Preliminary Results and Conclusions From the PRESSS Five ...

James R. ConleyGraduate Research AssistantDepartment of Structural Engineering University of California at San DiegoLa Jolla, California

S. (Sri) Sritharan, Ph.D.Assistant Project Scientist

Department of Structural EngineeringUniversity of California at San Diego

La Jolla, California

A large-scale five-story precast concrete buildingconstructed to 60 percent scale was tested undersimulated seismic loading as the culmination ofthe 10-year PRESSS (Precast Seismic StructuralSystems) research program. The buildingcomprised four different ductile structural framesystems in one direction of response and a jointedstructural wall system in the orthogonal direction.The test structure was subjected to seismic inputlevels equivalent to at least 50 percent higher thanthose required for UBC (Uniform Building Code)Seismic Zone 4. The behavior of the structure wasextremely satisfactory, with only minimal damagein the shear wall direction, and no significantstrength loss in the frame direction, despite beingtaken to drift levels up to 4.5 percent, more than100 percent higher than the design drift level. Thetest validated the Displacement-Based Design(DBD) approach used to determine the requiredstrength and confirmed the low damage and lowresidual drift expected of the building.

The Precast Seismic Structural Systems (PRESSS) re-search program has been in progress for ten years,with the final phase of the program nearly complete.

PRESSS, sponsored by the National Science Foundation(NSF), Precast/Prestressed Concrete Institute (PCI) andPrecast/Prestressed Concrete Manufacturers Association ofCalifornia, Inc. (PCMAC), has coordinated the efforts ofover a dozen different research teams across the UnitedStates to improve the seismic performance of precast/pre-stressed concrete buildings.

Preliminary Results andConclusions From the PRESSSFive-Story Precast ConcreteTest Building

42 PCI JOURNAL

M. J. Nigel Priestley, Ph.D.Professor of Structural EngineeringDepartment of Structural EngineeringUniversity of California at San DiegoLa Jolla, California

SPECIAL REPORT

Stefano PampaninFulbright Visiting Scholar

Department of Structural EngineeringUniversity of California at San Diego

La Jolla, California

November-December 1999 43

In the context of this paper, “build-ings” refer to low- and high-rise build-ings such as office buildings, parkingstructures, hotels, hospitals, multi-family housing, and other specialstructures. However, bridges andtransportation structures are excluded.

Since the very beginning of thePRESSS program, all of the researchteams involved in the program havefocused their sights on two primaryobjectives:• To develop comprehensive and ra-

tional design recommendationsneeded for a broader acceptance ofprecast concrete construction in dif-ferent seismic zones.

• To develop new materials, concepts,and technologies for precast con-crete construction in different seis-mic zones.The first and second phases of the

PRESSS research program have beencompleted. The third phase of the pro-gram comprises the design, erectionand testing of a five-story precast con-crete building using dry jointed con-struction. The purpose of this paper isto present the preliminary results andconclusions from this test.

As the key element of the finalphase of the PRESSS research pro-gram, a 60 percent scale five-storyprecast concrete building was con-structed and tested under simulatedseismic loading at the University ofCalifornia, San Diego (UCSD), Cali-fornia. The tests were carried out be-tween June and September 1999.

The design of the building wasbased on concepts developed andtested in the earlier two phases of thePRESSS research program, whichhave been summarized by Priestley.1

Details of the building design havealso been described by Nakaki, Stan-ton and Sritharan,2 and hence only abrief summary of the major designfeatures of the structure is includedhere.

Fig. 1 shows a plan layout of thelower three floors of the building,which had a two-bay by two-bay con-figuration, with bay size 15 x 15 ft(4.57 x 4.57 m). Two different precastframes, one with prestressed beams,and the other using mild steel reinforc-ing bars across the beam-to-columnconnections provided lateral resistance

Fig. 1. Plan view of test building showing lower three floors.

at opposite sides of the building in onedirection of response, with a centralstructural wall providing lateral resis-tance in the other direction. Non-seis-mic gravity frames without momentconnection between beams andcolumns framed the test buildingboundaries parallel to the structuralwall.

As shown in Fig. 1, the floor systemfor the lower three floors consisted ofpretopped double tees spanning be-tween the seismic frames, and con-nected to the center of each of the twowall panels comprising the structuralwall. In the upper two floors, toppedhollow-core slabs spanning betweenthe gravity frames and the wall pro-vided the floor system. The hollow-core slabs were connected to eachother and the seismic frames with acast-in-place topping.

Although the plan dimensions of thetest building were comparatively mod-est, it should be recognized that thetributary floor area in the prototypebuilding, on which the test buildingwas based, was substantially larger

than what is shown in Fig.1. Verticalsupport to the extra area was providedby additional gravity frames which didnot contribute to the seismic resistanceof the building.

In order to obtain the maximumpossible information from the testbuilding, five different structural sys-tems were investigated in the samestructure. The four different ductileframe systems tested are illustrated inFig. 2.

The lower three floors of the pre-stressed concrete frame consisted ofhybrid connections (see Fig. 2a) withsingle-bay beams connected betweencontinuous columns by unbondedpost-tensioning. Additional flexualstrength and damping was providedby mild steel reinforcing bars slidthrough corrugated ducts in the beamsand columns, and grouted solid withshort unbonded lengths in the beamsto reduce peak strains under seismicresponse.

In the upper two floors of the pre-stressed concrete frame, continuousprecast rectangular beams with preten-

44 PCI JOURNAL

sioned strands bonded in external stubbeams, but debonded over the interiorbeam length between the column facesprovided seismic resistance (see Fig.2b). No additional mild steel rein-forcement was provided across thebeam-column interfaces. The fulllength pretensioned beams werethreaded over column reinforcing barsextending from the top of the columns,with reinforcing bar splices providingcolumn continuity.

Moment resistance over the lowerthree floors of the non-prestressedframe was provided by TCY-gap con-nections (TCY denotes tension/com-pression yielding), as illustrated inFig. 2c. At the bottom of the connec-

tions, the beams were clamped to thecolumn by unbonded post-tensioningthreaded bars reacting through fibergrout pads over the bottom part of theconnection only. Over the top two-thirds of the connection, a 1 in. (25mm) gap was provided.

Mild steel reinforcement slidthrough corrugated ducts and groutedsolid in the beams and column was de-signed to yield alternately in tensionand compression, for negative andpositive moments, respectively, with-out unloading the precompressionacross the bottom grout pad. As a con-sequence, all rotation at the beam-col-umn interface would occur by openingor closing of the gap at the top of the

connection. Over the top two floors ofthis frame, moment capacity was pro-vided by the TCY connection usingmild steel reinforcing bars in groutducts at the top and bottom of the con-nection, thus approximating the be-havior of a conventional reinforcedconcrete connection with equal topand bottom steel (see Fig. 2d).

In the orthogonal direction, thestructural wall consisted of four pre-cast panels, each two and a half storiestall (see Fig. 3). The panels were verti-cally connected to each other and tothe foundation by unbonded verticalpost-tensioning, using threaded bars.A horizontal connection across thevertical joint was provided by stain-

Fig. 2. Seismic frame connection types.

(a) Hybrid post-tensioned connection (c) TCY-gap connection

(b) Pretensioned connection (d) TCY connection


less-steel energy-dissipating U-shapedflexure plates, welded to embed platesin both adjacent wall panels. In addi-tion to providing energy dissipation,these plates provided additional lateralresistance by shear-coupling betweenthe two structural walls.

