October 2013

Design and Analysis of a Network Arch Bridge

BERNARDO MORAIS DA COSTA
Instituto Superior Técnico, University of Lisbon, Portugal
Contact: bernardo.costa@tecnico.ulisboa.pt

Abstract
The present dissertation aims the design and analysis of the hanger arrangement and the structural
stability of a Network arch bridge – a tied-arch bridge with inclined hangers that cross each other at
least twice. A comparative analysis with other types of hanger arrangements is also performed.

Possible solutions with respect to spans, materials and deck cross-section typology are presented
and briefly discussed. Modeling using a tridimensional finite element model of the main bridge is
described.

A detailed analysis of the hanger arrangement influence on the structural behavior is performed for
the adopted solution. Four different arrangements of hangers – a vertical, a Nielsen and two
different Network arrangements – are compared in terms of stress distributions, deflections,
hangers’ relaxation and fatigue behavior.

The linear stability analysis is finally performed for the different models, comparing their buckling
modes and discussing the results with respect to different load patterns and load increments. The
critical loads are evaluated using the European standards formulation, a simplified method and
FEModel models.

Keywords: Network arch bridge, tied-arch bridge, bowstring bridge, roadway bridge design, hanger
arrangement, arch stability analysis

Introduction
Arch bridges have outwardly directed
horizontal forces on the arch ends. These
important forces, proportional to the weight
being carried out, the relation between
Figure 1 – Arch mechanism, expressed as a “will to
bending and axial stiffness of the arch, the open", when sustaining loads.
rise, and several other factors, can be visually
understood from Figure 1, by the “will” of the Tied-arch bridges, also known as Bowstring
loaded arch to “open”. bridges, get their name from the way they
withstand these forces. The deck is used as a
2. Main objectives
tie (string) in tension to “hold” the top
compressed arch (bow) (Figure 2). The first aim of this thesis consists on
designing a Network arch bridge that can
cross over Llobregat River, in Barcelona
(Spain), 170 meters wide. This bridge should
have a total length of around 300 m,
considering the approach spans on both sides,
for crossing also a set of railway and roadway
Figure 2 – Tied-arch bridge. Arch in compression, tie in lanes. For aesthetical reasons and
tension environmental integration of the bridge, these
approach spans are also studied. Moreover,
Network arch bridges are tied-arch bridges
this dissertation intends to identify the
with inclined hangers that cross each other at
advantages or disadvantages of adopting a
least twice. To better understand it, this
Network arrangement of hangers and in
arrangement can be disassembled into three
which situations should it be considered.
or more simpler sets of hanger arrangements,
as for example the Nielsen arrangement of A second aim of this work is to investigate the
hangers, from Figure 3 to Figure 5, with structural influence of the different hangers’
