Seismic Behavior of Reduced Beam Section Moment

Connections to Deep Columns

Xiaofeng Zhang1 and James M. Ricles, M.ASCE2

Abstract: An analytical study was conducted to investigate the seismic behavior of reduced beam section 共RBS兲 moment connections to
a deep wide flange column. Calibrated three-dimensional finite element models of RBS connections in perimeter special moment resisting
frames with deep columns were used to perform parametric studies under inelastic monotonic and cyclic loading. The parameters in the
study included: beam-to-column connection type, column section, composite floor slab, panel zone strength, and beam web slenderness.
The results from the parametric study show that a composite floor slab provides restraint to the top flange of the beams, reducing the
magnitude of beam top and bottom flange lateral movement in the RBS, column twist, and strength degradation due to beam instability
in the RBS. The effect of beam web slenderness contributing to an increase in the lateral movement of the beam flange in the RBS and
column flange is significantly reduced when a composite floor slab is present. A weaker panel zone results in higher potential for ductile
fracture in the connection region. An RBS connection to a deep column is shown to have less potential for ductile fracture in the
connection region than a welded unreinforced flange connection. The potential for fracture is increased in RBS connections when the
column section properties are reduced, leading to larger stress and strain.
DOI: 10.1061/共ASCE兲0733-9445共2006兲132:3共358兲
CE Database subject headings: Columns; Beams; Finite element method; Slabs; Seismic effects; Connections.

Introduction was performed on specimens with shallow wide-flange column

sections 共e.g., W12 and W14兲. The series of tests by Jones et al.
A reduced beam section 共RBS兲 beam-to-column moment connec- included a floor slab in some of the experiments, where it was
tion is often utilized in the design of special steel moment resist- found that specimens with a composite floor slab exhibited a
ing frames 共SMFs兲. The details of a typical RBS connection are more stable behavior by having the beam at the RBS restrained
shown in Fig. 1共a兲, where the beam flanges are reduced in width against lateral torsional buckling. The results of the above prior
over a prescribed distance from the column face. The beam experimental studies led to design recommendations 共Engelhardt
flanges are attached to the column using complete joint penetra- 1999; Uang and Fan 2001; Chi and Uang 2002兲 and prequalifica-
tion 共CJP兲 groove welds. The beam web is either welded to the tion of RBS moment connections for column sizes up to a W14
column flange with a CJP groove weld or bolted to a shear tab. In section 共FEMA 2000兲.
the latter case the shear tab is groove welded to the column For economical reasons, design engineers in the U.S. prefer to
flange. By design, the RBS connection develops inelastic defor- use columns deeper than a W14 section 共e.g., a W36 steel wide
mations primarily in the region where the beam flange width has flange section兲 in order to control seismic drift in SMFs. Recently,
been reduced 共referred to herein as the RBS兲, limiting the devel- three tests were performed by Chi and Uang 共2002兲 to investigate
opment of inelastic strain at the beam flange-to-column CJP the inelastic cyclic behavior of RBS moment connections to a
welds. W27 wide flange section utilized as a column. The test specimens
RBS connections have been shown in numerous experimental each consisted of a one-sided RBS connection to the column 共i.e.,
tests to exhibit ductile behavior under inelastic cyclic loading. only one beam was connected to the column兲. The section sizes
These tests include those performed by Chen et al. 共1996兲; Engel- used in the test matrix for the column included a W27⫻ 146 and
hardt et al. 共1997兲; Plumier 共1997兲; Tremblay et al. 共1997兲; Tsai et W27⫻ 194. It was observed during the tests that the column de-
al. 共1999兲; and Jones et al. 共2002兲. The majority of past testing veloped twisting. One of the three specimens in the test matrix
共which had a W27⫻ 194 column兲 was observed to develop a
1
Structural Engineer, Skidmore, Owings and Merrill LLP, One Front sudden fracture in the column k-area within the panel zone while
St., San Francisco, CA 94111. subjecting the specimen to a second cycle of 4% story drift. While
2 all three specimens achieved a 4% story drift for at least one cycle
Bruce G. Johnston Professor of Structural Engineering, Dept. of Civil
Engineering, Lehigh Univ., 117 ATLSS Dr., Bldg H, Bethlehem, PA without fracture or severe strength degradation below 80% of the
18015. E-mail: [email protected] nominal capacity 关which is the current requirements in the AISC
Note. Associate Editor: Gregory A. MacRae. Discussion open until Seismic provisions 共AISC 2002兲 for qualifying a connection detail
August 1, 2006. Separate discussions must be submitted for individual for seismic use兴, the column twist and fracture raised concerns.
papers. To extend the closing date by one month, a written request must
The twisting is caused by an eccentric beam flange force that
be filed with the ASCE Managing Editor. The manuscript for this paper
was submitted for review and possible publication on November 22, develops as a result of a lateral-torsional buckling of the beam in
2004; approved on June 3, 2005. This paper is part of the Journal of the RBS, as illustrated in Fig. 1共b兲. In Fig. 1共b兲 F1 and F2 repre-
Structural Engineering, Vol. 132, No. 3, March 1, 2006. ©ASCE, ISSN sent the flange compression forces developed in the top and
0733-9445/2006/3-358–367/$25.00. bottom flanges of the right- and left-hand beams, respectively.

