Proc. of Int. Conf.

on Advances in Civil Engineering 2010

Analysis of Behaviour of U-Girder Bridge Decks

V Raju1, Devdas Menon2
1
Research Scholar, Indian Institute of Technology Madras, Chennai, India
Email: [email protected]
2
Professor, Indian Institute of Technology Madras, Chennai, India
Email: [email protected]

Abstract—The concept of U-shaped bridge girder is now being (for maintenance).

increasingly adopted in urban metro rail projects and for
replacing old bridges where there is a constraint on vertical
clearance. These bridge decks are commonly designed in
practice using simplified methods that assume beam action of
the webs in the longitudinal direction and similar flexural
action of the deck slab in the transverse direction. However, Box-girder cross-section U-girder cross-section
such assumptions can lead to errors. This paper attempts to Figure 1. U-shaped girder versus Box-girder
assess the extent of error in the simplified analysis, by
comparing the results with a more rigorous three-dimensional
finite element analysis (3DFEA). A typical prototype railway
bridge girder has been taken as a case study. The results of U-girder apparent depth Beam & slab depth
the 3DFEA, in terms of load-deflection plots, have been
validated by field testing.
Index Terms— U-girder bridge deck, Simplified methods, Abutment Abutment
Three-dimensional finite element analysis
Increased clearance (U-girder) Beam & Slab clearance
I. INTRODUCTION
Figure 2. Comparison of U-girder concept with
The U-shaped girder bridge (also called ‘channel conventional “beam and slab” construction
bridge’) is a relatively new and innovative concept in
bridge deck design. U-shaped girder is appropriate when a
B. Evolution of ‘Channel Bridge’
new or modified alignment structure requires an increase in
the vertical clearance beneath the bridge. The bridge deck, The precast segmental concrete “channel” or “U-shaped”
made of prestressed concrete (PSC), has other important bridge was first developed by Jean Muller in 1990s for the
advantages, such as protection against traffic noise Champfeuillaet Overpass Bridge in France. Subsequently,
pollution, aesthetic appearance, reduced construction time, in the mid-1990s, there was an extensive research
durability and economy. This concept can be used for evaluation programme carried out in USA by the Highway
overpasses, under-crossings, viaducts, etc. Innovative Technology Evaluation Centre (HITEC).
Between 2001 and 2003, the Sorell Causeway Viaduct was
A. Description of U-girder bridge concept built in Australia. It became the first channel bridge
Structurally, the U-shaped girder bridge can be viewed viaduct built in the world [1].
as the conventional ‘single-cell box girder’ with its top The specialist rail consultancy firm, Systra, has
flange removed, as shown in Fig. 1. The two webs are developed a precast prestressed concrete U-shaped type
configured as beams positioned above and on either side of bridge based on the original channel bridge constructed in
the deck surface. The webs and the deck slab are post- France. The Wodonga Rail Bypass project in Austria,
tensioned with longitudinal tendons anchored at the two designed by Systra uses a simplified U-shaped bridge
ends of the bridge deck (with suitable ‘end blocks’). The concept as shown in Fig. 3; an international patent has been
longitudinal stiffness and strength are obtained from the taken for this concept [2].
two webs as well as the connecting passageway slab spans
4400
between the webs. The resulting requirement for the depth 450 450
3500
of girder section below the passageway level is very less Precast Channel
than that required for convectional beam-and slab type Beam
1115
designs, as shown in Fig. 2, and herein lies its main 1500
functional advantage. The U-girder is essentially a 250
‘through’ type girder where the train passage occurs on the 425
soffit slab; the side cantilevers serve as ‘keyman’ pathways
Bearing Deepened Section Beyond
2
Devdas Menon (corresponding author), Professor, Structural Figure 3. Cross section for the Wodonga Rail Bypass in Australia
Engineering Division, Indian Institute of Technology Madras, Chennai,
Inida.
© 2010 ACEE
DOI: 02.ACE.2010.01.32 28
 Proc. of Int. Conf. on Advances in Civil Engineering 2010

