1

COLUMN STIFFENER DETAILING FOR NON-SEISMIC AND SEISMIC DESIGN

Jerome F. Hajjar is an Associate Professor in the Department of Civil Engineering at the Uni-versity of Minnesota. He is on the AISC Specification Task Com-mittees on Composite Construction, Stability, and Loads, Analysis, and Systems; he is leading the editing of the 2005 AISC Commen-

tary; he is on the Building Seismic Safety Council Provisions Update Committee; and he is the current chair of the ASCE Technical Administrative Committee on Metals. He was awarded the 2000 Norman Medal and the 2003 Walter L. Huber Civil Engineering Research Prize from ASCE, and is a registered professional engineer.

Robert J. Dexter is an Associate Professor in the Department of Civil Engineering at the University of Minnesota. He is on the AISC Specification Task Com-mittees on Connections and Materials and the AISC Technical Ac-tivities Committee. He previously worked at South-west Research

Institute and Lehigh University. In addition to AISC, he is a member of ASCE, RCSC, and AREMA. He is a registered professional engineer in several states.

Daeyong Lee is a Researcher at the Re-search Institute of Indus-trial Science and Tech-nology (RIST) in South Korea. Prior to joining RIST in 2003, he was a Post-Doctoral Associate at the University of Minnesota and a Post-Doctoral Research Fellow at the University of

Michigan. He is a member of the Research Council on Structural Connections (RCSC), the Earthquake Engineering Research Institute (EERI), and the George E. Brown, Jr. Network for Earthquake Engineering Simulation (NEES) Consortium, Inc.

ABSTRACT New alternatives are presented for detailing column stiffeners in steel moment resisting girder-to-column connections that avoid welding in the k-area. These details were evaluated using monotonically-loaded pull-plate experiments, cyclically-loaded cruciform girder-to-column connection experiments, and three-dimensional nonlinear finite element analysis. The results confirm that existing AISC design equations for local flange bending (LFB) and local web yielding (LWY) are reasonable and conservative for non-seismic design applications and may be considered for seismic design applications. However, improved design provisions for LFB and LWY are presented that correlate better with test data from this research and the literature. Revisions are also recommended for panel zone design provisions. The results also indicate that the welded unreinforced flange-welded web (WUF-W) prequalified moment connection can perform well under cyclic loading provided that the weld metal used in the connection has a minimum notch-toughness; that unstiffened columns perform well in the connection region, both for monotonic and cyclic loading, if the column section is sufficiently large; and that the alternative fillet-welded continuity plate and doubler plate details explored in this research all performed well.

2

COLUMN STIFFENER DETAILING FOR NON-SEISMIC AND SEISMIC DESIGN

INTRODUCTION

The 1994 Northridge, California earthquake resulted in the fracture of complete joint penetration (CJP) welds connecting girder flanges to column flanges in steel moment-resisting connections in a number of steel frame structures ranging from one to twenty-six stories in height (FEMA, 2000a). These welds fractured due to a combination of reasons, including the weld metal having low fracture toughness combined with the presence of a notch from a backing bar and weld defects; the pre-Northridge connection geometry making the CJP welds more susceptible to high strain and stress conditions; and welding practices that resulted in inconsistent weld properties (FEMA, 2000a). As a result of these fractures, in the subsequent years there has been a tendency to be more conservative than necessary in the design and detailing of steel moment-resisting connections both in seismic and non-seismic zones within the United States. In particular, continuity plates and web doubler plates have often been specified when they are unnecessary and, when they are necessary, thicker plates have been specified than would be required according to the applicable specifications. In addition, the welds of the continuity plates to the column flanges have often been specified as being complete joint penetration welds, even though the use of more economical fillet welds may have sufficed. The design criteria for the limit states applicable to continuity plate and doubler plate design for non-seismic conditions are provided in Section K1 of Chapter K of the American Institute of Steel Construction (AISC) Load and Resistance Factor Design (LRFD) Specification for Structural Steel Buildings (AISC, 1999a). There are additional, more stringent provisions in the requirements for Special Moment Frames (SMF) in the AISC Seismic Provisions for Structural Steel Buildings (1992). However, the 1997 and 2002 AISC Seismic Provisions (AISC, 1997a, 1999b, 2001a, 2002) removed all design procedures related to continuity plates, requiring instead that they be proportioned to match those provided in the tests used to qualify the connection. As part of the SAC Joint Venture research program, preliminary post-Northridge seismic design guidelines and two advisories were published (FEMA, 1995; FEMA, 1996; FEMA 1999) that pertained to these column reinforcements in seismic zones. For example, the guidelines called for continuity plates at least as thick as the girder flange that must be joined to the column flange in a way that fully develops the strength of the continuity plate, i.e., this encouraged the use of CJP welds. However, more recent seismic guidelines (FEMA, 2000a) have reestablished design equations to determine whether continuity plates are required and, if so, what thickness is required. Recent research has revealed that excessively thick continuity plates are unnecessary. El-Tawil et al. (1999) performed parametric finite element analyses of girder-to-column joints. They found that continuity plates are increasingly effective as the thickness increases to approximately 60% of the girder flange. However, continuity plates more than 60% of the girder flange thickness brought diminishing returns. Furthermore, over-specification of column reinforcement may actually be detrimental to the performance of connections. As continuity plates were made thicker and attached with highly restrained CJP welds, they sometimes contributed to cracking during fabrication (Tide, 2000). Yee et al. (1998) performed finite element analyses comparing fillet-welded and CJP-welded continuity plates including heat input of the weld passes. Based on principal stresses extracted at the weld terminations, it was concluded that fillet-welded continuity plates may be less susceptible to cracking during fabrication than if CJP welds are used. The objectives of the research summarized in this paper are to reassess the design provisions for column stiffening, including both continuity and doubler plate detailing, for non-seismic and seismic design conditions, and to investigate new alternative column stiffener details. The project includes three components: monotonically-loaded pull-plate experiments to investigate the need for and behavior of continuity plates (Prochnow et al., 2000; Hajjar et al., 2003), cyclically-loaded cruciform girder-to-column joint experiments to investigate panel zone behavior and local flange bending (Lee et al., 2002, 2004), and parametric finite element analyses (FEA) to corroborate the experiments and assess the performance of various continuity plate and doubler plate details (Ye et al., 2000).

3

Throughout the project, new doubler plate and continuity plate details were investigated to explore economical detailing and minimize welding along the column k-line region, as per the recommendations of the AISC advisory for the k-line region (AISC, 1997b). New column stiffener details that were investigated included: continuity plates that were half the thickness of girder flange and that were fillet-welded to the column; doubler plates that were fillet-welded to the column flanges; and doubler plates that were offset from the column flanges by several inches to serve as both doubler and continuity plates. In addition, several unstiffened column specimens were analyzed and tested to verify where stiffeners are not required in steel connection design.

BACKGROUND The present AISC (1999a) non-seismic provisions that govern the design of continuity plates are based on three limit states: local web yielding (LWY), local flange bending (LFB), and local web crippling (LWC). Local web crippling is not discussed in this paper, as it rarely controls continuity plate design for typical W14 column sections (Dexter et al., 1999; Prochnow et al., 2000). A provision to restrict LWY of column sections was first defined in the 1937 AISC Manual (AISC, 1937). The provision remained the same until after the research of Sherbourne and Jensen (1957) and Graham et al. (1960), which was focused on investigating column web and flange behavior in moment-resisting frame connections and updating design specifications related to stiffener connection design. The outcome of the research generated guidelines for the use and sizing of continuity plates, which were first used in various forms in the AISC Allowable Stress Design (ASD) Specifications (e.g., AISC, 1989). This equation, as shown below, is still used in the current AISC LRFD non-seismic provisions for the limit state of local web yielding (AISC, 1999a):

(5 )u n yc cwR R k N F t < = + for interior conditions (1) (2.5 )u n yc cwR R k N F t < = + for end conditions (2)

where: = 1.0 Ru = required strength Rn = nominal strength k = distance from outer face of flange to web toe of fillet N = length of bearing surface (typically taken as the thickness of the girder flange) Fyc = minimum specified yield strength of the column tcw = thickness of column web The current non-seismic design equation for LFB is also based on the research work of Graham et al. (1960), in conjunction with limit load and buckling analyses of Parkes (1952) and Wood (1955). The equation is a result of using plastic yield line analysis to fit the experimental results regarding local flange bending of the tests and lower bound approximations of dimensions of common girder and column combinations of the time. A modified form of the LFB equation from Graham et al. (1960) is currently used in the LRFD Specification (AISC, 1999a). The design strength of the column flange for the local flange bending limit state is given as:

26.25n cf ycR t F = (3) where: = 0.9 tcf = thickness of the column flange The AISC (1999a) provisions corresponding to Equations (1) through (3) require that the design strength of the column flanges and webs for these limit states exceed the concentrated transverse force applied by the girder flange across the column flange. The design of continuity plates should also conform to Section K1-9 of AISC (1999a). Since the research of Graham et al. (1960), several researchers have examined the behavior of moment-resisting connections with and without continuity plates, and have recommended various methods of sizing continuity plates. Hajjar et al. (2003) provide a summary of significant past conclusions from researchers regarding continuity plate design.

