Engineering Structures 24 (2002) 1547–1559

www.elsevier.com/locate/engstruct

Non-linear shape-finding analysis of a self-anchored suspension

bridge
Ho-Kyung Kim a,∗, Myeong-Jae Lee b, Sung-Pil Chang c
a
Department of Civil Engineering, Mokpo National University, 61 Dorim-ri, Cheonggye-myeon, Muan-gun, 534 729 Jeonnam, South Korea
b
Yooshin Engineering Corporation, Yoksam-dong, Gangnam-gu, 135-936 Seoul, South Korea
c
School of Civil, Urban & GeoSystem Engineering, Seoul National University, San 56-1 Shilim-dong, Kwanak, 151-744 Seoul, South Korea

Received 6 November 2001; received in revised form 10 June 2002; accepted 25 June 2002

Abstract

A non-linear shape-finding analysis is proposed for a self-anchored suspension bridge, the Yongjong Grand Bridge, the main
cable of which is three-dimensional in shape. The proposed procedure consists of two successive steps of non-linear analysis. The
first step focuses on the cable-only system and the second on the total bridge system. A trial profile of a three-dimensional main
cable was calculated based on the conventional method utilizing a simplified force equilibrium at each node of the main cable. The
non-linear analysis is then repeated for this cable-only system until the profile converges to the target configuration that satisfies
the vertical design sag at the mid-point of the main span. Even when the best efforts were made to determine the configuration of
the cable-only system, the initial deformation of the stiffening truss is not avoidable due to the huge axial force that originates
from the main cable. Therefore, the initial frame forces are predicted and included in calculating the unbalance loads for the non-
linear analysis of the total bridge system so as to prevent dead load deformation. The possible differences between the conventional
and the proposed method are examined through numerical applications.  2002 Elsevier Science Ltd. All rights reserved.

Keywords: Self-anchored; Suspension bridge; Initial configuration; Shape-finding; Yongjong Grand Bridge, Non-linear

1. Introduction unstrained length when a finite element method is con-

sidered. For this reason, thoughtful consideration should
In the middle of 1970, Abdel-Ghaffar [1] developed be given prior to the static/dynamic analysis of cable
a suspension bridge element for the free vibration analy- structures for introducing the proper tension on each
ses of suspension bridges. A few years later, Arzouman- cable element under full dead loads.
idis [2] proposed a non-linear finite element method for More particularly, designers cannot determine the
the static and dynamic analyses of suspension bridges. initial shape arbitrarily when cable structures are con-
Since then, due to enhanced computer capability and sidered. The initial shape is determined according to the
analysis techniques, the finite element method has gradu- force equilibrium between the external dead loads and
ally come to be regarded as an attractive alternative the internal member forces. For the general tensegrity
instead of the traditional analytic method represented by structures, this process is referred to as “shape finding”
the so-called deflection theory [3]. [4,5] or “form finding” [6]. “Initial shape or initial con-
The analytic method, by solving the differential equi- figuration” [5] was also used to express the target con-
librium equations in the deformed position, naturally figuration at the initial equilibrium state under full
accounts for the stiffening effect of the tension in the dead loads.
main cable caused by dead and live loads. Similarly, the For the case of a suspension bridge, the planning of
load-displacement relationship of the cable element is the bridge pre-determines several design parameters
dependant on the introduced tension or equivalently the related to the geometry and main dimensions. The shape-
finding process should determine the coordinates of the
main cables as well as the initial tension or the
∗
Tel.: +82-61-450-2475; fax: +82-61-452-6468. unstrained length of the main cables and hangers prop-
E-mail address: hkkim@mokpo.ac.kr (H.-K. Kim). erly so that this analysis model satisfies the design para-

0141-0296/02/$ - see front matter  2002 Elsevier Science Ltd. All rights reserved.
PII: S 0 1 4 1 - 0 2 9 6 ( 0 2 ) 0 0 0 9 7 - 4
1548 H.-K. Kim et al. / Engineering Structures 24 (2002) 1547–1559

