You are on page 1of 9

See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/327838512

Non-conventional Cable System in Multi-span Cable-stayed Bridges

Conference Paper · September 2018

CITATIONS READS

0 82

3 authors:

Clara Cid Bengoa Aitor Baldomir


University of A Coruña University of A Coruña
8 PUBLICATIONS   6 CITATIONS    64 PUBLICATIONS   195 CITATIONS   

SEE PROFILE SEE PROFILE

Santiago Hernandez
University of A Coruña
220 PUBLICATIONS   637 CITATIONS   

SEE PROFILE

Some of the authors of this publication are also working on these related projects:

OPTISAFE - Probabilistic optimization of damaged and undamaged aircraft structures under dynamic and impact loads View project

Optimum crossing cable system in multi-span cable-stayed bridges View project

All content following this page was uploaded by Clara Cid Bengoa on 24 September 2018.

The user has requested enhancement of the downloaded file.


40th IABSE Symposium, 19-21 September 2018, Nantes, France.
Tomorrow’s Megastructures

Non-conventional Cable System in Multi-span Cable-stayed Bridges


Clara Cid, Aitor Baldomir, Santiago Hernández
Structural Mechanics Group, School of Civil Engineering, Universidade da Coruña, A Coruña, Spain

Contact: clara.cid.bengoa@udc.es abaldomir@udc.es

Abstract
The design of multi-span cable-stayed bridges requires the control of horizontal displacements on
the towers head when unbalanced loads are applied on deck. There are several strategies to
reduce these displacements, being one of them the use of crossing stay cables in a central section
of each main span. Based on this bridge typology, a general methodology to obtain the optimum
cable system has been developed. Number of cables, anchor position of cables on the deck, cable
areas and presstressing forces are considered design variables simultaneously into the
optimization process. Each cable anchor position is associated to an independent design variable,
allowing overlapping stay cables, different spacing of cables along the deck as well as different
number of cables on each side of the same tower. Queensferry Crossing has been chosen as
application example where the final design achieves an important saving in cable steel.
Keywords: multi-span; cable-stayed bridge; crossing stay cables; optimum design; overlapping stay
system

[3-15]. Although some major enhancements were


1 Introduction made on multi-span cable-stayed bridges in the
The construction of multi-span cable-stayed last years, no optimization strategies were
bridges has improved considerably in recent years. performed to improve their cable system, being
Relevant examples are the Erqi Yangtze River the cable positions as well as the number of cables
Bridge, the Jiashao Bridge, the Millau Viaduct, the fixed design parameters [16-18].
Ting Kau Bridge or the Queensferry Crossing. The The goal of this research is to propose a general
main difficulty when designing this bridge optimization approach to obtain the optimum
typology is the significant sway of the interior number, anchor positions, cross-sectional areas
towers resulting from unbalanced live loads and post-tensioning cable forces in the cable
applied on deck. Consequently, the moments at system. For that purpose, each cable anchor
the base of these towers increases and the vertical position on the deck is associated to an
deflection in the deck in aggravated. Several independent design variable allowing overlapping
strategies were proposed to address this issue stay cables in main spans. Therefore, the cables
[1,2]. may have different spacing along the deck and the
In the existent literature there is an extensive number of cables may not be the equal each side
research related to the optimization of different of the same tower, unlike pre-existing approaches.
parameters in single-span cable stayed bridges

S5-157
40th IABSE Symposium, 19-21 September 2018, Nantes, France.
Tomorrow’s Megastructures

Figure 1. Geometry of the generic finite element model

Figure 2. Nomenclature considered for the bridge model


connection from the mesh, making possible that
2 Problem definition the cable anchor position on the deck is
considered as design variable.
2.1 Bridge structural model
A generic model of a four span cable-stayed 2.2 General definition of the optimization
bridge was defined. The geometry definition as problem
well as the nomenclature used are shown in The mathematical optimization problem can be
Figures 1 and 2. expressed as follows:
As can be seen in Figure 2, the number of cables in NC
the half length of the bridge (NC) will be the sum min V ( x , x , x )  2    xiE  xiA  Li ( xiP )  (1a)
i
E
i
P
i
A

