Professional Documents
Culture Documents
Investigation On The Weld Damage Behavior of Steel Beam-To-column Connection
Investigation On The Weld Damage Behavior of Steel Beam-To-column Connection
DOI 10.1007/s13296-015-0070-8
ISSN 1598-2351 (Print)
ISSN 2093-6311 (Online)
www.springer.com/journal/13296
Abstract
A large number of steel moment frames fractured at the welds of beam-to-column connections in earthquake. As the weld
damage behavior is a crucial factor for aseismic performance of steel frame connections, 20 local welded connections
representing beam-to-column connections were tested under monotonic and cyclic loads to study the weld damage behavior
in this paper. The failure modes and deformation curves were recorded; the effects of material strength, load types, and beam
geometries on the connection damage behavior were analyzed. Three damage evolution models were calibrated, the relation
between weld damage and macro connection performance was established. The results indicate weld cracking and plate
buckling are the main causes of the connection damage; the range of cyclic loads is the key factor affecting the weld damage
level; the plastic strain based damage model illustrate the damage process with good accuracy. This study provide technical
basis for the damage evaluation and fracture prevention of the welded steel beam-to-column connections.
Keywords: weld damage, steel connection, hysteretic performance, damage model, welded joint, seismic load
holding parts are designed to fix the specimen to loading little as the Code GB50661-2011 (2011) “Code for welding
device. The test parts are T-stub connections, the beam of steel structures” required.
flange width W=80~120 mm, the width-to-thickness ratio To investigate the influences of the material strength,
W/t=2.83~4.5, which satisfies the requirement of local beam geometries and loading conditions respectively, these
buckling prevention. The local connections are shop-
welded through GMAW (CO2) welding procedure. The Table 2. Chemical compositions of the steel wt%
beam web is CJP welded to the column flange by double C Mn Si S P
grooves (Fig. 2(a)); the beam flange is welded by single
groove with backing bar left (Fig. 2(b)). The weld quality Q345 0.18 1.60 0.44 0.025 0.017
is evaluated as grade I, which means all the welds are Q390 0.16 1.50 0.41 0.002 0.010
tested with ultrasonic flaw detection and the flaws are as Q420 0.16 1.29 0.36 0.018 0.026
276 X. Y. Liu et al. / International Journal of Steel Structures, 17(1), 273-289, 2017
20 specimens are designed to nine types. The column size the first phase, the loading is controlled by loads with
of T150 mm×75 mm×12 mm×12 mm is constant, while increase of 200 kN until the specimen yielded. Then in
the steel strength, beam flange width W, beam flange the second phase, the loading is controlled by displacement
thickness t and loading procedures are varied (Table 1). with increase of 2Δy (the yield displacement) until the
The steels standard are in accordance with GB-T1591- specimen failure. There are three different cyclic loading
1994 (1994) “High strength low alloy structural steels”. types as shown in Fig. 4. In Cyclic I, the pressure and
The chemical compositions and mechanical properties of tension loads are asymmetric. While for Cyclic II, the
the tested material are listed in Tables 2 and 3. Nine pressure is controlled by loads in the second phase, the
specimens are tested under monotonic loading (SP1A~ applied cyclic pressure is 50 kN. Besides, in order to
SP9A), the other 11 specimens are tested under cyclic prevent the overall instability of specimen, the maximum
loading (SP1B~SP9B, SP1C, SP1D). pressure for Cyclic III is controlled within 4Δy. The
details of loading scheme are showed in Fig. 4.
2.2. Test setup and loading
The specimen is bolted to the self-balancing reaction 2.3. Test measurement
frame by nine high-strength bolts (18 φ 24, Gr. 10.9), and The extensometers and displacement meters are used
the load is applied by a 2000kN capacity servo-hydraulic for axial and lateral extension measurement respectively.
actuator. Figure 3 shows the experimental equipment. Two extensometers are attached to either side of the
The axial tension applied to SP-1A~SP-9A increases specimen to monitor the elongation over the central 180
linearly until the specimen failure. All the cyclic loading mm gauge length as shown in Fig. 5. The lateral displace-
procedures for SP-1B~SP-9B consists of two phases. In ment is measured by displacement meters T1-1 and T1-2.
Investigation on the Weld Damage Behavior of Steel Beam-to-column Connection 277
connections are provided in Fig. 8. three different cyclic load types are shown in Figs. 9, 10,
and 11, respectively.
3.3. Cyclic test results
3.3.1. Load capacity and deformation 3.3.3. Energy dissipation
The load capacity and deformation of specimens under The connection energy dissipation during cyclic loading
cyclic loading are listed in Table 7, where the ultimate is evaluated by the accumulated area of the hysteresis
strength is obtained by the absolute value of the maximum loops. The energy dissipation values of cyclic specimens
tension and compression loads, the ultimate displacement are listed in Table 8, as well as compared with that of the
corresponds to the specimen fracture point. corresponding monotonic specimens.
