ACI JOURNAL TECHNICAL PAPER

Title no. 80-29

Serviceability Behavior and Failure Mechanisms of Concrete

Inverted T-Beam Bridge Bentcaps

by SherAli Mirza and Richard W. Furlong

Structural behavior of inverted T-beams differs from that of the more

conventional top-loaded standard T-section. The response of pre-
stressed members may be significantly different from that of rein-
forced members. Due to the lack of recommendation available for the
design of beams to resist torsion, this study was conducted to provide
data for designing both conventional and prestressed concrete in-
verted T-beam ben/caps with special allention to web reinforcement,
both the amount and location of stirrups; service load cracking in re-
sponse to equal and opposite torsions due to traffic loads; strength
and stiffness of web and flange as the amount of combined flexure,
shear, and torsion increases to the maximum capacity. Strength and
serviceability behavior of inverted T-beams bentcaps is also exam-
ined with emphasis on experimental study.

Keywords: bents; bridges (structures); caps (supports); failure mechanisms;

flanges; prestressed concrete; reinforced concrete; serviceability; stirrups; T-
beams.

Inverted T -beams provide a natural and popular

structural form for use as a girder to support precast
beams. The flange of the inverted T serves as a shallow
shelf to support precast beams while the stem of the in-
verted T, rising to the height of the supported beams,
provides the needed depth to sustain flexure and shear
forces. Fig. 1 illustrates how an inverted T-beam per-
mits increased headroom beneath the bridge deck and
at the same time presents an attractive appearance by
Fig. 1 - Some inverted T-beam bentcap girders in
keeping to a minimum the visible size of transverse place
supporting elements.
The structural behavior of the inverted T shape dif-
fers from that of the more conventional top-loaded The maximum, total vertical load on a bentcap oc-
standard T-section in the following ways: curs when stringers on both sides of the web simulta-
1. Loads that are introduced from precast beams into neously support traffic. As traffic moves across an in-
the bottom rather than into the sides or the top of the verted T-beam bentcap, the stringers on one side of the
web must be supported by stirrups acting as hangers to web will create a web twist opposite to that which oc-
transmit vertical forces into the body of the web. curs when the stringers on the opposite side of the web
2. Flange reinforcement perpendicular to the web is are loaded by the traffic. Flange forces on one side of
necessary to deliver flange forces to the hangers in the the web are significantly larger than forces on the op-
web.
Received May 4, 1981, and reviewed under Institute publication policies.
3. Application of flange forces occurs at a greater Copyright © 1983, American Concrete Institute. All rights reserved, including
distance from the centerline of the web, thereby creat- the making of copies unless permission is obtained from the copyright propri-
etors. Pertinent discussion will be published in the May-June 1984 ACI JOUR·
ing greater twisting forces on the web. NAL if received by Feb. I, 1984.

