Engineering Fracture Mechanics 74 (2007) 431–450

www.elsevier.com/locate/engfracmech

Limit analysis of flaws in pressurized pipes and

cylindrical vessels. Part I: Axial defects
M. Staat *, Duc Khoi Vu
Aachen University of Applied Sciences, Campus Jülich, Ginsterweg 1, 52428 Jülich, Germany

Received 7 October 2005; received in revised form 13 April 2006; accepted 18 April 2006
Available online 13 July 2006

Abstract

Within the theory of plasticity recently developed primal-dual limit analyses with the finite element method (FEM) give
upper and lower bounds of the limit load. The method can be iterated until both bounds have converged to the same value
which is then considered the exact plastic collapse load. This new numerical method is used in part I of the paper to derive
improved lower bound limit load formulae for axial defects in pressurized cylinders for any defect geometry and loading.
Based on the observation that even long slits in thick pipes have a residual strength a simple formula for the stress mag-
nification factor is justified also for thick pipes. Global collapse loads are compared with primal-dual FEM limit analyses
and with a large number of burst tests. For axial defects it is possible to find corresponding local collapse loads. Part II of
the paper will discuss the FEM discretization of the limit load theorems and will consider circumferential defects.
Ó 2006 Published by Elsevier Ltd.

Keywords: Limit load; Global and local collapse; Axial crack; Pressure loaded crack face; Plasticity

1. Introduction

Integrity assessment of piping components is essential for safe and reliable operation of oil or gas pipelines
and all types of process plants or power plants. It is particularly important for nuclear power plants because of
the application of the leak-before-break concept which involves detailed integrity assessment accounting for
postulated cracks. It is also especially important for safe operation of nuclear power plants, that sufficient duc-
tility of critical materials can be guaranteed for the lifetime under operating conditions. Therefore a large
number of burst tests with flawed pressure vessels and piping have been focussed on ductile materials and pub-
lished around 1970–85.
The established methods of defect assessment for ductile materials are based on fracture mechanics and
plastic collapse analysis. Limit load formulae are therefore needed for defect assessment with the two criteria

*
Corresponding author. Tel.: +49 0 2461 99 3209; fax: +49 0 2461 99 3199.
E-mail address: [email protected] (M. Staat).

0013-7944/$ - see front matter Ó 2006 Published by Elsevier Ltd.

doi:10.1016/j.engfracmech.2006.04.031
432 M. Staat, D.K. Vu / Engineering Fracture Mechanics 74 (2007) 431–450

Nomenclature

a crack depth
c half of crack length
D constraint factor (Tresca, von Mises)
eps relative prognosis error
FY, FU yield function, bounding function
MFL, M stress magnification factor old, new
MFL1, MFL2 Folias factor for internal, external defect
M1, M2 stress magnification factor for internal, external defect
n exterior unit normal
P reference pressure
p0 ; p0 collapse pressure without defect thin, thick
pexp, pformula experimental, predicted burst pressure
pex;glob ; pex;glob global collapse pressure of external defect old, new
pex;loc ; pex;loc local collapse pressure of external defect old, new

pin;glob ; pin;glob global collapse pressure of internal defect old, new
pin;loc ; pin;loc local collapse pressure of internal defect old, new
pL limit load (pressure)
pS ; pS collapse pressure of slit old, new

R1 ; R1 distinction of crack-face loading old, new
r1, r2 interior and exterior radius, respectively
Rp0.2 0.2% strain limit
RF flow stress
Rm ultimate tensile strength
t wall thickness
u_ velocity vector
Wext, Wint exterior power, interior dissipated power
c; c, c+ limit load factor; lower, upper bound
e_ plastic strain rate
r stress tensor
ry, ru yield, ultimate stress
X body
oXr, oXu traction, displacement boundary

methods [1–7] or the engineering treatment method [6,8]. In the reference stress approach limit loads can be
used to estimate non-linear fracture mechanics parameters such as crack tip opening displacement (CTOD), J
and C* integrals [4,9]. The J–resistance (J–R) curve is one of the important material input parameters in the
integrity assessment of flawed structures. Experimental evaluation of the J–R curve requires the gpl and c func-
tions which can be obtained from closed form limit load expressions [10,11].
Limit loads for pressurized cylinders have been collected and published, e.g., in [1,6,8,12–14]. However it is
shown in [15,16] that none of these formulae is correct for axial and circumferential defects, respectively.
Therefore new and improved lower bound limit loads are derived in this paper starting from a method of con-
necting extreme cases proposed in [14]. This method applies the static theorem of limit analysis by combining
the limit loads for a penetrating crack (slit) with the limit loads for a long axial defect or for a fully circum-
ferential one. It is the purpose of the paper to correct or improve solutions of the extreme cases and to com-
bine them to consistent limit loads of pressurized thick cylinders with axially and circumferentially oriented
flaws of all sizes and locations: From shallow cracks to penetrating slits, from short defects to maximum
length, internal (sealed and with pressure on crack faces) and external.
 M. Staat, D.K. Vu / Engineering Fracture Mechanics 74 (2007) 431–450 433

At the limit load the capacity of the structure is achieved because the load–deflection curve has a horizontal
slope which results in ‘‘unrestricted’’ deformation. This plastic collapse load is computed as the load at, e.g.,
twice the elastic slope of the load–deflection curve [11]. In classical plasticity this ambiguity is replaced by the
exact lower and upper bound theorems of limit analysis which define a unique limit load as the solution of the
maximum plastically safe load or the minimum plastic failure load. Both problems are dual and have the same
solution. The authors have implemented a new primal-dual method of limit analysis in the finite element
(FEM) software Code_Aster [17]. It could be demonstrated in [18] for a design-by-analysis problem of a pipe
junction that this method is much more exact and reliable than previous methods of FEM limit analysis such
as the basis reduction approach [19] or the elastic compensation method [20]. Therefore all FEM limit loads
calculated with the basis reduction approach in [15,21] are recalculated for axial cracks in thick pipes.
The analytical method in [14] allows two definitions of collapse loads of surface cracks – local collapse load
corresponding to the yielding of the remaining ligament and global collapse load corresponding to the net-
section collapse. The FEM limit analyses of the extreme cases (slit and very long crack) guided the develop-
ment of new global collapse loads for axial defects in pipes and cylindrical vessels with any crack shape and
size. With the primal-dual limit analyses it is now possible to check limit load formulae for long defects and for
slits with a new level of precision. FEM analyses have been made for thick and thin pipes. It could be shown,
that the difference between rectangular and semi-elliptical crack shapes is generally small [15]. The global limit
loads are compared with FEM limit loads and with 328 burst tests with a wide range of materials and dimen-
sions of pipes and axial defects in this first part of the paper. The results are also used to derive local collapse
loads which cannot be understood as limit loads. The title of the paper [22] is clearly misleading.
The same approach is applied to circumferential defects in the second part [23] of the paper, where more
detail on the approach is presented. It also contains the so-called Karush–Kuhn–Tucker conditions which rep-
resent the FEM implementation of the limit load problem. New formulae which were not found in literature
before are derived with the method in [14] and compared to FEM limit analyses. Burst tests with circumfer-
ential defects will be presented in the second part [23] of this paper.

