Assessment of A Railway Concrete Arch Bridge by Numerical Modelling and Measurements
Assessment of A Railway Concrete Arch Bridge by Numerical Modelling and Measurements
Assessment of A Railway Concrete Arch Bridge by Numerical Modelling and Measurements
G. He & Z. Zou
Department of Civil, Environmental and Mining Engineering, Luleå University of Technology, Luleå,
Sweden & Civil Engineering, Architecture College, Central South Forestry University, Changsha,
Hunan, P. R. China
O. Enochsson, A. Bennitz & L. Elfgren
Department of Civil, Environmental and Mining Engineering, Luleå University of Technology, Luleå,
Sweden
A. Kronborg, B. Töyrä & B. Paulsson
Banverket, Borlänge & Luleå, Sweden
ABSTRACT: Two main railway concrete arch bridges have been investigated in northern Swe-
den: the Vindel Bridge over the Vindel River with a span of 110 m and the Långforsen Bridge
over the Kalix River with a span of 90 m. The bridges were constructed in the 1950s with simi-
lar designs. In order to investigate the possibility to increase the axle load from 225 kN to 250
kN, two 3-D Finite Element models with shell elements were developed and two field tests were
carried out with passing trains and ambient vibrations. The measurement frequencies and mode
shapes were studied with the stochastic subspace identification method in the time domain.
Based on a comparison of the predictions of the FE analyses and the results of the field tests, the
FE models were updated. It was found that it was possible to make good predictions of the ac-
tual dynamic response with reasonable assumptions of material properties and joint stiffness in
the structure.
1 INTRODUCTION
Because of increases in axle loads and speeds, tougher environments and deteriorating materi-
als, many existing bridges need to be assessed regarding their remaining service life. Conse-
quently, many bridge engineers and researchers in the world are now faced with the challenge to
develop innovative solutions to enhance the performance level or extend the service life of ex-
isting bridges.
In Sweden, the history of railroads started some 150 years ago. Many bridges have been in
service more than 50 years, some even 100 years.
Two railway concrete arch bridges are discussed in this paper. The owner, Banverket, wanted
to increase the maximum allowed axle load on them from 225 kN to 250 kN. The Vindel
Bridge crosses the Vindel River near Vindeln some 60 km NW of Umeå. It is a concrete arch
bridge with a span of 110 m and a height of 22 m, which was constructed in 1952, see Figure 1.
The box rib arch carries a concrete deck with columns, see Figure 2. The section of the arch is a
box with two cells. The width and height of the section varies with the distance from the abut-
ment to middle span. Figure 3 shows a typical section of the box arch and the deck. The Lång-
forsen Bridge, built in 1960, crosses the Kalix River at Långforsen. It is a similar concrete arch
bridge as Vindel Bridge, with a span of 90 m and a height of 13.7 m, see Figure 4.
Large transversal movement was reported from visual site inspections on the Vindel Bridge.
Arch bridges are complex structures whose structural form, which includes the shape and profile
of the arch rib and sup-structures, results in significant three-dimensional structural responses.
However, generally for assessment purposes conventional modelling methods use simplified
two-dimensional beam element models. Here instead, mostly 3D shell elements will be used.
When a train passes over a bridge, dynamic effects will be induced in the bridge structure,
especially for high speeds and heavy axle loads. Difficulties in the assessment of existing
bridges arise from two primary reasons, the lack of knowledge about the conditions of the
bridges, and the lack of accepted procedures for the analysis of existing bridges.
Figure 1. Bridge over the Vindel River in northern Sweden.
← South G F E D C B A North →
I H
4 span @ 14.35 = 57.4 m Free span width = 110.022 m 4 span @ 14.35 = 57.4 m
Teoretical span width = 112.070 m
Figure 2. Elevation of the Vindel Bridge with location points of the sensors A-I.
6000
6900
H-I
A-G
2430
1900
347 347
West East
Figure 3. Vindel Bridge Cross Sections. Deck (left) and typical hollow box section of the arch (right),
with location points for sensors A to I.