meters at the initial equilibrium state under full dead

loads.
For a typical earth-anchored suspension bridge, this
procedure is relatively simple because main cables are
fixed on immovable earth anchorages. Several studies
have examined this subject for the linear/non-linear finite
element analysis so far. Kim [7,8] established the initial
equilibrium state of a suspension bridge by non-linear
analysis utilizing the elastic catenary cable element for
the modeling of the main cables and hangers. The
unstrained lengths and nodal coordinates of the cable
elements were calculated assuming a uniform dead load
along the bridge, which led to a parabolic profile of the
main cable. Karoumi [9] utilized a trial-and-error
approach in determining the initial shape of main cables.
He used the similar cable element for the non-linear
finite element analysis of a suspension bridge. Recently,
Kim and Lee [10] also suggested an approach for analyz-
ing the target configuration of a suspension bridge by Fig. 1. The Yongjong Grand Bridge.
the Newton-Raphson method. They also used the elastic
catenary cable element for the modeling of non-linear
cables. In addition, they defined the terminology of the main cables curve three-dimensionally from the top of
“target configuration under dead loads” for referring to each tower to the outside of the girder at the main span,
the deformed shape of a structure under dead loads. and provide increased lateral stability [16,17]. The
While the above-mentioned studies utilized a rigorous bridge has double decks to carry a six-lane highway on
non-linear analysis for the shape-finding problem of sus- the upper deck and a four-lane highway and dual rail
pension bridges, a relatively simple approach was used lines on the lower deck. The upper deck of the girder
for the practical designs. Ohtsuki [11] developed a is a steel box that also functions as the upper chord of
shape-finding procedure for the design of the Yongjong the truss.
Grand Bridge. In his approach, the main cable was One of the distinguishing features of the self-anchored
assumed to be located straight between two adjacent system is the main cable anchorage [15]. This system
hangers in a piecewise manner, just as truss elements. does not require the massive end anchorages used for
The simultaneous equations were then established from the typical earth-anchored suspension bridge. Instead,
the force equilibrium at each separated node. The the main cables are anchored to each end of a stiffening
unknowns, which consisted of nodal coordinates and the girder. Few self-anchored suspension bridges have been
horizontal tension of the main cable, could be easily built because of their structural and erectional com-
obtained by putting one more compatibility condition, plexities [15]. The Konohana Bridge in Osaka, Japan,
which is related to the vertical sag of the center span, and several small bridges in Germany adopted this struc-
into the simultaneous equations related to the force equi- tural system, both for pedestrian and for highway bridges
librium. [17]. The Konohana Bridge, built in 1990, is the first
For the cable-stayed bridge, shape-finding problems modern large-scale, self-anchored suspension bridge
are closely related to the determination of the initial with a main span of 300 m. The main cable is single in a
cable force [12] rather than the cable configuration. plane and the inclined hanger system improves structural
Recently, several papers [10,13,14] demonstrated the rigidity and increases dynamic damping [18]. The ratio
shape-finding analysis of the cable-stayed bridges. of sag to span was increased to 1:6. Kamei et al. [18]
Even though general shape-finding procedures were stated that the accuracy control during erection was
suggested for typical earth-anchored suspension bridges focused primarily on the hanger tension as is the case
or cable-stayed bridges so far, they are not readily appli- for cable-stayed bridges.
cable to all the modern cable-supported bridges. One The following three issues should be considered for
example is the Yongjong Grand Bridge [15] opened to the shape-finding analysis of the Yongjong Grand
traffic at the end of 2000 in Incheon, South Korea. The Bridge. First, the three-dimensional coordinates should
Yongjong Grand Bridge is a self-anchored suspension be determined for the main cable that sags in the hori-
bridge, the structure of which is shown in Fig. 1. The zontal plane as well as the vertical plane. Secondly, for
550 m-long suspension bridge has a center span of 300 typical earth-anchored suspension bridges, the distri-
m and side spans of 125 m. The main cables of the bution of dead load is typically uniform along the bridge.
bridge take a three-dimensional shape; that is, the two In addition, the ratios of sag to span usually fall in the
 H.-K. Kim et al. / Engineering Structures 24 (2002) 1547–1559 1549

range of 1:8 or less. In these circumstances it is natural 2. Finite element model

to adopt a parabolic profile for the main cable [1]. The
Yongjong Grand Bridge is, however, a self-anchored Fig. 3 shows the major structural dimensions of the
suspension bridge and adopts a ratio of 1:5 to reduce the Yongjong Grand Bridge. Table 1 summarizes the
axial force of the stiffening truss system. In addition, the material properties used in the finite element model. In
stiffeners are concentrated around not only the anchor the prototype, the stiffening truss of the suspension
points of the main cables but also the temporary piers bridge supports one end of the approaching truss bridge
that were used to support the stiffening truss during con- to decrease the uplifting vertical component of the main
struction. For these reasons, the profile of the main cable cable tension as shown in Fig. 4. The reaction of the
cannot be parabolic in both the vertical and horizontal approaching span was determined by a structural analy-
planes. Therefore, a shape-finding analysis should be sis and applied as the counterweight force at the both
applicable for the case of non-uniformly distributed dead ends of the stiffening truss in the analysis model as
loads. Thirdly, for the case of a self-anchored suspension shown in Fig. 5.
bridge, the stiffening truss sustains the huge force of the Since the cross-section of the stiffening truss is sym-
main cables, and this results in a large axial force on the metric and the shape-finding analysis considers only
stiffening truss system. This axial forces on the stiffen- dead loads, two-dimensional analysis would be sufficient
ing truss system and the main towers result in the initial provided that the two main cables are parallel and the
deformation of a self-anchored suspension bridge as towers are H-type, which is the case with a typical sus-
shown in Fig. 2. This initial displacement of anchorages pension bridge. For the Yongjong Grand Bridge, how-
and saddles denotes, from the aspect of structural analy- ever, the profiles of the main cables are three-dimen-
sis, changes in the boundary conditions. Therefore, even sional with A-type towers. This is the reason why the
with the best analysis of initial coordinates and three-dimensional analysis model is used in Fig. 5.
unstrained lengths of cables, the initial equilibrium state The stiffening truss can be modeled in several ways
under full dead loads cannot be established when a self- from an equivalent beam element to sophisticated shell
anchored suspension bridge is being considered. elements. Considering the purpose of the current analy-
This study was motivated to develop a reasonable sis, the upper deck is modeled as an equivalent beam
shape-finding procedure for this peculiar type of suspen- element. The stiffening truss, which consists of two
sion bridge in terms of the profile and the anchoring planes in the prototype, is modeled as only one plane by
method of the main cables. The procedure consists of doubling the cross-sectional area of each chord because
two successive steps of non-linear analysis. The first step the cross-section of the stiffening truss is free from tor-
is for the cable-only system and the second for the total sional deformation. Anchor points of hangers are con-
bridge system. For the cable-only system, the trial coor- nected to the central beam by means of imaginary rigid
dinates of the main cable were calculated as was done beams for the modeling of spatial arrangement in the
in the design of the Yongjong Grand Bridge. Successive prototype as shown in Fig. 5.
non-linear analyses were then performed until the trial The end link in Fig. 4 is modeled with a beam element
nodal coordinates and unstrained lengths of cable as shown in Fig. 5. Flexural stiffness is assumed to be
elements result in the target vertical sag at the mid-point
of the main span. For the total bridge system, the initial
forces of the stiffening truss and the main towers were
calculated and included in checking the force equilib-
riums, in order to prevent dead load deformation. The
second procedure utilizes an automated iterative algor-
ithm.

