Sunteți pe pagina 1din 20

7.

01
An Overview of Failure Assessment
Methods in Codes and Standards
U. ZERBST
GKSS, Geesthacht, Germany
K.-H. SCHWALBE
GKSS, Geesthacht, Germany
and
R. A. AINSWORTH
British Energy Generation Ltd., Gloucester, UK
7.01.1 INTRODUCTION

7.01.2 FRACTURE MECHANICS IN DESIGN, FITNESS-FOR-SERVICE PHILOSOPHIES


7.01.2.1 Conventional Quality Assurance Concepts and/or Fitness-for-Service?
7.01.2.1.1 Combined application of fracture mechanics concepts and conventional quality assurance
measures
7.01.2.1.2 Complementary application of fracture mechanics concepts and conventional quality assurance
measures
7.01.2.2 Options of Fracture Mechanics in Design
7.01.2.2.1 Damage tolerant design
7.01.2.2.2 Fail-safe design
7.01.2.2.3 Periodic proof testing
7.01.2.2.4 Periodic removal of the crack
7.01.2.2.5 Fracture mechanics based fatigue limit
7.01.2.3 Reliability and Safety Factors

4
4

5
5
5
10
12
13
14
15

7.01.3 DETERMINATION OF THE END-OF-LIFE CONDITION


7.01.3.1 Failure Conditions
7.01.3.2 CDF Parameters
7.01.3.3 Material Resistance to Crack Extension
7.01.3.4 Input Information

17
17
18
20
21

7.01.4 TYPES OF ASSESSMENT METHOD


7.01.4.1 Overview
7.01.4.2 Design Curve Approaches
7.01.4.3 FAD Approach According to R6IBS 7910
7.01.4.4 Fully Plastic Solutions of the EPRI Type
7.01.4.5 Approaches Based on the CEGB Reference Stress Method
7.01.4.6 The Engineering Treatment Model
7.01.4.7 The European SINTAP Procedure

23
23
23
24
25
26
27
28

7.01.5 INDUSTRY SPECIFIC PROCEDURES


7.01S . 1 Aerospace Industrj

29
29

An Overview of Failure Assessment Methods in Codes and Standards

7.01.5.1.I Special features in aerospace applications


7.01.5.1.2 NASGROIESACRACK
7.01.5.2 Nuclear and Conventional Power Generation
7.01.5.2.1 Special features in nuclear and conventional power generation
7.01.5.2.2 The A S M E boiler and pressure vessel code
7.01.5.2.3 Further guides
7.013 . 3 Petrochemical and Chemical Industries
7.01.5.3.1 Special features in petrochemical and chemical industries
7.01.5.3.2 The API 579 guide
7.013 . 4 Pipelines
7.01J.4.1 Special features in the pipeline industry
7.013.4.2 The W R C method
7.01.5.4.3 API 1104 and C S A 2662
7.01S.4.4 EPRG guidelines
7.01.5.4.5 Further guides
7.01.5.5 Civil Engineering and Offshore Installations
7.01S.5.1 Special features of steel structures
7.013.5.2 Eurocode 3
7.01.5.5.3 DNV R P 404 and other guides
7.01.5.6 Special Applications: Welded Structures
7.01.5.6.1 Special features of weldments
7.01.5.6.2 IIW Fitness-for-Purpose guide
7.01.5.6.3 W E S 2805
7.01S.6.4 Further guides
7.01.5.7 Special Applications: Pressure Vessels
7.01s.7.1 Special features of pressurized components
7.01.5.7.2 PrEN 13445
7.01.5.7.3 Further guides

29
30
30
30
31
32
32
32
32
33
33
33
33
34
35
35
35
35
36
37
37
37
37
37
37
37
38
38

7.01.6 HORIZONTAL PROCEDURES


7.01.6.1 Introduction
7.01.6.2 R6IBS 7910:1999
7.01.6.3 The EFAM Method
7.01.6.4 The European SINTAP Procedure

38
38
39
39
40

7.01.7 SOURCES OF CALCULATIONAL INPUTS


7.01.7.1 Stress Intensity Factor Solutions
7.01.7.2 Limit Load Solutions
7.01.7.3 Residual Stress Distributions
7.01.7.4 Constraint Parameters

41
41
42
43
43

7.01.8 CONCLUSIONS .

43

7.01.9 GUIDE TO VOLUME 7

44

7.01.10 REFERENCES

44

NOMENCLATURE
a

a,
aeff

a0

crack size, half length of a center


crack, crack depth for a surface flaw
or half height for an embedded flaw
critical crack size
plasticity corrected crack size
initial crack size (referring to the
NDE detection limit in damage
tolerant design, Figure 3, in other
cases to some representative size
from the manufacturing process)
half crack length for a surface crack
or an embedded flaw
materials constant from the Paris
Law, Equation (l), Figure 5

c1, c2, c3

constants, design curve approaches,


Equation (8)

daldN
E
El

ALJ
F
FY

h
J
Je

Jmat
JP
Jssy

crack propagation rate per cycle


modulus of elasticity
modulus of elasticity for constraint
conditions (plane stress: E = E ;
plane strain: E = E/(1-v2))
plasticity correction function (R6,
BS 7910, ETM, SINTAP)
applied force, also general term for
load
yield load (limit load)
dimensionless shape function, EPRI
approach, Equation (17)
J-integral
elastic component of the J-integral
fracture resistance in terms of J ,
general
plastic component of the J-integral
J-integral, small scale yielding
linear elastic stress intensity factor

Nomenclature
plasticity corrected stress intensity
factor
linear elastic stress intensity factor,
crack opening mode I
fracture resistance in terms of K,
general
upper value of K factor range in
fatigue (Figure 6)
lower value of K factor range in
fatigue (Figure 6)
ratio of KI and Kmat (FAD approaches)
reference fracture toughness, ASME
curve, Equation (43)
characteristic dimension, Equation
(7), EPRI approach
ratio of applied load to yield load,
Lr = FIFY = Jreflay
maximum allowable value of Lr
materials constant from the Paris
Law, Equation (l), Figure 5
stress intensity magnification factors, Equation (42)
strain hardening exponent, Ramberg-Osgood description, Equation
(18)
strain hardening exponent (cyclic
stress-strain curve)
strain hardening exponent, various
definitions, Equations (29) and (40),
Figure 27
number of load cycles in fatigue
applied load, EPRI approach
failure probability
fracture mechanics based damage
parameter, used as ordinate of a
generalized S-N curve, Equation (3)
reference load, Equation (17), EPRI
approach
calibration factor for K as a function
of a/2c shape in surface flaw
hydrostatic @stress
radius of the plastic zone
stress ratio in fatigue, R = Kmin/Kmax
elasticity limit of steels with a Luders
plateau
tensile strength
yield strength at 0.2% plastic strain
reference temperature for KIR curve,
ASME, Equation (43)
time up to failure (Figure 7)
T-stress
minimum design temperature, Eurocode, Equation (45)
reference temperature, Master Curve
approach, in Eurocode designated as
TlOO

dimensionless shape function for K

3
distance through the thickness of a
component, Equation (47)
coefficient of the Ramberg-Osgood
description of the stress-strain
curve, Equation (1 8)
crack tip opening displacement,
CTOD, general
fracture resistance in terms of 6,
general
elastic component of CTOD
plastic component of CTOD
CTOD defined for a gauge
length of 5mm (Figure 28), ETM
approach
value of the CTOD-B5 at F= Fy
inspection interval, Figure 7
effective range of the J-integral in
fatigue
stress intensity factor range,
AK= Kmax-Kmin
effective stress intensity factor range
due to crack closure
threshold value of AK below which
there is no crack propagation
A

