FE-analysis of Vårby Bridge: Robert Bengtsson Mikael Widén

FE-analysis of Vårby Bridge
Investigation of fatigue damage in a composite bridge

Master of Science Thesis in the Master’s Programme Structural Engineering and
Building Performance Design
ROBERT BENGTSSON
MIKAEL WIDÉN
Department of Civil and Environmental Engineering

Division of Structural Engineering
Steel and Timber Structures
CHALMERS UNIVERSITY OF TECHNOLOGY
Göteborg, Sweden 2010
Master’s Thesis 2010:85
MASTER’S THESIS 2010:85

ROBERT BENGTSSON
MIKAEL WIDÉN

CHALMERS UNIVERSITY OF TECHNOLOGY
ROBERT BENGTSSON
MIKAEL WIDÉN
© ROBERT BENGTSSON & MIKAEL WIDÉN, 2010
Examensarbete / Institutionen för bygg- och miljöteknik,

Chalmers tekniska högskola 2010:85

Chalmers University of Technology
SE-412 96 Göteborg
Sweden
Telephone: + 46 (0)31-772 1000
Cover: Moving load in right traffic lane

Chalmers Reproservice / Department of Civil and Environmental Engineering
ROBERT BENGTSSON
MIKAEL WIDÉN
Chalmers University of Technology
ABSTRACT
Vårby Bridge was built in Stockholm in 1996 and serves as a highway bridge. The
bridge consists of two adjacent composite bridges with three lanes of traffic each.
During inspection of the bridge in the summer of 2006, only ten years after the bridge
was opened for traffic, fatigue cracks were observed in several welded connections
between the vertical web stiffeners and the upper flange of the main girders. Common
for all detected cracks is that they occur only where there is a single vertical stiffener,
i.e. not over support.
The purpose of this master thesis was to investigate the mechanisms behind crack
initiation, why the fatigue cracks only appear at spans and if cracking is deformation
or load induced.
The bridge was modelled with shell elements for the entire bridge, including the
concrete deck in the FEM software program Abaqus. First a global model was created
and used to verify the accuracy of the model by comparing the bending stresses at
certain locations in the main girders with the corresponding values from the on-site
strain measurements. In addition to this comparison, a parameter study was
performed, including load locations and stiffness.
To capture the stress gradients in the vertical web stiffener of interest, a refined model
was created by modifying the global model with a finer mesh in this area and by
modelling the applied truck load in a more detailed way.
In the literature study a number of possible mechanisms were found that might induce
fatigue cracks in the welded joints studied. A theory why cracks have been found in
Vårby Bridge and why only in span is presented. The result of the theory presented
argues that cracks are formed by load applied directly above the vertical web
stiffeners together with residual stresses causing fatigue cracks.
Keywords: composite bridge, fatigue cracks, vertical web stiffener, Abaqus, FEM,
global model, refined model
I
FE-analys av Vårbybron
Undersökning av utmattningsskador i en samverkansbro
Examensarbete inom Structural Engineering and Building Performance Design
ROBERT BENGTSSON
MIKAEL WIDÉN
Institutionen för bygg- och miljöteknik
Avdelningen för Konstruktionsteknik
Stål- och träbyggnad
Chalmers tekniska högskola
SAMMANFATTNING
Bron som leder E4-an över Vårby-fjärden färdigställdes under 1996 och består av två
nästan identiska samverkansbroar, en för vardera trafikriktning. Under en visuell
inspektion av bron sommaren 2006 hittades utmattningssprickor i flertalet svetsar
mellan vertikala livavstyvningar och överfläns på huvudbalkarna. Sprickorna hittades
i fältsnitt men inte i stödsnitt.
Målet med examensarbetet var att tillsammans med Ramböll Luleå undersöka varför
utmattningssprickorna i Vårbybron endast uppstår i fältsnitt. För att lösa detta gjordes
FE-analyser i programvaran Abaqus som jämfördes med töjningsmätningar utförda på
Vårbybron.
Först gjordes en global modell för att jämföra och verifiera böjstyvheten hos bron.
Detta gjordes med en grov elementindelning och simulering av fordon med hjälp av
punktlaster. Resultatet visade god överensstämmelse mellan uppmätta värden och
värden från Abaqus.
Därefter förfinades modellen med avseende på elementindelning och laster för att
erhålla bättre spänningsfördelning i området kring svetsarna som studerades.
Jämförelsen mellan töjningsmätningar och värden från modellen visade ingen god
överensstämmelse på grund av att samverkan mellan betong och stål i bron inte var
korrekt modellerat.
Resultat kunde ändå användas för att presentera en teori om varför sprickorna har
uppkommit och varför de bara finns i fältsnitt. Trafiklast placerad över en huvudbalk
leder till stora tryckspänningar vilket tillsammans med egenspänningar i svetsarna gör
att spänningsintervallet för en lastcykel kan vara upphovet till sprickorna. I
litteraturstudien presenteras även andra tänkbara teorier för uppkomst av sprickor
vilka kan utgöra en förhöjning av spänningsnivåerna när olika lastfall inträffar
samtidigt.
Nyckelord: samverkansbro, utmattning, sprickor, vertikal livavstyvning, Abaqus,

FEM
II
TABLE OF CONTENTS
ABSTRACT I
SAMMANFATTNING II
TABLE OF CONTENTS III
PREFACE V
1 INTRODUCTION 1
1.1 Background 1
1.2 Aim and objectives 1
1.3 Limitations 1
1.4 Methodology 1
2 LITERATURE STUDY 2
2.1 Studied reports 2
2.2 Background 2
2.3 Damage and cause 2
2.4 Retrofitting methods 4
2.5 FE-modelling 5
2.6 Experimental test 6
2.7 Summary 7
3 VÅRBY BRIDGE 9
3.1 Description 9
3.2 Dimensions and general features 9
3.3 Damage 12
3.4 Measurements 14
3.4.1 Location of measurement equipment 14
3.4.2 Strain measurement with known vehicle data 16
4 PRE-MODELLING PHASE 18
4.1 Pre-modelling phase 18
5 GLOBAL FE-ANALYSIS 21
5.1 Introduction 21
5.2 Steel details 21
5.3 The deck 23
5.4 Loads 24
CHALMERS Civil and Environmental Engineering, Master’s Thesis 2010:85

III
5.5 Boundary conditions 24
5.6 Mesh evaluation 26
5.7 Approximations 28
5.7.1 Stiffness of the cross section 28
5.7.2 Bridge dimensions 29
5.7.3 The connection between diaphragm and main girders 30
5.7.4 Global location of the main girders 30
5.7.5 Connections between the concrete deck and the top flanges 30
5.7.6 End-diaphragms 31
5.7.7 Concrete deck 31
5.8 Verification of the global bridge model 31
6 REFINED MODEL 37
6.1 Introduction 37
6.2 Comparison between refined model and measurement 37
6.3 Approach 1 38
6.4 Approach 2 39
7 CONCLUSIONS 41
7.1 Theory of why cracks have occurred 43
7.2 Further studies 43
8 REFERENCES 44
Appendix A: Drawings of Vårby Bridge

Appendix B: Hand calculations of small-scale model
Appendix C: Mesh evaluation
Appendix D: Additional diagrams from parameter study
CHALMERS, Civil and Environmental Engineering, Master’s Thesis 2010:85

IV
PREFACE
This master thesis was carried out between December 2009 and June 2010 at the
Division of Structural Engineering at Chalmers University of Technology together
with Ramböll Luleå.
We would like to thank all people involved in this project especially Mattias Nilsson,
supervisor at Ramböll, and Mohammad Al-Emrani, supervisor and examiner at
Chalmers.
Big thanks to Farshid Zamiri A. and Mustafa Aygül for valuable input and help along
the way.
Our roommates at the department, Nikolaus Ljungström and Olof Karlberg, also
deserve thanks for keeping a good and joyful working climate.
Göteborg October 2010

Robert Bengtsson & Mikael Widén
CHALMERS Civil and Environmental Engineering, Master’s Thesis 2010:85

V
VI
1 Introduction
1.1 Background
Vårby Bridge was built in Stockholm in 1996 and serves as a highway bridge. The
bridge consists of two adjacent composite bridges with three lanes of traffic each.
During inspection of the bridge in the summer of 2006, only ten years after opened for
traffic, fatigue cracks were observed in several welded connections between the
vertical web stiffeners and the upper flange of the main girders. In common for all
detected cracks is that they occur only where there is a single vertical stiffener, i.e. not
over support.
The Swedish road authority gave Ramböll AB the task to investigate why the detected
cracks have occurred and which measures need to be taken – if any – to prevent
possible further propagation of the cracks and ensure integrity of the structural system
of the bridge. This master thesis is a part of the investigation conducted by Ramböll,
and part of the European project “Brifag” in which Ramböll and Chalmers are
involved, why the fatigue cracks occur in the bridge and to find possible ways of
retrofitting if necessary.
1.2 Aim and objectives

The aim of the master thesis is to investigate the reasons for the formation of the
fatigue cracks found in the Vårby Bridge and the specific objectives are:
• To find possible stress concentrations and what effect(s) that generate
fatigue damage in the bridge
• To investigate why cracks only occur in span and not over the supports
• To find suitable retrofitting methods if necessary
1.3 Limitations
This master thesis will only investigate why fatigue cracks occur in the connection
between vertical web stiffeners and upper flanges of the main girders. Retrofitting
method of the studied welded joints will be examined if the cause of damage is
established.
1.4 Methodology
To reach the stated objectives the following methodologies are used:
• A literature study of similar bridges that suffer from comparable fatigue
damage
• A numerical analysis of the bridge by creating FE-models in the software
Abaqus/CAE 6.8-2

