Professional Documents
Culture Documents
www.elsevier.com/locate/ijfatigue
Received 10 December 2000; received in revised form 11 May 2001; accepted 12 June 2001
Abstract
A mesh-size insensitive structural stress definition is presented in this paper. The structural stress definition is consistent with
elementary structural mechanics theory and provides an effective measure of a stress state that pertains to fatigue behavior of
welded joints in the form of both membrane and bending components. Numerical procedures for both solid models and shell or
plate element models are presented to demonstrate the mesh-size insensitivity in extracting the structural stress parameter. Conven-
tional finite element models can be directly used with the structural stress calculation as a post-processing procedure. To further
illustrate the effectiveness of the present structural stress procedures, a collection of existing weld S-N data for various joint types
were processed using the current structural stress procedures. The results strongly suggests that weld classification based S-N curves
can be significantly reduced into possibly a single master S-N curve, in which the slope of the S-N curve is determined by the
relative composition of the membrane and bending components of the structural stress parameter. The effects of membrane and
bending on S-N behaviors can be addressed by introducing an equivalent stress intensity factor based parameter using the structural
stress components. Among other things, the two major implications are: (a) structural stresses pertaining to weld fatigue behavior
can be consistently calculated in a mesh-insensitive manner regardless of types of finite element models; (b) transferability of weld
S-N test data, regardless of welded joint types and loading modes, can be established using the structural stress based parameters.
2001 Elsevier Science Ltd. All rights reserved.
Keywords: Structural stress; Finite element analysis; Welded joints; Fatigue; Notch stress; Stress concentration; Mesh-size sensitivity
0142-1123/01/$ - see front matter 2001 Elsevier Science Ltd. All rights reserved.
PII: S 0 1 4 2 - 1 1 2 3 ( 0 1 ) 0 0 0 5 5 - X
866 P. Dong / International Journal of Fatigue 23 (2001) 865–876
assumption that the surface stresses on a structural mem- joints cannot be captured in this extrapolation scheme,
ber provides an indication of the stress state at a weld resulting in little stress concentration effects from this
fatigue prone location, such as a weld toe. This underly- calculation. On the other hand, for Type II joints, Fig.
ing assumption may become questionable if the struc- 1(b) shows that extrapolation from the two reference
tural member is not a dominant load-transfer member in positions (open circles) should provide some indication
a joint. Under such circumstances, the surface stresses of the concentrations at the weld toes. Then, one obvious
at some distance away from a weld toe may not be rel- question is if such calculation procedures provide a
evant to the stress state of concern. In addition, a refer- reliable stress concentration measurement or hot spot
ence nominal stress in such a structural member may not stresses. As discussed in Neimi [9], the results are often
be readily identified for conventional SCF calculations. questionable due to the fact that these stresses can be
Among the various extrapolation procedures proposed strongly dependent on mesh-size and loading modes.
in the open literature (e.g., [7,8]), a typical one is based To improve the S-N curve approach (using either
on a linear extrapolation from stress values at both 0.4t nominal stresses or hot spot stresses) for welded struc-
and 1t from a weld toe [8,9], as shown in Fig. 1, where tures, a relevant stress parameter must satisfy the two
t represents the plate thickness of a structural member. basic requirements: (a) mesh-size insensitivity in finite
The drawback in such an extrapolation scheme becomes element solutions; (b) ability to differentiate stress con-
immediately clear in view of Fig. 1 in which some of centration effects in different joint types (e.g., butt joints
the well-studied joints in the research community are versus T-fillet cruciform joints) in welded structures. In
illustrated. The stress concentration behaviors can be the following, such a stress parameter is presented and
categorized into two types [10]: one is rather localized the corresponding finite element procedures using both
stress concentration behavior (Type I) at weld toe, while solid and shell element models are given. The validation
the other is more global in length-scale (Type II). In of such a structural stress parameter is demonstrated by
order to correlate the fatigue behavior in various joint reprocessing a series of existing S-N data for joint types
types, stress concentration behavior at the weld toe of listed in Fig. 1(a).
various joint types must be captured. However, as shown
in Fig. 1(b), any stress concentration effects in Type I
2. Structural stress definition and formulation
冕
t
1
sm⫽ sx(y)·dy (2)
t
0
冕 冕
Fig. 2. Structural stresses definition for through-thickness fatigue t t
crack. (a) Local through-thickness normal and shear stress at weld toe, t2 t2
(b) Structural stress definition at weld toe. sm· ⫹sb· ⫽ sx(y)·y·dy⫹d txy(y)·dy. (3)
2 6
0 0
should eliminate or minimize the mesh-size sensi-
Eq. (2) represents the force balances in x direction,
tivity in the structural stress calculations. This is due
evaluated along B–B and Eq. (3) represents moment bal-
to the fact that the local stress concentration close
to a notch is dominated by self-equilibrating stress
distribution, as discussed by Niemi [9].
