Contents lists available at ScienceDirect

Thin-Walled Structures

journal homepage: www.elsevier.com/locate/tws

Full length article

Performance of flange-welded/web-bolted steel I-beam to hollow tubularcolumn connections under seismic load

Qian-Yi Songa,b,⁎,1, Amin Heidarpourb, Xiao-Ling Zhaob, Lin-Hai Hanc

a China Southwest Architectural Design and Research Institute Corp. LTD, Chengdu, Sichuan Province 610093, Chinab Department of Civil Engineering, Building 60, Monash University, VIC 3800, Australiac Department of Civil Engineering, Tsinghua University, 100084, China

A R T I C L E I N F O

Keywords:Flange-welded/web-bolted steel connectionStatic loadingCyclic loadingDamage levelExperimentFinite element modelling

A B S T R A C T

Earthquake causes wide and severe damage to building structures due to the great ground motion which is alsoone of inducements of other extreme events such as fire, tsunami and debris flow. The Multi-extreme loads suchas earthquake and the subsequent actions could be much more destructive than the earthquake action itself.Regarding the safety of properties and lives, in order to evaluate building structures subjected to post-earthquakeactions, it is important to accurately assess the partial damage caused by earthquake on structural components.By using experimental method, two groups of flange-welded/web-bolted steel I-beam-to-hollow-tubular-columnconnections were investigated under static and cyclic loadings at ambient temperature. The flexibility andstrength of the connections are measured, while the damage phenomena and failure modes are explored duringthe tests. The connection damage was found to be a cumulative process with fracture extending. The increasingdamage led to significant gradual degradation of the connection mechanical properties. The quantificationalevaluation was also carried out to investigate the connection damage degree according to different loadingintensities. Using finite element (FE) modelling, study was carried out and validated by the experimental resultswhile a good agreement was achieved. The test and FE modelling results in this study provided detailedunderstanding of the performance of the flange-welded/web-bolted steel tubular connection. It can be used forfurther investigation of the connections subjected to combined actions of extreme loading such as post-earthquake fire.

1. Introduction

Great ground motion in earthquake often causes wide and hugedamages to building structures. Earthquake would affect the residualproperties of construction materials which might be subjected to post-earthquake extreme events scenarios (e.g. fire) [1,2]. Additionally forstructural components, various seismic induced damage patterns withdifferent damage levels ranging from no damage to collapse have beenobserved and reported [3–5]. The structures are prone to be damagedby the low probability but high consequences earthquake eventsespecially in places of low to moderate seismic regions [6]. Thepartially damaged structures, which do not collapse, are much vulner-able under the possible secondary actions (e.g. post-earthquake fire[7–10] and aftershocks [11]). Considering potential extreme scenarioscombining earthquake and subsequent actions, investigating damagelevels and deterioration of load-carrying capacity of steel structuresafter seismic loading are necessary.

For isolated connections, increasing cyclic loading procedure isusually used to simulate the seismic actions in experimental andnumerical investigations. The steel beam-to-column frame connectionsare required to have sufficient ductility and avoid brittle failure underseismic loading. Damage assessment is critical to evaluate the cumula-tive damage states of steel structures for the possible post-earthquakeextreme events (e.g. fire, tsunami and debris flow). A general acceptedstructural failure criteria and quantified damage index are necessary todescribe damage degree of structures. These were summarized byCastiglioni and Pucinotti [12] for various structural components.

Although steel or steel-concrete connections are various in engi-neering practices [13,14], simple steel I-beam-to-hollow-tubular-col-umn connections are commonly used in low- to mid-rise buildings inareas of low seismic activity. Two types of simple steel I-beam-to-hollow-tubular-column connections [15,16] were investigated pre-viously. However, compared to the bolted connections and simplewelded connections investigated before, the flange-welded/web-bolted

http://dx.doi.org/10.1016/j.tws.2017.03.011Received 15 April 2016; Accepted 8 March 2017

⁎ Corresponding author.

1 Postdoc in Tsinghua University.E-mail address: soujy.qy@gmail.com (Q.-Y. Song).

Thin-Walled Structures 116 (2017) 250–264

Available online 06 April 20170263-8231/ © 2017 Elsevier Ltd. All rights reserved.

MARK

connections were enhanced in the beam-column joint area andsupposed to have better seismic performance. Especially for the possiblepost-earthquake actions, the bolted connections were not even aconsideration due to their low rotation stiffness. Seismic loadingperformance of flange-welded/web-bolted steel I-beam-to-hollow-tub-ular-column connections were studied through experimental andnumerical investigation in this study. Connections were fabricatedand tested under quasi-static and cyclic loadings at room temperature.The earthquake induced damage of the adopted connections wasassessed on the basis of the experimental test data. The experimentaland numerical investigation presented in this paper provided bench-mark data for further investigation of the connections under post-earthquake events.

2. Experimental tests

2.1. Specimens and fabrication

In total 4 connections were fabricated at the Monash CivilEngineering Workshop. The flange-welded/web-bolted connectionswere made of I-beams and tubular columns through welds, double-

angle and bolts, as well as the configurations of the simple weldedconnection and the bolt connection investigated previously [15,16] forcomparison, as shown in Fig. 1. The flanges of the I-beam were directlywelded on the column face through single-V welds. Two steel angleswere bolted to the I-beam web and welded on the column face throughfillet welds. All the I-beams were 200UB22.3 of Grade 250 structuralmild steel. Square tubular columns with cross-section of200 mm×200 mm, made of Grade 350 structural mild steel, wereadopted. Two column wall thicknesses, of 5 mm and 9 mm, were usedin this study. The cross-section of the adopted steel angle was65 mm×65 mm×6 mm whilst 8.8M10 bolts were utilised. Addition-ally, plates with nominal thickness of 8 mm were welded on the top andbottom of the columns, and two stiffeners with nominal thickness of5 mm were welded to the I-beam, 1 m away from the column panelsurface where the load was applied during testing. The geometricalproperties of the fabricated columns and I-beams are listed in Table 1.

