EFA-02604; No of Pages 13

Engineering Failure Analysis xxx (2015) xxx–xxx

Contents lists available at ScienceDirect

Engineering Failure Analysis

journal homepage: [Link]/locate/efa

Parameters affecting the weld defect acceptance criteria of clad submarine

pipelines installed by S-lay or reel-lay
Morten Hval ⁎, Trond Lamvik
Reinertsen AS, Leiv Eiriksson Senter, Postboks 6380 Sluppen, 7492 Trondheim, Norway

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

Article history: This case study presents a methodology to determine the criticality of defects in submarine pipe-
Received 16 January 2015 line girth welds made of clad carbon steel and to establish weld defect acceptance criteria. The
Received in revised form 23 June 2015 maximum target acceptance criteria are usually given by the applied design code often with addi-
Accepted 4 July 2015
tional requirements set by the client. ECA calculations are carried out in order to demonstrate that
Available online xxxx
the target criteria can be met using the relevant material data, geometry and load input. The ECA
methodology must follow the defect development from welding fabrication through pipeline in-
Keywords: stallation from the lay barge, as well as consider loads in the as-laid condition during pre-
Submarine pipelines
operational and operational phases until the end of design life and verify that leakage or failure
Clad material
will not occur.
ECA
S-lay © 2015 Elsevier Ltd. All rights reserved.
Reel-lay
Trawl interference
Global buckling
Vortex induced vibration
Defect acceptance criteria

1. Introduction and background

Many submarine pipelines systems are designed according to the DNV design code DNV-OS-F101 [1]. The code states how almost
all possible design scenarios shall be treated in order to maintain structural integrity through all phases of the design life. In general,
submarine pipelines can be subjected to different load scenarios from the welding fabrication through the installation phase and dur-
ing the operation phase to the end of the design life. This may include:

1. Large plastic bending strains during reel-lay (typically 1-2%)

2. Fatigue loads in the lay catenary during installation
3. Vortex induced vibration (VIV) fatigue loads in free spans at the seabed
4. Global buckling loads due to pressure and temperature variations
5. Interaction between fishing trawl gear or ship anchors during operations resulting in high local plastic strains.

In most cases each of these load scenarios alone may not be critical. However, when considering all scenarios together through the
whole lifetime of the pipeline, the initial weld defect indications may become critical in the end, when crack development from all
stages are included. Hence, all relevant load scenarios must be considered when determining the weld defect acceptance criteria.
Only recently a generally accepted methodology for ECA analyses of solid pipe material has been available, DNV-OS-F101 [1],
which is based on the principles given in BS7910:2005 [2]. In the latest revision of DNV-OS-F101, some general guidelines are

⁎ Corresponding author.
E-mail addresses: [Link]@[Link] (M. Hval), [Link]@[Link] (T. Lamvik).

[Link]
1350-6307/© 2015 Elsevier Ltd. All rights reserved.

Please cite this article as: M. Hval, T. Lamvik, Parameters affecting the weld defect acceptance criteria of clad submarine pipelines
installed by S-lay or reel-lay, Engineering Failure Analysis (2015), [Link]
2 M. Hval, T. Lamvik / Engineering Failure Analysis xxx (2015) xxx–xxx

Nomenclature

BS backing steel
D/t diameter to wall thickness ratio
CRA corrosion resistant alloys
ECA engineering criticality assessment
GMAW gas metal arc welding
OD outer diameter
SAC strained and aged with last strain cycle ending in compression
SAT strained and aged with last strain cycle ending in tension
SMTS specified minimum tensile strength
SMYS specified minimum yield stress
t/2 mid thickness location of defect
VIV vortex induced vibration
WT wall thickness
Δamax maximum ductile crack growth
amax maximum crack height

given also with respect to ECA and clad/lined pipe materials. This however, restricts the methodology to be used for clad pipes when
the weld is even-matched or over-matched relative to the carbon backing steel.
In order to extend the application also to partly under-matched weld cases, the finite element based program LINKpipe have been
applied in the ECA analyses. LINKpipe is developed by SINTEF and is a special purpose FE program for fracture mechanics analyses of
cracks and defect in pressurized pipeline girth welds [3]. The program is based on shell elements and uses line-spring elements for
simulating the fracture response in pipeline girth welds, whether driven by static loads, global buckling, VIV or corrosion. Ductile
crack growth through the thickness and along the length direction of the crack is included. The use of line-spring elements and the
theory behind it is demonstrated in Ref. [4]. LINKpipe has been limited to cases with relatively small deformations. Typical validation
cases are global bending of pressurized pipes with nominal strains up to 2–3%. The program has, however also demonstrated good
agreement with local buckling of cracked cylindrical shells [5]. For further description and background of the LINKpipe program in-
cluding validation cases vs. general purpose FE program such as ABAQUS, reference is given to Ref. [6]. Refs. [7] and [8] have also dem-
onstrated good agreement with full scale experiments. A more detailed description of the validation of the clad pipe implementation
in LINKpipe is given in Ref. [9]. LINKpipe also has an ANALYTICAL module that calculates according to BS7910:2005 that can be used
for direct comparisons between FE and analytical calculations for cases with single materials and uniaxial stress state.

