materials

Article
Improvement of Cyclic Void Growth Model for
Ultra-Low Cycle Fatigue Prediction of Steel
Bridge Piers
Shuailing Li, Xu Xie * and Yanhua Liao
College of Civil Engineering and Architecture, Zhejiang University, Hangzhou 310058, China;
11612057@zju.edu.cn (S.L.); 21512196@zju.edu.cn (Y.L.)
* Correspondence: xiexu@zju.edu.cn; Tel.: +86-0571-8820-6572




Received: 3 April 2019; Accepted: 7 May 2019; Published: 16 May 2019


Abstract: The cyclic void growth model (CVGM) is a micro-mechanical fracture model that has
been used to assess ultra-low cycle fatigue (ULCF) of steel structures in recent years. However,
owing to the stress triaxiality range and contingency of experimental results, low goodness of fit is
sometimes obtained when calibrating the model damage degradation parameter, resulting in poor
prediction. In order to improve the prediction accuracy of the CVGM model, a model parameter
calibration method is proposed. In the research presented in this paper, tests were conducted on
circular notched specimens that provided different magnitudes of stress triaxiality. The comparative
analysis was carried out between experimental results and predicted results. The results indicate that
the number of cycles and the equivalent plastic strain to ULCF fracture initiation by the CVGM model
calibrated by the proposed method agree well with the experimental results. The proposed parameter
calibration method greatly improves prediction accuracy compared to the previous method.

Keywords: ultra-low cycle fatigue; cyclic void growth model; circular notched specimens; steel bridge
piers; high stress triaxiality; moderate stress triaxiality

1. Introduction
After the 1994 Northridge earthquake in California and the 1995 Kobe earthquake in Japan, it was
observed that beam-to-column connections and baseplate connections in steel bridge piers undergo
ultra-low cycle fatigue (ULCF) in such events [1–3]. This ULCF damage has been shown to cause the
progressive collapse of entire structures [4]. The ULCF is characterized by a ductile crack that initiates
at the strain concentration position, then this crack expands stably under cyclic loading, and finally,
the catastrophic failure occurs in the brittle mode. ULCF has been shown to occur in the areas of strain
concentration in steel bridge piers and beam-column connections under cyclic loading [5–7]. Unlike
traditional high cycle and low cycle fatigue, ULCF experiences large plastic strain amplitude and is
usually characterized by few reverse loading cycles (in general less than 100). Therefore, ULCF is of
great significance in the seismic design of steel structures.
The Coffin-Manson formula [8,9] has been widely used to predict the low cycle fatigue life
of steel structures. In order to predict ULCF life, Ge et al. [10–12] introduced a damage index to
evaluate the ULCF life in steel bridge piers based on the Coffin-Manson formula and Miner’s rule [13].
Tateishi et al. [14] developed a new fatigue prediction model that can accurately predict the fatigue life
of plain material in an extremely large strain range. Xue [15] proposed a uniform expression to predict
low cycle fatigue and ULCF by introducing an exponential function and additional material parameters.
However, the above empirical models are derived under uniaxial strain conditions and cannot be
applied to a multiaxial stress condition. Micro-mechanism-based models have been proposed to solve
these problems in recent years and will be discussed below.

Materials 2019, 12, 1615; doi:10.3390/ma12101615 www.mdpi.com/journal/materials

Materials 2019, 12, 1615 2 of 17

The micro-mechanical models in the literature can be classified into coupled and uncoupled
models [16]. The coupled models consider the intercoupling between material constitutive properties
and damage. Mear et al. [17] and Leblond et al. [18] modified the Gurson–Tvergaard–Needleman (GTN)
model for cyclic loading. Tong et al. [19] proposed a model based on continuous damage mechanics
(CDM) to investigate the ULCF behaviour of beam-column connections. However, in coupled models,
the model parameter calibration is a complex task due to the interdependency among the parameters
as well as the high calculation cost. These shortcomings impede the application of the coupled
models. The uncoupled models can be efficient for crack initiation modelling. Since the uncoupled
models assume independence between the material constitutive properties and damage, the parameter
calibration is simpler compared to that of the coupled models, and the most accurate state-of-the-art
constitutive models can be used in the uncoupled models.
Several uncoupled models have been presented in the literature. Kanvinde and Deierlein [20]
proposed the cyclic void growth model (CVGM) to predict ULCF life of structural steel based on the
Rice-Tracey void growth theory [21]. Owing to some advantages of the CVGM model, such as predicting
fracture initiation at the continuous level and being suitable for multiaxial stress condition compared
to empirical models, it has received extensive attention in attempting to predict the ULCF damage of
steel structural members. Myers et al. [22] and Fell et al. [23] investigated the ULCF fracture initiation
of column baseplate connections and steel frame brace components, respectively. Zhou et al. [24]
investigated the ULCF behaviour of beam-column connections, and Liao [25] conducted ULCF fracture
predictions for welded connection combined square steel pipe column and H-shaped steel beam.
In general, the above research results are encouraging, and it is promising to predict the ULCF fracture
initiation of steel structural members using the CVGM model. However, the results predicted by
the CVGM model largely depend on the calibration values of the model parameters, especially the
calibration value of the damage degradation parameter. In some literature [25,26], the calibration results
of damage degradation parameters of the CVGM model is rather discrete. Therefore, it is significant to
investigate the effect of model parameter calibration methods on the prediction accuracy of CVGM
model. Additionally, it has been suggested that the Lode angle parameter should also be accounted for
ULCF modelling except for stress triaxiality (T = σh /σe ), defined as the ratio of the hydrostatic stress
(σh ) and the Mises stress (σe ), especially in the case of low stress triaxiality (T < 0.33) [27–29]. However,
high stress triaxiality (T > 0.70) and moderate stress triaxiality (0.33 < T < 0.70) are employed in the
present study, and the effect of Load angle parameter can be neglected.
In order to improve the prediction accuracy of the CVGM model, tests were conducted on circular
notched specimens made of Q345qC steel commonly used in the construction of steel bridges in China.
A model parameter calibration method was proposed, and model parameters of the CVGM for Q345qC
steel were calibrated at both high and moderate stress triaxiality based on the experimental results and
finite element analysis (FEA). Comparisons were made between the experimental results and predicted
results to verified the effectiveness of proposed parameter calibration method. Finally, the effect of
damage degradation parameter on ULCF life prediction in steel bridge piers was discussed.

2. Cyclic Void Growth Model and Parameter Calibration

2.1. Cyclic Void Growth Model

Rice and Tracey studied the growth of spherical void in infinitely large ideal elastoplastic materials
and deduced the formula of void growth based on the stress triaxiality and equivalent plastic strain [21]:
p
dr/r = C · exp(1.5T )dεeq (1)

where r represents the instantaneous void radius, C indicates a material constant, T represents the
q
p p p
stress triaxiality, and dεeq = (2/3)dεij ·dεij represents the equivalent plastic strain increment.
Materials 2019, 12, 1615 3 of 17

For cyclic loading, the sign of the stress triaxiality, T, changes, and Equation (1) can be revised into
a more generalized form [20]:
p
dr/r = sign(T ) · C exp(|1.5T|)dεeq (2)

where sign(T) represents the sign of the stress triaxiality. It should be noted that if the stress triaxiality
is positive, the void will grow, and sign(T) = 1. Conversely, if stress triaxiality is negative, the void will
shrink, and sign(T) = −1.
The void–void interaction is not be considered here. By integrating Equation (2) over the tensile
and compressive excursions of loading, the void radius during cyclic loading can be expressed
as follows:
X Z ε2 X Z ε2
p p
ln(r/r0 )cyclic = C1 exp(|1.5T|)dεeq − C2 exp(|1.5T|)dεeq (3)
tensile ε1 compressive ε1

where ε1 and ε2 represent the equivalent plastic strains at the beginning and end of the tensile and
compressive excursions, respectively. Due to the lack of data to confirm the relative rates of void
growth and shrinkage, it is assumed that C = C1 = C2 . The void growth index, VGIcyclic , for cyclic
loading, representing cyclic void growth “demand”, is defined as follows [20]:
X Z ε2 X Z ε2
p
VGIcyclic = exp(|1.5T|)dεp − exp(|1.5T|)dεeq (4)
tensile ε1 compressive ε1

crit , under cyclic loading is determined by a degraded

The critical void growth “capacity”, VGIcyclic
function of its counterpart under monotonic loading, as described Kanvinde and Deierlein [20].
 
crit crit
VGIcyclic = VGImon exp −λεaccu
p (5)

where VGImoncrit represents the monotonic void growth “capacity” [30], λ indicates the material damage

degradation parameter under cyclic loading and is fitted according to Equation (6), and εaccu
p represents
a damage variable that is the cumulative equivalent plastic strain at the beginning of each tensile
cycle [20].  
crit crit
f = VGIcyclic /VGImon = exp −λεaccu
p (6)

where f represents the material damage ratio.

