Engineering Fracture Mechanics 76 (2009) 1531–1547

Contents lists available at ScienceDirect

Engineering Fracture Mechanics

journal homepage: www.elsevier.com/locate/engfracmech

Numerical and experimental analyses of damage behaviour of steel

moment connection
W. Bleck a, W. Dahl a, A. Nonn b,*, L. Amlung a, M. Feldmann c, D. Schäfer c, B. Eichler c
a
RWTH Aachen University, Department of Ferrous Metallurgy, Intze-Straße 1, 52072 Aachen, Germany
b
Salzgitter Mannesmann Forschung GmbH, Mechanical Engineering, Ehinger-Straße 200, D-47259 Duisburg, Germany
c
RWTH Aachen University, Institute of Steel Construction, Mies-van-den-Rohe-Straße 1, 52074 Aachen, Germany

a r t i c l e i n f o a b s t r a c t

Article history: Plastic design allows the exploitation of the full resistance of steel structures by taking
Received 5 June 2008 advantage of stress–redistributions due to plastic strains exceeding the yield strain. Espe-
Received in revised form 12 January 2009 cially in seismic design the utilization of material reserves and the formation of plastic
Accepted 12 March 2009
hinges play an important role. In devastating earthquakes in Northridge (USA) and Kobe
Available online 21 March 2009
(Japan) brittle fracture of welded connections in steel moment frames occurred prior to for-
mation of plastic hinges and utilization of plastic material reserves. The subsequent
Keywords:
research works resulted in improved design rules and recommendations for these kinds
Beam column connections
Seismic loading
of failure. But to guarantee sufficient ductile performance of these connections also in
LPD model the upper shelf region, plastic and earthquake resistant design rules should take into
account degradation of strain capacity and toughness properties due to quasi static and
especially seismic loading.
In the scope of the current European project ‘‘Plastotough”, the main objective is to
derive quantified toughness design rules in the upper shelf based on the strain require-
ments opposed to strain capacities. This paper gives an overview over the research work
in performance and shows recent results from experimental and numerical analyses per-
formed within this project for monotonic and cyclic loading.
Ó 2009 Elsevier Ltd. All rights reserved.

1. Introduction

The strong seismic events in Northridge (USA, 1994) and Kobe (Japan, 1995) caused severe damages of beam-column con-
nections by brittle fracture. In Kobe during the 1995 Hyogoken-Nanbu earthquake, the damage of the steel frame buildings
was governed by the fracture process consisting of three parts, ductile crack initiation at the weld toe representing hot-spot
concentration, subsequent stable crack growth and sudden explosive crack propagation in the brittle mode [1]. The brittle
fracture occurred before the stress–redistribution in the beam and formation of plastic hinges were possible. The causes
of damage were too high assumed design values for the material resistance due to unconsidered strain rate effect, large scale
cyclic straining and residual stresses. These may lead to degradation of the material toughness and ductility especially in the
presence of mechanical and metallurgical defects and heterogeneities, e.g. welding defects or stress risers in the local brittle
zones of weld or heat affected zone.
In recent years many research studies [2–4] have been initiated in the USA and Japan with the objective to investigate
deformation and fracture behaviour of steel moment connections with respect to seismic loading. The results of these
investigations were improved design rules for connections, recommendations concerning quality control of welding, the

* Corresponding author. Tel.: +49 203 999 3174; fax: +49 203 999 4352.
E-mail address: [email protected] (A. Nonn).

0013-7944/$ - see front matter Ó 2009 Elsevier Ltd. All rights reserved.
doi:10.1016/j.engfracmech.2009.03.004
1532 W. Bleck et al. / Engineering Fracture Mechanics 76 (2009) 1531–1547

amendments of existing seismic codes [5] and development of new engineering methods, e.g. WES 2808 [6]. Parallel to re-
search activities in the USA and Japan, the European scientists have recognized the necessity to revise European seismic
codes especially Eurocode 8. Within the project called ‘‘Recos” [7,8] the seismic behaviour of beam-to-column connections
were examined in dependence of different influences, e.g. strain rate effects, connection typology, column size, etc.
The major aim of the European project ‘‘Plastotough” started in 2005 is to define the limit state for structures subjected to
certain plastic strain requirements and to derive design rules for the choice of material for these structures based on an
upper shelf criterion by applying fracture (FMA) and damage (DMA) mechanics approaches. One of the innovations com-
pared to previous projects is the use of damage mechanics models, which allow for prediction of ductile crack initiation
and propagation under monotonic and cyclic loading. In this context, it should be noted, that plastic design rules for mono-
tonic and cyclic loading conditions based on the earthquake situation will be derived for the limit states defined by types of
failure, such as local buckling or ductile tearing. The modifications of these rules by taking into account the results from
investigations on the strain rate effects and the process of transition from ductile to brittle fracture are planned within
the subsequent project.
The first part of the paper gives a short overview over the available ductility criteria in plastic design of steel structures
and highlights the lack of upper shelf toughness criteria in the European standards. In the second part the applied models are
presented where beside well established fracture mechanics models also damage mechanics models are shown. Afterwards,
the investigated plate and beam materials are characterized.
Subsequently the results from small scale tests and their numerical utilization for the calibration of damage models are
described in detail. Furthermore small scale tests are performed to obtain the basic characterization of the analysed steel
grades and to determine fracture mechanics characteristic values. The simulation of ductile fracture is performed by apply-
ing different models in dependence on the type of loading (monotonic or cyclic). The widespread GTN (Gurson–Tvergaard–
Needleman) damage model is used in case of monotonic loading. Once damage parameters are identified, this damage model
can be applied for different geometries. In case of cyclic loading features like cyclic hardening, softening and Bauschinger
effect are taken into account by the extended GTN model, so called LPD (Leblond–Perrin–Devaux) model. Additionally the
effective damage concept according to Toyoda [22,23] with ‘‘advanced two-parameter criterion” is also evaluated with re-
spect to predict ductile crack initiation in steel structures under cyclic loading. The parameters necessary for cyclic damage
models are obtained from deformation controlled tests on notched round bar tensile specimens and CT specimens with
blunted notch.
The first results from large scale tests are presented in the next part of the paper. These tests are performed with wide
plate specimens under tension and beam-to-column connection under bending to examine deformation and fracture behav-
iour in terms of strain, rotation and load bearing capacity. By means of these test results, the numerical models are calibrated
and optimised with respect to global load–deformation behaviour and crack initiation. Additionally results from large scale
tests are used to demonstrate the transferability of the damage parameters and fracture mechanics characteristic values
determined by small scale tests.
The last part of the paper deals with the numerical simulations with FMA and DMA. Both approaches will be used for
parametric studies to determine toughness and strain requirements by varying crack sizes, steel grades, type of loading
and geometry of joints and members. The results should allow for the development of general rules valid for the parameter
range investigated. Within DMA, the evolution of damage in the welded connection is demonstrated by the LPD model in
dependence on different crack size and loading amplitude.