2. Problem definition

The base case for this case study is a 16 in. OD pipeline internally clad with 3 mm 316 L stainless steel. Total wall thickness is
24 mm. The clad material is normally under-matched compared to the X60 carbon backing steel, something which must be consid-
ered in the ECA analyses. The pipeline weld was a narrow gap mainline girth weld with a cap and root width of 14.2 mm and
9.1 mm, respectively. If the installation method is S-lay, maximum bending strains over the stinger is in the range 0.25–0.3%. This
is considerably lower than for installation by reel-lay, which would have had maximum bending strains above 2%. When installed
on an uneven seabed free spans may occur that can result in VIV fatigue crack growth. During operation the internal hot, high pressure

Fig. 1. True stress strain curves at 115 °C for clad carbon steel (X60).

Please cite this article as: M. Hval, T. Lamvik, Parameters affecting the weld defect acceptance criteria of clad submarine pipelines
installed by S-lay or reel-lay, Engineering Failure Analysis (2015), [Link]
 M. Hval, T. Lamvik / Engineering Failure Analysis xxx (2015) xxx–xxx 3

Fig. 2. True stress strain curves at 115 °C for X65 carbon steel.

process fluid will heat the pipeline material and result in thermal expansion and global buckling. In order to control the buckling the
pipeline can be installed over buckle triggers. In this way the buckle location can be controlled and maximum buckle strains predicted.
Free spans will also be introduced on each side of the buckle trigger. A challenge at this location is VIV fatigue loads in the adjacent
free spans in combination with high local buckle strains and strain fluctuations at the buckle trigger due to pressure/temperature var-
iations during production shut-in/shut-downs.
If the pipeline is located in a fishery area, interaction with fishing trawl gear (over-trawling) may give local high plastic strains that
can result in leakage or fracture of the pipeline.
The below examples study the different possible load scenarios for a submarine pipeline made of 316 L clad carbon steel. In this
case the CRA material is metallurgically bonded to the carbon backing steel.

3. Material data

ECA analyses according to DNV-OS-F101 [1] require accurate description of material data. For strain-based analyses upper bound
material stress strain data shall be used for the backing steel and the clad material, while lower bound shall be used for the weld metal.
For installation analyses material data at ambient temperature are used, while for operational conditions the maximum operation
temperature must be used. The true stress strain curves at maximum operation temperature for the clad pipe case is shown in
Fig. 1. It can be seen that the clad material is under-matched compared to the backing steel and the weld metal, while the weld
metal is slightly under-matched at strains below approximately 0.5% and over-matched above. Material data as presented in Fig. 1
can be used for cases where no significant plastic strain is introduced during installation, such as for S-lay.
For reeling installation where plastic strains up to and above 2% can be introduced, the material stress strain response in operation
is influenced by the previous stress–strain history. To simulate the reeling strain cycles the material is strain cycled according to the
relevant reeling operation and aged at 250 °C for 1 h. One example is shown in Fig. 2 for X65 carbon steel. In the unstrained condition
both the lower and upper bound material stress strain curves are lower compared to the strained and aged curves (SAT and SAC). For
the strained and aged material with last strain cycle ending in tension (SAT), the material response has a high yield stress without any
significant plastic strain hardening. For the strain and aged material with last strain cycle ending in compression (SAC), a continuous

Fig. 3. Cumulative fatigue stress range distributions for the different phases.

Please cite this article as: M. Hval, T. Lamvik, Parameters affecting the weld defect acceptance criteria of clad submarine pipelines
installed by S-lay or reel-lay, Engineering Failure Analysis (2015), [Link]
4 M. Hval, T. Lamvik / Engineering Failure Analysis xxx (2015) xxx–xxx

yield curve is observed with a significantly lower yield stress, while the tensile strength for the two strain and aged conditions are
nearly similar.
The fracture toughness properties of the weld were characterized by CTOD-R curves determined by SENT tests according to DNV-
RP-F108 [10]. It appeared that the lowest tearing resistance was in the center of the weld compared to the fusion line which had a
significantly higher curve. The lower bound CTOD-R curve applied for all analyses were according to:

0:8
CTOD ¼ 0:38  Δa : ð1Þ

The lower bound curve was constructed by curve fitting of 6 SENT tests and adjusted down to intersect to lowest of the test result
points. The average CTOD-R curve was represented by the same equation, however with the constant 0.38 substituted with 0.48.