The ductility factor is defined as Eq. (1) (Han, 2004):
μ=Δu /Δy (1) 4 Discussion of the Test Results
where Δu denotes the connection ultimate deformation 4.1. Material strength effect
and Δy is the yield deformation. The material strength affects the local welded connections
strength, ductility and hysteretic behaviors, which can be
3.3.2. Hysteretic curves derived from the test results of connections SP1~SP3.
The hysteresis loops of the connections subjected to the
Investigation on the Weld Damage Behavior of Steel Beam-to-column Connection 279
4.1.1. Specimen strength seen that the connection strength is lower than the
As listed in Table 6, the yield strength of welded material strength under cyclic loading as listed in Table 7.
connection increases with the material strength increases The tested strength of connections made of Q345 and
under monotonic loading, while the ultimate strength is Q420 are lower than material strength by 10~20%.
not entirely dependent on the material strength. The Besides, similar with the results of monotonic loading,
ultimate strength of connection SP-2A made of Q390 is the connection fabricated from the high strength steel
lower than that of SP-1A made of Q345 by 0.8%. may have lower ultimate loading capacity than the
Compared the connections strength between SP-1D, ordinary steel connection. Connection SP-3B has lower
SP-2B and SP-3B which made of different material ultimate strength than SP-1D and SP-2B, which may be
strength under the same cyclic loading type, it can be correlated with the welding quality of high strength steel.
280 X. Y. Liu et al. / International Journal of Steel Structures, 17(1), 273-289, 2017
were tested under Cyclic I load. The influence of the weld ductility when the applied tensile force is similar. In
three cyclic loading procedures can be derived from the this test, Cyclic I adopts symmetrical loading, in this
test results. loading type the compressive plastic deformation is the
(1) The SP-4B, SP-1B and SP-9B subjected to Cyclic same as the tension plastic deformation. Cyclic II and III
II and Cyclic III load have better ductility than the are asymmetrical loads, the specimens are in elastic phase
connections subjected to Cyclic I load, which indicates when compressed under Cyclic II, while the maximum
that the compressive plastic deformation weakens the plastic deformation is controlled as four times the yield
282 X. Y. Liu et al. / International Journal of Steel Structures, 17(1), 273-289, 2017
the models are set up based on two damage factors: Table 11. Constant parameters of damage models evaluated
deformation (displacement, rotation, et al.) and energy. from experiments
Kravinkler and Zhorei (1983), Krawinkler and Nassar Deformation Energy Combination
(1992), Powell and Allchabadi (1988), Kato and Akiyama Damage model: model: model:
Specimen
(1975), Cosenza and Manfredi (1992) proposed the index Parameter Parameter Parameter
damage model based on the deformation respectively. C C β
Kumar and Usami (1994) proposed the energy based SP-1C 0.3 1.06 1.15 0.25
model. Park and Ang proposed the model combining the SP-1D 0.28 1.73 1.73 0.06
displacement and energy index (Park et al., 1985). SP-2B 0.17 2.24 2.15 -0.01
In this paper, three damage models are verified by the SP-3B 0.4 1.21 1.22 0.48
experimental data. In the damage models, the damage SP-5B 0.19 4.37 2.63 -0.03
index D is used (D=0 indicates no damage and D ≥1 SP-6B 0.07 3.13 2.69 -0.05
means collapse or failure). The models are listed as
Mean 2.09 1.73 0.23
follows:
COV 0.55 0.40 1.66
Cosenza and Manfredi model (Deformation model)
(Cosenza and Manfredi, 1992):
n
δi – δy ⎞ c
⎛ ------------- dissipated energy and β is constant. In this model, the
D= ∑ ⎝ δu – δ-y⎠ (4)
displacement and energy is combined linearly.
i=1
Some parameters in these models such as Emon and δ u
where δ i is the deformation amplitude of each half cycle, are calibrated by the monotonic test. In this test, four
δ y is the yield deformation, δ u is the ultimate deformation monotonic specimens SP-1A, SP-3A, SP-5A and SP-6A
under monotonic loading and c is constant. failed due to the whole section fracture, the others failed
Kumar and Usami model (Energy model) (Kumar and by web weld crack (local fracture). The ultimate displace-
Usami, 1994): ment δu and energy dissipation Emon of the connections
n
cracked at the whole section are taken as the model
Ei ⎞ c
⎛ ---------- parameters. For the connections fractured locally, these
D= ∑ ⎝ Emon⎠ (5)
two parameter values are scaled by comparing the results
i=1
of the standard specimen SP-1A (totally fractured), the
where Ei is the energy dissipated each half cycle, Emon is scaling factor is according to the ratio of specimen section
the energy dissipated under monotonic loading and c is area and material strength to the standard specimen.
constant. The damage process for the specimens subjected to
Park and Ang model (Combination model) (Park and Cyclic I load are simulated by applying the displacement
Ang, 1985): model, energy model and combination model respectively.