Fig. 3. Structural dimensions of the Yongjong Grand Bridge. (a) front

Fig. 2. Dead load deformation of a self-anchored suspension bridge view; (b) plane view; (c) main tower; and (d) cross section of the
in shape-finding analysis. stiffening truss. (Unit: m).
1550 H.-K. Kim et al. / Engineering Structures 24 (2002) 1547–1559

Table 1
Main structural parameters

Parameter Value Remark

Dead load of stiffening truss 499.1 kN/m/Br center span

Dead load of stiffening truss 507.7 kN/m/Br side span
Dead load of main cable 31.9 kN/m/Br center span
Dead load of main cable 32.2 kN/m/Br side span
Modulus of elasticity for steel 2.06 × 108 kN/m2 stiffening truss, main tower
Modulus of elasticity for main cable 1.99 × 108 kN/m2
Modulus of elasticity for hanger 1.37 × 108 kN/m2
Area of main cable 0.1355 m2/cable
Area of hanger 0.0233 m2/cable typical
Area of hanger 0.0156 m2/cable each end
Vertical sag 60 m mid point of center span

Fig. 5. Three-dimensional finite element model of the Yongjong

Grand Bridge.
Fig. 4. Structural details of the Yongjong Grand Bridge. (a) Location
of ‘A’ and ‘B’. (b) Details of ‘A’. (c) Details of ‘B’.
along the height, the towers are divided into several
beam elements and the averages of the sectional proper-
nearly zero for the longitudinal direction of the bridge ties are used for each element. The cross beam of the
while a very large value is used for the lateral direction tower supports the stiffening truss by vertical elasto-
to simulate the restraint by wind shoes. Fig. 4 also shows meric bearings as shown in Fig. 4. These are also mod-
the longitudinal elastomeric bearing that provides the eled with beam elements similar to the modeling of end
overall stability of the bridge for longitudinal movement. links as shown in Fig. 5.
Boundary springs are used to simulate these shoes as The main cables are modeled with a series of elastic
shown in Fig. 5. catenary cable elements. Each segment of the main
Since the cross-sectional properties of the towers vary cable, separated by two adjacent hangers, is modeled by
 H.-K. Kim et al. / Engineering Structures 24 (2002) 1547–1559 1551

one cable element. Each hanger is also modeled with i (0, 0, 0) and j (lx, ly, lz) as shown in Fig. 7. The relative
one elastic catenary cable element. distances between the two nodes along the global X⫺,
Y⫺, and Z⫺axis, which are denoted by lx, ly, and lz,
2.1. Three-dimensional thin-walled beam element respectively in Fig. 7, can be expressed as a function of
the global nodal force Fi; i ⫽ 1,2,3at node i as [20]
A thin-walled beam element, shown in Fig. 6, is used
for the stiffening truss and the main tower. The degrees
of freedom represented by triple arrows denote the warp-
lx ⫽
F1 F1
L ⫺ sinh⫺1
EcAc 0 w
F3
H 再 冉冊
⫺sinh⫺1
F3⫺wL0
H 冉 冊冎
ing deformation. This element was originally formulated ⫽ f(F1, F2, F3)
and is included in the current computer code to consider
the warping deformation as well as the flexural-torsional
coupling effect for a thin-walled cross-section in which
ly ⫽
F2 F2
L ⫺ sinh⫺1
EcAc 0 w
F3
H 再 冉冊
⫺sinh⫺1
F3⫺wL0
H 冉 冊冎
the centroid and the shear center do not coincide. The ⫽ g(F1, F2, F3)
shape-finding analysis, however, causes only symmetric F3 F1 H
deformations on the cross-section of the upper deck and lz ⫽ L⫺ ⫺
EcAc 0 2EcAc w

冋再 冉 冊 冎 再 冉 冊冎册
the truss system and, as a result, warping and torsional 1 1
deformations would not be induced. Furthermore, the F3 2
2 F3⫺wL0 2
2
1⫹ ⫺ 1⫹ ⫽ h(F1, F2, F3)
upper deck consists of a box with a closed symmetric H H
cross-section in which the locations of the centroid and
(1)
the shear center are nearly identical. Therefore, the same
result can be obtained even when a conventional beam where Ec is the elastic modulus, Ac the cross-sectional
element with six degrees of freedom at each node is used area, L0 the unstrained length, H the in-plane horizontal
instead of the thin-walled beam element if the current tension, and w the self-weight per unit length of the
shape-finding analysis with prescribed modeling of the cable, respectively. The Newton-Raphson method is util-
stiffening truss system is only concerned. ized as an iterative technique to solve Eq. (1) for the
The elementwise tangential stiffness matrix KT is unknown nodal forces.
expressed as the sum of the elastic stiffness matrix KE Suppose that the two nodes io and jo are moved to i
and the geometrically non-linear stiffness matrix KG, and j, respectively as shown in Fig. 7. If the infinitesimal
which is the function of elemental nodal forces Fi; i ⫽ changes in the relative distances between the two nodes
1,2, %, 14. The 14 × 14 components of the tangential are denoted as dlx, dly, and dlz, while the infinitesimal
stiffness matrix have previously been formulated by Kim changes in nodal forces are denoted as dFi ; i ⫽
and Chang [19] based on the Updated Lagrangian For- 1, 2, 3, the following equation can be obtained by dif-
mulation. ferentiating both sides of Eq. (1):
∂f ∂f ∂f
dlx ⫽ dF ⫹ dF ⫹ dF
2.2. Three-dimensional elastic catenary cable element ∂F1 1 ∂F2 2 ∂F3 3
∂g ∂g ∂g
dly ⫽ dF1 ⫹ dF2 ⫹ dF (2)
Elastic catenary cable elements are used for the main ∂F1 ∂F2 ∂F3 3
cables and the hangers. This element is derived from ∂h ∂h ∂h
dlz ⫽ dF ⫹ dF ⫹ dF
the exact solution of the elastic catenary cable equation ∂F1 1 ∂F2 2 ∂F3 3
deformed due to its self-weight.
Consider a cable suspended between points

Fig. 6. Three-dimensional thin-walled beam element. Fig. 7. Three-dimensional elastic catenary cable element.
1552 H.-K. Kim et al. / Engineering Structures 24 (2002) 1547–1559

or in matrix notation: 3.1. Shape-finding analysis for cable-only system

冦冧 冤 冥冦 冧 冦 冧
∂f ∂f ∂f Fig. 8 shows the cable-only system. The anchor points
of the main cables and hangers are treated as the fixed
∂F1 ∂F2 ∂F3
dlx dF1 dF1 boundary. The equilibrium state of the cable-only sys-
∂g ∂g ∂g tem, if the analysis model applies the coordinates and
dly ⫽ dF2 ⫽ F dF2 (3)
∂F1 ∂F2 ∂F3 the unstrained lengths of cable elements determined in
dlz dF3 dF3 this first step of the shape-finding analysis, should satisfy
∂h ∂h ∂h
the design vertical sag of the main cable after the non-
∂F1 ∂F2 ∂F3
linear analysis.
The profile of the main cable of the Yongjong Grand
where F (3 × 3) is the flexibility matrix. The stiffness Bridge is no longer parabolic and Ohtsuki [11] proposed
matrix K (3 × 3) can be obtained by inverting F . Since a practical shape-finding analysis method as mentioned
this stiffness matrix K (3 × 3) is defined for the relative in the introduction. This method provides a preliminary
displacement between nodes i and j, the tangential stiff- profile of the cable-only system. The target configuration
ness matrix KT (6 × 6) should be formulated for the six of the cable-only system is then determined by an iterat-
degrees of freedom representing three orthogonal dis- ive procedure based on a rigorous non-linear analysis of
placements at each nodes as follows: the cable net.

KT ⫽ aTKa ⫽ 冋 K
⫺K K
⫺K
册 (4) 3.1.1. Preliminary configuration of cable-only system
[11]
The preliminary profile of the three-dimensional main
where a is the transformation matrix defined as cable, which has horizontal as well as vertical sag, can
be determined for the vertical and the horizontal plane

冤 冥
⫺1 0 0 1 0 0 separately. The following assumptions are made for this
process. 1) The hangers are inclined only in the cross-
a⫽ 0 ⫺1 0 0 1 0 (5) sectional plane of the bridge. The horizontal tension of
0 0 ⫺1 0 0 1 the main cable, therefore, remains constant along the
bridge. 2) The main cable is assumed to be straight
between two adjacent nodal points.
Fig. 9 shows a part of the main cable that is deformed
due to dead loads of the bridge. In this figure, T is the
3. Shape-finding analysis tension along the main cable, Tx, Ty, and Tz the three
components of tension T, Pi hanger force at node i.
Fig. 10 shows the projected profile of the main cable
Two successive non-linear analysis steps are sug- on the vertical X⫺Z plane for a span. The main cable is
gested for the shape-finding analysis of the Yongjong divided into N straight members in a piecewise manner
Grand Bridge. The first step is for the cable-only system if the number of hanger is N⫺1 in the span being con-
and the second for the total bridge system. The shape- sidered. The force equilibrium in X⫺Z plane can then
finding analysis for the cable-only system is intended to be expressed as
determine the three-dimensional target profile of the
di di+1
main cable. In this process, only cable elements are acti- Ti ⫽ Ti+1 , i ⫽ 1,2,%,N⫺1 (6)
vated in the finite element model. li li+1
If the profile of the main cable is once converged by
the iterative non-linear analysis, all the remaining
elements are then activated for the shape-finding analysis
of the total bridge system. For the total bridge system,
the initial force is assumed for each member consisting
of the stiffening truss and the main towers. For the pur-
pose of preventing dead load deformations, this initial
force is regarded as pre-stress and is included in check-
ing the force equilibrium while the non-linear shape-
finding analysis is performed. This process also requires
an iterative procedure until an appropriate initial force
is determined for each frame member. The details are
described below. Fig. 8. Cable-only system.
 H.-K. Kim et al. / Engineering Structures 24 (2002) 1547–1559 1553

Fig. 9. Tension components of the main cable.

Fig. 11. Force equilibrium in a cross-section of the bridge.

Based on Eqs. (7) and (8), the following can be

obtained:

Fig. 10. Force equilibrium of the main cable projected on the verti- 冉
Tx ⫺
zi⫺1⫺zi zi⫺zi+1
di
⫹
di+1
⫽ Pi 冊
zGi⫺zi
hi
⫹ WCi ⫽ WSi ⫹ WCi,
(9)
cal plane.
i ⫽ 1,2,…,N⫺1

where Ti and li are the tension and the length of the where WSi denotes the vertical component of hanger ten-
straight member between node i⫺1 and i, respectively, sion at node i which is the portioned dead weight of the
and di ⫽ xi⫺xi⫺1, in which xi is the x-coordinates of the stiffening truss supported by this hanger.
main cable at node i. By applying Eq. (6) to all members, A total of N⫺1 simultaneous equations can be
the following equation can be obtained: obtained from Eq. (9), while the unknowns are zi, i ⫽
1,2,%,N⫺1 and Tx. One more equation is, therefore,
d1 d2 dN
T1 ⫽ T2 ⫽ … ⫽ TN ⫽ Tx (7) necessary to determine the preliminary profile of the
l1 l2 lN three-dimensional main cable. The mid-point vertical sag
From Figs. 10 and 11, the force equilibrium for the verti- in the center span f provides the following constraint.
cal direction at node i can be expressed as
1
zi⫺zi⫺1 zi+1⫺zi zGi⫺zi ZN ⫽ (zN ⫹ z0) ⫹ f (10)
Ti ⫺Ti+1 ⫽ Pi ⫹ WCi, 2 2
li li+1 hi (8)
From Eqs. (7), (9) and (10), the z-coordinates and the
i ⫽ 1,2,%,N⫺1
tension of the main cable at each node can be determ-
where hi is the length of the hanger, zi the z-coordinates ined.The same procedure can be applied to the projected
of the main cable, zGi the z-coordinates of the anchor cable to the horizontal X⫺Y plane. When the X⫺Y plane
point of the hanger on the stiffening truss at node i, and is considered instead of the X⫺Z plane in Fig. 9, the
WCi the self-weight of the main cable allotted to node force equilibrium in the horizontal plane can be
i. expressed as
1554 H.-K. Kim et al. / Engineering Structures 24 (2002) 1547–1559

