01
2
Ends 0.92 0.89 1.04
Center 0.95 0.92 1.08
3
Ends 0.93 0.94 1.05
Center 0.92 0.94 1.05
4
Ends 0.73 0.84 1.08
Center 0.74 0.84 1.09
Average 0.82 0.87 1.06
*
AASHTO: American Association of Highway Transportation Ofcials.|
. Because
the compressive strength of 6130 psi (42.3 MPa) was lower
than targeted, the maximum compressive stress at release was
0.82 f
ci
.
Similarly, Girder 3 was designed to have an allowable com-
pressive stress at release of 0.60 f
ci
. In all
cases, it is worth noting that the prestressing forces were
adequately transferred to the concrete (demonstrated by the
strain measurements that occurred within the end regions of
each girder) and no external signs of distress were visible.
Although the high release stresses did not appear to affect
the girder performance otherwise, higher prestress losses re-
sult from higher compressive stresses at release. This result
is in line with expectations. The total measured losses for the
four girders and the allowable compressive stress at release
are shown in Table 6. As shown in Fig. 8, the girders with the
greatest release stress had the most prestress loss.
The effects of large compressive release stresses are ex-
amined in Table 11. In the rst column, Table 11 lists the
maximum release stress as a percentage of the one-day com-
pressive strength of the concrete. As noted in Table 11 and
elsewhere, the compressive stresses ranged from 56.9% to
82.1% of release strength. The second column in Table 11
lists the measured prestress losses, which are the average
losses for the ends at each girder. Total losses ranged from a
low of 51.6 ksi (356 MPa) to a high of 72.0 ksi (496 MPa).
The third column of Table 11 lists the jacking stress (ksi) of
the strand immediately prior to release. The reported jacking
stress is based on the elongation measurements made during
stressing of the strands and includes the seating losses. The
jacking stresses listed in the third column, however, do not
include relaxation losses that occur in the 24 hours between
tensioning and release. The fourth column of Table 11 ex-
presses the measured losses as a percentage of the jacking
stresses. In other words, the 27.8% loss reported for Girder 1
is obtained by dividing the total measured loss in the second
column by the jacking stress in the third column.
The table clearly shows that increases in the maximum re-
lease stress result in increased prestress losses. The question
becomes, in the cases where the allowable stress provisions of
the specications were exceededas was the case for Gird-
ers 1, 2, and 3whether the losses indicate that a damaged
condition exists in the concrete due to excessive compressive
stresses at release.
The fth column addresses that issue by listing the ratio of
prestress losses divided by the release stress. Both losses in
the fourth column and the release stress in the rst column
are expressed as a percentage. According to the information
in the fth column, the ratio of losses to release stresses is
approximately the same for all four beams regardless of the
amount of compressive release stresses. This occurs over a
fairly broad range of values ranging from Girder 4, which
has release stresses within the allowable limits, to Girder 2,
with a release stress that is 82% of the compressive strength
of the concrete.
Table 9. Elastic Shortening Losses.
Girders Location
Prestress Losses Due to Elastic Shortening (ksi)
1
Measured AASHTO Predicted Zia et al. NCHRP 496
1
Ends
2
27.8 23.3 20.9 23.3
Center 27.8 23.2 20.8 23.2
2
Ends 32.8 27.4 24.6 27.4
Center 33.6 27.2 24.4 27.2
3
Ends 24.4 20.7 18.7 20.7
Center 23.4 20.6 18.6 20.6
4
Ends 25.6 20.4 18.4 20.4
Center 24.8 20.3 18.2 20.3
Average elastic shortening 27.5 22.9 20.6 22.9
Note: 1 ksi = 6.895 MPa.
0
10
20
30
40
50
60
70
80
0 50 100 150 200 250 300 350 400
Girde r 2 , f
b o t
/f
c i
= 8 2.1 %
Girde r 3, f
bo t
/f
c i
= 6 9 .3 %
G irde r 1 , f
bo t
/f
c i
= 6 4 .9 %
Girde r 4 , f
bo t
/f
c i
= 5 6 .9 %
Age (Days)
Fig. 8. Shown are the measured losses for the girders at ends.
24 PCI JOURNAL 24 PCI JOURNAL
It is the authors view that these data provide strong evi-
dence that the allowable release strength of 0.60 f
ci
can be
relaxed to allow higher compressive stresses to be imposed
on the concrete immediately after release.
CONCLUSION
Both the AASHTO LRFD Specications and the ACI
Building Codes currently limit the compressive stress at re-
lease stress to 0.60 f
ci
to 0.70 f
ci
.
ACKNOWLEDGMENT
The authors express their deep gratitude to Coreslab Struc-
tures Inc., Oklahoma City, for donating the materials and
fabricating the four prestressed concrete girders at its plant.
The authors also want to thank Oklahoma State University
in Stillwater for the use of its facilities and in particular to
express their appreciation to the personnel of the Fears Struc-
tural Engineering Laboratory for their care and diligence in
carrying out the testing.
Table 10. Girder Properties and Allowable Compressive Stresses.
Parameter Girder 1 Girder 2 Girder 3 Girder 4
f
ci
(psi) 8700 6130 6130 8700
E
eff
(ksi) 5022 3862 4159 5122
A
tr
(in.
4
) 173.38 175.71 173.40 169.19
y
tr
(in.) 13.32 13.20 13.35 13.48
I
tr
(in.
4
) 13,223.68 13,488.50 13,215.80 13,007.65
e
tr
(in.) 6.90 7.20 6.40 7.75
Jacking stress (ksi) 204.25 202.20 2000.76 204.47
f
bot
(ksi) 5.646 5.030 4.245 4.949
f
bot
/ f'
c
(%) 64.9 82.1 69.3 56.9
Note: 1 in. = 25.4 mm; 1 in.
2
= 645 mm
2
; 1 in.
4
= 416,230 mm
4
; 1 psi = 0.006895 MPa; 1 ksi = 6.895 MPa.
Table 11. Effects of Large Compressive Stresses at Release.
Girders
1 2 3 4 5
Release Stress
(%)
Measured Pre-
stress Loss (ksi)
Jacking Stress
(ksi)
(2)/(3) (%) (4)/(1)
1 64.9 56.8 204.25 27.8 0.43
2 82.1 72.0 202.20 35.6 0.43
3 69.3 59.2 200.76 29.5 0.43
4 56.9 51.6 204.47 25.2 0.44
Note: 1 ksi = 6.895 MPa.
MarchApril 2006 25
APPENDIX: NOTATION
A
g
= gross area of concrete section
A
tr
= transformed area of concrete section
I
g
= moment of inertia of gross concrete section
I
tr
= moment of inertia of transformed concrete section
y
g
= distance from bottom ber to center of gravity of gross
section
y
tr
= distance from bottom ber to center of gravity of transformed
section
e
tr
= eccentricity of prestress force of transformed section
f
ci