ARTICLE IN PRESS

International Journal of Pressure Vessels and Piping 83 (2006) 607–617

www.elsevier.com/locate/ijpvp

Structural integrity analysis of axially cracked pipelines using

conventional and constraint-modified failure assessment diagrams
Sebastian Cravero, Claudio Ruggieri
Department of Naval Architecture and Ocean Engineering, Av. Prof. Mello Moraes, 2231 (PNV–EPUSP), São Paulo SP 05508-030, Brazil
Received 5 January 2006; received in revised form 24 April 2006; accepted 25 April 2006

Abstract

This study explores applications of the failure assessment diagram (FAD) methodology to predict the failure behaviour for high
pressure pipelines with planar defects having different geometries (i.e., crack depth and crack length). One purpose of this investigation is
to assess the capability of FAD procedures in integrity analyses of high pressure pipelines with varying crack configurations. Another
purpose is to address the effectiveness of constraint-based FADs to predict burst pressure of low-constraint cracked pipelines. Full scale
burst testing of end-capped pipe specimens with axial surface flaws provide the data needed to compare the failure predictions derived
from the FAD procedures. The analyses reveal that the degree of agreement between predicted pressures and experimentally measured
values depends rather markedly on the crack size for the tested pipes. Moreover, the analyses also show a possible weak dependence of
the predicted pressures on the constraint-based correction scheme. Overall, the results validate the use of FAD-based methodologies for
defect assessments of axially cracked pipelines.
r 2006 Elsevier Ltd. All rights reserved.

Keywords: FAD procedure; Axial flaws; Burst pressure; Pipelines; Constraint; J–Q approach

1. Introduction
Fracture assessment procedures for pressurized compo-
nents play a key role in design, fabrication and safe
operation of pressure vessels, piping systems and storage
tanks. In particular, accurate predictions of the failure
pressure in damaged oil and gas pipelines remain essential
for the safety assessment of high pressure piping systems,
including onshore and offshore facilities. As the pipeline
infrastructure ages, robust procedures for integrity ana-
lyses become central to specifying critical flaw sizes which
enter directly into procedures for repair decisions and life-
extension programmes of in-service structural components.
Perhaps more importantly, these procedures must ensure
fail-safe operations which avoid costly leaks and ruptures
due to material failure to comply with the current stringent
environment-based regulations. Current codes and stan-
dards for oil and gas pipelines provide rules for welding,
inspection and testing of transmission pipelines (see, for
Corresponding author. Tel.: +55 11 30915350; fax: +55 11 3091 5717.
E-mail address: claudio.ruggieri@poli.usp.br (C. Ruggieri).
example, API 1104 [1], CSA Z662 [2]). While these codes
provide simplified acceptance criteria for fabrication
defects (such as slag inclusions and porosity in weldments)
based upon workmanship standards and fracture tough-
ness testing, they do not specifically address fitness-for-
service assessments of crack-like defects that form during
in-service operation.
Fracture mechanics-based approaches, also referred to
as engineering critical assessment procedures, provide a
means for constructing a correlation of crack size with
applied loading as measured by the linear elastic stress
intensity factor, K, or the elastic–plastic parameter defined
by the J-integral and its corresponding value of the crack
tip opening displacement, CTOD (see further details on
these fracture parameters in Anderson [3]). Further
developments in the engineering critical assessment meth-
odology include the effects of plasticity on crack tip
loading by adopting the concept of failure assessment
diagrams (FADs) to evaluate the severity of crack-like
flaws. A key feature of FAD-based approaches is the
introduction of a concise framework to explicitly address
the potential interaction between stress-controlled cleavage