4. Design load input data

For the global buckling case the maximum strain value used in the ECA was determined by the detail design analyses which result-
ed in a maximum strain at the buckle trigger of 0.48%. This was considered to be rather high as it was well above the yield strain. The
value is, however assumed to be rather conservative since the global analyses were based on a lower bound yield stress strain curve.
The cumulative fatigue stress range distributions in the different phases are shown in Fig. 3. Both the installation fatigue stress and
the buckle fatigue stresses are in the same range while the VIV stresses are at a much lower level, however with a high number of
cycles.
The design load data were the result of an iterative process that was carried out in order to eliminate unnecessary conservatism in
the analyses and arrive at reasonable defect criteria for welding fabrication.
The trawl load case is taken from a previous study [11], with a 5000 kg clump weight at a velocity of 5 knots. Load response anal-
yses were done by means of ANSYS Version 10.0, in accordance with DNV RP-F111 [12]. General time histories for trawl pull-over ap-
plied as a point load are shown in Fig. 4. Model tests have shown that the clump weight can be represented as a quasi-static load, and
dynamic loading effects are not significant during the pull-over. First the clump weight stops in the collision, then the warp line is
tightened until the clump weight is rotated over the pipeline. However, the pipeline response may be dynamic, e.g., if global buckling
is triggered. Thus, the following parameters are governing for the pull-over duration; trawl velocity, pipeline induced movement at
interaction, warp line stiffness and clump weight. The effect of possible denting of the pipe due to the trawl impact itself is not con-
sidered. The maximum strain during over-trawling is also depending on the material stress–strain curve applied in the analyses. A
lower bound Analysis, using the lower bound stress–strain curve, results in significantly higher maximum strain than when using a
more realistic stress strain curve for the operational case, such as for instance strained and aged material curve with last strain ending
in tension or compression.

Fig. 4. Typical time histories for trawl clump weight.

Please cite this article as: M. Hval, T. Lamvik, Parameters affecting the weld defect acceptance criteria of clad submarine pipelines
installed by S-lay or reel-lay, Engineering Failure Analysis (2015), [Link]
 M. Hval, T. Lamvik / Engineering Failure Analysis xxx (2015) xxx–xxx 5

The maximum axial strain due to trawl pull-over is summarized in Table 1. The results show that the maximum strain when using
lower bound curve is significantly higher than the more realistic curves such as strain aged or upper bound stress–strain curves. Typ-
ical load response curves are shown in Fig. 5, for the most extreme cases; the lower bound material curve and the strain-aged (ending
in tension) curve. While the lower bound curve shows significant plastic strains after pull-over, the SAT curve shows almost purely
elastic behavior (blue curves).

5. Description of methodology

There have been several attempts in the pipeline community during the last years to establish a consistent methodology for strain
based ECA. The latest edition of DNV-OS-F101 [1] has included a procedure for assessing single material cases with strains below 0.4%.
For strains above 0.4% a more general description using FE model is given. During operation of high pressure/high temperature pipe-
lines maximum strain may easily exceed 0.4% for buckling or trawling cases, and ECA therefore usually has to be performed using FE
modeling. In addition, application of internal/external overpressure makes the use of FE analyses necessary in order to describe the
crack driving force accurately.
In the present cases the special purpose finite element program LINKpipe [3] was applied to study weld defect in pipeline girth
welds, i.e., defects perpendicular to the axial direction. Load input has been taken from ABAQUS and ANYS global pipeline analyses.
Modeling of residual stresses with LINKpipe deviates from the method given in DNV-OS-F101. Further description and justification
of the used implementation of residual stresses is given below in Section 5.1.
The intention has been to follow the development of a weld indication/defect from fabrication/NDT/repair through the relevant
load scenarios until end of life. The obvious requirement is that an initial weld indication in a pipeline girth weld shall survive the
whole lifetime with all the relevant loading stages (static or fatigue) of the pipeline without resulting in leakage or structural failure.
Only the phase after installation will be addressed in this paper. It was considered that the probability of having the worst case flaw
size experiencing the worst case installation scenario and at the same time be at the worst possible location during operation was very
low (less than 10−5), and hence, regarded as negligible.
The weld/clad/backing steel mismatch was modeled in LINKpipe as close as possible to the real geometry using the material true
stress–strain curves from Fig. 1 and the tearing resistance curve of Eq. (1) above with a weld geometrical misalignment of 1 mm.
No distinction was made in terms of tearing resistance for the various flaw locations, external, embedded or internal, backing steel
or clad layer. The lowest CTOD-R curve, which was the weld metal of the backing steel part, was therefore used for all of the analyses.