Field tests were planned so that the bridge’s dynamic behaviour, i.e., frequencies and mode sha-
pes, could be investigated under ambient vibration and passing through train loading. Specific
purposes were to investigate actual dynamic responses, i.e., displacement and acceleration under
test train loading and different axle load distribution. Furthermore, the results of the field test
were used to validate the finite element modelling.
Figure 5 Instrumentation used in the two tests. Photo Anders Bennitz (2005).
3 NUMERICAL MODELLING
Railway arch bridges are complex structures involving interaction between the material, the su-
perstructures, ballast, sleeper, rail, and railway equipment. Furthermore, the structural form of
the bridge, which includes the rib, sup-structure and side spans, results in significant three-
dimensional structural responses. The contribution of these three-dimensional effects to the re-
sponse of railway arch bridges has been noted, but is rarely discussed as important element of
bridge assessment. Three-dimensional finite element models allow the analyst to account for the
extent and geometry of each of these constituents in constructing a representative numerical
model of arch bridge.
Y Y
X X
Z Z
In the beam models, arch ribs, columns and deck are modelled with 3-D beam elements. In the
shell models, the arch rib, deck and end columns are modelled with 3-D shell elements. How-
ever, the other columns are still modelled with beam elements. All of the bearings are modelled
by joint elements with different DOF (degree of freedom) and spring stiffness in both beam and
shell models. The ballast and sleeper are modelled as masses without stiffness. The connection
between the beams of the deck and the columns are modelled with 3-D beam elements with re-
leased end rotations. The ballast and sleeper are modelled as masses without stiffness. The con-
crete has a design compressive strength of 50 MPa in the arch and the deck, and 40 MPa in the
columns. The modulus of elasticity has been assumed to have values of 35 and 33 GPa respec-
tively. Two additional FE models have been developed to study the effect of the super structure
and the secondary dead load, such as ballast and railway equipment, on the dynamic behaviour,
He et al (2006). In total, 1011 nodes and 509 elements are included in the beam model. In the
shell model, 8624 nodes and 8514 elements are included.
Table 2. Ten first frequencies and mode shapes for beam and shell models
________________________________________________________________________________________
No. ____________________________________________________________________________________
Shell model Beam model
Freq.[Hz] Mode* Freq.[Hz] Mode*
________________________________________________________________________________________
1 1.13 T-1 1.18 T-1
2 1.93 T-2 2.47 T-2
3 2.23 T-3 2.52 V-1
4 2.46 V-1 3.23 T-3
5 3.25 T-4 3.59 V-2
6 3.46 V-2 4.38 T-4
7 4.19 T-5 4.72 V-3
8 4.71 V-3 5.91 V-4
9 5.26 T-6 6.39 T-5
10 5.64 V-4 6.49 T-6
________________________________________________________________________________________
* T: transverse global vibration.
V: vertical global vibration
From Table 2, it can be seen that the beam model is stiffer than the shell model. The reason may
be that the columns, the deck and beam are modelled in a different way in the beam model and
the shell model. The shell model is used to study the dynamic behaviour and to simulate the dy-
namic response of the bridge when a freight carriage passes through. The modelling of connec-
tions of components, especially the bearings, is important for the numerical model of a bridge
structure. Based on the above beam and shell models, the effects of different type of bearings
and the connection between the deck beam and the column on the dynamic behaviour have been
investigated, for more details see He et al (2006). The connection between the deck beam and
the shorter columns are modelled with a joint for which the rotation degree about the z-axis has
been released.
In order to calibrate the calculated results with measurement in September and December, the
joint element is used to simulate the bearings with movable and fixed joint at the two ends of the
deck beam in the FE models. The movable joint bearings are used in the shell model corre-
sponding to field test in September as S-Sep, whereas, the fixed joint bearings are used in the
shell model corresponding to field test in December as S-Dec. The calculated first 10 natural
frequencies of vibration mode with S-Sep and S-Dec are list in Table 3.