of cables in the side span (NS) and the cables in i 1


the main span (NM = NM1 + NMC1 + NMC2 + NM2). subject to:
A generic 2D finite element model was defined g j ( xiE , xiP , xiA , xiF )  0 i  1,..., N C 
using the commercial software Abaqus[19]  (1b)
through a Python script. The structural analysis j  1,..., m 
type adopted for the Abaqus FE model is
nonlinear taking into account the beam-column being V the steel volume in the cable system. The
effect (important in deck and towers), P-Δ effect design variable xiE represents a binary variable
and cable sag effect. used to define the existence of the ith cable, and
Cables-deck and cables-tower connections were xiP, xiA and xiF are the design variables associated
designed as “tie” constraints, realising the to the cable anchor position on the deck, cross
rotational degree of freedom of cable nodes but sectional area and prestressing force, respectively.
guaranteeing zero displacement between both Li is the length of the ith cable, which depends on
entities. These constraint entities make the anchor position on the deck and gj refers to
independent the cable-deck and cable-tower the jth design constraint.

S5-158
40th IABSE Symposium, 19-21 September 2018, Nantes, France.
Tomorrow’s Megastructures

strategy of using an initial design with a large


enough number of cables in each span. All the
cables that tend to its minimum area will be
considered as non-existent cables. As each pair of
cables in the crossing zone will share the cable
anchor position, the number of design variables
xiP will be lower than NC, being this number
NP=NC−NCM2. As a result, the total number of
design variables (position, area and prestressing
Figure 3. Nomenclature of design variables force) is NT=NP+2NC.
Figure 3 shows a general scheme of the bridge The Sequential Quadratic Programming algorithm
model and the corresponding design variables for (SQP) implemented in the MATLAB function
each cable. As can be seen, cables belonging to fmincon was used as optimization tool method. At
the crossing zone (LMC) are grouped by pairs (e.g. each iteration of the optimization process, a
x7A and x1A), sharing the same anchor position on parametrized Python script is firstly updated with
the deck (x7P). As a result, both cables can exist the new values of the design variables and then
(leading to a V-configuration), only one of them or externally run through the software Abaqus [19]
none of them. to obtain the structural responses and evaluate
the design constraints.
3 Optimization strategy
4 Application example: the
As presented in Section 2.2, four design variables
have been assigned to each cable. The most Queensferry Crossing
intuitive approach would be consider xiE as a The Queensferry Crossing Bridge is the longest
discrete variable, taking the values 0 (cable does three tower cable-stayed bridge in the world, with
not exist) and 1 (cable exists) and using a Mixed a total length of 2638 m. The part corresponding
Integer Nonlinear Programming (MINLP) method, to cable-stayed bridge is 1950 m long with two
such as the GA of MATLAB [20], that manages main spans of 650 m, two lateral spans of 325 m
discrete and continuous variables simultaneously. and three towers around 200 m high. A monolithic
The strategy would be to define a large number of connection exists between the deck and the
cables in each cable region defined in Figure 2. central tower while the lateral towers only
Doing so, some of the existence variables would restraint the vertical displacement of the deck
correspond to a zero value, meaning that the [21]. The deck is also anchored on V-piers placed
number of cables initially considered was enough. at the beginning and 104 m inwards the side span,
However, as stated in [13] the use of this specific allowing exclusively the rotational degree of
algorithm is inefficient when considering a large freedom (Figure 4).
number of design variables of different nature.
The projected bridge has a total of 288 cables
Consequently, a different approach consisting of distributed in two cable planes. The number of
using continuous variables exclusively was cables in each deck region is listed in Table 1. As a
followed. That is, the discrete variables xiE were 2D finite element model of the bridge was chosen,
suppressed and the cable existence was taking only one plane of cables was considered.
into account by setting the lower bound of the Therefore, the values of area and prestressing
cross-sectional areas to value very close to zero forces must be distributed in the two cables
(10−6 m2). This allows the use of a gradient-based defined in the bridge project.
optimization algorithm, following the previous

S5-159
40th IABSE Symposium, 19-21 September 2018, Nantes, France.
Tomorrow’s Megastructures

Figure 4. Distribution of cables in the real bridge.


Table 1. Cables in the half length of the bridge.

