53
Chapter 4
Design and Analysis of the Proposed Composite
Post and Related Structural Elements
In the previous section, the analysis of the steel post was presented. The present section differs
greatly from the previous in that the proposed composite post must be designed before any
type of analysis treatment can be done. In relation to this project, the original steel post has a
particular geometry, material, and boundary conditions which are unchangeable, that is to say
this post has been designed previously and all that is required in this project is an analysis of the
structure. This analysis brings the project up to the present point where the determined
behavioural response is utilised in the initial design of the proposed composite post.
The chapter defines the geometry of the composite post, its structural elements, and a number
of laminate combination sub-models to be analysed. The analysis is carried out using the FEM
with the behavioural response magnitudes of the original post, in particular displacement,
being implemented as the constraints imposed on the proposed post model. The authenticity
of the FEM model is confirmed by semi-analytic analyses which include CLT and a smeared
approach. Additional loading scenarios are also analysed in the chapter while although
conservative, they represent loading issues specific to the proposed structures service
environment. Finally, a design and analysis of a steel resistant moment base is realised. The
objective of this related structure is to provide the constrained boundary conditions necessary
to enable the post to execute its service requirements.
4.1 Problem Data
4.1.1 Geometry
While the overall geometrical constraints are maintained, the geometry of the proposed post
differs slightly from the original steel geometrical configuration so as to incorporate general
fabrication methods of structures made from composite materials. The original steel structure
is composed of two U-section beams (UPN) fixed together by plates. The plates are connected
by butt-welding. This specific geometry creates problems if it is to be directly applied in
composite materials. A more efficient geometry in terms of fabrication and joining was
54
therefore examined and implemented in the composite model. In all, there are two types of
geometrical configurations presented in this project. The only difference geometrically between
the two is the inclusion of holes
hole or cut-outs
outs in the width of the post in one of the models. Apart
from this, the overall dimensions are the same in both models.
Within each model there lies a second variation. Each model is split into two sub-models
sub
which
are a result of distinct laminate
lamin
configurations for both. The first sub--model (A) of the two
contains a ply mix of woven fibres (weave) and unidirectional fibres (tape) while the second
sub-model
model (B) in each model contains plies of unidirectional fibres only. These two different ply
types are described more
re extensively in Section 4.1.3
4.1.3 of Materials. Below in figure 4.1 is a
schematic representation of the two different types of models and their subsequent variations
within due to different laminate composition. In total there are four post designs and will be
denoted by their number/letter comb
combination, e.g. 1B is the post
st structure with no holes
composed of laminates containing unidirectional fibres
fibre only.
1A) Weave + Tape
1) No Holes
1B) Tape
2 Model Types
2A) Weave +Tape
2) Holes
2B) Tape
The new structure consists of four separate structural components which include two identical
U-profile
profile beams and two identical
identic plates of varying width. The components are fixed together
through adhesion. The beam flanges are fixed to the inside of the plate surface and are aligned
along the plate edges so as to converge
converg towardss the top of structure. The holes which define
the second model type are contained in the plate component (mechanized post-curing)
post
and are
incorporated in the free surface of the plate, i.e. the plate surface area not directly bonded to
the UPNs. Figure 4.2 shows firstly the four components (including holes in plates of model 2)
separately but in their correct orientation and secondly the components joined together to
form the complete post.. The diameters of the holes are contained within the area of
o the plate
not fixed to the beam flanges and appropriately reduce along the leng
length
th of the post towards
the top.
55
A box-type section for the structure was decided upon instead of using an open-type (I-type)
section as its sectional configuration have a more effective resistance to torsion than the opentype section as all its parts at the periphery of the box section are connected to one another.
Secondly, while no torsional conditions are applied in the project data, the box section is a
design variable that is maintained from the original steel post structure.
A detailed geometric representation of the structure is shown in figure 4.3. The thicknesses of
the UPNs and plate elements have been omitted as they vary with different types of layer
configuration of the laminate. In this report, there are three distinct types of composite
configurations examined. As with the original steel structure, the geometric sectional
boundaries are identical so as to incorporate the cantilever and the catenary elements without
making any additional adjustments at their connections. The structure stands at a height of 7 m
which is actually 1 m shorter than its steel counterpart. This is due to the distinct boundary
conditions of the two which are described in detail later. The structures width varies from 460
mm at its base to 200 mm at the top. The thickness of the post is 140 mm.
56
Pressure Load
P (N/m2)
Perpendicular to line
(Case 1)
3072.886
57
Parallel to line
(Case 2)
2188.072
As outlined in the geometry, the proposed composite post differs from the original steel post in
terms of its method of fixation. While the original post of overall length of 8 m is embedded in a
concrete foundation of a depth of 1 m, the proposed post is instead fixed onto the top of this
same type of concrete pad by a steel moment resistant base as shown in figure 4.4. The overall
height of the proposed post is therefore reduced to 7 m with the bottom 200 mm fixed in the
moment resistant base. For simplification purposes of the composite model this bottom section
of 200 mm is treated as completely fixed, i.e. translations and rotations impeded. A separate
analysis of the combined base and post is carried out later in Section 4.8.
Figure 4.4: Composite post (blue) fixed to steel moment resistant base (grey)
As well as being the overall lighter structure of the two, the composite post with holes
experiences a reduced resultant force R for load Case 2, in comparison to the composite post
with no holes, due to its reduced surface area. However, this reduced surface area
consequently reduces the moment inertia (about the y-axis) and therefore its mechanical
resistance is negatively affected. This effect is analysed in more detail in the second part of
Section 4.1.4. As previously stated in Section 3.1.2, the additional loading effect due to the
catenary assembly is considered as a separate numerical analysis in Section 4.7.2.
58
4.1.3 Materials
The post is a composite based structure that is composed of two principal components, the
fibre and the matrix. The fibre component consists of a carbon based material. There are two
processes in which to fabricate the carbon fibre, they include the PAN (polyacrylonitrile) and
pitch process. The principal objective of the fibre component is to support approximately 70%
to 90% of the applied loading, and to provide mechanical resistance and stiffness to the
material. The matrix component consists of a polymeric resin which in this case is epoxy. The
objective of the epoxy is to give cohesion and maintain the fibres principal direction(s),
transmit loads to the fibres, protect the fibres from environmental degradation and provide
shear resistance between plies. There are two types of carbon/epoxy composite plies utilised in
the analysis depending on the model type, they are defined, with their thickness included, in
table 4.2 below.
1
2
Ply type
Woven fibres (weave fabric)
Unidirectional fibres (tape)
Thickness (mm)
0.280
0.184
59
Property
E11 (GPa)
E22 (GPa)
G12 (GPa)
G13 (GPa)
G23 (GPa)
v12
Weave Fabric
Composite Ply
61
61
4.9
4.9
3
0.05
Unidirectional Tape
Composite Ply
131
9.75
4.65
4.65
3
0.3
Properties
XT
XC
YT
YC
S
Weave Fabric
Composite
MPa
460
435
460
435
155
60
Unidirectional Fibre
Composite
MPa
2220
1300
60
240
108
61
underlined define the weave ply while the remaining angles not underlined define the tape ply.
The subscript number relates to the multiple of the number of plies defined within the bracket
pair. Finally, subscript s implies that the configuration is symmetric.
It is worth noting at this point that all laminates shown are symmetric which implies a reduction
in the number of independent elastic constants from 21 which characterises anisotropic
materials to 13 independent constants which is termed as a monoclinic laminate material. From
viewing the configurations of the laminates in sub-models A (weave + tape) for both models, it
is evident that no 90o plies have been defined in the configuration. Mechanical resistance in the
transversal direction is gained through inclusion of the weave ply as it gives equal mechanical
resistance in the orthogonal angles so by only defining the 0o, it is implied that equal resistance
is attained in its orthogonal angle, which in this instance is 90o. Angled plies of +45o and -45o are
defined in equal quantities so as to maintain equivalent proportions of each angle in the
laminate.
On a final observation of the configurations, it is worth highlighting that only the unidirectional
ply is orientated at +/-45o in all laminates, this is due to a fabrication constraint imposed on the
weave ply. During lamination, the posts substantial length of 7 m causes an issue of ply join-up.
At an angle of 45o, a number of plies are required to be laid-up together at their edges in order
to completely cover the laminate. Plies cannot be jointly laid-up together end to end through
the direction of the fibre as this creates a discontinuity and stress are not transferred through
the entire layer. For this particular reason, the weave ply cannot be jointly laid-up as one of its
orthogonal axes of mechanical resistance would be rendered discontinuous. See Section 5.2 on
the lay-up process for additional information.
Model 1
A
Weave + Tape
Tape
Substructure
Plate
Beam
Plate
Beam
Ply Configuration
[(0/0)2/(0/45/0/-45)3/(0)2]s
[(0/0)2/(0/45/0/-45)4/(0/0)2]s
[(0/45/90/-45/(0)4)2]s
[(0/45/90/-45/(0)4/(0/45/90/-45)2/(0)6]s
Weave + Tape
Tape
Substructure
Plate
Beam
Plate
Beam
Ply Configuration
[(0/0)4/(0/45/0/-45)3]s
[(0/0)4/(0/45/0/-45)4/(0/0)2]s
[(0/45/90/-45/(04))2/(0)2]s
[(0/45/90/-45/(04))2/(0/45/90/-45)2/(0)8]s
62
Tables 4.7 and 4.8 presents a summarised view of the design considerations for the laminate
configurations including overall percentage thickness of both the plate and beam components
and percentage of fibres orientated off-axis, or at +/-45o. In the cases of combined weave and
tape laminates, the percentage off-axis laminates is not considered by the number of plies but
rather by the thickness of such orientated fibres. Other data presented include the number of
plies and thickness for each laminate.
The objective of the laminate configuration design is to maintain equivalent proportions (%
thickness and % +/-45o) between both sub-models (A and B) for each model (1 and 2) while also
maintaining equivalent magnitudes of the critical design control, which in this project is the
maximum displacement permitted. While the variation in the quantity of plies/thickness
between sub-models A and B is a result of the different mechanical resistance provoked by
distinct ply combinations, proportionally both sub-models are maintained approximately the
same. In all models, the overall percentage thickness of the plate and the beam are
approximately 42-43% and 57-58%, respectively. A slight difference in the proportion of +/-45o
occurs between both models but is maintained relatively equal for the sub-models A and B. The
slight difference in proportions arises due to the requirement of maintaining +45o and -45o at
equal quantities while the total number of plies in the laminate varies between both models.