The structural design was carriedout in accordance with Direct Dis-placement-Based Design (DBD) prin-ciples which have been described byPriestley et al.1,3,4 The chosen designcriterion was such that the structureshould achieve a maximum lateraldrift of 2 percent under seismic excita-tion equivalent to UBC (UniformBuilding Code) Zone 4, for an inter-mediate Sc soil. More complete detailsof the design procedure and the designdetails are available in a recent paperby Nakaki et al.2

Erection of the building was carriedout inside the Charles Lee PowellStructural Laboratory of UCSD, oneof the largest structural testing facili-ties in the United States. Despite therestricted space and special con-straints, erection of the test buildingwas completed within 12 workingdays. Because of page limitations,construction details regarding the as-sembly of the test building will not bedescribed here, but will be the subjectof a separate paper, to be publishedlater as part of a series of final paperson the project.

TEST SEQUENCELateral forces and displacements

were applied to the building by twoactuators at each floor level, each con-nected to the floor by actuator exten-sions and a triangulated connectiondetail, shown in Fig. 4. With this de-tail, the actuator load was equally dis-tributed to the two bays, and relativehorizontal displacement between thetwo connection points on the floor waspossible.

One consequence of the connectiondetail was that since the actuator waslocated above the floor level, the con-nection detail imposed vertical forceson the floors, in addition to the hori-zontal forces. Equal displacementswere applied to the two actuators ateach floor level at all stages of the test.

The objective of the experiment wasto test the structure under a series of

Fig. 3. Jointed structural wall elevation.

earthquake levels. These seismic lev-els were chosen to conform to differ-ent limit states recommended by theStructural Engineers Association ofCalifornia (SEAOC).6 As shown inFig. 5a, four different earthquake lev-els are defined, corresponding to 33,50, 100 and 150 percent of the designlevel earthquake.

The principal method of testing thebuilding was pseudodynamic testing,using spectrum-compatible earthquakesegments, scaled in frequency and am-plitude from recorded accelerograms tomatch the design spectrum at 5 percentdamping. An example of the agreementbetween the design spectrum and theaccelerogram segment for the designlevel event (EQ3 in Fig. 5a) is shown inFig. 5b. More details on the test pro-gram are available in Refs. 7, 8 and 9.

The concept of pseudodynamic test-ing is described in Fig. 6. The com-puter controlling the displacementinput to the actuators contained a sim-ple five-degrees-of-freedom model ofthe building, initially calibrated bylow-level flexibility tests, as describedsubsequently. This model was sub-jected to inelastic time-history analy-sis under the specified earthquakerecord.

The structural response is governedby d’Alembert’s equation of motion:

(1)

where [M], [C], and [K] are the mass,damping and stiffness matrices, and

are the relative accelera-tion, velocity and displacement vec-tors, and ag is ground acceleration.

˙̇ , ˙x x xand

[ ]˙̇ [ ] ˙ [ ] [ ]M x C x K x M ag+ + = −

46 PCI JOURNAL

Eq. (1) was solved for sequentialtime steps and the calculated displace-ment vector at the end of each timestep was applied to the building by theactuators. The forces required to applythese displacements represent the non-linear stiffness terms [K]x in Eq. (1),and were used as the appropriaterestoring force for the next time incre-ment. Thus, as the stiffness of the ac-tual structure is modified by inelasticaction, or strength degradation, the an-alytical procedure recognizes the stiff-ness changes, and modifies the struc-tural response accordingly.

Note that since the computer is notaware of the actual building weight,the mass vector is a required input tothe program. This meant that the cor-

rect scaled tributary mass for each di-rection of loading, which was substan-tially higher than the building mass of500 tons (454 t), but different in thetwo orthogonal directions, could beused. It also means that the results ofthe test can be interpreted for zonesother than Zone 4 seismicity. For ex-ample, the response of a similar struc-ture constructed in Zone 2 could beobtained, recognizing that the tributarymass for the structural system wouldbe greater than for Zone 4.

Viscous damping could not be di-rectly modeled in the test, and hence itwas a required input to the computerprogram controlling the test. It wasnecessary, for design and testingequivalence, for the same assumptions

Fig. 4. Connection of actuators to floor.

Fig. 5. Earthquake excitation and spectrum matching.

(a) Earthquake levels (b) Spectrum matching for Zone 4 earthquake

about elastic viscous damping to bemade for the design models, and thephysical test. Since the design wasbased only on the first inelastic modeof response, this equivalence was onlyrequired for the first mode. As dis-cussed subsequently, it was necessaryto apply higher damping values tohigher modes at later stages of thetesting.

Since the initial input to the buildingwas based on an initial estimate ofelastic stiffness, flexibility tests wereperiodically carried out at differentstages of the test sequence. In thesetests, actuator forces were applied se-quentially and independently to thefive floor levels, with displacements ateach of the floors being recorded foreach floor force input. This enabledthe flexibility matrix to be established,and after inverting and smoothing, theinitial stiffness matrix was defined.

The initial stiffness matrix was alsoneeded in the iteration procedureused to obtain the required displace-ments at each time step. Displace-ments defining the structural responsewere directly measured on the struc-ture, whereas the command displace-ments were applied and monitored byinternal transducers in the actuators.Inevitably, a discrepancy betweencommand and external response dis-placements existed, primarily due toflexibility in the load path.

To remedy this situation, an errorcorrection procedure was carried out,


multiplying the vector of displacementerrors by the initial stiffness matrix,and applying the resulting floor forcesas corrections. This procedure was it-eratively carried out until the displace-ment errors at each floor level werewithin a predetermined tolerance. Fur-ther details on multi-degree-of free-dom pseudodynamic testing are foundin Ref. 10.

The third type of test carried out in-volved cyclic testing under a forcevector in the form of an inverted trian-gle to specified roof-level displace-ments. After each earthquake input,one or two complete cycles of re-sponse were applied to the peak dis-placement achieved during the pseu-dodynamic sequence. This was donefor three reasons:

First, the pseudodynamic sequencesdid not exercise the structure to thesame displacements in each direction(positive and negative). The invertedtriangle test exercised the building tothe same peak displacement in bothdirections.

Second, the pseudodynamic se-quences, though matching the requiredspectrum were comparatively short, asa consequence of the time taken to ac-tually apply the record (typically, 20to 30 minutes per second of earth-quake record). As a consequence, onlyone displacement peak to maximumresponse was typically achieved in thetests. The inverted triangle tests en-abled the structure to be subjected tomore cycles at peak response.

Third, during the inverted triangletests, the test was halted at each dis-placement peak, to enable the physicalcondition of the building to be exam-ined, cracks marked, and photographstaken.

Initial testing was carried out at afraction of the EQ1 level (see Fig. 5a)to enable the test control to be refined.Following this, the seismic intensitywas gradually increased up to the max-imum level. Table 1 lists the tests car-ried out in each direction of response.

The wall direction was tested first,with actuators reacting against a spe-cially constructed precast reactionwall. During application of the maxi-mum earthquake segment of 1.5xEQ3(equivalent to EQIV in Fig. 5a), thedisplacements of the precast reaction

Fig. 6. Concept of pseudodynamic testing.

wall exceeded the safe maximumvalue, and as a consequence the subse-quent inverted triangle force input wasnot applied in this direction. Actuatorswere then moved to the frame direc-tion and testing was continued.

In the frame direction, it was foundthat higher mode story force levels atEQ3, the design level of input weresignificantly higher than anticipated,and the connections of actuator tofloor, and floor to frame were exceed-ing design ultimate values. As a con-

sequence, subsequent testing was onlycarried out using the inverted triangleforce pattern, which exercised thestructure to maximum displacements,but did not input the higher modestory force levels. This is discussedfurther in the results section of thispaper.

Dynamic characterization of thebuilding was carried out using ashaker attached to the top of the build-ing before the start of the frame testprogram, and also after its completion,

48 PCI JOURNAL

enabling the first three elastic modesto be determined, and the influence ofdamage on the frequencies and modeshapes to be examined.

