journal of constructional steel research -...

Journal of Constructional Steel Research 102 (2014) 204–216

Contents lists available at ScienceDirect

Journal of Constructional Steel Research

Seismic performance of prefabricated steel beam-to-column connections

Fangxin Hu a, Gang Shi a,⁎, Yu Bai b, Yongjiu Shi a

a Key Laboratory of Civil Engineering Safety and Durability of China Education Ministry, Department of Civil Engineering, Tsinghua University, Beijing 100084, P.R. Chinab Department of Civil Engineering, Monash University, Melbourne, VIC 3800, Australia

⁎ Corresponding author. Tel.: +86 10 6279 7420; fax:E-mail address: [email protected] (G. Shi).

http://dx.doi.org/10.1016/j.jcsr.2014.07.0120143-974X/© 2014 Elsevier Ltd. All rights reserved.

a b s t r a c t
a r t i c l e i n f o
Article history:Received 12 April 2014Accepted 12 July 2014Available online 1 August 2014

Keywords:Beam-to-column connectionBolted end-plate connectionNovel cover-plate connectionQuasi-static experimentSeismic performance

Three types of prefabricated steel beam-to-column connection different from common welded unreinforcedflange-boltedweb (WUF-B) connections are examined in this paper. Full-scale specimenswith specific joint con-figurations were prepared, so that the effects of joint detail on the failure mode, ultimate capacity and ductilitycould be identified. They were tested under cyclic loading to further investigate their seismic performance.Experimental results showed that the measured moment capacities of these connections at the face of a columnflange reached 120% to 140% of the beam's full plastic moment. The maximum plastic rotations of all connectiontypes were greater than 0.025 rad and the cumulative plastic rotations were 20 to 30 times themaximum plasticrotation. In this way, connection Types I and III were capable of accommodating a story drift angle of 0.04 rad.These results demonstrated excellent joint strength, and showed potential opportunities for connection TypesI and III to be used in special moment frames (SMFs) in AISC Seismic Provisions, and for connection Type II tobe used in intermediatemoment frames (IMFs). Finite-element analysiswas performed and showed good agree-ment with the experimental results. It was found that rigid end-plate assumption is not appropriate for connec-tion Type I, and higher likelihood of fracturewas detected in the cover plates than in beam flanges for connectionTypes II and III.

+86 10 6278 8623.

© 2014 Elsevier Ltd. All rights reserved.

1. Introduction

Steelmoment frames are currently widely used inmulti-story build-ings as lateral force resisting systems due to their superior ductility andenergy dissipation capacity. The use of steel moment frames for seismicdesign dates back to the 1960s to the 1970s when welding began to beadopted. A few buildings at that time were constructed with unrein-forced flange-welded web (WUF-W) connections. However, the prac-tice quickly evolved to unreinforced flange-bolted web (WUF-B)connections because of greater economy and also because of the satis-factory energy dissipation capacity of WUF-B connections under cyclicloading, as evidenced by research [1,2]. Generally, seismic provisions re-quire that energy dissipation should take placemainly at the beamends,and eventually at the base section of the columns. As a result, these con-nections should possess sufficient strength and rotational stiffness topermit the development of yielding and strain hardening in the zonesthat dissipate energy (such as the beam end, panel zone, and base sec-tion of column) before their final fracture. In the 1994 Northridge andthe 1995 Kobe earthquakes, however, unexpected brittle failures wereobserved between the weld metal and base metal at the toe of theweld access holes within steel moment frames [3–5]. After those earth-quakes, experimental and analytical studies were conducted to investi-gate the causes of such failures and to provide reliable and economical

solutions to avoid them. The efforts have led to improvements ofANSI/AISC 341 [6] and development of new seismic design criteria forbeam-to-column connections, such as FEMA-355D [7].

Improvements in traditional WUF-B connections were achieved intwo main ways. One was to reduce welding defects and the associatedstress concentration in the vicinity of weld access holes through im-proved welding techniques or modification of local details of connec-tions [8–10]. The other was to move the location of the plastic hingeaway from column-to-beam flange groove welds by allowing yieldingof the corresponding beam. This concept was achieved by reinforcingthe beam flanges as in haunched, cover-plate and flange-plate connec-tions [11], or by weakening the beam flanges as in reduced-beam-section (RBS) connections [12]. Research outcomes have shown thatthese methods are effective in providing satisfactory performance ofconnections in steel moment frames and that brittle failure of connec-tions may be avoided.

Connections developed previously were oftenmade on-site. Investi-gated in this paper are three types of connection extracted from the pro-totype frame as shown three-dimensionally in Fig. 1(a); subassembliescontaining those connections are prefabricated in the shop and then as-sembled on-site into the frame. As illustrated in Fig. 1(b), the connectionType I is similar to the traditional WUF-W connection in that the beamflanges (BFX) are connected to the column flange (CF) through com-plete joint penetration welds. The differences are that 1) the beamweb (BWX) is stiffened with a shear plate (SP) and angle (AG) oneach side of connection Type I whereas only one side of the WUF-W is

http://crossmark.crossref.org/dialog/?doi=10.1016/j.jcsr.2014.07.012&domain=pdf

http://dx.doi.org/10.1016/j.jcsr.2014.07.012

mailto:[email protected]

http://dx.doi.org/10.1016/j.jcsr.2014.07.012

http://www.sciencedirect.com/science/journal/0143974X

(b)

CO

CW CF

CF

BWX

BFX

BFX

RIB

EP

SP

BOLT

DP

EP

RIB

BFY

BFY CF

AG

CP

BFY

BWY

BFY

BWY

CW CF

(d)

CO

BFY

BFX

CW

SPAG

CP

CF

CF

BFX

BWX

BFY

BWY

(c)

(a)x y

z

Type III

Subassembly #2

Type II

Subassembly #1Subassembly #3 Type I

fillet weld

Fig. 1. Configuration and components of proposed connections: (a) prototype frame; (b) connection Type I; (c) connection Type II; and (d) connection Type III.

Table 1Results of tensile coupon test.

