Académique Documents
Professionnel Documents
Culture Documents
Article information:
For Authors
If you would like to write for this, or any other Emerald publication, then please use our Emerald for
Authors service information about how to choose which publication to write for and submission guidelines
are available for all. Please visit www.emeraldinsight.com/authors for more information.
The current issue and full text archive of this journal is available at
www.emeraldinsight.com/0264-4401.htm
Slack cables
in the FDM
Miltiades Elliotis
Department of Mathematics and Statistics, University of Cyprus,
Nicosia, Cyprus
1011
Received 20 March 2012
Revised 2 August 2013
Accepted 7 August 2013
Abstract
Purpose The purpose of this paper is to present a solution strategy for the analysis of cable
networks which includes an extension to the force density method (FDM) in an attempt to support
cable elements when they become slack. The ability to handle slack cable elements in the analysis is
particularly important especially in cases where the original cable lengths are predefined, i.e. the cable
structure has already been constructed, and there is a need for further analysis to account for
additional loading such as wind. The solution strategy is implemented in a software application.
Design/methodology/approach The development of the software required the implementation of
the FDM for the analysis of cable networks and its extension to handle constraints. The implemented
constraints included the ability to preserve the length in the stressed or the unstressed state of
predefined cable elements. In addition, cable statics are incorporated with the development of the cable
equation and its modification to be able to be handled by the FDM .
Findings The implementation of the solution strategy is presented through examples using the
software which has been developed for these purposes.
Originality/value The results suggest that for cable networks spanning large distances or cable
elements with considerable self-weight the neglect of the cable slackening effects is not always conservative.
Keywords Analytical solutions, Cables, Large deflection, Nonlinear analysis, Slack cables
Paper type Research paper
Nomenclature
xr, yr, zr
D
D*
GT
Q
qn 1
qn
Dq
lu
Lu
M
Engineering Computations:
International Journal for
Computer-Aided Engineering and
Software
Vol. 31 No. 5, 2014
pp. 1011-1030
r Emerald Group Publishing Limited
0264-4401
DOI 10.1108/EC-03-2012-0054
EC
31,5
U; V; W
1012
mni
n
li
Ti
Ai
Ei
hi
1. Introduction
Cables distinguish themselves in that they can be packaged in a compact volume
and yet retain the same structural properties as rod elements when extended. They
function as tension truss members when in a stretched state but cannot support any
compressive forces once they become slack. The applications which incorporate cable
elements are therefore very interesting and knowledge of their analysis and behavior is
a challenge to the structural engineer. This paper is concerned with the analysis of
cable networks which contain cables that can become slack, either during the iteration
procedure of a nonlinear scheme for the determination of the initial equilibrium state,
or because of the application of additional loading such as wind. The presented work
includes an extension to the Force Density Method (FDM) in order to handle cables
which may become slack under compressive loads.
2. Background
The analysis of cable networks is not always easy since traditional methods of
structural analysis such as the direct stiffness method cannot be applied without any
nonlinear modification. The difficulties in the analysis arise from both the material and
geometric nonlinearities of cable networks. The material nonlinearity is due to cable
slackening, while the geometric nonlinearity stems from the large displacements and
geometric stiffness due to the applied loads (Mitsugi, 1994). A cable network is different
from a conventional structure, such as a space truss, in that its initial shape is
stabilized by the prestress in the cables. Analysis methods of such structures include
shape finding and loading analysis. Shape finding of a cable network is a process of
obtaining the initial equilibrium configuration or initial shape, given a set of
requirements. These requirements may consist such aspects as the connectivity
pattern of the structure, its approximate overall dimensions, the support reactions and
the initial stresses in the cables. Loading analysis of a cable network is the process of
finding the member stresses and structural deflections under external loads with
respect to the initial shape.
Cable networks can be analyzed using any method developed for other structures
including the finite element method as presented by Varum and Cardoso (2005).