Model 1
A
Weave + Tape
Tape
Substructure
Plate
Beam
Plate
Beam
No. of
plies
36
48
32
44
Thickness
(mm)
8.352
11.136
5.888
8.096
Overall %
Thickness
43
57
42
58
% +/-45o
27
26.4
25
27.3
Model 2
A
Weave + Tape
Tape
Substructure
Plate
Beam
Plate
Beam
No. of
plies
40
56
36
48
Thickness
(mm)
9.28
12.992
6.624
8.832
Overall %
Thickness
42
58
43
57
% +/-45o
23.8
22.6
22.2
25
63
A number of elemental key options settings utilised in this model include applying a full
integration analysis (KEYOPT(3) = 2). This option is recommended as it is highly accurate (even
with coarse meshes) in that its application is well suited to cantilever-type problems which are
dominated by in-plane bending. The second key option setting utilised for this element stores
64
data for top and bottom for all layers in the laminate (KEYOPT(8) = 1). This permits the user to
obtain more accurate interlaminar results (e.g. stresses and strains) caused by different ply
types and orientations. Data for a specific layer is attained by calling on that layer through the
LAYER command. In relation to this project and briefly outlined in its objectives, this key option
setting enables the author to apply any type of desired failure criteria created in macro-style
parametric language (APDL), contained in the ANSYS program. This process is described in
detail later in Section 4.4. A typical laminate section with distinct interlaminar stress results (x)
is shown in figure 4.6.
Laminate Thickness
65
Figure 4.7: Post segment showing separated planar areas for lamination
The local coordinate system (LOCAL) defines the ply orientation of the laminate in question
with its relevant capabilities shown in figure 4.9. The origin of the local coordinate system is
defined by three points on the global Cartesian coordinate system and its orientation through
rotation of Euler angles. This permits each area to orientate the direction of the composite
fibres independently. Each area is meshed individually or collectively depending on whether or
not two or more areas share (1) the same configuration and (2) orientation. Before meshing of
the area, elemental attributes need to be assigned to that area and are done so by the AATT
command. From here the section information is associated with a section ID number through
the SECTYPE command which also defines the type of element (SHELL) and the name to be
given to that particular section, e.g. the first constructed area is AREA1. The geometry data
describing this section type is defined by the SECDATA command. It describes the lamina
configuration of the section by the thickness of each shell layer, material type and the angle (in
degrees) of the already-defined local coordinate system of the layer element.
A simplified sketch of the cross section of the structure is shown below figure 4.8. In total, there
are 8 planar areas that make up the outer boundary of the structure which highlighted in red in
the figure below. Their dimensions are equivalent to those of the original steel post, thereby
maintaining the same profile as has been outlined as a project constraint. From this, lamination
of all areas occurs from the areas surface towards the centroide of the cross section. As a
result, each sections defined shell layers or configuration originates on the plane marked in red
and advances or stacks up in accordance to the sections own local coordinate system. This
method of lay-up is achieved by offsetting each section to the origin of their local z-axis through
the SECOFFSET command. Note that the SECOFFSET command is determined by the node
orientation of the element (i, j, k, l).
66
Figure 4.8: Section of structure with individual sections according to lamina lay-up direction
Orientation issues arise when attempting to represent laminates in a 3-D model. An example of
this is incurred with the orientation of a single layer about the UPN beam section. As described
in Section 5.5.2 of fabrication, the plies of the beam are laminated onto and around a suitable
mould profile. Three planar surface areas of the mould are to be covered by the layer. Figure
4.9 shows the layer and its fibre orientation before and after lamination. Before lamination and
in the first image, one local coordinate system defines the complet layer. In order to maintain
fibre orientation throughout lamination of a single layer around the three planar surfaces of the
mould the layer needs to be subdived into same number of local axes as the amount of
distinctly orientated planar axes, and in this case are three axes which are shown in the second
image of figure 4.9. Each of the three local coordinate systems are different and are defined as
so through the LOCAL command. While attention is required in defining the local coordinates,
the ply configuration or stacking sequence for the three laminates remain the same.
A complication arises in the stacking sequence due to the simplification of the unique surface
defined earlier between the plate area and beam flange adhered together. Taking the example
in bottom figure again, the top local coordinate system (x1, y1, z1) equivalent to the coordinate
system for the combined top flange/plate area defines this laminate configuration adequately.
However, the combined bottom flange/plate area becomes problematic in that the flange
configuration is adequately defined by the local coordinate system (x3, y3, z3) while the plate
configuration is not. The orientation of the fibres in top and bottom areas of the third image
describes issue. According to both sets of local coordinates in these areas (1 and 3), if an
abitrary layer is defined the same for both plate areas the fibre orientation would be opposite
to each other on the global axes as clearly evident in the final image of figure 7. To overcome
this event, the plate stacking of off-axis plies must be altered by 90o. For example, if a layer in
both the top and bottom plate is orientated 45o on the global axes the top layer, with local
coordinate system (x1, y1, z1), is defined at 45o while the bottom layer is defined as -45o in
relation to its appropriate local coordinate system (x3, y3, z3).
67
In relation to meshing of the structure, both U-profile beams are regular in shape and their
areas that make up the sub-structure are mapped meshed (forming straight-sided elements)
with a specific number of divisions made on the selected lines. In relation plate structures in
Model 2 where holes are incorporated in the plates, meshing requirements are more
specialised. As shown previously in figure 4.7 and below in figure 4.8, the plate components are
divided into 3 areas: 2 of which coincide with the area of the beam flange (blue), and the third
is the remaining area between the two beam flanges (green). This third area shares the same
line number with the two bordering flange/plate areas and thus is divided into the same
number of elements in its longitude so that boundary nodes between the 3 plate areas
68
coincide. The centre section however, is free meshed so as to incorporate the holes in the
section. Figure 4.10 shows a plane view of a segment of the plate meshed in accordance to the
three described areas.
Holes in plate
Coinciding
Nodes
69
Tsai-Wu
Maximum Stress
Puck
These three criteria were chosen for a number of reasons. Firstly the three criteria represent a
number of fundamental concepts and assumptions that have been associated with the
evolution and development of composite failure criteria over the past 50 years. As previously
mentioned, they are classified principally into two categories: criteria that distinguish between
70
fibre failure and matrix failure (direct failure mode) and those that do not (non-direct). The
failure criteria directly associated with failure modes can be sub
sub-divided
divided further into two
categories: interactive and non
non-interactive
interactive limit criteria. The difference between these two
types is the existence (or non-existence)
non
of association between in-plane
plane stress components
and the influence of the transversal stress component for a particular failure mode. Figure 4.11
describes the formation of the discussed categories and given are the criteria associated with
each type.
Non-Direct
(Tsai-Wu)
Failure Mode
Non-Interactive
Interactive
(Criterion)
(Max. Stress)
Direct
Interactive
(Puck)
The second for choosing the above criteria is due to the presence of two types of composite
plies utilised in the design of the structure,
struc
namely woven fibre (weave)) and unidirectional fibre
(tape)
pe) composites. Issues arise with determining an appropriate failure criterion for each of the
two ply types. For instance, the direct-interactive
direct interactive criterion of Puck was originally developed for
the application to unidirectional composites and not for woven composites. Accuracy of results
may therefore be called into question with the application of this criterion for the analysis of
the mixed ply laminates of Models 1B and 2B. Of the three presented, the criterion most suited
for the application to the wove
woven fibre (in-plane stress) is the non-direct,
direct, fully interactive
inter
criterion of Tsai-Wu [13].
The final reason for choosing these particular criteria is down to the discretion of the author.
The high accuracy of the Puck criterion in the World Wide Exercise and
and the extensive use of the
Maximum Stress criterion in FEM commercial programs and industry contributed to these
criteria inclusion.
In this post-processing
processing analysis of the project, the above criteria are executed by two different
methods. They include analysis
alysis through
71
Of the three failure criteria to be applied in the analysis only the Maximum Stress and Tsai-Wu
criteria are readily available in the ANSYS FEM program. As a result, the third criterion (Puck) is
written as a macro file in APDL (ANSYS Parametric Design Language) and is executed in the
post-processing stage of the analysis. As stated previously, this parametric script coding is
described in detail in Section 4.4. For validity purposes of the macro-written criteria, the
Maximum Stress and Tsai-Wu criteria are also written in APDL and compared with failure
results of the predefined equivalent criteria in ANSYS.
All failure criteria results are presented as failure index values where a value greater or equal to
one signifies failure.
4.3.1 Tsai-Wu
The theory behind the Tsai-Wu criterion (1971) is a simplification of the Tsai-Hill criterion for
generalized failure theory of anisotropic materials. There are two forms of the Tsai-Wu failure
criterion presented in the predefined ANSYS failure analysis [8], [5]. The first form of the
criterion is the strength index or TWSI which expressed as the following.
IF = A+ B
(4.1)
The second form presented is the inverse of the strength ratio (TWSR) given as
1 B
IT =
=
+
R 2A
2
1
B
+
A
2A
(4.2)
( 1 )2
XT XC
( 2 )2 + ( 12 )2
YT YC
S2
+ c1
1 2
X T X C YT YC
(4.3)
where c1 is a coupling coefficient of the Tsai-Wu theory which by default is taken to be -1. The B
term is defined as
1
1
1
1
B = 2 2 1 + 2 2 2
XT XC
YT YC
(4.4)
72
(11 0)
(4.5a)
(22 0)
(4.6b)
12 < S
(4.6c)
11 < X C
(11 < 0)
(4.7a)
22 < YC
(22 < 0)
(4.7b)
Where one or more of the inequalities are not met fracture occurs in the material according to
the mechanism associated to that equation in which the inequality has not been met [5].
4.3.3 Puck
The Puck failure criterion is one type of criterion that is associated with failure modes. The
criterion distinguishes between fibre failure (FF) and inter-fibre failure (IFF). In relation IFF or
failure of the matrix, the in-plane failure parallel to the fibres is governed by the three stress
vector components associated with that plane which include the normal stress acting on the
plane, and two planar stresses, one acting parallel and the other perpendicular to the fibre. The
criterion proposes that the two shear stresses always promote fracture, while the normal stress
only promotes fracture if it is in a traction state and has the adverse effect in compression, i.e.
it impedes fracture [12].
