Anda di halaman 1dari 33

1

Cdric Lequesne
1a
, Andr Plumier
1b
, Laurent Duchne
1c
, Anne Marie Habraken
1d

Modeling of Crack Propagation in Weld Beam-to-Column Connections
Submitted to Cyclic Loading with a Cohesive Zone Model

ABSTRACT: During the earthquakes in Japan and California in the 1990s, cracks appeared in
some weld beam-to-column connections of heavy rigid frame steel buildings. This prompted the
necessary assessment of the performance of weld connections in terms of rotation capacity and
crack propagation. In the present study, experimental tests were performed where weld
connections were submitted to cyclic loadings with increasing amplitude until a macro crack
event was reached. However the crack phenomenon depends on many parameters: the geometry,
the material, the welding process. For this reason, it was interesting to develop a finite element
modeling of these connections in order to complete these experiments and perform a parametric
study. This paper describes the finite element model development, its material parameter
identification and its comparison with experimental results. The weld connections were modeled
by using three-dimensional mixed solid elements. The constitutive laws applied were
elastoplastic with isotropic hardening identified for the base metal and the weld metal. Crack
propagation was modeled by a cohesive zone model. The parameters of this cohesive zone model
were identified by an inverse method with the modeling of three point bending tests of pre-
cracked samples performed on the base and weld metals. The fatigue damage generated by the

1
ArGEnCo Department, University of Lige, Chemins des Chevreuils, 1, B-4000 Lige, Belgium.
a
Cedric.Lequesne@ulg.ac.be; tel.: +32(0)4-366-91-40; fax: +32(0)4-366-91-92
b
A.Plumier@ulg.ac.be; tel.: +32(0)4-366-91-30; fax: +32(0)4-366-91-92
c
L.Duchene@ulg.ac.be; tel.: +32(0)4-366-93-32; fax: +32(0)4-366-91-92
d
Anne.Habraken@ulg.ac.be; tel.: +32(0)4-366-94-30; fax: +32(0)4-366-91-92
2

cyclic loading was computed by the fatigue continuum damage model of Lemaitre and
Chaboche, which was coupled with the cohesive zone model.
KEYWORDS: Cohesive zone, Fatigue, Finite Element, Beam-to-Column Connection, Welding

3

Introduction
In seismic zones, steel moment-resisting frame buildings contain heavy welded beam-to-column
connections. By means of energy dissipation due to plastic strain, the ductility of steel makes it
possible to avoid the appearance of cracks in the connections. However, earthquakes in the 1990s
in the USA and Japan resulted in widespread and unpredicted damage in welded beam-to-
column connections in rigid steel frame buildings. These failures explain why the engineering
community decided to investigate the reasons for this unexpected behavior and to explore
alternative connection types. Research in many countries resulted in a number of changes to
building design codes and specifications. However, performance is affected by many factors
such as dimensions of beam-to-column components, connection design, manufacturing quality
and the mechanical properties of the different regions of the joint. A procedure for analyzing
these factors, called the Risk Assessment Procedure (RAP), was recently published under the
auspices of the International Institute of Welding (IIW) [1]. This procedure determines the risk
of fracture in seismically affected moment connections. It covers design, material, fabrication
and loading issues.
European steelmakers produce heavy sections used in multi-storey buildings in seismic zones. In
order to maintain the competitiveness of the European Union in this market, methods need to be
available for steel users to verify the specification of steel sections, to define connections and to
assess safety in service.
The VERAPS project (Validation and Enhancement of Risk Assessment Procedure for Seismic
connections) [2] aimed to validate and enhance the RAP for connections in seismic areas.
Fabrication, testing, modeling methods and reliability analyses were combined to achieve this
aim. A series of full-scale tests on connections with different parent material strength and weld
4

metal toughness was carried out by the University of Karlsruhe. The variables covered included
moment capacity ratios, welding methods, weld metal toughness and joint design details.
However these tests were expensive and required significant installation.
An objective of the VERAPS project was to model the crack propagation in heavy welded beam-
to-column connections submitted to cyclic quasi-static loading by finite elements. The objective
of the simulation is the assessment of the cyclic plastic rotation capacity of heavy welded beam-
to-column connections as a function of the mechanical properties of the beams, columns and
weld materials, and of the type of joint preparation. This modeling approach was compared with
tests performed by the University of Karlsruhe and a validated model made it possible to predict
the behavior of non-tested connections and to explore a larger field of possibilities. This paper
focuses on the description and the comparison on these finite element models.
The finite element code used was the Lagrangian non-linear finite element code Lagamine,
developed at the University of Lige since the 1980s. The damage in the connections was due to
a high level of deformation and to cyclic loading. The monotonic damage and the crack process
were modeled by a cohesive zone model, which was used to simulate initiation and propagation.
This model required few parameters and was easy to implement. The classical fatigue damage
due to cyclic loading was evaluated by the fatigue continuum damage model of Lemaitre and
Chaboche [3]. The computation was performed in two steps. A first simulation evaluated the
fatigue damage without the cohesive zone model. The crack path was identified with the zone
where the damage was significant. Then a second simulation was performed with cohesive zone
elements to model the crack propagation. After the Karlsruhes tests, some samples were
machined to perform some material tests, such as tensile tests, Charpy tests, and three-point
bending tests. These tests allowed the authors to identify the material parameters of their model.
5

This paper firstly presents the weld beam-to-column connections and the test characteristics.
Secondly, it explains the model of crack propagation. Thirdly, it sets out the modeling of the
connection tests. Finally the results from the numerical model and the experiments are compared.
Connection Tests
Eight large-scale tests were performed with a 6300kN alternating load test frame at the
University of Karlsruhe [2]. All beams, columns and plates were 355 J0 and J2 grade steel.
Varying their dimensions made it possible to test different ratios between joint plastic moment
and beam plastic moment.
In all specimens, the beam flanges were welded to the column flanges with butt welds. The
flange butt welds transmitted the full plastic moment of the section area of the beam flange. The
beam web was bolted to a shear tab, which was welded to the column flange. For some
specimens, the shear tab was welded and bolted to the beam web. The shear tab was welded to
the column flange for all specimens (see FIG. 1). The shear tab transmitted the plastic moment of
the beam web and the corresponding shear.
Regarding the welding, the VERAPS connections were manufactured using a range of
consumables to cover the full toughness range, together with three different types of process:
manual metal arc welding, tubular cored metal arc welding with active gas shield or metal arc
welding without gas shield. Due to the significant thickness of the flange, the number of passes
was ranging from ten and thirty. The dimensions of the weld access holes complied with FEMA
[4] requirements. Some characteristic flaws, such as a lack of fusion, were artificially added
during the welding process. During the welding process of the beam flange to the column flange,
6

a backing bar in steel or ceramic was installed below the butt weld to support the welding at the
beginning. This was removed at the end of the welding process.
Two types of stiffener were also manufactured for the column panel (see FIG. 1):
30-mm-thick transverse stiffeners acting as continuity plates for the beam flanges but
bringing no increase in the shear resistance of the panel zone.
40-mm-thick extended doubler plates placed parallel to the column web. These acted
as stiffeners and as continuity plates and significantly increased the shear resistance
of the panel zone.
The frame was oriented for a classic compression-tension test, which resulted in the specimen
geometry presented in FIG. 2. Each test sample was subjected to cyclic loading with step-wise
deformation cycles increasing up to failure according to the loading procedure of FEMA 350.
The protocol imposes the number and the amplitude of cyclic rotations defined in TABLE 1 until
macrocracks are observed. If no macrocrack appears after the total number of cycles, the rotation
amplitude is incremented by 0.01 radians, with two cycles of loading at each rotation level.
During the test, the displacement was imposed and the force of the machine was measured. For
this purpose, some actuators and strain gauges were installed on the connection. The rotation is
obtained from the imposed displacement and the beam length. The beam end moment comes
from the measured force and the beam length. The results of the tests are available in [2] and [5].
For this paper, only one connection test was modeled. The time of the crack observation are
described in TABLE 2. The operator observed the connection at different cycles of the loading in
order to check if a macro crack event was occurred. An abrupt failure appeared at the first cycle
of the step n5 at the beam bottom flange (see FIG. 3). Perhaps some micro crack initiated
through the thickness during the step 4 which were not observed.
7

TABLE 1- Cyclic rotation imposed on the connection according to FEMA 350
Step Number of cycles
Rotation
amplitude
(rad)
1 6 0.00375
2 6 0.00500
3 6 0.00750
4 4 0.010
5 2 0.015
6 2 0.020
7 2 0.03

FIG. 1 - Details of the connections from [2]


