See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/300815967

Fatigue and Reliability of Welded Structures

Conference Paper · April 2007

DOI: 10.4271/2007-01-1657

CITATION READS

1 159

3 authors, including:

G. Glinka Mohamad S El-Zein

University of Waterloo John Deere
172 PUBLICATIONS   4,004 CITATIONS    47 PUBLICATIONS   178 CITATIONS   

SEE PROFILE SEE PROFILE

Some of the authors of this publication are also working on these related projects:

Symposium on Risk analysis and Safety of Technical Systems (ECF22 conference, Serbia) View project

International Symposium on Fatigue Crack Growth – Experimental, Theoretical and Numerical Approach (ISFCG2017) View project

All content following this page was uploaded by G. Glinka on 06 August 2018.

The user has requested enhancement of the downloaded file.

 STRESS ANALYSIS and FATIGUE of welded structures

STRESS ANALYSIS and FATIGUE of welded

structures
A. Chattopadhyay, G. Glinka, M. El-Zein, J. Qian
and R. Formas

Fatigue analyses of weldments require detailed knowledge of the stress fields in critical regions. The stress
ABSTRACT

information is subsequently used for finding high local stresses where fatigue cracks may initiate and for cal-
culating stress intensity factors and fatigue crack growth. The method proposed enables the determination of
the stress concentration and the stress distribution in the weld toe region using a special shell finite element
modelling technique. The procedure consists of a set of rules concerning the development of the finite ele-
ment mesh necessary to capture the bending and membrane structural stresses. The structural stress data
obtained from the shell finite element analysis and relevant stress concentration factors are subsequently
used to determine the peak stress and the non-linear through-thickness stress distributions. The peak stress
at the weld toe is subsequently used for the determination of fatigue crack initiation life. The stress distribution
and the weight function method are used for the determination of stress intensity factors and for the analysis
of subsequent fatigue crack growth.

IIW-Thesaurus keywords: Structural analysis; Stress distribution; Finite element analysis.

1 Introduction
drawback if various multiple loads are applied to the same
weldment or welded structures.

The purpose of the method discussed below is to find such

Fatigue processes originate at stress concentration points, an approach that would require only stress concentration
such as the weld toe in weldments. Both the fatigue crack factors independent of the load configuration and appro-
initiation and propagation stages are controlled by the priate reference stresses to be used. The only parameters
magnitude and the distribution of stresses in the potential needed for estimating the stress peak and the stress
crack plane. The peak stresses at the weld toe can be distribution induced by any combination of loads are only
calculated using stress concentration factors, available in the geometrically unique stress concentration factors and
the literature, and appropriate reference stresses. These appropriate reference or nominal stresses.
stress concentration factors are unique for given geom-
etry and mode of loading. However, weldments are often
subjected to multiple loading modes, and therefore it is
not easy to define a unique nominal or reference stress.
For this reason, the use of classical stress concentration
2 The nature of the stresses
factors is limited to simple geometry and load configura- in the weld toe region
tions for which they were derived. This problem can be
resolved by using the hot spot or structural stress, σhs con- The stress state at the weld toe is multi-axial in nature. But
cept applied initially in the offshore structures industry [1]. the plate surface is usually free of stresses, and therefore
If the stress concentration factors, based on the hot spot the stress state at the weld toe is in general reduced to
stress, σhs as the reference (or nominal stress), are known one non-zero shear and two in-plane normal stress com-
then the shell or coarse 3D finite element mesh models ponents (Figure 1). Due to stress concentration at the
[2] can be used to determine only the hot spot stress at weld toe the stress component, σyy normal to the weld
the weld toe and subsequently to determine the peak toe line is the largest in magnitude and it is predominantly
stress by using appropriate stress concentration factors. responsible for the fatigue damage accumulation in this
Unfortunately, the hot spot stress based stress concen- region. Therefore, it is sufficient in practice to consider
trations factors vary for the same geometry depending for the fatigue analysis of welded joints only the stress
on the type of loading, i.e. such stress concentration fac- component, i.e. its magnitude and distribution across the
tors are not unique for a given geometry. This is a serious plate thickness.

Doc. IIW-2201, recommended for publication by Commission XIII “Fatigue of Welded Components and Structure.”

1106_0037_WELDING_7_8_2011.indd 2 14/06/11 10:35:01

 STRESS ANALYSIS and FATIGUE of welded structures

distributions are defined as follows; (A) represents the

remote normal through-thickness stress distribution away
from the weld, (B) the actual through-thickness normal
stress distribution in the weld toe plane, (C) the statically
equivalent linearized normal stress distribution in the weld
toe plane, i.e. the stress distribution (C) yields the same
resultant force and bending moment as the actual stress
a) The overall geometry b) The stress state distribution (B). The linearized stress distribution (C) is
at the weld toe independent of the micro-geometrical weld parameters
Figure 1 – Stress state in the weld toe region such as the weld radius, r, and the weld angle, Θ, contrary
of a welded joint to the stress distribution (B) which does depend on these
features. The statically equivalent linearized stress distri-
bution (C) can be characterized by two parameters, i.e. the
magnitude of the hot spot stress, σhs and the slope.
3 The hot spot stress and the stress The stress concentration factor and the peak stress are
dependent on the magnitude and also the slope (gradi-
concentration factor
ent) of the linearized stress field C (Figure 3). Therefore
The nominal stress, σn in a plate without any attachments the same hot spot stress (V hs 1,a
V hs
1,b
), as seen in Figure 3,
or notches [Figure 2 a)] would be equal to that one deter- may ‘produce’ different stress concentration factors and
mined using the simple tension or/and bending stress different peak stresses, σpeak. For this reason the hot spot
formula. The existence of the attachment changes the stress alone is not sufficient for the determination of the
stiffness in the weld toe region resulting in the stress con- load independent stress concentration factors. In order to
centration and non-linear through-thickness distribution define a unique stress concentration factor dependent on
as shown in Figure 2 b). However the nominal membrane the geometry only both the magnitude and the gradient
and bending stresses, actually nonexistent in the welded of the linearized (hot spot) stress must be accounted for.
joint, are the same as in the unwelded plate. Unfortunately, Therefore, Niemi [3] has proposed to decompose the lin-
determination of meaningful nominal stress in complex earized through-thickness stress field (Figure 3) into the
welded structures is difficult and often non-unique. uniformly distributed membrane (axial) stress field, V hs m
3
and the anti-symmetric bending stress field, V hs . This is a
b

Therefore the structural stress, σhs often termed as the ‘hot very useful concept because it captures the stress gradi-
spot stress’, is used in some cases. The hot spot stress has ent (V hs m
V hsb ) around the hot spot stress location. However,
the advantage that it accounts for the effect of the global in order to determine appropriate magnitude of the peak
geometry of the structure and the existence of the weld, stress, σpeak, the stress concentration for pure axial load
but it does not account for the micro-geometrical effects ( K tm,hs ) and pure bending load ( K tb,hs) need to be known.
[Figure 2 b)] such as the weld toe radius, r, and weld angle, The advantage of using two stress concentration factors
Θ. Typical stress distributions in a welded connection with K tm,hs and K tb,hs lies in the fact that they are independent of
fillet welds are shown in Figure 2 b). These various stress the load magnitude and are unique for a given geometry.
In addition, the nominal stresses and the hot spot stresses
for pure axial loading are the same, and analogously the
same applies to bending load. Therefore the classical

a) Stress fields in an unwelded plate

b) Stress fields in a plate with non-load carrying one-sided Figure 3 – The effect of load configuration on through-
attachment with fillet welds thickness stress fields in two geometrically identical
weldments having the same hot spot stress magnitude
Figure 2 – Stresses in unwelded and welded plate but different stress contributions

