You are on page 1of 12

Engineering Structures 201 (2019) 109767

Contents lists available at ScienceDirect

Engineering Structures
journal homepage: www.elsevier.com/locate/engstruct

Cracking reasons and features of fatigue details in the diaphragm of curved T


steel box girder

Yixun Wanga, Zhongqiu Fua, , Hanbin Geb, Bohai Jia, Norihiko Hayakawab
a
College of Civil and Transportation Engineering, Hohai University, No. 1 Xikang Road, Nanjing, China
b
Department of Civil Engineering, Meijo University, 1-501, Shiogamaguchi, Tempaku-ku, Nagoya, Japan

A R T I C LE I N FO A B S T R A C T

Keywords: The cracking reasons and features of fatigue details in the diaphragm of curved steel box girder were studied
Curved steel bridge based on a real curved steel bridge with three spans located in Nagoya, Japan. The cracking conditions of certain
Steel box girder fatigue details were summarized according to in situ monitoring data. The in-plane and out-of-plane deformation
Diaphragm was analyzed considering structural features and loading conditions of the bridge as a whole. The influence of
Fatigue
the curve on fatigue characteristics was investigated and reasons for cracking were concluded. Moreover, the
Cracking features
fatigue test was carried out on cracking details to verify cracking behaviours in the real bridge, based on which
the cracking laws were summarized. The results indicated that the unevenly distributed loadings caused by the
curvature resulted in a greater stress on outside details than on inside details in the curved steel box girder,
which was the major difference in the fatigue characteristics between curved and straight steel box girders. The
eccentricity of vehicle loadings and structural characteristics were the major reasons behind the antisymmetric
propagation of the cracks in the diaphragm-web rib weld. Apparent interference could be observed in the fatigue
cracks of the diaphragm-U rib weld on both sides of an identical U rib.

1. Introduction [7,8]. A lot of theoretical analysis and experimental verifications have


been conducted to investigate the reasons for such cracking and to
The curved steel box girder is known for its favourable continuity explore features of common fatigue details. Kainuma et al. [9] believed
and bearing capacity, and has been widely applied in bridges on that significant out-of-plane deformation of the deck-U rib weld was
highways, railways, and overpasses for its light weight, high strength, caused by the effects of the transverse distribution of vehicle loadings.
and rapidity of construction [1,2]. Fatigue cracking has become one of Cui et al. [10] found that fatigue cracks were usually initiated from the
the most significant problems impairing the safe operation of steel weld root or weld toe, and Kim et al. [11] concluded that after the
bridges under long-term cyclic vehicle loadings [3,4]. In particular, initiation, cracks propagated through the thickness of the deck until the
coupling effects between the bending and torsion moments caused by deck was penetrated. Based on the in situ monitoring and FE analysis,
the shape of curve cannot be neglected in the curved steel box girder Connor and Fisher [12] found that the effective acting area of the ve-
[5,6]. The stress distribution of structural details is influenced by un- hicle loadings was decreased by a scallop cut-out from the diaphragm,
evenly distributed loadings, making the problems of fatigue damage where high in-plane, and out-of-plane secondary stresses could be ob-
more complicated than in a straight steel bridge. According to in situ served. Fu et al. [13] figured out that fatigue cracks initiated from the
monitoring data from a curved steel box girder in Nagoya, Japan, col- weld toe of the diaphragm-U rib weld and then propagated at a certain
lected in 2017, fatigue cracks were found in several structural details angle with the weld, while Zhang et al. [14] also pointed out that a
while the distribution of cracks was irregular. Most cracks initiated in certain number of cracks started at the edge of the scallop cut-out and
the outside region of the curved bridge, which indicated the significant developed through the base metal of the diaphragm. According to the
difference in fatigue features between curved and straight steel box experiment results, Zong et al. [15] found that initial deficiency, geo-
girders. metric discontinuity, and rough transition of the U rib butt weld were
Current studies of fatigue cracking are usually conducted on straight often caused by in situ welding, which resulted in cracks originating
steel box girders, and especially for the long-span flat steel box girders from the weld root on the bottom of the U rib butt weld and


Corresponding author.
E-mail addresses: emitakei@hhu.edu.cn (Y. Wang), fuzhongqiu@hhu.edu.cn (Z. Fu), gehanbin@meijo-u.ac.jp (H. Ge), bhji@hhu.edu.cn (B. Ji),
183433006@ccmailg.meijo-u.ac.jp (N. Hayakawa).

https://doi.org/10.1016/j.engstruct.2019.109767
Received 22 March 2019; Received in revised form 2 September 2019; Accepted 5 October 2019
Available online 22 October 2019
0141-0296/ © 2019 Elsevier Ltd. All rights reserved.
Y. Wang, et al. Engineering Structures 201 (2019) 109767

Fig. 1. Research flow chart.

propagating along the weld until the U rib fractures, proposed by Yang diaphragm-floor rib weld is marked by pentagon. As seen in Fig. 4,
et al. [16]. The curved steel box girder is usually applied to the simply three cracks were found in the diaphragm-U rib weld, five cracks were
supported and continuous bridges of short to medium span [17], the found in the diaphragm-web rib weld and one in the diaphragm-floor
stress conditions in which vary from those prevailing in a flat steel box rib. The distribution of fatigue cracks was localized around three dia-
girder under the support of cables [18]. The structure of fatigue details phragms as shown. Structural details were classified into two types
also varies: discrepancies in stress state and structural form result in based on the longitudinal axis of the bridge, namely the outside details
different cracking characteristics for similar fatigue details, thus re- and inside details as seen in Fig. 5: all cracking diaphragm-web rib
search results from tests on flat steel box girder cannot be directly ap- welds were located on the outside of the curve, the cracking diaphragm-
plied to the curved steel box girder. U rib welds were observed on both sides of the curve and the only one
The cracking reasons and features of fatigue details in the dia- crack found in the diaphragm-floor rib weld was situated on the inside
phragm of curved steel box girder were studied in this research. of the curve. The cracking details of diaphragm-U rib welds No. 1.1 and
According to the cracking conditions and data obtained by stress No. 1.2 were located on adjacent U ribs, thus synergistic effects from
monitoring on a certain curved steel bridge, the cracking features of these two cracks on the fatigue performance of diaphragm could not be
different fatigue details were analyzed. The stress characteristics and ignored. Moreover, the cracking details of diaphragm-web rib welds No.
reasons for cracking of the studied fatigue details were investigated by 2.4 and No. 2.5 were observed on both sides of a certain diaphragm:
finite element (FE) analysis. The results by FE analysis were verified by these propagated together on the diaphragm after initiating from the
fatigue testing to furtherly conclude the cracking rules of the curved welding end and significantly diminished the bearing capacity of the
steel box girder. The research flow chart was shown in Fig. 1. diaphragm.