However, the material and geometric nonlinearities create numerical difficulties in
solving the equations created from the analysis of such structures. Several methods
were developed to overcome the aforementioned difficulties and solve the required
equations including numerical algorithms for computer programming in order to
evaluate simpler three dimensional structures (e.g. Hangai and Wu, 1999; Volokh et al.,
2003). These methods can be divided in two major categories based on the initial
assumptions of the analysis. The first category of analysis methods, considers the initial
stress-free state as a reference. Methods in this category include various iterative finite
element methods, beam analogy of the cables, as well as the use of nonlinear constitutive
equations to mathematically denote slackening. In order to obtain the solution Buchholdt
and McMillan (1971) and Buchholdt (1985) presented several minimization techniques
such as minimization of the total potential energy. Mitsugi and Yasaga (1991) presented a
simple scheme for cable analysis in which the exact axial strain and a nonlinear
constitutive equation of cable elements were used. The scheme turned out to be quite
useful because the initial cable length can be specified exactly for the actual hardware
fabrications. This method requires the cable lengths and tensions in the initial stress-free
state as input for the analyses. Similar methods are employed by Yamamoto (1990) and
by Miura and Miyazaki (1990). These methods, however, show some problems when
detailed and accurate analyses are required. Mitsugi (1994) presents a nonlinear static
cable analysis method that avoids all the above problems. In his paper the equilibrium
equation which virtually provides the unbalanced force at the designated variables and
the stiffness matrix at the current configuration are derived for the Newton iteration.
Cables under compressive strain are canceled at the integration for the global equilibrium
to account for the effect of slackening. Krishna (1978) has used the Newton-Raphson
method and its variations, such as reducing load increments, to solve the nonlinear
stiffness equations. Herbert and Bachtell (1986) introduced the Green-Langrange strain of
a cable and solved for the tension state in a cable network in conjunction with the
displacement fields using Newtons method of iteration. The method successfully
eliminates rigid body motions of the cable elements but the effects of cable slackening are
not explicitly mentioned. Ren (1999) accounted for the cable slackening by considering an
equivalent straight chord member with an equivalent modulus of elasticity.
When the final geometry and tension in the equilibrium state in a cable network is
known, it is difficult to back-calculate the initial lengths and tensions in the cables at the
loading free state. The presence of large deformations, even if the strains are small, adds to
the complication of the calculations. When the cables constitute a complicated network,
which is a typical case, the initial values are even more difficult to be prescribed. Modeling
initial conditions, e.g. initial slack ratio and pre-tension, of a cable is important in nonlinear
structural analysis of complex cable networks, which may have significant effect on the
accuracy in the numerical prediction of a final shape of cable networks subjected to
external loading. Schek (1974) addressed this modeling challenge by using optimization of
the initial configurations of the cable network. The difference between the proposed shape
by an architectural viewpoint, and the equilibrium shape under the loading condition, and
the differences between states of real tensions and prescribed tensions by a designer for
some of the elements, are optimized. Optimization is achieved by treating the force
densities (FD), defined by internal tension of each element divided by its length, as
independent variables. The coordinates of the free nodes are eliminated from the
optimization problem by introducing the FD in the equilibrium equations. The
formulation is efficient when determination of a shape close to the proposed shape under
the major loading is the sole target of the optimization problem.
3. FDM
The implementation of the FDM as described by Schek (1974) is presented in the
following paragraphs. The linear FDM is presented first, in 3D Cartesian coordinates,
together with a two dimensional (2D) example of a cable network, followed by the
nonlinear extension to it to allow for external constraints. This presentation is
Slack cables
in the FDM
1013
EC
31,5
1014
considered necessary to familiarize the reader with the FDM and to aid to the
understanding of the implementation of the slack cable element which is the main
concept used in this study.
Force equilibrium equations in 3D Cartesian coordinates
Schek (1974) managed to transform the analysis of general networks into a linear
problem. He only considered that the final network shape should be in an equilibrium
state. The linearization is accomplished by assigning force-to-length ratios to each
member of the network. These force-to-length ratios proved to be good parameters for
the description of the network and they are referred to as the FD. Any equilibrium state
can be obtained by solving a linear set of equations which are expressed as functions of
the FD and the external loading. The solution that is provided by the method consists
of the final coordinates of the free nodes and a set of tensions in the cables that are
required to maintain the cable network in equilibrium.
The FDM essentially solves similar equations for each cable network under
consideration. Obviously the number of equations increases for large cable networks
which makes the use of the computer necessary. The FDM can be classified as a matrix
method since the use of matrices for the solution of large systems of equations is
essential. The following sections provide a description for the basic matrices that are
used as well as the equilibrium equations which are expressed in terms of these matrices.
Branch node matrix [C,C*]. The first step in the FDM is to define the network
connectivity. For this purpose the branch node matrix, as defined by J.H. Argyris (1964)
and by Fenves and Brannin (1963), is incorporated. The branch node matrix consists of
m-rows, equal to the number of branches (cable elements) in the network and ncolumns, equal to the number of nodes. The matrix is subdivided into two regions. The
C part of the matrix contains the columns which represent the free nodes while the C*
part contains the columns which represent the fixed nodes. The branch node matrix is
a graph of the cable network denoting which branches are connected to which nodes.