N° 07
08 2011 Vol. 55 WELDING IN THE WORLD Peer-reviewed Section

1106_0037_WELDING_7_8_2011.indd 3 14/06/11 10:35:02

 STRESS ANALYSIS and FATIGUE of welded structures

b noting that the definition of the classical nominal stress

stress concentration K tm,hs and K t ,hs factors based on the around point B (Figure 4) is very vague in this case.
nominal stress can be used. The peak stress, σpeak, neces-
sary for the prediction of fatigue crack initiation can be In the case of shell finite element analysis the linearized
finally determined as the sum of the membrane and pure through-thickness stress is the final result of the analysis
bending load contribution. and can be easily extracted from the final output data.
V peak V hsm ˜ K tm,hs  V hsb ˜ K tb,hs
(1)
Thus, in order to determine the peak stress, σpeak, the
axial and bending hot spot stresses V hsm
and V hs
b
respec-
tively and appropriate stress concentration factors K t ,hs
m
4 Stress concentration factors
b
and K t ,hs must be known. Therefore, it may be informa- for fillet welds
tive at this moment to clarify the difference between the
linearized stress field and the classical nominal stress The hot spot stresses under pure axial and pure bending
and various stress concentration factor definitions used loads are the same as the nominal stresses (V hsm
V nm and
in practice. The difference between the classical nominal V hs V n ) and therefore the classical stress concentration
b b

stress and the original hot spot stress definition lies in the factors can be used in Equation (1). Extensive literature
fact that the hot spot stresses V hsm
and V hs
b
are uniquely search was carried out [4-7] for this reason and several
defined at any point along the weld toe (Figure 4). They stress concentration factor solutions were compared with
can be determined by linearization of the through-thick- each other and verified using in-house finite element
ness stress below the point on the weld toe line where the data. The most universal were the stress concentration
peak stress needs to be determined. factors supplied by Japanese researchers [7] and they
are discussed below. The generic geometrical configura-
The relationships between the actual through-thickness tions used for producing these stress concentration fac-
stress distribution and hot spot stresses V hs
m
and V hs
b
tors were the T-butt and cruciform welded joints shown in
are given by Equations (2) and (3) respectively and they (Figure 5) and (Figure 6) respectively.
represent the average membrane and bending stress
contributions.
4 0
³t V  x , y 0 dx
V hs
m
(2)
t
0
6 ˜ ³V  x, y 0  xdx
V hs
b t
2
t (3)

Definitions and the method of determination of stresses

V hsm and V hsb at points A and B (Figure 4) on the weld
toe line are the same and the stress concentration factors
K tm,hs and K tb,hs at those points are also the same if the Figure 5 – Geometry and dimensions of a T-butt welded
weld geometry and dimensions are the same. Therefore, joint subjected to axial and bending load
in order to determine the peak stress σpeak, at point A or B
the same stress concentration factors K tm,hs and K tb,hs can
be used when correctly associated with the hot spot mem-
brane V hsm
and bending V hs b
stresses at those points. Worth

Figure 4 – The actual through-thickness stress distributions

along the weld toe line and the linearized statically Figure 6 – Geometry and dimensions of a cruciform welded
equivalent stress fields joint subjected to axial and bending load

1106_0037_WELDING_7_8_2011.indd 4 14/06/11 10:35:03

 STRESS ANALYSIS and FATIGUE of welded structures

– Stress concentration factor near one-sided fillet weld

under axial load (Figure 5, point A)
§ W · ª 0.65
5 The shell finite element model
1 exp ¨ 0.94 º
¸ « » The linear through-thickness stress field is naturally
© 2h ¹ 1 h
K m
1 u« u » embedded in as the property of most basic shell finite
t ,hs
§ W · « 2.8 § ·  2 r »
W (4)
1 exp ¨ 0.45S ¨ ¸ elements. The output stresses are the stress components
¸ « © t ¹ »
© 2h ¹ ¬ ¼ V hs
1
and V hs
2
(Figure 7) acting on each side of the plate
thickness. Therefore the determination of the membrane
– Stress concentration factor near one-sided fillet weld and bending hot spot stresses requires only simple post-
under bending load (Figure 5, point A). processing as shown below.
§ W · V hs
1
 V hs
2
1 exp ¨ 0.94 ¸ V hsm (8)
© 2h ¹ § 2t p 2r · 2
K tb,hs 1 u 1.9 tanh ¨  ¸u
§ W · © t  2h t ¹ V hs  V hs
1 2
1 exp ¨ 0.45S ¸ V hsb (9)
© 2h ¹ 2

ª § 2h ·0.25 º ª § r·
4 º (5) Unfortunately, such a simple finite element model as that
« ¨ ¸ » « 0.13  0.65 ¨1 ¸ »
« © t¹ » one shown in Figure 7 b) is not capable of supplying suf-
tanh « © ¹ » u «
t
« r » 1 » ficiently accurate stresses in the weld toe region. This is
« 1 t » « § r ·3
¨ ¸
»
¬ ¼ «¬ ©t ¹
»
¼
due to the fact that the critical cross-section in the actual
welded joint is located at the weld toe (sections A and B,
where Figure 7) being away from the mid-planes intersection. In
addition, the magnitude of stresses V hs 1
and V hs
2
and the
W = (t + 2h) + 0.3 (tp + 2hp ) resultant slope of the linear stress field depend on the dis-
tance from point ‘O’ [Figure 7 b)] and the size of the shell
Equations (4) and (5) are semi-empirical in nature and
element. The shell stresses in the weld toe region depend
have been derived using analytical solutions for stress
strongly on the local stiffness of the joint and, therefore,
concentrations at corners supplemented by extensive
they are sensitive to how the weld stiffness is accounted
finite element stress concentration database. Their appli-
for in the finite element shell model. It is important to model
cation was verified for a range of geometrical configura-
the weldment and any welded structure in such a way that
tions limited to 0.02 ≤ r/t ≤ 0.16 and 30° ≤ Θ ≤ 60°.
the hot spot shell membrane V hs m
and bending stress V hs b
in 5

Similar expressions have also been derived for stress con- critical cross-sections [Figure 7 b)] are the same as those
centration factors in cruciform welded joints [7] and they which would be determined from the linearization of the
apply in general to weldments with two symmetric fillet actual 3D stress fields, obtained analytically or from fine
welds placed on both sides of the load carrying plate. mesh 3D finite element model [Figure 7 a)] of the joint. In
other words the shell model of the weld needs to be also
– Stress concentration factor for a cruciform joint sub- included.
jected to an axial load (Figure 6, point A).
Fayard, Bignonnet and Dang Van [8] have proposed a
§ W · ª º
0.65
1 exp ¨ 0.94 ¸ shell finite element model with rigid bars simulating the
2h ¹ « h»
K tm,hs 1 © u 2.2 «
1
u » (6)
§ W · « 2.8 § W · r»
1 exp ¨ 0.45S ¸ ¨ ¸  2
2h ¹ «¬ © t ¹ »¼
©

– Stress concentration factor for cruciform joint sub-

jected to bending load (Figure 6, point A).
§ W ·
1 exp ¨ 0.94 ¸
© 2h ¹ § 2t p 2r ·
K tben 1  u tanh ¨  ¸u
§ W · © t  2h t ¹
,n

1 exp ¨ 0.45S ¸
© 2h ¹
(7)
ª § 2h ·0.25 º ª § r·
4 º

« ¨ ¸ » « 0.13  0.65 ¨ 1 ¸ »
« © t¹ »
tanh « © ¹ » u «
t
« r » 1 »
« 1  » « § ·
r 3 »
¬ t ¼ « ¨ ¸ »
¬ © t ¹ ¼

where
a) Welded joint and stress distributions
W = (t + 4h) + 0.3(tp + 2hp) in critical cross-sections
b) Shell finite element model and resultant stress
Equations (6) and (7) are also empirical in nature and distributions
have been derived using extensive finite element stress
data. The range of application for these expressions is: Figure 7 – A welded joint and its simple shell finite
r/t = 0.1-0.2, h/t = 0.5-1.2, Θ = 30°-80°. element model

N° 07
08 2011 Vol. 55 WELDING IN THE WORLD Peer-reviewed Section

1106_0037_WELDING_7_8_2011.indd 5 14/06/11 10:35:07

 STRESS ANALYSIS and FATIGUE of welded structures

weld. They have also formulated a set of rules concerning b. The first and the second row of elements
the finite element meshing in order to capture correctly adjacent to the theoretical intersection line
the properties of the linear stress field. However, using of mid-thickness planes must be of the size
shell elements and rigid bars was found not very conve- (tp + h)/4 in the ‘x’ direction for elements in
nient in practice. Therefore, a new model involving only the main plate and (t + h)/4 in the ‘y’ direction
shell elements of the same type in the entire structure for elements in the attachment. The shell ele-
was proposed. ments simulating the weld are subsequently
attached to each plate in the middle of the
There are two important issues concerning the shell FE weld leg length and they are spanning the first
modelling of welded joints namely: the simulation of the two rows of elements in each plate. The thick-
local weld stiffness and the location of the stress refer- ness of the shell elements simulating the weld
ence point where the stress corresponding to the actual is recommended to be equal to the thickness
weld toe position is to be determined. Therefore, the shell of the thinner plate being connected by the
finite element model has to be constructed in such a way weld (i.e. either t or tp whichever is less). All
that the location of the stress reference point coincides shell elements simulating the weld are of the
with the actual position of the weld toe (Figure 7). same thickness.
All shell elements in the weld region have the
In order to assure that the global effects of the joint same dimension in the z direction and it is
geometry and the weld are adequately modelled, a set equal or less than the half weld leg length, i.e.
of rules have been formulated concerning the construc- ‘h/2’ or less.
tion of appropriate finite element shell model as shown in
Figure 8. The meshing principles of the model are illus- c. The shell elements in the third row simulat-
trated using as an example a T-welded joint. The following ing the main plate should have the size equal
steps need to be carried out while creating appropriate to the half weld leg length ‘h/2’ in the ‘x’ and
shell finite element mesh. ‘z’ directions and the same ‘h/2’ dimension
in the ‘y’ and ‘z’ directions for elements in the
a. Connect the mid-thickness plate planes attachment plate [Figure 8 b)]. The choice of
[Figure 8 a) and 8 b)], and add one layer (blue) such element dimensions enables to locate
6
of inclined shell elements representing the the reference points A at the nodal points of
weld. elements from the third row. The location of

a) Side view of a T-joint with single fillet weld

b) The shell finite element model
c) Superposition of the actual welded joint and its shell finite element model

