You are on page 1of 247

Retrofitting of Angle Legs

of Transmission Towers to
Increase Load Capacity

by

Chirawat Tongkasame

A thesis submitted for the degree of

Doctor of Philosophy (Civil Engineering)

School of Natural and Built Environments

University of South Australia

January, 2008
Contents
Contents .................................................................................................................................i
List of figures........................................................................................................................v
List of tables ........................................................................................................................ix
Abstract.................................................................................................................................x
Declaration .........................................................................................................................xii
Acknowledgments .............................................................................................................xiii
Chapter 1............................................................................................................................1
1 Introduction and objectives .............................................................................................1
1.1 Introduction..................................................................................................................1
1.2 Research questions.......................................................................................................2
1.3 Research program ........................................................................................................2
Chapter 2............................................................................................................................5
2 Background and literature review ..................................................................................5
2.1 Introduction..................................................................................................................5
2.2 Current analysis and design methods for self-supporting lattice towers .....................5
2.2.1 First-order linear elastic analysis ..........................................................................6
2.2.2 Second-order elastic analysis................................................................................6
2.2.3 General nonlinear analysis....................................................................................6
2.3 Current approaches to reinforcing existing transmission towers.................................7
2.3.1 Replacement of members......................................................................................7
2.3.2 Horizontal bracing diaphragms.............................................................................8
2.3.3 Upgrading of tower leg members .......................................................................11
2.4 Single angle leg members in compression.................................................................14
2.5 Compound cruciform leg members ...........................................................................16
2.5.1 Nominal flexural buckling strength about v-v axis, Pn,v .....................................17
2.6 Browns method in calculating capacity of reinforced columns ...............................18
2.7 Shear forces in connections .......................................................................................22
2.7.1 End connections..................................................................................................22
2.7.2 Intermediate connections ....................................................................................22
2.7.3 Shear force calculation........................................................................................24
2.8 Requirement and arrangement of intermediate connectors .......................................25
2.8.1 Requirement for intermediate connections .........................................................25
2.8.2 Arrangement of intermediate connectors............................................................26
2.9 Modelling background and concepts for use .............................................................27
2.9.1 Bolt pretensioning in connection models............................................................28
2.9.2 Use of beam elements for economy....................................................................32
2.10 Development of multi-panel tower models .............................................................32
2.11 Conclusion ...............................................................................................................35
Chapter 3..........................................................................................................................36
3 One-panel experimental study Methodology ............................................................36
3.1 Introduction................................................................................................................36
3.2 Test series and modifications.....................................................................................36
3.2.1 Initial tests...........................................................................................................36
3.2.2 Modification to top end connection ....................................................................37
3.2.3 Bolt pretension....................................................................................................37
i
3.2.4 Number of bolts at intermediate joints ...............................................................38
3.2.5 Special case of welded connections....................................................................38
3.3 Terminology used for test specimens ........................................................................38
3.3.1 Connection types.................................................................................................38
3.3.2 Connecting methods ...........................................................................................39
3.3.3 Testing sets .........................................................................................................39
3.4 Test specimens, assembly, and apparatus..................................................................39
3.4.1 Specimens general details..................................................................................39
3.4.2 Mechanical properties of component members ..................................................42
3.4.3 Bolts and bolt tensioning ....................................................................................42
3.4.4 Bottom support ...................................................................................................43
3.4.5 Top support .........................................................................................................44
3.4.6 Measurement of internal forces ..........................................................................45
3.4.7 Loading ...............................................................................................................47
3.4.8 Displacement measurement ................................................................................48
3.5 Test procedure............................................................................................................48
3.6 Single angle tests .......................................................................................................49
3.7 Example of determination of internal forces .............................................................50
3.8 Conclusions................................................................................................................51
Chapter 4..........................................................................................................................53
4 Experimental study Results and discussion ..............................................................53
4.1 Introduction................................................................................................................53
4.2 Test result summary...................................................................................................53
4.3 Physical observation during tests...............................................................................55
4.3.1 Plate connectors at all joints, Set 1 .....................................................................55
4.3.2 Aligned plate connection, Set 1 ..........................................................................57
4.3.3 Alternating plate connection, Set 1.....................................................................59
4.3.4 Cruciform angle connection, Set 1 .....................................................................61
4.3.5 Alternating plate connection, Set 2 (pretensioned bolts)....................................63
4.3.6 Cruciform angle connection, Set 2 .....................................................................64
4.3.7 Welded alternating plate connection...................................................................66
4.3.8 Conclusions from physical observations ............................................................67
4.4 Internal force analysis................................................................................................67
4.4.1 Load vs. vertical displacement relationship........................................................68
4.4.2 Internal force development: Member forces at sub-panels and shear forces at
connectors ....................................................................................................................69
4.4.3 Conclusions from internal force analysis............................................................83
4.5 Recommendations arising from one-panel tests ........................................................84
Chapter 5..........................................................................................................................86
5 Development of one-panel finite element models of reinforced tower legs ...............86
5.1 Introduction................................................................................................................86
5.2 General material properties........................................................................................87
5.3 One-panel single angle models ..................................................................................87
5.3.1 General details ....................................................................................................88
5.3.2 Buckling analysis and results..............................................................................89
5.3.3 Models with geometric imperfection and results................................................91
5.4 Bolt pretensioning......................................................................................................96
5.4.1 Bolt design ..........................................................................................................96
5.4.2 Meshed geometries and damping constant values..............................................97
5.4.3 Steps involved and energy curves.......................................................................98
5.4.4 Resultant pretension..........................................................................................100
ii
5.5 Reinforced models ...................................................................................................100
5.5.1 Stages 1 and 2, Type-1 Model (using detailed bolted connections + continuum
elements for sub-panel portions)................................................................................101
5.5.2 Stage 3, Type-2 Model (using detailed bolted connections + beam elements for
sub-panel portions) ....................................................................................................114
5.5.3 Stage 4, Type-3 Model (using simplified connections + beam elements for sub-
panel portions) ...........................................................................................................117
5.6 Conclusions..............................................................................................................122
Chapter 6........................................................................................................................124
6 Extension to multi-panel tower models Reinforced compression leg ...................124
6.1 Introduction..............................................................................................................124
6.2 Observation of Indian and Silva Towers .................................................................126
6.2.1 Indian Tower.....................................................................................................126
6.2.2 Silva Tower.......................................................................................................130
6.2.3 Summary...........................................................................................................132
6.3 Geometries of tower models ....................................................................................132
6.4 Design concept of the models (use reinforcement for the vertical leg only) ...........135
6.5 Determining target member forced by truss analysis and location to terminate the
reinforcement .................................................................................................................135
6.6 Unreinforced pilot models .......................................................................................138
6.6.1 General material properties...............................................................................139
6.6.2 Support conditions ............................................................................................139
6.6.3 Type of analysis procedure, loading time period, and loading method............139
6.6.4 Member types and shapes and tie addition .......................................................140
6.6.5 Unreinforced models with cores .......................................................................146
6.7 Reinforced models ...................................................................................................147
6.7.1 Reinforcing member .........................................................................................147
6.7.2 Cases, interaction assigning, runs, convergence, and forces ............................148
6.7.3 Change in support conditions ...........................................................................150
6.7.4 Results and discussions.....................................................................................150
6.8 Capacity increase potential based on Browns method ...........................................174
6.8.1 Combined code method and Browns method..................................................174
6.8.2 Examples...........................................................................................................174
6.8.3 Discussion.........................................................................................................178
6.9 Conclusions..............................................................................................................181
Chapter 7........................................................................................................................185
7 Summary, conclusions, and recommendations ..........................................................185
7.1 Reinforced one-panel leg testing .............................................................................185
7.1.1 Objectives and summary...................................................................................185
7.1.2 Conclusions drawn from the one-panel testing.................................................186
7.2 One-panel finite element models .............................................................................188
7.2.1 Objectives and summary...................................................................................188
7.2.2 Conclusions.......................................................................................................189
7.3 Multi-panel FE modelling........................................................................................190
7.3.1 Objectives and summary...................................................................................190
7.3.2 Conclusions.......................................................................................................193
7.4 Summary recommendations for designers...............................................................197
7.5 Summary recommendations for future research......................................................199
References.........................................................................................................................201
Appendix...........................................................................................................................204

iii
A. Single panel test resultsLoad-displacement curves, sub-panel force
development curves, and connection shear development curves ............................204
Cru-2B-S1, Test 1 ......................................................................................................204
Cru-2B-S1, Test 2 ......................................................................................................206
Cru-2B-S1, Test 3 ......................................................................................................207
Cru-2B-S1, Test 4 ......................................................................................................208
Cru-3B-S1, Test 1 ......................................................................................................209
Cru-3B-S1, Test 2 ......................................................................................................210
Cru-3B-S1, Test 3 ......................................................................................................211
Cru-3B-S1, Test 4 ......................................................................................................212
Ali-2B-S1, Test 1.......................................................................................................213
Ali-2B-S1, Test 2.......................................................................................................214
Ali-3B-S1, Test 1.......................................................................................................215
Ali-3B-S1, Test 2.......................................................................................................216
Alt-2B-S1, Test 1.......................................................................................................217
Alt-2B-S1, Test 2.......................................................................................................218
Alt-3B-S1, Test 1.......................................................................................................219
Alt-3B-S1, Test 2.......................................................................................................220
Alt-2B-S2, Test 1.......................................................................................................221
Alt-2B-S2, Test 2.......................................................................................................222
Alt-2B-S2, Test 3.......................................................................................................223
Alt-W-S2, Test 1........................................................................................................224
Alt-W-S2, Test 2........................................................................................................225
Cru-2B-S2, Test 2 ......................................................................................................226
Cru-2B-S2, Test 3 ......................................................................................................227
B. Member selections for multi-panel tower models ................................................228
B.1 Dimensions and properties of 75 x 75 x 6 equal angle .......................................228
B.2 Dimensions and properties of round-hollowed sections used in the selection of
tension legs and bracing members .............................................................................228
B.3 Forces and adjusted forces on the legs, Panels 1 to 10 .......................................229
B.4 Selection of round-hollowed sections for tension legs and bracing members based
on ANSI/AISC 360-05 formulae ...............................................................................231
B.5 Formulae used in determining the compressive strengths of round-hollowed
section bracing members (from ANSI/AISC 360-05) ...............................................232

iv
List of figures
Figure 2.1 Tower deflected shape of 220 kV as-built tower (Albermani & Kitipornchai,
2003) ......................................................................................................................................8
Figure 2.2 A modification scheme to strengthen 220 kV tower (Albermani & Kitipornchai,
2003) ......................................................................................................................................9
Figure 2.3 Tower sub-structure with type 1a diaphragm bracing (Albermani et al., 2004)10
Figure 2.4 Upgrading of TV tower using type 2c diaphragm bracing (left pair); Deflected
shapes at collapse before and after the upgrade (right pair) (Albermani et al., 2004).........11
Figure 2.5 Leg reinforcement by attaching additional angles to existing legs (Courtesy of
Powerlink Queensland: Dan Gleeson Microwave Structure) ...........................................13
Figure 2.6 A tower reinforced at legs and certain bracing members (CIGRE Meeting
Upgrading, uprating, and refurbishing, 20042005) ...........................................................14
Figure 2.7 Typical compound compression member..........................................................16
Figure 2.8 Doubly symmetric compound angles ................................................................17
Figure 2.9 Reinforced column model and loading mechanism (Brown, 1988)..................19
Figure 2.10 Assumption made when plastic hinge forms on core column (member CGD)
(Brown, 1988)......................................................................................................................20
Figure 2.11 (a) A double-angle bracing specimen; (b) Mid-span model for maximum shear
force at an intermediate connection (Astaneh-Asl et al., 1985)...........................................23
Figure 2.12 Equilibrium of forces of a built-up member (Zandonini, 1985)......................25
Figure 2.13 Slender specimens with one to five intermediate connections (Temple et al.,
1986) ....................................................................................................................................26
Figure 2.14 Arrangement of intermediate connectors: (a) Aligned; (b) Alternating; and (c)
Cruciform (Temple et al., 1994) ..........................................................................................27
Figure 2.15 Continuum element beam-to-column connection model (Citipitioglu et al.,
2002) ....................................................................................................................................29
Figure 2.16 Bolt pretension calibration curves used in the FE model (Citipitioglu et al.,
2002) ....................................................................................................................................30
Figure 2.17 General configuration of joint model (Reid & Hiser, 2005) ...........................31
Figure 2.18 Combined use of beam and shell elements for a compression member (Prasad-
Rao & Kalyanaraman, 2001) ...............................................................................................32
Figure 2.19 Tower geometry layout (Silva et al., 2005) .....................................................33
Figure 2.20 Dummy bars added to the structure: (a) Vertical plan; (b) Transverse plan,
section A-A (Silva et al., 2005) ...........................................................................................35
Figure 3.1 Dimensions and arrangement of test specimens: (a) Alternating; (b) Cruciform;
(c) Aligned ...........................................................................................................................37
Figure 3.2 Bottom support of plate connection type specimens (general views) ...............44
Figure 3.3 Semi-spherical top support and its parts............................................................45
Figure 3.4 Locations and numbers of strain gauges on (a) cross-section of the first sub-
panel of Set 1 specimens, (b) profile of the main leg of Set 1 specimens, (c) cross-section
of the first sub-panel of Set 2 specimens, and (d) profile of the main leg of Set 2 specimens
.............................................................................................................................................46
Figure 3.5 Baldwin loading controller and data logging equipment ..................................47
Figure 3.6 A cruciform specimen under test in the loading machine .................................48
Figure 4.1 Summary of strength increase percentages of the specimens of different
connection types and bolt pretensioning forces from Sets 1 and 2 steels............................55
Figure 4.2 Failure mode of Pt-Ali-3B-S1, Test 2 ...............................................................56
Figure 4.3 Deformed shapes of (a) Ali-2B-S1, Test 1, and (b) Ali-3B-S1, Test 1.............58
Figure 4.4 Deformed shapes of (a) Alt-2B-S1, Test 1, and (b) Alt-3B-S1, Test 1.............60

v
Figure 4.5 Deformed shapes of (a) Cru-2B-S1, T1, (b) Cru-2B-S1, T2, and (c) Cru-3B-S1,
T2 .........................................................................................................................................62
Figure 4.6 Deformed shape of Alt-2B-S2, Test 1. (a) Small overall bending. (b) Slight
local buckling of the main members unconnected face at the upper intermediate
connection............................................................................................................................63
Figure 4.7 Deformed shape of Cru-2B-S2, Test 1. Left: A small bent at the top portion,
beginning from the first sub-panel upward. Right: A slight local buckling immediately
below the top connection .....................................................................................................65
Figure 4.8 Deformed shape of Alt-W-S2, Test 1................................................................66
Figure 4.9 Typical patterns of load vs. vertical displacement curves of weld- and bolt-
connected specimens and single-angle tests ........................................................................68
Figure 4.10 Axial force development curves of Alt-W-S2, Test 2 specimen.....................71
Figure 4.11 Shear development curves of Alt-W-S2, Test 2 specimen..............................72
Figure 4.12 Axial force development curves of Cru-2B-S2, Test 2 specimen ...................73
Figure 4.13 Shear development curves of Cru-2B-S2, Test 2 specimen ............................74
Figure 4.14 Axial force development curves of Cru-2B-S1, Test 3 specimen ...................75
Figure 4.15 Shear development curves of Cru-2B-S1, Test 3 specimen ............................76
Figure 4.16 Shear development curves of Cru-2B-S1, Test 1 specimen ............................77
Figure 4.17 Axial force development curves of Alt-2B-S2, Test 2 specimen ....................78
Figure 4.18 Shear development curves of Alt-2B-S2, Test 2 specimen .............................79
Figure 4.19 Axial force development curves of Alt-2B-S1, Test 1 specimen ....................80
Figure 4.20 Shear development curves of Alt-2B-S1, Test 1 specimen .............................81
Figure 4.21 Axial force development curves of Pt-Ali-3B-S1, Test 2 specimen ...............82
Figure 4.22 Shear development curves of Pt-Ali-3B-S1, Test 2 specimen ........................83
Figure 5.1 General configuration of a single-angle model and when meshed (on the right)
.............................................................................................................................................89
Figure 5.2 The first four buckled modes of single-angle model using C3D8 elements .....90
Figure 5.3 Load-time curves of 500E1, 500R1, and 500S1 models ...................................94
Figure 5.4 Load-displacement curves of 500E1, 500R1, and 500S1 models and of three
single-angle tests..................................................................................................................95
Figure 5.5 Load-time curves of all cases and runs showing convergence..........................96
Figure 5.6 (a) Bolt design; (b) and (c) Steps 1 and 2 of bolt pretensioning process...........99
Figure 5.7 Changes in total internal energy and total kinetic energy from bolt
pretensioning process of Models 1190_3D and 1195_3D...................................................99
Figure 5.8 Complete assembly of a reinforced one-panel model with cruciform bolted
connections in meshed view ..............................................................................................102
Figure 5.9 Load-time curves of five runs of Model 1199_3D_R .....................................104
Figure 5.10 Deformed shape at maximum capacity of (a) Model 1199_3D_R, Run 5; (b)
Model 1199_3D, Run 4; and (c) Specimen Cru-3B-S1, Test 1.........................................104
Figure 5.11 Load-time curves of four runs of Model 1199_3D .......................................105
Figure 5.12 Load-displacement curves of representative runs of Models 1190_3D,
1195_3D, 1199_3D, and 1210_3D ....................................................................................107
Figure 5.13 Load-displacement curves of models and test specimens with bolts having
moderate, near-zero, and zero pretension ..........................................................................109
Figure 5.14 Deformed shapes of Models 1190_3D, 1199_3D, and 1210_3D .................111
Figure 5.15 Sub-panel force development curves of Cru-3B-S1, Test 3..........................112
Figure 5.16 Sub-panel force development curves of Model 1195_3D.............................112
Figure 5.17 Sub-panel force development curves of Cru-3B-S1, Test 3 vs. Model
1195_3D.............................................................................................................................113
Figure 5.18 Model 1210_Beam: (a) Figure of one angle member and (b) Full assemblage
of the reinforced model......................................................................................................114

vi
Figure 5.19 Profile and coordinates of beam members for sub-panel portions of Models
1210_Beam and Simplified_Beam ....................................................................................115
Figure 5.20 Deformed shapes of Models 1210_3D (left), 1210_Beam (middle), and
Simplified_Beam (right) ....................................................................................................116
Figure 5.21 Load-displacement curves of Models 1210_3D, 1210_Beam, and
Simplified_Beam vs. Cru-2B-S1, Test 1 ...........................................................................117
Figure 5.22 Components of a simplified connection: (a) the Core Part with Locking
Stripes; (b) the Grooved Locking Part; and (c) the Smooth Wrapping Part......................118
Figure 5.23 Model Simplified_Beam: (a) Figure of one angle member and (b) Full
assemblage of the reinforced model ..................................................................................119
Figure 5.24 Slip between the two members of Model Simplified_Beam.........................120
Figure 5.25 Sub-panel forces of Simplified_Beam model................................................121
Figure 6.1 Indian Tower: Elevation views and dimensions of (a) original tower; (b) the
model used in investigation; and (c) the model used in truss analysis; (d) Loads applied to
the tower model .................................................................................................................128
Figure 6.2 Silva Tower: Elevation views and dimensions of (a) original tower; (b) the
model used in this investigation; and (c) the model used in truss analysis .......................131
Figure 6.3 Cross-braced tower model: (a) general view; (b) top view; and (c) vertical face
...........................................................................................................................................133
Figure 6.4 Warren braced tower model: (a) general view; (b) vertical face; and (c) inclined
face.....................................................................................................................................134
Figure 6.5 Truss-typed bracing members on (a) the vertical face and (b) the inclined face
of unreinforced No Cores Cross S120 model going out of plane ..................................141
Figure 6.6 Ties added to the bracing members on the vertical and inclined faces to prevent
them from moving out of plane for (a) Cross models and (b) Warren models..................142
Figure 6.7 Failure mode of (a) unreinforced No Cores Cross models and (b)
unreinforced No Cores Warren models when having additional ties.............................142
Figure 6.8 (a) Local coordinates of the integration points and segment thickness used in
defining 75 75 6 EA section using arbitrary shape function; (b) Orientations of the
vertical leg and the inclined legs with respect to the global axis system; (c) Showing seeds
that subdivide a member into a number of elements .........................................................144
Figure 6.9 Unreinforced Cores model ..............................................................................147
Figure 6.10 Outlook when the reinforcing member was added next to the existing leg ..148
Figure 6.11 Force-time curves of Panel 8 from the five runs of Cross S80 model ..........151
Figure 6.12 Cross S80 model, Case 2C 6/9, showing failure by buckling of Panel 7 ......152
Figure 6.13 Cross S80 2C 7/9 model ................................................................................154
Figure 6.14 Force distribution on the compression leg of Cross S80 2C FE model, Cases
6/9 and 7/9 .........................................................................................................................156
Figure 6.15 (a) A portion of the reinforced model created for simple elastic analysis; (b) A
portion of the reinforced leg in 3D showing the member orientations..............................159
Figure 6.16 Force distribution on the compression leg of Cross S80 2C 7/9 Elastic
ConI=1.0 ............................................................................................................................160
Figure 6.17 Force-time curves of Panel 10 from three cases of Cross S80 FE model .....162
Figure 6.18 Detail of buckling of Cross S80 FE model, C1 7/9 case ...............................163
Figure 6.19 Force-time curves of Panel 10 from five runs of the four cases of Warren S80
FE model............................................................................................................................165
Figure 6.20 Failure modes of Warren S80 FE model: (a) Case 1, 2C 6/9, buckling at Panel
7; (b) Case 2, 2C 7/9, yielding at Panel 10; and (c) Case 4, 1C 7/9, buckling of the existing
leg at Panel 2......................................................................................................................166
Figure 6.21 Force distribution on the compression leg of three cases (2C 6/9, 2C 7/9, and
2C 8/9) of Warren S80 FE model ......................................................................................168

vii
Figure 6.22 Force distribution on the compression leg of Cross S80 2C 7/9 FE and Warren
S80 2C 7/9 FE....................................................................................................................170
Figure 6.23 Force-time curves of Panel 10 showing convergence from two runs of Cross
S100 2C 6/9 FE and two runs of Cross S120 2C 6/9 FE ...................................................171
Figure 6.24 Force distribution on the compression leg of Cross S100 FE model, case 2C
6/9 ......................................................................................................................................173
Figure 6.25 Force distribution on the compression leg of Cross S120 FE model, case 2C
6/9 ......................................................................................................................................174
Figure 6.26 Nominal capacity of cruciform reinforced leg (2 75 75 6 EA) from
Browns method.................................................................................................................179
Figure 6.27 Capacity increase potential of cruciform reinforced leg (2 75 75 6 EA)
from Browns method........................................................................................................179

viii
List of tables
Table 3.1 Connection types and components......................................................................41
Table 3.2 Strength properties of Set 1 and Set 2 steels from tensile coupon tests..............42
Table 3.3 Calculation example of internal forces for Cru-2B-S2, Test 3 ...........................51
Table 4.1 Summary of failure loads and strength increase percentages of Set 1 and Set 2
tests ......................................................................................................................................54
Table 4.2 Sum of tension force in bolts at each connection and average strength increase of
each connection type............................................................................................................79
Table 5.1 Chosen element details for the investigation in this section (in highlighted areas)
.............................................................................................................................................88
Table 5.2 Summary of all cases and runs of the single-angle model with initial
imperfection .........................................................................................................................93
Table 5.3 Stages of model development and comparisons which lead to the final simplified
reinforced model ................................................................................................................101
Table 5.4 Details of the models specified by their names in Table 5.3 ............................101
Table 5.5 Capacities and capacity increase percentages of Type-1 models .....................107
Table 5.6 Capacities and capacity increase percentages of models with zero bolt pretension
...........................................................................................................................................116
Table 6.1 Panel forces on the compression leg of Indian Tower model ...........................129
Table 6.2 Panel forces on the compression leg of Silva Tower model .............................132
Table 6.3 Basic dimensions and panel lengths of the four tower models .........................135
Table 6.4 Panel forces on the compression leg of Cross S80 model corresponding to the
assumed loads, and when adjusted for different strength increase demands, in comparison
with the design buckling strengths.....................................................................................136
Table 6.5 Panel forces on the compression leg of Warren S80 model corresponding to the
assumed loads, and when adjusted for different strength increase demands, in comparison
with the design buckling strengths.....................................................................................137
Table 6.6 Comparing design buckling strengths and reduced analysis buckling strengths to
check the validity of the created models............................................................................143
Table 6.7 Sections and areas assigned to the members according to their groups of
arrangement similarity or functions...................................................................................146
Table 6.8 Analysis of forces of Panels 8 and 10 from the fifth run (0.085m, 1.8s) of Cross
S80 2C 7/9 case, in comparison with their expected forces and predetermined strengths 154
Table 6.9 Sub-panel forces and connection shears from FE analysis of Cross S80 2C 7/9
FE at the target capacity stage ...........................................................................................157
Table 6.10 Variation of the connectors sizes and shear force results at the top connection
from simple elastic reinforce model simulating Model Cross S80 2C 7/9........................161
Table 6.11 The expected design forces and analysis forces of Panel 10 from five runs of
the four cases of Warren S80 FE model ............................................................................165
Table 6.12 Comparison of shear forces at the top connection from elastic models with
varying connection sizes and from FE models ..................................................................169
Table 6.13 The expected design forces and analysis forces of Panel 10 from two runs of
Cross S100 2C 6/9 FE and two runs of Cross S120 2C 6/9 FE.........................................172
Table 6.14 Critical loads of the existing leg, reinforcement, and reinforced leg made of 75
75 6 EAs, and % of capacity increase determined by the suggested method..............178

ix
Abstract
As a result of changes in design code requirements for wind loads, the increase in power
demand, and the growing communication industry, many existing steel lattice transmission
and communication towers now require strengthening to meet increased load requirements.
Various strengthening methods are being utilised by engineers in this area, with leg
reinforcement being one of the most common methods. However, even though it has been
used in practice for some time, there has been little research done into its effectiveness,
until this study.

The design of the study consisted of three parts: 1) Experimental study on reinforced one-
panel leg specimen; 2) Developing an economical finite element model which was able to
simulate the reinforced one-panel leg specimens well; and 3) Extension of the reinforced
one-panel leg model to multi-panel leg-reinforced tower models.

The first part aimed to determine the most effective connection types among Cruciform,
Aligned, and Alternating types, and examine the effect of bolt pretension on improvement
of load carrying capacity. The specimens had two setsthe first with snug-tight bolts and
the second with pretensioned bolts. It was found that Cruciform type was the most
effective type; and although bolt pretensioning did not help increase the capacity much on
average, it did bring more consistent performance.

In the second part, three finite element models of reinforced one-panel legs were created.
Simplification was gradually done to get the most computationally economical but still
reliable model. A simple slip connection design was originated in this thesis. It used
locking stripes and locking grooves for force transferring, and encasing rings to hold the
reinforcing member to the main member. This connection design in combination with
using beam elements for sub-panel portions created a very economical reinforced leg
model. ABAQUS software was used in modelling work. Buckling analysis could not be
performed because the models contained contact interactions. As a result, they had to be
analysed with perfect geometry, with the use of Dynamic, Explicit (Quasi-static) analysis
procedure. Since the models had only one panel, the disadvantage of using perfect
geometry was not significant.

In the third part, the model was extended from single to multiple panels so that the effects
of slenderness ratio and bracing pattern on the capacity increase potential could be
examined. Considering L / r = 80, 100, and 120 and Cross- and Warren-bracing patterns,
four nine-panel tower models were formed. Truss analysis was used to approximate the
x
minimum reinforcement required. The connection design and modelling techniques from
the second part were then applied to create the four leg-reinforced FE tower models. The
results showed that: 1) Bracing pattern was not an important factor, but the slenderness
ratio was; 2) A leg with slender panels provided with the approximated minimum
reinforcement tended be able to double its capacity better than with short panels; 3)
Although a leg with short panels provided with the approximated minimum reinforcement
might not gain 100% capacity increase, it could be achieved easily by extending the
reinforcement for a few panels; and 4) With slip connections, not all connections helped
transfer the force, and the maximum force transfer happened at the top connection. It was
also demonstrated that normal structural elastic analysis software could be used to
determine the maximum connection shear force.

The study has demonstrated that the use of retrofitted angles bolted to existing
transmission tower legs is an effective method of increasing load capacity, up to as much
as double the original capacity. The thesis concludes with a series of recommendations for
practicing designers who wish to use this reinforcing method.

xi
Declaration
This thesis presents work carried out by myself and does not incorporate without
acknowledgment any material previously submitted for a degree or diploma in any
university; to the best of my knowledge it does not contain any materials previously
published or written by another person except where due reference is made in the text; and
all substantive contributions by others to the work presented, including jointly authored
publications, are clearly acknowledged.

Chirawat Tongkasame

January 2008

xii
Acknowledgments
The great guidance, encouragement, and assistance of my principal supervisor, Associate
Professor Julie Mills, and associate supervisor, Dr. Yan Zhuge, is gratefully acknowledged.
Without their kindness to allow me to go through many things my own way with full
freedom, and their patience given when I was repeatedly behind the schedule, there would
not be this thesis. Again, thanks Julie for her cheering me up and long contributed time,
and thank Yan for her keen advices.

I would like to thank Mr. Graham Brown, a senior transmission design engineer of
ODonnell Griffin, who originated the idea of this thesis and dedicated plenty of his time
for consultation.

I thank the Energy Networks Association and member organisations, who through their
Australian Strategic Technology Program, contributed the funds supporting the research
program.

I also would like to thank the following placesUniversity of South Australia (Mawson
Lakes Campus), Mawson Lakes district, Adelaide City, and Australiafor providing me
all the good and warm feelings throughout these four years of my staying.

I dedicate this thesis to my father and mother who waited patiently at home, and in
particular, to my father who was always a strong example and constantly gave mental
support which helped me through many difficult periods when I felt discouraged and
wanted to pull the plug.

xiii
CHAPTER 1

1 Introduction and objectives


1.1 Introduction
Retrofitting of existing steel lattice towers has become an important area in the
transmission and communication tower industry for the last two decades. Existing
transmission towers are increasingly required to transfer greater loads than those for which
they were originally designed. One direct reason for this is from higher power demand for
public use. When constructing a new transmission line is considered uneconomical or
obtaining a new Right of Way is difficult or prohibited, usually due to environmental
considerations, one good solution is to increase the transfer rate of power. To do this, the
old conductors and power lines on the transmission towers will be replaced with larger
conductors and higher-capacity lines, which impose higher loads on the towers. In addition
to higher power demand, in some cases installation of antenna systems for communication
purposes on existing towers may be required several years after construction has finished.
This installation also imposes higher loads on towers, both dead load and wind load. In
addition, an indirect cause of increased loads is from the new loading code for strength
limit state (AS 1170, 1989) which imposes higher loads than the previous load code for
allowable stress design (AS 1170, 1971). This can result in exceeding the towers strength
if the tower was designed to the former code. For these reasons, the existing towers are
being pushed beyond their original design capacity and this has led to collapse of towers in
many countries. To prevent them from collapsing, there is a need to improve their
structural capacity in order to sustain the increased load requirement.

At present there are two main reinforcing methods that are used in practice for lattice
transmission towers. The first method is adding a number of horizontal braces
(diaphragms) to mid-length of the slender, inadequate bracing members along the height of
the tower. This will reduce the slenderness ratio of the bracing members, create more unity
of the members, and increase stability of the whole tower. The study of this method was
1
undertaken by a group of researchers in Australia at the University of Queensland
(Albermani & Kitipornchai, 2003, and Albermani, Mahendran & Kitipornchai, 2004).

The second method is focused on the capacity of the existing legs only. If the existing legs
of a tower appear to be inadequate, they are reinforced by attaching additional angle
members, usually called reinforcing members, parallel to the existing legs from the base
upward. The reinforcing members are usually angles of the same size as the existing legs,
and are connected corner-to-corner to the existing legs by bolted connections at intervals,
normally two or maximum three connections per panel. The cross-section of the reinforced
leg is now cruciform. This method for reinforcing the legs is popular and has been widely
adopted for many years in Australia. Practicing engineers view the effectiveness of the
method in two different ways. The first group of engineers will assume that the method can
double the strength of the existing leg because the total area of the leg has been doubled,
which may be unconservative. The second group of engineers will consider a reinforced
leg as being a compound compression member loaded concentrically, which is more
reasonable and conforms to the design code AS 3995 (1994) for steel lattice towers but
may be too conservative. The results of the two adoptions can be quite different. Worse,
not even the mechanism of force distribution from the existing leg to the reinforcing
member is known. These reasons have made engineers uncertain about the true safety and
economy of their design, and is the reason for this study.

1.2 Research questions


To enable a better understanding of the effectiveness of the leg reinforcing method to be
developed, this thesis has examined the following research questions:

1 How does the leg reinforcing method work, i.e. what is the force transfer mechanism
behind it?

2 How effective is the leg reinforcing method in achieving its desired capacity increase?

3 Are there any simple means that can be used to improve or optimise the current
industrial practice of design and construction of the leg reinforcement method?

1.3 Research program


The three research questions above cannot not be answered by conducting investigation by
either experiment or computer modelling using reinforced leg of a few panels long where
2
the leg stands alone, independently free of other members, in the whole experimental setup
or computer modelling setup. This is because (1) when an existing leg needs to be
reinforced, it will be reinforced continuously several panels, (2) there are force increments
between the panels of the existing leg which come from the bracing members that are
connected to the existing leg only, and (3) the reinforcing member is connected to the
existing leg only, which means panel forces on the reinforcing member come from the
existing leg, not directly from the bracing members. (Unlike a built-up or compound
compression member where there is only one panel concerned and only one magnitude of
force applied at both ends from the gusset plates.) Therefore the study of reinforced legs
needs to use a lattice tower structure with several leg panels similar to the real towers,
which means there must be more than one leg as well as bracing members in the structure
to be created. Two choices regarding the creation of the structure are available: create a
physical model in reduced scale or create a computer model. The second choice was
chosen due to the limited budget and the ease by which changes could be made. However,
the computer modelling study of reinforced multi-panel tower models could not be
completed until initial studies were conducted on single leg panels. These first two stages
were preliminary stages necessary for developing a simple, computationally effective
reinforced leg model and modelling techniques which were to be applied in the third stage.

In the first stage (Chapters 3 and 4), laboratory testing of reinforced one-panel leg
specimens was undertaken. The purpose was to get the test results which were to be used
for verification of the finite element models in the second stage, and also to find the most
effective connection types for the top connection and intermediate connections.

In the second stage (Chapter 5), finite element modelling was used to model the reinforced
one-panel leg specimens of the first stage, with the primary focus on finding a model
design which was computationally economical and at the same time was able to produce
reliable results.

In Chapter 6, after the simple and computationally effective design for a reinforced leg
model was developed, the geometries, dimensions, and member profiles of four multi-
panel tower models were decided. They differed with respect to the use of three
slenderness ratios of the common panels of the existing compression leg and two different
bracing patterns of the bracing members. Truss analysis was used to approximate the
minimum reinforcement for the existing compression leg, assuming the additional capacity

3
required was 100%. Four finite element tower models were then created with the existing
compression leg reinforced to the minimum as had been approximated. The design of
reinforced leg developed in the second stage used two connections per panel. The capacity
increases of the four models were determined. For any model whose 100% target capacity
increase was not met, improvement was done by extending the reinforcement for one or
two panels upwards and the model reanalysed. For a few cases, the number of connections
per panel was reduced from two to one in order to see whether this could be economically
effective. The study went on to examine how well a normal elastic structural analysis
software could predict the force distribution between the existing leg and the reinforcing
member in comparison with that of the FE models, and whether it could accurately
determine the maximum connection shear force. At the very end of the chapter, the method
of Brown (1988) in calculating the capacity of columns reinforced under load was
considered applied to reinforced tower legs. The method showed a fixed pattern of capacity
increase potential which was opposite to the prediction by the FE models.

In summary, the research objectives were:

1 To find the most effective connection types for the top and intermediate connections
through experimental study using reinforced one-panel leg test specimens, and to
determine the critical factors that influence the effectiveness.

2 To develop a simple and effective (reliable and computationally economical) finite


element model design of a reinforced leg, that is capable of being applied further to
develop reinforced multi-panel FE tower models.

3 To develop reinforced multi-panel FE tower models and determine the capacity


increase potential of a reinforced leg, considering factors such as slenderness ratio,
bracing pattern, and the minimum quantity of reinforcement required.

4 To investigate a simple means to determine the maximum connection shear force from
using a normal structural elastic analysis software to construct leg-reinforced tower
models.

5 To provide design recommendations and limitations for the leg reinforcing method, and
recommendations for future research work.

4
CHAPTER 2

2 Background and literature review


2.1 Introduction
This chapter will give the necessary background of the transmission towers from basic
analysis assumptions to how they are reinforced. The design strength formulae for single-
angle tower members and compound (or built-up) leg members will be provided. The
formulae for single-angle members were used later to prepare two tower models when the
investigation on the potential of leg reinforcing method took place. Although the load-
carrying condition of a compound compression member (normally used as web members
in trusses) seems not to match the condition of a compound leg member in transmission
tower where the leg member is continuous for several panels, the available design formulae
and requirements of the former should still be useful for the latter if adapted properly.

The chapter will then focus on the prior work of others, especially the ideas which
gradually led to the final design of the test specimens and setups in Chapters 3 and 4, and
the finite element models and modelling concepts in Chapters 5 and 6. The later chapters
will show how they were combined and implemented.

2.2 Current analysis and design methods for self-supporting


lattice towers
In analysing member forces of steel lattice towers, only the axial forces are generally
considered. Although slightly eccentric loads, lateral wind load, framing eccentricities, and
so on, normally cause bending moments in the members, they are believed to be small and
insignificant, compared with the axial forces. In addition, it is impractical to model every
eccentric detail. Therefore, lattice towers are analysed almost exclusively as ideal three-
dimensional space trusses, where all angle members are straight and pin-connected at
joints, permitting only tension or compression in the members and joint displacement.
Although not considered in analysis, the moments from normal framing eccentricities of
5
angles can be accounted for in the design of members by derating the angles load
capacity.

Usual industry practice is to conduct a first order linear elastic analysis in tower design.
However, other methods are possible and these are also discussed below.

2.2.1 First-order linear elastic analysis


This analysis method treats all solid members as linearly elastic (the members are capable
of carrying tension as well as compression) and assumes that the secondary moment effect
( P ) of the deflected shape of the leg members and (primary) bracing members can be
ignored. The redundant members, so-called secondary bracing members, are not
included in this type of analysis since they have no effect on the forces in the leg and
bracing members. This type of analysis is generally used for self-supporting lattice towers.

2.2.2 Second-order elastic analysis


Second-order elastic analysis takes into account the additional forces created by member
displacements calculated from the first-order analysis. A second-order analysis is one that
produces forces that are in equilibrium in the deformed geometry. The analysis is normally
performed as a succession of the first-order analysis; the deformed geometry of the
structure is updated at the end of each iteration. According to ASCE (1988), Guide for
Design of Steel Transmission Towers, conventional self-supporting towers are sufficiently
rigid, and a nonlinear analysis is not required, except for flexible towers or guyed towers.

2.2.3 General nonlinear analysis


Due to the fact that, in construction, leg members and (primary) bracing members are not
purely axially loaded mainly because of the way angles are connected (for example, one-
face of a bracing member to one face of a leg member, or one face of a secondary bracing
member to one face of a bracing member or a leg member) which caused eccentricity, the
redundant forces can have a significant effect on the tower. In this case, Prasad-Rao &
Kalyanaraman (2001) suggested that, if possible, a nonlinear analysis should be carried
out, with the following factors to be taken into account.

The leg members are usually continuous through the joints.


More than one bolt is usually used in the connections and hence the joints are semi-
rigid (not pinned).

6
The bracing members and secondary bracing members are normally bolted through
only one of their faces and hence the force transfers in the members are eccentric.
The joints are flexible due to the local deformation of the face of the angles under the
concentrated bolt forces.
The towers with high electric ratings tend to be flexible and hence equilibrium in the
deformed configuration has to be considered (large deformation effects).
The compression member deformation increases the bending moments ( P effect).

2.3 Current approaches to reinforcing existing transmission


towers
The majority of towers, which include most towers of 20 years age or older, would require
a major retrofit if required to carry additional loads. The retrofit method ranges from
removing and rebuilding large portions of the tower, to replacing or adding additional
bracing members, to reinforcing tower legs. A common improvement scenario for the
existing bracing members is replacing them with larger sections. For the legs, a few means
of upgrading are available. The methods are described below.

2.3.1 Replacement of members


For the bracing members (horizontals and diagonals), upgrading can be done in several
ways. A simple method is using bigger sections. The process typically begins with bracing
the tower with temporary supports first, then the undersized or overstressed bracing
members will be removed and replaced with the new stronger/stiffer members. This
method can be achieved relatively easily since the bracing members usually carry only a
small amount of load. For example, if after reanalysing an existing tower with a new set of
design loads some bracing members appear inadequate, they will be removed and replaced
with adequate sections. They can also be strengthened by either bolting or welding new
members, such as cover plates, onto the existing bracing members.

This method is usually avoided in practice for leg members because the process is clumsy.
If utilised, before removing an existing leg, longitudinal temporary member(s) must be
applied to hold all bracing members that are connected to the leg into their positions. The
leg is then removed from the concrete footing or both are removed together, and a new
section of leg is placed into position. The process for the legs needs high caution, is time-
consuming, and as a result is very costly.

7
2.3.2 Horizontal bracing diaphragms
Another possible method is adding internal bracing diaphragms to the overstressed
members. This approach has been studied by a group of researchers in Queensland,
Australia. In the second case study of Albermani & Kitipornchai (2003), they performed a
nonlinear analysis to evaluate the response of an existing 220 kV transmission-line tower if
subjected to higher wind loads than the original wind loads the tower was designed for.
Nine new revised loading conditions were used. The results revealed that the tower would
collapse under three of the nine loading conditions. The collapse was due to either spread
of plasticity or premature buckling. Figure 2.1 shows the magnified deflected shape of the
tower at collapse under a loading condition. Plastic hinges initiated in the compression leg
at the lower part of the common body and spread down in this leg. The nonlinear analysis
results indicated that the tower would fail at loads significantly below the revised ultimate
design loads. One modification scheme to strengthen the tower in the investigation was
adding a diaphragm at two locations with some horizontal and secondary bracing members
(Figure 2.2). With these upgrades, the analysis results showed that the tower response
improved but still did not reach the revised ultimate design loads under two loading
conditions. However, this computer modelling was not verified by physical testing.

