Post-Northridge connections with slit dampers to enhance strength

Journal of Constructional Steel Research 80 (2013) 138–152

Contents lists available at SciVerse ScienceDirect

Journal of Constructional Steel Research

Post-Northridge connections with slit dampers to enhance strengthand ductility

H. Saffari a,⁎, A.A. Hedayat b, M. Poorsadeghi Nejad b

a Department of Civil Engineering, Shahid Bahonar University of Kerman, 22 Bahman Blvd. P.O. Box 76175-133, Kerman, Iranb Department of Civil Engineering, Islamic Azad University–Kerman Branch, Joopar Road, P.O. Box 7635131167, Kerman, Iran

⁎ Corresponding author. Tel.: +98 9131411509; fax:E-mail addresses: [email protected] (H. Saffari),

(A.A. Hedayat), [email protected] (M. Poorsadeg

0143-974X/$ – see front matter © 2012 Elsevier Ltd. Alhttp://dx.doi.org/10.1016/j.jcsr.2012.09.023

a b s t r a c t
a r t i c l e i n f o
Article history:Received 25 March 2012Accepted 29 September 2012Available online xxxx

Keywords:DuctilityStrengthSlit damperPre-Northridge connection

In 1994 and 1995, Northridge and Kobe earthquakes caused unexpected damages to beam-to-column connec-tions which were followed by extensive investigations on the connections' behavior to resolve problems intheir performance and to improve their strength and ductility. During these years methods were proposedby researchers which were mainly based on reducing the beam section or strengthening the connection.Adding slit dampers at the top and bottom of the beam flanges plates is another way to improve the connectionbehavior. These dampers are able to absorb and dissipate a significant amount of energy. Slit dampers caused aremarkable reduction in the plastic strain at the column face area and consequently kept plastic hinge formationaway from the column face. In this study a total of 8 small slit damperswere used at the column face area. Tofindout best configuration for slit dampers which can only be used for newly designed connections, a parametricstudy was carried on their geometry. Generally the slit damper yields in shear or in flexure. Hence, fortwo different yield mechanisms (shear yielding and flexural yielding) details for designing of slit damperswere proposed and connection's strength and ductility were compared.

© 2012 Elsevier Ltd. All rights reserved.

1. Introduction

Ductile rigid connections must be able to provide both adequatestrength and rotational capacity. The main concern for most of rigidconnections (e.g. pre and post-Northridge connections) is their lowplastic rotational capacity rather than their strength. Low plasticrotational capacity of a connection causes a premature fracture ofbeam flange at the weld metal before developing plastic hingesinto the beam length. Other factors that are believed to contributeto the brittle fractures of Northridge connections are discussed inRefs. [1–4]. Fig. 1 shows a typical pre-Northridge connection. Themodified pre-Northridge connections which are now called thepost-Northridge connections use smooth weld access holes, highfracture toughness weld metal E70-TGK2, long and thick shear tabplates and no backing bar at the bottom beam flange [3]. However,post-Northridge connections did not achieve adequate plastic rotationas required by the seismic codes [3]. As a result further modificationswere applied on post-Northridge connections. During previous years,various modifications were proposed by researchers to resolve theductility problem of post-Northridge connections. Generally thesemodifications were based on only two methods: strengthening ofthe connection or weakening of the beam section. In the first methodsome additional elements including cover plates [5], triangular

+98 [email protected] Nejad).

l rights reserved.

haunches [6], straight haunches [7], upstanding ribs [8], lengthenedribs [9], side plates [10] and bolted brackets [11,12] are added tothe connection at the column face level. These elements are intendedto keep the beam end at the column face level in an elastic range whenplastic hinges are developed. In the secondmethod, a reduction ismadein the beam section at a distance away from the column face. It enforcesthe formation of plastic hinge at the reduced beam section. Weakeningthe beam section can be done either by cutting a portion of the beamflange (reduced beam section, RBS, connections [13]) or the beamweb (RBW connections). RBW connections include the wedge designbeam connections [14,15] and reduced beam web with circular voids[16,17], rectangular long voids [18], sinusoidal voids [19] and drilledvoids [20]. Slotted web connection is another solution where the levelof strains is decreased in the beam flange to column connection, usinga reduced secondary stress effect [21]. It should be noted that in thepost-Northridge connections, there are only two sources to dissipatethe seismic energy, beam end and panel zone (PZ).

All methods mentioned in the preceding paragraph are intended toincrease the energy dissipation capacity of the main source; beam end.However, the connection ductility could also be increased by addinganother energy dissipation source to the other two sources. One exampleis the use of slit dampers (Fig. 2), originally proposed by Oh [22]. Presentstudy aimed to use welded slit dampers as additional energy dissipationsource to increase the ductility of newly designed connections. To findout the best configuration of slit dampers, a parametric study usingthe finite element method was carried on the geometry of slit damperswith respect to different beam length–beam-depth ratios. Moment–

http://crossmark.dyndns.org/dialog/?doi=10.1016/j.jcsr.2012.09.023&domain=f

http://dx.doi.org/10.1016/j.jcsr.2012.09.023

mailto:[email protected]



http://dx.doi.org/10.1016/j.jcsr.2012.09.023

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

Fig. 1. Typical pre-Northridge beam-to-column moment connection.

139H. Saffari et al. / Journal of Constructional Steel Research 80 (2013) 138–152

rotation curve of the modified connections were compared and thebest design parameters to achieve adequate connection ductilitywere proposed.

2. Slit dampers and their application in rigid connections

Fig. 2 shows a slit damper and its lateral displacement induced byforce P. Lateral displacement of slit damper develops two differentyield mechanisms; shear yielding and flexural yielding. The yieldingtype depends directly on the geometric properties of slit damper.Fig. 3 shows the geometric properties defined by Chan and Albermani[23]. In this reference, the required force P to yield the slit damper wasdefined as the lesser value obtained from Eqs. (1) and (2), which werecorresponding to the achievement of shear or flexural yielding of slitdamper respectively.

Py ¼n⋅σy⋅t⋅B3

ffiffiffi3

p ð1Þ

Py ¼n⋅σy⋅t⋅B2

2H′: ð2Þ

Fig. 2. Slit damper and deformation due to loading.

