Engineering Failure Analysis 18 (2011) 271–283

Contents lists available at ScienceDirect

Engineering Failure Analysis

journal homepage: www.elsevier.com/locate/engfailanal

Corrosion defect assessment on pipes using limit analysis

and notch fracture mechanics q
M. Hadj Meliani a,c,⇑, Y.G. Matvienko b, G. Pluvinage c
a
Laboratoire de Physique Théorique et Physique des Matériaux (LPTPM), FSSI, Université Hassiba Benbouali-Chlef-02000, Algeria
b
Laboratory of Modelling Damage and Fracture, Mechanical Engineering Research Institute of the Russian Academy of Sciences,
4 M. Kharitonievsky Per., 101990 Moscow, Russia
c
Laboratoire de Mécanique, Biomécanique, Polymères et structures, LaBPS-ENIM, île de Saulcy 57045, Université Paul Verlaine de Metz, France

a r t i c l e i n f o a b s t r a c t

Article history: Two methods of corrosion defect assessment on pipes are described: limit analysis (LA) and
Received 16 May 2010 notch modified failure assessment diagram (NMFAD). Limit pressure analysis are based on
Received in revised form 13 September 2010 ASME B31G, modified ASME B31G, DNV RP-F101 codes and recent proposed formulation.
Accepted 13 September 2010
The notch stress intensity factor concept and SINTAP structural integrity procedure are
Available online 22 September 2010
combined to assess pipelines integrity into a notch-based assessment diagram so-called
‘NFAD’. Defect assessment is made by comparing safety factor to a prescribed value (deter-
Keywords:
ministic approach) or probability of failure to a conventional level.
Limit analysis
Notch modified failure assessment diagram
Ó 2010 Elsevier Ltd. All rights reserved.
Safety factor
Probability of failure

1. Introduction

Pipelines have been employed as one of the most practical and low price method for large oil and gas transport since
1950. The pipe line installations for oil and gas transmission are drastically increased in last three decades. Consequently,
the pipeline failure problems have been increasingly occurred. The economical and environmental and eventually in human
life considerations involve the current issue as structural integrity and safety affair. The explosive characteristics of gas
provide high wakefulness about the structural integrity. Therefore, the reliable structural integrity and safety of oil and
gas pipelines under various service conditions including presence of defects should be warily evaluated. The external defects,
e.g., corrosion defects, gouge, foreign object scratches, and pipeline erection activities are major failure reasons of gas pipe-
lines. A typical example of a corrosion defect is given in Picture 1. According to numerous design codes, this kind of defects is
considered as a semi-elliptical crack-like surface defect of aspect ratio a/c. The aspect ratio varies in range [0.1–1] depending
on corrosion rate anisotropy. Several types of pipes failures can be distinguished as longitudinal, circumferential or helicoi-
dally failures. These types depend mainly on pipe diameter. For small diameter pipes, where bending stresses are predom-
inant, circumferential failure occurs. For large diameters, hoop stresses are more important than bending stresses and
longitudinal failure appears. When bending and hoop stresses are of the same importance, fracture path becomes spiraled.
Pipe steels have yield stress up to 700 MPa for the most recent quality in order to ensure enough ductility and weldability.
Failures emanating from corrosion defect are elasto plastic fracture or plastic collapse. For these two situations, defect
assessment is made generally by two tools: failure assessment diagram (FAD) and limit analysis (LA).

q
The Paper was presented by Prof. G. Pluvinage at the Nato-Workshop Biskra, Algeria, 26–28 April 2010.
⇑ Corresponding author.
E-mail address: [email protected] (M. Hadj Meliani).

1350-6307/$ - see front matter Ó 2010 Elsevier Ltd. All rights reserved.
doi:10.1016/j.engfailanal.2010.09.006
272 M. Hadj Meliani et al. / Engineering Failure Analysis 18 (2011) 271–283

Picture 1. Example of corrosion defect on pipe.

In this paper the two major corrosion defect assessment tools for pipes are presented:

(i) limit analysis

(ii) a notch adapted failure assessment diagram by modification of the SINTAP procedure using the volumetric method, a
comparison of the methods is given as a conclusion.

