Professional Documents
Culture Documents
by
Ashkan Bassandeh
Master of Engineering
Sydney, Australia
August 2012
COPYRIGHT STATEMENT
‘I hereby grant the University of New South Wales or its agents the right to
archive and to make available my thesis or dissertation in whole or part in the
University libraries in all forms of media, now or here after known, subject to the
provisions of the Copyright Act 1968. I retain all proprietary rights, such as patent
rights. I also retain the right to use in future works (such as articles or books) all
or part of this thesis or dissertation.
I also authorise University Microfilms to use the 350 word abstract of my thesis in
Dissertation Abstract International (this is applicable to doctoral theses only).
I have either used no substantial portions of copyright material in my thesis or I
have obtained permission to use copyright material; where permission has not
been granted I have applied/will apply for a partial restriction of the digital copy of
my thesis or dissertation.'
Signed …………………………………………….......................
Date ……………………………………………...........................
AUTHENTICITY STATEMENT
‘I certify that the Library deposit digital copy is a direct equivalent of the final
officially approved version of my thesis. No emendation of content has occurred
and if there are any minor variations in formatting, they are the result of the
conversion to digital format.’
Signed …………………………………………….......................
Date ……………………………………………...........................
Page | ii
ORIGINALITY STATEMENT
‘I hereby declare that this submission is my own work and to the best of my
knowledge it contains no materials previously published or written by another
person, or substantial proportions of material which have been accepted for
the award of any other degree or diploma at UNSW or any other educational
institution, except where due acknowledgement is made in the thesis. Any
contribution made to the research by others, with whom I have worked at
UNSW or elsewhere, is explicitly acknowledged in the thesis. I also declare
that the intellectual content of this thesis is the product of my own work,
except to the extent that assistance from others in the project's design and
conception or in style, presentation and linguistic expression is
acknowledged.’
Signed ……………………………………………..............
Date ……………………………………………..............
Page | iii
ABSTRACT
Load paths can define how a structure performs its intended load carrying
recent years, a novel method to compute these paths has emerged in the
literature using finite element analysis yet common FEA packages do not have
MATLAB or FORTRAN).
been achieved by overlaying of link and shell elements in the commercial finite
Page | iv
The work is then extended to membrane structures. Static structural analysis
using finite element geometric nonlinear analysis due to the large deflection is
conducted for a rectangle membrane under uniform pressure. Shell and link
elements in the model represent the main cloth and possible carbon fibre paths
link elements bridging the corners remained in the corner fixed structure. The
final application of this research is cruising boat sails which are analyzed with
the proposed nonlinear programming formulation. The results from this work are
Finally, this work concludes that defining load paths using overlaid link and shell
for reinforcing carbon fibre paths. This thesis has focused on developing the
Page | v
ACKNOWLEDGMENTS
There are a number of people without whom this thesis might not have been
supervisor, Dr Garth Pearce for his inspiration, and his great efforts to explain
good ideas.
My heartful thanks go to my co. supervisor Emeritus Professor Don Kelly for the
provided tireless support and advice across all aspect of the project and has
personally.
My sincere thanks also go to Ravi Shankar for his help and supports and for the
many discussions we have had over a year on static analysis and also to Ehsan
Chavoshi from IRANSYS who has provided me many insights on the modelling
of membrane structures.
My final and most sincere thanks go to my family, Sedighe and Baraatali who
Page | vi
school and have shared their love and advice with me in all the time constantly.
To my sister, Mojgan and her lovely husband, Mehrab, I thank them for their
wonderful support not only during this research but also over my entire life
being far from home country. To my sister, Mandana, who has always widened
Page | vii
TABLE OF CONTENTS
Abstract ............................................................................................................................ iv
Acknowledgments ............................................................................................................ vi
Table of Contents ...........................................................................................................viii
list of Figures ................................................................................................................... xi
list of tables ..................................................................................................................... xv
Abbreviations and Acronyms ......................................................................................... xvi
Nomenclature ................................................................................................................ xvii
CHAPTER 1. .................................................................................................................... 1
1.1. Thesis background ........................................................................................................ 1
1.2. Thesis objective ............................................................................................................ 4
1.3. Thesis outline ................................................................................................................ 4
1.4. Research papers ............................................................................................................ 6
CHAPTER 2. .................................................................................................................... 7
2.1. Introduction ................................................................................................................... 7
2.2. Load path ...................................................................................................................... 8
2.3. Implementation of load paths ...................................................................................... 16
2.4. Structural Optimization:.............................................................................................. 20
2.4.1. Structural Optimization method .......................................................................... 21
2.4.2. History of structural optimization ....................................................................... 24
2.4.3. Techniques of topological optimization .............................................................. 25
2.5. Conclusion .................................................................................................................. 33
CHAPTER 3. .................................................................................................................. 35
3.1. Introduction ................................................................................................................. 35
3.2. Using load path to determine force transfer within the solution boundaries............... 37
3.2.1. Load path method clarification: .......................................................................... 41
3.2.2. Load path vector: ................................................................................................ 41
3.2.3. Load path integration: ......................................................................................... 47
3.2.4. Cantilever plate example:.................................................................................... 50
Page | viii
3.2.5. Example of an L-Shape structure: ....................................................................... 53
3.2.6. Properties of the paths ......................................................................................... 56
3.2.7. Structural configuration ...................................................................................... 62
3.2.8. Discussion ........................................................................................................... 63
3.3. Exploration of load paths in trusses ............................................................................ 63
3.3.1. Gradual reduction based on force level ............................................................... 64
3.3.2. Example of a one bay truss ................................................................................. 69
3.3.3. Example of a two bay truss: ................................................................................ 71
3.3.4. Example of a multi-bay truss .............................................................................. 76
3.3.5. Discussion ........................................................................................................... 78
3.4. Conclusion .................................................................................................................. 79
CHAPTER 4. .................................................................................................................. 81
4.1. Introduction: ................................................................................................................ 81
4.2. Theory of shells and plates: ........................................................................................ 84
4.2.1. General concept................................................................................................... 84
4.3. Modelling of a clamped circular plate: ....................................................................... 86
4.3.1. Material characteristics ....................................................................................... 87
4.3.2. Element ............................................................................................................... 87
4.3.3. Boundary conditions, and loading ...................................................................... 89
4.3.4. The solution ........................................................................................................ 91
4.3.5. Post processing.................................................................................................... 92
4.4. Overlaying meshing method ....................................................................................... 94
4.5. Modelling of a clamped rectangle model:................................................................... 95
4.5.1. Material characteristics ....................................................................................... 96
4.5.2. Element ............................................................................................................... 96
4.5.3. Boundary conditions and loading ....................................................................... 98
4.5.4. The solution ........................................................................................................ 99
4.5.5. Post processing without overlaying link elements ............................................ 100
4.5.6. Post processing with overlaying link elements ................................................. 101
4.5.7. Determining the load paths ............................................................................... 103
4.6. Modelling of a corner fixed rectangle model: ........................................................... 105
4.6.1. Material characteristics ..................................................................................... 106
4.6.2. Element ............................................................................................................. 106
4.6.3. Boundary conditions, and loading .................................................................... 107
4.6.4. Determining the load paths ............................................................................... 108
Page | ix
4.7. Conclusion: ............................................................................................................... 109
CHAPTER 5. ................................................................................................................ 111
5.1. Introduction: .............................................................................................................. 111
5.2. Sail as a membrane structure .................................................................................... 112
5.3. Finite element model of the sail ................................................................................ 113
5.3.1. Definition of the geometry ................................................................................ 113
5.3.2. Material characteristics ..................................................................................... 115
5.3.3. Element ............................................................................................................. 116
5.3.4. Boundary conditions, and loading .................................................................... 118
5.3.5. Results and Post processing .............................................................................. 122
5.4. Load path determination ........................................................................................... 123
5.4.1. Overlaid elements ............................................................................................. 123
5.4.2. Effect of boundary conditions on load paths determination ............................. 127
5.5. Conclusion: ............................................................................................................... 139
CHAPTER 6. ................................................................................................................ 140
6.1. Summary of findings................................................................................................. 140
6.2. Recommendations for Future Work .......................................................................... 143
References ..................................................................................................................... 145
APPENDIX.A……………….…………………...………………………………….….......…151
APPENDIX.B……………….…………………...………………………………….……....…160
Page | x
LIST OF FIGURES
Figure 2-1: Load path in a pinned connection (Kelly and Tosh, 2000) ............................ 9
Figure 2-2: Principal stress directions (Kelly and Tosh, 2000) ...................................... 11
Figure 2-3: Zero force components in the 𝑥 direction on a load path wall (Kelly and
Tosh, 2000) ..................................................................................................................... 12
Figure 2-4: Load path in a plate with a hole under tension in 𝑥 direction (a) and 𝑦
direction (b) (Kelly and Tosh, 2000)............................................................................... 13
Figure 2-7: Test for straightening of load path (Kelly et al., 2010) ................................ 18
Figure 2-8: Transferred force to a node for 3D truss (Harasaki and Arora, 2001) ......... 19
Figure 2-9 Typical size optimized structure in initial given structure (a) and final
optimized design (b)(Christensen and Klarbring, 2008) ................................................. 22
Figure 2-10: Typical shape optimized structure (Christensen and Klarbring, 2008) ...... 22
Figure 2-11: Two dimensional optimization (Christensen and Klarbring, 2008) ........... 23
Figure 2-12: Topology optimization of a two bay truss (Harasaki, 2000) ...................... 24
Figure 2-13: Design domain (Querin et al., 1998; Huang and Xie, 2011) ...................... 28
Figure 2-14: Topology optimized design with BESO (Huang and Xie, 2011) ............... 29
Figure 3-1: Load transfer in a pinned joint(Kelly and Tosh, 2000) ................................ 38
Figure 3-7: sample pointing vector at the solid element centre ...................................... 49
Figure 3-30: Reduction process for the one bay truss ..................................................... 71
Figure 3-32: The reduction process of the two bay truss ................................................ 73
Page | xii
Figure 3-33: Clarification of force equilibrium in node 4............................................... 75
Figure 3-36: Final design for the multiple bay structure ................................................ 78
Figure 4-4: Point load in the centre and enforced displacement in the edge .................. 90
Figure 4-9: Shell element size convergence for the rectangle plate................................ 97
Figure 4-18: load path for a masonry dome (O’Dwyer, 1999) ..................................... 105
Figure 4-21: Possible load path in the corner fixed rectangle ....................................... 108
Figure 4-22: Alternative load paths for a groined vault (O’Dwyer, 1999) ................... 109
Page | xiii
Figure 5-1: Components of a boat sail, adopted from (Jazzmanian, 2006) .................. 114
Figure 5-11: Link elements remaining after the 50 iterations ....................................... 125
Figure 5-16: (a) Mesh generation and boundary conditions (b) Load paths defined by
the link removal algorithm. ........................................................................................... 130
Figure 5-18: load path with force in the head ............................................................... 132
Figure 5-19: load path with force in the head and clew ................................................ 133
Page | xiv
LIST OF TABLES
Table 5-1: Some common materials used for sails ....................................................... 115
Page | xv
ABBREVIATIONS AND
ACRONYMS
Term Definition
TF Transferred force
WP Woven Polyester
Page | xvi
NOMENCLATURE
Term Definition
Celsius
𝑎 Radius
𝐸 Young’s modulus
𝐸𝑅 Evolutionary rate
𝐸 Scalar field
𝑒 Element
𝐹 Force
𝐹 Transferred force criterion
Internal Force
h Thickness
𝐼 Load flow
Unit vector
𝑗 Unit vector
𝐿 Length
𝑁 Node
𝑞 Pressure
𝑅𝑅 Rejection Ratio
Page | xvii
𝑠 Increment value
Load path contribution scalar
Developed Load path contribution scalar
𝑉 Pointing vector
Velocity
𝑤 Deflection
𝑥 X direction
𝑦 Y direction
𝑧 Z direction
Normal stress
𝑒 Local stress
Shear stress
Change
Removal ratio
Boundary condition
Solution Domain
Poisson’s ratio
Air density
Velocity
, , , Angle
Page | xviii
CHAPTER 1.
INTRODUCTION
often used in conjunction with structural analysis and design since the
load, the identification of the load flow from the structure to the support aids to
Investigation of the load path in a structural design also broadens the vision of
the designers to anticipate how a load flow will be altered if the structure is
damaged.
A review of the literature however reveals that historically there has been no
components such as a joint subject to a single axial load, the force results can
be followed from one component to the next. In texts such as (Flabel, 1997) the
insight gained is used to warn of bending introduced by offset of the loads from
al., 1969) is to use the stress flow directly to predict and interpret the load
paths. Principal stress vectors can give some qualitative information but plots of
the principal stress in a structure do not identify the path for transfer of a shear
assemblies. Simplified beam and frame finite element models allow force
sections cut across the structure and the load transfer can be followed manually
by the designer reviewing the results. However, all these methods are labour
intensive and rely on the designer’s skill to create the models and interpret the
stress fields. As a result, it is hard to find two methods that give similar
defining load paths (Ullman, 1992; Kelly et al., 2001; Waldman et al., 2002).
segment of the path between the components of shear stress and normal
stress. This theory is able to calculate the local orientation using finite element
Page | 2
Chapter1Introduction
The aim of this thesis is to investigate the application of load paths to identify
trajectories for reinforcement in tension structures such as the shade sails and
wind sails. These structures are typically made from cloth that creates a large
surface to shade an area from the sun or harness pressure from the wind. Both
applications have local supports to which forces must be drawn from the global
structure. This means the fabric of the surface needs to be locally thickened to
carbon fibre is the preferred reinforcing material due to its high stiffness and
strength. Part of this thesis will therefore focus on fibre reinforcement of these
structures.
The first step in this research is to develop method to utilise the results of a
finite element analysis to define the fibre trajectories. In the early stages of this
continuum solids and plates to allow load paths to be followed through truss
membrane and the fibres represented by axial force elements. The aim of the
The application to shade and wind sail structures is set as the goal to define the
structures are dependent on the topology of the truss, the algorithm to find the
evolutionary design procedure (Xie and Steven, 1997) is set up. Relative nodal
Page | 3
Chapter1Introduction
structure. The aim of the algorithm is to provide a topology that follows the
This work therefore aims to determine the load path in tension structures under
The outline of the remaining chapters of the thesis is given in this section.
and describes how these methods have been developed and implemented in
the available literature in the area of the current study on this topic and
Page | 4
Chapter1Introduction
and integrated into ANSYS. ANSYS is selected because it is the main FEA
package utilized for teaching and research at the University of New South
Once the analysis is performed stress resultants of the elements are exported
to the user defined codes and the load path vectors are generated for plotting.
Link elements are created and imported back into ANSYS. The links enable the
vectors to be visualized by plotting the finite element mesh without requiring the
algorithm to provide a contour plot capability. This chapter then extends the
membranes.
element and four node shell element is developed to define the load paths in
uniform static pressure for the wind sail. Shell and link elements in the model
represent the main cloth and possible carbon fibre paths respectively.
reinforce cruising boat sails using the overlaying method and structural
conditions using over population of the link elements and then the algorithm of
work.
Page | 5
Chapter1Introduction
Kelly, D., Reidsema, C., Bassandeh, A., Pearce, G. and Lee, M. (2011).
"On interpreting load paths and identifying a load bearing topology from
finite element analysis." Finite Elements in Analysis and Design 47(8):
867-876.