In these equations: σy, t, B, n and H′ are tensile yield stress of slitdamper material, thickness of struts, width of struts, number of strutsand equivalent height of slit damper (Fig. 3) respectively.

Application of slit dampers in rigid connections was first proposedby Sang Hoon Oh et al. [24] (Fig. 4). In this type of connection, slitdampers were used beneath the bottom beam flange. The top of theslit dampers were welded to a thick plate where this plate was boltedto the bottom beam flange. The bottom of the slit dampers wereseated on a strong split-T, where the split-T was bolted to the columnflange. Also a strong split-T was used to connect the top beam flange tothe column flangewith high strength bolts. The limited plastic deforma-tion developed in the upper split-T, does not require repairing after anearthquake. In Ref. [24], three specimenswith slit dampers were tested.Two of themwere tested without concrete slab (specimens D1 and D2)and the third one with concrete slab (specimen D2C). In all specimens,slit dampers were designed to keep the beam in the elastic region. Thismeans that the maximum bending moment in the beam developed bythe slit dampers when the slits were in their ultimate state was limitedbelow the beam plastic moment. The ratio of maximum bendingmoment to the beam plastic moment (parameter α) was 0.34 and0.63 for specimens D1 and D2, respectively.With respect to this discus-sion, the geometric properties of slit dampers (Fig. 3) were determinedusing Eq. (3). In this equation σyd and σud are yield stress and ultimatestrength of slit damper's material respectively, while Mpb is the beamplastic moment. Lb is the beam length measured from the beam tip tothe column face, Lb1 is the beam length measured form the beam tipto the middle of slit damper and dst is the distance between the webcenter of upper and lower T sections.

Py ¼ ασyd

σud

Mpb

dst

Lb1Lb

: ð3Þ

All tested specimens in reference [24] were beams having overalldepth up to 582 mm and the beam length (distance between twoadjacent column faces) was 7000 mm, indicating the beam length–beam overall depth ratio (2Lb/db) equal to 12. Test results showed thatboth D1 and D2 specimens could reach to the minimum requiredductility (total rotation=4%). The maximum strength of specimenD1 was not developed at the beam plastic moment. Although theratio of the beam yield strength to the beam plastic moment wasequal to 0.63, the maximum strength of specimen D2 was almostthe same as the beam plastic moment. This was due to the effect ofslit damper strain hardening [24].

3. Aim of study

The present study benefit slit dampers to enhance the ductility ofpost-Northridge connections and proposes a simpler connection detail.Using slit dampers, several connection configurations are proposed andassessed to find out the best configuration of slit dampers in combinationwith post-Northridge connections. Parametric studies are conductedusing five non-modified post-Northridge connections. These specimenscomprised of three pre-tested specimens, SAC3 (beam: W24×68;column: W14×120), SAC5 (beam: W30×99; column: W14×176) andSAC7 (beam: W36×150; column: W14×257) utilized by Lee et al. [3],a designed specimen by Hedayat et.al. [18], SPE1 (beam: W18×46;column: W14×82) and a newly designed specimen by the authors,SPE2 (beam: W12×30; column: W14×43) which are modeled usinga multipurpose finite element program ANSYS [25]. The beam depthsvaried from 300 mm to 912 mm to consider a wider range of beamdepths. These specimen sizes were chosen since they may be goodrepresentatives of the conventional pre/post Northridge specimensizes range, small, medium and large [3]. These specimens were alsotested in Phase 1 of SAC Steel Projects [26] and connectionmodificationprocedures presented by Engelhardt and Uang [27].

Fig. 3. Geometric properties defined by Chan and Albermani [23] for slit damper.

140 H. Saffari et al. / Journal of Constructional Steel Research 80 (2013) 138–152

4. Finite element modeling

To better understand the behavior of post-Northridge and slit damperconnections, and to verify the accuracy of finite element modeling, atfirst, specimens SAC3, SAC5 and SAC7 of reference [3] and specimenD1 of reference [24] was remodeled using finite element method. For

Fig. 4. Proposed connection by

instance, Fig. 5 shows the finite element mesh of specimen D1ofwhich details are shown in Fig. 6. In this model, all connection's partswere modeled using shell elements except the plate on the top of slitdampers and the T sections which were modeled using solid elements.Based on the experimental results presented in reference [24], thefailure of specimen D1 was not due to the failure of bolts (also it was

Sang Hoon Oh et al. [24].

Fig. 5. Finite element model of specimen D1.


the case for SAC specimen of reference [3]). Bolts were not exactlymodeled but their effects were considered using coupling method inANSYS software. Also, interaction between plates (e.g. interactionbetween flange of beam and flange of split-T) was modeled usingcontact elements.

Totally three different elements, SHELL43, SHELL181 and SOLID95were used. SHELL43 and SHELL181 are a single-layer four-node andmulti-layer eight-node shell elements respectively. SOLID95 is asingle-layer twenty-node brick element. Shell elements and brickelements have six (three translation and three rotation) and three

Fig. 6. Detail of spe

(translation) degrees of freedom at each node respectively. All ofthese elements have the ability to model plasticity, large deflection,and large strain phenomena. SHELL43was used tomodel columnplatesand SHELL181was used tomodel beam and slit damper plates. SOLID95was also used to model split-T plates and the plate on the top ofslit-damper. In the case of using SHELL181, each element was separatedinto five layers across the thickness. The number of layers was selectedbased on the finite element study carried by Gilton and Uang [28]. Inorder to determine the appropriate mesh density, a study on meshsensitivity was carried based on the recommendation given by ANSYS

cimen D1 [24].

Fig. 7. Comparison between experimental and analytical results of specimen D1.


software and then results compared with experimental results ofreference [24].

To perform material nonlinearity analyses, plasticity behavior wasbased on the Von-Mises yielding criteria and the associated flow rule.Isotropic hardening was assumed for the monotonic analysis, whereaskinematic hardening was assumed for the cyclic analysis as used byMao et al. [29] Ricles et al. [30]. A bilinear material response was usedfor base metals based on the material properties given in reference[24], while for weld metals (welds were used for modeling of post-Northridge connections) a multi-linear material response based onmaterial property given by Mao et al. [29] and Ricles et al. [30] wasused.