4

The demands, Ru, for the above two limit states (i.e., LFB and LWY) are based on the forces delivered to the column flanges in the connection. Several possibilities exist for the calculation of these demands, including (but not limited to):

u yg gfR F A= (4)

1.8u yg gfR F A= (5)

1.1

0.95yg yg g g

ug

R F Z V aR

d

+= (6)

1.1u yg yg gfR R F A= (7)

where: Fyg = specified minimum yield stress of the girder Agf = area of one girder flange Ryg = ratio of expected yield strength of girder to specified minimum value Zg = girder plastic section modulus Vg = shear force in girder at plastic hinge location a = distance from column face to girder plastic hinge location dg = girder depth Equation (4) is typically used for non-seismic design, representing the nominal yield strength of one girder flange. Equation (5) was included in the 1992 AISC Seismic Provisions (AISC, 1992). The 1.8 factor includes a strain-hardening factor of 1.3 to account for the probability of having strength much greater than the minimum specified value and for the increase in strength after significant yielding, and another factor of 1.4 (1.4*1.3 1.8) that is related to the assumption that the full plastic capacity of the girder is carried by a force couple of the flanges only. This factor of 1.4 is the approximate upper bound ratio of the girder plastic section modulus, Zg, to the flange section modulus, Zgf (Bruneau et al., 1998). Although the stress state in the girder flange is not uniaxial, the girder flange demand predicted by Equation (5) can be put in perspective by comparing to the maximum possible uniaxial tensile strength of A992 steel. Equation (5) predicts stresses in the flange of 90 ksi for Fyg of 50 ksi, well above the likely tensile strength of A992 steel. For example, a survey of more than 20,000 mill reports from (Dexter, 2000; Dexter et al., 2001; Bartlett et al., 2003) showed that A992 steel has a mean tensile strength of 73 ksi. The 97.5 percentile tensile strength was 80 ksi, and the maximum value reported was 88 ksi. Equation (6) is included in AISC Design Guide No. 13 (AISC, 1999c). This equation, or slightly modified forms of it, has been widely used for the design of column stiffeners in steel moment connections (FEMA, 2000a). Equation (7) was presented by Prochnow et al. (2000) to provide a more realistic representation of the girder flange force in steel moment connections, mostly for use in assessment of the pull-plate experiments (Hajjar et al., 2003). FEMA (2000c) discussed the results of the research conducted by the SAC Joint Venture. The continuity plate requirements presented were essentially the same equations as the AISC Seismic Provisions (AISC, 1992). FEMA (2000c) thus concluded that it was appropriate to return to the requirements of the 1992 AISC Seismic Provisions (AISC, 1992). Hajjar et al. (2003) show that there is some consensus that continuity plates may be fillet-welded and may not always be required in non-seismic connections. However, there is a prevailing consensus that continuity plates are generally required for connections in seismic zones, although there are differing conclusions on the required width and thickness of the plate, on the type of weld that should be used to connect the stiffener to the column flange, and on whether very thick column flanges always require continuity plates. These prevailing trends are explored in this research.

5

The current AISC Seismic Provisions for the panel zone (PZ) design (AISC, 2002) include two design equations. The first, based largely on research by Krawinkler et al. (1971) and Krawinkler (1978), specifies the shear strength of the panel zone and the second places a limitation on panel zone slenderness:

230 6 1 cf cfv v v yc c p

g c p

b tR . F d t

d d t = +

(8)

where: v = 1.0 dc = column depth tp = panel zone thickness bcf = column flange width

( ) 90z zt d w / + (9) where: t = column web or doubler plate thickness; or total thickness if doublers are plug welded dz = panel zone depth wz = panel zone width In Equation (8), the term in parentheses accounts for the post-yield strength of the panel zone, as proposed by Krawinkler (1978). General procedures for the seismic panel zone design demand calculation are described in AISC (2002). The general formula can be given as:

1.1

0.95yg yg g g

u cgirders g

R F Z V aR V

d

+= (10)

where: Vc = shear force in the column This equation represents a modified approach from AISC Design Guide No. 13 (AISC, 1999c) by using v = 1.0 and removing the factor of 0.8 from the girder moments. While Equations (8) and (10) are commonly used at present, Lee et al. (2002, 2004) summarize a wide range of specification provisions that have been proposed in the past for computing panel zone design strength and demand. After the 1994 Northridge, California earthquake, a number of experimental and computational studies were conducted to further understand the role of panel zone shear deformation in seismic connection performance (FEMA, 2000a). Based on observations of the seismic behavior of pre-Northridge Welded Unreinforced Flange-Bolted Web (WUF-B) moment-resisting connections, and experimental and computational studies of several pre- and post-Northridge moment connection details, several conclusions regarding the effects of large panel zone shear deformation were drawn. First, panel zone yielding is stable under large inelastic cyclic loads and is thus potentially an excellent seismic energy dissipater. Nonetheless, excessive panel zone deformation can lead to localized column flange kinking, which may increase the potential for Low Cycle Fatigue (LCF) fracturing in the girder flange-to-column flange groove welds (Krawinkler et al., 1971; Krawinkler, 1978; Popov et al., 1986; Roeder and Foutch, 1996; El-Tawil et al., 1999; El-Tawil, 2000; Mao et al., 2001; Roeder, 2002). Second, large panel zone shear deformation in the connection also increases the inelastic stress and strain demands in other locations such as the edges of the shear tab (Mao et al., 2001; Ricles et al., 2002). Third, panel zone flexibility can significantly influence global stiffness and seismic response of the moment frame, and thus should be considered in frame analyses when flexible panel zones are used (e.g., Charney and Johnson, 1986; Liew and Chen, 1995; Biddah and Heidebrecht, 1998; Schneider and Amidi, 1998; Biddah and Heidebrecht, 1999).

6

As a consequence of these conclusions, the panel zone design demand and capacity included in the 1997 AISC Seismic Provisions (AISC, 1997a) were modified in AISC (1999b, 2001) to provide a panel zone strength and stiffness level sufficient to prevent excessively weak panel zones. The AISC (2001a) provisions were retained in the 2002 AISC Seismic Provisions (AISC, 2002). A stiffer panel zone forces most of the yielding in the connection region into the girder and thus causes a plastic hinge to form in the girder if the Strong Column-Weak Beam (SCWB) criteria (AISC, 2002) are satisfied. Many specimens tested after the Northridge earthquake and satisfying the prequalification requirements of AISC (2002) have shown good cyclic performance and higher ductility using a wide range of panel zone strengths. The few recently tested specimens that have had weak (i.e., under-designed) panel zones have also performed satisfactorily (e.g., Choi et al., 2000; Wongkaew et al., 2001). The panel zone deformations of these specimens were much larger than 4y (where y is panel zone shear yield strain), which was assumed by Krawinkler (1978) as the panel zone shear deformation at which the ultimate shear strength of the panel zone was developed in a joint. Nevertheless, those specimens completed the SAC loading history (SAC, 1997) up to the 4.0% interstory drift cycles without significant strength degradation.

DESIGN OF PULL-PLATE TEST SPECIMENS

A substantial parametric study was conducted (Prochnow et al., 2000) to assess appropriate girder and column combinations for the pull plate tests. The pull plate was chosen to represent a W27x94 girder flange. Using the girder minimum specified yield strength of 50 ksi, the girder flange demands were approximately 375 and 450 kips from Equations (4) and (7) for non-seismic and seismic design, respectively. Using these demands, a range of column sections were identified as being on the cusp of the LWY and LFB limit states, as defined by Equations (1) and (3), respectively. Finite element models of the proposed pull-plate test setup were then analyzed and compared for these sections to identify likely failure modes in the specimens (Ye et al., 2000). Table 1 summarizes the ratios of the design strength to required strength for the limit states of LWY and LFB using the three different girder flange demands from Equations (4), (7), and (5) and, for the design strengths, nominal dimensions from AISC (1995) as shown in the table, and a column nominal yield strength of 50 ksi. Note that AISC (1995) did not distinguish between design and detailing k-dimensions, but the values from AISC (1995) are close to the corresponding new design k-dimensions discussed in AISC (2001b). As seen in Table 1, within the test matrix, only the W14x159 satisfies the limit states of LWY and LFB for non-seismic design [i.e., using Equation (4)], and none of the sections satisfy these limit states using the larger values for demand. Figures 1 and 2 show the typical details of one of the pull-plate specimens used in this research (Prochnow et al., 2000; Hajjar et al., 2003). The pull-plate specimens consisted of a three-foot-long section of a column with pull plates welded to both column flanges. The pull-plates were in. by 10 in. in section, approximately the same as the flange of a W27x94 girder. None of the girder-to-column combinations satisfy the strong column-weak beam criteria from AISC (1997a, 2001a, 2002), as the overriding objective was to impart a large flange force onto a relatively weak column section so as to test the extreme limits of the associated limit states. Nine pull-plate specimens were tested based on these column sections: 1. Specimen 1-LFB: W14x132 without continuity plates, with 1/2 thick doubler plates, examined LFB 2. Specimen 2-LFB: W14x145 without continuity plates, with 1/2 thick doubler plates, examined LFB 3. Specimen 1-LWY: W14x132 without any continuity or doubler plates, examined LWY and LFB 4. Specimen 2-LWY: W14x145 without any continuity or doubler plates, examined LWY and LFB 5. Specimen 3-UNST: W14x159, without any continuity or doubler plates, examined LWY and LFB 6. Specimen 1-FCP: W14x132, with full-thickness (3/4 thick) continuity plates and CJP welds 7. Specimen 1-HCP: W14x132, with half-thickness (3/8 thick) continuity plates and fillet welds 8. Specimen 1B-HCP: repeat of 1-HCP to verify results 9. Specimen 1-DP: W14x132, with doubler plate box detail with 3/4 thick doubler plates Three doubler details were tested in this research (see Figure 3). Specimens 1 and 2 included a new doubler plate detail in which beveled doubler plates were fillet-welded to the column flange to avoid welding in the column k-line

