Finite Element Analysis of The Design and Manufacture of Thin-Walled Pressure Vessels Used As Aerosol Cans

Swansea University E-Theses
_________________________________________________________________________
Finite element analysis of the design and manufacture of thin-walled

pressure vessels used as aerosol cans.
Abdussalam, Ragba Mohamed
How to cite:
_________________________________________________________________________
Abdussalam, Ragba Mohamed (2006) Finite element analysis of the design and manufacture of thin-walled pressure
vessels used as aerosol cans.. thesis, Swansea University.
http://cronfa.swan.ac.uk/Record/cronfa42323
Use policy:
_________________________________________________________________________
This item is brought to you by Swansea University. Any person downloading material is agreeing to abide by the terms
of the repository licence: copies of full text items may be used or reproduced in any format or medium, without prior
permission for personal research or study, educational or non-commercial purposes only. The copyright for any work
remains with the original author unless otherwise specified. The full-text must not be sold in any format or medium
without the formal permission of the copyright holder. Permission for multiple reproductions should be obtained from
the original author.
Authors are personally responsible for adhering to copyright and publisher restrictions when uploading content to the
repository.
Please link to the metadata record in the Swansea University repository, Cronfa (link given in the citation reference
above.)
http://www.swansea.ac.uk/library/researchsupport/ris-support/
University of W ales, Swansea
Finite Elem ent Analysis of the Design and M anufacture of
Thin-walled Pressure Vessels used as Aerosol Cans
By
Ragba M ohamed Abdussalam
B. Sc Eng (honours) M. Sc Eng
Thesis submitted to the University of Wales Swansea
For the degree of Doctor of Philosophy, September 2006
Civil and C om putational Engineering C entre
School of Engineering
ProQuest Number: 10798031
All rights reserved
INFORMATION TO ALL USERS

The quality of this reproduction is d e p e n d e n t upon the quality of the copy subm itted.
In the unlikely e v e n t that the a u thor did not send a c o m p le te m anuscript

and there are missing pages, these will be noted. Also, if m aterial had to be rem oved,
a n o te will ind ica te the deletion.
uest
ProQuest 10798031
Published by ProQuest LLC(2018). C opyright of the Dissertation is held by the Author.
All rights reserved.

This work is protected against unauthorized copying under Title 17, United States C o d e
M icroform Edition © ProQuest LLC.
ProQuest LLC.
789 East Eisenhower Parkway
P.O. Box 1346
Ann Arbor, Ml 4 8 1 0 6 - 1346
^ <
LIBRARY
SUMMARY
Thin-walled cylinders are used extensively in the food packaging and cosmetics
industries. The cost o f material is a major contributor to the overall cost and so
improvements in design and manufacturing processes are always being sought.
Shape optimisation provides one method for such improvements.
Aluminium aerosol cans are a particular form o f thin-walled cylinder with a complex
shape consisting o f truncated cone top, parallel cylindrical section and inverted dome
base. They are manufactured in one piece by a reverse-extrusion process, which
produces a vessel with a variable thickness from 0.31 mm in the cylinder up to 1.31
mm in the base for a 53 mm diameter can. During manufacture, packaging and
charging, they are subjected to pressure, axial and radial loads and design
calculations are generally outside the British and American pressure vessel codes.
‘Design-by-test’ appears to be the favoured approach. However, a more rigorous
approach is needed in order to optimise the designs.
Finite element analysis (FEA) is a powerful tool for predicting stress, strain and
displacement behaviour o f components and structures. FEA is also used extensively
to model manufacturing processes. In this study, elastic and elastic-plastic FEA has
been used to develop a thorough understanding o f the mechanisms o f yielding,
‘dome reversal’ (an inherent safety feature, where the base suffers elastic-plastic
buckling at a pressure below the burst pressure) and collapse due to internal pressure
loading and how these are affected by geometry. It has also been used to study the
buckling behaviour under compressive axial loading. Furthermore, numerical
simulations o f the extrusion process (in order to investigate the effects o f tool
geometry, friction coefficient and boundary conditions) have been undertaken.
Experimental verification o f the buckling and collapse behaviours has also been
carried out and there is reasonable agreement between the experimental data and the
numerical predictions.
Declaration
This work has not previously been accepted in substance for any degree and is not
being concurrently submitted in candidature for any degree.
Signature: .
(Candidate)
D a te : ........ / . . 2 . . r ....... J..Z ....... *.........
Statement 1
This thesis is the result o f my own investigation, except where otherwise stated.
Where correction services have been used the extent and nature o f the correction is
clearly marked in footnote(s). Other sources are acknowledged by footnotes giving
explicit references. A bibliography is appended.
Signature: .
(Candidate)
D a te :............... L .^ ,
Statement 2
I hereby give a full consent for m y thesis, if accepted, to be available for
photocopying and for the title and summary to be made available to outside
organisations.
Signature: .
(Candidate)
D a te :.........
ii
ACKNOWLEDGEMENT
I would like to express my sincere appreciation and gratitude to my advisor Dr Steve.

J. Hardy for his support, guidance, encouragement and his understanding during my
research years and I would like to thanks Professor David G.T. Gethin. For his
professional guidance advise. Many thanks to Dr. M.K. Pipelzadeh for his technical
support in my experimental work; I also want to thanks all the technicians for their
assistance and patient during the experimental work.
I feel a deep sense o f gratitude for my late parent who formed part o f my vision and
taught me the good things that really matter in life. The happy memory o f my parent
still provides a persistent inspiration for my journey in this life. I am grateful to my
brother and sisters for all the opportunities they have given me along with their
loving support without them, I would not have achieved my success today.
Finally I would like to give a special thanks to my husband for his encouragement,
patient and moral support.
In my opinion, doing a PhD is a holy task and this was definitely one o f the best
decisions o f my life, it takes a lot o f effort and time but it’s worth it.
CONTENTS
Page
Summary ............................................................................................................................... i
Declaration ...........................................................................................................................ii
Acknowledgments.............................................................................................................. iii
List o f contents.................................................................................................................... iv
List of Table ...................................................................................................................... xii
List of Figures................................................................................................................... xiii
Nomenclature ........................................................................ xix
Chapter 1 INTRODUCTION
1.1. Background.............................................................................................................. 1
1.2. Thin walled pressure vessel...................................................................................2
1.3. Aims Objectives o f the p ro je c t............................................................................ 5
1.4. Structure o f th e sis ..................................................................................................7
Chapter 2: GENERAL BACKGROUND AND LITERATURE
REVIEW
2.1 Introduction..............................................................................................................9
2.2 The design and manufacture o f aerosol cans.....................................................10
2.2.1 Top and valves..........................................................................................11
2.2.2 Main b o d y ................................................................................................ 13
2.2.3 Base............................................................................................................ 14
2.2.4 Principles o f operation.............................................................................15
2.2.5 D esign........................................................................................................17
2.2.5.1 Internal pressure............................................................................ 18
2.2.5.2 Axel loading.................................................................................. 18
2.2.5.3 Dome reversal o f the inverted base............................................ 20
2.2.6 M anufacturer........................................................................................... 20
2.3 Basic concepts o f elasticity and plasticity..........................................................21
2.3.1 E lasticity........................................................-..........................................21
2.3.2 Plasticity.................................................................................................... 23
2.3.2.1 Yield criterion................................................................................25
2.3.2.2 Flow rule (normality principl).....................................................27
2.3.2.3 Material hardening m odels.......................................................... 27
2.4 Overview o f non-linear finite element analysis................................................. 29
2.4.1 Explicit and implicit m ethods................................................................. 30
2.5 Thin cylinder formulae......................................................................................... 32

2.6 BS5500................................................................................................................... 33
2.7 ASME V III ................................. 34
2.8 Buckling................................................................................................................. 35
2.8.1 Bifurcation buckling.................................................................................37
2.8.2 Pre-buckling deformation........................................................................ 38
2.8.3 Post buckling deformation.......................................................................38
2.8.4 Eigenvalue analysis................................................................................... 39
2.8.5 R ik's method for modelling snap through.............................................40
2.8.6 Buckling o f thin-wall tubes......................................................................40
2.9 Upper and lower bound analysis.......................................................................... 42
2.9.1 Elastic compensation method................................................................. 43
2.9.2 Application to lower bound limit load................................................... 45
2.9.3 Application to upper bound limit load................................................... 45
2.10 Extrusion process and modelling............................................................ 49
2.10.1 Process description.................................................................................... 49
2.10.2 Constant volume analytical approach.....................................................53
2.10.3 Finite element modelling......................................................................... 54
2.10.4 Friction considerations............................................................................ 55
2.10.5 Unloading (spring back).......................................................................... 57
2.10.6 Other issu e s............................................................................................... 59
2.10.6.1 Effects o f punch speed.................................................................. 59
2.10.6.2 Thermal effects.............................................................................. 59
2.11 O ptim isation..............................................................................................60
2.11.1 Previously published w ork...................................................................... 62
2.12 C losure...................................................................................................... 63
Chapter 3 I N T E R N A L P R E S S U R E L O A D I N G
3.1 Introduction............................................................................................................65
3.2 Elastic finite element analysis using axisymmetric models............................... 66
3.2.1 Geometry and finite element models...................................................... 66
3.2.2 Loading and boundary conditions..........................................................68
3.2.3 Materials m odels..................................................................................... 69
3.2.4 Constant thickness models...................................................................... 69
3.2.4.1 Results for geometry G4 (t = 1 m m ).........................................71
3.2.4.2 Effects o f wall thickness............................................................ 81
3.2.4.3 Limiting pressures....................................................................... 89
3.2.5 Can with varying thickness.................................................................... 91
3.2.5.1 Results 92
3.3 Elastic-plastic finite element analysis using axisymmetric m o d e l................ 95
3.3.1 Constant thickness model....................................................................... 95
3.3.2 Material m o d e l.........................................................................................96
3.3.3 Finite element results for geometry G 4................................................ 98
3.3.3.1 Elastic perfectly-plastic model...................................................99
3.3.3.2 Multi-linear work-hardening model......................................... 103
3.3.4 Effects o f wall thickness.................................................................... 105
3.3.5 Can with varying thickness................................................................... 106
3.4 Elastic-plastic finite element analysis using 3D m odels............................... 109
3.4.1 Finite element m odel.............................................................................. 109
3.4.2 Material model and loading.................................................................. I l l
3.4.3 Eigenvalue analysis.................................................................................I l l
3.4.4 Results...................................................................................................... 112
3.5 Upper and lower bound pressures................................................................... 115
3.5.1 Material models, loading and boundary conditions........................... 116
3.5.2 Constant thickness m odel......................................................................116
3.5.2.1 Geometry G4 (t = 1 m m )...........................................................117
3.5.2.2 Method o f implementation o f elastic compensation method
.......................................................................................................119
3.5.2.3 Effects o f wall thickness............................................................122
vi
3.5.3 Can with varying thickness....................................................................124
3.6 Experimental testing............................................................................................ 125
3.7 Closure...................................................................................................................127
Chapter 4: A X IA L L O A D I N G
4.1 Introduction........................................................................................................... 129

4.2 Potential failure modes........................................................................................130
4.3 Case (a) Compressive behaviour during neck formation................................ 131
4.3.1 Geometry and finite element model..................................................... 131
4.3.2 Loading and boundary conditions.........................................................132
4.3.3 Material model........................................................................................ 132
4.3.4 Finite element predictions.....................................................................133
4.3.5 Experimental testing...............................................................................142
4.3.6 Analytical solution..................................................................................149
4.3.7 Comparison and discussion o f re su lts................................................ 150
4.3.7.1 Load-displacement characteristics............................... 150
4.3.7.2 Buckling mode s h a p e ...................................................150
4.4 Case (b)compressive behaviour during valve insertion and charging .... 151
4.4.1 Geometry and finite element model......................................................151
4.4.2 Loading and boundary conditions........................................................151
4.4.3 Material model....................................................................................... 152
4.4.4 Finite element predictions..................................................................... 157
4.4.5 Experimental te stin g ..............................................................................161
4.4.6 Analytical solution................................................................................. 162
4.4.7 Comparison and discussion o f results.................................................. 167
4.4.7.1 Load displacement characteristics............................... 167
4.4.7.2 Deformed sh a p e .............................................................167
4.5 C losure..................................................................................................................167
Chapter 5: MODELLING OF THE EXTRUSION PROCESS
5.1 Introduction.......................................................................................................... 168

5.2 Stage 1 modelling the base and side w a ll........................................................ 169
5.2.1 Geometry and finite element m odel........................................169
5.2.2 Loading and boundary conditions. ...................................... 170
5.2.3 Material models......................................................................... 172
5.2.4 Finite element predictions (p = 0.25)..................................... 174
5.2.5 Effect o f coefficient o f friction................................................ 181
5.2.6 Comparisons with analytical solution.....................................182
5.3 Stage 2 modelling................................................................................................ 182
5.3.1 Geometry and finite element models......................................182
5.3.2 Loading and boundary conditions........................................... 183
5.3.2.1 Stage 2(a) - base formation before decoration 183
5.3.2.2 Stage 2(b) - base formation after decoration............. 183
5.3.2.3 Stage 2 loading.............................................................186
5.3.3 Material models............................................. ;.......................... 188
5.3.4 Finite element predictions (p = 0.25)......................................189
5.3.4.1 Stage 2(a):' pre-decoration boundary conditions.......189
5.3.4.2 Stage 2(b) post-decoration boundary conditions.......194
5.3.5 Effect o f friction coefficient....................................................197
5.3.6 Comparisons with experimental data......................................197
5.4 Closer................................................................................................................. 198
Chapter 6: O PTIM ISA TIO N
6.1 Introduction........................................................................................................... 199

6.2 Simplistic approach.............................................................................................. 199
6.2.1 Geometry and finite element models...................................... 199
6.2.2 Loading and boundary conditions........................................... 200
6.2.3 Material models........................................................................ 200
6.2.4 Results for axisymmetric model............................................. 201
6.2.5 Elastic-plastic buckling results using a 3-D model...............203
6.3 Optimisation procedure....................................................................................... 208
6.3.1 Objective function and constraints.......................................... 208
6.3.2 DOT optimisation program...................................................... 210
6.3.3 Geometry and finite element model........................................ 211
6.3.4 Loading and boundary conditions................................................ 211
6.3.5 Material Model............................................................................212
6.3.6 Results........................................................................................212
6.4 C loser................................................................................................................215
Chapter 7: DISCUSSION
7.1 Introduction......................................................................................................... 216
7.2 Internal pressure loading (Chapter 3 ).............................................................. 218
7.2.1 Elastic analyses................................................................... 218
7.2.2 Elastic plastic analyses....................................................... 219
7.2.3 Upper and lower bound pressures.................................... 220
7.3 Axial loading (Chapter 4 )..................................................................................221
7.3.1 Axial loading during neck forming.................................. 221
7.3.2 Axial loading during valve insertion/charging...............222
7.4 Modelling o f the extrusion process (Chapter 5).............................................223
7.4.1 Stage 1 simulation..............................................................224
7.4.2 Stage 2 simulation..............................................................225
7.5 Optimisation studies(Chapter6)........................................................................225
7.5.1 Simplistic approach............................................................226
7.5.2 Structured approach using DOT..................... 226
7.6 C losure................................................................................................................227
Chapter 8: CONCLUSIONS AND RECOMMENDATIONS
8.1 Conclusions........................................................................................................228
8.2 Recommendations for further w ork................................................................230
References..................................................................................................................232
Appendix A Drawing o f extrusion tooling............................................................236
Appendix B Properties o f aluminium for impact extrusion................................238
Appendix C Fortran program..................................................................................240
LIST OF TABLLES
Page
3.1 Geometric parameter 67
3.2 Mechanical Properties o f 1050 Aluminium 69
3.3 The variation o f maximum equivalent stress index with wall thickness 87
3.4 Limiting pressures for constant thickness cans 90
3.5 Plastic stress-strain data for multi-linear material model 98
3.6 First yield and collapse pressures for different wall thickness 106
3.7 Upper and lower bound pressures using elastic compensation method 123
3.8 Results o f elastic compensation and finite element analyses 123
3.9 Comparison between measured and predicted burst pressure cans 126
4.1 Comparison o f actual failure load to buckling and compressive method 163
6.1 Thickness at each iteration 212
LIST OF FIGURE
Page
1.1 Typical aluminium aerosol can 5
2.1 Aerosol cans 10
2.2 Aerosol can valve 12
2.3 Aluminium aerosol can valve 13
2.4 Aerosol can bottom 14
2.5 Bottom forming process 15
2.6 Liquid and compressed propellant 17
2.7 Axial loading during valve insertion and filling 19
2.8 The buckling o f the can top 19
2.9 Axial loading during neck forming 20
2.10 Schematic diagram o f aerosol can production line 21
2.11 Stress-strain curve for a simple one-dimensional tension (or
compression) 24
2.12 Projection o f the von Mises yield surface onto the 7r-plane 26
2.13 Isotropic and kinematic hardening models 28
2.14 Hoop stress diagram 33
2.15 Load-deflection curves showing limit and bifurcation points 36
216 Variation in strain energy and dissipation energy with applied load,
used in the calculation of upper bound limit load 48
2.17 Direct and indirect extrusion 51
2.18 Extrusion load/displacement curves for direct and indirect extrusion 52
2.19 Tool setup used for the manufacturing o f aerosol cans 53
2.20 Effect o f friction on the thickness o f the sheet metal 56
2.21 Effect o f friction on the final forming time and pressure 57
3.1 Can base geometry (constant thickness) 67
3.2 Basic finite element model o f can base (constant thickness) 69
3.3(a) Finite element mesh for geometry G4 70
3.3(b) Mesh convergent 71
^ a] contour plot (G4, p = 0.1 MPa) ^
G l contour plot (G4, p = 0.1 MPa) ^
xi
73
0-3 contour plot (G4, p = 0.1 MPa)
3.7 Principal stress distributions around the inside surface (G4, p=0.1 75
MPa)
3.8 Principal stress distributions around the outside surface (G4, p=0.1
MPa) 76
3.9 Equivalent stress contour plot (G4, p = 0.1 MPa) 79
3.10 Equivalent stress distribution (G4, p = 0.1 MPa) 80
3.11 Equivalent stress contour plot (G l, p = O.IMPa) 82
3.15 Equivalent stress contour plot (G6, p = 0. IMPa) 86
3.16 The relationship between wall thickness and maximum equivalent 88
stress
3.17 The relationship between maximum elastic equivalent stress index
and the wall thickness 88
3.18 Maximum equivalent stresses, nominal stresses and maximum
equivalent stress indices versus D/t ratio 89
3.19 Variation of limiting pressure with wall thickness and D/t 90
3.20 Finite element model for can with varying thickness 91
3.21 Finite element mesh for can with varying thickness 92
3.22 Equivalent stress contour plot at internal pressure o f 0.1 MPa 93
3.23 Equivalent stress distribution around inside surface 94
3.24 Equivalent stress distribution around outside surface 94
3.25 Parallel spring plasticity Vs multi-linear model 96
3.26 Stress-Strain relationship for an elastic-perfectly-plastic material
model 97
3.27 Parallel-springs plasticity model 98
3.28 Von Mises stress contour plot for G4 with internal pressure = 1.0
MPa and an EPP material model 100
3.29 Von Mises stress contour plot for G4 with pressure = 1.2 MPa and an
EPP material model 100
3.30 Von Mises stress contour plot for G4 with internal pressure = 1.4
xii
Von Mises stress contour plot for G4 with internal pressure = 1 .6
Equivalent stress distribution around the inside surface for G4 and an
Equivalent stress distribution around the outside surface for G4 and
Equivalent stress contour plot (P = 1.50 MPa) for a multi-liner
hardening material mode 104
Equivalent stress contour plot (collapse, p = 1.59 MPa) 105
The relationship between wall thickness first yield and collapse
pressure 106
Equivalent stress contour plot (pre-buckling, pressure =1.50 MPa) 107
Equivalent stress contour plot (collapse, pressure =1.53 MPa) 108
3D finite element model geometry 110
Finite element constrains 110
3-D Finite element model mesh 111
Von Mises Stress Contour Plot at internal pressure o f 1.50 MPa 112
Von Mises stress contour plot at internal pressure o f 1.70 MPa 113
Final von Mises Stress Prediction at pressure o f 2.02 MPa 113
Von Mises stress contour plot at internal pressure o f 0.83MPa and 0.6
mm constant thickness 114
Von Mises stress contour plot at internal pressure o f 1.20 MPa and
1.0 mm constant thickness 115
Simple finite element meshes for geometry G4 117
Equivalent stress contour plot for iteration 0 118
Maximum equivalent stress at the end o f each iteration for t = 1 mm 119
Steady state equivalent stress contour plot for t = 1 mm 121
Comparison o f finite element method and compensation method 123
Deformation and burst pressure o f can base 126
Deformation and burst pressure 127
Axial compression loading on the can 130
Effect o f cylinder length on the buckling modes 131
xiii
4.3 Cross-section geometry for the analysis o f Case (a) axial loading
during neck forming 134
4.4 Three-dimensional model for Case (a) axial loading analysis 135
4.5 Structural constraints for Case (a) axial loading analysis 136
4.6 Applied loading for Case (a) axial loading analysis 137
4.7 Finite element mesh for Case (a) axial loading analysis 138
4.8 Pre-buckling equivalent stress counter plot for case(a) axial loading
analysis 139
4.9 Equivalent stress contour plot at the point o f buckling for Case (a)
axial loading analysis 140
4.10 Buckling mode shape for case (a) axial loading analysis 141
4.11 predicted rim load-displacement curve for Case (a) axial loading
analysis 141
4.12 Aluminium aerosol can used in experimental testing for case (a) axial 143
loading analysis
4.13 Zwick 20 KN tensile test machine 144
4.14 Steel insert and jubilee clip arrangement used in experimental testing 145
for case (a) axial loading analysis
4.15 Experimental rim load-displacement curve for Case (a) axial loading 146
analysis
4.16 Buckled can for Case (a) axial loading analysis (load = 2800 N) 147
4.17 Extended experimental rim load-displacement curve for Case (a) axial
loading analysis 148
4.18 Buckled can for Case (a) axial loading analysis 149
4.19 Cross-section geometry for the analysis o f Case (b) axial loading
during valve insertion and charging • 153
4.20 Three-dimensional model for Case (b) axial loading analysis 154
4.21 Structural constraints for Case (b) axial loading analysis 155
4.22 Applied loading for Case (b) axial loading analysis 154
4.23 Finite element mesh for Case (b) axial loading analysis 157
4.24 Pre-buckling equivalent stress contour plot for Case (b) axial loading
analysis 158
4.25 Predicted rim load-displacement curve for Case (b) axial loading
xiv
analysis 159
4.26 Exaggerated equivalent stress contour plot at the point o f buckling for 160
case(b) axial analysis
4.27 Deformed shape for Case (b) axial loading analysis 161
4.28 Aluminium aerosol can used in experimental testing for case (b) axial
loading analysis 164
4.29 Experimental rim load-displacement curve for Case (b) axial loading
analysis 165
4.30 Buckled can for Case (b) axial loading analysis 166
5.1 Die and punch geometry 170
5.2 Finite element model boundary conditions 171
5.3 Displacement loading and contact objects 172
5.4 Finite element mesh 173
5.5(a) Stage 1 partially deformed mesh 175
5.5(b) Stage 1 partially deformed mesh 175
5.5(c) Stage 1 partially deformed mesh 176
5.5(d) Stage 1 partially deformed mesh 176
5.5(e) Stage 1 partially deformed mesh 177
5.6(a) Stage 1 at the end o f the punch travel 178
5.6(b) When the punch retracted 179
5.7 Von Mises max equivalent stress counter plot at the end o f punch
travel 179
5.8 Comparison o f punch load Vs punch travel displacement for various
coefficient of friction 180
5.9 Stage 1 model with force loading 180
5.10 Comparison o f the effect o f coefficient o f friction on the Stage 1
extruded thickness profile with that predicated by result from[2] 181
5.11 Impact extrusion dome base 184
5.12 Stage 2(a) finite element model boundary condition 185
5.13 Stage 2(b) finite element model boundary condition 186
5.14 Finite element loading 187
5.15 Finite element mesh 188
5.16(a) Stage 2 deformed mesh 190
XV
5.16(b) Stage 2 deformed mesh 191
5.16(c) Stage 2 deformed mesh 192
5.16(d) Stage 2 fully deformed (max.punch travel 8.5 mm) 193
5.16(e) Stage 2 with the punch removed 194
5.16(f) Stage 2(a) punch travel = 8 mm 195
5.16(g) Stage 2(a) punch travel =7.5 mm 195
5.17 Thickness displacement characteristic 196
5.18 Stage 2(b) fully deformed mesh (max. punch travel = 8.5 mm) 196
6.1 Simplistic approach finite element mesh before reduction (centreline
thickness =1.25mm) 200
6.2 Simplistic approach finite element mesh after reduction (centreline
thickness =0.75mm) 201
6.3 Simplistic approach equivalent stress counter plot for p= l .20MPa
(centreline thickness =0.75mm) 202
6.4 Simplistic approach equivalent stress counter plot for p = l .35MPa
(centre line thickness =0.75mm) 203
6.5 Simplistic approach 3Dmodel (centreline thickness =0.75) 204
6.6 Simplistic approach 3D constraints (centreline thickness =0.75mm) 204
6.7 Simplistic approach 3D loading (centreline thickness =0.75mm) 205
6.8 Simplistic approach 3D mesh (centreline thickness =0.75mm) 205
6.9 Simplistic approach equivalent stress counter plot at the point o f
elastic-plastic buckling with p=2MPa (centreline thickness =1.25mm) 207
6.10 Simplistic approach equivalent stress counter plot at the point o f
elastic-plastic with p= l .80MPa (centreline thickness =0.75mm) 208
6.11 Optimisation analysis basic model with six super-element and six
design variables 211
6.12 Optimisation analysis geometry after optimisation 213
6.13 The convergence o f the solution 213
6.14 Equivalent stress counter plot (pre-buckling pressure=0.50MPa) 214
6.15 Optimisation analysis equivalent stress counter plot at the point of
elastic-plastic with p=0.62MPa) 215
NOMENCLATURE
Area o f can walls
The inside diameter
Rate o f dissipation o f energy per unit volume
Dissipation energy
Young’s modulus (MPa)
True modulus o f elasticity
Elastic modulus o f each element
Finite element
Finite element analysis
Objective function
Acceleration due to gravity
Inequality constraints
Geometric non- linearity
Equality constraints
Second Moment o f Area

Current iteration number
Can length
Load Factor
Internal pressure
Burst pressure o f the can
Yield pressure
Upper bound limit load
Lower bound limit load
Collapse pressure
Punch head radius, Radial co-ordinate direction, Can radius