All critical phases of testing oc-curred at night, primarily for reasonsof thermal stability of instrumenta-tion, but also because other activitiesin the Structural Systems Laboratorycould not be carried out while thebuilding was under test. Thermalstability was of particular impor-

tance in the wall direction, since thestructure of the laboratory frame wasused as the reference system for lat-eral displacements.

It was found, prior to testing, thatdisplacements up to 0.1 in. (2.5 mm)occurred at different levels of the lab-oratory frame due to direct solar radia-tion on the east side in the morning.The thermal displacements peaked atabout 1 p.m., and gradually reduced tozero by about 8 p.m.

INSTRUMENTATION AND DATA RECORDINGPrimary data recorded included the

actuator forces and displacements, andthe displacements at each side of thebuilding at each floor level. These lat-ter displacements were recorded at thetransverse centerline of the buildingfor the test direction considered.

Wall uplift displacements at thebase, and relative vertical displace-

Duration ∆-Max Base ShearBeam Designation Test No. (pseudo-time) (in.)** (kips)**

Wall Test

IT 0 023 0.15 71.4

0.25 EQ1 025 2.952 0.20 80.5

0.5 EQ1 032 3.972 0.41 142.31.0 EQ1 033 3.972 1.22 315.3

IT 1 034* 1.25 186.6036 1.25 174.3

1.0 EQ1 038 3.0 1.70 286.4-1.0 EQ1 039 3.972 1.78 292.6

1.0 EQ2 040 5.304 3.00 294.7

IT 2 041* 3.00 221.3

-1.0 EQ2 046 5.208 2.84 299.4

1.0 EQ3 047 0.6 2.10 302.51.0 EQ3mod 049 0.6 2.07 214.51.0 EQ3mod2 050 0.606 1.94 260.9

1.0 EQ3mod5-10 051 5.772 8.31 321.5

IT 3 052 8.31 278.2053 2.00 217.9054 8.31 278.2

-1.5 EQ3mod5-10 055* 5.688 11.6 464.9

Frame Test

0.25 EQ1 107 3.972 0.57 120.70.5 EQ1 108* 3.972 1.17 201.21.0 EQ1 110 3.972 2.63 340.8

IT 1 111 2.63 258.1

1.0 EQ2mod 112 5.304 6.84 333.2

IT 2 113* 6.86 342.8

1.0 EQ3mod5-10 116 3.408 9.85 353.5

IT 3 117 10.00 349.3

IT 4 119 11.88 347.1

IT 5 120 18.58 369.9

Table 1. Summary of time duration, maximum displacement and base shear for wall and frame tests.

* Flexibility test performed.

** Absolute value.

Note: 1 in. = 25.4 mm, 1 kip = 4.45 kN.


ments across the central joint werealso recorded at the base and top ofthe wall. Potentiometers were placedto measure overall beam elongations,and also, in the form of transducer ar-rays on selected joints, to monitorjoint deformations. In addition, a largenumber of electrical resistance straingauges were placed on reinforcingbars in a number of locations.

At the time of writing this prelimi-nary report, many of these data remainunexamined, and will be reported onat a later stage. In this paper, overallstructural performance, recorded in theform of forces and displacements, willbe of prime interest.

In addition to electronic instrumen-tation, the condition of the buildingwas recorded by photographers, aftermarking cracks with felt pens. To fa-cilitate crack detection, the entirebuilding was painted with a white un-dercoat prior to testing. Time-lapsevideo recording, using three cameras,was made at all the critical loadingstages. Fig. 7 shows a general view ofthe building during the test.

TEST RESULTSDiscussed herein are the general ob-

servations and force-displacementcharacteristics in the wall directionand frame direction of loading.

Wall Direction of Response

General Observations — In thewall direction of loading, the designdrift under the design level earthquake(Zone 4 intensity) represented a rooflevel displacement of 9 in. (228 mm).Under EQ3, the peak recorded dis-placement was 8.3 in. (211 mm), 8percent below the target design dis-placement. This is considered to be avery satisfactory result, and a goodverification of the displacement-basedprocedure used to design the structure.

Damage to the building when testedin the wall direction was minimal,even after subjecting the structure toan earthquake of intensity 50 percenthigher than the Zone 4 level input mo-tion. No cracking developed in thewall, except at the base connection tothe foundation. Fig. 8a shows the con-dition of the walls at the location ofthe lower U-shaped flexure plates con-

Fig. 7. Overall view of five-story building under test.

necting the walls. The relative verticaldisplacement at the maximum rooflevel displacement of 11.6 in. (294mm) is clearly apparent from the off-set of the lower ends of the plate re-cesses in the walls. Minor crushing de-veloped at the wall base at each end,shown in Fig. 8b, for a height of about6 in. (150 mm) above the foundation.This damage was essentially cosmeticand could have been easily repairedwithout disrupting the normal opera-tions of the building.

Minor cracking was developed atthe base of all columns, but it wasclear from the small width of thesecracks that no yielding of reinforce-ment in the columns had occurred.

Fig. 8c shows the crack pattern at thebase of an external gravity load col-umn at the end of testing in the walldirection. The only other damage ob-served was hairline cracks in the floorsystems, particularly in the cast-in-place topping above the joints betweenthe hollow-core slabs.

Force-Displacement Characteris-tics — It is impossible within thespace limitations of this paper to pro-vide the results for all of the tests car-ried out in the two directions of load-ing. As a consequence, only the mostimportant tests will be discussed.However, the displacement profiles upthe height of the building at the peakroof displacement for different levels

50 PCI JOURNAL

of input excitation in the wall direc-tion are shown in Fig. 9.

These results provide some indica-tion of the nonlinear relationship be-tween excitation intensity and peakdisplacement. For example, at EQ1excitation, the intensity was 33 per-cent of the design level, but the rooflevel displacement was only 14.7 per-cent of that recorded under EQ3, thedesign level event.

Initial testing at the design levelEQ3 had to be aborted after 1.0 sec-onds of scaled earthquake record be-cause of unacceptably high floor forcelevels. Before resuming the testing,the earthquake record was filtered toreduce the severity of third to fifthmode effects. Analyses of the struc-ture using models described subse-quently in this paper confirmed thatthis filtering would not influence thepeak displacements or base shearforces, which were primarily influ-enced by the first (inelastic) and sec-ond modes of response.

The measured displacement re-sponse of the building in the wall di-rection to 1.5xZone 4 (i.e., 1.5xEQ3input) is shown in Fig. 10. Displace-ments for each floor level are plottedagainst time, where the time is basedon the input accelerogram time scale,and needs to be divided by 0.6 to pro-vide prototype time.

The displacement response appearsto be dominated by first mode defor-mation patterns, with displacements atdifferent floor levels approximatelyproportional to the height at all times.However, examination of the storyshear force time-history response forthe same test, shown in Fig. 11, indi-cates that there was a considerableamount of higher mode response, par-ticularly in the earlier part of therecord. During the maximum displace-ment response pulse, the force re-sponse is closer to that expected of thefirst inelastic mode.

This difference in sensitivity tohigher modes between the displace-ment and force response is to be ex-pected, since only the first mode is sig-nificantly modified by ductility.Higher modes remain elastic and thedisplacement response for these modesis very small because of their largestiffness. It should be noted that peak

Fig. 8. Damage to building

after testing to1.5xZone 4 intensity.

(a)Wall conditionat lower U-plates

(b)Crushing atwall base

(c)Cracking atbase of gravitycolumns


shear forces for the lower floors of thebuilding occur in the early stages of re-sponse, not when the structure reachedthe peak lateral displacement.

The overall hysteretic characteristicsof the building in the wall directionare best illustrated by the base mo-ment/roof-level displacement hystere-sis response. Two examples are givenfor the wall direction in Figs. 12 and13, respectively.