2. Ductility and toughness requirements in plastic design of steel structures

Plastic design is used to take advantage of stress–redistributions due to plastic strains exceeding the yield strain. Such
plastic strains may occur with different spatial extensions and magnitudes, as it is shown in Table 1 for beams in bending.
The plastic design of steel structures is mainly applied in two areas:

Table 1
Levels of plastic strain requirements.

Level Type of Resistance of beams Resistance of connections

analysis
1 Elastic Elastic distribution of stresses in beams Local plastic strains resulting from
Plastic strains only local e.g. adjacent to No moment assumptions for simplified ‘‘elastic”
holes redistribution distribution
2 Elastic Plastic distribution of stresses at one point of the Local plastic strains yielding from
Plastic strains limited to the No moment structure at a time that receives the maximum plastic redistributions of forces in a
exploitation of full cross sectional redistribution elastic moment connection
resistance
3 Plastic Plastic distribution of stresses at more than one Plastic distribution of forces
Plastic strains limited to plastic rotation Moment point of the structure at a time effecting plastic together with rotations permitting
requirements for exploitation of full redistribution moment redistributions by rotations also plastic redistributions
resistance of structure
 W. Bleck et al. / Engineering Fracture Mechanics 76 (2009) 1531–1547 1533

1. For structures subjected to quasi static loads where the loading at ultimate limit state (from persistent, transient or acci-
dental load combinations) is assumed to be applied so slowly, that no time dependent and repetitive effects occur, and
resistances are therefore determined in tests with a slow increase of loads (often in load increments with rest time in
between to allow for complete yielding).
2. For structures subjected to loads varying with time where loads are not increased monotonously to the ultimate limit
state, but may change repetitively the direction or magnitudes, so that the following effects may occur:
(a) action effects may be influenced by acceleration of masses (e.g. for seismic actions or impact loads),
(b) time history effects on resistances may occur, e.g.
– plastic strain increments may accumulate to extreme values,
– cyclic strain effects may lead to low cycle fatigue due to the Bauschinger effect on the stress–strain-behaviour,
– strain rate effects (strain velocity) may reduce the material toughness,
– cyclic plastic strains may increase the temperature of the structure.

For quasi static actions no specific toughness criteria so far exist for steels to be used in plastic design. EN 1993-1-1 [19]
only recommends the following ductility criteria from tensile tests of steels used in plastic design:

– fu/fy P 1.10,
– elongation at failure not less than 15%,
– eu P 15 ey, where ey is the yield strain (ey = fy/E).

These recommendations are still debated as they are based on engineering judgements and so far cannot be justified by
comparison between requirements and performances.
Beside these ductility criteria concerning the material itself the European standards give global requirements for the
endurable plastic rotation of a connection. In structures where the full resistance of the structure is exploited (level 3) plastic
hinges occur at the structural points where the bending moment exceeds the maximum plastic moment of the member. To
exploit the full plastic resistance of the structure the members at these locations have to endure a certain plastic rotation
without failure. This amount of this rotation is called rotation capacity.
The requirements for the rotation capacity are defined in EN 1993-1-8 [21] for the design of connections and EN 1998-1
[20,21] for seismic design. For welded connections in EN 1993-1-8, 6.4.3 (1) the required rotation capacity is defined by
hc
/Cd ¼ 0:025 
hb
with hc is the height of column section, and hb is the height of beam section.
In EN 1998-1, 6.6.4 (3) the required rotation capacity for earthquake design for beam-to-column connections is defined in
dependency on the ductility class. For ductility class DCM the rotation capacity has to be more than hp = 25 mrad. For duc-
tility class DCH the required value is hp = 35 mrad.
From these global requirements it is not possible for the design engineer to derive the required material toughness as the
rotation capacity depends on the steel grade, the selected member and the structural configuration.
For both design situations an upper shelf toughness criterion for the choice of material is still missing. Thus, plastic design
both for quasi static actions and for time varying actions is limited to structures not exhibited to low temperatures and
where for normal temperatures plastic design is justified by a set of severe test requirements, that penalise the use of steel.
The possible toughness criterion in the upper shelf region can be written in terms of a limit state
appl;d  Mat;d ð1Þ
where eappl,d is the applied plastic strain which occurs under a given design load at the hot-spot of a structural component
and eMat,d is the design value of the material resistance which will be determined as a critical equivalent plastic strain at a
certain point (e.g. crack initiation). The applied plastic strains eappl,d can be determined by numerical simulations (e.g. ABA-
QUS). The material resistance eMat,d can be determined either by small scale tests or by application of damage models.

3. Applied models

For the derivation of a toughness criterion in the upper shelf region both fracture mechanics and damage mechanics mod-
els are applied. While for the fracture mechanics approach the assumption of a crack-like defect is necessary damage
mechanics can be applied with or without integration of an initial flaw.
For both approaches the failure criterion ‘‘crack initiation” is used, either via J-integral (here JIc or Ji) or damage mechanics
(eu).

3.1. Fracture mechanic models for monotonic loading

According to the lower shelf region the toughness criteria shall be based on fracture mechanics assuming a structural de-
tail containing an initial crack from fabrication and/or fatigue loading. The existence of plastic strains leads to the utilization
of elastic plastic fracture mechanics by using the well established J-integral.
1534 W. Bleck et al. / Engineering Fracture Mechanics 76 (2009) 1531–1547

The limit state for the fracture mechanics approach is given in terms of the J-integral by
J appl;d  J Mat;d ð2Þ

but can also be transferred into a limit state of plastic strains.

The action effect Jappl,d arises at the crack tip of the assumed crack which is integrated at the hot-spot of a typical com-
ponent. These crack driving forces (CDF) can be calculated numerically by the use of the finite element code ABAQUS. The
CDF is dependent on the structural detail, the (nominal) deformation to which the structure is subjected, the type of loading
(monotonic or cyclic) and the material.
The material resistance JMat,d is set to either the physical crack initiation value Ji or the technical initiation value JIc. Ji char-
acterises the actual beginning of crack growth and is determined experimentally in small scale tests, e.g. by the potential
drop method according to ASTM. Ji shows no dependency on geometry and can therefore be transferred from small scale test
result to a structural detail. However this value is very conservative since no crack growth and constraint effects are utilised.
The technical initiation value JIc is less conservative, because it considers stable crack growth of about 2 mm. It is determined
from crack resistance curves obtained by the multi-piece test method and shows a dependency on many parameters, e.g.
crack length, test piece dimension and thus stress state. Therefore it is not easily transferred to the structural component.
Another possibility to quantify the material resistance is the crack resistance curve JR (Da) which characterises the tough-
ness resistance against instable crack growth and can be determined experimentally in small scale tests. The crack resistance
curve is a material value dependent on the stress triaxiality (specimen geometry, size, loading, etc.) and therefore cannot be
easily transferred to the structural component.