Inner can radius
Outer can radius
Can wall thickness
xvii
•
u Displacement rates
u. Strain energy
Longitudinal stress
^2 Circumferential stress
^3 Radial stress
G char Characteristic stress within the element
° e Converged stress
Equivalent stress
( v eq)nom Nominal stress
a’ Stress increment
a Arbitrary constant
° L Limiting stress
G UTS
Ultimate tensile stress o f aluminium
° Yield Yield stress
°cr Critical stress
£ Strain
•
£ Strain increment
P The constant o f proportionality
K Elastic stress concentration factor
Coefficient o f friction
i Sum
Subscripts
Char Characteristic
d Associated with the arbitrary load(set) Pfl
e Elastic
eg Equivalent
I Iteration number
L Lower bound
r Resident
xviii
U Upper bound
y Yield
1, 2, 3 Principle values
Superscript
* Incomplete solution
Abbreviations
EIH Elastic isotropic hardening
EKH Elastic kinematic hardening
EPP Elastic perfectly plastic
Kt Stress concentration factor
xix
Chapter 1
INTRODUCTION
1.1 Background
The use o f thin-walled cylinders in the food packaging and cosmetics industries is
extensive and the demand for steel and aluminium containers is such that extremely
high-volume manufacturing processes have been developed over the past two
decades. The cost o f material is a major contributing factor to the overall cost o f the
container and so improvements in design and manufacturing processes are always
being sought.
The early cans were hand-made by practiced artisans who could produce up to six
each hour. The process was laborious and required considerable skill and strength.
The craftsman would cut a rectangular body and two circular pieces (for the lids)
from a sheet o f tinned iron. The rectangular body would be bent around a cylindrical
mould and the sides soldered together before affixing the ends. But the can opener
was not invented until 1930 [1].
This project continues on from work carried out by Patten [2] in conjunction with a
major manufacturer o f aluminium aerosol cans for the cosmetics industry. The
manufacturing process for these cans is described in detail in Chapter 2 but, in brief,
they are manufactured from a cylindrical billet o f almost pure aluminium using a
‘back-extrusion’ process. Prior to Patten’s work, ‘design by test’ was the recognised
method o f proving the designs and it was acknowledged that certain regions o f the
can cross-section were ‘over-designed’ and that potential savings were to be made.
1
Patten carried out an analytical study o f the manufacturing process and developed a
constant volume model to predict the thickness profile based on billet, punch and die
dimensions. He also carried out finite element analyses in order to identify regions o f
the cross-section where stresses were low and hence potential material savings could
be made.
An aerosol can is a thin-walled cylinder with a complex shape (see Section 1.2)
which limits the amount o f ‘design’ that can be undertaken using simple thin cylinder
equations to estimating the burst pressure o f the can. In reality, an aerosol can is
subjected to a number o f loading patterns including internal pressure, axial and radial
loading and although the behaviour o f a plain cylinder with constant wall thickness is
well understood, very little research has been conducted into the design o f these
more complex shapes. What is required is an analysis method that can be used to
accurately predict the elastic and elastic-plastic stresses and deformation o f these
cylinders due to internal pressure, axial and radial loading, as well as providing
details o f the modes and behaviour during failure, including buckling. Finite element
analysis (FEA) is such a powerful and comprehensive analysis method and has been
used comprehensively in this project, supported by experimental validation.
1.2 Thin walled pressure vessels
A pressure vessel is a closed structure containing liquids or gases under pressure. An
aerosol cans are one example o f a cylindrical pressure vessel categorised as a shell
structure due to its thin wall in comparison to its radius and length. The current
practice o f pressure vessel design by analysis is most commonly based on elastic
finite element analysis and the rules defined in codes such as BS5500 (the British
standard for unfired fusion welded pressure vessels) [3] and Section VIII o f the
ASME Boiler and Pressure Vessel Code [4]. This approach gives rise to two
significant problems in the design: elastic analysis is used to assess possible inelastic
failure mechanisms and the design by analysis rules is essentially based on shell
theory. These problems introduce the concept o f stress categories into the design
procedure. Some designers argued that plastic analysis should be the preferred
method for assessing failure modes associated with gross distortion due to a single
application o f pressure. Plastic and limit analysis can be performed using non-linear
finite element analysis, which is much more difficult to perform than elastic analysis.
Furthermore, an aerosol can has additional complexities due to its shape which, for a
one-piece aluminium can, consists o f an inverted base, a nominally constant
thickness parallel cylinder and a sloping top, as shown in Figure 1.1. The nature o f
the manufacturing process is such that there is a significant variation in thickness,
particularly between the base and cylinder, which limits the usefulness o f simple
design rules. The thickness profile o f an aerosol can is such that a number o f design
requirements have to be met:
Cylinder - must be thick enough to withstand bursting due to overpressure
- must be thick enough to withstand radial buckling during
bundling/packaging
- must be thick enough to avoid axial buckling/collapse under
compressive axial load during manufacture and charging
Base - must be thick enough to withstand bursting due to overpressure
3
- must be thin enough to facilitate ‘dome reversal’ at a specified
pressure below the burst pressure
Top - must be thick enough to withstand bursting due to overpressure
- must be thick enough to avoid collapse under compressive axial load
during manufacture and charging
and hence the design o f such components presents some interesting problems and
complex balancing o f structural integrity and manufacturing economy, for which
little detailed analysis has previously been undertaken.
4
Figure 1.1: Typical alum inium aerosol can
1.3 Aims and objectives of the research
The aims of this research project have been to investigate the factors influencing the
characteristic mechanical behaviour o f these thin-walled pressure vessels subjected
to internal pressure and axial loading and to provide a reliable analysis tool for future
optimisation studies. Included in this has been an investigation into the modelling of
the extrusion process by which such vessels are manufactured and an investigation
5
The aims o f this research project have been to investigate the factors influencing the
characteristic mechanical behaviour o f these thin-walled pressure vessels subjected
to internal pressure and axial loading and to provide a reliable analysis tool for future
optimisation studies. Included in this has been an investigation into the modelling o f
the extrusion process by which such vessels are manufactured and an investigation.
The numerical analysis used the ELFEN Non-linear finite element program, which is
an established commercial package [5] apart from the development o f (FE) analysis
method.
The specific objectives are:
• to further the understanding o f the process o f dome reversal (elastic-plastic
buckling) o f thin-walled cylinders with inverted bases subjected to internal
pressure
• to investigate the application o f the elastic compensation method for
estimating upper and lower bound pressure loads
• to accurately predict axial buckling loads and buckling modes using
experimental results for comparison
• to use finite element analysis to model the back-extrusion process and
compare the predictions with analytical solutions and experimental evidence
• to carry out a preliminary investigation into the optimisation o f the can
profile
6
1.4 Structure of thesis
This thesis consists o f eight main chapters:
Chapter 1 has provided an introduction to the project, stating the aims and
objectives o f the research.
Chapter 2 reviews the background information and literature relevant to all areas o f
the project.
In Chapter 3, using constant thickness models and a realistic thickness profile for an
axisymmetric and a full three-dimensional model, elastic and elastic plastic finite
element analysis predictions for the vessel subjected to internal pressure are
presented. The application o f the elastic compensation method to provide upper and
lower bound pressure estimates is also investigated.
The analyses in Chapter 3 are extended in Chapter 4 to predict the buckling and
collapse behaviour under axial compressive loading. Experimental tests to validate
the predictions are also described.
Chapter 5 describes the modelling o f the two-stage extrusion process, using billet,
tool and die data from a local manufacturer, have been considered and comparisons
with experimental measurements.
Chapter 6 describes a preliminary investigation into the optimisation o f the
thickness profile, as it is anticipated that material savings are possible, particularly in
the base region. The results from a reduced thickness model are presented and an
7
introduction to structured optimisation, using the DOT optimisation program, is
described.
Chapter 7 presents an overall discussion o f the research findings and conclusions
and recommendations for further study are provided in Chapter 8.
The appendices contain information not included in the main body o f the thesis.
8
Chapter two
GENERAL BACKGROUND AND LITERATURE REVIEW
2.1 Introduction
This Chapter reviews the background information and the appropriate literature
relevant to this research work on one-piece aerosol cans, which are a specific form o f
thin-walled pressure vessel. It starts by considering the design and manufacture o f
these pressure vessels, the basic concepts o f elasticity and plasticity (with particular
reference to finite element analysis) and appropriate theory for thin-walled cylinders
subjected to internal pressure or axial loading (i.e. buckling loading). The relevant
British and American standards for pressure vessel design are also reviewed. Also,
approximate methods o f determining pressure vessel limit loads, in particular the
elastic compensation method, are reviewed.
An important element o f the research has been the modelling o f the extrusion process
using finite element analysis and this subject is reviewed in Section 2.10. An initial
investigation into the optimisation o f a typical vessel profile is described in Chapter 6
and relevant background information is provided here. A considerable amount o f
research has already been carried out in the field o f extrusion and optimisation o f
aerosol cans and this is discussed in Sections 2.10 and 2.11.
9
2.2 The design and manufacture of aerosol cans
Aerosol cans are generally made of tin-plated steel (normally constructed from three
components; the base, the cylinder and the top, which are joined) or aluminium
(normally produced in one piece [6] from a curved billet, using the ‘back extrusion’
process - see Section 2.10). Examples o f aerosol cans are shown in Figure 2.1.
Figure 2.1: Aerosol cans [6]
The thickness of the tinplate steel varies, depending on the size, the pressure o f the
contents and the location (i.e. cylinder or ends). For the cylindrical section, the
thickness is typically in the range 0.18 mm to 0.25 mm whereas the tops and bottoms
are made from material that is typically between 0.28 mm and 0.43 mm thick [6].
Aluminium cans produced by the back-extrusion process have a typical thickness
variation of 0.31 mm to 0.41 mm in the cylindrical section and 0.7 mm to 1.31 mm at
the ends [6].
10
An aerosol can is a pressurised system and, as such, is governed by legislation. This
not only covers the design and manufacture o f the empty can, but also its subsequent
filling [6].
Legislation governs the amount o f product that may be contained in an aerosol can
since, for safety reasons there must always be some space in the can, which does not
contain liquid. The propellant occupies this empty space, which is greater when a
compressed gas, such as air, is used since it operates at higher pressures than those
for liquefied propellants.
2.2.1 Top and valves
A typical top with valve is shown in Figure 2.2. The components are:
• Valve Cup: - typically constructed from tinplate steel or aluminium
• Outer Gasket: - this is the seal between the valve cup and the aerosol can
• Valve Housing: - contains the valve stem, spring and inner gasket
• Valve Stem: - the tap through which the product flows
• Inner Gasket: - covers the hole in the valve stem
• Valve Spring: - usually stainless steel
• Dip Tube: -allows the liquid to enter the valve
• Actuator: - fitted to the top o f the valve stem.
When the actuator (red in the figure) is depressed it pushes the valve stem through
the inner gasket, and the hole is uncovered, allowing liquid to pass through the valve
and into the actuator and out to atmosphere.
11
VALVE CUP
.INNER G ASK ET
O UTER G ASK ET
VALVE H O USIN G
SP RING
DIP TU BE
Y\
Figure 2.2: Aerosol can valve [6]
Figure 2.2 shows the top of typical tinplated steel can which is pressed from flat
sheet. A typical one-piece aluminium top, shown in Figure 2.3, is less complex and is
often a simple tapered section with central rim into which a valve system insert is
added at a later stage.
12
Figure 2.3: A lum inium aerosol can valve
2.2.2 Main body
The main body o f a tinplated steel can is a constant-thickness rolled section, which is
joined using the welded process. And the round end pieces (pressed from another
sheet o f steel) are then fitted by a clinching process known as double seaming-
welded process [7]. Alternatively, the back-extrusion process for aluminium cans
produces a thickness profile in the cylinder, which is discussed in more detail in
Chapter 5.
13
2.2.3 Base
Most cans have bases that curve inwards and this shape strengthens the structure o f
the can. The inverted base design is also an inherent safety feature as it provides a
natural pressure release mechanism in the event o f a pressure overload, with ‘dome
reversal’ (which is a form of elastic-plastic buckling) o f the base occurring. This
sudden change in geometry (a) results in an immediate fall in pressure and (b)
provides a visual indication, since the can is no longer stable. In order for this
pressure release mechanism to be effective, the design must be such that ‘dome
reversal’ occurs at a pressure lower than the burst pressure [2].
Also, the last bit of the product collects in the small area around the edges o f the can
and this makes it easier to empty almost all o f the liquid as shown in Figure 2.4.
Figure 2.4: Aerosol can bottom [6]
The bottom dome-shape of the can base is produced by the forming process. This is
produced by supporting the can on a mandrel and forming the can base with a punch
as shown in Figure 2.5. This process has a direct effect on the pressure that the can
will withstand. The bottom forming process increases the can strength and provides a
14
safety feature that is required according to customer specifications [2] This states that
the can base must pop out at a pressure 20% lower than the can burst pressure.
The formation o f the base can take place either before or after the can is decorated
and the support provided to the can during base formation is different for the two
cases. The two types o f formation described and obtained by using finite element
method in Chapter five.
Figure 2.5: Bottom forming process
2.2.4 Principles of operation
The basic principle o f an aerosol can is very simple: One fluid stored under high
pressure is used to propel another fluid out o f the can [8]:
• A fluid is any substance made up o f free-flowing particles. This includes
substances in a liquid state, such as the water from a faucet, as well as
substances in a gaseous state, such as the air in the atmosphere.
15
• The particles in a liquid are loosely bound together, but they move about with
relative freedom. Since the particles are bound together, a liquid at a constant
temperature has a fixed volume.
• If the applied energy to a liquid is high enough (e.g. by heating it), the
particles will vibrate so much that they break free o f the forces that bind them
together. The liquid changes into a gas. This is the boiling process, and the
temperature at which it occurs is referred to as a substance’s boiling point.
The force o f individual moving particles in a gas can add up to considerable pressure.
An aerosol contains two essential components (see Figure 2.6): -
• The product, in the form o f a liquid, emulsion or suspension
• The propellant, which can be a liquefied or compressed gas
Liquefied propellants are gases that exist as liquids under pressure. Because the
aerosol is under pressure, the propellant exists mainly as a liquid, but it will also be
in the headspace as a gas. As the product is used up, some o f the liquid propellant
turns to gas and keeps the head space full o f gas. In this way the pressure in the can
remains essentially constant and the spray performance is maintained through the life
o f the aerosol. Compressed gas propellants occupy the headspace above the liquid in
the can. When the aerosol valve is opened the gas pushes the liquid out o f the can.
The mass o f gas in the headspace remains the same but it has more space, and as a
result the pressure will drop during the life o f the can.
16
LIQUEFIED PROPELLANT COMPRESSED GAS
J <€— l-P R O PE L L A W T
k i« n
S V A PO U R S
CO M PRESSED G A S — > «
/
/ product
I
I f l paFB
and
/
L I Q U I D P R O D U C T — I t t p H
| / LP R O P E L L A N T I f I
si l^ s l
Figure 2.6: Liquid and compressed propellant [8]
2.2.5 Design
The integrity of the can is the key condition, since a failure (e.g. burst or leakage)
could have catastrophic consequences. For a one-piece aluminium can, the thickness
profile is the principle design consideration.
At the same time, overall weight should be minimised in order to keep material costs
low (see Chapter 6). Experimental results from burst tests are discussed in Section
3.6.
17
2.2.5.1 Internal pressure
In practice, the customer generally specifies the minimum internal pressure, without
showing any visible signs o f deformation or failure. For the aluminium can
geometries used in this project, internal pressures o f 12, 15 or 18 bar, depending on
diameter, have been adopted.
2.2.5.2 Axial loading
The aerosol cans are required to support an axial load that is applied when the valves
are inserted as part o f the filling and charging process (see Figure 2.7). The cans
must support this load and show no visible signs o f deformation (buckling). Any
deformation o f the can will take the form o f flattening (collapse) o f the top, as shown
in Figure 2.8.
18
Axial Load
Figure 2.7: Axial loading during valve insertion and filling
Figure 2.8: The buckling of the can top
Also, during the forming o f the top, the plain can rim is also subjected to an axial
force which may cause the cylindrical section to buckle, as shown in Figure 2.9.
19
Figure 2.9: Axial loading during neck forming
2.2.5.3 Dome reversal of the inverted base
As previously stated, the dome reversal (or plastic buckling or plastic snap-through)
o f the inverted base is an important safety feature and design consideration. The
pressure-deformation response o f an aerosol can is carefully monitored (by
experiment) to ensure that this plastic buckling occurs at a pressure that is at least
20% below the corresponding burst pressure for that geometry [2]. This requires a
specific thickness profile, which must be controlled closely.
2.2.6 Manufacture
The production o f aluminium cans starts off in the form o f aluminium curved billets.
The process is shown diagrammatically in Figure 2.10 which is taken from [2].
Firstly the billet is coated in dry lubricant (graphite powder) and secondly the back
extrusion process forms the basic can shape with a flat base. By using tapered
extrusion dies the extrusion process allows the wall thickness to vary from the top o f
20
the can to the bottom. The cylinder is then coated internally with a protective
lacquer, which is cured in an oven. The base o f the can is formed, either before or
after decoration and drying ovens, in the bottom-forming machine. A series o f dies,
for the purpose o f producing the shoulder and neck on the extruded cans, are
designed to work within a tolerance range o f ± 0.01mm on the thickness o f the top o f
the extruded can walls. This means that any changes to the thickness o f the upper
third o f the cans may require a complete new set o f tooling for the necking machine.
Accl Acc2
Trim/Bottom Wash IX. Ovenl

form/ Brush
Oven2 Ovcn3 Oven4

Acc3
Base Print O.V. Neck Pack

Cost
Acc = Accumulator, LL, = Inside lining, O.V = Over varnish
Figure 2.10: Schematic diagram of aerosol can production line [2]
2.3 Basic concepts of elasticity and plasticity
2.3.1 Elasticity
For a perfectly elastic material, the removal o f the loads returns the component to its
original form with no permanent deformation. Most o f the equations used in design
engineering are derived from such an assumption, where stress and strain have a
21
linear relationship defined by Hooke’s law, [9] which is independent o f time and
load history. This relationship can be expressed in a generalized form as:
where [cr] is the stress matrix, [s] is the strain matrix and [d ] is the elasticity matrix
For a generalised three-dimensional state-of-stress:
<Jxx ^ XX
£
yy yy
°zz *zz
and [f] = . . . ( 2 .2 )
r*y
V r>*
Jxz _ y xz _
where cr^ ,<7^, and cr^ are the normal stresses and x ^ , r and x xz are the shear
stress also £ „ , Syy and s „ are the normal strains and y v , Y „ and y o are the shear
strains
For an isotropic material with a two-dimensional plane stress assumption:
u 0
[D] = 1 0 ...(2 .3 )
1 - 1/
0 0 ^
22
where E is Young’s modulus and v is Poisson’s ratio. Similarly, for plane strain,
V
1 0
T -y
E(\-v) v
\p] 1 0 ...(2 .4 )
2 u)
1-2 o
0 0
2(1 - v )
For an axisymmetric analysis:
O’z
w = (2.5)
c,
er
H = (2 .6)
£0
Y rz
V V
1- u 1- V
V
[Z)] = —— 1 (2.7)
l-o (l-2 u ) 1- v 1- v
V V
1
l^ u \^ v
2.3.2 Plasticity
Figure 2.11 shows a typical stress-strain curve for simple one-dimensional tension
(or compression) for an elastic-plastic material [10]. The stress at point A, which
separates the curve into an elastic portion and a plastic portion, is defined as the yield
23
stress crY. Because the yield stress is not always clearly identified, it is often taken as
the limit o f proportionality.
15
a
Y
Figure 2.11: Stress-strain curve for a simple one-dimensional tension (or
compression)
The general theory o f plasticity requires the following:
i. A yield criterion, which defines the onset o f plastic deformation under
multi-axial conditions o f stress;
ii. A flow rule, which relates the stress to the increments o f plastic strain;
iii. A hardening rule, which describes the work hardening o f the material and
how the yield condition changes with progressive o f plastic deformation.
The hardening rule also describes the material behaviour under cyclic
loading conditions;
24
2.3.2.1 Yield criterion
The purpose o f the yield criterion is to define the point o f yielding for a material
subjected to general 3-dimensional multi-axial stress system [16]. In the case o f uni
axial loading, yielding occurs when the axial stress reaches the uni-axial yield stress
for the material. However, for multi-axial loading, the effect o f all stress components
must be considered. Yielding o f an elastic-plastic material is defined by a scalar
function termed the yield function F, which is a function o f the stress invariants. The
yield function is written in the form, which leads to the conditions.
F<0 for elastic behaviour
F=0 for initial yielding and plastic
The most commonly used criterion for metals and that adopted by most finite
element programs (including the program used in this work) is the von Mises
effective stress criterion [12]. As early as 1913, von Mises suggested a yield criterion
o f this type, which is applicable to metal plasticity. The yield criterion has been
verified by a series o f experiments mostly on thin metal tubes under biaxial stress
states. According to von Mises:
F = ^ • { ( cV <t2 ) 2 + 0 2 - o-3) 2 + ( ( T 3- £T,)2 } - ( a , ) 2 ...(2.8)
where <ry is the uni-axial yield stress o f the material. Yielding is assumed to be
unaffected by the hydrostatic stress. Thus, yielding occurs when F = 0 and
...(2 .9 )
25
where creff is an effective stress for a mult-iaxial state-of-stress.
.^[(<7, -<r2)2 +(<t2 -< t3)2 +(<t3 - c r ,) 2] ...(2 .1 0 )
In principal stress space, the yield condition F ( cTj, cr2, <j 3 ) = 0 defines a yield
surface. The von Mises yield criteria is independent o f the hydrostatic stress and the
infinitely long cylinder shown in Figure 2.12 defines its surface. The axis o f the
cylinder makes equal angles with the coordinate axes. Stress points, which lie inside
the cylindrical yield surface, are associated with elastic stress states whereas those
that lie on the surface represent yielding . The TT-plane is defined by:
j 1 + a 2 + cr3 = 0
< ...(2 .1 1 )
and the intersection o f the 7T-plane with the von Mises yield surface, termed the yield
curve, appears as a circle.
03
02
.yield locus
02
oi n p la n e
yield surface
Figure 2.12: Projection of the von Mises yield surface onto the TT-plane
26
23.2.2 Flow rule (normality principle)
A flow rule defines the relationship between the stress components and the
corresponding plastic strain components after initial yielding. The direction o f the
plastic strain components is also defined through the flow rule by the plastic
potentials expressed as follows [12]:
. . . ( 2.12)
d{cr}
Associating equation (2.12) with particular a yield criterion (in order to obtain the
plastic strain increments) is generally known as a flow rule. The above rule is known
as the normality principle because equation (2.12) requires the plastic strain rate
components to be normal to the yield surface. In 1924, Prandtl [13] proposed stress
and strain relationships for an elastic-perfectly-plastic material under plane strain
conditions and later, in 1930, Reuss [14] generalized these relationships which
became known as the Prandtl-Reuss flow rule. Thus the Prantl-Reuss flow rule is the
rule associated to the von Mises yield criterion and, again, this flow rule is used
extensive by finite element codes (including the one used in this work) to predict
plastic strain increments.
2.3.23 Material hardening models
The most common material models used to analysis the behaviour o f a material
under elastic-plastic loading are elastic-perfectly-plastic (EPP), elastic-isotropic
hardening (EIH) and elastic-kinematic hardening (EKH) [15, 16]. As shown in
Figure 2.13.
27
Initial yield S u b s e q u e n t yield
Subseqent yield Initial yield
Isotrobic hardening
Kinematic hardening
Figure 2.13: Isotropic and kinematic hardening models
Most engineering materials work-harden if taken beyond their elastic limit. If a stress
reversal from tension to compression then takes place, there is a clear reduction in
the compressive yield stress when compared to the original tensile yield stress o f the
material. This is also true for stress reversal from compression to tension. This is
referred to as the Bauschinger effect. As can be seen in Figure 2.13, the isotropic
hardening model is based on the assumption that the hardening effect is the same in
both tension and compression, in other words ignoring the Bauschinger effect. The
onset o f compressive yielding will be initiated when:
o = -o y
...(2 .1 3 )
where a y is the current yield stress. For isotropic hardening the yield surface
increases in size but maintains its original shape under loading conditions. It can be
seen in Figure 2.12 that the von Mises yield appears as a set o f concentric circles.
28
On the other hand, the kinematic hardening model assumes a constant elastic stress
range o f 2 a y and can be used to model the Bauschinger effect (see Figure 2.13)
2.4 Overview of non-linear finite element analysis
Many texts provide detailed information on the modelling o f non-linear problems
(e.g.[12,17]) and only a brief summary is given here.
The majority o f general engineering components and structures are considered to
exhibit linear elastic behaviour under load and small deflection finite element theory
is used where the response o f the structure or material is directly proportional to the
load applied. Hooke’s law [16], which is illustrated by a simple spring problem,
givers a simple linear relationship between the applied force, F, and the resulting
deflection, u:
F = k .u ...(2 .1 4 )
where k is the spring stiffness. The deflection can be calculated easily by dividing F
by k. This is valid so long as the spring remains linear-elastic and the deflection is
such that they do not cause the spring material to yield. Therefore, doubling the force
doubles the deflection. In a finite element model, F and u are replaced' by matrices
and K becomes a square stiffness matrix. However, in many practical situations, the
force is not equal to K.u and these are referred to as non-linear problems. There are
three types o f non-linear finite element modelling [17]:
29
• Geometric non-linearity (GNL) - where large deformations and large strains
may be present. This includes snap-through buckling (see Section 2.8). The
deformations are large enough to cause the loading direction and stiffness to
change throughout the analysis.
• Material non-linearity - where plasticity, creep or visco-elasticity is present
in the material model and stress is not directly proportional to strain.
• Boundary non-linearity - where a status-dependant problem exists, in which
two surfaces come into or out o f contact.
The problems being studied in this work contain geometric non-linearity, material
non-linearity and boundary non-linearity.
The application in the finite element method to non-linear problems involves
replacing the non-linear loading history o f the structure by a sequence o f linear or
weakly non-linear increments. This means that instead o f applying the full load in
one load step, it is applied as a number o f small increments. An iterative procedure is
used within each load increment to ensure that the solution has converged within an
acceptable level.
2.4.1 Explicit and implicit methods
Numerical solution is often referred to as being explicit or implicit. When a direct
computation o f the dependent variables can be made in terms o f known quantities,
the computation is said to be explicit. In contrast, when the dependent variables are
30
defined by coupled sets o f equations, and either a matrix or iterative technique is
needed to obtain the solution, the numerical method is said to be implicit.
The principal reason for using implicit solution method which are more complex to
program and require more computational effort in each solution step, is to allow for
large time-step sizes.
In an explicit numerical method would be evaluated in terms o f known quantities at
the previous time step n. An implicit method, in contrast, would evaluate some or all
o f the terms in terms o f unknown quantities at the new time step n+1.
The choice o f whether an implicit versus explicit method should be used ultimately
depends on the object o f the computation. When time accuracy is important, explicit
methods produce greater accuracy with less computational effort than implicit
methods. Also the implicit options are important for other methods.
Explicit is a dynamic finite element tool specifically designed for application to
complex non-linear finite element simulations [5].
Explicit may be utilized for multi-phase analysis, for example a produce made from
sheet steel may be formed using four sets o f tools [5].
The implicit neutral file contains the entire model data associated with the
application
The explicit solver is more suitable for forming simulation. The analysis cost
increases in direct proportion to the size o f the mesh, whereas the implicit solver cost
31
increased with the square o f the matrix bandwidth o f the mesh. In this thesis the
implicit and explicit are used.
2.5 Thin cylinder formulae
Cylinders are usually considered to be either thick, where stress gradients due to
relative curvature are significant, or thin in which case, stress gradients are
negligible. If the ratio o f thickness to internal diameter is less than about 1/20 (D ;/t
>20) it is considered to be a thin cylinder.
By symmetry the three principal stresses in a thin cylinder subjected to internal
pressure are the circumferential (or hoop) stress, the longitudinal stress and the radial
stress [18] (see Figure 2.12) where:
hoop stress = <j { = ~ “ » longitudinal stress = cr2 = and radial stress = cr3 = -
P_
2
These stresses only depend on pressure and the cross-section o f the cylinder. The
length o f the cylinder has no effect, so long as the cylinder is long enough for ’end
effects’ to be ignored in which case these formulae are correct away from the
cylinder ends. Either the inside diameter Di, outside diameter Dc or the mean
diameter, Dm, can be used in these equations since the difference between them is
very small.
Using equation 2.10, yielding occurs when:
32
...(2 .1 5 )
2.6 BS5500
BS5500 is the British Standard for Unfired Fusion Pressure Vessels. It states wide-
ranging requirements for design, construction, inspection, testing and verification o f
compliance for this type o f pressure equipment. Grip the fluid under pressure is the
mean function o f pressure vessels. BS5500 provides equations to calculate minimum
thickness for vessels required to withstand a given internal pressure [3].
Figure 2.14: Hoop stress diagram
For equilibrium conditions from figure 2.14:
2 <retl = p l r l
33
This is a reasonable approximation o f the circumferential stress, which is used in
design because it is the largest, BS5500 takes cre to be the design stress / .
Substituting mean diameter D m = D + t into Equation (2.16):
£ (£ + 0 * t =_ P » _ . . . (217)
21 2 f - p
For cylindrical shells:
, =^ l . or t= ^ _ (2 1 8 )
2f - p 2 f - p
2.7 AS ME VIII
The American Society o f Mechanical Engineers set up a committee in 1911 for the
purpose formulating standard rules for the construction o f steam boilers and other
pressure vessels [4]. This committee is now called the Boiler and Pressure Vessel
Committee. To determine minimum thickness o f shells under internal pressure by the
following formulas:
For cylindrical shells:

where
t = minimum thickness
P = design pressure
R = inside radius
S = maximum allowable stress
E = efficiency o f appropriate joint in cylindrical shells
This, ASME VIII is very similar to BS5500
2.8 Buckling
When a component or structure is subjected to compressive loading, it may
experience visibly large displacements a direction perpendicular to the load at a force
well below the force required to cause the material to yield. This is known as elastic
buckling.
The primary path (curve oac) in Figure 2.15 load-deflection curve is the perfect
structure [19]. The second phenomenon is known as bifurcation buckling and this is
a very different kind o f failure. At the buckling load or bifurcation point the curve
will branch away from the primary path and continue on a secondary path (curve obd
on Fig 2.15). The subsequent deformation will follow a new path, which differs
considerably from the pre-buckling pattern.
This path will be followed if the post-bifurcation load deflection curve has a negative
slope and the applied load is independent o f the deformation amplitude.
To clarify this somewhat, bifurcation buckling is what could be termed elastic
buckling, as the branching point may be within the linear elastic region in which case
at least part o f the secondary path will be elastic.
35
In the case o f real structures, which contain unavoidable imperfections, true
bifurcation buckling occurs infrequently. In fact the structure will generally fail in
the snap-through manner describe later on. Imperfections will reduce the structure’s
strength and thus it will fail at a lower load than the perfect structure (represented by
curve oac on Fig 2.15) and thus curve oef shown the response for an imperfect
structure. Figure 2.15 also illustrates the varying buckling loads, where X CR is the
limit load o f a perfect shell, X CRJ is the limit load o f an imperfection structure and
XB1R is the bifurcation load. [19].
X CR
X BIR
— - IMPERFECT
STRUCTURE
PERFECT
X CRI STRUCTURE
BIFURCATION
0
Displacement
Figure 2.15: Load-deflection curves showing limit and bifurcation points [19]
36
2.8.1 Bifurcation buckling
A t a certain stage during the compressive loading o f structures, the equilibrium state
o f an ideal structure may reach a point beyond which two possible equilibrium paths
exist. The point at which these two paths diverge is known as the ‘bifurcation’ point
[20]. Beyond this point, the load-displacement characteristic o f the structure may
either follow the initial equilibrium regime (corresponding to the stress-strain curve
for the material) or follow a new path (associated with a different form o f
deformation). In practice, the characteristic follows the path that minimizes the total
potential energy o f the system. An axially compressed column that fails by Euler
buckling is an example o f this type o f failure. In a similar way, an ideal cylinder
subjected to an axial compressive load reaches a critical load at which the
deformation mode bifurcates from uniform axial compression into a pattern o f
diamond-shaped radial indentations.
The elastic buckling load for a cylindrical shell in axial compression, based on
classical theory, has been determined by many researchers and reviews o f early
theoretical work are presented by Timoshenko [9] and Bushnell [19]. The critical
stress <jcr, resulting from these calculations is:
2Et
cr.c r ... ( 2 .20 )
where E is Young’s modulus, t is the wall thickness, D is the cylinder diameter and
v is Poisson’s ratio.'
37
The critical stress is the minimum axial stress for buckling in the cylindrical shell
depending on the minimum buckling load which is a function o f X CR as can be shown
in Figure 2.15.
2.8.2 Pre-buckling deformation
Farshad [21] emphasises the importance that the pre-buckling solution has on the
prediction o f the bifurcation buckling loads by considering the pre-buckled state in
an axially compressed cylinder. Bushnell [19] describes a bifurcation buckling
failure mechanism o f straight-walled collapse due to edge effects developed in the
pre-buckled state. This type o f failure is common in straight-walled cans. The edge
buckling is mainly due to local hoop compression, which is greater nearer to the
cylinder end. Bushnell states that in a near-perfect shell, where imperfections and
end effects are negligible, edge buckling occurs before general instability remote
from the edge or axisymmetric collapse near to the edge. (i.e. the plastic collapse
observed in ‘thick’ shells). This study has shed light on the plastic failure in the can
base by highlighting the mechanism o f plastic hinge development.
2.8.3 Post buckling deformation
Experimental results suggest that actual collapse loads for axially compressed
cylindrical shells may be as low as 10 to 20 % o f the theoretical values. In 1932,
Flugge [22] carried out experimental tests on cylindrical shells under axial
compression in order to investigate this discrepancy. He found that his experimental
buckling loads were approximately one-half o f the theoretical values. Later, in 1941,
von Karman and Tsies [23] provided a major contribution to the understanding o f the
effects o f initial imperfections in cylindrical shells on the subsequent post-buckling
38
compressive behaviour. Their analytical results showed that the secondary
equilibrium path drops sharply downward from the bifurcation point.
2.8.4 Eigenvalue analysis
Eigenvalues are a set o f scalar values that are used in the solution o f a linear system
o f equations. They are also known as characteristic roots. (Eigenvalues and
eigenvectors) have particular significance in science, particularly in physics and
engineering. For example, in the context o f this research, they can be used to predict
the critical load at which a structure will bifurcate and also the ‘shape’ o f the
subsequent buckling pattern. An eigenvalue buckling prediction is based on the
determination o f singularities in a linear perturbation o f the structure’s stuffiness
matrix [24]. Because the lowest buckling mode is expected to be non-symmetrical,
an initial perturbation (small change) to the geometry is required to promote non-
symmetric deformation. A finite element eigenvalue analysis can be used to provide
the data necessary to locally perturb a perfectly symmetric geometry o f a structure in
order to create the non-symmetry o f loading required for buckling.
However, the eigenvalue approach is only useful if the perturbation is a realistic
representation o f the structural displacement prior to buckling. Consequently, the
method can only be used when displacements are small (and elastic) and, therefore,
the structure is stiff [24, 25].
Robotham e t a l [24, 25] used this method to investigate the elastic-plastic buckling
o f shafts (thin-walled tubes) subjected to torsion, using finite element analysis. They
39
demonstrated that accurate predictions for the collapse behaviour can be obtained
using this method, which has significant advantages over existing analytical theories.
2.8.5 Rik's method for modelling snap through
‘Snap through’ behaviour is associated with large elastic displacements, which result
in large changes in geometry prior to collapse. Structures that exhibit ‘snap through’
tend to reach a maximum sustainable load, which will then decrease or increase in
the post-buckling regime. At the point o f ‘snap through’, zero stiffness is reached and
a standard finite element analysis, based on the Newton-Raphson method, will
predict an unbounded displacement increment which often causes the program to
stop prematurely not allowing further prediction o f the load deflection
characteristics. The modified Riks method [26] is one approach that can be used to
overcome this problem o f zero stiffness.
A small imperfection (or perturbation) in the geometry is required and this is applied
to the structure prior to loading. As discussed above, this comes from an eigenvalue
analysis o f the structure. An incremental loading process is adopted and the modified
R ik’s method is used to determine the quasi-static equilibrium state at each
increment. However, unlike a traditional static non-linear analysis, the size o f the
load increment is variable in order to satisfy equilibrium conditions.
2.8.6 Buckling of thin-walled tubes
The buckling o f thin-walled cylinder under axial compression and lateral pressure
has been investigated by Flugge [22] who found that the effect o f the internal
40
pressure on the buckling load is negligible. He considered a thin walled cylindrical
shell o f length L, and wall thickness t.
The strength and stability o f a thin cylinder depends on a number o f factors including
the Young’s modulus and yield stress o f the material, the plate thickness and the
cylinder diameter. The mode o f failure may be buckling or yielding, whichever
occurs at the lower level o f applied force [7]. The compressive yield strength o f the
cylinder subjected to a uniform compressive force around its rim can be estimated
using:
F^, = 2 n R t c r y . . . ( 2.21)
where F = yield force
R = cylinder radius
t = cylinder thickness
<jy = yield stress o f cylinder material
But it is suggest that measured values are typically between 40 and 60% o f this
theoretical value [7]. Then the predicted collapse load in buckling:
F' =0.4619 E ... ( 2 .22)
where F = buckling force
41
2.9 Upper and lower bound analysis
The design o f pressure vessels and related components is usually based on a
combination o f finite element analysis and rules contained within the appropriate
codes o f practice such as BS5500 [3] and ASME VIII [4] where yielding is generally
considered to be the upper bound. Post-yield design is becoming more extensive,
with techniques such as elastic-plastic finite element analysis being used in order to
study shakedown and ratchet-ting regimes as well as collapse conditions. To avoid
the added complexities o f non-linear analysis, a limit load approach has been
suggested [27]. The lower limit is based on the lower-bound limit load theorem:
“If for a given load PL, a statically admissible stress field exists in which the stress
nowhere exceeds the yield stress o f the material, then P l is a lower bound limit load”
Correspondingly, the upper limit is based on the upper-bound limit-load theorem
“If, for a given load set, the rate o f dissipation o f internal energy in a body is equal to
the rate at which external forces do work in any postulated mechanism o f
deformation, the applied load set will be equal to or greater than the plastic collapse
load”
Direct calculation o f limit loads using upper and lower bound theories is very
difficult because it requires a statically admissible stress field and a kinematically
admissible strain field. In order to determine the equilibrium equations between the
external forces and internal stresses and the stress-strain relationships, a complicated
collapse solution is required. To avoid this, several alternative approaches have been
investigated see review in [30]. The reduced modulus method (see, for example,
[29]) has been modified [30] such that the elastic-plastic solution is replaced by a
series o f elastic solutions. After each elastic computation, the modulus o f elasticity is
42
reduced until the conditions o f admissible stress and strain fields, as lower and upper
bound criteria respectively, are satisfied.
This method has been further developed by Mackenzie and Boyle [31] and
Mackenzie e t a l [32], who have presented an elastic compensation method, where a
series o f elastic finite element analyses are used to predict a converged solution,
which meets either the lower or upper bound criteria. Applications such as beams in
bending and/ or tension, nozzles in spheres and torispherical heads are considered.
Gowhari-Anaraki and Adibi-Asl [33] have used the method to estimate upper and
lower limit and shakedown loads for beam members and a thick sphere.
Hardy e t a l [34] have used the method to estimate upper and lower bounds for
hollow tubes with axisymmetric internal projections under axial loading. They found
that this method could be used successfully to determine upper and lower bounds for
both limit and shakedown loading, when compared with elastic-plastic finite element
predictions.
Seshadri and Kizhatil [35] have suggested that if the procedure could not be verified
for simple components, it was unsafe to use it for more complex design. Hence, in
this work, a relatively simple geometry is used to further investigate the validity o f
the method.
2.9.1 Elastic compensation method
The aim o f the method, as described in [31], is to systematically re-distribute the
predicted stress field, while still remaining statically admissible, by carrying out an
iterative elastic analysis and modifying the local elastic modulus at each stage. An
43
initial elastic finite element analysis is performed with an arbitrary load set (e.g., Pd),
using the true modulus o f elasticity for the material, E 0. This is taken to be the zero
th iteration in a series of linear elastic analyses. In each o f the subsequent analyses,
the elastic modulus o f each element is modified according to the equation:
E, = E „ <j l t a ci„ ... (2.23)
where subscript ‘i’ is the current iteration number, crL is a limiting value o f stress and
<jchar is some characteristic stress within the element. It is suggested that this limiting
stress is related to the material yield stress, <rY, by:
crL= a <jy ... (2.24)
where a is an arbitrary constant between 0 and 1 ( li being found to provide suitable
convergence. It is also suggested that the characteristic stress is the maximum
(unaveraged) nodal equivalent stress associated with the element calculated in the
previous iteration, defined ascrM . Hence the iteration on element modulus o f
elasticity becomes:
£ , . = 2 £ M a y /(3< rM) ... (2.25)
The iterative procedure redistributes the stresses in the component and, over a
number o f iterations; the net effect is to decrease the maximum stress in the model to
reach a converged constant value a d .
44
2.9.2 Application to lower bound limit load
The lower bound limit load is calculated by applying the lower bound limit load
theorem. The converged elastic compensation solution satisfies the first requirement
o f the lower bound theorem in that it is statically admissible. Because the solution is
linear elastic, there is a linear relationship between stress and applied load. A lower
bound load, PL, can therefore be established as the load required giving a maximum
(nodal equivalent) stress in the component/structure that is equal to the uni-axial
yield strength o f the material, a y. for the worst point in the model and using
proportionality:
v* = pPi
and g y = p PL
(where p is the constant o f proportionality)
hence:
PL = Pd ° r / c d . . . ( 2 . 26)
The applied load setPrf, is not restricted to single loads and may represent multiple
forces, moments, pressure etc., in the form o f proportional loading.
2.9.3 Application to upper bound limit load
The upper bound limit load theorem for a complete plastic collapse solution can be
expressed as:-
45
£ / > it = \ b d V . (2.27)
v
where D is the rate o f dissipation o f energy per unit volume, P is the set o f
equilibrium external loads and u is the compatible set o f displacement rates, which
requires details o f that complete plastic collapse solution.
Alternatively, an upper bound solution can be found when an incomplete or partial
plastic collapse solution is available [32] and Equation (2.27) can be re-written in the
form:
... (2.28)
V
where the asterisk denotes an incomplete solution (i.e. a geometrically possible mode
o f deformation in which the stress field is not necessarily defined).
For this incomplete solution, compatible sets o f displacement and strain rate
increments are required and an iterative elastic finite element analysis, employing the
elastic compensation method, will provide such information. However, the finite
element predictions required to obtain the left hand side o f Equation (2.28) are not
always readily available. However, since the solutions are elastic, the elastic strain
energy increment can be substituted, i.e.:
V V
46
where & and s* are the elastically calculated stress and strain increments,
respectively. Also, the increment o f energy dissipation per unit volume for an elastic-
perfectly plastic material, using the von-Mises yield criterion, can be expressed as
[32]:
D = c r J j ( s , 2 + s 22 + s 32 ... (2.30)
where e i (i = 1, 2, 3) are the three principal strain rates.
Equation (2.28) can be re-written in simple form as:
U <D
and, as shown in Figure 2.16, the dissipation energy, D , is linearly related to the
applied load whereas the strain energy, £/, varies with the square o f the load.
Furthermore, the intersection o f the two lines represents the upper bound on the limit
load.
47
Pd Pu p
Figure 2.16: Variation in strain energy and dissipation energy with applied
load, used in the calculation of the upper bound limit load
The upper bound limit load is therefore obtained using predictions from the
converged elastic compensation finite element solution with the arbitrary load set, Pd
,i.e.
since U a P2 and D a P
then Ud a Pd2 and D ^ a P ^
and because the solutions are elastic:
48
U = P 2 and D = P for any load set, P
Pd2 Pd
Equating U and D at the upper bound limit load, Pu , gives:
Pu = ^ - P d ...(2 .3 1 )
Ud
where D d and U d are found from the converged elastic compensation finite element
solution.
2.10 Extrusion process and modelling
2.10.1 Process description
In 1797, Bramah [see [36]] described a press in which lead, maintained molten in an
iron pot, was forced by a pump into a long projecting tube, which served as a die.
This was the earliest example o f the extrusion process. A tapered mandrel was
supported concentrically with the tube by bridge in its enlarged end.
Extrusion is a forming method that is widely used in industry for producing a large
variety o f products such as window frames, tubes, cans and cables. The cross-
sections that can be produced vary from solid round, rectangular, to L shapes, T
shapes, tubes and many other different shapes. The extrusion process is a simple
method, which involves using a punch to press a ductile material through a die, thus
causing gross plastic deformation and forming the required shape.
49
The essential feature o f the extrusion process is the occurrence o f extremely high
pressures during the process, this being due to the constraints imposed by rigid tools.
This high pressure may increase the ductility o f the material, which in turn enables
large deformations to take place in one operation without the material cracking,
achieving at the same time precise dimensional accuracy and shape o f the product.
The process is particularly suited to high volume requirements and produces
components free from porosity. Working the metal in the cold state creates a fine
grain structure, which improves toughness, strength and hardness, and the high
quality finish is ideal for polishing and anodising. These features combine to give
price, quality and delivery advantages over other methods o f manufacture such as
turned parts, castings and deep drawn components [37].
Although extrusion is a modem process it precedes the development o f aluminium,
which was only commercially available since 1886. There are two types o f extmsion
commonly used in industry: direct and indirect extmsion as shown in Fig 2.17. In
direct extmsion, the die is located at one end o f the container and the metal to be
extmded is pushed towards it, hence moving relative to the container. In the case o f
indirect extmsion, the die is placed on the end o f the ram, and moves through the
container from one end (see Figure 2.17).
50
♦ «
I P «*-
l— l m
— r
* *
D ir e c t
I n d ir e c t
Figure 2.17: Direct and indirect extrusion [36]
In indirect extrusion of aluminium alloys, the process is characterized by the absence
of friction between the billet surface and the container. The load required is therefore
always decreased, compared with direct mode (as illustrated in Figure 2.18) and can
be reduced by as much as 50%. The advantages o f indirect extrusion are partly
related to the lower load needed and partly to the more uniform flow pattern
developed because of the absence of relative motion between the billet and the
container such that no heat is produced by friction [36].
51
Direct
indirect
Displacement
Figure 2.18 Extrusion load/displacement curves for direct and indirect
extrusion [36]
Impact (or back) extrusion is a type o f indirect extrusion process that produces
components by striking a cold billet, or slug, o f metal contained in a die cavity. The
metal slug is forced to flow around a punch by a single high-speed blow. The wall
thickness is controlled by the clearance between the punch and die [37]. This type o f
extrusion process is used to form aluminium aerosol cans (see Section 2.2.6). The
aluminium aerosol cans are manufactured in one piece by reverse-extrusion process
and the tool used for manufacturing is shown in Figure 2.19.
52
Upper corvamer
Figure 2.19: Tool setup used for the m anufacturing of the alum inium cans
2.10.2 Constant volume analytical approach
Patten [2] developed a program to predict the height and thickness variation in the
first stage of the back-extrusion process for aluminium aerosol cans, using a constant
volume approach. The mathematical method involved the development o f a
volumetric model of the extrusion process based on billet, punch and die dimensions
coupled with information on the punch travel. The profile is split into sections and a
cumulative volume conservation principle applied. The results produced were
validated through experimental measurement.
53
2.10.3 Finite element modelling
The application o f the finite element method to metal forming problems began as an
extension o f structural analysis techniques into the plastic deformation regime. Thus
early applications o f the finite element method to metal forming problems were
based on the plastic stress-strain matrix developed from the Prandtl-Reuss equations.
Hydrostatic extrusion, compression, and indentations were analysed using this matrix
and the infinitesimal variation formulations [38]. In recent years, a trend can be
observed towards a more objective documentation o f the empirical knowledge
available on extrusion components. The development o f automated design
applications or expert systems can be seen as a part o f this trend. These systems
require the explicit formulations o f the design rules. To formulate such rules more
knowledge o f the mechanics behind the extrusion process is required. The finite
element method can be a valuable tool in obtaining such knowledge, providing
insight into the process that cannot easily be obtained in any other way [39].
The use of the finite element method (FEM) is becoming increasingly important in
understanding the processes that occurs during aluminium extrusion.
Joeri [39] in 2000 described the finite element simulation o f the extrusion process for
aluminium prismatic sections with some new developments in the simulation of
aluminium extrusion for complex sections and simple sections also hollow profiles
was produced using the finite element method being reported. The simulation can be
used to investigate particular aspects o f the extrusion process. The simulation can
also be used directly in the design process to improve the design o f specific dies in
order to improve the performance o f these dies.
54
Joachim in 2005 [40] studied the backward cans extrusion process. The study
discusses the type o f punch used in the backward cans extrusion process, which is
commonly made with a cylindrical punch land. Using finite element analysis, the
radial contact force o f the punch has been determined. The results the finite element
simulations o f the process employing a new punch design show that a slight change
in the angle o f the punch land causes a drastic change in the contact conditions
between the punch land and can wall and the change in contact condition gives rise
to a net radial force on the punch, which will deflect the punch off centre leading to
variations in the can wall thickness. He does not consider the effects o f friction
coefficient or the relationship between punch force and punch travel.
2.10.4 Friction considerations
The effect o f friction in metal forming operations is fairly complex. Friction occurs
between the processed specimen and the forming tool in the appearance o f surface
shears, and therefore directly affects the position o f the planes o f principal stresses.
The effect increases with the increasing area o f contact between the specimen and the
tools, and with the reduction thickness o f the processed material [41].
According to the Coulomb friction law [39] the standard coulomb friction model
assumes that no relative motion occurs if the equivalent frictional stress is less than
the critical frictional stress. In the rough friction model for non-slipping case, it can
be further assumed that there seems to be no relative motion as long as the two
surfaces remain in contact. A penalty contact algorithm in the Lagrangian multiplier
method was adopted to remove the relative motion by dividing the friction force by
the penalty stiffness [42]. The effect on thickness o f friction coefficient is shown in
Figure 2.20. The thickness decreased with increasing coefficient o f friction.
55
1.2
0.9
0.0 0.1 0.2 0.3 0.4 0,5

Frictional coefficient
Figure 2.20: Effect of friction on the thickness of the sheet model [42]
The effect o f friction in the direct extrusion process is important in the commercial
process because it determines the billet size, either by pressure limitation or by the
surface at the end o f the ram stroke. During the extrusion process the normal pressure
on the interface between the aluminium and the die is so high that no slipping friction
occurs [39].
The higher the contact friction the higher the forming costs, see Figure 2.21.
Therefore, it is necessary to ensure that the contact friction is as low as possible in
the forming process [43]. This is achieved by applying a dry lubricant to the billet
before it enters the die.
56
40
3200
n 2400 2.4
CO
: 1600
800 —■— Panning tknc 0,8