7

(Detail I, Figure 3). The doubler plates stiffened the web of the two specimens in order to isolate local flange bending as the governing limit state. The fillet weld sizes were chosen to satisfy both the strength requirement (i.e., full development of the doubler plate in shear) and geometric requirements as outlined in AISC (1999c). Specimens 3 through 5 were unstiffened connections that looked at the interaction between LWY and LFB. Specimens 6 through 8 tested connections either with full-thickness continuity plates and CJP welds, replicating details often seen in current practice, or half-thickness continuity plates with fillet welds. Specimen 8 repeated the experiment of Specimen 7 to help verify the results of this economical continuity plate detail. The continuity plates all had clips, and for Specimens 7 and 8, the fillet weld along the web was terminated an additional from the toe of the clip to help mitigate stress concentrations near the column k-line. Specimen 9 included no continuity plate, but rather two doubler plates placed out away from the column web, as shown in Detail III, Figure 3. These plates thus act both as continuity and doubler plates, and in this detail it is the intent that two doubler plates would always be used. This detail provides an economical alternative to connections that require two-sided doubler plates plus four continuity plates. The CJP welds joining the pull plates to the column sections were made using the self-shielded FCAW process and E70T-6 filler metal. The E70T-6 wire had a diameter of 0.068 in. The filler metal used for the pull-plate specimens had a measured ultimate tensile strength of 77 ksi, and Charpy V-Notch (CVN) values of 63.7 ft-lbs at 70F and 19.0 ft-lbs at 0F. Figure 2 shows the detail of the girder tension flange-to-column flange connection. Continuity plates and web doubler plates were fillet-welded using the 100% carbon dioxide gas-shielded FCAW process and E70T-1 filler metal with a 0.0625 in. diameter. For Specimen 6, CJP welds were used to join the continuity plate to the column flanges, and for Specimen 9, CJP welds were used to join the web doubler plate to the column flanges. These CJP welds were also made with the gas-shielded FCAW process and E70T-1 filler metal. All plate material of the same thickness and columns with the same sizes were produced from the same heats. Table 2 presents a comparison of the average values of the coupon test results and the mill reports, as well as key measured dimensions of the actual cross sections tested [Prochnow et al. (2000) presents details of coupon specimens]. The columns were made from A992 steel, and the girders from A572 Gr. 50 plate. Along the column k-lines, the hardness values (Rockwell B-scale) were shown to range from 78 to 90 for the W14x132, from 86 to 96 for the W14x145, and from 75 to 81 for the W14x159. Measured notch toughness in the k-line region for the three specimens ranged from 100 to 200 ft-lbs at 70F for all column sections. A high strain rate of 0.004 sec-1 was used, which approximates the strain rate from seismic loading at about a 2 second period. The high strain rate increases the yield strength of the materials and increases the probability for brittle fracture, thereby testing the specimens under more severe conditions.

ESTABLISHMENT OF FAILURE CRITERIA FOR LOCAL FLANGE BENDING

AND LOCAL WEB YIELDING LIMIT STATES Before testing began, connection failure criteria were developed for the LWY and LFB limit states. The primary indicator of failure was whether the weld fractured prematurely. In these experiments, none of the welds fractured prior to the pull plate fracturing, so secondary failure criteria were established based on excessive deformation to identify problematic limit states. For each specimen, the column section was examined for failure at non-seismic and seismic girder demand load levels, uR , of 375 kips and 450 kips calculated as per Equations (4) and (7), respectively. The pull-plate load of 450 kips (which corresponds to approximately 1.5% specimen elongation) was used as the primary target for demand (Prochnow et al., 2000; Hajjar et al., 2003). The connection was classified as failing by LWY if at 450 kips the strain in the column k-line directly under the pull-plate was greater than 3.0%, or the strain in the column k-line was greater than the yield strain for the entire 5k+N area. The connection was defined as failing by LFB if at 450 kips the separation of the two column flange tips on the same side of the web and located at the column centerline was greater than in. The LFB failure

8

criterion was based on the permissible variations in cross section sizes ASTM (1998), which specifies that the flanges of a W-shape may be up to in. out-of-square.

PULL-PLATE EXPERIMENTAL RESULTS Table 3 summarizes the pertinent test results, including loads and specimen elongations when the specimen failed and when the different failure criteria were exceeded. Nominal strengths are computed using nominal material properties and the nominal section dimensions given in Table 1. Actual design strengths are calculated using the measured dimensions and material strengths given in Table 2. Prochnow et al. (2000) show that none of the specimens had strain levels exceeding 3% in the column web directly under the pull-plate at a load level of 450 kips, and only the unstiffened W14x145 specimen (2-LWY) had strain values greater than yield for the entire 5k+N region, as seen in Figure 4, which shows the strain distributions along the k-line of the column web. The W14x145 exhibits these higher strains relative to the W14x132 because measurements showed that the specific W14x145 section used in the test actually had a thinner web than the specific W14x132 section (see Table 2). There is no tolerance on web thickness in ASTM A6; the tolerance is only on the weight per foot (ASTM, 1998). The strain distribution also shows a much steeper gradient for the W14x132 (1-LWY) than the other two unstiffened sections. This gradient is likely due to its thinner column flange. The thicker column flanges of the W14x145 and W14x159 act to distribute the load more evenly into the column web. Equation (1) for LWY assumes a constant distribution of stress equal to the column web yield strength across the k-line for a distance 5k+N. Prochnow et al. (2000) and Hajjar et al. (2003) show both the experimental and finite element strain distributions in the column web in the direction of loading of Specimens 1-LWY, 2-LWY, and 3-UNST. The rectangular stress block inherent in Equation (1) for LWY only approximates the actual nonlinear stress distribution predicted both in the experiments and nonlinear analyses. Hajjar et al. (2003) thus propose a new equation for the LWY limit state that provides a more accurate representation of the nominal strength of this limit state. However, the simpler equation currently in AISC (1999a), Equation (1), was compared both to the pull plate results in this research and to past work by Graham et al. (1960) in Prochnow et al. (2000) and was found to be both reasonable and conservative for the non-seismic loading exhibited in these pull plate tests. Figure 5 shows the separation of the flanges near the tips of the flanges along the column length for all nine specimens. The W14x132 unstiffened and the W14x132 with doubler plates on the web (1-LFB) both failed this LFB criterion. By comparing the specimens without continuity plates but with web-doubler plates (1-LFB and 2-LFB) to those with no stiffeners at all (1-LWY and 2-LWY), it can be seen that a significant portion of the flange separation is due to web deformation, as confirmed by the finite element results of Ye et al. (2000) and experimental observations (Prochnow et al., 2000; Hajjar et al., 2003). In the case of the W14x145 (2-LWY and 2-LFB), which has a stiffer flange and, as it turns out, a thinner web, half of the flange separation is due to web deformation. The derivation of Equation (3) is based on the research of Graham et al (1960). Failure of their pull plate specimens was determined based upon fracturing. This is an unreliable failure mechanism for comparison with LFB predictions because it is based upon the toughness of the weld and base metal; the weld metal in particular was likely to be less tough than those used in the current pull-plate tests. Prochnow et al. (2000) thus show that the scatter in the test-to-predicted ratio of both the results of Graham et al. (1960) and in the current work is significant. Therefore, a new nominal strength equation for local flange bending has been derived in this work based upon a procedure similar to that used by Graham et al. (1960), in which a plastic yield mechanism in the flange is assumed. Prochnow et al. (2000) conducted a substantial parametric study of the range of yield mechanism parameters exhibited in typical girder-to-column connections, and thus simplified the proposed new local flange bending nominal strength equation to [see Prochnow et al. (2000) for details of this derivation]:

( )22 5.9n gf cf ycR t k t F= + (11a) where: tgf = thickness of girder flange

9

Using the mean values minus one standard deviation of tgf and k for common column-girder combinations simplifies this equation to:

( )20.8 5.9n cf ycR t F= + (11b) when using units of kips and inches (Prochnow et al., 2000; Hajjar et al., 2003). Table 4 exhibits the predicted (using both nominal and measured properties as listed in Tables 1 and 2, respectively, without a factor) and experimental strengths using the LFB yield mechanism reported in Table 3 and Equations (3) and (11b). The test-to-predicted ratio of Equation (11b) is consistently closer to 1.0 than Equation (3), and the standard deviations are similar. Prochnow et al. (2000) show even more dramatic results when comparing to the test results of Graham et al. (1960). However, it should be emphasized that Equation (3) was compared both to the pull plate results in this research and to past work by Graham et al. (1960), and Equation (3) was found to be both reasonable and conservative for the non-seismic loading exhibited in these pull plate tests. The results of the stiffened specimens (1-HCP, 1B-HCP, and 1-FCP) showed that, at least for monotonically loaded connections, a half-thickness continuity plate was adequate to avoid web yielding and flange bending. Results such as Figures 4 and 5 show a significant difference between the unstiffened and stiffened specimens. In particular, the specimens with fillet-welded half-thickness continuity plates (1-HCP and 1B-HCP) are well below the LWY and LFB failure criteria. The failure criterion for the continuity plates was complete yielding across the full-width section of the plates at 450 kips. The full-width section of the continuity plates was defined as the area just outside of the in. clips. Hajjar et al. (2003) show that neither Specimen 1-HCP nor 1-FCP fully yielded across the width of the continuity plates, and therefore both were still capable of resisting load and had not failed. The half-thickness continuity plate fillet welds also did not fracture, thus validating the integrity of this detail. Specimen 1-DP, the box detail, also performed well. Strains in the doubler plate and web did not exceed yield for the entire 5k+N region, and neither the LWY nor the LFB limit states were breached. However, the doubler plates used in Specimen 1-DP were rather thick, based on the observation of Bertero et al. (1973) that the doubler plates are less effective when they are moved away from the web. Equations (1) and (3) predict that two 3/32 doubler plates would be required to mitigate the LWY and LFB limit states, respectively [using Equation (4) to compute girder flange demand], whereas 3/4 thick plates were used in the specimen. Thus, further research is required to determine an appropriate sizing procedure for the box detail. Note also that the backing bars used for the CJP welds of the doubler plates were not removed (and would not normally be removed for this detail), and the welds performed well in this test.