The 4 connections were divided into two groups according to thecolumn wall thickness. In each group, one of the two connections wastested under monotonic loading until failure while another one, withthe same configurations, was tested under cyclic loading. The speci-mens were labelled in terms of the column wall thickness and loading

Nomenclature

b A parameterb0 Width of chord or width of column of connectionsb1 Width of brace or width of beam of connectionsbe Effective plate widthbp Width of transverse platesBc Width of column cross-sectionBf Flange width of I-beamEi Energy dissipated by a single loading cycleEs Initial elasticity Modulus of steelEtotal The total energy assumed to be dissipated until failurefu Steel ultimate strengthfy Steel yield stressfy0 Yield strength of the chord steel in connectionsfyp Yield strength of transverse platesFy Yield strength of connectionh1 Height of brace or beamH Flange height of I-beamK Connection stiffness used for joint classificationld Damage indexMf Moment strength of connection at failureMpl

beam The design plastic moment resistance of a beamMR Moment strength of connectionMu Connection ultimate moment strengthMu, + or - Ultimate moment strength of connection in positive or

negative loading directionMu,c Calculated ultimate connection strengthMu,t Ultimate connection strength in testMy Connection yield moment strengthMy, + or - Yield moment strength of connection in positive or

negative loading directionMy,t Connection yield strength from in testMy, cc The calculated yield strength of connection from column

panel plastificationMy,cf The calculated yield strength of connection from I-beam

flangesMy

−Average yield moment strength in the positive andnegative directions

NR Transverse load capacityS1,t Initial stiffness of connectionS2,t Elongation stiffness of connection

Sj,ini Initial stiffness of connectionsSs,i The Secant stiffness of each loading cyclet0 Wall thickness of chord or columntc Thickness of column walltf Beam flange thicknesstp Thickness of single-V weldstw Design thickness of the welds or web thickness of I-beamZx The plastic section modulus of steel I-beamβ Width ratio, b1/b0βi The energy ratio dissipated by a single loading cycleδi The imposed displacement amplitude of each loading

cycleδu,mon The ultimate displacement of connection under monotonic

loadingδy Yield displacementεy Steel yield strainθ Joint or connection rotationθf Connection rotation at failureθf, + or - Connection rotation at failure in positive or negative

loading directionθi The imposed rotation amplitude of each loading cycleθmax The maximum connection rotationθmax, + or - The maximum connection rotation in positive or

negative loading directionθu Ultimate connection rotation corresponding to Mu

θu, + or - Ultimate connection rotation in positive or negative load-ing direction

θy Yield connection rotation corresponding to My

θy, + or - Yield connection rotation in positive or negative loadingdirection

θy−

Average yield connection rotation of the positive andnegative directions

μ Ductility coefficientμ+ or - Ductility coefficient from the positive or negative loading

directionμ Average ductility coefficient of the positive and negative

directionsμsi Ratio of imposed displacement amplitude of each cycle δi

to the yield displacement δyνs Poisson's ratio of steel

Q.-Y. Song et al. Thin-Walled Structures 116 (2017) 250–264

251

scenario; for example CCH5S and CCH9S represent the flange-welded/web-bolted connections subjected to quasi-static loading, while CCH5Cand CCH9C represent connections subjected to cyclic loading. Thenumbers 5 and 9 in the abovementioned specimen labels represent thecolumn wall thickness. It is noting that the welding metal was AWSER70S-6 wires while the average size of the fillet welds between thecolumn panel and the legs of double-angle was measured as tw=6 mm.

The design thickness of the single-V welds was taken as tw=5.5 mm.

2.2. Setup and loading protocol

The loading setup and boundary conditions for both the monotonicand cyclic loading tests are similar as shown in Fig. 2a. The bottom ofthe columns was fixed on a heavy foundation plate through welding,

Fig. 1. Configuration details of steel I-beam to hollow square column connections (Unit: mm): a) unstiffened welded connection; b) Double-angle bolted shear connection; c) flange-welded/web-bolted connection.

Q.-Y. Song et al. Thin-Walled Structures 116 (2017) 250–264

252

while all degrees of freedom of the column top end were restrainedexcept in the axial direction. The boundary conditions of the columnwere carefully chosen so that the effect of the overall deformation of thecolumn was minimised. A constant axial load of 50 kN was exerted onthe top of the column during testing. For the connection with thecolumn wall thicknesses of 5 mm and 9 mm, the axial compressionratios were only 0.037 and 0.017, respectively, so that the exerted axialload did not affect the connection performance. The cantilever I-beamwas extended to a reaction frame where an Instron actuator with250 kN tension and compression loading capacity hanged on. Theactuator was connected to the I-beam through a uniform joint andused to exert displacement controlled loads. It is worth noting thatcompressing the beam towards ground (downward direction) was takenas positive loading direction. The cyclic loading tests were carried outaccording to the AISC-341 [17] cyclic loading protocol, as shown inFig. 2b. The tests were terminated when a significant and sudden loaddrop of about 30% occurred or the Instron actuator achieved its travelor load capacity.

2.3. Measurements and gauges

In order to investigate the connection behaviour, various dataincluding moment vs. rotation relationship, deformation of differentcomponents and strain values were measured using LVDTs and straingauges attached on the specimens. The arrangement of LVDTs andstrain gauges are shown in Figs. 3a and b, respectively. The gaugearrangement was the same for both the monotonic and cyclic loadingtests, except that the LVDT 4# was removed in the cyclic loading tests.ARAMIS Photogrammetry system was used to monitor the typicalfailure at the joint area.

2.4. Material properties

Standard tensile coupons, according to AS1391 [18], were cut fromthe flat area of each part including the I-beam flange &web, tubes,angles and other plates. The bolts were also tested for the yield andultimate strengths. The gauge width of all the plate coupons was13 mm. The material properties, such as yield strength fy, ultimatestrength fu, initial modulus of elasticity Es, yield strain εy and Poisson'sratio νs, were obtained from uniaxial tensile tests. The magnitudes ofthese parameters are listed in Table 1. Steel materials of all thecomponents were ordinary low-carbon steel and have obvious yield

platform except the tubular columns and bolts where 0.2% offsetstrength was reported as the yield strength. The ultimate strength ofthe welds was conservatively taken as 480 MPa [19].

3. Experimental results

3.1. Monotonic test results

The joint rotation of the connections was calculated from themeasurements of the LVDTs. For monotonic loading tests, the momentvs. rotation relationships are shown in Fig. 4a. The results, includingfailure phenomena, initial stiffness S1,t, elongation stiffness S2,t, yieldmoment strength My, yield rotation θy, ultimate moment strength Mu

and the corresponding ultimate rotation θu, obtained from the mono-tonic loading tests are listed in Table 2. Fig. 4b depicts how abovementioned parameters are determined. The damage phenomena of theconnections under monotonic loading are shown in Figs. 4c and d.Punching shear fracture was obtained on the column panel of theconnection CCH5S at the tensile side while tearing fracture wasobserved at the I-beam tensile flange of the connection CCH9S. Duringthe tests, obvious deformation of the column panel and the lateralcolumn walls was measured as shown in Fig. 5. It is worth noting thatthe direction of column wall movement towards the outside isconsidered positive whereas movement towards the inside is considerednegative. The pull-out deformation indicated a severe plasticity flow ofthe column panel region. Among all these phenomena, the discontin-uous fracture phenomena were considered as the most critical damagewhich would affect the connection performances significantly.