Figure 2.1 Tower deflected shape of 220 kV as-built tower (Albermani & Kitipornchai,
2003)

8
Figure 2.2 A modification scheme to strengthen 220 kV tower (Albermani & Kitipornchai,
2003)

In the view of design engineers, there are some issues that this strengthening scheme (by
applying a few diaphragms and some bracing members to increase load carrying capacity
of an existing tower) does not cover, which make them feel reluctant to exclusively
implement it without applying other reinforcing methods as well.

The concept of columns that the shorter they are, the higher capacity they have has been
used with transmission tower design for a long time. If a leg panel appears to be inadequate
to carry its new design loads, its unsupported length will be reduced. This is done by
adding new secondary bracing member(s) to such a leg panel somewhere between the
panels two ends, normally at the middle. (The secondary bracing member(s) will be
connected to the leg on one end and to the main bracing member on the other end.)

One issue for reluctance is at the leg members of the lowest leg extension (that with K-
bracing pattern) which carry the highest force. The question that arises is if larger design
loads are to be applied to an existing tower, should not the leg members of the lowest leg
extension be reinforced too? The work of Albermani & Kitipornchai (in Figure 2.2) did not
address thisthe panel lengths of the leg members of such portion remained the same, not

9
reduced. To engineers, this means that the capacity of that leg portion has not increased.
Another issue is that most engineers in commercial practice do not routinely do, or would
avoid doing, nonlinear analysis. Therefore, it is hard for them to know where exactly to
add horizontal diaphragms. If the method of adding diaphragms is chosen, it is much more
likely they will be added abundantly, not selectively, but the exact capacity increase to be
expected is not known.

In 2004, Albermani et al. pursued a similar study again. This time a series of horizontal
bracing diaphragms were added to at mid-height of the slender diagonal members. The
study had two parts. In the experimental part, a tower sub-structure assembly was created
and strengthened with a variety of diaphragm bracing types. Figure 2.3 shows the sub-
structure with bracing diaphragm Type 1a. Two load cases designed to cause bending and
torsion respectively were considered. The test results showed that, with the addition of any
pattern of the additional diaphragms, buckling capacity of the slender diagonals could
improve considerably. In the computer modelling part, to determine the best solution of
how to strengthen a 30-year-old TV tower, a nonlinear analysis of the tower model was
organized with the use of Type 2c diaphragms equipped along the tower height. Trial and
error was used to find the best number and positions of the diaphragms. Figure 2.4 shows
the upgrading arrangement which led to the highest strength increase for the tower (nearly
+40%). This methodadding a series of horizontal diaphragms to mid-length of bracing
memberswas said to not only help improve performance, but also use less steel and be
easier to implement than the replacement method.

Figure 2.3 Tower sub-structure with type 1a diaphragm bracing (Albermani et al., 2004)

10
Figure 2.4 Upgrading of TV tower using type 2c diaphragm bracing (left pair); Deflected
shapes at collapse before and after the upgrade (right pair) (Albermani et al., 2004)

2.3.3 Upgrading of tower leg members


For the leg members, there are a few methods which can be used in upgrading. The method
of replacing the existing leg members is unpopular because it a difficult process in
construction.

2.3.3.1 Reducing panel length with secondary bracing members

One upgrading method is the addition of more secondary bracing members to the tower.
They will be connected to the leg members at one end and to the bracing members at the
other end. This reduces the unsupported length of the leg members and can increase their
axial load capability effectively. This method is believed to be a cost-effective option.
There is no strict rule when this method should be applied, and at what range of
slenderness ratio of the inadequate leg panel. The author believes it is chosen whenever it
can help accomplish the additional capacity demand quickly and economically. From
common sense, the benefit is high when the inadequate leg panel is long, and vice versa. A
11
disadvantage that can be seen quite obviously is that if a leg needs to be reinforced for
several panels continuously, using this method will involve adding a lot of new secondary
bracing members which means a lot of on-site workmarking and punching (or drilling)
bolt holes needs to be done at many locations on both the leg and the main bracing
members on site.

2.3.3.2 Adding reinforcing members to the existing legs

Another commonly used method is bolting or welding additional membersangles,


channels, or platesto the existing leg members on the exterior face of the tower. Figure
2.5 shows an example of this method applied to a lattice tower. This method increases
cross-sectional area of the leg members. Considering the process of making bolt holes
which can be quickly done with hydraulic puncher, this method can save time
considerably. This reinforcing method has been recognized and used in practice by tower
engineers in many places around the world for many years. However, there has been no
solid evidence of a study that details the effectiveness of the method. This is the aim of the
current study. As opposed to the previous reinforcing method, the advantage of this
method is that using only one reinforcing member can reinforce many panels of an existing
leg. On-site work will require marking and punching bolt holes on the existing legs only,
but bottom ends of the reinforcing members need to be cast into the concrete footing as
well.

An example of a tower where leg members and some bracing members have been
reinforced by means of adding a reinforcing member parallel to an existing member as
described above is given in Figure 2.6. Both members are connected together at intervals
using two short angle cleats and eight bolts (two on each connecting face) at each joint.
The existing member and reinforcing member become a compound member with
cruciform cross-section.

For hollow leg towers, reinforcing the legs can be simply achieved by filling the legs with
structural high-strength non-shrinking grout. After curing, the concrete fill will help
increase load-carrying capacity of the legs, provide better stability and protection from
local buckling. This option is easy to implement and more cost-effective than the two
traditional reinforcing methods above, while a sound knowledge of composite structures is
required. However, since this method is applicable to tube sections only, it is beyond our
concern in this study.

12
Figure 2.5 Leg reinforcement by attaching additional angles to existing legs (Courtesy of
Powerlink Queensland: Dan Gleeson Microwave Structure)

13
Figure 2.6 A tower reinforced at legs and certain bracing members (CIGRE Meeting
Upgrading, uprating, and refurbishing, 20042005)

2.4 Single angle leg members in compression


The background knowledge of the practical design concepts used for single angle
compression members has been discussed below because it was used later to help set up
the multi-panel tower models in Chapter 6.

An important design requirement based on ASCE (1988) and AS 3995 (1994), Design of
Steel Lattice Towers and Masts, is the limit of the effective slenderness ratio e . The limit

is 150 for leg members (e 150) ; 200 for bracing members (e 200) ; and 250 for

secondary bracing members (e 250) .

The e is related to the member slenderness ratio L / r . The length L for leg members

with symmetrical bracing and for other members is the actual unsupported length l of the
member for buckling about the relevant axis.

14
For leg members, the e is equal to the member slenderness ratio, or e = L / r . For

bracing members and secondary bracing members, the e depends on the end conditions.

If the condition is unrestrained against rotation at both ends of the support length, the e

is equal to L / r as well.

In Chapter 6, in finding the sizes of the members when constructing the tower models, the
bracing and secondary bracing members were assumed to have such end conditions for
simplicity, so that the L / r value of a trial section could be used in calculating the
compression capacity and, at the same time, indicate whether it was within the slenderness
ratio limit.

Below are the formulae taken from AS 3995 (1994), Design of Steel Lattice Towers and
Masts, which is based on ASCE (1988), Guide for Design of Steel Transmission Towers.
They were used to calculate the capacity of a single angle compression member. There are
formulae of the same purpose in other design codes too, but since they are not specifically
for transmission towers, they were neglected. The design axial compression capacity N c*
shall satisfy the following:

N c* N c Eq. 2.1

and N c = c Af y Eq. 2.2

where = 0.9 for compression members

N c = nominal capacity of the member in compression

c = member slenderness reduction factor

A = area of cross-section

f y = yield stress of steel

The member slenderness reduction factor is determined from the e as follows:

e
2

c = k f 1 0.5 for e c Eq. 2.3


c

2E
c = for e > c Eq. 2.4
f y (e ) 2
15
Where kf = form factor for members subject to axial compression

= 1.0 in this study

c = factor for the determination of c

2E
c = Eq. 2.5
kf fy

E = Youngs modulus of elasticity


= 2.05 105 MPa
Wakabayashi & Nonaka (1965) conducted an experimental study on buckling strength of
angles in transmission towers. A number of specimens of steel angles were tested. The
useful information was the approach used to fix the angle specimens to the end supports
(holes made near the ends of the specimens) and the design of the end supports which were
adjustable for producing different eccentricities. In the experimental part of this thesis,
Chapters 3 and 4, the ideas gained from such designs were modified and implemented.

2.5 Compound cruciform leg members


AS 3995 (1994) specifies that compound leg members built up with two or more angles in
cruciform section bolted together with stitch bolts at a distance less than 40rmin (Figure
2.7) can be considered as fully composite and treated as single members. ( rmin is the
minimum radius of gyration of a component angle.)

But when the distance is in excess of 40rmin , a consideration will need to be taken when
determining nominal strength based on flexural buckling about the axis that causes relative
deformations between the component members (flexural buckling about the v-v axis in
Figure 2.8, left).

Figure 2.7 Typical compound compression member


16
As compound cruciform columns have a doubly symmetric section, they fail either by
flexural buckling about the weak axis or by torsional buckling about the shear centre.
Because the columns are built-up section with stitch bolts, it is necessary to also take the
built-up section effect into account. Three nominal compressive strengths to be considered
are nominal strength based on flexural buckling about u-u axis ( Pn ,u ) , nominal strength

based on flexural buckling about v-v axis ( Pn ,v ) ; and nominal strength based on torsional

buckling ( Pn ,t ) . (Figure 2.8) From the practical design point of view, the torsional buckling

is not the major concern. Therefore, the smallest between Pn ,u and Pn ,v is the controlling

strength.

Figure 2.8 Doubly symmetric compound angles

2.5.1 Nominal flexural buckling strength about v-v axis, Pn,v

When the distance between stitch bolts is in excess of 40rmin , for the buckling mode which
involves relative deformations that produce shear forces in the connectors between
component members (which is flexural buckling about the v-v axis in Figure 2.8, left), the
effective slenderness ratio to be substituted in Equation 2.3 or 2.4 is:

e = 1 2 + 2 2 Eq. 2.6

where 1 = slenderness ratio of the compound member acting as a unit, about the
axis that causes relative deformations between the component
members

2 = slenderness ratio of one component angle

= C / rmin

C = distance between the stitch bolts


17
rmin = minimum radius of gyration of a component angle

Equations 2.1, 2.2, and 2.5 still apply for calculating nominal or design flexural strength.

2.6 Browns method in calculating capacity of reinforced columns


Brown (1988) developed a method for determining the ultimate capacity of a column
reinforced under load, based on a rational analysis. It was claimed that the analysis is
applicable to any kind of column and reinforcement, but it was not verified by extensive
testing.

The model consists of two separate columns representing the core column and the
reinforcement (Figure 2.9), connected together by rigid links to ensure they follow the
same axial deformation and have the same deflection curve. The model has three stages of
loading:

Stage 1: The core column carries an initial load when the reinforcement is welded to it.
Stage 2: The core column reaches its ultimate capacity
Stage 3: Failure occurs when the reinforcement is unable to stabilize the core column.

The loading in the core column, reinforcement, and the reinforced column at Stage 1 is:
Load on the core column Q is equal to its initial load Qo when applying reinforcement.

Load on the reinforcement P is zero. Hence, the total load on the reinforced column W is
Q + P = Qo + 0 = Qo .

As the load W is increased, the core column will approach its buckling load of Qcr and the
load on the reinforcement will increase. When Stage 2 is reached, the core column has
reached its maximum capacity Qcr with a plastic hinge at G. Any further load will be
carried by the reinforcement. The mechanism against failure at Stage 2 is shown in Figure
2.9(b). The total load on the reinforced column is W = Qcr + P .

As W increases, a plastic hinge will form at E and the reinforcement loses it stiffness. This
is the failure stage of the reinforced column.

18
Figure 2.9 Reinforced column model and loading mechanism (Brown, 1988)

Figure 2.10 shows the assumed behaviour of both members when the plastic hinge has
formed on the core column. From plastic analysis (Allen & Bulson, 1980), the lateral
deflections y and w of the core member and the reinforcement are:

y = FL / 4Qcr Eq. 2.7

FPe L3
w= Eq. 2.8
48 EI r ( Pe P)

where F = internal lateral force at mid-height between the core member and
reinforcement

Qcr = critical load of the core column only

Ir = moment of inertia of the reinforcement

Pe = Eulers load of the reinforcement = 2 EI r / L2

P = load carried by the reinforcement


Equating these two deflections, the load P on the reinforcement is obtained as:

19
Figure 2.10 Assumption made when plastic hinge forms on core column (member CGD)
(Brown, 1988)

P = Pe 0.822Qcr Eq. 2.9(a)

2 EI r
or P= 0.822Qcr Eq. 2.9(b)
L2

By taking P in Equation 2.9(b) equal to the critical load of the reinforcement Pcr , a critical

length L1 results. It is the maximum column length for which the properties of an unloaded
reinforced column can be used, given as:

2 EI r
L1 = Eq. 2.10
Pcr + 0.822Qcr

The alternative assumption is that the reinforcement will buckle first. Equation 2.9(b) is
equated to zero, a critical column length L2 results. It is the minimum length of column for
which the reinforcement provides additional stiffness to the core column.

12 EI r
L2 = Eq. 2.11
Qcr

20
Browns conclusions and how they would be applied to reinforced tower legs are described
as follows:

1 For a reinforced column with L < L1 , the initial load Qo on the core column has no
effect on the ultimate capacity of the reinforced column and can be assumed as zero.
The properties of the reinforced column can be used to calculate its ultimate capacity.

Likewise, this should mean that a reinforced tower leg with the panel length less than
L1 should be considered as a compound leg. The total capacity is equal to the ultimate
capacity of the compound leg (W = Pu ) , and Equation 2.9 need not be considered.

Calculating the ultimate capacity of the compound leg Pu will require the modified
effective slenderness ratio from Equation 2.6.

2 For L1 < L < L2 , the reinforced column uses the post-buckling strength of the
reinforcement after the column core has failed. Load on the reinforcement at the critical
stage is P given in Equation 2.9. The total capacity of such reinforced column is
W = Qcr + P .

This conclusion can be applied directly to reinforced tower legs of this L1 < L < L2
category.

3 For L > L2 , the reinforced column should be designed for elastic behaviour only.

Since tower leg members do not fall or are too short to fall in this case due to the
limiting slenderness ratio requirement L / r 150 , calculating this capacity need not be
considered in this study.

Based on the assumptions used which are understandable, especially the force-failure
mechanism model (Figure 2.10) which is rationally very much like what can be expected
to happen to reinforced tower legs, Browns method looks promising and is believed to be
applicable for predicting the capacity of reinforced tower legs. The use of this method is
demonstrated in Chapter 6, Section 6.8.

21
2.7 Shear forces in connections
2.7.1 End connections
When a concentrically loaded built-up column buckles about the axis that involves relative
deformations between the component members, there will be shear forces in the
connectors. Aslani & Goel (1992) and Sherman & Yura (1998) proposed equations for
determining the magnitude of shear force in the end connections. However, they have not
been adopted into a design code. ANSI/AISC 360-05 (2005) does provide a user note that
it is acceptable to design a bolted end connection of a built-up compression member for the
full compressive load with bolts in shear and bolt values based on bearing values; however,
the bolts must be pretensioned with Class A or B surfaces in order to prevent relative
movement between the component members at the ends as the built-up member takes a
curved shape. (Class A surface = unpainted clean mill scale steel surfaces or surface with
Class A coatings on blast-cleaned steel and hot-dipped galvanized and roughened surfaces;
Class B surface = unpainted blast-cleaned steel surfaces or surfaces with Class B coatings
on blast-cleaned steel)

2.7.2 Intermediate connections


Another issue is that there is no clear source that specifies how large a shear force the
intermediate connections should be designed for, or whether their design should be
identical to the design of the end connections but without pretension. A recommendation
was given by Astaneh-Asl et al. (1985) on the investigation of the behaviour of double-
angle back-to-back bracing members subjected to out-of-plane buckling due to severe
cyclic load reversals (Figure 2.11(a)). It recommended that the largest force transferred
through a stitch (or an intermediate connection) is equal to Py / 4 , where Py is the total

yield capacity of the double-angle bracing member. The recommendation was obtained
from a simple mechanism of the mid-span as shown in Figure 2.11(b), where after the
overall buckling has occurred, the cyclic load still continued.

22
Figure 2.11 (a) A double-angle bracing specimen; (b) Mid-span model for maximum shear
force at an intermediate connection (Astaneh-Asl et al., 1985)

Sherman & Yura (1998) stated that the bolted intermediate connections do not need to be
pretensioned, as long as the local buckling of a component member between the
connections prior to the overall buckling is prevented.

23
With the ANSI/AISC 360-05s user note and the recommendation of Astaneh-Asl et al.
(1985), it is now possible to design the end and intermediate connections for built-up
columns more confidently. Possibly, the concepts may be applicable for reinforced tower
leg members too, but physical investigations will be required. If that is the case, the end
connection will mean the very top connection where the reinforcing member is terminated
(see Figure 2.5), and the intermediate connections mean the other connections along the
leg.

2.7.3 Shear force calculation


Zandonini (1985) studied stability of compact built-up struts by experimental investigation
and numerical simulation. The compact built-up struts may be called buttoned struts
where generally they are made of two channel sections (chords) stitched together back-to-
back with filler plates that divide the whole length into sub-panels. Their use can be as web
members in trusses, bracing members, or even columns. In the experimental part,
Zandonini measured axial forces in sub-panels ( N k ) almost exclusively. The reason given
was that they were enough to be used to indicate the overall and local behaviour of the
specimens, and it helped avoid excessive measuring work for other values. After the axial
forces were obtained, they could be used to find shear forces at the connections (Vi )

directly. The relationship between the N k and Vi is shown in Figure 2.12.

The shear force at connection ith, determined from the axial forces of two adjacent sub-
panels of the top chord, N k and N k 1 , is:

Vi ' = N k N k 1 Eq. 2.12

Similarly, using those from the bottom chord, the shear force at the ith connection is:

Vi '' = N k*1 N k* Eq. 2.13

From the equilibrium condition, the Vi ' and Vi '' values must be identical. However, since

the readings of the axial forces N k and N K* can be uncertain in the actual measurement
process, Zandonini used the average value, or

Vi = (Vi ' + Vi '' ) / 2 Eq. 2.14

24
Figure 2.12 Equilibrium of forces of a built-up member (Zandonini, 1985)

The presentation of the axial forces and shear forces that Zandonini used ( N / Py , V / Py ,

and P / Pc ) was excellent because it makes the interpretation easy. Therefore it was
adopted in Chapter 4 of this thesis. Because the yield strengths of different pieces of steels
for different specimens were not equal, the specimens failed at different loads. The force
curves of each specimen were normalized, so that they could be compared with those of
other specimens.

2.8 Requirement and arrangement of intermediate connectors


2.8.1 Requirement for intermediate connections
Temple, Schepers & Kennedy (1986) studied the requirements for the number of
intermediate connections of starred (cruciform) angle compression members. Slender
starred angle members were considered in the study. The results from theoretical analyses
indicated that the buckled shape changes when the number of intermediate connections is
increased from one to two, resulting in an increase in the critical load; and the addition of
more than two intermediate connectors has no significant effect on the critical load. These
theoretical results were verified by experiments, where the specimens used welding as the
fastening method at all the connections. The study recommended that, for all starred angle
compression members made with equal angles, two intermediate connections should be
used, one at each of the third points, in addition to the end connections. (Figure 2.13
illustrates slender specimens with different numbers of intermediate connections used in
the study.)

25
Figure 2.13 Slender specimens with one to five intermediate connections (Temple et al.,
1986)

In Chapters 3 and 4 of this thesis which involved experiments on reinforced one-panel leg
specimens, the above recommendationusing two intermediate connections in addition to
end connectionswas adopted. The whole panel length was divided into three sub-panels.
Later, it was decided that such provision is unlikely to happen in real practice. This is
because the general panel length of leg members is relatively short and because the
reinforcement needs to be bolt stitched to the existing leg member intermittently and
continuously for several panels, there may not be enough space to use two intermediate
connections per panel. Or if it is implemented anyway the amount of work load for
marking and punching bolt holes for a structure can be excessive and the project becomes
uneconomical. The real practice regarding the number and positions of the connections as
in Figure 2.6 was adopted in the computer modelling study in Chapter 6.

2.8.2 Arrangement of intermediate connectors


Temple et al. (1994) conducted another study on slender built-up starred angle
compression members. This time it was to find a preferred arrangement for the
intermediate connectors. Three arrangementsAligned, Alternating, and Cruciform, all
using plates as the intermediate connectorswere applied (Figure 2.14). Again, the

26
specimens used welding as the fastening method at all of the connections. The study
concluded that as long as the component angles of a starred angle compression member are
firmly connected at the points of intermediate connections, the arrangement type of the
intermediate connectors does not have a significant effect on the load-carrying capacity.

What remains unclear from Temples study is that if the starred compression members are
short or have an intermediate length and if snug-tight bolting is used at the intermediate
connections instead of welding, will different arrangements and types of intermediate
connectors produce different results? For this reason, the aligned and alternating plate
arrangements were still considered in the reinforced one-panel leg experiment in Chapters
3 and 4. The cruciform arrangement was modified to using two angle cleats instead of two
plates per connection.

Figure 2.14 Arrangement of intermediate connectors: (a) Aligned; (b) Alternating; and (c)
Cruciform (Temple et al., 1994)

2.9 Modelling background and concepts for use


The purpose of Chapter 5 was to find a simple but effective finite element model that could
simulate the one-panel reinforced model satisfactorily in terms of performance. In addition,

27
that model was required to allow expansion of the study to the creation and analysis of
tower models with leg reinforced for several panels in Chapter 6.

To do this in Chapter 5, three different types of models which simulated the one-panel
reinforced specimens were created. The first model type was the most detailed onethe
longitudinal members, the connectors, and the bolts were created in separate pieces, all
using 3D continuum elements, and assembling was somewhat complicated. The second
model type was similar to the first type for the portions that represented the connections,
but beam elements were introduced for the portions that represented sub-panels. The third
model type used a new form of connection, while the sub-panel portions still used beam
elements. Some issues relevant to the FEM are discussed below.

2.9.1 Bolt pretensioning in connection models


The first model type had every part of the assembly (the longitudinal members, the
connectors, and the bolts (including nuts and washers)) built individually using 3D
continuum elements, and assembled later. This way of creation was based on most prior
modelling work where FE modelling was used in studying the behaviour of bolted beam-
to-column connections.

With respect to the modelling of a friction-type connection, most prior works found used
3D continuum elements, continuum elements or brick elements, for every part of the
model. The most important difference among them was the technique used in applying
pretension to the connection, which is discussed below.

2.9.1.1 Modelling of bolts and pretensioning techniques

Citipitioglu et al. (2002) studied FE modelling of partially restrained connections including


slip (Figure 2.15). The technique they used in pretensioning was easy to follow. A bolt part
consisted of two bolt heads and shank in the middle. They were created as one piece for
convenience. To clamp the parts of the connection, the pretension was achieved in two
steps. The first step employed the bolt parts with a shorter length shaft (or shank length)
compared to the total thickness of the connecting plates. Therefore, one side of the bolt
heads was initially in contact with its respective surface, while the other side representing
the nut was displaced by the pre-specified amount which would clear the bolts hole. In the
second step, the contact between the displaced bolt head and its respective surface was
activated and the imposed displacement was released, thus forming the desired clamping
between the parts via the bolt. The overall stiffness seen by a bolt was affected by several
28
factors, such as the deformation of the connected members, the bolt heads and nuts, the
interaction between the bolts and boundary conditions among others. Therefore it was not
possible to use the elastic force displacement equation of the bolt shaft to determine the
displacement needed to induce the desired pretension value in the bolts for the method
described above. To induce accurate pretension values, a method illustrated in Figure 2.16
was introduced. The forces at the contact between the bolt head and the plate were
computed by displacing the bottom of the bolt shaft. This gave a close approximation of
the actual pretension force in the connection model. One curve was plotted for one total
thickness of the plates connected. The advantage of this method was that the bolt geometry
was the same as that in the actual connection model.

It may be said that Citipitioglu et al. (2002) used this method with an Example Bolt which
had no nut connected (Figure 2.16), to determine the amount of displacement required that
correspondingly gave the desired tension in the bolt shank of the example bolt. Then this
amount of displacement was converted and calibrated to find the shank length of another
set of bolts which had nuts created as one piece. These bolts were then used in the
assembly of the beam-to-column connection model.

Figure 2.15 Continuum element beam-to-column connection model (Citipitioglu et al.,


2002)

29
Figure 2.16 Bolt pretension calibration curves used in the FE model (Citipitioglu et al.,
2002)

Reid & Hiser (2005) created FE models that simulated bolted joints of roadside structures
that allowed for slippage (Figure 2.17). They developed detailed modelling techniques for
bolted joints that slip in one direction when loaded in shear. To preload the bolts and thus
clamp the connection, two models were created and two pretensioning techniques were
applied, one for each. The models were called Discrete-Spring Based Clamping (DBC)
model and Stress Based Clamping (SBC) model. Both models were geometrically identical
but different in details of element types assigned. The former model had several parts
created with rigid solid elements, while the latter model had all parts created with
deformable solid elements (= continuum or brick elements).

The DBC model utilized a discrete spring element to preload the bolt and generate the
clamping force between the slip base plates. The slip base plates were created with
deformable solid elements, but the bolts, nuts, and washers were created with rigid solid
elements. Clamping forces were achieved with a single centrally located discrete spring
element. The spring was defined to act along the axis of the rigid body shaft, connecting
the centre of the bolt head to the centre of the nut. In order to produce a desired preload,
the spring was given an initial offset, which induced an initial force within the spring. The
slip base plates were then compressed slightly. (Only the base plates were compressed

30
because they were the only parts that were deformable.) A few iterations on the length of
the spring were required to achieve the desired preload.

For the SBC model, the objective of its creation was said to be to develop a bolt model that
would capture a higher sense realism regarding the bolt preload method and the physical
behaviour of the bolt shaft itself. The bolt part was created using deformable solid
elements with the material properties of steel, which when stretched through an initial
displacement would produce the desired preload in a manner consistent with an actual bolt.
Different from the design of the DBC model, but similar to that of Citipitioglu et al., the
bolt part here had the bolt head, bolt shaft, and nut created together in one piece. The bolt
shaft length was equal to the installed length which was the combined thickness of slip
base plates and washers between the bolt head and the nut. A manual operation was
required to enter a prestress amount that was then applied to every integration point of
every element of the bolt shaft. This caused the bolt shaft to shrink and compress the
deformable plates. A number of iterations were required to attain the desired preload after
the static equilibrium was reached.

Figure 2.17 General configuration of joint model (Reid & Hiser, 2005)

Hong, Yang & Lee (2001) did a parametric study of double angle framing connections
subjected to shear and tension. In their FE models, prestressing forces were applied to bolts
using a built-in loading option namely Pre-tension Section in the boundary module of
ABAQUS. This option is available in ABAQUS/Standard only and models were analysed
using ABAQUS/Standard. However, there were reasons that made ABAQUS/Standard not

31
applicable for the modelling work of this thesis in Chapters 5 and 6. The reasons will be
given in the chapters. ABAQUS/Explicit was applied instead.

2.9.2 Use of beam elements for economy


To create the second type of model in Chapter 5, the first type of model was modified. For
the sub-panel portions, the continuum elements were replaced with beam elements. The
connection portions and the beam-element sub-panel portions were linked together with
coupling constraints.

This arrangement was based on the modelling work of Prasad-Rao & Kalyanaraman
(2001) on the study of non-linear behaviour of lattice panel of angle towers, from the
section that dealt with single angles under compression. A model using mixed element
types was presented, as shown in Figure 2.18. The left and right portions of the angle were
modelled using beam elements. At the middle, over a short length of 0.2 times the total
length where plastification was expected to occur, the portion was modelled using (flat)
shell elements. At a transition between the beam element and the shell elements, rigid
elements (or coupling constraints) were used to connect them together.

Figure 2.18 Combined use of beam and shell elements for a compression member
(Prasad-Rao & Kalyanaraman, 2001)

2.10 Development of multi-panel tower models


Chapter 6 applied the economical and effective modelling approach and connection design
developed from Chapter 5 to create FE tower models with compression leg reinforced for
several panels continuously. Two tower models were created from the geometry layouts of
two actual towers from Hayward & Weare (1989) and Silva et al. (2005). The two created
models were elastic truss-type only (the members used one-dimensional axial-force-
32
carrying truss elements and elastic truss analysis was used). The models were used to
determine the minimum number of panels that a smaller-size tower model for FE study
should have, to allow us to probe the actual potential of the leg reinforcing method. This
was necessary because the number of panels relates to the magnitude of force increment on
the leg panels. The relationship is that the more panels a tower has, the smaller the
magnitude of force increment on the leg panels is. So if the FE tower model to be created
in this study had only a few panels, the force increment magnitude would be impractically
large, and the understanding obtained from the FE study would be incorrect or impractical.

Silva et al. (2005) assessed the current modelling and analysis methods of steel
transmission and telecommunication towers. The geometry layout of a tower used in their
study is given in Figure 2.19. Three modelling strategies to construct a tower were
investigated. The first strategy used truss finite elements exclusively, the second used beam
finite elements, and the third used combined truss and beam finite elements.

Figure 2.19 Tower geometry layout (Silva et al., 2005)


33
The first modelling strategy presented problems with respect to the loss of structural
continuity in some parts of the structure due to the presence of hinges associated with the
used truss finite elements. Although the in-plane behaviour could be considered flexible,
the out-of-plane behaviour disregarded the torsion and bending continuity presented in the
structure. The general concept is that, to overcome the out-of-plane problem, dummy bars
with low axial stiffness values (slender bars) needed to be introduced to the structure at
elevations along the height. These elevations were where the bracing members come to
join one another at the middle of the tower faces. The nodes that the bracing members
come to join, namely mid-set nodes (black nodes in Figure 2.20), are connected together
with four dummy bars which form a diamond shape. The bars restrain the displacements of
the mid-set nodes and help improve the structural torsional stiffness; otherwise the tower
would collapse by means of a simple structural mechanism. Lastly, another dummy bar is
provided inside the previous set of bars to complete a typical triangular truss system, which
was to prevent another possible in-plane horizontal static mechanism. Considering that this
traditional modelling strategyusing truss finite elements for every memberwould
require a large number of horizontal dummy bars to be added in order to make the analysis
run successfully, Silva et al. did not recommend it.

The second modelling strategy used beam finite elements for every member. This made
every member have effective length factor less than 1.0. Each member, or the whole tower
as a unit, became stronger than it should be. Therefore, this strategy was not considered in
the work of this thesis.

34
Figure 2.20 Dummy bars added to the structure: (a) Vertical plan; (b) Transverse plan,
section A-A (Silva et al., 2005)

The third modelling strategy used beam finite elements for the leg members and truss finite
elements for the bracing and secondary bracing members. The analysis could run
successfully with no need for the dummy bars, and the results were more appreciable than
the first modelling strategy. The study concluded that this mixed modelling strategy is able
to produce more realistic and trustworthy results.

The first and the third modelling strategies mentioned above were combined and adapted
to create two nine-panel tower models with the compression leg reinforced, in Chapter 6.

2.11 Conclusion
Although references on the exact topic of leg reinforcing for transmission towers were not
available, those selected and presented in this chapter were helpful. They provided the
basic knowledge of lattice tower structures and single-angle and compound compression
members. They also provided the experimental techniques, FE modelling techniques, and
recommendations which were adopted and modified for the study program of this thesis.

35
CHAPTER 3

3 One-panel experimental study Methodology


3.1 Introduction
The purpose of the one-panel experimental study was to investigate the effectivenesses of
three different connection designs or typesAlternating, Aligned and Cruciform
generally used in leg reinforcing design. In the early stage, these three terms indicated the
types and arrangements of the connectors used for all connection joints. The alternating
means that single plates were used as the connectors, one plate per joint, and placed on the
perpendicular planes alternately (Figure 3.1(a)). The aligned used single plates as well, but
they were placed on the same plane at every joint (Figure 3.1(c)). The cruciform means
that the specimens used angle cleats as the connectors, two cleats per joint (Figure 3.1(b)).
The specimens were designed to use three sub-panels, as was recommended by Temple et
al. (1986). Each of these connections was tested using both two bolts and three bolts per
connecting face.

3.2 Test series and modifications


3.2.1 Initial tests
In current construction practice when a connection type is selected for attaching the
reinforcing member, all connection joints are made identical, the top end joint and
intermediate joints will look the same. For example, if an alternating plate connection type
is selected, all the connecting plates will have the same length and use the same number of
bolts. This was used as the starting point for the experimental tests, i.e. all connection
joints of a specimen were designed to be identical.

36
Figure 3.1 Dimensions and arrangement of test specimens: (a) Alternating; (b) Cruciform;
(c) Aligned

3.2.2 Modification to top end connection


The results of the initial tests of the specimens with plate connection types (both
alternating and aligned) indicated that the top connection was very important in load
transfer. This led to a modification at the top connection joint of every subsequent
specimen, whereby the single plate connector at this joint was replaced with a cruciform
connection, using two angle cleats. All fasteners used up to this stage were galvanized
tower bolts, tightened snug-tight as they normally are in real construction. Tests in these
two stages were termed Set 1.

3.2.3 Bolt pretension


Following the modification at the top connection, another improvement was introduced to
increase friction force at contact surfaces of the component members and connectors at the
joints. Here, high-strength bolts were adopted as the fasteners instead of tower bolts. They
were tightened to reach the minimum tension force as specified in ANSI/AISC 360-05
(2005). It was expected that doing so would help reduce relative movements between the

37
component members induced by unfilled bolt holes, even though they were a standard size.
It was hoped that this could help transfer more force to the reinforcing member as well,
and consequently increase the total load-carrying capacity. Tests with tensioned high
strength bolts were termed Set 2.

3.2.4 Number of bolts at intermediate joints


One minor change also introduced was reducing the number of bolts used at the
intermediate joints, from three bolts per connecting face to two bolts. This was to test the
assumption that since the intermediate joints were shown in early tests to have lesser shear
force conveyed through them than the end joints, it may be more economical to use fewer
bolts just to provide adequate shear forces, without leading to a side effect of reducing the
load carrying strength of the specimens. This modification was made during Set 1 and 2
tests.

3.2.5 Special case of welded connections


To understand how effective the specimens can be when the slip (relative movements)
between the members is completely prevented, a set of specimens was constructed having
welded connections at the top and intermediate joints. As it was anticipated that shear
forces at the intermediate joints would be small, single plate connectors were applied at
these joints, while at the top joint, double angle cleats were still used.

3.3 Terminology used for test specimens


Designations used for each test specimen were based on 1) the type of connectors and their
arrangement for the intermediate joints, 2) connecting method, and 3) the testing set. For
example, Ali-2B-S1 was a specimen that had aligned plate connectors, with two bolts in
each connecting face of the plate, and was part of Set 1 tests. Detailed explanations of each
abbreviation are given below.

3.3.1 Connection types


The first design was called cruciform connection type. Each intermediate joint used two
angle cleats to connect the component members together, and therefore four angles
appeared at the cross-section. (Figure 3.1(b)) Its abbreviation is Cru.

38
The second design was alternating plate connection type, Alt. Single plates were used as
the connectors at the intermediate joints, one plate per joint, and placed on perpendicular
planes alternately. (Figure 3.1(a))

The third design was similar to the second, but the plate connectors were on the same plane
for every intermediate joint, and hence it was called aligned plate connection type, Ali.
(Figure 3.1(c))

3.3.2 Connecting methods


Abbreviation B means that that specimen used bolts as the connection method, and the
number in front of it is the number of bolts used per connecting face at the intermediate
joints.

Abbreviation W means that the specimen used welded connections.

3.3.3 Testing sets


The specimens that used galvanized tower bolts tightened snug-tight were part of test Set
1, abbreviated as S1. Those specimens using high-strength bolts tightened fully tensioned
and those with welded connections were Set 2 specimens, abbreviated as S2. It must be
noted that the angle steels of these two sets were ordered from different manufacturers, so
they had slightly different average yield strength values, which are detailed later.

3.4 Test specimens, assembly, and apparatus


3.4.1 Specimens general details
Each test specimen consisted of one main member and one reinforcing member. They were
each made of 75 75 6 equal angle (EA) steels, grade 300.

In the actual configuration of a reinforced tower leg at the top end joint where the
reinforcing member stops, there is only the existing leg member above this joint and it
alone carries all the force. The force coming from the upper portion of the tower through
the existing leg will start spreading partly down to the reinforcing member at this first
joint. To imitate such a condition, the main members of the specimens were cut slightly
longer than the reinforcing members. The cut length of the main members was 1180 mm,
giving an approximate slenderness ratio of 80 (a quite common value for the unsupported
length of tower legs). The reinforcing members were cut 10 mm shorter, or 1170 mm in
length. Loading was applied only to the longer member.
39
Following the recommendation given in the study of Temple et al. (1986), the test
specimens were designed to use two intermediate connection joints. (Figure 3.1 and Figure
3.4) Each specimen was divided into three sub-panels. The top and middle sub-panels were
332 mm long and had a local slenderness ratio a / ri equal to 22.6, while the bottom sub-

panel was 382 mm long with a / ri = 26.0. The total length L measured from the bottom
supports centre to the main members top end was 1130 mm.

For connectors, generally when the connections are cruciform-typed, it is quite a common
practice in industry to use connectors of the same section size and thickness as the
component members. Therefore, for the Cru specimens in this experiment the angle cleats
were cut from the same stocks of angles used for the component members, at 150 mm in
length each. For the Ali or Alt specimens, the plate connectors were cut from 150 6 mm
flat bar, grade 300, 150 mm long. These dimensions were checked against the minimum
edge and spacing distance requirements of AS 3995 (1994) when M12 bolts are used.

As mentioned earlier in this section, before the first modification at the top connection took
place, all connections of a specimen (top, first intermediate, and second intermediate) were
made alike, depending on what type of connection was being considered. Each connecting
face used three M12 bolts. But when the modification was introduced, double angle cleats
were used exclusively as the connectors at the top connection for every specimen of all
connection types. At the top connection, three bolts were applied at each connecting face,
totalling 12 bolts over the connection. Details of the connections and number of tests of all
groups of specimens are summarized in Table 3.1.

40
Table 3.1 Connection types and components
Set 1 Tests
Connection Pt-Ali-3B-S1 Pt-Alt-2B-S1 Pt-Alt-3B-S1
No. From Connectors Bolts/ Connector Bolts/ Connector Bolts/
Top Made From Qnt. Face Made From Qnt. Face Made From Qnt. Face
1 PT 1 3 PT 1 2 PT 1 3
2 PT 1 3 PT 1 2 PT 1 3
3 PT 1 3 PT 1 2 PT 1 3
No. of Tests 2 1 2
Connection Ali-2B-S1 Ali-3B-S1 Alt-2B-S1
No. From Connectors Bolts/ Connector Bolts/ Connector Bolts/
Top Made From Qnt. Face Made From Qnt. Face Made From Qnt. Face
1 AC 2 3 AC 2 3 AC 2 3
2 PT 1 2 PT 1 3 PT 1 2
3 PT 1 2 PT 1 3 PT 1 2
No. of Tests 2 2 2
Connection Alt-3B-S1 Cru-2B-S1 Cru-3B-S1
No. From Connectors Bolts/ Connector Bolts/ Connector Bolts/
Top Made From Qnt. Face Made From Qnt. Face Made From Qnt. Face
1 AC 2 3 AC 2 3 AC 2 3
2 PT 1 3 AC 2 2 AC 2 3
3 PT 1 3 AC 2 2 AC 2 3
No. of Tests 2 4 4
Set 2 Tests
Connection Alt-2B-S2 Cru-2B-S2 Alt-W-S2
No. From Connectors Bolts/ Connector Bolts/ Connector Bolts/
Top Made From Qnt. Face Made From Qnt. Face Made From Qnt. Face
1 AC 2 3 AC 2 3 AC 2 -
2 PT 1 2 AC 2 2 PT 1 -
3 PT 1 2 AC 2 2 PT 1 -
No. of Tests 3 3 2
Note: Pt - Plate connectors were used at every connection.
Ali - Aligned plate connection type with modified top connection. Plate connectors were
used at intermediate connections, and put on the same plane.
Alt - Alternating plate connection type with modified top connection. Plate connectors
were used at intermediate connections, but put on perpendicular planes.
Cru - Cruciform connection type. Double angle cleats were used at intermediate
connections.
B - Bolted connection was used.
nB - Number of bolts used per connecting face for intermediate connections.
W - Welded connection was used.
S1 - Set 1; Angle steels had fy = 328 MPa. M12 tower bolts, Grade 5.6, were used and
tightened to 50 N.m torque.
S2 - Set 2; Angle steels had fy = 354 MPa. M12 Grade 8.8 bolts were used and tightened
to 150 N.m torque.
AC - Angle cleat made of 75 75 6 EA, 150 mm long.
PT - Plate connector having a 150 150 6 mm dimension.

41
3.4.2 Mechanical properties of component members
Mechanical properties of the steel angles were determined by the tensile testing method
according to Australian Standard, AS 1391 (1991). The angles Set 1 and Set 2 had yield
stresses f y equal to 328 MPa and 354 MPa respectively. (Table 3.2) Both sets had elastic

modulus E of 205,000 MPa, and therefore this value was adopted for both sets.

Table 3.2 Strength properties of Set 1 and Set 2 steels from tensile coupon tests

Set 1 Angles Set 2 Angles


Coupon Test fy fu Coupon Test fy fu
Piece No. (MPa) (MPa) Piece No. (MPa) (MPa)
1 324 464 1 349 512
2 328 464 2 354 508
3 328 453 3 361 494
4 331 453 4 348 501
5 360 508
6 354 507
Average 328 459 Average 354 505

3.4.3 Bolts and bolt tensioning


In Set 1 of testing, all bolts used at the top and intermediate joints were galvanized tower
bolts to AS 1559 (1997), grade 5.6 ( f y = 320 kN and f u = 480 kN). They were selected

because it was considered to be most appropriate to use the same bolt type as that used on
real towers. Since the tower bolts are medium-strength grade, not high-strength, they had
to be tightened snug-tight only. Here, a 50 Nm torque was applied to tighten the M12
tower bolts. The connections in Set 1 were categorised as being shear-bearing type.

