You are on page 1of 11

International Journal of Fatigue 27 (2005) 383–393

www.elsevier.com/locate/ijfatigue

Fatigue of web-flange weld of corrugated web girders:


2. Analytical evaluation of fatigue strength of corrugated web-flange weld
Kengo Anamia,1,*, Richard Sauseb,2
a
Department of Infrastructure Systems Engineering, Kochi University of Technology, Tosayamada-cho, Kochi 782-8502, Japan
b
ATLSS Center, Lehigh University, 117 ATLSS Drive, Bethlehem, PA 18105, USA
Received 10 March 2004; received in revised form 15 June 2004; accepted 6 August 2004

Abstract
The challenges in evaluating the fatigue strength of the web-flange weld of corrugated web girders are the complexity of the stress field and
uncertainty regarding the appropriate fatigue strength category. In this paper, using finite element analysis and crack propagation analysis,
the fatigue strength of this web-flange weld is analytically examined. The main conclusions are that the fatigue strength is negatively affected
by the existence of the longitudinal folds of the corrugated web and that it is necessary to have a large bend radius so that the fatigue critical
point on the inclined folds are far from the longitudinal folds to eliminate the influence of the longitudinal folds. In addition, simple methods
to estimate the predict fatigue life of the web-flange welds of corrugated web girders are proposed.
q 2004 Elsevier Ltd. All rights reserved.

Keywords: Corrugated web girder; Geometry effect; Fatigue strength prediction; Crack propagation analysis; Weld toe stress; Hot spot stress

1. Introduction show that secondary in-plane bending and out-of-plane


bending in the flange plate occur even in a region of constant
The application of steel corrugated web girders to primary bending moment in a corrugated web girder. This
bridges has been a focus of recent research. Several studies secondary bending behavior is influenced by the corrugation
of the fatigue strength of the weld between the corrugated geometry, defined by parameters such as the corrugation
web and the steel flange plates of a corrugated web girder angle (q), bend radius (R) and corrugation depth (H), and the
have been carried out. However, the applicability of the flange geometry, defined by the flange thickness (T) and
results of these studies were limited to the specific width (W). Thus, the nominal stress range calculated from
corrugation geometries used in these studies, and a general beam theory does not accurately describe/represent the
method for evaluating the fatigue strength of the corrugated relevant stress conditions.
web-flange weld has not been established. Evaluation of the The difficulty in establishing the fatigue strength
fatigue strength of this weld is difficult because of category is illustrated in Fig. 1, which compares corrugated
the complex stress field in the flange plate, and because web-flange welds to several other welded joints. Usually,
the fatigue strength category is unknown. The authors have the fatigue strength of the corrugated web-flange weld is
analytically investigated the stress condition in the vicinity recognized to be between the fatigue strengths of a
of the web-flange weld toe along the inclined fold of the longitudinal web-flange weld (joint a-1 in Fig. 1), which
corrugated web (in Part 1 of this report [1]). The results falls in AASHTO fatigue strength Category B or B 0 [2], and
a cruciform or T-form attachment weld (joint a-2), which
* Corresponding author. Fax: C81 887 57 2420. falls in Category C. However, experimental and analytical
E-mail addresses: anami.kengo@kochi-tech.ac.jp (K. Anami), results in [1] indicate that the maximum stress appears at the
rs0c@lehigh.edu (R. Sause).
1 end of the flat part of the inclined fold (point S, see Fig. 2),
Formerly, Research Associate, ATLSS Center, Lehigh University, 117
ATLSS Drive, Bethlehem, PA 18105, USA. because the curved bend region and the longitudinal web
2
Fax: C1 610 758 5902. fold influence the stress conditions near point S.
0142-1123/$ - see front matter q 2004 Elsevier Ltd. All rights reserved.
doi:10.1016/j.ijfatigue.2004.08.007
384 K. Anami, R. Sause / International Journal of Fatigue 27 (2005) 383–393

List of symbols
q corrugation angle, angle of attachment inclined Ac coefficient (ZNfs3n) expressing Category C
to primary stress fatigue strength
R bend radius of transition between inclined web A0c coefficient (ZNfs3n) depending on definition of
fold and longitudinal web fold hot spot stress
H corrugation depth Nf100 fatigue life from crack propagation analysis
H1 distance between point S and longitudinal web under 100 MPa stress range applied
fold Nf fatigue life
H01 magnitude of H1 to eliminate influence of a0 initial crack size
longitudinal fold on stress at point S M,c constants used for crack propagation analysis
T flange thickness or base plate thickness DK, DKth stress intensity factor range and threshold level
W flange width of stress intensity factor range
r toe radius N number of loading cycles
qtoe toe angle b,b1 constants used for fatigue life prediction
SCF, SCF-1, SCF-2 hot spot stress concentration factor, d(q) fatigue life of a-3 weld joint
SCF-1 is based on [12] and SCF-2 is based on sn,slongitudinal nominal stress, longitudinal stress (range)
[13]