With the column wall thickness increased from 5 mm (CCH5S) to9 mm (CCH9S), the ultimate moment strength increased from38.04 kN m to 48.52 kN m and the failure mode changed from punch-ing shear fracture of column panel to tearing fracture of tensile I-beamflange. The initial stiffness of the connection CCH9S was also higherthan that of the connection CCH5S. According to definitions given byEC3-1–8 [20], a typical connection can be classified as pinned joints,rigid joints or semi-rigid joints which depends on the magnitude of theinitial stiffness (Fig. 6a); meanwhile, connections can also be classifiedas full strength, partial strength or pinned connections by using theultimate strength (Fig. 6b). It is note that both CCH5S and CCH9S areclassified as semi-rigid connections by the stiffness criterion as shown inTable 3. Both connections are also classified as partial strengthconnection according to strength criterion as shown in Table 3.

Table 1Geometrical and material properties of the connections.

Geometrical properties (unit: mm)

No. Column Beam Bf/Bc Angle Bolt

Bc tc Bc/tc Bf×H tf tw

CCH5S 200×200 5.07 39.45 131.6×203.0 6.64 5.14 0.658 65×65×5.96 G8.8M10CCH9S 8.75 22.86CCH5C 5.07 39.45CCH9C 8.75 22.86

Material properties

Components Column Beam-web Beam-flange Angle Bolt

Thickness (mm) 5.07 8.75 5.14 6.64 5.96 —fy (MPa) 401.43 421.63 361.85 337.97 340.95 1042.77fu (MPa) 472.02 452.95 507.87 496.92 488.24 1164.05Es (MPa) 20,0433 20,1267 19,8250 19,8750 20,2900 23,5392εy (με) 4006.7 4007.2 1825.2 1700.5 1680.4 4429.9νs 0.293 0.3

Note: In the specimen number, the meaning of characters is as follows: C-connection, C-combined type (flange-welded/web-bolted), H-hollow section box column, S-static loading, C-cyclic loading.

Q.-Y. Song et al. Thin-Walled Structures 116 (2017) 250–264

253

Additionally, both connections satisfy the plastic rotation capacity of atleast 0.03 rad which is recommended by FEMA interim guideline [21].

3.2. Cyclic loading test results

The moment vs. rotation relationships resulted from cyclic loadingtests are shown in Fig. 7. The skeleton curve, connecting all the peakpoints of each loading cycle, reflects the strength and ductilityperformance of connections under cyclic loading, as shown in Fig. 8a.The measured yield moment strength My, ultimate moment strength Mu

and related rotations (i.e. θy, θu; in which θ represents joint rotation)and the maximum rotation amplitude θmax achieved in positive andnegative directions are shown in Table 2. Fig. 8b depicts how theseparameters are determined. The total loading cycles, failure modes andthe loading cycle number at which initial failure occurred are reportedin Table 4. The ductility coefficient μ (= θf/θy) in the positive andnegative directions and the average value μ are also reported in

Table 4. The final failure of the connections was defined at the pointwhere the load (the failure strength Mf) corresponding to the rotationamplitude of a cycle decreased to less than 85% of the ultimate strength(Mu) [22]. Due to the travel capacity of the loading apparatus, thefailure load Mf might be still higher than 0.85Fu at the end of the test,and thus it was conservatively assumed that the failure rotation θf isequal to θmax. Additionally, in the cyclic loading tests, the measuredpulling out deformations of the column panel and the lateral bucklingdisplacements of the column walls were hysteretic curves but smallerthan the corresponding values resulted from the monotonic loadingtests.

The connection CCH5C was subjected to 43 cycles in total. Similarto the monotonic loading test, the initial fracture was observed afterabout 25 loading cycles on the face of the column panel around the I-beam flange tips. The cracks extended through the column wallthickness quickly and then further extended along the edges of beamflanges and angle outstand legs with the increase of the cyclic loading.

Fig. 2. Setup and the cyclic loading protocol: a) Setups; b) Cyclic loading protocol (AISC 2005).

Q.-Y. Song et al. Thin-Walled Structures 116 (2017) 250–264

254

The final fracture failure mode is shown in Fig. 9a. Due to the cyclicloading, the punching shear fractures of the column panel wereobserved symmetrically on both sides of the I-beam. The differencebetween the connections in this paper and the simple welded connec-tions investigated by authors before [16] is that fracture on the columnwall propagated not only along outside edges of the beam flangestowards the centre of the column panel but also along the edges of theangle outstand legs. The test of CCH5C ceased when the maximumrotation amplitude achieved 0.110 rad in the positive direction. Themaximum positive moment strength was 17.72 kN m during the 35thloading cycle at the rotation of 0.070 rad. The maximum negativemoment strength was −16.36 kN m during the 34th loading cycle atthe rotation of −0.060 rad. The load dropped gradually after the peakpoint until the end of the test when load in both positive and negativedirections dropped to below 85% of the corresponding ultimatestrengths. The calculated average ductility coefficient μ was 5.21 asshown in Table 4. The failure mode of CCH5C was the punching shearfracture on the column panel around the cross-section formed by theedges of the I-beam flanges and angle outstand legs.

The connection CCH9C was stronger than CCH5C under cyclicloading. It was subjected to 46 cycles in total. As shown in Fig. 9b, theinitial fracture was observed on edges of the I-beam flange after about26 cycles. Then the fracture extended towards the centre of the beamflanges with the increase of the cyclic loading. The fracture finallyextended fully through cross-section of the I-beam flanges. Whiledifferent with the simple welded connections [16], the punching shearcracks were observed on the centre of the column panel on both sides ofthe I-beam discontinuous to the tearing fractures of the I-beam flanges.The maximum positive moment strength was 39.20 kN m during the

29th loading cycle at the rotation of 0.039 rad. The maximum negativemoment strength was −39.48 kN m during the 39th loading cycle atthe rotation amplitude of −0.039 rad. The load was found decreasinggradually after the peak point until the final failure. The test of CCH9Cceased with a load drop over 15% at the 46th loading cycle in thepositive direction. The calculated average ductility coefficient μ was4.76 conservatively as shown in Table 4. The failure mode of CCH9Cwas the tearing fractures of I-beam flanges with the partial punchingshear fracture of column panel on both sides of the I-beam.