3.2. Damage mechanics models for monotonic loading

A widespread constitutive model for damage evolution is the GTN (Gurson–Tvergaard–Needleman) model. In this model
the damage effects are taken into account by a softening term in the plastic potential, which is usually related to volume
fraction of voids in the material. This internal variable induces a progressive shrinkage of the yield surface until failure occurs
due to loss of stress carrying capability.
Based on the von Mises yield condition extended with material damage parameters, the GTN damage model describes the
formation, development and the coalescence of the voids during stable crack growth [9–11]
 
  r2eq 3q2 rm 2
U rm ; req ; f  ¼ þ 2q1 f 
cosh  ð1 þ q3 f  Þ ¼ 0; ð3Þ
r2 2r

where req is the von Mises equivalent stress, rm the hydrostatic stress. This relation includes isotropic hardening stress by
taking into account the flow stress r of the ‘‘voidless” matrix material, which is a function of the accumulated plastic strain
epleq .
r ¼ ry þ Cðepleq Þn ð4Þ

The coefficients q1, q2, and q3 are phenomenologically based parameters, introduced to consider the interaction between
adjacent voids. The modified void volume fraction f* is defined by
8
> f f 6 fc
< 

f ¼ f c þ jðf  fc Þ fc < f < ff with j ¼ ffuf f
fc
ð5Þ
>
: 
c

fu f P ff

The modified void volume fraction f* corresponds to the void volume fraction f, unless the critical porosity fc is exceeded.
Thereafter the accelerated void growth, quantified by the factor j, sets in till final damage values quantified by ff and fu are
reached.
The evolution equation consists of terms for growth of existing voids and nucleation of new voids

f_ ¼ f_ growth þ f_ nucleation ð6Þ

The growth term is based on the assumption of incompressible matrix material

f_ growth ¼ ð1  f Þ  e_ pl
kk ; with plastic volume dilatation rate e_ pl
kk ð7Þ
The finite element code ABAQUS [12] is used for numerical simulation with GTN model, which is included as a user-sub-
routine [13].
A tool to predict local failure behaviour of structures is also given by damage curves which present a quantitative relation
between equivalent plastic strain and stress triaxiality, Fig. 1. The main difference between this approach and the GTN model
lies in avoidance of a ‘‘microscopical” damage parameter and exclusive application of continuum mechanics or ‘‘macro-
scopic” values.
 W. Bleck et al. / Engineering Fracture Mechanics 76 (2009) 1531–1547 1535

asticc strain εeqpl

p
crack initiation

E uiva
Equ ent pla
ale
Stress triaxiality h

Fig. 1. Typical damage curve.

3.3. Damage models for cyclic loading

The failure of the structures subjected to cyclic loading resulting form seismic events occurs usually after less than ten
cycles. This fatigue conditions, characterized by a few reverse (in general less than 20) loading cycles with large amplitude
scale straining is termed as ultra low cycle fatigue (ULCF). The damage mechanism during ULCF is a combination of fracture
process and fatigue mechanism. The fracture process during monotonic loading is governed by nucleation, growth and coa-
lescence of micro-voids and thus depends strongly on microstructure. The so called slip and decohesion mechanism behind
low cycle fatigue results from blunting and re-sharpening of the crack tip.
Fig. 2 shows the leading damage mechanisms in dependence on type of loading.
The main controlling factor for the ductile crack initiation under cyclic loading is the nucleation of micro-voids in a soft
phase. This nucleation results from the development of dislocation density near ferrite–pearlite interface due to the alternate
loading. Therefore the kinematic hardening described by the Bauschinger effect influences not only the load–deformation-
behaviour but also the material damage aspect. The latter is the consequence of the accumulation of long-range internal
stresses due to pile of dislocations near obstacles like inclusions and phase boundaries.
In [14] a cyclic void growth model (CVGM) has been introduced to simulate ductile fracture initiation due to ULCF in steel
connections. This model is an extension of the Rice and Tracey’s void growth model to cyclic loading and takes into account
cyclic accumulation of the void growth and coalescence with progressive damage.
Instead of CVGM two different models are applied within the damage analysis for cyclic loading. The first model is the LPD
model, which presents the extension of the GTN model [15]. The extension refers to kinematic hardening, which replaces the
Caushy stress tensor by the difference of the Cauchy stress and backstress tensor in Gurson’s yield function. The cyclic hard-
ening behaviour is described with the model of Chaboche, which combines kinematic and isotropic hardening. An analytical
relation with the difference between saturation Q value and initial yield stress r0 is applied for the isotropic flow stress

r0 ¼ r0 þ Q 1 ð1  ebep Þ ð8Þ
The kinematic hardening of the model is described by equation (9)
1
a~_ ¼ C e_ p ðr
~ a ~ ep
~ Þ þ ca ð9Þ
r0
log( σ)

MONOTONIC FRACTURE
ULTRA LOW-CYCLE FATIGUE
void growth and coalescence

LOW-CYCLE FATIGUE
intergranular fracture

HIGH-CYCLE FATIGUE

log(N)

Fig. 2. Definition of failure mechanism in dependence of number of cycle.

1536 W. Bleck et al. / Engineering Fracture Mechanics 76 (2009) 1531–1547

This is the basic Ziegler law, to which a recall term has been added to introduce the nonlinearity in response. This model
allows reproducing most of the features of cyclic plastic behaviour like the Bauschinger effect, cyclic hardening or softening,
ratcheting of strain under constant stress amplitude and relaxation of mean stress. The model parameters have to be ad-
justed, until the best correspondence between numerical and experimental results is achieved.
The LPD model has been implemented as user-subroutine in FE code ABAQUS [16]. In [17] this model has been examined
with respect to accurate damage prediction due to cyclic plastic loading under different stress triaxiality levels. Based on this
study, it can be concluded that the LPD overestimates void growth but on the other hand it is able to capture some quali-
tative features of deformation.
Additionally to the LPD model, the effective damage concept [18] according to Toyoda with the ‘‘advanced two-parameter
criterion” is also applied for damage analyses. It is based on the idea that the material damage is only controlled by the effec-
tive plastic strain. The crack initiation occurs when the cumulative effective plastic strain depending on the stress triaxiality
reaches the two-parameter critical condition, which is defined as a damage curve presented in Fig. 1. The ‘‘advanced two-
parameter criterion” is proposed for evaluation of ductile crack initiation in steel structures under cyclic loading. It includes
the following ideas:

1. Material damage is only affected by the effective plastic strain, where the long-range internal stress is higher than the
maximum internal stress under preceding loading.
2. Ductile cracking occurs when the cumulative effective plastic strain as a function of stress triaxiality reaches the two-
parameter critical conditions.

