Post-Damage Repair of Prestressed Concrete Girders
Thomas H.-K. Kang
Concrete is an economical construction material and for that reason it is widely used in buildings and infrastructures. The use of deicing salts, expansion joint failure, and freeze-thaw cycles have led to concrete bridge girders experiencing corrosion of steel reinforcement and becoming unsafe for driving. The goal of this research is to assess the effectiveness of current and possible repair techniques for the end region of damaged prestressed concrete girders. To do this, three American Association of State Highway and Transportation prestressed concrete girders were tested to failure, repaired, and retested. Three different repair materials were tested including carbon fiber, glass fiber, and surface mounted rods. Each different repair material was also tested with and without injected epoxy. Comparisons were then made to determine if injecting epoxy had a positive effect on stiffness and strength recovery as well as which repair type regained the largest percentage of original strength.
The national transit system plays a large role in economy
and society. More than 60 % of the total bridge inventory in
the United States are made of concrete and is reinforced
concrete (U.S. Department of Transportation 2004). Over the
past 40 years the rate of deterioration of concrete has
increased due to increased use of deicing salts. Although
deterioration of concrete structures over time is normal and
expected, the rate at which this has occurred for highway
bridges since the 1960s, when officials began applying
deicing salts in the winter, has been abnormally advanced and
has posed significant challenges, both economically and
technically. The deterioration is a consequence of the
aggressive nature of chloride ions. This has led to a direct
annual cost of 5.79.7 billion dollars (Koch et al. 2002). If
indirect costs are included the cost increases roughly ten
times (Yunovich et al. 2003). An investigation conducted by
the Michigan Department of Transportation indicated that for
prestressed concrete I-beams, repair ranges from 35 to 69 %
the cost of the superstructure replacement (Needham 2000).
Concrete is generally considered a durable material but is
subject to deterioration caused by mix design, environmental
effects, and corrosion of the embedded reinforcement.
Improper aggregate type and sizing, calcium chloride, and
alkali-aggregate interactions are a few of the material
problems. Environmental effects include expansion and
contraction due to absorption of moisture followed by a freezethaw
cycle. This causes the concrete to crack and may lead to
significant damage or failure. End region concrete
deterioration of prestressed concrete girders tends to be exacerbated
when the expansion joint located at the deck level above the
beam end fails and all of the deicing salts drain over the beam
ends (Yeager 2007). A partially fixed girder end, such as one
created by a frozen bearing, may impose additional stress at
the girder end. When the stress build-up is relieved, tension
cracks or shear cracks may result (Emmons 1994).
Moisture absorbed by concrete expands and contracts with
temperature changes and the resulting mechanical action can
cause cracks, fractures, and spalling. Airborne compounds,
such as carbon dioxide, can also cause adverse chemical
reactions which can lead to surface deterioration (Freeman
et al. 1999). When a concrete structure is often exposed to
deicing salts, salt splashes, salt spray, or seawater, chloride
ions from these will slowly penetrate into the concrete. The
chloride ions eventually reach the steel and then accumulate
beyond a certain concentration level. The protective passivity
is destroyed and the steel begins to corrode, since oxygen and
moisture are present at the steelconcrete interface. Cracking
and spalling may occur in the concrete due to the formation of
voluminous corrosion products, which are up to six times the
volume of the original steel. Corrosion of the reinforcing steel
and prestressing strands at the end of the girder is the most
destructive deterioration mechanism for reinforced concrete
bridges in the United States (Weyers et al. 1993).
The objective of the research is to investigate the
feasibility of methods to repair end damage or strengthen against
end damage in American Association of State Highway and
Transportation (AASHTO) prestressed concrete bridge
girders. Note that to the authors knowledge, no research has
been conducted, nor are there any prediction models
available on the shear strength of post-repaired beams that
experienced shear failure prior to the repair. Figure 1 shows
damage in the end region of an AASHTO bridge girder on
I-244 over the Arkansas River in Tulsa County, Oklahoma,
Several methods of repair for AASHTO girders are being
used at the present time. These include wraps, U-jackets, and
near surface mounted rods. Wraps are impractical because
they require access to the top of the beam which is most
likely covered by the deck and are also expensive. U-jackets
are also expensive but do not require access to the top of the
beam. Near surface mounted rods are a less expensive
alternative if they can recover strength to the beam, but they
also require access to the top of the beam.
