Basile G. RabbatManager

Structural CodesCodes and Standards Department

Portland Cement AssociationSkokie, Illinois

Koz Sowlat*AssociateThe Datum/Moore Partnership

Consulting EngineersDallas, Texas

Testing of SegmentalConcrete Girders With

External Tendons

C onstriction of Interstate 526 overthe Cooper and Wando Rivers,

near Charleston, South Carolina in-cludes 8 miles (13 km) of bridges andviaducts. In the preliminary designstage of this project some innovativeconcepts for segmental concrete con-struction were proposed. Because ofthe large investment involved, itwas imperative to evaluate the struc-tural performance and economy ofthe proposed systems.

The owner of the bridge, South Caro-lina Department of Highways and Pub-lic Transportation, in cooperation withthe Federal Highway Administration,recommended that tests be conducted tocompare the performance of two in-novative segmental construction sys-tems with a conventional system, under

*Formerly, Structural Engineer, StructuralExperimental Section, Construction TechnologyL&iboratories, Skokie, Illinois.

short term static loading. This was ac-contplished through fabrication andtesting of three match cast segmentalconcrete girders.

SIGNIFICANCE OFRESEARCH

The conventional system for post-ten-sioning segmental concrete bridges usesinternal tendons. In this system, ductsthat house the post-tensioning tendonsare located within the main concretecross section. At all joints, the ducts ofadjacent segments have to he carefullypositioned to meet precisely. Securingthese ducts in place inside the forms re-quires extra care and time. In addition,the ducts interfere with casting and con-solidation of concrete. Special precau-tions are required to prevent concreteintrusion into the ducts at both ends ofeach segment.

86

On the other hand, the external ten-don system facilitates construction andmaintenance. Advantages of the externalprestressing system have been dis-cussed by Virlogeux. 2 They include thefollowing:

1. Manufacturing time, a major eco-nomic concern for all contractors, is re-duced by eliminating duct laying andadjustment operations.

2. A more streamlined tendon layoutcan be achieved, imparting a reductionin frictional losses and thereby improv-ing the efficiency of the prestressingsystem.

3. Placing and consolidation of con-crete are simplified.

4. Web thickness and therefore theweight of the superstructure may be re-duced.

5. Potential problem of concrete in-trusion in the ducts during casting iseliminated.

6. Grout injection is simplified. It iseasier to check the grouting process andto assess and repair inadequatelygrouted areas.

7. With proper design, replacementof broken or corroded tendons is possi-ble.

Construction speed and economy as-sociated with external tendon systemsmake them more attractive than internaltendon systems. In external prestressingsystems, although the ducts may begrouted, the tendons act as if unbonded.As a result, when strength consid-erations govern, a slightly larger tendonarea may be needed to obtain a nominalflexural strength equal to that of internalbonded systems.

SynopsisThree 30 ft long (9.14 m) match cast

segmental concrete girders wereloaded statically to destruction. Thetest variable was the location of thetendon ducts. In the first girder, theducts were embedded within thegirder cross section. The ducts of thesecond girder were external to theconcrete cross section except at piersegments and intermediate deviationblocks. The third girder was similar tothe second except that portions of theexternal ducts were embedded in asecond stage concrete cast. Thesegments included multiple shearkeys and were dry jointed. All ductswere grouted.

The first and third girders attainedthe flexural strength predicted by theclassic bending theory for bondedtendons. The flexural strength of thesecond girder (external tendons) waslower than that of the first and thirdgirders but considerably higher thanpredicted by the AASHTO equationfor flexural strength of members withunbonded tendons.

Until more refined analytical proce-dures become available, it is recom-mended that the tendons of segmen-tal members with external groutedpost-tensioning systems be designedusing the procedures for unbondedtendons given in the 1983 AASHTOSpecifications for Highway Bridges.

OBJECTIVESThe objectives of this experimental

investigation were to confirm theoreti-cal analyses and to compare the be-havior of partially bonded external ten-don systems with a completely bondedinternal tendon system. The effects of

potential damage to anchorages at girderends due to a severe earthquake were tobe evaluated through releasing the an-chor wedges of selected tendons prior toloading the girders to destruction. Lossof anchorages in the event of a severeearthquake relates to structural redun-dancy.

PCI JOURNAL/March-April 1987 87

mOD

SegmentNumber le ^2ri 8 11 I8 e 10 EfSegment Type P I II I II I II I II T P

A- B A +-)

2 41

-A

t- I B-

-0'_ Nine Segments 01 3'- 0 2 -DI

— —__

4-0" 4 -0

lJ I __'

2'-4 2-4

i4

Metric Equivalents:

Section A-A I ft =0305m Section B-BTypes hand P I in.= 25.4 mm Typel

Fig. 1. Principal dimensions and details of girders tested.