—C— Fanning pressure
0.0
0.0 0.2 0.3 0.4 0.5
Frictional coefficient
Figure 2.21: Effect of friction on the final forming time and pressure [42]
2.10.5 Unloading (spring back)
The unloading process following extrusion is primarily elastic. However, because the
sheet is bent and unbent around the die and punch comers, some secondary yielding
may occur. Spring back is additional deformation o f the material that happens during
unloading. In the extrusion process, spring back is a phenomenon, which takes place
when the work-piece is removed from the tools after completing forming [44]. The
degree o f change in the shape depends on the material properties as well as the
technological parameters: restraining force, friction between the sheet and the tools.
It is an important consideration in both pressing and forming since the final shape o f
the component is (slightly) different to that created by the punch and die geometry.
R. Akbari, et a l [45] described the finite element code used to simulate the spring
back and sidewall curl in 2-D draw bending. Five stages have been applied for the
simulation o f the whole process.
57
The results showed that the springback and sidewall curl phenomena could be
completely simulated by ELEFEN software for sheet metal forming o f high strength
steel.
Mercer et a l [46] have illustrated the effective use o f different solvers in the
simulation o f material forming processes. By combining the advantages o f the
explicit and direct solvers, an effective solution scheme is obtained for simulating the
complete sheet forming process as well as subsequent analysis o f in service loads.
Joannic and Glin [47] used finite element analysis to simulate stamping or deep
drawing operations. A 3-D simulation code was used to design appropriate tools in
sheet metal forming. They found that the springback procedure proposed can
evaluate the amount o f springback in deep drawing processes in a realistic manner.
Narasimhan [48] studied numerical techniques for predicting springback deformation
in sheet metal forming. He used implicit and explicit finite element methods to
analyse the formation o f an actual automotive module. He found that a finite element
procedure that couples the implicit and explicit finite element methods accurately
predicted spring back in the stamping o f an automotive component. The overall
number o f iterations involved in the design o f die components could be significantly
reduced. Hence the simulation saved design and production time for manufacturers.
Arwidson [49] studied the numerical simulation o f sheet metal forming for high
strength steels very deep drawing situation was investigated both experimentally and
by numerical simulation for four high-strength steels. He observed that the
58
simulation is highly sensitive in the critical bending region. Also, varying the friction
coefficient between 0 and 0.1 had a significant influence on the results.
In the case o f the back extrusion o f aluminium aerosol cans and after these vessels
have been formed, there will be a small amount o f elastic strain left within the
aluminium. This will cause a very slight reduction in the dimensions and change in
shape. Since these are thin-walled cylinders, where the wall thickness is very much
less than the cylinder diameter, the mechanical elastic effects are very small and
therefore can be neglected.
2.10.6 Other issues
2.10.6.1 Effects of punch speed
During the extrusion process, the punch speed should be selected to make sure that
the dynamic effect on the deformable body is minimal. The final kinetic energy o f
the blank should be less than 5% o f the internal energy. Even in the initial contact
stage, the kinetic energy should not exceed 10% o f the internal energy. Usually the
peak punch speed is chosen to be in the range o f lm/s-5m/s. [39].
2.10.6.2 Thermal effects
During the extrusion process, the temperature o f the aluminium increases. A
significant portion o f this heating takes place during the early stages. The rest occurs
when the material flows around the punch. Because the thermal conductivity o f
aluminium is high, it is expected that localised heating is limited.
59
When the aluminium is plastically deformed, there is considerable heat generation.
This will affect the tooling dimensions. However, it is considered that these changes
in dimensions are small and are generally assumed to be negligible
2.11 Optimisation
Optimisation is the act o f obtaining the best results under the given or prevailing
circumstances [50]. In the design, construction, maintenance and operation o f
process plant, technological and managerial decisions must be taken at several
stages. The ultimate goal o f such decisions is either to minimise the effort required or
maximum the desired benefit. The effort required or benefit desired can generally be
expressed as functions o f certain decision variables. Thus optimisation can be
thought o f as the process o f finding the conditions that produce the maximum or the
minimum o f such functions. (Note that maximisation and minimisation are
interchangeable since the maximum o f a function can be determined by seeking the
minimum negative o f the same function).
There is no single method for solving all optimisation problems successfully or
indeed efficiently. Hence, it is important to identify the type o f optimisation problem
involved and then apply the appropriate procedure for its solution. Some o f the basic
building blocks o f optimisation were developed in the time o f Newton, Lagrange and
Cauchy [50], including the use o f calculus to obtain maxim and minima. There have
been spectacular advances since the advent o f digital computers. Currently,
optimisation in engineering covers a wide range o f applications.
The general optimisation problem may be stated as [50]:
60
X,
X.
X = which minimises /(X ) ...(2.33)
X.
subject to the constraints:
g<- ( X ) < 0 i= 1,2 ...., m ...(2 .3 4 )
h j (X) = 0 j = l ,2 ,....,p
where X is a non-dimensional vector
/(X ) is the objective function or performance index
g . (X) are inequality constraints
h y (X) are equality constraints
An objective function is a function associated with an optimization problem that
determines how good a solution is.
One particular area o f optimisation, which is o f particular interest to this project, is
the area of shape optimisation where optimised shapes are investigated in order to
reduce stress variations in components and to reduce the amount o f material used.
61
2.11.1 Previously published work
In 1984 Sodeik [51] published literature is not particularly helpful towards the
specific the design issue o f optimisation o f aerosol cans. He developed an equation
for axial collapse based o f a bead on the application o f the theory o f the point o f
metal yield. In the development o f equation he considers a single triangular bead
with three yield points but does not take into account the circular nature o f the
problem and the development o f hoop stresses.
In 1992, Jing Han e t a l [52] studied the application o f structural optimisation
techniques to aluminium beverage bottle design to investigate the influence o f the
design parameters to the buckling strength and rigidity o f the base under an axial
load and internal pressure. His paper dealt with the shape optimisation o f the bottom
o f aluminium beverage bottles by applying the structural optimisation technique. He
used non-linear finite element analysis to study the influence o f the design parameter
on the buckling strength and rigidity o f the bottom under an axial load and internal
pressure. The thickness o f the bottom and the top parts o f sidewall are t = 0.4 mm
and 0.135 mm. The objective function which he used for optimisation to maximize
the column strength o f the bottom is:
F = P{X)
With design variables
X = {X i ] , i = l...n
where n is the number o f design variable
62
The results obtained a 50% increase in column strength. Also he established that
using the progressive optimisation method can help designers to understand the
optimisation problem more clearly and the computational cost was reduced.
In 1999, Benjamin [53] studied the computational strategies for the design and
optimisation o f three-piece steel food cans. He described the performance o f the
finite element models across a range o f bead-depths. He used the suite to provide
consultancy services to a number o f can-makers to study the effect o f geometry and
material changes. This study calculated that using the current can material and
geometry specifications finite element models should be constructed. He studied the
effect o f imperfections on the post buckling behaviour o f food cans and the axial
collapse of the beaded cans in 2-D and 3-D model
2.12 Closure
The background information and published literature relevant to this project have
been reviewed. In particular, information on thin-walled cylinders under internal
pressure and thin-walled cylinders subjected to axial loading has been presented and
will be referred to in later chapters o f this thesis. The areas o f extrusion modelling
and optimisation analysis have also been explored, since they also feature in this
work.
Several observations are made based on this review:
1. there is a requirement for further understanding o f the process o f dome reversal
(elastic-plastic buckling) o f thin-walled cylinders with inverted bases subjected to
internal pressure;
63
2. similarly, a need for greater accuracy in the modelling and prediction o f axial
buckling loads has been identified;
3. the British and American standards do not cover such design considerations;
4. further applications o f the elastic compensation method are required in order to
validate the approach;
5. there is little evidence o f finite element analysis being applied to the modelling o f
the back-extrusion process;
6. the raw material costs associated with the manufacture o f aluminium aerosol cans
are very high and, at the same time, the cans are often considered to be over-
designed. Consequently, there is a requirement for the optimisation o f the can
thickness profile such that material can be reduced while, at the same time,
maintaining the same critical pressures for dome reversal and bursting.
These six issues are investigated further in the remaining chapters.
64
Chapter three
INTERNAL PRESSURE LOADING
3.1 Introduction
This chapter describes the results o f an analysis o f the aerosol can under internal
pressure loading using finite element analysis, together with details o f the
experimental validation o f the predictions. The analysis focuses on the base o f the
can, since this is the critical area for improvements to be made. Elastic analysis is
used to study the elastic stress distributions and the onset o f yielding and elastic-
plastic analysis is used to investigate the post-yield behaviour up to dome reversal
and eventual failure. The application o f approximate methods, such as the upper and
lower bound techniques, to this type o f component and loading is also investigated.
Full details o f geometries, loading, boundary conditions and material models
considered are given in Sections 3.2.1 to 3.2.3. The mechanisms are described in
detail using one particular geometry-load combination and an elastic-perfectly-
plastic material model. The maximum internal pressure that the charged pressure
vessel will withstand is presented together with upper and lower bound estimates.
The finite element predictions have been obtained using ELFEN Version 3.0.4, [5] a
finite element program for Microsoft Windows NT. The program allows pre
processing, analysis and post-processing stages to be completed within a single
application. The program can be used to model a large number o f situations
including buckling, plastic deformation, forming and stress analysis problems, etc.
65
3.2 Elastic finite element analysis using axisymmetric models
3.2.1 Geometry and finite element models
In order to fully understand the stress response, predictions have been obtained for a
series o f constant thickness can profiles as well as for an actual profile. This has been
developed using the measured outside profile o f an actual can [2] and assuming a
constant thickness throughout. The model used for the actual profile analysis and the
corresponding results are presented in Section 3.3.5.
The basic shape is assumed to be axisymmetric about the Y-axis. The geometry o f
the base is described using six dimensions; H (the dome depth), t (the wall
thickness), L (the flat base length), R (the major arc radius), r (the minor arc radius)
and the angle#, as shown in Figure 3.1. Seven geometries have been considered in
this analysis and the relevant geometric parameters are listed in Table 3.1. A detailed
investigation o f geometry G4 is described and a summary o f results presented for the
other geometries.
66
Figure 3.1: Can base geom etry (constant thickness)
Geometries H (mm) t (mm) L (mm) R(mm) r (mm) <9(degree)

G1 7.5 0.4 8.924 13.75 3.20 87.8°
G2 8.5 0.6 8.924 13.75 3.18 86.7"
G3 8.5 0.8 8.924 13.75 2.73 67.3“
G4 8.5 1.0 8.924 13.75 1.94 50.5"
G5 8.5 1.2 8.924 13.75 1.62 30.4”
G6 8.5 1.4 8.924 13.75 1.50 25.2“
G7 8.5 vary 8.924 14.13 3.23 64.2°
Table 3.1: Geom etric p aram eter
The basic finite element model is shown in Figure 3.2. During mesh generation,
these "super elements" are sub-divided to create a suitable mesh, a typical mesh
being presented in Figure 3.3 and discussed later in this chapter.

No Y (U tplat'nnnil
Figure 3.2: Basic finite element model of can base (constant thickness)
The top section of the can is not included at this stage to simplify the model. This
will have little effect on the results, as it is known from experimental testing that the
base of the aerosol can deforms first. On the basis o f preliminary predictions, the
cylindrical section was made long enough to ensure that a uniform stress distribution
was reached away from the comers. Additionally, by removing the top section o f the
can, the model accuracy and computation time is increased. 8 noded, axisymmetric,
isoparametric elements have been used because o f their efficiency and increased
accuracy.
3.2.2 Loading and boundary conditions
The axisymmetric model shown in Figure 3.2 is constrained in the X direction along
the plane AA to represent symmetry. It is also constrained in the Y direction along
the plane GG. These constraints on the displacement are sufficient to prevent a
singularity occurring in the finite element solution. Elastic finite element calculations
68
have been performed for a pressure load o f 1 bar (0.1 MPa) applied uniformly on the
inner surface o f the model.
3.2.3 Materials models
The material assumed for the elastic analysis is aluminium 1050, which is commonly
used in this type o f application. This means that the aluminium is 99.50% pure, with
0.5 % natural impurities [54] (and therefore no added impurities). The mechanical
i
| properties are given in Table 3.2 [2]. The results are generally normalized with
i
|
respect to these material properties.
Material properties Value

Density, p (kg/m3) 2700
Young’s modulus, E (GPa) 68.3
Poisson’s ratio, v 0.33
Yield stress (MPa) 100
Ultimate tensile strength (MPa) 156
Table 3.2: Mechanical properties of 1050 aluminium [2]
3.2.4 Constant thickness models
The region of the component under investigation is the can base, which is subjected
to uniform pressure loading. In addition, the whole component is analysed in order to
obtain ‘nominal’ predictions. Geometry G4, having t = 1 mm, is selected for a full
review and a summary o f the results is given for the other geometries. The mesh was
generated using the ELFEN mesh generator and the mesh for G4, containing 1267
elements, is shown in Figure 3.3(a).
69
In finite element modelling, a better-quality mesh typically results in a more accurate
solution. However, as a mesh is made better, the computation time increases. There
is one way to perform a mesh convergence study as follows:
• Create a mesh using the fewest reasonable number o f elements and analyze
the model.
• Recreate the mesh with a denser (biggest number o f element) element
distribution, re-analyze it and compare the results to those o f the previous
mesh.
• Keep increasing the mesh density and re-analyzing the model until the results
converge satisfactorily.
This type of mesh convergence enables an accurate solution with a mesh that is
sufficiently dense and not overly demanding o f computing resources, the mesh
convergence for this analysis as shown in Figure 3.3(b).
Figure 3.3(a): Finite elem ent mesh for geom etry G4
70
26.5 i
26.4
26.3
26.2
26.1
X
(0 25.9
S 25.8
25.7
25.6
25.5
500 1000 1500 2000
No of element
Figure 3.3(b): Mesh convergence
3.2.4.1 Results for geometry G4 (t = 1 mm)
Elastic principal stress contour plots for G4 for an internal pressure o f 0.1 MPa are
presented in Figures 3.4, 3.5 and 3.6. It can be seen that crx has a maximum localized
value o f +15.72 MPa on the inside surface close to the intersection between the base
and vertical sides (section EE in Figure 3.2). Elsewhere, cr1 is reasonably uniform
and o f low value. <j2 varies between +2.02 and -14.46 MPa with the maximum
compressive value on the inside surface between sections CC and DD in Figure 3.2.
<r3 varies between +9.26 and -11.43 MPa with a maximum tensile value close to
section DD in Figure 3.2 and generally compressive stresses in the uniform base
71
region, cii is the hoop stress, <Ji is the longitudinal stress and <53 is the radial stress
approximately.
1 5 . 72 1 7 0
14.28344
12 . 84518
1 1 . 40 6 9 2
9 . 968660
8 . 530399
7.092140
5.653880
4.215619
2.777359
1.339099
- 0 . 0 99 1 6
-1
Figure 3.4: cr, contour plot (G4, p =0.1 M Pa)
2.021949
0 . 64 83 6 8
- 0 . 72521
-2.09879
-3.47238
-4.84596
-6.21954
-7.59312
-8.96670
-10.3403
-11.7139
-13.0874
-14.4610
Figure 3.5: (7-, contour plot (G4, p = 0.1 MPa)
72
Figure 3.6: cr3 contour plot (G4, p = 0.1 MPa)
The corresponding principal stress distributions around the inside and outside
surfaces between points A and G (see Figure 3.2) are shown in Figures 3.7 and 3.8
respectively. It can be seen from figure 3.8 that the highest principal stress ( crt ) is at
Section FF. Also cr1 is higher at Section CC then cr3at Section EE. As can be seen
from the figure, there are very sharp rise in cr0 at point E compared to the other
stresses. The flat sections of these curves occur when two o f the principal stresses are
negative and the other (maximum) must therefore be zero.

1.20 *
1.00 A
0.80 Y it
iI
1
i
0.60 T
0.40
f k
II
f
0.20
y>
rÊ“
Ih
X
i
1
0.00 -Uc
■IB 1
- 0.20
0.00 0.40 0.80 1.20 1.60 2.00 2.40 2.80 3.20 3.60 4.00
A B C D E F G
• IB 4
0.20
0.00
- 0.20
-0.40
-0.60
- 1.00
- 1.20
.-1.40
-1.60
0.40 0.80 1.20 1.60 2.00 2.40 2.80 3.20 3.60 4.00
A C E G
74
a* 1
-------
0.80 .....
.—...J A
0.60
J f t
0.40 i *
0.20 1
1
0.00 j
4*f
-
9* S s /1
w
0.20 /
f
-0.40 f
J
-0.60 /*
- 0.00
r
y*
- 1.00
.1*1 1
. - 1.20
0.00 0.40 0.80 1.20 1.60 2.00 2.40 2.80 3.20 3.60 4.00
A C E G
Figure 3.7: Principal stress distributions around the inside surface (G4, p = 0.1
M Pa)
8.00
7.00 h
6.00 >
I 7
% *
5.00 1 I
*
f \
\ i m
4.00
. \
\ i
2.00 \ ■"1.................
I V
»
1 3^
\ i
1.00
\ /
i
\ 1
Jf
0.00 J
K tf 1
-1.00
0.00 0.50 1.00 1.50 2.00 2.50 3.00 3.50 4.00 4.50
A C E G
75
1.00
0.00 1 - m
11
V m I
I v f
- 1.00
\ f
If \
V
i
-
-2.00
1.00
\ M
I
j
m
I I
r f
i
if
T- '
f \
2 i.oo *
• f.
-5.00
< f \ *
j
I
-6.00 IT
J
7
\
-7.00 V r
u
-8.00 V 2
«!• 1
-9.00
0.00 0.50 1.00 1.50 2.00 2.50 3.00 3.50 4.00 4.50
A C E G
8.00 TT
A
7.00 i *
6.00 7
/
t1
*
5.00 "■T
4.00 j k i
i.
i i
3.00
I \i i
»
2.00 I*
V 4
/
1.00 T’ !T j f / -
0.00 \ ... 1 J
M \ t4
-1.00 1i
■an 1
- 2.00
0.00 0.50 1.00 1.50 2.00 2.50 3.00 3.50 4.00 4.50
A C E G
Figure 3.8: Principal stress distributions around the outside surface (G4, p = 0.1
M Pa)
76
The von Mises equivalent stress contour plot, for p = 0.1 MPa, is shown in Figure
3.9. The equivalent stress distributions around the inside and outside surface are
shown in Figure 3.10. The maximum equivalent stress is 13.88 MPa and occurs on
the inner surface close to the point E in Figure 3.2. The maximum equivalent stress
index, I e9(or elastic equivalent stress concentration factor), is obtained by dividing
the maximum equivalent stress by the nominal stress:
-( 3 .1 )
( Oeq / nom
where the nominal stress is found from:
Vt(CTl ~ a 2>2 +(<J 2 - C 3) 2 +(<T, -<T3)2] ...(3 .2 )
where a. = 0 ,= ^ - , <7, = - — ...(3 .3 )

1 It 2 At 3 2
and D is the inside diameter.
Using p = 0.1 MPa, D = 53 mm and t = 1 mm:
<Ti = 2.65 MPa, cj2 = 1.325 MPa, <33 = -0.05 MPa, ( a eq)nom = 2.34 M Pa and hence Ieq
= 5.93.
It is clear that yielding will occur here first at a pressure which is well below the
yield pressure for the plain tube region o f the can with 1 mm wall thickness, which is
77
when ((Jeq)nom = 1 0 0 MPa. Hence, scaling up these elastic results, first yield occurs
when:
p v = - ^ --.0.1 = 0.720 MPa -fo r the base

y 13.88
compared with:
p v = - ^ - .0 .1 = 4.27 MPa - for the plain tube

y 2.34 F
78
13.87696
12.73571
11.59446
10.45321
9.311964
8.170716
7.029469
5.888221
4.746973
3.605725
2.464477
1.323230
0.181982
1 3 .8 7 6 9 6
1 2 .7 3 5 7 1
1 1 .5 9 4 4 6
1 0 .45321
9 .3 1 1 9 6 4
8 .1 7 0 7 1 6
7 .0 2 9469
5 .8 8 8 2 2 1
4 .7 46973
3 .6 0 5 7 2 5
2 .4 6 4477
1 .3 2 3230
f l . 18198?
Figure 3.9: E quivalent stress contour plot (G4, p = 0.1 M Pa)

1.40
1.20
1.00
E q u iv a le n t s t r e s s
0.80
0.60
0.40
0.20
0.00
0.00 0.40 0.80 1.20 1.60 2.00 2.40 2.60 3.00 3.20 3.60 4.00
A C E G
(a) Inside surface
9.00
8.00
7.00
s tr e s s
6.00
5.00
E tjc j/v a le n t
4.00
3.00
2.00
1.00
0.00
0.00 0.50 1.00 1.50 2.00 2.50 3.00 3.50 4.00 4.50
A C E g
(b) O utside
Figure 3.10: E quivalent stress distribution (G4, p =0.1 M Pa)
80
3.2.4.2 Effects of wall thickness
Equivalent stress contour plots for t = 0.4, 0.6, 0.8, 1.2 and 1.4 (which represents the
variation in thickness seen in actual cans), with an internal pressure o f 0.1 MPa are
shown in Figures 3.11, 3.12, 3.13, 3.14 and 3.15 respectively.
The maximum equivalent stresses in each case are in the comer region close to the
point E (see Figure 3.2) and decrease with increasing thickness. The equivalent
stress for t = 0.4 mm (Figure 3.11) varies between 62.2 and 0.13 MPa close to the
Sections DD and EE in Figure 3.2. For t = 0.6 mm, (Figure 3.12), the equivalent
stress varies between 43.5 and 0.27 M Pa close to the Section EE. For t = 0.8 mm
(Figure 3.13), the equivalent stress varies between 24.9 and 0.03 MPa close to the
Section DD and EE. For t = 1.2 mm (Figure 3.14) the equivalent stress varies
between 8.48 and 0.06 MPa close to the Section DD and EE, and for t = 1.4 mm
(Figure 3.15) the equivalent stress varies between 6.72 to 0.02 MPa close to the
Section DD and in the Section AA in Figure 3.2. The relationship between thickness
and maximum stress is clearly non-linear.
81
02-23291
44.49974
40.46656
36.43339
32.40021
20.36703
24 .2 3 3 8 6
20.30060
16.26751
12.23433
0.201157
4.167981
0.134805
Figure 3.11: Equivalent stress contour plot (G l, p = O.IMPa)
82
43.59052
39.98090
36.37128
32.76167
29.15204
25.54243
21.93281
18 . 32319
1 4 . 71 3 5 7
11.10395
7 . 494327
3 . 884708
0.275088
42.69052
99.90090
36.31126
32.16161
29.
25.54243
21.93261
18.32319
14.11351
11.10295
1.49432?
3.884108
0.215086
Figure 3.12: Equivalent-stress contour plot (G2, p = 0.1 M Pa)
83
r 24.96066
- 22.88348
- 20.80630
- 18.72912
- 16.65194
- 14.57476
b 1 2 . 49 7 58
10 . 4 2 0 3 9
8. 343214
6. 266034
4. 1 8 8 0 5 3
2.111673
0. 0 3 4 4 9 2 0
18.72912
16.65194
14.57476
12.49758
10.42039
8.343214
6.266034
4.188853
2.111673
0.0344921
Figure 3.13: Equivalent stress contour plot (G3, p = O.IMPa)
84
8.489853
7.787560
- 7.085268
- 6.382975
- 5.680682
- 4.978389
- 4.276096
- 3.573803
- n Q T 1 Q 1 1
0 .4 8 8 6 4 1
7 .7 8 6 4 1 7
7 .0 8 4 1 9 3
6 .3 8 1 9 6 9
5 . 6 7 974C
4 .9 7 7 5 2 2
4 .2 7 5 2 9 0
3 .5 7 3 0 7 4
2 .8 7 0 8 5 0
2 .1 6 8 6 2 7
1 .4 6 6 4 0 3
0 .7 6 4 1 7 9
0 .0 6 1 9 5 5
Figure 3.14: E quivalent stress contour plot (G5, p = 0.1 M Pa)
85
6.727580
6.169023
5. 610465
5. 051908
4 . 493351
3. 934793
3 . 376236
2. 817679
2. 259122
1.700565
1.142007
0.583450
0. 0248930
Figure 3.15: Equivalent Stress C ontour plot (G6, p = 0.1 M Pa)
A summary o f the maximum equivalent stresses, nominal stresses and maximum
equivalent stress indices, together with the number o f elements in each mesh, is
86
presented in Table 3.3. These results are presented graphically in Figures 3.16 and
3.17. It is clear that the relationship between I and thickness is reasonably linear for
t in the range 0.4 to 1.4 mm, whereas changes in thickness have a more marked effect
on maximum equivalent stress particularly when the value o f thickness is at the
higher end o f the range considered.
These results are presented in an alternative form in Figure 3.18. Here, the diameter
has been normalised by dividing by the thickness to give the dimensionless
parameter D/t.
Geometry G1 G2 G3 G4 G5 G6
Thickness (mm) 0.4 0.6 0.8 1.0 1.2 1.4
D/t 132.50 88.33 66.25 53 44.16 37.85
62.23 43.59 24.96 13.88 8.48 6.72
)
(V< j eq / nom 5.77 3.86 2.91 2.34 1.95 1.65
A
13.83 11.29 8.58 5.93 4.34 4.07
K
No o f elements 424 889 894 1267 1514 1787
Table 3.3: The variation of maximum equivalent stress index with wall
thickness
i
[
87
40
0.2 0.4 0.6 0.8 1
Thickness (mm)
Figure 3.16: The relationship between wall thickness and m axim um equivalent
stress
0.2 0.4 0.6

Thickness (m m )
Figure 3.17: The relationship between m axim um elastic equivalent stress index
and the wall thickness

100 120 140
D/t
Figure 3.18: M axim um equivalent stresses, nom inal stresses and m axim um
equivalent stress indices versus D/t ratio
3.2.4.3 Limiting pressures
Using the method of analysis discussed in Section 3.2.4.1, the elastic predictions
have been scaled linearly in order to obtain values o f the limiting pressure (the
pressure at which yielding will first occur) for the can base as well as for the plain
tube region. These predictions are presented in Table 3.4 and Figure 3.19. It can be
seen that the relationship is reasonably linear over the range 0.4 mm < t < 1.4 mm.
89
Lim iting pressure (M Pa)
Thickness D/t
(mm) C an Base Plain Tube
0.4 132.50 0.16 1.73
0.6 88.33 0.22 2.59
0.8 66.25 0.40 3.43
1.0 53 0.72 4.27
1.2 44.16 1.17 5.12
1.4 37.85 1.49 6.06
T able 3.4: Lim iting pressures for constant thickness cans
C an b a se
P la in tu b e
02 0.4
Plain tube
100 120 140
Figure 3.19: V ariation of limiting pressure with wall thickness and D/t
3.2.5 Can with varying thickness
In practice, the actual thickness profile o f a can is highly non-uniform for a number
of reasons. Based on experimented measurements, Patten [2] obtained a thickness
profile and this has
been used here to obtain realistic values o f stress and limiting pressure. It was found
that the thickness varies in the range from 0.31 mm to 1.31 mm. A variable thickness
model, which reflects the true thickness profile o f measured cans has been used and
it is clear that there is a significant difference between the thickness o f the cylindrical
section (0.31 mm minimum) and that o f the base (1.31 mm maximum). The basic
finite element model is shown in Figure 3.20. The constraints, loading conditions and
material properties are as discussed in Sections 3.2.2 and 3.2.3 for the constant
thickness models. A suitable mesh of 8 noded, axisymmetric, isoparametric elements
was generated using the ELFEN mesh generator and is shown in Figure 3.21.
Y No Y -d isp la ce m e n t
Axi8sym etric about Y-axis
N o X -d is p la c e m e n t
Figure 3.20: Finite elem ent model of can with varying thickness
91
Figure 3.21: Finite elem ent mesh for can with varying thickness
3.2.5.1 Results
The predicted variation in maximum equivalent stress, for an internal pressure of
O.IMPa, is shown in the form of a contour plot in Figure 3.22. There are large stress
gradients close to the regions CC and FF in Figure 3.2 and the maximum value of
equivalent stress is 25.68 MPa at the interface between the base and the cylinder. The
distribution of equivalent stress around the inside and outside surfaces is shown in
Figures 3.23 and 3.24 respectively. Inner surface equivalent stresses are generally
greater than the corresponding outer surface values.
From the experimental measurements, it is seen that the thickness in the plain region
of an actual can is approximately 0.31 mm and this gives a nominal stress o f 7.43
MPa (Equation 3.2) and, therefore, a maximum equivalent stress index o f 3.45.
92
25.67913
■ 23.54039
21.40164
19.26290
17.12416
14.98542
12.84668
10.70794
8.569198
6.430457
4.291716
2.152975
0.0142334
014233'
Figure 3.22: Equivalent stress contour plot at internal pressure of 0.1 M Pa