358 / JOURNAL OF STRUCTURAL ENGINEERING © ASCE / MARCH 2006

 Grade 50 steel 共which has a nominal yield stress of 345 MPa兲 was
used for the beams, column, doubler plates, and continuity plates.
The member section sizes for the models in the analysis matrix
were based on representing the range of anticipated member sec-
tion sizes of the test specimens. The beam section size was se-
lected for each model to ensure a weak beam-strong column con-
figuration in order to comply with the AISC Seismic provisions
共AISC 2002兲.
The prototype connection for the study is similar to that shown
in Fig. 1共a兲, where the beam flanges are joined to the column
flange with complete joint penetration 共CJP兲 field groove welds
using the flux core arc welding procedure and E70T-6 electrodes.
In all RBS-type connections in this study the beam flange width
was reduced by 50% in the RBS with a circular cut over a
686 mm length. The E70T-6 weld metal has a typical toughness
of 27– 95 J at −28°C 共Ricles et al. 2000兲, and can satisfy the
AISC Seismic provisions 共AISC 2002兲 of 27 J at −28°C and 54 J
at 21°C 共Ricles et al. 2004兲. The bottom backing bar is removed,
back-gouged, and reinforced with a notch tough fillet weld, and
the top backing bar is left in place and reinforced with a closure
fillet weld. The beam web is connected to the column flange using
a CJP groove weld with a supplemental fillet weld around the
edges of the shear tab. The shear tab serves as an erection device
Fig. 1. 共a兲 RBS connection details and 共b兲 RBS local buckling and during construction 共using erection bolts兲 and as a backup bar for
lateral beam flange movement the web CJP groove weld. Continuity plates are located on both
sides of the column web. The composite floor slab 共not shown in
Fig. 1兲 was 133 mm thick; the prototype connection was assumed
Deeper wide flange sections for columns tend to have thinner to be in a perimeter SMF, hence a 1,524 mm wide floor slab
flanges and a thinner web, resulting in a reduced torsional resis- extended 1,219 mm and 305 mm from the center of the beam
tance 共combined St. Venant and Warping torsion兲 that can lead to cross section 共as it did in the setup for the experimental study兲.
greater column twist and warping stresses 共Seaburg and Carter While this did not model exactly the conditions in a prototype
1997兲. Consequently, there is a concern that the use of an RBS SMF 共i.e., the floor slab did not extend to an adjacent parallel
connection to a deeper wide flange column in a SMF can lead to interior beam兲, it did provide restraint to the beam against lateral
inferior seismic performance because of it being susceptible to torsional buckling and insight into the effectiveness of the floor
torsional loading from the beams. In this paper, a deep column is slab when compared to analysis models without the floor slab.
considered to be a column with a depth that exceeds that of a The effects of the parameters in the study were investigated by
W14 wide flange section. There is a lack of experimental data for varying the details in a model to create different finite element
the seismic performance of RBS connections to wide flange col- models. This included varying the section size of the beams and
umn that is deeper than a W14 wide flange section to address this column 共to investigate the effect of a different column depth and
concern. beam web slenderness, respectively兲; connection type; composite
The lack of knowledge of the performance of RBS beam-to- floor slab; and panel zone strength. The analysis matrix for the
deep column connections under seismic loading led to a study on study is summarized in Table 1.
this topic. This study involved both analytical and experimental The three-dimensional finite element model is shown in Fig. 2,
investigations. The objective of the study was to evaluate the where a global model and a submodel are shown. The floor slab is
effects of the following parameters on the inelastic behavior of an not shown in the global model in Fig. 2共a兲. The models include
RBS connection to a deep wide flange column: 共1兲 connection details such as all welds, shear tab, RBS cut, bolt holes and bolts,
type; 共2兲 column section; 共3兲 composite floor slab; 共4兲 panel zone continuity plates, doubler plates, top backing bars 共the bottom
strength; and 共5兲 beam web slenderness. The analytical investiga- backing bar is removed in the connection兲, weld access holes, and
tion is presented in this paper. composite floor slab. The global model consisted of a cruciform
interior connection subassembly with a beam attached to both
sides of the column with respect to the column’s strong axis. The
Finite Element Model column height between the bottom pin to the top loading point
and length between the far ends of the beams are 3,962 and
8,992 mm, respectively. The inflection points in the prototype
Finite Element Model Details
SMF were assumed to exist at the ends of the beams and column
The analytical study involved developing finite element models in the model. The boundary conditions for a global model con-
of connection subassemblies for the purpose of evaluating the sisted of roller boundary conditions at the end of each beam and
effect of the above parameters on connection behavior. Three- a pin boundary condition at the bottom of the column, as shown
dimensional nonlinear finite element models of the connection in Fig. 2共a兲. The top of the column was subjected to in-plane
subassemblies were created using the ABAQUS computer pro- lateral displacement, where an actuator was attached to the col-
gram 共ABAQUS 2001兲. The geometry 共i.e., member span lengths兲 umn through a clevis. Out-of-plane movement of the beams and
and boundary conditions of the connection subassemblies were the column was restrained at their flanges near the ends of the
based on a test setup used in the companion experimental study. beams, and at the top and bottom of the column to simulate the

JOURNAL OF STRUCTURAL ENGINEERING © ASCE / MARCH 2006 / 359

Table 1. Analysis Matrix
Connection Floor ␴col,flg CU
Case type Column Beam Rv / V pz slab
␴col,flg FEA

1 RBS W14⫻ 398 W36⫻ 150 1.22 No 1.10

2 WUF W14⫻ 398 W36⫻ 150 1.36 No —
3 RBS W36⫻ 230 W36⫻ 150 1.09 No 1.74
4 WUF W36⫻ 230 W36⫻ 150 1.09 No —
5 RBS W27⫻ 194 W36⫻ 150 1.05 No 2.13
6 RBS W36⫻ 230 W36⫻ 150 0.83 No 2.89
7 RBS W36⫻ 230 W36⫻ 150 1.34 No 1.67
8 RBS W27⫻ 146 W30⫻ 108 1.05 No 1.37
9 RBS W27⫻ 194 W36⫻ 150 1.35 No 1.38
10 RBS W14⫻ 398 W36⫻ 150 1.22 Yes 1.03
11 RBS W36⫻ 230 W36⫻ 150 1.09 Yes 1.24
12 RBS W27⫻ 194 W36⫻ 150 1.05 Yes 1.53
13 RBS W27⫻ 194 W36⫻ 150 0.65 No 3.35
14 RBS W27⫻ 194 W36⫻ 150 1.25 No 2.23