2. Assessment of corrosion defect by limit analysis

The structural integrity of corrosion defects is substantially studied. In Fig. 1, a list of methods available for corrosion de-
fect assessment is presented. They are grouped vertically by their type, codified methods or others, and horizontally by their
applicability, pressure or combined loading, etc.

2.1. ASME B31G and modified ASME B31G

ASME B31G [1] is a code for evaluating the remaining strength of corroded pipelines. It is a supplement to the ASME
B31 code for pressure piping. The code was developed in the late 1960s and early 1970s at Battelle Memorial Institute
and provides a semi-empirical procedure for the assessment of corroded pipes. Based on an extensive series of full-scale

Pressure only Combined loading

Pressure and
Length and depth Area and depth Area and depth
bending
Coded methods

ASME B31G
Modified
ASME B31G

DNV F101 DNV F101 DNV F101 DNV F101

RSTRENG 0.85 RSTRENG Effective Bubenik FEM

Safe-SwRi
Mok et al. Leis-PCORRC
Stress model
Other methods

Andrew correction
Hopkins factor

Wang-SwRi
Rosenfeld Strain model

Choi et al. SINTAP

SINTAP

Fig. 1. Methods for corrosion defect assessment based on limit analysis.

 M. Hadj Meliani et al. / Engineering Failure Analysis 18 (2011) 271–283 273

tests on corroded pipe sections, it was concluded that pipeline steels have adequate toughness and the toughness is not
a significant factor. The failure of blunt corrosion flaws is controlled by their size and the flow stress or yield stress of
the material. The input parameters include pipe outer diameter (D) and wall thickness (t), the specified minimum yield
strength (rY ), the maximum allowable operating pressure (MAOP), longitudinal extent of corrosion (Lc) and defect depth
(d).
According to the ASME B31G code, a failure equation for corroded pipelines was proposed by means of data of burst
experiments and expressed with consideration of two conditions below. First, the maximum hoop stress cannot exceed
the yield strength of the material (rhh 6 rY ). Second, relatively short corrosion is projected on the shape of a parabola
and long corrosion is projected on the shape of a rectangle. The failure pressure equation for the corroded pipeline is clas-
sified by parabola and rectangle as shown in Fig. 2.
Parabolic defects:
 
2ð1:1rY Þ  t 1  ð2=3Þ  ðd=tÞ
Pf ¼ ; ð1Þ
D 1  ð2=3Þ  ðd=tÞ=M
qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
 2   qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
 2  
where M ¼ 1 þ 0:8 DL Dt ; for 0:8 DL Dt 6 4:
Rectangular defects

2ð1:1rY Þ  t
Pf ¼ ½1  ðd=tÞ; ð2Þ
D
where M ¼ 1,
sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
 2  
L D
for 0:8 > 4;
D t
where Pf, D, d, t, M, rY and L are the failure pressure, outer diameter, maximum corrosion depth, wall thickness, bulging
factor, yield stress and longitudinal corrosion defect length, respectively. Due to some problems associated with the defini-
tion of flow stress a new flow stress was proposed as:
rf ¼ 1:1  rY þ 69ðMPaÞ: ð3Þ
The modified ASME B31G including this new modified flow stress and bulging factor is a follows:
 
2ð1:1rY þ 69Þ  t 1  0:85  ðd=tÞ
Pf ¼ ;
D 1  0:85  ðd=tÞ=M

D/2
t

(a)
Longitudinal axis of corroded gas pipeline Longitudinal axis of corroded gas pipeline

L L
t d t d

Parabolic defect shape assumption for short defects Rectangular defect shape assumption for long defects Corroded regions
Corroded regions

(b) (c)
Fig. 2. (a) Typical illustration of corrosion defects (longitudinal axis). (b) Short corrosion defect simplified as a parabolic curve (ASME B31G). (c) Long
corrosion defects simplified as a rectangular defect.
274 M. Hadj Meliani et al. / Engineering Failure Analysis 18 (2011) 271–283

sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
 2    4  2
L D L D
where M ¼ 1 þ 0:6275  0:003375 ; ð4Þ
D t D t