AJ

safety factors, Ezirocode, Equation


(45)
stress range in fatigue, AJ=J,,,~~-Jmin

A&
E
Ea
&ref
CY
60

EO
EC1

c1
V

Jb
Jm
Jmax
Jmin
JO

JO

Jcl
J

Jref
JY

@D

Luders strain
strain
applied strain
reference strain, Equation (24)
yield strain, strain at cy
reference strain, Ramberg-Osgood
description of the stress-strain
curve, Equation (17)
strain at crack opening in fatigue
strain at crack closure in fatigue
constraint factor, Equation (21)
Poissons ratio
bending stress component (Figure
31)
membrane stress component (Figure
31)
upper value of stress range in
fatigue, Figure 6
lower value of stress range in fatigue,
Figure 6
stress at crack opening in fatigue
reference stress, Ramberg-Osgood
description of the stress-strain
curve, Equation (17)
stress at crack closure in fatigue
standard deviation, Equation (4)
reference stress
yield strength, general
normalized
CTOD, @D = Jmat/
(2zeya), design curve approaches

A n Overview of Failure Assessment Methods in Codes and Standards

7.01.1 INTRODUCTION
This volume provides comprehensive up-todate information on the assessment of the
integrity of engineering structures containing
crack-like flaws, in the absence of effects of
creep at elevated temperatures (see Volume 5 )
and of environment (see Volume 6 ) . Key
methods are extensively reviewed and background information as well as validation is
given. However, it should be kept in mind that
for actual detailed assessments the relevant
documents have to be consulted.
In classical engineering design, an applied
stress is compared with the appropriate material resistance expressed in terms of a limit stress,
such as the yield strength or fatigue endurance
limit. As long as the material resistance exceeds
the applied stress, integrity of the component is
assured. It is implicitly assumed that the
component is defect-free but design margins
provide some protection against defects. Modern design and operation philosophies, however, take explicit account of the possible
presence of defects in engineering components.
Such defects may arise from fabrication, e.g.,
during casting, welding, or forming processes,
or may develop during operation. They may
extend during operation and eventually lead to
failure, which in the ideal case occurs beyond
the design life of the component.
Failure assessment methods are based upon
the behavior of sharp cracks in structures, and
for this reason all flaws or defects found in
structures have to be treated as if they are sharp
planar cracks. Hence the terms flaw or defect
should be regarded as being interchangeable
with the term crack throughout this volume.
7.01.2 FRACTURE MECHANICS IN
DESIGN, FITNESS-FOR-SERVICE
PHILOSOPHIES
The various chapters of Volume 2 gives
detailed accounts of the theories underlying the
failure of materials and the principles of
fracture mechanics. This volume addresses the
application of these theories and principles to
real structures and components.
7.01.2.1 Conventional Quality Assurance
Concepts and/or Fitness-for-Service?
In general, fracture mechanics concepts are
applicable to the safe design of any component
and structure containing flaws such as cracks.
Their practical use is, however, restricted by
certain limiting conditions, which vary from
case to case. Limits are set by economic and
organizational aspects as well as by engineering

feasibility. A fracture mechanics based defect


assessment concept will have to accommodate
all these aspects and it should also include
proven conventional concepts of quality assurance, The relation between these concepts and
fracture mechanics is briefly discussed in the
following paragraphs.
Conventional design concepts as a rule are
aimed at producing components free of any
imperfections. Associated component tests and
the guarantee of satisfying certain manufacturing standards as well as final quality control
measures are intended to meet this requirement. The common term for this philosophy is
Good Workmanship. It is, however, quite
obvious that production of a perfectly defect
free component is unlikely. Instead, a structure
is regarded as defect free if no defect is found
by nondestructive evaluation (NDE). It has to
be noted, that both the detection limit and the
probability of detection (POD) are a matter of
the test method applied including subjective
factors such as training and experience of the
operator and the practical conditions under
which the evaluation has to be carried out.
In contrast to the Good Workmanship
philosophy a fracture mechanics based Fitnessfor-Service (FFS) analysis, sometimes called
Fitness-for-Purpose analysis, provides an individual treatment of the component, which is
always regarded as containing a crack. Due to
this, its results will vary from case to case in the
same way as a conventional design for strength
usually does. Frequently the fracture mechanics
based critical crack size will be larger than the
detection limit of the nondestructive evaluation.
In those cases the FFS analysis will yield results
that are less conservative than the outcome
of a Good Workmanship application. But the
opposite case is possible, such that a component
designed according to Good Workmanship
principles is proven as unsafe on the basis of
a fracture mechanics FFS analysis.
Nowadays a trend is observed towards an
increasing need for FFS analyses. This is
promoted by features such as a permanent
increase of the level of performance of engineering components and structures; by the
introduction of new materials, joining technologies, and design principles; the problem of
ageing components and life extension; an
increased emphasis on nondestructive inservice inspection combined with improved
NDE techniques; and also a number of failure
events caused by fatigue and fracture.
In principle, Good Workmanship and FFS
are independent and competing concepts.
However, it is often reasonable to apply them
in a combined or complementary manner as
discussed below.

Fracture Mechanics in Design, Fitness-for-service Philosophies


7.01.2.1.1 Combined application of fracture
mechanics concepts and
conventional quality assurance
measures
An inherent problem of a fracture mechanics
analysis is that there are limitations to NDE in
achieving an adequate POD and in sizing a
crack. In particular, the determination of the
crack depth in the wall thickness direction may
create difficulties and as a consequence the
advantage of the individual treatment of the
component may be devalued. In cases where
NDE crack sizing is a major problem the design
can be based on a combination of conventional
quality assurance and fracture mechanics. This
may be illustrated by an example: Following a
Good Workmanship criterion cracks are unacceptable. As already mentioned, in practical
terms this requirement means that no crack is
acceptable whose dimensions exceed the detection limit of the NDE technique applied.
Possible flaws below the detection limit are
not detected. At that point a fracture mechanics
approach may be taken. Starting from an initial
flaw size defined by the NDE detection limit,
fatigue or other mechanisms of crack extension
can be simulated up to the critical stage of the
component. This way, the residual lifetime is
obtained and inspection intervals can be
defined such that a crack will be found prior
to failure. If a crack is actually detected during
an inspection it will, however, not be tolerated
even if it is not yet in a critical state. Instead,
the component has to be repaired or retired.
Thus, in that case the aim of the fracture
analysis is not to allow for further service of the
flawed component but is restricted to the
development of a meaningful inspection plan.
Performing a fracture mechanics analysis
does not necessarily mean that a subcritical
crack will be accepted after it is detected by
NDE. If no crack is allowed at all, fracture
mechanics, nevertheless, has to be applied in
order to guarantee that a crack that might be
overseen in inspection cannot extend to a
critical size before it will finally be found.
7.01.2.1.2 Complementary application of
fvacture mechanics concepts and
conventional quality assurance
measures
When applying Good Workmanship and
FFS in a complementary way the latter has to
serve as a second level of safety. Basic safety is
guaranteed by imposing the requirement that
the component is defect-free and this is sometimes accompanied by component tests.
A second level of safety is necessary due to
inevitable shortcomings of the basic level. In