Kelly, D., Pearce, G., Ip, M. and Bassandeh, A. (2011). Plotting load
paths from vectors of finite element stress results. NAFEMS, Boston
The following paper being prepared for the ICCM2012 conference summarises
Page | 6
CHAPTER 2.
LITERATURE SURVEY
2.1. Introduction
This literature survey aims to review how load paths and the application of load
the history and the evolution of the load path theory. An attempt is also made to
review techniques that have been developed for the visualisation of load paths
survey that cites all references relating to the concept of load path. However, it
does aim to identify the current status of the load path method as a useful
Load paths are used to identify the flow of a force from the application point to
the reaction point in the solution domain. Engineers have been able to visualize
how a load is transferred from its application point to the support and have
have been derived and various visualizations and interpretations have been
defined. However, in spite of all these attempts and conducted studies, there
the load paths in the structure. In addition, none of these methodologies have
thorough insight into how a structure is performing its intended load carrying
functions.
Load paths in structural design have been studied over the last five decades. In
and Osgood (Kermode, 1964; Osgood, 1970; Osgood, 1982) however, a clear
method for visualization was not defined by these engineers. The concept of
force flows was further introduced by French (French, 1992) as “A very helpful
emphasized that making the paths as straight lines can minimize bending. This
explain the problem. Later Juvinall in 1991 reviewed force flows or “load paths”
for “line of forces” in a pin loaded structure under tension (Juvinall and Marshek,
Page | 8
Chapter2Background Theory
1991) however, the finding did not have mathematical basis. Figure 2-1 depicts
the load path in a pinned connection structure (Kelly and Tosh, 2000):
Figure 2-1: Load path in a pinned connection (Kelly and Tosh, 2000)
Here Fxa and Fxb represent the tension force in x axis. A similar approach was
conducted study, forces were treated as fluids and it was concluded that
through shear (Ullman, 1992). This method was employed in another work to
minimize the stress concentration (Budynas, 1977) and later, the concept of
load flow was discussed by (Singh, 1996) where the force transfer through a
structure occurs as “flux”. The “load path” of a force that flows through a
Page | 9
Chapter2Background Theory
Any curvature of the load lines represents shear and the creation of
bending moments in structures
use load paths as an essential criteria in the design of a structure and its
identify load paths. The first successful study of how to plot load path
Kelly and Tosh (Kelly and Tosh, 2000) and it was later implemented and further
discussed in a typical aircraft structure by Kelly 2001 (Kelly et al., 2001). Some
insight for a load path through a continuum structure can be obtained from
stress vectors are the vectors aligned with the maximum stress contours and
can be found from the post processing from the available commercial finite
element packages in the market. However principal stress vectors do not align
with the direction of transfer of shear loads and a more general technique has
to be developed.
From the theory of elasticity, in plane normal stress and the shearing
with 𝑥 axis, can be derived from Equation 2-1 & Equation 2-2 (Boresi et al.,
2010):
𝑠 𝑠 𝑠 𝑠 Equation 2-1
Page | 10
Chapter2Background Theory
𝑠 𝑠 𝑠 𝑠 Equation 2-2
Equating the shear stress to zero delivers the principal direction which is
𝑎 Equation 2-3
From this definition, the stress paths or stress trajectories are defined as the
lines parallel to the maximum normal stress vectors with the angle of ,
Figure 2-2, and can reflect reasonable paths if either the shear stress
Tosh, 2000). The load path is correct to the left and beside the hole but does
not visualize the transfer of the force to a bearing load behind the hole.
The idea of visualization of load flow in the structures which was introduced
initially by Kelly and Tosh (Kelly and Tosh, 2000) and was comprehensively
reviewed later (Kelly et al., 2010) indicate that the flows of loads can define
paths along which, loads remain constant as they traverse the solution domain.
Although stress analysis does not obey the continuity law in the sense of fluid
flow, continuity can be applied to the components of the force in an arbitrary set
Page | 11
Chapter2Background Theory
𝐹
𝐹
-3: Zero force components in the 𝑥 direction on a load path wall (Kelly and Tosh, 2000)
Figure 2
Load paths define boundaries in such that no load contribution in the specified
𝐹 𝐹 Equation 2-4
Accordingly, the equilibrium in any segment (along load path wall) is achieved
𝑠 𝑠 Equation 2-5
When the equation is resolved for 𝑦 direction, the equilibrium can be modified
as (Equation 2-6):
𝑠 𝑠 Equation 2-6
The theory of load paths to calculate load flow orientations using finite element
solutions was implemented by Kelly and Elsley (Kelly and Elsley, 1993) and
Page | 12
Chapter2Background Theory
plates. This method was linked to elastic energy flow. Where the orientation
𝑎 Equation 2-7
In finite element analysis (FEA), the vectors field can be evaluated at the centre
the constraints required to suppress rigid body motion but it could not explain
From the definition of load path, the load trajectories can be mapped in two
path can be determined in the direction that designer desires. Figure 2-4
depicts the load paths in and 𝑦 directions in a plate with a hole under tension in
the 𝑥 direction.
(a) (b)
Figure 2-4: Load path in a plate with a hole under tension in 𝑥 direction (a) and 𝑦 direction (b) (Kelly
and Tosh, 2000)
Page | 13
Chapter2Background Theory
and displacement tensors were visualized based on elastic energy flow which is
Equation 2-8
where represents load flow in the structure and is the displacement tensor.
Load paths have been also focussed for structural optimization purposes. The
stiffness and transferred load inside structures, was refined by (Harasaki and
Arora, 2001; Arora and Harasaki, 2004). The authors introduced new concepts
these key concepts. The concepts were also studied and implemented for
topology optimization in the work by (Harasaki and Arora, 2001; Harasaki and
Arora, 2002). This work was also continued by Wang (Wang et al., 2010)
measuring the transferred force in rivets and gussets on joints of truck frames.
Page | 14
Chapter2Background Theory
actual vehicle, the distribution of 𝐸 was measured by obtaining the load transfer
from the rear suspension to the front. It was concluded that the load transfer
initiated from rear did not coincide with the load transfer initiated from the front
under torsional loading. This concept was developed to tailor stiffness to reduce
method can only provide qualitative directions of the load flow and with no
ablility to trace any trajectories from the application point to the reaction point.
force. However, this does not show critical stress regions under loading of
tension or compression.
load path determination based on a scalar . This term could measure the
calculate , the solution domain is initially loaded under an external force and
be obtained at any point in the structural domain by the following method. The
displacement degree of freedom at that point is fixed and then the problem is
solved for the other displacements at other points. The gradient of this scalar
field is defined as the stiffness lines and the steepest lines are considered as
Page | 15
Chapter2Background Theory
Load paths can be incorporated into most design processes to aid the designer
to determine the transfer of load through a structure. Load path contours may
indicate the optimal fibre orientation in a fibre reinforced composite layup and
geometric defects. In the early twentieth century, Michell developed a theory for
designers. Figure 2-5 illustrates the Michell structure for a tip loaded
cantilevered beam.
(Richards and Chan, 1966) and was applied to composite materials. The
feasibility of using this theory was investigated later through numerical studies
can reduce these stress concentrations by 25% in the pin-loaded hole cases
(Lackman and Ault, 1967), they have been employed this technique for fibre
Page | 16
Chapter2Background Theory
1972; Jones and Platts, 1998). The fibre steering was achieved by placing dry
fibre onto prepared fabric or resin film in a highly optimized laminate along the
stress trajectories obtained from FEA. At the most efficient joints (up to 25 % of
fibres are placed in load paths patterns) the fibres were placed in these
in carbon fibre or fibreglass laminates were also examined by (Li et al., 2006)
Figure 2-6 shows a fibre steering pattern by this method. The authors could
fibre along the load paths based on FEA analysis. It was found that the load
reducing the ratio of the joint width to the bolt hole diameter.
combine the pointing vectors in order to define a topology to carry the loads.
Based on the definition of load path considered by the authors, a vector plot of
Page | 17
Chapter2Background Theory
stress components determined the load flow across the domain and a contour
plot tangent to these pointing vectors identified paths along which a component
of the load remained constant. Since stress is a second order tensor and is
defined on an orthogonal set of axes, the vector plots could define separate
paths for load transfer in each direction. The algorithm modified the modulus of
the material based on the magnitude of the load path pointing vector on each
element. The algorithm was able to straighten the path indicated in the left hand
image in Figure 2-7, reducing bending moments and giving the straight path in
the second image. The form of the path during the iteration is indicated in the
third image. The significance of the plots and feasibility of the algorithm were
examples such as bolted joint, racing car and yacht hull (Kelly et al., 2010).
Figure 2-7: Test for straightening of load path (Kelly et al., 2010)
An algorithm was also developed into sketching the topology of a structure from
plots of the load paths. Sketching topological optimization for the given structure
from load paths was a helpful achievement based on the definition of load flow
in the structures.
structures was established by Harasaki and Arora (Harasaki and Arora, 2001).
Page | 18
Chapter2Background Theory
and also rectangle frames. The force transfer through the structure was
structure. Figure 2-8 shows the transferred force to node 5 for 3D truss in the
𝑧
𝑦
-8: Transferred force to a node for 3D truss (Harasaki and Arora, 2001)
Figure 2
This work revealed that the transferred force magnitude was different from the
force applied to the structures, though the models engaged were quite simple.
Some suggestions about re-designing of the structures were given that could be
method in order to obtain size and topology optimization. Indirect methods were
also found to deliver similar solutions for different starting designs and
Arora, 2001).
Wang and others (Wang et al., 2010) presented load paths in a truck cab under
Page | 19
Chapter2Background Theory
load paths along the floor member. To perform linear analysis using , the
stage of collision using LS-DYNA and NASTRAN as the finite element package.
The new term allowed the authors to estimate the effect of newly designed
stiffeners in the cab which could convey the load from the wall surface of the
tunnel into the rear part of the cab leading to better performance of the structure
under frontal collision. The load paths presented in the study were quite
different to path reported by other methods from what was previously reported.
the design problems across all the fields. In the current competitive international
market, the companies can only survive if they can present cost optimized
methods that can fulfil and improve the quality and reliability (such as light
weight and high performance) below the certain cost limit should be used
Page | 20
Chapter2Background Theory
constraints (Yang and Chuang, 1994). As this structural optimisation can offer
The general trend in vehicle and aerospace design has been towards more
the designer to supply designs with sufficient strength, weight and safety
distribution) within the design domain such that it minimizes the cost and meets
find the optimal design by changing the size variables such as the cross-
sectional dimensions of truss and frames or the thickness of plates. This is the
structure and final size optimized design have been shown in Figure 2-9(a) and
(a) (b)
Figure 2-9 Typical size optimized structure in initial given structure (a) and final optimized design
(b)(Christensen and Klarbring, 2008)
conceptual designs of structures, and has been investigated for many years. It
has been enhanced extensively to a viable level and has been implemented in
CAE (Bendsøe and Kikuchi, 1988). Figure 2-10 indicates a typical shape
and can be more cost effective than shape optimization. Figure 2-11
search for the optimal spatial order and connectivity of the bars compare to
location and geometries of cavities and holes in the design domains (Huang
and Xie, 2010). Figure 2-12 shows an example of topology optimization of a two
bay truss. Figure 2-12(a) and Figure 2-12(b) present the original domain and
Page | 23
Chapter2Background Theory
(a) (b)
Figure 2-12: Topology optimization of a two bay truss (Harasaki, 2000)
1996).
Structural optimization has been pointed out for more than 100 years; however
there was no significant progress for more than five decades since the initial
introduction of the idea. In fact the improvement started with investigating the
were used to gain a weight optimum design for a rail road car with stress and
attention must be paid for buckling effects if non uniform design is considered
(Haftka and Prasad, 1980). In 1981, the procedures were extended to solid
plastic plates and several methods to optimize axisymmetric plates with design
1981).
Page | 24
Chapter2Background Theory
use of load path methods have been most common techniques in this area.
Homogenization method:
One of the main approaches for structural design that considers variable
with micro- scale voids is introduced and the topology optimization problem is
defined. Then by seeking the optimal porosity using the optimal and satisfactory
criteria, the final design would be achieved (Wang et al., 2003). This method
technique was introduced which did not need re-meshing in the topology
material having microscopic voids (Bendsøe and Kikuchi, 1988). This approach
the design was verified by Suzuki (Suzuki, 1991). This method was extended
later for multiple material layout problems (Bruggi, 2008). This area was also
studied on bicycle and floor beam problem (Haber et al., 1996) and was
1994). An algorithm called Solid Isotropic Material with Penalization (SIMP) for
Page | 25
Chapter2Background Theory
suggested in 1988 which was the starting point of the outstanding work by
limited since this theory uses the asymptotic expansion and the assumption of
may vary slowly when the solution is close to the initial equation. Although this
method is valuable for composite materials, it may not yield the intended results
design with pores in the material that make the structure not manufacturable
and numerical instabilities may also introduce ‘ non- Physical ‘ artefacts in the
results and thus make the design sensitive to variation under loading (Wang et
al., 2003).
A new era of optimization called Evolutionary Method (EM) was opened in 1993
(Xie and Steven, 1993). This method was simple for shape and layout
optimal design. The method has been developed for various problems of the
structure and subsequently to remove some elements with the least contribution
(Huang and Xie, 2010). Additionally, this study was based on Von Mises stress
in which elements with the magnitude lower than a certain level (rejection ratio),
Page | 26
Chapter2Background Theory
were removed. The ESO method was expanded later to more complicated and
multiple loaded cases employing different rejection ratio( RR ) for each iteration
trusses with stress and local constraints was investigated by (Ohsaki and
recent research, the algorithms proposed previously by (Xie and Steven, 1993)
minimize the weight of structures with stress constraints (París et al., 2010).
FEA. Low value of stress or strain in the structure would be reliable indicators
for inefficient material use. Ideally the stress in the entire structure should have
even margin of safety. This concept leads to a rejection criterion based on the
local stress level where the low stressed material is assumed to be under-
(Equation 2-9):
Equation 2-9
𝑅𝑅
Where, RRi is the current rejection ratio and and are the local and
Page | 27
Chapter2Background Theory
The same process with the increased rejection ratio takes place until a new
any element should be in such that the discarded element can be added further
at any stage of the process if needed (Querin et al., 1998). This method is
that it can deliver a more reliable and suitable final optimum design. Figure 2-13
and Figure 2-14 illustrate the initial design domain and optimal topology of the
5m
20m
-13: Design domain (Querin et al., 1998; Huang and Xie, 2011)
Figure 2
Page | 28
Chapter2Background Theory
-14: Topology optimized design with BESO (Huang and Xie, 2011)
Figure 2
define load transfer within bars and solids. These methods were called
Transferred Force (𝑇𝐹) method. Since these methods do not use any formal
and reaction points by defining an index for two and three dimensional models.
The methods were based on the concept of transferred forces, the part of the
Page | 29
Chapter2Background Theory
transferred force and another one which goes through the remaining region
called potential transferred force which is the indication of the stiffness of the
region.
The displacements are calculated by applying the forces on the load region.