IFF contains three distinct modes of failure: mode A is when perpendicular transversal cracks
appear in the ply under transversal tensile stress with or without in-plane shear stress; Mode B
also implies transversal cracks, but are a caused by in-plane shear stress with small transverse
compression stress; and finally Mode C denotes oblique cracks onset (typically of angle 53o in
73
carbon/epoxy composites) when the material is under significant transversal compression. The
FF yields one failure index which assumes that fibre failure only depends on longitudinal
tension [14]. It is defined as
11 < XT
11 < X C
(11 0)
(4.8a)
(11 < 0)
(4.8b)
The three failure modes in the IFF imply that it yields three separate index failures. The
particular IFF mode to be activated for analysis depends on the stress state of the lamina in
question. For instance, mode A is activated if the transversal stress is positive. It is expressed as
Y
12
+ 1 p12T T 2 + p12T 2 = 1
S YT
S
S
2
(22 0)
(4.9)
where p12 T is the slope of the failure curve for 22 0 at the point 22 = 0, also known as a
fitting parameter. Without experimental values, the parameter is assumed to be 0.3 which is
representative of carbon based composites. In relation to negative transverse stress 22 < 0,
either Mode B or Mode C is activated, depending on the relationship between in-plane shear
stress and transversal stress. The selection of either mode is defined by the limit of the relation
YA/SA where
YA =
S
YC
1
1 + 2 p12 C
2 p12 C
S
S A = S 1 + 2 p2 C
(4.10)
(4.11)
and
p2C = p12C
YA
SA
(4.12)
where p12 C is another fitting parameter but which corresponds to the interlaminar shear
strength S. A value of 0.2 is selected to represent this parameter [14]. The failure index and its
limitation YA/SA for Mode B are defined as
1
122 + p12C 2 2 + p12C 2 = 1
22 < 0
2
Y
A
12 S A
74
(4.13)
12
YC
+ 2
2 2 1 + p 2 C S YC
=1
22 < 0
2
Y
A
12 S A
(4.14)
75
areas) for time-reduced, localised failure analysis. The failure analysis is carried out by using
vector functions and *DO loops of nodal data for each layer of the laminate. Figure 4.12 shows
the manner in which the nodal results are stored in the matrix. Extrapolation of elemental
results rather than nodal results is more facilitated to this type of written parametric analysis
due to the fact that elements of the selected area would have maintained its consecutive
numeric order during merging while the nodal numeric order of the selected area(s) can vary
due to the merging of the boundary nodes at adjoining areas. However, initial APDL macro
analysis which utilised elemental results was deemed to be outside a tolerable difference for
comparison purposes with those same criteria results yielded in the predefined ANSYS failure
analysis which uses nodal data. To avoid such differences, nodal result data would have to be
obtained and utilised in the failure criteria written in APDL. The following is the process
developed by the author to obtain nodal result data from the analysis, store it and call it for the
APDL macro.
As a result of the inconsecutive numerical order of the nodes in the selected areas caused by
merging, the required nodal result data is obtained and coherently recalled for the failure
analysis through the following steps. Information from the model is retrieved through the *GET
command. *VGET has a similar functionality, but it act upon an array rather than a single
parameter which is a more rapid form of obtaining information than by looping the single
parameter command. Because *VGET acts upon a vector, *VMASK typically is used in
conjunction with *VGET. *VMASK is a masking array which tells ANSYS to perform vector
operations only on certain items in the array, which in this case are the nodes of the selected
areas. The desired areas are selected by the ASEL command and the nodes associated with
these areas (NSLA) are then selected. The objective of this part of the macro is to store the
required nodal results for all layers in a matrix array. The matrix dimensions are determined by
retrieving the number of nodes in the structure (rows) and defining the number of layers in the
equivalent laminate in the selected area(s) (columns). An extra column is defined in the matrix
which is occupied by the node number for that particular row. After the masking vector has
been applied at the stage of retrieving the nodal results, the node number will be retrieved also
thereby insuring the data for each row of the matrix are defined by their appropriate nodal
number. This coherent method of defining the rows of data by their equivalent node number is
vital for retrieving information such as the location most critical in the particular failure
analysis. Figure 4.12 shows the manner in which the nodal results are stored and their
equivalent number.
76
Figure 4.12: Matrix layout for nodal stress results of each layer
Below is a segment of the macro that defines the number of layers (36) in the laminate, the
maximum amount of nodes in the structure, the relative areas to be selected and their
associated nodes.
*SET,NUMPLY,36
*SET,NUMCOL,37
!36 LAMINAS
*GET,NMAX,NODE,,NUM,MAX
!SELECCIONAR AREAS: PLATES
ASEL,S,AREA,,12
ASEL,A,AREA,,16
NSLA,S,1
At this point, the nodal stresses are to be retrieved from the models results database. This is
done, as previously mentioned through the *VGET command. To use *VGET with masking, the
following steps need to be performed:
1.
2.
3.
4.
5.
Define the masking vector by its dimensions with the *DIM command
Define the regular array to hold results of interest with the *DIM command
Fill the masking vector with the selected nodes *VGET,,NODE,1,NSEL
Activate the masking array with *VMASK
Fill the regular array with the nodes selected using *VGET
The following is a continuation of the macro above where the previously outlined steps are
executed to obtain the nodal stress data in the x-direction of the models global axes. The script
below combines vector function and *DO loop capabilities where the function commands
*VMASK and *VGET are repeated for each layer of the laminate through the LAYER command
thereby only obtaining data for each layer of the selected areas and leaving the rest a null. This
segment of the macro is repeated for obtaining stresses in the y and xy directions.
77
*DEL,SXMASK
*DEL,SXARRAY
*DIM,SXMASK,ARRAY,NMAX
*DIM,SXARRAY,ARRAY,NMAX,NUMCOL
*VFILL,SXARRAY(1,1),RAMP,1,1
*VGET,SXMASK(1),NODE,1,NSEL
*DO,j,1,NUMPLY,1
LAYER,j
*VMASK,SXMASK(1)
*VGET,SXARRAY(1,j+1),NODE,1,S,X
*ENDDO
The results are directed to a file by the /OUTPUT command. The masking command is once
again applied here so as to preserve the file for only information of the selected nodes.
*MWRITE writes the obtained results to the file in a formatted sequence. Below is the segment
which writes the stresses (x-direction) into a file. The /NOPR command suppresses the
expanded interpreted input data listing. This command reduces the file to the leave only the
matrix of results which is directly applicable to the failure criterion analysis part of the macro.
The precision of results is handled by the FORTRAN format F contained in the brackets.
/NOPR
/OUTPUT,SXARRAY,FILE
*VMASK,SXMASK(1)
*MWRITE,SXARRAY(1,1),,,,JIK
(37F12.6)
/OUTPUT
The failure criteria analysis initiated by recalling the result data from the previously written files
through the *VREAD command. The array is defined by its dimensions with the *DIM
command. The process of reading a specific results file (x) is described below.
*DEL,SX
*DIM,SX,ARRAY,NODOS,NUMCOL
*VREAD,SX(1,1),SXARRAY,FILE,,JIK,NUMCOL,NODOS
(37F12.6)
Before the criterion can be applied, the stresses must be rotated to their principal directions. In
order to rotate the stresses from the orientated global coordinates to the principal axis (1,2)
the angle (in degrees) of each layer needs to be defined in the macro by using the *VFILL
command. The angles are then converted into radians. This is realised by using an APDL vector
operation command *VOPER where the vector of angles is multiplied by its radian equivalent.
Again, this vector operation reduces the calculation time that would be incurred if a do-loop
78
process was to be used instead. The transformation of orientated stresses to principal stresses
is performed by the following matrix expression.
2
1 cos
2 = sin 2
12 cos .sin
sin 2
cos2
cos .sin
x
1
2 cos .sin x
2 = [T ]. y
2 cos .sin . y
;
2
2
xy
12
cos sin xy
(4.15)
where T is the matrix of transformation. Resolving the transformation above, the stresses in the
principal axis in plane stress are.
(4.16)
where sine and cosine of each angle is carried out by another vector operation (*VFUN). These
three equations above are incorporated in the macro in the form of a do-loop process which
incorporates the off-axis stresses for each node (i) of all layers (j) at their respective angle of
orientation (j).
At this stage of the macro the principal stresses have been calculated for each layer of each
element and the failure criterion can now be applied. Below is shown a segment of the
Maximum Stress criterion in which it determines if failure occurs due to breakage of the fibre
caused by traction. As this part of the criterion determines the possibility of failure caused by
traction, it therefore implies that nodal stress results need to be separated in terms of their
state (in compression/tension) before they can be applied to the criterion. The separation is
achieved by an if-else statement which simply defines the stress state as being negative or
positive and is then applied accordingly to the criterion in question. For example, in the
criterion of failure of the fibre in tension the if-else statement utilises the stresses greater than
zero i.e. tensile stresses, and set the compressive stresses as null.
All of the three criteria applied in the APDL macro have been written so as to locate whether
failure occurs in each of the mechanisms and in the event of failure occurring in two or more
mechanisms, the script would locate in which of the mechanisms failure would occur initially.
After determining the primary mechanism of failure the script then relays to the user the most
critical node and layer within the laminate.
79
80
4.5 Results
Tabulated below are the relevant results of the behavioural response for both models. Table
4.9 and 4.10 present the nodal stress results in global coordinates for Model 1 and Model 2,
respectively. Nodal stress results include stress in the directions of the global axes (x, y, z)
and shear stresses with respect to the two planes in which laminates of the structure lie (xy,
xz). Both tables are subdivided by the type of applied load case with also results shown for
both laminate composition sub-models A (weave + tape) and B (tape). Both traction (T) and
compression (C) stresses are tabulated for each stress component. The maximum displacement
experienced in each load case is tabulated with their respective displacement direction.
Model 1
xy
xz
MPa
MPa
MPa
MPa
MPa
A (Weave +Tape)
B (Tape)
5.952 5.922
13.279 13.210
A (Weave +Tape)
B (Tape)
24.267 24.229
52.39 52.678
Displacement
mm
0.074
0.103
0.075
0.102
6.432 (Uy)
6.500 (Uy)
0.002
0.007
0.895
0.976
54.470 (Uz)
53.439 (Uz)
Table 4.9: Model 1 (no holes) nodal stress and displacement results (global coordinates) for Case 1 and 2
Model 2
T
A (Weave +Tape)
B (Tape)
A (Weave +Tape)
B (Tape)
xy
xz
MPa
MPa
MPa
MPa
MPa
Displacement
mm
0.109
0.118
0.104
0.112
5.482 (Uy)
5.881 (Uy)
0.000
0.001
1.313
1.185
57.368 (Uz)
58.202 (Uz)
Table 4.10: Model 2 (holes) nodal stress and displacement results (global coordinates) for Case 1 and 2
The laminate configurations of tables 4.5 and 4.6 and their design formats of tables 4.7 and 4.8
are directly associated with the results of the above tables. In terms of the design of the
laminates of each model, the principal constraint imposed was the magnitude of maximum
deflection permitted for each type of structure. These constraint values were taken from the
analysis of the original steel post where equivalent loading conditions were applied. As a result,
displacement correlation is the means by which the composite models are designed in this
81
project. The displacements shown in the above table of results confirm this type of correlation.
Values of displacement for the set of sub-models A and B, in their respective case types, are
approximately equal with the largest variation between a sub-model set being just over 1 mm.
However, while the displacements of sub-model sets are equal, differences emerge in terms of
their maximum stresses experienced. With regards to the stress maximums along the direction
of the beam caused by bending (x), each sub-model set demonstrate differences in magnitude
of approximately 100% between the A and B laminate types. Such variations in magnitude are
caused by variations in
1. Laminate stiffness
2. Sectional inertia
Laminate sub-model A for each case is composed of plies of woven and unidirectional fibres
whereas sub-model B is entirely compose of plies of unidirectional fibres. The differences in the
elastic constants of both types of plies can be appreciated by recalling table 3. It must be noted
that the elastic modulus is not the same as stiffness. It is instead a property of the constituent
material whereas stiffness is a property of the laminate structure. That is to say, the modulus is
an intensive property of the material whereas stiffness is an extensive property which is
dependent on the material, ply orientation, shape and boundary conditions of the structure. As
a consequence, the higher stiffness of the laminate B in the direction of the beam length
induces an increased load transfer and higher stress magnitude than that of the mixed ply
laminate of A. In order to maintain a comparable maximum displacement magnitude in the
lower stiffness laminate A to that of B, two approaches need to be performed: the first is by
adding additional woven plies to the laminate thereby increasing the inertia of the section to a
certain extent and the second is orientating unidirectional fibres along the direction of the
beams length.