order to simulate the most critical conditions,


the tested components would have to include
flaws just below the NDE detection limit. This
is, however, unlikely for the limited number of
components tested. The number of structures
in service is much larger and the probability
that one of the in-service components meets the
critical condition is much higher. In addition, a
problem that the philosophy of component
tests is faced with, is that many structures, such
as axles or wheels of vehicles, experience an
extraordinarily high number of loading cycles
over their lifetime. It is state-of-the-art today
that even ferritic steels do not exhibit a fatigue
limit above lo6 or lo7 cycles (Miller and
O'Donnell, 1999; Bathias, 1999). A component
test, which includes 10'' or more loading
cycles, is usually not realistic. The gap between
component tests and practical applications can
be bridged by fracture mechanics analyses
which complement the basic safety arguments.
7.01.2.2

Options of Fracture Mechanics in


Design

No unique scheme exists on how to apply


fracture mechanics to different components.
One of the most important aspects is the time
taken for a detected crack to extend to its
critical size. If this time is sufficiently large, a
design concept can be applied which is based on
inspection intervals. If it is rather small the
concept has to aim at avoiding crack initiation.
In the remainder of this subsection, a number of
different options of fracture mechanics application in design are briefly discussed. These are:
(i) damage tolerant design;
(ii) fail-safe design (redundant design, crack
arrest, leak-before-break);
(iii) periodic proof testing;
(iv) periodic removal of cracks; and
(v) durability design using a fracture mechanics based fatigue endurance curve.
7.01.2.2.1 Damage tolerant design
Damage tolerance is the classical concept of
fracture mechanics application. It is illustrated
in Figure 1. A typical application field is the
aerospace sector. The steps of a damage
tolerant analysis are:
(i) Establishing an initial crack shape and
size for the further analysis.
(ii) Simulating subcritical (e.g., fatigue)
crack extension.
(iii) Determining the critical crack size for
component failure.
(iv) Determining the residual lifetime of the
component.
(v) Establishing inspection intervals.

An Overview of Failure Assessment Methods in Codes and Standards


Detection limit of NDT

Apply another design

Initial crack size a,

Fracture mechanics
analysis

Is the initial crack size


already critical?

If necessary, consideration of retardation effects

crack extension

Fracture mechanics
analysis

critical(a = a,)?

On the basis adS


S: safety margin (>1)

Yes

Determination of

Is the critical crack size


n,lS adequately detectable

and sizeable under service


conditions?
Take into account, that a
crack may be missed in one
or more inspections!
Work out an
inspection plan!

Application of another NDT


technique with lower detection
limit and better POD possible?

Establishing inspection
intervals

Are the inspection intervals


realistic under service
Yes
conditions?

no

When a crack is detected

.
Are subcritical cracks

component

As above, a, is replaced
with the actual crack size

Establish new
inspection intervals
I

Figure 1 Fracture mechanics analysis within the framework of Damage Tolerant Design.

In the context of a damage tolerant design


the initial crack size will be based on the
capability of the applied NDE technique. In
principle, this is given by the NDE detection
limit which is, however, a statistical quantity
rather than a fixed value. Essential factors,
which affect both the detection limit and the
POD are:

the degree of automation;


the skill of the operator with respect to the
test method;
the skill of the operator with respect to the
component;
redundancy ( e g , two or four eyes testing);
the accessibility of the potential defect site;
and
the material of which the component is
made.

the applied NDE method (visual inspection,


ultrasonic, etc.);

The minimum residual lifetime of a cracked


structure as the basis of any further analysis

( i ) Establishing an initial crack shape and size

Fracture Mechanics in Design, Fitness-for-service Philosophies


depends essentially on the detection limit of
NDE as illustrated in Figure 2.
The basic approach for establishing the
initial crack size based on information relating
to the POD is illustrated schematically in
Figure 3. Each combination of the factors
listed above refers to one curve relating the
POD to the crack size. It is evident that a 100%
POD would refer to an unrealistically large
crack size. Therefore, the initial crack size has
to be defined for a smaller value of POD. PODa curves have to be determined empirically.
Because of the high effort involved in this kind
of investigation only a few curves are available.
Instead, the initial crack size is often based
arbitrarily on operator experience.
For reasons of practicality it is prudent not
to choose initial crack sizes on a case-by-case
basis but to establish reference values for
certain application fields. The guidelines have
to be slightly conservative, Le., they have to
overestimate the actual detection limit and the

NDE tester has to ensure that he will meet the


requirements under in-service conditions.
Up to here the crack size has been designated
uniformly as a. For embedded and surface
cracks it is, however, necessary to distinguish
between two dimensions (2D), the crack depth
in the wall thickness direction and the crack
length (for surface cracks at the outer surface).
In the following, these dimensions will be
referred to as a and c or 2a and 2c,
respectively. This is illustrated in Figure 4.
Whereas the smaller of a or c has to be based
on the NDE detection limit the other dimension has to be derived from the first by using a
representative ajc ratio of the initial crack. This
is necessary because the initial a/c ratio
significantly affects the residual lifetime of the
component.
When the initial crack size is selected, a
check has to be made to ensure that it is not
already critical (cf. (iii)). If it is already of
critical size, damage tolerant design is not
possible and another fracture mechanics based
design philosophy has to be chosen (see
Sections 7.01.2.2.2-7.01.2.2.5).

(ii) Simulation of crack extension

-I

Design Life

4-

I I
Time

Figure 2 Determination of acceptance criteria in


terms of a, for given design life. Curve 1 : a, is above
the limit of detection, uNDE. Curve 2: provides
residual life for a, = UNDE.

The initial crack can extend due to various


mechanisms: fatigue, stress corrosion, high
temperature creep, and various combinations
of these mechanisms.
Because the crack growth does not immediately cause failure of the component, it is
usually designated as subcritical crack extension. The following remarks are restricted to
fatigue crack extension. In Figure 5 the crack
extension rate (crack extension per load cycle)
is plotted versus the cyclic stress intensity
factor range AK ( = Kmax-Kmin,for definition
see Figure 6). The da/dN-AK curve consists of
three stages:

function

Assumed
step function
Defined initial
cracksize

c,xj,/

a0

Crack Size, a
(a)
Crack Size, a
(b)
Figure 3 (a) POD as a function of crack size and (b) establishing the initial crack size, a,, for a damage
tolerance analysis on the basis of a POD-a curve.

