Case Studies in Structural Engineering 3 (2015) 19–32

Contents lists available at ScienceDirect

Case Studies in Structural Engineering

journal homepage: www.elsevier.com/locate/csse

An extended force density method for form finding

of constrained cable nets
Ghada Aboul-Nasr ⇑, Sherif A. Mourad
Structural Engineering Department, Faculty of Engineering, Cairo University, Egypt

a r t i c l e i n f o a b s t r a c t

Article history: The force density method (FDM) is a classical method used in linear and nonlinear form.
Available online 19 February 2015 The linear approach presents a quick tool for finding cable net new shapes by solving a
set of linear equilibrium equations for certain topology, boundary conditions and assumed
Keywords: cables force density. The nonlinear approach was introduced to solve cable nets under con-
Force density method straints (assigned certain distance between nodes, limit force or unstressed length in some
Cable net structures elements). Any type of constraint introduces nonlinearity.
Tension structures
This paper studied the prestressed cable nets and the loaded cable nets. For prestressed
Form finding
cable nets, coordinate constraints to all nodes of the cable net are introduced to modify the
shape after graphically examining the preliminary shape. This preliminary shape resulted
from linear analysis of assumed distribution of cable force densities. For analyzing cable
nets under different load cases, the first load case is analyzed to achieve the coordinate
constraints assigned to nodes. Analysis results are node coordinates, cable forces and
lengths. Young’s modulus and areas of cables are used to calculate the unstressed length
of all cables using materialization equations, those lengths are used as constraint in the
analysis of other load cases. Forces in all cables under different load cases/combinations
are calculated. By using this approach, design of cable net under static load is simplified.
Ó 2015 The Authors. Published by Elsevier Ltd. This is an open access article under the CC
BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/).

Introduction

Cable nets are structures consisting of cables or bars connected by pins or hinges, where cable elements can carry axial
forces only. They have several advantages over traditional construction such as, light weight, high stiffness, elastic behavior
and innovative forms. They also allow natural light in case of roofs. Its form depends on its forces and vice versa. This means
strong relationship between force and form. Accordingly it could not be dealt with as conventional structures. The cable net
shapes, whether pretensioned or not, should be analyzed to get the initial form which is a basic form for the design. Before
1970’s, the only possible way was to build a physical model and use photogrammetric tool to measure the deformation
under external loads. Linkwitz and Scheck (1971) [8], and Scheck (1974) [5] introduced the force density method. It is a sim-
ple and very powerful tool to get the initial form for a given cable connectivity, fixed points and cable force densities by using
a system of linear equations to get the exact form and improve it to get the desired shape. Frank Baron (1971) [4] introduced
nonlinear analysis of cable structures with the finite element method using stiffness matrix including geometrical matrix.
Pellegrino and Calladine (1986) [13], Pellegrino (1990) [14] and Pellegrino (1993) [15] used singular value decomposition

⇑ Corresponding author.
E-mail addresses: ghada_kh31@live.com (G. Aboul-Nasr), smourad@eng.cu.edu.eg (S.A. Mourad).

http://dx.doi.org/10.1016/j.csse.2015.02.001
2214-3998/Ó 2015 The Authors. Published by Elsevier Ltd.
This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/).
20 G. Aboul-Nasr, S.A. Mourad / Case Studies in Structural Engineering 3 (2015) 19–32

in mixed structure analysis. Dynamic Relaxation method developed by Day (1965) [2] to analyze concrete pressure vessels,
Rayleigh (1967) [11] used the DR concept to obtain static equilibrium at the steady state.
Since its introduction, the form finding using FDM has been extensively researched. Some of the research points are topo-
logical mapping method [3], equal distribution of static or quasi-static life loads in the composites retainers of a foot bridge
[1], equating sum of coordinate difference/length = 0 to obtain minimal way net [9], tension structures with integrated, lin-
ear, actively bent elements [16] and form-finding of cable domes making full use of symmetry [17], mixed cable and strut [6],
tensegrity [7], membrane structures using force density and stress density and Dome erection analysis [18].
The great advantage of the FDM is that it is the only method among form findings that does not need cable geometry in
advance. Thus, it enables the creation of lots of structural shapes based on their topology by converting the nonlinear relation
between force and form to linear relation between cable force densities (force/length) and free joints coordinates.