2. Project description 2.2. Cracking features

2.1. Conditions of fatigue cracking The cracking features of three fatigue details are illustrated in Fig. 6.
With regard to the diaphragm-U rib weld and diaphragm-floor rib weld,
The reasons for cracking of fatigue details in the diaphragm of the cracks initiated from the welding end of the connection between the
curved steel box girder were studied using data from an overpass bridge rib and diaphragm, and then propagated at a certain angle with the
(Pier 61 to Pier 64) in the central line running through Nagoya, Japan. weld in the diaphragm. The cracks in the diaphragm-web rib weld
The bridge (Fig. 2) was completed in July 1987 and opened to traffic in usually started at the welding end where stress concentration was sig-
April 1988. It has been safely operated from then on. It was designed as nificant. Crack tips on both sides developed asymmetrically in the
a one-way highway with three lanes and a three-span continuous steel diaphragm, making a certain angle with the web rib. To avoid further
box girder, which featured a single box and a single cell. The total propagation of these fatigue cracks, repairs involving a stop-hole and
length of this bridge is 223.675 m and its radius of horizontal curvature steel plate reinforcement were applied to cracks in the diaphragm-U rib
is 350 m. The main girder is composed of the orthotropic steel bridge weld, and the stop-hole method was applied to cracks initiating from
deck (OSD), diaphragm, web, and floor. The connection between the the diaphragm- web rib weld. No more propagation thereof was seen
web and diaphragm is reinforced by a local stiffener to constrain out-of- after these repairs.
plane deformation. The width and height of the section are 12,750 mm To analyze the cracking conditions of fatigue details, the changes in
and 2750 mm. The thickness of deck and pavement are 12 mm and crack length with service life are shown in Fig. 7 according to the an-
80 mm, respectively. The vehicles travel from Pier 61 to Pier 64 and the nual detection records of Nagoya Expressway. As seen in
traffic volume has reached a total of 4.148 × 109 vehicles by 2017 Figs. 4, 5, and 7(a), cracks No. 1.1 and No. 1.2 were located on the
according to statistics collected by the Nagoya Expressway operators, adjacent scallop cut-outs of an identical diaphragm while crack No. 1.1
among which the proportion of heavy vehicles is 12.6%. The daily was more affected by eccentric vehicle loading. The length and pro-
average traffic volume is shown in Fig. 3. It can be observed that the pagation rate of crack No. 1.1 were greater than those of crack No. 1.2.
traffic volume has significantly increased until the end of 20th century The cracking rate of crack No. 1.1 continued to increase with the in-
with peak daily flow reaching 479,000 vehicles. The increase in number creasing propagation of crack No. 1.2. Effects of fatigue repair on crack
of vehicles has stopped yet while peak flows persist to this day. The No. 1.1 were favourable because no further propagation has been seen
high traffic volume and sophisticated constitution of vehicles have re- since. The length of crack No. 1.2 decreased by 2 mm after cutting and
sulted in fatigue damage to several structural details and cracks have it then propagated at a similar rate as crack No. 1.1. The crack propa-
already been found in the steel box girder, threatening the safety and gation had stopped developing in recent years after the repair of crack
durability of the bridge. No. 1.1. Crack No. 1.3, which was discovered several years later, had
The crack distribution is shown in Figs. 4 and 5 (data from recent in propagated rapidly and should be repaired timely to slow its propaga-
situ monitoring). The cracking diaphragm-U rib weld is marked by tri- tion.
angle, the diaphragm-web rib weld is marked by quadrangle and the As seen in Figs. 4, 5 and 7(b), cracks No. 2.1 and No. 2.2 were

2
Y. Wang, et al. Engineering Structures 201 (2019) 109767

(a) Shape of the curve

(b) Appearance of the bridge


Fig. 2. The curved steel bridge.

located on adjacent diaphragms. These two cracks were detected at


50
about the same time and their rates of cracking were similar. It was
deemed necessary to repair them as soon as possible because of their
45
cracking lengths. The stop-hole method was applied to crack No. 2.3
Daily vehicles (Thousand)

40 and no propagation was detected after drilling the necessary holes.


Cracks No. 2.4 and No. 2.5 were detected on both sides of an identical
35 diaphragm and the stop-hole method was also applied to improve local
fatigue performance. No cracks have been found initiating from the
30 hole edge at time of writing.
As seen in Figs. 4, 5, and 7(c), only one crack was observed in the
25 diaphragm-floor rib weld. The cracking time was later than in the other
two fatigue details, the cracking length was shorter and rate of cracking
20 was slower; it was assumed that the fatigue performance of the dia-
phragm-floor rib weld was more favorable than the diaphragm-U rib
15 weld and diaphragm-web rib weld.
New cracks will inevitably continue to initiate in this three-span
10 continuous and curved steel bridge with increasing of service life. The
1990 1995 2000 2005 2010 2015
existing fatigue cracks will also threaten the safe operation of the bridge
Time (Year) if not repaired timely. Therefore, it is necessary to obtain the stress
characteristics of such fatigue details, analyze the reasons for cracking,
Fig. 3. Average daily vehicle flows.
and explain influence of the curvature on the stress distribution in the
bridge. The conclusions can be of significance to decide suitable repair
methods for cracked details and when proposing schemes to monitor
critical details with no cracks for the time being.