TEST SPECIMENS

Three segmental girders were fabri-cated and tested. The first girder, calledthe Bonded Tendon Girder, was post-tensioned with internal tendons. Thesecond girder, called the UnbondedTendon Girder, was post-tensioned withexternal tendons. The third girder,called the Modified Unbonded TendonGirder, was similar to the second girder,except that a second stage concrete castwas placed on top of the bottom flange,covering portions of the external ducts,for bond development.

The tested girders represented an ap-proximate one-fifth scale model of thebridge prototype. The bridge prototypeconsists of a single cell trapezoidal box.However, for simplicity and economy, adecked bulb T section was selected forthe test specimens. Shear keys betweenadjacent segments were not fully scaleddown in the test specimens.

Test VariableThe controlled test variable was the

location of the strand ducts in the testgirders. This variation can be seen insome of the illustrations.

GeometryEach girder included eleven match

cast segments of three types as shown inFig. 1. Type I segment was a prismaticsection. Type II segment was similar toType I but included a full depth dia-phragm in the middle of the segment.Strands were deflected at these dia-phragms.

The diaphragms incorporated saddlesas used in deviation blocks of segmentalbox girders. The pier segment, Type P,was a solid T section. Size and location ofshear and alignment keys are shown inFig. 2. A midspan cross section for eachgirder is given in Fig. 3. Tendon profilesof the three girders are plotted in Fig. 4.

Material PropertiesThe concrete mix was designed for a

target one-day compressive strength of2500 psi (17.2 MPa), for stripping offorms, and a target 14-day compressivestrength of 5500 psi (37.9 MPa), forpost-tensioning. Maximum aggregatesize was % in. (9.5 mm) pea gravel. Themix used for grouting the tendon ductsconsisted of Type I portland cement,water, and a water-reducing admixture.

At girder testing time, concrete com-pressive strength of each segment wasdetermined from 6x12 in. (152x305 m)standard cylinders. Average concretecompressive strength ranged between5810 and 7660 psi (40.0 and 52.8 MPa).Average concrete compressive strengthof the secondary cast of the ModifiedUnbonded Tendon Girder was 5250 psi(36.2 MPa) at test time.

Low-relaxation seven-wire strandsconforming to ASTM Designation: A416, Grade 270 (1862 MPa) were used.Two 0.5 in. (12.7 mm) and four 0.6 in.(15.2 mm) diameter strands were used ineach girder as shown in Fig. 3. Mea-sured tensile strength of 0.5 and 0.6 in.(12.7 and 15.2 mm) diameter strandswere 276 and 278 ksi (1903 and 1917MPa), respectively.

Reinforcing steel conformed to ASTMDesignation: A 615, Grade 60 (414MPa). Shear reinforcement in each seg-ment consisted of four vertical No. 4(12.7 mm) reinforcing bars at each faceof the web.

Specimen Fabrication.Techniques used in fabrication of the

test girders simulated actual construc-tion procedures for full scale match castsegmental concrete bridge girders. Ashort line match casting system wasused. Appropriate steps were taken toensure geometry control. Casting startedfrom both ends with pier segments.They were cast against a bulkhead. Sub-sequent segments were match cast

PCI JOURNAL March-Apnl 1987 89

Metric Equivalent*:Ift =0.305 mLin. = 25.4 mm

yp.)

a

8 Thick DeviationDiaphragm for TypefSegments Only —

12

„{TYRI

4' - 0'

__

I3„

i 72-4'

”

Fig. 2. Cross-sectional dimensions including shear and alignment keys.

vC-)

C-0C30z

n

v

CD

(a) Bonded Tendon Girder

Fig. 3. Midspan cross sections of girders. ® ® 4' LL2L2.

Legend:•0.6-in. Dia Strand0 O.5-in. Dia Strand

Second Stage Cast

41 4:' 2 I^ 44 42..

2Metric Equivalent:

(b) Unbonded Tendon Girder I in.= 254 mm(c) Modified Unbonded Tendon Girder

against an adjacent segment and a bulk-head. Finally, the middle segment wasmatch cast against two adjacent seg-ments.

After casting, the top surface of eachsegment was troweled and covered withpolyethelene sheets. On the next day,the polyethelene sheets were removed,forms were stripped, and the two matchcast segments were separated by pullingthem apart carefully. All segments werestored under laboratory conditions of73°F (23°C) and 50 percent relativehumidity for about 5 weeks, beforetesting.

In the Modified Unbonded Tendon

Girder, shear connectors were added tothe bottom flange to ensure compositeaction between the second stage castand the original concrete. G-shaped No.4 (12.7 rum) reinforcing bars served asshear connectors to the bottom flange oneither side of the web. They werewelded to the web reinforcement at oneend and to the bottom flange reinforce-ment at the other end. To accommodatewelding of shear connectors, prior tocasting, expanded polystyrene stripswere secured to the reinforcement.