1
2 Literature study
2.1 Studied reports
In the study focus has been to gather relevant findings from literature available.
Interesting information for this thesis is why fatigue cracks occur in the welded area,
results from experimental tests, how FE-modelling has been performed and how to
retrofit cracks. Only three reports of interest were found and the main reasons are little
investigations have been made of the specific bridge construction and language
difficulties.
2.2 Background
One of the reports is from Austria (Greiner & Taras, 2009) and investigates a
continuous composite bridge, Figure 2.1, which suffers from fatigue cracks. The
bridge is 560 m long with spans varying between 70-84 m. In the report a comparison
by numerical analysis was made with a second bridge from Austria with structural
conditions that were similar to the main object. Also retrofitting methods are tested
and evaluated.
Figure 2.1: Studied bridge by Greiner & Taras, 2009
The second report (Sakano, Kawakami, Sakai, & Matsushita, 2007) states that one of
the most common fatigue damages in orthotropic steel deck bridges in Japan is
cracking in the weld between the deck and the vertical stiffener. The purpose with the
report was to evaluate a retrofitting method when cracking has occurred. A fatigue
test was done to see the performance of the proposed retrofitting method.
The third report (Okura, Shiozaki, Fukumoto, & Nanjyo, 1995) also has its
background in Japanese highway bridges suffering from fatigue cracking at welded
vertical stiffeners. Focus in this report lies on how the arrangement of stud shear
connectors in a composite bridge influence fatigue cracking in the welded joints
between a connection plate and web and top flange.
2.3 Damage and cause

The damages found in the Austrian bridge are both root and toe cracks in the welded
joint between vertical stiffener and the upper flange of the main girders, see Figure

2
2.2. Cracks are found in about 70 % of all these joints. The main causes for the fatigue
cracks are change in the bridge construction due to more traffic and that fatigue was
not considered in code regulations at the time of design for the bridge.
(a) (b)
Figure 2.2: (a) root crack; (b) toe crack, Greiner & Taras (2009)
Originally the deck cantilever was smaller, but due to the more traffic the deck was
broadened, Figure 2.3, which lead to a larger cantilever and greater hogging moment.
Additionally when the deck was broadened, the deck thickness was decreased to keep
the self weight constant.
Figure 2.3 Original and broadened deck, Greiner & Taras (2009)
In the report by Sakano et al. (2007) weld toe cracking is considered to initiate when
heavy traffic pass over the location of the vertical stiffeners creating severe stress
concentrations.
The investigation by Okura et al. (1995) was also regarding the welded joint between
vertical stiffeners and the top flange. In Figure 2.4 the different crack types considered
in the report can be seen. With consideration to this thesis only one of the types
investigated are of interest, type 1 which can be either root or toe crack. No particular
cause for damage was mentioned in the report other than the rotation of the slab
played a dominant role for type 1 fatigue cracks.

3
Figure 2.4 Cracks considered by Okura et al.(1995)
2.4 Retrofitting methods

Greiner & Taras (2009) proposed four methods of retrofitting seen in Figure 2.5 and
explained below.
a) Replacing fillet welds with full-penetration welds
b) Weld a plate between the flange and the vertical stiffener
c) Make a cut-out at the top of the vertical stiffener
d) Use gusset plates and pre-stressed bolts to be able to remove pre-damaged

material close to the cracked welds and place new full-penetration welds in the
joint.
(a) (b) (c) (d)
Figure 2.5: Four methods of retrofitting; (a)-(d), Greiner & Taras (2009)
Method a) increases the fatigue strength in the joint with a factor 2, though it is stated
that it is not beneficial to use when repairing cracks in bridge due to the fact that only
small portions of the base material can be removed without causing other damage.
Method b) did not work since fatigue cracks developed in the fillet welds connecting
the plate to the top flange.
Method c) works in principal, when aiming to transfer the stress maximum to the cut-
out in the vertical stiffener and thereby to higher fatigue strength. It is though stated
that the geometry of the cut-out should be made in an optimised way and that that is
very difficult to obtain when repairing fatigue cracks.
Method d) works though the use of pre-stressed bolts appeared necessary to not cause
local stress concentrations due to slip of the plates.

4
Also Sakano et al. (2007) proposes the use of method c) above and concludes that the
stress concentration in the area can be lowered but the propagation of cracks does not
stop, though it slows down.
2.5 FE-modelling
In the report by Greiner & Taras, (2009) a comparison was made by numerical
analysis between the investigated bridge and another bridge with similar structural
conditions. Field measurements were used to verify the model. The results presented
stated the magnitude of stresses (tension) at the end of the top edge of vertical
stiffener to be between 70-115 MPa in the model and 60-80 MPa from field
measurements. The difference was regarded to occur due to difficulties to obtain the
correct stiffness in the area. Conclusions in the report stated that the stresses in the
model decreases to about 40 % of the value above, when top bracing was not attached
to the stiffener and a decrease to 60 % when the studs were out of alignment with the
stiffener. The rotation of the concrete slab over the main girder was 0.10° for the
investigated bridge and 0.037° for the other bridge. The difference between the
bridges was concluded to be the greater slab thickness in the second bridge even
though it had a larger cantilever.
A FE-analysis was also performed by Sakano et al. (2007) to evaluate how the stress
concentration changes when making a cut-out in the vertical web-stiffener, see Figure
2.6. The result from the analysis stated the magnitude of stresses (compression),
before the cut-out, to -961 MPa at the weld-end at the stiffener-side. After the cut-out
was made the stress was decreased to -543 MPa and the magnitude of stress in the
crown of circle.
Figure 2.6: Stress concentration, cut out hole in the vertical web stiffener, Sakano
et al. (2007)

5
In the report by Okura et al. (1995) FE-modelling was performed to establish a load-
transfer model between the steel deck and the top flange of the girder, Figure 2.7. The
connection was modelled using beam elements as stud, with the proper stiffness, and
rigid beams with hinges at one side of the flange depending on the rotation of the
cross beam and deformation of the deck, Figure 2.8. The load-transfer model was the
used to evaluate the influence of stud spacing and arrangement. The conclusions,
made by Okura et al. (1995), stated that the spacing and arrangement of studs had no
influence when looking on fatigue cracks at the top of the vertical stiffener for this
specific case.
(a) (b)
Figure 2.7: Definitions of rotation and deformation; (a) positive; (b) negative,
Okura et al. (1995)
(a)
(b)
Figure 2.8: Definitions of rotation and deformation; (a) positive (b) negative,
Okura et al. (1995)
2.6 Experimental test

The experimental test presented by Sakano et al. (2007), and the only of interest, was
a fatigue test with the aim to run until toe cracks were formed at the fillet welds on
both deck-side and stiffener-side. When cracks had formed the retrofitting method
mentioned above was performed and the fatigue test continued and evaluated to see
the effects of the retrofitting. The test was performed on a full-size specimen, Figure
2.9. It is an orthotropic steel deck with trough ribs welded to it and three vertical
stiffeners were used, denoted R1-L, R1-R and R2-L

6
Figure 2.9: The tested specimen by Sakano et al (2007)
After testing was finished it was stated that the crack propagation on the deck side of
the fillet weld was stopped, while on the stiffener side propagation continued, though
very slow.
2.7 Summary
All the studied reports have the background from bridges suffering fatigue cracks in
the main girders at the welded joint between the top flange and the vertical web
stiffener.
The cracks found are both weld toe and root cracks.
The causes of initiation of the cracks from the literature are:
• Hogging moment when traffic runs on the cantilever of the deck causing too
high tensile stresses in the weld.
• When the traffic load is applied over the vertical stiffener causing high
compressive stresses in the top edge of the vertical stiffener together with
tensile residual stresses in the welds.
Two causes for crack initiation were found during the literature study. This might not
be the effects causing fatigue cracking in the bridge investigated in this thesis.
Therefore a few additionally possible causes are stated below:
• When traffic load is applied in the centre of the deck transversally, causing
rotation of the deck which generates high compressive stresses together with
tensile residual stresses in the welds.
• Another cause of cracking when traffic load is applied on the main girders,
apart from the one above, is the influence of the rotational rigidity of the cross
section. If the rigidity is too small the displacement of the main girders will be
uneven i.e. the loaded girder will rotate around the other girder causing
compressive stresses in the weld at the stiffener on the unloaded girder.
Together with tensile residual stresses this can cause cracking. In addition to
compressive stresses also shear stresses can occur in the weld due to the
influence of cross beams attached. This behaviour will though not be at the
support since the rotation is prevented.
The FE-analysis performed can be summarised with the conclusion that it is difficult
to model the stiffness of the cross section in a correct way and that the results tend to
be higher in the model than from field measurements.

7
Regarding retrofitting method a cut-out in the stiffener would work to lower the
magnitude of the stress concentration according to Sakano et al. (2007), though
Greiner & Taras, (2009) suggests it is difficult in practice due to the cut-out needs to
be made with geometrically good precision and therefore difficult and time
consuming to apply on a bridge. Further Greiner & Taras, (2009) proposes two
methods that work both in theory and for use in practice.
Additionally a thought about the necessity to apply retrofitting methods is if the
fatigue cracking is load or deformation induced. The methods above is to prevent load
induced cracks, if deformation induced cracks are formed it will perhaps be enough to
make control measurement of the cracks in certain intervals to ensure that they does
not propagate further.