ances with respect to Section A–A at y =0. The integral at failure is assumed to be t1. Without losing generality,
term on the right-hand side of Eq. (3) represents the the structural stress procedures [10,11] can be effectively
transverse shear force as an important component of the demonstrated using the example in Fig. 4. Note that for
structural stress definition. It is then follows that if convenience, the local y coordinate is defined as shown
element size (d) is small or transverse shear is negligible, in Fig. 4(a), different from that in Fig. 3, At a horizontal
the integral representations of sb and sm in Eqs. (2) and cross section of depth t1 from the top surface, both nor-
(3) can be directly evaluated at Section A–A in Fig. 3(a). mal stress (sy) and shear stress (tyx) are present in gen-
eral.
2.2. Solid model with finite fatigue crack depth By imposing equilibrium conditions between Sections
A–A and B–B, as well as the horizontal cross section
Often, a fatigue crack of a finite depth is used as a in between, it can be shown that the structural stress
fatigue failure criterion (e.g. [9]), the corresponding components (sb and sm) must satisfy the following equa-
structural stress can be then defined in a similar manner tions:
to that in Fig. 3. In Fig. 4, the depth of the fatigue crack
冕 冕
t1 d
1 1
sm⫽ sx(y)·dy⫹ tyx(x)·dx (4)
t1 t1
0 0
冕 冕
t1 t1
t21 t21
sm· ⫹sb· ⫽ sx(y)·y·dy⫹d txy(y)·dy (5)
2 6
0 0
冕
d
⫹ sy(x)·x·dx.
0
冕 冕
t⫺t1 d
1 1
sm⬘⫽ sx(y)·dy⫺ t (x)·dx. (7)
t−t1 t−t1 yx
0 0
2.4.1. Using stresses and stress resultants described by Eq. (9) are automatically satisfied within
As shown in Fig. 7, stresses and nodal quantities from the accuracy of the finite element solutions. In view of
shell or plate element models are often defined in a glo- typical finite element procedures in commercial codes,
bal coordinate system (x, y, z), depending on the finite a general structural stress calculation procedure is
element codes used. Given the definition of the structural presented below:
stress components in Eq. (1), it is the local coordinate With respect to the global coordinate system (x, y, z),
system (x⬘, y⬘, z⬘) that is convenient for calculating the the element stiffness matrix {Ke} can be obtained either
structural stresses with respect to a weld, with local x⬘ directly from a finite element solution or formulated sep-
and y⬘ being perpendicular and parallel to the weld direc- arately afterwards. The nodal displacements at a node
tion, respectively. Consistent with the solid element within the reference element are typically described in
model approach (e.g., see Fig. 3), three components of the form of:
the stress resultants (sectional forces and moments), i.e.,
fx⬘, fz⬘, and my⬘, at Section B–B in Fig. 7 can be used to {u}Ti ⫽{uxi, uii, uzi, qxi, qyi, qzi}
calculate the structural stress components at Section where uxi, uyi, uzi represent the three translational dis-
A–A: placements in x, y, and z directions at node i and qxi,
fx⬘ 6(my⬘+d·fz⬘) qyi, qzi three rotational displacements, respectively. The
ss⫽sm⫹sb⫽ ⫹ . (9) subscript i takes 1, 2, …, n, with n being the number of
t t2
the nodes in the element. The element nodal force vector,
In the above, a finite element formulation with six
degrees of freedom at each node is assumed, i.e., six {Fe}Ti ⫽{Fxi, Fyi, Fzi, Mxi, Myi, Mzi, …}, i⫽1,2,…,n
components of generalized forces at each node (three
can be obtained by:
translational and three rotational). If stresses in the glo-
bal coordinate system (x, y, z) are used, they must be {Fe}⫽{Ke}{u}. (10)
transformed to the local coordinate system (x⬘, y⬘, z⬘)
before Eqs. (2) and (3) can be used for the structural The element nodal forces in the local coordinate system
stress calculations. (x⬘, y⬘, z⬘) can then be computed as
{Fe⬘}⫽{T}{Fe} (11)
2.4.2. Using element nodal forces
In some applications, the reference section B–B in where the matrix {T} is the coordinate transformation
Fig. 7 may not be available. This situation arises if welds matrix built up of directional cosines of angles formed
are rather close to each other or load transfer at a weld between the two sets of axes in Fig. 7.
of interest is very localized. If the element sectional Once the element nodal forces are obtained for the
forces and moments (with respect to the reference nodal positions along Section A–A in Fig. 7, the corre-
element in Fig. 7) at Section A–A are available from sponding sectional forces and moments (fx⬘, fz⬘, and
a finite element solution, the equilibrium requirements my⬘) can be calculated using appropriate shape functions.
