See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/267614022

Improvements in Adapting mα-Tangent Method for Fitness for Service

Evaluation of Local Thin Areas in Storage Tanks

Conference Paper  in  American Society of Mechanical Engineers, Pressure Vessels and Piping Division (Publication) PVP · July 2013
DOI: 10.1115/PVP2013-98068

CITATIONS READS

3 2,087

2 authors, including:

Sridhar Sathyanarayanan
Flex-N-Gate
10 PUBLICATIONS   108 CITATIONS   

SEE PROFILE

All content following this page was uploaded by Sridhar Sathyanarayanan on 21 June 2019.

The user has requested enhancement of the downloaded file.

IMPROVEMENTS IN DESIGN AND FITNESS
EVALUATION OF
ABOVE GROUND STEEL STORAGE TANKS

by

© Sridhar Sathyanarayanan

A Thesis submitted to the

School of Graduate Studies

in partial fulfillment of the requirement for the degree of

Doctor of Philosophy in Engineering

Faculty of Engineering and Applied Science

Memorial University of Newfoundland

May, 2014

St. John’s Newfoundland and Labrador

 ABSTRACT

Above ground steel storage tanks are widely used to store liquids in a variety of

industries. The design and fitness for service procedures for such tanks are a concern for

international standards and need to be continually improved upon to ensure better safety

and serviceability. Several important aspects about tank design and assessment are

studied in this thesis.

The bottom plate material near the shell to bottom joint in the tank is generally in

plastic range. It is a critical failure point in many modes of tank failure. The effect of

increasing the bottom plate projection length at this joint for tanks with rigid ring wall

foundations is studied both theoretically and numerically. A theoretical beam model is

validated using finite element analysis (FEA) and extended to determine the length of

bottom plate projection needed for maximum effect. The formation of plastic hinges in

the bottom plate on the inside and outside of this joint is discussed in detail using FEA.

Tanks operating at elevated temperatures (200F to 500F) need to consider

additional stresses due to thermal expansions and restraints from the tank shell and

bottom plate interactions. The frictional forces from the foundation cause significant

stresses at the tank bottom. The design guidelines by API 650 standard address this issue

using a factor named ‘C’ that defines the ratio of actual expansion against free expansion

of the tank bottom. At present, an empirical range of ‘C’ values (0.25 – 1.0) is allowed

without clear guidelines for selecting a suitable value. This thesis evaluates the current

procedure and suggests an alternate method by incorporating the friction coefficient

ii
directly in the stress equations, instead of the C-factor. The fill/draw down cycle of the

stored liquid could lead to low cycle fatigue near shell to bottom joint. The peak

alternating stress (strain) at this location determines the fatigue life of the tank. The

widely used API 650 procedure employs beam on elastic foundation theory to determine

the fatigue life for all tanks. The thesis shows that this is incorrect for tanks on concrete

ring wall. The appropriateness of using this theory is studied and an alternative beam

model is proposed. It is verified using FEA.

Damage due to corrosion in the form of local thin area (LTA) is a widespread

problem in storage tanks. Fitness for service (FFS) methods are quantitative engineering

evaluations used to demonstrate the structural integrity of an in-service tank containing

damage like LTA and make run, repair or replace decisions. The m-tangent method is a

simplified limit load procedure that can be used for such FFS evaluations. This thesis

uses a modified reference volume for m-tangent method applied to tanks and reports

initial results for FFS evaluations. The study also finds that for large cylinders like tanks

with very high R/t ratio, the circumferential decay lengths will be smaller than those

previously reported ( 2.5 Rt rather than 6.3 Rt ).

iii
 ACKNOWLEDGEMENTS

I would like to express my sincere gratitude to my supervisor, Dr. Seshu M. R.

Adluri for his intellectual guidance, support and valuable discussions during the course of

my Doctoral program. I also wish to express my gratitude to Dr. R. Seshadri for all his

supervision, generous support and patience.

I would also like to thank Dr. A. S. J. Swamidas and Dr. Katna Munaswamy for

all their help and support.

The financial support provided by the School of Graduate Studies and the Faculty

of Engineering and Applied Science, Memorial University is gratefully acknowledged. I

also thank Dr. Leonard Lye, Associate Dean of Graduate Studies, Ms. Moya Crocker and

Ms. Mahoney Colleen at the Office of Associate Dean, for all the administrative support

during the course of my program. I also wish to thank the computing services at the

engineering faculty for their continuous assistance throughout my program.

I would like to thank all my friends for their support and encouragement. I also

wish to thank my friend Dr. Geeta Iyer for her motivation and support during difficult

times. I owe my deepest thanks to my wife, Hemalatha, for her unconditional love,

constant encouragement, sacrifice and understanding during all these years. I also express

my heartfelt gratitude to my beloved parents and sister for their love, support and

encouragement at all times.

Lastly, I profoundly thank the almighty for giving me the strength and

perseverance to pursue and complete this program successfully.

iv
 TABLE OF CONTENTS

ABSTRACT ............................................................................................................................................ II
ACKNOWLEDGEMENTS ................................................................................................................... IV
TABLE OF CONTENTS ........................................................................................................................ V
LIST OF TABLES ...............................................................................................................................VIII
LIST OF APPENDICES ......................................................................................................................XIV
LIST OF SYMBOLS AND ABBREVIATIONS ..................................................................................XV
CHAPTER 1 INTRODUCTION ............................................................................................................ 1
1.0 BACKGROUND ...................................................................................................................................1
1.1 NEED FOR RESEARCH .........................................................................................................................4
1.2 OBJECTIVES OF RESEARCH.................................................................................................................7
1.3 SCOPE OF THE STUDY ........................................................................................................................7
1.4 STRUCTURE OF THE THESIS ...............................................................................................................8
CHAPTER 2 LITERATURE REVIEW............................................................................................... 11
2.0 THEORIES OF FAILURE .........................................................................................................................11
2.0.1 Maximum Shear Stress Criterion .................................................................................................12
2.0.2 Distortional Energy Density Criterion ..........................................................................................13
2.1 SHELL THEORY .....................................................................................................................................14
2.2 BEAM-ON-ELASTIC FOUNDATION .......................................................................................................18
2.3 SHAKEDOWN ANALYSIS ......................................................................................................................21
2.4 LOW CYCLE FATIGUE ...........................................................................................................................22
2.5 TANK RESEARCH ..................................................................................................................................24
2.6 FITNESS FOR SERVICE ..........................................................................................................................29
2.6.1 Levels of FFS Assessment ............................................................................................................30
2.6.2 FFS Procedures for Local Thin Areas ...........................................................................................32
2.6.2.1 Limit Load Analysis ......................................................................................................................... 34
2.6.2.2 The mα- Tangent Method ............................................................................................................... 35
2.6.2.3 Remaining Strength Factor Method .............................................................................................. 36
2.6.3 API 579-1/ASME FFS-1 Metal Loss Assessment Procedure .........................................................37
2.6.3.1 Data Requirements for Characterizing Metal Loss ..................................................................... 38
2.6.4 FFS Research for Locally Thinned Areas ......................................................................................41
2.7 SUMMARY...........................................................................................................................................48
CHAPTER 3 FINITE ELEMENT MODELING ................................................................................. 50
3.0 BASIC ASSUMPTIONS ..........................................................................................................................50
3.1 MODELING USING PLANE AXISYMMETRIC ELEMENTS ........................................................................52
3.2 MODELING USING SHELL ELEMENTS...................................................................................................55
3.3 CONTACT ELEMENTS...........................................................................................................................59
3.3.1 Contact Algorithm .......................................................................................................................61
3.4 ISSUES IN FEM MODELING ..................................................................................................................64
3.4.1 Mesh Convergence Study ............................................................................................................66
3.5 DIMENSIONS OF TANKS USED IN THE ANALYSIS .................................................................................72

v
 3.6 SUMMARY...........................................................................................................................................77
CHAPTER 4 ISSUES WITH BOTTOM PLATE PROJECTION AT SHELL-TO-BOTTOM
JOINT ................................................................................................................................................... 78
4.0 SHELL-TO-BOTTOM JOINT ...................................................................................................................78
4.1 ANNULAR PLATE .................................................................................................................................80
4.2 PLASTIC HINGES ..................................................................................................................................83
4.3 TEMPERATURE EFFECTS ......................................................................................................................85
4.4 ANALYTICAL MODEL AS PROPOSED BY DENHAM ET AL. .....................................................................86
4.5 VERIFICATION OF DENHAM’S MODEL USING FEA ...............................................................................92
4.6 EFFECT OF PROJECTION OF ANNULAR PLATE BEYOND SHELL .............................................................95
4.7 DETERMINATION OF FULL PROJECTION LENGTH ................................................................................98
4.7.1 Ratio of Lengths “b” and “a” .....................................................................................................100
4.8 RELATIONSHIP BETWEEN MOMENTS MO AND MFX FOR FULL PROJECTION LENGTH .........................102
4.8.1 Influence of the Term (1-1/𝛃H) .................................................................................................105
4.9 VERIFICATION OF THEORETICAL RESULTS WITH FEA ........................................................................109
4.10 EFFECT OF SELF-WEIGHT ON STRESSES (COMPARISON USING FEA) ...............................................113
4.11 CLOSURE .........................................................................................................................................117
CHAPTER 5 INCORPORATION OF FRICTION COEFFICIENT IN ELEVATED
TEMPERATURE TANK DESIGN .................................................................................................... 118
5.0 BACKGROUND ...................................................................................................................................119
5.1 CURRENT PRACTICE ..........................................................................................................................122
5.1.1 Stresses in Shell (Tank Wall) ......................................................................................................123
5.1.2 Stresses in Annular Plate ...........................................................................................................125
5.2 INCORPORATION OF FRICTION COEFFICIENT IN DESIGN EXPRESSIONS ............................................129
5.2.1 Incorporation of Friction Coefficient in Shell Equations ............................................................133
5.2.2 Thermal Stresses in Cylinders ....................................................................................................135
5.3 RESULTS AND DISCUSSION ................................................................................................................139
5.3.1 Influence of Friction on Fatigue Life ..........................................................................................151
5.3.2 Influence of Filling Procedure ....................................................................................................152
5.4 CONCLUSIONS...................................................................................................................................153
CHAPTER 6 FATIGUE ANALYSIS OF SHELL TO BOTTOM JOINT OF TANKS ................... 155
6.0 INTRODUCTION .................................................................................................................................156
6.1 FATIGUE AT SHELL TO BOTTOM JOINT ..........................................................................................157
6.2 DESIGN FOR FATIGUE ........................................................................................................................158
6.3 ISSUES IN EXISTING PROCEDURE .......................................................................................................160
6.3.1 Necessity of the Condition in Eq.6.2 ..........................................................................................160
6.3.2 Use of Beam-on-elastic Foundation Theory ..............................................................................162
6.4 PROPOSED CHANGES IN THE DETERMINATION OF PEAK ALTERNATING STRESS ..............................164
6.5 RESULTS AND DISCUSSION ................................................................................................................167
6.6 INFLUENCE OF PLASTIC HINGES ........................................................................................................173
6.7 CONCLUSION ....................................................................................................................................176
CHAPTER 7 LOCALLY THINNED AREAS ................................................................................... 177
7.0 CURRENT PRACTICE ..........................................................................................................................178
7.1 CONCEPT OF DECAY LENGTH AND REFERENCE VOLUME ..................................................................182
7.2 CIRCUMFERENTIAL DECAY LENGTH FOR TANKS ...............................................................................184
7.3 DETERMINATION OF RSF USING M-TANGENT METHOD ..................................................................188

vi
 7.3.1 RSF using Analytical Approach ..................................................................................................189
7.3.1.1 RSF Based on Upper Bound Multiplier ...................................................................................... 191
7.3.1.2 RSF Based on the mα-Tangent Multiplier .................................................................................. 191
7.3.1.3 RSF Based on Classical Lower Bound Multiplier ..................................................................... 192
7.3.2 RSF Based on Non Linear Finite Element Analysis (NLFEA) .......................................................192
7.4 REFERENCE VOLUME FOR LTA IN TANKS ...........................................................................................193
7.5 MODIFIED REFERENCE VOLUME APPROACH ....................................................................................196
7.6 ILLUSTRATIVE EXAMPLE ....................................................................................................................202
7.7 RSF USING ANALYTICAL APPROACH ..................................................................................................207
7.8 RESULTS AND DISCUSSION ................................................................................................................207
7.8.1 Results from m-Tangent Method Using LEFEA .......................................................................208
7.8.2 Results from m-Tangent Method Using Analytical Approach .................................................210
7.8.3 Influence of Hydrostatic vs. Uniform Pressure Loading on RSF from m -Tangent Method ......211
7.9 SUMMARY.........................................................................................................................................215
CHAPTER 8 CONCLUSIONS AND RECOMMENDATIONS ....................................................... 217
8.0 SUMMARY.........................................................................................................................................217
8.1 CONCLUSIONS...................................................................................................................................220
8.1.1 Guidelines for Designers ...........................................................................................................223
8.1.2 Recommendations for Future Study..........................................................................................224
REFERENCES ................................................................................................................................... 227
BIBLIOGRAPHY ............................................................................................................................... 238
APPENDIX A .................................................................................................................................... 243
APPENDIX B ..................................................................................................................................... 252
APPENDIX C .................................................................................................................................... 262

vii
 LIST OF TABLES

Table 3.1 Element Types used in Finite Element Method ................................................ 63

Table 4.1 Annular Plate Thickness in millimetres [API 650, Cl.5.5.3] ............................ 84

Table 4.2 Design Details for Several Typical Tank Sizes .............................................. 103

Table 4.3 Variation of (1-1/H) with Radius and Height ............................................... 106

Table 4.4 Variation of Mfx/Mo with (1-1/H) Value ..................................................... 106

Table 4.5 Comparison of Theoretical and FEA Results for Uplift Lengths ‘a’ and ‘b’ . 110

Table 4.6 Effect of Self-Weight on Parameters ‘a’ and ‘b’ ............................................ 111

Table 5.1 Implied Coefficient of Friction for Fixed Values of C-Factor........................ 148

Table 5.2 Implied C-Factor for Fixed Values of Coefficient of Friction........................ 150

Table 7.1 RSF Values from Nonlinear FEA and m-Tangent Method Using xc = 2.5 Rt

......................................................................................................................................... 208

Table 7.2 Comparison of RSF Values from m-Tangent Method (xc= 6.3 Rt & 2.5 Rt )

......................................................................................................................................... 209

Table 7.3 RSF Values from m-Tangent Method Considering Entire Volume as

Reference Volume [Ahmad, et al., 2010] ....................................................................... 210

Table 7.4 RSF from m-Tangent Method using Analytical Approach ........................... 211

Table 7.5 Comparison of RSF Values for Hydrostaic Pressure and Uniform Pressure

Loading ........................................................................................................................... 212

viii
 LIST OF FIGURES

Fig. 1.1 Tank with Annular Plate ....................................................................................... 5

Fig. 2.1 Yield Criteria in σ1 - σ2 plane: (a) Tresca Criterion, (b) von-Mises Criterion .. 14

Fig. 2.2 Forces and Moments Acting on a Cylindrical Shell Element ............................ 15

Fig. 2.3 Deflections of Elastic Foundations under Uniform Pressure ............................. 19

Fig. 2.4 Beam-on-Elastic Foundation .............................................................................. 19

Fig. 2.5 Manson-Coffin Strain – Life relation for AISI 304 Stainless Steel [Manson &

Halford, 2006] .................................................................................................... 23

Fig. 2.6 Pseudo Elastic Tank Model ................................................................................ 27

Fig. 2.7 Critical Thickness Profile [API 579-1/ASME FFS-1, 2010] ............................. 40

Fig. 3.1 Elastic - Plastic Material Model for Steel ............................................................ 52

Fig. 3.2 Axisymmetric Tank Model (Not to Scale) .......................................................... 53

Fig. 3.3 Element Geometry for 8 Noded Plane Element (PLANE 183) ........................... 55

Fig. 3.4 Geometry of Axisymmetric Shell Element (SHELL 209) .................................. 56

Fig. 3.5 Deformation Profile of a Tank Modeled Using Shell Elements .......................... 57

Fig. 3.6 Eight Noded Shell Element (SHELL 281) ......................................................... 58

Fig. 3.7 Target Element Geometry [ANSYS Reference Manual, 2009] .......................... 60

Fig. 3.8 Contact Element Geometry [ANSYS Reference Manual, 2009] ........................ 61

Fig. 3.9 Localised Deformation at the Bottom Corner of Bottom Plate ........................... 67

Fig. 3.10 Deformation in the Bottom Plate (with Extra Projection) ................................. 68

Fig. 3.11 Zones Specifying Different Mesh Densities ...................................................... 68

ix
Fig. 3.12 Typical Mesh at Shell to Bottom Joint .............................................................. 69

Fig. 3.13 Stress Distribution for Partial Plasticity ............................................................ 69

Fig. 3.14 Stress Distribution at Limit Load ...................................................................... 70

Fig. 3.15 Deformation Profile of Tank with LTA ............................................................ 70

Fig. 3.16 Stress Profile of Tank with LTA ....................................................................... 71

Fig. 3.17 LTA of Tank at Failure ...................................................................................... 71

Fig. 3.18 Dimensions of Tank (Case 1) ............................................................................ 73

Fig. 3.19 Dimensions of Tank (Case 2) ............................................................................ 75

Fig. 3.20 Dimensions of Tank (Case 3) ............................................................................ 76

Fig. 4.1 Shell to Bottom Joint in the Tank ........................................................................ 80

Fig. 4.2 (a) Annular Plate on Concrete ring wall (b) Plastic Hinges at the Shell to Bottom

Joint [Jones and Seshadri, 1989] ........................................................................ 84

Fig. 4.3 Denham’s Model of Shell to Bottom Joint .......................................................... 87

Fig. 4.4 Freebody Diagram of Shell.................................................................................. 89

Fig. 4.5 Moment vs. Slope Graph to Find Uplift Length .................................................. 91

Fig. 4.6 Uplift in the Bottom Plate.................................................................................... 94

Fig. 4.7 Bending Stress on the Bottom Side of Bottom Plate ........................................... 95

Fig. 4.8 Bending Stress in the Bottom Side of Bottom Plate for 100mm Projection Length

............................................................................................................................ 96

Fig. 4.9 Bending Stress in Bottom Plate for 71mm Projection ......................................... 97

Fig. 4.10 Bending Stress in Bottom Plate for 50mm and 71mm Projection ..................... 98

Fig. 4.11 Beam Model of Shell to Bottom Joint for “Full” Projection ............................. 99

x
Fig. 4.12 Bottom Moment for Thickness Ratio ............................................................. 108

Fig. 4.13 Mfx/Mo vs. ts/tp for Different Values of (1-1/H)............................................. 107

Fig. 4.14 Bending Stresses in Bottom Plate from FEA .................................................. 109

Fig. 4.15 Bending Stress in the Bottom Plate of Elevated Temperature Tank ............... 113

Fig. 4.16 Bending Stress in Bottom Plate (Bottom Side) – With and Without Self-weight

.......................................................................................................................... 114

Fig. 4.17 Bending Stress in Bottom Plate – With and Without Self wt. – Expanded View

Near Tank Wall ............................................................................................................... 115

Fig. 4.18 von Mises Stress in Shell Wall – With and Without Self-weight Consideration

......................................................................................................................................... 116

Fig. 5.1 Meridional Moment in the tank [Karcher, 1978a] ............................................. 127

Fig. 5.2 Annular Plate Bending Stress Caused by Hydrostatic and Thermal Loads

[Karcher, 1978a] ............................................................................................... 128

Fig. 5.3 Friction Forces Below Bottom Plate ................................................................. 130

Fig. 5.4 Forces in Tank ................................................................................................... 134

Fig. 5.5 Expansion of Tank ............................................................................................. 135

Fig. 5.6 Deflection in the Tank Wall Including the Effect of Friction (=0.3) .............. 140

Fig. 5.7 Bending Stress (Sx) in Tank Wall Including Friction Effects (=0.3) .............. 141

Fig. 5.8 Hoop Stress in Tank Wall Including Friction Effects (=0.3) .......................... 141

Fig. 5.9 Bending Stress in the Tank Wall Including Friction Effects – (Elastic-Plastic

FEA) ................................................................................................................. 143

Fig. 5.10 Tank Wall Bending Stress for Various Values of ‘C’ and  ........................... 144

xi
Fig. 5.11 Tank Wall Hoop Stress for Various Values of ‘C’ and ................................ 144

Fig. 5.12 Tank Wall Bending Stress at Different Temperatures for  = 0.7 .................. 145

Fig. 5.13 Temperature Influence on C-Factor ................................................................ 147

Fig. 5.14 Influence of C-Factor on Fatigue Life ............................................................. 151

Fig. 6.1 Shell to Bottom Joint in the Tank on a Rigid Base ........................................... 157

Fig. 6.2 Free Body Diagram of Shell .............................................................................. 160

Fig. 6.3 Uplift at Shell to Bottom Joint of a Tank on Concrete Ring Wall .................... 162

Fig. 6.4 Uplift of Bottom Plate from FE Model with 2D Axisymmetric Elements ........ 163

Fig. 6.5 Idealized Beam Model ....................................................................................... 166

Fig. 6.6 Uplift Deformation of Bottom Plate .................................................................. 168

Fig. 6.7 Uplift Deformation from FE Model with Shell Elements ................................. 169

Fig. 6.8 Stress Sx in Bottom Plate near Shell Joint ........................................................ 169

Fig. 6.9 Bending and von Mises Stress in 6mm Thick Bottom Plate (On the Inside) .... 171

Fig. 6.10 Bending Stress in 6mm Bottom Plate (on the Outside) ................................... 172

Fig. 6.11 Bending and von Mises Stress in 8 mm Thick Bottom Plate .......................... 173

Fig. 6.12 Bending of Bottom Plate in Shell to Bottom Joint [Zick and McGrath, 1968] 174

Fig. 6.13 Influence of Bottom Moment on Tank Wall Bending Stresses ....................... 175

Fig. 7.1 Critical Point ...................................................................................................... 181

Fig. 7.2 Reference Volume and ‘Dead Volume’ ............................................................ 184

Fig. 7.3 Hoop Stress Variation with Height of LTA....................................................... 186

Fig. 7.4 Hoop Stress Variation with Width of LTA ....................................................... 187

Fig. 7.5 Tank with Locally Thinned Area [Ahmad, et al., 2010] ................................... 194

xii
Fig. 7.6 Tank with LTA and Reference Volume as used by Ahmad et al.,[2010] ......... 195

Fig. 7.7 RSF using Analytical approach -Ahmad [2010] ............................................... 196

Fig. 7.8 mα-Tangent Value for Tank with Hydrostatic Pressure ..................................... 198

Fig.7.9 mα-tangent value for tank with hydrostatic and uniform pressure ...................... 199

Fig.7.10 Reference Zone for LTA .................................................................................. 201

Fig.7.11 Modified Reference Volume (Decay Lengths Smaller than Flaw) ................. 201

Fig.7.12 Modified Reference Volume (Decay Lengths Larger than Flaw) .................... 202

Fig.7.13 Evaluation of Hydrostatic Equivalent Pressure ................................................ 205

Fig.7.14 RSF Values for Hydrostatic and Uniform Pressure Loadings .......................... 212

Fig.7.15 RSF Using Proposed Reference Volume Approach ......................................... 213

Fig.7.16 RSF Using Three Different Approaches for Reference Volume ...................... 215

xiii
 LIST OF APPENDICES

Appendix A : ANSYS APDL Input for performing EPFEA of Example Tank- CASE1

Appendix B : ANSYS APDL Input for performing EPFEA of Example Tank- CASE2

Appendix C : ANSYS APDL Macro incorporating m-tangent method

xiv
 LIST OF SYMBOLS AND ABBREVIATIONS

Symbols

a Projection of bottom plate outside the tank wall

b Length of bottom plate inside the tank wall

C Reduction factor to account for the partial expansion of tank bottom due to

temperature change

D Diameter of tank

Ds Flexural Rigidity of tank wall

E Young’s Modulus

G Specific gravity of liquid

Gm Shear modulus

EIP Flexural rigidity of bottom plate/unit width

H Tank filling height

K Foundation modulus

Ka Width of annular plate

Kc Stress concentration factor at shell to bottom joint

L Total uplift length of the bottom plate

Lmsd Shortest distance between the edge of corrosion area and any closest structural

discontinuity

M Folia’s factor

MT Moment due to thermal loading at bottom joint

xv
Mb Moment due to pressure loading at bottom joint

Mx Long moment at a distance ‘x’ from tank bottom

Mbp Inelastic moment in the bottom plate

Mo Moment at the bottom of the shell

MP Plastic moment capacity of the bottom/annular plate

m0 Upper-bound limit load multiplier

mL Lower-bound limit load multiplier

mT m-tangent multiplier

N Number of load cycles estimated for the design life capacity of tank

Nx Membrane Force Per unit length in the meridional direction

N Circumferential (hoop) Force per unit length in tank wall

P Load

Pf Failure load (pressure) of damaged component

Pfo Failure load (pressure) of undamaged component

qo Uniform internal pressure

q Uniformly distributed load (Pressure)

Qo Shear force at bottom of tank wall

Qxs Shear force in the tank wall at distance ‘x’ from bottom

QT Shear force due to thermal loading at bottom joint

R Nominal (mean) radius of tank wall at bottom

r Radial distance from the centre of the tank

S Alternating stress range in the bottom plate near shell to bottom joint (Sb/2)

xvi
s Length of flaw in height direction ( = 2b)

Sb Pseudo elastic bending stress (strain range) in the bottom plate near shell to

bottom joint

Sx Stress in bottom plate in the radial direction

SY Specified minimum yield strength at the design temperature

SY Yield Strength

Sf Flow stress

sij0 Statically admissible deviatoric stress for impending plastic flow

T Shell self weight

t Nominal thickness of the tank wall

ts Thickness of the tank bottom shell course

ta Annular plate thickness (is equal to tb if not given exclusively)

tb Thickness of bottom plate

tp Bottom plate thickness (assuming uniform thickness and friction coefficient for

the entire tank; same as tb unless specifically noted)

tmm Minimum measured remaining wall thickness

tc Corroded thickness of the tank wall

tmin Minimum required wall thickness

u Radial displacement of any point in bottom plate

ux Displacement in the xs direction of shell

UD Distortional energy density

Uo Strain energy density

xvii
w Hydrostatic pressure at tank bottom

x Distance from the tank shell towards center of the tank

xs Distance from the tank bottom along the height of tank

Xc Decay length in the circumferential direction

Xl Decay length in the longitudinal direction

y Deformation/Deformation in annular plate

ys Radial deformation of tank shell

Z Pressure acting normal to shell surface

 Coefficient of thermal expansion

β Shell parameter

 Strain

a Applied strain

p Plastic strain

y Yield strain

 Principal strains

T Change in temperature (C)

TL Limiting temperature (C)

δ Final radial displacement of tank wall bottom

0 Point function

γ Unit weight of stored liquid

 Coefficient of friction

xviii
L Limiting Coefficient of friction

0 Flow parameter

ν Poisson’s ratio (0.3 for carbon steel)

 Stress/Normal stress

 Principal stress

e von Mises equivalent stress

y Yield strength of the material

τ Shear stress

τmax Maximum shear stress

Ψ Moment ratio (Mfx/Mo)

s Slope of the shell at the bottom

2a Width of LTA

2b Height of LTA

Abbreviations

2-D Two dimensional

3-D Three dimensional

APDL ANSYS parametric design language

API American petroleum institute

ASME American society of mechanical Engineers

BKIN Bilinear kinematic hardening

CTP Critical thickness profile

COV Coefficient of variation

xix
DOF Degree of freedom

EPFEA Elastic-Plastic finite element analysis

FCA Future corrosion allowance

FEA Finite element analysis

FFS Fitness for service

LEFEA Linear elastic finite element analysis

LF Load factor

LTA Locally thinned area

LTAH LTA height

LTAW LTA width

MAWP Maximum allowable working pressure

MAWPr Reduced maximum allowable working pressure

MFH Maximum fill height

MPC Multi point constraint

PTR Point thickness reading

RSF Remaining strength factor

RSFa Allowable remaining strength factor

xx
CHAPTER 1

INTRODUCTION

1.0 BACKGROUND

Storage tanks are part of many industries. Tanks are used for storing petroleum

products or any fluids/gases at ambient/elevated/low temperatures. Tanks could be either

above ground or below ground. Above ground storage tanks (AST) are constructed just

above the ground or at a higher elevation using structural supports. In an industrial setup,

both above and below ground tanks are widely used to store water, petroleum or other

chemical fluids. Generally tanks above ground have much higher capacity than below

ground tanks. The above ground steel tanks can either be vertical or horizontal. Most

large ASTs are generally vertical cylindrical vessels which are built with flat bottoms that

rest directly on prepared subgrades or concrete ring wall with suitable infill. Ambient

temperature tanks store and operate the infill liquid at atmospheric temperature whereas

elevated temperature tanks have arrangements to heat the liquid while in storage. In

petroleum industries, tanks are extensively used to hold liquids like asphalt, residuum,

high pour point hydrocarbons, etc., at higher temperature than the ambient. In the U.S

alone it is estimated that over 700 million gallons of petroleum is consumed daily. The

number of ASTs in the US is in the range of 950,000 to 1,100,000. In Canada there could

be another 100,000. Among these, shop built ASTs account for approximately 85% and

the remaining 15% are field erected. For ease of definition, shop-built ASTs are

1
generally considered to be available in capacities less than 1000 barrels (42,000 gallons)

and less than 15 feet in diameter [Digrado and Thorp, 2004]. Tanks of larger capacity are

fabricated at site. Design life and failure of such storage tanks has a major impact on the

process cycle and the environment. The design and fitness-for-service procedures are a

concern for many international standards and are continually improved upon to ensure

better safety and serviceability. This thesis focuses on the design and fitness for service

issues of above ground cylindrical (vertical) steel tanks used in petroleum and other

industries.

The analysis and design of such tanks in North America is generally based on

American Petroleum Institute standard API 650 [2012]. This standard is also used or

referenced to extensively in other countries and is sometimes seen as “de facto”

international standard. It governs the design, erection and testing requirements of above

ground storage tanks operating at ambient atmospheric pressure or slightly higher

pressures. The wall of such a tank can have a single thickness throughout its height or

several shell courses with varying thicknesses. The design principle is based on small

deflection theory of thin walled cylindrical shells. The wall is predominantly designed for

the hoop stresses due to the tank liquid, with some considerations for discontinuity

stresses at joints. The bottom of the tank is a thin plate generally considered to be a

membrane without any structural requirements. However this thin plate is subjected to

very high bending near the shell joint due to shell rotation and can cause failure.

Although API 650 has been in use for a long time, just like any other design standard, it

needs periodic evaluation to examine or re-examine its provisions. The present thesis is

2
aimed at this objective. Some of the current technical issues in tanks are briefly discussed

below.

The bottom plate of the tank extends to a small length outside the shell wall. This

projection length is generally prescribed for welding requirements. It influences the

stresses in the tank wall/ bottom plate. Hence a detailed study is needed for selecting this

projection length and examining relevant issues. At present there are no detailed studies

on this issue. In many situations, tanks are used at operating temperatures ranging from

ambient to 500 F (260 C). Such ‘elevated temperature tanks’ have significant thermal

stresses especially near the shell to bottom discontinuity. The magnitude of thermal stress

occurring in the tank wall and the bottom plate is influenced by the amount of radial

restraint, which in turn depends upon factors like liquid pressure, friction between the

bottom plate and the foundation and differential settlement, if any. It is reported that this

zone undergoes low cycle fatigue [Karcher, 1978a, b] and hence needs to be analyzed in

detail for such situations. API 650 only provides minimum guidelines for elevated

temperature tanks and hence the designer needs to do a rigorous analysis if the conditions

are critical.

Like any other process equipment, damage due to erosion/corrosion is a common

problem in the tanks. Localized corrosion damages and thermal hot spots are typical of

damage that occurs in ageing pressure vessels, piping or storage tanks. In the case of

above ground storage tanks, an average size tank may cost around one-half million

dollars to construct/replace and substantial portion would go for repair and rehabilitation,

if required [Andreani, et al., 1995]. In addition to the cost, issues concerning logistics, the

3
effect on the process cycle due to out of service tanks, safety permits, etc., cause huge

losses in resources and time that cannot be exactly accounted for. Hence, periodic

inspection of the tanks for structural health and safety is part of the engineer’s duties. The

engineer or tank operator is expected to periodically monitor the structural health and

integrity, asses the degradation and take decisions regarding repairs/replacement. Fitness

for service assessments are the primary tools adopted to carry out this task. They are

quantitative and qualitative engineering evaluations performed to determine the structural

integrity of an in-service component containing damage and estimate the remaining life

of degraded components and to make run/repair/replace decisions. They can be applied to

storage tanks, pressure vessels, boilers, piping systems and pipelines, etc. American

Petroleum Institute (API) standards API 579 & 653 provide a general procedure for

assessing fitness for service in tanks. Based on the problem, different levels of fitness for

service need to be employed to ensure the safety and serviceability of tank. Practitioners

usually proceed sequentially from Level 1 to Level 3 analysis (unless otherwise directed

by the assessment techniques). Level 1 is the most conservative, but is easiest to use. The

engineer proceeds to the next level, if the current assessment level does not provide an

acceptable result or a clear course of action cannot be determined.

1.1 NEED FOR RESEARCH

As mentioned above, the shell to bottom joint in tanks is an important location

where the discontinuity stresses are very high and tanks have historically failed at this

location [Cornell and Baker, 2002]. In large tanks, the bottom plate near this joint is

usually thickened for a small width (Fig.1.1), whereas in small tanks the same thickness

4
is provided throughout. The thickened portion of bottom plate is termed as “annular

plate”.

Radius (R)

Height (H)
Tank Shell

Hydro static Pressure

Bottom plate Annular Plate

tb a ta
Ka
ts
Fig. 1.1 Tank with Annular Plate

The designer needs to determine the width of the annular plate from the inner

surface of the shell towards the tank centre (Ka), the thickness of the plate (ta) and the

projection length of the plate outside the shell (a).

Wu and Liu [2000], report that codes and standards in many countries do not have

equations to design the annular plate. Denham, et al., [1968 a, b] proposed a model for

5
finding the elastic stresses near the shell to bottom joint. This model, based on beam and

shell theory, was developed for a fixed annular plate projection length (3”) and the

theoretical results were compared with strain gage measurements. In case of shell, the

measured stresses confirmed the validity of the theoretical approach. For annular plate,

the field results indicated that it could reach yield close to the shell, and hence a low

cycle fatigue situation could be expected near this joint. The study indicated that the

projection could influences the stresses, but the effect of changing the projection length

and its upper limit on the stresses was not studied.

In case of elevated temperature tanks, the thermal stresses induced due to partial

expansion of the tank was studied by Karcher [1978a]. The restraint causing partial

expansion at the bottom was addressed using an expansion factor called ‘C’. However,

there are no clear guidelines to choose a particular value for this factor for a given

situation. Hence the influence of ‘C’ on tank stresses and fatigue life has to be clearly

understood and explicit guidelines are needed to choose a value for this factor. It can be

shown that ‘C’ is directly related to coefficient of friction between tank bottom and the

foundation. At present this is not explored in any systematic manner. Another significant

issue in the current practice is the use of beam-on-elastic foundation theory with tanks

resting on concrete ring walls. The design community has largely ignored this

discrepancy. This needs closer examination.

As mentioned earlier, locally thinned areas (LTA) due to corrosion is a common

occurrence in pressure vessels and tanks. Application of simplified limit load methods in

fitness for service (FFS) evaluation of LTA is still in the initial stages and can be

6
explored further to improve the accuracy of the results. Moreover in the current practice,

the FFS procedures are mainly calibrated for pipes and merely used for tanks without

much verification. It would be good to develop procedures specifically for large tanks

and their special type of geometry and loading. This is especially so since the R/t ratios

of tank could well exceed 1000 whereas most pipes fall in the range of 20~100.

1.2 OBJECTIVES OF RESEARCH

The following objectives have been selected for the current research:

1. Study the influence of annular plate on the internal moment at the shell to

bottom joint of tanks and determine the upper design limit of the annular plate

projection length outside tank wall.

2. Study the influence of friction forces below the tank bottom and propose a more

rational procedure to determine the value of the expansion factor.

3. Improve the existing procedure for fatigue design of tanks on concrete ring wall

foundation vis-à-vis the use of beam-on-elastic foundation theory.

4. Explore the use of simplified fitness for service methods for tanks with LTA

using the mα-tangent multiplier.

1.3 SCOPE OF THE STUDY

This study applies to tanks in non-refrigerated service with internal pressures

approximating atmospheric pressure. The analyzes of the tanks are carried out for

hydrostatic test condition, i.e., the infill liquid is assumed to be water for finding relevant

stresses. However the treatment is also valid for any other liquid. For elevated

7
temperature tanks the temperature range is considered between 200F (93C) and 500F

(260C). The foundation of the tank is assumed to have concrete ring wall with well

compacted infill. In the finite element analysis, the tank is modeled using elastic-plastic

ductile material which is able to absorb significant deformation beyond the elastic limit

without the danger of fracture. Strain hardening is assumed to be small.

For the study of locally thinned areas, the flaw area is considered with the

remaining thickness ratio (defined as the ratio of the corroded wall thickness to the

nominal thickness) not less than 0.5. The discontinuity of corrosion damage is assumed to

be tapered down smoothly and corrosion damage is considered as blunt patch (non-crack

like flaw). For ease of analysis, the shape of the flaw is assumed to be rectangular with

different aspect ratios. The location of the flaw is assumed to be remote from other major

structural discontinuities, such as nozzles, or geometry changes. The tanks studied in the

current work are assumed to be designed and constructed in accordance with a

recognized code or standard such as API 650.

1.4 STRUCTURE OF THE THESIS

This thesis is organized in eight Chapters. The first Chapter introduces the

background to the problem, the need for research, objectives and scope of the study.

The second Chapter presents a general review of literature. It briefly reviews the basic

theories associated with tank analysis and design. The tanks are designed similar to any

mechanical component or pressure vessel, and hence a short summary of different failure

theories of material is provided. Even though the tank bottom is not designed for

8
structural loads, the zone near the shell should safely carry the localized bending stresses

and be checked for fatigue due to cyclic loading. The local bending at the shell to bottom

joint, causes yielding and shakedown in the bottom plate. The formulas used for this

evaluation are obtained considering the bottom plate as a beam-on-elastic foundation and

hence the basics of beam-on-elastic foundation are included. The tank wall is a thin shell

structure and all the design and analysis is carried out using shell theory and therefore

basic shell theory is reviewed. The concept of fitness for service and its application to

AST is introduced. Finally the survey of previously reported research works relevant to

the scope of the current research is included.

In Chapter 3 finite element modeling details of the present study including

geometry, constraints, material models and samples of typical mesh are provided. Details

about the friction model used for interaction between bottom plate and foundation are

explained. The problems encountered during modeling and the solution techniques

adopted are discussed briefly.

Chapter 4 introduces the basic tank design procedures and describes the behavior

of shell to bottom joint, the purpose of annular plate and its design basis. The mechanism

of formation of plastic hinges near the shell to bottom joint is explained. The influence

of the annular plate projection length on tank stresses is studied. The procedure to

determine the length of the annular plate projection (required to minimize the stresses in

the shell) is provided and the theoretical results are compared with the finite element

Model results.

9
 Chapter 5 discusses the friction interaction between bottom plate and foundation.

The procedure by Karcher that uses a factor called ‘C’ to account for the partial

expansion of the tank is explained. A more rational alternative procedure using friction

coefficient instead of C factor is proposed. Theoretical results are validated using finite

element analysis.

Chapter 6 presents the procedure for fatigue evaluation of shell to bottom joint

in tanks. It discusses about the suitability of using beam-on-elastic foundation theory for

determining the peak stress values in tanks on concrete ring walls. The current procedure

is modified by replacing the beam-on-elastic foundation theory with beam theory to

reflect the bottom plate uplift near this zone. The results of the modified procedure are

compared with the existing procedure and finite element analysis.

Chapter 7 discusses a modified reference volume scheme for determination of

remaining strength factor (RSF) in tanks using the mα-tangent method. The RSF

determined using modified reference volume is compared with the API procedure and the

elastic-plastic FEA results. The circumferential decay length for large tanks with high R/t

ratio is studied using FEA and a suitable value is recommended.

Chapter 8 summarizes the current work and gives conclusions and

recommendations for future work.

10
CHAPTER 2

LITERATURE REVIEW

This Chapter reviews briefly the basic theories associated with tank analysis and

design. Since tanks are basically thin cylindrical shells, the basic shell theory is presented

here. A brief summary of failure theories, basics of beam-on-elastic foundation theory

and concepts of shakedown analysis are presented. Pertinent concepts about fitness for

service procedures and their application to storage tanks are introduced. The concept of

limit analysis, limit load multipliers and the estimation of remaining strength factor of

damaged tanks using limit analysis are also explained. A survey of previously reported

research works relevant to the scope of current research is included.

2.0 THEORIES OF FAILURE

Structural members subjected to loads may fail to perform their intended function

in various ways. Depending on the loading, geometry and material properties they may

fail because of excessive deformation, material yielding, fracture, instability, etc.

Yielding of the material is an important mode of failure in many components. For

unidirectional stress field, the yield strength obtained from a standard uniaxial test can be

the criterion. But for multi-axial state of stress, the yielding is governed by some quantity

representing the state of strain, stress, components of strain energy, etc. Hence, the yield

criterion is usually expressed in mathematical form by means of a yield function ƒ(σij,

11
σy,), where σij is the state of stress and σy is the uniaxial yield strength. ƒ(σij, σy) = 0

implies that yield boundary has been reached and ƒ(σij, σy) < 0 indicates that the stress

state is elastic. The yield function is developed such that the components of multi-axial

stress can be combined in to a single quantity and termed as effective stress σe. The

effective stress is then compared with the yield strength σy (obtained from uniaxial test),

in some appropriate form, to determine if yield has occurred. The following are some of

the yield criteria that are often used in engineering practice. The following is a simple

summary of the material found in solid mechanics text books.

2.0.1 Maximum Shear Stress Criterion

The maximum shear-stress criterion, also known as the Tresca criterion, states

that yielding begins when the maximum shear stress at a point in the structure equals the

maximum shear stress at yield in uniaxial tension (or compression). The maximum shear

stress (τmax) is given by half of the difference between maximum and minimum principal

stress components.

e  max = (σ1- σ3)/2 (2.1)

ƒ = (τmax – σy/2) (2.2)

In ductile metals, the crystals have slip planes along which the resistance to shear

force is the weakest. Hence yield criterion based on shear stress is more appropriate for

ductile metals. The Tresca criterion generally gives conservative results for metals.

Because it is simple to use, many engineers and codes (e.g., ASME B&PV) prefer it for

12
metal structures. The yield surface for this criterion is a regular hexagon in 2D principal

stress space as shown in Fig. 2.1(a).

2.0.2 Distortional Energy Density Criterion

This criterion, also referred as von Mises criterion or octahedral shear stress

criterion states that yielding begins when the distortional strain energy density at a point

in the structure equals the distortional strain energy density at yield in uniaxial tension (or

compression). This criterion assumes that the energy associated with volume change has

negligible effect on yielding and the energy associated with distortion alone is

responsible for the yielding of material. The distortional energy density UD is expressed

as

1
UD  (1  2 )2  (2  3 )2  (3  1 ) 2 
12G m (2.3)

where, Gm is the shear modulus. Using this, the effective (von Mises stress) is given by

1
e  (1  2 )2  (2  3 )2  (3  1 )2 
2 (2.4)

The yield function then becomes

ƒ=  e2   y2
(2.5)

13
 2 2 = y 2
y
y
1 = y
y y
y y
1
1

y 2 = 1 - y 1  4 y  31
2 2

y 2 
2

(a) (b)

Fig. 2.1 Yield Criteria in σ1 - σ2 plane: (a) Tresca Criterion, (b) von-Mises Criterion

The von Mises or Tresca criteria are more suitable for ductile metals and predict

the initiation of yielding quite well and are the most popular and hence are presented

here. However the von Mises criterion is slightly more accurate than Tresca criterion and

has a smooth profile unlike the Tresca function. Since the von Mises yield function is

continuously differentiable, it is preferred in computational plasticity studies in which

plastic flow and strain hardening are considered. Similarly there are many other theories

of failure suitable for different types of materials. The current work uses von Mises

exclusively.

2.1 SHELL THEORY

In the present research, Above Ground Storage Tanks (AST) of cylindrical shape

are studied. The theory of cylindrical shells is used in the analysis and design of AST.

Hence a brief summary of this theory is presented here. Detailed analysis of this theory is

presented by many authors e.g., Timoshenko and Woinowsky-Kreiger [1959].

14
 Mx M
ys
Nx N
.
z
M dM x xs
Mx  dxs
dxs
dN x
Nx  dxs
dxs dxs
dQx
N
Qx  dxs
dxs
R

d

Fig. 2.2 Forces and Moments Acting on a Cylindrical Shell Element

Consider the equilibrium of an element of the shell as shown in Fig. 2.2. The term

N x indicates membrane force in the plane perpendicular to ‘xs’ axis and in the direction

of φ. Because of axi-symmetry, the membrane shearing forces N x & Nx vanish and

hence the hoop force N  is constant along the circumference. Similarly, because of

symmetry the twisting moments M x & Mx vanish and M  remains constant along the

circumference. Assuming that the external forces consist only of pressure (Z) normal to

the surface, the following equilibrium equations are obtained

dN x
R dx s d  0 (2.6)
dx s

dQx
R dx s d  N  dx s d  Z R dx s d  0 (2.7)
dx s

15
dM x
R dx s d  Qx R dx s d  0 (2.8)
dx s

Eq. 2.6 indicates that force Nx is constant. It can be taken as zero if the self-weight

and stress due to forces applied axially on the cylinder is neglected. If they are

significant, they can be superimposed on top of other stresses appropriately. Equations

2.7 and 2.8 give, respectively,

dQ x 1
 N  Z (2.9)
dx s R

dM x
 Qx  0 (2.10)
dx s

Since the problem is axisymmetric, the displacement along the circumferential

direction is zero. Hence, only the displacements ux and ys in the xs and y directions,

respectively, exist.

Using Hooke’s law,

Et s Et s du x y
Nx  ( x   y )  (  s )  0 (2.11)
1  2
1   dx s
2
R

Et s Et s y du
N  (   x )  ( s   x ) (2.12)
1  2
1  2
R dx s

where ts is the thickness of the shell. From Eq. 2.11 and 2.12,

Et s ys
N   (2.13)
R

16
 Since there is no change in curvature in the circumferential direction, the
d 2 y s
curvature in the xs-direction is equal to . Hence
dx s 2

d 2 ys Et s 3
M x  D , M   M , where, Ds = (2.14)
12(1   2 )
x
dx s 2

Using Eqns. (2.8) and (2.9) and eliminating Qx,

d2 d 2 ys Et s
(Ds )  2 ys  Z . (2.15)
dx s 2 dx s 2 R

If Ds doesn’t vary along xs,

d 4 ys Z 3(1   2 )
 44 ys  where,  
4
, (2.16)
dx s 4
Ds R 2 t s2

This equation is similar to that of a beam-on-elastic foundation, as explained in

the next section. The general solution of Eq. 2.16 is given by

ys  exs (C1 cos xs  C2 sin xs )  ex (C3 cos xs  C4 sin xs )  f (x s ) (2.17)

where, f(xs) is the particular solution and C1 to C4 are constants of integration. These

constants have to be determined from the boundary conditions of the shell. For

cylindrical tanks with uniform wall thickness, the pressure acting normal to the shell

surface is the hydrostatic load Z = (H  x s ) , where,  is the specific weight of the

liquid and H is the height of the liquid. Eq. 2.16 can be written as

d 4 ys  (H  x s )
 4 4
y s 
dx s 4 Ds (2.18)

17
  (H  x s ) (H  x s )R 2
The particular solution for the above equation will be ys     .
44 Ds Et s

This expression represents the radial expansion of the cylindrical shell with free edge

under the action of hoop stresses. The constants C1 and C2 are zero if the shell is long.

Hence Eq. 2.17 can be written as

x s  (H  x s )R 2
ys  e (C3 cos x s  C4 sin x s )  (2.19)
Et s

The constants C3 and C4 can be obtained from the boundary condition at the bottom of the
tank.

2.2 BEAM-ON-ELASTIC FOUNDATION

Beam-on-elastic foundation theory is used to analyze structures that can be

idealized as a beam of relatively low stiffness placed on a flexible foundation and loads

are applied on the beam. This theory finds applications in a variety of practical

engineering problems like a rail road placed on soil subgrade, floor systems with beams

(as used in ships), buildings, bridges and components made of thin shells of revolution

like tank walls, boilers, etc. In ASTs, this theory is also used to determine the minimum

length of the annular plate [Karcher, 1978b], the fatigue life of the shell to bottom joint of

tanks, etc.

The theory is based on the assumption that the reaction forces of the foundation

are proportional at every point to the deflection of the beam at that point. This assumption

was first introduced by Winkler in 1867. The vertical deformation characteristics of the

foundation are defined by means of identical, independent, closely spaced, discrete and

18
linearly elastic springs. The constant of proportionality of these springs is known as the

modulus of sub grade reaction (K). The Winkler model, which has been originally

developed for the analysis of railroad tracks, is very simple and does not accurately

represent the characteristics of many practical foundations. One of the most important

deficiencies of the Winkler model is that a displacement discontinuity appears between

the loaded and the unloaded part of the foundation surface. In reality, the soil surface

does not show any discontinuity as shown in Fig. 2.3(b).

(a) Winkler Foundation (b) Practical Soil Foundations

Fig. 2.3 Deflections of Elastic Foundations under Uniform Pressure

q(x)x

M V x M+M V+V
M

(k w)x

Fig. 2.4 Beam-on-Elastic Foundation

19
 Various models have been proposed by researchers [Wang, 2005] to overcome

this deficiency in Winkler’s model by introducing the interactions between the individual

springs using interconnecting elements like beams or plates. In the model proposed by

Hetényi [1947], interaction between the independent spring elements is accomplished by

incorporating an elastic beam (2-D) or an elastic plate (3-D), that can deform only in

bending.

Considering the equilibrium of an infinitesimal element of the beam-on-elastic

foundation as shown in Fig. 2.4, the vertical equilibrium leads to

dV
=Ky- q(x)
dx (2.20)

Similarly, moment equilibrium leads to

dM
V
dx (2.21)

From the small deflection beam theory, the moment curvature relationship is

d2 y
M  EI (2.22)
dx 2

Hence

d4 y q K
4
 44 y  where   4
dx EI 4 EI (2.23)

The general solution of the equation is given by

y  ex (C1 cos x  C2 sin x)  ex (C3 cos x  C4 sin x)  w(q) (2.24)

20
where, y(q) is the particular solution associated with the applied load q(x), and disappears

when q = 0. It can be seen that both the beam-on-elastic foundation theory and the shell

theory gives the same governing equation. The beam on elastic foundation theory will be

used later for discussing annular plate behavior of tanks.

2.3 SHAKEDOWN ANALYSIS

In studies on tanks, researchers [Karcher, 1981a, Long, 2004] have assumed that

shakedown occurs at the shell to bottom plate joint. A brief review of shakedown in

general is given here.

Shakedown is a phenomenon, that can occur in structures/components (of ductile

material) when they are subjected to repetitive loads. Apart from the regular modes of

failure that are expected in static loading conditions, failure due to fatigue, alternate

plasticity or ratcheting can occur due to such repetitive loads. When the cyclic loads

cause stresses beyond the yield strength, some sort of re-adjustment of stress distributions

can happen through limited plastic flow which results in residual stresses that will

minimize the plastic fatigue strains in subsequent cycles [Calladine, 2000]. Hence the

study of shakedown performance of the structure is needed for design considerations.

For example, if a load ‘q’ causes the structure to begin yielding, and if the load is

further increased to 1.25q and brought down to zero load to complete the first load cycle,

then the residual stresses set up in the structure will enable the structure to respond to the

second load cycle by purely elastic action completely up to the load of 1.25q. The stress

distributions tend to adjust themselves under cyclic loading, to an elastic state whenever

21
possible, since for a given amount of external work performed by the applied loading, the

internal strain energy will be a minimum when the stress distribution is linearly elastic.

Hence the structure ‘shakes down’ to elastic behavior.

The shakedown limit, i.e., the load up to which the structure will shakedown to

elastic behavior depends upon the material, the structure geometry and the loading

pattern. Similar limits exist for alternate plasticity and ratcheting. Generally in pressure

components, the plasticity/shakedown occurs near the discontinuities because of the

secondary stress effects. Since the secondary stresses are self-equilibrating stresses, the

shakedown limit can be found using elastic analysis. During the initial loading of the

structure, if the stresses computed on elastic basis are everywhere less than twice yield

stress, then it can be ensured that the structure will shake down [Burgreen, 1975].

2.4 LOW CYCLE FATIGUE

Fatigue is a progressive and localized structural damage that occurs when a

material is subjected to cyclic loading. The nominal stress values that cause such damage

need not necessarily be equal to the yield/ultimate strength of the material. Fatigue is

classified as High cycle (HCF) and Low cycle fatigue (LCF). Failure in HCF is caused by

lower amplitude of stress level and higher number of load cycles (>104) whereas in LCF,

it is higher amplitude of stress level (generally higher than yield strength) and lower

number of load cycles (<104). For the LCF case, when the plastic deformation occurs, the

accounting of the loading in terms of stress is less useful and hence the strain in the

22
material offers a simpler and more accurate description. Low-cycle fatigue is usually

characterized by the Manson-Coffin relation as given below [Wiki, 2013]:

 p
 

  'f 2 N f (2.25)
2

where, Δεp /2 is the plastic strain amplitude; εf' is an empirical constant known as the

fatigue ductility coefficient, the failure strain for a single reversal; 2Nf is the number of

reversals to failure (Nf cycles), η is an empirical constant known as the fatigue ductility

exponent, commonly ranging from -0.5 to -0.7 for metals.

Fig. 2.5 Manson-Coffin Strain – Life relation for AISI 304 Stainless Steel [Manson &
Halford, 2006]

The above relationship was obtained by Manson and Coffin independently in the

1950s from the results of experiments which suggested that a logarithmically linear

relation between plastic range and fatigue life exists as shown in Fig. 2.5. Since

determination of plastic strain alone was difficult for practical engineering problems, the

23
plastic strain was later replaced by the total strain appropriately [Manson & Halford,

2006]. This work gave the fundamental insight that fatigue life as a function of plastic

strain reversal and formed the basis of all the future work in fatigue. Langer [1962]

proposed the following expression relating the total alternating strain and the fatigue life.

εa = A1(N)-n1 + A2 (2.26)

where, Δεa is the applied strain amplitude, N is the fatigue life and

A1, A2,n1 are coefficients of the model. The Equation 2.26 can also be expressed in

terms of stress amplitude Sa as

E  100 
Sa  In   B (2.27)
4 N  100  A 

where E is the elastic modulus, A and B are constants related to reduction in area in a

tensile test and endurance limit of the material at 107 cycles. The strain amplitude is

converted to stress amplitude using material modulus i.e. Sa = E εa.

The ASME fatigue design curves were obtained by strain controlled experiments

based on Langer’s expression. The design curves have been developed as the best fit

curves to the experimental data [Chopra, K., and Shack, W. J., 2003].

2.5 TANK RESEARCH

A number of researchers have worked on various aspects of tank design and

analysis. The current section discusses the works that are relevant to the proposed

research.

Zick and McGrath [1968] carried out detailed theoretical studies to understand

the stresses in the shell wall of cylindrical storage tanks and proposed the variable design

24
point method of shell design which was adopted by the API. In their study, they also

considered the effect of bottom plate on shell stresses and concluded that both the

restraining/resisting moment (offered by the bottom plate) and radial growth at the joint

reduce the circumferential stresses in the shell near the bottom, but to a small degree.

This implies that the higher the restraining moment, the (slightly) lower the stresses in the

shell. Regarding the plastic hinges in the bottom plate, Zick and McGrath [1968] say

“where the resisting moment of the tank bottom is to be evaluated, a reasonable

approach would be to use the full yielding moment in the bottom plate on one side for an

earth grade foundation and to use two moments (one on each side) for a ring wall

foundation.”

In their analysis, for the bottom boundary condition of the shell, they assumed a

restraining moment equal to one yielding moment (MP) of a bottom plate thickness

(supposedly standard thickness, although the details were not specified) without any

radial growth. This assumption will be conservative since occurrence of the second hinge

will further reduce the stresses in the shell for the most part, which will be explained later

in this report.

Denham, et al. [1968a, b], proposed a method to determine the restraining

moment at the shell to bottom joint and the stresses in shell and bottom plate, when the

stresses are elastic. The method is based on beam and shell theory and the derivation is

based on the concept of establishing slope compatibility for the shell and bottom plate at

the shell to bottom joint. The authors compared their theoretical data with actual strain

measurements from an instrumented tank and stated that measured shell stresses

25
confirmed the validity of theoretical approach. In the annular plates, the theoretical

calculations indicated that the stresses would reach yield close to the shell and this was

also confirmed from the field data. But for locations away from the shell to bottom joint,

the field data showed considerable variation. The theoretical model was not verified with

numerical models like FEA.

Karcher (1978a, 1981a) proposed a method to find the stresses in the shell and the

bottom plate for elevated temperature tanks. The method is based on the elastic shell

theory for the tank wall and beam-on-elastic foundation theory for the bottom plate. It

assumes that complete plastic hinges are formed in the bottom plate near the shell to

bottom joint and that the shell stresses are elastic. This is justified since the tanks are

constructed such that bottom plate is very thin compared to shell wall and, it is assumed

that during loading the bottom plate completely yields and forms plastic hinge before the

shell wall yields. The equations as given by Timoshenko [1959] for cylindrical tanks with

clamped end conditions were suitably modified to have the plastic moment at bottom as

the boundary condition instead of the clamped condition. The plastic hinge condition was

adopted from the work by Zick and McGrath [1968]. A single plastic hinge is assumed

for earthen foundation, whereas two hinges are assumed to form in case of concrete ring

wall foundation. The recommendations by Karcher are included in API 650 [2012],

Appendix M “Requirements of Tanks Operating at Elevated Temperatures” for

evaluating the design cyclic life of tanks.

Jones and Seshadri [1989] studied the validity of Karcher’s equations using linear

finite element method and found them to be satisfactory in predicting the stresses in shell.

26
They concluded that Karcher’s analysis based on a single plastic hinge is conservative for

deflection and fatigue life estimates. The authors opined that, to properly assess the

formation of plastic hinges in the annular-plate, a pseudo-elastic analysis be carried out

by modeling the subgrade. The authors have used a pseudo elastic tank wall model to

compare the stresses with Karcher’s theory, i.e., plastic moments are applied at the tank

bottom as shown in Fig.2.6 Though this will give a good understanding of the wall

stresses for a given bottom moment, it may not reflect the exact way in which the hinges,

if any, are formed in bottom plate and loads are distributed in the actual tank.

MP MP

Fig. 2.6 Pseudo Elastic Tank Model

Wu [1996 and 2000] proposed a method similar to Denham’s method but had

taken into account the effect of elastic foundation beneath the bottom plate near the joint.

The inside part of the annular plate towards the tank centre was considered as a semi-

infinite plate flexibly supported by the foundation. The reaction force of the subgrade was

assumed to be linear with the annular plate vertical displacement. He also compared the

27
theoretical results with field data. The author also commented that, based on the proposed

method and Denham’s model, the wider the width of the bottom plate that projects

beyond the outer surface of the shell, the smaller is the maximum stress in the annular

plate. However, guidelines to find the maximum projection length were not given. The

other important aspect mentioned by the study was that in practice the shell is assumed to

coincide with the centre line of the ring wall and hence the load from the shell acts on the

centre line of the wall, but in actually the load gets applied at the peripheral point of the

annular plate causing eccentricity in load application. This was also confirmed from field

data on the soil pressure beneath the ring wall. The study did not report any numerical

analysis to compare with the field data and the proposed theory.

Kim, et al., [2009] studied failure of oil storage tank due to crack in the fillet

weldment between the reactor shell and annular plates. Failure studies such as

fractography, tensile test, hardness test, corrosion test and chemical analysis were

performed. They conclude that although cracks were initiated by corrosion, failure was

generated by propagation of crack caused by stress concentration. It was recommended

that the local stresses be reduced by improved weld toe geometry and reinforcement be

used to reduce the stress concentration.

Chen, et al., [2007] proposed a simplified long-short shell method to determine

the stresses in multiple layer shell wall. The classical short shell method as proposed by

Timoshenko and others is cumbersome when the number of layers are more, while the

long shell method is simple but the results are erroneous near the joints. Hence the

authors have proposed a model in such a way that, the first shell is regarded as a short

28
shell while the others are long cylindrical shells. The authors support the results by

comparing the field measured strain data from a 150,000 m3 oil tank during its hydro

static test.

2.6 FITNESS FOR SERVICE

In the case of above ground storage tanks, a typical tank may cost in the millions

of dollars to construct/replace and a substantial portion of that would be needed for repair

and rehabilitation [Andreani, et al., 1995]. In addition to the cost, issues concerning

logistics, environmental impact, effect on the process cycle due to out of service tanks,

safety permits, etc., cause huge losses in resources and time that cannot be exactly

accounted for. If the tank fails by flooding, explosion, etc., it creates a severe impact on

the environment and the surrounding community. It sets in a series of crisis like situations

at multiple levels adversely affecting the lives of the population [Comfort, et al., 1989].

Hence, periodic inspection of the tanks for structural health and safety is a part of the

engineering duties. The engineer is expected to periodically monitor the structural health

and integrity, asses the degradation and take decisions regarding repairs/replacement.

Fitness for service assessments are the primary methods to do the task. They are

quantitative and qualitative engineering evaluations performed to demonstrate the

structural integrity of an in-service component containing damage, determine the

remaining life of degraded components and make run/repair/replace decisions.

Common reasons for assessing the fitness for service of equipment include the

discovery of a flaw such as a locally thin area or crack or corrosion, settlements, failure to

29
meet current design standards, and plans for operating under more severe conditions than

originally expected. Fitness for service assessment applies analytical methods to evaluate

flaws, damage, and material aging. The analytical methods are based on stress analysis,

but they also require information on equipment operations, nondestructive examination

(NDE), and material properties. Stress analysis can be performed using standard

handbook or design code formulas or by means of finite element analysis (FEA), which is

increasingly becoming more common. The main results of the assessment are (1) a

decision to run, alter, repair, monitor, or replace the equipment and (2) guidance on

inspection interval for the equipment. Engineers and tank users can take decisions based

on engineering inputs, whether to carry out repairs immediately or at a later time or go

for immediate replacement. Fitness for service assessment requires both knowledge of

past operating conditions and a forecast of future operating conditions. American

Petroleum Institute (API) codes of practice API 579-1/ASME FFS-1/653 provide a

detailed procedure for assessing fitness for service in tanks.

2.6.1 Levels of FFS Assessment

Three Levels of assessment are provided in API 579-1/ASME FFS-1 [2009]

standard. In general, each assessment level provides a balance between conservatism, the

amount of information required for the evaluation, the skill of the personnel performing

the assessment, and the complexity of analysis being performed. Level 1 is the most

conservative, but is easiest to use. Practitioners usually proceed sequentially from a

Level 1 to a Level 3 analysis (unless otherwise directed by the assessment techniques) if

30
the current assessment level does not provide a clear result, or a course of action cannot

be determined.

Level 1 Assessment

The assessment procedures included in this level are intended to provide

conservative screening criteria that can be utilized with a minimum amount of inspection

or component information. A Level 1 assessment may be performed either by plant

inspectors or engineering personnel. The only load considered is internal pressure, and a

single thickness with one or two surface area dimensions are used to characterize the

local metal loss. Level 1 assessments are limited to components covered by a recognized

code or standard which have a design equation that specifically relates pressure (or liquid

fill height for tanks) to a required wall thickness. Hence it is not applicable for complex

loading or damage conditions.

Level 2 Assessment

The assessment procedures included in this level are intended to provide a more

detailed evaluation that produces results that are more precise than those from a Level 1

assessment. In a Level 2 assessment, inspection information similar to that required for a

Level 1 assessment is needed; however, more detailed calculations are used in the

evaluation.. More general loading is considered (e.g., net-section bending moments on a

cylindrical shell), and rules are provided for the evaluation of local metal loss at a nozzle

connection. Level 2 assessments would typically be conducted by plant engineers, or

engineering specialists experienced and knowledgeable in performing FFS assessments.

31
The Level 2 assessment rules provide a better estimate of the structural integrity of a

component when significant variations in the thickness profile occur within the region of

metal loss. Hence a component that fails to be fit for service from Level 1 assessment can

pass from a detailed Level 2. However, the Level 2 procedures still have some limitations

regarding the component type, location of damage, loading and damage type that can be

assessed.

Level 3 Assessment

The assessment procedures included in this level are intended to provide the most

detailed evaluation which produces results that are more precise than those from a

Level 2 assessment. In a Level 3 assessment, the most detailed inspection and component

information is typically required, and the recommended analysis is based on numerical

techniques such as the finite element method or physical testing when appropriate.

Level 3 assessment rules are intended to evaluate components with complex geometries,

regions of localized metal loss and/or components with details where only limited design

rules are provided in the original construction code or standard. A Level 3 assessment is

primarily intended for use by engineering specialists experienced and knowledgeable in

performing FFS assessments. Since advanced numerical procedures are used for stress

analysis, the limitations of level 1 or level 2 procedures are surpassed in this level.

2.6.2 FFS Procedures for Local Thin Areas

Damages due to corrosion/erosion in the form of blunt metal loss are a

widespread problem in pressure vessels. Elaborate research had been carried out to

32
address the FFS issues due to locally thinned areas. These LTA damages can be global

(over the entire area) or local (at certain locations). These damages are progressive and

may or may not adversely affect the safety of the equipment at a particular instant. Hence

FFS assessments are periodically carried out to evaluate the damages and ensure the

required safety and serviceability of the equipment. The fitness for service procedures for

wall thinning or metal loss is generally divided in to three categories:

1. General Metal Loss (GML)

2. Local Metal Loss (also referred to as local thin area or LTA)

3. Pitting

The objective of performing FFS assessment is to check for rupture, bulging and leakage.

The procedures ensure that the corroded (or eroded) component has sufficient strength to

resist applied loads and is sufficiently thick to prevent pinhole leaks. The rupture

prevention is based on the concept of metal reinforcement, i.e., the weak thin metal area

is reinforced by surrounding sound metal provided the thin metal area is not too large and

leak prevention is ensured by keeping the remaining wall thickness above a minimum

threshold.

The principal failure mode for a pressure vessel (including tanks) with LTA

subject to constant internal pressure is plastic collapse [WRC 465, 2001]. Plastic collapse

may be evaluated using elastic stress analysis, limit load analysis, plastic collapse

analysis or using the concept of remaining strength factor. The remaining strength factor

method has been adopted in API 579-1/ASME FFS-1 standard. It has proven to be

33
effective in several applications [WRC 505, 2005]. The method using limit analysis and

remaining strength factor are reviewed below.

2.6.2.1 Limit Load Analysis

Limit analysis is a design philosophy used for designing mechanical components

and engineering structures. The method is based on maintaining equilibrium of the

structure at all stages of loading and thereby determining the safe load called the limit

load just prior to plastic failure (unconstrained plastic flow) of the structure/component. It

is the load, that triggers overall structural instability or plastic collapse, corresponding to

a point in the load-displacement curve at which the rate of external work of applied load

does not balance the rate of plastic dissipation [WRC 505]. Limit analysis offers a more

realistic and economical design than the methods based purely on elastic analysis.

The limit load can be determined using analytical methods, detailed numerical

procedures or simplified methods. The analytical procedures use the bounding theorems

of limit analysis, while the numerical procedures employ the widely used finite element

analysis (or similar methods) to determine the limit load. The analytical procedure will

not be feasible for complicated structures and is restricted to simple geometry and

loading. Hence, nonlinear FEA is a widely adopted and recognised procedure for detailed

limit analysis. However, performing nonlinear FEA involves huge computing resources,

time and expertise. Simplified methods derived from the basic limit theorems and

variational plasticity concepts might be able to predict limit load using the much simpler

34
linear FEA. Limit analysis using simplified procedures can also be used in the fitness for

service procedures for pressure vessels and tanks [Seshadri, 2004 & 2005].

2.6.2.2 The mα- Tangent Method

Among numerous simplified procedures for estimation of limit load, this thesis

adopted the mα-tangent method developed by Seshadri and co-workers [e.g., Seshadri and

Hossain, 2008]. The procedure is briefly reviewed here. Detailed explanation is provided

in Chapter 7. The method is based on variational principles in plasticity and makes use of

a statically admissible stress field based on linear elastic analysis. Limit load multiplier is

a factor that scales the applied load proportionately to obtain the limit load. The mα -

tangent multiplier is a limit load multiplier that depends on the upper bound multiplier mo

and the lower bound multiplier mL and can be expressed as

m0
mT  (2.28)
1 0.2929 ( 1)

m0
where  
mL

The mo depends on the ‘reference volume’ of the structure whereas the mL is

based on the magnitude of maximum stress in the structure. The reference volume is the

part of the entire volume that actually participates in the plastic failure of the structure or

component. Recently, Ahmad and Seshadri [2010] demonstrated the use of this

simplified limit analysis procedure in FFS methods for damages in above ground tanks.

The detailed review of the procedure is given in Chapter 7.

35
2.6.2.3 Remaining Strength Factor Method

The use of remaining strength factor (RSF) for the assessment of LTA in pressure

vessels and tanks was proposed by Sims, et al., [1992]. The RSF factor is defined as:

Collapse Load of Damaged Component

RSF =
Collapse Load of Undamaged Component

In API standards, the RSF is used to define the acceptability of a damaged

component for continued service. The acceptability criteria can be quantitatively

established in terms of plastic collapse load using RSF in combination with closed form

solutions, limit analysis or elastic-plastic analysis depending on the complexity of the

assessment. The RSF can be used to estimate the failure pressure at the plastic collapse or

the reduced (or rerated) working pressure called the Maximum Allowable Working

Pressure (MAWP) of damaged components. The failure pressure of the damaged

component (Pf) is given by:

Pf = Pfo (RSF) (2.29)

where, Pfo is the failure pressure of the undamaged component. Similarly, the reduced

Maximum Allowable Working Pressure (MAWPr) can be calculated as follows:

 RSF 
MAWPr  MAWP   for RSF < RSFa (2.30)
 RSFa 

MAWPr = MAWP for RSF  RSFa (2.31)

where RSFa is the allowable Remaining Strength Factor. It is a function of the allowable

stress and the design criteria used in the original construction code. The API recommends

36
an RSFa value of 0.9 for equipment in process services. If conservative design procedure

with lower allowable stresses were used in the vessel construction, then a lower RSFa

value can be used.

2.6.3 API 579-1/ASME FFS-1 Metal Loss Assessment Procedure

Before any standardized procedures were introduced, regions of metal loss were

assessed using thickness averaging techniques. In these methods, since the flaw depth is

generally irregular and varies over the entire area, it is averaged over the flaw length (or

width) and a uniform depth is assumed for that length of LTA. Assessment techniques

were developed by Texas Eastern Transmission Corporation and AGA Pipeline research

committee in the late 1960s and later incorporated in the ASME codes as B31.G method.

Though the averaging technique is not accurate for complex damage profiles, it gives the

most conservative results of all the proposed methods [Janelle, 2005]. This method forms

the foundation for most of the local thin area assessments that are currently in use.

API Standard “Fitness for Service - API 579-1/ASME FFS-1, 2012” is a

compendium of consensus methods for reliable assessment of the structural integrity of

equipment containing identified flaws or damage. It provides standardised and

technically sound consensus approaches. It is written as a Recommended Practice rather

than as a mandatory standard or code and is to be used in conjunction with the refining

and petrochemical industry’s existing codes for pressure vessels, piping and aboveground

storage tanks (API 510, API 570 and API653).

37
 API 579-1/ASME FFS-1 uses modified thickness averaging rules as well as

specific problem based procedures. This assessment method has two main classifications

namely the Local Metal Loss and Global Metal Loss. The procedure has three levels of

assessment as mentioned in Section 2.5.1

2.6.3.1 Data Requirements for Characterizing Metal Loss

Several non-destructive techniques are available to inspect the extent of the metal

loss/leakage in tanks and pressure vessels. The choice depends upon the material, type of

flaw, access to surface, availability, cost, etc. Some of the inspection techniques used in

the process industry are visual examination, magnetic particle testing, radiographic

testing, ultrasonic testing, acoustic emission testing and thermography.

Fitness for service assessments for wall thinning cannot be performed from

thickness measurement at a single sample point. Usually ultrasonic thickness readings are

measured in a grid with a minimum spacing equal to twice the nominal wall thickness of

the vessel [Antaki, 2005]. The region of metal loss can be characterised by two thickness

measurement techniques, namely point thickness reading (PTR) and critical thickness

profile (CTP). PTR is random sampling of thickness measurement that can be used only

if the variation in thickness readings is statistically small. The variation in the thickness

reading is expressed using Coefficient of Variation (COV), which is defined as the

standard deviation of a sample divided by the mean of the sample. If the COV of the

thickness reading population minus the Future Corrosion Allowance (FCA) is less than

10%, then the metal loss can be considered as uniform over the area and hence the

38
average thickness value calculated directly from the population can be used for FFS

calculations. If COV is not less than 10%, then critical thickness profiles are required to

determine the average thickness value [API 579-1/ASME FFS-1, 2012].

The CTP is established using thickness reading from a measurement grid (Fig.2.7)

with suitable interval to allow for accurate characterisation of the metal loss. In Fig.2.7.,

gridlines M1-M5 and C1-C7 represent the meriodional and circumferential inspection

planes respectively. For most of the pressure vessels and pipe lines, the CTP should be

established in both the meridional (longitudinal) and circumferential directions. But for

atmospheric tanks, only the CTP along the meriodinal direction is required by the

standard and hence the meriodional inspection planes are sufficient. The circumferential

inspection planes are not required because the stress in the direction normal to the hoop is

considered negligible and does not govern the design thickness calculation [API 579,

2012, pp.4-26]. After obtaining the thickness readings of individual locations in the grid,

the CTP in the meriodional direction is established by projecting the minimum thickness

point at each interval along the M1-M5 inspection planes on to a common plane. The

detailed procedure to establish the CTP is given in API 579-1/ASME FFS-1 Cl.4.3.3.3. In

Fig.2.7(b), tc represents uniform wall thickness (after general metal loss resulting from

corrosion, erosion, or both) away from the local metal loss, to be used in FFS assessment

procedures and tmm represents minimum measured wall thickness in the LTA.

39
 (a) Inspection Planes and the Critical Thickness Profile

(b) Critical Thickness Profile (CTP) - Longitudinal Plane (Projection of Line M)

(c) Critical Thickness Profile (CTP) - Circumferential Plane (Projection of Line C)

Fig. 2.7 Critical Thickness Profile [API 579-1/ASME FFS-1, 2010]

40
 After establishing the CTP, FFS procedures are used for three different levels of

assessment. A detailed description of the LTA assessment is given in Chapter 7.

2.6.4 FFS Research for Locally Thinned Areas

Determination of fitness for service condition of damaged equipment is not yet an

exact science [Janelle, 2005]. Researchers have continuously improved the safety and

reliability of the FFS procedures using full scale burst tests, finite element analysis and

statistical methods. The following is a brief list of research by various authors.

Folias, [1969, 1999] formulated detailed analytical expression to correlate stress

field in flat and curved plates with finite crack. He extended the study and proposed

analytical expressions to determine the state of stress near the crack tip in a

spherical/cylindrical shell. Modified versions of the expression proposed by him are now

widely used as “Folias factor” or “Folias bulging factor” in many fitness for service

procedures involving LTA or crack like flaw. His theory is probably the most influential

of the many available ideas.

Kiefner, et al.,[1973, 1989, 1990] has published a series of papers on LTA

assessment in pipes. His work forms the basis of many FFS procedures that are in

practice. He has reported experimental data from a large number of burst tests on pipes

with LTA. He has made important contribution to the development of the widely used

ASME B31.G, RSTRENG FFS criterion.

Kanninen, et al., [1991 and 1994] developed a theoretical model based on elastic

shell theory to address the LTA assessment in pipes. The objective of the work is to

41
provide a theoretical basis to include combined loading cases which are not usually

addressed by the empirical procedures used in practice. The authors expect that this

approach provides the flexibility to reduce excess conservatism that is known to exist in

B31.G guidelines.

Sims, et al., [1992] proposed the use of RSF factor for LTA in pressure vessels

and tanks. The authors developed an empirical equation to conservatively estimate the

remaining strength of spherical and cylindrical shells with LTAs by curve fitting the

results obtained from elastic-plastic FEA. Additional criteria for acceptance of LTA were

also presented. The results were compared with ASME B31G method prevalent at that

time.

Stephens, et al., [1997] presented an overview, comparison and evaluation of

acceptability criteria for LTA in cylindrical pressure vessels and pipes. Eleven different

models were compared and evaluated against experimental database from the literature.

The authors comment that although all the models have adequate safety, there is

variability in predicting the failure and hence additional work is recommended to reduce

the conservatism. The study also suggests which criteria are most desirable based upon

adequate levels of safely and minimum levels of variability.

Chen, et al., [2000] proposed an empirical formula for obtaining the load carrying

capacity of pressure vessels with two part-through defects. The article discusses the

effects of the distance between the two defects on the load carrying capacity of pressure

vessels. The authors conclude that their numerical results confirm the applicability of the

simplified numerical method.