This analysis is followed by eliminating 3 region from the solution domain and
force on the support region can then be recorded. The difference between initial
force and defines the transferred traction through the region 3 to the
_
Equation 2-11
Fi fi f f ij n j ij n j
where 𝐹 is the difference between initial and final force and ̅ is the new and
Page | 30
Chapter2Background Theory
The elements with low transferred force can be removed from the design.
method only provides qualitative information about the direction of load transfer
In the last decade, load path based algorithm methodology has drawn attention
of the researchers. This is due to the reason that load path not only can lead to
statically determinant design but also may provide a tool for new technologies
structures. A novel fibre placement pattern was defined for highly loaded joints
by (Li et al., 2006) earlier. Optimal design of machine tool bed by load bearing
this novel method to identify optimal layout of the stiffener plates conducted
Boat sail design has evolved over the years along many development paths in
1996; Shankaran, 2005). The design of sails is a very old activity principally
based on the practice and the experience of sail makers and users, prototype
building and testing on the water, in constant search of performance and safety.
The use of a computer aided design and simulation package can significantly
Page | 31
Chapter2Background Theory
reduce the time and cost associated with design and testing by optimizing the
Unlike the aircraft industry where the development has been essentially driven
by commercial consideration, boat sail design needs to comply with the sailing
rules. These rules may impose some limitation in regard to the weight or size of
the components; they have to be as light and stiff as possible (Gilliam, 2006). It
was this issue which led to the development of a number of alternate sailcloth
regarding the deformation and bending issues (Trimarchi et al., 2011). Carbon
fibre can be used to reinforce a structure with highest lightness and relatively
high stiffness. Since bending is the prevalent stress in a cursing yacht hull, they
Carbon has excellent fatigue life and is extremely strong and rigid
When combined with a core material Carbon Fibre will deliver substantial
weight reduction
Curing at 90⁰C (200⁰F) delivers a very stable part with a very low
coefficient of expansion
Results in a 25% lighter hull, deck, and structure that is 10-15% stiffer
then wet-preg
Page | 32
Chapter2Background Theory
In sail structures the primary stress is membrane tension but many of the
advantages of using carbon fibre remain. Clearly the industry is used to using
commercially viable since users want high performance and look to new and
The aim of this thesis will be to show that the possible carbon fibre paths in the
shade and wind sail cloth can be determined iteratively using ANSYS APDL as
2.5. Conclusion
This literature review studied the history and development of load transfer in the
the load paths in structures. It also underscored the concept of the structural
methods that can quantitatively characterize load paths in a structure. All these
methods provide different results and insights into the load transfer. However,
efficiency of load transfer along load paths. This study also highlighted the
problem that, in spite of the existence of all the available methods, no single
Page | 33
Chapter2Background Theory
quantify, visualize and tailor load paths. Thus the question of exactly what the
best possible approach is for defining “true” load directions in the structures still
remains unanswered.
Page | 34
CHAPTER 3.
BACKGROUND THEORY
3.1. Introduction
The literature review has identified the need to develop tools to aid the designers
to define load transfer in structures. Although it is clear why load paths are
approach to determine the paths and display them to the design engineers. Load
transfer based on the stress trajectories, load path based on the concept of load
flow and finally potential transferred force are the main contributors in this area.
Ideally, a load path method should be able to reflect some specific behaviour of the
Visualize the overall path of forces from the application points to the
reaction points
Chapter3Background Theory
Chapter 2 concluded that load paths that define a region in which a load
stress vectors. Therefore tracing contours following these stress trajectories would
not define the load paths that this research aims to identify.
Load paths derived from the load flow analogy are able to show the direction of the
force within the solution domain and highlight the zones with high stress magnitude
such as stress concentration regions. This method has been used for the
placement of the fibres in composite material and has proved its ability to be used
in industry (Li et al., 2006). On the other hand, this method has not been applied to
truss structures and the application to truss structures therefore became the aim of
this research.
The third possibility for determination of load paths is the potential transferred force
method (Harasaki and Arora, 2001). Although this technique can be used to
qualitative information about where the loads flow. No trajectories can be traced
from the point of loading to the reaction points and it has not been used for shell
structures and defines procedures to use the paths in the design of truss
structures. It also describes procedures to integrate the new methods into the
Page | 36
Chapter3Background Theory
load paths in FEA packages, an algorithm where stress results are determined
using ANSYS and FORTRAN codes has been developed to read the results and
define the paths. ANSYS is then used to plot these paths for three-dimensional
applications. The load paths provide better understanding of the required layout of
the structure in the initial stage of design. For some simple beam configurations
the general pattern of paths for different loading conditions will be described
regardless of the detailed geometry of the structure. The last part of the chapter
applies the idea of transferred force to trusses and describes how they can
Load path originally was proposed as a method to design Strut and Tie models in
reinforced concrete structures (Schlaich et al., 1987). The load path method has
also become a simple and effective instrument to understand the behaviour of the
structures under different loading conditions. Figure 3-1 provides the information
about how a simple load can flows through a bolted joint from the point of origin to
its end.
Page | 37
Chapter3Background Theory
As it can be seen from Figure 3-1, a tension load is applied to the upper plate and
passes through the hole to become a bearing load on the pin. It then transfers in
shear form to reach the lower plate and finally ends in compression form in the
opposite end.
membrane loads. Such structures include shade cloth erected to provide shade
over large areas and sails used on boats (Figure 3-2). Besides, these structures
deform to carry load as membrane forces. The literature survey does not identify
any analytical procedures based on load paths for placing reinforcing fibres in the
membrane structures. Tension structural designers have used trial and error
strategy to find the “Fibre-path” sails and shades for many years. The ’fibre path’
technology would allow the designers to align the fibres along the primary loads in
a sail and build in multiple fibre patterns to address secondary loads. Sails with
reinforcing fibres have a better wind range than traditional radial sails and also
Page | 38
Chapter3Background Theory
technology handles secondary loads, the sail holds its design shape throughout
The use of major principal stress trajectories for engineering applications has been
are aligned in directions for which the shear stress is zero. A tensile principal
stress will be carried out by reinforcing fibres and the membrane connecting the
fibres will be lightly loaded. Figure 3-3 and Figure 3-4; however, indicate the main
drawback using the principal stress directions to define the direction of the
direction for the minimum principal stress are two sets of principal stresses that
can be observed in Figure 3-3 and Figure 3-4 respectively. According to these
identified directions, two sets of fibres would therefore be required; one set needs
tension. It is noteworthy; the load parallel to the fibres varies along the path.
Page | 39
Chapter3Background Theory
For that reason similar to other methods, principal stress trajectories experience
some limitations. The placement of fibres along the tensile principal stress
trajectories alone may result in strength improvement but may not represent the
investigation of load paths, when the loads create a stress field dominated by
tension (such as the stresses around a hole under uniaxial tension such as
Figure 3-3 and Figure 3-4) the resultant dominant paths from the stress trajectories
are very similar plot to the paths obtained from load path method. However, tracing
load trajectories based on these orientations would not result in trajectories where
the loads being transferred are constant in the sense of streamlines in a fluid flow.
Load paths determined by the concept of load flow results trajectories similar to
streamlines in a fluid flow. As the patterns obtained from this method can be
compared to the benchmarks from the paths originated from stress trajectories,
Page | 40
Chapter3Background Theory
Load paths have been generally accepted by many researchers (Karkauskas and
Norkus; Chan, 1963; Waldman et al., 1999; Kelly and Tosh, 2000; Arora and
Harasaki, 2004; Sakurai et al., 2007; Wu et al., 2009) as a useful tool for
how remote loads are equilibrated through a structure and can also provide insight
into how well a structure is performed its intended load carrying functions. There
was no significant research to define load paths until late 90s when all the
mathematical form to determine load transfer was found until Kelly and Elsley
proposed a method for computing load flow orientation (Kelly and Elsley, 1993).
This method was based on the stress resultants obtained from finite element
As previously described, load paths are the regions carrying an applied load in a
prescribed direction within the structure. This concept of load flow comes from fluid
flow analogy, so it needs to originate from one prescribed location and end in
the 𝑥, 𝑦 or 𝑧 axes. For some simple cases such as the tensile loads (Figure 3-5),
there is one direction as the dominant direction while in other cases, such as
Page | 41
Chapter3Background Theory
bending from a shear force, there is more than one main way that the load flows,
this will be addressed later in the current chapter. For load flow in a single
dominant direction, the concept is illustrated with a hypothetical force “stream tube”
θ
x
The load path defined is bounded by contours in such that there is no contribution
simplified to:
There are two stress components that are normal and tangent to the side wall of
the force tube, namely n and nt . n is the local stress acting normal to the plane
and nt is the shear stress acting along the plane. nt acts on the face whose
normal is in the n direction and is positive in the positive t direction. Equation 3-1
will be satisfied if the orientation of n with respect to 𝑥 axis at each point along the
Page | 42
Chapter3Background Theory
(Equation 3-2):
or
sin cos
n nt
The solution to Equation 3-2 can be achieved by expanding the terms in the
Figure 3-6:
y
t
xy n
nt n
xy
x x
xy
xy
y
Page | 43
Chapter3Background Theory
is the shear stress component on the same plane. Similarly, is the normal
and
According to Figure 3-6, can be calculated from dividing shear stress by normal
stress (Equation 3-5). Having n and nt from the equations above, allows updating
the Equation 3-5 to the new form as Equation 3-6 and Equation 3-7:
nt Equation 3-5
tan
n
giving:
Page | 44
Chapter3Background Theory
x Equation 3-7
tan 3 tan 2 x tan 1 0
xy xy
x
tan (tan 2 1) (tan 2 1) 0
xy
Solving Equation 3-7 shows that with only value of for tan the equation will be
satisfied. Since the value of tan 2 is always equal greater than zero, tan 2 1 1 .
In addition there are two solutions for in the range of 0 360 , in which each
one will differ 180 from the other. This noticeably reveals that load path is
directionless and can travel from left to right or right to left in an identical direction
regardless of the origination of the paths. A pointing vector (𝑉 tangent to the wall
Vx xi xy j Equation 3-8
A load path in 𝑦 direction (𝑉 can be derived similarly. If the axes are transformed
in such that 𝑥 axis is in the direction of the dominant applied load, these load paths
Vy y i yx j Equation 3-9
One of the significant aspects of load path is that singularity has no effect on
Page | 45
Chapter3Background Theory
Equation 3-10
tan xy
1
x
xy
lim (tan 1 ) tan 1
x 0 x
90
In Equation 3-10 if x tends to zero (the case where singularity happens) then
tends to 90 with respect to 𝑥 axis, so singularity cannot effect the load path
direction. Knowing the direction and the extent of the load is a high priority in the
design discipline. The extent of the load being carried between adjacent contours
in Figure 3-5 is given by Pxa as the external applied force. This force can be
The stresses n and nt can be extracted from the finite element analysis packages
such as ANSYS. In the standard version of these packages, load paths cannot be
plotted in the post processing section directly from the solution and thus an
number of codes written by Kelly (Kelly, 2005) could define these paths, additional
improvement was made since this research has begun. These improvements were
Page | 46
Chapter3Background Theory
to find a unique attribute of load path within different types of structures regardless
The main platform for the finite element package was ANSYS APDL 12.1 due to its
wide use and proved acceptance in the simulation world. FORTRAN was also
selected as the programming code due to its compatibility with the FEA package.
CAD modelling
contours
Importing back the contours to ANSYS in order to plot the load paths
define the load paths contours. These contours illustrate continuous load paths
through a structure and are traced by placing this contours tangent to the load path
based on the visualization of the pointing vectors and alternatively, load path
Page | 47
Chapter3Background Theory
sections until Equation 3-1 is satisfied. This approach however requires user
interaction and does not lend itself to a global plot for all points on the domain.
The initial stage to achieve the pointing vector approach is to obtain the elemental
information such as nodal stresses from the solution mainstream. The element
used for this purpose has 8 nodes and translational movement at each node.
Solid185 elements in ANSYS are used to interface with the codes. In this chapter it
In every element, the solution of the model can produce 8 nodal stress vectors and
stress can be reported. These stresses are averaged only internally between the
nodes within the same element. This process must be repeated for all the
elements within the boundaries and once this collection has been made, as
starting element of the path can be selected. By selecting an arbitrary element and
recording the element number, the starting point for the path the general pattern of
the paths can be plotted. The stresses and initiation points for the load paths are
sent to the FORTRAN codes to calculate the pointing vectors at the centroid of
each element. The pointing vector is then defined using a non-structural link
(Figure 3-7). Each link is drawn in such that angular orientation of the link is
tangent to the load path direction. The length of the link can be scaled up to
indicate the qualitative magnitude of axial stress or can be given a unit value to
Page | 48
Chapter3Background Theory
8 7
4 3
5
6
1 2
Applying this method to each element produces the required links that can be
plotted in ANSYS to define the pointing vectors and once these links are
generated, they exported back as an input file to ANSYS. The links are then
plotted with the required ANSYS format in the CAD section over the initial
geometry.
A fourth order Runge-Kutta method was used predict the contours through the
vector field by repeatedly projecting forward from one location on the contour to the
next in a similar fashion to plotting streamlines in fluid flow. Runge- Kutta method
The use of Runge – Kutta method is defined based on the increment of s as the
increment along the load path. Using this step length can ensure that all the spatial
increments are fixed to a unit magnitude during the generation of the path through
As an evidence, for a normalized vector field, V is defined over the mesh domain
dp1 V pi
s
dp2 V 1
pi dp1
s
2
dp3 V 1
pi dp2
s
2
Equation 3-12
dp4 V 1
pi dp3
s
2
1
pi 1 pi ( dp1 dp2 dp3 dp4 )
6
𝑃.The vector can be defined at any arbitrary point by first associating the point with
As an example of the application of load path visualization using the vector plots
approach, a cantilever plate with a tip shear load is examined (Figure 3-8). The
model is a three-dimensional plate under a tip shear load of 100 N in its right wall.
This load is applied to a node in 𝑦 direction downward and is constant and non-
changeable during the simulation. The left surface of the plate is fully fixed. The
length to width ratio of the plate is 3, and the length, height and depth are 0.6, 0.2
and 0.1 , correspondingly. The plate is meshed with 693 nodes in 400 elements.
The size of the elements is completely subjective and depends on the complexity
Page | 50
Chapter3Background Theory
of the structure and accuracy of the results. A finer mesh can generate more
contours hence, resulting more accurate paths. This is a simple shear loaded
example and thus 400 nodes would be enough in terms of the visualization of the
load transferred within the boundaries. All elements are quadratic Solid185 with 8
nodes with isotropic behaviour. Young’s modulus of 210GPa and Poisson’s ratio of
Regarding selection of the element for initiation of the load path, 20 random
elements are considered. The recorded element numbers accompanying the nodal
stresses in each element are then defined in the FORTRAN code. The FORTRAN
codes produce 3200 links with individual length of 0.01 and at the final stage the
links are imported back to ANSYS in the post processing environment to visualize
the load paths. The load path contours for loads parallel to the 𝑥 axis, are shown in
Figure 3-9.
Page | 51
Chapter3Background Theory
It can be clearly observed that some of the regions are not traversed by the load
paths since they are not traversed by paths selected in the random selection. The
FORTRAN codes are written in a way to be able to show high stress zones in the
structures. In Figure 3-9, the paths are given a spectrum of 10 colours. The colour
presentation can reflect the stress level of each zone in the domain and therefore,
structure in their further study. Red colour represents the highest stress level while
blue shows the lowest and the rest of colours in the spectrum indicate the stress
with the extent between the maximum and minimum level. The looping of the paths
shows existence of bending in the structure which will be discussed further in the
next section.
The contours plotted in the Figure 3-9 have the dominant load transformation
within the structure. Due to the boundary condition top half bear tension and
bottom half bear compression and transfer the shear load from left to right.
Figure 3-10 gives the y direction load paths and 𝑦-Path load paths show shear
transfer from right to left. Although load paths can be generalised to three
Page | 52
Chapter3Background Theory
dimensions, examples in this section are restricted to cases with non-trivial paths
in only 2 dimensions for clarity. However, only the front view of the model can be
An L-shaped domain with vertical shear load is the employed example to test
requirement of the code, this model is a three-dimensional block with height and
width of 0.2 m and depth of 0.1 m . The applied load as shear applied to the middle
Figure 3-11 and is constant during the analysis. Boundary conditions constrain the
top wall of the model. All other nodes are free to move in 3 degree of freedom.
Finer meshing has been chosen for this model as it has more complicated
geometry compare to the previous model. 1936 nodes in 1470 elements have
Page | 53
Chapter3Background Theory
been created. All of the elements are 8 node Solid185 elements in ANSYS. The
Young’s modulus and Poisson’s ratio is 210 GPa and 0.3 respectively.
Similar procedure is applied to transfer data between the FORTRAN codes and
ANSYS and the outcome for 𝑦 direction load has been plotted in Figure 3-12. The
red colour identifies zones in which the magnitude of the pointing vector is high.
Similarly blue colour contours present lower value of pointing vector. As expected,
the corner in the geometry is stressed more than the other zones because of the
Page | 54
Chapter3Background Theory
existence of loop in the paths reflects the existence of bending. This bending is
In three dimensional applications with an element with free face, the paths are
projected parallel to the free surface when the angle of vector to the surface is
small and this implies on order to prevent the path leaving the solution domain due
to numerical errors. The amount can be set in the program and is totally
Page | 55
Chapter3Background Theory
projection.
geometry of the model and it is impossible to have a general idea how the force
can flow in every structure from the application of the load to the support.
Nevertheless, some patterns can be pre-defined for some simple cases. Following
Pure tension
Pure compression
Pure bending
Cantilever bending
Torsion
For the purpose of generalization of load flow, the pure tension case is considered
first and without any discontinuities in the structure, the load can trace straight from
left to right. Since the load is only in x direction, no paths can be expected in the
normal direction, 𝑦 direction. Since the path carries the same load, they are equally
Page | 56
Chapter3Background Theory
Pure compression has a similar pattern to pure tension but the difference is that
the load is transferred from left to right and as the geometry is totally symmetric,
load paths are straight and equally spaced. As the load transfer is directionless
Pure bending in Figure 3-16 provides a linear stress distribution from top to bottom.
Therefore if the paths bound regions transferring the same load, the contours will
be closer together at the top and bottom of the beam in Figure 3-17 where the
Page | 57
Chapter3Background Theory
y xx
z x
Mz
A more complex case occurs when a shear load is applied to a beam. Shear load
creates a bending moment in the beam and make the load paths lose their uniform
linearity and form loops or eddies in the solution domain. Since shear stress
straight.
Figure 3-18 (b) shows 𝑉 contours in a cantilever bending. In more details; a sub-
domain that is bounded by two load paths including the mid-plane of the beam and
the loaded right hand wall (Figure 3-19) can be created. 𝑥-force equilibrium is
yx
maintained by a shear stress at the mid-plane, , and a direct stress at the right
hand wall,
xx .
Page | 58
Chapter3Background Theory
𝑥-Path
(a) (b)
𝑦-Path
(c)
The two edges defined by the curved contours are free from any 𝑥-force. This case
demonstrates that the stress that ensures the subdomains, are in equilibrium can
vary between direct stress and shear stress as the load bearing surfaces are re-
Px
Px
The loops in simply supported beam are similar in cantilevered beam; the loops
form oval/circular shape depends on the ratio of the block. In this loading condition,
Page | 59
Chapter3Background Theory
𝑥-Path
𝑦-Path
The most efficient load path is a straight path of constant cross-sectional area
parallel to the load (such as a cable in tension) carrying load from the point of load
application to the point of reaction. However, load paths are forced to follow curved
moments. The force is constant along the path identified in Figure 3-21, a stress
concentration narrows the path so that the ratio of the force to the cross-sectional
area gives the higher stress experienced at the stress concentration. This can
Page | 60
Chapter3Background Theory
If the path is not straight, the bending moments are introduced that are equilibrated
by tractions in the orthogonal direction acting on the boundaries. For the path
segment, for example, Figure 3-22, the tractions on the top and bottom boundaries
direction. The bending moment due to the offset in the loads pxa can be
Px
Px
Py
Py
Page | 61
Chapter3Background Theory
Typical structural configurations to transfer the load for each case identified in the
The geometries proposed are suggested by the contour patterns of the load paths.
bending moments and shear. For axial load, the relevant property is the cross-
the case of compression. For bending, the second moment of area is essential and
only requires a path for the shear and bending but also includes connecting
Page | 62
Chapter3Background Theory
Once a clear interpretation of the types of load and the load paths are defined,
3.2.8. Discussion
To determine the load paths, the three-dimensional structures with solid elements
were studied in this chapter. A formulation to define the load path from the theory
defined by Kelly was covered and further refined to plot the load paths in the
mentioned earlier they are completely subjective and totally depend on the
In this section of the study, determination of load paths in trusses will be covered.
In trusses the load path is determined by the member that carries majority of the
load (Kelly and Elsley, 1993). These members are crucial to the structure since
any damage or failure to them would collapse the entire design. However, the
other members only play a complementary role in the transformation of the load in
the structure and can be removed from the domain. The internal force can be set
Page | 63
Chapter3Background Theory
to a criterion to determine the load paths from the FEA analysis. The criteria then
will deduce the elements based on an iterative procedure through the solution.
The reduction of the members can follow the method introduced by Xie and Steven
(Xie and Steven, 1997) called Evolutionary Structural Optimization Method (ESO).
defined as the force must follow the members in the truss. It is then possible to
determine which parts of the structures carry the most applied load. This technique
can provide a useful tool for engineers and architects in their design processes.
The internal force in members can be used to represent force transfer from the
load point to the support. The force magnitude in any part of the structure can be
significant load, it can be removed from the structure in such that its removal does
not allow a mechanism to form and lead to failure in the structure. The element
removal from the structure can be based on a rejection scalar parameter that is
The internal force value of each single element can be determined by comparing;
e
for example, an element with f i as the local internal force with the maximum
Page | 64
Chapter3Background Theory
max
force value of fi as the maximum global force value. After each finite element
analysis, element which satisfies the following condition can have their modulus
reduced:
fi e Equation 3-13
max <r
fi
In ANSYS all elements for trusses have chosen to be “Link 180”. Link180 is a bar
that can be used in a variety of engineering applications. This element can be used
to model trusses, sagging cables, links, springs, etc. This 3-D bar element is a
deflection, and large strain capabilities are considered in order to model these
The algorithm implemented for determination of the load path in truss structures
consists of eight iterative stages. These stages can be repeated for a number of
times depending on the set up removal ratio. For more clarification, the load path in
Figure 3-24 which is a part of multi bay truss can be obtained from the following six
stages:
Page | 65
Chapter3Background Theory
II. If entering force is tension one side of section – exiting force must be
5. Select member with largest force component (𝑥 or 𝑦) and obeying sign rule
N5 2T N4
2
1.414C 1.414T
2
N1 1C N2 1C N3
1 1
In Figure 3-24 Ni i 1...4 are the nodes which are connected to each other with
Page | 66
Chapter3Background Theory
12 1C 1 0
23 1C 1 0
14 1.414C 1 1
24 0 0 0
34 1.414T 1 1
45 2T 2 0
Start at node with external x-force (ie. N1 and member 𝑁 𝑁 ). Consider vertical
N5 2T N4
2
(1)
1.414C
(1)
2
N1 N2 N3
Page | 67
Chapter3Background Theory
Now follow load in member 𝑁 𝑁 . Applying rules above gives the second 𝑥-paths,
Figure 3-26.
N5 2T N4
2
(1)
(1) (1)
2
N1 N2 N3
For the 𝑦-path start at N1 or N3 . Here start at N1 and apply rules, (Figure 3-27)
N5 N4
1.414T
(1) (1)
N1 N2 N3
1 1
Once the algorithm of the determination of load path is set up, it can be
then can be compared with the benchmarks in the literature on truss optimization
Page | 68
Chapter3Background Theory
To find a possible load path for the loading and support conditions shown in
Figure 3-28, a one bay truss with unit width and height producing a square is
employed. A vertical load of 100 N is applied to the bottom right side of the truss.
The value of force has been taken from the force applied by Xie and Steven (Xie
and Steven, 1993) in their paper. This load has to be constant during the analysis
at all time. Young’s modulus of the elements has been chosen to 210 GPa .
The truss is analysed by ANSYS APDL in the static structural environment. The
means that stress distribution is uniform along the length of the element. Two full
supports at left side do not allow any movements to the node N1 and N 4 . However,
the other joints ( N 2 and N3 ) are free to move in any translational directions.
𝑁 𝑁
𝑁
𝑁
Page | 69
Chapter3Background Theory
The axial forces obtained from each element are tabulated and stored as a
parameter in ANSYS temporary memory. Then the lowest value from the table is
extracted and the member number associated with the called value is written as a
new parameter. Once this parameter is generated, the reduction procedure starts.
The reduction can be performed in two ways; the ratio method or the cutting
method. In the ratio method as explained earlier in this chapter, Young’s modulus
is reduced by to certain extent. In this example the removal is set to just one single
element reduction in one iteration and the designer can interact with the design
process to approve which element is removed from the structure in every step. The
contribute nothing to the overall mass (or capacitance, etc.) of the matrix.
The reduction is performed for four iterations and then a statically determinant
Page | 70
Chapter3Background Theory
(1) (2)
(3)
In this reduction the elements shown with lighter colours are more desired to be
eliminated from the domain. These colours depict the lower internal force value
method can be obtained via either the reduction ratio or constraining the number of
statically determinate design criterion to fix the ending time. In this thesis the
Similar approach was taken for a two bay truss with cross section area of 1 cm2 and
unit length for each bay. A load of 100 N was applied to the bottom right corner.
Page | 71
Chapter3Background Theory
And then FEA and other procedure were applied in ANSYS APDL program,
Figure 3-31:
The reduction of the element was set to be the member with the lowest axial force.
So it was expected to have only one element removed at every step unless
multiple elements with identical values of axial forces were observed. The
Page | 72
Chapter3Background Theory
(1) (2)
(3) (4)
(5) (6)
(7) (8)
In Figure 3-32 elements are coloured in spectrum of 10 colours where red depicts
the highest value and Blue the lowest extent of internal force.
Page | 73
Chapter3Background Theory
1 2 3 4 5 6 7 8 9 10 11 12 13
1
-1302.5 -766.95 443.71 639.53 1291 0 406.85 -39.804 488.81 -350.56 -432.52 471.04 -437.87
1 2 3 4 5 7 8 9 10 11 12 13
2
-1302.5 -766.95 443.71 639.53 1291 406.85 -39.804 488.81 -350.56 -432.52 471.04 -437.87
1 2 3 4 5 7 9 10 11 12 13
3
-1290.7 -757.04 458.14 649.43 1302.8 377.38 461.99 -377.38 -461.99 481.16 -427.75
1 2 3 4 5 9 10 11 12 13
Iterations
4
-1014.4 -1014.4 276.57 553.15 1418 611.55 3.22E-13 -611.55 650.68 -618.44
1 2 3 4 5 9 11 12 13
5
-1014.4 -1014.4 276.57 553.15 1418 611.55 -611.55 650.68 -618.44
1 2 4 5 9 11 12 13
6
-1335.3 -1335.3 0 1329.4 940.06 -940.06 749.71 0
1 2 5 9 11 12
7
-1335.3 -1335.3 1329.4 940.06 -940.06 749.71
1 2 5 9 11
8
-1000 -1000 2000 1414.2 -1414.2
Page | 74
Chapter3Background Theory
It is noteworthy; the final truss has two members connected at node N 2 . This would
give a singular stiffness matrix if the killed members are actually removed. Instead,
the killed members are remained with reduced stiffness and prevent singularity in
the stiffness matrix. In iteration 6 of the reduction, two members indicate no carried
load since their recorded axial force is zero. Removal of two members at the same
time occasionally is possible if two members meet each other in a node and lack
Figure 3-33:
In the node N 4 , three elements 𝑒 𝑒 and 𝑒 meet each other. Element e3 is vertical
and thus carries the force only in 𝑦 direction. Element e4 is angled and theoretically
carries the load in both 𝑥 and 𝑦 directions. Similar interpretation would be true for
element e13 as well. It means that it can carry the load in 𝑥 direction. During the
reduction process, e3 was removed from the solution domain while the other two
Page | 75
Chapter3Background Theory
members are still active. Removing element e3 causes the equilibrium of the node
to fail since there is no more elements to cancel out the load caused by e4 in 𝑦
direction and hence e4 will be automatically removed from the domain. Similar
The procedure obtained from both one and two bay truss can be compared and
Harasaki, 2000; Harasaki and Arora, 2001; Arora and Harasaki, 2004). This
bays with the ratio of three and different loading conditions is considered. This
example validates the results by comparison with benchmarks in the literature. The
width with the unit value and the height of the truss with the half times to the length
is examined. Two full supports are placed at the very bottom of the structure.
These supports forced zero displacement on the associated nodes in any direction.
A tip vertical load of 100N is located in the middle of the structure in the bottom
flange. The material property of 210 GPa as Young’s modulus is considered for this
model.
Page | 76
Chapter3Background Theory
This model consists of 10 nodes and 21 elements. The nodes are connected to
each other via two ways; regular and overpopulation meshing method. In the
regular method the nodes are connected to each other with horizontal, vertical and
(Figure 3-35).
After FEA analysis and the reduction procedure are conducted, the total number of
Figure 3-36.
Page | 77
Chapter3Background Theory
As it is observed, the result from both meshing methods is identical with no effect
on the final plot of the remaining members. The achieved result can be also
validated from the result from the benchmarks in the literature (Karnovsky and
Lebed, 2009). Figure 3-37 reveals illustrates the weight reduction in every step of
the elimination:
200
150
Weight (kg)
100
50
0
1 2 3 4 5 6
Iteraion Number
3.3.5. Discussion
Optimized design can be generated from the initial structure by different methods
design can interpret the load path in the discrete structures as they comprise a
finite number of clearly defined members which carry only axial load. The above
Page | 78
Chapter3Background Theory
numerical approach.
proposed by Xie and Steven. The Evolutionary Method acts via a reduction
criterion which was explained and utilized in a number of examples. The achieved
results were similar to the optimized structures defined in the literature using shape
path in the boat sails and shade structures which will be investigated in the next
chapter.
3.4. Conclusion
Chapter three discussed two formulations to find load paths in order to illustrate the
concept of load path in both continuum and discrete structures and then some idea
of generating load path for sail structures were explained. Formulation by Kelly et
al. (Kelly et al., 2010) earlier to this research could determine the load paths in the
continuum structures while it was not applicable through ANSYS as the FEA
package. The development of the algorithm and the process of obtaining load
paths and the integration into ANSYS were therefore explained. The formulation
was applied to a cantilever beam and L-Shape structure. Regardless of the type of
the structure in a rectangle solution domain, the idea which was examined through
path that can be generalized into the other applications. This idea was classified
Pure tension
Pure compression
Pure bending
Cantilever bending
The extension of load paths to trusses was also explained later in this chapter and
the idea of considering the axial force as the parameter was proposed. This was
the structure. The combination of the axial force method and EM provided
criterion, 𝐸𝑀 method and membrane analysis may be a great step for this
covered later in the next chapter and the application of the formulation and the
Page | 80
CHAPTER 4.