crit . To quantify the extent of ULCF
ULCF is considered to occur when VGIcyclic exceeds VGIcyclic
damage here, a damage index, D, has been defined as follows:



 D = max{D
 n−1 , Dth } 
(7)
 crit −VGI
 VGIcyclic cyclic
 Dth = 1 −


crit
VGImon

where n indicates the number of incremental steps during finite element calculation, and Dth represents
the value of the damage index calculated by the current incremental step. During the time-history
calculation, if Dth exceeds Dn−1 of the previous step, then D is updated to Dth , otherwise it remains
constant. When D reaches one, ULCF is considered to occur.
The ULCF fracture initiation is not the failure of a material point but involves a critical volume
of material. The characteristic length is defined to reflect the critical volume and can be determined
from the microscopic fracture surfaces of specimens by scanning electron microscopy. The proposed
characteristic length is commonly determined from two boundary values and a mean value [31].
The lower bound is twice the average diameter of the dimples, the upper bound is the maximum value
Materials 2019, 12, 1615 4 of 17
Materials 2019, 12, x FOR PEER REVIEW 4 of 17

of a plateau or trough, and the mean value is taken as an average of about ten plateaus or troughs, that
of a plateau or trough, and the mean value is taken as an average of about ten plateaus or troughs,
is, the most likely value of characteristic length.
that is, the most likely value of characteristic length.
2.2. Parameter Calibration of Q345qC Steel
2.2. Parameter Calibration of Q345qC Steel
2.2.1. Material Property
2.2.1. Material Property
Uniaxial tensile tests of three smooth round bar specimens were carried out using MTS 880
(MTSUniaxial
Systemstensile tests of Eden
Corporation, three smooth round
Prairie, MN, bar specimens
USA) were carried
to obtain mechanical out usingofMTS
properties 880 (MTS
Q345qC steel.
Systems Corporation, Eden Prairie, MN, USA) to obtain mechanical properties of Q345qC
The dimensions of the smooth round bar specimens are presented in Figure 1, and the gauge lengthsteel. The
of
dimensions of the smooth
the extensometer is 50 mm. round bar specimens are presented in Figure 1, and the gauge length of
the extensometer is 50 mm.

Figure 1. Dimensions
Figure 1. Dimensions of
of smooth
smooth round
round bar
bar specimens (unit: mm).
specimens (unit: mm).

The deformation in the extension gauge was uniform before necking, and the stress–strain
stress–strain curve
of the material was fitted according to Equation (8) [32].

σσ==KK
(ε(pε)pn )
n
(8)
(8)

where
where K K represents
represents the thestrain
strainhardening
hardening coefficient, n indicates
coefficient, n indicatesthe the
strain hardening
strain index,index,
hardening and
represents the plastic strain.
and represents the plastic strain.
When
When thethe specimen
specimenbegan
begantotoneck,
neck,the
thedeformation
deformation inin
thethe extension
extension gauge
gauge waswas concentrated
concentrated in
in the
the necking
necking region.
region. It assumed
It was was assumed that
that the the stress–strain
stress–strain relationship
relationship linearlylinearly increased
increased from thefrom
neckingthe
necking to the fracture. The true stress, σ f, and true strain, εf, of the specimen when fractured can be
to the fracture. The true stress, σf , and true strain, εf , of the specimen when fractured can be calculated
calculated
by Equation by(9).
Equation (9). The mechanical
The mechanical properties of properties of the
the material arematerial
providedarein provided in Table
Table 1. The 1. The
stress–strain
stress–strain curve of the material is presented in Figure 2, and the key parameters
curve of the material is presented in Figure 2, and the key parameters are listed in Table 2. are listed in Table
2.
pf
σ f = p2
πd /4
σ f = "2ff 2 # (9)
π d fd0/ 4
ε f = In d
f
 d 2
 (9)
Materials 2019, 12, x FOR PEER REVIEW  5 of 17
ε f = In  0  
 d f  
 

Table 1. Mechanical properties of Q345qC steel.

(ε f , σ f )
Method E (MPa) σy(MPa) σu (MPa) εf σf (MPa) A (%)
σ / Mpa

Mean 198,221 σ =351.10

K (ε p ) n
508.57 1.14 1104.57 40.60
Cov/% 0.83 1.16 1.51 0.75 1.74 3.57
(ε 2 , σ 2 )
Note: E indicates elastic modulus; σy and σu denote the yield strength and ultimate strength
σy
respectively; A represents elongation
(ε , σ )ratio.
1 1

εp /%

Figure 2.
Figure Calibrated true
2. Calibrated true stress–plastic-strain
stress–plastic-strain curve.
curve.

Table 2. Constitutive model parameters under tensile loading.

Method σy (MPa) ε1 σ1 (MPa) ε2 σ2(MPa) εf σf (MPa) K (Mpa) n

Mean 351.10 0.02 364.31 0.17 606.84 1.14 1104.57 906.80 0.22
Cov/% 1.16 3.14 0.39 2.52 0.87 0.75 1.74 0.58 1.20
 (ε f , σ f )
(ε f , σ f )

σ / Mpa
σ = K (ε p )n

σ / Mpa
σ = K (ε p )n
Materials 2019, 12, 1615 5 of 17
(ε 2 , σ 2 )
σy (ε 2 , σ 2 )
σ y (ε1.1 , σ
Table 1)
Mechanical properties of Q345qC steel.
(ε1 , σ 1 )
Method E (MPa) σ y (MPa) σ u (MPa) εf σ f (MPa) A (%)
Mean 198,221 351.10 εp /%
508.57 1.14 1104.57 40.60
Cov/% 0.83 1.16 εp /%
1.51 0.75 1.74 3.57
Figure 2.σyCalibrated
Note: E indicates elastic modulus; and σu denote
true the
stress–plastic-strain
yield strength and curve.
ultimate strength respectively; A
represents elongation ratio. Figure 2. Calibrated true stress–plastic-strain curve.
Table 2. Constitutive model parameters under tensile loading.
Table 2.
Table Constitutive model
2. Constitutive model parameters
parameters under
under tensile
tensile loading.
loading.
Method σy (MPa) ε1 σ1 (MPa) ε2 σ2(MPa) εf σf (MPa) K (Mpa) n
Method σ y (MPa) ε1 σ1 (MPa) ε2 σ2 (MPa) εf σ f (MPa) K (Mpa) n
Method
Mean σy351.10
(MPa) ε
0.02
1 σ1364.31
(MPa) ε
0.17
2 σ606.84
2(MPa) ε
1.14f σ1104.57
f (MPa) K906.80
(Mpa) n
0.22
Mean 351.10 0.02 364.31 0.17 606.84 1.14 1104.57 906.80 0.22
Mean
Cov/%
Cov/% 351.10
1.16
1.16 0.02
3.14
3.14 364.31
0.39
0.39 0.17
2.52
2.52 606.84
0.87
0.87 1.14
0.75
0.75 1104.57
1.74
1.74 906.80
0.58
0.58 0.22
1.20
1.20
Cov/% 1.16 3.14 0.39 2.52 0.87 0.75 1.74 0.58 1.20
2.2.2. Calibration of Monotonic Void Growth Growth Capacity
Capacity
2.2.2. Calibration of Monotonic Void Growth Capacity
Uniaxial tensile tests of circular notched specimens were carried out using MTS 880, as presented
Uniaxial
in Figure
Figure tensile
3.3.The
The tests ofof
dimensions
dimensions circular
of
thethe notchedare
specimens
specimens specimens
are were
presented
presented carried 4. out
in Figure
in Figure 4.using
Loading MTSapplied
Loading
was 880, applied
was aswith
presented
with
strain
in Figure
strain
control, 3. The
control,
and dimensions
theand gaugeoflength
the length
gauge ofthe
thespecimens are presented
of the extensometer
extensometer is 50 mm. in
is 50 Figure
mm.
Since the 4.
SinceLoading wastriaxiality
thetriaxiality
stress stress applied with
may
may change
strain
changecontrol,
significantly and the
significantly
during gaugeloading,
during
loading, alength
concept of the
a of extensometer
concept
average ofstress
average isstress
50 mm.
triaxiality Since
triaxiality
was the
was
introducedstress triaxiality
introduced
[33], may
[33],
as defined as
change
defined significantly
Equation as(10): during loading, a concept of
Equation (10): Z average stress triaxiality was introduced [33], as
defined as Equation (10): 1
Tm = 1 T (εp )dεp (10)
Tm =εF  T (ε p )dε p (10)
ε1F
εF 
where εF represents the fracture strain of notched T (ε p )dε pat the instant of crack initiation, and T(ε
Tm = specimens (10)
p)
represents the loading history of stress triaxiality obtained by FEA.