Figure 8 – Rules for constructing the shell finite element mesh of a welded T-joint

1106_0037_WELDING_7_8_2011.indd 6 14/06/11 10:35:08

 STRESS ANALYSIS and FATIGUE of welded structures

m b
the reference point A must coincide with the associated stress concentration factors K t ,hs and K t ,hs are
physical position of the weld toe. Thus stresses sufficient for the determination of the through-thickness
at the reference point A are the same as the distribution denoted as distribution B in Figure 2 b).
nodal stresses and they can be extracted
without any interpolation or additional post- The stress distribution, needed for the stress intensity fac-
processing. tor K calculation, can be determined by using universal
stress distributions proposed by Monahan [13]. Both equa-
d. The dimension ‘z’ of the first two rows of ele- tions shown below were derived for the through-thickness
ments adjacent to the intersection of plate stress distribution in a T-butt weldment (Figure 5) but they
mid-thickness planes is dictated by the small- can also be applied over half of the thickness in the case
est element in the region, i.e. it should not be of cruciform weldments (i.e. for a weldment with symmet-
greater than half of the weld leg length ‘h/2’. It ric welds located on both side of the plate as shown in
means that the first and the second row of ele- Figure 6).
ments in the main plate [Figure 8 b)] have the
dimension of [(h/4 + tp/4) × h/2]. The first two For pure tension loading the through-thickness stress
rows of elements in the attachment counted distribution can be sufficiently accurate approximated by
from the mid-plane intersection should have Equation (10).
the size of [(h/4 + t/4) × h/2]. The elements m ª
1
 º
3
K tm, hsV hs 

Vm x «§¨ x  1 ·¸  1 §¨ x  1 ·¸ » 1
2 2
in the third row are (h/2 × h/2) in size. The
2 2 «© r 2 ¹ 2 © r 2 ¹ » Gm (10)
spacing in the ‘z’ direction might need to be ¬ ¼
smaller than half of the weld leg length ‘h/2’ where
while modelling corners of non-circular tubes x
or weld ends around gusset plates. Gm 1 for d 0.3
r
0.94 u e EmTm x
5.1 Determination of the peak stress Gm 0.06   EmTm1.1
for ! 0.3
at the weld toe 1  E m Tm u e
3 0.8 r
q
In order to determine the peak stress σpeak at the weld toe §r ·
Em 1.05 u 4 0.18 ¨ ¸
it is necessary to determine the membrane V hs m
and bend- ©t ¹ 7
ing V hs stress from the shell finite element model using
b
q = −0.12Θ −0.62
Equations (8-9). Then the stress concentration factors
x r
K tm,hs and K tb,hs for tension and bending need to be calcu- Tm  0.3
lated from Equations (4-5) or (6-7) by using actual dimen- t t
sions of the weld. The peak stress σpeak can be finally cal- For pure bending load Equation (11) is recommended:
culated from Equation (1).
§x·
b ª  º 1 2 ¨ ¸
1 3
K tb, hsV hs 

Vb x «§¨  ·¸  §¨  ·¸ »
x 1 1 x 1 ©t ¹
2 2
The knowledge of the linear elastic peak stress σpeak (11)
2 2 «© r 2 ¹ 2 © r 2 ¹ » Gb
enables subsequently the assessment of the fatigue ¬ ¼
crack initiation life by applying the local strain-life method where
widely used in the automotive industry [9-12].
x
Gb 1 for d 0.4
r
5.2 Determination of the through-thickness
stress distribution 0.93 u e EbTb x
Gb 0.07   E bTb1.2
for ! 0.4
1  E b Tb u e
3 0.6 r
Because welded structures are known as having high
stress concentration at weld toes and roots the fatigue §
 ¨ 0.0026 
0.0825 ·
¸
crack initiation period might be relatively short and there- §r · © 4 ¹
Eb 0.9 ¨ ¸
fore fatigue life assessment based on the fatigue crack ©t ¹
growth analysis is often required. In order to carry out a x r
meaningful fatigue crack growth analysis appropriate
Tb  0.4
t t
stress intensity factor solutions are needed. Because of
wide variety of possible configurations of the global joint Equations (10-11) can be used to predict through-thick-
geometry, the weld geometry, the crack geometry and ness stress distributions near fillet welds joining plates,
loading reliable ready-made stress intensity solutions are tubes and other structural elements providing that param-
seldom available. Therefore the weight function method eters Θ, r/t, and x are within the following limits:
(discussed later) seems to be a convenient and efficient S S 1 r 1
solution but the through-thickness stress distribution in the d4 d and d d and 0 d x d t (12)
6 3 50 t 15
prospective crack plane must be known in such a case.
If both membrane and bending stresses are present the
It has been found that the same information, i.e. the resultant through-thickness stress distribution can be
membrane V hs
m
and bending V hs
b
hot spot stresses and obtained by superposition of Equations (10) and (11).

N° 07
08 2011 Vol. 55 WELDING IN THE WORLD Peer-reviewed Section

1106_0037_WELDING_7_8_2011.indd 7 14/06/11 10:35:09

 STRESS ANALYSIS and FATIGUE of welded structures

ª § x ·º The same ESED rule but combined with the expanded by

« K mV m 1 K bV b 1 2 ¨ t ¸ » factor of 2 stress-strain curve is applied for calculating the
V x V x  V x  «
m b t hs
˜  t hs
˜ © ¹»
elastic-plastic strain and stress range.
« 2 2 Gm 2 2 Gb »
«¬ »¼
 'V   'V 
1
2 a 2
peak 2 ˜ 'V a § 'V a · n '
ª 
1
 º
3  ' ¨ ¸ – the ESED rule
«¨§ x  1 ¸·  1 §¨ x  1 ·¸ »
2 2
(13) 2E 2E n  1 © 2K ' ¹ (20)
«© r 2 ¹ 2© r 2¹ »
¬ ¼ 1
'V a § 'V a · n '
'H a
 2¨ ' ¸ – the expanded material stress-
Appropriate weight functions [14-15] and the stress E © 2K ¹
distribution (13) make it possible to determine stress strain curve (21)
intensity factors and to subsequently simulate the fatigue
crack growth in any welded structure without the labour The actual maximum stress V max a
at the weld toe and the
and time consuming extensive FE numerical analyses of actual strain range Δεa obtained from Equations (14-17)
cracked bodies. In addition the method is self-consistent or (18-21) formulate the base for calculating the number
and does not require making any arbitrary adjustments. of cycles, Ni, to initiate a fatigue crack at the weld toe.

The basic fatigue material property is the strain-life equa-

tion proposed [11] by Manson and Coffin.

