Download as pdf or txt
Download as pdf or txt
You are on page 1of 18

Accepted Manuscript

The effect of residual stresses on fatigue crack propagation in


welded stiffened panels

Željko Božić, Siegfried Schmauder, Hinko Wolf

PII: S1350-6307(17)30431-4
DOI: doi: 10.1016/j.engfailanal.2017.09.001
Reference: EFA 3283
To appear in: Engineering Failure Analysis
Received date: 7 May 2017
Revised date: 1 September 2017
Accepted date: 1 September 2017

Please cite this article as: Željko Božić, Siegfried Schmauder, Hinko Wolf , The effect
of residual stresses on fatigue crack propagation in welded stiffened panels, Engineering
Failure Analysis (2017), doi: 10.1016/j.engfailanal.2017.09.001

This is a PDF file of an unedited manuscript that has been accepted for publication. As
a service to our customers we are providing this early version of the manuscript. The
manuscript will undergo copyediting, typesetting, and review of the resulting proof before
it is published in its final form. Please note that during the production process errors may
be discovered which could affect the content, and all legal disclaimers that apply to the
journal pertain.
ACCEPTED MANUSCRIPT

The Effect of Residual Stresses on Fatigue Crack Propagation in Welded


Stiffened Panels

Željko Božić
University of Zagreb, Faculty of Mech. Eng. And Nav. Arch., Croatia
E-mail: zeljko.bozic@fsb.hr

Siegfried Schmauder

T
Universität Stuttgart, IMWF, Pfaffenwaldring 32, D-70569 Stuttgart
E-mail: siegfried.schmauder@imwf.uni-stuttgart.de

IP
Hinko Wolf

CR
University of Zagreb, Faculty of Mech. Eng. And Nav. Arch., Croatia
E-mail: hinko.wolf@fsb.hr

US
Abstract: This paper presents a method for predicting fatigue crack propagation in welded stiffened
panels accounting for the effects of residual stresses. Well known power law models were employed to
simulate fatigue life for a welded stiffened panel specimen damaged with a central crack. Stress intensity
AN
factor values (SIF) were obtained in finite element (FE) analyses by a linear superposition of the SIF
values due to the applied load and due to weld residual stresses. The FE models included idealized,
rectangular or triangular, residual stress distribution profiles. The effect of welding residual stresses on
M

the crack propagation rate is taken into account by replacing the nominal stress ratio R in the power laws
by the effective stress intensity factor ratio Reff. The considered stiffened panel specimen, made of mild
steel commonly used in ship structures and manufactured by electric arc welding process, was subjected
ED

to the fatigue test with constant amplitude loading until failure occurred. By using the presented method
remaining fatigue life of welded stiffened panel specimen was analyzed. The simulated crack growth rate
was relatively higher in the region of tensile residual stresses and lower between the stiffeners where
PT

compressive residual stresses prevail, which is in good agreement with physical principles and
experimentally obtained results.

Keywords: Crack propagation, Residual stress, Remaining fatigue life, Finite element analysis, Stiffened
CE

panel

NOMENCLATURE
AC

a = half crack length; crack parameter


a0, afin = initial crack length and final crack length
C, m = the Paris-Erdogan equation constants
da/dN = fatigue crack growth rate
E = Young’s modulus
Fmax = maximum fatigue load
Fmin = minimum fatigue load
K = stress intensity factor (SIF)
Kapp = stress intensity factor due to the applied load
Kmax = maximum stress intensity factor
ACCEPTED MANUSCRIPT

Kmin = minimum stress intensity factor


Kres = stress intensity factor due to weld residual stresses
Kth = stress intensity factor threshold
Ktot = total stress intensity factor
N = number of stress cycles for the fatigue crack propagation
R = nominal stress ratio due to external load
Reff = effective stress intensity factor ratio
U = factor in the Elber or Schijve model
F = applied force range
K = stress intensity factor range

T
Keff = effective stress intensity factor range

IP
Δ = average applied stress range
ν = Poisson’s ratio

CR
𝜎0 = Yield strength

1. Introduction

US
Stiffened panels are commonly used in ship, aircraft and other thin-walled structures which
require high stiffness and strength combined with low weight. They are mostly implemented in
AN
the fuselage and wings of aircrafts and the deck and side ship structure. At stress concentration
sites in thin walled structures fatigue cracks may initiate and further grow due to cyclic service
loads. This can finally lead to unstable fracture of the structural element, which may result in a
loss of structural integrity and structural failure. Extensive studies on crack initiation and
M

propagation in riveted and welded aircraft stiffened panels have been conducted [1,2,3,4].
Hobbacher [5] presented in his recommendations general methods for the assessment of fatigue
ED

damage in welded components. The book provides a basis for the design and analysis of welded
components subjected to fluctuating forces. It is also intended to those who are establishing
fatigue design codes.
PT