2.75
2.50
1 I
2.25 r
i
2.00
1.75
j
f
¥
it
I\
I
f
1.50 T
1.25 1|T“--------
1.00
Uj 0.75 • ij 11f
t
I
■
X
0.50 f
3
P *
T \ ¥
0.25 U a ll ft
0.00
0.00 0.40 0.80 1.20 1.60 2.00 2.40 2.80 3.20 3.60 4.00
A B C D E F G
Figure 3.23: Equivalent stress distribution around inside surface
1.80
1.60
1.40
.20
■5 100
1
| 0.80
0.60
0.40
0.20
0.00 0.40 0.80 1.20 1.60 2.00 2.40 2.80 3.20 3.60 4.00
A B C D E F G
Figure 3.24: Equivalent stress distribution around outside surface
The limiting pressure (the pressure at which yielding will first occur) has been
obtained using the method described in Section 3.2.4.1 and a value o f 0.38 MPa is
predicted. This result is seen to compare to an average thickness o f 0.73 mm. It is
clear from Figure 3.22 that there are very low stresses in the central region o f the
base of the actual can, compared to the intersection region, and thus the amount of
94
the aluminium in this region can be reduced with no significant effect on the integrity
o f the can. This will be the subject o f the optimisation analysis discussed in Chapter
6.
3.3 Elastic-plastic finite element analysis using axisymmetric models
The finite element models and loading conditions considered here are the same as
those used for the elastic analysis discussed in 3.2.land 3.2.2.
3.3.1 Constant thickness model
The objective o f this model is to predict the maximum possible internal pressure the
pressurised can will withstand before plastic buckling (plastic snap-through) o f the
base occurs. This model was used initially to investigate the stress concentrations,
optimise the mesh density and provide a better understanding o f how to improve the
can base strength. The basic geometry is identical to that used in the elastic analysis
and shown in Figure 3.1. An identical mesh to that used for the elastic analysis and
seen in Figure 3.3 has been used here. The finite element mesh was generated
automatically using the ELFEN mesh generation. In the modelling o f aluminium
cans, the deformation o f the can base is non-linear. The material deforms plastically
and also the deformations are large enough to cause the loading direction and
stiffness to change throughout the analysis. This change in loading direction and
geometry-dependent stiffness is referred to as a geometric non-linearity (GNL) and
the GNL option within ELFEN was selected for these analyses. The non-linear finite
element analysis is achieved by incrementing the applied load in very small steps
allowing the stresses to be calculated at each load increment.
95
3.3.2 Material models
The multi-linear uni-axial stress-strain characteristic for the aluminium used in can
production is shown in Figure 3.25 and was derived by Patten [2] from experimental
results. Two types of finite element model have been used to represent this data:
180
160 -
140 •
120 ■
I 1“ -
60 ■
40 ■
P arallel spring plasticity
20 -
0 0.2 0.4 0 6 0.8 1 1 2 14 1.6 18

Strain (°«)
Figure 3.25: Parallel spring plasticity Vs m ulti-linear model
a) Elastic-perfectly-plastic (Figure 3.26)
Using this model, the von Mises equivalent stress cannot exceed yield stress ( cry =
100 MPa). Once yielding occurs across a section o f the tube, the finite element
procedure will no longer converge since the model predicts infinite strains and the
‘plastic collapse’ condition is reached. This model is extremely conservative and
under predicts stresses.
96
1
Stress
crv
strain
Figure 3.26: Stress-Strain relationship for an elastic-perfectly-plastic material
model
b) Bi- and multi-linear work hardening
In order to improve the accuracy of the model, a simple solution is to assume a bi
linear relationship. This is known as a parallel spring model [17] as can be seen in
Figure 3.27. The model will now predict stresses above the yield stress and the
material is said to work (or strain) harden. However, this model will not predict
collapse since the stress can continue to increase with increasing load.
The most accurate model is the one that uses a series o f straight lines to model the
true a -e behaviour up to the ultimate tensile strength o f the material (156 MPa). The
data for this multi-linear o -e curve is shown in Table 3.5.
97
Stress
M wyv-
Figure 3.27: Parallel-springs plasticity model
Plastic strain (%) Stress (MPa)

0 100
0.08 106
0.18 115
0.28 124
0.38 132
0.48 138
0.58 143
0.68 148
0.78 151
0.88 153
0.98 155
1.08 156
1.18 156
Table 3.5: Plastic stress-strain data for multi-linear material model [2]
3.3.3 Finite Element results for geometry G4
A typical geometry (i.e. Geometry 4) having t = 1mm is selected for a full review.
The pressure load was increased from 0 up to failure (collapse). The load was
incremented from an initial time factor o f 0.1 to a total stage time 1.0. A Newton-
Raphson iteration method [17] was used to perform an equilibrium check, to ensure
that the predicted results satisfy the underling differential equation.
98
3.3.3.1 Elastic-perfectly-plastic model
The results o f the analysis are shown as stress contours in Figures 3.28 to 3.31 for
pressure increments o f 0.22 MPa and starting at 1.0 MPa. Equivalent stresses slightly
higher than the uni-axial yield stress (100 MPa) are predicted due to the convergence
criteria within the ELFEN program. These figures indicate that major plastic zones
develop in the regions AA to BB and DD to FF (see Figure 3.2) and that failure
ultimately occur when a plastic hinge forms between DD and FF. From these
predictions, it was established that first yield and plastic hinge occur at pressures o f
0.72 and 1.60 MPa respectively.
These figures indicate the regions o f high stress and also the growth o f the plastic
zone. Yielding first occurs when p = 0.72 MPa, (Figure 3.28). As the pressure is
increased, four plastic zones are clearly seen to develop, at points labelled A, B, C &
D in Figure 3.2, for p = 1.2 MPa. A further increase in pressure to 1.4 MPa results in
a ‘plastic hinge’ where the whole o f Section XX (see Figure 3.30) has yielding
because o f the merger o f zones B, C, & D. The size o f zone A has also increased.
However, the pressure can be further increased to 1.6 MPa before final collapse
occurs (Figure 3.31) with further growth o f the plastic zones.
99
1 0 9 . 35 7 0
1 0 0 . 39 8 7
9 1 . 44 0 3 9
82 . 48209
73 . 52379
64.56549
55.60719
46.64889
37.69059
28.73230
1 9 . 77 4 0 0
1 0 . 81 5 7 0
1 . 85 7 3 9 8
Figure 3.28: Von Mises stress contour plot for G4 with internal pressure = 1.0
M Pa and EPP m aterial model
104.5157
95.99315
87.47057
78.94798
7 0 .4 2 5 3 8
61.90280
53.38021
44.85762
36.33503
27 . 8 1 2 4 4
19.28986
10. 76727
2.244680
M Pa and an EPP m aterial model
100
104.3394
95.68108
87.02274
78.36440
69.70607
61.04774
52.38940
43.73107
3S . 0 7 27 3
26.41440
17.75606
9.097729
0.439394
Figure 3.30: Von Mises stress contour plot for G4 with pressure = 1.4 M Pa and
an EPP m aterial model
108.2539
99 . 2 7 1 0 6
90.28825
81.30544
72.32262
63.33581
54.35701
45.37420
36.39138
27.40857
18.42576
9.442951
0.460140
M Pa and an EPP m aterial model
LIBRARY
101 ' ^
The corresponding equivalent stress distributions around the inside and outside
surfaces are shown in Figures 3.32 and 3.33 respectively. It can be seen that the
maximum stresses not exceed the yield stress when the pressure is 1.6 MPa. This is
to be expected since the equivalent stress (which cannot exceed the yield stress) is a
combination o f the three principal stresses, which can therefore individually be
greater that the yield stress.
jji
gj 0.80
tr
Uj
1.6 MPa
1.2 M P a
00 0.80 1.20 1.60 2.40 2.80 3.20 3.60 4.00 20
Distance
Figure 3.32: E quivalent stress distribution aro u n d the inside surface for
G4 and an EPP m aterial model
102
X l0 1
11.00
12 5.00
^ 4-B“
1.4 MPa
1 .2 MPa
0 .0 0 0.50 1.50 2.00 2.50 3.00 3.50 4.00 4 .5 0 x i a 1

A B Distance C D
Figure 3.33: Equivalent stress distribution aro u n d the outside surface
for G4 and EPP m aterial model
3.3.3.2 Multi-linear work-hardening model
The results of the elastic-plastic analysis using a work-hardening material model are
shown as equivalent stress contour plots in Figures 3.34 and 3.35 for pressures o f
1.50 MPa (just before collapse) and 1.59 MPa (at collapse) respectively. The plastic
hinge is clearly seen in Figure 3.34 to occur at the sharp radius close to the
intersection of base and cylinder. The shape after collapse is also clearly seen in
Figure 3.35. Very large deformation has taken place in the region o f the plastic
hinge, allowing the base to plastically buckle from a convex shape to a concave one.
103
■- 116.8156
I - 107.4243
I - 98.03309
— 88.64185
I - 79.25060
j - 69.85935
I - 60.46811
51.07686
41.68561
32.29436
22.90312
13.51187
4.120626
Figure 3.34: E quivalent stress contour plot (P = 1.50 M Pa) for a multi- linear
hardening m aterial model
104
137.3391
126.2598
115.1805
104.1012
93.02189
81.94259
70.86329
- 59.78399
- 48.70469
- 37.62540
A
- 26.54610
15.46680
4.387500
Figure 3.35: Equivalent stress contour plot (collapse, p =1.59 M Pa)
3.3.4 Effects of wall thickness
The above analyses have been repeated for the other thickness values, using the
multi-linear hardening model. The EPP model was not considered because it is not
realistic and was only included for t = 1 mm for illustration. A summary o f the
results for the range of wall thickness is presented in Table 3.6 and Figure 3.36.
Whereas the variation in limiting pressure is reasonably linear, the curve for collapse
pressure shows a clear increase in slope with increasing thickness. This is important
for material optimisation, as discussed in Chapter 6.
105
W all thickness (mm) Lim iting pressure (MPa,) Collapse pressure
(M Pa)
0.4 (G l) 0.16 0.73
0.6 (G2) 0.22 0.93
0.8 (G3) 0.40 1.18
1.0 (G4) 0.72 1.59
1.2 (G5) 1.17 2.37
1.4 (G6) 1.49 3.56
Table 3.6: F irst yield and collapse pressures for different wall thickness
3. 5
2.5
0. 5
0.2 0.4 0.6

Thick ness(mm)
Figure 3.36: The relationship between wall thickness first yield and collapse
pressure
The geometry of the finite element model has previously been described in Section
3.2.5 and the finite element model is shown in Figure 3.20. The equivalent stress
distribution just prior to collapse (at p = 1.50 MPa) and at collapse (p = 1.53 MPa)
using the multi-linear work hardening material model previously described in
106
Section 3.3.2 are shown in Figures 3.37 and 3.38 respectively, superimposed on the
displaced shape.
119.5338
110.2195
100.9051
91.59081
82.27647
72.96214
63.64780
54.33347
45.01913
35.70480
26.39047
17.07613
7.761796
Figure 3.37 Equivalent stress contour plot (pre-buckling, pressure =1.50 M Pa)
128 . 94 3 5
118.7351
108 .5 2 6 7
98.31827
88.10986
77.90146
67.69305
57.48465
47.27625
37.06784
26.85944
16.65104
6.442631
107
128.9435
118.7351
108.5267
98.31827
88.10986
77.90146
67.69305
57.48465
47.27625
37.06784
26.85944
16.65104
6.442631
Figure 3.38: equivalent stress contour plot (collapse, pressure =1.53 M Pa
As for the constant thickness models, regions o f high stress are apparent in the DD to
EE region of the base (see Figure 3.2). Also, the stress levels close to the axis of
symmetry o f the base are generally low because this region is significantly thicker
than elsewhere. It is considered that shape optimisation could result in a significant
reduction in material, while still retaining the plastic buckling and collapse
characteristic of the can. This has been investigated in Chapter 6.
The internal pressure is a function of volume therefore any large deformation will
cause the pressure inside the can to reduce.
108
This cannot easily be modelled, therefore the assumption is made that the pressure in
the pressure test is increased very slowly such that the water pump will prevent a
reduction in pressure due to increased volume.
3.4 Elastic-plastic finite element analysis using 3D models
The analyses discussed previously in this chapter, although useful in studying the
mechanisms involved and the accuracy o f the upper and lower bound estimates is not
truly representative in one important respect. Experimental evidence suggests a
slightly unsymmetrical buckling mode, due to minor radial variations in profile and
there is a clear distinction between the elastic-plastic buckling o f the base and burst
(collapse) pressures, where bursting occurs in the plain tube region. This behaviour
cannot be predicted using an axisymmetric model. Therefore a full three-dimensional
model was developed and elastic-plastic buckling o f the base replicated by the
introduction o f a small imperfection in a similar way to that reported by Robotham e t
a l [25] for plain shafts in torsion.
3.4.1 Finite element model
The basic cross-section shown in Figure 3.20 using 6 super-elements has been used
to create a three-dimensional model as shown in Figure 3.39 (half model shown for
clarity). The boundary conditions are shown in Figure 3.40. The model was
constrained along its line o f symmetry in the X direction see Figure 3.39 (plane
ABCD). This does not allow X displacement o f these elements, to model the can as
symmetrical. The top section o f the can was constrained in the Z direction (plane
ADE) to simulate the gripping o f the can in the pressure testing equipment. In reality
the can is gripped at the shoulder during the pressure tests not in the midsection as in
109
the model. Again, due to the large deformations, it was necessary to use a geometric
non-linear analysis since the loading will change direction during the buckling
process and the stiffness o f the base changes significantly.
Y-
Figure 3.39: 3D finite elem ent model geom etry
//
Figure 3.40: Finite Elem ent C onstraints
110
3.4.2 Material model and loading
The multi-linear material model for aluminium 1050, shown in Table 3.5, has been
used. An incremental uniform pressure load was applied to the internal surface o f the
can. The mesh made up o f 6315 four-noded three-dimensional elements, the finite
element mesh is shown in Figure 3.41.
Figure 3.41: 3-D Finite elem ent model mesh.
3.4.3 Eigenvalue analysis
From a preliminary eigenvalue analysis (i.e. lowest mode) and supported by
experimental evidence, a small perturbation was introduced into the geometry to
prevent a symmetrical deformation mode giving an increased load prediction at the
limit point. This was achieved by increasing the radial coordinates o f the nodes lying
on one side of the half-model cutting plane from the centre to the edge o f the base by
0.1mm (-10% o f the wall thickness at that point). This provided a bifurcation point
and enabled the snap-through buckling mode to be investigated. Robotham e t a l [24]
showed that imperfections in the range 1 to 10% produced very similar results. In the
111
snap-through model, the load/displacement function will be cubic having a maximum
at the point of buckling and a minimum after complete snap-through is achieved.
3.4.4 Results
The analysis resulted in stress contours plots for a number o f incremental pressures.
Finite element predictions of yield and elastic-plastic buckling pressures were
predicted. Unlike the axisymmetric model, this model is able to resist a further
increase in pressure, prior to collapse and the predicted collapse pressure is 2.02
MPa.
It can be seen from Figure 3.42 that the stresses are high enough such that
The can base yields when the internal pressure is 1.50 MPa since the yield stress for
the aluminium is 100 MPa.
100. 6710
94. 68992
88. 70883
82. 72775
76. 74667
70.76558
64. 78450
58.80341
52.82233
46.84124
40.86016
34.87907
28.89799
Figure 3.42: Von Mises Stress C ontour Plot at internal pressure of 1.50 M Pa
When the internal pressure is increased to 1.70 MPa Figure 3.43 shows that the
deformation due to this pressure is clearly unsymmetrical, since it is not possible for
an object to be perfectly symmetrical and will ultimately enable plastic collapse
(snap-through) to occur at a pressure o f 2.02 MPa, as shown in Figure 3.44, this
112
shows that the finite element analysis predicts that the aerosol can base will be fully
deformed at 2.02 MPa and that the stresses are now concentrated in the lower section
of the can walls.
120 . 5 5 9 8
113.4004
106.4011
99.32178
92.24245
85.16312
78 . 0 0 3 8 0
71.00447
63.92515
56.84582
49.76649
42.68717
35.60785
Figure 3.43: Von Mises stress contour plot at internal pressure of 1.70 M Pa
before snap-through
122 . 1 9 0 4
117 . 3 0 7 7
112.4250
107 . 5 4 2 3
102.6597
97 . 7 7 7 0 0
92.89433
■ a .01166
83 . 1 2 8 9 9
78.24632
7 3 . 3 6 3 65
68 . 4 8 0 9 7
63 . 5 9 8 3 0
122.1904
117.3077
h 112.4250
- 107.5423
102.6597
h 97.77700
-
-
-
92.89433
88.01166
83 . 1 2 8 9 9
78.24632
73.36365
68 . 4 8 0 9 7
mmsm
63 . 5 9 8 3 0
Figure 3.44: Final von Mises Stress Prediction at pressure of 2.02 M Pa
113
For further investigation and information, two 3-D constant thickness models, for t =
0.6 mm and 1.0 mm were created and analysed, using the same approach but without
the nodal perturbations to produce the geometrical asymmetry. It can be seen from
Figures 3.45 and 3.46 that the plastic collapse (snap-through) occurs at a pressure of
0.83 MPa and 1.20 MPa for t = 0.6 and 1.0 mm respectively. These results show that
the deformations due to this pressure are clearly symmetrical. This confirms the need
for the original asymmetry in order to generate a realistic response.
106-1584
93 - 53969
85-42101
77-30233
69-18365
61-06497
52-94629
44-82761
- 36-70892
- 28-59024
- 20-47156
12-35288
42 34199
Figure 3.45: Von Mises stress contour plot at in tern al pressure of 0.83 M Pa and
0.6 mm constant thickness
114
i— 1 1 6 - 5 0 0 5
107 - 1033
97-70600
88-30873
78 - 9 1 1 4 6
69-51419
60-11692
50-71965
41-32238
31-92511
225 . 278 4
131.3058
37.33307
|bu
116-5005
107 - 1033
97-70600
88-30873
78 - 9 1146
69-51419
60-11692
50-71965
41-32238
31-92511
225.2784
131.3058
37. 33307
r
Figure 3.46: Von Mises stress contour plot at internal pressure of 1.20 M Pa and
1.0 mm constant thickness
3.5 Upper and lower bound pressures
In this section, the elastic compensation method proposed by Mackenzie and Boyle
and discussed in Chapter 2 is used to estimate the upper and lower bound limit
(collapse) loads for the one-piece aluminium aerosol cans subjected to internal
pressure loading. As in Sections 3.2 and 3.3, the wall o f the can is initially assumed
115
to be o f constant thickness and results for six thickness values are presented. A
realistic thickness profile is also used in a seventh model. Upper and lower bound
pressures are found using axisymmetric models.
3.5.1 Material models, loading and boundary conditions
Since the analyses are elastic, only values for Y oung’s modulus and Poisson’s ratio
o f 68.3 GPa (zeroth iteration) and 0.33 respectively (as before) are required. The
loading and boundary conditions are as discussed in Section 3.2.2.
3.5.2 Constant thickness model
The basic finite element model, made up o f six ‘super elements’ and shown in Figure
3.2 for a can section that has a constant thickness o f 1 mm, has been used. Since the
methodology involves an iterative finite element procedure, it was necessary to
choose a mesh that meets both the condition o f convergence and that o f economy o f
the solution. A preliminary investigation, starting with a mesh o f 296 elements (four
through-thickness), was undertaken in order to establish a suitable mesh for which
mesh convergence had been reached. For the elastic compensation method analysis,
296 8-noded, axisymmetric elements were generated manually from the basic mesh
in Figure 3.2 and the mesh for this analysis is shown in Figure 3.47.
116
Figure 3.47: simple finite element mesh for geom etry G4
3.5.2.1 Geometry G4 (t = 1 mm)
Figure 3.48 shows the von-Mises equivalent stress contour plot for the initial elastic
solution (i.e. zero-th iteration in the elastic compensation method) for an arbitrary
pressure of 0.1 MPa. Regions of above-average stress occur in the transition region
between cylinder and base and at the base centre. On the basis o f the results shown in
Figure 3.48, an internal pressure at which first yield occurs was found to be 0.75
MPa.
117
13.25290
12.16671
11.08051
9 .994317
8.908122
7 .821928
6.735733
5.649539
4.563344
3.477150
2 .390955
1.304761
0.218566
Figure 3.48: Equivalent stress contour plot for iteration 0
The iterative procedure described in Section 2.9 has then been employed (with the
aid of a FORTRAN program) and the modulus o f elasticity in each element modified
according to Equation (2.25). The maximum equivalent stress at the end o f the each
subsequent iteration is shown in Figure. 3.49, from which it is clear that a converged
solution occurs after 4 iterations with ad = 10.72 MPa. The elastic compensation
method may, depending on the function used, caused the maximum stress to increase
or decrease, but by careful selection o f the function it is generally found that over a
number of iterations there will be a net decrease in maximum stress with respect to
the initial elastic solution.
118
’s matx
Q.5 2.5 3.5

Iteration
Figure 3.49: Maximum equivalent stress at the end o f each iteration for t = 1
mm
The steady-state (converged) equivalent stress contour plot is shown in Figure 3.50.
A redistribution o f stress has occurred with an initial stress range o f 0.21 - 13.25
MPa (see Fig. 3.48) reducing to 0.02 - 10.72 MPa. It is also apparent that the stress
discontinuities at element boundaries have become more pronounced since the values
o f elastic modulus can now significantly vary from element to element.
3.5.2.2 Method of implementation of elastic compensation method
The procedure used in this approach is as follows:
(a) zero-th iteration. The initial elastic analysis is carried out with an arbitrary
pressure, Pd, using E 0 throughout.
(i) store the elastic stress field, cre.
119
(ii) identify the maximum stress in each element and use them to update
elemental E values, using Equation (2.25)
(iii) identify the maximum stress in the model, <rd
(iv) re-create the finite element program input data file, using the new E
values
(b) i th iteration
(i) - (iv) as above
(v) compare <rd with the value from the previous iteration (i.e. for
convergence)
(c) converged solution. This occurs when Gd becomes constant
(i) calculate PL using Equation (2.26)
(ii) obtain U d and D d from the finite element program output (see note
below)
(iii) calculate Pv using Equation (2.31).
Note that:
1. Strain energy values are obtained directly from the finite element program output
file. The dissipation energy for each element is obtained from the three principal
strains, the yield stress and the element volume, using a version o f Equation (2.30)
based on total values, not rates. A simple FORTRAN program was therefore written
to perform this calculation.
120
2. The procedure in (i) to (iv) above is time consuming and prone to error, when
performed manually. A FORTRAN program was written to perform these tasks
automatically.
10.71782
9.826537
8 .935252
8 .043966
7 .152681
6.261396
5.370111
4.478825
3.587539
2.696254
1.804968
0.913683
0 .0223976
Figure 3.50: Steady state equivalent stress contour plot for t = 1 mm
From Equation (2.26), it follows that:
<T 100
P l = — Pd = — xO.l =0.93 MPa
o\, 10.72
In order to obtain an upper bound estimate, values o f dissipation energy and strain
energy, for the converged solution, are required. A FORTRAN program was written
to extract the stress and strain predictions from ELFEN and from which the
121
dissipation energy was derived, using the method described in Section 2.9. Having
done this and using Equation (2.30) and (2.31):
D. 0.001405
x 0.1 = 2 .2 0 MPa
0.00006384
3.5.2.3 Effects of wall thickness
This process was repeated for constant thickness models o f 0.4, 0.6, 0.8, 1.2 and 1.4
mm, using 296 elements in each case. The resulting upper and lower bound pressures
are summarised in Table 3.7. A comparison between the upper and lower bounds
pressures and the yield and collapse pressures are presented in Table 3.8 and Figure
3.51.
The results presented in Figure 3.51 show that the FE predicted collapse pressures lie
between the upper and lower bound estimates, closer to the upper bound, and this
provides a degree o f confidence in these approximate methods. However, the range
between the upper and lower bounds is large and, furthermore, the lower bound is
always greater than the yield stress. This limits the use o f these approximate methods
for this type o f geometry and loading to collapse pressure estimates. Nevertheless,
the elastic compensation method is useful since it only requires elastic analyses.
122
C om pensation m ethod
t Pl Pu
(mm) (MPa) (MPa)
0.4 0.52 1.16
0.6 0.61 1.37
0.8 0.72 1.66
1.0 0.93 2.20
1.2 1.64 2.72
1.4 2.65 3.95
Variable 0.81 2.59
Table 3.7: U pper and lower bound pressures using elastic com pensation method
Com pensation m ethod Finite elem ent

t (mm) P7 (MPa) PM(MPa) Py(MPa) ? c (MPa)
0.4 0.52 1.16 0.16 0.73

0.6 0.61 1.37 0.22 0.93
0.8 0.72 1.66 0.40 1.18
1.0 0.93 2.20 0.72 1.59
1.2 1.64 2.72 1.17 2.37
1.4 2.65 3.95 1.49 3.56
Variable 0.81 2.59 0.38 1.53
Table 3.8: Results of elastic com pensation and finite elem ent analyses
4 .5
3.5
CL.
2 2.5
0.5
0 0.2 0.4 0.6 08 1.2 14 1.6

T hicness(m m )
Figure 3.51: C om parison of finite elem ent m ethod and com pensation m ethod
123
A finite element mesh containing 296 8-noded, axisymmetric elements was
generated manually from the basic mesh shown in Figure 3.21. The iterative
procedure previously described for constant thickness models was repeated using this
variable thickness model. A steady-state maximum equivalent stress o f 12.4 MPa
was predicted and from Equation (2.26):
P l = — Pd - 7J 7 0.1 =0.81 MPa

co 12.4
Values for the steady state dissipation and strain energies were obtained using the
procedure described above and using Equation (2.30) and (2.31):
Dd 0.001712 ocniV/m
p TI = — p A = --------------- 0.1 = 2.59 MPa
Ud 0.0000659
The upper pressure bound estimate o f 2.59 MPa is higher than the predicted yield
pressure o f 0.38 MPa (Section 3.2.5) and higher than the elastic-plastic buckling
pressure prediction o f 1.53 M Pa (Section 3.3.5). Therefore, the upper bound estimate
has good application. In this variable thickness example, however, the lower bound
estimate of 0.81 MPa is higher than the yield pressure.
By comparing variable thickness results with those for constant thickness models, it
is apparent that the upper and lower bound estimates for the variable thickness model
fall between the 0.7 to 1.0 mm constant thickness results, which might be considered
to be reasonable since the region with the highest stress concentration and where
buckling ultimately occurs (i.e. the comer between cylinder and base) has a thickness
124
varying between 0.7 and 1.3 mm. However, the upper bound pressure estimate
exceeds the collapse pressure.
3.6 Experimental testing
Experimental pressure testing of cans having various dimensions has previously been
carried out by Patten and full details o f the test procedure, the test equipment and
results can be found in [2]. A typical burst can is shown in Figure 3.52, which also
shows the buckling o f the base, prior to failure. The non-symmetric nature o f the
deformed shape is clear and comparable with finite element predictions (see Figure
3.44). There is a requirement that the buckling pressure is at least 20% below the
actual burst pressure. In this practical situation, minimum burst pressures are
specified by the customers.
125
Figure 3.52: Deform ation and b u rst pressure of can base
A typical pressure-time curve, taken from [2], is shown in Figure 3.53 and a
summary o f Patten’s findings are presented in Table 3.9. The results for the 53 mm
can (which has been modelled here) shown very good agreement between the
experimental burst pressure obtained by Patten and the analytical solution and finite
element predictions obtained here. Similarly, the experimental buckling pressure o f
1.6 MPa compares favourably with the finite element prediction o f 1.7 MPa.
D iam eter W all thickness Average actual Predicted b u rst

(mm) (mm) b u rst pressure pressure(bar)
(bar)
38 0.32 28 29.6
44 0.30 24 23.4
50 0.33 23 23.1
53 0.30 21 21.0
59 0.35 21 20.8
59 0.41 24 24.4
60 0.44 24 23.4
Table 3.9: Com parison between m easured and predicted b u rst pressures of cans
[2]
126
26 i
Burst Pressure
24
22
20
Deformation Pressure
i 16
T 14
3 12
0 50 100 150 200 250 300 350 400 450 500 550 600 650 700 750 800
Tim e
Figure 3.53: Deformation and b u rst pressure [2]
3.7 Closure
This chapter has described the elastic and elastic-plastic analysis o f the thin
cylindrical component under internal pressure loading. Initially, axisymmetric
constant-thickness models have been used to investigate the stress distributions that
are set up, the yield pressures and the way in which the plastic zones develop, after
yielding, leading up to elastic-plastic buckling. In addition, a realistic thickness
profile has been modelled in order to more accurately study the pre- and post-yield
characteristics. Emphasis has been placed on the base o f the cylindrical can, since
this is where the major deformation occurs.
However, the axisymmetric models are not capable of distinguishing between the
elastic-plastic buckling of the base and the ultimate bursting o f the can. In fact, these
two events are predicted to be coincident, whereas experimental evidence suggests a
127
slightly unsymmetrical buckling mode and a clear distinction between the elastic-
plastic buckling o f the base and burst (collapse) pressures. A three-dimensional half
model was developed in order to investigate the elastic-plastic buckling o f the base.
Finite element predictions o f yield, elastic-plastic buckling and collapse (burst)
pressures have been compared with experimental evidence and analytical solutions
and there is generally good agreement between them. Reasons for any discrepancies
will be discussed in Chapter 7.
Finally, the elastic compensation method has been adapted in order to estimate upper
and lower bounds on pressure for this type o f geometry and loading conditions. The
method has been found to be o f limited use since the lower bound pressure is
generally higher than the yield pressure.
In Chapter 4, the elastic-plastic behaviour o f these cylinders, when subjected to axial
compressive loading, will be investigated.
128
Chapter Four
AXIAL LOADING
4.1 Introduction
This chapter describes the analysis o f a thin-walled cylinder with inverted base
(typically used as aerosol cans) under axial compressive loading in order to predict
the limit and failure loads for this type o f loading. Such an analysis is important
because aerosol cans are subjected to axial compressive loading:
a) when the neck is formed during the manufacturing process
b) when the valves are inserted and the vessel is charged
as shown in Figure 4.1. Elastic and elastic-plastic finite element analysis is used to
predict the buckling behaviour and results are compared with those obtained from
experimental testing, which is also described in this chapter. Case (a) is discussed in
Section 4.3 and Case (b) in Section 4.4.
129
Axial Load
Ar 4 ^
Figure 4.1: Axial compressive loading on the can
4.2 Potential failure modes
The straight section is basically a thin-walled cylinder and it is thought that the base
has little or no effect on the buckling process. Most o f the previous research has been
conducted on plain cylinders and is directly applicable to cans. The collapse o f a
cylinder under an axial load may occur in a number o f ways [21]. If the cylinder is
slender (i.e. if the height to radius ratio is sufficiently large) then it will fail in a long
wave bending mode over its entire length (see Figure 4.2(a)). If the cylinder is
moderately long with sufficiently thick walls, failure occurs plastically with an
axisymmetric ‘diamond type’ buckling mode (see Figure 4.2(b)). For short cylinders
with adequately thin walls, failure occurs elastically with an axisymmetric ‘ring type’
buckling mode as shown in Figure 4.2(c).