Ifthe corrugated web is assumed to behave as an attachment stiffness, can diminish the secondary bending stresses. In the
and not carry longitudinal bending stress, then the present study, a 450 (W)!40 (T) mm flange plate, which is
corrugated web-flange weld is similar to a T-form welded a practical flange plate size for a bridge girder, is used for
joint with an inclined attachment (joint a-3) in the inclined the analysis model. Other geometric parameters are shown
web fold region, and is similar to an out-of-plane gusset in Fig. 2, with an example FEM model, which uses 8-node
welded joint (joint b-1) with bend regions at the ends (joint solid elements with a minimum mesh size at the weld toe of
b-3) in the longitudinal web fold region. However, the about 0.2 mm. The parameters examined are q and R, which
fatigue strength of joint a-3 and joint b-3 are also unknown, are the parameters dominating the stress conditions at the
even though fatigue strength categories are established for weld toe [1]. The toe radius (r) and angle (qtoe) are assumed
joints a-1 and a-2, and joint b-1 and b-2. Fisher [3] recently to be 1 mm and 1358, respectively. Uniform tensile loading
investigated the fatigue strength of the joint with an H-shape is applied at the edge of the model, where no web exists, as
welded attachment (joint b-2) and categorized it as shown in Fig. 2. This model is reasonable assuming the
AASHTO fatigue strength Category D or E. Recently, corrugated web does not carry longitudinal stress. This
Yamada [4,5] carried out fatigue tests of joint a-3, but the model eliminates the secondary flange bending behavior
fatigue strength, especially when the inclined angle is small, that occurs in a corrugated web girder and allows the fatigue
is not well established. Consequently, the fatigue strength of strength category of the joint itself to be discussed.
the corrugated web-flange weld joint has not been fully
established. This paper addresses the fatigue strength
category for corrugated web-flange welds using finite
element method (FEM) analysis and a simple crack
propagation analysis. In addition, simple methods to
estimate the fatigue strength of corrugated web-flange
welds are proposed.

2. Analytical models

2.1. FEM models

It was reported in [1] that secondary in-plane bending


and out-of-plane bending behavior occurs in the flange plate
of a corrugated web girder. This secondary bending occurs
both in a region of combined shear and bending moment and
in a region of constant bending moment. Increasing the W Fig. 1. Schematic fatigue strength of corrugated web-flange joint compared
and T of the flange plate, which increases the flange bending with AASHTO fatigue strength categories.
K. Anami, R. Sause / International Journal of Fatigue 27 (2005) 383–393 385

(meter) for crack length and MPa m1/2 for DK. In the
analyses, a threshold level of crack intensity factor (DKth)
was not considered. For the calculation of DK, the method
proposed by Albrecht [7] was used. Since, according to the
fatigue test reported in [1], all fatigue cracks were found to
initiate from the weld toe, only fatigue cracking from the
weld toe was considered in the study. In addition, the study
in [1] indicated that the maximum stress at the web-flange
weld toe always appears at point S, at the intersection of the
inclined fold and the bend region (see Fig. 2), and
the fatigue crack initiating from this point led to failure of
the test specimens. Thus, it was also assumed that a unique
crack leading to failure initiated at point S in the crack
propagation analyses.
The initial crack shape was assumed to be semi-circular.
The initial crack size, a0 (Z0.17 mm), was determined by
calibration using the fatigue strength of the T-form weld
joint (joint a-2), where the fatigue strength of this joint was
selected to be the lower bound design fatigue strength of
AASHTO Category C weld joints. The thickness of the base
Fig. 2. Simple tension FEM model.
plate and the thickness of the attached plate of this T-form
weld joint were the same as the flange thickness and web
This model is called the ‘simple tension model’ in this
thickness of the simple tension model, respectively. The
study. Although the finite element (FE) mesh was not fine
mesh discritization in the vicinity of the weld toe of this T-
enough to very accurately calculate the stress at the weld
toe, a comparison of the toe stresses between models is form weld joint was also the same as that of the simple
possible because the mesh size near the weld toe, especially tension model. In some of the crack propagation analyses of
around point S, was kept the same in all FEM models. the simple tension model, it was assumed that the crack
initiating at point S propagated perpendicular to the applied
stress (longitudinal stress) direction, and this crack propa-
2.2. Crack propagation model gation path is called ‘path-1’. In this case (path-1), as q is
changed in the analyses, the surface length of the initial
Only mode-I crack propagation was considered and crack was determined by projecting the initial crack to a
Paris’ law (Eq. (1)) was used for the crack propagation plane perpendicular to the applied stress [8]. In other crack
analyses propagation analyses, it was assumed that the crack
propagates along the toe of the weld, which is called
da
Z CðDKÞm (1) ‘path-2’. These other analyses were carried out to show the
dN
influence of the propagation direction on fatigue strength. In
where a is the crack length, N is the number of loading both cases, the stresses perpendicular to the crack
cycles, C and m are constants, and DK is the stress intensity propagation path obtained from FEM analyses were used
factor range. for the calculation of DK.
For the constants C and m in Eq. (1), the values of The unsymmetrical stress distributions at the flange
the most conservative design curve recommended by the surface obtained from FEM analyses were simplified to
Japan Society of Steel Construction [6] were used, which symmetrical distributions as shown in Fig. 3. In Fig. 3, ‘dist.
are 2.7!10K11 for C and 2.75 for m, using units of m 1’ and ‘dist. 2’ are the stress distributions obtained by FEM.