All the cyclic loading specimens fractured at the web
δ max β
∫
weld and partly damaged. The damage index is calculated
D = --------- + ---------- dE (6)
δu Fyδu by Eq. (3). The parameters of these damage models are
obtained by the test data, as shown in Table 10.
where δ max is the maximum deformation under cyclic The constant parameters of the three damage models
loading, δ u is the ultimate deformation under monotonic are evaluated by test data fitting. Some conclusions can
∫
loading, F y is the yield strength, dE is cumulative be drawn from the fitting results as follows:
286 X. Y. Liu et al. / International Journal of Steel Structures, 17(1), 273-289, 2017
Figure 18. Comparisons of damage curves between damage models and test results.
(1) From Table 11 it can be observed that some β model is not used for SP-2B, SP-5B and SP-6B. The
values of combination model are negative, which indicates damage curves calculated by the three damage models are
the effect of the cycle with maximum displacement is illustrated in Fig. 17 and some conclusions can be made
overestimated in that model. as follows:
(2) The dispersion of the three damage models is The combination model is a linear accumulated model;
different when simulating the damage course of welded the linear fitting gives the same weighting for each cycle,
connections. From the coefficient of variation (COV), it except for the cycle with maximum displacement. While
can be perceived all the three models has large scatter, the displacement model and energy model are accelerated
moreover, the dispersion of combination model is larger damage models, the parameter c has large effect on the
than the other two models. connection damage process.
(3) As the steel material can be assumed to be rigid- All the displacement model and energy model curves
perfectly plastic, the Eq. (4) can be simplified as Eq. (7): have obvious threshold. In this experiment, all the critical
n
values are near the yield point. The damage curves show
δ pi⎞ c
∑ ⎛⎝ -----
that the obvious damage occurred after the specimen
D= - (7)
δu ⎠ yielded, which is in accordance with the models physic
i=1
meaning. On the other hand, the terraced curve of
where δ pi denotes the plastic deformation of each half combination model shows a different damage process
cycle, δ u is the ultimate deformation under monotonic with the other two, the critical points are not apparent.
loading. That curves show the damage at the elastic cycle and
Comparing Eq. (7) with Eq. (4), it shows great agreement cannot reflect the real situation at the initial stage.
between the energy model and displacement model. The The effects of connection geometry on damage process
fitting parameters of these two models are similar to each are not significant for displacement model and energy
other as listed in Table 11, which also indicates that the model. Although specimen SP-1, SP-5, SP-6, SP-7 and
two models can be converted to each other approximately. SP-8 has different geometries, the two damage model
To avoid the conflict between the definition of parameter curves all agree well with each other.
β and the negative values of the specimen, the combination The scatter of the fitting results could be resulted from
288 X. Y. Liu et al. / International Journal of Steel Structures, 17(1), 273-289, 2017
the following reasons: The connection made of high strength steel probably
The welding defects and residual stress existed in the has a lower ultimate loading capacity, which may be
welded connection causes the connection has different correlated with the welding quality of high strength steel.
welding quality. The connection with larger beam flange width-thickness
The welded connection may be eccentric loaded ratio has lower ultimate strength and deformation. Cyclic
differently during loading procedure, which leads to the loading results in crack initiate and propagate in the
different lateral buckling for the different specimen. The connection weld, which weakens the bearing capacity and
lateral buckling makes the connection damage unilaterally, deformation capacity of the welded connections. The
while the extensometers are installed at two sides of range of cyclic loads is the key factor affecting the weld
specimen, they cannot measure the displacement distribution damage level.
at different locations of the specimen. So the mean values Weld cracking and plate buckling are the major causes
are adopted to evaluate the connection performance, of connection damage. The load capacity decreases greatly
which leads to the difference between test results and real with the crack propagating and plate buckling. For the
values. welded connection designed to have higher load capacity
The fitting curves obtained by the damage models are than components, the damage of its bearing capacity and
compared with the test curves of specimen SP-1D, SP-5B stiffness is mainly caused by the crack evolution in weld.