Fig. 4. Crack distribution in the steel box girder.

3
Y. Wang, et al. Engineering Structures 201 (2019) 109767

3250 1000
3250 3250
1000 First lane Shoulder
Third lane Second lane
Shoulder

500 500 1750 500


1750 500 500 1750 Outside
500
Inside
D1.1
D1.2 No.1.1 No.1.2 No.1.3
U rib
Web rib D2.1
D2.2 Diaphragm-U rib weld
Diaphragm Diaphragm-web rib weld
D i aphragm

No.2.1~2.5 Diaphragm-floor rib weld


Web Floor rib
D3.2 No.3.1 The outside fatigue details
D3.1
The inside fatigue details

Fig. 5. Crack distribution (cross-section).

160

140 Stop-hole and steel plate reinforcement


No more propagation

Crack length(mm)
120

100
No.1.1
No.1.2
80
No.1.3
60 Cutting
Decrease by 2mm
40

2004 2006 2008 2010 2012 2014 2016


(a) Diaphragm-U rib weld Year
(a) Diaphragm-U rib weld
250

200
Crack length(mm)

150

No.2.1
100 No.2.2 Stop-hole
No more propagation
No.2.3 Stop-hole
No.2.4 No more propagation
50 No.2.5

2004 2006 2008 2010 2012 2014 2016


(b) Diaphragm-web rib weld Year
(b) Diaphragm-web rib weld
95
No.3.1

90
Crack length(mm)

85

80

75
2010 2011 2012 2013 2014 2015 2016 2017
(c) Diaphragm-floor rib weld Year

Fig. 6. Fatigue details. (c) Diaphragm-floor rib weld


Fig. 7. Crack length-service life.

4
Y. Wang, et al. Engineering Structures 201 (2019) 109767

3. FE model of the full-scale bridge

3.1. Parameters of the FE model

A finite element (FE) model was built for this curved steel bridge to
analyze the stress characteristics of fatigue details under vehicle load-
ings and investigate the reasons for cracking thereof. The model of a
full-scale steel box girder was established while elements of the sub-
structure, such as piers and bearings, were neglected because all cracks
detected were in the diaphragm. The elastic modulus of steel E was set
to 206 GPa and the Poisson’s ratio υ to 0.3. The thicknesses of the deck,
diaphragm, web, floor, U rib, web rib and floor rib were 12 mm, 9 mm,
11 mm, 25 mm, 6 mm, 12 mm, and 14 mm, respectively. The elastic
modulus of the pavement was 1000 MPa and its Poisson’s ratio was 0.3.
The thickness of the pavement was 80 mm.
The boundary conditions were set on the basis of the constraint
offered by the bearings in the real bridge. All degrees of freedom (DOF)
of the main girder at Pier 64 were restrained except the translational
freedom longitudinal to the bridge and the constraint area was
270 mm × 270 mm. All DOFs of the main girder at Pier 63 were re-
strained except the translational freedom longitudinal to the bridge and Fig. 8. FE model of steel box girder.
rotational freedom transverse to the bridge. The constraint area was
798 mm × 580 mm. All DOFs of the main girder at Pier 62 were re-
3.2. Verification of accuracy
strained except the rotational freedom longitudinal to the bridge. The
constraint area was 860 mm × 860 mm. All DOFs of the main girder at
The grid size, boundary conditions, and constraints exert a sig-
Pier 61 were restrained except translational and rotational freedom
nificant influence on the calculated FEA results, especially for the full-
longitudinal to the bridge. The constraint area was 860 mm × 860 mm.
scale bridge with sophisticated structures. It is necessary to verify the
According to the Specification for Fatigue Design of Highway Steel
accuracy of the FE model: a strain checker was set on the actual bridge
Bridges [19], a wheel loading with an area of 500 mm × 200 mm was
(Fig. 9) to collect the strain history of fatigue details, which was then
applied to the steel bridge deck. The equivalent applied stress was
compared to the FEA results. A 14 min-long strain history of the dia-
1 MPa. The diaphragm in the mid-span of the bridge was studied and
phragm-web-rib weld was taken as an example (Fig. 10).
stress characteristics of the diaphragm-U rib weld, diaphragm-web rib
The peak tensile strain was 34 με in this strain history, at which
weld, and diaphragm-floor rib weld were investigated. Sub-routine
point the peak stress was 7.0 MPa. This value was of the same order of
DLOAD was compiled to allow simulated wheel loading and a total of
magnitude as the peak stress predicted by FE analysis (Section 4.2),
26 working conditions transverse to the bridge were decided at a dis-
which was 4.2 MPa. Besides, the peak compressive strain was −141 με
tance of 500 mm between them. The working conditions were num-
in this strain history, at which point the peak stress was −29.0 MPa:
bered G0 to G25 from the outside of the curve to the inside. The ap-
this was also of the same order of magnitude as the stress predicted by
proximate scope of the most unfavourable loading position was first
FE analysis, which was −57.1 MPa. The discrepancy between measured
determined and then distance between loading cases was reduced to
and FE data can be caused by the differences in wheel loading applied
100 mm to refine the location of the worst-case loading position. The
to the real bridge and the standard fatigue wheel applied in the FE
loading scope longitudinal to the bridge was 6400 mm (i.e., the distance
analysis. Changes in wheel positions or superimposed effects induced
between diaphragms). The loading path followed the curve of the real
by several wheels acting in parallel could also cause such differences.
bridge. The distance between loading steps was 200 mm and a total of
Here, it was assumed that the measurement agreed with the FE analysis
32 steps were applied to the steel bridge deck (Fig. 8).
in terms of the order of magnitude of the peak stress, thus the FE model
ABAQUS 6.16 software was used to build the FE model. The span of
was deemed to have been both reasonable and correct.
the main girder was so great that components could be deemed to act as
shell structures. Therefore, shell elements were used to model compo-
nents below the steel bridge deck and solid elements were used to re-
present the steel bridge deck and pavement. A TIE constraint was set
between the steel bridge deck and other parts which were connected by
welds. As for the pavement and steel bridge deck, the C3D8R element
was globally assigned and the grid size was 500 mm. The grid size, in
the region within 6400 mm from the diaphragm of interest, was refined
to 50 mm. The C3D10 element was applied in the transition of the mesh
from coarser regions. The total number of elements assigned to the
pavement and steel bridge deck was 345,222.
As for the components below the steel bridge deck, the S4R element
was globally assigned and the grid size was 500 mm. The overall ele-
ment dimension of the diaphragm under investigation was 50 mm. The
diaphragm-U rib weld (D1.1), diaphragm-web rib weld (D2.1), and
diaphragm-floor rib weld (D3.1) on the outside of the curve were re-
fined, as were the corresponding fatigue details D1.2, D2.2, and D3.2 on
the inside of the curve (Fig. 5). The grid size in the region within 50 mm
from the weld under analysis was refined to 1 mm. The total number of
elements assigned to components below the steel bridge deck was
329,712. Fig. 9. Strain monitoring.