Fig. 3. Stress simplification for crack propagation analysis.


386 K. Anami, R. Sause / International Journal of Fatigue 27 (2005) 383–393

However, ‘dist. 3’, which is obtained by simplifying ‘dist. 2’ difference between these stresses increases with a decrease
and making it symmetric with ‘dist. 1’ about the weld toe, in q. In some cases where q is large (e.g. qZ60 and 908) and
and ‘dist. 1’ were used for the crack propagation analyses. R is small, the stresses are higher than the stresses of the
The calculated fatigue life was defined as the number of T-form joint (a-3: qZ908). Thus, it can be expected that the
cycles until the crack depth reached 80 percent of T. In the fatigue strength of these cases is lower than Category C,
cases where ‘path-1’ was used, crack propagation into the which is the fatigue strength category of the T-form joint
weld bead and web was not considered, and crack (a-3: qZ908). The stresses decrease with an increase in R
propagation along the weld toe, which was observed in and a decrease in q. In addition, the decrease in stress with
the fatigue tests only when the crack length was very small the increase in R is more significant when q is large.
[1], was not considered. According to fatigue test results Crack propagation analysis results for crack propa-
reported in [1], the fatigue cracks macroscopically propa- gation along path-1 and along path-2 are summarized in
gated perpendicular to the principal stress direction, which Fig. 5. The vertical axis is the calculated fatigue life
was inclined about 108 to the longitudinal direction. under a 100 MPa stress range (Nf100). The arrows in the
However, the FEM analyses reported in [1] indicate that figure indicate that the fatigue life is over 40 million
the difference between the longitudinal stress and the cycles. The results for crack propagation along path-1,
principal stress at the weld toe was not significant. which corresponds with the observations from previous
Therefore, it was assumed in the present study that the fatigue tests [1,9–11], are discussed first. Similar to the
difference in fatigue life between crack propagation along decrease in stress at the weld toe, Nf100 increases with a
path-1 and crack propagation perpendicular to the principal decrease in q and an increase in R. When q is 908 and R
stress was not significant. is about 11 mm, which is quite small, Nf100 is close to
In addition to these crack propagation analyses of the
the fatigue life for Category D. This result is reasonable
simple tension model, crack propagation analyses of joint
because this case is similar to joint b-2 in Fig. 1, which
a-3 with inclined angles of 35, 45, 60 and 908 were carried
is a welded joint with an H-shape attachment and has a
out in order to evaluate the influence of the bend region
Category D or E fatigue strength. By increasing R, Nf100
and the longitudinal web fold. Joint a-3 with an inclined
reaches the fatigue life of Category C, which is
angle of 908 is identical to the T-form weld joint a-2 used for
the fatigue strength category for the T-form joint (a-3:
determining a0. The results of this analysis are labeled (a-3:
qZ908). When q is 358 or smaller, Nf100 exceeds the
qZ908).
fatigue life of Category B 0 , and Nf100 increases with an
increase in R. To reach the fatigue strength of Category
B, which is the fatigue strength category of a
3. Analytical results conventional web-flange fillet weld joint (qZ08), q
needs to be very small such as 258 (in Fig. 5) and a
The longitudinal stress at the weld toe and the stress large R is needed. However, if the target fatigue strength
perpendicular to the weld toe at point S, where the category is Category B 0 , which is one category higher
maximum stress is observed, are summarized in Fig. 4. than Category C, the fatigue strength category for a
Fig. 4 shows the stresses normalized by the externally conventional transverse stiffener joint, this goal can be
applied uniform nominal stress. The longitudinal stress is achieved with q equal to 358 and a range of R values.
higher than the stress perpendicular to the weld, and the Since these discussions of fatigue strength are relative to