After stripping the forms, the polysty-rene strips were removed. The detail ofshear connectors proposed for the actual

Applied Load

{a) Bonded Tendon Girder

Metric Equivalents'Ift=0.305mIin.254mrn Apphed Load

Two 0.6" 0 Strand

Support-Two 0.5 i Strand

-6°15'- 6'

(b) Unbonded and Modified Unbonded Tendon Girders

Fig. 4. Tendon profiles of Bonded Tendon Girder (top) and Unbonded and ModifiedUnbonded Tendon Girders (bottom).

92

bridge girders differed from the oneused in the laboratory and was moresuited for large scale production. Afterpost-tensioning the Modified UnbondedTendon Girder, a 2.5 in. (63.5 mm) thicksecond stage concrete cast was placedon top of the bottom flange to covershear connectors and tendon ducts.

Prior to post-tensioning each girder,all segments were pulled togethertightly, on a flat and level surface, usingtemporary connectors. At the joints ofthe Bonded Tendon Girder, special pre-cautions were taken to prevent groutleakage. At the interface of adjacentsegments, a thin bead of caulking com-pound was applied around each tendonduct, a distance of 1 in. (25.4 mm) awayfrom the duct edge.

Each test girder included two 0.5 in.(12.7 mm) and four 0.6 in. (15.2 mm) di-ameter seven-wire low-relaxationstrands. Each strand was housed insidea 1 in, (25.4 mm) inside diameter duct.Jacking force for 0.5 and 0.6 in. (12.7 and15.2 mm) diameter strands was 28.9 and

41.0 kips (129 and 182 kN), respectively.The eccentricity of prestress was similarfor all three girders.

Strain gauges installed on the tendonswere monitored during prestressing.Average midspan strand stresses, basedon measured strand strains, corre-sponded to 161 and 140 ksi (1110 and965 MPa) for 0.5 and 0,6 in. (12.7 and15.2 mm) diameter strands, respectively.The difference in effective prestressmay be attributed to the difference inseating losses and the sequence ofpost-tensioning.

In each girder, tendon ducts weregrouted after tensioning all six strands.The joints between segments were dryand free of grout or adhesives.

ii*11 J:I.Iei;1_1►yi

Test SetupA photograph of the test setup is

shown in Fig. 5. Each girder was simplysupported over a 30 ft (9.14 m) span.

^.,.Z z i TLoad Cell-^ Loading Beama41 \

'Wf 1

Fig. 5. Test setup of experimental girder with external tendons.

PCI JOURNALIMarch-April 1987 93

AppliedLoad

1.1I._ SS rSS

2 -a•

Support—

6 r- FF-- 6 -0'

I5-6'

Legend

r S3 Section for Strand Strain Measurement r - Joint Opening

X Concrete Surface StrainMetric Equivalents:

I ft = 0.305 mDeflection I in. = 25.4 mm

Fig. 6. Instrumentation plan for experimental girder.

8Cycle 2 Bonded Tendon

7

6 Cycle I '^^ I ^- Modified UUnbondenbonded Tendon

dApplied 5 /^/^..-^

, / /Moment,i00Q kip —in. 4 7 ^^ ! i

3/ /

2 Metric Equivalents:

/

000 i

hip-in = 113 kN.mfin. = 25.4mm

0 I0 I 2 3 4 5 6 7 8 9 10

Deflection , in.

Fig. 7. Midspan applied moment versus deflection relationships.

94

Loads were applied along two sectionslocated 6 ft (1.83 m) apart and equidis-tant from midspan.

Instrumentation

Measurements recorded during eachtest included applied loads, deflections,joint openings, concrete surface strains,and strand strains. The instrumentationplan was identical for all three girders,and is schematically shown in Fig. 6.

Data Acquisition

All measurements were scannedthrough an electronic data acquisitionsystem and transmitted to a mi-crocomputer. During the test, all mea-surements were reduced to engineeringunits and displayed continuously on aCRT screen. This allowed monitoring ofthe test progress. At selected load in-crements, reduced data were printed ona hard copy and stored on a computerdisk. A continuous plot of midspanapplied moment versus deflection wasalso recorded on an X-Y plotter.

Test Procedure

Each girder was subjected to twoloading cycles. In the first cycle, eachgirder was loaded in small incrementsuntil an inelastic behavior was observedand a midspan deflection of approxi-mately 3 in. (76 mm) was reached. Thegirder was then completely unloaded.Subsequently, anchor wedges for thetop two strands were burned at bothends to simulate anchorage loss due to asevere earthquake. In the second load-ing cycle, each girder was loaded in in-crements up to destruction. Each testwas concluded when a significant dropin load carrying capacity of the girderwas observed.