DESIGN OF CRUCIFORM TEST SPECIMENS A total of six full-scale, girder-to-column cruciform specimens were tested in this research program. A Welded Unreinforced Flange-Welded Web (WUF-W) connection detail was used; this is a pre-qualified steel moment connection detail in FEMA (2000a). The original connection topologies were selected mainly based on a parametric study of panel zone stiffening requirements and on using sizes that would highlight specific aspects of the column limit states being investigated in this research (Lee et al., 2002). In particular, a relatively shallow girder depth (and thus a relatively large flange force) and relatively weak column panel zone were used so as to: 1) create specimens with different characteristics from the large body of tests that have deeper girders and stronger panel zones, for which a base of results has to some extent already been established; and 2) generate large strains in the connection region and thus create severe tests for the limit states being investigated in this research. The test matrix is outlined in Table 5. Due to unexpected premature brittle failure of girder flange-to-column flange welds in one of the five cruciform specimens (i.e., Specimen CR4), one additional specimen was replicated with new base metal and weld consumables and retested. This new specimen was identified as Specimen CR4R. Lee et al. (2002, 2004) document the brittle failure of Specimen CR4 in detail. As shown in Table 5, all three doubler details of Figure 3 were tested in this experimental study. One continuity plate detail was also tested, in which the plate thickness was approximately equal to half the girder flange thickness,

10

and the plate was fillet-welded to both the column flanges and doubler plates. The size of the fillet welds needed for both doubler and continuity plate details were calculated using procedures given in the AISC Design Guide No.13 (AISC, 1999c). A 5 in. wide plate, which is slightly larger than half of the girder flange width, was selected in compliance with the requirements of 1.79 yt b F E= and 3 2gf pb b t contained in AISC (1999a). In addition, 1 in. clips were provided to avoid interference with the fillet welds connecting doubler plates to the column flanges. One completely unstiffened specimen with a W14x283 column (Specimen CR1) was also tested to verify the cyclic response of a specimen without continuity plates and doubler plates. In addition, Specimens CR2 and CR5 had no continuity plates although continuity plates were required as per AISC (1992) for Specimen CR2 and as per AISC (1992, 1999a, 1999c) for Specimen CR5. These specimens were used to examine the response of a column flange subjected to the LFB limit state under cyclic loading. As one focus of the cruciform tests is to investigate the design provisions for panel zones and column stiffeners and to test new stiffening alternatives, the panel zones were designed with the intent to exceed the shear deformation of 4y, where y is the panel zone shear yield strain. The design shear deformation level of 4y was suggested by Krawinkler et al. (1971) and Krawinkler (1978) and is implied in the AISC non-seismic and seismic panel zone design equations (AISC, 1999a; AISC, 2002). Designing the specimens with relatively weak panel zones ensures that all column stiffening details are rigorously tested through large localized cyclic strains, and provides a means for evaluating the strength of the panel zone at the level of design deformation. It is also desirable, however, to ensure the panel zones are strong enough to allow for development of the plastic moment strength of the girders. This is necessary to develop large girder flange forces, thereby placing high force demands on the column flanges and continuity plates. To meet this balance of girder and panel zone strength, a method of estimating the relative strengths was used for the design of panel zones in this experimental study. The quantity of most interest for the purpose of the panel zone design was the ratio of nominal panel zone strength (Pz) to nominal girder strength (Pg). These strengths were calculated as the total girder tip loads required to reach the strength level under consideration, and are given in Lee et al. (2002, 2004). A baseline value of Pz/Pg equal to 1.0 was targeted for the design of the panel zones. This implies that the panel zone strength (at an average shear distortion of 4y) is achieved at the same time the girders reach their plastic moment capacities. By selecting this ratio, the intent is to achieve the goals of exceeding both plastic moment Mp in the girders and shear distortion of 4y in the panel zones. Table 6 presents the design strength-to-demand ratios using nominal material properties for the LFB, LWY, and panel zone limit states using various methods of demand calculations, i.e., using Equations (4) through (7) for LFB and LWY and Equation (10) for panel zone yielding. Note that the design strengths for LFB and LWY are the design strengths of the column shape alone and do not include the column reinforcement, if any. Also shown in Table 6 are the column-girder moment ratios calculated from the 2002 AISC Seismic Provisions (AISC, 2002), assuming no axial compression in the column. As mentioned above, panel zone strengths are based on the AISC Seismic Provisions (AISC, 1997a, 1999b, 2001a, 2002), with a resistance factor of v = 1.0, while the LFB and LWY strengths are based on the AISC LRFD Specification (AISC, 1999a). Table 6 shows that Equation (5) and Equation (6) provide similar demand values for these specimen sizes. As with the pull-plate specimens, it was anticipated during specimen design that the box detail may be less than fully effective, based on the results of Bertero et al. (1973). Thus, the doubler plates provided were approximately 30% thicker in Specimens CR4 and CR4R than those of Specimen CR3, which has the same member sizes. The capacity-to-demand ratios given in Table 6 reveal that the panel zones of most of the specimens are weak. Typical connection topologies for the cruciform specimens are shown in Figures 6 and 7. Specimen CR1 represents a relatively large, unreinforced interior connection with a relatively weak panel zone. It is intended primarily to study the panel zone strength provision for thick column flanges. The relatively thick column flange of 2.07 in. coupled with the unreinforced panel zone results in a post-elastic panel zone strength contribution of approximately 40% in Equation (8), representing a high, yet typical value. This specimen meets the Strong Column-Weak Beam (SCWB) criteria of the 2002 AISC Seismic Provisions (AISC, 2002), also shown in Table 6. No continuity plates are needed as per the 1992 AISC Seismic Provisions (AISC, 1992) [i.e., using Equation (5)] or the AISC Design

11

Guide No. 13 (AISC, 1999c) [i.e., using Equation (6)] (see Table 6). Specimen CR1 is also intended to show that continuity plates are not necessarily needed for all seismic moment connection details. Specimen CR2 represents a moderately-sized, reinforced interior connection with a single-sided doubler plate. It is intended primarily as a verification of the LFB criteria of the 1992 AISC Seismic Provisions (AISC, 1992) and the AISC Design Guide No. 13 (AISC, 1999c). This specimen is also intended to confirm that continuity plates are not always needed for seismic moment connections. A relatively weak panel zone, similar to Specimen CR1, is provided. A square-cut fillet-welded doubler plate detail (Detail II, Figure 3) is utilized; this detail was deemed by the fabricator to be more efficient to fabricate than Detail I. Note that welding across the top and bottom of the doubler plate is not done in Detail II. The SCWB moment ratio is close to unity in Specimen CR2. The presence of the doubler plate and thinner column flanges reduces the value of the post-elastic panel zone strength contribution in Equation (8), but it is still a moderate magnitude of approximately 17%. Specimen CR3 represents a moderately-sized, reinforced interior connection with both doubler plates and continuity plates. This is the second test of doubler plate Detail II (Figure 3). Specimen CR3 requires continuity plates as per the 1992 AISC Seismic Provisions (AISC, 1992) and the AISC Design Guide No. 13 (1999c), and is intended to show that a fillet-welded continuity plate detail using a continuity plate that is approximately half as thick as the girder flange can perform satisfactorily in cyclic loading applications. The nominal strength of this continuity plate, computed as:

( ) ( ) ( ) ( )[ ]2 2 2 0.9 50 0.5 5.0 1.0 180t n t g yP A F = = = kips was approximately 30% less than the required strength:

( ) ( )( ) ( ) ( )( ) ( )221.8 6.25 1.8 50 9.07 0.875 0.9 6.25 1.31 50 232yg gf cf ycF A t F = = kips (The nominal strength is approximately equal to the required strength with a factor of 1.0 used for local flange bending). The value of 1.0 in the above parenthesis accounts for 1 in. clip in the continuity plate. The SCWB moment ratio is lower than unity in this specimen. The panel zone strength is similar to Specimens CR1 and CR2. Because of the thinner column flanges and heavier panel zone reinforcement as compared with the above two specimens, the predicted post-elastic strength of the panel zone is reduced to approximately 12%. Specimen CR4 (and CR4R) represents a moderately-sized, reinforced interior connection with relatively heavy panel zone reinforcement using the box (offset) detail and no continuity plates in a situation in which continuity plates would be required according to the 1992 AISC Seismic Provisions (AISC, 1992) and the AISC Design Guide No. 13 (AISC, 1999c). The panel zone design is just below the design strength from AISC (2002). As with Specimen CR3, this specimen does not meet the SCWB criteria of AISC (2002) for the case of no axial compression. The thick doubler plates result in a low post-elastic panel zone strength contribution, just below 10%. Specimen CR5 represents the smallest column section tested, with fillet-welded doubler plates and no continuity plates. Doubler plate Detail I (Figure 3), the back-beveled fillet-welded detail, is tested in Specimen CR5. This specimen requires continuity plates as per the non-seismic (AISC, 1999a) and is well below the LFB limit states for seismic design (AISC, 1992, 1999c), but no continuity plates were used. While tested cyclically, the non-seismic details of this specimen (i.e., lack of continuity plates) were intended to investigate the LFB design criteria of the AISC LRFD Specification (1999a) as well as to provide further evidence that continuity plates may not be required in all seismic moment connections. The panel zone strength is similar to Specimens CR1, CR2, and CR3. Because a smaller column was needed to breach the non-seismic LFB limits, the SCWB moment ratio is much lower than unity in Specimen CR5. A low post-elastic panel zone strength of approximately 8% is expected in this specimen. The weld details were as recommended in FEMA (2000a) for the prequalified Unreinforced Flange-Welded Web (WUF-W) connection, as modified in connections tested by Ricles et al. (2002). Figures 6 and 7 illustrate the typical connection welding details used to fabricate the specimens in this experimental study [represented by the details for Specimens CR1 and CR3, respectively; all specimen details are presented in Lee et al. (2002, 2004)]. All welding was done with the Flux-Cored Arc Welding (FCAW) process. The welding was done in two stages, with shop welds made at the fabricator and field welds made in the Structural Engineering Laboratory at the University of Minnesota by experienced erection welders. E70T-1 (Lincoln Outershield 70) wire with 100% carbon dioxide gas

12

shielding was used for all shop welding, including the shear tab welding to the column flange, the welding of the doubler plates, and the welding of the continuity plates (see Figure 7). The notch toughness of the E70T-1 wire is required by AWS A5.20 (AWS, 1995) to be 20 ft-lbs at 0F and, according to the Lincoln Electric product family literature, the typical values for Outershield 70 are 28 ft-lbs at 0F. The shear tab was welded to the column flange using 1/4 in. fillet welds on each side of the plate, although this deviated from the recommended WUF-W connection welding details, which require Partial Joint Penetration (PJP) groove welds at this location. The field welds were made with the self-shielded FCAW process. The girder flange-to-column flange CJP groove welds were made in the flat position with E70T-6 (Lincoln Innershield NR-305) wire. Welds made with E70T-6 wire are required by AWS A5.20 (AWS, 1995) and AISC (2002) to have notch toughness of 20 ft-lbs at -20F and 40 ft-lbs at 70F. FEMA (2000a) has recommended minimum notch toughness requirements at two temperatures, 20 ft-lbs at 0F and 40 ft-lbs at 70F. According to the Lincoln Electric Company product family literature, the typical values for NR-305 are 21 to 35 ft-lbs at -20F and 21 to 54 ft-lbs at 0F. For the first two specimens that were fabricated (i.e., Specimens CR1 and CR4), 5/64 in. diameter NR-305 wire was used for girder flange-to-column flange CJP groove welds. However, this wire and the weld procedures were subsequently found to produce weld metal with only 2 to 3 ft-lbs at 0F and 2 ft-lbs (Specimen CR4) to 19 ft-lbs (Specimen CR1) at 70F that did not meet the FEMA (2000a) or AISC (2002) recommended minimum notch toughness requirements (Lee et al., 2002, 2004). For this reason, for the remaining specimens, it was decided to use a lot of NR-305 weld wire with 3/32 in. diameter that was previously characterized by the Edison Welding Institute (EWI) and was known to have good notch toughness (Lee et al., 2002). This particular lot of 3/32 in. diameter NR-305 wire was used for the CJP welds in Specimens CR2, CR3, CR4R, and CR5. In addition, different welding equipment and procedures were also used for the remaining specimens (details are reported in Lee et al., 2002, 2004). All CJP welds were ultrasonically tested by a certified inspector in conformance with Table 6.3 of AWS D1.1-2000 (AWS, 2000) for cyclically loaded joints. The out-of-position field welds, including the CJP welds connecting the girder web to the column flange and all reinforcing fillet welds were made with 0.068 in. diameter E71T-8 (Lincoln Innershield NR-203MP) wire for Specimens CR1 and CR4, and 5/64 in. diameter E71T-8 (Lincoln Innershield NR-232) wire for the other specimens (see Figure 6). Welds made with E71T-8 wire are required by AWS A5.20 (AWS, 1995) to have notch toughness of 20 ft-lbs at -20F. According to the Lincoln Electric Company product family literature, the typical values for NR-203MP are 50 to 200 ft-lbs at -20F and the typical values for NR-232 are 20 to 69 ft-lbs at -20F. The shear tab was designed to extend approximately 0.25 in. into the top and bottom access holes and acted as the backing bar for the CJP welds of the girder web to the column flange. This extension acted as a short runoff tab, allowing the weld to extend the full depth of the girder web. Ricles et al. (2002) recommended that these runoff tabs of the vertical web weld be ground smooth, which is labor intensive. Since it was felt that this might not be necessary, these runoff tabs were not ground smooth in the specimens tested in the present study. As shown in Figure 6, 5/16 in. reinforcing fillet welds were placed under the top girder flange backing bar and below the back-gouged region under the bottom girder flange. Supplemental fillet welds were also provided between the shear tab and girder web, with 5/16 in. fillet welds being placed along the full height of the shear tab using the E71T-8 wire. The weld access holes for the test specimens were designed based on the research of Mao et al. (2001) and Ricles et al. (2002). This weld access hole or similar details are now required for many of the prequalified steel moment connections outlined in FEMA (2000a), including the WUF-W connection. Lee et al. (2002, 2004) illustrate the dimensions of the access hole used in this experimental study. Material testing was performed on all wide-flange shapes and available stiffener plates used for the test specimens. All rolled sections were fabricated from A992 wide-flange sections, and A572 Grade 50 steel was selected for all stiffener materials. Lee et al. (2002) present detail of the coupon specimens and locations from which they were cut. Table 7 summarizes the tensile test results (average values of the coupons are shown) and mill certificate values for the W-shapes. Lee et al. (2002, 2004) present similar properties for the stiffener plates.

13

In order to verify the material properties of the girder flange-to-column flange CJP groove welds, one weld test plate was made as per Figure 2A of AWS A5.20-95 (AWS, 1995) for Specimens CR2, CR3, CR4R, and CR5 at the time of the connection welding. CVN specimens were also made for Specimens CR1 and CR4. These CVN specimens were machined from one of the girder flange-to-column flange groove welds that showed no signs of fracture after the test. The test results, which followed ASTM E23 for the CVN test and ASTM E8 for the tensile coupon test, are presented in Lee et al. (2002, 2004). Specimens CR2, CR3, CR4R, and CR5 satisfied the minimum requirements recommended by FEMA (2000a), i.e., filler metals providing CVN toughness of 20 ft-lbs at 0F and 40 ft-lbs at 70F. However, only the CJP welds in Specimen CR2 satisfied the supplemental requirements in FEMA (2000b) of the minimum filler metal yield strength of 58 ksi. Figure 8 shows a schematic of the test setup. The load pins placed at the top and bottom of the column were designed to allow free rotation of the column ends during loading, simulating inflection points at the mid-height of the column in a steel moment frames. Four bracing members, not shown in Figure 5, were attached to the diagonal load frame members to restrict the out-of-plane deformation of the girders due to lateral-torsional buckling. Quasi-static, anti-symmetric, cyclic loads were applied to the girder tips by using four MTS hydraulic actuators, i.e., two actuators for each side. Each actuator was capable of 77 kips at a stroke of +/- 6.0 in. The SAC (1997) loading history was applied to ensure results could be compared to numerous other SAC girder-to-column tests conducted, in which six cycles were applied at each interstory drift level of 0.375%, 0.5%, and 0.75%, and four cycles were applied at 1.0% interstory drift level, and two cycles were applied at each interstory drift level of 1.5%, 2.0%, 3.0%, and 4.0%. Further cycles at 4.0% interstory drift were then applied until specimen failure to provide information on low cycle fatigue response of these connections.