 2  
L D
for 6 50:
D t
 
2ð1:1rY þ 69Þ  t 1  0:85  ðd=tÞ
Pf ¼ ;
D 1  0:85  ðd=tÞ=M
 2  
L D
where M ¼ 3:3 þ 0:032 ; ð5Þ
D t
 2  
L D
for > 50:
D t

It is necessary to recall that ASME B31G is limited to low stress concentration factors and internal pressure loading con-
ditions. In the assessment procedure, one considers the maximum depth and longitudinal extent of the corroded area, but
ignores the circumferential extent and the actual profile. If the corroded region is found to be unacceptable, B31G allows the
use of more rigorous analysis or a hydrostatic pressure test in order to determine the pipe remaining strength. Alternatively,
a lower maximum allowable operating pressure may be imposed.

2.2. DNV RP-F101 [2]

DNV RP-F101 is the first comprehensive and extensive code for pipeline corrosion defect assessment. It provides guidance
for internal pressure and combined loading. It covers all loading types e.g., pressure only and combined loading. Further-
more, it provides codified formulations for pressure, bending and area depth. DNV RP-101 proposes two methods to find
the failure pressure. The first method is based on the partial safety factor and the second is classified as allowable stress
design. Both methods entail information on the pipe outside diameter (D), wall thickness (t), ultimate tensile strength
(ru ), maximum allowable operating pressure (MAOP), longitudinal extent of corrosion (Lc) and defect depth (d). The allow-
able stress design method considering non-interacting defects is discussed here. The exact procedures for the partial safety
factor method and interacting defects can be found within the DNV code. To pursue the design procedure via DNV RP-101, it
is required to determine the loading type (pressure only and combined loading) and consequently, the failure pressure can
be obtained as:
 
2ðru Þ  t 1  ðd=tÞ
Pf ¼ ; ð6Þ
ðD  tÞ 1  ðd=tÞ=Q
sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
 2
1
where Q ¼ 1 þ 0:31 pffiffiffiffiffiffi ;
Dt

where Pf, D, d, t, Q and ru are the failure pressure, outside diameter, corrosion depth, wall thickness, correction factor and
ultimate tensile strength, respectively. According to DVN RP-101, the failure pressure should not exceed the maximum
allowable stress design operating pressure (MAOP), otherwise, the corroded pipe will be repaired or replaced before return-
ing to service.

2.3. Choi’s method

Based on limit load analysis assumptionspffiffiffiffiffi and finite element analysis of corroded pipelines, Choi et al. [3] proposed a limit
load solution as a function of R/t, d/t, L= Rt as follows:
8  
2
>
> 0:9  2ðru Þt
C þ C pLffiffiffiffi þ C 2 pLffiffiffiffi ;
>
> Di 0 1 Rt Rt
>
<
Pf ¼ pLffiffiffiffi < 6 ; ð7Þ
> Rt
>
>
> h i
>
: 1  2ðru Þt C þ C pLffiffiffiffi ; pLffiffiffiffi P 6
Di 3 4 Rt Rt

where Pf, ru, Di, d, t and R are the failure pressure or maximum pressure, ultimate tensile strength, inside diameter, defect
depth, wall thickness and average pipe radius, respectively. In general, the corrosion pits are idealized into a semi-elliptical
shape rather than rectangular and semi-spherical shapes.
 M. Hadj Meliani et al. / Engineering Failure Analysis 18 (2011) 271–283 275

Fig. 3. Geometry of the specimens used for burst test.

8 d2 d 9
> >
< C 0 ¼ 0:06 t  0:1035 t þ 1;
> >
=
   
where C 1 ¼ 0:6913 d 2 þ 0:4548 d  0:1447; C 2 ¼ 0:1163 d 2  0:1053 d þ 0:0292; : ð8Þ
>
> t
 t
 t t >
>
: ;
C 3 ¼ 0:9847 dt þ 1:1101; C 4 ¼ 0:0071 dt  0:0126