Nonlinear geometric analyses were performed through a smallstrain, large displacement formulation. The monotonic analyses wereconducted by applying a monotonic vertical displacement load to thebeam tip until more than 4% total rotation at column web center wasachieved. Load history recommended by FEMA-355D [31] was usedfor cyclic analyses.When applied loads are in only the vertical direction,then the out of plane deformations (normal to the web) may not occur.Therefore, in order to ensure that buckling would occur due to instabilityof model, the imperfect model subjected to cyclic or monotonic loading

Fig. 8. Plastic equivalent strain d

was used. In this study, in order to determine the imperfect model, firstthe bucklingmode shapeswere computed in a separate buckling analysisand then were implemented to perturb the original perfect geometry ofthe model.

In order to verify the accuracy of themodel, the experimental resultsof D1 from reference [24] and all SAC specimens from reference [3]were compared with the analytical results obtained from the finiteelement models in terms of moment and rotation. For instanceFig. 7 shows this comparison for specimen D1. As this figure showsanalytical result is in good agreement with experimental result. Fig. 8shows the plastic equivalent strain distribution in specimen D1. As thisfigure shows almost all of plastic equivalent strains are concentrated atthe slit dampers indicating that the slit dampers act as the main energydissipation source.

It should be noted that the fracture prediction is the most question-able part of afinite element study. Because it is inherently a complicatedphenomenon and is dependent on many parameters such as weld andbase metal properties, weld defects, notch effects, weld quality andweld toughness. In this study it was assumed that the qualifiedweldersand fabricators are employed and high fracture toughness weld metalsare used (as these should be, based on AISC 2010, [32]). Also finite

istribution of specimen D1.


element results obtained from specimen D1 are compared with theexperimental results presented by Oh [24]. Considering the factthat the locations of high level of strains have significant probabilityof premature fracture, a failure criterion was assumed as follows:

The connection fracture occurs when the Von-Mises strains at thewhole width of beam flange or the flange plates or at all struts of theslit damper plates exceeds the strain associated with the ultimatestrength of the materials, based on the material properties reportedby Lee [3] and Oh [24].

A similar failure criterion alsowas used by other researchers (e.g. [33],[34]). Note that the connection failure can also be predicted using ruptureindex, RI (e.g. [35], [36]). This index is, in fact the ratio between the plasticequivalent strain index and the ductile fracture strain. However, as statedby most of researchers (e.g. [36]), none of the mentioned methods areintended to predict exact rotation capacities for connections; rather,they provide a tool for comparing various models.

5. Proposed connection configurations SDS1 to SDS3

Fig. 9 shows the first proposed connection's detail named as SDS1.In this connection, slit dampers were added to the specimen SAC7,which have the deepest beam and the lowest beam length–beam tooverall depth ratio among all other SAC specimens presented in Ref.[3]. The top side of the slit dampers was welded to the lower face ofthe bottom beam flange. The bottom side of slit dampers was weldedto the web of an I-section member where the flange and web ofI-section were welded to the column flange using groove welds.Several lengths of I-section members and various thicknesses of slitdampers were examined. Generally results showed the unacceptableperformance of the SDS1 connection detail. However, using weakerslit dampers which a higher energy dissipation capacity caused a slightlyincrease in the connection ductility. Early top beam flange fracture at thecolumn face area (Fig. 9a) caused the average of the maximum totalrotation at the column web center of the SDS1 specimens to bearound 1.16% which was below the minimum required ductility, fourpercent total rotation. The main problem for this connection detail

Fig. 9. Proposed connection SDS1 with Von-Mis

was the low participation of slit dampers in the dissipation of imposedloads (Fig. 9b). Low participation of slit dampers was due to thelow lateral displacement demands at the top of slit dampers whichwas, in turn, due to the presence of groove weld column flange.

To enhance the participation of slit dampers to dissipate the seismicloads, connection details SDS2 and SDS3 were examined. All thesemodifications were applied on the specimen SAC7. In SDS2 specimen(Fig. 10) energy dissipation system comprised of a slit damper and anend thick plate. The left side of slit damper was welded to the columnflange and its right side was welded to the end thick plate. Slit damperlengthwas selected 500 mm. The thickness of slit damperwas designedas such that the yielding of slit damper in shear occurswhen the verticaldisplacement of the end tick plate reach to a value corresponds to 1 to2% total rotation at the column web center. The maximum achievedductility was 2.1% and connection failed due to the fracture of the topbeam flange at the column face level. To increase the relative verticaldisplacement between the end points of slit damper, and hence to in-crease the strain demands in the slit damper, connection detail SDS3was proposed (Fig. 11). The details and the design procedure of thisconnectionwere similar to those defined for SDS2, except that a verticalstiffener was added beneath the bottom beam flange. This stiffener re-duced the vertical displacement of the beam end at the column facearea and consequently increased the relative displacement of the endsof slit damper. However, this modification was not able to increasethe connection ductility. The maximum total rotation achieved was2.2% and failure mode was the same as that observed for SDS2 detail.

6. Proposed rigid connection SDS4

Fig. 12 shows the details of SDS4. Applied moments and shearforces were transferred to the column face level through a dissipativeload transferring system and to shear tab respectively. This load trans-ferring system was intended to reduce the amount of stress and straindemands at the column face level as follows.

In this connection unlike the conventional post-Northridge connec-tion detail, the beam flanges were not directly welded to the column

es strain distribution at 2.7% total rotation.

Fig. 10. Details of proposed slit damper connection SDS2.


flange. However the beam flanges along the beam length were firstwelded to the middle of two slit dampers (plates A in Fig. 12) by eithergroove or fillet welds. Top and bottom sides of slit dampers were thenwelded to a wider and two narrower flange plates (plates B andC respectively) where these plates were themselves groove welded to

Fig. 11. Details of proposed slit

the column flange. In this connection, tensile and compressive forcesdeveloped by applied moment were first transferred to two slitdampers. Slit dampers were designed as such to yield under these lateralloads. Yielding of slit dampers reduced the amount of lateral loads. Finallya reduced tensile or compressive force was transferred to the flange

damper connection SDS3.