Figures 4.14 and 4.15 show the stress variation in the x-direction x around post section of
Model 1 for load Case 1 and load Case 2, respectively. Results for both sub-models A and B are
together for each load case. The section shown is equal to the section 40 mm above the fixed
end of the beam. Figure 4.13 depicts a sketch of the equivalent section with the dimensional
path related to the graphical figures defined also.
82
The figure below represents the loading perpendicular to the rail line in which stress maximums
in the x-direction are experienced over the thickness segments of the post structure with one
side in a state of tension and the opposing side of equal magnitude in compression. Between
these, the stress varies approximately linearly along the width of the structure with stresses
equal to zero occurring for both laminate A and B at the centre of the posts width.
Stress Variation in Post Section for Case 1
15
10
Stress (MPa)
-5
-10
1A:Weave & Tape
1B: Tape
-15
200
400
600
Length (mm)
800
1000
Figure 4.14: Stress Variation in post section at 40 mm above constraint boundary conditions
(Case 1)
83
Figure 4.15 represents the loading parallel to the rail line in which stress maximums in the xdirection are experienced over the width segments of the post structure. For both sides of the
structure which are in opposing states of stress, a maximum value is found for both at the
corners of the structure which reduce towards the centre of the width. This variation in stress
across the width represents the different load carrying capabilities of the structural elements
with the beam flange sections being the greatest of all elements. It is worth recalling that the
flange elements were determined as strong design drivers in the project outset.
Stress Variation in Post Section for Case 2
50
40
30
Stress (MPa)
20
10
0
-10
-20
-30
1A:Weave & Tape
1B: Tape
-40
-50
200
400
600
Length (mm)
800
1000
Figure 4.15: Stress Variation in post section at 40 mm above fixed end (Case 2)
Figure 4.16 shows the stress distribution (x) over the bottom region of the post Model 1A
respectively, under loading conditions of Case 2.
84
Large variations are presented between the maximum stresses between Model 1 and Model 2
for load Case 2. The cause for such differences is the presence of holes in the plate elements of
the structure. These cut-outs provoke a stress concentration factor at the edge of the laminate.
As describe in Section 2, the stress concentration is caused by the combination of the fibre
direction and the load direction in which the material is subjected. The complexity of stress
analysis around the holes edge increases with the introduction of additional plies in the
laminate with various orientations. The stress concentrations occurring in such regions are
areas of concern and are therefore analysed using various strength criteria for failure analysis
as outlined previously in Section 4.3. Figure 4.18 shows the variation of nodal stress in the xdirection around the hole edge for laminate sub-models A and B. The sketch in figure 4.17
depicts where the starting point and direction of the graphical stress distribution in figure 4.18.
Figure 4.17: Sketch of circumferential path for stress variation around hole
Stress (sigmax ) Distribution Around Bottom Plate Hole
90
2A) Weave + Tape
2B) Tape
80
70
Stress (MPa)
60
50
40
30
20
10
0
0
100
200
300
400
500
600
Circumference (mm)
700
800
Figure 4.18: Stress variation in laminate around bottom hole for Model 2A and 2B, load case 2
85
Figure 4.19 shows the stress distribution (x) over the bottom region of the post Model 2A
respectively, under loading conditions of Case 2.
Tables 4.11 to 4.16 present the failure criteria results for both the predefined ANSYS failure
criteria and for the APDL macro failure criteria. Criteria include Tsai-Wu and Maximum Stress
for the predefined FEM program, and Tsai-Wu, Maximum Stress and Puck for the macro. Only
the most critical failure criteria results of the two models are presented below so as to reduce
the quantity of result information displayed. These critical failure values occur in Model 2
where the cut-outs/holes in the plate components creates stress concentrations in the laminate
bordering these regions as can be seen in the stress distribution of figure 4.17.
The failure criteria results are presented as failure index values where a value of one or above
signifies failure in one of the laminates layers. Both sets of results are obtained from nodal
stress data of each layer. Regions of high stress concentrations are analysed through one of the
defined failure criteria firstly by the APDL macro and secondly, by the predefined criteria in
ANSYS. By initially carrying out the failure analysis by the criteria written in the macro,
additional information such as the failure mode type and particular layer of failure are
attainable. The equivalent criterion analysis is then carried out in the ANSYS program in which
the user can define the desired layer for analysis, i.e. the layer at which failure has occurred in
the macro analysis. Failure index values and their node location are obtained and compared to
those of the macro.
Table 4.11 and 4.12 both present the most critical failure index values for the Tsai-Wu and
Maximum Stress, respectively, for laminate sub-model A analysed by both the ANSYS program
and the macro, with their % difference calculated and presented in the final column. Also
included are the node and layer number at which the critical index value occurs. Additional
86
information can be obtained from the APDL macro for the Maximum Stress criterion that
includes the mode in which the critical index value occurs. Table 4.13 presents the failure
analysis results for the Puck criterion defined in the APDL macro. Information including critical
node and layer number are obtained from the analysis. The macro also outputs the failure type
(FF or IFF) and in the case of the IFF, its mode number (1, 2, 3) for the most critical index value.
Tsai-Wu: Model 2A
FPF Index Value
Node
Layer
ANSYS
0.13800
15322
34
Macro (APDL)
0.13816
15322
34
% Difference
0.1%
Table 4.11: Tsai-Wu Index failures for Model 2A in FEM (ANSYS) and macro program (APDL)
ANSYS
0.12216
15322
34
Macro (APDL)
0.12223 (Yt)
15322
34
% Difference
0.05%
Table 4.12: Maximum Stress Index failures for Model 2A in FEM (ANSYS) and macro program (APDL)
Puck: Model 2A
FPF Index Value
Failure Mode
Node
Layer
Macro (APDL)
0.12303
1 (IFF)
15322
34
Table 4.13: Puck Index failure for Model 2A in macro program (APDL)
Tables 4.14 to 4.15 below present the equivalent result format as for the set of tables above,
however this time the results presented are for the model composed of laminate sub-model B.
Again, the Tsai-Wu and Maximum Stress criteria are given in the first two tables for both the
ANSYS and APDL analysis. The third table contains the results of the Puck criterion for the APDL
analysis only.
Tsai-Wu: Model 2B
FPF Index Value
Node
Layer
ANSYS
0.11787
12728
1
Macro (APDL)
0.11770
12728
1
% Difference
0.14%
Table 4.14: Tsai-Wu Index failures for Model 2B in FEM (ANSYS) and macro program (APDL)
ANSYS
0.10917
12728
1
Macro (APDL)
0.10902
12728
1
87
% Difference
0.13%
Table 4.15: Maximum Stress Index failures for Model 2B in FEM (ANSYS) and macro program (APDL)
Puck: Model 2B
FPF Index Value
Mode
Node
Layer
Macro (APDL)
0.10902
FF
12728
1
Table 4.16: Puck Index failure for Model 1B in macro program (APDL)
All of the above criteria results demonstrate that failure does not occur in any of the structures
(i.e. index failure < 1). However, it is worth noting that the more critical index failures occur for
the sub-model A (weave + tape) even though the stress concentrations around the holes shown
in figure 4.19 are greater in sub-model B (tape).
The index result values for the predefined ANSYS criteria and APDL macro criteria are identical
for each criterion with only minute differences between the results occurring due to variation
of precisions.
88
N y = N xy = M y = M xy = 0
Ky 0
(4.17)
89
The Classical Laminate Theory (CLT) of Chapter Two is applied in this section to verify the
validity of the numeric model of the FEM. Comparable response results such as the axial
stiffness, bending stiffness and stresses
stresses are calculated using the process shown as a flowchart in
figure 4.21. The process includes the CLT described in Chapter Two up to determination of the
extension, coupling and bending stiffness. From that point, the narrow beam conditions are
implemented
d and the responses are subsequently determined in the order shown in the
flowchart. The CLT validation is written using MATLAB parametric language.
Material Data +
Structural
Configuration
Compliance
Matrix [S]
Reduced
Stiffness [Q]
Transformed
Stiffness [Q]
Inverse
Compliance
Matrix
Narrow Beam
Theory
Extension [A]
Coupling [B]
Bending [D]
X, Y, XY
= Q.
Figure 4.21:: Classical Laminate Theory (CLT) process for narrow beam
(4.18)
o a
=
k bT
(4.19)
or as its inverse
b N
.
d M
90
Applying the initial conditions of the narrow beam, where Ny= Nxy= My= Mxy = 0, the relation is
reduced to the following equations
xo = a11 N x + b11 M x
(4.20)
k x = b11 N x + d 11 M x
xo a11 b11 N x
=
.
k x b11 d11 M x
(4.21)
M x b11 d11
xo
.
k x
=>
N x A'1
=
M x B '1
B'1 xo
.
D1 k x
(4.22)
where
A'1 =
B'1 =
1
a11
b112
d11
1
a d
b11 11 11
b11
D '1 =
(4.23)
(4.24)
1
d11
b112
a11
(4.25)
The load-deformation relation of equation (4.22) for the plate and flange laminates in the xyplane (i = 1, 2, 3, 4) can be expressed by the following two equations.
(4.26)
k x ,5 = 0
91
(4.27)
which implies that the force and moment in the x-direction for the web are reduced to
(4.28)
Figure 4.23: Sketch of reconfigured section about y-axis which structural elements numbered
92
Centroide (ZC)
In relation to figure 4.23, the centroide of the structure is defined as the average location of the
forces acting on each part of the cross-section. The y-coordinate is located at the bottom of the
section and the z-coordinate, due to symmetry, is located at the centre of the section passing
through the centre of the web thickness. The net force acting on the centroide is given by
5
N x z c = bi N x , i z i
(4.29)
i =1
where bi is the breadth of the elements, zi is the distance from the y-axis to the centre of the
element, and N x is
5
N x = bi N x ,i
(4.30)
i =1
zc =
i =1
bi A'1,i
o
x ,i
+ B'1,i k x ,i
(4.31)
As the strain along the x-axis of the structure is the same for all elements, it can be deduced
that
xo,i = xc
(4.32)
where xc is the strain at the centroide in the x-direction. As well as that, with only load
application at the centroide of the section along the x-axis, no radius of curvature exists.
k x ,i = 0
(4.33)
Which implies the expression for determining the centroide location is simplified to
5
bi A'1, i zi
i =1
bi A'1, i
zc =
(4.34)
In relation to the composite post sectional detail of the project, it is evident that the
configuration is symmetric about the centre of the web element in the y-direction. This implies
that the centroide is located at this point.