CRUCIFORM TEST RESULTS All specimens, excluding Specimen CR4, completed the SAC (1997) loading history up to 4.0% interstory drift without noticeable strength degradation. The plots of moment versus total plastic rotation for two representative connections (West Connections of Specimens CR2 and CR5) are presented in Figures 9 and 10. After completing the two cycles at 4.0% interstory drift required by the SAC (1997) protocol, additional 4.0% interstory drift cycles were applied until each specimen failed. Specimens CR1, CR2, CR3, and CR4R were subjected to 14, 16, 14, and 12 cycles, respectively, before significant strength degradation was noticed. It cannot be determined that there is any significance to the variation in number of cycles in the range from 12 to 16 cycles. Thus, it can be assumed that these four specimens performed equally well. Specimen CR5 also performed satisfactorily, completing six cycles at 4.0% interstory drift even though the this specimen was significantly under-designed for the limit states of local flange bending and panel zone yielding as per AISC (1992, 1997a, 1999a, 2001, 2002) (see Table 6). In these five successful tests, the primary failure mode was low cycle fatigue cracking and rupturing near the girder flange-to-column flange boundary region. Visible cracking in the connections typically first occurred at the top or bottom edge of the shear tab in the 3.0% interstory drift cycles in some specimens, but the connections suffered no strength degradation until the girder flanges were locally buckled or until the low cycle fatigue cracks in the girder flange were significant after several 4.0% drift cycles. Except for Specimen CR5, the initial girder flange cracks in these specimens originated in the center of girder flange width (in both the top and bottom flanges in the various specimens) at the toe of the girder-to-column fillet welds that reinforced the topside of the CJP welds. The major crack in Specimen CR5 was observed in the middle of the CJP welds instead of at the toe of the CJP welds. Details of the progression of failure in each specimen are given in Lee et al. (2002). Specimen CR5, reinforced by doubler plate Detail I (back-beveled fillet-welded detail), satisfactorily completed the SAC (1997) loading history. Specimens CR2 and CR3 showed better cyclic connection performance, when compared with the test results of Specimen CR5, with the doubler plate Detail II (square-cut fillet-welded detail), although the column flanges and panel zones in general were stronger in these specimens, which more likely contributed to the difference in performance. In addition, the comparable performance of Specimen CR4R relative to Specimen CR3, which consisted of the same girder and column sizes, showed that the doubler plate Detail III (box detail) can also provide a similarly ductile connection performance and is equally effective in functioning as continuity plates. This finding was also indicated in the pull-plate tests, and the cruciform tests verify that cyclic loading test does not change those conclusions. Within the limited number of experiments, it has been found that

14

the above three different variations had no significant impact on the cyclic connection performance. Thus, it cannot be concluded that any of those details are more advantageous. Instead, the most economical detail should be recommended. The continuity plates of Specimen CR3 were heavily instrumented and largely showed strains below yield for the full loading history. At 4.0% interstory drift, some yielding was just beginning on the continuity plate near the 1 in. clip (see Figure 7). Note that because of the clip, this cross section of the continuity plate has reduced area relative to the remainder of the continuity plate, which remained elastic. While the multi-axial strain state in a continuity plate is complex, these results imply that the plastic moment may be achieved in the girder without significantly yielding the continuity plate. The fact that this specimen performed well provides two important conclusions. First, when continuity plates are used, it should not be necessary to use full thickness continuity plates that are groove-welded to the column flanges. Second, the use of smaller fillet-weld sizes is feasible for attaching thinner continuity plates to the column flanges. For Specimen CR3, only 3/8 in. fillet welds on each side were required to connect the 1/2 in. continuity plates to the column flanges, and 5/16 in. fillet welds were required to connect to the doubler plates (see Figure 7). Fillet welds, especially relatively small fillet welds such as these, pose a less significant risk of causing k-area cracking in the column due to the lower restraint and resultant tensile residual stress. It is thus proposed that it may be sufficient to design the continuity plate size and associated fillet welds using procedures given in the AISC Design Guide No.13 for both non-seismic and seismic design (AISC, 1999c). The pull-plate tests also support these conclusions. Present AISC LRFD specifications (AISC, 1999a) explicitly require that the fillet welds develop the full strength of the continuity plate. This implies that the welds are required to essentially remain elastic when the plate is fully plastified. In this way, it is assured that the fillet welds are not the weak link in the column details. Although the fillet welds in these test specimens were designed for this criterion, the fact that the continuity plates are not fully plastified across their gross section indicates that the fillet welds need not necessarily develop the full plastic capacity of the continuity plate. This issue often arises when continuity plates are sized greater than they need to be for LFB, to use a particular standard thickness for example or to accommodate a weak axis connection. In these cases the continuity plate will clearly not be yielded and the welds need not develop the plate but rather need only to provide strength greater than the difference between the demand and the LFB capacity of the flanges without continuity plates. In order to understand the complex stress and strain distributions and force flows near the girder-to-column junction, the distributions of strains in the longitudinal direction of girder flanges near the CJP welds were also investigated (Lee et al., 2002). In all five successfully tested specimens, the maximum longitudinal tensile strains measured in the middle of West girder top flanges were within the range of 19,000 to 26,500 at the first peak of 4.0% interstory drift. Similarly, the maximum longitudinal tensile strains in the middle of the East girder bottom flange were within the range of 10,000 to 33,000 at the first peak at 4.0% interstory drift. Figure 11 shows typical strain gradients in the East girder bottom flange in Specimens CR2 and CR3, without and with a continuity plate, respectively. In Figure 11a, three-dimensional nonlinear static finite element results from Ye et al. (2000) are shown for interstory drift levels of 2.0% and 4.0%. At 4.0% interstory drift, the computed strains agree fairly well with the measured strains. However, the computed strains were somewhat higher than measured strains at 2.0% interstory drift. Although the peak strain levels are approximately the same, Specimens CR3 and CR4R [not shown in the figure; see Lee et al. (2002)] showed relatively lower strain gradients along the girder flange width at higher drifts as compared with the other three specimens. It is believed that the trend towards having lower strain gradients in Specimens CR3 and CR4R girder flanges is primarily due to their column stiffening details. These results reinforce that both the half-thickness continuity plates (in case of Specimen CR3) and the box (offset) doubler plate detail (in case of Specimen CR4R) are effective as column stiffeners to mitigate the local flange bending in column flanges. However, while the general trends in the results are as discussed above, Figure 11 shows that the actual difference in strain gradients between a specimen with and without a stiffened flange may often be subtle. In addition, while Specimens CR1, CR2, and CR5 did generally have greater girder flange strain gradients, there is no evidence that these high strain gradients were detrimental to the performance of the connection. For example, Specimen CR1 had low notch toughness [far less than the FEMA (2000a) requirements], yet the high strain gradient did not cause a

15

brittle fracture. In addition, the strain gradient was worse in the West girder top flange of Specimen CR1 than in Specimen CR2 [not shown in the figure; see Lee et al. (2002)], whereas Specimen CR1 did not require continuity plates and Specimen CR2 did, based upon the seismic girder demand. Therefore, some of the variation in the strain gradient among the different specimens may be somewhat random, based upon local residual stresses, etc., and this research indicates that having a strain gradient in the girder flange does not necessarily precipitate premature fracture.

LOCAL FLANGE BENDING AND LOCAL WEB YIELDING CRITERIA FROM CRUCIFORM TESTS While the pull-plate tests investigated both the LFB and LWY limit states, the investigation of both non-seismic and seismic design criteria for LWY was limited with the cruciform tests. As shown in Table 6, the selected five cruciform specimens satisfied all the LWY design criteria considered in this research program, as would be customary for girder-to-column cruciform connections. In the pull-plate tests, the LFB yield mechanism was defined by limiting the column flange separation, measured between the column flanges at the edges of the pull-plates (simulating the girder flanges). A flange separation limit of 1/4 in. between the two flanges was established for the non-seismic design. In the cyclic cruciform experiments, this limiting deformation was taken as one-half the pull-plate separation value of 1/4 in., as the deformation of only one flange measured. The measured maximum column flange displacements in Specimens CR1 and CR5 were thus compared with a limit of 1/8 in. Specimen CR1 developed relatively small column flange deformations up to 4.0% interstory drift cycles (just 26% of the 1/8 in. limit), as was expected from the large LFB design strength/demand ratios presented in Table 6. In Specimen CR5, an unexpectedly low maximum column flange deformation (49% of the 1/8 in. limit) was measured during 4.0% interstory drift cycles even though Specimen CR5 did not even meet the non-seismic LFB design criteria. These tests thus indicate that the LFB strength predicted by Equation (1) (AISC, 1999a) may be suitable for use for seismic design as well. With respect to seismic demand, Specimen CR5 was the most substantially under-designed cruciform specimen with respect to LFB. As shown in Table 6, the design strength-to-demand ratio was only 0.84 for non-seismic design demand as per the AISC LRFD Specification (1999a), and it equaled to 0.47 and 0.46 for the seismic design demands within the 1992 AISC Seismic Provisions (AISC, 1992) and the AISC Design Guide No. 13 (AISC, 1999c), respectively. However, Specimen CR5 performed satisfactorily up to 4.0% interstory drift cycles without any damage induced by the lack of LFB mitigation. Therefore, at a minimum, observation of the five successful cruciform tests and the pull-plate tests indicated that the seismic demand given by Equation (6) (AISC, 1999c) is adequate and conservative for determining seismic LFB demand when the capacity presented in Equation (3) is used. Equation (6) provides a more rational basis for calculation of the required force than does Equation (5) (both equations often yield similar values). Table 6 indicates that both Specimens CR2 and CR5 would require continuity plates if Equation (3) is used in conjunction with Equation (6) (using nominal properties). Placing a reduction factor of 0.85 on Equation (6), particularly for use with interior connections, would change these values to 0.94 and 0.54 respectively, putting Specimen CR2 just over the cusp of breaching the LFB limit state. Use of Equation (6) with a reduction factor of 0.85 for interior connections, coupled with Equation (3) to compute strength, also compares favorably when compared to other test results. For example, Specimens C1 and C3 from Ricles et al. (2002), which each had two W36x150 (50) girders framing into a W14x398 (50) column (for Specimen C1) and a W27x258 (50) column (for Specimen C3) using the WUF-W connection with no continuity plates, performed adequately through the SAC (1997) loading history. Their ratio of LFB strength to demand using Equation (6) with a reduction factor of 0.85 was 2.32 and 0.76, respectively. Thus, Specimen C1 clearly should not need continuity plates. Specimens CR2 from this work and Specimen C3 (Ricles et al., 2002) could also both go without continuity plates, thus showing even the proposed 0.85 factor on Equation (6) would often be conservative.