Rather few studies have been conducted on the propagation of fatigue cracks in stiffened panels
that are used in ship structures. Božić [6] carried out experimental and finite element studies on
the propagation of multiple long cracks in stiffened panels under cyclic tension loading, related
CE

to fatigue problems in ship deck structures. Welded stiffened panel specimens damaged with
multiple cracks were subjected to fatigue tests with constant amplitude loading. Corresponding
fatigue crack propagation simulations were carried out by using the Paris-Erdogan law. Fricke
AC

contributed extensively to the fatigue life assessment of welded ship structures [7,8,9]. In [7]
Fricke et al. investigated two structural details, the web frame corners being typical for ro/ro
ships and the intersection between longitudinals and transverse web frames, which showed
fatigue failures in containerships. The investigations give a good insight into the fatigue strength
of complex welded structures and into current numerical analyses possibilities.
Dexter et al. [10] and Mahmoud et al. [11] carried out cyclic tension fatigue tests on half-scale
welded stiffened panels in order to study propagation of long cracks and their interaction with
the stiffeners. The authors developed models to predict the growth of long fatigue cracks in
welded stiffened panels, where stress intensity factors were computed based on linear-elastic
fracture mechanics superposition assumption, taking account of the typical residual stress
distribution in stiffened panels. Idealized residual stress patterns, similar to what was measured
ACCEPTED MANUSCRIPT

in the panels, were introduced in the finite element model to calculate stress intensity factors.
Crack propagation in box girders with welded stiffeners was predicted with reasonable accuracy.
Feng et al. [12] developed a probabilistic crack growth model for a stiffened panel, allowing for
the existence of multiple cracks both in the stiffeners and in the plate. The developed
probabilistic model can be used for fatigue crack growth analysis and for reliability assessment
of a stiffened panel. Garbatov et al. [13] analysed the effect of corrosion deterioration on a
transversely stiffened welded joint subjected to cyclic loading. The fatigue tests were carried out
on the specimens cut from a box girder and initially corroded in real sea water conditions. A
direct relation between the initial corrosion surface roughness and estimated fatigue life was
observed.

T
The influence of residual stress on fatigue strength of welded joints is a topic largely studied in

IP
recent literature. Ferro [14] analyzed by the numerical simulation the residual stress
redistribution near a V-notch tip induced by cyclic loads and introduced a method for the fatigue

CR
resistance estimation of pre-stressed components, which is based on the local strain energy
approach. Ferro et al. [15] studied the effect of fatigue loading on the asymptotic residual stress
redistribution near the weld toe of a butt-welded joint by means of numerical simulations. A
model was proposed to estimate the influence of residual stresses on the fatigue strength of

US
welded joints. The results predicted by the proposed model were in good agreement with
experimental results taken from the literature. Ferro and Berto [16] estimated the influence of
residual stresses on the fatigue resistance of Al-alloy butt-welded joints is by using the local stain
AN
energy density approach. Predicted values by means of the proposed method were in a good
agreement with experimental data taken from literature.
Residual stresses introduced by the welding process can significantly influence the crack growth
M

rate in stiffened panels [6, 10, 11]. Predan et al. [17] analysed the effect of residual stress on the
fatigue crack propagation for a loading regime close to threshold stress intensity factor range.
ED

Fatigue crack propagation experiments were performed on notched bending specimens machined
from a welded plate. It was observed in the experiment that the residual stresses induced a
variation in the crack propagation rate along the crack front. Leitner et al. [18] studied crack
propagation in welded structures and rehabilitation after pre-fatigue loading by the high-
PT

frequency mechanical impact (HFMI) post-treatment technique. A thin-walled longitudinal


stiffener specimen made of mild structural steel S355 was investigated. Fracture mechanics
analyses were mainly performed on the basis of the weight-function approach. The influence of
CE

effective mean stresses on fatigue crack propagation in welded structures was emphasized in the
paper. Nitschke-Pagel at al. [19] proposed a concept which considers the combined effect of load
mean stresses and cyclically stabilized residual stresses. It was recommended in the paper to
AC

evaluate residual stress effects based on the effective stress ratio reflecting local loading
conditions.
In a recent study Božić et al. [20] presented a multiscale fatigue crack growth model by which
the total fatigue life of a welded stiffened panel can be analysed. It was demonstrated that from
the Molecular Dynamics (MD) simulation the critical resolved shear stress (CRSS) can be
inferred. CRSS is then the input parameter for a micromechanics model such as for example the
Tanaka-Mura model. The micromechanical model provides information on the number of
loading cycles to initiate a small crack and its size, which is further an input for the macroscale
fatigue crack propagation model based on power law equations, by means of which a total
fatigue life up to fracture as a final event can be assessed. It was demonstrated in the paper that
ACCEPTED MANUSCRIPT

tensile residual stresses may significantly accelerate crack propagation in welded stiffened
panels.
In this paper the fatigue crack propagation in a welded stiffened panel specimen is studied by
using power law models, taking into account the effect of residual stresses. The considered
stiffened panel specimen is made of mild steel and manufactured by electric arc welding process
commonly used in shipbuilding industry. The finite element method is used to determine the
Mode I total stress intensity factor Ktot by a linear superposition of the SIF values due to the
applied load, Kappl, and due to welding residual stresses, Kres. Idealized residual stress distribution
profiles, such as rectangular or triangular distributions, are implemented in FE analyses by using
the software package ANSYS [21]. Remaining fatigue crack propagation life of the specimen

T
under consideration has been simulated by using the Paris-Erdogan law [22]. The Elber [23] and

IP
Schijve [24] models are employed to simulate fatigue crack propagation in the analysed stiffened
panel specimen. The influence of welding residual stress on the crack growth rate is taken into

CR
account in the power laws by introducing the effective stress intensity factor ratio Reff [20, 25].
Simulated fatigue crack propagation life for the test specimen was compared with experimental
results of cyclic tension fatigue tests [6, 26].