An Overview of Failure Assessment Methods in Codes and Standards

ence values for C and m is sometimes


possible within certain material classes.
Stage (3). Crack extension is strongly
accelerated. The component fails when the
value of K,,, approaches the fracture
resistance of the material, K,,,. As in the
curve for stage (l), crack extension is
strongly dependent on the stress ratio, R.
An important factor with respect to the load
at which final failure occurs is the crack tip
constraint, which is dependent on the specimen or component dimensions.
Figure 4 Crack dimensions for: (a) through-thickness cracks, (b) surface cracks, and (c) embedded
cracks.

The cyclic stress intensity factor range, AK,


affects the cyclic crack tip plasticity. It is a
function of the applied load or stress range, ACT,
the crack size, a, and the geometry given by a
shape factor, Y, according to

Stage ( I ) . Below a threshold value (AK,) of


AK, the crack is not able to extend. For
AK = A o m Y
values of AK just beyond AK, the curve is
strongly affected by the mean stress, om, and
For surface cracks, the crack extension has to
the stress ratio R ( = C T ~ ~ ~ / C T =
, ~ Kmin/Kmax).
,
be calculated simultaneously in the thickness
Other important factors are the environment
direction and along the surface.
(corrosive media) and microstructure.
Stage ( 2 ) . Crack extension can be described
by the Paris equation
da = C(AK)"'
dN

where C and m are constants. Factors such


, stress ratio R, and
as the mean stress C T ~the
the environment are of minor influence. Due
to the fact that the microstructure of the
material is also of moderate importance the
application of material independent refer-

Cyclic Stress Intensity Factor Range

Figure 5 Crack extension rate da/dN as a function


of the cyclic stress intensity factor range (see
Chapter 4.04).

omin

Time

Figure 6 Definition of the various K values for

fatigue loading.

Fracture Mechanics in Design, Fitness-for-service Philosophies


The Paris law (Equation (l), see Chapter
4.04)is restricted to stage (2) of the da/dN-K
curve. Its integration over the whole lifetime
would yield over-conservative predictions of
the crack extension in stage (1) but nonconservative predictions in stage (3). This is the
reason why numerous alternative equations
have been proposed instead of Equation (l),
which are able to describe two or even all three
sections of the curve in a more appropriate way
for specific applications.
So far the discussion has been restricted to the
relatively simple case of so-called long cracks
growing at constant load amplitude. In reality,
however, more complicated features such as
short cracks, real load spectra, and mixed
mode loading of the crack and superimposed
corrosion often have to be accounted for.
For a more comprehensive discussion of
fatigue crack extension, see Volume 4. Other
mechanisms of subcritical crack extension such
as creep or stress corrosion cracking are
described in Volumes 5 and 6.

(iii) Determination of the critical crack size of


the component
Subcritical crack extension is terminated by
the final failure of the component. Failure may
occur as brittle fracture or as unstable ductile
fracture. Critical states may, however, also be
defined by other events such as stable ductile
crack initiation or the breakthrough of a surface
crack in a wall. Additionally, failure modes
such as buckling or cavitation may occur which
are not the subject of a fracture mechanics
analysis. At a certain point in ligament yielding,
the stress peak at the crack tip is leveled off to a
degree which does not allow for application of
fracture mechanics crack tip parameters. Failure is controlled by the size of the net section
rather than by the local stress field ahead of the
crack tip. It occurs by plastic collapse. The
determination of end-of-life conditions is the
subject of Section 7.01.3.

defect should be detected before it reaches its


critical size. The fracture control plan has to be
based on the following assumptions:
(a) the component under consideration already contains a crack of size a, at the
beginning of its lifetime; and
(b) this crack has to be detected before it
reaches its critical size, a,.
It follows immediately that the time interval
between two inspections has to be smaller than
the time necessary to reach a,. This refers to a
time span tc-to. Taking into account that there
is a limited probability for a crack of a certain
size to be detected during an inspection (see
Figure 3), the inspection intervals, AZ, have to
be smaller than t,-to. Frequently, inspection
intervals are chosen as AZ=O.5 (tc-to), i.e., up
to t, a minimum of two inspections is required
(Figure 7) (Broek, 1988, 1996). The idea behind
this is that a crack will be detected prior to
failure even if it is missed in one inspection.
It is, however, obvious that even two or
more inspections cannot guarantee the crack
being detected because this would require a
100% overall POD. Nevertheless, increasing
the number of inspections or reducing the time
intervals between the inspections significantly
increases the POD.
The determination of the residual lifetime
might be subject to further uncertainty, especially when the component experiences a
spectrum of loading.
With respect to damage tolerant design the
effect of NDE is important for different
reasons:
(a) If the allowable crack size is relatively
small not all NDE techniques will be able to
detect it with a sufficiently high probability. In
those cases damage tolerant design is not
possible.
(b) For a required minimum detection
probability the inspection intervals can be

(iv) Determination ofthe residual lifetime ofthe


component
The residual lifetime is that time or number
of loading cycles which a crack needs in the
simulation for extending from the initial crack
size, a,, up to the allowable or critical crack
size a,/S with S being an additional safety
factor, the choice of which is up to the user.

( v ) Establishing inspection intervals


Establishing inspection intervals is the usual
aim of a damage tolerance analysis. A potential

1 NDE Techniaue V,

fa

(v*)

I/

fa & I )

tC

Time or number of load cycles

Figure 7 Establishing inspection intervals for two


different NDE techniques. The choice of two
inspections (AI) between to and t, is due to the
possibility that a crack may be missed at one
inspection.

10

An Overview of Failure Assessment Methods in Codes and Standards

adjusted by choosing a suitable NDE technique. This might be crucial within the economic
and organizational constraints in industrial
practice if, e.g., the inspection has to fit into
existing maintenance programs.
(c) The term NDE technique as used above
is defined as the practical realization of NDE
under service conditions. Besides the test
methods (visual inspection, ultrasonic, magnetic particle testing, etc.) questions of training
the testers, automation, and organizing the
inspections may play a major role.
Establishing inspection intervals is essentially
a statistical task acting as a link between the
fracture mechanics analysis, NDE, and the
constraints and requirements of industrial
practice. Therefore, within the overall concept
of damage tolerant design it holds a key
position. Although the empirical determination
of POD-a curves is often an expensive task,
spending this effort is worthwhile in many cases.
7.01.2.2.2 Fail-safe design
Damage tolerant design is implicitly based on
the assumption that some damage due to the
occurrence of cracks may be tolerated, but a
crack will cause catastrophic failure of the overall component after it has reached its critical
size. In contrast to this the aim of a fail-safe
design philosophy is to make sure that the
failure of a component will occur in a noncatastrophic way. In other words, the damage of the
component or the fracture of a substructure is
partial and locally restricted and will be detected prior to global failure. Sometimes fail-safe
analyses can be regarded as a second safety level
complementary to damage tolerance.
Important fail-safe concepts are: redundant
design, design for crack arrest see Chapter
7.09, and design for leak-before-break see
Chapter 7.10.
Many aspects of a damage tolerance analysis
may also be utilized in the context of a fail-safe
philosophy. Those points will not be repeated
here. Instead the aspects and facets in which
fail-safe differs from damage tolerance are
emphasized.