冉
Tx ⫺
yi⫺1⫺yi yi⫺yi+1
di
⫹
di+1 冊
⫽ Pi
yGi⫺yi
hi
(11)
앫 x⫺coordinates of the cable bands along the main
cable xi,(i ⫽ 1,%,N⫺1)
앫 coordinates of the tower saddle
yGi⫺yi 앫 distribution of dead loads
⫽ WSi , i ⫽ 1,2,%,N⫺1
zGi⫺zi
where yGi is the y-coordinate of the anchor point of the Step 1: Set the cable-only system as shown in Fig. 8.
hanger on the stiffening truss at node i. Since y0 and For y⫺ and z⫺coordinates of each nodal point, use the
yN in Eq. (11) are given parameters, the y-coordinates y⫺ and z⫺coordinates of the preliminary configuration
can be obtained using the previously determined Tx obtained in Eqs. (9)–(11). Note that x⫺coordinates are
given design parameters, not unknowns.
3.1.2. Target configuration of cable-only system Step 2: Calculate the unstrained length L0 of each
One of the possible approaches to suspension bridge cable element. According to Eqs. (1)–(4), the tangential
analysis is using a truss element for the modeling of the stiffness matrix of the elastic catenary cable element is
segmented main cables and hangers. In this case, the the function of relative distances between end nodes lx,
elementwise stiffness matrix of the pre-tensioned truss ly, lz and unstrained length L0. The relative distances can
element can be formulated with the member length and be calculated with the current nodal coordinates but the
the initial tension, which can be determined in Eqs. (9)– unstrained length remains unknown and should be
(11). In general, linear analyses are then performed for determined prior to structural analysis. If at least one
design loads using the assembled tangential stiffness component of the main cable tension is known, the
matrix even though the preliminary configuration in the unstrained length L0 can be determined. As the x⫺
last section shows a certain discrepancy with the target component of main cable tension remains constant along
configuration due to assumption 2). the bridge, F1 in Eq. (1) is constant for all the main cable
In fact, Kim and Chang [21] showed that live load elements at the initial equilibrium state. The Tx, which
responses are not sensitive to the accuracy of the initial was determined in Eqs. (9) and (10) as the result of the
equilibrium state of a suspension bridge. The preliminary preliminary analysis, can be used as the target compo-
configuration, therefore, may be sufficient for general nent of tension in calculating the unstrained length by
live load analyses. the Newton-Raphson method. The following are the
For the backward construction step analysis, however, computational algorithms used to calculate the
geometrically non-linear analysis is indispensable due to unstrained length L0 for each main cable element.
the large deformation of the main cable. The accuracy
of the initial equilibrium state may play a more
important role in this case because not only for guaran- 1. Calculate lx, ly, and lz from the preliminary configur-
teeing a more reliable configuration of the cable but also ation.
for freeing the analysis model from the unbalanced load 2. Assume l, L0, F2, and F3 using F1 ( ⫽ Tx) as follows.
prior to the backward analysis. The considerable change
of configuration during construction can then be exam- 2F1
ined starting from this non-linear-analysis-based initial l⫽⫺
wlx
equilibrium state.

冉 冊
For this reason, this study suggests an iterative pro-
冪
2
sinhl
cedure that is capable of upgrading the preliminary con- L0 ⫽ (l2x ⫹ l2y ) ⫹ l2z
figuration by the geometrically non-linear analysis. The l
converged configuration is referred to the ‘target con-
figuration’ of the cable-only system compared to the pre- wly
F2 ⫽ ⫺
liminary configuration. 2l
The following are the detailed computational algor-
ithms used to determine the target configuration. It
should be noted that the following parameters are given
prior to the shape-finding analysis of the cable-only sys-
F3 ⫽ ⫺ lz 冉
w coshl
2 sinhl
⫹ L0 冊
tem: 3. Calculate updated l⬘x, l⬘y, and l⬘z in Eq. (1) using current
l, L0, F2, and F3.
앫 vertical sag of the main cable in the center span 4. Calculate the incompatibility vector of relative dis-
앫 anchoring positions of the main cables on the stiffen- tances {ds} ⫽ {dlx dly dlz}T wheredlx ⫽ lx⫺l⬘x,
ing truss dly ⫽ ly⫺l⬘y, and dlz ⫽ lz⫺l⬘z.
앫 anchoring positions of each hanger on the stiffening 5. If {ds} is within the tolerance, current L0 is the
truss yGi, zGi,(i ⫽ 1,%,N⫺1) unstrained length. Otherwise, continue to next step.
 H.-K. Kim et al. / Engineering Structures 24 (2002) 1547–1559 1555

冤 冥
6.
∂f ∂f ∂f
∂F2 ∂F3 ∂L0
∂g ∂g ∂g
Calculate [C] ⫽ by differentiating
∂F2 ∂F3 ∂L0
∂h ∂h ∂h
∂F2 ∂F3 ∂L0
Eq. (1).
7. Calculate {dF2 dF3 dL0}T ⫽ [C]⫺1{ds}.
8. Update F2, F3, and L0 as follows and go to (3).
F2 ⫽ F2 ⫹ dF2, F3 ⫽ F3 ⫹ dF3, and L0 ⫽ L0 ⫹ dL0