II. NEED FOR PRESENT STUDY

6700
The U-shaped girders are analysed and designed in 100 Thick pipe hand rail
practice using simplified methods. In simplified methods 1000 1000
4500
of analysis, the bridge deck behaviour is conveniently
350
divided into longitudinal analysis and transverse analysis.
In longitudinal analysis, the whole U-girder is treated as a Wearing coat
1800 75mm thick (avg)
simply supported beam in the longitudinal direction; it is
known as ‘simple beam analysis (SBA)’. For bending in 500
the transverse direction, the deck slab alone is analysed as
being simply supported between the two webs, as shown in 600 Section at mid span 600
Fig. 4. Such analyses do not account for the interaction
Figure 5. Cross section of U-girder bridge deck with cable locations
between the longitudinal and transverse bending, as well as
warping, distortion and shear lag effects under possible
eccentric loading (encountered in two-lane railway decks IV. DESIGN BASIS
and in highway decks). In the present paper, the study is The loads considered for analysis and design of this
limited to single track railway deck, now being used in bridge are based on Indian Railway Standards (IRS) Bridge
railway bridges in India. The results obtained by simplified Rules [3]. The analysis had been carried out by the
methods are compared with the more rigorous three- designers using the concepts of simplified analysis
dimensional finite element method of analysis (3DFEA). (described earlier) for the longitudinal and transverse
actions. The prestressing in the bridge is designed for ‘no
Depth of tension’ conditions under service loads according to IRS
deck slab Concrete Bridge Code [4]. In the present study, a rigorous
analysis of the bridge has been carried out using three
Width of deck slab
dimensional finite element analysis (3DFEA) under service
Figure 4. Deck slab is modelled as a beam for transverse analysis loads.
A. Finite element model descriptiont
III. DETAILS OF MODEL BRIDGE DECK The linear finite element model is developed using
SAP 2000 package [5], a commonly available finite
A prototype U-shaped railway bridge deck is considered element program. All the components of girder are
for analysis, which is a real structure at Villupuram, Tamil modelled using four-noded quadrilateral shell element and
Nadu, designed and constructed recently. The effective with aspect ratio of around one. To ensure accuracy of
span and overall span of the bridge are 18.5 m and 19.7 m results, a convergence study of the solution has been
respectively between the piers. The overall width of the performed. Material properties are specified as isotropic
cross section is 6.7 m and the overall depth is 1.8 m. The and the following values are used in modelling: Modulus
section of the bridge deck is a U-girder, which comprises of elasticity (M45 Grade), E = 3.35×107 MPa, Poisson’s
two webs (0.6 m wide and 1.3 m deep) with a deck slab ratio = 0.2 and Unit weight = 24kN/m3. The prestressing
(0.5 m thick) having a clear spacing of 4.5 m between the cables are modelled as tendon loads with tendon section
webs, to cater to a single lane Modified Broad Gauge element using feature in SAP 2000. A 3D view of one-
(MBG). The projections on top of the webs have an overall half of the bridge model is shown in Fig. 6. The supports
width of 1.1 m, acts as footpaths or keyman walkways (Fig. are modelled as linear elastic translational springs with
5). The proposed design is that of a post-tensioned girder specified elastomeric bearing stiffness. Live (EUDL) loads
with 15 parabolic profile cables, each cable comprising 12 are assigned as uniformly surface pressure on top of bridge
strands (each strand having an area of 98.7 mm2). The deck. Prestressing losses are manually calculated and the
arrangement of cables at the mid-span location is shown in corresponding effective prestressing force is applied as
Fig. 5. At the pier locations, the U-girder bridge deck is loads at the ends of the parabolic tendon, using the ‘load
supported on elastomeric bearings, under each web. balancing’ concept.

Figure 6. Half span of a 3D model of U-girder in SAP2000

© 2010 ACEE
DOI: 02.ACE.2010.01.32 29
 Proc. of Int. Conf. on Advances in Civil Engineering 2010