Simulation of anchorage loss was alast minute addition to the test program.It was an attempt to "hit two birds withone stone." Production of segments was

in progress when the decision was madeto bum the anchor wedges of the top twostrands. Had the decision been madeearlier to release the anchor wedges,pier segments would have been de-signed differently. Further, the testspecimens had monostrand anchors.The actual bridge includes "trumpet"anchors for multiple strands. The coni-cal shape of a trumpet anchor providesbetter confinement and slip resistancefor the strands than a straight anchor,once the grout hardens. Enhanced be-havior would have been anticipated hadthe pier segments been longer and/ortrumpet anchors been used,

BEHAVIOR OF SPECIMENSRelationships of midspan applied

moment versus deflection for the threetested girders are illustrated in Fig. 7.Peak applied moments and corre-sponding deflections for first and secondloading cycles are summarized in Table1.

In the initial portion of the first load-ing cycle, the three girders exhibitedsimilar stiffnesses. As loading pro-gressed, the Bonded Tendon and theModified Unbonded Tendon Girdersbehaved similarly while the UnbondedTendon Girder exhibited a larger de-flection for the same applied moment.Loads resisted by the test girders duringthe first loading cycle were compared ata midspan deflection of 2.65 in. (67 mm).This was the maximum deflection of theUnbonded Tendon Girder during thefirst loading cycle. The Bonded Tendonand the Modified Unbonded TendonGirders required 22 and 28 percentmore load than the Unbonded TendonGirder, respectively, to reach thismidspan deflection.

After the first loading cycle, eachgirder was completely unloaded and theanchor wedges for the top two strandswere burned. In the Bonded TendonGirder, because the ducts were em-

PCI JOURNALJMarch-April 1987 95

Table 1. Summary of test data.'

Dead load

Peak moment duringfirst loading cycle

Peak moment duringsecond loading cycle

Applied Residual AppliedTest moment moment Deflection deflectiont moment Deflection

girder (kip-in.) (kip-in.) (in.)_-- (in.) (kip-in.) (in,)

BondedTendon Girder 690 6220 2.95 0.13 6910 7.68

UnbondedTendon Girder 69U 4980 2,65 0.05 4670 4.31

ModifiedUnhonded 720 6530 3.26 0,18 5080 4,34

Tendon Girder

* All tabulated moments and deflections correspond to midspan. Applied moments exclude dead loadmoment due to weight of girder and loading hardware,

t After first loading cycle.Metric Equivalents: 1000 kip-in. = 113 kN-m; 1 in. = 25.4 mm_

bedded in the concrete cross section andwere grouted, the tendons were bondedand transfer of prestress occurred in theend regions. Although the anchorageswere burned, the top two strands re-mained bonded. For the two externaltendon girders, measurement of strandstrains indicated a loss of bond for thetop two strands at the end regions, im-mediately after burning of the anchor-ages. As a result, the effectiveness ofabout one-third of the tendons was lost.Applied load versus strand strain re-lationship for a top strand of the Un-bonded Tendon Girder is illustrated inFig. 8. This strand strain was measuredat a section 10.5 ft (3.2 m) away frommidspan.

Fig. 8 indicates a slight increase in thestrain of the top two strands during thesecond loading cycle. The ducts of thesestrands were embedded in the piersegments and deviation diaphragms.The friction between the strands and thegrout was probably adequate to stabilizethat small tensile force in the strands.However, the remaining force in thesestrands was negligible. In the secondloading cycle, loss of bond of the top two

strands lead to considerable reductionin the load carrying capacity of the ex-ternal tendon girders.

All three girders were assembled dryjointed. During the tests, opening of thejoints between adjacent segments wasmonitored at the level of the bottom fi-bers. Opening of joints during the firstand second loading cycles is plotted inFigs. 9 and 10, respectively. In Fig. 9,maximum opening of joints in the Mod-ified Unbonded Tendon Girder waslarger than that of the Unbonded Ten-don Girder because each girder wasIoaded up to a different deflection dur-ing the first loading cycle.

With all three specimens, Joints 1-2and 10-11, identified in Fig. 1, remainedclosed throughout testing. Similarly,Joints 2-3 and 9-10 of the UnbondedTendon and Modified Unbonded Ten-don Girders remained closed through-out testing. In the Bonded Tendon Gir-der, maximum opening of Joint 2-3 was0.028 and 0.066 in. (0.7 and 1.7 mm) atthe peak moments of the first and sec-ond loading cycles, respectively. AtJoint 9-10, the corresponding openingswere 0.018 and 0.049 in. (0.5 and 1.2

96

Applied Moment, kip-in.

—Second Loading Cycle

3000

i,ss of Prestress Due tcBurning of Anchorage

-4000 -3000 -2000 -1000

2000

e

--First Loading Cycle

Metric Equivalent:1000 kip-In- = 113 kN. m

1000 2000

Strand Strain, millionths

Fig, 8, Flepresentative strand strain for a top strand in the external tendon girder.

5000

4000

mm), respectively.The following sections include a de-

scription of the behavior of each testgirder. Segment numbers are identifiedin Fig. 1. Each joint is identified by thenumbers of adjacent segments.

