Leak Before Break Fracture Mechanics Thesis

The mechanism of Leak-Before-
Break fracture and its application in

Engineering Critical Assessment
A thesis submitted for the degree of

Doctor of Philosophy
By
Renaud Bourga
Department of Mechanical, Aerospace and Civil Engineering

Brunel University London, Uxbridge
United Kingdom
March 2017
Abstract
This thesis investigated the different aspects and mechanisms of leak-before-break

(LBB) assessment. The main objective was to improve the understanding of the tran-
sition between surface and through wall defects. While existing procedures generally
idealise the through-wall crack into a rectangular shape, in reality a crack propagates
with a shape depending on the loading. Comparison between the related solutions from
established procedures have been undertaken. The apparent variation depending on the
solutions used in the assessment has been highlighted. Two different methodologies have
been employed to investigate the transition of flaw: (i) non-ideal through-wall and (ii)
surface-breaking flaw propagation. The first approach consists of numerical models of
non-idealised flaws in order to assess the effect on LBB parameters. For the second ap-
proach, experiments have been first carried out to visualise the shape of defect growths.
To further study surface-breaking flaws, both experimental and numerical studies were
performed. Fatigue tests on deeply notched plates with two crack aspect ratios were car-
ried out. Strain evolutions on the back surface were recorded along the axes parallel and
perpendicular to the crack. Numerical models have been prepared to investigate a larger
scope. Behaviour of growing surface-breaking defects was examined. Based on the work
conducted in this research, the major findings can be summarised as follows:
• The existing solutions to carry out a LBB assessment using available procedures were
reviewed and discussed. For axial flaws, SIF solutions were found similar and in good
agreement with FEA values. Reference stress solutions showed significant difference
between BS 7910 and API 579-1/ASME FFS-1. When compared to experimental
data, API’s solutions were able to distinguish between leak and break cases.
• Flaw geometry assumption for through-wall crack yet to become idealised did not
always reflect the actual behaviour, especially for COA calculation. In this case,
FEA can be used as a good predictive tool for LBB to estimate margins when
assessing leak rate.
• The experiment using metallic specimens showed that high stress/strain on back
surface would provide a good estimate of the crack propagation as it approached
break-through. This offers a more accurate monitoring mechanism. Strain-mapping
devices such as gauges could be used.
i
Acknowledgements
Writing these acknowledgements marks the end of my adventure at TWI / Brunel

University London as a student.
I would like to express my gratitude to my Supervisor, Dr Bin Wang for his precious
guidance and help.
Next, I would like to thank Dr Yin Jin Janin for her excellent supervision, valuable
guidance throughout this research and for going through several drafts of my thesis! Her
support before being my industrial supervisor, during her maternity leave and up to the
end has been a source of comfort and inspiration and has made this research possible.
I would like to thank Dr Simon Smith for the really valuable discussions, for sharing
me his knowledge and answering my questions on FEA in general and experiment. His
support was greatly appreciated.
Furthermore, I would like to thank everyone that I had the pleasure to share the office
with for their continuous support during the period of my PhD studies. It was a great
time and I learned a lot from all of you. Thank you very much.
Lastly, and most importantly, I would like to thank my family for its continuous and
unconditional support.
ii
Contents
Abstract i
Acknowledgements ii
Symbols list vii
Abbreviations viii
1 Introduction 1
1.1 Background . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2
1.2 Research hypothesis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
1.3 Aims and objectives . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4
1.4 Outline of thesis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4
1.5 Contribution to new knowledge . . . . . . . . . . . . . . . . . . . . . . . . 6
1.6 List of publications . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6
2 Literature Review 7
2.1 Typical defects in pressurised equipment . . . . . . . . . . . . . . . . . . . 8
2.1.1 Fabrication induced defects . . . . . . . . . . . . . . . . . . . . . . 8
2.1.2 Environmentally induced defects . . . . . . . . . . . . . . . . . . . . 9
2.1.3 Third-party caused defects . . . . . . . . . . . . . . . . . . . . . . . 11
2.2 Non-Destructive Testing (NDT) . . . . . . . . . . . . . . . . . . . . . . . . 12
2.2.1 Conventional NDT (Surface/Embedded defects) . . . . . . . . . . . 13
2.2.2 Leak Testing (Through-wall defects) . . . . . . . . . . . . . . . . . 14
2.3 Fracture mechanics tools . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17
2.3.1 Crack behaviours . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17
2.3.2 Energy release rate, G (Energy-based) . . . . . . . . . . . . . . . . 20
2.3.3 Stress Intensity Factor, K (Stress-based) . . . . . . . . . . . . . . . 21
2.3.4 J-Integral (Energy-based) . . . . . . . . . . . . . . . . . . . . . . . 24
2.3.5 Crack-Tip Opening Displacement (CTOD) (Strain-based) . . . . . . 26
2.4 Fitness-for-Service and Engineering Critical Assessment . . . . . . . . . . . 27
2.4.1 Definitions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 28
2.4.2 Assessment procedures available and application . . . . . . . . . . . 29
2.4.3 Input data and assessment . . . . . . . . . . . . . . . . . . . . . . . 30
2.5 Pressure vessel design . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31
2.5.1 Stresses in cylindrical components . . . . . . . . . . . . . . . . . . . 32
2.5.2 Pressure vessel codes . . . . . . . . . . . . . . . . . . . . . . . . . . 34
2.6 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 36
iii
CONTENTS iv
3 Leak-Before-Break concept 38
3.1 Background . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 39
3.2 Definition and application . . . . . . . . . . . . . . . . . . . . . . . . . . . 40
3.2.1 Basic design formulation . . . . . . . . . . . . . . . . . . . . . . . . 40
3.2.2 Detailed procedures . . . . . . . . . . . . . . . . . . . . . . . . . . . 41
3.2.3 Limitations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 44
3.3 Leak-Before-Break Procedures . . . . . . . . . . . . . . . . . . . . . . . . . 45
3.3.1 Nuclear industry . . . . . . . . . . . . . . . . . . . . . . . . . . . . 45
3.3.2 Other industries (Non-nuclear) . . . . . . . . . . . . . . . . . . . . 47
3.4 Applying LBB to industries other than nuclear . . . . . . . . . . . . . . . . 48
3.5 Crack Opening Area . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 49
3.5.1 Analytical model . . . . . . . . . . . . . . . . . . . . . . . . . . . . 49
3.5.2 Leak rate and COA . . . . . . . . . . . . . . . . . . . . . . . . . . . 52
3.6 Past and future research in LBB . . . . . . . . . . . . . . . . . . . . . . . . 52
3.7 Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 54
3.8 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 56
4 LBB parameters - Analysis of current solutions from codes 57

4.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 58
4.2 Comparison of formulae for wall-thickness determination . . . . . . . . . . 59
4.2.1 Calculations of the required thickness . . . . . . . . . . . . . . . . . 59
4.2.2 Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 61
4.3 Comparison of widely used COA models . . . . . . . . . . . . . . . . . . . 62
4.3.1 Axial cracks . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 62
4.3.2 Circumferential cracks . . . . . . . . . . . . . . . . . . . . . . . . . 63
4.3.3 Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 65
4.4 Comparison of crack stability assessment solutions . . . . . . . . . . . . . . 66
4.4.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 66
4.4.2 Failure Assessment Diagrams . . . . . . . . . . . . . . . . . . . . . 66
4.4.3 Stress Intensity Factor solutions . . . . . . . . . . . . . . . . . . . 70
4.4.4 Reference Stress solutions . . . . . . . . . . . . . . . . . . . . . . . 71
4.4.5 Cases study . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 84
4.5 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 96
5 Numerical Methods - Investigation of Ideal Through-Wall Cracks in

Plates and Effect of Crack Shape in Cylindrical Components 99
5.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 100
5.2 Modelling methods - Benchmark using through-wall cracks in plates . . . . 101
5.2.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 101
5.2.2 Presentation of models . . . . . . . . . . . . . . . . . . . . . . . . . 102
5.2.3 Parameters investigated . . . . . . . . . . . . . . . . . . . . . . . . 104
5.2.4 Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 113
5.3 Investigation of crack shape influence (cylinders) . . . . . . . . . . . . . . . 115
5.3.1 Models transformation - From plate to cylinder . . . . . . . . . . . 115
5.3.2 An ideal axial through-wall crack in cylinders . . . . . . . . . . . . 116
5.3.3 Non-ideal axial through-wall crack in cylinders . . . . . . . . . . . . 126
5.3.4 Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 130
5.3.5 Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 135
5.4 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 137
CONTENTS v
6 Experimental Work - Stress-Strain Field on the Back Surface of Surface

Cracked Plate During Crack Transition 138
6.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 139
6.2 Semi-qualitative study of crack transition - Perspex specimens . . . . . . . 139
6.2.1 Preparation of Perspex samples . . . . . . . . . . . . . . . . . . . . 139
6.2.2 Test setup . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 141
6.2.3 Observations from tested plates . . . . . . . . . . . . . . . . . . . . 142
6.2.4 Summary of the test results . . . . . . . . . . . . . . . . . . . . . . 144
6.2.5 Crack shape re-categorization . . . . . . . . . . . . . . . . . . . . . 146
6.3 Evaluation of stress-strain field on back surface using high strength steel
specimens . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 147
6.3.1 Material properties . . . . . . . . . . . . . . . . . . . . . . . . . . . 148
6.3.2 Test specimens . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 149
6.3.3 Test arrangement . . . . . . . . . . . . . . . . . . . . . . . . . . . . 150
6.3.4 Static tests . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 151
6.3.5 Fatigue tests . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 154
6.3.6 Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 157
6.3.7 Observations from tested plates . . . . . . . . . . . . . . . . . . . . 160
6.4 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 165
7 Numerical Simulation - Stress-Strain Field on the Back Surface of Sur-

face Cracked Plate During Crack Transition 167
7.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 168
7.2 Models description . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 168
7.2.1 Plate and cracks geometry . . . . . . . . . . . . . . . . . . . . . . . 168
7.2.2 The effect of crack geometry on back surface stress field . . . . . . . 169
7.2.3 Mesh sensitivity study . . . . . . . . . . . . . . . . . . . . . . . . . 170
7.3 Validation against experimental data . . . . . . . . . . . . . . . . . . . . . 174
7.3.1 Model geometries and results . . . . . . . . . . . . . . . . . . . . . 174
7.3.2 Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 175
7.4 Stress analyses along axes parallel and perpendicular to crack . . . . . . . 178
7.4.1 Stress analyses along X-axis . . . . . . . . . . . . . . . . . . . . . . 179
7.4.2 Stress analyses along Y-axis . . . . . . . . . . . . . . . . . . . . . . 184
7.4.3 Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 188
7.5 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 190
8 Concluding remarks and future work 191

8.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 192
8.2 Summary and Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . 192
8.3 Main PhD Achievements Findings . . . . . . . . . . . . . . . . . . . . . . . 196
8.4 Recommendations for future work . . . . . . . . . . . . . . . . . . . . . . . 197
Bibliography 199
A Complementary photographs and plots from experiments 211

A.1 Fracture surface from Perspex samples . . . . . . . . . . . . . . . . . . . . 212
A.1.1 Series 1 (10x50x100mm) . . . . . . . . . . . . . . . . . . . . . . . . 212
A.1.2 Series 2 (15x50x100mm) . . . . . . . . . . . . . . . . . . . . . . . . 214
A.1.3 Series 3 (10x75x100mm) . . . . . . . . . . . . . . . . . . . . . . . . 216
CONTENTS vi
A.1.4 Series 4 (15x75x100mm) . . . . . . . . . . . . . . . . . . . . . . . . 218

A.2 Recorded strain data from metallic specimens . . . . . . . . . . . . . . . . 220
A.2.1 Static test . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 220
A.2.2 Fatigue test . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 222
A.3 Photographs of high strength steel specimens . . . . . . . . . . . . . . . . . 227
A.3.1 Fracture surface . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 227
A.3.2 Re-characterisation into through-wall crack . . . . . . . . . . . . . . 228
B Equations for crack opening area, stress intensity factor and reference
stress 229
B.1 Equations of widely used COA models . . . . . . . . . . . . . . . . . . . . 230
B.1.1 Axial cracks . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 230
B.1.2 Circumferential cracks . . . . . . . . . . . . . . . . . . . . . . . . . 233
B.2 Stress Intensity Factor solutions . . . . . . . . . . . . . . . . . . . . . . . . 234
B.2.1 Axial through-thickness flaws . . . . . . . . . . . . . . . . . . . . . 234
B.2.2 Circumferential through-thickness crack . . . . . . . . . . . . . . . 236
B.2.3 Axial surface flaws . . . . . . . . . . . . . . . . . . . . . . . . . . . 236
B.2.4 Circumferential surface defect . . . . . . . . . . . . . . . . . . . . . 237
B.3 Reference Stress solutions . . . . . . . . . . . . . . . . . . . . . . . . . . . 238
B.3.1 Axial through-thickness flaws . . . . . . . . . . . . . . . . . . . . . 238
B.3.2 Circumferential through-thickness crack . . . . . . . . . . . . . . . 239
B.3.3 Axial surface flaws . . . . . . . . . . . . . . . . . . . . . . . . . . . 239
B.3.4 Circumferential surface defect . . . . . . . . . . . . . . . . . . . . . 240
Symbols list
Symbol Name Unit
a Crack depth (surface crack) / Half crack length (through-wall mm

crack)
c Half crack length (surface crack) mm
C Fatigue parameter —
Di Internal diameter mm
Dm Mean diameter mm
Do External diameter mm
E Young modulus MPa
E Weld efficiency factor — √
J Rice’s J-Integral M P a.√m
KI , KII , KIII Stress Intensity Factor mode I, II and III M P a. m
Kr Fracture ratio —
Lr Load ratio —
N Number of cycles (Fatigue) Cycles
P Pressure MPa
Pb Primary bending stress MPa
Pm Primary membrane stress MPa
Ri Internal radius mm
Rm Mean radius mm
Ro External radius mm
S Allowable stress MPa
SF Safety Factor —
t Thickness mm
W Width mm
w Strain energy density —
δ Crack Tip Opening Displacement mm √
∆ Stress Intensity Factor range M P a. m
ν Poisson ratio —
σf Flow stress MPa
σref Reference stress MPa
σuts Ultimate Tensile Stress MPa
σys Yield Stress MPa
vii
Abbreviations
Symbol Name
CFP Crack Face Pressure

CMOD Crack Mouth Opening Displacement
COA Crack Opening Area
CSA Cross Section Area
CTOD Crack Tip Opening Displacement
ECA Engineering Critical Assessment
EPFM Elastic Plastic Fracture Mechanics
FAD Failure Assessment Diagram
FEA Finite Element Analysis
FFS Fitness For Service
LBB Leak-Before-Break
LEFM Linear Elastic Fracture Mechanics
RS Reference Stress
SF Safety Factor
SG Strain Gauge
SIF Stress Intensity Factor
UTS Ultimate Tensile Stress
YS Yield Stress
viii
Chapter 1
Introduction
Contents
1.1 Background . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2
1.2 Research hypothesis . . . . . . . . . . . . . . . . . . . . . . . . . 3
1.3 Aims and objectives . . . . . . . . . . . . . . . . . . . . . . . . . 4
1.4 Outline of thesis . . . . . . . . . . . . . . . . . . . . . . . . . . . 4
1.5 Contribution to new knowledge . . . . . . . . . . . . . . . . . . 6
1.6 List of publications . . . . . . . . . . . . . . . . . . . . . . . . . 6
1
CHAPTER 1. INTRODUCTION 2
1.1 Background
Different industries (i.e. oil and gas, nuclear, process, etc) commonly use pressurised
components for static or circulating fluids. It may be an isolated component or a complex
system consisting of a number of different engineered structures. With an ever increasing
demand in energy coupled with potential economical restraint, these structures tend to be
used up to their maximum capacity. For this reason, increasingly demanding operating
conditions and extended service life are sought for pressurised components.
A good example is the oil and gas sector. Resources are becoming more and more
difficult to extract. They need to be transported and processed as quickly as possible.
Rising demands have pushed suppliers to extract resources from deeper source and often
require extensive refinement process/treatments. Thus, more intensive operations are
required with structures working under harsher conditions.
Numerous countries, such as USA or France, began constructing their nuclear power
plants (NPPs) in 1960’s - 1970’s. These countries are now planning for large-scale refur-
bishment programs to extend their operations for another 20-30 years. This has increased
the need to assess the structural integrity of these components beyond their initial design
life. Life extension is used to bridge the gap between ageing and new plants. Extending
the operating life of existing NPPs will help to reduce the short term need for new gen-
erating capacity without new capital costs. However, these extensions must take place
with careful safety analysis and monitoring of equipment of ageing concerns. Ageing in
NPPs must, therefore, be efficiently assessed to ensure the availability of design functions
throughout the extended service life. It is crucial to demonstrate that adequate safety
margins (i.e. integrity and functional capability) remain in excess of minimum safety
requirements.
Fracture mechanics-based structural integrity assessments, most commonly referred

to as either Engineering Critical Assessments (ECAs) or Fitness-For-Service (FFS) as-
sessments, are not a new concept. Established since the 1960s, they have found more
widespread acceptance over the years. The nuclear and offshore oil and gas industries
were the main drivers behind the development of them. BS 7910 [1], a British Standard
focusing on the assessment of brittle fracture and fatigue, offers three levels of assessment
for fracture, based on the failure assessment diagram (FAD) concept.
Structural integrity assessments can be used at the design stage to: (i) estimate the
maximum flaw size that will not grow to an intolerable size during the life of the compo-
nent, or (ii) assess defects that have grown after some time in service. Such task requires
information about defect tolerance, which itself relies on the availability of representative
and reliable experimental data. For structures operating in inert environments, primary
material properties that are needed for this type of assessment are tensile properties,
fracture toughness and fatigue crack growth rate (FCGR).
A typical requirement for average component would be to stipulate the maximum

allowable flaw size lasting two design service lives (DSL), or twice the expected life of
the component. However, in the case of pressure vessels, a further requirement is added
that the body must be able to leak for 0.5 DSL before failure, giving the maintenance
team sufficient time to notice the leak and replace the part. Satisfying this leak before
break (LBB) requirement means placing a through-wall flaw in any location receiving a
damage tolerance inspection and ensuring that a crack with a depth equal to the wall
thickness of the pressure vessel and a length of twice the wall thickness will survive for
the required number of cycles. LBB concept essentially demonstrates through fracture
mechanics analysis there is negligible chance of any catastrophic break without giving
earlier indication of leakage.
One area where there is still significant uncertainty is the behaviour of deep flaws and
their transition from part-penetrating to through thickness. It is known that there can
be substantial uncertainty in calculating leak rates from crack opening area, and often
safety factors are used to account for this. The work described in this thesis serves the
purpose to try to reduce this uncertainty. The methods of assessing these flaws (with a
more accurate crack shape) is crucial and the main motivation for this research.
1.2 Research hypothesis
Traditional crack evaluation uses conservative methods to predict maximum loading

that a flawed component can sustain. When performing an analysis, it is appropriate to
use the current standards and assume an idealized flaw shape of either constant depth (i.e
through-wall) or semi-elliptical. However, when cracks grow from surface to through-wall,
their shapes are more complex than the ideal models assume. These cracks can grow as
very long surface cracks before breaking through the surface and complex irregular shape.
Deterministic LBB assessments often consider an already through-wall cracked compo-
nent. This configuration is often assumed to be ideal as the solutions available have been
derived for these geometries. However, the existing procedures do not assess components
with active degradation mechanisms such as fatigue, while some have been granted for
LBB such as nuclear piping systems affected by Primary Water Stress Corrosion Cracking
(PWSCC). An understanding of the effect of crack shape hypothesis can help reducing
safety margins applied and saving significant resources by not having to develop complex
FE models. This research determines the actual shape of flaws breaking through the wall
and compares their behaviour in terms of leak and break to the predictions assuming an
”ideal” through-wall flaw.
1.3 Aims and objectives
The overall purpose of this research project is to improve fracture assessment methods
for leak-before-break procedures available in BS 7910 (2013) or API 579-1/ASME FFS-1
(2007) for non-nuclear industries. Through this research it is intended that the safety and
integrity assessment of pressurised components could be improved and it would help to
optimise operations. This aim is achieved in this thesis by performing the following:
• Review available integrity assessment / leak-before-break procedures available;

• Determine the best available solutions up to date to carry out a full assessment;
• Provide guidance on performing LBB calculations and improved and/or revised
expression for surface breaking or existing through-thickness flaws in pressurised
components;
• Give an increased knowledge of the behaviour of cracked components during LBB
transition.
1.4 Outline of thesis
This thesis is divided into eight chapters including this introductory chapter.
Chapter 2 focuses on literature review. Different types of defects that can be present in
metallic structures, especially pressurised components are reviewed and discussed (Section
2.1). Conventional and leak-related non-destructive testing (NDT) methods are discussed
in order to determine how FFS input parameters are obtained (Section 2.2). These ex-
isting defects (i.e. cracks), once detected and characterised, are assessed using fracture
mechanics (Section 2.3). The assessment of the criticality of a defect is performed using
fitness-for-service procedures (Section 2.4). The basis of Fitness-For-Service and associ-
ated advantages are reviewed. Since leak-before-break concept is mainly applicable for
pressurised components, a review of related solutions and design codes is discussed in
Section 2.5.
Chapter 3 focuses on the literature review on Leak-Before-Break. Its underlying basic

principles are discussed (Section 3.2) and existing procedures are reviewed (Section 3.3).
Application in industries is discussed (Section 3.4). Parameters specific to this concept
(i.e. Crack Opening Area (COA)) are evaluated (Section 3.5). Review of work carried
out in the last 30 years and scope for future work is provided in section 3.6.
Chapter 4 focuses on the analysis of current LBB solutions from procedures. As this
concept can be applied at either design or in-service stage, both design codes and FFS
procedures have been reviewed. A short discussion on the effect of pressure vessel design
codes on the required minimum thickness is given in Section 4.2. Leak-Before-Break
assessment can be divided into two separated analyses, namely crack stability and leakage
analysis. These two analyses require calculations of different parameters. The solutions
for Crack Opening Area models (Section 4.3), Failure Assessment Diagram (Section 4.4.2),
Stress Intensity Factors (Section 4.4.3) and Reference Stress (Section 4.4.4) solutions have
been examined. Existing solutions are compared to each other. When possible, they are
compared against experimental data from literature.
Chapter 5 presents the numerical work carried out on through-thickness defects. Both
plate and cylindrical component geometries have been evaluated. The main objective of
this chapter is to investigate the flaw geometrical parameters and validate current FE work
with benchmark cases (Section 5.2). Different modelling strategies have been employed
and the validation of existing solution is included. Finite element work undertaken to
assess the effect of crack idealisation in cylinders is provided in section 5.3. Investigation
on stress intensity factor and crack opening area for ’short’ and ’long’ crack has also been
accounted for in this chapter. The effect of crack face pressure on existing solutions is
also discussed.
Chapter 6 presents the experimental work undertaken. Effort to assess crack shape
evolution when a crack propagates through wall is given in section 6.2. As a qualita-
tive approach, plates made of Perspex have been studied. Four geometries have been
considered (2 widths / 2 thicknesses). This experiment provides an insight of the trans-
formation of a surface crack of 80% through-thickness when it is subject to fatigue loading
(i.e. break-through). Using the experience gained from Perspex experiments, tests have
been reproduced with metallic specimens (Section 6.3). Two different notch geometries
have been considered to include ’short’ and ’long’ c racks. Stress/strain field evolution
on the back surface of a surface cracked plate has been investigated. A set of strain
gauges on the back surface recorded the strain evolution during the transition from sur-
face to through thickness crack. Further discussion was made on the fracture surface from
specimens.
Chapter 7 covers the numerical work undertaken to examine the transition from surface-
breaking to through-thickness crack. Experimental work validation from the static tests
was done in section 7.3. The evolution of stresses on the axis of the crack and the axis
perpendicular to the crack were assessed (Section 7.4). Investigation of the stress field on
the back surface of a surface-cracked plate is done for crack depth ranging from 20% to
95% thickness. Models were prepared in order to increase the range of observations from
experiments.
The final conclusions, summary of the findings and highlights of future research topics
related to this research is presented in Chapter 8. Two appendices are provided to in-
clude the complementary files from experiment (App.A) and a compendium of solutions
compared in Chapter 4 (App.B)
1.5 Contribution to new knowledge
The main contributions from this work are:
• Comparison of the different LBB available procedures and highlight of the short-
comings when they are applied to non-nuclear industries. Solutions for SIF, RS
and COA from literature for different parameters are discussed. The best available
solutions are highlighted when compared to experimental data;
• Validation of existing solutions with FE models of plates and cylinders containing
through-thickness defects for a wide range of geometries. The effect of crack shape
idealisation in cylinders with axial through-wall defect using numerical model results
(both ’short’ and ’long’ cracks inclusive) was investigated;
• Analysis of stress/strain field on the back surface of a plate during the crack growth
to provide a better understanding of crack transition from surface to through-
thickness. Experimental observation on shape evolution from deep surface-breaking
crack to through thickness crack in Perspex and metallic specimens;
• Draft improved LBB procedure for BS 7910 Annex F based on research findings.
1.6 List of publications
- R. Bourga, P. Moore, Y.-J. Janin, B.Wang, and J. Sharples. Leak-before-break: Global

perspectives and procedures. International Journal of Pressure Vessels and Piping, 129-
130:43-49, 2015.
- R. Bourga, P. Moore, Y.-J. Janin, B.Wang. Comparison of BS 7910:2013 and API

579-1 / ASME FFS-1 solutions with regards to Leak-Before-Break. In Proceedings of the
ASME 2016 Pressure Vessels & Piping Conference, Vancouver, 17-21 July 2016. Paper
No.PVP2016-63876
- R. Bourga, P. Moore, Y.-J. Janin, B.Wang. The effect of crack shape idealisation
on Leak-Before-Break assessment. In Proceedings of the ASME 2016 Pressure Vessels &
Piping Conference, Vancouver, 17-21 July 2016. Paper No.PVP2016-63877
- R. Bourga, Y.-J. Janin, B.Wang. Transition from part-penetrating to through thick-

ness cracks: experimental and numerical study. International Journal of Pressure Vessels
and Piping (Submitted on 02/04/2017)
Chapter 2
Literature Review
Contents
2.1 Typical defects in pressurised equipment . . . . . . . . . . . . 8
2.1.1 Fabrication induced defects . . . . . . . . . . . . . . . . . . . . 8
2.1.2 Environmentally induced defects . . . . . . . . . . . . . . . . . 9
2.1.3 Third-party caused defects . . . . . . . . . . . . . . . . . . . . 11
2.2 Non-Destructive Testing (NDT) . . . . . . . . . . . . . . . . . 12
2.2.1 Conventional NDT (Surface/Embedded defects) . . . . . . . . 13
2.2.2 Leak Testing (Through-wall defects) . . . . . . . . . . . . . . . 14
2.3 Fracture mechanics tools . . . . . . . . . . . . . . . . . . . . . . 17
2.3.1 Crack behaviours . . . . . . . . . . . . . . . . . . . . . . . . . . 17
2.3.2 Energy release rate, G (Energy-based) . . . . . . . . . . . . . . 20
2.3.3 Stress Intensity Factor, K (Stress-based) . . . . . . . . . . . . . 21
2.3.4 J-Integral (Energy-based) . . . . . . . . . . . . . . . . . . . . . 24
2.3.5 Crack-Tip Opening Displacement (CTOD) (Strain-based) . . . 26
2.4 Fitness-for-Service and Engineering Critical Assessment . . . 27
2.4.1 Definitions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 28
2.4.2 Assessment procedures available and application . . . . . . . . 29
2.4.3 Input data and assessment . . . . . . . . . . . . . . . . . . . . 30
2.5 Pressure vessel design . . . . . . . . . . . . . . . . . . . . . . . . 31
2.5.1 Stresses in cylindrical components . . . . . . . . . . . . . . . . 32
2.5.2 Pressure vessel codes . . . . . . . . . . . . . . . . . . . . . . . . 34
2.6 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 36
7
CHAPTER 2. LITERATURE REVIEW 8
2.1 Typical defects in pressurised equipment
Any metallic structure can contain defects that are introduced during the manufacture
of the material, during the fabrication of the piece (i.e. defects of machining or welding)
or in service. Pipes and pressure vessels are examples of such structures. They are widely
used for fluid transport or storage in various environments. However, they are also safety-
critical structural elements of which the risk of a malfunction cannot be accepted. In
the case of pipes, their increased diameters and operating pressures increase the risk of
fracture initiated from defects. The presence of a defect coupled with the effect of the
loading (i.e. internal pressure) can cause a localized fracture inducing leakage or in the
worst case a general fracture or explosion. Defects can be divided into three general
categories: manufacturing defects, service induced defects or defects caused by external
forces.
2.1.1 Fabrication induced defects
Numerous defects can be introduced in components during fabrication, especially weld-

ing [2]. There are many different welding processes, thus different types of defect can be
introduced depending on the methods and the materials. Physical discontinuities around
a weld bead may also alter the local stress distribution and metallurgical differences be-
tween the parent metal and weld metal affect the mechanical and chemical properties of
the weld joint. Thermal gradients from fusion welding give rise to a range of microstruc-
tures and properties that varies from the parent material through the heat-affected zone
(HAZ) associated with the weld. The different types of possible discontinuity may be
categorized as:
• Porosity: More commonly in the form of spherical defects, some gas inclusions
may have elongated forms (wormholes).
• Inclusions: These are external solid particles that may be introduced during the
welding, casting and rolling processes.
• Lack of fusion or incomplete root penetration: Lack of complete melting and
fusing between the molten weld bead and other parts of the joint. The atom-to-atom
bond is incomplete reducing the effective contact area carrying the load.
• Cracks: Although many cracking mechanisms exist for many different materials,
the most common types of cracks in steel welds are hydrogen cracks, sometimes
called cold cracks. This common mechanism can form cracks in several different
orientations and locations; transverse in high strength weld metal, longitudinal along
the HAZ or weld root, or crater cracks at the end of a weld run . Figure 2.1 presents
a view of different locations and orientations of cracks which can occur in a weld.
After fabrication, welds are inspected to ensure that no unacceptable defects remain
in the welds. This is usually done 24 or 48 hours after welding, since hydrogen cracks can
occur up to several hours after cooling. This work considers the subsequent degradation
that occurs in-service in piping and pressure vessels, whether or not that occurs as a leak
or as a consequence of catastrophic failure.
Figure 2.1: Different types of cracks located in and around a welded T-joint [2]
2.1.2 Environmentally induced defects
Defects can develop in components depending on the operating conditions, such as the
loads imposed (static or cyclic), service temperatures, locations of components, etc. Three
types of environmentally induced defects are discussed here: corrosion, fatigue and fluid-
related defects. These defects have been widely observed in pressurised components.
2.1.2.1 Corrosion
Corrosion is the degradation of a metal and its properties because of a reaction with
the surrounding environment. It is a significant factor in the failure and damage of metals,
including pressurized components [3]. Due to the large amount of metals and applications,
many factors contribute to the corrosion process [4]. The nature of corrosion products,
the rate of corrosion and the mechanism can vary widely depending on the environment
[5]. The full complexities of corrosion engineering (general corrosion, galvanic corrosion,
erosion corrosion, fatigue corrosion, pitting corrosion and stress corrosion cracking (SCC))
are beyond the scope of this thesis but the last two are of the most significance to the
risk of leak and/or break in pressure equipment [6]:
• Pitting corrosion is localized and can result in small holes on the surface of a metal.
The metal loss is negligible compared to the total mass of metal of the component
but it can be rapid. Nevertheless, equipment failure by perforation, particularly
where a large number of pits occur in a localised region is the usual outcome of
pitting corrosion.
• Stress Corrosion cracking (SCC) is a localized phenomenon and occurs mostly when
residual stress remains high [6]. When high stresses are applied to metals in a
corrosive environment, cracking can be accelerated in the form of SCC. A way to
reduce this effect by redistributing the load and reducing the residual tensile stress
is post-weld heat treatments [7]. In industrial plants, SCC is responsible for a third
of all damage due to corrosion [6]. The average age to encounter a first SCC incident
in gas transmission pipelines has been evaluated after 20-25 years of service life [4].
2.1.2.2 Damage related to the fluid
The interaction between the fluid in movement and the structure can lead to damage
mechanisms. This failure-related phenomenon depends on the operating conditions and
the fluid conveyed. Two phenomenon will be discussed here:
• Erosion is an undesired removal of material due to the action of the fluid on the
component-wall. In service, the friction due to the contact between the fluid in
movement and the static component wall may lead to an erosion of the wall, also
known as wall-thinning. Another aspect may be noted when a leak occurs. Erosion
of the area surrounding the leak site also increases the crack opening area.
• Water hammer is a phenomenon that takes place when a sudden change in flow
velocity occurs resulting in a sudden increase in load and possible negative pressure
in a very short time. It may lead to other degradation mechanisms such as fatigue.
2.1.2.3 Fatigue
Another mechanism that can lead to failure of pressurized components is fatigue. When
components are subjected to repeated changes in their loading conditions, fatigue cracks
will initiate at locations of stress concentration, often weld toes. These cracks progressively
propagate over many thousands of load cycles until they can cause failure at stresses below
the material’s yield strength with no sign of plastic deformation. The number of cycles
leading to fracture at a given stress is generally referred to as fatigue strength. This
phenomenon begins generally from a crack that continues to grow as a result of the stress
fluctuations to its critical length. Fatigue is greatly influenced by the type of alloy, and
the magnitude and frequency of loading cycles. Different models have been developed
to assess fatigue based on fracture mechanics, from the simplest, such as the well-known
Paris’ law (Eq.2.1):

da
= C × (∆K)m (2.1)
dN
to the most complex such as the NASGRO equation (Eq.2.2) which incorporates the
effects of fatigue crack closure:
n ∆Kth p

1−

da 1−F ∆K
=C× × q (2.2)
dN 1−R 1− Kmax
KIc
In these models, the crack increment, a, per cycle N , is expressed as a function of the
stress intensity factor range ∆K.
2.1.3 Third-party caused defects
Third-party caused defects may be unforeseeable in the general design and operation
of equipment, but may need to be assessed after the event for their impact on structural
integrity. During installation or maintenance, components may be subjected to an exter-
nal mechanical impact. Two mechanical defects are summarized here: dents and gouges
depending on whether there is material removal or not.
• Dents are a permanent plastic deformation due to an impact. In case of pressurized

components, a dent is a change of the curvature in the wall without changing wall-
thickness. Dents are distinguished as plain or kinked depending if there is a change
in wall thickness or not [8]. It causes local concentration of stress and strain and a
local reduction of the pipe diameter.
• A gouge or scratch is a superficial surface damage due to contact with a foreign
object causing material removal. The dimensions of a scratch are defined so that the
length is greater than the width. Figure 2.2 presents an illustration of a combination
of a dent and a gouge, which can be resulting from an external impact.
Figure 2.2: Illustration of a dent and gouge combination [9]

2.2 Non-Destructive Testing (NDT)
Non-Destructive Testing (NDT) is the inspection of materials without causing any

subsequent damage to the material. In contrast, destructive testing (involving the extrac-
tion of test specimens for mechanical or metallurgical testing) usually makes testing more
costly and is also inappropriate in many circumstances once a component is in service.
Discontinuities as surface or internal (embedded) defects, discussed in Section 2.1, have
to be evaluated to ensure product integrity and reliability. During manufacturing, NDT
methods are applied to ensure the quality of materials and joining processes before the
component is used. In-service, NDT inspections are planned at regular intervals during
service to ensure that the components still exhibit the integrity necessary to be safe. In
order to make a Leak-Before-Break case, NDT techniques have an important application
to detect defects as early as possible and characterize their geometries. No single method
will satisfy all requirements of sensitive flaw detection and accurate geometry measure-
ment), each of the NDT methods has advantages and disadvantages when compared to
other methods. Figure 2.3 provides some guidance on the selection of NDT methods for
common flaw detection and measurement applications [10], [11].
Figure 2.3: Comparison of NDT methods available for different defect type [10],[11]
Having in mind that Leak-Before-Break procedures are of interest here and will be
discussed in detail, both penetrating and through-wall defects will be evaluated. The first
part focuses on partially-penetrating defects. Methods to detect and size both surface,
sub-surface or embedded defects are reviewed. These methods are generally known as
conventional NDT methods. The leak testing method for through-wall defects is a collec-
tion of different methods which can be considered more as proof testing or performance
testing, although they are non-destructive in their application. These methods, specific
to particular industries using pressurized components, are generally less commonly known
as NDT techniques.
2.2.1 Conventional NDT (Surface/Embedded defects)
According to pressure vessel design codes such as BS PD 5500 [12], Radiographic

Testing (RT) / Ultrasonic Testing (UT) should be used for examination of embedded flaws
and Penetrant Testing (PT) / Magnetic Particle Testing (MT) for surface breaking flaws.
Numerous standards exist for each NDT method. Relevant ASME and BS standards
are given in the title of each part. Further guidance can also be found in Section V of
ASME Boiler and Pressure Vessel Code [13]. However, before applying one of these NDT
techniques, the first approach is to simply look at the component and visually examine
the locations of concern for evidence of damage.
• Visual Testing (VT): VT is done to check the surface condition (for evidence of
corrosion products, weld appearance and fusion, coating condition etc), the shape
of the component (and the presence of dents or gouges) or evidence of leakage. Only
surface-breaking defects can be detected (and access to the surface is required) and
sizing the crack geometry is not easy.
• Penetrant Testing (PT) ([13]-6, [14], [15]): PT can be used for the detection
of discontinuities that are open to the surface such as cracks, seams, laminations,
through leaks or lack of fusion in non-porous materials. However the surfaces tested
must be free of all contaminants (dirt, oil, grease, paint, rust, etc.). This is a good
alternative method to the VT for large components
• Magnetic Particle Testing (MT) ([13]-7, [16], [17]): MT can be used to
locate surface and near-surface discontinuities in ferromagnetic materials. Surface
and subsurface flaws approximate size and shape can be detected rapidly on large
surface areas of complex parts. Surface preparation is less critical than it is in
PT. The orientation and strength of magnetic field is critical, a proper alignment
of magnetic field and defect is necessary. Large currents are needed for very large
parts.
• Ultrasonic Testing (UT) ([13]-4/5, [18], [19]): Ultrasonic testing has a high
accuracy to position and size embedded as well as surface defects. It has a higher
penetrating power than RT which allows examination of extremely thick sections
(up to about 7 meters of steel) and its fast response permits rapid and automatic
inspection. Moreover, access to only one surface of the specimen is needed. However,
the geometry of the part being tested, the defect orientation or material properties
affects the detectability (i.e. materials having coarse grain microstructure such as
austenitic stainless steel can be difficult to inspect).
• Radiographic Testing (RT)([13]-2, [20]): Radiographic testing is used for the
detection of surface breaking and embedded flaws in many different materials. Sur-
face preparation may be necessary to avoid confusion with discontinuities. Moreover,
planar defects and the depth of defect cannot be detected easily.
2.2.2 Leak Testing (Through-wall defects)
Leak testing methods group together a number of conventional and unconventional

NDT techniques to detect, locate and size leaks. A leak is generally due to an unintended
crack, hole or porosity in an enveloping wall which must contain fluids or gases. A section
about leak testing is present in most pressure vessel codes as quality control step in terms
of hydrostatic or pneumatic pressure test where the pressure vessel is pressurized above its
design pressure in order to assure that there is no leak. Many modern designs are based
on the “leak before break” principle, in which critical thickness and other dimensions
are selected so that a cracked vessel will leak before the crack grows to catastrophic
proportions. In these applications, the ability to detect the leak at a very early stage is
therefore important. Numerous methods are available in the field of leak testing. However,
due to the huge range of methods for the determination of the leakage rate, it is often
not easy to select the right method. Some standards give advices for the selection of
the appropriate leak testing technique and for calibration (i.e ASTM E432 [21] and BS
EN 1779 [22]). Figure 2.4 reviews some of these methods with their sensitivity. Some
general guidance are given in Section V of ASME Boiler and Pressure Vessel Code [13]
Article 10. The surfaces to be tested have to be as clean as possible in order to avoid
leaks clogging. Pressure gauges range shall be not less than 1/2 or more than 4 times
the maximum pressure. All openings have to be sealed and components to be tested
shall not be pressurized at more than 25% of the maximum design pressure if no other
indications are specified. As seen previously, visual examination is a basic non-destructive
test to detect surface defects. It can also be used for leaks detection. Prior to employing
a sensitive leak testing method, a preliminary test to detect and eliminate potential large
leaks may be useful to perform in order to avoid missing small leaks.
Figure 2.4: Sensitivity ranges of various leak detection methods [23] [24]
• Liquid penetrants and chemical tracers ([13]-6, [2]): Although this test is
time consuming, it requires no significant equipment and no pressure differential is
necessary. It is easy to apply and the results are visual. However, access to both
sides of the envelope is necessary and it may require extensive cleaning before and
after the test. It cannot be applied while the component is in operation.
• Ultrasonic leak testing ([13]-10, [25]): Ultrasonic gas leak detector are used to
detect the ultrasonic energy produced by the flow of a gas leaking through a pressure
boundary. Mainly used for outdoor environments where weather conditions can
easily dissipate the escaping gas before allowing it to reach more conventional gas
leak detectors. This simple, fast and non-contacting method is useful for locating
and estimating the size of pressurized gas leaks in service. However, due to the low
sensitivity of this technique, it should not be utilized for the acceptance testing of
vessels that will contain lethal or hazardous substances [13], [25]. Sensitive to other
environmental noises, it is qualified as a semi-quantitative method used to detect
and locate leaks. It may be used as a pre-test or monitoring technique before other
more time consuming and sensitive leak test methods are employed.
• Acoustic Emission (AE) leak testing ([13]-12/13, [26], [27], [28], [29]):
Acoustic emission leak detection detects high frequency sound waves travelling
within the wall thickness of the component. Using this method it is possible to
detect, locate, and characterize AE sources to evaluate their significance relative to
integrity. It is the most used method in the water industry to locate leaks. AE test-
ing is considered to be more sensitive than ultrasonic leak testing, but does require
access to the component surface.
• Testing by means of pressure ([13]-10, [30]): Another common method is
the observation of pressure change that can reveal leakage of a pressurized compo-
nent.The test object is initially pressurized to establish a reference pressure. After a
stabilizing period for temperature and moisture equalization (not less than 1h [30]),
the pressure is monitored again. As this method tends to detect only large leaks,
this procedure is generally used as a preliminary leak test and is often combined
with another leak detection method such as bubble testing or tracer gas detection
to detect and locate smaller leaks. The limit of sensitivity for pressure decay test is
in the range of 10−1 to 10−3 mbar.l/s.
• Bubble testing ([13]-10, [31], [32]): Bubble testing for leak location is one of
the most widely used tests because of it is quick, simple and has relatively low
cost to perform [2]. Immersion and liquid application are possible, however liquid
application is the most useful on large components on which the immersion tech-
nique is impractical. The sensitivity of the method is dependent on the pressure
differential, the gas used to create the differential, and the liquid used for test-
ing. Unless otherwise specified, the test gas is air; but, inert gases may be used.
This provides immediate indications of the existence and location of large leaks
(10−2 to 10−4 mbar.l/s). Smaller leaks may be found with longer inspection times
(up to 10−5 mbar.l/s). A quantitative measure is not possible with this method but
it can be used as pretest for the hydrostatic test.
• Tracer gas method ([13]-10, [33], [34], [35], [36], [37]):Tracer gas methods
describe a group of test methods used to detect and measure a tracer gas flowing
through a leak. It is one of the most efficient test methods because of its high
sensitivity [13]. The basic principle is to create a tracer gas pressure differential
across the pressure boundary and detect it on the low pressure side due to its
physical or chemical properties [37]. Two methods are generally described: Halogen
Diode Detection and Helium Mass Spectrometer Test. However, halogen detection
methods are losing their appeal as a tracer gas, due to environmental issues. An
important factor for any tracer gas detector is the background concentration in air.
Helium, being present in only small amounts (5 ppm) in the ambient air is preferred
for different reasons. Firstly, it penetrates small leaks rapidly. The only smaller
molecule is hydrogen, which is not inert. Helium is also non-toxic and relatively
inexpensive to produce. Typically a helium leak detector will be used to measure
leaks in the range of 10−6 to 10−13 mbar.l/s. A basic limitation to this method is
therefore that a large leak will saturate the detection. A pre-test method can be
employed to weed out the gross leaks.
• Radiotracer testing ([38], [39]): Radiotracers provide a powerful alternative
to other leak detection techniques. It uses the principle of detecting radioactive
material introduced from one side of the pressure boundary. They are the most
sensitive tools available and they can be used in-service, providing information in
the shortest possible time. Only very small concentrations of radioactive isotope
can provide an extremely detectable level of measurable radiation. The emission of
radiation is a specific property of the radioisotope and is not affected by interfer-
ence from other materials in the system. Thus radiotracers have strong resistance
against severe process conditions. As the characteristics of the radiation differs from
one radioisotope to another, multiple radiotracers may be employed and measured
simultaneously if needed to locate the leaks. The amounts of radiotracer used are
virtually insignificant. For this reason, when injected, they do not disturb the fluid
dynamics inside the vessel under investigation and do not interfere with product
quality. The radiotracer technique is very sensitive and enables the measurement of
leak flows up to 10−10 mbar.l/s.
2.3 Fracture mechanics tools
As mentioned in Section 2.1, any structures can contain defects introduced during
different stages of the component’s life. These defects must be taken into account because
they affect the reliability of the structure. Once detected and characterised (Section 2.2),
defects criticality must be assessed. Fracture mechanics is a field of solid mechanics that
analyses the mechanical behaviour of cracked bodies taking into account the stress and
strain fields in the area of a crack. Its objective is to characterize the crack behaviour
with measurable parameters (i.e. stress field, flaw size or material toughness) and can be
applied to a large range of scale, from the atomistic level to the full structure. Fracture
mechanics is used to evaluate the strength of a structure or component in the presence
of a defect. Historically, the first major development in modern fracture mechanics was
initiated by Inglis [40] in 1913. He published a stress analysis for an elliptical hole in an
infinite linear elastic plate. The next decades brought developments in the establishment
of methods to measure fracture toughness in laboratory specimens from wide plate test
to smaller standard notched specimens [41].
Figure 2.5: Illustration of a circumferential crack in a weld (TWI Ltd)
2.3.1 Crack behaviours
In fracture mechanics, two main modes of rupture are considered: brittle or ductile.
During crack propagation, new free surfaces are generated with a specific surface energy.
Not all available energy provided by the external load is used for the generation of new
crack surfaces. It is also transformed into other energies, like kinetic energy or dissipative
heat in the plastic zone at the crack tip. When a large part of this energy is used for crack
growth and little energy is absorbed in other ways, the fracture is said to be brittle.
Here, the component fails unstably and no measurement of the fracture resistance
beyond initiation is possible. When a lot of energy is transformed into other energies,
mainly due to dissipative mechanisms, the fracture is said to be ductile. The plastic
deformation dominates at the crack-tip, and the material fracture resistance increases as
the crack grows. A crack can be defined as a plane defect that separates a solid locally. The
displacement field is then discontinuous across this surface and three vector components
of this discontinuity form the three possible crack surface displacements to allow a crack
to propagate, known as Mode I, Mode II and Mode III (See Figure 2.6):
• Mode I (Opening mode): The crack tip is subjected to a tensile stress normal
to the crack plane. Crack faces separate at the crack front and the displacements
of the crack surfaces are normal to the crack plane.
• Mode II (In-plane shear mode): The crack tip is subjected to a shear stress
parallel to the plane of the crack and perpendicular to the crack front. Crack faces
slide relative to each other so that the displacements of the crack surfaces are in the
crack plane and normal to the crack front.
• Mode III (out-of plane shear mode): The crack tip is subjected to a shear
stress parallel to the plane of the crack and parallel to the crack front. Crack faces
move relative to each other so that the displacements of the crack surfaces are in
the crack plane and parallel to the crack front.
The superimposition of these three modes is sufficient to describe the most general
case of a loaded cracked body (mixed mode). However, most of developments have been
proposed for mode I, this being the usual mode for isotropic materials brittle fracture.
Figure 2.6: Illustration of the three modes of crack opening
Most formulae in fracture mechanics are derived considering these modes by assuming
either plane stress or plane strain conditions as defined below. However, mixed-modes
analyses is limited. In the case of a thick body, the plastic zone size depends on the
position of the crack tip through the thickness (Figure 2.7). There is a progressive decrease
in the size of the plastic zone when moving from the outer surface, where the plane
stress conditions are dominant, to within the field, where plane strain conditions are
dominant.
• Plane Stress Condition: In a thin body, the stress through the thickness (σzz ) is
assumed to be equal to zero (σzz = 0) because it cannot change significantly due to
the thin section.
• Plane Strain Condition: In a thick body, the material is constrained through
the thickness and strain in this direction is equal to zero (zz = 0) and σzz =
ν × (σxx + σyy ).
Figure 2.7: Illustration of the plane stress / plane strain conditions
For a century, a considerable amount of work has been done to study crack propaga-
tion under different loading conditions (constant and variable) and environments. Most of
these studies were conducted to develop efficient characterizing parameters. One of them
is fracture toughness which represents the material’s resistance to fracture. In general,
fracture toughness of a standard fatigue pre-cracked specimen depends on temperature,
environment, loading rate, the composition of the material and its microstructure, to-
gether with geometric, thickness and constraint effects. Crack stability assessment meth-
ods have been developed in three different frameworks: Linear Elastic Fracture Mechanics
(LEFM), Elastic-Plastic Fracture Mechanics (EPFM) and Limit Analysis (LA) [42]:
• LEFM methods are restricted to small-scale yielding configurations (i.e., geome-

try, loading. . . ) for brittle materials without significant ductility or plastic zone.
• EPFM methods do not have these limitations but require FE techniques which are
expensive and time consuming to perform to obtain correct and precise evaluations.
• Limit analysis methods are effective only for deeply cracked structures for which
initiation and maximum load are close to each other.
In simple terms, fracture occurs when the crack driving force exceeds the material
resistance to fracture. A description is given in the following subsections. The fracture
toughness which can be established from a single specimen [41] in terms of any or all of
the four parameters below:
• Energy release rate, G