Fig. 4. Dependence of toe stresses at point S on inclined angle q and bend radius R.
K. Anami, R. Sause / International Journal of Fatigue 27 (2005) 383–393 387

Fig. 5. Crack propagation analysis results: dependence of fatigue life at 100 MPa, Nf100, on inclined angle q and bend radius R.

the fatigue strength of the T-form joint for a specific case 4. Influences of corrugation depth and weld toe radius
of r, H, etc. the generality of these results is arguable.
However, these results could be used as indices for The results explained in Section 3 can be influenced by
evaluating the influence of q and R on the fatigue other parameters such as the corrugation depth (H) and the
strength of the corrugated web-flange weld joint. weld toe geometry. Thus, influences of those parameters are
When considering the results for crack propagation along discussed here only for the case of qZ358, which is a typical
path-2, the increase of Nf100 with a decrease in q is more value of q for corrugated web girders. The crack
significant. These Nf100 values are considerably greater than propagation analyses were carried out only for crack
those calculated for path-1 crack propagation especially propagation along path-1. Although the initial crack size
when q is small. These results and previous fatigue test a0 was not changed when H was changed, a0 was
results indicate that path-1 is the most likely direction of recalculated when the toe radius (r) was changed.
crack propagation and path-2 is unrealistic especially when
q is small.
4.1. Influence of corrugation depth
Fig. 6 is similar to Fig. 5, however, the data for Nf100
is plotted against H1 instead of R. H1 is the distance
A comparison between the cases of HZ150 mm and
from the longitudinal web fold to point S. The fatigue
HZ225 mm is shown in Fig. 7. Small differences in Nf100
life increases almost proportionally to an increase in H1.
are observed between the case of HZ225 mm and
These results indicate that the role of R with regard to
HZ150 mm, where N f100 is slightly larger when
fatigue strength can be represented by H1. The results
HZ225 mm, especially when H1 is small. This is because
also show it is important to locate point S as far as
the stress concentration factor at point S is slightly
possible from the longitudinal fold by increasing R to
decreased with an increase in H as explained in [1].
eliminate the influence of the longitudinal fold and to
increase the fatigue strength. In Fig. 6, Nf100 of joint a-3
models with q values of 35, 45, 60 and 908, are shown
by bold broken lines labeled, for example (a-3: qZ358)
for the case where q is 358. While the fatigue life is
much lower than these lines when H1 is almost 0 mm,
the fatigue life approaches these lines with an increase in
H1. This tendency is independent of q, and indicates that
increasing H1 can change the fatigue behavior of a
corrugated web-flange weld from behavior similar to the
joint-b series in Fig. 1 to behavior similar to the joint-a
series. Although the analysis cases are limited, the
analytical results suggest that H1 larger than 60–70 mm
is needed to eliminate the influence of the longitudinal
web fold and to enable the corrugated web-flange weld
joint to be treated as a T-form joint with an inclined Fig. 6. Crack propagation analysis results: replacing bend radius R by H1,
attachment (joint a-3). which is distance from longitudinal fold to point S.
388 K. Anami, R. Sause / International Journal of Fatigue 27 (2005) 383–393

is originally high. However, the difference in fatigue


strength due to changes in r is not so significant for the
limited analysis cases of r between 0.25 and 5.0 mm when a
recalculated a0 is used.

5. Fatigue strength evaluation

Based on the results in the previous sections and Ref.