Side span Main span (NM)


NC
(NS) NM1 NMC1 NMC2 NM2

72 24 14 10 10 14

The material used for cables and deck is steel with


E=200 GPa and n=0.3, and concrete for towers
and piers, with E=35 GPa and n=0.2. The
mechanical properties of the deck and tower
sections are shown in Table 2. The dimensions of
the tower vary linearly between the sections at Figure 5. Load cases.
the base (C-C), at the deck level (B-B) and at the
As initial design equally spaced cables have been
top (A-A).
assumed in each span and a unique value of area
Table 2. Mechanical properties of the bridge and prestressing force is assigned for all the
cables. The initial number of cables is presented in
A-A B-B C-C Cables Deck Table 3 as well as the initial area value and
Des. prestressing force. The total number of design
A(m2) 22.10 32.81 56.65 2
Var. variables for each optimization problem are listed
in Table 4.
Iz(m4) 118.83 411.61 1658.71 - 200
Table 3. Cable system used as starting point
Iy(m4) 57.91 193.21 1386.67 - 7.6
NM A0
NC NS F0 (kN)
Load cases considered are shown in Figure 4. A NM1 NMC1 NMC2 NM2 (m2)
uniform load of 200 kN/m was applied in the deck 60 20 2 18 18 2 0.03 11,160
to represent the dead load as well as a live load of Table 4. Number of design variables considered
102.5 kN/m (Spanish regulation of bridges [22]).
xiP xiA xiF Total
4.1 Definition of the optimization problem 42 60 60 162

The real bridge has a length crossing zone 4.2 Formulation of the optimization
LMC=146m with equally spaced cables, which problem
results in a total amount of steel in the cable
system of 759 m3 (5958 tons). The optimization problem is formulated in Eq. (2).
In Table 5 are summarized the limit values defined
The purpose of this research is to find out if by below for each design constraint, being k the load
modifying the value of the crossing length new case and LS=325 m, LM=650 m, and HT=200 m. The
cable distributions appears and the cable weight minimum distance between cables in deck is
can be decreased. For that purpose, a value of dmin=2 m.
length crossing zone (LMC) almost equal to the
total length of the main span (LM) was considered
(LMC=640 m).

S5-160
40th IABSE Symposium, 19-21 September 2018, Nantes, France.
Tomorrow’s Megastructures

Table 5. Limit values for design constraints


Side Span Main Span
u0max ukmax sM sC,deck sT,deck sC,tower sT,tower
w0max wkmax w0max wkmax
(m) (m) (MPa) (MPa) (MPa) (MPa) (MPa)
(m) (m) (m) (m)
LS/7500 LS/500 LM/7500 LM/500 HT/7500 HT/500 0.45fpk 200 300 35 3.2

NC length of the bridge (NC*) corresponds to 23, value



min V  2   xiA  Li ( xiP )  (2a) substantially lower compared to the real bridge
i 1
(NC=72).
subject to:
Table 6. Minimum volume of cable steel
0   k ( xi )   M (2 b)
Real bridge Optimum design
wkj  wmax
k
(2c) 759 m3 634.15 m3
k
utower ,p
 umax
k
(2d) Table 7. Final number of cables at the optimum
k  0,..., 5 i  1,..., N C j  1,..., N D p  1, 2, 3 Main Span (N*M)
* Side Span
N C
 C , deck   top
k
, deck , j
  T , deck  (N*S) N*M1 N*MC1 N*MC2 N*M2
 (2e)
23 6 0 7 10 0
 C , deck   bottom
k
, deck , j
  T , deck 
Figure 6 shows the initial configuration and the
j  1,..., ED k  0,..., 5
resulting layout of the cable system. In Table 8
 C ,tower   left
k,p
tower , l
  T ,tower  appears the initial and final anchor position of the
 (2f ) cables that survives after the optimization.
 C ,tower   right
k,p
  T ,tower 
tower ,l
Significant changes are observed between initial
l  1,..., ET k  0,..., 5 p  1, 2, 3 and final values of xiP. In that table are also
summarized the numerical values of the optimum
xi 1  xi  d min i  1,...N P  1 (2g)
areas and prestressing forces.
Regarding to the final cable arrangement, a subset
The tensile stress in cables and normal stresses in of cables emerge from the central tower top until
deck and towers (Eqs. (2b), (2e) and (2f)) are the border of the length crossing zone. This is
evaluated in ultimate limit-state (ULS) (Eq. (3b)) consistent from an engineering point of view,
while vertical displacements of the deck and since these cables contribute to reduce the sway
horizontal displacements of the tower tops (Eqs. of the central tower. Moreover, some cables tend
(2c) and (2d)) must satisfy the serviceability limit- to cluster near the V-piers placed 104 m inwards
state (SLS) (Eq. (3a)). the side span keeping the minimum distance of 2
SLS :1.0× (Dead Load) +1.0× (Live Load) (3a) m (Eq. (2g)). These piers act as anchor points
allowing high stability of lateral towers. The center
ULS :1.35× (Dead Load) +1.35× (Live Load) (3b)
of main span is supported by two groups of cables
that emerge from the lateral and central tower,
4.3 Numerical results without crosses between them. This cable layout
has many similarities with the Ting Kau Bridge.
The optimum cable steel volume appears in Table
6. As can be seen, the volume of cable steel The main difference is that not all cables are
reduces up to a 16.45%, corresponding to a placed at the boundary of the length crossing zone
volume of 634.15 m3, substantially lower than the (LMC), but two other cables appear defining V-
759 m3 of the projected bridge. Table 7 configurations. Additionally, cables are not equally
summarizes the total number of cables that spaced along the deck. Thus it can be concluded
remain in each deck zone after the optimization that the optimum solution leads to a hybrid design
process. The total number of cables in the half between the adopted in the Ting Kau Bridge and