hangers not necessarily with the same slope. arrangements on the bridge behavior. Four
different hangers arrangements are studied
using tridimensional SAP2000 FEModels,
namely: i) a Vertical hangers arrangement,
ii) a Nielsen hangers arrangement, iii) a
Figure 3 – Nielsen arrangement of hangers. 1 set of
hangers.
Network hangers arrangement with constant
slope, and iv) a Network hangers arrangement
with variable slope. The influence of the
following aspects are investigated: i) resulting
Figure 4 – Hangers cross each other once. 2 sets of stresses distributions on the arch, ties and
hangers. hangers, ii) total stiffness of the structure and
expected deflections, iii) number and
importance of relaxing (compressed) hangers,
and iv) global stability of the structure.
Figure 5 – Network arrangement of hangers – most
hangers cross each other twice. 3 sets of hangers. Finally, it is also a main objective of this work
to investigate the stability of the arch,
Using the Network arrangement of hangers in describing and comparing the multiple
[1]
a tied-arch bridge, Per Tveit refers that is possible approaches. A linear stability
possible to save between 40 % and 50 % of analysis is performed, for the different
the cost of the superstructure, when models and arrangements studied,
comparing with other steel bridges. considering five different load patterns, and
discussing the different ways of incrementing
the bridge loads up to the bucking load. The h = 1.8 m, b =0.8 m, tw = 0.012 m, tftop =
different procedures to obtain this buckling 0.020 m, tfbottom = 0.040 m. End-cross-
load are also investigated using the proposed sections: h=1.0m, b= 0.4 m, tw = 0.012m,
formulation from the European standards, a tftop = 0.020m, tfbottom = 0.040 m]; iii) a
simplified method proposed by Outtier et concrete (C40/50) slab, 250 mm thick, with
[2]
al. and comparing the results with the ones φ25//0.10 longitudinal reinforcement bars
obtained using FEModel linear and nonlinear (A500); iv) 2 steel (S420N) arch box-sections
analysis. (1.400 x 1.200 x 0.040 m) leaning 79⁰
inwards; v) 2 x 70 steel (S460N) hangers, with
a 80 mm diameter, a Network arrangement
3 Adopted Solution with hangers equidistant (5 m distant) ate the
The adopted solution (Figure 6) has a total deck level, and coincident with the ribs / ties
length of 350 m, with two approach viaducts intersection, and a constant 65⁰ slope; and
with balanced spans of 25.5 – 34 – 25.5 m, vi) other secondary steel elements such as 7
and a tied-arch span of 180 m, separated by bracing box-section beams linking the arches
expansion joints. A composite steel-concrete and 2 box-section end-cross-girders (Fig. 6).
deck, 26.6 m wide, is the most economical
The adopted constructive procedure includes
solution. The main tied-arch span comprises
the launching of the steel part of the arch
the following structural elements: i) 2 steel
structure through the approach bridge decks,
(S420N) ties with a box-section (1.411 x 1.344
until its final position, with the aid of a
x 0.030 m); ii) 35 steel (S355N) ribs with a
floating pontoon, as can be seen in Figure 7.
variable I cross-section [middle cross-section:

Figure 6 – Elevation view, plan view and deck’s cross-section view of the designed bridge.
 of the deck; and b) tension of the slab due to
the referred arches’ “will to open”.

Figure 7 - Pushing the arch through the access viaduct.

Sequence by Per Tveit [1].

Then precast slabs are then put in place

supported by the ribs’ top flanges and finally
concreting operations are done in a
symmetrical form with respect to the arch.
Headed studs on the ribs’ top flange and ties’
inner web assure the deck’s composite
behavior.
A database collected by Gonçalves, P. [3] gives
useful information on several built tied-arch
bridges. The comparison of these results with
the adopted solution (a star in the graphics) is
presented in Figure 8.

4. Structural analysis
4.1 Overview

The network arch can be compared to a beam

with a compression (arch) and a tension (tie)
zone. A higher arch decreases axial forces in
the chords and it’s mainly aesthetic reasons
that limit this height. The network hangers’
arrangement act like a web, taking some of
variation of the shear, while most of it, is Figure 8 – Adopted solution comparison with built tied-
taken by the vertical component of the arch arch bridges. Database collected by Gonçalves, P. [3]
axial force.
Taking into consideration the constructive
4.2 Deck procedure, tension of the slab comes only
from the SDL (Superimposed Dead Loads) and
The tied-arch span imposes two important
variable loads, as before that, the precast
events to its deck: a) global bending moment
slab’s do not have stiff solid concrete
And 𝜒 × 𝐴 × 𝑓𝑦 (1)
connecting them and only the ties are there 𝑁𝑏,𝑅𝑑 = = 67588 𝑘𝑁 ≥ 𝑁𝐸𝑑
1,0
to resist the horizontal forces from the arch. With: Χy ~ Χz = 0.790 and NEd = 44926 kN
The same occurs with the transversal
While the instability occurs with an out-of-
compressions of the composite behavior slab-
plane movement for the all observed buckling
rib.
modes, in-plane movements are
Ribs, which are a class 4 section, must resist a simultaneously visible in every single buckling
constructive process without the aid of mode, since the arches are inclined at 79⁰
scaffolding. Therefore, as usual the composite with the horizontal plan and cannot buckle
section is only activated for SDL and variable outwards without some vertical movement.
loads, and for the DL (Dead Loads), only the Since instability in both axis of the arch cannot
steel section resists. be easily divided, in a conservative and simple
For the same reason, the ties, for the DL, will way, λy is assumed to be equal to λz, such as χy
support alone the arch “will” to open, and, at and χz. The resultant ULS verification is
both ends of the span (at the corners), it also presented according to EN1993-1-1[4]:
supports basically alone all following loads N𝐸𝑑 𝑀𝑦,𝐸𝑑 𝑀𝑧,𝐸𝑑
+ 𝑘𝑦𝑦 × + 𝑘𝑦𝑧 ×
since the connection to the slab is barely 𝛸𝑦 × N𝑅𝑘 𝑀𝑦,𝑅𝑘 𝑀𝑧,𝑅𝑘 (2)
𝛸𝐿𝑇 × 𝛾𝑀1
𝛾𝑀1 𝛾𝑀1
mobilized. Moreover, the interaction with the 44926 4455 8985
+ 1.059 × + 0.529 × = 0.963 ≤ 1.0
arch causes important bending moments in 67588 36000 28462

these corners.
4.4 Hangers
The resulting stresses on the deck, at the
conditioning cross-sections, for the Ultimate Hanger’s verifications are the following:
Limit States, are summarized in Table 1. 1) Maximum axial force.
Table 1 – Deck’s main conditioning stress results 2) Characteristic axial force (limited to
Structural Element fEd 𝜏Ed fyd 50% of the ultimate resistance).
Slab's rebars 403 - 435 3) Fatigue.
Ribs 213 161 355
4) Relaxation.
Ties 392 59 420
Units MPa 1), 2) and 3) are easily checked, by correctly
adapting the hangers’ characteristics, and 2)
4.3 Arches resulted the most conditioning. The 4),

The arch springs (corners) are also over however, presents great interest to be
discussed, since it is entirely related to the
stressed from the interaction between both
hangers' arrangement and live loads patterns
chords (arch & tie). Though, the highest
bending moments are in-plane, they occur in and significance.

the same arch sections and are due to wind Relaxation can be seen as the consequence of
action. The elastic stress verification, at the the hangers’ inability to sustain compression
arch spring, are presented in Table 2. forces, and, at both ends of the span, hangers
Table 2 – Arch corner elastic verification (MPa) do compress due to a truss-beam like
Stress NEd MEd 3-3 MEd 2-2 Total Limit behavior. Notice the similarities in Figure 9
σx,Ed -221 52 133 405 420
τEd - - - 20 242 and Figure 10.
 right combination of prestress that prevents
relaxation. Two prestress combinations,
applied in different periods were required to
improve the structure behavior during the
construction stages.
Figure 9 – Hangers’ axial forces for DL (half span
shown). Leaning inwards’ hangers are compressed.
5. Hangers arrangements and
arch instability investigations
5.1 Hangers arrangement study
Figure 10 – Truss-beam scheme. Leaning inwards Different arrangements are studied, namely:
diagonals compressed.
“Vertical”, “Nielsen”, “Network” and
Then, for half-span loading, other central “Optimized Network”, and are respectively
hangers tend to relax (dashed hangers in illustrated in Figure 12.
Figure 11).

Vertical

Nielsen

Figure 11 - Influence line for shear force over A-A,

related to the influence line of the dashed hangers. Network
.axial force.