The advantage of this concept is that the derived limit curves for monotonic loading (damage curves) are also valid for
cyclic loading.

4. Materials

Within the research work both plate and beam materials are investigated. For the plate material three different steels
according to EN 10025 were investigated in different thicknesses: S355J2 + N, S460N and S690QL. Table 2 gives the chemical
composition and Table 3 the mechanical properties of the plate material.
For the beam material two different steel grades (S355J2 and S460M) with two different sections (IPE 500 and HEA 300)
are examined. Table 4 gives the chemical composition and Table 5 the mechanical properties of the beam material.

5. Small scale tests – experimental and numerical results

The experimental and numerical investigations are performed on the small scale specimens in order to calibrate damage
parameters for monotonic and cyclic loading and to determine damage curves.

Table 2
Chemical composition of the plate material (mass%).

Heat C Mn Si P S Cu Ni Cr Mo V Ti Al N Nb
Steel S355J2 + N
16,620 0.17 1.49 0.34 0.017 0.005 0.03 0.02 0.06 0.01 0.004 0.002 0.033 0.008 0.003
24,914 0.18 1.52 0.35 0.018 0.009 0.04 0.04 0.08 0.01 0.004 0.002 0.048 0.007 0.003
Steel S460N
21,967 0.18 1.60 0.54 0.014 0.005 0.05 0.72 0.06 0.01 0.057 0.025 0.033 0.008 0.031
12,937 0.20 1.60 0.52 0.015 0.004 0.04 0.75 0.05 0.01 0.062 0.028 0.039 0.006 0.034
Steel S690QL
24,827 0.15 0.87 0.31 0.013 0.005 0.06 0.85 0.77 0.33 0.004 0.020 0.050 0.009 0.003
16,989 0.15 0.90 0.28 0.009 0.003 0.03 0.79 0.79 0.37 0.004 0.004 0.059 0.007 0.003

Table 3
Mechanical properties of the plate material.

Steel Heat Thickness (mm) ReH (MPa) Rm (MPa) A5 (%) KV (J) T (°C) Orientation
S355J2 + N 16620 40 359 546 30.0 221 20 L
S355J2 + N 24914 20 400 584 28.6 95 20 L
S460N 21967 31 471 637 29.6 44 0 T
S460N 12937 21 470 626 26.4 38 0 T
S690QL 24827 40 764 834 15.2 131 40 T
S690QL 16989 20 753 823 16.4 126 40 T
 W. Bleck et al. / Engineering Fracture Mechanics 76 (2009) 1531–1547 1537

Table 4
Chemical composition of the profile material (mass%).

Profile C Si Mn P S N Al V Nb Ti Cu Cr Ni Mo
Steel S355J2
IPE 500 0,09 0,18 1,50 0,011 0,008 0,007 0,025 0,06 0,01 0,00 0,06 0,03 0,05 0,00
HEA 300 0,09 0,16 1,48 0,019 0,004 0,008 0,029 0,05 0,01 0,00 0,09 0,05 0,05 0,00
Steel S460M
IPE 500 0,09 0,23 1,63 0,014 0,001 0,01 0,025 0,11 0,03 0,00 0,01 0,03 0,03 0,00
HEA 300 0,09 0,22 1,62 0,012 0,002 0,010 0,026 0,11 0,02 0,00 0,02 0,03 0,06 0,00

Table 5
Mechanical properties of the profile material.

Profile Steel grade ReH (MPa) Rm (MPa) A5 (%) KV1 (J) KV2 (J) KV3 (J) KV (J)
IPE 500 S355J2 442 525 32 223 239 233 232
HEA 300 447 520 33 259 254 259 257
IPE 500 S460M 503 548 32 271 266 276 271
HEA 300 483 557 30 270 256 265 264

For monotonic loading the damage parameters are calibrated by comparing results from numerical and experimental
analyses with respect to the fracture resistance curve determined for the CT specimens see Fig. 3. Except the crack initiation
value which is distinctly underestimated by the GTN model, the best agreement with experimental data is achieved for a

900

750
Exp.: J i=176.8 N/mm
J-Integral [N/mm]

GTN: J i=120 N/mm

600

450

300
GTN S355

150 Exp. S355

0
0.00 0.25 0.50 0.75 1.00 1.25
Crack growth Δa [mm]

Fig. 3. Numerical and experimental R-curve for S355.

20
2.0
R01(at surface) experiment
pl

-1 2807h
-1.2807h
p

ε= 2.492e
2 492 GTN
n εeq

smooth
1.5 Damageg curve
s rain

R04(at surface)
p astiic str

R05( t surface)
R05(at f )
10
1.0 smooth
nt pla

R2 R04(at
( centre))
valen

R05(at
( centre))
E uiv

05
0.5 R2 R1
Equ

R1

0.0
0.0
0 0
0.5
5 1
1.0
0 1
1.5
5 2
2.0
0
Stress triaxiality h

Fig. 4. Damage curve for S355.

1538 W. Bleck et al. / Engineering Fracture Mechanics 76 (2009) 1531–1547

crack length smaller than 0.5 mm. With growing crack the GTN model provides increasingly underestimated crack
resistance.
Fig. 4 shows the experimentally obtained damage curve from tensile tests on notched 8  40 round bar specimens. Dif-
ferent levels of triaxiality at the crack initiation are reached by varying the notch radius q = 0.4, 0.5, 1, and 2 mm indicated by
R04, R05, R1 and R2 in the diagram, respectively. The location of the crack initiation in the notched round bar specimen de-
pends strongly on the notch radius. With decreasing radius, e.g. q = 0.4 and 0.5 mm, the deviatoric stresses increase causing
the failure to occur at the surface of the notch region by formation of local shear bands. With a smoother notch geometry,
q = 1, 2 mm, the failure in the centre of the specimen is primarily controlled by the stress triaxiality.
Additionally, damage curves are determined by applying the GTN damage model in order to examine the capability of the
GTN model to predict failure in triaxiality range lower than 1.0, see Fig. 4. As expected, for h > 1.0 good agreement between
the numerical and experimental results is achieved with respect to strain level at the initiation and damage location. How-
ever for sharper notch radii like q = 0.4 mm and 0.5 mm, damage predicted by the GTN model takes place at the centre of the
specimen as opposed to experimental results. The other difference refers to the level of triaxiality at the initiation. With
h = 1.5 the triaxiality level gained by the GTN is ca. 1.0 higher than the experimental value. Only by using a much smaller
radius (q = 0.1 mm) the predicted crack initiation can be reached at the surface of the specimen at lower triaxiality
(h = 0.6) but therefore at the higher plastic strains compared to experiments. Since the damage progress resulting from
the GTN model is primarily influenced by the level of hydrostatic stresses, the effects from shear stresses, which contribute
to the damage localisation on the surface, cannot be accounted for.
The calibration of damage parameters required for the LPD model can be fulfilled by conducting tests on two types of
specimens, round bar and 1CT specimen with blunted notch, see Fig. 5. In order to test these specimens existing software
for displacement controlled monotonic testing had to be extended for cyclic loading.
In contrast to the 1CT specimen employed for the fracture mechanics tests, the thickness of 1CT specimen used for cyclic
loading is reduced to 12.5 mm. Instead of a sharp fatigue pre-notch the modified CT specimen contains additionally a
blunted notch with the radius of 1.5 mm and the depth of 1.5 mm in order to prevent overlapping of fracture surfaces during
compressive cycles. With reduced thickness the out-of-plane constraint decreases leading to locally higher plastic strains
preceding failure occurrence.
The experimental and numerical results achieved with the LPD model for the CT specimen with blunted notch are pre-
sented in Fig. 6. Due to kinematic hardening the increase of the absolute load value in the tension and compression region is
evident after two cycles. Within the further cycles the damage increases, whereas the material softening and subsequent
failure occur at the fifth cycle. The cyclic deformation behaviour predicted by the LPD model corresponds very well with