• Stress Intensity Factor, K
• Crack Tip Opening Displacement, δ
• J-Integral, J
2.3.2 Energy release rate, G (Energy-based)
Based on the work of Griffith [43] on brittle materials in 1920s, energy release rate
criterion is based on the idea is that when a crack initiates or extends, it releases energy
and crack growth occurs when this energy is greater than the energy necessary to create
the surface relative to this extension. Using Inglis’ stress analysis [40], Griffith showed
that the change of energy, U , due to a through-thickness crack in an infinitely plate equals
to:
−2πa2 σ 2 πa2 σ 2
Ucracked − Uuncracked = + + 4aγs (2.3)
E E
where a is the crack half-length, σ the remote tensile stress, E the Young’s modulus and
γs is the surface energy of the material. Minimizing the energy in relation to the crack
length gives:
πa2 σ 2

∂
− + 4aγs = 0 (2.4)
∂a E
That leads to σf , the failure stress, given as:
r
2Eγs
σf = (2.5)
πa
This equation is significant because it relates the size of the imperfection to the tensile
strength of the material and predicts that small defects are less damaging than large ones.
Irwin [44] generalized Griffith’s energy approach (Eq.2.5) for ductile materials. He added
plastic deformation energy γp to the surface energy introduced by Griffith. He concluded
that rupture thus occurs for a critical stress defined by:
s
E(γs + γp )
σf = (2.6)
b0
He also proposed the energy release rate, based on Griffith’s work, defined by the
variation of the potential energy Π stored in the structure when increasing the free surface
of the corresponding extension of the crack:
∆Π ∂Π
G = lim =− (2.7)
∆A ∂A
where Π is the potential energy obtained by the subtraction of energy provided by external
forces and the internal strain energy and A the crack area. Using the energy required to
advance per unit of length by the crack, known as crack growth resistance R and the energy
release rate G, the crack propagation can be defined. The so-called R-Curve is a rising
curve, nearly independent of the initial crack [41], that describes that the plastic zone
growth as the crack extends. The propagation may be stable or unstable, depending on
how parameters vary with crack size. It can be noted that R-curve is geometry dependent
and decreases with an increasing thickness. The conditions for stable crack growth can
be expressed as follows:

G = R, dG ≤ dR
for stable crack growth
da da
(2.8)
dG
 > dR
da da
for unstable crack growth
Figure 2.8: Schematic driving force vs. rising R curve
2.3.3 Stress Intensity Factor, K (Stress-based)
The stress intensity factor, K, defines the magnitude of the local stresses around the
crack tip depending on loading, crack size, crack shape, and geometric boundaries. As-
suming isotropic linear elastic material behaviour, some authors [45] showed that the
stress field in a cracked body exhibit a √1r singularity at the crack-tip. Employing polar
coordinates (r, θ) centered at the crack tip, stresses can be expressed as:
∞
k X m
σij = √ × fij (θ) + Am × r 2 × gijm (θ) (2.9)
r m=0
Figure 2.9: Stresses at an arbitrary distance (in terms of distance and angle) from the
crack-tip under linear elastic assumptions [45]
Irwin [44] showed that it was convenient at this point to replace k by the stress intensity
√
factor K, where K = k × 2π . The √1r singularity being dominant, this led to:

 lim σ I = √K2πrI
× fijI (θ)
 r→0 ij


limr→0 σijII = √K2πr
II
× fijII (θ) (2.10)


KIII
σ III = √ × f III (θ)

lim
r→0 ij 2πr ij
He defined the stress intensity factor, K, a single parameter which characterises the
stress field near the crack-tip. It is a stress-based measure proportional to the square root
of the crack size, multiplied by a function Y which depends on both the geometry and
the loading mode. A more general form of K is given as:
√
K = σY πa (2.11)
where σ is the remote stress applied to the component, a, the crack length and Y a geo-
metrical correction factor. Many solutions for stress intensity factor have been developed
for various geometries and loading to assess the Y factor and can be found in many hand-
books. When the stresses at crack-tip reach a critical material value, known as Kmat the
materials’ generic fracture toughness, the crack will grow. Note that KIc is often used
for brittle fracture. Irwin [44] also introduced a simple relationship between K and the
elastic energy release rate G. The contributions to G from the three modes are additive
because energy release rate is a scalar quantity and can be written as:
2
KI2 KII 2
KIII
G= + + (2.12)
E0 E0 2µ
where E 0 = E for plane stress and E 0 = 1−ν

E
2 for plane strain. It is important to note
that if “the three modes K” contribute to G, they are not additive to calculate the total
SIF (Ktot 6= KI + KII + KIII ). But it is possible to add different sources of stress in
the same mode to obtain the total SIF (Ktot = KIa + KIb + KIc + ...)according to the
principle of superposition [45]. However, most fracture mechanics analyses are based on
K solutions for crack opening Mode I loading (KI ). So the previous relation is generally
limited to:
KI2
G= 0 (2.13)
E
Fracture occurs when the driving force KI (which is load and geometry dependent)
reaches its critical intensity KIc (which is material, temperature and rate of loading de-
pendent), the plane strain fracture toughness which represents the material resistance to
fracture, expressed as:
KI ≥ KIc (2.14)
However as r (defined in Figure 2.9) approaches zero, the stress at the crack-tip tends
to infinity because of the singularity at r = 0. As infinite stresses cannot exist physically,
the elastic solution must be modified to take some crack tip plasticity into account. This
plastic zone formed permits a relaxation of stresses near the crack-tip. Westergaard [46]
developed a first-order estimation of the size of the plastic zone on the axis of the crack
where the presence of a plastic zone is assumed small compared to the length of the crack
and yielding occurring when σ = σys the yield strength of the material. It leads to a first
order estimate of the plastic zone size for perfectly elastic materials and is presented as
follows:  2
ry = 1 × KI for plane stress
2π σ
ys 2 (2.15)
r = 1 × KI × (1 − 2ν 2 )for plane strain
y 2π σys
When yielding occurs, stresses must redistribute at the crack-tip in order to reach
equilibrium. The plastic zone must increase in size in order to accommodate these forces.
Irwin [47] developed the second order approximation for the confined plastic zone based
on the redistribution of stresses around the crack tip. A simple force balance leads to a
second-order estimate of the plastic zone size rp :
Z ry Z ry
K
σys rp = σdr = √ I dr (2.16)
0 0 2πr
Integrating and solving this leads to:

 2
1 KI
rp =
π
× σys
for plane stress
2 (2.17)
r =
p
1
3π
× σKysI for plane strain
This plastic effect, applicable only to low levels of loading when there is no generalized
plasticity, is shown on Figure 2.10, coloured area (*) representing loads that must be
redistributed, resulting in a larger plastic zone.
Figure 2.10: Irwin’s plastic zone correction for the stresses at the tip of a crack [45]
A similar result was obtained by Dugdale [48] (Irwin’s solution was π1 = 0.318 and
Dugdale instead used π8 = 0.392) who developed another model known as strip yield
model: 2
π KI
ρ= × (2.18)
8 σys
An important restriction to the use of LEFM (K) is that plastic zone size at the crack
tip must be small relative to the crack length as well as the geometrical dimensions of the
specimen or part. Otherwise, a plasticity correction is required for the stress intensity
factor, K, or elastic-plastic fracture mechanics may be needed.
2.3.4 J-Integral (Energy-based)
An alternative fracture parameter, developed by Rice is called ’J’ and characterizes

the energy required to propagate a crack based on an integral around the crack tip.
The J-Integral calculation is applicable to elastic or elastic-plastic material treated by a
deformation theory of plasticity [49] and is related to the previous work of Griffith [43].
A general definition of the J-Integral can be expressed as the rate of decrease of potential
energy with respect to crack area:
dΠ
J =− (2.19)
dA
where Π is the potential energy obtained by the subtraction of energy provided by

external forces and the internal strain energy and A the crack area. In case of a linear-
elastic material and small scale yielding, J is assumed to be equal to G:
KI2
J =G= (2.20)
E0
The mathematical definition of the J-Integral proposed in 1968 by Rice [50], is cal-
culated on a closed curve Γ surrounding the crack tip. A physical interpretation on a
contour Γ, which represent any curve surrounding the crack-tip, is that the J-Integral is
equal to the rate of decrease of the potential energy stored in the field surrounded by Γ
when the crack grows. It describes the crack face-opening work when a flaw grows in a
linear or non-linear material. The original familiar form of the J-Integral is:
Z
∂ui
J= wdy − Ti ds (2.21)
Γ ∂x
where ui are displacement vector components, Ti traction vector components are defined
R
by Ti = σij .nj , w the strain-energy density defined by w = 0 σij dij and ds and y are
defined in Figure 2.11.
Figure 2.11: Representation of parameters in J-Integral (Adapted from [41])
An important aspect to highlight is the path-independence of J-Integral for linear

elasticity [50]. This property remains true for elastic-plastic behaviour if the loading
increases monotonically without unloading. J always increases with an increase of the
crack length under constant load conditions, suggesting that no crack arrest is possible.
As a result, the integration path can be taken through regions where the elastic-plastic
state is determined with sufficient accuracy and possible sources of numerical inaccuracy
(due to the possible singular state in the region surrounding the crack tip) can be avoided.
Rice [50] also showed that the J-integral may be interpreted as the energy release rate
on a nonlinear elastic body containing a crack per unit fracture surface area. The elastic
energy release rate is defined as the rate of change in potential energy with the crack area
whereas much of the strain energy absorbed by an elastic-plastic material is not recovered
when the crack grows. Therefore, rather than defining the energy released from the body
when crack grows, the following equation, expressed in displacement control condition,
relates J to the difference in energy absorbed during load:

dU
J =− (2.22)
Bda ∆
where U is the strain energy, a, the crack length, B the thickness and ∆ the associ-
ated load-point displacement, kept constant during partial differentiation. An alternate
form has been developed by Rice [51] to evaluate the J-Integral directly from the load-
displacement curve from a single specimen. This approach is relatively easy to apply in
an experimental procedure and does not require multiple specimen tests. It allows J to
be related the total load applied (P ) or the associated load-point displacement (∆):
Z P Z ∆
1 ∂∆ 1 ∂P
J =− dP = − d∆ (2.23)
B 0 ∂a P B 0 ∂a ∆
A more general relationship for estimating the J-integral was proposed by Sumpter
and Turner [52] as a sum of elastic and plastic components:
ηel Ael ηpl Apl

J= + (2.24)
Bb Bb
where B is the thickness, b = (W − a) the ligament length, Ael and Apl the elastic
and plastic area under the load–displacement curve and ηel and ηpl two geometry factors
dependent on a/W but independent of material properties.
2.3.5 Crack-Tip Opening Displacement (CTOD) (Strain-based)
Crack-tip opening displacement (CTOD) proposed by Wells [53] in 1963 is used to

quantify fracture toughness. It represents the physical opening of the crack faces to
characterize the fracture toughness. In the first instance, it was approximated for a
centre-cracked plate as:
4 KI2
δ= (2.25)
π Eσys
Two definitions of CTOD can be found in the literature: (i) the opening displacement
of the original crack tip calculated from the crack mouth opening displacement (CMOD)
and (ii) the displacement at the intersection of a 90◦ vertex with the crack edges when it
comes from numerical modelling (Figure 2.12).
Figure 2.12: Two different definition for Crack Tip Opening Displacement (CTOD)
(Adapted from [54])
Rice and Rosengren [55] showed that a general relationship can be made between J
and CTOD :
J = m × σys × δ (2.26)
where m is an empirical plastic constraint factor set as 1 for plane stress conditions
and varying from 1 to 2 for actual stress states. Under linear elastic conditions (σ σys ),
fracture toughness parameters can therefore be linearly related so that they are equivalent
as fracture characterizing parameters:
KI2
G=J = = m × σys × δ (2.27)
E0
CTOD is a strain-based criterion separated into two components: elastic and plastic.
While the elastic part of CTOD is obtained from the stress intensity factor, K considering
LEFM, the plastic component is derived from the crack mouth opening displacement
which is measured with the help of a clip gauge (See Figure 2.13). A general estimate of
CTOD was proposed by Hellman and Schwalbe [56] as:
KI2 (1 − ν 2 ) [rp (W − a) + ∆a]Vpl

δ= + (2.28)
mσys E [rp (W − a) + a + Z]
where a is the crack length, ∆a the crack extension, Z the distance of knife edges
from the front face of specimen, Vpl the plastic component of the crack mouth opening
displacement and rp a plastic rotation factor.
Figure 2.13: CTOD measurement using clip-gauge (Adapted from [54])
CTOD is a single parameter which also accounts for the crack-tip plasticity. From a
practical point of view, CTOD is easier to conceptualise and measure compared to the
similar techniques like J-integral, being a fracture parameter which has more physical
meaning. However, the concept of CTOD is harder to expand the for large deformation
conditions and it is easier to calculate J-integral in case of a design process using finite
element modelling (FEM) techniques
2.4 Fitness-for-Service and Engineering Critical As-

sessment
Conventional design and fabrication codes/standards aim to ensure that the structure
is safe and fabrication quality is high, assuming the components to be defect-free. Specific
codes exist for a range of different engineering structures, from pipelines and bridges to
pressure vessels and nuclear equipment. However, engineering structures may contain
defects and mechanical properties can change over time depending on the type of the
component and operating conditions [10]. Even when the components are manufactured
with due care they may still contain or develop defects. As a result, many items of
operating equipment may not still satisfy the original requirements after some years in
service. These defects can occur during manufacture of metal components and during
service life, with different origins and are governed by various growth mechanisms as
discussed in Section 2.1 (e.g. corrosion, pores, flaws, cracks or welding defects). In cases
where safety is a primary issue (such as aircraft, pipelines, pressure vessels, etc.), the
failure of a single part due to the presence of a defect can threaten human lives and can
result in severe economic or environmental consequences. The structural importance of
such defects needs to be assessed to prevent component failure during its service life. If
a defect is found, an assessment is required in accordance with a recognised and codified
procedure. This will determine if it should be removed, can be repaired, or whether the
component has to be replaced. Conversely, some defects may be harmless and do not
lead to failure during the lifetime of the component. The cost implication of replacement
or repair of such defects is a financial loss and could even lead to the introduction of
more significant defects. In this case, defects can be left in the component to continue
its use. Five major failure modes, namely fracture, fatigue, creep, plastic collapse and
corrosion have been identified as the most common phenomena in engineering structures
[57]. A procedure for the Fitness-For-Service (FFS) assessment based on the principles of
fracture mechanics are used to evaluate planar defects consistently through an Engineering
Critical Assessment (ECA), and make rational decisions about the remaining life, whether
to repair, replace a component or reduce operating conditions. For this reason, different
FFS methods have been developed to cover each of these failure mechanisms.
2.4.1 Definitions
Fitness-For-Service procedure is defined [58] as a quantitative engineering analysis pro-

cess incorporating calculations and assessments, which are performed to assess the struc-
tural integrity of an in-service item of a plant, containing defects or has the potential to
develop defects. An item is considered to be fit for the intended service, if it can be clearly
demonstrated (with acceptable safety margin) that the conditions to cause failure will not
be reached within a predetermined time period or lifetime. In practice, the use of FFS
procedures involves performing an Engineering Critical Assessment (ECA) of a compo-
nent containing a defect to assure its structural integrity during the intended service-life
or until the next planned inspection period. Generally speaking, FFS assessment methods
allow the users to determine whether the structural components are safe and suitable for
operating conditions, to make informed decisions in the operation (repair, reclassification,
alteration or removal) to assess the remaining life of the equipment, set up inspection at
regular intervals, reduce downtime and extend life time. FFS assessments are also used
as a preventive measure to demonstrate the tolerance and existing safety margins for a
postulated defect.
2.4.2 Assessment procedures available and application
A large survey was conducted among FFS-users in 2003 which provided an insight
into FFS assessment methods available [57]. FFS procedures are applicable to most
metallic structures including pressure vessels, pipelines, offshore rigs, storage tanks, ships,
bridges and other structural components. Some common types of defects assessed using
FFS evaluations are uniform/localized or pitting corrosion, blisters, laminations, creep,
gouges/dents, cracks, fatigue or fire damages. A number of industrial sectors have estab-
lished FFS standards in place for the assessment of flaws found in-service depending on
the field of application (e.g. nuclear industry, oil and gas, etc.). Some of the widely used
standards are:
• BS 7910 - Guide to methods for assessing the acceptability of flaws in metallic

structures [1]
• FITNET - FITness-for-service NETwork [59]
• API 579-1/ASME FFS-1 - Fitness for Service [9]
• R6 - Assessment of the integrity of structures containing defects [60]
FFS assessment methods are used in different ways to assess the reliability of com-
ponents. There are typically four distinct ways of using FFS assessments. In terms of
design, FFS assessments can be used to determine material properties requirements or
stress levels that would be suitable for the design. A defect is postulated to assess the
critical condition of the new component based on NDT detection limit. It can assist in
material selection, design and decision to select the appropriate NDT technique (Section
2.2). Secondly, in terms of fabrication, FFS analyses can be used to determine flaw ac-
ceptance criteria and provide guidance on the fabrication route to assure the quality of
component. In-service, FFS can be used to assess the structural integrity of components
when defects are detected. Criticality of defects and remaining lifetime of the component
can be assessed. Thirdly, FFS can also help in deciding the repair/replacement strategy
of the component, for example a reduction of the operating loads may be required in
order for the component to remain in-service. It can also be used to support cases for
life extension or change of service. Finally, FFS methods can also be used to investigate
and clarify the root cause of failures and evolution of defects and thus avoiding similar
failures in the future.
2.4.3 Input data and assessment
Traditional design codes and standards generally concentrate only on two main sets
of parameters: the applied stresses and material properties, assuming a certain level of
fabrication quality. However, the integrity of a structure depends on a combination of
factors taking into account the actual condition (i.e. presence of defects or flaws). This
is illustrated with the FFS triangle in Figure 2.14. Therefore, a typical FFS assessment
requires interdisciplinary inputs for the components in-service [9]. Defects have to be
identified (type, location, size, etc.) and information about operating conditions, compo-
nent geometries and material properties have to be known. The quality of these inputs
data has a significant effect on the value of the FFS analysis results.
Figure 2.14: Illustration of the three factors affecting structural integrity
The following guidance is given in BS 7910 (Annexe H) [1] to report FFS assess-
ment:
1. Provide analysis details, including the Option or Level of analysis carried out
2. Input data
(a) Loading/operating conditions: loads (normal, transient, fault,. . . ), tempera-

ture, environment, etc
(b) Material properties: material spec., stress-strain data, fracture toughness,

source and validity of all data, etc
(c) Defect characterisation: type, location, orientation, shape, allowance for sizing
errors, etc
(d) Defect growth: allowance for defect extension (i.e fatigue, tearing, creep, stress
corrosion), growth laws employed, etc
(e) Plastic limit load and stress intensity factor solution employed: from published
solutions or finite element analysis.
3. Results
4. Sensitivity analysis
The framework described permits to have an overview of a classical assessment proce-

dure. Note that depending on the defects and mechanisms under investigation, additional
steps may be required. In every case, care has to be taken with the input data and jus-
tification made at each step. Sensitivity analysis are undertaken to assess the effect of
possible uncertainty on final results.
2.5 Pressure vessel design
Pressure vessels are widely used, both in everyday life as well as in industry, for storage
and transportation of liquids or gases. They can be found in a wide range of applica-
tions from a fire extinguisher to a reactor pressure vessel (RPV) in the nuclear indus-
try or as components of aerospace and marine vehicles. Most common are cylindrical,
hence hoop(circumferential) or axial (longitudinal) stresses, and spherical (circumferen-
tial/meridional). Depending on the ratio of the wall-thickness to diameter, pressure vessels
are classified in terms of design as thin- or thick-walled pressure vessels. Thin-wall the-
ory is developed from strength of materials solutions which yields the state of stress as
an average over the pressure vessel wall thickness whereas thick-wall theory is developed
from the theory of elasticity which yields the state of stress as a continuous function of ra-
dius over the pressure vessel wall thickness [61]. For thin-walled pressure vessels, the wall
thickness is assumed to be small compared to other dimensions. According to theory, for a
wall thickness (t) and radius (R), a pressure vessels is assumed as thin-walled if R/t > 20.
In practice, a less conservative rule is typically used and a pressure vessel is assumed to
be thin-walled if R/t > 10 and thick-walled if R/t < 10. In terms of thin-wall theory, it
is assumed that the hoop and longitudinal stresses are constant across the wall thickness,
and the radial stress, being small in comparison with the hoop and longitudinal stresses,
is negligible. This is an approximation since it will vary from zero at the outside surface
to a value equal to the internal pressure at the inside surface. Note that depending on the
curvature, they might also be considered to have the same behaviour as flat plates.
2.5.1 Stresses in cylindrical components
For a cylinder subjected to internal pressure, the coordinate system is cylindrical with
axial, circumferential and radial direction as shown in Figure 2.15. The cylinder is defined
by its inside radius (Ri ), outside radius (Ro ), thickness (t) and length (L). When subjected
to an internal pressure (Pi ), stresses develop in the cylinder. Each element of pipe is
subjected to the following stresses:
• Circumferential stress (or hoop stress σh )

• Axial stress (or longitudinal stress σl )
• Radial Stress (σr )
Figure 2.15: Geometry and stresses in a cylinder
The following expressions are meant for internal pressure. If the external pressure is
higher, the solutions should be applied with caution because other failure modes such as
instability due to wall buckling may come into play. Thin-walled cylinders are calculated
by Barlow’s formula and thick-walled cylinder with Lame’s formula.
2.5.1.1 Stress in Circumferential Direction - Hoop Stress
Circumferential or hoop stress (σh ) acts around the circumference of a cylinder. It is

assumed to be uniform across the wall thickness. Very high hoop stress may split the pipe
into two halves across any plane, including the diameter and axis of the pipe. The hoop
stress at a given point in the cylinder wall can be expressed as:
• Thin-walled cylinder
Pi Ri
σh = (2.29)
t
• Thick-walled cylinder
Pi Ri2 − Po Ro2 Ri2 Ro2 (Po − Pi )

σh = − 2 2 (2.30)
Ro2 − Ri2 r (Ro − Ri2 )
2.5.1.2 Stress in Axial Direction – Longitudinal stress
Longitudinal stress (σl ) is parallel to the longitudinal direction of the cylinder. It is

only valid when at a distance far away from end caps where bending, non-linearities and
stress concentrations are not significant. The stress in axial direction at a given point in
the cylinder wall can be expressed as:
Pi Ri
σl = (2.31)
2t
Pi Ri2 − Po Ro2
σl = (2.32)
Ro2 − Ri2
2.5.1.3 Stress in Radial Direction
Radial stress (σr ) acts in directions coplanar with but perpendicular to the symmetry
axis. In most pipe design, the radial component of normal stresses (σr ) is negligible
compared to the other two components (σh and σl ). For a thick-walled pipe it may be
taken equal and opposite to the gauge pressure on the inside surface, and zero on the
outside surface. The stress in radial direction at a point in the cylinder wall can be
expressed as:
σr = p (Generally σr = 0 by definition) (2.33)
Pi Ri2 − Po Ro2 Ri2 Ro2 (Po − Pi )

σr = + 2 2 (2.34)
Ro2 − Ri2 r (Ro − Ri2 )
2.5.1.4 Discussion
The relationships presented earlier are sufficiently accurate to design cylindrical pres-
sure components if no defect is present. But presence of defects such as presented in
Section 2.1 can affect component integrity. In pressure vessels codes, various shapes are
detailed including cap ends. For a comparable radius, wall thickness and internal pressure,
a spherical pressure vessel is the optimum geometry with its maximum normal stress in
is one half as large as that in a cylindrical one.
In a cylindrical shaped vessel, the internal pressure is resisted only by the hoop stress
in “arch action” whereas in the spherical vessel, the double curvature means that all stress
directions around the pressure point contribute to resisting the pressure. However, cylin-
drical geometry may be more convenient to fabricate and transport. Moreover cylinder
and the ends would in general expand by different level. If thick plates are used instead of
relatively flexible hemispherical ends, juncture stresses would increase considerably. For
this reason, the ends of cylindrical pressure vessels must be designed carefully, and flat
ends should be avoided if possible.
2.5.2 Pressure vessel codes
Pressure vessels design methods can be divided into two basic categories regardless of
which code is used: the Design by Rule (DBR) method and the Design by Analysis (DBA)
method. In the first one, overall dimensions and loads are known and the wall-thickness
is calculated in consequence. In the latter, the vessel geometry and dimensions are known
and allowable loads are evaluated. The structural analysis is divided into elastic analysis
and inelastic or plastic analysis depending on anticipated material behaviour. Differences
between the material behaviours will be explained with reference to the stress-strain
diagram. Various pressure vessels codes exist, depending on industrial regulations and
laws. Amongst the most known and used there are:
• USA - ASME Section VIII-Boiler & Pressure Vessel Code (Division 1, Division 2)
[62]
• EU - EN 13445 - Harmonised European code for unfired pressure vessels [63]
• UK - PD 5500 - Specification for unfired pressure vessels [12]
These codes can be applied to all pressure equipment except such pressure equipment
that are critical to nuclear safety which should be designed and manufactured in accor-
dance with specific codes not detailed here. Two main approaches are used for the design
of pressure vessels. In Europe, pressure vessels have to be designed and built according
to the Pressure Equipment Directive (PED), whereas the ASME Boiler & Pressure Vessel
Code [62] prevails in North America. The philosophy is different between the two. Some
aspects are detailed here. While PED is a European legal requirement applied for all Eu-
ropean Union Member States, the ASME BPVC is “only” a construction standard even if
it is mandatory in USA and Canada. In PED, Essential Safety Requirements have to be
met. The recommendation is to follow harmonized standards (EN-13445, EN-13480, etc.)
to reach PED’s requirements but other codes can be used including ASME, providing
conformance with the PED can be demonstrated. Welding procedures and NDT require
recognized third-party certification to meet the PED’s requirements whereas in ASME
code it is approved by the manufacturer himself. Moreover the qualification of welders
is valid without time limit to ASME IX whereas the welder’s certificate to EN 287-1 ) is
only valid for 2 years. There are also differences on acceptable materials between PED
and ASME. Some authorized materials are listed in ASME Section II can also be used
in harmonized standards for the PED. The British pressure vessel code, PD 5500 [12] ,
can be used under the PED to design and manufacture a pressure vessel. Another aspect
that can differ is the safety factors to apply on material properties. Table 2.1 summarizes
values of safety factors used to obtain the allowable stress for a general case. Note that
different safety factors can be found in relation to the material used in the design or the
application.
Table 2.1: General safety factors used in pressure vessel codes

Codes Value of safety factor n Value of safety factor n
(U T S/n) (Y S/n)
ASME Section VIII Div.1 3.5 1.5
ASME Section VIII Div.2 2.4 1.5
EN 13445-3 2.4 1.5
PD 5500 2.35 1.5
Figure 2.16: Allowable stress based on the material stress-strain curve behaviour
A study comparing the costs of pressure vessels designed using the EN 13445-3 [63]
and ASME Boiler & Pressure Vessel (Section VIII) [62] codes has been carried out un-
der a European Commission contract [64]. Nine examples of widely used pressure vessels
were considered, including a gas storage tank, hydrogen reactor, stirring vessel, autoclave,
water separator, air cooler header and three different types of heat exchangers. Overall it
was demonstrated that EN 13445-3 offers an economically competitive design for unfired
pressure vessels. Material costs are frequently greater when using the ASME code. How-
ever, savings attributable to lower material costs with EN 13445-3 are partly offset by
additional costs of weld testing and NDT when compared with ASME requirements.
2.6 Summary
This literature study has covered different aspects of the subject. It has been shown
that metallic components are unlikely to be defect-free. A wide range of defects can be
present at different stages of their life (Section 2.1). These defects that can be introduced
during fabrication or in-service due to the operating conditions might be harmful and are
often covered by design codes safety margins.
Advances in knowledge of material behaviour and Non-Destructive Testing (NDT)

technologies permit these defects to be predicted and detected, and can help reduce the
number of defects that may occur at the different stages (Section 2.2). A better knowledge
about material behaviour permits a better material selection depending on the application.
Improved NDT technologies and wide use enables improvements in a component’s safety.
Note that as a result of methods or equipment’s accuracy, they are accompanied by some
uncertainty of measurement which have to be taken into account. An understanding of
suitable NDT techniques is important to help evaluate the final shape and dimensions of
defects. They are particularly important in a leak-before-break assessment that requires
a certain monitoring of a detected defect. Both part-penetrating and through-wall defects
might be assessed via conventional NDT or leak testing.
Numerous techniques are available in terms of leak detection. Conventional leak testing
methods such as hydrostatic testing, visual inspection or bubble testing are only quali-
tative (which means they are of limited use to provide input for a LBB assessment) and
offer limited leak detection capability. In terms of in-service monitoring, hydrostatic test-
ing and visual inspection have limited practical potential. Relying on visual inspection
requires large leaks in order for the change to be noticeable, and it may be hours or days
before the leak is discovered. However, visual detection is widely used to detect leaks in
long pipelines. These tests are used on a go/no-go basis to assure that no large leaks
are present. Liquid penetrant testing permits qualitative detection and location of leaks
without application of a pressure boundary however access to both sides is necessary and
the measurement of flow rates is not practical. Tracer gases (or radiotracers) and Acous-
tic Emissions methods on another hand permit testing of the component in the same
conditions as in service, or directly in-service monitoring. The former permits detection
of very small leaks however, the precise detection can be difficult. The latter gives better
results in identifying the location of the leak but the sensitivity range is limited and will
depend on the material and fluid properties. Common practice is to use a combination
of the different methods to reach the final objective of leak detection, identification of
the location, and measurement of defect size and leak rate. Due to the range of sensitiv-
ity, some methods are used as pre-test to assure that no large leaks exist, more sensitive
studies being carried out afterwards.
To assess these detected defects, fracture mechanics solutions have been developed
(Section 2.3). Depending on the materials under investigation or assumptions made,
solutions are generally divided into linear elastic (LEFM) or elastic-plastic (EPFM). Dif-
ferent parameters have been defined to characterise the stress field at proximity of the
defects to assess their criticality. These parameters provide a link between the local crack
tip response (Stress/Strain) to the global conditions (Loads, material properties, geome-
try). Many solutions have been derived for different cases and can be found in standards
or handbook.
For unstable crack propagation to occur, the crack driving force must exceed the re-
sistance of the material to crack growth. Fracture toughness is a critical input parameter
for fracture mechanics based fitness-for-service assessments. For example some standards
as BS 7910 [1] recommend to use the lowest value measured from three small-scale speci-
mens. It can sometimes be estimated from the literature or materials properties databases
but it is preferable to base this input on experimental measurement for the particular
material and condition being assessed. It is important when establishing appropriate
fracture toughness data that the service conditions to which the component is subject to
are adequately considered including the dimensions of the crack, the section thickness,
temperature, strain rate and environment.
Moreover, for particular cases, solutions can be derived using finite element analysis.
Fitness-For-Service (FFS) assessments based on the principles of fracture mechanics are
used to evaluate the defects consistently through an Engineering Critical Assessment
(ECA) (Section 2.4). This permits to make rational decisions about the remaining life,
whether to repair, replace a component or reduce operating conditions. As the concept
of leak-before-break is mainly employed for pressurised components, a brief review of
pressure vessel design codes is discussed in section 2.5. Depending on the design code
employed, the requirement in terms of materials or geometrical dimensions might differs.
This has to be taken into account while applying FFS assessment.
Chapter 3
Leak-Before-Break concept
Contents
3.1 Background . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 39
3.2 Definition and application . . . . . . . . . . . . . . . . . . . . . 40
3.2.1 Basic design formulation . . . . . . . . . . . . . . . . . . . . . . 40
3.2.2 Detailed procedures . . . . . . . . . . . . . . . . . . . . . . . . 41
3.2.3 Limitations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 44
3.3 Leak-Before-Break Procedures . . . . . . . . . . . . . . . . . . 45
3.3.1 Nuclear industry . . . . . . . . . . . . . . . . . . . . . . . . . . 45
3.3.2 Other industries (Non-nuclear) . . . . . . . . . . . . . . . . . . 47
3.4 Applying LBB to industries other than nuclear . . . . . . . . 48
3.5 Crack Opening Area . . . . . . . . . . . . . . . . . . . . . . . . . 49
3.5.1 Analytical model . . . . . . . . . . . . . . . . . . . . . . . . . . 49
3.5.2 Leak rate and COA . . . . . . . . . . . . . . . . . . . . . . . . 52
3.6 Past and future research in LBB . . . . . . . . . . . . . . . . . 52
3.7 Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 54
3.8 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 56
38
CHAPTER 3. LEAK-BEFORE-BREAK CONCEPT 39
3.1 Background
Structural integrity of components containing fluids is critical for economic, environ-

mental and safety issues. Any risk of catastrophic failure, in either a brittle or ductile
manner, is not acceptable across the majority of industries. Consequently, many efforts
have been invested in the structural integrity aspect to improve the assessment method-
ologies of pressure containing equipment. One of the ways to aid the decision whether
or not to live with the defect is through the demonstration of Leak-Before-Break (LBB).
LBB is a well-established practice in the nuclear industry around the world, albeit as a
defence-in-depth argument or to justify the elimination of pipe whip restraints, but it also
finds applicability in other industries. The principle of Leak-Before-Break (LBB) is that
a flaw in a structure will become sufficiently large to grow through the thickness of the
material wall, and allow leakage of the fluid contained within the pipe or pressure vessel.
The safety principle is to ensure the leak rate is sufficiently large to ensure the flaw is
detected (and can then be repaired) before the flaw becomes sufficiently large to cause
catastrophic global fracture of the structure. A review of the available procedures, their
associated limitations and the research carried out in the last thirty years is presented
below. Application of this concept within non-nuclear industries is also discussed.
Since 1950, numerous investigations have been performed to assess the mechanical
and structural behaviour of pressurized components, such as loading capacity and failure
behaviour of piping. One of the first few cases associated with LBB was presented by
Irwin [65] in the 1960s. According to his work, leakage was predicted to occur due to an
axial flaw if the defect length was less than twice the thickness of the pressure vessel. In
that case, crack driving forces in the radial direction exceed those in the axial direction
resulting in a through-wall crack which could exist up to a significant size without any
risks of pipe burst. After that, most research on LBB has been carried out for nuclear
applications. Historically, an instantaneous double-ended guillotine break (DEGB) of
the largest heat transport pipe was used as the design basis in nuclear power plant,
assuming that the pipe would break in a brittle manner [66]. This led to the installation
of numerous pipe-whip restraints to hold ruptured pipes in place. However this criterion
was restrictive [67], due to the risk of loose pipe ends jamming under certain conditions
and the difficulties of carrying out inspection. Advances in fracture mechanics allowed a
better understanding of piping behaviour and it has been demonstrated that postulated
small through-wall flaws could be detected by leakage long before the flaws could grow to
unstable sizes which might cause a DEGB [66]. For this reason, developing an alternate
design criterion was necessary [68], [69]. Further studies have expanded the elaboration
of LBB procedures, which were adopted in 1986 by the United States Nuclear Regulatory
Commission (USNRC), for the assessments of high energy pipes in Pressurized Water
Reactors (PWRs), which provided guidelines (revised in 2007 [70]) for safety evaluation
of the operating and design of Nuclear Power Plant. LBB assessment methods have
contributed to a new approach of pressure equipment design. Details about guidance for
the implementation, limitations and acceptance criteria for LBB were provided in the late
1980s by the American regulatory authority [70], [71], [72]. Nowadays, this criterion is
widely used in the nuclear industry as either validation to remove pipe-whip restraints
and jet-impingement shields or as defence-in-depth argument. Outside of the nuclear
industry, LBB arguments are sometimes included as part of Fitness-for-Service (FFS)
assessments.
3.2 Definition and application
The European Commission [73] defines LBB as “a failure mode of a cracked piping
leaking through-wall crack which may by timely and safely detected by the available
monitoring systems and which does not challenge the pipe’s capability to withstand any
design loading”. Although inelegantly described, this concept is related to pipe failures
and their safety implications and it has been presented as a way to partially relax the
common requirements to the postulated DEGB failure. Fracture mechanics principles are
used to demonstrate that a flaw will develop through-wall allowing sufficient and stable
leakage that it can be detected before catastrophic rupture of the component occurs. This
concept may therefore be applied to structures containing a fluid such as pipes or pressure
vessels. LBB is applicable to ductile materials which exhibit high toughness and are
fracture resistant [73]. These material properties permit a through-wall defect of a certain
length to be stable under specified conditions and allow sufficient time for the detection
of the resulting leak. A combination of ductile material, benign fluid environment and a
reliable leak detection system is therefore necessary.
3.2.1 Basic design formulation
A basic analysis to show the balance of leak conditions and break conditions is presented
in [74]. These formulations are used for the design of pressurized thin-walled structures.
For example, a thin-walled cylindrical pressure vessel of radius R and thickness t is subject
to an internal pressure P . In the basis of design, the maximum hoop stress cannot exceed
2/3 of the yield strength of the material (σh ≤ 2σys /3), but the theoretical thickness t to
preclude yielding can be written as:
P ×R
t≥ (3.1)
σys
In the case of a through-wall crack (2ac = t) where a leak may be detected, the crack will
remain stable if :
KIc
σh ≤ q (3.2)
π × 2t
where KIc is the plane strain fracture toughness. These two equations lead to a limiting
pressure:
2 K2
P ≤ × Ic (3.3)
πR σys
These basic conditions (more detailed procedures will be discussed in paragraph 3.2.2)
are used at the design stage to select materials and they may also be used to ensure
that a leak-before-break condition can be feasibly reached. Design engineers select the
material yield strength and thickness according to conventional formulae of stress analysis
so that the wall thickness is sufficient to withstand the internal pressure. The next step
is the selection of the minimum required fracture toughness to meet the leak-before-break
criterion. This is followed by the evaluation of cost of material, fabrication, certification,
and other technical and economic decisions [75].
3.2.2 Detailed procedures
In a detailed LBB assessment a number of different calculations is required, including

those to establish the limiting length of a through wall defect (based on well-established
fracture mechanics principles) and those of crack opening area (which allows the fluid to
escape). Validation of methods for the calculation of the limiting length of a through wall
defect is included in the validation of flaw assessment procedures [76]. The formulation
of a Leak-Before-Break argument can be explained with the aid of the following diagram
(see Figure 3.1) where (1) and (2) represent the margins applied on leak detection and
crack length and flowchart (see Figure 3.2).
During stable crack growth (Crack length < Critical crack length) the penetrating
crack will grow to a through-wall crack and form a leak until it reaches the critical length.
Catastrophic failure occurs when the crack length reaches its critical length leading to
unstable crack growth assuming stresses are load-controlled (generally true for pipes con-
taining high energy fluids). Under fatigue crack growth, defects will grow under the action
of cyclic stress mainly due to changes in internal pressure or due to thermal transient load
cycles. For example, circumferential defects will grow under cyclic axial stresses and are
subject to axial pipe end load, internal pressure and external pipe bending moments
[77].
Figure 3.1: LBB Diagram based on [1] (where (1) & (2) represent margins on leak
detection and crack length)
Figure 3.2: Leak-Before-Break flowchart based on [77]

