See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/313786083

Ductile fracture of pipe-ring notched bend specimens – Micromechanical

analysis

Article  in  Engineering Fracture Mechanics · February 2017

DOI: 10.1016/j.engfracmech.2017.01.022

CITATIONS READS

2 35

7 authors, including:

Bojan Medjo Nenad Gubeljak

Faculty of Technology and Metallurgy, University of Belgrade University of Maribor
41 PUBLICATIONS   169 CITATIONS    147 PUBLICATIONS   489 CITATIONS   

SEE PROFILE SEE PROFILE

A. Sedmak Marica B. Rakin

University of Belgrade University of Belgrade
269 PUBLICATIONS   852 CITATIONS    147 PUBLICATIONS   1,439 CITATIONS   

SEE PROFILE SEE PROFILE

Some of the authors of this publication are also working on these related projects:

Effects of complex heterogeneous constitutive properties on the structural integrity of a flawed weld View project

Symposium on Hydrogen Embrittlement Understanding and Future Research Framework (ECF22 conference, Serbia, 26-31 August 2018). View project

All content following this page was uploaded by Nenad Gubeljak on 08 October 2018.

The user has requested enhancement of the downloaded file.

 Engineering Fracture Mechanics 175 (2017) 247–261

Contents lists available at ScienceDirect

Engineering Fracture Mechanics

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

Ductile fracture of pipe-ring notched bend specimens –

Micromechanical analysis
W. Musraty a, B. Medjo a,⇑, N. Gubeljak b, A. Likeb b, I. Cvijović-Alagić c, A. Sedmak d, M. Rakin a
a
University of Belgrade, Faculty of Technology and Metallurgy, Belgrade, Serbia
b
University of Maribor, Faculty of Mechanical Engineering, Maribor, Slovenia
c
University of Belgrade, Institute of Nuclear Sciences ‘‘Vinča”, Belgrade, Serbia
d
University of Belgrade, Faculty of Mechanical Engineering, Belgrade, Serbia

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

Article history: Integrity of pipes is typically assessed by testing fracture mechanics specimens, such as
Received 16 August 2016 compact tensile (CT) or single-edge notched bending (SENB). However, for some pipe
Received in revised form 25 January 2017 dimensions it is not easy or possible to fabricate a specimen conforming to the require-
Accepted 30 January 2017
ments of standard procedures. A new type of specimen is proposed recently, which can
Available online 15 February 2017
be advantageous for relatively small pipe diameters and axial defects - the pipe ring notch
bend specimen - PRNB.
Keywords:
In this work, criteria for failure by ductile fracture of PRNB specimens are determined
Pipe failure
Pipe ring notched bend specimens
experimentally and by application of micromechanical analysis. The influence of size of
Micromechanical analysis the specimen, as well as size and shape of the stress concentrator, is analysed. The results
Ductile fracture of this study, along with previous authors’ results, lead to the conclusion that the pipe ring
specimens can be applied in integrity assessment of pipes with defects. Also, the benefits of
their application include much simpler fabrication and the same material history as the
pipe itself.
Ó 2017 Elsevier Ltd. All rights reserved.

1. Introduction

Many recent studies have dealt with the analysis of deformation behaviour, failure, integrity assessment and remaining
service life of pipelines. Most of them, e.g. [1–3], deal with fracture initiation, i.e. determining the critical loading level for
crack growth initiation. The crack growth in pipelines is modelled so far using different techniques, such as cohesive zone
model [4], the extended finite element method X-FEM [5] or micromechanical analysis [6]. In addition to fracture, the pipe-
line components with cracks can also fail by plastic collapse under the exploitation loads; straight pipes are considered in [7],
while pipe elbows with different geometries are examined in [8,9]. Both of the failure types are considered in [10], on the
pipes with axial surface cracks.
Also, in addition to crack-like defects, blunt volumetric defects (which resemble the shape of local corrosion damages),
are also often analysed in the literature, [11–14]. Plastic collapse of the remaining ligament is considered as failure condition
in [11,12], typically by tracking the value of von Mises stress. On the other hand, failure by fracture initiation at the position
of the defect (bottom of the blunt stress concentrator - machined notch) is analysed in the papers [13,14]. Other approaches
for assessment of load carrying capacity of pipes with simulated corrosion defects have also been applied, such as notch

⇑ Corresponding author.
E-mail address: bmedjo@tmf.bg.ac.rs (B. Medjo).

http://dx.doi.org/10.1016/j.engfracmech.2017.01.022
0013-7944/Ó 2017 Elsevier Ltd. All rights reserved.
248 W. Musraty et al. / Engineering Fracture Mechanics 175 (2017) 247–261

Nomenclature

a0 initial crack (or notch) length [mm]

af final crack length [mm]
Da crack length increment [mm]
f porosity or void volume fraction [
]
f⁄ damage function or modified void volume fraction [
]
f0 initial porosity [
]
fc critical porosity [
]
fF porosity at final fracture [
]

fu value of damage function at the moment of fracture [
]
fv volume fraction of non-metallic inclusions [
]
K parameter which describes how the load capacity loss develops after the critical porosity has been reached in
Gurson-Tvergaard-Needleman (GTN) model [
]
n hardening exponent [
]
q1, q2 constitutive parameters in the GTN model and Complete Gurson model (CGM) [
]
r void space ratio [
]
R ring specimen external radius [mm]
S distance between supports [mm]
W ring specimen width [mm]
B ring specimen thickness [mm]
Sij stress deviator [MPa]
Rf stress ratio for cyclic loading, i.e. ratio of the minimum and maximum load [
]