8
3 Vårby Bridge
3.1 Description
In the summer of 2006 an inspection of the bridge was made and several fatigue
cracks were found. To find the origin of the cracks Ramböll AB was given the task to
investigate the bridge. As a part of the investigation strain measurements were
performed on the bridge (James & Kärrfelt, 2009). Below the bridge dimensions are
presented followed by a description of the damage found and also how the
measurements were carried out.
Since the bridge consists of two adjacent composite bridges only one of these will be
described below. The bridge chosen is the one called the “North Bridge”, because the
strain measurements were performed there.
3.2 Dimensions and general features

The bridge is a continuous bridge with six spans over seven supports. The four inner
spans are of equal length, 44m, and the end spans are about 38 m long, Figure 3.1.
The main girders in the spans are of varying height, where the height is at maximum
over the support and minimum at mid-span. Between support and mid-span the girder
height varies linearly, naturally with the top of the girder kept horizontal.
Figure 3.1: Elevation of the six spans in Vårby Bridge. (1) and (7) corresponds to
end supports and (2) – (6) mid-supports.
The composite bridge consists of steel I-girders and a concrete deck. Between the
main girders there are cross-beams in the shape of I-beams working as diaphragms.
The functions of the cross-beams were both during the casting of the deck and when
the bridge was finished. During casting they were used to prevent lateral buckling of
top flanges in compression and when the bridge was finished to provide resistance
against lateral forces and to distribute hogging bending moment from main girders.
The cross-beams are attached to the vertical web-stiffeners by bolts through the
vertical web stiffeners, Figure 3.2. Although the girder heights vary, the distance
between the top flanges in the main girders and the top of the cross-beams remains
constant to 400 mm which can be seen in Figure 3.3.

9
Figure 3.2: Diaphragm in span where the two main girders are connected by a
cross-beam through bolts. The two main girders are displaced in
vertical direction due to the road building of the highway
In total, there are 28 cross-beams in the 6 spans and also one cross-beam over each
support. Three different cross section types are used all over the bridge at the
positions of the vertical stiffeners in the spans. Over the intermediate supports one
cross section type is used and also one type for the two end supports.
Figure 3.3: A constant distance of 400 mm between the top flange and cross beam
for all diaphragms in Vårby Bridge
There is one difference between the cross sections for the spans compared to the
supports, over the supports there are double vertical web stiffeners and in the spans
there is a single stiffener. Figure 3.4 show a diaphragm at support with two vertical
web stiffeners. The geometry of the different types of cross sections and the
remaining dimensions for the two main girders etc, are presented in Appendix A.
Figure 3.4: Diaphragm at support with double vertical web stiffener
The concrete deck has a free width of 14000 mm. The deck varies in thickness
between 350 mm, which is at the area connected to the top flanges of the main

10
girders, 170 mm which are at the end in the close proximity to the end beams and 280
mm along the centre line of the bridge.
Figure 3.5: The concrete deck with dimensions
To receive the composite action between the concrete deck and the main girders studs
are used in Vårby Bridge visible in figure 3.3. The studs are placed in two rows along
the bridge with spacing of 550 mm in transversal direction and in longitudinal
direction the spacing varies between 150, 190 and 260 mm.

11
3.3 Damage
During the visual inspection in 2006 cracks were found at several welded connections
between the vertical web stiffeners and main girder top flange, see Figure 3.6. The
cracks found were weld toe cracks, see Figure 3.7, caused by fatigue.
Figure 3.6: Location of fatigue cracks, James & Kärrfelt, (2009)
Figure 3.7: Toe crack at a vertical web stiffener, James & Kärrfelt, (2009)
The damages were observed all over the bridge without any particular pattern, Figure
3.8. One thing though when observing the location of the cracks is important, there
are no cracks found in the connections at the supports.

12
Figure 3.8: Fatigue cracks observed at the visual inspection 2006, James &
Kärrfelt, (2009)
In May 2009 a more thorough inspection of one span was made which lead to an
increase in cracks found. Previously 2 cracks were noticed but the new investigation
found 7, including the old ones, see Figure 3.9. Also the total lengths of the cracks
were measured and the magnitude lies in the range of 15-30 mm. The inspection was
made because the measurements were later performed in this span. No more
inspection has been made on the bridge to the awareness of the authors of this thesis
though the additionally found cracks in the more investigated span indicates that there
are more cracks than shown in Figure 3.9
Figure 3.9: Fatigue cracks observed in the inspection 2009 in the span between
support 5 and 6 where the strain measurements are performed, James
& Kärrfelt, (2009)
To investigate the initiation and propagation one sample was deducted from a crack.

13
The conclusions drawn by Eriksson, (2009) were:
• No particular defects of the weld was found
• No defects found in the heat affected zone (HAZ)
• Multiple initiations, which means that two separate cracks initiated
independently of each other. This behaviour should appear after a large
amount of cycles.
• The cracks propagated and were merged together.
• The origin of initiation cannot be fully concluded due to severe corroded crack
surfaces.
• The corrosion decreases with depth of the cracks
• The propagation was determined to be disturbingly rapid with typical beach
marks and striations. The speed of the propagation was due to the fact that two
cracks were closing in on each other, leaving smaller and smaller thickness of
the base material.
• It was concluded that the cracks was a result of fatigue.
3.4 Measurements
The strain measurements performed on the bridge were made in June 2009 by James
& Kärrfelt, (2009) and consisted of three different types of measurements:
• The first measurement was done when the equipment were placed on the
bridge. The outcome from this occasion was to be used to evaluate if the
position of the points of measurements were satisfying or needed correction.
Below the positions chosen and the locations of the equipment are presented.
• The second type of measurements was done when loading the bridge with a
truck of known weight. The idea behind this was to use the result when
calibrating a FEM-model of the bridge.
• A free flow measurement was also done during approximately 5 days to
evaluate what magnitude of stresses and stress range that could be expected
caused by ordinary traffic.
3.4.1 Location of measurement equipment

The measurement equipment was placed at three different diaphragms, position 1, 2
and 3 which can be seen in Figure 3.9 above. Each diaphragm in the bridge is denoted
with a letter and a number. Below this notation are used to describe which diaphragm
regarded. In Appendix A the notations are stated.
Position 1 is at diaphragm F23 where no cracks in the welded joints of interest were
found. The locations of the equipment are presented in Figure 3.10. As can be seen in
the figure there are strain gauges applied at the bottom of the main girders to measure
the global behaviour of the bridge and to obtain influence lines.

14
Figure 3.10: Measurement position 1, in span; no cracks observed, James &
Kärrfelt, (2009)
Position 2 is over support B6, which is also an cross section without cracks, and
chosen to use as comparison with the measurements in position 1 to answer the
question why cross sections above support do not suffer from fatigue cracks. The
locations of the equipment are seen in Figure 3.11.
Figure 3.11: Measurement position 2, at support; no cracks observed, James &

Kärrfelt, (2009)

15
Figure 3.12 Measurement position 3, in span; cracks observed, James & Kärrfelt,
(2009)
Position 3 is in the same span as position 1 at diaphragm F21. The cross sections at
F21 and F23 are identical geometrically, but F21 suffers from cracks in both welded
joints of interest and was therefore chosen. In Figure 3.12 the locations of the
equipment are shown.
3.4.2 Strain measurement with known vehicle data

As mentioned above the second type of measurements were made with a truck of
known weight and distances between tyres. After loading the truck with stone dust, of
0-8 mm size, the front axle and rear bogie were weighed to 7.38 and 14.26 tonnes
respectively. The distances between the tyres are shown in Figure 3.13.
(a) (b)
Figure 3.13: Distances between the tyres of the truck with known weight: (a) wheel
base, (b) wheel tracks
The measurements were performed during night due to that the bridge was not
allowed to be closed down for traffic. As mentioned before, the bridge has three lanes
and also a strip of width 2.5 m which is not intended for traffic. The additionally strip

16
width was measured on site while the width of the ordinary traffic lanes were
approximated, see Figure 3.14.
Figure 3.14: Width of each traffic lane, James & Kärrfelt, (2009)
In total 21 different measurements were performed and the difference between them
are the location, which lane the truck drove in, and the speed of the truck
Three measurements were chosen, one for each traffic lane with an additional
measurement for the lane not intended for traffic (outer lane), to use when verifying
the FE-model. This is further presented in chapter 5.

17
4 Pre-modelling phase
4.1 Pre-modelling phase
In order to be able to model the Vårby Bridge in a proper way and to obtain reliable
results, a test were performed in the software Abaqus to evaluate which type of
element that should be used when modelling the entire bridge. For this, a number of
simple small-scale models with two I-girders and a concrete deck were created. In
addition to good results, computational speed and simplicity when constructing the
models were taken into consideration. One additional aim with the small-scale models
was to find a suitable way to model the composite action between the concrete deck
and the I-girders.
For the deck 3-D deformable shell elements were chosen for all models as beam
elements are only a line assigned a certain cross-section in Abaqus and is therefore
not an option for modelling the deck in the bridge due to difficulties when applying
loads on the deck. Solid elements were rejected for both the I-girders and the deck due
to the desire of keeping the model small with regard to the computational speed
The two I-girders were decided to be modelled in four different ways with a mixture
of 3-D deformable shell and beam elements inspired by Chung & Sotelino (2006) who
used these types in their report. Element type B32 is used for the beam elements and
QUAD S8R for the shell elements including the deck. The four models are presented
in Figure 4.1
(a) (b)
(c) (d)
Figure 4.1: The 4 small scale test models (a) shell elements, (b) flanges as beam
elements and webs as shell elements, (c) flanges as shell elements and
webs as beam elements, (d) beam elements

18
To get interaction between the upper flanges and the deck in the models a “tie
constraint” was used. Using this type of constraint the flanges are, whether they are
made of beam or shell elements, fully bonded to the adjacent area in the deck
corresponding to the width of the flanges.
During the pre-modelling phase, problems occurred considering how to construct the
final bridge model. Since the I-girders in the Vårby Bridge have varying height and
flange thickness this should be modelled as well in order to obtain the correct stiffness
in the bridge model. In Abaqus it is not possible to construct a beam element with
varying cross section, which makes it inappropriate to use. The only way to use beam
elements with varying cross section is to divide the element in satisfyingly small parts
and assign each part the correct cross section geometry. This is very time consuming
and by means of this; both model 3 and 4 were excluded as possible modelling
techniques in the bridge model. Model 2 with the webs as shell elements and the
flanges as beam elements is possible to use by creating “stringers” at the
corresponding edges on the webs. These can be chosen as beam elements and
assigned different cross-sections, such as plates. However, there is a need of using
offset for the flanges (further explained below) and that is only possible for the
stringers using “trapezoidal section” which makes it inconvenient to use for the bridge
model. Due to the shortcomings of model 2-4, the fact that the analysis in Abaqus are
linear static and that model 1 are easy to construct, model 1 was chosen for the bridge
model. Modelling with shell elements also gives a good overview of the geometry
when creating the parts and makes it possible to partition the shells in order to attach
loads or different thicknesses and it is easy to collect results in the right nodes which
are important features in the global analysis.
The shell elements for the flanges were all placed directly on the edge on the web
which is not the correct point of gravity centre. To solve this offsets were made to
move the flanges (though not physically in Abaqus) to get the true stiffness of the
cross section, Figure 4.2.
(a) (b) (c)
Figure 4.2: Cross-section of an I-girder with; a) infinite thin shell elements, b)

elements assigned a thickness, the red marks show the decreased lever
arm of the flanges leading to reduced stiffness for the cross-section, c)
flanges attached by using offset giving the correct stiffness
The deck was placed in its gravity centre leading to a gap between the top flanges and
the deck in the model which can be seen in Figure 4.3.

19
Figure 4.3: The deck distanced by half its thickness plus the thickness of the top
flange away from the top flange
The girders were modelled as simply supported and the boundary conditions were
placed at the end of each bottom flange in x-direction. To model a simple supported
boundary condition one end was free to move in x-direction and rotate around z-axis,
while the other end only was allowed to rotate around z-axis. The load used to test the
model was of the magnitude of 0.05 MPa and uniformly distributed on the deck in the
area over the girders. The mesh varied in size for different parts of the bridge as a
consequence of using tie constraints. When using tie constraints in Abaqus the
software requests the slave surface to have a finer mesh than the master surface in
order to obtain better results and to shorten the run time for the analysis.
To ensure that model 1 function as expected a comparison between deflections
obtained from Abaqus and hand calculated ditto were made, see Appendix B. Also the
general looks of the loaded model was checked to see that the deformation was
reasonable and that the tie constraint between the top flanges and the deck worked
satisfyingly.

20
5 Global FE-analysis
5.1 Introduction
The Vårby Bridge is modelled with 3-D deformable shell elements following the
outcomes in the previous chapter. The composite bridge model is build up from 4
different parts for the steel details and one part representing the concrete deck. The
analyses performed of the global bridge model are linear static and verified by the on-
site strain measurements presented in chapter 3.4.
The materials used in the bridge model are linear elastic with isotropic behaviour,
displayed in Table 5.1 below.
Table 5.1: Linear elastic materials used in the bridge model
Material Density Young’s modulus of Poison´s Ratio

[kg/m3] elasticity[GPa]
Steel 7850 210 0.3
Concrete 2300 50 0.25
5.2 Steel details

The steel parts were modelled by first sketching the web plate in one of the main
girders with regard to the coordinates for the vertical web stiffener, Figure 5.1. These
coordinates was used in the following step to extrude the vertical web stiffener and
the stiffeners at the support, Figure 5.2.
Figure 5.1: The web plate in one of the main girders in sketch mode

21
The top and bottom flanges were attached to the web plate using offset; the girder was
then laterally transposed in order to create the adjacent main girder.
Figure 5.2: The web plate with vertical web stiffeners and stiffeners at support
Cross beams were inserted between the main girders to finalise the part. The
described procedure was used to create all the parts for the steel details. Steel parts 1-
4 are presented in Figure 5.3.
(a) (b)
(c) (d)
Figure 5.3: The steel details in the bridge model: (a) steel 1 (b) steel 2 (c) steel 3
(d) steel 4
Varied compositions of these parts were then assembled together to create the
different spans which can be seen in Figure 5.4. The spans assembled together have a
total length of 252.6 m.

22
Part Steel 1 Steel 2 Steel 3 Steel 4
Colour Green Red White Blue
Figure 5.4: The steel parts assembled together in varied composition

denoted by different colours for each part
5.3 The deck

The deck is divided into a number of different strips along the bridge in order to
simulate the different thicknesses of the concrete deck, Figure 5.5. The composition of
the strips is symmetrically arranged around the centre line in longitudinal direction in
the deck. The outer most strip is representing the edge beam of the concrete deck and
the fourth strip, counting from every long side, have the same width as the top flanges
in the main girders due to the connection between the deck and the top flanges.
Figure 5.5: The figure shows 1/5 of the bridge deck. The different strips represent
varying thicknesses in the deck
Because of the thickness of the deck, (fourth strip), and the thickness of the top
flanges, the deck is distanced by 195 mm in order to receive the correct centre of
gravity for the co-operating parts. The dimension of the deck is 252.6 m x 14.9 m

23
5.4 Loads
To be able to verify the global behaviour of the bridge model, a number of load steps
were created in order to simulate the moving truck load from the on-site
measurements. The loads from the tires, wheel track distance 2010 mm, were
represented by 2 point loads for the front tires and 4 point loads for the bogey. The
load was applied for each traffic lane in the beginning of the bridge and moved
towards the mid-span. The distances between each load step along the bridge varies
between 1.0 m and 2.126 m. The shorter distance were used in the interesting area
were the largest number of strain gauges (diaphragm F23) is located. Partition lines
were sketch with the chosen load step lengths along the bridge. These lines containing
64 nodes (vertices), one for each wheel track.
The vertices were imported into a script-file who distributes the point loads between
the vertices for every load step. The bridge models were loaded at a total distance of
139.2 m consisting of 68 load steps. 4 additional steps were created in the beginning
and in the end due to that not all of the truck is on the bridge deck at these steps. If a
point load is located between two vertices, the script-file distributes the load between
these two which can be seen in Figure 5.6 where 4 point loads in the bogey are spread
out to 8.
Figure 5.6: Distributed load on the bridge deck representing the moving truck in
the on-site measurements
5.5 Boundary conditions

In the Vårby Bridge there are 7 supports which for the two main girders give 14
boundary conditions which can be seen in Figure 5.7. The bridge is free to move in
the bridge axis (x-axis) for both main girder C and D, but in transversal direction (z-
axis) only for main girder C. This applies for all supports, except in the mid supports
where the bearings are fixed for main girder D and partially fixed for main girder C.

24
Figure 5.7: Boundary conditions in the bridge model; y-axis represent the vertical
direction
The 4 groups of different boundary conditions used in the bridge model are presented
in Table 5.2.
Table 5.2: Table over the different boundary conditions for main girder C and D
-x -z -y Rotations
C Free Free Fixed Free Free Free
C;mid support Fixed Free Fixed Free Free Free
D Free Fixed Fixed Free Free Free
D; mid support Fixed Fixed Fixed Free Free Free
The boundary conditions are attached to one node located directly under the web in
the main girders and in line with the vertical web stiffeners. To represent the bearings
in a reasonable way the horizontal plate under the vertical support stiffeners in the
bottom flange are free to rotate around the node using a Multi point constraint (“MPC
Link”), Figure 5.8.

25
Figure 5.8: Multi point constraint representing the bearings in the supports
5.6 Mesh evaluation

To see that the results are reasonable and capture the expected results in a correct
way, a mesh evaluation was performed using a quadratic 8-node doubly curved thick
shell element with reduced integration (S8R). The quad-shaped element is used with
free technique, medial axis with minimizing the mesh transitions when appropriate.
The mesh is seeded by using global seeds for the entire part. To study the differences
in the results for different element sizes a range of 200 to 1750 mm was chosen for the
steel part in the composite span.
Other parameters that the study investigates are computer usage to run a certain job,
smaller elements increase the computer run time, and the difficulty level (time) to
achieve a certain mesh. It is preferable if the mesh could be generated in an easy and
fast way but still give good results. Table 5.3 summaries the chosen element sizes
with related number of elements and run time for a certain job
The study is performed at one of six free spans in the Vårby Bridge. The part of the
bridge (in the following denoted as the span) are simply supported with a total length
of 45045 mm and with a distance between the supports of 44000 mm. The span is
loaded in the middle by a concentrated point load placed at three positions; at the
console, over one main girder and at mid strip in transversal direction. The chosen
load locations are selected so that different actions of the span can be captured due to
the three traffic lanes on the bridge. Two positions are checked with regard to stresses
and one of them for deflection; see Appendix C.