294 ACI JOURNAL I July-August 1983

ACI member SherAli Mirza is an associate professor of civil engineering at
PHYSICAL TESTS
Lakehead University, Thunder Bay, Ont. His industrial and consulting experi- The behavior of inverted T -beam elements subjected
ence includes high-rise buildings, industrial structures, and composite steel- to combined flexural and torsional loading was ob-
concrete bridge design. Dr. Mirza has conducted research on structural con-
crete in areas of safety and limit states design, bridge bentcap girders, and bond-
served in 27 tests to failure performed on three rein-
creep and combined torsion in beams. forced concrete and four prestressed concrete speci-
mens. The specimens represented one-third scale
Richard W. Furlong, FACJ, is a professor of civil engineering at the Univer-
sity of Texas at Austin. With a lifelong interest in application of theory to de-
models of inverted T-beam bentcap girders similar to
sign, he has served on jour AC!technical commitlees: Commitlee 340, Strength those used by the Texas Department of Highways and
Design Handbook; joint AC!-ASCE Commillee 428, Inelastic Behavior of Public Transportation (TDHPT). The prototype bent-
Reinforced Concrete Structures; joint ACI-ASCE Commitlee 44/, Reinforced
Concrete Columns; and Commitlee 104, Notation. He has served as ACI Di-
cap girders had been proportioned to support dead load
rector. stringer reactions near 90 kips (400 kN) and live load
plus impact reactions near 135 kips (600 kN) at each
stringer location. For one-third scale models, the load
similitude factor becomes one-ninth, such that nominal
posite side of the web, and the passage of traffic tends dead load stringer reactions on test specimens could be
to make such twisting an alternating phenomenon. taken as 10 kips (44 kN) and live load plus impact re-
Conventional reinforced concrete members tend to actions could be taken as 15 kips (67 kN).
crack at a relatively low level of principal tension stress, Four basic test arrangements were employed to ob-
thereby redirecting tensile force paths through rein- serve the response of specimens to:
forcement in cracked portions. The precompression of 1. Loads that created positive moment combined
a concrete cross section can reduce and even eliminate with torsion and shear.
tension stress. Consequently, the response of pre- 2. Loads that caused negative moment combined
stressed concrete members may be significantly differ- with torsion and shear.
ent from that of conventionally reinforced concrete 3. Loads that created pure torque in torsionally
members subjected to the same conditions of combined clamped cantilevers.
torsional and flexural loading. 4. Loads that caused pure torque in torsionally
The Standard Specification for Highway Bridges by clamped single spans. The principal variables employed
American Association of State Highway and Transpor- in the experimental study were the reinforcement de-
tation Officials (AASHTO) contain general design cri- tails associated with web shear and torsional strength.
teria for reinforced and prestressed concrete members.' Loading procedures involved careful monitoring of
The 1977 edition does not contain recommendations for response to service load forces before loads were in-
the design of beams to resist torsion, and the guidance creased enough to create failure conditions. The nomi-
associated with bracket design is cursory. Thus, evalu- nal span-depth ratios for prototype inverted T-beam
ations of capacity to resist torsional forces on rein- bentcap girders appeared to be 4 to 8 for spans sup-
forced concrete members and the capacity of inverted ported at each end and 2 to 4 for cantilevered mem-
T-beam flanges acting as brackets was based on regu- bers. Similar proportions were maintained for the ar-
lations contained in the 1977 edition of the American rangement of test specimens. Test loadings representa-
Concrete Institute (ACI) Building Code. 2 The design tive of stringer forces on the prototype girders required
advice concerning the assignment of an effective region about 12 ft. (3. 7 m) long laboratory specimens. Instead
over which the transverse reinforcement in the flange of fabricating 12 ft long specimens, each to be ioaded
can be expected to help support a concentrated load one time to failure, considerable savings were envi-
applied to the flange was taken from recent research. 3· 5 sioned from the use of fewer specimens, each subjected
The evaluation of torsional strength of prestressed con- to several different tests until failure in local regions.
crete beams cannot be based on either the ACI Code or Failure regions involved approximately one depth of
the AASHTO Specifications and, therefore, constitutes the specimen, or something near 2 ft in length of the
one of the goals of the research reported here. Several specimen. Consequently, seven specimens having ap-
other aspects of the structural behavior of inverted T- proximately 24 ft (7 .3 m) lengths were fabricated, and
beam bentcap girders cannot be classified to fit existing 27 individual tests were conducted altogether.
specifications for the design of reinforced and pre- All specimens had the same nominal cross-sectional
stressed concrete members. These aspects are: dimensions except for flange depth of Specimen TCI,
I. Web reinforcement, both the amount and location which was one in. shallower than other specimens. A
of stirrups. nominal Vs in. (16 mm) concrete cover of reinforce-
2. Service load cracking in response to equal and op- ment in test specimens corresponded to approximately
posite torsions due to traffic loads. 2 in. cover for the prototype bentcaps. Most of the
3. Strength and stiffness of web and flange as the specimens were furnished with closed rectangular stir-
amount of combined flexure, shear, and torsion in- rups. Test Regions TC21 and TC24 were provided with
creases to the maximum capacity. Hence, this study was additional closed triangular stirrups placed alternately
conducted to provide data for designing both conven- with closed rectangular stirrups at the bottom of the
tional and prestressed concrete inverted T-beam bent- web. Reinforcement near the upper corner of the flange
caps with special attention given to the enumerated as- was welded to provide anchorage of these bars. The
pects. specific cross section and reinforcement details are
ACI JOURNAL I July-August 1983 295
Table 1 - Reinforcing details of test sections* shown m Table I and l'Jg. 2.
Longitudinal bars Specimens TP3 and TP5 were prestressed such that
Test number Top Bottom Web stirrups' the centroid of a total initial prestressing force of 248
TCII,TC22,TC25,TC71 10 #6 II #6 #3@ 6 in.
kips (1100 kN) was 2.67 in. (68 mm) below the centroid
TCI2,TCI3,TC72 10 #6 6 #6 #3@ 4 in. of the cross section to prestress the specimen for posi-
TC21,TC24 10 #6 II #6 #3@ 6 in. + tive flexural loading. Specimens TP4 and TP6 were
#3!1@ 6 in.
TC23 10 #6 II #6 #3 @ 3 in.
prestressed to resist negative flexural loading with the
TC73 10 #6 II #6 I #3@ 4 in. centroid of a total prestressing force .of 214 kips (952
TP31,TP32,TP33 2 #4 4 #3 2 #3@ 4Y' in. kN) being 6.83 in. (173 mm) above the centroid of the
TP51,TP55 2 #4 4 #3 2 #3@ 9 in.
TP52 2 #4 4 #3 #3@ 4Y, in.
cross section. All of the prestressed concrete specimens
TP53,TP54 4 #4 8 #3 2 #3 @ 4'12 in. were post-tensioned with grouted tendons.
TP41,TP43 2 #4 4 #3 2 #3@ 4 in. Concrete mixes were designed on the basis of
TP42,TP65 2 #4 4 #3 #3@ 3 in. TDHPT specifications used for inverted T-beam bent-
TP61,TP63 2 #4 4 #3 #3@ 2 in.
TP62,TP64 2 #4 4 #3 #3@ 4 in. caps. The model proportioned at one-third scale em-
•Specimens TP3 and TP5 contained three prestressing tendons, each having ployed a coarse aggregate with a maximum size of Vs
10!4 in. diameter wires (total initial prestressing force = 248 kips), in addition in. (10 mm). High-early-strength Type III portland ce-
to reinforcing bars shown above. Similarly, specimens TP4 and TP6 were pre-
stressed with two tendons, each having 13'.4 in. diameter wires (total initial ment was used for concrete. The average values of
prestressing force = 214 kips).
'Closed rectangular stirrups unless noted otherwise. compression strength of 9 to 15 control cylinders per
I in. = 25.4 mm. specimen are given in Fig. 2.
Loads were monitored by readings from the load
cells. The distribution of load within reinforced con-
crete specimens could be traced by means of strain