1.1. Limit analysis

Stresses r are admissible in a perfectly plastic material model, if they satisfy the Tresca or the von Mises
yield condition F(r) 6 ry. With the equality F(r) = ry in one point, the elastic limit (0.2% strain limit)
ry = Rp0.2 is assumed and yielding can begin there. In the context of the two-surface theory of plasticity
the yield surface FY(r) 6 ry = Rp0.2 can harden kinematically within a bounding surface FU(r) 6 ru with some
ultimate stress ru. In the two criteria methods only partial use is made of the hardening by setting ru equal to
the flow stress RF,
Rp0:2 þ Rm
ru ¼ RF ¼ : ð1Þ
2
In the simplest theory the bounding surface is assumed as fixed in size, form, and location in stress space. Usu-
ally the same function is used for both surfaces, i.e. F(r) = FY(r) = FU(r).
The structure X is loaded monotonically by the surface traction p on the traction boundary oXr. For sim-
plicity of the presentation the body forces are neglected. The tractions are proportional to the internal pressure
p. Starting from some reference pressure P one may ask for the load factor c > 1 by which P can be increased
up to the collapse pressure pL = cP. The structure is plastically safe if it is in equilibrium and the yield con-
dition F(r) 6 ru is satisfied. The limit load pL of a bounded kinematic hardening material can be computed by
the static or lower bound theorem, [19],
c ¼ max c;
such that:
F ðrÞ 6 ru in X; ð2Þ
divr ¼ 0 on X;
rn ¼ cP n on oXr :
434 M. Staat, D.K. Vu / Engineering Fracture Mechanics 74 (2007) 431–450

The structure fails if the power Wext done by the externalR loading on a kinematically admissible strain rate field
pffiffiffiffiffiffiffiffi
e_ is larger than the internal plastic dissipation W int ¼ X p2ffiffi3 ru e_ : e_ dX. With the normalized external power
Wext = 1 the kinematic or upper bound theorem can be formulated as
cþ ¼ min W int ;
such that:
R
W ext ¼ oXr P nu_ dA ¼ 1; ð3Þ
 
_ T
e_ ¼ 12 Du_ þ ðDuÞ in X;
u_ ¼ 0 on oXu ;
+
where c denotes the upper bound of the limit load multiplier, oXu denotes the displacement boundary of X, e_
and rE denote the plastic strain rate and the fictitious elastic stress for an infinitely elastic material, u_ denotes
the velocity vector. To the maximum problem (2) which is stated in static quantities r and P there is a dual
minimum problem (3) which is stated in the kinematic quantities e_ and u. _ It is shown in the theory of optimi-
zation that by duality cP 6 pL 6 c+P.
The lower bound theorem states that cP is a safe pressure if a stress field can be found which is in equi-
librium with cP and satisfies F(r) 6 ru. The theorem is used in this paper to derive analytical lower bounds
for the global collapse pressure of axially flawed cylinders by finding appropriate stress fields. The limit load is
exact if an upper bound c+P can be found so that cP = c+P. The check by the upper bound is replaced by
primal-dual finite element limit analyses which iterate until cP = c+P is achieved.
Upper and lower bound theorems can be discretized by the FEM. The discretized minimum and maximum
problems are combined in the so-called Karush–Kuhn–Tucker conditions, which have been implemented in
the FEM software Code_Aster [17]. This primal-dual limit analysis seems to be more reliable and precise than
other methods such as the elastic compensation or the basis reduction [18]. One iterates until equality is
achieved in c 6 c 6 c+, which guarantees the exact solution pL of a given FEM discretization. The fast con-
vergence to a unique solution is demonstrated for the primal-dual FEM limit analyses in [16]. Note however,
that c itself depends on the type of the FEM discretization. All limit analyses in this paper are performed by
the primal-dual method using kinematically admissible finite elements with reduced integration. The numerical
error of volume elements is found in the range of 2–3% and is much smaller for membrane elements. The dis-
cretization and the Karush–Kuhn–Tucker conditions are discussed in the second part of this paper for the case
of circumferentially oriented cracks [23].
The optimization problems (2) and (3) show that the burst pressure pL is homogeneous of first order in ru.
Therefore, one may write in non-dimensional similarity variables
 
pL a a t r2
¼ DL ; ; ; ; . . . ; ð4Þ
ru t c r1 r1

where L is a function of crack geometry. Here the crack size is characterized by crack depth a and crack length
2c as shown in Fig. 1. The pipe geometry is characterized by r1, r2 and t which are internal and external radius
and wall thickness. The constraint factor D distinguishes the yield condition
8
<1 for Tresca;
D¼ 2 ð5Þ
: pffiffiffi for von Mises
3
for problems of pressurized pipes without defects.
The static or lower bound theorem states that any stress field in a structure is plastically safe if it is in equi-
librium with the loading and if F(r) 6 ru. This is used to derive lower bounds of the global collapse pressures.
First limit loads for extreme cases of infinitely long axial cracks and penetrating slits are considered in Section
2. In Section 3 the pipe is divided into two pipes which safely carry these extreme cases. The global collapse
pressure is found from the equilibrium of the interior pressure with the superposition of the stress fields in the
two pipes. Only rectangular crack shapes are considered for simplicity of the analysis and for a simple FEM
mesh. To any other shape of a crack an enveloping rectangular crack can be obtained by removing material.
 M. Staat, D.K. Vu / Engineering Fracture Mechanics 74 (2007) 431–450 435

2c

t a
r2
r1

(a)

a
t
r2
r1

(b)

Fig. 1. External semi-elliptical surface defect and infinitely long defect and in a pipe. (a) Axial, external surface defect. (b) Long axial
external defect in a pipe.

It is well known that the removal of material cannot increase the limit load. Therefore consideration of a rect-
angular defect instead of a semi-elliptical one is conservative. It is shown in [15] that it is only slightly so.
Limit analysis predicts only global collapse loads. The local collapse load is sometimes related to ligament
(or net-section) collapse. As a ! t the ligament thickness approaches zero and it is characteristic for the local
collapse load that it also approaches zero. However, in burst tests the pipe still carries the limit load, i.e. the
global collapse load, [21]. The local collapse pressure is found in Section 4 by the superposition of a stress field
which corresponds to an alternative partition of the piecewise continuous stress filed in the pipe.