Figure 3. Test setup of notched specimen.

Figure 3.
Figure Test setup
3. Test setup of
of notched
notched specimen.
specimen.

Figure 4. Dimensions of notched specimens for tensile tests (unit: mm).

Figure 4. Dimensions of notched specimens for tensile tests (unit: mm).
Figure 4. Dimensions of notched specimens for tensile tests (unit: mm).
The axisymmetry of the specimen geometry and loading procedure allowed for the establishment
of a half axisymmetric two-dimensional finite element model of the specimen in ABAQUS 6.14,
as presented in Figure 5. The reduced integration element (CAX8R) was adopted, and the element size
in the notched area was approximately 0.20 mm in order to be consistent with the characteristic length
of Q345qC steel [34].
T(εp) represents
ABAQUS 6.14, asthe loadingin
presented history
Figureof5.stress triaxiality
The reduced obtained element
integration by FEA.(CAX8R) was adopted, and
The axisymmetry
the element of the area
size in the notched specimen geometry and
was approximately 0.20 loading procedure
mm in order allowed with
to be consistent for the
the
establishment of a half axisymmetric
characteristic length of Q345qC steel [34]. two-dimensional finite element model of the specimen in
ABAQUS 6.14, as presented in Figure 5. The reduced integration element (CAX8R) was adopted, and
the element size in the notched area was approximately 0.20 mm in order to be consistent with the
Materials 2019, 12, 1615 6 of 17
characteristic length of Q345qC steel [34].

Figure 5. Axisymmetric finite element model of notched tensile specimen (R = 3.75 mm).

FigureFigure
6 presents the comparison
5. Axisymmetric of force-displacement
finite element model of notched tensile curvesspecimen
obtained(Rfrom
= 3.75the tensile tests
mm).
Figure 5. Axisymmetric
and FEA, respectively. It can be finite
observed element
thatmodel of notched
test curves are tensile
in strong specimen (R = 3.75
agreement mm).
with FEA curves.
The sudden
Figure 6change
presents in the
the slope of the force-displacement
comparison of force-displacement curve corresponds
curves obtainedtofromthe instant
the tensileof crack
tests
Figurerespectively.
and FEA, 6 presents the comparison
It can be observed of force-displacement
that test Δ
curves are curves
in strong obtained
agreementfromwith
the tensile
FEA tests
curves.
f
initiation
and
The FEA,
sudden
[30], and its corresponding
respectively.
change in the It can be observed
slope
displacement,
that test curves
of the force-displacement
, is used
are inasstrong
curve
the control deformation
agreement
corresponds to thewith in the
FEA
instant
FEA
curves.
of crack
The sudden
initiation
to
change
[30],
calculate andmonitsin
VGI crit
T of the
the slope
. corresponding
The m and displacement,
force-displacement
calibration results
curve
of VGI
∆ f , is used as thecorresponds
crit
mon control
to the instant
at the deformation
center of net theofFEA
in sectioncrack
to
of
calculate VGI crit . The T and calibration results of Δ
VGI f crit at the center of net section
crit of specimens are
initiation [30],
specimens and
aremon its corresponding
presented
m
in Table 3. The displacement,
monotonic void , isgrowth
mon used ascapacity, VGIdeformation
the control mon = 2.03 , of in the FEA
Q345qC is
presented in Tablecrit3. The monotonic void growth capacity, VGI crit = 2.03, of Q345qC is smaller than
to VGI T
The (mVGI
crit
mon = 2.55 ) [35].results VGI mon
crit

thatcalculate
smaller than that
of Q345B (VGI moncrit. Q345B
of and calibration Thus, itofcan be mon at the center of net section of
deduced that Q345B has greater
mon = 2.55) [35]. Thus, it can be deduced that Q345B has greater fracture toughness
crit
fracture
than that ofare
specimens presented in Table 3. The monotonic void growth capacity, VGI mon = 2.03 , of Q345qC is
toughness
Q345qC. than that of Q345qC.
crit
smaller than that of Q345B ( VGI mon = 2.55 ) [35]. Thus, it can be deduced that Q345B has greater
fracture toughness than that of Q345qC.

Figure 6.
Figure Comparison of
6. Comparison of force-displacement
force-displacement obtained
obtained by
by tests
tests and
and FEA
FEA under
under tensile
tensile loading.

Table 3. Calibration results of monotonic void growth capacity.

Table 3. Calibration results of monotonic void growth capacity.
Figure 6. Comparison of force-displacement
Notch Size No.
obtained
∆f (mm) by tests
¯ and FEAVGI
under
crit tensile loading.
Tm
Δ f (mm) T
mon crit
Notch Size No. VGI mon
BM-1 1.16 1.23 m 2.05
Table1.80
3. CalibrationBM-2
results of monotonic
1.20
void
1.23
growth capacity.
2.10
BM-1 1.16 1.23 2.05
1.80
Notch Size BM-2
No.
3.75
BM-3
BM-4
Δ f 1.20
1.64
(mm)
1.80
0.89
0.92 T
1.23 1.92
m 2.10
2.10
VGI crit
mon

BM-1 BM-5 3.18

1.16 0.69 1.23 2.30 2.05
1.80 7.50
BM-6 3.00 0.68 2.15
BM-2 1.20 1.23 2.10
BM-7 4.35 0.54 1.98
30.0
BM-8 4.36 0.54 1.99
BM-9 5.05 0.49 1.78
60.0
BM-10 5.32 0.50 1.98
Mean value 2.03
Cov/% 6.56
 BM-7 4.35 0.54 1.98
30.0
BM-8 4.36 0.54 1.99
BM-9 5.05 0.49 1.78
60.0
BM-10 5.32 0.50 1.98
Materials 2019, 12, 1615 Mean value 2.03 7 of 17
Cov/% 6.56

2.2.3.
2.2.3.Calibration
CalibrationofofDamage
DamageDegradation
DegradationParameter
Parameter
Circular
Circularnotched
notchedspecimens
specimenswereweresubjected
subjectedtototwo twotypes
typesofofcyclic
cyclicloading
loadinghistories.
histories.TheThecontour
contour
ofofthe specimens is presented in Figure 7, and the dimensions are provided in Table
the specimens is presented in Figure 7, and the dimensions are provided in Table 4. In one loading 4. In one loading
history,
history,the
theapplied
applieddisplacements
displacementswere werecycled
cycledbetween
betweentwo twopredetermined
predeterminedvaluesvaluesuntil
untilfracture
fracture
occurred,
occurred,which
which is is referred
referred to asas the
thecycle
cycletotofailure
failure(CTF)
(CTF)loading.
loading.In In
thethe second
second loading
loading history,
history, five
five cycles were applied at specified loading amplitudes, then monotonic tensile
cycles were applied at specified loading amplitudes, then monotonic tensile pulling to fracture was pulling to fracture
was performed.
performed. ThisThis second
second loading
loading history
history is called
is called the cycle
the cycle andand
pullpull to failure
to failure (C-PTF)
(C-PTF) loading.
loading. The
The loading
loading waswas controlled
controlled byby strain,the
strain, thegauge
gaugelength
lengthof ofthe
the extensometer
extensometer of of specimens
specimens(R (R11== 60.00
60.00mm)mm)
isis50
50mm,
mm,and
andthat
thatofofthe
theother
otherspecimens
specimensisis12.5 12.5mm.
mm. Similarly,
Similarly,the
thefollowing
followingaverage
averagestress
stresstriaxiality
triaxiality
was
wasintroduced,
introduced,defined
defined asas
Equation
Equation (11)
(11)below.
below. TheTheloading
loadingprotocol andand
protocol average stress
average triaxiality
stress of
triaxiality
specimens for cyclic loading are provided in
of specimens for cyclic loading are provided in Table 4. Table 4.

1
Z 
 (Tε(pε)pdε
1 )dεp p
Tc =Tc =

T (11)
(11)
εF ε F

R2
R1

50 d3 50

Figure7.7.Contour
Figure Contourofofnotched
notchedspecimens
specimensfor
forcyclic
cyclictests
tests(unit:
(unit:mm).
mm).

Table 4. Dimensions and loading protocols of circular notched specimens.