Greek symbols
a, b constants (in the CGM) introduced by Thomason [
]
e1, e2, e3 principal strains [
]
g plastic correction factor [
]
k mean free path between non-metallic inclusions [lm]
r1 maximum principal stress [MPa]
rm mean stress [MPa]
r current flow stress of the material matrix [MPa]
/ yield function of the GTN model [
]

fracture mechanics in [15] or using the sharp pre-crack shape to model the blunt local corrosion damages, [16]. Of course,
some of the initial defects can be prevented in the pipe fabrication procedure; quality testing of the welded (seam) pipes is
considered in [17].
Experimental determination of fracture toughness of the fabricated pipes is often difficult, because the standard require-
ments regarding the specimen/crack geometry cannot be fulfilled for all wall thicknesses. For the circumferential cracks, the
authors of [6,18] have shown that the SENT specimen (single-edge notched tension specimen) is more adequate for assess-
ment of fracture development in the pipe wall than SENB specimen (single-edge notched bending). Testing of SENB specimen
is defined by ASTM 1820 standard [19], while British standard BS 8571 [20] deals with examination of SENT specimen.
Another two studies [21,22] also deal with the defects in circumferential direction; the authors of [21] consider determining
the stretch zone width (SZW) on the newly proposed compact pipe specimen, while the fracture conditions of these spec-
imens are compared with those for a cracked pipeline in [22].
In this work, a recently proposed testing technique for determining the fracture toughness on defects in axial direction is
applied, which includes the use of new pipe-ring notch specimens for bending (PRNB) [23,24], with stress concentrator in the
form of fatigue pre-crack or machined notch with root radius = 0.25 mm, Fig. 1. The main aim is to determine the ductile
fracture conditions under external loading. These specimens can be easily produced (i.e. cut) from a pipe, which enables
a quick and efficient testing. Moreover, the specimen material has undergone the same manufacturing and thermal treat-
ment as the pipe itself.
Some other examples of non-standard specimens containing cracks in axial direction of the pipe can also be found in lit-
erature, such as curved compact tension (CCT) specimens examined in [26], or different non-standard configurations for ten-
sion and bending shown in [27]. However, they are typically characterised by a relatively complex preparation and/or testing
procedure.
 W. Musraty et al. / Engineering Fracture Mechanics 175 (2017) 247–261 249

Fig. 1. Pipe-ring notch bend (PRNB) specimens after testing [25].

A schematic representation of a PRNB specimen exposed to bending is given in Fig. 2; stress concentrator can be either
machined notch or fatigue pre-crack. A more detailed drawing of the specimen, which includes the dimensions, is given in
the following section.
Previous examinations [23–25] have dealt with fracture parameters of the PRNB specimens. The procedure for determin-
ing the plastic correction factors (g) for pipe-ring notched bend specimens, in order to calculate the J integral based on load
versus displacement test, is considered in [23]. Comparison of fracture of PRNB specimen and SENB specimen, as standard
fracture mechanics specimen, is shown in [24,25]. Also, it has been shown in [24] that fatigue pre-cracking of PRNB speci-
mens is actually not a necessary condition for testing, i.e. that a machined narrow notch can serve as the initial defect in this
geometry. In this work, fracture of PRNB specimens with different geometries, as well as defect size and shape, is analysed.
Micromechanical analysis is used to assess the damage development in the specimens.

2. Specimens and testing

An experimental study has been conducted, [24], with an aim to determine the fracture behaviour of PRNB specimens. In
[24,25], the fracture of this specimen is compared with the fracture of standard SENB specimen; it was chosen for compar-
ison due to the same loading type, i.e. bending, and crack type (passing through thickness).
In this work, fracture conditions of several PRNB specimens, cut by Electrical discharge machining (EDM) from a single
plate of a high-strength low alloyed steel, are analysed; three of the specimens are shown in Fig. 1. It should be emphasised
that this study is performed entirely on the plate material, i.e. all specimens - Ring (PRNB) and standard SENB - were cut from
a single plate. PRNB specimens were fabricated with the same cross-section size as standard SENB specimen, which is often
impossible for pipe testing, especially if thin-walled pipes are considered. The idea is to provide the same material conditions
for the specimens cut from the same plate, with an aim to continue with further testing on specimens cut from the seamless
and seam pipes (which has recently been started).
The tested material is a micro alloyed high strength steel - grade HT50, with commercial mark NIOMOL 490K (producer -
Železarna Jesenice, Slovenia). Chemical composition of the material is shown in Table 1, while tensile properties are given in
Table 2. Based on these properties, it is comparable with S460 steel (according to EN 10025-6 standard).
Dimensions of the specimens are shown in Fig. 3 and Table 3: Ring external radius R, specimen width W and thickness B,
distance between the supports S, as well as crack or notch length a0. The specimens with radius 60 mm and 40 mm were

Fig. 2. PRNB specimen - testing scheme.

250 W. Musraty et al. / Engineering Fracture Mechanics 175 (2017) 247–261

Table 1
Chemical composition of the material, weight %.

C Si Mn P S Cr Ni Mo
0.123 0.33 0.56 0.003 0.002 0.57 0.13 0.34

Table 2
Tensile properties of the material.

Yield strength (ReH) [MPa] 460

Tensile strength [MPa] 575
Elongation at fracture [%] 21

Fig. 3. Specimen dimensions and testing scheme.

Table 3
Dimensions of specimens (shown in Fig. 3).

PRNB R85 PRNB R60 PRNB R40

R = D/2 [mm] 85 60 40
B [mm] 10.04 7.5 5
W [mm] 20 15 10
a/W [
] 0.44 (notch radius = 0.25 mm) 0.41 and 0.71 (pre-crack) 0.6 (pre-crack)
S ( = 0.9 D) [mm] 153 108 72

notched and subsequently pre-cracked. The pre-crack was obtained by cyclic loading, with the stress ratio value Rf = 0.1
(ratio of the minimum and maximum load), in accordance with the requirements of ASTM E1820 standard for SENB spec-
imens [19]. The specimens with radius 85 mm were notched, without fatigue pre-cracking. Notch radius for all specimens
was 0.25 mm. Stable crack growth by ductile fracture mechanism is observed in all specimens, regardless of the shape of
the stress concentrator (fatigue pre-crack or machined notch).
Due to the cylindrical geometry, cyclic loading produced uneven cracks, i.e. the fatigue crack growth was more pro-
nounced on the interior surface of the specimen, left side of Fig. 4. In the same figure, it can also be seen that the plasticity
 W. Musraty et al. / Engineering Fracture Mechanics 175 (2017) 247–261 251

Fig. 4. Fracture surfaces of the fatigue pre-cracked (a) and notched (b) PRNB specimen [24].

and necking are more pronounced on the notched ring geometry, Fig. 4b. This can be attributed to the fact that larger defor-
mation is needed to initiate the crack at the notch tip (in comparison with the fatigue pre-cracked geometry).
It is discussed in [24] that machined notch is a suitable stress concentrator for this geometry and that it gives reliable and
repeatable results. The main problem with the fatigue pre-cracked ring specimens is the fact that the fatigue crack growth is
more prominent on the internal surface, which is one of the arguments to favour the notched specimen. The stable crack
growth (by ductile mechanism) from the machined notch does not show similar deviation, i.e. the crack does not exhibit
more pronounced crack growth on the internal surface. Therefore, it will be mainly used in the future examinations of
the rings cut from the pipes manufactured by different technologies. In addition to the fact that a more precise and straight
defect shape can be obtained by notch fabrication (when compared to fatigue pre-cracking), this further simplifies the prepa-
ration for testing.

3. Numerical analysis and material modelling

Numerical analysis is performed by the finite element method (FEM), using the software package Simulia Abaqus (www.
simulia.com). Three-dimensional models consist of 8-node full integration elements, and the mesh is shown in Fig. 5, includ-
ing the enlarged view of the region ahead of the crack front. Due to the symmetry of the geometry and loading, one quarter of
PRNB specimen is considered, i.e. two symmetry planes are applied. In the part of the ligament ahead of the initial crack
front, where the crack growth occurs, finite elements have the same size and similar dimensions in all directions (cubic-
shaped), in order to capture the crack growth adequately. The element size in this region is approx. 0.5  0.5  0.5 mm,
which will be discussed in detail in the following section. The models of all considered ring specimens were formed with
a pre-crack as stress concentrator, which will also be further commented.

Fig. 5. PRNB specimen - finite element mesh.

252 W. Musraty et al. / Engineering Fracture Mechanics 175 (2017) 247–261

Loading is applied through the contact with the rigid bodies, by prescribing the displacement of the loading pin in the
vertical direction and completely fixing the support, which corresponds to the experimental testing. Surface-to-surface con-
tact with a finite-sliding formulation is defined between the contact surfaces in both contact pairs.
Micromechanical material model is applied for determining the crack growth initiation and stable growth. Micromechan-
ical models, including those based on the Gurson yield criterion [28], take into account the influence of voids on the material
deformation and yielding, by extending von Mises plasticity theory. The porosity f (volume fraction of voids) is introduced as
the damage parameter into the expression for plastic potential; this expression corresponds to one of the most-often used
micromechanical models - the GTN or Gurson-Tvergaard-Needleman model [28–30]:


3Sij Sij  3q2 rm  2
/¼ þ 2q1 f cosh
½1 þ ðq1 f Þ  ¼ 0 ð1Þ
2r2 2r
where r is the current flow stress of the material matrix, rm is the mean stress, Sij is the stress deviator, q1 and q2 constitutive
parameters (in this work, values q1 = 1.5 and q2 = 1 are used), and f⁄ is the damage function, or modified void volume
fraction:
 
 f for f 6 f c fu
fc
f ¼ ;K ¼ ð2Þ
f c þ Kðf
f c Þ for f > f c fF
fc
K is parameter which describes how the load capacity loss develops after the critical void volume fraction (fc) has been
reached. This critical value of f is an important parameter of the model, and it corresponds to the onset of void coalescence.
 