In Set 2 of testing, grade 8.8 high-strength bolts to AS 1252 (1996) were used, tightened to
a 150 Nm torque. This torque was expected to provide an equivalent tension of 62.5 kN to
each bolt (see the next paragraph). This tension is larger than 70% of the minimum yield
strength of the boltthe requirement of ANSI/AISC 360-05 for a friction-type
connectionor in number equal to 52 kN. Doing this ensured that the connections in Set 2
had friction-type condition.

When a torque wrench is used to tighten a bolt, the tension in the bolt can be approximated
by the following relationship:

Tension = Torque / (K diameter)


42
where K is a friction factor that depends on the bolts thread and nut condition. The most
commonly used K values are 0.22 for zinc plated or galvanized bolts, and 0.2 for plain
finished bolts. According to this formula, the M12 galvanized tower bolts of the specimen
Set 1, tightened with a 50 Nm torque would carry a tension of 18.9 kN. Similarly, the high-
strength bolts of the specimen Set 2 tightened with a 150 Nm torque should carry a tension
of 62.5 kN.

Bolt holes were made by using a hydraulic punching machine for convenience and speed.
It gave an average hole diameter of 13.5 mm, which is a standard size suitable for M12
bolts.

3.4.4 Bottom support


In site construction when a reinforcing member is to be added to an existing leg, that leg
will be held stationary by a temporary support system. Then the concrete cap casing
around the existing leg at the tower base will be removed, the reinforcing member added,
and the concrete cap recast again. This practice provides a fixed-end condition to the
bottom ends (base) of both members. To achieve a fixed-end condition for the test
specimens, the study used a supporting block as shown in Figure 3.2 as the bottom support.
It consisted of four 16-mm-thick mild-steel vertical plates, arranged in a star shape to fit
the specimens cross-section. They were butt welded together at corners and to the 20-mm-
thick base plate. Eight M16, grade 8.8 bolts (two on each vertical plate) were used to
transfer force between the bottom support and the component members through the bolts
shear capacity. The bolt holes were marked on the vertical plates at the positions that
would provide concentric force to each component member, which were half width, 37.5
mm, from the corner. The supporting block appeared to work satisfactorily; the bolts size,
strength, number and positions on the plates, together with the vertical plates stiffness
against bending, could provide an absolute fixity to the bottom ends of the component
members. As the arrangements of the specimens with plate connection type and cruciform
type were different, in terms of the space between the component members, two supporting
blocks were fabricated. Figure 3.2 shows the general view of a supporting block when
installed.

43
Figure 3.2 Bottom support of plate connection type specimens (general views)

3.4.5 Top support


It was suspected prior to testing that the stiffness of the top end connection joint was
important for the leg strengthening purpose. The main member was designed as being pin-
connected at its top end so as to allow it to twist or rotate freely (thus modelling the
situation in practice), and at the same time force the top connectors to more easily show
their ability in providing additional restraint against those movements. Figure 3.3 shows
the top support used in the study. To allow rotation and/or twisting to occur spontaneously,
the top support was designed to have a semi-spherical shape. Force transfer occurred at the
touching surface between the supports flat side and the main members top end. The
radius of the support was approximately 110 mm. Since the rotating point of the support
was at the touching point, the radius distance was not taken into account when calculating
the total length of the main member.

44
Figure 3.3 Semi-spherical top support and its parts

3.4.6 Measurement of internal forces


To measure internal axial forces in the specimens, 5-mm-long strain gauges were placed at
the middle of each sub-panel, on the outer faces of the component members. For Set 1
specimens, four gauges were applied to each section, one on each face. (Figure 3.4(a)) For
Set 2 specimens, because of the improvement attempted by increasing the bolt tightening
force to 150 Nm, a more accurate force measurement was desired. Here, eight strain
gauges were used per section, two on each face. (Figure 3.4(c)) It is important to note that
the strain gauges used had the manufacturers specified acceptable displacement of 10%,
or 0.05 mm only. Raw strain data showed that this size of strain gauges may have been a
bit too smallsome strain gauges at the first sub-panel of some specimens became
overstressed when load underwent close the failure load of the specimens. However, the
strain readings obtained before that were still usable, and therefore, strain gauges of the
same size were used throughout the experiment. Figure 3.4(b) and (d) illustrate the
locations and numbers of the strain gauges on the main leg of a specimen in Set 1 and Set 2
respectively.

45
Figure 3.4 Locations and numbers of strain gauges on (a) cross-section of the first sub-
panel of Set 1 specimens, (b) profile of the main leg of Set 1 specimens, (c) cross-section
of the first sub-panel of Set 2 specimens, and (d) profile of the main leg of Set 2
specimens
46
3.4.7 Loading
A Baldwin loading machine was used in this experiment. Load was applied by moving the
lower cross head of the machine upward, while the upper cross head was stationary. The
speed of the lower cross head was maintained at approximately 0.005 strain per second
throughout the loading period. This speed was slow enough not to disturb the specimens
load carrying behaviour when weak spots started to deform in their inelastic range
allowing the individual members to respond to the internal force smoothly without instant
impact effects. The applied load was measured by the 1000 kN built-in load cell
underneath the bottom crosshead. Outputs were recorded by data logger. In Figure 3.5, the
picture on the left shows the controlling panel of the Baldwin loading machine, and the
picture on the right shows the data logging equipment. Figure 3.6 shows a cruciform
specimen under test in the loading machine.

Figure 3.5 Baldwin loading controller and data logging equipment

47
Figure 3.6 A cruciform specimen under test in the loading machine

3.4.8 Displacement measurement


In this study, only vertical displacement was measured. A displacement transducer was
used to measure the upward movement of the lower crosshead. Lateral displacements (or
deflections) of the specimens were deliberately not measured as the axial capacity of the
members was the point of interest of the study. Plots of vertical displacement versus
applied load were sufficient and reliable in representing the true effectiveness of each
specimen and connection design.

The load vs. vertical displacement curves from the test results were highly valuable to
verify the finite element models and modelling techniques developed in Chapter 5.

3.5 Test procedure


After a specimen was completely assembled and strain gauges were attached, the specimen
was put onto the bottom support block. Eight M16 bolts were used to connect the
48
specimens to the support block. The other bolts on the intermediate and top connections
were then tightened to the desired torque. This procedure was carried out with great care
for Set 2 specimens since the effect of bolt tightening force on the ultimate strength was of
particular interest.

At the top end, a 1-mm-thick zinc plate was placed between the top end of the main
member and the semi-spherical top support. This was to help achieve a good distribution of
load across the cut surface of the main member, in case the cut was slightly uneven.

The wires from the strain gauges and the displacement transducer were hooked to the
channels on the data logger. The readings at this stage before loading was started were
checked for errors. When no errors existed, the logger was set to start recording data. The
specimen was then loaded by moving the lower crosshead of the loading machine up at an
approximate speed of 0.005 strain/second from the beginning to failure state.

The data (load, vertical displacement, and strain values) were recorded automatically at
every five seconds. The data were exported to an Excel file for use in analysis of the test
results.

3.6 Single angle tests


To provide strength references for the reinforced specimens, single angle tests were
required. As there were two sets of steels due to different batches of supplied materials,
two sets of single-angle tests were used. They were termed SA-S1 for single angle Set 1,
and SA-S2 for single angle Set 2. The angles used here were 1180 mm in length and had
four 17.5-mm-diameter holes punched at the bottom end for M16 bolts, similar to the main
leg of the reinforced specimens. No other treatments were provided. The same top and
bottom supports were used, and a single angle was put on the bottom support at the
position of the main leg which was designed for a concentric loading condition. Four M16
bolts were then added to fasten the angle to the bottom support. After that, load was
applied until the angle member failed. Only the applied load and vertical displacement
were recorded. The results of these tests are given in Chapter 4.

49
3.7 Example of determination of internal forces
An example is given in Table 3.3 to demonstrate how the internal forcesmember forces
at sub-panels and shear forces at connectionswere determined from the set of strain
gauge readings.

In Table 3.3, the second column lists a set of recorded strain readings of Cru-2B-S2, Test 3
when the applied load was at 237 kN. Gauges No. 1 4, 9 12, and 17 20 were on the
main member at the top, middle, and bottom sub-panels (Figure 3.4), while Gauges No. 5
8, 13 16, and 21 24 were on the reinforcing member at the sub-panels of the same
order. These strain values were then averaged to determine the representative strain values
on the members. This simple averaging method should be a relatively sensible method if
we assume that strain or stress distribution across a section of a member is perfectly linear
at any elevation, especially at the middle of sub-panel where the strain gauges were placed.
By averaging, an inclined stress distribution will become an even one.

Next, the axial forces were determined using the basic elastic stress-strain relationship
/ = E , or N = AE , where A = 867 mm2, the cross-sectional area of a 75 75 6
equal angle, and E = 205,000 MPa. In Chapter 4, the results were presented by dividing
the axial forces of the sub-panels determined by the relationship just explained with the
cross-sectional yield strength of a component member ( N i / Py or N i* / Py ). The values

were lower than 1.0, and this indicated that the axial forces were in elastic range and
determining the axial forces using elastic stress-strain relationship was applicable.

50
Table 3.3 Calculation example of internal forces for Cru-2B-S2, Test 3

Applied Load = 237 kN


Gauge Strain Avg. Strain Member Forces Total Force Shear at Joint Avg. Shear
-6 -6
No. (x 10 ) (x 10 ) at Sub-Panel at Sub-Panel Det. From at Joint
(kN) (kN) Ni or Ni* (kN) (kN)
st
G1 1050 G 1-4 N1 1 or Top Sub- V1 V1
G2 1164
52
G3 1102 1145 188
(240-188)
G4 1263
G5 437 G 5-8 N1* 240 (188+52) V1* 52
G6 192
G7 436 314 52 52
G8 189
G9 861 G 9-12 N2 2nd or Mid. Sub- V2 V2
G10 1061
31
G11 858 957 157
(188-157)
G12 1048
G13 538 G 13-16 N2* 233 (157+76) V2* 28
G14 363
24
G15 554 463 76
(76-52)
G16 396
G17 738 G 17-20 N3 3rd or Bot. Sub- V3 V3
G18 916
26
G19 698 797 131
(157-131)
G20 837
237
G21 750 G 21-24 N3* V3* 28
(131+106)
G22 518
30
G23 761 645 106
(106-76)
G24 552

Note: Refer to Figure 3.4 for gauge locations


G1 = strain gauge No. 1
N1 = axial force on the main member in top sub-panel
N1* = axial force on the reinforcing member in top sub-panel
V1 = shear force at joint 1 determined from axial force on the main member
V1* = shear force at joint 1 determined from axial force on the reinforcing member

3.8 Conclusions
There exist three connection designs used in leg reinforcing work for transmission towers.
They can be termed Cruciform Angle Cleat connection, Alternating Plate connection, and
Aligned Plate connection. Through common sense, it is understood that the Cruciform
design should be the most effective design. However, to confirm the assumption, all three

51
designs were tested in this experimental study. This chapter has explained the ideas that led
to the configuration details of the test specimens and end support conditions.

A one-panel reinforced specimen consisted of two longitudinal membersmain member


and reinforcing member. The L / r ratio of each was approximately 80. With connectors,
the whole length (or the whole combined panel) was divided into three portions (or three
sub-panels). End support conditions created were Pinned for the top end and Fixed for the
bottom end, which was the same for every specimen.

At the same time, to investigate whether pretensioning the bolts could have a significant
improvement on the effectiveness of the reinforced legs, the specimens were also divided
into two setsSet 1 and Set 2. Bolts were tightened snug-tight in Set 1 and fully tensioned
in Set 2.

A number of single angle tests were conducted to provide the strength reference point for
the reinforced specimens. In all tests, three basic resultant valuesthe applied load, the
vertical displacement, and the strain values of each sub-panelwere measured for
evaluation.

52
CHAPTER 4

4 Experimental study Results and discussion


4.1 Introduction
This chapter will discuss the results of the cruciform-angle, alternating-plate, and aligned-
plate, Set 1 and Set 2 test specimens. The first half of the chapter will discuss the observed
failure pattern of each specimen and the possible causes for such outcomes. Some
conclusions will be made from this physical observation. In the second half, an evaluation
of the effectiveness of the connection designs will be made using internal forces (sub-panel
forces and connection shear forces) in relation to the applied load.

4.2 Test result summary


The failure loads and percentages of strength increases of the reinforced specimens
compared with the single-angle tests are summarised numerically in Table 4.1 and
graphically in Figure 4.1. See the notes of Table 4.1 for the strengths of single-angle tests.

53
Table 4.1 Summary of failure loads and strength increase percentages of Set 1 and Set 2
tests

Set 1 Tests Set 1 Tests


Specimen Test Failure Strength Std. Specimen Test Failure Strength Std.
category No. Load (kN) Increase % Div. category No. Load (kN) Increase % Div.
1 343 29.4 (1) 1 280 5.7
Pt-Ali-3B-S1
Ali-2B-S1(1) 2 311 17.2 2 292 10.2
(1)
Avg. 327 23.3 22.8 Pt-Alt-2B-S1 1 284 7.2
1 312 17.7 1 284 7.2
Pt-Alt-3B-S1(1)
Ali-3B-S1(1) 2 339 28.0 2 277 4.5
Avg. 326 22.9 19.2 Avg. 283 6.9 5.6
1 320 20.6
(1)
Alt-2B-S1 2 317 19.6
Avg. 318 20.1 1.9 Set 2 Tests
1 347 30.9 Specimen Test Failure Strength Std.
(1)
Alt-3B-S1 2 342 29.1 category No. Load (kN) Increase % Div.
Avg. 345 30.0 3.5 1 356 21.9
1 328 23.6 2 366 25.3
Alt-2B-S2(2)
2 384 44.7 3 368 26.0
Cru-2B-S1(1) 3 349 31.9 Avg. 363 24.4 6.4
4 352 32.7 1 392 34.2
Avg. 353 33.2 23.1 2 388 32.9
Cru-2B-S2(2)
1 345 30.1 3 389 33.2
2 400 50.8 Avg. 390 33.4 2.1
(1)
Cru-3B-S1 3 348 31.3 1 405 38.7
(2)
4 367 38.6 Alt-W-S2 2 417 42.8
Avg. 365 37.7 25.2 Avg. 411 40.8 8.5

Notes: (1) % of strength increase was obtained by comparing the failure load with the average
ultimate capacity of 265 kN from three single-angle tests of steel Set 1.
(2) % of strength increase was obtained by comparing the failure load with the average
ultimate capacity of 292 kN from four single-angle tests of steel Set 2.

54
Note: Those specimens connected with dotted line arrows have the same connection type, but
different bolt pretensioning forces.

Figure 4.1 Summary of strength increase percentages of the specimens of different


connection types and bolt pretensioning forces from Sets 1 and 2 steels

4.3 Physical observation during tests


4.3.1 Plate connectors at all joints, Set 1
The test results showed that all five specimens of this connection category failed by the
same failure mode, which was twisting of the main member at the point where it extended
above the reinforcing member. A typical example is illustrated in Figure 4.2 which shows
the failure pattern of Pt-Ali-3B-S1, Test 2. The main members extreme top end twisted
away from the plate connector. The unconnected face showed severe deformation, while
on the connected face the deformation stopped at the level of the top bolt. No twisting,
flexural buckling, or bending appeared anywhere else on the component members. All
plate connectors, including the top one, showed no sign of deformation, and the bolts were
still intact without shear damage. The strength increases varied from +4.5% to +10.2%,
with an average of +6.9%.

55
Figure 4.2 Failure mode of Pt-Ali-3B-S1, Test 2

The failure mode, due mainly to twisting of the main members top end, can be explained
as the result of using a single plate as the top end connector. In this arrangement, the top
plate connector provided direct restraint against twisting and vertical deformation to the
main members connected face only, but provided no restraint to the unconnected face.
Since the unconnected face had stiffness of only half of the connected face, when loaded it
buckled and failed first.

Considering the percentage of strength increases in Table 4.1, the values obtained are very
low, compared with those of all other cases that used double angle cleats as the top end
connectors. This resulted from the use of the semi-spherical top support in which when the
unconnected face started to twist or deform, the support rotated itself more onto the
unconnected face, making the unconnected face absorb more load. Once the unconnected
face reached the plastic state and failed, the connected face was left to carry all the applied
force alone. It therefore failed almost immediately thereafter. However, since in the real
condition the existing leg members run continuously to the towers top end, this failure
mechanism may not happen, or if it did happen the unconnected face may not be this
vulnerable, and the strength increase would possibly be higher. Nevertheless, the study has
demonstrated a very important precaution against using a single plate as the connector at
the top end joint, that it should not be used if a significant and reliable strength increase is
sought after. It is highly recommended that reasonably long double angle cleats be used at
this position.
56
Following the completion of this group of tests, the top connection was modified to a
cruciform with angle cleats and three bolts per connecting face for all subsequent tests.

4.3.2 Aligned plate connection, Set 1


4.3.2.1 Observations

Ali-2B-S1, Test 1 The specimen failed by moderate twisting of the main member over
the top sub-panel, below the top connection, in the direction away from the top plate
connector. (Figure 4.3(a)) Only the top portion, from the twisting path up, showed a small
bend about the built-up (or compound) sections strong axis. (See note in the next
paragraph.) No twisting or bending appeared anywhere else. The plate connectors showed
no deformation or twisting, no bolts failed by shear force. The failure load was 343 kN.

Note: As has been explained that the main member was intentionally cut 10 mm longer
than the reinforcing member and this longer part was at the top end of the specimen. When
loaded, only the top end of the main member touched the top support. This arrangement
induced loading eccentricity to the top end (or top portion) of the specimen. The
eccentricity in this place was the distance from the centroid of the main members section
to the centroid of the built-up section, and it was perpendicular to the major axis. The
eccentricity was large enough to make the top portion of the specimen bend or experience
flexural buckling about the major axis.

Ali-2B-S1, Test 2 The specimen had the same failure mechanism as that of Ali-2B-S1,
Test 1, except the main member twisted in the direction toward the top plate connector.
The failure load was 311 kN.

Ali-3B-S1, Test 1 The specimen failed by small twisting of the main member over the
top sub-panel, in the direction towards the plate connector. (Figure 4.3(b)) The top portion
showed a very small bend about the built-up sections strong axis. The two plate
connectors were intact, no bolts failed, and no twisting appeared anywhere else. The
specimen failed at 312 kN.

Ali-3B-S1, Test 2 The specimen failed with the same mechanism as that of Ali-3B-S1,
Test 1, except that the main member had a small twist in the direction away from the top
plate connector and the top portion had a small bend. The failure load was 339 kN.

57
Figure 4.3 Deformed shapes of (a) Ali-2B-S1, Test 1, and (b) Ali-3B-S1, Test 1

4.3.2.2 Discussion

It was observed that all four specimens using aligned plate connection type failed by one
failure mode only, namely twisting (or torsion) of the main member over the top sub-panel
below the top connection joint. This twisting also caused a small bend to the top portion of
the specimen.

Looking more closely, it was seen that when the main member twisted in the direction
away from the top plate connector, as in the cases of Ali-2B-S1, Test 1 and Ali-3B-S1,
Test 2, their corresponding failure loads were higher. It was opposite in the case of Ali-2B-
S1, Test 2 and Ali-3B-S1, Test 1, where the main member twisted in the opposite
direction.

Therefore, it would be possible for us to conclude that, for a starred reinforced leg having
aligned plate connection type, the twisting of the main member is likely to be the
governing failure mode, whereby the twisting direction also has some influence on the

58
ultimate carrying strength. The number of bolts used seems not to be an important factor as
far as they can provide adequate shear force.

4.3.3 Alternating plate connection, Set 1


4.3.3.1 Observations

Alt-2B-S1, Test 1 This specimen failed by small twisting of the main member over the
top sub-panel, in the direction away from the top plate connector. (Figure 4.4(a)) The top
portion showed a small bend. Plate connectors were intact and no bolts failed. The
specimen failed at 320 kN.

Alt-2B-S1, Test 2 Failed with the same pattern as of Alt-2B-S1, Test 1. The maximum
capacity was 317 kN.

Alt-3B-S1, Test 1 The specimen failed by flexural buckling of the built-up section about
its strong axis, with a plastic hinge formed at the top plate connectors level. (Figure
4.4(b)) The main member bent inward at the plastic hinge and its unconnected face twisted
severely, while the reinforcing member bent outward, but showed no sign of twisting of its
section. As a part of the plastic hinge, the top plate connector was deformed out of plane
considerably. Nevertheless, no bolt-shear failure was observed. The failure load obtained
was 347 kN.

Alt-3B-S1, Test 2 The specimen failed in the same manner as that of Alt-3B-S1, Test 1.
Its failure load was 342 kN.

59
Figure 4.4 Deformed shapes of (a) Alt-2B-S1, Test 1, and (b) Alt-3B-S1, Test 1

4.3.3.2 Discussion

The results showed that both the failure loads and the failure mechanisms for each of the
two cases, Alt-2B-S1 and Alt-3B-S1, were very consistent. However, due to the limited
number of test specimens in this set, it may not be totally correct to conclude that if
alternating plate connection is adopted, using three bolts per connecting face can make the
reinforced leg behave more effectively than using two bolts per connection face, and lead
to flexural buckling failure of the member.

Instead, it would be more appropriate to say that when an alternating plate connection type
is used for leg reinforcement, two failure modes are likely to happen. One is twisting of the
main member, and the other is flexural buckling of the built-up section about its strong
axis.

Although the second mode appears to have a higher probability to happen with alternating
plate connection type than with aligned plate connection type, it may not make a
significant difference in terms of the effectiveness. As can be noticed from the test results,
60
even though Ali-2B-S1, Test 1 and Ali-3B-S1, Test 2 failed by a different failure mode
from that of Alt-3B-S1, Test 1 and Test 2, they still had comparable ultimate strengths.

4.3.4 Cruciform angle connection, Set 1


4.3.4.1 Observations

Cru-2B-S1, Tests 1, 3, and 4 These three specimens failed by large twisting of the main
member over the top sub-panel, below the top intermediate connection, in a clockwise
direction. (Figure 4.5(a)) The twisting caused a moderate bend to the top portion of the
specimen only, while the lower portion was still straight. Angle cleats at the first
intermediate connection and elsewhere were intact, and no bolt-shear failure was detected.
The specimens failed at 328, 349, and 352 kN respectively.

Cru-2B-S1, Test 2 From observation during the test, the specimen began by deflecting
laterally in the weak axis direction (bending about the strong axis) as the load increased,
with the centre of curvature located at about mid-length of the specimen. As the load
continued to increase, twisting occurred to the main member over the top sub-panel, and
the specimen failed consequently. Figure 4.5(b) shows the specimen at failure having a
small overall deflection and moderate twisting on the main section in the clockwise
direction. No angle cleats were impacted and the bolts were undamaged. The specimen
failed at 384 kN.

Cru-3B-S1, Tests 1, 3, and 4 These three specimens failed in the same manner as those
of Cru-2B-S1, Tests 1, 3, and 4, except that the twisting direction was counter-clockwise
for Test 1 specimen, and clockwise for Tests 3 and 4. The failure loads were 345, 348, and
367 kN respectively.

Cru-3B-S1, Test 2 The specimen failed by flexural buckling of the built-up section
about the strong axis. The curvature was obvious by its centre (plastic hinge) located
immediately below the first intermediate connection. (Figure 4.5(c)) Additionally, careful
observation revealed that there was some local buckling on the main member below the top
connection as wellthe two faces of the main member squeezed towards each other very
slightly. The angle cleats at the first intermediate connection showed some deformation,
but only by a small degree. No bolt-shear failure was observed. The failure load was 400
kN.

61
Figure 4.5 Deformed shapes of (a) Cru-2B-S1, T1, (b) Cru-2B-S1, T2, and (c) Cru-3B-S1,
T2

4.3.4.2 Discussion

Similar to the case of alternating plate connection type, the same two failure modes existed
here. The most common mode was twisting of the main member, and the less common
mode was flexural buckling of the built-up section about its strong axis.

To determine the effect of the number of bolts, the average strength values were calculated.
(Refer to Table 4.1) The values show that when the cruciform (double angle) connection
type was used, the specimens that used three bolts per connecting face were more effective
than the specimens that used two bolts per connecting face, by approximately 4% (365 kN
from Cru-3B-S1 to 353 kN from Cru-2B-S1). If those specimens which failed by flexural
buckling (Cru-2B-S1, Test 2 and Cru-3B-S1, Test 2) are considered as rare, and then
discounted from the calculation, the new average strength values still provide the same
result. Therefore, it is possible to conclude, with more confidence than in the case of plate
connection types, that the effectiveness of a reinforced leg using cruciform connection type

62
can be increased somewhat by applying more bolts per connecting face. However, whether
such an application is cost effective is beyond the scope of this study.

4.3.5 Alternating plate connection, Set 2 (pretensioned bolts)


4.3.5.1 Observations

Alt-2B-S2, Tests 1, 2 and 3 All three specimens had the same failure mode, which was
bending about the strong axis of the built-up section. The centre of the bend was located at
the upper intermediate connection. (Figure 4.6(a)) At this location a slight local buckling
showed on the unconnected face of the main member, buckling away from the connected
face. (Figure 4.6(b)) No traces of local buckling or twisting were found anywhere else. The
maximum strength of the specimens were 356, 366, 368 kN respectively. The average
strength increase was +24.4%.

Figure 4.6 Deformed shape of Alt-2B-S2, Test 1. (a) Small overall bending. (b) Slight local
buckling of the main members unconnected face at the upper intermediate connection.

63
4.3.5.2 Discussion

In this case where the intermediate connectors were flat plates in alternating arrangement
with bolts pretensioned, only one failure mode was found. The failure was originated by
local buckling of the unconnected face of the main member at the upper intermediate joint,
which led to flexural bending failure of the whole specimen.

With bolts pretensioned, the capacities of the specimens should have been larger than the
obtained values. But since flat plates were used at the intermediate connectors, the
unconnected face of the main member became the weak portion and the unconnected face
buckled at the middle of its free length. Here the clear length term was the distance
between the top connection joint and the lower intermediate joint. As a result, the average
strength increase of this design (+24.4%) was not much higher that that of its counterpart,
case Alt-2B-S1, having bolts snug-tightened (+20.1%). So, if flat plates are applied as the
(intermediate) connectors, pretensioning the bolts (or making friction-type connections)
would not yield a significantly higher strength benefit than without pretensioning (or
making snug-tight connections).

4.3.6 Cruciform angle connection, Set 2


4.3.6.1 Observations

Cru-2B-S2, Tests 1, 2, 3 All three specimens experienced twisting failure mode. Both
faces of the main member immediately below the top connection showed a slight twist.
(Figure 4.7) This caused the top portion of the specimens to bend by a small degree. This
was the only failure pattern found; where the four specimens of Cru-2B-S1 showed two
failure patterns. No other traces of twisting or local buckling were found anywhere else.
The maximum capacities of the specimens were 392, 388, and 389 kN respectively,
making the average strength increase of +33.4%.

64
Figure 4.7 Deformed shape of Cru-2B-S2, Test 1. Left: A small bent at the top portion,
beginning from the first sub-panel upward. Right: A slight local buckling immediately below
the top connection

4.3.6.2 Discussion

With two angle cleats at each joint, the clear length mentioned in the case of Alt-2B-S2
was reduced by half to sub-panel length, which enabled the specimens of this case (Cru-
2B-S2) to carry higher loads. The twist of the first sub-panel of the main member indicated
that the amount of force passing it should be equal to or almost equal to its section yield
strength.

From seven specimens of cases Cru-2B-S1 and Cru-2B-S2, six of them (three from each)
experienced twisting failure mode. Based on this mode, their average strength increases
were 29.4% and +33.4% respectively, which were about the same level but showed some
improvement. These two coincidences (strength and failure pattern) were likely to be
because both cases used bolts as the fasteners. Assuming this assumption is correct, it
means that when bolted connections are used, whether they are made friction-type (bolts
tensioned to 0.7 times their minimum tensile strength) or bearing-type (bolts tensioned
65
snug-tight), the slip can still take place, and the reinforced legs will have the same level of
performance. However, tensioning the bolts brings consistence to results.

4.3.7 Welded alternating plate connection


4.3.7.1 Observations

Alt-W-S2, Tests 1 and 2 Both specimens had the same failure pattern. Both main and
reinforcing members appeared to twist together as one unit, at the location just below the
top connection. (Figure 4.8) Bending of the top portion was almost unnoticeable. Their
strengths were 405 and 417 kN, or equal to an average strength increase of +40.8%.

Figure 4.8 Deformed shape of Alt-W-S2, Test 1

4.3.7.2 Discussion

The twisting of both main and reinforcing members as one unit should be because the
connections were welded type. For the same reason, there was obviously no slip between
the component members. Compressing the main member at the top end means also having
to compress heavily on the reinforcing member at the same time too. As a result, this

66
design gained a very high strength increase (+40.8%) although the intermediate connectors
were flat plates, compared with that of Cru-2B-S2 (+33.4%) whose intermediate
connections used angle cleats. However, it may not be practical in either economy or safety
aspect to do welding on site (with towers being in service and welds being required at
heights above the ground.)

4.3.8 Conclusions from physical observations


There were two basic conditions that made the specimens with cruciform connection type
superior to those with aligned or alternating plate connection type. First, the cruciform
arrangement allowed bolts to be applied on all four faces of the built-up section. This
condition helps facilitate the distribution of force from the main member to the reinforcing
member, and helps minimize the concentration of force on the connecting faces around the
joints. In addition, it allows the load path to stay on the same face throughout the member
without having to go across the corner from one connection joint to another like in the case
of alternating plate connection type. Second, the angle connectors put in cruciform shape
are effective enough to prevent free twisting of the component members at the connections
in both clockwise and anti-clockwise directions. As happened in Alt-3B-S1, Test 1 and
Test 2, the unconnected face of the first intermediate joint became the weak spot that later
induced severe flexural buckling to the specimens. In summary, the cruciform angle
connection type is able to create a better integrity to the built-up reinforced legs.

In Set 1, all connections were bearing-type (tower bolts were used and tensioned snug-tight
only), two failure modes were found which were local buckling or twisting mode and
flexural buckling mode. The local buckling mode was dominant and the strength increase
accompanied with the mode was much less or moderately less than that of the flexural
buckling mode.

In Set 2, bolted connections were changed to friction-type (high-strength bolts were used
and pretensioned). This change helped improve the strength but by a little amount more
than that achieved in Set 1. It also helped limit the failure mode of the specimens to local
buckling only and bring more consistent failure strengths.

4.4 Internal force analysis


To investigate the effectiveness of each set of tests and connection design in increasing
strength capacity, typical examples of load vs. displacement, axial force development, and

67
shear force development of some specimens will be presented and some discussion given,
in the following subsections.

4.4.1 Load vs. vertical displacement relationship


Figure 4.9 shows load vs. vertical displacement curves of Alt-W-S2, Test 2, Cru-2B-S2,
Test 2, and Cru-2B-S1, Test 3, as well as those of two single-angle tests. The curve of Alt-
W-S2, Test 2 progresses constantly with a steep slope up to one point, and almost
immediately with only a small displacement increase, the specimen reached its ultimate
capacity. For Cru-2B-S2, Test 2 and Cru-2B-S1, Test 3, the curves increase with a
comparable constant steep slope to one point, and after that they progress with a much
lower slope but more curvature until failure.

Figure 4.9 Typical patterns of load vs. vertical displacement curves of weld- and bolt-
connected specimens and single-angle tests

Among these samples, the pattern of the curve of Alt-W-S2, Test 2 is closest to that of the
single-angle tests. This indicates that welding is the most effective method in bringing the
component members to work together like a single member because it can completely
prevent relative movement between the members from bolt slippage. It has to be kept in

68
mind that this conclusion is drawn from considering the similarity between the shapes of
the load vs. vertical displacement curves only; other parameters and effects are not
considered.

By contrast, if the component members are bolt-connected, absolute rigidity and/or fixity
at the connection joints cannot be attained. The built-up member will have a load vs.
vertical displacement curve consisting of two distinct slope intervals, indicating that some
relative movement will occur during the failure process. In this case, whether the relative
movement will be high or low depends on the connectors type and arrangement as well as
the friction from the bolt pretensioning force.

4.4.2 Internal force development: Member forces at sub-panels and shear


forces at connectors

For each test, development curves of the member forces over the three sub-panels N i ( N 1 ,

N 2 , N 3 , N 1* , N 2* , and N 3* ) were plotted. The subscripts 1, 2, and 3 denote top, middle,

and bottom sub-panels respectively; while without the asterisk indicates that the force was
of the main member, and with the asterisk indicates that the force was of the reinforcing
member. Since the steel angles Set 1 and Set 2 had different yield stress values, to make
the results from these two sets comparable, normalization was adopted. See Figure 4.10 for
an example. N i values on the vertical axis were normalized by the cross-sectional yield

strength of a component member Py = Ai f y . Similarly, since ultimate capacities of the

specimens Pc were different, on the horizontal axis the applied load P was normalized by

the specimens Pc value. (This method of normalization was used by Zandonini (1985).)

Shear forces at the three connection joints were determined from the applied load and the
measured axial forces, and plotted in the same manner as that of the sub-panel forces.
Subscripts 1, 2, and 3 denote connection joint numbers from the top end down. See Figure
4.11 for an example.

In this subsection, the development curves of some specimens will be analysed and
discussed in order, according to the effectiveness they provided, starting from welded
connection type to cruciform connection type, alternating plate connection type, and all
plate connectors type. The aligned plate connection type has not been included separately
since it had the same degree of effectiveness as the alternating plate connection type.

69
4.4.2.1 Alt-W-S2 (T2)

Since Alt-W-S2 specimens had the most rigid connection type (welds) which enabled the
component members to act together in a way closest to being a single member, the
behaviours of their internal forces will be used as our reference of the capacity that it is
desired to achieve from bolted connections. The Test 2 specimen has been selected for the
following discussion.

Axial forces Figure 4.10 shows that the axial forces on sub-panels of both members
increased constantly from the beginning to failure point. They can be sorted from the
highest to the lowest as follows: N 1 > N 2 > N 3 > N 3* > N 2* > N 1* . Within the force band,

between N 1 and N 1* , there is a good distribution of the other forces ( N 2 , N 3 , N 3* , and

N 2* ) across the band width from the beginning to failure. It can be noticed that N 3 and N 3*
are very close to each other, indicating nearly equal load sharing between the main
member and the reinforcing member. This implies that if the specimen were made longer
and more connections used, then equal sharing of force between the members would
possibly be achieved in the fourth or fifth sub-panel. In the figure, the N1 / Py line near

failure point increases abruptly due to the slight deformation of the main angle at the top
sub-panel that caused a disproportionally high strain increase to gauges No. 1 and 2. If we
neglect this effect and extend the line further using the last observed slope, at failure point,
the N1 / Py value will be about 1.0 or slightly higher, which means that the main member in

the top sub-panel had reached its cross-sectional yield capacity. This analysis from strain
readings agreed well with the observed result.

70
Figure 4.10 Axial force development curves of Alt-W-S2, Test 2 specimen

Shear forces Figure 4.11 illustrates the development of shear forces at the three
connection joints. It shows that shear forces increase relatively constantly, with V1 at the
top connection being significantly higher than V2 and V3 . Such behaviour can be
interpreted as the top connection having the most important role in transferring force to the
reinforcing member. It is interesting to see that V3 is materially higher than V2 , which
suggests that if all connection joints were made fully rigid using welded connection, the
importance of the next joint will be less. Thus a design using welded connection for every
joint may be overdesigned. It may be more economical, while still comparably effective, to
use bolted connections for the intermediate joints. In such case, friction-type bolted
connections would be preferable to the snug-tight with regard to the provided fixity. The
V1 / Py line has a maximum value of 0.278 and 0.294 for Alt-W-S2, Test 1 and Test 2

respectively before an abrupt change in slope starts to take place. The corresponding shears
V1 of these maximum values are 85 kN and 90 kN respectively.

71
Figure 4.11 Shear development curves of Alt-W-S2, Test 2 specimen

4.4.2.2 Cru-2B-S2 (T1, T2, T3) vs. Alt-W-S2 (T1, T2)

Axial forces Figure 4.12 represents the axial force development of Cru-2B-S2, Test 2.
Its pattern is close to that of Alt-W-S2 specimens in the sense that there is a good
distribution of forces throughout the force band width, and the N 3* line is very close to the

N 3 line like it is in Alt-W-S2 specimens. From this viewpoint, it can be concluded that
using cruciform connection type with bolts pretensioned has the potential to provide equal
load-sharing like when using welded connections. However, the inferior failure loads
obtained in Table 4.1 suggest that its efficiency resulting from the fixity and/or rigidity of
the connection joints cannot surpass that of the welded connection type.

It can also be seen in the figure that when the applied load approaches close to the overall
ultimate capacity, N 1 increases very slightly (almost flat), while N 1* still rises up
increasingly. This point confirms the earlier prediction that when the main member comes
close to its local ultimate compression capacity, the reinforcing member will be
responsible for almost all of the additional load. In that stage, the eccentricity of the top
sub-panel will be doubled, measuring from centroid of the main member to centroid of the
72
reinforcing member. Such induction of eccentricity will then significantly reduce the local
load-carrying capacity of the reinforcing member.

Figure 4.12 Axial force development curves of Cru-2B-S2, Test 2 specimen

Shear forces Figure 4.13 shows the development curves of shear at the connections of
Cru-2B-S2, Test 2. Compared with its axial force development curves, the changes in shear
forces can be observed more apparently, especially for the V1 / Py line. This line clearly

indicates where the main member starts to weaken and the reinforcing member starts to
take up most of the additional load. That point is understood to be where it starts to re-
emerge from zero slope and then accentuates up, which is 0.215 in the figure. The V2 / Py

and V3 / Py lines increase smoothly and constantly, and are inclined to increase further

constantly like in the case of Alt-W-S2 specimens. This means that if the main member
had not failed first, the intermediate connections should still be able to transfer force
further. From this point of view, it may be possible that the cruciform connection type with
bolts pretensioned can work as well as the welded connection type when used for the
intermediate joints.

73
Figure 4.13 Shear development curves of Cru-2B-S2, Test 2 specimen

4.4.2.3 Cru-2B-S1 (T3, T4) & Cru-3B-S1 (T1, T3) vs. Cru-2B-S2 (T1, T2, T3)

Axial forces Specimens Cru-2B-S1, Tests 3 and 4, and Cru-3B-S1, Tests 1 and 3 failed
by twisting of the main member over the top sub-panel and had strength increases in the
range of 30 to 33 %. The axial force development curves of these tests show that the N 1* ,

N 2* , and N 3* lines stay close together in a group throughout the loading period and are

quite independent from the group of N 1 , N 2 , and N 3 lines which also stay close together.

(See Figure 4.14 for N i / Py curves of Cru-2B-S1, Test 3 as an example.) This

development pattern suggests that force transfer is likely to occur mainly at the top end
connection, and in a small amount at the intermediate joints. This comparison proves true
the general perception that snug-tight bolted connection is more flexible than pretensioned
bolted connection. Although the bolt holes are standard-sized, the relative movement (slip)
between the members seems to be able to reduce the force transfer capability significantly.
Therefore, the equal load sharing will be difficult to achieveas shown in Figure 4.14 the
N 3 and N 3* lines are still far from each other.

74
Figure 4.14 Axial force development curves of Cru-2B-S1, Test 3 specimen

Furthermore, if the connection arrangement is prepared this wayusing short connectors


spaced evenly from the top downwhether the equal load sharing is successfully achieved
or not may not be important. This is because if equal load sharing can happen, it will
happen in the fourth or fifth sub-panel; but in fact the main member will get twisted and
fail first in the first sub-panel. Therefore, to solve this premature failure problem, it is
desirable that at the base of the top connection, half of the force is transferred to the
reinforcing member. Using a sufficiently long top connection in conjunction with more
pretensioned bolts may be a possible solution to the problem.

Shear forces Despite the higher importance of the top end connection when bolts are
only snug tightened as discussed in the previous paragraph, the V1 lines of Cru-2B-S1 and
Cru-3B-S1 specimens are still lower than the V1 line of Cru-2B-S2 with bolts
pretensioned. The developments of V2 and V3 are quite unsystematic and hard to clarify.

For example, for Cru-2B-S1, Test 3 and Test 4, each of the V2 and V3 lines increases to

one level and remains unchanged until failure, while for Test 1, they rise beyond the V1
line. (Compare Figures 4.15 and 4.16 for shear curves of Cru-2B-S1, Test 3 and Test 1.)
75
Figure 4.15 Shear development curves of Cru-2B-S1, Test 3 specimen

76
Figure 4.16 Shear development curves of Cru-2B-S1, Test 1 specimen

Discussion By looking at the average percentage of strength increase of the four


specimens above only, namely +31.5%, (especially if Cru-2B-S1, Test 2 and Cru-3B-S1,
Test 2 having higher strength increases are included) and comparing it with that of Cru-2B-
S2, Tests 1, 2, and 3, namely +33.4%, it may be concluded that using snug-tight bolted
connection for all joints is as comparatively efficient as using pretensioned bolted
connection. However, the unpredictability of the slippage effect is a significant reason that
made the strength increases of the specimens in Cru-2B-S1 and Cru-3B-S1 cases vary
greatly. (Figure 4.1)

Looking from the other side on the simplicity of creating the connections in relation to the
strength increase potential, the lowest bound values of Cru-2B-S2 and Cru-2B-S1 are
compared. They are +32.9% from Cru-2B-S2, Test 2 and +23.6% from Cru-2B-S1, Test 1
respectively. Although that of Cru-2B-S1, Test 1 is lower, it is still appreciable. In other
words, using snug-tight (bearing-type) connections is simple and still gives appreciable
outcome.