These methods are known to work; however, full-scale
structural performance testing on the repair of the prestressed
concrete bridge end region (both undamaged and damaged)
is still lacking and there is also not much direct comparison
between the methods. The research program examined the
effectiveness of three post-damage repair methods including
near surface mounted rods, and two types of fiber reinforced
polymers. Also, the research investigates whether or not the
injection of a high strength-high strain cement-epoxy into
the cracked section will help increase the bearing capacity.
Additionally, the research is devised to test the effects of
adding a distribution plate along with the surface mounted
rods. The ultimate goal of this experiment is to find the best,
cost effective, most convenient way to repair damaged
AASHTO prestressed concrete girders.
2. Tests of Undamaged Prestressed
2.1 Bending Tests
For the bending test setup, full-scale Type II AASHTO
girders with a depth of 915 mm (36 in.) and an area of
2,380 cm2 (369 in2) were placed on simple supports located
Fig. 2 Beam during bending test.
at 6.3 m (20 ft. 9 in.) from the center of the girder. Nine
15.2 mm (0.6 in.) diameter Grade 270 seven-wire strands
were used as tensile longitudinal reinforcement, and the
specified concrete strength was 40 MPa (6,000 psi). The
beams were supported on the rollers with one end above
the roller and the other end extending 914 mm (three feet)
past the support to keep the ends independent of each other
(Figs. 2, 3). This independence was necessary due to the
destructive shear testing of the beam ends later in the test
series. Note that all the tested beams had the same conditions
The load was applied using a 1,780 kN (400 kip)
hydraulic cylinder mounted to a testing frame. A grout pad
was placed directly on the beam surface and a steel plate was
placed on top of the leveler. A 445 kN (100 kip) load cell
was put in place with a steel cylinder on top of the load cell
to provide good contact and protect the cell. Finally, a metal
box-frame extension was included to close the gap between
the hydraulic cylinder and the rest of the testing equipment.
This extension was loosely supported by chains as a safety
precaution in case the loading apparatus happened to fall
(Fig. 4). The vertical deflection measurements were taken
using a string potentiometer which was attached to the top
flange of the beam by a clamp and small diameter bent metal
rod. Deflection and load measurements were collected and
recorded by the data acquisition system.
The hydraulic cylinder was used to load the beams until a
deflection of 7.5 mm (0.3 in.) was measured, at which point
a moderate degree of flexural damage in the elastic stress
range was expected to occur. Both load and deflection were
recorded simultaneously over the course of the test. After
reaching 7.5 mm (0.3 in.) of deflection, the beams were
unloaded at a controlled pace to keep the beams from
rebounding past their normal state and possibly causing
damage. The beam would then be moved 914 mm (3 ft) and
the test repeated for the other end.
2.2 Shear Tests
After the bending tests had been completed for both ends
the ultimate shear capacity was tested. The beam end was
failed in shear to determine their strength and also to
simulate a corrosively failed end region. For this test each beam
Fig. 3 Location for loading points and supports during
bending and shear tests.
Fig. 4 Beam during shear test.
was supported with the same distance from each end, as was
done in the bending test; however, the load was applied only
914 mm (3 ft) from the end in question (Figs. 3, 4).
The loading stack used in the shear test was similar to that
of the bending test, except a load cell with a larger capacity
(1,335 kN; 300 kip) was used. Vertical deflection
measurements were also taken during this test and the set up was the
same as it was for the bending tests.