US
2. Fatigue test specimen and loading conditions

Under cyclic service loading in deck structure of a ship a fatigue crack may initiate at the sniped
AN
end of the face plate of a deck longitudinal stiffener [6]. The initiated crack can further propagate
through the stiffener and then it penetrates into the deck plate ship structure.
Test specimen used in this study is a stiffened panel made of mild steel for welding commonly
M

used in ship structures. The specimen is manufactured from a plate of 4 mm thickness by


employing manual arc welding process. An initial through-wall sharp notch was machined in the
middle of the specimen. Figure 1 shows the specimen geometry and initial sharp notch details.
ED

The sharp notch was manufactured in the following way. At first a hole with one millimeter
radius was drilled through the middle stiffener and the plate. Then the notch was cat by electric
arc with the beam thickness of 0.2 mm. After the notch of the length 2a=8mm was introduced
PT

into the plate, the notch fronts at both ends of the notch were sharpened manually by a very thin
saw.
One should note that the middle stiffener was completely cut at the crack location, while the
CE

other two stiffeners were intact. The specimen geometry was selected with the intention to model
the situation where a fatigue crack has already propagated through the stiffener into the deck
plate of a ship and it further propagates towards the intact stiffeners on the both sides. This is a
AC

realistic scenario for crack propagation in aged ship structures as the accident of the Castor
tanker ship showed [11].
In Table 1 are given material properties and chemical composition of the mild steel under
consideration, [27]. Conditions of the constant stress amplitude fatigue test are given in Table 2.
Here, Δσ and Ao, denote the average stress range far away from the notch and the related cross
sectional area of the intact section, respectively. The force range is denoted as ΔF = Fmax - Fmin,
and the stress ratio is given as R = Fmin/Fmax. The applied constant amplitude stress range in the
fatigue test was Δσ = 80MPa. The fatigue test was carried out at a loading frequency of 3 Hz.
The given nominal stress range of 80MPa is higher than the IIW-recommended value of FAT36
for welded steel joints exhibiting cracks; hence, a crack growth with a total number of cycles
below two million cycles can be expected, which actually was the case in this particular
ACCEPTED MANUSCRIPT

experiment.
So-called crack gauges were used in the fatigue test to measure the length of the propagating
crack. The grid of crack gauges, which are bonded to the specimen in front of a crack tip, is cut
with crack development. This results in resistance change, which is measured and transferred
into propagated crack length. A crack gauge type with the crack propagation measuring range of
25 mm and the resolution of 1 mm was used in the present experiment. In order to measure
longer ranges several gauges can be bonded sequentially. In this experiment the objective was to
measure crack propagation from the initial crack length 2a=8mm until crack tips approach the
side stiffeners, which are distanced 160mm.

T
IP
CR
US
AN
M

Figure 1. Stiffened panel specimen with a central sharp notch


ED

Table 1. Material’s properties


Mechanical properties
PT

E – Young's modulus ν – Poisson’s coefficient σo – Yield strength


206 000 MPa 0.3 235 MPa
CE

Chemical composition (%)


Cmax Simax Mnmin Pmax Smax
0.18 0.35 0.70 0.035 0.035
AC

Table 2. Fatigue test loading conditions


Ao ΔF Δσ R
[mm2] [N] [MPa]
1200 96000 80 0,02

3. Finite element analysis and stress intensity factor calculation

3.1 Finite element model of the stiffened panel specimen


ACCEPTED MANUSCRIPT

Wawrzynek et al. presented in their paper [28] the developed fracture analysis program FRANC
which incorporates the concepts of finite element analysis, fracture mechanics, computer
graphics, automatic mesh generation and finite element postprocessing. The program allows a
practicing engineer or a researcher to perform an incremental crack propagation analysis. A
number of example problems of incremental crack propagation analysis was presented.
Under tension load in damaged stiffened panel specimen out-of-plane displacements occur with
local bending as a result, due to the geometric eccentricities introduced by the cracked cross-
section. For such a thin-walled structure it is appropriate to use eight node quadratic
isoparametric shell elements in the FE analysis. The region surrounding the crack tip was

T
meshed with singular elements, having midside nodes adjacent to the crack tip placed at the

IP
quarter points [21, 29 - 32].
Due to the symmetry of the specimen geometry and loading conditions it is sufficient to model

CR
one quarter of the specimen. The FE mesh of the stiffened panel specimen is shown in Figure 2.
Loading conditions are given in Table 2 and the applied boundary conditions are depicted in
Table 3. The cracked surface is represented by the line AG, where G is a crack tip. The stiffener
at location A is completely cut, as shown in Figure 2. Therefore, the AG part of the model

US
boundary is free to deform. The rectangular surface between corners B, C, D and E represents the
thick plate as depicted in Figure 1. This part of the model can move in axial direction only,
because of the settings of the testing equipment.
AN
M
ED
PT
CE
AC

Figure 2. FE mesh of the stiffened panel specimen

With shell elements employed in the analysis the stress intensity factor values were extracted
from FE results for nodal displacements of singular elements in a standard post processing
procedure by using ANSYS software package [21]. The well-known extrapolation method for
determining SIF values from nodal displacements of singular crack tip elements was proposed by
Barsoum [29] and by Henshell and Shaw [30] and further investigated by Bloom [31] and
Hibbitt [32].
ACCEPTED MANUSCRIPT

Table 3. FE mesh and boundary conditions.