Other characteristics such as the energy dissipation, the secantstiffness degradation, the residual deformation also reflect the connec-tion performances. Energy dissipation can be used to indicate thecapacity of connections subjected to cyclic/seismic loading while thesecant stiffness degradation after each loading cycle can indicate thebehaviour deterioration of the connections [16]. The energy dissipationof each loading cycle increased with the increase of cyclic loading asshown Fig. 10a. However, the secant stiffness after each loading cycledecreases within the increase of cyclic loading as shown in Fig. 10b.Moreover, the residual rotation increased with the increase of loadingcycles. The values associated with the load of 0 kN after each halfloading cycle are shown in Fig. 10c.

4. Mechanical analysis

4.1. Analysis of test results

For the flange-welded/web-bolted steel I-beam-to-hollow-tubular-column connections investigated in this study, the experimental testsindicated that the failure mode is dependent on the configuration

Fig. 3. Arrangement gauges: a) LVDTs; b) Strain gauges.

Q.-Y. Song et al. Thin-Walled Structures 116 (2017) 250–264

255

details and loading procedures. CCH5S and CCH5C failed due to thepunching shear fracture of the column panel without showing thetearing fracture of the tensile I-beam flanges because the I-beam flangeswere relatively stronger than the column panel. When the thickness ofthe tubular column wall increased from 5 mm to 9 mm, CCH9S and

CCH9C failed mainly due to the tearing fracture of I-beam flanges.Partial punching shear fracture at the centre of the column panel wasobserved for CCH9C connection at the end of testing. It was seen fromthe ultimate moment strength Mu and the related rotation θu that thecolumn wall thickness contributed to the strength and stiffness of the

Fig. 4. Performance of welded connections under static loading: a) Moment vs. rotation relationships; b) Parameter definitions; c) Failure phenomenon of CCH5S; d) Failure phenomenonof CCH9S.

Table 2Experimental results obtained from monotonic and cyclic loading tests.

Monotonic loading test results

Specimens S1,t (kN m/rad) My (kN m) S2,t (kN·m /rad) θy (rad) Mu (kN m) θu (rad) Failure

CCH5S 1008.9 15.19 112.6 0.0313 38.04 0.2324 Column wall shear fractureCCH9S 2390.6 36.44 107.1 0.0411 48.52 0.1333 Beam flange tearing fracture

Cyclic loading test results

Specimens My,+ (kN m) θy,+ (rad) My,- (kN m) θy,- (rad) Mu,+ (kN m) θu,+ (rad) Mu,- (kN m) θu,- (rad) θmax,+ (rad) θmax,- (rad)

CCH5C 11.97 0.0191 −11.73 −0.0197 17.72 0.0702 −16.36 −0.0602 0.1107 −0.1101CCH9C 27.52 0.0196 −27.18 −0.0198 39.20 0.0393 −39.48 −0.0390 0.1107 −0.1101

Q.-Y. Song et al. Thin-Walled Structures 116 (2017) 250–264

256

connections. The connections with thinner column wall were moreflexible.

Connection damage caused by the cyclic loading was more severethan those caused by monotonic loading, which led to lower connectionstrength. For a group of loading cycles with the same travellingamplitude, damage caused by the first loading cycle was observedmore severe than the following ones. Under cyclic loading, theincreasing energy dissipation greatly contributed to the damage ofthe welded connections. The dissipated energy was not apparent in thefirst 22 loading cycles because the loading amplitudes exerted on theconnections were very small so that cracks had not occurred yet and the

Fig. 5. Deformation of the column walls: a) Pulling out deformation of the column panel of the two connections; b) Deformation of the lateral column walls of CCH5S; c) Deformation ofthe lateral column walls of CCH9S.

Fig. 6. Classification of connections: a) Stiffness criterion; b) Strength criterion.

Table 3Classification of the connections according to monotonic loading tests.

Specimens Sj,ini (kN m/rad) By Stiffness Mu (kN m) By strength

CCH5S 1008.9 Semi-rigid 38.04 Partial strengthCCH9S 2390.6 Semi-rigid 48.52 Partial strength

Note: The material properties used in the calculation of beam section characteristics arethose of the beam flange when the properties of the beam flange differ from those of thebeam web.K=1.90×103 kN m/rad, 25 K=47.5×103 kN m/rad, 0.5 K=0.95×103 kN m/rad;Mpl

beam=fy·Zx=70.30 kN m, 0.25 Mplbeam=17.57 kN m.

Q.-Y. Song et al. Thin-Walled Structures 116 (2017) 250–264

257

materials were still in elastic. It implies that the damage was relativelyignorable in the first 22 cycles with elastic behaviour. The energydissipation became apparent in the following loading cycles. The secantstiffness Ss,i of each loading cycle decreased significantly during theloading (Fig. 10b) especially after 20 loading cycles. Thus due to theinelastic behaviour caused by the material plasticity, as well as crackpropagation, the connections showed inelastic behaviour. The increas-ing residual rotation (Fig. 10a) also reflected the deterioration of theconnection performance.

During both the monotonic and cyclic loading tests, the local strainvalues of the I-beam flanges and the column panel were also measured.As shown in Fig. 11 as an example, values from the strain gauges1#~4# on the column panel of the connections CCH5S and CCH9Sunder monotonic loading showed quite small. The strain measurementsfrom the centre of the I-beam flanges were even smaller than the shownvalues from the column panel. The strains measured from the cyclicloading tests were relatively smaller than the monotonic loading testresults. Regarding to the strain/stress states together with the failurephenomena, uneven stress distribution indicates that the fracturestarted from areas around the beam flange tips where the stressconcentrated.

It can be concluded that fracture is the main damage phenomenonof the tested connections under both monotonic and cyclic loadings.The fractures always started at the joint regions around the I-beamflange tips; however they might occur at the weakest and intensivelystressed components among the column panel, welds or beam flanges.

4.2. Ultimate strength and failure mode

Failure at connections under cyclic loading occurred at much lowerloads compared with those under monotonic loading due to thecumulative damage process. Several possible damage modes for theflange-welded/web-bolted steel connections such as column panelpunching shear fracture, the column panel plastification, tearingfracture of I-beam flange, lateral buckling of column wall, I-beamflange buckling, and failure at welds or bolts may occur. A givenconnection may fail due to one of these modes or shown multipledamage modes at ultimate states such that the weakest component isdamaged first and leads to the failure of structures. The connectionstrength under monotonic loading can be predicted according to thefailure mode. The fracture at welds would occur when the loads exceedthe weld strength. Welds are required to be carefully checked by visualinspection and non-destructive examination [23]. Calculation accord-ing to AS4100 [24] showed that weld strength of the tested connectionswas significantly larger than the strengths corresponding to otherfailure modes for CCH5S but close to the ultimate strength for CCH9S.