Computation model

Cable nets are geometrically nonlinear and elastically linear. They sustain large displacements with small deformation in
elements. Force density analysis assumes that cables are weightless and their joints are pinned. The analysis may be classi-
fied into linear and nonlinear.
The linear analysis is applied when a certain topology, known coordinates of fixed nodes and force densities in elements
are known. The stability equations of force densities in Cartesian Coordinates forming a set of linear equations. Those coordi-
nates enable the display of cable net and enable for improvement of this shape by changing fixed point location or position
and/or force densities. Distribution of force densities not their values governs the form.
The nonlinear analysis comes from any limitation or constraints applied to cable nets. The constraints are typically not
applied to all elements or nodes. The number of nonlinear equations is usually less than the number of cables or nodes
and there might be more than one solution. Among those solutions, the one that achieves the constraints with a given tol-
erance is chosen. The previous researches dealt with fixed distance between nodes, limited force and unstressed length of
some cables [5] and reactions [10,12] with defined values in chosen fixed nodes constraints.
Also, minimum way net is a cable net in which forces in all cables are equal. In this analysis, after calculating the cable
force using LFDM, all forces are restrained to a chosen value and nonlinear analysis is performed.
In this research, the coordinate constraints are used to enhance and control the form of the initially found form from the
LFDM. The coordinate constraints are divided into three different functions in x, y and z or more depending on coordinate
combinations. The NLFDM with least square method is used to analyze the constrained cable net. The coordinate constraint
is used in both prestressed and loaded cable nets. For loaded cable net, the coordinate constraint is used to control the defor-
mation under loads in first load case and, in the following load cases the unstressed length calculated from the first load case
is used as a constraint for the other load cases/combinations.

Force density method

FDM utilizes a specified net topology, fixed point coordinates and force densities of elements to get the coordinates of the
free joints at equilibrium state. In this research, plan graph at z equal zero is used to define the topology of the net. Only fixed
joints coordinates should be exactly defined. The notations used in describing stability equations are consistent with those
used by Scheck (1974) as follows:
Lower case is a vector whereas upper case is a matrix. The same vector symbol in upper case refers to a diagonal matrix
having vector elements as its diagonal. Element joints are i and j.

Stability equation

Fig. 1 shows system axis and coordinates. In this system only three stability equations exist and sum of forces in the three
Cartesian directions are zero.
X X X
Fx ¼ 0 Fy ¼ 0 Fz ¼ 0 ð1Þ

Stability equations at a free joint K are as follows:

Sa cosða; xÞ þ Sb cosðb; xÞ þ Sc cosðc; xÞ þ Sd cosðd; xÞ ¼ Px ð2aÞ

Sa cosða; yÞ þ Sb cosðb; yÞ þ Sc cosðc; yÞ þ Sd cosðd; yÞ ¼ Py ð2bÞ

Sa cosða; zÞ þ Sb cosðb; zÞ þ Sc cosðc; zÞ þ Sd cosðd; zÞ ¼ Pz ð2cÞ

Where Sa is the force in element a, cos (a, x) is the cosine of the angle between element a and X axis, and Px is the x compo-
nent of the applied force at K. Similar equations are derived for Y and Z directions.
Substituting cosine by the normalized projection length, Equations 2 become:
 G. Aboul-Nasr, S.A. Mourad / Case Studies in Structural Engineering 3 (2015) 19–32 21

Fig. 1. System axes and coordinates.

Sa ðxai  xaj Þ=La þ Sb ðxbi  xbj Þ=Lb þ Sc ðxci  xcj Þ=Lc þ Sd ðxdi  xdj Þ=Ld ¼ P x ð3aÞ

Sa ðyai  yaj Þ=La þ Sb ðybi  ybj Þ=Lb þ Sc ðyci  ycj Þ=Lc þ Sd ðydi  ydj Þ=Ld ¼ Py ð3bÞ

Sa ðzai  zaj Þ=La þ Sb ðzbi  zbj Þ=Lb þ Sc ðzci  zcj Þ=Lc þ Sd ðzdi  zdj Þ=Ld ¼ P z ð3cÞ
The above equations are nonlinear. To linearize them, q is introduced as force density S/L. If q and P are known, the result
is a set of linear equations in x, y, and z and it can be formed as follows:
qa ðxai  xaj Þ þ qb ðxbi  xbj Þ þ qc ðxci  xcj Þ þ qd ðxdi  xdj Þ ¼ Px ð4aÞ