Therefore, for branch i, spanning between nodes j and k, the entries in the branch
node matrix are as follows:
*
ci; j or k
1
1
0
+
1
where C(i,j or k) is the coefficients in the branch node matrix; i indicates the row number
(cable number); j or k indicates the column number (node number). Once the branch
node matrix is defined then it can be used to calculate the projections of the branches
on the coordinate axes (Schek, 1974):
Dx u Cxr C xf
Dy v Cyr C yf
Dz w Czr C zf
9
Dx; Dy; Dz =
where
or
is the vectors that contain the projections of the branches on the
;
u; v; w
coordinate axes; xr, y r, zr the vectors that contain the coordinates of the released nodes;
and xf, y f, zf the vectors that contain the coordinates of the fixed nodes.
According to Schek (1974) the equilibrium equations at each joint can be expressed
in matrix form as shown in the following equation:
Slack cables
in the FDM
C T UL1 t px
C T VL1 t py
T
1
C WL t pz
1015
where U,V,W is the diagonal matrices which contain the vectors u, v, w, respectively;
L1 the diagonal matrix which contains the inverse of the lengths of the branches; t the
vector which contains the tensions in the cable elements; and px, p y, pz the vectors with
the applied nodal loads.
Substituting with L1 t q, Uq Qu, Vq Qv, Wq Qw and the element
projections (u, v, w) on the axes in the above equations, the equilibrium equations can
be further modified to take their final form which is shown in the following equation
(Schek, 1974):
C T QCxr C T QC xf px
C T QCyr C T QC yf py
T
C QCzr C QC zf pz
where Q is the diagonal matrix which contains the FD of the branches.
Force density matrix (D). The product CTQC is a square matrix which is defined as the
force density matrix, D. The force density matrix has number of rows and columns equal
to the number of free nodes. It turns out that the D matrix can be assembled directly
instead of multiplying the individual matrices. Its entries consist of the FD of the branches
connected to the free nodes and its form depends on the branch node matrix which is
defined above. The entry D(j,j) on the diagonal is equal to the sum of the FD of the branches
that are framing into node k; the off-diagonal term D(k,j) is equal to the negative value of the
FD of the branch connecting nodes k and j. The D matrix in the FDM is analogous to
the stiffness matrix in the direct stiffness method. In fact, it is actually equivalent to the
linear geometric stiffness matrix for the given structure. Therefore, it is not surprising that
the two matrices share the same properties; they are both symmetric and positive definite.
In the same manner the two matrices can be assembled in a similar way.
Fixed force density matrix (D*). The product CTQC* is defined as the fixed force density
matrix, D*. The D* matrix has number of rows equal to the number of free nodes and
number of columns equal to the number of fixed nodes. Its entries consist of the negative
values of the FD of the branches that connect the free node in the corresponding row to the
fixed node in the corresponding column. For example, the term D*(k,j) equals the negative
value of the FD of the branch that connects the free node k, to the fixed node j.
Once D* is obtained the form of the structure (fixed and free nodes) is defined. Then
an equilibrium shape exists for a given set of FD in each branch and a given set of
point loads applied to the released (free) nodes. The equilibrium equations are solved
for the unknown vectors xr, yr, zr as shown in the following equation (Schek, 1974):
xr D1 px D xf
yr D1 py D yf
zr D1 pz D zf
EC
31,5
1016
cable network using li Dx2i Dy2i Dz2i . The element forces can be calculated
using the set of original FD and the equation t Lq. Once the equilibrium state of the
network is obtained, the original unstrained length of each member can be determined
by using the principles of elasticity theory. These are the original lengths of the
branches to be used during the construction period.
Dy1
Dy2
T1
T2 PyA
l1
l2
Dx2
Dx3
T2
T3 PxB
l2
l3
Dy2
Dy3
T2
T3 PyB
l2
l3
where Dx1, Dx2, Dx3 is the difference in X-coordinates for each cable; Dy1, Dy2, Dy3 the
difference in Y-coordinates for each cable; l1, l2, l3 the cable length; T1, T2, T3 the cable
tension; PxA, PxB the applied load in X-direction at nodes A and B; and PyA, PyB the
applied load in Y-direction at nodes A and B.
In order to keep the two-dimensional structure of Figure 1 in equilibrium, the above
equations have to be satisfied. The system of equations consists of four unknowns
which are the coordinates (x, y) of the free nodes (Note: the FD ratios Ti/l i are assigned
for each cable). The system of equations is linear and since there are four equations it
can be solved for the unknown four coordinates of the free nodes (xA, yA, xB, yB).