[1], it can be concluded that the primary parameters
which dominate the stress conditions and the fatigue
strength of the corrugated web-flange weld joint in a
region of constant bending moment are q and R. The
Fig. 7. Example of dependence of fatigue strength on corrugation depth effect of R is captured in the variable H1. In another
(qZ358). words, the fatigue strength changes from a very low
fatigue strength category like Category D or E to
However, the difference is very small and the two cases Category B, by changing q and H1. While the crack
show a very similar trend in the change of Nf100 with H1. propagation analyses indicate that the calculated fatigue
Therefore, it can be said when H1 is large enough, the strength could exceed Category B when q is very small,
influence of the longitudinal fold is eliminated and the embedded weld defects including weld root defects are
fatigue strength is independent of H. likely to dominate the fatigue strength of the weld joint
in this case. Therefore, the appropriate range of fatigue
4.2. Influence of toe radius strength is between D or E and B. The remaining focus
of this study is how to determine the fatigue strength of
The observed change in fatigue strength due to a change a corrugated web-flange weld with an arbitrary corru-
of q could be a result of either of two mechanisms: (1) a gated web geometry and arbitrary flange geometry. In
change in the structural stress concentration, and (2) a this section, simple fatigue life prediction methods for
change in the stress concentration due to an apparent change corrugated web-flange weld joints are considered. A
of the weld toe geometry, which is expressed in terms of the nominal stress approach and a hot spot stress approach
variables r and qtoe. While the former mechanism was based on the results of the crack propagation analyses
included in the discussion in the Section 3, the latter reported in the previous sections are studied.
mechanism is discussed in this section. The influence of the
apparent change of the local toe geometry due to the change 5.1. Arrangement of crack propagation analysis results
of q depends on the original toe geometry. For example, if (nominal stress approach)
the original r is large, the stress concentration due to the
local toe geometry is small and a change in the severity of It is assumed that the influences of q and H1 can be
the toe geometry due to the change of q will have a small treated separately and that the dependence of the fatigue life
effect. On the other hand, if the original r is small enough, a on H1, i.e. the slope of the relationship between the
change in the toe geometry might have a significant effect. A calculated fatigue life and H1 shown in Fig. 6, is
similar statement can be made for qtoe, but the effect of independent of q, although there is some influence of q on
change in qtoe on stress concentration is relatively small this slope. As mentioned before, if H1 is enough large, the
compared to that of r [12]. Therefore, only the influence of r effect of the longitudinal web fold can be eliminated and the
is examined here, with cases of rZ0.25 mm and rZ5 mm corrugated web-flange weld can be treated as joint a-3 from
studied in addition to the original model, which had rZ Fig. 1. Therefore, the influence of q on fatigue life can be
1 mm. The results are shown in Fig. 8. In Fig. 8(a) the discussed using the fatigue life of joint a-3. The relationship
original value of a0, which was calculated based on the T- between Nf100 for joint a-3 models with different values of q
form joint with rZ1 mm was used. In Fig. 8(b) a for the cases of rZ1 and 0.25 mm is shown in Fig. 9, which
recalculated a0 is used, which is recalculated for each r shows log(Nf100) is inversely proportional to sin q. The
value using the procedure used previously for the case of slope depends on r but is almost independent of a0. This
rZ1 mm. It is obvious that Nf100 decreases with an increase relationship can be expressed by the following equations
in r when a0 is constant. On the other hand, when a log Nf100 Z b K a sin q Z b1 C að1 K sin qÞ (2a)
recalculated a0 is used, Nf100 slightly increases with a
decrease in r, because the influence of changing q from 90
Nf100 Z 10b1 !10að1Ksin qÞ (2b)
to 358 on toe stress is larger when the original r is small,
since the stress concentration due to the local toe geometry where b and b1 are constants, and a is the slope.
K. Anami, R. Sause / International Journal of Fatigue 27 (2005) 383–393 389

Fig. 8. Dependence of fatigue strength on local toe radius (qZ358).

The fatigue strength must correspond to Category C and the corresponding results are shown in Fig. 10
when qZ908, and therefore Eq. (2b) can be modified as    
H1 p
follows log Nf100 Z dðqÞ C b sin K 1 (4)
H10 2
Nf100 Z 10b1 !10að1Ksin qÞ 5 Nf100
where d(q) is the fatigue life of joint a-3 and given by Eq.
A (2b), b is the slope of the lines in Fig. 10, H1ZH 0 1 if H1O
Z c 3 f10að1Ksin qÞ g (3a) H 0 1, and H 0 1 is the value of H1 for which the joint can be
100
treated as joint a-3.
Ac Although the slopes, b, slightly vary around 0.2–0.35, an
Nf Z f10að1Ksin qÞ g (3b) almost linear relationship is observed in Fig. 10. Then, by
s3n
combining Eqs. (3b) and (4), Eq. (5) is obtained for
where Ac is a constant for Category C (i.e. Ac Z s3n Nf ), Nf is predicting the fatigue life of the corrugated web-flange
the fatigue life and sn is the nominal stress range. models which results from the crack propagation analyses
As shown in Fig. 8, the influence of r on the relationship Ac
f10að1Ksin qÞ gf10bðsinððH1 =H1 Þðp=2ÞÞK1Þ g
0
between Nf100 and H1 is not significant. This result is Nf Z (5)
s3n
because the influence of H1 is mostly on the structural stress
and r does not influence the structural stress. Since the effect Eq. (5) has a very simple form and considers the influences
of the longitudinal fold can be eliminated when H1 is of q and H1 separately in the first and second parentheses,
sufficiently large and the fatigue strength of the corrugated respectively. The comparison between the fatigue life
web-flange weld can be estimated from that of joint a-3, the predicted by Eq. (5) and Nf100 calculated by the crack
following sinusoidal-expression is examined to explain propagation analyses for the case of rZ1 mm is shown in
the relationship between the fatigue life and H1 Fig. 11.