V. FIELD TESTING Fig. 9 shows the load-deflection curves at the mid-span

section of the U-girder, the deflection being measured at
It is assumed that the 3DFEA method is more rigorous
the centre of deck slab and under the webs. To facilitate
and accurate, compared to the simplified methods of
comparison with the results of 3DFEA, the effect of pre-
analysis, but the accuracy of the results need to be verified
camber in the deck on account of prestressing, is also
using available experimental results. This has been
included in Fig. 9, after compensating dead load
facilitated in this instance by the field testing carried out on
deflections. The results show close correspondence
the prototype U-girder bridge deck.
between the experimental and numerical results. The slight
Load testing was conducted at the Villupuram site of the
non-linearity in the plots is attributable to the non-linear
bridge. The test carried out mainly to assess the flexural
stiffness in the elastomeric bearings.
capacity of the structure at working loads in the elastic
range, through measuring the deflections of the super
VI. BEHAVIOUR OF U-GIRDER
structure. The static load test was performed using sand
filled bags. The load testing was conducted as per IRS It is instructive to compare the behaviour of the U-
Concrete Bridge Code [4]. girder, as obtained from simplified methods, with the more
accurate 3DFEA results for the bridge under consideration.
A. Test method
The dashed line in Fig. 10(a) shows the deflected shape (as
A test load is calculated as per IRS code, per simple beam analysis) of the U-girder section. It is
clause 18.2.3 [4], 4808 kN, for limit states of deflection. seen that in the simplified longitudinal analysis, the
Dial gauge are located with independent staging at three deflection is implicitly assumed to be constant throughout
different locations along span length, i.e., mid-span and the cross section. In Fig. 10(b), the dashed line indicates
under each support, as shown in Fig. 7. The deflections of the actual deflected shape as captured by the 3DFEA,
the girder caused due to variations in ambient temperature which includes transverse bending of plates along with
are monitored at one hour intervals for 24 hour. combined effect of longitudinal and transverse bending
Deflections are measured by using mechanical dial gauges curvatures.
together with the ambient temperature at each stage of
loading and unloading respectively. A view of the bridge 8000
span loaded with sand bags is shown in Fig. 8. The total 7000
load was applied in four increments of 1202 kN each, and 6000
Test load (kN)

deflections measured at these four stages of loading as well 5000

as unloading. All deflection data was corrected for 4000

3000 EXP - under Web
temperature effects. Exp - Centre of deck slab
2000
FEA - under Web
1000 FEA - Centre of deck slab
Foot path 0
-1 0 1 2 3 4 5 6 7 8 9 10
Deflection (mm)

Centre line of passage Figure 9. Load-deflection curves across the bottom deck

Foot path

Bearing line Midspan Bearing line

Mechanical dial gauge
Figure 7. Dial gauge locations along span and across width of girder (a) Simple beam analysis (b) 3DFEA (actual behaviour)
in plan view
Figure 10. Deflection of cross-section

Fig. 11 shows comparison of the normal force

distribution as per simple beam analysis and 3DFEA. It is
seen that the 3DFEA brings out the nonlinear variation of
longitudinal forces across the width of the section in the
flanges (deck slab and cantilevers), which is not captured
by simple beam analysis. This nonlinear variation induces
slightly higher stresses at the web-flange junction,
compared to the middle of the flange, and is attributable to
shear lag effect in flanges.

Figure 8. Load applied with sand bags

© 2010 ACEE
DOI: 02.ACE.2010.01.32 30
 Proc. of Int. Conf. on Advances in Civil Engineering 2010

1471 (3DFEA) 1471 (3DFEA)

3DFEA in the web and cantilever portion are very less
1404 (SBA) 1404 (SBA)
1399 (3DFEA) 1404 (SBA) 1404 (SBA)
1399(3DFEA)
(Fig. 13). Fortunately, the nominal reinforcement provided
2538 (3DFEA) 2538 (3DFEA) in the transverse direction in the webs and top of slab is
2406 (SBA) 2406 (SBA) adequate to give the necessary flexural strength due to this
transverse bending.
1117(3DFEA) 1117(3DFEA)
973 (SBA) 973 (SBA) –0.91(3DFEA) –0.91(3DFEA)
811 (SBA) 784 (3DFEA) 811 (SBA)
915 (3DFEA) 811 (SBA) 915 (3DFEA)
3DFEA
SBA