4.4.2.4 Alt-2B-S2 (T1, T2, T3) vs. Cru-2B-S1 (T1, T3, T4) & Cru-2B-S2 (T1, T2, T3)

Axial forces Figure 4.17 shows the axial force development curves of Alt-2B-S2, Test 2.
As the main member starts to buckle and/or twist, the strain at gauge number 2 increased
disproportionally, resulting in an abrupt increase of the N 1 line. Although bolts were
pretensioned in this case, the force distribution pattern is more similar to the case of Cru-
2B-S1 rather than Cru-2B-S2the sub-panel forces on each member stay together in a
group and are quite independent from the forces on the other member.

77
Figure 4.17 Axial force development curves of Alt-2B-S2, Test 2 specimen

Shear forces Figure 4.18 shows the shear development curves of Alt-2B-S2, Test 2, in
which their patterns agree well with what we can expect from the axial force curves. In
about the first half of the loading period ( P Pc ratio up to 0.5-0.6), force transfers at the
three connection joints increase relatively constantly, with the top end connection
responsible for most of the transferred force. After that, V2 and V3 values stay unchanged

until the failure point, whereas V1 still increases further.

78
Figure 4.18 Shear development curves of Alt-2B-S2, Test 2 specimen

Discussion Considering the sum of bolt tensioning force applied to a joint of the
specimens (Table 4.2), it can be seen that although Alt-2B-S2 had the total amount of
tension force applied in bolts at each connection much higher than of Cru-2B-S1, the
average strength increase was significantly lower (+24.4% versus +33.2%). This point
emphasizes the importance of the stiffness of the intermediate connectors over the
importance of fixity obtained from bolt tensioning when bolting is selected as the method
of connection where the relative movement between members (slip) will certainly occur.

Table 4.2 Sum of tension force in bolts at each connection and average strength increase
of each connection type

Alt-2B-S1 Cru-2B-S1 Alt-2B-S2 Cru-2B-S2


Total Tension in Bolts / Connection (kN) &
Connection No. From Top
(Total Number of Bolts / Connection)
1 227 (12) 227 (12) 750 (12) 750 (12)
2 76 (4) 152 (8) 250 (4) 500 (8)
3 76 (4) 152 (8) 250 (4) 500 (8)
Avg. Strength Increase (%) 20.1 33.2 24.4 33.4

79
Note: For Set 1, the total amount of tension force in bolts of a connection is obtained by
multiplying the number of bolts per connection in the bracket with 18.9 kN (see Section
3.4.3). For Set 2, the multiplier is 62.5 kN.

4.4.2.5 Alt-2B-S1 (T1, T2) vs. Alt-2B-S2 (T1, T2, T3)

As can be expected when bolts are tensioned to just the snug-tight condition, the force
transfer occurs mostly at the top end cruciform connection. Alt-2B-S1 specimens have
axial force and shear development curves similar to those of Alt-2B-S2, except that the
shear forces which occurred at all three joints were lower. (See Figures 4.19 and 4.20 for
axial force and shear development curves of Alt-2B-S1, Test 1 specimen.) Considering the
development curves of these two cases and their average strength increases, we can again
conclude that the capacity of a built-up member can be improved by applying more tension
force to the bolts, and therefore the friction-type connection is recommended if
economically possible.

Figure 4.19 Axial force development curves of Alt-2B-S1, Test 1 specimen

80
Figure 4.20 Shear development curves of Alt-2B-S1, Test 1 specimen

4.4.2.6 Plate connectors at all joints, Set 1

Axial forces The axial force development curves of the five specimens are similar to
those of Ali-S1 and Alt-S1 specimens, except that the sub-panel forces are lower for both
individual members. The distribution of force to the reinforcing member is very poor. As
shown in Figure 4.21, the N 3 and N 3* lines of Pt-Ali-3B-S1, T2 specimen are far away
from each other. If this connection type is selected, to achieve the equal load-sharing
condition, the reinforcing member will need to be extended several sub-panels beyond the
required zone.

81
Figure 4.21 Axial force development curves of Pt-Ali-3B-S1, Test 2 specimen

Shear forces Figure 4.22 shows shear development curves of Pt-Ali-3B-S1, T2


specimen which is typical of the other specimens of this type. The V1 , V2 , and V3 values
were almost the same throughout the loading period. They increased together to about
0.05 0.08Py and stay constant throughout the rest of the loading period. If the use of

semi-spherical support at the top end has no direct effect in making force become
concentrated onto the main members unconnected force as discussed in Section 4.3.1,
such behaviour of the shear curves indicates that if the connector of the top end joint is
made of a single plate, it can induce force to transfer only by the same amount as the plates
at the intermediate joints. In this case, whether the arrangement of the intermediate plate
connectors is aligned or alternating, or two or three bolts are used per connecting face, is
not particularly important due the poor force transfer at the top connection.

82
Figure 4.22 Shear development curves of Pt-Ali-3B-S1, Test 2 specimen

4.4.3 Conclusions from internal force analysis


The important understandings found in this section are as follows.

1 Among the welded, the cruciform, the alternating, and the aligned connection types, the
welded connection type can make a reinforced leg perform best because there is no slip
between component members.

2 The distribution of force to the reinforcing member occurs most at the top connection,
or in other words the top connection is very important for a reinforced leg. The amount
of force transfer is lesser and lesser on the lower connections.

3 If a single plate is used as the top connector, the distribution of force to the reinforcing
member will be very poor. In this case, whether the arrangement of the intermediate
plate connectors is aligned or alternating, or two or three bolts are used per connecting
face, is not particularly important.

4 Using cruciform connection type with bolts pretensioned has the potential to provide
equal load-sharing similar to when using welded connections. However, its efficiency

83
resulting from the fixity and/or rigidity of the connection joints is less than that of the
welded connection type.

5 The snug-tight bolted connection is more flexible than the pretensioned bolted
connection. Although the bolt holes are standard-sized, the relative movement (slip)
between the members can reduce the force transfer capability of the connections
significantly. However, using snug-tight (bearing-type) connections is simple and still
produces appreciable outcomes.

6 For intermediate connections, using cruciform connections with bolts tensioned snug-
tight is more effective than using plate connections with bolts pretensioned.

4.5 Recommendations arising from one-panel tests


After conducting the one-panel leg test, the following recommendations which are
intended to increase the reinforcing ability as much as possible are made.

1 A cruciform arrangement consisting of two angle cleats of the same size as that of the
legs is recommended to be used for all connection joints.

2 The top end joint is crucial in the distribution of force to the reinforcing member. The
most desirable (ideal) condition is that both individual members carry equal
magnitudes of force in the first sub-panel, immediately below the top connection. Here
it is suggested that the slip-critical bolted connection type (friction-type) be used at the
top end joint, to ensure a more controllable rigidity or stiffness and less relative
movement between the individual members that will reflect throughout the whole
length of the reinforced leg.

3 Also, at the top end joint, the angle cleats used should be continuously long enough to
provide adequate space for the bolt row. The number of bolts required can be
approximated by dividing the maximum load an individual member will carry with the
slip resistance (friction force) a pretensioned bolt can provide. In this case, all the bolts
to be used at this joint should be of high-strength type, Grade 8.8.

4 If possible, a number of connections from the top end should be made with the same
quality as mentioned in points 2) and 3).

84
These recommendations will be examined using the finite element modelling as discussed
in Chapters 5 and 6.

85
CHAPTER 5

5 Development of one-panel finite element models of


reinforced tower legs
5.1 Introduction
The first objective of the finite element modelling was to develop a model that could
accurately replicate the experimental results discussed in Chapter 4. This chapter will
discuss the means used to model and assemble the connections and sub-panels
economically and reliably, as they are the two major components that constitute a
reinforced leg. A series of FE models which simulated the one-panel specimens was
created and the modelling results were compared with the available test results for
verification. Once this was satisfactorily achieved, the next significant objective of this
research was to investigate the potential of the leg reinforcing method by conducting a
study on tower models having leg(s) reinforced for several panels. This latter study will be
discussed in Chapter 6.

It was concluded in Chapter 4 that bolt slip had an important effect on the potential of the
reinforcing method. Hence, most of the literature reviewed in relation to FE modelling (in
Chapter 2) was related to models with bolt pretensioning and bolt slip problem. The
majority of the studies found were about bolted beam-to-column connections and many of
them used solid continuum (or 3D) elements for every part. Therefore, in the early stage of
this study, models were constructed using 3D continuum elements. Only one connection
typecruciform angle typewas modelled, given that it was the most effective type and
is most likely to be adopted in practice. Using 3D continuum elements, the models could
be created to have exactly the same geometries as those of the test specimens. The
components (the main member, reinforcing member, angle cleats, and bolts (including nuts
and washers)) were created as separate parts and assembled together later. This means of
modelling gave satisfactory results. However, it involved too many parts and interaction
definitions to be input, for it to be used to create multi-panel reinforced tower models in
86
the same way. This limitation forced the study to find a more flexible model or method that
still was able to produce acceptably the same level of result but was able to be extended to
multiple panels. To do this, the connection design was simplified principally as follows:
two angle cleats forming a cruciform shape were replaced by an encasing ring; and bolts
and bolt holes were replaced with locking stripes and locking grooves respectively. To
reduce the analysis time further, the sub-panel portions that were continuum elements were
replaced with beam (1D) elements as appropriate, and their results were evaluated.

Another important issue was that when 3D continuum elements were used, there were
different hourglass control types to choose and each type had a different effect on the
stiffness of the model. The most appropriate type was chosen using a single-angle model
whose results were compared with the results of single-angle tests.

5.2 General material properties


The material strength property of the longitudinal angle members and the parts that
constituted them and the angle connectors were assumed to be elastic-perfectly-plastic ( f y

= 328 MPa and E = 205,000 MPa). The strength hardening property was not included, in
order to reduce the analysis time.

Bolts, nuts, and washers, as well as the ring connectors in Section 5.5.3, were assumed to
be completely elastic ( E = 205,000 MPa) to make them stronger than the longitudinal
members and the connectors.

5.3 One-panel single angle models


One-panel single angle (or unreinforced) models were developed at the beginning to find
out which parameters needed to be used in order to get reasonable results.

Since the load was concentric, a geometric imperfection was provided. A buckling analysis
in ABAQUS/Standard was first performed to find the first four buckling modes. Then the
modes and their deflected coordinates were scale factored to make initial coordinates of the
imperfect models. The imperfect models were then loaded to failure using Dynamic,
Explicit (Quasi-Static) procedure in ABAQUS/Explicit. In this section, 3D continuum
elements were used.

When assigning the details of the elements in Mesh Module, it was necessary to choose the
right type of Element Control and the right type of Hourglass Control in order to obtain
87
reliable results. For Standard Linear Hexagonal elements (8-node linear bricks), besides
the ordinary linear element type, there were three types of element controls to choose from:
Hybrid Formulations, Reduced Integration, and Incompatible Modes. (Table 5.1) However,
for Explicit linear hexagonal elements, only the Reduced Integration element control type
was available, but it had three main hourglass control types to choose from: Enhanced,
Relaxed Stiffness, and Stiffness. To examine the differences they could make, each of
them was assigned to the same single-angle model one by one.

For the buckling analysis, whether the element control was specified or not, or if specified,
whether it was hybrid formulation type or incompatible modes type, the results
(eigenvalues) were the same. Therefore, only the model with no element control will be
discussed in Subsection 5.3.2. For the dynamic/explicit analysis, using different hourglass
control types gave different results. Their strength results were compared with the code
strength to select the best type to be used for the reinforced models (see Subsection 5.3.3).

Table 5.1 Chosen element details for the investigation in this section (in highlighted areas)

Subsection 5.3.2 Subsection 5.3.3


Buckling Analysis Dynamic, Explicit Procedure
in ABAQUS/Standard in ABAQUS/Explicit
Selected Type of
Element Library, Geometric Order, Family, Element Shape
Standard, Linear, 3D Stress, Hexahedral Explicit, Linear, 3D Stress, Hexahedral
Element Controls
1) No Element Controls 1) With Element Controls
2) With Element Controls - Reduced integration
- Hybrid formulation - Hourglass control
- Reduced integration - Enhanced
- Hourglass control - Relax stiffness
- Enhanced - Stiffness
- Relax stiffness
- Stiffness
- Incompatible modes

5.3.1 General details


A single angle model was created using 3D continuum elements. Its geometry and section
are shown in Figure 5.1. The fillet at the inner corner of the cross-section was ignored.
Four circular cuts, 17.5 mm in diameter, were made at the bottom to represent bolt holes.
The length from the centre of bolt group to the top end was 1.13 m.

88
Figure 5.1 General configuration of a single-angle model and when meshed (on the right)

Since the bottom end of the angle test pieces was bolt-connected to the bottom support and
the force was in the axial direction, only the upper half or less of each bolt hole was
responsible for bearing the force. Therefore, in the models the upper half area of each bolt
hole was specified with boundary condition. The condition was zero displacement in the X,
Y, and Z axes (U1 = U2 = U3 = 0) at the initial stage.

For the top end, a square rigid plate was placed centre-to-centre with the angle section, and
they were tied together surface-to-surface. A zero displacement (U1 = U2 = U3 = 0)
boundary condition was given to the centre of the rigid plate to provide a pin-end
condition.

5.3.2 Buckling analysis and results


To perform the buckling analysis, an assumed concentrated force of 1000 kN was applied
to the centre of the rigid plate at the top end in the Z-direction (CF3 = -1000 kN). The

89
boundary condition of the top end mentioned earlier was changed to U1 = U2 = 0; while
the condition of the bottom end at the bolt holes remained the same.

Since buckling analysis was only available in ABAQUS/Standard, it required that Standard
element types be assigned. For the single angle model using 3D continuum elements, the
ordinary 8-node linear brick element type (C3D8) was used. The linear geometric order,
instead of quadratic, was chosen in order to save analysis time. However, to compensate
for accuracy the model was subdivided into small elements as shown in Figure 5.1.

While usually only the first buckled mode was of interest, four buckled modes were
requested here. The four eigenvalues of the four modes in sequence were 0.4651, 0.5159,
0.57967, and 0.61407; and Figure 5.2 illustrates buckled shapes of these four modes in
Von Mises Stress outputs. The deformed coordinates of nodes of each buckled mode were
saved to be scaled by a factor later to make base coordinates of the models with geometric
imperfection.

Figure 5.2 The first four buckled modes of single-angle model using C3D8 elements

90
The first to the third modes showed local buckling or twisting failure, but with different
number of twisting curves. Since they all were twisting failure, the first mode dominated.
In case of twisting failure, the scale factor must be determined by using the percentage of
the maximum initial displacement to the angles thickness. The node with maximum
displacement of the first mode was searched and its displacement magnitude was found to
be 1.001 m. By assuming that the percentage above was 3%, as recommended by
ABAQUS, the scale factor was calculated as follows:

Max. Initial Displacement (1.001 m) Scale Factor


= 0.03 =
Section' s Thickness 0.006 m

The scale factor of the first mode was 1.8 10-4. If the percentage was 2%, the scale factor
would be 1.2 10-4.

The fourth mode was buckling about the sections weak axis. For this failure mode, the
scale factor was determined using the ratio of the maximum initial displacement to the
angles length. Von Mises Stress output mode was used to help locate the inflection point.
(Figure 5.2, Mode 4) The inflection point was located at the middle of the dark-blue zone
where the stress was lowest, and the coordinate was queried. The effective length L'
obtained was 0.789 m. (With the total length L of 1.13 m, the effective length factor was
0.789/1.13 = 0.698, or equal to the theoretical value 0.7 for pinned-fixed columns.) This
0.789-m effective length, not the total length, was used in finding the scale factor. To
determine the maximum displacement, the spatial displacement output was used. The
cross-section that had the maximum displacement was searched and the average
displacement of the nodes on that section was calculated. The average value was 1.389 m.
Assuming that the imperfect angle had the ratio of the maximum initial displacement to the
effective length equal to 1/1000, the scale factor was calculated from:

Max. Initial Displacement 1 (1.389 m) Scale Factor


= =
Effective Length 1000 0.789 m

The scale factor of the fourth mode was 5.68 10-4. And if the ratio was 1/500, the scale
factor would be 1.136 10-3.

5.3.3 Models with geometric imperfection and results


For a concentrically loaded single angle, it is usual, as design codes or standards generally
do, to consider that the angle will fail by buckling about the weak axis primarily.

91
Therefore, the imperfect models were created (scale factored) based on the shape of the
fourth buckling mode rather than any of the first three modes, even though the fourth mode
had a higher eigenvalue. The deformed coordinates of the nodes of the fourth buckling
mode were carried over to this section and scale factored by 5.68 10-4 and 1.136 10-3 to
make base coordinates of two imperfect models with maximum-initial-displacement-to-
effective-length ratios of 1/1000 and 1/500 respectively.

The Dynamic/Explicit analysis procedure was used in this place to take advantage of its
computational speed and flexibility. Although Static/General and Static/Risk procedures
could also be used to analyse single-angle models here, they were not chosen because of
their lack of capability to handle reinforced models in the next sections which involved a
complicated contact problem.

To load the angle models, the top end was moved (or displaced) toward the bottom
support. The loading process required two input values which were the displacement
magnitude of the top support at the centre of the rigid plate and the loading time period.
These two values were unknown at the beginning. A large target displacement magnitude
and a short time period were selected by trial for the first analysis run. Then its load-time
curve and its displacement-time curve were checked to determine at what new target
displacement and time period the second run should be set. The second run was (or the
later runs, in general, would be) given the same target displacement magnitude as that of
the first run or slightly less, while the time period was increased. The load-time curve was
evaluated and the processes repeated for the third and fourth runs and so on, until the
difference in buckling strengths of two adjacent runs was within 3%. This method may be
called trial and error or convergence check. In addition, to smooth the loading process
which in return helped produce smooth results, the Smooth Step amplitude in Loads
procedure was used for every run.

The last step was to choose the element type and its details. As the analysis procedure was
Explicit, the Explicit element library was chosen. As explained previously for explicit 8-
node linear brick elements, only Reduced Integration element control was available, but
there were three types of Hourglass Control to choose from. These were designated as
follows:

92
C3D8RE = 8-node linear brick, reduced integration, enhanced hourglass control

C3D8RR = .., relaxed stiffness hourglass control

C3D8RS = .., stiffness hourglass control

To test the differences they made, each of them was assigned to separate models. Table 5.2
summarizes all the models, cases, and runs that were created and analysed in this section.

Table 5.2 Summary of all cases and runs of the single-angle model with initial
imperfection

Run No. (Assigned Displacement, Loading Time Period)


Imperfection Element Type Run 1 Run 2 Run 3 Run 4
Ratio and Controls (2mm, 0.03s) (2mm, 0.06s) (1.8mm, 0.09s) (1.7mm, 0.14s)
Model Name (Strength (kN))
1/500 C3D8RE 500E1 (256.6) 500E2 (250.6) 500E3 (245.2) 500E4 (242.4)
C3D8RR 500R1 (179.3) - - -
C3D8RS 500S1 (256.3) 500S2 (250.5) - 500S4 (241.6)

5.3.3.1 Results and discussion

A. Using 1/500 imperfect models, Set 1 to find the best hourglass control type

First, the case of 1/500 imperfection ratio was considered. It had three models assigned
with three hourglass control types mentioned above. All were given the same set of target
displacement magnitude and loading time period (2 mm in 0.03 s). This set was called
Run 1. The three models were named 500E1, 500R1, and 500S1 (Table 5.2). Their
results presented by load-time curves are in Figure 5.3. The model using relaxed stiffness
hourglass control (500R1) had the maximum load of 179.3 kN, while the models using
enhanced hourglass control (500E1) and stiffness hourglass control (500S1) had the same
maximum load of 256 kN. From the strength viewpoint, these results apparently indicated
that using relaxed stiffness hourglass control will make the model weaker (not as stiff as)
than using the other two types.

To check which control type was the appropriate one to use, the maximum analysis loads
obtained were compared with the code compressive strength. The code strength N c was

244 kN according to AS 3995 (1994) (taking L' = 0.7 L , L = 1130 mm, r = 14.7 mm, f y

= 320 MPa, and A = 864 mm2). Because the 500E1 and 500S1 models had the maximum
loads closer to the code strength, it was concluded that either Enhanced or Stiffness
93
hourglass controls were suitable to be used, while the Relaxed Stiffness hourglass control
was not.

Figure 5.3 Load-time curves of 500E1, 500R1, and 500S1 models

Figure 5.4 shows load-displacement curves of the three models above. It can be noticed
that the curves of the 500E1 and 500S1 models perfectly overlay each other until they
reach the same maximum load value. But after that (sometimes called post-buckling
portion) while the curve of the 500E1 model drops suddenly almost vertically, the curve of
the 500S1 model slowly drops.

A comparison of the post-buckling portion of both models with those of the single-angle
test specimens was plotted in Figure 5.4 where it can be seen that the 500S1 model shows a
better result. In other words, although the Enhanced type and Stiffness type of hourglass
control produced the same pre-buckling behaviour and maximum load value, the Stiffness
type gave the post-buckling behaviour more similar to the actual test specimens. Hence,
the Stiffness hourglass control was selected to be used in further models.

94
Figure 5.4 Load-displacement curves of 500E1, 500R1, and 500S1 models and of three
single-angle tests

B. Finding the actual buckling strengths of 1/500 imperfect models

To determine the actual buckling strength of any model or case, a series of identical
models with lower loading speeds were created and analysed. Their load-time curves were
checked for convergence as in the previous section. Here, three more runs (Run 2, Run 3,
and Run 4) of the 1/500 imperfection case were created. Their displacement magnitudes
and loading time details are given in Table 5.2.

Similar to the load-time curves of Models 500E1 and 500S1, the load-time curves of
Models 500E2 and 500S2 in Figure 5.5 overlay each other up to the point of maximum
load. The curves of Models 500E4 and 500S4 were also the same. This confirms that the
Enhanced and Stiffness hourglass control type when assigned to a model will produce
exactly the same performance to the point of maximum capacity. Since Runs 3 and 4
(Models 500E3 and 500E4) had the maximum capacity of 245.2 and 242.4 kN which made
a 1.2% difference, no further run was created. Therefore, the actual strength of this single-
angle model with 1/500 initial imperfection was assumed to be 242 kN, which was an
average value from 500E4 and 500S4. Since this strength value was very close to the 244
95
kN calculated strength, it was considered unnecessary to create and determine the strength
of single-angle models with 1/1000 imperfection. With regard to this matching-well
strength result, the objective of this subsectionto learn and exercise appropriate
modelling method in order to obtain reasonable resultshad been achieved.

No other runs of the case using Relaxed Stiffness hourglass control type, except 500R1,
were created because they would only produce lower strengths than 179.3 kN of 500R1.

Figure 5.5 Load-time curves of all cases and runs showing convergence

5.4 Bolt pretensioning


5.4.1 Bolt design
The bolt, nut, and washer were modelled together as one piece to save input data. This
piece was simply called Bolt. (Figure 5.6(a)) The bolts were designed to have the clear
distance between the bolt head and the washer slightly less than 12 mm which was the
combined thickness of a connector (an angle cleat or a flat plate) and a main angle
member. Four clear distances11.90 mm, 11.95 mm, 11.99 mm, and 12.10 mm
representing four bolt designs were used to examine the effect of using different bolt
pretensioning forces. For material properties, although in reality bolts had elastic and
96
inelastic property ranges, they were assumed to be exclusively elastic in this place. This
was to prevent them from yielding at some locations along the shank and not being able to
contract back elastically after being stretched and released, which would happen if they
were assumed as having elastic-inelastic property that would cause difficulty in
determining the resultant pretensioning forces.

5.4.2 Meshed geometries and damping constant values


To clamp the angle components and the connectors together, the bolts were stretched and
then released. The process will be explained in Section 5.4.3. After releasing, trying to
maintain their original length, the bolts immediately contracted and tried to clamp the
connectors and longitudinal members in between the bolt head and washer. This caused
energy exchange between them. The kinetic energy would dissipate, but without
introducing some general damping the dissipation (or stabilization) would take a very long
time. Therefore, some general damping was given to every part of the model. Here, the
stiffness proportional damping R was specified as part of the material definition.

The R factor introduces damping proportional to the strain rate, which can be thought of
as damping associated with the material itself. ABAQUS states that R must be used with
caution because it may significantly reduce the stability time, and to avoid a dramatic drop
in the stability time increment the R should be less than or of the same order of
magnitude as the initial stability time increment without damping. It is mentioned that if
the size of the smallest element dimension, Le , and the wave speed of the material, c d , are

known, the stability time increment can be estimated using t stable = Le / c d and

c d = E / , where E is Youngs modulus and is the mass density.

The smallest element dimension Le of a part was searched from its meshed figure. The Le
of the bolts was 0.85 mm, and that of the connectors and longitudinal members was 1.687
mm. Using the estimating relationships above, the corresponding t stable values were 1.658

10-7 and 3.291 10-7 sec (taking E = 2.05 1011 N/m2 and = 7800 kg/m3).
According to the above statement that the R value should be less than or equal to the
t stable value, the finally specified R values were assumed to be approximately half of the

calculated t stable values, or 8 10-8 sec for bolts and 1.8 10-7 sec for the connectors and
longitudinal members.
97
5.4.3 Steps involved and energy curves
Bolt pretensioning involved three main steps: 1) Pulling the bolt heads, 2) Releasing the
bolt heads, and 3) Stabilizing. The interactions, restraints, and loads were assigned in the
format that required the smallest size of input data.

In the first step, after every part was assembled, there were two initial contact areas.
(Figure 5.6(b)) The first one was between the leg member and the connector (Contact A),
and the second was between the leg member and washer (Contact B). Their interactions
were created. In the Load module, the free surface of the connector (Surface C) was
restrained while the inner surface of the bolt head (Surface D) was moved (pushed)
smoothly for a distance that made both surfaces become the same level in a time period.
(For example, in the case that used 11.95-mm-clear-distance bolts, the pushing distance
was 0.05 mm (12 mm minus 11.95 mm)). This caused the washer (plus the nut) on the
other end to squeeze the angle leg and the connector in the middle and itself together.

In the second step, after the free surface of the connector (Surface C) and the inner surface
of the bolt heads (Surface D) were at the same level, their interaction was created. Then the
bolt heads (Surface D) were released or inactivated and the connections were left to
stabilize for a short period. (Figure 5.6(c))

In the third step, the restraints on Surface C were inactivated and the connections were left
to stabilize for a certain length of time. This allowed the bolts, the angle legs, and the
connectors to adjust their position, length, and thickness spontaneously. Figure 5.7 gives an
example of the changes in total internal energy and total kinetic energy of these three steps
of Models 1190_3D and 1195_3D. (See Table 5.4 for the meanings of these names.)

Now having completed the bolt pretensioning process, the reinforced model was ready. In
entering the fourth step where the model was to be loaded, interactions between bolt
shanks and bolt holes on the angle members and on the connectors were created so that
bolt bearing could take place and force could be transferred.

98
Figure 5.6 (a) Bolt design; (b) and (c) Steps 1 and 2 of bolt pretensioning process

Figure 5.7 Changes in total internal energy and total kinetic energy from bolt
pretensioning process of Models 1190_3D and 1195_3D
99
5.4.4 Resultant pretension
The resultant force in a bolt was determined at the end of Step 3. Eight elements around the
shank were chosen and their stress outputs were averaged and converted to the resultant
force. The resultant forces of the first three bolt designs with the clear distances of 11.90,
11.95, and 11.99 mm were 47.0, 23.6, and 4.7 kN respectively. The actual resultant forces
in the bolts of the test specimens were not known, but they had been approximated by a
simple formula in Section 3.4.3, as 62.5 kN for a pretensioned bolt and 18.9 kN for a snug-
tight bolt. Even though these approximate tension forces were not equal to the values
obtained from the models as given above, the models still had the merit to demonstrate the
effect of the degree of bolt pretension on strength reinforcing.

5.5 Reinforced models


It is important to restate here that all modelling work presented in this thesis was carried
out in a series of trials in a logical way, where the most important purpose was to find a set
of modelling ways that allowed the study of the efficiency of leg reinforcing method for
multi-panel tower structures to be possible in a manageable time. These ways covered
two important issues. The first was how to model the connections, and the second was
what element type to use for the steel of the sub-panels. The two issues above and the
causes of difficulties will be explained partly in the following subsections and also in
Chapter 6.

The trial series had four stages as summarised in Table 5.3. Stages 1 and 2 used a similar-
looking model whose connection portions and sub-panel portions were made of continuum
elements. Stage 1 was set up to confirm that the plain Stiffness type of Hourglass Control
was the suitable type to use. Stage 2 was to evaluate whether the model and especially the
means used in modelling were alright, giving reasonable results.

In Stage 3, the second model type was created by modifying the sub-panel portions of the
first model type from being continuum elements to being beam elements.

In Stage 4, the third model type was created by using a simplified connection design and
beam elements for the sub-panel portions. Each stage had at least one test result to compare
it with. The details of the models presented in this section are summarized in Table 5.4.

100
Table 5.3 Stages of model development and comparisons which lead to the final
simplified reinforced model

Comparison Between
Stage
Model (Tension Level in Bolts) Test Set (Bolted Condition)
1199_3D_R (Nearly Zero Tension)
Stage 1 Set 1 (Snug-Tight)
1199_3D (Nearly Zero Tension)
1190_3D (High Tension) Set 2 (Fully-Tensioned)
1195_3D (Moderate Tension)
Stage 2
1199_3D (Nearly Zero Tension) Set 1 (Snug-Tight)
1210_3D (Zero Tension)
1210_3D (Zero Tension)
Stage 3 Set 1 (Snug-Tight)
1210_Beam (Zero Tension)
1210_3D (Zero Tension)
Stage 4 1210_Beam (Zero Tension) Set 1 (Snug-Tight)
Simplified_Beam (Zero Tension)

Table 5.4 Details of the models specified by their names in Table 5.3

Model Connection Bolt Head-to-Washer Element Type Hourglass


Model Name Type Design Clearance (mm) & for Sub-Panel Control Type
Tension Level in Bolts (Abbreviation)
11.99 Conntinuum
1199_3D_R 1 Detailed Bolted Relax Stiffness
Nearly Zero Tension (C3D8RR)
11.90 Conntinuum
1190_3D 1 Detailed Bolted Stiffness
High Tension (C3D8RS)
11.95 Conntinuum
1195_3D 1 Detailed Bolted Stiffness
Moderate Tension (C3D8RS)
11.99 Conntinuum
1199_3D 1 Detailed Bolted Stiffness
Nearly Zero Tension (C3D8RS)
12.10 Conntinuum
1210_3D 1 Detailed Bolted Stiffness
Zero Tension (C3D8RS)
12.10
1210_Beam 2 Detailed Bolted Beam (B31) -
Zero Tension
Simplified -
Simplified_Beam 3 Beam (B31) -
(Locking System) Zero Tension

5.5.1 Stages 1 and 2, Type-1 Model (using detailed bolted connections +


continuum elements for sub-panel portions)
The general view of Type-1 Modelone-panel reinforced model with cruciform bolted
connections having every part and portion created from continuum elementis shown in
Figure 5.8. The element type C3D8R (8-node linear brick, reduced integration) was
assigned to the whole model. However, it required an hourglass control type among the

101
Enhanced, Relaxed Stiffness, or Stiffness type to be chosen too. From Section 5.3 with the
single-angle model, the results obtained from using Enhanced or Stiffness type were the
same, at least until they reached the maximum strength value. Therefore, in this section the
Enhanced type was neglected.

Figure 5.8 Complete assembly of a reinforced one-panel model with cruciform bolted
connections in meshed view

5.5.1.1 Stage 1 (1199_3D & 1199_3D_R)

In Stage 1, only the Stiffness and Relaxed Stiffness hourglass control types were
considered. Its objective was to ensure that the plain Stiffness hourglass control type was
the suitable type to use. In this stage the case of bolt parts having the head-to-washer clear
distance of 11.99 mm from Section 5.4 which created a near-zero tension force (4.7 kN) to
the bolts. Two models were created. They were assigned with Stiffness and Relaxed
Stiffness types, and named as 1199_3D and 1199_3D_R respectively.

Each model used a number of runs with decreasing loading speeds (the target displacement
decreased but the loading time period increased) in which their strength results were
102
checked for convergence. The concept used was the same as that used with the single-
angle models.

A. Results and discussion

Model 1199_3D_R used five runs in total and their load-time curves are presented in
Figure 5.9. The figure shows that the maximum capacity of the model is still on a declining
trend. The fifth run (Run 5) had the capacity of 213.6 kN which was significantly lower
than the worst strength value of the Set 1 Cruciform-type specimens (328 kN from Cru-2B-
S1, Test 1), and even lower than the average strength value of the Set 1 single-angle tests
(265 kN). This indicated that using the Relaxed Stiffness hourglass control type caused the
continuum elements, and consequently the model, to be weak. The deformed shape of
1199_3D_R, Run 5 at the maximum capacity stage is shown in Figure 5.10(a). It shows an
obvious twisting of the main member at the first sub-panel, which was very much like the
dominant failure pattern of Set 1 Cruciform-type specimens. (Figure 5.10(c))

In the above paragraph, the result of Cru-2B-S1, Test 1 specimen was referred to. The
2B term means that each connecting face of the intermediate connections used two bolts
(so each intermediate connection had eight bolts in total). This condition, regarding the
number of the bolts used, was not the same as that of Type-1 Model where each connecting
face used three bolts. However, since the specimen gave the lowest load-carrying capacity
in the Set 1 Cruciform-type specimens, it was brought forward and used as a reference
here.

103
Figure 5.9 Load-time curves of five runs of Model 1199_3D_R

Figure 5.10 Deformed shape at maximum capacity of (a) Model 1199_3D_R, Run 5; (b)
Model 1199_3D, Run 4; and (c) Specimen Cru-3B-S1, Test 1
104
Similarly, Model 1199_3D used four runs in total. Their load-time curves show an obvious
convergence. (Figure 5.11) The third and fourth runs had the same maximum capacity
(320.8 kN and 320.6 kN respectively). Hence, no more runs were created. The model was
assumed to have the maximum capacity of 320.6 kN. This value was not much lower than
the 328 kN of Cru-2B-S1, Test 1. Comparing these two numbers, it was understood that
using the plain Stiffness hourglass control type was likely to make the models perform in
terms of strength as well as the real structures. However, the deformed shape of 1199_3D,
Run 4 (when it reached the maximum capacity stage) (Figure 5.10(b)) showed no sign of
twisting, but only flexural bending of the compound (or built-up) section about its major
axis at the top portion of the model.

Figure 5.11 Load-time curves of four runs of Model 1199_3D

From the results and comparisons above, two conclusion points were made. First, if
Relaxed Stiffness hourglass control type is used, it will take several runs before the
strength results will be constant. At that time, the models maximum strength will be
dramatically lowmuch lower than what the real structure should have. This indicated
that the Relaxed Stiffness type should not be used, although the deformed shape it
produced looked very convincing. Second, if Stiffness type is used for a model, the model
105
will end up with a run that gives a quite reasonable strength result although the deformed
shape may not look so convincing.

In ABAQUS, since the default hourglass control type for C3D8R (8-node linear brick,
reduced integration) elements was Relaxed Stiffness type, it therefore required that the
type be changed to only Stiffness type for every part made of the continuum C3D8R
elements.

5.5.1.2 Stage 2 (1190_3D & 1195_3D & 1199_3D & 1210_3D)

Now it was known that the Stiffness hourglass control type should be used, Stage 2 was to
confirm that Type-1 Model also gave reasonable results in other bolt tensioning cases,
which would mean that the modelling approach, the material properties, the interaction
properties, the meshing, and the element properties adopted were reasonable too.

Here, three new models1190_3D, 1195_3D, and 1210_3Dwere introduced. The 3D


term means that the elements hourglass control type was Stiffness type as before. Model
1190_3D had high tension (47.0 kN) in each bolt after the bolt pretensioning process,
while Model 1195_3D had moderate tension (23.6 kN). (See Section 5.4 and Table 5.4)
Due to such levels of tension, the results of these two models were compared with those of
Cruciform specimens, Set 2 and Set 1 respectively.

Model 1210_3D was an interesting case. The bolts of the model bolts were set to have the
bolt head-to-washer clearance equal to 12.1 mm, or 0.1 mm longer than the total thickness
of the members the bolts held. The bolts therefore had no pretension, and the 0.1-mm gap
allowed the main and reinforcing members to behave more individually. This case was
added in to imitate the real condition of the bolted connections where nuts will loosen after
some time of service. The result of this model was compared with the poorest test result of
Cruciform specimens, Set 1.

A. Results and discussion

Models 1190_3D, 1195_3D, and 1210_3D used five, three, and four runs respectively.
Since their presentations of load-time curves for convergence check are similar to Figures
5.9 and 5.11, they are not presented here. The discussion below will refer to the result of
the last run, the representative run, of each model only.

106
A.1 Models 1190_3D, 1195_3D, 1199_3D, and 1210_3D

Load-displacement curves of the representative runs of Models 1190_3D, 1195_3D,


1199_3D, and 1210_3D are plotted in Figure 5.12. The maximum capacities and their
corresponding displacements and the capacity increase percentages are summarised in
Table 5.5.

Figure 5.12 Load-displacement curves of representative runs of Models 1190_3D,


1195_3D, 1199_3D, and 1210_3D

Table 5.5 Capacities and capacity increase percentages of Type-1 models

Connection Element Type Tension in Max Capacity Displacement Increased


Model
Design Type for Sub-Panels Bolts (kN) (kN) (mm) Capacity (%)
1190_3D Detailed Bolted Continuum 47.0 366.0 6.6 51.2
1195_3D Detailed Bolted Continuum 23.6 350.4 7.7 44.8
1199_3D Detailed Bolted Continuum 4.7 320.8 7.3 32.6
1210_3D Detailed Bolted Continuum -- 303.5 7.8 25.4

While Model 1210_3D and 1199_3D with zero and near-zero pretension in the bolts
gained capacity increase of 25.4% and 32.6%, Models 1195_3D and 1190_3D with
moderate and high pretensions gained +44.8% and +51.2% respectively. These results

107
indicated that applying moderate or high pretensioning force to the bolts could help
improve the capacity of reinforced legs appreciably more than applying low or no
pretensioning force. However, the benefits from applying moderate and high pretensioning
force were just about the same level. (See Figure 5.12 for a clearer view.) Hence, if one
wishes to specify bolt pretensioning into the specification for field operation for upgrading,
a moderate pretension is recommended.

Model 1210_3D had the maximum load of 303.5 kN, or a capacity increase of 25.4%
which was lower than, but not much different from, the +32.6% of Model 1199_3D. (See
Table 5.5.) So, even in this worst case the potential of the leg reinforcing method was
impressive.

It has to be borne in mind that the above numbers were obtained from the models which
simulated reinforced one-panel leg. They need to be viewed for comparison purposes only.
The actual potential of the reinforcing method can be quite different because when an
actual tower leg needs to be reinforced, it will be reinforced for several panels. Also the
slenderness ratio of each panel and the size of connectors can be different from what was
used in this study.

To verify these models, their load-displacement curves from Figure 5.12 were compared
with test results as guided in Table 5.3. (Figure 5.13)

In Figure 5.13, the curves were divided in three groups. The first had moderate pretension
in bolts (Model 1195_3D vs. Cru-3B-S1, Test 3), the second had near-zero pretension
(Model 1199_3D vs. Cru-2B-S1, Test 1), and the third had zero pretension (Model
1210_3D).

For the moderate-pretension group, the curves of Model 1195_3D and Cru-3B-S1, Test 3
specimen had about the same slope in their elastic range and entered inelastic range at
about the same applied load value. Then the two curves overlaid each other for some
period and separated. The curve of Cru-3B-S1 continued with its higher slope until it
reached the peak at the displacement of about 4.5 mm and then dropped abruptly. In
contrast, the curve of Model 1195_3D continued to rise at a decreasing rate until the
displacement of 8.0 mm and then declined slowly. As seen in Figure 5.13, the gap or
difference of the two curves in the inelastic range after separation was small.

108
Figure 5.13 Load-displacement curves of models and test specimens with bolts having
moderate, near-zero, and zero pretension

In the second group, the pretension level was near-zero, the model (Model 1199_3D) was
able to imitate the performance of the specimen (Cru-2B-S1, Test 1) well again in the
elastic rangethe curves of the model and the specimen almost overlaid each other. When
entering the inelastic range, the two curves showed the same patterns of progress as the
curves of the moderate-pretension group. However, they separated upon entering the
inelastic range, and the gap or difference between them was larger than that of the
moderate-pretension group.

Such difference in the inelastic range was likely to result from the assumption that the
material property of the models was elastic-perfectly-plastic only, no hardening property
was provided. The specimens, having hardening property, were stronger, and hence their
curves rose higher. The magnitude of difference (or in other words, the effect of not
specifying hardening property) was small when the pretension level was high, but became
larger when the pretension level was lower. (Note again that hardening property was
intentionally not included, in order to save some analysis time.) The reason that the curves

109
of the models did not show an immediate drop will be discussed after the next two
paragraphs.

Model 1210_3D had zero pretension in the bolts. Because there was no test specimen in
this category, the load-displacement curve of the model was compared with the curve of
the poorest-performance specimen which was Cru-2B-S1, Test 1. The slope of the model
in the elastic range was lower than those of the other models and the specimens. This
reflected the lowest effectiveness of the connections when the bolts had no tension. The
integrity of the main member and reinforcing member was lowest, and as a result the load-
carrying capacity was lowest.

To explain why the curves of the models did not show immediate drop, the deformed shape
of the models (Models 1190_3D, 1199_3D, and 1210_3D) in Figure 5.14 was observed.
They look alike even though the pretension levels in bolts and the produced maximum
capacities were different. They showed only bowing of the compound section. The
interesting point is that the first sub-panel of the main leg of every model did not twist as it
should, even though the stress over the sub-panel had already reached yield stress. The
most likely cause that attributed to such behaviour was the modelling approach adopted.
Because a model was formed by assembling individual parts together which involved
several contacts among the parts, only ABAQUS/Explicit could be used. And since
buckling analysis can be performed with ABAQUS/Standard only, conducting buckling
analysis with the reinforced models was not possible. (To be able to perform buckling
analysis, all parts or members of the desired model have to be created as one piece. But by
doing so, the member slip which is crucial to the performance of a reinforced leg will not
be modelled. Therefore, the models were not specified in this way.)