5.1. Residual stresses

The common approach to treat weld residual stresses as described in BS7910:2005 is to add a secondary stress equal to the yield
stress to the applied primary stress. When the primary stress approaches the yield stress it is allowed to relax the secondary residual
stresses, Qm, by a factor according to the following equation:

!
σ re f
Qm ¼ 1:4−  σY ð2Þ
σf

where σref, σf and σY are the local reference stress, the flow stress and the yield stress of the base material, respectively.
Weld residual stress equal to the yield stress reduces the elastic capacity of the cross section area, i.e., yielding occurs at loads
below the yield limit compared to a corresponding structure without residual stress. This approach does not work for LINKpipe. There-
fore, an alternative approach has been applied in LINKPipe where the residual stresses are implemented for the line-spring elements
only. In this way the residual stresses contribute directly to the crack driving force (CTOD). This method is strictly not in accordance
with DNV-OS-F101 [1]. However, several initial sensitivity analyses and comparison with calculations according to BS7910 performed
as basis for this study, showed that the method described above was considered to be the most appropriate method when using
LINKpipe. The theoretical basis for the LINKpipe weld residual stress model is further described in Refs. [3] and [13]. Relaxation of
the residual stress according to Eq. (2) is also possible when using LINKpipe. However, in order to maintain some degree of conser-
vatism this was not applied in the analyses.

Table 1
Summary of strains due to trawl pull-over.

Case/material Maximum axial strain [%] Maximum equivalent plastic strain [%]

1-LB 0.52 0.31

2-UB 0.40 0.16
3-SAT 0.30 0.02
4-SAC 0.41 0.19

Please cite this article as: M. Hval, T. Lamvik, Parameters affecting the weld defect acceptance criteria of clad submarine pipelines
installed by S-lay or reel-lay, Engineering Failure Analysis (2015), [Link]
 6
installed by S-lay or reel-lay, Engineering Failure Analysis (2015), [Link]
Please cite this article as: M. Hval, T. Lamvik, Parameters affecting the weld defect acceptance criteria of clad submarine pipelines

M. Hval, T. Lamvik / Engineering Failure Analysis xxx (2015) xxx–xxx

Trawl pull-over in free span - Lower Bound material curve Trawl pull-over in free span - SAT material curve
600 0.6 600 0.6
AXIAL STRESS AXIAL STRESS
500 0.5 500 0.5
AXIAL STRAIN AXIAL STRAIN

400 EQUIVALENT PLASTIC STRAIN 0.4 400 EQUIVALENT PLASTIC STRAIN 0.4
Stress [MPa]

Stress [MPa]
Strain [%]

Strain [%]
300 0.3 300 0.3

200 0.2 200 0.2

100 0.1 100 0.1

0 0.0 0 0.0

-100 -0.1 -100 -0.1

0 10 20 30 40 50 60 0 10 20 30 40 50 60
Time [s] Time [s]

Fig. 5. Stress and strain response in the pipeline due to trawl pull-over. Left: Lower bound material curve. Right: SAT material curve.
 M. Hval, T. Lamvik / Engineering Failure Analysis xxx (2015) xxx–xxx 7

5.2. Loading sequence

After being installed on the sea bottom the pipeline will be subjected to temporary loads before start-up of production. This period
can last up to 2 years and needs to be included in the analyses, particularly the VIV loads in free spans which may result in fatigue
crack growth. When production starts, the pipeline will be exposed to high pressure and temperature resulting in expansion and glob-
al buckling at the buckle triggers. At these locations the maximum strain of 0.48% will occur together with the VIV fatigue stress ranges
as well as the stress fluctuations from the planned and unplanned production shut-in/shut-downs (buckling fatigue). The loading se-
quence was as follows:

1. Temporary VIV fatigue

2. Axial stress and internal pressure
3. Buckling strains (bending)
4. Operational fatigue (buckling and VIV).

Trawl interaction may also occur during operation. The trawling case is performed in order to define the maximum defect size be-
fore the over trawling incident occurs. This will form basis of the maximum allowable crack growth due to static high strains or fatigue.
The loading sequence for the trawling case was:

1. Axial stress and internal pressure

2. Trawl-pullover strains (bending)

5.3. Definition of the end-of-life failure criteria

The coinciding occurrence of high maximum strain and high fatigue stresses may affect the fatigue life because high stress ranges
at high maximum strain (at yield level) can result in higher fatigue crack growth than what is obtained according to Paris' law which
in most cases is used for ECA. An alternative approach would be to use the CTOD-R relation to calculate the fatigue crack growth based
on ductile tearing in each stress cycle. This approach was considered to be very impractical and extremely time consuming. Instead a
low maximum allowable ductile tearing, Δamax, was specified as a “stop-criterion” in the analyses. Through sensitivity analyses this
was shown to give the ECA analyses sufficient conservatism. The selection of an appropriate Δamax is discussed in more detail below.
Another important stop criterion was that the cracks were not allowed to grow from the inside beyond the CRA-Carbon Steel in-
terface, resulting in a maximum allowable crack height of 3 mm equal to the thickness of the CRA layer. Furthermore, embedded de-
fects were not allowed to break the internal surface.

5.4. Target defect criteria

The aim of the ECA was to verify that the weld defect acceptance criteria were equal or better than the target defect criteria defined
by the workmanship criteria. The target defect criteria were defined according to Table 2. Particularly the internal weld root defect
criterion of 1.0 x 15 mm is very stringent due to the fact that fatigue cracks initiating from the weld root are not allowed to grow
through the CRA layer neither by fatigue nor ductile tearing.