Fig. 9. Relationship between calculated fatigue life of joint a-3, Nf100, and inclined angle, q.
390 K. Anami, R. Sause / International Journal of Fatigue 27 (2005) 383–393

Fig. 12. Application of Eq. (5) for fatigue test results from Ref. [1].
Fig. 10. Relationship between calculated fatigue life, Nf100, and H1: H 0 1Z
80 mm, rZ1.0 mm. sn for Eq. (5) must be addressed. In developing Eq. (5), an
externally applied uniform stress was used in the analytical
The conditions used for the results Fig. 11 are aZ1.0, models. Secondary in-plane and out-of-plane flange plate
bZ0.3, and H 0 1Z80 mm. For most cases, the predicted bending occur in both a constant bending moment region
fatigue life coincides with the crack propagation analysis and a region of combined shear and bending moment, and
results. The reason why the cases of qZ258 do not show the stresses from secondary bending are not negligible, as
agreement between Eq. (5) and crack propagation analysis reported in [1]. Fig. 12 is an example where Eq. (5) is
results is that first parenthesis term in Eq. (5), which applied to the fatigue test specimens reported in [1], where
includes the influence of q, does not work well for the case the specimens had qZ36.78, RZ120 mm and the flange
of qZ258 as shown in Fig. 9. However, Fig. 11 indicates plate was 225 (W)!20 (T) mm. The nominal stress, sn, was
that the second parenthesis term captures the effect of H1 calculated from beam theory without including load
even for the case of qZ258. Although it is necessary to do
carrying contribution by web and secondary flange bending
further work to prove the applicability of Eq. (5), it is
stresses. Design curves predicted by Eq. (5), with bZ0.2
possible to suggest that Eq. (5) could be used to determine
and 0.3, lie between the Category B and B 0 design curves,
the fatigue strength category of corrugated web-flange
and the margin between the fatigue test results and the lines
welds.
predicted by Eq. (5) is quite small. It is worth mentioning
that the web-flange weld toe conditions for the test
5.2. Hot spot stress approach specimens are likely to be, at worst, the average conditions
for actual bridge girder welds. Significant weld defects were
Although Eq. (5) predicts the fatigue life from the crack not observed on the crack surfaces. Hence, the application
propagation analyses, it is necessary to obtain an appro- of Eq. (5) with the above defined sn has the possibility of
priate nominal stress range (sn) to use this equation. When providing unsafe fatigue design curve for the corrugated
Eq. (5) is applied to the corrugated web-flange welds web-flange weld joint when secondary bending in the flange
in actual bridge girders, the problem of how to determine plate is not negligible. Therefore, it is necessary to consider
an appropriate sn, which corresponds to the externally
applied uniform stress examined in this study, or to include
such secondary flange bending. Although not considered
further here, secondary bending must be quantitatively
evaluated. Here, the hot spot stress approach or structural
stress approach is used to approximately evaluate the
secondary flange bending stress.
The crack propagation analysis results (rZ1 mm) are
plotted on S–N diagram in Fig. 13 using SCF times the
nominal stress range for the vertical axis. SCF is the stress
concentration factor, and two definitions of SCF are used,
namely SCF-1, the hot spot stress concentration factor
recommended by the International Institute of Welding
Fig. 11. Comparison of crack propagation analysis and Eq. (5) for case of (IIW) [12], and SCF-2, the stress concentration factor that
rZ1 mm at 100 MPa: in Eq. (5), aZ1, bZ0.3, H 0 1Z80 mm. estimates the stress at 6 mm far from weld toe proposed by
K. Anami, R. Sause / International Journal of Fatigue 27 (2005) 383–393 391

Fig. 13. Influence of inclined angle on fatigue strength by stress concentration factor.