5
Y. Wang, et al. Engineering Structures 201 (2019) 109767

Fig. 10. A typical strain history.

4. Stress characteristics of fatigue details G9 GX1 GX2 GX3


50 GX4 G10
4.1. The diaphragm-U rib weld
0
The wheel loading was applied sequentially from working condition

Stress(MPa)
G0 to G25. The loading path was compiled by FORTRAN and matched
with the curve pertinent to that load case. The normal stress history of -50
diaphragm-U rib weld D1.1 (Fig. 11) was obtained by the function of
Field Output provided by ABAQUS. The tensile stress was close to zero -100
during the movement of the wheel loading under most working con-
ditions while it changed significantly and reached a peak value under
working conditions G7 to G11. The distance between the center of the -150
wheel loading and the fatigue detail ranged from −900 mm to
1100 mm (measured transverse to the bridge) under the aforemen- -200
tioned working conditions. The stress distribution in the diaphragm-U -6000 -4000 -2000 0 2000 4000 6000
rib weld followed a symmetrical double-peaked shape. The compressive Longitudinal position(mm)
stress increased as the wheel loading moved closer to the fatigue detail
and the peak compressive stress reached −174 MPa, but decreased Fig. 12. Sub-loading of diaphragm-U rib weld.
significantly when the wheel was close to the fatigue detail and became
tensile. The worst-case loading position transverse to the bridge was at GX1, thus GX1 was assumed to have been the most unfavourable
located between working conditions G9 and G10 according to the peak working condition transverse to the bridge. The center of the wheel
tensile stress, thus the increment in load positioning between working loading coincided with the diaphragm-U rib weld transverse to the
conditions G9 and G10 was refined to locate the most unfavourable bridge under this condition, and the most unfavourable loading position
loading case more accurately. longitudinal to the bridge was 600 mm from this fatigue detail. The
Divide the loading distance between working conditions G9 and cracking reason for the diaphragm-U rib weld was analyzed based on
G10 into 100 mm intervals, named GX1 to GX4, respectively. The wheel working condition GX1 in terms of its stress characteristics.
loading was applied to sub-conditions GX1 to GX4 and the stress history The out-of-plane and in-plane stresses were defined according to the
was shown in Fig. 12. The peak tensile stress (42.4 MPa) was achieved Recommendations for Fatigue Design of Welded Joints and Components
[20]. According to the IIW recommendation, it was assumed that the
direction of bending deformation which was out of the plane was per-
pendicular to the thickness of the plate and the direction of tensile
deformation which was in the plane was parallel to thickness of the
plate. Therefore, normal stress on components in the steel bridge deck
was equal to the vector superposition of the out-of-plane stress and in-
plane stress (Eqs. (1) and (2)):
σleft + σright
σin =
2 (1)
σleft − σright
σout =
2 (2)
where σleft denotes the normal stress on the left of the component, σright
denotes the normal stress on the right of the component, σin denotes an
in-plane stress, and σout denotes an out-of-plane stress. The fatigue de-
tails were defined as being located on the left of the component. Taking
the thickness center of the component as the axis of symmetry, the
symmetric side was defined as being right of the component. It could be
speculated that, when out-of-plane stresses dominated in the section, if
the in-plane stress was neglected, the compressive stress was observed
on one side of a fatigue detail, and a tensile stress must be detected on
Fig. 11. Stress cycles of diaphragm-U rib weld. the other. Besides, the stress state on both sides of the fatigue details

6
Y. Wang, et al. Engineering Structures 201 (2019) 109767

60 generated as the wheel loading approached this detail. The ratio of in-
σleft
plane stress to out-of-plane stress was less than 1 under most conditions
40 σright during movement of the wheel loading. The in-plane stress was greater
σin than the out-of-plane stress only when the out-of-plane stress changed
20 from compressive to tensile, and the ratio reached −6.5. As the center
σout
of the applied wheel loading coincided with the diaphragm, the ratio
Stress (MPa)