qa ðyai  yaj Þ þ qb ðybi  ybj Þ þ qc ðyci  ycj Þ þ qd ðydi  ydj Þ ¼ Py ð4bÞ

qa ðzai  zaj Þ þ qb ðzbi  zbj Þ þ qc ðzci  zcj Þ þ qd ðzdi  zdj Þ ¼ Pz ð4cÞ

A matrix Cs is defined to be used in obtaining the final equilibrium solution. Matrix Cs is a matrix that describes the ele-
ment connectivity. So it has m rows and ns columns, m is the number of elements and ns is the number of free nodes (n) plus
the number of fixed nodes (nf). Fixed nodes should come after free nodes in the node numbering. All elements of Cs are zero
except nodes connecting elements take a value 1 or 1 as follows:
1 for element first node
cðj; iÞof element :
1 for element second node
By numbering fixed nodes at the end, Cs is divided into C and Cf (Cs = [C| Cf]).
The coordinate difference vectors are u, v, and w
u ¼ Cx þ C f xf ð5aÞ

v ¼ Cy þ C f yf ð5bÞ

w ¼ Cz þ C f zf ð5cÞ
From Eqs. (4a–c and 5a–c) we get:

C t Q ðCx þ C f xf Þ ¼ Px ð6aÞ

C t Q ðCy þ C f yf Þ ¼ Py ð6bÞ

C t Q ðCz þ C f zf Þ ¼ Pz ð6cÞ
Eqs. (6a–c) are linear in x, y and z unknown coordinates.

Prestressed cable net

In case of prestressed cable net the stability equation will be:

22 G. Aboul-Nasr, S.A. Mourad / Case Studies in Structural Engineering 3 (2015) 19–32

C t QCx ¼ C t QC f xf ð7aÞ

C t QCy ¼ C t QC f yf ð7bÞ

C t QCz ¼ C t QC f zf ð7cÞ
t
Defining the Gaussian transformation matrix of C (D = C QC), it could be formed directly from force densities, diagonals
are the sum of all force densities connected to the node and off diagonals are negative value of element force density
connecting the two nodes.
Df = CtQCf, it has n rows and nf columns, it could be formed directly by setting negative value of cable force density con-
necting free node to fixed node, and zero to all other matrix elements.
Multiplying both sides of Eqs. (7a–c) by D1, we get:

x ¼ D1 Df xf ð8aÞ

y ¼ D1 Df yf ð8bÞ

z ¼ D1 Df zf ð8cÞ
Linear Eqs. (8a–c) are solved to get x, y and z.

General constraint

Constraints are expressed as a function of coordinates and force densities which should be satisfied with an acceptable
tolerance. Since coordinates are function of force densities, constraints are function of force densities. Assume r is the num-
ber of constraint of n nodes or of m elements.
g i ðx; y; z; qÞ ¼ 0 iði ¼ 1 : r; r < mÞ ð9Þ
For all r condition
gðx; y; z; qÞ ¼ 0 gðxðqÞ; yðqÞ; zðqÞ; qÞ ¼ 0 gðqÞ ¼ 0 ð10Þ
The starting set of force densities q0 will not satisfy the constraint conditions, to solve for q the equations are nonlinear
and are solved by iteration using Dq.

@gðq0 Þ
gðq0Þ þ Dq ¼ 0 ð11Þ
@q
By using:

@gðq0 Þ
Jacobian matrix GT ¼ ð12Þ
@q

and misfit r ¼ gðq0 Þ ð13Þ

And substituting Eqs. (12) and (13) in Eq. (11) we get:

GT Dq ¼ r ð14Þ
The system of r equations is less than the number of nodes/elements. So, it is under determinate and has mr solutions.
The solution that satisfies the minimum sum of Dq square and in the same time minimum misfit value is chosen.

Dq:DqT ! min: r:r T ! min: ð15Þ

Eq. (14) will be:

GT GDq ¼ Gr ð16Þ
Then
1
Dq ¼ GðGT GÞ r ð17Þ
For the next iteration

qðnewÞ ¼ qðoldÞ þ Dq ð18Þ

(new) T (new)
q is used to get new coordinate Eqs. (6a–c). Compose G to get q and check for convergence or maximum number
of iterations allowed by the program.
 G. Aboul-Nasr, S.A. Mourad / Case Studies in Structural Engineering 3 (2015) 19–32 23

Coordinate constraint

Scheck (1974) introduced node distance, force and length constraints, Malerba (2012) introduced the reaction component
constraint.
Coordinate constraint assumes a certain position x, y, and z or one/two or combination of them to some nodes or to all
nodes.