Fig. 12. Configuration of proposed rigid connection SDS4.


plates (plates B and C) and their groovewelds as it was intended to causea delay in the failure of connection.

Because of the details proposed for SDS4 specimen, it appears tobe only applicable for newly designed connections. In addition, inthis section it was assumed that the fracture of connection occurswhen the Von-Mises strains at the whole width of beam flange or theflange plates (plates B and C, Fig. 12) or at all struts of the exterior slitdamper plates exceed the strain corresponding to the ultimate strengthof the materials.

6.1. Assumptions and design procedures

6.1.1. Length of slit damperFig. 13 shows the slit damper used in the SDS4 type specimen. The

length of slit damper (Lslit) is a function of the required length of filletwelds A, LFWA, to connect the slit plate (plate A) to the beam flange

Fig. 13. Geometry of used slit d

plates (Lslit=LFWA+50 mm). LFWA can be determined using Eq. (4).The design of the fillet welds was based on the maximum momentdeveloped at column face (Mf) which is a function of limiting moment,Mpd, and the associated shear force, Vpd, at critical plastic section. Thecritical plastic section was assumed to be at the end of slit damperand itwas estimated at a distance equal to 80 to 90% of the beamoveralldepth away from the column face (SC=0.8db to 0.9db).Cpr factor accountsfor the peak connection strength, including strain hardening, localrestraint and other connection conditions. In FEMA350 [37], the Cprfactor is given by equation (Fy+Fu) /2Fy, where Fy and Fu are thespecified minimum yield and tensile stress of material respectively.FEMA350 proposes the use of value 1.2 for any case ofmodified connec-tions exceptwhere otherwise noted in the individual connection designprocedures. In the present study the Cpr factor was chosen as 1.4. Thereason for choosing such value is explained in Section 6.6. Fye, is theexpected yield stress of beamflanges and Ry is amultiplier that accounts

amper in SDS4 specimen.

0.0

0.4

0.8

1.2

1.6

2.0

0.00 0.01 0.02 0.03 0.04 0.05

M/M

p

Rotation(rad)

W36x150 , L/d=7.92

α=0.3

α=0.4

α=0.5

SAC7

Fig. 14.Moment–rotation curves of SAC7–SDS4 specimenswith Lb/db=7.92 and differentvalues of parameter α.


for material over strength [38]. FEXX, aw, w, and L′ are electrode classifi-cation number, fillet weld leg size, factored uniform load on the beam,and beam span between critical plastic sections respectively.

Lslit ¼ Mf = 0:89db⋅aw⋅FEXXð Þ: ð4Þ

Mf ¼ Mpd þ Vpd⋅Sc: ð5Þ

Mpd ¼ Cpr⋅Z⋅Fye: ð6Þ

Vpd ¼ 2Mpd=L′þw⋅L′=2: ð7Þ

Fye ¼ Ry⋅Fy: ð8Þ

6.1.2. Details of top and bottom flange plates (plates A and B)The lengths of top and bottom flange plates were defined by Lslit. The

widths of top and bottom flange plates were assumed to be equal to thecolumn flange width and to be more than one quarter of beam flangewidth respectively. The thickness of plate B was computed by dividingthe beam flange area by the flange plate width. However, its thicknesswas selected as such to be at least equal to the minimum requiredweld leg size to connect the slit plates (plate A) to the flange plateB. Analytical results showed that the level of stresses at flange plateC is significantly less than those of flange plate B. Hence, it can bededuced that despite of the smaller width of the bottom flange plate Ccompared to the top flange plate B, their thicknesses were selected tobe same.

6.1.3. Design force of slit dampersAs shown in Fig. 12, at the level of each beam flange, four slit

dampers were used. Slit dampers must be designed as such to yieldunder the lateral force developed in each slit due to the plasticizationof the beam section. This lateral force can be estimated by the followingequation:

Fslit ¼ 0:25P ¼ α⋅Mp= 4dbð Þ ð9Þ

whereMp and db are the beam plastic moment and beam overall depthrespectively. In Eq. (9),α is amultiplier and limited to one. This parameteraccount for the yielding of slit dampers to occur before the formation ofplastic hinge in the beam section. Smaller values of parameter α leads toearly yielding of slit damper plates and causes early connection failure.On the other hand, large values of this parameter, lead to having thickerslit damper plates and consequently reduce the energy dissipationcapacity of the slit dampers. Ref. [24] proposes the use of 0.6 for this pa-rameter. However, there is a big difference between the connectionpresented in Ref. [24] and the one proposed in this research from theconnection configuration, load transferring system and connection de-tails points of view. Hence, in this study a parametric study was carriedto find out the optimum value of parameter α to achieve the best con-nection performance with respect to the different beam overall depths,different beam length–beam overall depth ratios and different yieldmechanisms of slit damper plates (shear yielding or flexural yieldingof slit dampers due to force Fslit). This parametric study led to a totalof model of which their details are presented in the following sections.

6.1.4. Details of slit damperPreliminary finite element results showed that connection strength

and ductility is increased by reducing the slit damper height. However,excessive reduction in the slit damper height drastically limits theaccess space for welding process. Hence, distance between the centroidsof flange plates B and C was taken as 200 mm.

Other slit damper's details such as Vw, B, n, t, (Figs. 3 and 13) canbe determined based on either shear yielding of slit damper (equatingEqs. (1) and (9)) or flexural yielding of slit damper (equating Eqs. (2)

and (9)). Preliminary investigations also indicated that shear yielding ofslit damper occurs when n≥5 and 10 mm≤B≤50 mm while flexuralyielding of slit dampers governs the design if n≥10 and B is around10 mm. It should be noted that in any case the thickness of slit damperis a function of parameter α. Hence, using a parametric study, theoptimum value of αwas determined for each design criteria separately.