8


FIG. 2 - Test specimen (from [2])
9

TABLE 2 - Loading record of the experimental test reference
Step Cycle
Imposed Displacement
(mm)
Crack control is indicated with grey
1
1
17.6

2
3
4
5
6 No crack detected
2
1
24

2
3
4
5
6 No crack detected
3
1
35
No crack detected
2
3
4
5
6 No crack detected
4
1
47
No crack detected
2 No crack detected
3 No crack detected
4 No crack detected
5 1 70.6 Abrupt failure of the weld of the beam bottom flange

FIG. 3 - Experimental crack observations
10


Crack Propagation Modeling
The Fatigue Continuum Damage Model (CDM)
In multiaxial state, the fatigue stress limit is defined by a fatigue surface. The criterion of fatigue
limit depends on the amplitude and the mean values per cycle. As observed experimentally, the
mean shear does not affect the fatigue limit, as opposed to the mean tension [3]. Therefore it was
defined:
the second invariant of the amplitude of the deviator of the stress tensor, :

( )( )
II ijmax ijmin ijmax ijmin
1 3
A = - -
2 2
(1)
where
ij ij kk
k
1
= -
3


,
ijmax
and
ijmin
are the maximum and minimum components of the
deviatoric stress tensor during one cycle.
the mean hydrostatic stress,
Hm
:

kk
Hm
k
T
1
= dt
T 3

(2)
The multiaxial fatigue criterion is defined by a damage yield locus, noted f
D,
as in plasticity:

*
D II II
f =A -A (3)
where A
II
*
represents the fatigue limit.
Therefore if the loading, A
II
, is smaller than the threshold value, the structure does not undergo
any damage. Although different criteria exist, the Sines criterion [6] was chosen because of its
11

simplicity. No test was available to evaluate the performance of other criteria. The fatigue limit
of the Sines criterion is:

*
II l0 Hm
A = (1-3b ) (4)
where
l0
is the fatigue stress limit.
For multiaxial loading, Lematre and Chaboche [3] stated that the evolution law for the fatigue
damage variable, D
f
, during the N
th
cycle, , is as follows:


( )
( )
D

+1 D
II
f
*
II II II
II 0 Hm
f u eqmax
0 if f <0
D
= if f 0
A
N 1- 1-D
M
A A -A
A ; M=M 1-3b ; =1-a
1 D -

| |
(

|

\
=

(5)
This damage evolution law depends on the maximum von Mises stress per cycle,
eqmax
.
u
is
the ultimate tensile stress of the material. M
0
, b, a and are other material parameters. The
symbol x corresponds to the following definition: if x is negative, then its value is null and if x
is positive, then its value is x. The fatigue continuum damage model makes it possible to
compute the non-linear evolution of the damage with different levels of loading amplitude.
The parameters of this model were identified from Wohlers curves, which were determined by
cyclic tensile tests for the base and weld metal samples extracted from different locations of the
connection (see TABLE 3). The procedure is described in [7].
12

TABLE 3 - Parameters of the identified constitutive law

Fatigue continuum damage parameters Cohesive zone parameters

l0

(MPa)

u
(MPa)
a b
M
0
(MPa)

max

(MPa)

n0

(mm)

nc

(mm)
Base metal 275 585 7.38 0.15 0.00171 1069 1700 1.70x10
-4
1.89x10
-1

Weld metal 228 650 4.08 0.15 0.00155 2255 1650 1.65x10
-4
6.06x10
-2

The Cohesive Zone Model (CZM)
In the early 1960s, Dugdale [8] and Barenblatt [9] introduced the concept of the cohesive zone
model. This model has been used for monolithic and composite materials. During crack
propagation, there is a fracture process occurring behind the crack tip, where microcracks and
microvoids nucleate, grow and then coalesce. Thus the behavior of this zone is different from the
sound bulk material, due to its progressive degradation. The cohesive zone model describes this
behavior. The potential crack is modeled by two interface areas connected by cohesive stresses
(see FIG. 4). The degradation process is described by the constitutive law linking the cohesive
stress, T, and the separation, u. According to a literature review, there are different forms of
this law, but they all have common features. Cohesive stress, during the increase in separation,
begins to increase until it reaches a maximum stress value
max
; it then decreases and vanishes
after full rupture.
13