Hanger’s relaxation may or may not have

significant consequences on the structure Optimized Network
since it changes considerably its structural Figure 12 – Hangers arrangements studied.
behavior. Therefore it might be of interest to The same materials, deck and arch cross-
prevent it by adopting an appropriate sections were adopted in the different
[1]
hangers’ arrangement. Per Tveit and Brunn models. The hangers’ cross-section area was
[5]
& Schanack give multiple useful advices on defined as inversely proportional to the
this subject, leading in the present case to the number of hangers (Table 3). Two load
65⁰ slope adopted, which prevents relaxation distributions, matching the LM4 preconized in
for a half-span live load patterns. EN1991-2[6], were defined: i) “LD-All” – Load in
For the full-span load case, it is decided to all span length, and ii) “LD-Half” – Load in the
study the required prestress forces in the first left-half of the span (see the first two schemes
8 leaning inwards hangers of each arch corner of Figure 13). The main forces results and
to avoid relaxation. For that, an influence displacements are listed in Table 4, and lead
matrix of the effects of pre-stressing each to the following comments:
hanger was built. This allows obtaining the
 Table 3 – Hangers characteristics on the different models
Vertical Nielsen Network Opt. Network
Nº 2 x 35 2 x 34 2 x 70 2 x 80
A (m2) 0.0101 0.0104 0.0050 0.0044
fyd (Mpa) 460 460 460 460
NRd (kN) 4646 4784 2312 2021

Nº - 2 (arches) x number of hangers (per arch)

 Vertical arrangement is extremely was defined inversely proportional to the

vulnerable to half-span loading. It balances number of hangers.
unsymmetrical loads by bending both the 2. Even axial forces (low Nvariance). This
arch and the tie, since hangers do not prevents overdesigned hangers and/or
connect different sections of the deck to different solutions for different hangers.
the arch. This is connected with what is Nvariance is simply defined, in Eq.3.
observed with deflections.
Nmax − Naverage
𝑁𝑉𝑎𝑟𝑖𝑎𝑛𝑐𝑒 = (3)
 The distance between hanger’s nodes on Naverage
the arch does not influence its results.
Therefore, the axial force results on the
 A higher number of hangers fairly decrease hangers are presented next, in Table 5, which
the resultant bending moments on the
raises the following considerations:
ties.
 To compensate, and clearly related to the
 The adopted solution (Network) has the
disturbing results obtained previously, the
best results on the arch and on deflections.
Vertical arrangement shows the best
 The Optimized Network arch has the best results here.
results on the tie.  The Optimized Network arrangement
 With almost the same number of hangers finally reveals its benefits, having virtually
as the Vertical, the Nielsen arrangement no compressed hangers and a considerable
behaves seemingly well, in respect to the even axial force between hangers.
main axial forces and bending moments.  Nielsen arrangement is by far the most
penalized, in contrast to the previous fairly
The next step is to evaluate the hangers’ good results. The Nmin = -889 kN indicates
behavior. An optimal arrangement solution an alarming compression value. In fact, this
accomplishes two goals: compression force alone will exceed the
tension forces from the permanent loads.
1. Low maximum axial force. Since models
Moreover, an even bigger Nmax leads to the
differ on the number of hangers, the way
very demanding results observed.
to better access this is to measure Nmax /
NRd. NRd is proportional to the area, which

Table 4 – Main forces and displacements on the different hanger arrangements models.
Arrangement: Vertical Nielsen Network Opt. Network
Units
LD: All Half All Half All Half All Half
M33,max 1631 -12203 937 916 848 688 996 817 kNm
Arch
Nmax -8492 -4513 -7767 -5633 -7664 -5576 -7811 -5635 kN
M33,max -1563 -10955 934 1114 732 737 607 656 kNm
Tie Nmax 5339 9721 4931 3653 4967 3677 4738 3517 kN
δmax 132 860 62 78 57 38 65 38 mm
 Table 5 – Hanger axial forces on the different hanger arrangements
Arrangement: Vertical Nielsen Network Opt. Network
Units
LD: All Half All Half All Half All Half
Nmin 215 58 46 -889 -118 -168 14 -65 kN
Nmax 344 222 584 1114 283 304 248 211 kN
Hangers