Typical inputs for LBB evaluation include pipe geometry, material properties, crack
morphology, cyclic loads, operating pressure and temperature. The different procedures
available are explained in the next section. However, having a common origin, some major
steps can be summarized as follows (from [1], [60], [78]):
1. Characterise/Postulate the initial flaw: Flaw dimensions have to be defined

for surface or through-wall flaws. Depending on the procedure used, surface defects
may be assessed in addition to through-wall defects by taking crack growth into
account.
2. Determine critical length of the through-wall flaw: This refers to the length
at which the through-wall defect becomes unstable, based on fracture mechanics
calculations, assuming stresses are load-controlled.
3. Estimate the flaw length at breakthrough: This is carried out by calculat-

ing the surface flaw length at which ligament failure is predicted to occur and re-
characterising this flaw as a through-wall flaw.
4. Determine detectable leakage length of the through-wall flaw: This includes

calculation of the Crack Opening Area (COA) associated with the crack length and
calculation of the resulting detectable leak rate appropriate to the leak detection
system capabilities. Time to detect the leak should be taken into account. The leak
rate may be estimated from relevant experimental data if available or computer
codes which predict leakage rates for single- or two-phase flows for a wide range of
through-wall defects that appropriately account for the surface roughness, number
of turns, etc for the crack mechanisms of interest.
5. Assess the results: A case for LBB is established provided the calculations in
previous steps show that:
• the flaw length at breakthrough is less than the critical length of the through-
wall flaw.
• the time to detect the leak is less than the time for the flaw to grow to the
critical length.
Guidance and established procedures are given in different standards and procedures
to resolve each of these steps. Depending on the procedure used, different methodologies
can be found with various levels of assessment and explicit margins may be given for each
step. This will be further elaborated in Section 3.3.
3.2.3 Limitations
According to the US requirement [70], LBB methodology has to be applied to the worst
location (lower material properties, higher stresses, etc.) in an entire system and should
not be applied to a particular location along the pipe system alone. The initial crack may
arise from different types of defect, and may grow as a result of loading or environment.
However, it is generally recommended to demonstrate that certain in-service degradation
mechanisms are not present: LBB is not usually applied to systems which experience
excessive or unusual loads and is generally limited to piping that is not susceptible to fail
from degradation mechanisms such as the following [79]:
• Water hammer
• Creep/ Creep fatigue damage
• Erosion, corrosion or erosion/corrosion
• Stress Corrosion Cracking (SCC) or Inter Granular SCC (IGSCC)
• Thermal ageing
• Brittle fracture
• Potential indirect sources of pipe rupture
Current LBB procedures only consider the case of a single crack. The scenario of
multiple cracks was considered by means of Monte Carlo simulation in [80], [81]. It was
concluded that this leads to a shorter leak-free time, a shorter allowable response time and
that catastrophic failure may also occur without a detectable leak where a large number
of initial cracks have a strong crack interaction. Therefore, a multiple crack LBB case
would be analysed case by case, and in most cases would probably not be possible. In
practice, many service cases of LBB occur with multiple cracks (generally due to Stress
Corrosion Cracking), but how to establish the guidelines for assessment of multiple cracks
is difficult. Another significant limitation is the requirement on minimum pipe radius. In
order to have sufficient margin between the leakage crack size and the critical crack size,
pipe dimensions have to be appropriate. In the case of a small pipe, the critical crack
length may be reached before the leak detected. In terms of pressure, LBB is generally
restricted to high energy components (such as in in the nuclear industry [70] Class 1 and
2, which are the classes containing the highest pressure) in order to improve the leak
detection aspect. However, it may also be applied to other significant components when
defence-in-depth is invoked.
3.3 Leak-Before-Break Procedures
3.3.1 Nuclear industry
In the last thirty years, LBB has received increased consideration as a criterion for as-
sessing or upgrading the safety of existing plants [82] and has also been applied to optimize
the design when large modifications were made. Various countries operating nuclear plant
have developed their own procedures regarding LBB applicability taking historic and reg-
ulatory aspects into account. Two reports published by the International Atomic Energy
Agency (IAEA) in 1993 and 1994 present the application of LBB in different countries
[69], [83]. Other publications also provide examples of LBB applications or developments
in other countries, for example France [84], [85], Belgium [86], USA [79], Canada [87],
[88], Korea [89], India [90] or Czech Republic [91] where the following procedures were
used with slight modifications. According to a recent report [92], published in 2012, four
distinct approaches of LBB are used throughout Europe in the nuclear industry. All these
approaches share some degrees of similarities:
• Germany - Integrity Concept (IC):The LBB procedure is a part of the prin-

ciple of break preclusion, now included in the overriding safety philosophy for the
assessment and structural integrity maintenance of nuclear power plants [78]. The
assessment itself is applied only to the most susceptible locations, such as welds, sec-
tions of increased load or material degradation. It employs the detectable leak rate
to determine the crack size required for confident detection and compares this to
the critical crack size. This deterministic approach and defence in depth argument
renders probabilistic approaches unnecessary [73].
• USA - SRP-3.6.3 [70]: This flaw tolerance approach requires a screening criterion
to be met (limitations stated earlier, Section 3.1) to ensure an incredibly low pipe
break probability. Similar to other LBB procedures, the basic principle is to use
the detectable leak rate to determine the crack size required for confident detection
and compares this to the critical crack size to ensure suitable margins exist. LBB
is only applicable to the entire section of piping, not individual components or
welded joints. The LBB procedure in SRP-3.6.3 can only be applied once the
reliability of the detection system has been demonstrated. The US procedure also
defines margin coefficients [93]: 10 on leak detection, and either 2 on crack size or
1.4 for (normal and seismic) stresses. These margins have been defined in order
to increase the degree of conservatism required in nuclear industry, and account
for other uncertainties not explicitly called out in the analyses (i.e., variability of
crack morphology parameters, not including cyclic effects from seismic loading on
toughness, etc.). Although published in the U.S., this procedure has been adopted
by some European countries [94].
• UK - R6 [60]: Two approaches are available. The first is a simplified ‘detectable

leakage approach’ used in design and safety studies of a hypothetical through-wall
flaw. It aims to demonstrate that a leaking flaw is detectable long before it grows
to a limiting length. The second approach is a ‘full LBB procedure’ that considers
the growth of an initial part penetrating surface flaw. In this case, it is necessary to
show that the flaw penetrates the pressure boundary before it can lead to disruptive
failure and that the resulting through-wall flaw leaks at a sufficient rate to ensure
its detection before it grows to a limiting length at which disruptive failure occurs.
Heavily based on sensitivity analyses within deterministic calculations, R6 does not
apply specific safety margins. The LBB procedure in R6 can usually only be used
as part of a defence in depth argument.
• France - RCC-MRX Appendix 16: Originally based on the American NUREG-
1061 [71], LBB estimates are determined by comparing the crack size that allows
the leak to be detected and that which causes component failure. This comparison
is quantified by a safety factor of 2 on crack sizes and 10 on the detection capability.
Some consideration of the crack dimensions are provided by taking into account
different internal and external crack lengths.
According to [92], [94], the German Integrity Concept (IC) and the American Standard
Review Plan (SRP) 3.6.3 form the basic foundation of LBB assessment for most countries
whereas the UK and French approaches are country-specific. Besides European and USA
procedures, Japan has also developed its own procedure:
• Japan – JSME S ND1: This guideline provides some unique features but the
overall principles are consistent with those previously mentioned. Stability analysis
is performed for the larger of the penetrating crack and leakage crack. Guidelines
on the crack growth are given and correction of the stress intensity factor at the
surface interaction has been included. A factor of 5 is applied to the leak rate and
no particular safety factor is specified for the crack length or applied stresses [95].
All LBB procedures described here share the same basis and are underwritten by a
large number of studies conducted in the US or Germany. However some differences are
noticeable. This may be explained by the different scope of applications. While LBB has
been developed in USA as a means to avoid DEGB behaviour and eliminate protections
such as pipe-whip restraints and jet-impingement shields; in most European countries
it has been used as a defence-in-depth argument. One important aspect to note is the
difference in safety factors used. Most countries use a safety factor of 10 on leak detection
and 2 on crack length (or applied stresses). Japanese guidelines recommend a factor of 5
on leak detection and 1 on crack length. In contrast, the UK procedure does not provide
explicit guidance on the safety factors.
3.3.2 Other industries (Non-nuclear)
In other industries, BS 7910 and API 579 are the most comprehensive, structured and
widely accepted FFS standards. They cover a broad range of equipment, and include a
LBB methodology which is linked to the assessment of crack-like defects:
• API 579-1/ ASME FFS-1 (Clause 9.5.2) [9]: Guidance is given to conduct a
simplified LBB assessment. Surface defects have to be re-characterized as through-
wall defects. Unique Crack-Opening Areas (COA) solutions are explicitly provided
for both elastic and elastic-plastic conditions in cylinders and spheres.
• BS 7910 (Annex F) [1]: LBB procedures are the same than those available in R6
(Detectable leakage approach and full LBB procedure) and guidelines are given for
each step. Heavily based on sensitivity analyses, no particular margins are explicitly
required.
Both API 579-1/ ASME FFS-1 and BS 7910 provide a procedure for LBB analysis
based on Failure Assessment Diagrams (FADs) for evaluating critical crack sizes and
three levels of assessment are available. However, there are some differences with respect
to parameters such as reference stress solutions and crack-opening-area solutions, both
of which can lead to different results. A notable difference is the starting point of the
analysis. In API 579-1/ ASME FFS-1, the starting point is a part-penetrating defect,
which is then re-characterized as a through-wall defect whereas in BS 7910, depending on
the procedure used, the starting point is either a through-wall defect or a part-penetrating
defect (consistent with R6). Procedures in these standards are generally limited by their
applicability. The assessor is often redirected to other publications or experimental work.
The leakage aspect is generally less detailed than the flaw assessment due to fields of
applications of these standards. For example, guidance on leak rate calculations is not
explicitly provided in API 579-1/ ASME FFS-1. The re-characterization rules used to
transform surface flaws to fully through-wall flaws also differ between API 579-1/ ASME
FFS-1 and BS 7910. In BS 7910, the length of the re-characterized through-wall flaw,
2cT W , will be equal to:
2cT W = 2cs + t (3.4)
where 2cs is the length of initial surface flaw and t the wall-thickness. In API 579-1/
ASME FFS-1, a more restrictive rule is used:
2cT W = 2cs + 2t (3.5)

3.4 Applying LBB to industries other than nuclear
Some available publications provide illustrations where LBB has been applied outside
the nuclear industry. A noticeable difference lies in the application of the concept. In
the nuclear industry, procedures are strictly followed step-by-step, whereas outside the
nuclear industry, each application seems to follow a hybrid procedure, taking aspects from
various sources. Some examples are described below:
• A simple case study on a spherical helium storage vessel was conducted [96], where
the loading was well within the yield limit of the material (301 stainless steel).
LBB demonstrated that the critical depth of a surface defect was larger than the
component thickness, for the crack lengths of interest.
• An LBB assessment of vertical cylindrical tanks for oil storage is provided in [97].
Investigations on crack growth are presented to show time margins between inspec-
tions. Critical defect length was calculated by means of simple fracture mechanics
formulation and time-factor in terms of fatigue (with a safety factor of 20 on fatigue
life).
• A large set of experiments were conducted on aluminium beverage cans (ratio of
diameter to thickness up to 590) in [98]. A good correlation between plates and
cans for axial crack growth was observed and the radial fracture toughness was
minimum and constant for crack aspect ratios greater than 70.
• Fitness-for-Service assessment on an ammonia storage tank in [99] used a LBB
argument in order to justify an inspection plan.
• A LBB case was compared with a fatigue analysis for an offshore structure [100].
The full development and growth of a circumferential crack in a tubular member
from a long deep surface crack to final failure was presented.
• The effect of pre-straining on subsea pipelines was also studied with small/full scale
testing in [101] to demonstrate that a LBB case could be made since the initial
assessment failed to provide sufficient margins.
• A recent published thesis looked at LBB applications on aerospace components such
as high pressure fuel lines and the fuel-to-oil cooler [102].
When LBB is used outside the nuclear field, the major steps detailed in Section 3.2
are followed. However none of established procedures such as those described in Section
3.3 has strictly been followed. Various sources are used and referred to while performing
assessments and this can lead to uncertainties (under- or over-estimation) on the final
results. Despite such assessment procedures being readily available in general standards
such as API 579-1/ ASME FFS-1 or BS 7910, the examples listed in the previous para-
graph did not appear to utilise them. It appears that there is a lack of awareness or
experience with full LBB general procedures outside the nuclear industry. This might
reflect a perception that they are either too complicated to apply or not sufficiently thor-
ough. In either case, there is a need to expose both established methods more to other
industries and to continue research and continuous improvements to these procedures
wherever possible.
3.5 Crack Opening Area
Crack Opening Area (COA) is a key element in a LBB assessment of pressurized

components. One important parameter in a LBB assessment is the current crack length
which governs the leak rate. During service or inspection, through-wall cracks can be
detected with a leak within the pressure envelope. Depending on the crack geometry, the
leak rate will vary and the relation between the leak-rate and the crack geometry has
to be established. This link is given through the Crack Opening Area, which is used to
link the leak-rate to the crack length. Therefore, for an accurate establishment between
leak-rate and current crack length, the COA has to be accurately predicted for different
loading conditions. Different models have been proposed to estimate the values of the
COA. They can be classified into three categories:
1. Linear elastic models
2. Elastic models including a small scale plasticity correction
3. Elastic-plastic models
The use of elastic solutions is deemed conservative in terms of leak detection.
3.5.1 Analytical model
The COA is calculated in terms of the crack tip opening displacement (δ) and crack
half-length (c) [103]. Most COA models for cylindrical shells are obtained from those for
flat plates, with some modifications for curved shells. Using linear elastic theory for a
thin-walled component, the COA is determined from the SIF as described below [104].
Its shape is often assumed to be elliptical (See Figure 3.3):
πδc
COA = (3.6)
2
A diamond-shaped profile may be more appropriate for circumferential cracks of length

greater than half the pipe circumference:
COA = δc (3.7)
Figure 3.3: Representation of CTOD and COA
The definition for strain energy release rate (G) is related to the potential energy (U )
and crack length (a). This can be replaced by two times the crack half-length (c):
1 ∂U 1 ∂U
G=− =− (3.8)
t ∂a 2t ∂c
Membrane forces, which play a dominant role in pressure vessels and pipes, can be
substituted by surface forces acting along the crack faces as presented on Figure 3.4.
Figure 3.4: Decomposition of forces acting on the crack faces [104]

The strain energy may, therefore, be expressed in the form of the surface integral:
Z
1
W =− (Ti × ui ) dSc (3.9)
2 Sc
where Ti and ui are respectively the surface traction and displacement vectors. The
integral extends over the surfaces Sc with non-zero forces, i.e. over the two crack sur-
faces. In the case of constant membrane stress, σ, and internal pressure, P , the integral
becomes: Z
1
W = − × (σ + P ) (ui × ni ) dSc (3.10)
2 Sc
The forces caused by the medium flowing through the crack are usually much lower
than the membrane forces and may be neglected for thin-walled vessels. By introducing
the average of COA, A, over the wall thickness, it gives:
Z
1
A= (ui × ni ) dSc (3.11)
t Sc
and the relation between the strain energy release rate, GI , and the stress intensity factor,
KI , is defined as:
K2
GI = I0 (3.12)
E
With this, a relation between the COA and SIF can be established:
(σ + P )E 0 ∂A
KI2 = (3.13)
4 ∂c
where E is the Young’s Modulus. If the SIF is known, COA can be readily calculated
from the integral: Z c
4
A= K 2 (x)dx (3.14)
(σ + P )E 0 0 I
The derivation shown here does not take into account the fact that, for curved shell
geometry, the SIF is dependent on a variable perpendicular to the shell surface [104]. This
variable accounts for the bulging effect, which causes the COA to be different on each side
of the shell. Using a mean value for the SIF can give an approximate value. In reality
COA is affected by geometrical, material and loading parameters. There are extensive
literature surveys dedicated to quantifying these effects. For a cylindrical shell, the SIF
for a crack is given often expressed in the form of:
p
KI = M (λ) × (σ + P ) (πa) (3.15)
where the bulging factor, M (λ), is a function of the shell parameter which depends on
the crack dimension (c), the thickness (t) and the mean radius (Rm ):
c
λ = Coef f icient × √ (3.16)
Rm t
Incorporating (Eq.3.15) into (Eq.3.14), the following expression for COA is obtained:
Z a
4π(σ + P )
COA = M 2 (λ) × a da (3.17)
E0 0
3.5.2 Leak rate and COA
In order to carry out an LBB assessment, it is important that the leak is detected
before the crack length reaches its critical value. The minimum value of the leak rate
Qmin is defined by the capability of the leak detection in place. This value is accompanied
by a safety factor of 10 in US regulation [93] to allow sufficient margins between the crack
length at leak detection and the critical crack length. Therefore the leak rate has to
be sufficiently large to be detected by the leak detection system. The leak rate can be
estimated as a function of the COA and the pressure difference using the following model
[93]: s
2∆P
Q=C ×A (3.18)
ρ
where C is an empirical flow coefficient set as 0.6 in terms of l/sec, mm2 , MPa and kg/m3
units. ∆P is the pressure difference, ρ is the mass density of the fluid and A is the COA,
recommended to be determined by using Zahoor’s or Paris-Tada’s models. Another model
is provided in R6 [60] for a single phase flow where Cdis a discharge coefficient, W and L
are the crack width and length of an equivalent rectangular crack respectively:
p
Q = Cdis P ×ρ×W ×L (3.19)
3.6 Past and future research in LBB
Over the last 30 years, a large number of research activities have been conducted in
regards to LBB in the nuclear industry. Various experimental and analytical studies
have been carried out on LBB behaviour of nuclear reactor components of non-cylindrical
geometries such as tees [105], piping elbows [106], [107] and shell nozzle junctions [108],
[109]. A summary of the different development and future research may be found in
[94] including attempts to validate LBB using the Integrity Concept, helium environment
experiments, crack shape development, crack length, crack opening area and leak rate. A
selection of other research activities is discussed in the following paragraphs.
On one aspect of a LBB procedure, fracture mechanics principles are used to assess
important parameters such as crack length at which the leak will be detectable and critical
crack length of a through-wall crack. Finite element modelling [110], large scale plate
experiments [111] and pipe experiments [112] have been conducted to assess crack shape
development. Complex shaped crack analysis can also be found in [113],[114].
On the other aspect of the assessment, fluid mechanics and fluid-structure interactions
are also important parameters in a LBB assessment. In order to detect the leak, leak
rates through cracks have to be evaluated to validate the leak detection system. Explicit
equations for leak rates through narrow cracks (single phase flow) have been developed for
four distinct flow regimes [115] using experimental work. The effect of crack morphology
and surface roughness are also studied in [116],[117]. It is concluded that improper mor-
phology parameters can result in large errors in the determination of leakage crack size,
leading to possible non-conservative margins. Different software codes such as DAFCAT,
PICEP, SQUIRT have been developed to calculate leak rates for various crack shapes or
fluids as described in [60], [93]. A recent doctorate dissertation [118] shows the effects of
thermal interaction between fluid and structure for a leaking fluid. A new finite element
was presented to give a convenient way to analyse this effect quickly with good accuracy
[119].
The interaction between structural integrity and fluid mechanics is evident in the crack
opening area (COA) parameter. Numerous models have been developed for plates, spheres
and cylinders with axial and circumferential defects. Three models are available: elastic,
elastic with plasticity correction [104], and elastic-plastic models [120]. High temperature
effects on the rate of change of the COA have been studied with the help of 2D finite
element modelling [121]. Complex geometries such as welded attachments are also taken
into account [122]. Simple plate estimations can provide conservative results, which imply
that detailed FE analysis may not always be required. A large amount of work has
been carried out to improve these deterministic approaches but an important effort has
also been undertaken in terms of probabilistic assessments. Probabilistic assessments
have been developed to strengthen deterministic assessments (i.e. ProLBB [123]) or to
enlarge their applicability (i.e. xLPR (extremely Low Pipe Rupture) [124]), taking into
account active degradation mechanisms, such as Primary Water Stress Corrosion Cracking
(PWSCC), that are generally out of scope for a standard LBB assessment. A statistical
treatment of material data, such as mechanical properties, crack resistance and fatigue
crack growth curves is presented in [125]. An example of the application of a probabilistic
assessment in accordance with the R6 procedure using Monte Carlo method is provided
in [126]. Application of LBB in CANDU (CANada Deuterium Uranium) reactors is given

in [127] or more recently in [128] to assess delayed hydride cracking.
Recent work undertaken to assess the safety aspects of Atucha II nuclear power plant
piping system under beyond design basis seismic loading [129], [130] showed surprising
results. It was found significant margins due to the LBB assumption that all stresses are
load-controlled while the pipe system behaved more like if it was displacement-controlled.
The critical flaw size changed from 20% of the circumference with a traditional LBB anal-
ysis, to 95% when the full FE model was developed, consistent with IPIRG experimental
work [131]. For circumferential flaws in a plant system, there are large margins with the
established methodologies where LBB analysis are based on a load-controlled behaviour
compared to detailed FE models.
As far as the continuity of LBB assessment is concerned, less work is currently being
undertaken and planned to be performed [94] in the near future. Present research is
generally limited to the following areas:
• Extension of COA solutions for R6 (high temperature creep and constraint effects)
• Dissimilar Metal Welds application as part of an EPRI/NRC programme
• Extension on probabilistic analysis
• Application of LBB to components which are not suitable for In-Service Inspection
including weld overlay
This leaves significant scope for further research to build on the existing knowledge,
particularly in the areas of comparing and optimising LBB procedures, assessing the
influence of welds on LBB and transition between part-penetrating and through-wall
flaws.
3.7 Discussion
LBB is widely recognised as an important methodology for supporting structural in-

tegrity safety cases in the nuclear industry. To use LBB as a fail-safe criterion, it has to
be demonstrated that that any credible defect would grow through the wall in a stable
way and create a detectable leak. The LBB criterion provides an extension to fracture
mechanics assessment for pressurized components, by demonstrating that leakage can oc-
cur in such components for a certain time period. It is important to note that even for
simple cases, LBB assessments may not be straightforward [132]. Detailed procedures
are established and numerous applications as defence-in-depth or to allow the non-use of
protection equipment can be found. However, performing a robust LBB assessment does
not seem to be a regular practice in industries outside nuclear, despite procedures being
available.
One reason for this may be that leakage of fluid through a penetrating defect will often
not be tolerable, for instance, when the fluid is toxic or flammable, or its leakage has
environmental consequences [133]. In the nuclear industry, processing fluids are generally
water, a mix of water and air or carbon dioxide. In other industries (such as oil and
gas, refinery or petrochemicals), leaking fluids are less acceptable in some circumstances
due to their nature: either environmental (pollution) or safety (risk of fire or explosion)
aspects cannot be accommodated with these procedures. The time-factor present in LBB
procedure, which allows the crack to grow from a detectable length to a critical length, is
unacceptable due to the increasing likelihood of ignition and explosion when flammable
fluids leak over a period of time. However, water-based products and steam are also
common fluids in refineries and these would suit LBB consideration. Another unique
feature within the nuclear industry is the confinement of the components. LBB is applied
mainly on primary circuit components. In other industries, for example petrochemical
and/or oil and gas, fluid-carrying components are generally in the open-air. This difference
has a strong influence on the leak detection capacity, since vapour cloud detection will
be affected by wind and weather conditions. Insulation of components can also increase
the difficulty of leak detection. As far as the fracture mechanics aspect is concerned,
examples presented in Section 3.4 show that evaluations are generally limited to estimation
of critical length by means of simple formulations, such as the one presented in Section
3.2. Established procedures clearly consider two aspects, namely fracture mechanics and
leakage assessment. This second aspect is generally less frequently addressed.
The margins used in LBB assessments are also subject to discussion. Most of the
established procedures, based on the American procedures, adopt a margin of 10 on leak
detection and a margin of 2 on the crack length. However, Japanese guidelines do not
apply any margins on crack length and a margin of 5 on leak detection. Some other pro-
cedures do not provide any explicit margins for LBB assessment. The higher the safety
margins adopted, the more conservative the assessment will be. However it will be more
difficult to satisfy LBB requirements if margins are too high. Load-controlled stresses
assumption and moment reductions with the presence of large circumferential flaws also
have to be considered in established LBB procedures large margins. The main use of the
LBB concept is to provide grounds for the assessment to determine the stability of pene-
trating defects and hence to use as a forewarning of catastrophic failure for components
especially when inspection is not possible or practicable. For the assessment to remain
valid the calculated leakage must be detectable and the consequences must be manage-
able within the context of the overall safety case. It may be useful to determine an upper
bound for a part-through flaw that is growing at an unknown rate. In this case, detection
of a leak will be an early warning. It is also worth considering the solutions/treatments
available. Reference stress and COA solutions are available for homogenous material.
However cracks are mostly found in weldments and such effects have to be taken into
account.
3.8 Summary
This chapter focused in details on the different aspects of Leak-Before-Break. Most

of the research on Leak-Before-Break has been carried out in the nuclear industry which
requires a high level of safety (Section 3.1). This industry has also developed more estab-
lished procedures (Section 3.3). Numerous situations and types of component have been
studied and LBB is generally applied to pressurized components containing benign fluids
such as water, steam and carbon dioxide (Section 3.6). Only a few published applications
can be found in non-nuclear industries (Section 3.4), certainly due to the limitations dis-
cussed in Section 3.7. However, LBB could be further developed in some areas of these
industries. When applied to non-nuclear industries, case studies did not appear to follow
established LBB procedures such as those presented in API 579-1/ ASME FFS-1 or BS
7910. A mix of different sources was generally used, which could lead to non-consistent
results. The margin between the smallest detectable leak size and the critical crack size
must be adequate to support LBB. Reliable leak detection methods must be employed to
ensure the ultimate success of the technique in order to prevent catastrophic failure.
Chapter 4
LBB parameters - Analysis of

current solutions from codes
Contents
4.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 58
4.2 Comparison of formulae for wall-thickness determination . . 59
4.2.1 Calculations of the required thickness . . . . . . . . . . . . . . 59
4.2.2 Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 61
4.3 Comparison of widely used COA models . . . . . . . . . . . . 62
4.3.1 Axial cracks . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 62
4.3.2 Circumferential cracks . . . . . . . . . . . . . . . . . . . . . . . 63
4.3.3 Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 65
4.4 Comparison of crack stability assessment solutions . . . . . . 66
4.4.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . 66
4.4.2 Failure Assessment Diagrams . . . . . . . . . . . . . . . . . . . 66
4.4.3 Stress Intensity Factor solutions . . . . . . . . . . . . . . . . . 70
4.4.4 Reference Stress solutions . . . . . . . . . . . . . . . . . . . . . 71
4.4.5 Cases study . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 84
4.5 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 96
57
CHAPTER 4. ANALYSIS OF CURRENT LBB SOLUTIONS FROM CODES 58
4.1 Introduction
A comparison study has been carried out to explore various concepts and standards for
pressurised equipment. The objectives of this study are to review and evaluate current
analytical solutions at design stage and to conduct Leak-Before-Break analyses on pipes
or pressure vessels with both longitudinal and circumferential cracks.
In the first section (Section 4.2), the solutions and relevant safety factors from the most
used pressure vessel codes have been evaluated to study their influence on the component
thickness at design stage. Important parameters in an LBB assessment have also been
reviewed and solutions available in standards and literature have been compared. As
discussed in Chapter 3, LBB can be seen as two independent aspects: (i) leakage and (ii)
crack stability assessment. Section 4.3 presents the comparison of available COA solutions
for both longitudinal and circumferential cracks in cylindrical components. Crack stability
assessment solutions are presented in Section 4.4: (i) Failure Assessment Diagram (Section
4.4.2), (ii) Stress Intensity Factor (Section 4.4.3) and (iii) Reference Stress (Section 4.4.4)
solutions are discussed. Finally, different cases studies are presented in Section 4.4.5.
For illustration and comparison purpose, a case of an internal pressure of 7.5 MPa is
presented. As only elastic COA are presented, the loading has to be kept low in order
to reduce the possible error due to the elastic assumption. The following geometric and
material parameters have been used.
Geometric parameters
Internal Radius Thickness Length Ratio

Parameters t
(Ri ) (t) (W ) Ri
Values 400 mm 20 mm 1000 mm 0.05
Material parameters
Yield Stress Ultimate Tensile Stress Young’s Modulus Poisson ratio

Parameters
(σys ) (σuts ) (E) (ν)
Values 450 MPa 530 MPa 210000 MPa 0.3
For every solutions presented, a Matlab script has been developed. This allowed to
calculate each solutions independently or a full assessment.
4.2 Comparison of formulae for wall-thickness deter-

mination
Various pressure vessels design codes are available across the industries. They aim to
ensure the component can operate in safe conditions. While the philosophy is similar
for all codes, solutions and safety factors provided may differ. As discussed in Chapter
2, pressure vessels have to be designed and built according to the Pressure Equipment
Directive (PED) in Europe and ASME Boiler and Pressure Vessel Code in North Amer-
ica. An example of the required wall-thickness calculation using the Design by Rule
(DBR) method where the overall dimensions and loads are known is given in this section.
Procedures provided by the following codes have been used:
• ASME Section VIII-Boiler & Pressure Vessel Code (Division 1, Division 2) [62]
• EN 13445 - Harmonised European code for unfired pressure vessels [63]
• PD 5500 - Specification for unfired pressure vessels [12]
Table 4.1 lists the different solutions provided by various design codes to calculate the
required thickness in cylinders. Also included is the equilibrium equation. Solutions from
the codes are compared against the hoop stress equation assuming a maximum stress
equal to the yield stress. This will represent the minimum thickness at which plasticity is
reached for a given pressure. Distance from this limiting thickness is calculated for each
code in order to show the margins applied.
Table 4.1: Required thickness for thin-wall cylinder according to different pressure vessel
codes
Codes Equations (Cylinder, internal pressure) Conditions

P Di
ASME Section VIII Div.1 t = 2.E.S−1.2P P < 0.385SE
q
Di SE+P
(UG-27 + Appendix 1) t= 2 SE−P − 1 P > 0.385SE
P Di
ASME Section VIII Div.2 t = 2S−P P < 0.4S
Di
exp P2 − 1

(AD-201) t= 2 P > 0.4S
P Di
EN 13445-5 t = 2ES−P −−−
P Di
PD 5500 t = 2S−P −−−
P Di
Cylinder (equilibrium equation) t = 2σys
−−−
4.2.1 Calculations of the required thickness
In order to calculate the minimum thickness required, it is necessary to determine the

allowable stress based on material properties and safety factors (See Table 2.1). Table
4.2 provides details of allowable stresses, S, for each code using minimum tensile material
properties and general safety factors (SFys , SFuts ) as described in Eq.4.1. The allowable
stress to consider for design is defined as:

σys σuts
S = min , (4.1)
SFys SFuts
Table 4.2: Allowable stress calculation
Codes UTS/SF YS/SF Allowable Stress (S) Units

ASME Section VIII Div.1 151.43 300 151.43 MPa
ASME Section VIII Div.2 220.83 300 220.83 MPa
EN 13445-5 220.83 300 220.83 MPa
PD 5500 225.53 300 225.53 MPa
Once the allowable stresses have been determined, the minimum required thicknesses
can be calculated with the different equations given in Table 4.1. Internal diameter is
assumed to be equal to 800 mm (consistent with dimensions given in Section 4.1), the
weld efficiency factor, E, equals to one and the internal pressure ranges from 1 to 20 MPa.
Results are presented in Figure 4.1 for an increasing internal pressure and Table 4.3 for
four selected pressures. The y-axis represents the minimum required thickness calculated
for each code normalized to the thickness calculated with the equilibrium equation.
Figure 4.1: Required thickness for different internal pressure normalized to cylinder
equilibrium solution
Table 4.3: Required thickness for four internal pressure normalized to cylinder equilib-
rium solution
Codes P=2 MPa P=5 MPa P=10 MPa P=20 MPa

ASME BPVC VIII - Div I 3.00 3.03 3.09 3.23
ASME BPVC VIII - Div II 2.05 2.06 2.08 2.13
EN-13445-3 2.05 2.06 2.08 2.13
PD 5500 2.00 2.02 2.04 2.09
Cylinder Equilibrium 1.00 1.00 1.00 1.00
4.2.2 Discussion
Depending on the design code used, safety factors applied to the yield or ultimate stress
have an significant impact on the allowable stress as well as on the minimum required
thickness. A large amount of comparison work has been conducted between the different
codes to update their procedures and to be consistent with each other, which has led
to identical values obtained using ASME BPVC Section VIII-Div.II, EN-13445-3 and PD
5500. On the other hand, ASME BPVC VIII-Div.I is known to provide more conservative
solutions and higher safety margins, thus requiring a greater minimum thickness. The
value of thickness is greater as per ASME BPVC Section VIII-Div.I. This is due to the
safety factor used which reduces the allowable stress.
The thickness of the component at a design stage has a great importance in terms of
leak-before-break assessments. As seen in Section 3, early concepts were based on the
defect size and thickness of the wall. Another remark is that for a given defect size, the
fracture behaviour is dependent on the thickness. For a thin structure containing a surface
crack, the collapse limit of a ligament is usually the main cause of structure rupture. On
the other hand, for a thicker structure, brittle fracture will have to be considered. In a
two-parameter assessment such as Failure Assessment Diagram described in the following
sections, this can have a significant effect. The final results will be more or less dependent
on the couples stress intensity factor/fracture toughness or reference stress/yield strength
parameters.
4.3 Comparison of widely used COA models
As discussed in Chapter 3, Leak-Before-Break assessment procedures consider two in-

dependent aspects to the potential failure which are respectively fracture mechanics and
leakage assessment. On the leakage side of the assessment, one of the main parameters
is the calculation of the crack opening area (COA). This parameter links the fracture
mechanics side of the assessment to the leakage side. It allows the user to investigate
the leak rate to ensure the crack detection or to show that the jet forces do not exceed a
critical value. The determination of leak rates is crucial for cracked pressure vessels and
pipes and requires the evaluation of the COAs. This has been a widely focused area of
research resulting in a wide range of models. This part of the assessment method review
focuses on three different elastic models that are available in BS 7910 [1] and API 579-
1/ASME-FFS-1 [9] as well as models from the literature recommended by the USNRC
guidelines [93]. Both longitudinal and circumferential models are discussed. Plasticity for
small scale yielding is also discussed. Equations for the crack opening area discussed in
this section are provided in Annex B.1.
4.3.1 Axial cracks
Axial cracks cracks develop along the longitudinal axis of the component. A repre-
sentation of the different geometrical parameters is given in Figure 4.2. Ri , Rm and Ro
being the inner, mid-thickness and outer radius respectively, 2a, the crack length and
δ the crack opening displacement. Considering a long pipe under internal pressure, the
membrane stress equals to the hoop stress with the following definition:
P Rm
Pm = (4.2)
t
Figure 4.2: Schematic representation of a pipe with a longitudinal through-wall crack

Results of the COA calculations for pipes/cylinders containing axial through thick-
ness defects using the three models discussed earlier are given in Figure 4.3. The x-axis
represents crack length, 2a, normalized to component’s length, W , and the y-axis repre-
sents the resulting COA. Results are plotted up to the limit of applicability of BS 7910.
Note that BS 7910 and Zahoor [134] give unique solutions at mid-thickness, representing
the average COA through-thickness. On the other hand, API 579-1/ASME FFS-1 pro-
vides solutions for both inner and outer surface. Both BS 7910 and Zahoor’s solutions
produce similar results at mid-thickness until half of the limit of applicability. These re-
sults are also coherent with API 579-1/ASME FFS-1 solutions at inner and outer surfaces.
However, when approaching this limit, both solutions tend to give a slightly greater COA.
The values from BS 7910 solution approach those of API’s solutions at the outer surface.
The effect of plasticity have been also evaluated for these axial models (See App.B). Only
elastic models are presented and compared against each other in this section. These are
known to under-estimate the COA for ductile materials.
Figure 4.3: Comparison of elastic COA solutions for a longitudinal through-wall crack
from API 579-1/ASME FFS-1, BS 7910 and Zahoor
4.3.2 Circumferential cracks
Circumferential cracks cracks develop along the axis of revolution of the component.
A representation of the different geometrical parameters is given in Figure 4.4. Ri , Rm
and Ro are the inner, mid-thickness and outer surface radius respectively, 2a, is the crack
length and δ is the crack opening displacement. Considering a long pipe under internal
pressure, the membrane stress is assumed to be equal to the longitudinal stress with the
following definition:
P Rm
Pm = (4.3)
2t
Figure 4.4: Schematic representation of pipe with a circumferential through-wall crack
The results of COA calculations for pipes/cylinders containing circumferential through-

thickness defects from the three models discussed earlier are summarised in Figure 4.5.
The x-axis represents the crack length, 2a, normalized to the component’s circumference,
2π/Rm , and the y-axis represents the calculated COA. Results are plotted up to the limit
of applicability of BS 7910. Note that both BS 7910 and Paris-Tada give an unique so-
lution at mid-thickness, representing the average COA through-thickness. On the other
hand, API 579-1/ASME FFS-1 provides solutions for both inner and outer surfaces. As
for the axial case, both BS 7910 and Paris-Tada solutions give coherent results as that
of API 579-1/ASME FFS-1 solutions until half of the applicability limit. After this, the
solutions tend to provide a lower COA values. All over the range, Paris-Tada model is
similar to BS 7910 but slighly lower. When approaching this limit, both solutions tend to
give a slightly greater COA. BS 7910 solution produces the same values os those of API’s
at the outer surface. It can be seen that BS 7910’s solution is well comprised between the
two bounds defined by API 579-1/ASME-FFS-1’s solutions.
Figure 4.5: Comparison of COA solutions for circumferential through-wall crack from
API 579-1/ ASME FFS-1, BS 7910 and Paris-Tada
4.3.3 Discussion
Crack Opening Area models for both axial and circumferential cracks have been com-
pared. In both cases, BS 7910 and the recommended solutions from the USNRC report
provide an averaged COA solution at mid-thickness whereas API 579-1/ ASME FFS-1
solutions are available for inner and outer surfaces. The pressure based equations in API
579-1/ ASME FFS-1 solutions,always give a slightly greater values than stress based equa-
tions. All solutions are coherent up to about half of the limit of applicability. After this
limit, solutions begin to diverge. For a given axial crack, BS 7910 and Zahoor solutions
tend to give a greater values with a COA at mid-thickness which is similar to API 579-1/
ASME FFS-1 at the outer surface. For circumferential cracks, BS 7910 and Paris-Tada
solutions tend to give lower values than that of API 579-1/ ASME FFS-1 with a COA at
mid-thickness lower than API 579-1/ ASME FFS-1 at the inner surface.
The effect of plasticity has been evaluated for axial models. BS 7910 and Zahoor
models include correction factors for small-scale yielding while API 579-1/ ASME FFS-1
propose a correction factor based on the plastic collapse. This results in a wider range
of application for the latter. It is important to remember that the COA solutions are
required to determine leak rate, which is a critical parameter in LBB assessment. An
over-estimation of the COA can lead to a non-conservative result (assuming there will be
sufficient leakage that could be detected when in reality there is very little) and inversely
an under-estimation could lead to a very conservative result. Elastic models are presented
here. These are known to under-estimate the real COA for ductile materials. For the
same crack length, the COA for circumferential defects is smaller. This explains why
circumferential defects are more critical in a LBB assessment than longitudinal defects.
For the same defect length, it will be easier to detect the leak from a longitudinal defect
than from a circumferential one.
4.4 Comparison of crack stability assessment solu-

tions
4.4.1 Introduction
Defect assessment procedures such as API 579-1/ASME FFS-1 [9] or BS 7910 [1] can
be used to assess the integrity of structures containing flaws, for example pipelines or
pressure vessels. The procedures used to assess crack-like flaws require the calculation of
the likelihood of failure by both fracture (brittle or ductile) and plastic collapse. In both
standards, the interaction between the two is expressed by the use of a Failure Assessment
Diagram (FAD). This diagram plots the proximity to fracture (via the fracture ratio, Kr )
vertically, and the proximity to plastic collapse (the collapse ratio Lr ) horizontally, and
an interaction line between the two. An assessment point inside the line is safe, while
outside the line is potentially unsafe. Different assessment levels or options are available
depending else on the level of qualifications of the assessor or the data available. A default
FAD is presented in Figure 4.6
With the use of high toughness materials and relatively low wall-thickness, fracture
tends to be a less concern than plastic collapse in modern pipes and pressure vessels.
The reference stress solutions used to assess the likelihood of failure by plastic collapse
should be safe, but as accurate as possible. With plastic collapse being the subject of
recent research, FAD Option 1 equation has been modified in the new BS 7910:2013. This
section presents the evaluation of the FAD (Section 4.4.2), SIF (Section 4.4.3) and RS
(Section 4.4.4) solutions provided in both procedures.
4.4.2 Failure Assessment Diagrams
One of the ways to assess the safety of structures containing flaws is to evaluate
the interaction between the likelihood of failure by both fracture (brittle or ductile) and
plastic collapse. In both standards, this is expressed by the use of a Failure Assessment
Diagram (FAD). It allows users to assess the critically of a given flaw. Different assess-
ment levels or options are available. Following some harmonization with R6 procedure,
BS 7910:2013 offers now three alternative routes, known as “Options”, to carry out frac-
Figure 4.6: Default Failure Assessment Diagram
ture assessments. Option 1 requires basic information and is divided into continuous or
discontinuous yielding material, whereas Option 2 requires full stress-strain data for the
material under consideration. Option 3 which is recommended for specific cases as an
alternative to the two previous, but requires both elastic and elastic-plastic analysis with
the aid of numerical analysis to derive crack driving forces. The current edition of API
579-1/ASME FFS-1 offers three “Levels” depending on the available material properties
and assessor skills. Level 1 is a quick screening method for the assessment, whereas Level
2 is a more sophisticated analysis using a generic FAD. A Level 2 analysis requires stresses
in terms of membrane and bending components and partial safety factors are applied to
the independent variables to account for uncertainty. Level 3 offers five methods to pro-
vide the best estimate of structural integrity including a method using material’s specific
FAD similar to BS 7910’s Option 2 or elastic plastic analysis similar to Option 3.
The different routes to calculate FADs are presented below. It is worth noting that
the equation of API’s FAD for a Level 2 assessment is the same than the one in previous
edition of BS 7910 (2005). BS 7910 Option 1 (Eq.4.4) and API 579-1/ASME FFS-1 Level
2 (Eq.4.5) can be regarded as equivalent in terms of data requirements. However, it can
be observed that BS 7910 formulation is dependent on material properties while API’s is
only defined via Lr .
BS 7910:2013 (Option 1):

 6
0.3+0.7e−µLr )
 √ for Lr ≤ 1
1+0.5L2r



N −1
Kr = f1 (1) × Lr2N for 1 < Lr < Lrmax (4.4)



0 for Lr ≥ Lrmax


0.001E
where N is the work hardening coefficient and µ defined by µ = min σys
, 0.6.
API 579-1/ASME-FFS-1:2007 (Level 2):


 0.3+0.7e−0.65L6r for L < L
1−0.14L2r r rmax
Kr = (4.5)
0 for L ≥ L
r rmax
In both codes, the cut-off value in terms of Load Ratio is defined in the same way as:
σf low
Lrmax = (4.6)
σys
Parametric studies show the influence of Young’s modulus (Figure 4.7), yield strength
(Figure 4.8) and ratio of yield strength to ultimate tensile strength (Figure 4.9) on the
FAD’s shape. In most cases, BS 7910 FAD curve tends to be slightly lower for 0 < Lr <
0.8, higher for 0.8 < Lr <1 and lower for 1 < Lr than API 579-1/ASME FFS-1. This
allows potential larger flaws in the region just before plasticity when using that of BS
7910, while more conservatism is applied when Lr > 1.
Figure 4.7: Influence of the Young modulus (E) on the shapes of FADs
Figure 4.8: Influence of the Yield Stress (YS) on the shapes of FADs
Figure 4.9: Influence of the ultimate tensile strength (UTS) on the shapes of FADs
4.4.3 Stress Intensity Factor solutions
Stress Intensity Factor, KI , solutions for a wide range of geometry and loading config-
urations are available in many handbooks or standards. Alternatively, numerical analysis
methods may be used to derive solutions. This helps to establish the state of the stress at
the location of a crack. A comparison of SIF solutions for various geometries can be found
in Ref.[59]. For this reason, only solutions for an axial through-wall crack in a cylinder
will be presented and calculations provided in this section. Some remarks will be added
to assist the understanding of the assessment method.
Stress Intensity Factor solutions are functions of the applied primary stresses (Pm , Pb ),
the defect size (2a, 2c) and the structural dimensions (t, W , Ri ). Procedures as API 579-
1/ASME FFS-1 or BS 7910 contain solutions for many crack geometries that are likely
to occur in large range of components including pressurized components. All the SIF
equations from API 579-1/ASME FFS-1 or BS 7910 can be found in Annex B.2
Axial through-thickness flaws
Axial through-thickness defects are defined in the same way in both standards.
Geometrical parameters such as crack length (2a), internal radius (Ri ), length (W ) and
thickness (t) are required. Geometrical parameters are presented in Figure 4.11.
The results of SIF calculations for pipes/cylinders containing a through-thickness defect

are summarised in Figure 4.10. All solutions provided similar results. Table 4.4 provides
the numerical values for some crack lengths. BS 7910 solution has been taken as reference
and compared with API solutions. A maximum difference up to 4% for the stress input
solution and up to 2% for the pressure input solution can be observed on the range of
study.
Table 4.4: Stress Intensity Factor solutions from API 579-1/ASME FFS-1 and BS 7910
for an axial through-wall crack
BS 7910 API 579-1 API 579-1

2a/W Eq.B.35
√ Stress - Eq.B.33
√ Diff. (%) Pressure - Eq.B.36
√ Diff. (%)
KI (M P a. mm) KI (M P a. mm) KI (M P a. mm)
0.05 1575.6 1587.6 -0.8 1579.5 -0.2
0.1 2576.4 2668.9 -3.6 2622.3 -1.8
0.25 4949.7 5093.1 -2.9 4973.2 -0.5
0.5 13441 13586 -1.1 13527 -0.6
Figure 4.10: Stress Intensity Factor solutions for an axial through-wall crack (API
579-1/ASME FFS-1 and BS 7910)
4.4.4 Reference Stress solutions
The reference stress, σref , is a measure of the applied stress in the un-cracked region
of a cracked component. It accounts for the loss of cross sectional area in the region
containing a flaw. Reference stress solutions are functions of the applied primary stresses
(Pm , Pb ), the defect size (2a, 2c) and the structural dimensions (t, W , Ri ). API 579-
1/ASME FFS-1 or BS 7910 provide reference stress solutions for many crack geometries
that are likely to occur in pressurized components. As no comparison work has been
previously performed for this parameter, the following section includes such discussion
for the different geometries. All the SIF equations from API 579-1/ASME FFS-1 or BS
7910 can be found in Annex B.3
4.4.4.1 Axial through-thickness flaws
Axial through-thickness defects are defined in the same way in both standards. Ge-
ometrical parameters such as crack length (2a), internal radius (Ri ), length (W ) and
thickness (t) are required. Geometrical parameters are presented in Figure 4.11. Equa-
tions from BS 7910 and API 579-1/ASME FFS-1 can be found in Annex B.
Figure 4.11: Geometry of an axial through-thickness crack in a cylinder
To assess the reference stress for through-thickness defects, both codes refer to Willoughby
and Davey [135] which is a solution derived for plates with a correction factor to fit
to pipes/cylinders geometries. Two important points may be noted here. In BS 7910
(Eq.B.51), a factor of 1.2 applied to the membrane stress is intended to produce a similar
level of conservatism as that is inherent in the flat plate solutions. This increases the ref-
erence stress in BS 7910 by 20% compared to that of API 579-1/ASME FFS-1. Assuming
a negligible bending stress (Pb = 0), reference stress solutions are reduced to:
• BS 7910 (From Eq.B.51)

σref (Pb =0) = 1.2MT Pm (4.7)
• API 579-1/ASME-FFS1 (From Eq.B.53)
σref (Pb =0) = Mt Pm (4.8)
Moreover, the use of surface correction factors leads to further divergences. In BS

7910, a unique surface correction factor, MT , is provided (Eq.B.52) whereas three different
factors are available in API 579-1/ASME FFS-1. Note that (Eq.B.54) is recommended
for use in all assessments, this being a best fit to data from different references, (Eq.B.55)
is an approximate expression and (Eq.B.56) an upper bound expression. The latter being
the same surface correction factor than in BS 7910, (Eq.B.52).
From (Eq.B.57):
a2
λ2 = 3.305 (4.9)
Ri t
Inserting (Eq.4.9) into (Eq.B.56) leads to:
s s
√ a2 ∼ a2
Mt(AP I579) = 1 + 0.4845λ2 = 1 + 1.5997 = 1 + 1.6 = MT (BS7910) (4.10)
Ri t Ri t
The results of reference stress calculations for pipes/cylinders containing through-wall

defect are presented in Figure 4.12 to Figure 4.14. The length of defect, 2a, was normalized
to the length of the pipe/cylinder, W . BS 7910 solution gives higher value of reference
stress than all API 579-1/ASME FFS-1 solutions. Removing the 1.2 factor, the modified
solution matches with API 579-1/ASME FFS-1 solution using the third surface correction
factor (Eq.B.56) and API 579-1/ASME FFS-1 equation. API’s two first surface correction
factors (Eq.B.54 & Eq.B.55) give similar result. Taking the recommended correction
factor, a difference of around 20% is shown for small crack length between BS 7910
and API 579-1/ASME FFS-1. This difference is due to the factor 1.2 on the membrane
stress component. As the crack length increases, this difference increases due to the
different surface correction factors used. In an example presented here with membrane
stress loading only, the difference between BS 7910 and API 579-1/ASME FFS-1 solutions
reaches 40% at Lr =1. Assuming only bending stress is present, it has to be noted that
API 579-1/ASME FFS-1 equation results in a constant value equal to 0.66 Pb while BS
7910 equation is varying with the defect length.
This means that for all loading cases, BS 7910 is more conservative than API 579-
1/ASME FFS-1. It will give a smaller critical flaw sizes before plastic collapse is predicted
to occur.
Figure 4.12: RS - Pipe/cylinder containing through-thickness axial defect - Membrane

stress only
Figure 4.13: RS - Pipe/cylinder containing through-thickness axial defect - Membrane

and Bending stress
Figure 4.14: RS - Pipe/cylinder containing through-thickness axial defect - Bending

stress only
4.4.4.2 Circumferential through-thickness crack
Circumferential through-thickness defects are defined in the same way in both stan-
dards. Geometrical parameters such as crack length (2a) or crack angle (2θ), internal
radius (Ri ) and thickness (t) are required. They are presented in Figure 4.15.
Figure 4.15: Geometry of a circumferential through-thickness crack in a cylinder
The second through-wall geometry examined a circumferential defect in a pipe/cylinder.

