You are on page 1of 11

Proceedings of the ASME 2010 Pressure Vessels & Piping Division / K-PVP Conference

PVP2010
July 18-22, 2010, Bellevue, Washington, USA

PVP2010-25169
DRAFT

INFLUENCE OF STRENGTH MISMACHT AND WELD RESIDUAL STRESS ON


FATIGUE CRACK GROWTH BEHAVIOR FOR PIPELINE STEEL INCLUDING
CLOSURE EFFECTS

Diego Felipe Sarzosa Burgos Claudio Ruggieri


Department of Naval Architecture and Ocean Department of Naval Architecture and Ocean
Engineering, University of São Paulo, São Paulo, Engineering, University of São Paulo, São Paulo,
SP 05508−900, Brazil SP 05508−900, Brazil
Email:dsarzosa@gmail.com Email:claudio.ruggieri@usp.br

ABSTRACT joints have detrimental effects on fatigue crack propagation


The integrity of mechanical components, particularly when rate. The fatigue life can be reduced by more than 70% for a
they undergo significant fatigue damage for the duration of condition of 50% overmatch when compared with the
operating life, can be strongly influenced by the presence of evenmatch condition.
residual stress fields and mechanical heterogeneity. Premature
closure of crack flanks greatly influences fatigue crack growth 1. INTRODUCTION
rate. To estimate the crack closure in any welded structure, the
residual stresses and strength mismatch in the vicinity of the Fatigue assessments of steel weldments remain a key
crack tip should be considered. Extensive elastic-plastic finite issue in design, fabrication and safe operation of critical
element analyses have been carried out to investigate detailed engineering structures including pressure vessels, storage tanks
crack closure behavior in heterogeneous compact tension (CT) and piping systems among others. Typical welding processes
specimens with three levels of weld strength mismatch and (e.g., hot or cold cracking, lack of penetration, undercut) almost
imposed uniform tensile residual stress field. The restriction on invariably introduce crack-like defects in the welded region,
uncraked ligament imposed by E-647 is rather unconservative most often in the form of planar flaws, which substantially
because it does not ensure linear elastic behavior at the crack reduce the fatigue strength of the welded joint.
tip. A relationship between the crack opening loads, the
mismatch level and maximum applied stress intensity factor Current design and in-service fatigue assessments for
was obtained for small scale yielding (SSY) condition. This welded components in pressure vessels and piping systems,
equation shall be used for fast estimations of closure effects for including the previsions given in the ASME code [1], employ
welding joints. A homogeneous, soft material has the largest conventional procedures adopting stress-life approaches in
crack opening loads, while a heterogeneous material with 50% which fatigue life predictions are made based upon an implicit
overmatch conditions has the smallest opening load under SSY crack initiation criterion and the simple knowledge of nominal
condition. Residual tensile stresses have detrimental effects on stresses acting on the weldment. While used effectively in
the fatigue resistance of the material. On average, residual many structural applications, it is now generally recognized
tensile stresses, with magnitude equal to 0.5σy of the base that crack propagation from pre-existing defects dominates the
metal, increase the Fatigue Crack Growth Rate (FCGR) by fatigue life of welded joints. Moreover, there is a complex
40% when compared to the case without residual stress under interplay of key factors affecting fatigue crack growth rates in
SSY condition. Moreover, overmatch conditions in welded welds, such as weld geometry, strength mismatch and weld