Plane Boundary conditions
ABC symmetry with respect to yz plane
CD axial load applied
BCDE displacements in xz plane allowed
GF symmetry with respect to xy plane
AG free (cracked surface)

T
IP
3.2 Welding residual stress modelling and stress intensity factor calculation

CR
The FEM is broadly employed for calculation of SIFs. In this study the SIFs were calculated in a
linear elastic FE analysis by using the above described model and the loading and boundary
conditions. Singular elements were used in the FE model to mesh the crack tip region. For this

US
purpose well-known “quarter-point” elements described by Barsoum [29] and by Henshell and
Shaw [30] were applied. The crack tip displacements extrapolation method was implemented for
evaluating SIFs by using ANSYS program [21], as explained in the previous paper [20].
AN
The SIF values in welded stiffened panels may be significantly influenced by the residual
stresses, which in turn can influence the fatigue crack propagation rate [20, 33, 34]. For the
prediction of fatigue crack propagation rates in residual stress fields the so-called residual stress
M

intensity factor Kres is required. According to linear elastic fracture mechanics (LEFM)
superposition concept, the total SIF value, Ktot, consists of the part due to the applied external
load, Kapp, and the part due to residual stresses, Kres, as given by the equation:
ED

Ktot = Kapp + Kres (1)


PT

For accurate computing the Kres, inputting correct initial stress conditions into numerical models
is essential [25, 35]. In the finite element software package ANSYS the initial stress conditions
can be defined by using the command “INISTATE” [21]. Residual stresses are applied as initial
CE

stresses in the first load step. The imposed initial stresses should be equilibrated in the first
analysis step in order to achieve zero stresses in free edges of the model and self-equilibrium
condition [20, 35]. In the next step the external load is applied. Based on the superposition rule
AC

of LEFM the total SIF value Ktot is calculated.


In a welded stiffened panel, commonly used in ship structures, the residual stress distribution is
such that tensile stresses, which can reach yield strength, appear at stiffeners site, and between
the stiffeners compressive stresses occur [36]. Dexter et al. [10, 11] measured residual stresses in
welded stiffened panel specimens after fatigue tests. It was demonstrated that these stresses can
be idealized by the triangular distribution [10] or the rectangular distribution [11], with peak
residual tensile stress equal to the yield strength of the steel under consideration. In the present
study the distribution of residual stresses in welded stiffened panel specimen is modelled as
depicted in Figures 3a and 4b.
ACCEPTED MANUSCRIPT

T
IP
a) b)

CR
Figure 3. Welding residual stress distribution: a) rectangular; b) triangular.

The tensile regions around the stiffeners are idealized as rectangular or triangular shapes with a

US
base width equal to 10mm, where the stress level reaches the yield strength of the considered
material, σo = 235 MPa. The estimated base width is in accordance with Dexter’s models [10, 11]
with roughly 25 % of the plate between the two stiffeners exposed to tensile residual stress. The
magnitude of the compressive residual stresses is determined by the equilibrium conditions.
AN
One should note that in the FE modeling for each considered crack length residual stresses were
implemented in the above described manner in the model which includes a crack, as depicted in
M

Fig. 2, before external load was applied in the next load step and Ktot was calculated in the
postprocessing procedure. In this way the effect of residual stresses was taken into account with
respect to increasing crack length, as well as the change in global stiffness of the model.
ED

Glinka [37] proposed the superposition method which is based on the principle of LEFM. In the
residual stress field, under the cyclic loads, the total SIF range ΔKtot and effective SIF ratio Reff
PT

are given as:

∆𝐾tot = (𝐾app,max + 𝐾res ) − (𝐾app,min + 𝐾res ) = ∆𝐾app (2)


CE

(𝐾 +𝐾res )
𝑅eff = (𝐾app,min +𝐾 (3)
res )
AC

app,max

Under cyclic loading, the Reff changes as the crack tip propagates through the residual stress
field. It is important to note that Reff is the crack tip stress intensity factor ratio which is
determined based on the calculated values of Ktot,max and Ktot,min. The value Reff is not the same as
the nominal applied stress ratio R, which was in the present experiment equal to zero, as
indicated in Table 2.

The mode I SIF values, KI, were calculated for the considered stiffened panel specimen by using
the FE model depicted in Figure 2. Determined SIF values versus to half crack length a are
shown in Fig. 4. Kapp,max curve represents the maximum SIF values due to the applied external
stress range Δσ = 80MPa, without considering residual stresses. Kapp,max initially increases with
ACCEPTED MANUSCRIPT

the increasing crack length and as the crack tip approaches the intact stiffener Kapp,max begins to
decrease, which shows the effect of crack arrest. Ktot,max curve represents the maximum SIF
values for the case when the residual stresses are taken into account along with the applied
external loading stress range. One can see that residual stresses significantly increase Ktot,max
values for shorter crack lengths, in the area where tensile residual stresses prevail. Between the
stiffeners compressive residual stresses reduce the Ktot,max values. The Ktot,max values associated
with the rectangular residual stress distribution are higher in the region close to the stiffener, and
lower between the stiffeners, in comparison to the Ktot values associated with the triangular
distribution.