Without buckling analysis, the geometric imperfection pattern could not be obtained, and
the analysis had to be conducted with perfect geometry. Then the only failure pattern
possible was bending of the compound section, instead of twisting of the first sub-panel of
the main leg.

110
Figure 5.14 Deformed shapes of Models 1190_3D, 1199_3D, and 1210_3D

The development curves of the sub-panel forces of Cru-3B-S1, Test 3 and Model 1195_3D
are given in Figures 5.15 and 5.16. The forces have been normalized (see Section 4.4.2).
The figures show similar development and distribution of the force lines. To make a
clearer comparison, the forces in the first and second sub-panels of Cru-3B-S1, Test 3 and
Model 1195_3D are plotted together in Figure 5.17.

111
Figure 5.15 Sub-panel force development curves of Cru-3B-S1, Test 3

Figure 5.16 Sub-panel force development curves of Model 1195_3D


112
Figure 5.17 Sub-panel force development curves of Cru-3B-S1, Test 3 vs. Model
1195_3D

Since the models truly imitated the test specimens in the aspect that all the parts were
created in separate pieces and assembled together later, to have obtained the modelling
results that could perform appreciably well compared with the test results was
encouraging. The important parameters that also contributed to such good agreement
should be the specified friction coefficients for contacts ( = 0.3 ) and the C3D8RS
element type selected. In brief, the satisfactory results gave confidence in the modelling
approaches and parameters selected which were to be used further.

Models with other connection arrangementsalternating plate, aligned plate, and two bolts
per connecting facewere not created because it had already been demonstrated that these
connection types were not the most efficient and hence were not going to be extended to
multi-panel towers. Again, the objective of creating these reinforced one-panel FE models
was to find a simple and flexible way that enabled a successful creation and analysis of
reinforced multi-panel tower models, while also yielding reliable results.

113
5.5.2 Stage 3, Type-2 Model (using detailed bolted connections + beam
elements for sub-panel portions)
Continuing from Section 5.5.1, a search for a more economical model which would
produce the same accurate result was pursued. At this stage, the easiest means appeared to
be changing the member type or section of the sub-panel portions from continuum type to
beam type, while the detailed connection design remained the same. An example of one
single member is given in Figure 5.18(a). It shows that the portions for forming
connections which were solid sections and the sub-panel portions which were beam-type
members were in separate pieces. They were linked together by coupling constraints (the
thin grey threads in the figure), kinetic-typed, on all six degrees of freedom. Figure 5.18(b)
shows the complete figure after assembling. The profile of beam members for the sub-
panel portions is shown in Figure 5.19. Each beam member was subdivided into sixteen
beam elements. They were assigned with B31 (2-node linear beam in space) element type
in Explicit element library.

Because this model was aimed to help bridge to an even more economical model, only one
useful case was considered. That case was in general when the bolt and nut are loosely
connected. Here the bolt parts having the bolt head-to-washer clear distance of 1210 mm
were used again. The model was named 1210_Beam. Hence, the connections of this model
were identical to the connections of Model 1210_3D.

Figure 5.18 Model 1210_Beam: (a) Figure of one angle member and (b) Full assemblage
of the reinforced model
114
Figure 5.19 Profile and coordinates of beam members for sub-panel portions of Models
1210_Beam and Simplified_Beam

5.5.2.1 Result and discussion

The deformed shape and load-displacement curve of this model (Model 1210_Beam) are
given in Figures 5.20 and 5.21. The model failed in bending mode with a very small degree
of bending, similar to the previous models, although the element types used for sub-panel
portions were different. Load-displacement curves of this model and Model 1210_3D are
plotted together in Figure 5.21. They overlay each other perfectly, except for a tiny
difference at around the maximum load position. The maximum capacities of both models
were 303.6 kN (Table 5.6). In the aspect of running or calculation time consumption, while
this model (having target displacement of 8.0 mm and loading time of 0.034 sec) used 14
hours, Model 1210_3D (8.3 mm in 0.032 sec) used up to 31 hours.

Initially these very similar results were not expected, but the cause was likely to be because
both models had no imperfect geometry pattern introduced which resulted from the
complicated contact problem they had that prevented successful buckling analysis. It is still
possible to create the models or any models to have imperfect geometry, but the node
coordinates have to be specified manually, which would require much more consideration
and time.

115
Table 5.6 Capacities and capacity increase percentages of models with zero bolt
pretension

Connection Element Type Tension in Max Capacity Displacement Increased


Model
Design Type for Sub-Panels Bolts (kN) (kN) (mm) Capacity (%)
1210_3D Detailed Bolted Continuum -- 303.5 7.8 25.4
1210_Beam Detailed Bolted Beam -- 303.6 7.3 25.5
Simplified_Beam Simplified Beam -- 294.1 2.0 21.5

Figure 5.20 Deformed shapes of Models 1210_3D (left), 1210_Beam (middle), and
Simplified_Beam (right)

The above comparisons on capacities and calculation time of Model 1210_Beam and
Model 1210_3D indicated that beam elements were able to replace continuum elements for
sub-panel portions very well, given that the replacement still produced the same level of
result, i.e. strength and displacement, and saved considerable analysis time. Therefore, this
selection was applied to reinforced multi-panel tower models in the next chapter.

116
Figure 5.21 Load-displacement curves of Models 1210_3D, 1210_Beam, and
Simplified_Beam vs. Cru-2B-S1, Test 1

5.5.3 Stage 4, Type-3 Model (using simplified connections + beam elements


for sub-panel portions)
Although a considerable amount of time could be saved by using beam elements for the
sub-panel potions, the design of the connection was still too detailedinvolving many
individual parts and numerous interactions. A complete connection required a sizable space
for input data, and when more connections were required for a multi-panel model, the size
of input data would exceed the softwares limitation. Therefore, it was necessary to create
a new design of connection that required less space for input data, but was still able to
provide reliable results to the same level. Here details of the new (simplified) connection
will be given first, and then the result will be discussed.

5.5.3.1 Details of simplified design

The new connection design (Figure 5.22) used a simple locking system between Locking
Stripes on the component members and Grooves on the Enclosing Ring which together
functioned as bolts, bolt holes, and connectors respectively of a bolted connection.

117
The design used three locking stripes to represent the three rows of bolts used in the
experimental connections. The stripes had a dimension of 12 mm x 6 mm (width x
thickness). The width was equal to the diameter of the bolts. The stripes extruded out from
the front side of the member angle faces, termed Core part, and were at the same positions
as the bolts (Figure 5.22(a)). Each ring was an assembly of two parts, namely the grooved
locking part (Figure 5.22(b)) and the smooth wrapping part (Figure 5.22(c)), tied together
at four edges using tie constraints. Each groove had a dimension of 13.5 mm x 6.1 mm
(width x depth), hence allowing for a slip of 1.5 mm between the groove and the stripe.
Analogously, the locking stripes worked as the bolts (the media to transfer force) and the
grooves worked as the bolt holes (to keep the longitudinal slip within limit). Since the ring
wrapped around the component members, individual lateral movement of the component
members was prevented.

Figure 5.22 Components of a simplified connection: (a) the Core Part with Locking
Stripes; (b) the Grooved Locking Part; and (c) the Smooth Wrapping Part

The core parts, including the locking stripes, were defined as having elastic-perfectly-
plastic properties as before ( f y = 328 MPa and E = 205,000 MPa); but the grooved

locking parts and smooth wrapping parts were assumed to have elastic-only property (E =
118
205,000 MPa) to be stiffer, to prevent them from behaving plastically and opening up too
much. Any possible contacts were defined with two contact propertiestangential
behaviour and normal behaviour. The tangential behaviour used penalty friction
formulation with the friction coefficient of 0.3, and the normal behaviour used hard contact
pressure-overclosure mode. This new design used only eight interaction definitions for a
connection, where the old design used forty-eight. Again, since the parts were solid and
Explicit analysis had to be used, they were assigned with C3D8RS (8-node linear brick,
reduced integration, stiffness hourglass control) element type.

To form a complete longitudinal member, the connections core parts and the sub-panel
portions made of beam elements were connected together by coupling constraints. The
model was named Simplified_Beam. Its complete assemblage is shown in Figure 5.23.
Having no pretensioning force in the connections, the model was comparable to Model
1210_Beam and Model 1210_3D of Sections 5.5.2 and 5.5.1.2 respectively.

Figure 5.23 Model Simplified_Beam: (a) Figure of one angle member and (b) Full
assemblage of the reinforced model

5.5.3.2 Results and discussion

The model used a total of four runs to achieve the strength convergence of less than 3%
difference from two adjacent runs. The last run had target displacement of 4.8 mm and
loading time period of 0.072 sec. In Figure 5.20, its deformed shape showed only a very
119
small bending, similar to those of Models 1210_3D and 1210_Beam. In Figure 5.24 the
slip between the main and reinforcing members is shown (the ring connectors are hidden to
show the slip more clearly).

Figure 5.24 Slip between the two members of Model Simplified_Beam

The load-displacement curve of the model is compared with its counterparts in Figure 5.21.
The maximum capacity was 294.1 kN, or a 21.5% capacity increase, occurring at the
displacement of 2.0 mm. One reason that probably attributed to this low amount of
displacement was the design of the connectors where they completely encased the
component members for their whole length. This somewhat reduced the flexibility (or
length, in other words) of each sub-panel, where it was more (or longer) for the case that
used bolted connections. Therefore, the amount of displacement was less (see Table 5.6).
But the reason that caused the model to have the same level of capacity as those of Models
1210_Beam and 1210_3D should be because of its connections had no pretensioning
forces, just like its counterparts.

120
Sub-panel force curves of the model are plotted in Figure 5.25. From observation, forces
on the reinforcing member start to occur after 1.5 mm of displacement which is equal to
the slip distance allowed in the connections. This means, in that period the connection
design allowed the members to slip freely and at the same time all load applied was carried
by the main member only. Then after that, the stripes and grooves started to lock, and force
began to transfer to the reinforcing member.

Figure 5.25 Sub-panel forces of Simplified_Beam model

Finally, since the final strength result of the model was not very different from those of
Models 1210_3D and 1210_Beam, it proved that the simplified connection design created
had the potential to replace the detailed bolted connection design, under a condition that
pretensioning force must not be brought into consideration. The last run of this model
(having target displacement of 4.8 mm and loading time period of 0.072 sec) used only
half an hour to complete analysis by comparison with the last run of Model 1210_Beam
(8.0 mm in 0.034 sec) which used 14 hours. This indicated a significant time-saving
benefit in using the simplified connection design.

121
5.6 Conclusions
The objective of this chapter was accomplished by that a combined design approach to
modelling reinforced legs which saved considerable input data and analysis time while still
producing reliable strength results was found. The following are important findings of the
chapter.

1 Using buckling analysis on a perfect model in order to obtain and introduce imperfect
coordinates to the model is applicable only when the whole model is created in one
piece, such as the case of a single-angle model. Or if a model of a built-up member is
to be created, the connection type of the member must be welded-type so that the
model can be created accordingly using Tie Constraints in ABAQUS to tie the
component parts together to become completely one piece. The principle is that there
must be no slip allowed between the parts in the model, to have successful buckling
analysis. Therefore, in this development of the reinforced models where slip was an
important factor and must be allowed to happen, the models had to be created and
analysed in perfect geometry. As a result, it was necessary to accept the disadvantage
about some degrees of error of the modelling results as compared with the test results.

2 The default type of hourglass control that ABAQUS/Explicit assigns to continuum


elements is Reduced Stiffness type. If that is used, the FE model will have load-
carrying capacity much less than what should be obtained from test specimens of the
same category. It is crucial to change the default type to Stiffness or Enhanced type in
order to get appropriate results. Here the plain Stiffness type is recommended rather
than the Enhanced Stiffness type, as the latter tended to make the models become a
little stronger.

3 The general perception that using different element types (which in this place meant
3D continuum and beam element types) for sub-panel portions will make different
results is not correct (see Section 5.5.2), at least for this study where slip was a
requirement for the design of connections. With no buckling analysis, there is no
failure mode to be factored to form imperfect geometry shape for the models; and then
using continuum, shell, or beam elements for sub-panels will give almost identical
results. So, to spend least analysis time, beam elements should be used. To introduce
geometrical imperfection, node coordinates need to be specified manually, but before
that, a very careful consideration will be required.

122
4 The first connection design had bolts, connectors, main and reinforcing members
created in separate pieces. This design demanded numerous interactions to be defined,
which as a result required a very large space for input data. The second (simplified)
connection design had much fewer details. It used encasing rings to wrap around the
main and reinforcing members and unify them together. The rings had locking grooves
while the main and reinforcing members had locking stripes extruding out, to function
as bolt holes and bolts respectively in transferring forces. The second connection
design required much less space for input data. Also, in the case that the connections
are loose (bolts having no pretension), the second design proved to be able to replicate
the first design satisfactorily.

5 It may be possible to provide pretension to the simplified connection design, for


example, by applying spring elements to the encasing rings to make them squeeze the
longitudinal members tighter. That will make the reinforced legs become more
effective. However, it will require more input data, longer analysis time, computational
resources, and above all, making friction connections and maintaining the bolt
pretension is not practical in the field.

123
CHAPTER 6

6 Extension to multi-panel tower models Reinforced


compression leg
6.1 Introduction
As illustrated in Chapter 5, the modified modelling method using the combination of beam
elements and continuum core parts had the potential to be applied in examining the
effectiveness of leg reinforcing method in multi-panel tower models.

The next question was how many panels would be sufficient to gauge and obtain the real
potential of the method, as it was a concern that having too many panels involving a lot of
interaction definitions would require the input file to be larger than that allowed by
ABAQUS, and take a very long analysis time. The answer was found by observing the
geometry and member force distribution of two real towersIndian Tower (Hayward &
Weare, 1989) and Silva Tower (Silva et al., 2005)to be discussed in Section 6.2, together
with considering the additional capacity demand of 100% more than the original design
capacity as likely to be the demand required by the transmission tower industry. The
observation results showed that both towers needed leg reinforcement for at least six to
seven panels. Therefore, the models to be created in this study part must have at least six
panels, and once the compression leg was reinforced for six panels, the design capacity of
the models should double.

Another observation from the same two tower models above was that the reinforcement
had to be extended to and terminated in the zone where the bracing members were
crossing-typed (X-typed). This observation point is believed to be very common for the
majority of transmission towers. Hence, it might be enough to adopt only cross-bracing
type for the bracing members in this study. When the bracings are crossing-typed, the
forces on the leg at adjacent panels must of course be different, but will gradually increase
from top panel to bottom. In this case if we reinforce the leg by providing two connectors

124
per panel, it is expected that in each panel the upper connector would carry shear force that
is higher than the lower one.

However, if the tower happens to use Warren-bracing type or happens to have the
reinforcement terminated in the zone where Warren-bracing type exists, the actual force
panel consists of two panels. (These two adjacent panels carry the same amount of force,
but are separated by a secondary bracing member.) In this case, when the leg is reinforced
and two connectors are provided per panel again, there are four connectors to transfer the
force in each actual force panel. It was suspected that most of the force transfer amount
would occur at the upper two connectors only. If this perception is correct, even though the
force increments created by the use of Warren-bracing pattern are high, and higher than
those using cross-bracings, if the transfer relies mainly on two connectors out of four, it is
likely that the effectiveness of the leg reinforcing method would be lower.

Hence, it was believed that the arrangement (or type) of the bracing members in the zone
of termination of the reinforcing member played a significant role on the performance of
the leg reinforcing method. It was therefore necessary to consider using both cross-bracing
type and Warren-bracing type to study their effects in this study part.

The second factor that might have a moderate to high influence on the reinforcing
methods effectiveness was the slenderness of the panel being considered. The term
moderate influence might be appropriate to use when the slenderness ratio of an existing
leg member is high, say L / r 100, and term high influence when the ratio is lower, say
L / r 100.

They were termed so in regard to the fact that a compression member with a high
slenderness ratio would have a low compressive strength. In parallel, the amount of
additional strength that we can expect from the reinforcing would also be low
proportional to the initial design strength. Therefore, the concentration of force transferred
at the top connector(s) of each actual force panel would not be so considerable, and a
simple and short connector would be adequate. Also, any extra development length
(extension of the reinforcement for a few panels more above the panel where it could be
technically terminated for strength requirements) may not be necessary. All the
expectations mentioned here would be opposite if the leg member has a low(er)
slenderness ratio. For this reason, the study models were varied with three leg slenderness
ratios, namely L / r = 80, 100, and 120.
125
To cover the effects of the two bracing patterns and three slenderness ratios mentioned
above, four reinforced tower models were created and studied. While the evaluation of the
leg reinforcing method and design recommendations were the major issues which would be
drawn from the four models, they happened at the later stages. The same level of
importance was first given to developing the proper modelling techniques that must yield
reliable results. In this process, the same four tower models were also used, but without leg
reinforcement. The correctness of the techniques was checked by observing the buckling
mode and strength of the compression leg as compared to the design code. If the model
strength appeared to be somewhat lower than the code strength, that means the techniques
used provided some degrees of conservativeness, which was more preferable than the
opposite.

The study then moved forward to using normal structural elastic analysis software to create
elastic reinforced tower models. This allowed us to see how different the force distribution
on the reinforced leg of the elastic models and FE models would be. The main objective
was to examine the possibility of using elastic models to find the maximum connection
shear force.

At the end of the chapter, the application of Browns method in calculating the capacity of
reinforced legs will be given, plus two calculation examples provided. The norm of the
capacity increase potential from using his method will be presented. The author believes
that his method is reliable and will give conservative results.

6.2 Observation of Indian and Silva Towers


It is useful to describe the beginning step where we used Indian Tower and Silva Tower to
help make a decision about how many panels the models to be created in this part of study
should have. Also, we will explain how we identified the position to terminate the
reinforcement according to a strength increase demand.

6.2.1 Indian Tower


Figure 6.1(a) shows the original profile in elevation of a 45 m high transmission tower in
India. This profile was taken from Steel Detailers Manual by Hayward & Weare (1989).
The leg members of the lowest five actual force panels being K-braced in the figure were
made of 120 120 15 EA ( ry = 23.5 mm), with the unsupported length of 2000 mm.

The slenderness ratio was equal to 85.1. Since the ratio was close to 80 which was the ratio
126
that had been considered exclusively in Chapters 3, 4, and 5 and we were familiar with
from using 75 75 6 EA with the unsupported length = 1.2 m, we then imitated and re-
proportioned the profile and dimensions of the original Indian Tower into a new tower with
the unsupported length for the lowest five actual force panels equal to 1.2 m. The new
tower is shown in Figure 6.1(b). It was reasonably similar to the original tower and it
certainly had enough panels to keep the force increments low. This is an important issue
because if the tower has only three or four or five panels in total, the increments will be
large, and its legs will need reinforcement for only two or three panels to accomplish the
expected strength increase, which is unrealistic. For the sections used for the legs, while
the original tower used three different sections, the new tower was configured with 75 75
6 EA only for simplicity. The straight part at the top of the original tower was not
included. In this investigation, truss analysis was used, and the model used in truss analysis
was that in Figure 6.1(c) with a few secondary bracing members used as horizontal
diaphragms. (The model in Figure 6.1(b) cannot be solved by truss analysis because it has
many redundancies.)

Since the information about the loads and load cases of the original tower was not
available, to be able to carry out the investigation, some assumptions were made. First, for
any four-leg tower with square cross-section, it was assumed that the critical case for the
legs occurs when only one leg is in compression. It means that the opposite leg in the
oblique direction must be in tension, and the other two legs carry no force. To create this
condition, the resultant of the applied loads can only be in the 45 oblique direction. Two
horizontal 40 kN loads were applied to the top end of the new tower. (Figure 6.1(d)) (This
load magnitude could be of any value because its corresponding member forces were not to
be used directly.)

127
Figure 6.1 Indian Tower: Elevation views and dimensions of (a) original tower; (b) the
model used in investigation; and (c) the model used in truss analysis; (d) Loads applied to
the tower model
128
The corresponding panel forces on the compression leg from truss analysis are given in
Table 6.1, Column 4. Since 75 75 6 EA section was planned to be used for the legs, its
design strengths for different unsupported lengths were calculated using AS 3995 (1994)
Standard. The calculated values are in Column 10. From observation of the panel forces in
Column 4, Panel 1 (the lowest panel) carried the highest force, at 227.9 kN, and therefore it
was the critical panel. Because we wanted the force on Panel 1 to be equal to the design
strength of 182.3 kN in Column 10, the 227.9 kN was reduced to 182.3 kN by multiplying
it with a multiplier equal to 182.3/227.9, or 0.8. The multiplier in general terms is equal to
the design strength of the critical panel divided by the force on the critical panel obtained
from analysis. Forces on the other panels in Column 4 were also multiplied by the same
multiplier to get the forces at the initial design stage. They are given in Column 5. This
initial stage was also called 0% target capacity increase. The forces at the initial design
stage were again increased by 50%, 60%, 70%, and 100% to obtain the forces at the
expected 50%, 60%, 70%, and 100% capacity increase stages as shown in Columns 6 to 9
respectively.

Table 6.1 Panel forces on the compression leg of Indian Tower model

Indian Tower
Panel Length L Panel Force % of Capacity Increase Demand 0.9Nc of
No. of Unsupported from 0 50 60 70 100 75x75x6EA
Panel Length Assumed Adjusted or Expected Panel Force fy=320MPa
(m) (m) Loads (kN) (kN) (kN)
1 2.40 1.20 227.94 182.3 273.5 291.7 309.9 364.6 182.3
2 2.40 1.20 219.83 175.8 263.7 281.3 298.9 351.6 182.3
3 2.40 1.20 210.06 168.0 252.0 268.8 285.6 336.0 182.3
4 2.40 1.20 198.07 158.4 237.6 253.5 269.3 316.8 182.3
5 2.40 1.20 183.00 146.4 219.5 234.2 248.8 292.7 182.3
6 2.00 1.00 172.38 137.9 206.8 220.6 234.4 275.7 202.9
7 1.70 0.85 161.90 129.5 194.2 207.2 220.1 259.0 215.9
8 1.50 0.75 138.39 110.7 166.0 177.1 188.2 221.4 223.4
9 1.30 0.65 126.26 101.0 151.5 161.6 171.7 202.0 229.9
10 1.10 1.10 113.22 90.6 135.8 144.9 153.9 181.1 193.0
11 0.90 0.90 83.58 66.8 100.3 107.0 113.6 133.7 211.8
12 0.85 0.85 62.82 50.2 75.4 80.4 85.4 100.5 215.9
13 0.80 0.80 39.56 31.6 47.5 50.6 53.8 63.3 219.7
14 0.74 0.74 13.63 10.9 16.4 17.4 18.5 21.8 224.1

In other words, the force values in Columns 6 to 9 were the forces that the panels needed to
carry if the reinforcing could make the tower accomplish the respective capacity increase
129
demands. By comparing them with the design strengths of the single section in Column 10,
we could identify which panels needed to be reinforced, i.e. the panels where their
expected forces were higher than their single-section design strengths. For example, in the
case of 50% capacity increase demand, Panels 1 to 6 had expected forces higher than their
single-section design strengths, and hence, they needed to be reinforced. The double-
underlines were used to separate the panels that needed reinforcement from those that did
not.

An interesting observation was that while aiming for a 60% capacity increase needed six
panels of reinforcement, if the reinforcement was extended for one more panel to cover
Panel 7, it was possible to aim for as much as a 100% increase. (In reality, the sections for
the legs tend to be smaller in the upper portions. In that case the gap of improvement will
be smaller.) If achieving the maximum potential is the main goal of the reinforcing work,
the 100% increase should be aimed for; as was done in this study. Another interesting point
was that the pattern of the main bracing members over the zone of Panels 6, 7, and 8 where
the reinforcement would be terminated was Cross-bracing; while in fact there were
secondary bracings members used too.

6.2.2 Silva Tower


Silva Tower in Figure 6.2(a) was taken from the 40 m high transmission tower that Silva et
al. used in their study. Again, the towers original dimensions were re-proportioned to suit
the use of 75 75 6 EA section and L / r = 80. The new towers layout is shown in
Figure 6.2(b). The tower had ten actual force panels as indicated by the numbers on the
left. While the first actual force panel consisted of five short panels of equal force
magnitude, the second and the third actual force panels consisted of four panels, the fourth
and the fifth consisted of three panels, the sixth had one, and the seventh to the tenth each
consisted of two panels. To prevent confusion, and because the simplified tower as shown
in Figure 6.2(c) was used in truss analysis, the panel numbers were renumbered as
displayed on the right of the tower. The same two assumed 40 kN loads were applied to the
top of the new tower in the same directions as before. The same presentation of panel
forces on the compression leg at different expected capacity increases is presented in Table
6.2.

130
Figure 6.2 Silva Tower: Elevation views and dimensions of (a) original tower; (b) the
model used in this investigation; and (c) the model used in truss analysis

For Silva tower, Panel 2 was the critical panel because although it carried the same amount
of force as Panel 1, it had longer unsupported length (see Columns 2, 3, and 11). The table
shows that if we wanted to increase the capacity of the tower for 60%, 70%, and 100%, the
reinforcement needed to be applied to cover Panels 7, 8, and 9, respectively. (These panels
were the actual force panel numbers 4, 5, and 6.) Over this zone, the pattern of the main
bracing members was Cross-bracing again.

131
Table 6.2 Panel forces on the compression leg of Silva Tower model

Silva Tower
Panel Length L Panel Force % of Capacity Increase Demand 0.9Nc of
No. of Unsupported from 0 50 60 70 80 100 75x75x6EA
Panel Length Assumed Adjusted or Expected Panel Force fy=320MPa
(m) (m) Loads (kN) (kN) (kN)
1 3.30 1.10 226.94 182.3 273.5 291.7 309.9 328.1 364.6 193.0
2 2.40 1.20 226.94 182.3 273.5 291.7 309.9 328.1 364.6 182.3
3 2.40 1.20 215.39 173.0 259.5 276.8 294.1 311.4 346.0 182.3
4 2.40 1.20 215.39 173.0 259.5 276.8 294.1 311.4 346.0 182.3
5 2.40 1.20 199.42 160.2 240.3 256.3 272.3 288.3 320.4 182.3
6 2.40 1.20 199.42 160.2 240.3 256.3 272.3 288.3 320.4 182.3
7 2.70 0.90 181.69 146.0 218.9 233.5 248.1 262.7 291.9 211.8
8 2.70 0.90 164.77 132.4 198.5 211.8 225.0 238.2 264.7 211.8
9 1.00 1.00 146.49 117.7 176.5 188.3 200.0 211.8 235.3 202.9
10 2.00 1.00 122.37 98.3 147.4 157.3 167.1 176.9 196.6 202.9
11 2.00 1.00 96.38 77.4 116.1 123.9 131.6 139.4 154.8 202.9
12 2.00 1.00 49.45 39.7 59.6 63.6 67.5 71.5 79.4 202.9
13 2.00 1.00 23.46 18.8 28.3 30.2 32.0 33.9 37.7 202.9

6.2.3 Summary
The observation on the Indian and Silva towers indicated that the models to be created and
studied in this chapter should have at least six or seven actual force panels, and they
possibly could achieve 100% capacity increase when those actual force panels were
reinforced. This consideration was important to having a useful and realistic study. It was
decided that the models would have nine panels as will be discussed in the next sections.

6.3 Geometries of tower models


The tower models in the following sections were designed to have three legs, instead of
four, in order to save computational resources and analysis time. The cross-section had a
right-triangular shape. The two faces on the perpendicular sides were vertical, and the face
on the opposite side (hypotenuse) was inclined. They formed a tapered shape from the base
to the top. Two configurations were considered for the models, based on the pattern of the
bracing members on the two vertical faces. The first configuration had cross-bracing
pattern, and the second had Warren-bracing pattern. (Figures 6.3 and 6.4) Nine panels were
used as mentioned in the summary of Section 6.2. Two horizontal point loads of equal
magnitude were applied to the top right-angle corner in the perpendicular directions that
created 1) compression on the vertical (right-angle corner) leg, 2) symmetrical force

132
distribution on the bracing members of the two vertical planes, and 3) equal tensile forces
on the two inclined legs.

Figure 6.3 Cross-braced tower model: (a) general view; (b) top view; and (c) vertical face

As 75 75 6 EA section was used in the experimental part, it was used here again
exclusively for the vertical (main) leg and its reinforcement. As mentioned earlier it is
likely that in general the reinforcement would be terminated in the zone where the pattern
of the bracing members of the zone is cross-bracing, so the study then focused on the
cross-bracing pattern. Three models with the cross-bracing pattern as in Figure 6.3 were
created with three different slenderness cases, based on the length of the common panels
of the vertical leg. (The three models had the same dimensional scale and length ratios
among the members, but different length magnitudes of the same members.) The common
panels of the vertical leg here were Panels 1 to 6 in the picture, where their lengths were
equal. For the first slenderness case, the length of the common panels was 1.2 m. It was
named S80 according to the L / r ratio which was 81.6, or approximately 80. The second
case, namely S100, had the length of the common panels equal to 1.5 m ( L / r = 102.0
133
100). And the third case, namely S120, had the length equal to 1.8 m ( L / r = 122.4
120). For Warren-bracing pattern, only one model was used and the slenderness case was
S80. The models were named Cross S80, Cross S100, Cross 120, and Warren S80. See
Table 6.3 for dimension details.

On the inclined face (hypotenuse), a K-bracing pattern was used. Since it had a symmetric
arrangement and its plane was perpendicular to the resultant of the two applied loads, the
K-bracing arrangement ensured that there would be no force distributed on its bracing
members. It is noted here that using cross-bracing arrangement on this plane would provide
the same result; while using Warren-bracing arrangement would create unsymmetrical
force distribution and consequently might cause the structure to twist, which is a situation
too complicated to analyse at this stage.

Figure 6.4 Warren braced tower model: (a) general view; (b) vertical face; and (c) inclined
face

134
Table 6.3 Basic dimensions and panel lengths of the four tower models

Model
Cross S80 & Cross S100 Cross S120
Warren S80
Length (m)
Base Width 2.32 2.90 3.48
Top Width 0.80 1.00 1.20
Total Height 10.24 12.80 15.36
1-6 1.20 1.50 1.80
7, 8 1.04 1.30 1.56
9 0.96 1.20 1.44
Panel No. 10 1.31 1.64 1.97
11 - 15 1.20 1.50 1.80
16, 17 1.04 1.30 1.56
18 0.96 1.20 1.44

6.4 Design concept of the models (use reinforcement for the


vertical leg only)
For real towers, if they are to be reinforced by adding/stitching additional members parallel
to the existing members, not only do all four legs need to be reinforced, but also some
bracing members. (The term some bracing members is used because it is possible that
the designer will replace the inadequate members with larger sections, rather than
reinforcing them).

Adopting such practical aspect to this study can be highly computationally expensive or
require an input file larger than that allowed by the software. Hence, it was decided to use
a reinforcing member on the vertical compression leg only, which made it become a
compound member. The other membersthe inclined tension legs and all bracing
memberswere not reinforced, so they were sized beforehand to have sufficient strength
on their own for the expected forces.

6.5 Determining target member forced by truss analysis and


location to terminate the reinforcement
With the detailed geometries and bracing members arrangements of all four models now
available, elastic truss analysis was used to determine axial member forces. Using truss
analysis means that the towers could have pin supports only. A section shape and size was
defined for every member. (The section shape and size selected had no importance in this
place because truss analysis was used, and the purpose of its use was to obtain the member

135
forces only.) Two assumed 20 kN loads were applied at the top end in the right directions
as shown in Figure 6.3. The corresponding panel forces on the vertical leg from analysis of
Cross S80 and Warren S80 models are given in Column 3 of Tables 6.4 and 6.5
respectively. The complete sets of forces and adjusted forces of the four models are given
in Appendix, Section B3.

Table 6.4 Panel forces on the compression leg of Cross S80 model corresponding to the
assumed loads, and when adjusted for different strength increase demands, in
comparison with the design buckling strengths

Cross S80 Model


Panel L Panel Force % of Capacity Increase Demand 0.9Nc of
No. Unsupported from 0 50 60 70 100 75x75x6EA
Length Assumed Adjusted or Expected Panel Force fy=320MPa
(m) Loads (kN) (kN) (kN)
1 1.20 172.33 182.3 273.5 291.7 309.9 364.6 182.3
2 1.20 163.58 173.0 259.6 276.9 294.2 346.1 182.3
3 1.20 153.48 162.4 243.5 259.8 276.0 324.7 182.3
4 1.20 141.35 149.5 224.3 239.2 254.2 299.1 182.3
5 1.20 126.25 133.6 200.3 213.7 227.0 267.1 182.3
6 1.20 107.26 113.5 170.2 181.5 192.9 226.9 182.3
7 1.04 84.38 89.3 133.9 142.8 151.7 178.5 199.0
8 1.04 56.43 59.7 89.5 95.5 101.5 119.4 199.0
9 0.96 20.10 21.3 31.9 34.0 36.1 42.5 206.5
10 1.31 -87.3 -92.4 -138.5 -147.8 -157.0 -184.7 -

Note: Panel 10 is the maximum loaded panel on the tension leg

136
Table 6.5 Panel forces on the compression leg of Warren S80 model corresponding to the
assumed loads, and when adjusted for different strength increase demands, in
comparison with the design buckling strengths

Warren S80 Model


Panel L Panel Force % of Capacity Increase Demand 0.9Nc of
No. Unsupported from 0 30 40 50 100 75x75x6EA
Length Assumed Adjusted or Expected Panel Force fy=320MPa
(m) Loads (kN) (kN) (kN)
1 1.20 168.14 182.3 237.0 255.2 273.5 364.6 182.3
2 1.20 168.14 182.3 237.0 255.2 273.5 364.6 182.3
3 1.20 148.02 160.5 208.6 224.7 240.7 321.0 182.3
4 1.20 148.02 160.5 208.6 224.7 240.7 321.0 182.3
5 1.20 117.93 127.9 166.2 179.0 191.8 255.7 182.3
6 1.20 117.93 127.9 166.2 179.0 191.8 255.7 182.3
7 1.04 72.36 78.5 102.0 109.8 117.7 156.9 199.0
8 1.04 72.36 78.5 102.0 109.8 117.7 156.9 199.0
9 0.96 17.34 18.8 24.4 26.3 28.2 37.6 206.5
10 1.31 -89.24 -96.8 -125.8 -135.5 -145.1 -193.5 -

It can be seen from Column 3 of Tables 6.4 and 6.5 that, for the vertical leg, the force was
highest on Panel 1 (the lowest panel) and gradually decreased on the upper panels. Thus, it
was sensible to expect Panel 1 to fail first by buckling, and as that happened, having lost
the strength of the compression leg, it should mean the failure of the structure as well.

The member force results obtained from analysis were not used directly. They were
adjusted to prepare for the expected member forces at different percentages of target
capacity increases. The same process as used in Section 6.2 was used here again.

In determining where to terminate the reinforcement of the vertical leg, the same method
as used in Section 6.2 was applied. For the Cross S80 model, if the target capacity was
100%, Panel 7 was the first panel from the base that had the expected force to carry (178.5
kN) less than its single-section design strength (199.0 kN). Therefore, the vertical leg
would be reinforced for the lowest six panels (Panels 1 to 6). After FE analysis, in case that
the target capacity increase could not be achieved, which means that the model
experienced a significant shear lag, the reinforcement would be extended for one or two
panels and the model would be reanalysed to determine the new capacity.

137
6.6 Unreinforced pilot models
After having set up the concepts in creation of the models that only the vertical leg would
be reinforced, where to stop the reinforcement, and that the other members were to be pre-
designed, the next steps were to choose appropriate member types, member sections and
support conditions. However, before immediately embarking on constructing and studying
the reinforced models, it was necessary to develop the modelling methods that could
produce reliable results. For this purpose, a number of pilot models were created and
analysed. The pilot models had identical dimensions and configurations to the reinforced
models, but without the reinforcing member. Without the reinforcing member, the pilot
models had fewer members and parts and no interactions for the connections involved.
Hence, they required considerably less computational resource and run time, which
allowed for faster evaluation and adjustments.

Two groups of pilot models, namely No Cores and Cores, were used. For No Cores
group, every panel of the vertical leg used only beam elements, and there were no core
parts involved. (See the definition of Core Part and its use in Section 5.5.3.1) The
appearances of No Cores Cross and No Cores Warren models are identical to those in
Figures 6.3 and 6.4 respectively. For Cores group, every panel, except Panel 9, used a
combination of beam elements and core parts, similar to that developed in Chapter 5. The
core parts were used as a part of the existing leg and reinforcing member at the locations of
the connections (see Section 6.6.5 for detailed configuration). They were designed to be
used with the encasing- ring connectors to form the slip connections. (Again, see Section
5.5.3.1 for the details of the slip connections and assembling.)

No Cores models were created and analysed, then the resultant buckling mode and
buckling strength were evaluated. A series of modifications were carried out until the
buckling mode was near the expected mode which was buckling at Panel 1 because it
carried the highest compression force, and until the buckling strengths were close to the
code strengths.

The modelling and loading methods that worked with the No Cores models were then
applied to Cores models in exactly the same way. Having everything the same, both No
Cores and Cores models should have the same capacity. However, because the Cores
models used two element types, some difference was expected. The following subsections
describe the details of the pilot models.

138
6.6.1 General material properties
The material properties of the members and parts were the same as those in Chapter 5. All
leg members including the core parts on the vertical leg and its reinforcement, and all
bracing members had elastic-perfectly-plastic property, E = 2.05 105 and f y = 328 kN.

The encasing-ring connectors had purely elastic property, E = 2.05 105.

6.6.2 Support conditions


Both No Cores and Cores models used pinned supports (U1 = U2 = U3 = 0), which were
intended to be the same as that used in the truss analysis, Section 6.5, so that forces would
distribute in the same manner. In addition, it was to make sure that no additional strength
was provided by other types of support and buckling would happen at the lowest panel of
the vertical leg.

A set of restraints against rotation (UR2 = 0) were applied to the ends of each panel of the
vertical leg. (They are dark spots in Figures 6.3 and 6.4.) Their purpose was to prevent the
vertical leg from freely turning around itself (this does not mean twisting or bucking) while
loads were applied, which otherwise could lower the vertical legs true strength potential.

6.6.3 Type of analysis procedure, loading time period, and loading method
Because of the highly complicated contact problem from numerous surface-to-surface
interactions encountered in the reinforced models, it was necessary to use the Dynamic
Explicit analysis procedure. Hence it was also used for the unreinforced models. However,
to use dynamic explicit analysis to imitate static analysis, the total kinetic energy of the
system must be kept low, compared with the total internal energy. This was done by
specifying a sufficiently long imaginary time period for the loading process to finish. (This
type of analysis is sometimes called Quasi-Static analysis.)

Loads were applied by specifying a target displacement value of the top right-angle corner
node of the towers in the X and +Z directions that must be achieved at the end of the time
period specified earlier. Smooth amplitude function was also used to smooth the loading
process, which in turn created smooth analysis and results.

These two valuesloading time period and target displacement amountwere not known
at the beginning. The first trial set was done by using a short time period ( t1 ) and a quite
large target displacement ( d 1 ). Checking the force-time curve of the critical panel told us

139
at what time the panel buckled. This time, or t1 ' , was used to check back its corresponding
amount of displacement ( d1 ' ). Then in the second trial, a time period ( t 2 ) longer than t1 '
and a displacement ( d 2 ) slightly larger than d1 ' were used. (Usually a t 2 between 1.5 t1 '
and 2 t1 ' , and d 2 equal to d1 ' plus 5 - 10 millimetres more were used.) By this means, the
kinetic energy could be reduced significantly. A new force-time curve was plotted to find
t 2 ' and d 2 ' . All these steps were repeated for a few times until the buckling mode and
strength became constant. For the third and fourth trials and so on, the target displacement
values were likely to be the same, but we still needed to increase their time periods to
check the convergence.

6.6.4 Member types and shapes and tie addition


The vertical leg and inclined legs used beam type elements, while the bracing members
used truss type elements. (During the first stage of creating the models, all bracing
members were beam elements. Hence, they had beam-to-beam connection type with the
legs. The length of analysis became less than the actual panel length. This resulted in
altering the buckling mode of the tower from the expected mode, and created buckling
strength of the critical leg panel significantly higher than the code strengths expected.)

To correct the above condition, the bracing members were redefined using truss typed
elements. Since they could not transfer moment, they ensured that the critical analysis
length was the same as the panel length. Although, the buckling strengths dropped down to
the level close to the code strengths, the convergence was not goodthe longer loading
time period was used, the lower the buckling strength was produced. Observations revealed
that using truss elements for the bracing members had one disadvantage, which was that
they would go out of plane if there were more than one truss element joined together at a
node at the middle. They were Cross-bracings or K-bracings in this study. As the joined
nodes moved out of plane more, forces on the bracing members were not in the directions
they should be. This encouraged the models to lose their stability and in the end lose their
potential capacity. Figure 6.5 illustrates the bracing members on a vertical face and the
inclined face of Cross S120 model going out of plane.

140
Figure 6.5 Truss-typed bracing members on (a) the vertical face and (b) the inclined face
of unreinforced No Cores Cross S120 model going out of plane

To prevent the truss-typed bracing members from going out of plane, ties were added to
the models. These ties might be considered as another set of secondary bracing members
whose duty was to brace the structures only (carrying zero force). These ties were defined
as truss elements as well. Figures 6.6 shows the positions of ties added to the Cross and
Warren models. The analysis results converged well. The models failed by the wanted
mode which was buckling of Panel 1 (Figure 6.7). The buckling strengths from analysis are
compared with the calculated strengths in Table 6.6. Their differences were small for S120
models, but became larger for S100 and S80 models.

141
Figure 6.6 Ties added to the bracing members on the vertical and inclined faces to
prevent them from moving out of plane for (a) Cross models and (b) Warren models

Figure 6.7 Failure mode of (a) unreinforced No Cores Cross models and (b)
unreinforced No Cores Warren models when having additional ties

142
Table 6.6 Comparing design buckling strengths and reduced analysis buckling strengths
to check the validity of the created models

Design or Reduced Analysis Buckling Strength,


Nc, at Panel 1 (kN)
Cross Warren
S80 S100 S120 S80
AS 3995 182.3 144.3 102.7 182.3
Unreinforced No Cores 242.6 167.1 117.6 232.7
Unreinforced No Cores + Imperfection
L/590 170.3 -- 95.2 --
L/1000 185.9 137.3 101.4 179.8
L/1500 198.0 -- 104.9 --
Unreinforced Cores 128.4 119.0 109.7 129.7

To check the workability of all the models developed for the last time, three initial
imperfections based on L / 590 , L / 1000 , and L / 1500 were applied to the models. While
the new buckling strengths of L / 590 case were slightly lower than the code strengths, and
vice versa for L / 1500 case, those of the L / 1000 case matched very well with the code
strengths (see Table 6.6). These latest outcomes ensured that the modelling methods that
had been developed were in the right direction and could be applied confidently for the
reinforced models.