1 Copyright © 2010 by ASME


residual stresses, which is not explicitly taken into account in such as a high pressure pipeline or vessel, can be planned based
conventional stress-life methodologies. Consequently, on the knowledge of the fatigue damage evolution
advanced procedures for fatigue assessments of critical
weldments must include the effects of crack extension under Evaluation of the safe service life for structures containing
cyclic loading on fatigue life predictions for pressurized cracks undergoing cyclic loading has been usually conducted
components. Moreover, accurate estimates of the effective based on semi-empirical relationships connecting key
crack-tip driving force (here characterized by the stress parameters characterized by the crack growth rate da/dN, stress
intensity factor (SIF) range, as measured by ∆K ) with crack intensity factor range ∆K and R-ratio, in addition to some
extension become central in robust correlations of fatigue material properties whose values are commonly determined by
behavior between these conventional test specimens and experimental measurements. The standard approach to predict
structures. fatigue crack growth adopts the Paris law [4] to assess the
crack growth rate per cycle of load (da/dN). The rate of crack
Since the seminal work of Elber [2, 3] on the closure growth is assumed to be a function of the effective stress
mechanism, a considerable amount of investigation has been intensity factor ( ∆K eff ) which is defined as the difference
focused on incorporating the effects of crack closure on fatigue between the maximum stress intensity factor, K max and the
assessment procedures. Most of the recent research specifically
addresses numerical models to simulate plasticity-induced opening stress intensity factor, K op . Another more refined
fatigue crack closure, most of them based on a 2D framework approach that successfully models the R dependence is the
(either plane-strain or plane-stress conditions). While these Walker model [4]. Both approaches will be used to estimate the
previous research initiatives represent major advancements in fatigue crack growth rate.
describing the crack closure mechanism for fatigue life
evaluation scheme, full understanding of fatigue crack Regarding the treatment of the influence of residual
propagation to support development of robust fatigue stresses in the calculation of FCGR there are two typical
assessment procedures is still lacking. In particular, rather little methodologies applied to model the effects of residual stress
further effort has been expended in the systematic investigation fields on fatigue crack growth. The first one employs the
of the potential strong effects of weld strength mismatch and principle of superposition. The stress intensity factor associated
residual stresses on the crack closure behavior and their with the residual stress ( K res ) is added to the stress intensity
interconnection with fatigue life predictions. factor due to the applied external cyclic load ( K L ). Thus, a
resultant stress intensity factor is obtained for the maximum
As a step in this direction, this work provides an
and minimum loads. This procedure will result in a range of
investigation of the coupled effects of strength mismatch and
stress intensity factor SIF, ∆K , unaffected by the residual
weld residual stresses on Fatigue Crack Growth Rate (FCGR)
stress. However, the R-ratio will change as the crack
behavior for pipeline steel weldments. One purpose of this
propagates through the residual field. This will affect the mean
study is to develop a framework while, at the same time,
gaining additional understanding of key factors affecting the stress thereby resulting in an effective stress ratio ( R e ff ). The
fatigue life of structural welds. Another purpose is to assess the second approach employs the effective stress intensity factor
potential strong effects of plasticity-induced crack closure on range, ∆K eff = K max − K op , where K op is the crack opening
fatigue crack extension of weldments. Very detailed non-linear stress intensity factor. This procedure is known as the crack
finite element analyses using plane-strain models of compact closure approach.
tension (CT) fracture specimens with center cracked, square
groove welds provide the evolution of crack growth with cyclic Both methods have been successfully applied to predict
stress intensity factor which is required for the estimation of the fatigue crack growth rates through uniform tensile and
closure forces. This exploratory investigation provides further compressive residual stresses fields [8]. The equivalence of
understanding on the effects of crack closure of fatigue both methods is due to the fact that the behavior of the crack tip
propagation behavior for mismatched welds. opening stress is distinctively controlled by R e ff for a specified
K max , and constant tensile and compressive residual stress
2. FATIGUE CRACK GROWTH MODELING fields [7]. Many other researchers have applied both methods
successfully in the linear elastic regime [[6, [8, [9].
The finite element method has been increasingly used as a
very effective tool to model fatigue crack growth aiming at An important issue related to fatigue crack propagation
estimating the number of cycles for a crack to grow from a through residual stress fields is the redistribution and possible
certain initial length to a maximum tolerable (critical) crack relaxation of stresses. In the low-cycle fatigue regime the high
size. Thus, a safe inspection interval for a specific structure, level of plastic deformation achieved at the crack tip causes the
relaxation of the initial residual stresses such that they are

2 Copyright © 2010 by ASME


virtually equal to zero. In the medium-cycle to high-cycle There are two common criteria used to define the opening
fatigue regime, which does not involve excessive cyclic plastic and closure of the crack. The first criterion is based on node
deformation at the crack tip, the relaxation almost does not contact and assumes that the crack opens when its free surface
exist. Thus, a high residual stress is usually said to be stable for loses contact as described by the nodal displacement value:
low applied fatigue loads. However, as the crack tip advances positive nodal displacements indicate opening of the crack
the initial residual stresses get redistributed by the presence of a whereas negative nodal displacements indicates closure of the
new surface. Here, this redistribution will be neglected as a first crack. Another criterion which is adopted in the present work
approximation of this phenomenon. How this redistribution establishes that the crack opens when the (opening) stress state
affects the fatigue crack propagation is an open issue and ahead of crack tip, which is compression in the previous cycle,
subject of continuous research [9]. changes from a state of compression to tension. Similarly, crack
closure is defined when the crack tip opening stress changes
Several researchers have done important studies about from tension to compression. Such criterion appears more
plasticity-induced crack closure phenomena using finite realistic because the first one is strongly dependent on the mesh
element method [12, [14, [15]. The basic algorithm involves size, since the node behind the advancing crack tip can
building a refined enough mesh for the loads to be applied; it experience a positive (negative) displacement but the crack tip
means enough elements on the crack tip in order to simulate the node would still carry compressive (tensile) stresses [14].
effects of forward and reverse plastic zone. Remote loads are
applied cyclically, this is time-dependent, and the crack tip is 1400

released at an arbitrary point on the cycle, either during the Nodal


1200 release
loading or unloading portion. This is repeated for each applied
cycle. 1000 Relaxation of nodal forces
Remote Load P [ N ] Accurate
The growing crack creates a plastic wake which is 800 determination
of opening load
responsible for the premature contact of the faces of the crack,
even for loads greater than zero as shown in Fig. 1. The process 600

is repeated until stable values of K op are obtained. 400

In this study a constant cyclic loading was imposed on the 200

model under load control conditions. The cyclic loading Crack face contact
0
consisted of repeatedly increasing the values of applied load 0 10 20 30 40 50 60 70 80 90
from zero to the maximum load and then decreasing back to # Steps

zero as shown in Fig. 2. The crack propagates uniformly by an


amount equal to ∆a =0.02mm in each cycle. This is achieved Figure 2. Variable incremental load applied by each cycle.
by releasing all (current) crack front nodes after the maximum
load is applied. Here the amount of crack growth was set equal In the present analysis the contact between crack flanks was
to the size of plastic zone, rp , for a plane strain condition modeled using a rigid surface at the crack symmetry plane in
calculated as a second order estimation using Irwin’s approach. order to prevent crack surface penetration during unloading.
The contact between a soft surface (crack flanks) with a rigid
surface (symmetry plane) was considered. For more detailed
information on the numerical modeling of crack closure used
here the reader is remitted to [10].