The flange-welded/web-bolted steel connections investigated in thisstudy can be considered as simple welded connections [16] strength-ened by double-angle with which the joint region between the I-beamflanges becomes more rigid. For CCH5S connection, the chord faceplasticity and punching shear fracture are considered as the character-istic modes from which to calculate the corresponding yielding andultimate moment-resisting strengths. It is mentioned in AISC-360 [25]that the limit states of column sidewall under tension and compression

Fig. 7. Moment vs. rotation relationships under cyclic loading: a) CCH5C; b) CCH9C.

Fig. 8. Skeleton curves and cyclic loading performance parameters: a) Skeleton curves; b) Determination of parameters in cyclic loading tests.

Q.-Y. Song et al. Thin-Walled Structures 116 (2017) 250–264

258

do not need to be checked except the beam member and the columnmember have the same width. Considering the rigid region enclosed bythe edges of the I-beam flanges and the double-angle legs, the designmoment resistance according to the chord face failure (β≤0.85) ofwelded joints between RHS brace members and RHS chords is adoptedhere [20],

⎛⎝⎜⎜

⎞⎠⎟⎟M f t h

b t βb t

β= 1

2 /+ 2

1 −+ /

1 −g y0 02

0 0

0 0

(1)

in which fy0 is the yield strength of the chord steel; t0 is the chordthickness; h1 is height of the brace; β(=b1/b0) is the width ratiobetween brace and chord; b1 and b0 are the width of the brace andchord, respectively. The axial load effect is ignored in the calculation

because the axial load ratio of the chord is quite small in the test.Similar to the calculation of simple welded steel connections [16],

the connection strength due to punching shear fracture of the columnpanel and the tearing fracture of the I-beam flanges can be calculatedaccording to transverse plate connections. When the transverse loadcapacity NR of a single plate is calculated, the connection moment canbe obtained from [20].

M N h t= ( − )R R f1 (2)

where h1 is the depth of the I or H beam and tf is the correspondingflange thickness.

From the test of CCH5S, punching shear fracture of the columnpanel was observed not only through the whole cross-section along the

Table 4Failure and ductility characteristics of the connections under cyclic loading.

Observed failure characteristics

Specimen Totalcycles

Failure phenomena Cycle of initialfailure

CCH5C 43 Column panel punchingshear fracture

25

CCH9C 39 Beam flange tearingfracture

26

Ductility characteristic

Specimens θf,+ (rad) θf,- (rad) μ+ μ- μ My−

(kN m) θy−

(rad)

CCH5C 0.1030 −0.1019 5.25 5.17 5.21 11.85 0.0194CCH9C 0.0900 −0.0975 4.59 > 4.92 > 4.76 27.35 0.0197

Fig. 9. Failure phenomena of welded connections under cyclic loading: a) CCH5C; b) CCH9C.

Q.-Y. Song et al. Thin-Walled Structures 116 (2017) 250–264

259

tensile I-beam flange but also extending to the cross-sections along theedges of the legs outstand of the double-angle. Simply, considering thecolumn panel cross-section along the I-beam flange fully effective forpunching shear only, the punching shear strength of a transverse platecan be calculated by

N f t b= 0.6R y0 0 1 (3)

When calculating the ultimate moment-resisting strength, the fy0 inEq. (3) is replaced by fu0 which is the ultimate strength of the material.

The case of CCH9S is relatively complex. Unlike the simple weldedconnections with fillet welds, single-V welds were adopted between theI-beam flanges and the column panel. The design thickness of the weldswas taken as 5.5 mm. From the failure mode (Fig. 4d), it is clear that thefracture was close to the welds such that the damage and failure weresignificantly affected by the strength of the welds. After a series ofcalculation and analysis, the yielding of the tensile I-beam flange (the

single-V weld cross-section) was found to be close to the yieldingstrength of the connection while the ultimate strength of the welds wasclose to the maximum-resisting strength of the connection. The localyielding of a transverse plate is obtained from [26].

N f t b=R yp p e (4)

where fyp is the yield strength of transverse plates but replaced by theyield strength of the welds. tp is replaced by the thickness of the single-V welds. The effective plate width be is taken as

bb t

f tf t

b b= 10/

but ≤ey

yp pp p

0 0

0 0

(5)

The predicted yield and ultimate moment-resisting strength of theconnections CCH5S and CCH9S are shown in Table 5, which agree wellwith the test results.

Fig. 10. Performance characteristics of the connections: a) Energy dissipation; b) Secant stiffness degradation; c) Residual rotation.

Fig. 11. The lateral displacement of the column and strain status: a) Lateral displacement of CCH5S; b) Strain Status of CCH9S.

Q.-Y. Song et al. Thin-Walled Structures 116 (2017) 250–264

260

4.3. Damage analysis

The rotation limitation corresponding to the collapse prevention forwelded unreinforced flanges-welded web (WUF-W) connections is givenas 0.064 rad [21]. The rotation capacity of the tested connectionssatisfies this limit without loss of load-carrying capacity under bothmonotonic and cyclic loadings. The test results showed that theconnection damage was accumulative during the loading process fromthe initial damage to the final failure particularly under cyclic loadingscenario. The damage caused by each loading cycle led to furtherdeterioration in the strength, stiffness and energy dissipation capacityof the connections. The damage induced by cyclic loading can beconsidered as the results of material plasticity and fracture.

The assessment of structural members after earthquake was on thebasis of the partially damaged, but still surviving, structures. A damageindex which is normalized by a failure point states to the range between0 and 1 is supposed to evaluate the seismic loading induced damagedegree. The value of 0 refers to no damage, while the value of 1 refers toa fully damaged case in which structures cannot resist any more loads[15,16].

The kinematic ductility ratio μsi (=δi/δy), which is the ratio of theimposed displacement amplitude δi of each cycle to the yield displace-ment δy, is an important parameter to evaluate the connection damagecontributed by the exceeding travel amplitude in several damageindexes [26,27]. In the calculation of μsi in this paper, the displace-ments δi and δy were replaced by the joint rotation values θi and θy. Asstructures are assumed to dissipate a certain amount of energy beforefailure, another widely used damage indicator is the normalized energydissipation ratio such that

∑ ∑D β EE

= =cei

N

ii

Ni

total (6)

in which, Ei represents the energy dissipated by a single cycle and Etotalrepresents the total energy that is assumed to be dissipated until failureoccurs. In this study Etotal is measured from the static loading tests until

Table 5Theoretical static strength of the connections.