6.2. Parametric study

As mentioned above totally eighty SDS4 type specimens of differentbeam and column sections, beam length–beam depth ratios, yieldingcriteria of slit damper and parameter α were analyzed. For instance,Fig. 14 compares the moment–rotation curves of SAC7–SDS4 speci-mens, with Lb/db=7.9 and designed based on the shear yieldingcriteria. As this figure shows all modified connections have higher con-nection strength and ductility when compared to the non modifiedspecimen SAC7. By increasing parameter α and consequently increasein slit damper thickness, the connection initial rotational stiffness, con-nection strength and ductility increases. However, as can be seen inFig. 15a, excessive increase in this parameter caused a reduction in theenergy dissipation capacity of slit dampers and promotes the beamflange fracture at the end of slit dampers. It caused a reduction in theconnection ductility. It should be noted that small value of parameterα is also not desirable. However it increases the energy dissipationcapacity of slit dampers, but it promotes the early fracture of outerslit dampers, possibly the one which is closer to the flange plate B(Fig. 15b). Therefore, it can be concluded that there is an optimumvalue of parameter α to achieve best connection performance. Theresults of all SDS4 specimens are summarized in Tables 1 and 2 forthe shear and the flexural yielding of the slit damper respectively.

By comparing the results presented in Table 1 it can be concludedthat for the shear yielding criteria, an optimum value of parameter αequal 0.5 is chosen to achieve both minimum required strength andductility (M/Mp=0.8 and θ=4%). Such a value for α corresponds toan average connection strength and ductility of 1.7 and 4.35% radianrespectively. The failure of these connections was estimated to bedue to the beam flange fracture at the end of slit dampers. Resultsshowed that for a given values of parameter α (e.g. α=0.5, specimenSAC7), an increase in Lb/db, cause the connection ductility to increases di-rectly. The minimum ductility was achieved for specimens of minimumbeam length–beam depth ratio (Lb/db=7.92) which was 3.85% radian.Table 1 also shows that for given values of parameterα and Lb/db, connec-tion ductility reduces as the beam height increase.

Considering the fact that specimen of lowest Lb/db had minimumductility, for flexural yielding criteria, only specimen of minimumLb/db (7.92) was investigated. Similar to the case of shear yielding,specimens of α=0.5 showed the highest ductility. In this case, the

Fig. 15. Von-Mises strain distribution of SAC7–SDS4 specimens at failure time for: (a) α=0.5 and (b) α=0.3.


average strength and ductility were 1.71 and 4.46% radian respectivelywhile the minimum achieved ductility was 4% radian.

6.3. Comparison of the two design criteria

Comparison of results presented in Tables 1 and 2 shows that usingflexural yielding criteria leads to the achievement of higher connectionstrength and ductility when compared with the use of shear yieldingcriteria. When slit dampers were designed based on the shear criteria,yielding of the struts started from their middle height and propagated

towards their ends rapidly. It therefore caused the yielding of the entireheight of almost all of struts at the occurrence of the connection failure(Fig. 16a). In contrast, in the case of using flexural yielding criteria,yielding started from the end of struts and its transmission rate to themiddle parts of struts was relatively low. In this case connection failurehappened before entire yielding over the strut height (Fig. 16b). For agiven slit damper height, the number of voids based on the flexuralcriteria is higher when compared with the one which was designedbased on the shear criteria. In comparison to using the shear criteria,the flexural criteria always lead to the use of thicker struts. Hence, the

Table 1Results of SAC-SDS4 specimens based on shear yielding criteria.

Spe. name Lb/db α Failure mode θ(%)

M/Mp

SAC7 7.92 0.3 OS 2.92 1.540.4 BF 4.15 1.740.5 BF 3.85 1.74

9 0.4 BF 4.31 1.770.5 BF 4.00 1.77

11 0.4 BF 4.62 1.710.5 BF 4.38 1.710.6 BF 4.15 1.71

SAC3 7.92 0.3 BF 2.92 1.370.4 BF 4.23 1.620.5 BF 4.00 1.61

9 0.4 BF 4.39 1.600.5 BF 4 1.61

11 0.4 BF 4.77 1.580.5 BF 4.39 1.59

SPE2 7.92 0.3 OS 4.46 1.470.4 NF 5.00 1.700.5 BF 4.92 1.80

9 0.3 OS 4.62 1.460.4 NF 5.00 1.660.5 BF 4.54 1.84

11 0.3 OS 5.00 1.450.4 NF 5.00 1.600.5 NF 5.00 1.70

SAC5 7.92 0.3 OS 3.23 1.560.4 BF 3.54 1.670.5 BF 4.00 1.66

9 0.3 OS 3.77 1.590.4 BF 4.15 1.680.5 BF 4 1.7

11 0.3 OS 4.54 1.490.4 BF 4.92 1.580.5 BF 4.62 1.58

SPE1 7.92 0.4 OS 3.46 1.600.5 BF 4.39 1.77

9 0.4 OS 3.69 1.540.5 BF 4.46 1.72

11 0.4 OS 4.23 1.530.5 BF 4.46 1.68

Summary of results.Average ductility when α=0.4 is 4.36%.Average ductility when α=0.5 is 4.35%.Average M/Mp when α=0.4 is 1.64.Average M/Mp when α=0.5 is 1.70.Minimum ductility when α=0.4 is 3.46%.Minimum ductility when α=0.5 is 3.85%.OS: outer slit; BF: beam flange.NF: no fracture.

Table 2Results of SAC–SDS4 specimens based on flexural yielding criteria, Lb/db=7.92.

Spe. name α Failure mode θ (%) M/Mp

SAC7 0.4 OS 2.77 1.540.5 BF 4.23 1.820.6 BF 3.92 1.72

SAC3 0.4 OS 4.46 1.640.5 BF 4.00 1.640.6 BF 3.69 1.64

SPE2 0.5 NF 5.00 1.650.6 NF 5.00 1.760.7 NF 5.00 1.84

SAC5 0.4 OS 2.00 1.260.5 BF 4.08 1.670.6 BF 3.69 1.67

SPE1 0.5 OS 5.00 1.780.6 BF 4.77 1.83

Summary of results.Average ductility when α=0.5 is 4.46%.Average M/Mp when α=0.5 is 1.71.Minimum ductility when α=0.5 is 4%.


flexural criteria lead to higher fabrication and material cost for slitdampers when compared to using the shear yielding criteria. The resultsalso showed that the failure of slit dampers is promoted when the shearand flexural yielding of slit dampers occur simultaneously (i.e. Eqs. (1)and (2) be equal). Hence, it is recommended to avoid such designs.