gðxÞ ¼ 0 gðyÞ ¼ 0 gðzÞ ¼ 0 ð19Þ

g(x + y) = 0 as an example for coordinate combination function.
Taking z coordinate function as an example:

@gðqÞ @gðqÞ @z
¼  ð20Þ
@q @z @q
Using Eqs. (5c and 6c)

C t Wq ¼ Pz ð21Þ
dq and dz should keep the equilibrium of the system

@ðC t WqÞ @ðC t WqÞ

dðC t WqÞ ¼ dq þ dz ð22Þ
@q @z

@ðC t WqÞ
¼ CtW ð23Þ
@q

@ðC t WqÞ @ðC t QwÞ @w

¼ ¼ CtQ ¼ C t QC ð24Þ
@z @z @z

@w
CtQ ¼ C t QC ð25Þ
@z

@z
¼ DC t W ð26Þ
@q

@gðzÞ
¼1 ð27Þ
@z
Substitute Eqs. (26) and (27) into Eq. (20)

@gðzÞ
¼ D1 C t W ð28Þ
@q
And the same for x and y

@gðxÞ
¼ D1 C t U ð29Þ
@q

@gðyÞ
¼ D1 C t V ð30Þ
@q
When r < ns, GT could be formed directly by considering only the constrained nodes, so GT has dimension rxm.
The Jacobian matrix of the whole coordinate constraints is obtained by sequentially add rows of each class after another.

@gðx; yÞ
For g ¼ x þ y ¼ D1 C t U  D1 C t V ð31Þ
@q

Materialization

In order to extend the force density method to be used for cable net analysis for different load cases, the system should be
materialized to maintain the unstressed length of each element constant. This is done by applying Hooke’s law to the equi-
librium set of forces in the elements where first load case is applied. During materialization, the system is no longer linear.
The set of nonlinear equations in coordinates and force densities needs to be linearized as previously mentioned in General
Constraint.
24 G. Aboul-Nasr, S.A. Mourad / Case Studies in Structural Engineering 3 (2015) 19–32

Unstressed constraint

Stress
Hooke’s Law E ¼ Strain
where E is Young’s Modulus

S DL
E¼ ð32Þ
A L

Substituting S ¼ q  L, DL = L  Lu and EA = H in Eq. (32), the result is qL ¼ H LL

Lu
u
, using this result to get:

H
Lu ¼ L ð33Þ
HþS
Eq. (33) is used to calculate the unstressed length for all cable net elements of the final equilibrium resulting from first
load case. Then this length is considered as a constraint for the other load cases.

GT ¼ L2u H1  L2u H3 ðUCD1 CU þ VCD1 CV þ WCD1 CWÞ ð34Þ

Eq. (34) is the Jacobian matrix of unstressed length constraint.

Program

A computer program is developed to implement the equations formulated in the above sections. The extended force den-
sity method (EFDM) can solve two system of cable nets; prestressed and loaded nets. For both systems, the topology and
coordinates of fixed nodes of the net (the free nodes are drawn in z = 0 plane) are read from drawing exchange format
(dxf). Fig. 2 outlines a flowchart for the procedure, identifying the solution algorithm for the prestressed net as well as
the loaded net.
For the prestressed cable net, the distribution of cable force densities is read from a text file, the program form the branch
node matrix Cs, matrix D and Df, and solve the set of linear equations 8 to get the free nodes coordinates. The resulted form is
displayed after forming dxf file to investigate the shape and enhance it. The enhancement is done by changing free node
coordinates. Changes made to the net are transferred to the program as coordinate constraints which could be applied to
all nodes. Now the stability equations are no longer linear, but underdetermined nonlinear. Least square method is used
to solve these equations to get the force densities in all elements and calculate the cable lengths and forces.
For the loaded cable net, the shape is known in advance. The force densities and loads on the free nodes are read for each
load case from text file. For the first load case the linear set of equations 6 is solved to get the initial equilibrium. Coordinate
constraints are applied to the initial form, a set of nonlinear equations is solved and dxf file is formed. Then lengths,
unstressed lengths and forces are calculated. For the following load cases, the unstressed length calculated from the first load
case is set as a constraint for all load cases.

Program verification

To verify the linear part of the program two finite element programs were used; SAP2000 and STAAD.Pro. A rectangular
cable net covering an area of 8  6 m is analyzed. Nodes on perimeter are fixed and the entire net is in zero z plane. All free
nodes are loaded by 0.5 ton concentrated load in gravity direction, the force divided by length from each program output is
used as force density input in EFDM to compare the z coordinate. Fig. 3 shows a plan of the net used for verification and
identifies the free nodes in one quarter of the mesh. Table 1 compares nodal displacements in ‘z-direction.
From Table 1 it could be seen that the maximum difference between z computed by EFDM, and both SAP2000 and STAAD
are 0.50505% and 0.37422%, respectively.