Bonded Tendon GirderFirst surface cracks appeared in the

center portion of Segment 6 at anapplied moment of 3420 kip-in. (386kN•m). These were flexural cracks thatextended up 13 in. (330 mm) from thebottom surface of the girder. As theapplied load was increased, flexuralcracks grew in size and number. In ad-dition, diagonal web cracks appeared inthe shear spans.

Throughout the first loading cycle,joint openings were largest at Joints 5-6and 6-7. The bottom shear key on Seg-ment 7 at Joint 6-7 broke off at an

applied moment of 6070 kip-in. (686kN•m). Upon further loading, amaximum applied moment of 6220kip-in. (703 kN•m) was attained. At thispoint, an extensive and well distributedpattern of diagonal web cracks hadformed in the shear spans, outside theconstant moment region.

The specimen was completely un-loaded. Then, anchor wedges for the toptwo strands were burned at both ends.Strand strains did not change duringwedge burning.

Next, the second loading cycle wasstarted. At an applied moment of 5800kip-in. (655 kN •m), new diagonal webcracks were observed. Throughout thesecond loading cycle, joint openingswere largest at Joints 5-6 and 6-7. Themaximum applied moment leveled off at6910 kip-in. (781 kN •m), Upon furtherloading, the load carrying capacity of thegirder decreased as a flexural failure

PCI JOURNALMarch -April 1987 97

e

6

4

2

r 0i¢B

m s

.4C

^ 2

uQ Opaning of Joints, In.a B

6

4

a

0

Fig. 9. Joint opening during first loading cycle.

HondadTendon

UnhandedTendon

Mad If!edUnbanded

Tendon

T 1 ^

Applied Load I .1. Applied Load

G

o,

cc,m

r.

Metric EquivalentI in. = 25.4 mm

7

C)

OC

zI-I-

:7

03mV

Metric Equivalent:I in. = 25.4 mm

B

B

c80 .5

— B

mmm

r unbondedca TendonEar

m^e 5 ..^

Opening of Joints, In.Q g ^ r F r I-

Modified

2 V J'-' UTen don d

e B

Fig. 10. Joint opening during second loading cycle.cOcfl

Fig. 11. Joint 5-6 of Bonded Tendon Girder after testing.

mechanism formed at Joint 5-6. Failurestarted with localized crushing of con-crete along the top surface of the jointresulting in a load decrease. Then, sev-eral strands broke leading to a total lossof load carrying capacity. Fig, 11 is aphotograph of the failure zone at Joint5-6.

Unbonded Tendon GirderFirst surface cracks appeared at an

applied moment of 3170 kip-in. (358kN•m). These were diagonal web cracksmainly concentrated at Joints 4-5 and7-8. As the applied load was increased,diagonal web cracks grew in size andnumber. Throughout the first loadingcycle, joint openings were largest atJoints 4-5 and 7-8. The maximumapplied moment was 4980 kip-in. (563kN•m). At this point, an extensive pat-tern of diagonal cracks had formed atJoints 4-5 and 7-8. At each of these twojoints, cracks mainly formed on the sideof the joint closer to the loading points.

Few cracks occurred on the side of thejoint closer to the support.

The specimen was completely un-loaded. Then, anchor wedges for the toptwo strands were burned at both ends.Strand strains indicated a large drop inthe stress of the two top strands in thedraped portions of the strands. Thisstress drop was a result of releasing theanchors of the two top strands. Transferlength in the pier segments was inade-quate for transfer of prestress. Atmidspan, no noticeable drop in straindue to burning of wedges was observed.

During the second loading cycle, newdiagonal web cracks were observed atan applied moment of 2970 kip-in. (336kN•m). Throughout the second loadingcycle, joint openings were largest atJoints 4-5 and 7-8. At an applied momentof 4550 kip-in. (514 kN•m), two shearkeys at Joint 4-5 broke off, causing aminor load drop.

Subsequently, load was increasedgradually. The maximum applied mo-ment attained was 4670 kip-in. (528

,II]

Fig. 12. Joint 4-5 of the Unbonded Tendon Girder after testing.

kN•m). Upon further loading, the loadcarrying capacity of the girder decreasedas more individual shear keys broke off.Finally, a shear compression failure oc-curred in the top flange at Joint 4-5. Thetest was concluded. Fig. 12 is a photo-graph of the failure zone at Joint 4-5.

Modified Unbonded Tendon GirderThe first observed cracks were

flexural and occurred in the secondstage concrete cast. As the applied loadwas increased, flexural cracks in thesecond stage cast grew in size andnumber. In addition, diagonal webcracks appeared in the shear spans. Jointopenings were largest at Joints 5-6 and6-7 up to an applied moment of 565()kip-in. (638 kN•m). Then, up to anapplied moment of 6170 kip-in. (697kN•m), joint openings were largest atJoints 4-5 and 6-7. Subsequently, jointopenings were largest at Joints 4-5 and7-8.