42
 Janelle, et al., [2005, 2007] presented an extensive review on the technical basis

of FFS procedures for global and local metal loss. Most of the existing procedures were

compiled and statistically compared with burst test data to arrive at the most accurate

method. A new method is proposed and guidelines were provided for standardizing the

margin on calculated MAWP against failure pressure. The authors have recommended

best practices to be included in API 579-1/ASME FFS-1 standard. The study is reported

as welding research council bulletin [WRC 505].

Konusu [2009] proposed an assessment procedure for multiple volumetric flaws

(locally thin areas). He comments that in the current practice as prescribed in ASME

BPVC, BS 7910, API 579 and similar standards, multiple flaws are characterized as a

single larger flaw enveloping the individual smaller flaws, which may provide unrealistic

assessment in some cases. Hence a new assessment procedure is proposed in this article,

which is based on the interaction between differently sized flaws.

Mukaimachi [2009] proposed a simplified evaluation procedure for locally

thinned areas under plastic condition in cylinders. It is based on the IBARAKI FFS

procedure where the internal pressure and external bending moment are considered

simultaneously in the cylinder using the pressure and moment interaction diagram to

predict the plastic collapse. But, since this method has limitation with respect to flaw

depth, it cannot be applied to deep flaws. This paper proposes a solution to include deep

flaws. The author validated his work by comparing it with nonlinear FEA results.

Seshadri [2004] used shell equations to obtain decay length in circumferential

direction and proposed a Level 2 FFS assessment for hot spots based on elastic analysis

43
and the mα – method. The localized effect of discontinuities on the cylindrical shell is

discussed and the concept of reference volume is introduced as the “kinematically active”

portion that participates in plastic action. The reference volume is defined by using a

characteristic length called “decay length” for local load effects on the cylinder. In order

to verify the evaluation assessment, inelastic finite element analyzes were carried out on

the basis that the maximum strain is limited to 1%.

Indermohan and Seshadri [2004] proposed a Level 2 FFS methodology for

evaluating corrosion in cylindrical shells. Thin cylindrical shells with radius to thickness

ratio of about 50 subject to internal corrosion of various sizes were studied. The

remaining thickness ratios were taken as 0.9, 0.8, and 0.6. The results were compared

with inelastic finite element analysis results. It was concluded that the RSF obtained were

conservative and compared to nonlinear FEA results. The work was later extended by

Ramkumar and Seshadri [2005] to a thicker cylindrical shell with radius to thickness ratio

of around 30 with either internal or external corrosion.

Tantichattanont, et al., [2006a, 2006b, 2007a, 2007b, and 2009] presented three

methods of Level 2 FFS assessment for thermal hot spots and corrosion damages in

cylindrical/spherical pressure components. They were based on variational principles in

plasticity, the m method, reference volume and the concept of decay lengths in shells.

The study proposes decay lengths for cylindrical shells subject to local forces, in terms of

shell geometric properties. The decay length in circumferential direction was different

from the earlier studies. The extent of “local” damage in cylindrical shells and

interactions of damage effects in longitudinal and circumferential directions of the shells

44
were investigated based on these decay lengths. The stretching and bulging effects due to

damage were studied.

Seshadri and Hossain [2008] developed the m-tangent method, an extension of

the m-method, to predict the limit load for a general class of mechanical components and

structures. Based on the m-tangent method they have also presented a Level 2 FFS

procedure (as per API 579-1/ASME FFS-1) to evaluate thermal hot spot and corrosion

damage in pressure vessels and piping systems [Hossain, 2009]. The authors report that

the proposed procedure provides reasonably accurate estimate of the remaining strength

of ageing pressure components. It was compared to the results obtained Level 3 inelastic

FEM based FFS procedures.

Ahmad and Seshadri [2010] proposed a level 2 FFS procedure for structural

integrity assessment of storage tank with LTAs in tank wall. The authors proposed two

methods based on mα-tangent multiplier to determine the RSF, the first one was an

analytical method involving limit load multipliers and the second one was based on

elastic finite element analysis. The study is an initial step in the use of mα type methods

for tanks. Even though the article claims the applicability of the method; the results were

comparatively limited to small LTAs in smaller tanks. Also the entire volume was taken

as the reference volume for the method using LEFEA. It is further discussed in Chapter 7.

Peng, et al., [2011] studied about multiple wall thinning defects in pipes. The

interaction of multiple LTAs under different geometric arrangements, loading conditions

45
and depth ratios were studied. An index based on the limit load from twice elastic slope

procedure was presented to define the interaction effects quantitatively.

American Petroleum Institute’s Standard - API 650 is the premier standard for

tank design in North America. The standard establishes minimum requirements for

material, design, fabrication, erection, and testing for vertical, cylindrical, aboveground,

closed and open-top, welded storage tanks in various sizes and capacities for internal

pressures approximating atmospheric pressure. This standard applies only to tanks whose

entire bottom is uniformly supported and to tanks in non-refrigerated service that have a

maximum design temperature of 93C (200 F) or less. Separate guidelines are given as an

Appendix if the tank has to be designed for higher temperature of pressure. This standard

provides procedures for design of tank shell (wall), annular plate, bottom plate, roof,

nozzles, stiffeners, wind girders etc., Most of the work in the current thesis is geared

towards evaluation of design procedures of this standard with a view to provide

additional insights and improve the guidelines.

The European Standard EN 14015-2004 is the construction standard equivalent to

API 650 with a similar scope. The applicability of this standard with respect to internal

pressure and metal temperature is specified as 20 to 500 mbar and -40 to 300 C. This

standard restricts the maximum design strength of tank material to be 260 MPa. Unlike

API, this standard does not have any explicit equations or guidelines regarding design of

elevated temperature tanks.

API Standard “Tank Inspection, Repair, Alteration and Reconstruction - API 653”

[2009], is a post construction standard that provides minimum requirements for

46
maintaining the integrity of tanks after they have been placed in service and addresses

inspection, repair, alteration, relocation, and reconstruction issues. This standard covers

steel storage tanks built in accordance to API 650 (or its predecessor API 12C). It

recognizes fitness-for-service assessment concepts for evaluating in-service degradation

of tanks in a limited way. But API 579-1/ASME FFS-1, Fitness-For-Service standard

[2012], provides detailed assessment procedures or acceptance criteria for specific types

of degradation referenced in API 653 standard. When API 653 standard does not provide

specific evaluation procedures or acceptance criteria for a particular type of degradation,

API 579-1/ASME FFS-1 may be used to supplement or augment the FFS requirements in

API 653.

The British standard BS 7910 "Guide to methods for assessing the acceptability of

flaws in metallic structures" is equivalent to API 579-1/ASME FFS-1 addressing similar

issues. For assessment of single LTA flaws, the BS 7910 provides general guidelines to

assess remaining strength factor (similar to Residual Strength factor of API) based on the

work by Batte, et al., [1997] and Fu and Kirkwood [1995]. The remaining strength factor

is determined based on the strength of the remaining ligament in the LTA portion to

prevent plastic collapse due to bulging. Since the guidelines provided are not exhaustive,

it refers to the work by Sims, et al., [1992] for further details. It should be noted that Sims

was responsible for developing the RSF acceptability criterion adopted in API 579 as

well.

The literature review about FFS procedures given in this section gives the global

picture of the research carried out in developing the general procedure for determination

47
of RSF in corroded pressure vessels. The objective of this thesis work is to improve the

FFS procedure using mα-tangent multiplier method. This approach is different from most

of the earlier research works in this area. The use of mα-tangent multiplier for FFS

evaluations was proposed by Seshadri and coworkers. An attempt was made recently by

Ahmad et al [2010] to use this technique for LTAs in tanks. Hence only the publications

using this approach are critically reviewed here and others are presented to give an

overall idea about the development of general FFS methods.

2.7 SUMMARY

The basic theories of mechanics pertaining to the area of this research work are

briefly presented. The current thesis aims to re-evaluate and improve specific design aspects

researched mainly by Karcher, Denham, Zick & McGrath, Seshadri, and Tantichattanont.

Publications by these authors are critically reviewed. Even though, Zick and McGrath [1968]

did a seminal work in tank design, their research is focussed on shell stresses and has limited

results regarding bottom plate behaviour. Denham [1968] studied the bottom plate stresses

without the influence of temperature or friction. His work had a fixed projection length of

bottom plate beyond the tank wall. He did not consider the effect of changing this projection

length. There could be beneficial aspects for such a study. Karcher [1978a, 1981a] was the

first person to consider elevated temperature tanks systematically. He accounted for the

restrained expansion using a reduction factor without clear guidelines for predicting its value.

However friction forces were not directly considered by Karcher. This needs to be re-

48
evaluated to determine if friction can be included directly and if so, whether it will be easier

than the current approach for estimating the expansion factor.

The fitness for service of pressure vessels and pipelines is a well-researched area for

past several decades. The salient articles in the development of this area are briefly reviewed.

Most of the assessment procedures till now are empirical in nature and hence there is a

continuous attempt to improve the accuracy of these models. The widely used FFS standard

API 579-1/ASME FFS-1 still has several limitations in its methods regarding the

applicability to the type of component, damage, location, and loading conditions. Using

LEFEA and simplified limit load procedures for Level 2 FFS procedures is a recent

development in this area. The interest in this thesis work is to study the improvements in

using mα-tangent multiplier method with LEFEA for FFS assessments of tanks with blunt

corroded patches.

49
CHAPTER 3

FINITE ELEMENT MODELING

This Chapter describes the Finite Element Model used to analyze the tank. A brief

description of the material model, type of the elements used and their properties is

provided. A summary of the contact algorithms used for modeling the friction interaction

between the tank bottom plate and foundation is presented. Assumptions used in

modeling and the solution control parameters for nonlinear analyzes are specified. The

convergence issues faced and the solution procedure adopted are also described.

3.0 BASIC ASSUMPTIONS

For the studies described in this thesis, the finite element modeling was carried

out using ANSYS software (ANSYS version 12, SASIP Inc., 2011). ANSYS is a

multipurpose, commercially available finite element analysis software with inbuilt

pre/post processors. The tank is modeled as an axisymmetric structure (Fig. 3.2) for cases

when the overall behavior of the tank is studied, like the shell to bottom joint behavior or

friction interaction at tank bottom. Tanks with quarter/half symmetry or no symmetry are

modeled as 3D components when localized behavior is studied for LTAs. The foundation

is generally assumed to be a concrete ring wall with suitable infill. The interface between

the tank bottom and the foundation is modeled using contact elements to understand the

influence of friction forces. The contact model has the capability to stick, separate, and

50
slide over the contact surfaces depending upon the forces developed. The friction

coefficient between the tank bottom and the subgrade depends upon the characteristics of

the sand used, level of compaction, presence of moisture and other factors. Friction

coefficient values ranging from 0.3 – 1.0 are used in this thesis to check for worst case

scenarios depending upon the type of analysis. In general, a coefficient of 0.3 is used

unless mentioned otherwise. It is to be noted that the aim of this thesis is not to establish

a proper friction coefficient value at the tank bottom. More importantly the findings from

this research work are independent of the actual value of friction coefficient. The friction

coefficient of 0.3 is only used as an example value to evaluate and compare the results

from mathematical procedures with the numerical models.

The tank material is assumed to be A841M grade steel with Young’s modulus of

200,000 MPa and Poisson’s ratio of 0.3. The yield strength is 345 MPa with a product

design stress of 194 MPa for operating pressure. For hydrostatic test, the value is

208MPa. For the nonlinear analysis, the constitutive relationship is taken as bilinear

kinematic hardening (BKIN). This option as used in ANSYS assumes that the total stress

range is equal to twice the yield stress, so that the Bauschinger effect is included. BKIN

may be used for materials that obey von Mises yield criterion. The material behavior is

described by a bilinear stress - strain curve starting at the origin and with positive stress

and strain values. The initial slope of the curve is taken as the elastic modulus of the

material. At the specified yield stress, the curve continues along the second slope defined

by a tangent modulus. Elastic-perfectly plastic material model (Fig.3.1) is used for most

of the analyzes except for cases where the material model is changed with a post yield

51
plastic modulus of 1000 MPa. SI units are used throughout the thesis. The stress and

deformation data shown in ANSYS plots have units of MPa and mm.

For tanks with locally thinned area for FFS studies, plastic failure of the tank is

assumed to have occurred when equilibrium (primary) plastic strain is 1%. In addition,

5% peak plastic strain condition is imposed as recommended by API. The peak stress is

the highest stress produced locally in a region due to stress concentration caused by any

structural discontinuity or some form of thermal stresses. The associated strain in that

region is the peak strain. This approach is conservative and is consistent with the work

reported by earlier researchers on FFS assessment of thermal hot spots and corrosion

damage [Seshadri, 2004; Tantichattanont et al., 2008].

σ Post Yield Modulus 1000 MPa

345 MPa
Elastic Perfectly Plastic
Material
E = 200 000 MPa

Fig. 3.1 Elastic - Plastic Material Model for Steel

3.1 MODELING USING PLANE AXISYMMETRIC ELEMENTS

Quadrilateral elements with four corner nodes with 2x2 integration points

(PLANE 182) were used for the tank wall and bottom plate with axisymmetric option.

52
The bottom plate is modeled using flexible contact elements (Fig. 3.2), wherever

applicable, concrete foundation is assumed to be rigid. The blue and red lines indicate the

target and contact modules of the interacting friction model. There is no physical gap

between the elements and the surfaces of bottom plate and the top of the foundation.

SYMM

Hydrostatic
Pressure

tp

Highly Compacted Sand/Gravel Concrete

Contact Ring Wall
Wall
Elements
R ts Projection
Length

Fig. 3.2 Axisymmetric Tank Model (Not to Scale)

On performing the analysis with these assumptions, it was found that the solution

does not converge easily. For the examples studied, the solution was found to converge

without much difficulty if thicker bottom plates (say, 8 mm and above) were used (where

53
localized bending strain was reduced). For thinner plates (e.g., 6 mm), using four-node

plane elements subjected to very high bending near the shell-to-bottom joint could result

in shear/volumetric locking Volumetric locking is an overly stiff behavior that is due to

incompressibility of plastic material. This occurs when plastic deformation is dominant.

Shear locking causes the elements to be too stiff in bending (also known as parasitic

shear), especially in thin members, or in elements with large aspect ratios. Shear strains

(the change in the element angle at a node) for pure bending loads should be zero.

Because a lower-order element can only have straight edges, the element angle at each

node cannot be maintained when the nodes deform, inducing an artificial strain.

Increasing the mesh density significantly and using higher order elements would reduce

the problem. In order to compare the efficiency, some plane element models were solved

with different conditions and element types. However such trials could not be used for

the bulk of analyzes, since they took far too long to be completed.

Because the bottom plate tends to be very thin, to avoid shear locking a higher

order 8 noded quadrilateral element (PLANE 183) with Mixed U-P option was used

(Fig.3.3). The solution with this element converged well although it still took long to

complete the analysis. The Mixed U-P (also called Hybrid, or Hermann) formulation

solves the problem of incompressible material behavior by solving the hydrostatic stress

(volumetric) as an additional degree of freedom. The stiffness matrix is broken up into

displacement and pressure terms, and instead of solving just for the displacements,

ANSYS also solves for pressure degrees of freedom for each element. Because the

hydrostatic pressure in the element is a separate DOF, it is no longer dependent on

54
Poisson’s ratio. The element has quadratic displacement behavior and is defined by 8

nodes having two degrees of freedom at each node: translations in the nodal x and y

directions for axisymmetric models. Using these elements and fine tuning the friction

model parameters, convergence was achieved even with 6 mm thick bottom plate.

Fig. 3.3 Element Geometry for 8-Node Plane Element (PLANE 183)

The element may be used as a general plane element (plane stress, plane strain

and generalized plane strain) or as an axisymmetric element. This element has plasticity,

hyper elasticity, creep, stress stiffening, large deflection, and large strain capabilities.

3.2 MODELING USING SHELL ELEMENTS

Because of the inordinate amount of run time and other modeling issues

associated with plane elements, modeling options using shell elements were considered

for repeat analyzes (whereas detailed analyzes with plane 183 elements were reserved for

typical cases and special problems). For the present problem with all types of nonlinearity

and very high bending strain in a small portion, the element SHELL 209 was found

suitable. It can be used for analyzing thin to moderately thick axisymmetric shell

structures. It is a 3-node element with 3 DOFs at each node: translations in the X, Y

direction (radial and axial direction), and a rotation about the Z-axis. A fourth

55
translational degree of freedom in Z direction can be included to model uniform torsion.

When membrane option is used, the rotational degree of freedom is excluded. This

element is well suited for linear, large rotation, and/or large strain nonlinear applications.

Fig. 3.4 shows the geometry, node locations (I, J, K), and element coordinate system for

this element. For material property labels, the local x-direction corresponds to the

meridional direction of the shell element. The local y-direction is the circumferential

direction. The local z-direction corresponds to the through-the-thickness direction.

As a first step, the tank wall was assumed with fixed boundary condition and

modeled with shell elements. The results were verified using a similar model with

PLANE 183 elements and found to be matching. Later, the tank wall was modeled using

a hinged joint with a moment equal to the plastic moment of the bottom plate. This

exercise was done to verify the validity of Karcher’s model [Karcher, 1978a, 1981a]

without changing any of the assumptions involved. The results from these analyzes are

discussed in Chapter4.

Local
Global Coordinates Coordinates

Fig. 3.4 Geometry of Axisymmetric Shell Element (SHELL 209)

56
 The tank bottom was modeled using the contact elements between the bottom

plate and the foundation. Figure 3.5 shows the deflection profile of a tank modeled using

shell elements and subjected to both temperature change and hydro static load.

When elastic-perfectly plastic material models were used for the material, there

were convergence issues due to localised yielding near the shell to bottom joint which

was overcome by increasing the mesh size and number of layers in the element. In

general the analysis was carried out using up to 7 integration points across each layer.

Multiple layers were used along the thickness when needed. A macro was developed in

ANSYS to incorporate decay lengths in the post processing analyzes and extract stress

values from different layers of the shell element, if needed.

Original
Geometry

Deflected
Geometry

Fig. 3.5 Deformation Profile of a Tank Modeled Using Shell Elements

57
 For fitness for service studies, quarter or half symmetric models were used with

SHELL 281 element (Fig.3.6). It was suitable for analyzing thin to moderately-thick

shell structures. The element has eight nodes with six degrees of freedom at each node:

translations in the X, Y, and Z axes, and rotations about the X, Y, and Z-axes.

Fig. 3.6 Eight-Node Shell Element (SHELL 281)

Even though the PLANE elements took more computational time, the main

advantage of using them (rather than SHELL elements) is that the model clearly shows

the yielding behavior, formation of plastic hinges and stress/strain profile across the plate

thickness. The results for stresses are straigthforward to interpret. Shell elements need

much less computation effort and have less convergence problems compared to plane

elements. However, shell models need a fair bit of post processing in order to interpret

the results.

58
3.3 CONTACT ELEMENTS

Contact problems are highly nonlinear and require significant computer

resources. They present two significant difficulties. Firstly, in general the regions of

contact are not known until the problem is analyzed. Depending on the loads, material,

boundary conditions and other factors, surfaces can come into and go out of contact with

each other in a largely unpredictable manner. Secondly, most contact problems need to

account for friction. There are several friction laws and models to choose from and all are

nonlinear. Frictional response sometimes becomes chaotic, making solution convergence

difficult. In addition to these two difficulties, many contact problems must also address

multi-field effects, such as the conductance of heat, etc. [Ansys Contact Technology

Guide, 2009]

Contact elements can be grouped into four general categories based on increasing

levels of sophistication or complexity: Point-to-point gap elements, Point-to-line (or

slide-line) contact elements, Point-to-surface contact elements, Surface-to-surface contact

elements. For the present work, surface to surface element CONTA172 is used as the

contact source element and TARGE169 is used as the contact target element.

TARGE169 (Fig. 3.7) is an element used to represent various 2-D "target"

surfaces for the associated contact elements. The contact elements themselves overlay the

underlying solid/shell elements describing the boundary of a deformable body and are

potentially in contact with the target surface. This target surface is discretized by a set of

target segment elements (TARGE169) and is paired with its associated contact surface

59
via a shared real constant set. Any translational or rotational displacement constraint can

be imposed on the target segment element, if required.

CONTA172 (Fig. 3.8) is used to represent contact and sliding between 2-D

target surfaces (TARGE169) and a deformable surface. The element is applicable to 2-D

structural and coupled field contact analyzes. This element is located on the surfaces of 2-

D solid elements or shell elements with mid side nodes that can be used with a specific

set of elements that includes PLANE183 and SHELL 209, both of which are used for

modeling the tank wall and base plate. The contact element has the same geometric

characteristics as the solid/shell element face with which it is connected. Contact occurs

when the element surface penetrates one of the target segment elements (TARGE169) on

a specified target surface. Coulomb and shear stress friction is allowed. This element also

allows separation of bonded contact to simulate interface delamination.

Fig. 3.7 Target Element Geometry [ANSYS Reference Manual, 2009]

60
 Fig. 3.8 Contact Element Geometry [ANSYS Reference Manual, 2009]

Figure 3.7 show that the normals ‘n’ of the contact and target surface should

meet each other for the contact to happen. Hence the numbering of nodes and positioning

of the normal should be appropriate for the contact to occur. The nodes must be ordered

such that the target must lie to the right side of the contact element when moving from

the first contact element node to the second contact element node. The 2-D contact

surface elements are associated with the 2-D target segment elements (TARGE169) via a

shared real constant set. ANSYS looks for contact only between surfaces with the same

real constant set. For either rigid-flexible or flexible-flexible contact, one of the

deformable surfaces must be represented by a contact surface. This element supports

various 2-D stress states, including plane stress, plane strain, and axisymmetric

conditions. The stress state is automatically detected according to the state of the

underlying element.

3.3.1 Contact Algorithm

For surface-to-surface contact elements, ANSYS offers several different contact

algorithms: Penalty method, Augmented lagrangian, Lagrange multiplier on contact

61
normal and penalty on tangent, Pure Lagrange multiplier on contact normal and tangent,

Internal multipoint constraint. The details about these models can be found in any

ANSYS reference manual. [ANSYS reference manual, 2009]

Initially many trial runs were performed with different types of contact algorithms

and it was found that the augmented Lagrangian type is suitable. By default ANSYS uses

the material property of the underlying elements to calculate appropriate contact stiffness

and these values need to be tuned for proper convergence. Many guidelines and

theoretical descriptions are provided by ANSYS and other sources for properly selecting

the contact algorithms and contact parameters, reviewing all the literature is beyond the

scope of current report. In the augmented Lagrangian method which is used here, the

factors for the normal and tangential stiffness are required to be specified. The amount of

penetration between the contact and target surfaces depends upon the normal stiffness,

while the amount of sliding depends upon the tangential stiffness. Higher stiffness values

decrease the amount of penetration or slip, but can lead to ill-conditioning of global

stiffness matrix and hence cause convergence difficulties. Lower stiffness value can lead

to greater amount of penetration and slip and hence the results may be erroneous. Even

though ANSYS uses default values, some experimentation is necessary to determine an

optimum stiffness and tolerance factors. In general, a low contact stiffness value is used

in the beginning and the analysis is carried out with the fraction of the total load. Based

on the resulting contact penetration/slip and the number of equilibrium equations, the

normal/tangential stiffness is adjusted till penetration/slip is less the tolerance levels.

Using this procedure, the appropriate stiffness and tolerance values for the tank were

62
determined after several trials. The Penetration Tolerance Factor (FTOLN) is kept at -

0.02 and the Normal Penetration Stiffness Factor (FKN) value is kept at 1 for most cases

and as 10 for a few situations. Table 3.1 gives the gist of element types used in this study

and their associated characteristics.

Table 3.1 Element Types used in Finite Element Method

Element Element
FE Model Analysis type Objective of analysis
type Name

2D - Nonlinear Analysis with To study effect of

Axisymmetric Shell SHELL 209 friction interaction at tank friction, projection
model bottom length and fatigue
2D - PLANE Nonlinear Analysis with
Axisymmetric Plane 182/PLANE friction interaction at tank To study plastic hinge
model 183 bottom formations
3D- Half or Nonlinear Analysis without
quarter friction interaction -i.e
Shell SHELL 209 To study LTAs
symmetry bottom fixed boundary
problem condition

These solution control values are extremely important in determining if an

analysis will succeed or fail. If there are too few sub steps, the contact nodes may be

driven through the target elements before ANSYS "realizes" it has happened. In this case

the solution will resemble that of an analysis that didn't have contact elements defined at

all. Therefore it is important to choose a relatively large number of sub steps initially to

ensure the model is defined properly. The time step size must be small enough to capture

the proper contact zone. The smooth transfer of contact forces is disrupted if the time step

size is too large. Once everything is working, the number of sub-steps can be reduced to

optimize the computational time. Also, if the maximum number of sub steps or iterations

63
is left too low, ANSYS may stop the analysis before it has a chance to converge to a

solution.

3.4 ISSUES IN FEM MODELING

In the present work, modeling the problem includes material, geometric and

contact nonlinearity. As modeled, it may take 3 to 4 days (in a dual core 2.4Ghz

processor) with for a single detailed nonlinear analysis run. After several trials, many of

the solution parameters were finalised. It was observed that the plane element model may

need more than 1500 sub-steps for convergence to happen. Also, the storage space

required for a single analysis was more than 30 Giga Bytes. Refining the meshes near the

discontinuity increased the size of the problem. With the same condition of geometry and

loading, for cyclic load conditions, it took almost a week to complete a single run. In case

of tanks with both water and temperature load applied in a cyclic manner, it takes longer

to complete the analysis. It should be noted that, separate thermal analysis is not

necessary for the thermal stress case since the rise in temperature is uniform across the

wall thickness and full depth of liquid. The assumption of uniform temperature rise is

appropriate and it is explained in Chapter 6.

Since the tank wall, for the most part is predominantly a membrane structure

governed only by hoop forces, the size of the mesh can be increased as the height

increases from bottom. Similarly in the bottom plate, the portion away from the shell to

bottom joint is not subjected to bending and hence doesn’t need a refined mesh. It is not

necessary to model the entire width of bottom plate from the centre to periphery; a major

64
portion of the bottom plate except for the small length near the shell is subjected to

uniform pressure and is fully supported by the foundation. Hence there is very low stress

in this region. This portion can be curtailed in the model which will considerably reduce

the size of the problem. Studies were performed using models with full width and

curtailed width of bottom plate and it was found that there was no significant difference

in using this assumption, for the hydraulic load alone. However, in case of tank with

thermal loading, the full length of the radius has to be modeled since the thermal

expansion (RT) is the product of radius.

The other important issue observed was modeling the tip/periphery of the bottom

plate using plane elements. As mentioned earlier, when the tank (with the standard 50

mm bottom plate projection) is loaded, the portion of the bottom plate below the shell

lifts up, and the outer edge of the tank is supported by the foundation only at a single

point. The concentrated application of reaction force locally at this point, results in very

high strains and distorts the element shape (Fig. 3.9). The FE analysis sometimes fails to

converge because of this reason. The alternatives used are (i) increasing the mesh density

(ii) increasing the projection length and (iii) increasing the strength of the material near

the tip. Increasing the mesh density helps to an extent, but the same problem arises at a

higher load. Increasing the strength (Young’s modulus) locally is not very effective,

while increasing the projection length proved to be very effective (Fig.3.10). Apart from

the above issues, modeling contact behavior as explained in the previous section is a

laborious task that needs fine tuning several parameters.

65
 The shell to bottom joint as shown in Fig. 3.10 is not modeled with weld details.

The residual stress and the change in material property after welding are not included in

the analysis. The residual stress effects are self-equilibrating and hence can only alter the

load at which yielding starts, but not the final load at which the plastic hinges are fully

formed. Ignoring the weld size increases the lever arm for moments slightly. This is

conservative. Also, ignoring the slight increase in material strength due to the weld is

likely to be conservative. On the other hand, welding would have an increase in stress

concentration effect. This is addressed separately by API 650 in the fatigue evaluation

equation (see Chapter 6).

3.4.1 Mesh Convergence Study

In order to reduce the size of the problem, different mesh sizes are adopted at

appropriate locations. Seven different areas are used as shown in Fig. 3.11 which

included areas for smoother transition of mesh size from fine to coarse or vice versa.

Figure 3.12 shows typical areas where different mesh densities are adopted. The stress

distribution plots at the shell to bottom joint for the hydrostatic load are shown in

Fig.3.13 and 3.14. Limit load analysis is carried out by proportionately increasing the

loading till material failure stops further increase.

For tanks with locally thinned areas, the LTA and the surrounding area is finely

meshed while the element size is increased in the remaining area since hoop is the

predominant force away from the LTA (Fig. 3.15). The von Mises stress contours of a

typical tank with LTA are shown in Fig. 3.16. The von Mises stress at the time of failure

(1% plastic strain) of LTA is shown in Fig. 3.17. If quarter model of a tank is used to

66
study the LTA, it must be ensured that the decay length [6.3 ] should be less than

one fourth of the circumferential length as explained in Chapter 7.

Tip of the projected portion

of the bottom plate with
excessive stress and
deformation

Fig. 3.9 Localized Deformation at the Bottom Corner of Bottom Plate

67
 Inner side Outer side
of tank of tank
the tank

Fig. 3.10 Deformation in the Bottom Plate (with Extra Projection)

Fig. 3.11 Zones Specifying Different Mesh Densities

68
(a) Shell to Bottom Joint
(b) Tip of Bottom Plate Projection

Fig. 3.12 Typical Mesh at Shell to Bottom Joint

Fig. 3.13 Stress Distribution for Partial Plasticity

69
Fig. 3.14 von Mises Stress Distribution at Limit Load

Fig. 3.15 Deformation Profile of Tank with LTA

70
Fig. 3.16 von Mises Stress Profile of Tank with LTA

Fig. 3.17 LTA of Tank at Failure

71
3.5 DIMENSIONS OF TANKS USED IN THE ANALYSIS

The following details describe some of the different cases of tanks used in the

present study. The dimensions of the tanks (Figures 3.18-3.20) are chosen such that they

represent lower, middle and upper end of the design range for typical ASTs.

Case 1

Geometric Data:

Diameter : 60,000 mm

Height : 12,000 mm

Thickness of shell wall : 18 mm

Thickness of Bottom plate : 6 mm

Projection of Bottom plate : 50 mm (Beyond shell)

Material Data:

Young’s Modulus : 200,000 MPa

Yield Strength : 345 MPa

Product Design Stress : 194 MPa

Hydro Static Test Stress : 208 MPa

Density of liquid (water) : 9.81 kN/m3

Density of steel : 77.1 kN/m3

Poisson’s ratio of steel : 0.3

The required shell thickness is determined using 1-foot method as 18 mm (as prescribed

in Clause 5.6.3.2 of API 650 [2012].

72
 SYMM

HYDRO STATIC PRESSURE

H =12,000 mm

6 mm
R =30,000 mm 50 mm
18mm

Fig. 3.18 Dimensions of Tank (Case 1)

Case 2

Geometric Data:

Diameter : 90,000 mm

Height : 16,800 mm

Required thicknesses of shell wall :

t1 t2 t3 t4 t5 t6 t7
33.98 31.33 22.82 18.46
13.46 mm 10 mm 10 mm
mm mm mm mm

The suffixes of “t” represent the successive shell courses from bottom.

Thickness of Annular plate : 14 mm

73
Width of Annular plate : 735 mm

Thickness of Bottom plate : 6 mm

Projection of Annular plate : 50 mm (Beyond shell)

Material Data : Similar to Case 1

The thicknesses of the tank wall obtained above are the values at the end of design run as

described in Clause 5.6.4 of API 650 (variable design point method). It is not rounded off

nor is the corrosion allowance added to it.

74
 t7

t6

t5

H=16800mm

t4

HYDRO STATIC PRESSURE

t3

t2

R = 45000mm

t1

14mm
6mm

735mm 50mm

SYMM Centre line

Fig. 3.19 Dimensions of Tank (Case 2)

75
Case 3

SYMM

HYDRO STATIC PRESSURE H = 7200 mm

8mm
R =18,000 mm 50 mm
10mm

Fig. 3.20 Dimensions of Tank (Case 3)

Diameter : 36,000 mm

Height : 7200 mm

Thickness of shell wall : 10 mm

Thickness of Bottom plate : 8 mm

Projection of Bottom plate : 50 mm beyond shell (also with variable

projection lengths as explained in Chapter4)

Young’s Modulus : 200,000 MPa

Density of liquid (water) : 9.81 kN/m3

Density of Steel : 77.1 kN/m3

Steel Poisson’s ratio : 0.3

76
This tank is designed using the “Appendix A Optional Design Basis for small tanks” with

an allowable stress of 145 MPa and joint efficiency factor 0.85.

3.6 SUMMARY

In the present Chapter, the finite element modeling of tanks is discussed. The

details regarding elements used and the problems encountered during the analysis are

given. A review of contact elements and the contact algorithms applicable to the present

problem is provided. The details of finite element analysis and the interpretation of the

results are given in the following Chapters. Typical tank dimensions, material and

loading data used in the present research are included.

The FE model of the tank is initially checked with shell equations with fixed

boundary conditions at the bottom. Later the friction interaction at bottom is introduced

and the restraint friction deformation is compared with mathematical model developed in

this thesis. In order to establish the validity of using axisymmetric model, comparisons

were also made with half symmetry models. In order to achieve convergence, different

types of contact algorithms and their modelling parameters were tried before finalising

the model. For study of locally thinned areas the limits loads of damaged tanks from FEA

and their corresponding RSF values were checked against the empirical RSF procedures

prescribed by API 579 standard. The measures taken to optimise the mesh by zoning the

cross-section are also presented. It is also ensured that local changes in the model for the

sake of numerical stability do not affect the global behaviour of the tank.

77
CHAPTER 4

ISSUES WITH BOTTOM PLATE PROJECTION AT

SHELL-TO-BOTTOM JOINT

This Chapter describes the importance of shell to bottom joint in tanks and studies

the effect of bottom plate projection length at this joint. The beam model used by

Denham, et al., [1968] to analyze the bottom plate is introduced and the FEM results to

validate this model are presented. The effect of increasing the bottom plate projection

length is studied both theoretically and numerically. The beam model is modified to

determine the length of bottom plate projection needed for maximum effect. Nonlinear

failure behavior of the tank, specifically near the bottom is studied using FEA. The

formation of plastic hinges in the bottom plate on the inside and outside of this joint is

discussed in detail.

4.0 SHELL-TO-BOTTOM JOINT

The cylindrical shell of the tank is designed for hoop stress while the bottom plate

for the most part is a membrane without any structural function other than to act as a

barrier between liquid and foundation. However, the shell to bottom joint is a critical

location in many failure modes of the tank since the bending stresses in the bottom plate

78
are close to yield in most practical cases. The radial growth of the shell at the bottom is

restricted because of the bottom plate. This induces discontinuity stresses near this joint.

Even though the tank is basically designed for hoop stress, the bottom portion of the shell

(approximately till a height of (D ts) 0.5, where, D is the diameter and ts is the thickness

of the shell) is influenced by the bending stresses and the free radial expansion of the tank

wall is realized only after this height [Long and Garner, 2004].

When the tank is filled with the liquid, the shell tends to rotate as shown in Fig.

4.1. The welded joint is constructed as a rigid joint (fillet weld on both sides with the

requirements as specified in Cl.5.1.5.7 of API 650, 2010). Hence the bottom plate rotates

to the same amount as the shell at the joint. This rotation could cause the bottom plate to

lift off the foundation over a distance L as shown in Fig. 4.1, until the liquid pressure

acting on the plate is sufficient enough to balance the uplift. The internal moment (Mo)

causing the rotation in the plate is balanced by that in the shell. The magnitude of this

moment depends upon the parameters R, H, ts, tb and the specific gravity of the liquid

(G). In addition to these factors, the type of foundation directly beneath the tank shell and

the temperature of the stored liquid influence the magnitude of this moment. The plate

separation (uplift) from the foundation is especially significant for concrete ring walls

and slabs. Because of this, the present Chapter assumes that the tank wall is resting on a

foundation of high rigidity. Knowing the value of the bottom moment will help to

determine the stresses in the shell and bottom plate.

The stresses in the shell to bottom zone are also influenced to some extent by the

type of weld used (double fillet/full penetration). In the case of double fillet weld

79
 (Fig.4.1), the stresses in the bottom plate will be high near the toe of the weld, whereas in

full penetration weld the stresses will be high at the face of the shell. It is assumed that

the latter case is slightly conservative because of the small increase in the lever arm.

Hence it is adopted for the current work.

H Hydro static
Pressure

Mo
Annular Plate

Uplift Length (L)

Ka Bottom Plate
tb
(a) Full view (b) Expanded view

Fig. 4.1 Shell to Bottom Joint in the Tank

4.1 ANNULAR PLATE

As the bottom plate is subjected to very high bending stresses near the shell joint,

in practice, the bottom plate close to the joint is made slightly thicker than the rest of the

80
plate and is sometimes called as annular plate (Fig.4.1). API 650 recommends the annular

plate only if the local stresses near this joint will be substantial, otherwise the same

bottom plate which is used elsewhere will be extended below the shell wall. Similarly,

the width of the annular plate (Ka) as shown in Fig. 4.1, should be chosen such that it

would be sufficient to assure that all bending stresses by the moment have damped out.

This is based on the consideration that it is desirable to avoid bending in the conventional

lap welded tank bottom construction [Karcher 1978 a, 1978 b, 1981a].

Various aspects of storage tank analysis and design with regard to tank wall and

bottom plate had been studied by Zick and McGrath [1968], Denham et al.[1968a, b],

Karcher [1978, 1981a, b], Jones and Seshadri [1989], Wu and Liu [1996, 2000], etc.

Among them, Wu and Liu [1996, 2000] and Denham, et al. [1968a, b] proposed models

to predict the stresses in the annular plate near the shell to bottom joint.

API 650 recommends the following formula for the minimum width of annular plate:

Ka = 215 tb/(G H)0.5 (4.1)

The minimum width formula recommended by API can be established by

assuming that the annular plate behaves like a cantilever beam subjected to vertical

hydrostatic pressure and determining the length required to form a plastic hinge at the

shell-to-bottom joint [Jawad and Farr 1984].

M = w (Ka)2/2 ; SY = 4M/tb2 (4.2)

81
where, M is the moment at the shell to bottom joint (by assuming the annular plate as a

cantilever beam), w is the liquid pressure acting above the annular plate, SY is the yield

strength, tb is the bottom plate thickness and G is the specific gravity. Rearranging Eq.4.2

0.5
 S t2  (4.3)
K  Y b 
a GH
 

In Eq. 4.3 ‘γ’ represents the specific weight of water. From Eq. 4.3, by using a

suitable SY and a factor of safety, Eq. 4.1 can be arrived at. For example if SY = 345 MPa,

the constant coefficient from Eq. 4.3 will be 187.5 compared to 215 from Eq.4.1. Since

the coefficient of ‘215’ is given in Eq.4.1 without any parameter for the yield strength,

the factor of safety will change with respect to the SY value of the plate used. Plates with

higher yield strengths will have lower factor of safety and vice–versa. It is unclear if this

done deliberately by API 650. However, it will be good to change this practice and make

the margin of safety uniform for all plate materials.