The value of damage function at the moment of fracture is denoted as f u ¼ 1=q1 (in this work: f u ¼ 1=1:5 ¼ 0:67), while void
volume fraction at final fracture is denoted as fF. It should be noted that parameter q3 is sometimes included in the yield
2  2
potential, in the last term: q3 f instead of ðq1 f Þ . However, it is very often set as: q3 ¼ q21 , as in this work.
The GTN model has undergone many modifications until now. Also, it has been combined with other methods by some
authors; one such combination with the cohesive zone modelling was proposed in [31]. The modification of the GTN model
used in this work was proposed by Zhang et al. [32] - the Complete Gurson model or CGM. One of the main properties of this
model is the fact that it contains a criterion for the onset of void coalescence:


 
r1 1 b
> a
1 þ pffiffiffi ð1
pr 2 Þ ð3Þ
r r r
where r 1 is the maximum principal stress, a and b are constants fitted by Thomason [33] (a = 0.1 and b = 1.2). Instead of a
constant value, Zhang et al. proposed a linear dependence of a on the hardening exponent n, which is applied in the CGM.
The void space ratio r is given by [32]:
rffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
pffiffiffiffiffiffiffiffiffiffiffiffi
3 3f e e2 þ e3
r¼ e e1 þ e2 þ e3 ð4Þ
4p 2
e1, e2 and e3 are principal strains. Therefore, the critical void volume fraction fc can be calculated during the finite element
analysis, i.e. it is not a material constant in the CGM. If the GTN model is used, it is necessary to determine this value and use
it as material parameter for fracture assessment.
In some studies, e.g. [34–36], it is shown that critical void volume fraction depends on the stress state (triaxiality is
mainly considered), based on the unit cell analyses or testing the cylindrical geometries without a sharp stress concentrator,
such as notched tension specimen. However, Eq. (3) enables calculation based on the actual stress and strain state in each
finite element - the moment when fc is reached corresponds to the moment when this inequality is satisfied. As mentioned
before, this means that fc is not transferred from unit cell calculations or some other geometry. Also, it does not have to be
constant in a single model; variation of this parameter is shown in Results and Discussion section on the example of a ring
specimen, Fig. 14.
In the initial stage of ductile fracture of steel, the voids nucleate mainly around non-metallic inclusions. Therefore, the
initial porosity f0 is here set equal to the volume fraction of non-metallic inclusions fv, Table 4.
Another parameter will be commented here - fF or void volume fraction at final fracture, as very important for microme-
chanical analysis of the crack growth (i.e. ductile damage development) through the material. In general, it could be deter-
mined from microstructural observation of the fracture surfaces, but this is a very difficult and uncertain procedure. Zhang
[37] have shown, using the unit cell model, that fF depends on f0, and can be approximately calculated by Eq. (5). This value is
applied in the CGM, and this means that failure occurs when the porosity in an integration point within a finite element
reaches fF. This approach is chosen because it is based on the unit cell analyses where both initial damage (initial void vol-
ume fraction) and triaxiality have been varied, [37].

Table 4
Microstructural parameters of the material.

Volume fraction of inclusions fv [
] 0.0094

Mean free path between the inclusions k [lm] 578
 W. Musraty et al. / Engineering Fracture Mechanics 175 (2017) 247–261 253

f F  0:15 þ 2f 0 ð5Þ
The complete Gurson model is applied in the numerical analysis through user material subroutine (created by Zhang,
based on [32]).

4. Results and discussion

Ductile fracture of different PRNB geometries is analysed here experimentally and using the micromechanical model
(CGM). The application of the damage model enabled the assessment of the crack growth, through tracking the elements
ahead of the initial crack front which have lost their load carrying capacity. This can be seen in Fig. 6, where these elements
exhibit very low values of equivalent von Mises stress and form a typical crack shape.
The force - CTOD (crack tip opening displacement, obtained here using d5 concept [38]) curves for PRNB specimens are
shown in Fig. 7. It should be noted that three Ring specimens with radius 85 mm are tested, resulting in almost identical
curves; therefore, only one experimental line for this geometry is shown.
The diagram in Fig. 7 shows the influence of the ring radius and defect length. In [25], it is shown that these curves for
PRNB R85 and SENB specimens are very similar by shape, as well as by the highest CTOD values reached during the testing.
Here, a similar conclusion can be drawn for the ring specimens with different radii (60 mm), despite different shapes of the
stress concentrator (notch for R85 and pre-crack for R60 rings). Of course, the loading levels are different due to different
specimen sizes. The micromechanical model predicts somewhat more pronounced loss of the load carrying capacity, espe-
cially for the CTOD values exceeding 0.8 mm.
Stress triaxiality fields in front of the crack tip, at the moment of crack growth initiation, are shown in Fig. 8 for both PRNB
and SENB specimen. The model of SENB specimen is formed using symmetry with respect to the mid-thickness plane; hence
it contains one half of the crack front. It can be seen that all PRNB specimens (and also SENB specimen) have similar depen-
dences of triaxiality on the maximum principal strain, Fig. 9. These values, both triaxiality and maximum principal strain, are
tracked in the element in the middle of the initial crack front, i.e. in the position where the crack growth is initiated. Having
in mind that the triaxiality significantly affects ductile fracture, it can be said that the conditions in all specimens are rather
similar.
Crack resistance curves obtained by experimental testing are shown in Fig. 10a. Experimental curves for the pre-cracked
rings with radii R60 and R40, are obtained using the stereometric measurement along with normalization procedure, [39].
On the other hand, three sets of experimental data for R85 specimens in Fig. 10a correspond to three tested Ring specimens

Fig. 6. Equivalent von Mises stress field obtained by CGM.