2. The extreme cases

2.1. Thick pipe without defect

The burst pressure p0 of the thick-walled pipe without defects is

  "  2  3 #
p0 r2 t t 1 t 1 t
¼ D ln ¼ D ln 1 þ ¼D  þ   ; ð6Þ
ru r1 r1 r1 2 r1 3 r1

which realistic limit load solutions for the cracked pipe must assume asymptotically. Therefore the constraint
factor D is introduced in all equations below, although it is omitted in most of the equations that have been
cited from different references. The series expansion (6) converges for rt1 6 1. The solution for the Tresca yield
function applies independently of the conditions at the pipe end. The solution for the hypothesis after von
Mises applies for any plane strain state. Therefore it does not apply to the open pipe with free ends.
The approximation

p0 t
¼D ð7Þ
ru r1
for thin pipes over-estimates the load-carrying capacity of thick pipes, as the series expansion (6) shows. For
m = 0.3 the assumption of small deformations applies and therefore Eq. (6) remains valid with the Tresca
436 M. Staat, D.K. Vu / Engineering Fracture Mechanics 74 (2007) 431–450

hypothesis up to rr21 ¼ 5:43; 6:19 and 5.75 for closed end, open-end and plane-strain conditions, respectively
[24]. In the following analysis a closed pipe is assumed.

2.2. Long axial cracks in pipes

A lower bound for the limit load of a thick pipe with a long defect is obtained from a piecewise constant
stress field, if the pipe is divided into two coaxial pipes similar to Fig. 4. Pipe 1 carries the defect (slit pipe) and
is stress free. Pipe 2 is a pipe thinned by a and is at yield at all of its points.
For the interior crack, pipe 2 extends over r1 + a < r < r2. Using Eq. (6) the local collapse load
   

pin;loc r1 r2
lim ¼D ln ð8Þ
c!1 ru R1 r1 þ a
was presented with R1 ¼ r1 in [13]. The pressure loading of the crack faces can be considered with
r1 sealed crack faces;
R1 ¼ ð9Þ
r1 þ a pressure on crack faces:
Eq. (8) incorrectly assumes that the internal pressure acts on a cylinder of radius r1 + a. By correcting r1 + a
to the actual internal pipe radius r1, equilibrium was achieved in [15] at least for the hoop stress with
     
pin;loc r1 r1 þ a r2
lim ¼D ln : ð10Þ
c!1 ru R1 r1 r1 þ a
It is less conservative to replace R1 with R1,
(
r1 sealed crack faces;
R1 ¼ a ð11Þ
r1 þ pressure on crack faces:
2
A pipe with a penetrating defect over its full length will not collapse globally before a plastic hinge forms in the
wall opposite to the crack, see Fig. 4. The force F = 4cR1pin,glob has the lever arm R1 + a in Fig. 4. Equilibrium
of the bending moment alone leads to a global collapse at
    
pin;glob r1 r1 þ a r2 a2
lim ¼D ln þ ð12Þ
c!1 ru R1 r1 r1 þ a 4R1 ð2R1 þ aÞ
and with interaction between bending and normal force the global collapse pressure becomes
    
pin;glob r1 r1 þ a r2 1 h pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffii
lim ¼D ln þ ð2R1 þ aÞ þ 4R1 ðR1 þ aÞ þ 2a2 : ð13Þ
c!1 ru R1 r1 r1 þ a 2R1
For a conservative solution Eq. (13) is recommended although Eq. (12) falls nearly exactly on the FEM solu-
tion for sealed and pressurized crack faces. Both bending corrections assume that the normal stress in tension
and compression is ±ru. Strictly however the triaxial stress has the observed strengthening effect. The effect of
the radial stress rr and the axial stress rz is that for Tresca material in tension the normal stress achieves only
ru + rr and in compression only rz  ru. The complete consideration of all stresses is possible and will be
demonstrated for a circumferentially oriented slit in the second part [23] of the paper. But the analysis is long
and leads to complex formulae which must be solved numerically. The need for an iterative solution is not
justified here by the rather small improvement of the solution. The bending correction is only important
for thick pipes and for the fit of the global collapse pressure of a slit to the FEM limit analyses in the next
section.
The local collapse pressure for the external defect
 
pex;loc r2  a
lim ¼ D ln ð14Þ
c!1 ru r1
is correct in [13], except of the missing bending effect. The global collapse pressure due to the plastic hinge
caused by the moment is
 M. Staat, D.K. Vu / Engineering Fracture Mechanics 74 (2007) 431–450 437
 
pex;glob r2  a a2
lim ¼ D ln þ ð15Þ
c!1 ru r1 4r1 ð2r2  aÞ

and with interaction between bending and normal force

 
pex;glob r2  a 1 h pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffii
lim ¼ D ln þ ð2r2  aÞ þ 4r2 ðr2  aÞ þ 2a2 : ð16Þ
c!1 ru r1 2r1

2.3. Pipes with penetrating axial cracks

For the collapse load of wall-penetrating longitudinal cracks, semi-empirical formulae were set up, which
are often called Battelle formula or slit curve in the literature [25,26]. According to [21] it can be written in the
form

pS D r2 t
¼ ln  D ð17Þ
ru M FL r1 r1 M FL
which includes also thick pipes. In linear elastic fracture mechanics (LEFM) a bulging or Folias factor MFL is
used as a correction for the stress amplification due to the shell curvature. A simple relation is
sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
c2
M FL ¼ 1 þ k
r1 t
with the constant k = 1.61.
If the exact form of Eq. (17) holds, then a plot of
 2

pS c2
¼ M 2FL ¼ 1 þ k
p0 r1 t
2
as function of rc1 t must show a straight line passing through the ordinate at 1 and with slope k. This is not ob-
served however for the graphs of FEM limit analyses for slits in Fig. 2(a). Particularly the graph for the thick
a2
pipe with rr21 appears to be piecewise linear. The problem disappears with the bending correction 4R1 ð2R 1 þaÞ
in
Fig. 2(b) and the range of the slopes k narrows. The slope k is nearly constant if r1 is replaced by r2. The
approximation
sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
c2
M ¼ 1 þ 1:25 ð18Þ
r2 t
is found for the stress magnification factor M from Fig. 3.
It is suggested to generalize formula (17) as
 qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
pS D r2 1
¼ ln þ ðr1 þ r2 Þ þ 4r21 þ 4r1 t þ 2t2 ð19Þ
ru M r1 2r1
which includes the bending correction. It shows that the limit load of a cylinder with a deep axial crack or a slit
is not only influenced by the defect itself but also strongly by the opposite unflawed section of the wall. This
has not been recognized in [15] before.

3. Global collapse of pipes containing axial surface defects

3.1. Global collapse of pipes containing internal defects

A lower bound of the global limit load dependent on the defect position is obtained by dividing the pipe
into two coaxial pipes as shown in Fig. 4, which together are in static equilibrium with the internal pressure
[13]. Pipe 1 is the domain r1 6 r 6 r1 + a. It contains the part-axial surface crack as a penetrating defect. Pipe 2
is the domain (r1 + a 6 r 6 r2) with no crack. Alternatively it can be understood as a pipe with an axial crack
438 M. Staat, D.K. Vu / Engineering Fracture Mechanics 74 (2007) 431–450

10
9
8
7

(P0 /Pr)2
6
5
4 r2 /r1=1.10
3 r2 /r1=1.25
2 r2 /r1=1.50
r2 /r1=1.75
1
r2 /r1=2.00
0
0 5 10 15 20 25
(a) c2/(r1t)

18
16
14
12
(P0 / Pr)2

10
8
r2 /r1=1.10
6
r2 /r1=1.25
4 r2 /r1=1.50
2 r2 /r1=1.75
r2 /r1=2.00
0
0 5 10 15 20 25
(b) c2/(r1t)
2
Fig. 2. Collapse pressure from FEM limit analyses of slits; M 2FL ¼ 1 þ k rc1 t. (a) Formula without bending effect. (b) Formula with bending
effect.