For this example matrix C is of dimensions 3 2 and matrix Q is of dimensions
3 3. Thus the D matrix of the cable network in Figure 1, is of dimensions 2 2 and
has the following expression:
q q2 q2
10
D 1
q2 q2 q3
D
C
PyA
1
Figure 1.
Simple cable network
PyB
PxA
A
2
B
PxB
The fixed force density matrix D* of the cable network in Figure 1 is as follows:
D
q1
0
0
q3
Slack cables
in the FDM
11
12
where qn 1 is the force density of the next iteration; qn the force density of current
iteration; and Dq the incremental force density.
External constraint function ( ). The external constraint function , has the form:
gx; y; z; q 0
13
14
Jacobian matrix (GT). In order to express the coordinates as functions of the FD, the
Jacobian matrix is utilized which is obtained after applying the chain rule (Schek,
1974). Therefore, the Jacobian matrix can be viewed as a characteristic description of
each additional constraint and it can be obtained as follows:
GT
q
qg qx qg qy qg qz qg
qx qq qy qq qz qq qq
qq
15
1017
EC
31,5
1018
where:
qx
D1 C T U
qq
qy
D1 C T V
qq
qz
D1 C T W
qq
In the present work no damped conditions are considered. By using the definitions
of the external constraints and the Jacobian matrix, given in the previous sections,
Dq can be calculated from the following expression which is also described by
Schek (1974) in his classical paper:
GT Dq r
16
17
Once the incremental FD is obtained, the FD are updated and the next iteration of the
nonlinear procedure is performed, with the updated values of the FD substituted in
the set of linear equations. It is obvious that when the external constraint function ,
and the Jacobian matrix GT are defined, the implementation of an external constraint is
quite simple.
Unstrained length constraint
The implementation of the unstrained length constraint is presented here. As it can be
seen in a subsequent section the particular constraint is required for the implementation
of the slack cable in the FDM. The equilibrium shape that is obtained from the FDM is
based on the equilibrium of forces at each node. The original lengths of the branches can
be back calculated based on the principles of elasticity. Thus, the material properties of
each branch come into play only after the final shape is obtained. In order to introduce the
material properties of each element in the analysis, the unstrained length constraint is
required. This constraint accounts for the strains in the members. The external constraint
function which is shown in Equation (14) takes the form below:
lu m 0
18
2 3
19
where l u is the vector that contains the required unstrained lengths of the constraint
branches; m the vector that contains the current iteration unstrained lengths of
the constraint branches; Lu the diagonal matrix. Its entries are the values of l u ; M the
diagonal matrix. Its entries are the values of m; U ; V ; W the diagonal matrices which
respectively, which, in turn, are vectors that contain Dx, Dy,
contain the vectors u; v; w
Dz, respectively, for each branch that is constraint; B the matrix that contains the
inverse of the branch constants hi (AE)i for each branch i. The dimensions of B equal
the number of constrained branches by the number of all branches; and C the matrix
that contains rows of the branch node matrix which correspond to the constrained
branches. The dimensions of C equal the number of constrained branches by the
number of free nodes.
In each iteration a new set of coordinates is obtained which results in new lengths
for each branch. Based on these lengths, the current branch tensions and the branch
material properties, the current unstrained length m for each constraint branch, is
calculated. The required unstrained lengths lui, have known values. This constraint
ensures that the required unstrained length and the constrained length from the final
solution are equal (or within certain tolerance). This means that the strain in the
element is consistent with its tension. The unstrained length mni , for each branch i, in
each iteration, n, is calculated as follows:
hi
n
20
n li
hi T i
n
n
where l i is the length in the current iteration n of the constraint branch i; T i the force
in the current iteration n of the constraint branch i; Ai the cross-sectional area of branch
i; Ei the Youngs Modulus of elasticity of branch i; and hi Constant (AE)i of the
constraint branch i.
mni
Slack cables
in the FDM
1019
EC
31,5
1020
Depending on the nature of the load (e.g. wind load), it is possible that some of the elements
may be in compression. Trusses can support compression; however, this is not true for
cable elements. In reality, cable elements under compression become slack. This raises the
question: What happens when a cable becomes slack? If the cable element is considered
to be weightless then the element force is zero. However, this is not true. Leonard (1988)
states: The deflected geometry of a cable however, is sensitive to load patterns and
magnitudes. It is therefore, essential to consider the distributed load along the cable arc.
The linear element, which is used in the FDM in this work, is modified to model
cable behavior. This modified element is not able to support any compressive force.