Material fy/MPa εy/% εst/% fu/MPa εu/%

Hot-rolledsection

H400 × 400 × 13 × 21 292.3 0.147 2.013 449.0 20.173L100 × 10 306.4 0.151 1.621 464.7 14.798

Plate in weldedsection

12 mm thick 305.4 0.151 1.895 453.4 16.82216 mm thick 289.4 0.142 1.895 450.4 19.27020 mm thick 273.4 0.134 1.449 439.2 19.10325 mm thick 244.8 0.117 1.499 440.4 19.41330 mm thick 269.9 0.120 1.245 444.1 18.65635 mm thick 349.7 0.168 0.534 541.1 12.727

Note: H400 × 400 × 13 × 21 represents H-shaped sectionwith flangewidth and section-al height equal to 400 mm,web and flange thickness equal to 13 mm and 21 mm respec-tively; L100 × 10 represents equal-leg angle sectionwith leg length and thickness equal to100 mm and 10 mm respectively.

205F. Hu et al. / Journal of Constructional Steel Research 102 (2014) 204–216

connected to the shear plate; and 2) a bolted end-plate connection(EP, BOLT, RIB) lies near the face of the column flange in connectionType I. This end-plate connection is used for assembly. ConnectionType II is shown in Fig. 1(c). The strong-axis (x axis) configuration isthe same as in Type I but without the end-plate connection, and in theweak-axis (y axis) the beam flanges (BFY) are connected to novelcover plates (CP) using fillet welds along the long slot, nose and innersides of cover plates (see Fig. 1(c)). The connection of the beam web(BWY) to the column web (CW) is stiffened with angles (AG) on bothsides. This is quite different from the traditional weak-axis momentconnections in which beam flanges are connected to continuity plateswith complete joint penetration welds and the beam web is connectedto the shear plate with fillet welds [12]. Unlike connection Type II, con-nection Type III is cruciform, as shown in Fig. 1(d). It consists of twobeams attached to the column with the same weak-axis configurationas in connection Type II. These types of joint configuration differ fromcommonly-used connections prequalified in ANSI/AISC 358 [13], andallow prefabrication and on-site installation. Their mechanical perfor-mance has not been well studied previously.

In order to investigate their seismic behavior and failure mechanisms,three full-scale beam-to-column specimens were fabricated using corre-sponding connection types, and examined under cyclic load. The load car-rying capacity and rotational capacity, as well as ductility and energydissipation capacity are discussed in detail. Furthermore, finite-elementanalysis was conducted for comparison with the experimental resultsand to study the local stress distribution and likelihood of fracture of sev-eral key components in the connections. This work aims to provide valu-able experimental evidence and to develop reliablemodeling approachesfor the evaluation and design of these specific types of prefabricatedbeam-to-column connections for engineering applications.

2. Experimental program

2.1. Materials

Hot-rolled sections and several plates of low carbon steel Q235Bwere used to fabricate the specimens. Tensile tests were carried out

using three identical coupon specimens for each kind of section or thick-ness of the plates to obtain averaged results. The results are summarizedin Table 1, where εy is the yield strain, εst is the strain at the end of yield-ing plateau, and fu and εu are the ultimate strength and strain respec-tively. Note that measured yield strength (fy) exceeds the nominalyield strength of 235 MPa.

2.2. Specimens

Three specimenswere preparedusing the corresponding connectiontypes introduced above as shown in Fig. 2. The beam in Specimen I wasconnected to the prefabricated column-tree joint using tenM30 bolts ofclass 10.9 (see Fig. 2(a)), while Specimens II and III were bothprefabricated with beams and column as a whole (see Fig. 2(b) and(c)). The beam components used in all specimens were hot-rolledsections of H400 × 400 × 13 × 21 and the column component was awelded section of H500 × 500 × 16 × 25 in Specimen I and H700 ×500 × 20 × 35 in the other two specimens. The column height was cho-sen as 2430 mm for all specimens, and the resulting distance betweenthe top and bottom boundary constraints was 2210mm. The length be-tween the loading point at the end of the beam and the column axiswas chosen as 2650 mm for Specimen I, 2750 mm for the strong-axis

(a) (b) (c)

z

x y

2430

2430

2430

2650

2450 2450

2450

2750

West BeamSouth Beam

North Beam

South Beam

LC1

LC2

LC2 LC3

LC3

LC2LC1

LC1

Fig. 2. Test specimens: (a) Specimen I; (b) Specimen II; and (c) Specimen III.

206 F. Hu et al. / Journal of Constructional Steel Research 102 (2014) 204–216

connection in Specimen II and 2450mmfor theweak-axis connection inSpecimens II and III as shown in Fig. 2. The column height and beamlengths were chosen to present the loading points at the ends of thebeams as the inflection points of beams and columns in the prototypeframe shown in Fig. 1(a).

2.3. Experimental setup

The loading setup is shown in Fig. 3, where a 5000 kN jack (LC1)wasemployed to apply axial force on the top of column and two 750 kN ac-tuators (LC2 and LC3) were used to apply cyclic load at the ends ofbeams. Specimens I and III were planar joints and were tested in theloading frame colored light-gray (see Fig. 3). To apply load to SpecimenII as a 3D joint, another loading frame colored dark-gray (see Fig. 3) wasset up and two additional steel beamsweremounted to connect the twoloading frames (see Fig. 3(a)) in order to ensure global stiffness and sta-bility. To simulate the pin supports, the top end of the column in eachspecimen was fixed through two short beams (colored red in Fig. 3) torestrain the displacement of the top end of the column; these shortbeams were connected to the loading frame by bolts. The lower end ofthe column was bolted to a thick steel plate (see Fig. 3) which wasfixed to the strong floor. As well, four posts were erected near the actu-ators (LC2 and LC3) to restrain lateral-torsional buckling of the beams,colored green in Fig. 3(a) and (b).

2.4. Loading protocol

The axial load on the top of the column remained constant duringthe test, at 1000 kN for Specimen I, 2000 kN for Specimen II and 1500

North

Two short

beams

Thick

plate

yx

A Asouth beam

west beam

north beam

Ad

dit

ion

al

stee

l b

eam

Ad

dit

ion

al

stee

l b

eam

(a)

reac

tio

n w

all

Fig. 3. Test setup: (a) plan view

kN for Specimen III, which were determined by the effective seismicweight in the prototype frame shown in Fig. 1(a). For each specimen,loads were imposed on the ends of the beam(s) through displacementcontrol in accordance with the basic loading sequence [14] as shownin Fig. 4 and it can be characterized by the yield displacement (Δy)which corresponds to the occurrence of significant yield in each speci-men. The loading began with one cycle each of Δy/3 and 2Δy/3 in anelastic state, followed by three cycles of multiples of the yield displace-ment (i.e., Δy, 2Δy, 3Δy, etc.) until failure or a limitation of the actuatorsor the loading frame. The yield displacement (Δy) was estimatedthrough preliminary finite-element analysis as 25 mm, 20 mm, and20 mm for Specimens I, II, and III respectively. The loading patternbeganwith the axial load (LC1) applied at the top of column for all spec-imens (see Fig. 2). After the axial loading was stabilized, the actuators(LC2 and LC3, or only LC2 for Specimen I) were used to apply asymmet-ric displacements on the beam ends (i.e., one upward and the otherdownward) according to the loading program in Fig. 4. The displace-ment downward (negative of z direction in Fig. 2) was defined as posi-tive in this study.