Since the procedure requires applied loads to be expressed in terms of membrane and bend-
ing stresses, the following conversion formulae was used for pressure loading accounting
for internal pressure, P and axial force, F [9]:
P Ri2 F
Pm = 2
+ (4.11)
2
Ro − Ri π(Ro − Ri2 )
2
where Ri and Ro are respectively the inner and outer radii. Within this part of the
study, pipes/cylinders with the same dimensions and material properties than for axial
defects cases have been assumed. The loading conditions considered were membrane
and bending stresses of 150 MPa (Pm = 150 MPa, Pm = Pb = 150MPa and Pb = 150
MPa).
Results of the reference stress calculations for pipes/cylinders containing circumferen-

tial through-thickness defect are presented in Figure 4.16 to Figure 4.18. For the analysis,
BS 7910 (Eq.B.58) and API 579-1/ASME FFS-1 (Eq.B.59) formulae were employed to
carry out the assessment. The geometrical normalisation for circumferential flaws is made
by the division of the half crack angle, θ, by π. The assessment results of the pipe/cylinder
containing a through thickness defect subject to pure tension, combined tension and bend-
ing, and in addition to combined tension, bending and internal pressure are summarized
in Figure 4.16 , Figure 4.17 and Figure 4.18 respectively.
For the case where through-thickness defects are subject to pure tension (Figure 4.16),
both BS 7910 and API 579-1/ASME FFS-1 delivered very similar results. However, signif-
icant difference was observed in the assessment of combined tension and bending loading
(Figure 4.17). This is mainly due to the definition of α in API 579-1/ASME FFS-1, which
is two times lower than BS 7910 equivalent and the presence of the squared denomina-
tor (See equation below). Taking only bending stress into consideration, (Eq.B.58) and
(Eq.B.59) were reduced to:
2Pb 2Pb
σref −BS = 6= σref −AP I = (4.12)
2a
3(1 − πRi ) 3(1 − πRam )2
As a whole, BS 7910 produces higher reference stress results. However the evolution of
solutions is similar in the range of applicability for both standards. The same observation
can be made when internal pressure is taken into account together with tension and
bending (Figure 4.18).
Figure 4.16: RS - Pipe/cylinder containing through-thickness circumferential defect -

Pure tension only

Membrane and Bending stress

Combined tension, bending and internal pressure
4.4.4.3 Axial surface defect
Axial surface defects are defined in the same way in both standards. Geometrical
parameters such as crack length (2c), crack depth (a) internal radius (Ri ) and thickness
(t) are required. They are presented in Figure 4.19. Equations from BS 7910 and API
579-1/ASME FFS-1 can be found in Annex B.3.3. These solutions are generally applicable
for 0< a/t <0.8.
Figure 4.19: Geometry of an axial surface crack in a cylinder
Limit load solutions can be complicated for surface defect assessment as it can be
defined as the load corresponding to local yielding (‘local’ limit load), or net section
yielding, (‘global’ limit load). In the case of a surface defect, in contrast to API 579-
1/ASME FFS-1 (Eq.B.68), BS 7910 distinguishes internal and external defects even if
the same solution is provided (Eq.B.65). None of the above allows consideration of the
defect face pressure when an internal defect is present in a pressurized component. As per
through-thickness solution, BS 7910 provides a unique surface correction factor, while API
579-1/ASME FFS-1 provides different solutions. Information given in API 579-1/ASME
FFS-1 concludes that the first surface correction factor Ms1 does not produce consistent
results when the crack approaches a through-wall configuration. It is also mentioned that
Ms1 and Ms2 give similar results up to a ratio a/t < 0.5. The formulation Ms1 with
a rectangular shape is the surface correction factor used in BS 7910. Here, the second
surface correction factor in API 579-1/ASME FFS-1 Ms2 will be used to keep the same
formulation for different crack depths. Three different cases corresponding to each surface
correction factors Mt ((Eq.B.54) to (Eq.B.56)) have been assessed in this work.
The results of reference stress calculation for pipes/cylinders containing surface defect
can be seen in Figure 4.20 to Figure 4.23. The length of internal surface defect, 2c, is
normalized to the width, W , of the pipe/cylinder and the defect depth, a, is normalized to
the wall-thickness, t. Figure 4.20 shows the effect of surface correction factors available in
API 579-1/ASME FFS-1 on Lr . Calculations presented with Eq.B.71 have been performed
using C=0.85. One can see that the recommended equations tend to lower reference stress
results. Note that the couple Eq.B.54 and Eq.B.72 are recommended for all assessment.
Comparison with the BS 7910 solution will be made using this couple of surface correction
factors only.
In Figure 4.21 to Figure 4.23, Lr vs a/t values of membrane only, membrane and
bending and bending only cases were plotted against increasing 2c/W values. As long as
membrane stress is involved, BS 7910 gives higher reference stress values. This is partly
due to the 1.2 factor applied on membrane stress as per through-thickness case. Moreover
as the crack length increases, the values derived using BS 7910 solution tends to increase
more rapidly for defects with a depth superior to the half thickness (Figure 4.21). Another
observation can be made regarding the effect of bending stress (Figure 4.22 - Figure 4.23).
Values calculated from API 579-1/ASME FFS-1 solutions are lower due to the parameter
g. Without this parameter applied on bending stress, solutions give similar results.
Figure 4.20: RS - Pipe/cylinder containing surface axial defect - API 579-1/ASME

FFS-1 - Membrane stress only
Figure 4.21: RS - Pipe/cylinder containing surface axial defect - Membrane stress only
Figure 4.22: RS - Pipe/cylinder containing surface axial defect - Membrane & Bending
stress
Figure 4.23: RS - Pipe/cylinder containing surface axial defect - Bending stress only
4.4.4.4 Circumferential surface defect
Circumferential surface defects are defined in the same way in both standards. Geo-
metrical parameters such as crack length (2c) or crack angle (2θ), crack depth (a) internal
radius (Ri ) and thickness (t) are required. They are presented in Figure 4.24. Equations
from BS 7910 and API 579-1/ASME FFS-1 can be found in Annex B.3.4.
Figure 4.24: Geometry of a circumferential surface crack in a cylinder
Both BS 7910 (Eq.B.74) and API 579-1/ASME FFS-1 (Eq.B.76) distinguish internal
and external cases for circumferential surface defects. None of the above allows consid-
eration of defect face pressure in case of an internal defect in a pressurized component.

Only internal defect will be illustrated here. The results of reference stress calculation
for pipes/cylinders containing circumferential surface defect can be seen in Figure 4.25 to
Figure 4.27. The geometrical normalisation for circumferential defect is achieved by the
division of the half defect angle, θ, by π and the defect depth, a, is normalized to the
wall-thickness, t. The assessment results of pipe/cylinder containing a surface defect sub-
ject to membrane, combined membrane and bending, and bending only can be found in
Figure 4.25, Figure 4.26 and Figure 4.27 respectively. Considering membrane stress only,
both solutions give similar results for small defect depth but diverge significantly as the
defect depth increases (Figure 4.25). Solution in API 579-1/ASME FFS-1 increase much
slower than BS 7910’s one. For combined membrane and bending stress loading, BS 7910
solution gives higher values than that of API 579-1/ASME FFS-1 (Figure 4.26). When
only bending stress is considered, both solutions give similar results (Figure 4.27).
Figure 4.25: RS - Pipe/cylinder containing surface circumferential defect - Membrane

stress only
Figure 4.26: RS - Pipe/cylinder containing surface circumferential defect - Membrane

& Bending stress
Figure 4.27: RS - Pipe/cylinder containing surface circumferential defect - Bending

stress only
4.4.5 Cases study
Three relevant cases studies are provided in this section. In Section 4.4.5.1, a full
case study is described including comparison of reference stress, stress intensity factor
and final FAD solutions comparison for an axial through-wall defect. The influence of
different solutions are discussed. Section 4.4.5.2 provides more details on surface axial
defect analytical solutions. The third (Section 4.4.5.3), looks at analytical solutions for a
surface axial defect versus experimental data.
4.4.5.1 Axial Through-wall crack - Comparison of the different analytical

solution
In order to compare formulations found in both codes, a parametric study on the crack
length has been carried out. A case of a pipe with a ratio t/Ri = 0.05 under an internal
pressure of 2.5 MPa is shown here. In this particular case, the following parameters
describing the pipe geometry and loading were used to solve equations from BS 7910 and
API 579-1/ASME FFS-1.
Geometric parameters:
Parameters Internal Radius Thickness Part Length Ratio

t
(Ri ) (B) (W) Ri
Values 200 mm 10 mm 1000 mm 0.05
Loading parameters:
Parameters Pressure Membrane Stress Bending Stress

(P) (Pm ) (Pb )
Values 2.5 MPa 50 MPa 1.25 MPa
Material parameters:
Parameters Yield Stress Ultimate Tensile Stress Fracture Toughness Cut-off

(σys ) (σuts ) (Kmat ) (Lrmax )
√
Values 450 MPa 530 MPa 100 MPa m 1.09
Using different crack lengths, a parametric study can be conducted on the reference
stress solutions in API 579-1/ASME FFS-1 and BS 7910. Figure 4.28 shows the reference
stress solutions plotted against the normalized crack length, 2a/W . Five curves are plotted
here:
• (I) BS 7910:2013 (Eq.B.51)

• (II) BS 7910:2013-Modified (Eq.B.51 with the 1.2 factor removed)
• (III) API 579-1/ASME FFS-1:2007 (Eq.B.53 with Mt1 from (Eq.B.54))
• (IV) API 579-1/ASME FFS-1:2007 (Eq.B.53 with Mt2 from (Eq.B.55))
• (V) API 579-1/ASME FFS-1:2007 (Eq.B.53 with Mt3 from (Eq.B.56))
Figure 4.28: Reference stress solutions from BS 7910 and API 579-1/ASME FFS-1 for
an axial through-wall crack
Figure 4.28 shows that the BS 7910 solution gives higher values of reference stress than
all the API 579-1/ASME FFS-1 solutions. The modified BS 7910 solution matches with
API 579-1/ASME FFS-1 solution using the third surface correction factor (Eq.B.56). API
579-1/ASME FFS-1 solution using the two first surface correction factors ((Eq.B.54) &
(Eq.B.55)) give approximately the same results. Table 4.5 lists the values of reference
stress for some normalized crack lengths. Values obtained from API 579-1/ASME FFS-1
solutions are compared to those obtained by the BS 7910 solutions. A difference of around
20% is shown for small crack lengths. This difference is due to the factor of 1.2 on the
membrane stress part in Eq.B.51. As the crack length increases, this difference increases
due to the different surface correction factors used. In this example, the difference between
the formulation recommended for all assessments in API 579-1/ASME FFS-1 and BS 7910
reaches 40%. Removing the 1.2 factor from the BS 7910 formula, brings the values down
to the same as in API 579-1/ASME FFS-1 with the third surface correction factor.
Table 4.5: Comparison of reference stress between BS 7910 and API 579-1/ASME FFS-1
for axial through-wall crack
API 579-1/ASME FFS-1 BS 7910 - Mod BS 7910

Normalized crack length
σref (MPa) σref (MPa) σref (MPa)
(2a/W ) Mt1 % Mt2 % Mt3 % MT % MT
0.05 60.7 18.7 59.6 20.2 61.9 17.1 62.4 16.5 74.7
0.1 81.9 21.9 80.2 23.5 87.0 17.0 87.5 16.5 104.8
0.25 156.9 29.2 158.4 28.5 184.2 16.9 184.8 16.6 221.6
0.5 273.7 36.4 265.4 38.3 357.6 16.9 358.7 16.6 430.2
0.75 388.1 39.6 390.4 39.2 533.3 17.0 536.0 16.6 642.6
Figure 4.29 shows the result in terms of load ratio, the value of Lr cut-off being the
same in both codes (Eq.4.6). The difference between the reference stress solutions has an
effect on the crack length at which the cut-off value is reached.
Figure 4.29: Load ratio and cut-off solutions from BS 7910 and API 579-1/ASME FFS-1
for an axial through-wall crack
Here, the cut-off value is a normalised crack half-length of around 0.57 in BS 7910
whereas it reaches a value around 0.95 in API 579-1/ASME FFS-1. With the modified
equation of BS 7910, this value reaches approximately 0.68. This is mainly due to the
surface correction factor Mt definition. The next step in order to assess the defect in terms
of the FAD is the calculation of the fracture ratio Kr . Stress Intensity Factors solutions
have to be known for each code (Figure 4.30).
Figure 4.30: Stress Intensity Factor solutions from BS 7910 and API 579-1/ASME
FFS-1 for an axial through-wall crack
Table 4.6 gives the values of stress intensity factor for some lambda values obtained
from API 579-1/ASME FFS-1 solutions and are compared to those obtained by the BS
7910 solution. Different solutions give approximatively the same result with a maximum
difference of 5%.
√
Table 4.6: Comparison of stress intensity factor, KI (M P A m), between BS 7910 and
API 579-1/ASME FFS-1 for axial through-wall crack
API 579-1/ASME FFS-1 API 579-1/ASME FFS-1 API 579-1/ASME FFS-1

Lambda BS 7910
(Pressure) (Stress) (Stress+P)
(λ) KI % KI % KI % KI
0.25 242 1.98 235 4.77 247 0.09 247
0.5 369 2.32 364 3.45 383 1.46 377
1 610 0.55 613 0.09 644 4.97 613
2 1153 1.67 1170 0.24 1229 4.80 1173
4 2394 2.86 2474 0.38 2598 5.43 2464
6 4073 0.63 3998 2.47 4197 2.40 4099
8 6239 2.83 6115 4.76 6420 0.007 6420
10 8730 3.55 8579 5.22 9007 0.48 9051
12 11613 3.10 11511 3.95 12086 0.85 11984
With the previous assumption, the calculation of the fracture ratio Kr is limited to
Kr = KIp /Kmat . The following figure shows the evolution of Kr with the crack half-
length. The cut-off value in terms of Kr is obtained for Kr =1. The difference between
API 579-1/ASME FFS-1 and BS 7910 is small, with a cut-off value around 249 mm in
API 579-1/ASME FFS-1 (Pressure, Stress), 237 mm in API 579-1/ASME FFS-1 (Stress
+ CFP) and 250 mm in BS 7910.
Figure 4.31: Fracture ratio solutions from BS 7910 and API 579-1/ASME FFS-1 for an
axial through-wall crack
Having calculated the values of Lr and Kr for different values of crack length, the last
step is to plot the FAD. In order to compare results obtained by API 579-1/ASME FFS-1
and BS 7910, a material specific FAD could have been chosen as the same definition in
provided both codes. However, here the simplest definition from each code is chosen.
Figure 4.32: FAD comparison between API 579-1/ASME FFS-1 and BS 7910 for an
A critical value is reached when assessment point reaches the FAD curve. This critical
value is different depending on the code used as shown in Table 4.7:
Table 4.7: Critical length for each standard for axial through-wall crack
Critical length Critical length

Codes based on API 579-FAD based on BS 7910-FAD
BS 7910
(Eq.B.29 & Eq.B.51) 244 mm 240 mm
API 579-1/ASME FFS-1
(Eq.B.34 without CFP & Eq.B.53) 244 mm 242 mm
(Eq.B.34 with CFP & Eq.B.53) 234 mm 230 mm
(Eq.B.36 & Eq.B.53) 246 mm 244 mm
Based on API 579-1/ASME FFS-1 (pressure) the critical crack length is determined
as 246 mm whereas in BS 7910 the critical value is reached for a crack length of 240 mm.
Here a crack length of 243 mm will fail with an assessment using BS 7910 procedures
whereas it will be assessed as safe using API 579-1/ASME FFS-1 procedures. One can
also notice that using the same input parameters, API 579-1/ASME FFS-1 gives different
results depending on the solution used. This difference is due to the assessment of the
critical length of the defect and may be a major concern in a LBB case. In a LBB case,
the critical length value of a defect is one of the most important parameters to validate the
leak detection system and margins associated with the time remaining before failure.
When the different parameters are taken in order to be in the FAD area where the
solutions are the more diverging, the choice of the solution may be important. The fol-
lowing case presents the same component as previous in terms of size. However, both
material properties and loading have been changed so that plastic collapse becomes dom-
inant. The material is assumed to have a yield strength, σys = 275 MPa and an ultimate
tensile strength σys = 350 MPa. The component is subject to an internal pressure of 15
MPa. From these hypothesis, the following FAD can be computed:
Figure 4.33: FAD comparison between API 579-1/ASME FFS-1 and BS 7910 for an
Table 4.8: Critical length for each standard for axial through-wall crack
Critical length Critical length

Codes
(API 579-FAD) (BS 7910-FAD)
BS 7910 (Eq.B.29 & Eq.B.51) — —
BS 7910-Mod (Eq.B.29 & Eq.B.51 - 1.2 factor removed) 9.8 mm 5.3 mm
API 579-1 (Eq.B.34 without CFP & Eq.B.53) 10.3 mm 5.6 mm
API 579-1 (Eq.B.34 with CFP & Eq.B.53) 9.7 mm 5.1 mm
API 579-1 (Eq.B.36 & Eq.B.53) 9.7 5.1 mm
Table 4.8 presents the final results for this selected case. BS 7910 solutions assumes
that the component will fail by plastic collapse while API 579-1/ASME FFS-1 allows a
through-wall crack to a certain extend. The effect of the FAD shape is also important.
Depending on FAD definition, the crack length allowable can be doubled. Removing the
1.2 factor from BS 7910 equations leads to results that are similar to API 579-1/ASME
FFS-1.
4.4.5.2 Axial surface defect - Analytical solutions
The reference stress solutions for an axial surface crack are described in this section. In
order to maintain the same input parameters as described in Section 4.4.5.1, only internal
pressure will be considered for three cases based on the crack depth (a/t = 0.2, 0.5 and
0.8 ). As previously, the curve “BS 7910-Mod” is based on (Eq.B.65) with the factor 1.2
removed.
Figure 4.34: Comparison of reference stress between BS 7910 and API 579-1/ASME
FFS-1 for axial surface crack (a/t=0.2)
Table 4.9: Comparison of reference stress between BS 7910 and API 579-1/ASME FFS-1
for axial surface crack (a/t=0.2)
Lambda σref (MPa) σref (MPa) σref (MPa)
Mt1 % Mt2 % Mt3 % MT % MT
0.5 51.14 16.40 51.02 16.61 51.15 16.39 51.14 16.40 61.18
1 52.22 15.35 52.05 15.62 52.38 15.08 51.58 16.39 61.69
2 54.69 13.61 54.53 13.85 55.11 12.94 52.94 16.37 63.30
5 58.43 14.01 58.47 13.96 59.07 13.07 56.83 16.37 67.96
10 60.38 15.66 60.30 15.78 61.00 14.80 59.87 16.38 71.60
20 61.61 16.57 61.77 16.36 62.05 15.98 61.75 16.38 73.85
Table 4.10: Comparison of reference stress between BS 7910 and API 579-1/ASME

0.5 51.99 17.73 51.67 18.24 51.98 17.75 52.91 16.29 63.20
1 55.18 18.94 54.72 19.61 55.69 18.20 57.03 16.22 68.08
2 62.32 21.19 61.82 21.82 63.73 19.40 66.27 16.20 79.08
5 76.20 23.73 76.36 23.56 79.01 20.91 83.71 16.21 99.90
10 84.88 23.42 84.48 23.78 88.07 20.54 92.86 16.22 110.84
20 91.33 21.93 92.23 21.17 93.81 19.82 98.02 16.22 117.00
Table 4.11: Comparison of reference stress between BS 7910 and API 579-1/ASME

0.5 53.07 27.12 52.55 27.84 53.03 27.17 61.10 16.10 72.82
1 58.34 41.06 57.54 41.87 59.18 40.21 83.17 15.97 98.98
2 72.69 52.90 71.62 53.59 75.79 50.89 129.75 15.92 154.32
5 110.10 53.60 110.65 53.37 119.99 49.43 199.71 15.84 237.30
10 144.60 47.70 142.74 48.38 160.46 41.97 232.94 15.75 276.50
20 178.79 40.00 184.37 38.13 195.04 34.55 251.24 15.68 297.97
The BS 7910 formulation produces higher reference stress values of approximately 15-
20% above API 579-1/ASME FFS-1 for a crack depth between 0 < a/t < 0.5. This result
is contributed by the factor 1.2 on the membrane stress and the surface correction factor
based on a rectangular shape (Eq.B.65). Removing this factor, reference stress results
reach the API 579-1/ASME FFS-1 solutions. However, as shown on Figure 4.34, removing
this factor tends to give smaller reference stress results than that of API 579-1/ASME
FFS-1 for small crack length. Care has to be taken in this case. As mentioned in API
579-1/ASME FFS-1, for crack depth to thickness ratio a/t > 0.5, the surface correction
factor used in BS 7910 gives more conservative results. In the case of a/t = 0.8, there is
a difference of 35-50% on the reference stress between API 579-1/ASME FFS-1 and BS
7910 solutions. Here, the surface correction factor has a large influence on the final result.
Assuming negligible bending stress, (Eq.B.65) and (Eq.B.68) reduce to:
• BS 7910
σref (Pb =0) = 1.2Ms Pm (4.13)
• API 579-1/ASME FFS-1

σref (Pb =0) = Ms Pm (4.14)
As discussed previously, the factor 1.2 is applied to produce a similar level of conser-
vatism that is inherent to the flat plate solutions. This increases directly the reference
stress in BS 7910 by 20% compared to that in API 579-1/ASME FFS-1. This is shown
in Figure 4.34 and Figure 4.35 for 0 < a/t < 0.5. However for a ratio crack depth to
thickness ratio greater than 1/2, surface correction factors used in BS 7910 give different
results leading to the increasing difference in the reference stress.
4.4.5.3 Axial Surface Crack - Analytical solutions against experimental data

from [136]
Pressure tests have been conducted on cylinders made of 4134V and 4130X steel rated
for a 31 MPa service pressure [136]. Sharp, semi-elliptical surface axial cracks (0.076 mm
tip radius) were introduced by Electro-Discharge-Machining and data are summarised

in Table 4.12 and Table 4.13. The cylinders were pressurized monotonically to failure
with water or nitrogen gas and leak/break failure conditions are reported. The transition
between these two failure modes in the experimental data was compared to the predictions
based on the reference stress solutions in BS 7910 and API 579-1/ASME FFS-1. Cases
have been plotted in a plane of hoop stress versus a/2c. A failure boundary has been
developed using reference stress solutions, assuming a constant thickness, t, of 7.5 mm,
inner radius, Ri , of 110.94 mm and crack depth, a, of 5.75 mm for the first case. The
second case was prepared assuming a constant thickness, t, of 7.25 mm, inner radius, Ri ,
of 110.87 mm and crack depth, a, of 5.5 mm. Results are plotted in Figure 4.37 and
Figure 4.38.
Table 4.12: Case 1 - Experimental data of leak/break tests on 4134V steel from [136]
Test Ro t a c Re Rm Pexp Leak /

N◦ (mm) (mm) (mm) (mm) (MPa) (MPa) (MPa) Break
5 118.6 7.6 6.2 25.4 1097.1 1179.9 49.3 Leak
8 118.25 7.1 6.1 38.1 1097.1 1179.9 28.8 Leak
9 118.35 7.8 6.5 12.7 1097.1 1179.9 68.5 Leak
11 118.6 7.1 6.1 34.95 1097.1 1179.9 34.8 Leak
12 118.25 7.4 6.7 34.95 1097.1 1179.9 33 Leak
14 118.35 7.4 6.3 34.95 1097.1 1179.9 35.6 Leak
15 118.25 7.4 7.2 34.95 1097.1 1179.9 32.6 Leak
1 118.5 7.6 5.7 25.4 1097.1 1179.9 55.5 Break
2 118.5 7 4.3 12.7 1097.1 1179.9 64.2 Break
3 118.35 7.4 4.6 25.4 1097.1 1179.9 57.5 Break
4 118.35 7.5 5.2 25.4 1097.1 1179.9 52.3 Break
6 118.25 7.4 5.1 38.1 1097.1 1179.9 47.3 Break
7 118.5 7.7 5.5 38.1 1097.1 1179.9 45.1 Break
10 118.25 7.4 5.4 12.7 1097.1 1179.9 70.3 Break
13 118.1 7.5 5.2 34.95 1097.1 1179.9 48.4 Break
Table 4.13: Case 2 - Experimental data of leak/break tests on 4130X steel from [136]
Test Ro t a c Re Rm Pexp Leak /

N◦ (mm) (mm) (mm) (mm) (MPa) (MPa) (MPa) Break
30747 118.1 7.4 4.9 25.4 648 783 41.5 Break
30835 118.1 6.8 5.2 25.4 648 783 37.1 Break
38444 118.1 7.2 4.8 50.8 648 783 33.2 Break
42934 118.1 6.9 4.6 68.6 648 783 27.9 Break
55172 118.1 7.2 5.5 50.8 648 783 35.7 Break
56792 118.1 7.6 5.1 76.2 648 783 32.7 Break
65284 118.1 7.2 5.5 76.2 648 783 24.6 Break
84065 118.1 7.1 5.4 68.6 648 783 25.2 Break
82915 118.1 7.4 6.3 76.2 648 783 21.4 Leak
83905 118.1 7.2 6.1 68.6 648 783 26.2 Leak
30933 118.1 7.3 6.2 25.4 648 783 38.3 Leak
45491 118.1 7.4 6.4 50.8 648 783 29.5 Leak
Figure 4.37: Experimental data and failure boundaries as per BS 7910 and API 579-
1/ASME FFS-1 - Case 1 - Material = 4134V
Figure 4.38: Experimental data and failure boundaries as per BS 7910 and API 579-
1/ASME FFS-1 - Case 2 - Material = 4130X
The boundary created using recommended reference stress solution in API 579-1/ASME
FFS-1 was found to successfully distinguish leak and break cases of the literature test data
considered. On the other hand BS 7910 procedure was more conservative assuming leak-
age as failure. Removing the 1.2 factor applied on membrane stress (Eq.B.65) lead to a
less conservative assumption but still excluded leaks (‘BS 7910 – Mod’).
4.5 Summary
In this chapter, the solutions from both BS7910 and API 579-1/ASME FFS-1, at both
design or re-assessment stages have been analysed. All solutions from these standards
used in this chapter are summarised in Table 4.14. The influence of the design code and
associated safety margins is described in Section 4.2. Depending on the procedure used,
the minimum required thickness differs slightly. ASME BPVC Section VIII - Division 1
results in the greatest thickness due to the safety margins employed. The other codes
investigated gave similar results to each other. The design may affect the fracture be-
haviour. For a thin-walled structure containing a surface crack, the collapse limit of a
ligament is usually the main cause of structure rupture. On the other hand, for a thicker
structure, brittle fracture will have to be considered. In a two-parameter assessment such
as the Failure Assessment Diagram, this can have a important effect. The final results
will be more or less dependant on the couples stress intensity factor/fracture toughness
or reference stress/yield strength parameters.
Leak-before-break parameters as described in established procedures have been eval-

uated for leakage and crack stability assessment. The COA results (Section 4.3), were
provided for both inner & outer surfaces or mid-thickness (representing the averaged COA
through-thickness) depending on the solution. All solutions investigated provided coher-
ent results until half of the limit of applicability for both axial and circumferential cracks.
It should be noted that BS 7910 solutions were developed for plates and corrected for
cylindrical shapes. On the other hand, API 579-1/ASME FFS-1 have been developed for
cylinders by FEA. For both axial or circumferential cracks, the stress-based solution from
API 579-1/ASME FFS-1 gave greater values of COA than the pressure-based equation.
After this limit, solutions begin to diverge. For axial cracks, BS 7910 and Zahoor solutions
tend to give greater values of COA at mid-thickness (similar to API 579-1/ ASME FFS-1
at the outer surface). For circumferential cracks, BS 7910 and Paris-Tada solutions tend
to give lower values than API 579-1/ ASME FFS-1 with a COA at mid-thickness lower
than API 579-1/ ASME FFS-1 at the inner surface. The effect of plasticity has been
evaluated for axial models. BS 7910 and Zahoor models include correction factors for
small-scale yielding while API 579-1/ ASME FFS-1 propose a correction factor based on
the plastic collapse assessment. This results in a wider range of application for the latter.
Recalling that COA solution will be used to assess the leak rate which is a critical param-
eter in a LBB assessment, an over-estimation of the COA can lead to non-conservative
result and vice-versa an under-estimation lead to a very conservative result. Only elastic
models are presented compared to each other here. They are known to under-estimate
the real COA for ductile materials.
In terms of crack stability (Section 4.4), different solutions provided for FAD (Section
4.4.2), SIF (Section 4.4.3) and RS (Section 4.4.4) have been evaluated. Four relevant
geometries for LBB assessment have been chosen. For a given axial defect, BS 7910 tends
to give greater values of reference stress for both through-thickness and surface defects.
The multiplier of 1.2 introduced in the reference stress solutions, as such achieve a certain
level of conservatism partially contributed to this observation. It should also be noted that
in the original paper, this factor is not present. The other factor leading to an important
difference is the correction factor employed. Analyses of the reference stress solutions
for axial surface defects versus available experimental data show that API 579-1/ASME
FFS-1 is found to successfully distinguish between leak and break cases. On the other
hand, BS 7910 gives more conservative results, assuming leak as a fail or fracture result.
This is also observed even after the removal of factor 1.2. For circumferential defects,
when only membrane stress loading is considered, both standards give similar values in
terms of reference stresses. However, when bending stress loading is applied, BS 7910
solutions tend to give higher reference stresses. It is worth noting that reference stress
solutions recommended in BS 7910 are mainly from local limit load solutions, as they
are more conservative. In leak before break assessments, since that the failure mode is
pre-supposed to be ductile, the criteria for break will be based mainly on the reference
stress solution. Therefore, BS 7910 is likely to be significantly more conservative than
API 579-1/ASME FFS-1 for its break condition.
Table 4.14: Summary of API 579-1/ASME FFS-1 and BS 7910 solutions used in this
chapter
Flaw type Solutions API 579-1/ASME FFS-1 [9] BS 7910 [1]

Eq.B.12 or Eq.B.13 Eq.B.7 + Eq.B.8
COA (Anderson, [137]) (Wuthrich, [104])
Eq.B.34 or Eq.B.36 Eq.B.33
SIF (Green, [138]) (France, [139])
ATWC Eq.B.53 + (Eq.B.54, B.54 or Eq.B.56) Eq.B.51 + Eq.B.52
RS (Miller, [140]) (Willoughby, [135])
Eq.B.26 or Eq.B.27 Eq.B.24 + Eq.B.25
COA (Anderson, [137]) (Wuthrich, [104])
SIF (Green, [138]) (France, [139])
CTWC Eq.B.59 Eq.B.58
RS (Zahoor, [134]) (Kastner, [141])
SIF (Anderson, [142]) (Newman, [143])
Eq.B.68 + (Eq.B.71 or Eq.B.72) Eq.B.65 + Eq.B.66
ASC
RS (Willoughby, [135]) (Willoughby, [135])
SIF (Anderson, [142]) (Newman, [143])
Eq.B.76 Eq.B.74
CSC
RS (Zahoor, [134]) (Kastner, [141])
In the light of the results presented within this study, the reference stress solutions
for pipes/cylinders containing axial flaws should be re-evaluated in order to enhance the
assessment capability. Validation of the solutions against other data from industry or
research papers that include leak and break analysis would be beneficial in establishing a
more accurate assessment of LBB. Further work to assess the effect of these difference on
a full LBB case would be advantageous as well.
Chapter 5
Numerical Methods - Investigation

of Ideal Through-Wall Cracks in
Plates and Effect of Crack Shape in
Cylindrical Components
Contents
5.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 100
5.2 Modelling methods - Benchmark using through-wall cracks
in plates . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 101
5.2.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . 101
5.2.2 Presentation of models . . . . . . . . . . . . . . . . . . . . . . . 102
5.2.3 Parameters investigated . . . . . . . . . . . . . . . . . . . . . . 104
5.2.4 Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 113
5.3 Investigation of crack shape influence (cylinders) . . . . . . . 115
5.3.1 Models transformation - From plate to cylinder . . . . . . . . . 115
5.3.2 An ideal axial through-wall crack in cylinders . . . . . . . . . . 116
5.3.3 Non-ideal axial through-wall crack in cylinders . . . . . . . . . 126
5.3.4 Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 130
5.3.5 Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 135
5.4 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 137
99
CHAPTER 5. FEA I - THROUGH-WALL CRACKS IN PLATES/CYLINDERS 100
5.1 Introduction
As discussed in Chapter 3, a conservative approach of the fracture side of an LBB

assessment is to assume a larger flaw size than might be present. In other words, fracture
could occur earlier than in reality. However, this exaggeration of crack length would
over-estimate the leak rate from an actual crack size, whereas a smaller flaw size than
anticipated will give a conservative prediction of leakage. It is therefore very important to
determine the behaviour of crack size and shape accurately to ensure the most appropriate
fitness-for-service assessment using LBB principles.
In a simplified analysis [60], an idealized through-wall crack with straight crack sides
perpendicular to the surface is postulated. Such an assumption simplifies the analysis
significantly. However, in reality, a surface crack that grows through the wall thickness
and breaks through at its deepest point. At break-through, a given flaw may have a much
longer length on the inner surface, and only a small crack length on the outer surface.
The crack sides are not perpendicular to the component surface but are more likely to be
steeply angled or curved. This results in a through-wall crack with different crack lengths
on the internal and external surfaces. This crack eventually grows further to a shape
which does become closer to an idealized through-wall crack. This is described more in
details in Chapter 6. Experimental works to investigate crack growth shape due to fatigue
loading have been carried out [144], [145], [111]. However, stress intensity factors (SIFs)
and crack opening areas (COAs) of non-idealized through-wall cracks, which are essential
in a LBB assessment, have often been underlooked since there are no available formulae
to derive these parameters for non-idealised crack shape. These parameters are essential
to predict the growth of an initial surface defect through a pipe wall to final failure.
In this chapter, FE models of idealized through-wall defects will be investigated and

compared to existing solutions from BS 7910 and API 579-1/ASME FFS-1. These two
procedures provide structural integrity assessment routes with different SIF and COA
solutions, including LBB assessment for cylindrical components. The effect of through-
wall cracks with shapes yet to become ‘idealized’ are also examined for their significance
in an LBB assessment. Based on the findings, the significance of non-idealised crack is
discussed. Table 5.1 gives a summary of all simulation models that are described in this
chapter. It is worth pointing out that the consideration of the crack face pressure has
been included in the work described in this chapter. If not stated, it is safe to assume
that crack-face pressure has not been included.
Table 5.1: A summary of simulation models described in this chapter
Model N◦ Description
PTW02 Plate with a through wall crack of 0.2 x Width long
Cylindrical component with a through-wall crack of length
CTWLX1 -X2
equal to λ = X1 and ratio inner to outer crack length equal to X2
X-CFP Model X including crack face pressure
X-nCFP Model X not including crack face pressure
5.2 Modelling methods - Benchmark using through-

wall cracks in plates
5.2.1 Introduction
Finite Element Method (FEM) is a numerical technique that divides a whole problem
domain into simpler and smaller parts, called finite elements. Approximate solutions
are numerically integrated for partial differential or integral equations. A continuous
physical problem is therefore transformed into a discretised finite element problem with
unknown nodal values of a physical field. Values inside finite elements can be recovered
using nodal values. The main sources of error are often round-off or discretization errors.
Convergence study is often employed to overcome this limitation. Care has to be taken
while pre-processing results as errors or bad hypotheses in the input data can produce
highly incorrect or inaccurate results that may be easily overlooked or accepted as true.
For this reason, in the first part of this work, a simple through-wall cracked plate model
was prepared in order to validate the techniques used. All FEA work presented in this
thesis has been modelled using Abaqus/CAE software. As with most FE software, Abaqus
does not offer a built-in system of units. It allows flexibility to define the desired units.
SI-system of mm was employed in this work and the units are given in Table 5.2.
Table 5.2: SI units used in FEA models
Parameter Length Time Mass Density Force Pressure/Stress Energy

Unit mm s tonne tonne/mm3 N MPa mJ
FEA has proven to be useful for computing custom crack solutions when there are no
available solutions for component geometries. These custom solutions can provide more
accurate evaluation than a standard simple geometry does. In this section, the methodolo-
gies employed for crack modelling are presented. Results obtained from idealised cases are
compared to well-established solutions (i.e. BS 7910 or API 579-1/ASME FFS-1).
The parameters of interest are stress intensity factor (SIF), reference stress (RS) and
crack opening area (COA). As part of the effort to validate the techniques used and to
examine the means of deriving meaningful results and potential difficulties when creating
a cracked model, a plate with a through-wall defect has been selected as benchmark. The
geometry, material, meshing methodology, loading and boundary conditions are presented
in this chapter. Post-processing and validation methods are also given for each parameter
using various models. Elastic (for SIF and COA) and elastic-plastic (for RS) FEA have
been carried out. The difference in values between analytical solutions and FEA are
presented as:
|F EA − Analytical|
%Dif f erence = 100 × (5.1)
Analytical
5.2.2 Presentation of models
In order to validate the techniques used to simulate cracked structures, a plate with
a through-wall crack has been used as benchmark. The plate length, width and thick-
ness were 400 mm, 100 mm and 15 mm respectively. Four different through-wall cracks
were positioned at the middle of the plate, with lengths of 0.2W, 0.4W, 0.6W and 0.8W
respectively (Models PTW02, PTW04, PTW06 and PTW08). A far field pressure of
100 MPa equivalent to a force of 150kN was applied as a ramp to both top and bottom
surfaces. The material was assumed isotropic. For elastic properties, Young’s modulus
and Poisson’s ratio were estimated to be 217147 MPa and 0.3 respectively (See Chapter
6). Von-Mises yield criteria and isotropic hardening were assumed in all elastic-plastic
analyses (Section 5.2.3.2). At the same time, TWI ECA software CrackWISE R
5 has been
used to generate analytical solutions from BS 7910:2013. Elastic SIF, RS and elastic COA
values were extracted for these four cases and compared to handbook solutions.
The step-by-step approach used in the numerical simulation is given in detail in the
following section. In order to minimise calculation time but still maintaining sufficient
accuracy, different strategies were employed. A solid uncracked plate part and a swept
crack box were prepared independently (Figure 5.1). The crack box was then merged
with the uncracked plate at the desired location in the assembly module. This allows
a great flexibility to duplicate models in an efficient and reliable manner as compared
to conventional. The different parameters (i.e. component/crack dimensions, partitions,
material properties) can be changed rapidly and a new model re-constructed in seconds,
without having to rebuilt a model from early beginning each time. The use of crack
box also permits to keep similar features throughout the different models (i.e mesh or
partition) and reduce the possibility of making mistake. Python scripting were used to
develop plug-ins for this purpose. The use of symmetry planes allows users to model
only one fourth of the whole plate (Figure 5.2). Another symmetry plane could have
been used (Z-symm), however, for other models presented later in the thesis, only quarter
model reduction is possible. Another methodology is to impose a mesh transition from a
fine mesh along the crack and around the crack-tip to a coarser mesh far from the crack.
Partition toolset was used to divide the plate into different regions. Two swept regions
were created to generate mesh transition through the thickness (y-z plane) and on the
surface (x-y plane). This reduces significantly the number of elements and maintains a
fine mesh in the area of interest.
Figure 5.1: Modelling technique: (i) Uncracked plate and (ii) Crack Box
Figure 5.2: From full model to quarter model using symmetry plans - Model PTW02
For typical crack modelling, the most efficient mesh design for crack-tip region is spider-
web configuration [146]. The model in Figure 5.3 consists of concentric rings of struc-
tured hexagonal quadratic elements (C3D20R) surrounding the first ring of swept wedge
quadratic element (C3D15) at the crack tip. In the first ring, a singularity was introduced
to get more accurate stress/strain results at crack tip. Mid-side nodes were located at 1/4
element size from the crack tip for elastic models. When elastic-plastic conditions were
considered, this singularity was moved to 1/2 element size as recommended in Abaqus
documentation [146]. In addition, spider-web configuration was introduced to allow a
smooth transition from a fine mesh at the crack tip (required to get accurate displace-
ment fields) to a coarser mesh in the remaining bulk body. This resulted in concentric
contours (integration domains). Eight different contours have been employed to compute
J-integral, SIF and COA. As J-integral estimates may vary for different contours due
to the approximate nature of finite element solutions. For this reason, the accuracy of
these outputs had to be verified. One of the ways to do so was to examine the values of
J-Integral across a few contours. Good accuracy in the results was observed as J-Integral
values appeared approximately constant from one contour to the next. An example of con-
vergence study is given in Section 5.2.3.1. Significant variations in J-Integral values may
indicate errors in the definition of contour integral, and a finer mesh would be required
to produce a more gradual change in the values. If plasticity were taken into account,
the contour integral domain would have to include the whole plastic zone. Similar mesh
pattern was used for other models (PTW04, PTW06 and PTW08). A mesh convergence
study has been performed to verify the adequacy of this mesh.
Figure 5.3: Mesh refinement, elements and contours used to compute SIFs - Model
PTW02
5.2.3 Parameters investigated
5.2.3.1 Stress intensity factor (SIF)
Elastic SIF values were extracted at each node through the thickness for all the 8
contours using J-integral, assuming plane strain conditions, using the following equa-
tion: r
E.J
KI = (5.2)
1 − ν2
Convergence was also checked. For example, tabulated values of SIF for the model PTW02
with a crack length of 0.2W are given in Table 5.3 and the convergence for different
contours is shown in Table 5.4. In both tables, the results are given for a normalized
position through thickness. The nodes spacing is not constant due to the bias ratio
employed in the mesh seeds. This mesh repartition allows more elements approaching the
plate surfaces, where there is a greater variation in output values. This technique permits
the introduction of more elements (i.e. more data points) where needed. As expected,
Contours 1 and 2 do not present a stabilized solution as there is some difference in SIF
values between these contours. Stabilization and path independence can be considered as
acceptable when approaching Contours 3-4 with less than 0.1% variation and was fully
stabilized in the 8th contour.
Figure 5.4 presents the SIFs values from FEA (PTW02, PTW04, PTW06 and PTW08)
versus handbook solutions. These values were normalized at 1. The horizontal axis repre-
sents the normalized thickness and the vertical axis, the normalized stress intensity factors
from FEA to those of BS 7910. It can be seen that results obtained from FEA are well
in agreement with BS 7910 analytical solution. The mean difference is approximatively
0.2% excluding edge effects. Edge effects were expected as J-integral estimates may be
inaccurate at the crack front ends due to the skewness of the external layers of elements
[146] but this does not affect the accuracy at neighboring node sets. The last elements
(two nodes) on both surfaces were then excluded. Numerical values for all the four cases
are given in Table 5.5. The maximum values obtained using FEA are in good agreement
with those obtained using BS 7910 handbook solutions.
Table 5.3: SIF values for 8 integral contours for model PTW02 (2a=0.2W)
Location FEA CW5 Diff.

(Norm) Ctr 1 Ctr 2 Ctr 3 Ctr 4 Ctr 5 Ctr 6 Ctr 7 Ctr 8 BS 7910 (%)
0.00 625.8 642.8 645.2 645.7 645.7 645.5 645.2 644.9 703.9 8.38
0.03 645.0 670.9 671.9 672.4 672.7 672.9 673.0 673.1 703.9 4.37
0.06 662.0 691.8 693.9 694.2 694.3 694.3 694.3 694.3 703.9 1.36
0.11 668.1 697.2 699.5 700.1 700.3 700.4 700.4 700.4 703.9 0.49
0.16 671.5 700.6 702.9 703.6 703.9 704.0 704.1 704.1 703.9 0.03
0.23 672.7 701.7 704.2 704.9 705.2 705.4 705.5 705.5 703.9 0.23
0.32 673.0 701.9 704.5 705.2 705.6 705.7 705.8 705.9 703.9 0.28
0.43 673.0 701.7 704.3 705.2 705.5 705.7 705.8 705.9 703.9 0.28
0.57 673.0 701.7 704.3 705.2 705.5 705.7 705.8 705.9 703.9 0.28
0.68 673.1 701.9 704.5 705.2 705.6 705.8 705.9 705.9 703.9 0.29
0.77 672.7 701.7 704.2 704.9 705.3 705.4 705.5 705.5 703.9 0.23
0.84 671.5 700.6 702.9 703.6 703.9 704.0 704.1 704.1 703.9 0.03
0.89 668.2 697.3 699.5 700.2 700.4 700.4 700.4 700.5 703.9 0.49
0.94 662.0 691.8 693.9 694.2 694.3 694.3 694.3 694.4 703.9 1.35
0.97 645.0 670.9 672.0 672.4 672.7 672.9 673.1 673.2 703.9 4.37
1.00 625.9 642.8 645.2 645.8 645.7 645.5 645.2 644.9 703.9 8.38
Table 5.4: Contour independence validation for model PTW02 (2a=0.2W)
Absolute difference between contours (%)

Location
C1-C2 C2-C3 C3-C4 C4-C5 C5-C6 C6-C7 C7-C8
0.00 2.63 0.38 0.08 0.01 0.03 0.04 0.05
0.03 3.87 0.15 0.07 0.05 0.03 0.02 0.01
0.06 4.31 0.30 0.05 0.01 0.00 0.00 0.00
0.11 4.17 0.32 0.09 0.03 0.01 0.00 0.00
0.16 4.15 0.33 0.10 0.04 0.02 0.01 0.00
0.23 4.13 0.35 0.10 0.04 0.02 0.01 0.01
0.32 4.11 0.36 0.11 0.05 0.02 0.01 0.01
0.43 4.09 0.37 0.11 0.05 0.03 0.02 0.01
0.57 4.09 0.37 0.11 0.05 0.03 0.02 0.01
0.68 4.11 0.36 0.11 0.05 0.02 0.01 0.01
0.77 4.13 0.35 0.10 0.04 0.02 0.01 0.01
0.84 4.15 0.33 0.10 0.04 0.02 0.01 0.00
0.89 4.17 0.32 0.09 0.03 0.01 0.00 0.00
0.94 4.31 0.30 0.05 0.01 0.00 0.00 0.00
0.97 3.86 0.15 0.07 0.05 0.03 0.02 0.01
1.00 2.63 0.38 0.08 0.01 0.03 0.04 0.05
Table 5.5: Comparison of results from FEA and BS 7910 handbook solutions
R
Normalized FEA - Abaqus
√ 6.13 BS 7910 - CrackWISE
√ 5 Difference
crack length (M P a. mm) (M P a. mm) (%)
0.2 705.9 703.9 0.28
0.4 1088.9 1079.3 0.89
0.6 1569.3 1550.9 1.19
0.8 2522.9 2469.8 2.15
Figure 5.4: Comparison of normalized SIF determined from FEA and normalized BS
7910 handbook solutions
5.2.3.2 Reference stress (RS)
Calculation of reference stress solutions requires the elastic-plastic models. In order

to compute elastic-plastic J-integral along the crack front, actual stress-strain data of
the material are required. A common difficulty often encountered is to achieve sufficient
convergence in analysis. Different calculation tasks that were used to compute reference
stress from J-Integral values are presented in this section. Computed reference stress
was used to determine plastic collapse ratio, Lr . When a tolerable crack is found during
routine inspection, or when there is a need to evaluate a critical crack size for a given
service life, ECA is performed by means of FAD. Plastic properties have been curve-fitted
to be implemented into Abaqus. Analysis difficulties can occur when too many data points
are used. It is important that plastic strain must stay in an ascending order, otherwise
the analysis will not run. The best approach is to sample the strain-hardening data until
a representative and acceptable curve is achieved. Figure 5.5 presents of plot of plastic
strain versus true stress data used in Abaqus and the selected values are given in Table
5.6.
Figure 5.5: True stress – plastic strain curve for elastic-plastic finite element analysis
Table 5.6: Plastic properties employed for elastic-plastic models
True stress (MPa) 638.2 659.9 680.5 715.4 742.9 763.9 779.4 801.5 810.4
Plastic strain (mm/mm) 0 0.005 0.01 0.02 0.03 0.04 0.05 0.075 0.1
The elastic region of the total elastic-plastic J-integral plot. This was obtained by
applying curve-fitting to the first few load increments of J-Integral values at a specific
crack front location. The calculations can be performed for every node position at crack
front . Only at the crack front where J-Integral results were maximum at the final loading
step were of interest. A quadratic curve fit was expected since J 2 is proportional to SIF,
K, which is linear in the elastic region. Figure 5.6 shows a close-up of the first few
increments of J-Integral (0 to 50 MPa). Quadratic curve fit has been applied to the first
few data points (open red circles). The elastic J trend was computed using a second order
polynomial fit. It was compared to the next several J increments (blue cross data points)
to verify that these results were in the expected elastic range and that the curve-fit was
valid.
Figure 5.6: J-Integral quadratic curve fit to estimate elastic region - Model PTW02
As an effort to validate the curve fitting exercise, computed elastic J-Integral from an
elastic-plastic model has been compared with a fully elastic model. Figure 5.7 shows both
elastic J-Integral obtained from the first few data points of an elastic-plastic model with
curve fitting (blue line) and from a fully elastic model (red line).
In a typical elastic-plastic analysis without a crack, the initial load increments can
be large due to equilibrium convergence. However, for an elastic-plastic fracture analysis
with a crack, several small load increments are required at the beginning of the analysis to
ensure that it is posible to derive J-Integral values within the elastic range. The maximum
load must be sufficiently high to create yielding at the crack front, which usually requires
a much higher load than the operating or design load. In this work, the pressure applied
at the plate ends for the elastic-plastic case has been increased to 800 MPa (i.e. a pressure
of 100 MPa only was used for the elastic analysis). Figure 5.8 shows the elastic and elastic
plastic J-Integral which are used to estimate the reference stress.
Figure 5.7: Comparison of elastic J-Integral from curve fit and FEA - Model PTW02
Figure 5.8: Illustration of the Jtotal and computed Jelastic curves - Model PTW02
The nominal load value was obtained using the material-specific FAD equation evalu-
ated at Lr =1 (nominal plastic collapse) [1]:
−1/2
L3 σys

Eref
f (Lr ) = + r (5.3)
Lr σys 2Eref
When the material-specific FAD curve equation is evaluated at Lr =1, it results in:
Jtotal 0.002E 1
|Lr =1 = 1 + + (5.4)
Jelastic σys 2 1 + 0.002E
σys
The value of Jtotal /Jelastic evaluated at Lr =1 was calculated using modulus of elasticity,
E, and yield strength, σys . 0.002 is the nominal 0.2% offset strain at yield. This gives a
Jtotal /Jelastic ratio value of 1.978 using material properties described earlier. The nominal
load was obtained at the intersection of this ratio with the Jtotal /Jelastic ratio curve. In
Figure 5.9 the material-specific ratio is shown by the horizontal red line, J-Integral ratio
curve is the blue line, and the vertical green line represents nominal load value at the
intersection between blue and red lines. In this analysis, nominal load of 528.75 MPa was
applied to FEA models. If the maximum applied load was not high enough, elastic-plastic
J-Integral values would not sufficiently large to obtain the intersection with material-
specific J ratio. If the maximum J was not large enough, then elastic-plastic analysis
was repeated with a higher maximum load so that J increases accordingly to evaluate the
intersection point and thus obtain the nominal load. The nominal load was derived from
the expected load at which Lr =1, where nominal plastic collapse occurs. The nominal
load is typically higher than operating load. The nominal load was used to normalize the
plastic collapse ratios.
The reference stress geometry factor, Y , is defined as the ratio of yield strength, σys ,
to nominal load obtained at Lr =1, σnominal |Lr =1
σys
Y = (5.5)
σnominal |Lr =1
The nominal load value, σnominal |Lr =1 , obtained from the intersection point in Figure 5.9,
determines the reference stress that satisfies the material-specific FAD equation at Lr =1.
It was also used to normalize the values at Lr axis. In this analysis the geometry factor Y
is 1.207. Reference stress and Lr values can now be computed for each analysis increment
to obtain case-specific and material-specific FAD values. The reference stress, σref , at
each load increment is given by:
σref = Y σi (5.6)
Figure 5.9: Derivation of the intersection between Jtotal /Jelastic ratio and the material-
specific value - Model PTW02
where σi is the load value at each load increment i. Reference stress can be computed for
different loadings for a given geometry. Table 5.7 and Figure ?? present the results for
the four plate geometries of interest.
Table 5.7: Comparison of reference stress determined from FEA and BS 7910 handbook
solutions
Normalized Nom. Load Geom. factor FEA - Abaqus 6.13 BS 7910 - CW5 Diff.
crack length (MPa) Y (MPa) (MPa) (%)
0.2 528.75 1.21 120.70 125 3.44
0.4 406.42 1.57 157.03 166.67 5.78
0.6 271.08 2.35 235.43 250 5.83
0.8 137.60 4.64 463.81 500 7.24
5.2.3.3 Crack opening area (COA)
The calculation of COA is also not as direct as that of SIF. Displacements at every
nodes along the crack would need to be extracted. Figure 5.10 presents a selection of the
elements where the displacements were extracted. An accurate evaluation of the crack
tip opening displacement (CTOD) around the crack tip is crucial as most of the crack
opening will occur in this area. The mesh refinement in this area is therefore important.
On the other hand, variation in the displacement is limited as it goes further.
As an example, a detailed description of the post-processing of FEA outputs is given

for Model PTW02. In order to derive crack opening area, the CTODs have been extracted
Figure 5.10: Displacement field along the half-crack length from FEA
for each nodes along the half-crack length. As previously mentioned, the COA from FEA
was determined by integrating these CTOD values along the crack length. As the CTODs
values are dependent on the mesh, cubic data interpolation has been used to reduce the
distance between each extracted value. This allows a more accurate area calculation using
trapezoidale rule. An example is provided in Figure 5.11.
Figure 5.11: Crack opening displacements extracted from FEA and cubic spline inter-
polation along the crack half-length - Model PTW02
From the interpolated values of CTODs, a trapezoidale rule (Eq.5.7) has been used
to calculate the opening area between each extracted values along the crack half-length.
The distance between two location is defined by a segment [a,b] and its corresponding
CTODs by the couple [f(a),f(b)].

f (a) + f (b)
Area = (b − a) (5.7)
2
The sum of these small increment areas is only a quarter of the total value of crack opening
area. This required multiplication by four due to the symmetrical assumptions in the
FEA models. Assuming an elliptical shape, the crack opening area is often calculated
with Eq.3.6 (See section 3.5).
Table 5.8: Comparison of analytical solution from BS 7910 and FEA results for COA
calculation
Crack length Analytical COA COA from FEA Difference

2a/W (mm2 ) (mm2 ) (%)
0.2 0.68 0.67 0.84
0.4 2.96 2.94 0.51
0.6 7.77 7.77 0.10
0.8 18.50 18.73 1.25
5.2.4 Discussion
Different techniques have been employed to calculate the parameters of interest in

a LBB assessment. A plate with a through-wall crack has been taken as benchmark to
compare the results obtained using FEA with existing solutions from BS 7910. Four
different crack sizes have been considered (PTW02, PTW04, PTW06 and PTW08). A
summary of the findings is given below:
• Model preparation: Different strategies have been used in this validation section to
create FEA models. The utilization of symmetries and mesh strategy have been
detailed. Symmetries have been used to reduce model size. Partitions and mesh
swept also helped to reduce the number of elements in the models, focussing on the
region of interest. An efficient manner to prepare the models was to create two parts
independently, namely: (i) uncracked plate and (ii) crack box (See Section 5.2.2).
This allows quick modification to the position of the through-wall crack in the plate
without having to repeat the effort to create models. Development of Python script
offers users a great flexibility in model generation and is extremely time-effective
when similar models are required.
• Stress intensity factor: Stress intensity factors have been determined from J-Integral
values. These values have been compared with SIF from Abaqus/CAE output. Con-
vergence study on eight contours has been detailed. Results were similar to those
using analytical solutions of BS 7910 with a maximum difference of 2% for the
longest crack.
• Reference stress: Plastic properties were required for calculation of reference stress.
Loading was applied in excess to achieve the desired nominal load. The elastic region
of J-Integral was inferred by curve-fitting from elastic-plastic J-Integral. Material-
specific FAD curve equation was established at Lr =1 to determine the nominal load
required for the calculation of reference stress. The values were slightly lower than
those determined using handbook solutions of BS 7910 with a maximum difference
of 7% for the longest crack.
• Crack opening area: Crack opening area was calculated from crack opening dis-
placements along the crack front. Cubic spline data interpolation has been used to
sample the extracted displacement into a finer set of values. The area of the crack
opening was then estimated using trapezoidal rule. These values were similar to the
analytical solution in BS 7910 Annex F with a maximum difference of 1% for the
longest crack.
• Post-processing: All three parameters described above requires post-processing of

FEA output. Excel and Matlab spreadsheets have been prepared to assist the
synthesis of Abaqus outputs. SIF have been calculated using J-Integral values and
had to be converted using the relationship between them. RS calculation requires
several steps to determine the nominal load for the material and specific case. COA
were calculated from CTOD and integration was necessary.
This first section focussed on the modelling techniques and their validation compared
to analytical solution. The different parameters of interest have been calculated with
confidence from FEA with the methodologies described above. Post-processing of FEA
output was performed using Excel or Matlab. Based on observations from Chapter 4, SIF
and COA solutions from FEA were found in good agreement with analytical solutions.
However reference stress solutions showed more discrepancy in values. The development
of plug-ins using Python script proved to be highly time-saving and offers great flexibility
in model generation. The next section will focus on cylindrical components. Similar tech-
niques will be detailed to validate solutions for idealized through-wall cracks. Investigation
on the effect of crack shape on SIF and COA will also be discussed.
5.3 Investigation of crack shape influence (cylinders)
5.3.1 Models transformation - From plate to cylinder
Two techniques were employed to generate cylindrical component models: (i) a similar
plug-in for cylinder or (ii) transform plate model into cylinder by mapping the coordinates.
The first technique was presented in detail in Section 5.2.2. Similar plug-in was developed
for cylinders. The second technique is presented in this section. An initial plate model has
been created based on the final desired cylinder geometry using the plug-in developed and
described in Section 5.2.2. Once meshed, the plate model was transformed into cylinder
by mapping the mesh into a cylindrical system as seen in Figure 5.12.
Figure 5.12: Strategy from to transform plate geometry into cylinder geometry
Because the cylinder was generated from a plate model, the dimensions had to be
defined accordingly. The plate length, L, and thickness, t, had to be equal to the cylinder
dimensions. The plate width, W , should be half the circumference of the cylinder:
W = πRo (5.8)
Only the y–z cross section was transformed. The cracked face of the plate was fixed and
the other end was bent into a cylinder. This allows a convenient location as a reference
point for the local plate-coordinate system (x1 , y1 , z1 ) in the global pipe-coordinate system
(x2 , y2 , z2 ). Figure 5.12 gives an overview of the coordinate systems. The mathematical
transformations followed the steps below:
1. The coordinate system of the plate was moved from the corner of the plate to the
center of the cylinder:
y2 = y1 − Ro + t (5.9)
2. The nodes were converted from the plate coordinate system (dark blue) to the
cylindrical one (yellow):
r = Ri + y (5.10)
y1 π
θ= + (5.11)
W 2
3. The nodes coordinates must also be transferred from the cylindrical coordinate
system (yellow) to the global rectangular one (cyan):
y2 = rsin(θ) (5.12)
z2 = rcos(θ) (5.13)
As the cracked face was kept stationary, the coordinates on x-axis remained the
same:
x2 = x1 (5.14)
It is worth pointing out, that special attention must be paid to the final coordinates
of the nodes. Due to rounding errors resulting from the multiplication by pi, some values
have to be correct to construct a proper cylinder. Numerical adjustment had to be
implemented for very small values.
5.3.2 An ideal axial through-wall crack in cylinders
5.3.2.1 Presentation of models
In an LBB assessment, one of the recommended methods is to re-categorize a surface

crack approaching through-thickness (about 80% wall-thickness) as a through-wall crack.
This section will focus on axial through-wall cracks in cylindrical components. The same
material properties and techniques used in the assessment of plates in Section 5.2.2 were
employed. Numerical results were compared with handbook solutions for SIF and COA
given in both BS 7910 and API 579-1/ASME FFS-1. Three locations of concern are
referred as: (A) at internal radius, (B) at mid-thickness and (C) at external radius (see
Figure 5.13). TWI software CrackWISE R

5 was used to generate BS 7910 handbook
solutions, while API 579-1/ASME FFS-1 solutions were performed using Matlab.
Figure 5.13: Illustration of various geometrical parameters of interest
Three-dimensional, elastic FE analyses were carried out using ABAQUS 6.13. Figure
5.14 illustrates a cylinder with idealized and non-idealized axial through-wall defects of
different flaw dimensions. The inner radius,Ri , and wall-thickness, t, were assumed to be
200 mm and 10 mm respectively. The non-idealized defect was characterized by its length
on each side of the cylinder (2ai and 2ao ) while the idealized by its length, 2a. In both
cases, the crack front was assumed as a straight line and the crack length was normalized
as:
a
λ= √ (5.15)
Rm t
Three different crack lengths (λ = 1,2 and 4) have been studied with different ao to ai
ratios. A plug-in has been prepared using Abaqus GUI toolkit [146], to facilitate model
generation in a more time-efficient manner. This plug-in offers the advantage to define the
main parameters which the user wishes to modify without the need to re-construct a new
model. Mesh has also been optimized to reduce the number of elements and to increase
accuracy. Sensitivity study has shown that a better option was to reduce the number of
elements per contour and to use double bias seed (Abaqus function) along the thickness,
such that the stress distribution through the thickness could be captured.
Figure 5.15 shows a typical FE mesh used for the cylindrical models described in this
section. Quarter models were employed by exploiting the advantage of the symmetrical
boundaries. Reduced integration 20-nodes brick elements (C3D20R) were used for the
model. The crack-tip was modelled with a spider-web configuration consisting of swept
wedge quadratic element (C3D15). Homogeneous, isotropic linear elastic material was
assumed with a Young’s modulus of 217 GPa and a Poisson’s ratio of 0.3. Internal
pressure, P, of 2.5 MPa was applied to the inner surface of the cylinder. The elastic J-
integral along the crack front was determined as the mean value of the 4th–5th contours.
Figure 5.14: Schematics of ideal and non-ideal axial through-wall cracks in a cylinder
Path independence was achieved with a difference of up to 1%. The elastic SIF was then
determined assuming plane strain conditions.
Figure 5.15: Typical FE mesh employed in this Chapter
The elastic COA was determined at the inner, mid and outer surfaces by integrating the
crack opening displacements along the crack length. SIF and COA values obtained from
idealized through-wall crack were compared with the existing solutions. TWI software
CrackWISE R
5 and Matlab have been used to generate BS 7910 and API 579-1/ASME
FFS-1 solutions. Equations employed can be found in Appendix B respectively.
5.3.2.2 Discussion on handbook formulations
Based on the comparison work presented in Chapter 4, the following observations have
been made:
• Stress intensity factors solutions given in both procedures are expressed in the same
manner i.e. membrane and bending stresses with correction factors derived from
FEA. They are defined as M and G in BS 7910 and API 579-1/ASME FFS-1 re-
spectively.
• API 579-1/ASME FFS-1 solutions also include a crack face pressure term (pc ). BS
7910 explicitly states that its solutions do not account for this effect, considering
it negligible. The effects of crack face pressure were included and studied in some
cases.
• API 579-1/ASME FFS-1 solution for internal pressure only has included this crack
face pressure by incorporating external radius in its formulation.
• The formula given in BS 7910 employs mean radius whereas API 579-1/ASME
FFS-1 solutions are generally accounting for internal radius
The implication of using either mean radius or internal radius ought to be investigated
prior to progressing into more complicated solutions. As far as internal pressure is con-
cerned, the bending stress can be assumed to be negligible and the membrane stress is
equal to hoop stress using thin shell theory which is given as (PR)/t. However, depending
on the procedure, the use of which radius to use was questioned. While API 579-1/ASME
FFS-1 recommends calculating membrane stress using the internal radius, the mean ra-
dius is implicit in thin shell theory. Simulations have been conducted using the following
three formulae:
• Case 1: Pm = (P Ri )/t = 50 MPa

• Case 2: Pm = (P Rm )/t = 51.25MPa
• Case 3: Pm = (P Ro )/t = 52.5 MPa
The following trends can be observed from evaluations previously discussed:
• When comparing Pm derived from internal, mean or external radius, a higher mem-
brane stress produces a similar gradient in the SIFs through the component thickness
but with an offset at a higher value.
• Variation of SIF through the thickness does not exhibit the same behaviour when
assessed using both procedures. Using BS 7910 solution as reference, API 579-
1/ASME FFS-1 solutions show more significant variation through the thickness,
giving lower values than those of BS 7910 at the internal surface and larger at the
external surface.
• Assuming the stresses are higher at the external surface, it is apparent that BS
7910 solution (Case 2) and API 579-1/ASME FFS-1 solutions (Case 1) give almost
identical values at the outer surface.
• BS 7910 provides a list of different references for different COA models and a solution
at mid-thickness from the literature is given. SIF calculations were derived assuming
mean radius.
• API 579-1/ASME FFS-1 provides its own COA solutions, for both internal and
external surfaces of an idealized crack.
5.3.2.3 Validation of cylindrical models and CFP effect
Based on the geometrical and loading conditions described earlier, two different simu-
lations (models X-CFP and models X-nCFP) have been prepared with and without crack
face pressure (Figure 5.16). In model CTWL2-Id-nCFP, internal pressure was applied
only to the internal surface of the cylinder and the crack face was kept free of loading
.
Figure 5.16: Cylindrical component under internal pressure: (a) with crack face pressure
and (b) without crack face pressure
Abaqus allows the extraction of both SIF KI values directly or J-Integral along the
crack front. Figure 5.17 shows a typical through-thickness SIF profile. Both direct KI
and J-integral extraction are presented: the red line is SIF derived from J and the blue
line is the direct KI extraction. Both results show very identical values through section
thickness. As the values of KI and SIF derived from J are very close, only J-Integral
results based on the fifth contour are presented here.
Using the normalized values of FEA-derived SIFs as reference, Figure 5.18 presents the
solutions given in both procedures for Case 1 and Case 2 (as defined in Section 5.3.2.2).
BS 7910 solution (using mean radius, Rm ) gives the closest value to that of FEA all along
the thickness, while API 579-1/ASME FFS-1 solution(Rm ) solution agrees up to 60%
thickness. It is also evident that API 579-1/ASME FFS-1 (Ri ) solution converges to BS
7910 (Rm ) solution as it approaches the external radius.
Figure 5.17: Typical FEA computed SIF (KI and J) for a through-wall crack in a
cylindrical component
Figure 5.18: A comparison between FEA-derived and handbook SIF solutions (No CFP)
The crack face pressure term is often associated with membrane stress component in
API 579-1/ASME FFS-1 solution. The addition of crack face pressure was expected to
increase the SIF values compared to the solution without crack face pressure. Figure
5.19 presents the FEA results concerning the aspect of solutions through the thickness.
Values derived from both handbook solutions and FEA were normalized and plotted
along the thickness. Similar observations reported in Section 5.3.2.3 can be made. As
expected, if crack face pressure has been included, API 579-1/ASME FFS-1 solutions are
able account for it while BS 7910 solution do not. The values derived using BS 7910
solutions are therefore the lowest among those of FEA and API 579-1/ASME FFS-1 with
additional crack face pressure which produces the highest value.
Figure 5.19: A comparison between FEA-derived and handbook SIF solutions (CFP)
Current capabilities of Abaqus/CAE does not allow users to acquire the values of
COA directly from simulation model as in the case for SIF. In an effort to determine
COA from FEA, displacements have been extracted at every node along the crack length
at three locations (A, B and C) shown in Figure 5.13. Analytical COAs were calculated
at locations (A) and (C) using API 579-1/ASME FFS-1 solutions with and without crack
face pressure and using BS 7910 solution at location (B) without crack face pressure. As
shown in Figure 5.20, the results from both procedures are well in agreement with the
models with or without application of crack face pressure. A mid-thickness COA with
crack face pressure can be easily extrapolated. Although BS 7910 and API 579-1/ASME
FFS-1 offer dissimilar COA solutions for different locations and employ different principles
with respect to crack face pressure, the two procedures are essentially consistent with each
other. Both offer accurate predictions of crack opening area which could be used for a
LBB assessment.
Figure 5.20: A comparison between FEA-derived and handbook COA solutions
5.3.2.4 Results
For the three crack lengths under investigation, three ideal models have been prepared
and validated (CTWL1-Id, CTWL2-Id and CTWL4-Id). Figure 5.21 and Figure 5.22
present the values of SIF and COA obtained from FEA versus those determined using
handbook solutions respectively.
Stress intensity factor (SIF)
Figure 5.21 presents SIF values at inner (a) and outer (b) surfaces. As J-integral tends
to be unreliable at free surfaces [9], SIF at inner and outer surface were estimated using
a fourth-order polynomial fit. As shown in Table 5.9, the present FE results are in good
agreement with handbook solutions, given only a maximum difference of 3% and 5% at
inner and outer surfaces respectively. As the crack length increase, the SIF values increase.
The effect of crack face pressure is not presented in this part. However, it is worth noting
that when crack face pressure is considered, API 579-1/ASME FFS-1 solutions take this
effect into account while BS 7910 explicitly do not.
Table 5.9: A comparison of SIF values between FEA and handbook solutions (S: stress
based equations, P: pressure based equations)
Diff. Diff Diff.

λ FEA BS 7910 (%) API 579-1 - (S) (%) API 579-1 - (P) (%)
1 510.3 510 0.1 503.3 1.4 499.6 2.1
Inner
2 964.4 963.3 0.1 940.8 2.4 922 4.4

4 2310.7 2354.9 1.9 2315.3 0.2 2250.9 2.6
1 612.3 622.5 1.7 644.8 5.3 603.4 1.5
Outer
2 1190.3 1195.8 0.5 1230.5 3.4 1144.3 3.9

4 2458 2522.3 2.6 2594.5 5.6 2376.5 3.3
Figure 5.21: A comparison of SIF values between FEA and handbook solutions for an
ideal axial through-wall crack in a cylindrical component
Crack opening area (COA)
Figure 5.22 presents COA results at three selected locations, namely inner (a), mid (b)
and outer (c) surfaces. As shown in Table 5.10, FEA results agree well with handbook
solutions. Only a maximum difference of 2%, 1% and 2% at inner, mid and outer surfaces
respectively has been observed. As the crack length increases, the COA increases steeply.
The effect of crack face pressure has not been included in all FEA models. API 579-
1/ASME FFS-1 solutions take into account crack face pressure while BS 7910’s explicitly
do not.
Table 5.10: A comparison of COA values between FEA and handbook solutions (S:
stress based equations, P: pressure based equations)
Diff. Diff. Diff.

λ FEA BS 7910 (%) API 579-1 - (S) (%) API 579-1 - (P) (%)
1 1 — — 1 2 1.1 4.2
Inner
2 5.9 — — 5.8 1.5 6.1 4.2

4 51.6 — — 51.7 0.1 54.2 5
1 1.2 1.2 0.9 — — — —
Mid
2 6.9 6.9 0.3 — — — —

4 59.8 59.6 0.4 — — — —
1 1.3 — — 1.3 2.3 1.4 6.1
Outer
2 7.9 — — 8 0.6 8.3 5.2

4 68.1 — — 68.9 1.2 72 5.6
Figure 5.22: A comparison of COA values between FEA and handbook solutions for an
ideal axial through-wall crack in a cylindrical component
5.3.2.5 Discussion
As previously mentioned (See Section5.3.2.2), BS 7910 provides one solution for both
SIF and COA. API 579-1/ASME FFS-1 provides two solutions for both SIF and COA
depending on the input data (stress or pressure). For COA, BS 7910 provides a solution at
mid-thickness and API 579-1/ASME FFS-1 at inner and outer surfaces. The validation
of techniques for SIF extraction through the section thickness with and without crack
face pressure show that for an idealized flaw, FEA models are in good agreement with
the solutions from procedures. The validation of techniques for COA calculation with
and without crack face pressure show that for idealized cases, FEA models are also in
good agreement with the solutions from procedures. Other aspects such as non-idealized
crack shape were investigated using previously described techniques to study their effect
on LBB assessment in Section 5.3.3 of this chapter.
5.3.3 Non-ideal axial through-wall crack in cylinders
5.3.3.1 Preliminary study
Introduction
An assumption of an idealized crack shape with perfectly straight crack sides ori-
ented perpendicular to the wall is a common practice in a LBB assessment. This as-
sumption helps to reduce the amount of analytical effort needed and is intended to be
conservative. A typically postulated shape for a through-wall crack is a crack with its
front perpendicular to the component thickness. The largest dimension of the real crack
is often assumed (Figure 5.23).
Figure 5.23: Through-wall crack idealisation with API 579-1/ASME FFS-1 and BS 7910
However, assumptions that are valid and conservative for the fracture mechanics aspect
of a LBB assessment can have an opposite effect on the calculation of leakage. In reality, a
crack grows through the thickness and partially penetrates at the deepest point, resulting
in a through-wall crack with different crack lengths at the inner and outer surface. Both
scenarios of a crack developing from the inner or outer surface were considered in this
preliminary study. This crack idealization overestimates the SIF and underestimates the
COA. However, it is obvious that an idealized crack with the largest dimension of the real
crack size may well overestimate the SIF but will also overestimate the COA, therefore
the resulting leak rate prediction will be non-conservative.
SIF and COA solutions have been derived for five individual analysis cases:
• Three idealised through-wall cracks with crack half-lengths of 0.95a (Case 1), a
(Case 2) and 1.05a (Case 3) - See Figure 5.24
• One 45◦ slanted cracks from inside to outside with crack half-lengths through the
component thickness of 0.95a, a and 1.05a (Case 4) - See Figure 5.25
• One 45◦ slanted cracks from outside to inside with crack half-lengths through the
component thickness of 0.95a, a and 1.05a (Case 5) - See Figure 5.26
Figure 5.24: Three ideal through-wall cracks with crack half-length of 0.95a (Case 1),
a (Case 2) and 1.05a (Case 3)
Figure 5.25: Slanted through-wall crack with internal crack half-length=1.05a and ex-
ternal crack half-length=0.95a (Case 4)
Figure 5.26: Slanted through-wall crack with internal crack half-length=0.95a and ex-
ternal crack half-length=1.05a (Case 5)
Results
SIFs along the crack front from the inner to the outer surface were determined. The
three ideal cases, i.e. Cases 1, 2 and 3, have been normalized with the SIF values from
Case 1 (crack half-length = a) and Cases 4 and 5 have been normalized with SIFs values
from the three ideal cases in order to establish the significance of crack idealization (Figure
5.27). The calculted SIFs were higher at locations where the crack length was shorter,
which was consistent with experimental observations. Normalizing by the largest ideal
through-wall crack (which is equal to the largest crack length of a non-ideal crack) resulted
in a conservative assumption along the crack front.
Figure 5.27: SIFs of both idealized and developing through-wall cracks
Idealization of a through-wall crack leads to a conservative estimate of SIF. The highest

SIF along the crack front for a slanted crack will be covered by the SIF from the idealized
shape (See Figure 5.28).
Figure 5.28: Schematics of both idealized and developing through-wall cracks and equiv-
alent SIF position
However, as far as COA is concerned, this idealisation of crack as through-thickness

does not produce conservative results (Figure 5.29). While an idealized crack with a crack
half-length of 1.05a, COA of the slanted crack will be definitively lower, resulting in a
lower leak rate.
Figure 5.29: COAs of both idealized and growing through-wall cracks
Concluding remarks from this preliminary study
Based on this preliminary study, the following observations can be drawn:
• The values of SIF and COA of ideal through-wall crack agree well with handbook
solutions. For slanted through-wall cracks, idealization of a growing crack to an ideal
crack with the largest dimensions gives conservative SIF values up to 80% through
the thickness. Further work is needed to expand crack transformation behaviour on
the first and last 20% (simulation vs. experiment)
• For slanted through-wall cracks, idealization of a growing crack to an ideal crack
with the largest dimensions leads to possibly un-conservative results in term of
COA values. Assumption of a largest crack size overestimates the current COA,
and therefore the current leak rate.
• Taking the minimum crack size to idealize a through-wall crack is likely to produce
more realistic results. However as shown in Figure 5.29, a slanted crack also leads
to a variation of COA through section thickness compared to an ideal case. This is
further studied in the next section.
5.3.3.2 Numerical models
In order further evaluate the effect of these ideal/non-ideal crack shapes on LBB as-
sessments, various FE models were prepared with different crack sizes and ratios of inner
crack size (kept constant) to outer crack size. These cases are presented in Figure 5.30.
Both SIFs and COAs have been derived for each model. Note that as the crack length
increases, the slant angle from the inner and outer surface increases at a constant ra-
tio of aout /aint . However, the application of this technique was however limited due to
Abaqus/CAE model generation. This is the reason why as λ increases, this ratio is limited
as shown on Figure 5.31.
Figure 5.30: Principle of the different case studied
Figure 5.31: Summary of the different cases presented
5.3.4 Results
Both SIF values for the three reference crack lengths λ = 1 (Figure 5.32), λ = 2 (Figure
5.33) and λ = 4 (Figure 5.34)and COA solutions for λ = 1 (Figure 5.35), λ = 2 (Figure
5.36) and λ = 4 (Figure 5.37) from FEA are presented.
Stress intensity factors
In this section, stress intensity factors for non-idealized axial through-wall cracks
are investigated. To improve confidence in the present results, the SIF values derived
from FEA models for an idealized geometry have been compared to handbook solutions
and are presented in Section 5.3.2.3. Both BS 7910 and API 579-1/ASME FFS-1 provide
a general SIF solution and membrane stress due to internal pressure is given as:
PR
Pm = (5.16)
t
The following paragraph, discusses the results from non-idealized crack geometry. SIFs
derived for λ=1, 2 and 4 are presented in Figure 5.32 to Figure 5.34 respectively. The SIFs
values of non-idealized geometries (dotted lines) were normalized to the corresponding
idealized geometry (blue plain line). An idealized through-wall crack length is equal to
the longer crack length (inner crack length, 2ai ) for each model (CTWL1-Id, CTWL2-Id
and CTWL4-Id). The SIF results from idealized case were normalized to 1 and the results
of non-idealized models were directly normalized to the value of idealized models. The
horizontal axis represents normalized through-thickness distance. For all the cases, the
same trend has been observed i.e. SIF is greater at the outer surface whereas lowest value
at the inner surface. Moreover this effect became more pronounced with an increasing
outer to inner crack length ratio, ao to ai . As the difference in length between the inner
and outer surfaces increases, SIF values at the inner surface decrease and increase at the
outer surface. This can be explained by the fact that as the ratio between inner and
outer crack lengths increase, the inner surface undergoes more local constraint than at
the outer surface. This contributes to a greater stress field at the outer surface. Note
that at the outer surface, SIF values are greater than those of an idealized model. Further
observations have been given in Section 5.3.5.
Figure 5.32: Evolution of SIF through section thickness for λ = 1


Crack opening areas
In this section, the crack opening areas for non-idealized axial through-wall cracks
are investigated. To assure confidence in the present results, the COA values derived from
FEA for an idealized geometry have been compared to handbook solutions. Note that BS
7910 gives a solution for mid-thickness while API 579-1/ASME FFS-1 provides solutions
for inner and outer surface. CTODs were directly determined from FE displacements
along the crack length at the inner surface, outer surface and mid-thickness. These values
were then integrated and multiplied by four (due to the symmetric conditions) to get a
final COA value for each model under consideration. The results of the reference lengths
λ = 1, 2 and 4 are presented in Figure 5.35 to Figure 5.37. The horizontal axis represents
the difference between the inner and outer crack lengths. The COAs results obtained
for each model at inner (blue) mid-thickness (green) and outer surfaces (red) have been
normalised to the values derived from an ideal geometry. The COA values for an idealized
cases were normalized to 1 for each location. For an idealized through-wall crack, COAs
are larger at the outer surface than at the inner surface (see Figure 5.22 and Table 5.10).
However, COAs at outer surface and mid-thickness decrease with shorter outer crack
lengths while SIFs become greater (see Section 5.3.4). COAs also decrease at the inner
surface while the inner crack length is kept constant for each respective case. Due to the
higher local constraint at the inner surface, the displacement at the inner surface tends
to be larger than that at the outer surface. This is particularly apparent for relatively
short crack (Figure 5.35). Further observations are given in Section 5.3.5.
Figure 5.35: Evolution of COA at inner, mid and outer surfaces for λ = 1
5.3.5 Discussion
The SIF and COA solutions for idealized and non-idealized axial through-wall crack in
a cylindrical component under internal pressure were investigated.
Stress intensity factors
Figure 5.38 summarizes the SIFs values derived from axial through-wall cracks with
shapes yet to become ‘idealized’ (i.e. difference between inner/outer crack length ≤ 15%)
for three different reference lengths (λ). It can be seen that as the reference crack length
increases, the effect of non-idealization becomes more significant. In all models, as the
ratio of aout /aint decreases, the absolute values of SIF decrease but the gradient along the
thickness begins to develop. For longer cracks (i.e. λ=4), SIF at the inner surface is much
lower than at the outer surface. Under these circumstances, the crack is expected to grow
much faster on the outer surface. With a difference of only 15% between inner and outer
surface crack lengths, SIF is close to zero, resulting in a crack growth only at the outer
surface. At the outer surface, the SIF values can reach up to two times the values of an
idealized geometry. For shorter crack (i.e. λ=1), the variation of SIF along the thickness
is still present but at a much lower scale. It is safe to say that for a given non-idealized
crack, crack growth is dominated by the outer surface. Based on such observation, it
can be argued that a non-idealized crack will grow into an idealized through-wall crack
at some point during components service life, which is consistent with the experimental
observations (Chapter 6). The crack presented can be considered as relatively short. It
can be observed that the crack growth takes place at both inner and outer surface.
Figure 5.38: The effect of crack shapes on SIF profiles through-thickness for different
reference crack lengths, λ
Crack opening areas
In a typical LBB assessment, an estimation of crack length corresponding to a de-

tectable leak-rate is required. This is achieved by determining COA values that relates
leak rate to corresponding crack length. Figure 5.39 summarizes COAs values of through-
wall cracks with shapes yet to become ‘idealized’ (with a difference of less than 20%
between inner and outer crack length) for the three reference lengths (λ). It can be seen
that as the reference crack length increases, the effect of non-idealization becomes more
apparent but the difference in COA between inner and outer surface is less pronounced.
Assuming a linear trend for each curve, the slope of the inner and outer surface is more
pronounced for smaller cracks (λ=1) than that of longer cracks (λ=4). Shape idealization
for cracks ‘yet’ to become fully idealized can rapidly lead to an overestimation of COA.
This results in an overestimation of leak rate which is one of the most important param-
eters in a LBB assessment. There is a risk of producing un-conservative prediction. A
better evaluation of the COA would have helped to reduce margins applied on leak rate
determination.
Figure 5.39: The effect of crack shapes on COA profiles for different reference crack
lengths, λ
5.4 Summary
This chapter validated the solutions for through-wall cracks in plates and cylindrical
components. The calculation of various solutions has been detailed for plates, with step-
by-step methodologies and post-processing to generate models and results.
In order to generate numerous models in a time-efficient manner, different plug-ins have

been created using Python script. Cracks have been introduced into un-cracked bodies
(plate or cylinder) in the assembly step, using a crack-box. This permitted to generate
the models changing only the desired parameters. FE results in this work were found in
good agreement with solutions available in procedures for idealized shape models.
The effect of crack idealization on SIF and COA solutions has been investigated for
axial through-wall cracks in a cylindrical component under internal pressure using detailed
3D elastic FE analyses. Idealization of a through wall crack with its largest dimension
tends to be conservative when it comes to deriving SIF. However, in a LBB assessment,
both crack stability and leakage have to be evaluated. Crack idealization rapidly increases
COA estimation leading to an over-estimation of detectable leak rate. This implies that
for a given crack size, its idealisation will be conservative in terms of SIF but the estimated
resulting leak rate over-estimated. The leak detection can therefore be distorted.
Models and analyses employed in this work were validated against existing solutions
for SIF and COA given in BS 7910 and API 579-1/ASME FFS-1. The work presented
in the second part of this chapter (Section 5.3) permits to extend the understanding on
non-idealised cracks. Different reference crack length (inner surface) have been selected
in order to cover various crack sizes. The observations on SIF and COA for various crack
geometries can be used to carry out sensitivity studies while assessing LBB cases. The
effect of non-idealized geometry on both SIF and COA could be included to assess cracks
‘yet’ to become idealized using the observations presented in this work. The analysis
showed that a non-idealized crack at break-through (from surface to through-wall) will
grow and finally evolve into an ideal shape. LBB assessment in this transition region
is generally limited due to a lack of existing solutions. The effect of crack shape on SIF
solutions can be used in fatigue assessment to examine cracks growing under cyclic loading.
The modification on COA solutions can improve the leak rate calculation more accurately.
This is an important aspect of the assessment often overlooked, while the variation is
important in the transition region. Further work to investigate more realistic crack shapes
(i.e. non-straight crack front) and development of correction factors for existing solutions
would be beneficial in establishing a more refined assessment of LBB.
Chapter 6
Experimental Work - Stress-Strain

Field on the Back Surface of Surface
Cracked Plate During Crack
Transition
Contents
6.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 139
6.2 Semi-qualitative study of crack transition - Perspex specimens139
6.2.1 Preparation of Perspex samples . . . . . . . . . . . . . . . . . . 139
6.2.2 Test setup . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 141
6.2.3 Observations from tested plates . . . . . . . . . . . . . . . . . . 142
6.2.4 Summary of the test results . . . . . . . . . . . . . . . . . . . . 144
6.2.5 Crack shape re-categorization . . . . . . . . . . . . . . . . . . . 146
6.3 Evaluation of stress-strain field on back surface using high
strength steel specimens . . . . . . . . . . . . . . . . . . . . . . 147
6.3.1 Material properties . . . . . . . . . . . . . . . . . . . . . . . . . 148
6.3.2 Test specimens . . . . . . . . . . . . . . . . . . . . . . . . . . . 149
6.3.3 Test arrangement . . . . . . . . . . . . . . . . . . . . . . . . . . 150
6.3.4 Static tests . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 151
6.3.5 Fatigue tests . . . . . . . . . . . . . . . . . . . . . . . . . . . . 154
6.3.6 Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 157
6.3.7 Observations from tested plates . . . . . . . . . . . . . . . . . . 160
6.4 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 165
138
CHAPTER 6. STRESS-STRAIN FIELD ON THE BACK SURFACE 139
6.1 Introduction
This chapter describes the work carried out to examine the transition from surface-
breaking to through-thickness crack. Besides FE simulations, experimental tests were
performed using Perspex and high strength steel plates. The test on Perspex speci-
mens was used as a preliminary study to semi-qualitatively examine crack transition
behaviour. Analysis of the fracture surface provided an insight of the crack transition
between surface-breaking to through-thickness cracks. For the metallic samples, strain
gauges were attached to the back surface of the plates to capture the change in strain
at several locations adjacent to expected breakthrough position during the crack growth
and transition from 80% to 100% through-thickness. Crack lengths were recorded at the
same time to determine crack growth rate of both surfaces during transition.
6.2 Semi-qualitative study of crack transition - Per-

spex specimens
6.2.1 Preparation of Perspex samples
A set of 20 Perspex samples were prepared with semi-elliptical cracks. Figure 6.1
shows the schematic of sample geometry. Actual dimensions are given in Table 6.1. Four
different geometries have been selected, including two different thickness. All samples
were free of internal defects (pores, cracks,etc) and were assumed homogeneous under
visual inspection. The cracks have been artificially introduced with a 36mm diameter,
1mm thick acrylic cutting disc at 3500rpm. It was also necessary to introduce a pre-
crack, to simulate a natural flaw with a sharp crack tip. Two widely adopted methods for
polymers are to either tap or press a razor blade tool into the machined notches. As the
experiment aimed to look at surface cracks, assumed to have an elliptical shape, the latter
procedure has been selected for simplicity. Both procedures are known to produce similar
results when applied to PMMA [147]. Deep cracks and razor edges have been introduced
respectively to reach about 70% - 80% of the sample thickness.
Figure 6.1: Perspex sample
Table 6.1: Dimensions of Perspex samples
L c1 c2 2c a w t a/t RA RA
Test (mm) (mm) (mm) (mm) (mm) (mm) (mm) (—) (mm2 ) (%)
Specimen 100x50x10
S11 100.1 10.45 10.25 29.8 7 1.75 9.54 0.73 317.94 65.99
S12 100.1 10.5 10.2 30.0 6.9 1.5 9.85 0.7 336.82 67.45
S13 100.3 11.2 9.3 30.3 6.6 1.5 9.73 0.68 337.22 68.22
S14 100.3 10.2 9.4 31.2 7.1 1.5 9.64 0.74 315.73 64.47
S15 100.5 9.7 10.6 30.1 7.2 1.5 9.5 0.76 308.59 64.45
Specimen 100x50x15
S21 100.2 7.1 8.2 34.7 10.1 2.0 14.38 0.7 443.74 61.72
S22 100.3 8.5 7.7 33.8 9.7 2.0 14.35 0.68 460 v 64.11
S23 100.1 5.8 8.1 36.4 10.1 1.5 14.31 0.71 431.05 59.89
S24 100.4 7 6.7 36.4 11.1 2.0 14.08 0.79 388.08 55.01
S25 — — — — — — — — — —
Specimen 100x75x10
S31 100.1 21.9 23.8 30.1 7.1 1.5 9.72 0.73 568.93 77.22
S32 100.4 22.8 22.2 31.0 7.1 1.5 9.71 0.73 565.09 76.58
S33 100.3 21.9 21.6 32.0 7 1.5 9.48 0.74 539.81 75.42
S34 100.3 21.7 24.1 30.3 6.7 1.5 9.68 0.69 577.2 78.36
S35 100.4 21.7 22.4 31.7 7.2 1.5 9.65 0.75 552.21 75.49
Specimen 100x75x15
S41 100.3 19.6 20.5 35.2 10.1 1.5 14.36 0.7 802.08 74.18
S42 100.5 20.3 18.3 36.7 10.3 1.5 14.34 0.72 782.91 72.51
S43 100.2 20 20.0 35.2 10.8 1.5 14.42 0.75 785.81 72.47
S44 100.2 20.3 19.9 34.9 10.3 2.0 14.44 0.71 802.12 73.97
S45 100.2 20 19.8 35.5 10.3 1.5 14.15 0.73 778.31 73.05
6.2.2 Test setup
PMMA samples (cast Perspex) have been used as preliminary study specimens as they
provided the possibility to acquire a visual representation of crack propagation through
thickness. It is one of the few brittle amorphous thermoplastic materials with a glass
transition temperature Tg around 100◦ C. However, since PMMA is a viscoelastic mate-
rial, this means fatigue behaviour is different from conventional metallic materials. Two
notable differences are the sensitivity to temperature and cyclic frequency [148]. In order
to duplicate the fatigue behaviour of metallic material, the test frequency was maintained
at a very low level. This prevented localized plastic deformation and at the same time
allowed the generated heat (due to loading) to dissipate quickly enough to avoid crack-tip
blunting. It had been observed that the effect of temperature rise on fatigue behaviour
was insignificant at frequencies less than 0.5Hz. Moreover, a consistent conclusion from
experiments is that fatigue crack propagation decreases with frequency. To prevent sam-
ples from buckling under compressive loading, tension to tension loading was applied for
all tests. The cyclic plastic zone is known to be only about one quarter of the size of the
monotonic plastic zone [149].
Table 6.2: Assumed material properties for preliminary parametric calculations
Density Youngs’s modulus Tensile strength Poisson’s ratio Fracture Toughness

[150] [150] [150] [148] [147]
√
1.19 g/cm3 3210 MPa 75 MPa 0.41 1.1 MPa. m
As described previously, surface cracks and pre-cracks have been introduced in Perspex
samples (Stage 1 on Figure 6.2). Fatigue testing was carried out until the surface crack
grew through thickness, i.e. breakthrough (Stage 2 on Figure 6.2). All tests were con-
ducted in the air with an Instron machine Model 5967 and accompanying Instron Bluehill
software. Tensile cyclic loadings were applied in axial direction and applied stresses ranged
from 3 to 4 MPa. Test was continued (Stage 3 on Figure 6.2) until a through-thickness
crack had developed with its edges parallel to each other or until specimen ruptured (Stage
4 on Figure 6.2).
Figure 6.2: Expected fatigue crack growth in perspex samples
Tests parameters used were:

• Stress ratio, R = σmin /σmax =0.