254 W. Musraty et al. / Engineering Fracture Mechanics 175 (2017) 247–261

Fig. 7. F-CTOD curves for PRNB specimens - the influence of specimen and stress concentrator size.

with this size. These sets are characterised by two points each (the crack growth initiation and final failure); while the three
F-CTOD curves were almost identical, these points on CTOD-Da diagram are not coincident.
Micromechanical prediction of the crack growth curves is given in Fig. 10b. It can be seen that R60 and R40 specimens
have very similar curves. Somewhat lower crack resistance is obtained for the specimen with radius 85 mm, which is in
agreement with experimental results. It should be noted that the ratio of radius to wall thickness (R/B) is kept approximately
constant for all specimens analysed here; therefore, the thickness of the R85 ring is higher than the thickness of R40 and R60
rings and it has an effect on decrease of crack growth resistance. Isolated influence of wall thickness is also considered in
further text.
A comparison of the experimental results and prediction of the micromechanical model is given in Fig. 11, on the example
of two ring specimens with radius 60 mm.
From all the crack growth curves in Figs. 10 and 11, it can be seen that specimen radius, initial defect shape and size do
not have a very pronounced influence on the fracture behaviour. Regarding the simulations using the micromechanical
model CGM, it can be seen that the influence of changes of geometry (specimen radius and wall thickness, as well as crack
length) on the crack growth resistance predicted by the model corresponds to the experimentally determined trends, Fig. 10.
However, there is a difference between the experimental curves and those obtained by the model for early stages of crack
growth (from 0.2 to 0.4 mm), as shown in Fig. 11.
Measurement of the crack growth in experimental analysis is performed based on the ESIS (European Society for Struc-
tural Integrity) P2-92 procedure [40]. This means that the crack growth has been measured on fractured surfaces along nine
parallel lines (i.e. measurement positions), and Eq. (6) is used to determine the original and final crack length, a0 and af. After
these values are obtained, crack growth is calculated as their difference: Da ¼ af
a0 .

Fig. 8. Stress triaxiality field for a PRNB specimen and SENB specimen.
 W. Musraty et al. / Engineering Fracture Mechanics 175 (2017) 247–261 255

Fig. 9. Change of stress triaxiality with maximum principal strain for all ring specimens and SENB specimen.

( )
1 a1 þ a9 X 8
a¼ þ ai ð6Þ
8 2 i¼2

In the micromechanical analysis, the crack growth is determined by tracking the finite elements in the ligament which
have completely lost their load carrying capacity. This is accomplished by analysis of the damage parameter (porosity) ahead
of the initial crack front. Failure in an integration point within an element is defined as reaching the value fF - void volume
fraction at final fracture. In accordance with Eq. (5), this value is 0.15 + 2f0, i.e. around 0.17 for the analysed material.
Fig. 12a shows the porosity field at the moment of crack initiation and during the stable crack growth, while the increase
of porosity with maximum principal strain is shown in Fig. 12b. Porosity and principal strain are tracked in the element in
the middle of the initial crack front. The value fc, the critical porosity corresponding to the onset of void coalescence, is deter-
mined during FE analysis with CGM, and after it is reached the porosity continues to increase until the value fF.
The procedure for determining the crack growth is also based on [40], like in experimental analysis, except for the num-
ber of measurement positions which is set equal to the number of elements along the thickness (instead of nine positions
used in experiment). As mentioned before, reaching fF is the criterion for crack growth in the numerical model, and an exam-
ple of the crack front shape obtained by CGM is shown in Fig. 12a - failed elements are characterised by the porosity value
which reached fF.
For the micromechanical model, it is very important to set the appropriate finite element size in the region of crack
growth. All the models were formed with the finite element size, assessed as appropriate for this material (the range was
from 0.45 to 0.5 mm).

Fig. 10. Crack growth curves: experimental (a) and obtained using CGM (b).
256 W. Musraty et al. / Engineering Fracture Mechanics 175 (2017) 247–261

Fig. 11. Comparison of crack growth curves for rings R60 (experimental and obtained by CGM).

The element size at the crack front is selected based on the fitting with crack growth in standard fracture specimen (SENB
specimen), i.e. the FE size is transferred from SENB specimen to all ring geometries. Fig. 13 represents the influence of the
finite element size on F-CTOD curves obtained for SENB specimen, both experimentally and by the CGM. Of course, it can be
seen that the material model which does not include damage (von Mises) predicts much higher load carrying capacity, due to
the absence of crack growth. As for the results of the micromechanical model, finite element sizes below 0.5 mm (0.2 and
0.35 mm) result in more pronounced loss of load carrying capacity on this curve.
On the other hand, elements with dimensions larger than 0.5 mm lead to less pronounced crack growth in this material,
which is also not in accordance with the experimental results. This can be clearly seen by comparing the CTOD values cor-
responding to the fixed crack growth value Da = 0.7 mm, Table 5. The values predicted by the use of two element sizes (0.5
and 0.7 mm) differ significantly, and the latter overestimates the experimental value. Therefore, approximate element size
0.5 mm is selected as appropriate for the analysed material, and transferred to the ring geometries. Depending on particular
3D model geometry and direction, an element dimension may slightly vary from 0.5 mm, but in all considered geometries it
is within the range from 0.45 to 0.5 mm.
It turned out that the most suitable FE size approximates the value of the mean free path between non-metallic inclusions
(k, Table 4). The relation between the FE size and material microstructure, expressed through k, has been considered in many
studies, including [35,41–47]. In some of them, the finite element size is equal to k, as obtained here. However, this cannot be
taken as a rule; in the literature, the finite element size ranges from 1 to 6 times the mean free path. In the mentioned stud-
ies, especially [35,41], it has been concluded that these quantities are related, which means that FE size (as numerical param-
eter) depends on k (as microstructural parameter). An important outcome, which is applied here, is that the finite element
size can be regarded as material parameter and transferred from one geometry to another (i.e. once determined, it can be

Fig. 12. Porosity field (a) and change of porosity with maximum principal strain (b) for ring specimen.
 W. Musraty et al. / Engineering Fracture Mechanics 175 (2017) 247–261 257