8" 8"

~ ~ t 'n
~ E
PI?ESTRES~<
TENDONS
/+ ~
e•6.83" /NTROID
·- <Q

~-cP
N
2·3- c £:.STIRRUPS 2"3
/rc21 &. TC24
"31!'3 ONLY J./
.2"'3,"'- /"{ T ~3 ~
~4-
(f)
a:
\/_
'
w
-I
"-
/. ... ~
(.)I-
1-0
<of'.-
\a 2 ... 3 •
~2@>3TC73 22
3 OTHERS
SPECIMENS TP4 & TP6
22"
SPECIMENS TCI, TC2 ,TC7

8'

..
I
MATERIAL STRENGTH KSI
Fv
f~ .2 "3 "4 "'6
<X> =-~ TCI 4650 50 70 62 62
0 --
TC2 4.550 50 70 62 62
t")
TP3 5.480 50 70 62
I 2.3

~-$-/
PRESTRESS! NG CENTROID TP4 5.130 50 70 68 -
TENOONS,e·2.67 TP5 5.530 50 70 64 -
_'--..+ .3113 ;.,~ TP6 5.160 50 72 64 -
2 1-4 w TC7 4,560 50 82 68 62
~

L 4- . . "-$-·.
N
(}) PRESTRESSING STEEL F. •240
,..._ i'- I KSI•6.9MPa 1 I IN•2.54 MM.
i-t)
•
'211'3
22'
I SPECIMENS TP3 TP5 WELD ALL
SPECIMENS
BRACKET ANCHORAGE DETAIL

Fig. 2 - Details of test specimens

296 ACI JOURNAL I July-August 1983
gages attached to the reinforcement of the specimens.
Seventy to more than 100 strain gages were used for
each specimen. The strain gages on web stirrups were
placed near the same level as the top face of the flange
and on top flange reinforcement near the face of the
web. Longitudinal flexural steel received some gages,
particularly those bars located in the extreme corners of
the cross section.

SPECIMEN DESIGN
The maximum spacing of stirrups for reinforced
concrete specimens can be determined on the basis of
stirrups acting as hangers. The design ultimate load R
for one stringer in accordance with AASHTO
specifications 1 becomes
5
R = 1.3 [D + 3 (L + /)] (1)

For laboratory specimens having stringer dead load of

10 kips and stringer live load plus impact load of 15
kips, R is equal to 45.5 kips. The hanger capacity
within a stringer spacing should be as great as R. Thus

(2) (d) Live Load Increased to Foilure

Fig. 3 - Test load sequence

where A/2 cross-sectional area of one stirrup leg
h yield strength of stirrups
s stirrup spacing mm 2). At least this amount of transverse reinforcement
es spacing of stringers was provided at top of the flange for all specimens. In
addition, all reinforced concrete specimens (TC I, TC2,
To reduce the probability of hanger failure in the test TC7) contained supplementary "bracket" reinforce-
specimens, stirrup spacings were made small enough for ment at midheight of the flange providing an addi-
most specimens so that the computed values of R were tional 0.33 in. 2 (210 mm 2) of steel for A,.r·
more than 45.5 kips (202 kN). Due to prestressing, the longitudinal precompression
In flanges transverse reinforcement to resist bracket of the cross section should improve both the flexural
failure was made large enough so that the local bracket shear and bracket action of flange by delaying the for-
weakness would not interfere with observations of mation of tension cracking, thereby increasing the ef-
combined flexure and torsion behavior for overall fective amount of shear force resisted by concrete
specimen. In accordance with the ACI Building Code 2 alone. Hence, the transverse reinforcement across mid-
and a design ultimate load R( = 45.5 kips), the trans- height of the flange was omitted for prestressed speci-
verse flange reinforcement Av1 can be estimated as mens. The top transverse reinforcement used for the
flange remained unchanged. The stirrup spacings for
A=-
vf
1 [ R
64 6.5 (1 - 0.5 aid) c/> .JJ:
-bd
e
l (3)
prestressed concrete specimens were closer than those
for reinforced concrete specimens and the computed
values of R exceeded 45.5 kips (202 kN) for most pre-
in which stressed specimens.
a distance from face of web to center of bearing
plate [4 in. (100 mm) for all specimens] TEST LOADING SEQUENCE
b. effective width of flange for each load R Specimens subjected to flexure combined with tor-
ct> capacity reduction factor for shear = 0.85 sion had eight points of loading representing eight
f: must be expressed in units of psi stringer reactions. Dead loads of 10 kips (44 kN) were
applied to all eight load points, indicated by open rec-
A value of b. equal to the width of bearing plus five tangles in Fig. 3(a). Loads of 15 kips (66 kN), repre-
times the distance a has been recommended. 5 It would senting live load plus a 25 percent overload on live load
give b. = 6 + 5 X 4 = 26 in. (660 mm), but with an due to impact, were then added to four of the 10 kip
effective stringer spacing of about 23 in. (580 mm) the dead loads indicated by cross-hatched rectangles in Fig.
b. value must be reduced to 23 in. The amount of Av1 3(b). The 15-kip loads due to live load plus impact will
for the 7 in. (180 mm) thick flange [d = 6.2 in. (158 be hereafter referred to as live loads. The cycled tor-
mm)] is calculated from Eq. (3) to be 0. 78 in. 2 (500 sional loading proceeded from the load condition
ACI JOURNAL I July-August 1983 297
shown in Fig. 3(b). One complete cycle of load re- side of the web and on the other side of the web a load
versals would consist of the load stages shown in Fig. P representing one stringer dead load plus a variable
3(b), followed by that in Fig. 3(a), followed by that in stringer live load. It should be recalled here that the
Fig. 3(c), followed again by that shown in Fig. 3(a), stringer dead loads (10 kips each) were maintained at all
and a return to load condition illustrated in Fig. 3(b). load points both sides of the web, while stringer live
The cyclic loading was repeated until the deformed loads were added to pairs of load points only each side
condition and reinforcement strains showed negligible of the web. This was done to subject the specimen to
changes from the previous identical live load condition. torsion combined with flexure and flexural shear. The
This procedure was adopted to observe the nature of stringer live loads were 15 kips at service loads. Similar
cracks that might be expected under service load con- hanger stress variations were recorded for other test re-
ditions involving torsion reversals as traffic crosses the gions.
bentcap. After stabilization of deformations under ser- Transverse reinforcement in the top of the flange in-
vice load conditions, live loads were increased in stages dicated that stresses as high as 15 to 18 ksi (l 00 to 125
at locations indicated by solid rectangles in Fig. 3(d) MPa) were developed under service load conditions,
until failure took place. The test dead load in Fig. 3(d) but no flexural cracking specifically attributed to the
was maintained at 10 kips for stringer reactions that tensile behavior of flange steel could be observed. It is
were not increased. At each load stage a complete set of probable that the crack which appeared at the junction
readings was made and visible cracks were identified. of the web and the flange extended diagonally down-
ward within the specimen toward the intersection of the
vertical hanger and the flange reinforcement. However,
Behavior of reinforced concrete specimens there was no evidence of distress in the flange under
All of the reinforced concrete specimens with stirrup service loads.
spacings 4 in. (100 mm) or greater revealed under the Some flexural cracking could be observed in regions
dead load condition some cracking on the face of the of maximum moments in all of the reinforced concrete
web at the intersection of the flange. These cracks were specimens during the first cycle of live load plus dead
parallel to the flange and were visible only with bright load. Strain gages on the longitudinal reinforcement
light. Subsequent applications of the live load caused indicated that the maximum stress under repeated live
initial cracks to widen and to extend. Strain gage read- loads was in the order of 25 ksi (170 MPa). If the in-
ings indicated that the stirrups adjacent to applied live fluence of torsion on longitudinal stress in the steel is
loads yielded either under the initial application of live ignored, the nominal stresses in the flexural reinforce-
loads or under cyclic live loading. However, stirrups ment would have reached 20 ksi ( 140 MPa) theoreti-
located two or more spaces away from the point of load cally. Flexural cracks that initiated on one side of the
applications exhibited low stresses and the cracks men- flange generally propagated entirely across the flange
tioned earlier did not propagate far from the bearing when the torsional reversal of live loads was applied.
plates. In most cases these cracks tended to propagate The service load performance of the test specimens as
along the top face of the flange, diagonally away from designed appeared to be satisfactory in all aspects, with
the face of the web, an inch or so each side of the bear- possible exception of the high stress observed in stir-
ing plate. Fig. 4 shows the variation in hanger stresses rups near the point of application of load. However,
in Test Region TC22 with one stringer dead load on one even for the stirrups that yielded under the application

FAILURE LOAD=52K
F.,·70.5 KSI

60 60
STEEL
STRESS STEEL
KSI STRESS
KSI
40 40

20 20

20" 20" 20"

HALF SPAN TEST REGION
REA TION
(o) REINFORCED CONCRETE TEST REGION TC22 \b) PRESTRESSED CONCRETE TE.ST REGION TP52

Fig. 4 - Hanger stress distribution

298 ACI JOURNAL I July-August 1983
of live load, subsequent application of live load did not to positive moment test regions. It should be recalled
appear to generate higher st~ains in the once-yielded that the amount of stirrup reinforcement in test regions
stirrups. The measured strains of yielded stirrups were TP31 and TP53 was approximately double that used
not large enough to indicate significant strain harden- for the reinforced concrete specimens. Even though the
ing of steel. Apparently, the "rounded" characteristics stirrup size and spacing of TP52 was similar to that
of the stress-strain behavior of the steel in the initial in- which permitted some yielding of stirrup steel under
elastic stages of strains involved just enough "strain service loads on reinforced concrete specimens, there
hardening" to permit the reinforcement to stabilize un- was no evidence of overstress on the prestressed con-
der successive applications of "yield" strain at service crete member as indicated by Fig. 4. Strain gage read-
load conditions. ings showed that the maximum hanger stress under
cyclic live loads reached only 12 ksi (83 MPa). A com-
Behavior of prestressed concrete specimens parison of hanger stress variation in reinforced and
Flexural reinforcement of prestressed concrete speci- prestressed concrete specimens in Fig. 4 clearly demon-
mens was made adequate only to resist nominal ulti- strates superior performance of prestressing under ser-
mate test loads and to satisfy nominal requirements for vice load conditions. At ultimate loads, however, the
service load flexural stresses without torsion. The geo- hanger stress distributions in conventional and pre-
metric properties and details of test specimens are given stressed concrete inverted T-beams do not seem to dif-
in an earlier section and in Fig. 2. The specimens sub- fer significantly as indicated by Fig. 4.
jected to positive moment plus torsion (TP3, TP5) were Strain gages attached to transverse bars along the top
prestressed such that precompressive stress along the of the flange indicated some service load stresses as
cross section varied from 1475 psi (10 MPa) at bottom high as 24 ksi (165 MPa). This stress was associated
to 930 psi (6.4 MPa) at centroid to 29 psi (0.2 MPa) at with some cracking in the corner between the face of
top. Under service loads these stresses changed to av- the web and the top of the flange.
erage values of 130, 930, and 2250 psi, (0.90, Specimens TP4 and TP6 that were subjected to neg-
6.4, and 15.5 MPa), respectively, in compression. Sim- ative moments combined with torsion showed no
ilarly, for specimens subjected to negative moment and cracking, either flexural, torsional, or hanger, under
torsion (TP4 and TP6), the stresses due to prestressing service load conditions shown in Fig. 3. There were
force ranged from 378 psi (2.6. MPa) in tension at bot- enough stirrups to sustain hanger loads without reveal-
tom fiber of the cross section to 805 psi and 2760 psi ing tension cracks along the face of the web. In addi-
(5 .6 MPa and 19.0 MPa) in compression at the cen- tion, no cracking was observed due to transverse flex-
troid and the top fiber of the cross section. These ure of flange. It seems that the precompression of the
stresses changed to average values of 1535 psi (10.6 stem near the top of the flange tended to restrain the
MPa) compression, 805 psi (5.6 MPa) compression, formation of cracks in the web and the flange. Strain
and 400 psi (2.8 MPa) tension, respectively, under ser- gages on stirrups and flange transverse bars tended to
vice load condition. verify the absence of tensile strain high enough to cre-
Prestressed concrete specimens were designed for ate visible cracks.
probable failures in the web as a combination of flex- The absence and delay of initial crack formation in
ural and torsional stresses reached the capacity of the prestressed concrete specimens made their apparent
compressed concrete. Hence, the flexural prestressing performance under service load seem superior to that of
was kept light so that the influence of torsion on lon- specimens reinforced only with deformed bars. The
gitudinal stresses, cracking, and compression capacity prestressed concrete specimens behaved in a more sta-
would be more readily apparent. In addition, some test ble manner with smaller variations in response to alter-
regions contained an intentional excess of stirrups in the nate cycles of live loads. Strain gages that had been at-
web to avoid failure of stirrups acting as hangers near tached to the reinforcement tended to encourage the
the stringer loads. impression that prestressed concrete specimens behaved
The loading scheme for prestressed concrete speci- monolithically under the influence of combined flexure
mens subjected to positive flexural and torsional loads and torsion at least until service load levels had been
(TP3 and TP5) was the same as that for reinforced reached. Service load analysis of the prestressed con·
concrete specimens subjected to similar loads. When crete specimens, therefore, could be based on relations
specimens were loaded initially with dead load at all for an elastic continuum. In contrast, the behavior of
eight load points, no cracks were observed on the sur- ordinary reinforced concrete specimens involved crack-
face of the specimens. On application of torsional live ing so extensive that the overall member should be con-
loads, a flexural crack could be observed in the maxi- sidered analytically as an assemblage of fractured
mum moment region. This crack closed after the test mechanisms even at service load levels of response.
live load had been removed. Recycling of the live load The comparison of behaviors of prestressed concrete
revealed few new cracks. Once service load cracking and ordinary reinforced concrete specimens in the re-
had occurred, recycling of live loads created little ported tests is not exactly fair in that the prestressed
change in measurements of cracks and steel strains. concrete specimens contained considerably more stir-
No cracks appeared at the face of the web near the rup reinforcement than that which was used in the
top of the flange under service loads that were applied reinforced concrete specimens. But, then the pre-
ACI JOURNAL I July-August 1983 299
Table 2 - Hanger load analysis Crushing and

Mc--~JM
Effective
Hanger Stirrup* stirrup
Test yield load, /y, spacing,
number kips ksi in. 'Y
TC11 25 70 6 3.4
TC21 25 70 6 3.4
TC22 25 70 6 3.4
TC71 25 82 6 2.7 Numerous Iorge crocks in concrete,
TP51 40 70 4.5 4.3 Obvious stretching of steel.
TP52 45 70 4.5 5.1
TP53 55 70 2.25 2.5 (a) FLEXURAL FAILURE -POSITIVE MOMENT
TC12 25 70 4 1.7
TC24 17.5 70 6 1.9
TC25 35 70 6 5.3 {.Numerous Iorge crocks in concrete
TC72 35 82 4 2.4 "\Obvious stretching of steel.
TP41 50 70 2 1.7
TP42 42.5 70 3 2.6
TP61 50 72 2 1.7
TP62 45 72 4 4.2
Average 3.1
• All stirrups were made from Number 3 reinforced bars.
I k = 4.448 kN.
I ksi = 6.895 MPa.
I in. = 25.4 mm.
(b) FLEXURAL FAILURE- NEGATIVE MOMENT.
stressed concrete specimens were underdesigned for
longitudinal prestress compared to longitudinal rein-
Steel stretches and concrete crocks become larger.
forcement in ordinary reinforced concrete specimens. Failure occurs only after compressive resistance is
No adverse effect from the intentional underdesign of exhausted os revealed by spoiling or flaking on the
flexural prestress was observed. compressive surface.