Then, Eq. (9) simply becomes at Section A–A:
fx⬘ 6my⬘
ss⫽sm⫹sb⫽ ⫹ 2 . (12)
t t
Note that transverse shear effects in Eq. (9) are already
taken into account in the finite element solution in this
instance.
3. Numerical examples
3.1. Plate lap fillet weld with respect to remote nominal stress, i.e., F/A, where
F is remote loading and A the area of loaded member.
A typical lap fillet joint is shown in Fig. 8(a). A plane It is important to note that once the mesh-size insensi-
strain eight-node solid element model is used as shown tivity is ensured, the structural stress should serve as an
in Fig. 8(b) and (c), illustrating two representative intrinsic stress parameter for a given geometry and
meshes with drastically different element sizes at the boundary conditions, regardless of numerical procedures
weld toes, among those used for mesh-size sensitivity used. Then, it is natural to expect a similar structural
investigation. Since large element sizes are to be investi- stress value from a shell/plate element model for a geo-
gated, parabolic elements with reduced integration are metrically similar joint. Indeed, this is the case, as evi-
used for this joint, due to shear lock considerations if a denced in Fig. 9. The structural stress values obtained
strong bending action is present [11]. using the two shell element procedures, i.e., Eqs. (9) and
As the element size (a/b) varies from a/b=0.16t/0.1t (12), give the same results. The slightly higher structural
to a/b=2t/1t, the structural stresses calculated at the weld stress values than those obtained by solid element mod-
toe according Eqs. (2) and (3) remains essentially ident- els reflect, to a large extent, the simple representation of
ical. Fig. 8(d) summarizes the structural stress based the fillet weld geometry in the shell model, as shown in
SCF values calculated with different element mesh sizes. Fig. 9(a). A proper definition of the shell element thick-
The SCF values were calculated using the structural ness in the weld area should further improve the shell
stresses calculated using the present methods normalized element results. Attempts were not made here to optim-
ize some of the detailed modeling issues.
Fig. 11. Structural stress at weld toe for RHS joint (shell element
Fig. 10. Structural stress and mesh insensivity–double plate lap fillet model). (a) Stress distribution - Shell Element model, (b) comparison
weld: (a) model definition; (b) structural stresses calculated from four of local stress and structural stresses.
FE models with various mesh-sizes at weld toe.
detailed considerations of various other joint types and using existing solutions such as those given by Tada et
corresponding S-N data. al., [15]. It can be shown that the Mode I stress intensity
factor range for the crack geometry in Fig. 13 can be
4.3. Effects of bending and membrane components expressed as a function of the ranges of the structural
stress components using superposition principle:
With a S-N curve approach, fatigue lives (in cycles)
of welded joints are described as a function of either
nominal or hot spot stress ranges. In the present
approach as shown in Fig. 12(c), structural stress ranges
⌬K⫽⌬Km⫹⌬Kb⫽冑tr ⌬smfm 冋 冉冊 a
tr
⫹⌬sbfb 冉 冊册
a
tr
(13)
are used as a stress parameter to relate to fatigue lives The fm(a/tr) and fb(a/tr) are dimensionless functions of
of the welded joints under remote tension. For all the a/tr for the membrane and bending components,
joints analyzed in Fig. 12(c) under remote tension, the expressed as follows [15]:
bending content (sb) of the structural stresses at weld
toe varies from 0.214 to in Joint A to 0.522 to Joint F,
冪2tan2t
pa
冉冊 冋 冉冊 冉 冊册
normalized by remote tension stress. It is conceivable 3
that the peak surface stress range parameter (sm+sb), a a pa r
fm ⫽ 0.752⫹2.02 ⫹0.37 1⫺sin (14)
although it can be calculated in a mesh-insensitive man- tr tr 2tr pa
cos
ner, may not be adequate to consolidate S-N data from 2tr
drastically different remote loading modes, such as pure
remote tension versus pure remote bending. In such situ-
冪2tan2t .
pa
ations, the corresponding structural stress components
sm and sb will be drastically different as well. The
effects of the membrane and bending effects on S-N
fb冉冊 冋
a
tr
⫽ 0.923⫹0.199 1⫺sin
pa
冉
2tr 冊册
4
cos
pa
r
(15)
a/tr⫽1
⌬K⫽冑tr 冕冋 冉冊
a/tr⫽0
⌬smfm
a
tr
⫹⌬sbfb冉 冊册 冉 冊
a
tr
d
a
tr
.
⌬K⫽冑tr(33.17⌬sm⫹11.87⌬sb). (16)