To create the angle shape for the vertical leg, an arbitrary shape function was used with
nine integration points to constitute a 75 x 75 x 6 EA section. The points coordinates in
the local axis system are shown in Figure 6.8(a). This section was used for every panel of
the vertical leg. Since the members were beam-typed, their section orientation was
important. In the global axis system, they were assigned with the tangent vector t = 0, 1, 0
which was parallel to the axis of the leg in the vertical direction, and the first normal vector
n1 = 1, 0, 0 which was on a vertical plane of the tower and perpendicular to the tangent
vector (Figure 6.8(b)).

143
Point 1-coordinate 2-coordinate Segment Thickness
(m) (m) (m)
1 -0.01797 0.05403 - -
2 -0.01797 0.03603 1-2 0.006
3 -0.01797 0.01803 2-3 0.006
4 -0.01797 0.00003 3-4 0.006
5 -0.01797 -0.01797 4-5 0.006
6 0.00003 -0.01797 5-6 0.006
7 0.01803 -0.01797 6-7 0.006
8 0.03603 -0.01797 7-8 0.006
9 0.05403 -0.01797 8-9 0.006

Figure 6.8 (a) Local coordinates of the integration points and segment thickness used in
defining 75 75 6 EA section using arbitrary shape function; (b) Orientations of the
vertical leg and the inclined legs with respect to the global axis system; (c) Showing seeds
that subdivide a member into a number of elements

For the inclined legs being tension members, any shape applied presumably should provide
the same result. Hence, a circular hollow section (CHS) was chosen, for symmetrical
advantage, based on its design strength versus the expected force to carry and L / r 150
criterion for leg members. Every panel used the same section size. Because they were also
beam elements, the orientation must be assigned. Both legs were assigned with the same n1

144
value (0.703244, 0.104388, -0.703244) which was calculated from, and equal to, the
normal vector of the inclined face.

Primary bracing members, or bracing members, were those that transferred forces among
the legs. They were defined with CHS sections as well. Their section selections were based
on the required design strength and L / r 200 criterion. However, because the members
were truss elements, cross-sectional areas were the only entry data needed. The members
were sorted into groups, and after it had been found out which section size to use for each
individual member, the biggest area of each group was entered.

Secondary bracing members were the members that, from elastic truss analysis, carried
zero force, but were added to brace other members to reduce the unsupported lengths of
those members. Their sections were selected based on L / r 250 criterion. The members
were truss typed elements, sorted into groups, and the biggest cross-sectional area of each
group was entered, similar to the case of the primary bracing members. The sections and
areas assigned to the members are summarized in Table 6.7. The cross-sectional details
were entered for the legs being beam elements, while only the cross-sectional areas were
entered for all bracing members being truss elements. For the details of member selections
of the four models, see Appendix, Section B4.

In the mesh module, each panel of the vertical leg and inclined legs were subdivided into
sixteen and eight elements respectively (Figure 6.8c), and they were defined as explicit
linear beam elements in space, Code B31. (Prasad-Rao and Kalyanaraman (2001) reported
that eight to ten elements were enough to give accurate results.) For the bracing members,
all used one truss element from joined node to joined node.

145
Table 6.7 Sections and areas assigned to the members according to their groups of
arrangement similarity or functions

2
Element Section (Area mm )
Member Group on Model Type Cross
S80 S100 S120
Vertical Leg & Reinforcement Beam 75x75x6 EA 75x75 6 EA 75x75x6 EA
Inclined Legs Beam 48.3x5.0 CHS 48.3x4.0 CHS 42.4x3.2 CHS
26.9x3.2 CHS 26.9x3.2 CHS 33.7x2.6 CHS
All Braces on the Two Vertical Faces Truss
(238) (238) (254)
21.3x3.2 CHS 26.9x3.2 CHS 33.7x2.6 CHS
Horizontals of K-bracings Truss
(182) (238) (254)
26.9x3.2 CHS 33.7x2.6 CHS 42.4x2.6 CHS
Diagonals of K-bracings Truss
(238) (254) (325)
21.3x3.2 CHS 26.9x3.2 CHS 33.7x2.6 CHS
Ties Truss
(182) (238) (254)

Warren S80
Vertical Leg & Reinforcement Beam 75x75x6 EA
Inclined Legs Beam 48.3x5.0 CHS
42.4x3.2 CHS
Diagonals on the Two Vertical Faces Truss
(394)
26.9x3.2 CHS
Horizontals on the Two Vertical Faces Truss
(238)
21.3x3.2 CHS
Horizontals of K-bracings Truss
(182)
26.9x3.2 CHS
Diagonals of K-bracings Truss
(238)
21.3x3.2 CHS
Ties Truss
(182)

6.6.5 Unreinforced models with cores


Up to now, the modelling methods developed from unreinforced No Cores models were
capable of producing the results as wanted. Therefore, they should be on the right track and
well applicable for the next study process. However, we needed to know when the core
parts were used as part of the vertical leg, would the strength results be lower or higher
than without the core parts. It would be desirable for them to be equal or lower, for the
sake of safety when actually having to assess the results of the reinforced models in the
next stage.

Out of a total of nine panels, eight panels had core parts assembled with, two cores each
panelthe upper and the lower. Each core was made of continuum elements that represent
a 150-mm-long 75 x 75 x 6 EA section with locking stripes. They were used as part of a
part of the existing leg and reinforcing member at the locations of the connections. The
146
locations were set with the same generalization for every panel and every model as
follows. The upper cores centre was 200 mm below the upper panel point, the lower
cores centre was below the upper cores centre by half of the panel in length (see Figure
6.9). To assemble the complete vertical leg, the continuum cores and beam-element
members were connected together using kinetic-typed coupling constraints with all degrees
of freedom constrained. The vertical beam members were subdivided into elements (or
segments) with the numbers as shown.

Figure 6.9 Unreinforced Cores model

Including geometrical initial imperfection into the models had been considered. However,
later it was neglected, because even when it was not included, the models had lower
buckling strengths than those from the design code, as shown previously in Table 6.6.

6.7 Reinforced models


6.7.1 Reinforcing member
The reinforcing member had eight panels in total. It was assembled from beam members,
defined with the arbitrary 75 x 75 x 6 EA section, and the cores, two cores per panel,
147
located at the same positions as those on the main leg. Their orientation was assigned with
the tangent vector t = 0, 1, 0 and the first normal vector n1 = -1, 0, 0 in the global axis
system. Each beam member was subdivided into eight elements. Fig 6.10 shows where the
reinforcing member was added. The wrapping connectors are not displayed here for clarity.

Figure 6.10 Outlook when the reinforcing member was added next to the existing leg

6.7.2 Cases, interaction assigning, runs, convergence, and forces


A 100% strength increase of the compression leg (or might be stated as 100% capacity
increase of the whole structure) was set as the aim to achieve for every model. First, we
adjusted the member forces from truss analysis, Fassume , to original design member forces,

F1E.0, . Second, we increased the original design forces by 100% to F2E.0, . Then, for each

panel of the vertical leg, we compared its F2E.0, value with its code design strength, N c .

The comparison indicated that every model in this study (Cross S80, Cross S100, Cross
120, Warren S80) needed reinforcement for at least six panels.

148
Therefore, every model started with the case that had the existing leg reinforced for six
panels, with two connections per panel (2C 6/9), which required 12 connections. From the
16 pairs of cores and connectors available in total and already put in place, the interaction
definitions of the lowest 12 connections were activated for this 2C 6/9 case. This case had
a few runs as necessary with longer loading time periods to check for convergence. At
convergence, if the target capacity increase was not achieved, then another case with one
more panel being reinforced (2C 7/9) would be created and analysed for a few runs. Then
the interaction definitions of two more connections were then activated. If the case having
eight panel reinforced (2C 8/9) was attempted, all interaction definitions were activated.

To indicate the convergence, the axial force (SF1 in ABAQUS) on any panel or member of
the model, including the bracing members, could be used. An easy choice was to use force
on a panel of an inclined leg (for example, Panel 10) or force on an unreinforced panel of
the compression leg. But if we were to use the force of a reinforced panel, we needed to
use the sum of forces from the existing leg and reinforcing member of that panel instead.

It was important to mention here that in indicating a potential capacity (or strength)
increase, the increase must be based on the new potential design capacity after reinforcing
as referred to the initial design capacity. The new potential design capacity was either
adequate or deficit. The term adequate was used when the reduced ultimate/critical
capacity (0.9 x ultimate capacity) of the model was larger than the required design
capacity. And the term deficit was the opposite.

Here, four nomenclatures were used, listed by the order of their use:

Fassume = forces from truss analysis, based on the assumed two 20 kN forces

F1E.0, = forces at the original design stage ( Fassume multiplied by a specific factor
equal to N c / FAssume of the lowest panel of the vertical leg)

F2E.0, = expected forces at 100% capacity increase stage, equal to 2 F1E.0,

F2A.0 = force results from FE analysis at ultimate or critical stage, based on the
design intending to achieve 100% capacity increase

F2A.0, = 0.9 F2A.0 , reduced analysis force result, to be compared with F2E.0,

149
6.7.3 Change in support conditions
Previously in the pilot models, in order to have the same restraint condition as that used in
the truss analysis section, and to develop the modelling techniques whose results could be
checked, the vertical leg was pin-supported at the base. In this section, the vertical leg and
its reinforcing member were fully fixed (U1 = U2 = U3 = UR1 = UR2 = UR3 = 0) to
simulate the actual condition in practice. It was considered that the use of connectors
would provide sufficient self-restraint against free turning-around of the vertical existing
leg and the reinforcing member, therefore there was no need to add additional restraints
UR2 = 0 to the vertical existing leg at its panel points as was done for the unreinforced
models.

For the inclined legs, they were pin-supported as before. By not having moment restraints
at the base of the inclined legs, it was hoped that this condition would cause the reinforcing
member and the connectors to perform their duty a little more than if the moment restraints
had been provided. In other words, strength increase would rely on the connectors and
reinforcing member a little more. As a result, the results obtained (strength increase
potential) should be a little more conservative.

6.7.4 Results and discussions


6.7.4.1 Cross S80

A. Force-time curves

Five force-time curves of Panel 8 (the eighth panel from the bottom) of five runs of Cross
S80 model are shown in Figure 6.11. For the first three runs, the vertical leg was reinforced
for six panels out of nine with two connections per panel (2C 6/9 case); and for the other
two runs, seven panels out of nine (2C 7/9 case). For each case, the target displacement
was adjusted and the time period increased until the convergence was obvious, as can be
seen in the figure.

150
150
Panel 8, Cross S80 FE
2C 6/9 0.072m 1.0s
2C 6/9 0.067m 1.4s
125 2C 6/9 0.065m 1.8s
Axial Force, FA2.0, on Panel 8 (kN) 2C 7/9 0.075m 1.4s
2C 7/9 0.085m 1.8s

100

75

50

25

0
0 0.5 1 1.5 2
Time (s)

Figure 6.11 Force-time curves of Panel 8 from the five runs of Cross S80 model

While the expected design force F2E.0, of Panel 8 was 119.4 kN (see Table 6.4), the

reduced analysis force result F2A.0, of the first run (6/9, 0.072m, 1.0s) was only 104.0 kN.
This inadequacy, although small, suggested that if we still wanted to pursue the expected
force, some additional measures must be done to overcome shear lag, and adding one or
more additional panels of reinforcement was worth trying.

Before going further, the following description is useful for better understanding of the
later work and discussions. After having finished the first run, the second run (6/9, 0.067m,
1.4s), and the third run (6/9, 0.065m, 1.8s), a point came to mind which was that the
second run and the third run were not really necessary. What suggested this was the
relation between the loading speed (the target displacement over the loading time period)
and its capacity product of the structure. That is, assuming everything else is the same,
using a faster loading speed will produce a higher capacity than using a slower loading
speed. Since the result of the first run of 6/9 case showed inadequacy (104.0 kN < 119.4

151
kN), hence, there was no need to have the second run and the third run, unless we wanted
to find out the percent of inadequacy.

The third runs curve shows no drop of force, which means that the target displacement
was set a bit too low. (It was set too low that Panel 7 did not reach the buckling point.
Panel 7 was the critical panel of the 6/9 case, which will be talked about in the next
paragraph.) Therefore, among the three runs, the second run was the most representative in
terms of presenting the capacity of the 6/9 case. The second run had the reduced analysis
force F2A.0, of Panel 8 equal to 103.3 kN, almost the same as the 104.0 kN of the first run.

When compared with the original design force F1E.0, = 59.7 kN, the obtained capacity
increase was 73.0%. Figure 6.12 shows the buckled shape of Cross S80 2C 6/9 model,
second run, with the obvious buckling at Panel 7.

Figure 6.12 Cross S80 model, Case 2C 6/9, showing failure by buckling of Panel 7

In the above discussions, the F2E.0, and F2A.0, values of Panel 8 were referred to
exclusively. But in fact, the force of any unreinforced member can be referred to for
discussion purpose because the capacity of the whole structure was controlled by only one
152
critical member. Figure 6.12 shows an obvious buckling at Panel 7. This means that Panel
7 was the critical member of this Cross S80 2C 6/9 case. While Panel 7 had the code
design strength N c equal to 199.0 kN, it buckled at F2A.0 = 169.0 kN, or F2A.0, = 152.1 kN
only ( 23.6% lower). This lower strength performance was due to the effect of combined
use between beam members and the core parts, which agreed well with the same
observation on the unreinforced models with core parts in Section 6.6.5. Again, this lower
strength helped guarantee that if the models here could be reinforced to achieve the target
capacity increase, when the recommendations obtained from this study are practically
implemented, the desired capacity will be achieved too.

Having wondered whether it would be possible to obtain more capacity by only extending
the reinforcement for one or two more panels while everything else remained the same, the
fourth run (7/9, 0.075m, 1.4s) and the fifth run (7/9, 0.085m, 1.8s) were created and
analysed (see Figure 6.13). Similar to the third run, the fourth run was set with a little low
target displacement. Thus, there was no sign of failure such as 1) force dropping due to
buckling of any panel of the vertical leg, or 2) force going horizontally due to yielding of
any panel of the inclined (tension) legs. (Figure 6.11) Nevertheless, it shows that the force
could increase higher than that of the 6/9 case already.

The fifth run was set with enough target displacement to fail the model. The failure
occurred by yielding of the inclined legs at the lowest panel, or Panel 10. Because it failed
by yielding, its failure shape cannot be seen obviously (see Figure 6.13). Since only one
run was conducted for the 7/9 case, there was only one result and it did not allow us to
check for convergence. However, from experience the researcher had observed that in the
case of low slenderness ratio (S80) like this, having another run with a longer time period
would only produce the same results, especially when the time period used before was
already long. Therefore, it was assumed that the fifth run alone should be able to present
the true potential of Cross S80 2C 7/9 case. Table 6.8 compares the reduced analysis forces
of Panel 8 and Panel 10 of the fifth run (7/9, 0.085m, 1.8s) with their expected forces and
their predetermined strengths.

153
Figure 6.13 Cross S80 2C 7/9 model

Table 6.8 Analysis of forces of Panels 8 and 10 from the fifth run (0.085m, 1.8s) of Cross
S80 2C 7/9 case, in comparison with their expected forces and predetermined strengths

Approx. Predetermined Buckling or Yielding


Panel No. F2.0E, F2.0A,
Strength from FEM with Cores (kN)
8 119.4 123.7 152.1 (Buckling)
10 184.7 195.8 195.9 (Yielding)

Note: 1) Since Panel 8 did not buckle, its actual buckling strength was not known. However,
because it had the same length as Panel 7, Panel 8 was assumed to have the same
buckling strength, which was 152.1 kN.
2) For Panel 10, being a tension member, its yield strength was calculated from its cross-
sectional area and yield strength.

The comparison shows that force F2A.0, on Panel 10 (195.8 kN) reached the panels yield
strength (195.9 kN) before force on Panel 8 (123.7 kN) could reach the panels buckling
strength (152.1 kN). It can be said that now the designed reinforced model was controlled
by the yield strength of Panel 10. However, at this stage as Panel 10 started to yield, the

154
forces on both panels as well as the forces on other members of the whole structure had
increased higher than the expected forces already, although only slightly. This means that,
for this specific Cross S80 2C model, adding one more panel of reinforcement (from six
panels to seven panels) helped achieve the target 100% capacity increase (from +73.0% to
+100%).

Seeing the result of the 7/9 case where capacity of the structure was governed by Panel 10
which experienced failure by yielding first, it was concluded that there was no need to
conduct further study using eight panels of reinforcement because the results would be the
same.

B. Force distribution regarding member slips

In addition to the overall view of capacity increase potential, it is also interesting to know
the behaviour of force distribution between the vertical existing leg and its reinforcement,
sometimes called the reinforcing member.

First before plotting graphs, it was necessary to mention about the time step and output
intervals. In the whole study, each analysis of every run, case, and model type used only
one step which was loading the top right-angle corner node to a target displacement
amount. Analysis then proceeded automatically and the output was saved at 40 equally
spaced intervals within the time period. These 40 intervals had 41 time frames, from frame
0 to frame 40.

Since to reinforce the model to achieve the target or expected design capacity was our
main goal, it was rational to plot the force distribution at this capacity level.

In the successful case of Cross S80 2C 7/9, Run 5, Panel 10 had its F2A.0, = 195.8 kN,

which was larger than its F2E.0, = 184.7 kN. Therefore, we looked for the time frame when
the force on Panel 10 was closest to 184.7 kN. That frame was Frame 27 which had the
force equal to 182.7 kN. Then we used all the panel and sub-panel forces of the existing
leg and reinforcing member at Frame 27 to plot the force distribution lines of this case.
They are the solid lines in Figure 6.14. To make it clear in the presentation how forces
were distributed along the panels, they were plotted against the elevations. The amount of
force transferred to the reinforcing member at each connection, or shear force, was
calculated for this case and they are all presented in Table 6.9. Examining Figure 6.14, the

155
capacity improvement of the sum member (reinforcing + existing) compared with the
existing member alone is obvious.

Figure 6.14 Force distribution on the compression leg of Cross S80 2C FE model, Cases
6/9 and 7/9

However, for the case of Cross S80 2C 6/9, Run 2, that failed by buckling of Panel 7 at
F2A.0 = 169.0 kN, or F2A.0, = 152.1 kN, which was lower than its F2E.0, = 178.5 kN, choosing
the time frame that had the force on Panel 7 closest to 169.0 kN would be on the extreme
limit. Therefore, it was decided that the rational frame should be the one whose force was
closest to 152.1 kN. That frame was Frame 30 when the force on Panel 7 = 151.9 kN.
Force distribution lines of this case at this time frame are the dotted lines in Figure 6.14.

The two total force lines of both cases apparently tell that, for this particular model and its
member selection, extending the reinforcement for one more panel (to Panel 7) than had
been identified earlier could increase the models capacity. Although the increased amount
may look moderate, it came from just one additional panel, and most importantly it helped
lift the capacity to meet the target design capacity.

156
Table 6.9 Sub-panel forces and connection shears from FE analysis of Cross S80 2C 7/9
FE at the target capacity stage

Panel Sub-panel Force (kN)


No. No. Main Leg Connection Reinforcing Sum
Shear Member
9 45.3 45.3
8 115.8 115.8
7-1 175.7 175.7
7 7-2 139.3 36.4 36.4 175.7
7-3 110.6 28.7 65.1 175.7
6-1 157.2 65.1 222.3
6 6-2 134.3 23.0 88.1 222.3
6-3 116.2 18.1 106.1 222.3
5-1 153.9 106.1 260.1
5 5-2 140.6 13.3 119.5 260.1
5-3 131.4 9.2 128.7 260.1
4-1 164.2 128.7 293.0
4 4-2 157.3 7.0 135.7 293.0
4-3 153.5 3.8 139.5 293.0
3-1 179.3 139.5 318.8
3 3-2 177.8 1.5 141.0 318.8
3-3 177.7 0.1 141.1 318.8
2-1 199.8 141.1 340.9
2 2-2 199.7 0.2 141.2 340.9
2-3 199.5 0.2 141.5 340.9
1-1 218.7 141.5 360.2
1 1-2 218.6 0.1 141.6 360.2
1-3 217.4 1.2 142.7 360.2

The total force lines show that, in general, the force increment was large on the upper
panels and became smaller and smaller on the lower panels. How large or how small
depends on the arrangement of the bracing members. This force increment pattern was in
the same direction as had been predicted by truss analysis.

A surprising result was the force pattern of the reinforcing member. The force kept
increasing as it went down for a number of panels, but with lower increments, and after
that remained nearly constant down to the base. For both 6/9 and 7/9 cases, the force
stopped increasing after it passed Panel 4 and stayed constant over Panels 3, 2, and 1.
Since no more force went to (or could be transferred to) the reinforcing member over these
three panels, it all had to go to (or be carried by) the existing leg only, as can be observed
in Figure 6.14.

157
The force distribution behaviour above was attributed to the slip connections used, which
can be explained in general terms as follows. In whichever panels and sub-panels that
show no force increment on the reinforcing member, the connectors over those areas do
not function. This is because such areas of the existing leg do not displace enough to form
a complete bearing condition between the bolts and bolt holes on the related members (the
existing leg, the reinforcing member, and the connectors).

Another interesting point was the strength increase potential of the panels that were
reinforced. In this place, the maximum potential occurred at Panel 1 of the existing leg.
From Section 6.6.5, pilot model with cores without reinforcement, the model failed by
buckling of Panel 1 at a force of 142.7 kN, or a reduced force of 128.4 kN (Table 6.6).
Compare this force with the 218.6 kN on the existing leg of the 7/9 case in Table 6.9, the
amount of force it carried could increase up to 70.2% which was considerable. Although
over this panel the reinforcing member did not help carry force at all (zero force amount
was transferred from the existing leg to the reinforcing member), it was understood that its
existence plus the (inter)connectors used helped brace the existing leg, which brought
about such significant strength increase. Panel 1 had sub-panels with approximate local
slenderness ratio a / r of 60.

C. Critical connection shear and prediction

In the above subsection, it was observed that one result of using bearing-typed connections
was that at some lower panels there may be no force transfer happening at all at the
connections. In practice it is unlikely that engineers will both have time and be willing to
use a FE software to create a slip model for strength reinforcing projects. Therefore, this
study also attempted to use a normal structural elastic analysis software to create the same
set of reinforced 2C 7/9 model and examine its compatibility regarding the results.

Almost everything of the elastic tower models (the geometry, bracing arrangements,
member types, shapes, orientations, support conditions, and load directions) were the same
as that of the FE models, except the connection system and loads or loading used. Loading
was defined using two 42.3 kN loads in the perpendicular directions, that aimed to create
the 184.7 kN target design force on Panel 10. For connections, core parts were not used
and the slip connectors were replaced by ordinary beam-typed members (Figure 6.15)
connected directly to the leg members. Having moment restraints in all directions, beam-
typed members were considered as the fastest and most appropriate type to use as the

158
connectors. A problem arose next as to what section and size should be used for the
connectors. The easiest pick was to use the same one as was used for the existing leg and
reinforcing member, which was 75 75 6 EA in this place, oriented face down (or up)
with the sections longitudinal axis lying horizontally. For this selection, the moment of
inertia of the connectors and the existing leg were the same, or I y ,con = I y ,existing . This model

set was named Cross S80 2C 7/9 ConI=1.0. The term ConI=1.0 indicates that the ratio of
I y ,con / I y ,existing was equal to 1.0.

Figure 6.15 (a) A portion of the reinforced model created for simple elastic analysis; (b) A
portion of the reinforced leg in 3D showing the member orientations

Force lines of Cross S80 2C 7/9 ConI=1.0 are plotted in Figure 6.16. Because the model
did not allow slip, the lines show force transfer taking place at every connection, although
in a small amount at the lower panels. As a result, forces on both members become pretty
close. This force distribution pattern from using this simple elastic model proved that it
cannot replace the FE models with slip connections (compare Figures 6.16 and 6.14).
Adopting this simple method in the hope of getting the correct member forces and using
the results for design purpose would be misleading.

159
Figure 6.16 Force distribution on the compression leg of Cross S80 2C 7/9 Elastic
ConI=1.0

In designing the connections, the amount of shear force they are to carry is the primary
issue. Engineers who have been doing the design in practice have wondered whether their
design of the connections is stronger or weaker than it really needs to be. Despite the
conclusion made in the previous paragraph, the merit of this simple model may still exist.
Observing shear values of Cross S80 2C FE 7/9 in Table 6.9, it was noticed that the largest
shear occurred at the top connection at 36.4 kN. For the simple Cross S80 2C 7/9 ConI=1.0
model, the largest shear also occurred at the top connection, but at 53.1 kN. This
observation indicated that although using the simple model did not give the correct
maximum shear value, it still gave a conservative shear value that could be used to design
the connections safely. It should be kept in mind that such conservative value came from
using I y ,con = I y ,existing . If a smaller section was used, the shear values would be less.

To speculate what section size should be used for the connectors in order to produce the
shear force at the top connection close to the 36.4 kN of the 7/9 FE model, seven sections
were tried, based on the I y ,con / I y ,existing ratio, from 1.0 down to 0.02. The top connections

160
shear forces of the seven trials are shown in Table 6.10. The results were quite surprising
because the section that provided the shear value nearest to the 36.4 kN was very small.
The section was 25 25 5 EA which made I y ,con / I y ,existing = 0.025 only, and the

produced shear was 38.3 KN. Assuming that this quite matched result was not a
coincidence, it can become a design recommendation that if engineers want to get a non-
conservative shear value by using a simple elastic model like what we created here, a
section that makes I y ,con / I y ,existing ratio about 0.025 should be defined for the connectors;

while defining the same section as of the existing leg, making I y ,con / I y ,existing = 1.0, will

give a larger shear value. The author suggests using the ratio of 0.1, which in this study
gave the shear value of 46.3 kN that was about the middle between the two extremes.
However, this recommendation was obtained from the speculation of one case only. The
same speculation was also examined for other cases which will be discussed in Section
6.7.4.2.C.

Table 6.10 Variation of the connectors sizes and shear force results at the top connection
from simple elastic reinforce model simulating Model Cross S80 2C 7/9

Model Iy,con/Iy,existing Iy,con Section Iy,con Iy,con/Iy,existing Shear at Top


Cross S80 2C 7/9 Sampled Wanted Selected for Actual Actual Connection
6 4 6 4
10 mm Connectors 10 mm (kN)
ConI=1.0 1.0 0.187 75 x 75 x 6 EA 0.187 1.0 53.1
ConI=0.4 0.4 0.0748 55 x 55 x 6 EA 0.0723 0.387 51.1
ConI=0.2 0.2 0.0374 45 x 45 x 6 EA 0.0383 0.205 49.3
ConI=0.1 0.1 0.0187 40 x 40 x 5 EA 0.0209 0.112 46.3
ConI=0.05 0.05 0.0094 30 x 30 x 6 EA 0.0107 0.0572 43.8
ConI=0.025 0.025 0.00468 25 x 25 x 5 EA 0.00469 0.02508 38.3
ConI=0.02 0.02 0.00374 25 x 25 x 3 EA 0.00319 0.01706 34.9

D. Limitation of using one connection per panel

It was presented in Section 6.7.4.1.A that using two connections per panel was enough for
the reinforced legs with L / r 80 to achieve the 100% capacity increase target. Hence,
using three connections would produce the same or slightly better result, and therefore it
was unnecessary to be studied. The more interesting case is when only one connection is
used per panel, how good the performance would be, compared with when using two
connections per panel. Although this case was anticipated to be less effective, knowing its
actual potential and the cause of failure was desirable. However, if it turned out to be

161
equally effective, it would be suggested that using one connection per panel can give a
great engineering value in terms of the lower cost benefit.

For this new study, the model used was identical to that of Cross S80 C2 7/9 FE case, but
for each panel the interaction definitions of the upper connection were activated only,
while the interaction definitions of the lower connection were suppressed. This new case
was named Cross S80 C1 7/9 FE. The target displacement and time period were 85 mm
within 1.8 sec, the same as that of Cross S80 C2 7/9 FE case. The analysis force F2A.0 in
Panel 10 reached the highest value at 156.0 kN (Figure 6.17), or the reduced analysis force
F2A.0, = 140.4 kN. When compared with the initial design force ( F1E.0, = 92.4 kN), it gave a
capacity increase of +51.9% ( +50%). Since this amount of increase was about half of
the target amount only, it can be concluded that using one connection per panel is
ineffective and cannot substitute for using two connections per panel.

225
Panel 10, Cross S80 FE
200
2C 6/9
on Panel 10 (kN)

2C 7/9
175
1C 7/9

150

125
2.0,

100
A
Axial Force, F

75

50

25

0
0 0.5 1 1.5 2
Time (s)

Figure 6.17 Force-time curves of Panel 10 from three cases of Cross S80 FE model

The cause of such low potential was observed. Figure 6.18 shows that the existing leg
buckled at either Panel 1 or 2 over the length measured from each panels lower panel
162
point to the panels upper core, which was 1.0 m, just 0.2 m less than the original
unsupported length. (To differentiate which panel of the existing leg between Panels 1 and
2 did buckle first was not easy even though the panel force values were available in
number. This was because while Panel 1 carried larger force, it had shorter effective
unsupported length because of the fix-pin end condition it had; or while Panel 2 carried
lower force, it had longer effective unsupported length due to the pin-pin end condition it
had. However, it was not necessary to know which panel actually buckled.)

Figure 6.18 Detail of buckling of Cross S80 FE model, C1 7/9 case

But if two connections were used per panel, as had been proven in the case of Cross S80
2C 7/9 FE, the unsupported length was reduced by half. It was short enough to prevent the
reinforced parts or panels from experiencing local buckling before the target capacity was
achieved. In this case, the vulnerable part moved or migrated from Panel 1 or 2 either
upward to the first unreinforced panel (Panel 8) or to the lowest panel of the inclined
tension legs (Panel 10).

163
6.7.4.2 Warren S80

Warren S80 was another important case that was aimed to investigate whether the method
to indicate the minimum requirement of reinforcement (where to terminate the
reinforcement) using the results from elastic truss analysis worked well with the towers
with Warren bracings. If yes, it should mean that the method can be applied to towers of
any bracing pattern reliably. Also it was aimed to examine, with Warren-bracing pattern,
whether the model needed more or less extra panels of reinforcement than that needed by
Cross model to achieve the target capacity, would the forces distributed to the reinforcing
member be different, and what controls the transfer. The discussions will go through the
same kinds of analyses and observations as done with Cross S80 model.

Four cases in a total of five runs were conducted, all aimed for a 100% capacity increase.
The first and second cases had reinforcement for six and seven panels with two
connections each panel as before. They were named Warren S80 2C 6/9 FE and Warren
S80 2C 7/9 FE, respectively. The third case had reinforcement for eight panels, namely
Warren S80 2C 8/9 FE, which was intended to check the assumption given at the end of
Section 6.7.4.1.A that using reinforcement for seven or eight panels would give the same
result because the models performance would then be controlled by other members else,
such as Panel 10 yielding in tension. For the fourth case, seven panels of reinforcement
were applied with one connection per panel, namely Warren S80 1C 7/9 FE.

A. Force-time curves

Five force-time curves of Panel 10 of five runs of the four cases are shown in Figure 6.19,
and their important force values are summarized in Table 6.11. The result of the Case 1,
Warren S80 2C 6/9 FE, showed that by using six panels of reinforcement (the minimum
amount of reinforcement as predetermined in Section 6.5, Table 6.5), the capacity
increased by 80.1% only, which was still inadequate. The model failed by buckling of
Panel 7, the first unreinforced panel of the compression leg (Figure 6.20(a)), at the F2A.0 =

157.3 kN, or the F2A.0, = 141.6 kN. Similar to the case of Cross S80 2C 6/9 FE, this F2A.0,

was lower than the N c = 199.0 kN quite significantly ( 28.8% lower). This could be
reasoned as being the result of the combined usage between beam elements and the core
parts. Only one run of this case was conducted based on the judgement that the given 1.0
second time period should be long enough and that using any longer period would still give
the same result.
164
225 Panel 10, Warren S80 FE

200 2C 6/9 0.074m 1.0s


2C 7/9 0.084m 1.0s

on Panel 10 (kN)
2C 7/9 0.082m 1.5s
175
2C 8/9 0.082m 1.5s
1C 7/9 0.082m 1.5s
150

125
2.0,

100
A
Axial Force, F

75

50

25

0
0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6
Time (s)

Figure 6.19 Force-time curves of Panel 10 from five runs of the four cases of Warren S80
FE model

Table 6.11 The expected design forces and analysis forces of Panel 10 from five runs of
the four cases of Warren S80 FE model

Panel 10
Warren S80 FE Run No.Target Disp. &
% Capacity
Case No. Loading Time (m, s) F1.0E, F2.0E, F2.0A F2.0A,
Increase
1. 2C 6/9 1. (0.074 m, 1.0 s) 96.8 193.5 193.7 174.3 80.1
2. (0.084 m, 1.0 s) 217.6 195.8 102.3
2. 2C 7/9
3. (0.082 m, 1.5 s) 217.3 195.6 102.1
3. 2C 8/9 4. (0.082 m, 1.5 s) 217.5 195.8 102.3
4. 1C 7/9 5. (0.082 m, 1.5 s) 159.6 143.6 48.3

165
Figure 6.20 Failure modes of Warren S80 FE model: (a) Case 1, 2C 6/9, buckling at
Panel 7; (b) Case 2, 2C 7/9, yielding at Panel 10; and (c) Case 4, 1C 7/9, buckling of the
existing leg at Panel 2

For Case 2 (Warren S80 2C 7/9 FE), its first run (Run 2) turned out to have the capacity
above the target capacity (193.5 > 195.8 kN). Hence, the second run (Run 3) using a longer
time period was conducted for convergence check. It also showed the same result (see
Table 6.11, Case 2, Runs 2 and 3). The failure mode was yielding of the tension legs at
Panel 10. (Figure 6.20(b)) This could be reasoned that, as Panel 7 was reinforced, it was
not the weakest member of the model anymore, but Panel 10 became the weakest member
instead.

For Case 3 (Warren S80 2C 8/9 FE) having eight panels reinforced, while force
distribution between the existing leg and the reinforcing member of the compression leg
could be different from that of Case 2, which will be discussed later, the force-time curve
of Panel 10 showed no difference (see Figure 6.19). It reached the same yield strength
which was the members cross-sectional yield strength. This result confirmed the

166
assumption that, after the reinforcing member to a certain point, reinforcing the leg further
has no strength increase benefit.

For Case 4 (Warren S80 1C 7/9 FE) where seven panels were reinforced with one
connection per panel, the force on Panel 10 indicated that the capacity of the model could
increase by 48.3%, which was close to the +51.9% of Cross S80 1C 7/9 case. Due to
inadequate use of the connections, the sub-panel length was only a little less than the panel
length, and hence, the existing leg and the reinforcing member behaved relatively
individually from each other over a panel. This resulted in low capacity increase. The
model failed when force on the existing leg of the critical panel reached the buckling
strength of its sub-panel. In this case it was obvious that the existing leg buckled at Panel 2
(Figure 6.20(c)) because while Panel 1 and Panel 2 carried the same amount of force
(Panels 1 and 2 were separated by a secondary bracing member), Panel 2 had longer
effective panel length due to the pin-pin end condition it had, while Panel 1 had shorter
effective panel length due the fix-pin end condition it had.

B. Force distribution regarding member slips

As before, for the case that had the F2A.0, > F2E.0, , the time frame that had its corresponding

capacity closest to F2E.0, was identified and the panel or sub-panel force values at this time

were taken for plotting. On the other hand, if the F2A.0, < F2E.0, , the time frame that had its

corresponding capacity closest to F2A.0, was identified.

Force distributions on the compression leg of the model Case 1 (2C 6/9), Case 2 (2C 7/9),
and Case 3 (2C 8/9) are presented in Figure 6.21. The force lines of the reinforcing
member of these three cases showed the same characteristic, which was that the force
increased more and more from the upper to the lower connections, but with smaller and
smaller increments. At one point, the increment became zero and the force line dropped
vertically. If stated in terms of the force transfer at the connections, it can be said that the
transfer amount was largest at the top connection, and became smaller and smaller in the
lower connections until at one connection there was no transfer. The range of no transfer
could be from a few panels to several panels. For these three cases, it happened to the
lowest three panels. This kind of outcome has already been reasoned in the section of
Cross S80 FE that it was the result of slip connection usageto cause force transfer to
happen at a connection, there must be enough displacement between the two members in

167
the zone of that connection to complete all the contacts required for a slip shear-bearing
connection.

Figure 6.21 Force distribution on the compression leg of three cases (2C 6/9, 2C 7/9, and
2C 8/9) of Warren S80 FE model

Notice that while the force lines of the existing leg and the reinforcing member of Case 2
(2C 7/9) and Case 3 (2C 8/9) are different, their total-force lines are identical. For these
two cases, despite such different force distribution patterns between the two members of
the compression leg, it was the cross-sectional yield strength of Panel 10 that dictated the
capacity of the model after it was reinforced. This expression proves the previously made
assumption that providing more panels of reinforcement than really needed has no capacity
increase benefit for the tower. Nevertheless, it helps guarantee that the reinforced tower
will achieve the expected capacity.

C. Critical connection shear and prediction

From Figure 6.21, it can be observed that for every case the largest amount of force
transfer (shear) occurred at the top connection. In this section, simple elastic reinforced
models were created and used to predict the force transfer amount at the top connection
168
again. For Warren S80 model, the case considered was 2C 7/9. Two other cases of the
other two models (Cross S100 and Cross S120) were also investigated here. They were
Cross S100 2C 6/9 and Cross S120 2C 6/9. These cases were chosen because they were the
first case of each model that was able to achieve the target 100% capacity increase. The
section of the beam-typed connectors was varied by the same I y ,con / I y ,existing ratios from

1.0 to 0.02 as before. The two perpendicular applied forces were intended to create a 100%
capacity increase. The shear values of the top connection are summarized in Table 6.12.

Table 6.12 Comparison of shear forces at the top connection from elastic models with
varying connection sizes and from FE models

Cross 2C Warren 2C
S80 7/9 S100 6/9 S120 6/9 S80 7/9
Two Perpendicular Applied Forces (kN)
Iy,con/Iy,existing Selected Section Iy,con/Iy,existing 42.3 33.5 23.8 43.4 Analysis
Sampled for Connectors Actual Shear at Top Connection (kN) Model
1.0 75 x 75 x 6 EA 1.0 53.1 50.8 36.2 53.3 Elastic
0.4 55 x 55 x 6 EA 0.387 51.1 49.2 35.3 51.3 "
0.2 45 x 45 x 6 EA 0.205 49.3 47.8 34.4 49.5 "
0.1 40 x 40 x 5 EA 0.112 46.3 45.3 32.8 46.5 "
0.05 30 x 30 x 6 EA 0.0572 43.8 43.1 31.4 44.0 "
0.025 25 x 25 x 5 EA 0.02508 38.3 38.0 28.0 38.5 "
0.02 25 x 25 x 3 EA 0.01706 34.9 34.6 25.7 35.1 "
36.4 35.3 25.6 37.3 FEM

It can be seen that out of four cases here, the three cases that used I y ,con / I y ,existing about

0.025 obtained the top connection shear values closest to, but larger than, the same values
of their FE counterparts. They were Cross S80 2C 7/9, Cross S100 2C 6/9, and Warren S80
2C 7/9. And for the other case, Cross S120 2C 6/9, the ratio was 0.02. These similar
findings indicate that it is possible use elastic models to predict the shear force at the top
connection relatively precisely if the size of beam-element members used as the connectors
has the moment of inertia approximately equal to 0.025 times that of the existing leg
( I y ,con / I y ,existing = 0.025). However, if one wants to get a larger value for conservative

design, using a section that gives I y ,con / I y ,existing = 1.0 will do the job well.

6.7.4.3 Cross S80 2C 7/9 FE vs. Warren S80 2C 7/9 FE

This section considered whether different arrangements of bracing members created


different force distribution patterns on the reinforcing member. The issue was examined by
169
comparing the force distribution lines of Cross S80 and Warren S80 models. Both models
achieved their target capacities when the reinforcement was applied for seven panels, or
the case of 2C 7/9. The force distribution lines of Cross S80 2C 7/9 FE and Warren S80
7/9 2C FE are compared in Figure 6.22.

Figure 6.22 Force distribution on the compression leg of Cross S80 2C 7/9 FE and
Warren S80 2C 7/9 FE

It was expected that the total force lines of both models would be somewhat different due
to the difference in bracing arrangements. And as the bracing members were connected to
the existing leg only, the panel forces of the existing leg of both models should also be
different. These two points are well illustrated in the figure. The forces on the reinforcing
member were not expected to be the same due to the influence from different bracing
types, but from the FE results, they are almost identical. This similarity indicates that, if
the connections are slip-bearing type, the bracing pattern does not have significant
influence on the amount of force distributed to the reinforcing member at each connection.
The slip will take place; and based on the obtained data, almost all force transfer will take
place at the connections over about the upper half of the reinforced length.

170
However, the above conclusion was drawn from Cross S80 and Warren S80 models which
had the same geometry, the same panel length for every panel of the legs, and the same
positions of the connections. If one or more of these parameters are different, it is possible
that force distributed to the reinforcing member of both models will be different.

6.7.4.4 Cross S100 and Cross S120 vs. Cross S80

Cross S100 and Cross S120 models were created and analysed in the same way as before.
Both models started with the 2C 6/9 case. So the two cases were Cross S100 2C 6/9 FE
and Cross S120 2C 6/9 FE, respectively. Each case used two runs to check for
convergence. Since they were able to achieve a 100% capacity increase as desired (see
Figure 6.23 and Table 6.13), and since we had got the understanding of what would happen
if the reinforcing member was used for more panels or if only one connection was used per
panel, no other cases were attempted.

