Beruflich Dokumente
Kultur Dokumente
Abstract
The purpose of this paper is to compare some foregoing numerical methods for form-finding of tension structures.
Various numerical methods have already been proposed by many researchers. However, while each of them formulates a
mathematical expression to represent a self-equilibrium state, the form of them is different from each other.
As an analytical approach for form-finding, there are two types of mathematical expression to be solved. One is an
equilibrium equation based on a self-equilibrium state. The other is a stationary problem of a functional based on a
variational principle. In general, the equilibrium equations and the stationary problems of a functional are closely related
to each other.
In this paper, their related functionals are clarified for both of them, even some of them, e.g. the force density
method[1], are not related to the variational principle originally. The revealed functionals can be thought to take a minimal
number for their objective shape. The clarification helps us to understand the relation between foregoing methods and
enables us to select various non-linear computational methods instead of the original one proposed in each paper.
Additionally, in the last part of this paper, we introduce new functionals in the relation to the functional related to the
force density method. They can be considered as an extension of the force density method, and allow us to find the forms
of complex tension structures which combine cables, tension membranes and compression members. Some numerical
results by solving a stationary problem of the newly introduced functionals are also presented.
Keywords: Tension Structure, Force Density Method, Form-Finding, Functional, Variational Principle
1. Introduction
Tension structures have many advantages when we construct large spanned architecture. In general, the members of
them are light, thin and flexible, so that they usually achieve elegant forms.
Since the tension structures, such as cable-nets, suspended membranes, tensegrities, etc, are stabilized by introducing
pre-stress, they require a process in which ensures them to have a self-equilibrium state, so-called pre-stress state. While
the existence of pre-stress state relies on their form, the process is so-called “form-finding”. For this purpose, various
numerical methods have already been proposed by many researchers.
This paper compares some foregoing methods proposed in Ref. [1] to Ref. [6]. While each of them formulates a
mathematical expression to represent a self-equilibrium state, the form of them is different from each other.
As an analytical approach for form-finding, there are two types of mathematical expression to be solved. One is an
equilibrium equation based on a self-stress condition. The other is a stationary problem of a functional based on a
variational principle. In general, the equilibrium equations and the stationary problems a functional are closely related to
each other.
In Chapter 2, their related functionals are clarified for both of them, even some of them are not related to variational
principle originally. The revealed functionals can be thought to take a minimal number for their objective shapes. The
clarification helps us to understand the relation between foregoing methods, and enables us to select various non-linear
computational methods instead of the original one proposed in each paper.
From Chapter 3 to Chapter 5, some of the revealed functionals are investigated with their corresponding papers.
Additionally, in Chapter 6, we introduce new functionals which allow us to find the forms of complex tension
structures which combine cables, tension membranes and compression members.
2. Related functionals
Table 1 shows the functionals related to the foregoing methods proposed by Ref. [1] to [6]. Each method can be
considered as a method to solve the stationary problem of each functional.
For each functional, Lj denotes a function to give the length of j-th member or j-th linear element, Sj denotes a function
to give the j-th element area and wj is a weight coefficient assigned to each member or element. Since the weight
coefficients are assigned, we can obtain different forms by changing them.
As shown in Fig. 1, functional (a) suits form-finding of prestressed cable-nets. Functional (a) is related to the force
density method (FDM) [1], and investigated in Chapter 3.
As shown in Fig. 2, both functional (b) and (c) suit form-finding of membrane structures, while some differences are
exist. Functional (b) is related to the surface stress density method (SSDM) [2], and investigated in Chapter 4. Functional
(c) is the basic functional when we consider a numerical analysis of minimal surface problem, and investigated in Chapter
5 based on Ref. [3].
As shown in Fig. 3, both functional (d) and (e) suit the form-finding of tension membrane structures which
equilibrated by the reaction forces from cables. Functional (d) is the other functional related to SSDM. With (d), we can
change the forms by changing the weight coefficients wj assigned to linear elements. On the other hand, functional (e) is
the functional proposed in Ref. [4]. With (e), we can change the forms by changing Lk , the length of the k-th cable.