18
16
14
12
(P0 / Pr)2

10
8
r2 /r1=1.10
6
r2 /r1=1.25
4 r2 /r1=1.50
2 r2 /r1=1.75
r2 /r1=2.00
0
0 2 4 6 8 10 12 14
c2/(r2t)
2
Fig. 3. Collapse pressure from FEM limit analyses of slits, M 2 ¼ 1 þ k rc2 t.

that extends over its total length with the collapse load in Eq. (6). In this way the collapse load for the thick
pipe with an internal axial surface crack has been obtained in [13,14],
    

pin;glob a r1 r2
¼D þ ln ð20Þ
ru r1 M FL1 R1 r1 þ a
with the Folias factor MFL1
sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
c2
M FL1 ¼ 1 þ 1:61 ð21Þ
r1 a
using Eqs. (17) and (8).
 M. Staat, D.K. Vu / Engineering Fracture Mechanics 74 (2007) 431–450 439

Fig. 4. Partition of piecewise continuous sections in section with different continuous stress fields at plastic collapse of an internal axial
surface crack in a pipe; R1 after Eq. (11).

The global collapse pressure has been checked extensively against lower bound FEM limit analyses with
rectangular and semi-elliptical defects in [15]. The analyses of the rectangular cracks have been repeated with
the more precise primal-dual method in the present paper, which shows that some of the critical comments
about the formulae in [14,15] has to be revised. A comparison of the FEM solutions in the present paper with
[15] is recommended to the reader who is interested in the progress of FEM based limit analysis. A typical
FEM net of an internal defect in a thick-walled pipe with rr21 ¼ 2 is shown in Fig. 5. Only one quarter of
the pipe has been modelled, because of the symmetry of the problem. The normalized von Mises stresses in
Fig. 5(a) show the collapse mechanism of an internal and external crack, respectively. The rectangular defect
causes unloading behind the crack front. There is no need for special crack tip elements or for mesh refine-
ment, because theplastic collapse
 is not controlled by the crack tip. All analyses have been repeated with
thin-walled pipes rr21 ¼ 1:1 with both types of crack shape, rectangular and semi-elliptical. These show sim-
ilar trends but the differences between the FEM limit analyses, the old and the new collapse formulae are less

Collapse mechanism
Crack front Crack front

Collapse mechanism

BioMech FH-Aachen Stresses, sigma vm R a p s BioMech FH-Aachen Stresses, sigma vm R a p s

9.500E-01 9.571E-01 9.643E-01 9.714E-01 9.786E-01 9.857E-01 9.929E-01 1.000E+00 9.900E-01 9.914E-01 9.929E-01 9.943E-01 9.957E-01 9.971E-01 9.986E-01 1.000E+00

(a) (b)

Fig. 5. Von Mises stress at Gauss points normalized to ru for internal and external surface defect axially oriented in a pipe with at ¼ 0:6
and ac ¼ 0:4. White indicates r/ru = 1. (a) Internal defect with pressure on crack face. (b) External defect.
440 M. Staat, D.K. Vu / Engineering Fracture Mechanics 74 (2007) 431–450

pronounced for thin pipes [15]. The formulae have been derived for rectangular defects. Therefore, in this
paper they are only compared with FEM solutions for this crack shape.  
The global collapse pressure (20), normalized with the burst pressure p0 ¼ ru D ln rr21 of the pipe without a
defect, is compared with FEM limit analyses in Fig. 6 for a thick-walled pipe with rr21 ¼ 2. The long crack limit
(8), i.e. c ! 1 or ac ! 0, under-estimates the burst pressure strongly. In Eq. (20) R1 of Eq. (9) makes the dis-
tinction of cases of crack-face loading only for the long defect part of the solution. Therefore, the pressure on
the crack faces is not considered for deep cracks such that Eq. (20) over-estimates the burst pressure for deep
defects as is obvious from Fig. 6(b). More precise limit loads of long cracks with ac ! 0 are calculated with a
2D model with membrane elements.
By using the corrected equations (19) and (10) an improved global collapse load is found
         
pin;glob r2 r1 1 r1 þ a r1 þ a r2
¼ min D ln ;D ln þ ln
ru r1 R1 M 1 r1 r1 r1 þ a
1 h p ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
ffi i
þ ð2R1 þ aÞ þ 4R1 ðR1 þ aÞ þ 2a2 ð22Þ
2R1
with the stress magnification factor M1
sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
c2
M 1 ¼ 1 þ 1:25 : ð23Þ
ðr1 þ aÞa
Here the effect of pressure on the crack faces is also correctly considered on the slit part of the solution. As
long crack limit one obtains the new lower bound (10) for the global collapse, because M1 ! 1 for c ! 1. It
is a truly global collapse pressure for all crack geometries up to the slit with a = t.
Fig. 7 shows that Eq. (22) greatly improves the global collapse load for all limiting cases, for deep or long
internal defects. The new long crack limit (10) is a close lower bound. For sealed crack faces in Fig. 7(a) the
solution needs to be bounded by p0 ¼ ru D ln rr21 for small defects. Fig. 7(b) shows that the new solution (22)
is conservative for pressurized crack faces for all parameters.

Carter’s global
p/p0 FEM

1.0
0.8
0.6
0.4
0.2 1.0
0.8
0.0 0.6
0.2 0.4 a/c
0.4 0.2
0.6
(a) a/t 0.8
1.0

Carter’s global
p/p0 FEM

1.0
0.8
0.6
0.4
0.2 1.0
0.8
0.0 0.6
0.2 0.4 a/c
0.4 0.2
0.6
(b) a/t 0.8
1.0

Fig. 6. Carter’s global collapse pressure pin;glob Eq. (20) compared with FEM limit analyses for internal defects in a thick pipe with rr21 ¼ 2.
FEM —, formula   . The pressures are normalized by p0. (a) Sealed crack faces, R1 ¼ r1 . (b) Pressure on crack faces, R1 ¼ r1 þ a.
 M. Staat, D.K. Vu / Engineering Fracture Mechanics 74 (2007) 431–450 441

New global
p/p0 FEM

1.0
0.8
0.6
0.4
0.2 1.0
0.8
0.0 0.6
0.2 0.4 a/c
0.4 0.2
0.6
(a) a/t 0.8
1.0

New global
p/p0 FEM

1.0
0.8
0.6
0.4
0.2 1.0
0.8
0.0 0.6
0.2 0.4 a/c
0.4 0.2
0.6
(b) a/t 0.8
1.0

Fig. 7. New global collapse pressure pin,glob Eq. (22) compared with FEM limit analyses for internal defects in a thick pipe with rr21 ¼ 2.
FEM —, formula   . The pressures are normalized by p0. (a) Sealed crack faces, R1 = r1. (b) Pressure on crack faces, R1 = r1 + a/2.