Table 4. Dimensions and loading protocols of circular notched specimens.
R1 d1 d2 d3 R2 Cycles to ¯
No. Loading Loading Strain Cycles to Tc
(mm) (mm) (mm) (mm) (mm) Fracture Initiation
Loading
R1(mm) No.
BMC-1 d115.0
(mm) d2(mm)
7.50 80.0d3(mm) 12.5 R2(mm)
C-PTF Loading
[0, 0.020] 6 Fracture 0.44 Tc
Strain
BMC-2 15.0 7.50 80.0 12.5 C-PTF [0, 0.020] 6 Initiation0.45
BMC-3 15.0 7.50 80.0 12.5 CTF [0, 0.025] 24 0.53
BMC-1
BMC-4
15.0
15.0
7.50
7.50 80.0
80.0 12.5 12.5
CTF
C-PTF
[0, 0.025]
[0, 0.020] 20 6 0.54
0.44
60.00
BMC-2
BMC-5 15.0
15.0 7.50
7.50 80.0 80.0 12.5 12.5
CTF C-PTF
[0, 0.035] [0, 0.020] 10 6 0.51 0.45
BMC-6
BMC-3 15.0
15.0 7.50
7.50 80.0 80.0 12.5 CTF
12.5 [0, 0.035] [0, 0.025] 11
CTF 24 0.51 0.53
BMC-7 15.0 7.50 80.0 12.5 CTF [0, 0.040] 7 0.48
BMC-4
BMC-8 15.0
15.0 7.50
7.50 80.0 80.0 12.5 12.5
CTF CTF
[0, 0.040] [0, 0.025] 8 20 0.49 0.54
60.00
BMC-5
BMC-9 15.0
12.5 7.50 50.0 80.0 21.0 12.5
6.25 C-PTF CTF
[0, 0.020] [0, 0.035] 6 10 1.17 0.51
1.500
BMC-6
BMC-10 15.0
12.5 7.50 50.0 80.0 21.0 12.5
6.25 C-PTF CTF
[0, 0.020] [0, 0.035] 6 11 1.17 0.51
3.125
BMC-7
BMC-11 15.0
12.5 7.50 50.0 80.0 21.0 12.5
6.25 CTF CTF
[0, 0.035] [0, 0.040] 20 7 1.12 0.48
BMC-12 12.5 6.25
BMC-8 15.0 7.50 50.0 80.0 21.0 12.5 CTF [0, 0.035]
CTF [0, 0.040] 15 8 1.10
0.49
BMC-13
BMC-9 12.5
12.5 6.25
6.25 50.0 50.0 21.0 CTF
21.0 [0,
C-PTF 0.035] [0, 0.020] 25 6 0.99 1.17
4.500
BMC-14 12.5 6.25 50.0 21.0 CTF [0, 0.035] 24 0.98
1.500 BMC-
1.500
BMC-15 12.5
10.0 6.25 46.0 50.0 25.0 21.0
5.00 CTF C-PTF
[0, 0.045] [0, 0.020] 7 6 1.26 1.17
10
BMC-16 10.0 5.00 46.0 25.0 CTF [0, 0.048] 4 1.20
BMC-
BMC-17 10.0 5.00
3.500 12.5 6.25 46.0 50.0 25.0 21.0 CTF [0, 0.035]
CTF [0, 0.035] 11 20 0.98
1.12
11
BMC-18 10.0 5.00 46.0 25.0 CTF [0, 0.035] 9 0.97
3.125
BMC-
Note: For example, [0, 0.025] refers to the specimen cycled between strain 0 and 0.025.
12.5 6.25 50.0 21.0 CTF [0, 0.035] 15 1.10
12
The FEA of cyclic tests was carried out by ABAQUS 6.14, and the division of elements was similar
to that of the tensile loading analysis. The Lemaitre–Chaboche hybrid hardening model [36] was
used to simulate the cyclic plastic flow of the material, including isotropic and kinematic hardening.
The above calibrated true stress–plastic-strain curve was used to simulate the kinematic behaviour,
and the isotropic hardening is described by Equation (12).
h  i
σ0 = σ0 + Q∞ 1 − exp −bεp (12)
Materials 2019, 12, 1615 8 of 17
Materials 2019, 12, x FOR PEER REVIEW 8 of 17

where σ0 represents BMC- the size of the initial yielding surface;Q∞ indicates the maximum change value of
12.5 6.25 50.0 21.0 CTF [0, 0.035] 25 0.99
yielding surface; 13and b denotes the changing rate of yielding surface size as plastic strain develops.
4.500
BMC-
The parameters Q∞ and 12.5 b were determined
6.25 50.0from a21.0trial procedure
CTF based
[0, 0.035]on the24 best fit between
0.98 the
14
test curves and BMC- FEA curves.
10.0 5.00 46.0 25.0 CTF [0, 0.045] 7 1.26
Similar to tensile
15 loading, the sudden change in the slope of the last tensile cycle is the instant of
1.500
crack initiation BMC-[20]. Figure10.08 presents
5.00 the 46.0
force-displacement
25.0 curves[0,of
CTF partial specimens
0.048] 4 obtained
1.20 by
tests and FEA. It 16 can be observed that the test curves agree with the FEA curves. The other specimens
BMC-
have similar results. 10.0 5.00 46.0 25.0 CTF [0, 0.035] 11 0.98
17
3.500 crit
The VGIcyclic BMC-at the center of the section under cyclic loading was calculated, obtaining the material
10.0 5.00 46.0 25.0 CTF [0, 0.035] 9 0.97
18
damage ratio, f. The size of voids at the critical instant under cyclic loading was smaller than that of
monotonic loading, Note: For example,
thus f was[0,no0.025]
morerefers
thanto the
onespecimen cycled
in theory between
[31]. Thestrain
damage0 andratio
0.025. of BMC-15 and

BMC-17 was The greater than one

FEA of cyclic testsand
wasthus were
carried not
out byused
ABAQUSto calibrate
6.14, andthethedamage degradation
division of elementsparameter.
was
similar for
The reason to that
thisofmay
the tensile loadingFirst,
be twofold. analysis. The Lemaitre–Chaboche
although the constitutive model hybrid used
hardening
in themodel
FEA has[36] been
was used
calibrated, to simulate
there remains the cyclic difference
a certain plastic flowcompared
of the material,
to theincluding isotropic and
actual constitutive kinematicof the
behaviour
hardening.
material, Thedifference
and this above calibrated
may causetrue stress–plastic-strain curve was used
the error of the stress–strain to simulate
analysis. Second,the the
kinematic
calibration
behaviour, and the isotropic hardening is described by Equation (12).
of VGImon also lead to an error in calculating the damage ratio. The f and εp were fitted according
crit accu

to the exponential function in Equation σ 0(6) Q∞ 1 −the

= σto0 +obtain ( −bε p ) damage degradation parameter,
exp material (12) λ.
The experimental data scatter plot and the fitted curve are presented in Figure 9.
As shown
where
σ 0 represents
in Figure the9, the
sizecoefficient ofyielding
of the initial determination Q
surface;of∞the fitted curve
indicates is low. This
the maximum may
change result in
value
poorofULCF prediction. In order to increase the goodness of fit, the damage degradation
yielding surface; and b denotes the changing rate of yielding surface size as plastic strain develops. parameter
was separately calibrated Q at high and moderate stress triaxialities, as shown in Figure 10, and it can
The parameters ∞ and b were determined from a trial procedure based on the best fit between the
be observed that the experimental data and fitted curves agree well. It can be seen that the damage
test curves and FEA curves.
degradation parameter at moderate stress triaxiality is larger than that at high stress triaxiality. This may
Similar to tensile loading, the sudden change in the slope of the last tensile cycle is the instant of
be due to the dominant factor in the failure process. Recent experimental and computational studies
crack initiation [20]. Figure 8 presents the force-displacement curves of partial specimens obtained
haveby shown
tests that micro-void
and FEA. It can bedilation is dominant
observed that the testat high stress
curves agree triaxiality,
with the and FEA micro-void
curves. The elongation
other
gradually dominates
specimens the failure
have similar results.process of ULCF damage as stress triaxiality decreases [37–39].

(a) (b)

Figure 8. Cont.
Materials 2019, 12, 1615 9 of 17
Materials 2019, 12, x FOR PEER REVIEW 9 of 17

(c) (d)

Force(kN)

(e) (f)

(g) (h)

(i)

Figure 8. Comparison of force-displacement of specimens obtained by tests and FEA: (a) BMC-1;
(b) BMC-3; (c) BMC-5; (d) BMC-7; (e) BMC-9; (f) BMC-11; (g) BMC-13; (h) BMC-15; (i) BMC-17.
has been calibrated, there remains a certain difference compared to the actual constitutive behaviour
of the material, and this difference may cause the error of the stress–strain analysis. Second, the
crit ε accu
calibration of VGI mon also lead to an error in calculating the damage ratio. The f and p were fitted
according to the exponential function in Equation (6) to obtain the material damage degradation
Materials 2019, 12, 1615 10 of 17
parameter, λ. The experimental data scatter plot and the fitted curve are presented in Figure 9.