6. Results

The first analyses were done without any trawl interaction. However, all other loads were included such as VIV in temporary phase
(before operation), application of internal pressure and expansion forces, buckling strain and VIV and buckling fatigue stress fluctu-
ations. Both inner and outer surface defects and embedded defects were studied.
The next analyses were carried out for the over-trawling case. This case included only the application of internal pressure and
bending strain due to over-trawling in a free span. The purpose of these analyses was to calculate the critical defect size as a function
of bending straining caused by the trawl clump weight and thereby be able to determine the margin available for crack growth by
ductile tearing or fatigue.

Table 2
Target defect acceptance criteria.

Defect location Maximum defect height [mm] Maximum defect length [mm] Comment

Near outer surface1 3.0 25 Surface breaking defect not allowed

Embedded ½ WT 3.0 50 –
Embedded BS/CRA interface 3.0 50 Defect located in BS
Internal root defect 1.0 15 Surface breaking defect not allowed

Note: 1: embedded defect is to be re-categorized to surface defect if the ligament is less than half the defect height.

Please cite this article as: M. Hval, T. Lamvik, Parameters affecting the weld defect acceptance criteria of clad submarine pipelines
installed by S-lay or reel-lay, Engineering Failure Analysis (2015), [Link]
8 M. Hval, T. Lamvik / Engineering Failure Analysis xxx (2015) xxx–xxx

6.1. Inner and outer surface defects

The results of the ECA for surface defects are shown in Fig. 6. The ECA curve for outer surface defects is very close, but crosses the
target defect size slightly below a defect size of 3 x 25 mm, while the ECA curve for the inner defects is well above. Calculations for
25 mm long defect resulted in a critical height of 2.86 mm. One of the reasons why the ECA curve is lower for the inner surface defect
is that defects are not allowed to grow by fatigue or ductile tearing through the 3 mm CRA layer, which was defined as one of the stop
criteria in the calculations. Crack growth from the inside into the carbon backing steel may rapidly result in failure because of the cor-
rosive process medium inside the pipeline. For the outer surface defect the second stop criterion, Δamax, was governing. Hence, longer
initial defects could be accepted.

6.2. Embedded defects

Results for the embedded defects are shown in Fig. 7. Both defect types are above the target defect size of 3 × 50 mm. The defect
type denoted “BS/CRA”, which is located with the tip of the defect at the backing steel/CRA interface, is lower than the defect located in
mid thickness (denoted “t/2”). The reason for this is that these types of defects are not allowed to grow by fatigue or ductile tearing
through the CRA layer and reach the inner surface. The fatigue margin will therefore be lower and, hence, the allowable defect size is
therefore also lower.

6.3. Over-trawling

The maximum bending strain as a result of over-trawling is depending on several parameters, where pipe dimensions (outer
diameter and wall thickness) are the most important. The maximum tolerable defect size is also depending on the crack growth
after welding, either due to ductile tearing or fatigue, and the maximum tolerable ductile tearing during the over-trawling in-
cident. The effect of defect size on ductile tearing and maximum strain during over-trawling is shown in Fig. 8. If as a starting
point maximum 1 mm ductile tearing is selected, the target defect size of 3 × 25 mm can resist about 0.95% strain. If a crack
growth margin of 1 mm due to ductile tearing and/or fatigue during installation/operation is included, the maximum strain
is reduced to about 0.8%. For the trawl case described above in paragraph 4, the maximum allowable defect height for a
25 mm long defect can be more than 10 mm. In this case defect growth due to ductile tearing and fatigue during both installa-
tion and operation must be included in the final assessment to determine the initial maximum acceptable defect size after
welding.

7. Discussion

The purpose of this ECA study was to verify the applied target defect criteria and to study the sensitivity of different input param-
eters. Due to the complex interaction between the different input parameters, it may be very difficult to predict the outcome of the
analyses or the individual effect of each parameter. The selection of all input parameters is as required by the code [1] conservative
“worst case” input values. However, when the conservatism of all input values is added together, very conservative results are expe-
rienced with a low probability of occurence. This paragraph presents the result of sensitivity analyses of the most important variables
related to ECA of clad carbon steel pipelines for the operational phase.

Fig. 6. Defect criteria for inner and outer surface defects.

Please cite this article as: M. Hval, T. Lamvik, Parameters affecting the weld defect acceptance criteria of clad submarine pipelines
installed by S-lay or reel-lay, Engineering Failure Analysis (2015), [Link]
 M. Hval, T. Lamvik / Engineering Failure Analysis xxx (2015) xxx–xxx 9

Fig. 7. Defect criteria for embedded defects located in mid thickness (t/2) or with tip of the defect located at the backing steel/CRA interface (BS/CRA) 3 mm from the
internal surface.