2.5. Instrumentation

The instrumentation for the specimens consisted of displacementtransducers as shown in Fig. 5 and uniaxial and rosette strain gagesplaced at key locations, as introduced in the following discussion ofstrain distribution. It should be noted for Specimen I, two displacementtransducers (δ8 and δ9 in Fig. 5) were mounted at the top and lowerflange centerlines of the beam to derive the rotation produced by thebolted end-plate connection.

Post Post

ActuatorActuator

Two short beams

south beam

Thick

plate

y

z

Jack

north beam

(LC1) (LC3)(LC2)

(b)

reac

tio

n w

all

strong floor

and (b) cut view (A-A).

Load

ing d

ispla

cem

ent

Loading step

Fig. 4. Loading protocol.


Themeasured vertical displacement at the beam end δ1, as shown inFig. 5, is caused by the rigid bodymotionof the connection (asmeasuredby δ2, δ5), the deformations of the beam itself (as denoted by δ), the col-umn (asmeasured by δ3, δ4), the panel zone (asmeasured by δ6, δ7) andalso, for Specimen I only, the bolted end-plate connection (as measuredby δ8, δ9). Therefore the actual beam end displacement δ applied by theactuators is calculated from Eq. (1):

δ ¼ δ1−δ2−δ5

HL ð1Þ

where H is the distance from pin to pin of the column (i.e. 2210 mm forall the specimens), and L is the distance from the loading point to thecenterline of the column, as shown in Fig. 5. According to the SACSteel Project [15], the total rotation, or story drift angle θ is calculatedby Eq. (2):

θ ¼ δL: ð2Þ

The result reported in Ref. [16] suggested that the rotation of panelzone θpz can be calculated by

θpz ¼δ6−δ7

2

ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffib2pz þ h2pz

q

bpzhpzð3Þ

δ7 δ6

δ1

δ3

δ4

δ2

δ5

H d

L

Displacement transducerStrain gaugeStrain rosette

Pin

support

Pin

support

δ8

δ9

continuity plate

continuity plate

Fig. 5. Layout of displacement transducers.

where bpz and hpz are the dimensions of the rectangular panel zone area,δ6 and δ7 are panel zone deformation as shown in Fig. 5. The rotation θcoldue to column deformation is

θcol ¼δ3−δ4

d−θpz ð4Þ

where d is the distance between continuity plates as shown in Fig. 5.Therefore the beam rotation θbeam can be derived as

θbeam ¼ θ−θcol−θpz ð5Þ

Further, the rotation contributed by the bolted end-plate connectionθbep in Specimen I is calculated by Eq. (6):

θbep ¼ δ8−δ9d

ð6Þ

The plastic rotation of beam-to-column connection θp for all thespecimens, which represents the inelastic story drift, can be calculatedby Eq. (7) [8,9]:

θp ¼ δ−F=Kel

Lð7Þ

where F is the applied load, andKel is themeasured elastic stiffness of thetested specimen determined from F versus δ curve (see Eq. (1) for δ).

3. Experimental results

3.1. Experimental observations and failure modes

There were no significant observations in the elastic stage for all thespecimens. The end plates (see EP in Fig. 1(b)) near the beam flanges(see BFX in Fig. 1(b)) of Specimen I showed slight bending deformationduring the first cycle of 3Δy displacement. In this cycle, the bolts (seeBOLT in Fig. 1(b)) above the top flanges of the beam experienced visiblebending deformation due to combined bending and shear force. Slip be-tween end plates was detected in the third cycle of 3Δy displacementand gaps were observed between bolts and end plates due to large in-elastic deformation. In the third cycle of 4Δy displacement, the boltabove lower flange fractured in tension as shown in Fig. 6 and theother bolts located beside the beam flanges showed significant bendingand necking deformation when taken out.

For Specimen II, local bucklingwas noticed in the lower flange of thesouth beam (see Fig. 2) in the first cycle of 4Δy displacement. In the sec-ond cycle of this displacement, significant global torsional buckling oc-curred as shown in Fig. 7 and local buckling at the web and flange wasobserved in the south beam as shown in Fig. 8. In the first cycle of 5Δy

displacement, cracks in the fillet weldwere identified in a few locations,including the area between the top flange (BFX in Fig. 1(c)) and theweb

bolt fracturedin tension

BFX(top)

BFX(lower)

Specimen I

Fig. 6. Tensile failure of the bolt in Specimen I.

south beam

CP(lower)

CP(top)

localbuckling at

the web

localbuckling atthe flange

Specimen II

Fig. 8. Local buckling in south beam in Specimen II.


(BWX in Fig. 1(c)) of the west beam (see Fig. 2), the angle (AG inFig. 1(c)) and the column flange (CF in Fig. 1(c)), and between theshear plate (SP in Fig. 1(c)) and the column flange (CF in Fig. 1(c)). Inthe third cycle of 5Δy displacement, local web buckling of the westbeam was noticed and the torsion of the south beam became more se-vere. For safety reasons, the test was stopped manually.

In the third cycle of 3Δy displacement of Specimen III, cracks wereobserved due to extensive yielding in the tip of the weld fusion zonebetween the cover plate (CP in Fig. 1(d)) and column flange (CF inFig. 1(d)). In the subsequent loading process, those cracks extendedtowards the beam flanges (BFY in Fig. 1(d)). In the second cycle of 4Δy

displacement, the lower cover plate of the north beam (see Fig. 2) wasfractured in a suddenmoment as shown in Fig. 9, leading to an immedi-ate decrease of the applied load. Local buckling of the lowerflange of thenorth beamwas identified in the third cycle of 5Δy displacement. Withthe completion of the third cycle of 6Δy displacement, the crack in thelower cover plate of the north beam became more significant andcaused severe overall torsion of the beam. The test was thereforestopped for safety reasons. The crack pattern in the cover plates isshown in Fig. 10, indicating that the crack initiated in the tip of the filletweld between the cover plate and column flange, then propagated in-wards along the beam flange in subsequent cycles, and then towardsthe column web center along the fillet weld between the tip of thebeam flange and the cover plate.