2.4. Experimental verification and codes comparison

An experimental verification has been made by burst tests. Specimens are a rolled steel cylindrical pipes with an external
diameter of 219.1 mm and a thickness of 6.1 mm. A quasi-semi-ellipsoidal notch represents a corrosion external defect. The
major axis of the ellipsoid is parallel to the axis of the tube and the length of the notch is 30.5 mm. The depth of the notch is
3.05 mm and the width is also 3.05 mm. The radius of the notch tip is 0.15 mm, as shown in Figs. 3 and 4. A schematic view of
the notched tube is given in Fig. 3.
Some tubes containing a notch with the dimensions given here were loaded with a gas up to failure (Fig. 4). Fracture oc-
curs under a pressure of about 12 MPa. Only two results are available, but they give fracture pressures that are close together
(see Fig. 5). Moreover, this type of burst test with gas is rare, as are papers on this subject. So it is very difficult to properly

Fig. 4. Experimental device for burst test under gas pressure.

276 M. Hadj Meliani et al. / Engineering Failure Analysis 18 (2011) 271–283

Fig. 5. Specimen after fracture.

Table 1
Predicted burst pressures from codes.

Puk (MPa) Error compared between experimental results (%)

ASME B31 G 11.3 5.8
Modified ASME B31 G 10.8 10
DNV RP-F101 6.6 45

compare our results. If we consider the gross hoop stress, it appears that computation gives a maximum value that is less
than the yield strength. The hoop stress acts along the circular direction of the tube, but the stress in a perpendicular direc-
tion acts in the longitudinal direction of the tube and the strains in this direction are constrained by the great length of the
tube. Consequently, a triaxiality effect is induced, and the stress to be observed is the Von Mises stress. This triaxiality effect
induces an over stress to obtain the yield of the material. Here, the overstress factor is t = rVM =ry = 760/528 = 1.44. The max-
imum value is about 760 MPa for an internal pressure of 12 MPa.
Predicted burst pressure was determined with codes ASME B31G and DVNRP-F101 and results are reported in Table 1.
According to the experimental result, it seems that the ASME B31G code is the closest. The DNVRP-F101 is the most conser-
vative code.

3. Modified SINTAP procedure for fracture emanating from notches [4,5]

The structural integrity of corroded pipes can be made using failure assessment diagram. Classical failure assessment dia-
gram are established for crack-like defects and are not directly applicable to corrosion defect. However a notch-adapted pro-
cedure based on notch fracture mechanics and particularly volumetric method can be used.

3.1. Volumetric method

The volumetric method [6] is a local fracture criterion, which assumed that the fracture process requires a certain volume.
This volume is assumed as a cylindrical volume with effective distance as its diameter. Physical meaning of this fracture
process volume is ‘‘the high stressed region” where the necessary fracture energy release rate is stored. The difficulty is
to find the limit of this ‘‘high stressed region”. This limit is a priori not a material constant but depends on loading mode,
 M. Hadj Meliani et al. / Engineering Failure Analysis 18 (2011) 271–283 277

Fig. 6. Schematic elastic–plastic stress distribution along notch ligament and notch stress intensity virtual crack concept.

structure geometry and load level. The size of the fracture process reduced to the effective distance according to the above
mentioned assumptions is obtained by examination of the stress distribution. The bi-logarithmic elastic–plastic stress dis-
tribution (Fig. 6) along the ligament exhibits three distinct zones which can be easily distinguished. The elastic–plastic stress
primarily increases and it attains a peak value (zone I) then it gradually drops to the elastic plastic regime (zone II). Zone III
represents linear behaviour in the bi-logarithmic diagram. It has been proof by examination of fracture initiation sites that
the effective distance correspond to the beginning of zone III which is in fact an inflexion point on this bi logarithmic stress
distribution. A graphical method based on the relative stress gradient v associated the effective distance to the minimum of
v.
The relative stress gradient is given by:
1 @ ryy ðrÞ
vðrÞ ¼ ; ð9Þ
ryy ðrÞ @r
where v(r) and ryy ðrÞ are the relative stress gradient and maximum principal stress or crack opening stress, respectively. The
effective stress for fracture is then considered as the average volume of the stress distribution over the effective distance.
However stresses are multiply by a weight function in order to take into account stress gradient due to geometry and loading
mode. The stress distribution is given by:
Z X eff
1
reff ¼ ryy ðrÞ  ð1  r  vðrÞÞdr: ð10Þ
X eff 0

Therefore, the notch stress intensity factor is defined as a function of effective distance and effective stress:
qffiffiffiffiffiffiffiffiffiffiffiffiffiffi
K q ¼ reff 2pX eff ; ð11Þ

where K q , reff and X eff are notch stress intensity factor, effective stress and effective distance, respectively. Description of this
kind of stress distribution at notch tip and the procedure with the help of the relative stress gradient is given in Fig. 6.