radius smooth machined notch

M16x1.5
Ø8

15 20 15
40 50 40
[mm]
130
50
25

Fig. 5. Round bar specimen and CT specimen with blunted notch.

50
40
30
20
Load F [kN]

10
0
-10 0 1 2 3 4 5 6 7 8 9

-20
-30 Exp.
-40 LPD 1-4 cycles
LPD 5 cycle
-50
Displacement [mm]

Fig. 6. Load–displacement behaviour for CT specimen with blunted notch.

 W. Bleck et al. / Engineering Fracture Mechanics 76 (2009) 1531–1547 1539

the experimental data. While the LPD model underestimates the load level slightly, hardening behaviour can be well repro-
duced. Although no previous softening is apparent, the initiation in the FE-model occurs also during the fifth cycle.
Fig. 7 shows the damage development with the displacement. After the first cycle the maximum void volume fraction is
about 0.7%. After four cycles void volume fraction of 2.3% is reached, which corresponds to the critical void volume fraction fc.
Due to the accelerated void growth, a very steep increase of void fraction leads to the final failure with f* = 67% in the fifth
cycle.
The same damage parameters, determined from the tests on CT specimens with blunted notch are applied for the simu-
lation of damage behaviour of the round bar specimen, which contains a notch defined by 1 mm notch radius and 2 mm
notch depth. In Fig. 8 the experimental and numerical load–deformation behaviour is compared with respect to load level
and number of cycles to failure. With only 5 cycles to failure, the LPD damage model underestimates significantly the dam-
age resistance of the specimen. In contrast to the experiments with 10 cycles to failure, the damage progress in the FE-model
begins already with increasing material softening during the third tension cycle and leads to final fracture after two addi-
tional cycles. Since the reasons for the deviation of the numerical results are still not clear, the limits for the application
of the LPD model as well as the parameter sensitivity must be further investigated.
Similar results compared to the LPD model are obtained with the ‘‘advanced two-parameter criterion”. The failure of the
CT specimen with blunted notch occurs with the fifth cycle, see Fig. 9. With the 5 cycle the plastic strains increase almost
upto 100% at the same triaxiality level leading to failure.
The results from the notched bar specimen differ distinctly with respect to the damage development. Already after two
cycles the damage is very high with plastic strains reaching the 100% level. However with further cycles no damage progress
is observed. Finally failure occurs during the fifth cycle when the triaxiality –strain history path intersects the damage curve
with a high increase of stress triaxiality, whereas the strain level remains unchanged.

0.050
Modified void volume fraction f*

0.045 LPD
0.040
0.035
0.030
0.025
0.020
0.015
0.010
0.005
0.000
0 2 4 6 8 10
Displacement [mm]

Fig. 7. Development of the damage for the element at the notch.

15

10

10. Zyklus
5
Load F [kN]

0
-0.2 -0.1 0.0 0.1 0.2 0.3
-5
Exp. (10 cycles)
-10 LPD (5 cycles)

-15
Displacement [mm]
Fig. 8. Load–displacement behaviour for notched round bar specimen.
1540 W. Bleck et al. / Engineering Fracture Mechanics 76 (2009) 1531–1547

1.5

n εeqpl
p
strain
10
1.0
2. cycle

p stic s
Damage curve
D

ent plas
Notched
vale 05
0.5 round bar
specimen
E uiv
Equ

CT specimen
i
4 cycle
4.
0.0
0.0 0.5 1.0 1.5 2.0
Stress triaxiality h

Fig. 9. Damage development according to effective damage concept.

6. Large scale testing – experimental and numerical results

The set up for large scale testing is based on the selected structural details due to local and global strain requirements. The
local strain requirements refer to strain concentrations at welds or at net sections in bolted members. These requirements
are investigated by means of wide plates with welded stiffeners and holes, which represent sub-models of global structures.
The analysis of global strain requirements such as bending moment resistance, formation of plastic hinges and rotation
capacity is possible with bending and rotation tests, which are conducted by beam-to-column joints under monotonic
and cyclic loading.
Like in case of small scale testing, the focus within this paper is put on results from large scales testing, which serve pri-
marily for calibration and optimisation of the numerical models.

6.1. Wide plate tests

The transferability of the GTN damage model regarding the prediction of the crack resistance is examined on wide plates
subjected to monotonic loading under room temperature. The tests are performed by means of a servo-hydraulic universal
testing machine.
One type of test is carried out on DECT specimens for steel grade S355 with plate thickness of 40 mm. The specimen
geometry and measuring instrumentation are presented in Fig. 10.
Both experimental and numerical results from tensile tests of DECT specimens with respect to global deformation behav-
iour are given in Fig. 11. The load versus opening displacement curve is slightly overestimated by the FE-model. The deter-
mination of the crack initiation by ACPD method (Alternate-Current-Potential-Drop-Method) is demonstrated in Fig. 11, in
which the change of slope of the potential curve indicates the crack initiation Ji.
By comparison of experimental Ji values for CT and DECT specimens, it is evident that often no exact constraint indepen-
dent initiation values can be determined by ACPD method. The difference between these values for the analysed S355 lies in
the range of ca. 20 N/mm. The reason for this difference can be found in the delayed change of potential with beginning of

l0 =360 t=40
a1=30
500
W=300 Strain gage, front
extensometer
double clip
a2=30
R=200
l=1100
[mm]
L=2500

Fig. 10. Specimen geometry with instrumentation (schematic).