180 Panel 10,


Cross S100 and S120
160 2C 6/9 FE
Axial Force, FA2.0, on Panel 10 (kN)

S100 0.09m 1.0s


140 S100 0.09m 1.5s
S120 0.14m 0.9s
120 S120 0.1m 1.3s

100

80

60

40

20

0
0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6
Time (s)

Figure 6.23 Force-time curves of Panel 10 showing convergence from two runs of Cross
S100 2C 6/9 FE and two runs of Cross S120 2C 6/9 FE

171
Table 6.13 The expected design forces and analysis forces of Panel 10 from two runs of
Cross S100 2C 6/9 FE and two runs of Cross S120 2C 6/9 FE

Panel 10
Run No.Target Disp. &
Cross FE % Capacity
Loading Time (m, s) F1.0E, F2.0E, F2.0A F2.0A,
Increase
1. (0.09 m, 1.0 s) 73.1 146.2 172.3 155.1 112.2
S100 2C 6/9
2. (0.09 m, 1.5 s) 171 153.9 110.5
1. (0.14 m, 0.9 s) 52 104.1 126.1 113.5 118.3
S120 2C 6/9
2. (0.10 m, 1.3 s) 125.8 113.2 117.7

However, the two cases/models experienced different failure modes. Cross S100 2C 6/9 FE
failed by buckling of the compression leg at Panel 7, while Cross S120 2C 6/9 FE failed by
yielding of the tension legs at Panel 10. However, since they were able to achieve the
target capacity increase, it was not considered necessary to be worried about why they had
different failure modes. In fact, it depends on which member or section will reach its
maximum strength (buckling or yielding) first. But if the buckling failure mode is
considered undesirable, it is suggested that the reinforcement should be extended by at
least one panel more than the minimum quantity that the elastic truss analysis indicates.

Force distributions between the existing leg and the reinforcing member of the two cases
are shown in Figures 6.24 and 6.25. It can be noticed that force patterns on the reinforcing
member of both cases are the same as that of Cross S80 and Warren S80, except that the
magnitudes are smaller. This similarity indicates that even in the case where the panel
length is long, the slip will still take place. To consider the shear lag effect in relation to the
slenderness of the common panel length, the number of the connections that carried force
(shear not equal to zero) will be compared with the number of all connections used. For
Cross S80 2C 7/9 FE and Warren S80 2C 7/9 FE, that number is 9 out of 14, or a ratio of
0.64. For Cross S100 2C 6/9 FE, that number is 7 out of 12, or 0.58. And for Cross S120
2C 6/9 FE, the number is 6 out of 12, or 0.5. This comparison indicates that the shorter the
panel length is, the more connections are required to transfer the force, or the more shear
lag effect the reinforcing has to deal with. This does not mean that those connectors over
the lower half of the reinforced length that do not carry any shear force or carry very low
shear force do not have any function. They serve function by stitching the reinforcing
member to the existing leg, which results in increased strength of the leg.

172
Figure 6.24 Force distribution on the compression leg of Cross S100 FE model, case 2C
6/9

173
Figure 6.25 Force distribution on the compression leg of Cross S120 FE model, case 2C
6/9

6.8 Capacity increase potential based on Browns method


The reinforced models were developed step by step, from initially the unreinforced models
with geometrical imperfection introduced and the models were verified by code strengths.
However, when it came to the final design of the reinforced models, the geometrical
imperfection was not (or could not be) introduced. Therefore, it is possible that the actual
capacity increase potential of the leg reinforcing method is not as high as 100%. To know
this for sure, testing on reduced-scale or real-scale multi-panel towers is essential.

6.8.1 Combined code method and Browns method


For the moment, the author would like suggest the method of Brown (1988) in calculating
the capacity of a reinforced leg. The scenario that Brown used in developing his method
was that the core column is under a load (or the existing leg in this thesis) when the
reinforcement is added to it. As the applied load keeps increasing, to one point the core
column will reach it buckling capacity first. After the core column has failed, the stiffness
of the reinforcement will provide additional capacity (post-buckling capacity) to the
reinforced column. This scenario is quite similar to the scenario of reinforced tower legs,
therefore the method has been selected here. In calculating the effective slenderness ratio
of compound leg members for the buckling mode that involves relative deformations, the
modification equation given in AS 3995 (1994) will be adopted due to its easiness.

The explanation of the work of Brown is given in Chapter 2. The equations for calculating
the critical compression capacities of the core column, the reinforcement, and the
reinforced column can be those given in AS 3995 (1994) or in other design codes.
However, to be consistent of the subject of tower member strengths, the equations of AS
3995 (1994) are suggested. The application of the method is demonstrated by examples
below.

6.8.2 Examples

The examples will use 75 75 6 EA, f y = 320 MPa, E = 200,000 MPa, for both the

existing leg and the reinforcement. The reinforcement is placed corner to corner to the
existing leg, forming a cruciform shape. Eleven lengths of the existing leg according to
L / r = 50, 60, 70, , 150 are considered. The ratio is limited to 150 which is the
174
maximum limit for leg members. Assume two connections are used per panel, located at
the positions similar to what the models in this chapter presented, dividing a panel into two
sub-panels. Two complete calculation examples are given below, and the results of all are
tabulated in Table 6.14. Figure 6.26 shows the change in the calculated nominal capacity
of the reinforced leg at different L / r ratios of the existing leg. Figure 6.27 presents
relationship between the capacity increase potential of the reinforced leg and the L / r ratio
of the existing leg.

Properties of existing leg and reinforcement, single 75 75 6 EA

Ag = 867 mm2
Iy = 0.187 10 6 mm4
ry = 14.7 mm
Ix = 0.722 10 6 mm4
rx = 28.9 mm
x5 = distance from centroid to heel, parallel to the strong axis
= 29.0 mm

Properties of combined section

Ag = 2 867 = 1734 mm2


Iv = 2 [0.187 10 6 + 867(29) 2 ] = 1.832 10 6 mm4
rv = 1.832 10 6 / 1734 = 32.5 mm
Iu = 2 (0.722 10 6 ) = 1.444 10 6 mm4
ru = 1.444 10 6 / 1734 = 28.9 mm

Case 1L/r = 50, two connections per panel

e = 50
L = 50 14.7 = 735 mm

Determine the critical loads of the existing leg and reinforcement, Qcr and Pcr .
Since they are of the same section size, their critical loads are the same.

c = 2 E / f y = 2 (2 10 5 ) / 320 = 111.07
For e < c ,
c = [1 0.5(e / c ) 2 ] = [1 0.5(50 / 111.07) 2 ] = 0.899
Qcr = Pcr = c f y Ag = 0.899 320 867 / 1000 = 249 kN

175
Determine the ultimate load of the compound leg Pu , assuming load on the existing
leg at the time of reinforcing Qo is zero.

(e ) u = L / ru = 735 / 28.9 = 25.43


1 = L / rv = 735 / 32.5 = 22.62
2 = C / rmin = ( L / 2) / ry = (735 / 2) / 14.7 = 25
(e ) v = 1 + 2 = 22.62 2 + 25 2 = 33.71
2 2

Since (e ) v > (e ) u , therefore e = (e ) v = 33.71.


For e < c ,
c = [1 0.5(33.71 / 111.07) 2 ] = 0.954
Pu = 0.954 320 1734 / 1000 = 529 kN

Check critical length L1 for post buckling,

L1 = 2 EI r /( Pcr + 0.822Qcr )
= 2 (2 10 5 ) (0.187 10 6 ) /((249 + 0.822 249) 1000)
= 901 mm

Check critical length L2 for elastic buckling,

L2 = 12 EI r / Qcr
= 12 (2 10 5 ) (0.187 10 6 ) /(249 1000) = 1342 mm

Since ( L = 735 mm) < ( L1 = 901 mm), the reinforced leg has the capacity equal to
the compound leg, or

W = Pu = 529 kN

The percentage of capacity increase is


= (529 249) 100 / 249 = 112 %.

Case 2L/r = 120, two connections per panel

e = 120
L = 120 14.7 = 1764 mm

Determine the critical loads of the existing leg and reinforcement, Qcr and Pcr .
Since they are of the same section size, their critical loads are the same.

c = 2 (2 10 5 ) / 320 = 111.07
For e > c ,

176
c = 2 E /( f y (e 2 )) = 2 (2 10 5 ) /(320 120 2 ) = 0.428
Qcr = Pcr = 0.428 320 867 / 1000 = 119 kN

Determine the ultimate load of the compound leg Pu , assuming load on the existing
leg at the time of reinforcing Qo is zero.

(e ) u = 1764 / 28.9 = 61.04


1 = 1764 / 32.5 = 54.28
2 = (1764 / 2) / 14.7 = 60
(e ) v = 54.28 2 + 60 2 = 80.91
Since (e ) v > (e ) u , therefore e = (e ) v = 80.91.
For e < c ,
c = [1 0.5(80.91 / 111.07) 2 ] = 0.735
Pu = 0.735 320 1734 / 1000 = 408 kN

Check critical length L1 for post buckling,

L1 = 2 (2 10 5 ) (0.187 10 6 ) /((119 + 0.822 119) 1000)


= 1306mm

Check critical length L2 for elastic buckling,

L2 = 12 (2 10 5 ) (0.187 10 6 ) /(119 1000) = 1943 mm

Since ( L1 = 1306 mm) < ( L = 1764 mm) < ( L2 = 1943 mm), the reinforced leg is
in the transition stage,

Pe = 2 EI r / L2 = ( 2 (2 10 5 ) (0.187 10 6 ) / 1764 2 ) / 1000 = 119 kN


P = Pe 0.822Qcr = 119 0.822 119 = 21 kN

where Pe is Eulers load of the reinforcement, and P is load on the reinforcement.

P Pcr O.K.

Capacity of the reinforced leg is:


W = Qcr + P = 119 + 21 = 140 kN

The percentage of capacity increase is


= 21 100 / 119 = 18 %.

177
Table 6.14 Critical loads of the existing leg, reinforcement, and reinforced leg made of 75
75 6 EAs, and % of capacity increase determined by the suggested method

L C = L/2 Qcr Pcr Pu L1 L2 Pe P W % Capa.


L/ry Case
(mm) (mm) (kN) (kN) (kN) (mm) (mm) (kN) (kN) (kN) Increase
50 735 368 249 249 529 901 1342 1 - - 529 112
60 882 441 237 237 518 925 1376 1 - - 518 119
70 1029 515 222 222 505 955 1421 2 349 166 388 75
80 1176 588 205 205 489 993 1478 2 267 98 303 48
90 1323 662 186 186 472 1043 1552 2 211 58 244 31
100 1470 735 165 165 453 1108 1649 2 171 35 200 21
110 1617 809 141 141 431 1197 1782 2 141 25 166 18
120 1764 882 119 119 408 1306 1943 2 119 21 140 18
130 1911 956 101 101 382 1414 2105 2 101 18 119 18
140 2058 1029 87 87 355 1523 2267 2 87 15 103 18
150 2205 1103 76 76 325 1632 2429 2 76 13 89 18

Note: fy = 320 MPa used in calculation


Qcr = critical load of the existing leg
Pcr = critical load of the reinforcement
Pu = ultimate load of the compound leg, assuming load on the existing leg at the time of
reinforcing is zero.
L1 = first critical length (Equation 2.10)
L2 = second critical length (Equation 2.11)
Pe = Eulers load of the reinforcement
P = load on the reinforcement (Equation 2.9)
W = capacity of the reinforced leg
= Pu for L < L1
= Qcr + P for L1 < L < L2

6.8.3 Discussion
The method is easy to follow. From Figure 6.27, the capacity increase potential may be
divided into three ranges. The first range is when the panel length of the existing leg is
long, L / r 110 , the increase potential is low and constant, at 18%. The second range is
when the L / r is between 65 and 110. As the ratio decreases, the capacity increase
potential increases exponentially, from 18% to 100%. The third range is when the panel
length is very short, with L / r 65 , the increase potential is 100% or higher. In general,
the curve indicates that the larger the slenderness ratio is, the lesser the potential can be
expected. The curve in Figure 6.27 was created from using f y = 320 MPa. When other f y

values are considered, their capacity increased potential curves will also be in similar
pattern, but with different specific values.

178
Figure 6.26 Nominal capacity of cruciform reinforced leg (2 75 75 6 EA) from
Browns method

Figure 6.27 Capacity increase potential of cruciform reinforced leg (2 75 75 6 EA)


from Browns method
179
In the FE modelling part, the modelling results suggested that an existing leg with L / r
equal to either 80, 100, or 120 has the potential to double its capacity after reinforced.
However, using Browns method, the capacity increase potentials of those three L / r cases
are 48%, 21%, and 18% only. These results from the two sources show a big difference.
Hence, it is necessary to conclude that, to be able to know the true potential of leg
reinforcing method, experimental investigations on full-scale or reduced-scale multi-panel
towers are needed.

The method of Brown gives conservative results because of his assumptions used in
developing the method. With no supporting testing evidences, the author believes that his
method is reliable and recommends it. It is important to bear in mind that while it is
possible to expect a high strength increase from an existing leg with short panels, the
reinforcement should be provided adequately by extending the reinforcing member for
some extra panels beyond the location of technical termination approximated by the results
of the truss analysis.

Normally the tower legs are designed as single angle members. Hence, when we talk about
compound legs, they are likely to be reinforced legs. AS 3995 (1994) does not specify
clearly whether compound legs mean reinforced legs, but only uses the term compound
leg members. Observe the values of Qcr (capacity of the existing leg), Pu (capacity when

the leg is considered as a compound leg), and W (capacity when the leg is considered as a
reinforced leg) in Table 6.14. For a large L / r ratio, the difference between the Pu and

W values is very large (the larger the L / r ratio is, the larger the difference between the Pu

and W values is). Also comparing between Pu and Qcr , Pu is even larger than 2Qcr . For

example, at L / r = 80 , Pu = 2.39Qcr ; and at L / r = 120 , Pu = 3.43Qcr . Therefore, it can be


a mistake if an engineer perceives reinforced legs as compound legs, follows the design
code straightforwardly to determine the design capacity, and expects additional capacity
more than twice the original design without exercising rational judgement. Hence, future
studies should try to clarify the real meaning of compound legs whether they are reinforced
legs or not. If not, is the method of determining the capacity of compound legs still
applicable to reinforced legs?

180
6.9 Conclusions
The objectives of the chapter were to find 1) the effectiveness of the leg reinforcing
method for compression legs, 2) how to make the reinforced leg achieve the desired
capacity increase, 3) the effect of bracing patterns and slenderness ratios, and 4) whether it
was possible to use simple elastic-analysis models to predict the critical connection shear
force. Since the first three objectives had issues that are interrelated, their conclusions
cannot be separated. The last objective was 5) to suggest the method of Brown (1988) for
calculating the capacity of reinforced legs and demonstrate it through examples.

Two sets of models were used in the study. The tower models of the first set were created
using finite elements. The first set consisted of four models. Three models had the L / r
ratio of the common unreinforced leg panels equal to 80, 100, and 120, where their bracing
patterns on the vertical faces were cross-bracing. The other model had the L / r ratio of the
common unreinforced leg panels equal to 80, but the bracing pattern on the vertical faces
was Warren-bracing. The towers had three legs forming a right triangular section which
tapered from the base to the top. The leg at the right-angle corner was vertical and the legs
at the acute corners were inclined. Loads were applied to create compression on the
vertical leg and tension on the inclined legs. Truss analysis was used to approximate
member forces, and then the approximate member forces were used to select the section
sizes for the members in the FE models. Since the studys focus was on reinforcing the
compression leg only, only the vertical compression leg was reinforced. The other
members (the inclined tension legs and the bracing members) had no reinforcement, but
were pre-designed for the force levels they would carry if the capacity of the towers could
increase by 100%. The results from the truss analysis were also used to determine how
many panels from the base the compression leg needed to be reinforced. Slip connections
were applied to stitch-connect the reinforcing member to the compression leg. Then the FE
analysis was performed. In the case that the target 100% capacity increase was not
achieved with such provision of the reinforcing member, the reinforcing member was
extended upward for one or two panels and reanalysis was performed.

Based on the results of the four FE tower models, it was determined for each model, how
many panels of the compression leg needed to be reinforced in order to achieve the target
capacity increase. Then a normal structural elastic analysis software was used to create
another set of models which imitated the successful FE models. The vertical leg, the

181
reinforcing member, and the connectors were beam-typed members, connected together
node to node. This means of modelling did not allow connection slip. The size of the
connectors to be defined in the software was varied, based on the moment of inertia ratio
of the connector to the existing leg ( I y ,con / I y ,existing ) from 1.0 down to 0.2. Then the shear

forces at the top connection of these simple models were compared with those of the FE
models, to find which ratio gave the nearest shear value.

The important findings of this chapter are summarized below.

1 From the FE analysis results, for the model case whose L / r of the common panels
was equal to 80 and whose bracing pattern was cross bracing, while the target capacity
increase was 100%, applying the reinforcement to the minimum amount as predicted
by truss analysis helped increase the capacity by 73%. The structure failed by flexural
buckling of the existing leg at the first unreinforced panel. When the leg was reinforced
for one more panel, the target capacity increase was achieved, and the weakest or
controlling member became the lowest panel of the tension legs which experienced
yielding.

For another model with the same L / r = 80 case, but whose bracing pattern was
Warren bracing, the same kind of results was also obtained.

For the cases of L / r = 100 and 120 with cross-bracings, using the minimum amount
of reinforcement as predicted by the results from truss analysis was enough to provide
the target capacity increase.

These results reflect the shear lag effect that is critical when the slenderness of the
existing leg member is low (which is the case of most tower leg members). To make
the legs achieve the target capacity increase, more development length must be
provided. It can be simply done by providing the reinforcing member more than it is
technically needed for one or two panels. However, providing more than that will not
result in further capacity increase.

2 In general cases, the amount of force transfer to the reinforcing member is largest at the
top connection, and reduces at the lower connections until at one connection, and the
connectors below it, there is no transfer. Almost all force transfer takes place at the
connections over about the upper half of the reinforced length. The no transfer, or zero
connection shear force, happens because the relative displacement of the existing leg
182
and the reinforcing member over that zone is not enough to make the bolts touch the
bolt holes (the contacts required for the bearing condition are not complete). However,
the connectors over this zero shear zone do serve function by stitching the reinforcing
member to the existing leg. The unsupported length of the existing leg is reduced, and
its capacity increases.

The forces on the reinforcing member of the model with cross bracings and the model
with Warren bracings, both of the L / r = 80 case, were plotted against the elevation
and compared. The two force lines almost overlaid each other. This can be reasoned
that the difference in force distribution to the reinforcing member that a bracing pattern
can make is reduced or eliminated by the use of slip-bearing connections. In other
words, due to the slip, the bracing pattern may not be so significant (bearing in mind
that the two models in this comparison were of the same L / r case). The reinforced
length seems to be the important factor that will make the difference.

3 The above conclusions were obtained from the models that used two connections per
panel. Then, two cases were attempted with only one connection per panel. Both cases
gained more capacity by 50% only, and the structures failed by flexural buckling of the
existing leg over one of the lowest two panels. Therefore, using one connection per
panel is ineffective.

4 Since the elastic models could not simulate slip, there was force transferred from the
existing leg to the reinforcing member at every connection. The distribution of forces
on the existing leg and reinforcing member was different from that of the FE models.
Forces on the reinforcing member ended up larger than forces on the existing leg,
which is wrong. However, elastic models could be used to approximate the maximum
connection shear force.

In using the elastic models to predict the maximum shear force at the connections, the
results from all elastic models appeared to be the same. When the same section as that
of the existing leg was defined to the beam-member connectors ( I y ,con / I y ,existing = 1.0),

the maximum shear value obtained was conservativemuch higher than the maximum
shear value from the corresponding FE model. The value decreased when a smaller
section was defined. The maximum shear value from an elastic model was closest to
the maximum shear value from the corresponding FE model when a section that made

183
the I y ,con / I y ,existing ratio equal to 0.025 was defined to the beam-member connectors.

Therefore, this I y ,con / I y ,existing = 0.025 is recommended in selecting the section to be

defined to the beam-member connectors in elastic models. Although this approach of


getting the maximum shear value seems to be speculative, it can be helpful for practical
design engineers, who mostly use a structural elastic analysis software, rather than a
finite element analysis software.

5 Browns method uses two length values, L1 and L2 , to indicate how to consider a
reinforced leg with a length L as.

If L L1 , the reinforced leg is considered as a compound leg. The calculation results


indicated that this case has the capacity increase potential as much as 100%, when two
connections were used per panel and the panel was subdivided into two sub-panels.

If L1 < L < L2 , the reinforced leg is considered as a true reinforced leg. The calculation
results showed that in this case the capacity increase potential will decrease as the L / r
increases.

From the above two cases, the general trend from using Browns method is: the longer
the slenderness ratio is, the lesser the potential can be expected. This trend is opposite
to the trend observed from the results of the FE models.

For a leg member that meets the limiting slenderness ratio requirement ( L / r 150) ,
the case of L L2 will not need to be considered.

184
CHAPTER 7

7 Summary, conclusions, and recommendations


7.1 Reinforced one-panel leg testing
7.1.1 Objectives and summary
The objectives of the testing program that was described in Chapters 3 and 4 were:

To determine the most effective connection type among the three different types of
connections, why that connection type was better than the other two, and what
particular aspects the design should pay attention to.

To investigate how significantly using pretensioned bolted connections could improve


the performance of reinforced legs.

To obtain a set of test results that could be used to calibrate a FE model.

Every test specimen in the study consisted of two 75 75 6 equal angles, which were the
main (or existing) member and the reinforcing member. They were 1180 mm and 1170
mm in length respectively. The slenderness ratio of the main member was 76.9; and with
the fix-pin end condition, its theoretical effective slenderness ratio was 53.9. Both
members were put together in cruciform shape, with the main member extruding beyond
the reinforcing member for 10 mm at the top end.

Every specimen had four connectionsone at the bottom end, which was the bottom
support itself, another at the top end, and the other two at approximately the one-third
points. The top connection used two angle cleats put in cruciform shape for every
specimen. (In the early stage, the top connector was also tried with single plate.) The two
intermediate connections had three different types (or arrangements). The first type,
namely Cruciform, used two angle cleats as the connectors for each intermediate
connection, also put in cruciform shape. The second type, namely Aligned, used one single

185
plate as the connector for each intermediate connection and the plates of both connections
were aligned on the same plane. The third type, namely Alternating, used one single plate
for each intermediate connection and both were placed in perpendicular planes.

All specimens used bolted connections. Bolts were tensioned to provide two levels of
tension forces to the bolts. The first was snug-tight level, and the second was pretension
level (0.7 times the minimum tensile strength of bolts). This was done for the second
objective. In addition, to create a reference of how the performance would be if the slip
between the members was totally not allowed, another two specimens with welded
connections were built.

At the top end of every specimen, load was applied to the top end of the main member
only, at its cross-sections centroid. This was intended to simulate the real condition of a
reinforced tower leg at the area where the reinforcing member is terminated, leaving the
existing leg to carry load alone. The top and bottom supports were made to provide pin end
and fix end conditions respectively.

The physical behaviours were observed, while the internal forces were measured for a clear
evaluation in the respect to the objectives.

7.1.2 Conclusions drawn from the one-panel testing


7.1.2.1 Importance of different connections

The performance of the reinforced panel specimens relied greatly on the top end
connections and their designs, whilst the intermediate connections had secondary
importance. (Since the bottom end connection, being the bottom support itself, represented
the actual base condition of general reinforced tower legs, it was not considered.)

The amount of force that transferred from the main member to the reinforcing member was
largest at the top end connection, and decreased at the first and second intermediate
connections. However, this desirable force transfer mechanism did not happen when a
single plate was used as the connector at the top connection, which in this case created an
ineffective top connection because the main and reinforcing member were connected
together on two out of four faces only. With an ineffective top connection, force could not
transfer to the reinforcing member effectively, resulting in no or a very small strength
increase benefit.

186
To achieve the desirable force transfer mechanism, the top connection needed to be the
cruciform connection type (using two angle cleats as the connectors and putting them in
cruciform shape) to make it able to carry the majority of force transfer to the reinforcing
member effectively.

While the intermediate connections had secondary importance, they were able to help
improve the overall capacity of a reinforced leg for some extent if a good connection
design was used. Among the Cruciform, the Aligned, and the Alternating connection types,
the Cruciform type provided the best outcome. This was because all four faces of the main
and reinforcing member could be tied together with two angle cleats; where only two faces
(one face from each member) could be tied when the Aligned or Alternating connection
type was used.

7.1.2.2 Snug-tight bolted connections vs. pretensioned bolted connections

In the case where the connections were snug-tight bolted, the performance (or capacity
increase percentage) of the specimens varied over quite a wide range. Two failure modes
were experienced, which were flexural bucking and torsional buckling. The flexural
buckling mode had the two component members flex together. This mode was rare, but
provided the highest performance (as much as 50.8% capacity increase). For the torsional
buckling mode, the first sub-panel of the main member experienced a severe twisting. This
mode was dominant, but provided a lower or much lower performance than the flexural
buckling mode. (See Table 4.1 and Figure 4.1.)

In the case that the connections were pretensioned bolted, the specimens performed more
consistently. Only one failure mode was experienced which was from a slight torsional
buckling or twisting of the first sub-panel of the main member. This mode was compatible
to the torsional buckling mode of the snug-tight bolted case, but the degree of twisting was
much less.

Using pretensioned bolts connections had two benefits. One was that it helped ensure a
more consistent or predictable performance of a reinforced leg. Second, from comparison
between the average performance of the specimens with snug-tight bolted connections and
that of the specimens with pretensioned bolted connections which experienced the same
failure mode by twisting of the first sub-panel of the main member, using pretensioned
bolted connections helped increase the performance higher than that obtained from using
snug-tight bolted connections, between 13.6% and 21.4%.
187
Therefore, the perceptionthat snug-tight bolted connections are more flexible than
pretensioned bolted connections, and when used will bring about a lower performance (the
capacity increase percentage) to a reinforced leg than if the pretensioned bolted
connections are usedwas proved to be correct. However, from the construction point of
view, pretensioning the bolts is not practical or is not accepted due to the high fitting cost
that will incur. Hence, only the snug-tight bolted condition is being adopted due to its
simplicity. Although, not the best option, the test results showed that it could produce
appreciable performance. It is wise to expect for the worst performance such as when the
reinforced towers are in service, the nuts have loosened, and there is no tension left in the
bolts.

Among the welded, the cruciform, the alternating, and the aligned connection types, the
welded connection type gave the best performance because the connections did not allow
slip between the component members. But again, this option will not be adopted in
practice.

For the intermediate connections, using cruciform connections with bolts tightened snug-
tight was more effective than using plate connections with bolts pretensioned.

7.2 One-panel finite element models


7.2.1 Objectives and summary
The main objective of Chapter 5 was to develop a reliable and computationally economical
finite element modelling method (or design) of the reinforced one-panel leg model which
simulated the reinforced one-panel test specimens. In addition, the method was required to
have the ability to be applied further to create reinforced multi-panel tower models
successfully.

Three model types were created, starting from the most detailed and uneconomical
(resource-and-time-consuming) type to the simplest and most economical type.

The first model type had every partthe main member, the reinforcing member, the
connectors (plates or angle cleats), and the bolts (including nuts and washers)created
separately, all using continuum elements. Four models were created with four different sets
of the bolt parts (the bolt parts of all four sets had the same design, but different sets had
different shank lengths), used to create four levels of tension in the bolts, from high to
moderate, nearly zero, and zero tension. Slip was allowed by making the bolt holes larger
188
than the bolt shanks diameter by the standard size. The appearance after assembly was the
same as the test specimens.

The second model type had the same design as that of the first model type, except that the
sub-panel portions made of continuum elements were substituted by beam elements. The
connection portions made of continuum elements and the sub-panel portions made of beam
elements were linked together by coupling constraints. For this model type, only the case
of zero pretension was considered.

The third model type utilised a new connection design which used encasing rings to encase
and hold the main and reinforcing members together. The encasing rings functioned as the
connectors. The portions of the main and reinforcing members that were encased were
made of continuum elements. These portions were called Core parts. The encasing rings
were also made of continuum elements. To allow force transfer to happen, a locking
system was created. The core parts were provided with locking stripes which extruded out
from their surfaces; where the encasing rings had locking grooves to lock with the locking
stripes. In this place, the locking stripes and the locking grooves functioned as the bolts
and the bolt holes respectively. To allow slip to happen, the width of the locking grooves
was made wider than the width of the locking stripes. The new connection design required
much fewer interaction definitions than the detailed connection design of the first and
second model types. Again, the sub-panel portions used beam elements to save the
computational resource and analysis time, the same as the second model type.

Because member slip was a required property of the models, the parts representing the
main and reinforcing members, the connectors, and the bolts had to be connected together
by surface contact interactions. When a model had surface contact interactions, the
buckling analysis which was an analysis procedure in ABAQUS/Standard could not be
applied. Hence, the deformed shapes from the buckling analysis were not available to be
factored and introduced to create the geometrical imperfections for the model. As a result,
the models had to be analysed with perfect geometry. And since the models had complex
contact problem, to be able to handle it and run the analysis successfully, Dynamic Explicit
(Quasi-Static) analysis procedure in ABAQUS/Explicit was used.

7.2.2 Conclusions
According to the main objective of the chapter, as far as the reliable and economical FE
modelling method could be found, all the modelling techniques and parameters used
189
should be acceptable. Therefore, the important findings were related to the modelling
techniques and parameters used.

In the development of the reinforced models where slip was an important factor and must
be allowed to happen, the models had to be created and analysed in perfect geometry. As a
result, it was necessary to accept the disadvantage about some degrees of error of the
modelling results as compared with the test results.

The plain Stiffness type hourglass control in ABAQUS/Explicit is recommended rather


than the Enhanced Stiffness type, as the latter tended to make the models become a little
stronger.

The general perception that using different element types (which in this place meant 3D
continuum and beam element types) for sub-panel portions will make different results is
not correct (see Section 5.5.2), at least for this study where slip was a requirement for the
design of connections. With no buckling analysis, there is no failure mode to be factored to
form imperfect geometry shape for the models; and then using continuum, shell, or beam
elements for sub-panels will give almost identical results. So, to spend least analysis time,
beam elements should be used. To introduce geometrical imperfection, node coordinates
need to be specified manually, but before that, a very careful consideration will be
required.

The first connection design had bolts, connectors, main and reinforcing members created in
separate pieces. This design demanded numerous interactions to be defined, which as a
result required a very large space for input data. The second (simplified) connection design
used encasing rings to wrap around the main and reinforcing members and unify them
together, with locking grooves and locking stripes extruding out, to function as bolt holes
and bolts respectively in transferring forces. This design required much less space for input
data and was able to replicate the first design satisfactorily for bolts having no pretension.

7.3 Multi-panel FE modelling


7.3.1 Objectives and summary
The economical FE modelling method developed in Chapter 5 was used to build FE multi-
panel tower models with the compression leg reinforced. The objectives of this part of the
study were:

190
To investigate the relationship between the panel length (or slenderness) of the existing
legs and the effectiveness of the leg reinforcing method.

To check whether the existence of different bracing patterns especially in the zone of
termination of the reinforcing member would cause any difference to the effectiveness
of the leg reinforcing method.

To find the effectiveness of the leg reinforcing method when the reinforcement was
provided to the minimum length, and if the 100% capacity increase could not be
achieved, how many more panels the reinforcing member needed to be extended to
achieve the desired capacity increase.

To consider value engineering.

To investigate how well normal structural elastic analysis software could be used to
predict the force distribution on an existing leg and its reinforcing member, since the
elastic model would not allow connection slip.

To try to find the size of the members representing the connectors which could bring
the maximum connection shear force as close as possible to that from the FE models.

To examine the use of Browns method for calculating the capacity of a reinforced
column to see if it applied well in these cases.

First, two real towers were investigated by truss analysis to find the minimum number of
panels that the tower models should have if their compression legs were to be reinforced
with the expectation of a 100% capacity increase, disregarding the shear lag that might
occur at the connections of the reinforcing member to the existing compression leg. The
results of the truss analysis indicated that the minimum number was six or seven panels.
Hence, the models were developed with nine panels, allowing the reinforcing member to
be extended for two or three extra panels to handle the shear lag effect.

The general configuration of the tower models was that each had three legsone leg
vertical and two legs inclinedforming a right triangular cross-section with the vertical
leg at the right triangle corner. The cross-section tapered from base to the top. Loads were
applied to cause compression on the vertical leg only so that only the vertical leg needed to
be reinforced. All these arrangement were intended to save the computational resource.

191
For the first objective, three cases of slenderness ratios of the common panels of the
(unreinforced) vertical leg were considered. They were L / r = 80, 100, and 120, with
respect to 75 75 6 EA and its radius of gyration.

For the second objective, two bracing patternscross bracing and Warren bracingwere
used for the bracing members on the two vertical faces.

In total, four models were created. The first three models used cross bracing, with the L / r
ratios of the common panels of the vertical leg equal to 80, 100, and 120 respectively. The
fourth model used Warren bracing, with the L / r equal to 80.

With the geometry and outline of each model ready, first the model was created as an all-
truss-member structure. Then, assumed loads were applied and truss analysis was used to
get the assumed member force of every member. As single 75 75 6 EA was selected
for the vertical leg and its design single-member compression strength of every panel leg
was calculated. The assumed member forces of the whole model were multiplied by a
factor equal to two times the design compression strength of the lowest panel of the
vertical leg over the obtained assumed member force of the same panel. This action was to
approximate the corresponding or expected forces in the members, in case doubling
capacity of the vertical leg was successful. For the vertical leg, as it was to be reinforced,
the design single-member strength and the expected force of each panel were compared to
find at which panel the reinforcing of the vertical leg could be stopped. The 75 75 6
EA was also selected for the reinforcing member. The inclined legs and the bracing
members were independent members with no reinforcement and their section sizes were
selected to be adequate for the expected forces they were to carry.

With all the section sizes and where to stop the reinforcing member known, four FE
models of the same dimensions and bracing arrangements as the four truss models were
created. The vertical leg and the reinforcing member used the simplified design of the
connections and sub-panel portions developed in Chapter 5, but only two connections were
provided per panel. The other members (the inclined legs and bracing members) were
defined as truss elements to save analysis time, as they were not the members of interest.
Each model had a number of cases. The first case had the vertical leg reinforced to the
minimum amount (six panels). The second and the third cases had the vertical leg
reinforced for one more and two more panels (seven and eight panels) respectively. This
part was conducted for the third objective.
192
For the fourth objective, another two FE models of the L / r = 80 case, one with cross
bracings and one with Warren bracings, were created, and they were provided with only
one connection per panel.

For the fifth objective, four elastic models were created using a normal elastic structural
analysis software. The vertical leg, the reinforcing member, and the connectors were
assigned as beam-typed members, while the other members were assigned as truss-typed
members. For the sixth objective, the member size of the connectors was varied in relation
to the size of the existing vertical leg. The term used was I y ,con / I y ,existing , the moment of

inertia of the connector to the moment of inertia of the existing leg.

Lastly, the method of Brown (1988) was used to calculate the capacities of a number of
reinforced legs with different lengths from short to slender.

7.3.2 Conclusions
7.3.2.1 Effectiveness of using truss analysis in indicating the minimum
reinforcement

Unreinforced lattice towers are typically modelled as all-truss-member structures and


analysed by truss analysis. When a tower needs to be upgraded and the method of leg
reinforcing is selected, it is possible to use truss analysis results of that unreinforced tower
to approximate the minimum amount (minimum number of panels) of reinforcement to be
provided to the existing legs.

The results from the FE models showed that providing the reinforcement to the
approximated minimum amount was able to increase the capacity of the tower up to or
near the desired capacity level.

The panel force on a leg is largest at the lowest main force panel and is smaller and smaller
on the upper main force panels. This means when a leg needs to be reinforced, the strength
of the lowest main force panel is of concern. This implies that if we can successfully
reinforce the leg until the point that, when the tower is subjected to loads, the combined
force from the existing leg and reinforcing member of the lowest main sub panel can reach
two times the strength of the existing leg alone, the tower should be considered as having
achieved twice its original capacity as well.

193
7.3.2.2 Conclusions from the FE models

The panel forces on a leg increase from the top main force panel to the bottom main force
panel, but the magnitude of force increment decreases from the top to the bottom main
force panel. This pattern of panel forces on the legs is always true. It does not change even
when the (existing) legs are reinforced, because reinforcing the (existing) legs has nothing
to do with the bracing membersthe positions and arrangements of the bracing members
remain the same. As a result, the same proportions of force increments will still be
transferred from the bracing members to the existing legs.

However, when the existing legs are reinforced, the results of the FE models showed that
there was no force transferred from the existing leg to the reinforcing member at the
connections over the lower half of the reinforced length. Only the connectors over the
upper half of the total reinforced length were active. The amount of force transfer
(connection shear force) was largest at the top connection, decreased more and more at
lower connections until it became zero (zero shear force) at the connection located around
the middle of the reinforced length. This force transfer mechanism is known as shear lag
effect which comes with slip connections.

Therefore, if the additional capacity required from reinforcing a leg is high, such as in the
case that an existing leg with short panel length is demanded to be reinforced, enough
number of connections should be provided to allow half of them to worktransferring
some force from the existing leg to the reinforcing memberand finally after
accumulation deliver the required additional capacity. In other words, enough panels of the
existing leg must be reinforced for enough active connections to handle the shear lag
effect.

The results from the FE models indicated that bracing pattern was not an important factor.

The two FE models with the L / r of the common panels equal to 80, one with cross-
bracings and the other with Warren bracings, provided with the minimum amount of
reinforcement approximated from the truss analysis results (applying the reinforcement to
the existing leg for six panels out of nine), had close capacity increase percentages,
although not as high as 100% (73.0% and 80.1% from the cross-bracing and Warren-
bracing models respectively). The two models experienced the same failure mode which
was flexural buckling of the existing leg at the first unreinforced panel.

194
Also when the panel force lines of the reinforcing member of both models were plotted
together, the two force lines almost overlaid each other. This implied that the difference in
force distribution to the reinforcing member that different bracing patterns could make was
reduced or eliminated by the use of slip-bearing connections. In other words, due to the
slip, the bracing pattern may not be significant.

In this L / r = 80 case, by adding the reinforcement for one more panel than the minimum
amount approximated, both models could reach the 100% target capacity increase
percentage.

The panel length (or slenderness) for the existing leg appeared to be a very important
factor.

For the cases whose L / r of the common panel length were 100 and 120, with cross-
bracing, using the minimum amount of reinforcement as approximated by the results from
the truss analysis (applying the reinforcement to the existing leg for six panels out of nine)
was enough to double the capacity of the structures. This was reasoned to be because the
panel length of the existing leg of either case was in the range of intermediate to slender
length, the strength of the existing leg alone was not too high, and therefore the amount of
the additional strength expected from reinforcing was not too high. By providing the
reinforcement to the minimum amount as approximated by the results of the truss analysis,
the amount of the connections within the reinforced length was just enough that the upper
half, active connections were able to transfer sufficient force from the existing leg to the
reinforcing member to deliver the required additional strength.

Caution should be exercised when using the truss analysis results to approximate the
minimum amount of reinforcement. The FEM results showed that the shear lag effect is
critical when the slenderness of the existing leg member is low ( L / r 80) which is the
case of most tower leg members. To make the legs achieve the target capacity increase,
more development length must be provided than that indicated by the truss analysis. It can
be simply done by extending the reinforcing member more than the approximated
minimum amount for one or, for certainty, two panels. However, providing more than that
is unlikely to result in any additional strength gain.

The no transfer, or zero connection shear force, in connections below the approximate half
way point happened because the relative displacement of the existing leg and the

195
reinforcing member over that zone was not enough to make the locking stripes
(representing the bolts) touch the locking grooves (representing the bolt holes), which
made the contacts required for the bearing condition incomplete.

However, the connectors over this zero shear transfer zone did still serve the function of
holding the reinforcing member to the existing leg at intervals. The unsupported length of
the existing leg was thus reduced, and its own capacity increased.

When each panel used only one connection to connect the reinforcing member to the
existing leg, the capacity increased by 50% only, and the structures failed by flexural
buckling of the existing leg over one of the lowest two panels. Therefore, using one
connection per panel is ineffective.

7.3.2.3 Conclusions from elastic models

The elastic models could not simulate slip. There was force transferred from the existing
leg to the reinforcing member at every connection. The distribution of forces on the
existing leg and reinforcing member was different from that of the FE models. However,
elastic models could be used to approximate the maximum connection shear force.

When the same section as that of the existing leg was defined to the beam-member
connectors ( I y ,con / I y ,existing = 1.0) , the maximum shear value obtained was conservative

well higher than the maximum shear value from the corresponding FE model. The value
decreased when a smaller section was defined.

The maximum shear value from an elastic model was closest to the maximum shear value
from the corresponding FE model when a section that made the I y ,con / I y ,existing ratio equal

to 0.025 was defined to the beam-member connectors. Therefore, this


I y ,con / I y ,existing = 0.025 is recommended in selecting the section to be defined to the beam-

member connectors in elastic models.

7.3.2.4 Browns method: the pattern of capacity increase percentage in relation to


the slenderness or panel length of the leg, and its use

Browns method uses two length values, L1 and L2 , to indicate how to consider a
reinforced leg with a length L as.

196
If L L1 , the reinforced leg is considered as a compound leg. The capacity of the
reinforced leg is equal to the capacity when it is considered as a compound leg member
(W = Pu ) . The calculation results indicated that this case has a capacity increase potential
of as much as 100%, when two connections were used per panel and the panel was
subdivided into two sub-panels.

If L1 < L < L2 , the reinforced leg is considered as a true reinforced leg. The capacity of the
reinforced leg is equal to the critical load (or capacity) of the existing leg plus load on the
reinforcing member ( W = Qcr + P ). The calculation results showed that in this case the

capacity increase potential will decrease as the L / r increases.

From the above two cases, the general trend from using Browns method is: the longer the
slenderness ratio is, the lesser the potential can be expected. This trend is opposite to the
trend observed from the results of the FE models.