The tensile force is a function of the cable element weight which is uniformly
distributed along the cable arc, and the cable span.
The classic catenary solution for the analysis of a cable element subjected to
uniform load along the cable arc is presented. The tensile force in the cable element due
to the cable weight is obtained in this manner. When a cable supports a uniform load
per unit length, such as its own weight, it takes the form of a catenary. This section
presents the classic catenary solution of a cable element subjected to uniform load
along the cable arc (Leonard, 1988). The different cable parameters, i.e. tension, etc.
are expressed in terms of the horizontal component of the cable tension, Ho. Thus, all
the cable related parameters can be calculated when the horizontal component of the
tension is obtained. The material is considered to be linearly elastic. The cable itself
is assumed to be perfectly flexible so that the bending moment at any point of the cable
must be zero.
The typical cable (Figure 2) catenary solution is presented by Leonard (1988) and is
shown below. The cable element is supported at the cable ends which do not necessarily
have to be on the same level. The horizontal span is Lc. The load on the cable consists of
its own weight which is uniformly distributed along the cable arc. Also shown on Figure 2
is the local coordinate system of the cable and the cable related parameters.
Since the applied loads are vertical, the horizontal component of the cable tension
must be constant throughout the cable element. Therefore, for the differential length dx
in Figure 2 (Leonard, 1988):
dHc
0 ! Hc Ho
dx
21
Also, from the equilibrium of the vertical forces on the same differential length
(Leonard, 1988):
dVc
ds
22
wc
dx
dx
where wc is the uniform load along the cable arc.
The cable tension must be directed along the tangent of the arc which results to the
following relation between the horizontal and the vertical forces:
dz
23
Vc H o
dx
The next step is to obtain the catenary equation of the cable profile. The derivative of
the equation above yields:
dVc
d
dz
d2 z
Ho
Ho 2
dx
dx
dx
dx
24
Slack cables
in the FDM
Vo
x
Ho
V1
z
L tan
ds
1021
Ho
dx
wc = Load Intensity
Lc
Vc
Hc
ds
dy
dx
dHc
Hc + dx dx
dVc
Vc + dx dx
Figure 2.
Typical cable element
subjected to uniform
weight along its arc
wc
p
dx2 dz2 which yields
25
Finally, the catenary equation is obtained by substituting Equations (24 and 25) in
Equation (22). The catenary equation for the deflecting cable profile is shown below:
d 2 z wc
dx2 Ho
s
2
dz
1
0
dx
26
The solution of the above equation can be obtained by integrating it twice and
applying the boundary conditions z 0 at x 0 and z Lc tan y at x Lc. The solution
is shown below (Leonard, 1988):
Ho
x
coshg b cosh g b 1 2
zx
Lc
wc
27
EC
31,5
where:
b
wc Lc
2Ho
g sinh1 tan y
1022
b
sinh b
1
1
coshg b cosh g tan y
2b
2
Also the vertical component of the force is obtained from (Leonard, 1988):
x
V Ho sinh g b 1 2
Lc
29
30
Vo V 1
wc
31
where S is the cable stretched length; Vo the vertical force at x 0; andV1 the vertical
force at x Lc.
Finally, the unstretched cable length, So, is determined from the differential equation
(Leonard, 1988):
dso
T=Ho
T
Ho T
1
32
Ho
dx 1 T=AE
AE Ho
where AE is the cable constant.
The solution of the equation above can be obtained by integration. The boundary
conditions so 0 at x 0 and so So at x Lc are applied:
"
#
So Vo V1 wc L Ho
Vo To V1 T1
33
2AE wc Lc
Lc
w c Lc
wc Lc 2
As previously stated, all the cable related parameters are expressed in terms of the
horizontal component of the cable tension, Ho. Based on the current circumstances,
Ho can be determined in one of two possible ways. If the sag ratio at the cable mid-span
is specified, the nonlinear Equation (29) can be solved for b and subsequently for Ho.
On the other hand, if the total unstretched length So is specified, then Equation (33) is
solved numerically for Ho.
Slack cables
in the FDM
1023
T Ho coshg b
34
Also known from the discussion above the initial tension, T is:
T qS
35
Vo V1
wc
36
The relation between the tensions and the FD can be obtained from Equations (35
and 36) and it is as follows:
Tq
Vo V1
wc
37
It should be noted that Equation (37) is also expressed in terms of the yet unknown Ho
which is the horizontal component of the cable tension. In order to calculate Ho, the
tension from Equation (34) is substituted in Equation (37). Equations (37) is rearranged
to yield the equation below:
q
Vo V1
wc
Ho coshg b 0
38
Equation (38) can be solved numerically for Ho. When Ho is calculated, then all of the
cable parameters can be obtained by direct substitution in the equations that are
presented in the previous section. If the unstrained cable length, as an external
constraint, is implemented, the unstrained cable length So which is the unstrained
length that corresponds to the particular FD, is calculated from Equation (33). Also the
current tension in the cable can be calculated from Equation (34).