2.1.Incorporation of residual stress field.

Residual stresses fields have been handled quantitatively


with crack closure models or by applying the principle of
superposition. Both methods are described in the following
sections itens.

2.1.1 Effective stress ratio approach


Figure 1. Ilustration of plastic wake left by fatigue crack
growing It is known that residual tensile stresses increase the mean
stress and reduce the weldment fatigue strength. Compressive
residual stresses have an opposite effect, that is, they reduce the

3 Copyright © 2010 by ASME


mean stress and increase the weldment fatigue strength. A element analyses were performed on plane-strain models. The
quantitative assessment of the influence of residual stresses can research code WARP3D [13] was employed. WARP3D only
be made by the principle of superposition: supports 3D solid elements, so the plane strain condition is
K T = K L + K res (1) simulated by restraining the displacement of all nodes in the
thickness direction (w=0). Figure 3 shows the coordinated
where K L corresponds to the SIF due the applied remote cyclic
system employed, the geometry and specimen dimensions for
load and K res is the SIF due to residual stress. The SIF for the the analyzed crack configurations.
residual stress is usually obtained by means of weight function
solution by applying the residual stress field on the uncracked Two types of mesh were employed to investigate the
component. The weight function, m(x, a), depends only on the effects of the mesh refinement on the values of the opening
geometry of the component. The relationship to obtain the K res loads. The level of refinement at the crack tip is the same for
using m(x,a) is: both models but the mesh at the interface – base metal-weld
a metal – is different as can be seen in Figure 4. Symmetry
K res = ∫ σres (x) × m(x, a)dx ( 2) conditions permit modeling of only one-half of the specimen
0 with appropriate constraints imposed on the remaining ligament
Here, a is the crack size. The weight function employed here v=0). Thus, the number of nodal degree of freedom is reduced
was taken from [5]. The integration was made in two domains: and consequently the computational time is decreased. One
first in [0, 0.95a] employing a trapezoidal rule and second in layer was used to define the thickness of the specimens. A
[0.95a, a] using a Gauss’s rule with three points. The sequence of small and identical elements ahead of crack front
incorporation of residual stresses using the effective stress permits an equal amount of crack growth, as shown in Figure 4.
approach only changes the local stress ratio ( Reff ). This ratio
will change continuously with the increasing of crack length as A typical half-symmetric model has 1930 8-node, 3-D
shown below: elements (4178 nodes). The element material properties around
the loading point are defined as an elastic material to enhance
KT K L + K res
R e ff = Tmin = Lmin ( 3) numerical converge.
K max K max + K res

This effect will cause a mean stress modification at the crack


tip. y y
BM
2.1.2. Effective stress intensity factor range approach
WM
The residual near-tip strains associated with the cyclic x
loading form a plastic wake which is responsible for the BM
z
premature closure of the crack over some distance. This
phenomenon was first discovered by Elber [2] and may
strongly affect the effective load increment and reduce the
crack growth rate. The crack will be able to propagate only if
the residual stresses are overcoming to a degree that the crack
Figure 3: Schematic view of (CT) specimen geometry
tip is opened again [9]. When the closure phenomena are taken
employed in the analyses.
into consideration the driving force is reduced and effective
stress intensity factor range is defined as:
3.2. Computational procedures
∆K eff = ( K max + K res ) − ( K OP + K res ) = K max − K OP (4)
The finite element code WARP3D provides the numerical
solution for the plane-strain analyses reported here. The code
where K op is the crack opening stress intensity factor of the incorporates Mises (J2) constitutive model in both small-strain
structure affected by a residual stress field. and finite-strain framework. The global solution procedure uses
an implicit, incremental–iterative strategy with Newton
3. NUMERICAL PROCEDURE iterations to achieved equilibrium at each load increment. Key
features of the code employed of the code in this work include:
3.1. Finite element models (1) node release scheme, (2) contact procedure, and (3)
automatic nodal release forces. Each load cycle was set with 50
Numerical models for a conventional compact tension (CT) variable loads increments.
1T fracture specimen have been developed. Detailed finite

4 Copyright © 2010 by ASME


3.3. Material constitutive model 1.5, respectively, where the mismatch level is defined by
M y = σ WM
ys σ BM
ys with superscripts WM and BM denoting weld
The analyses use an incremental, isotropic hardening metal and base metal). The adopted yield stress and hardening
constitutive model to describe the cyclic, elastic-plastic properties of the baseplate material are the same in all analyses
response of the material in small geometry change setting. The
numerical solutions employ a simple power-hardening model to and assigned the following properties: n=12.4 and σ ys = 483
characterize the uniaxial true stress-logarithmic strain in the MPa.
form: 800
n
ε σ ε ⎛σ ⎞ 700
= ε ≤ εy ; =⎜ ⎟ ε > ε y (5)
εy σy ε y ⎜⎝ σ y ⎟⎠ 600

where σy and εy are the yield stress and strain, and n is the 500

σ [MPa]
True stress-strain
strain hardening exponent. The finite element analyses consider 400 Ramberg - Osgood
Engineering stress-strain
material flow properties of typical pipeline steel API 5L grade
300
X60 (20º C). The stress-strain response for this material is API 5L - X60 (20º C)
shown in Fig. 5. σy=483 MPa
200
E=206 GPa
n=0.3
100

0
0 0,05 0,1 0,15 0,2
ε
Mesh 1
Figure 5. Stress-strain response for a steel API 5L- X60 [14].