1.2
Experimental data
Fitted curve
1.0

f = exp(−0.12.ε paccumulated )
0.8
R 2 = 0.611
0.6

f 0.4

0.2

0.0
0 1 2 3 4 5 6 7 8 9 10
ε accumulated
p

Materials 2019, 12, x FOR PEER REVIEW 11 of 17

Figure 9. Scatter plot and fitted curve of damage degradation parameter for Q345qC.
Figure 9. Scatter plot and fitted curve of damage degradation parameter for Q345qC.

As shown in Figure 9, the coefficient of determination of the fitted curve is low. This may result
in poor ULCF prediction. In order to increase the goodness of fit, the damage degradation parameter
was separately calibrated at high and moderate stress triaxialities, as shown in Figure 10, and it can
be observed that the experimental data and fitted curves agree well. It can be seen that the damage
f = exp(−0.08ε paccumulated )
degradation parameter at moderate stress triaxiality is larger than that at high stress triaxiality. This
R 2 = 0.773
f

may be due to the dominant factor in the failure process. Recent experimental and computational
studies have shown that micro-void dilation is dominant at high stress triaxiality, and micro-void
elongation gradually dominates the failure process of ULCF damage as stress triaxiality decreases
[37–39]. R 2 = 0.779
f = exp(−0.18ε paccumulated )

ε paccumulated

Figure10.
Figure Calibrationof
10.Calibration ofdamage
damagedegradation
degradationparameter
parameterat
athigh
highand
andmoderate
moderatestress
stress triaxiality.
triaxiality.

3. ULCF Fracture Initiation Prediction

3. ULCF Fracture Initiation Prediction
In order to examine the accuracy of the CVGM model at high and moderate stress triaxiality,
In order to examine the accuracy of the CVGM model at high and moderate stress triaxiality,
cyclic tests were conducted on circular notched specimens providing both high and moderate stress
cyclic tests were conducted on circular notched specimens providing both high and moderate stress
triaxiality. The contour and dimensions of specimens are presented in Figure 7 and Table 5, respectively.
triaxiality. The contour and dimensions of specimens are presented in Figure 7 and Table 5,
Loading was controlled by strain with CTF loading, the gauge length of the extensometer was 50 mm,
respectively. Loading was controlled by strain with CTF loading, the gauge length of the
and the loading strain is provided in Table 5.
extensometer was 50 mm, and the loading strain is provided in Table 5.
Table 5. Dimensions and experimental results of circular notched specimens.
Table 5. Dimensions and experimental results of circular notched specimens.
¯
R1 (mm) No. d1 (mm) d2 (mm) d3 (mm) R2 (mm) Cycles to Fracture Initiation Loading Strain Tc
Cycles to Fracture Loading
R1 (mm) No.
BMC-19 d115.0
(mm) 7.50 d2 (mm)50.0 d3 (mm) 12.5 R2 (mm) 9 [0, 0.040] 0.85 T
7.50 Initiation [0, 0.040]Strain 0.85 c
BMC-20 15.0 7.50 50.0 12.5 9
BMC-19
BMC-21 15.0
15.0 7.50 7.50 50.0 50.012.5 12.5 9 9 [0, 0.040]0.77 0.85
[0, 0.040]
7.50 BMC-22 15.0 7.50 50.0 12.5 13 [0, 0.040] 0.80
10.0 BMC-20 15.0 7.50 50.0 12.5 9 [0, 0.040]0.77 0.85
BMC-23 15.0 7.50 50.0 12.5 7 [0, 0.045]
BMC-21
BMC-24 15.0
15.0 7.50 7.50 50.0 50.012.5 12.5 8 9 [0, [0, 0.040]0.76 0.77
0.045]
BMC-25
BMC-22 15.0
15.0 7.50 7.50 50.0 50.012.5 12.5 21 13 [0, 0.035]
[0, 0.040]0.72 0.80
10.0 15.0 BMC-26 15.0 7.50 50.0 12.5 25 [0, 0.035] 0.73
BMC-23
BMC-27 15.0
15.0 7.50 7.50 50.0 50.012.5 12.5 13 7 [0, 0.045]0.71 0.77
[0, 0.045]
BMC-28 15.0 7.50
BMC-24 15.0 7.50 50.0 50.012.5 12.5 15
8 [0, 0.045]
[0, 0.045]0.72 0.76
BMC-29 15.0 7.50 80.0 12.5 13 [0, 0.030] 0.52
BMC-25
BMC-30 15.0
15.0 7.50 7.50 80.0 50.012.5 12.5 13 21 [0, 0.035]0.52 0.72
[0, 0.030]
60.0
BMC-31
BMC-26 15.0
15.0 7.50 7.50 80.0 50.012.5 12.5 8 25 [0, 0.040]
[0, 0.035]0.50 0.73
15.0 BMC-32 15.0 7.50 80.0 12.5 8 [0, 0.040] 0.49
BMC-27 15.0 7.50 50.0 12.5 13 [0, 0.045] 0.71
Note: For example, [0, 0.040] refers to a specimen cycled between strain 0 and 0.040 until ULCF failure was observed.
BMC-28 15.0 7.50 50.0 12.5 15 [0, 0.045] 0.72
BMC-29 15.0 7.50 80.0 12.5 13 [0, 0.030] 0.52
BMC-30 15.0 7.50 80.0 12.5 13 [0, 0.030] 0.52
60.0
BMC-31 15.0 7.50 80.0 12.5 8 [0, 0.040] 0.50
BMC-32 15.0 7.50 80.0 12.5 8 [0, 0.040] 0.49
Note: For example, [0, 0.040] refers to a specimen cycled between strain 0 and 0.040 until ULCF failure
was observed.
Materials 2019, 12, 1615 11 of 17

The two-dimensional axisymmetric finite element model was established using ABAQUS 6.14
with the element division similar to that of the aforementioned tensile loading. The Lemaitre-Chaboche
hybrid hardening model [36] was used to simulate the cyclic plastic flow of material. Figure 11 presents
an example of the force-displacement plot for a specimen, and it can be seen that the test curve is in
good agreement with the FEA curve. The development of the damage index, D, of partial specimens at
the fracture
Materials initiation
2019, 12, location
x FOR PEER REVIEWis presented in Figure 12. ULCF fracture is predicted to occur12
inofthe
17
Materials
notched2019, 12, x FOR when
specimens PEER REVIEW
D reaches one. 12 of 17

Figure
Figure11.
11.Comparison
Comparisonof
offorce-displacement
force-displacementcurve
curveobtained
obtainedby
by tests
tests and
and FEA
FEA under
under cyclic
cyclic loading.
loading.
Figure 11. Comparison of force-displacement curve obtained by tests and FEA under cyclic loading.
1.2
1.2 Fracture Initiation
Fracture Initiation
1.0
1.0

0.8
0.8
0.6
0.6
0.4
0.4

0.2 D (Initially calibrated CVGM)

DD (Segmentally calibrated
(Initially calibrated CVGM)CVGM)
0.2
D (Segmentally calibrated CVGM)
0.0
0.00 2 4 6 8 10 12 14 16
0 2 4 6 Cycle
8 10 12 14 16
Cycle
(a) (b)
(a) (b)

(c) (d)
(c) (d)
Figure
Figure 12.
12. Evolution
Evolutionof ofdamage
damageindex
indexFigure:
Figure: (a)
(a)BMC-19;
BMC-19;(b)
(b)BMC-28;
BMC-28;(c)
(c)BMC-19;
BMC-19;(d)
(d)BMC-28.
BMC-28.
Figure 12. Evolution of damage index Figure: (a) BMC-19; (b) BMC-28; (c) BMC-19; (d) BMC-28.
Figure 13 presents the comparison of the experimental results and predicted results, including
Figure 13 presents the comparison of the experimental results and predicted results, including
ε
the number of cycles, Nf , and equivalent plastic strain, ε pp , to fracture initiation. To evaluate the
prediction accuracy of the CVGM model, relative error, γγ , is calculated according to Equation (13),
the number of cycles, Nf , and equivalent plastic strain, , to fracture initiation. To evaluate the
prediction
and accuracy
the results of the CVGM
are presented model,
in Table relative
6. The error,
results , is calculated
indicate according
that the predicted lifetoand
Equation (13),
equivalent
and the results are presented in Table 6. The results indicate that the predicted life
plastic strain by a segmentally calibrated CVGM model is nearer to the experimental results and equivalent
plastic strain by a segmentally calibrated CVGM model is nearer to the experimental results
Materials 2019, 12, 1615 12 of 17