6 Fatigue crack initiation analysis 'H

2
V f'
E
2Ni 
b
 H f' 2Ni 
c
(22)
The maximum of the peak stress σpeak,max and the peak
In order to account for the mean stress effect or the exis-
stress range Δσpeak at the weld toe can be subsequently
tence of residual stresses the SWT [12] fatigue damage
used for the calculation of the actual stresses and elastic-
parameter can be used in association with the Manson-
plastic strains at the weld toe. The most common method
Coffin curve.
is the use of the Neuber [9] or the ESED [10] method and
the material cyclic stress-strain curve in the form of the
V a 'H a V 
' 2
f
˜ 2Nf 
2b
 H f' V f' ˜ 2Nf 
b c
Ramberg-Osgood expression. max (23)
2 E
V 
2
8
peak ,max
V max
a
H max
a Unfortunately, the strain-life method does not specify
– the Neuber rule (14)
E what crack size is associated with the end of the initia-
1 tion period therefore an engineering definition needs to
V max
a
§ V a ·n' be adopted. Based on author’s experience a semi-circular
H a
max  ¨ max' ¸ – the material Ramberg-
E © K ¹ crack of the initial depth of ai = 0.5-0.8 mm seems to be a
Osgood stress-strain curve (15)
reasonable good assumption.
In order to Calculate the elastic-plastic strain range and
corresponding stress range the same Neuber rule associ- It is worth noting that there are several fatigue software
ated with the stress-strain curve expanded by factor of 2 packages for which the local elastic stress σpeak and
can be used. material curves (13) and (19) are the standard input data
for automatic fatigue crack initiation life assessment.
 'V 
2
Therefore, the method described above can be associated
'V a ˜ 'H a
peak

E – the Neuber rule (16) with any available computer fatigue software package.
1
'V a § 'V a · n'
'H a  2¨ ' ¸
E © 2K ¹ – the expanded material stress-

Because it is known that the Neuber rule has the ten-

strain curve (17)
7 The fatigue crack growth analysis
dency of overestimating the local elastic-plastic strains The fatigue crack growth period is often thought to be
and stresses at high stress concentration factors the representing almost the entire fatigue life of weldments
Equivalent Strain Energy Density (ESED) method can be because it is believed that the fatigue crack initiation
used in the form of analogous set of equations for obtain- period in welded joints is relatively short. The authors
ing the maximum elastic-plastic strain and stress at the experience is that the ratio of the fatigue crack initiation
weld toe. life to the length of the fatigue crack growth period var-
ies depending on the load and weld geometrical factors
V  V 
1
2
a 2 and its relative contribution to the total fatigue life of a
peak ,max max V max
a
§ V max
a
·n
'

 ¨ ¸ – the ESED rule weldment increases in with the decrease of the applied
2E 2E n  1© K ' ¹
'
cyclic load and the increase of the total life. Therefore
(18)
good assessments of both the fatigue crack initiation and
1
V max
a
§ V a ·n' propagation period are necessary for reasonably accu-
H a
max  ¨ max' ¸ – the material Ramberg-Osgood rate estimation of fatigue lives of weldments subjected to
E © K ¹ stress-strain curve (19) cyclic loading histories.

1106_0037_WELDING_7_8_2011.indd 8 14/06/11 10:35:10

 STRESS ANALYSIS and FATIGUE of welded structures

The first and the most popular expression characterizing

material properties from the point of view of the mate-
rial resistance to fatigue crack growth is the Paris [16]
equation.
da
C  'K 
m
(24)
dN
Unfortunately, the Paris equation in its original form (24)
does not account for the effect of the stress ratio R or
the mean stress. Therefore the recently proposed [17, 18]
Equation (25) is gaining popularity.

C  K max 'K p 
da 1 p J

dN (25)

where
C, p, m and γ are material constants, Kmax is the maximum
stress intensity factor and ΔK is the stress intensity range.
Figure 9 – Notation for the weight function
In the case of pulsating constant amplitude stress history
(R = 0) Equation (25) takes the well-known form of the
Paris equation. Parameters M1, M2 and M3 depend on the crack geometry
and they have been derived already [15] for a variety of
The fatigue crack propagation life is obtained by analyti- cracks. The Mi parameters for a single edge and surface
cal or numerical integration of the fatigue crack growth semi-elliptical crack are given in reference [15] and listed
equation. in the Appendix.
af af

´ da ´ da If the weight function is known, the stress intensity fac-

Np µ or Np µ
¶ C  K max 'K 
(26)
¶ C  'K  p J
m 1 p
tor K can be calculated by integrating the product of the
ai
stress distribution, σ (x), in the prospective crack plane and
ai

the weight function m(x,a): 9

However, in order to Integrate any fatigue crack growth
a
rate expression available in the literature the stress inten-
sity factors Kmax and ΔK need to be determined.
K AV ( x ) ³ V  x  ˜ m  x , a  ˜ dx
0
(28)

7.1 Application of the weight function Thus, the calculation of stress intensity factors by the
weight function method for any crack, including cracks in
method for efficient determination
weldments, requires the knowledge of the stress distribu-
of stress intensity factors
tion, σ (x), in the prospective crack plane in the un-cracked
In the case of simple geometry and load configurations body [Figure 10 a)] and then the stress distribution should
like the edge or semi-elliptical surface crack in a plate be virtually applied to the crack surfaces [Figure 10 b)].
subjected to pure bending or tension load ready-made
stress intensity factor expressions can be found [19] in
Handbooks of Stress Intensity Factors. Unfortunately,
there are no ready-made solutions for cracks in welded
structures except for a few crack configurations in simple
welded joints. Therefore the weight function technique
[14, 15] was employed in order to determine stress inten-
sity factors for cracks in real welded structures.

The weight function can be understood as the stress

intensity factor (Figure 9) induced by the simplest load
configuration, i.e. a pair of unit splitting forces F attached
to the crack surface.

There are many expressions for various weight functions but

it is possible [15] to write them in one general form (27).
­ 1
2F ° § x ·2
m  x,a K P
®1 M1 ¨1 ¸
2S  a  x  °̄
A
© a¹
(27)
3
½ Figure 10 – The use of the weight function
§ x· § x ·2 °
M2 ¨1 ¸  M3 ¨1 ¸ ¾ for an edge crack in plate for calculating the SIF
© a¹ © a¹ °
¿ for a crack in a T-butt weldment

N° 07
08 2011 Vol. 55 WELDING IN THE WORLD Peer-reviewed Section

1106_0037_WELDING_7_8_2011.indd 9 14/06/11 10:35:11

 STRESS ANALYSIS and FATIGUE of welded structures

Finally the product of the stress distribution σ (x) and the

weight function m(x,a) needs to be integrated over the
entire crack surface area [see Equation (28)]. It is to note
that the stress analysis needs to be carried out only once
and for an un-cracked body [Figure 10 a)].

The stress intensity factor calculations can be repeated

after each crack increment induced by subsequent load
cycles so the stress intensity factor is calculated for the
instantaneous (actual) crack size and geometry. Such a
method enables simulation of the crack growth and the
evolution of the crack shape in the case of 2D planar
cracks.

8 Stress and fatigue analysis Figure 12 – Solid model of the welded joint- configuration
of support points and the location of the loading force
of a tubular welded section
subjected to constant amplitude
and maximum uniformity of weld geometry along its entire
fully reversed torsion length. The average weld dimensions obtained from the
welding process were (see Figure 5):
and bending loads
Several welded structures and welded joints were stud- t = 0.312 in, tp = 3 x 0.312 in, h = 0.312 in, hp = 0.312 in,
ied experimentally and numerically in order to verify the Θ = 45°, r = 0.0312 in
validity of the proposed methodology. The welded struc-
The specimens were tested at two different cyclic load
10 ture shown in Figure 11, subjected to torsion and bend-
levels of ± 3 000 lb and ± 4 000 lb applied at the free
ing load, was chosen as an example for illustrating the
end of the large square profile (Figure 12). The bottom
stepwise procedure. The overall geometry of the welded
base plate of the fixture was held by six screws. In order
joint fixed in the testing rig is also shown in Figure 11.
to assess anticipated scatter of experimental fatigue lives
Dimensions of selected welded tubular profiles were
seven specimens were tested at each load level.
4 x 4 x 23.625 in and 2 x 6 x 14.313 in and the wall
thicknesses were equal to 0.312 in. The geometrical
constrains of the fixture and the location of the applied 8.1 The shell FE stress analysis
load is shown in Figure 12. The welded tubular profiles The shell model of the analysed welded tubular joint is
were made of A22-H steel (ASTM A500 Cold Formed presented in Figure 13. The overall stress field obtained
Steel for Structural Tubing) used often in the construc- from the FE analysis of the entire structure is also shown
tion and earth moving machinery. The manual Flux Cord graphically in Figure 13. Three high stress locations were
Arc Welding method was used to manufacture a series identified with the highest stresses found at Location 1.
of specimens. The welding parameters were typical for Therefore detail analysis of stresses at Location 1 was
manual welding aimed at the maximum weld penetration undertaken. The exact location of the maximum stress
was found in the region around the ending edge of the
rectangular tube. The stresses of interest were those
found at the reference point shown in Figure 14 of the
shell model. The location of the reference point was coin-
ciding as usual with the physical position of the weld toe
in the actual weld joint. The distance between the refer-
ence point and the tube wall mid-planes intersection point
in the shell FE model was equal to:
tp 0.312
h 0.312  0.468 in
2 2

The shell stresses V hs 1

and V hs
2
on opposite sides of the
tube wall (see Figure 7), obtained from the shell FE
model, were used for the determination of the mem-
brane and bending [Equations (8-9)] hot spot stresses,
Figure 11 – The tubular welded joint subjected to torsion V hsm and V hsb , respectively. The peak stress at the weld toe,
and bending load in the testing rig σpeak, was determined from Equation (1). The through-

1106_0037_WELDING_7_8_2011.indd 10 14/06/11 10:35:12

 STRESS ANALYSIS and FATIGUE of welded structures

Figure 14 – Details of the shell FE model and the location

of the reference point for determining the hot spot stress

Figure 13 – The shell finite element model of the welded

joint showing overall stress distribution with three distinct
regions of high stresses

thickness stress distribution, σ (x), was calculated using

Equations (10-11,13).