Fig. 13. F-CTOD curves for SENB specimen - influence of FE size.

used for different geometries). An approach aimed at decrease, or elimination, of the mesh dependence would be the appli-
cation of non-local models, such as modification of the GTN model shown in [48].
It should be noted that relatively large FE dimensions (around 0.5 mm in the ligament) influenced the model formation
for the ring R85 geometry. The notch with 0.25 mm radius could not be defined using so large elements, so the model con-
tains a sharp crack instead.
As mentioned in the Section 3, the critical value of void volume fraction fc is calculated during the analysis if the CGM is
applied. Dependence of this value on distance ahead of the initial crack front is shown in Fig. 14, for PRNB specimen with
radius 60 mm. It can be seen that there is a variation (unlike in the ‘classical’ approach using the GTN - Gurson-
Tvergaard-Needleman model, where fc is a material parameter transferred from another geometry or unit cell calculations),
which is caused by the stress and strain field gradients in accordance with Eqs. (3) and (4).
The influence of the pre-crack shape is also examined, having in mind that the fatigue pre-cracking of PRNB specimens
typically resulted in uneven crack fronts ([24] and Fig. 4 - left hand side). For the analysed geometries, this deviation was up
to 10%. In the numerical analysis, three models are compared: with a straight crack front, and with an uneven crack front -
differences between the pre-crack length on the internal ring surface and on the external ring surface are 10 and 30%. The
crack growth (damage parameter field - porosity) in these geometries is shown in Fig. 15. It can be seen that 10% difference,
corresponding to the examined specimens, does not cause significant influence on the crack front shape. However, for the
third model there is a very slight difference which can be seen through more pronounced growth on the external surface.
A better measure of this difference is less pronounced loss of the load carrying capacity for the models with uneven crack
front, but only after the crack has advanced (Fig. 16, larger CTOD values).
The influence of the ratio W/B on ductile fracture prediction using the micromechanical model is considered in Fig. 17. The
comparison includes two models with the same radius (60 mm), the same ratio a/W, with varied ratio W/B - values 2 and 4
are examined. It can be seen that this influence is not pronounced, having in mind that large difference in W/B did not result
in significant difference between the curves. The crack growth resistance only slightly decreases with this increase of W/B
ratio.
The influence of thickness on fracture behaviour is also examined on the rings with radius 60 mm. In addition to the
geometry with B = 7.5 mm and W = 15 mm, a model with increased thickness (12 mm) is created, Fig. 18. It can be seen that
the length with the constant crack growth is larger, i.e. the region with dominating plane strain conditions is much wider.
If the F-CTOD curves for these specimens are compared, Fig. 19a, an important detail can be noticed: The drop of the load
carrying capacity for the thicker ring is more significant for high CTOD values. This can be attributed to more pronounced
constraint effect due to the thickness, which is also considered from the point of view of crack growth. The crack growth
curves in Fig. 19b present a decrease of resistance to crack growth of the specimen with increased thickness (B = 12 mm,
W = 15 mm). Therefore, it can be said that the constraint effect caused by wall thickness can be simulated by testing of
the ring specimens.

Table 5
Values of CTOD for crack growth Da = 0.7 mm in SENB specimen - influence of FE size.

SENB - exp. SENB – CGM (FE = 0.5 mm) SENB – CGM (FE = 0.7 mm)
CTOD [mm] (for Da = 0.7 mm) 1.08 1.21 1.76
1.17
1.32
258 W. Musraty et al. / Engineering Fracture Mechanics 175 (2017) 247–261

Fig. 14. Variation of the critical void volume fraction ahead of the initial crack front (PRNB R60 specimen).

Fig. 15. Crack growth for the straight and uneven crack front (a), and a photo of fatigue pre-cracked ring fracture surface (b).

Fig. 16. Influence of the crack front shape on F-CTOD curves for PRNB specimens.
 W. Musraty et al. / Engineering Fracture Mechanics 175 (2017) 247–261 259

Fig. 17. Influence of specimen width (W) on crack growth curves.

Further, if we now compare the crack growth curves for two thick-walled models with different ratio W/B (B = 12 mm,
with ratios W/B = 1.25 and W/B = 2), also shown in Fig. 19b, it can be seen that very similar trends are observed. Therefore,
the assessment of crack growth for the ring with increased thickness also does not significantly depend on the ring width W.
The work presented here is carried out on the specimens cut from the same steel plate, to ensure the same material state
for all of them. The future work will include the failure analysis of Ring specimens (with and without the fatigue pre-crack)
cut from the pipes - both seamless and seam, with different wall thickness. Examination of thin-walled pipes will be empha-
sised in the future work, having in mind that this study involved relatively high thicknesses of all ring specimens. Another
important reason for analysis of fracture in thin-walled pipes is the fact that it is rather difficult to fabricate standard fracture
mechanics specimens from them.
A limitation for the diameter of the pipe used for testing is determined by the available testing machine - maximum load
and space. However, it should be mentioned that the key quantities are not pipe dimensions, but ratios R/B, W/B and a/W.
Until now, most of the examinations were made on specimens with ratios R/B = 8, W/B = 2 and a/W = 0.4–0.7. Also, expres-
sions for plastic correction factor (gpl) have been proposed by Likeb et al. [23] for a certain range of the mentioned ratios, in
order to determine the J integral values in the same way as for SENB specimen, in accordance with the standard ASTM 1820
[19]. We are currently trying to prove that pipes with thinner walls can also be tested using this approach. In other words, we
are trying to extend the range of the tested ring geometries and verify the applicability on different pipe sizes.