3.2. Notch adapted failure assessment diagram (NFAD)

An example of notch adapted failure assessment diagram is given by Matvienko [7]. The parameter kr is defined as the
ratio of the applied notch stress intensity factor K q and fracture toughness KIC.
rffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
2
r
Kq 1 rg
coh
kr ¼ ¼ sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
 : ð12Þ
K IC
r 2
1  r g : ðk1Þ2
coh t

Here, kt is the elastic stress concentration factor, rcoh is the cohesive stress ahead of the notch tip and rg the global stress.
The failure assessment curve is taken from the cohesive zone model and the criterion of average stress in the cohesive
zone head of the notch tip as indicates in [7]. However the critical stress intensity factor is shown as to be a decreasing func-
tion of the elastic stress concentration factor and consequently failure assessment curve is notch radius dependant. In order
to get a NFAD interpolation curve independent of the notch radius, the parameter kr is defined as follows:
278 M. Hadj Meliani et al. / Engineering Failure Analysis 18 (2011) 271–283

Table 2
SINTAP levels 0 and 1 description.

Level Data needed When to use

Default level
Level 0 Yield or proof strength When no other tensile data available
Standard levels
1. Basic Yield or proof strength: ultimate tensile Strength For quickest result. Mismatch in properties less than 10%

K q;app
kr ¼ ; ð13Þ
K q;c

Kq,app represents the applied notch stress intensity factor and Kq,c is the critical state of the notch stress intensity factor. An-
other presentation of the failure criterion can be made using a two parameters relationship, kr ¼ f ðLr Þ, where kr is the non
dimensional stress intensity factor and Lr the non dimensional load (for more detail see Ref. [4]. The fracture toughness Kq,c is
a function of notch radius. Lr parameter keeps the same definition. By assumption, the interpolation curve is independent of
notch radius and is the same that the crack’s one. Consequently, the SINTAP interpolation curves [8] were used in the NFAD.

3.3. Failure assessment curve

Two SINTAP levels are used (levels 0 and 1). Characteristics of these two levels are given in Table 2.
Level 0:
" #12  2:5
L2 h 6
i 150
f ðLr Þ ¼ 1 þ r 0:3 þ 0:7  eð0:6Lr Þ ; for 0 6 Lr 6 1 where Lmax
r ¼1þ : ð14Þ
2 rY
Level 1: 8h
> i1
< 1 þ L2r 2 ½0:3 þ 0:7  eðlL6r Þ ; 0 6 Lr 6 1
2
f ðLr Þ ¼ ;
>
: 1 N1
1 þ 12 2
0:3 þ 0:7  eðlÞ  Lr ; 2N
1 < Lr 6 Lmax
r

 
E
where l ¼ min 0:001  ; 0:6 ; ð15Þ
rY
   
1 rY þ rU rY
Lmax
r ¼ ; N ¼ 0:3 1  ;
2 rU rU
where, f(Lr), Lr, Lmax
r , rY, l, E, rU and N are interpolating function, non dimensional loading or stress based parameter, max-
imum value of non dimensional loading or stress based parameter, yield stress, first correction factor, modulus of elasticity,
ultimate stress, second correction facto, respectively.