(a) Long cylinder (fc>) Moderately long cylinder (c) Very short cylinder
Figure 4.2: Effect of cylinder length on the buckling modes [21]
4.3 Case (a) - Compressive behaviour during neck formation
At the very start o f the neck formation process, it can be assumed that the loading is
applied to the rim, as shown in Figure 4.1(a). At this point, the overall length is at a
maximum and the cylinder will fail at the lowest axial load.
As the load is increased there comes a point at which the deformation mode suddenly
bifurcates into a pattern running around the circumference o f the vessel (see Section
2.8.1) and the deformation o f the buckle pattern is near the cylinder base.
4.3.1 Geometry and finite element model
The finite element model was produced in three dimensions by rotation o f the cross
section geometry shown in Figure 4.3 through 360° about the Y-axis. The cross
section is based on measurements made by Patten [2] and so provides a realistic
131
model o f the cylinder thickness profile. As with the dome reversed failure mode
described in Chapter 3 the model needs to be ‘seeded’ with an imperfection in order
to reproduce the correct buckling behaviour. A small perturbation in the shape o f a
known buckling pattern to the geometry produces the corresponding failure mode
upon collapse. The cylinder geometry was modified by shifting the radial positions
o f the nodes as one proceeds around the circumference by 0.2 mm (see Section 2.8.4
for more information). The resulting three-dimensional shape is shown in Figure 4.4.
At the rim o f the cylinder, displacement is permitted along the axis but the rim is
restrained in the X and Z directions. The bottom o f the model is completely
constrained to maintain a circular cross-section, as shown in Figure 4.5.
A face loading is applied normal to the horizontal rim surface o f the vessel as shown
in Figure 4.6 to model axial loading. The loading is ramped up linearly with time.
The user need only specify the loading rate since all other loading data for the
analysis is generated automatically.
4.3.3 Material model
The material data described in Section 3.3.2 including the data for the multi-linear o -
e curve, as shown in Table 3.5, was used for this analysis. A finite element mesh o f
7488 rectangular 4-noded shell elements and one element through the thickness was
generated automatically using the ELFEN mesh generator and the mesh study result
is shown in Figure 4.7.
132
4.3.4 Finite element predictions
Predictions have been obtained using the elastic and elastic-plastic analysis facilities
within ELFEN [5]. A geometric non-linear (GNL) analysis was performed since
large deformations and strains, which can have a significant effect on the load-
deflection characteristics o f the component, were anticipated. GNL considerations
may have an influence on both the static and dynamic behaviour o f structures [17].
Also in snap-through buckling, deflections o f the structure are large compared with
the original dimension o f the structure. Changes in stiffness and load occur as the
structure deforms. Geometry non-linearity occurs when the change in the geometry
o f the structure due to its displacement under load are taken into account in analysing
its behaviour. The equilibrium equations take into account the deformed shape,
whereas in small strain analysis the equilibrium equations are based on the original
un-deformed shape.
The von Mises equivalent stress contour plot for a pre-buckling (elastic) face load o f
-29.3N/mm2 and with a total load o f 3247 N is shown in Figure 4.8. It is clear that
the highest stresses occur close to the base o f the cylinder. The corresponding von
Mises equivalent stress contour plot and deformed shape at buckling are shown in
Figures 4.9 and 4.10 respectively. It can be seen form these figures that the buckling
o f the cylinder is occur near the base and the mode shape has seven modes.
The load is increased and the load-displacement curve for the rim o f the cylinder is
shown in Figure 4.11. A reasonably linear response is seen up to a load o f -2500 N
and above this, the buckling process begins with failure occurring at a load o f 3247
N. The three stages o f pre-buckling, buckling and post-buckling are clearly seen.
133
Figure 4.3: Cross-section geometry for the analysis of Case (a) axial loading
during neck form ing
134
Figure 4.4: Three-dim ensional model for Case (a) axial loading analysis
135
.J * '
Figure 4.5: Structural constraints for Case (a) axial loading analysis
136
—-F -a c e
--Race
Figure 4.6: Applied loading for Case (a) axial loading analysis
137
Figure 4.7: Finite elem ent mesh for Case (a) axial loading analysis
138
155.2412
142.3227
129.4043
116.4859
103.5674
90.64899
77 . 7 3 0 5 6
64.81212
51.89368
38.97525
26.05681
13.13837
0.219934
Figure 4.8: Pre-buckling equivalent stress contour plot for Case (a) axial
loading analysis
139
Figure 4.9: E quivalent stress contour plot at the point of buckling for Case (a)
axial loading analysis
140
Figure 4.10: Buckling mode shape for Case (a) axial loading analysis
B uckling point
350.0
3000
O)
2500
O
2000 O)
-O
as
o 1500 Q_ o
_i
1000 CO
o
Q_
500
0 0.2 0.4 0.6 0.8 1 1. 2

Displacement (m m )
Figure 4.11: Predicted rim load-displacem ent curve for Case (a) axial loading
analysis
141
4.3.5 Experimental testing
Experimental testing on a can with outer diameter 53 mm, inner diameter 52.4 mm,
wall thickness 0.315 mm and length 125 mm (see Figure 4.12) has been carried out
in order to validate the finite element prediction in Section 4.3.4.
A Zwick 20 kN electrically driven tensile test machine has been employed with a
compressive load being applied, as shown in Figure 4.13. A compressive load was
applied to the rim o f the cylinder using a steel insert which was fixed to the
uppermost part o f the cylinder using a standard jubilee clip (see Figure 4.14).
During the test the crosshead movement and the applied load, using a 15000 N load
cell, were logged and a typical resulting load-displacement curve is shown in Figure
4.15. A maximum load o f 3230 N was noted at a rim displacement o f -0.86 mm,
after which the load began to reduce until reach 2800 N approximately. The resulting
deformed cylinder is pictured in Figure 4.16. The test was then repeated several
times and very similar results were obtained.
In one case, the test was continued well beyond the point o f first buckling and the
resulting load-displacement curve is shown in Figure 4.17. It is clear that once a
significant change in shape has occurred (i.e. for a displacement o f approximately 3
mm) the can stiffness starts to increase and an increase in load is seen up to
approximately 2000 N for a total displacement o f -5.5 mm. After this, a second
buckling mechanism is observed with the load decreasing until the test was stopped
when the rim displacement was approximately 6.8 mm. The final deformed shape is
shown in Figure 4.18. It can be seen from the figure that the buckling occurs near
the base and with a seven-lobed collapse pattern.
142
Figure 4.12: Alum inium aerosol can used in experim ental testing for Case (a)
143
Figure 4.13: Zwick 20 kN tensile test m achine
144
Figure 4.14: Steel insert and jubilee clip arrangem ent used in experim ental
testing for Case (a) axial loading analysis
145
□
Ui
3
CD
03
—i
■o
CD
cr
c
n
^ | Buckling point
Ol
u
CJ
Load in N
CJ P o stb u cklin g
Q — I— I— I— I— I— I— h-H — I— h " I— I— I— h
0.0 0.5 1.0 1.5

displacement in mm
Figure 4.15: Experimental rim load-displacement curve for Case (a) axial
loading analysis
146
Figure 4.16: Buckled can for Case (a) axial loading analysis
147
3000 -
Load in N
2000 ~
1000 -■
0 2 4 6
Displacement in mm
Figure 4.17: Extended experim ental rim load-displacem ent curve for Case (a)
148
Figure 4.18: Buckled can for Case (a) axial loading analysis
4.3.6 Analytical solution
For constant thickness thin walled tubes, the maximum buckling force can be
calculated from the following formula [9]:
F , = j 2^ 2 -(4.1)
13(1- o 2)
Using this equation and assuming the aluminium material properties in Table 3.2 and
a constant thickness of the 0.315 mm, the theoretical maximum buckling force is
26,043 N.
149
4.3.7 Comparison and discussion of results
4.3.7.1 Load-displacement characteristics
By comparing Figure 4.11 with Figure 4.15, it can be seen that the shape o f the
experimental load-deflection curve is very similar to that predicted by finite element
analysis. Furthermore, the predicted elastic-plastic buckling load o f 3247 N is less
than 1% greater than the experimentally observed buckling load o f 3230 N. Also, the
predicted buckling displacement o f -0.8 mm is similar to the observed value of
-0.86 mm. There is excellent correlation between the two characteristics.
However the analytical solution, which is for the elastic buckling o f a plain open
cylinder with a constant wall thickness o f 0.315 mm (based on experimental
measurements [Patten]), is 26,043 N. This higher value is to be expected as there is
clearly a stress concentration at the base o f the actual cylinder (see Figure 4.8) which
acts as the catalyst and causes elastic-plastic buckling at a load far less than that
estimated for elastic buckling o f the corresponding plain open cylinder.
4.3.7.2 Buckling mode shape
As the load increases there comes a point at which the collapse mode suddenly
initiates. A linear buckling analysis indicates that for the lowest modes obtained from
an eigenvalue analysis, the buckling mode for the open cylinder is a ‘diamond type’
inward and outward deformation see Figure 4.2(b) [21]. The predicted collapse mode
o f the cylinder with inverted base is shown in Figure 4.10 and the corresponding
experimental results are shown in Figure 4.18.
150
In both cases, the buckling modes are very similar but buckling occurs close to the
base, unlike that shown in Figure 4.2(b), with a seven-lobed collapse pattern.
4.4 Case (b)-Compressive behaviour during valve insertion and
charging
During valve insertion and charging o f an aerosol can, the loading is applied to the
inner rim, as shown in Figure 4.1(b). As the load is increased, it is anticipated that
the top will act as a belleville spring (washer) and will ‘flatten’ under load. Finite
element predictions o f this behaviour are compared with experimental results from
tests on an aluminium aerosol can.
The finite element model was produced in three dimensions by rotation o f the cross
section geometry shown in Figure 4.19 through 360° about the Y-axis. The cross
section is based on measurements made by Patten [2] and so provides a realistic
model o f the cylinder thickness profile. The resulting three-dimensional shape is
shown in Figure 4.20.
The nodes at bottom o f the model are completely constrained to maintain a circular
cross-section, as is see in Figure 4.21.
151
A face loading is applied normal to the rim o f the vessel as shown in Figure 4.22.
This load was incremented using the arc length method such that the pressure load
was increased from zero up to failure. The load was incremented from an initial time
factor o f 0.01 to a total stage time o f 1. For the non-linear solution o f the problem, an
arc load incrementing method was used to increase the applied load such that the
maximum load could be found. A Newton-Raphson iteration method was used to
perform an equilibrium check. A residual level o f 0.1 was specified which would
give sufficient accuracy for the analysis.
The material data described in Section 3.3.2 including the data for the multi-linear o -
€ curve, as shown in Table 3.5, was used for this analysis. A finite element mesh o f
8548 triangular 4-noded shell elements and one element through the thickness was
generated automatically using the ELFEN mesh generator and the mesh is shown in
Figure 4.23.
152
Figure 4.19: Cross-section geometry for the analysis of Case (b) axial loading
during valve insertion and charging
153
Figure 4.20: Three-dim ensional model for Case (b) axial loading analysis
154
B
Figure 4.21: S tructural constraints for Case (b) axial loading analysis
155
y
Figure 4.22: Applied loading for Case (b) axial loading analysis
156
Figure 4.23: Finite elem ent mesh for Case (b) axial loading analysis
4.4.4 Finite element predictions
Again, elastic and elastic-plastic GNL analyses have been performed using ELFEN
[5]. The von Mises equivalent stress contour plot for a pre-buckling (elastic) face
load o f - 4.5 N / m m 2 and with a total load o f 1617 N is shown in Figure 4.24. In this
case, the highest stresses occur at the rim and at the intersection o f the top and
157
parallel sections of the cylinder. The load is increased and the load-displacement
curve for the rim of the cylinder is shown in Figure 4.25. A reasonably linear
response is seen up to a load of 1617 N and above this, the top buckles and a
reduction in load is clear. The corresponding von Mises equivalent stress contour
plot and deformed shape at buckling are shown in Figures 4.26 and 4.27
respectively. It can be shown from the figures that the buckling occur in the ring top
until reached the shoulder of the can.
155.9000
143.0347
130.1695
117.3043
104.4390
91.57378
78 . 7 0 8 5 4
65.84329
52.97805
40.11281
27 . 2 4 7 5 7
14.38233
—
1.517082
Figure 4.24: Pre-buckling equivalent stress contour plot for Case (b) axial
loading analysis
158
1800
1600
1400
1200
•ooo
800
600
400
200
0
0 0.5 1.5 2 2.5 3 3.5
D isplacem ent in mm
Figure 4.25: Predicted rim load-displacement curve for Case (b) axial loading
analysis
159
' >>* V rlV -’ <2S ? ;
Figure 4.26: Exaggerated equivalent stress contour plot at the point of buckling
for Case (b) axial loading analysis
160
Figure 4.27: Deformed shape for Case (b) axial loading analysis
4.4.5 Experimental testing
Experimental testing on a can with outer diameter 53 mm, inner diameter 52.4 mm,
wall thickness 0.315 mm (see Figure 4.28) has been carried out in order to validate
the finite element predictions in Section 4.4.4. Again, the Zwick 20 kN electrically
driven tensile test machine was used with a compressive load being applied.
161
A typical resulting load-displacement curve is shown in Figure 4.29. As the load is
increased, the can is compressed until a maximum load of 1650 N is reached. Over
the first 2 mm of the displacement, the load appears to increase linearly with
displacement then the slope increases sharply until the point of buckling, with
reducing load, is reached for displacement of approximately 2.9 mm. A second stage
of buckling appears to start when the load is approximately 700 N. The load may
increase again when the necked section is completely crumpled as can be seen in
Figure 4.30. For this size of can, the minimum axial load that must be supported, as
required by the customer specifications, is 1180 N [2] This suggests that it may be
possible to make the top o f the can thinner, therefore leading to further saving in
material usage.
A number of compression test were carried out to investigate the axial loading that
the aluminium cans are able to support.
4.4.6 Analytical solution
For a simple larg cylinder under purly compressive axial loading, the buckling load
can be calculated from the following formula [55]:
P = - ( 4 . 2 )
The second moment of area for a tube section is given by:
162
I = j(r. 2 - r, 2 ) (4.3)
Also, the maximum longitudinal compressive load may be calculated from:
F = UTS * A WJs ...(4.4)
the predicted maximum compressive load based on Equations 4.2 to 4.4 and actual
failure loads for the cans is given in Table 4.1 below.
Wall Actual Predicted

thickness failure maximum
Diameter(mm) Length(mm)
(mm) load(N) compressive
load(N)
53 110 0.41 1355 10644.2
53 125 0.315 1617 8177.8
Table 4.1: Comparison of actual failure load to buckling and compressive
models
Table 4.1 shows that the cans are too short for failure to be caused by buckling and
also, the failure is not caused by compressive stress in the can walls. Inspection of
the aerosol cans after axial testing show that the compressive failure was
concentrated on the can shoulder, as shown in finite element analysis.
It has been shown that the simple stress analysis equations cannot be used to model
the deformation of the can base or axial loading.
163
Figure 4.28: Aluminium aerosol can used in experimental testing for Case (b)
164
~o
03
O
4
displacement in mm
Figure 4.29: Experimental rim load-displacement curve for Case (b) axial
loading analysis
165
Figure 4.30: Buckled can for Case (b) axial loading analysis
166
4.4.7 Comparison and discussion of results
4.4.7.1 Load-displacement characteristics
By comparing Figure 4.25 with Figure 4.29, it can be seen that the shape of the
experimental load-deflection curve is slightly different to that predicted by the finite
element analysis in the first 2 mm of the displacement no relation visible, although
they both shown buckling at a rim deflection o f ~3 mm with predicted and actual
buckling loads of 1617 N and 1650 N respectively.
4.4.7.2 Deformed shape
Both finite element predictions and experimental results show that, as expected,
progressive failure occurs with the top of the can/cylinder taking all the deflection up
to a point where the top becomes flat.
4.5 Closure
This chapter has dealt with the elastic-plastic analysis of a thin-walled cylinder (a)
with inverted dome base and (b) with inverted dome base and tapered top, in both
cases subjected to axial compressive loading, using a multi-linear material hardening
model. Predictions have been compared with the results of experimental testing on
aluminium aerosol cans and, for Case (a), with an analytical solution (which is found
to be inappropriate for reasons given).
In both cases, there is reasonable agreement between the predicted and experimental
collapse loads, although the Case (b) load-deflection curves are slight different. The
next chapter will consider modelling of the extrusion process.
167
Chapter five
MODELLING OF THE EXTRUSION PROCESS
5.1 Introduction
In this Chapter, the finite element modelling of the can extrusion process is
discussed. With reference to Chapter 1, there are two independent stages to the
extrusion process:
Stage 1 - formation of the side walls and (flat) base,
Stage 2 - formation of the inverted dome base.
Furthermore, Stage 2 can be undertaken either before or aftercan decoration. In
which case, although the punch and die geometries are identical,the boundary
conditions are different and this leads to different profiles.
The application of finite element analysis to the extrusion process is well established
and details of previous investigations are reported in Chapter 2. The aims of the
modelling work described here are:
1. to validate the approach by comparing the numerical predictions with analytical
predictions from Patten’s constant volume model [2] and with experimental data;
2. to study the influence of the coefficient of friction and the boundary conditions on
the resulting profile;
168
3. to obtain predictions for the forces required by the process;
4. to study the effects of punch and die slug (billet) geometry on the resulting profile.
Once a validated model has been produced, this will enable further investigation of
punch, die, and aluminium billet geometries in order to generate an optimised can
profile, an initial investigation in to which is presented in Chapter 6. Currently,
industrial practice is based on a ‘trial and error’ method and relies heavily on
extensive knowledge and experience to match the desired can profile with that of the
slug, punch and die geometry.
A 53 mm diameter can made from aluminium 1050. has been selected for analysis.
Clearly, there is a relationship between the accuracy of the predictions and the size of
the can since a larger can requires a larger slug of material and greater deformation
takes place.
5.2 Stage 1 modelling the base and side wall
The basic punch, die and billet geometry for a 53 mm can are shown in Figure 5.1,
based on information provided by Envases (UK) Ltd. 'Although it is virtually
impossible to produce a perfectly axisymmetric profile (due to tool wear, deflection
of the punch etc.), an axisymmetric model has been adopted because of the benefits
of reduced model size and consequent reduction in computing time that can be
achieved.
169
Punch 52.14 mm
£
$ 7.02 E
<£>
m
r>
CM
Billet 52.88mm
29.68mm
Figure 5.1: Die and punch geometry
Finite element predictions have been obtained using the large displacement elastic-
plastic facilities in the ELFEN [5] suite of programs. The geometry in Figure 5.1 was
then drawn into AutoCAD and the file was then transported in DXF format to
ELFEN. The resulting model has 2309 eight-noded axisymmetric isoparametric
elements.
With reference to Figure 5.2, the model is constrained as follows:
1. the surface ABCDE (which represents the die) was fully restrained
2. the surface FGHJ (which represents the punch) was restrained in the X
direction.
170
G H D C
bu
E c-ib-
A B
Figure 5.2: Finite element model boundary conditions
Three contact sets are created for this analysis:
• Die-Slug
• Punch-inner slug
• Punch-outer slug
Contact with friction was used in this analysis. Objects defining the contact between
the slug, punch and die were defined as:
Ob-die
Ob-punch_inner
Ob-punch_outer
Ob-slug_bottom
171
Ob-slug_inner-top
Ob-slug_outer-top
A negative displacement of the punch in the Y direction was used to model the punch
movement. A displacement in the Y direction o f -4.5 mm was therefore applied to
the punch using a rigid body load assigned to the top surface o f the punch as shown
in Figure 5.3.
A pp© - A ppD -A p p B -AppD
AppB AppD AppB~ ©-------
Figure 5.3: Displacement loading and contact objects
The elastic-plastic material properties of the slug (billet) are those for aluminium, as
discussed in Section 3.3.2 and presented in Tables 3.2 and 3.5. As before, yielding is
172
determined using the von Mises yield criterion and post-elastic behaviour is based on
the Prantl-Reuss flow rules (see Section 2.3). The material properties for the punch
and die are based on steel. Values for the coefficient o f friction at the contacting
surfaces o f 0, 0.1 and 0.25 have been assumed. The automatic mesh generation
facilities were used to create the mesh shown in Figure 5.4.
bu
Figure 5.4: Finite element mesh
After the cans have been formed, there will be a small amount o f elastic strain left
within the aluminium, which will cause a very slight reduction in the can size. Since
the cans are thin walled cylinders the diameters are small, thus the mechanical elastic
effects are very small and therefore can be neglected.
173
The aluminium is plastically deformed and then there is considerable heat
generation. This will affect the tooling dimensions, since there are considered in this
analysis however, it is very small and therefore can also be negligible.
5.2.4 Finite element predictions (p = 0.25)
Figure 5.5 shows the development of the extruded can for punch displacements of -
4.5 mm with a coefficient of friction of 0.25. When the punch is moved down it
pushes the aluminium billet down in to the die and the aluminium billet will start to
deform. The punch is now in contact with the billet and the billet is drawn through
the die to producing a profile as can be seen in Figure 5.5(a).
Also when the punch moves down by a distance the parallel section of the aerosol
can is formed. Although the tapered section of the can is formed, as can be shown in
Figures 5.5(b) and 5.5(c).
At the point when the can walls are about to be made, the aluminium fills the gap
between the extrusion punch and the die base. Also the tapered section of the can is
predicted see Figures 5.5(d) and 5.5(e).
174
Figure 5.5(a): Stage 1 partially deformed mesh
Figure 5.5(b): Stage 1 partially deformed mesh
175
Figure 5.5(c): Stage 1 partially deformed mesh
bu
Figure 5.5(d): Stage 1 partially deformed mesh
176
tu
Figure 5.5(e): Stage 1 partially deformed mesh
The profile at the end of the punch travel with load still applied is shown in Figure
5.6, which also includes the corresponding details when the punch is retracted. It can
be seen from the figures that the can walls and base are completely formed. The
maximum equivalent stress at the end of the punch travel is shown in Figure 5.7,
from which it is clear that gross yielding has occurred throughout the material. The
force-displacement characteristic for the punch is shown in Figure 5.8. It can be seen
from this figure that the displacement increases with increasing punch force and this
increase depends on the value of the coefficient o f friction, p. As the coefficient o f
friction increases, the force required for any given punch travel increases.
177
The force-displacement curve was obtained by re-running the analysis using an
applied force to the punch, as shown in Figure 5.9. An incrementally increasing force
was applied and the punch displacement noted after each increment. Hence it was
possible to determine the force at various stages o f the extrusion process, knowing
the punch displacement. The results of the analysis, shown in Figure 5.8, indicate
that a maximum force of -45.7 kN is required to produce this profile. Up to a
displacement of 1 mm the curve has a sharp rise with increasing force and
displacement. Until reaches the highest point.
Figure 5.6(a): Stage 1 at the end of the punch travel
178
Figure 5.6(b): When the punch retracted
156.0530
143.0486
130.0442
117.0398
104.0353
91.03093
78 . 0 2 6 5 1
65 . 0 2 2 0 9
52.01767
39 . 0 1 3 2 6
26.00884
1 13.00442
0 . OOe+OQO
Figure 5.7: Von Mises max equivalent stress contour plate at the end of punch
travel
179
H =025
1 0 -----
Punch travel Ornu)
Figure 5.8: Comparison of punch load Vs punch travel displacement for
various coefficient of friction
-Body
Body
Body
Body
Body
Botoy- B o dy Body 'Both -Both
bu
Figure 5.9: Stage 1 model with force loading
180
The thickness profile, starting at the centreline of the base and moving along the
base, around the comer and up the cylinder, is shown in Figure 5.10. This figure
includes finite element predictions for p = 0, 0.1 and 0.25 (discussed in Section
5.2.5). Figure 5.10 also included the results from [2], which is discussed in Section
2 . 10 . 2 .
e
S
Result from [2]