Apart from the recommendation for thickness and width, API 650 specifies that

the annular plate must extend at least 50 mm from the shell surface or 13mm beyond the

toe of fillet weld whichever is greater. This minimum projection seems to have been

advised for providing a proper fillet weld on the outer side of the shell. API is silent about

providing extra projection length beyond this requirement and its effects on shell or

bottom plate stresses. The effects of providing higher projection length on the stresses in

this zone, or the theory to determine the maximum projection length are not well known

to many practitioners. In this Chapter the effect of increased projection length is studied

both theoretically and numerically.

82
4.2 PLASTIC HINGES

Even though the bottom plate is subjected to bending to a distance L (uplift

length), this length is small compared to the radius of the tank. Hence, except for this

length, the bottom plates are simple membranes (without any bending stresses) used to

hold the tank liquid, fully supported by the subgrade underneath. The design procedure

implicitly allows the bottom plate to yield completely to form a plastic hinge near the

shell to bottom joint. The formation of plastic hinge here will not disturb the functions of

the tank or render it unsafe. A typical joint can have plastic hinge on the inside, outside or

both sides of the shell as shown in Fig. 4.2. This again depends on the thickness ratio

between the shell and bottom plate and the type of foundation beneath the shell, i.e.,

earthen foundation without ring wall, or earthen foundation with crushed stone/concrete

ring wall, or concrete slab foundation.

API 650 lists the minimum thickness of annular plate based on the hoop stress at

the lowest layer of shell (Table 4.1). Using FEA, it can be shown that the minimum

recommended dimensions are such that the bottom plate stresses are at or close to yield

near the joint when the tank is filled fully. It should be noted that API recommends only

the minimum required thickness. The designer has no restriction on adopting a thickness

higher than the prescribed minimum. In such a situation, the stresses in the annular plate

could be fully elastic.

83
 Annular
plate

Bottom (a)
plate (b)

Fig. 4.2 (a) Annular Plate on Concrete ring wall (b) Plastic Hinges at the Shell to Bottom
Joint [Jones and Seshadri, 1989]

Table 4.1 Annular Plate Thickness in millimetres [API 650, Cl.5.5.3]

Plate Thickness of Stress in First Shell Course (MPa)

First Shell Course
σ 190 σ 210 σ220 σ250
ts, mm (mm)

ts 19 6 6 7 9
19 < t  6 7 10 11
t  6 9 12 14
t 40 8 11 14 17
t 45 9 13 16 19

When the tank is filled to a height less than the design height or the specific

gravity of the liquid used is less than the design specific gravity of the liquid and in other

similar situations, the stresses in the bottom plate could be completely elastic.

84
 Proper understanding of hinge formation in this area is necessary for the design

and analysis of tanks, as the moments at this joint essentially form the boundary

condition for the shell analysis. Also, the shell stresses close to the bottom are important

since most of the fixtures, nozzles, openings, etc., are located in the lower portion of the

shell close to the bottom. In order to understand the stress distribution in shell and the

bottom plate, the following issues have to be studied.

1. The internal moment at the joint for operating loads, if the plastic hinge has not

formed

2. The load at which the hinge will be formed completely

3. If the plastic hinge is formed, then the number of hinges - one/two, if one hinge

– then is it inside or outside or if two hinges, do they form simultaneously

4. If one hinge is formed either inside or outside, then the moment from the other

side of the shell

5. Will the shell remain completely elastic till the plastic hinges are formed

completely?

4.3 TEMPERATURE EFFECTS

In many situations, storage tanks are required to hold liquids like asphalt,

residuum, high pour point hydrocarbons, etc., at higher temperature than the ambient and

such tanks are called as Elevated Temperature tanks. The temperature loading gives rise

to thermal stresses in the tank, especially near the shell to bottom joint. For tanks

85
operating at elevated temperatures (200F to 500F), in addition to the factors considered

for ambient temperature tanks, the stresses in the shell to bottom zone depend upon the

radial expansion of the tank bottom and the frictional forces acting in between the bottom

plate and the subgrade. In addition to the induced thermal stresses in the tank, the cyclic

loading due to continual filling and emptying of the tank is reported to cause low cycle

fatigue in the shell to bottom joint. Karcher [1978a, b] proposed a theoretical model to

determine the stresses in the bottom plate and tank wall and the safe design cycle life for

elevated temperature tanks. The details of this model and further information about the

elevated temperature tanks are provided in Chapter 5 and 6.

4.4 ANALYTICAL MODEL AS PROPOSED BY DENHAM ET AL.

In order to analyze the bottom plate, Denham, et al., [1968a] proposed a beam

model as shown in Fig. 4.3 that is compatible with the shell rotation at the shell to bottom

junction. They assumed a projection length as a fixed value of 3” while the uplift length

inside the tank is a variable that changes depending on the applied forces and stiffness of

the shell and plate. This elastic method was also compared with field measured data. The

strain gage data confirmed that the annular plate would reach yield close to shell and

hence a high stress low cycle fatigue situation can be expected at this joint. The field data

from the remaining portion of the annular plate showed considerable variation from the

theoretical results. The authors ascertained that the probable cause could be the annular

plates being uplifted from the foundation for a distance exceeding uplift length L before

water filling. There wasn’t any comparison with numerical models like FEM reported in

86
the article. In the current work, the model as proposed by Denham, et al., is verified with

FE model and is further extended.

T (shell self wt) Deflection

Pattern
y
Mo water pressure ‘w’/unit length
x B
A
C
a (= 3”) b
L
RA RB

Fig. 4.3 Denham’s Model of Shell to Bottom Joint

Denham used the shell theory for the tank wall and beam theory for the annular

plate with the following assumptions:

1. The annular plate lifts clear off the foundation for some distance L

2. Support points A & B do not yield (no settlement)

3. The deflection and moment are zero at the outer support (A)

4. The deflection, moment and also the slope are zero at the inside support (B)

5. The slope of the shell and the slope of the annular plate, at the shell to bottom

joint (C) are same, i.e., slope compatibility is assumed at the joint

6. There is no radial deformation of the shell at the bottom

87
 The procedure is to first consider a beam of unit width with hydrostatic load and

self-weight of shell wall acting on it as shown in Fig. 4.3. Because of the rigid joint at the

shell to bottom junction, an internal moment Mo from the tank wall will be applied. Its

value depends on the uplift length, weight of liquid and other parameters of geometry.

Hence, as a first step an arbitrary value for the uplift length L is assumed and the moment

and slope at the shell to bottom joint location are determined. For an assumed uplift

length L, the beam model is a determinate problem, hence the moment and slope can be

found using equilibrium conditions.

The slope of the shell bottom for this assumed moment (Fig.4.4) must be

compatible with the slope that was calculated from the beam model at this location. The

assumed uplift length is suitably changed to satisfy this condition. This will be a trial and

error procedure and was solved graphically by Denham, et al. [1968a]. The equations for

shell and annular plate are as given below:

Shell equations:

 ( H  xs ) R 2 e  xs
ys   [c1 cos(  xs )  c2 sin(  xs )] (4.4)
Ets E

3(1   2 )
where,   4 (4.5)
R 2t s 2

dys  R 2 e  xs
  [(c2  c1 ) cos(  xs )  (c2  c1 )sin(  xs )] (4.6)
dxs Ets E

88
where, ys is the radial deflection of shell wall and xs is the height of the location in the

shell wall from bottom, ts is the thickness of the shell and γ is the specific weight of the

liquid infill.

xs

γH
ys C Membrane forces
Mo
T
Fig. 4.4 Freebody Diagram of Shell

Annular plate equations:

Case (i): when x ≥ a

d2y w( L  x)2
EI  RB ( L  x )  (4.7)
dx 2 2

dy RB ( L  x)2 w( L  x)3
EI   (4.8)
dx 2 6

89
 RB ( L  x)3 w( L  x)4
EIy   (4.9)
6 24

The integrations constants in Eq.4.8 and 4.9 are eliminated using the boundary conditions

of zero slope and deflection at x = L

It should be noted that the parameter ‘y’ in the equations above is the vertical

deflection of the tank bottom plate, whereas ‘ys’ denotes the radial deformation of the

tank shell. Similarly ‘x’ is the horizontal distance from the outer edge of the

bottom/annular plate to a point under consideration in the bottom plate and ‘xs’ is the

vertical distance of any point in the shell wall from bottom.

Case (ii): when x ≤ a

d2y w( L  x)2 w(a  x)2

EI 2  RB ( L  x)   T (a  x)   Mo (4.10)
dx 2 2

dy RB ( L  x)2 w( L  x)3 T (a  x)2 w(a  x)3

EI      M o ( x)  M o (a ) (4.11)
dx 2 6 2 6

RB ( L  x)3 w( L  x)4 T (a  x)3 w(a  x)4 x2 a2

EIy      M o ( )  M o (ax)  M o ( )
6 24 6 24 2 2

(4.12)
The integration constants in Eq.4.11 and 4.12 are eliminated using compatibility

conditions of slope and deflection at x = ‘a’ from Eq.4.8 & 4.11 and from Eq.4.9 & 4.12.

Taking a moment of all forces about point A,

w 2 Ta M
RB  (L  a2 )   o (4.13)
2L L L

90
Similarly using the boundary condition of zero deflection at the outer edge of the bottom

plate (x=0, y =0), in Eq.4.12,

wL Ta3 w a 4 a2 M o
RB   3  3 3 (4.14)
4 L 4 L3 L

Subsequently the Moment Mo can be found explicitly using Eq. 4.13 & 4.14.

L2  a 2 w 2 
M0   4  L  a   T  a  
2
(4.15)
L2  3a 2

Moment, kN-m
Slope

Fig. 4.5 Moment vs. Slope Graph to Find Uplift Length

Finally, the iterative procedure can include a typical graph as shown in Fig.4.5.

The solid line indicates the slope values at the shell to bottom joint calculated for

different values of Mo from annular plate slope equation (Eq.4.8). Similarly the dashed

line is the slope from shell equation (Eq.4.6). For slope compatibility, a particular

91
moment value should produce same slope in both the annular plate and the shell wall.

This is determined graphically by Denham as shown in Fig.4.5. Though the slope to

moment relation is nonlinear, the graph seems to be fairely linear. This could be because

only a small portion of the lines near the intersection is plotted.

Denham, et al., [1968a], compared the theoretical results with actual strain

measurements from an instrumented tank. The theoretical calculations indicated that

stresses in the annular plate would reach yield close to the shell. This was confirmed from

the field data. However for locations away from the shell to bottom joint, the field data

showed considerable variation. The probable reason stated is that, the annular plate would

have been uplifted from the foundation for a distance exceeding the uplift length L prior

to water filling.

4.5 VERIFICATION OF DENHAM’S MODEL USING FEA

As mentioned above, Denham, et al. [1968a], in their paper, compared the

theoretical model with field data and found that the annular plate stresses are not

predicted satisfactorily for location away from the shell joint. Hence, in the present work,

a finite element model is used for verification of the theoretical model.

The dimensions of the tank are as per Case 3 Model as described in Section 3.5.

A detailed description of the Finite element modeling is also provided in Chapter 3. The

FEA was carried out using an axisymmetric model with plane elements (8 node

PLANE183 elements with u-p capabilities) and material nonlinearity. The bottom support

is modeled with contact elements resting on a rigid surface capable of

92
sliding/sticking/separation. The loading is the pressure due to fully filled tank. The

accuracy of modeling the tank using this type of finite element mesh was examined by

comparing the results for a fixed bottom tank with those obtained from using

Timoshenko and Woinowsky-Kreiger [1959]. For the example tank (H= 16.8m, ts =

30mm, R = 45m), maximum deflections differed by less than 0.05%. Similarly maximum

bending stress in the shell differed by less than 0.3%. The yield strength of the material is

taken as 345 MPa. The meshes are appropriately sized at different parts of the tank. The

outer tip of the bottom plate and the area near the shell to bottom joint is very finely

meshed to account for the local stress concentration and formation of plastic hinges,

respectively. Based on the importance of the location and convergence issues, several

trials were performed before the mesh sizes were finalised.

Fig. 4.6 and 4.7 compare the theoretical and numerical results for the bending

stress and the uplift in the bottom plate. The friction coefficient between the bottom plate

and the subgrade is assumed as 0.2. This value is used only for demonstration purpose;

the actual friction coefficient varies depending on the site conditions. The value of

friction coefficient does not play a significant role for the current problem since the tank

shell expansion at the bottom due to hydrostatic loading is negligible due to the presence

of membrane action of the bottom plate (due to high axial stiffness). The friction

coefficient will be of much greater significance if temperature expansion is fully or partly

curtailed due to frictional forces at the bottom. This issue is studied in detail in the next

Chapter.

93
 From the stress plots, it is shown that theoretical model and FEA results are in

reasonable agreement. The small difference in the peak stress locations can be

ascertained mainly because of the fact that, in the theoretical model the shell is idealized

as a single concentrated point and hence it causes two different stresses (the positive

maximum and the negative maximum) at the same ‘x’ location, whereas in the FE model

the tank wall has thickness and hence the positive and negative peaks are not in the same

‘x’ location. Although the beam model is simple looking, the results of FEA show that it

is reasonably good in predicting the stresses.

Fig. 4.6 Uplift in the Bottom Plate

94
 Location of Tank
Wall

Fig. 4.7 Bending Stress on the Bottom Side of Bottom Plate with 50mm Projection

Length

4.6 EFFECT OF PROJECTION OF ANNULAR PLATE BEYOND

SHELL

From the theoretical model, it can be observed that, the projected length of the

plate ‘a’ influences the peak bending stresses occurring on either side of the shell to

bottom joint. Figures 4.6 and 4.7 are for a fixed length of 3” projection as used by

Denham, et al. [1968a]. On further analysis it was found that, increasing the projected

length till a particular value, increases the bending stress on the outside face and hence

moves the maximum bending stress location from the inner side of the joint to the outer

side of the joint. Any further increase in the projection length will not alter the location

nor the value of the bending stresses. Figure 4.8 shows the bending stress from FEA for

95
the same tank described previously, but with 100 mm projection beyond the shell. Unlike

the previous plot (Fig. 4.7), it can be seen here that the maximum stress lies on the outer

side of the bottom joint.

134 MPa

A’ A C B

-118 MPa

Fig. 4.8 Bending Stress in the Bottom Side of Bottom Plate for 100mm Projection Length

In Fig. 4.8, Point A’ is the outer end of the plate and C is the junction where shell

joins the plate. It can be observed that the bending stress becomes zero in the portion A’A

where the uplift is zero. Similarly the portion of the bottom plate after point B lies on the

foundation and is not subjected to bending or uplift. Hence, it can be concluded that the

length of projection on the outer side, beyond a certain distance (AC in this case) will not

be of practical use and hence can be curtailed. It is of interest to determine the length AC

and CB as a function of tank geometry and material property.

96
 From the FEM results it was found that the length of AC was approximately

71 mm for the example tank. Using this, the theoretical stresses are calculated. The

following graph (Fig. 4.9) shows the FEA and theoretical stresses for 71mm projection.

Fig. 4.9 Bending Stress in Bottom Plate for 71mm Projection

As expected, it can be seen from the plot above that the theoretical stresses match

reasonably with the FEA results. The following graph (Fig. 4.10) shows the theoretical

stresses for the same tank with 50mm and 71mm projection.

97
 Fig. 4.10 Bending Stress in Bottom Plate for 50mm and 71mm Projection

It can be seen that because of the increased projection from 50mm to 71mm, the

maximum stress location which was occurring at the inner face (for 50mm) of the shell to

bottom joint has moved to the outer face (for 71mm). Hence within the elastic limit,

providing longer projection helps in moving the maximum stress location to outside the

tank shell.

4.7 DETERMINATION OF FULL PROJECTION LENGTH

It is easy to see that for a sufficiently wide outer plate (dimension “a” in

Fig. 4.11), the slope of the plate on the outer edge will be zero, just as the case at the end

98
of dimension “b”. Consider this to be “full” projection width, since a larger value of “a”

has no additional impact.

Deflection,
Moment and Deflection,
slope are Zero Moment and
T (Tank self wt) slope are Zero

Mo water pressure w/unit length

A B
C
a b
L
Ra Slope of shell Rb
and plate must
be compatible

Fig. 4.11 Beam Model of Shell to Bottom Joint for “Full” Projection

It should be noted that even though the slopes are zero at supports A and B, it

does not imply that there are fixed end moments at A and B. Proceeding as before and

setting A = 0,

1 L2  b2
A  [4M o (2 L2  6aL  3a 2 )  4Tb( )  wb 2 (2 L2  b 2 )]  0
24 DL 6 DL (4.16)

 4Tb( L2  b 2 )  wb2 (2 L2  b 2 ) 
Mo   
 4(2 L2  6aL  3a 2 )  (4.17)

Similarly for B =0,

1
B  [4M o (3a 2  L2 )  4Ta( L2  a 2 )  w( L2  a 2 )]  0
24 DL (4.18)

99
This gives,

 4Ta( L2  a 2 )  w( L2  a 2 )2 
Mo    
 4(3a 2  L2 )  (4.19)

As before,

 R2 Mo
c ( shell )  (  H  1) 
Ets 2 Ds  (4.20)

Mo Tb( L2  b2 ) Tba 2 ) wb2

c ( plate)  [3a 2  L2  3aL]    [2 L2  b 2  6a 2 ]
3DL 6 DL 2 DL 24 DL

(4.21)

1   R2  Tb( L2  b2  3a 2 ) wb2
Mo   (  H  1)   [2 L2  b 2  6a 2 ]
 3a  L  3aL
2 2
1   Ets  6 DL 24 DL
  
 3DL 2 Ds 

(4.22)

Equating the Moment (Mo) expressions (4.17) & (4.19) and equations (4.19) &

(4.22) results in a nonlinear equation of 6th degree which is not explicitly solvable. The

value of Mo, ‘a’, and ‘b’ should satisfy equations 4.17, 4.19 & 4.22.

4.7.1 Ratio of Lengths “b” and “a”

Let  = b/a and assuming T is negligible compared to the liquid weight acting at

the tank bottom, from Eq.4.17,

wa 2 (2   2 )2
Mo   (4.23)
4 (2  2   2 )

100
Similarly from Eq.4.19,

wa 2  2 (2  4   2 )
Mo  (4.24)
4 (2 2  2  1)

Equating the Eq. 4.23 and 4.24 and simplifying

1  33
  3.372 (4.25)
2

If the shell self-weight T is assumed negligible and the relation b = a is used,

then a simple relation between the unknowns Mo and a can be obtained. The effect of

self-weight on the bending stresses is discussed later in section 4.11.

1  33
Using the relation   in Eq. 4.16 gives:
2

Mo 2 Ra L2 w( L  a)4
(a  L ) 
2
  M oa  0 (4.26)
2L 6 24 L

Slope at B is zero gives:

Mo 2 R L2 w( L  a)3 w( L  a)4
(a  L2 )  a   MoL  0 (4.27)
2L 3 6 24 L

Simplifying the above equations,

Mo  5.091( wa 2 ) (4.28)

Note that the above equation needs to be used in conjunction with the value of “a”

obtained by solving equations 4.17, 4.19 and 4.22.

101
4.8 RELATIONSHIP BETWEEN MOMENTS Mo AND Mfx FOR FULL
PROJECTION LENGTH
It has been observed that the moment created in the shell bottom for fixed

boundary condition (Mfx) will have some relation to the moment (Mo) from the previous

section. From shell theory, the fixed end moment Mfx at the bottom of shell is expressed

as [Timoshenko and Woinowsky-Krieger, 1959]:

 RHts 1
M fx  (1  ) (4.29)
12(1  )2 H

Let   Mo / M fx

Using Eq. 4.22 and 4.29 and assuming   0.3 , the following relationship between

moment ratio (ψ )and thickness ratio (ts/ta) can be obtained

 1 
 3 1   6
  H   ts     2.2  1  0
2
 
35.97  ta  0.83 (4.30)

This is a cubic equation in terms of the Mo/Mfx ratio, which can be solved for a given

ratio of ts/ta and the value of βH.

Table 4.2 presents a range of tank dimensions obtained from the standard design

procedure as per API 650 [2012]. The table includes elastic analysis results for Mo and

Mfx for each tank. It must be pointed out that these Mo values are for ‘full projection

length’ of bottom plate outside the tank wall (not the standard minimum of 50 mm

projection length).

102
 Table 4.2 Design Details for Several Typical Tank Sizes

Radius R, Height H, Shell Ann.Plate Moment Moment

All. Design
Group Thickness thickness Mo, Mfx, ts/ta
Mfx /Mo
Stress, MPa mm mm
ts, mm ta, mm N-mm N-mm

I 159 30000 12000 22 6

4675 22273 4.76 3.67

37500 12000 27 6
4994 33716 6.75 4.50

40000 12000 28 6
5165 37161 7.20 4.67

57500 12000 38 8
9561 70379 7.36 4.75

27500 14400 24 6
4874 26976 5.53 4.00

37500 14400 32 6
5251 48262 9.19 5.33

45000 14400 38 8
9153 67936 7.42 4.75

25000 16800 26 6
4897 31207 6.37 4.33

32500 16800 33 8
8737 50924 5.83 4.13

38500 16800 39 9
11151 70635 6.33 4.33

25000 19200 30 6
5010 41248 8.23 5.00

31250 19200 37 8
8832 63032 7.14 4.63

II 208 37500 12000 21 7

7661 26440 3.45 3.00

52500 12000 29 12
19796 49898 2.52 2.42

60000 12000 31 12
21941 60402 2.75 2.58

103
 Radius R, Height H, Shell Ann.Plate Moment Moment
All. Design
Group Thickness thickness Mo, Mfx, ts/ta
Mfx /Mo
Stress, MPa mm mm
ts, mm ta, mm N-mm N-mm

32500 14400 22 7
7840 29169 3.72 3.14

50000 14400 32 12
22026 63723 2.89 2.67

60000 14400 37 14
30642 87262 2.85 2.64

30000 16800 24 7
8057 34498 4.28 3.43

45000 16800 34 14
28202 71935 2.55 2.43

60000 16800 44 16
40341 121759 3.02 2.75

32500 19200 29 12
20280 51607 2.54 2.42

42500 19200 38 16
35204 87312 2.48 2.38

50500 19200 44 16
39276 119000 3.03 2.75

III 236 32500 14400 20 10

13174 26576 2.02 2.00

50500 14400 30 14
27810 60457 2.17 2.14

60000 14400 33 18
42757 78207 1.83 1.83

30000 16800 21 11
15759 30265 1.92 1.91

42500 16800 29 14
27821 58309 2.10 2.07

60000 16800 39 22
61209 108436 1.77 1.77

30000 19200 24 11
17717 39628 2.24 2.18

104
 Radius R, Height H, Shell Ann.Plate Moment Moment
All. Design
Group Thickness thickness Mo, Mfx, ts/ta
Mfx /Mo
Stress, MPa mm mm
ts, mm ta, mm N-mm N-mm

45000 19200 35 17
40936 85207 2.08 2.06

55000 19200 42 19
54683 123557 2.26 2.21

IV 145 3000 14400 5 5

511 637 1.25 1.00

1500 9600 5 5
170 212 1.25 1.00

7500 19200 9 6
2458 3807 1.55 1.50

13500 14400 12 6
3344 6775 2.03 2.00

33000 4800 10 6
2557 4265 1.67 1.67

12000 16800 12 6
3480 7056 2.03 2.00

Notes for Table 4.2

 Tank dimensions are based on Tables 3.1, K1, K2, K3 of API 650 [2012]
 The thickness values are rounded to next higher mm
 In determining the annular plate thickness, for the few cases where the
hydrostatic stress values were beyond the limits in Table 3.1 of API 650 [2012],
the thickness were suitably assumed
 The last row of the table is not taken from API 650. The appropriate design
values are added for completeness of design range

4.8.1 Influence of the Term (1-1/𝛃H)

In Eq.4.30, the only variable other than thickness ratio and moment ratio is the

term (1-1/H). This term is a function of H and R (since the shell thickness ts, present in

the shell parameter β is also a function of R). Table 4.3 shows the variation of (1-1/H)

for the minimum and maximum values of R and H specified in API 650 tank data:

105
 Table 4.3 Variation of (1-1/H) with Radius and Height

R H (1-1/H)

1500 4800 0.9965

60000 16800 0.9225

1500 16800 0.9981

60000 4800 0.8584

It can be seen that only the case where the radius is extremely large and height is

relatively very small has the (1-1/H) value less than the 0.91-0.99 range. For practical

applications this is a rare situation.

To study the influence of (1-1/H) on Mfx/Mo ratio (for full projection length), an

example tank with the following geometry is considered. The geometry is chosen such

that it has a high moment ratio (assuming the higher moment ratio will be influenced

more by the (1-1/H) variation). Let R = 37500 mm, H = 14400 mm, ts = 32mm, tb =

6mm. Table 4.4 gives Mfx/Mo values for three different assumed values of (1-1/H) for

this tank using Eq.4.30.

Table 4.4 Variation of Mfx/Mo with (1-1/H) Value

(1-1/H) Mfx/Mo
0.999 9.362
0.900 9.068
0.858 8.936

106
The variation of (1-1/H) produces a variation of 4.6% between the extreme Mfx/Mo

values. For the dimensions given, the actual (1-1/H) = 0.9408 and Mfx/Mo = 9.1912.

The deviation of the extreme value from the actual Mfx/Mo is only 1.86%.

Fig. 4.12 Mfx/Mo vs. ts/tp for Different Values of (1-1/H)

Extending the analysis to the entire design data as given in Table 4.2, the above

graph (Fig.4.12) is plotted. Figure 4.12 shows the plots of Mfx/Mo ratio for a constant

value of (1-1/H) as 0.92, 0.945, and 0.98. It can be seen that, all these three cases

coincide very closely with the actual Mfx/Mo values.

Assuming an average value of 0.945 for the (1-1/H) term, Eq.4.30 can be

expressed as

107
 6
3  ts

    45.86 2  83.74  38.06  0 (4.31)
 ta 

Equation 4.31 can be used to find the value of the elastic moment that will be

induced in the shell to bottom joint. This moment could form the boundary condition to

find the stresses in the shell and bottom plate provided that plastic hinge formation does

not limit the bottom moment. It should be noted that the compression due to wall weight

is not included in this analysis but can be done if needed.

Fig. 4.13 Bottom Moment for Thickness Ratio

The values from Table 4.2 are plotted again in Fig. 4.13. Although Eqs. 4.31 and

4.32 are cubic, the figure shows that Mfx/Mo and ts/ta have a simple quadratic relationship.

It can be obtained from curve fitting the graph in Fig. 4.13 as follows:

108
 2
t  t 
  0.3138  s   0.1429  s   1.0514 (4.31a)
 ta   ta 

It is to be noted that even though this relationship is deduced from design data (Table 4.2)
using minimum required thickness for the annular plate, it works for other cases when the
annular plate thickness used is greater than the minimum required.

4.9 VERIFICATION OF THEORETICAL RESULTS WITH FEA

A typical tank Model, (Case 1 as described in Chapter 3) is used for study. Figure

4.14 shows the bending stress in the bottom plate due to hydrostatic loading. Due to

higher allowable stresses and larger tank size, the maximum stress values for this

example are higher than those for the previous example considered.

Fig. 4.14 Bending Stresses in Bottom Plate from FEA

The values of dimensions “a” and “b” (as per Fig.4.11) from FEA and theory are

compared as given below:

109
 Table 4.5 Comparison of Theoretical and FEA Results for Uplift Lengths ‘a’ and ‘b’

FEA Theory

a, mm 96 94

b, mm 332 317

The values are in good agreement with each other. The minor difference can be

due to the fact that in theoretical model the tank wall is idealized as a single concentrated

point, whereas in FE Model (PLANE element Model), thickness of tank wall is

considered. In the calculations above, the self-weight has been ignored both in FEA and

theoretical calculations. If the self-weight of the tank is included, the proposed theoretical

approach with increased projection gives a slightly lower value for the parameters, ‘a’

and ‘b’ and a slightly higher value for the moment at the shell to bottom junction.

In practical applications, the shell wall for this height will have different course

thicknesses along the height (Fig.3.19). The multiple thicknesses have two effects, viz.,

reduction in self-weight and reduction in stiffness along the height. From the theoretical

analysis it is found that reduction in self-weight alone will increase the value of ‘a’ & ‘b’

parameters and reduce the internal moment, while reduction in stiffness will increase the

value of ‘a’ and ‘b’ parameters and also increase the internal moment. Hence, ignoring

the self-weight will have conservative effect in majority of the cases.

In order to support the above statement, the influence of considering or ignoring

the self-weight and assuming uniform wall thickness (even in case of tanks with multiple

wall thicknesses) on the parameters ‘a’ & ‘b’ and the stresses on the inner and outer side

110
of the shell to bottom joint is studied in detail. In the theoretical model, the self-weight of

the tank wall is considered as a concentrated force acting at the shell to bottom joint. It is

easy to find the actual self-weight for a multiple layered or single layered tank and apply

it as force at the joint. But for stiffness considerations, a uniform thick wall is assumed

(for tanks with single/multiple wall thicknesses). The question arises, whether it is

acceptable to use the bottom layer thickness for the entire tank wall height for stiffness

purposes. The argument in support of using it is that the height of the bottom course is

generally 2400mm (except for rare cases where it is 1800mm) and this height, even

though less than 3.46 Dt s (i.e. 4.89 Rt s ) the change in thickness above it will not

influence the stiffness at the bottom joint significantly.

Table 4.6 presents data obtained from the CASE 2 model (tank with multiple

courses) as described in Chapter 3. The actual data obtained from the ANSYS is

compared with different scenarios in the theoretical model.

Table 4.6 Effect of Self-Weight on Parameters ‘a’ and ‘b’

Stress at shell to bottom Wall thickness used for Stiffness

S.No a, mm b, mm Self-weight
joint, MPa Calculations
Outer Side Inner Side

1 185 635 325 -318 Included -VT DATA obtained from FEM
2 132 591 386 -298 Included - UT Uniform thickness

3 183 618 300 -320 Excluded Uniform thickness

4 149 601 355 -306 Included -VT Uniform thickness

5 151 609 363 -314 Included -VT uniform thickness is reduced by 10%

6 152 613 368 -318 Included -VT uniform thickness is reduced by 25%

111
Notes:
 VT: Variable Thickness, - the self-weight is calculated for the tank with

multiple layers

 UT: Uniform Thickness - the self-weight is calculated for the tank assuming

that the entire height of tank wall has same thickness as that of the bottom layer

 “uniform thickness” in the stiffness calculations indicate that the thickness of

the entire wall is assumed same as bottom layer thickness

From examining the data in Table 4.6, it can be seen that, for finding the required

projection length of the bottom plate and the corresponding stress values on either side of

the joint, excluding the self-weight and assuming the bottom layer thickness for the entire

wall for stiffness is appropriate. This assumption will give a slightly lower value of

bending stress on the outer side of joint. Since it is a location outside the tank, and

yielding and plastic hinge are allowed in this zone, this can be ignored. Similar to the

case with ambient temperature tanks, providing extra projection length in elevated

temperature tanks reduces the plate stress on the inner side of the shell to bottom joint

and increases the stresses on the outer side. Apart from the joint, the stresses are also

generally reduced in the inner portion of the bottom plate as shown in Fig. 4.15.

112
 A

Fig. 4.15 Bending Stress in the Bottom Plate of Elevated Temperature Tank

The figure refers to a tank with 30m radius, 12m height, 18mm shell thickness

and 6mm bottom plate subjected to a uniform temperature rise of 347F (175C).

Axisymmetric shell elements are used to model the tank wall and bottom plate with

contact elements for bottom plate - subgrade interaction.

4.10 EFFECT OF SELF-WEIGHT ON STRESSES (COMPARISON

USING FEA)
This section discusses the effect of self-weight on the stresses in the tank wall and

bottom plate. A tank with multiple wall thicknesses (CASE 2 - described in Chapter 3) is

considered. It is a fairly big tank in the upper end of the design range. Two FE models

with and without the self-weight effect are used for the study

113
Fig. 4.16 Bending Stress in Bottom Plate (Bottom Side) – With and Without Self-weight

. Figures 4.16 and 4.17 show the bending stress in the bottom plate for three

different load factors (0.7, 1, and 1.4). “LF” indicates Load Factor; a value of 1 indicates

that the tank is filled fully with a liquid of specific gravity 1. Fig 4.17 is the same as

Fig.4.16 but zoomed and expanded near the shell-to-bottom joint.

114
 Wall Thickness

Fig. 4.17 Bending Stress in Bottom Plate – With and Without Self wt. – Expanded View
Near Tank Wall

From the plots above it can be seen that the self-weight affects the bending stress

of the bottom plate only in a small portion on the outer side of the shell to bottom joint.

Inner side, the stresses are not significantly affected by the self-weight. For the LF=1

case, the maximum difference occurs close to the joint (3mm from the shell on the outer

side). The stresses increase from 133 MPa to 152 MPa with the inclusion of self-weight

effect (14.3% increase). For small tanks or for lower loads, it can be expected that the

percentage change will be much less. However, it should be noted that this zone is

outside the tank and yielding is permissible in this region. Hence ignoring the self-weight

may not have adverse effects.

115
 D

Fig. 4.18 von Mises Stress in Shell Wall – With and Without Self-weight Consideration

Figure 4.18 shows the stresses in the tank wall. As expected, shell stresses are not

significantly affected by the self-weight. The maximum difference in stress for load case

LF=1 is only 3 MPa higher with self-weight. The zones marked as ‘D’ in the stress plot

for each load case are due to discontinuity resulting from change in tank wall thickness.

In case of openings at tank bottom, it can be qualitatively said that there will not

be uplift of the bottom plate directly below the opening and hence the projection length

determined using the proposed analysis will be conservative. However, the uplift length

may be larger near the edges of opening because of the extra moments transmitted from

the shell. This may cause the strains in the bottom plate to slightly increase on the inner

side. The nozzle loads may increase or decrease the corresponding bottom plate stresses

slightly based on the direction of the nozzle load.

116
4.11 CLOSURE

The shell to bottom joint is an important high stress location in the tank where the

material is close to yield or in plastic range. This joint is a critical failure zone for tank

failure. Hence proper evaluation of stresses in this zone is very important. Denham’s

beam model (where the projection length is assumed to be fixed as 3”) is reviewed and

validated using FEA. The model is extended to any arbitrary projection length. Using this

theoretical approach and FEA, the effect of bottom plate projection beyond the shell wall

has been studied for tanks on rigid ring wall foundations. It is found that the length of

projection changes the stresses in the shell as well as the bottom plate near the shell to

bottom joint. Increasing the length is beneficial. The stresses will not alter if the length is

increased beyond a certain limit, which is being termed as ‘full projection length’. An

extension to Denham’s model is proposed to determine this ‘full projection length’. The

theoretical results are validated using FEA. The effect of assuming a uniform thickness

for a variable thickness tank wall for the sake of determining the ‘full projection length’

is studied and found to be acceptable. Similarly the results of including or excluding the

self-weight of the tanks are compared. The self-weight of tank does not increase the

stresses in the tank wall or the bottom plate on the inner side of the shell to bottom joint,

whereas it may increase the stresses marginally on the outer side of the joint.

An expression for the tank wall bottom moment has been presented for ambient

temperature tanks. Using the expression the designer can readily find the moment at the

shell to bottom joint and hence the stresses in that zone. This equation is based on a fully

elastic model and may be used to find the stresses near the shell to bottom joint.

117
CHAPTER 5

INCORPORATION OF FRICTION COEFFICIENT IN

ELEVATED TEMPERATURE TANK DESIGN

Storage tanks operating at elevated temperatures (200F to 500F) need to

consider stresses due to thermal expansions and restraints to the expansion in addition to

the design requirements for ambient temperature tanks. Appendix M of API Standard 650

provides additional requirements and guidelines for the design of tanks operating at

elevated temperatures. These are based on Karcher’s method [Karcher, 1978a, b] which

gives a simplified procedure for determining the stresses in the tank wall and bottom

plate. During temperature cycle, the tank will expand radially at the bottom. This

expansion is partially or fully restrained by friction between the tank bottom and

foundation components. A factor named ‘C’ is used for defining the ratio of actual

expansion to free expansion of the tank. Such partial expansion causes significant thermal

stresses. API uses these stresses to estimate the low cycle fatigue life of the tanks. At

present, a range of ‘C’ values (0.25–1.0) is allowed by API without clear guidelines for

selecting a suitable value. If no other choice is available, API 650 specifies that ‘C’ be

taken as 0.85. API 650 does not provide any guidelines to determine the ‘C’ factor. The

present study is aimed at systematic estimation of the ‘C’ factor by relating it to the

friction coefficient. This Chapter evaluates the current procedure and suggests an

118
alternate method by incorporating the friction coefficient directly in the stress equations,

instead of the ‘C’ factor.

5.0 BACKGROUND
In many industries, storage tanks are needed to hold liquids like asphalt,

residuum, high pour point hydrocarbons, etc., at higher temperature than the ambient.

This gives rise to thermal stresses in the tank, especially near the shell-to-bottom joint.

The magnitude of thermal stress occurring in the tank wall and the bottom plate is

influenced by the amount of radial restraint, which in turn depends upon factors like

liquid pressure, friction between the bottom plate and the foundation, piping connections

and differential settlement. In addition, from low cycle fatigue point of view, the cyclic

loading due to continual filling and draw down is reported [Long and Garner, 2004] to be

more severe for elevated temperature tanks compared to that for ambient temperature

tanks.

The general basis of the tank wall design for tanks operating at ambient

temperature (any temperature less than 200F or 93C) is to limit the hoop stress to the

allowable limits using either the 1-Foot Method or the more accurate Variable-Design-

Point Method. American Petroleum Institute Standard, API 650 [2012] adopted these

design procedures which were originally proposed by Zick and McGrath [1986]. In the

case of elevated temperature tanks, API 650 provides guidelines in Appendix M

“Additional Requirements for Tanks Operating at Elevated Temperatures.” This gives

requirements for operation of large storage tanks at temperatures up to 500 F (260 C). It

119
is recommended that, for designing the tank wall, the same procedure as that of ambient

temperature tanks shall be used with modified allowable stress limits based on strength

reduction factors for the design temperature. This Appendix provides design

considerations for limiting the loadings and strains resulting from thermal effects, such as

differential thermal expansion, tank bottom plate buckling and thermal cycling.