NUMERICAL STUDY OF LOAD
PATHS IN SHELL AND PLATE
STRUCTURES
4.1. Introduction:
missiles and ships. It is essential to understand how the load travels in the tension
structures under boundary conditions. This additional information would open the
designers’ vision to have deeper insight through the design. The wide application
structural component efficiency: the larger this ratio, the more optimal is a
structure
High stiffness
Containment of space
In the area of tension structures, the sail and yacht construction industry has
become the front runner when it comes to implementation of fibre reinforced resin
building and testing on the water. The use of computer aided design and
simulation packages can significantly reduce the time and cost associated with
design and testing the design before a prototype is built. The optimum use of
carbon fibre in the sails can reduce the total cost of the product. Yet, there is no
establishing a finite element method not only can provide the designer stress
distribution and other stress resultants but also can deliver information of how
fibres distribute through the domain. To avoid any possible error in determination
of the carbon fibre through finite element analysis, the tension structure with their
pressure load in ANSYS. The analysis must be carried out in the static condition,
nonlinearity to tolerate the condition of large deflection imposed by wind load in the
real situation. In order to obtain the possible load paths in the yacht and cruising
boat sails, the first step is to simulate a shell as the main cloth and then validate
the results with the theory of shells and plates and then overlay the suggested
method to find the load paths. A shell element for the analysis of thin and moderate
plate and shell structures is formulated along with the compatible link element. The
overlaid elements are created for the analysis of laminated fibre reinforced boat
sails.
assumptions. Then the numerical results obtained from finite element analysis
(ANSYS APDL) can be compared with the theories. This chapter mainly highlights:
Page | 83
Chapter4Numerical Study of load paths in Shell and Plate Structures
Firstly, it is worth introducing some general terms used in the plate analysis
compared to the overall dimensions of the sail, so large deflection theory of the
3. The deflection of the mid-plane is fairly large compared with the thickness of
the plate. So the slope of the deflected surface is therefore very large in
4. The normal lines initially normal to the middle plane before deflection remain
straight and normal to the middle surface during the deformation and the
thicknesses:
a
1. THICK PLATES, when the ratio of 8:
h
Where a in this inequality is the area of the plate and h is the thickness of the
plate. The analysis of such bodies includes all the components of stresses, strains,
Page | 84
Chapter4Numerical Study of load paths in Shell and Plate Structures
and displacements as for solid bodies using the general equations of three-
dimensional elasticity.
a
2. MEMBRANES, when the ratio of 80 :
h
Membranes carry the lateral loads by in-plane axial tensile forces 𝑁 (and shear
forces). These forces are called membrane forces; they produce a projection on a
vertical axis and thus balance a lateral load applied to the plate-membrane.
a
3. THIN PLATES with 8 80 :
h
This group represents an intermediate type of the plates. Depending on the value
w
of the ratio the part of flexural and membrane forces may vary.
h
Generally, sail and shade sails can be considered as membranes, therefore the
section. The plate in Figure 4-1 illustrates the terms to be used later in this chapter.
𝑞 𝑎
𝑓𝑦𝐵 𝑓𝑦𝐴
𝑓𝐵 𝑓𝐴
𝑓𝑥𝐵 𝑓𝑥𝐴
w
𝐵 h A
Page | 85
Chapter4Numerical Study of load paths in Shell and Plate Structures
this part of the study validates finite element analysis approach by comparing with
shell theory. ANSYS classic is chosen as the platform for the analysis due to its
wide use and compatibility with third party applications. Thin and moderate thin
shell structures such as shades and sails are analysed usually by Shell elements
bending stiffness often leads to singular solutions of the stiffness matrix under
satisfy the enforced conditions in the membrane, distributing the pressure through
membrane and advance feature of ANSYS using Stabilizer and Line search are
circular model clamped in the radius edge. The radius of the plate is 8.70 and the
thickness is 4 . The model is used to explain the validation of the analysis steps
and capability of the modelling for the layered elemental creation which will be
Page | 86
Chapter4Numerical Study of load paths in Shell and Plate Structures
Materials characteristics used in the current analysis are adapted from the values
used in the sail construction industry (CyTec, 2011), woven polyester (WP)
sailcloth with and Young’s modulus of 873 MPa and a Poisson’s ratio of 0.3.
4.3.2. Element
The element used for this model is Shell181. This element is suitable to analyse
about the x , y and z axes. Shell181 can be used for linear, large rotation,
deflection and large strain nonlinear applications .This element is well suited for
sail modelling since it will give the designer the chance of stiffening the model as it
According to Zienkiewicz and Taylor (Zienkiewicz and Taylor, 2005), the elements
with reduced integration scheme can improve responses for thin plates and shells.
Since the predicted transverse shear stresses are found to be incorrect in shell
points, while more points may not increase the accuracy of the computed FE
the shell elements in this research. The shell elements used in this analysis do not
resist any bending and thus the feature of the “membrane only” is chosen for the
Page | 87
Chapter4Numerical Study of load paths in Shell and Plate Structures
element creation. This enforced limitation to the elements, assures the designer
that the structure deflects easily under the wind load and has only translational
The size of the element is selected based on final 𝑧 deflection compare to the
Table 4-1. The selected size is 51 cm which is the closest value to the theory. In
implemented like in this research, the deflection w may become very large in
neglected, hence the maximum deflection can be calculated from Equation 4-1
qa Equation 4-1
w 0.662a 3
Eh
Page | 88
Chapter4Numerical Study of load paths in Shell and Plate Structures
To confirm the analysis steps and the implemented parameters with Timoshenko
theory of plate and shells (Timoshenko et al., 1959), a clamped circle is used. The
clamped circle has no DOF (degree of freedom) in all directions in the edge
(Figure 4-3). This is simulated by applying the fully support in the edge nodes
within ANSYS. However, the other elements are free to move in x , y and z
directions.
Page | 89
Chapter4Numerical Study of load paths in Shell and Plate Structures
Modelling of the pre-tension in the plate is carried out with the radially stiffening of
the membrane in the first step of the loading step with 8 enforced
be acting in the normal direction to the plate at the centroid of the circle in the 2nd
loading step while it is removed from the boundary conditions later during the
analysis. This is due to the fact that one step load application would lead to a non-
convergent solution therefore; the membrane must be stiff enough to react with the
-4: Point load in the centre and enforced displacement in the edge
Figure 4
2008). The pressure is static and remains constant during the analysis. Due to the
distortion discussed earlier, the load is applied in 9 load steps of the increments for
Page | 90
Chapter4Numerical Study of load paths in Shell and Plate Structures
70
As Figure 4-5 shows, the pressure must be slightly increased up to loading step 8,
means after a certain point the membrane is stiff enough and let the platform to
To begin the analysis, a total number of 2352 Shell181 elements and 2409 nodes
geometrical nonlinear. Stress stiffening is also used to make the material stiffer
once stress is applied. Newton-Rophson iteration is chosen and the stepping time
may vary in each different steps. The reason of analysing plate through geometric
nonlinearity and different loading steps is due to the fact that the model does not
resist bending and thus the elemental formulation can easily be distorted if the
Page | 91
Chapter4Numerical Study of load paths in Shell and Plate Structures
It is evident that maximum 𝑧 deflection occurs in the centre of the plate. The
edge as it is fully constrained in all DOF. This is the result which was expected and
Figure 4-6 provides information of 𝑧 deflection contours over the entire model:
From Figure 4-6, the symmetrical behavior of the model can be understood since
Page | 92
Chapter4Numerical Study of load paths in Shell and Plate Structures
The stress distribution is also graphed in Figure 4-7. which illustrates that Von
Mises stress in the centroid is in maximum extent and the clamed edge is
experiencing the lowest extent. This is is expected from the distribution of load into
Above results indicate that the procedure of the loading steps and pretentioning of
the model do not affect the result and model remains symmetrical in 𝑧 direction
and maximum 𝑧 deflection agrees with shell theory, Table 4-1. From this promising
rectangle shade and cruising boat sail simulated by overlaying meshing method in
Page | 93
Chapter4Numerical Study of load paths in Shell and Plate Structures
are distinct at one length scale but together form a single structural entity at a
longer length scale. One example is long continuous fibres imbedded in a matrix
material. The fibres, either unidirectional or interwoven, are usually layered to form
a laminate.
to find the possible load paths in the tension structures. By combining these two
elements a finite element model for reinforced shells is formulated. The reinforced
finite element model employs elastic constitutive relationship for the main cloth,
discrete link elements for modelling of reinforcement and bonded joints for bonding
the slippage between the main cloth and reinforcement. The numerical results
show that the model can predict the behaviour of the laminated tension structures
The four and eight node isotropic plate elements for thin plates and shells are
generally used for thin and moderate thin shells and two node beam elements are
al., 1989; Ranjbaran, 1992). Due to the fact that beam elements can resist
Link181 is used as the replacement due to its similar characteristics but bending
resisting behaviour. Figure 4-8 illustrates overlaying of link and shell elements
clearly. As shown, a four node shell element consists of four nodes which is
Page | 94
Chapter4Numerical Study of load paths in Shell and Plate Structures
combined with six Link181 elements on top. The members are common in corner
nodes and since both elements do not resist bending and have three degree of
freedom, essentially they are bonded in the corner joints hence they move together
as the structure deflects. It means that Link181 and Shell181 can only move
Link180
Node2
Node1 Shell181
Node3
y
x Node4
and without over laying link elements in ANSYS. Rectangle model is chosen since
it has a regular shape and all the nodes and elements creation can be modelled
highlighted when the load path determination process is added to the analysis
number of loops which depends on the element size. Essentially the shell model is
modelled without overlaid link element first due to the possibility of convergent
issue then overlaid shell and link element is analysed together. The thickness of
the plate is the same as circular model with 4 mm and the width and height of the
shade is 2.4 m . The cross section area of the link elements is considered to be
1 .
model and are adapted from the values used in the sail construction industry
(CyTec, 2011), woven polyester (WP) sailcloth with and Young’s Modulus of 873
4.5.2. Element
The elements for modelling of shell are chosen to be Shell181 due to its ability
which was explored earlier in this chapter. The membrane model in this research
does not resist any bending hence membrane only attribute is picked through the
convergence which is depicts in the Figure 4-9. The model consists of 1600 nodes
and 6 cm of length in each element and 4641 shell elements all together.
Page | 96
Chapter4Numerical Study of load paths in Shell and Plate Structures
1.80
1.75
1.70
1.65
Z deflection (m) 1.60
1.55
1.50
1.45
1.40
1.35
1.30
0 0.1 0.2 0.3 0.4 0.5 0.6
Element size (𝑚)
Figure 4-9: Shell element size convergence for the rectangle plate
membrane attribute of the shell elements. The nodes in the Link181 have DOF to
move translationally in 𝑥 and 𝑦 and 𝑧 directions. The size of the links is identical to
the size of one side of a Shell181 element. The meshing pattern is in such a way
that between every pair of nodes in shell element a link member is overlaid. This
Page | 97
Chapter4Numerical Study of load paths in Shell and Plate Structures
means that there are 6 link members in a shell element in vertical, horizontal and
diagonal directions. This meshing type can let the load flows over the shell layers
as well as in the edges of the elements. Generating the link elements as explained
earlier creates 3120 Link181 elements in the model. Figure 4-11 shows the
meshing pattern.
This model is a clamped rectangle which is fixed in all DOF in the nodes in the
edges (Figure 4-12). All other inner nodes are free to move in all x , y and z
directions.
Page | 98
Chapter4Numerical Study of load paths in Shell and Plate Structures
membrane in 𝑥 and 𝑦 directions in the 1st step of the loading step with 5
acting in the normal direction to the plate at the centroid of the circle after the pre
tensioning process while it is removed from the boundary conditions later during
the analysis. And finally 10𝑃𝑎 is ramping on the membrane through the rest of the
analysis.
ANSYS. The simulation is performed twice, with and without overlaid link elements.
Starting with shell element only, and then with shell and link elements. Similarly,
Page | 99
Chapter4Numerical Study of load paths in Shell and Plate Structures
circular plate, large displacement and stress stiffening are also used to resist large
deformation in the structures. The analysis is performed though 12 load steps and
11 sub steps with Newton-Raphson iteration method. This model due to its
boundary condition is easily solved very quickly. However, for more unstable
As Figure 4-13 displays, the maximum 𝑧 deflection occurs in the centre and it is
rectangle which results in the typical shape of clamped shades. Along the edges of
the model there are some negative deflections; this indicates that wrinkling or
Page | 100
Chapter4Numerical Study of load paths in Shell and Plate Structures
rippling of the membrane is occurring at these points. This is generally due to the
translation of the material towards these edges and the inability of the nylon to
resist bending and wrinkling. Also the stress distribution can be found in
Figure 4-14. Similarly to the circular membrane stress in the 𝑧 direction can be
found significant compare to the other directions and hence Von Mises stress also
gets affected.
Compare to the model without overlaid link element, it is expected that deflection
from 1.7 m to 0.16 m and also Von Mises stress in the membrane is lowered from
1039 Pa to 11.63 Pa . (Figure 4-15, Figure 4-16). This can highlight the effect of
carbon fibre into the sail and tension structures performance. In this specific model,
Page | 101
Chapter4Numerical Study of load paths in Shell and Plate Structures
the plate with link element can bear 1000𝑃𝑎 while the plate without link element
Page | 102
Chapter4Numerical Study of load paths in Shell and Plate Structures
procedure. This technique involves modelling of the internal force in the link
lie within the shell elements. The time for elimination is dependent on the
complexity of the structure and the meshing pattern. The elimination procedure of
the rectangle plate with diagonal overlaid link element took seven hours and thirty
minutes over 700 iterations to reach the final pattern. The computational time is
based on the regular desktop with 3.00 𝐺𝐻𝑧 CPU and 4.00 𝐺𝐵 RAM and hard drive
of 150𝐺𝐵.
It must be highlighted that the elimination method as explained earlier does not
delete or simply removes the selected element from the solution domain while the
Young’s modulus is reduced to half of the magnitude of its initial value. The benefit
of this method is that stress singularity never occurs in the solution procedure due
to elimination of one element from the solution system. However, the reduced
Figure 4-17 provides information about how the links elements are remained in the
model after the elimination process. The starting point of modulus is from the
elements in the boundaries as they carry the load since all their nodes are fixed in
all DOF. Then the process kills the horizontal elements in the top and bottom sides
Page | 103
Chapter4Numerical Study of load paths in Shell and Plate Structures
as well as vertical elements in the right and left sides. As the process continues,
two strips of link elements connecting the top to bottom and left side to right side
remain in the structure. These two strips of link elements are joints in the centre of
the shade. The connections are where all the forces that act on and within the
It can be observed from Figure 4-17, that the force has the higher magnitude along
the edges and is reduced towards the centre. This means that if the elimination
process continues, the elements in the areas coloured in yellow are the latest
members to be eliminated from the solution domain. The line of links elements
represents the state of stress in a linear pattern in the shade. If the loads upon the
Page | 104
Chapter4Numerical Study of load paths in Shell and Plate Structures
shade are discretised into a set of point loads then the line of links can construct a
cross polygon. The remaining result can be compared with the load path for a
as the second numerical example. Under this circumstance, the rectangle is only
fixed in the corners while the other nodes are free to move in translational degree
tensioned from the corners and usually are perfect solution for shading any
outdoor area including pools, patios, playgrounds and schools. The dimensions are
similarity to the previous model with 2.4 m lengths in each side, 4 as thickness
Page | 105
Chapter4Numerical Study of load paths in Shell and Plate Structures
The material of the sail is woven polyester with 873 MPa Young’s modulus and 0.3
Poisson’s ratio. The carbon fibre resins are modelled with Young’s modulus of 150
4.6.2. Element
Due to the fact that increasing the number of elements may increase the solution
time, the Shell181 element length of 12 cm (the element size is two times of the
clamped circular model) is considered to minimise the modelling time. Figure 4-9
can verifies that increasing the element length would not lead into major difference
in the 𝑧 deflection. However it can decrease the solution time by approximate half
of the initial. So this model consists of 400 nodes and 380 Shell181 elements and
Page | 106
Chapter4Numerical Study of load paths in Shell and Plate Structures
The model is pre-tensioned in 𝑥 and 𝑦 directions in the 1st step of the loading step
direction to the plate at the centroid of the circle after the pre tensioning process
while it is removed from the boundary conditions later during the analysis. And
finally 10𝑃𝑎 is ramping on the membrane through the rest of the analysis.