For the hanger elements, the same procedure can also

be utilized in calculating the unstrained length if the
y⫺component of hanger tension Pi in Fig. 11 is used as
target component of cable tension instead of Tx.
Step 3: Perform the non-linear analysis for the cable-
only system. If the tried profile of the main cable shows
some discrepancy with the unknown target configur-
ation, the cable system will deform due to unbalanced
forces until equilibrium is satisfied.
Step 4: Check the convergence of the profile of the
main cable. If the displacements in all nodes are within Fig. 12. Iterative procedure to determine the configuration of the
the allowable limit, this profile can be regarded as the cable-only system.
target configuration of the cable-only system. In this
case, the shape-finding analysis for the cable-only sys- so that the analysis model at the initial equilibrium state
tem is finished. maintains the pre-determined cable configuration above.
Step 5: Update the nodal coordinates of the main The introduction of the initial force on the frame
cable. The nodal coordinates can be replaced with the elements, which consist of the stiffening truss and the
deformed profiles of main cable obtained as the result main towers, can solve this problem. The resultant
of the non-linear analysis in Step 3. Note that the vertical unbalance load could be zero for each degree of freedom
mid-point sag f is the pre-determined design parameter. if the initial forces are determined properly. Kim and
Therefore the z⫺coordinates of the corresponding node Chang [21] introduced this concept previously but the
need not be updated in this step. Go to step 2 and repeat initial force was determined manually. To minimize such
the procedure. Fig. 12 illustrates this procedure in a efforts in calculating the initial force as well as to
flowchart. increase the accuracy of the initial equilibrium state, an
automated iterative procedure is developed and utilized
in the computer code.
3.2. Shape-finding analysis for total bridge system Fig. 13 illustrates the main sources that can cause

The configuration of the main cable was determined

previously based on the principle that the cable ends
should remain on each anchoring position. This was ach-
ieved by replacing each anchorage with a fixed support
during the shape-finding analysis of the cable-only sys-
tem. If the position of any anchorage moves excessively
during the shape-finding analysis for the total bridge sys-
tem, the initial equilibrium state cannot be established
properly no matter how precisely the coordinates and
unstrained lengths of cable segments were determined.
The stiffening truss of a self-anchored suspension
bridge sustains the huge force of the main cable. As pre-
viously mentioned with regard to Fig. 2, this axial force
would introduce the initial deformation of the stiffening Fig. 13. Initial force introduced in the stiffening truss and main
truss. This dead load deformation should be prevented tower.
1556 H.-K. Kim et al. / Engineering Structures 24 (2002) 1547–1559

initial forces in the stiffening truss and the main towers. introduced in each frame element while the configuration
The superstructure generally has a longitudinal grade of the main cable remains unchanged. Once the initial
and adopts a somewhat complicated structural system. In forces in the beam elements converge, the finally
fact, the deck-stiffening system of the Yongjong Grand attempted analysis model becomes the unbalanced load-
Bridge consists of upper box girder and truss. Even free target structure of a self-anchored suspension
though the cable force acting on each anchoring point is bridge. This initial equilibrium state provides important
known, it is impossible to determine all the resultant information such as a reasonable configuration of a
initial forces introduced in this type of complicated deck- three-dimensional main cable, the tension and unstrained
stiffening system. For this reason, this study suggests length of the main cables and hangers, and the expected
a trial-and-error procedure, which utilizes a non-linear member force of the stiffening truss when the bridge is
structural analysis until the proper initial forces are constructed as designed. Fig. 14 illustrates this procedure
determined. in the form of a flowchart.
The followings are details of the computational algor-
ithm.
Step 1: Remove all fixed supports of the cable-only 4. Numerical examples
system in Fig. 8.
Step 2: Activate all the remaining members in the A shape-finding analysis was performed for the
analysis model, which consists of the stiffening truss and Yongjong Grand Bridge, in order to verify the feasibility
the main towers. Apply appropriate boundary conditions and effectiveness of the proposed procedure. This
and activate the connecting elements that simulate end example mainly examined two principal issues. The first
links and elastomeric bearings. Apply full dead loads of is how much the configuration of main cable can be
the structure as well. Note that the current shape is improved from the preliminary configuration when a rig-
nothing but the target configuration of the total bridge orous non-linear analysis is applied. The second is how
system if the force equilibrium is not of concern. Since
a self-equilibrating structure, however, is the main issue
of this study, the following iterative procedure is
intended to determine the analysis model that satisfies
the force equilibrium condition while maintaining the
current shape.
Step 3: Calculate the unbalanced load by comparing
the external dead loads and the internal member forces.
The internal member forces mainly consist of cable ten-
sions and initial forces in beam elements introduced in
the previous step. If this is the first iteration, initial forces
in beam elements are zero.
Step 4: Perform a non-linear analysis of the total
bridge system for the unbalanced load in Step 3. Calcu-
late the displacements for all degrees of freedom and the
internal resultant forces for all elements.
Step 5: Check the convergency of the structural
geometry. If the displacements in all nodes are within
the allowable limit, this analysis model can be regarded
as the target structure in self-equilibrium. In this case,
the shape-finding analysis for the total bridge system
is finished.
Step 6: Save the current resultant forces as the initial
forces on corresponding beam element for the next iter-
ation.
Step 7: Restore the initially tried coordinates in Step
2 for all structural nodes. Go to Step 3 and repeat the
procedure.
As the iterative procedure advances, the internal
resultants of the beam elements would come into better
balance with the other sources of force components,
which originate from the cable force and the dead load.
After a few attempts, the converged initial forces can be Fig. 14. Shape-finding analysis for the total bridge system.
 H.-K. Kim et al. / Engineering Structures 24 (2002) 1547–1559 1557