The frequency and the order of the mode shape are very different for S-Sep and S-Dec. The
movable joint bearing results in a reduction of frequencies of each vibration mode and a change
of order of the mode shapes. The effect of the bearing on the mode shapes and frequencies is
significant. However, the effect of the released rotation degree about the z-axis is just to bring
the local mode shapes related to the columns. It is noteworthy that the bridge has the lowest fre-
quency in relation to transverse bending, meaning that the bridge is affected more easily by lat-
eral loading than by vertical loading.
Table 3. Ten first frequencies and mode shapes estimated from the two tests.
________________________________________________________________________________
No. ____________________________________________________________________________
S-Dec (fixed joints) S-Sep (movable joints)
Freq.[Hz] Mode Freq.[Hz] Mode
________________________________________________________________________________
1 1.13 T-1 1.11 T-1
2 1.93 T-2 1.62 V-1
3 2.23 T-3 1.78 T-2
4 2.46 V-1 2.00 T-3
5 3.25 T-4 3.08 T-4
6 3.46 V-2 3.43 V-2
7 4.19 T-5 3.82 T-5
8 4.71 V-3 4.69 T-6
9 5.26 T-6 4.70 V-3
10 5.64 V-4 4.93 V-4
_________________________________________________________________________________
a) First transversal bending mode, T-1 b) First vertical bending mode, V-1
c) First transversal bending mode, T-1 d) First vertical bending mode, V-1
Figure 7. First mode shapes estimated with a-b) the beam model, and c-d) the shell model.
7.0
6.0
M-Dec
Frequency [Hz]
5.0
4.0 S-Dec
3.0
S-Sep
2.0
1.0 M-Sep
Transverse Vertical
0.0
1 2 3 4 5 6 1 2 3 4
Mode number
Figure 9 Comparison of frequencies calculated with FE models and measured in field tests.
The shapes of the modes calculated with model S-Dec and model S-Sep are the same as the
measured ones. The percentage errors of frequencies between the two calculated models and
measured results are about 10 %, except for the first frequency of vertical mode in model S-IV-
Sep. The first frequency of vibration is 1.11 Hz and 1.13 Hz in September and December, re-
spectively. The situation of the bridge is changed with the working state of bearings in different
seasons. So, the models of S-Dec and S-Sep will be used to calculate the dynamic response of
the Vindel River Bridge under different load cases.
In the field test, the bridge dynamic response under three types of passing trains has been meas-
ured: locomotives with 90 tons, steel freight train and timber freight train. Figure 11 shows the
typical responses of time-history for displacements.
The maximum measured and calculated displacements are summarized in Table 6 to 8.
1.0 1.0
Displacement [mm]
Displacement [mm]
0.0 0.0
-1.0 -1.0
-2.0 -2.0
-3.0 -3.0
-4.0 -4.0
10.0
12.5
15.0
17.5
20.0
0.0
2.5
5.0
7.5
10.0
12.5
15.0
17.5
0.0
2.5
5.0
7.5
Time[s] Time[s]
Figure 10. Vertical and transversal displacements in the centre of the deck in main span (point D) when
two test steel train (axle load = 225 kN) are passing the bridge with about 126 km/h.
1 0,8
Vertical Transversal
Displacement [mm]
Displacement [mm]
0.5 0,6
Laser sensor, filtered
0 0,4
-0.5 Vibration sensor 0.2
-1 0
-1.5 Laser sensor, filt. -0.2 Vibration sensor
-2 -0.4
-2.5 -0.6
-3 -0.8
0 5 10 15 20 25 30 35 0 10 20 30 40 50 60 70
Time [s] Time [s]
Figure 11 Vertical and transversal displacements in the center of the deck in main span (point D) when
two individual steel freight train (axle load = 225 kN) are passing the bridge with about 100 km/h. The la-
ser sensors measure both the static and dynamic part of the displacement while the vibration sensors only
measure the dynamic part, Bennitz (2006).