The shell stresses induced by the unit load of P = 1.0 lb

on the two sides of the tube wall and calculated using the
shell FE data were:
V hs
1
= 8.25 psi and V hs
2
= −3.05 psi 11

Therefore the hot spot stresses determined according to

Equations (8) and (9) were:
V hs
1
 V hs
2
8.25   3.05 
V hsm 2.6 psi
2 2
Figure 15 – Through-thickness stress distribution
V hs
1
 V hs
2
8.25   3.05 
V hsb 5.65 psi at Loc. 1 induced by the unit load P = 1 lb
2 2
m b
The stress concentration factors K t ,hs and K t ,hs obtained
from Equations (4) and (5) for geometrical dimensions carried out for the same welded joint. The through-thick-
t = 0.312 in, tp = 3t = 0.936 in, h = 0.312 in, hp = 0.312 in, ness stress distribution obtained from the 3D fine mesh
r = 0.0312 in and Θ = 45°, were K t ,hs = 1.784 and K t ,hs
m b FE model (Figure 16) is also shown in Figure 15. It can be
= 2.203. The hot spot stresses corresponding to the unit noted that the stress distribution obtained from the coarse
load P = 1 lb and the stress concentration factors input- mesh shell FE model and Monahan’s equations coincide
ted into Equation (1) resulted in the determination of the very well with the stress distribution obtained from the
peak stress at the weld toe. very detail and fine 3D FE mesh model of the same joint
showing good accuracy of the proposed methodology.
σpeak = V hs ⋅ K t ,hs + V hs ⋅ K t ,hs = 2.6 x 1.784 = 5.65 x 2.203 Similar results were obtained for a variety of other welded
m m b b

= 17.089 psi. In order to determine stresses correspond- structures.

ing to the actual load P = 3 000 lb or 40 000 lb the stress
σpeak = 17.089 psi was scaled by the factor of 3000 or In addition to the FE calculations of the through-thick-
4000 depending on the applied load magnitude. ness stress distribution induced by the applied load P
(Figure 12) several X-ray measurements of welding
The through-thickness stress distribution σ (x) was residual stresses were carried out on a few full scale
obtained using the universal Equations (10-11 and 13) welded joints. Based on the X-ray measurements and the
and geometrical parameters, hot spot stresses and stress force and bending moment equilibrium requirements the
concentration factors listed above. The through-thickness approximate through-thickness residual stress distribu-
stress distribution induced by the applied external force tion at the hot spot Location 1 has been estimated and it
P = 3 000 lb is shown in Figure 15. is shown in Figure 17. The actual maximum residual stress
distribution measured in the closed proximity of the refer-
In order to verify the accuracy of the shell FE based ence point (Figures 11 and 13) in as-welded joints was
method a very detail 3D FE stress analysis was also σr0 = 45 ksi.

N° 07
08 2011 Vol. 55 WELDING IN THE WORLD Peer-reviewed Section

1106_0037_WELDING_7_8_2011.indd 11 14/06/11 10:35:18

 STRESS ANALYSIS and FATIGUE of welded structures

Figure 16 – Details of the 3D fine mesh FE model Figure 17 – Through-thickness residual stress distribution
of the critical location at Location 1

9 Experimental and theoretical initiation. The fatigue crack initiation life calculations were
based on the material properties listed below.
fatigue analyses
The monotonic and fatigue (cyclic) properties of the
The fatigue life was predicted as a sum of the crack initia- A22-H steel material are listed in Table 1 and Table 2
tion life (the strain-life method) and crack propagation life respectively.
(the fracture mechanics approach). In order to compare
predicted lives with the experimental data, two series of The resultant amplitude of the fully reversed (R = -1) cyclic
tests were carried out. The first one was conducted at the stress at the weld toe Location 1 obtained from the linear
12 fully reversed cyclic load of ± 3 000 lb and the other at elastic analysis was σa = 51.27 ksi and σa = 68.36 ksi for
± 4 000 lb. the load 3 000 lb and 4 000 lb respectively.

All experiments were carried out under the load control The assessed fatigue crack initiation lives, Ni, are listed
conditions and the crack length 2c, visible on the surface, in Tables 3 and 4. Unfortunately comparison of calcu-
was measured versus the number of applied load cycles. lated fatigue crack initiation lives with experimental data
Optical measurements with the aid of magnifying glass is not very meaningful because the method itself does
were carried out throughout all tests. not specify what crack size corresponds to the end of
rather vaguely defined crack initiation period. Therefore
The peak stresses at the weld toe and the through- an engineering definition of the crack initiation size, based
thickness stress distributions induced by the applied on experimental data, needs to be employed. The experi-
loads were obtained by scaling the stress distribution of mental data analysed by the authors up to date indicated
Figure 12, obtained for the unit load P = 1 lb, by factor that the predicted fatigue crack initiation lives coincided
3000 and 4000. most often with lives (number of cycles) needed for the
creation at the weld toe of semi-elliptical cracks of depth
ai = 0.02 in with the aspect ratio ai/ci = 0.286.
9.1 Fatigue crack initiation life analysis
of tested welded joints
Table 2 – The cyclic and fatigue properties
The fatigue crack initiation life was predicted using the in- of the A22-H steel material
house FALIN fatigue software with implemented strain-life
fatigue live prediction procedure. The procedure is briefly Fatigue strength coefficient (σ ’f) [ksi] 169.98
described in Section 6. The elastic-plastic stresses and Fatigue strength exponent (b) -0.12
strains at the weld toe were calculated for each load cycle Fatigue ductility coefficient(ε’f) 0.648
based on the Ramberg-Osgood stress strain curve (13) Fatigue ductility exponent (c) -0.543
and the Neuber Equation (12). These strains and stresses
Cyclic strength coefficient (K’) [ksi] 155.2
were used in the Smith, Watson, and Topper (SWT) strain-
life Equation (20) for calculating the fatigue life to crack Cyclic strain hardening exponent (n’) 0.187

Table 1 – Monotonic mechanical properties of the A22-H steel material

Ultimate strength (Su) Yield strength (Sy) Elastic modulus (E)

[ksi] [ksi] [ksi]
79.0 68.89 29 938

1106_0037_WELDING_7_8_2011.indd 12 14/06/11 10:35:32

 STRESS ANALYSIS and FATIGUE of welded structures

The fatigue crack analyses were carried out for two load ΔKth = 3.19 ksi√in at R = 0 and KC = 72.81 ksi√in.
levels (3 000 lb and 4 000 lb) with and without residual
stresses. The residual stress was combined with the cyclic It should be noted that the crack was not growing with the
stress induced by the applied load by including it [20, 21] same rate in all directions. Therefore crack increments at
into the Neuber or ESED equation in such a way that only the deepest point A and those on the surface (point B
the actual maximum elastic-plastic strain and stresses at Figure A2) were determined separately for each cycle.
the weld toe were affected. The crack dimensions and the crack aspect ratio (a/c)
V  Vr0 
2 were updated after each load cycle. The fatigue crack
V max H max growth predictions were carried out first neglecting and
peak ,max a a
– the Neuber rule (29)
E secondly including the residual stress effect.
V  Vr0  V 
2 2 1
a
peak ,max max V max
a
§ V max
a
·n '