 W. Bleck et al. / Engineering Fracture Mechanics 76 (2009) 1531–1547 1541

4500 0.20

4000 0.18

3500 0.16

0.14
3000

Load F [kN]

Potential [V]
Exp.: J i=209N/mm 0.12
2500
GTN:Ji=198N/mm
0.10
2000
0.08
1500 crack initiation
0.06
Experiments
1000 0.04
Simulation GTN Model
500 Potential 0.02

0 0.00
0.0 0.2 0.4 0.6 0.8 1.0 1.2 1.4 1.6 1.8

Opening displacement [mm]

Fig. 11. Determination of the crack initiation with ACPD method.

stable crack growth. As a consequence the initiation value is evaluated for some amounts of crack growth (often less than
0.1 mm), which can have a distinct effect on the level of crack resistance especially in case of low constraint specimens.
A stronger discrepancy of the Ji values occurs when using GTN model. The Ji value calculated by the GTN model for DECT
specimens is ca. 80 N/mm higher than for the CT specimen. This is due to the definition of the crack initiation by the GTN
damage model, according to which the initiation can be evaluated, when all integration points of the element ahead of
the crack tip reach the final damage fu . Hence, the initiation by damage analysis is defined as J-integral determined for
the crack extension, which corresponds to the element width. This is an explanation, why the numerical crack initiation in-
creases with increasing constraint level.

6.2. Rotation tests on beam-column-connections (BCC)

The 3-point-bending tests are selected to determine strain and rotation capacity of beam-column-connections (BCC) con-
taining no initial flaw. The column consists of a HEM profile, the beams of HEA 300 or IPE 500 profiles. The tests are con-
ducted with two different welding details in order to differentiate the performance of a detail sophisticated for seismic
loading (butt weld with root fillet weld) from a non sophisticated detail (fillet weld). All structural configurations are inves-
tigated with steel grades S355J2 and S460M. Fig. 12 shows the test setup for a rotation test.
The tests are performed deformation controlled until reaching a vertical displacement restricted to the maximum piston
path. Beside the load and displacement strains are measured near the vicinity of the weld seams. During the monotonic tests
neither failure nor crack initiation occurred. The selected beam-column-connections show the typical moment-rotation
behaviour with formation of a plastic hinge after a certain amount of plastic rotation, see Fig. 13.
The FEA results show good agreement to test data with respect to global load–deformation behaviour.

Fig. 12. Setup for a beam-column-connection test.

1542 W. Bleck et al. / Engineering Fracture Mechanics 76 (2009) 1531–1547

800
Et/10
700 Fy = 691 kN

600
500

load, kN
400
Test result
300 FEA

ey = 31,4 mm
200
100 Et

0
0 20 40 60 80 100 120 140 160 180 200
vertical displacement, mm

Fig. 13. Experimental results: load vs. vertical displacement (HEA 300, S460M).

Fig. 14 shows a beam-column specimen after finalization of the test and the corresponding numerical results. During the
formation of the plastic hinge large deformations occur at the compression flange of the beam in form of plastic buckles near
the welding connection. The numerical calculated deformations are in good accordance with those obtained by the test. The
geometry and location of the plastic hinge match well and also the amplitudes of the buckles fit well with those measured at
the test specimen.
The tests under cyclic loading are in preparation and will be performed till mid of 2008.

7. Numerical models

7.1. Fracture mechanics approach

7.1.1. Monotonic tension tests

As shown before several monotonic tests have been carried out at room temperature where three different types of typ-
ical details with artificial flaws were subjected to tension and driven to failure. These details were the DECT specimen, a plate
with a bore hole with cracks and a plate with longitudinal stiffener with a semi-elliptical crack at the weld toe as given in
Fig. 15.
Numerical models of these specimens have been developed in ABAQUS [12] and were calibrated by means of the test re-
sults. Especially the plates with surface cracks and additionally with welded stiffeners required a sophisticated model build-
ing due to the complicated geometry with regard to the connection ‘‘crack region-welded stiffener”. Volume elements of
type C3D20R with 20 nodes were used and the near region of the crack detail is modelled with a cuboid sub-model, where
the elements adjacent to the crack line are collapsed elements. The development of this sub-model, which allows a param-
eterised variation of the crack size (a, c), provides a method to perform a crack analysis for any rectangular plate with surface

Fig. 14. Deformed geometry of beam-column connections after finalizing test (left) and from numerical simulation (right).

Fig. 15. Specimen geometry of a plate with bore hole and cracks and plate with longitudinal stiffener and semi-elliptical crack.
 W. Bleck et al. / Engineering Fracture Mechanics 76 (2009) 1531–1547 1543

Fig. 16. Experimental results vs. numerical results: plate (S460N) with longitudinal stiffener with semi-elliptical crack.

crack, as the model can be implemented in global structural models. The sub-model is linked with the remaining model by
TIE- commands. Slight inaccuracies resulting from extrapolations performed at the intersecting surfaces are negligible small.
Three steels (S355J2G3, S460N and S690QL) are investigated in the parametric studies using nonlinear material laws. As
elastic–plastic-calculations are performed, the true stress–strain curves are implemented in the FE-models. For example,
Fig. 16 shows the load–displacements curves of a S460N with longitudinal stiffener and semi-elliptical crack. Very good con-
formity can be registered due to a high-advanced parametric numerical model with a refined spider-web mesh which is
essential to evaluate quantified local stress–strain requirements in the vicinity of the elliptical crack front. The equivalent
plastic strain and the triaxiality are determined in the vicinity of the crack front (measured at the ellipse’s vertex in a dis-
tance lc which corresponds to the microstructural length) which was experimentally measured.
Furthermore, the ACPD method provides a useful tool to specify beginning crack growth in the experiments. One can use
the measured crack initiation parameters to transfer the appropriate displacement to the FE-results and thus to determine
the upper limit Jmat,d = Ji as the toughness requirement.
Since all FE-models showed good accordance with the experimental results, a numerical tool has been developed which
provides the opportunity to extensive parametric research. First investigations by varying the controlling geometric dimen-
sions have just begun.

7.1.2. Monotonic rotation tests

Using the same techniques as for the recalculation of the wide plate tests the beam-column-connections have been mod-
elled in ABAQUS. In a first step the tests with beam-column connections have been modelled and recalculated for calibration
and optimisation purposes. The calculated results of the derived models agree well with the results of the large scale tests.
To derive the crack driving force at the critical locations of the experimentally investigated beam-column connections
numerical investigations are performed by means of FEA. For all tested configurations numerical models have been designed.
A crack-like flaw in form of a semi-elliptical or quarter-elliptical surface crack is integrated at those positions of the tension
flange where the analysis with an uncracked model gave points of stress and strain concentration.
1544 W. Bleck et al. / Engineering Fracture Mechanics 76 (2009) 1531–1547

Fig. 17. Numerical model of HEA300 with integrated semi-elliptical surface crack.