tended to 0, indicating that little in-plane stress acted in the diaphragm-


0
U rib weld and influence of out-of-plane stresses on this fatigue detail
had reached a maximum. Contributions from out-of-plane stress to fa-
-20 tigue damage were greatest because the overall stress was higher under
this condition. When the wheel loading was far from the diaphragm, the
-40 in-plane stress was about half of the out-of-plane stress. The stress was
low and made little difference to the fatigue damage. In summary, the
fatigue damage suffered by a diaphragm-U rib weld was mainly caused
-60
by out-of-plane stresses. The in-plane stress was much lower than the
out-of-plane stress and had little effect on the fatigue performance of
-80
the diaphragm-U rib weld.
-6000 -4000 -2000 0 2000 4000 6000
Longitudinal position (mm)
4.2. The diaphragm-web rib weld
Fig. 13. Stress state of diaphragm-U rib weld.
The wheel loading was applied to the steel bridge deck as described
increased or decreased relative to changes in out-of-plane deformation. in Section 4.1: the normal stress history of diaphragm-web rib weld
When in-plane stresses dominated in the section, if the out-of-plane D2.1 was shown in Fig. 15. The stress cycles were entirely compressive
stress was neglected, the stress state on both sides of the fatigue details under working conditions G0 to G8 and the peak compressive stress was
must be identical. The stress state increased or decreased simulta- reached when the wheel loading was applied on the outside of the curve
neously with changes in in-plane deformation. (G0) transverse to the bridge. As the flanges of the diaphragm could be
The in-plane and out-of-plane stresses on the diaphragm-U rib weld seen as cantilevers, the cantilever length reached a maximum when the
were calculated under wheel loading based on the method mentioned wheel loading was located on the outside of the curve, thus the peak
above. As seen in Fig. 13, changes in σleft and σright were mutually op- compressive stress was reached under this condition. The cantilever
posed. With movement of the wheel loading along the steel bridge deck, length decreased as the wheel loading moved from the outside to the
the trend in σleft was to decrease, increase, decrease, increase, decrease middle of the diaphragm, resulting in the gradual decrease in com-
and increase while the opposite change rule was observed in σright as pressive stress. When the wheel loading was located close to the dia-
increase, decrease, increase, decrease, increase, and decrease. This be- phragm-web rib weld (measured transverse to the bridge), the com-
haviour embodied the predominant effects of out-of-plane stress on pressive stress tended to 0. The stress distribution on the diaphragm-
stress characteristics of fatigue details. Besides, when the wheel loading web rib weld followed a single-peak shape. The compressive stress
was close to the fatigue detail, measured longitudinally to the bridge, became tensile under working conditions G9 to G11. The distance be-
σleft and σout reached peak tensile values while σin remained in com- tween the center of the wheel loading and fatigue detail ranged from
pression. The major influence on the stress characteristics of a dia- −350 mm to 650 mm (transverse to the bridge) under the aforemen-
phragm-U rib weld was the out-of-plane stress. tioned working conditions. The most worst-case loading position
The ratio of in-plane stress to out-of-plane stress in the diaphragm-U transverse to the bridge was located between G9 and G10 according to
rib weld was shown in Fig. 14. When the wheel loading was applied far the peak tensile stress, thus the increment in load position was refined
from the fatigue detail (measured longitudinally), the in-plane stress therein.
and out-of-plane stress were similar. Significant deviations were Divide the distance between G9 and G10 into 100 mm increments
(GX1 to GX4, respectively): the wheel loading was applied at GX1 to

0
σin/σout

-2

-4

-6

-6000 -4000 -2000 0 2000 4000 6000


Longitudinal position (mm)
Fig. 14. Stress ratio of in-plane and out-of-plane stress of diaphragm-U rib
weld. Fig. 15. Stress cycles of diaphragm-web rib weld.

7
Y. Wang, et al. Engineering Structures 201 (2019) 109767

G9 40
4 GX1
GX2 30

GX3 20
3 GX4
G10 10
Stress (MPa)

σin/σout
0
2

-10

1 -20

-30

0 -40

-6000 -4000 -2000 0 2000 4000 6000


-6000 -4000 -2000 0 2000 4000 6000
Longitudinal position (mm)
Longitudinal position (mm)
Fig. 18. Stress ratio of in-plane and out-of-plane stress of diaphragm-web rib
Fig. 16. Sub-loading of diaphragm-web rib weld. weld.

σleft significance buckling occurred.


4 The ratio of in-plane stress to out-of-plane stress of the diaphragm-
σright
web rib weld was shown in Fig. 18: the ratio was slightly greater than 1
3 σin under most conditions and only as the wheel loading approached the
σout fatigue detail, or was located above it, did the ratio increased to a
maximum of 43.4. Little out-of-plane stress was observed under this
2
Stress (MPa)

condition and influence of in-plane stress on this fatigue detail had


reached its greatest extent: the ratio of in-plane stress to out-of-plane
1 stress was between 1.0 and 1.5, indicating that the in-plane stress was
generally greater than the out-of-plane stress, while the effects of out-
of-plane stress on the fatigue damage could not be ignored.
0

4.3. The Diaphragm-floor rib weld


-1

The wheel loading was applied to the steel bridge deck as described
-2 in Section 4.1. The normal stress history of diaphragm-floor rib weld
-6000 -4000 -2000 0 2000 4000 6000 D3.1 was shown in Fig. 19. Alternating tensile and compressive stress
Longitudinal position (mm) cycles were observed under working conditions G0 to G7 while stress
cycles were entirely compressive under working conditions G8 to G25.
Fig. 17. Stress state of diaphragm-web rib weld. The peak tensile stress was reached when the wheel loading was ap-
plied to the outside of the curve (G0) transverse to the bridge. As the
GX4 and the stress history was as shown in Fig. 16. The peak tensile wheel loading moved from the outside to the middle of the diaphragm,
stress (4.2 MPa) was reached at G10, thus G10 was assumed to have
been the worst-case working condition (measured transverse to the
bridge). The center of the wheel loading was at a distance of 650 mm
from the diaphragm- web rib weld (measured transverse to the bridge)
under this condition, and the worst-case loading position longitudinal
to the bridge coincided with this fatigue detail. The cracking reason for
the diaphragm-web rib weld was analyzed on the basis of working
condition G10 in terms of its stress characteristics. The in-plane and
out-of-plane stresses on the diaphragm-web rib weld were calculated
under wheel loading based on the changes in σleft and σright (Fig. 17).
Unlike the stress characteristics of the diaphragm-U rib weld, curves
of σleft and σright were not mutually opposed but changed asymme-
trically, taking the diaphragm as the axis. This indicated that both in-
plane and out-of-plane stresses affected the stress distribution. The out-
of-plane stress distribution was symmetrical about the diaphragm while
that of the in-plane stress was asymmetrical. The in-plane stress was
generally greater than the out-of-plane stress, which tended to 0 as the
wheel load approached the diaphragm-web rib weld. This was because
in-plane tension or compression caused most deformation when the
wheel loading was located above the fatigue detail. The out-of-plane
deformation of the diaphragm-web rib weld was small unless Fig. 19. Stress cycles of diaphragm-floor rib weld.