lateral-torsional bracing for the members in the test setup of the ment constraints were used in the vertical direction to avoid ver-
experimental study. For models with a composite floor slab, trans- tical separation between the beams and the floor slab. The details
verse floor beams at 3,048 mm 共for models with W36⫻ 150 of the shear stud modeling procedure are reported in Lee and Lu
beams兲 and 2,743 mm 共for models with W30⫻ 108 beams兲 spac- 共1989兲.
ing from the column centerline braced the main beams from one The analyses in the parametric study were conducted by ap-
side, which is similar to how the test specimens were braced in plying a prescribed displacement history to the top of the column
the experimental study. The distances of 3,048 and 2,743 mm of the global model 关see Fig. 2共a兲兴. Both a monotonic and cyclic
were based on the AISC Seismic provisions 共AISC 2002兲 for brac- displacement were applied, respectively, where the latter was
ing requirements for a W36⫻ 150 beam and a W30⫻ 108 beam, based on the story drift history given in Appendix S: “Qualifying
respectively. In the global model, the beams and the column, as cyclic tests of beam-to-column and link-to-column connections”
well as the connection attachments were modeled using a four- of the AISC Seismic provisions 共AISC 2002兲.
node shell element with standard integration 共element S4 in the The submodel shown in Fig. 2共b兲 was utilized to perform a
ABAQUS element library兲. The composite floor slab was modeled local analysis to obtain increased accuracy in the computed stress-
using a four-node shell element 共type S4 in the ABAQUS element strain state in the beam tension flange region of the connection.
library兲 with an elastic-plastic compression-only stress material, The submodels were composed of eight-node brick elements with
while a two-node three-dimensional cubic formulation beam ele- standard integration 共element C3D8 in the ABAQUS element li-
ment 共type B33 in the ABAQUS element library兲 was used to brary兲. The displacement results of the global model were used
model the transverse floor beams. This method of modeling of the for the boundary conditions around the perimeter boundary of the
composite floor slab was necessary since ABAQUS does not have submodel.
a concrete material model that enables cyclic analysis to be per- Geometric and material nonlinearities were included in both
formed involving concrete. Shear studs, which affix the compos- models, where the geometric nonlinearities were accounted for by
ite floor slab to the main beams as well as the transverse floor using a finite strain, large-displacement formulation. A metal plas-
beams, were modeled at a spacing of 305 mm using an inelastic ticity model in the ABAQUS material library based on the von
spring element 共type SPRING2 in the ABAQUS element library兲 Mises yield criterion with combined isotropic and kinematic
to account for their lateral flexibility. The spring elements were strain-hardening was used to account for material nonlinearities in
put in both horizontal directions of the floor slab, and displace- the members and welds. The monotonic stress-strain curves for
the metallic materials of the models are shown in Fig. 3. The floor

Fig. 2. Finite element model 共slab not shown for clarity兲 Fig. 3. Stress-strain relationships used in finite element models

360 / JOURNAL OF STRUCTURAL ENGINEERING © ASCE / MARCH 2006

slab was assumed to have 27 MPa compressive strength concrete.
Elastic and inelastic convergence studies were performed to
evaluate and arrive at the final mesh for the finite element models.
The effects of the stiffness of the column bracing restraining
column lateral and torsional movement at the brace points were
investigated in order to ensure that it did not influence the results
of the parametric study, whereby they would not be representative
of prototype member behavior. The investigation indicated that a
lack of torsional bracing stiffness can result in a significant reduc-
tion in strength and ductility of the connection subassembly due
to column instability. Excessively stiff torsional bracing did not
significantly affect the results. Therefore the use of roller bound- Fig. 4. Results of validation study for finite element model, SPEC-2
ary conditions to laterally and torsionally brace the ends of the 共data adapted from Ricles et al. 2004兲
column in the finite element models 关see Fig. 2共a兲兴 was reason-
able. Complete details of the modeling procedures mentioned
above, the mesh convergence and bracing stiffness studies, and
how the geometric and material nonlinearities were implemented
into the models are reported in Ricles et al. 共2004兲.
␧p = 冑 2 p p
␧ ␧
3 ij ij
共4兲

Unstable crack propagation was not addressed in the study. where ␧ijp = plastic strain components. The ratio of hydrostatic
Rather, the study was concerned with the potential of ductile frac- pressure-to-von Mises stress in the denominator of Eq. 共1兲 is
ture through the development of stress and strain states that would known as the triaxiality ratio 共TR兲. High triaxiality can cause a
facilitate fracture if a flaw or other irregularity exists, as well as large reduction in the rupture strain of a material, thereby limiting
the effects of the parameters on the cyclic performance of the its ductility 共Lemaitre 1996兲. Hence locations in a connection
connection. The evaluation of the potential for ductile fracture with larger negative values for TR lead to larger values of the RI
was based on the relative values of a rupture index 共RI兲 共ex- 共tension stress produces a negative hydrostatic pressure兲, which
plained below兲, where the stress-strain state for the RI was deter- imply a greater potential for fracture. The locations determined to
mined using the submodels shown in Fig. 2 subjected to the have the greatest fracture potential 共Ricles et al. 2004兲 include:
monotonic displacement history. Computational limitations of the the heat affected zone 共HAZ兲 at the interface of the CJP weld
software used to develop the finite element models did not enable metal and base metal of the outside surface of the beam tension
accumulated plastic strain under cyclic loading to be determined flange; the weld access hole region; the end of the beam web CJP
in the submodel. As in prior analytical studies on welded moment groove weld closest to the beam tension flange; and the beam
connections 共El-Tawil et al. 1998; Mao et al. 2001兲, and con- tension flange CJP groove weld. The fracture potential was evalu-
firmed by experimental studies by Ricles et al. 共2000兲, it was ated at these locations, in addition to the column k-area and
assumed that the conclusions drawn from the submodel analysis continuity plates, at a story drift of 4% 关which, as noted previ-
results are qualitatively applicable to cyclic loading conditions. ously, is the current drift requirement in the AISC Specifications
The evaluation of the effects of the parameters on cyclic perfor- 共AISC 2002兲 for connection qualification兴.
mance, including cyclic strength and column twist, involved the
application of the cyclic displacement history to a model.
Finite Element Model Verification

Response Indices The finite element global model was verified by comparing the
measured response of various connection specimens tested in
The RI was computed at different locations of the connection prior research by Ricles et al. 共2000兲; and Jones et al. 共2002兲 with
using the finite element analysis results. The RI was motivated by the response predicted by the finite element models. In addition to
the research of Hancock and Mackenzie 共1976兲 on the equivalent these test specimens, a test specimen with an RBS connection to
plastic rupture strain of steel for different conditions of triaxiality a deep column tested more recently by Ricles et al. 共2004兲 was
and is defined as also modeled and the response prediction compared with the mea-
␧ p/␧y sured response of the specimen during the test. The specimen

