PREDICTION OF THIRD PARTY DAMAGE FAILURE FREQUENCY

FOR PIPELINES TRANSPORTING MIXTURES OF NATURAL GAS

AND HYDROGEN
Zhang, L.1, Adey, R.A.2
1
C M BEASY Ltd, Ashurst Lodge, Southampton, SO40 7AA, UK, [email protected]
2
C M BEASY Ltd, Ashurst Lodge, Southampton, SO40 7AA, UK, [email protected]

ABSTRACT
As Europe is gradually moving towards a hydrogen based society it has been acknowledged that
adding certain amount of hydrogen, as a clean energy carrier, to the existing natural gas pipeline will
help reduce the CO2 emissions which contribute to the greenhouse effect. On the other hand, hydrogen
has been demonstrated to be able to change the behaviour of the pipeline steel such as lower toughness
and faster crack growth due to hydrogen embrittlement. Therefore, it is necessary that the risks
associated with the failure of the pipeline carrying mixtures of natural gas and hydrogen be assessed.

The study reported in this paper is part of European NATURALHY project, whose aim is to
investigate the possibility of using the existing natural gas transmission pipelines to convey natural
gas/hydrogen mixtures. According to the EGIG database, the most common cause of failure for the
existing natural gas pipelines is third party damage, which mainly refers to a gouge, a dent/gouge
combination of known geometry. Among third party damage failures, 90% are the result of immediate
failure i.e. leakage or rupture of the pipeline and only 10% of them are the result of delayed failure.
While its not expected that hydrogen will impact the immediate failure it could increase the
vulnerability of the pipe to delayed failure through the initiation or activation of crack like defects.

This paper will present a methodology to predict the probability of increased failures and describe a
software tool that has been developed to perform the calculations.

Nomenclature
Pm primary stress Pb half crack length
H dent depth D nominal pipe diameter
σH nominal hoop stress t pipeline wall thickness
M bending moment ∆K th threshold stress intensity factor range
m fatigue growth parameter C fatigue growth parameter
a crack depth 2c crack length
W pipe section length K mode 1 stress intensity factor
Kr ratio of applied elastic K to K IC Lr ratio of applied load to yield load
K IC toughness of material ρ plastic correction factor
KP primary stress intensity factor KS secondary stress intensity factor
Lr max permitted limit of Lr σ ref reference stress
σY yield strength of material σU ultimate tensile strength of material
Pf probability of failure Kf failure frequency

1
1.0 INTRODUCTION

Among all failure modes for gas pipelines third party damage poses the greatest threat to the safety of
the pipeline network because it accounts for more than 50% of the total incidents according to a recent
survey [1]. Some of the failures do not happen immediately after the pipeline is damaged but occur
years after the strike. The aim of the current study is to investigate if adding hydrogen to the existing
natural gas pipeline network will increase the risk of delayed failure.

Europe is investigating the adding of Hydrogen to the gas transmission system as it is believed to be
one of the most promising energy carriers in the 21st century. However, the transition from pure
natural gas pipelines to pipelines carrying natural gas/hydrogen mixtures entails extra risk to the
integrity of the pipeline as hydrogen has been shown to reduce the toughness of the pipeline and to
accelerate crack growth. The NATURALHY project has concluded that hydrogen molecules within
the pipeline presents no hazard to the pipeline steel unless the inner thin oxide layer on the pipeline
steel is disturbed and a crack-like defect exists which can cause disassociation of the hydrogen
molecules and allow hydrogen to permeate into the pipe body, thereby reducing the material
toughness. Crack like defects may be present on the inner wall of the pipe as a result of manufacturing
defects, or as a result of some kind of third party damage. In the case of third party damage a crack
may be initiated on the inner wall of the pipe due to the incident. Alternatively the damage may
activate an existing dormant crack (i.e. a crack which is not growing or growing so slowly that it is not
significant). The second case would be very unlikely but it is theoretically possible.

The following sections propose a methodology to predict the impact of hydrogen on the delayed
failure probability associated with the dent where a crack has been initiated/activated due to third party
damage.

2.0 ESTIMATING THE MAXIMUM PERMISSIBLE CRACK DEPTH

2.1 Set-up of the dent-crack model

The dent in a pipeline is a complex structural problem as the stress distribution at the root of the dent
is heavily influenced by the depth of the dent, the included angle of the dent, the length of the dent and
the location of the dent. A detailed description of the elastic analysis of a dent on a pipe can be found
in references [2, 3]. To solve this complex problem a simplified model has been adopted. The loading
applied to the crack embedded at the bottom of the dent can be described by following relationships
[4].