8
Y. Wang, et al. Engineering Structures 201 (2019) 109767

σleft
1.5
σright
σin

1.0
σout

Stress (MPa)
0.5

0.0

-0.5
-6000 -4000 -2000 0 2000 4000 6000
Longitudinal position (mm)

Fig. 20. Sub-loading of diaphragm-floor rib weld. Fig. 21. Stress state of diaphragm-floor rib weld.

the tensile stress gradually decreased and tended to 0 as the wheel


loading approached the fatigue detail. When the wheel loading moved 10
to the inside of the curve (measured transverse to the bridge), the
tensile stress became compressive and the peak value thereof continued
to increase at the same time. The stress distribution in a diaphragm- 8
floor rib weld followed a single-peak shape, while the stress varied
between −2.0 MPa to 1.5 MPa. The worst-case loading position (mea-
σin/σout

6
sured transverse to the bridge) was located between G0 and G1 ac-
cording to the peak tensile stress, and this region was refined to find the
accurate position of the most unfavourable loading case.
4
Divide interval G0 and G1 into 100 mm increments (GX1 to GX4,
respectively): wheel loading was applied to GX1 to GX4 and the stress
history was as shown in Fig. 20. The peak tensile stress (1.4 MPa) was 2
reached at G0, thus G0 was assumed to be the most unfavourable
working condition (measured transverse to the bridge). The center of
the wheel loading was 4640 mm from the diaphragm-floor-rib weld 0
(measured transverse to the bridge) under this condition, and the most -6000 -4000 -2000 0 2000 4000 6000
unfavourable loading position (measured longitudinal to the bridge) Longitudinal position (mm)
coincided with this fatigue detail.
The cracking reason for the diaphragm- floor rib weld was analysed Fig. 22. Stress ratio of in-plane and out-of-plane stress of diaphragm-floor rib
based on working condition G0 in terms of its stress characteristics. The weld.
in-plane and out-of-plane stresses on the diaphragm-floor rib weld were
calculated under wheel loading on the basis of the changes in σleft and the outside of the curve. The fatigue details on the outside of the curve
σright, as shown in Fig. 21. Curves of σleft, σright, and σin followed a single and corresponding details on the inside (as shown in Fig. 5 (D1.1, D1.2,
peak-shaped distribution while the out-of-plane stress remained near 0, D2.1, D2.2, D3.1, and D3.2)) were investigated. The worst-case loading
indicating the predominant effects of in-plane stress on the stress conditions were applied and the stress history was as shown in Fig. 23.
characteristics of these fatigue details. The peak of in-plane stress was As for the three cracking details examined in this study, the stresses
reached when the wheel loading approached, or was above, the fatigue on fatigue details on the outside were all greater than those on the
detail (measured longitudinal to the bridge). inside: the stress on a diaphragm-U rib weld on the outside was 1.8
The ratio of in-plane stress to out-of-plane stress in the diaphragm- times that on the inside, the stress on a diaphragm-web rib weld on the
floor rib weld was shown in Fig. 22: the ratio usually exceeded 4 under outside was 1.5 times that on the inside, and the stress on a diaphragm-
most conditions and the maximum value thereof reached 10.4. Only U rib weld on the outside was 3.0 times that on the inside. The results
when the in-plane stress changed from tensile to compressive, was the explained why the existing fatigue cracks in the diaphragm-web rib
ratio less than 1. It could be concluded that the in-plane stress was weld were all located on the outside of the curve.
much greater than the out-of-plane stress, and the effects of out-of-
plane stress on fatigue damage were negligible.
5. Verification of cracking features

4.4. Influence of the curvature on the stress state 5.1. Design of the experiment

The stress characteristics of fatigue details inside and outside the The cracking features in a diaphragm-U rib weld and a diaphragm-
curve were significantly different because the steel box girder was web rib weld were influenced by the eccentricity of vehicle loading and
usually under abnormal load. The development of fatigue cracks was structural characteristics according to their detection in an actual
also biased according to the cracking features in the real bridge, where bridge and FE analysis. Fatigue testing was conducted to verify the
the cracking details of the diaphragm-web rib weld were all located on results proposed above. Full-scale specimens were designed on the basis

9
Y. Wang, et al. Engineering Structures 201 (2019) 109767

40 D1.1
D1.2
Stress (MPa) 20

-20

-40

-60

-6000 -4000 -2000 0 2000 4000 6000


Longitudinal position (mm)
(a) Diaphragm-U rib weld
D2.1
4
D2.2
Fig. 25. Diaphragm-web rib weld specimen.
3
Stress (MPa)

specimen. The eccentricity of the applied wheel loading was simulated


2 by setting the fatigue machine eccentrically on the diaphragm. The
experiment equipment, and test specimen designs, were shown in
1 Figs. 24 and 25, respectively.
The nominal stress was measured 10 mm from the cracking weld toe
0 [21]. The nominal applied stress range was 100 MPa considering the
fatigue strength of those two details. The relationship between crack
-6000 -4000 -2000 0 2000 4000 6000 length and number of cycles was recorded in real-time, and cracking
Longing position (mm)
path was observed. The coordinates of a and b were determined after
(b) Diaphragm-web rib weld fatigue failure to obtain the tendency of fatigue cracking to propagate.
1.5 A paint marker was also used to sketch the cracking path. The nominal
D3.1
D3.2 stress range of both specimens was read from the strain gauge after
1.0
fatigue loading. The nominal stress range applied to the diaphragm-U
rib weld on the eccentric side was 100.3 MPa and that on the non-ec-
Stress (MPa)

centric side was 70.5 MPa, thus eccentric effects of wheel loading would
0.5 result in significant differences in the stress on the weld on both sides of
the U rib. The nominal stress range on the diaphragm-web rib weld was
100.1 MPa.
0.0 The cracking features of specimens were observed after fatigue
loading. As for the diaphragm-U rib weld, cracks initiated from the
welding end of the connection between the rib and diaphragm, and
-0.5
-6000 -4000 -2000 0 2000 4000 6000 then propagated at a certain angle with the weld in the diaphragm. The
Longitudinal position (mm) cracking length on the eccentric side was much greater than that on the
(c) Diaphragm-floor rib weld non-eccentric side, which was similar to the cracking found in the ac-
tual bridge. With regard to the diaphragm-web rib weld, the cracks
Fig. 23. Stress states of fatigue details outside and inside the curve.
therein usually started at the welding end, and then developed asym-
metrically in the diaphragm until fatigue failure, matching the actual
behaviour in the actual bridge. It was therefore assumed that the
loading method applied in the fatigue experiment could reflect the
stress state around fatigue details in the actual bridge.

