You are on page 1of 17

Journal of Constructional Steel Research 55 (2000) 325–341

www.elsevier.com/locate/jcsr

Critical issues in achieving ductile behaviour of


welded moment connections
Le-Wu Lu *, James M. Ricles, Changshi Mao, John W. Fisher
Department of Civil and Environmental Engineering, Lehigh University, Bethlehem, PA 18015, USA

Abstract

A comprehensive research program on large-size, welded moment connections conducted


after the 1994 Northridge earthquake in California has identified three critical issues that could
significantly affect the strength and ductility of the connections. They are: fracture toughness
of weld metal, geometry and size of weld access hole, and control of panel zone deformation.
Experimental and analytical studies on each of the issues have been performed at the Center
for Advanced Technology for Large Structural Systems at Lehigh University. This paper
presents brief descriptions of the studies, key results obtained and recommendations made to
insure ductile performance of the connections.  2000 Elsevier Science Ltd. All rights
reserved.

Keywords: Connection; Moment-resistant; Welding; Fracture; Access hole; Panel zone; Ductility

1. Introduction

The January 1994 Northridge earthquake in southern California caused widespread


damages to welded beam-to-column connections in steel moment-resisting framed
structures [1]. The type of connections used in these structures is the popular “flange
welded and web bolted” connections, as shown in Fig. 1. The beam flanges are
welded to the column flange with full penetration welds and the beam web is bolted
to the shear tab, which is shop welded to the column flange. In some cases, sup-
plementary welds are provided between the shear tab and web. Access holes and
back-up bars, which generally remain in place, are needed to make the field flange
welds. This connection has been “prequalified” as a ductile connection based on the

* Corresponding author. Tel.: +610-758-3537; fax: +610-758-6405.


E-mail address: lwl@lehigh.edu (L.-W. Lu).

0143-974X/00/$ - see front matter  2000 Elsevier Science Ltd. All rights reserved.
PII: S 0 1 4 3 - 9 7 4 X ( 9 9 ) 0 0 0 9 2 - 9
326 L.-W. Lu et al. / Journal of Constructional Steel Research 55 (2000) 325–341

Fig. 1. Typical “pre-Northridge” welded moment connection.

results of previous tests on moderate size laboratory specimens. It has been widely
adopted since the late 1960’s. Shortly after the earthquake, in a routine inspection
of a building under construction, cracks initiating from beam flange welds were
found in several connections. In some cases the cracks also extended into the column.
Subsequent inspections found similar damages in more than 100 low- and high-rise
buildings in the Los Angeles area.
Since the earthquake, many laboratory and field investigations have been carried
out on welded connections with the objectives to (1) determine the primary cause
of the observed fracture, (2) examine, in a systematic manner, the various issues that
can significantly affect the performance of the connections, (3) develop cost-effective
schemes to repair the damaged connections, and (4) explore new or alternative con-
nection designs which will have good ductility for use in seismic-resistant steel struc-
tures. Several of these investigations have been conducted at the Center for Advanced
Technology for Large Structural Systems (ATLSS) at Lehigh University. Among the
issues that have been examined, the following three are considered to have dominant
effects on the behavior of the connections: properties of weld metal, geometry and
size of the access holes, and the amount of panel zone deformation. The studies
conducted on these issues and some key results obtained are presented in this paper.

2. Fracture toughness of weld metal

A preliminary investigation of some of the damaged connections indicates that


the deposited weld metal of the fractured flanges had unusually low fracture tough-
ness, although the static strength properties of the metal met the specification require-
ments [2]. The welds were made using the AWS E70T-4 self-shielded, flux-cored
wire, which is for non-impact applications. The E70T-4 electrode has a typical
Charpy V-notch fracture toughness of 6.5 to 20 Joules at 21°C. The toughness of
L.-W. Lu et al. / Journal of Constructional Steel Research 55 (2000) 325–341 327

Fig. 2. CVN test data for various filler metals.

this and other electrodes are shown in Fig. 2 as a function of temperature. It is


apparent that the E70T-4 electrode has the lowest toughness amongst the electrodes
in Fig. 2, which include E70TG-K2, E71T-8, E7018, and E70T-7. Four full-size
beam-to-column assemblies were tested dynamically to study the effect of weld metal
toughness on connection performance and to establish a probable cause of the
observed fracture. Fig. 3 shows a schematic of the test assembly laid on the labora-

Fig. 3. Connection test assembly (A-series).