3.2. Strain distribution

To assist understanding of the load transfer within the cover plate inthe weak-axis configuration, Fig. 11 presents the axial (y direction)strain distribution along the depth of the beam cross-section of Speci-men II including the cover plate (CP in Fig. 11), normalized to theyield normal strain (εy) calculated using measured material propertiesin Table 1. Although the cover plate was located outside the beamflange, its axial strain remained below that of the beam flange. This re-sult suggested that the “planes remain plane” assumption was notvalid with this cross-section and the cover plate was not effectively uti-lized. A similar result was found for Specimen III, where the beamswerealso connected to the column through cover plates.

Fig. 12 shows the normalized measured axial strains along the axialdirection of the beam measured at the top beam flange (BFX, BFY inFig. 1(b), (c) and (d)), top cover plate (CP in Fig. 1(c) and (d)), andtop continuity plate (CO in Fig. 1(b) and (c)) for each specimen in sev-eral loading steps. It was found for the weak-axis connection of Speci-mens II and III that the maximum axial strains occurred at the topbeam flange just beyond the nose of the cover plate (i.e. measuredfrom strain gages SG1 and SG6, see Fig. 12(c) and (d)). These locationscoincided with those where the local buckling was observed, andbeams developed notable flexural yielding due to the reinforcementby cover plates. For the strong-axis connection of Specimen II, the

south beam

Specimen II

Fig. 7. Global torsional buckling of south beam in Specimen II.

strains in the continuity plate (see SG2 in Fig. 12(b)) were larger thanthose in the flange (see SG3 in Fig. 12(b)) of the west beam; whereasfor Specimen I the strains in the beam flange (see SG3 in Fig. 12(a))near the column flange were always larger than those in the continuityplate (see SG2 in Fig. 12(a)). This indicated that the 3D loading configu-ration in Specimen II changed the strain distribution in the strong-axisconnection andmade continuity platesmore critical than beam flanges,but the beam was still the component in which most of inelasticityoccurred.

Fig. 13 shows the development of measured shear strains in Speci-mens I and II, normalized to the yield shear strain (γy) calculatedusing measured material properties in Table 1. It should be noted thata 12 mm doubler plate (DP in Fig. 1(b)) was added to the column web(CW in Fig. 1(b)) when designing Specimen I to satisfy the panel zonestrength requirements [17], and strain rosettes were attached in themiddle and one corner of both the column web and the doubler plateto identifywhether the shear strains in the columnweb and the doublerplate were comparable (see Fig. 13(a)). The shear strains in the doublerplate (see R1 andR2 in Fig. 13(a)) and the columnweb (see R3 andR4 inFig. 13(a)) in Specimen I were almost equal to the yield strain when thebeam end displacement reached Δy but the former increased rapidly inthe subsequent loading steps. In Specimen II, strain rosettes were at-tached in the middle and one corner of both the column web and col-umn flange (see Fig. 13(b)). The shear strains in the column flange(see R3 and R4 in Fig. 13(b)) retained small values, and the shear strainsin themiddle of the columnweb (see R1 in Fig. 13(b))were substantial-ly larger than those in the corner (see R2 in Fig. 13(b)).

3.3. Hysteretic and skeleton curves

The shape of hysteretic curves of moment (M) versus rotation (θ) isclosely related to the seismic performance of specimen and its energydissipation capacity. As shown in Fig. 14, the abscissa is the rotation

CP(lower)

CF

CW

fractureSpecimen III

weld fusion zone weld fusion zone

Fig. 9. Fracture in the lower cover plate of the north beam in Specimen III.

20cm

4cm

22cm

3cm

x

y

CP

CP CF

CF

BFY BFY

(a)

5cm

30cm

9cm

47cm

x

y

CP

CP CF

CF

BFY BFY

(b)

Fig. 10. Crack pattern in the cover plates in Specimen III: (a) top and (b) bottom.


calculated from Eqs. (1) and (2), representing the story drift angle (θ),and the ordinate is the ratio of bending moment (M) at the columnface to the full plastic momentMp of the connected beam, whereMp iscalculated from the measured material properties in Table 1 as 1052kNm, and M is calculated as the product of the force applied by the ac-tuator and the distance (see L0 in Fig. 5) between the loading pointand the column face. Note that in the weak-axis configuration, ‘columnface’ here refers to the edges of the column flanges.

As shown in Fig. 14(a), the hysteretic performance of Specimen Ishowed a pinching behavior and the curve shape was narrow. Thismay be due to the slip between bolts and endplates, resulting in smallerreloading stiffness than unloading stiffness as well as smaller energydissipation capacity than the other specimens. Specimens II and IIIevidenced a plump shape for the hysteretic curves as shown inFig. 14(b) to (e), as a result of good welding quality. Specimen I failedby brittle fracture, and there was no apparent strength degradation be-fore the fracture (see Fig. 14(a)). Specimen II failed through progressivelocal buckling at the flange and web of the south beam (see Fig. 14(c))rather than excessive deformation or damage of the reinforcing coverplates (see CP in Fig. 8). In Specimen III, failure initiated because of thefracture of the cover plates that occurred in the second cycle of 4Δy dis-placement (see Fig. 9), but it experienced a few subsequent cycles afterthis failure (see Fig. 14(d) and (e)).