冉 冊
RI = 共1兲 共SPEC-2兲 was a cruciform-shaped specimen having the same
p
exp 1.5 spans and boundary conditions as that of the finite element model
q described above and shown in Fig. 2, with the same connection
where ␧ p, ␧y, p, and q = equivalent plastic strain, yield strain, hy- details 共shown in Fig. 1兲 and floor slab details for the prototype
drostatic pressure, and von Mises stress, respectively, with connection. Note that the floor slab is not shown in Fig. 1 for
clarity. The beams and column of SPEC-2 were W36⫻ 150 and
1 1 W27⫻ 194 sections, respectively, fabricated from Grade 50 steel.
p = − tr共␴ij兲 = − ␴ii 共2兲
3 3 The specimen was tested using the cyclic inelastic loading proto-
col defined in Appendix S of the AISC Seismic provisions 共AISC
and 2002兲.

q= 冑 3
2
SijSij 共3兲
The lateral load-story drift response predicted by the model is
given in Fig. 4 共where it is labeled as ABAQUS兲. The predicted
response is shown to be in good agreement with the experimental
In Eqs. 共2兲 and 共3兲, ␴ij and Sij = Cauchy stress and deviatoric results 共labeled as Test兲. The model was determined to properly
stress components, respectively. The equivalent plastic strain ␧ p is predict the occurrence of panel zone yielding, beam yielding, cy-
given as clic beam web and flange local buckling in the RBS, and strength

JOURNAL OF STRUCTURAL ENGINEERING © ASCE / MARCH 2006 / 361

deterioration of the specimen during the test. Good agreement
was also found to exist between the measured and predicted
strains, where at 4% story drift the experimental and predicted
strains in the beam flanges near the column face were within a 5%
difference of each other. The comparisons between the test and
finite element analysis results indicate that the finite element mod-
eling procedures produce predictions which should lead to accu-
rate results in the parametric study.

Analysis Matrix
Fig. 5. Column total flange maximum stress predicted by Chi and
The analysis matrix for the finite element parametric study con- Uang 共2002兲 versus section weight
sists of the 14 cases summarized in Table 1. For each analysis
case the following are identified in Table 1: connection type 关RBS
and welded unreinforced flange 共WUF兲 connections were stud- W27⫻ 194 column section, with the beam section ranging in size
ied兴; column section size; beam section size; panel zone strength from a W36⫻ 135 to a W36⫻ 210.
ratio Rv / V pz; and whether a composite floor slab existed in the
model. All cases consisted of a two-sided connection 共as shown in
Fig. 1兲, except for Case 9 which was a one-sided connection. Analysis Results
Rv = panel zone shear strength based on the AISC Seismic provi-
sions 共AISC 2002兲, as given below in Eq. 共5兲, while V pz = shear
force in the panel zone corresponding to the maximum expected Effect of Connection Type
plastic moment M pr developing at the middle of the RBS 共or at As noted above, to examine the effect of connection type on
two-thirds of the beam depth from the column face for a WUF behavior, two types of connections, namely WUF and RBS, were
connection兲 based on Eq. 共6兲: investigated. The results from Cases 1–4 are used to evaluate the
effect of connection type. Cases 1 and 2 in Table 1 both include a

冋
Rv = 0.6Fydct p 1 +
3bcf t2cf
d bd ct p
册 共5兲
W14⫻ 398 section for the column, where the former has a RBS
connection and the latter a WUF connection. Cases 3 and 4 both
have a deeper column section 共W36⫻ 230兲, with Case 3 having a
RBS connection and Case 4 a WUF connection. The elastic tor-
M pr = C prRyZRBSFy 共6兲 sional stiffness of the W14⫻ 398 is 1.23 times greater than that of
the W36⫻ 230 section. All of these cases had a W36⫻ 150 sec-
where in Eq. 共5兲 Fy; dc; t p; bcf ; tcf ; and db = panel zone yield tion for the beams and no floor slab 共the effect of a floor slab is
strength, column depth; combined total thickness of the column discussed later兲, with either a strong 共Cases 1 and 2兲 or balanced
web and doubler plate; column flange width; column flange thick- panel zone strength 共Cases 3 and 4兲 relative to the beam flexural
ness; and beam depth, respectively, and in Eq. 共6兲 C pr; Ry; ZRBS; capacity.
and Fy = strain hardening factor to account for peak connection The maximum values of the RI at the various critical regions
strength that is equal to 1.1 for Grade 50 steel 共FEMA 2000; in the connections for these cases are shown in Fig. 6 at 4% story
AISC 2002兲; ratio of expected yield strength of beam to minimum drift. The results in Fig. 6 indicate that the RBS connection has a
specified yield strength 共AISC 2002兲; plastic section modulus of lower value for the RI, and thus lower fracture potential, com-
the beam section at the RBS; and beam nominal yield strength, pared to a WUF connection to a similar column section. The
respectively. cause for the higher value of the RI in the WUF connection is due
The member section sizes were selected in order to have a to the larger plastic strains that develop in the connection region
varying degree of column elastic torsional stiffness K␪ 共including near the column face. The larger plastic strains are the result of
warping and St. Venant torsion兲 while also satisfying the weak the larger beam moment at the column face in the WUF connec-
beam-strong column criteria in the AISC Seismic provisions tion 共average value of 1.22M pn兲. The maximum beam moment at
共AISC 2002兲. The column sizes in the analysis matrix in-
cluded a W14⫻ 398 共K␪ = 3,940 kN m / rad兲, a W27⫻ 146
共K␪ = 900 kN m / rad兲, a W27⫻ 194 共K␪ = 1,400 kN m / rad兲, and a
W36⫻ 230 共K␪ = 3,200 kN m / rad兲. The beam section was a
W36⫻ 150 for all cases, except for Case 8. Case 8 had a W30
⫻ 108 beam 共and W27⫻ 146 column兲. The total maximum elastic
normal stress 共due to bending and warping torsion兲 in the column
flanges of these sections and others according to Chi and Uang
共2002兲 is shown in Fig. 5. The lighter sections are shown in
Fig. 5 to develop a larger predicted normal stress in their flanges.
The effects of beam web slenderness were investigated by
considering 32 additional cases 共not included in Table 1兲 involv-
ing a two-sided RBS connection, both with and without a com-
posite floor slab. Eighteen of these cases had a W36⫻ 230
column section, with the beam section ranging in size from a
W36⫻ 135 to a W36⫻ 256. The remaining 14 cases had a Fig. 6. Effect of connection type on rupture index at 4% story drift