Naverage 277 138 289 144 144 72 154 77 kN

NRd 4646 4784 2312 2021 kN
Nmax
7% 5% 12% 23% 12% 13% 12% 10% -
N𝑅𝑑
NVariance 24% 61% 102% 672% 96% 322% 61% 174% -

From these results, compressed hangers were Finally, during the analysis, a few more
removed from the Nielsen model, LD-Half was remarks were noted:
applied and a bending moment of 10606 kNm
was obtained, opposing to the previously  Within the same arrangement, the greater
obtained 688 kNm with compressed hangers. axial stiffness of the hangers, the more
Final conclusions are noted: uneven forces result.

 The Nielsen arrangement has severe  The higher the number of hangers, the

relaxation issues. For unsymmetrical loads lower bending moments on the ties.

it sees many of its hangers relaxed. This  More steep hangers give smaller hanger
changes the apparently good results forces but bigger stress variations and
obtained in Table 4, since hangers cannot relaxations problems.
mobilize compression. For this reason,
5.2 Arch Instability analysis
when there is a live load in the nearness of
relaxing hangers, the effects of The same arrangements are analysed
incrementing that live load are very much considering 5 different LD (load distibutions)
like the ones in the Vertical Model. So, of the same LM4 preconized in EN1991-2[6]
accordingly to a higher or lower (Figure 13). The results of the instability
importance of the live loads, the Nielsen analysis were obtained from incrementing
arrangement behaves respectively, more Live Loads only, on the FEModel. This is a
closely to the Vertical Model or more conservative approach which revealed to be
closely to the Network models. sufficiently accurate. The results are listed in
Table 6.
 The Vertical model, for unsymmetrical
loads, gets bending moments on the arch LD 1
(LD-All)
17 times greater than the ones of the
LD 2
Network arch. (LD-Half)

 The Network arrangement has the lowest LD 3

forces and bending moments on the

LD 4
arches. Its disadvantages to the Optimized
Network, regarding the hangers, can be LD 5
partially compensated by applying
Figure 13 – Load Distributions applied in this study.
appropriate prestress.
Blue color corresponds to a 5kN/m2 vertical uniform
distributed load.
 Table 6 – Instability Analysis Results on the FEModels
Arrangement LD: 1 2 3 4 5
Vertical λ 12.2 22.0 24.2 24.3 24.4
NEd [kN] 8492 4513 6200 5578 5111
NFE,el [kN] 103761 99316 150152 135562 124630
Nielson λ 16.9 28.2 33.6 31.6 32.8
NEd [kN] 7769 5162 5682 5062 4637
NFE,el [kN] 131215 145498 190989 159954 152288
Network λ 17.0 26.1 33.8 33.9 33.9
NEd [kN] 7677 5576 5604 4644 4636
NFE,el [kN] 130231 145478 189221 157435 157164
Opt. Network λ 16.3 26.2 32.5 32.6 32.6
NEd [kN] 7811 5635 5696 4675 4982
NFE,el [kN] 127343 147775 184872 152338 162277
λ = buckling factor NEd = maximum compression force the LD applied NFE,el = λ x NEd

According to Per Tveit [1], in a normal network i) and ii) predicts accurate results for the
arch the decisive load cases are maximum FEModels illustrated respectively in Figure 14
load on the whole span (LD1), which was and 15.
confirmed.

Remembering that all buckling modes are

significantly out-of-plane, the results were not
much affected by the hangers’ arrangement.
Figure 14 – FEModel with stiff wind bracing elements -
They were, in fact, affected and are mostly NFE,el = 226265 kN.
related to the previous stress distribution
results of Table 4.