Serviceability criteria for hanger performance Fig. 5 - Flexural failure modes

Even though all arrangements of stirrups acting as
hangers supported the nominal ultimate loads, some
yielding of hangers was observed under nominal service Av 3 (PD + PL)
(5)
loads. The force in any specific hanger is a function of s /y (B + 3a)
applied load, base plate size, and some "effective dis-
tribution zone" that incorporates hangers nearest the STRENGTH OBSERVATIONS
applied load. Thus, a relationship between the applied Failure mechanisms
load and the hanger resistance can be expressed as The strength of concrete inverted T -beams is limited
to the strength of the weakest of several possible com-
Avfs ponents or combinations of components that partici-
(4)
2 s pate in the retention of the applied loads. The compo-
where nents will be defined in terms of six failure mechanisms
Avfs/2 force in one leg of stirrup acting as a or failure modes identified from the physical tests.
hanger Among the 27 tests conducted very few involved fail-
s hanger spacing ures easily attributable to only one component, but
Pv + PL = applied load (stringer load) failure generally took place after weakness was appar-
B base plate width ent in more than one failure mode. The strength of the
a distance of applied load from face of web entire cross section of a specimen can be interpreted in
'Y effective width coefficient, an unknown terms of resistance to flexure, flexural shear, or tor-
sion. Any of the three or combination of three will be
Data from all torsion-plus-moment tests listed in Table reached only if there exists adequate local strength for
2 contain values of 'Y for fs equal to yield strength of hanger action, flange punching shear, and flange
hangers and Pv + P 1 values corresponding to the mea- bracket response. Sketches of typical failure modes are
sured load at first observed hanger yielding. The values provided in Fig. 5 through 10 with descriptions given in
of 'Y varied from a low of 1. 7 with 2 in. (50 mm) stir- the following paragraphs.
rup spacing to a high of 5.3 with 6 in. (150 mm) stirrup 1. A cross section is considered to have failed in
spacing. The average of all values computed was 3.1. flexure when its resistance to flexural deformation be-
As a serviceability control, it seems desirable to limit gins to decrease. Under the application of large flex-
the hanger stress to values less than Y, fY under service ural deformations, reinforcing bars and prestressing
load conditions. Taking the coefficient 'Y as 3.0 ap- strands tend to yield without a reduction in tensile
proximately to represent the average measured value, force. Simultaneously, the resistance of concrete and
Eq. (4) can be modified to serve as a serviceability con- compressive reinforcement can remain relatively con-
trol stant, but tl}e amount of compressive "yielding" is
300 ACI JOURNAL I July-August 1983
 Stirrups that cross the crock yield
before o shear failure. /
/
Fig. 6- Flexural shear failure mode /
/
/

limited by the amount of stress that can be redistri- TORSION-MODE I

buted by concrete before concrete cracks and spalls off Compression in Bottom
at the surface of maximum compressive strain as shown
in Fig. 5.
2. Flexural shear failure for specimens with an effec-
tive shear span as short as that employed for these tests
involves the yielding of all stirrups that cross a large
crack which extends diagonally along the side faces of
the web until the shear strength or compressive strength
of uncracked concrete is exhausted (Fig. 6). The maxi-
mum shear span near supports for any test region was
only 18 percent greater than the effective depth of the
member. Additional strength due to short shear span is
at least partially offset by application of the stringer
loads near the bottom rather than the sides or the top TORSION- MODE 2
of the web. Compression in Top
3. Torsional distress in concrete beams appears in di-
agonal cracks that extend in a spiral pattern from one Fig. 7- Tqrsion failure modes
face of the member to an adjacent face. Failure begins
in the form of diagonal tension cracks across which
both longitudinal and stirrup reinforcement must
transmit the tensile forces that were lost where the con-
crete cracked. The diagonal cracks extend in length and
in width as torsional forces increase until reinforce-
ment across the crack yields. If the anchorage strength
of reinforced bars is not lost due to edge cracking, tor-
sional deformation continues until there is a compres-
sion failure along the face nearest the center of tor-
sional rotation (Fig. 7).
4. Hanger failure of stirrups is revealed by the verti-
cal separation that occurs between the flange and the Stirrups acting as hangers yield near load
web at the top of the flange. The separation begins as points
a local phenomenon near the stringer bearing plate, but Fig. 8 - Hanger failure mode
as the stirrups closest to the bearing plate yield, the
flange deflects and causes more hangers to share the
concentrated stringer load. The failure occurs after all 6. The bracket failure mode is used here to identify a
available stirrups acting as hangers have yielded pro- local loss of resistance to load because the flange acting
vided that the flange is strong enough to distribute lon- as a bracket tends to deform outward and away from
gitudinally the stringer loads (Fig. 8). the web while also deforming downward from loss of
5. Flange punching failure can take place if stringer shear strength along the face of the web. As the trans-
forces are large enough to "punch out" the truncated verse bars acting in flexure across the top of the flange
pyramid of concrete beneath a bearing pad. The failure yield, the flange resistance in "shear friction" along the
is evidenced by the appearance of diagonal tension face of the web is unable to sustain the stringer forces
cracks emanating from the edges of bearing plate as in- and may allow sliding of the flange downward along
dicated in Fig. 9. the face of the web (Fig. 10).
ACI JOURNAL I July-August 1983 301
 side of the web and the bottom side of the flange on the
opposite side of the web to form couples. Thus, the
loads that formed torsional couples were applied 16 in.
apart in the transverse plane of the specimen for all
pure torsion tests, except for test regions TC13, TP32,
TP43, and TP54, where they were placed 14 in. apart.
The specimens were designed to support nominal
stringer reactions of 45.5 kips (200 kN) each at ulti-
mate. Table 3 indicated that this load was supported at
each live load bearing point for all tests subjected to
torsion plus flexure with the exception of Specimen
TC 1 which had a flange depth shallower than that of
the other specimens. However, the strength results that
were listed in Table 3 included six failures that were at-
tributed to hanger failure. In addition, hanger weak-
ness was felt to be significant for the failure of three
other test regions. Fig. 11 indicates how the flange re-
gion of an inverted T-beam can be cracked away from
the upper portion of the web after stringer loads exceed
Fig. 9 - Flange shear or punching failure mode the strength of stirrups acting as hangers.
With the exception of punching and bracket failures,
Test results and discussion the specimens exhibited considerable reserve strength
A summary of test results is given in Table 3. Theta- after the maximum load was attained. The reserve
ble displays for each test the recorded load at which strength could be realized only after propagation of
stirrups reached yield stress, the load at which cracks wide cracks as internal resistances were distributed to
opened to as much as 0.025 in., here referred to as wide less damaged regions of the specimen. The local and
cracks, and the ultimate load that could be resisted by sudden failure associated with punching and bracket
the specimen. The spacing of No. 3 web stirrups is shear precluded any capacity for redistribution of
given, and in the last column the letter code indicates stresses. The explosive nature of punching failure was
the most prominent mode of failure. characterized by wide cracks and, finally, by crushing
It should be noted that the loads shown in Table 3 and spalling of concrete as the flange failed due to
were applied at four live load bearing points for all the downward movement of a truncated pyramid beneath
torsion plus flexure tests. For these tests an additional a stringer load. This is illustrated by failure of test re-
set of 10 kip dead loads was maintained constant on gion TP41 shown in Fig. 12. It should be noted here
four bearing plates at the opposite side of the web from that adequate local strengths for hanger action, flange
each live load bearing plate as shown in Fig. 3(d). On punching shear, and flange bracket response are re-
the other hand, loads for pure torsion tests were ap- quired before full strength in combined torsion, flex-
plied in equal amounts to the top of the flange on one ural shear, and flexure can be realized.