362 / JOURNAL OF STRUCTURAL ENGINEERING © ASCE / MARCH 2006

Fig. 7. Effect of connection type on cyclic lateral load-drift response Fig. 9. Effect of column section size on rupture index

the column face in the RBS connection is an average value of beams. However, as noted above, the potential for fracture in the
0.95M pn. M pn = beam nominal plastic flexural capacity. For each WUF connection to the W36⫻ 230 column is greater than that of
case, the largest fracture potential is at the end of the beam web- the RBS connection to the same column size due to the larger
to-column flange CJP groove weld. The larger value for the RI at moment developed in the beam and demand imposed on the WUF
this location is associated with a large amount of local plastic connection. For the W14⫻ 398 column, there is a large increase
strain that develops in the connection region. in the column twist 共from 0.005 to 0.018 rad兲 for the RBS con-
The cyclic lateral load-drift response of the cases with the nection model 共Case 1兲 between the story drift of 3 and 4% due to
deeper column 共i.e., Cases 3 and 4, having a W36⫻ 230 column兲 the severe local buckling in the beam web and flanges at the RBS.
is shown in Fig. 7. The response shows that the WUF connection The column twist at 6% story drift was 0.035 rad for Case 1 and
model has a greater capacity than the RBS connection, as 0.01 rad for Case 2, where the latter has the W14⫻ 398 column
expected. Both appear to develop their respective maximum and WUF connection. The WUF connection model with the
capacity at about 3% story drift, where a cyclic deterioration in W14⫻ 398 column does not show a significant increase in twist
capacity occurred in both models due to local buckling in the since the beam local buckling was not as severe.
flanges and web of the beam. The WUF connection model
appears to suffer a greater amount of loss of capacity compared to Effect of Column Section
the RBS connection model. A summary of the column twist at
story drifts up to 6% for Cases 1–4 is shown in Fig. 8 共Case 5 is To evaluate the effect of the column section size on the perfor-
also included in Fig. 8, and discussed later兲. The cases with a mance of an RBS connection, five analysis cases involving
deeper column appear to develop a greater amount of column different column sizes were performed. These cases include
twist, where the model with the RBS connection and Cases 1, 3, 5, 8, and 9 in Table 1, which had column section sizes
W36⫻ 230 column 共Case 3兲 develops the largest amount of twist of W14⫻ 398, W36⫻ 230, W27⫻ 194, W27⫻ 146, and
among Cases 1–4. Column twist is defined as the twist in the W27⫻ 194, respectively. As noted above, Case 9 is a one-sided
column at the top beam flange location at the connection. At 3% RBS connection attached to a W27⫻ 194 column. Except for
drift the RBS connection to the deeper column 共Case 3兲 shows a Case 9, for each of these cases the panel zone was designed in
column twist of about 0.012 rad, which is almost three times that accordance with ASIC Seismic provisions 共2002兲, resulting in a
of the WUF connection model with the same W36⫻ 230 column strong 共Case 1兲 or balanced panel zone strength condition 共Cases
共Case 4兲. At 4% story drift the column twist in the models for 3, 5, and 8兲. For Case 9, the one-sided beam configuration led to
these two cases is almost equal, where the twist in the RBS con- a stronger panel zone condition. No composite floor slab existed
nection model has increased to about 0.021 rad. At 6% story drift among these cases. All cases except for Case 8 had W36⫻ 150
the twist in the column for Cases 3 and 4 was 0.05 and 0.04 rad, beams; Case 8 had W30⫻ 108 beams to maintain a weak beam-
respectively. The increase in twist in both of these cases is due to strong column configuration.
the significant local web and flange buckling that develops in the The values for the RI at the critical locations in the connection
region at 4% story drift are shown in Fig. 9. The greatest fracture
potential is again at the end of the beam web-to-column flange
CJP groove weld. The values of the RI are smallest in the column
k-area and continuity plates. Case 5 共W27⫻ 194 column兲 is seen
to have the largest value for the RI, which is due to the larger
plastic strain that develops in the connection. Although the
W27⫻ 146 column section for Case 8 has a smaller section
modulus and torsional stiffness K␪ than the W27⫻ 194 column
section, the RBS connection to the W27⫻ 146 involves a smaller
beam section 共W30⫻ 108兲. Consequently, the strength demand on
the connection is less in Case 8, as are the local plastic strains in
the column flange and at the beam-column interface. The one-
sided RBS connection, Case 9, is seen to have a lower value for
the RI compared to the two-sided RBS connection with similar
Fig. 8. Effect of connection type and column section on maximum section sizes 共i.e., Case 5兲 due to the lower demand from only one
cyclic column twist beam.

JOURNAL OF STRUCTURAL ENGINEERING © ASCE / MARCH 2006 / 363

 Fig. 11. Effect of composite floor slab on cyclic lateral load-drift
Fig. 10. Effect of floor slab on rupture index response, W27⫻ 194 column