Pm = σ H (1 − 1.8 H / D ) (1)

M = 0.85σ H tH (2)

Pb = 6 M / t 2 (3)

The only difference between a surface crack in an undamaged pipe and in a dented pipeline is that
there are additional stresses caused by bending around the dent but the membrane stress is lower. If the
bending stress is not relaxed when the dent is constrained by a rock, re-rounding is not possible. The
dent depth follows the Weibull distribution with α of 0.69 and β of 6.202 mm according to UKOPA
fault database [5]. For the current study, the dent depth is chosen to be 6mm as this represents a
typical dent depth in the pipeline (see Fig. 1).

2
 Dent depth Distribution
0.9 1.1
0.8 1

Cumulative Probability
0.9

Probability Density
0.7
0.8
0.6 0.7
0.5 0.6
0.4 0.5
0.3 0.4
0.3
0.2
0.2
0.1 0.1
0 0
0
0
0
0
0
0
0
0
0
0
.0
.0
.0
.0
.0
.0
0.
1.
2.
3.
4.
5.
6.
7.
8.
9.
10
11
12
13
14
15
Dent depth (mm)
Probability density Cumulative distribution

Figure 1. Dent depth distribution

The transmission gas pipeline considered is made of X70 of wall thickness 12.9 mm and diameter 900
mm. This represents a typical modern pipeline made after 1975. The geometry of the pipe, the dent
and the crack are shown in Fig 2. The operating pressure of the pipeline is assumed to be 60 bar. The
material properties for X70 pipeline in different natural gas/hydrogen mixtures are summarised in
Table 1. The data represents a typical X70 pipeline. It has to be noted that the pipelines manufactured
before 1975 might have significantly lower toughness values and more vulnerable to crack
propagation. In this case the figures shown in the table are not applicable and additional experimental
results are required to describe the material properties.

axial internal crack

Figure 2. Geometry of the pipeline with a crack embedded in a dent

3
 Table 1. Material properties of X70 (internal pressure = 60 bar)

Ultimate
Yield strength Toughness for
Gas contents tensile strength
(MPa) pipelines (MPa√m)
(MPa)
100% NG 482.6 565 200
50% NG+50%
482.6 565 150
H2
100% H2 482.6 565 100

The problem of a crack embedded in the dent can be solved using a flat plate model as shown in Fig.
3. W is the pipe section which is long enough to accommodate the crack defect. The crack is axially
oriented so that the hoop stress applied to the crack surface is the maximum. The pressure swing ratio
used here is 0.35, which is a fairly large number for an average gas pipeline network. However, this
number is used during the pipeline design process, so it is also used for fatigue calculation. Here the
failure of the pipeline refers to general failure, i.e. either leakage or rupture of the pipeline.

2c
a

W t

Figure 3. Geometry of the Semi-elliptical surface crack model

2.2 Calculation of the maximum permissible crack depth

For the current study BS7910 level 2 Failure Assessment Diagram (FAD) is adopted. As can be seen
from Fig 4 if a crack is initiated in the safe region it does not lead to failure immediately, but once it
grows outside the safety boundary, which is depicted on the graph, the pipeline becomes unstable and
leakage/rupture can occur. In this section a series of sensitivity tests are performed, for which the
maximum crack depth that will not cause immediate failure is calculated.

Brittle fracture
Kr dominated region

1.00
Kr =f(Lr )
0.80

0.60 Safe region End(fail)

Plastic collapse
0.40 cut-off
Start
0.20

0 0.20 0.40 0.60 0.80 1.00 1.20 1.40

Lr

Figure 4. BS7910 Failure Assessment Diagram (FAD)

4
According to BS7910 the failure boundary is defined as follows:

{
for Lr ≤ Lr max K r = (1 − 0.14 L2r ) 0.3 + 0.7 exp(−0.65L6r )

} (4)
for Lr > Lr max K r = 0

The failure criterion includes both brittle fracture and plastic collapse. K r measures the proximity to
brittle fracture and Lr represents the likelihood of plastic collapse. For BS7910 level 2A FAD, they
are given by:

 K P + KS
Kr = K +ρ
 IC
 (5)
 L = σ ref
 r σ Y

ρ is a parameter that takes plastic interaction between primary and secondary stress into
consideration. For materials, especially low strength materials that exhibit a yield discontinuity
(Lüders plateau) Lr is restricted to 1.0 [12]. Otherwise it is calculated through:

σY + σu
Lr max = (6)
2σ Y

The above equation is used for the analysis of the current study as the Lüders plateau is not visible for
high strength steel such as X70. The results of the sensitivity tests are displayed in Fig. 5. It can be
seen that for modern pipelines with a relatively high toughness hydrogen does not have much impact
on the integrity of the pipeline when the cracks have just formed in the dent. For these pipelines the
failure mechanism is plastic collapse rather than brittle fracture. It is also discovered that the critical
depth (i.e. the crack depth which will lead to immediate failure) declines as crack length increases and
the critical crack depth levels off as the crack length reaches about 130 mm. However, for old
pipelines both brittle fracture and plastic collapse could be the failure mechanism.