For the out-of-plane buckling, EN1993-2[7]

presents two procedures to obtain the β
(Buckling length factor): i) “Out-of-plane Figure 15 – FEModel with free standing arches - NFE,el =
25044 kN.
buckling of arches with wind bracing and end
portals”, and ii) “Out of plane buckling factors Finally, a study from Outtier et al [2], based on
for free standing arches”. The results from a database of more than 50 steel tied-arch
proceeding according to i) and ii), and the bridges spanning from 45 to 200 m, where
result of the FEModel, for the Network detailed linear and nonlinear elastic-plastic
arrangement, are resumed in Table 7. analysis were performed and compared to the
EN1993-2[7] procedures, led into a simplified
Table 7 – Instability procedures comparison
Procedure Ncr
method of assessing instability by proposing
i) 207310 kN an alternative formula to obtain β. The results
ii) 19021 kN of the Ncr obtained for the current Network
FEModel 130231 kN bridge, from this alternative β, were

It can be concluded that both EN1993-2[7] unexpectedly high (Ncr = 404551 kN). In fact, β

procedures provide a lower and upper factors, obtained for spans greater than

boundary of the FEModel result, but within an 150 m were suddenly low. After confirming

unsatisfying large interval. Actually, procedure with one of the researchers that this was not
expected to happen, it was concluded it is
worthy trying to improve the formula and to In the author’s opinion, for out-of-plane
extend its validity domain in the future since it conditioned tied-arch bridges, EN1993-2[7]
offers an easy and straightforward procedure. may be carefully used in the two situations
analyzed and successfully compared: i) when
no bracing beams exist, or ii) when the
6. Conclusions bracing beams form a really stiff structure.

The base case design here developed allowed Finally, this study hopes to have
substantial material savings when compared demonstrated that Network arch bridges can
to many other tied-arch bridges. be competitive and structurally efficient when
compared to other tied-arch bridges.
The pre-design of the hanger’s arrangement,
facilitated by Per Tveit[1] and Brunn &
Schanack[5], proved to be remarkably accurate 7. References
on the benefits it predicted.
[1]
Both Network arrangements analyzed Per Tveit - The Network Arch. Findings on
network arches during 54 years (2011). Available
evidenced clear structural advantages over
at:
the Vertical arrangement. The Vertical http://home.uia.no/pert/index.php/The_Network
arrangement only presented benefits for the _Arch [10/09/2013]
hangers’ forces, as a consequence of over [2]
Outtier et al. - Amelie Outtier, Hans De Backer,
requesting the bending stiffness of the Ken Schotte, Dries Stael, Philippe Van Bogaert -
chords. Design methods for buckling of steel tied arch
bridges - IABSE conference 2010.
On the Nielsen results, if significant [3]
Gonçalves, P. - Pedro Pereira Clemente Andrade
unsymmetrical live loads exist, the severe Gonçalves - Estudo Prévio de um Tabuleiro em
relaxation of hangers leads this solution to Arco Superior do tipo Bowstring, Tese de
behave similarly to the Vertical hangers’ Mestrado, IST, 2012.
arrangement model. [4]
EN1993-1-1 - European Committee for
Standardization (CEN), “Eurocode 3 – Design of
With the inclination of the arches and with
steel structures – Part 1-1: General rules and rules
the presence of the bracing beams it is for buildings”, December 2003.
extremely unlikely to occur a pure in-plane [5]
Brunn & Schanack - Benjamin Brunn & Frank
buckling. The inclination of the arches also Schanack - Calculation of a double track railway
reduces the wind portal frames and the network arch bridge applying the European
bracing beams length, resulting in a more standards (2003).

stable solution. [6]

EN1991-2 - European Committee for
Standardization (CEN), “Eurocode 1 – Actions on
Hangers’ arrangements don’t affect directly
structures – Part 2: Traffic loads on bridges”,
the stability of the arch significantly, but September 2003.
indirectly through the stress distributions. An [7]
EN1993-2 - European Committee for
integrated methodology for the simplified Standardization (CEN), “Eurocode 3 – Design of
analysis of in-plane and out-of-plane buckling steel structures – Part 2: Steel Bridges”, October
2006
of the arch still needs to be developed.