The cable element as shown above is 2D with respect to its local coordinate system.
The implemented cable element in the FDM is also 2D even if the cable network that
the element belongs to is in 3D. Therefore, the local coordinate system as shown in
EC
31,5
1024
Figure 2, is obtained for each cable element and all the cable parameters are expressed
in this system.
6. Case study
The following examples are provided to demonstrate the results of the implemented
method. The computer program SHAPE was developed by the authors to implement
the technique that is described in this paper. SHAPE provides a number of options
for the analysis of such structures. Of interest is the option (option 1) which considers
the network elements as links (trusses) thus allowing them to develop compressive
forces. The second option (option 2) is to treat the network elements as weightless
and thus they are only capable to support tension. If they become slack then the axial
load equals zero. The third option (option 3) is to consider the cable weight and thus
even if the cables become slack they still develop tension caused by their own
weight. The implementation and use of these options allow for the study of the
differences between the solutions. In the following examples the analysis with option 1
is used to demonstrate that some elements would become slack (i.e. they are
shown to develop compression). Then the same examples are analyzed again using
option 3 to incorporate slackness. It should be noted that there is no need for the
multiple analyses. In actuality option 3 can be used at all times accounting for
the weight of the cables and possible slackness. The reason that the analyses
with option 1 are presented is to identify the slack cables and therefore demonstrate the
difference when option 3 is used. Otherwise the slack cables would not be easy to identify
as the results will only provide elements with tensile forces. In particular example 1 shows
a trivial solution which is used to demonstrate in simple terms the presented technique.
The same network was analyzed using the commercial software SAP. Example 2 shows a
more complex structure which is again analyzed using SHAPE and also the commercial
software SAP. The comparison of the results in both examples shows that SHAPE
provides reasonable results at a favorable computational time.
6.1 Example 1
A simple cable network which consists of two cables is shown in Figure 3. Each cable
has a prescribed unstrained length of 2 m (6.55 ft). The area (A) is equal to 6.45 cm2
(1 in2) and the Youngs modulus (E) is equal to 200,000 MPa (29,000 ksi). The cable
weight per unit length equals to 0.04 KN/m (0.225E-03 kips/in). The structure is loaded
at node 1 as shown on Figure 3. The initial geometry of the structure is also shown
on Figure 3. If the cable elements were to be modeled as links which can support
compression (truss elements) the following results would be obtained.
Element forces:
Element No. 1 1.000 KN (0.225 kips)
Element No. 2 1.000 KN (0.225 kips)
3
Figure 3.
Initial configuration of a
two span cable structure
1
1
Slack cables
in the FDM
1025
6.2 Example 2
Example 2 is of increased complexity compared to example 1. The cable network
consists of eight cables. The material properties for each cable are the same as those
of example 1. This network represents a structure which is initially in equilibrium
under the applied vertical loads (Figure 4). SHAPE and SAP were used to obtain
the equilibrium state due to the vertical loads. The coordinates are shown in Table II.
Table III shows the internal forces obtained from the SHAPE and SAP analyses.
Once the equilibrium state was obtained then the network was subjected to an
additional horizontal (live) load as shown in Figure 5. As in example 1 the network
CPU time (SHAPE) sec
SAP KN (kips)
4.529 (1.018)
2.529 (0.569)
3.1 102
4.53 (1.018)
2.53 (0.569)
2.19
Table I.
Analysis results for
network in Figure 3
2
1.1 kN
1.1
kN
5
33
0.
kN
0.33 kN
1
2
SHAPE KN (kips)
1.1
k
Element
1.1 k
N
The results show that if true cables are used then element 1 is in tension whereas
element 2 is in compression satisfying the condition of equilibrium at node # 2.
However, if element 2 was a cable element then it would become slack.
SHAPE is then used with the option to account for the cable weight as well as cable
slackening (option 3). The results together with the required analysis time are provided
below in Table I.
It is worth noting that there is a considerable force in the cables that were supposed
to become slack. This force is a function of the cable self-weight which cannot be
ignored especially if the cable spans are large or if the cable weight is high (as in the
case of this example) or both.