Although preliminary sensitivity tests suggest modern pipelines with a high toughness are less
susceptible to hydrogen it will be shown that once the crack starts growing in the dent the associated
risks will increase as time elapses. Replacing part of natural gas with hydrogen does make pipelines
vulnerable and the severity of the damage depends on the content of hydrogen.

5
 Critical crack depth vs. crack length
6

5
crack depth (mm)
4

0
10 30 50 70 90 110 130 150 170 190 210 230 250 270 300
crack length (mm)

X70 pipeline in NG, 50%H2, 100% H2

Figure 5. Critical crack depth vs. crack length for an X70 pipeline (60 bar, dent depth=6mm)

3.0 ESTIMATION OF THE DELAYED FAILURE PROBABILITY OF DENTED PIPELINES

3.1 Crack growth in the dent

Among all third party damage related failures, 90% of them are immediate failures and the rest are
delayed failures [7]. Delayed failures refer to those that do not occur when the pipeline is damaged but
occur years after it was damaged. From the sensitivity test results as seen in the previous section it can
be concluded that any cracks present in the pipeline must be smaller than those shown in Fig. 5 as any
larger cracks would have led to immediate failure.

Since the cumulative probability of failure over a given timeframe is required, crack propagation due
to cyclic loading must be included to estimate how the growth of cracks will impact the number of
failures. The Paris law with a threshold ∆K th is selected to calculate the crack length and depth with
regard to the corresponding number of cycles.

da 0 for ∆K <∆K th

= (6)
dn C ∆K m for ∆K ≥ ∆K th

The actual calculation is performed by estimating the amount of crack growth during a loading cycle
and the detailed procedures can be found in authors' another paper [8]. According to Paris equation:

∆a = C (∆K a ) m
(7)
an +1 = an + ∆a

where an corresponds to the crack depth after n load cycles, C and m are fatigue growth
parameters.

3.2 Procedures for estimating the failure probability

The failure probability can be calculated through Monte-Carlo simulation. By generating a large

6
number N of independent repetitions, the probability of failure can therefore be estimated as the
quotient of the failure counts to the number of simulations performed, which is given as follows:

Nf
Pf =
N

where N f is the number of failures recorded. The whole process is shown in Fig. 6. The flow chart is
simplified as the real simulation involves fatigue calculation, inspection and repair program. Relevant
information can be found in reference [8]. Because crack propagation can lead to fracture or leakage
of the pipeline after a certain period of time, Pf is a function of load cycle n .

Pf = P(n) (8)

Start

No Generate
End analysis random
variables?

Yes

Yes
Fail? Nf=Nf+1

No
No failure

Figure 6. Flow chart of Monte-Carlo simulation

Pf denotes the cumulative probability which monotonically increases with load cycles. This value is
only associated with the dent which has a crack at the bottom of it and the crack is very small which
does not lead to failure at the beginning. As we do not know how many cracks are associated with
dents, we therefore have assumed that all dents under consideration contain a surface crack.

If we assume that all dents have a depth of 6mm. The failure frequency of a pipeline arising from such
defects in year 'u' can be approximated by:

u Pf ( i + 1) − Pf ( i )
k f (u ) = ∑ ×k (9)
i =0 1 − Pf ( i )

where k is the incident rate per km per year (kmy) and u for year. According to UKOPA database, the
dent and gouge related incident rate is 8.49×10-4 per kmy, but the dent related incident rate is around
1.68×10-4 per kmy. Equation 9 implies that the defects which do not lead to failure in previous years
may contribute to the failure frequency in the future.