0308-0161/$ - see front matter r 2006 Elsevier Ltd. All rights reserved.
doi:10.1016/j.ijpvp.2006.04.004
 ARTICLE IN PRESS
608 S. Cravero, C. Ruggieri / International Journal of Pressure Vessels and Piping 83 (2006) 607–617
fracture and plastic collapse to predict structural failure.
The methodology thus provides a highly effective, albeit
conservative, acceptance criterion for cracked structural
components which relates the operating conditions with a
critical applied load or critical crack size. Several flaw
assessment procedures based upon the FAD concept, such
as the R6 methodology [4], BS 7910 [5], SINTAP [6], API
579 [7] and ASME Code Section XI [8], among others,
are now well established and widely employed to analyze
the significance of defects in terms of assessment of
structural integrity.
The FAD methodology defines a two-criteria assessment
curve which incorporates a coupling relationship between
crack-tip loading (K, J or CTOD) describing the fracture
conditions and a limit-load solution describing plastic
collapse of the remaining crack ligament. The key to this
approach lies in the use of fracture toughness data
measured from deeply cracked specimens tested under
bend loading to guarantee high levels of stress triaxiality
which drive the fracture process. Under such conditions, a
single failure locus then suffices to provide geometry-
independent predictions. However, structural defects in
pressurized piping systems are very often surface cracks
that form during fabrication or during in-service operation
(e.g., blunt corrosion, slag and nonmetallic inclusions, weld
cracks, dents at weld seams, etc.) [9–11]. These crack
configurations generally develop low levels of crack-tip
stress triaxiality (associated with the predominant tensile
loading which develops in pressurized piping systems)
thereby contrasting sharply to conditions present in deeply
cracked specimens under bending. Moreover, high grade
pipeline steels currently used exhibit much higher cleavage
fracture resistance compared to older, lower grade pipeline
steels; under increased loading, these materials develop
extensive plastic deformation at the crack tip prior to
fracture. Consequently, assessments of defects in low
constraint structural components based upon conventional
FAD equations may be unduly conservative and overly
pessimistic. While such conservatism represents an extra
factor of safety, excessive pessimism in defect assessments
can lead to unwarranted repairs or replacement of in-
service pipelines at great operational costs.
The technological importance of fracture behaviour for
low-constraint cracked structures prompted the develop-
ment of more refined defect assessment procedures capable
of including effects of constraint variations on cleavage
fracture toughness. These approaches advocate the use of
geometry dependent fracture toughness values so that
crack-tip constraint in the test specimen closely matches the
crack-tip constraint for the structural component. In
particular, Ainsworth and O’Dowd [12] and Ainsworth
[13] proposed a constraint-based correction to the FAD
procedure which reflects the strong role of constraint on
correlations of cleavage toughness data for varying crack
configurations and loading modes (tension vs. bending).
The approach builds upon the constraint-based Q metho-
dology [14,15] to correct measured toughness values using
low constraint fracture specimens thereby modifying the
shape of the FAD assessment line. The predictive
procedure for defect assessments thus becomes a function
of structural constraint which should remove or alleviate
the inherent conservatism of the FAD philosophy.
This study explores applications of the FAD methodol-
ogy to predict the failure pressure for high pressure
pipelines with planar defects having different geometries
(i.e., crack depth and crack length). One purpose of this
investigation is to assess the capability of FAD procedures
in integrity analyses of high pressure pipelines with varying
crack configurations. Specifically, the present work com-
pares the burst pressure predictions for two widely used
FAD procedures: BS 7910 [5] and API 579 [7]. Another
purpose is to address the effectiveness of the Q-based
correction for the influence of constraint on the FAD
curves and, consequently, on failure predictions. Full scale
burst testing of end-capped pipe specimens with axial
surface flaws provides the data needed to compare the
failure predictions derived from the utilized FAD proce-
dures. The analyses reveal that the degree of agreement
between predicted pressures and experimentally measured
values for both FAD procedures depends rather markedly
on the crack size for the tested pipes. Our exploratory
application presented here provides a representative set of
results which provide further support for using the FAD
methodology in defect assessments of pressurized pipes
with axial flaws.
2. Overview of the FAD methodology
It is widely recognized that brittle fracture and plastic
collapse caused by overloading are competing failure
modes in cracked structural components made of materials
with sufficient toughness. Early work by Dowling and
Townley [16] and Harrison et al. [17] to address the
potential interaction between fracture and plastic collapse
introduced the concept of a two-criteria failure assessment
diagram (most often referred to as FAD) to describe the
mechanical integrity of flawed components. In the FAD
methodology, a roughly geometry and material indepen-
dent failure line is constructed based upon a relationship
between the normalized crack-tip loading, Kr, and the
normalized applied (remote) loading, Lr, in the form
K r ¼ f ðLr Þ, (1)
where
K I ðP; aÞ
Kr ¼ (2)
K mat
and
P
Lr ¼ . (3)
PL ða; sys Þ
Here P is the applied (remote) load, a is the crack size, KI
is the elastic stress intensity factor, Kmat is the material’s
fracture toughness, sys is the yield stress and PL is the value
 ARTICLE IN PRESS
S. Cravero, C. Ruggieri / International Journal of Pressure Vessels and Piping 83 (2006) 607–617
of P corresponding to plastic collapse of the cracked
component. Alternatively, the parameter Lr can be defined
in terms of a reference stress, sref, defining the plastic
collapse load solution of the remaining crack ligament as
sref
Lr ¼ . (4)
sys
Current defect assessment procedures based on Eq. (1)
include the R6 Methodology [4], BS 7910 [5], the SINTAP
procedure [6], API 579 [7] and ASME Code Section XI [8]
among others. Structural integrity assessment of a cracked
component is based on the relative location of the
assessment point with respect to the FAD curve defined
by each procedure. The component is simply considered
safe if the assessment point (Kr,Lr) lies below the FAD
line whereas it is considered potentially unsafe if the
assessment point (Kr,Lr) lies on or above the FAD curve.
An increased load or larger crack will move the assessment
point along the loading path towards the failure line. Fig. 1
provides a schematic illustration of the FAD methodology
on a Kr vs. Lr plot.
The present work focuses on integrity assessments of
axially cracked pipelines in terms of the BS 7910 [5] and
API 579 [7] procedures. Both methodologies have emerged
as engineering codes widely used for defect assessments
which incorporate a three-tiered (or three-level) FAD
criterion with increasing analytical sophistication and data
requirements and decreasing conservatism. The normal
crack-flaw assessment in BS 7910 is Level 2A and in API
579 is Level 2 which utilize the following FAD expression
f ðLr Þ ¼ ½1
0:14ðLr Þ2 
f0:3 þ 0:7 exp½
0:65ðLr Þ6 g; Lr pLmax
r , ð5Þ
where the cut-off parameter, Lmax
r , is most often defined in
terms of the flow stress, sf ¼ ðsu þ sys Þ=2, in the form


max 1 su þ sys
Lr ¼ , (6)
2 sys
Kr = KI / Kmat
FAD Assessment Line
Safe Region
Potentially Unsafe
Region
Loading Path
(Increasing Crack Size or Loading)
Lr =
ref /
ys
Fig. 1. Schematic illustration of the FAD methodology.
609
where su denotes the material’s tensile strength. The cut-off
parameter defined by Eq. (6) reflects a simplified criterion
to prevent localized plastic collapse while, at the same time,
taking account of the hardening effect above yielding
observed in ferritic carbon steels and austenitic stainless
steels.
While both defect assessment schemes previously de-
scribed essentially share a common basis, there are
differences which can have important practical implica-
tions on failure predictions. In particular, API 579 provides
new and more accurate KI solutions for cracked cylinders
and spheres with varying crack face loading. Moreover, the
reference stress solutions of API 579 also include addi-
tional refinements to assess plastic collapse of cracked
cylinders and spheres which produce more consistent
results with experimental data. Section 5 describes defect
assessments for high pressure pipelines with axial flaws
using the FAD methodology and compares the burst
pressure predictions based upon the BS 7910 and API 579
procedures.
3. Constraint-modified FAD curves
3.1. Description of constraint using the Q parameter
Fracture testing of ferritic structural steels in the ductile-
to-brittle transition region consistently reveals a significant
effect of specimen geometry and loading mode (bending vs.
tension) on fracture toughness values (see [18–21] for
illustrative data). These studies show significant elevations
in the fracture resistance for single edge notched specimens
under bending (SE(B) specimens) with shallow cracks and
tension geometries of ferritic steels tested in the transition
region, where transgranular cleavage triggers macroscopic
fracture. At increased loads in a finite body, such as a
cracked specimen or a structural component, the crack-tip
plastic zones increasingly merge with the global bending
plasticity on the nearby traction free boundaries which
relax the near-tip stresses well below the values corre-
sponding to small scale yielding (SSY) conditions. This
phenomenon, often termed loss of constraint [22–24],
contributes to the apparent increased toughness of shallow
cracked and tension loaded geometries observed in fracture
testing. Once high triaxiality conditions no longer apply,
larger crack driving forces in the finite body are necessary
to generate a highly stressed region ahead of the crack tip
sufficient to trigger cleavage. These features have enormous
practical implications in defect assessment procedures,
particularly repair decisions and life extension programmes
of in-service structures as well as structural design
specifications.
The above arguments motivated O’Dowd and Shih (OS)
[14,15] to propose an approximate two-parameter descrip-
tion for the elastic-plastic crack tip fields based upon a
triaxiality parameter more applicable under large scale
yielding (LSY) conditions for materials with elastic-plastic
response described by a power hardening law given by
 ARTICLE IN PRESS
610 S. Cravero, C. Ruggieri / International Journal of Pressure Vessels and Piping 83 (2006) 607–617