7
1 kN
0.33 kN
33
0.
1 kN
1 kN
kN
1 kN
Figure 4.
Equilibrium shape
under vertical loading
elements were modeled as links to demonstrate that some elements are in compression
and if cables would actually become slack. They are also provided here to demonstrate
the differences between the results. The coordinates of the equilibrium position are
shown in Table IV and the element forces in Table V. The SAP solution was not
obtained for this configuration as it was not considered necessary.
1026
Node
X (m)
Y (m)
Z (m)
Table II.
Coordinates (from
SHAPE) for the
network in Figure 4
1
2
3
4
5
6
7
8
0.0000
6.0000
0.0000
6.0000
1.5000
4.5000
1.5000
4.5000
0.0000
0.0000
6.0000
6.0000
1.5000
1.5000
4.5000
4.5000
0.0000
0.0000
0.0000
0.0000
4.5277
4.5277
4.5277
4.5277
Element
Table III.
Element forces for the
network in Figure 4
1
2
3
4
5
6
7
8
SHAPE KN (kips)
1.104
1.104
1.104
1.104
0.331
0.331
0.331
0.331
SAP KN (kips)
3.1 102
(0.248)
(0.248)
(0.248)
(0.248)
(0.074)
(0.074)
(0.074)
(0.074)
1.106
1.106
1.106
1.106
0.334
0.334
0.334
0.334
(0.249)
(0.249)
(0.249)
(0.249)
(0.075)
(0.075)
(0.075)
(0.075)
2.19
1
3
3.
45
kN
0.02 kN
4
5
1.3 kN
3k
5k
0.02 kN
3.4
1.0
1.3 kN
1 kN
1.6
5k
Figure 5.
Equilibrium stage of
network including truss
elements (allowing
compression)
1.6 1.3 kN
5k
N
kN
1 kN
0
.01
EC
31,5
1.3 kN
8
1 kN
1 kN
Figure 6 shows the geometry and the forces after the implementation of the slack
element using SHAPE. Figure 7 shows the element forces obtained from the analyses
of SHAPE and SAP. The horizontal axis refers to the element number whereas the
vertical axis shows the internal forces in KN.
The results obtained from SHAPE for example 2 exemplify the observations of
example 1. If links (truss elements) are considered in the analysis then a number
of elements would normally become slack (elements under compression) under the
application of the load. However, the implementation of the slack cable results to
all elements being in tension. It is also worth mentioning the difference in the
computational time between the analyses of SHAPE and SAP.
Slack cables
in the FDM
1027
7. Conclusions
In this paper a solution strategy for the analysis of cable structures is presented. Cable
elements cannot support any compressive forces but rather become slack. In this paper
the implementation of a slack cable in the FDM was presented. The implementation of
this particular element is necessary to completely model the true behavior of the cable
elements which may become slack depending on the nature of the applied loads.
The solution strategy is iterative and therefore highly computational which makes the
use of a digital computer necessary.
Two representative examples were also presented comparing the results of the
analysis using links (truss elements) and the implementation of slack elements.
Based on the results presented the following conclusions can be drawn:
(1)
The results from the examples presented verify the original assumption that
when a cable becomes slack it can still influence the final equilibrium shape
and element forces.
Node
X (m)
Y (m)
Z (m)
1
2
3
4
5
6
7
8
0.0000
6.0000
0.0000
6.0000
1.5000
4.5000
1.5000
4.5000
0.0000
0.0000
6.0000
6.0000
3.7703
3.7703
6.1351
6.1351
0.0000
0.0000
0.0000
0.0000
2.9217
2.9217
4.7678
4.7678
Element
1
2
3
4
5
6
7
8
SHAPE KN (kips)
3.449
3.449
0.016
0.016
1.035
1.650
1.650
0.005
(0.775)
(0.775)
(0.003)
(0.003)
(0.233)
(0.371)
(0.371)
(0.001)
Table IV.
Coordinates (from
SHAPE) for the
network in Figure 5
Table V.
Element forces for the
network in Figure 5
EC
31,5
1
3
1028
4k
3.
N
5
1.3 kN
0.9
1.
0.0
1 kN
1.3 kN
61
kN
1.
kN
9k
61
1 kN
Figure 6.
Equilibrium stage of
network including
cable elements
(allow cable slackening)
1.3 kN
8
1 kN
1 kN
SHAPE
SAP
AP
E
SA
P
Figure 7.
Element vs internal force
(KN) from the SHAPE
and SAP analyses
SH
1.3 kN
0.05kN
5k
N
0.05 kN
3.