7
3.3 Distributions of defect dimensions

The biggest difficulty in predicting the failure probability arises from the fact that information on the
distribution of cracks imbedded in the dent is not available because currently inspection tools are not
used to detect cracks in gas pipelines. Also there are many factors which influence the probability of
failure based on the assumed incident scenario. For example if the POF of a single dent is considered
the analysis must consider the severity of the dent, the probability that a crack is initiated on the inside
wall of the dented pipe, the probability that there is a pre existing crack on the inside wall of the pipe
near the dent and the frequency of such a dent causing incident occurring. In addition there is also the
scenario regarding the introduction of hydrogen gas mixtures into the pipeline. For example has the
pipeline been inspected and repaired? Was the dent pre-existing? Or did it occur after the introduction
of the hydrogen gas mixture etc.

While calculation can be performed for the different scenarios the results will be very much dependent
upon the defect distribution assumed in the pipeline. The following section attempts to simulate the
POF for the case where the pipeline has been previously operating safely with natural gas and now is
operating with a hydrogen gas mixture and the dent occurs and a crack is present on the inside wall
either because it is initiated or as a consequence of the dent was pre existing in the pipe at that
location.

Based on the sensitivity test results shown in section 2, it is possible to estimate the distribution of the
length and depth of the crack imbedded in a dent which is 6mm deep. It is proposed that the following
log-normal distributions for initial defect depth and length in dents are used to calculate the failure
probability. The estimation is based on the findings of the sensitivity tests which show shorter cracks
generally have larger critical depth up to a certain crack length where the critical depth stabilises. The
log-normal cumulative curves, which give the probability of the occurrence of a specified size of
damage or smaller, are plotted for crack depth and crack length for a long but shallow crack in Fig. 7
and Fig. 8. Fig. 9 and Fig. 10 are for short but deep cracks. Examining the curves it can be seen from
Fig. 7 and Fig. 8 that they have been designed to have a very small number of defects with a depth
greater than 1.5 mm which was the critical crack depth for a long crack in a dent in the pipeline. While
there are still a small number of defects above the critical crack depth they are not significant as they
only lead to a small increase in the POF in the early years.

Table 2. Distributions of crack depth and length

Standard
Type of Mean
Variable deviation
distribution (mm)
(mm)
Depth (long cracks) Log-normal 0.8 0.2
Length (long cracks) Log-normal 15 5
Depth (short cracks) Log-normal 2 0.3
Length (short cracks) Log-normal 5 2

8
 C rack depth distribution
2.5 1.1
1

Probability Density
2 0.9
0.8

Cumulative
Probability
1.5 0.7
0.6
0.5
1 0.4
0.3
0.5 0.2
0.1
0 0
01
11
21
31
41
51
61
71
81
91
01
11
21
31
41
51
61
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
1.
1.
1.
1.
1.
1.
1.
Depth (mm)

Figure 7. Crack depth distribution (long cracks)

Crack length Distribution

0.1 1.1
0.09 1
Probability Density

0.08 0.9
0.07 0.8

Cumulative
Probability
0.06 0.7
0.6
0.05
0.5
0.04 0.4
0.03 0.3
0.02 0.2
0.01 0.1
0 0
01
01
01
01

15 1
18 1
21 1
24 1
27 1
30 1
33 1
36 1
39 1
42 1
45 1
48 1
1
.0
.0
.0
.0
.0
.0
.0
.0
.0
.0
.0
.0
.0
0.
3.
6.
9.
12

Length (mm)

Figure 8. Crack length distribution (long cracks)

Log-Normal Distribution
1.6 1.1
1
Cumulative Probability

1.4
Probability Density

0.9
1.2 0.8
1 0.7
0.6
0.8
0.5
0.6 0.4
0.4 0.3
0.2
0.2 0.1
0 0
0. 1
22

0. 3
0. 4
85

1. 6
27

1. 8
69

2. 9
2. 1
32

2. 3
2. 4
95

3. 6
37
1.
0

4
6

5
7

1
0.

0.

1.

1.

2.

3.

Depth (mm)

Figure 9. Crack depth distribution (short cracks)

9
 Log-Normal Distribution
0.3 1.1
1

Probability Density
0.25 0.9
0.8

Cumulative
Probability
0.2 0.7
0.6
0.15
0.5
0.1 0.4
0.3
0.05 0.2
0.1
0 0
01
21
41
61
81
01
21
41

10 1
12 1
13 1
14 1
15 1
16 1
18 1
19 1
1
6
.8
.0
.2
.4
.6
.8
.0
.2
0.
1.
2.
3.
4.
6.
7.
8.
9.
Length (mm)

Figure 10. Crack length distribution (short cracks)

3.4 Comparing the delayed failure probability for natural gas pipelines and pipelines carrying
natural gas/ hydrogen mixtures

Based on the assumed defect distributions and the dent size the simulation results for a single dent are
shown in Table 3 and Table 4. It is obvious that the delayed probability of failure for pipeline carrying
hydrogen is much higher than that for the existing natural gas pipelines. It has to be noted that as the
defect size distributions proposed here are for a specific X70 pipeline and the failure probabilities will
change if a different pipeline is analysed. However, it can be expected that the general trends indicated
by these results will remain the same for other pipelines.