The results in Fig. 9 indicate that the fracture potential can be having a W14⫻ 398 column 共Case 10兲 due to an increase in both
larger in an RBS connection to a shallower column. It was found the local plastic strain 共by a factor of about 1.75兲 and the triaxi-
that the fracture potential of an RBS connection is highly depen- ality ratio 共TR兲 共by 25%兲. For cases with a deeper column, the
dent on the plastic strain that develops at the beam-column inter- increase in the RI at this location when adding the slab is less, and
face. For a given beam section size, a deeper column with a larger equal to 70% for the W36⫻ 230 column 共Case 3 versus Case 11兲
section modulus has lower stresses, leading to smaller local plas- and about 9% for the W27⫻ 194 column 共Case 5 versus Case 12兲.
tic deformations at the beam-column interface. Smaller beams can In these latter cases there is an increase in the local plastic strain
also reduce the fracture potential of the connection by imposing at the column-beam interface with the TR remaining relatively
smaller forces on the connection and column, leading to reduced constant when adding the floor slab.
plastic deformations in the connection. As noted, Fig. 10 shows that the RI increases in the beam
A summary of the maximum column twist-story drift response bottom flange when a floor slab is added to the model 共see results
of the models for Cases 1, 3, and 5 is included in Fig. 8. The for beam flange HAZ and beam flange weld兲. This increase
model with the W27⫻ 194 column 共Case 5兲 begins to develop ranged from 15 to 25% for the various cases, and was associated
considerably more column twist beyond the story drift of 3%. At with an increase in the plastic strain by a factor of about 2.0 in the
3% story drift an appreciable amount of local web and flange beam bottom flange when adding a floor slab.
buckling develops in the RBS that leads to the onset of strength The same analysis cases used for investigating the effects of a
degradation. The out-of-plane movement of the beam compres- composite floor slab on local performance were used to investi-
sion flanges in the RBS and the reduced torsional stiffness of the gate the global performance of an RBS connection. The lateral
W27⫻ 194 column 共a W27⫻ 194 section has a K␪ that is 2.8 and load-story drift hysteretic response for Cases 5 and 12 is given in
2.3 times less than that of a W14⫻ 398 and W36⫻ 230 column, Fig. 11. The floor slab stabilizes the beam in the RBS region,
respectively兲, resulted in a column twist of 0.081 and 0.145 rad in enhancing the cyclic performance of an RBS connection to a deep
the column at a story drift of 4 and 6%, respectively. The column column. As shown in Fig. 11, the floor slab slightly increases the
twist for Cases 1 and 3 are approximately 0.018 and 0.021 rad, strength of the connection, delays the onset of strength degrada-
respectively, at 4% story drift and 0.035 and 0.05 rad, respec- tion, and reduces the amount of strength degradation. It was
tively, at 6% story drift. The cyclic behavior of the models for found that the floor slab has about the same effect on a connection
Cases 1, 3, and 5 were similar. The cyclic lateral load-story drift to a shallower column section 共Cases 1 and 10, W14⫻ 398 col-
response of Case 3 was shown previously in Fig. 7 共Cases 1 and umn兲 as it does on a deeper W27⫻ 194 column section 共Cases 5
5 are not shown兲. The models each had about the same strength and 12兲. The increase in the maximum strength provided by the
and rate of cyclic strength degradation due to local buckling of floor slab in these two cases is 4 and 2%, respectively. The extent
the beam web and flanges in the RBS. of strength deterioration was evaluated by comparing the re-
sponse following local beam buckling that occurred at 3% story
drift. At 4% story drift there is a greater strength of 18 and 16%
Effect of Composite Floor Slab
for Cases 10 and 12, respectively, compared to their correspond-
The effect of a composite floor slab on the ductile fracture poten- ing bare steel models. The enhancement of the connection perfor-
tial of an RBS interior connection was investigated by performing mance 共i.e., reduction in the extent of strength degradation兲 is
the six analysis cases identified in Table 1 as Cases 1, 3, 5, 10, 11, consistent with the findings in the study by Jones et al. 共2002兲 on
and 12. All of these cases have W36⫻ 150 beams. Cases 1 and 10 RBS connections to W14 column sections, who observed that the
have a W14⫻ 398 column without and with a floor slab, respec- slab appears to have a stabilizing effect on the RBS moment
tively, while Cases 3 and 11 have a W36⫻ 230 column without connection, increasing its load and rotation capacity.
and with a floor slab, respectively. Cases 5 and 12 have a Fig. 12 shows the column twist angle at selected story drift
W27⫻ 194 column without and with a floor slab, respectively. levels for Cases 1, 5, 10, and 12. The results show a greater
The results for the values of the RI at the critical locations in column twist for the deeper W27⫻ 194 column 共i.e., Cases 5 and
the connection at 4% story drift are shown in Fig. 10. The floor 12兲 compared to the W14⫻ 398 column 共Cases 1 and 10兲. For the
slab is shown to increase the RI in all corresponding cases, where models without a floor slab, there is significantly more column
the maximum RI value occurs at the beam web-to-column flange twist in the model with the W27⫻ 194 column compared to
CJP groove weld. There is an increase in the RI by a factor of 2 at the W14⫻ 398 column. At 4% story drift the twist in the
this location when adding a floor slab to the RBS connection W27⫻ 194 column and W14⫻ 398 column without a floor slab is

364 / JOURNAL OF STRUCTURAL ENGINEERING © ASCE / MARCH 2006

 A summary of the maximum values of the RI at the critical
locations in the connection is given in Fig. 13 at 4% story drift.
The weak panel zone model 共Case 13兲 has the largest RI value,
which develops at the beam flange CJP weld. The large RI value
is associated with the large plastic strain and triaxiality that de-
velops in the CJP welds due to the shear distortion of the panel
zone. The shear distortion leads to a local bending of the beam
and column flanges adjacent to the panel zone. In the models with
the balanced panel zone 共Case 5兲 and strong panel zone 共Case 14兲
strength conditions the shear distortion of the panel zone is re-
duced. Consequently, the RI value at the beam flange CJP welds
Fig. 12. Effect of composite floor slab on maximum column twist for these cases is 2.5 and 5.0 times less, respectively, than that of
the weak panel zone model. For the balanced panel zone case, the
largest fracture potential is at the end of the beam web-to-column
0.081 and 0.02 rad, respectively. Adding a floor slab appears to flange CJP groove weld. The occurrence of a large local plastic
significantly reduce the column twist, particularly in the model strain developing in the beam flange CJP welds in weaker panel
with the deep column. The column twist reduced from zone connections was also found in studies by Ricles et al. 共2000兲
0.081 to 0.018 rad at 4% story drift when adding a floor slab to on WUF connections to W14 column sections. The results of the
the model with the W27⫻ 194 column. When adding a floor slab current study indicate that RBS connections with a deeper column
to the model with the W14⫻ 398 column, the column twist at 4% also appear to be susceptible to a greater fracture potential when
drift was reduced from 0.02 to 0.005 rad. At 6% story drift the the connection has a weak panel zone.
column twist is reduced from 0.028 to 0.019 rad when adding the The lateral load-story drift hysteretic response for all three of
floor slab to the model with a W14⫻ 398 column, and from these cases is shown plotted in Fig. 14. The hysteretic response
0.145 to 0.033 rad when adding the floor slab to the model with shows the weak panel zone model 共Case 13兲 developing the low-
the W27⫻ 194 column. est strength, but not suffering any degradation in capacity during
the analysis, while the balanced and strong panel zone models
共Cases 5 and 14, respectively兲 develop a deterioration in capacity
Effect of Panel Zone Strength during the analysis due to beam local web and flange buckling,
The effect of panel zone strength on the ductile fracture potential and an out-of-plane movement of the beam compression flange in
of an RBS connection to a deep column was evaluated by com- the RBS. The deterioration in capacity commenced at about 3%
paring the results for Cases 5, 13, and 14. Each of these cases had story drift. The balanced and strong panel zone models show an
a W27⫻ 194 column section and W36⫻ 150 beam section, but no almost identical lateral load-story drift hysteretic response. The
composite floor slab. Other cases in the analysis were also weak panel zone model had yielding and plastic deformations
compared to each other 共e.g., Cases 3, 6, and 7 involving a concentrated in the panel zone, leading to cyclic local buckling in
W36⫻ 230 column兲, resulting in the same conclusions noted the panel zone. Similar lateral load-story drift hysteretic behavior
below for Cases 5, 13, and 14. The cases involving the was observed in analytical studies by Ricles et al. 共2000兲 on WUF
W27⫻ 194 column were found to be more sensitive to the effects connections to a W14 column.
of panel zone strength. The values for the panel zone strength The column twist for all three cases at various story drifts is
ratio Rv / V pz are 1.05, 0.65, and 1.25 for Cases 5, 13, and 14, given in Fig. 15. The weak panel zone model 共Case 13兲 is shown
respectively. These values for Rv / V pz correspond to a balanced to have minimal column twist, with significantly more twist de-
panel zone strength condition 共with respect to the beam flexural veloping in the models with a balanced and strong panel zone
strength兲, weak panel zone strength condition, and strong panel 共Cases 5 and 14, respectively兲. The reason for this is because in
zone strength condition. The balanced panel zone strength design the weak panel zone model the inelastic deformations are concen-
is in accordance with the ASIC Seismic provisions 共2002兲. The trated in the panel zone, and the RBS does not develop significant
thickness of the panel zone doubler plate corresponding to these yielding in the beam that would lead to local buckling in the RBS.
strength conditions are 12.5, 0, and 19 mm, respectively.