Due to the complexity of the structure under the defined loading condition
convergence problem still appeared. Avoiding this issue, line search is activated in
the model due to the fact that 1- Structure is force-loaded and 2- oscillatory
due to an unstable problem is usually results from a large displacement for small
Page | 107
Chapter4Numerical Study of load paths in Shell and Plate Structures
nodes in the system. Any degree of freedom that tends to be unstable has a large
the degree of freedom so that stabilization can be achieved and these techniques
may support gaining convergent solution despite of large deflection in the model.
Possible load paths in the corner fixed rectangle are achieved by implementing the
two straps bridge the corners oppositely and double symmetrical in the pattern
The lighter colours indicate the elements with more possibility of remaining in the
domain through the iterations. The pattern can be compared with groined vault
(O’Dwyer, 1999).
cylindrical shells and arranged in a cross form in such that it forms arches on each
sides (Figure 4-22). This is one of the most ancient of masonry arch structures and
still is used for underground water reservoirs of concrete without any reinforcing.
4.7. Conclusion:
This chapter was a concise part of the application chapter in the final thesis and
aimed the variation of deflection and load path in shades, using woven polyester
as the main cloth and carbon fibre as the reinforcement resin. FE Analysis program
Page | 109
Chapter4Numerical Study of load paths in Shell and Plate Structures
was used to complete the modelling through geometric nonlinear static analysis. In
order to verify the accuracy of the analysis, the final Z deflection was compared to
Timoshenko’s Theory of Shells and Plates. The possible load paths were obtained
Link181 and Shell181 elements. The load paths obtained in the rectangle model
revealed that there are two possible strips in the shade, which can hold the
majority of the load caused by normal pressure. In the clamped rectangle model
two strips bridged the opposite sides while in the corner fixed model they
connected the corners and jointed in the centre of the shell. The procedure
contained within a tension structure. Identifying the possible load paths leads to
understand how a structure is likely to support its load and let the designer to
Page | 110
CHAPTER 5.
APPLICATION TO CRUISING
YACHT SAIL DESIGN
5.1. Introduction:
The development of modern sailing boats has been based almost entirely on the
slight contribution from scientists and technologists and using just basic scientific
principles (Beck et al., 2009). In recent years, urgent and strong requests for
improved performance, mostly for racing yachts, have guided the interest and the
light weight and high strength which are rarely studied in the literature. This
chapter focuses on the importance of the quantitative evaluation of the sail loads
and studies how these loads contribute to the improvement of the performance of a
Chapter5Application To Cruising Yacht Sail Design
sailboat through the development of an algorithm for aiding sail design. The
supplies load path design features in a sail configuration. The major achievement
load paths under distributed load and the consequent deformed sail-shape. Hence,
the outcome of the research improves sail performance and design of new sails.
The scope of the current study is to apply a load path determination approach to
sail design, to define the possible carbon fibre reinforcement directions for
laminated sails for any given shape using a numerical algorithm, this will lead to an
A shell structure is usually defined as a surface structure that supports only tensile
curvature and membrane stress. In an ideal case where the thickness tends to
zero, the material can only resist tension a shell is termed a membrane. Textile
and laminate structures such as boat sails are good examples of membranes. Boat
and constraints. However, in most cases deformation occurs such that the solution
Page | 112
Chapter5Application To Cruising Yacht Sail Design
Shape deformation of the sail system should be considered in order to predict the
correct load path determination in sail. However, the shape deformation in general,
depends on seaming and rig tension. Currently, effect of elastic deformation is not
and Kim, 2006). This work, however, has made an attempt to predict the deformed
shape of the sail in the static structural analysis with some simplified assumption
Initially, it is necessary to introduce some terms that will be used later in this
chapter to describe components of the sails. Figure 5-1 illustrates the components
of a mainsail. Prior to simulating the boat sail, following assumption and condition
is made:
The incident angle of apparent wind is set to 90º with respect to the mast.
Page | 113
Chapter5Application To Cruising Yacht Sail Design
cruising boat sail is presented. Static analysis will be performed in ANSYS APDL
as the finite element package. Due to the fact that sails behave similar to very thin
Page | 114
Chapter5Application To Cruising Yacht Sail Design
provide the most degree of freedom of common element types. The sail cloth is
discretised into finite elements with a set of quadrilateral membrane elements with
four nodes. These elements withstand all external forces through tension but they
cannot withstand bending moments. However, these elements are not easy to use
particularly in static analysis since they can distort easily. In this case the bending
due to the large deflection in the plate. In order to satisfy the enforced conditions in
the membrane, distributing the pressure through number of load steps, pre-
Stabilizer and Line search are implemented. An FEA static structural analysis is
conducted on a three dimensional rectangle model with 5.61 m mast length and
Table 5-1 depicts some common materials for most boat sails. The materials used
Blicblau et al., 2008). Woven polyesters can be used for sail cloth and carbon fibre
Page | 115
Chapter5Application To Cruising Yacht Sail Design
5.3.3. Element
FE mesh generation is carried out by using FEM modeller. This strategy is used to
quadrilateral elements with a side particularly coincident with mast and foot. The
quadrilateral elements are the proper choice to model yacht sails built by
composite materials with stiffeners since the elements in the side has been
assumed to be origin of the a local reference. They can define the composite
elements used earlier in Chapter 4 is implemented for sail modelling due to the
reasons were explained earlier. It is worth mentioning that the elements do not
resist any bending moment since “membrane only” is activated. The element
formulation is based on Midlin’s theory where ANSYS codes reduce “locking” effect
(Zienkiewicz and Taylor, 2005). It should be noted that sails are modelled without
considering their effective shape, but simply like shells to which loads are applied,
stressing the structure by means of foot, mast and shrouds. Particular care is spent
in finding a suitable mesh which could to achieve reliable results with spending
reasonable duration of computational time. The final mesh shown in Figure 5-2 is
Page | 116
Chapter5Application To Cruising Yacht Sail Design
50
45
40
Z displacement(mm)
35
30
25
20
15
10
5
0
4 5 6 7 8 9 10 11 12
Number of division
Page | 117
Chapter5Application To Cruising Yacht Sail Design
size of the element is reduced. Mesh loads should be refined carefully but this
increases the time of analysis (Figure 5-4). Therefore, time of analysis and
essential since in cases such as when the load paths algorithm is implemented;
560
510
460
410
360
Time(S)
310
260
210
160
110
60
4 5 6 7 8 9 10 11 12
Number of division
desktop computer with 3.00 GHz CPU and 4.00 GB RAM and hard drive of 150 GB .
Generally sails are modelled in such that the translational and rotational degrees of
freedoms along the foot of the main sail where the sail is connected to the boom
will be suppressed. This suppression fixes all the displacement in these regions.
Page | 118
Chapter5Application To Cruising Yacht Sail Design
Along the mast, the translational degrees of freedom must be inhibited. For the
head sail, the point of attachment of the foot to the rig must be constrained. The
leech of the main and head sail must be allowed to move freely to induce a
geometric twist due to the aerodynamic loading and the mast is assumed to be
rigid during the structural calculations. Due to the complexity of the simulation
through static structural analysis, some simplification is carried out in the boundary
Boundary conditions
Prior to applying any pressure to the sail, the cloth must be stiff enough to resist
membrane bending under the wind load. Thus, two scenarios are considered
the sail in installation stage and 2- stiffening of the membrane in normal direction
foot, mast and leech. The edge of the sail at the foot is pre-strained by 1 in the
downward 𝑦 direction, the edge of the sail at the mast is displaced 2.82 in
positive 𝑥 direction and the leech is displaced with -1.3 in 𝑥 direction and -
the sail elements to overcome the divergence issue in the solution (Figure 5-6).
It is followed by a ramp uniform pressure to the entire sail cloth uniformly. The
constant during the procedures. Essentially, the wind load is applied gradually
Page | 120
Chapter5Application To Cruising Yacht Sail Design
through the analysis to the model since ramp increment of pressure helps the
steps and 38 sub steps. The pressure is obtained from Equation 5-1 indicates
𝑞 Equation 5-1
where 𝑞 is the maximum dynamic pressure, is the air density in the temperature
of +25 and is the summation of boat and wind velocities (25 knots). The
60
50
Pressure(Pa)
40
30
20
10
0
1 2 3 4 5 6 7 8 9 10 11
Loading step number
of the analysis but also for better understanding of the behaviour of boat sail in
different boundary conditions along with their effects on load path determination.
As the initial trial, the presence of battens and shrouds and other structural
Page | 121
Chapter5Application To Cruising Yacht Sail Design
elements of the sail rig are neglected from the analysis and the exterior edges are
fully fixed and then the membrane is stretched. This implies that there would be no
degree of freedoms in the nodes coincides with these sides. However, there is free
movement for all other interiors nodes in all allowable DOF. Due to the necessity
for large deformation in the sail simulation, analysis must be performed with
step process that incrementally adds the wind load while obtaining a converged
displacement field for each step. Wrinkling of the structure, which is an important
consideration especially around the leading edge (luff) and at the sail tip, is not
Figure 5-8 illustrates the final result for the 𝑧 displacement normal to the plane of
the sail. It is obvious that the maximum displacement occurs in the centre of gravity
of the sail. The displacement in z direction decreases towards the edges of the sail
where the nodes are fully constrained. In some regions the displacement is
Page | 122
Chapter5Application To Cruising Yacht Sail Design
-8: 𝑧 deflection
Figure 5
Due to the fact that load path algorithm is mainly based on the boundary
conditions; variety of load path would be expected through different analysis. In the
To determine the possible load paths, Link180 elements are overlaid on the
Shell181 elements. Mesh generation needs to follow the previously defined pattern
Page | 123
Chapter5Application To Cruising Yacht Sail Design
(explained in Chapter 4). Link members are linked to the shell elements in the
outer edges and in the diagonal orientation. Figure 5-9 and Figure 5-10 show how
Figure 5-10 which is a zoom-in section from the entire sail model, visualizes how
Shell181 element
Link180 element
After setting up the link elements in the model, the evolutionary algorithm is
allowed to iterate for 50 times. The resulting pattern of links can be found in
Figure 5-11.
Page | 124
Chapter5Application To Cruising Yacht Sail Design
(a) (b)
Studying the generated paths reveals that they mainly connect the mast to the
leech and the foot similar to the expected pattern for shade cloths restrained at the
boundaries. The paths expected in sails are indicated by the black lines showing
the carbon fibre tapes (Figure 5-12). The final sail design should derive its stability
from the carbon fibre paths. The number of these paths must be adequate and
accurate enough to make the design far from instability failure. As Figure 5-12
depicts the population of the carbon fibres are much higher than found in
Figure 5-11. The paths obtained by the numerical algorithm do not reflect paths
Boundary condition in the real sail structures differ from the boundary
Page | 125
Chapter5Application To Cruising Yacht Sail Design
by link element in only diagonal and boundaries of the shell elements. There
would be some paths in the sail which do not conform to this limitation thus
trusses in Chapter 3
In a real sail as depicted in Figure 5-12, the leech is free to move and there is no
constraints limiting the movement of the membrane in this side. Since there is no
Page | 126
Chapter5Application To Cruising Yacht Sail Design
Many parameters can affect the final link distribution. For better understanding of
these parameters they should be included in the modelling separately. One of the
significant parameters is existence of battens which can support the main sail cloth
in the normal direction. Usually four to five battens are included in the sail,
horizontally normal to the mast, depending on the size of the main sail. In the
Besides, adding battens to the model affects both the mesh in the main cloth, and
the generation of link elements. Since the main cloth is divided into five different
sections more regular patterns of mesh with quadrangular shell elements can be
Page | 127
Chapter5Application To Cruising Yacht Sail Design
Battens physically presents larger cross-sectional area than the carbon fibre
reinforcing tape in boat sails. In this model the cross section area of 100 mm2 is
chosen for the batten while the links representing the carbon fibre tape are given a
cross-sectional area of 49 mm2 . Battens are modelled with beam elements which
withstand bending moment. They can remain until the end of the analysis and
should not be removed from the solution domain. Figure 5-14 shows the relative
Batten
Due to the great number of analysis with different structural factors, a summary of
boundary conditions for better understanding is provided in Table 5-2 which will be
Page | 128
Chapter5Application To Cruising Yacht Sail Design
Attempt 1
The new boundary conditions free the nodes in leech and foot to move in any
allowed direction, while, they are fixed to move only rotationally on the edge at the
mast. The sail is fixed at the head and clew as well as at the tack in all DOF. It is
(along the pressure application) from the centre of gravity in earlier model move to
the middle of the leech since no constraints limited the nodes in the related zones
and the nodes are free to move in all DOF. While it still has the minimum
magnitude in the corners and at the mast since nodes are fixed in all translational
Page | 129
Chapter5Application To Cruising Yacht Sail Design
Implementing the load path determination algorithm produce the paths plotted in
Figure 5-16. Figure 5-16 (a) depicts how boundary conditions are applied to nodes
in the model and Figure 5-16 (b) shows how load path directions are affected by
the imposed boundary condition and how are the load path patterns.
Figure 5-16 explains how the load travels between the head and clew/ foot. It is
expected that in real sail most of the forward thrust of the sail is applied to the clew
and push on the mast through the boom. The patterns in the links reflect this path
for the load from the sail (Figure 5-16). Nevertheless, the directions of the paths
are restricted to be parallel to the edges and diagonals of shell elements in the
mesh. Hence more paths would be observed if this restriction is eliminated .So an
(a) (b)
Figure 5-16: (a) Mesh generation and boundary conditions (b) Load paths defined by the link removal
algorithm.
Page | 130
Chapter5Application To Cruising Yacht Sail Design
Attempt 2
The link elements included in this initial model are shown in Figure 5-17. In the set
up stage, the sail in the head is tensioned with additional cable force representing
pre-tensioning that can be applied when the sail is raised. Similar forces can be
applied to the clew and the tack. These forces are added gradually in different
added to the head corner from the tension shrouds cables for this purpose.