and SP-7B as shown in Fig. 18. A variety of damage evolution equations applied in
As shown in Fig. 18, the displacement model and engineering are calibrated and compared, the comparison
energy model curves are approximate to the test curves results indicate that the damage model adopting the
when accumulated plastic deformation is large. While the connection plastic deformation index shows a reasonable
curve shape of the combination model agrees well with good accuracy to fit the test data. Moreover, that model
the test curve, especially for the nonuniform damage has less relation with the connection geometries so that it
acceleration process, the variation trend of the curve can be widely applied, which means adopting the plastic
slope is more similar to test curve. Accordingly, these deformation to illustrate damage process of metal material
three models have their own merits when simulating conforms to its physical mechanism. The damage models
welded connections damage, the result of displacement based on the macro mechanics index can reveal the
model is more accurate. connection deterioration behavior to a certain extent.
structure: theory and practice”, Beijing: Science Press. prediction by deterministic methods: Concept and
John W. F., Robert J. D., and Eric J. K. (1995). “Fracture procedures”, Earthquake Engineering and Structural
Mechanics of Welded Structural Steel Connections”, Dynamics, Vol. 16, pp. 719-734.
SAC report No. 95-09, Washington, DC. SAC. (1995). “Interim Guidelines: Evaluation, Repair,
Kato B. and Akiyama H. (1975). “Aseismic Limit Design of Modification and Design of Welded Steel Moment Frame
Steel Rigid Frames”, Proceeding of Architectural Institute Structures”, SAC-95-02, Sacramento, California.
of Japan, No.237, Nov., 1975. Shen Z.Y. and Shen S. (2002). “Seismic Analysis of Tall
Kaufmann E.J. and Fisher J.W. (1995). “A study of the Steel Structures with Damage Cumulation and Fracture
effects of material and welding factors on moment frame Effects”, Journal of Tongji University (Natural Science),
weld joint performance using a small scale tension Vol. 30, No. 4, pp. 393-398.
specimen”, Technical report SAC 95-08. Tawil S., Mikesell T., Vidarsson E. and Kunnath S.K.
Kravinkler H. and Zhorei M. (1983). “Cumulative Damge in (1998). “Strength and Ductility of FR Welded-Bolted
Steel Structures Subjected to Earthquake Ground Motions”, Connections”, SAC report No. 98-01, Washington, DC.
Computers and Structures, Vol. 16, pp. 1-4. Shi W.L, Xiao Y., Li G.Q. and Ye Z.M. (2008). “Pseudo-
Krawinkler H. and Nassar A. A. (1992). “Seismic Design dynamic tests on composite joints with flush end plate
based on Ductility and Cumulative Damage Demands connections”, Earthquake Engineering and Engineering
and Capacities”, Nonlinear Seismic Analysis and Design Vibration, Vol. 28, No. 6, pp. 124-133.
of Reinforced Concrete Buildings, Elsevier. Shi Y.J., Ao X.L., Wang Y.Q. and Shi G. (2009).
Kumar S. and Usami T. (1994). “A note on evaluation of “Experimental study on the seismic performance of
damage in steel structures under cyclic loading.” Journal beam-to-column composite connections in medium-high
of Structure Engineering”, JSCE, Vol. 40A, pp. 177-178. strength steel frame structures”, China Civil Engineering
Liu Y.M., Chen Y.Y. and Chen Y.J. (2002). “Hysterical Journal, Vol. 42, No. 4, pp. 48-54.
Model of Steel Frame Connection Considering Partial Su D. (2005). “Seismic Behavior of Beam-Column Connections
Fracture”, Journal of Tongji University (Natural Science), in Steel Structure with Composite Effect”, Tsinghua
Vol. 31, No. 5, pp. 525-539. University.
Matos C.G. and Dodds R.H. (2001). “Probabilistic modeling Toyoda M. (2002). “Properties of steel structures damaged in
of weld fracture in steel frame connections part I: quasi- Hanshin-Japan and Northridge-USA earthquake”, Seminar
static loading”, Engineering Structures, Vol. 23, pp. 1011- notes. Dokuz Eylul University, Izmir(Turkey).
1030. Yang T.S. and Popov E.P. (1995). “Experimental and
Matos C.G. and Dodds R.H. (2002). “Probabilistic modeling Analytical Studies of Steel Connections and Energy
of weld fracture in steel frame connections part II: Dissipators”, SAC report No. 95-13, Washington, DC.
seismic loading”, Engineering Structures, Vol. 24, pp. Zhang L. (2004). “Study on earthquake resistance of rigid
687-705. beam-to-column connection of steel structure”, Tianjin
Park A. J. and Ang H. S. (1985). “Mechanistic Seismic University.
Damage Model for Reinforced Concrete”, Journal of Zhao D.W., Shi Y.J. and Chen H. (2000). “Experimental
Structure Engineering, Vol.111, No.4, pp. 195-207. Research on Beam-column Connections Under Cyclic
Powell G. H. and Allchabadi R. (1988). “Seismic damage Loading”, Building Structure, Vol. 30, No. 9, pp. 3-6.