PANEL ZONE CRITERIA FROM CRUCIFORM TESTS All the WUF-W specimens shown in Table 6 had inadequate panel zone strengths as per AISC (1999b, 2002). In spite of the weaker panel zones, however, these specimens (excluding Specimen CR4) showed stable, ductile panel zone response. The first three specimens (Specimens CR1, CR2, and CR3) developed more than 10y of panel zone shear deformation during the 4.0% interstory drift cycles, while the other two specimens developed more than 6y

16

(Specimen CR4R) and approximately 8y (Specimen CR5) maximum panel zone shear deformations at the same interstory drift level. The analyses of the panel zone elastic and inelastic behavior indicated significant energy dissipation in this region for Specimens CR1, CR2, and CR3. Relatively mild energy dissipation was observed in Specimens CR4R and CR5 even though the measured maximum amount of the connection total plastic rotation was similar in all cases. The smaller panel zone energy dissipation in these two specimens were mostly caused by the design of a stronger panel zone in the case of Specimen CR4R, and by the larger column flange yielding around each girder flange in the case of Specimen CR5. Lee et al. (2002, 2004) show that Equation (8) over-predicts the panel zone shear strength at the design deformation of 4y for the five cruciform test specimens and a number of tests from the literature. A new panel zone shear strength equation is derived in Lee et al. (2002) and is shown to compare well with 49 tests from the literature:

3

2

150.55 1 cf cfv yc c p

g c p

b tR F d t

d d t= +

(12)

A resistance factor of 0.85 is recommended for use with the equation (Lee et al., 2002, 2004). Equation (12) will tend to yield lower panel zone nominal shear strengths as compared to Equation (8). Thus, if it is assumed that the AISC (2002) and FEMA (2000a) procedures result in adequate panel zone designs, use of Equation (12) should also include adoption of a lower panel zone required shear force (i.e., panel zone shear demand) to result in similar panel zone thicknesses as AISC (2002). Lee et al. (2004) discuss an alternative recommendation for computation of panel zone shear demand that permits panel zone shear deformation up to 8y. Lee et al. (2002, 2004) also document that observation of the yielding in the joint region at the higher interstory drift levels indicated that it is more realistic to calculate the panel zone design demand directly at the column face for connections that have weaker panel zones, without considering the moment increments due to the girder shear force at the assumed girder plastic hinge location.

CONCLUSIONS This paper has presented the results of a combined experimental and computational research program investigating non-seismic and seismic response and design of column stiffening details. The limit states of local flange bending, local web yielding, and panel zone yielding were studied in detail, and several alternative and economical column stiffening details were investigated that avoid welding in the column k-line region. Conclusions from this research include the following: Specimens CR1, CR2, CR3, CR4R, and CR5 completed the SAC (1997) loading history up to 4.0% interstory

drift cycles without any significant strength degradation in the connections, and ductile failure modes were observed in all specimens. The primary failure mode of these five specimens was Low Cycle Fatigue (LCF) crack growth and eventual rupture of one or more girder flange-to-column flange E70T-6 Complete Joint Penetration (CJP) groove welds. Visible cracking in the connections typically first occurred at the top or bottom edge of the shear tab in the 3.0% interstory drift cycles in some specimens, but the connections suffered no strength degradation until the girder flanges were locally buckled or until the low cycle fatigue cracks in the girder flange were significant after several 4.0% drift cycles. The weld access hole detail chosen for this experimental study showed good performance under repeated large cyclic connection deformations. No LCF cracking occurred at the toe of the weld access hole prior to significant cracking occurring elsewhere in the connection, particularly at the toe of the CJP welds of the girder flange to the column flange.

These experimental results showed that, when properly detailed and welded with notch-tough filler metal, the WUF-W steel moment connections can perform adequately under large quasi-static cyclic loads even though relatively weak panel zones and low local flange bending strengths were chosen as per the current design provisions and recommendations (AISC, 1992, 1997a, 1999a, 1999b, 1999c, 2001a, 2002; FEMA, 2000a). The pull-plate tests also support this conclusion. None of the E70T-6 CJP welds in the pull-plate tests fractured despite plastic deformation, even when the flange tip separation was over in, indicating that column stiffener details may have little influence on the potential for brittle weld fracture provided the weld is specified with minimum CVN requirements and backing bars are removed.

17

Specimen CR4 was unintentionally prepared with E70T-6 weld metal that had Charpy V-Notch (CVN) values that were much lower than the minimum requirements of FEMA (2000a). This was the only test that did not satisfy the connection prequalification requirement of completing two cycles at 4.0% interstory drift level. The premature brittle failure of this specimen reconfirmed that achieving the required minimum CVN toughness in the girder flange-to-column flange CJP welds is critical for good performance in the prequalified steel moment connections. Note that these low toughness welds occurred despite the certification of the filler metal, and that the certification is only required annually, unlike the way that each heat of steel is tested.

Application of the alternative column stiffener details (i.e., back-beveled fillet-welded doubler plate detail; square-cut fillet-welded doubler plate detail; groove-welded box doubler plate detail; fillet-welded 1/2 in. thick continuity plates) in the WUF-W steel moment connections was successfully verified. No cracks or distortions were observed in the welds connecting these stiffeners to column flanges before the rupturing of the girder flange-to-column flange CJP welds. Additionally, no cracking occurred in the k-area of the columns in these column-stiffened specimens.

From the results of the pull-plate tests and analyses, it was determined that the AISC non-seismic design provisions for local web yielding and local flange bending are reasonable and slightly conservative in calculating the need for column stiffening. However, new design equations are also proposed in this research for the limit states of local flange bending and local web yielding in connections consisting of wide-flange girders framing into wide-flange columns with fully-restrained connections. When compared to experiments both from this research and past work, both equations provide superior test-to-predicted ratios as compared to current provisions. Broadening the range of experimental tests to include a wider range of section sizes and loading configurations would strengthen the assessment of these new equations.

For the cruciform specimens, the measured maximum column flange deformation due to the concentrated girder flange force in the unstiffened specimens ranged from 26% of the assumed yield mechanism limit of 1/8 in. flange deformation in the case of Specimen CR1, to 49% of the 1/8 in. limit in the case of Specimen CR5. Specimen CR5 was the most substantially underdesigned specimen for local flange bending the resistance-to-demand ratio was only 0.84 for non-seismic demand as per AISC (1999a), and it equaled to 0.47 for seismic demand as per AISC (1992). Specimen CR5 met the requirements for two cycles at 4.0% interstory drift. Continuity plates may thus not be necessary in many interior columns in steel moment connections, and design provisions similar to those in AISC (1992) or FEMA (2000a) permitting the design, or lack of inclusion, of continuity plates are recommended for consideration for reintroduction into the AISC Seismic Provisions (note however that further research may be warranted to investigate the behavior of deep columns and other pre-qualified connection details for the limit state of local flange bending). Specimens CR1, CR2, CR4R, and CR5, none of which had continuity plates (although Specimen CR4R included the offset doubler plate detail), showed ductile connection behavior even though only Specimen CR1 met the seismic requirements of AISC (1992) and FEMA (2000a) with respect to continuity plates for the limit state of local flange bending. Specifically, it is recommended that the provisions for LFB strength of AISC (1992, 1999a) [Equation (3)] be adopted for seismic design, and that the calculation for LFB demand according to AISC (1999c) [Equation (6)] is conservative and may be adopted for use; reducing this LFB demand by a factor of 0.85 for interior connections is also recommended for consideration.

For a wide range of column sections and doubler plate detailing, strain gradients and strain magnitudes well above the yield strain in the girder flanges did not prohibit the specimens from achieving the connection prequalification requirement of completing two cycles at 4.0% interstory drift without significant strength degradation. This was even the case for one specimen that had notch toughness in the weld metal that was significantly below the requirements in the FEMA (2000a, 2000b) guidelines. These results indicate that the column reinforcement detailing may not have a significant effect on the potential for brittle fracture at the girder flange-to-column flange weld. Note that this is contrary to previous finite-element analyses reported in the literature using theoretical fracture criteria that have predicted a significant effect of using or omitting continuity plates.

If continuity plates are required, fillet-welded continuity plates that were approximately half of the girder flange thickness performed well. The results showed that only minor local yielding occurred in these continuity plates in part of the most stressed cross sections at peak drift level and that these strains were not sufficient to cause cracking or distortion in the continuity plate or to change the strain gradients in the girder flange substantially. Since continuity plates do not significantly yield, it may not be necessary to size the welds large enough to develop the continuity plate. Rather the weld and the plate may only need to be designed for the difference between the demand and the capacity of the column shape without the continuity plate.

18

The offset doubler plate detail was also found to function effectively as continuity plates while simultaneously serving as web doubler plates.

In all the five successful cruciform tests, the seismic performance of the relatively weak panel zones was stable and ductile, and the panel zones exhibited good energy dissipation. While the plastic moment was achieved in the five WUF-W specimens tested in this work, no full-depth girder plastic hinge formation was observed up to the 4.0% interstory drift cycles. With the weaker panel zones in these five specimens, a major portion of the girder yielding was observed near the girder flange-to-column flange boundary area even in Specimens CR1 and CR2, which satisfied the Strong Column-Weak Beam (SCWB) criteria of the 2002 AISC Seismic Provisions (AISC, 2002). The girder webs yielded only locally at the higher interstory drift cycles, just before the low cycle fatigue fracturing in the girder flange-to-column flange groove welds occurred in all specimens. Observation of the yielding in the joint region at the higher interstory drift levels indicated that it is more realistic to calculate the panel zone design demand directly at the column face for connections that have weaker panel zones, without considering the moment increments due to the girder shear force at the assumed girder plastic hinge location.