Base Metal Table 1 provides the material properties utilized in the


numerical analyses of these fracture specimens with square
groove welds which also consider E=206 GPa and ν = 0.3.
Y
h=7 mm

Weld Metal The strain hardening parameters for the weld metal are
Z X
estimated from a simple correlation between the yield stress
and hardening exponent applicable for typical structural steels:
n=5 and E / σ y = 800 (high hardening material), n=10 and
E / σ y = 500 (moderate hardening material), n=20 and
Mesh 2 E / σ y = 300 (low hardening material). These ranges of
properties also reflect the upward trend in yield stress with the
increase in strain hardening exponent characteristic of ferritic
materials, including pipeline steels.
Y
h=7 mm

Table 1: Material properties adopted in the analyses


Z X

Figure 4. Details of the mesh used in the analysis.The crack tip


area has been enlarged to ilustrate the mesh around the crack 4. RESULTS AND DISCUSSION
tip
4.1 Mesh refinement effects
The finite element analyses consider material flow
properties covering a wide range of moderate-to-high strength The crack opening values for the condition of 20%
mismatch: evenmatch, 20% and 50% overmatch ( M y =1.2 and overmatch, K max =25 MPa m and ratio a/W=0.635, are shown

5 Copyright © 2010 by ASME


in Figure 6. These values are presented as a function of the value of the applied force to open crack flanks. The general
normalized crack length normalized by the radius of the plastic trend of lower K op values with increasing constraint should be
zone. The plot indicates that there is no significant difference noted. Relative “steady” values can be selected when the crack
between the two mesh refinements. This result suggests that a
has grown one size of the plastic zone rp . On the other hand,
correct modeling of the crack tip region is sufficient for
estimating reliable values of K op . Therefore, the Mesh 1 was for a/W=0.5 the evenmatch condition has the lowest values of
K op and the highest values correspond to the condition of 50%
selected to perform all analyses in this study.
0,50 of overmatch. The situation of 20% overmatch is again in the
mesh2 middle of the two previous conditions. This behavior is in
mesh 1 contrast to that observed previously for a/W=0.25. This
0,40
behavior can be explained by the loss of elastic constraint at the
crack tip. The size requirements, ( W − a ) ≥ 2.5 ( K I × σ −y1 ) ,
2
0,30
Kop / Kmax

Kmax [MPa-m^0,5] = 25
CTspecime
according to ASTM to ensure that the nominal behavior is
0,20 a/W= 0.635 W=50.8 mm linear elastic and the SIF at the crack tip, K, characterizing the
R=0.0 ; rp=0.27 mm
2h= 14 mm
crack tip conditions were calculated and for the case a/W=0.5,
0,10 20% Overmatch it was determined that these conditions were not satisfied.
0,50

0,00 0,45
0% 50% 100% 150% 200% 250% 300% 0,40
(ai-ao)/rp [mm/mm]
0,35
Figure 6. Crack opening loads for the two types of mesh
0,30
Kop / Kmax
refinement adopted in the analyses.
0,25
Pmax [N] = 1230
The effect of weld width (2h) was analyzed. Two values of 0,20 CT Specimen EvenMatch
the width, 2h=7.5mm and 2h=14mm, were considered. 0,15 Plane Strain
20% OverMatch
a/W= 0.25 W=50.8 mm
Basically, both specimens display essentially similar values of 50% OverMatch
0,10 R=0
K op / K max with a little difference when the crack has grown to 2h= 14 mm
0,05
1.25rp. As expected the opening behavior only depends of the 0,00
crack wake left by the growing crack. The weld width does not 0% 20% 40% 60% 80% 100% 120%
affect the values of Kop, under SSY conditions. (ai-ao)/r p [mm/mm]

0,50
The first effort in this study was to understand the crack
0,45
closure behavior for welded specimens with three levels of
0,40
strength mismatch. Three models were employed in the
0,35
analyses: one model with evenmatch condition, and the
0,30
Kop / Kmax

remaining two models with 20% and 50% overmatch levels.


0,25 Pmax [N] = 1230
The loading corresponded to R=0 and maximum load, CT Specimen
Pmax =1230 N. The results of the two extreme conditions were 0,20 EvenMatch
Plane Strain
0,15 a/W= 0.5 W=50.8 mm 20% OverMatch
analyzed. The corresponding crack lengths, in terms of the R=0 50% OverMatch
0,10
ratio,a/W , were equal to 0.25 and 0.5, respectively. The 2h= 14 mm
loading configuration satisfied the requirement of ASTM E-647 0,05

for the uncraked ligament of a compact specimen, given by 0,00


0% 20% 40% 60% 80% 100% 120%
(W − a) ≥ 4× π −1
(K I ×σ Y )
−1 2
. Theoretically, this condition (ai-ao)/r p [mm/mm]

guarantees elastic behavior of the specimen. The evolution of Figure 7: Values of K op / K max for two cracks lengths and three
crack opening loads with the amount of crack growth is shown
levels of strength mismatch.
in Figure 7.
Figure 8 provides further support to the above argument.
For a/W=0.25 the evenmatch condition has the highest
The plastic zones for the evenmatch and 50% overmatch
values of K op and the lowest values correspond to the 50% conditions are shown in the figure. The plastic zones shown
overmatch condition. The K op values for 20% of mismatch are correspond to the minimum load of the applied cyclic loading.
closest to evenmatch configuration. As expected: the greater the
levels of plastic deformation at the crack tip the greater the