Prestressed unloaded example: Star net

Fig. 4 shows a plan view and node numbering of a star cable net. The net has 36 nodes and 63 members. The perimeter
nodes (14 nodes) are fixed, whereas an interior node (6) is z coordinate constrained and all free nodes z coordinate are pro-
portionally constrained also.
The used unit is tonf-m. The nodes 25–36 are fixed, all nodes are in zero z plane except 23, 24, 28 and 36 are in plane
z = 3 m and node 6 has z constraint of 2 m and all free nodes z coordinate are proportionally constrained too.
By using any value of equal force densities in all elements and solving according to Eq. 7 to get the initial equilibrium
position, the z constraint is extended to include z of all free nodes. The iteration stopped after 6 iterations or
Dq.DqT < 0.00001 and r.rT < 1.0E  7. Fig. 5 shows the LFDM and NLFDM equilibrium position. Fig. 6 shows quarter of the
net z position and force densities.
G. Aboul-Nasr, S.A. Mourad / Case Studies in Structural Engineering 3 (2015) 19–32 25

Fig. 2. Program flow chart.

Fig. 3. Cable net for program verification.

26 G. Aboul-Nasr, S.A. Mourad / Case Studies in Structural Engineering 3 (2015) 19–32

Table 1
Program verification nodes z coordinate.

Node EFDM z SAP2000 z % Diff. EFDM z STAAD z % Diff.

1 0.06999 0.06964 0.50505 0.06707 0.06723 0.24856
2 0.04489 0.04484 0.09824 0.04294 0.04310 0.37422
3 0.11786 0.11746 0.33938 0.11300 0.11313 0.11823
4 0.09376 0.09355 0.23091 0.09018 0.09035 0.18918
5 0.12633 0.12603 0.23589 0.12128 0.12141 0.11049
6 0.06314 0.06294 0.30647 0.06083 0.06091 0.13481

Fig. 4. Plan view and program node numbering of the star net.

Fig. 5. Prestressed unloaded star net initial and z constrained equilibrium shapes.

Fig. 6. Quarter of the star net z position and force densities.

 G. Aboul-Nasr, S.A. Mourad / Case Studies in Structural Engineering 3 (2015) 19–32 27

Loaded net examples

Four examples are presented to demonstrate the capabilities of the proposed algorithm to handle loaded nets. For each
example, a figure illustrates node numbering and element numbering. The first table identifies the displacements at four dis-
tinct equilibrium states. Symbol a is the equilibrium for load case 1, b is for load case 1 under z constraint, c and d are for load
cases 2 and 3 respectively based on the unstressed length obtained from b. The second table shows the unstressed length and
the corresponding force densities for b, c and d.

Star cable net

It is the same cable net in Section ‘‘Prestressed unloaded example: Star net’’ except nodes 23, 24, 28 and 36 are free. A
force density of 5 ton/m is assumed for all elements. The first load is assumed to be 0.4 tons, second load case is 0.54, the
third one is 0.7 in gravity direction applied to all free nodes. Fig. 7 shows the nodes and members numbering of a quarter.
Table 2 identifies the equilibrium states z coordinate and table 3 presents the unstressed length and corresponding force
densities for the three load cases.
First the solution determines the initial equilibrium state using LFDM, then, constrain the z position of all free nodes by
setting the maximum deflection to 0.14 m and the program take the ratio of the maximum z to the virtual one and pro-
portionally get the other free nodes z constraint. From the final equilibrium position of the first load case, the unstressed
length of all elements is the constraint for the following load cases.

Rhombic cable net

A rhombic cable net covers an area of 10  10 m. Force density of 5 ton/m is assumed for all elements and three load cases
are identified with 0.7, 0.54 and 0.4 tons in gravity direction assumed for all nodes. Maximum deflection is assumed equal to
(0.14 m). Fig. 8 shows a quarter of a rhombic cable net. Table 4 identifies the equilibrium states z coordinate and table 5
presents the unstressed length and corresponding force densities of the quarter.
The analysis results shows that node z coordinate in load cases b, c and d are slightly different due to the unstressed
length constraint applied. It also demonstrates that the unstressed length of each element in all load cases is the same.
The resulted force densities in c and d equal to the force densities in b multiplied by the ratio of the applied load. This rela-
tion is important and introduce new point to study.

