1. Introduction
Cable-supported structures, such as suspension bridges, have been recognized as the most appealing structures due to their aesthetic appearance as well as the structural advantages of cables [
1,
2,
3,
4]. It is well known that cables cannot behave as structural members until large tensioning forces are induced, such as pre-stressed cable in structures [
5]. Therefore, in order to design a cable-supported structure economically and efficiently, it is extremely important to determine the optimized initial cable tensions or unstrained lengths.
Generally, designers cannot determine the initial shape arbitrarily when the cable structures are considered. The initial shape is determined while satisfying the equilibrium condition between dead loads and internal member forces, including cable tensions in the preliminary design stage because cable members display strongly geometric nonlinear behavior as well as the configuration of a cable system cannot be defined in stress-free state. The process determining the initial state of cable structures is referred to as ‘‘shape finding ’’, ‘‘form finding ’’, or‘‘ Initial shape or initial configuration’’ [
6,
7,
8,
9,
10,
11].
Until now, nonlinear analysis procedures have been developed for shape finding problems of cable bridges: the trial-and-error method [
12], the initial force method [
10,
13], the analytical and iteration method [
14,
15], the target configuration under dead loads (TCUD) related methods [
9], the optimization method [
16,
17], and the combined method [
18].
Above mentioned various form-finding approaches are generally into three categories: (1) the simplified approach; (2) the Finite Element (FE)-based approach; and (3) the analytical method.
The simplified method assumes that the load acts uniformly along the span of the main cable, which follows a parabolic shape [
2,
17,
19]. To account for a cable’s sag effect, Ernst proposed the equivalent modulus of elasticity for a parabolic cable [
20]. The simplicity of Ernst’s formula has made it widely used not only in the research field, but also for the practical designs of suspension bridges. Owing to its simplicity, this approach has been adopted by several investigators [
21,
22,
23], and has been proved to be sufficient for some cases. Namely, when a cable has relatively high stress and small length, the Ernst equivalent modulus approach could achieve a good result. However, the parabolic approximation becomes inaccurate for cables with a large sag-to-span ratio (>1/8), which experience self-weight along the length of the cable and concentrated forces from the hangers.
To improve the accuracy and facilitate nonlinear analyses of suspension bridges, various FE-based approaches have been developed. In these approaches, most of the finite element packages are still lack of suitable cable elements. A sagging cable is often simulated as two-node element, multi-node element, and curved element with rotational degrees of freedom [
24,
25,
26]. The two-node element is only suitable for modeling the cables with high pretension and small length [
27,
28], and equivalent modulus are used to account for the sag effect. For cables with large sag, a series of straight elements is used to model the curved geometry of cables. The multi-node element is based on the higher order polynomials for the interpolation functions [
29,
30]. The tangent stiffness matrix and nodal force vector are obtained while using the iso-parametric formulation. These elements give accurate results for cables with small sag. For cable element with large sag, it is necessary to use a large number of elements to model the curved geometry of cable. Therefore, it causes computational costs.
These FE-based approaches identify the target configuration of main cable via updating nodal positions and internal tension of cable elements based on nonlinear structural analysis. However, these FE-based approaches elevate the computational effort, and their convergence depends to a large extent on the assumed initial cable configuration and forces.
The alternative approach is based on exact analytical expressions for the elastic catenary, since the equilibrium configuration of a hanging cable is a catenary in nature. This method was originally proposed by O’Brien and Francis [
31] and was later extensively developed [
32,
33,
34,
35,
36]. In particular, there are various catenary-type analytical elements available, which can be used to model large sag cables in suspension bridges:
(1) Inextensible catenary elements: The cable elements adopted are infinitely stiff in the axial direction and cannot experience any increment of length. In practice, computer applications that are based on this type of element encounter severe difficulties, solving procedures tend to experience large numerical instability, causing a very difficult or even impossible convergence.
(2) Elastic catenary elements: An elastic catenary curve is defined as the curve formed by a perfectly elastic cable, which obeys Hooke’s law and has negligible resistance to bending, when being suspended from its ends and subjected to gravity. It should be noted that the conventional formulations are based on the hypothesis of small deformations, meaning that the forces are integrated with respect to the initial configuration of the catenary. Hence, the weight per unit length does not vary consistently with the elongation of the catenary. This may result in an inaccurate equilibrium of forces in the deformed configuration.
The main advantages of the catenary-type cable elements are the reduction of degrees of freedom, the simplicity of finding the dead load geometry of the cable system, the exact treatment of cable sag, the exact treatment of cable weight as it is included in the equations used for element formulation, and the simplicity of including the effect of pre-tension of the cable by simply giving the unstressed cable length. However, the cable segment equation is unsolvable when the initial three components are not set properly because of the so-called initial value sensitivity.
The purpose of this paper is to develop a catenary cable element for the nonlinear analysis of cable structures that are subjected to static loadings. Firstly, the tangent stiffness matrix and internal force vector of the element are derived explicitly based on the exact analytical expressions of elastic catenary. Self-weight of the cables can be directly considered without any approximations. The effect of pre-tension of cable is also included in the element formulation. Then, a search algorithm with the penalty factor is introduced to satisfy the convergence requirement with high precision and fast speed. Finally, numerical examples are presented and discussed to illustrate the accuracy and efficiency of the proposed analytical algorithm.
2. Segmental Catenary Theory of Main Cable
To accurately simulate the realistic behavior of main cables, the catenary element exactly considering the effects of cable sags, cable self-weight, and cable pretension is used.