6 Copyright © 2010 by ASME


Only under overmatch conditions the crack tip has enough obtained using such an equation will be conservative and useful
elastic constraint to close the crack flanks. Therefore, it can be for fatigue life assessments for design/maintenance.
concluded that the restriction on uncraked ligament imposed by
E-647 is not conservative because it does not ensure linear 15 20 25 30 35 40
0,5 0,5
elastic behavior at the crack tip. Besides, plasticity-induced
0,45 0,45
crack closure becomes less significant as net section plasticity
increases and elastic constraint is lost. 0,4 0,4

0,35 0,35

In related article, Sarzosa and Ruggieri [10] verified that the 0,3 0,3

Kop/Kmax
values of K op are independent of defect size for constant value 0,25 0,25

of ∆K applied under SSY conditions, corresponding to a test 0,2 0,2


with a decreased applied cyclic load. In view of this, a high 0,15 0,15
constraint specimen, a/W>0.6, was employed to calculate the
0,1 50% OverMatch C(T)-1T Specimen 0,1
intermediary values of K op / K max for 0.25<a/W<0.5. This 20% OverMatch Plane Strain
0,05 EvenMatch R=0 2h= 14 mm 0,05
guaranteed SSY conditions and ensured that high plasticity at
0 0
the crack tip was avoided. 15 20 25 30 35 40
Kmax [MPa m^0.5]

Figure 9: Ratio K op / K max as function of maximum applied


stress intensity factor for R=0.

A non-linear regression was employed to obtain a


multidimensional fitting equation. The model employed uses a
second order equation on M y and K max . The final expression of
the equation is written as:
K op
= P1 + P2 M y + P3 M 2y (6)
K max
Where
(
P1 = 3.7655 − 0.213K max + 0.0032K 2max )
P2 = ( −7.2896 + 0.5006K max − 0.0081K 2
max )
P3 = ( 2.8456 − 0.2114Kmax + 0.0036K 2
max )
The previous equation is valid under SSY conditions.

3.4.Coupled effects of residual stress and strength


mismatch
Figure 8: Plastic zones for two levels of strength mismatch,
a/W=0.5. For two levels of tensile residual stress (0.5 σ yBM and
1.0 σ yBM ), the effective stress ratio, Reff , was calculated and
A curve for K op / K max versus K max , for the opening values
determined to be equal to 0.1202 and 0.215, respectively. The
when the crack has grown to one rp size, is shown in Figure 9.
finite elements models were run using these values. The
For low stress levels where the crack closure phenomena is evolution of opening loads for both cases is shown in Figure 10
more important, a high degree of parallelism between all the for two crack lengths.
curves can be seen. This behavior should be further
investigated to estimate the opening loads for a given level of Analytical work made by Sehitoglu [16] and recently by
strength mismatch with reference to the condition of evenmatch Codrington and Kotousov [19] showed increase in the values of
for example. An equation for the ratio K op / K max as a function K op / K max with increase in the applied stress ratio R, for R>0. It
of K max and the level of mismatch has been developed. As the is expected that for high R values K op / K max , will be equal to
analyses were made for plane-strain situation the values R, which implies that K op = K min and there are no closure

7 Copyright © 2010 by ASME


effects. As the level of mean stress has been increased the These values have been used to evaluate the FCGR using
relative separation of curves is reduced. This effect is more the Paris law and ∆K eff as the driving force.
pronounced for higher values of applied residual stress. For
a/W=0.5, the K op / K max ratio between evenmatch and 20% The variation of K op / K max with R eff is similar to that seen
overmatch and evenmatch and 50% overmatch differs by 3.7% in the previous case analyzed, with no residual stresses. For
and 9.7%, respectively, for σ res =241 and 3.2% and 6.4% for a/W=0.25, the highest value of K op / K max corresponds to the
σ res =483 MPa, when the crack has grown one rp . For higher evenmatch condition and the lowest value to 50% overmatch.
residual tensile stresses it is expected that there will be no In contrast the trend is reversed for a/W=0.5. The highest value
difference between the levels of opening loads for all levels of corresponds to the 50% overmatch condition and the lowest to
strength mismatch under SSY condition. the evenmatch.

0,60 fillet_symbol: σres=241 MPa x [mm]


0 0,05 50%MS 0,1 0,15
unfillet_symbol: σres=483 MPa
4,0E-03 4,0E-03
0,50 + bottom
opening
3,5E-03 Kmax [MPa-m^0,5] = 52.7 3,5E-03
a/W= 0.5 top
Σ ai =0.185rp
0,40 3,0E-03 R=0.2146 (σres=412 Mpa) 3,0E-03
2h= 14 mm
Kop / Kmax

2,5E-03 2,5E-03
0,30

Evenmatch

50%MS
v [mm]
v [mm]
EvenMatch 2,0E-03 2,0E-03
Kmax [MPa-m^0,5] = 52.8
0,20 CT specimen 20% OverMatch
a/W= 0.5 W=50.8 mm 50% OverMatch 1,5E-03 1,5E-03
2h= 14 mm
0,10 1,0E-03 1,0E-03
Base Metal: n=12.4 ; σy=483MPa

5,0E-04 5,0E-04
0,00
0% 20% 40% 60% 80% 100% 120%
0,0E+00 0,0E+00
0 0,1 0,2 0,3
(ai-ao)/rp [mm/mm]
x [mm]
0,60 fillet_symbol: σres=241 MPa Evenmatch
unfillet_symbol: σres =483 MPa
Figure 11: Crack tip profiles when the crack has grown 0.18rp
0,50 for evenmatch (close symbols) and 50% overmatch (open
symbols).
0,40
Kop / Kmax