intervals in damage tolerant design, but the


checks are less expensive and time consuming.
Depending on the individual application, a
fail-safe inspection plan may be based on a
visual survey, regular function tests, etc.
The essential requirement is that partial
failure will be detected before it can cause
significant damage to the whole structure
(Figure 8). This aim requires some information
on inspection intervals, the accessibility of the
vulnerable sections of the structure, and the
ability of the NDE technique, etc. Starting
from the maximum possible damage at the
time of an inspection, the task of fracture
mechanics consists of the prediction of the
residual lifetime of the member of the structure
that is still intact up to global failure.
In principle, this can be done as described
for damage tolerant design (Figure 1) but with
the difference that instead of the N D E detection limit the maximum crack size in the intact
member of the structure at the inspection time
must be known. However, establishing the
latter under service conditions is a rather
sophisticated task:
(a) It is of relevance how the local damage
will be detected. Is it still a subcritical crack, or
is it the complete failure of a substructure at the
detection time? Is it detected immediately after
it occurred or is it detected by chance in a
regular inspection? If the latter is true the
damage will most probably have advanced since
the occurrence of the partial failure. Anyway,
the initial crack size for establishing the residual
lifetime has to be chosen conservatively such
that its detection is guaranteed under service
conditions. Self-evidently, this requires a caseby-case strategy ideally incorporating experience from tests and case histories.
(b) Frequently, the residual lifetime of the
intact member will be very limited. In such
cases the component must be repaired or
retired as soon as possible. If a limited
continued service time is required from a
practical point of view it has to be guaranteed
that it will not exceed the residual lifetime
established in the fracture mechanics analysis.
Otherwise, a redundant design philosophy
should not be adopted.

(i) Redundant design

In contrast to damage tolerant design, for


redundant design (multiple load paths), it is
not the detection of a crack, but the detection
of local damage or partial failure of one or
more substructures that is the key task. This
can be realized in different ways, e.g., by in situ
monitoring of the service operation or by
regular inspections. The interval of these
checks is usually much shorter than inspection

f ii) Design for crack arrest

Another fail-safe concept is design for crack


arrest. An extending crack may arrest if
(a) the actual crack driving force (CDF)
decreases with the extending crack, or if
(b) the crack advances into a region of
increased crack resistance.
As a rule, the decrease in CDF is due to
certain design features such as crack stoppers

11

Fracture Mechanics in Design, Fitnessfor-service Philosophies


After partial failure

Fracture mechanics
analysis

Fracture mechanics
analysis

Is the inital crack size


alreadv critical?

Is the extended
crack critical (a= a,)?

no

Determination of the residual


lifetime of the intact member

Intervention must be
possible under service
conditions

Is there enough residual lifetime


left in order for intervention in the system to be made?

The component is

When partial failure of a

Interfere as soon as possible!

Repair or retire the component!

Figure 8 Fracture mechanics analysis within the framework of redundant design philosophy.

or stress shielding. Thermal or residual stress


fields may lead to similar effects. In transient
loading cases a temporary stress peak may be
followed by a time-dependent decrease of the
load magnitude, arresting the extending crack
after a certain time.
Typical examples of regions of higher crack
resistance are strength mismatch zones such as
weldments with different strength and toughness levels of the base plate and weld metal.
Increased fracture resistance properties can
also be the consequence of stationary thermal
gradients when the crack extends into a region
of increased temperature.
In Figure 9 the CDF decreases when
the crack reaches a certain size. Up to that

Plastic Collapse

E-,

$4

Crack Initiation Toughness

Ip\

tt 1
Crack Size, a

Figure 9 Schematic illustration of crack arrest, e.g.,


at a crack stopper.

12

An Overview of Failure Assessment Methods in Codes and Standards

point it was rising. After the C D F exceeds the


materials resistance to crack initiation it may,
in this case, extend by a stable ductile tearing
mechanism. Without a crack stopper the stable
crack extension would lead to unstable ductile
crack extension or plastic collapse and failure
of the component. However, with a crack
stopper the local C D F is reduced and crack
extension is retarded. For a C D F smaller than
the crack arrest toughness, the stable crack
extension is temporarily stopped. Under displacement controlled loading conditions, crack
arrest may occur even subsequently to a certain
amount of cleavage fracture. The difference in
the levels of crack initiation and arrest toughness, as schematically illustrated in Figure 9, is
pronounced for rapid crack extension.
As also illustrated in the figure, the crack may
continue to extend under cyclic loading even at
load levels below crack arrest which means that
the structure is safe only for a limited time.
Then, intervention from outside is necessary to
stop the crack before it reaches its critical size.
A fracture mechanics based prediction of crack
extension after the arrest incident is necessary
for establishing the residual lifetime.
Besides retarding the crack extension the
crack stopper may also be responsible for crack
angle deflection or forcing the crack into a
mixed mode status. Another important question is whether or not the crack stopper itself
will be damaged. A detailed treatment of crack
arrest mechanisms is given in Chapter 7.09.
(iii) Design for leak-before-break
Design for leak-before-break (Figure 10) is a
special fracture mechanics approach for pressurized components, the aim of which is to
provide assurance that a crack breaking
through the wall does not immediately cause
global failure. As for damage tolerant design,
the starting point of the analysis is predefining
an initial crack. However, while the crack size
corresponds to the detection limit for damage
tolerance, for a leak-before-break analysis it
will be representative of the defect type caused
by the manufacturing process.
The crucial point is the initial crack shape,
Le., the initial ajc ratio of the surface crack
rather than the initial crack size. Information
about this may be obtained from failure
experience or from pre-investigations.
The primary aim of the subsequent crack
extension simulation is not to establish the
residual lifetime of the component up to
leakage, but to determine the development of
the crack shape until the surface crack breaks
through the wall. If crack extension is driven by
a fatigue mechanism the analysis will be done

for points A and B (see Figure 4) of the crack.


If, however, ductile tearing is an issue, constraint conditions have to be taken into account,
which may cause the surface crack to exhibit
what is called the canoe-effect, Le., deviation
from its original semielliptical shape. In those
cases the analysis becomes very sophisticated.
Breakthrough is assumed to occur when the
crack becomes critical at its deepest point
(point A, Figure 4) or when the remaining
ligament ahead of the crack is reduced below a
certain value. The subsequent recharacterization of the surface crack as a through thickness
crack has to be done, e.g., as illustrated in
Figure 11 (see, also, Chapter 7.06).
This recharacterization approach is slightly
different from those commonly applied to
surface cracks which usually aim to be conservative from a fracture mechanics viewpoint. In
the context of a leak-before-break analysis, the
recharacterization should be as realistic as possible. The reason is that the resulting crack opening area controls the leak rate of the medium. In
order to be conservative with respect to the
detectability of the leak, this area should be
underestimated rather than overestimated.
The actual aim of a leak-before-break analysis is to provide assurance that the time between
detecting a potential leak and the global failure
of the component is large enough to allow the
system to be shut down. For this purpose,
fracture mechanics based determination of the
residual lifetime following leak detection is
necessary. For a more comprehensive discussion of leak-before-break see Chapter 7.10.