6.4. The effect of cyclic loading

Generally cyclic loading increases the level of developed strains.Hence to investigate the effect of cyclic loading on the connectionbehavior, two SDS4 type specimens were modeled and loaded undercyclic loading. The used beam section and value of parameter α forboth specimens were W30×99 and 0.5 respectively. One of them wasdesigned based on the shear yielding criteria with Lb/db equal to 11while the other one was designed on the basis of developing flexuralyielding in the slit dampers with Lb/db equal to 7.92. Figs. 17 and 18compare the moment–rotation curves of these specimens under cyclicand monotonic loadings. As these figures show, both specimens havestable hysteretic curves until failure occurs. There was no strength andstiffness degradation which usually occur due to the presence of localbuckling of connection elements. There were also a good agreementbetween the moment–rotation curves obtained based on cyclic loadingand the ones obtained using monotonic loading.

6.5. Plastic equivalent strain distribution

In a connection, the locations of high level of plastic equivalentstrain (PEEQ) can be evaluated as the locations of probable plastichinges and ultimate fracture location. For instance, Fig. 19 showsPEEQ distribution for a SDS4 type specimen designed based on theshear yielding criteria having Lb/db equal to 9. As this figure shows,there are three sources to dissipate the seismic loads, panel zone(PZ), beam end, and slit dampers. The contribution of the slit dampersto dissipate the seismic loads was more when compared with that ofthe other two sources. In addition, this figure shows that when themaximum capacity of the outer slit dampers is reached, the inner slitdampers have still a large potential to deform and to absorb energy.This ensures the success of connection to dissipate the seismic loadsand the ability of connection to be stable after primary failure of outerslits and avoid to the sudden failure of connection when outer slitdampers are failed.

6.6. Moment–rotation curve of a typical SDS4 design specimen

Based on thefinite element results, themoment–rotation characteris-tics of a SDS4 design connection can be estimated using the least squaremethod. Fig. 20 shows the moment–rotation curve of a typical SDS4design specimen which is idealized by a bilinear relationship withthe following key parameters: 1) initial rotational stiffness, K1; 2)post yielding rotational stiffness K2; 3) moment ratio at the yieldpoint, ry; and 4) moment ratio at the ultimate point, ru. The idealizedcurvewas obtained in such away that the area under both the nonlinearcurve and the bilinear relationship is to be approximately the same.

The key parameters used to idealize themoment–rotation curve canbe represented as a function of the beam overall depth (db), the beamlength–beam overall depth ratio (Lb/db) and parameter α (used inEq. (9)). Therefore, in this study the nonlinear models defined asEqs. (10) to (14) were used to estimate these key parameters. Thevariables C1 to C4 were determined using regression analysis and aresummarized in Table 3. For each key parameter, regression analyseswere performed using two different data bases. The first databaseconsisted of all data, considered all values of parameter α, whilethe second database consisted of only specimens of the best valuesof α (0.5), indicating for the second data base variable C2 is zero.

In Eq. (10), ES, Ib, Lb, Zb and fy, are the modulus of elasticity of steel,the beam moment of inertia, the beam length (distance between the

Fig. 16. Yield mechanism of struts designed based on: (a) shear criteria and (b) flexural criteria.


faces of the two adjacent columns), the beam plastic section modulusand the yield stress of material respectively. In Eqs (11) to (14) thebeam overall depth should be written in millimeter. The two lastcolumns of Table 3 give the errors observed for each key parameter.Err-1 is the average of absolute errors while Err-2 is the mean squareerror which shows the impact of large errors (∑1

n(real parameter−estimated parameter)2, n is total number of data).

K1 ¼ β⋅Kb

Mpb¼ β

ES⋅Ib=LbZb⋅f y

: ð10Þ

β ¼ C1⋅αC2⋅dC3b ⋅Lbdb

� �C4: ð11Þ

K2

K1¼ C1⋅αC2⋅dC3b ⋅

Lbdb

� �C4: ð12Þ

ry ¼My

Mpb¼ C1⋅αC2⋅dC3b ⋅

Lbdb

� �C4: ð13Þ

ru ¼ Mu

Mpb¼ C1⋅αC2⋅dC3b ⋅

Lbdb

� �C4: ð14Þ

As explained before, the Cpr factor (used in Eq. (6)) accounts for thepeak connection strength, including strain hardening, local restraintand other connection conditions. This factor should be adjusted basedon the test results conducted on the connections of the proposed type.

However, in the absence of the experimental data, this factormight be approximately evaluated using the finite element results.

Fig. 17. Moment–rotation curves of SDS4 type specimen designed based on shear criteria.


Eq. (15) can be used to estimate the Cpr factor, where ru is given inEq. (14) and SC is equal to 0.9db.

Cpr ¼ ru⋅Lb−2SC

Lb¼ ru⋅

Lb=db−2SC=dbLb=db

¼ ru⋅Lb=db−1:8

Lb=db: ð15Þ

In this study the beam length–beam overall depth ratio (Lb/db)varied between 7.9 and 11. Based on the finite element results, thepeak values of ru corresponding to the lower and upper bounds ofLb/db were 1.8 and 1.71 respectively, corresponding to a Cpr between1.386 and 1.43. Hence, in this study the Cpr factor was chosen as 1.4(the average of the extreme values). This value shows a high degree ofstrain hardening due to the presence of slit dampers (as also indicatedin Refs. [39] and [40]). It is also the same as the Cpr factor proposed for

Fig. 18. Moment–rotation curves of SDS4 type sp

the welded unreinforced flange–welded web (WUF–W) momentconnections presented in the ANSI/AISC 358–10 [41]. However,based on the values used for the design parameter α, Lb/db and db, Cprfactor can be estimated using Eqs. (14) and (15) with the factors C1 toC4 presented in Table 3.