The result of the model in Figure 5-18 indicates that adding tensioning cable effect
in the sail head introduces a force that flows downward from the head to the foot
and generating more paths compare to the previous analysis. In this design the
wind load travels from head to clew as it transfer shear stress in the sail. The loads
Page | 131
Chapter5Application To Cruising Yacht Sail Design
need to cause failure of the model depends on the boundary conditions and shape
of the sail and the point where load applied. This dependency of sail strength upon
boundary conditions and carbon fibre paths is an advantage since if a scale model
of a sail is stable, the full scale structure would also be stable and as a result the
(a) (b)
Attempt 3
In third attempt, only the effect of different combination of the tensioning cables in
the head and in the clew is investigated. The result arising from these changes is
Page | 132
Chapter5Application To Cruising Yacht Sail Design
(a) (b)
Figure 5-19: load path with force in the head and clew
In spite of adding the tension cable in the clew, the improvement is insignificant
although there are some concentration of remaining links in the head and transfer
of load to the clew. From these results, it is realized that the ideal pattern for
carbon fibre is needed to consist some straight patterns from head to foot to
transfer the load on the sail. Yet, effect of the shrouds in the battens has not been
studied. These tension cables loads are introduced to the sail at the ends of the
battens to control the shape of the sail which possibly affect the load path
Page | 133
Chapter5Application To Cruising Yacht Sail Design
Attempt 4
to determine the tension force caused by ropes, the reaction forces in the clew and
tack is calculated from the model. Then wards 10% of these magnitudes would be
added to leech and mast to model the shrouds effects. The reaction forces in the
Application of the load on the battens can be found by cable forces in the locations
Page | 134
Chapter5Application To Cruising Yacht Sail Design
Since there is no certainty about the magnitude of the cable force in the sail, two
Different value for mast and leech according to the table above
From the first scenario, magnitude of cable is identical in both sides, Figure 5-22.
These cable forces need to be distributed into the membrane representing the sail.
Some links can be seen at the corners where the forces are applied.
Page | 135
Chapter5Application To Cruising Yacht Sail Design
(1) (2)
In the second scenario, shrouds with different reaction forces value is applied to
the sail (Figure 5-23). As it was expected more link elements remain in the model
when the leech and mast cable forces are increased. The links present a similar
therefore expected that the boundary conditions are very close to the real
are those who will not fail under loading. Therefore the sail stability under the wind
load with combination of different boundary condition depends on the new lines of
Page | 136
Chapter5Application To Cruising Yacht Sail Design
(1) (2)
Attempt 5
The final boundary conditions are in such that all the nodes in the foot are fixed to
move in, x , y and z directions while those in the mast are fixed in x and y only and
all nodes in the leech are free to move in all allowable DOF (Shankaran, 2005).
The effect of shrouds differs from leech to mast where they experience higher
reaction forces (Figure 5-24) and the membrane bears 98 N from the tensioning
cable in the head and tack. The final boundary condition and result and from the
Page | 137
Chapter5Application To Cruising Yacht Sail Design
Figure 5-24: Boundary conditions Figure 5-25: Final possible load paths
In the final model a new pattern of links reveals more details of cruising yacht sails
carbon fibres. The lines of resins are obtained as a series of mainly straight lines.
This network of lines represents the network of force in the sail under loading
conditions and identify of how the structure is likely to support its load. The
accuracy with which model can model the forces in the structure in upon an ideal
meshing patterns and meshing density, the denser the mesh, the more accurate
the results. Nevertheless, due to the some limitations such as their manufacturing
cost and complexity (neither of them have been considered in this study) these
types of patterns might not be practical. However the validation of the method is
examined by finding pointing vectors which was discussed in Chapter 3 in the truss
structures. These pointing vectors then are numerically found in the sail for number
5.5. Conclusion:
In this chapter, load path theory was extended to the application of cruising boat
sail. Uniform distributed pressure representing wind load with speed of 20 knots
perpendicular to the mast was applied to the sail and the 𝑧 deflection was
calculated through the static structural analysis. Due to the large deflection, the
advance features in the FEA package ANSYS such as stabilizer and line search
for force convergence. The results were validated against the results in the
experimental use of carbon fibre resin in the industry via six different attempts. It
was found out that boundary conditions affect the final load path directions
Page | 139
CHAPTER 6.
CONCLUSION
element. It provides the designer with insight into how the structure is performing
its primary function of transferring load from the point of application to the supports.
structures.
The first objective was to integrate existing load path plotting algorithms into the
FEA package ANSYS though the development of new FORTRAN codes. Some
domain. It concluded that typical patterns observed for the paths can be classified
into five distinct categories: 1-pure tension, 2-pure compression, 3-pure bending,
Chapter6Conclusion
4-cantilever bending and 5-simply supported bending. Load paths were then
the elements. It was observed that this method tended to result in a statically
determinant design and terminated when the removal of the next member created
a mechanism.
ANSYS. The analysis was extended to geometric nonlinear analysis due to the
large deflection experienced by shade sail and wind sail structures under uniform
static pressure. The main load bearing members in the structure were identified by
removing elements from the domain using an evolutionary method. The design
language. The integrity of this design technique was verified by the development of
promising patterns for clamped and corner fixed rectangles. In the clamped model,
two straps connected the opposite sides and met each other in the centroid to
essentially create a cross pattern. In the corner fixed model, the opposite corners
The final objective was to apply the algorithms to cruising yacht sails. The results
presented have similarities between the paths defined by this method and the
Additionally, it was concluded that boundary condition can affect the final result
Page | 141
Chapter6Conclusion
considerably. Finally pointing vectors were calculated at the element centroid for a
number of elements in the main cloth and plotted on the boat sail. These pointing
vectors revealed connections between the fibre paths created by the proposed
This research program for a Master by Research Thesis has focussed on the
These included FORTRAN programs that interfaced with ANSYS by the export of
results from ANSYS and the return of new link elements to plot vectors and
The application to defining reinforcement patterns in shade sail and wind sail
structures was limited by the time allowed for this thesis. The results from these
One interesting outcome of the research has been the development of procedures
to plot load paths in simple trusses. The truss in Figure 6.1 has the minimum
number of axial force members before the truss becomes a mechanism. However
not all members are participating in the transfer of the Y-force from Node 3 to the
support. For structures supporting multiple load cases the identification of the path
for one of the applied loads is not achieved by simply optimizing the structure. A
Page | 142
Chapter6Conclusion
subset of the members will in general be associated with the load transfer. Here
the remaining members are required to provide a path for the bending loads -
except member 24 which carries no load. This member is required to prevent the
N4
1.414T
1.414C
(1) (1)
N1 N2 N3
1 1
-1: 𝑦-Force load path in a simple truss.
Figure 6
The recommendations arising from this work relate mainly to more in-depth
investigations of the applications that have been used to check the performance of
The design load cases for the shade sails will require a detailed study of
wind loads and will depend on the geometry of the sail. Probably the major
the shade sails and the wind sails and the use of more representative
design loads.
Page | 143
Chapter6Conclusion
The work completed showed that the load paths are also sensitive to the
manufacturing and cost constraints. A design requirement for the wind sails
would be the final aerodynamic shape of the sail. While the reinforcement
ensures the load enhancement near the supports does not cause failure,
away from the supports the fibres could help to control the shape of the sail.
project.
Page | 144
REFERENCES
Brezzi, F., Bathe, K. J. and Fortin, M. (1989). "Mixed‐interpolated elements for Reissner–
Mindlin plates." International Journal for Numerical Methods in Engineering 28(8):
1787-1801.
Bruggi, M. (2008). "On the solution of the checkerboard problem in mixed-FEM topology
optimization." Computers & Structures 86(19–20): 1819-1829.
Budynas, R. G. (1977). Advanced strength and applied stress analysis, McGraw-Hill.
Cappello, F. and Mancuso, A. (2001). "Lay-up optimization for the hull of a racing sailing
yacht." Advances in Engineering Software 32(2): 133-139.
Chan, H. (1963). Optimum Mitchell Frameworks for Three Parallel Forces, The colledge of
aeronautics cranfield: 26.
Chaperon, P., Heller, M., Jones, R., Pitt, S. and Rose, F. (1998). Load flow visualisation in
structural optimisation. Australasian Conference on Structural Optimisation,
Sydney; Australia.
Cheng, K. and Olhoff, N. (1981). "An investigation concerning optimal design of solid
elastic plates." International Journal of Solids and Structures 17(3): 305-323.
Christensen, P. and Klarbring, A. (2008). An introduction to structural optimization,
Springer Verlag.
Cook, R. D. (2001). Concepts and applications of finite element analysis, Wiley.
Cooper, A. (1972). "Trajectorial fiber reinforcement of composite structures(feasibility
analysis)[Ph. D. Thesis]."
Crosky, A., Kelly, D., Li, R., Legrand, X., Huong, N. and Ujjin, R. (2006). "Improvement
of bearing strength of laminated composites." Composite Structures 76(3): 260-271.
CyTec. (2011). "CyTec Engineered Materials." Retrieved July, 2011, from
http://www.cytec.com/business/EngineeredMaterials/marinemarket.html.
Duysinx, P. and Bendsøe, M. (1998). "Topology optimization of continuum structures with
local stress constraints." International Journal for Numerical Methods in
Engineering 43(8): 1453-1478.
Fallow, J. B. (1996). "America's Cup sail design." Journal of Wind Engineering and
Industrial Aerodynamics 63(1–3): 183-192.
Flabel, J. C. (1997). Practical Stress Analysis for Design Engineers: Design and Analysis of
Aerospace Vehicle Structures, Lake City Publishing Company.
French, M. (1992). "Form, Structure and Mechanism." Macmillan Education Ltd, Hong
Kong
Gilliam, j. (2006). History of Sailing Yacht Masts, Rigging and Sails 1900-Present day.
Haber, R., Jog, C. and Bendsøe, M. (1996). "A new approach to variable-topology shape
design using a constraint on perimeter." Structural and Multidisciplinary
Optimization 11(1): 1-12.
Haftka, R. and Prasad, B. (1980). "Optimum design of the sides of a railroad car."
Engineering Structures 2(4): 230-236.
Harasaki, H. (2000). Concept of Transferred Force and Its Use in Topology Design of
Structures, University of Iowa.
Harasaki, H. and Arora, J. (2002). "Topology design based on transferred and potential
transferred forces." Structural and Multidisciplinary Optimization 23(5): 372-381.
Harasaki, H. and Arora, J. S. (2001). "A new class of evolutionary methods based on the
concept of transferred force for structural design." Structural and Multidisciplinary
Optimization 22(1): 35-56.
Page | 146
References
Harasaki, H. and Arora, J. S. (2001). "New concepts of transferred and potential transferred
forces in structures." Computer Methods in Applied Mechanics and Engineering
191(3-5): 385-406.
Harasaki, H. and Arora, J. S. (2001). "Optimal structural design with indirect use of
Transferred Forces." Structural and Multidisciplinary Optimization 22(5): 384-393.
Heppel, P. (2002). Accuracy in sail simulation: Wrinkling and growing fast sails.
Hoshino, H., Sakurai, T. and Takahashi, K. (2003). "Vibration reduction in the cabins of
heavy-duty trucks using the theory of load transfer paths." JSAE Review 24(2):
165-171.
Huang, X. and Xie, Y. (2011). "Evolutionary topology optimization of continuum
structures including design-dependent self-weight loads." Finite Elements in
Analysis and Design 47(8): 942–948.
Huang, X. and Xie, Y. M. (2010). Evolutionary topology optimization of continuum
structures method and applications, John Wiley & Sons, 2010.
Jazzmanian. (2006). " Illustration showing various parts of a small sailboat." Retrieved
20/5/2012, 2012, from
http://upload.wikimedia.org/wikipedia/commons/b/bd/Parts_of_sailboat.jpg.
Jones, S. and Platts, M. (1998). "Practical matching of principal stress field geometries in
composite components." Composites Part A: Applied Science and Manufacturing
29(7): 821-828.
Juvinall, R. C. and Marshek, K. M. (1991). Fundamentals of machine component design,
John Wiley & Sons.
Karkauskas, R. and Norkus, A. "Truss optimization under stiffness, stability constraints and
random loading." Mechanics Research Communications 33(2): 177-189.
Karnovsky, I. A. and Lebed, O. (2009). Advanced methods of structural analysis, Springer
Verlag.
Kelly, D. and Elsley, M. (1993). "A procedure for determining load paths in elastic
continua." Engineering Computations 12(5): 415-424.
Kelly, D., Hsu, P. and Asudullah, M. (2001). "Load paths and load flow in finite element
analysis." Engineering Computations 18(1/2): 304-313.
Kelly, D., Hsu, P. and Asudullah, M. (2001). "Load paths and load flow in finite element
analysis." Engineering Computations 18(1-2): 304-313.
Kelly, D., Reidsema, C., Bassandeh, A., Pearce, G. and Lee, M. (2010). "Algorithms to
Define Load Paths and Load Bearing Topology from Finite Element Analysis."
Finite Elements in Analysis and Design.
Kelly, D., Reidsema, C. and Lee, M. (2010). On load paths and load bearing topology from
finite element analysis, IOP Publishing.
Kelly, D. and Tosh, M. (2000). "Interpreting load paths and stress trajectories in elasticity."
Engineering Computations 17(2): 117-135.
Kelly, G. (2005). "Load transfer in hybrid (bonded/bolted) composite single-lap joints."
Composite Structures 69(1): 35-43.
Kermode, A. C. (1964). The Aeroplane Structure (Second Edition ) Pitman
Lackman, L. and Ault, R. (1967). "Mollifying Stress Fields by Using Filamentary
Composite Materials." Advances in Structural Composite, Society of Aerospace
Materials and Process Engineers 12.
Page | 147
References
Li, B., Hong, J., Wang, Z., Wu, W. and Chen, Y. (2012). "Optimal Design of Machine Tool
Bed by Load Bearing Topology Identification with Weight Distribution Criterion."
Procedia CIRP 3: 626-631.
Li, R., Kelly, D., Crosky, A., Schoen, H. and Smollich, L. (2006). "Improving the
efficiency of fiber steered composite joints using load path trajectories." Journal of
Composite Materials 40(18): 1645.
McDoon, A. (2004). "Yachting and Boating World Forums " Retrieved 14/02/2012, 2012,
from http://www.ybw.com/forums/member.php?u=9824.
Michell, A. G. M. (1904). "The limits of economy of material in frame structures." Phil.
Mag 8(47): 589-597.
Nha, C. D., Xie, Y. and Steven, G. (1998). "An evolutionary structural optimization
method for sizing problems with discrete design variables." Computers & structures
68(4): 419-431.
O’Dwyer, D. (1999). "Funicular analysis of masonry vaults." Computers & Structures
73(1–5): 187-197.
Ohsaki, M. and Katoh, N. (2005). "Topology optimization of trusses with stress and local
constraints on nodal stability and member intersection." Structural and
Multidisciplinary Optimization 29(3): 190-197.
Osgood, C. C. (1970). Fatigue Design.
Osgood, C. C. (1982). Fatigue design. Oxford, England, Pergamon Press.
París, J., Martínez, S., Navarrina, F., Colominas, I. and Casteleiro, M. (2010). Topology
optimization of structures with stress constraints: Aeronautical applications, IOP
Publishing.
Payne, R. M., Mechanical (2008). A knowledge-based engineering tool for aiding in the
conceptual design of composite yachts.
Prager, W. (1977). "Optimal layout of cantilever trusses." Journal of Optimization Theory
and Applications 23(1): 111-117.
Querin, O., Steven, G. and Xie, Y. (1998). "Evolutionary structural optimisation (ESO)
using a bidirectional algorithm." Engineering Computations 15(8): 1031-1048.
Ranjbaran, A. (1992). A computer program for the stress analysis of reinforced concrete
structures, University of Manchester, Institute of Science and Technology.