328 L.-W. Lu et al. / Journal of Constructional Steel Research 55 (2000) 325–341

tory floor and the test setup used. The column was a W14×311, 3.89 m long, and
made of A572 Gr.50 steel with a yield stress of 400 MPa (58 ksi). It was supported
by a pin at its bottom and a roller at its top. The W36×150 beam, approximately 3.05
m long, was made of A36 steel with a yield stress of 262 MPa and was connected to
the column at the mid-height. Fully reversed dynamic displacement cycles were
applied at the free end of the beam by two servo-controlled actuators at pre-determ-
ined rates. Two issues were considered in determining the rates. One was the
response frequency of a typical multi-story steel structure and the other was the
appropriate kinetic deformation rate of steel in the plastic range.
Specimen A-1 was detailed to represent a typical connection fractured during the
Northridge earthquake. The details were similar to those shown in Fig. 1, but without
the continuity plates and added shear tab weld, and the low toughness E70T-4 wire
was used in making the flange groove welds. Ten high strength A325 bolts, 25.4
mm diameter, were used to connect the beam web. Specimen A-2 was the same as
A-1, but with the back-up bar removed and a small reinforcing fillet weld (AWS
E71T-8 wire) added to the weld root. This was an attempt to eliminate the fracture
initiation potential of the back-up bar gap. Specimens A-3 and A-4 had the details
of a fully welded connection fabricated with the highly tough AWS E7018 electrode
and the moderately tough E70TG-K2 flux-cored wire, respectively.
Specimen A-1 fractured at the bottom flange connection in a brittle manner. The
maximum bending moment achieved during the test was only 87% of the plastic
moment capacity of the beam. The fracture occurred when both the beam and column
were still in the elastic range, because no visible sign of yielding was observed,
except some flaking of the whitewash in small areas around the access holes. After
testing, a careful examination was made of the crack surface. The point of crack
initiation was at the incomplete fusion flaw contiguous with the weld backing. A
fractographic analysis indicated that the critical fracture initiated within the weld
metal. The condition for the crack initiation has been analyzed and reported in a
previous paper [2].
The removal of the back-up bars in Specimen A-2 improved the connection per-
formance, but brittle fracture of the flange welds again led to failure. Other than the
limited yielding around the access holes, the connection behaved elastically when
the welds of the top and bottom flanges fractured almost simultaneously during a
reversed loading cycle. The fracture surface was confined entirely in the weld. The
maximum bending moment resisted by the connection was 92% of the plastic
moment capacity of the beam. The load vs displacement relationship (at the load
point) of Specimen A-2 is shown in Fig. 4.
Much improved performance in terms of both strength and ductility was observed
in Specimen A-3. A typical “moment gradient” plastic hinge developed in the beam
near the column face with extensive yielding occurring in the beam flanges as well
as the web. There was also some yielding of the panel zone. The load vs displacement
hysteresis loops of the specimen are presented in Fig. 5. Fig. 6 is a photograph taken
after the application of large amplitude displacement cycles, showing yielding and
local buckling at the hinge region. The maximum displacement applied was 92.5
mm, and the corresponding maximum bending moment at the face of the column
L.-W. Lu et al. / Journal of Constructional Steel Research 55 (2000) 325–341 329

Fig. 4. Hysteresis loops of A-2.

Fig. 5. Hysteresis loops of A-3.

was 1.22 times the plastic moment capacity of the beam. The plastic hinge rotation
reached was about 0.026 radian (calculated by assuming that the hinge is located at
a distance d/2 from the column surface, where d is the depth of the beam). A crack
initiating at the flange tip and extending in a stable manner across the width of the
flange caused eventual failure of the specimen.
Specimen A-4 fabricated with E70TG-K2 flux-cored wire also showed ductile
behavior and achieved a plastic rotation of 0.037 radian. A more detailed discussion
of the four connection tests can be found in Refs. 3 and 4.
330 L.-W. Lu et al. / Journal of Constructional Steel Research 55 (2000) 325–341

Fig. 6. Yielding and local buckling in A-3.