Fig. 7. Nodes and elements numbering of quarter of the star cable net.

Table 2
Nodes (z) coordinates.

Element z Coordinate m
a b c d
1 0.1658 0.1411 0.1443 0.1472
2 0.1270 0.1081 0.1085 0.1091
3 0.0686 0.0584 0.0595 0.0606
4 0.1707 0.1452 0.1468 0.1483
5 0.0674 0.0573 0.0589 0.0603
6 0.1210 0.1029 0.1052 0.1072
7 0.1888 0.1607 0.1623 0.1638
28 G. Aboul-Nasr, S.A. Mourad / Case Studies in Structural Engineering 3 (2015) 19–32

Table 3
Element unstressed length and force density.

Element Unstressed length m Force density ton/m

b c d b c d
1 3.6159 3.6159 3.6159 5.9414 7.8856 10.0599
2 6.4742 6.4742 6.4742 5.7563 7.3810 9.1334
3 2.7020 2.7020 2.7020 5.8991 8.1858 10.8649
4 3.1429 3.1429 3.1429 5.9319 7.7444 9.7402
5 2.2087 2.2087 2.2087 6.0729 8.0596 10.2818
6 2.4533 2.4533 2.4533 5.2245 6.9370 8.8499
7 2.5136 2.5136 2.5136 5.5923 7.4632 9.5678
8 3.8255 3.8255 3.8255 5.5983 7.4160 9.4465
9 2.6828 2.6828 2.6828 5.2770 7.3766 9.8493
10 3.0847 3.0847 3.0847 5.0281 6.5663 8.2570
11 4.3759 4.3759 4.3759 5.6895 7.2557 8.9304
12 2.4912 2.4912 2.4912 6.1929 8.3170 10.7271
13 2.8567 2.8567 2.8567 5.5346 7.0109 8.5738
14 2.7447 2.7447 2.7447 5.6435 7.8207 10.3673
15 3.2231 3.2231 3.2231 5.5398 7.0034 8.5481
16 3.6920 3.6920 3.6920 6.0355 7.9573 10.0988
17 4.2600 4.2600 4.2600 5.7807 7.4947 9.3666

Fig. 8. Nodes and elements number of one-quarter of the rhombic cable net.

Table 4
Nodes (z) coordinates.

Element z Coordinate m
a b c d
1 0.1543 0.0792 0.0806 0.0791
2 0.2737 0.1405 0.1422 0.1403
3 0.2387 0.1225 0.1243 0.1223
4 0.1406 0.0721 0.0730 0.0720
5 0.2461 0.1263 0.1284 0.1261
6 0.1836 0.0942 0.0956 0.0941
7 0.0809 0.0415 0.0423 0.0414
8 0.1268 0.0651 0.0659 0.0650

Table 5
Element Unstressed length and force density.

Element Unstressed length m Force density ton/m

b c d b c d
1 1.4164 1.4165 1.4164 9.7745 7.5528 5.5903
2 1.4149 1.4149 1.4149 9.7745 7.4114 5.6004
3 1.4143 1.4143 1.4143 9.7745 7.5944 5.5593
4 1.4160 1.4161 1.4160 9.7745 7.4754 5.5762
5 1.4151 1.4151 1.4151 9.7745 7.5945 5.5593
6 1.4142 1.4142 1.4142 9.7745 7.4113 5.6005
7 1.4160 1.4161 1.4160 9.7745 7.5945 5.5593
8 1.4144 1.4144 1.4144 9.7745 7.4754 5.5763
9 1.4146 1.4146 1.4146 9.7745 7.5526 5.5905
10 1.4148 1.4148 1.4148 9.7745 7.4875 5.5951
11 1.4152 1.4152 1.4152 9.7745 7.5526 5.5905
12 1.4145 1.4145 1.4145 9.7745 7.4754 5.5763
13 1.4148 1.4148 1.4148 9.7745 7.5526 5.5905
14 1.4148 1.4148 1.4148 9.7745 7.4875 5.5951
15 1.4157 1.4157 1.4157 9.7745 7.5374 5.5717
 G. Aboul-Nasr, S.A. Mourad / Case Studies in Structural Engineering 3 (2015) 19–32 29

Fig. 9. Plan view and program numbering of circular net.

Fig. 10. Nodes and elements number for 1/8 of the circular net.

Table 6
Nodes (z) coordinates.