3.2. Global collapse of pipes containing external defects

3.2.1. Limit loads for with pipes containing external defects

In [1,14] the partition into sections as shown in Fig. 4 leads to a piecewise continuous stress field which has
been used to derive global collapse loads for the thick pipe with an axial surface crack at the external wall
  

pex;glob a r2  a
¼D þ ln ð24Þ
ru ðr2  aÞM FL2 r1
with the Folias factor MFL2
sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
c2
M FL2 ¼ 1 þ 1:61 : ð25Þ
ðr2  aÞa
This can be improved if Eq. (17) is replaced by Eq. (19) and the bending strength is included such that
    
pex;glob 1 r2 r2  a 1 h pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffii
¼D ln þ ln þ ð2r2  aÞ þ 4r2 ðr2  aÞ þ 2a2 ð26Þ
ru M2 r2  a r1 2r1
with the stress magnification factor M2
sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
c2
M 2 ¼ 1 þ 1:25 : ð27Þ
r2 a
With finite c, Eqs. (22) and (26) approach the solution (19) of the penetrating defect for a ! t. As the long
crack limit the lower bound of Eq. (10) for the local collapse is obtained, because M2 ! 1 for c ! 1. Eq.
(26) concerns the thick wall correction of the slit solution and the bending strength against the opening of
the long slit. The difference between Eqs. (24) and (26) can be observed for the wall penetrating defect
a
t
¼ 1 in Fig. 8.
442 M. Staat, D.K. Vu / Engineering Fracture Mechanics 74 (2007) 431–450

Carter’s global
p/p0 FEM

1.0
0.8
0.6
0.4
0.2 1.0
0.8
0.0 0.6
0.2 0.4 a/c
0.4 0.2
0.6
(a) a/t 0.8
1.0

New global
p/p0 FEM

1.0
0.8
0.6
0.4
0.2 1.0
0.8
0.0 0.6
0.2 0.4 a/c
0.4 0.2
0.6
(b) a/t 0.8
1.0

Fig. 8. Global collapse pressure formulae compared with FEM analyses for external defects in a thick pipe with rr21 ¼ 2. FEM —, formula
  . The pressures are normalized by p0. (a) Carter’s global formula pex;glob (24). (b) New global formula pex,glob (26).

The limit loads for the exterior and the interior long cracks have been calculated with 8 nodes plane finite
elements. All other limit analyses use 20 nodes volume elements. The difference can be observed for the no-
crack solution at ¼ 0 in Figs. 6–8. The limit loads for the 3D discretization ac ¼ 0:2; . . . ; 1:0 are 2–3% lower
than the exact 2D discretization at ac ¼ 0. The same limit loads are obtained even with a much coarser FEM
mesh. The accuracy of different volume elements in limit and shakedown analyses is discussed in [18]. The dif-
ference between different finite elements indicates the accuracy of the FEM discretization, because the primal-
dual method converges with no additional contribution to the numerical error. This is different from standard
step-by-step FEM analyses. For perfectly plastic material they tend to increase the yield stress numerically by
typically 2–3% during the iteration. The new global collapse pressure can be considered an exact limit load
within the numerical error bounds. The limit loads for interior defects seem to be good approximations, how-
ever slightly less accurate for some parameters (see Fig. 9).

under
over
eps [%]
a/t
0.0
100 0.2
0.4
0.6
0.8
50 1.0
0.8
0 0.6
0.4 a/c
-50 0.2

-100

Fig. 9. Relative error of prognosis with pex,glob (new global formula (26)) of 277 burst tests with external defects from [21]: + — burst
pressure under-rated,     burst pressure over-rated.
 M. Staat, D.K. Vu / Engineering Fracture Mechanics 74 (2007) 431–450 443

3.2.2. Burst tests with pipes containing external defects

In [21] 293 burst tests have been collected from the literature and carefully crosschecked. Aiming at a best
estimate, the new global formula (26) is compared to the 277 burst tests with external or penetrating defects in
Fig. 9. It shows that the relative prognosis error eps,
pexp  pformula
eps ¼  100% ð28Þ
pformula
increases by its definition for long, deep defects. The prognosis error for the uncracked cylinders is located at
the point ðat ; acÞ ¼ ð0; 1Þ. It indicates the influence of the uncertainty of pipe geometry and material strength
data. Fig. 10(a) shows
 the errors over the relative crack depth at together with the errors which would occur
r2
in thin cylinders r1 ¼ 1:1 with very long flaws (i.e. ac ! 0) with a ±5% and ±10% variation of t. The calcu-
lated uncertainties are not representative for the lengths of most of the tested flaws. But they indicated that
large uncertainties could be caused by uncertain input data. Probabilistic fracture mechanics could be a ra-
tional way to handle the stochastic uncertainties [28]. The prognosis error has a maximum around at ¼ 0:9
and it is falling to lower values for slits at ¼ 1:0 . The maximum of the error calculated as a function of
the only uncertain variable t decreases with shorter cracks. Fig. 10(b) shows the scatter of burst pressure over
the aspect ratio ac.
It is also expected that the scatter increases due to uncertainties of geometry and material data for crack
geometries for which a larger number of experiments has been made. A more detailed analysis of the data
in [21] shows that quite different burst pressures occur in some groups of experiments with steel from the same
heat with nominally identical data and close pipe and crack size. Typically the material of the tested pipes and
pressure vessels was neither isotropic nor particularly homogeneous. Therefore, conservative data has to be
used in defect assessment. The range of the data in these tests is indicated in Table 1.
The burst pressure of larger cracks cannot be explained by plastic collapse alone. On the one hand, it is
observed that the global collapse pressure is not achieved by most of the experiments. On the other hand it
is shown in [21] that the local collapse pressure cannot be considered as a valid structural model, because it
fails to show the residual strength of slits which has been observed in all tests. Therefore fitness-for-purpose
assessments should be made with global collapse loads in connection with two criteria methods such as the
SINTAP procedure, if values for fracture toughness were available [2–6]. For the tests in [21] fracture tough-
ness is strongly anisotropic and only given for a few tests in the source [27]. The new precise global collapse
loads are an important contribution to eliminate a potential source of systematic errors in the application of
the SINTAP procedure.
Fig. 11 shows the prognosis error for 68 burst tests with seamless gas cylinders made of modern high
strength steels [29,30] and aluminum [30]. The strength data range from Rm = 758 MPa to Rm = 1138 MPa
for the steels and Rm = 331 MPa for aluminum. Forty-one of these tests have been made for the International
Standards Organization (ISO) Technical Committee 58 (TC 58)/Subcommittee 4 (SC 4)/Working Group 1
(WG 1) on rejection criteria for metal cylinders. For at 6 0:4 the tests can be represented by plastic collapse
within the range of 7% to +15%. The variations for very long cracks have been calculated for rr21 ¼ 1:06
for Fig. 11. Again the plastic collapse over-estimates the burst pressures for deeper flaws. In fitness-for-pur-
pose assessments limit loads should therefore only be used as part of one of the two criteria methods men-
tioned above.
Additionally 219 plastic bursts tests with gas cylinders for ISO TC 58/SC 3/WG 14 on cylinder fracture in
[31] with deeper flaws with at P 0:34 have been assessed with the new global formula (26). The cylinders tested
in the WG 14 programme were manufactured over a period of about 80 years worldwide and the quality of the
steels, the uniformity of the cylinders and the provided data were highly variable. All global collapse loads are
found to be strongly unsafe with prognosis errors of 19% to 68%. Only Charpy-V notch impact energy Av
but no fracture toughness data is given in [31] for a fitness-for-purpose assessment of these tests.