26
Table 5.3: Chosen element sizes in the mesh evaluation with number of elements
in vertical direction, number of elements in the span and run time in
seconds
Element size 1750 1500 1250 1000 750 600
Web, y-axis 3
Nr of elements 1127 1272 1417 1584 2006 2366
CPU [s] 24 26 28
Element Size 500 400 300 250 220 200
Web, y-axis 4 5 6 7 8 9
Nr of elements 3115 3999 5580 6905 8442 10361
CPU [s] 32 34 40 49 52 63
The largest element size, 1750 mm, was found to be the maximum value in order to
keep the quadratic form. The mesh with the highest number of elements contains 9
elements in y-direction; smaller elements than this are not reasonable to use in the
global analysis phase. Figure 5.9 show the span without the concrete deck meshed
with element size 500 mm. The element size in the concrete deck (size 1000 mm;
deck dimension 45045x14900 mm) remains constant for the different element sizes in
the steel part.
Figure 5.9: The span with 4 elements in the vertical direction, quad-shaped mesh
generated by global seeds for the whole part
The evaluation of the different element sizes show that for deflection either one of the
different sizes can be used but for the stresses the size must be smaller in order to
receive good results, down to 200-300 mm.
The difference in CPU time for the different analysis is negligible and its only takes a
few seconds to generate the mesh. Due to this, an element size of 250 mm is a good

27
choice; especially when the smallest parts (vertical stiffeners) in the bridge model are
of 250 mm width, which will enhance the ability for the software to produce a good
quality mesh.
5.7 Approximations
5.7.1 Stiffness of the cross section
When modelling with shell elements it is of significant importance that the elements
are connected to each other in their respective (end) nodes in order to obtain the
correct behaviour. In the Vårby Bridge, the main girders are cambered and the sub
parts (web and flanges) vary in thickness along the bridge which also affects the
height of the web. The web plates have certain geometry so that the flanges with
different thicknesses can be welded to the web plate creating a straight line at the
outer surface of the flanges, see Figure 5.10.
Figure 5.10: Schematic figure of the web plate before the flanges with different
thicknesses are attached in Vårby Bridge
To get a model with the correct behaviour .i.e. the flanges are connected in their
respective end nodes, the web plate was sketched with a straight top and bottom edge,
Figure 5.11. The height of the web near the supports (left end in figure 5.11) was
increased in order to adjust for that the top flange in this area is 2 times thicker. The
web plate are sketched using the height of the vertical web stiffeners as reference, so
this increased height at the end of the web plate will only affect the first section
between the end and the first vertical web stiffener. The thicker part is given the
correct stiffness by using an offset ratio of 0.024, i.e. the centre of gravity is moved
slightly downwards. By means of this, the top surface is a straight line after the
flanges are attached to the web plate and the shell elements are attached in the same
node. The change in stiffness due to that the cut out areas for the bottom flange are
represented by a straight line are neglected.

28
Figure 5.11: Schematic figure over how the flanges are attached to web plate using
altered offsets. The figure shows the part “Steel 3” containing two
vertical web stiffeners
The changes in second moment of inertia for the main girder due to these small
geometrical changes are negligible and will not influence the global behaviour.
5.7.2 Bridge dimensions

Each main girder in Vårby Bridge is erected from 12 structural members, denoted in
Appendix A, 1-12 followed by C or D depending on which main girder it belongs to.
Main girder C and D for every number (1-12) have the same shape and geometry
except for some of the vertical web stiffeners, were the height differs 1-2 mm. From
these 12 structural members (including C and D) 4 main types can be seen. The 4
types have the same geometrical shape but with a slight difference in the dimensions,
see Table 5.4.
Table 5.4: Summarising table over the diverging dimensions for top and bottom
flanges in the main girders in Vårby Bridge, the members with bold
text are used in the bridge model
Structural 1CD 2CD 3CD 4CD 5CD 6CD

member
bf 28 47 39 39 42 42
47 50 50 45 45 47
tf 20 20 20 20 20 20
47 47 41 41 42
[mm] 7CD 8CD 9CD 10CD 11CD 12CD
bf 42 42 38 38 46 28
47 47 47 50 50 46
tf 20 20 20 20 20 20
42 42 42 47 41

29
In the bridge model 4 different parts are used (denoted with bold text in Table 5.4 ),
each part consisting of two main girders with the same geometry. Steel 1 and 3 are
created using the geometry for the span between support 5 and 6 (drawing 8CD and
7CD). Steel 2 are created using the same geometry as in steel 1 but with 3 diaphragms
instead of 4. Steel 4 are created from drawing 12CD and also cover 1CD. Steel 1
includes 4CD, 6CD, 8CD and 10 CD. Steel 3 includes 3CD, 5CD, 7CD, 9CD and
11CD
The geometrical differences within the structural members in the main girders in
Vårby Bridge are negligible and the approximation to use steel 1-4 when constructing
the bridge model does not affect the global behaviour.
5.7.3 The connection between diaphragm and main girders

In the Vårby Bridge the diaphragm beams are attached to the vertical web stiffener
with bolts, see figure 3.2 and 3.4 in chapter 3.2, and the diaphragm beams have cut
out areas which can seen in Appendix A; figure 9.5:C414 and in figure 9.10. In
Abaqus, the edge of the web in the end of the diaphragm beam is attached to the edge
of the vertical web stiffener of the main girders.
This approximation makes this connection stiffer due to that the diaphragm beam and
the vertical web stiffener are merged together. But due to the large number of heavy
bolts in this connection and that only the web is connected to the vertical web stiffener
in Vårby Bridge, this approximation in the bridge model is acceptable.
5.7.4 Global location of the main girders

Due to the road building of the highway that passes Vårby River, the Vårby Bridge is
shaped in a certain way. The top surface in the bridge is also somewhat higher in the
spans to compensate for the self weight. In Abaqus the bridge is modelled with the
two main girders starting in the same level, separated by the designated transversal
distance. The top surfaces in the main girders are modelled as a horizontal line.
These approximations do not have any influence of the static behaviour of the bridge
model.
5.7.5 Connections between the concrete deck and the top flanges
In Vårby Bridge, studs are used in the connection between concrete deck and the top
flanges. In Abaqus the connection is modelled with a “tie constraint” meaning that the
adjacent areas are bonded with full interaction in all degrees of freedom. The tied
connection is overestimating the interaction; it will not take any slip between the
adjacent surfaces into account, a possible elongation of the studs due to the uplifting
force is neglected and the deck are prevented to lift away from the top flange in the
main girders.
The tied connection are however acceptable in the global analysis when the structural
behaviour of the bridge is studied.

30
5.7.6 End-diaphragms
In Vårby Bridge both of the end-diaphragms, S1 and S7, are adjusted due to the
connection between the bridge and firm ground which can be seen in Appendix A;
main girder 1 and 12cd. In Abaqus, the end diaphragms are modelled without the cut
out area and with the transversal beam placed at the same distance from the top as in
the other diaphragms in the bridge model (400 mm).
This was done do make it easier to create the parts for the end sections and the change
in geometry will not influence the global behaviour of the bridge model.
5.7.7 Concrete deck

The thickness of the concrete deck in Vårby Bridge varies along the width; see figure
3.5, chapter 3.2, and the end beams do not have the same geometry. The thickness is
approximated in Abaqus with an appropriate number of strips along the deck where
the height of each strip is an average of the maximum and minimum value in the strip.
The end beams are modelled with an equal width and height and the dimensions are
approximated from the drawings.
The concrete deck is assumed to be un-cracked (State I) and therefore is a static
analysis performed in Abaqus. The deck is also modelled without any reinforcing bars
and with the same modulus of elasticity in the whole deck.
5.8 Verification of the global bridge model

The verification of the bridge model was made by comparing the stresses in the global
channels from the on-site strain measurements with the corresponding values obtained
in Abaqus. In addition to the comparisons, the deformed shape from the analysis for
each traffic lane were reviewed ocular to verify that the structural behaviour were as
expected. Figure 5.12 visualize the bridge model when loaded in right lane at a certain
step.
Figure 5.12: Moving load in right lane, passing diaphragm F23 which gives the
maximum stress in global channel 8
The global channels, Ch 4 and 8, where the strain measurements are compared to the
values obtained from analysis are located at diaphragm F23 on the lower surface of

31
the bottom flanges in the main girders. The global channels can be seen in figure 3.10,
chapter 3.4.1.
Measurement strain data were collected from right, left, mid- and the outer traffic lane
and these were plotted against the corresponding values from the analysis. Figure
5.13, Figure 5.14, Figure 5.15, Figure 5.16. The truck used during the measurement
was travelling when measuring the right- and outer lane at a speed of 30 km/h, in mid
lane at a speed of 50 km/h and in left lane at a speed of 90 km/h.
3
[MPa]
2
0 [m]
0 50 100 150
-1
-2
15
[MPa]
10
0 [m]
0 50 100 150
-5
Figure 5.13: Right lane Ch 4 and 8