Connections Calculation Test

My,cf My,cc Mu,c My,t Mu,t My,c/My,t Mu,c/Mu,t Failure mode(kN·m)

CCH5S – 14.41 36.91 15.19 38.04 0.949 0.970 Chord punching shearCCH9S 35.71 – 50.72 36.44 48.52 0.980 1.050 Beam flange tearing

Note: My,t and Mu,t are the yield strength and ultimate strength of the connections during tests, respectively. My, cf is the calculated yield strength from I-beam flanges while My, cc is thatfrom column panel plastification.

Fig. 12. Damage assessment: a) CCH5C; b) CCH9C.

Table 6Comparison ratios of the connection performances.

Specimens Mu (kN m) My (kN m) S1,t (kN m/rad) δmax μ

CCH5S 1.00 1.00 1.00 1.00 1.00CWH1S 0.68 0.64 0.53 0.90 1.08CBH1S 0.45 — 0.10 0.94 —CCH9S 1.00 1.00 1.00 1.00 1.00CWH2S 1.33 1.06 0.90 1.26 1.47CBH2S 0.29 — 0.08 1.42 —CCH5C 1.00 1.00 — 1.00 1.00CWH1C 0.75 0.73 — 1.44 1.01CBH1C 0.49 0.49 — 1.57 0.47CCH9C 1.00 1.00 — 1.00 1.00CWH2C 1.02 1.05 — 1.27 1.02CBH2C 0.38 0.26 — 1.90 0.54

Note: The comparison ratios of the connections subjected to the cyclic loading test arefrom the positive values. The maximum travel displacement ratio of the connectionsunder cyclic loading is the ratio of the maximum travel amplitude of the loading loops.

Fig. 13. Comparison of the total energy dissipation of the connections under cyclicloading.

Q.-Y. Song et al. Thin-Walled Structures 116 (2017) 250–264

261

failure. βi represents the energy ratio dissipated by a single cycle.There are several damage indexes, which were proposed in previous

researches, available to evaluate the damage degree of structures undercyclic/seismic loading. The damage assessment of the connectionsCCH5C and CCH9C is depicted in Fig. 12 which is based on the indexessummarized by Castiglioni and Pucinotti [12]. The shown damageindex ld in Fig. 12 was the generalized value by the monotonic loading

test data. The index proposed by Newmark and Rosenblueth obviouslyunderestimated the damage degree of the connection CCH5C butoverestimated the damage of the connection CCH9C. Most of otherdamage indexes overestimated the damage degree of the testedconnections. The Park-Ang's index [28] is found to reasonably describethe damage level of the welded connections, which is expressed by

Fig. 14. Finite element modelling: a) Finite element model; b) The von Mises stress distribution on tensile beam flange: CCH9S); c) The von Mises stress distribution on column wall:CCH9S.

Q.-Y. Song et al. Thin-Walled Structures 116 (2017) 250–264

262

∑D δδ

b EF δ

= max( ) +mlsi

u mon i

Ni

y u mon, =1 , (7)

where b is a parameter related to the contribution of the energydissipation to the damage. Although b=0.025 was recommended forsteel components [29] and it was found predicting the damage degreeof simple welded connections well [16], it is definite that a constantvalue of b cannot be used for all cases. In this study, the value ofb=0.05 is adopted after several trials. The δu,mon represents theultimate displacement of the same specimen under monotonic loading.Fy is the connection yield strength. The ld of Park-Ang's index shown inFig. 12 is normalized by using the static loading test results.

As shown in Fig. 12, the evaluation value of the Park-Ang's indexwas very close to those observed during the tests. It predicted thedamage index of the connection CCH5C as a value of 0.97 at the end ofthe last cycle and that of CCH9C as 1.05, when load already dropped toaround or below 85% of the corresponding ultimate strength. Bothconnections did not experience obvious damage before the initialfractures around the 25th cycle. The damage gradually increased fromthe occurrence of initial cracks until final failure.

4.4. Comparison of connection performance

Although there is a little difference in the material properties of thesteel columns of the connections, as a series of investigation, theperformance of the flange-welded/web-bolted connections was com-pared with the simple welded and bolted connections presentedpreviously in this section. The comparison ratios of connectionstrengths, initial stiffness, the maximum travel displacement at failureand ductility ratio are listed in Table 6. It can be seen that for theconnections of 5 mm column wall thickness, the ultimate strength, yieldstrength and initial stiffness of the flange-welded/web-bolted connec-tions are obviously higher than those of the simple welded and boltedconnections under both the monotonic and cyclic loading tests. Whilefor the connections of 9 mm column wall thickness, the ultimatestrength, yield strength and initial stiffness of the flange-welded/web-bolted connections are close to those of the simple welded connectionsbut except that the ultimate strength of CCH9S is lower than that ofCWH2S. The strengths and initial stiffness of the bolted connections aremuch lower. All the ductility ratio of the flange-welded connections isclose to that of the simple welded connections except that the ductilityratio of CCH9S is apparently smaller than that of CWH2S. The ductilityratio of the bolted connections under cyclic loading is only about 50%of the other corresponding connections. Moreover, for the maximumtravel displacement, except the CCH5S under monotonic loading isclose to the corresponding simple welded and bolted connections, allother flange-welded/web-bolted connections are smaller than the otherconnections under both monotonic and cyclic loading.

For the cyclic loading tests, the total energy dissipation of theconnections is compared with the flange-welded/web-bolted connec-tion CCH5C in Fig. 13. It can be seen that the connections with thickercolumn wall have better energy dissipation capacity than those withthinner column wall. The energy dissipation capacity of the boltedconnections is the lowest. CCH5C has bigger energy dissipationcapacity than CWH1C, while CCH9C has less energy dissipation

capacity than CWH2C because that the maximum travel displacementof CCH9C is smaller than that of CWH2C during the loading tests.

Overall, the flange-welded/web-bolted has better performanceespecially under seismic loading. The double angle used to connectthe beam web with the column face enhanced the joint stiffness and thecolumn face plastic deformation, while it reduced the ultimate defor-mation ability of the joint.