For the butt weld detail the crack is located at the weld interface between beam flange and column flange. For the fillet
weld detail the flaw is integrated at the toe of the fillet weld. In transverse direction of the beam the cracks are located either
in the middle or at the edge of the flange (see also Fig. 17).
For the integrated crack the ratio between crack depth a and crack width c is generally selected to a/c = 0.4. The dimen-
sions of the integrated cracks are starting with values according to the regulations of EN 1993-1-10 [24] and are increased up
to crack depths of 7 mm.
Fig. 18 gives the results of the crack depth variation for one profile and steel grade. Generally the J-integral increases with
growing crack size. Within the elastic range of the load–displacement curve the J-integral behaves linearly. By reaching the
elastic limit load the J-integral curve grows steeper until full plastification. After this point and during formation of the plas-
tic hinge the increase of J-integral stays linear.
By application of the limit state of the fracture mechanics approach by introducing the material resistance Jmat,d = Ji the
local strain requirement in terms of plastic equivalent strain or the global strain requirement in terms of vertical displace-
ment resp. global rotation at the point of crack initiation can be derived (see Fig. 19).

1000 1000
J-Integral (a0 = 1,354 mm) J-Integral (a0 = 7,0 mm)
900 J-Integral (a0 = 3,0 mm) J-Integral (a0 = 5,0 mm) 900
800 load 800
J-Integral, N/mm

700 700
600 600
load, kN

formation of plastic hinge

500 500
400 400
300 300
200 200
100 100
0 0
0 25 50 75 100 125 150 175 200
vertical displacement, mm
Fig. 18. Load vs. deflection and J-integral vs. vertical displacement (HEA 300, S460M, fillet weld).
 W. Bleck et al. / Engineering Fracture Mechanics 76 (2009) 1531–1547 1545

600 0.6 600 0.6

J-Integral J-Integral
J-Integral [N/mm]

500 0.5 500 0.5

J-Integral [N/mm]
eps.pl.eqv eps.pl.eqv

400 0.4 400 0.4

εpl.eqv [-]
εpl.eqv [-]
300 0.3 300 0.3
Ji = 200 N/mm
Ji = 200 N/mm
200 0.2 200 0.2

100 0.1 100 0.1

0 0.0 0 0.0
0 25 50 75 100 125 150 175 200 0 25 50 75 100 125 150 175 200
vertical displacment [mm] vertical displacment [mm]

Fig. 19. Numerical results: J-integral vs. vertical displacement and equivalent plastic strain vs. vertical displacement for beam-column connection (HEA
300, S460M, fillet welds), left side: crack size a0 = 1.354 mm, right side: crack size a0 = 7.0 mm.

By this means the fracture mechanics approach covers the monotonic loading situation. Further investigations will be
done in parametric studies varying profile, steel grade and detail.

7.2. Damage mechanics approach

The damage analysis of the beam-column-connection is carried out for S355 with the same parameter set used for the CT
and DECT specimens. There are two competing failure locations for the beam-column-connections, namely the flawed
welded connection in the tension flange with initiation of ductile or cleavage fracture and the compression flange with plas-
tic hinge formation due to the buckling, see Fig. 20. Which location will fail first depends on the material properties, the ini-
tial crack shape and size, as well as on the geometry of the steel construction (used profiles and welded connection).
The ductile fracture in the welded connection under monotonic loading is examined by the GTN model for two different
crack sizes a = 1.39 and 5 mm. The semi-elliptical surface cracks are located at the weld toe in the centre of the tension
flange. The validity of the GTN model in the crack area is ensured, since the triaxiality level ahead of the crack tip is higher
than 1.0 for both cracks at the initiation. For both cracks the location of the crack initiation is at the angle of 66° (0° at the
crack vertex and 90° at the free surface). The initiation is primary governed by the stress triaxiality, which reaches the high-
est value at the angle of 66°.

Fig. 20. Buckling of the compression flange and damage development in the tension flange, S355.

1000

800
J-Integral [N/mm]

Ji=598 N/mm
a=1.39 mm
600 a=5 mm

400 J =257 N/mm

i

200

0
0 0.2 0.4 0.6 0.8 1 1.2 1.4
Crack growth Δa [mm]

Fig. 21. Crack resistance curves, BCC.

1546 W. Bleck et al. / Engineering Fracture Mechanics 76 (2009) 1531–1547

The failure behaviour of BCC with a = 5 mm is characterized by the beginning of stable crack growth in the weld connec-
tion. In contrast, buckling of BCC is observed prior to ductile crack initiation for smaller cracks. In Fig. 21 the crack resistance
curves calculated for two crack sizes differ significantly. The dependence of numerical crack initiation on the constraint level
leads to more than two times higher crack initiation for a = 1.39 mm than for larger crack. The difference in crack resistance
for two cracks increases continuously with crack propagation. The reasons for such a strong influence of the crack size on the
damage behaviour leading to significant discrepancy between the crack resistance curves will be investigated by further var-
iation of the crack initial size, BCC geometry and applied steel grade.
The equivalent plastic strains and stress triaxialities at the crack initiation are summarized in Fig. 22 for all investigated
specimens under monotonic loading. The h values for DECT, CT specimens and BCC with a = 5 mm are set in the narrow range
between 2 and 2.5. As already indicated by the R-curve, the constraint level for the BCC with a = 1.39 mm is distinctly re-
duced to 1.25. When compared to the damage curve, the GTN enables the utilization of safety reserves for h > 2.0.
By applying the LPD model, the damage evolution in the flawed region of the BCC is also examined for two different crack
sizes a = 1.39 mm and a = 5 mm and additionally for two different levels of cyclic amplitudes (A = 80 mm and A = 120 mm),
see Fig. 23. In case of a = 5 mm, the damage level reached after 2.5 cycles is with 1.5% near the critical porosity (fc = 2%) and
distinctly higher than the damage level of 0.5% obtained for a = 1.39 mm after 4 cycles. For a = 5 mm the increase of the cyclic
amplitude from 80 to 120 mm leads to accelerated damage development already after a first cycle with a maximum damage
level of 12%. With the same basic parameters and the same flow curves concerning isotropic hardening used for the GTN and
the LPD model, the deflections at which the critical porosity is reached should be the same for monotonic and cyclic loading.
With a difference of 20 mm regarding the deflection at the onset of fc, it can be concluded that both models require separate
calibrations of the basic parameters. Due to the instability of the FE calculation with the LPD model, the final number of

0.7

06
0.6
n εeqqpl

BCC a=1
BCC, a=1.39
39 mm
0.5
s rain
p astiic str

0.4 DECT
BCC a=5 mm
BCC,
nt pla

03
0.3
CT
vallen

0.2
Equiv

GTN S355
GTN-S355
Eq

0.1 Damage
g curve

0.0
0 0
1.00 1.25 1.50 1.75 2.00 2.25 2.50 2.75
Stress triaxiality h

Fig. 22. Quantified local stress and strain state.