The first example (see Fig. 12) plotsthe response under the inverted trian-gle excitation carried out to the peakdisplacement response obtained duringthe EQ3 design-level input. This en-ables the shape, stability and symme-try of the response to be observed. Itwill be noted that the energy dissi-pated by the wall is considerable, andthat equal strength was obtained inboth directions of response.

Fig. 13 shows the moment-displace-ment response obtained during thefinal test in the wall direction, namely,the pseudodynamic test to 1.5xEQ3(1.5x Zone 4). The shape of the hys-teretic response is now much more ir-regular than for the inverted triangletest, but the high level of energy dissi-pation is still apparent. The low resid-ual drift at the end of testing is identi-fied by the large solid circle.

It is of interest to examine the en-velopes, with height, of various ac-tions developed in the wall. To thisend, Fig. 14 compares the envelopesof story overturning moment, storyshear force, and floor forces obtainedfrom testing the building at the EQ3design level. In this arrangement, theenvelopes correspond to the design in-verted triangular distribution of forces.The floor forces represent the actualforce levels applied to the floors bythe two actuators at each floor level.

The increase in the maximumrecorded base moment over the designlevel (see Fig. 14a), of approximately25 percent, represents the over-strength resulting from the U-shapedenergy dissipators having higherstrength than the design value; andalso, to a lesser extent, due to lateralresistance provided by the columns ofthe building, which were ignored inthe design approach.

At levels above the base, it will beobserved that the experimental distri-

Fig. 9. Peak displacement profiles up the wall at different levels of seismic excitation.

Fig. 10. Time-history of wall direction displacement response to1.5xZone 4 excitation.

Fig. 11. Time-history of wall direction story shear force response to1.5xZone 4 excitation.

52 PCI JOURNAL

bution of moment is almost linear withheight, whereas the design distributionis concave upwards. This discrepancyin shape is the result of higher modeeffects. It is noted that the linear distri-bution of moment with height con-forms to recommendations for can-tilever structural walls made by Paulayand Priestley.11

The discrepancy between the designand experimental shear force distribu-

tions is much more marked than for mo-ments. As can be seen in Fig. 14b, themaximum base shear at the design levelof excitation was 63 percent higher thanthe design level. The great majority ofthis excess shear demand is the conse-quence of higher mode effects. It is alsonotable that the shear demand reducedonly gradually with height.

Paulay and Priestley11 have recom-mended a dynamic shear amplification

Fig. 12. Wall base-moment/roof-level displacement hysteresis response to invertedtriangular force input to 8.3 in. (211 mm) peak displacement.

Fig. 13. Wall base-moment/roof-level displacement hysteresis response to1.5xZone 4 excitation.

for a five-story cantilever wall struc-ture of 40 percent to account forhigher mode effects. This shows a rea-sonable agreement with the experi-mental value, when the overstrengthdue to flexural enhancement, apparentin Fig. 14a, is included.

The biggest differences between ex-periment and design levels were in thefloor force levels. As seen in Fig. 14c,the floor force levels greatly exceededthe design values at all heights of thebuilding. It is also clear that the floorforces do not show any tendency to re-duce in the lower levels, as would beexpected from a predominantly firstmode response. For the wall directionof response, a rough approximation tothe floor force levels for design pur-poses would be to apply 75 percent ofthe design base shear as a floor forceat each level.

Note that these high floor force lev-els, while not significantly influencingthe moment demand at the base, areimportant since they represent themagnitude of diaphragm forces thatmust be transmitted from floors to lat-eral force-resisting elements. Thesehigh force levels caused considerableproblems in the testing of the PRESSSbuilding.

Recognizing the potential impact ofhigher mode effects, Ms. SuzanneNakaki and Professor John Stanton,the building designers,2 had designedall diaphragm-to-wall and diaphragm-to-frame connections for force levelscorresponding to 50 percent higherthan the top story design level. Actua-tor connections to the floors were de-signed for the same force levels.

In the test, the design ultimatestrength of these actuator connectionswas exceeded by up to 44 percent, andsignificant yielding of the connectionsoccurred. Without material over-strength in the connections, failurewould have occurred. However, as aconsequence of the unexpectedly highfloor forces, it was necessary to pro-vide extra damping to the fourth andfifth modes in the pseudodynamic testcontrol. Without this extra damping,floor forces would have been evenhigher.

It is possible that a part of the highermode effects resulted from the itera-tion procedure used to impose the cal-


culated displacements at each timestep, discussed previously. However, acomparison with analytical results in-dicated that similar levels of highermode effects were predicted. Further,as noted above, the floor forces werereduced in the pseudodynamic tests byapplying extra damping to the highermodes. It is clear that current Ameri-can design practice does not ade-quately consider the influence ofhigher mode effects.

As a consequence of the negligibledamage and the very low residualdrift, it was apparent that after beingsubjected to an earthquake 50 percenthigher than the design intensity, thebuilding had responded within the ser-viceability or “immediate occupancy”level of performance defined in Ref. 6.

Frame Direction of Response

General Observations — Responseof the building in the early stages ofloading was characterized by low levelsof damage, with all inelastic action oc-curring in the beam-to-column connec-tions, as intended. Minor flexuralcracking was noted in the beams of theTCY frame at locations other than thebeam-column interfaces, and minorshear cracking of the beam-to-columnjoints was noted on testing to EQ2, cor-responding to 50 percent of the designintensity. This joint cracking was muchmore apparent in the non-prestressedframe than the prestressed frame, andwas more noticeable in externalcolumns when subjected to seismicaxial tension than axial compression.

At EQ2, incipient spalling wasnoted at the base of the beams of theTCY-gap connections where theywere connected to the columns, as wasa slight tendency for the beams toslide up the column interface.

During testing to EQ3, the designlevel of ground motion, it was ob-served that the beams of the hybridframes experienced some inward rota-tion about their longitudinal axes. Thistorsional rotation and the cause of therotation are illustrated in Fig. 15. Theheavy double-tee floor members werehung off the side of the beams at thelower three floors, causing a torsionalmoment of significant magnitude.Under inelastic action, the torsion inthe hybrid beams was primarily re-

Fig. 14. Comparison of design and experiment force envelopes at design level(Zone 4) excitation.

(a) Story overturning moments

(b) Story shear force

(c) Floor forces

54 PCI JOURNAL

sisted by the compression zone at thebeam-column interface, with a smallcontribution from dowel action of themild steel damping bars crossing theinterface.

The torsional resistance of these ac-tions was insufficient to prevent therotation at the lower three floors of theprestressed frame. The very high floorforce levels also exacerbated the prob-lem since these provided additionalvertical forces to the floors, which inone direction of response more thandoubled the reactions of the doubletees on the prestressed beams, andhence increasing the torsion by a simi-lar amount.

Following testing at EQ3 input, thesteel base of the floor beams waswelded to the brackets on the side of

Fig. 15. Rotation of hybrid beam due to connection detail with double-teefloor member.

Fig. 16. Condition of frame connections after design level of response.

(a) Hybrid post-tensioned connection

(b) Pretensioned connection

(c) TCY-gap connection

(d) TCY connection


the prestressed beams, effectivelylocking in the rotation at the existinglevel, and prohibiting further rotation.It is clear that support of these heavydouble-tee floor members should beprovided as close to the beam center-line as possible to reduce the torsionaleffect, which would be importantwhether the frame was precast or castin place.

It should also be noted that al-though the test building was sub-jected to additional forces from theactuator reactions, as noted above,the actual dead load reactions wereless than in the prototype building be-cause of the scale factor involved.Calculations showed that the total re-action in the test building was similarto that to be expected in practice. De-

spite continuing the test with the hy-brid beams having a locked-in tor-sional rotation of more than 3 de-grees, there were no visible signs ofthis influencing the response.