As shown in Fig. 4, both functional (f) and (g) suit the form-finding of tensegrities, which consist as a combination of
tension (cable) and compression (strut) members. Functional (f) is identical with the functional proposed by Goto and
Noguchi [5]. Functional (g) is one of the functionals proposed by us [6]. For both of them, the first summation is taken for
every cable and the second, for every strut. While the first term is different from each other, both of them take the
objective length Lk as a given parameter for each strut. Fig. 5 shows another numerical result using (g).
In addition, functional (h) is the other functional proposed by us [6], which suit the form-finding of tension structures
that combine cables, membranes, and compression members, as shown in Fig. 6
Therefore, it is thus confirmed that various functionals have been used for the form-finding analysis of tension
structures. It is also pointed out that the selection of the functional has some arbitrariness.
Moreover, the functionals (g) and (h), which are proposed by us, are also confirmed as a natural extension of the
foregoing methods. These functionals are proposed as an extension of FDM. The extended force density method is
discussed in Chapter 6.
where nj denotes the axial force of j-th member. If we substitute the definition of the force density, Eq. (1), to the general
form of equilibrium state, Eq. (8), we obtain
r
q jLj Lj 0 , (9)
j
which is an alternative form of the equilibrium condition for FDM. Therefore, since Eq. (6) and Eq. (9) are
mathematically equivalent, Eq. (4) is the functional which is related to FDM. Moreover, wj in Eq. (4) is assumed that it
plays as same role as the force density does in a form-finding analysis.
The self-equilibrium state, Eq. (8), lead us another form. Let
T
x x1 L x n (10)
be an arbitrary chosen column vector, in other words, virtual displacement. Since any inner product of Eq. (8) and
x must be 0, we obtain the principle of virtual work for the general prestressed cable-nets by
w nj Lj 0, (11)
j
(a) Form finding analysis with SSDM([2] Fig. 8) (b)Equilibrium of triangular element([2] Fig. 1)
Figure 9. Surface stress density method
where wj is a given weight coefficient, Sj is a function to give the surface area, for j-th triangular element respectively. The
surface is approximated by sufficient triangular elements.
Let us consider the stationary problem of Eq. (17). The stationary condition of Eq. (17) is given by
r
r 2w j S j S j 0 . (18)
x j
Let S be a function to take the coordinates of three vertices, and to give the surface area of the triangle defined by them.
In this case, S can be illustrated as shown in Fig.10 (a). On the other hand, let a triangular element bears uniform surface
stress and be balanced with three nodal loads applied to each vertex. In this case, the three nodal loads can take just
unique configuration as shown in Fig. 10(b). By comparing Fig. 10(a) and (b), a self-equilibrium state of the triangular
elements is expressed by
r
j Sj 0, (19)
j
where j denotes the surface stress of j-th element. Substituting Eq. (16) to Eq. (19), we obtain
r
QjS j S j 0 . (20)
j
Since Eq. (18) and Eq. (20) are mathematically equivalent, therefore, Eq. (17) is the functional related to SSDM.
If S j S j x , Eq. (18) also lead us the principle of virtual work for SSDM by
w 2w j S j S j 0. (21)
j
where t denotes the uniform thickness of the membrane, u denotes the virtual displacement. If the surface stress field is
uniform and isotropic, i.e. σ ˆI , we get
w tˆ I: u da 0 . (24)
a
Thus we obtain a simple form:
w tˆ ( u )da 0 . (25)
a
Since the variation of the surface area is given by
1
S ( u )da , (26)
2 a
a well known theorem is confirmed that ‘the minimal surfaces and the uniform stress surfaces are identical’.