The maximum applied moment was

6530 kip-in. (738 kN•m). At this loadlevel, an extensive pattern of diagonalcracks had formed in the shear spans. Inone shear span, cracks were mainly con-centrated in Segment 5 at Joint 4-5.Fewer cracks occurred in Segment 4. Inthe other shear span, cracks were welldistributed and generally extended onboth sides of Joint 7-8.

The specimen was completely un-loaded. Then, anchor wedges for the toptwo strands were burned at both ends.Strand strains indicated a large drop inthe stress of the top two strands. Thisstress drop was a result of releasing theanchors of the two top strands. Transferlength in the pier segments was inade-quate for transfer of prestress. Atmidspan, no noticeable drop in straindue to burning of wedges was observed.

During the second loading cycle, at anapplied moment of 3490 kip-in. (394kN •m), new diagonal web cracks wereobserved. Throughout the second load-ing cycle, joint openings were largest atJoints 4-5 and 7-8. 'The maximum

PCI JOURNAUMarch-April 1987 101

Fig. 13, Joint 7-8 of the Modified Unbonded Tendon Girder after testing.

applied moment of 5080 kip-in. (574kN •m) was 9 percent higher than thatattained by the Unbonded TendonGirder.

Following the peak load, two shearkeys at Joint 4-5 broke off, causing aminor load drop. Upon further loading,more keys broke. Finall y, shear keysbroke off at Joint 7-8 and a shear com-pression failure occurred in the topflange at Joint 7-8 concluding the test.Fig. 13 is a photograph of the failurezone at Joint 7-8.

FLEXURAL STRENGTHANALYSIS

Flexural strength analyses were per-formed for midspan sections of the testgirders. Two analysis methods wereused. In the first method, the classicbending theory of plane sections wasused. The flexural strength was calcu-lated through strain compatibility andequilibrium of internal forces, assumingtendons were bonded. In the second

method, flexural strength was calculatedfrom the equations given in the 1983AASHTO Specifications for bonded andunbonded tendons.

All analyses were based on measuredmaterial properties and effective pre-stress. The effective prestress was de-termined from measured strand strains.Since the segments were assembledwith dry joints, concrete tensile strengthwas ignored. As the Unbonded Tendonand the Modified Unbonded TendonGirders had identical strand layout, bothanalysis methods would predict similarflexural strength for the midspan sec-tions. Results of the flexural strengthanalyses for the Bonded Tendon andUnbonded Tendon Girders are listed inTable 2.

Four cases were analyzed. Case 1 cor-responds to the midspan section of theBonded Tendon Girder with six strands.Cases 2 and 3 correspond to the midspansection of the Unbonded Tendon Gir-ders with six strands, The differencebetween Cases 2 and 3 is the concretecompressive strength. This difference

102

Table 2. Calculated flexural strength.

Calculated momentConcrete (kip-in.}

Number of Girder compressiveCase effective cross strength Computer

number strands section (psi) analysis AASHTO

BondedI Tendon 5810 7260 7320

All Girdersix

Unbonded 5810 7210` 43802 strandsTendon

6400 7220* 43903 Girders

Four Unbonded4 bottom Tendon 6400 4660* 2850

strands Girdersonly

* Strands assumed bonded to concrete for analysis purposes.Metric Equivalents: 1000 kip-in. = 113 kN•m; 1000 psi = 6.895 MPa.

allowed evaluation of the effect of minorvariation in the concrete compressivestrength on flexural strength. Concretecompressive strengths for Cases 2 and 3corresponded to values measured oncompanion cylinders from the failurezone concrete of the Bonded TendonGirder and Unbonded Tendon Girder,respectively. Case 4 corresponds to theUnbonded Tendon Girders with fourbottom strands only.

After the first loading cycle, the an-chorages for the top two strands wereburned. Burning of the anchorages didnot significantly affect the behavior ofthe Bonded Tendon Girder. For theUnbonded Tendon Girders, however,the contribution of the top two strandswas drastically reduced after burningthe anchorages. Therefore, Case 1 corre-sponds to the first and second loadingcycles of the Bonded Tendon Girder.Case 3 corresponds to the first loadingcycle of the Unbranded Tendon Girders,while Case 4 provides a lower boundflexural strength for the second loadingcycle of the Unbonded Tendon andModified Unbonded Tendon Girders.

Computer AnalysisA computer program was used to pre-

dict the flexural strength of the test gird-ers. For the Unbonded Tendon Girders,the strands were external to the concretecross section at midspan. However, foranalysis purposes, the strands were as-sumed fully bonded to the concrete.

Cases 1 and 2 were compared toevaluate the effect of the varying tendonconfigurations, assuming fully bondedstrands. The analysis shows that tendonlocations used in Bonded and Un-bonded Tendon Girders yielded similarflexural strengths. Therefore, the pre-stress eccentricity was the same for allthree girders.