K
w
e
P = 0 .2 5
0.6
2v
'9
0.4
H
0.2
Position from ran base (nun)
Figure 5.10: Comparison of the effect of coefficient of friction on the Stage 1
extruded thickness profile with that predicted by the result from [2]
5.2.5 Effect of coefficient of friction
The effect of friction in the direct extmsion process is important in the commercial
process because it determines the billet size and hydraulic pressure requirements,
either by pressure limitation or by surface at the end o f the ram stroke.
181
The basic extrusion process was simulated with three different values for the friction
coefficient on the contacting surface and the comparisons of punch force and
resulting thickness profiles are shown in Figures 5.8 and 5.10 respectively. From
Figure 5.8 it can be seen that the variation in punch load with p during the process
can be significant (up to 5 KN) although the maximum force variation is less
significant (45 KN). Figure 5.10 indicates that thicker sections in the base are
produced when p is low but that the thickness at the start o f the cylindrical section is
less affected by the friction. Also the effect increases with the increasing area of
contact between the specimen and the tools, and with the reduction thickness of the
processed material [43].
5.2.6 Comparisons with analytical solution
The results of this validation are shown in Figure 5.10. It can be seen from the figure
that the extrusion model is reasonably accurate with the lower friction giving the best
correlation with the constant volume results of [2].
5.3 Stage 2 modelling
The bottom forming process produces the dome in the can base. This is produced by
supporting the can on a mandrel and forming the can base with a punch.
The basic punch and die geometry for the base of a 53 mm can are shown in Figure
5.11, based on information provided by Envases (UK) Ltd. Again, an axisymmetric
182
model has been adopted because of the benefits of reduced model size and
consequent reduction in computing time that can be achieved.
Finite element predictions have been obtained using the large displacement elastic-
plastic facilities in the ELFEN [5] suite of programs. The geometry in Figure 5.11
was replicated in AutoCAD [56] and the file was then transported in DXF format to
ELFEN.
Two sets of boundary conditions have been considered. The formation of the base
can take place either before or after the can is decorated and the support provided to
the can during base formation is different for the two cases. Experimental
observations indicate that a different base profile is generated for these two cases and
an additional objective of this work is to confirm (or otherwise) this variation.
5.3.2.1 Stage 2(a) - base formation before decoration
The die is fully restrained in the X and Y directions, the punch is restrained in the X
direction, and the can was constrained in the X-direction along the centre line.
5.3.2.2 Stage 2(b) - base formation after decoration
183
The vessel is restrained in the X and Y directions in the inside o f the cylindrical
section. The punch is restrained in the X direction and the die restrained around the
edge.
Figure 5.11: Impact extrusion dome base
184
K ,
Figure 5.12: Stage 2(a) finite element model boundary conditions
185
hu
Figure 5.13: Stage 2(b) finite element model boundary conditions
5.3.2.3 Stage 2 loading
A positive displacement of the punch of 8 mm ± 0.5 mm. in the Y direction was used
to model the punch travel [57]. Three values o f punch displacement in the Y
direction o f 7.5mm, 8 mm and 8.5 mm were therefore applied to the punch using a
rigid body load assigned to the top surface o f the punch, as shown in Figure 5.14
Objects need to be defined to account for contact between the tooling and the can.
These were defined as:
■ Outer surface o f can
186
■ Inner surface o f can
■ Top surface o f mandrel
■ Top surface of punch
Two contact surfaces were defined using updated penalty:
• Punch can
• Can mandrel
AppR ppR p pR ppS -
bu
App&
Figure 5.14: Finite element loading
187
The material models used are the same as those discussed in Section 5.2.3.
The automatic mesh generation facilities were used to create the mesh shown in
Figure 5.15. The resulting model has 1325 four-noded axisymmetric isoparametric
elements.
i i t t n i l \Z )-rT V W 7 7 v >r \ ^ \ i~ x \ \ \ i m r r r r i
Figure 5.15: Finite element mesh
188
5.3.4 Finite element predictions (p = 0.25)
5.3.4.1 Stage 2(a):' pre-decoration boundary conditions
Figure 5.16 shows the bottom forming simulation for a 53 mm aerosol can base at
various stages during the loading process, with p = 0.25. The development of the
inverted base shape is clearly visible. It can be seen from figures 5.16(d) (i) and 5.16
(d) (ii) that the highest coefficient of friction the lower punch travel.
The deformed shape after unloading is shown in Figure 5.16(e). It can be seen from
this figure that spring back (see Section 2.10.5) does occur, although the level is
relative low. And it is clear in point 28 in both figures and nodes 20 that the Y
coordinated for this point in figure 5.16 (e) (~ 5.3mm) is bigger than Y coordinate of
the same point in figure 5.16(d)(i) (~4 mm).
The predicted thickness-displacement characteristic is shown in Figure 5.17, along
with the experimental measurements taken from [2]. Figure 5.17 is discussed in
Section 5.3.6.
189
Figure 5.16(a): Stage 2 deformed mesh
190
Figure 5.16(b): Stage 2 deformed mesh
tu
S S K ^ /^ ^ 'A V f !
MMM^WVMMm»mmmmmmmmmmmS*
Figure 5.16(c): Stage 2 deformed mesh
192
SSslsJjlSl^ss&S
(i) ju = 0.25
(ii) p = 0
Figure 5.16(d): Stage 2 fully deformed mesh (max. punch travel 8.5 mm)
193
Qfil
Figure 5.16(e) Stage 2(a) with the punch removed
The deformed mesh for the other two values o f maximum punch travel o f 8.0 mm
and 7.5 mm are shown in Figures 5.16(f) and 5.16(g) respectively. It can be shown
from the figures that there are slight differences between the figures.
5.3.4.2 Stage 2(b): post-decoration boundary conditions
The fully deformed shape for Stage 2(b) is shown in Figure 5.18. By comparing this
figure with Figure 5.16(d), it is clear that the final deformed shape is affected by the
boundary conditions.
194
!!!!« »
asS fe& w
Figure 5.16(f): Stage 2(a) punch travel = 8 mm
Figure 5.16(g): Stage 2(a) punch travel =7.5 mm
195
Thickness of ran base (nun)
Experimental
Predicted
0.6
0.4
0.2
Position (nun)
Figure 5.17: Thickness-displacement characteristic after Stage 2(a)
Figure 5.18: Stage 2(b) fully deformed mesh (max. punch travel = 8.5 mm)
196
5.3.5 Effect of friction coefficient
Figure 5.16(d) shows the punch travel changes by variation in friction coefficient
however, the punch travel is slightly different with the coefficient of friction. It can
be seen that the lower friction coefficient the higher punch travel.
According to Figures 5.16(f), 5.16(g) that the cans have diverse bottom forming at
various punch travel. On the other hand the different coefficient of friction produces
the different punch travel.
5.3.6 Comparisons with experimental data
According to the Figure 5.17 that the results from finite element analysis were then
validated by comparison with the results of experimental measurements taken from
[2] both in terms of the thickness values and the profile, the best correlation is
achieved when p is set to 0.25 in the simulation. The results provide good qualitative
agreement. This shows that the finite element analysis predicts that the bottom
forming of aerosol cans is accurate to within a maximum error of 0.015 on thickness.
Figures 5.16(d) and 5.18 shows that the different boundary condition predict
different can geometries.
197
5.4 Closure
In this chapter, finite element analysis has been used to model the two-stage back
extrusion process for one-piece aerosol cans. The effect of coefficient of friction on
the thickness profile and the extrusion punch forces has been investigated.
This work shows that reasonable predictions can be achieved, when compared with
experimental data and an analytical solution. The next chapter presents preliminary
findings from an optimisation exercise. Having identified an ‘optimal’ profile, the
work in this chapter can be expanded in order to investigate the punch, die and slug
geometry requirements to produce this ‘optimum’ design.

Chapter six
OPTIMISATION
6.1 Introduction
It has been shown in Chapter 3 that these thin-walled cylindrical pressure vessels
with inverted bases are ‘over designed’ in some respects. In particular, it is
considered that the base thickness profile could be reduced, while still maintaining
the integrity o f the vessel. However, care must be taken since the vessels are
designed to accommodate any over-pressure by the mechanism of ‘dome reversal’
(or elastic-plastic buckling) of the base and this inherent safety feature must be
retained in any revised design.
In this chapter, a preliminary investigation into the optimisation of the vessel base
when under internal pressure, in order to reduce the thickness profile, is described.
Initially, a simplistic approach of reducing the base thickness is used to examine the
effect on stress distribution and elastic-plastic buckling pressure. Secondly, the DOT
optimisation program [58] has been used in conjunction with elastic finite element
analysis to provide a more structured approach to optimisation.
6.2 Simplistic approach
The basic geometry used in Chapter 3 has been modified by removing a horizontal
slice of material from the inside of the inner section of the can base. The original and
‘sliced’ axisymmetric finite element meshes are shown in Figures 6.1 and 6.2
199
respectively. The thickness at the centreline has been reduced from 1.25 mm to 0.75
mm in increments of 0.1 mm.
The three-dimensional model used to determine the elastic-plastic buckling pressure
o f the ‘sliced’ model is discussed in Section 6.2.5.
The loading and boundary conditions are identical to those used for the analysis of
pressure loading in Chapter 3 and described in Section 3.2 (axisymmetric model) and
Section 3.4 (3-D model).
Elastic and elastic-plastic analyses have been carried out with values for Young’s
modulus, yield stress and Poisson’s ratio o f 68.3 GPa, 100 MPa and 0.33
respectively. The multi-linear material hardening curve described in Table 3.5 has
been used to model the post-yield stress-plastic strain behaviour.
Figure 6.1: Simplistic approach, finite element mesh before reduction
(centreline thickness = 1.25 mm)
200
Figure 6.2: Simplistic approach, finite element mesh after reduction (centreline
thickness = 0.75 mm)
6.2.4 Results for axisymmetric model
The equivalent stress contour plot for the ‘sliced’ model with an internal pressure of
1.20 MPa is shown in Figure 6.3 it can be seen from the figure that the equivalent
stress varies between 122.66 MPa and 8.35 MPa this result comparison with the
corresponding Figure 3.37 in chapter three which has equivalent stress varies
between 119.53 MPa and 7.76 MPa and the maximum equivalent stresses in each
case are in the comer region. Although Figure 6.4 compared with Figure 3.38 it can
be seen from the figures that the collapse load is changes from 1.35 MPa to 1.53
MPa when the thickness of the sliced model decreased by 0.75 mm.
201
r- 122.6648
112.1386
103.6125
---------- 94.08633
— — - 84.56017
---------- 75.03402
- 65.50786
- 55.98170
- 46.45554
36.92938
27.40322
17.87707
8.350908
Figure 6.3: Simplistic approach, equivalent stress contour plot for p =1.20 MPa
The loading was then increased and collapse is predicted when the pressure reaches
1.35 MPa (compared to 1.53 MPa for the original geometry). The corresponding
equivalent stress contour plot is shown in Figure 6.4.
202
132.2284
124.7480
117.2676
109.7872
102.3069
94.82648
87.34611
79.86573
72.38535
64.90498
57.42460
49.94422
42.46384
Figure 6.4: Simplistic approach, equivalent stress contour plot for p = 1.35 MPa
6.2.5 Elastic-plastic buckling results using a 3-D model
Using the approach described in Section 3.4, a three-dimensional model o f the
‘sliced’ geometry has been created and the basic 3D model, constraints, loading and
finite element mesh are shown in Figures 6.5, 6.6, 6.7 and 6.8 respectively.
203
.mr'
><
//
Figure 6.5: Simplistic approach, 3D model (centreline thickness = 0.75 mm)
Figure 6.6: Simplistic approach, 3D constraints (centreline thickness = 0.75 mm)
204
sb e
Figure 6.7: Simplistic approach, 3D loading (centreline thickness = 0.75 mm)
Figure 6.8: Simplistic approach, 3D mesh (centreline thickness = 0.75 mm)
205
The internal pressure load was increased incrementally until elastic-plastic buckling
took place at a pressure o f 1.80 MPa (compared to 2.02 MPa for the original model)
and the corresponding equivalent stress contour plots for the original model and the
‘sliced’ model are shown in Figures 6.9 and 6.10 respectively.
In both cases and due to time restrictions, the buckling mode is symmetrical because
a symmetrical model (without small-scale perturbations - see Section 3.4) was used.
This is an area for further investigation. Figures 6.9 and 6.10 show that the finite
element analysis predicts that the aerosol can base will be fully deformed at 2.02
MPa for the actual thickness and 1.80 MPa for the modified geometry and the
maximum stresses are now concentrated in the lower section o f the can walls in both
cases.
1 06- 7 69 8
9 8- 2686 5
89-76746
81- 26626
7 2- 7 6 5 0 6
64 -26387
- 5 5- 7 6 2 6 8
- 4 7- 261 48
38 - 7 6 0 2 8
- 30- 2 5 9 0 9
- 21-75789
- 13- 256 70
55501
206
106-7698
98-26865
89-76746
81- 2 6 6 2 6
72-76506
64 - 26387
55-76268
47-26148
38-76028
30-25909
21- 7578 9
13-25670
55501
r ____________ ______
Figure 6.9: Simplistic approach, equivalent stress contour plot at the point of
elastic-plastic buckling with p = 2.02 MPa (centreline thickness = 1.25 mm
110-5005
107 -1033
97-70600
88-30873
78 -91146
69-51419
60-11692
50 -71965
41-32238
31- 92511
2 2 5 .2 7 8 4
131.305 8
37.33307
207
-
^k'^.l4
M N nH M
XM5^»*®V*Y HwV« »i»nT*w®*w®wiB'™
«® w M B B B wB*^&WSE5lSS!iSwî
H M iH H S n
k ^ M § ss< S « gg
Figure 6.10: Simplistic approach, equivalent stress contour plot at the point of
elastic-plastic buckling with p = 1.80 MPa (centreline thickness = 0.75 mm)
6.3 Optimisation procedure
6.3.1 Objective function and constraints
The choice o f a suitable design variable (s) is very important as it can affect the
degree of non-linear of the objective function or the constraints. It can also result in
other implicit constraints, which are not necessarily obvious at first sight. It is
recommended to have a direct connection between the values o f the design variables
and actual geometry [50]. In this work, the volume is indirectly used as the objective
function.
An optimisation problem is stated as:
Minimize or maximize the objective function (F(x)), subject to:
208
g,(X)<0 }
} j = 1, NCON Inequality constraints
h,(x) = c,. }
and
X. < X, < X f }
} i = 1, NDV Side constraints
x^x<x„}
If the objective functions to be minimised is the cross-sectional area:
F(X )=24 . . . (6. 1)

1=1
where A,, is the CSA of the i th FE.
and n is the number o f elements.
Subjected to the constraints that the stresses in each element must be less than the
yield stress:
Gj (x) < a y (j = 1 to n) .. .(6.2)
and hence:
-O ’J/-C * ) ■< i -► O 1’ / C */) - j < ...(6.3)

Gy Gy
209
where <jy is the yield stress
The selected design variables are the thickness values at the centre of each element
around the profile and the objective function to be minimised becomes:
F(X) = ti + 12 + 13.................. + tn ... (6.4)
6.3.2 DOT optimisation program
DOT is a computer program for optimisation. Specifically, it is used to automatically
adjust to maximize or minimize a calculated quantity, while satisfying a number of
constraints [58].
A computer-based procedure was written using the FORTRAN programming
language that reads the output file from an ELFEN analysis and extracts the
equivalent stresses at each node in each super-element and determines the maximum
values in each of these elements. The DOT program is called and these n values of
equivalent stress together with the n thickness values are then used by DOT to
generate a new set of thicknesses, based on the constraints in equation (6.3). This
information is then used to manually generate a new set of nodal co-ordinates for the
n+1 nodes on the inner surface of the profile (i.e. the outer profile is fixed and the
cross-section is modified by moving the outside nodes). The new model then
provides the input for the next iteration.
Details of the FORTRAN coding can be found in Appendix C.
210
The basic geometry, using 6 super-elements, is that shown in Figure 6.11. A mesh of
187 four-noded axisymmetric elements was created from this model, using the
automatic mesh generator within ELFEN.
Figure 6.11 Optimisation analysis, basic model with six super-elements and six
design variables
The loading and boundary conditions are identical to those used for the analysis of
pressure loading in Chapter 3 and described in Section 3.2. An arbitrary pressure of
0.50 MPa has been used in the pre-buckling analyses.
211
Elastic and elastic-plastic material data are as discussed in Section 6.2.3.
6.3.6 Results
An initial (zero iteration) analysis using the original model in Figure 6.11 was
performed and the procedure described in Section 6.3.2 was used to generate a
revised model (with new thickness values from DOT) in this analysis the thickness of
the base only considered due to the high stress in this region and the wall thickness is
constant. ELFEN then re-generated the mesh and the process repeated.
After five iterations, DOT indicated that convergence had been achieved and the
resulting ‘optimised’ shape is shown in Figure 6.12. The original and ‘optimised’
thickness values are given in Table 6.1 also the optimisation convergence is shown in
Figure 6.13.
Original Iteration (1) Iteration2 Iteration3 Iteration4 Iteration5

1.25 1.15 1.05 0.95 0.82 0.81
1.1 1 0.9 0.8 0.71 0.70
1.07 0.95 0.85 0.75 0.63 0.59
1.13 1.03 0.93 0.83 0.70 0.68
1.05 0.94 0.84 0.75 0.61 0.59
I 5.6 5.07 4.58 4.08 3.47 3.37
Table 6.1: Thickness at each iteration
212
Figure 6.12: Optimisation analysis geometry after optimisation
+£
+Li
+£->
U -\-tr.
Iteration
Figure 6.13: The convergence of the solution
213
The ‘optimised’ shape was then analysed using the incremental elastic-plastic
facilities within ELFEN to establish the pressure at which collapse occurs. (Note that
axisymmetric models cannot distinguish between elastic-plastic buckling and
collapse, as discussed in Section 3.4).
The equivalent stress contour plot, corresponding to a collapse pressure o f 0.62 MPa,
is shown in Figure 6.15 also the pre-buckling equivalent stress contour plot of
pressure 0.50 MPa is shown in Figure 6.14. The equivalent collapse pressure for the
original profile is 1.53 MPa (see Section 3.3.5). It can be seen from the figures that
the thickness is reduced then the amount o f materials will reduce hence, the collapse
pressure also decrease.
126.913 6
1 1 7 .7 516
1 0 8 .5 896
99.42763
90.26563
- 81.1 036 4
- 71.9 416 4
- 62 .77965
- 53.61766
- 44 .45567
- 35.293 67
- 2 6 .131 68
- 16.96968
Figure 6.14: Equivalent stress contour plot (pre-buckling, pressure =0.50 MPa)
214
165.0404
153.0080
140.9756
128.9433
116.9109
104.8786
92.84621
80.81385
68.78149
56.74913
44.71677
32.68441
20.65204
Figure 6.15: Optimisation analysis, equivalent stress contour plot at the point of elastic-
plastic buckling with p = 0.62 MPa
6.4 Closure
The results from a preliminary investigation into shape optimisation applied to these
thin-walled cylinders have been presented. This work was carried out at a late stage
in the project and, therefore, only provides a starting point for further, more detailed
analysis. It is clear that significant reductions in the cross-section o f the vessel base
are possible, within the limits of acceptable burst pressure (which occurs in the plain
cylinder region). At the same time, however, the elastic-plastic buckling pressure is
significantly affected and this may adversely affect the lower operational pressure
limits. The choice o f model, objective function and constraints is an area for further
investigation and this is discussed in more detail in the final chapters.
215
Chapter seven
DISCUSSION
7.1 Introduction
The project has looked in-depth at the design and manufacture of aluminium aerosol
cans, as a specific form o f thin-walled pressure vessel which, due to its complex
shape, cannot simply be designed around the traditional British (BS5500) and
American (ASME VIII) codes for pressure vessel integrity. Also, the structural
integrity of this vessel shape, which consists of an inverted dome end, parallel
cylindrical section and truncated cone top, due to external loading is beyond the
scope of these codes.
Furthermore, it has been established, from collaboration between the University and
a local manufacturer of such vessels (through Knowledge Transfer Partnerships), that
competition is fierce and material costs contribute at least 50% of the cost of
manufacture. Therefore, it is essential that both the design and the manufacturing
processes can be simulated in order to optimise on material usage, whilst still
maintaining the integrity of the vessels. This requires:
• a thorough understanding of the yield, elastic-plastic buckling and ultimate
failure of the vessels under internal pressure;
• accurate modelling of the buckling behaviour under compressive axial load;
216
• accurate modelling of the back-extrusion process itself, thus providing the
opportunity to investigate the effects of tooling geometry changes on the
resulting vessel profile;
• optimisation studies which integrate with the above in order to reduce costs
with no loss of integrity and still maintaining the inherent safety feature
provided by the inverted base at an acceptable pressure below that at which
burst occurs.
These four requirements have formed the basis of the work described in this thesis.
The literature review has concluded that little specific research has been carried out
on the structural integrity of these complex vessels (in fact, ‘design by test’ appears
to be the preferred approach) and there is little evidence of finite element analysis
being applied to the modelling of the back-extrusion process. Where work has been
carried out, it avoids the issue o f friction and its influence on thickness profile and
extrusion force requirements.
The research work reported in this thesis investigates the linear and non-linear, large
displacement behaviour of aluminium thin-walled pressure vessels, in the form of
aerosol cans, subject to internal pressure and axial loading using the elastic and
elastic-plastic facilities of the ELFEN finite element program. Extensive elastic and
elastic-plastic analyses have been preformed using both constant thickness and
realistic thickness profile models to obtain a better understanding of the mechanisms
of buckling and failure. Similarly, the yielding and flow of material under pressure
during the extrusion process has been modelled using ELFEN explicit.
217
7.2 Internal pressure loading (Chapter 3)
7.2.1 Elastic analyses
Initially, constant thickness (0.4 mm < t < 1.4 mm) axisymmetric finite element
models have been used to study the stress patterns that develop and hence establish
the conditions for yielding and the variation with thickness profile. A typical
geometry (i.e. Geometry 4) with t = 1mm is selected for a full review and the
summary results of other geometries are presented. The predictions show that initial
yielding will, as expected, occur on the inside surface at the relatively sharp comer
close to the plain tube region, which acts as a significant stress concentration feature
with Kt values between 4.07 and 13.83 (depending on thickness) being predicted. A
non-linear relationship between maximum stress and thickness is also predicted,
although when plotted against D/t, the predictions become more linear. However, the
relationship between limiting (yield) pressure and thickness, for both the base and
plain tube, appear to be reasonably linear. This clearly helps in any investigations
into material optimisation.
A similar response is seen for the realistic thickness profile. This is based on
experimental observations made by Patten [2], who found that although the
cylindrical section is reasonably parallel (-0.31 mm), the variation in thickness along
the base is significant (0.7 and 1.31 mm). This results in a reduced elastic stress
concentration factor of 3.45 and a greater ‘spread’ of the stress contours, which is
similar to that predicted for the 0.8 mm constant thickness model. The predictions
also confirm that the base stresses are very low, compared with the comers and
cylinder and this is considered to provide the impetus for material optimisation.
218
7.2.2 Elastic plastic analyses
Elastic-plastic finite element predictions from the constant thickness axisymmetric
models, using both elastic-perfectly-plastic (EPP) and multi-linear kinematic work-
hardening (EKH) models for aluminium, have enabled the investigation of the
development of the plastic zones as the pressure is increased above that required for
initial yielding. A plastic hinge, where the complete cross-section has yielded, is seen
to develop and (for the EPP model) no further increases in pressure can be applied.
For the EKH model, further increases in pressure are possible until the UTS are
reached. A similar response is seen for the realistic thickness model however, it is
clear that these axisymmetric models cannot be used to simulate the elastic-plastic
buckling (form of elastic-plastic ‘snap-through’, also referred to as ‘dome reversal’)
characteristic exhibited by actual vessels under pressure.
In reality, the thickness profile is not entirely axisymmetric due to tolerances in the
extmsion process and although variations in thickness are very small, they are
sufficient to cause a slightly unsymmetrical buckling mode, due to minor radial
variations in profile and there is a clear distinction between the elastic-plastic
buckling of the base and burst (collapse) pressures, where bursting occurs in the
plain tube region.
Consequently, a full three-dimension model was created and a small imperfection
was introduced in the base, using results from an eigenvalue analysis (lowest mode)
and based on the method described by Robotham et al [24]. There is good agreement
between the predicted buckling mode (Figure 3.44) and that obtained experimentally
(Figure 3.52) and there is reasonable agreement between the predicted buckling
219
pressure (1.7 MPa) and the experimental value (1.6 MPa). Possible reasons for the
discrepancy include:
• Variations in the level of material strain hardening that have occurred during
the extrusion process (the material data used is from tests on specimens taken
from the cylindrical section, which has been subjected to greater strain
hardening than the base)
• The softening effects of temperature increase on material properties. A
significant increase in temperature is apparent both during extrusion and
internal coating which will cause the aluminium to soften. This effect has not
been investigated.
• The approximate nature of the finite element method, particularly for non
linear analysis.
7.2.3 Upper and lower bound pressures
The elastic compensation method proposed by Mackenzie and Boyle has been used
to estimate the upper and lower bound limit loads, using only elastic finite element
analysis, for both constant thickness axisymmetric and realistic thickness profile
models. The use of this method in a limited number of relatively straightforward
components and loading arrangements has been reported in the open literature and it
is considered that the application described in this thesis provides further, more
detailed, information on the nature and limitations of the method which will be of
interest and benefit to Engineers and Designers.
220
The results for the constant thickness models show that the predicted collapse
pressures are within the upper and lower bound estimates, closer to the upper bound,
and this provides a degree o f confidence in these approximate methods. However,
the range between the upper and lower bounds is large and, unfortunately, the lower
bound is always greater than the yield stress. This limits the use of these approximate
methods for this type of geometry and loading to collapse pressure estimates. The
upper and lower bound range for the realistic thickness model includes the predicted
buckling and collapse pressures but, again, the yield pressure is outside the range.
Care should therefore be taken when using this approximate method.
7.3 Axial loading (Chapter 4)
In practice, these components are subjected to axial loading during neck formation
and valve insertion/charging. Under these conditions, the can must not collapse and
this requirement provides the need for a study of thin-walled, complex shape
pressure vessels subjected to axial loading. Again, the results will have direct
implications to any subsequent material optimisation study. Also, the effect of strain
hardening and temperature on material properties and finite element predictions,
discussed in Section 7.2.2, is also relevant here.
7.3.1 Axial loading during neck forming
For this analysis, a model of the plain open cylinder with inverted base (to simulate
the first stage of necking) was used with a multi-linear kinematic work-hardening
model for aluminium. Small perturbations were introduced into the model to enable
buckling, rather than compressive collapse.
221
Experimental validation tests were also performed and the predictions compare
favourably with the experimental results in a number of ways:
• Elastic-plastic buckling occurs at the base, which demonstrates the
significance o f the sharp comer on the stress distributions. The analytical
solution for the equivalent plain cylinder in compression is clearly
inappropriate.
• The predicted and experimental load-displacement characteristics are very
similar with linear pre-buckling behaviour.
• There is excellent correlation between the predicted and experimental
buckling loads.
One experimental test was extended in order to show the post-buckling behaviour
with an apparent increase in stiffness prior to a secondary buckling process.
7.3.2 Axial loading during valve insertion/charging
In this case, the top and cylindrical sections of the thin-walled cylinder were
modelled. Again, a multi-linear material hardening model was assumed. Small
perturbations to the geometry were not necessary since the experimental results show
the deformation to occur in the truncated cone-shaped top in preference to the plain
cylinder or base.
222
The finite element predictions indicate a reasonably linear load-displacement curve
up to a deflection of ~3 mm, during which time the convex top is being flattened and
the stiffness should increase slightly, although noted. After this, there is a reduction
in load as the top becomes flatter towards becoming concave. A very different
experimental response is observed with low loads up to a deformation of -2 mm and
a rapid increase up to -3.5 mm. However, the angle of the cone on the finite element
model is shallower than that for the components used in the experimental tests and a
different response is, therefore, not surprising. A peak load is again shown and it
appears that secondary stiffening may be taking place as the displacement
approaches -5 mm.
It is surprising therefore that the predicted and experimental ‘collapse’ loads are
reasonably similar. This cannot be explained and further investigation is necessary.
7.4 Modelling of the extrusion process (Chapter 5)
The modelling of manufacturing processes is a more recent application of non-linear
finite element analysis and, in particular, research into the modelling of the back-
extrusion process is limited. In this chapter, particular attention has been paid to the
effects of friction and boundary conditions on the forces required to extrude the
material and the resulting thickness profile, for which no previous results could be
found.
There are three independent stages to the deformation process:
• Stage 1 - formation of the side wall and flat base
• Stage 2 - formation of the inverted dome base
• Stage 3 - formation of the truncated cone top
223
and the first two stages have been simulated in this project. Also, Stage 2 can take
place either before or after decoration, in which case the boundary conditions and the
subsequent thickness profile are different.
7.4.1 Stage 1 simulation
The effect of friction coefficient on the finite element predictions for the thickness
profile has been investigated for 0 < p <0.25, based on discussions with colleagues
and industrialists. The maximum difference in predicted thickness (~ 0.2 mm) is in
the base region close to the sharp comer and round into the first part of the plain
cylindrical section. The results suggest that thicker sections in the base and cylinder
are produced when p is low this is described in Section 2.10.4.
Predictions are compared with the predicted profile from a simple model of the
punch and die geometry when the punch is fully inserted. It would appear that finite
element predictions with p = 0 provides the best comparison. This seems reasonable
as, in practice, a graphite powder is applied to the billet prior to extmsion.
The force-displacement curve for the punch clearly predicts two slopes and it is
considered that the change in slope corresponds to the point at which the material
starts to flow around the punch comer and up the die. The effect of friction on punch
load is significant during the process (~ 5 KN) but the maximum force varies by less
than 5% with the range of p considered. In practice, the predicted machine power
requirements will vary little with friction.
224
Unfortunately, the predicted length of the plain cylindrical section is significantly
less than that for the actual cans. This is an area requiring further investigation.
7.4.2 Stage 2 simulation
The effect of friction coefficient on the finite element predictions for the bottom
forming process produces the dome in the can base has been investigated for 0 < p <
0.25. The results suggest that the lower coefficient of friction the higher punch travel.
The finite element result at different boundary conditions shows that the different
boundary conditions produce different cans.
The predictions suggest that the effects of springback, due to elastic recovery when
the punch is removed, are minimal. This information is useful in the design of dies
and punches however, if an optimised can had a thinner base then higher levels of
springback would be expected.
7.5 Optimisation studies (Chapter 6)
The need for an ‘optimised’ thickness profile has been identified in Section 7.1. The
results of the research presented in this thesis provide important background
information on how thickness affects both the integrity and structural response of
these complex thin-walled cylinders under typical loading conditions and it is clear
that preferential thinning of the section is possible.
In this chapter, a preliminary look at optimisation has been carried out and it the
results are far from conclusive. However, they do provide a valuable insight for
future investigation.
225
7.5.1 Simplistic approach
Assuming that the base is the main section where material could be removed without
affecting the integrity, the simplest form of optimisation is to remove a ‘slice’ of the
material from the inner section o f the base. In this way, the thickness at the base
centreline was reduced from 1.25 mm to 0.75 mm in increments of 0.1 mm.
Using an axisymmetric finite element model, the reduction in ‘collapse’ pressure is
relatively small (from 1.53 MPa to 1.35 MPa for a centreline thickness reduction
from 1.25 mm to 0.75 mm). Similarly, using a three-dimensional model (but without
any geometrical perturbation), the predicted elastic-plastic buckling pressure falls
from 2.02 MPa to 1.80 MPa for the same thickness reduction. These results are
encouraging and suggest that material savings may be possible without loss of
structural integrity.
7.5.2 Structured approach using DOT
Section 6.3 has described a more structured approach to optimisation, using the DOT
optimisation procedure. In this exploratory work, the sum of the thicknesses at
particular sections in the model has been the objective function to be minimised with
constraints on maximum equivalent stress.
The procedure requires interaction with the finite element program by way of a
separate FORTRAN program (see Appendix), which acts as the interface between
DOT and ELFEN.
226
The optimised shape from this preliminary study shows some irregularities. This is
probably due to the original stress distribution, since the regions where the thickness
has been reduced significantly compare with the regions of high stress in the original
model and a more ‘smoothed’ approach is needed. Also, the ideal situation of a
constant stress cannot be achieved in this type of problem because the pressure
loading produces a bending moment on the shell and so a stress variation between
inside and outside surfaces will always exist in the base and comer. Further work on
the choice o f objective function and constraints is necessary.
7.6 Closure
The final chapter, Chapter 8, provides a summary of conclusions from the research
and some recommendations for further studies.
227
Chapter eight
CONCLUSIONS AND RECOMMENDATIONS
8.1 Conclusions
The following conclusions are drawn from the research:
1. The finite element method provides a significant advantage over traditional
experimental testing methods for proving and improving designs, as the method has
the advantages.of repeatability, rapid re-analysis of geometry and loading changes