Ideally, probabilistic analyses should have been carried out. This would, however require a large amount of analyses and investi-
gations with respect to determining the statistical distribution of several parameters such as:

• Stress–strain curve for the different materials

• Geometrical weld misalignment
• Tearing resistance curve
• Location of defect around the circumference and in the weld
• Distribution of the static and fatigue loads

Sensitivity analyses are presented below to demonstrate the inherent safety levels in the analyses as a function of the most impor-
tant input parameters. Since outer surface defects seems to be the most critical defect type as compared to the target defect criteria,
most of the sensitivity analyses were carried out with an outer surface defect. Where relevant, also cases with inner surface defect are
included.

7.1. Influence of weld misalignment

The base case ECA analyses were carried out with a default weld misalignment of 1 mm. The default misalignment of 1 mm was
compared with an alternative eccentricity of 0.5 mm. Since the pipes had very good dimensional tolerances 0.5 mm misalignment
might be more realistic in many cases.
Fig. 9 shows the effect of comparing 1 mm eccentricity with 0.5 mm. At 25 mm defect length, the ECA curve is raised by about
0.5 mm. At longer cracks the difference is smaller. In case the misalignment turns out to be larger than 1 mm, which might happen
if the welding is performed outside the requirements, the opposite effect with lower ECA curve would be the result.

Fig. 8. Ductile tearing as a function of defect size during over-trawling.

Please cite this article as: M. Hval, T. Lamvik, Parameters affecting the weld defect acceptance criteria of clad submarine pipelines
installed by S-lay or reel-lay, Engineering Failure Analysis (2015), [Link]
10 M. Hval, T. Lamvik / Engineering Failure Analysis xxx (2015) xxx–xxx

Fig. 9. Effect of weld misalignment for outer surface defect.

7.2. Influence of circumferential flaw location

The maximum stress and the stress ranges due to buckling and in-line VIV occur at the buckle crown at 3 or 9 o'clock position
around the pipe circumference. If the weld defect is located offset compared to 3 o'clock (or 9 o'clock), the stresses and strains will
be reduced correspondingly. Fig. 10 shows the result if the flaw is located at 2 or 4 o'clock (or 8 and 10 o'clock). In this case the max-
imum strain and the stress ranges will be reduced by 13.4%. Fig. 10a shows that the effect of clock position is very large, in the range of
1.0 mm at 25 mm length for outer defects. The effect will be even higher for higher offset values. For inner defects, Fig. 10b, the dif-
ference is less significant (approximately 0.1 mm at 15 mm defect length).

7.3. Influence of fracture toughness

The base case analyses were performed with the lower bound CTOD-R curve given in Eq. (1). If the CTOD-R curve is replaced by the
average CTOD-R curve, where the constant 0.38 of Eq. (1) is replaced by 0.48, the ECA curve is significantly raised, as shown in Fig. 11.
The difference at 25 mm defect length is about 1.0 mm, and shows that the tearing resistance curve is a very important parameter.

7.4. Effect of maximum strain in buckle crown

Initial analyses carried out showed that the maximum strain in the buckle crown affected the ECA curve to a certain extent de-
pending on type of defect. In the present sensitivity analysis the maximum strain at the buckle crown has been reduced from 0.48%
to 0.4%, corresponding to a reduction of 17%. For outer surfaces defects as presented in Fig. 12, the ECA curve is markedly improved,
corresponding to more than 1.0 mm at the target defect length of 25 mm. Great effort was therefore put on having as accurate design
calculations as possible, since too conservative analyses giving too high maximum strain level will greatly affect the defect criteria.

7.5. Sensitivity of maximum allowable ductile crack growth, Δamax

Important parameters in the ECA are the “stop-criteria”, which determine when the analysis reaches a limit state. One of these
limit states are the maximum allowable ductile tearing, Δamax. This will determine the fatigue crack growth allowance for the initial
defects.

a) b)

Fig. 10. Effect of circumferential flaw location. a) Outer surface defect (left) and b) Inner surface defect (right).

Please cite this article as: M. Hval, T. Lamvik, Parameters affecting the weld defect acceptance criteria of clad submarine pipelines
installed by S-lay or reel-lay, Engineering Failure Analysis (2015), [Link]
 M. Hval, T. Lamvik / Engineering Failure Analysis xxx (2015) xxx–xxx 11

Fig. 11. Effect of fracture toughness. Lower bound vs. average CTOD-R curve.