3.4. Safety factor obtained from different defect geometry

In the failure assessment diagram, service conditions of a structure exhibiting a corrosion defect is represented by the

assessment point A of coordinates [Lr , kr ] in the NFAD plane (Lr, kr). Due to the definition, Lr and kr parameters are propor-
tional to pressure and the loading path is linear passing through A and intercepting the failure assessment curve at point C
(see Fig. 7). The safety factor is then defined as:
fs ¼ OC=OA: ð16Þ
Safety factor on corrosion defect have been determined for pressure service conditions of 70 bars for a gas pipe made in
X52 steel (yield stress 410 MPa). Three kinds of surface defects are examined, semi-spherical, semi-elliptical and long semi-
elliptical defects.
In Figs. 8a–8c, the geometrical configuration of these defects is presented. The defect depth for all models is equal to one-
half of pipe wall thickness and the defect length over defect depth ratio is considered as 10 (L=d ¼ 10).
The obtained notch stress intensity factors and applied internal pressure are utilized to define the required assessment
points which are used in the failure assessment prediction (Table 3).
The safety factor has been determined on the Failure Assessment diagram according to the procedure described in Fig. 6.
In Table 4, different safety factors according to the SINTAP procedure are presented.
We note that all values of the safety factor are above the conventional value of 2 and consequently all the defect sizes are
acceptable. The failure pressure is extracted by means of the maximum failure pressure according to the above-mentioned
codes (ASME B31G, Modified ASME B31G, DNV RP-101 and Choi’s method). The safety factor is determined by means of the
applied pressure Papp over failure pressure Pf as:
 M. Hadj Meliani et al. / Engineering Failure Analysis 18 (2011) 271–283 279

Fig. 7. Failure assessment diagram and definition of safety factor.

Fig. 8a. Central semi-spherical defect (t = 6.1 mm, d = t/2).

Fig. 8b. Central semi-elliptical defect (t = 6.1 mm, d = t/2, d/L = 0.1).

SectionA - A
45
45º

d
ρ t d
B
B
L
67.5°
SectionB - B

Fig. 8c. Central longitudinal semi-elliptical defect (t = 6.1 mm, d = t/2, d/L = 0.1, q = 0.15 mm).

P app
Fs ¼ : ð17Þ
Pf
In Table 5, different safety factors according to the SINTAP procedure and limit load analysis methods are computed using
implemented MATLAB code. As expected earlier, the SINTAP 0B is more conservative than SINTAP 1B. Nevertheless, ASME
B31G, modified ASME B31G, DNV RP F-101 and Choi’s method do not offer any structural integrity formulae for blunt notch
280 M. Hadj Meliani et al. / Engineering Failure Analysis 18 (2011) 271–283

Table 3
Effective stress, effective distance and notch intensity factors along radial and longitudinal direction using 70 bars as applied internal pressure.
p
Orientation of defect Defect type Effective distance (mm) Effective stress (MPa) Kq (MPa m)
Radial direction Semi-spherical 0.42 202.7 10.4
Semi-elliptical 0.67 343.4 22.3
Long 0.38 539.6 26.3
Longitudinal direction Semi-spherical 0.72 184.4 12.4
Semi-elliptical 0.53 252.1 14.5
Long blunt notch 0.63 311.4 19.7

Table 4
Safety factors according to the SINTAP.

Type SINTAP 0B SINTAP 1B

Semi-spherical 4.095 4.106
Semi-elliptical 3.407 3.750
Blunt notch 3.186 3.583

defects and DNV RP F-101 does not exhibit any variation in safety factor for the chosen semi-spherical and semi-elliptical
defects. The comparison of computed safety factors emphasizes that DNV RP F-101 and Choi’s method provide the upper
bound and lower bound margins.

3.5. Probabilistic safety factor

On the probabilistic point of view, the failure assessment diagram can divide into zone of iso failure probabilities. The
material failure curve is a particular case for which the failure probability is equal to 1 because failure is then a certainty.

Any assessment point of coordinates [Lr  kr ] is situated on an equi probability curve Pr . The following conventional failure
probability are often used Pr = 10 if there is no human life risk and Pr = 106; if there is. These conventional iso failure prob-
4

abilities divided the failure assessment diagram into three zones: the unsafe zone with Pr = 1; the safe zone with mainte-
nance Pr > 104 or 106 and the safe zone without maintenance P r < 104 or 106. This type of failure assessment diagram
is called probabilistic NFAD (Fig. 9).
The pipes located in a water network which consists of a pump, a reservoir and five pipe sections are submitted to a sto-
chastic water hammer. Fracture toughness, yield stress and corrosion depth are assumed to be randomly distributed to allow
determination of safety factor by Monte-Carlo and Form/Sorm methods. Within the chosen procedure, the following param-
eters are treated as random parameters and introduce into the notch failure assessment diagram:

 notch fracture toughness Kr,c,

 yield strength Re,
 ultimate tensile strength rult,
 defect depth a,
 maximum pressure pmax.