7.01.2.2.3 Periodic proof testing


Periodic proof testing is another design
philosophy, which requires fracture mechanics
modeling. If a structure fails in a proof test
facility this happens in a controlled manner. If
it survives it is assumed to be safe under the
lower design load.
The proof test by itself may create or
enhance some damage, which can cause the
load carrying capacity to be decreased to well
below the proof test load and even below the
design load at a later time. This is the point
where fracture mechanics has to come into
action.
A critical crack size is determined according
to the proof test load. In Figure 12 this crack
size is designated as aco where the subscript o
stands for overload. The fracture mechanics
analysis is then performed as for damage
tolerant design, but with aco being used as the
initial crack size. Starting from aco, fatigue
crack extension is simulated under design load
conditions. The result from this analysis gives a

Fracture Mechanics in Design, Fitness-for-service Philosophies


Representative initial
crack
size and shape

13

Yes

Fracture mechanics
analysis

already critical?

Along both axes of


the surface crack
Fracture mechanics
analysis

locally critical (u=uc, 2c=2cc)?


CVeS

Fig. 1 I

Recharacterization as

Fracture mechanics
analysis

Is the recharacterized crack


(2c) globally critical?
I

Cno

Is the leak detectable in a


suitable manner?

In length direction
of the through crack
Fracture mechanics
analysis

no

Iterative simulation of

Is the extended crack


globally critical
(2c = 2CJ?

Determination of the residual


lifetime after detecting

Is the residual lifetime big


enough to interfere ?

leak-before-break
condition is given
Figure 10 Fracture mechanics analysis within the framework of leak-before-break analysis.

life prediction. However, this is applied in a


different way from damage tolerance. It is not
used to establish an inspection plan but for
establishing a proof test plan. The next proof
test has to be carried out in due time before the
crack becomes potentially unsafe. For a more
comprehensive discussion of handling proof
testing see Chapter 7.08.

7.01.2.2.4 Periodic removal of the crack


Figure 11 Leak-before-break analysis: recharacterization of the surface crack after breakthrough
(according to SINTAP, Anon., 1999). The subscript
b stands for breakthrough.

Periodic removal of the crack is an alternative


design PhilosoPhY in cases where the critical
crack size is potentially smaller than realistic
detection limits of common NDE techniques. A

14

An Overview of Failure Assessment Methods in Codes and Standards


Proof test in a
test facility

Repair or
retirement

Application of an overload

Alternative: lower proof


test load; in the design
state: change the design

Does the component fail?

Fracture mechanics
analysis

Determination of the critical


crack size, aco,for the overload

a,, = initial crack

Crack extension
under
service load
Fracture mechanics
analysis

size

k-

Iterative simulationof
crack extension

Is the extended crack


t
~~

Determination of residual
lifetime of the component

1
Establish the time for
the next proof test

Continue as
above

Figure 12 Fracture mechanics analysis within the framework of a proof test concept.

special problem might be the predetermination


of a representative (metallurgical) initial crack
with respect to its shape (a/c) as well as its size.
This information has to be provided experimentally. Because the initial crack size will
usually show a certain scatter band, statistical
processing will be necessary in many cases. As
in damage tolerance, the extension of the initial
crack will be simulated until the crack reaches
its critical size. The residual lifetime then defines
the time between periodic removals of the
crack, e.g., by grinding (Figure 13).
The fracture mechanics analysis has to be
performed for short cracks on the basis of
elastic-plastic crack tip parameters such as the
J-integral or the crack tip opening displacement (CTOD). Applying the linear elastic
stress intensity factor is not possible for this
class of crack.
7.01.2.2.5 Fracture mechanics based fatigue
limit
Fracture mechanics analyses can be applied
in order to determine crack initiation S-N

curves (see Chapter 4.01). In principle the


initial crack size can be adjusted such that the
simulated number of load cycles up to a crack
extension of 0.5 or l m m is identical to that
obtained in the S-N curve. Alternatively, if the
initial crack size is known, S-N curves can be
predicted not only for specimens but also for
component geometries. The advantage of a
fracture mechanics based S-N curve is that
parameters such as the component geometry
and the loading type can be treated individually; simply by applying the equations used in
the fracture mechanics approach. As in the
concept of periodic removal of the crack the
assessment has to be done for short cracks.
An example of a fracture mechanics application to initiation S-N curves is the
PJ-parameter (Vormwald and Seeger, 1991),
which may be interpreted as a normalized stress
amplitude (ordinate axis in the S-N curve):
(a, - a,$
p J = -AJdF
- 1.24
U
E

1.02

+ -d2
(a,

- acl)

(3)

Fracture Mechanics in Design, Fitness-for-service Philosophies

15

Representative crack size


and shape from the manufacturing process
1

Revise the design


I

Is the initial crack size


already critical?

ln0
crack extension

I
I
critical (a =a,)?

Determination of the residual


lifetime of the comDonent

1
no

Is removal of the
crack a, possible?

Take into account wall


thickness reduction due to
grinding.

Figure 13 Fracture mechanics analysis in combination with periodic removal of cracks

In Equation (3) E is Youngs modulus, AJeffis


the effective range of the J-integral (see Section
7.01.3) and the terms (T, and cCland E, and cC1
mark the upper stresses and strains due to the
applied load and the crack closure stresses and
strains, respectively. The quantity n designates
the cyclic strain hardening exponent.

7.01.2.3 Reliability and Safety Factors


The input parameters of a fracture mechanics
based design contain uncertainties, which have
to be considered in order to avoid significant
pitfalls in the analysis. In practice, variations
and scatter bands in the input parameters tend
to be of higher relevance than intrinsic inaccuracies in the fracture mechanics models.
All input parameters are concerned. With
respect to the critical crack size the scatter in
the fracture resistance of ferritic and bainitic
steels especially in the ductile-to-brittle transition regime is a very important factor, but
variations and uncertainties in the loads and
component dimensions may also play a role.
With respect to the residual lifetime of a
component, factors such as the scatter band
of the daldN-AK curve, the adequate predefinition of the initial crack size, environmental
effects, and crack extension retardation are
very important. If inspection intervals have to

be established, characteristic parameters of the


NDE techniques such as the detection limit
and the POD are of additional influence on the
results of the analyses.
In Table 1 some important input parameters
are summarized whose uncertainties and scatter bands are relevant for residual lifetime
analysis. A rough weighting of the factors is
attempted. However, which of the listed parameters is really the most important has to be
determined for each application. If the statistical aspects are not adequately taken into
account the result can be a nonconservative
assessment, Le., an overestimate of the real
load carrying capacity, or the critical crack
size, or an unrealistically high residual lifetime
of the component. Conversely the assessment
may be over-pessimistic leading to unnecessarily short inspection intervals, etc.
The problem of variations in input parameters and scatter bands can be approached in
different ways:
(i) Representative lower (or upper) bound
values are chosen in order to guarantee a
conservative result from the fracture mechanics
analysis.
(ii) The input parameters are introduced into
the analysis as statistical quantities, usually as
probability distributions. The outcome of the
analysis will also be a statistical quantity, e.g., a

16

An Overview of Failure Assessment Methods in Codes and Standards

Table 1 Influence of uncertainties and scatter bands of different input parameters on the assessment of the
residual lifetime of a component.
Input parameters

Effect
Moderate
Pronounced
Very pronounced

Material
Crack resistance (Klc,Ji, 6i,
J- or 6-R curve)
Resistance to subcritical crack
extension (daldN-AK curve)
Initial crack size

failure probability of the component, a probability function of residual lifetime, etc.