2.1. Basic Equations
An elastic catenary cable element has been derived from the exact solution of the elastic catenary cable equation, deformed due to its self-weight [
32,
33]. It can be formulated in three dimensional coordinates, but only two-dimensional formulation is described in this study.
Consider a cable segment suspended between points
i(
xi,
yi) and
j(
xj,
yj), as shown in
Figure 1. It is assumed that the cable:
- (1)
is perfectly flexible and can sustain only tensile forces;
- (2)
is composed of a homogeneous material which is linearly elastic;
- (3)
is subjected to a uniform distributed load q along the cable length; and,
- (4)
the tensile stiffness of the cable is calculated using the cross-section before deformation.
The relative distances between two nodes (
i,
j) along the global
x,
y axis, are denoted as
l (
l =
xj −
xi) and
h (
h =
yj −
yi), respectively, in
Figure 1, which can be expressed as a function of the global nodal force
Hi and
Vi at the node
i as:
The force equilibriums of the elastic catenary cable require that:
Equations (1) and (2) are defined as the basic equations for segmental catenary cable, showing the relation between the segmental forces and geometric parameters. Generally, the main cable is divided into several segments (number N), each segment establishes two basic equations, in total 2 times of N equations are obtained for the whole main cable system. In Equations (1)–(3), E is the elastic modulus; A is the cross sectional area, q is the self-weight of the unstressed main cable; l represents the span length of the cable segment, h represents the elevation difference of two ends, and S0 represents the unstressed length of cable segment; Ti, Tj are the cable tension at the left (i) and right (j) ends of the cable segment, respectively; Hi and Hj are the horizontal component of cable tension at the left (i) and right (j) ends of the cable segment, respectively; and, Vi and Vj are the vertical component of cable tension of the left(i) and right (j) ends of the cable segment, respectively.
From Equations (1) and (2), it can be found that for a cable segment with determined S0, H, and Vi, the length l, and high difference h can be easily obtained; similarly, for a cable segment with determined S0, l, and h, the internal forces H and V can be easily solved. Thus, only three independent variables exist in these five variables (S0, H, Vi, l and h).
2.2. Stiffness Formulation
Following describe the procedure of stiffness formulation of the elastic catenary cable element. Considering
q,
S0,
EA as constants, partial differentiation of both sides of Equations (1) and (2) yield the following incremental relationships between the relative nodal displacements and nodal forces.
where: [
K] is the stiffness matrix due to cable shape change from end point (e.g., left end) to segment point
i; if the segment point
i become the other end point (e.g., right end), [
K] is the stiffness matrix of the main cable for the whole span;
dx,
dy are the cumulative amount of change in span and elevation respectively from end point to segment point
i; and,
dHi(k),
dVi(k) are the increment horizontal and vertical component of cable force at segment
i, respectively.
2.3. General Solution Procedure
The tangent stiffness matrix and internal force vector of cable element are determined while using an iterative procedure. This procedure requires the initial values of end forces (H, V). The iterative procedure for obtaining tangent stiffness matrix and internal force vector of cable element is briefly presented, as follows:
- (1)
input q, E, A, S0, nodes I (xi, yi) and J (xj, yj);
- (2)
calculate l0 = xj − xi, h0 = yj − yi;
- (3)
initialize end forces (H, V);
- (4)
update (l, h) using Equations (1) and (2);
- (5)
calculate incompatibility vector of relative distances ds = {dl dh}T;
- (6)
if ds is smaller than the permissible tolerances, calculate [K] using Equation (5) and internal forces using Equation (3), otherwise continue to next step;
- (7)
calculate the correction vector of end forces {dH, dV}using Equation (4);
- (8)
update the end forces Hi+1 = Hi + dH, Vi+1 = Vi + dV and go to Step (4).
2.4. No Solution Cases for Cable Segment Equation
The solution to the governing equation requires the Newton-Raphson type iteration while using initial trials of the force vector of the left node in the first cable element. However, the convergence of the gradient-based Newton-Raphson approach strongly depends on the initial value, and the estimation of initial value remains a challenge.
Generally, there are two states for numerical analysis of main cable system: one is the main cable system at finished state for the whole bridge; the other is at construction state, only the main cable installation is finished [
37]. The tension force at one end need to be assumed (or determined) for the main cable system calculation, the coordinates are iterated with convergence conditions. At the finished state, the tension force at one end and the horizontal distance between two ends are given, the unstressed cable length and the elevation between two end points can be solved, that is,
l,
Hi,
Vi are known, to solve
S0,
h. If the end tension force is assumed unreasonably, then there will be no solution for Equations (1) and (2).
To solve unstressed length
S0, Equation (1) is rewritten as:
Suppose that
l,
Hi,
EA are constants, and
EA > 0,
q > 0, 0 <
S0 < 5000 m (the length of main cable for single-span suspension bridge is currently less than 5000 m), there will be no solution for Equation (1) in the following three conditions:
Condition 1. When Vi is positive and the absolute value of Vi is large enough, l and Hi have the same sign, there will be no solution for Equation (1). It can be proved, as follows: Condition 2. When Vi is negative and the absolute value of Vi is large enough, l and Hi have the same sign, there will be no solution for Equation (1). It can be proved, as follows: Condition 3. When the absolute value of Vi is small enough, l and Hi have the same sign, Equation (14) is obtained from Equation (11), there will be no solution for Equation (1), It can be proved, as follows: