-
i
SHAPE ANALYSES FOR ISOTONIC SURFACE WITH COMBINATION OF
SOAP FILM ELEMENTS AND AXIAL LINES
by
Wu Yusheng
A thesis submitted in partial fulfillment of the requirements
for the degree of Master of
Engineering in Civil Engineering, Saga University
Examination Committee: A. Prof. H. OBIYA (Chairperson)
Prof. K. IJIMA
Prof. ISHIBASHI
Previous Degree: Bachelor of Engineering in Civil
Engineering
Zhe Jiang University of Science and Technology
Department of Civil Engineering and Architecture
Graduate School of Science and Engineering
Saga University
Japan
August, 2010
-
ii
ACKNOWLDEGEMENTS
This thesis has been carried out at the Division of Structural
Engineering, Department of
Civil Engineering, Saga University, Saga, Japan from October
2008 to September 2010. I
have completed my research under the supervision of Professor
Hiroyuki OBIYA.
First, I would like to express my appreciation to my advisor
Hiroyuki OBIYA. Thanks
for offering me the opportunity to be part of his group. He is
such a good advisor. He
always was there to listen to my problems and gave me good
advices. He taught me how to
ask questions and express my ideas. He showed me different ways
to approach a research
problem and the need to be persistent to achieve any goal. Under
his encouragement and
constant guidance, I overcame some very tough time and completed
my research during
two years staying in Japan.
Second, I also would like to thank the other two members in
division of structural
engineering, Professor Katsushi IJIMA and Technician N.Kawasaki.
Thanks for everything
you offered during my studying period. Without your supports, I
would have not able to
finish my research.
Last but not least, I would like to thank my colleagues in Obiya
Laboratory as well as
Ijima Laboratory for interesting discussions and being fun to be
with.
Saga, Japan 2010
-
iii
SHAPE ANALYSES FOR ISOTONIC SURFACE WITH COMBINATION OF
SOAP FILM ELEMENTS AND AXIAL LINES
08547004 Wu Yusheng
1 Introduction
The shape analysis is to find the mechanically rational form for
many kinds of structures,
for example, membrane structures, suspended cables, and so on.
The isotonic surfaces can
be adopted for the primary shape of these structures. Purpose of
this study is to discuss
about the discrete solutions obtained by soap film analyses.
Under the assumption of
"triangle isotonic element", the determined solution should be
obtained as "a polyhedral
equilibrium shape".
Physically soap film element has no stiffness to the tangent
direction of its surface.
When 1D analysis only to normal direction is applied for shape
analysis of soap film,
dispersion of the element area is a fatal problem. However, 2D
axial lines (The axial line
elements are projected to tangent direction of its surface, and
have 2 degree of freedom)
have an ability to optimize the node distribution even large
increment of compulsory
displacement is adopted. Therefore, this study proposes a 3D
analysis to obtain the
polyhedral solutions by parallel calculation of soap film
analysis only to normal direction
and the 2D axial line analysis in tangent plane of the surface.
By using this technique, node
distribution becomes more rational and large incremental of
compulsory displacement can
be adopted when the simultaneous control is used as the
incremental method. Moreover,
there is another usage of axial lines that can be converted to
cable elements with real
stiffness after form-finding. This type of axial line elements
which is named "3D axial
line" has degree of freedom to all of 3 dimensions. When 3D
analysis with 3D axial lines is
applied for shape analysis of soap film, the polyhedral
solutions can be expected to
determine composite structures of membranes and cables.
2 Analytical Method
2.1 3D Analysis
In this thesis, this study proposes a technique which is the
parallel calculation of the
soap film analysis and the axial line analysis. This technique
defines two independent and
parallel coordinate, as shown in Fig.1. One has only the one
axis toward normal direction
of the surfaces and that is prepared for the isotonic surface
analysis. The other has two axes
and perpendicular each other and both are to tangent direction
of surfaces. These two
axes are prepared for 2D axial line analysis.
-
iv
2.2 2D axial line and 3D axial line
The Fig.2 shows an axial line element. The element forces of the
axial line elements are
projected to and coordinate axes. Thus this type of axial line
elements has only 2
degree of freedom, so the author calls it "2D axial line" in
this thesis. According to the
effect of the 2D axial lines, the mesh distribution becomes more
uniform, and the
polyhedral equilibrium solutions become more similar to the
isotonic surfaces.
Moreover, there is another usage of the axial line elements that
can be converted to cable
elements with real stiffness after form-finding. This type of
axial line elements has degree
of freedom to all of 3 dimensions, and the author calls it "3D
axial line". 3D axial lines can
restrict the deformation of soap film during the process of
form-finding. Therefore, the
combination of the soap film elements and 3D axial lines can be
expected to determine a
composite structure of membranes and cables.
3 Numerical Examples and Discussions
The geometry of initial shape is shown in Fig.3. The green lines
denote 3D axial lines
and the 2D axial lines to unify the mesh distribution are
located on 3 sides of all triangular elements. The tensile
of
the soap film elements is 3.0kN/m, and the coefficient of
line elements C=1. Simultaneous control is used as an
incremental method in this example. The increment of the
compulsory displacement is given as Dc=2m and Dc=0.1m
respectively for 1D soap film analysis (Fig.4 [a], [b]). In
contrast, for 3D analysis with 3D axial lines, the increment
of the compulsory displacement is given as 2m (Fig.4 [c]).
Because of the effect of 2D axial lines, node distribution
becomes more rational and
i
Fig.3 Initial plane
N
N
Wi
Ui
Vi
i
Wj
Uj
Vj
j
i
i
i
i
R=8m
Fig.1 Coordinate system with normal direction
and tangent direction Fig.2 Axial line element
-
v
large increment of compulsory displacement can be adopted.
Meanwhile, 3D axial line can
restrict the deformation of soap film during the process of
form-finding. The determined
shape can be expected to composite structure of membranes and
cables.
The mesh connectivity is shown in Fig.5, and the power of 3D
axial lines n is 4. The
Fig.6 shows the equilibrium solutions compared with two kinds of
mesh connectivity.
Fig.4 The equilibrium shapes
[a] 1st connectivity
Fig.5 Initial plane
[b] 2nd
connectivity
Fig.6 The equilibrium shapes
[a] 1st connectivity [b] 2
nd connectivity
n:Power of 3D axial lines, Dc: Increment of compulsory
displacement
[a] 1D analysis, Dc =2m
[b] 1D analysis, Dc =0.1m
[c] 3D analysis with 3D axial
lines, Dc=2m, n=5
R=8m
-
vi
18m
3@6m
On the next example, the geometry of initial shape is shown in
Fig.7, and the green lines
denote the 3D axial lines. Further, the 2D axial lines to unify
the mesh distribution are
located on 3 sides of all triangular
elements. The tensile of the soap film
elements is 3.0kN/m, and the
coefficient of line elements C=1.
The Fig.8 shows the equilibrium
shapes in cases with 3D axial lines
and without 3D axial lines
respectively.
4 Conclusion
In this thesis, a technique for form finding of isotonic
surfaces is shown, and it is to use
the combination of triangular soap film elements and the axial
lines. The characteristics of
this technique and results from the numerical examples are
concluded as follows:
1) Because of the effect of 2D axial lines, node distribution
becomes more rational and
large increment of compulsory displacement can be adopted when
simultaneous control
is used as the incremental method. Therefore, the polyhedral
equilibrium solutions
become more similar to the isotonic surfaces.
2) 3D axial lines can restrict the deformation of soap film
during the process of form
finding. Furthermore, these axial lines can be converted to
cable elements with real
stiffness after form-finding. Therefore, the combination of the
soap film elements and
3D axial lines can be expected to determine a composite
structure of membranes and
cables. In addition, in some cases of boundary shapes, we have
to notice that the mesh
connectivity has influence to determined shape.
Fig.7 Initial plane
[a] Without "3D axial line" [b] With "3D axial line", n=6
Fig.8 The equilibrium shapes
-
vii
TABLE OF CONTENTS
CHAPTER TITLE PAGE
Title Page i
Acknowledgements ii
Abstract iii
Table of Contents vii
List of Figures ix
List of Tables xi
1 Introduction 1
1.1 Motivation and objective 1
1.2 Background 3
2 Tangent Stiffness Method 10
2.1 Outline 10
2.2 Tangent Stiffness Equation 11
2.3 Derivation of Tangent Geometrical Stiffness Matrix with
Energy Principle
11
2.4 Iterative procedure of Tangent Stiffness Method 14
3 Shape Analysis 18
3.1 Element Potential Function 18
3.2 Triangular Soap Film Element 19
3.3 Axial Line Element 20
3.4 Nodal Coordinate Definition 21
3.5 Simultaneous Control 22
4 Numerical Examples and Discussions 26
4.1 Low-rise equilibrium shape by combination of soap film
elements and axial
lines (Load control) 26
4.2 Equilibrium shape by combination of soap film elements and
axial lines
(Simultaneous control) 33
5 Case Study by Using 3D Axial Lines 50
5.1 Case 1 50
5.2 Case 2 54
-
viii
5.3 Case 3 57
6 Minimal Surface 60
6.1 Introduction 60
6.2 Catenoid 61
6.3 Scheck surface 68
7 Conclusion 73
8 Reference 74
-
ix
List of Figures
Fig.1.1 Soap film constrained by a rectangular frame with a
moveable side AB ................. 5
Fig.1.2 Surface tension forces acting on a tiny patch of surface
........................................... 6
Fig.1.3 One curvature of A soap bubble with two surfaces
................................................... 8
Fig.1.4 The process of inflating with air pressure increasing
................................................ 8
Fig.1.5 The process of inflating with air pressure decreasing
............................................... 9
Fig.1.6 Inner pressure Volume curve
................................................................................
9
Fig.2.1 Energy vectors relationship in Tangent stiffness Method
.................................... 12
Fig.2.2 Iterative procedure of Tangent Stiffness Method
.................................................... 16
Fig.2.3 Flow chart of the program for the form finding
...................................................... 18
Fig.3.1 Triangular soap film element
..................................................................................
20
Fig.3.2 Normal direction at a node #i
..................................................................................
22
Fig.3.3 Coordinate system with normal direction and tangent
direction............................. 23
Fig.3.4 The process of simultaneous control
.......................................................................
25
Fig.3.5 Elements connected to node #i
..............................................................................
27
Fig.4.1.1 Initial plane and connectivity
...............................................................................
28
Fig.4.1.2 ~ Fig.4.1.12 Equilibrium shapes and convergent process
of unbalanced force
respectively
..................................................................................................................
29~34
Fig.4.2.1 Initial plane and connectivity
.............................................................................
35
Fig.4.2.2 Inner pressure-volume curve
................................................................................
36
Fig.4.2.3 Equilibrium shapes (Left: 1D analysis; Right: 3D
analysis) .............................. 36
Fig.4.2.4 Equilibrium shapes when the total compulsory
displacement = 6 m ................... 39
Fig.4.2.5 Initial plane and mesh connectivity
......................................................................
40
-
x
Fig.4.2.6 Equilibrium shapes when the total compulsory
displacement= 5m ..................... 41
Fig.4.2.7 Straight line 1 and straight line 2
.......................................................................
42
Fig.4.2.8 The curve of the straight line 1
.............................................................................
43
Fig.4.2.9 The curve of the straight line 2
.............................................................................
44
Fig.4.2.10 Straight line 1 and straight line 2
.......................................................................
45
Fig.4.2.11 The value of n for straight line 1
........................................................................
45
Fig.4.2.12 Equilibrium shapes when the total compulsory
displacement= 5m ................... 46
Fig.4.2.13 The curve of the straight line 1
...........................................................................
47
Fig.4.2.14 The curve of the straight line 2
...........................................................................
48
Fig.4.2.15 Initial plane and connectivity
.............................................................................
49
Fig.4.2.16 Inner pressure-volume curve
............................................................................
50
Fig.4.2.17 Equilibrium shapes of two mesh connectivity
................................................... 51
Fig.5.1.1 Initial plane and connectivity
...............................................................................
52
Fig.5.1.2 Inner pressure-volume curve
..............................................................................
53
Fig.5.1.3 Equilibrium shapes (Left: 3D analysis without 3D axial
lines; Right: 3D analysis
with3D axial lines)
............................................................................................................
55
Fig.5.2.1 Initial plane and connectivity
...............................................................................
56
Fig.5.2.3 Equilibrium shapes
...............................................................................................
58
Fig.5.3.1 Initial plane and connectivity
...............................................................................
59
Fig.5.3.2 Inner pressure-volume curve
..............................................................................
60
Fig.5.3.3 Equilibrium shapes
.............................................................................................
61
Fig.6.1 The surface of catenoid
...........................................................................................
63
Fig.6.2 The relation between Z*and C
.................................................................................
64
Fig.6.3 The relation between Z and r
...................................................................................
64
-
xi
Fig.6.4 Primary shape and the connectivity
........................................................................
65
Fig.6.5 Minimal surfaces of Catenoid
.................................................................................
66
Fig.6.6 Stable solution of Catenoid
...................................................................................
67
Fig.6.7 Unstable solution of Catenoid
...............................................................................
68
Fig.6.7 Element mesh of Scheck surface
...........................................................................
70
Fig.6.8 Minimal surface of Scheck surface
.........................................................................
71
Fig.6.9 Curve comparison of Scheck surface
......................................................................
73
List of Tables
Table 1.1 Surface Area of each polyhedron
...........................................................................
4
Table 6.1 Comparison of c value in stable solution
.............................................................
67
Table 6.2 Comparison of c value in unstable solution
......................................................... 69
-
1
1. INTRODUCTION
The general research question of the study is related to the
design problem of membrane
structures while the aim is to adopt the tangent stiffness
method for generating minimal
surfaces, and form finding of the initial shape with given
boundary shape by combining
axial line elements and triangular soap film elements.
The main concept is based on the simulation of a potential
tensioned membrane such as
a soap film within a pregiven boundary. This study proposes a
new technique to obtain the polyhedron solutions which can express
smooth surfaces by parallel calculation of soap
film analysis only to normal direction and the "2D axial line"
analysis in tangent plane of
the surface. By using this technique, node distribution becomes
more rational and large
compulsory displacement can be adopted when the simultaneous
control is used as the
incremental method. Further, when using the "3D axial lines"
that have degree of freedom
to all of 3 dimensions, we can assume it to be converted to the
real cable elements after
form-finding.
1.1 Motivation and Objective
Plateau's problem is to find the minimal surface with a boundary
shape, and many
researchers have spent their effort to solve the shape of soap
film for over 100 years.
On the other hand, for practical use of isotonic surfaces as an
initial shape, such as for
suspended membrane roofs or pneumatic structures, computational
form-finding methods
have been developed since the end of last century. Recently, we
can determinate an
isotonic surface as precise solution by general type of the
finite element method, in the
case of the application to membrane structures with low-rise and
with simple in-plane
boundary shape.
However, application of membrane structures is not only for
large span roofs. We can
expect the possibility to adopt the isotonic surfaces to the
primary shape of more various
styles of structures, for example, the underwater structures
such as the storage tank of pure
water floating on ocean, underground structures which is so
called geo-membrane', and
space structures on some extraterrestrial planet.
If we apply the shape of the isotonic surface to these new uses
for membrane structures,
the technical consideration for computation will be required as
follows:
(1) Loading condition should be applied not only to constant
inner pressure but also to the
location dependency load such as water or soil pressure.
(2) Sure convergence property should be guaranteed even if in
cases of complicated
boundary shapes.
-
2
(3) Development of a rational mesh division technique to express
smooth shape of curved
surface is expected.
The authors' former paper has concluded that the tangent
stiffness method is very useful
for the form-finding of soap films [1]-[5]. The method uses the
iterative process of the
tangent stiffness equation that derived from the equilibrium
equation between the nodal
forces and the element edge forces. Furthermore, the authors
mentioned that the
simultaneous control, which is an incremental technique, makes
it possible to apply to the
location dependency load listed in above (1). For the problem of
(2), the primary
equilibrium shape constituted by the axial line elements instead
of the soap film elements
makes it easier to get the solutions, even if in case of the
complicated boundary shape. The
axial line elements can produce the solutions very close to
isotonic surfaces, and can let the
mesh distribution be uniform [2], [4].
With regard to problem (3), for instance, when the deformation
of the surface from a
primary shape to the determined shape is too large, dispersion
of the element area also
grows so large. In such a case, we cannot keep 'the
significance' as approximate solutions
for curved surfaces anymore. Now, we discuss about an isotonic
surface as a discrete
solution. Under the assumption of "triangle isotonic element",
the determined solution
should be obtained as "a polyhedral equilibrium shape".
Moreover, originally, the isotonic
surface does not have any stiffness toward normal direction of
the surface, and the position
of every node as discrete solution can be located anywhere on
physical isotonic surface.
Namely, there is the discrete equilibrium solution innumerably
on an isotonic surface.
Therefore, we should keep the proportionate mesh distribution
all through the incremental
process of shape analysis.
Based on the above background, in this paper, the authors
propose another technique
which is the parallel calculation of the soap film analysis and
the axial line analysis. The
technique defines two independent and parallel coordinate. One
has only one axis toward
normal direction of surfaces and that is prepared for the
isotonic surface analysis by
triangular soap film elements. The other has two axes
perpendicular each other and both
are to tangent direction of surfaces. The element forces of the
axial line elements are
projected to these two coordinate axes. Thus this type of axial
line elements has only 2
degree of freedom, so we call it "2D axial line", in this paper.
According to the effect of the
2D axial lines, the mesh distribution becomes more uniform, and
the polyhedral
equilibrium solutions become more similar to the isotonic
surfaces.
In this paper, we also mention another usage of the axial line
elements that can be
converted to cable elements with real stiffness after
form-finding. This type of axial line
elements has degree of freedom to all of 3 dimensions, and we
call it "3D axial line" to
distinguish these axial line elements from 2D axial line.
Therefore, the combination of the
soap film elements and 3D axial lines can be expected to
determine a composite structure
of membranes and cables.
-
3
1.2 Background
1.2.1 Soap bubble and soap film
Soap bubble, like a balloon, is a very thin soap film
surrounding a volume of air. Now,
imagine a simple experiment. If dip a curved metal wire into a
mixture of water and dish-
washing detergent, and then pull it out. After that you blow a
slight air into the soap film, a
beautiful soap bubble forms. No matter what shape a bubble has
initially, it will try to
become a sphere. The sphere is the shape that minimizes the
surface area of the structure,
which makes it the shape that requires the least energy to
achieve.
The Table 1.1 illustrates the surface area of each polyhedron in
nature.
Table 1.1 Surface Area of each polyhedron
Soap films are the classical examples of minimal surfaces. A
minimal surface is more
properly described as a soap film with zero mean curvature. A
soap bubble, due to the
difference in outside and inside pressure, is a surface of
constant mean curvature.
In addition, when a soap film is in a state of equilibrium, the
surface tension is the same
on the surface. There would be an extremely valuable application
in architecture. As we all
know, the knowledge of the form is a very important aspect of
design of structures. Having
the least area property minimal surface is most used for
light-roof constructions, and form-
finding models for membrane structure. Among the surfaces having
the same boundary,
minimal surface is the surface of the least area. Therefore, its
weight is less and the amount
of material is reduced on minimum.
Shape # of sides Volume Surface Area
Tetrahedron 4 1 cubic meter 7.21 square meter
Cube 6 1 cubic meter 6.0 square meter
Octahedron 8 1 cubic meter 5.72 square meter
Dodecahedron 12 1 cubic meter 7.21 square meter
Icosahedrons 20 1 cubic meter 7.21 square meter
Sphere infinite 1 cubic meter 7.21 square meter
-
4
1.2.2 Minimal surface
The Plateau's problem is to find the shape of the minimal
surface constrained by a given
boundary. In 1873 a physicist named Joseph Plateau observed that
soap film bounded by
wire appeared to form minimal surfaces. The problem named after
him.
Minimal surfaces are defined as surfaces of the smallest area
spanned by a given space
curve. The theory of minimal surfaces is a branch of mathematics
that has been highly
developed, particularly recently.
In spite of the fact that it seems that soap film easily solves
mathematical problem of
finding minimal surface for the boundary curve, many
mathematicians devoted much of
their life to solve some basic problems as well as to give
description of minimal surfaces in
mathematics for over 100 years.
The main fields of mathematics contributing to minimal surface
theory are differential
geometry, theory of partial differential equations and calculus
of variations.
In the recent time, as computer has made a great progress in
science, this new
technology enabled researchers, especially in the field of
differential geometry of
mathematicians, to study infinitely periodic minimal surface.
This field has enjoyed a
lively renaissance with the help of modern visualization
techniques.
Theoretical investigation of these surfaces is useful for
application of this knowledge in
further investigation of forms in architecture.
-
5
1.2.3 Energy consideration
In order to obtain an expression for the energy of a soap film,
let us consider a soap film
contained by a rectangular wire frame (Fig. 1.1), with one side
of the frame AB, free to
move in the direction perpendicular to AB. The thickness of soap
film is t. If this side is
initially at a distance x from the parallel fixed side and
undergoes a further displacement
x, maintaining the temperature of the film constant, the work
done against the film
tension force is
(1.1)
Where, is the film tension and is the increase in the volume of
the soap film. From
Eq. (1.1), is seen to be the energy gained per unit increase in
volume. Thus the total
energy necessary to increase the volume of a soap film from zero
to V is given by F,
(1.2)
F is called the free energy at constant temperature. In the case
of is constant and result
becomes
(1.3)
The free energy of the soap film is thus proportional to the
area of the film under the
conditions of constant and t.
When a soap film is in stable equilibrium, any small change in
its area A will produce a
corresponding change in its free energy F, if remains
constant.
Fig.1.1 Soap film constrained by a rectangular frame with a
moveable side AB
A
B
A
B
Y
X X
-
6
1.2.4 Equilibrium of soap film
A soap bubble consists of a thin soap film surrounding a volume
of air. Inside the film is
air or gas at a greater pressure than the external atmospheric
pressure. By examining the
equilibrium on a tiny patch surface of a soap bubble, we can
relate the excess pressure
inside the bubble over the surrounding pressure P, and the film
tension .
If no force acts normal to a tensioned surface, the surface must
remain flat. But if the
pressure on one side of the surface differs from pressure on the
other side, the pressure
difference times surface area results in a normal force. In
order for the surface tension
forces to cancel the force due to pressure, the surface must be
curved, as shown in Fig.1.2.
The weight of a tiny patch surface of fluid can be neglected
compared with either of these
forces. When all the forces are balanced, the resulting equation
is known as the Young
Laplace equation.
(1.4)
Where:
P is the pressure difference.
is surface tension.
Rx and Ry are radii of curvature in each of the axes that are
parallel to the
surface.
Fig.1.2 Surface tension forces acting on a tiny patch of
surface.
x and y indicate the amount of bend over the dimensions of the
patch.
FT
FT
FT
FT
Rxx
Ryy
-
7
In the case of soap bubble, the principal radii of curvature, Rx
and Ry, are equal to its
radius, that is Rx = Ry = r. So the excess pressure across each
surface of the bubbles is,
from the Eq. (4)
(1.5)
For a sphere, the radius of curvature is simply its radius.
(1.6)
Where, is the curvature of the soap bubble.
-
8
1.2.5 Inner pressure-volume curve
When a soap film with a given boundary is inflated under the
inner pressure, one
curvature of a soap bubble has two curved surfaces, as shown in
Fig.1.3.
When the air pressure is applied on the soap film from inside,
as the inner pressure
increases, the volume of soap film and curvature of curved
surface also increase. When the
soap film is in the state of half sphere, the curvature of
curved surface became a maximum
in Fig.1.4. Therefore, before the soap film is in the state of
half sphere, the inner pressure
and volume of soap film is increasing. In this stage, we call it
"expansion with inner
pressure increasing".
Initial state Half sphere
Curvature0 Maximum
Inner pressure0 Larger
Volume0 Larger
Fig.1.3 One curvature of a soap bubble with two surfaces
Fig.1.4 The process of inflating with air pressure
increasing
-
9
However, when the inner pressure is over the state of the
maximum point of its
path, volume is increasing while the curvature of curved surface
and inner pressure is
decreasing, as shown in Fig.1.5. In this stage, we call it
"expansion with inner pressure
decreasing".
Half sphere Determined shape
CurvatureMaximum Smaller
Inner pressureMaximum Smaller
VolumeLarge Larger
By the means of simultaneous control, we can get the determined
shape which are over
the maximum point and have high-rise. The Fig.1.6 shows inner
pressure volume curve
analyzed by simultaneous control.
Fig.1.5 The process of inflating with air pressure
decreasing
Volume (m3)
Expansion with Ip increasing
Inner p
ressure (k
pa)
Fig.1.6 Inner pressure Volume curve
Expansion with Ip decreasing
-
10
2. TANGENT STIFFNESS METHOD
2.1 Outline
This chapter presents the derivation of Tangent Stiffness
Equation and Tangent
Geometrical Stiffness Matrix; also the chapter descibes
iterative process for unbalanced
force equation solution and geometrical nonlinear structure
analysis technique for Tangent
Stiffness Method.
The Tangent Stiffness Equation can be easily derived by
differencial calculus of
Equilibrium Equation that derived from the equilibrium equation
between the nodal
forces and the element edge forces. Element edge force that
works on element end is
completely prescribed by Element Force Equation. In this
equation, element stiffness is
independent from the element displacement; it is caused by the
nonlinearity of the Tangent
Geomtrical stiffness. Also the Element Force Equation precisely
represents element
behavior, avoiding approximation even for complex cases.
Furthermore, by applying Principle of Stationary Potential
Energy, it is possible to
express a symmetric matrix for Tangent Geometric Stiffness.
Using this Tangent
Geometric Stiffness matrix makes possible to express strict
element behavior that
prescribed in the Element Force Equation.
Tangent Stiffness Method avoids the derivation of complicated
nonlinar element
stiffness equation with nonlinear material properties, which is
quite difficult to formulate.
Only strcit Compatibility Equation and Equilibrium Equation are
used for iterative
process to converge the Unbalanced Force.
Itervative solution in Tangent Stiffness Method is
mathematically equal to Newton-
Raphson Method technique. Comparing the Tangent Stiffness Method
to Newton-
Raphson Method applied in Finite Element Method(FEM), it shows
overwhelmingly
high efficiency of Tangent Stiffness Method in convergence
performace.
-
11
2.2 Tangent Stiffness Equation
Let the vector of the element edge forces independent of each
other be indicated by S,
and let matrix of equilibrium which relates S to the general
coordinate system by J. Then
the nodal forces D expressed in the general coordinate follow
the equation:
(2.1)
The tangent stiffness equation is expressed as the deferential
calculus of Eq. (2.1),
(2.2)
In which, K0 is the tangent element stiffness, and it is caused
by the relation between the
increments of axial forces and the nodal displacements. Namely,
K0 provides the elements
own behavior inside of the local coordinate moving with the
rigid body displacements.
KG is the tangent geometrical stiffness caused by the relation
between the changing of
the equilibrium conditions and nodal displacements. The cosine
vectors and the length of
elements construct the equilibrium equation. Changing of cosine
vector means the rotation
of the element. Therefore, KG provides the rigid body rotation
and displacements. is
nodal displacement vector in general coordinate.
2.3 Derivation of Tangent Geometrical Stiffness Matrix with
energy principles
Regarding to Eq. (2.1) and Eq. (2.2), Tangent Geometrical
Stiffness KG can be expressed
as
(2.3)
According to expansion of Principle of Stationary Potential
Energy, and element
force vector obtained from primary balance condition leads into
expression of Tangent
Geometrical Stiffness Matrix.
-
12
V#cV
V#
V
#U
U
U
Stiffness Equation Equilibrium Equation
Nodal
Displacement
Element Force
Element Force Equation Compatibility Equation
Element Deformation
Fig.2.1 Energy vectors relationship in Tangent stiffness
Method
Nodal Force (External Force)
-
13
The Fig.2.1 shows Nodal displacement vector, Element Deformation
vector, Element
Force vector and Nodal Force vector, expressed as scalar
quantities versus energy. The first
quadrant represents Tangent Stiffness Equation, the second
quadrant Equilibrium
Equation, the third quadrant Element Force Equation, and the
last quadrant
Compatibility Equation. In addition, the inner rectangle shows
primary balance condition,
and the outer rectangle presents balance condition after nodal
force increment when
external force has been applied. Here, if Stain Energy is U and
External Energy V is, the
Total Potential Energy is
(2.4)
After deformation, Total Potential Energy can be expressed as
follow:
(2.5)
If balance condition before and after deformation is assumed to
be constant, according to
the Principle of Stationary Potential Energy,
(2.6), (2.7)
In Fig.2.1 shows that is not being influenced by the increment
of nodal displacement
. Therefore,
(2.8)
The differential for increment of nodal displacement for V#, ,
U# and becomes
(2.9), (2.10), (2.11)
(2.12)
Therefore, Eq. (2.7) can be rewritten as follows.
(2.13)
Thus, it is clear that Eq. (2.13) and Eq. (2.1) are
identical.
(2.14)
-
14
According to Eq. (2.2) and Eq. (2.3), the Tangent Geometrical
Stiffness is
(2.15)
If a finite increment of element deformation vector expands as a
quadratic function
, the work quantity that performed by element force S and
increment .
Tangent Geometrical Stiffness KG can easily be expressed by
second differential of ,
according to geometrical quantity and dynamic quantity in
primary balance condition.
Therefore, the Tangent Geometrical Stiffness Matrix from Eq.
(2.15), expressed in
particular form is
(2.16)
2.4 Iterative procedure of Tangent Stiffness Method
The procedure of Tangent Stiffness Method can be implemented in
the way describe
below. The iterative technique proceeds until convergence of
unbalanced force to a strict
balance position by using loading steps. The common procedure
steps are:
Primary displacement:
Primary load:
Load increment:
Displacement for step of iteration (r):
Element deformation vector:
Element force vector:
Element force Nodal Force Conversion Matrix:
Tangent Stiffness Matrix:
The Fig.2.2 shows a convergence concept diagram for Tangent
Stiffness Method. The
calculation goes clockwise. The first quadrant represents
relation of displacement and load.
In Tangent Stiffness Method, nonlinear stiffness equation is not
involved in calculation
process and it is marked as a dotted line on the graph. The
fourth quadrant represents the
condition of compatibility, which expresses relation between
nodal displacements in
-
15
global coordinate system and element deformation vector S from
element deformation
vector. The second quadrant represents Equilibrium Equation,
which is obtained from
Element Force vector and the coordinate transformation for the
current displacement. This
is necessary to calculate in order to seek Nodal Force
vector.
In Fig.2.2, iterative process in Tangent Stiffness Method begins
form given state of
balance, which starts from the point O, and the following steps
explain the iterative
solution process until convergence.
O
A
B
C
D
EF
G
H
I
J
KL
M Z
dS
s
2s
1s
1d
2d
1D
2D
2S 1S
1d
2d
1D
2D
1D
2D
0D
D
O
A
B
C
D
EF
G
H
I
J
KL
M Z
dS
s
2s
1s
1d
2d
1D
2D
2S 1S
1d
2d
1D
2D
1D
2D
0D
D
Fig.2.2 Iterative procedure of Tangent Stiffness Method
Stiffness Equation Equilibrium Equation
Element Force Equation Compatibility Equation
-
16
is obtained by solving Tangent Stiffness Method at point O, for
given value of load increment . (O A B)
Calculate by adding to displacement in primary state.
From displacement obtain strict compatibility equation that is
used to solve element deformation vector . (B C)
From the Element Force Equation , the state of equilibrium, F is
provided by calculation for using the Equilibrium Equation. (C D E
F)
Unbalanced force is calculated by cyclic calculations, from the
difference of load condition and load vector which is in balance
state in displacement position .
In this stage, condition F is considered as the primary state,
and it finds solution by Tangent Stiffness Method which corresponds
to . (F G H)
New displacement for nodes is calculated using the previous
Calculation for the next state of balance L is performed in the
similar way.
(H I J K)
The rest of the steps is repeated continuously until the state
of balance in first quadrant get s closer to Z, and the convergent
solution is obtained for this calculation procedure.
Summary, iterative process in Tangent Stiffness Method can be
expressed as
(2.17)
Thus, from the comparison between iterative process of Finite
Element Method, Eq.
(2.17) and calculation performed by Tangent Stiffness Method, it
is necessary to
formulate or apply any approximation concept. Although it is
almost impossible to express
strict solution using nonlinear stiffness equation in reality,
(actually, it is quite impossible
to be expressed with any analytic technique, and furthermore, it
is not necessary to express
any of it in this method). The results from the iterative
process which was explained
previously, is presented by the dotted line in Fig.3.2 that
shows nonlinear stiffness equation
which is solved strictly while passing through the curved line
in a random step.
In addition, Fig.2.3 is a flow chart for a geometrical nonlinear
structure analysis
program performed by Tangent Stiffness Method.
-
17
INPUT PRIMARY VALUES START
INPUT COMPULSORY DISPALCEMENTS
CAL LOAD INTENSITY
CAL UNBALANCED FORECES
REACTION FORCES AT CONTROL POINT
CONVERGE?
COMPOSE TANGENT SRIFFNESS MATRIX
SOLVER SIMULTANEOUS EQUATIONS
RENEW NODAL POSITIONS
YES
NO
Fig.2.3 Flow chart of the program for the form finding
-
18
3. SHAPE ANALYSIS
3.1 Element potential function
In general type of finite element analysis, the strain is
measured with respect to an
element (local) coordinate and relationship between displacement
and strain is formulated
as a nonlinear equation. In contrast, in the tangent stiffness
method, the relationship
between the vector of the element edge forces and nodal forces
is formulated as linear
geometric stiffness. Therefore, when we need to calculate the
finite strain in all elements, it
is not necessary to solve any complicated nonlinear equation.
Instead, we just calculate
nodal displacements after deformation, and the elements
behaviour in any conditions can
be tracked.
In order to regulate the element behaviour in element (local)
coordinate, we define the
element measure potential which is expressed as the function of
measurement such as
element length or element area. Defining element measure
potential is equal to assuming
the "virtual" elemental stiffness. And it has no relationship
with material's stiffness.
Let element measure potential is P, and let the vector of
elements' measurements whose
component is independent of each other is s
(3.1)
Then we can get the element edge force S.
-
19
3.2 Triangular soap film element
Let the area of triangle is A, and unit membrane tensile is t
(=constant, where is
tensile stress of the soap film element, t is the thickness of
element). The element measure
potential is expressed as the function of element area,
tAP (3.2)
We can get the element edge force by differential calculus of
element measure potential,
)3,2,1(2
1
itr
l
PN i
i
i (3.3)
Where, ri is the distance from the orthocentre of triangle
(point 0) to each node.
As the element edge force can be set to the direction along the
side of triangle, and
therefore, we can use the tangent geometrical stiffness matrix
as same form as one of
triangular truss block. So, the form of KG becomes as
follow:
G2 G3 G3 G2
G G3 G1 G3 G1
G2 G1 G1 G2
k + k -k -kK -k k + k -k
-k -k k + k (3.4)
( ), ( 1,2,3)i
i
Ni
l TGik e - (3.5)
In which, are the components of cosine vector alongside
direction, and e is 33 unit
matrix.
Fig.3.1 Triangular soap film element
)3,2,1()( il
N Tii
i
i
Gi ek
-
20
Considering that the value of element stiffness K0 is much less
than the value of tangent
geometrical stiffness KG [1], we ignore K0, and substitute K0 =
0 to the Eq. (2). From the
above equations, we get the tangent stiffness equation,
1 1S
2 G 2
3 3
U uU K uU u
(3.6)
3.3 Axial line element
The line element is connected with nodal point 1 and nodal point
2. Supposing that the
element measure potential is proportional to the power of length
of line element, the
element measure potential can be expressed as:
nClP (3.7)
The axial line element force can be obtained by differential
calculus of element measure
potential:
1 nnClN (3.8)
Where, C is the coefficient of axial lines.
Let are the components of cosine vector alongside direction, and
we can rewrite the Eq.
(1) as:
N
1
2
U -U
(3.9)
Substituting the Eq. (11) to the above Eq. (2), and make it
matrix.
L1 1T
2 2
U uK
U u (3.10)
2( 2) ( 2)( 2) ( 2)
n n nnCln n
T TL
T TT
e e K
e e (3.11)
For the Eq. (10), in the case of n=1, the element forces become
constant, and for the Eq.
(11), the tangent geometrical stiffness of line elements becomes
the same form as truss
element's. Therefore, form-finding with fixed axial forces can
be realized.
-
21
In addition, in the case of n=2, axial force is proportional to
the length of line element,
and Eq. (12) is linear. However, in the case of n>2,
iterative steps are required because of
nonlinearity. The magnitude of n become larger, the length of
all line elements on the
solution surface tend to be more uniform [2]
3.4 Nodal coordinate definition
3.4.1 Surface normal definition
As shown in Fig.3.2, the node i was connected with m (in this
case, m=6) pieces of
elements (j=1~m). The area of element #j can be written as aj.
When we consider that node
i shares 1/3 of each triangular area where inner pressure
affects, we can express area vector
of each element as,
3
ij
ij
aA (3.12)
And normal directional cosine vector of surface at node i can be
written as follows:
1 1
m m
i
j j
ij ijA A (3.13)
Therefore, the tangent stiffness equation Eq. (8) can be
rewritten as
1 11 1
2 1 1 2
3 31 1
T
T
T
Z zZ zZ z
S
GK (3.14)
Thus, the resultant force on each node can be projected to
normal direction.
Fig.3.2 Normal direction at a node #i
-
22
3.4.2 Tangent plane direction definition
In this paper, the line elements which can make the node
distribution more uniform are
located along the 3-sides of triangular elements. We propose the
independent coordinate
which has two axes perpendicular each other and both are to
tangent direction of surfaces.
By using this coordinate, the axial forces of line elements can
be projected to tangent plane
directions.
As shown in Fig.3.3, the normal direction of node i is i , and
two tangent directions are
i and i , which are perpendicular with normal direction. The
tangent stiffness equation Eq.
(12) can be rewritten as follows:
11 1
1 1 1 1 1
2 22 2 2
2 22
T
T T T
T T T
T
X xY yX xY y
S
GK (3.15)
The Eq. (16) and Eq. (17) are perfectly independent tangent
stiffness equations. We
parallelly solve the Eq. (16) and Eq. (17) to get the
incremental displacement of every node.
After that, we renew the nodal displacements, and calculate the
unbalance forces in new
state. The new iterative calculation is continued until the
unbalanced forces converge.
3.5 Simultaneous Control
3.5.1 Outline
When a soap film with a boundary is inflated under the inner
pressure, the normal load
control is used to find shapes of curved surfaces. As the volume
increases, the inner
pressure and curvature of curved surface also increase. However,
when the inner pressure
is in the state of the maximum point of its p- path, we cannot
get the solutions in
exceeding the maximum point of its path by using this method.
Therefore, we need to
figure out another way to input the increments.
Fig.3.3 Coordinate system with normal direction and tangent
direction
-
23
Simultaneous control is an incremental method, and it is
effective to find shapes which
are over the maximum point and have high-rise. The technique
controls both of
compulsory displacements on a control point and the inner
pressure derived from the shape
in each iteration step at the same time, when it is applied to
shape analyses under inner
pressure. When the unbalanced forced on every node as well as
the average of converted
inner pressure converge, we can get the equilibrium shape with
large volume.
3.5.2 The process of Simultaneous Control
Here, the Fig.3.4 is an example that simultaneous control is
applied to isotonic surface
analysis. The primary shape in Fig.3.4 [a] is a low-rise
equilibrium shape which is the
perfect balanced solution at all the nodes. However, our current
purpose is to obtain the
equilibrium shape with very large volume and high-rise in
Fig.3.4 [c]. This can be
considered as a post-buckling problem in wide sense, because
this large volume solution is
on over the maximum point of its equilibrium path.
To begin with, we give the compulsory displacement of Dc to only
this control point,
and the Fig.3.4 [b] is the shape of the first step of iteration.
Next, we need to calculate the
intensity of inner pressure of P1, which is depending on this
current shape. After this, shape
is changed continually, and also load intensity is modified in
each iterative calculation.
Eventually, when both of the unbalanced force of all nodes and
the reaction force of the
control point converged almost to zero, and the load intensity
also should converge to a
constant value, we can find the perfect equilibrium shape in
Fig.3.4 [c].
Fig.3.4 The process of simultaneous control
[a] Primary shape [b] First step of iteration [c] High-rise
equilibrium shape
-
24
3.5.3 Converted inner pressure
In case of form finding of soap film, which is inflated by inner
pressure, we calculate the
intensity of load in every iterative step as the average of the
resultant of element edge
forces concentrated at every node. Namely, we set the inner
pressure to be balanced for
every node.
As shown in Fig.3.5, the node #i was connected with m (in this
case m=6) pieces of
elements (j=1~m). Let the vector of element edge forces of
element #J connected to node
#i be indicated by Nij, and let matrix of equilibrium which
relates Nij to the general
coordinate system by Jij. Then the nodal forces Uij expressed in
the general coordinate
follow the equation,
(3.16)
Therefore, resultant of element edge force at node #i
(3.17)
The soap film has stiffness only towards normal direction of the
surface and normal
directional cosine vector of surface at node #i can be written
as follows:
(3.18)
Therefore, the resultant of element edge force can be projected
to normal direction as,
(3.19)
The inner pressure P is a distribution load which is applied on
the soap film from inside.
The direction of P is always normal to the changing surface
during all stages of
deformation. So, balance equation to normal direction can be
expressed as,
(3.20)
-
25
Therefore the converted inner pressure PCi (i =1, 2n) can be
written as,
(3.21)
Let n be the number of all the nodes, the average of the
converted inner pressure can be
obtained as,
(3.22)
We use this PAV as the inner pressure of each iteration step, so
eventually the load term
becomes follower force. But we never need special modification
for this follower force.
And it is enough to use the simple symmetric tangent stiffness
matrix as same form as
when we use normal load increment.
i
j=1
j=2
j=m
j=m1
j= j=3
j=
i
j=1
j=2
j=m
j=m1
j= j=3
j=
Fig.3.5 Elements connected to node #i
-
26
4. NUMERICAL EXAMPLES AND DISCUSSIONS
4.1 Low-rise equilibrium shape by combination of soap film
elements and 3D axial
lines (Loading control)
Fig.4.1.1 shows the primary shape and the connectivity. One side
of initial shape is
divided into 18. The green lines denote the 3D axial lines to be
converted into cables after
form finding. Further, the 2D axial lines to unify the mesh
distribution are located on 3
sides of all triangular soap film elements. The tensile of the
soap film element is 3.0kN/m,
and the coefficient of line element C=1. The magnitude of n is
set 1, 2, 4, 6, and 10
respectively in 3D analysis with 3D axial lines.
Fig.4.1.2 ~ Fig.4.1.12 show the equilibrium shapes and
convergent process of
unbalanced force respectively.
Fig.4.1.1 Initial plane and connectivity
18m
3@6m
-
27
3-D analysis (without 3D axial lines)
Tensile of the soap film element : 3.0kN/m
Acceptable minimum unbalanced force: 10-5
kN
1.E-06
1.E-05
1.E-04
1.E-03
1.E-02
1.E-01
1.E+00
1.E+01
1.E+02
1 4 7 10 13 16 19 22 25 28 31 34 37 40
Volume=346.3 m3, Ip=1.0 kpa
Fig.4.1.3 Convergent process of unbalanced forces
Iterative steps
Unbalan
ced fo
rces
Normal direction
Tangent plane direction
Fig.4.1.2 Equilibrium shapes when the volume close to 347m3
-
28
3-D analysis (with 3D axial lines, n=1)
Tensile of the soap film element : 3.0kN/m
Acceptable minimum unbalanced force: 10-5
kN
Volume=347.3m3, Ip=1.077 kpa, n=1
Fig.4.1.4 Equilibrium shapes when the volume close to 347m3
n: Power of 3D axial lines; Ip: Inner pressure
1.E-06
1.E-05
1.E-04
1.E-03
1.E-02
1.E-01
1.E+00
1.E+01
1.E+02
1 4 7 10 13 16 19 22 25 28 31 34 37
Fig.4.1.5 Convergent process of unbalanced forces
Iterative steps
Unbalan
ced fo
rces
Normal direction
Tangent plane direction
-
29
3-D analysis (with 3D axial lines, n=2)
Tensile of the soap film element : 3.0kN/m
Acceptable minimum unbalanced force: 10-5
kN
Volume=347.4m3, Ip=1.19 kpa, n=2
Fig.4.1.6 Equilibrium shapes when the volume close to 347m3
n: Power of 3D axial lines; Ip: Inner pressure
1.E-06
1.E-05
1.E-04
1.E-03
1.E-02
1.E-01
1.E+00
1.E+01
1.E+02
1 4 7 10 13 16 19 22 25
Fig.4.1.7 Convergent process of unbalanced forces
Iterative steps
Unbalan
ced fo
rces
Normal direction
Tangent plane direction
-
30
3-D analysis (with 3D axial lines, n=4)
Tensile of the soap film element : 3.0kN/m
Acceptable minimum unbalanced force: 10-5
kN
1.E-06
1.E-05
1.E-04
1.E-03
1.E-02
1.E-01
1.E+00
1.E+01
1.E+02
1 4 7 10 13 16 19 22
Volume=347.0m3, Ip=1.425 kpa, n=4
Fig.4.1.8 Equilibrium shapes when the volume close to 347m3
n: Power of 3D axial lines; Ip: Inner pressure
Fig.4.1.9 Convergent process of unbalanced forces
Iterative steps
Unbalan
ced fo
rces
Normal direction
Tangent plane direction
-
31
3-D analysis (with 3D axial lines, n=6)
Tensile of the soap film element : 3.0kN/m
Acceptable minimum unbalanced force: 10-5
kN
Volume=348.0m3, Ip=1.6 kpa, n=6
Fig.4.1.10 Equilibrium shapes when the volume close to 347m3
n: Power of 3D axial lines; Ip: Inner pressure
1.E-06
1.E-05
1.E-04
1.E-03
1.E-02
1.E-01
1.E+00
1.E+01
1.E+02
1 4 7 10 13 16 19 22 25
Fig.4.1.11 Convergent process of unbalanced forces
Iterative steps
Unbalan
ced fo
rces
Normal direction
Tangent plane direction
-
32
3-D analysis (with 3D axial lines, n=10)
Tensile of the soap film element : 3.0kN/m
Acceptable minimum unbalanced force: 10-5
kN
1.E-06
1.E-05
1.E-04
1.E-03
1.E-02
1.E-01
1.E+00
1.E+01
1.E+02
1 4 7 10 13 16 19 22 25 28 31 34
Volume=347.6m3, Ip=1.8 kpa, n=10
Fig.4.1.12 Equilibrium shapes when the volume close to 347m3
n: Power of 3D axial lines; Ip: Inner pressure
Fig.4.1.13 Convergent process of unbalanced forces
Iterative steps
Unbalan
ced fo
rces
Normal direction
Tangent plane direction
-
33
4.2 Equilibrium shape by combination of soap film elements and
3D axial lines
(Simultaneous Control)
4.2.1 Effect of the 3D axial lines
As mentioned above, this thesis also mentioned another usage of
the axial line elements
that can be converted to cable elements with real stiffness
after form-finding. This type of
axial line elements has degree of freedom to all of 3
dimensions, and we call it "3D axial
line". In this chapter, we will make three discussions. The
first one is the comparison of the
proposal with 1D soap film analysis without 3D axial lines, and
the sencond is the
influence of power n in the 3D axial lines. At last is influence
of mesh connectivity in
determined shape.
As shown in Fig.4.2.1, the radius direction of initial shape is
divided into 8, and the
circumferential direction is divided into 8. The tensile of the
soap film element is 3.0kN/m,
and the coefficient of line element C=1. For 3D analysis, the
green lines denote the 3D
axial lines to be converted into cables after form finding.
Further, the 2D axial lines to
unify the mesh distribution are located on 3 sides of all
triangular soap film elements.
Simultaneous control is an incremental technique adopted in this
example. The control
point is set in the center of the initial shape, and compulsory
displacement is given along
vertical direction. Compulsory displacement of each step is
given as Dc=0.1m for the 1D
soap film analysis. In contrast, for the 3D analysis, the
compulsory displacement of each
step is given as Dc=1m.
Fig.4.2.1 Initial plane and connectivity
R=8m
Control point
-
34
The Fig.4.2.2 denotes inner pressure volume curve. The
tendencies of curves are so
similar. However, "3D axial line" can restrict the deformation
of soap film during the
process of form-finding. Therefore, larger air pressure is
required to get the equal volume.
The Fig.4.2.3 shows the equilibrium shapes in cases with 3D
axial lines and without 3D
axial lines respectively.
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0 5000 10000 15000 20000 25000
Inner p
ressure (k
pa)
1 DOF, Dc=0.1m
3 DOF, Dc=1m, n=2
n:Power of 3D axial lines, Dc: Increment of compulsory
displacement
Fig.4.2.2: Inner pressure-volume curve
Volume (m)
-
35
R=3.0m, Ip=0.36kpa, V=453.6m
R=3.0m, Ip=0.39kpa, V=467.1m
R=6.0m, Ip=0.61kpa, V=1029.6m
R=9.0m, Ip=0.66kpa, V=1782.8m
R=6.0m, Ip=0.55kpa, V=990.2m
R=9.0m, Ip=0.61kpa, V=1688.7m
-
36
R=12.0m, Ip=0.64kpa, V=2821.1m
R=15.0m, Ip=0.60kpa, V=4241.4m
R=18.0m, Ip=0.55kpa, V=6143.4m
R=12.0m, Ip=0.60kpa, V=2615.3m
R=15.0m, Ip=0.57kpa, V=3830.3m
R=18.0m, Ip=0.52kpa, V=5388.4m
Fig.4.2.3 Equilibrium shapes (Left: 1D analysis; Right: 3D
analysis)
R: Rise of the control point, Ip: Inner pressure, V: Volume
-
37
In Fig.4.2.4, the increment of the compulsory displacement is
given as Dc=2m and
Dc=0.1m respectively for 1D soap film analysis (Fig.4.2.4 [a],
[b]). In contrast, for the 3D
analysis with 3D axial line elements, the increment of the
compulsory displacement of
each step is given as Dc=2m (Fig.4.2.4 [c]).
When the 1D analysis is applied, too large increment makes the
area of triangular
elements around the control point large. Further, because of no
stiffness to the tangent
direction, the uneven distribution of the size of triangles
would be kept all through the
process of the analysis. If the magnitude of increment becomes
larger than 2m, it causes
divergence. Therefore, the compulsory displacement in each
iterative step should be less
than 0.1m in case of the 1D analysis.
On the other hand, the 3D analysis brings the stable calculation
process, even though the
increment is larger than 2m. Compare with 1D soap film analysis,
3-Danalysis can obtain
better equilibrium solution because 2D axial line elements
optimize the node distribution.
Meanwhile, 3D axial line can restrict the deformation of soap
film during the process of
form-finding. Therefore, the determined shape can be expected to
composite structure of
membrane and cables.
[c] 3-D analysis with "3D axial line"
Dc=2m, V= 990 m, n=5 [a] 1-D analysis
Dc=2m, V = 1070 m
[b] 1-D analysis
Dc =0.1m, V= 903 m
Fig.4.2.4: Equilibrium shapes when the total compulsory
displacement = 6 m
n:Power of 3D axial lines, Dc: Increment of compulsory
displacement, V:Volume
-
38
4.2.2 The power of n in 3D axial lines
As previously mentioned, the 2D axial line elements located on 3
sides of all triangular
soap film elements can unify the mesh distribution, and 3D axial
line elements to be
converted into cables after form finding which can restrict the
deformation of soap film
during the process of form finding. Here, we will discuss the
influence of power n in the
3D axial line elements.
Case 1 Assuming that the power of n in the 3D axial line element
force equation is kept the
same in all of 3D axial line elements.
In this situation, every 3D axial line element has the same
stiffness. Here, the green lines
denote the 3D axial lines to be converted into cables after form
finding. Further, the 2D
axial lines to unify the mesh distribution are located on 3
sides of all triangular soap film
elements, as shown in Fig.4.2.5.
In addition, the tensile of the soap film element is 3.0kN/m.
Simultaneous control is
used as an incremental technique in this example. Compulsory
displacement of each step is
given as Dc=1m along the vertical direction and the total
compulsory displacement is 5m.
Here, we will discuss the influence of soap film by 3D axial
lines. The power of n and
coefficient of 3D axial lines C are discussed in this example.
The equilibrium shapes are
shown in Fig.4.2.6.
Fig.4.2.5 Initial plane and mesh connectivity
-
39
[a] Without 3D axial line
V = 800.6m, Dc=1m
[b] With 3D axial line, n=1
V= 809.6m, Dc=1m, C=1
[c] With 3D axial line, n=4
V= 893.9m, Dc=1m, C=1
[d] With 3D axial line, n=1
V= 912.5m, Dc=1m, C=10
[e] With 3D axial line, n=6
V= 1155.5m, Dc=1m, C=1
Fig.4.2.6 Equilibrium shapes when the total compulsory
displacement= 5m
n: Power of 3D axial lines, C: Coefficient of 3D axial lines
V: Volume, Dc: Increment of compulsory displacement
[f] With 3D axial line, n=1
V= 1010.5m, Dc=1m, C=30
-
40
In the case of n=1 and C=1, axial forces of 3D axial line
elements become unit force.
The equilibrium shape is almost the same as in the condition of
without 3D axial lines
(Fig.4.2.6 [a]). However, as n or C become large (Fig.4.2.6[c] ~
[f]), the restriction of soap
film by 3D axial lines become more strong. Therefore, the
combination of 3D axial lines
and soap film elements can be expected to determine a composite
structure of membrane
and cables.
In the Fig.4.2.7, the green lines denote the 3D axial lines, and
straight line 1 is composed
of 3D axial line elements. Here, we pick up straight line 1 and
straight line 2 to discuss the
influence by the power of n, and n is set from 1 to 5.
The Fig.4.2.8 and Fig.4.2.9 denote the curves on the straight
line 1 and straight line 2
respectively compare with different value of power of n in the
3D axial lines. Here,
(4.2.1)
In which, x, y is the coordinate of nodes on straight line 1 and
straight line 2.
Straight line 1
Straight line 2
Fig.4.2.7: Straight line 1 and straight line 2
-
41
In the Fig.4.2.8, as the magnitude of n increases, the
deformation of underneath curve is
restricted away from the green curve which is in the condition
of without 3D axial line
elements. However, the deformation of upper curve is away from
the green curve to
contrary direction. This is because the nodes are perfect fixed
in the boundary of initial
shape, and power of n in the 3D axial lines is kept the same,
which causes the restriction of
lower part is larger than uppers.
0
1
2
3
4
5
6
0 1 2 3 4 5 6 7 8 9
n = 0
n = 5
n = 4
n = 3
n = 2
n = 1
R
Z
Fig.4.2.8 The curve of the straight line 1
Here, "n = 0" means without
3D axial line elements
Underneath
Upper
-
42
Next, the deformation of curve of straight line 2 is
discussed.
In the Fig.4.2.9, as the magnitude of n increases, the curvature
of curve becomes larger.
3D axial lines are only located on 3 straight lines. The soap
film is divided into 3 parts.
When the soap film with a given boundary is inflated by air
pressure from inside, some
parts of surface without 3D axial lines are expanded much
larger.
0
1
2
3
4
5
6
0 2 4 6 8 10 12
n = 5
n = 4
n = 3
n = 2
n = 0
R
Z
Fig.4.2.9 The curve of the straight line 2
Here, "n = 0" means without
3D axial line elements
-
43
Case 2 Assuming that the power of n is a linear function of 3D
axial line element number
m on a straight line and incremental coefficient k.
In the Fig.4.2.10, the green lines denote the 3D axial lines,
and straight line 1 is
composed of 3D axial line elements. Therefore, for straight line
1, the magnitude of n can
be obtained according to the Fig.4.2.11. Where, k is set from
0.1 to 0.5, and m = 8 in this
example.
The Fig.4.2.12 shows the equilibrium shapes with different
magnitude of k when the
total compulsory displacement is 5m.
Straight line 1
Straight line 2
1 2 3 4 5 6 7 8 9
k = 0.1
m = 1 m = 2 m = 3 m = 4 m = 5 m = 6 m = 7 m = 8
n = 1.7 n = 1.6 n = 1.5 n = 1.4 n = 1.3 n = 1.2 n = 1.1 n =
1
k = 0.2 n = 2.4 n = 2.2 n = 2.0 n = 1.8 n = 1.6 n = 1.4 n = 1.2
n = 1
k = 0.3 n = 3.1 n = 2.8 n = 2.5 n = 2.2 n = 1.9 n = 1.6 n = 1.3
n = 1
k = 0.4 n = 3.8 n = 3.4 n = 3.0 n = 2.6 n = 2.2 n = 1.8 n = 1.4
n = 1
k = 0.5 n = 4.5 n = 4.0 n = 3.5 n = 3.0 n = 2.5 n = 2.0 n = 1.5
n = 1
Node 3D axial line element
Fig.4.2.11 The value of n for straight line 1
Fig.4.2.10 Straight line 1 and straight line 2
-
44
[a] Without 3D axial line
V = 800.6m, Dc=1m
[b] With 3D axial line
V= 822.2m, k=0.1, Dc=1m
[c] With 3D axial line
V= 839.6m, k=0.2, Dc=1m
[d] With 3D axial line
V= 861.7m, k=0.3, Dc=1m
[e] With 3D axial line
V= 888.3m, k=0.4, Dc=1m
[f] With 3D axial line
V= 918.5m, k=0.5, Dc=1m
Fig.4.2.12: Equilibrium shapes when the total compulsory
displacement= 5m
n: Power of 3D axial lines, Dc: Increment of compulsory
displacement, V: Volume
-
45
As previously mentioned, if power of n in the 3D axial line
element force equation is
kept the same, the restriction of lower part is larger than
uppers. Thus, the deformation of
underneath curve is different with the deformation of upper
curve.
However, in this example, we adjust the power of n in the 3D
axial line element force
equation to make the restriction of lower part is equal to the
restriction of upper part. The
deformation of curve is shown in Fig.4.2.13. As the magnitude of
n increases, the curvature
of curve becomes larger.
0
1
2
3
4
5
6
0 1 2 3 4 5 6 7 8 9
n = 0
k = 0.1
k = 0.2
k = 0.3
k = 0.4
k = 0.5
Here, "n = 0" means without
3D axial line elements
Fig.4.2.13 The curve of the straight line 1
Underneath
Upper
R
Z
-
46
Next, the deformation of curve of straight line 2 is
discussed.
In the Fig.4.2.14, as the magnitude of n increases, the
curvature of curve becomes larger.
The tendency of deformation of curve is similar to the situation
which assuming that power
of n is kept the same in all of 3D axial line elements.
0
1
2
3
4
5
6
0 2 4 6 8 10 12
n = 0
k = 0.1
k = 0.2
k = 0.3
k = 0.4
k = 0.5
Here, "n = 0" means without
3D axial line elements
Fig.4.2.14 The curve of the straight line 2
R
Z
-
47
4.2.3 In case that different shape is obtained depending on mesh
connectivity
The Fig.4.2.15 shows the same initial shape is meshed by two
kinds of different
connectivity. The radius direction of initial shape is divided
into 8, and the circumferential
direction is also divided into 8. The control point is set in
the centre, and compulsory
displacement is given along vertical direction. The green lines
denote 3D axial line and its
power is n=4. Also in this example, the 2D axial lines to unify
the mesh distribution are
located on 3 sides of all triangular soap film elements. The
tensile of the soap film element
is 3.0kN/m. Fig.4.2.16 illustrates the inner pressure-volume
curve.
Fig.4.2.17 shows equilibrium solutions compared with two kinds
of mesh connectivity.
For 1st mesh connectivity, the surface is inflated with rotation
and it looks "twisted". In
contrast, when the 2nd
mesh connectivity is adopted, the surface is inflated without
rotation.
This is why the shape of the polyhedral solution is depending on
the mesh connectivity.
Therefore, we have to be careful when we prepare the mesh
connectivity in some cases like
this example.
[a] 1st mesh connectivity
Fig.4.2.15 Initial plane and connectivity
[b] 2nd
mesh connectivity
Control point
R=8m
Control point
R=8m
-
48
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
0 1000 2000 3000 4000 5000 6000 7000 8000
1st mesh connectivity, Dc=0.1m, n=4
2nd
mesh connectivity, Dc=0.1m, n=4
Inner p
ressure (k
pa)
n:Power of 3D axial lines, Dc: Increment of compulsory
displacement
Fig.4.2.16 Inner pressure-volume curve
Volume (m)
-
49
R=6.0m, Ip=0.75kpa, V=1139.5m R= 6.0m, Ip= 0.68kpa,
V=1062.5m
R=12.0m, Ip=0.82kpa, V= 4496.3m R= 12.0m, Ip=0.80kpa, V=
3804.2m
Fig.4.2.17: Equilibrium shapes of two mesh connectivity
R= 9.0m, Ip=0.86kpa, V= 2233.8m R= 9.0m, Ip= 0.80kpa, V=
1967.8m
[a] 1st mesh connectivity [b] 2
nd mesh connectivity
R: Rise of the control point, Ip: Inner pressure, V: Volume
-
50
5 Equilibrium shape by combination of soap film elements and 3D
axial
lines Case study (1~3)
5.1 Case 1
Fig.5.1.1 shows the primary shape and the connectivity. One side
of initial shape is
divided into 8. The green lines denote the 3D axial lines to be
converted into cables after
form finding. Further, the 2D axial lines to unify the mesh
distribution are located on 3
sides of all triangular soap film elements. The tensile of the
soap film element is 3.0kN/m,
and the coefficient of 3D axial line C=1 in this example.