(iii) Based on predefined target failure probabilities, partial safety factors can be established statistically.
(iv) Safety factors (partial for individual
input parameters or groups of input parameters, or global, e.g., for the residual lifetime
of the component) can also be determined on a
deterministic basis whereby service experience
plays a major role. A certain arbitrariness
cannot be completely avoided.
In practical applications different combinations of (i)-(iv) above will usually be implemented, whereas lower bounds and
deterministic safety factors (i) and (iv) play
an important role in cases where the database
is too limited to perform a complete statistical
analysis. The penalty for this consists, however, in more or less conservative results.
In the following, the options (i) and (ii)
which require the input parameters as statistical distributions will be discussed in more
detail. Detailed discussion of margins in
fracture assessment is contained in Chapter
7.15. Here, the discussion will be limited to the
critical state of the component at the end of its
lifetime. The most straightforward but also
most time consuming type of analysis is based
on a Monte Carlo simulation. A large number
of random data points are generated on the
basis of the probability distributions of the
various input parameters. For each of these
data sets a fracture mechanics analysis is
carried out in a deterministic way. Finally,
the output data set is processed statistically
yielding a failure probability of the component.
More advanced, but approximate approaches
are the first- and second-order reliability
methods (FORM/SORM), respectively.
A crucial question is which target probability is really justifiable in engineering practice.
With respect to this a possible clue might be
provided by the recently published I S 0 standard 2394: General Principles on Reliability
for Structures, (International Organization
for Standardization (ISO), 1998). The docu-

Component
Load

Environment
Loading history

Fracture mechanics model

Model for component


failure.
Model for subcritical
crack extension.
Model for considering
retardation effects.

ment defines reliability indices, which refer to


certain target failure probabilities in the range
lO--lO-
dependent on the consequences of
failure and the relative costs of safety measures. In the British flaw assessment standard
BS 7910:1999 (British Standards Institution
(BSI), 2000), the relevant information is
adapted as summarized in Table 2.
An alternative approach to the principle of
reliability analysis as described above is the
determination of statistically based partial
safety factors (iii). For this the target failure
probabilities of Table 2 are used as input
parameters. The partial safety factors are
defined such that they lead to an overall failure
probability no larger than the predefined target
probability. For practical determination, each
input parameter is separately varied while the
remaining parameters are kept constant.
The limitation of this approach is that
different combinations of safety factors of
various input parameters may yield identical
overall probabilities. In order to limit this
ambiguity it is common practice to divide the
input parameters into two groups-load and
material resistance parameters-and to treat
them separately.
The advantage of the partial safety factor
approach as compared to the approach according to (ii) above is that the data are provided in
tabular format. The user can apply these
tabular data without the need to perform
individual reliability analyses. Tables of partial
safety factors are given in documents such as BS
7910 (BSI, ZOOO), SINTAP (Anon., 1999), API
579 (American Petroleum Institute (API),
2000), and WES 2805 (Japan Welding Engineering Society (WES), 1997). The tables refer
to the applied stress, the yield strength, the
crack size, and the fracture resistance of the
material. For each of these input parameters, a
partial safety factor can be determined based on
a given target probability for the risk of failure
and the coefficient of variation (COV) which is
defined as the ratio of the standard deviation
and the mean value of the applied probability
distribution. An example is given in Table 3.

Determination of the End-of life Condition

17

Table 2 Target failure probabilities depending on the potential consequences of component failure.
Potential consequences of
component failure

Moderate
Severe
Very severe

Redundant design

Nonredundant design

P f ~ 2 . x3 lo-
p f = 1.0 IO-^
pf =7.0 x 1 0 - ~

p f = 1.0 x 1 0 - ~
p f = 7.0 x 1 0 - ~
pf = 1.0 1 0 - ~

Table 3 Partial safety factors for the fracture resistance of the material in terms of a critical CTOD 6 ( y a is
the divider to the mean minus one standard deviation value of a Weibull distribution).

Toughness 6,,,

pf=2.3 x io-*

p f = 1.0 x 1 0 - ~

p f =7.0 x io-5

(C0V)s

Ya

Ya

Ya

YS

0.2
0.4

1.oo
1.oo

1.69
3.20

2.25
6.75

2.89
10.00

0.6

1.oo

p f = 1.0

10-~

8.00

Source: BS 7910:1999 (BSI, 2000).

The following example is provided. Given a


COV value of (COV)6=0.6 and a target
probability of Pf= 1.0 x loF3 (redundant design and severe consequences of component
failure or nonredundant design and moderate
consequences, see Table 2) the design value of
the crack resistance is obtained as
(4)

where 8 is the mean value and og is the


standard deviation of the critical CTOD. For a
more comprehensive discussion of this issue,
see Chapter 7.11.

7.01.3 DETERMINATION OF THE


END-OF-LIFE CONDITION
7.01.3.1 Failure Conditions
This volume deals with assessment of end-oflife conditions, Le., with the quantitative treatment of final failure; the significance of which
can be demonstrated by the following items:
(i) The final failure condition is one of several
parameters needed for the determination of the
residual life of a component, see Figure 14.
(ii) The final failure condition providesamong other items-the critical flaw size, a,,
Figure 15, which is the largest flaw, which is not
allowed to be missed in an NDE. The POD of a
crack increases with its size (see Figure 3), hence
a large critical crack size is highly desirable.
(iii) A further point concerns safety margins
(see Section 7.01.2.3).
If the maximum stress applied to a component, either a regularly occurring stress or a
stress due to an overload, is plotted versus an
actual or assumed crack size, Figure 15, then
its location in comparison with the failure line
representing the end of life condition-which
for simplicity is drawn in Figure 15 for linear

elastic fracture mechanics conditions-gives


the safety margins in terms of stress or applied
force and in terms of crack size.
The task of determining critical (or failure)
conditions resembles that used in classical
design: an applied parameter is compared
with the appropriate material resistance. The
equality of the stress intensity factor, K, the Jintegral, or the CTOD, 6, respectively, with the
appropriate material property Kmat, Jmat, or
amat, respectively, thus defines the failure
condition (see Chapter 2.03):
K = Kmat

(5a)

J = Jmat

(5b)

mat

(5c)

6=

For the applied parameters, designations


such as CDF, crack tip parameter, crack field
parameter, or loading parameter are used. In
this volume, the terms crack driving force or
driving force will be used exclusively. Here the
same dilemma as with classical design analyses
can be observed: the applied parameters and
the material properties are designated with the
same symbols. Therefore, in this volume the
symbols for the material properties have the
subscript mat.
Equation (5) describes a failure condition
based on the equality of the C D F with fracture
toughness, which is measured at zero crack
extension or a small amount of crack extension. If, however, material response is ductile
and crack extension during loading is taken
into account, credit can be taken from the
increase in fracture resistance with crack
extension as represented by the crack extension
resistance curve (R-curve). In that case the
CDF is calculated for a number of constant
forces as a function of stable crack extension,

18

An Overview of Failure Assessment Methods in Codes and Standards

Time

Figure 14 Fracture control plan with inspection intervals.