On testing the frame with the hybridbeams under the design level EQ3input, a maximum roof-level displace-ment of 10 in. (254 mm) was sus-tained, corresponding to an averagepeak drift 10 percent higher than thedesign level of 2 percent, althoughlocal interstory drifts were higher.Again, this is seen to be a good verifi-cation of the displacement-based de-sign procedure. It is believed that theslightly larger than expected drift wasa consequence of effective dampinglevels being somewhat less than thoseassumed in the design.

Fig. 16 shows the condition of typi-cal frame connections after subjectingthe building to the EQ3 design levelinput, and two subsequent cycles ofinverted triangle testing to the samepeak displacement sustained in thepseudodynamic test. As is clear fromFig. 16a, the hybrid prestressed con-nections exhibited little or no damageat this stage, despite the torsional ro-tation noted above. Joint crackingwas minor, with maximum crackwidths of about 0.005 in. (0.12 mm).Very little cracking had developed inthe beams except at the connection tothe column.

Similarly, excellent behavior wasexhibited by the pretensioned connec-tions provided at the upper two levelsof the prestressed frame (see Fig.

Fig. 17. Condition of frame connections after being subjected to drift levels more than twice design level.

(a) Hybrid post-tensioned connection

(b) Pretensioned connection

(c) TCY-gap connection

(d) TCY connection

56 PCI JOURNAL

16b). The inelastic action for theseconnectors was concentrated in singlecracks that developed at or near the re-veals formed in the beams in line withthe column faces. Joint cracking in thepretensioned connections was less ap-parent than at the hybrid connectionsprovided at the lower three levels. Atthe design stage of seismic response,no repair would have been needed tothe prestressed frame.

Fig. 16c shows the condition of atypical interior TCY-gap connectionafter the design level excitation. It isto be noted that there is significantspalling from the soffit of the beamsimmediately adjacent to the connec-tion. Also, there is some crushing anddeterioration of the fiber-grout pad atthe base of the gap. It had been ob-served during testing to this level thatthere was upwards sliding of thebeams on the grout pads during maxi-mum response levels.

Analysis showed that this was theconsequence of an insufficient clamp-ing force in the post-tensionedthreaded bars at the base of the con-nection. These bars had been sized toprovide a sufficient clamping force toresist the anticipated shear at the beaminterface when the top mild steel con-nection bars yielded in compression.Under this condition, the clampingforce across the grout pad is reducedby an amount equal to the top rein-forcing bar compressive force.

Although an overstrength factor of1.25 was applied to the top compres-sive nominal yield force in accor-dance with American practice, thisdid not provide an accurate represen-tation of the actual overstrength re-sulting from yield strength exceedingthe nominal minimum value, coupledwith the strength increase resultingfrom strain hardening. In actual prac-tice, higher overstrength factorsshould be provided.

The condition of the upper TCYconnections was good, as illustrated inFig. 16d, this despite the observationthat incipient sliding was occurring atthe interface. It should be emphasizedthat this sliding or shear deformationin the plastic hinge region has beencommonly observed in tests of rein-forced concrete beam-to-column con-nections at moderate ductility levels. Fig. 18. Condition of various structural elements at end of testing.

(a) Hollow-core slab topping to seismic frame connection

(b)Welded X-platesbetween double-tee floor andseismic frames

(c)Seismiccolumn/foundationconnection


Fig. 17 shows the condition of thevarious connections after cycling tomore than twice the design level drifts.The hybrid prestressed connections(see Fig. 17a) were performing ex-tremely well at this stage, with onlyminor damage in the form of coverspalling, and some crushing and incip-ient break-down of the fiber groutpads between the beams and columns.No significant increase in width of thejoint shear cracks above that sustainedat the design level was observed. It isprobable that the joint strain gauge re-sults will show that substantial reduc-tions in joint reinforcement will bepossible with this detail.

The performance of the preten-sioned connections in the upper floorsof the prestressed frame was also ex-cellent, as shown in Fig. 17b. Damagein these connections was also limitedto superficial cover spalling at, or ad-jacent to the beam-column interface.

Sliding of the TCY-gap connectionshad continued during testing levelsabove the design level. As a conse-quence of the high tension strainsfrom the seismic response coupledwith the dowel bending caused by theinterface sliding, a few of the mildsteel reinforcing bars crossing the in-terfaces fractured in the latter stages oftesting. It is clear that sliding of theTCY-gap beams could have beenavoided if the clamping force providedby the post-tensioned threaded bars atthe base of the beams had been higher.

It was also observed that earlycrushing of the cover concrete may beinevitable with this design detail be-cause of the high compressive forceacross the grout pad when the top barsare in tension. Stregthening the beamends or some other detail to inhibitcrushing may be advisable in futuredesigns.

Fig. 17c shows the condition of atypical TCY-gap interior-column con-nection at the end of testing. Damageis still much less than would be ex-pected from a conventional reinforcedconcrete beam-to-column joint at thislevel of drift. Design recommenda-tions for improving this detail will fol-low in a subsequent paper.

Damage to the TCY connections inthe upper levels of the reinforced con-crete frame was not significantly dif-

Fig. 19. Peak displacement profiles up the frames at different levels ofseismic excitation.

Fig. 20. Time-history of frame direction displacement response to Zone 4 excitation.

Fig. 21 Time-history of frame direction story shear force response toZone 4 excitation.

58 PCI JOURNAL

Fig. 22. Frame direction gross base-moment/roof-level displacement hysteresisresponse to Zone 4 excitation.

ferent from that at the design level. Asmall amount of interface sliding con-tinued to occur at these higher levelsof response, but crack patterns andspalling did not noticeably change.Some loss of bond between the toplevel reinforcing bar in the grout ductswas noted at the exterior connections,with the headed reinforcing bar, whichwere exposed at the ends of the con-nection being pushed out by up to 1 in.(25 mm). There did not appear to beany significant strength reduction as-sociated with this action. Fig. 17dshows the condition of a typical TCYconnection at the end of testing.

The condition of selected otherstructural elements at the end of thetest program is shown in Fig. 18. Theconnection detail between the cast-in-place hollow-core topping and theframes, formed by drag bars from thetopping connecting into pockets in thetop of the frame beams, shown in Fig.18a performed very well. There wasonly minor cracking along the inter-face, despite the very high levels offloor forces.

The welded X-plate connectors be-tween the double-tee floor membersand the frame beams also performedwell, although these exhibited signifi-cant inelastic action and permanentdistortion, as illustrated in Fig. 18b.Damage to the column bases was min-

imal, with very minor spalling beingobserved at some, but not all of theseismic columns (see Fig. 18c)

Force-Displacement Characteris-tics — Peak displacement profiles upthe frames at different levels of seis-mic excitation are plotted in Fig. 19. Itwill be noted that despite the differentcharacteristics, particularly relating toinitial stiffness of the two frames, andto a lesser extent, relating to strength,the test procedure ensured that the twoframes were subjected to the same dis-placement at all times. The displace-ment profiles show a pronounced cur-vature, with reduction in drift at theupper stories, particularly in the earlystages of testing. This is believed to bea consequence of the excess strengthprovided at the upper levels.

Time histories of displacement re-sponse and story shear force at differ-ent levels of the building are shown inFigs. 20 and 21, respectively, underthe EQ3 design level of response. Thistest ended slightly prematurely due toexcessive story force levels, despitefiltering the earthquake record, as dis-cussed earlier. However, at this stage,the peak displacement response hadalready occurred and the record waswithin 0.5 seconds of entering the freedecay section of input, i.e., the inputrecord had a trailing section of zeroacceleration to enable free decay to be

observed. The abrupt termination ofthe record is apparent in both Figs. 20and 21.

The story shear force plots representthe sum of the shear force in the twoframes. It will be observed that a verysignificant higher mode response isparticularly apparent in the earlystages of the story shear forces, but ismuch less apparent in the displace-ment response, or during the peak dis-placement pulse, as was observed forthe shear wall direction of testing.