During the shear test, the beam was incrementally loaded
and cracks were identified between loading increments. The
beam was then loaded until the beam reached a maximum
load after which the beam yielded so much that the beam
could not hold any more load. The process was then repeated
for the other end.
3. Post-Damage Repair and Post-Repair Testing
3.1 Post-Damage Repair
The fiber and epoxy repairs to the beams were done by a
professional concrete service firm in Oklahoma, and the
surface mounted rod repairs were installed by the research
assistants of the University of Oklahoma. Repairs were
begun by removing the loose concrete from the ends of the
beams. Rapid set cement was added to ends where there was
a significant amount of missing concrete (Fig. 5).
One end of each beam was epoxy-injected. To accomplish
this, toothpicks were placed in the cracks then small tubes
were placed over the toothpicks. Putty was added to seal off
the cracks and hold the tubes in place. When the putty was
cured the toothpicks were removed and epoxy was then
injected under pressure through the tubes and into the cracks.
The tubes were then capped off. Once the epoxy had cured
the putty and tubes were ground off so the surface was
smooth. Two types of FRP sheets were applied: carbon FRP
and glass FRP. The fiber was soaked in a saturate then placed
over the primer and putty. Rollers were used to remove
surface imperfections. Grooves were cut in the beams and
metal rods were placed in them to hold the fiber sheets in
place. A diagram showing FRP dimensions is presented
in Fig. 6.
While the FRP repairs were curing the surface mounted rod
repair was conducted for the last beam. It was decided to use
five sets of stirrups on each end, evenly spaced out over the
crack area, to complete the repairs. Threaded steel rods with a
diameter of 19 mm (3/4 in.) were used in conjunction with
76 9 76 9 6.35 mm (3 9 3 9 1/4 in.) angle steel to create
the stirrups. The angle steel was cut into 406 mm (16 in.)
lengths and had two 0.8125 mm (13/16 in.) holes drilled into
one side so that the closest edges of the holes were 254 mm
(10 in.) apart. This hole placement allowed the steel rods to
pass just on either side of the bulb on the beam (Fig. 7).
A template was made to allow for quick hole placement.
Then using the hammer drill, a hole was drilled through the
top flange of the concrete so that the rods could be connected
to a piece of angle steel on top of the flange. The rods were
then hand-tightened until flush against both the bottom of the
bulb and the top of the flange. With the repairs complete, the
repaired sections was retested.
3.2 Post-Repair Tests
Once the repairs were finished, the bending and shear tests
were repeated as previously described. The load was not
paused during the loading process to mark the cracks
because they could not be seen due to the repairs (Fig. 5).
Also, on the final tests a piece of neoprene was placed on top
of the plate that was on the roller to even out the uneven
surface of the bottom of the beam.
3.3 Additional Testing with Distribution Plate
During the testing of the repaired ends, a trend was
noticed. In every beam, bulb deterioration was the first sign
of (bearing) failure or, in a couple of cases, the only
noticeable failure point. To resolve this, a distribution plate
was added to the beam with the surface mounted rod repair
to decrease the concentration of stress in the 152 mm (6 in.)
of the bulb resting on the support. A 25.4 mm (1 in.) thick
steel plate with the same width as the lower bulb that
extended 533 mm (21 in.) from the end of the beam was
secured between the steel rods and the bottom of the beam.
The beam was then re-retested as described in the preceding
(a) Missing concrete
(b) Rapid set cement repair
(c) Epoxy injection tubes
(d) Primer and putty
(e) Trench cut in top of beam web
(f) Carbon FRP being applied
Fig. 5 Beam post-damage repair.
Fig. 6 Configuration of FRP application to the AASHTO prestressed concrete girder.
4. Results and Discussion of Experimental
The results of the initial tests and tests after post-damage
repair are presented in Table 1. The percent of strength
regained by each repair method is presented in Table 2. Note
that all the tested prestressed concrete AASHTO girders had
the same dimensions and materials and were tested under the
same conditions except for the repair method; thus, the test
results provide a direct comparison between the different
post-damage repair methods. As soon as the first bending
post-repair test was conducted on one end (end A) of the first
(a) Beam after steel repairs
(b) Steel repair sand support
Fig. 7 Post-damage repair using surface mounted rods.