The skeleton curves constructed from the hysteretic curves byconnecting the peak point of the third loading cycle for each displace-ment amplitude [18] are also shown in Fig. 14. Table 2 summarizesthemajor experimental results characterized from hysteretic and skele-ton curves for the associated beams in each specimen, including themaximum moment capacity (Mm) and its ratio to Mp, the maximumstory drift angle (θm) in the last cycle prior to failure, the yield momentcapacity (My) and yield rotation (θy), which were identified fromthe end of the linear stage of the skeleton curves, the initial elastic stiff-ness (Kel) calibrated by a least squares method, the ductility coefficient(μ= θm/θy), themaximumplastic rotation (θp) calculated by Eq. (7), thetotal energy dissipation normalized by the product of yieldmoment and

-5 -4 -3 -2 -1 0 1 2 3 4 5-300

-200

-100

0

100

200

300

Alo

ng t

he

sect

ional

hei

ght

(mm

)

Axial strain (εy)

2Δy/3Δy(1)2Δy(1)3Δy(1)4Δy(1)5Δy(1)4

40

CP

BFY

cross-section

south beam

west beam

CP

BFY BFY (bottom)

BFY (top)

CP (top)

BWY

CP (bottom)

BWY

BWY

Δy/3

Fig. 11. Strain distribution of beam cross-section including cover plates in Specimen II.

yield rotation (ΣAi/Myθy), and the ratio of cumulative plastic rotation(Σθp) to θp. The cumulative plastic rotation (Σθp) was calculated bythe method shown in Fig. 15.

Table 2 shows that the maximum moment capacity (Mm) prior tofailure for each specimen reached 120% to 140% of the beam full plasticmoment (Mp), which means all those specimens can be treated as fullstrength connections. This is because of the strain hardening effect ofthe connection components, including the cover plates and beamflanges. Therefore, such overstrength is attributed to both the beamand connection components. In order to propose design method forthose full strength connections, it's necessary to further investigate theflexural overstrength experienced only by the beam, such as the workby D'Aniello et al. [19] and Güneyisi et al. [20].

Table 2 also shows that Specimens I and III both accommodated amaximum story drift angle (θm) larger than 0.04 rad, and their momentcapacities at the face of column flange (M) weremore than 80% ofMp ata story drift angle of 0.04 rad (see Fig. 14(a), (d) and (e)), satisfying thespecifications for special moment frames (SMFs) in ANSI/AISC 341 [6].Specimen II, though not qualifying for this requirement, was capableof accommodating a maximum story drift angle (θm) greater than0.02 rad (see Table 2), satisfying the specifications for intermediatemo-ment frames (IMFs). Table 2 also shows the lower yield moment capac-ity (My) and larger yield rotation (θy) in the south beam of Specimen IIIthan that of Specimen II. Consequently, the 3D loading configurationon Specimen II led to higher elastic stiffness (Kel) in the weak-axisconnection.

According to previous studies [21], the deformation capacity of aconnection specimen is deemed sufficient if its ductility coefficient (μ)reaches 4.0 and its maximum plastic rotation (θp) reaches 0.03 rad.Thus, Specimens I and III demonstrated excellent deformation capaci-ties (see Table 2). Nevertheless, the normalized energy dissipation(ΣAi/Myθy) in Table 2 indicates the unsatisfactory energy dissipation ca-pacity of Specimen I comparedwith that of Specimens II and III. Anotherprimary indicator for evaluating the energy dissipation capacity andsafety under earthquakes is cumulative plastic rotation (Σθp), whichgenerally is 5 to 8 and even 10 times greater than the maximum plasticdeformation (θp) [21]. However, the Σθp/θp ratio as presented in Table 2was even greater than 20 for all specimens, illustrating the excellent de-formation and energy dissipation capacities.

3.4. Contribution of rotation capacity

To illustrate the rotation capacity of each specimen, Fig. 16 showsthe contribution of each component including beam rotation (θbeam)calculated by Eq. (5), column rotation (θcol) by Eq. (4), panel zone rota-tion (θpz) by Eq. (3), and also bolted end-plate connection rotation(θbep) by Eq. (6) for Specimen I only, to the total rotation or story driftangle (θ) in the last cycle of each loading displacement. Beam rotation(θbeam) with its value shown in Fig. 16 was predominant for SpecimensII and III, whereas rotation of the bolted end-plate connection (θbep)in Specimen I contributed the most to the story drift angle (θ)

0

2

4

6

Along the axial direction of the beam

∆y/32∆y/3∆y(1)2∆y(1)3∆y(1)

∆y/32∆y/3∆y(1)2∆y(1)

∆y/32∆y/3

2∆y(1)

∆y(1)∆y(2)∆y(3)

∆y/32∆y/3

4∆y(1)

∆y(1)2∆y(1)3∆y(1)

xy

SG1 SG2 SG3 SG4

Ax

ial

stra

in (

ε y)

SG1 SG2

(a)

SG4SG3

(b)

0

1

2


Axia

l st

rain

(ε y

)

xy

SG1 SG2 SG3

west beam

SG1 SG2 SG3

SG2

-

6

-

4

-

2

0


Ax

ial

stra

in (

ε y)

SG3 SG2 SG1

xy

south beam

(c)

SG3 SG1 SG5 SG4 SG3 SG2 SG1-2

0

2

4

6

8

10

12

Along the axial direction of the beamA

xia

l st

rain

(ε y

)

xy

SG6SG4 SG3

SG1

SG5 SG2

south

beam

north

beam

(d)

SG6

Fig. 12. Strain response along the axial direction of the beams: (a) Specimen I; (b) west beam of Specimen II; (c) south beam of Specimen II; and (d) Specimen III.


(see Fig. 16(a)). The panel zone rotation (θpz) was negligible in theweak-axis configuration of Specimens II and III (see Fig. 16(c) and(d)) compared with the strong-axis configuration of Specimens I andII (see Fig. 16(a) and (b)). This is because the two column flanges con-sidered as the panel zone in the weak-axis configuration had greaterstrength and were much stiffer than the column web, i.e. the panelzone in the strong-axis configuration. However, greater column rotation(θcol) was observed in the weak-axis configuration of Specimens II andIII (see Fig. 16(c) and (d)) than the strong-axis configuration of Speci-mens I and II (see Fig. 16(a) and (b)) due to the much lower weak-axis bending stiffness of the column. Moreover, in Specimen I, the in-crease in story drift angle (θ) after yield displacement was mainly con-tributed by the bolted end-plate connection and panel zone (seeFig. 16(a)). This indicated that the bolted end-plate connection andpanel zone, rather than the beam section, were critical in the evaluationof Specimen I's rotation capacity.