5. Finite element modelling

5.1. Modelling details

Finite element modelling was also carried out to predict thebehaviour of the connections CCH5S, CCH9S, CCH5C and CCH9Cunder monotonic and cyclic loadings using ABAQUS [30]. For carbonmild steel under monotonic loading, an elasto-plastic stress-strainrelationship model consisting of five stages with the isotropic hardeningcriterion was used, while a nonlinear combined isotropic/kinematicmodel with a von Mises yield surface and an associated plastic flow rulewere applied for steel material under cyclic loading [31]. The measuredmaterial properties at ambient temperature were from tensile materialtests as listed in Table 1.

Geometrical dimensions of the single-V welds between the I-beamand column panel were considered in the model. The 3D hexahedronelement (C3D8R) was adopted for all components. Mesh analysis wascarried out to make sure the modelling is converged. “Tie” was used tosimulate the welding assemblages among different parts. “Hard con-tact” and “Coulomb friction” with a coefficient of 0.25–0.6 were usedamong the surfaces of the steel angles, bolts, column panel and I-beamweb. The boundary conditions were set the same as those in tests.Details of the developed finite element model as well as the boundaryconditions are shown in Fig. 14a.

5.2. Verification and analysis

The moment vs. rotation relationships of CCH5S and CCH9S undermonotonic loading are shown and compared with test results in Fig. 4a.The punching shear fracture on the column panel of CCH5S occurred atearly stage during the test so that the agreement between the FEmodelling and the test data is only good before the fracture appearance.As fracture modelling is not covered in this stage, the FE curves in thefigure are cut at somewhere when they exceed the ultimate points of thetest curves. The FE curves reveal the plastic development of theconnection materials under loading while the gap between the FEcurves and the corresponding test curves indicate the release of theconnection load bearing capacity caused by material fracture. Differ-ences increase when the joint rotation exceeds about 0.090 rad. Theresults given by FE modelling of CCH9S agree well with test data of theelasto-plastic behaviour; however, differences between the numericaland experimental results increase when the tearing fracture failureoccurred on the tensile I-beam flange from the joint rotation of about0.085 rad. The ultimate connection strength of the FE models is takenaccording to the ultimate rotations measured in the tests. The compar-ison of the ultimate strength Mu, the yield strength My and the initialloading stiffness S1,t are listed in Table 7, which shows good agreementbetween the modelling and the experiment results, except that theultimate strength Mu of CCH5S obtained from EF modelling which isobviously higher than that of obtained from the test due to punchingshear fracture explained above.

The moment vs. rotation relationships of CCH5C and CCH9C undercyclic loading are shown in Fig. 7. A good agreement between theresults of FE modelling and experiment results is achieved at the initialloading cycles before fracture occurs. The differences between numer-ical and experimental results are obvious particularly when pinch effectbecomes apparent from tests mainly because of fracture damagephenomena.

Table 7Comparison between FE modeling and experimental results under monotonic loading.

Specimens Parameters Test FEM FEM/Test Rotation (rad)

CCH5S Mu (kN) 38.04 47.85 1.26 0.232My (kN) 15.19 10.69 0.71 ——S1,t (kN/m) 1008.9 907.5 0.90 ——

CCH9S Mu (kN) 48.52 53.32 1.10 0.133My (kN) 36.44 33.03 0.91 ——S1,t (kN/m) 2390.6 2291.6 0.96 ——

Q.-Y. Song et al. Thin-Walled Structures 116 (2017) 250–264

263

However, it is shown by the FE modelling results that the stressdistributes unevenly at the joint area of the connections under bothmonotonic and cyclic loading scenarios. For an example, the stressdevelopment on the beam flange and the column panel of CCH9S areshown in Fig. 14b and c. When the load increases, the area of the I-beam flange and the column panel around the flange tips yields first.Thus yielding extends to the adjacent areas until the entire cross-sectionis fully yield. Although steel is a kind of ductile material, the fracturewould occur at the most stress concentrated areas (e.g. the I-beamflange tips and the column panel regions around the I-beam flange tips).

In conclusion, the FE modelling can describe the elasto-plasticbehaviour of the connections well, but the ductile fracture phenomenoncannot be reflected by the modelling with conventional materialconstitutive model. In order to predicate ductile fracture phenomenon,material damage models, which can be used to predicate the elementfailure due to fractures, should be investigated and considered in the FEmodel.

6. Conclusions

An experimental investigation of the flange-welded/web-boltedsteel I-beam-to-hollow-tubular-column connections was presented inthis paper. Two sets of steel connections were tested under monotonicand cyclic loadings at ambient temperature. During the tests, themoment vs. rotation relationships and strengths were measured. Thedamage phenomena were observed and analysed. The quantifieddamage analysis was carried out on the basis of the test data.Additionally, the FE modelling with conventional material constitutivemodels was carried out and compared to the test data. The followingconclusions can be made:

• Several possible failure modes of the flange-welded/web-boltedsteel connections could occur depending on configuration detailsand loading procedures. Failure always occurs at the criticalcomponent which is the weakest among all components. For thetested connections in this paper the weakest component is thecolumn panel for the connections CCH5S and CCH5C and the tensileI-beam flanges for the connections CCH9S and CCH9C. Fracture isfound to be the main damage phenomenon of the tested connectionsunder monotonic and cyclic loading.

• The tested connections have adequate rotation capacity and goodductility. The observed fracture damage is a cumulative propagatingprocess particularly under cyclic loading. Different damage degree isinduced by different loading intensity, which would lead to thedegradation of the mechanical behaviour of the connections underpossible secondary actions such as fire.

• The ductility ratio and the energy dissipation ratio are considered asthe main damage indicators. After comparison, the Park-Ang's indexwas found to predict the damage degree of the flange-welded/web-bolted steel I-beam-to-hollow-tubular-column connections well. Theparameter b of the Park-Ang's index considering the contribution ofthe energy dissipation is a variable for different structures andshould be determined according to test results carefully.

• Finite element modelling was also carried out to investigate theelasto-plastic behaviour of the tested connections under monotonicand cyclic loadings. The modelling results show uneven stressdistribution around the joint regions. The severe stress concentra-tion occurred at the areas of the I-beam flange and column panelaround the beam flange tips where fractures were initially observed.Ductile fracture model needs to be incorporated into the materialmodel for FE analysis in order to make a better agreement betweennumerical and experimental results.

Acknowledgement

The research work presented in this paper was supported by

Australian Research Council through a Discovery Project(DP1096454) awarded to the second author. The authors would liketo express gratitude to the laboratory manager Mr Long Goh whoprepared materials and upgraded equipment for the experimental tests.The authors thank the lab technician Mr Andres Cortes.