=
ys / ðs=sys Þn . Here, n denotes the strain hardening
exponent, sys and eys are the (reference) yield stress and
strain, respectively. Guided by detailed numerical analyses,
OS identified a family of difference fields, sdiff, in the form
sij ¼ ðsij ÞSSY þ ðsij Þdiff


r O’Dowd [12] and Ainsworth [13]. The failure line is then
¼ ðsij ÞSSY þ f ij ; y; Q ,
J=sys
where the dimensionless second parameter Q defines the
amount by which sij in fracture specimens differ from the
adopted reference SSY solution, which is usually asso-
ciated with an infinite cracked plate. Here r and y are polar
coordinates centred at the crack tip with y ¼ 0 correspond-
ing to a line ahead of the crack.
Limiting attention to the forward sector ahead of the
crack tip between the reference (SSY) and the fracture
specimen fields, OS showed that Qsys corresponds effec-
tively to a spatially uniform hydrostatic stress, i.e., the
difference field relative to the high triaxiality reference
stress state. For Mode I fracture (which is associated with
the opening stresses, syy), OS defined Q as
syy
ðsyy ÞSSY
Q at r ¼ 2J=sys ,
sys
where the difference field is evaluated at the microscale
distance r ¼ 2J=sys which represents the location of the
triggering cleavage mechanism ahead of the crack tip.
Further details on the J–Q methodology can be found in
Refs. [22–24].
3.2. Modified FAD
A key feature of the FAD methodology previously
described lies in the adoption of a geometry independent
failure line to provide a single failure locus applicable in
general integrity assessments. However, as already ob-
served, crack-like defects in structural components are very
often surface cracks that form during fabrication or during
in-service operation. These crack configurations generally
develop low levels of crack-tip constraint which contrasts
sharply to conditions present in deeply cracked specimens
upon which the failure line given by Eq. (5) is based.
Constraint effects can be accommodated in the FAD
methodology by making the failure assessment curve a
function of structural constraint. Following the work of
Ainsworth and O’Dowd [12] and Ainsworth [13], these
effects are incorporated into the FAD approach by
quantifying constraint in terms of the load ratio, Lr. The
load dependence of constraint, here characterized by the
hydrostatic parameter Q previously defined, is then
expressed by
Q ¼ bQ Lr ,
where the parameter bQ is a function of geometry, crack
size, material properties (particularly strain hardening
exponent, n) and load.
When constraint is lost, a constraint-dependent tough-
ness, K Q
mat , can be related to structural constraint by
KQ m
mat ¼ K mat ½1 þ að
bQ Lr Þ , (10)
where a and m are constants which describe the elevation of
toughness with constraint loss as defined by Ainsworth and
ð7Þ
written as
K r ¼ f ðLr Þ½1 þ að
bQ Lr Þm . (11)
This modified failure equation is equivalent to retaining
the definition of Kr given by Eq. (2) in terms of a fracture
toughness derived from a highly constrained geometry,
but with the failure assessment curve, f(Lr) as defined by
Eq. (5), being corrected by a constraint-based factor
expressed by [1+a(
bQLr)m].
4. Experimental programme
To investigate the failure behaviour of axially flawed
pipelines, a series of full scale burst tests was performed on
end-capped, seam-welded pipe specimens with external
diameter, De ¼ 508 mm (20 in), wall thickness, t ¼ 15:8 mm
(8)
and length, L ¼ 3 m at room temperature [25]. These
experimental tests are part of a pipeline integrity pro-
gramme conducted by the Brazilian State Oil Company
(Petrobrás). The material is an API 5L Grade X60 pipeline
steel with 483 MPa yield stress at room temperature and
moderate hardening properties (su =sys ¼ 1:24). Rectangu-
lar tensile specimens (ASTM A370) with 13 mm thickness
were extracted from the circumferential orientation of the
pipe (which is also the transverse direction of the steel
plate) to provide the mechanical properties [27]. Table 1
summarizes the mechanical properties obtained from these
tests. Other mechanical properties for the material include
Young’s modulus, E ¼ 210 GPa and Poisson’s ratio,
n ¼ 0:3.
A set of 15 Charpy-V notch impact specimens was also
extracted in the TL plate orientation [27]. This set was
tested following the requirements of the ASTM E23
standard. The test pieces were broken at six different
temperatures:
60,
40,
20, 0, and 25 1C, to measure the
absorbed energy. At room temperature this material
presents fully ductile fracture with a Charpy impact energy
of E135 J. This value is selected as the Charpy-V notch
energy to correlate with a lower bound value to the upper
shelf fracture toughness, KIc. Estimation of the fracture
toughness for the tested material at the temperature of the
Table 1
Mechanical properties of tested API 5L X60 steel at room temperature
[27]
(9) sys (MPa) su (MPa) et (%) su/sys
483 597 29 1.24
sys: 0.2% proof stress; su: ultimate tensile strength et: uniform elongation
(gauge length ¼ 50 mm).
 ARTICLE IN PRESS
S. Cravero, C. Ruggieri / International Journal of Pressure Vessels and Piping 83 (2006) 607–617
integrity assessment followspfromffiffiffiffi Appendix F of API 579
[7] yielding K Ic  177 MPa m which is then taken as the
Kmat value to be employed in the integrity assessments for
the tested pipes describedpffiffiffiffi next. The adopted toughness
value of K Ic ¼ 177 MPa m falls within common values of
typical pressure vessel and pipeline steels. For example,pffiffiffiffi
Appendix F suggests adopting KIc -values of 110 MPa m
for steels with unknown chemistry (which is clearly a very
pffiffiffiffi
conservative value for structural steels) and 220 MPa m
for low sulphur steels (such as API 5L pipeline steels).
Further, using the Charpy correlation curve