5.2. Analysis of cracking path

The cracking path in a diaphragm-U rib weld on the eccentric and


non-eccentric side was shown in Fig. 26(a): the tangents on both sides
of the specimen (the bold, dashed lines) were obtained for the cracks
upon initiation and after fatigue failure. The propagation angles were
calculated based on these tangents. As for the crack on the eccentric
side, the angle between the crack and weld was 10.73° upon initiation
and 33.31° after fatigue failure. With regard to the crack on the non-
eccentric side, the angle between the crack and weld was 17.21° upon
Fig. 24. Fatigue testing of diaphragm-U rib weld.
initiation and 47.62° after fatigue failure. The angle between the crack
weld was smaller upon initiation, indicating that the crack initially
of the geometric dimensions of fatigue details in the actual bridge. Eight propagated along the weld. The angle increased with increasing of
holes were drilled on one side of the deck, by which the specimen was crack length and the crack gradually deviated from the weld and de-
fixed to the loading frame using high-strength bolts. Four holes were veloped to the edge of the diaphragm. Besides, the eccentricity of
drilled on the other side to fix the fatigue testing machine to the loading had great influence on the cracking angle, and the angle

10
Y. Wang, et al. Engineering Structures 201 (2019) 109767

120
Non-eccentric side 160 Non-eccentric side
Eccentric side 140 Eccentric side
100

Crack length/mm
120
80 100
80

b/mm
60
60
40 40
20
20
0
0 100 200 300 400
0 Cycles(×104)
-120-100 -80 -60 -40 -20 0 20 40 60 80 100 120 (a) Diaphragm- U rib weld
a/mm 150
Non-eccentric side
(a) Diaphragm-U rib weld 125 Eccentric side
80

Crack length/mm
Non-eccentric side 100
60
75
40
50

20
b/mm

25

0
0
Eccentric side 50 100 150 200 250 300 350
-20 Cycles(×104)
(b) Diaphragm-web rib weld
-40
Fig. 27. Rates of crack propagation.
-150 -100 -50 0 50 100 150
a/mm
as soon as the crack was initiated on the non-eccentric side. Specifically,
(b) Diaphragm-web rib weld if the fatigue cracks on the eccentric side propagated rapidly, so would
Fig. 26. The cracking path. the fatigue cracks on the non-eccentric side. If the fatigue cracks on the
eccentric side developed slowly, so would the fatigue cracks on the non-
eccentric side. Therefore, the propagation of a crack on the non-ec-
between the crack and weld on the eccentric side was smaller than that
centric side would be greatly influenced by that on the eccentric side,
on the non-eccentric side.
the trend in which was the same as for cracks No. 1.1 and No. 1.2 in the
The cracking path of a diaphragm-web rib weld on the eccentric and
actual bridge. In practice, it is necessary to repair the cracked dia-
non-eccentric side was shown in Fig. 26(b): the cracking lengths par-
phragm-U rib weld timely in case rapid crack propagation on one side
allel and perpendicular to the weld were different for cracks on the
results in rapid crack development on the other.
eccentric, and non-eccentric, sides. The eccentric loading resulted in
The rate of crack propagation in the diaphragm-web rib weld was
adverse directions of the principal stresses on both sides of the welding
shown in Fig. 27(b). The cracks on the eccentric and non-eccentric side
end, and cracks on the eccentric side showed greater deviation from the
presented similar rates of propagation with increasing numbers of cy-
weld. This cracking feature in the real bridge could be explained as
cles. The fatigue lives under the same crack length were similar.
following: the welding end connecting the diaphragm and the deck or
Comparing this experimental behaviour to that in the actual bridge
floor could be seemed as the cantilever. The in-plane and out-of-plane
(Fig. 6(b)), it could be observed that crack lengths on both sides were
deformation of the welding end was caused under the wheel loading.
similar to those in the actual bridge, so it could be concluded that fa-
The upside welding end of the diaphragm was comparatively farther
tigue test data were consistent with reality.
from the diaphragm-web rib weld, resulting in the greater cantilever
length. As the cracks usually propagated perpendicular to the normal
6. Conclusions
stress, the upside welding end of the diaphragm had significantly
changed the direction of the normal stress, thus cracks on the eccentric
The cracking reasons and features of fatigue details in the dia-
side deviated more from the diaphragm-web rib weld.
phragm of curved steel box girder were studied based on an actual
curved steel bridge with three spans. The influence of the curvature on
5.3. Analysis of cracking rate the fatigue characteristics was investigated and the reasons for cracking
behaviours were deduced from in situ monitoring, FE analysis, and fa-
The rate of crack propagation in a diaphragm-U rib weld was shown tigue tests. The following conclusions were drawn:
in Fig. 27(a). The number of cycles required for initiation of the crack
on the eccentric side exceeded that on the non-eccentric side because (1) The unevenly distributed loadings caused by the curvature resulted
stress ranges on the eccentric side were greater than those on the non- in greater stresses of details on the outside than that on the inside,
eccentric side. Interference could be observed between the fatigue which was the major difference in the fatigue characteristics be-
cracks on the diaphragm-U rib weld on both sides of an identical U rib. tween curved and straight steel box girders. This also explained
Fatigue cracks on the eccentric and non-eccentric sides presented the why existing fatigue cracks in a diaphragm-web rib weld were all
same trend, and the rate of crack propagation in a diaphragm-U-rib located on the outside of the curve.
weld on the non-eccentric side was similar to that on the eccentric side