FIG. 4 - Cohesive zone model description
In mixed mode fracture, stress and separation components are defined by (u
n
, u
p,
u
s
) and (
n
,

p
,

s
), where n, p and s correspond to crack modes I, II and III respectively. Mode I is a normal
opening of the interface, while modes II and III are shear slipping of the crack interfaces.
The Crisfields model [10] was chosen here. The initial stiffness can be regulated so that the
global stiffness of the structure does not interfere with the presence of the cohesive zone in the
model [11]. Crisfields model has a bilinear shape and is used to define the initial stiffness E
co
:

max
n0
max
co
t 0
max
t 0
0 0
E 0 0
0 0
(
(

(
(
=
(

(
(
(


(6)
where
max
and
max
are the maximum normal and shear stresses, and
n0
and
t0
are the normal
and shear separations when the normal and shear stresses reach their maximum values,
max
and

max
.
14


FIG. 5 - Crisfields cohesive zone model a) for mode I and b) for modes II or III
A cohesive damage tensor, D
c
, is defined as a function of the separation, u. At the beginning of
the loading phase, the damage is equal to zero. If one separation component, u
i
, is beyond the
separation,
i0
, where the stress is equal to the maximum stress, then the damage begins to grow
until the separation is equal to a critical displacement,
ic
, when the damage is equal to 1. In
unloading, the damage stops growing. Then, in reloading, the damage grows again when the
separation reaches the value when the unloading began.
For the mixed fracture mode, the cohesive damage tensor is defined by:

1
p
s n
n0 p0 s0
0
t
0
nc
c
0
nc n0
tc
tc t 0
tc
tc t 0
u
u u
-1
max( )
D F with
0 0
1
F 0 0
0 0

(
| | | | | |
(
= + + |
| |
|
(
\ \ \

(
=

(
+

(
=
(

(


(7)
15

where
tc
and
nc
are the separations when the normal and the shear stresses are null, is a
material parameter and t the time.
The condition when the monotone damage rate is positive is:

c
0
0
D >0
if (- >0), then

(8)
Otherwise,
c
D

is zero, since no decrease in damage is possible.


Thus, the constitutive law of the cohesive zone model is (see FIG. 5):

c co
T I D E u = (

(9)
where I the unity matrix.
In compression (u
n
<0), the damage tensor becomes null to model penalty contact. As a result,
the cohesive stiffness matrix is:
( )
c co c
1 1 n0
n n p n s
1 1
p n p p s 1
c co co c
t 0 1 1
s n s p s
t 0
I D E if D 0
1
0 0
u u u u u
C
1 1
I D E - FE u u u u u 0 0 if D 0
1
u u u u u
1
0 0

= (

(
(

( (
=
( (
> (
( (

+
( (


(
(


(10)

During the project, the projects partner Corus carried out three-point bending tests on fatigue
precracked base metal samples extracted from the connection in the base location. They plotted
the crack growth by function of the J-integral. For the weld metal, the Charpy energy was
measured and the J-integral/crack growth curve was plotted by power-law description. This
16

curve was correlated to the Charpy V-notch energy [12]. The parameters of the cohesive zone
model were identified by inverse method in modeling the three-point bending tests (see TABLE
3). Different parameters of the cohesive zone model were tested until they reached the equivalent
of the J-integral/crack growth curve [13].
Coupling between the Cohesive Zone and Fatigue Damage
Fatigue damage can be taken into account using the same method as in [11]. By adding the
fatigue damage parameter, D
f
, computed by Equation (5) to the cohesive damage tensor, D
c
, (see
FIG. 6).

f
co c
T I D E u where D D D I = = + (

(11)
However, in this way the fatigue damage reduces the separation when the material is totally
debonding (see FIG. 6a). Therefore, the stiffness of the cohesive law, which is negative when the
cohesive stress decreases, strongly decreases, and this can affect the convergence of the
computation. Roe and Sigmund [14] suggested taking the fatigue damage into account by
affecting the stiffness by the following relationship (see FIG. 6b).


f
c co co co
T I D E u where E =(1-D )E = (


(12)
17


FIG. 6 - Coupling between the cohesive zone model and fatigue damage suggested by a) [11] and b) [14]
Finite Element Modeling
A three-dimensional cohesive element was developed in the Lagamine code. In this element, the
interface side of each crack is modeled by 4 nodes (see FIG. 7). They are called, respectively,
CZM3D,
Foundation segment.
CZM3D and the foundation segment are 4-node plane elements. As a result, they can be
computed with 1, 4 or 9 integration points.
18


FIG. 7 - Description of CZM3D
The fatigue damage was computed according to the fatigue Continuum Damage Model of
Lematre and Chaboche [3] from the stress tensor components in the solid elements. The method
of transferring the fatigue damage from the solid elements to the cohesive elements is illustrated
in FIG. 8. The fatigue damage transferred to the integration points of the cohesive element was
the mean of the occurrences of fatigue damage computed in the integration points of the solid
elements linked to the cohesive element and to the foundation segment. In the studied cases, the
solid elements contained only one integration point for two-dimensional and three-dimensional
simulations.