Cases 2 and 3 were compared toevaluate the effect of concrete compres-sive strength on flexural strength. Theanalysis showed that flexural strengthwas not sensitive to small variations inconcrete compressive strength.

For the Bonded Tendon Girder, themaximum total moment attained duringthe second loading cycle, 7600 kip-in.(859 kN•m), exceeded the calculatedstrength inCase 1 by 5 percent. The

PCI JOURNAL/March-April 1987 103

Unbonded Tendon and Modified Un-bonded Tendon Girders were notloaded to destruction during the firstloading cycle while all strands werefully effective. Nevertheless, themaximum total moment attained by theModified Unbonded Tendon Girderreached the computed strength forbonded tendons.

For the Unhonded and Modified Un-bonded Tendon Girders, the maximumtotal moment attained during the secondloading cycle, 5360 and 5800 kip-in.(606 and 655 kN•m), exceeded the cal-culated strength in Case 4 by 15 and 24percent, respectively. This excesscapacity indicates that loss of bond atgirder ends for the top two strands didnot totally eliminate their contributionto the flexural strength at midspan.

Analysis by AASHTOSpecifications

The AASHTO Specifications' allow ahigher strength for bonded tendonscompared to unbonded tendons. Theaverage strand strength at ultimate isspecified as follows:

For bonded tendons:

^`* (_o.si)J=fa

For unbonded tendons:

f,*, =f„.+ 15,0(X)

wheref= average stress in prestressing

steel at ultimate load, psif.9 = ultimate strength of prestressing

steel, psip* = ARlbd, ratio of prestressing steelA*^= area of prestressing steel, sq in.b = width of flange of flanged mem-

ber or width of rectangularmember, in.

d = distance from extreme compres-sive fiber to centroid of pre-stressing force, in,

f : = compressive strength of con-

crete at 28 days, psi= effective steel prestress after

losses, psiTherefore, the AASHTO Specifica-

tions recognize that members with un-bonded tendons have a lower strengththan identical members with bondedtendons.

For the Bonded Tendon Girder, themaximum total moment attained duringthe second loading cycle, 7600 kip-in.(859 kN•m), exceeded the strength pre-dicted by the AASHTO equation (Case1) by 4 percent. For the Unbonded Ten-don Girder, the maximum total momentattained in the first loading cycle, 5670kip-in. (641 kN-m), exceeded the pre-dicted strength for unbonded tendons(Case 3) by 29 percent.

It is concluded that the AASHTOprovisions for unbonded tendonssignificantly underestimated theflexural strength of the Unbonded Ten-don Girder. Despite the loss of bond ofthe top two strands, the maximum totalmoment attained during the secondloading cycle, 5360 kip-in. (606 kN•m),exceeded the strength predicted byAASHTO for six unbonded tendons.

The Modified Unbonded Tendonpost-tensioning system is not addressedby the AASHTO Specifications. The"Modified Unbonded Tendon" systemtested was effectively a combination ofbonded tendons in the center portion ofthe girder where all ducts were embed-ded in concrete and external tendonswhere the ducts were outside the girdercross section. For the tested condition,the Modified Unbonded Tendon Girderreached during the first loading cyclethe strength predicted by AASHTO forbonded tendons.

Analysis by ACI CodeThe ACT 318-83 Code' equation for

unbonded tendons yields a flexuralstrength higher than that predicted bythe AASHTO Specifications. For theUnbonded Tendon Girders with six ef-

104

fective strands, the flexural strengthpredicted by ACI 318-83 was 5600kip-in. (633 kN•m). This strength wasreached by the Unbonded Tendon Gir-der during the first loading cycle, wherethe maximum moment attained was5670 kip-in. (641 kN •m). It is concludedthat the ACI Code equation for un-bonded tendons can also be used to pre-dict the flexural strength of girders withexternal tendons.

CONCLUSIONS ANDOBSERVATIONS

Based on the test results, the follow-ing conclusions were drawn and obser-vations noted:

1. Although the segments were dryjointed, the Bonded Tendon Girder at-tained the strength predicted by theclassic bending theory for bonded ten-dons.

2. During the first loading cycle, theModified Unbonded Tendon Girderreached the strength predicted by theclassic bending theory for bonded ten-dons.

3. Despite the loss of bond of the twotop strands, the Unbonded Tendon andthe Modified Unbondcd Tendon Gir-ders exceeded the flexural strength pre-dicted by the AASHTO Specificationsfor members with unbonded tendons.

4. During the first loading cycle, theUnbonded Tendon Girder reached theflexural strength predicted by ACI 318for unbonded tendons,

S. The Unbonded Tendon Girder wasthe easiest to construct. The ModifiedUnbonded Tendon Girder ranked se-cond in ease of construction.

6. During the first loading cycle, theBonded Tendon and the Modified Un-bonded Tendon Girders behaved simi-larly. The Bonded Tendon Girder ex-hibited a slightly larger deflection atcomparable applied moments.