6.7. Simplicity in connection fabrication and the effect of the beam compositeaction

As discussed in the previous sections, based on the finite elementresults, there might be a high possibility for a specimen of SDS4 designto achieve adequate connection strength and ductility. However, twoother important points should also be noted: 1) simplicity of theconnection details to minimize the difficulties during the connection

ecimen designed based on flexural criteria.

Fig. 19. PEEQ distribution of a typical SDS4 type specimen.


fabrication and 2) the effect of concrete slab on the seismic performanceof the connection.

Comparison of the details for specimen D1 proposed by Oh [24](Fig. 6) and the SDS4 design presented in this study (Fig. 12), showsthat the original proposal has a simpler detail. However, in this specimen(specimen D1) totally 50 bolts were used. Fabrication of these staggeredbolt holes on the relatively thick plates of the split-T sections and thebeam flange seems to be time consuming and costly. In addition, fordeeper beam sections where bigger sizes of bolts are required, makingthese bolts holes may reduce the flexural strength of the beam section.The beam section used in specimen D1 had an overall depth equalto 582 mm. The most comparable specimen to D1 specimen is SAC3specimen where the used beam section is W24×68 (overall depth is600 mm). Using the SDS4 design, the length of slit dampers for SAC3specimen was 350 mm with 15 mm-thick flange plates whereas thelength of the slit dampers used in specimen D1 was 750 mm with the22 mm-thick plates, indicating a heavier connection details. Despitethese disadvantages, it should be noted that specimen D1 can be morereliable and easier to implement in the field since there is no any fieldwelds. However, for the SDS4 design, and adopting suitable fabrication

0.0

0.4

0.8

1.2

1.6

2.0

2.4

Nonlinear curve

Bilinear relationship

Yield point

Ultimate point

K1

K2

M/M

p

0.00 0.01 0.02 0.03 0.04 0.05

Rotation(rad)

Fig. 20. Idealized moment–rotation curve of a typical SDS4 design specimen.

steps, the difficulties involved in the implementation of such design inthe field can be reduced to an acceptable level.

As reported by numerous researchers ([24], [42] to [45]) by placingthe concrete slab on the top of beam flange, the beam becomes unsym-metrical and the neutral axismoves toward the top beamflange. It causesa higher strain demands on the bottom beam flange and in most casesleads to the fracture of the bottombeamflange. The behavior of a connec-tionmaynot be as intendedonce a compositefloor slab is put in place. Forinstance, proposed connectionswith composite beams in references [24],[42] and [43] achieved adequate ductility whilst it was not the case forspecimens tested by Kim [44] and Okada [45]. Hence, to realize the truebehavior of the SDS4 design, connections of this design should be testedexperimentally in the presence and absence of concrete slabs.

7. Conclusion

This study was aimed to use slit dampers to enhance the ductility ofthe newly designed and existing connections. For this purpose several de-signs were investigated. These led to propose a modified connectionnamed SDS4 type connection. Finite element results showed that thisconnection typewhich can only be specifically used for a newly designedconnection has high energy dissipation capacity. The main energy dissi-pation source is the slit dampers. Slit dampers were designed as such toyield under either shear or flexure, defined as shear or flexural designcriteria. Since the amount of load transmitted to each slit damper due

Table 3Variables C1 to C4 to predict key parameters β, K2/K1, ry and ru.

Type Data base C1 C2 C3 C4 Err-1 (%) Err-2

β All data 0.0108 0.1470 0.0556 0.6326 8.56 0.000α=0.5 0.0095 0.0000 0.0657 0.6110 8.40 0.182

K2/K1 All data 0.1768 0.3293 −0.0410 −0.1919 5.48 0.050α=0.5 0.1540 0.0000 −0.0418 −0.0196 3.66 0.169

ry All data 0.7750 0.3504 0.1366 0.0846 5.08 0.009α=0.5 1.2503 0.0000 0.0245 −0.0888 3.93 0.012

ru All data 2.2075 0.2457 0.0190 −0.0971 3.34 0.005α=0.5 2.7334 0.0000 −0.0347 −0.1194 3.09 0.004


to the plasticification of the beam sectionwas not known, it was taken asa variable denoted byparameterα. A parametric studywas carried tofindout the optimum value of parameter α for each design criteria to achievethe highest performance of connection. Results showed that α equal to0.5 is the optimum value for both design criteria. However, slit dampersdesigned based offlexural yielding caused the achievement of a relativelyhigher connection strength and ductility. Their fabrication and materialcosts were more when compared to slit dampers designed based onshear yielding criteria. Hence, using shear yielding criteria to design slitdampers was recommended from performance and fabrication costspoints of view. In addition, to realize the true behavior of the SDS4 design,connections of this design should be tested experimentally in thepresence and absence of concrete slabs.

References

[1] Mahin SA. Lessons fromdamage to steel buildings during theNorthridge earthquake.Eng Struct 1998;20(4–6):261–70.

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

[3] SAC/BD −00/01. Parametric tests on unreinforced connections, volume I-finalreport. Lee K-H, Stojadinovic B, Goel SC, Margarin AG, Choi J, Wongkaew A, ReyherBP, Lee D-Y.

[4] Hedayat AA, Celikag M. Fracture moment and ductility of welded connection. ProcInst Civ Eng Struct Build December 2009;162(SB6):405–18.

[5] Engelhardt MD, Sabol TA. Reinforcing of steel moment connections with coverplates: benefits and modifications. Eng Struct 1998;20(4–6):510–20.

[6] Chia B, Uang CM, Chen A. Seismic rehabilitation of pre-Northridge steel momentconnections: a case study. J Constr Steel Res 2006;62:783–92.

[7] SAC. Experimental investigations of beam-column sub assemblages. Technical reportSAC-96-01, parts 1 and 2. Sacramento, CA: SAC Joint Venture; 1996. p. 739–46.

[8] Popove EP, Tsai KC. Performance of large seismic steel moment connections undercyclic loads. Eng J AISC 1998;26(2):51–60.

[9] Chen CC, Lee JM, Lin MC. Behavior of steel moment connections with a singleflange rib. Eng Struct 2003;25:1419–28.