(a)

Fig. 13. Development of shear strains in panel

4. Finite-element analysis

4.1. Modeling approach

FE models were established for all specimens using the ABAQUSprogram [22]. This program is capable of large-deformation nonlinear(including material and geometric nonlinearities) three-dimensionalanalysis. C3D8R solid elements with reduced integration were used inthe modeling. Previous studies have suggested that 13 solid elementsacross the beam flange width and 4 elements through the thicknesswere sufficient for accurate evaluation of stress or strain with consider-ation of inelastic behavior [23]. In this study, 24 elements across thebeam width and 4 elements through the beam or column flange wereadopted. Translations in x, y, and z directions were restrained at bothends of the column to represent the pin support condition in the exper-iments. A fine mesh was used in the regions of beam-to-column

(b)

zone: (a) Specimen I and (b) Specimen II.

(a)

(b) (c)

(d) (e)

Fig. 14.Moment ratio (M/Mp) versus story drift angle (θ): (a) Specimen I; (b)west beamof Specimen II; (c) south beamof Specimen II; (d) north beamof Specimen III; and (e) south beamof Specimen III.


connections. The resulting meshed FE models are shown in Fig. 17 forall the specimens, with mesh details of connection components.Multi-linear true stress-strain relationship of Q235 steel was used inthe FE analysis as shown in Fig. 18(a), with the points defining the

Table 2Summary of test results.

Specimen My

(kNm)θy(%rad)

Mm

(kNm)θm(%rad)

I 848.4 0.94 1293.1 4.45II: west beam 765.1 0.76 1314.7 3.56II: south beam 918.1 0.79 1420.3 3.96III: north beam 739.2 0.96 1326.8 4.96III: south beam 738.5 0.90 1415.0 4.97

stress-strain relationship measured from coupon tests. The materialproperties of class 10.9 high-strength bolts used in this study aregiven in Fig. 18(b) according to a previous experimental study of thesame bolt type [24], and bolt shanks were modeled using the effective

Mm/Mp Kel

(kNm)μ θp

(%rad)ΣAi/(Myθy)

Σθp/θp

1.23 14.9 4.7 3.5 53.2 22.21.25 16.7 4.7 2.6 107.3 24.81.35 19.9 5.0 2.9 114.5 26.11.26 15.5 5.1 3.7 154.0 28.51.35 16.3 5.5 3.7 171.0 28.4

Fig. 15. Definition of cumulative plastic rotation [21].


diameter. The Poisson ratio of steel was taken as 0.3. Material plasticitywas considered by the von Mises yielding criteria and the associatedflow rule with combined hardening.

The contact behavior was taken into account by the modeling of thebolted end-plate connection in Specimen I. The behavior included threepairs of contacts, i.e. between the end plates, between the bolts and theend plate on the left side, and between the bolts and the end plate onthe right side, as shown in Fig. 17(a). For each contact pair, a friction co-efficient of 0.35 for tangential behavior and hard contact for normal be-havior were used. The pre-tension of the bolts in Specimen I was takeninto account in the FE model.

First, the axial load defined in the loading protocol was applied onthe top of the column and remained constant. Then, a vertical displace-ment history as described in Fig. 4 was implemented at the free ends ofthe beams in the modeling, using the displacement-control feature in

(a)

(c)

Fig. 16. Development of components within story drift angle: (a) Specimen I; (b) w

ABAQUS. The corresponding history of the applied load was back-calculated from the reactions at the supports. The lateral deformationof the beamswas restrained at the loading point, considering the lateralsupports provided in the experiments.

4.2. Comparison with experimental results

Comparisons of the load-displacement hysteretic curves from thefree ends of the beams are shown in Fig. 14(a) for Specimen I, inFig. 14(b) and (c) for Specimen II, and in Fig. 14(d) and (e) for SpecimenIII. The initial stiffness obtained from the modeling was in good agree-ment with that from the experiments for all the specimens. Althoughthe pinching phenomenon from the modeling hysteretic curves forSpecimen I was not so significant as the experimental results after cer-tain load steps as shown in Fig. 14(a), the loading capacity comparedreasonably well with the experimental ones especially in negative dis-placements. Such discrepancy between experimental and modelinghysteretic curves for Specimen I might be attributed to the inaccuracyof the stress–strain relationship of the high-strength bolt adopted inthe FEM by referring to the results of previous experimental study.The actual strength of those bolts might be lower than that specifiedin the FEM, in which case in testing the bolts could developmore plasticdeformation as well as residual deformation, resulting in a larger gapbetween the bolts and the end plates than that found in the numericalresults. The presence of fillet welds in the bolted end-plate connectionwas also disregarded, which might affect the performance. Themodeling and experimental hysteretic curves for Specimen II were inrelatively good agreement as shown in Fig. 14(b) and (c); however,the discrepancy for Specimen III as shown in Fig. 14(d) and (e) werenotable andmight be explained as a result of local high residual stressesin the weld fusion zone around the cover plates (see Fig. 9), andconsequently, rapid strain hardening exhibited by the experimentalhysteretic curves. The strength deterioration caused by fracture ofcover plates in the test of Specimen III could not be simulated by the cor-responding FEM either.

(b)

(d)

est beam of Specimen II; (c) south beam of Specimen II; and (d) Specimen III.

(c) Specimen III

fine mesh

z

x y

z

x y

(b) Specimen II

fine meshcover plate

shear plate angle

contact

bolts

end plate

fine mesh

z y

x

(a) Specimen I bolted connection end plate

pretension

bolt

Fig. 17. Finite-element models: (a) Specimen I; (b) Specimen II; and (c) Specimen III.


5. Discussion

5.1. Tensile forces in bolts during loading

In addition to the global load and displacement responses, the FEanalysis can provide local mechanical responses at bolted positions,where instrumentation becomes difficult. Fig. 19 shows the history oftensile forces in the bolts (#1–#3) on the left side (positive x direction)of the beam web centerline in Specimen I. Bolt #3 experienced almostconstant tensile force close to the specified initial pre-tension force(i.e. 355 kN) applied in the FEM, due to its location in the neutral axisof the beam. Bolts #1 and #2 above the neutral axis experienced nearlythe same history of tensile force, whichwas in conflict with the conven-tional designmethod that assumes rigid end-plate and thus a linear dis-tribution of axial forces along the height of the section induced by themoment in bolts as shown in Fig. 20. The same result was found forbolts #4 and #5. This suggests that rigid end-plate assumption is not ap-propriate for Specimen I, and the new proposal of a distribution of axialforces in those bolts induced by the moment, as shown in Fig. 20, ismore reasonable.