T
IP
55 0.6
Ktot,max(rectangular) Rectangular

CR
50 Triangular
Ktot,max(triangular)
0.4
45 Kapp,max

40 0.2
)
1/2

US
35
K (MPa m

Reff
0
30

25 -0.2
AN
20
-0.4
15
M

10
0 0.02 0.04 0.06 0.08 0 0.02 0.04 0.06 0.08
a (m) a (m)

Fig. 4 SIF values Ktot,max and Fig. 5 The effective stress ratio Reff.
ED

Kapp,max

Calculated Reff values versus half crack length a are plotted in Figure 5. Reff values are higher
PT

around the cut stiffener and have positive values close to 0.5. In this region the Reff values are
very high compared to the nominal R ratio associated with the applied external cyclic load,
CE

which equals to zero, as indicated in Table 2. Therefore, in this region a higher fatigue crack
growth rate (FCGR) is to be expected. In the region between the stiffeners, where compressive
residual stresses prevail, Reff values are negative, which could decrease FCGR, as it will be
discussed in subsequent sections.
AC

4. Crack propagation models and fatigue life simulation

4.1 Power laws and crack propagation modeling

In order to support the damage tolerance concepts in cyclically loaded metallic structures,
various FCG prediction models based on LEFM have been developed [33, 34]. Paris and
Erdogan introduced the well-known power law model for predicting FCG under a constant stress
range [22], as given by equation (4). The Paris-Erdogan equation material constants C and m can
be obtained from the log-log plot of Δa/ΔN versus ΔK data. As long as the data follow a straight
ACCEPTED MANUSCRIPT

line relationship, the Paris-Erdogan law is commonly used. The major limitation of the Paris-
Erdogan equation is that it does not account for the load ratio, R = Kmin /Kmax. For the Paris-
Erdogan equation ΔK = Kmax - Kmin.

𝑑𝑎
= 𝐶(∆𝐾)𝑚 (4)
𝑑𝑁

T
The number of constant amplitude loading cycles due to which a crack propagates from its initial

IP
crack length, a0, to a final crack length, afin, can be determined by the integration of equation (4),
which becomes:

CR
𝑎fin
d𝑎
𝑁=∫ (5)
𝑎0 𝐶[∆𝐾]𝑚

US
The integration of equation (5) can be performed numerically [6, 20, 26, 38-40]. For a certain
number of discrete crack lengths the SIF values are calculated by FEA. Between the two discrete
AN
SIF values, a linear interpolation is used, the so-called piecewise linear function. Then, for each
segment fatigue life is calculated numerically and the total fatigue life, N, is calculated as the
sum of fatigue lives of each segment. In the present study the a-K curves given in Figure 4 are
M

obtained by calculating the K values for about 70 discrete a values.


ED

Donahue [41] suggested to use the effective stress intensity factor range ΔKeff = ΔK - ΔKth, in the
power law crack growth model given by equation (4), instead of ΔK. The ΔKth is the so called
SIF threshold range, a value below which no crack propagation should occur or it occurs at a
PT

very low FCGR, and ΔK is the applied SIF range in a loading cycle. The threshold range ΔKth
can be determined experimentally, and it often depends on the R ratio [42, 43]. The effect of R
ratio on the threshold stress intensity range ΔKth for various steels was investigated by Tanaka
CE

[42]. Tanaka showed that for mild steels, such as used in the present experiment, the ΔKth
decreases with respect to increasing R.
AC

Elber observed that crack closure decreases the fatigue crack growth rate by reducing the
effective SIF range [23]. He proposed the following model:

∆𝐾eff = 𝑈 ⋅ ∆𝐾app = (0.5 + 0.4𝑅)∆𝐾app (6)

Schijve improved the Elber model, based on test data of wider range of R ratios [24], and
proposed the following equation:

∆𝐾eff = 𝑈 ⋅ ∆𝐾app = (0.55 + 0.33𝑅 + 0.12𝑅 2 )∆𝐾app (7)


ACCEPTED MANUSCRIPT

In this study the Elber and Schijve models, which take the load ratio R into account are
implemented to simulate fatigue crack propagation in the test specimen. The nominal R ratio in
the original equations was replaced by the effective ratio Reff.

Using the Reff values from Figure 5, to replace the nominal R in equations 6 and 7, the factor U
was calculated. Calculated U values for the Elber and Schijve models, considering the
rectangular and triangular residual stress distributions, are depicted in Figures 6 and 7. The
model with rectangular residual stress distribution has higher U values, which is due to higher
Ktot,max values.

T
IP
0.8 0.8
Elber Elber
0.75 Schijve 0.75 Schijve

CR
0.7 0.7

0.65 0.65

0.6 0.6

US
0.55 0.55

U
U

0.5 0.5

0.45 0.45
AN
0.4 0.4

0.35 0.35

0.3
0 0.02 0.04 0.06 0.08 0 0.02 0.04 0.06 0.08
a (m) a (m)
M

Fig. 6 Factor U for the rectangular Fig. 7 Factor U for the triangular
ED

residual stress distribution residual stress distribution

According to equations 6 and 7 the effective SIF range ΔKeff was calculated for the Elber and
PT

Schijve models as given in Figures 8 and 9, respectively.

Table 4. Power laws’ constants*.


CE

Model C m
Elber 1.67e-10 2.75
Schijve 1.14e-10 2.75
*The units for ΔK and Δa/ΔN are [MPa ·m1/2 ] and [m], respectively, [20].
AC

The power laws’ constants given it Table 4 were determined in a previous study by means of
crack growth rate diagrams based on a–N data obtained for centrally cracked plate specimens [6,
20, 38-40]. The C constants were estimated in the manner to provide a good fit of computed
fatigue lifetime with experimentally determined a–N curve. The values given in Table 4 are
different from the Paris-Erdogan power law constants C and m given by Hobbacher’s IIW
recommendation, C=1.65e-11 and m=3.0, which is to expect for different power laws, as stated
by Broek [34]. Since the influence of the effective local stress ratio Reff and SIF values, and
therefore SIF threshold values, are implicitly included in the Elber and Schijve model, the
exponent m used in this study [20, 38-40] is slightly lower compared to the value recommended
by Hobbacher [5].
ACCEPTED MANUSCRIPT