pffiffiffiffi from Annex J
of BS 7910 [5] results in K Ic ¼ 150 MPa m. Consequently,
the KIc value adopted in our work seems fairly representa-
tive of a lower bound estimate of the material’s toughness
value.
Testing of the pipe specimens included both internal and
external longitudinal notches with different sizes measured
by notch depth and notch length, a  2c: (1) 3  60 mm, (2)
7  140 mm and (3) 10  200 mm. To avoid any effect of
the seam welding, the notches were machined at a 1801
position with respect to the weld. The present work focuses
on failure assessments for the tested pipes with external
cracks which are schematically illustrated in Fig. 2. The
pipe specimens were notched along their length using an
electrical discharge machine to create the required notch
shape. While the initial semi-elliptical defects were not
subjected to a pressure cycle to propagate a fatigue crack
from the original notch, the highly accurate machining
process allows considering them as initially blunted cracks.
Table 2 provides the burst pressures experimentally
measured in the tests, denoted PB
exp. During loading of
the pipes, ductile crack extension was monitored by using
an ultrasonic pulse technique to measure the crack
growth at the deepest point of the surface defect [26].
These measurements revealed small amounts of ductile
tearing (DaE0.5–1.0 mm) prior to pipe failure. Such
behaviour indicates intense plastic straining directly
ahead of the crack tip thereby resulting in pipe failure
controlled by plastic collapse rather than unstable fracture.
Within the present context, however, inclusion of ductile
tearing effects into the FAD assessment procedure is not
a does not alter significantly the stress fields ahead of the
t crack tip. Again, this approximation appears satisfactory
2c
(b)
De
(a)
Fig. 2. Schematic illustration of the pipe specimen with external surface
crack employed in the burst tests.
611
Table 2
Comparison of measured and predicted burst pressures for the pipe
specimens based upon BS 7910 Level 2A and API 579 Level 2 procedures
Pipe specimen PB-exp BS 7910 L2A API 579 L2
(MPa) PB-pred (MPa) PB-pred (MPa)
3  60 31.5 28.0 32.1
7  140 25.0 19.2 24.0
10  200 21.0 11.1 18.2
pursued. This issue is addressed in more detail in the next
Section 5.
5. Predictions of burst pressure using the FAD procedure
Fracture assessments for the tested pipe specimens with
external cracks follow the FAD methodology outlined in
Section 2. The integrity analyses described here compare
burst pressure predictions, hereafter denoted as PB-pred,
based upon the FAD curve given by Eq. (5) derived from
using the BS 7910 Level 2A [5] and API 579 Level 2 [7]
procedures with Lmax r E1.12 (see Eq. (6)). As already
mentioned, experimental measurements conducted during
the test programme revealed small amounts of ductile
tearing prior to pipe failure. To predict the burst pressure
for the tested pipes given the present context, the effect of
ductile crack extension on pipe failure is not considered
in this exploratory study. Rather, we pursue a line of
investigation which involves direct application of
the normal assessment levels in BS 7910 and API 579 to
position the assessment point for cleavage fracture (Kr)
and plastic collapse (Lr) on the FAD diagram. Since
the primary interest lies in examining the predictive
capability of both procedures within a simplified assess-
ment framework, this approximation proves satisfactory
to define failure pressure values for the tested pipes
referenced to a common basis. Moreover, the constraint-
based FAD curve using the Q-parameter is also con-
sidered unaffected by the small amount of ductile crack
extension observed in the tests. Here, the J–Q approach
can still be considered valid provided ductile tearing
given the measured values of Da in the range of
0.5–1.0 mm.
To provide a consistent basis of comparison for both
methodologies, the safety factors which would be
applied to flaw dimensions, acting stresses and fracture
toughness are set to unity. Because the testing programme
enables an accurate control of these parameters (particu-
larly the flaw geometry and loading), such approximations
should introduce no major effect on the integrity assess-
ments. The present section describes essential features of
the prediction scheme applied to determine PB-pred for both
procedures (full details may be found in [5,7]).
 ARTICLE IN PRESS
612 S. Cravero, C. Ruggieri / International Journal of Pressure Vessels and Piping 83 (2006) 607–617
5.1. BS 7910
Consider first predictions of burst pressure for the tested
pipes based upon the BS 7910 procedure. Key steps to
determine (Lr,Kr) at failure include evaluation of the stress
intensity factor and reference stress based upon the
primary load for a pressurized cylinder containing an axial
surface flaw. Following Annex M of BS 7910, the stress
intensity factor solution employed in the present study is
given by
pffiffiffiffiffiffi
K I ¼ Y s0 pa, (12)
where a is the crack depth, s0 is the applied primary stress
(due to membrane loading on the crack faces) and the
factor Y is defined as
Y ¼ Mf w M m ,
where M is the bulging correction factor (which is also
known as the Folias factor for thin-walled cylinders [28]),
Mm is the stress magnification factor and fw is the finite
width correction. Annex M of BS 7910 provides analytical
expressions for M, Mm, and fw as functions of a/t and c/a.
Now, following Annex P of BS 7910, the reference stress,
sref, for a surface flaw oriented axially in a cylinder is
given by
sref ¼ 1:2M s s0 ,
where Ms is again the Folias bulging factor defined as
1
½a=ðtM t Þ
Ms ¼ , (15)
1
ða=tÞ
with parameter Mt expressed by
sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi

2
c configuration on the Kr vs. Lr plot. Within the present
M t ¼ 1 þ 1:6 , (16)
Ri t
where Ri is the internal radius of the pipe.
5.2. API 579
Consider now application of API 579 to predict the burst
pressure for the tested pipes. The key steps to determine
(Lr,Kr) at failure remain essentially similar but with new
stress intensity factor solutions and somewhat improved
expressions for the reference stress. By adopting a uniform
crack face loading, the stress intensity factor has the form
rffiffiffiffiffiffi
pa
K I ¼ s0 G 0 , (17)
x
where G0 is the influence coefficient (which also represents
a nondimensional stress intensity factor) and x is the flaw
shape parameter defined for cXa in the form
a1:65
x ¼ 1 þ 1:464 . (18)
c
The influence coefficient, G0, in Eq. (17) is expressed as
sixth-order fitting polynomials which have been derived
from finite element analyses for cracked cylinders with
varying parameters Ri/t, a/t and c/a. Appendix C of API
579 outlines the procedure to determine G0 and provides
the fitting polynomials in tabular form.
Now, a formulation similar to the previous Eq. (14)
applies to determine the reference stress which derives from
Appendix D of API 579 as follows
sref ¼ M s s0 , (19)
where the Folias bulging factor, Ms, is now given by
1
Ms ¼ , (20)
1
ða=tÞ þ ½a=ðtM t Þ
with the parameter Mt expressed by

0:5
1:02 þ 0:4411l2 þ 0:006124l4
Mt ¼ , (21)
1:0 þ 0:02642l2 þ 1:533  10
6 l4
(13)
where the shell parameter l is defined as
1:818c
l ¼ pffiffiffiffiffiffiffi . (22)
Ri a
The above expressions to determine the reference stress
for a surface flaw oriented axially in a cylinder are
associated with a net section collapse solution. Conse-
quently, they provide slightly different results than the
reference stress solution given by BS 7910 previously
described which is based on a local ligament criterion.
(14)
5.3. Failure pressure predictions
Assessments of structural integrity for the tested pipe
specimens with varying crack geometries follow from
constructing the loading path (which derives from the
procedure outlined in the previous sections) for each crack
context, the intersection of the loading path with the failure
line described by Eq. (5) defines the predicted failure
pressure for the analyzed cracked pipes.
Figs. 3 and 4 display the FAD curves and analysis
predictions from the BS 7910 and API 579 procedures and
for each of the pipe specimens. The potential effect of the
parameter Lmax
r is also included in the plots. Here, we note
Fig. 3. Prediction of burst pressure for the tested pipe specimens based
upon BS 7910 Level 2A.
 ARTICLE IN PRESS
S. Cravero, C. Ruggieri / International Journal of Pressure Vessels and Piping 83 (2006) 607–617
Fig. 4. Prediction of burst pressure for the tested pipe specimens based
upon API 579 Level 2A.
that Lmax
r ¼ 1:25 corresponds to a typical cut-off value for
common structural steels. Table 2 compares the predicted
burst pressures for both procedures with the experimentally
measured failure pressures; here the predicted values are
marked as open circles in the plots.
While the integrity analysis conducted here demonstrates
the effectiveness of the FAD methodology in defect
assessments, predictions for the BS 7910 procedure show
excessive conservatism for the deeply cracked pipe speci-
men with a 10  200 mm crack (and perhaps also for the
specimen with a 7  140 mm crack). In contrast, the
predicted results from the API 579 procedure for these
specimens display much better agreement with the experi-
mental data.
The analyses for the shallow crack pipe specimen
(3  60 mm) reveal that both procedures provide predic-
tions of the failure pressure within a small (relative)
deviation from the experimental value. The API 579
procedure slightly overestimates the burst pressure for this
case with obvious potential implications on the safety
margin. However, we emphasize that the failure mode here
is primarily due to overload fracture of the remaining crack
ligament (note the slope of the loading path for this
specimen on the Kr vs. Lr plot displayed in Figs. 3 and 4
and its intersection with the failure line). Consequently,
predictions of the failure pressure for such analyses are
strongly dependent upon the value assigned to the cut-off
parameter, Lmax
r . Indeed, decreased values for Lmax
r yield
reduced values of PB-pred when plastic collapse of the
remaining crack ligament is the controlling failure mode.
These observations suggest that an alternative definition
for the flow stress may be required to modify the current
definition of Lmaxr given by Eq. (6). Nevertheless, the
overall deviations for this case appear sufficiently small to
support use of the FAD methodology for this range of
crack sizes.
Moreover, further examination to investigate the source
of the more conservative predictions obtained by the BS
7910 procedure revealed that both solutions for the stress
intensity factor, given by Eq. (12), and for the reference
613
stress, defined by Eq. (14), provide higher values at the
same applied pressure than the corresponding values
derived from the API 579 procedure. Consider first the
deeply cracked pipe. At P ¼ 21 MPa (which corresponds to
the burst pressure value for this specimen), the stress
intensity factor and the reference
pffiffiffiffi stress obtained from BS
7910 are K BS7910
I ¼ 240 MPa m and s BS7910
ref ¼ 773 MPa
whereas for the API pffiffiffiffi 579 procedure these values are
K API579
I ¼ 125 MPa m and s API579
ref ¼ 525 MPa. Con-
sider now the shallow cracked pipe. At P ¼ 31:5 MPa
(which is the burst pressure value for this pipe configura-
tion), p the 7910 procedure provides K BS7910
ffiffiffiffi BS BS7910 I ¼
59 MPa m and sref ¼ 609 MPa whereas for p theffiffiffiffi API
579 procedure these values are K API579I ¼ 56 MPa m and
sAPI579
ref ¼ 504 MPa.
6. Predictions of burst pressure based upon a constraint-
based FAD procedure
6.1. Load dependence of constraint for SE(T) specimens
Predictions of burst pressure for the tested pipe speci-
mens based upon the modified FAD methodology require
evaluation of the load dependence of constraint for the
analyzed structural component as described by Eq. (9). For
pressurized pipelines and cylindrical vessels, Cravero and
Ruggieri [29] advocate the use of geometry dependent
fracture toughness values measured using single edge notch
tension (SE(T)) specimens. Under increased remote loading
(as measured by J), Cravero and Ruggieri also demon-
strated strong similarities between J–Q trajectories for pin-
loaded SE(T) fracture specimens and axially cracked pipes
with common crack sizes relative to pipe wall thickness and
specimen width. At similar values of the continuum, scalar
parameters (J, Q), the crack-tip strain–stress fields which
drive the local process have similar values as well.
Consequently, an SE(T) specimen and a crack-like surface
flaw in a pipeline with similar J–Q trajectories exhibit
similar conditions for cleavage fracture. Such a feature is
central to the choice of SE(T) fracture specimens in
deriving a constraint-based FAD curve which is applicable
to an axially cracked pipeline.
The dependence of load on crack-tip constraint for the
SE(T) specimens is then obtained through numerical
computation of the evolution of loading with the hydro-
static Q parameter as given by Eq. (8). Nonlinear finite
element analyses are performed on plane-strain models for
pin-loaded 1
T SE(T) specimens with a=W ¼ 0:2, 0.5 and
0.65 with H=W ¼ 6. Here, a is the crack size specimen, W
is the specimen width and H is the distance between the pin
loads. The crack size to specimen width (a/W) ratios for
these specimens match the crack depth to wall thickness
(a/t) ratios for the precracked pipe specimens. Fig. 5(a)
shows the geometry and specimen dimensions for the
analyzed crack configurations.
The numerical computations for the fracture specimens
and cracked pipelines reported here are generated using the
 ARTICLE IN PRESS
614 S. Cravero, C. Ruggieri / International Journal of Pressure Vessels and Piping 83 (2006) 607–617