(a)

Fig. 18. Stress–strain curves: (a) Q235

5.2. Damage indices

Several characterizations ofmechanical responses have been used inprevious studies to identify critical locations of failure (brittle or duc-tile): the pressure index,Mises index, equivalent plastic strain index, tri-axiality index (TI), and rupture index (RI) [25]. In this paper the lattertwo were used for evaluation. The TI is defined as the hydrostatic stress(σm) divided by the Mises stress (σ),

TI ¼ σm

σð8Þ

TI values between−0.75 and−1.5 can cause large reductions in therupture strain of metals and values less than −1.5 can initiate brittlefailure [25].

The RI is defined in Eq. (9):

RI ¼ aPEEQεr

ð9Þ

(b)

steel and (b) high-strength bolt.

Fig. 19. History of bolt tensile forces in Specimen I.

Fig. 20. Distribution of axial forces in bolts.


where a is a material constant, PEEQ is the equivalent plastic straindefined in ABAQUS, εr is the rupture strain and is defined in Eq. (10):

εr ¼ a exp 1:5σm

σ

� �: ð10Þ

The RI can be used to compare the likelihood of fracture in criticalregions [25].

For the beamof Specimen I and thewest beamof Specimen II, TI andRIwere computed along four paths, as shown in Fig. 21(a), of which twopaths (A and C)were located at the interface between the beam and thecolumn flange and two paths (B andD)were at themid-thickness of thebeam flange below theweld access hole. Note that taking into consider-ation the modeling constraints, the two paths (A and C) at the interfacewere not located at mid-thickness but at one-fourth of the thickness ofthe beam flange, where crackingwasmore likely to initiate [23]. For thesouth beam of Specimens II and III, TI and RI were computed along two

(a)

Fig. 21. Location of paths to compute the indices: (a) stron

paths as shown in Fig. 21(b), one (E) at the nose of the cover plate andthe other (F) at the edge of the column flanges.

In Specimen I, the minimum value of TI was recorded a short dis-tance from the center of the lower flange at the interface (seeFig. 22(a)), andwas still larger than−0.7. In other words, brittle failurein the beam flanges was unlikely to occur. The maximum value of RIwas recorded in the center of the lower flange at the interface (seeFig. 22(b)), indicating that it was likely to fracture in this location first.In Specimen II, the distribution of RI in the west beam (see Fig. 22(d))was similar to that in Specimen I. However, the minimum value of TIin the west beam was recorded about 100 mm from the center of thetop flange at the interface (see Fig. 22(c)). The value reached −0.8,which could cause a large reduction in the rupture strain and conse-quently in the ductility of the steel. This difference between the westbeam of Specimen II and Specimen I resulted from the 3D loading con-figuration on Specimen II. Theminimumvalue of TI in the south beamofSpecimen II, whichwas about−0.6 (see Fig. 22(e)), was recorded in thebeam flange at the nose of the cover plate, while themaximum value ofRI, whichwas about 3.5 (see Fig. 22(f)), was recorded in the cover plateat the edge of the column flange. A distribution of TI and RI in the southbeam of Specimen III similar to that in Specimen II was observed(see Fig. 22(g) and (h)). It should be noted that in the south beam ofSpecimen III the maximum value of RI was about 4.5 (see Fig. 22(h)).This demonstrated a much higher likelihood of fracture initiating inthe cover plate at the edge of the column flange in Specimen III thanin Specimen II, which to some extent explained the different failuremodes (i.e. fracture in the cover plates for Specimen III but local buck-ling in the south beam for Specimen II, see Figs. 8 and 9) observed intesting.

6. Concluding remarks

Three full-size prefabricated steel beam-to-column connectionswere tested to investigate their cyclic behavior and evaluate seismicperformance. Specific joint configurations were utilized in these con-nections, which were quite different from the commonly used WUF-Bconnections or other specific connection forms prequalified in ANSI/AISC 358. The following conclusions can be drawn from the experimen-tal study and the FE analysis conducted in this work.

(1) Comparedwith Specimens I and III, the 3D loading configurationin Specimen II changed the strain distribution in the strong-axisconnection and made continuity plates more critical than beamflanges. This resulted in a different failure mode in the weak-axis connection, i.e. local buckling failure in Specimen II ratherthan the fracture failure that occurred in Specimens I and III. Itis apparent, therefore, that the effects of the 3D loading configu-ration on the seismic performance of connections need to beinvestigated.

(2) The maximum moment capacities of the specimens with differ-ent connection configurations at the face of the column flangereached 120% to 140% of the beam's full plastic moment. Suchoverstrength exhibited by the whole connection indicates a

(b)

g-axis configuration and (b) weak-axis configuration.

0 200 400 6000

1

2

3

4

5

RI

Distance along beam flange (mm)

path Epath F

south

beam

max value

F E

(h)

0 200 400 600-1.00

-0.75

-0.50

-0.25

0.00

0.25

TI


path Epath F

south

beam

min value

EF

(g)

0 200 400 600-1.00

-0.75

-0.50

-0.25

0.00

0.25

TI


path Epath F

south

beam

min value

EF

(e)

0 200 400 6000.0

0.5

1.0

1.5

2.0

2.5

3.0

3.5

RI


path Epath F

south

beam

max value

F E

(f)

0 200 4000.0

0.2

0.4

0.6

0.8

1.0

1.2

1.4

1.6

1.8

RI


path Apath Bpath Cpath D

west

beam

max value AB

CD

(d)

0 200 400-0.9

-0.8

-0.7

-0.6

-0.5

-0.4

-0.3

-0.2

-0.1

0.0

TI



west

beam

min value

AB

CD

(c)

0 200 400-0.7

-0.6

-0.5

-0.4

-0.3

-0.2

-0.1

0.0

TI



min value

AB

CD

(a)

0 200 4000.0

0.2

0.4

0.6

0.8

1.0

1.2

RI



max valueA B

C D

(b)

Fig. 22.Distribution of TI and RI in the third cycle of 4Δy displacement: (a), (b) Specimen I; (c), (d)west beamof Specimen II; (e), (f) south beam of Specimen II; and (g), (h) south beam ofSpecimen III.



necessary further investigation on flexural overstrength of thebeam alone for the sake of the capacity design between thebeam and the connection. Specimens I and III were capable of ac-commodating a story drift angle greater than 0.04 rad, and Spec-imen II accommodated a story drift angle greater than 0.035 rad.These results demonstrated the excellent joint strength of theproposed connection configurations, and indicated potential op-portunities for connection Types I and III to be used for specialmoment frames (SMFs) in ANSI/AISC 341, and for connectionType II to be used for intermediatemoment frames (IMFs). How-ever, since the test in this paper using a single specimen for eachtype of connectionmaygive only limited information,more com-plete and reliable qualifications should be conducted in future tostudy the effect of different design variables such as the height ofbeams on the seismic behavior of those connections.