Element z Coordinate m
a b c d
1 0.50000 0.30000 0.30053 0.30105
2 0.45000 0.27000 0.27042 0.27083
3 0.35000 0.21000 0.21041 0.21080
4 0.20000 0.12000 0.12038 0.12074

Circular cable net

Fig. 9 shows a circular cable net with read node and element numbers from dxf file. It covers an area of radius 19.66 m.
Outer nodes are fixed. Force density is assumed 1 ton/m for radial elements, 3 ton/m for diagonals and 5 ton/m for inner
radials. Load cases are 0.15, 0.20 and 0.25 tons in gravity direction for all free nodes. Maximum deflection is assumed equal
to (0.30) m. Fig. 10 shows node and element numbers given to one eighth of the cable net to show results. Table 6 identifies
the equilibrium states z coordinate and table 7 presents the unstressed length and corresponding force densities for all load
cases.
30 G. Aboul-Nasr, S.A. Mourad / Case Studies in Structural Engineering 3 (2015) 19–32

Table 7
Element unstressed length and force density.

Element Unstressed length m Force density ton/m

b c d b c d
1 1.18732 1.18732 1.18732 5.00000 6.64225 8.27314
2 1.56454 1.56454 1.56454 5.00000 6.66497 8.32891
3 2.12490 2.12490 2.12490 5.00000 6.66498 8.32874
4 2.93378 2.93378 2.93378 5.00000 6.64585 8.28209
5 1.55081 1.55081 1.55081 5.00000 6.64223 8.27310
6 2.45927 2.45927 2.45927 1.00000 1.34733 1.70097
7 3.65584 3.65584 3.65584 1.00000 1.33301 1.66566
8 5.28071 5.28071 5.28071 1.00000 1.31910 1.63187

Fig. 11. Plan view and program numbering of orthogonal net.

Fig. 12. Node and element numbers of quarter of the orthogonal net.
 G. Aboul-Nasr, S.A. Mourad / Case Studies in Structural Engineering 3 (2015) 19–32 31

Table 8
Nodes (z) coordinates.

Element z Coordinate m
a b c d
1 0.9048 0.5000 0.5005 0.5010
2 0.8460 0.4675 0.4679 0.4684
3 0.6811 0.3764 0.3767 0.3770
4 0.4506 0.2490 0.2492 0.2494
5 0.8635 0.4772 0.4777 0.4782
6 0.7991 0.4416 0.4420 0.4425
7 0.6138 0.3392 0.3395 0.3398
8 0.3430 0.1896 0.1897 0.1898
9 0.7512 0.4151 0.4156 0.4161
10 0.6731 0.3720 0.3724 0.3728
11 0.4320 0.2387 0.2390 0.2393

Table 9
Element unstressed length and force density.

Element Unstressed length m Force density ton/m

b c d b c d
1 2.0854 2.0854 2.0854 3.9137 5.2819 6.8446
2 1.9769 1.9769 1.9769 3.8537 5.2013 6.7407
3 1.6627 1.6627 1.6627 3.5740 4.8254 6.2558
4 2.1686 2.1686 2.1686 4.1355 5.5815 7.2332
5 2.1276 2.1276 2.1276 4.0521 5.4694 7.0885
6 1.9277 1.9277 1.9277 3.8893 5.2512 6.8080
7 2.3963 2.3963 2.3963 6.7013 9.0316 11.6856
8 2.5305 2.5305 2.5305 6.8252 9.1986 11.9017
9 2.9708 2.9708 2.9708 7.0810 9.5434 12.3479
10 1.0267 1.0267 1.0267 3.1993 4.3132 5.5828
11 0.8691 0.8691 0.8691 2.8728 3.8717 5.0094
12 1.1256 1.1256 1.1256 3.3912 4.5719 5.9174
13 0.9848 0.9848 0.9848 3.1617 4.2611 5.5132
14 1.4292 1.4292 1.4292 3.9840 5.3712 6.9520
15 1.4139 1.4139 1.4139 3.7931 5.1123 6.6148
16 1.9483 1.9483 1.9483 7.1900 9.7074 12.5850
17 2.5119 2.5119 2.5119 7.4364 10.0402 13.0165

For circular shape assumed force densities for radial elements should be more than the force densities of circular ele-
ments to get reasonable flat shape. This example results shows the same points as the rhombic cable net.