a
H

H /2

Fig. 6. Evolution of crack-tip constraint for the SE(T) specimens in terms

W of Q with increased normalized loading, Lr.

and
PL h a  a 2  a 3  a
¼ 1:16 1

1:23 þ ; p0:55.
a
sys W W W W W
(25)
W

(a) (b)
In the above expressions, the plastic collapse load, PL,
derives from a plane-strain solution using the von Mises
Fig. 5. (a) Geometry of the analyzed pin-loaded SE(T) specimens; (b) yielding criterion.
plane-strain finite element model used in the analyses of the SE(T)
In all plots, Q is defined by Eq. (8) at the normalized
specimen with a=W ¼ 0:5.
distance ahead of the crack tip given by r ¼ 2J=sys (note
that we plot Lr vs.
Q to maintain positive scales). The
research code WARP3D [30]. The analyses utilize an research code JQCRACK [32] is employed to compute the
elastic–plastic constitutive model with J2 flow (incremental) evolution of Q with increased loading for each fracture
theory and conventional Mises plasticity in a small geometry specimen.
change setting. The analyses utilize a piecewise-linear
approximation of the measured engineering stress–strain
curve for the tested APIX60 steel [27]. A typical plane-strain 6.2. Modified FAD
finite element model employed in the analyzes of the SE(T)
specimens has 1500 elements and 3200 nodes. Fig. 5(b) shows To verify the predictive capability of the constraint-
the half-symmetric model employed in the finite element based approach adopted in the present work, this section
analyses for the SE(T) specimens with a=W ¼ 0:5. Cravero describes application of the modified FAD methodology
and Ruggieri [29] also provide additional details on the finite using the Q-based correction to predict the measured burst
element models constructed for the plane-strain analyses of pressure for the tested precracked pipes with external
the pin-loaded SE(T) specimens. cracks. Figs. 7–9 provides the loading path for the pipe
Fig. 6 shows the evolution of crack-tip constraint for the specimens with 10  200, 7  140 and 3  60 mm crack
SE(T) specimens in terms of Q with increased normalized configurations on modified Kr vs. Lr plots. The solid lines
loading, Lr, as defined by define the modified failure assessment curves; for compar-
P ison, the dashed lines represent the BS 7910 Level 2A
Lr ¼ , (23) procedure as previously described by Eq. (5). Following
PL
Ainsworth [13], the parameters a and m appearing in the
where P is the applied load per unit thickness and PL is the constraint-modified FAD expression given by Eq. (11) is
plastic collapse load for the pin-loaded SE(T) specimen taken as unity. The values a ¼ 1 and m ¼ 1 derive from
defined as [31] simple linear fit to cleavage fracture toughness data (see
( 1=2 Ainsworth [13]) and are assumed to characterize relatively
PL  a 2 a
¼ 1:97 1:59
1:59 þ 0:63 well the elevation of fracture toughness with constraint loss
sys W W W for typical structural steels. Since our primary interest lies
)
a  a in an exploratory application of the constraint-corrected