The gross base moment/roof dis-placement hysteresis responserecorded during the EQ3 response isplotted in Fig. 22. The moment here isthe overturning moment found fromthe sum of the response of the twoframes. It will be noted that althoughsignificant energy dissipation is appar-ent in Fig. 22, it is somewhat less thanthat for the wall direction of response.

In Fig. 23, the hysteretic base mo-ment/roof displacement responseunder the inverted triangle loading to12.2 and 17.9 in. (310 and 454 mm)peak displacements is plotted with theresponse of the two frames being sep-arated. Since two actuators applied thelateral forces at each level, with oneactuator adjacent to each frame, andsince the torsional resistance providedby the central wall was insignificant,the determination of the forces carriedby each frame was straightforward.

It will be noted that the peak mo-ments sustained by the two frameswere very similar, but that the hys-teretic response of the frames was verydifferent, with the non-prestressedframe exhibiting more energy dissipa-tion, and more residual displacementthan the prestressed frame, as was ex-pected. The lack of significant strengthreduction on testing through two com-plete cycles at the lower displacementsof Fig. 23 is apparent.

On the second cycle to peak dis-placement at 17.9 in. (454 mm), thetest control system developed a severeinstability resulting in the buildingbeing subjected to violent shaking, andan emergency shut-down occurred. Al-though this instability, which was elec-tronic, rather than structural, did notappear to induce any further damage tothe building, the test program was ter-minated at this stage.


Fig. 24 plots the envelopes of over-turning moment, story shear force, andstory force levels as functions ofheight, for the EQ3 design level earth-quake, and compares them with thedesign envelopes. Trends apparent inFig. 24 are similar to those for the walldirection of response (see Fig. 14), al-though the extent of overstrength isless, particularly for overturning mo-ments and story shear. To some extentthis can be attributed to the strengthreduction of the TCY-gap connectionsdue to premature spalling and inter-face sliding, which was noted earlier.

ANALYTICAL MODELINGOF RESPONSE

Prime responsibility for the analyticalmodeling of the building rests with a re-search team from Lehigh University,who will be reporting separately on ex-periment versus prediction comparisonsat a later date. Their modeling has beenbased on fiber-element representation ofthe various structural components.However, as part of the test control, itwas necessary to have the ability to pre-dict the response of the building prior toperforming pseudodynamic tests.

Fig. 23. Frame base-moment/roof-level displacement hysteresis response to inversetriangle loading.

To this end, simple models were de-veloped to represent the wall andframe directions of response, based onone-dimensional inelastic elementswhose properties were based on sec-tion analyses. Essentially, these weretypical design-level representations ofthe building.

Wall Direction Model

The analytical model for the wall di-rection response prediction is shown inFig. 25. The two walls were separatelyrepresented by elements at their center-lines and given structural propertiesbased on the uncracked section. Thewalls were supported at the base onrigid horizontal beams supported ateach end by compression-only springslocated at the calculated position of thecenter of compression at each wall endunder rocking response. Properties ofthese springs are indicated in Fig. 25.

As shown in Fig. 25, the springswere preloaded by forces representingthe gravity weight and unbonded pre-stressing force at the wall base. As thewall rocks, the unbonded prestressingthreaded bars extend, resulting in anincreased clamping force at the base.This was modeled by an additionalspring directly under the wall element,as shown in Fig. 25.

The two wall elements were con-nected by vertical springs acting be-tween rigid horizontal links extendingat each floor level. These inelasticsprings represented the characteristicsof the U-shaped flexural yielding en-ergy dissipators connecting the wall,which are apparent in Fig. 3. Theproperties for these springs were de-termined by testing spare dissipatorsprior to the building test. It should bementioned that any additional strengthand stiffness in the wall direction re-sulting from column stiffness were ig-nored in the analysis.

This simple model enabled realisticestimates of the force displacement re-sponse to be obtained prior to testing.Prior to each earthquake record beingimposed on the test building, an in-elastic time-history analysis was runon the analytical model using the Ru-aumoko computer program.12 Thus,the results from these analyses weretrue predictions (rather than post-dic-tions) of response. No modification of

(b) Peak displacement = 17.9 in. (454 mm)

(a) Peak displacement = 12.2 in. (310 mm)

60 PCI JOURNAL

the analytical model was made afterthe test program started.

Wall Direction Prediction andExperiment Comparisons

Space limitations do not permitmore than a limited comparison be-tween the analytical predictions andthe experimental results. A full de-scription of the analytical model de-tails and analysis results will shortlybe available from UCSD.8 A compari-son of the predicted and observed timehistories of roof-level displacement,base shear force, and base momentfor the 1.5EQ3 maximum responselevel are shown in Fig. 26.

The results show that the agreementbetween the predicted and experimen-tal values is extremely close for thedisplacement time-history shown inFig. 26a. This is especially true whencorrection is made for the initial ex-perimental offset displacement, indi-cating that the model accurately cap-tured the fundamentals of the firstmode, including inelastic action, andwas capable of excellent prediction ofpeak displacement response.

The agreement for base shear forceis also very good (see Fig. 26b) withthe level of base shear contributed bythe first mode response being pre-dicted with an accuracy equivalent tothat for the displacements. The com-ponent of base shear due to highermode effects was also well predicted,although the experimental resultsshow somewhat higher shear forcesfrom higher modes, apparent as per-turbations on the first mode response,than predicted.

As discussed earlier, it is possiblethat this is a result of an artificial em-phasis of the higher modes as a conse-quence of the displacement error-correction procedure used in thepseudo-dynamic test method. It is em-phasized that the analytical modelmade the same damping assumptionsas incorporated in the pseudodynamictime-history control procedure.

A comparison between the exper-imental and predicted values for thebase moment time-history (see Fig.26c) is also satisfactory. Note thatthe base moment is less affected byhigher mode effects than is the baseshear force. It is clear, however,

Fig. 24. Comparison of design and experiment force envelopes at design level(Zone 4) excitation.

(a) Story overturning moment

(b) Story shear force

(c) Floor forces


that the strength was underesti-mated by about 20 percent at peakresponse.

A final comparison between the ex-perimental and predicted values is pro-vided in the base moment/roof leveldisplacement envelopes of Fig. 27.This figure includes the experimentalresults from the full test sequence andcompares them with the analytical re-sults from a pushover analysis of thebuilding, using the simple analyticalmodel.

It will be seen that the agreement isreasonably good, with experimentalpeak values tending to exceed the pre-dicted envelope by up to 25 percent.The initial stiffness of the model issomewhat higher than the experimen-tal value.

This will be partly due to the sup-port conditions assumed for themodel, which consisted of point sup-ports close to the wall extremities (seeFig. 25), and also due to the fact thatthe floor level displacements in thetest were measured on the floors rather

than on the wall. The analytical modelassumed rigid diaphragm behavior forthe floors, whereas there was a smallbut significant flexibility between thefloors and the wall resulting from theconnection details.

The strength increase apparent inFigs. 26 and 27 can be attributed totwo causes. In the wall direction, thesix seismic frame columns and the twogravity columns would have con-tributed a small but significant amountto the base moment capacity of thestructure. This was ignored in the ana-lytical model.

Also, it is expected that the prestresslosses in the tendons providing post-tensioning to the wall elements wasless than the design value, which wasused in the predictions. The influenceof these two effects is sufficient to ex-plain the overstrength.

Overall, the agreement between theexperimental and predicted results isbelieved to be extremely satisfactory,considering the simplicity of the ana-lytical model.

Frame Direction Model

The analytical model for the framedirection of response consisted of twoplanar frames slaved to have the samedisplacement at each floor level. Again,all members were represented by one-dimensional elements. Since the inelas-tic response was only expected at thebeam-to-column or column-to-founda-tion connections, the members weremodeled by elastic elements, represent-ing the appropriate stiffness.