4. Local collapse of pipes containing axial surface defects

The two-criteria fracture assessment procedures [4,7] conservatively recommend the use of a local limit load.
For part-through-thickness cracks the limit load may be taken as a local limit load which ‘‘. . . is defined as the
444 M. Staat, D.K. Vu / Engineering Fracture Mechanics 74 (2007) 431–450

under
50
over
10%
5%

0
eps [%]

-50

-100
0.0 0.2 0.4 0.6 0.8 1.0
(a) a/t

under
50
over

0
eps [%]

-50

-100
0.0 0.2 0.4 0.6 0.8 1.0
(b) a/c

Fig. 10. Relative error of prognosis with pex,glob (new global formula (26)) of 277 burst tests with external defects from [21]: + burst
pressure under-rated,  burst pressure over-rated. (a) Error as function of relative crack depth with uncertainties of very long flaws with a
±5% and ±10% variation of t. (b) Error as function of aspect ratio with uncracked pipes at ac ¼ 1:0.

load needed to cause a local crack ligament yielding somewhere along the crack front’’ in [22]. The present
paper follows a more operable approach presented by Carter in [14]. Similar to ligament instability it is char-
acteristic for Carter’s local collapse pressure that it is zero for the penetrating crack (slit). It has been demon-
strated in [21] that local collapse is hardly observed in burst tests. Local collapse loads may lead to prognosis
errors above +1000% for very deep cracks and slits because they ignore the residual strength of such flaws.
 M. Staat, D.K. Vu / Engineering Fracture Mechanics 74 (2007) 431–450 445

Table 1
Range of database of 293 burst tests (including 16 with internal defects) in [21]
Rp0.2 (MPa) Rm (MPa) Y/T ratio (Rp0.2/Rm) T (°C) Av (J) r2 r2 a a
r1 t c
min 155 416 0.364 70 6 30 1.02 0.0 0.002
max 1451 1611 0.956 370 261 711 2 1 1
Av denotes the Charpy-V notch impact energy.

40

30

20

10
eps [%]

-10

-20
under
-30 over
10%
-40 5%

0.0 0.2 0.4 0.6 0.8 1.0

a/t

Fig. 11. Relative error of prognosis with pex,glob (new global formula (26)) of 68 burst tests with external defects in seamless gas cylinders
from [29,30] with uncertainties of very long flaws with a ±5% and ±10% variation of t: + burst pressure under-rated,  burst pressure
over-rated.

4.1. Local collapse of pipes containing internal defects

For the thick pipe with an internal semi-elliptical surface crack in the longitudinal direction local collapse
loads are given1 in [1,14]
      

pin;loc D r2 r1 r2
¼ s1 ln þ c  ln ð29Þ
ru s1 þ c r1 R1 r1 þ a
along with
ca 1  a
s1 ¼ h    t  i ð30Þ
M FL1 r1 ln rr21  Rr1 ln r1rþa
2
a
1

and the distinction of cases of crack-face loading (9). This result is based on the old solutions of the extreme
cases.
Carter has obtained the local collapse load from the global one by reducing the additionally carrying width
w in Fig. 12 to the value s1 ¼ wð1  atÞ for the local solution. The width w is derived by comparing two global

1
A misprint in [1], p. AII.36, has been corrected in Eq. (29).
446 M. Staat, D.K. Vu / Engineering Fracture Mechanics 74 (2007) 431–450

w 2c w

t a
r2
r1

3 4 5

Fig. 12. Partition of piecewise continuous sections in section with different continuous stress fields at plastic collapse of an external axial
surface crack in a pipe.

solutions. The procedure is presented below with the new solutions of the extreme cases such that an improved
new local collapse pressure is found. Only the bending correction is neglected for simplicity.
A first global solution is achieved by dividing the pipe into two concentric pipes as already discussed with
reference to Fig. 4. Pipe 1 is the domain r1 6 r 6 r1 + a. It contains the surface crack as a penetrating slit and is
in equilibrium with
   
p1 1 r1 r1 þ a
¼D ln : ð31Þ
ru M 1 R1 r1

Pipe 2 is the domain r1 + a 6 r 6 r2. It has no crack and carries ru homogenously. It is in equilibrium with (the
simple solution (10) is preferred over the new Eq. (13))
    
p2 r1 r1 þ a r2
¼D ln : ð32Þ
ru R1 r1 r1 þ a

The superposition pI = min {p0, p1 + p2} yields the global collapse load (22) (for simplicity without the bend-
ing correction).
A second global solution is obtained if the pipe is divided into three domains as depicted in Fig. 12. The
domain 4 contains the crack along its total length c. It is in equilibrium with F4 = 4cr1p4 = 4cr1p2,
    
p4 p2 r1 r1 þ a r2
¼ ¼D ln : ð33Þ
ru ru R1 r1 r1 þ a

Additionally on the left and on the right, the domains 3 and 5 carry over their length w as uncracked pipes of
thickness t. They are in equilibrium with F3 = F5 = 2wr1p3,
 
p3 p5 r2
¼ ¼ D ln : ð34Þ
ru ru r1
Superimposed the global collapse load is FL = 4(w + c)r1pII = F3 + F5 + F4 = 4r1(wp3 + cp2),
wp þ cp2
pII ¼ 3 : ð35Þ
wþc
The additionally carrying length w is determined from the condition pII = pI = min{p0, p1 + p2} as
p  p2
w¼c I : ð36Þ
p3  pI
This approach shows that different stress fields may lead to the same lower bound of the limit load. Neverthe-
less the maximum value of the lower bound is unique. Eq. (35) becomes a local solution if the additionally
carrying length S 1 ¼ wð1  atÞ, is used. Then
 M. Staat, D.K. Vu / Engineering Fracture Mechanics 74 (2007) 431–450 447
       
pI r1 1 r1 þ a r1 þ a r2
¼ ln þ ln ð37Þ
ru D R1 M 1 r1 r1 r1 þ a
and
8    
pI
>
> D ln r2
if P ln r2
;
>
> r1 ru D r1
>
>
>
> h      i
>
< D
S 1 ln rr21 þ c r1Rþa ln r1rþa 2
pin;loc S 1 þc 1
¼ ð38Þ
ru >
>     
>
> pI r1 r1 þa r2  
>
> r D  R1 r1 ln r þa
>
> with S ¼ c 1  a u   1
if pI
< ln r2
:
>
: 1 t r p ru D r1
ln 2  I
r1 ru D