j
and reduced costs.
2. The aim o f the research project was to develop a predictive tool that facilitates the
can design and optimisation process and, in this respect, the objectives have been
achieved. The models are capable of reasonably accurate quantitative assessment of
j the effect of varying geometry and material properties and are invaluable aids in
| aerosol can design.

i
I 3. Constant thickness (axisymmetric) models can be used in a limited way to study

I
qualitatively the effects of changes in thickness on material and structural response
but the actual thickness profile is far from constant and quantitative information can
only be obtained when more realistic models are used.
4. Axisymmetric models generally provide some useful information on the behaviour
of such structures under pressure and/or compressive axial load. However, they lack
228
the ability to predict elastic-plastic buckling of the base due to internal pressure and
of the side-walls due to axial loading.
5. Full three-dimensional models, with realistic thickness profiles, can be used
successfully to predict the buckling and collapse conditions for both pressure and
axial loading. However, the models need to be modified (by means of geometric
imperfections) to enable the buckling mode to be simulated.
6. The elastic compensation method provides a straight-forward and useful approach,
without the need for complex elastic-plastic analysis, which requires knowledge and
modelling of the post-yield non-linear material behaviour. However, the method has
limited application to this type of geometry. Although estimated elastic-plastic
buckling and collapse pressures are below the upper bound estimates, the lower
bound estimates are higher than the pressure at which first yield occurs.
7. Finite element analysis can be used successfully to model the back-extrusion
process and good comparisons between predicted and experimental data have been
demonstrated. The effects of the choice of coefficient of friction and boundary
conditions on the extruded profile have been investigated and it appears that the best
correlation between experiment and prediction is achieved with a low coefficient of
friction being used.
8. In all cases, an accurate model for the non-linear material behaviour is necessary.
It has been noted that strain rate variations and elevated temperature may affect the
229
stress-strain characteristics and lead to variations in material properties across the
section.
9. Material optimisation is an important consideration for the manufacturer.
However, design constraints limit what can be achieved. The preliminary
optimisation study has highlighted both the opportunity for reduced material and the
complexities of using a structured approach.
8.2 Recommendations for further work
A number of recommendations for further investigation are drawn from the research:
1. Further studies of the extrusion process are needed in order to identify why the
length of the plain cylinder is under-predicted.
2. Stage 2 of the extrusion process should be investigated further (and modelled more
accurately) to understand the differences between the experimental results and finite
element predictions o f load-displacement for the punch.
3. A detailed investigation into shape optimisation is required where further thought
and detailed analysis is given to the choice of objective function and constraints.
Also the choice of element type and number of elements should be reviewed.
4. The spatial variation in material properties, due to strain rate and temperature
effects, may be significant and require further investigation.
230
5. Finally, there are areas outside the scope of this study, which should be addressed.
For example, such components are subjected to radial loading during packaging into
bundles. Radial buckling is a possible failure mechanism that needs to be considered,
particularly if material optimisation is considered.
231
REFERENCES
1 htt://www.cancentral.com - Can Central is the web site of the Can

Manufacturers Institute.
2 Patten, S. design and optimisation of aluminium aerosol cans
produced by a back extrusion process MSc Thesis, University of
Wales Swansea, 2001
3 British Standard Institution BS5500; Specification for Unified Fusion
Welded Pressure Vessels 1991
4 American Society of Mechanical Engineers, ASME Boiler and
Pressure Vessel Code, 1991
5 ELFEN User Manual Version 3.0.4, 2001, Rockfield Software,
Swansea, UK
6 http ://science.howstuffworks.com/aerosol-can.htm
7 Cheers, C. F. Design and optimisation of an ultrasonic for forming
metal cans PhD. Thesis, University of Loughborough Leicestershire
UK 1995
8 http://www.yorks.karoo.net/aerosol/link3.htm
9 Timoshenko, S. P. and Gere, J. M Theory of elastic stability
McGraw-Hill, New York 1961
10 NAFEMS- Finite element primer, NAFEMS, 1992
11 Prinja, N. and Clegg. R. Non linear behaviour of 3-D beams and
shells shells NAFEMS BENCHMARK 1993
12 Zienkiewicz, O. C. The finite element method, McGraw -Hill, 1984
13 Prandtl, L. Spannungsvert eilung in plastischen koerpem proc O f the
1st Int. congress on App.Mech., Delft, Tecnische Boekhandel en
Druker, J. Waltman, Jr., 1925, pp 43-54
14 Reuss, E. Beruecksichtigung der elastischen formaenderon in der
plastizifacts theorie Z. Angew. Math. Mech, 1930, Vol. 10, pp. 266-
274
15 Valliappan, S. Continum Mechanics Fundemantals, Rotterdam 1981
232
16 Owen, D. R. and Hinton, E. Finite element in plasticity theory and
practice pineridge press Lid. 1980
17 Owen, D. R. and Hinton, J. E. NAFEMS- Introduction to Non-linear
finite element analysis, NAFEMS, 1992
18 Ryder, G. H. Strength of Materials 1969
19 Bushnell, D. Computerized buckling analysis of shell Martinus
Nijhoff Publishers 1985
20 Brush, D. O. Buckling of bars, plates, and shells 1975
21 Farshad, M. Designed and analysis of shell structures Kluwer
Academic Publisher 1992
22 Flugge, W. Stresses in shells, 2 nd Edn., Springer, Berlin, 1973
23 Karman, T. and Tsien, H. The buckling of thin cylindrical shells
under axial compression, J. Aeronout. Sci., 1941, Vol. 8, pp. 303-312
24 Robotham, W.S., Hyde, T. H. and Williams, E. J. Finite element
tli
torsional buckling analysis and prediction for plain shafts MPSVA 5
25 Robotham, W. S. The elastic-plastic buckling behaviour of shafts
PhD. Thesis, University of Nottingham 2000
26 Rik’s, E. An Incremental Approach to the Solution of snapping and

Buckling Problem, Int. J. Solid Structures, 1979, 15,7, pp. 529-551
27 Mendleson, A. Plasticity Theory and Application 1968 (Macmillan,
London)
28 Berak, E. G. and Gerdeen, J. C. Finite element technique for limit
analysis of structure. Trans. ASME, J. Pressure Vessel Technol, 1990,
112, 138, 144
29 Dhalla, A. K. Verification of an elastic procedure to estimate follow-
up. In design o f elevated temperature piping (Eds R.H. Mallet and
R.M. Mello). 1984, pp. 81-96 (American Society of Mechanical
Engineers New York)
30 Marriot, D. L. Evaluation of Deformation or Load Control of
Stresses under Inelastic Conditions using Elastic Finite Element Stress
Analysis. In Proceedings of the Pressure Vessel and Piping
Conference, Pittsburgh, Pennsylvania, 1988, PVP-Vol. 136 (American
Society of Mechanical Engineers, New York
233
31 Mackenzie, D and Boyle, J.T. A method of estimating limit loads by
iterative elastic analysis. I: simple examples. Int. J Pressure Vessel
and piping, 1993, 53, 77, 95
32 Mackenzie, D., Nadarajah, C., Shi, J. and Boyle, J.T. Simple
bound on limit load by elastic finite element analysis. Trans. ASME,
J. Pressure Vessel Technol., 1993, 115,27-31
33 Gowhari-Anaraki, A. R. and Adibi-Asl, R. Estimation of Upper and
Lower bound limit loads and shakedown load for structural frames
based on reduced modulus approach. In Proceeding of the 6th
International Conference on Civil Engineering Isfahan, Iran, 2003,
399-406
34 Hardy, S. J., Gowhari-Anaraki, A. R., and Pipelzadeh, M. K.
Upper and Lower bound limit and shakedown loads for hollow tubes
with exisymmetric internal projections under axial loading. J. Strain
Analysis, 2001, 36(6), 595-604
35 Seshadri, R. and Kizhatil, R.K. Inelastic analysis of pressure
components using the 'GLOSS' diagram. Proc. ASME, 1990, pp.
186
36 Sheppard, T. Extrusion of Aluminium Alloys, Kluwer Academic
Publisher 1999
37 Htenger, H. Extrusion Process, Machinery, Tooling 1981
38 Kobayashi, S. and Altan. T Metal forming and the finite element
method 1989
39 Joeri, 1. Developments in finite element simulations of aluminium
extrusion. PhD. Thesis, University o f Twente 2000
40 Joachim. D. Backward can extrusion, PhD. Thesis, University of
Aalborg Denmark 2005
41 Blazynski, T. Z. Metal forming Tool profiles and flow 1979
42 GAO, C. and FANG. Y. Investigation on the factors influencing the
thickness distribution of super- plastic-formed components. Journal of
Zhejiang University Science, 2005, Vol. 6A (7), PP. 711-715
43 Zienkiewicz, O. C. and Godbolet, P. O. Flow of plastic and visco
plastic solids with special refemce to extrusion and forming process.
234
International Journal for Numerical Methods in Engineering, 8-16,
1974
44 Mattiasson, K. and Strange. A. Simulation of spring-back in sheet
metal forming NUMIFORM, 1995, PP. 115-124
45 Akbari, R., Hardy, S. J., Kadkhodayan, M., and Pipelzadeh, M. K.
Simulation of spring back in 2-D draw bending, 2001
46 Mercer, C., Nagtegaal, J., and Rebelo, N. Effective application of
different solver to formings NUMIFORM 1995, PP. 469-474
47 Joannic, D and Gelin, J. Accurate simulation of sprinback in 3-D
sheet metal forming process. NUMIFORM 1995, PP. 729-734
48 Narasimhan, N. Predicting springback in sheet metal forming an
implicit and explicit sequential solution proceder, Finite element in
analysis and design, 1999, Vol. 33, pp. 29-42
49 Arwidson, C. Numerical simulation of sheet metal forming for high
strength steels PhD. Thesis, University of Lulea Sweden 2005
50 Hinton, E. S. and Ozakca, M. Analysis and Optimisation of
Prismatic and Axisymmetric Shell Structures UK 2003
51 Sodeik, M. and Sauer. R. Mechanical behaviour of food cans under
radial and axial load. 3rd international tinplate conference London,
1984, paper 11
52 Jing, H. R. Application of structure optimisation technique to
aluminium beverage bottle design, 4th Congress on structural and
Multidisciplinary optimisation, 2003, PP. 103-105
53 Iestyn, B. J. Computational strategies for design optimisation of food
cans, PhD, Thesis, University of Wales Swansea, 1999
54 Davis, J. R. Aluminium &Aluminium Alloys, ASME International,
1993
55 Chilver, L. C and Rose, C. T. Strength of materials and structures,
Edward Arnold 1993
56 Auto CAD User Manual Version, 2002
57 Dr Elias, L. Envases (UK) Ltd, January 2006
58 DOT User Manual Version 4.20, Vanderplaats Research and
Development, Inc Colorado Springs, 1995
235
Appendix A
DRAWING OF EXTRUSION TOOLING
i-v
PLATE- a \rINGOT
BLOCK INGOT d ie 1 \ DIE fBACKER
EXTRUSION
i FLOW
RAM HOLLOW
RAM
CONTAINER EXTRUSION CONTAINER

FLOW
DIRECT EXTRUSION INDIRECT EXTRUSION
Figure 1: Tooling and metal flow for direct and indirect extrusion process [41]
236
o o
55505
in
20,5±1
in
cu
zi— i
DIN
cu
o
in
Figure 3: Aerosol can dimension
237
Appendix B
PROPERTIES OF ALUMINIUM FOR IMPACT EXTRUSION
In order to calculate the strength o f the extruded cans and the required thickness of
the can walls, the following mechanical properties are required:
» Ultimate Tensile Strength (UTS)

• Yield Strength
Knowledge o f the UTS will be used to predict the burst pressure of the cans and the
yield strength used to predict the deformation pressure. Both these properties will
then determine the required can wall thickness and therefore the cost of each can.
For commercial aluminium, the yield strength is not always a clearly defined point.
For this reason, most textbooks refer to percentage proof strains of aluminium rather
than yield strengths.
The mechanical properties of aluminium from different manufacturers can vary by

large amounts. This is due to differences in alloy content, grain structure and
processing (heat treatment, rolling etc.).
At present, Envases (UK) Ltd. use three types of aluminium to produce extruded
aerosol cans. These are summarised in the table below:
Supplier BS Code Purity Si, wt% Fe, wt% Hardness (Brunell)

Alucenca 1080 99.8 0.076 0.203 16-19
Hydroslug 1070 99.7 0.076 0.127 18-20
Inespal
Rhienfelden 1050 99.5 0.064 0.243 20-21
The following additional data is assumed about all three types:
238
Density p =2700 K g / m 3
Poisson’s Ratio o = 0.33
Cost £6700 per m3 (£2.50 per kg)
Thermal Conductivity K = 230 W/m°C
Coefficient of Linear Expansion a =24x10"6
Y oung ’s Modulus E = 68.3x10* N/mz
Chemical Properties
Alloy Si Fe Cu Mn Mg Zn Ga Ti %Pure
1080 0.15 0.15 0.03 0.02 0.02 0.06 0.03 0.02 99.80
1070 0.2 0.25 0.03 0.03 0.03 0.07 —
0.03 99.70
1050 0.25 0.4 0.05 0.05 0.05 0.07 —
0.05 99.50
Manufacturing Processes
Process Rhienfelden Alucenca

1 Melting Melting
2 Rotary Strip Casting Rotary Strip Casting
3 Hot Rolling («400°C) Hot Rolling ( « 400°C)
4 In line cooling and Cooling
lubrication
5 Cold Rolling ( « 60°C) Coiling
6 Shear Cool for 48 Hrs Minimum
7 Coil De-coil
8 Cool for 48 Hrs Minimum Cold Rolling ( » R.T.)
9 De-Coil Coil
10 Blank De-coil
11 Wash Blank
12 Anneal (500-520°C) Wash
13 Tumble Anneal (420-450°C)
14 Pack Tumble
15 Pack
239
Appendix C
SIMPLE FORTRAN PROGRAM USING TO CALCULATE

YOUNG’S MODULUS (E) AT EACH ITERATION
C PROGRAM ECM
INTEGER NNODE(8),NELEMENTS,NELEM,IBLANK,NEL
REAL EFSTRS(8),ENEW(296),EFFMAX,SIGY,SIGMAD
NELEMENTS=296
OPENUjFILE-ecnÔtempdat.dat',STATUS-OLD’)
DO J= 1,NELEMENTS
READ(1,'(F4.1)') ENEW(J)
PRINTXEIOA)', ENEW(J)
ENDDO
SIGY=100.0
SIGMAD=0.0
DO K=1,NELEMENTS
EFFMAX=0.0
DO 1=1,8
READ( 1,'(21 X,I3)') NELEM
READ(1,’(I1)') IB LANK
READ(1,'(91X,F10.5)') EFSTRS(I)
READ(1,’(I1)') IBLANK
PRINT*, I
IF(EFSTRS(I).GT.EFFMAX) THEN
EFFMAX=EFSTRS(I)
ELSE
ENDIF
ENDDO
IF(EFFMAX.GT.SIGMAD)THEN
SIGMAD=EFFMAX
NEL=K
ELSE
240
ENDIF
PRINT*, EFFMAX
ENEW (K)=2 *ENEW (K) *SIGY/(3 *EFFMAX)
PRINT*, ENEW(K)
ENDDO
DO K= 1,NELEMENTS
WRITE( 1,’(E 10.4)') ENEW(K)
ENDDO
WRITE( 1,'(A,F 10.5,A,I3)')'SIGMAD = ’,SIGMAD,'IN ELEMENT ’,NEL
CLOSE(l)
END PROGRAM ECM
241
C SIMPLE PROGRAM FOR CAN THICKNESS OPTIMISATION
DIMENSION X(7),XL(7),XU(7),G(7),SIG(7)
DIMENSION WK(800),IWK(200),RPRM(20),IPRM(20)
NRWK=800
NRIWK=200
DO 101=1,20
RPRM=1
10 IPRM=1
C TRY SQP METHOD
METHOD=3
NDV=7
NCON=7
C INTIAL THICKNESS VALUES
X(l)=1.25
X(2)=1.12
X(3)=1.34
X(4)=0.71
X(5)=0.62
X(6)=0.31
DO 201=1,NDV
XL(I)=0.0
20 XU(I)=20.0
SIG(1)=10.766
SIG(2)=6.641
SIG(3)=20.648
SIG(4)=8.265
SIG(5)=27.078
SIG(6)=25.035
SIGMAY=100
IPRINT=1
242
I
MINMAX=-1
INFO=0
100 CALL DOT (INFO, METHOD, IPRINT, NDV, NCON, X,XL,XU
* OBJ, MINMAX, G,RPRM,IPRM,WK,NRWK,IWK,NRIWK)
IF(INFO.EQ.O)STOP
call system ("./elfendyn elastic3t 40");
CALL EVAL (OBJ,X,G)
GO TO 100
END
SUBROUTINE EVAL(OBJ,X,G)
DIMENSION X(*),G(*)
OB J=X( 1)+X(2)+X(3)+X(4)+X(5)+X(6)
| DO 30 1=1,
| 30 G(I)=1-SIG(I)/SIGMA
RETURN
END
!
i
i
[
243

Finite Element Analysis of The Design and Manufacture of Thin-Walled Pressure Vessels Used As Aerosol Cans

Uploaded by

Copyright:

Available Formats

Finite Element Analysis of The Design and Manufacture of Thin-Walled Pressure Vessels Used As Aerosol Cans

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Finite Element Analysis of The Design and Manufacture of Thin-Walled Pressure Vessels Used As Aerosol Cans

Uploaded by

Copyright:

Available Formats

Swansea University E-Theses

Finite element analysis of the design and manufacture of thin-walled

Abdussalam, Ragba Mohamed

Finite Elem ent Analysis of the Design and M anufacture of

Thin-walled Pressure Vessels used as Aerosol Cans

Ragba M ohamed Abdussalam

B. Sc Eng (honours) M. Sc Eng

Thesis submitted to the University of Wales Swansea

For the degree of Doctor of Philosophy, September 2006

Civil and C om putational Engineering C entre

All rights reserved

INFORMATION TO ALL USERS

In the unlikely e v e n t that the a u thor did not send a c o m p le te m anuscript

Published by ProQuest LLC(2018). C opyright of the Dissertation is held by the Author.

All rights reserved.

I would like to express my sincere appreciation and gratitude to my advisor Dr Steve.

Chapter 2: GENERAL BACKGROUND AND LITERATURE

2.5 Thin cylinder formulae......................................................................................... 32

4.1 Introduction........................................................................................................... 129

Chapter 5: MODELLING OF THE EXTRUSION PROCESS

5.1 Introduction.......................................................................................................... 168

Chapter 6: O PTIM ISA TIO N

6.1 Introduction........................................................................................................... 199

Chapter 8: CONCLUSIONS AND RECOMMENDATIONS

^ a] contour plot (G4, p = 0.1 MPa) ^

G l contour plot (G4, p = 0.1 MPa) ^

Area o f can walls

The inside diameter

Rate o f dissipation o f energy per unit volume

Young’s modulus (MPa)

True modulus o f elasticity

Elastic modulus o f each element

Geometric non- linearity

Second Moment o f Area

Upper bound limit load

Lower bound limit load

Punch head radius, Radial co-ordinate direction, Can radius

Outer can radius

Can wall thickness

G char Characteristic stress within the element

( v eq)nom Nominal stress

° Yield Yield stress

°cr Critical stress

P The constant o f proportionality

K Elastic stress concentration factor

container and so improvements in design and manufacturing processes are always

was not invented until 1930 [1].

used comprehensively in this project, supported by experimental validation.

1.2 Thin walled pressure vessels

A pressure vessel is a closed structure containing liquids or gases under pressure. An

practice o f pressure vessel design by analysis is most commonly based on elastic

one-piece aluminium can, consists o f an inverted base, a nominally constant­

the manufacturing process is such that there is a significant variation in thickness,

requirements have to be met:

Cylinder - must be thick enough to withstand bursting due to overpressure

- must be thick enough to withstand radial buckling during

- must be thick enough to avoid axial buckling/collapse under

compressive axial load during manufacture and charging

Base - must be thick enough to withstand bursting due to overpressure

one-piece aluminium can, consists o f an inverted base, a nominally constant