The superior performance of Specimens A-3 and A-4 confirms that the premature
fracture of Specimens A-1 and A-2 was due to the low fracture resistance of the
E70T-4 weld metal. It is therefore necessary to impose a minimum fracture toughness
requirement on the weld metal used in order to insure ductile behavior. A Charpy
V-notch fracture toughness of 27 Joules at ⫺29°C was imposed by the AISC [7]
in 1997.

3. Geometry and size of access holes

Specimens A-3 and A-4 provided good plastic deformation and energy dissipation
capacity. Near the end of the tests, small cracks were observed at the weld toes of
the beam flange and web juncture, as shown in Fig. 7. These cracks were, however,
very stable and did not propagate rapidly. Similar weld access hole fractures have
also been reported in the post-Northridge earthquake inspections and in laboratory
tests conducted at the University of Texas, Austin [5] and the University of Michigan
(results not yet published), where conventional access hole configurations were
employed in the connections.
Finite element analysis conducted by El-Tawil et al. [6] has shown that small
access holes result in less strain concentration around the holes. However, the
reduced access hole tends to increase the size of the weld defects resulting from
incomplete fusion at the root of the flange groove weld. For the lower flange, the
critical location of major defects often occurs near the access hole, where fracture
may initiate. A detailed analytical study was therefore undertaken to examine the
effect of the geometry and size of the access hole on the fracture potential of the
material near the hole. The particular beam-and-column assembly on which the study
was conducted is shown in Fig. 8. The assembly consists of a W36×150 beam having
L.-W. Lu et al. / Journal of Constructional Steel Research 55 (2000) 325–341 331

Fig. 7. Stable crack at access hole of A-4.

Fig. 8. Connection test assembly (T-series).


332 L.-W. Lu et al. / Journal of Constructional Steel Research 55 (2000) 325–341

a span length of 4.49 m and made of A572 Gr.50 steel with a yield stress of 386
MPa and a 3.96 m W14×311 column made of the same grade steel but with a lower
yield stress of 324 MPa. A lateral load is applied to the column at the height as
indicated. Four such assemblies (T-series) have been tested in a subsequent experi-
mental program investigating the details of connecting the beam web to the column
flange (direct welding or through bolted shear tab, with and without added web
weld). Fig. 9 shows the connection details of Specimen T-1. Computer models of
the test assembly were developed using the general-purpose nonlinear finite element
analysis (FEA) program ABAQUS. The three-dimensional finite element model is
shown in Fig. 10. The FEA results are very sensitive to the type of elements used
and mesh size and orientation. A mesh convergence study was first made for the
purpose of selecting a suitable brick element and to examine the level of mesh
refinement. The study included three types of an 8-node brick element: (1) brick
element C3D8I with an incompatible deformation mode; (2) brick element C3D8R
with reduced integration; and (3) regular brick element C3D8. The results indicate
that the C3D8 element can produce good results and also that a fine mesh can provide
higher peak values than a coarse mesh. Therefore, it is important to employ a fine
local mesh near the access hole. A sub-modeling technique was applied to the access
hole region in order to obtain results that were sufficiently accurate to allow close
examinations of the ductile fracture potential of the various access hole configur-
ations and sizes. The sub-model analysis utilized directly the results of the global
analysis. The global model used 14 elements across the 305 mm wide beam flange
and 2 elements through the 23.9 mm thick beam flange. The mesh was comprised

Fig. 9. Connection details of specimen T-1.


L.-W. Lu et al. / Journal of Constructional Steel Research 55 (2000) 325–341 333

Fig. 10. Three-dimensional finite element model.

of approximately 7200 brick elements and 13,100 nodes; it had 37 000 degrees of
freedom. The sub-model had 26 elements and 6 elements through the width and
thickness of the beam flange, respectively. Six elements were used through the thick-
ness of the beam web. The mesh was comprised of approximately 6200 brick
elements, 7800 nodes, and 24 000 degrees of freedom. Geometric and material non-
linearities were included in the analyses. The assembly was loaded horizontally in
the manner shown in Fig. 8 and the analysis continued until a beam plastic rotation
of 0.03 radian was reached.
Nine different weld access hole configurations, as shown in Fig. 11, were investi-
gated. Note that all the dimensions in the figure are given in the Imperial unit (in),
as currently practiced in the U.S. The configurations are as follows:

1. standard access hole (minimum size permitted by AISC [7])