Dexter [13]. When SCF-1 was used, the thickness effect is observed in fatigue test results from Yamada [4,5], even
considered by the following equation proposed by IIW [12] though the test data is scattered more widely than analytical
 0:3 results due to local toe condition variations present in the
T test specimen. Reasons why the use of (SCF!sn) does not
shs Z shs;eff (6)
Teff work well include: (1) the change of stress concentration at
the weld toe due to the change in local toe geometry
where shs is the hot spot stress, shs,eff is the hot spot stress resulting from the change of q is not included, and (2) the
for TZ25 mm, and Teff is 25 mm. application of the hot spot stress approach for welded joints
Ref. [12] states that the fatigue strength of a welded having attachments inclined to the primary stress has not
attachment joint can be categorized as IIW-FAT100, which been fully examined.
almost corresponds to AASHTO Category B 0 , by using To address the first reason mentioned above, that the hot
SCF-1, and Ref. [13] states the joint can be categorized as spot stress approach does not include the change of the local
AASHTO Category C using SCF-2. In addition to the toe geometry due to the change of q, the first parenthesis in
analytical results, the fatigue test data from Yamada [4,5] Eq. (5), which captures the influence of q on local toe
for specimens of joint a-3, are also plotted for the purpose of geometry, is used to modify the vertical axis of Fig. 13 as
the comparison. The test data are plotted with analytical follows, with the results shown in Fig. 14
results even though the analytical results are calibrated to a
lower bound design curve (for Category C) and the test data stress range ðvertical axis in Fig: 14Þ
are expected to be scattered around a mean-value S–N
sn !SCF
curve. Regardless of the definition of the SCF, the difference Z ; where F Z f10að1Ksin qÞ g (7)
in fatigue strength between the cases with the same q is F 1=3
reduced, but is not eliminated completely. On the other Fig. 14 shows that the dispersion of the crack propagation
hand, while most of data plots between the Category C and analysis results due to q becomes very small, and the results
B design curves, regardless of the definition of SCF, are close to the design curve of Category C when SCF-2 is
differences in the results with different q values still remain used, and Category B 0 when SCF-1 is used. In addition,
even after the SCF is used. These tendencies are also the fatigue test results from Yamada [4,5], evaluated with

Fig. 14. Re-evaluation of Fig. 12 using F in Eq. (8), which can account for the effect of q on local toe geometry.
392 K. Anami, R. Sause / International Journal of Fatigue 27 (2005) 383–393

in the flange plate. In addition, Eq. (8) accounts for


the change in fatigue life from the influence of q on
the local weld toe geometry. The application
and limitations of these equations are summarized as
follows:

1. The present study that developed Eqs. (5) and (8) is


based on the limited analytical results only, which
were calibrated to a fatigue design curve. More
experimental and analytical data are needed to verify
the applicability of the proposed equations.
2. Eq. (5) needs an accurate nominal stress, sn, which
includes the secondary bending that occurs in the
flange plate of a corrugated web girder. Once we
obtain an accurate sn, the influences of q and H1 are
Fig. 15. Application of Eq. (8) to fatigue test results for corrugated web
girder specimen used in Ref. [1]. accounted for by this equation, and the fatigue life of
the corrugated web-flange weld can be estimated with
reasonable accuracy.
SCF-1, become almost independent of q by using Eq. (7), 3. Since an accurate sn is difficult to be determined for a
especially when the applied stress range is high. However, corrugated web girder using simple beam theory, Eq.
Fig. 14 shows the crack propagation analysis results do not (8) uses the hot spot stress, which can include
completely agree with the design curve due to conditions secondary bending stress in the flange plate. However,
(e.g. varieties in H1) that are not fully accounted in Eq. (7). this equation does not fully account for effect of H1
If this error can be neglected, Eq. (5) can be modified into even though SCF (hot spot stress) is calculated from
Eq. (8) FEM, as shown in Fig. 14. Therefore, further research
is needed to define the hot spot stress for an
Ac0 attachment inclined to the primary stress direction so
Nf Z f10að1Ksin qÞ g (8)
ðSCF !sn Þ3 that the applicability of Eq. (8) is improved.