The case with fatigue crack growth simultaneously with plastic deformation is a very special load case when end-of-life is consid-
ered. Normally, during reeling or over-trawling without simultaneous fatigue loads, maximum 1 mm ductile tearing is considered as
acceptable. Using up to 1 mm maximum ductile crack growth during operation when the pipe is bent up to and above yield with si-
multaneously occurring fatigue loads is not recommended, as this may give non-conservative results. The case can be illustrated as
shown in Fig. 13. In Fig. 13a high buckling stress ranges due to periodic shut-ins/shut-downs during operation occur while at maxi-
mum stress equal to yield stress. In addition to a high R ratio, the maximum stress is always at a high level caused by the global buck-
ling. It is doubtful that this case can be accurately described by Paris' law when the defect size increases and is approaching the critical
stage. In addition to the high stress ranges shown in Fig. 13a, also the VIV fatigue loads shown in Fig. 3 are superimposed, which can
make it even worse. In Fig. 13b the mean stress is 50% of the maximum stress and the fatigue crack growth can be reasonably well
represented by Paris' law. Too low Δamax values will on the other hand give too conservative results. A Δamax of 0.2 mm is based on
sensitivity trials and experience from similar applications and is considered to be a reasonable max value.
The sensitivity analyses demonstrated the effect of selecting alternative Δamax values from 0.1 mm–0.3 mm. For outer surface de-
fect the effect is considerable, Fig. 14. An increase from 0.2 mm to 0.3 mm increases the acceptable defect height for a 25 mm long
defect by more than 1 mm, while a reduction to 0.1 mm results in a very significant reduction of more than 1 mm.

7.6. Effect of weld metal mismatch

All analyses above were performed with the slightly under-matched weld metal stress strain curve in Fig. 1. For all practical pur-
poses, however, this curve is nearly even-matched up to about 0.5% strain from where it is increasingly over-matched. In order to
study the effect of weld mismatch, analyses were performed with 10% over-match and 10% under-matched weld metal. (Stress–strain
curve of given weld metal was reduced or increased by 10%.) The analyses were done for the trawl case with two different defect sizes;
3 × 25 mm and 6 × 25 mm outer surface defect. Fig. 15 shows that the effect is marginal below 0.5% strain for both defect sizes. For the
larger defect it appears that the under-matched weld has an increasing deviation from the even-matched and over-matched cases
when the strain is increasing. For the smaller defect, the ratio between the different curves seems to be about similar. Furthermore,
the welds with under-matched and even-matched weld metal have lower strain capacity than the over-matched weld metal. One rea-
son why the effect of weld under-match was not larger than this is the constraint effect due to the relatively narrow weld width. For
wider weld bead, the under-match effect would probably be larger.

Fig. 12. Effect of maximum strain at the buckle crown.

Please cite this article as: M. Hval, T. Lamvik, Parameters affecting the weld defect acceptance criteria of clad submarine pipelines
installed by S-lay or reel-lay, Engineering Failure Analysis (2015), [Link]
12 M. Hval, T. Lamvik / Engineering Failure Analysis xxx (2015) xxx–xxx

a) b)

Fig. 13. Illustration of high stress ranges with a) maximum stress equal to the yield stress (left) and b) mean stress level of Δσmax/2 (right).

8. Conclusions

ECA of pressurized bimetal pipelines with different material stress strain behavior in the clad material, the backing steel and the
weld metal has made it necessary to apply a finite element program such as LINKpipe to determine reliable and safe weld defect ac-
ceptance criteria. The present cases have shown that there are several input parameters that may greatly affect the defect criteria. It is
essential to have as accurate input data as possible. Particularly, it is very important to carry out the design analyses to establish ac-
curate input loads for the analyses (static stresses and strain as well as fatigue stresses during operation).
Global buckles with high static strain above yield level with superimposed fatigue stresses has been a demanding load case, since
standard Paris law data may give non-conservative results. To be on the conservative side a reduced maximum ductile crack growth to
0.2 mm was used as one of the “stop”-criteria in the analyses.
The load case with clump weight over-trawling in a free span was studied in order to determine the final critical defect size. The
case showed that for normal maximum strain levels during over-trawling up to 0.5% quite large defects sizes are tolerable. Further-
more, the crack growth margin available for ductile tearing and fatigue for a target defect of 3 × 25 mm is considerable since up to
more than 10 mm defect height is tolerable.
The main conclusion from the analyses was that even with conservative worst case selection of some of the input parameters (cir-
cumferential defect location, weld misalignment, maximum buckling strain, lower bound tearing resistance), it was possible to obtain
acceptable weld defect criteria. Sensitivity analyses showed that the inherent safety factor is large for many of the input parameters
when deviating slightly from the worst case input value. Ideally, a probabilistic approach should be used for the ECA. This would, how-
ever require statistical distributions of the most important input parameters and the use of a Monte Carlo simulation tool. Even if some
promising work related to probabilistic approach to ductile fracture has been performed [14,15], no recognized recommended prac-
tice is at present available for probabilistic ECA of submarine pipelines. A probability based methodology should, however be devel-
oped in order to establish weld defect criteria based on a given probability of failure.

9. Recommendations

It is difficult to give a specific or universal procedure for ECA of pipelines including design, fabrication and operation phases. Some
important items are, however outlined below:
1. Identify the critical load scenarios during installation and operation.

Fig. 14. Outer surface defect. Effect of Δamax.

Please cite this article as: M. Hval, T. Lamvik, Parameters affecting the weld defect acceptance criteria of clad submarine pipelines
installed by S-lay or reel-lay, Engineering Failure Analysis (2015), [Link]
 M. Hval, T. Lamvik / Engineering Failure Analysis xxx (2015) xxx–xxx 13

Fig. 15. Effect of weld metal mismatch.