Figure 15 Service condition of component relative


to failure. K,,, means any suitable definition of a
critical stress intensity factor. For simplicity, the
fracture line has been drawn for the conditions of

linear elastic fracture mechanics.


Aa. The event of failure is then given not only
by Equation (5) but also when a tangency
condition is exceeded, defined as
dK - 3Kmat
da

da

(6a)

then CDF curves are calculated for constant


force values, Figure 16(a), and the tangency
condition in Equation (6) provides the maximum force (or stress) a cracked component
can sustain. This is the maximum in the forcedisplacement curve shown in Figure 16(b).
(ii) There are, however, also conditions
where loading occurs under deformation or
displacement control. In those cases, the CDF
is calculated for constant values of displacement, q (or constant strain, E ) , and the pattern
shown in Figure 17 is observed, where the
tangency condition cannot be obtained.
The achievement of instability conditions is
consequently much more difficult under displacement control than it is under force
control. The reason is that as the crack
advances the force acting on the component
drops, thus leading to a reduction of the CDF.
This is the reason that R-curve tests can be
conducted beyond maximum load as they are
generally performed under displacement control. The structure is still stable after the
maximum load in Figure 16(b).
Under any circumstances, an accurate treatment of Equations (5) and (6) requires accurate
knowledge of both the driving force and the
resistance of a material to crack extension,
either expressed as a point value of fracture
toughness or as an R-curve. This condition is
generally not fulfilled as will be discussed in the
following.

or

7.01.3.2 CDF Parameters


The actual instability behavior depends
largely on the type of loading:
(i) If the component is under force control,
which is equivalent to dead weight loading,

The linear elastic stress intensity factor can


be expressed in the form
K

=o

G Y ( a /W )

(7)

Determination of the End-of-life Condition

19

Displacement

(b)

Figure 16 Instability behavior under force control: (a) determination of instability and (b) force-

displacement diagram.

Crack Length
Figure 17 Determination of instability under displacement control.

where Y(a/W) is a dimensionless calibration


function which-apart from the crack size, a,
depends on the size (W is a characteristic size
parameter, frequently equal to the width) and
geometry of the component, and on the stress
distribution in the cross-section containing the
crack. Comprehensive compendia of K solutions (e.g., Murakami et al., 1987, 1992; Tada
et al., 2000) have been compiled, containing
solutions for countless practical cases. Linear
superposition allows the derivation of solutions
not found in the compendia. Furthermore, K

solutions can be easily generated using the


weight function method.
Whereas the stress intensity factor depends
only on the loading and geometrical parameters, the J-integral and the CTOD are also
affected by the stress-strain curve of the
material. Therefore, comprehensive compendia
of J and CTOD solutions would require
extensive finite element (FE) analyses for a
range of material tensile properties, an often
prohibitive expense. Therefore, the EPRI Handbook (Kumar et aE., 1981; Zahoor, 1989, 1990,
1991) representing a compendium for J and 6
solutions, contains only a very limited number
of solutions (for details see Section 7.01.4).
A further problem is related to the ability of
K, J , or 6 to accurately describe the stress and
strain fields near the crack tip. Detailed
analyses of the crack tip field (e.g. Brocks
and Schmitt, 1993; Shih et al., 1993; Hancock,
1992) demonstrate that a single parameter,
such as K, J , or 6 characterizes the field only
under very limited circumstances. Higher order
terms, such as the hydrostatic Q-stress and the
elastic T-stress are needed to account for
effects of variations in in-plane constraint.
Another aspect which deserves attention is
the 3D nature of the crack tip fields. The
analyses referred to above have exclusively

Next Page

An Overview of Failure Assessment Methods in Codes and Standards

20

been carried out for 2D cases, with plane strain


and plane stress as limiting conditions. In
practice, behavior will often lie between the
limiting extremes.
Thus, it has to be recognized that a description of the CDF by a single parameter can only
approximately represent the real CDF. Details
on the determination of the CDF are given in
Chapter 7.04.
7.01.3.3 Material Resistance to Crack
Extension
Fracture processes are highly dependent on
constraint (stress triaxiality); this is true for
fracture due to void extension and coalescence
as well as cleavage fracture. Traditionally, the
geometry of a specimen or component (which
includes its size) is regarded as the main source
of constraint variations. However, as Figure 18
demonstrates, there are a range of parameters
affecting constraint and hence fracture. A
particular point of interest is the observation
that constraint tends to fall rapidly when the
yield force, Fy, of a cracked structure is
approached. This is of importance in so far
as fracture properties are determined on specimens which are frequently much smaller than
the component to be assessed. Since for a given
fracture event, such as initiation of a crack
extension, the degree of plasticity is higher and
constraint is lower in smaller bodies, the
resistance to crack extension in the larger
component may be overestimated.

Figure 19 shows schematically the effect of


constraint on the R-curve, together with some
parameters affecting constraint. It is seen that
tensile loading and thin walls promote high
resistance to crack extension. Bearing in mind
the parameters depicted in Figure 18, it is
obvious that the resistance to crack extension is
a complicated function of several parameters.
This makes it impossible to predict the crack
extension properties in an actual component in
a general way within the framework of classical
single parameter fracture mechanics. This is
highlighted by Figure 20, which shows tests on
two types of specimens each of which was
made of two different materials. It is clearly
seen that the specimen geometry alone is not
sufficient to describe the constraint conditions
and hence the resistance to crack extension; it
seems that in the low strength material 2024FC, conditions of high constraint cannot be
achieved even in compact tension specimens,
which is due to the high degree of plasticity
during crack extension, see bottom right
diagram in Figure 18.
These effects demonstrate the transferability
problem in fracture mechanics. Figure 21
depicts the dilemma: the hatched area is
supposed to cover the whole range of crack
extension resistance for a given material. An
actual structural configuration may be located
anywhere within this area. However, standard
test methods are designed such that they
provide lower bound values of crack extension
resistance. The use of data from such specimens usually leads to conservative assessments
of structural behavior, which is satisfactory in
numerous applications of fracture mechanics.
However, there are cases where such an
assessment is unduly conservative, and it may
be beneficial to reduce existing large safety

Relative Crack Depth, ult


T

x
u

/-Foi-

F,IF,
Surface Crack

Tension ab io, Bending

u
0

Crack Front Angle, 0 900

FIFy

Figure 18 Effect of various parameters on triaxi-

ality (the ratio of mean stress to equivalent stress


close to the crack tip).

Crack Extension

Figure 19 Schematic illustration of constraint

effects on crack extension (R-curve) in structural


components, indicating the factors affecting the Rcurve.

S-ar putea să vă placă și