32
8
6 [MPa]
4
2
0 [m]
-2 0 50 100 150
-4
8
6 [MPa]
4
2
0 [m]
-2 0 50 100 150
-4
Figure 5.14: Mid lane Ch 4 and 8
15
[MPa]
10 ]
5
0 [m]
0 50 100 150
-5
-10
1,5 [MPa]
1
0,5
0 [m]
0 50 100 150
-0,5
-1
-1,5
Figure 5.15: Left lane Ch 4 and 8

33
0,8 [MPa]
0,6
0,4
0,2
0 [m]
-0,2 0 50 100 150 200
-0,4
-0,6
-0,8
15
[MPa]
10
0 [m]
0 50 100 150 200
-5
-10
Figure 5.16: Outer lane Ch 4 and 8
Overall there is a good agreement between the on-site measurements and the values
obtain in analysis. The shapes of the curves match each other quite well with top and
bottom values on the same positions although with diverging values in some cases. As
can be seen in figure 3.14, chapter 3.4.2, the global channel 8 are located at the main
girder under the right traffic lane and global channel 4 under the left lane. However,
as can be seen in the figure, the traffic lanes are not symmetrically arranged on the
concrete deck which leads to that channel 4 are further away from left lane then
channel 8 are from right lane. Below comments about each traffic lane are presented.
• Right lane Channel 8 agrees well, the analysis values in channel 4 are higher
than the measurement values which means that more load have been
transferred to the opposite side in the bridge model compared to values from
on-site measurements.
• Mid lane Channel 8 agrees rather well, and as above, the analysis values are
higher than the measurement data for channel 4. The top and bottom values
from the measurement data are almost identical which could indicate that the
truck in the on-site measurement run very close to the centre line of the bridge,
while the analysis show higher values for channel 4 which is reasonable
because the mid lane are closer to channel 4.
• Left lane The measurement data from channel 8 is irregular; the reason could
be that the truck run in 90 km/h and creates thereby a dynamic effect. The two
curves follow however the same pattern but it is uncertain if that measurement
is reliable since it is small stresses. The analysis values for channel 4 are

34
higher than those from measurements who show almost the same values as
channel 8 when loaded in the right lane. This is surprisingly due to that the left
lane is placed partially outside the main girder, on the console, and more loads
should go to this strain gauge (channel 4) than if the load is placed inside the
main girders (which is the case for right lane).
• Outer lane The outer lane is located outside the right main girder and the
majority of the load goes to channel 8 which can be seen in the curves who
match very well. The value for channel 4 is very small and is fairly needed in
the comparison.
Adding together the comparisons for each traffic lane with the ocular judgment of the
deformed shape a good picture of the performance of the bridge model are
established. However, further research was performed by a parameter study to ensure
the correctness of the bridge model. (The main reason to the parameter study was due
to the fact that channel 4 consequently showed higher values from analysis compared
to these from the measurements).
Due to the observations mentioned above, it is possible that the given geometry
received for the concrete deck in Vårby Bridge (width of each traffic lane and location
on deck) is approximated values. By reason of the suspicious that the given widths of
each traffic lane maybe not are correct, an analysis with reduced number of load steps
were perform in the right lane. 9 load steps were moved 100, 200- and 500 mm
towards the outer lane in transversal direction i.e. away from the centre line of the
bridge. Stresses in both Ch 4 and 8 were collected and compared to the strain
measurements from Vårby Bridge but also with load in original position (one bogey
wheel in location directly over right main girder and the other bogey wheel between
the main girders) Table 5.5.
Table 5.5: Sensitivity when the load is moved in 100, 200- and 500 mm in
transversal direction, right lane
Measurement Original position 100 [mm] 200 [mm] 500 [mm]
Ch 4 2,03 2,72 2,60 2,48 2,12

[MPa]
Ch 8 10,56 9,98 10,10 10,22 10,57

[MPa]
The table show that when the load are moved 500 mm in transversal direction, which
means that the resultant of the truck load are moved closer to the right main girder, the
values from Abaqus are in close proximity to the values from the strain
measurements.
The sensitivity of the load location was further investigated by changing the distance
between the wheel tracks in transversal direction. The original distance were
compared to a larger distance, see Appendix D. The difference between the two wheel
tracks distances was negligible.
The load where also moved to the mid span in transversal direction to check the
symmetry of the bridge model; the main girders have the same geometry and the deck
are laterally transposed around the centre line in longitudinal direction. The curves for

35
Ch 4 and 8 were plotted against each other and showed identical results; see Appendix
D.
Due to the approximation that the diaphragm beams are merged to the vertical web
stiffeners in the bridge model, an analysis when the stiffness of the diaphragm beams
is decreased was performed. The goal was to see what influence the diaphragm beams
have for the global behaviour of the bridge model. The comparison showed that there
were no difference for the stresses in Ch 4 and 8 when reducing the diaphragm beam
stiffness markedly, see Appendix D.
The influence of different young´s modulus for the deck was investigated by changing
it from 20 to 70 GPa, were 50 GPa was the original value. This was made for the right
traffic lane and the plot is visible in Appendix D. The plot showed that a more stiff
concrete deck transfer more load in the transversal and longitudinal direction of the
deck compared to a deck with lower stiffness.
During analysis of the results the tensile stresses in the concrete deck were checked.
The maximum value reached was 1.4 MPa and the concrete deck was kept in the
assumed state I.
Summing up the verification above, the global model established is representative for
the performance of Vårby Bridge in the field of interest for this master thesis.

36
6 Refined model
6.1 Introduction
With the global model established and verified the next step was to model the area of
interest in a better way since the mesh, application of loads and also shell elements
may be too coarse to capture the local effects. Refinement of the model could be done
in various ways, for instance the area of interest can be made as a sub model which
might consist of solid elements with fine mesh but still keeping it rather small
considering computer usage time. As input, outcomes from global analysis should be
applied on the sub model to get correct results. Another alternative is to use the global
model and try to make a fine mesh in the area of interest and also apply the loads in a
different way. This thesis used the later proposed alternative since the saving of time
was crucial, a sub model would have taken too long time to create.
The actions taken to refine the model were to make a finer mesh and changes how the
loads were applied. The element size of the mesh was decreased from 250 mm to 25
mm in the span where the measurements were performed. The load was applied
making areas of 0.2x0.6 m modelling the 6 tyres of the truck and using pressure load
instead of concentrated forces. 5 load steps were chosen to make it possible to capture
the highest stresses in the interesting area.
6.2 Comparison between refined model and measurement
The results from the refined model were then compared to the on-site measurement
data; one measurement from each lane was used. Data was taken from channels 1-3,
4-6 and 9-10 at the span section and channels 12-15 at the support section. Due to the
changes of the mesh the nodes corresponding to channel locations could be found in
the model, which was not the case before since the mesh then was too coarse. The
changes in application of the loads were also beneficial obtaining a better stress
distribution, though the difference was not large.
The difference was large between measurement data and results from Abaqus when
comparing, Table 6.1. All models, right, mid and outer lane, had the same behaviour,
but with difference in magnitude of stresses, which indicated that something was not
properly modelled.
Table 6.1: Results from comparison between Abaqus and measurements, right
lane. Only values from channels on one vertical stiffener is presented
since both sides experienced the same behaviour
Ch 5 [MPa] Ch 6 [MPa] Ch 7 [MPa]
Abaqus -26 -17 -12,5
Measurement -9,0 -3,0 -1,5
Several things influenced the results and the ones considered were:
• The stiffness of the diaphragms and how they were connected to the vertical
stiffeners
• How the stiffeners were represented with regard to the stiffness

37
• The possibility that it was a local problem at the top corner of the stiffener on
the diaphragm side where it could be a large stress concentration due to the top
flange bending over the corner
• A poor modelling of the composite action between the I-girder and the
concrete deck
The diaphragms were ruled out due to the comparison with channel 9 and 10 showed
good agreement with the result from measurements compared to Abaqus in all
models, Figure 6.1. When looking at the stresses in the stiffeners and especially the
principal stresses the distribution was as expected with only one principal stress near
the edge of the diaphragm side and no shear stresses. The possibility that it was a local
problem causing too high stresses further down in the stiffener was rejected, even
though a stress concentration of large magnitude was found. The area of the high
stresses was though indeed local at the corner and faded out before reaching the nodes
of interest. The problem with the model, which had been of great concern since
starting the modelling in the thesis, was the modelling of the composite action.
Globally it worked very well, which was seen in previously chapter, but when looking
locally the forces in the studs were of great significance.
5 6
4 5
4
3
3
2 2
1 1
0 0
-1 0 50 100 150 200 -1 0 50 100 150 200
Figure 6.1: Influence lines for channel 9 and 10, mid lane
Since the studs were placed in pairs at the bridge one of the studs were loaded in
tension and the other in compression creating a force couple. This force couple had
large influence on the stresses in the stiffener. The composite action was modelled as
previously said with tie constraints between the entire width of the top flange and an
equal width at the deck. This worked incorrectly and probably the cause for
unsatisfying results in the local comparison.
To solve the problems two approaches was formed; approach 1 was to change the
model not using tie constraints and try to make a better representation of the
composite action. The other approach (2) was to keep the model but move the
application of the loads in order to make the use of tie constraints work.
6.3 Approach 1
To model the studs more appropriate the suggestion was to use axial springs, which
was in-built parts in Abaqus. The idea to use springs was to model the studs
experienced in tension and that they should not influence the structure when
compressed. When compressed they should not make an influence since the concrete
deck would put pressure on the I-girder below using contact surfaces. These new