93
N x = E A xc
(4.35)
where E A is the equivalent axial stiffness of the entire section. Substituting the equations
(4.26) and (4.28) into (4.30), the total force is
(4.36)
and recalling the conditions of the axial force in (4.32) and (4.33), the total force can be written
as
N x = bi ( A'1,i ) xc
i =1
(4.37)
E A = bi ( A'1,i )
5
(4.38)
i =1
M x = Dxc k xc
(4.39)
The total moment is made up of moment and force components, in the x-direction, of the five
sectional elements.
4
M x ,i = bi M x ,i + bi N x ,i ziC + M x ,5
i =1
(4.40)
M x ,5 =
94
d5 2
zN
(
)
x ,5
dz
(4.41)
d5 2
Thus
4
M x = bi M x , i + N x ,i ziC +
i =1
d5 2
zN
(d 5 2 )
x ,5
dz
(4.42)
Recalling the constitutive equations in (4.26), and considering that the curvature is equal for all
sectional elements
kx,i = kxc
(4.43)
xc = 0
(4.44)
The moment M x for the flange and plate elements (i = 1, 2, 3, 4) can be expressed as
(4.45)
(4.46)
(4.47)
M x ,5 = A'1,5 k
d5 2
c
x
dz
(4.48)
d5 2
With the curvature k xc common to all elements, the total moment in (4.40) can be expressed as
d5 2
4
2
M x = bi A'1,i z iC + 2 B'1,i z iC + D'1,i + A'1,5 z 2 dz k xc
i =1
d5 2
(4.49)
i =1
2
iC
d5 2
d5 2
dz
(4.50)
95
Knowing the bending stiffness and total moment, the curvature can be found by manipulating
the above equation into the form below.
k xc =
Mx
Dxc
(4.51)
The stresses and strains are calculated using equations in (2.44) and (2.45) of Chapter Two, and
are recalled as
xo,i
x ,i
k x ,i
y ,i = yo,i + z k ,i k y ,i
xyo ,i
xy ,i
k xy ,i
(4.52)
[ i ]k = [Q ]k [ i ]k
(4.53)
The relation of the laminate force and moments to strains and curvatures, which is expressed in
the equations of (4.26), is used to determine the force and moments at a specific laminate. The
bending stiffness conditions of (4.43) and (4.44) are implemented in the equations of (4.26).
The strains and curvatures are calculated by the inverse constitute relation in (4.19). The
transformed stiffness Qij is calculated for each layer of the laminate and subsequently
multiplied by the strain of the corresponding layer. Note that the curvature component of the
strain in (4.52) is dependent on the distance of the particular layer of interest from the
centroide (zk). This process is carried out for each laminate element in the section. Using the
bending stiffness approach, the steps in determining the stresses and strains are summarised as
follows.
1. Calculate the bending stiffness D xc
2. Knowing the moment M x , calculate the curvature k xc from equation (4.51)
3. Applying bending stiffness conditions in (4.43) and (4.44), find the force Nx and
moments Mx and Mxy using equations in (4.26)
4. Determine strains and curvatures from the deformation-load relation in (4.19)
5. Multiply the transformed stiffness matrix Qij of a particular layer by its corresponding
strains and curvatures to determine the stresses in that layer.
96
1
cos 2 ik sin 2 ik +
sin 4 ik
E
22
(4.54)
where i is the laminate element (plate, flange, or web) and k is the particular layer in the
laminate. The equivalent axial Elastic Modulus is calculated by multiplying the transformed
stiffness in (4.54) in each layer by its corresponding layer thickness and dividing by the overall
thickness of the laminate.
5
E xk,i t ik
i =1
t ik
E x ,i =
(4.55)
The total equivalent axial stiffness is gotten by summing the equivalent stiffness of each
element (i = 1, 2, 3, 4, 5).
5
E A = E x ,i Ai
(4.56)
i =1
where Ai is the sectional area. Similarly, the bending stiffness is calculated by summing the
individual sectional elements.
5
E I = E x ,i I i
(4.57)
i =1
As this approach is based on the assumption that the equivalent Elastic Modulus is constant
throughout the laminate, the centroide zi is always located at the mid-plane of each laminate.
The centroide of the entire section is found by the summation of all the laminate stiffnesses
multiplied by their corresponding centroide, and dividing them by the total stiffness equivalent.
5
zc =
i =1
(E
x ,i
Ai )z i
E x ,i Ai
(4.58)
97
At this point, the maximum stresses for a given point in the section are determined. To
incorporate the variation of equivalent stiffness between the sectional elements, the equivalent
width technique is used. From equation (4.55), the equivalent axial Elastic Modulus for the
plate element (sections 1 and 4) is determined to be the stiffer than the beam elements
(sections 2, 3, and 5). That implies that to maintain an equivalent stiffness throughout the
entire section, the geometric sectional properties of the plate elements must be altered by a
factor of n where
n=
E plate
Ebeam
(4.59)
Consider the two sectional drawings in figures 4.23 and 4.24, the first figure shows the original
section with elements of varying stiffness (Ex,plate, Ex,beam) whereas the second figure depicts the
equivalent section in which all elements have the same stiffness value (Ex,beam) and also shows
the necessary geometric transformation to maintain the mechanical properties of the original
section. Note that due to symmetry, the section elements below the plane y-y are considered
to take on the same transformation of their symmetric counterparts above. This also implies
that the centroide zc remains at the centre of the entire section.
98
The inertia Iyy of the entire section is recalculated, implementing this time the transformation in
figure 4.24. The stress at the plate extremity is calculated using the beam bending relationship
multiplied by the transformation factor n.
M . yi
i = n
I
(4.60)
Units
Centroide (zc)
Axial Stiffness (EA)x
Curvature (kc)x
Bending Stiffness (EI)x
mm
N
mm-1
Nmm2
99
Smeared
Approach
70
8.8244e8
5.3755e-6
3.2908e12
% Difference
0.0%
5.1%
7.1%
7.6%
Table 4.17: Behavioural response result comparison for CLT for narrow beams and the smeared approach
As both the entire structural section and each laminate element are symmetric about the yplane, the centroide distances are the same for both the CLT and Smeared Approach analyses.
Variation between both validation models occurs for the axial stiffness, curvature and bending
stiffness. This occurs due to the presence of extension, coupling and bending parameters (Ai,
Bi, Dj) in the CLT model while the Smeared Approach is an analogous method of elastic
equivalents (Ex) and sectional properties (area and moment inertia) used to determine intrinsic
laminate response. As all the laminates are symmetric about one plane (y-plane) they can be
classified as monoclinic, which causes the coupling terms BI to be reduced to zero meaning that
only extension and bending terms remain. The difference between the two models increases if
the ply stacking sequence is altered to an un-symmetric configuration where coupling returns
to the load-deformation relation.
The maximum stresses of the plate element in a state of tension are tabulated below in table
4.18. They include results for the FEM, CLT, and Smeared Approach. Both the FEM and CLT
results are presented at a layer level while the Smeared Approach stress result is of a laminate
level. Both the FEM and CLT results are the maximum tensile stresses experienced and occur in
the top layer (0o) of the top laminate, i.e. the layer furthest from the centroide of the section.
The higher stiffness caused by the ply orientation and its distance from the centroide confirms
that maximum stresses would develop in this zone of the cross-section. The minute difference
between both results (1.24%) would indicate that the approximation made in the FEM model of
the composite post is valid. The large variation between the Smeared Approach and FEM/CLT
models indicates the limitations incurred by assuming equivalent elastic properties for a
laminate as opposed to a layer analysis made in the CLT for Narrow Beams. The Smeared
Approach does not account the for the stress variations between layers and consequently
underestimates the maximum stress incurred in the section.
FEM
CLT for Narrow Beams
Smeared Approach
Location
Top Plate (1) 0o Ply
Top Plate (1) 0o Ply
Top Plate (Laminate)
x (N/mm2)
52.390
53.046
38.637
Table 4.18: Stress result comparison for FEM, CLT for narrow beams, and the smeared approach
100
101
(4.18)
where is the air density, S is the frontal area of the train, VA is the train velocity relative to the
air, and CD is the drag coefficient. The value of CD is influenced by the trains geometric profile
where values of CD for S on the order of 10 m2 and L on the order of 300 m can range from
approximately 1, or less, for highly streamlined trains to 10 to 15 for freight trains. For analysis
based on data correlations, the drag coefficient, CD, is broken down as:
(4.19)
where CDL is the drag coefficient of the leading car or locomotive, CB is the base drag at the
trains tail, T is the friction along the train, which includes the bogies, wheels, interference and
underbelly effects, and lT and lL are the lengths of the total train and the leading car,
respectively, [21].
When the train closes onto the stationary object and in this case the post, it applies a pulse of
pressure on the object. Figure 4.27 shows graphically the pressure history subjected on a
proximate structure due to a passing train and is taken from experimental data in [20]. It is felt
that this figure and the results provided are suitable to be interpreted in the dynamic analysis
of this project. The experimental results were collected by using a prototype high-speed
locomotive namely the JetTrain built by Bombardier. This passenger train was built in an
attempt to make European-style high-speed service more financially appealing to passenger
railways in North America. The validity of the experimental results is due to the similarity in the
trains profile to a number of locomotives/leading cars in service in Spain which include, for
example, Alstom S-100 AVE, Siemens S-103 AVE and Talgo-Bombardier S-130 Alivia. Figure 4.26
shows similarities between the profiles of the Bombardier JetTrain (left) and the Alstom S-100
AVE (right). An additional validation for the use of the experimental results in the numerical
model is the velocity at which the train carried out the tests. The maximum velocity reached by
the Bombardier JetTrain was 130 mph or approximately 210 kph which is in line with the
velocities of the train fleet of the High-Velocity Variable-Width section of the Spanish railway
system and it is in this section which the modelled structure would be designated to perform.
102
In relation to the graph in figure 4.27, above the zero mark represents the positive pressure
exerted on an adjacent structure and below zero represents the negative pressure or vacuum
created. Maximum values on the graph range from approximately 0.06 psig to 0.09 psig or
413.68 Pa and 620.58 Pa, respectively. The concluding information gathered by the reports
allowed the author to make a more conservative representation of the pressure exerted on the
post as train velocities. It is worth highlighting that this proposed post would not be designated
for such rail High-Velocity only corridors, can reach up to 350 kph (Siemens S-103 AVE) but for
the general rail lines (via general) of reduce velocity capacities.
Aerodynamic pressures, which include a Factor of Safety of 2.0, utilised in the numerical
program include a maximum pressure value of 1.2 MPa at the point at which the nose of the
leading car reaches the structure and, in a relatively instantaneous moment, a pulse of vacuum
of 1.4 MPa which has the opposite directional loading effect and occurs when the nose of the
train immediately passes the structure. As a result, the numerical model intends to simulate an
equivalent pressure change applied on the structure. Figure 4.28 is a representation of the load
history inputted into the transient analysis of the numerical model. The pressure history below
also highlights an interesting phenomenon that consists of a number of relatively large pressure
variations contained within the passing of the nose and tail of the train. The figure shows two
pressure spikes sustained in the negative portion of the pressure curve. These variations are
directly related to the discontinuities of the train, i.e. the inter-car gaps of the passing train.