30 30
Keff(rectangular) Keff(rectangular)
Keff(triangular) Keff(triangular)
25 25
)

)
1/2

1/2
20 20
K (MPa m

K (MPa m
15 15

T
10 10

IP
5 5
0 0.02 0.04 0.06 0.08 0 0.02 0.04 0.06 0.08
a (m) a (m)

CR
Fig. 8 The ΔKeff values for the Fig. 9 The ΔKeff values for the
Elber model Schijve model

US
Based on the ΔKeff values and material’s constants C and m given in Table 3, fatigue crack
propagation lives were calculated for the Elber and Schijve models by integrating the equation
AN
(5), as given in Figures 10 and 11, respectively.
0.08 0.08
Rectangular Rectangular
Triangular Triangular
M

0.07 0.07
Experiment Experiment
0.06 0.06
ED

0.05 0.05
a (m)

a (m)

0.04 0.04
PT

0.03 0.03

0.02 0.02
CE

0.01 0.01

0 0
0 1 2 3 4 5 0 1 2 3 4 5
N (cycles) 5
x 10 N (cycles) 5
x 10
AC

Fig. 10 Comparison of simulated a-N Fig. 11 Comparison of simulated a-N


values with experimental values with experimental
results for the Elber model results for the Schijve model

5. Discussion of results

The linear elastic FE analysis showed that high tensile residual stresses at the site of a cut welded
stiffener significantly increase total SIF values, Ktot, for the considered stiffened panel
specimens, as shown in Figure 4. The model with rectangular residual stress distribution (RRSD)
yielded higher total SIF values Ktot in the area with tensile residual stresses, and lower Ktot values
in the area of compressive residual stresses, compared with TRSD case, as can be seen in Figure
ACCEPTED MANUSCRIPT

4. Correspondingly, the effective stress ratio values Reff are higher in the area with tensile
residual stresses, and lower Reff values appear in the area of compressive residual stresses, as
Figure 5 shows.
The U functions for both the Elber and Schijve models have similar shape, where the Schijve
model provides higher U values. In case of RRSD for both the Elber and Schijve models the U
function reaches higher values in the area of tensile residual stresses and lower U values in the
area of compressive residual stresses, compared with the TRSD case, as can be seen in Figures 6
and 7. The effective SIF values ΔKeff depend on the U values and the applied SIF values ΔKapp
according to equations 6 and 7. Calculated effective SIF values for the Elber and Schijve models
obtained based on the U function are given in Figures 8 and 9. One can see that both the Elber

T
and Schijve models for RRSD case provide higher effective SIF values ΔKeff in the area with

IP
tensile residual stresses, and lower effective SIF values ΔKeff between the stiffeners, compared
with the TRSD case. The Elber and Schijve models for both RRSD and TRSD cases in the area

CR
with compressive residual stresses yield effective SIF values ΔKeff, which increase
monotonically, almost linearly, with the increasing crack length. Compressive welding residual
stresses between the stiffeners decreased the total SIF values, and correspondingly, the effective
SIF values.

US
The Elber model shows better agreement of simulated a–N curve with the experimental results,
compared to the Schijve model, particularly in RRSD case, as can be seen in Figure 10. In this
case the curve of the effective SIF values ΔKeff versus half crack length a, increase at a low rate
AN
and in turn provides FCGR very similar to that which occurred in the experiment. As a result, the
simulated a–N curve follows very well the experimentally obtained fatigue life curve.
In the case of the Schijve model effective SIF values ΔKeff increase at a slightly higher rate
M

compared with the Elber model, as can be seen in Figure 9. As a consequence of this, the Schijve
model yields a–N curves which grow at a slightly higher FCGR than observed in the experiment.
In summary, both the Elber and Schijve model managed to describe well crack propagation in the
ED

region of high total SIF values, which appear up to the half crack lengths a=20mm, where very
high FCGR was observed in the experiment. This is a big step forward compared with the
results obtained by the Paris-Erdogan model and the Donahue model without considering
PT

welding residual stresses, as reported in [6] and [26], respectively. High tensile residual stresses
in this region reduce crack closure effects and increase crack growth rates, as was observed in
previous studies [20, 25]. FCG models which do not take effects of residual stresses into account
CE

could not describe high FCGR in regions of high welding residual stresses.
The presented methodology of crack propagation simulation and analysis may have important
applications in fitness for service assessment of aged ship and offshore structures, and other thin
AC

walled structures, as well as rational inspection and maintenance planning. Determined crack
growth rates and a-N curves in conjunction with regularly spaced inspections can help
maintenance engineers and surveyors to decide whether some cracks require immediate repair or
not and how long it would take a crack to propagate from the detected length to an unacceptable
large size.
Since calculated total and effective SIF values were within the crack propagation region II where
the Paris-Erdogan law applies, the application of LEFM is appropriate. The crack tip after a few
millimeters passes the heat affected zone and the crack further propagates in the base material.
For a more precise FCG analysis a nonlinear fracture mechanics parameter such as J-integral can
be considered as the crack propagation force.
In further work, the effect of secondary bending on FCGR in cracked, axially loaded stiffened
ACCEPTED MANUSCRIPT

panels, should be considered. It is well known that secondary bending may increase FCGR, [44].
In addition, residual stresses in welded stiffened panel specimens should be measured after
fatigue tests, and measured values should be introduced in the existing FE models. Along with
the here discussed Elber and Schijve models, other well-known power law models which can
take into account the effective stress intensity factor ratio values, Reff, can be examined.