Orthogonal cable net

Fig. 11 shows an orthogonal cable net of 8  15 m with the program node and element numbering. Only corner nodes are
fixed. Force density is assumed equal to 2 ton/m for inner elements and 4 ton/m for outer elements. Load cases are assumed
equal to 0.4, 0.54 and 0.70 tons in gravity direction for all free nodes, maximum deflection is assumed (0.50 m). Fig. 12
shows the node numbers and element numbers to show results for one-quarter of the cable net. Table 8 identifies the equi-
librium states z coordinate and table 9 presents the unstressed length and corresponding force densities for all load cases.
For shapes with fixed nodes only at corners, the more the ratio of the assumed force densities between outer and inner
elements, the more flat the shape is. The results show the same points as the rhombic cable net and circular cable net.
In all examples, the results show that unstressed length is kept unchanged while the load change.

Conclusion

This research presents the procedure and computer program to determine cable net form by setting coordinate con-
straints to the initial equilibrium state of assumed force densities, fixed nodes and given distribution of force densities.
The coordinates are the main variable which controls the form of cable net of a certain topology. For loaded cable nets
the coordinate constraint is applied only in the first load case and the unstressed length is calculated. For the remaining load
cases the first load case unstressed length is the constraint applied to all net elements. The illustrative examples demonstrate
the capabilities of the program, thus simplifying the design of cable nets under static loads.
32 G. Aboul-Nasr, S.A. Mourad / Case Studies in Structural Engineering 3 (2015) 19–32

References

[1] Jülich AS, Caron JF, Baverel O. Cable structure with load-adapting geometry. In: Composites in construction 2005 – third international conference. Lyon,
France; 2005. p. 1–8.
[2] Day AS. An introduction to dynamic relaxation. The Engineer 1965.
[3] Hernández-Montes E, Jurado-Piña R, Bayo E. Topological mapping for tension structures. ASCE 2006:970–7.
[4] Baron Frank, ASCE F, Venka!esan Mohadev S. Nonlinear analysis of cable and truss structures. J Struct Div 1971:679–710.
[5] Scheck HJ. The force density method for form-finding and computation of general networks. Comput Methods Appl Mech Eng 1974;3:115–34.
[6] Tran Hoang Chi, Lee Jaehong. Advanced form finding for cable strut. Int J Solids Struct 2010;47:1785–94.
[7] Zhang JY, Ohsaki M. Adaptive force density method for form finding problem. Int J Solids Struct 2006:5658–73.
[8] Linkwitz K, Scheck H. Einige bmerkungen zur brechnug vonorgespannten seilnetz-konstruktionen (remarks concerning the analysis of prestressed
structures). Ingenieur-Archiv 1971;40:145–58.
-
[9] Fresl Krešimir, Gidak Petra, Vrančić Romana. Generalized minimal nets in form finding of prestressed cable nets. Gradevinar 2013:707–20.
[10] Manuel Quagliaroli. An extended force density method for the form finding of suspended pedestrian bridges. In: Footbridge 2011, 4th international
conference. Wrodaw, Poland; 2011. p. 892–903.
[11] Otter JRH, Cassell AC, Hobbs RE. Dynamic relaxation, discussion. In: Proc. ICE, v. 723–750, 1967.
[12] Malerba PG, Patelli M, Quagliaroli M. An extended force density method for the form finding of cable systems with new forms. Struct Eng Mech
2012;42(2):191–210.
[13] Pellegrino S, Calladine CR. Matrix analysis of statically and kinematically indeterminate frame-works. Int J Solids Struct 1986;22:409–28.
[14] Pellegrino S. Analysis of prestressed mechanisms. Int J Solids Struct 1990;26:1329–50.
[15] Pellegrino S. Structural computations with the singular value decomposition of the equilibrium matrix. Int J Solids Struct 1993;21:3025–35.
[16] Van Mele Tom, De Laet Lars, Veenendaal Diederik, Mollaert Marijke, Block Philippe. Shaping tension structures with actively bent linear elements. Int J
Space Struct 2013;28(3&4):127–35.
[17] Xi Ying, Xi Zhuo, Qin Wei Hong. Form finding of cable domes by simplified force density method. Struct Build 2011;164:181–95.
[18] Lin Zhibin, Qian Ruojun, Yuan Xingfei. Erection analysis of a large-scale radial cable net. Int J Civil Environ Eng 2012;164:47–51.

Further reading

[19] SAP2000 14.1.0 Program. Copyright Computers and Structures, Inc. USA (2009–07).
[20] STAAD.PRO 2007 Program. Research Engineers International, a Bentley Solutions Center, USA (2007–09).