The AISC seismic panel zone shear strength equation from the 2002 AISC Seismic Provisions (AISC, 2002) [Equation (8)] overestimated the panel zone shear strengths in the WUF-W specimens tested in this work when compared with the experimental panel zone responses at 4y of shear deformation, which is implied in Equation (8) as the design shear deformation of the panel zone. A new panel zone design shear strength equation is thus presented and compared to the results from this work and to other experiments from the literature that have a wide range of panel zone characteristics. The new equation provided a more accurate prediction of panel zone strength. A resistance factor of 0.85 is recommended for use with the equation. If used with the panel zone shear demand of AISC (2002), the new equation likely provides thicker panel zones than the design procedure of AISC (2002). Future work should be conducted to investigate more comprehensively the corresponding panel zone shear demand that is best used in conjunction with the proposed equation.

These experimental results indicate that, with improved connection detailing and notch-tough weld material, more than 4y may be assumed as the acceptable maximum panel zone shear deformation in prequalified steel moment-resisting connections, even in the presence of higher stress and strain demands due to weaker panel zones. Lee et al. (2004) present a new panel zone shear demand equation for use with the WUF-W and similar connections that are to be designed with weaker panel zones.

Within the limited number of WUF-W experimental study, a strong correlation between the panel zone strength and fracturing of the shear tab welded to the column flange at its top and bottom edges was not observed. Instead, fracturing in the shear tab edges seems to be more directly affected by girder flange local buckling, LCF crack opening in the girder flanges, or both, under large connection deformations. Local buckling in the bottom girder flange, for example, can increase the inelastic demand in the top girder flange, which may cause the initial crack at the shear tab top edge.

As a result of this study, the following additional research is recommended. First, these conclusions should be validated for deep columns. Second, continuity plates with undersized fillet welds should be tested to confirm that the weld need not develop the full continuity plate strength. Third, because of the possibility of obtaining brittle weld metal despite the fact that weld certifications show adequate toughness, a study should be conducted to fully characterize the typical variability in the CVN and other properties of the weld. An evaluation of the need for lot testing should be performed. Consideration should also be given to use of filler metals with a distribution of CVN such that there is a sufficiently small probability of not meeting the minimum required values, and therefore lot testing may not be required. Finally, the local flange bending and local web yielding criteria, if they are to be adopted, should be validated for a wider range of concentrated loading cases found in typical practice.

ACKNOWLEDGEMENTS This research was sponsored by the American Institute of Steel Construction, Inc. and by the University of Minnesota. In-kind funding and materials were provided by LeJeune Steel Company, Minneapolis, Minnesota; Dannys Construction Company, Minneapolis, Minnesota; Braun Intertec, Minneapolis, Minnesota; Nucor-Yamato Steel Company, Blytheville, Arkansas; Lincoln Electric Company, Cleveland, Ohio; and Edison Welding Institute, Cleveland, Ohio. Supercomputing resources were provided by the Minnesota Supercomputing Institute. The authors thank S. C. Cotton, S. D. Ojard, and Y. Ye for their extensive research contributions to this project. The

19

authors also thank T. V. Galambos, P. M. Bergson, and J. C. Nelson of the University of Minnesota, L. A. Kloiber of LeJeune Steel Corporation, and the members of the external advisory group on this project for their valuable assistance.

REFERENCES American Institute of Steel Construction (AISC) (1937). A Manual for Architects, Engineers and Fabricators of Buildings and Other Steel Structures, Third ed., AISC, New York, New York.

AISC (1989). Manual of Steel Construction Allowable Stress Design, Ninth ed., AISC, Chicago, Illinois.

AISC (1992). Seismic Provisions for Structural Steel Buildings, AISC, Chicago, Illinois.

AISC (1997a). Seismic Provisions for Structural Steel Buildings, AISC, Chicago, Illinois.

AISC (1997b). AISC Advisory on Mechanical Properties Near the Fillet of Wide Flange Shapes and Interim Recommendations, January 10, 1997, Modern Steel Construction, AISC, Chicago, Illinois, February, p. 18.

AISC (1999a). Load and Resistance Factor Design Specification for Structural Steel Buildings, AISC, Chicago, Illinois.

AISC (1999b). Seismic Provisions for Structural Steel Buildings - Supplement No. 1, AISC, Chicago, Illinois.

AISC (1999c). Stiffening of Wide-Flange Columns at Moment Connections: Wind and Seismic Applications, AISC Design Guide No. 13, AISC, Chicago, Illinois.

AISC (2001a). Seismic Provisions for Structural Steel Buildings - Supplement No. 2, AISC, Chicago, Illinois.

AISC (2001b). AISC Advisory on Changes to T, k, and k1 Dimensions for W-Shapes, February 14, 2001, AISC, Chicago, Illinois.

AISC (2002). Seismic Provisions for Structural Steel Buildings, AISC, Chicago, Illinois.

American Society of Testing and Materials (ASTM) (1998). ASTM A6/A6M: Standard Specification for General Requirements for Rolled Structural Steel Bars, Plates, Shapes, and Sheet Piling, ASTM, Conshohocken, Pennsylvania.

American Welding Society (AWS) (1995). Specification for Carbon Steel Electrodes for Flux Cored Arc Welding, AWS A5.20-95, AWS, Miami, Florida.

AWS (2000). Structural Welding Code Steel, AWS D1.1-2000, AWS, Miami, Florida.

Bartlett, F.M., R.J. Dexter, M.D. Graeser, J.J. Jelinek, B.J. Schmidt, and T.V. Galambos (2003). Updating Standard Shape Material Properties Database for Design and Reliability, Engineering Journal, AISC, Vol. 40, No. 1, 1st Quarter, pp 2-14.

Becker, R. (1975). Panel Zone Effect on the Strength and Stiffness of Steel Rigid Frames, Engineering Journal, AISC, Vol. 12, No. 1, pp. 19-29.

Bertero, V. V., Krawinkler, H., and Popov, E. P. (1973). Further Studies on Seismic Behavior of Steel Beam-Column Subassemblages, Report No. EERC-73/27, Earthquake Engineering Research Center, University of California, Berkeley, California.

Biddah, A. and Heidebrecht, A. C. (1998). Seismic Performance Moment-Resisting Steel Frame Structures Designed for Different Levels of Seismic Hazard, Earthquake Spectra, EERI, Vol. 14, No. 4, pp. 597-627.

Biddah, A. and Heidebrecht, A. C. (1999). Evaluation of the Seismic Level of Protection Afforded to Steel Moment Resisting Frame Structures Designed for Different Design Philosophies, Canadian journal of Civil Engineering, Vol. 26, pp. 35-54.

Bruneau, M., Uang, C., and Whittaker, A. (1998). Ductile Design of Steel Structures, 1st Ed., McGraw-Hill, New York, New York.

20

Charney, F. A. and Johnson, R. (1986). The Effect of Panel Zone Deformations on the Drift of Steel Framed Structures, Proceedings of the Fourth ASCE Structures Congress 86, ASCE, Reston, Virginia.

Dexter, R. J., Hajjar, J. F., Cotton, S. C., Prochnow, S. D, and Ye, Y. (1999). Reassessment of Design Criteria and New Alternatives for Column Transverse Stiffeners (Continuity Plates) and Web Doubler Plates: Interim Report, Report No. ST-99-3, Department of Civil Engineering, University of Minnesota, Minneapolis, Minnesota.

Dexter, R. J. (2000). Structural Shape Material Property Survey, Final Report to Structural Shape Producer's Council, University of Minnesota, Minneapolis, Minnesota.

Dexter, R. J., Hajjar, J. F., Prochnow, S. D., Graeser, M. D., Galambos, T. V., and Cotton, S. C. (2001). Evaluation of the Design Requirements for Column Stiffeners and Doublers and the Variation in Properties of A992 Shapes, Proceedings of the North American Steel Construction Conference, AISC, Chicago, Illinois.

El-Tawil, S., Mikesell, T., Vidarsson, E., and Kunnath, S. (1998). Strength and Ductility of FR Welded-Bolted Connections, Report No. SAC/BD-98/01, SAC Joint Venture, Sacramento, California.

El-Tawil, S., Vidarsson, E., Mikesell, T., and Kunnath, S. K. (1999). Inelastic Behavior and Design of Steel Panel Zones, Journal of Structural Engineering, ASCE, Vol. 125, No. 2, pp. 183-193.

El-Tawil, S. (2000). Panel Zone Yielding in Steel Moment Connections, Engineering Journal, AISC, Vol. 37, No. 1, pp. 120-131.

Engelhardt, M. D., Shuey, B. D., and Sabol, T. A. (1997). Testing of Repaired Steel Moment Connections, Proceedings of the ASCE Structures Congress, Kempner, L. and Brown, C. (eds.), Portland, Oregon, April 13-16, 1997, pp. 408-412.

Engelhardt, M. D. (1999). Design of Reduced Beam Section Moment Connections, Proceedings of the AISC North American Steel Construction Conference, Toronto, Ontario, May 19-21, 1999, AISC, Chicago, Illinois.

Federal Emergency Management Agency (FEMA) (1995). Interim Guidelines: Inspection, Evaluation, Repair, Upgrade an