much deformation occurs in the bridge if initial forces regarding the vertical sag of the base strand in the center
are not introduced in frame members. span during the air-spinning construction. In this con-
struction stage, the two main cables are temporarily par-
4.1. Shape-finding analysis of cable-only system allel in the center span and suspended freely due to their
own weight. The configuration of the main cable con-
Table 2 sets the target configuration against the pre- verges to the target configuration as the hangers are
liminary configuration of the three-dimensional main installed. In general, the admittable tolerance in this
cable. Each line in Table 2 shows the three-dimensional stage would be larger than that for the base strand.
coordinates at each cable band. Note that the x-coordi- As the railway parts are not yet constructed, the
nates of each cable band are given parameters prior to Yongjong Grand Bridge is currently opened to only
the shape-finding analysis. The z-coordinates are related highway traffic. According to an unofficially reported
to the vertical position of the cable bands, while the y- surveying result, after the construction of the bridge
coordinates are related to the horizontal position. Note except for the railway parts, the levels of two main
that the coordinates for the anchor points of main cables cables at the sag point showed upward differences of
and the tower saddles are given parameters. The vertical 15 cm and 16 .7cm compared to the target level. Here,
sag of the main cable is also given while the horizontal the target level implies the adjusted level of the main
sag is unknown. The differences for such given points cable at the sag point calculated by removing the corre-
are therefore zero between the preliminary and the tar- sponding dead loads from the analysis model.
get configuration. It is, however, the author’s opinion that a direct com-
The difference for the vertical position, however, parison with this raw field result should be more careful.
increases over 9 cm around one-sixth point of the center The field test result generally requires a calibration pro-
span. The y-coordinates at the mid-point of the center cess considering possible environmental factors such as
span, in other words, horizontal sag, can be improved temperature. Nevertheless the surveying result provides
by 6 cm when the proposed non-linear analysis is meaningful insights. Possible construction errors remain
applied. The maximum difference for the horizontal pos- in the range of the same order with the upgraded
ition amounts to 8.7 cm between the preliminary and the amounts of configuration according to the proposed non-
target configuration. linear analysis. Accordingly, a rigorous analysis would
For reference, the Special Specification [22] for the be preferable to the shape-finding analysis of this type
Yongjong Grand Bridge admitted a ± 3cm tolerance of cable structure, utilizing every possible means.

Table 2
Calculated nodal coordinates of the cable-only system

X (m) Y (m) Y (m) Difference Z (m) Z (m) Difference Remark

preliminary final (mm) preliminary final (mm)
configuration configuration configuration configuration

6.250 17.429 17.429 0 ⫺48.062 ⫺48.062 0 anchor point

25.000 15.711 15.693 18 ⫺53.479 ⫺53.462 17
37.500 14.510 14.478 32 ⫺58.145 ⫺58.111 34
50.000 13.175 13.132 43 ⫺63.660 ⫺63.609 51
62.500 11.679 11.629 50 ⫺70.024 ⫺69.956 68
75.000 10.008 9.957 51 ⫺77.236 ⫺77.157 79
87.500 8.158 8.110 48 ⫺85.297 ⫺85.215 82
100.000 6.125 6.086 39 ⫺94.207 ⫺94.134 73
112.500 3.906 3.883 23 ⫺103.960 ⫺103.917 43
125.000 1.500 1.500 0 ⫺114.573 ⫺114.573 0 saddle
137.500 3.661 3.632 29 ⫺104.980 ⫺104.938 42
150.000 5.637 5.586 51 ⫺96.239 ⫺96.160 79
162.500 7.428 7.360 68 ⫺88.322 ⫺88.230 92
175.000 9.034 8.955 79 ⫺81.239 ⫺81.147 92
187.500 10.454 10.369 85 ⫺74.989 ⫺74.905 84
200.000 11.688 11.601 87 ⫺69.572 ⫺69.502 70
212.500 12.735 12.650 85 ⫺64.989 ⫺64.935 54
225.000 13.594 13.514 80 ⫺61.239 ⫺61.201 38
237.500 14.264 14.191 73 ⫺58.322 ⫺58.300 22
250.000 14.745 14.678 67 ⫺56.239 ⫺56.228 11
262.500 15.034 14.973 61 ⫺54.989 ⫺54.985 4
275.000 15.131 15.071 60 ⫺54.573 ⫺54.573 0 center
1558 H.-K. Kim et al. / Engineering Structures 24 (2002) 1547–1559

4.2. Shape-finding analysis of total bridge system tiveness of the proposed method. Deformations at sev-
eral check positions were successfully suppressed. This
Based on the profile of the main cable determined indicates that the configuration of the main cable at this
above, this section continues the shape-finding analysis state conforms precisely to the target configuration of
for the total bridge system. Two different analyses were the cable-only system. The compressive frame forces are
performed to examine the necessity of introducing initial also summarized in Table 3 for the mid-point of center
forces in the stiffening truss and the main tower for this and side spans, respectively. These frame forces are
type of a self-anchored suspension bridge. The first nothing but the axial components of the iteratively intro-
analysis case does not consider initial forces in any duced initial forces. The current analysis model, there-
frame member, while the second case introduces the fore, satisfies not only the target configuration but also
appropriate initial forces in all frame members. the self-equilibrium between the full dead loads and
For the first case, even though the unstrained length internal member forces.
and nodal coordinates for each segment of the cable were
rigorously calculated in the previous example, defor-
mations cannot be avoided over the entire bridge due to 5. Conclusions
the huge axial force arising from the main cables. As
shown in Table 3, the tower saddle deforms toward the A non-linear shape-finding analysis is proposed for a
center span by almost 8.7 cm, while the anchor point of self-anchored suspension bridge, the Yongjong Grand
the main cable on the stiffening girder moves a distance Bridge, of which the main cable adopts a three-dimen-
of over 6 cm toward the center span. These support sional profile. For the cable-only system, the preliminary
movements result in a change in the tension and con- configuration of the main cable, which is generally used
figuration for the cable system. According to Table 3, for the purpose of design, is upgraded by an iterative
the vertical sag shows a 15.6 cm discrepancy compared non-linear analysis to obtain the target configuration.
to the target. In addition, the horizontal tension of the The unstrained lengths of main cables and hangers are
one main cable decreases by the amount of 3,335kN, also rigorously calculated using the elastic catenary
which is 6.7% of the target tension. cable element. According to numerical examinations, the
For the second case, however, the initial forces are upgraded amounts of the configuration according to the
introduced in frame members. This procedure utilizes an proposed procedure remain within the meaningful range
iterative technique as previously stated. After a few iter- of geometry control in construction. For the total bridge
ative calculations, the target configuration is determined system, by introducing the proper initial force in frame
for the total bridge system. Table 3 illustrates the effec- members, deformations at several check positions can be