2. no access hole (as a basis of comparison)
3. 51 mm (2 in) long standard access hole
4. 127 mm (5 in) long standard access hole
5. 76 mm (3 in) long access hole, Type-I
6. 76 mm (3 in) long access hole, Type-II
7. 76 mm (3 in) long access hole, Type-III
8. 51 mm (2 in) long access hole, Type-I
9. 51 mm (2 in) long access hole, Type-II

The standard access hole is the minimum hole size for rolled sections permitted
by the AISC-LRFD specification. The diameter of the circular portion of the hole
is 19 mm with a length equal to 1.5 times the thickness of the beam web. The size
and geometry of the other configurations were developed from the standard hole
with the intent to minimize strain concentration and reduce the potential for fracture
of the beam flange near the hole region. The diameter of the holes in all the configur-
334 L.-W. Lu et al. / Journal of Constructional Steel Research 55 (2000) 325–341

Fig. 11. Weld access hole configurations.

ations was 19 mm, which is probably the smallest diameter that still permits proper
welding of the beam flanges.
Fig. 12 shows the results of longitudinal stresses, σ11, along the centerline of the
top and bottom faces of the tension flange. These results have been normalized by
the yield stress Fy of the beam material. The results from only some typical con-
figuration cases are given in the plot. Configuration (1) has the highest longitudinal
stresses. Fig. 13 gives the normalized longitudinal stresses, σ11/Fy, and the effective
plastic strain index, PEEQ/εy (see Table 2 for the definition of PEEQ, εy=yield strain
of beam material), along the direction of the flange width at the root of the access
hole (point A in Fig. 11). In configurations (1), (2), (3), (4), and (5) the maximum
PEEQ occurs at the root of the hole, while in configurations (6), (7), (8), and (9)
the maximum is at a point on the semi-circle of the access hole. The maximum
PEEQ locations are shown in Fig. 11. Fig. 14 summarizes the maximum PEEQ
indices of all the nine configurations studied. Note that the PEEQ index of the stan-
dard hole (configuration (1)) is twice that of configuration (6), which appears to be
the best configuration. This configuration is recommended for use in seismic-resistant
structures and has been adopted in fabricating the T-series test assemblies, referred
to previously. This series of tests have already been completed and there has been
no problem of fracture near the access holes.
L.-W. Lu et al. / Journal of Constructional Steel Research 55 (2000) 325–341 335

Fig. 12. Longitudinal stresses along centreline of tension flange of top and bottom faces.

4. Control of panel zone deformation

The T-series test specimens were fabricated using E70TG-K2 wire for the flange
connections and E71T-8 wire for the web connection, as shown in Fig. 9. The
strength and fracture toughness properties of the two wires are similar. The access
holes had the new configuration (i.e., configuration 6) as described before. In these
specimens, a relatively large portion of the total plastic deformation of the assembly
was due to panel zone deformation (weak panel). The assemblies were tested by
applying a series of increasing lateral displacement cycles to the column, as shown
in Fig. 8. The lateral displacements applied correspond to certain pre-selected story
drifts (defined as the ratio of the lateral displacement to the story height, expressed
as percent). Each test assembly was carefully instrumented to permit separate deter-
minations of the beam, column and panel zone rotations.
In Specimen T-1, the shear tab was shop welded to the column flange using a
fillet weld. The beam web was welded directly, in a vertical position, to the column
flange with a full-length groove weld. Supplementary fillet weld was then applied
336 L.-W. Lu et al. / Journal of Constructional Steel Research 55 (2000) 325–341

Fig. 13. Longitudinal stresses and PEEQ indices transverse through root of access hole.

Fig. 14. Maximum PEEQ indices at access hole.