As limit value for c ! 1 one finds (10). The presentation in [15] is equivalent but may lead to S1 < 0.
Little distinction can be made between global and local collapse for long defects with c ! 1, i.e. ac ! 0.
Therefore the local collapse can be checked against FEM limit analysis for this case. The new local formulae
are close to the FEM solution whereas Carter’s formulae are unnecessarily conservative for internal defects. A
check against limit analysis is not possible for finite crack length. However, the local formulae are derived
from the same extreme cases. Therefore, it is expected that all corrections of the solutions of these extreme
cases will improve the local collapse pressures in the same way as it was proved for the global collapse.
The same corrections as for the global collapse can be observed for the local collapse in Fig. 13. Also the local
formulae (37) need to be bounded by p0. It is characteristic for the local collapse that there is no residual
strength for wall penetrating defects at ¼ 1 . The largest difference between the new formulae (37) and the
old local collapse pressure (29) and (30) is found for the pressure loaded defect. The more complete consid-
eration of the pressure loading on the crack faces in the new local formulae is considered to be more safe.

Carter’s local
p/p0 New

1.0
0.8
0.6
0.4
0.2 1.0
0.8
0.0 0.6
0.2 0.4 a/c
0.4 0.2
0.6
(a) a/t 0.8
1.0

Carter’s local
p/p0 New local

1.0
0.8
0.6
0.4
0.2 1.0
0.8
0.0 0.6
0.2 0.4 a/c
0.4 0.2
0.6
(b) a/t 0.8
1.0

Fig. 13. New local collapse pressure pin,loc Eq. (37) compared with Carter’s local collapse pressure pin;loc Eqs. (29) and (30) for internal
defects in a thick pipe with rr21 ¼ 2. New formula —, Carter’s formula   . The pressures are normalized by p0. (a) Sealed crack faces.
(b) Pressure on crack faces.
448 M. Staat, D.K. Vu / Engineering Fracture Mechanics 74 (2007) 431–450

Carter’s local
p/p0 New local

1.0
0.8
0.6
0.4
0.2 1.0
0.8
0.0 0.6
0.2 0.4 a/c
0.4 0.2
0.6
a/t 0.8
1.0

Fig. 14. New local collapse pressure pex,loc Eqs. (41) and (42) compared with Carter’s local collapse pressure pex;loc Eqs. (39) and (40) for
external defects in a thick pipe with rr21 ¼ 2. New formula —, Carter’s formula   . The pressures are normalized by p0.

4.2. Local collapse of pipes containing external defects

The local collapse load for the external defect is derived in [14] from the piecewise continuous stress field
with the partition into sections as shown in Fig. 12,
    

pex;loc D r2 r2  a
¼ s2 ln þ c ln ð39Þ
ru s2 þ c r1 r1
with
ca 1  a
s2 ¼ h  t  i : ð40Þ
M FL2 ðr2  aÞ ln rr21  ln r2ra
1
 a

Repeating all arguments that lead to the improved global limit load for the internal crack the improved local
collapse load is obtained by replacing Eq. (17) by Eq. (19) as
    
pex;loc D r2 r2  a
¼ S 2 ln þ c ln ð41Þ
ru S2 þ c r1 r1
with
 
c ln r2ra
2
1  at
S2 ¼ h    i  : ð42Þ
M 2 ln rr21  ln r2ra
1
 ln r2
r2 a

One finds Eq. (14) as the limit function for c ! 1. The new and the old local collapse formulae (41) and (42)
and the local collapse pressure (39) and (40), respectively, are compared in Fig. 14.

5. Conclusions and Comments

Improved lower bound limit loads have been derived for extreme cases of axial defects. The extreme cases
of long surface defects and penetrating slits both show that even a long slit has a residual strength against
opening by bending in the wall opposite to the crack. This observation shows that the limit load not only con-
siders the complete state of stress around the crack but also the structural behaviour of the cracked pipe. It
also leads to a consistent approximation of the stress amplification factor for slits. The new solutions for
the extreme cases have been combined to new global and local collapse loads for internal and external axial
defects for all cases: long, short (to defect free) or deep cracks including slits, thin and thick pipes. The new
limit loads are truly global collapse pressures, because they now include the residual strength of a long slit. The
new local collapse pressures are always below the global ones. They are not limit loads.
 M. Staat, D.K. Vu / Engineering Fracture Mechanics 74 (2007) 431–450 449

The new formulae are particularly recommended as safe global and local solutions for internal cracks with
and without pressure applied to crack faces because previously existing solutions are found to largely under-
estimate the global collapse pressure. In the case of pressure loading on the crack faces also a severe over-esti-
mation of the limit load is made by the old formula. Reliable limit loads are important, because over-estimat-
ing the burst pressure is clearly non-conservative. However, under-estimating the limit loads could also be
non-conservative, because it leads to under-predicting the CTOD and the crack opening area and conse-
quently also to under-predicting the leak rates.
The new global collapse loads have been checked against primal-dual finite element (FEM) limit analyses
with rectangular defects in thick pipes. Defect assessment with two criteria methods will benefit from the new
accurate collapse loads. Particularly the global collapse load for external defects appears to be nearly exact.
The comparison with 345 burst tests shows a large statistical uncertainty for long and deep defects with a ten-
dency of limit loads to under-predict the burst pressure. Further 219 burst tests with seamless steel gas cylin-
ders seem to show a more brittle failure. Therefore, the global collapse loads should be used as part of a two
criteria method if fracture toughness data is available. Local collapse loads do not represent the residual
strength which is observed in burst tests with deep axial cracks and slits.
A structural mechanics method has been proposed as a general route to high quality lower bound limit load
solutions for all configurations and sizes of axial cracks. The transfer of the method to circumferential cracks
will lead to a new class of lower bound limit load formulae in the second part or the paper.

Acknowledgements

Some preliminary primal-dual limit analyses with Code_Aster have been performed by N.P.K. Pamujula as
part of his Master Thesis. The second author would like to thank Aachen University of Applied Sciences
(AcUAS) and Alexander von Humboldt Foundation (AvH) for their generous support during his research
stay in Germany. J.H. Smith gave some background info on the ISO TC 58 burst tests, J. Hoffmann helped
with pstricks for LaTeX. We would like to thank U. Zerbst for pointing out some numerical problems in the
local collapse load of interior cracks.