Table 1 Repair type and loadings before and after repair.
Peak load during Peak load during
initial bending test initial shear test
(kN; kips) (kN; kips)
Type of repair
Peak load during Peak load during
post-repair bending post-repair shear
test (kN; kips) test (kN; kips)
beam (beam 1) strengthened with carbon FRP but without
epoxy, the end of the beam began to fail (Fig. 8). This might
have been due to the bottom surface of the beam being
uneven, causing stress concentrations. As a result, a
neoprene pad was added between the beam and its support to
provide a better contact surface and reduce stress
concentrations on the bottom bulb of the beam. However, at the end
of the testing program cracking had still occurred at all of the
ends not repaired with epoxy. These results indicate that the
existing unrepaired cracks due to the absence of epoxy
injection ended up being the cause of the problem.
Due to the damage that beam 1 end A endured, its reaction
point was moved 152 mm (6 in.) further away from the end
for the rest of its testing so that it was possible to get a full
test from that end of the beam. Since this moved the new
load partially out of the previously failed zone, it led to the
maximum load much higher than that noted in the previously
Cracks on beam 2 end A began to form during bending
testing but nowhere near the extent of beam 1 end A, so it
was left the same for later testing. The beam (beam 3) end
strengthened with surface mounted rods but without epoxy
(end A) cracked a considerable amount during its bending
test and so it was stopped slightly early at a deflection of
6.9 mm (0.27 in.). Due to its repair setup a move would
have been impossible with the angle steel in the way so all
other testing also took place from the initial reaction point.
Several of the graphs show an initial deflection opposite of
the expected (Fig. 9). This was likely a result of a
combination of the beam not being exactly flush with its reaction
point and the load not being applied in the direct center of
the beam causing the beam to roll to one side slightly before
coming to a rest, giving the initial deflection.
Tables 1 and 2 provide the data on the maximum loads
that each beam end achieved for each test as well as how
each was repaired and the percentage of initial loading
returned after post-damage repair. The graphs shown in
Figs. 9, 10 and 11 give an overview of how each beam
reacted in deflection as load was applied until its maximum
was reached and unloaded. The graphs also provide a
comparison between how the beam reacted on its initial test
and how it reacted during the tests after repairs (Fig. 10).
From the tables and graphs many comparisons can be
made about the effect that epoxy injection had on the repairs.
For beam 1 end A repaired with carbon FRP, the results are
not as conclusive due to the aforementioned problem in
testing; however, comparisons can still be made. The
postdamage repair job on end B (epoxy-injected end) provided a
peak bending load of 300 kN (67.3 kips) and a peak shear
load of 391 kN (87.9 kips) during the shear test, giving a
recovery percentage of 91.1 and 71.2 % respectively. When
this is compared to end A (non-epoxy end), the peak loads of
300 kN (67.3 kips) and 362 kN (81.4 kips) were obtained
from the bending and shear tests, respectively, providing
Type of repair
Surface mounted rods
Surface mounted rods
Table 2 Repair type and percentage of loading regained.
Percent of initial bending
Percent of initial shear
Fig. 8 Premature crack on beam 1 end A.
recovery of 96 % in bending and 82.1 % in shear. In this
case, while end B is better in final shear load, end A ties in
bending load and does better in percent recovery. However,
once again this is believed to be contributed to having to
move the reaction point in for end A especially since this is
the only end for which the end without epoxy exceeds the
epoxy-injected end at all.