For the prestressed frame, thismeant using a gross-section stiffnessfor the beams and a cracked sectionstiffness for the columns. For the non-prestressed frame, beam stiffness wasinitially based on the gross-section, re-ducing to 35 percent of the gross-sec-tion stiffness after cracking. Inelasticaction at the beam-column interfacesof both frames was represented by in-elastic moment-rotation springs at thebeam-column interfaces.

Fig. 28 illustrates the fundamentalsof this modeling for a hybrid pre-

Fig. 25. Analytical model for wall direction response prediction.

62 PCI JOURNAL

stressed connection. The location ofthe centroid of concrete compressionwas established by hand calculationsas a function of joint rotation. As thejoint opened, the extension and hencethe force of the prestressing tendoncould be calculated.

Similarly, making appropriate as-sumptions about strain penetration ofthe mild steel damping reinforcingbar, its force could be determined.The separate relationships betweenmoment and rotation provided sepa-rately by the prestress tension forceand the reinforcing bar forces couldthus be determined.

For a pushover analysis, the twocomponents of moment were simplyadded to provide a composite mo-ment-rotation curve. For an inelastictime-history analysis, separate hys-teretic relationships for the prestressand reinforcing bar were modeled byusing two inelastic springs in parallel,as indicated in Fig. 28.

The moment component from theprestress tension force was modeledby a nonlinear elastic spring (i.e., themodel unloaded down the loadingcurve), without hysteretic energy dis-sipation. A Takeda degrading stiffnessspring13 was used to model the mo-ment component resulting from theyielding reinforcing bar.

The principles implied in Fig. 28 forthe hybrid prestressed connectionwere applied, with modifications, tothe other connection types. Thus, themoment-rotation behavior of the pre-tensioned connection at the upper lev-els of the prestressed frame was mod-eled solely by a nonlinear elasticspring, since there was no additionalenergy dissipation from mild steel re-inforcing bars at these locations. Mo-ment-rotation springs for the non-pre-stressed frame were modeled solely byTakeda degrading stiffness springs.Ruaumoko, the time history programused for the analyses contained bothspring types within its extensive ele-ment library.

Frame Direction Prediction andExperiment Comparisons

Unlike the wall direction of re-sponse, the frame direction predictionswere not calculated prior to testing thewall. Two rather minor modifications

Fig. 26. Comparison of predicted and observed wall response at1.5xZone 4 excitation.

(a) Roof-level displacement

(b) Base shear force

(c) Base moment


Fig. 27. Comparison ofpredicted andobserved walldirection moment-displacement enveloperesponse.

Fig. 28. Analytical model for hybrid prestressed frame connection.

64 PCI JOURNAL

were made subsequent to the tests toimprove agreement between the ex-perimental and predicted values.

The first of these involved the hys-teretic rule adopted for the energy-dissipating inelastic springs. In thepretest predictions, a bilinear elasto-plastic hysteresis rule was used. Thiswas subsequently modified to theTakeda degrading stiffness model, asnoted above, which is known to pro-vide better representation of rein-forced concrete behavior.

The second modification involvedthe definition of the initial conditionsfor the model. In the pretest analyses,each test was predicted independently,as though the frames were initially inthe undamaged condition, withouthaving been subjected to the earlierlevels of seismic excitation, eventhough significant inelastic actionmight have occurred.

With a modified Takeda representa-tion of response, the initial stiffnessfor subsequent tests should reflectdamage and cracking in the earlierstages. As a consequence, the finalanalyses for the frame direction wereobtained by a sequential analysis,where the entire sequence of earth-quake records were run in the sameorder as occurred during testing. Thus,the starting condition for one test de-pended on the state at the end of theprevious test. These two modificationssignificantly improved (by about 15percent) the “predictions.”

Note that no changes to strength,stiffness, or moment-rotation charac-teristics were made subsequent to thestart of testing. Thus, the envelope ofresponse from a pushover analysis wasunaffected by the modifications, and itis felt reasonable to still describe theanalyses as predicted results.

Comparisons of experimental andpredicted values for time-histories ofroof-level displacement, base shearforce, and base moment are plotted inFig. 29 for the EQ2 excitation level.Good agreement is obtained for rooflevel displacements, with the peak ex-perimental displacements exceedingthe predicted level by less than 10percent.

It was earlier noted that experimen-tal drifts at the design level of re-sponse exceeded the design drift byFig. 29. Comparison of predicted and observed frame response to Zone 4 excitation.

(a) Roof-level displacement

(b) Base shear force

(c) Base moment


about the same amount. It is probablethat the reason for this increase is thelevel of damping exhibited by themodel.

The moment-displacement hystere-sis plots of Figs. 22 and 23 indicateequivalent viscous damping ratios,based on a secant stiffness approxima-tion of about 12 percent with lowervalues applying for the prestressedframe than for the reinforced frame.This is significantly lower than the de-sign assumption of 20 percent, result-ing in increased experimental dis-placements, despite the increasedlateral strength.

Agreement for base shear and mo-ment, as shown in Figs. 29b and 29c isalso very satisfactory. The effects ofhigher mode response in both figuresis very similar for both predicted andexperimental curves.

A final comparison between the pre-dicted and experimental values is pro-vided by the base moment/roof-dis-placement envelope response of Fig.30. In this figure, the results from theanalytical pushover analysis are com-pared with the full set of experimentalresults, including pseudodynamic testsand inverted triangle tests.

The agreement is satisfactory in thesmall to moderate displacement range.However, a discrepancy is seen at

large lateral displacements, with theanalytical model predicting higher lat-eral strengths than the experimentalvalues. The reason for this discrep-ancy is mainly due to not incorporat-ing softening behavior of the beam-column systems, as a result ofcrushing of concrete and coverspalling in the analytical model.

CONCLUSIONSThe structural response of the

PRESSS five-story precast concretetest building under simulated seismictesting was extremely satisfactory.The following summarizes the re-sponse and conclusions available atthis stage. It is emphasized that addi-tional information will become avail-able as data reduction continues.

1. Damage to the building in thewall direction was minimal, despitebeing subjected to seismic intensities50 percent above the design level.Only minor spalling at the wall base,and fine cracking in floor slabs, andnear the column bases were observed.The wall was essentially uncrackedexcept at the base during wall direc-tion response. The wall showed addi-tional flexural cracking when sub-jected to the 4.5 percent out-of-planedrift in the frame direction of re-

sponse. However, these cracks closedup to be invisible to the naked eye atthe end of testing.

2. Damage to the building in theframe direction of response was muchless than could be expected for anequivalent reinforced concrete struc-ture, subjected to the same drift levels.The performance of the prestressedframe was particularly good, withdamage being limited to minorspalling of cover concrete in thebeams immediately adjacent to thecolumns and some crushing of thefiber grout pads at the beam-columninterfaces. Crack levels in the beam-to-column joints due to shear were ex-tremely small, and it was evident thatthe amount of joint shear reinforce-ment provided, which conformed tocurrent code levels, could have beensubstantially reduced.

3. Although the non-prestressedframe also performed well, the TCY-gap connections showed more dam-age, occurring at an earlier stage, thanthe other connection types. This wasdue to the inadequate clamping forceprovided by the bottom post-tension-ing, resulting in upwards slip at thebeam-column interface under seismicshear forces. Minor changes to the de-sign criteria would correct this prob-lem, but some attention will also be

Fig. 30. Comparison ofpredicted andobserved framedirection moment-displacement enveloperesponse.

66 PCI JOURNAL

needed to improve detailing at thebeam ends to reduce the tendency forpremature spalling of cover concrete.