0:21 ; 40:55 ð24Þ FAD curve, the choice of these values simplifies the
W W
analyses and appears satisfactory given the present context.
 ARTICLE IN PRESS
S. Cravero, C. Ruggieri / International Journal of Pressure Vessels and Piping 83 (2006) 607–617
Fig. 7. Prediction of the burst pressure for the tested pipe specimens with
10  200 mm crack based upon constraint corrected BS 7910 Level 2A.
Fig. 8. Prediction of burst pressure for the tested pipe specimens with
7  140 mm crack based upon constraint corrected BS 7910 Level 2A.
Fig. 9. Prediction of burst pressure for the tested pipe specimens with
3  60 mm crack based upon constraint corrected BS 7910 Level 2A.
Consistent with the dependence of load on the parameter
Q, the results show a strong dependence of the FAD curves
on structural constraint. While the FAD assessment line is
essentially unchanged for the correction based on the
deeply cracked specimen displayed in Fig. 7 (a=W ¼ 0:65),
615
there is a pronounced effect on the FAD curve presented in
Fig. 9 when the correction procedure utilizes the shallow
crack specimen (a=W ¼ 0:2). However, it can be seen that
the constraint-based correction of the FAD curve has little
or no effect on the burst pressure predictions for all cases
analyzed. While such behaviour could be anticipated for
the deeply cracked pipe specimen with a 10  200 mm crack
(note the relatively small effect of constraint, as measured
by parameter Q, on parameter Lr displayed in Fig. 6), we
observe no significant differences in fracture predictions for
other pipe specimens. The predicted failure pressure for the
pipe specimen with a 7  140 mm crack is now E20.2 MPa
(compare with the predicted value of 19.2 MPa displayed in
Table 2 using the normal integrity assessment) whereas the
prediction for the shallow crack pipe specimen (3  60
mm) remains unchanged. Here, this pipe specimen deserves
more attention as this crack configuration has the lowest
crack-tip constraint. Because the failure mode is mainly
controlled by plastic collapse of the remaining crack-tip
ligament, the corresponding loading path intersects the
FAD curve at the line Lr ¼ Lmaxr (which has been assigned
a value of 1.12 in the present work).
The behaviour displayed by the constraint-based ana-
lyses conducted in this study clearly underlies the central
role played by the collapse failure mode in the tested pipes.
As previously discussed, these axially cracked pipe speci-
mens give rise to low levels of crack-tip constraint,
particularly in the case of the shallow crack pipe specimen
(3  60 mm) but, at the same time, display almost no
sensitivity to the Q-based correction of fracture toughness.
Consequently, accurate plastic collapse solutions remain
essential to correctly quantify the failure pressure based
upon the FAD methodology.
7. Discussion and concluding remarks
This study explores applications of the FAD procedure
in defect assessments of high pressure pipelines with axial
surface cracks. Full-scale burst tests conducted on X60
end-capped pipe specimens with varying crack configura-
tion provided the measured failure pressures to verify the
applicability of FAD approaches. Conducted as part of the
pipeline integrity program at Petrobras, these pipe speci-
mens cover a range of crack length and crack depth. These
structural defects under predominant tensile loading which
develops in pressurized piping systems are associated with
low levels of crack-tip stress triaxiality. Such conditions
contrast sharply to the high levels of stress triaxiality
implicitly assumed in the construction of geometry-
independent FAD curves.
The predictions of the burst pressure for the pipe
specimens based upon the BS 7910 and API 579 procedures
(both utilizing normal levels of assessment) depend rather
markedly on the crack size. The analyses reveal larger
margins between experimental values and predicted results
for the deeply cracked pipe specimen (10  200 mm),
particularly for the BS 7910 procedure. In contrast, the
 ARTICLE IN PRESS
616 S. Cravero, C. Ruggieri / International Journal of Pressure Vessels and Piping 83 (2006) 607–617