S5-161
40th IABSE Symposium, 19-21 September 2018, Nantes, France.
Tomorrow’s Megastructures

Figure 6. Initial configuration (upper) and optimum cable distribution (bottom) with LMC=640 m
Table 8. Initial and final cable positions in deck, cable areas and prestressing forces at the optimum
Side Span (NS) Main span crossing zone - tower 1 (NMC1)
Cable
5 6 7 8 11 15 2 3 5 6 8 9 10
0,P
xi (m) 65.00 81.25 97.50 113.75 162.50 227.50 397.37 431.05 498.42 532.11 599.47 633.16 666.84

xi f,P(m) 63.02 86.96 102.74 104.74 162.65 229.46 414.45 436.34 499.88 523.35 589.31 618.95 649.01
A 2
xi (m ) 0.0337 0.0036 0.1000 0.1000 0.0230 0.0594 0.0281 0.0311 0.0122 0.0514 0.0624 0.0784 0.0322
F
xi (kN) 17025 1435 38433 38143 11492 34700 11535 11260 3103 14246 16122 24469 13569

Main span crossing zone - tower 2 (NMC2)


Cable
1 2 3 11 12 13 14 16 17 18
xi0,P(m) 363.68 397.37 431.05 700.53 734.21 767.89 801.58 868.95 902.63 936.32
xi f,P(m) 330.00 414.45 436.34 708.63 726.14 769.85 798.65 873.05 892.62 932.32
xi A(m2) 0.1000 0.0217 0.0421 0.0593 0.0457 0.0487 0.0121 0.0403 0.0275 0.0036
F(kN) 51114 11301 21837 23853 18556 20362 4369 20128 15636 2227
xi

the Queensferry Crossing.


5 Conclusions
The final values of design variables are presented
in Figure 7. The blue colour makes reference to This research presents a general approach to
the cables anchored to the lateral tower and the optimize the cable system in multi-span cable-
green colour to the cables anchored to the central stayed bridges. The optimum number of cables as
tower. The final area values and presstresing well as their anchorage positions, areas and
forces of the existent cables are presented prestressing forces were obtained, guaranteeing a
through bar graphs, using the same colour code. lower steel volume in cables regarding the
solution of overlapping cables in center of main
The design constraints at the optimum are shown spans. The main advantage of this methodology is
in Figure 8. The active constraints correspond to to formulate the problem avoiding the use of
axial stresses for the LC 1, 2, 3, 4 and 5, vertical binary variables (cable existence). Therefore we
displacement in the deck in lateral and main spans can take advantage of the efficiency of gradient-
for DL and exclusively in main spans for the LC 4 based optimization algorithms (SQP) which
and 5. The horizontal displacement of the central include a considerable number of design variables
tower head is also active for LC 2, 4 and 5. (162) and design constraints (16293).
Deformed shape of the deck for each LC appears
in Figure 8 where the maximum vertical
displacement (1.3 m) is an active constraint for
many points in main spans (LC4 and LC5).