The data for beam 2 repaired with glass FRP follow what
was initially expected, that is, end B (epoxy-injected end)
had a peak load of 222 kN (49.9 kips) and 481 kN (108
kips) during bending and shear tests, respectively, resulting
in a recovery percentage of 77 % in bending and 101.6 % in
shear. This only outperformed itself on a test after
postdamage repairs. End A on this beam performed reasonably
well but not nearly as well as end B. The repairs to end A
provided a peak bending load of 162 kN (36.5 kips) and a
peak shear load of 374 kN (84.1 kips) with a percent load
return of 64.6 % in bending and 85.9 % in shear.
The beam (Beam 3) repaired with surface mounted rods
also behaved as was expected with the epoxy-injected end.
The epoxy-injected end (end B) was able to provide a peak
load of 218 kN (49.1 kips) during bending testing and a
peak load of 328 kN (73.8 kips) during shear testing,
recovering 69.9 % of its initial bending load and 67.8 % of
its initial shear strength. While end A for beam 3 only
provided a peak bending load of 170 kN (38.2 kips) and a peak
shear load of 247 kN (55.4 kips) and was therefore able to
recover 60.1 % in bending and 40.8 % in shear.
Figures 9, 10 and 11 show that during the bending tests
both of the ends (epoxy or no epoxy) varied from their initial
lines by approximately the same amount as each other of the
course of the test. However, the graphs of the shear testing
reveal a large difference between the injected and
noninjected ends. The epoxy-injected ends were able to keep
their deflection much closer to that of the original test than
the ends without the injection.
Another important thing that the graphs reveal is the load
(the peak of the post-repair test) a beam still supported after
the loading cycle ended on its shear test. When the peaks of
the repaired tests are compared to the peaks of the initial
tests it can be seen that the initial shear strength was not
re-gained after the FRP repair. However, the effect of adding
the distribution plate was prominent. The significantly larger
shear strength was obtained for the repaired beam compared
with the initial shear strength (Fig. 11). It is the only one that
showed potential for an increase in strength of the beam.
One of the two tests with a distribution plate showed an
increase in shear strength of 43 % (end B) and the other test
(end A) did not show an increase; however, that end after
two different failure runs was noticeably deteriorated to the
point of almost being unusable. The two different failures
also made the results of the end showing strength increase
much more impressive. It was not stabilized, but it still
outperformed all other methods.
5. Summary and Conclusions
From the data of undamaged and damage-repaired
prestressed concrete girders described in this paper, it can be
concluded that epoxy injections as an aid to repair methods
provide a degree of help over beam ends not injected with
epoxy. The epoxy supplies an amount of stiffness and
stability to an end that has been injected. In two of the three
cases the data show that the epoxy-injected beam end
outperformed the regular non-injected end in every facet. In the
(d) Shear test (end B)
0.1 0.2 0.3 0 0.2 0.4 0.6 0.8
Displacement [in.] Displacement [in.]
Fig. 10 Data from bending and shear tests of beam 2.
(a) Bending test (end A)
(c) Bending test (end B)
(b) Shear test (end A)
Fig. 11 Data from bending and shear tests of beam 3.
remaining case there was equality in the bending load
achieved while the injected end provided a higher overall
shear strength, supporting the conclusion that epoxy
injections as an aid to repair methods provide a degree of help
over beam ends not injected with epoxy.
Reviewing the different repairs, it was concluded that the
carbon FRP has the greatest amount of stiffness recovery and
the glass FRP has the highest percentage of overall strength
recovery. Finally it was found that the use of a distribution
plate along with the surface mounted rods provides the
greatest potential for strength recovery, or in the best case
scenario, strength gain. To validate the concept, however,
numerical analyses should be carried out as a future study.
The research presented in this paper was supported by the
Oklahoma Turnpike Authority and by the National Research
Foundation of Korea (NRF) Grant funded by the Korea
government (MEST) (No. 2012005905). The opinions,
findings and conclusions in this paper are those of the authors
and do not necessarily represent those of the sponsors.
This article is distributed under the terms of the Creative
Commons Attribution License which permits any use,
distribution, and reproduction in any medium, provided the
original author(s) and the source are credited.