Compulsory displacement of each step is given as Dc=1m along
vertical direction for
the 3D analysis with 3D axial lines and 3D analysis without 3D
axial lines respectively.
The Fig.5.1.2 denotes inner pressure volume curve. Obviously,
the tendency of curves is
quite different. "3D axial line" can restrict the deformation of
soap film during the process
of form-finding. Therefore, larger air pressure is required to
get the equal volume.
Control point
8m
Fig.5.1.1 Initial plane and connectivity
-
51
The Fig.5.1.3 shows the equilibrium shapes in cases with 3D
axial lines and without 3D
axial lines respectively.
0
0.2
0.4
0.6
0.8
1
1.2
0 500 1000 1500 2000 2500 3000 3500 4000
Fig.5.1.2: Inner pressure-volume curve
Volume (m)
Inner p
ressure (k
pa)
3 DOF without 3D axial lines, Dc=1m
3 DOF with 3D axial lines, Dc=1m, n=4
n:Power of 3D axial lines, Dc: Increment of compulsory
displacement
-
52
R=5.0m, Ip=0.53kpa, V=784.3m R=5.0m, Ip=0.77kpa, V=873.6m
R=6.0m, Ip=0.57kpa, V=986.7m R=6.0m, Ip=0.86kpa, V=1126.5m
R=7.0m, Ip=0.59kpa, V=1213.0m R=7.0m, Ip=0.94kpa, V=1434.5m
-
53
R=8.0m, Ip=0.60kpa, V=1466.3m
R=9.0m, Ip=0.61kpa, V=1749.6m
R=10.0m, Ip=0.60kpa, V=2065.7m
R=8.0m, Ip=1.0kpa, V=1820.7m
R=9.0m, Ip=1.05kpa, V=2323.4m
R=10.0m, Ip=1.08kpa, V=3010.3m
Fig.5.1.3 Equilibrium shapes
(Left: 3D analysis without 3D axial lines; Right: 3D analysis
with3D axial lines)
R: Rise of the control point, Ip: Inner pressure, V: Volume
-
54
5.2 Case 2
As shown in Fig.5.2.1, the radius direction of initial shape is
divided into 7, and the
circumferential direction is divided into 7. The tensile of the
soap film element is 3.0kN/m,
and the coefficient of line element C=1. The green lines denote
the 3D axial lines to be
converted into cables after form finding. Further, the 2D axial
lines to unify the mesh
distribution are located on 3 sides of all triangular soap film
elements.
Simultaneous control is an incremental technique adopted in this
example. The control
point is set in the center of the initial shape, and compulsory
displacement is given as Dc =
0.1m along vertical direction.
The Fig.5.2.2 shows Inner pressure volume curve.
The Fig.5.2.3 shows Equilibrium shapes in case with 3D axial
lines.
Fig.5.2.1 Initial plane and connectivity
Control point
-
55
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
0 1000 2000 3000 4000 5000 6000
Fig.5.2.2 Inner pressure-volume curve
Volume (m)
n:Power of 3D axial lines, Dc: Increment of compulsory
displacement
Inner p
ressure (k
pa)
3 DOF, Dc=1m, n=4
-
56
R=6.0m, Ip=0.79kpa, V=845.5m R=8.0m, Ip=0.84kpa, V=1308.3m
R=10.0m, Ip=0.84kpa, V=1976.5m R=12.0m, Ip=0.80kpa,
V=2991.3m
Fig.5.2.3 Equilibrium shapes
R: Rise of the control point, Ip: Inner pressure, V: Volume
-
57
5.3 Case 3
Fig.5.3.1 shows the primary shape and the connectivity. One side
of initial shape is
divided into 8. The green lines denote the 3D axial lines to be
converted into cables after
form finding. Further, the 2D axial lines to unify the mesh
distribution are located on 3
sides of all triangular soap film elements. The tensile of the
soap film element is 0.5kN/m,
and the coefficient of 3D axial line C=1 in this example.
Simultaneous control is an incremental technique adopted in this
example. The control
point is set in the center of the initial shape, and compulsory
displacement is given as Dc =
0.1m along vertical direction.
The Fig.5.3.2 shows Inner pressure volume curve.
The Fig.5.3.3 shows Equilibrium shapes in case with 3D axial
lines.
Fig.5.3.1: Initial plane and connectivity
Control point
W =
40 m
L = 40 m
-
58
0
0.02
0.04
0.06
0.08
0.1
0.12
0.14
0.16
0.18
0 5000 10000 15000 20000 25000 30000 35000
3 DOF, Dc=0.1m, n=4
Fig.5.3.2 Inner pressure-volume curve
Volume (m)
n:Power of 3D axial lines, Dc: Increment of compulsory
displacement
Inner p
ressure (k
pa)
-
59
R=8.0m, Ip=0.14kpa, V=16382m R=10.0m, Ip=0.15kpa, V=21718m
R=12.0m, Ip=0.16kpa, V=27316m R=14.0m, Ip=0.17kpa, V=31099m
Fig.5.3.3 Equilibrium shapes
R: Rise of the control point, Ip: Inner pressure, V: Volume
-
60
6. MINIMAL SURFACES
6.1 Introduction
Minimal surfaces are defined as surfaces of minimal area which
are enclosed by a given
fixed boundary. Since centuries, they are one of the oldest toys
of mathematicians. There
exists an enormous amount of knowledge of how to describe and to
generate them. From
the mechanical point of view minimal surfaces are determined by
an isotonic stress field
that can be experimentally simulated by the soap film
analogy.
Minimal surfaces have gradually been translated from the field
of mathematics into the
architectural design research due to their remarkable geometric
properties. As we all know,
the knowledge of the form is a very important aspect of design
of structures. Having the
least area property minimal surface is mostly used for light
roof constructions, and form-
finding models for membrane structure. Among the surfaces having
the same boundary,
minimal surface is the surface of the least area. Therefore, its
weight is less and the amount
of material is reduced on minimum.
The minimal surface form-finding analysis of membrane structures
is a nonlinear
problem. Its strong nonlinearity makes its computation to be a
great challenge, because
calculation usually needs a lot of iterations and sure
convergence of the calculation process
should be guaranteed. In this chapter, the author is to adopt
the tangent stiffness method for
generating minimal surface. The method uses the iterative
process of the tangent stiffness
equation that derived from the equilibrium equation between the
nodal forces and the
element edge forces. This method can be used either for an
approximate form of the
membrane surface in the primary design stage or for an initial
solution for further
computation of real membrane analysis.
Two classical minimal surfaces (Catenoid and Scheck surface)
will be preferred in
follow examples to compare with analytical solutions which are
produced by this method.
The results show that this method is simple, efficient and
reliable with highly satisfactory
accuracy.
-
61
6.2 Catenoid
The Fig.6.1 shows a soap film between two circular rings. The
rings are pulled apart
until the distance between the rings is about 0.6627 times the
diameter, and the soap film
pops. I use the term soap film to denote a surface that does not
enclose a volume, the way a
soap bubble does. A soap film forms a minimal surface, which
means that any small
deformation of the surface would have more area. The mean
curvature of a surface is the
average of the two principal curvatures. Every minimal surface
has zero mean curvature.
The converse, however, is not true. A surface of zero mean
curvature corresponds to a
critical point (a local minimum or a saddle point). In other
words, some surfaces with zero
mean curvature are not minimal.
The Catenoid is a minimal surface of rotational symmetry which
can be expressed by
formulation as follows:
(6.1)
Where, c > 0 is a constant and cosh(x) = (ex + e
-x)/2 is the hyperbolic cosine.
Thus, every surface of revolution with r = c cosh (z/c) has zero
mean curvature. Assume
that the rings have radius r =1, and here z are replaced at z =
z*. Then z* and c are
related by
(6.2)
While we cannot solve for c as a function of z*, we can solve
for z* as a function of c:
(6.3)
Fig.6.1 The surface of catenoid
-
62
The Fig.6.2 plots
. Note that for z* < 0.6627, there are two values of c.
The blue curve shows c corresponding to catenoids that are not
minimal surfaces, even
though they have zero mean curvature. These soap films are
unstable and will pop. The red