In order to account for the residual stress effect the method

 ¨ ¸ – the ESED rule
2E 2E n  1© K ' ¹
'
described in reference [21] and the modified Paris equa-
(30)
tion accounting for the stress ratio R were used.
The magnitudes of the elastic-plastic strain and stress
ranges were not affected by the static residual stress and da / dN = C (U ⋅ ΔK)m (31)
they can be determined according to Equations (16-17)
or (20-21). The U parameter accounting for the stress ratio effect was
given by Kurihara [22] for a wide range of stress ratios in
It is noticeable (see Tables 3 and 4) that the residual the form of Equation (32)
stress had profound effect on the fatigue crack initiation 1
U for  5.0 d R d 0.5
life. The analysis indicates that the tensile residual stress 1.5  R (32)
at the weld toe may decrease the fatigue crack initiation U=1 for R > 0.5
life by approximately factor of 3.
First the maximum and minimum stress intensity factors
9.2 The fatigue crack growth analysis for a given loading cycle were determined using the weight
function Equation (A9) and Equation (A33) and the load
The second part of the fatigue life assessment was devoted induced stress distribution σ (x) shown in Figure 15.
to the fracture mechanics based analysis of fatigue crack a 13
growth. The fatigue crack growth analysis was carried out
using the in-house FALPR software package enabling
K max ³ V max  x  ˜ m  x , a  ˜ dx
0 (33)
the calculation of stress intensity factors based on the a
weight function method and subsequent cycle by cycle
fatigue crack growth increments. The observed fatigue
K min ³ V  x  ˜ m  x , a  ˜ dx
0
min

cracks were semi-elliptical in shape (Figure A2) with initial

dimensions of ai = 0.02 in and 2ci = 0.14 in, i.e. the initial Then the stress intensity factor contributed by the residual
aspect ratio was a/c = 0.238. The semi-elliptical surface stress σr(x) shown in Figure 17 was also calculated using
crack in a finite thickness plate was assumed to be the the same weight functions given by Equation (A9) and
appropriate model for fatigue crack growth simulations. Equation (A33).
a

The stress intensity factors for the actual crack shape Kr ³ V  x  ˜ m  x , a  ˜ dx

r (34)
(a/c) and depth, a, were calculated using the weight func- 0

tion given by Equations (A1-A50). Stress intensity factors

Finally all stress intensity factors were combined and the
and crack increments at points A and B (Figure A2) were
effective stress ratio was determined.
simultaneously calculated on cycle by cycle basis. As a
result the crack growth and the crack shape evolution K min  K r
Reff (35)
were simultaneously simulated. K max  K r

The through-thickness stress distribution induced by the The effective stress ratio enabled to determine the actual
external load shown in Figure 15 and the residual stress value of the U parameter in Equation (31) and calculate
of Figure 17 were used for the determination of stress the crack increment induced by analyzed loading cycle.
intensity factors. The crack increments induced by sub- This process was carried out simultaneously for both
sequent load cycles were calculated by using the Paris points A and B (Figure A2) of the semi-elliptical surface
fatigue crack growth Equation (24) valid for R = 0.5 with crack.
parameters:
An example of the crack depth growth versus the num-
m = 3.02 and C = 2.9736 x 10-10 for ΔK in [ksi√in] and ber of applied loading cycles (a vs. N), the evolution of
da/dN in [in/cycle]. the crack aspect ratio (a/c vs. N) and the evolution of
the crack from its initial to the final shape are show in
The threshold stress intensity range and the fracture Figure 18. The experiments and fatigue crack calcula-
toughness for this material were: tions were carried out until the crack reached approximate

N° 07
08 2011 Vol. 55 WELDING IN THE WORLD Peer-reviewed Section

1106_0037_WELDING_7_8_2011.indd 13 14/06/11 10:35:33

 STRESS ANALYSIS and FATIGUE of welded structures

Figure 19 – Experimental surface crack measurements

of the crack length ‘2c’ on seven specimens and predicted
‘2c vs. N’ curves obtained under the cyclic load of ± 4 000 lb
Figure 18 – Fatigue crack growth curve, a vs. N; evolution
of the crack aspect ratio a/c vs. N and evolution It is clear that assuming constant aspect ratio for easier
of the crack shape; obtained for load 3 000 lb and σr0 = 0 fatigue crack growth analysis may contribute to significant
error. Moreover, the surface crack measurements without
depth of af = 0.14 in. The numerical analyses were car- any supporting theoretical crack growth simulations are
ried out using the following geometrical dimensions (see not sufficient for reliable estimation of the crack depth ‘a’.
Figure 5) of the weld:
According to the data above the ratio of the crack initia-
t = 0.312 in, tp = 3t = 0.936 in, h = 0.312 in, hp = 0.312 in, tion to the crack propagation life Ni / Np ≤ 0.303 and the
r = 0.0312 in and Θ = 45°. crack initiation life to the total fatigue life Ni / Nf ≤ 0.233
were rather low (and) indicating that majority of the fatigue
Statistical analysis was performed on large number (more life of analysed weldment was spent on propagating the
14 than 100 measurements) of measured real weld toe
radii and weld angles and the most frequent values given
above were used for the numerical analyses. All calculated
fatigue crack growth lives are summarized in Tables 3 and
4. In addition fatigue crack lengths 2c versus number of
applied load cycles N measured on seven specimens and
those calculated ones are shown in Figure 19.

Summaries of predicted fatigue crack growth periods are

given in Tables 3 and 4.

The a/c vs. N curve shown in Figure 18 indicated that

the crack was initially growing faster into the depth direc-
tion but after reaching the aspect ratio of approximately
a/c = 0.5 it started growing faster along the weld toe
tending to the single edge crack geometry. This conclu-
sion is supported by the actual shape of the final crack Figure 20 – The final shape of the fatigue crack initiated
(Figure 20) after break-opening one of the specimens. at the weld toe around the edge of the rectangular tube

Table 3 – Summary of predicted fatigue lives under load P = 3 000 lb

Ni (Cycles) N P (Cycles)
Residual stress [ksi] Ni/N P Nf (Cycles) Ni/Nf
ai = 0.02 in af = 0.14 in
σr0 = 0 93 105 683 000 0.136 776 105 0.12
σr0 = 45 27 939 92 000 0.303 119 939 0.233

Table 4 – Summary of predicted fatigue lives under load P = 4 000 lb

Ni (Cycles) NP (Cycles)
Residual stress [ksi] Ni/NP Nf (Cycles) Ni/Nf
ai = 0.02 in af = 0.14 in
σr0 = 0 25 039 286 500 0.087 311 539 0.08
σr0 = 45 10 602 49 975 0.212 60 577 0.175

1106_0037_WELDING_7_8_2011.indd 14 14/06/11 10:35:34

 STRESS ANALYSIS and FATIGUE of welded structures

crack from its initial crack depth ai = 0.02 in to the final distribution and the weight function method can be used
one af = 0.14 in. for simulating the growth of fatigue cracks.

The residual stress effect seems also to be significant. The validation of the proposed technique resulted in
The final fatigue life was reduced by the residual stress by confirming good accuracy of the proposed method. In
approximately factor of 5. the case of the tube-on-tube welded joint subjected to
torsion and bending load the shell finite element model
The theoretical fatigue life assessments were generally in underestimated the peak stress at the analysed location
good agreement with the experimental data for both the by approximately 5 % in comparison with very detail 3D
low and high load levels. The predicted fatigue lives were fine mesh finite element analysis.
well inside the 95 % reliability scatter band.

It was also found that the predicted fatigue crack initiation

lives were sensitive to the choice of the weld toe radius r. References
However, it has been found that the variability of the weld
toe radius is not the only factor influencing the fatigue [1] Marshall P.W.: Design of welded tubular connections,
crack initiation life. It was also found that the frequency Elsevier, Amsterdam, 1992.
of occurrence of the smallest radius per unit length of
the weld toe line, i.e. how close or how frequent along [2] Dong P.: A structural stress definition and numeri-
the weld toe line were located the spots with the smallest cal implementation for fatigue analysis of welded joints,
weld toe radii, had also visible effect on the final fatigue International Journal of Fatigue, 2001, vol. 23, no. 10,
lives of tested welded joints. The distribution of spots with pp. 865-876.
the smallest weld toe radius, determining the proximity of
early fatigue crack initiation sites, had noticeable effect on [3] Niemi E.: Stress determination for fatigue analysis of
the initial fatigue crack shape evolution and subsequent welded components, Doc. IIS/IIW-1221-93, Abington
fatigue crack growth life. Publishing, Cambridge, UK, 1995.

[4] Ryachin V.A. and Moshkariev G.N.: Durability and sta-

bility of welded structures in construction and earth moving 15

10 Conclusions machinery (and road building machines), Mashinostroyenie,

Moskva, 1984 (in Russian).