Figure 11. Comparison of normal force (kN/m) distribution by

3DFEA versus SBA –8.45(3DFEA) –8.45(3DFEA)
0.00(TA) 0.00(TA)
The percentage of error, based on 3DFEA, is calculated 91.34 (3DFEA)
from the following equation: 99.66 (TA)
Percentage of error = Figure 13. Comparison of transverse bending moements
(Simplified analysis results – 3DFEA results) (kN-m/m) by 3DFEA versus transverse analysis
× 100 (1)
(3DFEA results)
120
As per Eq. (1), the negative sign indicates that the

Transverse moment (kNm/m)

100
simplified method of analysis is under-estimating while the
80
positive sign indicates that the simplified method of
60 centre of deck slab
analysis is over-estimating, as comparison with 3DFEA.
40 Web-deck slab junction
Fig. 12 shows percentage of error in longitudinal force
20
estimated by simple beam analysis over 3DFEA, at mid-
0
span. From this figure (Fig. 11), it can be found that 0 5 10 15 20
simple beam analysis under-estimate the maximum stress -20
Span (m)
in the web-deck slab junction by about 12 percent; the
disparity elsewhere is within 5 percent. Figure 14. Variation of transverse moments along span

6 CONCLUSIONS
4
2
The longitudinal and transverse behaviour of a simply
A B C E F G supported U-girder bridge deck have been studied using the
Percentage of error

0
D
-2 simplified methods of analysis (used in practice), and
-4 compared with more accurate 3DFEA results. The
-6
A B F G deflections in the bridge girder, as predicted by 3DFEA,
-8 under various stages of loading, has been validated against
-10 C E
D field test results on a typical prototype U-girder railway
-12
-14
bridge, recently constructed. It is seen that simple beam
analysis generally predicts good results, except for some
Figure 12. Percentage of error in longitudinal forces estimated by local stress concentrations. The simplified longitudinal
simple beam analysis at mid span
analysis under-estimates the maximum stress in the web-
deck slab junction by about 12 percent, because it is not
Fig. 13 shows the comparison of the transverse bending
able to capture the effects arising from shear lag and
moments as predicted by transverse analysis and 3DFEA at
transverse bending. The simplified transverse analysis
the mid-span of U-girder. The transverse moments
obtained by 3DFEA and transverse analysis are shown (in over-estimates the sagging moments in the deck slab by
Fig. 13) with solid line and dotted line respectively. The about 9 percent, but fails to capture the hogging moments
sagging moments in the slab are over-estimated by about near the webs and the transverse bending in the webs.
9 percent in the simplified analysis. According to 3DFEA, These errors can be compensated for by the designer, by
some marginal hogging moments also get induced near the adopting appropriate corrections while designing and
web-slab junctions and in the webs; but these are detailing.
completely missed out in the simplified method of analysis. However, the use of 3DFEA is recommended in cases
Fig. 14 shows the variation of transverse bending moment where eccentric loading occurs on the U-girder, as in
along the span at centre of deck slab and web-deck slab double track railway and highway bridges, because the
junction. Transverse bending moments obtained by

© 2010 ACEE
DOI: 02.ACE.2010.01.32 31
 Proc. of Int. Conf. on Advances in Civil Engineering 2010

effects of torsion and distortion (not accounted for in [3] Indian Railway Standards (IRS) “Rules specifying the loads
simplified analysis) can be significant. for Design of super-structure and sub-structure of bridges.”
(Bridge Rules). Research designs and standards organization,
Lucknow, India, 2005.
REFERENCES
[4] Indian Railway Standards (IRS) “Indian Railway Standards
[1] B. Gibbens, P. Selby Smith, and G. Joynson, “Design- code of practice for plain, reinforced and prestressed
Construction of Sorell causeway channel bridge, Hobart, concrete for general bridge construction.” (Concrete Bridge
Tasmania”, PCI Journal, Vol. 49, pp. 56-66, May 2004. Code) Research designs and standards organization,
[2] B. Shepherd and B. Gibbens, “The evolution of the concrete Lucknow, Inida, 2003.
‘channel’ bridge system and its application to road and rail [5] SAP2000 NL CSI Reference manual, Computers and
bridges”, fib concrete structures (symposium), Avignon, Structures. Inc. University Avenue. Suite 540. Berkeley,
France, pp. 26-28, April 2004. California, USA, 2000.

© 2010 ACEE
DOI: 02.ACE.2010.01.32 32