5 CONCLUSIONS
Measurements carried out in September and December 2005 gave the lowest frequency to 1.11
and 1.15 Hz respectively for the transverse global vibration, the higher frequency in December
probably due to increased stiffness caused by frozen ballast and concrete. The largest vertical
displacement was caused by a steel freight train and the largest transversal displacement was
caused by a timber freight train. The transverse, as well as the vertical dynamic vibration is
highly dependent on the composition of the carriers. A light, but bad composed train, can cause
larger transverse motion than a heavy steel train.
Railway arch bridge under passing trains and ambient vibrations exhibit a complex three-
dimensional response. Commercially available three-dimensional finite element software can be
implemented to predict the effect of bearings, connection of structures and load distribution on
the dynamic behaviour and response. However, the FE model should be updated with field test
results to decrease the modelling error. The FE shell models of S-Sep and S-Dec in this paper
are updated with the field test in September and December, 2005. The calculated frequencies
and dynamic responses agree well with the measured results. It shows also that the bridge was
affected more easily by a lateral loading than by a vertical loading, especially under dynamic
load cases.
The displacements in vertical and transverse directions are acceptable under the test train and
BV-3 train with axle load of 250 kN, at least, in the limit of the applied code. The dynamic re-
sponses show that the vertical stiffness is enough to allow an increase of the axle load from 225
kN to 250 kN.
The transverse stiffness of the bridge can be improved distinctly by an increase of the stiff-
ness of the end columns. Based on the calculated and measured results, the transverse stiffness
should also allow an increase of the axle load from 225 kN to 250 kN.
Because of the stochastic character in magnitude and direction of the transverse axle load ap-
plied to the rail, further investigations are suggested with field tests for the bridges. Based on the
probabilistic method, the stochastic properties of a structure, such as geometric dimensions, ma-
terial strength, cracks distribution, mass density, bearing conditions, connection components,
and Young’s modulus can be considered in assessment of the bridges.
6 REFERENCES
BBK 04 (2004): Boverkets handbok om betongkonstruktioner (2004). (The Swedish Building Admini-
stration’s Handbook on Concrete Structures). Stockholm, Sweden: The Swedish Building Administra-
tion, Division of Buildings, August 2004 (In Swedish).
Bennitz, Anders (2006): Dynamic Behaviour of the Vindel River Railway Bridge – Measurements and
Evaluation for Displacements and Eigenfrequencies. Div. of Structural Engineering, Luleå University
of Technology. Master of Science Thesis 2006:154 CIV. ISSN: 1402-1617.
BV Bärighet (2005): Bärighetsberäkning av järnvägsbroar (Assessment of Railway Bridges. In Swedish).
Handbok BVS 583.11. Banverket, CB, Borlänge 2005-06-01, 108 pp + 9 app.
Frýba, L. (1996): Dynamics of Railway Bridges. London: Thomas Telford
He, G.J., Bennitz, A., Enochsson, O. & Elfgren, L. (2006a). The Vindel Railway Bridge. Analysis and
Measurements. Tech. Report. Div. of Struct. Eng, Luleå University of Technology (to be published).
He, G. J.; Bennitz, A.; Enochsson, O.; Elfgren, L.; Paulsson, B.; Töyrä, B.; Olofsson, P. and Kronborg, A.
(2006b): Numerical modeling and dynamic behavior of a railway concrete bridge over the Vindel
River in Sweden. Proc 3rd Int Conf on Bridge Maintenance, Safety and Management (IABMAS’06),
Porto, Portugal, 16-19 July 2006. Taylor & Francis Group, London, Abstract pp 327-328. (Full paper
P-104, 8 p, on attached CD). ISBN 0 415 40315 4.
LUSAS (2005): Powerful FE technology for specialist application, Modeller User Manual, United King-
dom
Peeters, B. and De Roeck, G. (2001): Stochastic system identification for operational modal analysis: a
review, Journal of Dynamic Systems, Measurement, and Control, 123, 659-667.
Stokes, G.G. (1849): Discussion of a differential equation relating to the braking of railway bridges,
Transactions of the Cambridge Philosophical Society, 8(5): 707-735.
Willis, R. (1849): Appendix to the report of the Commissioners Appointed to Inquire into the Application
of Iron to Railway, H. M. Stationary Office, London, United Kingdom.