Figure 13 presents the comparison of the experimental results and predicted results, including the
number of cycles, Nf , and equivalent plastic strain, εp , to fracture initiation. To evaluate the prediction
accuracy of the CVGM model, relative error, γ, is calculated according to Equation (13), and the results
are presented in Table 6. The results indicate that the predicted life and equivalent plastic strain by a
segmentally calibrated CVGM model is nearer to the experimental results compared to the originally
calibrated CVGM model. For a small part of specimens, the errors are larger than or close to 25%
between experimental results and predicted results by segmentally calibrated CVGM model, which
Materials 2019, 12, x FOR PEER REVIEW 13 of 17
may be related to some assumptions introduced in the model, but the error is acceptable. In general,
the mean value of γ calculated by the segmentally calibrated CVGM model is approximately 50%
CVGM model is approximately 50% smaller than that of the initially calibrated CVGM model.
smaller than that of the initially calibrated CVGM model. Therefore, separately calibrating the CVGM
Therefore, separately calibrating the CVGM model at high and moderate stress triaxiality greatly
model at high and moderate stress triaxiality greatly improved the prediction accuracy of ULCF.
improved the prediction accuracy of ULCF.
  analytical exp erimental 

 εεpanalytical
p −εp exp erimental
−εp


 exp erimental
εpexp erimental

γ= ε p −Nexp erimental 

 Nanalytical (13)
 
γ =
  f f 
(13)
 N analytical  − N exp erimental

 exp erimental
f Nf f

 N exp
f
erimental

±25% ±25%
N analytical

ε panalytical
f

N exp
f
erimental
ε experimental
p

(a) (b)

Figure 13. Comparisons

Figure 13. Comparisonsofof experimental
experimental andand
FEAFEA results:
results: (a) Number
(a) Number of cyclesoftocycles to(b)
fracture; fracture; (b)
Equivalent
Equivalent plastic
plastic strain up tostrain up to fracture.
fracture.

Table 6. Relative error of prediction results by CVGM model.

Table 6. Relative error of prediction results by CVGM model.
Relative Error
Relative Error
Life Equivalent Plastic Strain
Specimen
Initially Calibrated LifeSegmentally Calibrated Equivalent Plastic
Initially Calibrated Strain
Segmentally Calibrated
Specimen CVGM Model CVGM Model
Segmentally CVGM Model
Initially CVGM Model
Segmentally
BMC-19 Initially Calibrated
0.222 0.111 0.272 0.102
BMC-20 0.222Model
Calibrated
0.111
CVGM Calibrated
0.291
CVGM Calibrated
0.125
CVGM
CVGM
BMC-21 0.000 Model
0.111 Model
0.031 Model
0.174
BMC-22 0.308 0.231 0.339 0.247
BMC-19
BMC-23 0.222
0.000 0.111
0.143 0.272
0.021 0.102
0.199
BMC-24
BMC-20 0.125
0.222 0.000
0.111 0.135
0.291 0.016
0.125
BMC-25 0.238 0.095 0.264 0.103
BMC-21
BMC-26 0.000
0.360 0.111
0.240 0.031
0.390 0.174
0.258
BMC-27
BMC-22 0.231
0.308 0.077
0.231 0.253
0.339 0.076
0.247
BMC-28 0.333 0.200 0.367 0.216
BMC-23
BMC-29 0.000
0.154 0.143
0.000 0.021
0.219 0.199
0.010
BMC-30 0.154 0.000 0.212 0.004
BMC-24
BMC-31
0.125
0.125
0.000
0.000
0.135
0.258
0.016
0.067
BMC-25
BMC-32 0.238
0.125 0.095
0.000 0.264
0.180 0.103
0.000
Mean value 0.186 0.094 0.231 0.114
BMC-26 0.360 0.240 0.390 0.258
BMC-27 0.231 0.077 0.253 0.076
BMC-28 0.333 0.200 0.367 0.216
BMC-29 0.154 0.000 0.219 0.010
BMC-30 0.154 0.000 0.212 0.004
BMC-31 0.125 0.000 0.258 0.067
BMC-32 0.125 0.000 0.180 0.000
Mean value 0.186 0.094 0.231 0.114

4. Effect of Damage Degradation Parameter

Materials 2019, 12, 1615 13 of 17

Materials 2019, 12, x FOR PEER REVIEW 14 of 17

4. Effect of Damage Degradation Parameter

widthTheof the different

flange,values of the damage
W represents the width degradation
of the web, parameter
t represents for the
Q345qC steel of
thickness were
the obtained
flange and at
different stress triaxiality ranges. To investigate the effect of the damage degradation
the web, bs represents the width of the stiffener, ts represents the thickness of the stiffener, td represents parameter on
predicting
the 2019,ULCF
thickness
Materials 12,of fracture
the
x FOR initiation
diaphragm
PEER REVIEW and of steel structural
baseplate, members,the
and a represents thespacing
single-column steel bridge
of the diaphragms. pier
14 The
of 17
was applied
geometric as the research
dimensions of theobject.
steelInbridge
this study,
pier two are fillet
listedwelds were 7.
in Table used
Thewith
axialgroove anglesP,ofand
pressure, 45◦
width
and 56of ◦ , as
the flange, Winrepresents the
In width of14,
the web,
horizontal presented
forced Figure δ,
displacement, 14.were Figure
applied aththe topt of
representsrepresents
thethe the of
height
pier, and thickness
thethe pier,
axial ofBthe flangeratio
represents
compression and
the
the
width
is web,
0.15.ofTheb represents
flange, W
theloading
s the width
represents
pattern of the stiffener,
the width
is shown t
of the15,
in Figure s represents
web,
where the
t representsthickness
δy is the the of the
thickness
horizontal stiffener,
of the
yield t represents
flange and the
displacement
d of
the
web,
the thickness
bs represents
pier. of thethediaphragm
width of and baseplate,
the stiffener, ts and a represents
represents the spacing
the thickness of theofstiffener,
the diaphragms.
td representsThe
geometric
the thickness dimensions of the steel
of the diaphragm and bridge
baseplate, pierand area listed
representsin Table 7. The of
the spacing axial pressure, P, and
the diaphragms. The
horizontal forced displacement,
geometric dimensions steelδ,
Table
of the were
7.bridge applied
Geometric are at
pier dimensions the top
listed of the
inofTable
steel 7.pier,
bridge and
axialthe
Thepiers. axial compression
pressure, P, and horizontalratio
is 0.15. displacement,
forced The loading pattern δ, wereis applied
shown in at Figure
the top15, of where
the pier, δy and
is thethehorizontal yield displacement
axial compression ratio is 0.15. of
h/mm B/mm W/mm t/mm a/mm bs/mm ts/mm td/mm
The loading pattern is shown in Figure 15, where δy is the horizontal yield displacement of the pier.
the pier.
2.5 825 825 20 412 81 10 15
Table 7. Geometric dimensions of steel bridge piers.

h/mm B/mm W/mm t/mm a/mm bs/mm ts/mm td/mm

2.5 825 825 20 412 81 10 15

Figure 14. Schematic diagram of single-column steel bridge pier.

Figure 14. Schematic diagram of single-column steel bridge pier.
Table 7. Geometric dimensions of steel bridge piers.

h/mm B/mm W/mm t/mm a/mm bs /mm ts /mm td /mm

2.5 825 14. Schematic
Figure 825 20 of single-column
diagram 412 81 bridge pier.
steel 10 15
δ /δy
δ /δy

Figure 15. Loading pattern.