103
1000
Load (N/m2)
500
-500
-1000
-1500
0.5
1.5
Time (sec)
2.5
104
1A
1B
(Hz)
(rad/s)
(Hz)
(rad/s)
6.335
19.617
158.415
39.805
123.257
995.349
8.768
15.355
176.572
55.089
96.478
1109.433
Table 4.19: Modal analysis, first three natural frequencies of post in Model 1
Mode
1
2
3
2A
2B
(Hz)
(rad/s)
(Hz)
(rad/s)
4.249
12.941
45.721
26.694
81.311
287.272
5.151
15.375
64.859
32.364
96.607
407.522
Table 4.20: Modal analysis, first three natural frequencies of post in Model 2
105
(4.20)
where M is the mass matrix, C is the damping matrix, K is the stiffness matrix, u is the nodal
& is the nodal acceleration vector.
displacement vector, u& is the nodal velocity vector, and u&
In this project, the transient analysis is used to determine the time-varying stresses and
displacements in the structure as it responds to a prescribed variable load history. The
experimental load-history curve shown in figure 4.27 is applied in the transient analysis in
ANSYS. Figure 4.28 shows the approximated, more conservative load history converted from
psig to N/m2. The transient analysis involves loads which are functions of time, and thus are
divided into load steps. For each load step, the load and time values are specified with the
option to specify whether the load is applied as a ramp function (KBC,0) or as a step function
106
(KBC,1). The time value is specified after the load is applied via the TIME command. The loadhistory curve is solved by employing time integration of the equations of motion with the
integration time step (ITS) used as the time increment in this process. The size of time
increment step determines the accuracy of the transient solution where the smaller the
increment the more accurate the solution [22]. There are three methods of transient analysis
available which include Full, Reduced and Mode Superposition Method and are implemented
through the TRNOPT command. The one used in this analysis is the Full Method as it permits all
types of non-linearities however it is much more CPU intensive than the other two methods as
the full system matrices are used [23].
In most structures, damping is present in some form. The problem of dissipating the energy in
structures is an important feature in mechanical design in terms of vibration control, noise and
fatigue endurance. The amplitude and frequency of vibration in a structure is controlled by the
applied excitation and the response of the structure to that particular excitation. Damping
becomes important in the structure when the response is close to the natural/resonant
frequency or if the vibration response is short term. Energy dissipation mechanisms are
separated into classes: the first is damping in the structure due to friction at joints. While this
source is usually dominant in metal structures it does not contribute significantly to composite
structures as joints are tended to be kept to a minimum or are adhesively bonded. The second
class and more predominant of the two is inherent material damping [24]. For fibre reinforced
composite materials, material damping is dependent on two mechanism types: at a microscale
level, the energy dissipation is induced by viscoelastic behaviour of the matrix, damping at the
fibre-matrix interface, and damping due to damage. The second type is at the laminate level
where energy dissipation is dependent on the constituent properties, the ply orientation,
interlaminar effects, and stacking sequence [25].
In relation of the finite element model, damping is provided by Rayleigh damping equation
which is represented by the expression
C = M + K
(4.21)
where C is the damping matrix, M is the mass matrix and K is the stiffness matrix. The
parameters and are inputted into the program through the ALPHAD (mass damping) and
BETAD (material damping) commands, respectively. The parameters and are unknowns but
are related to the modal damping ratio . The modal damping ratio is the ratio of actual
damping to the critical damping for a particular mode of vibration. If i is the natural frequency
(as calculated previously in the modal analysis) of mode i, then and satisfy the relation
i =
i
+
2i
2
107
(4.22)
Figure 4.32 below describes the influence of each of the parameters over a range of natural
frequencies. For damping of at low frequencies, plays a more dominant role whereas higher
frequencies are damped more by the parameter and to a lesser extent by -damping.
To specify both and for a given damping ratio, it can be assumed that the sum of the two
damping functions is nearly constant over a range of frequencies. Therefore, given a modal
damping ratio and a frequency range i and j, two simultaneous equations can be solved for
and [8].
=
=
i
+
2i
2
(4.23)
+
2 j
2
In relation to the transient analysis presented in the project, only one model type and material
configuration was analysed. This is predominantly due to the intensive computational demands
required to carry out a full transient analysis on models with a large quantity of elements such
is the case in this project. The structure without holes composed of only laminates with
unidirectional material (1B) was analysed. To specify both and for a given damping ratio,
the first two modal natural frequencies (1 and 2) of the model were taken from table 4.19
and solved in the above simultaneous equations. The damping ratio of carbon fibre
composites is known to approximately 0.01 [26]. The values yielded from the above
simultaneous equations include 0.7 and 0.00132 for the -damping and -damping,
respectively.
108
It must be stressed that the relevant data for damping of a particular composite laminate
structure requires a certain degree of vibration damping experiments at a laminate and possibly
a structural level. The above damping parameter values accumulated are approximations and
their response results obtained in the transient analysis are viewed with caution. A number of
additional transient analyses were carried out with varying and values so as to show the
influence of both types of damping on the structure. Figure 4.33 shows the vibration response
of Model 1B caused by the given excitation in figure 4.28 for different values of the and
parameters. Figure 4.34 shows the stress variation in the x-direction at the bottom of the
structure (node 4553), close to the fixed end of the beam caused by the same load history.
The behaviour responses of the figures below indicate the importance of dynamic response
analysis. The initial two loadings applied to the structure occur in opposite directions, which
simulates a positive-negative pressure condition caused by the passing train. This load pair
induces an initial displacement of 32.14 mm and a subsequent displacement in the opposite
direction of 79.2 mm from its stationary position. The range of displacement of the pair is equal
to 111.34 mm and occurs over a time interval of 0.11 s whereas the maximum displacement for
this model in the static analysis was 53.439 mm. Dynamic ranges of this magnitude over a large
number of cycles can merit the need for a fatigue analysis. As this structure is intended for only
temporary substitution of a fail post structure, its service life duration is relatively low and for
that reason a fatigue analysis is not considered.
In terms of the effect and damping values, the analysis demonstrates that both damping
parameters influence the overall damping of the structure. The vibration response below shows
that in the case of reduced and damping values, the damping is significantly less than that
of the two cases where alternately one of the damping values is reduced. In both these cases,
the damping response is quite similar.
109
Displacement (m)
0.05
-0.05
ALPHAD=3.0
ALPHAD=0.7
ALPHAD=0.7
ALPHAD=3.0
0
3
4
Time (sec)
BETAD=0.00132
BETAD=0.000132
BETAD=0.00132
BETAD=0.000132
Figure 4.33: Vibration response for different values of the and parameters (Model 1B)
7
x 10
3
2
Stress (N/m )
1
0
-1
-2
-3
ALPHAD=3 BETAD=0.002
ALPHAD=0.7 BETAD=0.000132
ALPHAD=0.7 BETAD=0.00132
ALPHAD=3.0 BETAD=0.00-132
-4
-5
3
4
Time (sec)
Figure 4.34: Stress (x-direction) response for different values of the and parameters (Model 1B)
110
As part of the posts dynamic design-analysis, the frequency response generated by the
excitation is determined and compared with the natural modal frequencies of that structure.
The objective at this stage of the design-analysis is to ensure that the response frequency does
not coincide with the natural frequencies of the structure, as resonance would occur and the
behavioural response maximums of the structure would be amplified.
The frequency is determined by Fast Fourier Transform (FFT). A sample of vibration or discrete
signal is taken from the response analysis in figure 4.33. In this case, the sample taken extends
over a time domain of 4.5 s that begins at the moment the post is free to vibrate, i.e. after the
application of the transient load history of figure 4.28. Figure 4.35 and 4.36 shows the extent of
the sample used in the FFT taken from the original vibration response. Of the four amplitudevarying responses in figure 4.33, the response with the damping originally calculated from the
above simultaneous equations (4.23) is analysed in the FFT, recalling that the -damping and damping values are equal to 0.7 and 0.00132, respectively.
Displacement Variation of Node 14495 for Load History
Displacement (m)
0.05
-0.05
ALPHAD=0.7 BETAD=0.000132
0
3
4
Time (sec)
Figure 4.35: Extent of sampling frequency taken from the dynamic response (Model 1B)
111
y(t)
0.01
0
-0.01
-0.02
-0.03
ALPHAD=0.7 BETAD=0.000132
-0.04
2.5
3.5
4.5
5
time (seconds)
5.5
6.5
Figure 4.36: Sampling frequency from the dynamic response (Model 1B)
The FFT is a faster version of the Discrete Fourier Transform (DFT) which utilizes alternative
algorithms to carry out the same operation as DFT but in a substantially reduced time period.
The FFT function in MATLAB is an effective tool for computing the discrete Fourier transform of
a signal and is therefore used to carry out the discretisation of the sample taken here where the
transform is given by the following expression.
N 1
X ( k ) = x (t n )e jwk tn
n =0
(4.24)
where x(tn) is the input signal amplitude (real or complex), tn is sampling (sec), X(k) spectrum
of x at frequency k, and N is the number of time samples. Additional variables of the algorithm
include the sampling interval T (sec) and its inverse, the sampling rate fs (Hz) [27].
Only the positive frequency spectrum is considered in the analysis. Figure 4.37 presents the
discrete frequency domain representation of the sampled signal. Note that windowing, filtering
among other techniques were not applied in this analysis. These techniques are relevant for
analysis of more complex discrete signals [28].
-3
112
x 10
6
X= 5.9867
Y= 0.0060523
Amplitude
5
Freq (Hz)
10
Figure 4.37: Discrete frequency spectrum from vibration response of Model 1B in transient analysis
From initial examination of the dynamic response sample in figure 4.36, it is quite evident that
the periodic response composes of approximately six cycles per second, or 6 Hz. This
observation is confirmed by the FFT analysis of the sample where figure 4.37 above shows the
discrete frequency components with maximum amplitude occurring at 5.9867 Hz. While they
do not coincide, both the response frequency and first natural mode frequency 1 lie within
close range of each other resulting in possible amplification of the behavioural response.
However, other relative structural components such as the attached cantilever assembly and
electrification equipment are not considered in this dynamic analysis. Such components would
impede the structure from suffering from detrimental resonance.
113
114
Figure 4.38:: Post structure complete with cantilever and catenary assembly
The catenary assembly is the overhead lines used to transmit electrical energy to the train
through a pantograph connected to the locomotive. The power systems employed in Spain are
600, 750, 1500 and 3000 V in direct current (DC) and 25 kV in alternate current (AC). The typical
catenary assembly consists of a mechanical support wire and an electrical contact wire which
are connected together by a series of droppers which are all supported by the posts. Both the
support and contact wire are kept in a state of tension independent of temperature. The
spacing between the droppers is 4.50 m with the distance between two consecutive posts
known as the span. The span can vary from between 30 m to 70 m and depends on the degree
of curvature of the rail line.. Figure 4.39 shows a simplistic view of the catenary assembly. Other
components of the assembly include electrical insulators,
insulators stabiliser and bracing arm.