References

[1] S. Sinaie, A. Heidarpour, X.L. Zhao, Mechanical properties of cyclically-damagedstructural mild steel at elevated temperatures, Constr. Build. Mater. 52 (2014)465–472.

[2] Q.Y. Song, Behaviour of Steel I-beam to Box Column Connections under Post-earthquake Fire, Monash University, Australia, 2015.

[3] B. Kato, A. Tanaka, H. Yamanouchi A field work investigation of steel buildingdamage due to the 1978 miyagiken-oki earthquake, in: Proceedings of the SeventhWorld Conference on Earthquake Engineering, Istanbul, Turkey, 1980, pp.479–486.

[4] N.F. Youssef, D. Bonowitz, J.L. Gross, A Survey of Steel Moment-resisting FrameBuildings Affected by the 1994 Northridge Earthquake, US National Institute ofStandards and Technology, Gaithersburg, Maryland, 1995.

[5] R. Tremblay, M. Bruneau, N. Nakashima, H.G.L. Prion, A. Filiatrault, R. DeVall,Seismic design of steel buildings: lessons from the 1995 Hyogo-ken Nanbu earth-quake, Can. J. Civ. Eng. 23 (1996) 727–756.

[6] J.L. Wilson, N.T.K. Lam, L. Pham, Development of the new australian earthquakeloading standard, Electron. J. Struct. Eng. (2008) 25–31.

[7] J. Mohammadi, S. Alyasin, D. Bak, Analysis of post-earthquake fire hazard, in:Proceedings of the 10th World Conference on Earthquake Engineering, 1992, pp.5983–5988.

[8] S. Chen, G.C. Lee, M. Shinozuka Hazard mitigation for earthquake and subsequentfire, in: 2004 ANCER Annual Meeting: Networking of Young EarthquakeEngineering Researchers and Professionals. Honolulu, Hawaii, 2004, pp. 29–30.

[9] C. Scawthorn, Analysis of Fire following Earthquake Potential for San Francisco,California, 2010.

[10] Q.Y. Song, A. Heidarpour, X.L. Zhao, L.H. Han, Post-earthquake fire behaviour ofwelded steel I-beam to hollow column connections: an experimental investigation,Thin-Walled Struct. 98 (2016) 143–153.

[11] N. Luco, C.A. Cornell, G.L. Yeo, Annual limit-state frequencies for partially-inspected earthquake-damaged buildings, Struct. Saf. 24 (2002) 281–296.

[12] C.A. Castiglioni, R. Pucinotti, Failure criteria and cumulative damage models forsteel components under cyclic loading, J. Constr. Steel Res. 65 (2009) 751–765.

[13] Y. Chen, Y. Wu, Behaviour of square hollow section brace-H-shaped steel chord T-joints under axial compression, Thin-Walled Struct. 89 (2015) 192–204.

[14] J. Yang, T. Sheehan, X.H. Dai, D. Lam, Experimental study of beam to concrete-filled elliptical steel tubular column connections, Thin-Walled Struct. 95 (2015)16–23.

[15] Q.Y. Song, A. Heidarpour, X.L. Zhao, L.H. Han, Performance of double-angle boltedsteel I-beam to hollow square column connections under static and cyclic loadings,Int. J. Struct. Stab. Dyn. (2015) 16.

[16] Q.Y. Song, A. Heidarpour, X.L. Zhao, L.H. Han, Performance of unstiffened weldedsteel I-beam to hollow tubular column connections under seismic loading, Int. J.Struct. Stab. Dyn. (2014) 15.

[17] A. ANSI, AISC 341: Seismic Provisions for Structural Steel Buildings, AmericanInstitute of Steel Construction, American Institute of Steel Construction Inc.,Chicago, IL, 2005.

[18] AS. AS1391-2007, Metallic Mateials–tensile Testing at Ambient Temperature.Sydney, Australia, 2007.

[19] AWS, Specification for Carbon Steel Electrodes and Rods for Gas Shielded ArcWelding, American Welding Society, Miami, 2005.

[20] CEN, Eurocode 3, Design of Steel Structures, Part 1-8. Design of Joints, EuropeanCommittee for Standardization, Brussels, 2005.

[21] S.J. Venture, Recommended Seismic Design Criteria for New Steel Moment-FrameBuildings, Federal Emergency Management Agency, Sacramento, California, 2000.

[22] L.H. Han, W. Li, Seismic performance of cfst column to steel beam joint with RCslab: experiments, J. Constr. Steel Res. 66 (2010) 1374–1386.

[23] AS. AS/NZS 1554.1: Structural Steel Welding, Part 1: Welding of Steel Structures.Standards of New Zealand, 2011.

[24] AS, A. AS 4100-1998, Steel Structures, Standards Association of Australia, Sydney,Australia, 1998.

[25] A. ANSI, AISC 360-05: Specification for Structural Steel Buildings, AmericanInstitute of Steel Construction Inc., Chicago, IL, 2005.

[26] J.A. Packer, J. Wardenier, X.L. Zhao, G.J. van der Vegte, Y. Kurobane, Design Guidefor Rectangular Hollow Section (RHS) Joints under Predominantly Static Loading-Cidect, 2nd ed., TÜV Rheinland Gmbtt, Verlag, 2009.

[27] H. Krawinkler, Performance assessment of steel components, Earthq. Spectra 3(1987) 27–41.

[28] Y.J. Park, A.H.S. Ang, Y.K. Wen, Damage-limiting aseismic design of buildings,Earthq. Spectra 3 (1987) 1–26.

[29] E. Cosenza, G. Manfredi, R. Ramasco, The use of damage functionals in earthquakeengineering: a comparison between different methods, Earthq. Eng. Struct. Dyn. 22(1993) 855–868.

[30] D. Systèmes, Abaqus 6.10: Analysis User's Manual, Dassault Systèmes Simulia Corp,Providence, RI, 2010.

[31] L.H. Han, X.L. Zhao, Z. Tao, Tests and mechanics model for concrete-filled SHS stubcolumns, columns and beam-columns, Steel Compos. Struct. 1 (2001) 51–74.

Q.-Y. Song et al. Thin-Walled Structures 116 (2017) 250–264

264

本文献由“学霸图书馆-文献云下载”收集自网络，仅供学习交流使用。

学霸图书馆（www.xuebalib.com）是一个“整合众多图书馆数据库资源，

提供一站式文献检索和下载服务”的24 小时在线不限IP

图书馆。

图书馆致力于便利、促进学习与科研，提供最强文献下载服务。

图书馆导航：

图书馆首页 文献云下载 图书馆入口 外文数据库大全 疑难文献辅助工具