4k
(2)
(3)
References
Argyris, J.H. (1964), Recent Advances in Matrix Methods of Structural Analysis, Progress in
Aeronautical Sciences, Pergamon Press, Oxford, London, New York, NY and Paris.
Buchholdt, H.A. (1985), Introduction to Cable Roof Structures, Cambridge University Press,
Cambridge.
Buchholdt, H.A. and McMillan, B.R. (1971), Iterative methods for the solution of prestress cable
structures and pinjointed assemblies having significant geometrical displacements,
Proceedings of International Association of Shell and Spatial structures, Pacific Symposium
Part II on Tension Structures and Space Frames, Tokyo and Kyoto, pp. 305-316.
Christou, P., Michael, A. and Anastasiou, C. (2010), Accounting for cable slackening and
material non-linearity in the analysis of structures with cables, Proceedings of the 9th
HSTAM International Congress on Mechanics, pp. 469-475.
Fenves, S.J. and Brannin, F.H. (1963), Network topological formulation of structural analysis,
Proceedings of the ASCE, Journal of Structural Division, Vol. 89 No. ST4, pp. 483-514.
Hangai, Y. and Wu, M. (1999), Analytical method of structural behaviours of a hybrid structure
consisting of cables and rigid structures, Engineering Structures, Vol. 21 No. 8, pp. 726-736.
Herbert, J.J. and Bachtell, E.E. (1986), Comparison of tension stabilized structures for large space
antenna reflectors, Proceedings of 27th AIAA/ASME/ASCE/AHS Structures, Sructural
Dynamics and Materials Conference, pp. 72-756.
Krishna, P. (1978), Cable Suspended Roofs, McGraw-Hill Inc., New York, NY.
Leonard, J.M. (1988), Tension Structures, McGraw-Hill, New York, NY.
Mitsugi, J. (1994), Static analysis of cable networks and their supporting structures, Computers
and Structures, Vol. 51 No. 1, pp. 47-56.
Mitsugi, J. and Yasaga, T. (1991), Nonlinear static and dynamic analysis method for cable
structures, AIAA Journal, Vol. 29 No. 8, pp. 150-152.
Miura, K. and Miyazaki, Y. (1990), Concept of the tension truss antenna, AIAA Journal, Vol. 28
No. 6, pp. 1098-1104.
Ren, W.X. (1999), Ultimate behavior of long-span cable-stayed bridges, Journal of Bridge
Engineering, Vol. 4 No. 1, pp. 30-37.
Schek, H.J. (1974), The force density method for form finding and computation of
general networks, Computer Methods in Applied Mechanics and Engineering, Vol. 3
No. 1, pp. 115-134.
Varum, H. and Cardoso, R.J.S. (2005), A geometrical non-linear model for cable systems analysis,
International Conference on Textile Composites and Inflatable Structures, Stuttgart, October 2-4.
Slack cables
in the FDM
1029
EC
31,5
1030
Volokh, K.Y., Vilnay, O. and Averbuh, I. (2003), Dynamics of cable structures, Journal of
Engineering Mechanics, Vol. 129 No. 2, pp. 175-180.
Yamamoto, C. (1990), Shape control of cable net structures, Theoretical and Applied Mechanics,
Vol. 39 No. 9, pp. 203-208.
About the authors
Dr Petros Christou is an Assistant Professor in the Civil Engineering Department at the Frederick
University in Cyprus. His research interests include the analysis and design of frame and cable
supported systems, computer aided structural analysis, structural modeling and assessment of
structural integrity, and structural analysis software development. He is also interested in soil
structure interaction and bridge analysis. Besides the research engagement Dr Christou is a
Registered Professional Engineer practicing structural engineering as a Consultant. Dr Petros
Christou is the corresponding author and can be contacted at: p.christou@frederick.ac.cy
Dr Antonis Michael is a Lecturer in the Civil Engineering Department at the Frederick
University in Cyprus. His research interests include repair and strengthening of wood and
concrete structures with FRP composites, material characterization, and confinement of concrete
with FRP composite grids and load testing of bridges. He is also interested in the experimental
determination of forces in cable systems. Besides the research engagement Dr Michael is a
Registered Professional Engineer practicing structural engineering as a Consultant.
Dr Miltiades Elliotis is a Scientific Associate in the Department of Mathematics and Statistics,
University of Cyprus. His research interests include the finite element method, boundary
method, structural analysis, numerical methods and applied mathematics. Besides the research
engagement Dr Elliotis is a Registered Professional Engineer practicing structural engineering.