19


FIG. 8 - Transfer of fatigue damage from solid elements to cohesive elements
Connection Modeling
A special mesh generator was developed for this test modeling. The input of this module was the
beam and the column dimensions and the type of reinforcement. In this study, the meshes
generated were composed of approximately 13000 nodes and 8500 elements. The elements are
mechanical solid 8-node BWD3D [15] of mixed type available in the Lagamine code. FIG. 9 and
FIG. 10 present the meshes of a test with transverse stiffeners and a test with an improved
doubler.

FIG. 9 - Mesh of two beam-to-column connections tested by the University of Karlsruhe
20


FIG. 10 - Mesh of the connection around the welding
A plate was added to the end of the beam in order to stiffen it and to avoid yielding due to the
imposed displacement. The beam support was equivalent to a rolling bearing, whereas the
column support was equivalent to a hinge. Displacements were imposed on node lines at the
centre of the web at the end of the beam. Therefore, the rotation was free and no physical
plasticity was allowed at this point. The boundary conditions are described in FIG. 11.
21


FIG. 11 - Boundary condition of the connection modeling
The constitutive laws for each material are elasto plastic with isotropic hardening. These were
calibrated from tensile tests on specimens extracted at different locations of the connection. As a
result, the constitutive laws used were different for the flanges, the web and the welding (see
FIG. 12). The strain-stress curves are described in FIG. 13.

FIG. 12 - Position of the different sets of hardening coefficient
22


FIG. 13 - Strain-stress curves for the different materials
The mesh did not model the bolts of the shear tab. Instead, the connection between the shear tab
and the beam web was complete because the nodes were merged at the interface between the two
components. The weld metal material of the shear tab-to-column flange and the beam flange-to-
column flange connections were modeled.
This paper presents the results for the connection shown in FIG 1 a. Three simulations were
performed and their characteristics are summarized in TABLE 4. A first computation,
(simulation n1) was performed without the cohesive zone model but with the fatigue damage
model. The latter being a decoupled approach did not affect the mechanical behavior of the
structure. It was just used to find the damage evolution and localization defined by the fatigue
damage, D
f
. The first aim of this simulation was to compare the beam end moment versus
rotation curve with the experimental measurement. The second was to identify the potential
crack path to add cohesive zone elements in the simulation.
23

As the crack location was identified by the first simulation, the second simulation contained
cohesive elements coupled with the fatigue damage. The aim of the simulation was to model the
propagation of the crack at the connection and to observe its impact on the moment rotation
curve. This modeling was compared with the experimental results. Finally, a third simulation
was performed with cohesive elements where the fatigue was neglected. The aim was to quantify
its effect on the crack initiation.
TABLE 4 - Simulation characteristics
n Model Goal
1 Connection + CDM
Identify location of fatigue damage
defining where to put CZM
elements
2 Connection + CZM +CDM
Model crack propagation with
fatigue effect
3 Connection +CZM
Model the crack propagation if
fatigue damage is neglected

Results and Discussion
Identification of the Crack Path (simulation n1)
FIG. 14 shows the comparison of the beam end moment versus rotation curves between the finite
element simulation n1 results and the experimental measurement for the first three steps. The
elastic stiffness of the connection curve was equivalent for the two results curves (see steps 1 and
2). For positive rotation angles which correspond to a tensile state in the studied specimen, the
numerical and experimental values are similar. A small difference is observed for negative
rotation angles. The experimental curves show a slight shift towards negative rotation angles due
to the sliding of the connection around the bolts, which was not modeled. Moreover the
numerical curves present a cyclic hardening in contrast to the experimental ones for step 3.
24

Regarding the beam flange end (see FIG. 15), the damage variable was different from zero and
was concentrated at the interface between the weld metal and the column flange. In fact, damage
grew when the moment was at its highest, as the longitudinal stress. In addition, an experimental
macrocrack observed at this location validated the model prediction. In the weld flange, the
damage increased in the beam flange near the interface between the base metal and the weld
metal (see FIG. 15). FIG. 3 presents the location of the macro crack at the end of the experiments
in accordance with the numerical simulation. The cracks appeared at the beam bottom flange
weld.

25


FIG. 14 - Beam end moment versus rotation curves for FE and experimental results

FIG. 15 - Evolution of the fatigue damage near the lower weld flange



26

Simulation with the Cohesive Zone Model (n2 and n3)
The simulation n2 (see TABLE 4) took into account both the fatigue damage model and the
cohesive zone model. The cohesive zone elements were defined at the interface between the
weld metal and the base metal in the beam flange end.
During the computation, a crack initiated at the root of the welding on the beam bottom flange
end and propagated along the column flange. FIG. 16 gives the crack propagation: it happened
at cycle 2 of step 3 and quickly propagated during the first part of the loading cycle. The crack is
the zone where the longitudinal stresses are released. FIG. 17 a) is a contour plot of the fatigue
damage and FIG. 17 b) gives the longitudinal stress on a cross section at the mid width close to
the beam bottom flange. During this second simulation, the fatigue damage is significant at the
crack initiation location but it did not develop during the short propagation time.
FIG. 18 shows the evolution of the beam end moment versus the rotation for the finite element
computation and the experimental test. The curves of the first two steps are similar to the curves
of the previous analysis. However, the cyclic hardening of the finite element simulation for step
3 is less significant than for the simulation without the cohesive element so that the numerical
results are closer to the experimental measurements.

27


FIG. 16 Cohesive elements (black if cracked) during the second cycle of the step 3

FIG. 17 - Fatigue damage a) and longitudinal stress b) at the mid width on the beam bottom flange in the middle of step 3
28


FIG. 18 - Beam end moment versus rotation curves for FE with CZM and experimental results
The experimental crack event was observed at cycle 1 of step 5 (see TABLE 2). Its exact
propagation was not recorded. The operator just reported an abrupt failure. This abrupt event can
follow a less visible crack not detected by the operator eye. However it seems clear that the
model is too conservative and predicts a crack event earlier than in the experiment.
The simulation n3 where only cohesive zone elements were present without coupling with the
fatigue damage modeling did not predict any crack. It proves the significant effect of the
decrease of the maximum cohesive stress due to fatigue damage (see FIG. 6).
Conclusions
The objective of this study was the modeling of welded steel beam-to-column connection
cracking submitted to cyclic loading by the finite element method. The cohesive zone model is a
practical model due to its ease of implementation and its small number of parameters. It makes it
possible to model crack initiation and propagation. The fatigue damage was calculated by the
29

classical Lemaitre and Chaboches model [3]. The addition of these two models improved the
results by enabling the numerical analysis to get close to the experiments. It was observed in the
simulation n2 that the crack initiation was driven by the fatigue damage, D
f
, while the
propagation was driven by the cohesive damage, tensor D
c
, generated by the high level of
deformations.
Moreover the drawback of the cohesive zone model is that the crack path must be known during
the meshing. An initial idea was to perform a first analysis without cohesive elements because
the fatigue damage field gave an idea of the crack location. However as the fatigue damage
drove only the crack initiation, this model gave only the root of the crack, without validating the
crack path. Future studies are needed in order to perform a finite element analysis with a
remeshing step where some cohesive elements are added as a function of a crack bifurcation
criterion.
Finally it should be noted that, due to welding, the connection contains some residual stresses,
which can affect the crack propagation. In the present study, these residual stresses were not
taken into account. Another study has evaluated such residual stresses, and the results are
presented in [7]. It would be interesting to implement the residual stresses in the present
simulation in order to observe their impact on predicted results.
Acknowledgments
This study was carried out thanks to the supply of the European Community (VERAPS project
n RFS-CR-03035) and the assistance of the project partners (Corus, ISQ, The University of
Karlsruhe). The authors of this article would like to thank the Belgian Federal Science Policy
30

Office (Contracts IAP P6-24) for its financial support. A.M. Habraken and L. Duchene
acknowledge the Fund for Scientific Research (FRS-FNRS, Belgium).
31

References
[1] International Institute of Welding, IIW recommendations for assessment of risk of
fracture in seismically affected moment connections, IIW-X-1504-02, IIW-XV-1102-
02, IIW-XV-SCG-103-02, 2002.
[2] Bannister A., Report EUR, Publication Office, Publications.Europa.eu, VERAPS:
Validation and Enhancement of Risk Assessment Procedure for Seismic connection,
RFS-CT-2003-00035 project, 2007.
[3] Lemaitre J. and Chaboche J.L., Mcanique des matriaux solides, Dunod, 1996.
[4] FEMA, Recommended seismic design criteria for new steel moment-frame buildings,
FEMA 350, 2000.
[5] Plumier A., Lequesne C., Degee H., Bannister A. and Hoebling W., Behaviour of heavy
sections welded moment connections, STESSA09, Philadelphia, United States, 2009.
[6] Sines G., Behavior of metals under complex static and alternating stresses G. Sines and
J.L. Waisman, Metal Fatigue, McGraw Hill, New York, 1959.
[7] Lequesne C., Modeling of fracture in heavy steel welded beam-to-column connection
submitted to cyclic loading by finite elements, PhD thesis, University of Liege, 2009,
available from http://bictel.ulg.ac.be/ETD-db/collection/available/ULgetd-07062009-
125347/.
[8] Dugdale D.S., Yielding of steel sheets containing slits, Journal of the Mechanics and
Physics of Solids, 8, pp. 100108, 1960.
[9] Barenblatt G.I., Mathematical theory of equilibrium cracks in brittle fracture, Advances
in Applied Mechanics, VII, 55-129, 1962.
[10] Mi Y., Crisfield M.A., Davies G.A.O. and Hellweg H.-B., Progressive delamination
using interface elements, Journal of composite materials, 32 (14), 1246-1272, 1998.
[11] Bouvard J.L., Modlisation de la propagation de fissure dans les aubes de turbines
monocristalline, PhD thesis, Ecole Nationale Suprieur des Mines, 2006, available
from.
[12] Wallin K., Low-cost J-R curve estimation based on CVN upper shelf energy, Fatigue &
Fracture of Engineering Materials & Structures, 24 (8), 537-549, 2001.
[13] Lequesne C., Plumier A., Degee H. and Habraken A.M., Numerical study of the fatigue
crack in welded beam-to-column connection using cohesive zone model, M.H.
Aliabadi, et al., Fracture and Damage Mechanics V, Trans Tech Publications, Key
Engineering Materials, Harbin, China, 2006.
32

[14] Roe K.L. and Siegmund T., An irreversible cohesive zone model for interface fatigue
crack growth simulation, Engineering Fracture Mechanics, 70 (2), 209-232, 2003.
[15] Li X.K. and Cescotto S., A mixed element method in gradient plasticity for pressure
dependent materials and modelling of strain localization, Computer Methods in
Applied Mechanics and Engineering, 144 (3-4), 287-305, 1997.

Figure Captions
FIG. 1 - Details of the connections from [2] ................................................................................... 7
FIG. 2 - Test specimen (from [2]) ................................................................................................... 8
FIG. 3 - Experimental crack observations ...................................................................................... 9
FIG. 4 - Cohesive zone model description .................................................................................... 13
FIG. 5 - Crisfields cohesive zone model a) for mode I and b) for modes II or III ...................... 14
FIG. 6 - Coupling between the cohesive zone model and fatigue damage suggested by a) [11]
and b) [14] ..................................................................................................................................... 17
FIG. 7 - Description of CZM3D ................................................................................................... 18
FIG. 8 - Transfer of fatigue damage from solid elements to cohesive elements .......................... 19
FIG. 9 - Mesh of two beam-to-column connections tested by the University of Karlsruhe ......... 19
FIG. 10 - Mesh of the connection around the welding ................................................................. 20
FIG. 11 - Boundary condition of the connection modeling .......................................................... 21
FIG. 12 - Position of the different sets of hardening coefficient .................................................. 21
FIG. 13 - Strain-stress curves for the different materials.............................................................. 22
FIG. 14 - Beam end moment versus rotation curves for FE and experimental results ................. 25
FIG. 15 - Evolution of the fatigue damage near the lower weld flange ........................................ 25
FIG. 16 Cohesive elements (black if cracked) during the second cycle of the step 3 ............... 27
33

FIG. 17 - Fatigue damage a) and longitudinal stress b) at the mid width on the beam bottom
flange in the middle of step 3 ........................................................................................................ 27
FIG. 18 - Beam end moment versus rotation curves for FE with CZM and experimental results 28

Anda mungkin juga menyukai