L.-W. Lu et al. / Journal of Constructional Steel Research 55 (2000) 325–341 337

continuously to the edges of the shear tab (Fig. 9). In testing this specimen, a new
fracture problem appeared. During the application of one of the 2% drift cycles,
cracks were observed initially at the top and bottom edges of the fillet weld con-
necting the shear tab to the column flange. The beam web groove weld at the edges
also started to crack. The cracks grew in size and propagated vertically in the sub-
sequent cycles of testing (see Fig. 15). The beam flanges near the column buckled
locally and the panel zone showed extensive yielding during the 3% drift cycles.
Ductile fracture started to develop at an edge of the bottom flange in the base metal
(heat-affected zone) when the drift reached 5% and eventually caused complete sep-
aration of the flange from the column. The vertical crack in the web weld at the
bottom had extended to about 76 mm when this happened. Fig. 16 shows the yielding
in the panel zone, local buckling of the flanges and fracture of the web weld and
the beam bottom flange. The load vs displacement hysteresis loops of the specimen
are presented in Fig. 17. The maximum bending moment at the column face was
1.36 times the plastic moment capacity of the beam. The total plastic story drift
reached was 0.035 radian, 70% of which was due to panel zone shear deformation.
The beam plastic rotation was about 0.01 radian. A comparison of the results of this
test with those of the A-series specimens, which had a relatively strong panel zone
would suggest that the cause of the early web weld fracture was due to the excessive
panel zone deformation in Specimen T-1. To provide a confirmation to this, a detailed
study was made of the effect of panel-zone strength and stiffness on the magnitude

Fig. 15. Beam bottom flange base metal and web weld fracture.
338 L.-W. Lu et al. / Journal of Constructional Steel Research 55 (2000) 325–341

Fig. 16. Specimen T-1 after completion of testing, fracture occurred in the beam bottom flange HAZ
and beam web weld.

Fig. 17. Force-displacement response of specimen T-1.


L.-W. Lu et al. / Journal of Constructional Steel Research 55 (2000) 325–341 339

Fig. 18. Critical locations and the initiation of ductile fracture.

and distribution of stress and strain and the ductile fracture potential at the edge of
the shear tab fillet weld (Fig. 18(a)) and at the critical locations of the beam bottom
flange (Fig. 18(b)). The study used Specimen T-1 as the reference or benchmark.
Two ways to increase the strength and stiffness of the panel zone were included.
One was to add doubler plates to the panel; the other was to use a column of higher
yield stress (higher than 324 MPa, the measured yield stress of the T-1 column).
The following discussion pertains to the case when a 13 mm thick doubler plate
with a yield stress of 345 MPa (50 ksi) was added to the column web panel.
Finite element analyses were performed on the original T-1 and a modified T-1.
In the analyses, a monotonic horizontal load causing tension in the beam bottom
flange was applied to the column top to reach a drift of 5%, assuming no fracture
occurring anywhere in the assemblies. Selected results are presented in Tables 1 and
2. Comparisons of the effective plastic strain, PEEQ, Von Mises stress, σe, hydro-
static stress, σm, and rupture index, RI, at the edge of the shear tab weld and at the
two critical locations on the interface between the weld and base metal of the bottom
flange (edge and center) are shown in these tables, where the four quantities are
also defined. The rupture index, as defined, was based on the work of Hancock and
MacKenzie [8] on ductile fracture in steel under multi-axial state of stresses. Table
1 shows that the PEEQ and the rupture index at the edge of the shear tab weld are
decreased by more than 50% when the panel zone is strengthened by addition of
the doubler plate. This could significantly reduce the potential of cracking developing
in the web weld, as observed in the T-1 test. The rupture index at the two critical
locations of the beam flange are also reduced by the doubler plate. The contribution

Table 1
Stress and strain at the bottom edge of shear tab (Specimen T1)

Case Effective plastic Von Mises stress Hydrostatic stress Rupture index
strain PEEQ σe (MPa) σm (MPa)

No doubler plate 0.0925 703 917 0.654


With doubler 0.0417 627 807 0.285
340 L.-W. Lu et al. / Journal of Constructional Steel Research 55 (2000) 325–341

Table 2
Stress and strain at the interface of base and weld metals (Specimen T-1)a

Case Flange edge Flange center

PEEQ σe σm Rupture PEEQ σe σm Rupture


(MPa) (MPa) index (MPa) (MPa) index

No doubler plate 0.0816 490 193 0.147 0.0565 469 276 0.137
With doubler plate 0.0848 496 172 0.143 0.0458 427 276 0.121

a
Effective plastic strain, PEEQ=√2/3⑀pij ⑀pij ; Von Mises stress, se=√2/3Sij Sij ; Hydrostatic stress, sm=
PEEQ
⫺1/3trace(σij ). Rupture Index, RI= . Plastic strain, εp. Deviatoric stress, S.
sm
exp(−1.5 )
se

of the panel zone deformation to the overall plastic rotation of the connection is
reduced from 70% to 30%.
The results of this study suggest that it is important to control the amount of panel
zone deformation in order to avoid early occurrence of fracture of the web weld. A
tentative recommendation is to limit the rotation contributed by the panel zone to
50% of the total connection rotation. Further studies of this issue are required and
on-going.