11
Y. Wang, et al. Engineering Structures 201 (2019) 109767

(2) Based on the structural parameters of the actual bridge, out-of- doi.org/10.1016/j.engstruct.2019.109767.
plane deformation caused fatigue cracking in the diaphragm-U rib
weld. The cracking in the diaphragm-web rib weld was induced by References
the combination of in-plane and out-of-plane deformation. The
predominant factor affecting fatigue crack initiation in the dia- [1] Thanh L, Itoh Y. Performance of curved steel bridge railings subjected to truck
phragm-floor rib weld was in-plane deformation. collisions. Eng Struct 2013;54:34–46.
[2] Seo J, Linzell DG. Use of response surface metamodels to generate system level
(3) The eccentricity of wheel loadings and structural characteristics fragilities for existing curved steel bridges. Eng Struct 2013;52:642–53.
were major reasons that caused asymmetric crack propagation in [3] Guo T, Liu J, Zhang Y, Pan S. Displacement monitoring and analysis of expansion
the diaphragm-web rib weld. The eccentric loading resulted in ad- joints of long-span steel bridges with viscous dampers. J Bridge Eng
2015;20(9):04014099.
verse directions of the principal stresses on both sides of the [4] Maljaars J, Vrouwenvelder T. Fatigue failure analysis of stay cables with initial
welding end, and the greater deviation of cracks on the eccentric defects: Ewijk bridge case study. Struct Saf 2014;51(6):47–56.
side of the weld was influenced by the structural characteristics, but [5] Sennah KM, Kennedy JB. State-of-the-art in design of curved box-girder bridges. J
Bridge Eng 2001;6(3):159–67.
the crack lengths on both sides were similar. [6] Siringoringo DM, Fujino Y. Dynamic characteristics of a curved cable-stayed bridge
(4) Interference could be observed between fatigue cracks in the dia- identified from strong motion records. Eng Struct 2007;29(8):2001–17.
phragm-U rib weld on both sides of an identical U rib. The eccen- [7] Guo T, Frangopol DM, Chen Y. Fatigue reliability assessment of steel bridge details
integrating weigh-in-motion data and probabilistic finite element analysis. Comput
tricity of wheel loading caused the early propagation of cracks on
Struct 2012;112:245–57.
the eccentric side while a similar rate of crack propagation was [8] Fu Z, Ji B, Ye Z, Wang Y. Fatigue evaluation of cable-stayed bridge steel deck based
maintained on the non-eccentric side upon onset of cracking. In on predicted traffic flow growth. KSCE J Civ Eng 2017;21(4):1400–9.
engineering practice, it is necessary to repair cracking in the dia- [9] Kainuma S, Jeong YS, Yang M, Inokuchi S. Welding residual stress in roots between
deck plate and U-rib in orthotropic steel decks. Measurement 2016;92:475–82.
phragm-U rib weld timely to prevent rapid crack propagation on [10] Cui C, Zhang Q, Luo Y, Hao H, Li J. Fatigue reliability evaluation of deck-to-rib
one side resulting in the rapid development of cracks on the other. welded joints in osd considering stochastic traffic load and welding residual stress.
Int J Fatigue 2018;111:151–60.
[11] Kim YC, Hirohata M, Shibata K, Chang KH. Detection and monitoring by FSM for
Declaration of Competing Interest fatigue crack at invisible location of steel plate deck with U-rib. Int J Steel Struct
2013;13(1):183–9.
We declare that we have no financial and personal relationships [12] Connor RJ, Fisher JW. Consistent approach to calculating stresses for fatigue design
of welded rib-to-web connections in steel orthotropic bridge decks. J Bridge Eng
with other people or organizations that can inappropriately influence 2006;11(5):517–25.
our work, there is no professional or other personal interest of any [13] Fu Z, Wang Y, Ji B, Jiang F. Effects of multiaxial fatigue on typical details of or-
nature or kind in any product, service and/or company that could be thotropic steel bridge deck. Thin-Walled Structures 2019;135:137–46.
[14] Zhang QH, Cui C, Bu YZ, Liu YM, Ye HW. Fatigue tests and fatigue assessment
construed as influencing the position presented in, or the review of, the approaches for rib-to-diaphragm in steel orthotropic decks. J Constr Steel Res
manuscript entitled, “Cracking reasons and features of fatigue details in 2015;114:110–8.
the diaphragm of curved steel box girder” [15] Zong L, Shi G, Wang Y. Experimental investigation on fatigue crack behavior of
bridge steel q345qd base metal and butt weld. Mater Des 2015;66:196–208.
[16] Yang M, Kainuma S, Jeong YS. Structural behavior of orthotropic steel decks with
Acknowledgements artificial cracks in longitudinal ribs. J Constr Steel Res 2018;141:132–44.
[17] Zhou Q. Effect of small radius on the performance of curved steel box girder bridge.
The authors would like to appreciate the support of Nagoya Appl Mech Mater 2014;484–485:727–30.
[18] Li X, Cai J, Qiang S. Models of cable-girder anchorage for long-span cable-stayed
Expressway Association, The National Key Research and Development bridges with steel box girder. Eng Mech 2004;21(6):84–90.
Project (No. 2017YFE0128700), The Fundamental Research Funds for [19] Japanese society of highway. The specification for fatigue design of highway steel
the Central Universities (Grant 2018B56814) and Qing Lan Project. bridge; 2002 [In Japanese].
[20] International Institute of Welding (IIW). Recommendations for Fatigue Design of
Welded Joints and Components. Paris; 2008.
Appendix A. Supplementary material [21] Fu Z, Ji B, Kong X, Chen X. Grinding treatment effect on rib-to-roof weld fatigue
performance of steel bridge decks. J Constr Steel Res 2017;129:163–70.
Supplementary data to this article can be found online at https://

12

You might also like