An efficient shell finite element technique for obtain- [5] Trufyakov V.I. (editor): The strength of welded joints
ing stress data in welded structures relevant for fatigue under cyclic loading, Naukova Dumka, Kiev, ed. V. I., 1990
analyses has been proposed. According to the proposed (in Russian).
method the entire welded structure can be modelled using
a relatively small number of large shell finite elements. The [6] Young J.Y. and Lawrence F.V.: Analytical and graphical
modelling technique captures both the magnitude and the aids for the fatigue design of weldments, Fracture and
gradient of the hot spot stress near the weld toe which are Fatigue of Engineering Materials and Structures, 1985,
necessary for calculating the stress concentration and the vol. 8, no. 3, pp. 223-241.
peak stress at critical cross-sections, e.g. at the weld toe.
[7] Iida K. and Uemura T.: Stress concentration factor
A procedure for the determination of the magnitude of the formulas widely used in Japan, IIW Doc. XIII-1530-94,
peak stress at the weld toe using the classical stress con- 1994.
centration factors (one for axial load and one for bending)
[8] Fayard J.L., Bignonnet A. and Dang Van K.: Fatigue
has been laid proposed. The approach is based on the
design criteria for welded structures, Fatigue and
decomposition of the hot spot stress into the membrane
Fractures of Engineering Materials and Structures, 1996,
and bending contribution. The method can be success-
vol. 19, no. 6, pp. 723-729.
fully applied to any combination of loading and weldment
geometry. The stress concentration factors are used [9] Neuber H.: Theory of stress concentration for shear-
together with the hot spot membrane and bending stress strained prismatic bodies with arbitrary non-linear stress-
V hsm and V hsb at the location of interest in order to determine strain law, ASME Journal of Applied Mechanics, ASME,
the peak stress at the weld toe and the through-thickness 1961, vol. 28, pp. 544-551.
non-linear stress distribution.
[10] Molski K., Glinka G.: A method of elastic-plastic stress
The knowledge of the peaks stress at the weld toe and strain calculation at a notch root, Materials Science
enables application of the strain-life methodology for and Engineering, 1981, vol. 50, no. 1, pp. 93-100.
the assessment of the fatigue crack initiation life. The
through-thickness stress distribution is the base for calcu- [11] Manson S.S.: Behaviour of materials under condi-
lating stress intensity factors with the help of appropriate tions of thermal stress, 1953, NACA TN-2933 and Coffin
weight functions. Therefore the through-thickness stress L.F. Jr., Transactions of the ASME, 1954, vol. 76, p. 931.

N° 07
08 2011 Vol. 55 WELDING IN THE WORLD Peer-reviewed Section

1106_0037_WELDING_7_8_2011.indd 15 14/06/11 10:35:35

 STRESS ANALYSIS and FATIGUE of welded structures

[12] Smith KN, Watson P, Topper TH.: A stress–strain [18] Mikheevskiy S. and Glinka G.: Elastic–plastic fatigue
function for the fatigue of metals, Journal of Materials, crack growth analysis under variable amplitude loading
1970, vol. 5, no. 4, pp. 767-778. spectra, International Journal of Fatigue, 2009, vol. 31,
no. 11-12, pp. 1828-1836.
[13] Monahan C.C.: Early fatigue cracks growth at welds,
Computational Mechanics Publications, Southampton UK, [19] Murakami Y.: Stress intensity factors handbook,
1995. 1987, vol. 2, Pergamon Press, Oxford.

[14] Bueckner H.F.: A novel principle for the computa-

[20] Glinka G.: Residual stresses in fatigue and frac-
tion of stress intensity factors, Zeitschrift fur Angewandte
ture: Theoretical analyses and experiments, Advances in
Mathematik Und Mechanik, 1970, vol. 50, pp. 529-546.
Surface Treatments, vol. 4, A. Niku-Lari, Ed., International
Journal of Residual Stresses, 1987, pp. 413-454.
[15] Glinka G. and Shen G.: Universal features of weight
functions for cracks in Mode I, Engineering Fracture
Mechanics, 1991, vol. 40, no. 6, pp. 1135-1146. [21] Glinka G.: Effect of residual stresses on fatigue
crack growth in steel weldments under constant and vari-
[16] Paris P.C. and Erdogan F.: A critical analysis of crack able amplitude loading, Fracture Mechanics, ASTM STP
propagation laws, Journal of Basic Engineering, 1963, 677, C.W. Smith, Ed., American Society for Testing and
no. D85, pp. 528-534. Materials, 1979, pp. 198-214.

[17] Noroozi A.H., Glinka G. and Lambert S.: A two [22] Kurihara M., Katoh A. and Kwaahara M.: Analysis on
parameter driving force for fatigue crack growth analysis, fatigue crack growth rates under a wide range of stress
International Journal of Fatigue, 2005, vol. 27, no. 10-12, ratio, Journal of Pressure Vessel Technology, Transactions
pp. 1277-1296. of the ASME, May 1986, vol. 108, no. 2, pp. 209-213.

16

1106_0037_WELDING_7_8_2011.indd 16 14/06/11 10:35:35

 STRESS ANALYSIS and FATIGUE of welded structures

11 Appendix - Selected one-dimensional (1D) weight functions for cracks in plates

Figure A1 – Weight function notations for a single edge and central through crack in finite with plate

Single edge crack in a finite with plate (Figure A1, valid for 0 < a/t < 0.9) 17

­ 1 3
½
2F ° § x ·2 § x· § x ·2 °
m  x, a K F
®1  M1 ¨ 1  ¸  M2 ¨ 1  ¸  M3 ¨ 1  ¸ ¾ (A1)
2S  a  x  °
A
© a¹ © a¹ © a¹ °
¯ ¿

a§ a§ a§ a ···
0.029207  ¨ 0.213074  ¨ 3.029553  ¨ 5.901933  2.657820 ¸ ¸ ¸
t© t© t© t ¹¹¹
M1
a § a § a§ a§ a ···· (A2)
1.0+ ¨¨ 1.259723  ¨ 0.048475  ¨ 0.481250  ¨ 0.526796  0.345012 ¸ ¸ ¸ ¸¸
t© t© t© t© t ¹¹¹¹

a§ a§ a§ a ···
0.451116  ¨ 3.462425  ¨ 1.078459  ¨ 3.558573  7.553533 ¸ ¸ ¸
t© t© t© t ¹¹¹
M2 (A3)
a§ a§ a§ a§ a ····
1.0+ ¨¨ 1.496612  ¨ 0.764586  ¨ 0.659316  ¨ 0.258506  0.114568 ¸ ¸ ¸ ¸¸
t© t© t© t© t ¹¹¹¹

a§ a§ a§ a ···
0.427195  ¨ 3.730114  ¨ 16.276333  ¨ 18.799956  14.112118 ¸ ¸ ¸
t© t© t© t ¹¹¹
M3 (A4)
a§ a§ a§ a§ a ····
1.0+ ¨¨ 1.129189  ¨ 0.033758  ¨ 0.192114  ¨ 0.658242  0.554666 ¸ ¸ ¸ ¸¸
t© t© t© t© t ¹¹¹¹

Central through crack under symmetric stress field (Figure A1, valid for 0 < a/t < 0.9)

­ 1 3
½
2F ° § x ·2 § x· § x ·2 °
m  x, a F
®  1¨      ¾ (A5)
a ¸¹
2¨
a ¸¹
3¨
a ¸¹ °
K 1 M 1 M 1 M 1
2S  a  x  °
A
© © ©
¯ ¿
4 5 6 7
§a· §a· §a· §a·
M1 m1  200.699 ¨ ¸  395.552 ¨ ¸  377.939 ¨ ¸  140.218 ¨ ¸ (A6)
©t ¹ ©t ¹ ©t ¹ ©t ¹

N° 07
08 2011 Vol. 55 WELDING IN THE WORLD Peer-reviewed Section

1106_0037_WELDING_7_8_2011.indd 17 14/06/11 10:35:35

 STRESS ANALYSIS and FATIGUE of welded structures

2 3
§a· §a· §a·
m1 0.06987  0.40117 ¨ ¸  5.5407 ¨ ¸  50.0886 ¨ ¸
©t ¹ ©t ¹ ©t ¹
4 5 6
§a· §a· §a·
M2 m2  210.599 ¨ ¸  239.445 ¨ ¸  111.128 ¨ ¸ (A7)
© ¹
t © ¹
t ©t ¹
2 3
§a· §a· §a·
m2 0.09049  2.14886 ¨ ¸  22.5325 ¨ ¸  89.6553 ¨ ¸
©t ¹ ©t ¹ ©t ¹

4 5 6 7
§a· §a· §a· §a·
M3 m3  347.255 ¨ ¸  457.128 ¨ ¸  295.882 ¨ ¸  68.1575 ¨ ¸ (A8)
©t ¹ ©t ¹ ©t ¹ ©t ¹
2 3
§a· §a· §a·
m3 0.427216  2.56001¨ ¸  29.6349 ¨ ¸  138.4 ¨ ¸
©t ¹ ©t ¹ ©t ¹

Surface semi-elliptical crack (valid for 0 < a/t < 0.8, and 0 < a/c < 1, Figure A2)

18

Figure A2 – Weight function notation for a semi-elliptical crack in finite thickness plate