5. Conclusion

The stiffened panel specimen made of mild steel commonly used in ship structures and
manufactured by electric arc welding processes, was subjected fatigue testing with constant

T
amplitude loading until failure occurred. The Elber and Schijve power law models were

IP
employed to simulate fatigue life for the welded stiffened panel specimen, taking into account
the effect of residual stresses introduced during the welding process. The FEM models included

CR
typical idealized residual stress distribution patterns such as rectangular and triangular
distributions. The FE analysis showed that high tensile residual stresses at site of the cut stiffener
significantly increase the total stress intensity factor Ktot, and correspondingly the simulated
FCGR in this region, while the compressive welding residual stresses between the two stiffeners

US
decreased the total SIF values Ktot, and the crack growth rate. The Elber model showed better
agreement with the experimentally obtained a-N curve, particularly for the rectangular residual
stress distribution. The rectangular residual stress distribution provided better agreement with the
AN
experiment regarding the crack growth rate and computed fatigue lifetime for both models. It
was demonstrated that residual stresses should be taken into account for proper fatigue crack
growth modelling in welded stiffened panels. The developed procedure can be applied to real
M

thin-walled structures with welded stiffeners.


Acknowledgements
ED

This work was supported by the Deutsche Forschungsgemeinschaft (DFG) under Grant No.
Schm 746/132-1 and by the Croatian Science Foundation Grant No. 120-0362321-2198. The
PT

support is gratefully acknowledged.

References
CE

[1] Poe CC, Jr. Fatigue crack propagation in stiffened panels. Damage tolerance in aircraft
structures, ASTM STP 486. Conshohocken (PA): American Society for Testing and
Materials; 1971. p. 79-97.
AC

[2] Schijve J, Campoli G, Monaco A. Fatigue of aircraft materials and structures. International
Journal of Fatigue 1994; 16(1):21-32.
[3] Sabelkin V, Mall S, Avram J. Fatigue crack growth analysis of stiffened cracked panel
repaired with bonded composite patch. Engng Fract Mech 2006;73:1553–67.
[4] Uz MV, Koçak M, Lemaitre F, Ehrström JC, Kempa S, Bron F. Improvement of damage
tolerance of laser beam welded stiffened panels for airframes via local engineering.
International Journal of Fatigue 2009;31:916–926.
[5] Hobbacher A F: Recommendations for Fatigue Design of Welded Joints and Components.
Book IIW Collection. Second Edition. Springer International Publishing, 2016. ISBN: 978-
3-319-23756-5 (Print) 978-3-319-23757-2 (Online).
[6] Božić Ž. Fatigue and fracture of multi-site cracks in stiffened panels. Doctoral dissertation;
ACCEPTED MANUSCRIPT

Department of Naval Architecture and Ocean Engineering, Yokohama National University,


Japan, 1997.
[7] Fricke W, Lilienfeld-Toal A, Paetzold H. Fatigue strength investigations of welded details
of stiffened plate structures in steel ships. International Journal of Fatigue 2012; 34 (1):17–
26. DOI: 10.1016/j.ijfatigue.2011.01.021.
[8] Fricke W. Fatigue analysis of welded joints: state of development. Marine Structures 2003;
16:185–200.
[9] Fricke W, Gao L, Paetzold H. Fatigue Assessment of Local Stresses at Fillet Welds Around
Plate Corners. International Journal of Fatigue 2017; 101(2):169-176.
[10] Dexter RJ, Pilarski PJ, Mahmoud HN. Analysis of crack propagation in welded stiffened

T
panels. International Journal of Fatigue 2003;25:1169-1174.

IP
[11] Mahmoud HN, Dexter RJ. Propagation rate of large cracks in stiffened panels under
tension loading. Marine Structures 2005;18:265-288.

CR
[12] Feng GQ, Garbatov Y, Guedes Soares C. Probabilistic model of the growth of correlated
cracks in a stiffened panel, Engineering Fracture Mechanics 2012;84:83–95.
[13] Garbatov Y, Guedes Soares C, Parunov J. Fatigue strength experiments of corroded small
scale steel specimens, International Journal of Fatigue 2014;59:137–144.

US
[14] Ferro P. The local strain energy density approach applied to pre-stressed components
subjected to cyclic load. Fatigue and Fracture of Engineering Materials and Structures.
2014; 37(11):1268-1280.
AN
[15] Ferro P, Berto F, James N M. Asymptotic residual stresses in butt-welded joints under
fatigue loading. Theoretical and Applied Fracture Mechanics. 2016; 83:114–12.
[16] Ferro P, Berto F. Quantification of the influence of residual stresses on fatigue strength of
M

Al-alloy welded joints by means of the local strain density approach. Strength of Materials.
2016; 48(3):426–436.
[17]
ED

Predan J, Pippan R, Gubeljak N. Fatigue crack propagation in threshold regime under


residual stresses, International Journal of Fatigue 2010; 32:1050–1056.
[18] Leitner M, Barsoum Z, Schäfers F. Crack propagation analysis and rehabilitation by HFMI
of pre-fatigued welded structures. Weld World 2016; 60(3): 581–592. DOI:
PT

10.1007/s40194-016-0316.
[19] Hensel J, Nitschke-Pagel T, Dilger K. Engineering model for the quantitative consideration
of residual stresses in fatigue design of welded components. Weld World 2017; 61(5): 997–
CE