The constant A 0 c in Eq. (8) is similar to Ac, but depends


on the definition of SCF. If A 0 c is determined as the average 6. Conclusions
of the analysis data plotted in Fig. 14, A 0 cZ1.90!1012 for
SCF-1, which is slightly lower than Category B 0 , and A 0 cZ Based on the FEM analyses and simple crack propa-
Ac for SCF-2. Then, if Eq. (8) is applied the fatigue test gation analyses, which only considered mode-I crack
specimen described Ref. [1], which has 1.23 for SCF-1 and propagation of a single crack initiating at point S, the
1.21 for SCF-2, the predicted fatigue design curve is following conclusions were obtained:
obtained as shown in Fig. 15. The predicted design curve for
(1) Since point S (Fig. 2) is the most critical point for
this corrugated web girder specimen under constant bending
fatigue crack propagation, the corrugated web-flange
moment coincides with Category B 0 when SCF-2 is used
weld can be treated as an assembly of the out-of-plane
and is slightly less than Category B 0 when SCF-1 is used.
gusset weld joint (joint b-1) and the T-form joint with
Thus, one of conclusions in Ref. [1] that Category B 0 was an
inclined attachment (joint a-3) with a certain transition
appropriate fatigue design category for a corrugated web-
region between them.
flange weld joint with a similar geometry similar to the test
(2) By increasing R and locating point S, at least, 80 mm
specimen is also supported by the procedure for predicting
from the longitudinal web fold, the influence of the
the fatigue life discussed in the present study.
longitudinal fold can be eliminated and the fatigue
strength of this joint can be determined from that of T-
5.3. Application and limitations of results form joint with an attachment inclined to the primary
stress. When q is small, a large R is needed to make H1
The equations (Eqs. (5) and (8)) for calculating the large enough to eliminate the influence of the
fatigue life of the corrugated web-flange weld were longitudinal fold.
derived by calibrating the crack propagation analysis (3) The influence of the local toe geometry on the fatigue
results to a design curve (AASHTO Category C). Eq. (5) strength is affected by changes in q. The severity of
uses the nominal stress, sn, and accounts for the local toe geometry decreases with a decrease in q.
influence of the web corrugation geometry using the (4) The influences of H and the toe radius, r, were also
parameters, q and H1. Eq. (8) uses the hot spot stress, examined, but these influences on the relationship
which can include the effect of secondary bending stress between H1 and the fatigue life are not significant.
K. Anami, R. Sause / International Journal of Fatigue 27 (2005) 383–393 393

In addition, based on the results of those analyses, simple [2] American Association of State Highway and Transportation Officials.
formulae Eqs. (5) and (8), which include the influences of q AASHTO LRFD Bridge Design Specification; 1998.
[3] Fisher JW, et al. Fatigue resistance of large welded AL-6XN stainless
and H1, to estimate the fatigue strength of corrugated web-
steel components with fillet, groove, and attachment welds. ATLSS
flange weld joints were proposed. The applicability of these report no. 01-04.: Lehigh University; 2001.
formulae was discussed. Since these results are based on [4] Yamada K, et al. Fatigue behavior of inclined non-load-carrying fillet
assumptions, and limited FEM and crack propagation welded joints. J Jpn Soc Civil Eng 2001;682(156):383–90.
analyses, further experimental and analytical research, [5] Yamada K. Fatigue test of tensile plate with out-of-plane gussets
inclined to applied stress. J Struct Eng (Jpn Soc Civil Eng)
including research to establish the constants a, b, A 0 c, is
2001;1039–45 [in Japanese].
needed to fully validate these formulae. [6] Japanese Society of Steel Construction. Fatigue design recommen-
dation for steel structure; 1995.
[7] Albrecht P, Yamada K. Rapid calculation of stress intensity factors.
Acknowledgments J Struct Div, ASCE 1997;103(ST2):377–89.
[8] The Japan Welding Engineering Society. Japan Welding Engineering
Society standard-method of assessment for flaws in fusion welded
This research was supported by the Federal Highway joints with respect to brittle fracture and fatigue crack growth. WES
Administration and Pennsylvania Infrastructure Technology 2805; 1997.
Alliance (funded by a grant from Pennsylvania Department [9] Harrison JD. Exploratory fatigue test of two girders with corrugated
of Community and Economic Development). This financial webs. Br Weld J 1965;121–5.
[10] Korashy M, Varga J. Comparative investigation of fatigue strength of
support is gratefully acknowledged. The contents of the beams with web plate stiffened in the traditional way and by
paper reflect the views of the authors, and do not necessarily corrugation. Acta Tech Acad Sci Hung 1979;89(3–4):309–46.
reflect the views of those acknowledged herein. [11] Elgaaly M, et al. Bridge girders with corrugated webs. Fifth
international bridge engineering conference, transportation research
record, 1696 2000 p. 162–70.
[12] Niemi E, Marquis G. Introduction to the structural stress approach to
References fatigue analysis of plate structures. Proceedings of the IIW fatigue
seminar 2002 p. 73–90.
[1] Anami K, Sause R, Abbas HH. Fatigue of web-flange weld of [13] Dexter R, et al. Application of hot-spot stress fatigue analysis to
corrugated web girders (part-1), influence of web corrugation attachments on flexible plate. Proceedings of the 13th international
geometry and flange geometry on web-flange weld toe stresses. Int J conference on offshore mechanics and Arctic engineering conference
Fatigue. doi: 10.1016/j.ijfatigue.2004.08.006. (OMAE) 1994.

You might also like