After flange
flexure allows
a separation at
the face of the web,
the shaded region
con foil in loss of
"friction." Only possible
with several odjocent loads.

Fig. 10 - Bracket failure mode

302 ACI JOURNAL I July-August 1983
 Table 3 - Strength results
Yield Wide crack Ultimate Stirrup
Test load, load, load, spacing Failure mode
number kips kips kips (number 3) (see code below)
Torsion plus positive moment
TCII* 25.0 35.0 39.0 6 in. H
TC21* 25.0 37.5 52.5 6 + extra l1 @ 6 in. T,S
TC22* 25.0 37.5 47.5 6 in. H
TC23* 25.0 No record 57.5 3 in. p
TC71* 25.0 35.0 52.5 6 in. H
TP31 45.0 30.0 52.0 Pairs @ 4.5 in. T,F,P
TP51 40.0 No record 55.0 Pairs@ 9 in. H
45.0 52.0 4.5 in. H
I
TP52 45.0
TP53 55.0 52.5 57.5 Pairs @. 4.5 in. T,S,F
Torsion plus negative moment
TC12* 25.0 37.0 39.0 4 in. T,S,F
TC24* 17.5 30.0 55.0 6 + extra l1 @ 6 in. H,P
TC25* 35.0 25.0 55.0 6 in. H
TC72* 35.0 35.0 52.5 4 in. T,S,F
TP41 50.0 17.5 50.0 Pairs@ 4 in. p
TP42 45.0 42.5 52.5 3 in. H,P
TP61 50.0 50.0 55.0 2 in. p
TP62 45.0 40.0 50.0 4 in. T,S,F
Pure torsion in torsionally clamped cantilevers
TC13* 25.0(22.0)' 20.0(17.5)' 32.5(28.5)' 4 in. T
TC73* 25.0 20.0 32.3 4 in. T
TP32 No record 25.0(22.0)' 38.8(34.0)' Pairs @ 4.5 in. T
TP33 32.0 30.0 34.0 Pairs @ 4.5 in. T
TP55 No record 25.0 30.0 Pairs@ 9 in. T
TP64 30.0 32.5 34.2 4 in. T
TP65 30.0 No record 33.0 3 in. T
Pure torsion in torsionally clamped single spans
TP43 No record No record 78.4 Pairs@ 4 in. H,P,B
TP54 No yield No record 95.0 Pairs @ 4.5 in. P,B
TP63 No record No record 81.0 2 in. p
*These specimens contained reinforcing bars only. All other specimens had prestressing steel in addition to reinforc-
ing bars.
'The values shown in parentheses for TC13 and TP32 are the hypothetical loads that could be applied 16 in. apart
in the transverse plane of the beam and correspond to other tests on torsionally clamped cantilevers. Other values in
tests TC13 and TP32 are the actual loads placed 14 in. apart.
Failure code: F = flexural failure H z hanger failure
S = flexural shear failure P = punching failure
T = torsion failure B = bracket failure