5. Conclusions

Pipe ring notched bend (PRNB) specimens for determining the fracture resistance of pipes are examined experimentally
and by application of the micromechanical model (Complete Gurson model - CGM) for assessment of crack growth. Different
specimen sizes, as well as stress concentrator lengths and shapes (machined notch and fatigue pre-crack), are considered.

Fig. 18. Crack growth for two models with different wall thickness and otherwise identical geometry.
260 W. Musraty et al. / Engineering Fracture Mechanics 175 (2017) 247–261

Fig. 19. Influence of wall thickness on F-CTOD curves (a) and crack growth curves (b).

It is concluded, based on the presented results, along with previously obtained ones, that PRNB specimens are a good can-
didate for pipe material testing. Crack growth curves are not severely influenced by the specimen radius, defect shape and
defect initial size, provided that the ratio of specimen width and thickness is kept constant. The influence of the pre-crack
shape (uneven crack front), wall thickness, as well as specimen width, is assessed. Micromechanical analysis predicts the
change of fracture resistance with the change of specimen geometry and crack geometry; however, there are some devia-
tions from experimental results in the early stages of crack growth. Micromechanical parameters have been determined
by quantitative microstructural analysis and transferring from another geometry - single edge notch bend specimen, SENB.

Acknowledgement

The authors gratefully acknowledge the financial support from the Ministry of Education, Science and Technological
Development of the Republic of Serbia under the projects ON 174004 and TR 35006, as well as the financial support for
young researcher funding by Slovenian Research Agency. The authors would also like to thank Z.L. Zhang for the CGM user
subroutine.

References

[1] Dutta BK, Saini S, Arora N. Application of a modified damage potential to predict ductile crack initiation in welded pipes. Int J Pres Ves Pip
2005;82:833–9.
[2] Gubeljak N, Vojvodić Tuma J, Šuštaršić B, Predan J, Oblak M. Assessment of the load-bearing capacity of a primary pipeline. Eng Fract Mech
2007;74:995–1005.
[3] Marenić E, Tonković Z, Skozrit I. On the calculation of stress intensity factors and J-integrals using the submodeling technique. J Pres Ves Technol
2010;132(041203):1–12.
[4] Pati PK, Shrivastava SK, Basu S. Numerical analysis of crack initiation and growth in cylindrical geometries with an axial flaw. Int J Fract
2007;148:291–301.
[5] Zhang B, Ye C, Liang B, Zhang Z, Zhi Y. Ductile failure analysis and crack behavior of X65 buried pipes using extended finite element method. Eng Fail
Anal 2014;45:26–40.
[6] Xu J, Zhang ZL, Østby E, Nyhus B, Sun DB. Constraint effect on the ductile crack growth resistance of circumferentially cracked pipes. Eng Fract Mech
2010;77:671–84.
[7] Kim YJ, Shim DJ, Huh NS, Kim YJ. Plastic limit pressures for cracked pipes using finite element limit analyses. Int J Pres Ves Pip 2002;79:321–30.
[8] Huh NS, Kim YJ, Kim YJ. Limit load solutions for pipes with through-wall crack under single and combined loading based on finite element analyses. J
Pres Ves Technol - Trans ASME 2007;129:468–73.
[9] Dimić I, Arsić M, Medjo B, Stefanović A, Grabulov V, Rakin M. Effect of welded joint imperfection on the integrity of pipe elbows subjected to internal
pressure. Tech Gaz 2013;20:285–91.
[10] Arafah D, Madia M, Zerbst U, Beretta S, Cristea ME. Instability analysis of pressurized pipes with longitudinal surface cracks. Int J Pres Ves Pip
2015;126–127:48–57.
[11] Choi JB, Goo BK, Kim JC, Kim YJ, Kim WS. Development of limit load solutions for corroded gas pipelines. Int J Pres Ves Pip 2003;80:121–8.
[12] Chiodo M, Ruggieri C. Failure assessments of corroded pipelines with axial defects using stress-based criteria: numerical studies and verification
analyses. Int J Pres Ves Pip 2009;86:164–76.
[13] Oh CK, Kim YJ, Baek JH, Kim YP, Kim WS. Ductile failure analysis of API X65 pipes with notch-type defects using a local fracture criterion. Int J Pres Ves
Pip 2007;84:512–25.
[14] Medjo B, Rakin M, Arsić M, Šarkoćević Ž, Zrilić M, Putić S. Determination of the load carrying capacity of damaged pipes using local approach to
fracture. Mater Trans - JIM (Jpn Inst Metals Mater) 2012;53:185–90.
[15] Hadj Meliani M, Matvienko Y, Pluvinage G. Corrosion defect assessment on pipes using limit analysis and notch fracture mechanics. Eng Fail Anal
2011;18:271–83.
 W. Musraty et al. / Engineering Fracture Mechanics 175 (2017) 247–261 261