estimated failure pressures agree well with the experiments

for the shallow crack pipe specimen (3  60 mm); here, we
note that the API 579 procedure somewhat overpredicts
the burst pressure. Overall, the API 579 procedure appears
to provide slightly better agreement with experimental
data. While the analyses suggest a potential excessive
conservatism in some cases, the results validate the use of
FAD-based methodologies for defect assessments of axially
cracked pipelines.
Moreover, the results reported here support our previous
simplification in which ductile tearing is not considered in
the adopted assessment procedures. While the experimental
measurements revealed small amounts of ductile crack
extension preceding pipe failure, predictions of structural
response for the cracked pipes were nonetheless in fair Fig. 10. Predicted burst pressures for the tested pipes using the FAD
agreement with experimental data without recourse to methodology for all analyses.
crack growth resistance data. A more refined defect
assessment which incorporates ductile crack extension
prior to pipe failure involves the use of a tearing equality between the experiments and predictions, i.e.,
(Level 3) analysis in both BS 7910 and API 579 procedures. PB-exp ¼ PB-pred. The analyses also encourage further
While this analysis was not conducted, we would anticipate investigations in direct correlations between full-scale tests
that a Level 3 tearing assessment would bring the predicted of cracked pipelines and integrity assessments based upon
values for the deeply cracked pipes into better agreement FAD curves. In particular, we believe a key feature in the
with experimental data since crack extension results in success of the constraint-based FAD methodology lies in
additional load capacity not accounted for in a stationary the controlling failure mode for the structural component.
crack analysis. However, it appears unlikely that a tearing At moderate levels of crack-tip constraint, elastic–plastic
analyses would improve the failure pressure predictions for fracture controls the material’s failure. While there are
the shallow cracked specimen since the failure mode for sound arguments which can be used to justify the proposed
this crack configuration is predominantly controlled by Q-based approach in defect assessment procedures for such
plastic collapse. On-going work addresses this issue in cases, the full robustness of the methodology remains
connection with a micromechanics approach to predict untested. Nevertheless, the present analyses, when taken
ductile tearing in cracked pipelines [33]. together with previous studies, provide a fairly extensive
To incorporate the potential effects of structural con- body of results against which the applicability of FADs in
straint on the FAD approach, the present work adopts the defect assessments of pressurized pipelines can be weighed.
hydrostatic parameter Q to construct the dependence of
load on crack-tip stress triaxiality. Based upon the
Acknowledgements
similarities in J–Q trajectories, the functional relationship
of Lr vs. Q for pin-loaded SE(T) specimens is employed to
This investigation is supported by the State of São Paulo
derive constraint-modified FAD curves applicable for
Research Foundation (FAPESP) through Grant 03/02735-6
integrity assessments of the pipe specimens. While the
and by the Brazilian Council for Scientific and Technolo-
modified FAD assessment lines display a strong geometry
gical Development (CNPq). The authors acknowledge the
dependence (which is associated with varying levels of
Brazilian State Oil Company (Petrobrás) for making
crack-tip constraint), the constraint-based corrections have
available the experimental data for the API 5L X60 pipeline
little or no effect on the burst pressure predictions for all
steel plate and the burst pressure data. The authors are
cases analyzed. Specifically for the shallow crack pipe
indebted to José Claudio Guimarães Teixeira and Eduardo
specimen, the correction of constraint loss is completely
Hippert Jr. (CENPES-Petrobrás) for the many useful
offset by the (relative) position of the loading path on the
discussions and contributions. Helpful suggestions by the
Kr vs. Lr plot which underlies a predominant plastic failure
reviewers are gratefully acknowledged.
mode of the remaining crack ligament for this specimen.
While we have not explored an extensive range of crack
configurations, the results presented here provide strong References
support to use FAD-based approaches in defect assess-
ments of pressurized pipelines. Fig. 10 summarizes the [1] American Petroleum Institute. Welding of pipelines and related
predicted burst pressures for the tested pipes using the facilities. API–1104, 19th ed., 1999.
[2] Canadian Standards Association, CSA–Z662–99. Oil and gas pipeline
FAD methodology for all analyses (normal assessment and systems. 1999.
constraint-modified assessment). The symbols in the plots [3] Anderson TL. Fracture mechanics: fundaments and applications. 3rd
represent the predicted values whereas the solid line defines ed. New York: CRC Press; 2005.
 ARTICLE IN PRESS
S. Cravero, C. Ruggieri / International Journal of Pressure Vessels and Piping 83 (2006) 607–617
[4] British Energy. Assessment of the integrity of structures containing
defects. R6 Revision 4, 2003.
[5] British Standard Institution. Guide on methods for assessing the
acceptability of flaws in metallic structures. BS7910, 1999.
[6] SINTAP: Structural Integrity Assessment Procedure for European
Industry. Final Procedure, 1999.
[7] American Petroleum Institute. Recommended practice for fitness-for-
service. API RP-579, 2000.
[8] American Society of Mechanical Engineers. ASME boiler and
pressure vessel code. Section XI, 2004.
[9] Eiber RJ, Kiefner JF. Failure of pipelines. In: Metals handbook. 9th
ed. vol. 11-Failure analysis and prevention. American Society for
Metals; 1986. p. 695–706.
[10] American Welding Society. Welding handbook: welding technology.
8th ed., vol. 1. Miami: American Welding Society; 1987.
[11] National Energy Board. Stress corrosion cracking on canadian oil
and gas pipelines. Report MH–2–95, Calgary, 1996.
[12] Ainsworth RA, O’Dowd NP. Constraint in the failure assessment
diagram approach for fracture assessment. J Appl Mech 1995;117:
260–7.
[13] Ainsworth RA. A constraint-based failure assessment diagram for
fracture assessment. Int J Pressure Vessels Piping 1995;64:277–85.
[14] O’Dowd NP, Shih CF. Family of crack-tip fields characterized by a
triaxiality parameter: Part I—Structure of fields. J Mech Phys Solids
1991;39(8):989–1015.
[15] O’Dowd NP, Shih CF. Family of crack-tip fields characterized by a
triaxiality parameter: Part II—Fracture applications. J Mech Phys
Solids 1992;40:939–63.
[16] Dowling AR, Townley CHA. The effects of defects on structural
failure: a two-criteria approach. Int J Pressure Vessels Piping 1975;3:
77–107.
[17] Harrinson RP, Loosemore K, Milne I. Assessment of the integrity of
structures containing defects. CEGB Report R-H-R6. UK: Central
Electricity Generating Board; 1976.
[18] Sorem WA, Dodds RH, Rolfe ST. Effects of crack depth on elastic
plastic fracture toughness. Int J Fracture 1991;47:105–26.
[19] Wiesner CS, Goldthorpe MR. The effect of temperature and specimen
geometry on the parameters of the local approach to cleavage
fracture. In: International conference on local approach to fracture
(MECAMAT 96), Fontainebleau, France. 1996. p. C6–295–304.
[20] Ruggieri C, Dodds RH. A transferability model for brittle fracture
including constraint and ductile tearing effects: a probabilistic
approach. Int J Fracture 1996;79:309–40.
[21] Joyce JA, Link RE. Ductile-to-brittle transition characterization
using surface crack specimens loaded in combined tension and
617
bending. In: Underwood JH, et al., editors. Fatigue and fracture
mechanics: 28th volume, ASTM STP 1321. Philadelphia: American
Society for Testing and Materials; 1997. p. 243–62.
[22] Dodds RH, Shih CF, Anderson TL. Continuum and micro-
mechanics treatment of constraint in fracture. Int J Fracture 1993;64:
101–33.
[23] Nevalainen M, Dodds RH. Numerical investigation of 3-D constraint
effects on brittle fracture in SE(B) and C(T) specimens. Int J Fracture
1995;74:131–61.
[24] Dodds RH, Ruggieri C, Koppenhoefer K. 3-D constraint effects on
models for transferability of cleavage fracture toughness. In: Under-
wood JH, et al., editors. Fatigue and fracture mechanics: 28th
volume, ASTM STP 1321. Philadelphia: American Society for
Testing and Materials; 1997. p. 179–97.
[25] Brazilian State Oil Company–Petrobrás. Burst Pressure tests in 20’’
O.D. API grade 5L X60 pipelines. Private report, 2002
(in Portuguese).
[26] Brazilian State Oil Company–Petrobrás. Ultrasonic measurements in
burst pressure tests for a 20’’ O.D. API grade 5L X60 pipelines.
Private report, 2002 (in Portuguese).
[27] Silva MS. Fracture toughness and R-curve measurements for an API
X60 pipeline steel using a direct current potential technique. M.Sc.
Thesis. Faculty of Engineering (COPPE), Federal University of Rio
de Janeiro, 2002 (in Portuguese).
[28] Folias ES. An axial crack in a pressurized cylindrical shell. Int J
Fracture Mech 1965;1:104–13.
[29] Cravero S, Ruggieri C. Correlation of fracture behavior in high
pressure pipelines with axial flaws using constraint designed test
specimens—Part I: plane-strain analyses. Eng Fracture Mech 2005;
72:1344–60.
[30] Koppenhoefer K, Gullerud A, Ruggieri C, Dodds R, Healy B.
WARP3D: Dynamic nonlinear analysis of solids using a precondi-
tioned conjugate gradient software architecture. Structural Research
Series (SRS) 596. UILU–ENG–94–2017. Champaign: University of
Illinois at Urbana; 1994.
[31] Miller AG. Review of limit loads of structures containing defects. Int
J Pressure Vessels Piping 1988;32:197–327.
[32] Cravero S, Ruggieri C. JQCRACK—numerical computation of the
hydrostatic parameter Q for 2D cracked structural components.
Technical report BT–PNV–59, Politechnic School, University of Sao
Paulo, 2003 (in Portuguese).
[33] Dotta F, Ruggieri C. Fracture assessments of high pressure pipelines
with axial flaws including constraint and ductile tearing effects. In:
9th European Conference on Mechanics of Materials—EMMC9,
Fontainebleau, 2006.