(3) The maximum plastic component of story drift angle reached0.035 rad for Specimen I and III and 0.025 rad for Specimen II.The cumulative plastic rotation was 20 to 30 times greater thanthe maximum plastic rotation, and the ductility coefficient wasgreater than 4.0 for all specimens. Normalized energy dissipationin Specimens II and III was about two to three times higher thanthat in Specimen I. This result demonstrated the good deforma-tion capacity of Specimens I and III and the excellent energy dis-sipation capacity of Specimens II and III.

(4) The bolted end-plate connection in Specimen I played the role ofmoving the location of the plastic hinge away from the column-to-beam flange groove welds. However application of this con-nection configuration in earthquake-resistant steel momentframes should be undertaken with care, considering its limitedenergy dissipation and brittle failure observed in the experiment.Prefabricated all-welded connections as in Specimens II and IIIare recommended for seismic design by virtue of their satisfacto-ry energy dissipation and ductility.

(5) Beam rotation predominantly contributed to the total story driftangle in Specimens II and III, whereas the rotation of bolted end-plate connection in Specimen I was the most significant contrib-utor in the total story drift angle. Rotation of the panel zone wasnegligible in the weak-axis configuration in Specimens II and IIIbecause of the higher stiffness of their panel zone than that inthe strong-axis configuration in Specimens I and II.

(6) The states of stress and strain in the weak-axis configuration inSpecimens II and III reinforced with cover plates were complex.A higher likelihood of fracture initiation in the cover plates ratherthan in the beam flanges was determined in that configuration.Further detailed examination of the mechanical performance ofthe connection components is needed, especially the loading

capacity and failure mechanism of the components in theweak-axis configuration.

References

[1] Popov EP, Pinkney RB. Cyclic yield reversal in steel building connections. J Struct Div1969;95(3):327–53.

[2] Bertero VV, Popov EP, Krawinkler H. Beam-column subassemblages under repeatedloading. J Struct Div 1972;98(5):1137–59.

[3] Miller DK. Lessons learned from the Northridge earthquake. Eng Struct 1998;20(4–6):249–60.

[4] Popov EP, Yang TS, Chang SP. Design of steel MRF connections before and after 1994Northridge earthquake. Eng Struct 1998;20(12):1030–8.

[5] Nakashima M, Inoue K, Tada M. Classification of damage to steel buildings observedin the 1995 Hyogoken-Nanbu earthquake. Eng Struct 1998;20(4–6):271–81.

[6] ANSI/AISC 341-10. Seismic provisions for structural steel buildings. Chicago, IL: AISC;2010.

[7] FEMA-355D. State of the art report on connection performance. Washington, DC:FEMA; 2000.

[8] Stojadinovic B, Goel SC, Lee KH, Margarian AG, Choi JH. Parametric tests on unrein-forced steel moment connections. J Struct Eng 2000;126(1):40–9.

[9] Ricles JM, Mao CS, Lu LW, Fisher JW. Inelastic cyclic testing of welded unreinforcedmoment connections. J Struct Eng 2002;128(4):429–40.

[10] Han SW, Kwon GU, Moon KH. Cyclic behaviour of post-Northridge WUF-B connec-tions. J Constr Steel Res 2007;63(3):365–74.

[11] Kim T, Whittaker AS, Gilani ASJ, Bertero VV, Takhirov SM. Experimental evaluationof plate-reinforced steel moment-resisting connections. J Struct Eng 2002;128(4):483–91.

[12] Gilton CS, Uang CM. Cyclic response and design recommendations of weak-axis re-duced beam section moment connections. J Struct Eng 2002;128(4):452–63.

[13] ANSI/AISC 358-10. Prequalified connections for special and intermediate steel mo-ment frames for seismic applications. Chicago, IL: AISC; 2010.

[14] JGJ 101–96. Specificating of testing methods for earthquake resistant building. Bei-jing: China Architecture & Building Press; 1997 [in Chinese].

[15] SAC/BD-97/02. Protocol for fabrication, inspection, testing, and documentation ofbeam-column connection tests and other experimental specimens. Sacramento, Cal-ifornia: SAC Joint Venture; 1997.

[16] Shi G, Fan H, Bai Y, Yuan F, Shi YJ, Wang YQ. Improved measure of beam-to-columnjoint rotation in steel frames. J Constr Steel Res 2012;70(3):298–307.

[17] GB50017–2003. Code for design of steel structures. Beijing: China Planning Press;2003 [in Chinese].

[18] Suita K, Nakashima M, Morisako K. Tests of welded beam-column subassemblies II:detailed behavior. J Struct Eng 1998;124(11):1245–52.

[19] D'Aniello M, Landolfo R, Piluso V, Rizzano G. Ultimate behaviour of steel beamsunder non-uniform bending. J Constr Steel Res 2012;78:144–58.

[20] Güneyisi EM, D'Aniello M, Landolfo R, Mermerdas K. A novel formulation of the flex-ural overstrength factor for steel beams. J Constr Steel Res 2013;90:60–71.

[21] NakashimaM, Suita K, Morisako K, Maruoka Y. Tests of welded beam-column subas-semblies I: global behavior. J Struct Eng 1998;124(11):1236–44.

[22] ABAQUS. Analysis user's manual, vol. I-V. Providence, RI: Dassault Systèmes SimuliaCorp.; 2010 [Version 6.10].

[23] SAC/BD-98/01. Strength and ductility of FR welded-bolted connections. Sacramento,California: SAC Joint Venture; 1998.

[24] Shi G, Shi YJ, Wang YQ, Bradford MA. Numerical simulation of steel pretensionedbolted end-plate connections of different types and details. Eng Struct 2008;30(10):2677–86.

[25] Kim T, Whittaker AS, Gilani ASJ, Bertero VV, Takhirov SM. Cover-plate and flange-plate steel moment-resisting connections. J Struct Eng 2002;128(4):474–82.

http://refhub.elsevier.com/S0143-974X(14)00211-9/rf0005





















































journal of constructional steel research -...

Documents