Table 3
Results of the shape-finding analysis for the total bridge system

Parameter Without introducing initial With introducing initial

frame force frame force

Horizontal displacement at both ends of stiffening truss (cm)a 5.623 0.000

Vertical displacement at both ends of stiffening truss (cm)a 0.093 0.000
Horizontal displacement at top of tower (cm)a 8.693 0.000
Vertical displacement at top of tower (cm)a 6.335 0.000
Horizontal displacement at anchor of main cable (cm)a 6.025 0.000
Vertical displacement at anchor of main cable (cm)a 0.091 0.000
Vertical displacement at mid-point of center span (cm)a 22.000 0.000
Vertical sag of main cable at center span (cm)b 60.156 (⌬f = 0.156) 60.000 (⌬f = 0.000)
Horizontal sag of main cable at center span (cm)c 13.634 13.631
Horizontal tension of main cable (kN/cable)c 46,206~46,281 49,541
Tension of hanger at each end of bridge (kN/cable)c 3,638 3,917
Tension of hanger in side span (kN/cable)c 2,784~3,024 3,194~3,266
Tension of hanger in center span (kN/cable)c 2,753~2,997 3,185~3,218
Compressive force on upper deck at mid-point of center span (kN)c 88,192 81,835
Compressive force on lower chord at mid-point of center span (kN/Br)c 783 15,304
Compressive force on upper deck at mid-point of side span (kN)c 69,553 83,120
Compressive force on lower chord at mid-point of side span (kN/Br)c 17,285 15,539

a
The target values for these parameters are all zero as designed.
b
The vertical sag of the main cable is 60 m as represented in Table 1.
c
These parameters are determined as the results of the shape-finding analysis.
 H.-K. Kim et al. / Engineering Structures 24 (2002) 1547–1559 1559

successfully suppressed. The proposed analysis model, sidering construction stages. Doctorial dissertation, Seoul
therefore, represents a theoretically possible target National University, Seoul, 1993.
[9] Karoumi R. Some modeling aspects in the non-linear finite
bridge structure in self-equilibrium of the dead loads and element analysis of cable-supported bridges. Comput Struct
the internal resistant member forces. 1999;71:397–412.
[10] Kim K-S, Lee HS. Analysis of target configurations under dead
loads for cable-supported bridges. Comput Struct
2001;79:2681–92.
Acknowledgements [11] Ohtsuki S. Lecture and private discussions, 1996.
[12] Chen DW, Au FTK, Tham LG, Lee PKK. Determination of initial
cable forces in prestressed concrete cable-stayed bridges for given
The authors are particularly grateful to Dr. Ohtsuki, design deck profiles using the force equilibrium method. Comput
of Chodai Co. in Japan, for very valuable lectures and Struct 2000;74:1–9.
[13] Wang P-H, Yang C-G. Parametric studies on cable-stayed
private discussions on his configuration analysis applied bridges. Comput Struct 1996;60(2):243–60.
for the design of Yongjong Grand Bridge. [14] Wang P-H, Lin H-T, Tang T-Y. Study on non-linear analysis of
a highly redundant cable-stayed bridge. Comput Struct
2002;80:165–82.
[15] Gil H, Cho C. Yong Jong Grand Suspension Bridge, Korea.
References Struct Engrg Int 1998;8(2):97–8.
[16] Kwon SD, Chang SP, Kim YS, Park SY. Aerodynamic stability
of self-anchored double deck suspension bridge. J Wind Engrg
[1] Abdel-Ghaffar AM. Dynamic analyses of suspension bridge and Ind Aerodyn 1995;54/55:25–34.
structures. California Institute of Technology Earthquake Engin- [17] Ochsendorf JA, Billington DP. Self-anchored suspension bridges.
eering Research Laboratory, EERL 76-01, California, 1976. J Bridge Engrg 1999;4(3):151–6.
[2] Arzoumanidis SG. Finite-element analysis of suspension bridges. [18] Kamei M, Maruyama T, Tanaka H. Konohana Bridge, Japan.
Doctorial dissertation, Columbia University, 1980. Struct Engrg Int 1992;2(1):4–6.
[3] Steinman DB. A practical treatise on suspension bridges. New [19] Kim C-Y, Chang S-P. Coupled flexural-torsional analysis of
York: Wiley, 1953. cable-stayed bridge subjected to earthquake loading. In: Proceed-
[4] Hanaor A. Prestressed pin-jointed structures-flexibility analysis ings of the First KOREA-JAPAN Joint Seminar on Steel Bridges,
and prestress design. Comput Struct 1988;28:757–69. Seoul, 1990, pp. 119–130.
[5] Murakami H. Static and dynamic analyses of tensegrity struc- [20] Chang S-P, Park Y-Y. Dynamic analysis of the cable-suspended
tures. Part II. Quasi-static analysis. Int J Solids Struct roof system by using the three-dimensional cable element. In:
2001;38:3615–29. Proceedings of the Second KOREA-JAPAN Joint Seminar on
[6] Motro R. Tensegrity systems and geodesic domes. Int J Space Steel Bridges, Tokyo, 1992, pp. 203–216.
Struct 1990;5:341–51. [21] Kim H-K, Chang S-P. Determination of initial configuration of
[7] Kim HK, Chang SP. Motion of suspension bridge subject to mov- a self-anchored suspension bridge by introducing initial axial
ing load using finite element method. In: Proceedings of the force. J Korean Society of Civil Engineers 1997;17(1–2):197–
Second KOREA-JAPAN Joint Seminar on Steel Bridges, Tokyo, 205 in Korean.
1992, pp. 183–196. [22] New Airport Hiway Co Ltd. Special Specification for the
[8] Kim H-K. Three dimensional analyses of suspension bridges con- Yongjong Grand Bridge, Incheon, Korea, 1998.