The numerical analyses were conducted using ABAQUS 6.14. Three-dimensional finite element
models of the steel bridge pier were established, as shown in Figure 16. The beam element (B31) was
employed to simulate the upper partFigure Figure 15.
of the15. Loading
steel bridge
Loading pattern.
pier. The lower part of the pier, Ld, was
pattern.
simulated by the shell element (S4R) where the length of Ld was determined according to the
The numerical
The numerical analyses
analyses were
were conducted
conducted using ABAQUSABAQUS 6.14. Three-dimensional
Three-dimensional finite element
element
empirical formula for calculating the damagedusing domain [40]. To6.14. obtain the stress–strainfinite
history at the
models
models ofof the
of the steel
the boundary bridge
steel bridge pier were
pier were established, as shown in Figure 16. The beam element (B31)
bottom between the established,
flange and web, as shown in Figure
the solid element 16.(C3D8R)
The beam waselement (B31)
used. The was
MPC-
was employed
employed to simulate
to simulate the
the upper upper part
part of of the steel bridge pier. The lower part of the pier, Ld ,
beam connection was adopted between thethe
beamsteel bridge
and pier. TheThe
shell element. lower
shellpart
andof the element
solid pier, Ld,were
was
was simulated
simulated by shell
the shell element (S4R) where thethe length ofLdLdwas
wasdetermined
determinedaccording
accordingto to the
the
coupled byby the
Shell-Solid. element
The most (S4R) where
dangerous element length
in theofpoint of connection between the flange
empirical
and formula
web was for calculating
considered to be thethe damaged element,
calculating domain [40]. and To
theobtain
size ofthe
thestress–strain
calculating history
elementatwas
the
bottom
0.20 mm,ofwhich
the boundary between
is consistent withthe
theflange and web,length
characteristic the solid element (C3D8R)
of Q345qC [34]. The was used. The MPC-
Lemaitre–Chaboche
beam connection was adopted between the beam and shell element. The shell and solid element were
coupled by Shell-Solid. The most dangerous element in the point of connection between the flange
and web was considered to be the calculating element, and the size of the calculating element was
Materials
Materials 2019, 12, 1615
2019, 12, x FOR PEER REVIEW 15
14 of
of 17
17

hybrid hardening model was used to simulate the cyclic plastic flow of material, and the model
empirical formula
parameters used arefor calculating
provided the damaged
in Table 8 [34]. domain [40]. To obtain the stress–strain history at
the bottom of the boundary between the flange and web, the solid element (C3D8R) was used. The
MPC-beamTable 8. Lemaitre–Chaboche
connection hybrid hardening
was adopted between the beammodel parameters
and shell element. of Q345qC
The shell steel
and[34].
solid element
were coupled by Shell-Solid.
σ|0 The most
Q∞ dangerous element
C1 in the point of
C2 connection between
C3 the flange
Material
and web was considered to be the b element, andγ1
calculating the size of the γ2
calculating elementγ3 was
(MPa) (MPa) (MPa) (MPa) (MPa)
0.20 mm,
Base which
metal is consistent
351.10 with13.2
the characteristic
0.60 length 523.8
44373.7 of Q345qC [34]. The
9346.6 120.2Lemaitre–Chaboche
946.1 18.7
hybrid hardening
Weld deposit metal model was used
428.45 17.4to simulate
0.40 the cyclic 160.0
12752.3 plastic flow
1111.2of material,
160.0 and the model
630.5 26.0
parameters used areNote:
provided in Tableof8parameters
The meaning [34]. can be found in the literature [34].

Figure 16. Analytical model of steel bridge piers.

Figure 16. Analytical model of steel bridge piers.
Table 8. Lemaitre–Chaboche hybrid hardening model parameters of Q345qC steel [34].
The ULCF life of steel bridge piers with different groove angles was calculated using the above
Material σ|0 (MPa) Q∞ (MPa) b C1 (MPa) γ1 C2 (MPa) γ2 C3 (MPa) γ3
calibrated CVGM model, and the calculation results are shown in Table 9. The predicted ULCF life
Base metal 351.10 13.2 0.60 44373.7 523.8 9346.6 120.2 946.1 18.7
by segmentally
Weld deposit metalcalibrated
428.45 CVGM17.4 model varied
0.40 by12752.3
a maximum 160.0 of 37.5%
1111.2 as compared
160.0 to the predicted
630.5 26.0

life by initially calibrated Note:CVGM model.

The meaning Therefore,
of parameters based
can be foundon theliterature
in the calculation
[34]. results, the calibrated
damage degradation parameter has a significant influence on the ULCF fracture prediction, and it is
The ULCF
important life of steelcalibrate
to accurately bridge piers the with
damagedifferent groove angles
degradation was calculated
parameters under using the above
different stress
calibrated ranges.
triaxiality CVGM Additionally,
model, and the calculation
due to greaterresults
strain are shown in Table
concentration, 9. The
the steel predicted
bridge ULCF
pier with life by
a groove
segmentally
angle of 56° is calibrated
prone toCVGM model varied
ULCF damage compared by a maximum
with the steelof 37.5% as compared
pier with a groovetoanglethe predicted
of 45°. life
by initially calibrated CVGM model. Therefore, based on the calculation results, the calibrated damage
degradation parameterTable has a 9.significant
Calculation influence
results ofon the ULCF
ULCF fracture
life of steel bridge prediction,
piers. and it is important
to accurately calibrate the damage degradation parameters under different stress triaxiality ranges.
f 1 − N f angle
Predicted Cycles to ULCF Fracture Initiation
Additionally, due to greater strain concentration, the steel bridge pier with N a groove 2 of 56◦ is
Groove Angle
prone to ULCF damage compared ◦ Tc
Nf1 with the steel pier withNaf2 groove angle of 45 .N
f1
45° 45
Table 9. Calculation results of ULCF life29 of steel bridge piers. 35.5% 0.68
56° 8 11 37.5% 0.94
Note: Nf1 indicates thePredicted Cyclesby
cycles to ULCF toinitially
ULCF Fracture Initiation
calibrated CVGM model; N|Nf2f1indicates
−Nf2 | the cycles
¯ to
Groove Angle Nf1 Tc
ULCF by segmentally calibrated N CVGM model.
f1 N f2

45◦ 45 29 35.5% 0.68

5. Conclusions
56◦ 8 11 37.5% 0.94
In theNresearch
Note: f1 indicatespresented
the cycles to in thisbypaper,
ULCF initiallyQ345qC
calibratedsteel
CVGM was selected
model; for study
Nf2 indicates as ittoisULCF
the cycles commonly
by
segmentally calibrated CVGM model.
used in the construction of steel bridges in China. A model parameter calibration method that
separately calibrates the damage degradation parameter at high and moderate stress triaxiality was
5. Conclusions
proposed. The validity of the CVGM model calibrated by the proposed method was verified based
on the tests
In the and FEA.
research The effect
presented in thisofpaper,
the damage
Q345qC degradation parameter
steel was selected for studyon as
predicting
it is commonlythe ULCF
used
fracture initiation of steel
in the construction steel bridges
bridge in piers wasA investigated.
China. model parameter Based on the conducted
calibration method that studies, the
separately
following conclusions can be drawn:
calibrates the damage degradation parameter at high and moderate stress triaxiality was proposed.
The validity of the CVGM model calibrated by the proposed method was verified based on the tests
(1) The goodness of fit is improved when the damage degradation parameter is calibrated
separately at high and moderate stress triaxiality. It is shown that the predicted number of cycles
Materials 2019, 12, 1615 15 of 17

and FEA. The effect of the damage degradation parameter on predicting the ULCF fracture initiation
of steel bridge piers was investigated. Based on the conducted studies, the following conclusions can
be drawn:

(1) The goodness of fit is improved when the damage degradation parameter is calibrated separately
at high and moderate stress triaxiality. It is shown that the predicted number of cycles and
equivalent plastic strain to fracture by the segmentally calibrated CVGM model agree well with
the experimental results. The proposed parameter calibration method greatly improved the
predictive accuracy of the CVGM model compared to the previous method.
(2) From the numerical simulation of steel bridge piers, the value of the damage degradation
parameter has a relatively significant influence on the predicted cycles to ULCF fracture initiation.
Therefore, it is significant to separately calibrate the damage degradation parameters under
different stress triaxiality ranges.
(3) The different values of the damage degradation parameter were obtained at high and moderate
stress triaxiality, and it was determined that the value of the damage degradation parameter
depends on stress triaxiality. Therefore, it is clear that the relationship between damage
degradation parameter and stress triaxiality deserves further study in the future.

In future work, additional experimental data are needed to determine the relationship between
damage degradation parameter and stress triaxiality. Additionally, the different values of damage
degradation parameter at different levels of stress triaxiality may be related to the fracture mechanism,
and as such it requires further research. In general, the proposed parameter calibration method that
separately calibrates parameters under different stress triaxiality ranges is promising.

Author Contributions: Conceptualization, S.L. and X.X.; Data curation, Y.L.; Funding acquisition, X.X.;
Investigation, S.L., X.X. and Y.L.; Methodology, S.L. and X.X.; Software, Y.L.; Supervision, X.X.; Validation,
X.X.; Writing: original draft, S.L.; Writing: review & editing, S.L.
Funding: This research was funded by Natural Science Foundation of China (No. 51878606).
Acknowledgments: The authors would like to thank the Linjun Xie and Xiaojing Cai for their experimental
support and papergoing for checking the spelling of the manuscript.
Conflicts of Interest: The authors declare no conflict of interest. The funders had no role in the design of the study.