2. In case of high plastic strains during installation such as during reel-lay, large ductile tearing may increase the initial defect size
considerably.
3. Fatigue crack growth either during installation in the lay catenary between the pipelay vessel and the sea bottom, or in the oper-
ation phase at the sea bottom either due to VIV in free spans or due to stress variations due to shut-ins/shut-downs, can restrict the
initial acceptable defect size.
4. Pipelines are also subjected to high internal pressure which results in a biaxial stress state in the welds. This may increase the crack
driving force and result in reduced fracture capacity. In cases with high plastic strains, FE based analyses will become necessary
since common analytical equations such as given in for instance BS7910 [2] may give non-conservative results. Alternative general
purpose FE programs to LINKpipe such as ABAQUS or ANSYS may also be applied. These programs will in most cases require sig-
nificantly longer computer time compared to LINKpipe and large sensitivity analyses may take unrealistic long time to complete
or may not be feasible in practical situations.
5. High strains during operation, when the pipeline is pressurized either due to pipeline buckling, trawl pull-over or a combination of
both may dramatically reduce the acceptable defect size particularly when the pipeline is subjected to fatigue at the same locations.
The stop criteria for the fatigue crack growth should be carefully selected.
6. It is advisable to evaluate the probability of the possible critical incidents occurring at the same locations. For instance what is the
risk of having the worst case defect size going through the worst fatigue scenario during installation and ending up in the worst
case location for VIV, pipeline buckling and trawl pull-over. Worst case input data for all parameters may result in very conservative
results. If possible, and if statistical distributions of the most important input parameters are available, a probability based approach
is recommended in order to obtain a more quantified probability of failure.

References

[1] Det Norske Veritas, DNV-OS-F101, Submarine Pipeline Systems, October 2013.
[2] BS 7910, Guide to Methods for Assessing the Acceptability of Flaws in Metallic Structures, 2005.
[3] LINKpipe Theory Manual, Rev. v3.0.2, Aug. 2012.
[4] E. Berg, B. Skallerud, C. Thaulow, Two-parameter fracture mechanics and circumferential crack growth in surface cracked pipelines using line-spring elements,
Eng. Fract. Mech. 75 (2008) 17–30.
[5] B. Skallerud, E. Berg, K.R. Jayadevan, Two-parameter fracture assessment of surface cracked cylindrical shells during collapse, Eng. Fract. Mech. 73 (2005)
264–282.
[6] LINKpipe Validation Manual, Rev. v3.0.2, Aug. 2012.
[7] E. Berg, E. Østby, C. Thaulow, B. Skallerud, Ultimate fracture capacity of pressurised pipes with defects — comparisons of large scale testing and numerical sim-
ulations, Eng. Fract. Mech. 75 (2008) 2352–2366.
[8] E. Østby, A.O. Hellesvik, Large-scale experimental investigation of the effect of biaxial loading on the deformation capacity of pipes with defects, Int. J. Press. Ves-
sel. Pip. 85 (2008) 814–824.
[9] E. Olsø, E. Berg, C. Thaulow, E. Østby, A new assessment approach for ECFA of clad and lined pipes based on shell and line-spring elements, The Proceedings of The
Twenty-first (2011) International Offshore and Polar Engineering Conference, ISOPE, 2011.
[10] DNV Recommended Practice DNV-RP-F108, Fracture Control for Pipeline Installation Methods Introducing Cyclic Plastic, Strain, January 2006.
[11] M. Hval, E. Olsø, L.L. Brækstad, Structural integrity of submarine pipelines subjected to large strains caused by trawl pull-over, The Proceedings of The Nineteenth
(2009) International Offshore and Polar Engineering Conference, ISOPE, Osaka, 2009.
[12] DNV Recommended Practice DNV-RP-F111, Interference Between Trawl Gear and Pipelines, October 2006.
[13] LINKftr, Usage Guidance: The Modeling of Weld Residual Stresses in LINKpipe, 2013. (dated 16.04).
[14] A. Sandvik, E. Østby, E. Berg, C. Thaulow, An efficient FE-based probabilistic model for ductile fracture assessment of pipelines with surface defects, The Proceed-
ings of the Twenty-first (2011) International Offshore and Polar Engineering Conference, ISOPE, Maui, Hawaii, USA, 2009.
[15] D. Zhou, A. Mirzaee-Sisan, Engineering critical assessment of embedded flaws in a pipeline using a probabilistic approach — a case study, Proceedings of the
ASME 2011 30th International Conference on Ocean, Offshore and Arctic Engineering OMAE2011, June 19–24, 2011 (Rotterdam, The Netherlands).

Please cite this article as: M. Hval, T. Lamvik, Parameters affecting the weld defect acceptance criteria of clad submarine pipelines
installed by S-lay or reel-lay, Engineering Failure Analysis (2015), [Link]