5. Conclusions and recommendations

This paper has identified three critical issues that will have a strong effect on the
ductility of welded moment connections and which should be carefully considered
in design. The issues are: (1) fracture toughness of deposited weld metal, (2)
geometry and size of weld access holes, and (3) control of panel zone deformation.
They have been studied in detail in the research programs conducted at the ATLSS
Center of Lehigh University after the 1994 Northridge earthquake.
The key conclusions and recommendations may be summarized as follows:

1. The results of the A-series tests indicate that the type of connection fracture
observed after the Northridge earthquake is primarily due to the low toughness
of the weld metal used in making the flange welds. A Charpy V-notch toughness
comparable to that of E70TG-K2 wire, 47 Joules at ⫺29°C, should be adequate
for properly enclosed building structures.
2. In some connection assembly tests, ductile fracture initiating at the edges of weld
access holes was observed. An effort was made to study the influence of access
hole geometry and size on the potential of ductile fracture initiation near the holes.
Nine different access hole configurations were included in the study and results
indicate the importance of selecting a proper hole configuration. The particular
configuration shown as configuration (6) in Fig. 11 is recommended. This con-
L.-W. Lu et al. / Journal of Constructional Steel Research 55 (2000) 325–341 341

figuration was adopted in fabricating the T-series test assemblies. No fracture near
the access holes was observed during the tests.
3. In some of the T-series tests, initial fracture occurred in the vertical welds con-
necting the shear tab or beam web to the column flange. The test specimens in
this series had relatively weak column panels. The problem of web weld fracture
can be controlled by limiting the amount of panel zone deformation. In some
applications, this would require strengthening of the panel zone. A tentative rec-
ommendation limiting the connection rotation contributed by panel zone defor-
mation to 50% of the total rotation has been made.

The results and conclusions presented are based on studies on exterior connection
assemblies. Further research is necessary to study the behavior of interior connections
as well as connections having different details. Additional issues may emerge from
this research.

Acknowledgements

The research described in this paper was supported by grants from the U.S.
National Science Foundation, the SAC Joint Venture, and the Department of Com-
munity and Economic Development of the Commonwealth of Pennsylvania through
the Pennsylvania Infrastructure Technology Alliance. Dr. Eric J. Kaufmann of the
ATLSS Center, a specialist in welding metallurgy, was most helpful in providing
insights to some of the welding related issues.

References

[1] Youssef N, Bonowitz D, Gross JL. A survey of steel moment-resisting frame buildings affected by
the 1994 Northridge earthquake. NISTIR 5625, National Institute of Standards and Technology, 1995.
[2] Lu LW, Xue M, Kaufmann EJ, Fisher JW. Cracking, repair and ductility enhancement of welded
moment connections. Proceedings, NEHRP Conference and Workshop on Research on the Northridge,
California Earthquake of January 17, 1994, Vol. III, pp. 637–646, Los Angeles, 1997.
[3] Kaufmann EJ, Xue M, Lu, LW, Fisher JW. Achieving ductile behavior of moment connections.
Modern Steel Construction, AISC, pp. 30–39, January 1996.
[4] Xue M, Kaufmann EJ, Lu LW, Fisher JW. Achieving ductile behavior of moment connections —
Part II. Modem Steel Construction, AISC, pp. 38–42, June 1996.
[5] Engelhardt M, Sabol TA. Reinforcing of steel moment connections with cover plates: benefits and
limitations. Engineering Structures 1998;20(4-6):510–20.
[6] El-Tawil S, Mikesell T, Vidarsson E, Kunnath SK. Strength and ductility of FR welded-bolted connec-
tions. Report SAC/BD — 98/01, SAC Joint Venture, 1998.
[7] American Institute of Steel Construction (AISC). Seismic provisions for structural steel buildings.
1997.
[8] Hancock JW, MacKenzie AC. On the mechanisms of ductile failure in high-strength steels subjected
to multi-axial stress states. Journal Mech Phys of Solids 1976;24:147–69.

You might also like