38
properties of the model were done in the span where the comparisons were made.
When trying this it did not work as planned, the springs did not compress and
overruled the contact surfaces of the top flange and the concrete deck, causing strange
results. The contact surface was made to function such that a gap could appear
between the concrete deck and the main girder, this would appear on the side where
the studs were in tension, but then closing when the studs where released from
tension.
The next step made was to keep the contact surfaces but take away the row of studs
that were compressed keeping only the studs in tension to take away the errors with
compressed springs. Still no good results were obtained and one other possibility to
solve this was found. All the above tests in approach 1 were with the truck modelled
in the right lane which could cause difficulties in interpretation how the distribution
would be. In other words, were the studs really in tension? To ensure this the truck
was modelled driving in the mid lane, which would mean that the studs were in
tension.
The result from the last model was none; it ended up with convergence problem and
no solution. Probable causes for this could be that the springs and contact surfaces
were placed in a too small area of the bridge model and had to be increased. The area
was increased and the model tested in several steps until the area was the entire
bridge. Despite all the efforts put in still no convergence nor solution and the
approach to change the model was abandoned.
6.4 Approach 2
The second approach was to use the model with refined mesh and tie constraint. To
get different results compared to earlier models the position where the load was
applied were changed. As shown previously in the report the use of tie constraints
worked well for global studies of the bridge but not locally, since the composite action
comes into play. The main problem suffered with the first approach was how to model
the force couple created when the concrete slab rotated, i.e. the composite action.
Knowing that the tie constraint worked well under the influence from pressure, the
load was moved to a position more to the right of the original position when
modelling a truck in the right-lane. This was done since earlier, from the global
analysis; it was known that the results were sensitive when moving the application of
load transversally on the bridge. With regard to that the aim was to get the entire right
upper flange in compression and obtain better results without needing to model the
composite action in two directions.
The load was moved from the original position to the right with 500 and 650 mm
showing the results in Figure 6.2.

39
5
-5 0 50 100 150
-10
-15
-20
-25
-30
Figure 6.2: Sensitivity of stresses channel 5 when moving load tranversally
Channel 1 experienced the same behaviour as channel 5 and the results confirmed, as
for the global analysis, the sensitivity when moving the load transversally. Also, the
results presented above came quite close to the measured values. When looking on
results from the other channels, on the vertical stiffeners, they experienced the same
behaviour, coming closer to the measured values. However, the decrease in stresses
was not as large as for the highest placed channels, meaning that the stress distribution
in the vertical stiffeners still was not properly modelled.
The results from the support were similar in behaviour as the span models, but the
changes in magnitude of stresses were not as large. In Table 6.2 the difference in
sensitivity between span and support is presented.
Table 6.2: Difference of sensitivity when moving load in transversal direction,

right lane. Comparison between the corresponding values in support
and span
-300 [mm] 0 500 [mm] 650 [mm]
Ch 5, span [MPa] -33,0 -26,0 -13,3 -8,3
Ch 12, support [MPa] -26,6 -22,2 -13,5 -10,

40
7 Conclusions
The questions to answer in the thesis were; why do the fatigue cracks occur and why
only in span? Also how the cracks should be retrofitted were a task stated.
The results from the refined modelling phase were not as good as hoped for. The
modelling of the composite action was not solved and the stress distribution obtained
was not good either. However, based on the results collected and the comparisons
made with measurements one theory have been formed on why the cracks occur.
When comparing the results from the models with the measurements there are not
good agreement when looking at the magnitude of the stresses. When looking at the
behaviour there agreement is very good, which means that the results can be of use in
comparisons between values from the models since they have the same modelling
errors.
The first approach when drawing conclusions was to see how high stresses that the
bridge experienced from the measurements. Even though residual stresses exist there
is no sign of any stresses reaching the levels of stress range causing fatigue cracks.
Residual stresses must be a part of causing the fatigue cracks since the area where the
cracks are formed is loaded only in compression. This was confirmed when looking
on where the traffic lanes are located on the bridge.
Previously in the report a summary of the literature study was made. In the summary
four different theories that could cause fatigue cracks were proposed. Below figures
describing each theory and conclusions of what influence they have for inducing
fatigue cracking in Vårby Bridge are presented.
Figure 7.1: Hogging moment created whit vehicles driving in outermost lane (on
the outside of the main girders). The hogging moment then creates
high tensile stresses in the area where the fatigue cracks occur.
Hogging moment (Figure 7.1) can occur if there are vehicles on the outermost lanes,
but the stresses are not that large and the number of cycles is very small, since these
lanes are not intended for traffic.

41
Figure 7.2: Rotation of the concrete slab which occurs with vehicles driving
between the main girders which influence compressive stresses at the
vertical stiffener. In itself this cannot cause fatigue cracks, but together
with residual stresses in the
The rotation of the concrete slab (Figure 7.2) will occur but does not create
magnitudes of stresses which leads to formation of fatigue cracks and the same yields
for the last two theories presented (Figure 7.3), (Figure 7.4). To be noticed however is
that the highest stresses are found when placing the load directly above the area of
interest.
Figure 7.3: When vehicles drive directly above the vertical stiffener causing
compressive stresses together with the residual stresses fatigue cracks
may form
Figure 7.4: The bridge rotating around one main girder. This might happen as a
consequence if the main girder, with load applied directly above,
suffers from large deflection. This could influence a rotation around
the other main girder causing compressive stresses at the welded joint.

42
7.1 Theory of why cracks have occurred
The theory we present what might have caused the cracks were created when
comparing results from the models. When looking on the sensitivity when moving the
load transversally in the near of the main girder there was one difference found
between span and support.
The sensitivity is larger in the span compared to the support which might depend on
the double vertical stiffeners at the supports. This is not something that can be found
in the measurements from the bridge, but might still be valid. When doing the global
analysis we actually stated that the data we got of lane locations were wrong. When
comparing load location from our global and local analysis we found that the location
when the measurements and the results coincided the best was about the same for both
analyses. This leads to the conclusion that during the measurements the most severe
loading case can have been missed.
We think that the fatigue cracks are caused by vehicles driving directly over the area
of interest with one wheel and the other between the main girders. If any other
vehicles would be driving on the bridge at the same time this is beneficial for the
cracks forming, since the compressive stress will increase. With this theory presented
it can be stated that the theories dismissed above will also influence.
The main cause is the one mentioned above, but there is a mixture of effects leading
to the large number of fatigue cracks. All the other theories, except hogging moment,
will amplify the magnitude of stresses if these occur at the same time, for instance
when one vehicle is in the right lane and another vehicle in the mid lane causing
larger rotation of the concrete slab. Also the residual stresses play an integral part
since these make it possible for fatigue cracks to be formed in compression.
With this theory presented above the question of why fatigue cracks only occur in
span and not at support can be answered. This has not been confirmed in the report
since the model is not fully correct. The answer to the question is perhaps no answer
or that we cannot know. The support section might suffer from damageable stresses
but lower than the ones in the span which then demands more cycles for fatigue
cracks to be formed. The stresses might be low enough such that no damage will
occur. The important to point out is that cracks may occur at support also.
Since we do not have confirmed the theory fully, no investigation in retrofitting
methods has been performed.
7.2 Further studies
To confirm the theory presented more work needs to be done. First of all a complete
model must be formed, especially to model the composite action properly is of
essence. The actual stiffness of the concrete would also be of interest as input for the
models. Retrofitting methods need to be evaluated since, if the theory is correct, there
is a mixture of effects causing the fatigue cracks.
Since the aim of this thesis were to model the bridge and compare results with
measured data no regard of the measured data itself has been taken. Evaluation and
calculation of stress levels from the measured values should be performed.
A new measurement of the actual bridge would be very good if possible. The input
from the measurements to this thesis was not good not giving the correct measure
conditions, loading positions or position of lanes.

43
8 REFERENCES
Chung, W., & Sotelino, E. D. (2006). Three-dimensional finite element modeling of
composite girder bridges. Engineering Structures , 28.
Eriksson, K. (2009). Materialprov från vägbro i Vårby. Luleå: Luleå Tekniska
Universitet.
Greiner, R., & Taras, A. (2009). Fatigue failures at stiffener connections in highway
composite bridges. Graz, Austria: Graz University of Technology.
James, G., & Kärrfelt, U. (2009). Bro över Vårbyfjärden: Sprickor i avstyvningar.
Projektengagemang i Stockholm AB, Anläggningsunderhåll, Stockholm.
Okura, I., Shiozaki, T., Fukumoto, Y., & Nanjyo, A. (1995). Stud arrangement to
reduce fatigue cracking in vertical stiffeners.
Sakano, M., Kawakami, Y., Sakai, Y., & Matsushita, H. ( 2007). Retrofitting method
against fatigue cracking in welded joint between steel deck and vertical stiffeners.
Pacific Structural Steel Conference 2007, Steel Structures in Natural Hazards.
Wairakei, New Zealand.