4. At high levels of response dis-placements, beam rotation about thelongitudinal axis was noted, caused bythe high torsional moment induced bythe vertical load from the eccentricallysupported double-tee floor members,and the reduced torsional resistance inthe beam-end plastic hinges. Modifiedsupport details should be developedtransferring the double-tee reactioncloser to the support beam centerline.

5. As anticipated, residual drift afterthe design level excitation was verylow. In the wall direction, the residualdrift was 0.06 percent after sustaininga peak drift of 1.8 percent. This corre-sponds to only 3 percent of the maxi-mum drift. The low residual drift is acharacteristic of the unbonded pre-stressing system used to providestrength in the wall direction,5 and is asignificant advantage over conven-tional cast-in-place reinforced con-crete construction, where very highresidual drifts are possible. The lowresidual drift was also apparent in theprestressed frame which was based onthe same unbonded prestress philoso-phy. The non-prestressed frame dissi-pated more energy, but suffered higherresidual drifts and somewhat higherdamage levels.

6. The test provided an excellentconfirmation of the direct displace-ment-based design approach used todetermine the required strength of thebuilding. The structure was designedto achieve drifts of 2 percent under thedesign level earthquake (Zone 4, UBCintensity), and actually sustained aver-age drifts of 1.8 and 2.2 percent in thewall and frame directions, respec-tively. This drift is within the accuracyof spectrum-matching for the test ac-celerograms. The required base-shearstrength using direct displacement-based design was only 45 and 60 per-cent of the strength required by con-ventional force-based design usingUBC provisions for wall and frame di-

rections, respectively.2

7. In the pseudodynamic tests, muchhigher floor forces were experiencedthan anticipated. These high force lev-els were confirmed by analytical stud-ies and were the result of higher modeeffects. These high floor force levelsrepresent diaphragm force levels thatare significantly higher than currentlyconsidered in design and are a conse-quence of the comparative insensitiv-ity of the higher mode force levels toductility. Current designs, where forcereduction factors are applied equallyto higher modes as well as the funda-mental mode result in a critical under-estimation of the diaphragm force lev-els. The influence of higher modefloor forces is also translated intostory shear force levels, and momentdistributions that are more severe thancurrently considered in American de-signs using an inverted triangular dis-tribution of the base shear force. How-ever, the distributions of peak storyshear force and overturning momentwere well predicted by establishedprocedures.6

8. Simple analytical models, suit-able for the design office, were used topredict the response of the building.These were found to provide verygood simulation of response, with ex-cellent correlation between predictedand observed displacements, storyshears, and overturning moments.

NOTE: This paper must be consid-ered a preliminary report on the re-sponse of this substantial research pro-gram. Additional experimental resultswill be presented in subsequent pa-pers. There will also be further paperspresented on the design and construc-tion details, the design implications ofthe test results, and more detailed andsophisticated comparisons of predictedand experimental results.

ACKNOWLEDGMENTSThe project described in this paper

has involved a large number of indi-viduals and organizations, all of whom

deserve individual thanks and ac-knowledgment. A full list would beimpossibly long. Of particular impor-tance are Dr. Chris Latham (UCSD)and Professor Akira Igarashi (KyotoUniversity) whose efforts in solvingthe extremely difficult problems ofcontrolling the pseudodynamic testswere essential to the test success.

The efforts of the building design-ers, Professor John Stanton (plus Uni-versity of Washington graduate stu-dents), and Ms. Suzanne Nakaki, whonot only did a superb job of the build-ing design, as evidenced by its excel-lent performance, but also were pre-sent during most of the long-nighttesting sessions, are particularly ac-knowledged.

Primary financial support for thePRESSS research program was pro-vided by the PCI (Precast/PrestressedConcrete Institute), the NSF (NationalScience Foundation), and the PCMAC(Precast/Prestressed Concrete Manu-facturers Association of California).The extent of industry support, interms of financial assistance, materialdonation, technical advice, and provi-sion of precast products is unparal-leled in major United States structuralresearch projects.

Special thanks are due to Mario J.Bertolini, chairman of ATLSS andPRESSS Ad Hoc Committee, andThomas J. D’Arcy, chairman of thePRESSS Phase III Advisory Group.

In addition, contributors to the test-ing program include A. T. Curd Struc-tures, Inc.; BauTech, Co.; CaliforniaField Iron Workers AdministrativeTrust; Charles Pankow Builders, Ltd.;Clark Pacific; Coreslab Structures,L.A.; Dayton Superior; Dywidag Sys-tems International; ERICO; FloridaWire & Cable, Inc.; Fontana Steel;Gillies Trucking; Headed Reinforce-ment Corporation; Horizon HighReach; JVI, Inc.; LG Design; MasterBuilders, Inc.; NMB Splice Sleeve;Pomeroy Corporation; Precision Im-agery; Spancrete of California; Sumi-den Wire; and White Cap.


1. Priestley, M. J. N., “The PRESSS Program — Current Statusand Proposed Plans for Phase III,” PCI JOURNAL, V. 41, No.2, March-April 1996, pp. 22-40.

2. Nakaki, S. D., Stanton, J. F., and Sritharan, S., “An Overviewof the PRESSS Five-Story Precast Test Building,” PCI JOUR-NAL, V. 44, No. 2, March-April 1999, pp. 26-39.

3. Priestley, M. J. N., “Displacement-Based Approaches to Ratio-nal Limit States Design of New Structures,” Keynote Address,11th European Conference on Earthquake Engineering, Paris,September 1998.

4. Priestley, M. J. N., and Kowalsky, M. J., “Direct Displace-ment-Based Design of Buildings,” Bulletin of the NZ NationalSociety for Earthquake Engineering, 1999.

5. Priestley, M. J. N., and Tao, J. T., “Seismic Response of Pre-cast Prestressed Concrete Frames with Partially DebondedTendons,” PCI JOURNAL, V. 38, No. 1, January-February1993, pp. 58-69.

6. SEAOC Performance Based Seismic Design Ad Hoc Commit-tee, “Tentative Performance Based Seismic EngineeringGuidelines - SEAOC Blue Book, Appendix G,” 1998.

7. Sritharan, S., Igarashi, A., Priestley, M. J. N., and Seible, F.,“Test Design of the PRESSS Five-Story Precast ConcreteBuilding,” Proceedings, 68th Annual Convention, StructuralEngineers Association of California, Santa Barbara, CA, 1999,pp. 255-261.

8. Conley, J., Priestley, M.J.N., and Sritharan, S., “Wall Direc-tion Modeling of the Five-Story PRESSS Precast Test Build-ing,” Report SSRP 99/19, University of California, San Diego,CA, 1999.

9. Pampanin, S., Priestley, M. J. N., and Sritharan, S., “Frame Di-rection Modeling of the Five-Story PRESSS Precast TestBuilding,” Report SSRP 99/20, University of California, SanDiego, CA, 1999.

10. Igarashi, A., Seible, F., Hegemier, G., and Priestley, M. J. N.,“The U.S.-TCCMAR Full-Scale Five-Story Masonry ResearchBuilding Test, Part III - Seismic Load Simulation,” ReportSSRP-94/03, University of California, San Diego, CA.

11. Paulay, T., and Priestley, M. J. N., Seismic Design of Rein-forced Concrete and Masonry Buildings, John Wiley and Sons,New York, NY, 1992, 786 pp.

12. Carr, A. J., “Ruaumoko Program for Inelastic Dynamic Analy-sis — Users Manual,” Department of Civil Engineering, Uni-versity of Canterbury, Christchurch, New Zealand, 1998.

13. Takeda, T., Sozen, M., and Nielsen, N., “Reinforced ConcreteResponse to Simulated Earthquakes,” Journal Structures Divi-sion, ASCE, V. 96, No. 12, December 1970.

REFERENCES

Preliminary Results and Conclusions From the PRESSS Five ...

Documents

Transcript of Preliminary Results and Conclusions From the PRESSS Five ...