0.040 0.14
A=80 mm Monotonic a=1.39 mm
Monotonic a=5 mm a=5 mm
Modified void volume fraction f*
Modified void volume fraction f*

0.035 0.12
Cyclic a=1.39 mm
Cyclic a=5 mm
0.030 Monotonic
0.10
Cyclic A=80 mm
0.025
Cyclic A=120 mm
0.08
0.020 1 cycle
2.5 cycles
0.06
0.015
0.04
0.010
4 cycles 2.5 cycles
0.005 0.02

0.000 0.00
0 50 100 150 200 250 0 20 40 60 80 100 120
Deflection U [mm] Deflection U [mm]

Fig. 23. Damage development due to cyclic loading in dependence on cyclic amplitude level and crack size.
 W. Bleck et al. / Engineering Fracture Mechanics 76 (2009) 1531–1547 1547

cycles to failure has not been determined. In these cases the application of the ‘‘advanced two-parameter criterion” is
recommended.

8. Conclusions and outlook

In this paper the recent results from the European project ‘‘Plastotough” are presented. The main objective of this project
is to provide recommendations and practical criteria for plastic design of steel moment connection subjected to monotonic
and seismic loading based on extensive numerical and experimental work. This paper deals primary with fracture and dam-
age mechanics tools employed to quantify the strain capacity and strain requirements for typical structural details.
Damage models are presented, which differ in the approach (micromechanical or phenomenological) and the application
field (monotonic or cyclic loading). The small scale tests serve to determine required parameters. The great advantage of the
micromechanically based model is the possibility to predict the damage development in arbitrary steel construction once the
parameters are determined.
The main conclusions, which can be drawn, are:

– For the low constraint level quantified by the stress triaxiality h < 1, the damage curves for monotonic and ‘‘advanced two-
parameter criterion” for cyclic loading are suitable to describe the damage evolution. The damage curves can be deter-
mined by means of tests on the notched specimens with different notch radii. Although the derivation of damage curves
might be time consuming, it is indispensable not only for damage analyses due to monotonic loading but also for the
application of the ‘‘advanced two-parameter criterion.
– For the high constraint level (h > 1) which is typical for flawed structures, either the fracture mechanics (FMA) or the dam-
age mechanics approach (DMA) can be used. Within the DMA, micromechanics-based models such as the GTN and the LPD
model are applicable for damage analyses.

The transferability of the models for cyclic loading will be demonstrated on the rotation tests under cyclic loading. Fur-
thermore, the ability of the DMA will be examined regarding the prediction of the location and failure mechanisms depend-
ing on the alternating level of amplitudes and their succession during cyclic loading. These investigations will be carried out
further on for the unflawed steel moment connections. At the end of the project the design rules will be derived based on the
achieved results.
This paper shows the current research work of the European Project ‘‘Plastotough – Modern plastic design for steel struc-
tures”. The financial support by the Research Fund for Coal and Steel of the European Commission is gratefully
acknowledged.

References

[1] Kuwamura H. Fracture of steel during an earthquake-state-of-the-art in Japan. Engng Struct 1998;20:310–22.
[2] Matos CG, Dodds RH. Modeling the effects of residual stresses on defects in welds of steel frame connections. Engng Struct 2000;22:1103–20.
[3] Shama AA, Zarhamee MS, Ojdrovic RP, Schaefer BW. Sesmic damage evolution of a steel building using stress triaxiality. Engng Struct 2003;25:271–9.
[4] Stojadinovic B. Stability and low-cycle fatigue limits of moment connection rotation capacity. Engng Struct 2003;25:691–700.
[5] FEMA. Recommended seismic design criteria for new steel moment-frame buildings, FEMA-350; 2000.
[6] Minami F, Ohata M. Fracture mechanics assessment of beam-to-column joints subjected to cyclic and dynamic loading. Weld World 2007;51(5/
6):22–33.
[7] Mazzolani FM. Moment resistant connections of steel frames in seismic areas: design and reliability. E&FN Spoon; 2000.
[8] Mazzolani FM. Steel and composite structures in European seismic areas: research, codification, design and applications. Earthquake Spectra
2003;19(2):415–52.
[9] Tvergaard V. Influence of void nucleation on ductile shear fracture at a free surface. J Mech Phys Solid 1982;30:399–425.
[10] Tvergaard V, Needleman A. Analysis of the cup-cone fracture in a round tensile bar. Acta Metall 1984;32:157–69.
[11] Gurson AL. Continuum theory of ductile rupture by void nucleation and growth: part I – yield criteria and flow rules for porous ductile media. J Engng
Mater Technol-Trans ASME 1977;99:2–15.
[12] Hibbitt, Karlsson & Sorensen Inc. Abaqus Finite Element Program, Version 6.5; 2004.
[13] Siegmund T, Brocks W. A user-material subroutine incorporating the Gurson–Tvergaard–Needleman model of porous metal plasticity into the ABAQUS
finite element program, Technical report GKSS/WMG/97/2. Geesthacht: Institut für Werkstofforschung, GKSS; 1997.
[14] Kanvinde AM, Deierlein GG. Cyclic void growth model to assess ductile fracture initiation in structural steels due to ultra low cycle fatigue. J Engng
Mech 2007;133(6):701–12.
[15] Leblond JB, Perrin G, Devaux J. An improved Gurson-type model for hardenable ductile metals. Eur J Mech A-Solid 1995;14:499–527.
[16] Mühlich U, Brocks W. On the numerical integration of a class of pressure-dependent plasticity models including kinematic hardening. Comput Mech
2003;31:479–88.
[17] Steglich D, Pirondi A, Bonora N, Brocks B. Micromechanical modeling of cyclic plasticity incorporating damage. Int J Solid Struct 2005;42:337–51.
[18] Ohata M, Toyoda M. Damage concept for evaluating ductile cracking of steel structure subjected to large-scale cyclic straining. Sci Technol Adv Mater
2004;5:241–9.
[19] EN 1993-1-1: Eurocode 3: Design of steel structures – Part 1-1: General rules and rules for buildings.
[20] EN 1998-1: Eurocode 8: Design of structures for earthquake resistance – Part 1: General rules, seismic actions and rules for buildings.
[21] EN 1993-1-8: Eurocode 3: Design of steel structures – Part 1-8: Design of joints.
[22] Toyoda M, Ohata M, Ayukawa N, Ohwaki G, Ueda Y, Takeuchi I. Ductile fracture initiation behavior of pipe under a large scale of cyclic bending. In:
Proceedings of the third international conference, vol. 2, Brugge, Belgium; 2000. p. 87-102.
[23] Toyoda M, Ohata M, Yokota M, Yasuda O, Hirono M. Criterion for ductile cracking for the evaluation of steel structure under large scale cyclic loading.
In: Proceedings of the 20th international conference on offshore mechanics and arctic engineering, Rio de Janeiro, Brazil, 2001, OMAE01/MAT-3103.
[24] EN 1993-1-10: Eurocode 3: Design of steel structures, Part 1-10: Selection of materials for fracture toughness and through-thickness properties.