Richards, D. and Chan, H. (1966). Developments in the theory of Michell optimum
structures, Advisory Group for Aeronautical Research and Development.
Sakurai, T., Takahashi, K., Kawakami, H. and Abe, M. (2007). "Reduction of Calculation
Time for Load Path U* Analysis of Structures." Journal of Solid Mechanics and
Materials Engineering 1(11): 1322-1330.
Schlaich, J., Schäfer, K. and Jennewein, M. (1987). "Toward a consistent design of
structural concrete." PCI journal 32(3): 74-150.
Shankaran, S. (2005). Numerical analysis and design of upwind sails, Citeseer.
Shinobu, M., Okamoto, D., Ito, S., Kawakami, H. and Takahashi, K. (1995). "Transferred
load and its course in passenger car bodies." JSAE Review 16(2): 145-150.
Sigmund, O. and Petersson, J. (1998). "Numerical instabilities in topology optimization: a
survey on procedures dealing with checkerboards, mesh-dependencies and local
minima." Structural and Multidisciplinary Optimization 16(1): 68-75.
Page | 148
References
Signorini, G. F., Barrat, J. L. and Klein, M. L. (1990). "Structural relaxation and dynamical
correlations in a molten state near the liquid–glass transition: A molecular dynamics
study." The Journal of chemical physics 92: 1294.
Singh, K. (1996). Mechanical design principles, Nantel Publications.
Suzuki, K. (1991). "Shape and layout optimization using homogenization method(Ph. D.
Thesis)."
Svanberg, K. and Werme, M. (2007). "Sequential integer programming methods for stress
constrained topology optimization." Structural and Multidisciplinary Optimization
34(4): 277-299.
Takahashi, K. (1986). "Relative Rigidity of Structures and Saint Venant's Principle."
Transactions of the Japan Society of Mechanical Engineers, Series A 52(484):
2615-2621.
Timoshenko, S., Woinowsky-Krieger, S. and Woinowsky, S. (1959). Theory of plates and
shells, McGraw-hill New York.
Trimarchi, D., Vidrascu, M., Taunton, D. J., Turnock, S. and Chapelle, D. (2011). Fluid
structure interactions applied to downwind yacht sails. 10e Colloque National en
Calcul des Structures. Gien: 6.
Ullman, D. G. (1992). The mechanical design process, McGraw-Hill New York.
Viola, I. M. and Flay, R. G. J. (2011). "Sail pressures from full-scale, wind-tunnel and
numerical investigations." Ocean Engineering 38(16): 1733-1743.
Waldman, W., Heller, M., Kaye, R. and Rose, F. (2002). "Advances in two-dimensional
structural loadflow visualisation." Engineering Computations 19(3): 305-326.
Waldman, W., Heller, M., Kaye, R. and Rose, L. (1999). "Advances in structural loadflow
visualisation and applications to optimal shapes."
Wang, E., Yoshikuni, Y., Guo, Q., Nohara, T., Ishii, H., Hoshino, H. and Takahashi, K.
(2010). "Load Transfer in Truck Cab Structures under Initial Phase of Frontal
Collision." International Journal of Vehicle Structures & Systems 2(2).
Wang, M., Wang, X. and Guo, D. (2003). "A level set method for structural topology
optimization." Computer Methods in Applied Mechanics and Engineering 192(1-2):
227-246.
Wightsails. (2012, August 2012). "Cradle to Grave." Retrieved, August, 2012 from
(http://www.wightsails.com/recycling-sails/sailcloth-cradle-to-grave/)
Wu, H., Ying, X., Jin-gang, J. and Yang, L. (2009). "Dynamic Model of Alternative Load
Path for Structural Progressive Collapse." Earthquake Resistant Engineering and
Retrofitting.
Xie, Y. and Steven, G. (1993). "Optimal design of multiple load case structures using an
evolutionary procedure." Engineering Computations 11(4): 295-302.
Xie, Y. and Steven, G. (1993). "A simple evolutionary procedure for structural
optimization." Computers & structures 49(5): 885-896.
Xie, Y. and Steven, G. P. (1993). "A simple evolutionary procedure for structural
optimization." Computers & Structures 49(5): 885-896.
Xie, Y. M. and Steven, G. P. (1997). "Evolutionary structural optimization."
Xie, Y. M. and Steven, G. P. (1997). Evolutionary structural optimization, Springer.
Yang, R. and Chuang, C. (1994). "Optimal topology design using linear programming."
Computers & structures 52(2): 265-275.
Page | 149
References
Page | 150
APPEDNDIX.A
fini
/clear
Geometry set up
! Nodes creation
/Prep7
*do,i,1,n_max
*do,j,1,m_max
n,,L*i,L*j,L*i*(L*i-1000)*L*j*(L*j-1000)/1E10
*enddo
*enddo
mp,ex,4,y4
mp,prxy,3,v3
*enddo
e,i*m_max,(i+1)*m_max
*enddo
*do,j,1,m_max-1
e,(n_max-1)*m_max+j,(n_max-1)*m_max+j+1
*enddo
Page | 152
Appendix.A
Boundary conditions
nsel,s,loc,x,L
nsel,a,loc,x,n_max*L
nsel,a,loc,y,L
nsel,a,loc,y,m_max*L
cm,nextr,node
nsel,s,loc,x,L
cm,nextr_x_L,node
nsel,s,loc,x,n_max*L
cm,nextr_x_r,node
nsel,s,loc,y,L
cm,nextr_y_B,node
nsel,s,loc,y,m_max*L
cm,nextr_y_T,node
LOCAL,11,1,L,L,0, , , ,1,1,
LOCAL,12,1,m,m,0, , , ,1,1,
LOCAL,13,1,m,L,0, , , ,1,1,
LOCAL,14,1,L,n,0, , , ,1,1,
Csys,11
Page | 153
Appendix.A
nsel,s,loc,x,0,4*L
csys,12
nsel,a,loc,x,0,4*L
csys,13
nsel,a,loc,x,0,4*L
csys,14
nsel,a,loc,x,0,4*L
cm,ncorn,node
Csys,11
nsel,s,loc,x,0,L
csys,12
nsel,a,loc,x,0,L
csys,13
nsel,a,loc,x,0,L
csys,14
nsel,a,loc,x,0,L
cm,ncorn1,node
cmsel,,ncent
!Defining center element
ESLN,S,1
esel,r,type,,2
cm,ecent,elem
allsel
nplot
eplot
/user
fini
Save,modell,db ! Saving the geometry
Solution
/solu
antype,static
nlgeom,on ! (Non Linear geometric
solution)...
sstif,on ! Stress stiffening...
outres,all,all
! Constraining all DOFs and then redefine and stiffening the shell
time,1.0
nsubst,10,100,2
d,nextr,all
d,nextr_x_L,ux,-m_max*L/2000.0
d,nextr_x_r,ux,m_max*L/2000.0
d,nextr_y_B,uy,-n_max*L/2000.0
Page | 154
Appendix.A
d,nextr_y_T,uy,n_max*L/2000.0
allsel
solve
time,2.0
nsubst,10000,100000,10
sfe,ecent,,pres,,pressure
allsel
solve
time,3.0
nsubst,100,1000,1
d,nextr_x_L,ux,0
d,nextr_x_r,ux,0
d,nextr_y_B,uy,0
d,nextr_y_T,uy,0
allsel
solve
time,5.0
nsubst,10000,100000,10
sfe,ecent,,pres,,pressure*1E-6
allsel
solve
Page | 155
Appendix.A
time,6.0
nsubst,1000000,100000000,10
esel,s,type,,2
sfe,all,,pres,,pressure*1E-10
allsel
solve
time,7.0
nsubst,1000000,100000000,10
esel,s,type,,2
sfe,all,,pres,,pressure*1E-6
allsel
solve
time,8.0
nsubst,1000000,100000000,10
esel,s,type,,2
sfe,all,,pres,,pressure*1E-4
allsel
solve
time,9.0
nsubst,1000000,100000000,10
esel,s,type,,2
sfe,all,,pres,,pressure*1E-2
allsel
solve
time,10.0
nsubst,1000000,100000000,10
esel,s,type,,2
sfe,all,,pres,,pressure*1E-1
allsel
solve
time,11.0
nsubst,1000000,100000000,10
esel,s,type,,2
sfe,all,,pres,,pressure
allsel
solve
Post Processing
plnsol,u,z
! Preparing etable for determining
of the internal force
etable,eras
esel,s,type,,1
etable,mforx,smisc,1
SMIN,Min,MFORX,MFORX,1,1,
SABS,1
ESORT,ETAB,Min,1,1, ,
*GET,MForVal,SORT,,Min
esel,s,etab,min,0,MForVal+0.01*Mforval,,1
cm,emin,elem
allsel
CMWRITE,emin,cm
Finish
*Do,ii,1,10
resume,modell,db
/input,emin,cm
/prep7
allsel
cmsel,s,emin
emodif,all,mat,3
allsel
Finish
/solu
antype,static
nlgeom,on ! (Non Linear geometric
solution)...
sstif,on ! Stress stiffening...
outres,all,all
! Constraining all DOFs and then
redefine and stiffening the shell
time,1.0
nsubst,10,100,2
d,nextr,all
d,nextr_x_L,ux,-m_max*L/2000.0
d,nextr_x_r,ux,m_max*L/2000.0
d,nextr_y_B,uy,-n_max*L/2000.0
d,nextr_y_T,uy,n_max*L/2000.0
allsel
solve
time,2.0
nsubst,10000,100000,10
sfe,ecent,,pres,,pressure
allsel
Page | 157
Appendix.A
solve
time,3.0
nsubst,100,1000,1
d,nextr_x_L,ux,0
d,nextr_x_r,ux,0
d,nextr_y_B,uy,0
d,nextr_y_T,uy,0
allsel
solve
time,5.0
nsubst,10000,100000,10
sfe,ecent,,pres,,pressure*1E-6
allsel
solve
time,6.0
nsubst,1000000,100000000,10
esel,s,type,,2
sfe,all,,pres,,pressure*1E-10
allsel
solve
time,7.0
nsubst,1000000,100000000,10
esel,s,type,,2
sfe,all,,pres,,pressure*1E-6
allsel
solve
Page | 158
Appendix.A
time,8.0
nsubst,1000000,100000000,10
esel,s,type,,2
sfe,all,,pres,,pressure*1E-4
allsel
solve
time,9.0
nsubst,1000000,100000000,10
esel,s,type,,2
sfe,all,,pres,,pressure*1E-2
allsel
solve
time,10.0
nsubst,1000000,100000000,10
esel,s,type,,2
sfe,all,,pres,,pressure*1E-1
allsel
solve
time,11.0
nsubst,1000000,100000000,10
esel,s,type,,2
sfe,all,,pres,,pressure
allsel
solve
set,last
etable,eras
etable,mforx,smisc,1
SMIN,Min,MFORX,MFORX,1,1,
set,last
etable,eras
esel,s,type,,1
cmsel,r,emin
esel,inve
esel,r,type,,1
etable,mforx,smisc,1
SMIN,Min,MFORX,MFORX,1,1,
SABS,1
ESORT,ETAB,Min,1,1, ,
*GET,MForVal,SORT,,Min
esel,s,etab,mforx,-
ABS(3*MForVal+0.01*Mforval),ABS(3*MForVal+0.01*Mforval),,1
cmsel,a,emin
cm,emin,elem
allsel
CMWRITE,emin,cm
Finish
*Enddo
Page | 159
APPEDNDIX.B
In the last section of the thesis, plotting vectors and paths defined previously in
Chapter 3 is developed for truss structures. Procedure for plotting pointing vectors
at nodes using sections in regular grids to define an average stress on the section
at the node is explained. From the forces in the members average stresses
representing the load transfer is also calculated. The pointing vectors by internal
force in the member then iterated for numbers of nodes in the sail and it is found
out that it can be validated against the experimental carbon fibre path in industry.
Appendix.B
The aim of this method is to focus on a node and use the forces in the members
and the dimensions of the grid to define average stresses representing the load
transfer. A four bay truss in a regular grid pattern with 2 Lx width and 2 Ly height
Lx
Ly
2 1 Ni
3 7
6
4 5
𝑉 𝑗 Equation B- 1
where yx is the shear on a plane whose normal is in the 𝑦 direction, positive in the
positive 𝑥 direction and links carry only axial load. Then a cutting plane normal in 𝑥
Page | 161
Appendix.B
F2
2
2 Ly Ly
α1
3 F3 Px
6
2
α2
4
6
𝐹 F4
2
B- 2:
∑𝐹 𝑃 𝐹 𝐹𝐶 𝑠 𝐹𝐶 𝑠 Equation B- 2
Hence, is the summation of force in 𝑥 direction over the length of the member
as formulated in Equation B- 3:
𝑃 𝐹 𝐹𝐶 𝑠 𝐹𝐶 𝑠
Equation B- 3
𝐿 𝐿
Figure B. 3 :
Page | 162
Appendix.B
3 6 7
𝐹4 𝐹
2
5 𝐹
𝐹
2
2
6
2
𝐹 6
6
Lx
Px
F3/2 F7/2
β2 β1
F4 F6
𝐹 𝐹
𝑃 𝐹𝐶 𝑠 𝐹𝐶 𝑠 Equation B- 4
𝐿 𝐿
Similarly, 𝑦-path pointing vectors can be determined from load path theory as
Equation B- 5:
𝑉 𝑗 Equation B- 5
axis in Figure B. 4:
Page | 163
Appendix.B
4 6
𝐹 5 𝐹
2
2
𝐹
6
2
6
Lx
Py
γ1 γ2
F4 F6
F5
𝑃 𝐹 𝐹𝐶 𝑠 𝐹𝐶 𝑠
Equation B- 6
𝐿 𝐿
F1/2
1 F2 θ1
2
𝐹
Ly 3 𝐹 2 Py
2
4𝐹
6
θ2
2
6𝐹5 F4
2
𝐹 6 F6/2
𝐹
6 2
6
2
𝐹 𝐹
𝑃 𝐹𝐶 𝑠 𝐹𝐶 𝑠 Equation B- 7
𝐿 𝐿
Page | 164
Appendix.B
Procedure for obtaining 𝑥-path and y-path vectors in a regular grid is explained.
These procedures have been implemented in the finite element package to define
Y-Path and X-Path pointing vectors for yacht sails without battens (Figure B. 6 and
Figure B. 7)
(a) (b)
Figure B. 6: y-path pointing vectors in yacht boat sail (a), Zoom in section (b)
Page | 165
Appendix.B
The pointing vectors in 𝑦-path direction tend to form paths between the clew and
the head and the pointing vectors in the 𝑥-path direction form an arch between the
clew and the mast. These images show that in the vicinity of the clew the pointing
vectors for both the 𝑥-path and the y-path are directed radially away from the
support point. This indicates that radial direction of fibre reinforcement near the
clew and near other load introduction points on the sail can reinforce the cloth for
If the grid is skewed as in Figure B. 8, the procedure of finding pointing vectors can
Page | 166
Appendix.B
Stresses in the global axes can then be defined using transformation of the stress.
(Figure B. 9)
Ni
σyx
σx’y’
ϕ
σx’x’
𝐶 𝑠 𝐶 𝑠
𝐶 𝑠 𝐶 𝑠 Equation B- 8
𝐶 𝑠 𝐶 𝑠 𝐶 𝑠
Then from the first equation above, , the second equation, and the final
equation value of can be checked. Once all of the parameter calculated the
Page | 167