These random parameters are treated as not being correlated with one another. The exponential distribution generally
governs for defect size analysis. Consequently, the probability density function has the following form:
FðXÞ ¼ k expðkaÞ; ð18Þ
where k is the exponential distribution parameter. The mechanical properties of the studied steel materials are presented in
Table 6.
In probability theory and statistics, the variational coefficient (CV) is a normalized measure of dispersion of a probability
distribution. It is defined as the ratio of the standard deviation r to the mean l:
r
CV ¼ : ð19Þ
l
Table 5
Calculated safety factors using mentioned coded and other methods.

Type of defect SINTAP 0B SINTAP 1B ASME B31G MASME B31G DNV RP F-101 Choi et al.
Semi-spherical 4.0 4.1 3.5 4.0 4.2 3.3
Semi-elliptical 3.4 3.7 3.4 3.9 4.2 2.8
Blunt notch 3.1 3.5 N/A N/A N/A N/A
 M. Hadj Meliani et al. / Engineering Failure Analysis 18 (2011) 271–283 281

Fig. 9. Probabilistic notch failure assessment diagram (PNFAD) with definitions of material failure curve, assessment point, safe and unsafe zones and safety
factor.

This is only defined for non-zero mean, and is most useful for variables that are always positive. It is also known as unit-
ized risk or the variation coefficient. It is expressed as percentage. We have chosen the lower bound for variational coeffi-
cient, i.e. CV = 0.1. Using Monte-Carlo method, several assessment points (40–50) were generated using the characteristic
parameters of the distribution. The obtained NFAD are presented in Fig. 10.
The NFAD can be presented into polar coordinates (r, h). Two particulars value are noted in this polar diagram h1 and
h2. The first polar angle h1 corresponds to the angle of the intercept of the failure curve at abscissa Lr = 0.62. This cor-
responds to a conventional value of gross failure stress of 62% of the yield stress. The second polar angle h2 corresponds
to the intercept of the vertical line of Lmaxr abscissa. These two angles determine three domains in the NFAD diagram
(Fig. 11).

– If h < h1 brittle fracture.

– If h1 < h < h2 elastoplastic fracture.
– If h > h2 plastic collapse.

These three different failure zones are presented in Fig. 11. For any assessment point A we can define the safety factor
according to Eq. (20).

fs ¼ OA=OB: ð20Þ
However, and mainly for steel and cast iron, the assessment points lay in the plastic collapse zone. Another way to define
safety, is to consider that we have plastic collapse only for h = 0.
In this case the safety factor is define as:

fs ¼ OA =OB : ð21Þ

Values of hi and h2 are respectively hi = 55 and h2 = 22°. Muhammed et al. [9] have shown that the general trend that
emerges is that on average the margin of safety on the FAD is minimum in the middle (elastic–plastic) region, slightly higher
in the ‘plastic collapse region and maximum in the ‘fracture’ region. However, this overall trend is complicated by varying
degree of scatter in the different regions. For this reason, we have examined the evolution of the safety factor with the h
angle. It has been shown that the h angle is in range [0–7°] for steel, [0–15°] for cast iron and [0–4°] for polyethylene. All

Table 6
Mechanical properties of pipe steel and used distribution.

Mechanical properties Yield strength (Re) Ultimate strength Circumferential stress Fracture toughness Defect
p
Mean 410 MPa 528 MPa 41.8 MPa 116 MPa m 2 mm
CV 0.1 0.1 0.4 0.1 –
p
Stresses/SIF 41 MPa 52.8 MPa 18.44 MPa 11,6 MPa m –
Distribution Normal Normal Normal Weibull Exponential
282 M. Hadj Meliani et al. / Engineering Failure Analysis 18 (2011) 271–283

Fig. 10. Example of probabilistic notch failure assessment diagram for steel.