The permanent loading of the cantilever assembly consists of the total weight of the bronze
support wire (8300 kg/m3) of sectional area 65 mm2, the droppers (38.5 mm2) and the contact
wire (107 mm2) both of which are made from copper (8960 kg/m3). Knowing the uniformly
distributed load W of the contact wire (kg/m) and the length of two adjacent spans, li and li+1,
115
the total weight supported by one cantilever corresponds to half of that of the two adjacent
spans.
Pcontact =
Wcontact.(li + li +1 )
2
(4.24)
The weight of the support wire is similarly calculated to that in equation (4.24). Because the
wire is not linear but parabolic in shape, an additional 5% of the span is added to the support
wires length so as to account for this non linear shape.
Psup port =
.(1.05)
(4.25)
The final element of the catenary assembly load is the dropper. The number of droppers
supported by each cantilever is calculated simply in equation (4.26). A span of 45 m is chosen
for this analysis and with a dropper spacing of 4.50 m, the number of droppers supported by
the cantilever is 11.
n=
span
+1
dropperspacing
(4.26)
The total weight contributed by the droppers is calculated by multiplying the dropper linear
load (kg/m) by the average dropper length, and subsequently by the number of droppers per
span n.
Pdropper = n.Wdropper.laverage
(4.27)
These three load parts of the catenary assembly are summed together to form a total vertical
load applied on the cantilever as depicted in figure 4.41.
4.7.2.2 Temporary Loading
As already defined, the temporary loading accounted for in this analysis includes wind and ice
loading. The wind loading consists of two parts: the first part, which has been dealt with in the
preliminary analysis, is the horizontal loading on the post in both directions; the second part is
the horizontal loading of the catenary assembly, transversal to the post. The ice loading also
consists of two parts: the first is ice loading on the vertical surface of the post due to blizzardtype conditions; and the second is ice loading on the catenary wires. In relation to the wind
loading, the horizontal wind pressure acting on the surface of one of the cylindrical wire
elements of the catenary assembly and its resulting force component are shown in figure 4.40.
116
1
Pwind = .Cd .air.v2
2
(4.28)
where Cd is the drag coefficient where cables and wires have a value between 1.0 and 1.3 with
a value of 1.2 chosen in this
his instance, air is the air density (kg/m3), and v is the wind velocity
(m/s) which has a maximum of 120 km/hr in the project outline. The horizontal force vector is
given by
Fwind = Pwind.A
(4.29)
l +l
A = D. i i +1
2
(4.30)
where A is
The total horizontal wind load is divided into two components as dictated to the boundary
conditions of the catenary assembly. The wires are supported by an insulator above the
horizontal part of the cantilever and by a stabiliser/insulator midway up the angled
a
part of
cantilever. These components are shown in figure 4
4.41.
Ice loading on the vertical post surface is a phenomenon due to precipitation driven
horizontally by wind and is defined as ice accretion.
accretion. The winds velocity and profile of the
structure dictates the surface area on which the snow collects: winds of low velocity tend to
have accumulation on the leading surfaces while higher velocities cause accumulation on the
trailing surface due to wake turbulence. Persistent freezing precipitation and ccold temperatures
can cause this snow to compact and form ice. This part of the loading analysis is quite
conservative as a sheet of ice (960 kg/m3) of 40 mm depth is adopted to be frozen to the posts
vertical surface for cases in either direction. This lo
loading
ading type affects the symmetry about one
of the sectional axes and thereby affects the corresponding moment
moment inertia. In relation the
117
numerical analysis, the load is applied as its total (N) distributed vertically over the nodes that
form the relevant surface.
In terms of the vertical loading due to ice on the wires of the catenary assembly, the European
Standards EN 50119:2009 sets out a number of cases where the load magnitude depends on
altitude ranges or zones. The most critical of these is zone C of altitude superior to 1000 m
above sea level where the maximum loading due to ice (g/m) is 160. D where D is the diameter
of the wire (mm) [29]. This vertical loading type applies only to the horizontal cables such as the
support and the contact wires and not to the dropper.
All the above permanent and temporary loadings are shown schematically in figure 4.41 in
terms of the positioning and direction on the cantilever, and also shown is the horizontal
reaction of the strut T and the horizontal and vertical reactions RH and RV respectively, at the
point where the cantilever is connected to the post.
Figure 4.41: Loading and reaction system for cantilever and catenary assembly
The cantilever is fixed by an angle steel and bolt assembly as shown in figure 4.42. The
cantilever and catenary assembly, through the surface area of the angle steel in contact with
the post, applies a pressure onto the post surface. In relation to this structure, the pressure is
applied by the right-hand-side plate horizontally and vertically downwards. The strut is similarly
connected to the post, however as the strut is in tension, the pressure is applied onto the post
by the angle steel piece on the left-hand-side.
118
In terms of the FEM model, the horizontal loading is applied as a pressure over the equivalent
sized areas through elements of the meshed model so as to represent the boundary conditions
more accurately thereby avoiding the possible development of local stress spikes often
associated with force application to nodes. Vertical loading is distributed into vertical force
vectors on the nodes of the corresponding elements.
4.7.2.3 Cases of Load Combinations
Both the permanent and temporary loadings described above are combined with the previously
analysed post wind loading to form various possible loading scenarios that can affect the
entirety of the structure. The possible loading scenarios or cases are outlined in table 4.21.
There are five in total with the cantilever and catenary self weight being common to all of the
five. The cases are placed in order of increasing severity of loading culminating at Case E. The
wind loading cases of table 1 also contribute to the load hypothesis in this section. Case B and
Case D consist of the wind loading Case 1 while Case C and Case E contain wind loading Case 2
among additional loadings as describe above. The cases propose loading combinations that are
realistic for example, post wind loading perpendicular to the rail line (Case 1) is accompanied by
wind loading of the catenary assembly while post wind loading parallel to the rail line (Case 2)
does not include wind loading of the catenary assembly as the assembly runs parallel to the rail
line.
Cases
119
Case Description
Case A
Case B
Case C
Case D
Case 1 + Ice (web of UPN) + Catenary assembly (self weight + wind + ice)
Case E
4.7.2.4 Results
Table 4.22 and 4.23 shows the nodal results for the stresses (MPa) in global coordinates x, y,
z, xy and xz, and the maximum displacement (mm) of the structure for the five cases of the
load hypothesis in Model 1 and Model 2, respectively. Both the tension (T) and compression (C)
maximum values are given for each stress component. Depending on the load case type, the
maximum displacement is given for one or two directions.
x
Model 1
MPa
T
MPa
C
xy
MPa
C
Case A
Case B
Case C
Case D
Case E
5.107
7.915
25.544
10.786
27.100
10.374
12.052
25.750
21.551
27.596
5.742
6.602
6.405
11.862
10.866
10.577
12.116
11.867
21.803
20.123
Case A
Case B
Case C
Case D
Case E
15.200
17.490
55.060
31.095
58.450
29.915
34.558
55.639
61.951
59.678
2.273
2.618
2.516
4.700
4.274
3.708
4.254
4.154
7.650
7.047
xz
MPa
C
A) Weave + Tape
13.400 14.251
15.323 16.302
15.103 16.089
27.597 29.357
25.551 27.197
B) Tape
5.133
5.353
5.872
6.126
5.782
6.037
10.574 11.029
9.783 10.209
MPa
Uz
mm
Uy
mm
0.800
0.915
0.909
1.648
1.532
0.217
0.234
0.734
0.433
0.730
0.585
0.662
0.660
1.198
1.116
0.592
0.670
0.910
1.212
1.140
NG
NG
54.768
NG
55.144
6.686
13.837
7.563
19.989
12.783
1.083
1.237
1.228
2.229
2.072
0.425
0.480
0.813
0.870
0.813
1.123
1.269
1.269
2.298
2.143
1.133
1.282
1.295
2.320
2.179
NG
NG
53.852
NG
54.315
6.876
14.117
7.776
20.442
13.143
Model 2
120
xy
xz
MPa
MPa
MPa
MPa
MPa
Case A
Case B
Case C
Case D
Case E
4.379
7.023
41.885
9.290
43.613
8.432
9.785
46.247
17.505
48.789
4.201
4.830
12.168
8.678
12.322
8.089
9.255
12.592
16.673
15.218
Case A
Case B
Case C
Case D
Case E
12.755 25.937
15.840 29.921
92.593 102.480
26.093 53.671
96.663 108.540
1.577
1.812
2.674
3.251
2.951
2.888
3.313
3.231
5.958
5.483
A) Weave + Tape
10.538 11.335
12.049 12.963
11.876 12.592
21.701 23.347
20.091 21.631
B) Tape
4.256
4.469
4.868
5.113
4.795
5.043
8.767
9.206
8.113
8.525
Uz
mm
Uy
mm
0.850
1.022
2.918
1.801
2.818
0.393
0.379
3.436
0.739
3.618
0.462
0.522
0.520
0.945
0.880
0.465
0.526
1.365
0.953
1.400
NG
NG
61.596
NG
61.977
5.755
12.083
6.509
17.378
11.002
1.195
1.437
3.777
2.532
3.557
0.547
0.530
4.469
1.031
4.723
0.938
0.020
1.059
1.920
1.790
0.974
0.020
1.268
1.940
1.822
NG
NG
60.522
NG
60.574
6.149
12.858
6.951
18.516
11.749
121
FH
Fv
The application of loading combinations in directions parallel and perpendicular to the rail line
causes the post to suffer a twisting effect as can be observed from the stress distribution (x)
diagram of the structure below in figure 4.44. Such loading combinations are found in Case C
and Case E where the wind pressure is parallel to the rail line and the self-weight of the
catenary assembly causes a moment about the z-axis at the top of the structure. The stress
variation across the post section is captured graphically in figure 4.45. As before, the section is
taken 40 mm above the fixed end making again the sketch of the section in figure 4.13
applicable in this instance. The stress path demonstrates the effects caused by the twisting in
the section where the stress variation is not symmetrical about the centre of the posts width as
was observed for load Case 2 in figure 4.15.
122
40
Stress (MPa)
20
-20
-40
A) Weave & Tape
B) Tape
-60
200
400
600
Length (mm)
800
1000
Figure 4.45: Sectional stress variation of Model 1 post for load Case C
123
As briefly described in Section 4.1.2, the proposed composite post differs from the original steel
post in terms of its method of fixation which subsequently alters the overall height of the
structure and the boundary conditions imposed. The proposed structural fixation consists of
mounting the post onto the top of the original posts concrete foundation by the steel moment
resistant base. The base itself is secured to the concrete pad and to the composite post by bolt
connections. The model brings together three different element types with the motivation for
their implentation and their description outlined below.
4.8.1 Proposed Base Description
The bases preliminary design is based on the classical moment resistant base design for post
type structures where overturning of the structure is impeded by plate components fixed
transversally to the axis of overturning. Carbon steel S275JR similar to the original post, is the
material of choice because of its mechanical properties as outlined previously in table 3.3 of the
original steel post analysis. Steel was also chosen for the relative ease in which it can be
fabricated into complex mechanical structures including. Figure 4.46 shows an image of the
proposed base structure developed in the numerical model. Also included are the separate area
components which determine the required meshing characteristics in the latter stages.