0,30 0,60
fillet_symbol: a/W=0.5
unfillet_symbol: a/W=0.25
Kmax [MPa-m^0,5] = 26.8 0,50
0,20 CT specimen EvenMatch
a/W= 0.25 W=50.8 mm 20% OverMatch
2h= 14 mm 50% OverMatch 0,40
0,10 Base Metal: n=12.4 ; σy =483MPa
Kop / Kmax

0,30
Pmax [N] = 1230
0,00 CT Specimen EvenMatch
0% 20% 40% 60% 80% 100% 120% 0,20 Plane Strain 20% OverMatch
(ai-ao)/r p [mm/mm] W=50.8 mm 50% OverMatch
Σai=rp 2h= 14 mm
0,10
Figure 10: Evolution of K op / K max as function of the crack
length for R=0.12 ( σ yBM =241MPa) and R=0.214 0,00
0,1 0,12 0,14 0,16 0,18 0,2 0,22 0,24
(σ BM
y =483MPa). Reffective

Figure 12: K op / K max as function of the effective R for two


Figure 11 shows the behavior of the crack profile during the crack lengths.
cyclic loading. Upon loading the crack opens from the crack-
mouth side to the tip continuously. There is no partial opening
behavior and the typical K op definition can be used to predict 3.4. Fatigue crack growth evaluation and fatigue life
predictions.
FCGR as propose in [7]. When the crack has grown one rp only
a “small” region next to the crack actual tip is closed. For the case R=0, the effective driving force and the rate
da/dN as a function of applied remote load is shown for each
Figure 12 displays the variation of the ratio K op / K max as a level of mismatch analyzed in Table 2.
function of the R eff .

8 Copyright © 2010 by ASME


Table 2: (a) Values of ∆K eff as a function of the ∆K , and (b) Table 3: values of ∆K eff as a function of the ∆K and da/dN for
da/dN as a function of ∆K . R=0.1202(a), and R=0.2146 (b)
(a) (a)
evenmatch 20%over 50%over R=0,12024
∆K ∆Keff ∆Keff = ∆Keff = Keff [MPa m^0,5]
K 0% 20% 50% Walker
20 13,44 14,92 17,00
27 17,03 17,82 20,90 Equation
25 14,50 15,81 18,19
53 36,20 35,31 33,90
27 15,19 15,99 18,41
da/dN
29 16,10 16,82 18,91 27 1,777E-05 2,036E-05 3,286E-05 8,473E-05
35 20,79 20,55 20,86 53 1,707E-04 1,585E-04 1,402E-04 6,422E-04
53 37,91 36,04 33,79
(b)
(b) R=0,2146
evenmatch 20%over 50%over ⎝ Keff [MPa m^0,5]
∆K da/dN da/dN da/dN ⎝K 0% 20% 50% Walker
20 8,73E-06 1,195E-05 1,769E-05 27 17,810 19,779 19,895 Equation
25 1,098E-05 1,423E-05 2,166E-05 53 35,603 34,595 33,587
27 1,261E-05 1,473E-05 2,247E-05 da/dN
29 1,5E-05 1,713E-05 2,434E-05 27 2,034E-05 2,785E-05 2,835E-05 6,99E-05
53 1,625E-04 1,491E-04 1,364E-04 5,299E-04
35 3,235E-05 3,122E-05 3,268E-05
53 1,961E-04 1,685E-04 1,389E-04
Fatigue life predictions (FLP) have been performed for a
For lower values of the applied load corresponding to range of applied cyclic loads. Since only two values of opening
situations where the closure phenomena is more relevant, loads, K max =26.8MPa m and K max =52.8 MPa m , for the
considerable reduction of the rate da/dN upto 102% for the condition including residual stress fields were calculated the
case of evenmatch when as compared to 50% of overmatch FLP were performed only for the condition of no residual stress
condition and 37% as compared to 20% of overmatch can be fields.
seen. The above values are for an applied ∆K =20MPa m .
As the level of the applied load increases, the differences in Since ∆K increases with crack length during constant
da/dN rates is reduced until a minimum value is reached at amplitude loading the growth rate is not constant so it is
necessary to use an integration procedure to obtain the life
∆K =35 MPa m , under similar driving forces for each case
required for crack growth. The procedure employed to perform
of mismatch. For higher applied loads the elastic constraint at
the integration of FCGR was taken from Dowling [4].
the crack tip is lost for the case of evenmatch and 20%
Basically, the life in cycles is determined by the integration of
overmatch resulting in higher ratios of da/dN than 50%
the inverse of FCGR from an initial size ai (taken in all cases as
overmatch where the level of triaxiality is preserved and the
1 mm) to a final size af (taken as the smallest value between a
closure behavior is controlled by the elastic elements
critical crack size that produce plastic collapse or brittle
surrounding the crack tip material.
fracture). The inverse of FCGR, dN/da, represents the rate of
accumulation of cycles per unit increase in crack length [4].
For the case with R=0.1202 and R=0.2146, the effective
driving force and the rate da/dN as a function of applied remote
The numerical results are compared with two theoretical
load are shown for each level of mismatch analyzed in Table 3.
solutions presented by Codrington and Kotousov [19] and
Newman [20], Model 1 and Model 2 respectively, which
The da/dN rate evaluated using the Walker’s model is
employed the strip yield model to calculate the opening loads.
presented below. As the Walker’s model does not take into
As these models do not include the mismatch effects only
account the level of heterogeneity, direct comparisons can be
comparison with the condition of evenmatch will be possible.
made only in evenmatch case. The effective stress intensity
Newman’s equation for calculation of opening loads has been
factor approach gives da/dN rates that are 70% lower,
employed extensively by industry and academia. Further,
approximately, than those provided by Walker’s model. On
Newman’s equation is implemented in FASTRAN software
average the residual tensile stress, for R=0.1202, increase the
principal numerical code used to predict fatigue life for
FCGR by 40% for all levels of strength mismatch as compared
constant and variable cyclic load so that the employed
to the case without residual stress for cases under SSY
methodology can be evaluated with direct comparison with this
condition. For the case R=0.2146, on average the FCGR is
theoretical solution.
increased by 58% as compared to the case without residual
stress (R=0).
Figure 15 shows dN/da and ΣN (accumulation of cycles
with crack growing) curves obtained by the finite element
method and the two theoretical solutions previously mentioned,