[10] Engelhardt MD, Sabol TA. Testing of welded steel moment connections in response ofthe Northridge earthquake. Northridge steel update 1. American Institute of SteelConstruction; 1994.

[11] Chen CC, Lin CC, Tsai CL. Evaluation of reinforced connections between steelbeams and box columns. Eng Struct 2004;26:1889–904.

[12] Kasai K, Hodgon I, Mao C. Bolted repair methods for fractured welded momentconnections proceedings. Behavior of steel structures in seismic areas (Stessa 97).Kyoto, Japan; 1997. p. 939–46.

[13] Popove EP, Yang T, Chang S. Design of steel MRF connections before and after 1994Northridge.

[14] Wilkinson S, Hurdmanb G, Crowtherb A. A moment resisting connection forearthquake resistant structures. J Constr Steel Res 2006;62:295–302.

[15] Hedayat AA, Celikag M. Wedge design: Reduced BeamWeb (RBW) connection forseismic regions. Adv Struct Eng 2010;13(2).

[16] Ascheheim M.-A. Moment–resistant structure, sustainer and method of resistingepisodic loads. United State Patent, Patent number: 6,012,256.

[17] Hedayat AA, Celikag M. Reduced beam web (RBW) connections with circularopenings structural steel: shapes and standards, properties and applications.New York, USA: Nova Science Publishers, Inc.; 2010.

[18] Hedayat AA, CelikagM. Post-Northridge connectionwithmodified beamend config-uration to enhance strength ductility. J Constr Steel Res 2009;65:1413–30.

[19] Hedayat AA, SaffariH,HadiA. Behaviour of steel reducedbeamweb (RBW)connectionswith sinusoidal voids. 5th SASTech 2011. Mashhad, Iran: Khavaran Higher-educationInstitute; May 12–14 2011.

[20] Hedayat AA, Saffari H, Eghbali A. Behaviour of steel reduced beamweb (RBW) connec-tions with drilled voids. 5th SASTech 2011. Mashhad, Iran: Khavaran Higher-educationInstitute; May 12–14 2011.

[21] Allen J, Richard RM, Partridge J. Seismic connection designs for new and existingsteel moment frame structures. Proc., 2nd World Conf. on Steel in Constr. Oxford:Elsevier Science Ltd; 1998.

[22] Oh SH, Seismic design of energy dissipating multi-story frame with flexible-stiffmixed type connection. Ph.D. Thesis, Japan:Tokyo University; 1998.

[23] Chan RWK, Albermani F. Experimental study of steel slit damper for passive energydissipation. Eng Struct 2008;30:1058–66.

[24] Oh SH, Kim YJ, Ryu HS. Seismic performance of steel structure with slit dampers.Eng Struct 2009;31:1997–2008.

[25] ANSYS. ANSYS user manual. ANSYS, Inc.; 2007[26] SAC. Experimental investigations of beam-column sub assemblages; technical report

SAC-96-01, parts 1 and 2. Sacramento, CA: SAC Joint Venture; 1996.[27] Engelhardt MD, Uang CM, Gross JL, Kasai K, Iwankiw N. Modification of existing

welded steel moment frame connections for seismic resistance. AISC; 2003.[28] Gilton CS, Uang CM. Cyclic response and design recommendations of weak-axis

reduced beam section moment connections. J Struct Eng 2002;128(4).[29] Mao C, Ricles J, Lu L, Fisher J. Effect of local details on ductility of welded moment

connections. J Struct Eng 2001:1036–44.[30] Ricles JM, Mao C, Lu LW, Fisher JW. Ductile details for welded unreinforced moment

connections subject to inelastic cyclic loading. Eng Struct 2003;25:667–80.[31] FEMA. State of the art report on connection performance. Report No. FEMA 355D,

Washington; September 2000.[32] Specification for Structural Steel Buildings. Specification for structural steel buildings.

ANSI/AISC 360–10; June 22, 2010.[33] Berman JW, Okazaki T, Hauksdottir HO. Reduced link sections for improving the

ductility of eccentrically braced frame link-to-column connections. J Struct EngMay 1 2010;136(5).

[34] Dusicka P, Itani AM, Buckle LG. Finite element investigation of steel built-up shearlinks subjected to inelastic deformations. Earthquake Eng Eng Vib 2004;3(2).

[35] Chao SH, Khandelwal K, El-Tawil S. Ductile web fracture initiation in steel shearlinks. J Struct Eng August 1 2006;132(8).

[36] Prinz GS, Richards PW. Eccentrically braced frame linkswith reducedweb sections.J Constr Steel Res 2009;65:1971–8.

[37] Recommended seismic design criteria for new steel moment-frame buildings.FEMA-350; 2000.

[38] Seismic provisions for structural steel buildings. ANSI/AISC 341–10; June, 2010.[39] Benavent ClimentA,Oh SH,AkiyamaH.Ultimate energy absorption capacity of slit-type

steel plates subjected to shear deformations. J Struct Constr Eng 1998;503(1):139–47.[40] Chan RWK, Albermani F. Experimental study of steel slit damper for passive energy

dissipation. Eng Struct 2008;30(4):1058–66.[41] Prequalified connections for special and intermediate steel moment frames for

seismic applications. ANSI/AISC 358–10; 2010.[42] Chen SJ, ChaoYC. Effect of composite action on seismic performance of steelmoment

connections with reduced beam sections. J Constr Steel Res 2001;57:417–34.[43] Uang CM, Yu QS, Noel S, Gross J. Cyclic testing of steel moment connections rehabili-

tated with RBS or welded haunch. J Struct Eng 2000;126(1):57–68.[44] Kim YJ, Oh SH, Moon TS. Seismic behavior and retrofit of steel moment connections

considering slab effects. Eng Struct 2004;26(13):1993–2005.[45] Okada K, Oh SH, Yamada S, Imaeda T, YamaguchiM,Wada A. Experimental study on

deformation capacity of composite beams with conventional-type beam-to-columnconnections. J Struct Constr Eng 2001;547(9):161–8.

Documents

Post-Northridge connections with slit dampers to enhance strength