Fig. 13. Effect of panel zone strength on rupture index, W27⫻ 194 Fig. 14. Effect of panel zone strength on cyclic lateral load-drift
column response, W27⫻ 194 column

JOURNAL OF STRUCTURAL ENGINEERING © ASCE / MARCH 2006 / 365

Fig. 15. Effect of panel zone strength on maximum column twist, Fig. 17. Effect of beam web slenderness on maximum column twist
W27⫻ 194 column at 4% cyclic story drift

In the other two models significant inelastic deformations develop same beam web slenderness. The reason for this is associated
in the RBS, leading to local and lateral buckling in the RBS and with the lower column torsional flexibility and greater amount of
torsional loading and twisting of the column. local column flange yielding that occurs in the models involving
the W27⫻ 194 column. The addition of the floor slab is shown to
Effect of Beam Web Slenderness significantly decrease the transverse movement of the beam
flange at the RBS, where the transverse movement is less than the
The effect of beam web slenderness on the behavior of an RBS value of 0.2b f 共where b f = beam flange width兲, suggested by Chi
connection to a deep column was investigated by considering and Uang 共2002兲 in their design procedure for RBS connections
the 32 analyses mentioned previously. The parameters in the to a deep column.
analysis matrix consisted of beam web slenderness 共W36 sections The column twist developed in the models at 4% story drift is
ranging in weight from 201 to 381 kg/ m兲, column section size plotted against beam web slenderness in Fig. 17. It is apparent in
共W36⫻ 230 and W27⫻ 194兲, and a floor slab. The proportioning Fig. 17 that the increase in beam flange area offsets the reduction
of the beam and column sizes in all of the analysis cases satisfied in beam web slenderness, leading to an increase in column twist.
the weak beam-strong column criteria in the AISC Seismic pro- Because the W27⫻ 194 is more flexible in torsion compared to
visions 共2002兲. the W36⫻ 230 section, models with the W27⫻ 194 column have
The out-of-plane movement of the beam bottom flange at the a greater amount of column twist. The composite floor slab is
RBS is plotted against the beam web slenderness in Fig. 16 at 4% shown to cause a significant reduction in the column twist, with
story drift. The results from the analysis show that the movement an almost constant value compared to the results without a floor
of the beam flange increases with a reduction in beam web slen- slab. Fig. 17 shows that the column twist for cases involving the
derness for the models without a floor slab. This is due to the W27⫻ 194 section is about twice that involving a W36⫻ 230
effect of an increase in beam flange force. The area of the beam section when the floor slab is present.
flange increases in wide flange sections 共AISC 2001兲 as the beam
section becomes heavier and the beam web slenderness is re-
duced. A larger beam flange area results in a larger flange com- Comparison of Finite Element Analysis with Chi and
pressive force, leading to a greater amount of torque imposed to Uang Stress Prediction
the column by the RBS connection. Models with a W27⫻ 194 A comparison of the stress at the tip of the column flange
column are shown in Fig. 16 to develop more beam flange move- predicted by the finite element models of RBS connections at
ment in the RBS than those with a W36⫻ 230 column for the 4% story drift 共␴col,flg FEA兲 with that predicted using the recom-
mendations of Chi and Uang 共2002兲 is given in the last column of
Table 1. The comparison shows that the stress predicted by Chi
and Uang 共␴col,flg CU兲 in all cases is larger than that developed in
the finite element models. The column flange stress predicted by
Chi and Uang is 1.03–3.35 times larger than that in the finite
element model, with the largest overprediction relative to the fi-
nite element model occurring in cases where no floor slab exists
and the panel zone is one with a weak panel zone strength con-
dition. For cases involving a floor slab and a balanced panel zone
strength condition, Chi and Uang’s recommendations lead to a
column flange stress that is 1.03–1.53 times larger than that in the
finite element models.