A typical complete thermal cycle for an elevated temperature tank can be

described as follows [Jones and Seshadri, 1989]:

 Gasoil is pumped in to the storage tank and then heated to a specified

temperature

 The Gasoil is heated slowly to allow any excess water to be removed by

drainage or blow off. This step will also allow gradual thermal expansion of the tank

and shell

 The tank is filled with a high pour point hydrocarbon. When there is a demand

for stored fluid, the tank is drawn down to a specified level and the procedure is

repeated as needed.

For tanks with maximum design temperature greater than 93C (200°F), particular

consideration should be given to the following thermal effects [API 650]:

 Temperature differences between the tank bottom and the lower portion of the

shell. Such thermal differences may result from factors such as the method and

sequence of filling and heating or cooling, the degree of internal circulation, and

heat losses to the foundation and from the shell to the atmosphere. With such

120
 temperature differences, it may be necessary to provide for increased piping

flexibility, an improved bottom-to-shell joint, and a thicker annular ring or bottom

sketch plates to compensate for increased rotation of the bottom-to-shell joint.

 The ability of the bottom to expand thermally, which may be limited by the

method of filling and heating. With such a condition, it may be necessary to

provide improved bottom welding in addition to the details suggested above.

 Temperature differences or gradients between members, such as the shell and

the roof or stairways, the shell and stiffeners, the roof or shell and the roof

supports, and locations with insulation discontinuities.

 Whether or not the contents are allowed to solidify and are later reheated to a

liquid, including the effect on columns, beams, and rafters. The possible build-up

of solids on these components and the potential for plugging of the vent system

should also be considered.

 The number and magnitude of temperature cycles the tank is expected to

undergo during its design life.

The shell-to-bottom joint is a critical location vulnerable in many modes of

failure. Denham, et al. [1968a, b], Wu and Liu [2000] and Sathyanarayanan and Adluri

[2011] have proposed methods to determine the stresses in the bottom plate near this joint

for ambient temperature tanks. As mentioned above, this joint is subjected to low cycle

fatigue due to cyclic loading of the stored liquid (fill–draw down cycle) or the fluctuation

in the temperature of the stored liquid or both. Karcher [1978a, 1978 b, 1981a, 1981b]

carried out a ground breaking and very useful study on elevated temperature tanks. He

121
provided stress equations based on an assumed plastic hinge boundary condition at the

shell-to-bottom joint. Based on these, he proposed a simplified procedure to find the

fatigue design life of the tank considering the maximum stress developed at the shell-to-

bottom joint and a relevant stress concentration factor. API 650 has adopted Karcher’s

procedure for determining the design cyclic life of the tank.

5.1 CURRENT PRACTICE

The current practice assumes uniform temperature increase in the tank, i.e.,

without any temperature gradient across the thickness of the shell wall. This is a

reasonable assumption since the shell wall thickness is relatively thin and carbon steel is

a conductive material. At steady state, the difference in temperature should not be

significant. In addition to these, the temperature difference (if any) between the bottom

region of the shell and the bottom plate that might be caused by heat loss is ignored. The

stresses are determined using shell theory for tank wall while beam-on-elastic foundation

theory is used for the bottom plate. It is assumed that complete plastic hinges are formed

in the bottom plate near the shell-to-bottom joint [Karcher 1978 a, 1978b, 1981a, 1981b],

while the shell stresses are assumed to be elastic. Essentially the equations as given by

Timoshenko and Woinowsky-Kreiger [1959] for cylindrical tanks with clamped end

conditions are suitably modified to have the plastic moment at the bottom as the

boundary condition. The hydrostatic stresses caused by liquid infill are combined with

the stresses that may be induced due to heating of the filled liquid. Two plastic hinges are

122
assumed at the bottom plate for a rigid foundation (such as concrete ring wall or pile cap)

while one hinge is used for flexible subgrade below the tank wall.

5.1.1 Stresses in Shell (Tank Wall)

Based on the shell theory [Timoshenko and Woinowsky-Kreiger, 1959], the

following equations were given by Karcher [1981a, b] with an assumption of a single

plastic hinge at the shell bottom (applicable for the earthen foundation case),

The radial deflection of the tank wall (ys) is given by

R 2 
 TEt s  CTEt s  22 M o 

ys   H  x s   exs  H   cos  x  sin  x s 
R R 
s
Et s 
    (5.1)


where,  is the coefficient of thermal expansion, T is the uniform temperature rise, C is

the reduction factor for T to account for the friction restraint, Mo is the moment at the

shell to bottom joint and ‘xs’ is the height of a point along the tank wall at which the

deflection is calculated.

The bending moment, hoop force and shear force of the tank wall at a distance

‘xs’ from the tank bottom are given by

 xs  CTEt s  22Mo 

M xs   e  H   sin  x  cos x s 
R 
s
2 2
   (5.2)


  CTEt s  22Mo  
N  R H  x s  exs  H   cos x s  sin x s  

  R    
 (5.3)

 xs 
 CTEt s  22Mo 
Q xs   e  H    cos x s  sin x s    sin x s  cos x s 
2 
 R    (5.4)


123
 If T in the equations above is assumed to be zero, then these equations reduce to

those for a cylindrical shell loaded by hydrostatic pressure as given by Timoshenko and

Woinowsky-Kreiger [1959]. These equations can be used to find maximum stress

intensities in the shell and hence fatigue evaluation can be carried out, if required, for the

loading cycles that are expected during the design life of the tank.

All the above equations were developed based on the assumption that the boundary

condition at the shell to bottom joint is represented by a moment Mo and zero radial

deflection. Both of these assumptions (with Mo replaced by a single plastic hinge) were

first used by Zick and McGrath [1968] for finding the stresses in the shell wall for an

ambient temperature tank. The basis of these assumptions is given in their report to the

API. The validity of the first assumption will be discussed in Chapter 6. Regarding the

second assumption, Karcher [1978a, 1978 b] explains that even though the assumption of

zero deflection seems to be open to question, in actual practice, an elevated temperature

tank is usually filled initially to a level of several meters with only partially heated or

cold product. Then internal heaters and mixers are used to increase and maintain product

temperature. This start up procedure (with a smaller liquid head being heated to full

temperature) results in the tank bottom expanding more freely than the case where the

full liquid head is heated from ambient temperature. This results in lower frictional

resistance between tank bottom and the foundation. Preliminary observations on several

tanks indicate that the annular plate expands radially by about one half of the calculated

value of RT. Therefore assuming that the annular plate expansion is between zero and

124
half of RT is a reasonable approximation with the former being more conservative

[Karcher, 1981a].

Karcher accounted for the partial radial expansion of bottom plate by employing a

reduction factor (‘C’) in the equations above. A ‘C’ value of 1 assumes a fully restrained

bottom to shell junction (0% radial expansion) and a value of zero assumes free (100%)

radial expansion. The bottom will have free expansion (equal to RT) if there is

essentially no friction. But in reality, there are friction forces under the tank

bottom/annular plate (including backing plates, etc.) and hence the plate expands partially

(0.25≤C≤1.0). The magnitude of friction forces also depends on the liquid head in the

tank available at the time of heating.

5.1.2 Stresses in Annular Plate

The shell-to-bottom joint location is a region of high strains and plastic hinges are

assumed to form in the bottom plate near this location. The stress created here can be

classified as secondary bending stress and hence should be limited to twice the yield

strength of the annular plate material to assure shakedown to an elastic action. The

pseudo bending stress representing the strain range that occurs in the tank annular plate at

the shell-to-bottom joint is obtained by the following procedure [Karcher, 1978a]:

 From Eq. 5.1, assuming a fully plastic annular plate moment at the shell bottom

(i.e., Mo = Mp), the slope at the bottom of the tank wall is determined as

dys 
 R2  
 CTEts  3 SY tb2 
S   1-  H -  
dxs xs 0
Ets 
 R 2   (5.5)

125
 In order to find the strain range in the annular plate at the shell-to-bottom

junction, it is assumed that annular plate reacts like a beam-on-elastic foundation.

3K
Following Hetenyi [1971], the moment in the plate is M = K/43, where, Ω = 4
Etb3

and K is the modulus of the subgrade.

 The value of slope as obtained from Eq. 5.5 is used in M = K/43 resulting in

0.658R 2 Ktb 
 CαTEts β 3 SY tb2 
Sb   4
1  H   
ts E 
 γR 2γ   (5.6)

where, Sb is the pseudo bending stress representing the strain range at the bottom

plate

Equation 5.6 is valid only if the following condition is true.

 CαTEts  β SY t b
3 2
 H    
 R  2
(5.7)

This condition ensures that the tank is loaded thermally or hydrostatically such that a

single plastic hinge forms in the bottom plate near the shell-to-bottom joint. The safe

design cycle life, N, is given by,

2.44
 9700 
N   
 Kc S  (5.8)

126
where, S is the alternating stress range in MPa (=Sb/2), Kc is the stress concentration

factor. This procedure to ensure the safe cyclic life of tank can be avoided if a significant

head of heated liquid (close to design temperature) is maintained in the tank between

cycles. A significant liquid head can be defined as being equal to or greater than

0.4(Dts)0.5 ft, where, diameter D is in feet and ts is in inches [API 650].

Fig. 5.1 Meridional Moment in the tank [Karcher, 1978a]

Figures 5.1 and 5.2 demonstrate the effect of temperature rise on moments and

equivalent stresses (strain range) in the tank wall and the bottom plate. The results show

that these stresses can be significant on the lower shell courses and annular plate regions

close to shell bottom. Equations 5.5 – 5.8 depends on the C factor and hence using a

proper C value becomes important.

127
 Fig. 5.2 Annular Plate Bending Stress Caused by Hydrostatic and Thermal Loads
[Karcher, 1978a]

Equations 5.1-5.4 are valid for ambient and elevated temperature tanks. The basic

assumptions involved in these equations regarding radial growth characteristics and the

bottom moment of the tank wall have significant influence on the stresses in the tank

wall. Regarding the radial deformation, as mentioned earlier, it is obvious that bottom

plate can expand anywhere between zero to full radial growth (RT). No guidelines for

selecting the actual value (within the allowable range) are available either from API or

other sources. Although the current practice has not been publicly reported to have

resulted in field failures and hence can be construed to be safe, use of a set value of C for

all situations is not appropriate as discussed later in the Chapter. An equivalent but

considerably less problematic approach using friction coefficient is discussed below.

128
5.2 INCORPORATION OF FRICTION COEFFICIENT IN DESIGN
EXPRESSIONS
If the tank foundation does not restrain the expansion due to temperature (at least

partially), then the thermal stresses will be minimal. However the foundation generally

offers significant frictional resistance that restrict the free radial expansion of elevated

temperature tanks which in turn induces additional thermal stresses in the tank wall. The

following model obtains the partial radial expansion by first finding the effective

contraction due to the friction forces and subtracting it from the free thermal expansion.

The treatment is similar to that of a disc subjected to centrifugal forces. Considering the

equilibrium of an infinitesimal element in the bottom plate of the tank as shown in Fig.

5.3 (a).

 rr  d rr r  dr  d t p   rr (r d ) t p  2   d  dr t p  r d dr   H (5.9)

 2

where, σrr and σ are the stress components in the radial and circumferential directions,

 is the coefficient of friction between the bottom plate and the foundation as appropriate

to the given conditions. It is generally assumed that the tank bottom is resting on well

compacted dry sand/gravel. The equation assumes that the friction forces are

predominantly in the membrane direction of the bottom plate [Sathyanarayanan and

Adluri, 2013c].

Neglecting higher order terms in Eq.5.9,

d rr 1 H
  rr       (5.10)
dr r tp

129
 dr

r d


Bottom
Plate
xs

(a)
σrr + d σrr

σ

Tank wall

Mo
σrr
σ Hydrostatic
(b) Tank weight
℄ Pressure H Qo

℄
Annular Plate Friction Forces (c)

Fig. 5.3 Friction Forces Below Bottom Plate

For axisymmetric plates the stress displacement relationships are given by [Boresi and

Schmidt, 1993]

130
 E  du u
 rr  2 
  (5.11)
1    dr r

 du u 
E
      (5.12)
1   dr r 
2

where, u is the radial displacement at any point in the bottom plate due to friction forces.

The negative sign in Eq. 5.10 indicates the contraction of the plate.

Substituting (5.11) and (5.12) in (5.10)

E  d 2u 1 du u  H
   2 (5.13)
2  
1   dr 2
r dr r  tp

Solving Eq.5.13,

 1 2    H r 2 C
u  C1r  2
E tp 3 r

where, C1 and C2 are constants of integration. Using appropriate boundary conditions for

the given loading,

u
1     H r 2
(5.14)
3E tp

It must be pointed out that in addition to the above, there will be other effects

due to shear at the bottom, self-weight of bottom plate and tank wall stiffness (in relation

to that of the bottom plate). These can be shown [Adluri, 2012] to have only a small

influence. Other effects such as those due to local loss of contact with the foundation,

local bending of bottom plate due to bulging (or buckling), bottom plate lap welds, weld

131
reinforcements, backing strips, horizontal soil pressure due to uneven bottom surface of

the plate, etc., have also been omitted in the present treatment. Although the bottom plate

is treated with a single thickness here, multiple thickness plate along with other effects

omitted above can be incorporated. In a recent report, Adluri [2012] extended the

treatment given in the present chapter and discussed these issues at length along with the

effects of non-uniform friction below the bottom due to soil shearing under parts of the

bottom plate. The treatment in the present thesis is a general approach that can be used

with a single equivalent value for  for the entire bottom plate.

Hence for any tank subjected to uniform temperature increase T, the partial

radial expansion of the tank bottom after filling the liquid to a height H and heating it to

the required temperature, will be given by

  RT 
1     H R 2
(5.15)
3E tp

Note that H indicates the actual liquid height and not the constructed height of the tank

wall. The ‘C’ factor as used currently can be interpreted to represent the ratio between

the restraint deformation due to friction forces and the free expansion of the bottom plate.

Hence,

C
1     H R
(5.16)
3 E t p  T

132
The ‘C’ value will be 1.0 when the restraint deformation as calculated is equal to or more

than the free thermal expansion. The C-factor will be zero if the restraint deformation is

zero.

5.2.1 Incorporation of Friction Coefficient in Shell Equations

The governing equation for circular cylindrical shell (Fig.5.4) loaded

symmetrically with respect to its axis [Timoshenko and Woinowsky-Kreiger, 1959] is as

follows

d 4 ys w
4
 4 4 ys  (5.17)
dxs Ds

where, ys is the radial deflection, w is the liquid pressure acting inside the cylinder, β and
Ets3
Ds are shell parameters, where Ds 
12 1 2  
The general solution for the above equation is

ys  e xs  c1 cos  xs  c2 sin  xs   e  x  c3 cos  xs  c4 sin  xs   f ( xs ) (5.18)

For circular tanks with uniform wall thickness, Eq. 5.17 is modified as

d 4 ys   H  xs 
 4 4 ys  (5.19)
dxs4 Ds

A particular solution of Eq.5.19 is

  H  xs  R2
ys  (5.20)
Ets

The complete solution of Eq. 5.19 is

  H  xs  R2
ys  e xs  c1 cos  xs  c2 sin  xs   e  xs  c3 cos  xs  c4 sin  xs  
Ets (5.21)

133
Considering the shell as infinitely long; c1 = c2 = 0

H
xs
w
ys (outward
positive) Qo
Mo

Fig. 5.4 Forces in Tank

Using the boundary condition that there is no radial expansion at the bottom of the

tank (for hydraulic loading only) and the moment at the tank bottom is Mo, Eq. 5.22 is

obtained. The moment (Mo) at the tank bottom can be a maximum of one/two times the

plastic moment capacity of the bottom plate. It is denoted by Mb.

In case of earthen foundation, with a single plastic hinge on the

S y t 2p
inside Mb  and in case of concrete ring wall foundation, with a plastic hinge on
4

S y t 2p
either side of shell Mb  . This gives,
2

  HR2 Mb    H  xs  R2
ys  e  xs  cos  xs  sin  xs  (5.22)
 Et  2  Ets
 s 2 Ds 

134
dys  Mb  HR2   R2
  e  xs   cos  x  sin  x    cos  x  sin  x 
s  
dxs  2D  2 s s
Ets
s  Et (5.23)
 s  s

d 2 ys   HR2 Mb 
 2 2e  xs  sin  xs  cos  x  …….. (5.24)
2  Et 2 Ds  2
s
dxs  s 

5.2.2 Thermal Stresses in Cylinders

δ = RT δ = RT

MT MT
QT QT
(a) Free Radial Expansion of Tank  1    H r 2
under Temperature Change δ = E tp 3

(b) Thermal Expansion of Tank with

Frictional Forces
Fig. 5.5 Schematic Expansion of Tank

In Fig. 5.5, δ is the free expansion of cylinder for a uniform temperature increase

of T. If friction is present, δ is the expansion at the bottom of the cylinder (and the

bottom plate) for a uniform temperature increase. From Eq.5.17 & 5.18 and considering

the shell as infinitely long with only temperature loading, Eq.5.25 can be obtained. The

term f(xs) representing the particular solution in Eq.5.18 becomes zero since there is no

pressure loading for pure temperature loading (which can be added to the separate

solution obtained for hydrostatic loading).

135
 ys  e  xs  c3 cos  xs  c4 sin  xs  (5.25)

dys
  e-  xs c4  cos  xs -sin  xs  - c3  cos  xs  sin  xs  . (5.26)
dxs

... (5.27)
d 2 ys
 2 2e  xs  c3 sin  xs  c4 cos  xs 
dxs2

d 3 ys …
 2 3e  xs  c3 cos  xs  c3 sin  xs    c4 sin  xs  c4 cos  xs 
(5.28)
dxs3

Let MT and QT be the moment and shear force induced at the bottom due to thermal

loading,

 d 2 ys 
 M x x 0
s
  Ds 
 dx 2   MT (5.29)
s
 s 

 dM xs   d 3 ys 
 x 0
Qxs 
dx
   Ds
 dx 3 
(5.30)
s  s x0  s xs 0

From Eq. 5.29 and Eq. 5.27

MT
c4  (5.31)
2 Ds  2
From Eq. 5.30 and Eq. 5.28

1
c3    Qt   MT 
2 Ds  3 (5.32)

Substituting the constants,

136
 e  xs e  xs MT
ys   QT   MT  cos  xs  sin  xs (5.33)
2 Ds  3 2 Ds  2

dys e-  xs   QT   MT  
 MT  cos  xs -sin  xs  -   cos  xs  sin  xs 
(5.34)

dxs 2 Ds      
Considering the boundary condition as shown in Fig. 5.5 (b),

1  R2  H
 ys  xs 0   (5.35)
3t p E

Equating Eq. 5.33 and Eq. 5.35 at xs = 0

1  1  2
 QT   MT   R  H (5.36)
2 D 3 3t p E

Using Eq. 5.36 in Eq. 5.33

 1  2  e  xs MT
ys  e  xs  R  H  cos  xs  sin  xs
 3t p E   2
  2 D s (5.37)

dys  M  1  2  
 e-  xs   T
 cos  x -sin  x  -  R  H   cos  x  sin  x   (5.38)
dxs  2 Ds 
2 s s  3t p E  s s

 

d 2 ys   1  2  MT 
 2e-  xs  2   R  H  sin  xs  cos  xs (5.39)
dxs2   3t p E   2

  2 Ds

Using the principle of superposition, displacement (w), moment (Mx) and shear

(Qx) due to the liquid infill and the temperature rise is given by

ys  e   xs
 1 1    R 2 H
 cos  xs 
 MT  M b     H  xs  R 2
sin  xs  
 
 ts  E  2
3t p  2 Ds  E ts (5.40)

137
   MT  M b  
  cos  xs  sin  xs  
 2 Ds 
2
dys   R2
  e βxs    Et
dxs   
1 1     R 2 H
cos  xs  sin  xs  
  ts

 E
 
s ..(5.41)
3t p
   

  MT  M b  
 cos  xs 
 2 Ds 
2
d 2 ys 
M xs   Ds  2 2 Ds e  βxs   
   1  1     R  H sin  x 
dxs 2  2 (5.42)
  ts 3t p  E s
   

 1 1    R 2 H 
    cos  xs  sin  xs  
d 3 ys  ts 3t p  E
 
Qxs   Ds  2 3 Ds e βxs  
   M T  M b  sin  x  cos  x
3
dxs 
  s s 
(5.43)
 2 Ds  2


 1 1    R 2 H 
     cos  xs 
e βxs Ets  ts 3t p  E
 
N       H  xs  R
R   M T  M b  sin  xs  (5.44)
 2 Ds  2 
 

where, MT and Mb represent the moment induced at the bottom due to thermal and

hydrostatic loadings. The sum of these moments cannot exceed 2Mp for rigid foundations

and Mp for flexible foundations. These equations (Eq. 5.40- 5.44) are similar to Eq. 5.1 -

5.4, where, C is replaced with the expression from Eq. 5.16. Equations 5.40 – 5.44 are

valid only if ‘δ’ from Eq.5.15 is greater than zero i.e. the thermal expansion of bottom

plate is not less than the restraint offered by the frictional forces at the tank bottom. If this

is not so, it implies that the bottom of the tank does not expand at all because of high

138
amount of frictional resistance. For this case, replace the term (1-)/3tp with ET/RH

in the eq. 5.40-5.44.

With respect to stresses in the tank wall, Karcher’s equations will yield the same

results as those of the proposed model provided an exact C-factor is used. The value of

‘C’ has a significant influence on tank wall stresses. Predicting this factor during the

design stage or even during operation without any field measurements is a difficult task.

Hence using standard friction coefficients between known materials and determining the

stresses using Eq. 5.40 - 5.44 will be an appropriate choice.

5.3 RESULTS AND DISCUSSION

Finite element analyzes were carried out using ANSYS to explore the effect of

friction and validate the theory above. The foundation is assumed to be a concrete ring

wall with highly compacted infill. Axisymmetric SHELL elements were used for

modeling the tank wall and bottom plate. Both elastic and elastic-perfectly plastic

material models were used as described in Chapter3.

The example tank (CASE 1) from Chapter3 with T = 347F (175C) and  = 0.3

is considered for the study. Figure 5.6 shows theoretical deflection in tank wall with a

double hinge condition for bottom plate as specified by Zick and McGrath [1968] and

Karcher [1978 a,b,1981 a,b]. The shell stresses from nonlinear FEA (which does not rely

on the double hinge assumption) match well with theoretical stresses. This implicitly

139
confirms the formation of two plastic hinges in the bottom plate near shell bottom in the

FE model. The bending and hoop stresses for this analysis are shown in Figs.5.7 and 5.8.

It is to be noted that eventhough the bottom plate can reach plastic stage, the shell

stresses are designed to be within elastic limits for operational loads.

Fig. 5.6 Deflection in the Tank Wall Including the Effect of Friction (=0.3)

140
Fig. 5.7 Bending Stress (Sx) in Tank Wall Including Friction Effects (=0.3)

Fig. 5.8 Hoop Stress in Tank Wall Including Friction Effects (=0.3)

141
 The same tank with  = 0.5 is analyzed using elastic-plastic FE model. Figure 5.9

compares the FEA results and the theoretical values. The theoretical radial displacement

(horizontal movement) at the outer edge of the annular plate (Eq. 5.15) is 52.7 mm. The

corresponding radial displacement as obtained from FEA (including friction effects) is

53.4 mm. If there were no friction forces, the expansion would be RT=63mm (with

=12x106). For this example the value of C can be calculated as 0.15 (=[63-53.4]/63). It

can be seen that this value is far smaller than the value of 0.85 recommended by API 650.

Even though, the analysis indicates that two plastic hinges form at the bottom

plate for this case, a complete double hinge may not always form at the bottom for

general tanks. For the tank considered, two full hinges will not form if either the friction

coefficient or the temperature is less than the values used above. Hence using two plastic

hinges as boundary condition for lower temperature tanks or those with smaller friction

coefficient will underestimate the maximum stresses in the tank wall.

142
 115 MPa

Fig. 5.9 Bending Stress in the Tank Wall Including Friction Effects – (Elastic-Plastic
FEA)

Hoop and Bending stress values are obtained for a tank with the same dimensions

as before but with different values of  and C. The comparisons are shown in Fig. 5.10

and 5.11. From the figures it can be seen that the bending stress in tank wall for even a

friction coefficient as high as 1.0 is considerably less than the stress obtained using a C

value of 0.5. Hoop stress in the tank wall also shows a similar trend. Hence for the

example tank dimensions, using a C-factor of more than 0.5 is excessively conservative.

143
Fig. 5.10 Tank Wall Bending Stress for Various Values of ‘C’ and 

Fig. 5.11 Tank Wall Hoop Stress for Various Values of ‘C’ and 

144
 For a value of  in the range 0.30.5 the difference will be even more

pronounced. Figure 5.12 shows the variation of shell stresses as temperature increases

with friction being constant (for a friction coefficient of = 0.7).

Fig. 5.12 Tank Wall Bending Stress at Different Temperatures for  = 0.7

For a given tank, the bending stress caused by friction will be maximum when

u=RT or C =1. Hence from Eq. 5.14, for a particular friction coefficient

TL 
1     H R (5.45)
3 E tp

where, TL is the limiting temperature beyond which increase in temperature will not

increase the stresses. Conversely, the limiting friction coefficient can also be obtained as

below.

145
 3 E T t p
L  (5.46)
1    H R

where, L is the friction coefficient that will cause maximum stresses for a particular

temperature. If a friction coefficient value is chosen higher than L, there will be no

further increase in the stresses. In other words, this is the ‘’ at which the C becomes 1.

For any  greater than this value, C continues to be 1 as it represents zero expansion

condition.

For the example tank used earlier, TL is 40C, it can be seen from Fig. 5.12 that

the stresses (from FEA) in the tank wall do not increase when the temperature is

increased beyond this limit. Basically, if the restraint deformation due to friction is equal

to or greater than the free radial expansion due to temperature, the stresses in the tank

wall will reach a maximum and further increase in temperature does not have any impact.

For the same tank as above but with a friction coefficient of 0.7, the C-factor

changes with temperature as shown in Fig. 5.13. The plot indicates that C value is 1.0 till

the temperature is 40C, which is the temperature limit TL for this tank. The stresses in

the tank bottom increase till the temperature reaches the limit TL and remain constant

after this limit. Hence after TL, the C value reduces inversely with temperature such

that the state of stress at TL is maintained throughout. It should be noted that though C

value reduces, the product of T and C will remain constant. This is a unique value for

each tank and is directly proportional to the friction coefficient. The pattern for C-factor

as shown in Fig. 5.13 holds good for any typical tank. The tank dimensions will only alter

146
the value of TL limit. This observation implies that the present recommendation of

C=0.85 [API 650] for any tank may or may not be conservative (although no failures are

reported to the best of the author’s knowledge in general literature). If the temperature is

relatively low and the friction coefficient of the tank is high, then its C value could

sometimes be in the region A-B of Fig. 5.13 (before TL limit) and hence C = 1. For this

case, using C < 1 is unconservative. On the other hand, stresses predicted using C = 0.85

for temperatures at the end of region B-C will be very highly conservative.

C = 0.85

Fig. 5.13 Temperature Influence on C-Factor

Using a flat value for the C-factor (be it 0.5 or 0.85) gives inconsistent results. At

present there are no clear guidelines on choosing a value of C other than 0.85. The

margin of safety varies with operating temperature and could result in slightly

unconservative estimates temperature ranges less than TL and grossly conservative

predictions at higher temperatures. This is further illustrated in Table 5.1 which shows

147
the implied coefficient of friction for fixed values of C equal to 0.5 and 0.85 for typical

elevated temperature tanks.

Table 5.1 Implied Coefficient of Friction for Fixed Values of C-Factor

H D tb T Implied
m m mm
o
C 
11 16 6 93 3.32
11 16 6 260 9.29
11 90 7 93 0.69
C = 0.5

11 90 7 260 1.93
19.2 16 8 93 2.54
19.2 16 8 260 7.10
19.2 61 8 93 0.67
19.2 61 8 260 1.86

11 16 6 90 5.47
11 16 6 250 15.19
11 90 7 90 1.13
C = 0.85

11 90 7 250 3.15
19.2 16 8 90 4.18
19.2 16 8 250 11.60
19.2 61 9 90 1.23
19.2 90 11 90 1.02
19.2 61 9 250 3.42

The implied coefficient of friction in Table 5.1 is calculated using Eq. 5.16. As

can be seen, except for very large tanks at relatively low temperatures, the implied values

of  are unacceptable if the C-factor is fixed as 0.5. It is reported that C = 0.5 was based

on anecdotal observations on some tanks [Karcher, 1978a]. No Data regarding those

tanks is available. It is possible that they were very large tanks at temperatures in the 200

- 250F range. For these tanks, a C-factor of 0.5 may sometimes be acceptable. However,

148
as can be seen from the table, for other ranges of H, R and T, the implied values of 

(for C = 0.5) are too high. API 650 recommends a value of C = 0.85 in case other

information is not available. This is generally assumed to be a conservative

recommendation. However for most tanks, the implied values of  for this value of C are

even larger than those for C=0.5. Acceptable vales of  are implied only for large

diameter tanks with low temperatures. Therefore, instead of using a seemingly

conservative value for the C-factor, it may be preferable to use a conservative estimate

for the friction coefficient . The actual value depends upon the type of subgrade and

can be obtained from established literature or other means. Adluri [2012] provided

extensive discussion on this issue.

It should also be understood that friction resistance depends on liquid heights,

foundation and bottom plate/annular plate characteristics. The value of the friction

coefficient is not likely to alter significantly within the temperature range being

considered (unlike the change in yield strength based on temperature change as given in

Table M-1 of API 650). Because of this, thermal effects do not necessarily increase

linearly with temperature. The extraordinary values for implied  in Table 5.1 can be

understood by examining Eq. 5.15. Free thermal expansion is proportional to R and T

whereas the restraining displacement due to friction is proportional to R2 and H (and does

not depend significantly on T).

For smaller tanks in the Table, the friction resistance is not sufficient to provide

appreciable restraint against thermal expansion. As liquid height increases, friction

149
resistance proportionately increases while free thermal expansion (RT) stays the same.

When the radius increases, this effect is compounded. Hence as the tank size increases in

radius and height, the total friction force builds up and the restraining displacement starts

to approach the free thermal expansion. The opposite happens with increasing

temperature. These points can be further illustrated through Table 5.2 which lists the

implied C-factor for a fixed value of  = 0.85. For smaller tanks, the implied C-factor is

quite small. Only for large tanks with relatively low temperature increase, the implied

C-factor approaches the values currently being used. At the outer edge of the range of

parameters, the implied C-factor can become slightly unconservative if a lower C value is

chosen for design. Hence, it is more rational to choose a conservative estimate for the

coefficient of friction than for the C-factor.

Table 5.2 Implied C-Factor for Fixed Values of Coefficient of Friction

H D tb T Implied
o
m m mm C C-factor
11 16 6 90 0.13
11 16 6 250 0.05
11 90 7 90 0.64
0.85

11 90 7 250 0.23
19.2 16 8 90 0.17
=

19.2 16 8 250 0.06

19.2 61 9 90 0.59
19.2 90 11 90 0.71
19.2 61 9 250 0.21

150
5.3.1 Influence of Friction on Fatigue Life

In the current practice, the alternating stress required for fatigue life estimation of

tank (number of fill/draw cycles the tank can safely withstand) is determined by Eq. 5.6.

In this equation, the friction forces are indirectly represented by the C-factor. For the tank

example considered earlier, Fig. 5.14 shows the influence of C-factor on the fatigue life

of the tank (for T = 175C). The fatigue life reduces drastically if the C-factor is

increased.

1300 cycles

Fig. 5.14 Influence of C-Factor on Fatigue Life

As per API 650, the tank has to safely withstand at least 1300 fill/draw cycles. If C is

assumed to be 0.85 (and K=2) as recommended by the API, the tank fails by a significant

margin (N=369). If the same tank is considered with a (most likely conservative) friction

coefficient estimate, say, 0.85, the number of design life cycles (N) that the tank can

151
withstand becomes 1896 (which is well above the required limit). This is a case where the

API recommendation is overly conservative. If the tank is larger and is operating at

temperatures below 100 oC, using a smaller C value could sometimes be slightly

unconservative.

5.3.2 Influence of Filling Procedure

API 650 recognises (in Clause M.4.2) that the tank filling procedure influences

the amount of radial expansion and hence the stresses and fatigue life. Generally if the

tank is initially filled to a low height (at least 0.3(Dt)0.5, where, D is in ft and t is in

inches, [API 650] and heated to the required temperature as a first step, the tank will

radially expand almost to its full expansion and hence the thermal shell stresses induced

will not be very significant. However it must be noted that it may be difficult to guarantee

that this will always be followed when the tank is likely to be in operation for many

decades. On the other hand if the tank is heated with a significant liquid head, it induces

thermal stresses due to the friction forces. These stresses have been explored throughout

the paper. In order to verify the API recommendation of preheating the tank with low

liquid head, a finite element analysis was carried out. Using Eq. 5.15 for the example

tank used in finite element analysis, the radial expansion is 62 mm for a 2 m liquid head

heated to 175C. FEA results give 62.2 mm confirming the theoretical prediction. The

corresponding free expansion (RT) is 63 mm (with =12x106). The results show that

with very low liquid head, the radial expansion of the bottom is almost the same as that of

free radial expansion. The stresses for this analysis are very close to those of the same

152
tank at ambient temperature thus indicating that the thermal stresses can be neglected if

the above procedure for liquid filling and heating is used.

5.4 CONCLUSIONS

This present study examines the influence of bottom plate friction forces on the

stresses in the wall of elevated temperature tanks. Karcher’s equations used for finding

the stresses in the tank wall are verified using a finite element model with contact

elements. The equations use a value called as C-factor to include the effect of partial

expansion due to temperature change. It is observed that the equations give correct results

if the exact C-factor is used. Since the C-factor is highly sensitive in influencing the tank

wall stresses, the designer needs expertise and judgement in deciding on the factor during

the design stage. The current practice does not have clear guidelines for selecting a C

value. In the absence of such guidelines, a set value (like 0.85) for the C-factor is being

used irrespective of the tank dimensions and temperature change. The present study

shows that this approach is grossly conservative for most of the range of design

parameters (H, R & T). On the other hand, for very large tanks at relatively low

temperature changes, this could become slightly unconservative. This study shows that

the C-factor is directly related to the friction coefficient () and tank parameters.

Appropriate equations for stresses and by extension, fatigue life are presented. It is shown

that for a given tank with a particular friction coefficient, there is a limiting temperature

(TL) beyond which any further increase in temperature will not increase the tank wall

153
stresses. Similarly for a given temperature increase in a tank, the stresses will not increase

further even if the friction coefficient is higher than a particular value ( L). Several

related issues are discussed. The results are verified using finite element analysis

incorporating friction forces through contact elements between foundation and tank

bottom plate. The study shows that it is much more rational to use a conservative

estimate for the coefficient of friction than for the C-factor.

154
CHAPTER 6

FATIGUE ANALYSIS OF SHELL TO BOTTOM JOINT OF

TANKS

Shell to bottom joint of hydrocarbon storage tanks is a critical failure location

which needs careful evaluation especially in the case of elevated temperature tanks. The

fill/draw down cycle of the stored liquid causes low cycle fatigue near this joint and

hence a fatigue evaluation is recommended as explained in previous Chapters. The peak

alternating stress at this location, used to enter the fatigue curves is currently determined

using a pseudo elastic stress that represents strain range due to inelastic deformations. For

this, API 650 employs beam-on-elastic foundation theory. This theory is being used for

tanks resting fully on earthen foundation as well as those on concrete ring wall. This

Chapter studies the validity of using this theory for tanks with concrete ring wall

foundation which are much more rigid compared to earthen foundations. Some of the

difficulties in the current practice are highlighted. An alternative to the current model is

proposed to determine the peak stress at the shell to bottom joint of tanks resting on

concrete ring wall. The results are validated using finite element analysis.

155
6.0 INTRODUCTION

The shell (tank wall) to bottom joint of storage tank is subjected to low cycle

fatigue due to fill/draw down cycle and the temperature fluctuation of the stored liquid.

The storage tank standards API 650 and 653 consider this aspect in detail and provide

guidelines for determining the safe cyclic life of the tank based on the fatigue analysis of

this joint.

For the elevated temperature tanks, API 650 provides guidelines to determine the

appropriate thicknesses of structural members and determine the design cycle life of the

tank (based on the fatigue evaluation of the shell to bottom joint). As per API 650, it is

expected that the tank withstands at least 1300 load cycles in its lifetime. In Appendix M

of API 650, the peak alternating stress to be used for fatigue analysis of this joint is

determined using the original work by Karcher [1978, 1981 a, b]. He proposed a set of

design equations to determine the tank wall stress and the fatigue life of elevated

temperature tanks as given in Chapter 5. Jones and Seshadri [1989] studied the validity of

Karcher’s model using elastic finite element analysis of the shell with an assumed hinge

condition at the bottom.

In the current practice, the bottom plate is analyzed using beam-on-elastic

foundation theory. The resulting equations were derived basically for tanks on earthen

foundation by Karcher [1981a]. The same are applied to concrete ring wall foundations

except that two plastic hinges are assumed in the bottom plate (or annular plate) instead

of one plastic hinge in the case of earthen foundation. However, it must be noted that the

156
bottom plate resting on concrete wall loses contact in the immediate vicinity of the inside

face of the tank wall. This was recognized by the previous research works [Denham,

1968; Wu, 2000; Sathyanarayanan and Adluri, 2011] as well as the seminal work by Zick

and McGrath [1968]. The uplift, although small, is sufficient to cause clear separation of

bottom plate from the top of the concrete ring wall on the inside. Hence the stresses in

this region are not directly governed by beam-on-elastic foundation model.

6.1 FATIGUE AT SHELL TO BOTTOM JOINT

Figure 6.1 shows the rotation of shell to bottom joint of a tank resting on a rigid

base. As explained in Chapter 4, this rotation causes the bottom plate to lift off the

ground over a small distance (say, uplift length L). Beyond this, the downward liquid

pressure acting on the plate is sufficient to make the plate fully rest on the foundation.

s

Mo

Uplift Length

Fig. 6.1 Shell to Bottom Joint in the Tank on a Rigid Base

157
 The internal moment that forces the rotation is caused by the hydrostatic

pressure on the wall and is balanced by the moments (on either side of the wall) in the

bottom plate. Because of this bottom moment, the bottom plate on the inner and outer

side of the shell is subjected to high bending stresses in the radial direction. The thickness

ratios of the shell and bottom plates are such that in most of the tanks the bottom plates

undergo substantial yielding near this joint. The fatigue strength of this joint has to be

evaluated in order to determine the number of load cycles the tank can safely withstand.

6.2 DESIGN FOR FATIGUE

For short projection lengths (such as the minimum prescribed value of 50mm), the

bending moment in the bottom plate on the inside will be larger than that on the outside.

The inside portion of the shell to bottom joint needs to be evaluated for fatigue. Since the

material yields due to high stress in the bottom plate, strain range analysis (pseudo stress)

is used for fatigue evaluation. The rotation of the shell as shown in Fig.6.1 induces the

bending strain. The magnitude depends upon the tank parameters R, H, ts, tb and the

specific gravity of the stored liquid (G). In addition to these factors, the type of

foundation directly beneath the tank shell and the type of weld used (double fillet/full

penetration) will influence the fatigue evaluation.

The design cycle evaluation procedure of the elevated temperature tank as

described by the API 650 assumes uniform temperature increase in the tank, i.e., without

any temperature gradient across the thickness of the shell wall. This is a reasonable

assumption since the shell wall thickness is quite small. At steady state, the difference in

158
temperature between inner and outer surfaces of the wall may not be significant. The

same is adopted in this study. The current procedure [API 650] is outlined briefly below:

Using the Eq.5.6 from Chapter 5 and substituting the following coefficients [Karcher

1981a] used by API 650, the peak alternating stress (S) can be obtained as:

γ = G x 9.81 x10-6 N/mm3; E = 190986 MPa; K = 0.2715 N/mm2/mm;  = 11.7 x 10-6/C;

1.285
 ;
500 Dts

Sb 0.028 D2tb0.25  58HG 26.2C Tts0.5 4.8BSY tb2 

S      G 
  Dts   Dts 
0.5 1.5
2 ts D1.5  (6.1)

where, S is one-half of the maximum stress range (Sb) that occurs in the annular plate at

the shell to bottom junction, in MPa. The dimensions of D & H are in metres whereas the

dimensions of ts and tb are in mm.

Similarly using the above coefficients in the Eq.5.7, the following (Eq. 6.2)

inequality condition can be obtained. Equation 6.1 is valid only if this condition is true.

 58HG 26.2C Tts0.5   4.8BSY tb2 

   G  1.5 
  Dts     Dts  
0.5
D1.5 (6.2)

As stated earlier, this condition is prescribed to ensure that the tank is loaded such that

boundary conditions (plastic hinge) assumed for developing these equations are valid.

159
 The stress created at the shell to bottom joint can be classified as secondary

bending stress (as per ASME section VIII DIV 2) and hence should be limited to twice

the yield strength of the annular plate material to assure shakedown to an elastic action.

The shell theory used in this method basically assumes that the influence of the radial

expansion of bottom plate due to the hydrostatic load is negligible.

6.3 ISSUES IN EXISTING PROCEDURE

6.3.1 Necessity of the Condition in Eq.6.2

Karcher [1978,b] has mentioned that the current practice is applicable only when

Eq.6.2 is valid. This can be interpreted that Eq.6.2 is specified as a necessary condition

for the assumed plastic hinges to form. Satisfying Eq.6.2 ensures that the stress

determined from Eq.6.1 is not negative. Essentially it determines whether the slope of the

shell at the bottom, after the application of load, is inward or outward. Although API 650

does not say so, satisfying Eq.6.2 is not a sufficient condition to confirm the formation of

two complete hinges at the bottom [Sathyanarayanan and Adluri, 2012a, 2013a]

(a) (b) (c)

+S -S

M
o

Fig. 6.2 Free Body Diagram of Shell

160
 The sign of stress S is determined by the sign of the term

 58HG 26.2C Tts0.5 4.8BSY tb2 

    G  in Eq. 6.1. Figure 6.2 (a) represents free body
  Dts   Dts 
0.5 1.5
D1.5 

diagram of the tank wall. Fig. 6.2 (b) represents positive value for the stress S from

Eq.6.1 while Fig. 6.2 (c) represents negative value for the stress. The condition from

Eq.6.2 can be rearranged in non-dimensional terms as

   
 
 H   SY   E   CT   3
1   H   SY   tb 
2
 H   H   1   (6.3)
     0
 R    H   SY   H  2  H    H   ts 
 t   t 
 s   s 

H  Sd   H 
H H

where,  H      3 1 2 
0.25
and   
 R   ts  ts   H   R 

The validity of this condition for various situations can be easily examined. The

following ranges of parameters are considered:

t   E  H 
0.3   b   0.8, 550     800, 0.3     6,
 ts   SY   R 
S
1300  Y  7300, 0.25  C 1,
H

 12106 / C , 93 C  T  260 C

A simple calculation shows that the condition is satisfied for practical ranges of

variables. The only exception seems to be tanks where the design thickness of tank wall

comes to be less than what the minimum thickness rule prescribes. For all other tanks,

this rule need not be checked.

161
6.3.2 Use of Beam-on-elastic Foundation Theory

The existing procedure assumes that beam-on-elastic foundation theory is valid

for the immediate vicinity of tank wall bottom. In the current practice, the same

procedure is used for tanks on earthen foundation and concrete ring wall foundation

except that in the former case, a single plastic hinge is used and in later case two plastic

hinges are used as the shell bottom moment. It is clearly applicable to tanks on earthen

foundation with appropriate conditions. However, it is not suitable for concrete ring wall

foundation as explained below:

If the shell to bottom joint is supported on a rigid concrete ring wall, the internal

moment (Mo) lifts the bottom plate slightly thereby loosing contact with the foundation

(Fig. 6.3). This uplift is present even after accounting for the effect of liquid (hydrostatic)

head, self-weight of plate and compression from the wall (including roofing weight, wind

girder, etc.).

Compacted
Sand/Gravel
s Concrete Ring
Wall

Uplift Length

Fig. 6.3 Uplift at Shell to Bottom Joint of a Tank on Concrete Ring Wall

162
 The uplift length has been studied variously by Denham,et al [1968],

Sathyanarayanan and Adluri [2011], among others [Long and Garner, 2004]. Typical

uplift lengths range from 150 – 400mm. Annular Plates (Cl.5.5.2 of API 650) has lengths

typically larger than this. In addition to theory, a detailed finite element analysis of a tank

(as described in Chapter3 – Case1 with 8 mm bottom plate), along with roof load has

been carried out. Fig. 6.4 shows the uplift and von Mises stress profiles at the shell to

bottom joint from the nonlinear FEA. The figure is captured at the end of a time step

analysis performed with factored loads to see formation of plastic hinges.

Tank Wall Outside

Projection
Region

Uplift Region

Fig. 6.4 Uplift of Bottom Plate from FE Model with 2D Axisymmetric Elements

163
 The region (Point F in Fig.6.3) in which the peak alternating stress is determined

for fatigue calculation is not in direct contact with the earthen subgrade. Hence the beam-

on-elastic foundation theory as used in the present method is not directly applicable. The

direction of the internal moment Mo is such that it uplifts the bottom plate at point ‘F’

since the moment cannot push the projection part in to the concrete ring wall. Because of

this, the deflection of the plate is not significantly influenced by the foundation modulus.

Only if there is a significant uneven settlement, the foundation modulus might have some

influence on tanks resting on concrete ring wall. This is not the objective of the present

investigation. Since the uplift zone is supported only at the ends, a regular beam model is

more appropriate than beam-on-elastic foundation model.

6.4 PROPOSED CHANGES IN THE DETERMINATION OF PEAK

ALTERNATING STRESS

The safe cyclic life (number of safe load cycles) is determined from the strain

range of the bottom plate at point ‘F’. The stress (Sb) in the existing procedure is only a

pseudo stress used to quantify the strain range. The rotation of the shell to bottom joint

during the cyclic loading determines the strain range at ‘F’, which in turn controls the

fatigue life of the joint. The calculation of this rotation depends on the loading, geometry

of the tank and the double plastic hinge assumption used in the model. It should be noted

that, for a given geometry and loading, a single plastic hinge in bottom plate will cause

more rotation and hence shorter fatigue life estimation than the assumption of double

plastic hinge. However, the current practice is to use two plastic hinges (one on the inside

and another on the outside) for the estimation of peak alternating stress. This is retained

164
in the current study and follows the recommendation by Karcher [1981a, b] and Zick and

McGrath [1968]. The bottom plate near the shell can be idealized as a beam shown in

Fig. 6.5. Using beam theory and shell theory, the uplift length (L) and the moment (Mbp)

in the bottom plate near the shell can be determined as follows:

qL2
Mbp  (6.4)
4

24 EIPs
L 3 (6.5)
q

The basic idea of the procedure is that the strain in the bottom plate is governed

by the rotation of the shell bottom, and the maximum rotation of shell bottom occurs at

the limiting condition of a double plastic hinge of the bottom plate. Following on the

same lines as the current practice except for the equations above, the peak alternating

pseudo-elastic stress is given by

2/3
3 2 
2/3
 R2 
  Ets  3SY tb2 
Sb     H    H  CT  1 (6.6)
2  1 2   Hts 
  R  


This equation generally gives higher stress than that predicted by Eq.6.1 because

this equation does not have the benefit offered by the elastic subgrade.

The safe design cycle life, N, is then given by [API 650],

165
 2.44
 9700 
N  
 Kc S  (6.7)

where, S is the alternating stress range (Sb/2), Kc is the stress concentration factor which

can be conservatively taken as 4.0 for lap welded bottom plates/ butt-welded annular

plate (examined as per API 650 Specification). In case of butt welded annular plate with

100% surface examination and blend grinding, Kc = 2.0 [Karcher, 1981b].

Plastic Hinge with

Moment of 2Mp

s

Mbp
Uplift Length
T (shell self wt)
(a)
UDL water pres. q/unit length
Mbp

L
Rb Ra
(b)

Fig. 6.5 Idealized Beam Model

166
6.5 RESULTS AND DISCUSSION

As shown in Fig. 6.3, finite element analyzes were carried out assuming the

foundation to be a concrete ring wall with highly compacted infill. In addition,

axisymmetric SHELL elements were also used for modeling the tank wall and bottom

plate. Both elastic and elastic-perfectly-plastic material models were examined. The

advantage of using plane elements is that, the elasto-plastic stress/strain profiles and the

formation of plastic hinges can be explicitly viewed, whereas in shell models which are

line elements, the through the thickness stress profile cannot be seen. Unless otherwise

mentioned, the example tank – Case 1 as given in Chapter 3 with 6mm bottom plate and a

friction coefficient of 0.8 is used for all the finite element analysis.

For the same tank operating at an elevated temperature of 175C (T), Fig.6.6

shows the uplift deformation of the bottom plate and Fig.6.7 shows deformation pattern

at the shell to bottom joint as shown by FE shell model. It is the expanded view of

deformation profile at the shell to bottom joint.

167
 Tank
Wall
Location

Fig. 6.6 Uplift Deformation of Bottom Plate

Figure 6.8 shows the radial direction stress (Sx) of the bottom plate at the joint. It

must be noted that stress (Sx) can exceed yield as long as the equivalent von Mises stress

is within the yield limit. The bending stress (Sx in the radial direction) and von Mises

stress in the bottom plate near the shell are plotted in Fig.6.9 for two complete loading

cycles. As can be seen, even though the yield strength of the material is 345 MPa, Sx

reached 400 MPa indicating the effect of multi-axial stress field. The combined effect of

all the stresses at that point represented by von Mises equivalent stress is within the yield

limit. The tank is gradually loaded to full height, unloaded gradually and the cycle is

repeated one more time.

168
 Tank Inside
Shell to Bottom Joint
Tank Wall
Bottom Plate

Tank
Outside

Tank centre
Bottom Plate
Projection

Fig. 6.7 Uplift Deformation from FE Model with Shell Elements

Fig. 6.8 Stress Sx in Bottom Plate near Shell Joint

169
 Each loading or unloading is considered to be one load cycle stage. Therefore,

stage 4 ends when the tank is fully loaded and unloaded twice. As can be seen, the first

stage of full loading ends with the bottom plate (on the inside) yielding. At the end of

stage 2, the tank is unloaded where the bottom plate shows residual stresses. The next

cycle (shown by stages 3 & 4) is fully elastic indicating shake down behavior. The radial

direction stress SX varies from -219 to 398 MPa and hence the range from FEA is 617

MPa, whereas the predicted stress range using API (Eq.6.1) is 419 MPa. For the same

tank Eq.6.6 gives a stress range of 612 MPa. The partial radial expansion from FEA is 47

mm (against the free expansion of 63 mm with =12x106) and hence the C factor is

computed as 0.25. The von Mises stress values show the yield limit of 345 MPa at load

stage 1 and the formation of residual stress at stage 2. Fig.6.10 shows the stress Sx in the

bottom plate near the outer face of shell to bottom joint.

Similarly for the same tank, but with 8mm bottom plate the bending stress (Sx)

and von Mises Stress (on the inside) are plotted in Fig.6.11. The stress Sx varies from -75

to 398 MPa. Hence the range of Sx from FEA is 473 MPa. The stress range predicted

using current API (Eq.6.1) is 305 MPa. The stress range obtained by the proposed

equation (Eq.6.6) is 470 MPa

170
 Yielding
Stage

Fig. 6.9 Bending and von Mises Stress in 6mm Thick Bottom Plate (On the Inside)

The partial radial expansion of bottom plate due to temperature from the FEA

results is 51 mm against the free expansion of 63 mm (with =12x106) and hence the C

factor can be computed as 0.19 which is used with both API and beam equations. The

idea that the strain and hence the stress range of the bottom plate is controlled by the shell

rotation is appropriate for concrete ring wall foundations as well as earthen foundations.

However, since the bottom plate uplifts near the joint for the concrete ring wall, the

beam-on-elastic foundation theory could be under predicting the stresses. As can be seen

from the results, the beam model used in the current research predicts stress results quite

close to FEA results.

171
 Fig. 6.10 Bending Stress in 6mm Bottom Plate (on the Outside)

It is to be noted that the stress range computed in FEA is indicative of the elastic

strain range during all the shakedown load cycles after the first cycle. These results

indicate that the model used in the present work is more appropriate for concrete ring

walls than that of the current API 650. However, the proposed model predicts higher

stresses than those given by API 650 and hence a smaller fatigue life cycle estimate. It

must be pointed out that, the API procedure, originally developed by Karcher has been in

practice for about 30 years. To the best of our knowledge, no major failures due to fatigue

have been reported in public domain. This is likely because no failures actually occurred

or because failures were not recognised to be due to this issue or due to the embedded

margin of safety.

172
 Fig. 6.11 Bending and von Mises Stress in 8 mm Thick Bottom Plate

The current procedure although not reported (publicly) to be problematic in the

field is not very rational in comparison to the model presented here. On the other hand,

the model presented here seems to reduce the life cycle estimates from current levels. In

view of this, the current procedures as well as the model developed in this thesis must be

looked at thoroughly by relevant API committees to arrive at a rational, safe and

economical procedure.

6.6 INFLUENCE OF PLASTIC HINGES

Apart from the influence of the foundation, the other important factor that

contributes to the fatigue design is the assumption of single/double hinge boundary

173
condition at the shell bottom. Figure 6.12 shows the locations where high bending

moments are induced due to the liquid infill.

(a) In Concrete Ring Wall

FOUNDATION

(b) In Earthen Foundation

Fig. 6.12 Bending of Bottom Plate in Shell to Bottom Joint [Zick and McGrath, 1968]

Zick and McGrath [1968] suggested that “where the resisting moment of the tank

bottom is to be evaluated, a reasonable approach would be to use the full yielding

moment in the bottom plate on one side (inner side) for an earth foundation and to use

two moments (one on each side) for concrete ring wall foundation.” The influence of the

174
assumed bottom moment on tank wall bending stress is shown in Fig.6.13 for the

example tank from earlier. For a concrete ring wall foundation, two plastic hinges

assumption is used. But in actual field conditions, two complete hinges may or may not

form at the bottom because the loading and geometry do not necessarily allow it to

happen. The smaller the bottom moment, greater is the rotation of the shell due to

hydrostatic load. All other factors being same, the maximum bending stress is less if the

bottom moment is more and vice – versa. Hence, if an assumed bottom moment is used,

it is more conservative to assume smaller than the “exact” moment as opposed to a larger

than “exact” moment. This issue also needs to be looked at more thoroughly by relevant

API committees.

Fig. 6.13 Influence of Bottom Moment on Tank Wall Bending Stresses

175
6.7 CONCLUSION

The Chapter traces the development of the current API 650 procedure for life

cycle evaluation of elevated temperature steel storage tanks subjected to low cycle

fatigue. The current procedure uses beam-on-elastic foundation model for tanks resting

on earthen foundations as well as those using concrete ring walls. In the vicinity of the

ring wall, the inside portion of the bottom/annular plate looses contact with the

foundation due to the outward rotation of the tank wall at the bottom. Hence, the theory

of beam-on-elastic foundation is not applicable to tanks with concrete ring wall supports

below the wall plate joint. Modifications have been carried out in this study to arrive at

an alternative estimate for pseudo elastic peak alternating stress (strain range) for fatigue

calculations. Detailed finite element analysis using non linear models and friction-contact

elements were carried out. The results show that the alternative model is quite close to

FEA results where as the current procedures under predict the stress. The current

procedure is clearly less rational than that developed in this thesis. However, the current

procedure is long standing without major problems. In light of this, it is recommended

that API committees take a close look at this issue in order to arrive at a rational and yet

acceptably safe decision on the matter. Related issues regarding the inequality condition

and the assumption of two full plastic hinges at the bottom are also discussed.

176
CHAPTER 7

LOCALLY THINNED AREAS

Various types of damages can occur to tanks due to environmental and in-service

factors. The issues dealt in the previous chapters, like the effect of bottom plate friction

and low cycle fatigue at shell to bottom joint were structural integrity considerations

during operation of the tank. Similarly degradation due to wall thinning is an important

and most prevalent structural integrity issue that occurs due to either corrosion or erosion

or both together.

Local thin area (LTA) is type of wall thinning degradation wherein the metal loss

is constrained to patches of local zones in the tank wall. LTAs unattended can cause

local failure by leakage or rupture if the remaining wall thickness is not sufficient to

withstand the liquid pressure [Antaki, 2005]. Also, depending on the geometry and

location, it can also lead to buckling and related issues. As mentioned in Chapter 2,

replacing a slightly damaged tank is not a cost effective option compared to safely

operating the tank and extending its life. This Chapter deals with Fitness for Service

(FFS) assessments for LTAs in tanks. A level 2 procedure is proposed for the purpose of

estimating remaining strength factor (RSF). The use of m-tangent multiplier method

with modified reference volume assessment procedures for determining the RSF is

explored.

177
7.0 CURRENT PRACTICE

Structural integrity assessment is of considerable importance in many industrial

sectors. Fitness for service assessments are periodically performed for in service tanks to

ensure their structural integrity and safety. A number of FFS procedures are available in

practice, e.g., API 579-1/ASME FFS-1(2012), R5 (2003) and R6 (2001), Structural

INTegrity Assessment Procedures for European Industry (SINTAP, 1999), etc. Among

these, the API 579-1/ASME FFS-1 procedure is widely used in the industry. As

explained in Chapter 2, the fitness for service assessments are quantified using remaining

strength factor (Clause 2.4.2.2 of API 579-1/ASME FFS-1),

Limit or Plastic Collapse Load of Damaged Component

RSF =
Limit or Plastic Collapse Load of Undamaged Component

The following is a brief description of the Level 1 procedure given by API to

determine the RSF of pressure vessels/tanks/pipes with Locally Thinned Areas (LTAs). It

forms the basis for all other advanced procedures used for complex loading or damage

conditions. Based on the Critical Thickness Profile (CTP), the remaining thickness ratio

(Rt) and the shell parameter () are computed as

tmm  FCA
Rt 
tmin (7.1)

1.285s

Dtmin (7.2)

where, tmm is minimum measured remaining wall thickness, FCA is future corrosion

allowance, and tmin is minimum required wall thickness determined in accordance with

178
the original construction code. Similarly ‘s’ is the length of the flaw in the axial direction

of the tank and ‘D’ is the diameter of the tank. Then the RSF and stress at failure of LTA

are computed as

 y RSF
 fail  (7.3)
0.9

Rt
RSF 
1  (1  Rt ) M 1 (7.4)

where M, is the Folias factor equal to 1 0.48 2 .

The remaining strength factor (RSF) given above is based on a Dugdale

plastic-zone-size model, “Folias” factor and an empirically established flaw depth to wall

thickness relationship [Janelle, 2005]. Folias factor is a stress magnification factor due to

bulging caused by an axial crack in a pressurised cylinder [Folias, 1969]. An empirical

flaw-depth-to-pipe-thickness relationship is used to modify the Folias factor to account

for part-through wall effects based on “effective” cross sectional area. This method

assumes that the flaw fails when the stress in the flaw reaches the flow stress σflow. The

flow stress is assumed to be equal to y/0.9 (= 1.111y) similar to that of ASME B31G.

The procedure also assumes that the region of metal loss is not influenced by any

structural discontinuity, welded joints, change in geometry, etc. To ensure this, API 579-

1/ASME FFS-1 specifies that Lmsd 1.8 Dtmin where, Lmsd is the shortest distance

between the edge of corrosion area and the discontinuity. Additionally the remaining

thickness should be such that, tmm - FCA  2.5mm and Rt  0.20. The former condition

179
is to ensure safety against leakage and mechanical damage while the latter can possibly

be due to the absence of validation results for very deep flaws greater than 80% of design

thickness.

The Folias factor or bulging factor is basically a factor that relates stress near the

cracked portion of a flat plate to that of cylindrical plate. The factor is derived from the

following governing differential equations where the pressure is assumed to be uniform.

Et 2 (7.5)
2 F  0
R x2

1 2 F qo
2   (7.6)
RD x2 Ds

where 4 is the biharmonic operator, (x,y) is the displacement function, F(x,y) is the

stress function, E the young’s modulus, t the thickness of the vessel, Ds the flexural

rigidity, R the radius and qo the internal pressure.

In the equations above, it can be noted that the pressure qo is assumed to be a

constant (uniform internal pressure). This is obviously the case for most pipes and

pressure vessels. However in the current practice, the same Folias factor is used for

determining the RSF of tanks with hydrostatic pressure as well. The product of the RSF

and the existing pressure or wall thickness will give the rerated pressure or wall

thickness. In case of tanks, since the pressure is varying along the height (axial

dimension), the minimum acceptable thickness and the Maximum Fill Height (MFL) are

180
calculated at one foot above the bottom end of LTA (Clause A.6.1 of API 579-1/ASME

FFS-1).

Tank Wall

LTA

Critical Point

1 Foot or Variable
Design Point

Fig. 7.1 Critical Point in LTA for a Tank

The concept of considering the pressure one foot above the junction of two

different thicknesses in the tank wall (see Fig.7.1), also known as ‘one foot method’, is a

long standing practice used in design of tanks. The one foot is an approximate location

where the tank stresses are assumed to be maximum in any particular shell course of the

tank wall or LTA in consideration. In lieu of the one foot rule, a more precise location of

maximum stress point can be calculated using the variable design point method given by

Clause 5.6.4 of API 650 [2012]. If the height of the flaw is small, the pressure can be

simply considered at the bottom of the flaw (Clause 4.3.3 of API 653), which will give a

181
conservative estimate. A detailed procedure for calculating the minimum thickness and

the actual length of flaw (to be used for determining RSF) are discussed in API 653. To

summarize the current procedure, RSF for tank is determined in exactly the same manner

as that for cylindrical vessel with uniform pressure. Although the tank has hydrostatic

pressure, no special discussion seems to have taken place about applying the uniform

pressure procedure for tanks except that while rerating the tank, the pressure to be used is

taken at one foot above the bottom.

7.1 CONCEPT OF DECAY LENGTH AND REFERENCE VOLUME

The concept of “Decay Length” is widely used in pressure vessel theory. It is a

known fact that the effect of a localized force (such as that due to discontinuities)

dissipates as the distance from the point of application of the force increases. The decay

length of cylinders is the distance required for the effect of a force/moment applied on the

shell to dissipate (decay) to a negligible amount. It is a characteristic length based on the

geometry of the shell and the nature of force and is independent of the magnitude of the

force [Tantichattanont, 2009]. Generally larger decay lengths have larger volume for

energy dissipation and hence indicate higher load carrying capacity. For cylindrical shells

the decay length in the axial direction (xl) is shown by many sources [ex. Gill, S.S., 1970]

as,

xl  2.5 Rt (7.7)

where, ‘t’ is the nominal thickness of the cylinder

182
 Tantichattanont, et al., [2009] have discussed this in detail and showed the

circumferential decay length of cylindrical shell to be

xc  6.3 Rt (7.8)

The above expressions are obtained using shell theory applied to axisymmetric

cylindrical shell subjected to line loads along circumferential or longitudinal directions,

respectively. They are generally applicable for thin shells with R/t ratios in the range of

20 to 100.

When the tank fails by plastic collapse due to an LTA or any other localized

damage in the tank wall, it is well known that only a portion of the tank wall participates

in the plastic action and the remaining regions (Dead Volume) that do not participate in

the plastic action may remain unaffected. The “kinematically active” portion of the

component or structure that participates in plastic action due to the presence of the

damage is termed as “Reference Volume” by Seshadri and coworkers [Adibi-Asl, et al.,

2011]. The total volume of the component is the sum of this reference volume and the

“dead volume” (Fig.7.2). The accuracy of the limit loads predicted by m-tangent method

depends primarily upon the accuracy of the reference volume assumed.

183
 Fig. 7.2 Reference Volume and ‘Dead Volume’

7.2 CIRCUMFERENTIAL DECAY LENGTH FOR TANKS

As mentioned earlier, using the expression ( 2.5 Rt ) for decay length in axial

direction of cylinders is a well-established practice. It is shown theoretically and verified

numerically [Tantichattanont, et al., 2009] that the decay length in circumferential

direction is larger than that in the axial direction at 6.3 Rt . The reason for this seems to

be the change in curvature from axial to circumferential direction. However in case of

tanks, because of very large radius compared to pipes and pressure vessels, the curvature

in circumferential direction is substantially smaller (R/t for tanks can be substantially

larger than 1000 as opposed to R/t of around 50 to 100 for typical pressure vessels).

Hence, the circumferential decay lengths could be smaller than 6.3 Rt . In order to

184
study this, the example tank as specified in Chapter 3 (Sec.3.5 Case1) is analyzed using

FEA for varying sizes of LTA.

The finite element modeling was carried out using eight noded shell elements

(SHELL 281 of ANSYS). The element has six degrees of freedom at each node and is

suitable for analysing thin to moderately-thick shell structures. An elastic-perfectly

plastic material model is used and the nonlinear analysis is carried out as per the

guidelines given in Clause B1.2.4.2 of API 579-1/ASME FFS-1 [2012] standard. The

plastic collapse load is taken as the load corresponding to 1% plastic strain at the middle

fiber of any location in the LTA. This approach is consistent with the work reported by

earlier researchers on FFS assessment [Seshadri, et al., 2004, Tantichattamont and

Adluri, 2008]. The analysis is carried out using up to 7 integration points across each

layer. Multiple layers are used across the thickness. For example, an 8mm bottom plate

has 8 layers (with 3 integration points across each layer) and an 18mm shell wall has 3

layers (with 7 integration points across each layer). Since the yielding occurs in the

bottom plate and it is the zone of interest in this research work, the bottom plate is

modeled with more layers than the shell plate. This approach is adopted to optimize the

computational effort and data handling in FEA. A macro is developed in ANSYS to

incorporate decay lengths in the post processing analyses and extract stress values from

different layers of the shell element if needed.

The following stress plots (Fig.7.3 & 7.4) obtained from the analysis show the

hoop stress values at the centre of LTA as the ordinate and the height or width of LTA as

185
the abscissa. The notation LTAW 100% represents LTA width equal to 6.3 Rt and

LTAH 100% represents height equal to 2.5 Rt. Various lines in the plots refer to hoop

stress at the centre of LTAs of different sizes. For example, in Fig.7.3, the series LTAW

60% shows the hoop stress for different heights of LTA with the LTA width being

constant at 60% of 6.3 Rt . From this figure, it can be observed that the hoop stress

increases as height of LTA increases, but the width of LTA doesn’t seem to influence the

hoop stress beyond 40% of 6.3 Rt , which is approximately 2.5 Rt . All the series

from 40% -100% LTAW show the same pattern.

Fig. 7.3 Hoop Stress Variation with Height of LTA

186
 Fig. 7.4 Hoop Stress Variation with Width of LTA

Similar observation can be made from Fig.7.4., i.e., no significant change in the

hoop stress when the width is larger than 40% of 6.3 Rt . Hence, in the present study,

for the RSF calculations in tanks, a decay length of 2.5 Rt (≈ 40% of 6.3 Rt ) is used

in the circumferential direction as well as the axial direction. This leads to the

recommendation that 2.5 Rt be used instead of 6.3 Rt as circumferential decay

length for tanks and other vessels that have very large R/t ratio. More shells can be

analyzed to establish the limits of this recommendation for various R/t ratios.

187
7.3 DETERMINATION OF RSF USING m-TANGENT METHOD

As mentioned earlier, the remaining strength factor (RSF) is used as a basis for

the evaluation of locally thinned areas in storage tanks. Detailed discussion about RSF

and the methods for determining it are already presented in Chapter 2. For tankage, the

RSF acceptance criterion is [API 579-1/ASME FFS-1, 2012]:

MFHr=MFH (RSF/RSFa) for RSF< RSFa (7.9)

MFHr=MFH for RSF RSFa (7.10)

where,

MFHr = Reduced permissible maximum fill height of the damaged tank

MFH = Maximum fill height of the undamaged tank

RSF = Remaining Strength Factor of the damaged tank

RSFa = Allowable Remaining Strength Factor

If the calculated RSF is higher than the allowable RSFa, the tank is safe for full

operation. The recommended value for the allowable Remaining Strength Factor is 0.90

for tanks in process services [API 579-1/ASME FFS-1, 2012]. This value can be reduced

based upon the type of loading (e.g., normal operating loads, occasional loads or short-

time upset conditions) and/or the consequence of failure.

The RSF of a structure/component can be determined in different ways as already

mentioned in Chapter 2. The material/geometry/loading conditions of a given problem

can be too complicated to have a closed form solution to determine the RSF. Similarly,

performing ‘nonlinear’ FEA in those situations can also become impractical. For these

188
cases, performing a FFS assessment using linear FEA and simplified limit load

procedures is very useful. Using the simplified methods, the RSF can be estimated to a

reasonable accuracy in a shorter time and with much lesser effort.

The m-tangent multiplier method is a robust simplified limit load procedure, that

can be used for FFS assessments utilizing linear analyzes instead of full scale nonlinear

FEA. The linear analyzes can be done either using shell theory, FEA or any other linear

methods. The only requirement is that the method needs a statically admissible stress

field for the tank being analyzed. Hossain and Seshadri [2009] have used this method for

fitness for service assessments of pipes/pressure vessels with corrosion damages and

thermal hotspots. Recently Ahmad, et al., [2010] have made a preliminary study to use

this procedure for hydrocarbon tanks.

7.3.1 RSF using Analytical Approach

Seshadri, et al., [2005] have shown the use of Mura’s “integral mean of yield

criterion” for determining limit loads of components and hence the RSF of damaged

components.

For incipient plastic flow condition, the integral mean of yield criterion is expressed as

 f ( s 0 )   0 2  dV  0

0
(7.11)
VR
 ij 

where, sij0 is the statically admissible deviatoric stress for impending plastic flow, µ0 is

flow parameter,  0 is a point function which takes in a value of zero if sij0 is at yield and

189
remains positive below yield, and VR is the reference volume. The yield criterion can be

expressed as

f ( sij0 )   mdo e    y2   0

2

  (7.12)

where, mdo is the appropriate statically admissible upper bound multiplier for the

damaged component,  e is the equivalent stress and  y is the yield stress of the material.

For components containing LTAs, the integral mean of yield criterion can be expressed

as [Seshadri, 2005]

[(md0 eU ) 2   y2 ] VU  [(md0 eD ) 2   y2 ] VD  0
(7.13)

where, suffix U(or u) refers to the uncorroded region of the reference volume and suffix

D(or d) refers to the corroded (damaged) region, σeU is the equivalent stress in the

original shell and σeD is the equivalent stress in the corroded area of the shell.

Rearranging Eq. 7.13,

 y2VR
m 
0
(7.14)
 eU VU   eD
d 2 2
VD

Based on the above multiplier, the following three approaches are used to determine the

RSF of damaged tanks.

190
7.3.1.1 RSF Based on Upper Bound Multiplier

The remaining strength factor RSFU is based on the integral mean of yield

criterion, along with the von Mises failure criterion. The upper bound RSFU is obtained

using md0 as

md0
RSFU  (7.15)
mu0

where, mu0   y /  eU is the upper bound multiplier for the undamaged tank, and md0

(Eq. 7.14) is obtained from the integral mean of yield criterion. Since RSFU is the ratio of

an upper bound multiplier of a damaged tank to that of an undamaged tank, the RSF will

be an upper bound estimate.

7.3.1.2 RSF Based on the mα-Tangent Multiplier

The second RSF is obtained by using the m-tangent multiplier ( mT ), proposed by

Seshadri and Hossain [2009]. The m-tangent multiplier based on Eq. 2.28, can be used

to calculate the RSF as

mTα of damaged component

RSFT = (7.16)
mTα of undamaged component

while using Eq. 2.28 in order to evaluate RSF T , the classical lower bound multiplier can

be obtained as mL  mLd (  y  eD ) and the upper bound multiplier m 0 as equal to m d0

as defined in Eq. 7.14.

191
 For mT calculations, the stress distribution is obtained within the reference

volume of a damaged tank either using shell theory or from a linear elastic analysis as

mentioned earlier. The m-tangent multiplier is determined (Eq.2.28) for the reference

volume with and without damage, i.e., for the undamaged case the same reference

volume as that of the damaged case is used but with the undamaged thickness. The value

of mL is calculated from the maximum stress value in the reference volume and m0 is

determined using a macro incorporating Eq.7.14 in the postprocessor of ANSYS. Using

these multipliers, the mT value and RSF can be determined. The same procedure is

repeated for different sizes of LTA.

7.3.1.3 RSF Based on Classical Lower Bound Multiplier

The third remaining strength factor RSFL is based on the classical lower bound

limit load multiplier m L and is given by,

m Ld
RSFL  (7.17)
mu0

where, the classical lower bound multiplier mL  mLd (  y  eD ) for corrosion damage.

This RSF gives very conservative estimation of RSF.

7.3.2 RSF Based on Non Linear Finite Element Analysis (NLFEA)

The RSF can be determined more accurately using Nonlinear FEA. In this thesis,

a Level 3 FFS evaluation is performed using full scale elastic-plastic FEA (EPFEA) for

comparison purposes. The RSF is calculated as the ratio of the limit pressure (pressure

corresponding to 1% membrane plastic strain in the LTA) of damaged tank to the

192
corresponding limit pressure of the tank without loss of thickness. The limit pressure of

the tank without LTA is calculated at the same damage location as used in the tank with

LTA. Although the location of maximum plastic strain in the LTA is slightly below the

centre of LTA, it can be approximated to be at the centre for most cases

7.4 REFERENCE VOLUME FOR LTA IN TANKS

The reference volume is computed using the shell decay length and damage

geometry. At present, the m-tangent method considers the reference volume as the full

volume of LTA plus a portion of the undamaged tank bound by the appropriate decay

lengths on all sides of LTA [Seshadri, et al., 2009; Hossain, et al., 2010; Ahmad, et al.,

2010]. Figures 7.5 & 7.6 show the damaged zone and the corresponding reference

volume used by Ahmad, et al., [2010]. The actual LTA profile is generally irregular both

in shape and thickness. However, the existing level 2 practice [API 579, 2012] is to use

representative uniform rectangles with averaged thicknesses across several strips to

evaluate the irregular flaw. The treatment in the present study is a demonstration for one

such rectangular stip. This can be extended in a manner similar to that in the current API

practice. This is in line with earlier work by Tantichattnont, et al. [2008] and others.

For a damaged area of width ‘2a’ in circumferential direction and length ‘2b’ in

the axial/longitudinal direction (Fig.7.5), the volume of damaged zone (VD) is given by

VD = 4abtmm (7.18)

193
where, tmm is the minimum measured remaining wall thickness of the tank in the corroded

area. The adjacent undamaged zone (VU) bounded by the decay lengths and assumed to

be a part of reference volume is given by

VU = 4tc((xc+a)(xl+b)-ab) (7.19)

where, tc is the corroded thickness of the tank wall away from the region of LTA, xl and

xc are the decay lengths in the axial and circumferential direction (Fig.7.6). The reference

volume (VR) is therefore the sum of damaged and undamaged volumes.

VR = VD +VU (7.20)

Rectangular
Corrosion

LTA

Fig. 7.5 Tank with Locally Thinned Area [Ahmad, et al., 2010]

194
 xl
2b
LTA

Reference Volume

xl
xc 2a xc

Fig. 7.6 Tank with LTA and Reference Volume as used by Ahmad et al. [2010]

As mentioned earlier in Chapter.2, the mα multiplier is a function of mo and mL

where, mo depends upon reference volume. The accuracy of the multipliers depend on the

choice of reference volume. Using the reference volume adopted by Ahmad, et al.,

[2010], RSF for various heights of damage zone sizes are computed using the analytical

approach. These are shown in Fig.7.7. The method is quite conservative for small damage

sizes. However, the results show that the approach is significantly unconservative for

larger damage sizes. Even more importantly, the approach does not taper down to the

bottom limit that the RSF must converge to as the damage becomes large.

195
 Fig. 7.7 RSF using Analytical Approach –Ahmad, et al. [2010]

In the above procedure, the centre of the LTA also forms the centroid of the

reference volume, i.e., the reference volume is symmetric in both directions with respect

to the centroidal axes of LTA. The effect of pressure being non-uniform (Hydrostatic) on

the reference volume is not included. Also, the decay lengths used in the procedure are

based on elastic theory. However, it must be noted that the reference volume based on

elastic theory may not exactly be the same as the volume participating in plastic collapse.

7.5 MODIFIED REFERENCE VOLUME APPROACH

Ahmad’s reference volume taken as LTA plus decay length on the outside is

reasonable from the point of view that it must be equal to the kinematically active

volume. However, as outlined above, this approach seems to give inconsistent results that

196
are most likely unacceptable for larger damages. One of the reasons for this could be the

choice of reference volume. A brief explanation is given below. For a simple cylinder,

subject to hydrostatic loading, the limit pressure of the tank is given by

 yt
pLimit 
R (7.21)

 y2VR
m0 
  eq dv
2

VR (7.22)

y
mL 
 max (7.23)

Since the pressure is varying with respect to height,

h2

  eq dv    eqtdh
2 2

VR h1
(7.24)

Since the tank wall is subjected only to hoop stress

2
h2 h2
  hR     R 2  h2
  eqtdh    t  tdh   t   h dh
2 2
(7.25)
h1 h 1  1h

h
  R 2 h3  2
m  0
 (7.26)
 3t  h 1

Using the tank geometry as given in Chapter2 CASE 2 (H = 12000mm; R =

30000mm; t = 18mm) and employing the mα-tangent method (Eq.7.22 - 7.26), Fig. 7.8

shows the effect of changing reference volume. The actual limit load multiplier for the

197
hydrostatic loading case is 2.29. This is the same as that of the uniform pressure case

since the limit load depends on the highest pressure at the bottom. The true reference

volume for this case is a point at the bottom of the tank wall. Hence, for tanks with

hydrostatic loading, considering a larger volume or the entire volume as reference

volume deviates the mα-tangent values away from the actual limit load values. However,

in case of uniform pressure components such as pipes, even assuming a slightly larger

volume as reference volume has less impact on the mα-tangent values.

Fig. 7.8 mα-Tangent Value for Tank with Hydrostatic Pressure

A more detailed study of a tank (with hydrostatic pressure) whose bottom is fixed

has been carried out with ANSYS [2011]. Some of the results are shown in Fig.7.9. The

deviation from the full scale nonlinear analysis is very pronounced for small reference

198
lengths (less than 2.5 Rt from maximum stress location). The error is quite small at

2.5 Rt beyond which the error increases again. Therefore, a good choice for reference

volume estimate plays a major role in the computation of limit load

2.5√(Rt)

Fig.7.9 mα-Tangent Value for Tank with Hydrostatic and Uniform Pressure

From the concepts of decay length, it is obvious that the influence of discontinuity

moments at the bottom of the tank wall becomes insignificant beyond a height of 2.5 Rt .

Hence considering the volume till this height, as reference volume gives results close to

nonlinear FEA. Beyond this limit, the values diverge (increases) because the stress in the

tank wall decreases with height. In case the bottom hydrostatic pressure is applied as a

uniform pressure throughout the height, the mα-tangent value doesn’t change

199
significantly. Hence for tanks with hydrostatic pressure, considering larger volumes as

reference volume is unconservative.

Earlier Researchers (Seshadri [2005], Tantichattonont [2007]) had considered the

damaged volume and the undamaged volume enclosed by decay lengths outside the

damage as the reference volume. For tanks with hydrostatic pressure, this approach

assumes a relatively large volume to be part of the reference volume and hence the m-

tangent multiplier value was higher than that from non-linear FEA.

In light of the above, in this current study the use of a modified reference volume

estimate has been explored. This assumes that the most active volume is covered by the

zone obtained by overlapping decay lengths from either side of the discontinuity. There

may be some adjacent volume that is also kinematically active to some extent. However,

it is neglected for the purpose of simplicity. This is shown in Fig. 7.10. If the flaw is

large and the decay lengths from the discontinuity do not overlap, the most active volume

is the in-between zone of the highly stressed portion of LTA. It can be shown

theoretically that this is guaranteed to taper down the RSF to its lower limit (t mm/tc) for

large flaws. Hence this zone is considered as reference volume although this idea is not

fully in line with the original thinking behind the reference volume concept. It must be

noted that for small damage, the decay lengths from the edges will spill over on the other

edge of LTA. The reference volume will then include an appropriate undamaged zone as

well.

200
 Overlap area

2.5√(Rt) M
M
2.5√(Rt)
(a)

Highly Stressed
point

M 2.5√(Rt) 2.5√(Rt) M
(b)

Fig.7.10 Reference Zone for LTA

2.5√(Rt)
2.5√(Rt)
2.5√(Rt)

Overlap area
2.5√(Rt)

(Reference
Overlap area Volume) 2.5√(Rt)

2.5√(Rt)

LTA

(a) (b)

Fig.7.11 Modified Reference Volume (Decay Lengths Smaller than Flaw)

Figure 7.11 shows the overlap area obtained by considering the decay lengths in

both longitudinal and circumferential direction for flaw sizes that are larger than decay

201
lengths. In case the decay lengths are larger than the size of the flaw and spill beyond the

flaw (Fig.7.12), the entire volume of the flaw is included in the reference volume.

2.5√(Rt)
2.5√(Rt)

Overlap area
(Reference Fig.7.12 Modified
2.5√(Rt)
Volume) Reference Volume (Decay
LTA Lengths Larger than Flaw)

2.5√(Rt)

It must be noted that the failure of the tank with LTA damage happens only in the

region of reduced thickness, except for situations where the flaw is small and located

away from the bottom. In those exceptional situations, the damage does not affect the

limit capacity of the tank and the bending zone above the tank bottom will become

critical even if there is a minor damage elsewhere, as in the case of any undamaged tank.

7.6 ILLUSTRATIVE EXAMPLE

The example tank as given in Sect.3.5 CASE1 with the following specifications is

used for illustration. The tank is analyzed for hydrostatic pressure due to water at ambient

temperature.

202
Geometric Data:

Diameter : 60,000 mm

Height : 12,000 mm

Thickness of shell wall : 18 mm

Thickness of Bottom plate : 6 mm

Projection of Bottom plate : 50 mm (Beyond shell)

Material Data:

Young’s Modulus : 200,000 MPa

Yield Strength : 345 MPa

Product Design Stress : 194 MPa

Specific weight of liquid (water) : 9.81 kN/m3

Density of Steel : 77.1 kN/m3

Poisson’s ratio : 0.3

The required shell thickness (td) is determined using 1-foot method as prescribed in

5.6.3.2 of API 650

4.9 D( H  0.3)G
td  = 17.75mm, say t = 18mm
Sd

where Sd is the allowable product design stress (Table 5.2 of API 650) and G is the

specific gravity of the liquid.

Axial decay length: X l  2.5 Rt = 1837mm

Circumferential decay length: X C  6.3 Rt = 4630mm (as per the existing guidelines)

203
X C  2.5 Rt = 1837mm as per decay length proposed

LTA Dimensions:

For this illustrative calculation the longitudinal dimension (LTAH) of the LTA is

taken as 1.25 times 2.5 Rt (i.e., 2b =2296mm) and the circumferential dimension

(LTAW) is taken as 6.3 Rt (i.e., 2a = 4630 mm). The corroded thickness (tmm) is

assumed to be 9mm (50% corrosion).

Calculation of Reference Volume – Current Procedure [Ahmad, et al., 2010]:

 Volume of LTA (VD) = 4abtmm = 95.67 x 106 mm3

 Uncorroded Volume (VU) = [(2XC+2a)(2XL+2b)-4ab]tc = 13.96 x 108 mm3

 Reference Volume (VR) = (VU +VD) = 14.92 x 108 mm3

Evaluation of Hydrostatic Equivalent Pressure and Corresponding Stresses:

In order to account for the pressure varying hydrostatically over the reference

volume zone, Ahmad, et al. [2010], used the equivalent pressure which be determined by

centroidal calculations of pressure area as shown in Fig.7.13.

204
 Fig.7.13 Evaluation of Hydrostatic Equivalent Pressure

Triangular Area of LTA A1 = [(½) {γ (H-H1)} (2XL +2b)] = 174.832 N/mm

Rectangular Area of LTA A2 =[(γ H1) (2XL +2b)] = 353.15 N/ mm

Equivalent distance  A1 x1  A2 x2  = 9344mm

XG 
A1  A2
Hydrostatic equivalent pressure on LTA:

205
pavg = X G ρg = 91.66 kPa

Hoop stress in corroded region:

= (pavg Rc)/tmm = 305.56 MPa

where, Rc is the outer radius of the corroded thickness.

Hoop stress in uncorroded region:

= (pavg R)/tc = 152. 8 MPa

Only hoop stress is considered here as the tank is subjected to only hydrostatic pressure,

without any blanket pressure. Tank blanketing, or padding, refers to applying a cover of

gas (usually nitrogen) over the surface of a stored liquid to prevent it from vaporizing

into the atmosphere. It can maintain the atmosphere above a flammable or combustible

liquid to reduce ignition potential. It can make up the volume caused by cooling of the

tank contents, preventing vacuum and the ingress of atmospheric air. Hence in this

situation the equivalent stress is the same as hoop stress.

Therefore, σeu = 152. 8 MPa ; σec = 305.6 MPa

Evaluation of Multipliers:
y
Upper bound multiplier for the undamaged tank: mU0  = 2.258
 eu

y
Lower bound multiplier for the damaged tank: mLd  = 1.129
 ec

206
  y2VR
Upper bound multiplier for damaged tank: md0  = 2.068
 eU
2
VU   eC
2
VD

The mα-tangent multiplier for damaged tank using Eq. 2.28 is

m0 m0
mT  = 1.663; where,   d
1  0.2929(  1) m Ld

7.7 RSF USING ANALYTICAL APPROACH

Three different RSFs are evaluated here using the aforementioned limit load

multipliers.

md0
Using Eq. 7.15, RSFU  = 0.916
mu0

mT
Using Eq. 7.16, RSF  T
= 0.736
mu0

m Ld
Using Eq. 7.17, RSFL  = 0.5
mu0

All of the above RSF values are determined without using FEA.

7.8 RESULTS AND DISCUSSION

The RSF values from different approaches are compared with those from

nonlinear FEA results. The results are for different damage heights as indicated in the

tables. For all damages, the width is kept as 4630 mm ( 6.3 Rt ). As mentioned earlier,

the stress values required as input for mT calculations can be determined theoretically

207
(using shell theory or any other acceptable method) or from a linear elastic FEA

(LEFEA). The approach using LEFEA is more precise since the variation of stress in the

damage zone and the surrounding reference volume is taken in to account, whereas in a

simplified analytical approach the maximum stress is assumed for the entire damage zone

or the surrounding reference volume.

7.8.1 Results from m-Tangent Method Using LEFEA

In all the following tables, RSF1 indicates remaining strength factor determined

using mT method with proposed reference volume aproach and LEFEA using 2.5 Rt as

circumferential decay length. RSF2 is similar to RSF1 but the circumferential decay

length is taken as 6.3 Rt . RSF3 results are from LEFEA with the full volume taken as

reference volume in accordance wih Ahmad, et al. [2010].

Table 7.1 RSF Values from Nonlinear FEA and m-Tangent Method Using xc = 2.5 Rt

mT Method ( )
LTAH RSF
(mm) mαT mαT (NLFEA)
Damaged Undamaged RSF1
Tank Tank
735 1.584 2.152 0.736 0.899
1102 1.54 2.22 0.694 0.824
1470 1.55 2.282 0.679 0.752
1837 1.59 2.346 0.678 0.703
2296 1.328 2.408 0.551 0.666
2756 1.255 2.459 0.51 0.649

In Tables 7.1-7.5, the ‘mαT Damaged (or Undamaged) tank’ column shows the

limit load multiplier for the damaged (or undamaged) tank. Table 7.1 compares the RSF

208
value from nonlinear FEA and RSF1. The RSF1 values are conservative compared to the

nonlinear FEA results but follow the same trend.

Table 7.2 compares the RSF values (from mT method) calculated using two

different circumferential decay lengths of 2.5 Rt (RSF1) and 6.3 Rt (RSF2). The RSF

values from both the cases are similar except for the RSF value corresponding to a

damage height of 1837mm. The LTA height of 1837mm is at the transition zone

(= 2.5 Rt ) for the decay length and a further adjustment in the proposed reference

volume may smooth this transition. The results show that the RSF results from the mT

method do not change significantly if the circumferential decay length is increased from

2.5 Rt to 6.3 Rt . Based on this observation and from the results given in Section 7.3,

it can be said that the decay length for large tanks can be taken as 2.5 Rt instead of

6.3 Rt . This is based on the current flaw sizes only. However, it is reasonable to

assume that this holds for smaller flaw sizes as well.

Table 7.2 Comparison of RSF Values from m-Tangent Method (xc= 6.3 Rt & 2.5 Rt )

LTAH mα -Tangent Method (xc= 6.3 Rt ) RSF1

(mm) (xc= 2.5 Rt )
Damaged Undamaged
RSF2
Tank Tank
735 1.561 2.152 0.726 0.736
1102 1.540 2.220 0.694 0.694
1470 1.550 2.282 0.679 0.679
1837 1.652 1.997 0.827 0.678
2296 1.328 2.408 0.551 0.551
2756 1.255 2.459 0.510 0.510

209
 Table 7.3 gives the RSF3 values (using entire volume as reference volume as used

by Ahmed, et al. [2010]) in comparison with elastic plastic FEA results. The results show

that RSF3 approach overpredicts the RSF significantly. The current results show that it is

necessary to use an appropriate reference volume even if LEFEA is used to determine the

stress field for m-tangent method.

Table 7.3 RSF Values from m-Tangent Method Considering Entire Volume as
Reference Volume [Ahmad, et al., 2010]
m-Tangent considering entire volume
LTAH RSF
mαT mαT
(mm) (NLFEA)
Damaged Undamaged RSF3
Tank Tank

735 2.197 2.223 0.988 0.899

1102 2.183 2.443 0.894 0.824
1470 2.101 2.495 0.842 0.752
1837 2.084 2.449 0.851 0.703
2296 2.090 2.424 0.862 0.666
2756 2.102 2.408 0.873 0.649

7.8.2 Results from m-Tangent Method Using Analytical Approach

The RSF values obtained from the analytical approach without using FEA

[Ahmad, et al., 2010] are given in Table 7.4. The RSF-U and RSF-L values represent

RSF based on upper bound and lower bound multipliers whereas RSF-T is based on mαT

multiplier. The results show a very broad range within which the actual RSF value will be

located.

210
 Table 7.4 RSF from m-Tangent Method using Analytical Approach
LTAH RSF
RSF - U RSF - T RSF-L
(mm) (NLFEA)
735 0.961 0.757 0.5 0.899
1102 0.947 0.750 0.5 0.824
1470 0.935 0.745 0.5 0.752
1837 0.926 0.741 0.5 0.703
2296 0.916 0.736 0.5 0.666
2756 0.907 0.733 0.5 0.649

7.8.3 Influence of Hydrostatic vs. Uniform Pressure Loading on RSF from

m-Tangent Method

The m family of methods were initially applied to pressure vessels with uniform

pressure by Tantichattanonot and others where it works reasonably well. A brief study is

carried out in this thesis to understand if there are any significant differences in RSF

obtained from tank loaded with hydrostatic pressure and uniform pressure. The same tank

geometry and LTA dimensions as used previously in this Chapter are considered. RSF1

values (from m-tangent method using LEFEA) of the tank for the two types of loading

are compared in Fig. 7.14 and Table 7.5. The uniform pressure for an LTA is taken as

equivalent to the hydrostatic pressure at the centre of the flaw. As can be seen there is

very little difference in RSF between these two cases.

The results suggest that eventhough there is a stress gradient in the LTA due to

hydrostatic pressure, the average stress in the LTA (i.e., the stress at the centre of LTA) is

the prime factor that governs the RSF obtained using m-tangent method. Hence taking

211
the maximum stress at the centre of LTA (even though in reality it is slightly lower than

the centre) for m-tangent analytical method may be a reasonable assumption.

Table 7.5 Comparison of RSF Values for Hydrostaic Pressure and Uniform Pressure
Loading

Uniform Pressure Loading

RSF1
LTAH mαT mαT RSF (Hydrostatic
(mm) Damaged Undamaged (Uniform Pressure)
Tank Tank Pressure)

735 2.124 1.553 0.731 0.736

1102 2.172 1.491 0.687 0.694
1470 2.214 1.482 0.669 0.679
1837 2.256 1.500 0.665 0.678
2296 2.334 1.265 0.542 0.551
2756 2.334 1.220 0.523 0.510

Fig.7.14 RSF Values for Hydrostatic and Uniform Pressure Loadings

212
 1
0.9
0.8
0.7
0.6
R
S 0.5
F 0.4 RSF1
0.3 RSF(Elastic Plastic FEA)
0.2
RSF - API
0.1
0
0 500 1000 1500 2000 2500 3000
LTAH(mm)

Fig.7.15 RSF Using Proposed Reference Volume Approach

Figure 7.15 compares the RSF calculated using nonlinear FEA, mα-tangent

method (RSF1) and API 579-1/ASME FFS-1 procedure. It must be pointed out that

ASME FFS-1 is based on curvefitting adjustments to Folias formula using a massive

number of nonlinear finite element analyses for certain types of flaw geometries in pipe

like structures. The flaw geometry that is used in the current study falls into this

category. Although no special studies have been conducted for tanks in this regard, the

results for pipe like structures have been extended to tank structures by API 579-1 and

API 650. It seems to be a reasonable thing to do in light of the results shown in Table

7.5. This explains the very close agreement between API 579 values and nonlinear FEA

carried out in the current thesis. Such closeness may or may not be present for all

213
geometries. On the basis of assuming nonlinear FEA values to be accurate, the RSF1

values are conservative However, the method is still quite general in its scope and is

applicable for a wide variety of situations and structures.

One of the aims of the present work is to test its applicability for FFS assessment

of LTAs in tanks. An important issue in applying mα-tangent method is the selection of a

good reference volume. The choice employed in the present work seems to be quite

promising in the sense that the results are significantly improved from the previous study

using the mα-tangent method. However, there is also scope for improvement in order to

bring the results close to those from NLFEA. This clearly has the potential to provide an

alternative approach to current FFS methods as well as extend the general applicability of

mα-tangent method.

Figure 7.16 summarizes the full set of RSF results from before. The RSF3 values

obtained by taking the entire volume as reference volume [Ahmad, et al., 2010] are, as

expected, higher than NLFEA values and do not follow the common trend seen in other

methods. Again the RSF1 values from the proposed method relate favourably with

nonlinear FEA. It can be observed that, even though determining the exact reference

volume for the tank is a difficult task, incorporating the proposed approach for reference

volume will improve the results compared to the previous approach. However, some

modification to the reference volume is needed to avoid the transition spike seen when

the decay length overlaps transition from one pattern to another. This could be the

subject of further study [Sathyanarayanan and Adluri, 2013b].

214
 Fig.7.16 RSF Using Three Different Approaches for Reference Volume

7.9 SUMMARY

Locally thinned areas in tank wall are analyzed to determine the remaining

strength factor. The `mα-tangent method based on ‘integral mean of yield’ and variational

priciples to determine limit loads has been shown to be very useful for a range of

applications by earlier researchers. The current work adapts this method for RSF

estimates of partially damaged above ground storage tanks and presents preliminary

results. The tanks typically have very large R/t ratio (1000 or more) compared to regular

pipes and other cylindrical pressure vessels with much lower R/t ratios (50 to 100). In

view of this, the applicability of using 6.3 Rt as circumferential decay length is

215
examined. Based on FEA results of tank models, it is observed that the stress levels do

not change significantly beyond a value of 2.5 Rt and hence this value is recommended

as circumferential decay length as well as longitudinal decay length for practical tank

sizes. A modified procedure for reference volume estimate is examined. The RSF results

from mα-tangent method using this modified reference volume are conservative and

compare reasonably well with API 579-1/ASME FFS-1 and full scale nonlinear results.

The modified reference volume seems to be considerably better than that used before.

The present work shows the applicability of mα-tangent method for FFS assessments of

LTAs in tanks. The proposed method can be refined further to improve the performance

near transition zones (of decay length overlaps), and can be extended to include

interaction of two or more flaws.

216
CHAPTER 8

CONCLUSIONS AND RECOMMENDATIONS

8.0 SUMMARY

Large steel storage tanks with flat bottoms placed above ground are part of

several industries. Design and construction of these tanks involves significant

engineering expertise and cost. This thesis examines several issues regarding the design

and evaluation of such tanks with respect to the widely used design standard API 650.

The shell to bottom joint is a critical failure zone in the tanks. The hydrostatic

pressure from the liquid infill (combined with temperature loading in some cases) induces

internal moment at this joint. The high magnitude of this moment results in yielding of

the bottom plate near shell wall in most of the tanks. The fill-draw cycle, especially in

case of tanks operating in elevated temperatures causes low cycle fatigue in this region

which subsequently decides the overall life of the tank. Even though the problem of

plastic yielding and fatigue at this joint are recognized by earlier researchers and the

current design standards, very few works have been directed to study this problem.

Denham, et al., proposed a beam model with a fixed bottom plate projection

length beyond shell wall to analyze the stresses in this zone. In this thesis, Denham’s

model is reviewed and validated using FEA. It is extended to include arbitrary projection

lengths. The effect of increasing projection length beyond the shell wall has been studied

217
for tanks on ring wall foundations.It is found that the projection length influences the

stresses in the shell as well as the bottom plate near the shell to bottom joint. Increasing

the length is beneficial to a certain extent beyond which the stresses will not be altered.

This maximum projection length is termed as ‘full projection length’ and an extension to

Denham’s model is proposed to determine this length.

The influence of bottom plate friction forces on the stresses in the wall of elevated

temperature tanks is studied. The current API 650 procedure based on Karcher’s work

employs a factor called C to represent the bottom plate restraint due to friction forces.

Karcher’s equations used for finding the stresses in elevated temperature tank wall are

verified using a detailed finite element model with contact elements. It is observed that

the equations give correct results if the exact C-factor is used. Since the current practice

does not have clear guidelines for selecting a C value, a set value (like 0.85) is being used

irrespective of the tank dimensions and temperature change. The present study shows that

this approach is grossly conservative for most of the range of design parameters (H, R &

T) and could become slightly unconservative for very large tanks at relatively low

temperature changes. It is shown that the C-factor is directly related to the friction

coefficient () and tank parameters. Using plate theory the frictional restraint to bottom

plate expansion is formulated as function of . Appropriate equations for stresses and by

extension, fatigue life are presented incorporating  instead of C.

The bending stress in the shell increases with increasing friction forces. It is found

that for any tank geometry with a defined operating temperature, there is a limiting

218
friction coefficient (L) that will cause maximum stresses in shell. A friction coefficient

value greater than L will not increase the stresses any further. On the other hand,

increase in operating temperature also increases stresses in shell. Similar to limiting

friction coefficient, there exists a limiting temperature (TL) beyond which increase in

temperature will not induce additional stresses. Expressions were formulated to obtain L

and (TL).

As mentioned earlier, the low cycle fatigue at the shell to bottom joint caused by

the fill/draw down cycle of the stored liquid determines the cyclic life of the tank. The

peak alternating stress at this location, used to enter the fatigue curves is currently

determined using a pseudo elastic stress that represents strain range due to inelastic

deformations. The API 650 procedure for life cycle evaluation of elevated temperature

steel storage tanks subjected to low cycle fatigue is reviewed. The current procedure uses

beam-on-elastic foundation model to determine the peak alternating stress. In the current

study, it is observed that this model is not appropriate for tanks resting on concrete ring

walls and hence an alternate approach is recommended based on beam model (instead of

beam on elastic foundation model). The results are validated using finite element

analysis.

Apart from the damage issues at the shell to bottom joint, corrosion damage in the

tank wall as a locally thin area (LTA) is a widespread problem. In this thesis, a modified

reference volume procedure for a Level 2 fitness for service assessment method using mα-

tangent method is proposed for tanks containing LTA. The current method for reference

219
volume determination is modified to improve the accuracy of results. The volume inside

the flaw overlapped by the decay lengths from the edges of the flaw appears to give better

estimates. The circumferential decay length for tanks with very large R/t ratio is

examined and new recommendations are made.

8.1 CONCLUSIONS

The main conclusions from present research are as follows:

1. Increasing the projection length of annular plate outside the tank up to a certain

limit termed as ‘full projection length’ is beneficial to the shell stresses. A model

is proposed to determine this ‘full projection length’. Relevant equations are

presented. Since most of the large tanks are constructed with multiple wall

thicknesses, the effect of assuming a uniform thickness for a variable thickness

tank wall for the sake of determining the ‘full projection length’ is studied and

found to be acceptable.

2. The results of including or excluding the self-weight of the tanks in the procedure

to determine the ‘full projection length’ are compared and found that including

the self-weight does not increase the stresses in the tank wall or on the inner side

of bottom plate significantly, whereas it may increase the stresses marginally on

the outer side of the bottom plate.

3. Based on the slope compatibility at shell to bottom joint, the following new

equation is proposed to determine the elastic moment (Mo) at the bottom when

the bottom plate has ‘full projection length’ outside the shell.

220
 3
 M o   t s 6
   45.86 2  83.74  38.06  0
 M fx   ta 
 

4. In the current design procedure for elevated temperature tanks, the friction has not

been considered. This is sometimes quite confusing and leads the designer to

arbitrarily choose a C-factor value. The C-factor from the present practice can be

replaced using the following

C
1     H R
3 E t p  T

Appropriate equations for stresses and by extension, fatigue life are also

presented.

5. The current practice for shell to bottom joint rotation uses beam-on-elastic

foundation theory. It is shown that this is not valid for concrete ring walls.

Appropriate modifications have been proposed to estimate the peak alternating

stress (strain range) for fatigue design as below

2/3
3 2 
2/3
 R2 
  Ets  3SY tb2 
Sb     H    H  CT  1
2  1 2   Hts 
  R  


Detailed finite element analysis using nonlinear models and friction-contact

elements show that the alternative model is quite close to FEA results whereas the

current procedures under predict the fatigue stress significantly. The current API

procedure is clearly less rational than that developed in the present study. The

study recommends that the relevant API committee take a close look at this

practice.

221
6. The inequality condition in the fatigue design procedure of API 650 is examined

and found that though it is a necessary condition, it is not a sufficient condition to

ensure the formation of plastic hinges. It is also found that this inequality is

satisfied automatically for most of the design range and hence becomes

superficial unless the shell or plate thicknesses are more than the values necessary

from stress point of view.

7. Circumferential decay length for large tanks is studied using FEA and a value of

2.5 Rt is recommended instead of 6.3 Rt for reference volume calculations of

tanks. This reduction is likely the result of very large R/t ratios for tanks.

8. In order to improve the accuracy of fitness for service estimates using the

versatile m-tangent method, a modified reference volume scheme for

determination of remaining strength factor is examined. The RSF determined

using this modified reference volume follows elastic-plastic FEA results much

better compared to the results of previous studies. The reference volume estimate

can be refined to improve the performance further.

222
8.1.1 Guidelines for Designers

The following table gives guidelines for tank designers based on some of the main

results of this thesis work.

Existing Procedure Proposed Procedure

Obtain C from the following expression by

substituting the standard friction coefficient
C =0.25 to 1; C=0.85 if no value is given by
values
purchases

The peak alternating stress range is found Using the proposed beam model the peak
using alternating stress is given by
2/3
3 2 
2/3
 R2 
  Ets  3SY tb2 
  Sb     H   H  CT  1
0.658 R 2
Ktb  CTEts 3
SY tb2 2  1 
2
 Hts 
 R  

Sb   4 1   H   
ts E  R 2 
The peak alternating stress equation as given
above can be used only if the following is
true The condition is superfluous and need not be
 58HG 26.2C Tts0.5   4.8BSY tb2  checked
 0.5  G    1.5 
  Dts     Dts  
1.5
D
The projection shall be “full projection
length (a)” given by the following
expressions
M o where
Projection length of Annular plate beyond a 
5.09w
shell shall be minimum 50 mm.  2  t  t   RHts  1 
M o   0.3138  s   0.1429  s   1.0514  1  


 tp



 tp 
 


 12 1  
2   H

223
8.1.2 Recommendations for Future Study

1. In this thesis, an elastic beam model with plastic hinge boundary condition is used

to determine the peak alternating stress for fatigue life evaluation. This model can

be improved to elastic-plastic beam model to account for partial plastic hinge

conditions at the shell to bottom joint

2. A field test using instrumented strain gages on an elevated temperature tank can be

performed to confirm the stresses in the tank shell due to bottom plate friction. This

will show the influence of fill/draw cycle on the friction stresses. The influence of

lap joints, backing strips and other constructional joints underneath the bottom plate

on the shell stresses can be studied.

3. In this work, the foundation underneath the bottom plate is assumed to have very

high stiffness. The analysis can be extended to a flexible foundation and the

influence of the foundation deformations on the frictional stresses can be studied.

4. The influence of R/t ratio on the decay length of vessels with larger R/t ratio can be

studied theoretically. Since the decay lengths are used in FFS procedures to assess

plastic collapse, the decay length values obtained from the elastic theory can also be

compared using elastic-plastic analysis.

5. The LTA is assumed to be rectangular with uniform thickness loss. Complex shapes

and varying thickness profiles can be studied. The proposed modifications to the

reference volume can be extended to include interaction effects from multiple

flaws.

224
6. The author has attempted to evaluate and improve specific design issues currently

adopted by the widely used American Petroleum Institute standard – API 650. The

relevant loadings to address these issues have been used. The issues addressed are

specific to the dead, live and thermal loadings only. The tank can obviously be

subjected to other loading types. The standard addresses them in an entirely

different manner and hence they are not examined in this thesis. They could be part

of further study.

225
PUBLICATIONS AND PRESENTATIONS
FROM THE PH.D. PROGRAM
1. Sathyanarayanan, S. and Adluri, Seshu M. R., 2013, “Fatigue Stress Evaluation at
Shell to Bottom Joint in Elevated Temperature Steel Tanks on Concrete Ringwalls,”
ASME Journal of Pressure Vessel Technology, (Provisionally Accepted).

2. Sathyanarayanan, S. and Adluri, Seshu M. R., 2013, “Improvements in Adapting

mα–Tangent Method for Fitness for Service Evaluation of Local Thin Areas in
Storage Tanks,”Proc. of ASME 2013 Pressure Vessels & Piping Division
Conference (PVP2013), Paris, France.

3. Sathyanarayanan, S. and Adluri, Seshu M. R., 2013, “Incorporation of Friction

Coefficient in the Design Equations of Elevated Temperature Tanks,” ASME
Journal of Pressure Vessel Technology, 135(2), March 2013.

4. Sathyanarayanan, S. and Adluri, Seshu M. R., 2012, “Modified Fatigue Stress at

Shell-To-Bottom Joint of Steel Tanks on Ring Walls,” Proc. of ASME 2012
Pressure Vessels & Piping Division Conference (PVP2012), Toronto, Canada.

5. Sathyanarayanan, S. and Adluri, Seshu M. R., 2012, “Locally Thin Areas in Above
Ground Steel Storage Tanks,” Presented at the Aldrich Conference, St. John’s,
Canada.

6. Sathyanarayanan, S., Adluri, Seshu M. R., 2011, “Effect of Annular Plate

Projection Length on the Stresses in the Above Ground Steel Storage Tanks on
Rigid Ring Wall Foundations,” Proceedings of the ASME 2011 Pressure Vessels
& Piping Division Conference (PVP2011), Baltimore, Maryland, USA.

226
REFERENCES

Adibi-Asl, R. and Seshadri, R., 2011, “Thermal Hot Spot and Corrosion Damage in

Conical Pressure Components,” ASME Journal Pressure Vessel Technology, 133.

Adluri, Seshu M. R., 2012, “Effects of Friction on the Bottom Plate of Large Elevated

Temperature Steel Tanks,” Research Report, Faculty of Engineering and Applied

Science, Memorial University of Newfoundland, Canada.

Ahmad, F., Hossain, M. M., and Seshadri, R., 2010, “Fitness for Service Assessment of

Hydrocarbon Storage Tanks,” ASME Journal Pressure Vessel Technology, 132(2).

American Petroleum Institute Standard, 2012, Welded Tanks for Oil Storage –API 650,

Eleventh Ed.

American Petroleum Institute Standard, 2012, Fitness-For-Service API 579-1/ASME

FFS-1, Second Ed.

American Petroleum Institute Standard, 2009, Tank Inspection, Repair, Alteration, and

Reconstruction API 653, Fourth Ed.

American Society of Mechanical Engineers (ASME), 2002, ASME Boiler and Pressure

Vessel Code, Section III and VIII, New York.

Andreani, J. L., Osage, D. A., Parikh, P. D., Horwege, J. A., 1995, “An Evaluation of

Procedures for Determining the Fitness-for-Service of Settled Aboveground Storage

227
Tanks,” Proceedings of the ASME Pressure Vessel and Piping Conference-ASME PVP

1995 , 315, pp.45-66.

ANSYS, 2009, University Research Version 11.0, SASIP Inc.

Ansys Contact Technology Guide, 2009, Release 11.0, SASIP Inc.

Antaki, G., 2005, Fitness-for-Service and Integrity of Piping, Vessels, and Tanks, ASME

CODE SIMPLIFIED, McGraw-Hill Inc., New York.

Batte,A. D., Fu, B., Kirkwood, M.G., and Vu, D., 1997, “Advanced Methods for Integrity

Assessment of Corroded Pipelines,” Pipes and Pipelines International, 42 (1).

Boresi, A. P., and Schmidt, R. J., 1993, Advanced Mechanics of Materials, Fifth Ed.,

John Wiley & Sons, Inc., New York.

Digrado, B. D., and Thorp, G. A., 2004, The Aboveground Steel Storage Tank Handbook,

John Wiley & Sons, Inc., New Jersey.

Burgreen, D., 1975, Design Methods for Power Plant Structures, First Edition, C. P.

Press, Jamaica, New York.

Calladine, C. R., 2000, Plasticity for Engineers: Theory and Application, Horwood

Publishing, Chichester, England.

228
Chen, H. F., and Shu, D. W., 2000, “The Effects of the Distance Between Two Defects

on the Load Carrying Capacity of a Pressure Vessel,” ASME Journal of Pressure Vessel

Technology, 122, pp.198-203.

Chen, Z. P., Duan, Y. Y., Shen, J. M., and Jiang, J. L., 2007, “A Simplified Method for

Calculating the Stress of a Large Storage Tank Wall,” Proceedings of the Institution of

Mechanical Engineers, PART E: Journal of Process Mechanical Engineering, 221(3),

pp.119-127.

Chopra, O. K., Shack, W. J., 2003, “Review of the Margins for ASME Code Fatigue

Design Curve - Effects of Surface Roughness and Material Variability,” Report for U.S.

Nuclear Regulatory Commission, Washington, D.C.,USA (Available online

http://www.nrc.gov/reading-rm/doc-collections/nuregs/contract/cr6815/cr6815.pdf -

excerpted on 19-Dec-2013)

Comfort, L., Abrams, J., Camillus, J., and Ricci, E., 1989, “From Crisis to Community:

the Pittsburgh Oil Spill,” Industrial Crisis Quarterly, 3(1), pp.17-39.

Cornell, J. R., and Baker, M. A., 2002, “Catastrophic Tank Failures: Highlights of Past

Failures along with Proactive Tank Designs,” Proceedings of the US EPA Fourth

Biennial Freshwater Spills Symposium (FSS 2002), Ohio, USA.

Denham, J. B., Russell, J., and Wills, C. M. R., 1968a, “How to Design a 600 000-BBL.

Tank,” Hydrocarbon Processing, 47(5), pp.137-142.

229
Denham, J. B., Russel, J., and Wills, C. M. R., 1968b, “Comparison of Predicted and

Measured Stresses in a Large Storage Tank,” Proceedings of Division of Refining – API,

pp.1034-1074.

European Committee for Standardization, 2004, EN 14015 - Specification for the design

and manufacture of site built, vertical, cylindrical, flat-bottomed, above ground, welded,

steel tanks for the storage of liquids at ambient temperature and above.

Folias, E. S., 1969, “On the Effect of Initial Curvature on Cracked Flat Sheets,”

International Journal of Fracture Mechanics, 5(4), pp. 327-346.

Folias, E. S., 1999, “Failure Correlation between Cylindrical Pressurized Vessels and Flat

Plates,” International Journal of Pressure Vessels and Piping, 76, pp.803–811.

Fu, B., and Kirkwood, M. G., 1995, “Determination of Failure Pressure of Corroded

Pipelines using the Non-Linear Finite Element Method”, Proc. of 2nd International

Pipeline Technology Conference, Ostend, Belgium.

Gill, S. S., 1970, The Stress Analysis of Pressure Vessels and Pressure Vessel

Components, Pergamon Publishing, London, UK.

Hetenyi, M., 1971, Beams on Elastic Foundation, University of Michigan Press, Ann

Arbor, MI, USA.

230
Indermohan, H., and Seshadri, R., 2004, “Fitness-for-service Methodology based on

Variational Principles in Plasticity,” ASME Pressure Vessels Piping Div Publ PVP 2004,

473, pp. 83-91.

Janelle, J., Osage, D. A., and Burkhart, S. J., 2005, An Overview and Validation of the

Fitness-For-Service Assessment Procedures for Local Thin Areas in API 579, Welding

Research Council (WRC) Bulletin, 505.

Janelle, J. L., and Osage, D. A., 2007, “Validation and Development of Fitness for

Service Procedures for Local Thin Areas in API RP579,” Proceedings of 2007 ASME

Pressure Vessel and Piping Division Conference, 6, pp.809-824.

Jawad, M. H., and Farr, J. R., 1984, Structural Analysis and Design of Process

Equipment, John Wiley & Sons

Jones, R., and Seshadri, R., 1989, “Analysis and Design of High Temperature Storage

Tanks,” Design and Analysis of Pressure Vessels and Components – 1989 (PVP), 175,

pp. 45-52.

Kanninen, M. F., Pagalthivarthi, K. V., and Popelar, C. H., 1991 “A Theoretical Analysis

for the Residual Strength of Corroded Gas and Oil Transmission Pipelines,” Proceedings

of the Symposium on Corrosion forms and Control for Infrastructure, California, USA.

Kanninen, M. F., Roy, S., Couque, H. R. A., Grigory, S. C., and Smith, M. Q., 1994,

“Generalized Guidelines for Determining the Residual Strength of Corroded Oil and Gas

231
Pipelines,” Proceedings of The Energy Transportation, Transfer and Storage Conference,

pp. 391-40.

Karcher, G. G., 1978a, “Thermal Stresses in Tanks Operating at Elevated Temperatures,”

Proceedings - API Division of Refining, American Petroleum Institute, New York, 57,

pp. 515-521.

Karcher, G. G., 1978b, “New Design Calculations for High Temperature Storage Tanks,”

Hydrocarbon Processing, October 1978, pp.137-140.

Karcher, G. G., 1981a, “Simplified Stress Equations for Elevated Storage Tanks,”

Hydrocarbon Processing, July 1981, pp. 515-521.

Karcher, G. G., 1981b, “Stresses at the Shell-To-Bottom Junction of Elevated

Temperature Tanks,” Proceedings - API Division of Refining, American Petroleum

Institute, New York, 60, pp. 154-159.

Kiefner, J. F., Maxey, W. A., Eiber, R. J., and Duffy, A. R., 1973, “Failure Stress Levels

of Flaws in Pressurized Cylinders,” Progress in Flaw Growth and Fracture Toughness,

ASTM STP 536, American Society for Testing and Materials, pp. 461-481.

Kiefner, J. F. and Vieth, P.H., 1989, A Modified Criterion for Evaluating the Remaining

Strength of Corroded Pipe (with RSTRENG), American Gas Association, Catalog No.

L51609, PR3-805.

232
Kiefner, J. F., 1990, “The Remaining Strength of Corroded Pipe,” API Pipeline

Conference, American Petroleum Institute.

Kim, J., An, D., Lee, S., and Lee, B., 2009, “A failure analysis of fillet joint cracking in

an oil storage tank,” Journal of Loss Prevention in the Process Industries, 22, pp. 845-

849.

Konusu, S., Kano, M., Mukaimachi, N., and Kanamaru, S., 2010, “Validity of Assesment

Procedure in p-M method for Multiple Volumetric Flaws,” ASME Journal of Pressure

Vessel Technology, 132, 021402 1-10.

Long, B., and Garner, B., 2004, Guide to Storage tanks and equipments, Professional

Engineering Publishing.

Osage, D. A., Janelle, J., and Henry, P. A., 2000, “Fitness-for-Service Local Metal Loss

Assesment Rules in API 579,” Proceedings of ASME Pressure Vessel and Piping

conference 2000 – Service Experience and Fitness-for-Service in Power and Petroleum

Processing, 411, pp. 143-176.

Osage, D. A., Krishnaswamy, P., Stephens, D. R., Scott, P., Mohan, R., Janelle, J., and

Wilkowski, G. M., 2001, Technologies For The Evaluation Of Non-Crack-Like Flaws In

Pressurized Components - Erosion/Corrosion, Pitting, Blisters, Shell Out-Of-Roundness,

Weld Misalignment, Bulges And Dents, Welding Research Council (WRC) Bulletin, 465.

233
Peng,J., Zhou, C., Xue, J., Dai, Q., and He, X., 2011,"Safety Assessment of Pipes with

Multiple Local Wall Thinning Defects Under Pressure and Bending Moment," Journal of

Nuclear Engineering and Design, 241, pp.2578-2765.

Sathyanarayanan, S. and Adluri, Seshu M. R., 2013a, “Fatigue Stress Evaluation at Shell

to Bottom Joint in Elevated Temperature Steel Tanks on Concrete Ringwalls,” ASME

Journal of Pressure Vessel Technology, (Provisionally Accepted).

Sathyanarayanan, S. and Adluri, Seshu M. R., 2013b, “Improvements in Adapting mα–

Tangent Method for Fitness for Service Evaluation of Local Thin Areas in Storage

Tanks,”Proc. of ASME 2013 Pressure Vessels & Piping Division Conference(PVP2013),

Paris, France.

Sathyanarayanan, S. and Adluri, Seshu M. R., 2013c, “Incorporation of Friction

Coefficient in the Design Equations of Elevated Temperature Tanks,” ASME Journal of

Pressure Vessel Technology, 135(2), March 2013.

Sathyanarayanan, S. and Adluri, Seshu M. R., 2012a, “Modified Fatigue Stress at Shell-

To-Bottom Joint of Steel Tanks on Ring Walls,” Proc. of ASME 2012 Pressure Vessels

& Piping Division Conference (PVP2012), Toronto, Canada.

Sathyanarayanan, S. and Adluri, Seshu M. R., 2011, “Effect of Annular Plate Projection

Length on the Stresses in the Above Ground Steel Storage Tanks on Rigid Ring Wall

Foundations,” Proceedings of the ASME 2011 Pressure Vessels & Piping Division

Conference(PVP2011), Baltimore, Maryland, USA.

234
Seshadri, R., 2004, “Integrity Assessment of Pressure Components with Local Hot

Spots,” Proc. of ASME 2004 Pressure Vessels Piping Division Conference, 480, pp. 177-

185.

Seshadri, R., 2005, “Integrity Assessment of Pressure Components with Local Hot

Spots,” ASME Journal of Pressure Vessel Technology, 127, pp.137-142.

Seshadri, R. and Hossain, M. M., 2008, “Simplified Limit Load Determination Using the

mα-Tangent Method,” Proc. of 2008 ASME Pressure Vessels and Piping Division

Conference, Paper No. PVP2008-61171, July, Chicago, USA.

Sims, J. R., Hantz, B. F., and Kuehn, K. E., 1992, “A Basis for the Fitness for Service

Evaluation of Thin Areas in Pressure Vessels and Storage Tanks,” Proc. of ASME 1992

Pressure Vessels and Piping Division Conference, 233, pp. 51-58.

Tantichattanont, P., 2006, “Fitness-for-Service Assessment for Thermal Hot Spots and

Corrosion in Pressure Vessels”, Ph.D. Thesis, Memorial University of Newfoundland,

Canada.

Tantichattanont, P., Adluri, Seshu M. R., and Seshadri, R., 2006a, “Integrity Assesment

of Spherical Components with Local Corrosion and Hot spots,” Proc. of ASME 2006

Pressure Vessels and Piping/ICPVT-11 Conference, 3, pp.653-662.

Tantichattanont, P., Adluri, Seshu M. R., and Seshadri, R., 2007a, “Fitness for Service

Evaluation of Thermal Hotspots and Corrosion Damage in Cylindrical Pressure

Components,” Proc. of ASME 2007 Pressure Vessel and Piping Conference, 2, pp. 71-82.

235
Tantichattanont, P., Adluri, Seshu M. R., and Seshadri, R., 2007b, “Structural Integrity

Evaluation for Corrosion in Spherical Pressure Vessels,” Int. Journal of Pressure Vessels

& Piping, 84, pp. 749–761.

Tantichattanont, P., Adluri, Seshu M. R., and Seshadri, R., 2007c, “Fitness for Service

Assessment of Spherical Pressure Vessels with Hotspots,” Int. Journal of Pressure

Vessels & Piping, 84, pp. 762-772.

Timoshenko, S., and Woinowsky-Kreiger, S., 1959, Theory of Plates and Shells,

McGraw-Hill, New York, Second Edition.

Wang, Y. H., Tham, L. G., and Cheung, Y. K., 2005, “Beams and Plates on Elastic

Foundation: a Review,” Journal of Progress in Structural Engineering and Materials, 7,

pp. 174-182.

Wiki, 2013, “Fatigue (material),” excerpted from online Wikipedia, the free encyclopedia

on 18- Dec-2013 at: http://en.wikipedia.org/wiki/Fatigue_(material)

Wu, T. Y., and Liu, G. R., 2000, “Comparison of Design Methods for a Tank-Bottom

Annular Plate and Concrete Ring Wall,” Int. Journal of Pressure Vessels and Piping, 77,

pp. 511-517.

Wu, T. Y., 1996, “More Accurate Method Devised for Tank-Bottom Annular Plate

Design,” Oil and Gas Journal, 94(21), pp. 81-83.

236
Zick, L. P. and McGrath, R. V., 1968, “Design of Large Diameter Cylindrical Shells”,

Proceedings - API Division of Refining, American Petroleum Institute, New York, 48, pp.

1114-1140.

237
BIBLIOGRAPHY

The following have been consulted during the research although they have not been

directly referred to in the thesis.

Ahari, M.N., Eshghi, S., and Ashtiany, M.G., 2009, “The Tapered Beam Model for

Bottom Plate Uplift Analysis of Unanchored Cylindrical Steel Storage Tanks,” Journal of

Engineering Structures, 31, 623-632.

American Petroleum Institute Standard,2004, Design and Construction of Large, Welded,

Low-Pressure Storage Tanks, API – 620, Tenth Ed.

Anderson, T. L., and Osage, D. A., 2000, “API 579: a Comprehensive Fitness-for-Service

Guide,” International Journal of Pressure Vessels and Piping, 77, pp. 953-963.

Baltay, P., and Gjelsvik, A., 1990, “Coefficient of Friction for Steel on Concrete at High

Normal Stress,” Journal of Materials in Civil Engineering-ASCE, Vol.2 (1), pp. 46-49.

Berardi, R., and Lancellotta, R., ,2002, “Yielding from Field Behavior and its Influence

on Oil Tank Settlements”, ASCE Journal of Geotechnical and Geo environmental

Engineering, 128 (5), pp. 404-415.

Brownell, L. E., and Young, E. H., 1959, Process Equipment Design, John Wiley & Sons

Inc.

238
Carucci, V. A., and Delahunt, J. F., 2002, “Corrosion Considerations for Aboveground

Atmospheric Storage Tanks,” Proc. of NACE CORROSION/2002, Denver, USA, Paper

No. 02487.

Chen, Z. P., Sun, B., Yu, C. L., and Zeng, M., 2009, “Finite Element Analysis of Liquid

Storage Tank Foundations using Settlement Difference as Boundary Condition,” Proc. of

the Institution of Mechanical Engineers, PART E: Journal of Process Mechanical

Engineering, 223(4), pp. 225-231.

Cinquini, C., Lamblin, D. O., and Guerlement, G., 1984, “Limit Analysis of Circular

Cylindrical Shells under Hydrostatic Pressure,” Journal of Structural Mechanics, 12, pp.

263-278.

Daviea, M., Nolana, P. F., and Hoban, T. W. S., 1994, “Case Histories of Incidents in

Heated Bitumen Storage Tanks,” Journal of Loss Prevention in Process Industries, 7(3),

pp. 212-221.

Gorochov, Y., Muschanov, V., Kulik, A., and Tsyplukhin, A., 2005, “Vertical Cylindrical

Tank with Angular Geometrical Imperfection”, Journal of Civil Engineering and

Management- Lithuania Academy, 11(3), pp. 175-18.

Greiner, R., and Ofner, R., 2003, “Large Cylindrical Flat-Bottom Storage Tanks – Design

Calculations for Seismic Effects,” Proc. of the Conference on Design, Inspection,

Maintenance and Operation of Cylindrical Steel Tanks and Pipelines, Prague.

239
Hamdan, M. N., 2002, “A Simplified Analysis of Edge Settlement of a Large

Aboveground Liquid Storage tank,” Proc. of the 6th Saudi Engineering Conference,

KFUPM, Dhahran, 3, pp. 19-35.

Hamdan, M. N., Abuzeid, O., and Al-Salaymeh, A., 2007, “Assesment of an Edge type

Settlement of Above Ground Storage tanks using a Simple Beam model,” Journal of

Applied Mathematical Modeling, 31, pp. 2461-2474.

James, R. W., and Raba, G. W., 1991, “Behavior of Welded Steel Water-Storage Tank,”

ASCE - Journal of Structural Engineering, 117, pp. 61-79.

Jaske, C. E., 2001, “Process Equipment Fitness-for-Service Assesments Using API RP

579,” Proc. of the Process and Power Plant Reliability Conference.

Jonaidi, M., and Ansourian, P., 1998, “Harmonic Settlement Effects on Uniform and

Tapered Tank Shells,” Journal of Thin Walled Structures, 31, pp. 237-255.

Jones, R. D, 1991, “Stress Analysis of High Temperature Storage tanks,” Master’s

Thesis, University of Regina, Regina, Saskatchewan, Canada.

Mackerle, J., 2004, “Finite Elements in the Analysis of Pressure Vessel and Piping, an

Addendum: A bibliography (2001-2004),” International Journal of Pressure Vessel and

Piping, 82, pp. 571-592.

Malhotra, P. K., and Veletsos, A. S., 1994, “Uplifting Analysis of Base Plates in

Cylindrical Tanks,” Journal of Structural Engineering, 120 (12), pp. 3489-3505.

240
Malhotra, P. K., 1995, “Base Uplifting Analysis of Flexibly Supported Liquid-Storage

Tanks,” Journal of Earthquake Engineering and Structural Dynamics, 24 (12), pp. 1591-

1607.

Marr, W. A., Ramos, J. A., and Lambe, T. W., 1982, “Criteria for Settlement of Tanks,”

ASCE Journal of Geotechnical Engineering Division, 108, pp. 1017-1039.

May, I. L., 2007, “Structural Integrity and Petrochemical Industry,” Proc. of 9th

International Conference on Engineering Structural Integrity Assesment – ESIA9,

Beijing.

McGrath, R. V., 1963, “Stability of API Standard 650 Tank Shells,” API Division of

Refining Report, 43(3), pp. 458-469.

Melerski, E. S., 1992, “An Efficient Computer Analysis of Cylindrical Liquid Storage

Tanks under conditions of Axial Symmetry,” Journal of Computer and Structures, 45(2),

pp. 281-295.

Myers, P. E, 1997, Aboveground Storage Tanks, McGraw-Hill Inc.

Peek, R., and Jennings, P. C., 1988, “Simplified Analysis of Unanchored Tanks,” Journal

of Earthquake Engineering and Structural Dynamics, 16(7), pp. 1073-1085.

Rabbat, B. G., and Russet, H. G., 1985, “Friction Coefficient of Steel on Concrete or

Grout,” Journal of Structural Engineering – ASCE, 111(3), pp. 505-515.

241
Rasiulis, K., Sapalas, A., Vadluga, R., and Samofalov, M., 2006, “Stress/strain State

Investigations for Extreme Points of Thin Wall Cylindrical Tanks,” Journal of

Constructional Steel Research, 62, pp. 1232-1237.

Slagis, G. C., 2006, “ASME Section III Design by Analysis Criteria Concepts and Stress

Limits,” ASME Journal of Pressure Vessel Technology, 128, pp. 25-32.

Sun, X., Liu, Y., Wang, J., and Cen, Z., 2008, “Stress and Deflection Analyzes of

Floating Roofs Based on Load-Modifying Method,” International Journal of Pressure

Vessels and Piping, 85, pp. 728-738.

Tin-Loi, F., and Pulmano, V. A., 1991, “Limit Loads of Cylindrical Shells under

Hydrostatic Pressure,” Journal of Structural Engineering, 117(3), pp. 643-656.

Trebuna, F., Simcak, F., and Bocko, J., 2009, “Failure Analysis of Storage Tank,”

Journal of Engineering Failure Analysis, 16, pp. 26-38.

Wozniak, R. S., and Mitchell, W. W., 1978, “Basis of Seismic Design Provisions for

Welded Steel Oil Storage Tanks,” API Division of Refining Report – Advances in Storage

Tank Design.

Yoshida, S., 2002, “Behavior of Localised Bottom Bulge in Above Ground Oil Storage

Tanks under Liquid Pressure,” ASME Journal of Pressure Vessel Technology, 124 (1),

pp. 59-65.

242
APPENDIX A

This Appendix gives the ANSYS APDL Macro for performing nonlinear analysis of the
example tank given in Chapter 3-CASE 1. The load is hydrostatic pressure due to water with
material properties specified in Ch.3. Shell elements are used to model the tank wall and
bottom plate. Friction interaction between the tank bottom and the foundation is taken in to
account using friction elements. The information given after exclamatory mark (!) is for
understanding of the user and is not part of the APDL.

/PREP7 ! Enter the Preprocessor module

ET,1,SHELL208 ! Element Type selected (Type 1)- Axisymmetric shell element
KEYOPT,1,8,2 ! This kepoint enables storing top,mid and bottom surface
results for all layers
sect,1,shell,, ! Sectional Properties (Geometry) - Type 1
secdata, 6,1,0.0,7 ! Each layer is 6mm with seven integration points
secdata, 6,1,0.0,7 ! totally three layers, each with 6mm thickness
secdata, 6,1,0.0,7
secoffset,TOP ! the centre line is at the top of thickness
seccontrol,,,, , , ,
sect,2,shell,, ! Sectional Properties - Type 2
secdata, 1,1,0.0,3 ! 8 layers, each 1mm thick with 3 integration points
secdata, 1,1,0.0,3
secdata, 1,1,0.0,3
secdata, 1,1,0.0,3
secdata, 1,1,0.0,3
secdata, 1,1,0.0,3
secdata, 1,1,0.0,3
secdata, 1,1,0.0,3
secoffset,TOP

243
seccontrol,,,, , , ,
MPTEMP,,,,,,,, ! Material Property Table
MPTEMP,1,0
MPDATA,EX,1,,2e5 ! Youngs' Modulus
MPDATA,PRXY,1,,0.3 ! Poisson's Ratio
MP,DENS,1,7.86e-6 ! Density
MP,ALPX,1,12E-6 ! Coefficient of Linear Expansion
MP,KXX,1,60.5 ! Thermal film coefficient - used for thermal analysis
*SET,PM,0 ! PM is the tag used for Plastic modulus; here the value is zero
indication perfectly plastic model
*SET,YS,345 !Define Yield strength
TB,BKIN,1,,,1 !Bilinear Kinematic material model (see the option in Ansys at
Material properties)
TBDATA,,YS,PM !since Plastic modulus has no value it indicates elastic-
perfectly plastic material model
K,1,200,1000 ! Keypoints defining the axisymmetric geometry
K,2,29200,1000
k,3,30009,1000
k,4,30068,1000
k,5,30009,2000
K,6,30009,13000
k,7,200,999.5 ! Keypoints defining the base
k,8,31768,999.5
LSTR, 1, 2 ! constructing straight lines from keypoints
LSTR, 2, 3
LSTR, 3, 4
LSTR, 3, 5
LSTR, 5, 6
LSTR, 7, 8
LESIZE,1,50 ! Assigning element size for lines
LESIZE,2,1
LESIZE,3,1

244
LESIZE,4,0.5
LESIZE,5,100
LSEL, , , ,1 !Selecting Line 1
LATT,1, ,1, , , ,2 !Assigning attributes to Line 1 for meshing here sectional properties -
Type 2 is assigned
LSEL, , , ,2
LATT,1, ,1, , , ,2
LSEL, , , ,3
LATT,1, ,1, , , ,2
LSEL, , , ,4
LATT,1, ,1, , , ,1
LSEL, , , ,5
LATT,1, ,1, , , ,1
allsel
LMESH, 1 ! Meshing Line 1
LMESH, 2
LMESH, 3
LMESH, 4
LMESH, 5
!--------CONTACT MODELING---------------
! creating a entity " source"
NSEL,R,LOC,X,0,32000 ! Selecting the nodes at the bottom of tank that will be in
contact
NSEL,R,LOC,Y,1000,1000.5
CM,source,NODE ! Creating an entity "source" from the selected nodes
Nsel,all
! /COM, CONTACT PAIR CREATION - START
CM,_NODECM,NODE
CM,_ELEMCM,ELEM
CM,_KPCM,KP
CM,_LINECM,LINE

245
CM,_AREACM,AREA
CM,_VOLUCM,VOLU
! /GSAV,cwz,gsav,,temp
MP,MU,1,0.85 !GIVE THE VALUE OF FRICTION HERE MU=0.2 (from greiner paper)
MAT,1
MP,EMIS,1,7.88860905221e-031
R,3
REAL,3
ET,2,169
ET,3,171
R,3,,,1,-.02,0, ! SPECIFY FKN HERE and the tolerance
RMORE,,,1.0E20,0.0,1.0,
RMORE,0.0,0,1.0,,1.0,0.5
RMORE,0,1.0,1.0,0.0,,1.0
KEYOPT,3,3,0
KEYOPT,3,4,0
KEYOPT,3,5,1
KEYOPT,3,6,2
KEYOPT,3,7,0
KEYOPT,3,8,0
KEYOPT,3,9,0
KEYOPT,3,10,2
KEYOPT,3,11,0
KEYOPT,3,12,0
KEYOPT,3,2,0 ! contact algorithm 0 - augmented lagran(default),1-penalty,4-pure
lagrangian
KEYOPT,2,2,0
KEYOPT,2,3,0
! Generate the target surface
LSEL,S,,,6 ! specify the target line here
CM,_TARGET,LINE

246
TYPE,2
LATT,-1,3,2,-1
TYPE,2
LMESH,ALL
! Generate the contact surface
NSEL,S,,,SOURCE ! selecting the entity "source"
CM,_CONTACT,NODE
TYPE,3
ESLN,S,0
ESURF
*SET,_REALID,3
ALLSEL
ESEL,ALL
ESEL,S,TYPE,,2
ESEL,A,TYPE,,3
ESEL,R,REAL,,3
LSEL,S,REAL,,3
! The normals of the contact and the target surface has to meet in order to make the contact.
! Else the normals of contact or target or both has to be reversed based on the situation
!-----------reverse contact elements normals-----------------
CM,_CWZ_EL,ELEM
CM,_CWZ_ND,NODE
CM,_CWZ_KP,KP
CM,_CWZ_LN,LINE
CM,_CWZ_AR,AREA
CM,_CWZ_VL,VOLU
ESEL,NONE
ESEL,A,REAL,,3
ESEL,R,ENAME,,171,177
NSLE
*GET,_z1,ELEM,,NUM,MAX

247
KSLN,S
LSLK,S,1
ASLL,S,1
*CREATE,cwzplot,mac
/COM,
/COM,PLOT CONTACT PAIR(S)
~eui,'::apdl::noprint 1'
~eui,'::apdl::nooutput 1'
/PNUM,REAL,1
/NUM,1
/PSYMB,ESYS,1
EPLOT
/PSYMB,ESYS,0
/NUM,0
/PNUM,TYPE,0
/PNUM,REAL,0
/mrep,cwzplot
~eui,'::apdl::nooutput 0'
~eui,'::apdl::noprint 0'
*END
cwzplot
*SET,_REALID,3
FLST,5,4168,2,ORDE,2
FITEM,5,16170
FITEM,5,-20337
CM,_ELMCM,ELEM
ESEL,S, , ,all
ESURF,,REVERSE
! /REPLOT
CMSEL,S,_ELMCM
CMDELE,_ELMCM

248
CMSEL,S,_CWZ_EL
CMDEL,_CWZ_EL
CMSEL,S,_CWZ_ND
CMDEL,_CWZ_ND
CMSEL,S,_CWZ_KP
CMDEL,_CWZ_KP
CMSEL,S,_CWZ_LN
CMDEL,_CWZ_LN
CMSEL,S,_CWZ_AR
CMDEL,_CWZ_AR
CMSEL,S,_CWZ_VL
CMDEL,_CWZ_VL
! /MREP,EPLOT
!-----------------end of reversing---------------
! /REPLOT
ESEL,ALL
ESEL,S,TYPE,,2
ESEL,A,TYPE,,3
ESEL,R,REAL,,3
LSEL,S,REAL,,3
ESEL,ALL
ESEL,S,TYPE,,2
ESEL,A,TYPE,,3
ESEL,R,REAL,,3
LSEL,S,REAL,,3
CMSEL,A,_NODECM
CMDEL,_NODECM
CMSEL,A,_ELEMCM
CMDEL,_ELEMCM
CMSEL,S,_KPCM
CMDEL,_KPCM

249
CMSEL,S,_LINECM
CMDEL,_LINECM
CMSEL,S,_AREACM
CMDEL,_AREACM
CMSEL,S,_VOLUCM
CMDEL,_VOLUCM
CMDEL,_TARGET
CMDEL,_CONTACT
! /COM, CONTACT PAIR CREATION - END
FINISH

/SOL ! Entering solution phase to adjust contacts

ANTYPE,0 ! Static analysis
ALLSEL
SOLCONTROL, ON, ON, !Specifies to use optimized nonlinear solution defualts and some
enhaced internal solution algorithms
! The time step is adjusted automatically based on the non linearity due to
contact element
CNCHECK,ADJUST ! Adjust the initial status of the contact pair
NEQIT,46 ! Specifies the maximum no. of equilibrium for the nonlinear analysis (
Default is 15-25)
FINISH
/SOL ! Entering solution phase to solve the problem
ANTYPE,0 ! Static analysis
ALLSEL
SOLCONTROL, ON, ON,
NEQIT,96
NROPT,UNSYM
ACEL,,9.81 ! Set the constant for accelaration due to gravity i.e to include self-weight of the
model
RESCONTROL,DEFINE,ALL,-2,2 ! Restart control
TREF,273 ! Reference base temperature for thermal strain calculations

250
DK,1,UX, , , ! DOF of Node 1 constrained in x dirn.
*SET,LF,1 ! Set LF as 1 i.e the load factor(LF) that is used to multiply the loads
! LF is set to 1 for analysis with hydrostatic pressure and set to 5 (or
any suitable value for limit load analysis
!LOAD STEP 1
R,3,,,10,-.3,0,
SFL,1,PRES,0.11772*LF ! Apply pressure on line 1
SFL,2,PRES,0.11772*LF
SFL,4,PRES,0.11772*LF,0.10791*LF
SFL,5,PRES,0.10791*LF,0
BFL,4,TEMP,523 ! Apply temperature to line 4
BFL,1,TEMP,523
BFL,2,TEMP,523
BFL,3,TEMP,523
BFL,5,TEMP,523
OUTRES,ALL,ALL
/OUTPUT,OUT,DAT,,APPEND !APPENDS THE TEXT OUTPUT TO THE EXISTING
OUTPUT FILE
SOLVE ! Solve the routine
SAVE ! save the database with results

!LOAD STEP 2
BFL,4,TEMP,0 ! In the second load step the temp is brought down to zero , hence
apply zero temp
BFL,1,TEMP,0 ! at same locations
BFL,2,TEMP,0
BFL,3,TEMP,0
BFL,5,TEMP,0
OUTRES,ALL,ALL ! write all the solution data for all substeps in the database
/OUTPUT,OUT,DAT,,APPEND !Appends the text output to the existing output file
SOLVE
SAVE

251
APPENDIX B

This Appendix gives the ANSYS APDL Macro for performing nonlinear analysis of the
example tank given in Chapter 3-CASE 2 i.e. the tank with multiple wall thicknesses. The
load is hydrostatic pressure due to water with material properties specified in Ch.3. Plane
elements are used to model the tank wall and bottom plate. Friction interaction between the
tank bottom and the foundation is taken in to account using friction elements. The
information given after exclamatory mark (!) is for understanding of the user and is not part
of the APDL.

/PREP7 ! Enter the Preprocessor Module

ET,1,PLANE183 ! Define Element Type
KEYOPT,1,3,1 ! axisymmetric option
KEYOPT,1,1,0 ! indicates quad & not triangle element type
KEYOPT,1,6,1 ! Mixed U-P Formulation –Refer ANSYS help for more details
*SET,LF,2.0 ! Set a value of 2.0 to variable LF i.e here LF is meant for Load Factor
MP,EX,1,2e5 !Material Property – Youngs Modulus
MP,PRXY,1,.3 !Poissons ratio
MP,DENS,1,7.86e-6 ! Density
*SET,PM,1000 ! PM is the variable used for Plastic Modulus and in this example it is
given a value of 1000 MPa
*SET,YS,345 !Define Yield strength (YS) as 345 MPa
TB,BKIN,1,,,1 !Bilinear(see the option in ANSYS at Material properties)
TBDATA,,YS,PM !Here if Plastic modulus has no value it means elastic perfectly plastic
material model
! Providing key points for shell wall construction
k, 1 , 45016.99 , 0

252
k, 2 , 45000 , 0
k, 3 , 45000 , 2400
k, 4 , 45001.325 , 2400
k, 5 , 45001.325 , 4800
k, 6 , 45005.58000 , 4800
k, 7 , 45005.58 , 7200
k, 8 , 45007.76 , 7200
k, 9 , 45007.76 , 9600
k, 10 , 45010.26 , 9600
k, 11 , 45010.26 , 12000
k, 12 , 45011.99 , 12000
k, 13 , 45011.99 , 14400
k, 14 , 45011.99 , 16800
k, 15 , 45016.99 , 16800
k, 16 , 45021.99 , 16800
k, 17 , 45021.99 , 14400
k, 18 , 45021.99 , 12000
k, 19 , 45023.72 , 12000
k, 20 , 45023.72 , 9600
k, 21 , 45026.22 , 9600
k, 22 , 45026.22 , 7200
k, 23 , 45028.4 , 7200
k, 24 , 45028.4 , 4800
k, 25 , 45032.655 , 4800
k, 26 , 45032.655 , 2400
k, 27 , 45033.98 , 2400
k, 28 , 45033.98 , 0
!ANNULAR PLATE KEY POINTS
K, 29 , 45083.98 , 0
K, 30 , 45083.98 , -14
K, 31 , 40000 , -14

253
K, 32 , 40000 , -8
K, 33 , 44265 , -8
K, 34 , 44265 , 0
! EXTRA KEY POINTS TO DIVIDE AREAS TO OPTIMISE MESH
K, 35 , 43000 , -8
K, 36 , 43000 , -14
K, 37 , 45005.58 , 5000
K, 38 , 45028.4 , 5000
! CREATING LINES FOR SHELL CONSTRUCTION USING KEY POINTS
l, 2 , 3
l, 3 , 4
l, 4 , 5
l, 5 , 6
l, 6 , 7
l, 7 , 8
l, 8 , 9
l, 9 , 10
l, 10 , 11
l, 11 , 12
l, 12 , 13
l, 13 , 14
l, 14 , 15
l, 15 , 16
l, 16 , 17
l, 17 , 18
l, 18 , 19
l, 19 , 20
l, 20 , 21
l, 21 , 22
l, 22 , 23
l, 23 , 24

254
l, 24 , 25
l, 25 , 26
l, 26 , 27
l, 27 , 28
! CREATING BOTTOM & ANNULAR PLATE
l , 28 , 29
l , 29 , 30
l , 30 , 36
l , 31 , 32
l , 32 , 35
l , 33 , 34
l , 34 , 2
!EXTRA LINES
L,2,28
L,6,24
L,35,36
L,36,31
L,35,33
LPLOT
! CREATING AREAS FROM CONSTRUCTED LINES
AL,30,31,36,37 ! area 1
AL,27,28,29,36,38,32,33,34 !area 2
AL,1,2,3,4,35,23,24,25,26,34 !area3
lsel,s,line,,5,22,1 ! Selecting lines for creating area
lsel,a,line,,35
AL,all !area 4
AESIZE,1,2 ! Sizing the area for element formulation
AESIZE,2,1
AESIZE,3,2
AESIZE,4,6
MSHAPE,0,2D ! Instructing to form 2D Trapeziodal Element

255
MSHKEY,2
AMESH,ALL ! Mesh the areas to create Elements
!Creating Contact pair
K, 39 , 40000 , -14.1 ! Creating key points for constructing bottom rigid support
K, 40 , 45084.98 , -14.1
L,39,40 ! Creating Line 4 as bottom rigid support
! creating a entity " source"
NSEL,R,LOC,X,40000,45083.98 ! Node Selection
NSEL,R,LOC,Y,-14
CM,source,NODE ! Creating a module with Selected Nodes
Nsel,all
! /COM, CONTACT PAIR CREATION - START
CM,_NODECM,NODE
CM,_ELEMCM,ELEM
CM,_KPCM,KP
CM,_LINECM,LINE
CM,_AREACM,AREA
CM,_VOLUCM,VOLU
! /GSAV,cwz,gsav,,temp
MP,MU,1,0.2 !The Friction Coefficient is given here are 0.2
MAT,1
MP,EMIS,1,7.88860905221e-031
R,3
REAL,3
ET,2,169 ! Specifying Contact Elements
ET,3,172
R,3,,,10,-.09,0, ! Specify FKN here and the tolerance
RMORE,,,1.0E20,0.0,1.0,
RMORE,0.0,0,1.0,,1.0,0.5
RMORE,0,1.0,1.0,0.0,,1.0
KEYOPT,3,3,0

256
KEYOPT,3,4,0
KEYOPT,3,5,1
KEYOPT,3,6,2
KEYOPT,3,7,0
KEYOPT,3,8,0
KEYOPT,3,9,0
KEYOPT,3,10,2
KEYOPT,3,11,0
KEYOPT,3,12,0
KEYOPT,3,2,0 ! contact algorithm 0 - augmented lagran(default),1-penalty,4-pure
lagrangian
KEYOPT,2,2,0
KEYOPT,2,3,0
! Generate the target surface
LSEL,S,,,39 ! SPECIFY THE TARGET LINE HERE
CM,_TARGET,LINE
TYPE,2
LATT,-1,3,2,-1 ! Providing Line Attributes
TYPE,2
LMESH,ALL ! Meshing the Line for Contact element- Target Elements
! Generate the contact surface
NSEL,S,,,SOURCE
CM,_CONTACT,NODE
TYPE,3
ESLN,S,0
ESURF
*SET,_REALID,3
ALLSEL
ESEL,ALL
ESEL,S,TYPE,,2
ESEL,A,TYPE,,3

257
ESEL,R,REAL,,3
LSEL,S,REAL,,3
! /PSYMB,ESYS,1
! /PNUM,TYPE,1
! /NUM,1
! EPLOT
! Reverse target normals – since the normals of Source and Target contact elements should
meet
CM,_Y,LINE
LSEL,S,,,39 ! SPECIFY THE TARGET LINE HERE
CM,_YEL,ELEM
CM,_YND,NODE
NSLL,S,1
ESLN,S,1
ESEL,R,REAL,,_REALID
ESURF,,REVERSE ! Reversing the surface of elements i.e the normals
CMSEL,S,_Y
CMSEL,S,_YEL
CMSEL,S,_YND
CMDELE,_Y
CMDELE,_YEL
CMDELE,_YND
! /REPLOT
!*
ESEL,ALL
ESEL,S,TYPE,,2
ESEL,A,TYPE,,3
ESEL,R,REAL,,3
LSEL,S,REAL,,3
! /PSYMB,ESYS,1
! /PNUM,TYPE,1

258
! /NUM,1
! EPLOT
ESEL,ALL
ESEL,S,TYPE,,2
ESEL,A,TYPE,,3
ESEL,R,REAL,,3
LSEL,S,REAL,,3
CMSEL,A,_NODECM
CMDEL,_NODECM
CMSEL,A,_ELEMCM
CMDEL,_ELEMCM
CMSEL,S,_KPCM
CMDEL,_KPCM
CMSEL,S,_LINECM
CMDEL,_LINECM
CMSEL,S,_AREACM
CMDEL,_AREACM
CMSEL,S,_VOLUCM
CMDEL,_VOLUCM
! /GRES,cwz,gsav
CMDEL,_TARGET
CMDEL,_CONTACT
! /COM, CONTACT PAIR CREATION - END
! /MREP,EPLOT
FINISH
/SOLUTION
ANTYPE,0 ! Static analysis
NLGEOM,ON ! Include Nonlinear Geometry effect during Analysis i.e include Large
Strain/Rotation effect – See ANSYS help for more details
SOLCONTROL, ON, ON, ! Activate the default solution control techniques in ANSYS to
achieve convergence

259
CNCHECK,ADJUST ! Contact check and adjust for any small gaps
NROPT,FULL ! Emply full Newton Raphson technique to achieve convergence
neqit,35 ! Maximum no. of equilibrium equations to try for convergence is 35
ACEL,,9.81 ! Set gravity constant
DL,30,1,SYMM, ! UX=0 FOR LINE 30 ..LINE AT THE CENTRE OF TANK
!DL,30,1,UY, ! UY=0
LSEL,ALL ! Re-select all lines
! PRESSURE ON SHELL WALL MULTIPLIED BY LOAD FACTOR
SFL, 1 , PRES, 0.164808000*LF , 0.141264000*LF
SFL, 3 , PRES, 0.141264000*LF , 0.117720000*LF
SFL, 5 , PRES, 0.117720000*LF , 0.094176000*LF
SFL, 7 , PRES, 0.094176000*LF , 0.070632000*LF
SFL, 9 , PRES, 0.070632000*LF , 0.047088000*LF
SFL, 11 , PRES, 0.047088000*LF , 0.023544000*LF
SFL, 12 , PRES, 0.023544000*LF , 0.000000000*LF
SFL, 2 , PRES, 0.141264000*LF
SFL, 4 , PRES, 0.117720000*LF
SFL, 6 , PRES, 0.094176000*LF
SFL, 8 , PRES, 0.070632000*LF
SFL, 10 , PRES, 0.047088000*LF
!PRESSURE ON BOTTOM AND ANNULAR PLATE
SFL, 31 , PRES, 0.164886000*LF ,
SFL, 38 , PRES, 0.164808000*LF ,
SFL, 33 , PRES, 0.164808000*LF ,
SFL, 32 , PRES, 0.164886000*LF , 0.164808000*LF
TIME,1 !Time will go to 1 at the end of loading
!AUTOTS,ON ! This command is to instruct ANSYS for automatic time step increase. But
not used in this particular program since the solution is controlled by NSUBST command
NSUBST,5000 !Number of sub steps in total time is 5000
OUTRES,ALL,ALL ! write all results to database
/OUTPUT,OUT,DAT !Creates output files in the directory

260
SOLVE ! solve the model
SAVE ! save the database
FINISH ! close the solution module

261
APPENDIX C

The following is the post processing APDL routine incorporating the m-tangent method.
The routine has to be run after solving the FE model and results are stored in database. The
routine writes the m-tangent multiplier values in a file named “tank_mul” in the same
working directory of ANSYS.
The thickness of the shell element can be divided in to any number of layers (say 5 layers in
this routine) to improve the accuracy of results. Layering especially improves the results
where high bending strain is expected. But layering increases the computational effort and
the volume of results data. In this thesis layering was used in few situations to check the
accuracy of results obtained without layers (i.e., the entire shell thickness is a single layer).
The following routine can also be modified to run for shell elements without any layers by
removing the LAYER commands. Even though, the following macro has worked previously,
it is added here without final checking (using ANSYS), with latest version.

/POST1 ! Enter Postprocessing module

PLDISP,2 ! Plots the displaced structure
SHELL,MID ! choose the Middle plane of the shell layer
*SET,YS,206.8430 !YIELD STRENGTH
ETABLE,Sig,S,EQV !SIG is user given name, S,EQV is VON Misses Equivalent stress
ETABLE,Vol,VOLU !Element volume
!Counting total number of elements
*GET,Emax,ELEM,0,COUNT !Emax = No of total elements will be stored in Emax
!Evaluate total volume (sum of all element volume using do loop)
*SET,Vtot,0 !initial value of volume

*DO,i,1,Emax ! Starting the Do loop with i =1 to Emax with inc. of 1

*GET,Evol,ELEM,i,ETAB,Vol ! Get the volume of Each element inside the do-loop bracket
*SET,Vtot , Vtot + Evol ! summation of individual element volumes
*ENDDO

*SET,NLR,5 ! Enter No. of Layers here

262
!Evaluate the multipliers
*SET,SumRef,0 !initial value of Stress

*DO,j,1,NLR ! FINDING SUMREF FOR ONE LAYER

LAYER,j
*DO,i,1,Emax !i=1;i<=Emax;i++
*GET,SigEq,ELEM,i,ETAB,Sig
*GET,EvoL,ELEM,i,ETAB,Vol
*SET,SumRef , SumRef + ((SigEq*SigEq)*(Evol/5)) ! (SigEq*SigEq) =
(SigEq**2)
*ENDDO
*ENDDO

*SET,SIGref , SQRT(SumRef/Vtot) !Reference stress

*SET,m0_1 , (YS/SIGref)
*SET,m_0,m0_1
*SET,SIGmax,0 !initial value of Max.Stress

*DO,j,1,NLR
LAYER,j
ESORT,ETAB,Sig,0 !Getting max value of Sig from ETABLE
*GET,SMAX,SORT,0,MAX
*IF,SMAX,GT,SIGmax,THEN
*SET,SIGmax,SMAX
*ENDDO

*SET,m_L,(YS/SIGmaX)
*SET,JETA,(m_0/m_L)
*SET,Tan_theta,0.2929 !this is the fixed value shown in the formula of paper

263
 *SET,m_tangent , m_0/(1+(Jeta-1)*Tan_theta)

!Open file
*CFOPEN,tank_mul !tank_mul...file created
*CFWRITE,m0,m_0 !write m0=m_0 in the file
*CFWRITE,mL,m_L !write m0=m_0 in the file
*CFWRITE,m_tangent,m_tangent !Write m_tangent = m_tangent in file
*CFCLOS
FINISH

264

View publication stats