Leg members whose slenderness ratios are less than the limiting slenderness ratio
L / r = 150 will be shorter than L2 , therefore the case of L L2 will not need to be
considered.

Since up to date there has been no experimental study on full-scale or reduced-scale leg-
reinforced towers, it is not possible to conclude that the FE modelling results and
understandings gained from this chapter were completely proven, except to say that they
were reasonable. Therefore, an existing method of Brown (1988) was examined.
Application of the method is easy, and the results can be conservative based on the
assumptions that Brown used. While the calculation will indicate a high capacity increase
potential for an existing leg with short panels, it must also be remembered that the
reinforcement should be provided adequately beyond the location of technical termination
approximated from the results of the truss analysis.

7.4 Summary recommendations for designers


The most important finding from this thesis for designers of transmission tower retrofitting
is that the method of reinforcing legs using additional angles bolted to them has been
shown to be effective. In addition, some significant design guidance can be provided from
the thesis as follows:

197
The angle reinforcement configuration to be used should be of a cruciform
arrangement, using two angle cleats and a minimum of two bolts per face. The
approach to determine the maximum connection shear force by using a structural
elastic analysis software to create elastic leg-reinforced models has been proposed.
Once the maximum shear force is known, the number of bolts can be determined from
the design shear strength and ply-in-bearing strength. All connections may have the
same design (as is normally done in common practice).

The first or top connection of the reinforcing member to the existing leg is critical for
load transfer and must be carefully designed and constructed. It is recommended that to
ensure that a reinforcing project has achieved its full benefit, the practice given in the
user note of ANSI/AISC 360-05 for built-up compression members should be adopted
(see Chapter 2, Section 2.7.1). Applying it, the top end connection shall be designed for
the full compressive load. For example, Panel 7 of Cross S80 tower model is expected
to carry the total force of 178.5 kN (Table 6.4). If the reinforcing member is extended
to cover Panel 7, the very top connection should be designed for 178.5 kN. The number
of bolts at the top will be determined from the design shear strength and ply-in-bearing
strength; but in this case the bolts must be pretensioned with Class A surfaces prepared.
The other connections should be safe if designed for the maximum connection shear
force (38.3 kN in Table 6.12); the number of bolts determined from the design shear
strength and ply-in-bearing strength; and the bolts just tightened snug-tight.

Two intermediate connections per panel should be provided, three are not necessary
and one is not sufficient.

Pretensioning of the bolts will provide more consistent strengthening outcomes and
slightly increase the capacity compared with snug-tight bolts. However, the current
field practice which does not pretension bolts is still satisfactory and if pretensioning is
used the most benefit would be at the top connection, as mentioned above.

The bracing pattern of a tower does not alter the effectiveness of the leg reinforcing
method, it worked equally well for cross-bracing and Warren-bracing.

A standard truss analysis of a tower can be used to determine the minimum number of
panels of reinforcement required for the existing legs. This was shown to be adequate
where the leg panel lengths were in the range of L / r = 100 to 120. However, for panel
198
slenderness ratios less than this, it is recommended that the reinforcing should be
provided for two more panels than predicted by the truss analysis.

Browns (1988) method can be used to predict the load carried by both the existing leg
and reinforcing member (two calculation examples are given in Section 6.8, they are
clear and easy to follow).

If a designer wishes to construct a FE model for designing leg reinforcing of towers,


then the following advice is provided:

The slip connection design using locking grooves and stripes with an encasing ring
that was developed in this thesis was recommended for future use as it proved to be
both accurate and computationally efficient.

With the design of the reinforced multi-panel tower, only the parts representing the
connections (the Cores and the Encasing Rings) were continuum elements. In Mesh
module, when assigning element type, the element library needs to be Explicit type
because the Dynamic Explicit analysis needs to be used. For geometric order, using
linear order is good enough. Hexagonal element type is recommended. When the
hexagonal element type is assigned, it is necessary to change the default hourglass
control type to plain Stiffness to make the parts have the strength equivalent to the
actual steel members.

7.5 Summary recommendations for future research


For those involved in the next revision of the Australian Standard AS 3995 (1994), it is
recommended that additional attention should be paid to clarify the real meaning of
compound legs as to whether they mean reinforced legs or not. Guidance should be
specifically provided in the code for the design of reinforced legs and the findings of this
thesis could be useful in developing that guidance.

For any additional future research on the topic of retrofitting of angle legs for increased
load capacity, it is recommended that the next stage should be conducting of full scale
and/or multi-panel laboratory testing of retrofitted tower legs. This will add to the findings
from this thesis.

For future FE modeling of retrofitted towers, the following recommendations are made:

199
FE tower models for future study should have initial geometrical imperfection. The
node coordinates for the beam elements of the sub-panel portions of both existing leg
and its reinforcing member need to be predetermined and entered manually. The
predetermination of the node coordinates will require a sound common sense of the
researchers. Once the models are initially imperfect, the results should be close to
reality.

For convenience and time saving for future researchers, they are welcome to duplicate
the reinforced FE tower models presented in this thesis, but make them initially
geometrically imperfect, so that the results can be compared.

It is recommended that the method of Brown in predicting strength increase be used as a


guideline for future researching on both physical models and computer models. And after
the actual potential of the leg reinforcing method, the force distribution mechanism, and
failure mechanism are more understood from the results of the physical models, Browns
existing method should be revised to fit the type of the job and structures.

200
References
ABAQUS Users Manuals, Version 6.5. Hibbitt, Karlsson and Sorensen Inc., 2005.

Albermani, F.G.A. & Kitipornchai, S. (2003). Numerical simulation of structural


behaviour of transmission towers. Thin-Walled Structures, Vol. 41, 167-177.

Albermani, F.G.A., Mahendran, M., & Kitipornchai, S. (2004). Upgrading of transmission


towers using a diaphragm bracing system. Engineering Structures, Vol. 26, 735-744.

Allen, H.G. & Bulson, P.S. (1980). Background to Buckling. New York: McGraw-Hill.

ANSI/AISC 360-05 (2005). Specification for Structural Steel Buildings. American Institute
of Steel Construction, March 9, 2005.

AS 1170 (1971). Australian Standard: SAA Loading Code. New South Wales, Australia:
Standards Australia.

AS 1170 (1989). Australian Standard: SAA Loading Code. New South Wales, Australia:
Standards Australia.

AS 1252 (1996). Australian Standard: High-Strength Steel Bolts with Associated Nuts and
Washers for Structural Engineering. New South Wales, Australia: Standards Australia.

AS 1391 (1991). Australian Standard: Methods for Tensile Testing of Metals. New South
Wales, Australia: Standards Australia.

AS 1559 (1997). Australian Standard: Hot-Dip Galvanized Steel Bolts and Associated
Nuts and Washers for Tower Construction. New South Wales, Australia: Standards
Australia.

AS 3995 (1994). Australian Standard: Design of Steel Lattice Towers and Masts. New
South Wales, Australia: Standards Australia.

201
ASCE (1988). Guide for Design of Steel Transmission Towers, Manuals and Reports on
Engineering Practice, No. 52 (2nd ed.). New York: American Society of Civil
Engineers.

Aslani, F. & Goel, S.C. (1992). Analytical criteria for stitch strength of built-up
compression members. Engineering Journal, Vol. 29, 102-110.

Astaneh-Asl, A., Goel, S.C., & Hanson, R.D. (1985). Cyclic out-of-plane buckling of
double-angle bracing. Journal of Structural Engineering, Vol. 111, 1135-1153.

Brown, J.H. (1988). Reinforcing loaded steel compression members. Engineering Journal,
Vol. 25, 161-168.

Citipitioglu, A.M., Haj-Ali, R.M., & White, D.W. (2002). Refined 3D finite element
modelling of partially-restrained connections including slip. Journal of Constructional
Steel Research, Vol. 58, 995-1013.

Hayward, A. & Weare, F. (1989). Steel Detailers Manual. Oxford, England: BSP
Professional Books.

Hong, K., Yang, J.G., & Lee, S.K. (2001). Parametric study of double angle framing
connections subjected to shear and tension. Journal of Constructional Steel
Research, Vol. 57, 997-1013.

Prasad-Rao, N. & Kalyanaraman, V. (2001). Non-linear behaviour of lattice panel of angle


towers. Journal of Constructional Steel Research, Vol. 57, 1337-1357.

Reid, J.D. & Hiser, N.R. (2005). Detailed modelling of bolted joints with slippage. Finite
Elements in Analysis and Design, Vol. 41, 547-562.

Sherman, D.R. & Yura, J.A. (1998). Bolted double angle compression members. Journal
of Constructional Steel Research, Vol. 46, 470-471.

Silva, J.G.S., Vellasco, P.C.G.S., Andrade, S.A.L., & Oliveira M.I.R. (2005). Structural
assessment of current steel design models for transmission and telecommunication
towers. Journal of Constructional Steel Research, Vol. 61, 1108-1134.

202
Temple, M.C., Sakla, S.S.S., Stchyrba, D., & Ellis, D. (1994). Arrangement of
interconnectors for starred angle compression members. Canadian Journal of Civil
Engineering, Vol. 21, 76-80.

Temple, M.C., Schepers, J.A., & Kennedy, D.J.L. (1986). Interconnection of starred angle
compression members. Canadian Journal of Civil Engineering, Vol. 13, 606-619.

Wakabayashi, M. & Nonaka, T. (1965). On the buckling strength of angles in transmission


towers. In: Bulletin of the Disaster Prevention, Research Institute, Koyoto University,
Japan, Vol. 15, No. 91, 1-18.

Zandonini, R. (1985). Stability of compact built-up struts: Experimental investigation and


numerical simulation. Construzione Metalliche, Milan, Italy, November 4, 202-224.

203
Appendix
A. Single panel test resultsLoad-displacement curves, sub-
panel force development curves, and connection shear
development curves
Cru-2B-S1, Test 1
400
Cru-2B-S1, Test 1, 328 kN, +23.6%
Cru-2B-S1, T1, 328 kN, +23.6%
350

300
Applied Load (kN)

250

200

150

100

50

0
0 1 2 3 4 5 6
Vertical Displacement (mm)

1.2
Cru-2B-S1, T1, 328 kN, +23.6%

1.0
N1
N1*
0.8 N2
N2*
N3
N/Py

0.6 N3*

0.4

0.2

0.0
0.0 0.2 0.4 0.6 0.8 1.0
P/Pc

204
0.40

Cru-2B-S1, T1, 328 kN, +23.6%


0.35

0.30 V1
V2
0.25 V3

V/Py
0.20

0.15

0.10

0.05

0.00
0.0 0.2 0.4 0.6 0.8 1.0
P/Pc

205
Cru-2B-S1, Test 2
400
Cru-2B-S1, Test 2, 384 kN, +44.7%
Cru-2B-S1, T2, 384 kN, +44.7%
350

300

Applied Load (kN)


250

200

150

100

50

0
0 1 2 3 4 5 6
Vertical Displacement (mm)

1.2
Cru-2B-S1, Test 2, 384 kN, +44.7%
Cru-2B-S1, T2, 384 kN, +44.7%

1.0 N1
N1*
N2
0.8 N2*
N3
N3*
N/Py

0.6

0.4

0.2

0.0
0.0 0.2 0.4 0.6 0.8 1.0
P/Pc

0.40
Cru-2B-S1, Test 2, 384 kN, +44.7%
Cru-2B-S1, T2, 384 kN, +44.7%
0.35
V1
0.30 V2
V3

0.25
V/Py

0.20

0.15

0.10

0.05

0.00
0.0 0.2 0.4 0.6 0.8 1.0
P/Pc

206
Cru-2B-S1, Test 3
400
Cru-2B-S1, Test 3, 349 kN, +31.9%
Cru-2B-S1, T3, 349 kN, +31.9%
350

300

Applied Load (kN)


250

200

150

100

50

0
0 1 2 3 4 5 6
Vertical Displacement (mm)

1.2
Cru-2B-S1, T3, 349 kN, +31.9%

1.0
N1
N1*
0.8
N2
N2*
N3
N/Py

N3*
0.6

0.4

0.2

0.0
0.0 0.2 0.4 0.6 0.8 1.0
P/Pc

0.40
Cru-2B-S1, T3, 349 kN, +31.9%
0.35

0.30
V1
V2
0.25
V3
V/Py

0.20

0.15

0.10

0.05

0.00
0.0 0.2 0.4 0.6 0.8 1.0
P/Pc

207
Cru-2B-S1, Test 4
400
Cru-2B-S1, Test 4, 352 kN, +32.7%
Cru-2B-S1, T4, 352 kN, +32.7%
350

300

Applied Load (kN)


250

200

150

100

50

0
0 1 2 3 4 5 6
Vertical Displacement (mm)

1.2
Cru-2B-S1, Test 4, 352 kN, +32.7%
Cru-2B-S1, T4, 352 kN, +32.7%

1.0

N1
N1*
0.8
N2
N2*
N/Py

N3
0.6
N3*

0.4

0.2

0.0
0.0 0.2 0.4 0.6 0.8 1.0
P/Pc

0.40
Cru-2B-S1, Test 4, 352 kN, +32.7%
Cru-2B-S1, T4, 352 kN, +32.7%
0.35

0.30 V1
V2
0.25 V3
V/Py

0.20

0.15

0.10

0.05

0.00
0.0 0.2 0.4 0.6 0.8 1.0
P/Pc

208
Cru-3B-S1, Test 1
400
Cru-3B-S1, Test 1, 345 kN, +30.1%
Cru-3B-S1, T1, 345 kN, +30.1%
350

300

Applied Load (kN)


250

200

150

100

50

0
0 1 2 3 4 5 6
Vertical Displacement (mm)

1.2
Cru-3B-S1, Test 1, 345 kN, +30.1%
Cru-3B-S1, T1, 345 kN, +30.1%

1.0

N1
N1*
0.8
N2
N2*
N3
N/Py

0.6
N3*

0.4

0.2

0.0
0.0 0.2 0.4 0.6 0.8 1.0
P/Pc

0.40
Cru-3B-S1, Test 1, 345 kN, +30.1%
Cru-3B-S1, T1, 345 kN, +30.1%
0.35

0.30

V1
0.25
V2
V3
V/Py

0.20

0.15

0.10

0.05

0.00
0.0 0.2 0.4 0.6 0.8 1.0
P/Pc

209
Cru-3B-S1, Test 2
400
Cru-3B-S1, Test 2, 400 kN, +50.8%
Cru-3B-S1, T2, 400 kN, +50.8%
350

300

Applied Load (kN)


250

200

150

100

50

0
0 1 2 3 4 5 6 7
Vertical Displacement (mm)

1.2
Cru-3B-S1, Test 2, 400 kN, +50.8%
Cru-3B-S1, T2, 400 kN, +50.8%

1.0
N1
N1*
0.8 N2
N2*
N3
N/Py

0.6 N3*

0.4

0.2

0.0
0.0 0.2 0.4 0.6 0.8 1.0
P/Pc

0.40
Cru-3B-S1, Test 2, 400 kN, +50.8%
Cru-3B-S1, T2, 400 kN, +50.8%
0.35

0.30
V1
0.25 V2
V3
V/Py

0.20

0.15

0.10

0.05

0.00
0.0 0.2 0.4 0.6 0.8 1.0
P/Pc

210
Cru-3B-S1, Test 3
400
Cru-3B-S1, Test 3, 348 kN, +31.3%
Cru-3B-S1, T3, 348 kN, +31.3%
350

300

Applied Load (kN)


250

200

150

100

50

0
0 1 2 3 4 5 6
Vertical Displacement (mm)

1.2
Cru-3B-S1, Test 3, 348 kN, +31.3%
Cru-3B-S1, T3, 348 kN, +31.3%

1.0 N1
N1*
N2
0.8 N2*
N3
N3*
N/Py

0.6

0.4

0.2

0.0
0.0 0.2 0.4 0.6 0.8 1.0
P/Pc

0.40
Cru-3B-S1, Test 3, 348 kN, +31.3%
Cru-3B-S1, T3, 348 kN, +31.3%
0.35

0.30 V1
V2
V3
0.25
V/Py

0.20

0.15

0.10

0.05

0.00
0.0 0.2 0.4 0.6 0.8 1.0
P/Pc

211
Cru-3B-S1, Test 4
400
Cru-3B-S1, Test 4, 367 kN, +38.6%
Cru-3B-S1, T4, 367 kN, +38.6%
350

300

Applied Load (kN)


250

200

150

100

50

0
0 1 2 3 4 5 6
Vertical Displacement (mm)

1.2
Cru-3B-S1, Test 4, 367 kN, +38.6%
Cru-3B-S1, T4, 367 kN, +38.6%

1.0
N1
N1*
0.8 N2
N2*
N3
N/Py

0.6 N3*

0.4

0.2

0.0
0.0 0.2 0.4 0.6 0.8 1.0
P/Pc

0.40
Cru-3B-S1, Test 4, 367 kN, +38.6%
Cru-3B-S1, T4, 367 kN, +38.6%
0.35

0.30
V1
V2
0.25
V3
V/Py

0.20

0.15

0.10

0.05

0.00
0.0 0.2 0.4 0.6 0.8 1.0
P/Pc

212
Ali-2B-S1, Test 1
400
Ali-2B-S1, Test 1, 343 kN, +29.4%
Ali-2B-S1, T1, 343 kN, +29.4%
350

300

Applied Load (kN)


250

200

150

100

50

0
0 1 2 3 4 5 6
Vertical Displacement (mm)

1.2
Ali-2B-S1, Test 1, 343 kN, +29.4%
Ali-2B-S1, T1, 343 kN, +29.4%

1.0

N1
N1*
0.8 N2
N2*
N3
N/Py

0.6 N3*

0.4

0.2

0.0
0.0 0.2 0.4 0.6 0.8 1.0
P/Pc

0.40
Ali-2B-S1, Test 1, 343 kN, +29.4%
Ali-2B-S1, T1, 343 kN, +29.4%
0.35

0.30 V1
V2
0.25 V3
V/Py

0.20

0.15

0.10

0.05

0.00
0.0 0.2 0.4 0.6 0.8 1.0
P/Pc

213
Ali-2B-S1, Test 2
400
Ali-2B-S1, Test 2, 311 kN, +17.2%
Ali-2B-S1, T2, 311 kN, +17.2%
350

300

Applied Load (kN)


250

200

150

100

50

0
0 1 2 3 4 5 6
Vertical Displacement (mm)

1.2
Ali-2B-S1, Test 2, 311 kN, +17.2%
Ali-2B-S1, T2, 311 kN, +17.2%

1.0
N1
N1*
0.8 N2
N2*
N3
N/Py

0.6 N3*

0.4

0.2

0.0
0.0 0.2 0.4 0.6 0.8 1.0
P/Pc

0.40
Ali-2B-S1, Test 2, 311 kN, +17.2%
Ali-2B-S1, T2, 311 kN, +17.2%
0.35

0.30

0.25 V1
V2
V3
V/Py

0.20

0.15

0.10

0.05

0.00
0.0 0.2 0.4 0.6 0.8 1.0
P/Pc

214
Ali-3B-S1, Test 1
400
Ali-3B-S1, Test 1, 312 kN, +17.7%
Ali-3B-S1, T1, 312 kN, +17.7%
350

300

Applied Load (kN)


250

200

150

100

50

0
0 1 2 3 4 5 6
Vertical Displacement (mm)

1.2
Ali-3B-S1, Test 1, 312 kN, +17.7%

Ali-3B-S1, T1, 312 kN, +17.7%

1.0
N1
N1*
0.8 N2
N2*
N3
N/Py

0.6 N3*

0.4

0.2

0.0
0.0 0.2 0.4 0.6 0.8 1.0
P/Pc

0.40
Ali-3B-S1, Test 1, 312 kN, +17.7%
Ali-3B-S1, T1, 312 kN, +17.7%
0.35

0.30

V1
0.25 V2
V3
V/Py

0.20

0.15

0.10

0.05

0.00
0.0 0.2 0.4 0.6 0.8 1.0
P/Pc

215
Ali-3B-S1, Test 2
400
Ali-3B-S1, Test 2, 339 kN, +28.0%
Ali-3B-S1, T2, 339 kN, +28.0%
350

300

Applied Load (kN)


250

200

150

100

50

0
0 1 2 3 4 5 6
Vertical Displacement (mm)

1.2
Ali-3B-S1, Test 2, 339 kN, +28.0%
Ali-3B-S1, T2, 339 kN, +28.0%

N1
1.0
N1*
N2
N2*
0.8
N3
N3*
N/Py

0.6

0.4

0.2

0.0
0.0 0.2 0.4 0.6 0.8 1.0
P/Pc

0.40
Ali-3B-S1, Test 2, 339 kN, +28.0%
Ali-3B-S1, T2, 339 kN, +28.0%

0.35

0.30 V1
V2
V3
0.25
V/Py

0.20

0.15

0.10

0.05

0.00
0.0 0.2 0.4 0.6 0.8 1.0
P/Pc

216
Alt-2B-S1, Test 1
400
Alt-2B-S1, Test 1, 320 kN, +20.6%
Alt-2B-S1, T1, 320 kN, +20.6%
350

300

Applied Load (kN)


250

200

150

100

50

0
0 1 2 3 4 5 6
Vertical Displacement (mm)

1.2
Alt-2B-S1, Test 1, 320 kN, +20.6%
Alt-2B-S1, T1, 320 kN, +20.6%

1.0 N1
N1*
N2
0.8 N2*
N3
N3*
N/Py

0.6

0.4

0.2

0.0
0.0 0.2 0.4 0.6 0.8 1.0
P/Pc

0.40
Alt-2B-S1, T1, 320 kN, +20.6%
0.35

0.30 V1
V2
0.25
V3
V/Py

0.20

0.15

0.10

0.05

0.00
0.0 0.2 0.4 0.6 0.8 1.0
P/Pc

217
Alt-2B-S1, Test 2
400
Alt-2B-S1, Test 2, 317 kN, +19.6%
Alt-2B-S1, T2, 317 kN, +19.6%
350

300

Applied Load (kN)


250

200

150

100

50

0
0 1 2 3 4 5 6
Vertical Displacement (mm)

1.2
Alt-2B-S1, Test 2, 317 kN, +19.6%
Alt-2B-S1, T2, 317 kN, +19.6%

1.0

N1
0.8
N1*
N2
N2*
N/Py

0.6
N3
N3*

0.4

0.2

0.0
0.0 0.2 0.4 0.6 0.8 1.0
P/Pc

0.40
Alt-2B-S1, Test 2, 317 kN, +19.6%
Alt-2B-S1, T2, 317 kN, +19.6%
0.35

0.30

0.25
V1
V2
V3
V/Py

0.20

0.15

0.10

0.05

0.00
0.0 0.2 0.4 0.6 0.8 1.0
P/Pc

218
Alt-3B-S1, Test 1
400
Alt-3B-S1, Test 1, 347 kN, +30.9%
Alt-3B-S1, T1, 347 kN, +30.9%
350

300

Applied Load (kN)


250

200

150

100

50

0
0 1 2 3 4 5 6
Vertical Displacement (mm)

1.2
Alt-3B-S1, Test 1, 347 kN, +30.9%
Alt-3B-S1, T1, 347 kN, +30.9%

1.0
N1
N1*
0.8 N2
N2*
N3
N/Py

0.6 N3*

0.4

0.2

0.0
0.0 0.2 0.4 0.6 0.8 1.0
P/Pc

0.40
Alt-3B-S1, Test 1, 347 kN, +30.9%
Alt-3B-S1, T1, 347 kN, +30.9%
0.35

0.30
V1
V2
0.25 V3
V/Py

0.20

0.15

0.10

0.05

0.00
0.0 0.2 0.4 0.6 0.8 1.0
P/Pc

219
Alt-3B-S1, Test 2
400
Alt-3B-S1, Test 2, 342 kN, +29.1%
Alt-3B-S1, T2, 342 kN, +29.1%
350

300

Applied Load (kN)


250

200

150

100

50

0
0 1 2 3 4 5 6
Vertical Displacement (mm)

1.2
Alt-3B-S1, Test 2, 342 kN, +29.1%
Alt-3B-S1, T2, 342 kN, +29.1%

1.0
N1
N1*
0.8 N2
N2*
N3
N/Py

0.6 N3*

0.4

0.2

0.0
0.0 0.2 0.4 0.6 0.8 1.0
P/Pc

0.40
Alt-3B-S1, Test 2, 342 kN, +29.1%
Alt-3B-S1, T2, 342 kN, +29.1%
0.35

0.30

V1
0.25
V2
V3
V/Py

0.20

0.15

0.10

0.05

0.00
0.0 0.2 0.4 0.6 0.8 1.0
P/Pc

220
Alt-2B-S2, Test 1
400
Alt-2B-S2, Test 1, 356 kN, +21.9%
Alt-2B-S2, T1, 356 kN, +21.9%

350

300

Applied Load (kN)


250

200

150

100

50

0
0 1 2 3 4 5 6
Vertical Displacement (mm)

1.2
Alt-2B-S2, Test 1, 356 kN, +21.9%
Alt-2B-S2, T1, 356 kN, +21.9%

1.0

N1
0.8 N1*
N2
N2*
N/Py

0.6 N3
N3*

0.4

0.2

0.0
0.0 0.2 0.4 0.6 0.8 1.0
P/Pc

0.40
Alt-2B-S2, Test 1, 356 kN, +21.9%
Alt-2B-S2, T1, 356 kN, +21.9%
0.35

0.30

V1
0.25 V2
V3
V/Py

0.20

0.15

0.10

0.05

0.00
0.0 0.2 0.4 0.6 0.8 1.0
P/Pc

221
Alt-2B-S2, Test 2
400
Alt-2B-S2, Test 2, 366 kN, +25.3%
Alt-2B-S2, T2, 366 kN, +25.3%

350

300

Applied Load (kN)


250

200

150

100

50

0
0 1 2 3 4 5 6
Vertical Displacement (mm)

1.2
Alt-2B-S2, Test 2, 366 kN, +25.3%
Alt-2B-S2, T2, 366 kN, +25.3%

1.0

N1
0.8 N1*
N2
N2*
N/Py

0.6 N3
N3*

0.4

0.2

0.0
0.0 0.2 0.4 0.6 0.8 1.0
P/Pc

0.40
Alt-2B-S2, Test 2, 366 kN, +25.3%
Alt-2B-S2, T2, 366 kN, +25.3%
0.35

0.30

V1
0.25 V2
V3
V/Py

0.20

0.15

0.10

0.05

0.00
0.0 0.2 0.4 0.6 0.8 1.0
P/Pc

222
Alt-2B-S2, Test 3
400
Alt-2B-S2, Test
Alt-2B-S2, T3, 3,
368368kN,
kN, +26.0%
+26.0%

350

300

Applied Load (kN)


250

200

150

100

50

0
0 1 2 3 4 5 6
Vertical Displacement (mm)

1.2
Alt-2B-S2, Test 3, 368 kN, +26.0%
Alt-2B-S2, T3, 368 kN, +26.0%

1.0
N1
N1*
0.8 N2
N2*
N3
N/Py

0.6 N3*

0.4

0.2

0.0
0.0 0.2 0.4 0.6 0.8 1.0
P/Pc

0.40
Alt-2B-S2, Test 3, 368 kN, +26.0%
Alt-2B-S2, T3, 368 kN, +26.0%
0.35

0.30
V1
V2
0.25
V3
V/Py

0.20

0.15

0.10

0.05

0.00
0.0 0.2 0.4 0.6 0.8 1.0
P/Pc

223
Alt-W-S2, Test 1
450
Alt-W-S2, Test 1, 405 kN, +38.7%
Alt-W-S2, T1, 405 kN, +38.7%
400

350

300

Applied Load (kN)


250

200

150

100

50

0
0 1 2 3 4 5 6
Vertical Displacement (mm)

1.2
Alt-W-S2, Test 1, 405 kN, +38.7%
Alt-W-S2, T1, 405 kN, +38.7%

1.0

N1
0.8 N1*
N2
N2*
N/Py

0.6 N3
N3*

0.4

0.2

0.0
0.0 0.2 0.4 0.6 0.8 1.0
P/Pc

0.40
Alt-W-S2, Test 1, 405 kN, +38.7%
Alt-W-S2, T1, 405 kN, +38.7%
0.35

0.30

V1
0.25 V2
V3
V/Py

0.20

0.15

0.10

0.05

0.00
0.0 0.2 0.4 0.6 0.8 1.0
P/Pc

224
Alt-W-S2, Test 2
450
Alt-W-S2,
Alt-W-S2, Test T2, 417
2, 417 kN,kN, +42.8%
+42.8%

400

350

300

Applied Load (kN)


250

200

150

100

50

0
0 1 2 3 4 5 6
Vertical Displacement (mm)

1.2
Alt-W-S2, T2, 417 kN, +42.8%

1.0
N1
N1*
0.8 N2
N2*
N3
N/Py

0.6 N3*

0.4

0.2

0.0
0.0 0.2 0.4 0.6 0.8 1.0
P/Pc

0.40
Alt-W-S2, T2, 417 kN, +42.8%
0.35

0.30
V1
V2
0.25
V3
V/Py

0.20

0.15

0.10

0.05

0.00
0.0 0.2 0.4 0.6 0.8 1.0
P/Pc

225
Cru-2B-S2, Test 2
400
Cru-2B-S2, Test 2, 388 kN, +32.9%

350

300

Applied Load (kN)


250

200

150
Cru-2B-S2, T2, 388 kN, +32.9%

100

50

0
0 1 2 3 4 5 6
Vertical Displacement (mm)

1.2
Cru-2B-S2, T2, 388 kN, +32.9%

1.0
N1
N1*
N2
0.8
N2*
N3
N3*
N/Py

0.6

0.4

0.2

0.0
0.0 0.2 0.4 0.6 0.8 1.0
P/Pc

0.40
Cru-2B-S2, T2, 388 kN, +32.9%
0.35

0.30
V1
V2
0.25
V3
V/Py

0.20

0.15

0.10

0.05

0.00
0.0 0.2 0.4 0.6 0.8 1.0

P/Pc

226
Cru-2B-S2, Test 3
400
Cru-2B-S2, Test 3, 389 kN, +33.2%

350

300

Applied Load (kN)


250

200

Cru-2B-S2, T3, 389 kN, +33.2%


150

100

50

0
0 1 2 3 4 5 6
Vertical Displacement (mm)

1.2
Cru-2B-S2, Test 3, 389 kN, +33.2%
Cru-2B-S2, T3, 389 kN, +33.2%

1.0

N1
0.8
N1*
N2
N2*
N/Py

0.6
N3
N3*

0.4

0.2

0.0
0.0 0.2 0.4 0.6 0.8 1.0
P/Pc

0.40
Cru-2B-S2, Test 3, 389 kN, +33.2%
Cru-2B-S2, T3, 389 kN, +33.2%
0.35

0.30

V1
0.25
V2
V3
V/Py

0.20

0.15

0.10

0.05

0.00
0.0 0.2 0.4 0.6 0.8 1.0
P/Pc

227
B. Member selections for multi-panel tower models
B.1 Dimensions and properties of 75 x 75 x 6 equal angle

B.2 Dimensions and properties of round-hollowed sections used in the


selection of tension legs and bracing members
D t d A I r
2 4
(mm) (mm) (mm) (mm ) (mm ) (mm)
21.3 3.2 14.9 182 7684 6.50
26.9 3.2 20.5 238 17033 8.46
33.7 2.6 28.5 254 30927 11.03
3.2 27.3 307 36047 10.84
4.0 25.7 373 41898 10.60
42.4 2.6 37.2 325 64645 14.10
3.2 36 394 76200 13.91
4.0 34.4 483 89908 13.65
48.3 3.2 41.9 453 115857 15.99
4.0 40.3 557 137676 15.73
5.0 38.3 680 161527 15.41
60.3 3.2 53.9 574 234682 20.22
4.0 52.3 707 281729 19.96
5.0 50.3 869 334766 19.63

D = Exterior diameter
t = Thickness
d = Interior diameter
228
B.3 Forces and adjusted forces on the legs, Panels 1 to 10
Cross S80 model

Cross S80 Model


Panel L Panel Force % of Capacity Increase Demand 0.9Nc of
No. Unsupported from 0 50 60 70 100 75x75x6EA
Length Assumed Adjusted or Expected Panel Force fy=320MPa
(m) Loads (kN) (kN) (kN)
1 1.20 172.33 182.3 273.5 291.7 309.9 364.6 182.3
2 1.20 163.58 173.0 259.6 276.9 294.2 346.1 182.3
3 1.20 153.48 162.4 243.5 259.8 276.0 324.7 182.3
4 1.20 141.35 149.5 224.3 239.2 254.2 299.1 182.3
5 1.20 126.25 133.6 200.3 213.7 227.0 267.1 182.3
6 1.20 107.26 113.5 170.2 181.5 192.9 226.9 182.3
7 1.04 84.38 89.3 133.9 142.8 151.7 178.5 199.0
8 1.04 56.43 59.7 89.5 95.5 101.5 119.4 199.0
9 0.96 20.10 21.3 31.9 34.0 36.1 42.5 206.5
10 1.31 -87.3 -92.4 -138.5 -147.8 -157.0 -184.7 -

Cross S100 model

Cross S100 Model


Panel L Panel Force % of Capacity Increase Demand 0.9Nc of
No. Unsupported from 0 50 60 70 100 75x75x6EA
Length Assumed Adjusted or Expected Panel Force fy=320MPa
(m) Loads (kN) (kN) (kN)
1 1.50 172.33 144.3 216.5 230.9 245.3 288.6 144.3
2 1.50 163.58 137.0 205.5 219.2 232.9 273.9 144.3
3 1.50 153.48 128.5 192.8 205.6 218.5 257.0 144.3
4 1.50 141.35 118.4 177.5 189.4 201.2 236.7 144.3
5 1.50 126.25 105.7 158.6 169.1 179.7 211.4 144.3
6 1.50 107.26 89.8 134.7 143.7 152.7 179.6 144.3
7 1.30 84.38 70.7 106.0 113.0 120.1 141.3 170.6
8 1.30 56.43 47.3 70.9 75.6 80.3 94.5 170.6
9 1.20 20.10 16.8 25.2 26.9 28.6 33.7 182.3
10 1.64 -87.3 -73.1 -109.7 -117.0 -124.3 -146.2 -

229
Cross S120 model

Cross S120 Model


Panel L Panel Force % of Capacity Increase Demand 0.9Nc of
No. Unsupported from 0 50 60 70 100 75x75x6EA
Length Assumed Adjusted or Expected Panel Force fy=320MPa
(m) Loads (kN) (kN) (kN)
1 1.80 172.33 102.7 154.1 164.3 174.6 205.4 102.7
2 1.80 163.58 97.5 146.2 156.0 165.7 195.0 102.7
3 1.80 153.48 91.5 137.2 146.3 155.5 182.9 102.7
4 1.80 141.35 84.2 126.4 134.8 143.2 168.5 102.7
5 1.80 126.25 75.2 112.9 120.4 127.9 150.5 102.7
6 1.80 107.26 63.9 95.9 102.3 108.7 127.8 102.7
7 1.56 84.38 50.3 75.4 80.5 85.5 100.6 135.7
8 1.56 56.43 33.6 50.4 53.8 57.2 67.3 135.7
9 1.44 20.10 12.0 18.0 19.2 20.4 24.0 152.6
10 1.97 -87.3 -52.0 -78.0 -83.2 -88.4 -104.1 -

Warren S80 model

Warren S80 Model


Panel L Panel Force % of Capacity Increase Demand 0.9Nc of
No. Unsupported from 0 30 40 50 100 75x75x6EA
Length Assumed Adjusted or Expected Panel Force fy=320MPa
(m) Loads (kN) (kN) (kN)
1 1.20 168.14 182.3 237.0 255.2 273.5 364.6 182.3
2 1.20 168.14 182.3 237.0 255.2 273.5 364.6 182.3
3 1.20 148.02 160.5 208.6 224.7 240.7 321.0 182.3
4 1.20 148.02 160.5 208.6 224.7 240.7 321.0 182.3
5 1.20 117.93 127.9 166.2 179.0 191.8 255.7 182.3
6 1.20 117.93 127.9 166.2 179.0 191.8 255.7 182.3
7 1.04 72.36 78.5 102.0 109.8 117.7 156.9 199.0
8 1.04 72.36 78.5 102.0 109.8 117.7 156.9 199.0
9 0.96 17.34 18.8 24.4 26.3 28.2 37.6 206.5
10 1.31 -89.24 -96.8 -125.8 -135.5 -145.1 -193.5 -

230
B.4 Selection of round-hollowed sections for tension legs and bracing
members based on ANSI/AISC 360-05 formulae
Cross S80 model

Member Category Member Nc* Nt* L' (L/r)lim Round-Hallowed L'/r 0.9Nc 0.9Nt
No. (kN) (kN) (m) Section (kN) (kN)
Tension Legs 10 -184.7 1.31 150 48.3 x 5.0 85.0 -195.9
28 8.6 1.25 200 26.9 x 3.2 147.8 17.0
30 9.4 1.22 " 26.9 x 3.2 144.3 17.8
32 10.6 1.15 " 26.9 x 3.2 136.0 20.1
34 12.0 1.07 " 26.9 x 3.2 126.5 23.2
Bracing Members on the 36 14.0 1.00 " 26.9 x 3.2 118.3 26.6
Two Main Faces 38 16.6 0.94 " 26.9 x 3.2 111.2 29.7
40 18.9 0.81 " 26.9 x 3.2 95.8 36.8
42 23.0 0.76 " 26.9 x 3.2 89.9 39.7
44 28.2 0.68 " 26.9 x 3.2 80.4 44.2
46 -21.2 0.80 " 21.3 x 3.2 123.1 -52.4
Horizontals of K-Bracings 66, 67 1.50 = L3/2 250 21.3 x 3.2 230.8 5.3
Diagonals of K-Bracings 84, 85 2.10 = L 250 26.9 x 3.2 248.4 6.0
Ties 1.50 = L 250 21.3 x 3.2 230.8 5.3

Cross S100 model

Member Category Member Nc* Nt* L' (L/r)lim Round-Hallowed L'/r 0.9Nc 0.9Nt
No. (kN) (kN) (m) Section (kN) (kN)
Tension Legs 10 -146.2 1.64 150 48.3 x 4.0 104.3 -160.3
28 6.8 1.56 200 26.9 x 3.2 184.5 10.9
30 7.5 1.53 " 26.9 x 3.2 181.0 11.3
32 8.4 1.43 " 26.9 x 3.2 169.1 13.0
34 9.5 1.34 " 26.9 x 3.2 158.5 14.8
Bracing Members on the 36 11.1 1.25 " 26.9 x 3.2 147.8 17.0
Two Main Faces 38 13.1 1.17 " 26.9 x 3.2 138.4 19.4
40 15.0 1.02 " 26.9 x 3.2 120.6 25.5
42 18.2 0.95 " 26.9 x 3.2 112.4 29.1
44 22.3 0.85 " 26.9 x 3.2 100.5 34.6
46 -16.8 1.00 " 21.3 x 3.2 153.9 -52.4
Horizontals of K-Bracings 66, 67 1.88 = L3/2 250 26.9 x 3.2 222.3 7.5
Diagonals of K-Bracings 84, 85 2.62 = L 250 33.7 x 2.6 237.5 7.0
Ties 1.88 = L 250 26.9 x 3.2 222.3 7.5

231
Cross S120 model

Member Category Member Nc* Nt* L' (L/r)lim Round-Hallowed L'/r 0.9Nc 0.9Nt
No. (kN) (kN) (m) Section (kN) (kN)
Tension Legs 10 -104.1 1.97 150 42.4 x 3.2 141.7 -113.5
28 4.8 1.87 200 33.7 x 2.6 169.5 13.8
30 5.3 1.83 " 33.7 x 2.6 165.9 14.4
32 6.0 1.72 " 33.7 x 2.6 155.9 16.3
34 6.8 1.61 " 26.9 x 3.2 190.4 10.2
Bracing Members on the 36 7.9 1.51 " 26.9 x 3.2 178.6 11.6
Two Main Faces 38 9.3 1.41 " 26.9 x 3.2 166.8 13.3
40 10.7 1.22 " 26.9 x 3.2 144.3 17.8
42 13.0 1.14 " 26.9 x 3.2 175.4 20.4
44 15.9 1.02 " 26.9 x 3.2 120.6 25.5
46 -12.0 1.20 " 21.3 x 3.2 184.7 -52.4
Horizontals of K-Bracings 66, 67 2.26 = L3/2 250 33.7 x 2.6 204.8 9.4
Diagonals of K-Bracings 84, 85 3.14 = L 250 42.4 x 2.6 222.7 10.2
Ties 2.26 = L 250 33.7 x 2.6 204.8 9.4

Warren S80 model


Member Category Member Nc* Nt* L' (L/r)lim Round-Hallowed L'/r 0.9Nc 0.9Nt
No. (kN) (kN) (m) Section (kN) (kN)
Tension Legs 10 -193.5 1.31 150 48.3 x 5.0 85.0 -195.9
28 17.5 2.49 200 42.4 x 3.2 179.1 19.1
30 21.7 2.18 " 42.4 x 3.2 156.8 25.0
Diagonals of Warren-Bracings 32 28.7 1.89 " 42.4 x 3.2 135.9 33.2
34 38.6 1.52 " 42.4 x 2.6 107.8 42.6
36 32.8 0.97 " 33.7 x 2.6 87.9 43.3
38 2.13 250 26.9 x 3.2 251.9 5.8
Horizontals of Warren-Bracings
45 15.8 0.94 200 26.9 x 3.2 111.2 29.7
Horizontals of K-Bracings 66, 67 1.50 = L3/2 250 21.3 x 3.2 230.8 5.3
Diagonals of K-Bracings 84, 85 2.10 = L 250 26.9 x 3.2 248.4 6.0
Ties 1.51 = L 250 21.3 x 3.2 230.8 5.3

B.5 Formulae used in determining the compressive strengths of round-


hollowed section bracing members (from ANSI/AISC 360-05)

The nominal compressive strength, N c , determined base on the limit state of flexural

buckling stress, f cr , is:

N c = f cr Ag Eq. B.1

When f e 0.44 f y

fy

f cr = 0.658 f e f y Eq. B.2

and when f e < 0.44 f y

232
f cr = 0.877 f e Eq. B.3

2E
where fe = 2
Eq. B.4
KL

r

233

You might also like