[16] Kozak D, Ivandić Z, Konjatić P. Determination of the critical pressure for a hot-water pipe with a corrosion defect. Mater Technol 2010;44:385–90.
[17] Šarkoćević Ž, Rakin M, Arsić M, Sedmak A. Fabrication of high strength seam welded steel tubes and quality indicator testing. Struct Integr Life
2008;8:81–98.
[18] Xu J, Zhang ZL, Østby E, Nyhus B, Sun DB. Effects of crack depth and specimen size on ductile crack growth of SENT and SENB specimens for fracture
mechanics evaluation of pipeline steels. Int J Pres Ves Pip 2009;86:787–97.
[19] ASTM E1820-15 - Standard test method for measurement of fracture toughness, American Society for Testing and Materials ASTM; 2015.
[20] BS 8571-14 - Method of test for determination of fracture toughness in metallic materials using single edge notched tension (SENT) specimens. British
Standards Institution BSI; 2014.
[21] Mahajan G, Saxena S, Mohanty A. Numerical characterization of compact pipe specimen for stretch zone width assessment. Fatigue Fract Eng Mater
Struct 2016;39:859–65.
[22] Koo JM, Park S, Seok CS. Evaluation of fracture toughness of nuclear piping using real pipe and tensile compact pipe specimens. Nucl Eng Design
2013;259:198–204.
[23] Likeb A, Gubeljak N, Matvienko Y. Finite element estimation of the plastic gpl factors for pipe-ring notched bend specimen using the load separation
method. Fatigue Fract Eng Mater Struct 2014;37:1319–29.
[24] Gubeljak N, Likeb A, Matvienko Y. Fracture toughness measurement by using pipe-ring specimens. Proc Mater Sci (20th European Conference on
Fracture ECF20, Trondheim) 2014;3:1934–40.
[25] Medjo B, Rakin M, Gubeljak N, Matvienko Y, Arsić M, Šarkoćević Ž, et al. Failure resistance of drilling rig casing pipes with an axial crack. Eng Fail Anal
2015;58:429–40.
[26] Gajdos L, Sperl M. Evaluating the integrity of pressure pipelines by fracture mechanics. In: Belov A, editor. Applied Fracture Mechanics. InTech; 2012. p.
283–310.
[27] Bergant M, Yawny A, Perez Ipiña J. Numerical study of the applicability of the g-factor method to J-resistance curve determination of steam generator
tubes using non-standard specimens. Eng Fract Mech 2015;146:109–20.
[28] Gurson AL. Continuum theory of ductile rupture by void nucleation and growth: part I - yield criteria and flow rules for porous ductile media. J Eng
Mater - Trans ASME 1977;99:2–15.
[29] Tvergaard V. Influence of voids on shear band instabilities under plane strain conditions. Int J Fract 1981;17:389–407.
[30] Tvergaard V, Needleman A. Analysis of cupe–cone fracture in a round tensile bar. Acta Metall 1984;32:157–69.
[31] Scheider I. Derivation of separation laws for cohesive models in the course of ductile fracture. Eng Fract Mech 2009;76:1450–9.
[32] Zhang ZL, Thaulow C, Odegard J. A complete Gurson model approach for ductile fracture. Eng Fract Mech 2000;67:155–68.
[33] Thomason PF. Ductile Fracture of Metals. Oxford: Pergamon Press; 1990.
[34] Koplik J, Needleman A. Void growth and coalescence in porous plastic solids. Int J Solids Struct 1988;24:835–53.
[35] Steglich D, Brocks W. Micromechanical modeling of damage and fracture of ductile metals. Fatigue Fract Eng Mater Struct 1998;21:1175–88.
[36] Benseddiq N, Imad A. A ductile fracture analysis using a local damage model. Int J Pres Ves Pip 2008;85:219–27.
[37] Zhang ZL. A complete Gurson model. In: Aliabadi MH, editor. Nonlinear Fracture and Damage Mechanics. Southampton: WIT Press; 2009. p. 223–48.
[38] GKSS. Displacement gauge system for applications in fracture mechanics. Patent publication. Geesthacht: GKSS Research Center Publications; 1991.
[39] Gubeljak N, Chapetti M, Predan J, Landes J. CTOD - R curve construction from surface displacement measurements. Eng Fract Mech 2011;78:2286–97.
[40] ESIS procedure for determining the fracture behaviour of materials. European Structural Integrity Society ESIS publication P2-92; 1992.
[41] Bron F, Besson J. Simulation of the ductile tearing for two grades of 2024 aluminum alloy thin sheets. Eng Fract Mech 2006;73:1531–52.
[42] Ruggieri C, Panontin TL, Dodds RH. Numerical modeling of ductile crack growth in 3-D using computational cell elements. Int J Fract 1996;82:67–95.
[43] Nonn A, Dahl W, Bleck W. Numerical modelling of damage behaviour of laser-hybrid welds. Eng Fract Mech 2008;75:3251–63.
[44] Rakin M, Medjo B, Gubeljak N, Sedmak A. Micromechanical assessment of mismatch effects on fracture of high-strength low alloyed steel welded
joints. Eng Fract Mech 2013;109:221–35.
[45] Penuelas I, Betegon C, Rodriguez C. A ductile failure model applied to the determination of the fracture toughness of welded joints. Numerical
simulation and experimental validation. Eng Fract Mech 2006;73:2756–73.
[46] Dutta BK, Guin S, Sahu MK, Samal MK. A phenomenological form of the q2 parameter in the Gurson model. Int J Pres Ves Piping 2008;85:199–210.
[47] Younise B, Rakin M, Gubeljak N, Medo B, Burzić M, Zrilić M, et al. Micromechanical analysis of mechanical heterogeneity effect on the ductile tearing of
weldments. Mater Design 2012;37:193–201.
[48] Linse T, Hütter G, Kuna M. Simulation of crack propagation using a gradient-enriched ductile damage model based on dilatational strain. Eng Fract
Mech 2012;95:13–28.

View publication stats