• Frequency kept under 0.5 Hz
• Amplitude from 0 to σmax
• Triangular waveform
• Test carried out in Load Control
6.2.3 Observations from tested plates
Fracture surface for all tested samples can be found in Annex A. A representative photo
of the different behaviour observed is provided in the discussion.
6.2.3.1 Series 1 (10x50x100mm)
Series 1 was the smallest specimen size series to be tested with a thickness and width of
10 mm and 50 mm respectively. All samples suffered from non-symmetric crack growth.
This was assumed to be caused by a misaligned gripping of the samples in the testing
machine. Sample S12 is given as reference for this series in Figure 6.3. Similar behaviour
was observed for The other samples.The blue area (Area 1) represents the initial crack
introduced in the sample. Fatigue crack growth is represented by the red area (Area 2).
As a result of the misaligned gripping, on all specimens, crack growth mainly took place in
one direction. Crack growth on one side continued until it reached the sample edge. Test
was stopped when sudden rupture occurred (Area 3) and the fracture surface was found
to be flat and clear. Specimen S15 was used for calibration purpose. As a result, only
a small amount of fatigue crack growth has been observed. Few cycles have been tried
under an increasing loading of 125N, 250N, 500N, 1000N, 2000N and 3000N. Specimen
suffered from sudden rupture has at loads just under 3000N. A flat and clear fracture
surface could be observed. A summary of the specimens is given below:
• Sample S11 - Crack growth only on one side until it reached edge
• Sample S12 - Crack growth mainly on one side until it reached edge, no growth in
thickness direction
• Sample S13 - Crack growth mainly on one side until it reached edge
• Sample S15 - Crack growth only one side until instability
6.2.3.2 Series 2 (15x50x100mm)
The specimens thickness of series 2 was greater, i.e. 15mm, while the other dimensions
remained the same than series 1. With this increased thickness the samples were able to
sustain higher load with an increased cross section area. Samples S21 suffered from the
misaligned gripping leading to a crack growth only on one side (Figure 6.3). Different
observations can be made on Sample S23 (Figure A.8). Firstly, the crack was not exactly
located in the middle of the sample (d1 < d2 ). Moreover, the initial crack shape was not
exactly semi-elliptical as shown with the green dots on the initial crack shape. This caused
crack growth to reach the edge prematurely, leading to a sudden rupture. However, com-
pared to previous tests, as loading was applied perpendicular to the crack, the resultant
crack growth was fairly uniform. Both specimens S22 and S24 (Figure 6.4) achieved the
expected fatigue growth pattern (Figure 6.2). The initial crack (Area 1) grew by fatigue
until it reached the back surface (Area 2). Then rapid crack growth occurred and the
crack became an idealised through-wall crack shape (Area 3). Final rupture occurred at
the boundary between Areas 3 and 4, giving a clear and flat surface. A summary of the
specimens is given below:
• Sample S21 - Crack growth only on one side until it reached edge
• Sample S22 - Expected crack growth mainly in thickness direction
• Sample S25 - Not tested
6.2.3.3 Series 3 (10x75x100mm)
The specimens width from series 3 was greater (75mm) while thickness was similar
to series 1. Different level of loading were investigated, leading to a large number of
specimens failure. Sample S35 was used to calibrate the test for this series. As a result
no fatigue can be observed. The failure observed in the specimens was assumed to have
initiated from the area shown on the figure (Figure A.13). Similar to series 1, Sample
S34 also suffered from misaligned gripping (Figure 6.3). This results in a non-symmetric
crack growth reaching the sample’s edge and fails. The next two samples suffered from
bending due to the geometry. As a result, fatigue crack growth was first observed in the
thickness direction up to a certain stage during the test. Then fatigue crack growth was
observed to occur only on surface direction until instability. A summary of the specimens
is given below:
• Sample S31 - Crack growth quicker on one side until it reached edge
• Sample S32 - Not tested
• Sample S33 - Crack growth quicker on one side until it reached instability
• Sample S34 - Crack growth only on one side until instability
• Sample S35 - Fracture without fatigue crack growth
6.2.3.4 Series 4 (15x75x100mm)
The specimens from series 4 were the largest tested with a thickness of 15 mm and width
of 75 mm. Like specimens from series 1, Sample S45 also showed effect due to misaligned
gripping (Figure 6.3). Non-symmetric crack growth was evident from initial crack until
the sample’s edge. Sample S42 was the first to reach the expected crack evolution (Figure
6.4). Fatigue crack growth was observed until it reached the back surface (Area 2).
Breakthrough was observed as well (Area 3). However, the sample failed rapidly after
breakthrough (Area 4). Loading has been reduced for the next samples tested. Other
samples from this series achieved the expected crack shape evolution. A first stage of
fatigue crack growth can be observed (Area 2). When the crack reached the back surface,
it breakthrough and a rapid crack extension may be observed (Area 3). From this point,
another stage of fatigue crack growth could be identified (Area 4) before the final failure
(Area 5). Another observation on samples S41 and S44 was that after breakthrough (Area
3), another area could be observed (Area 4). The fracture surface was similar to sample
S42 where overloading was assumed. Compared to FEA models prepared for non-ideal
through-wall crack, this was realistic as the SIF was found higher on the back surface.
Once the crack shape reached an “ideal” through-wall crack shape, fatigue surface (Area
5) was observed until failure (Area 6). A summary of the specimens is given below:

• Sample S42 - Expected crack growth mainly in thickness direction, rupture soon
after break-through
• Sample S43 - Expected crack growth mainly in thickness direction, through-wall
crack growth mainly in one direction
• Sample S44 - Expected crack growth
• Sample S45 - Crack growth quicker on one side until it reached instability
6.2.4 Summary of the test results
Different observations have been noted based on the experimental work on Perspex
samples:
• If the sample was not loaded strictly perpendicular to the applied load, fatigue crack
growth resulted in a non-symmetric pattern. This lead to rapid extension of the
crack size on one side. It reached rapidly the specimen edge and failed prematurely
(S11, S12, S13, S14, S15, S21, S34, S45).
• Thicker samples allowed the crack growth through-thickness. Desired crack shape
evolution has been observed for Samples from Series 2 and 4 which were 15mm
thick.
• Samples from Series 3 experienced from bending during the tests. This resulted
in an increase in crack length on surface and crack growth arrest in the thickness
direction.
• Samples from Series 4 achieved expected crack shape evolution. Once the crack had
broken through, a rapid crack growth was observed until the crack shape reached an
“idealised” through-wall shape. This has been confirmed by the FEA work carried
on previously, (see Chapter 5), of which a high SIF values was predicted at the back
surface of a non-idealised through-wall crack.
Figure 6.3: Fracture surface of samples with a non-symmetric fatigue crack growth
(Samples S12, S21, S31 & S45)
Figure 6.4: Fracture surface of samples with an expected fatigue crack growth pattern
(Samples S22, S24, S41 & S44)
Table 6.3: Summary of test results
Specimen Non-Symmetric loading Fatigue+Breakthrough Fatigue+Break Break

S11 X
S12 X
S13 X
S14 X
S15 X
S21 X
S22 X
S23 X
S24 X
S31 X (Bend)
S33 X (Bend)
S34 X
S35 X
S41 X
S42 X
S43 X
S44 X
S45 X
6.2.5 Crack shape re-categorization
Once the surface crack depth reached 80% of the wall thickness by fatigue crack growth,
current available procedures recommend re-characterization as through-thickness to eval-
uate its acceptance when considering LBB condition. The re-categorization as a through-
wall crack consists of the combination of the surface crack length plus an additional length
based on the thickness or remaining ligament. The formulation for re-characterization in
three procedures are given:
• General method in API 579-1/ASME-1 FFS [9]:
2ct = 2cs + 2(t − as ) (6.1)
where 2ct is the re-categorized through-wall crack length equal to the sum of the
surface crack length 2 cs plus twice the remaining ligament with as the surface crack
depth, and t is the wall thickness.
• The recommended formula for LBB analysis in API 579-1/ASME FFS-1 [9]:
2ct = 2cs + 2t (6.2)
The re-categorized crack length is longer because the thickness is added twice to
the surface crack length, instead of using the remaining ligament length.
• The re-categorization formula given in Annex E of BS 7910 [1]:
2ct = 2cs + t (6.3)
This re-categorization of a surface crack falls in between the two API 579-1/ASME
FFS-1 equations by adding one thickness to the surface crack length.
Using the three equations above to re-categorize the surface crack, the recommended
through thickness crack lengths, are given in Table 6.4. Only samples that resulted in the
expected penetration were assessed.
Table 6.4: Re-categorization through-wall crack length as per BS 7910 and API 579-
1/ASME FFS-1
Specimen API 579 - Eq.6.1 API 579 - Eq.6.2 BS 7910 - Eq.6.3

S22 41.10 62.50 48.15
S24 41.36 64.56 50.48
S41 41.72 63.92 49.56
S42 42.78 65.38 51.04
S43 40.44 64.04 49.62
S44 41.18 63.78 49.34
6.3 Evaluation of stress-strain field on back surface

using high strength steel specimens
Several experimental works to investigate the crack growth shape due to fatigue loading
have been carried out [144], [145], [111]. The focus is often on stress or strain field at
crack-tip and stress intensity factor (SIF) derivation [151], [152], [153], [154]. With regard
to the influence of the geometric characteristics of the crack, in the case of tension, the
SIF was shown to be maximum at the centre of the crack front for cracks with low aspect
ratio and at the surface point for cracks with high aspect ratio. However, for deeper cracks
for (a/t > 0.8) the accuracy of solutions has not been well established. The crack front
progression was also studied by means of numerical methods [155], [156] based on Paris
law [157]. It was shown that cracks would try to grow with a SIF remaining constant along
the front, but such behaviour fail when the cracks become deeper and approach the free
surface [158]. This section deals with the propagation of deep surface cracks (a/t > 0.8)
in plates subjected to cyclic tension. Four flat plate specimens were produced from high
strength steel samples (400x150x22mm). All specimens notched with part-penetrating
cracks were subject to uni-axial tensile fatigue loading. Propagation from 80% thickness
surface cracks to fully through-thickness shape was investigated. Strain variation during
the fatigue crack growth has been analysed for two different crack aspect ratios. Material
properties, test arrangement and results are presented in the following sections.
6.3.1 Material properties
This section summarises the properties of the specimens. Material properties are re-
quired as basic input for integrity assessment. All the tests were carried out using facilities
at TWI, in accordance with the respective testing standards and accreditations. The sam-
ples for chemical analysis were analysed using optical emission spectroscopy, OES (also
known as direct spark analysis). The composition is given in Table 6.5.
Table 6.5: Chemical composition of high strength steel
Element C Si Mn P S Cr Mo Ni
Wt. (%) 0.2 0.31 1.43 0.007 0.005 0.22 0.53 0.59
Mechanical properties have been measured using tensile test in accordance with BS EN
ISO 6892-1:2009 B [159]. Round tensile samples were machined from one of the available
plates as shown in Figure 6.5. The specimen dimensions and results of tensile tests are
given in Table 6.6 and the associated stress/strain curve is presented in Figure 6.6.
Figure 6.5: Schematic of cutting plan for tensile specimens

Figure 6.6: Stress-strain curve derived from tensile specimens
Table 6.6: Mechanical properties derived from tensile test
Yield strength Ultimate tensile strength Young’s Modulus Poisson’s ratio

Load (kN) Stress (MPa) Load (kN) Stress (MPa) MPa —
78.07 638.2 93.42 763.7 217147 0.3
6.3.2 Test specimens
Test specimens (400x150x22mm) consist of flat plates machined to dog-bone shape with
semi-elliptical surface-breaking notches. All specimens were notched to a depth about 80%
of the plate thickness using electro-discharge machining (EDM). A pure copper electrode
with thickness of 0.1 mm has been prepared to be used in the spark procedure. Figure
6.7 presents the dimensions of machined specimens and details of notch. Specimens edges
have been radiied to avoid fatigue initiation at an undesired location. A total of four
plates have been manufactured. Two different crack geometries were examined (later
described as ”short” and ”long” cracks). Numbering and respective geometric details of
the test specimens are given in Table 6.7.
Figure 6.7: Schematic drawing of dog-bone shaped specimens with artificially-induced

cracks
Table 6.7: Test specimens numbering and dimensions
Specimen Width Thickness Crack depth Crack length Ligament

Geometry
No (mm) (mm) a (mm) 2c (mm) p (mm)
Long M01-01 100.6 14.80 12.1 57.9 2.7
Crack M01-02 100.1 15.79 11.9 58.9 3.9
Short M01-03 100.4 14.67 12.1 23.8 2.6
Crack M01-04 100.1 15.82 12.4 24.0 3.5
6.3.3 Test arrangement
Electrical resistance strain gauges were attached on to the back surface of the plates. A
6 mm uni-axial strain gauge (TML FLA-6-11) and sixteen 1 mm uni-axial strain gauges
(TML FLA-1-11) were positioned as shown in Figure 6.8 as seen in specimen M01-01.
The distance between strain gauges was established to be 10 mm. The strain gauge at
the centre carried a resistance of 120 Ω with a gauge length and width of 6 mm and 2.2
mm respectively. The other strain gauges carried a resistance of 120 Ω with a length and
width of 1 mm and 1.3 mm respectively. They have been adhesively bonded to the plates
using cyanoacrylate adhesive. They were wired with a quarter bridge configuration to a
NI 9235 module (Logging rate = 10kS/s/ch) for data acquisition. Three types of strain
were monitored during testing:
• Strain along the plane parallel to the crack - X-axis (SG 1/7/8/9/10/11/12/13/14)
• Strain along the plane perpendicular to the crack - Y-axis (SG 2/3/4/5/6)
• Strain along the diagonal axis (SG 15/16/17)
Figure 6.8: Set-up for fatigue test with strain gauges attached onto the back surface of
specimen. Notch was introduced on the opposite surface
A static test was performed for each specimen for the purpose of calibrating the strain
gauges. Specimens were then subject to fatigue loading at room temperature in air using
a Mayes servo-hydraulic test machine with a 500kN capacity load cell. All tests were
conducted using load control with a constant sinusoidal fluctuating tensile mean load of
135kN with a load ratio R=0.1. This resulted in a cross section tensile mean stress of 82.5
MPa (≈13% x material’s yield strength). The maximum applied stress of 150kN (100
MPa), was well below the yield strength (≈ 15% x σys ). The tests were monitored using
a computer software and all loading history was recorded. Fatigue tests were considered
achieved when the through-wall crack reached one of the last strain gauges along X-axis
(SG13/SG14). The specimen was then removed from the machine and pulled apart to
have access to the fracture surface. Results of both static and fatigue test are provided
in the following two sections.
6.3.4 Static tests
A static test had been performed for each specimen for the purpose of calibrating the
strain gauges. Specimens were subject to different loads and the strains were recorded.
Calibration was divided into five stages: ie (i) 0kN, (ii) 15kN, (iii) 50kN, (iv) 100kN
and (v) 150kN. The strains measured at each stage were logged and readings were used
to set the various trip values implemented to stop the fatigue tests at desired crack
development. Far field elastic strain was estimated for each specimens based on their
respective dimensions:
F
• M 01−01 = AM 01−01 E
= 463.96µ
F
• M 01−02 = AM 01−02 E
= 437.04µ
F
• M 01−03 = AM 01−03 E
= 469.24µ
F
• M 01−04 = AM 01−04 E
= 436.21µ
where F is the force applied, E the Young’s modulus and A the cross section for each
specimen. Static test recorded strain are provided for long crack geometry (M01-01/M01-
02). Data from short crack geometry (M01-03/M01-04) are provided in AnnnexA.2.1.
The recorded values of strain in the three directions for M01-01 & M01-02 are given in
the following figures:
• Figure 6.9 - X-axis

• Figure 6.10 - Y axis
• Figure 6.11 - Diagonal axis
Figure 6.9: Evolution of strains along X-axis under increasing static loads: 0kN, 15kN,
50kN, 100kN and 150kN for Specimens M01-01 & M01-02
Figure 6.10: Evolution of strains along Y-axis under increasing static loads: 0kN, 15kN,
Figure 6.11: Evolution of strains along the diagonal direction under increasing static
loads: 0kN, 15kN, 50kN, 100kN and 150kN for Specimens M01-01 & M01-02
The profiles of static tests were observed to be similar for both geometries (short”/”long
cracks). However, recorded values differed greatly. For short cracks, the variation of
strains is limited due to the dimensions (c=a) and area of the notch (≈ 15% of the
specimen CSA). For long cracks, the area of the crack is larger (≈ 35% of the specimen
CSA) and the length longer. This led to some bending (not measured) that needs to be
balanced. It is worth to note that strain values are measured at a particular location along
each axis. Further details are provided in the next chapter (Section 7.3) where numerical
model were developed.
6.3.5 Fatigue tests
All tests were carried out using load control with a constant sinusoidal fluctuating
stress where σmax was 100 MPa, σmin was 15 MPa and σmean was 82.5 MPa, as presented
in Figure 6.12 and Table 6.8.
Table 6.8: Cyclic loading parameters
Max. Load Min. Load Load range Mean Load Load Ratio Freq.
Parameter
(kN) (kN) (kN) (kN) (–) (Hz)
Value 150 15 135 82.5 0.1 3
Figure 6.12: Schematic graph illustrating cyclic loading applied during testing
The plots obtained from specimens fatigue test on M01-01 are presented in this section.
Strain evolution during tests for the other specimens can be found in Annex A.2. Strain
gauges numbering can be found in Figure 6.8. Strain evolution on the back surface of the
plates is given in:
• Figure 6.13 - Strain evolution along the X-axis

• Figure 6.14 - Strain evolution along the Y-axis
• Figure 6.15 - Strain evolution in the diagonal direction
Figure 6.13: Evolution of the strains along the X-axis during fatigue test - M01-01
Figure 6.14: Evolution of the strains along the Y-axis during fatigue test - M01-01
Figure 6.15: Evolution of the strains for the diagonal strain gauges during fatigue test
- M01-01
The final number of cycles for each specimens tested is given in Table 6.9. For ”long”
crack specimens, the number of cycles is fairly similar (between 140k-160k cycles). For
”short” crack specimens, a large discrepancy can be observed (ranging from 470k to 710k
cycles). A stabilisation of the strain evolution for about 200k cycles has been observed
from the strain data for sample M01-04. This was assumed to be due to the notch insertion
resulting in a lighter condition at crack-tip.
Table 6.9: Number of cycles until failure for all tests
Geometry Specimen No N (Cycles)

M01-01 159629
Long crack M01-02 139704
M01-03 468656
Short crack M01-04 709870
Figure 6.13 presents the strain variations along X-axis as the crack propagated. Initial
strains were in good agreement with static tests carried out earlier. The difference in the
peak maximum and plateau observed for the strain gauges SG1, SG7, SG8, SG9 and SG10
is due to the failure of the adhesive before the strain gauge. When rupture of the strain
gauge occurs before debonding, the maximum strain dropped instantly (i.e. SG11, SG12,
SG13 and SG14). The first variation has been observed in SG1, located where the ligament
was the thinnest (i.e. X=0mm), after ≈20000 cycles. Other strain gauges remained stable
until SG1 value reached ≈3000 µm/m. Break-through was estimated to occur when strain
values from SG1 were ≈3000-3500 µ (≈80000 cycles). Afterwards, all recorded values
from strain gauges started to increase along X-axis. It can be observed for the couple
SG7/SG8 showed an increase in recorded strain values at the approximate cycles where
breakthrough occurred. This implies that break-through did not occur exactly at the
centre of the plate where SG1 was attached (i.e X=0mm) but at a distance offset from it.
This effect was less pronounced for the couple SG9/SG10, indicating that soon after crack
penetration, crack growth occurred at a higher rate on one side to grow in a symmetric
way for the remaining.
Figure 6.14 presents the strain variations along Y-axis as the crack propagated. Similar
to the observation along X-axis, the break-through was estimated to have occurred after
≈80000 cycles. The higher variation rate for strain gauges close to the crack plane (SG2
and SG3) and a sudden change in the slope of the recorded values for strain gauges far from
the crack plane (SG4, SG5 and SG6) has been observed. Opposite to observations along
X-axis, strain values decreased as the crack grew. Close to the crack plane (SG2/SG3),
the first variations were observed after ≈ 20000 cycles (similar to SG1 behaviour along
X-axis) with descending values of recorded strain. While recorded strains from SG1
showed a sharp increase from 60000 to 80000 cycles, strain values for SG2/SG3 showed
increased descending rate. For strain gauges located further away from the crack plane
(SG4/SG5/SG6), there was a gradual decrease in strain between 60k and 80k cycles
but a swifter decrease only occurred after breakthrough (similar to observations along
X-axis).
Figure 6.15 presents the strain variations along the diagonal direction as the crack
propagated. Unlike the observations along X and Y axes, the effect of break-through
was not significant and really captured. It is provided based on previous observations.
However, similar behaviour as that along Y-axis has been observed. The recorded strain
for the strain gauge closest to the crack plane (SG15) showed decrease first, while for
strain gauges located further, the drop in recorded strain was more gradual, with SG16’s
strain profile being between SG15 and SG17.
6.3.6 Discussion
Based on the observations made on the strain data recorded during the tests for each
geometry, different aspect have been investigated:
• From Figure 6.13, break-through was estimated to occur when strain values from
SG1 were ≈3000-3500 µm/m. The strains at yield (Eq.6.4) and ultimate tensile
stress (Eq.6.5) were calculated using uni-axial Hooke’s law:
σys
ys = = 2984µ (6.4)
E
σuts
uts = = 3518µ (6.5)
E
The graphically estimated location of break-through corresponds to a value located

between yield and ultimate tensile strains. These closest recorded values were used
as boundaries to estimate the number of cycles range when break-through happened:
– Strain=2984 ⇔ Cycle=80093
– Strain=3522 ⇔ Cycle=80865
• From Figure 6.14, break-through was estimated to occur when:
– Strain gauges close to the crack plane (SG2 and SG3) shown their higher
variation rate
– Strain gauges far from the crack plane (SG4, SG5 and SG6) shown a sudden
change in strain variation
Variation in strains was calculated for all strain gauges along Y-axis. Value of
the tangent was calculated at each data point based on strain values within a de-
fined set of 50 cycles. Since the recorded reading of strain exhibited high level of
noise, the strain variation rates presented in Figure 6.16 were filtered using MatLab-
incorporated moving average function (rloess method). Actual data were used to
estimate the number of cycles where minimum variation was observed. Similar cal-
culations were carried out for SG1 along X-axis. In both cases, SG1 (i.e. X=0mm)
and SG2/SG3 (i.e. Y=10mm) exhibited greater variation rate in strain values with
a maximum at cycle N=80596. As seen previously, this corresponds to a strain value
between yield and ultimate tensile strains.
Figure 6.16: Evolution of the strain variation along Y-axis

• As discussed previously, strain variation profiles were similar for both geometries but
the recorded values were different. The initial aspect ratio was found to influence the
number of cycles for the crack to grow from one strain gauges to another. Table 6.10
provides the recorded values for Specimen M01-01. Figure 6.17 shows the number
of cycles when strain gauges measurement reached the estimated yield strain along
X-axis for both geometries. The number of cycle was normalised against the central
strain gauges (SG1) and calibrated at 0 cycles. The averaged crack growth rate
between strain gauges based on the yield strain value is also provided in Figure
6.18.
Table 6.10: Number of cycles against strains at yield and ultimate tensile stress along
X-axis - M01-01
YS UTS
M01-01 Cycles Strains Cycles Strains
SG1 80093 2984 80865 3522
SG7 89875 2983 90235 3514
SG8 84603 2978 85053 3505
SG9 109730 2989 110150 3536
SG10 111410 2990 111800 3535
SG11 139294 2984 139535 3513
SG12 141127 2981 141382 3513
SG13 154656 2986 154971 3500
SG14 156276 2988 156561 3531
Figure 6.17: Evolution of crack growth rate along X-axis for all specimens (Cycles)
Figure 6.18: Evolution of the crack growth rate along X-axis for all specimens (Cy-
cles/mm)
6.3.7 Observations from tested plates
Fatigue tests were terminated when the through-wall cracks reached one of the last
strain gauges along X-axis (SG13/SG14). The specimen was then removed from the
machine and pulled apart to examine to the fracture surface.
6.3.7.1 Photographs of fracture surfaces
Photographic record of each geometry is presented in this section. Figure 6.19 presents
the fracture surface from M01-01 which is a long crack and Figure 6.20 presents the
fracture surface from M01-03 which is a short crack. Photographs of the remaining two
specimens can be found in Annex A.3.1.
Four areas were identified on the associated fracture surfaces:
• 1 - Initial crack (blue)

• 2 - Estimated shape at breakthrough (red)
• 3 - Through-wall crack growth (green)
• 4 - Final rupture (pink)
Tests were stopped at different strain levels to introduce a mix of soap and water
solution in the crack. This offered the advantage of identifying the stage at which the crack
penetrated the back surface by the formation of bubbles. This also introduced oxidation
on crack surfaces, leaving beach marks to give a history of the crack shape.
This is particularly relevant to specimen M01-03 where the test has been stopped many
times and sufficient time was allowed before to restart.
For long cracks (M01-01/M01-02) as seen in Figure 6.19, the crack growth occurred
primarily in the depth direction. The initial crack length remained the same until break-
through and followed by some through-thickness growth. Upon break-through, the crack
length at the back surface continued to grow from the region bounded by the red line in
the width direction (B to B’) but the initial crack length only underwent very little exten-
sion (A to A’). Fatigue crack growth continued until an almost uniform through-thickness
flaw shape was achieved, with a similar length on both side of the plate.
For short cracks (M01-03/M01-04) as seen in Figure 6.20, the crack growth was uniform
in both depth and length direction until breakthrough. Crack length increased at a higher
rate on the back surface (B to B’) until a similar length was observed on both side.
However, unlike the long crack, the initial surface crack length always increased during
the test (A to A’).
It was evident from Figure 6.19 and Figure 6.20 that the evolution from break-through
to non-idealised flaw (post break-through) was dependent on the aspect ratio/geometry
of the initial flaw size.
Figure 6.19: Fracture surface of Specimen M01-01

Figure 6.20: Fracture surface of Specimen M01-03
6.3.7.2 Scanning Electron Microscope (SEM) photographs
As part of the effort to get a better understanding of the behaviour during the crack
evolution, SEM analysis were carried out with a Zeiss EVO LS15 microscope. Specimens
M01-02 and M01-04 were chosen to represent long and short cracks respectively and were
less corroded than those presented in Figures 6.19 and Figure 6.20. Fracture surface was
carefully cleaned before analysis. However, the use of water/soap solution has oxidised
the fracture surface significantly. For this reason, most of the physical feature of the
fatigue crack growth were either not available or difficult to detect. Majority of the areas
could not be thoroughly examined so only general observations were made. Observation
of crack growth on both specimens were similar to those of M01-01 and M01-03. Figure
6.21 gives a schematic of the location where the photographs where taken. Four areas
were considered:
1. Transition between initial notch and cross section
2. Break-through location
3. Some distance after break-through
4. Region away from initial notch

Figure 6.21: Schematic of the location of photographs from SEM analysis
(a) Location 1 (b) Location 2
(c) Location 3 (d) Location 4
Figure 6.22: Photographs from SEM analysis

Figure 6.22a shows a general view (200 µm) between the initial crack and the cross
section at initial crack-tip front. Typical initiation of fatigue can be observed along the
boundary. Severe corrosion can be noted with the presence of numerous pits. Figure
6.22b shows the area adjacent to the break-through. Three distinct regions were ob-
served. Classic fatigue pattern could be seen until the last stage of propagation. Once the
crack broke through, another area of transition (flat, blunt) was observed as well as the
formation of a lip. Figure 6.22c shows the transition area. Classic fatigue pattern could
still be observed from the initial crack. A transition of classic fatigue, crack blunting and
corrosion was also evident. Figure 6.22d was taken at a region away from the transition
stage. Propagation was dominated by classic fatigue only before the end of the test.
6.3.7.3 Crack shape recategorisation
Similarly to calculations in Section 6.2.5, the dimensions of recategorised through-wall

cracks as per BS 7910 and API 579 were determined. Using equations Eq.6.1 to Eq.6.3,
the calculated values are given in Table 6.11. A photograph of M01-02 (Figure 6.23) and
M01-03 (Figure 6.23) are provided to represent both geometries. Other specimens can be
found in Annex A.3.2 .
Table 6.11: Recategorised through-wall crack length (mm) as per BS 7910 and API
579-1/ASME FFS-1
Specimens
Equations M01-01 M01-02 M01-03 M01-04
API 579 - Eq.6.1 63.3 66.6 29.0 30.9
API 579 - Eq.6.2 87.5 90.5 53.2 55.6
BS 7910 - Eq.6.3 72.7 74.7 38.5 39.8
Figure 6.23: Recategorisation into through-wall crack for Specimen M01-02

Figure 6.24: Recategorisation into through-wall crack for Specimen M01-03
As described in Section 6.2.5, recategorisation as per BS 7910 falls between the values
derived using API 579’s general and LBB formulae. For both aspect ratios, the general
equation in API 579-1/ASME FFS-1 predicts a smaller through-wall crack when compared
to the observed actual shape, especially for short cracks (i.e. Figure 6.24). On the other
hand, API’s solution recommended in a LBB assessment appears to over-estimate the
flaw size for long cracks. The same has been observed in Perspex samples. In terms
of recategorisation, BS 7910 rule appears to be more adequate than those of API 579-
1/ASME FFS-1. However, based on the observations made in Chapter 5, assessment of
COA could still be over-estimated.
6.4 Conclusions
In this chapter, experimental work undertaken was presented. As a semi-qualitative

test, fatigue testing on surface cracked Perspex plate specimens showed that: (a) If the
sample was not loaded strictly perpendicular to the applied load, fatigue crack growth
resulted in a non-symmetric pattern. This led to rapid extension of the crack size on
one side and the specimen failed prematurely; (b) Thinner samples experienced bending
during the tests. This resulted in an increase in crack length on surface and crack growth
arrest in the thickness direction; and (c) Thicker samples allowed the crack to grow
through-thickness. They achieved expected crack shape evolution. Once the crack had
broken through, a rapid crack growth was observed until the crack shape reached an
“idealised” through-wall shape. This has been confirmed by the FEA work carried on
previously, (see Chapter 5), of which a high SIF values was predicted at the back surface
of a non-idealised through-wall crack.
Two different crack geometries were examined using metallic specimens to represent
short and long cracks and the following are concluded: (a) The measured strains al-
lowed the accurate prediction of break-through when recorded strains where between
yield and ultimate strain of the material and (b) Observations along three axes (i.e. par-
allel/perpendicular/diagonal) of the crack showed that break-through can be estimated
with a rapid increase of the strain variation on the back surface.
Fracture surfaces from both Perspex (Series 2 & 4) and metallic specimens showed
similar behaviour at break through. While the materials are different, further test on
Perspex would permit an easier way to characterise the transition from part-penetrating
to through thickness crack with direct observations during the test. Metallic specimens
showed a change in propagation behaviour during the transition but could not be precisely
established due to the corrosion. Similar observation was made on Perspex specimens.
Recent advance in testing such as DIC measurement [160] would permit a better under-
standing on metallic specimens with a full field characterisation of the strain evolution on
back surface.
Observations made on the fracture surfaces also showed the importance of the aspect
ratio in the LBB assessment: (a) For long cracks, the propagation was primarily in the
depth direction. Upon break-through, the crack length increased rapidly on the back
surface. Once similar length was reached on both sides, propagation of through-thickness
crack occurred at same rate; (b) For short cracks, the initial shape remained the same until
break-through. The change in shape occurred during the through-thickness propagation
and (c) The transition and growth from a part-penetrating flaw to through-thickness is
dependent on the aspect ratio of the initial flaw.
Chapter 7
Numerical Simulation - Stress-Strain

Field on the Back Surface of Surface
Cracked Plate During Crack
Transition
Contents
7.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 168
7.2 Models description . . . . . . . . . . . . . . . . . . . . . . . . . 168
7.2.1 Plate and cracks geometry . . . . . . . . . . . . . . . . . . . . . 168
7.2.2 The effect of crack geometry on back surface stress field . . . . 169
7.2.3 Mesh sensitivity study . . . . . . . . . . . . . . . . . . . . . . . 170
7.3 Validation against experimental data . . . . . . . . . . . . . . 174
7.3.1 Model geometries and results . . . . . . . . . . . . . . . . . . . 174
7.3.2 Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 175
7.4 Stress analyses along axes parallel and perpendicular to crack178
7.4.1 Stress analyses along X-axis . . . . . . . . . . . . . . . . . . . . 179
7.4.2 Stress analyses along Y-axis . . . . . . . . . . . . . . . . . . . . 184
7.4.3 Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 188
7.5 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 190
167
CHAPTER 7. FEA 2 - SURFACE CRACKS IN PLATES 168
7.1 Introduction
This chapter describes the work undertaken to examine the transition of crack from
surface-breaking to through-thickness. The validation of available solutions for through-
thickness cracks and investigation on the crack shape have been discussed in Chapter 5.
In this chapter, FEA were carried out to investigate the stress/strain field on the back
surface of a plate with a part-penetrating crack. Various FE models have been constructed
to evaluate the stress field on the back surface of cracks from 20% to 95% thickness. Two
crack geometries have been selected to represent ”short” and ”long” cracks. The change in
stress along the axis of the crack and the one perpendicular to it were also examined.
Figure 7.1: Schematic of a cracked plate with crack growth through-thickness
7.2 Models description
7.2.1 Plate and cracks geometry
Various models with crack depths ranging from 20% to 95% thickness have been cre-
ated for cracks to capture all the possible scenarios during crack evolution. Two crack
geometries have been selected to represent ’short’ and ’long’ cracks. The dimensions of
the cracks and plates have been designed to correspond to the experiments described in
the previous chapter. Plates were 15 mm thick and 100 mm wide. Crack dimensions
were varied. Taking advantage of the symmetries, only 1/4 of the plate of interest was
modelled. A typical model used in this work for both geometries is illustrated in Figure
7.2. A crack was introduced in the middle of the plate. Different simulations were pre-
pared for each geometry to include crack depth ranging from 20% to 90-95% of thickness.
Crack aspect ratios were kept constant: (i) c=a for Geometry 1, (ii) c'2a for Geometry
2. A tensile load of 100 MPa (150 kN) was applied to the ends of the plate. The key
outputs derived from crack simulations include Von Mises stresses along two axes (as seen
on Figure 7.2):
• (i) X-axis which is in plane with the crack

• (ii) Y-axis which is perpendicular to the crack
(a) Geometry 1 - Short crack (b) Geometry 2 - Long crack
Figure 7.2: FE models for the two crack aspect ratios
7.2.2 The effect of crack geometry on back surface stress field
As a preliminary study, models have been prepared with a crack depth of 80% thickness.
This was selected to duplicate the dimensions and geometry of the metallic specimens in
Chapter 6. The crack-tip radius was modelled to investigate the effect of a possible
blunted notch that can occur in experiments, depending on the notch introduction. Four
different notch radii thought to be relevant were included (Figure 7.3):
• Sharp crack
• Blunt notch - Radius=0.125mm
The resultant stress profiles along X-axis and Y-axis are presented in Figure 7.4.
Figure 7.3: Schematics of sharp crack and blunt notches geometry investigated
(a) X-axis (b) Y-axis
Figure 7.4: Comparison between sharp crack and blunt notches
It can been seen that models with sharp crack and blunt notches displayed very similar
back surface stress profiles on the of the plate. Models with blunt notches showed a slight
increase in the stress field on back surface. However this was not thought to have a
significant impact on the work described in this chapter. The larger the notch radius
produced higher Von Mises stress on back surface.
7.2.3 Mesh sensitivity study
Sensitivity studies have been carried out to improve confidence in the output data. For
both aspect ratios, four mesh sizes have been included for a 80% deep crack. Table 7.1
provides a list of models, mesh size and number of elements employed. Mesh sensitivity
studies were carried out on both X- and Y-axes. A coarse mesh (i.e. 2 mm) was first
introduced. The mesh size was then refined from 2 mm to 0.25 mm to highlight the
inaccuracy issues when using a coarser mesh. This also helped to decide which mesh size
was adequate.
Figure 7.5 and Figure 7.6 present the plots of Von Mises stresses along X-axis when
different mesh sizes were employed, whereas Figure 7.7 and Figure 7.8 on the Y-axis, for
Table 7.1: Number of elements in the various models used for mesh sensitivity study
Number of elements
Mesh size (mm) Geometry 1 Geometry 2
(Short crack) (Long crack)
2 2634 2546
1 16054 15837
0.5 50272 44036
0.25 191293 173892
both crack aspect ratios respectively. Numerical values of Von Mises stress are given in
Table 7.2 for X-axis and Table 7.3 for Y-axis.
From the sensitivity studies, models with a mesh size of 0.5 mm were created. The
stress values derived from models with this mesh size were identical to those from the finest
mesh (i.e. 0.25mm) with a maximum variation of 0.5% on X-axis (See Table 7.2) and 1%
on Y-axis (See Table 7.3). For all mesh sizes under consideration, the maximum deviation
in stress along X-axis has been observed in the first 15 mm in a ’short’ crack. Beyond 15
mm from the crack tip, as seen in Figure 7.5, similar results were obtained from all four
mesh sizes. The model with a ’long’ crack (Figure 7.6) demonstrates that consistency
in Von Mises stress was achieved when a mesh of 0.5 mm or finer was used and mesh
refinement was required within 25 mm of crack tip. With a mesh size of 0.5 mm, models
were still manageable in terms of running time and considered to be accurate.
Table 7.2: Mesh sensitivity study along X-axis - Values of Von Mises stresses
X (mm) 2 mm Diff. (%) 1 mm Diff. (%) 0.5 mm Diff. (%) 0.25 mm

0 121.7 1.51 116.1 3.09 119.3 0.45 119.9
Geometry 1
10 98.6 0.66 98.9 0.43 99.2 0.12 99.3

20 101.6 0.17 101.7 0.07 101.8 0.02 101.8
30 102.6 0.1 102.7 0.05 102.7 0.02 102.7
40 102.1 0.07 102.2 0.03 102.2 0.01 102.2
50 100.4 0.002 100.4 0.02 100.4 0.004 100.4
0 142.8 1.02 136.8 3.14 140.8 0.28 141.2
Geometry 2
10 129.5 3.05 124.5 0.92 125.4 0.21 125.6

20 103.3 0.75 103.4 0.57 104.0 0.07 104.0
30 104.7 0.36 104.9 0.13 104.7 0.34 105.0
40 107.7 0.20 107.9 0.09 107.9 0.02 107.9
50 101.2 0.06 101.1 0.02 101.2 0.002 101.2
Table 7.3: Mesh sensitivity study on Y-axis - Values of Von Mises stresses
Y (mm) 2 mm Diff. (%) 1 mm Diff. (%) 0.5 mm Diff. (%) 0.25 mm

0 121.662 1.51 116.139 3.09 119.311 0.45 119.847
Geometry 1
10 116.656 0.61 114.378 1.35 115.002 0.82 115.947

25 96.129 1.60 94.431 0.20 94.675 0.06 94.618
50 97.637 0.08 96.414 1.18 97.575 0.015 97.561
100 99.546 0.014 99.549 0.018 99.535 0.003 99.532
0 142.67 1.02 136.794 3.14 140.834 0.28 141.234
Geometry 2
10 151.774 2.10 147.205 0.98 148.728 0.05 148.657

25 98.034 4.05 97.525 3.51 95.277 1.13 94.215
50 93.852 0.16 93.766 0.07 93.717 0.014 93.705
100 98.788 0.03 98.772 0.012 98.762 0.0025 98.760
Figure 7.5: Mesh sensitivity study on X-axis - Geometry 1 (Short crack)

Figure 7.6: Mesh sensitivity study on X-axis - Geometry 2 (Long crack)
Figure 7.7: Mesh sensitivity study on Y-axis - Geometry 1 (Short crack)

Figure 7.8: Mesh sensitivity study on Y-axis - Geometry 2 (Long crack)
7.3 Validation against experimental data
7.3.1 Model geometries and results
In this section, the recorded strain values from static test undertaken prior to the fatigue
testing (See Section 6.3.4) for the four plates have been compared against numerical values.
Experimental values were recorded by the strain gauges mentioned on figures and location
can be found in Figure 6.8. Numerical values have been obtained from two models: (i)
FEAgeom1 and (ii) FEAgeom2. Models dimensions and properties are defined in Table
7.4.
Table 7.4: Models for validation against experimental results
Crack length Crack depth Min. element size

Nb. of elements
Model 2c (mm) a (mm) X-axis (mm) Y-axis (mm)
FEAgeom1 24 12 0.5 0.5 48980
FEAgeom2 56 12 0.5 0.5 49218
The continuous FEA estimated strain values are presented in :
• Figure 7.9 - X-axis (Density=0.5mm all along the axis)