Summary and Conclusions

Fig. 16. Effect of beam web slenderness on maximum RBS Parametric studies were conducted in order to evaluate the seis-
transverse movement at 4% cyclic story drift mic performance of an RBS connection to a deep wide flange

366 / JOURNAL OF STRUCTURAL ENGINEERING © ASCE / MARCH 2006

column and to evaluate the effects that various parameters have nia Infrastructure Technology Alliance 共PITA兲 program. The sup-
on connection performance. Based on the analytical study, the port provided by these funding agencies is greatly appreciated.
following main conclusions are noted.
1. The ductile fracture potential and column twist in an RBS
connection is dependant on the section modulus and torsional References
stiffness K␪ of the column section, respectively. A column
with a smaller section modulus has larger stresses and a ABAQUS user’s manual, Ver 6.1. 共2001兲. Hibbit, Karlsson & Sorenson,
greater amount of plastic strain develop in the column flange, Pawtucket, R.I.
which leads to a higher ductile fracture potential. Thus a AISC. 共2001兲. Manual of steel construction-load and resistance factor
deep column with a section modulus that is larger than that design, 3rd Ed., Chicago.
of a shallower column, such as a W12 or W14, will have a AISC. 共2002兲. Seismic provisions for structural steel buildings, Chicago.
smaller potential for fracture. The column with the largest K␪ Chen, S.-J., Yeh, C. H., and Chu, J. M. 共1996兲. “Ductile steel beam-to-
will have the smallest amount of column twist. column connections for seismic resistance.” J. Struct. Eng., 122共11兲,
2. RBS connections tend to have a reduced ductile fracture po- 1292–1299.
Chi, B., and Uang, C.-M. 共2002兲. “Cyclic response and design recom-
tential compared to a WUF connection due to the reduced
mendations of reduced beam section moment connections with deep
amount of plastic strain developed in the vicinity of the RBS columns.” J. Struct. Eng., 128共4兲, 464–473.
connection region. The twist in deeper wide flange columns El-Tawil, S., Mikesell, T., Vidarsson, E., and Kunnath, S. 共1998兲.
with an RBS connection was comparable to that of a WUF “Strength and ductility of FR welded-bolted connections.” Rep. No.
connection at story drifts of 4% and more. SAC/BD-98/01, SAC Joint Venture, Sacramento, Calif.
3. A composite floor slab can significantly reduce the lateral Engelhardt, M. D. 共1999兲. “The 1999 T. R. Higgins lecture: Design of
displacement of the beam bottom flange in the RBS and the reduced beam section moment connections.” Proc.: 1999 North
amount of twist developed in the column. The slab appears to American Steel Construction Conf., American Institute of Steel
be effective in reducing the twist in deeper columns attached Construction, Toronto, Canada, 1-1–1-29.
to an RBS connection, enabling the cyclic strength of the Engelhardt, M. D., Winneberger, T., Zekany, A. J., and Potyraj, T. J.
beam with an RBS connection to be better sustained. As a 共1997兲. “Experimental investigation of dogbone moment connec-
result, a larger amount of plastic strain develops in the con- tions.” Proc., 1997 National Steel Construction Conf., AISC, Chicago.
nection when a composite floor slab is present, which in- Federal Emergency Management Agency 共FEMA. 共2000兲. “Recom-
mended seismic design criteria for new steel moment-frame build-
creases the fracture potential of the RBS connection. RBS
ings.” Rep. No. FEMA 350, Washington, D.C.
connections with larger beam web slenderness cause the
Hancock, J. W., and MacKenzie, A. C. 共1976兲. “On the mechanisms of
amount of column twist to increase. The effect of beam web ductile failure in high-strength steels subjected to multi-axial stress-
slenderness is significantly reduced when a composite floor states.” J. Mech. Phys. Solids, 24, 147–169.
slab exists. Jones, S. L., Fry, G. T., and Engelhardt, M. D. 共2002兲. “Experimental
4. A weak panel zone in a deep column RBS connection will evaluation of cyclically loaded reduced beam section moment connec-
not develop as much column twist and strength degradation tions.” J. Struct. Eng., 128共4兲, 441–451.
as a connection with a stronger panel zone. However, a Lee, S., and Lu, L. 共1989兲. “Cyclic tests of full-scale composite joint
weaker panel zone can significantly increase the potential for subassemblages.” J. Struct. Eng., 115共8兲, 1977–1998.
ductile fracture of the connection. It is recommended that Lemaitre, J. 共1996兲. A course on damage mechanics, Springer, Berlin.
connections be designed with a balanced panel zone strength Mao, C., Ricles, J., Lu., L. W., and Fisher, J. 共2001兲. “Effect of local
condition. details on ductility of welded moment connections.” J. Struct. Eng.,
5. The effects of axial column load were not presented in this 127共9兲, 1036–1044.
Plumier, A. 共1997兲. “The dogbone: Back to the future.” Eng. J., 34共2兲,
paper, but were investigated in the study. It was found that
61–67.
axial load has only a slight effect on connection behavior, Ricles, J. M., Mao, C., Lu, L.-W., and Fisher, J. W. 共2000兲. “Development
including both local fracture potential and global behavior. and evaluation of improved details for ductile welded unreinforced
Axial load on a perimeter moment frame column does not flange connections.” Rep. No. SAC/BD-00/24, SAC Joint Venture,
significantly degrade the connection performance when the Sacramento, Calif.
axial stress due to axial load is less than 20% of the column Ricles, J., Zhang, X., Lu, L. W., and Fisher, J. W. 共2004兲. “Development
material yield strength. of seismic guidelines for deep-column steel moment connections.”
Results from the experimental program involving specimens Rep. No. 4–13, ATLSS Engineering Research Center, Lehigh Univ.,
with a floor slab 共Ricles et al. 2004兲 agree with the analytical Bethlehem, Pa.
results reported in this paper. The experimental results indicate Seaburg, P., and Carter, C. 共1997兲. “Torsional analysis of structural steel
that a floor slab is effective in reducing column twist in a RBS members.” Steel design guide series 9, AISC Chicago.
connection to a deep column. Furthermore, the test results show Tremblay, R., Tchebotarev, N., and Filiatrault, A. 共1997兲. “Seismic per-
formance of RBS connections for steel moment resisting frames:
that column sections with a larger section modulus reduce the
Influence of loading rate and floor slab.” Proc., Stessa ’97, Kyoto,
plastic strain in the column flange.
Japan, 664–671.
Tsai, K. C., Chen, W. Z., and Lin, K. C. 共1999兲. “Experimental responses
of steel reduced beam section to weak panel zone moment connec-
Acknowledgments tions.” Proc., Workshop on Design Technologies of Earthquake-
Resistant Moment-Resisting Connections in Steel Buildings, National
The research reported herein was supported by a grant from the Taiwan Univ., Taipei, Taiwan.
American Institute of Steel Construction 共Mr. Tom Schlafly Uang, C.-M., and Fan, C.-C. 共2001兲. “Cyclic stability criteria for steel
program manager兲 and from the Pennsylvania Department of moment connections with reduced beam section.” J. Struct. Eng.,
Community and Economic Development through the Pennsylva- 127共9兲, 1021–1027.

JOURNAL OF STRUCTURAL ENGINEERING © ASCE / MARCH 2006 / 367