Discretization procedure for S 0 is summarized as follows:
Let an n-dof function represents an approximated form of the surface by
x( 1 , 2 ) ~ x ( 1 , 2 , b1 , b2 Lbn ) , (27)
~
and let the variation of x be derived from x , we obtain an n-dof virtual displacement vector:
~ ~ b1
x x ~
u x ,L M x b. (28)
b1 bn
bn
Substitute Eq. (28) into Eq. (26), we get an approximation of S by
~ 1
S ( ~x b) da . (29)
2 a
Since the operators ,L and b are independent from the operators ada and i
g i , we derive
b1 bn
1~ ~
S ( ~ x da ) b S b. (30)
2 a
~ ~ ~ ~
Here, a relation a x da 2S is used, where S denotes the surface area of the approximated form. Therefore, if S can
be calculated as a summation of each element area, the variation of Eq. (22), which is given by
r r
S j ( x) S j ( x) x , (31)
j j
~
is mathematically equivalent to S . Note that we no longer need an explicit function of ~
x but only need a function S j
for each element.
Here, we extend FDM by considering various functionals as a generalization of eq. (32), e.g.
r r 4 r r
( x, λ ) w j L j ( x) k ( Lk ( x ) Lk ) . (34)
j k
Without the great advantage of the linear formulation of FDM, the key features of FDM are summarized as follows:
The coordinates are assigned to each fixed node as a known parameter.
Force densities q j n j / L j are assigned to each cable as a known parameter.
On the other hand, when we analyze a stationary problem of Eq. (34), the key features of the extended FDM are:
The lengths Lk are assigned to each strut as a known parameter.
Extended force densities w j n j /(4 L j 3 ) are assigned to each cable as a known parameter.
Therefore, the extended FDM can be said to be very similar to the FDM. Fig. 11 (a) shows some tensegrities which can
be found by solving a stationary problem of Eq. (34).
Furthermore, we can find the forms of complex tension structures, which combine cables, tension membranes and
compression members, by using various functionals, e.g.
r r 4 r 2 r r
( x, λ ) w j L j ( x) w j S j ( x) k ( Lk ( x ) Lk ) , (35)
j j k
where, the first summation is taken for every linear element, the second, for every triangular element, and the third, for
every strut. The forms of the cables are represented by linear elements, and the forms of the membranes are represented by
triangular elements. Fig. 11 (b) and (c) show some form-studies of such structures using Eq. (35).
7. Conclusion
Comparing the foregoing numerical methods for form finding of tension structures, we showed availability of
clarification of functionals related to each method. While some methods are not related to the variational principle
originally, such as FDM and SSDM, they also have related functionals. We also introduced new functionals. They can be
considered as an extension of FDM, and allow us to find the forms of complex tension structures which combine cables,
tension membranes and compression members.
Reference
[1] Schek, H. J., The force density method for form finding and computation of general networks, Comput. Methods Appl.
Mech. Engrg., 3, pp. 115– 134, 1974
[2] Maurin, B., Motro, R., The surface stress density method as a form-finding tool for tensile membranes, Eng. Struct.,
20(8), pp. 712–719, 1999
[3] Bletzinger, K-U., Ramm, E., Structural optimization and form finding of light weight structures, Comput. Struct., 79,
pp. 2053-2062, 2001
[4] Kawaguchi, K., Ke, W., Miki, M., Minimal surface with constraint conditions and steepest descent method, J. Struct.
Constr. Eng., Vol. 73, No. 632, pp.1773-1777, 2008
[5] Goto, K., Noguchi, H., Form finding analysis of tensegrity structure based on variational method, Proceedings of The
Forth CJK-OSM, pp. 455-460, 2006
[6] Miki, M., Kawaguchi, K., Fundamental study of extension of force density method on form-finding of tension
structures, J. Struct, Eng., V. 56B, pp. 533-538, 2010
[7] Tibert, A. G. and Pellegrino, S., Review of form-finding methods for tensegrity structures, Int. J. Space Struct., 18(4),
pp. 209-223, 2003
[8] Zhang, JY., and Ohsaki, M., Adaptive force density method for form-finding problem of tensegrity structures, Int. J.
Solids Struct., 43, pp. 5658-5673, 2006
[9] Vassart, N., Motro, R., Multiparametered formfinding method: application to tensegrity systems. Int. J. Space Struct.
14(2), pp. 147– 154, 1999