• Figure 7.10 - Y-axis (Density=0.5mm at crack-tip and increasing gradually)
and the averaged values from FEA are given in:
• Figure 7.11 - X-axis for FEAgeom1 (M01-03/M01-04)

• Figure 7.12 - Y-axis for FEAgeom1 (M01-03/M01-04)
• Figure 7.13 - X-axis for FEAgeom2 (M01-01/M01-02)

• Figure 7.14 - Y-axis for FEAgeom2 (M01-01/M01-02)
7.3.2 Discussion
Figure 7.9 and Figure 7.10 are included to show the continuous form of the FE-derived
strains. It can be seen that values are close to the far field strains for position far from
the deepest point of the crack (i.e [-50 -20] and [20 50]) especially along Y-axis. When it
comes closer to the deepest point of the crack (i.e 0 mm), the strain is varying the more.
These continuous values could not be compared directly to the experimental readings
as the sampling areas were different, so FE values were averaged with respect to strain
gauges sampling areas. This is particularly important for the strain gauge SG1 at the
centre, which was 6 mm long. Averaged FEA values were then compared to the static
experimental values (see Section 6.3.4). It can be seen that FE models and experiments
are in good agreement. The difference could be explained by the fact that FE models
simulated idealised dimensions for both plates and cracks. However the exact dimensions
from the specimens varied slightly from these idealised values.
Figure 7.9: FEA continuous values on X-axis - Geometry 1 & Geometry 2 (Both crack)
Figure 7.10: FEA continuous values on Y-axis - Geometry 1 & Geometry 2 (Both crack)
Figure 7.11: Averaged strain values derived from FEA vs. Experimental readings on
X-axis - Geometry 1 (Short crack)
Y-axis - Geometry 1 (Short crack)
X-axis - Geometry 2 (Long crack)
Y-axis - Geometry 2 (Long crack)
7.4 Stress analyses along axes parallel and perpen-

dicular to crack
A matrix of the cases investigated is given in Table 7.5.
Table 7.5: FEA matrix of the models for part-penetrating crack
Geometry 1 Geometry 2
(Short crack) (Long crack)
a/t
a c a c
0.20 3 3 3 7
0.40 6 6 6 14
0.50 7.5 7.5 7.5 17.5
0.60 9 9 9 21
0.65 9.75 9.75 9.75 22.75
0.70 10.5 10.5 10.5 24.5
0.75 11.25 11.25 11.25 26.25
0.80 12 12 12 28
0.825 12.375 12.375 12.375 28.875
0.85 12.75 12.75 12.75 29.75
0.875 13.125 13.125 13.125 30.625
0.90 13.5 13.5 13.5 31.5
0.925 13.875 13.875 13.875 32.375
0.95 14.25 14.25 14.25 33.25
7.4.1 Stress analyses along X-axis
This section discusses the analysis of stresses evolution along the X-axis of both crack
aspect ratios. This is the axis which is in plane with the crack. The stresses on the back
surface of the plate are plotted along this axis from the deepest point at the crack tip (X
= 0 mm) to the edge of the plate (X = 50 mm).
7.4.1.1 Geometry 1 - Short crack with c=a
This first case considers a crack having an half crack length equal to its crack depth
(c=a). A schematic of a typical crack evolution is given in Figure 7.15 and an example
of a typical mesh is given in Figure 7.16 for a crack equal to 80% thickness. This aspect
ratio of c=a has been kept constant for crack depths ranging from 20% to 95% through-
thickness.
Figure 7.15: Schematics of the first geometry crack depths
Figure 7.16: Typical mesh for the first geometry where crack depth a=0.8t
Figure 7.17 presents the Von Mises stresses along the X-axis for different crack depths.
The horizontal axis represents the distance along the X-axis from the deepest point at
the crack tip (X = 0 mm) to the edge of the plate (X = 50 mm). The vertical axis
shows the Von Mises stresses derived from FEA. A close-up of the first 10 mm is provided
as the stresses further away from the crack tip do not exhibit significant changes with
increasing crack depth. Below 0.7t crack depth, the difference in the far field stresses is
hardly visible. The stresses at X = 0 mm tend to be even slightly lower than that of the
far field stress. The elastic stress field at advancing crack tip only became visible on the
back surface when the crack was at least 75% deep.
Figure 7.17: Von Mises stresses along the X-axis for various crack depths - Geometry 1
(Short crack)
The maximum Von Mises stress value on the back surface of the plate was obtained at
the front of the deepest point of the crack tip (X = 0 mm). The results are given in Table
7.6 for a/t ranging from 0.2 to 0.95. The difference between the estimated Von Mises
stress and far field stress has also been included. Figure 7.18 presents the values plotted
against a/t ratios. As expected, there is a sharp increase in stress field at the back surface
when the crack tip advances deeper (70%) into the component section.
Table 7.6: Von Mises stresses along the X-axis for various crack depths - Geometry 1
(Short crack)
a/t (%) 20 40 50 60 70 75 80 82.5 85 87.5 90 92.5 95

Von Mises (MPa) 99 97 97 98 104 110 116 123 137 150 165 191 231
Diff. (%) 1 3 3 2 4 10 16 23 37 50 65 91 131
Figure 7.18: Maximum Von Mises stresses along the X-axis for various crack depths -
Geometry 1 (Short crack)
7.4.1.2 Geometry 2 - Long crack with c'2a
The second FEA model considered a crack having an half-length close to two times
the crack depth (c'2a). The schematic of a typical crack evolution is given in Figure
7.19 and an example of a typical mesh is given in Figure 7.20 for a crack depth of 80%
thickness . The crack ratio has been kept constant for crack depth ranging from 20% to
92.5% through-thickness.
Figure 7.19: Schematics of the second geometry crack depths
Figure 7.20: Typical mesh for the second geometry where crack depth a=0.8t
Figure 7.21 presents the Von Mises stresses along the X-axis for various crack depths.
The horizontal axis represents the distance along the X-axis from the deepest point at
the crack tip (X = 0 mm) to the edge of the plate (X = 50 mm). The vertical axis shows
the Von Mises stresses. A close-up of the first 20 mm is provided as the stresses further
away from the crack tip do not experience significant changes with increasing crack depth.
Similar to ’short’ crack model (Section 7.4.1.1), below 0.7t crack depth, the difference in
the far field stresses was limited. The stresses at X = 0 mm tend to be even slightly lower
than that of the far field stress. The elastic stress field at advancing crack tip reaches a
maximum of about 260 MPa when the crack is at 0.925t. Again, as observed in short
crack models, the effect of crack tip stress field is visible at the back surface when crack
depth is at least 0.7t.
Figure 7.21: Von Mises stresses along the X-axis for various crack depths - Geometry 2
(Long crack)
The maximum value of Von Mises stress was obtained in front of the deepest point of
the crack tip (X = 0 mm).The results are presented in Table 7.7. The difference between
the estimated Von Mises stress and far field stress has also been included. Figure 7.22
presents these values plotted against increasing a/t ratios.
Table 7.7: Maximum Von Mises stresses along the X-axis for various crack depths -
Geometry 2 (Long crack)
a/t (%) 20 40 50 60 70 75 80 82.5 85 87.5 90 92.5

Von Mises (MPa) 98 94 96 101 113 124 137 148 163 191 217 258
Diff. (%) 2 6 4 1 13 24 37 48 63 91 117 158
Figure 7.22: Maximum Von Mises stresses along the X-axis for various crack depths -
Geometry 2 (Long crack)
7.4.1.3 Discussion
Stresses on the X-axis have been investigated for two crack geometries (i.e. ’short’
and ’long’ cracks). In both analyses, it has been observed that for a depth below ap-
proximatively 70-75% of thickness, the presence of a crack had a negligible effect on the
back surface of the plate. For a depth greater than 70-75% of thickness, the maximum
stress was observed at the deepest point (X = 0 mm) with the elastic stress field ahead of
crack tip being visible in the back surface. This effect was more evident in longer crack
with greater predicted Von Mises stresses. This can be seen in Figure 7.23 in which the
maximum Von Mises stresses have been normalised to the applied stresses on the y-axis
and crack depths normalised to component thickness on x-axis for both geometries.
Figure 7.23: Maximum Von Mises stresses along X-axis for the different crack depths -
Geometry 1 & 2
Another observation was made to examine the extend of the area affected by the
increase of stress on X-axis. In contrast to Geometry 1, the effect on back surface due to
the crack stress field is stretched further in Geometry 2. This presumingly was due to the
spread of the cracked area on the X-axis, which is greater in the second geometry. The
extend to which the stresses are varying was limited to the first 10 mm in a short crack
(See Figure 7.17). On the other hand, for the second geometry, this area was increased
to about 20 mm (See Figure 7.21).
7.4.2 Stress analyses along Y-axis
This section looks into the analyses of the stress evolution along the Y-axis (i.e. in
the plane perpendicular to the crack - See Figure 7.2). Stress values were extracted along
this axis from the deepest point at the crack tip (Y = 0 mm) to the edge of the plate (Y
= 100 mm). An example of the model is given in Figure 7.24 for a crack depth equal to
80%.
Figure 7.24: Typical mesh on the Y-axis (a/t=0.8)

7.4.2.1 Geometry 1 - Short crack with c=a
The first model considers a crack having an half-length equal to the crack depth (c=a).
Figure 7.25 presents the predicted Von Mises stresses along the Y-axis for different crack
depths. The horizontal axis represents the distance along the Y-axis from the deepest
point on the crack tip to the edge of plate. The vertical axis shows the Von Mises
stresses. A close-up of the first 10 mm is provided as it is in this region that the greatest
fluctuations of stress occur.
Figure 7.25: Von Mises stresses along the Y-axis for various crack depths - Geometry 1
(Short crack)
The maximum value of stress was obtained at a different position along the Y- axis
compared to that along the X-axis. This peak value along with its associated position (i.e.
distance from crack tip) is given in Table 7.8. Figure 7.26 presents the results in Table
7.8 on a plot to illustrate the shift of maximum value as the crack depth increases.
Table 7.8: Maximum Von Mises stresses and associated positions along Y-axis for various
crack depths - Geometry 1 (Short crack)
a/t (%) 20 40 50 60 70 75 80 82.5 85 87.5 90 92.5 95

Y (mm) 10.4 8.8 8 6.5 5.2 4.5 3.1 3.3 2.7 2.1 2.1 1.6 1
Von Mises (MPa) 101 104 108 114 125 134 148 157 170 187 211 251 322
Figure 7.26: Position of maximum Von Mises stresses along Y-axis for various crack
depths - Geometry 1 (Short crack)
Along the X-axis, maximum values were observed at similar positions regardless of
the crack depths, at the deepest point of the crack. Along the Y-axis, when the crack
is shallow, it has been observed that it has little effect on the stresses. However, along
the axis perpendicular to the crack (i.e Y-axis), the maximum value has been observed
at some distance away from the crack tip, even if the value is quite negligible. As the
crack depth increases, the maximum value also increases and its associated position moves
closer to the crack tip. Initially located about 10 mm away from the origin (i.e. Y = 0
mm), it has moved to 1 mm close to the origin when the crack depth reaches 95%.
7.4.2.2 Geometry 2 - Long crack with c'2a
The second case considers a crack having an half-length equal to twice the crack depth
(c'2a). Figure 7.27 presents the Von Mises stresses along the Y-axis for various crack
depths. The horizontal axis represents the distance along the Y-axis from the deepest
point on the crack tip to the edge of the plate. The vertical axis shows the Von Mises
stresses. A close-up of the first 20 mm is provided as this is the region that greatest
fluctuation in stress has been observed.
Figure 7.27: Von Mises stresses along the Y-axis for various crack depths - Geometry 2
(Long crack)
The maximum value of Von Mises stress occurs at different positions along the Y-axis.
This value is given in Table 7.9 for all crack depths. The evolution of the position is given
in Figure 7.28 for 50%, 75% and 90%. Figure 7.29 presents a plot of positions at which
maximum Von Mises stress occurs against a/t ratios.
Table 7.9: Maximum Von Mises stresses and associated positions along Y-axis for various
crack depths - Geometry 2 (Long crack)
a/t (%) 20 40 50 60 70 75 80 82.5 85 87.5 90 92.5

Y (mm) 10.4 9.6 8.8 7.3 5.8 4.5 3.9 3.3 2.7 2.7 2.1 1.6
Von Mises (MPa) 102 112 121 135 157 174 201 218 241 274 321 396
(a) Geometry 2 - 50% a/t (b) Geometry 2 - 75% a/t (c) Geometry 2 - 90% a/t
Figure 7.28: Evolution of the position of maximum stress value along Y-axis
depths - Geometry 2 (Long crack)
7.4.3 Discussion
Stresses on the Y-axis have been investigated for two crack geometries. In both models,
it has been observed that the presence of a crack did have an effect on the strain field
on the back surface of the plate for shallow cracks. This effect was more significant for
longer cracks (Geometry 2) which Von Mises stress field ahead of crack tip increased more
sharply than those of shorter cracks. This can be seen in Figure 7.30 where the ratio of the
maximum Von Mises stresses are normalised to the applied stresses on vertical axis and
crack depth normalised to thickness on x-axis for both geometries. However the location
of the maximum stresses along Y-axis was not at the origin as for X-axis.
Opposite to the observations for the X-axis, the area affected by the stress rise due to
the crack on the back surface was not extended for Geometry 2. Both geometries having
the same depth from the Y-axis point of view, no difference was expected. Another
difference was the location of the maximum values. While they were located at (X =
0 mm) along the X-axis, they were progressing with the crack depth along Y-axis. The
maximum values were observed to be at some distance from the crack tip, even if negligible
(Slightly higher than far field stress). As the crack depth increased, the maximum value
increased and moved closer to the crack tip. Initially located about 10 mm far from the
origin (' 70% of the plate thickness), it moved to 1 mm when the crack depth reached
95%. Both geometries had the same trends as shown in Figure 7.31.
Figure 7.30: Maximum Von Mises stresses along Y-axis for various crack depths -
Geometry 1 (Short crack) & 2 (Long crack)
depths - Geometry 1 (Short crack) & 2 (Long crack)
7.5 Conclusions
In this chapter, different FEA models using a 0.5mm mesh with semi-elliptical cracks
in a plate were used to examine the stress field developing on the back surface along two
axes: (i) along the crack and (ii) perpendicular to the crack. Several observations could
be drawn:
(a) While the notch radius increased, the stress field on back surface increased accord-
ingly. In this chapter, a sharp notch model has been employed.
(b) FEA results agreed well with experimental results for both axes and geometry
investigated giving confidence in the numerical models. The minor difference between
FEA and experiments is due to the small discrepancy in flaw geometry.
(c) Along X-axis (Section 7.4.1), stresses on back surface were observed to be maximum
at the deepest point of the crack (i.e X=0). Stresses were close to the far field stress value
until for a crack depth less than 60% of the thickness. This back surface stress increased
when the crack depth increased and at a greater rate in a long crack.
(d) Along Y-axis (Section 7.4.2), stresses were observed to have a maximum at a
position varying with the crack depth. The maximum stresses along Y-axis came closer
the origin (i.e Y=0) as the crack depth increased.
Chapter 8
Concluding remarks and future

work
Contents
8.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 192
8.2 Summary and Conclusions . . . . . . . . . . . . . . . . . . . . . 192
8.3 Main PhD Achievements Findings . . . . . . . . . . . . . . . . 196
8.4 Recommendations for future work . . . . . . . . . . . . . . . . 197
191
CHAPTER 8. CONCLUDING REMARKS AND FUTURE WORK 192
8.1 Introduction
The work discussed in the previous chapters presents the investigation into the leak-
before-break approach. This chapter gives a summary of the study carried out in these
previous chapters and key observations are addressed. Possible directions of the future
research to improve upon the current work are also outlined in Section 8.4.
8.2 Summary and Conclusions
In Chapter 2, a literature study has covered different aspects of the subject. An

overview of the defects that can be present at different stages of components’ life ((i)
during fabrication, (ii) in-service and (iii) due to third-party) was presented. The root
of the damages was shown to be important with regards to the criticality assessment.
Non-Destructive Testing (NDT) technologies used to help evaluate the final shape and
dimensions of defects for both part-penetrating and through-wall defects were detailed.
Conventional NDT or leak testing procedures were presented for the different type of
defects. To assess their criticality, fracture mechanics parameters defined to characterise
the stress field at proximity of the defects were shown. They provide a link between the
local crack tip response (stress/strain) to the global conditions (loads, material properties,
geometry). Fitness-For-Service (FFS) practices are used to evaluate the defects tolerance
/ acceptance either through an Engineering Critical Assessment (ECA) or other meth-
ods such that rational decisions about the remaining life, whether to repair, replace a
component or reduce operating conditions could be made.
Chapter 3 covers an extensive literature review on the concept of leak-before-break.

The different approaches, including basic formulations and available established proce-
dures, their differences and limitations were elaborated for nuclear and other industries.
The margin between the smallest detectable leak size and the critical crack size must
be adequate and reasonable to support LBB. Reliable leak detection methods must be
employed to ensure the ultimate success of the technique in order to prevent catastrophic
failure. Numerous situations and types of component studied have been looked at. It
has been noted that LBB is generally applied to pressurized components containing be-
nign fluids such as water, steam and carbon dioxide. Only a few published addressed
such applications in non-nuclear industries. However, LBB could be further developed
in some areas of these industries. When applied to non-nuclear industries, case stud-
ies did not follow strictly the established LBB procedures such as recommended in API
579-1/ ASME FFS-1 or BS 7910. Various sources were generally used, which could lead
to non-consistent results. High margins were found in screening procedures and detailed
assessment procedures lacked solutions/guidance.
Chapter 4 presents a detailed comparison of LBB parameters as described in established

procedures for leakage and crack stability assessment. The COA results, were provided
for inner & outer surfaces or mid-thickness depending on the solution. All solutions
investigated provided coherent results until half of the limit of applicability for both axial
and circumferential cracks. It should be noted that BS 7910 solution were developed for
flat plates and corrected for cylindrical shapes. On the other hand, API 579-1/ASME
FFS-1 have been developed for cylinders by FEA. For both geometries, the stress-based
solution from API 579-1/ASME FFS-1 gave greater values of COA than the pressure-
based equation. For axial cracks, BS 7910 and Zahoor solutions tend to give greater values
of COA at mid-thickness (similar to API 579-1/ ASME FFS-1 at the outer surface). For
circumferential cracks, the handbook solutions from BS 7910 and Paris-Tada produced
the lowest COA values when compared with the API 579-1/ ASME FFS-1 (for both inner
and outer surfaces). BS 7910 and Zahoor models include correction factors for small-scale
yielding while API 579-1/ ASME FFS-1 propose a correction factor based on the plastic
collapse assessment. This results in a wider range of application for the latter. Recalling
that COA solutions are used to assess the leak rate which is a critical parameter in a
LBB assessment, an over-estimation of the COA can lead to non-conservative result and
vice-versa an under-estimation lead to a very conservative result. Only elastic models,
known to under-estimate the COA values when assessing ductile materials, were presented
for comparison purpose. In terms of crack stability, different solutions provided for FAD,
SIF and RS have been evaluated. Four relevant geometries for LBB assessment were
chosen: (1a) through-wall axial, (1b) circumferential, (2a) part-penetrating axial and
(2b) circumferential. For a given axial defect, BS 7910 tends to produce greater values in
terms of reference stress for both through-thickness and surface defects. The multiplier
of 1.2 introduced in the reference stress solutions, as such to achieve a certain level of
conservatism partially contributed to this observation. The other factor leading to another
significant difference is the correction factor employed. Validation of the reference stress
solutions for axial surface defects using available experimental data show that API 579-
1/ASME FFS-1 is found to successfully distinguish between leak and break cases. On the
other hand, BS 7910 approach is more conservative, treating leak as failure or unacceptable
result. This is also observed without the inclusion of factor 1.2. For circumferential
defects, when only membrane stress loading is considered, both standards produce similar
values in terms of reference stresses. However, when a bending stress loading is applied,
BS 7910 solutions tend to give higher reference stresses. It is worth noting that reference
stress solutions recommended in BS 7910 are mainly from local limit load solutions which
are more conservative. In leak before break assessments, since that the failure mode is
pre-supposed to be not dominated by brittle fracture, the criteria for break will be based
mainly on the reference stress solution. Therefore, BS 7910 is likely to be significantly
more conservative than API 579-1/ASME FFS-1 for its break condition.
Chapter 5 validated the solutions for through-wall cracks in plates and cylindrical
components. The calculation of various solutions has been detailed for plates, with step-
by-step methodologies and post-processing to generate models and results. In order to
generate numerous models in a time-efficient manner, different plug-ins have been created
using Python script. Cracks have been introduced into un-cracked bodies (plate or cylin-
der) in the assembly step, by inserting a crack-box. This permitted to generate the models
changing only the desired parameters (tractable approach). FE results in this work were
in good agreement with solutions available in procedures for idealized shape models. The
effect of crack idealization on SIF and COA solutions has been investigated using detailed
3D elastic FE models of axial through-wall cracks in a cylindrical component under inter-
nal pressure. Idealization of a through wall crack with its largest dimension tends to be
conservative when it comes to deriving SIF. However, in a LBB assessment, both crack
stability and leakage have to be evaluated. Crack idealization rapidly increases COA
estimation leading to an over-estimation of detectable leak rate. This implies that for a
given crack size, its idealisation will be conservative in terms of SIF but the estimated
resulting leak rate over-estimated. The leak detection can therefore be distorted. Models
and analyses employed in this work were validated against existing solutions for SIF and
COA given in BS 7910 and API 579-1/ASME FFS-1. Different reference crack lengths
(inner surface) have been selected in order to cover a range of crack sizes. The observa-
tions on SIF and COA for various crack geometries can be used to carry out sensitivity
studies while assessing LBB cases. The effect of non-idealized geometry on both SIF and
COA could be included to assess cracks ‘yet’ to become idealized using the observations
presented in this work. The analysis showed that a non-idealized crack at break-through
(from surface to through-wall) will grow and finally evolve into an ideal shape. LBB
assessment in this transition region is generally limited due to a lack of existing solutions.
The effect of crack shape on SIF solutions can be used in fatigue assessment to examine
cracks growing under cyclic loading. The modification on COA solutions (taking a more
realistic shape into account) can improve the leak rate calculation more accurately. This
is an important aspect of the assessment often overlooked, while the variation is important
in the transition region. Further work to investigate more realistic crack shapes (i.e. non-
straight crack front) and development of correction factors for existing solutions would be
beneficial in establishing a more refined assessment of LBB.
In Chapter 6, the experimental work undertaken on surface notched plate specimens

was presented. As a semi-qualitative test, fatigue testing on Perspex plate specimens with
surface cracks was carried out. Different observations were made from the examination
of fracture surface. If the sample was not loaded perfectly perpendicular to the applied
load (non-uniform distribution of membrane stress), fatigue crack growth showed a non-
symmetric growth pattern. This led to rapid extension (length-wise) of the crack size on
one side. It propagated rapidly to the specimen edge and failed prematurely. Thinner
samples experienced bending during the tests (not enough remaining cross section to
sustain the load). This resulted in an increase in crack length on surface and crack
arrest in the thickness direction. On the other hand, thicker samples allowed the crack
to grow through thickness. They achieved the expected crack shape evolution. Once
the crack had broken through, a rapid crack growth in the width direction was observed
until the crack shape reached an “idealised” through-wall shape. Tests were duplicated
with metallic material to study the resulting stress/strain on the back surface of a deeply
notched plate subject to cyclic tensile loading. Experiments consisted of four high strength
steel flat plates with surface-breaking notches to a depth about 80% thickness. Two
different crack geometries were examined to represent short and long cracks. Strain gauges
were attached on to the back surface of the plates and strain measured during crack
growth. Break-through was estimated to occur when recorded strains where between
yield and ultimate strain of the material. Observations were made on different axes (i.e.
parallel/perpendicular/diagonal to the crack).
Chapter 7 presented different FEA models with semi-elliptical cracks in a plate. Focus
has been on the stress field developing on two axes on the back surface of the plate: (i)
along the crack (X-axis) and (ii) perpendicular to the crack (Y-axis). A preliminary study
has been carried out to study the effect of blunt notch radius on the back surface stress
field. Results from a sharp crack were compared with those of blunt notched cracks using
three different notch radii, in order to determine the effect of crack-tip / notch geometry.
It has been shown that with the greater notch radii, the greater stress field on back
surface. The effect of notch geometry was considered negligible. Sharp crack geometry
has been used for subsequent investigations. To warrant confidence in the models used,
the FE results have been validated with experimental test values. FEA results agreed well
with experimental results for both axes and geometries investigated. For both geometries,
various models have been generated to study the effect of the crack depth on the back
surface stresses. The main observation was made on X-axis, parallel to the crack as well
as Y-axis which is perpendicular to the crack. Along the X-axis, stresses on back surface
were found to be maximum at the deepest point of the crack (i.e X=0). Von Mises
stresses derived from FEA were close to the far field values for a crack depth less than
60%. They started to increase when the crack grew deeper and this was more rapid in a
long crack (Geometry 2). Along the Y-axis, stresses exhibited a peak value at a position
which varied with the crack depth. In other words, as the crack grew deeper into the
section, the location at which peak stress occurred also moved closer to the origin along
the Y-axis. Similar trends in the position of the maximum stresses were observed in both
geometries. Same observations can be made in terms of the maximum value of the stresses
as for X-axis.
In summary, it was found high margins in screening methods and a lack of solu-
tions/guidance for detailed assessment methods in existing procedures. Depending on
the procedure employed, some parameters (i.e. reference stress) provide different values.
Overall, BS 7910 solutions were found more conservative, treating leak as failure or unac-
ceptable result. The effect of non-idealized geometry should be included to assess cracks
in the transition region. The effect on SIF and COA solutions can be used to improve
crack stability assessment and resulting leak rate estimation. Further work to investigate
more realistic crack shapes (i.e. non-straight crack front) and development of correction
factors for existing solutions would be beneficial in establishing a more refined assessment
of LBB. To counter over-conservatism in the estimate of LBB parameters, non-idealised
flaw at breakthrough needs to be considered to take into account the evolution/transition
of part-penetrating to through-thickness. It is recommended that refinement to exist-
ing LBB procedure should include the analysis of non-idealised flaw at break-through as
work presented in this thesis has demonstrated that the evolution/transition of flaw from
part-penetrating to through-thickness was significant. It is also beneficial to perform both
experimental and numerical analysis to gain some insight of the effect of aspect ratio and
to estimate the possible level of back surface stress/strain.
8.3 Main PhD Achievements Findings
The previous section presented the observations and conclusions achieved from the work
carried out in for this research. A list of the main conclusions and findings is given:
• The existing solutions, available to industry, to carry out a LBB assessment were
reviewed and discussed. Reference stress solutions showed important difference. BS
7910 solutions were found more conservative, treating leak as failure or unacceptable
result, due to different aspects. API 579-1/ASME FFS-1 on the other hand offered
more options (i.e. account of crack-face pressure or stress/pressure based solutions).
Further work on parameters should focus on reference stress solutions.
• Shape idealisation for through-wall crack yet to become idealised did not always
reflect the actual behaviour, especially for COA calculation. This effect should be
included to assess cracks in the transition region. FEA can be used as a good
predictive tool to investigate more realistic crack shapes (i.e. non-straight crack
front) and develop correction factors for existing solutions. This would be beneficial
in establishing a more refined assessment of LBB.
• The experiment using metallic specimens showed that high stress/strain on back
surface offer an accurate monitoring mechanism to determine breakthrough and
transition behaviour. Strain-mapping devices would help for a better understanding
of the transition mechanisms.
8.4 Recommendations for future work
The work discussed in this thesis addressed different aspects. There are several lines
of research arising from this work which should be pursued.
An extensive literature review on the concept of leak-before-break concept was drawn

in Chapter 3. It was shown that the concept of LBB could be further developed in
some areas of non-nuclear industries. When applied to these industries, case studies did
not appear to follow established LBB procedures such as those presented in API 579-1/
ASME FFS-1 or BS 7910. A mix of different sources was generally used, which could lead
to non-consistent results. A refined procedure giving more recommendation and detailed
would be beneficial.A simplified procedure for screening use and a detailed procedure with
guidance on the transition stage would make assessments more accessible.
In Chapter 4, the differences in the solutions related to LBB have been highlighted.
It was shown difference from the comparison of the reference stress solutions, especially
for pipes/cylinders containing axial flaws. The effect of different factors should be re-
evaluated in order to enhance the assessment capability. Validation of the solutions against
other data from industry or research papers that include leak and break analysis would
be beneficial in establishing a more accurate assessment of LBB. Data presented in this
thesis were limited to axial surface cracks, however data for circumferential cracks were
found in the literature and other defects geometry should be evaluated. Further work to
assess the effect of these difference on a full LBB case would be advantageous as well.
Simplified procedures considering through-thickness crack are known to be conservative.
They should be used as a screening criteria. The conservatism resulting from the safety
margins employed in flaw size should be evaluated for a wide range of conditions to reduce
discrepancy. Current methods to perform a detailed assessment are lacking of solutions
in terms of non-idealised flaw geometry. Further numerical studies or experiments should
be performed for crack propagation to assess non-idealised cracks. Two stages would be
of interest: (i) close to the ideal shape to evaluate the size uncertainty and (ii) at the
transition between part-penetrating to through thickness to assess the early stages. This
would help to achieve a better estimate of LBB parameters. Another point worth noting
is that flaw size should be designed in line with the critical flaw size determined using
ECA.
The effect of idealisation on axial through-wall crack in cylindrical components has

been highlighted in Chapter 5. The numerical models presented some limitations in terms
of component geometry, loading and crack geometry. Further models could be developed
using the techniques presented to increase the applicability range and develop a correction
factor that would permit to correct the current existing solution. It is worth noting that
non-idealised crack models assumed straight crack front for simplicity. More realistic crack
shape modelling based on experiment or experience would increase the accuracy of the
corrections. This would affect the safety margins applied in LBB, especially on the leak
side, with the calculation of crack opening area. All models were developed using elastic
material properties. However it is known that elastic-plastic consideration tends to lower
COA. Models with non-idealised cracks using elastic-plastic material properties would be
valuable. Other factors, such as temperature gradient through the wall, are known to have
an effect on the COA. Investigation of theses parameters would be valuable for a better
accuracy in the developed formulations. Study on the crack idealisation in this thesis was
limited to axial cracks considering internal pressure loading. Another aspect that could
be investigated would be a coupled mechanical and fluidic using CFD tools. This would
permit to investigate the direct effect on leak-rate from non-idealised defects.
Experimental work in Chapter 6 described the strain variation on the back surface of
a cracked plate subjected to tensile loading. Transition from surface to through-thickness
crack was observed with the recorded values. However due to the distance between strain
gauges, only ”local” evolution was analysed. Use of more sophisticated technology such
as digital image correlation (DIC) for the analysis of the displacement fields could be used
to improve the understanding on the back surface behaviour. This would allow to have
a global understanding of the strain field variation on the back surface. The strain field
presented in the thesis is for low loading to assume linear elastic behaviour. Based on
similar reasoning, future research work is needed to investigate the strain field for strain
hardening materials. As similar breakthrough behaviour was observed on both Perspex
and metallic specimens, further study on Perspex specimens would permit to have further
visual observations with live record.
Numerical simulations presented in Chapter 7 were performed to duplicate experimen-

tal geometries. Only two crack aspect ratios were considered. A wider range of crack
aspect ratios would permit a mapping of its effect on back surface strain field. Higher
level of loading and elastic-plastic material properties would also permit a better under-
standing.
Bibliography
[1] BSI. BS 7910:2013 - Guide to methods for assessing the acceptability of flaws in
metallic structures . British Standards Institution, London, 2013.
[2] IAEA. Training guidelines in non-destructive testing techniques: Leak testing

at level 2. Technical report, IAEA, 2012. IAEA Training Course Series No. 52,
IAEA/TCS/52, ISSN 1018–5518, Vienna.
[3] G. Koch, M. Brongers, N. Thompson, Y. Virmani, and J. Payer. Corrosion costs and
preventive strategies in the united states, 2002. Publication No FHWA-RD-01-156.
[4] M. Nuhi, T. Abu Seer, A.M. Al Tamimi, M. Modarres, and A. Seibi. Reliability
analysis for degradation effects of pitting corrosion in carbon steel pipes. Procedia
Engineering, 10:1930 – 1935, 2011. 11th International Conference on the Mechanical
Behavior of Materials (ICM11).
[5] A. Rajabipour and R.E. Melchers. A numerical study of damage caused by combined
pitting corrosion and axial stress in steel pipes. Corrosion Science, 76:292 – 301,
2013.
[6] J. Lee, S. Han, K. Kim, H. Kim, and U. Lee. Failure analysis of carbon steel pipes
used for underground condensate pipeline in the power station. Engineering Failure
Analysis, 34:300 – 307, 2013.
[7] G. Yang, K.B. Yoon, and Y.C. Moon. Stress corrosion cracking of stainless steel
pipes for methyl-methacrylate process plants. Engineering Failure Analysis, 29:45
– 55, 2013.
[8] K.A. Macdonald, A. Cosham, C.R. Alexander, and P. Hopkins. Assessing mechan-
ical damage in offshore pipelines – two case studies. Engineering Failure Analysis,
14(8):1667 – 1679, 2007. Papers presented at the Second International Conference
on Engineering Failure Analysis (Toronto, Canada, 12–15 September 2006) Part
{II}.
[9] API. API 579-1/ASME FFS-1 2007 Fitness-For-Service . American Society of

Mechanical Engineers, New-York, 2007.
[10] T.L. Anderson and D.A. Osage. API 579: A comprehensive fitness-for-service guide
. International Journal of Pressure Vessels and Piping, 77(14–15):953 – 963, 2000.
[11] M. Willcox and G. Downes. A Brief Description of NDT Techniques .

Technical report, Insight NDT Equipment Ltd. Technical paper available at
“http://www.insightndt.com”, copyright Insight
c NDT.
199
BIBLIOGRAPHY 200
[12] BSI. PD 5500:2012 - Unfired fusion welded pressure vessels . British Standards
Institution, London, 2012. ISBN: 9780580849756.
[13] ASME. ASME Boiler & Pressure Vessel Code, ”Section V - Nondestructive Exam-
ination” . American Society of Mechanical Engineers, 2010.
[14] ASTM. E165/E165M-12 - Standard Practice for Liquid Penetrant Examination

for General Industry . American Society for Testing and Materials, ASTM Interna-
tional, West Conshohocken, PA, 2012.
[15] ASTM. E433-71(2013) - Standard Reference Photographs for Liquid Penetrant

Inspection . American Society for Testing and Materials, ASTM International,
West Conshohocken, PA, 2013.
[16] ASTM. E709-14 - Standard Guide for Magnetic Particle Testing . American Society
for Testing and Materials, ASTM International, West Conshohocken, PA, 2014.
[17] ASTM. E1444/E1444M-12 - Standard Practice for Magnetic Particle Testing

. American Society for Testing and Materials, ASTM International, West Con-
shohocken, PA, 2012.
[18] ASTM. E213-14 - Standard Practice for Ultrasonic Testing of Metal Pipe and
Tubing . American Society for Testing and Materials, ASTM International, West
Conshohocken, PA, 2014.
[19] ASTM. E273-10 - Standard Practice for Ultrasonic Testing of the Weld Zone of
Welded Pipe and Tubing . American Society for Testing and Materials, ASTM
International, West Conshohocken, PA, 2010.
[20] ASTM. E94-04(2010) - Standard Guide for Radiographic Examination . American

Society for Testing and Materials, ASTM International, West Conshohocken, PA,
2010.
[21] ASTM. E432-91(2011) - Standard Guide for Selection of a Leak Testing Method
[22] BSI. BS EN 1779:1999 - Non-destructive testing. Leak testing. Criteria for method
and technique selection . British Standards Institution, London, 1999. ISBN:
0580329313.
[23] Report. Leak Detection Theory and Practise - Comparison among leak test-
ing techniques. Technical report, VTech, 2006. VTech Report, Available at
“www.vtechonline.com/pdf/vtech-leak-detection.pdf.
[24] A. Pregelj and M. Mozetic. Leak detection methods and defining the sizes of leaks .
In The 4th International Conference of Slovenian Society for Nondestructive Testing
”Application of ContemporaryNondestructive Testing in Engineering”, Ljubljana, 24
- 25 April 1997.
[25] ASTM. E1002-11 - Standard Practice for Leaks Using Ultrasonics . American
2011.
BIBLIOGRAPHY 201
[26] ASTM. E1211/E1211M-12 - Standard Practice for Leak Detection and Location
Using Surface-Mounted Acoustic Emission Sensors . American Society for Testing
and Materials, ASTM International, West Conshohocken, PA, 2012.
[27] ASTM. E1139/E1139M-12 - Standard Practice for Continuous Monitoring of

Acoustic Emission from Metal Pressure Boundaries . American Society for Testing
[28] ASTM. E569/E569M-13 - Standard Practice for Acoustic Emission Monitoring

of Structures during Controlled Stimulation . American Society for Testing and
Materials, ASTM International, West Conshohocken, PA, 2013.
[29] R. Rintamaa, H. Keinanen, K. Törrönen, H. Talja, A. Saarenheimo, and K. Ikonen.

Fracture behaviour of large scale pressure vessels in the hydrotest. International
Journal of Pressure Vessels and Piping, 34(1–5):265 – 291, 1988.
[30] BSI. BS EN 13184:2001 - Non-destructive testing. Leak testing. Pressure change

method . British Standards Institution, London, 2001. ISBN: 0580372170.
[31] ASTM. E515-11 - Standard Practice for Leaks Using Bubble Emission Techniques
[32] BSI. BS EN 1593:1999 - Non-destructive testing. Leak testing. Bubble emission

techniques . British Standards Institution, London, 1999. ISBN: 0580351661.
[33] ASTM. E1603/E1603M-11 - Standard Practice for Leakage Measurement Using

the Mass Spectrometer Leak Detector or Residual Gas Analyzer in the Hood Mode
[34] ASTM. E499/E499M-11 - Standard Practice for Leaks Using the Mass Spectrom-
eter Leak Detector in the Detector Probe Mode . American Society for Testing and
Materials, ASTM International, West Conshohocken, PA, 2011.
eter Leak Detector or Residual Gas Analyzer in the Tracer Probe Mode . American
2011.
eter Leak Detector in the Inside-Out Testing Mode . American Society for Testing
[37] BSI. BS EN 13185:2001 - Non-destructive testing. Leak testing. Tracer gas method
. British Standards Institution, London, 2001. ISBN: 0580372162.
[38] IAEA. Radiotracer Technology as Applied to Industry . Technical report, IAEA,

2001. IAEA-TECDOC-1262, ISSN 1011–4289, Vienna.
[39] IAEA. Leak Detection in Heat Exchangers and Underground Pipelines Using
Radiotracers . Technical report, IAEA, 2009. IAEA Training Course Series No. 38,
IAEA-TCS-38, ISSN 1018-5518, Vienna.
BIBLIOGRAPHY 202
[40] C. Inglis. Stresses in a plate due to the presence of cracks and sharp corners. Trans.
Inst. Nav. Archit., 55:219 – 230, 1913.
[41] X.K. Zhu and J.A. Joyce. Review of fracture toughness (G, K, J, CTOD, CTOA)
testing and standardization . Engineering Fracture Mechanics, 85:1 – 46, 2012.
[42] P. Gilles and F.W. Brust. Approximate fracture methods for pipes — part I: theory
. Nuclear Engineering and Design, 127(1):1 – 17, 1991.
[43] A. Griffith. The phenomena of rupture and flow in solids. Philos. Trans. R. Soc. A
Math. Phys. Eng. Sci., 221:163–198, 1921.
[44] G.R. Irwin. Analysis of Stresses and Strains Near the End of a Crack Traversing a
Plate. Journal of Applied Mechanics, 24:361 – 364, 1957.
[45] T. Anderson. Fracture Mechanics: Fundamentals and Applications. CRC Press,

third edition. edition, 2005. ISBN 9780849316562.
[46] H. Westergaard. Bearing pressures and cracks . Journal of Applied Mechanics,

61:A49–A53, 1939.
[47] G. Irwin. Plastic zone near a crack and fracture toughness . In 7th Sagamore
Conference, 1961. Volume 4, pp 63–76.
[48] D. Dugdale. Yielding of steel sheets containing slits . Journal of the Mechanics
and Physics of Solids, 8(2):100–104, 1960.
[49] J. Begley and J. Landes. The J integral as a fracture criterion. ASTM, 1972. Am.
Soc. Test. Mater. Spec. Tech. Publ. ASTM STP 514, pp 1-20.
[50] J. Rice. A Path Independent Integral and the Approximate Analysis of Strain
Concentration by Notches and Cracks . Journal of Applied Mechanics, 35:379–386,
1968.
[51] J. Rice, P. Paris, , and J. Merkle. Some further results of J-integral analysis and
estimates. ASTM, 1973. Am. Soc. Test. Mater. Spec. Tech. Publ. ASTM STP 536,
pp 231–245.
[52] J. Sumpter and C. Turner. Method for laboratory determination of Jc. ASTM, 1976.
Am. Soc. Test. Mater. Spec. Tech. Publ. ASTM STP 601, pp 3–18.
[53] A. Wells. Application of fracture mechanics at and beyond general yielding. British
Welding Journal, 10(5):63–70, 1963.
[54] TWI. Fracture Toughness Testing at TWI. Technical report, TWI, 2015. Internal
training course notes.
[55] J. Rice and G. Rosengren. Plane strain deformation near a crack tip in a power law
hardening materials. Journal of the Mechanics and Physics of Solids, 16(1):1–12,
1968.
[56] D. Hellmann and K.-H. Schwalbe. On the experimental determination of ctod based
r-curves. In K.H. Schwalbe, editor, The Crack Tip Opening Displacement in Elastic-
Plastic Fracture Mechanics, pages 115–132. Springer Berlin Heidelberg, 1986.
BIBLIOGRAPHY 203
[57] C. Filiou, N. Taylor, P. Lejuste, and R. Houghton. Survey of current application and
future requirements for european fitness-for-service technology. Technical report,
Technical Report No. FITNET/TR2/03, 2003.
[58] R. Selva. Living with Defects:Replace/Repair or prove Fitness-For-Service (FFS)?

. In 13th International Conference on Pressure Vessel and Piping Technology, Lon-
don, 20-23 May 2012. Keynote Paper – Technical Session: Replacement and Fitness
For Service.
[59] M. Koçak, S. Webster, J. Janosch, R. Ainsworth, and R. Koers. FITNET Fitness-

for-Service (FFS) - Procedure (Volume 1) . British Standards Institution, London,
2008. ISBN 978-3-940923-00-4, printed by GKSS Research Centre, Geesthacht.
[60] CEGB. R6 - Assessment of the integrity of structures containing defects . Central

Electricity Generating Board, 2007. CEGB Document No. R/H/R6-Rev. 4.
[61] J. Shigley, C. Mischke, and R. Budynas. Mechanical Engineering Design. McGraw-

Hill, 7th edition edition, 2004. ISBN 9780072520361.
[62] ASME. ASME Boiler & Pressure Vessel Code, ”Section VIII - Rules for Con-
struction of Pressure Vessels (Division 1 and Division 2)” . American Society of
Mechanical Engineers, 2010.
[63] BSI. EN 13445-3:2014 - Unfired pressure vessels. Design” . British Standards

Institution, London, 2010. ISBN: 9780580868030.
[64] European Commission. Comparative Study on Pressure Equipment Standards.