9 Copyright © 2010 by ASME


for a constant amplitude load ∆P =1000 N. For this level of and 2.41 for the FEA, Model 1 and Model 2, respectively, are
applied load the Finite Element Approximation (FEA) and obtained.
Newman’s model of the effective driving force resulted in FLP
nearly the same (difference less than 1%). Codrington and 1
Kutuzov’s model produces fatigue life estimation 15.3% higher 0,9
than FE model. The FCGR was expressed as a function of an

U= (Kmax-Kop)/(Kmax-Kmin)
0,8
K max − K op
effective stress intensity ratio ( U = ) as proposed 0,7
K max − K min 0,6
by Elber [2]. The variation between the FLP showed in Figure 0,5
13 lies in the different values of U estimated for each model.
0,4 FE
This can be shown in Figure 14. The finite element results of U C(T)-1T Specimen
0,3
Plane Strain Model 1
were fitting by a quadratic function. As the fatigue life is R=0 2h= 14 mm Model 2
Base Metal: n=12.4 ; σy=483MPa
mostly consumed during the initial crack growth, see the area 0,2
under dN/da curve in Figure 13, the values of U that has more 0,1
impact in the fatigue estimations correspond to low K max . A 0
balance between the higher values of U, for the FEA, at the 5 15 25 35 45 55 65 75 85 95
beginning of the applied cyclic load than the Newman’s model Kmax [MPa m^0.5]
and the lower values of U, for FEA after some crack growth Figure 14: Variation of the effective stress intensity ratio
( ∼ 7mm ), than the Newman’s model produces fatigue life with K max for a constant amplitude loading ∆P =1000kN.
estimation almost equal between these two models.

Finite element model Analytical Model 1 Analytical Model 2 Evenmatch


6,0E+05 2,5E+06 1,6
20% Overmatch
∆P=1000 N 50% Overmatch
5,5E+05 R=0 1,4 Without Closure
Analytical Model 1
5,0E+05 Analytical Model 2
2,0E+06 1,2
4,5E+05

4,0E+05
+ 1
∆P [kN]
dN/da [cyles/mm]

x 1,5E+06 0,8
Σ N [cycles]

3,5E+05

3,0E+05 O 0,6
Fatigue Crack Propagation Life
CT specimen
ai =1 mm
2,5E+05 1,0E+06 0,4 af=min(plastic collapse, brittle fracture)
2,0E+05
W=50.8 mm
0,2 2h= 14 mm ; R=0
1,5E+05 Base Metal: n=12.4 ; σy=483MPa
5,0E+05 0
1,0E+05 1E+05 1E+06 1E+07
5,0E+04 Nf [cycles]

0,0E+00 0,0E+00 Figure 15: Comparisons of fatigue life estimations for constant
1 6 11 16 21 26 31 36
aj [mm]
amplitude loading.
Figure 13: Comparison of curves dN/da (left ordinates) and Although the FLP for the conditions 20% and 50%
ΣN (right ordinates) vs. crack length for applied constant cycle overmatch have the same tendency as the case evenmatch,
load obtained by finite element method and two theoretical appreciable reduction in the life can be shown in Figure 15. In
solutions. average, for the range ∆P ≤ 1100 N , the decreases in FLP are
equal to 37% and 73% for the 20% overmatch and 50%
Figure 15 shows fatigue life estimations for several values
overmatch, respectively, when compared to evenmatch case.
of the applied cyclic constant load. Besides, the curve of
fatigue life without taking in consideration the closure effect
4. CONCLUSIONS
has been included as reference. A good agreement between the
finite element and theoretical results can be seen at the region
A methodology to evaluate the influence of strength
around 850N ≤ ∆P ≤ 1150N for the Model 2, and at the region
mismatch and residual stresses was described. Finite element
∆P ≥ 1150 N for the Model 1 with difference less than 10% in
analysis provides the necessary information to include the
both cases. From a structural point of view, the loads higher crack closure phenomena into the problem. The conclusions
than ∆P ≥ 1100 N produce plastic zones that violate the SSY drawn for this study are as follows:
condition for the initial crack size assumed here. When the FLP
are compared with those values that do not take in account the
closure effects, average Safety Factors (SF) equals to 2.3, 2.82