References

[1] Al-Laham S. Stress intensity factor and limit load handbook. British Energy Report EPD/GEN/REP/0316/98, Barnwood Gloucester
UK, Issue 2, April 1999.
[2] Webster S, Bannister A. Structural integrity assessment procedure for Europe – of the SINTAP programme overview. Engng Fract
Mech 2000;67:481–514.
[3] Zerbst U, Schwalbe H-H, Ainsworth RA. An overview of failure assessment methods in codes and standards. In: Ainsworth AR,
Schwalbe K-H, editors. Comprehensive structural integrity – practical failure assessment methods, vol. 7. Oxford: Elsevier; 2003.
p. 1–48.
[4] R6: Assessment of the integrity of structures containing defects. British Energy, Rev 4, 2001.
[5] BS7910-1999. Guide on methods for assessing the acceptability of flaws in structures. British Standards Institution, London,
December 1999.
[6] Berger C, Blauel JG, Hodulak L, Wurm B. Bruchmechanischer Festigkeitsnachweis. Abschlussbericht 2001, FKM Richtline Heft Nr.
258, Forschungskuratorium Maschinenbau, Frankfurt am Main: VDMA Verlag; 2004.
[7] API RP 579. Recommended practice for fitness-for-service. American Petroleum Institute, Washington DC, 2000.
[8] Delale F, Schwalbe K-H, Zerbst U, Kim Y-J, Brocks W, Cornec A, Heerens J, Amstutz H. EFAM ETM 97 – the ETM method for
assessing the significance of crack-like defects in engineering structures, comprising the versions ETM 97/1 and ETM 97/2. Report
GKSS 98/E/6, Geesthacht, 1998.
[9] R5: An assessment procedure for the high temperature response of structures. British Energy, Rev 3, 2003.
[10] Staat M, Heitzer M, Yan A-M, Duc Khoi Vu, Nguyen D-H, Voldoire F, Lahousse A. Limit analysis of defects. Report Jül–376,
Forschungszentrum Jülich, March 2000.
[11] Chattopadhyay J, Kushwaha HS, Roos E. Some recent developments on integrity assessment of pipes and elbows, Part I: Theoretical
investigations. Int J Solids Struct 2006;43(10):2904–31.
[12] Görner F, Munz D. Plastische Instabilität. In: Munz D, editor. Leck-vor-Bruch-Verhalten druckbeaufschlagter Komponenten.
Fortschr. Ber. VDI-Z. Reihe 18, vol. 14. Düsseldorf: VDI-Verlag; 1984.
[13] Miller AG. Review of limit loads of structures containing defects. Int J Pres Ves Pip 1988;32:197–327.
[14] Carter AJ. A library of limit loads for fracture two. Nuclear Electric, Internal Report TD/SID/REP/0191, 1991/92.
450 M. Staat, D.K. Vu / Engineering Fracture Mechanics 74 (2007) 431–450

[15] Staat M. Local and global collapse pressure of longitudinally flawed pipes and cylindrical vessels. Int J Pres Ves Pip
2005;82(3):217–25. doi:10.1016/j.ijpvp.2004.07.022.
[16] Staat M, Vu Duc Khoi. Limit loads of circumferentially flawed pipes and cylindrical vessels under internal pressure. Int J Pres Ves Pip
2006;83(3):188–96. doi:10.1016/j.ijpvp.2006.01.002.
[17] EDF R&D. Manuel d’utilisation: introduction au Code_Aster. Electricité de France, 2000.
[18] Vu Duc Khoi, Staat M, Tran Ich Thinh. Analysis of pressure equipment by application of primal-dual theory of shakedown,
Commun Numer Meth Engng, 2006, in press, doi:10.1002/cnm.891.
[19] Heitzer M, Staat M. Basis reduction technique for limit and shakedown problems. In: Staat M, Heitzer M, editors. Numerical
methods for limit and shakedown analysis – deterministic and probabilistic problems. NIC-Series, vol. 15. Jülich: John von
Neumann Institute for Computing; 2003. p. 1–55. Available from: <http://www.fz-juelich.de/nic-series/volume15/nic-series-
band15.pdf>.
[20] Taylor N et al. The design-by-analysis manual. Report EUR 19020 EN, European Commission, Joint Research Centre, Petten, The
Netherlands, 1999. With error corrections at: <http://info.tuwien.ac.at/IAA/news/dba1_engl.htm>.
[21] Staat M. Plastic collapse analysis of longitudinally flawed pipes and vessels. Nucl Engng Des 2004;234(1–3):25–43. doi:10.1016/
j.nucengdes.2004.08.002.
[22] Sattari-Far I, Dillström P. Local limit load solutions for surface cracks in plates and cylinders using finite element analysis. Int J Pres
Ves Pip 2004;81(1):57–66. doi:10.1016/j.ijpvp.2003.11.015.
[23] Staat M, Duc Khoi Vu. Limit analysis of flaws in pressurized pipes and cylindrical vessels, Part II: Circumferential defects. Engng
Fract Mech, submitted for publication.
[24] Koiter WT. On partially plastic thick-walled tubes. In: Koch JJ, Koiter WT, Legger RJ, van der Neut A, editors. Anniversary volume
on applied mechanics, dedicated to C.B. Biezeno by some of his friends and former students on the occasion of his sixty-fifth birthday,
March 2, 1953. Haarlem, Antwerpen, Djakarta: N.V. De Technische Uitgeverij H. Stam; 1953. p. 232–51.
[25] Hahn GT, Sarrate M, Rosenfield AR. Criteria for crack extension in cylindrical pressure vessels. Int J Fract Mech 1969;5:187–210.
[26] Kiefner JF, Maxey WA, Eiber RJ, Duffy AR. Failure stress loads of flaws in pressurized cylinders. Philadelphia: ASTM STP 536;
1973. p. 461–81.
[27] Hippelein K, Muz J, Schiedermaier J, Sturm D, Stoppler W. Forschungsvorhaben Phänomenologische Behälterberstversuche –
Traglast- und Berstverhalten von Rohren mit Längsfehlern. Förderkennzeichen 150 279, Phase 1, Forschungsbericht MPA Stuttgart,
1985.
[28] Staat M. Probabilistic assessment of the fracture mechanical behaviour of an HTR-module primary circuit pressure boundary. Nucl
Engng Des 1996;160:221–36. doi:10.1016/0029-5493(95)01105-6.
[29] Toughiry M, Smith J, Yolken T, Coughlin C. Development of accept reject criteria for requalification of high pressure steel and
aluminum cylinders. Nondestructive testing information analysis center (NTIAC), TAT No. NT-01-0002/0022, Austin, TX, July
2002.
[30] Webster CTL, Bhuyan GS. Fracture performance of NGV cylinder designs. Surrey BC, Canada: Powertech Labs Inc.; 1998.
[31] ISO/TR 12391-2:2002(E). Gas cylinders – refillable seamless steel – performance tests – Part 2: Fracture performance tests –
monotonic burst tests. Technical Report, Geneva, Switzerland: International Organization for Standardization, 2002-12-15.