There were no flexure or flexural shear failures

alone, although each of these phenomena was observed
to be a part of the failure mode in six of the tests. It
was apparent both from the pattern of applied forces
and the response of the specimens that these modes of
failure were accompanied by a torsional mode of fail-
ure.
The maximum moment that was applied to rein-
forced concrete Test Regions TC12 and TC72 was con-
siderably lower than the estimated "pure" flexural ca-
pacity of each. Obviously, the longitudinal effect of
torsional loading was significant for both tests, because
some of the longitudinal bars yielded and large cracks
could be observed under maximum loads. On the other Fig. 11 - Failure of test region TC11 due to overload-
hand, the changes in flexural behavior due to torsion ing of hangers
were less apparent in prestressed concrete Test Regions
TP31, TP53, and TP62, even though the applied mo- continually change for each new condition of load. The
ments were within 10 percent of the ultimate flexural superposition of torsional forces is expected to alter the
capacity under flexure alone. The initial cracking mo- distribution of longitudinal forces more so in exten-
ment for prestressed concrete members was almost 40 sively cracked, reinforced elements than in prestressed
percent of the ultimate moment compared to only 7 beams with limited cracking.
percent for reinforced concrete specimens. After crack- The interaction of flexural shear and torsional shear
ing begins, the internal mechanisms that resist flexure appeared to be influenced significantly by concrete
ACI JOURNAL I July-August 1983 303
 Fig. 13 - Failure of test region TC72 due to combined
torsion, shear, and flexure

ment perpendicular to the web is necessary to deliver

Fig. 12- Failure of test region TC41 due to punching flange forces to the hangers in the web. Recommenda-
of flange tions contained in the North American standards can be
applied to some design aspects of inverted T-beams,
strength and concrete stiffness only prior to the devel- while other design aspects remain within the discretion
opment of diagonal cracking. After cracks developed, of the designer.
the concrete served primarily to equilibrate in compres- The behavior of inverted T-beams subjected to com-
sion the tensile forces on both transverse and longitu- bined flexure and torsion was observed from 27 tests to
dinal reinforcement. All test regions that failed in flex- failure conducted on three reinforced concrete and
ural shear combined with torsional shear developed ul- four prestressed concrete specimens. The possible fail-
timate loads at least as high as calculated "pure" flex- ure mechanisms are identified and described. Based on
ural shear capacity based on equations of the ACI this study, the inverted T-beams should be designed to
Building Code. 2 A typical failure of reinforced concrete have adequate strength against possible failure due to
inverted T-beam under combined torsion, shear, and flexure, flexural shear, torsion, hanger forces, flange
flexure is shown in Fig. 13. punching shear, and flange bracket action. Observa-
The seven pure torsion failures indicated in Table 3 tions from physical tests indicated that the present ACI
were obtained in test regions adjacent to a "clamped" Building Code procedures were adequate in estimation
support with loads applied to the top and bottom of the of strength of inverted T -beams subject to flexure,
flange at a cantilevered portion near the end of each flexural shear, or torsion. The tests also showed that
specimen. There seems to be a remarkable similarity the prestressed concrete members exhibited under ser-
among the torsional failure loads, with the maximum vice loads less cracking and lower stresses in transverse
value of 34.2 kips (152 kN) only 20 percent higher than reinforcement than did the nonprestressed members
the minimum value of 28.5 kips (127 kN) at failure. with identical proportions.
This seems to confirm the belief that for inverted T-
beam bentcaps there may be a limit to torsion strength
that is independent of the amount of transverse rein- ACKNOWLEDGMENT
This study was sponsored by the U.S. Bureau of Public Roads and
forcement and independent also of the amount and the Texas Department of Highways and Public Transportation under
form of longitudinal reinforcement. An analysis of the a cooperative research program. The writers wish to acknowledge the
test results showed that the equations of the ACI financial assistance provided by these organizations.
Building Code2 can be conservatively used to calculate
the pure torsional capacities of inverted T-beams.
REFERENCES
SUMMARY I. Standard Specifications for Highway Bridges, 12th Edition,
American Association of State Highway and Transportation Offi-
Bentcap girders in the shape of an inverted T have cials, Washington, D. C., 1977, 496 pp.
been used with considerable success, particularly in 2. ACI Committee 318, "Building Code Requirements for Rein-
bridge structures at grade separations. As traffic ap- forced Concrete (ACI 318-77)," American Concrete Institute, De-
proaches such bentcaps, stringer reactions on one troit, 1977, 102 pp.
flange cause twisting of the bentcap toward the ap- 3. Mirza, S. A., "Concrete Inverted T-Beams in Combined Tor-
sion, Shear, and Flexure," PhD dissertation, University of Texas at
proaching load. After traffic has crossed the bentcap, Austin, May 1974, 181 pp.
it causes twisting in the opposite direction. The struc- 4. Furlong, R. W., and Mirza. S. A., "Strength and Serviceability
tural behavior of the inverted T shape differs from that of Inverted T-Beam Bent Caps Subject to Combined Flexure, Shear,
of the more conventional top-loaded standard T-sec- and Torsion," Research Report No. 153-IF, Center for Highway Re-
tion in that the loads introduced into the bottom rather search, University of Texas at Austin, Aug. 1974, 78 pp.
5. Furlong, R. W.; Ferguson, P.M.; and Ma, J. S., "Shear and
than into sides or the top of the web must be supported Anchorage Study of Reinforcement in Inverted T-Beam Bent Cap
by stirrups acting as hangers to transfer vertical forces Girders," Research Report No. I 13-4, Center for Highway Research,
into the body of the web. In addition, flange reinforce- University of Texas at Austin, July 1971, 73 pp.

304 ACI JOURNAL I July-August 1983