7. In the Bonded Tendon Girder, re-leasing of the anchor wedges of the twotop strands did not significantly affect

the behavior of the girder. Adequate de-velopment length for the two top strandswas available at the ends of the BondedTendon Girder.

8. Releasing the anchor wedges of thetwo top strands in the Unbonded Tendonand the Modified Unbonded Tendongirders affected the behavior of thesegirders. Strains measured on thesestrands indicated Ioss of prestress andtherefore loss of bond in the pier seg-ments and draped portions of the Un-bonded Tendon and the Modified Un-bonded Tendon Girders.

9. During the loading cycle to de-struction, the Modified Unbonded Ten-don Girder exhibited 8 percent higherstrength compared to the UnbondedTendon Girder. The Bonded TendonGirder, with virtually 59 percent moreprestressing, exhibited 42 percenthigher strength compared to the Un-bonded Tendon Girder.

10. The Bonded Tendon Girder failedin a flexural mode where simul-taneously concrete crushed in the com-pression zone and strands fractured inthe tensile zone.

11. In the Unbonded Tendon and theModified Unbonded Tendon Girders,where there was loss of prestress in thetwo top strands, joint openings werelarger in the shear span. As a result, theshear keys broke progressively leadingto loss of shear carrying capacity. Ulti-mately, shear compression failure oc-curred in the top flange at a joint whereshear keys had broken.

DESIGNRECOMMENDATIONS

Based on the test results reported inthe paper and the conclusions notedabove, the following is recommended:

1. The tendons of segmental memberswith external grouted post-tensioningsystem can be conservatively designedby the 1983 AASHTO Specifications orACI 318-83 Code equations for mem-

PCI JOURNAL'March-April 1987 105

hers with unbonded tendons.2. The tendons of segmental members

with internal grouted post-tensioningsystems can be reliably designed by theclassic bending theory for bonded ten-don s.

CLOSING REMARKSBased on the test results, the external

tendon system with the secondary con-crete cast was incorporated in the finaldesign. An addendum to the final designallowed contractor options for the pre-stressing system. The lowest bid was fora concrete option that included a com-bination of external (unhonded) and in-ternal (bonded) tendons. This systemproved to be the most economical solu-tion for 1.5 miles (2.4 km) of bridges forPhase I of this three-phase project. Thelowest bid for the steel alternate was 37percent higher than that of the selectedconcrete alternate.

Further research on analytical pro-cedures verified by these and other ex-periments on segmental concrete gir-ders with external tendons is currentlyunderway in France under the directionof Professor Michel Virlogeux and at theUniversity of Texas at Austin under thedirection of Professor John E. Breen.When completed, such analytical toolsshould yield more accurate and eco-nomical designs. Until such tools be-come available, however, external ten-dons can be conservatively designed by

the AASHTO equation for flexuralstrength of members with unbondedtendons.

ACKNOWLEDGMENTSThe test specimens were manu-

factured and tested in the StructuralLaboratory of Construction TechnologyLaboratories under the direction of Mr.Donald M. Schultz, Manager, StructuralExperimental Section, and Dr. Henry G.Russell, Director, Structural Develop-ment Department.

This research program was performedto provide experimental evidence toback tip the design of a match cast seg-mental post-tensioned concrete bridgeon I-526 over Cooper and Wando Riv-ers, near Charleston, South Carolina.Funding was provided by the SouthCarolina Department of Highways andPublic Transportation, Columbia, SouthCarolina, Mr. John R. Coleman, P.E.,Bridge Engineer—Design. ConsultingEngineers were Figg and Muller En-gineers, Inc., Tallahassee, Florida. Rep-resenting Figg and Muller Engineers,Inc. were Mr. James M. Barker, S.E.,P.E., Assistant Vice-President/Projects,and Mr. David C. O'Hagan, P.E., SeniorBridge Engineer.

The authors are indebted to all re-viewers of this paper and in particularMr. James M. Barker and Dr. WalterPodolny. All comments helped enhancethe value of the paper.

106

REFERENCES

1. Podolny, W., and Muller, J. M., Construc-tion and Design of Prestressed ConcreteSegmental B ridges, John Wiley & Sons,New York, N.Y., 1982,561 pp.

2. Virlogeux, M., "External Prestressing,"IABSE Proceedings, International Associ-ation for Bridge and Structural Engineer-ing, P-62J82, Zurich, Switzerland, 1982, pp_101-108.

3. Standard Specifications for HighwayBridges, American Association of StateHighway and Transportation Officials,Thirteenth Edition, Washington, D.G.,1983.

4. ACI Committee 318, `Building Code Re-quirements for Reinforced Concrete (ACI318-83)," American Concrete Institute,E etroit, Michigan, 1983, 111 pp.

NOTE: Discussion of this paper is invited. Please submityour comments to PCI Headquarters by December 1, 1987.

PCI JOURNAUMarch-April 1987 107