1002. DOI: 10.1007/s40194-017-0467-4.


[20] Božić Ž, Schmauder S, Mlikota M and Hummel M. Multiscale fatigue crack growth
modelling for welded stiffened panels. Fatigue Fract. Engng. Mater. Struct. 2014;
AC

37(9):1043-1054.
[21] Swanson Analysis System, Inc. ANSYS User's Manual Revision 11.0, 2009.
[22] Paris P, Erdogan F. A critical analysis of crack propagation laws. Journal of Basic
Engineering 1963;85:528–534.
[23] Elber W. The significance of fatigue crack closure. Damage tolerance in aircraft structures.
ASTM STP 486. American Society for Testing & Materials; 1971. p.230–242.
[24] Schijve J. Some formulas for the crack opening stress level. Engng Fract Mech
1981;14:461–5.
[25] Servetti G, Zhang X. Predicting fatigue crack growth rate in a welded butt joint: the role of
effective R ratio in accounting for residual stress effect. Engng Fract Mech 2009; 76:1589–
602.
ACCEPTED MANUSCRIPT

[26] Božić Ž, Bitunjac V, Semenski D. Interaction Modelling of Multiple Fatigue Cracks in


Stiffened Panels, Transactions of FAMENA 2010, 34(4):11-19.
[27] Croatian Register of Shipping. Rules for the classificationof ships. 2012. Part 25 - Metallic
Materials.
[28] Wawrzynek P A, Ingraffea A R. (): Interactive finite element analysis of fracture processes:
An integrated approach. Theoretical and Applied Fracture Mechanics. 1987; 8(2):137–150.
DOI: 10.1016/0167-8442(87)90007-3
[29] Barsoum RS. On the Use of Isoparametric Finite Elements in Linear Fracture Mechanics.
International Journal for Numerical Methods in Engineering 1976;10:25–37.
[30] Henshell RD, Shaw KG. Crack Tip Finite Elements are Unnecessary. International Journal

T
for Numerical Methods in Engineering 1975;9:495–507.

IP
[31] Bloom J M. An evaluation of a new crack tip element—the distorted 8-node isoparametric
element. International Journal of Fracture. 1975; 11(4):705–707.

CR
[32] Hibbitt H D. Some properties of singular isoparametric elements. Int. J. Numer. Meth.
Engng. 1977; 11(1):180–184. doi:10.1002/nme.1620110117
[33] Anderson TL. Fracture mechanics, fundamentals and applications. 3rd ed. Norway: CRC
Press, Taylor and Francis Group; 2005.

US
[34] D. Broek, The Practical Use of Fracture Mechanics. Kluwer Academic Publishers,
Dordrecht, The Netherlands, 1989.
[35] Bao R, Zhang X, Yahaya NA. Evaluating stress intensity factors due to weld residual
AN
stresses by the weight function and finite element methods. Engineering Fracture
Mechanics 2010;77:2550–2566.
[36] Faulkner D. A review of effective plating for use in the analysis of stiffened plating in
M

bending and compression. J Ship Res 1975;19:1–17.


[37] Glinka G. Effect of residual stresses on fatigue crack growth in steel weldments under
ED

constant and variable amplitude load. In: Fracture mechanics, ASTM STP 677, American
Society for Testing and Materials; 1979. p. 198–214.
[38] Božić Ž, Schmauder S, Mlikota M. Fatigue growth models for multiple long cracks in
plates under cyclic tension based on ΔKI, ΔJ-integral and ΔCTOD parameter. Key
PT

Engineering Materials 2012;488-489:525-528.


[39] Božić Ž, Mlikota M and Schmauder S. Application of the ΔK, ΔJ and ΔCTOD parameters
in fatigue crack growth modelling. Technical Gazette 2011;18(3):459–466.
CE

[40] Božić Ž, Wolf H, Semenski D. Fatigue Growth of Multiple Cracks in Plates under Cyclic
Tension. Transactions of FAMENA 2010;34(1):1 – 12.
[41] Donahue RJ, Clark HM, Atanmo P, Kumble R, McEvily AJ. Crack Opening Displacement
AC

and the Rate of Fatigue Crack Growth. International Journal of Fracture Mechanics
1972;8:209–219.
[42] Tanaka K. Mechanics and Micromechanics of Fatigue Crack Propagation. ASTM STP
1020, American Society for Testing and Materials, Philadelphia, PA, 1989, pp. 151–183.
[43] Liu Y, Mahadevan S. Threshold stress intensity factor and crack growth rate prediction
under mixed-mode loading. Engineering Fracture Mechanics 2007;74:332–345.
[44] Lanciotti, A., Polese, C., Fatigue crack propagation of through cracks in thin sheets under
combined traction and bending stresses, Fatigue & Fracture of Engineering Materials &
Structures, 2003;26:421-428.
ACCEPTED MANUSCRIPT

Highlights
- A method for predicting fatigue crack propagation in welded stiffened panels accounting
for the effects of residual stresses was introduced. Well known power law models, Elber
and Schijve, were employed to simulate fatigue life for the stiffened panel specimen.
- Stress intensity factor values (SIF) were obtained in finite element (FE) analyses by a
linear superposition of the SIF values due to the applied load and due to weld residual
stresses.
- The FE models included idealized, rectangular or triangular, residual stress distribution
profiles.

T
- The effect of welding residual stresses on the crack propagation rate is taken into account
by replacing the nominal stress ratio R in the power laws by the effective stress intensity

IP
factor ratio Reff.

CR
US
AN
M
ED
PT
CE
AC

You might also like