10 Copyright © 2010 by ASME


• The results indicate that heterogeneity in mechanical [9] Prawoto, Y. “Linear Elastic Fracture Mechanics (LEFM)
properties between the base plate and weld metal has a Analysis of the Effect of Residual Stress on Fatigue”.
great effect on crack opening behavior. The Journal of Failure Analysis and Prevention, Vol 2, number
homogeneous, soft, material has the largest crack 5, October 2002.
opening loads, while the 50% overmatch condition has [10] Sarzosa, B. D. F. and Ruggieri, C. “Numerical Modeling
the smallest one under SSY condition. of Crack Closure Effects on Fatigue Crack Growth
• Under SSY, the trend in the values of K op / K max , Propagation in C(T) Fracture Specimens”. 20th
International Congress of Mechanical Engineering,
condition, showed a high parallelism between each
COBEM, November 15-20, 2009, Gramado, RS, Brazil
level of mismatch, suggesting the possibility of
[11] LaRue, J. and Daniewicz, S. “Predicting the Effect of
determination of an equation relating the opening
Residual Stress on Fatigue Crack Growth”. International
loads as a function of the level mismatch.
Journal of Fatigue, 29, pp. 508-515, 2007.
• Walker’s model provides a more pessimistic approach
[12] McClung, R.C and Sehitoglu, H . “On the finite element
than the effective stress range model.
analysis of fatigue crack closure-1: Basic Modeling Issues”.
• Residual tensile stresses have detrimental effects on
Engineering Fracture Mechanics, Vol. 33, No. 2, pp 237 –
the fatigue resistance of the material.
252 , 1989.
• Overmatch conditions in welded joints have [13]Gullerud,A., Koppenhoefer,K., roy,A., Roychowdury,S.,
detrimental effects on the fatigue crack propagation Walters, /m., /bichon, B., Cochran, K and Dodds, R., “
life. WARP3D: Dynamic Nonlinear Fracture Analysis of Solids
using a parallel computer and workstations”, Structural
5. ACKNOWLEDGMENTS research series (SRS), 607. UILU-ENG-95-2012.
University of /Illinois at Urbana-Champaign.2004
This investigation is supported by Fundação de Amparo à [14] Wu,J., Ellyin, F. “A study of fatigue crack closure by
Pesquisa do Estado de São Paulo (FAPESP) through a elastic-plastic finite element analysis for constant-amplitude
scholarship (08/54130-4) provided to the first author. loading”, International Journal of Fracture 82, 43-65,1996.
[15] Solanki, Kiram. , Daniewicz, S.R, Newman J.C. Jr. “Finite
6. REFERENCES element modeling of plasticity-induced crack closure with
emphasis on geometry and mesh refinement effects”,
[1] American Society of Mechanical Engineers, Boiler and Engineering Fracture Mechanics, 70, pp. 1475-1489,
Pressure Vessel Code, New York, 2004. 2003.b.
[2] Elber,W., Eng. Fract. Mech., 2 (1970) 57. [16] Sehitoglu, Huseyin. “Crack Oepning and Closure in
[3] Elber, W., ASTM STP-486, American Society of Testing Fatigue”. Engineering Fracture Mechanics, Vol. 21, No. 2,
and Materials, Philadelphia PA, 1970, p. 230. pp 329-339 , 1985.
[4] Dowling, Norman. E., “Mechanical behavior of materials: [17] Silva, M.S., “Fracture Toughness and R-curve
Engineering methods for deformation, fracture and fatigue”. Measurements for an API X60 Pipeline Steel Using a Direct
3rd Edition. 2005 Current Potential Technique.” M.Sc. Thesis. Faculty of
[5] Glinka,G. and Shen. G, E. “Universal Features of Weight Engineering (COPPE), Federal University of Rio de
Functions for Cracks in mode I”. Engineering Fracture Janeiro, 2002 (in Portuguese).
Mechanics, Vol. 40, No. 6, pp 1135-1146 , 1991. [18] Silva, M.S., “Fracture Toughness and R-curve
[6] Liljedahl, C., Brouard, J., Zanellato, O., Lin, J., Tan, J., Measurements for an API X60 Pipeline Steel Using a Direct
Ganguly, S., Irving, P., Fitzpatrick, M., Zhang, X and Current Potential Technique.” M.Sc. Thesis. Faculty of
Edwards L. “Weld Residual Effects on Fatigue Crack Engineering (COPPE), Federal University of Rio de
Growth Behaviour of Aluminium Alloy 2024-T3”. First Janeiro, 2002 (in Portuguese).
International conference on Damage Tolerance of Aircraft [19] Codrington, J. and Kotousov, A. “A crack closure model
Structures, R. Benedictus, J. Schijve, R.C Alderliesten, J.J of fatigue crack growth in plates of finite thickness under
Homan (Eds). small-scale yielding conditions.” Mechanics of Materials
[7] Choi, H. and Song J. “Finite Element Analysis of Closure 41 (2009) 165–173.
Behaviour of Fatigue Cracks in Residual Stress Fields”. [20] Newman Jr, J.C. “A crack-opening stress equation for
Fatigue Fract. Engng Matter. Struct. Vol 18, N 1, pp 105- fatigue crack growth,” Int J Fracture, Vol. 24, 1984, R131-
117, 1995. R135
[8] Kang K. J, Song J.H and Earmme Y.Y. “Fatigue Crack
Growth and Closure Behaviour Through Compressive
Residual Stress Field”. Fatigue Fract. Engng Matter. Struct.
Vol 13, N 1, pp 1-13, 1990.

11 Copyright © 2010 by ASME

You might also like