44
Appendix A: Drawings of Vårby Bridge
Figure 8.1: Elevation of the two main girders
Figure 8.2: Plan of the two main girders
Diaphragms in Vårby Bridge

In Vårby Bridge there are three types of diaphragms in the spans, one type
representing the end diaphragms and one type for the intermediate supports. Figure
9.3, 9.6-9.8 and 9.12 show were in Vårby Bridge the diaphragms are located. The
height of each diaphragms (F1-F27, B2-B6 and S1, S7) can be seen in the drawings
showing the main girders.
Diaphragms in span (type 1,2 and 3)
The different types of diaphragms in the spans have the same geometry except for the
height that differs depending on where in the spans they are located.

Type one
Figure 8.3: Type 1, diaphragms F2 F7 F12 F17 F22 F27
Figure 8.4: Cross-beam 51: web 12x370x7420, flanges 14x350x7420
Figure 8.5: B414 Section of the bridge (type one) showing the two main girders
and the cross beam in span, C414 Plan view of the cross-beam with a
modified top flange in order to fit the vertical web stiffener in the main
girder, H414 Section of the cross-beam with the fillet welds

Type 2
Figure 8.6: Type 2, diaphragms F4 F5 F9 F10 F14 F15 F19 F20 F24 F25
Cross-beam 53, web 12x710x7420, flanges 14x350x7420

Geometry same as for cross beam 51
Type 3
Figure 8.7: Type 3, diaphragms F1 F3 F6 F8 F11 F13 F16 F18 F21 F23 F26 F28
Cross-beam 52, web 12x540x7420, flanges 14x350x7420

Geometry same as for cross beam 51
Diaphragm in end supports (S1, S7)
Figure 8.8: Diaphragms in end support S1 and S7
Figure 8.9: Cross-beam 66, web 12x700x7442, top flange 15x300x7424, bottom
flange 20x400x7424

Figure 8.10: D413, Plan view of the cross-beam with the cut out areas in the cross
beam
Figure 8.11: C413, Section of the bridge at end support S1 and S7; Vertical web
stiffener marked with (10),(13),(14)= 25x250

Diaphragms at intermediate supports (B2-B6)
Figure 8.12: Diaphragms B2, B3, B4, B5, B6 at intermediate supports
Figure 8.13: B415, Section of the bridge at intermediate support B2-B6; Vertical
web stiffener, 25x300

Figure 8.14: Cross beam 71, web 20x900x7420, top flange 25x500x7420, bottom
flange 25x600x7420
Figure 8.15: C415, Plan view over cross-beam 71 with different widths of the
flanges

Main girders
1c and 1d
Figure 8.16: The two main girders with dimension at each position. Difference
in starting position between the two main girders (193 mm)
Pos 9 and 12 F1 (F followed by a number in bold text represents the supports in span)
Pos 16 and 17 F2
End support S1 (S followed by a number in bold text represent the end supports)

2c and 2d
between the two main girders (77 mm)
Pos 36 and 42 F3
Pos 37 and 43 F4
Pos 38,39,44,45 B2 (B followed by a number in bold text represent the end
supports)

3c and 3d
in starting position between the two main girders (0 mm)
Pos 91 and 93 F5
Pos 92 and 94 F6

4c and 4d
Figure 8.19: The two main girders with dimension at each position
Pos 104 and 109 F7

Pos 105 and 110 F8
Pos 106 and 111 F9
Pos 107, 108, 112, 113 B3

5c and 5d
Pos 127 and 129 F10

Pos 128 and 130 F11

6c and 6d
Pos 139 and 144 F12

Pos 140 and 145 F13
Pos 141 and 146 F14
Pos 142, 143, 147, 148 B4

7c and 7d
Pos 162 and 164 F15

Pos 163 and 165 F16

8c and 8d
Pos 176 and 181 F17

Pos 177 and 182 F18
Pos 178 and 183 F19
Pos 179, 180, 184, 185 B5

9c and 9d
Pos 205 and 207 F20

Pos 206 and 208 F21

10c and 10d
Pos 230 and 235 F22

Pos 231 and 236 F23
Pos 232 and 237 F24
Pos 233, 234, 238, 239 B6

11c and 11d
Pos 260 and 262 F25

Pos 261 and 263 F26

12c and 12d
Pos 278 and 282 F27

Pos 279 and 283 F28
End support S7

Appendix B: Hand calculations of small-scale model
b 1 := 250 mm
z f1 := 10 mm
b 2 := 250 mm
z f2 := 690 mm
t 1 := 20 mm 700
z NA := mm
2
t 2 := 20 mm
zw := 350 mm
t 3 := 15 mm h w := 660 mm
b deck := 300 mm L := 4m
t deck := 250 mm
34
kN α 1 :=
q d := 1MPa⋅ b1 = 250⋅ P := 1kN E := 200GPa 210
m
t deck  t1 
b deck ⋅ t deck ⋅ ⋅ α1 + b 1⋅ t 1⋅  t deck +  ...
2  2
 hw   t2 
+ h w⋅ t 3⋅  t deck + t 1 +  + b 2⋅ t 2⋅  t 1 + t deck + h w + 
zna :=
 2   2
= 0.42m
b deck ⋅ tdeck ⋅ α1 + b 1⋅ t1 + h w⋅ t 3 + b 2⋅ t 2
3 2
b deck ⋅ t deck  t deck  −3 4
Iy := ⋅ α 1 + b deck ⋅ t deck ⋅ α1⋅  − zna  ... = 3.28 × 10 m
12  2 
3 2
b 1⋅ t 1  t1 
+ + b 1⋅ t 1⋅  t deck + − zna  ...
12  2 
3 2
t 3⋅ h w  hw 
+ + t 3⋅ h w⋅  t deck + t 1 + − zna  ...
12  2 
3 2
b 2⋅ t 2  t2 
+ + b 2⋅ t 2⋅  t deck + t 1 + h w + − zna 
12  2 
4
5⋅ q d ⋅ L −3 −3 δABAQUS
δhand := = 1.27 × 10 m δABAQUS := 1.215 ⋅ 10 m
384⋅ E⋅ Iy ⋅ 100 = 95.658
δhand

Appendix C: Mesh evaluation
Table 8.1: Element sizes in the mesh evaluation; The table show deviation from
the size 200 mm in percent
[%]
Check Size mm 1750 1500 1250 1000 750 600
point
1 S1.1 Load 1 0.85751 0.58014 0.5874 0.33908 -0.1975 -0.7101

Load 2 -0.0684 -0.3612 -0.3488 -0.6033 -1.1636 -1.6659
Load 3 -2.3271 -2.6449 -2.6087 -2.8768 -3.4602 -3.9219
2 S1.1 Load 1 2.11443 2.2811 2.42607 2.56365 2.90241 3.13417

Load 2 9.09014 9.10561 9.08085 9.05146 8.85809 8.70958
Load 3 21.8394 21.4912 21.1081 20.7509 19.5605 18.6854
1 U2 Load 1 0.10245 0.10245 0.09944 0.10305 0.10607 0.10667

Load 2 -0.9053 -0.9111 -0.9166 -0.9168 -0.9216 -0.9123
Load 3 -2.8064 -2.8255 -2.8358 -2.8468 -2.8696 -2.8445
500 400 300 250 220 200
1 S1.1 Load 1 0.35796 -0.5424 0.57869 -0.077 -0.3805 _

Load 2 -0.4049 -1.0513 0.13611 -0.4244 -0.6446 _
Load 3 -2.1913 -2.2384 -0.8936 -1.2295 -1.2511 _
2 S1.1 Load 1 2.87825 3.09225 1.39647 1.80525 1.62675 _

Load 2 7.03418 6.75959 2.55486 2.3816 1.61661 _
Load 3 14.6342 13.4664 4.71016 3.5144 1.69129 _
1 U2 Load 1 0.07774 0.04882 0.03556 0.02531 0.01205 _

Load 2 -0.7802 -0.4538 -0.3964 -0.2883 -0.168 _
Load 3 -2.4105 -1.4067 -1.2169 -0.8837 -0.5093 _

Figure 8.28: The black dot representing checkpoint 1 and the red checkpoint 2
Figure 8.29: Applied load; Load 1 in centre, load 2 over top flange and load 3 at
console

Appendix D: Additional diagrams from verification
12
10
8
6
4
2
0
-2 0 50 100 150
-4
-6
3,5
3
2,5
2
1,5
1
0,5
0
-0,5 0 50 100 150
-1
-1,5
-2
Figure 8.30: Different concrete stiffness, Ch 8 and 4, right lane

Left lane, wheel tracks Ch 8
2
1,5
0,5
MPa
2010 mm
0 2460 mm
0 20 40 60 80
-0,5
-1
-1,5
Number of steps
Figure 8.31: Different wheel tracks distances of the moving load
Load in mid span, transversal

8
4
MPa
2 Ch 8
Ch 4
0
0 10 20 30 40 50 60 70
-2
-4
steg
Figure 8.32: Load in mid span, symmetry check of the bridge model

Right lane Ch 4
3
2,5
1,5
Abaqus normal
1
mpa
0,5
0 Markedly lower value for

the diaphragm beam
-0,5 0 50 100 150 stiffness
-1
-1,5
-2
m
Figure 8.33: Lower stiffness of the diaphragm beam compared the normal value


FE-analysis of Vårby Bridge: Robert Bengtsson Mikael Widén

Uploaded by

FE-analysis of Vårby Bridge: Robert Bengtsson Mikael Widén

Uploaded by

FE-analysis of Vårby Bridge

Investigation of fatigue damage in a composite bridge

Department of Civil and Environmental Engineering