– For the deepest point A

­ 1 3
½
2F ° § x ·2 § x· § x ·2 °
mA  x , a  ®  1¨      ¾
a ¸¹
2¨
a ¸¹
3¨
a ¸¹ °
1 M 1 M 1 M 1 (A9)
2S  a  x  ° © © ©
¯ ¿
S 24
M 1A = (4Y 0  6Y 1 )  (A10)
2Q 5

M2A = 3 (A11)

§ S ·
M 3A = 2 ¨ Y 0  M 1A  4 ¸ (A12)
© 2Q ¹

where, for 0 < a/c < 1:

1.65
§a·
Q = 1.0 + 1.464 ¨ ¸ (A13)
©c ¹
2 4 6
§a· §a· §a·
Y 0 = B 0 + B1 ¨ ¸ + B 2 ¨ ¸ + B 3 ¨ ¸ (A14)
©t ¹ ©t ¹ ©t ¹
2 3
§a· §a· §a·
B 0 = 1.0929 + 0.2581¨ ¸  0.7703 ¨ ¸ + 0.4394 ¨ ¸
©c ¹ ©c ¹ ©c ¹ (A15)

1106_0037_WELDING_7_8_2011.indd 18 14/06/11 10:35:36

 STRESS ANALYSIS and FATIGUE of welded structures

2
§a· §a· 1.0
B1 = 0.456  3.045 ¨ ¸ + 2.007 ¨ ¸ +
©c ¹ ©c ¹
0.688
§a· (A16)
0.147 + ¨ ¸
©c ¹
9.953
1.0 § a·
B 2 = 0.995  + 22.0 ¨1 ¸
§a· © c¹ (A17)
0.027 + ¨ ¸
© ¹
c
8.071
1.0 § a·
B 3 =  1.459   24.211¨1 ¸
0.014 +
a © c¹ (A18)
c

and
2 4 6
§a· §a· §a·
Y 1 = A 0 + A1¨ ¸ + A2 ¨ ¸ + A 3 ¨ ¸ (A19)
© ¹
t © ¹
t ©t ¹
2 3
§a· §a· §a·
A 0 = 0.4537 + 0.1231¨ ¸  0.7412 ¨ ¸ + 0.4600 ¨ ¸ (A20)
© ¹
c © ¹
c ©c¹
2
§a· §a· 1.0
A1=  1.652  1.665 ¨ ¸  0.534 ¨ ¸ 
©c ¹ ©c ¹
0.846
§a· (A21)
0.198  ¨ ¸
©c ¹
9.286
§a· 1.0 § a·
A2 3.418  3.126 ¨ ¸   17.259 ¨ 1.0  ¸
© ¹ 0.041
c § a · © c ¹ (A22)
¨ ¸
©c ¹
9.203
§a· 1.0 § a·
A3  4.228  3.643 ¨ ¸   21.924 ¨1.0  ¸ 19
© c ¹ 0.020  a © c¹ (A23)
c

and for 1 < a/c < 2

1.65 2
§c · §a·
Q = 1.0 + 1 .464 ¨ ¸ ¨ ¸
©a¹ ©c¹ (A24)
2 4
§a· §a·
Y 0 = B 0 + B 1¨ ¸ + B 2 ¨ ¸ (A25)
© ¹
t ©t ¹
2
§a· §a·
B 0 = 1.12  0.09923 ¨ ¸  0.02954 ¨ ¸ (A26)
©c ¹ ©c¹
2
§a· §a·
B1 = 1.138  1.134 ¨ ¸ + 0.3073 ¨ ¸ (A27)
©c ¹ ©c ¹
2
§a· §a·
B 2 = 0.9502  0.8832 ¨ ¸  0.2259 ¨ ¸ (A28)
© ¹
c ©c ¹
2 4
§a· §a·
Y 1 = A 0 + A1 ¨ ¸ + A 2 ¨ ¸ (A29)
©t ¹ ©t ¹
2
§a· §a·
A 0 = 0.4735  0.2053 ¨ ¸  0.03662 ¨ ¸ (A30)
© ¹
c ©c¹
2
§a· §a·
A1 = 0.7723  0.7265 ¨ ¸  0.1837 ¨ ¸ (A31)
©c ¹ ©c ¹
2 3
§a· §a· §a·
A2 = 0.2006  0.9829 ¨ ¸  1.237 ¨ ¸  0.3554 ¨ ¸ (A32)
©c ¹ ©c ¹ ©c ¹

N° 07
08 2011 Vol. 55 WELDING IN THE WORLD Peer-reviewed Section

1106_0037_WELDING_7_8_2011.indd 19 14/06/11 10:35:37

 STRESS ANALYSIS and FATIGUE of welded structures

– For the surface point B

ª 1
1
3
º
2F « § x ·2 §x· § x ·2 »
mB  x , a  1  M1B ¨ ¸  M2 B ¨ ¸  M3 B ¨ ¸
Sx« ©a¹ ©a¹ ©a¹ » (A33)
¬ ¼
S
M 1B (30F1  18F0 )  8
4Q (A34)

S
M 2B = (60F0  90F1 ) + 15
4Q (A35)

M3B = − (1 + M1B + M1B) (A36)

where for 0 < a/c < 1:

ª §a·
2
§a· º a
4

F0  
«C 0 C 1¨ ¸ C 2 ¨ ¸ » (A37)
¬« ©t ¹ © t ¹ ¼» c
2
§a· §a·
C 0 1.2972  0.1548 ¨ ¸  0.0185 ¨ ¸ (A38)
©c ¹ ©c¹
2
§a· §a·
C1 1.5083  1.3219 ¨ ¸  0.5128 ¨ ¸ (A39)
© ¹
c ©c ¹
0.879
C2  1.101
a (A40)
0.157 
c
and
20
ª §a·
2
§a· º a
4

F1 « 0
D  D1¨ ¸  D2¨ ¸ » (A41)
¬« ©t ¹ © t ¹ ¼» c
2 3
§a· §a· §a·
D0 1.2687  1.0642 ¨ ¸  1.4646 ¨ ¸  0.7250 ¨ ¸ (A42)
© ¹
c © ¹
c ©c ¹
2
§a· §a·
D1 1.1207  1.2289 ¨ ¸  0.5876 ¨ ¸ (A43)
© ¹
c ©c¹
§a· 0.199
D2 0.190  0.608 ¨ ¸ 
© c ¹ 0.035  a (A44)
c

and for 1 < a/c < 2

ª §a·
2
§a· º a
4

F0 «C 0  C 1¨ ¸  C 2 ¨ ¸ » (A45)
«¬ ©t ¹ © t ¹ »¼ c
2
§a· §a·
C 0 = 1.34  0.2872 ¨ ¸  0.0661¨ ¸ (A46)
©c ¹ ©c ¹
2
§a· §a·
C 1 = 1.882  1.7569 ¨ ¸  0.4423 ¨ ¸ (A47)
©c¹ ©c¹
2
§a· §a·
C 2 =  0.1493  0.01208 ¨ ¸  0.02215 ¨ ¸ (A48)
©c ¹ ©c ¹

and
ª §a·
2
§a· º a
4

F 1 «D 0  D 1¨ ¸  D 2 ¨ ¸ »
«¬ ©t ¹ © t ¹ »¼ c (A49)

1106_0037_WELDING_7_8_2011.indd 20 14/06/11 10:35:38

 STRESS ANALYSIS and FATIGUE of welded structures

2
§a· §a·
D 0 = 1.12  0.2442 ¨ ¸  0.06708 ¨ ¸ (A50)
©c ¹ ©c¹
2
§a· §a·
D 1 = 1.251 1.173 ¨ ¸  0.2973 ¨ ¸ (A51)
©c ¹ ©c ¹
2
§a· §a·
D 2 = 0.04706  0.1214 ¨ ¸  0.04406 ¨ ¸ (A52)
©c¹ ©c ¹

About the authors

Mr. Aditya CHATTOPADHYAY ([email protected]) is with University of
Waterloo, Faculty of Engineering, Waterloo, Ontario (Canada). Dr. Grzegorz GLINKA
([email protected]) is also with University of Waterloo, Faculty of
Engineering, Waterloo, Ontario (Canada) as well as with Aalto University, Helsinki
(Finland). Dr. Mohamad EL-ZEIN ([email protected]), Dr. Jin
QIAN ([email protected]) and Mr. Rodrigo FORMAS (FormasRodrigoG@
JohnDeere.com) are all with Deere & Company World Headquarters, Moline,
Illinois (United States)

21

N° 07
05
06 2011 Vol. 55 WELDING IN THE WORLD
08 Peer-reviewed Section

1106_0037_WELDING_7_8_2011.indd
View publication stats 21 14/06/11 10:35:40