Fig. 11. Definition of zones of brittle fracture, elasto plastic failure and plastic collapse in NFAD and definition of theta angle.

data are in a narrow scatter band of range [l  3r; l + 3r] and in the region of plastic collapse. For this reason the safety
factor fs computed from the ultimate pressure done by code ASME B31G is also reported (Fig. 12). Mean values and standard
deviation for the safety factor of the three material are reported in Table 7.
The safety factor computed for fully plastic collapse is less than the mean value of fs.
The safety factor distribution is represented with a Weibull distribution. Kolmogorov–Smirnov test results indicates (see
Table 7) that the Weibull distribution is significant at 57%.

Fig. 12. Distribution of the safety factor with q angle for pipe steel.
 M. Hadj Meliani et al. / Engineering Failure Analysis 18 (2011) 271–283 283

Table 7
Mean values and standard deviation for
the safety factor of the material.

Material Steel
Mean l 12.32
fs 11.63
Standard deviation r 1.26
Variational coefficient CV 0.12

4. Conclusion

The structural integrity of corroded pipelines subjected to internal pressure is studied in this paper. The semi-spherical,
semi-elliptical and blunt notch defects are examined under this loading and safety factors are evaluated by means of the
SINTAP procedure which is modified using a notch-based failure assessment diagram or so-called ‘‘NFAD”. The ASME
B31G, modified ASME B31G, DNV RP F-101 and Choi’s method have been also utilized.
The values of deterministic safety factor are codified in codes such as EUROCODE 3 [10]. However, the actual trend is to
adopt partial safety factor according to the degree of uncertainties of the material properties in order to avoid over conser-
vatism. The design material properties are then defined as some percentile of the material resistance distribution (the mean
values is generally used for metals and alloy but for wood the admissible stress is defined as the 5th percentile of the dis-
tribution). The use of probabilistic notch failure assessment diagram allows getting global safety factor without introducing a
series of partial safety factors on material properties, defect size and applied loading. In addition, the probabilistic safety
factor is associated to a failure probability. Then, with a conventional allowed failure probability, it is possible to have a
maintenance policy and to compute the economic cost of the risk of structure failure. Under stochastic service pressure,
the safety factor is distributed randomly according to Weibull distribution. It has been seen that the Weibull modulus of
the distribution is about 10 and the confidence interval of the safety factor has a satisfactory value.

References

[1] American National Standard Institute (ANSI)/American Society of Mechanical Engineers (ASME). Manual for determining strength of corroded
pipelines, ASME B31G; 1984.
[2] DNV RP-F101 (2) DNV-RP-F101: Corroded pipelines. Det Norske Veritas; 1999.
[3] Choi JB, Goo BK, Kim JC, Kim YJ, Kim WS. Development of limit load solutions for corroded gas pipelines. Int J Pres Ves Piping 2003;80(2):121–8.
[4] Hadj Meliani M, Azari Z, Pluvinage G, Capelle J. Gouge assessment for pipes and associated transferability problem. Eng Fail Anal 2010;17:1117–26.
[5] Moustabchir H, Azari Z, Hariri S, Dmytrakh I. Experimental and numerical study of stress–strain state of pressurised cylindrical shells with external
defects. Eng Fail Anal 2010;17:506–14.
[6] Pluvinage G. Fracture and fatigue emanating from stress concentrators. Kluwer; 2003.
[7] Matvienko YG. Local fracture criterion to describe failure assessment diagrams for a body with crack/notch. Int J Fract 2003;124:107–12.
[8] Structural integrity assessment procedure (SINTAP). Final report EU project BE95-1462 Brite Euram Programme Brussels; 1999.
[9] Muhammed A, Pisarski HG, Sanderson RM. Calibration of probability of failure stimates made from probabilistic fracture mechanics analysis. Offshore
technology report, 2000/021. TWI Ltd.; 2000.
[10] Eurocode 3. Design of steel structures. General rules and rules for buildings division 1. Class 1; 2005.