S5-162
40th IABSE Symposium, 19-21 September 2018, Nantes, France.
Tomorrow’s Megastructures

Figure 7. Cable positions in deck, cable areas and prestressing forces at the optimum

Figure 8. Constraints at the optimum

S5-163
40th IABSE Symposium, 19-21 September 2018, Nantes, France.
Tomorrow’s Megastructures

composite cable-stayed bridges based on


6 References genetic algorithms. Struct Infrastruct Eng
2015; 11(8):1054–68.
[1] Gimsing NJ. Cable supported bridges,
concept & design. 2nd ed. West Sussex [13] Cid C, Baldomir A, Hernández S, Cid
PO19 1UD, England: John Wiley & Songs Montoya M. Optimum size, position and
Ltd; 1997. number of cables in cable-stayed bridges.
In: HPSM/OPTI international conference on
[2] Virloguex M. Bridges with multiple Cable- high performance and optimum design of
Stayed Spans. In: Structural engineering structures and materials; 2016.
international. IABSE; 2001.
[14] Samim F, Nakamura S. Static and seismic
[3] Janjic D., Pircher M., Pircher H. Optimization characteristics of cable-stayed bridges with
of cable tensioning in cable-stayed bridges. new stay systems. In: IABSE conference
J Bridge Eng 2003; 8:131–7. geneva. structural engineering providing
[4] Hassan M., Nassef A., El Damatty A. solutions to global challenges; 2015.
Determination of optimum posttensioning [15] Shao X, Hu J, Deng L, Cao J. Conceptual
cable forces of cable-stayed bridges. Eng design of superspan partial ground-
Struct 2012; 44:248–59. anchored cable-stayed bridge with crossing
[5] Baldomir A., Hernández S., Nieto F., Jurado stay cables. J Bridge Eng 2014;
JA. Cable optimization of a long span cable- 19(3):6013001-1–06013001-5.
stayed bridge in La Coruña (Spain). Adv Eng [16] Romo J. Four spans continuous cable stayed
Softw 2010; 41(7–8):931–8. bridges without extra cables. In: Pacheco,
[6] Simões LMC, Negrão JHO. Sizing and Magalhães, editors. Multi-span large
geometry optimization of cable-stayed bridges. Taylor & Francis Group; 2015.
bridges. Comput Struct 1994; 52(2):309–21. [17] Arnaud S, Matsunaga N, N.S., J.-P. Ragaru,
[7] Simões LMC, Negrão JHO. Optimization of Behavior of a multiple spans cable-stayed
cable-stayed bridges with box-girder decks. bridge. In: Walraven, Stoelhorst, editors.
Adv Eng Softw 2000; 31:417–23. Tailor made concrete structures. Taylor &
Francis Group; 2008.
[8] Negrão JHO, Simões LMC. Optimization of
cable-stayed bridges with three- [18] Baldomir A, Tembrás E, Hernández S.
dimensional modelling. Comput Struct Optimization of cable weight in multi-span
1997; 64:741–58. cable- stayed bridges. Application to the
forth replacement crossing. In: Proceedings
[9] Martins AMB, Simões LMC, Negrão JHO.
of multi-span large bridges; 2015.
Optimization of cable forces on concrete
cable-stayed bridges including geometrical [19] Abaqus, Abaqus 6.14.2. documentation.
nonlinearities. Comput Struct 2015; 155:18– [20] MATLAB. Matlab R2016b documentation.
27.
[21] Carter M, Kite S, Hussain N, Minto B. Forth
[10] Martins AMB, Simões LMC, Negrão JHJO. replacement crossing: scheme design of the
Optimum design of concrete cable-stayed bridge. In: IABSE symposium bangkok.
bridges. Eng Optimiz 2016; 48(5):772–91. sustainable infrastructure. Environment
[11] Hassan M, Nassef A, El Damatty A, bridges friendly, safe and resource efficient; 2009.
Optimal design of semi-fan cable-stayed. [22] IAP-11. Instrucción sobre las acciones a
NRC research press. Can J Civ Eng 2013; considerar en el proyecto de puentes de
40:285–97. carretera, Ministerio de Fomento, Gobierno
[12] Hassan M, El Damatty A, Nassef A. Database de España; 2011. p. 23–5.
for the optimum design of semi-fan

S5-164

View publication stats

You might also like