References
1. Miller, D.K. Lessons learned from the Northridge earthquake. Eng. Struct. 1998, 20, 249–260. [CrossRef]
2. Nakashima, M.; Inoue, K.; Tada, M. Classification of damage to steel buildings observed in the 1995
Hyogoken-Nanbu earthquake. Eng. Struct. 1998, 20, 271–281. [CrossRef]
3. Subcommittee on Investigation of Seismic Damage of Steel Structure. Investigation of causes of damage to
steel structure on Hanshin-Awaji earthquake disaster. Proc. JSCE 2000, 64, 17–30. (In Japanese)
4. Kuwamura, H.; Yamamoto, K. Ductile Crack as Trigger of Brittle Fracture in Steel. J. Struct. Eng. 1997, 123,
729–735. [CrossRef]
5. Tateishi, K.; Chen, T.; Hanji, T. Extremely low cycle fatigue assessment method for un-stiffened cantilever
steel columns. Doboku Gakkai Ronbunshuu A 2008, 64, 288–296. [CrossRef]
6. Ge, H.; Kang, L.; Tsumura, Y. Extremely Low-Cycle Fatigue Tests of Thick-Walled Steel Bridge Piers.
J. Bridg. Eng. 2013, 18, 858–870. [CrossRef]
7. Ge, H.B.; Ohashi, M.; Tajima, R. Experimental study on ductile crack initiation and its propagation in steel
bridge piers of thick-walled box sections. J. Struct. Eng. JSCE 2007, 53, 493–502. (In Japanese)
8. Coffin, L.F.J. A study of the effects of cyclic thermal stresses on a ductile metal. Trans. ASME 1954, 76,
931–950.
9. Manson, S.S. Behavior of Materials under Conditions of Thermal Stress; National Advisory Commission on
Aeronautics Report 1170; Lewis Flight Propulsion Laboratory: Cleveland, OH, USA, 1954.
10. Ge, H.; Kang, L. A Damage Index-Based Evaluation Method for Predicting the Ductile Crack Initiation in
Steel Structures. J. Earthq. Eng. 2012, 16, 623–643. [CrossRef]
Materials 2019, 12, 1615 16 of 17

11. Ge, H.B.; Luo, X.Q. A seismic performance evaluation method for steel structures against local buckling and
extra-low cycle fatigue. J. Earthq. Tsunami 2011, 5, 83–99. [CrossRef]
12. Kang, L.; Ge, H.B. Predicting ductile crack initiation of steel bridge structures due to extremely low-cycle
fatigue using local and non-local models. J Earthq Tsunami 2013, 17, 323–349. [CrossRef]
13. Miner, M.A. Cumulative damage in fatigue. J. Appl. Mech. 1945, 67, 159–164.
14. Tateishi, K.; Hanji, T.; Minami, K. A prediction model for extremely low cycle fatigue strength of structural
steel. Int. J. Fatigue 2007, 29, 887–896. [CrossRef]
15. Xue, L. A unified expression for low cycle fatigue and extremely low cycle fatigue and its implication for
monotonic loading. Int. J. Fatigue 2008, 30, 1691–1698. [CrossRef]
16. Pereira, J.; De Jesus, A.M.; Xavier, J.; Fernandes, A.; Fernandes, A. Ultra low-cycle fatigue behaviour of a
structural steel. Eng. Struct. 2014, 60, 214–222. [CrossRef]
17. Mear, M.; Hutchinson, J. Influence of yield surface curvature on flow localization in dilatant plasticity.
Mech. Mater. 1985, 4, 395–407. [CrossRef]
18. Bonora, N. A nonlinear CDM model for ductile failure. Eng. Fract. Mech. 1997, 58, 11–28. [CrossRef]
19. Tong, L.; Huang, X.; Zhou, F.; Chen, Y. Experimental and numerical investigations on extremely-low-cycle
fatigue fracture behavior of steel welded joints. J. Constr. Steel Res. 2016, 119, 98–112. [CrossRef]
20. Kanvinde, A.M.; Deierlein, G.G. Cyclic Void Growth Model to Assess Ductile Fracture Initiation in Structural
Steels due to Ultra Low Cycle Fatigue. J. Eng. Mech. 2007, 133, 701–712. [CrossRef]
21. Rice, J.R.; Tracey, D.M. On the ductile enlargement of voids in triaxial stress fields. J. Mech. Phys. Solids 1969,
17, 201–217. [CrossRef]
22. Myers, A.T.; Deierlein, G.G.; Kanvinde, A.M. Testing and Probabilistic Simulation of Ductile Fracture Initiation
in Structural Steel Components and Weldments; Blume Center TR 170; Stanford University: Stanford, CA,
USA, 2009.
23. Fell, B.V.; Kanvinde, A.M.; Deierlein, G.G. Large-Scale Testing and Simulation of Earthquake-Induced Ultra
Low Cycle Fatigue in Bracing Members Subjected to Cyclic Inelastic Buckling; Blume Center TR172; Stanford
University: Stanford, CA, USA, 2010.
24. Zhou, H.; Wang, Y.; Shi, Y.; Xiong, J.; Yang, L. Extremely low cycle fatigue prediction of steel beam-to-column
connection by using a micro-mechanics based fracture model. Int. J. Fatigue 2013, 48, 90–100. [CrossRef]
25. Liao, F.F. Study on Micromechanical Fracture Criteria of Structural Steels and Its Applications to Ductile
Fracture Prediction of Connections. Ph.D. Thesis, Tongji University, Shanghai, China, 2007. (In Chinese).
26. Xing, J.H.; Guo, C.L.; Li, Y.Y.; Chen, A.G. Damage coefficient identification of micromechanical fracture
prediction models for steel Q235B. J. Harbin Inst. Technol. 2017, 49, 184–188. (In Chinese)
27. Pereira, J.; De Jesus, A.M.; Fernandes, A. A new ultra-low cycle fatigue model applied to the X60 piping steel.
Int. J. Fatigue 2016, 93, 201–213. [CrossRef]
28. Pereira, J.; Van Wittenberghe, J.; De Jesus, A.; Thibaux, P.; Correia, J.; Fernandes, A. Damage behaviour of
full-scale straight pipes under extreme cyclic bending conditions. J. Constr. Steel Res. 2018, 143, 97–109.
[CrossRef]
29. Wen, H.J.; Mahmoud, H. New Model for Ductile Fracture of Metal Alloys. II: Reverse Loading. J. Eng. Mech.
2016, 142, 08216002. [CrossRef]
30. Kanvinde, A.M.; Deierlein, G.G. The Void Growth Model and the Stress Modified Critical Strain Model to
Predict Ductile Fracture in Structural Steels. J. Struct. Eng. 2006, 132, 1907–1918. [CrossRef]
31. Kanvinde, A.M.; Deierlein, G.G. Micromechanical Simulation of Earthquake-Induced Fracture in Steel Structures;
Blume Center TR 145; Stanford University: Stanford, CA, USA, 2004.
32. Ramberg, W.; Osgood, W. Description of Stress–Strain Curves by Three Parameters; NACA Techical Note No.902;
National advisory committee for aeronautics: Washington, DC, USA, 1943.
33. Bai, Y.; Wierzbicki, T. A new model of metal plasticity and fracture with pressure and Lode dependence.
Int. J. Plast. 2008, 24, 1071–1096. [CrossRef]
34. Liao, Y.H. Research on Ultra Low Cycle Fatigue Properties and Fracture Mechanism of Steel Bridge Welded
Joint. Master’s Thesis, Zhejiang University, Hangzhou, China, 2018. (In Chinese).
35. Liao, F.F.; Wang, W.; Chen, Y.Y. Parameter calibrations and application of micromechanical fracture models
of structural steels. Struct. Eng. Mech. 2012, 42, 153–174. [CrossRef]
36. Lemaitre, J.; Chaboche, J.L. Mechanics of Solid Materials; Cambridge University Press: Cambridge, UK, 1990.
Materials 2019, 12, 1615 17 of 17

37. Barsoum, I.; Faleskog, J. Rupture mechanisms in combined tension and shear—Experiments. Int. J.
Solids Struct. 2007, 44, 1768–1786. [CrossRef]
38. Barsoum, I.; Faleskog, J. Rupture mechanisms in combined tension and shear—Micromechanics. Int. J. Solids
Struct. 2007, 44, 5481–5498. [CrossRef]
39. Danas, K.; Castañeda, P.P. Influence of the Lode parameter and the stresstriaxiality on the failure of
elasto-plastic porous materials. Int. J. Solids Struct. 2012, 49, 1325–1342. [CrossRef]
40. Zhuge, H.Q.; Xie, X.; Tang, Z.Z. Study on length of damaged zone of steel piers under bidirectional horizontal
earthquake components. China J. Highway Transp. 2019, in press.

© 2019 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access
article distributed under the terms and conditions of the Creative Commons Attribution
(CC BY) license (http://creativecommons.org/licenses/by/4.0/).