124
Figure 4.46: Image of proposed steel base taken from the numerical model
The geometry of the structure is shown as CAD drawings below in figures 4.47 to 4.49. The
main features of the base include a flat plate in contact with the surface of the foundation,
vertical thin-walled section at a height of 200 mm and an inner section with the equivalent
outer dimensions of the post structure. The word equivalent is used with caution as the inner
dimension of the vertical walls of the base are fabricated to the same outer dimensions of the
post plus a tolerance. This extra-dimensional tolerance factors in discrepancies that come about
during fabrication, these include residual stresses and therefore deformations experienced in
the steel base caused during fabrication (welding and cutting), and deformations and
curvatures in the composite post which can occur during the curing process as a result of
unsymmetrical laminates, and fibre alignment discrepancies.
As already mentioned above, the base includes plate components connected at their thickness
to the vertical wall face and to the horizontal plate. The majority of moment resistance in the
structure is provided by these plates. In terms of dimensions, the plate component has a width
of 75 mm which tapers at one side inwards towards its height of 150 mm. These moment plates
are fabricated such that the taper does not reach its bottom or top. This blunting of the taper
removes the likely occurrence, during loading, of stress concentrations of infinitesimal
magnitudes caused by severe geometric discontinuities. Bolt holes in the horizontal plate are
situated outside the section wall and between the moment plates. In total, there are ten bolt
holes in the horizontal plate. The bolt holes that connect the base and post are situated in the
bases vertical walls spaced between the moment plates. Each of the longitudinal and
transversal walls consists of four and two bolts holes, respectively.
125
126
As can be seen in the above figure, the geometry of the element is defined by its eight nodes
with its coordinate system directly related to the configuration of the nodes. Accuracy of the
model is enhanced by solving the system by full integration (KEYOPT(2) = 0) with extra
displacement shapes (KEYOPT(1) = 0) although it is more CPU intensive. The full integration
analysis allows the capture of bending behaviour of single layer elements [8].
127
The problem of contact caused by bordering elements of distinct identities is omitted from the
numerical analysis. As a result, the associated non linearities are not considered and the overall
computational time is reduced. The elements of both the post (shell) and base (solid) in contact
within the base wall are assumed to be adhered with one another thereby sharing equivalent
transformations and rotations while retaining their own material properties within their
respective geometries. Note that contact in italics refers to the physical state and not to the
numerical problem. In order to maintain complete adherence between both sections, the
tapered width of the post is rendered vertical for its bottom 200 mm which is within the bases
walls. Above this the post begins to taper towards its final width of 200 mm at the top.
However, this limitation creates a new section variation along the posts length, thereby
changing its overall behavioural response from the analysis made for Cases 1 and 2 in tables 4.9
and 4.10. In continuation of the above assumptions, the bolts on the bases vertical walls are
not considered. The connection between both structures is achieved by merging their
bordering nodes. Figure 4.51 shows the geometry of the numerical model before and after the
bolts are removed from the analysis.
Figure 4.51: Bolts on vertical walls removed due to merging of both solid and shell elements
128
There are numerous methods to simulate the bolt assembly which include a complete solid
bolt, hybrid bolt, spider bolt, Rigid Body Element (RBE) bolt, coupled bolt, and no-bolt
simulation, with each method varying in their degree of accuracy. The most accurate approach
is to model the complete 3D solid bolt assembly however, with regards to the present analysis,
a large number of bolt assemblies are required (10 in total) and as a result, modelling of solid
bolts is impractical. Therefore, the analysis of the bolt assembly is carried out using line
elements (LINK10) and coupled nodes. This method reduces drastically the number of elements
required in the bolt assembly analysis with the LINK10 element adequately simulating the nut.
The typical bolt joint assembly and its numerical model equivalent are shown in figure 4.53.
129
Master Node
Link Element
Base surface in
contact with bolt
Slave Nodes
Plate thickness
The figure above shows the bolt element (LINK10) passing through the bolt hole and coupled to
surface nodes that represent the contact area between the head of the bolt and the base plate.
The LINK10 element is a 3-D spar element with three degrees of freedom at each node
(translations x, y, and z directions). The element simulates the nut of the bolt assembly which
has uniaxial tension capabilities only. The head is represented by coupled nodes. Coupling is
carried out between the outermost node of the LINK10 element (master node) and the
circumferential nodes of the flange at the hole (slave nodes). These circumferential nodes are
created by defining separate circular areas around the holes during the model development
and are mapped meshed. The bottom node of the link element simulates the point at which the
bolt enters the foundation. The displacement at this node is impeded.
In order to perform the iterative process of design, a number of the design drivers were defined
parametrically in the numerical input code of the model. Design drivers of the base structure
include the thickness of the horizontal base plate, the walls, and the moment plates which are
all defined separately. Also included as parameters are the bolt hole size and the area of the
bolt head in contact with the base. After completing the iterative design process, it was
determined that for the analysis stage the thickness parameters of the horizontal base plate
and the walls/moment plate would be equal to 9 mm and 7 mm, respectively.
130
Taking into account that the steel base structure has two planes of symmetry, it is initially
modelled as one-quarter of its entire size. Due to the bases isotropic properties, the REFLECT
command is adequately used in terms of overall time reduction in the geometric and meshing
development of the solid element. Figure 4.54 describes the progression of the models
development from one-quarter to complete structure using the REFLECT command.
However, the REFLECT command is not utilised for the shell element structure (i.e. the
composite laminates of the post) due to its anisotropic nature. This command would create
conflicting lamina configurations at areas either side of the planes of symmetry. In order to
merge the nodes of both element types at the interface of the base/post, the elements of both
structures must be identical in the areas of contact. To achieve this, the complete base
developed and meshed, a copy of the keypoints at the bases internal vertical surface (in
contact with composite laminates) is made at their same location. The shell areas and their
orientation are defined by these new keypoints and their sequence (i, j, k, l). The newly meshed
SHELL181 elements are dimensionally identical as to those of their corresponding SOLID45
elements. Figure 4.55 shows firstly the complete meshed base model (grey) and secondly the
coinciding meshed shell areas (green) of the post that adjoin the interior faces of the bases
walls.
131
Figure 4.56 shows the meshed base and included in its interior is the section of the meshed
post that is merged with the base which effectively anchors the post. Also shown are the
stacked layers that create the UPN beam and plate laminate components of the post.
Figure 4.56: Meshed base plus meshed post section joined to base
132
The remaining length of the composite post is created in the same manner as described
previously in the method statement of Section 4.2.2. As the base structure is of the most
importance in this analysis, the unconstrained post length is coarsely meshed so as to
reduce computational time. While the meshing maybe relatively coarse, the response
obtained remains at a high degree of accuracy in its approximation due to the full
integration option for the SHELL181 element used in the analysis (KEYOPT(3) = 2). Figure
4.51 shows a segment of the coarsely meshed unconstrained post (green) connected to the
finely meshed post segment contained within the base structure.
133
translation and rotation so as to prevent them pushing down into the foundation. As
previously described, all the bolts are constrained by the bottom node of the LINK10 element.
4.8.5 Results
The most significant stress and displacement results are shown in table 4.24 below. They
include principal stress 1, 2, and 3, and Von Mises equivalent stress.
Case
1
2
1
MPa
87.869
232.52
2
MPa
41.847
102.10
3
MPa
27.174
72.910
Von Mises
MPa
75.461
196.30
Displacement
mm
0.030
0.110
Table 4.24: Principal stresses maximums and equivalent Von Mises stresses
The final bolt assembly dimensions are shown in figure 4.58 and were determined from the
British Standard Whitworth (BSW) system. The bolt is of Grade A2/A4 - DIN 965 DIN EN ISO
7046-2. The assembly includes also a lock nut which resists loosening under vibration by locking
the first nut in position. An alternative method to prevent loosening is to utilise only one nut
which has a nylon sleeve insert. As the nut is fed and tightened onto the bolt the thread cuts
into the nylon thereby holding the assembly together by means of friction.
Bolts
1
2
3
4
5
6
7
8
9
10
Load Case 1
Reaction Force Tonnes-Force
N
T
33.922
0.00347
33.736
0.00344
34.473
0.00351
33.922
0.00346
33.736
0.00344
2348.0
0.23943
2413.6
0.24612
7649.3
0.78001
2414.5
0.24621
2348.5
0.23948
134
Case 2
Reaction Force
Tonnes-Force
N
T
36.143
0.00369
36.142
0.00369
1086.3
0.11077
5874.3
0.59901
14832.0
1.51244
14832.0
1.51244
5872.7
0.59885
1085.9
0.11073
36.143
0.00369
36.142
0.00369
Table 4.25: Bolt traction force reactions and equivalent force in tonnes
The maximum reaction force is experienced in Case 2 its equivalent tonnes-force equal to 1.512
T. A Factor of Safety of 1.8 is applied to this maximum reaction force giving a value of 2.7216 T.
The maximum tensile force applicable to the bolt specified is 3.01 T implying that bolt election
satisfies the design-analysis.
In viewing the results in table 4.25, the most critical stresses are deemed to occur due to load
Case 2. The maximum stress values given are caused by stress concentrations occurring at the
bolt head connection and to a lesser magnitude, at the vertical wall corner and horizontal base
plate connection, both of which occur in the lift zone of the structure.
Figure 4.59: Stress concentrations located at bolt head-base plate contact area
135
While it is accepted that both types of stress concentrations occur at these regions, they are
however viewed as being overestimations of the actual stresses incurred. The overestimation of
the stress magnitudes are a result of a number of the limitations outlined at the beginning of
the base design. Two of which are of most relevance are recalled as being
The stress response for both limitations is shown above in figures 4.59 and 4.60. The method of
the nodal coupling used to represent the bolt assembly has its limitations. All translations and
rotations of the coupled nodes which represent the bolt head in contact with the base are
completely constrained. This in turn causes high local stresses in its proximate unconstrained
region. The severe change in geometry at the wall-base plate interface of the model is reduced
by the presence of welded connections between the two.
Figures 4.61 and 4.62 show the stress distribution and deformed shape of the base structure for
load Case 2. It must be highlighted that the displacement scale is amplified and the stress range
is reduced in order to appreciate the behavioural response. A number of interesting
observations can be made from figure 4.55 which include the sectional shape of the walls and
the response of the moment connections. The inward deformation of the bottom wall is
connected to the composite post width in compression while conversely the top wall is adhered
to the post area in tension. Consequently, the outer moment plates connected to the
longitudinal walls are forced into flexure while the centre moment plate, which coincides with
136
the symmetrical plane, does not contribute any significant behavioural response. This overall
sectional response coincides with the type of behavioural response experienced by narrow
beams where the axial strain distribution in the post gives rise to a significant amount of
deformation of the cross-section.
137
Figure 4.63 shows the behavioural response of the base structure for load Case 1 which again
contains an amplified displacement scale and a reduced stress range.
138