Technical report, 2004. Contract No FIF.20030114.
[65] G. Irwin. Fracture of pressure vessels. Mater. Missiles Spacecraft, page 204–229,
1963. McGraw-Hill.
[66] G.S. Holman. Double-ended breaks in reactor primary piping. Nuclear Engineering
and Design, 89(1):1 – 12, 1985.
[67] USNRC. Safety evaluation of Westinghouse topical reports dealing with the elim-
ination of postulated pipe breaks in PWR primary main loop. Technical report,
USNRC, 1984. (Generic Letter 84-04).
[68] USNRC. State-of-the-Art Report on Piping Fracture Mechanics. Technical report,

USNRC, 1998. NUREG/CR-6540.
[69] IAEA. Applicability of the Leak-Before-Break Concept . Technical report, IAEA,

1993. IAEA-TECDOC-710, Vienna.
[70] USNRC. Standard Review Plan 3.6.3 Rev.1 - Leak-Before-Break Evaluation Pro-
cedures. Technical report, USNRC, 2007. NUREG-0800.
[71] USNRC. Report of the U.S. Nuclear Regulatory Commission Piping Review Com-
mittee - Evaluation of Potential for Pipe Breaks. Technical report, USNRC, 1984.
NUREG-1061.
BIBLIOGRAPHY 204
[72] USNRC. NRC Leak-Before-Break (LBB.NRC) Analysis Method for Circumferen-

tially Through-Wall Cracked Pipes Under Axial Plus Bending Loads. Technical
report, USNRC, 1986. NUREG/CR-4572.
[73] European Commission. European Safety Practices on the application of Leak Be-
fore Break (LBB) concept. Technical report, European Commission, 2000. Report
EUR18549 EN.
[74] S. Chattopadhyay. Material Selection for a Pressure Vessel . In Materials in Design

and Manufacturing, Conference & Exposition, ASEE, 2008.
[75] A. Blake. Practical Fracture Mechanics in Design. CRC Press, 1996.
[76] U. Zerbst and R. Ainsworth. Experiences with some European flaw assessment
procedures . In Proceedings of the 11th International Congress on Fracture, 2005.
[77] B.J.L. Darlaston. Some aspects of leak before break; their quantification and appli-
cation. Nuclear Engineering and Design, 84(2):225 – 232, 1985.
[78] D. Rieck and R. Wernicke. Experiences with fracture mechanics Leak-Before-Break

evaluations within the German Integrity Concept for NPPs . In Proceedings of the
18th European Conference on Fracture, 2010.
[79] B.F. Beaudoin, D.F. Quiñones, and T.C. Hardin. Leak-before-break application in
US light water reactor balance-of-plant piping . International Journal of Pressure
Vessels and Piping, 43(1–3):67 – 83, 1990.
[80] L. Xie. The effect of characteristic crack sizes on the leak-before-break case of
pressure vessels and piping with multiple cracks. International Journal of Pressure
Vessels & Piping, 76(7):435 – 439, 1999.
[81] L. Xie. Multi-crack growth/coalescence simulation and its role in passive component
leak-before-break concept. Nuclear Engineering and Design, 194(2–3):113 – 122,
1999.
[82] European Commission. 30 Years of NRWG activities towards harmonization of

nuclear safety criteria and requirements. Technical report, European Commission,
2002. Report EUR20818 EN.
[83] IAEA. Guidance for the application of the Leak-Before-Break Concept . Technical
report, IAEA, 1994. IAEA-TECDOC-774, Vienna.
[84] C. Faidy. Leak before break application in French PWR plants under operation
. In Proceedings of the seminar on leak before break in reactor piping and vessels,
Lyon, 1995. pp. 145–150.
[85] C. Faidy, S. Bhandari, and P. Jamet. Leak-before-break in French nuclear power

plants . International Journal of Pressure Vessels and Piping, 43(1–3):151 – 163,
1990.
[86] R. Gerard, C. Malekian, and O. Meessen. Belgian experience in applying the ‘leak-
before-break’ concept to the primary loop piping . In Proceedings of the seminar on
leak before break in reactor piping and vessels, Lyon, 1995. pp. 137–144.
BIBLIOGRAPHY 205
[87] M.L. Aggarwal, M.J. Kozluk, T.C. Lin, B.W. Manning, and D.K. Vijay. A leak-
before-break strategy for CANDU primary piping systems . International Journal
of Pressure Vessels and Piping, 25(1–4):239 – 256, 1986.
[88] J.S. Nathwani and J.D. Stebbing. Ontario hydro’s leak-before-break approach to
Darlington NGS heat transport system piping . International Journal of Pressure
Vessels and Piping, 43(1–3):113 – 127, 1990.
[89] Y.W. Park and Y.K. Chung. Leak-before-break assessment of CANDU pres-
sure tube considering leak detection capability . Nuclear Engineering and Design,
191(2):205 – 216, 1999.
[90] J. Chattopadhyay, B.K. Dutta, and H.S. Kushwaha. Leak-before-break qualifica-

tion of primary heat transport piping of 500 MWe Tarapur atomic power plant .
International Journal of Pressure Vessels and Piping, 76(4):221 – 243, 1999.
[91] J. Zdarek, L. Pecinka, and P. Kadecka. Leak-before-break criterion applied to

VVER 440/230 unit . Theoretical and Applied Fracture Mechanics, 23(2):117 – 123,
1995.
[92] E. Keim and D. Lidbury. Review of assessment methods used in nuclear

plant life management. Technical report, VTT, 2012. Availableathttp :
//www.vtt.f i/proj/nulif e/nulif ed 1d 2r eportm aster1 1may2012.pdf .
[93] USNRC. Development of Technical Basis for Leak-Before-Break Evaluation Proce-

dures. Technical report, USNRC, 2002. NUREG/CR-6765.
[94] J. Sharples. STYLE: Comparison of Leak-Before-Break Methodologies Applied in

Europe . In Proceedings of the ASME 2012 Pressure Vessels & Piping Conference,
Toronto, 15–19 July 2012. Paper No. PVP2012-78303, pp. 763–768.
[95] Y. Takahashi. Leak-Before-Break Assessment in JSME Standard . In Proceedings

of the ASME 2011 Pressure Vessels & Piping Conference, Baltimore, 17–21 July
2011. Paper No. PVP2011-57978, pp. 147–151.
[96] N. Drozdziel. Leak-before-break characterization and demonstration of a high-

pressurant helium storage vessel . In Proceedings of the 5th International Congress
on Fracture, 1981. pp.2145–2151.
[97] VV Knysh, AY Barvinko, YP Barvinko, and AN Yashnik. Substantiation of Leak-

before-Break criterion for vertical cylindrical tanks for oil storage. The Paton
Welding Journal, #9:26–29, 2012.
[98] M.R. Hackworth and J.M. Henshaw. A pressure vessel fracture mechanics study
of the aluminium beverage can. Engineering Fracture Mechanics, 65(5):525 – 539,
2000.
[99] M. Walter and R. Lesicki. Measures taken to ensure safe operation of an ammonia
storage tank. Process Saf. Prog., 17(4):288–296, 1998.
[100] K.P. Kou and F.M. Burdekin. The applicability of leak-before-break on flooded
member detection for offshore structures under fatigue. Engineering Fracture Me-
chanics, 75(1):31 – 40, 2008.
BIBLIOGRAPHY 206
[101] R. Andrews, N. Millwood, S. Tiku, N. Pussegoda, M. Hoekstra, and S. Smith.

Analysis and Testing of a 13Cr Pipeline to Demonstrate ‘Leak Before Break . In
Proceedings of the 9th International Pipeline Conference, 2012. pp.83–91.
[102] C. Bonn. Estimation of Crack Opening Displacement and Leak Rates Through a
Through-Flaw in Aerospace Pressure Vessels. Technical report, Rensselaer Poly-
technic Institute, 2008. Master Thesis.
[103] Y. Takahashi. Evaluation of leak-before-break assessment methodology for pipes
with a circumferential through-wall crack. part iii: estimation of crack opening
area. International Journal of Pressure Vessels and Piping, 79(7):525 – 536, 2002.
[104] C. Wüthrich. Crack opening areas in pressure vessels and pipes. Engineering Frac-
ture Mechanics, 18(5):1049 – 1057, 1983.
[105] P. Chellapandi, R. Srinivasan, A. Biswas, S. Chetal, and S. Bhoje. Leak Before
Break Investigation on Sodium Piping for Prototype Fast Breeder Reactor . In
Transactions of the 17th International Conference on Structural Mechanics in Re-
actor Technology (SMiRT 17), Prague, 17-22 August 2003. Division G10-7.
[106] P. Nagapadmaja, V. Kalyanaraman, S.R. Satish Kumar, and P. Chellapandi. Ex-
perimental study on LBB behaviour of LMFBR pipe elbows . International Journal
of Fatigue, 30(3):574 – 584, 2008.
[107] S. Bhandari, M. Fortmann, L. Grueter, J. Heliot, P. Meyer, B. Percie Du Sert,
A. Prado, and H. Zeibig. Crack propagation in a LMFBR elbow . Nuclear Engi-
neering and Design, 91(2):107 – 119, 1986.
[108] N. Mani, G. Thanigaiyarasu, and P. Chellapandi. Static stress analysis of steam
generator shell nozzle junction for leak before break analysis . Int. J. Des. Manuf.
Technol., 3(1):45–50, 2009.
[109] V. Ukadgaonker, Y. Khairnar, P. Vaidya, and P. Chellapandi. Leak-before-break
analysis of shell-nozzle junction of steam generator . Indian J. Eng. Mater. Sci.,
9:237–249, 2002.
[110] S. Burande and Raju Sethuraman. Computational simulation of fatigue crack
growth and demonstration of leak before break criterion. International Journal
of Pressure Vessels and Piping, 76(5):331 – 338, 1999.
[111] B. Brickstad and I. Sattari-Far. Crack shape developments for LBB applications .
Engineering Fracture Mechanics, 67(6):625 – 646, 2000.
[112] E. Roos, K.H. Herter, P. Julisch, G. Bartholomé, and G. Senski. Assessment of
large scale pipe tests by fracture mechanics approximation procedures with regard
to leak-before-break. Nuclear Engineering and Design, 112:183 – 195, 1989.
[113] R. Lukess. Predicting the Crack Response for a Pipe with a Complex Crack. Tech-
nical report, University of South Carolina, 2013. PhD Thesis.
[114] S. Yellowlees, K. Lynch, A. Warren, J. Sharples, and P. Budden. Development of
Guidance for Leak-Before-Break Analysis for Complex Shaped Cracks . In Proceed-
ings of the ASME 2005 Pressure Vessels & Piping Conference, Denver, 17-21 July
2005. Paper No. PVP2005-71382, pp. 361–370.
BIBLIOGRAPHY 207
[115] S.B.M. Beck, N.M. Bagshaw, and J.R. Yates. Explicit equations for leak rates
through narrow cracks. International Journal of Pressure Vessels and Piping,
82(7):565 – 570, 2005.
[116] D.L. Rudland, G. Wilkowski, and P. Scott. Effects of crack morphology parameters
on leak-rate calculations in {LBB} evaluations. International Journal of Pressure
Vessels and Piping, 79(2):99 – 102, 2002.
[117] T.C. Chivers. The influence of surface roughness on fluid flow through cracks.
Fatigue & Fracture of Engineering Materials & Structures, 25(11):1095–1102, 2002.
[118] P. Gill. Investigating leak rates for ‘leak-before-break’ assessments. Technical report,
University of Manchester, 2013. EngD Thesis.
[119] P. Gill and K. Davey. Analysis of thermo-mechanical behaviour of a crack using

{XFEM} for leak-before-break assessments. International Journal of Solids and
Structures, 51(11–12):2062 – 2072, 2014.
[120] Y. Takahashi. 7.10 - leak before break. In R.O. Ritchie B. Karihaloo, I. Milne,
editor, Comprehensive Structural Integrity, pages 427 – 464. Pergamon, Oxford,
2003.
[121] S. Yellowlees D. Hootonand J. Sharples P. Budden and D. Dean. Crack Opening

Areas During High Temperature Operation . In Proceedings of the ASME 2002
Pressure Vessels & Piping Conference, Vancouver, 5-9 August 2002. Paper No.
PVP2002-1192, pp. 25–31.
[122] J. Sharples, C. Madew, R. Charles, and P. Budden. Further Studies to Evaluate

Crack Opening Areas for Through-Wall Cracks in the Vicinity of Welded Attach-
ments . In Proceedings of the ASME 2010 Pressure Vessels & Piping Conference,
Bellevue, 18–22 July 2010. Paper No. PVP2010-25257, pp. 733–739.
[123] P. Dillström and W. Zang. ProLBB - A Probabilistic Approach to Leak Before

Break Demonstration. Technical report, SKI, 2007. Report 2007:43.
[124] USNRC. xLPR Pilot Study Report. Technical report, USNRC, Washington, DC,
and EPRI, Palo Alto, CA, 2012. NUREG-2110 and EPRI 1022860.
[125] I. Varfolomeev, D. Ivanov, D. Siegele, and G. Nagel. Probabilistic Leak-Before-

Break Assessment of a Main Coolant Line . In Proceedings of the ASME 2010 Pres-
sure Vessels & Piping Conference, Bellevue, 18–22 July 2010. Paper No. PVP2010-
25676, pp. 371–378.
[126] D.W. Beardsmore, S.F. Yellowlees, J.K. Sharples, R.A. Ainsworth, and P.J. Bud-
den. A Probabilistic Application of the R6 Detectable Leakage Leak-Before-Break
Procedure . In Proceedings of the ASME 2002 Pressure Vessels & Piping Confer-
ence, Vancouver, 5-9 August 2002. Paper No. PVP2002-1180, pp. 97–104.
[127] M.P. Puls, B.J.S. Wilkins, G.L. Rigby, J.K. Mistry, and P.J. Sedran. A probabilistic
method for leak-before-break analysis of CANDU reactor pressure tubes . Nuclear
Engineering and Design, 185(2–3):241 – 248, 1998.
BIBLIOGRAPHY 208
[128] Y.J. Oh and Y.S. Chang. Integrated probabilistic assessment for DHC initiation,
growth and leak-before-break of PHWR pressure tubes . Nuclear Engineering and
Design, 275:30 – 43, 2014.
[129] G. Wilkowski, B. Brust, T. Zhang, G. Hattery, S. Kalyanam, E. Kurth, Y. Hioe,
M. Uddin, J.J. Johnson, O. R. Maslenikov, J.J. Johnson, A.P. Asfura, W. Creek,
B. Sumodobila, A.A. Betervide, O. Mazzantini, and B. Aires. Robust LBB Analyses
for Atucha II Nuclear Plant . In Proceedings of the ASME 2011 Pressure Vessels
& Piping Conference, Baltimore, 17–21 July 2011. Paper No. PVP2011-57939 , pp.
107–120.
[130] T. Zhang, F. Brust, G. Wilkowski, H. Xu, A. Betervide, and O. Mazzantini. Beyond
Design Basis Seismic Analysis for Atucha II Nuclear Plant . In Proceedings of the
20th International Conference on Nuclear Engineering, Anaheim, July 30–August
3 2012. Paper No. ICONE20-POWER2012-54241 , pp. 625–634.
[131] USNRC. International Piping Integrity Research Group (IPIRG) Program. Tech-
nical report, USNRC, 1997. NUREG/CR-6233.
[132] P. Bouchard. Practical applications of the R6 Leak-before-Break procedure . In
Proceedings of the seminar on leak before break in reactor piping and vessels, Lyon,
1995. pp. 127–136.
[133] M.S. Mannan. The 100 Largest Losses (in the Hydrocarbon Industry) 1974-
2013. Technical report, 2014. Available at https://uk.marsh.com/Portals/18/
Documents/100_Largest_Losses_23rd_Edition_2014.pdf.
[134] A. Zahoor. Ductile fracture handbook. Technical report, EPRI, 1989. Report EPRI
NP-6301-D.
[135] A.A. Willoughby and T.G. Davey. Plastic Collapse in Part-Wall Flaws in Plates.
ASTM, 1989. Am. Soc. Test. Mater. Spec. Tech. Publ. ASTM STP 1020, pp 390-410.
[136] M.D. Rana and R.J. Selines. Structural integrity assurance of high-strength steel
gas cylinders using fracture mechanics. Engineering Fracture Mechanics, 30(6):877
– 894, 1988.
[137] T.L. Anderson. Stress Intensity And Crack Growth Opening Area Solutions For
Through-wall Cracks In Cylinders And Spheres. Technical report, WRC Bulletin
478, Welding Research Council, Inc. New York, NY, January 2003.
[138] D. Green and J. Knowles. The treatment of residual stress in fracture assessment
of pressure vessels. Pressure Vessel Technology, 116(4):345 – 352, 1994.
[139] D. Green C.C. France and J.K. Sharples. New stress intensity factor and crack
opening area solutions for through-wall cracks in pipes and cylinders. Technical
report, AEA Technology Report AEAT-0643, 1996. Addendum, April, 1999.
[140] A.G. Miller. Review of limit loads of structures containing defects. International
Journal of Pressure Vessels and Piping, 32(1):197 – 327, 1988.
[141] W. Kastner, E. Röhrich, W. Schmitt, and R. Steinbuch. Critical crack sizes in
ductile piping. International Journal of Pressure Vessels and Piping, 9(3):197 –
219, 1981.
BIBLIOGRAPHY 209
[142] T.L. Anderson et. al. Development of Stress Intensity Factor Solutions for Surface
and Embedded Cracks in API 579 . Technical report, WRC Bulletin 471, Welding
Research Council, Inc. New York, NY, May 2002.
[143] J.C. Newman and I.S. Raju. An empirical stress-intensity factor equation for the
surface crack. Engineering Fracture Mechanics, 15(1):185 – 192, 1981.
[144] K.W. Nam, K. Ando, S. Yuzuru, and N. Ogura. Leak-before-break conditions of
plates and pipes under high fatigue stresses. Fatigue and Fracture of Engineering
Materials and Structures, 15(8):809–824, 1992.
[145] K. W. Nam, K. Ando, and N. Ogura. The effect of specimen size on the behaviour
of penetrating fatigue cracks. Fatigue and Fracture of Engineering Materials and
Structures, 16(7):767–779, 1993.
[146] Abaqus. Abaqus 6.13 documentation. Technical report, Dassault Systemes.
[147] J. Souza, F. Peres, and C. Schön. Methods for Introducing Pre-cracks in Fracture
Mechanics Testing of Polymers . In 17th European Conference on Fracture, Brno,
2 -5 September 2008. pp. 2220–2227.
[148] A. Huang, W. Yao, and F. Chen. Analysis of Fatigue Life of PMMA at Different
Frequencies Based on a New Damage Mechanics Model. Mathematical Problems in
Engineering, Volume 2014, 2014. Article ID 352676.
[149] A. Almar-Naes. Fatigue handbook - Offshore Steel Structures. TAPIR, TAPIR,
Norges tekniske högskole, 7034, Trondheim-NTH, Norge, 1985.
[150] ThePlasticShop.co.uk. A guide to the technical properties and data
of Perspex Sheet (cast). Technical report, 2015. Available at http:
//www.theplasticshop.co.uk/plastic_technical_data_sheets/perspex_
technical_properties_data_sheet.pdf.
[151] J. C. Jr. Newman and I. S. Raju. Analysis of surface cracks in finite plates under
tension or bending loads, nasa-tp-1578. Technical report, NASA Langley Research
Center; Hampton, VA, United States, 1979.
[152] I.S. Raju and J.C. Newman Jr. Stress-intensity factors for a wide range of semi-
elliptical surface cracks in finite-thickness plates. Engineering Fracture Mechanics,
11(4):817 – 829, 1979.
[153] M. Nagai, N. Miura, and M. Shiratori. Stress intensity factor solution for a surface
crack with high aspect ratio subjected to an arbitrary stress distribution using the
influence function method. International Journal of Pressure Vessels and Piping,
131:2 – 9, 2015. The 10th International Workshop on the Integrity of Nuclear
Components.
[154] Z. Wu. The shape of a surface crack in a plate based on a given stress intensity
factor distribution. International Journal of Pressure Vessels and Piping, 83(3):168
– 180, 2006.
[155] A. Hosseini and M.A. Mahmoud. Evaluation of stress intensity factor and fa-
tigue growth of surface cracks in tension plates. Engineering Fracture Mechanics,
22(6):957 – 974, 1985.
BIBLIOGRAPHY 210
[156] X.B. Lin and R.A. Smith. Finite element modelling of fatigue crack growth of surface
cracked plates: Part i: The numerical technique. Engineering Fracture Mechanics,
63(5):503 – 522, 1999.
[157] P.P. Paris and F.F. Erdogan. A critical analysis of crack propagation laws. ASME.
J. Basic Eng., 85(4):528 – 533, 1963.
[158] X.B. Lin and R.A. Smith. Finite element modelling of fatigue crack growth of
surface cracked plates: Part iii: Stress intensity factor and fatigue crack growth life.
Engineering Fracture Mechanics, 63(5):541 – 556, 1999.
[159] BSI. BS EN ISO 6892-1:2009 - Metallic materials. Tensile testing. Method of test
at ambient temperature . British Standards Institution, London, 2009.
[160] J.R. Yates, M. Zanganeh, and Y.H. Tai. Quantifying crack tip displacement fields
with {DIC}. Engineering Fracture Mechanics, 77(11):2063 – 2076, 2010. Interna-
tional Conference on Crack Paths 2009.
[161] P. Paris and H. Tada. The application of fracture proof design methods using tearing
instability theory to nuclear piping postulating circumferential throughwall cracks.
Technical report, USNRC, 1983. USNRC Report, NUREG/CR-3464.
[162] E. S. Folias. An axial crack in a pressurized cylindrical shell. International Journal

of Fracture Mechanics, 1(2):104–113, 1965.
Appendix A
Complementary photographs and

plots from experiments
Contents
A.1 Fracture surface from Perspex samples . . . . . . . . . . . . . 212
A.1.1 Series 1 (10x50x100mm) . . . . . . . . . . . . . . . . . . . . . . 212
A.1.2 Series 2 (15x50x100mm) . . . . . . . . . . . . . . . . . . . . . . 214
A.1.3 Series 3 (10x75x100mm) . . . . . . . . . . . . . . . . . . . . . . 216
A.1.4 Series 4 (15x75x100mm) . . . . . . . . . . . . . . . . . . . . . . 218
A.2 Recorded strain data from metallic specimens . . . . . . . . . 220
A.2.1 Static test . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 220
A.2.2 Fatigue test . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 222
A.3 Photographs of high strength steel specimens . . . . . . . . . 227
A.3.1 Fracture surface . . . . . . . . . . . . . . . . . . . . . . . . . . 227
A.3.2 Re-characterisation into through-wall crack . . . . . . . . . . . 228
211
APPENDIX A. COMPLEMENTARY INFORATIONS FROM EXPERIMENTS 212
A.1 Fracture surface from Perspex samples
A.1.1 Series 1 (10x50x100mm)
Figure A.1: Test on Perspex plate - Fracture surface - Sample S11


A.1.2 Series 2 (15x50x100mm)


A.1.3 Series 3 (10x75x100mm)


A.1.4 Series 4 (15x75x100mm)


A.2 Recorded strain data from metallic specimens
A.2.1 Static test
Static test recorded strain data for specimens M01-01 and M01-02 were provided in Chap-
ter6, Section6.3.4 and recorded strain data for specimens M01-01 and M01-02 are given
in Figure A.19, Figure A.20 and Figure A.21.
Figure A.19: Evolution of strains along X-axis under increasing static loads: 0kN, 15kN,
Figure A.20: Evolution of strains along Y-axis under increasing static loads: 0kN, 15kN,
Figure A.21: Evolution of strains along the diagonal direction under increasing static
loads: 0kN, 15kN, 50kN, 100kN and 150kN for Specimens M01-03 & M01-04
A.2.2 Fatigue test
The plots obtained from specimens fatigue test on M01-01 were presented in details in
Chapter 6. Other specimens plots are provided in this appendix:
• M01-02
– Figure A.22 - Strain evolution along the X-axis.
– Figure A.23 - Strain evolution along the Y-axis.
– Figure A.24 - Strain evolution along the diagonal axis.
• M01-03
• M01-04
Figure A.22: Evolution of the strains along the X-axis during fatigue test - M01-02
Figure A.23: Evolution of the strains along the Y-axis during fatigue test - M01-02
Figure A.24: Evolution of the strains for the diagonal strain gauges during fatigue test
- M01-02
- M01-03
- M01-04
A.3 Photographs of high strength steel specimens
A.3.1 Fracture surface
Figure A.31: Test on high strength steel plate - Fracture surface - Sample M01-02
Figure A.32: Test on high strength steel plate - Fracture surface - Sample M01-04
A.3.2 Re-characterisation into through-wall crack
Figure A.33: Re-characterisation into through-wall crack for Specimen M01-01
Figure A.34: Re-characterisation into through-wall crack for Specimen M01-04

Appendix B
Equations for crack opening area,

stress intensity factor and reference
stress
Contents
B.1 Equations of widely used COA models . . . . . . . . . . . . . . 230
B.1.1 Axial cracks . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 230
B.1.2 Circumferential cracks . . . . . . . . . . . . . . . . . . . . . . . 233
B.2 Stress Intensity Factor solutions . . . . . . . . . . . . . . . . . 234
B.2.1 Axial through-thickness flaws . . . . . . . . . . . . . . . . . . . 234
B.2.2 Circumferential through-thickness crack . . . . . . . . . . . . . 236
B.2.3 Axial surface flaws . . . . . . . . . . . . . . . . . . . . . . . . . 236
B.2.4 Circumferential surface defect . . . . . . . . . . . . . . . . . . . 237
B.3 Reference Stress solutions . . . . . . . . . . . . . . . . . . . . . 238
B.3.1 Axial through-thickness flaws . . . . . . . . . . . . . . . . . . . 238
B.3.2 Circumferential through-thickness crack . . . . . . . . . . . . . 239
B.3.3 Axial surface flaws . . . . . . . . . . . . . . . . . . . . . . . . . 239
B.3.4 Circumferential surface defect . . . . . . . . . . . . . . . . . . . 240
229
APPENDIX B. COA, SIF AND RS SOLUTIONS 230
B.1 Equations of widely used COA models
B.1.1 Axial cracks
B.1.1.1 Axial Crack Opening Area Models
Zahoor’s model [134]

Zahoor’s model for axial crack is applicable when λ ≤ 5 and σ/σys ≤ 0.25. This model
comes from COA model of plates and has been modified to take into account the curvature
of the shell on the crack length.
2πP Rm a2e V0
COA = (B.1)
E 0t
where
V0 = 1 + 0.64935λ2 − 0.0089683λ4 + 0.000133873λ6 (B.2)
a
λ= √ (B.3)
Rm t
When small-scale yielding is taken into account λ has to be replaced by λe with the
following definition:
ae
λe = √ (B.4)
Rm t
where
F σh
ae = a 1 + (B.5)
2 σys
F = 1 + .2987λ2 − 0.026905λ4 + 0.00053549λ6 (B.6)
The effect of this parameter is displayed in Figure B.1. As plasticity increases, COA
becomes larger. On the other hand, as the loading is far below the yield stress, the
corrected expression is closer to that of the elastic model.
Using parameters presented earlier and the formulation of λ in Zahoor’s solution
(Eq.B.3), the limit of applicability is reached for a crack length of 2a=454 mm.
Figure B.1: Plasticity correction effect - Zahoor solution
BS 7910:2013 [1]
In BS 7910, a table is provided with different models of COA available in the literature
depending on the structure and loadings. The solution presented in BS 7910 has been
adopted from other codes in R6 [60] or FITNET [59]. The solution given here is originally
taken from [104] which is the solution for plate corrected for cylindrical shape:
2πσm a2
COA = α(λ) (B.7)
E0
Where α(λ) is a correction factor derived using thin-wall and elasticity theories to
allow for bulging in terms of the shell parameter (λ). The solution is strictly valid only
when flaw lengths do not exceed the least radius of curvature of the shell. For axial cracks
with λ ≤ 8, α(λ) is defined as follow:
α(λ) = 1 + 0.1λ + 0.16λ2 (B.8)
p a
λ = 4 0.75(1 − ν 2 ) √ (B.9)
Rm t
When small-scale yielding is taken into account, an Irwin-type correction factor is

introduced. The radius of the plastic zone is added to the physical crack size and gives
a simple estimate of the influence of plastic deformation [141]. The terms in brackets
represents a first-order correction for the effects of crack-tip plasticity increasing with the
load level. S is the ratio of the membrane stress to the flow stress:
3/2 2 3/2
S2

S
γ(S) = 1 + − (B.10)
2 2
σm
S= (B.11)
σf
Using the input parameters presented earlier, the following crack-opening area to crack
length relation can be established. Remembering the formulation of λ in BS 7910, the
limit of applicability is reached for a crack half-length of a=400 mm, corresponding to

a normalised length of 0.399. This model includes a small-plasticity correction factor,
S. Influence of this parameter is presented in Figure B.2. As plasticity increases, COA
becomes larger. However, this model is only acceptable for materials with low plasticity
effects, i.e. low strain hardening.
Figure B.2: Plasticity correction effect - BS 7910 solution
API 579-1/ASME-FFS-1:2007 [9]

The recommended solutions for the calculation of axial crack opening areas in API 579-
1/ASME-FFS-1 are based on a wide range of linear elastic fracture mechanics FEA vali-
dation. Two solutions are provided depending on the type of loading: pressure (Eq.B.12)
or stresses (Eq.B.13). They also account for the crack taper resulting from through-wall
bending loads. They are evaluated using tabulated values, given for different ratios of
Ri /t. Note that when compared to other solutions, the solutions are provided for both
inner and outer surfaces and not at mid-thickness.
P Ro 2πa2
COA = Hp (B.12)
t E
or
2πa2
COA = (Pm H0 + Pb (H0 − 2H1 )) (B.13)
E
where:
A0 + A1 λ + A2 λ2 + A3 λ3
Hp = (B.14)
1 + A4 λ + A5 λ2 + A6 λ3 + A7 λ4
1.818a
λ= √ (B.15)
Ri t
These elastic solutions can be corrected for elastic-plastic conditions with a plasticity
correction factor computed from the load ratio as follows:
COAp = γp × COA (B.16)
where the plasticity modifier γp , is determined using the following equation, valid for
0 ≤ Lr ≤ 1.2:
γp = 1.008 − 0.33015L2r + 5.53696L4r − 3.96974L6r + 2.00844L8r (B.17)
Influence of this parameter is displayed on Figure B.3. As plasticity increases, COA

becomes larger.
Figure B.3: Plasticity correction effect - API 579-1/ASME FFS-1 solution
B.1.2 Circumferential cracks
B.1.2.1 Circumferential Crack Opening Area Models
Paris-Tada’s model [161]

Paris-Tada’s model is used for a circumferential crack when λ ≤ 5:
2
πP Rm
COA = ×B (B.18)
E
where: (
λ2 + 0.16125λ4 for 0 < λ < 1
B= (B.19)
0.02 + 0.81λ2 + 0.30λ4 + 0.03125λ4 for 1 < λ < 5
r
Rm
λ=θ× (B.20)
t
When small-scale yielding is taken into account λ is replaced with λe using the following
definition: r
Rm
λe = θe × (B.21)
t
Fm2 P 2 Rm2

θe = θ × 1 + (B.22)
8t2 σys2
(
1 + 0.1501λ1.5 for 0 < λ < 1
Fm = (B.23)
0.8875 + 0.2625λ for 1 < λ < 5
BS 7910:2013 [1]
The solution used for circumferential through-wall crack is the same as that for axial
through-wall crack (Eq.B.7) except for the correction factor α(λ) , which is only valid for
λ ≤ 5:
2πσm a2
COA = α(λ) (B.24)
E0
√
α(λ) = 1 + 0.117λ2 (B.25)
API 579-1/ASME-FFS-1:2007 [9]

The recommended solution for crack opening areas calculation in API 579-1/ASME-
FFS-1 is based on a wide range of FEA validation, with λ given in Eq.B.15.
P Ro2 2πa2
COA = H0 2 (B.26)
Ro − Ri2 E
or
2πa2
COA = (Pm H0 + Pb (H0 − 2H1 )) (B.27)
E
where:
A0 + A1 λ + A2 λ2
H0 = (B.28)
1 + A3 λ + A4 λ2 + A5 λ3
B.2 Stress Intensity Factor solutions

Stress Intensity Factor, KI , solutions for a wide range of geometry and loading config-
urations are available in many handbooks or standards. Alternatively, numerical analysis
methods may be used to derive solutions. This helps to establish the state of the stress at
the location of a crack. A comparison of SIF solutions for various geometries can be found
in Ref.[59]. For this reason, only solutions for an axial through-wall crack in a cylinder
will be presented and calculations provided. Some remarks will be added to assist the
understanding of the assessment method. Stress Intensity Factor solutions are functions
of the applied primary stresses (Pm , Pb ), the defect size (2a, 2c) and the structural dimen-
sions (t, W , Ri ). Procedures as API 579-1/ASME FFS-1 or BS 7910 contain solutions
for many crack geometries that are likely to occur in large range of components including
pressurized components.
B.2.1 Axial through-thickness flaws
BS 7910
A general expression is provided in BS 7910 for all assessments:
√
KI = Y σ πa (B.29)
where Y σ is divided into two parts to take into account the contribution of primary and
secondary stresses:
Y σ = (Y σ)p + (Y σ)s (B.30)
These are calculated as follows, taking into account local stress concentration, misalign-
ment or residual stresses.
(
(Y σ)p = M fw (ktm Mkm Mm Pm + ktb Mkb Mb [Pb + (km − 1)Pm ])
(B.31)
(Y σ)s = Qm Mm + Qb Mb
In a simple case of an axial through-wall crack (M ,fw =1) under internal pressure with
no stress concentration factor (ktm , ktb , Mkm , Mkb = 1), no misalignment (km = 1) and no
residual stresses (Qm , Qb = 0) stresses, Eq.B.31 can be reduced to :
(
(Y σ)p = Mm Pm + Mb Pb
(B.32)
(Y σ)s = 0
The Stress Intensity Factor is therefore written in terms of membrane and bending
primary stresses corrected by Mm and Mb respectively. These coefficients are given at
both inner and outer surfaces from tabulated values for a range of Rm /t ratios. The
stress intensity factor solution is reduced to:
√
KI = (Mm Pm + Mb Pb ) πa (B.33)
Note that all BS 7910 solutions exclude the effect of crack face pressure for pressurized
components.

Several definitions are available in API 579-1/ASME FFS-1. In terms of membrane
and bending stress or polynomial stress distribution the two following equations are
given: √
KI = {(Pm + Pc ) G0 + Pb (G0 − 2G1 )} πa (B.34)
√
KI = {(σ0 + Pc ) G0 + σ1 G1 } πa (B.35)
Another simplified formulation is available for internal pressure only:
P Ro √
KI = Gp πa (B.36)
t
where:
A0 + A1 λ + A2 λ2 + A3 λ3
G0,1,p = (B.37)
1 + A4 λ + A5 λ2 + A6 λ3
A0 −→ A6 are influence coefficients given at both inner and outer surfaces from tabulated
values for a range of t/Ri .
B.2.2 Circumferential through-thickness crack
BS 7910
General formulation of the stress intensity factor is used here (Eq.B.29). However
primary membrane stress, Pm in the general form, should be multiplied by a factor β
defined as: v
u 2Rm
β=t (B.38)
u
a
atan 2Rm
Other previous assumptions are also taken into account. The Stress Intensity factor
can be written in terms of membrane and bending primary stress corrected by Mm and Mb
respectively. These coefficients are given at both inner and outer surfaces from tabulated
values for a range of Rm /t ratios. Introducing Eq.B.38 into Eq.B.33 leads to:
√
KI = (Mm βPm + Mb Pb ) πa (B.39)

Several definitions are available in API 579-1/ASME FFS-1. In terms of membrane
and bending stress or polynomial stress distribution the two following equations are
given: √
KI = {(σm + Pc ) G0 + σb (G0 − 2G1 )} πa (B.40)
√
KI = {(σ0 + Pc ) G0 + σ1 G1 } πa (B.41)
Another simplified formulation is available when internal pressure only is considered:
P Ro2 √
KI = 2
G πa
2 0
(B.42)
Ro − Ri
where:
A0 + A1 λ + A2 λ2 + A3 λ3
G0,1 = (B.43)
1 + A4 λ + A5 λ2 + A6 λ3
Here A0 −→ A6 are influence coefficients given for some ratio t/Ri .
B.2.3 Axial surface flaws
BS 7910
The general formulation of stress intensity factor is used here (Eq.B.29). Other previous
assumptions are also taken into account. The Stress Intensity factor can be written in
terms of membrane and bending primary stress corrected by Mm and Mb respectively.
These coefficients are given at both inner and outer surfaces from tabulated values for a
range of Rm /t ratios: √
KI = (Mm Pm + Mb Pb ) πa (B.44)

Several definitions are available in API 579-1/ASME FFS-1. The two following equa-
tions are given depending on the loading. For internal pressure loading, with R being the
internal or external radius depending on the position of the defect:
P R2
a 4 r πa
a a 2 a 3
KI = 2 2G0 − 2G1 + 3G2 − 4G3 + 5G4 (B.45)
Ro − Ri2 R R R R Q
when Through-Wall Fourth Order Polynomial Stress Distribution is available:
r
a a2 a3 a4 πa
KI = G0 (σ0 + Pc ) − σ1 G1 + σ2 G2 − σ3 G3 + σ4 G4 (B.46)
t t t t Q
Where: (
a 1.65

1 + 1.464 c
for a/c ≤ 1
Q= a 1.65
(B.47)
1 + 1.464 c
for a/c > 1
Here G0 −→ G4 are influence coefficients given for a wide range of ratio t/Ri . Influence
coefficients are given for internal and external crack, so a different Stress Intensity Factor
may be evaluated depending on the crack location.
B.2.4 Circumferential surface defect
BS 7910
The general formulae of stress intensity factor is used here (Eq.B.29). Other previous
assumptions are also taken into account. The Stress Intensity factor can be written in
terms of membrane and bending stress corrected by Mm and Mb given for the deepest point
in the flaw and the point where the flaw intersects with the free surface from tabulated
coefficients given for some ratio Rm /t.

Several definitions are available in API 579-1/ASME FFS-1. The two following equa-
tions are given depending on the loading. For internal pressure and net section axial force
loading, with R being the internal or external radius depending on the position of the
defect:
P R2
r
F πa
KI = 2 2
+ 2 2
G0 (B.48)
Ro − Ri π(Ro − Ri ) Q
Another formulation is given based on Through-Wall Fourth Order Polynomial Stress
Distribution:
r
a a2 a3 a4 πa
KI = G0 (σ0 + Pc ) − σ1 G1 + σ2 G2 − σ3 G3 + σ4 G4 + σ5 G5 (B.49)
t t t t Q
Where: (
a 1.65

1 + 1.464 c
for a/c ≤ 1
Q= a 1.65
(B.50)
1 + 1.464 c
for a/c > 1
Here G0 −→ G6 are influence coefficients given for some ratio t/Ri . Influence coefficients
are given for internal and external crack, so a different Stress Intensity Factor may be
evaluated depending on the crack location.
B.3 Reference Stress solutions

The reference stress, σref , is a measure of the applied stress in the un-cracked region
of a cracked component. It accounts for the loss of cross sectional area in the region
containing a flaw. Reference stress solutions are functions of the applied primary stresses
(Pm , Pb ), the defect size (2a, 2c) and the structural dimensions (t, W , Ri ). API 579-
1/ASME FFS-1 or BS 7910 provide reference stress solutions for many crack geometries
that are likely to occur in pressurized components. As no comparison work has been
previously performed for this parameter, the following section includes such discussion
for the different geometries.
B.3.1 Axial through-thickness flaws
BS 7910
BS 7910 refers to both Willoughby and Davey [135] and Folias [162] solutions. The
original solution from Willoughby and Davey has been developed for plates. To apply
the solution to cylinder, the conversion factor developed by Folias is associated with
membrane stress. A factor of 1.2, not present in the original paper [135], is also applied
to the membrane stress. It is intended to produce a similar level of conservatism as that
inherent in the flat plate solution.
2Pb
σref = 1.2MT Pm + 2a (B.51)
3(1 − W )
where s
a2
MT = 1 + 1.6 (B.52)
Ri t

The solution for an axial through-thickness flaw in a pipe available in API 579-1/ASME
FFS-1 is given as: p
Pb + Pb2 + 9(Mt Pm )2
σref = (B.53)
3
where three definitions are given for Mt , reported here as Mt1 , Mt2 and Mt3 :
r
(1.02 + 0.4411λ2 + 0.006124λ4
Mt1 = (B.54)
1 + 0.02642λ2 + 1.533 × 10−6 λ4
(√
1 + 0.3797λ2 − 0.001234λ4 for λ ≤ 9.1
Mt2 = (B.55)
(3.3 + 0.01936λ)2 for λ > 9.1
√
Mt3 = 1 + 0.4845λ2 (B.56)
and
a
λ = 1.818 √ (B.57)
Ri t
B.3.2 Circumferential through-thickness crack
BS 7910 The following solution is available in BS 7910:
π(Pm,a + Pm,p ) πPm,b (Ro4 − Ri4 ) 2Pb,l

σref = + +
a a sin2 Ra sin 2a
2a
π− Ri
− 2arcsin 0.5sin Ri 2 t) π − a − 2
(4Ro Rm i
− Ri 3 1− πRi
Ri π− a
Ri
2
(B.58)
where the primary membrane stress is divided into:
• Pm,a primary membrane stress due to global axial loads
• Pm,p primary membrane stress due to internal pressure
• Pm,b primary membrane stress due to global bending moments
• Pb,l primary through-wall bending stress

The following solution is available in API 579-1/ASME FFS-1:
p
Pb + Pb2 + 9(ZPm (1 − α)2 )2
σref = (B.59)
3(1 − α)2
where:
a
θ= (B.60)
Rm
a
α= (B.61)
πRm
t
τ= (B.62)
Ro
sinθ
φ = arcos( ) (B.63)
2
π(Ro2 − Ri2 )
Z= ) (B.64)
(2 − τ )(Ro t(2φ − θ))
The reference stress solution in API 579-1/ASME FFS-1 is based on the crack angle
measured on the mean radius of the pipe or cylinder shell.
B.3.3 Axial surface flaws
BS 7910
The following solution is available in BS 7910:
2Pb
σref = 1.2Ms Pm + (B.65)
3(1 − α00 )2
where: a
1− tMT
Ms = with MT from Eq. (B.66)
1 − at
a
α00 = t
t for W ≥ 2(c + t) (B.67)
1+ c
In BS 7910, the surface correction factor is based on the original Folias solution for
thin walled cylinders [162]. The surface correction factor corresponds to the one called
Ms1 in API 579-1/ASME FFS-1 for a rectangular crack shape. Assuming that the crack
shape is rectangular will lead to conservative results.

p
gPb + (gPb )2 + 9(Ms Pm (1 − α)2 )2
σref = (B.68)
3(1 − α)2
where: a 0.75
g = 1 − 20α3 (B.69)
2c
a
t
α= t (B.70)
1+ c
Two definitions are available for the surface correction factors:


1 − C tM a 1 for rectangular shape

t
Ms1 = a where C= 0.67 for parabolic shape (B.71)
1−Ct 
0.85 for a best fit to the data

1
Ms2 = a (B.72)
1− t
+ at Mt (λ
1
a)
where:
c
λa = 1.818 √ (B.73)
Ri a
B.3.4 Circumferential surface defect
BS 7910
The following solution is available in BS 7910 where R represents the internal or ex-
ternal radius depending on the crack location:
Pm π 1 − at + 2 at sin Rc

2Pb
σref = a
a c
+ (B.74)
1− t π− tR 3(1 − α00 )2
where ( a
00
t
1+ ct
for πR ≥ c + t
α = (B.75)
a c
t πR
for πR < c + t

p
Pb + Pb2 + 9(ZPm (1 − α)2 )2
σref = (B.76)
3(1 − α)2
where: (
πc
4Ro
for external crack
θ= πc
(B.77)
4Ri
for internal crack
a
x= (B.78)
t
t
τ= (B.79)
Ro
(1 − τ )(2 − 2τ + xτ )(1 − τ + xτ )2

A=x (B.80)
2(1 + (2 − τ )(1 − τ )
a
t
α= t (B.81)
1+ c
ψ = arcos(Asinθ) (B.82)
1
Z= 2ψ xθ 2−2τ +xτ
(B.83)
π
− π 2−τ

Leak Before Break Fracture Mechanics Thesis

Uploaded by

Copyright:

Available Formats

Leak Before Break Fracture Mechanics Thesis

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Leak Before Break Fracture Mechanics Thesis

Uploaded by

Copyright:

Available Formats

What is the main objective of the thesis?

What is the main objective of the thesis?

What are the two methodologies employed to investigate the transition of flaws?

What are the two methodologies employed to investigate the transition of flaws?

The mechanism of Leak-Before-

Break fracture and its application in

A thesis submitted for the degree of

Department of Mechanical, Aerospace and Civil Engineering

This thesis investigated the different aspects and mechanisms of leak-before-break

Writing these acknowledgements marks the end of my adventure at TWI / Brunel

Symbols list vii

4 LBB parameters - Analysis of current solutions from codes 57

5 Numerical Methods - Investigation of Ideal Through-Wall Cracks in

6 Experimental Work - Stress-Strain Field on the Back Surface of Surface

7 Numerical Simulation - Stress-Strain Field on the Back Surface of Sur-

8 Concluding remarks and future work 191

A Complementary photographs and plots from experiments 211

A.1.4 Series 4 (15x75x100mm) . . . . . . . . . . . . . . . . . . . . . . . . 218

Symbol Name Unit

a Crack depth (surface crack) / Half crack length (through-wall mm

CFP Crack Face Pressure

Fracture mechanics-based structural integrity assessments, most commonly referred

A typical requirement for average component would be to stipulate the maximum

1.2 Research hypothesis

Traditional crack evaluation uses conservative methods to predict maximum loading

1.3 Aims and objectives

• Review available integrity assessment / leak-before-break procedures available;

1.4 Outline of thesis

Chapter 3 focuses on the literature review on Leak-Before-Break. Its underlying basic

1.5 Contribution to new knowledge

The main contributions from this work are:

1.6 List of publications

- R. Bourga, P. Moore, Y.-J. Janin, B.Wang, and J. Sharples. Leak-before-break: Global

- R. Bourga, P. Moore, Y.-J. Janin, B.Wang. Comparison of BS 7910:2013 and API

- R. Bourga, Y.-J. Janin, B.Wang. Transition from part-penetrating to through thick-

2.1 Typical defects in pressurised equipment

2.1.1 Fabrication induced defects

Numerous defects can be introduced in components during fabrication, especially weld-

2.1.2 Environmentally induced defects

2.1.2.2 Damage related to the fluid

Paris’ law (Eq.2.1):

2.1.3 Third-party caused defects

• Dents are a permanent plastic deformation due to an impact. In case of pressurized

Figure 2.2: Illustration of a dent and gouge combination [9]

2.2 Non-Destructive Testing (NDT)

Non-Destructive Testing (NDT) is the inspection of materials without causing any

2.2.1 Conventional NDT (Surface/Embedded defects)

According to pressure vessel design codes such as BS PD 5500 [12], Radiographic

2.2.2 Leak Testing (Through-wall defects)

Leak testing methods group together a number of conventional and unconventional

2.3 Fracture mechanics tools

Figure 2.5: Illustration of a circumferential crack in a weld (TWI Ltd)

2.3.1 Crack behaviours

Figure 2.6: Illustration of the three modes of crack opening

Figure 2.7: Illustration of the plane stress / plane strain conditions

• LEFM methods are restricted to small-scale yielding configurations (i.e., geome-

• Energy release rate, G

2.3.2 Energy release rate, G (Energy-based)

Figure 2.8: Schematic driving force vs. rising R curve