Table 3. Cumulative POF values for pipelines carrying different content of Hydrogen (a long but
shallow crack initiated at the bottom of a dent)

50% NG-
Year 100% NG 100% H2
50% H2
0 1.54E-08 1.54E-08 9.92E-06
1 1.75E-08 3.70E-08 2.77E-04
2 2.32E-08 9.00E-08 2.77E-03
3 2.57E-08 2.16E-07 1.47E-02
4 3.47E-08 4.86E-07 5.94E-02
5 4.05E-08 9.48E-07 1.43E-01
6 5.85E-08 1.91E-06 2.47E-01
7 7.13E-08 4.96E-06 3.67E-01
8 8.07E-08 1.12E-05 4.78E-01
9 9.58E-08 2.07E-05 5.55E-01
10 1.26E-07 3.64E-05 6.22E-01

Table 4. Cumulative POF values for pipelines carrying different content of Hydrogen (a short but deep
crack initiated at the bottom of a dent)

50% NG-
Year 100% NG 100% H2
50% H2
0 1.21E-08 1.21E-08 3.39E-05
1 1.36E-08 1.47E-08 2.44E-04
2 1.39E-08 2.48E-08 1.10E-03
10
 3 1.41E-08 6.96E-08 2.38E-03
4 1.45E-08 1.41E-07 3.70E-03
5 1.48E-08 2.30E-07 4.82E-03
6 1.76E-08 3.30E-07 5.71E-03
7 2.02E-08 7.79E-07 6.47E-03
8 2.46E-08 1.46E-06 7.69E-03
9 2.61E-08 2.63E-06 9.51E-03
10 3.58E-08 5.73E-06 1.15E-02

By comparing the results in Table 3 and Table 4 it can also be discovered that for very short cracks in
the pipeline the associated risks are lower than those for long defects even if the short cracks are
deeper. Since the short cracks are less dangerous than the long cracks and we do not know the real
distribution of crack sizes in the dent, the results shown in Table 3 are used to compute the failure
frequency. However, adopting this approach will result in conservative results as the small cracks are
implicitly treated as long cracks and thereby having the same level of contribution to the total failure
frequency as the longer cracks do.

The failure frequency can be obtained by solving equation 9. Fig. 11 shows corresponding failure
frequency curves for the same X70 pipeline. It has to be noted that it was assumed the dent depth is
6mm, but in fact this is not accurate as the dent depth does follow certain distribution. However,
through this simplification the simulation results have already shown that adding hydrogen will
increase the delayed failure frequency but it is difficult to provide an accurate estimate without more
accurate data on the type of cracks initiated. Obviously the results show that reducing the amount of
hydrogen added to the natural gas pipeline system reduces the risk but if inspection tools are available
which can detect crack like defects the target risk level could be achieved through a carefully designed
integrity management programme.

Failure Frequency vs. Hydrogen Content

1.0E+00

1.0E-01
Failure frequency (1000km.year)

1.0E-02

1.0E-03

1.0E-04

1.0E-05

1.0E-06

1.0E-07

1.0E-08

1.0E-09
0 1 2 3 4 5 6 7 8 9 10
Year
100% NG 50% NG - 50% H2 100% H2

Figure 11. Failure frequency for a cracked dent in pipelines carrying NG and NG/H2 mixtures

4.0 CONCLUSIONS

A methodology has been presented to predict the probability of failure of a pipeline subjected to third
party damage is presented. It is based on the concept of the damage either initiating a crack like defect

11
or activating a dormant crack like defect on the internal wall of the pipeline thus leading to delayed
failure. The procedure is based on a probabilistic fracture mechanics approach.

The following conclusions have been drawn with regard to the determination of the failure probability
for transmission pipelines conveying natural gas/hydrogen mixtures.

1) From the initial calculations the results suggest that the existence of hydrogen in a pipeline
significantly increases the delayed failure probability. The reason for this is that the hydrogen
penetrates the parent material through the crack surface and changes the material properties,
which reduces the ability of the pipeline to resist crack growth.

2) An important element of the POF calculation is the proposed distribution of crack sizes located
near the dent. Since no data is available on such distribution, some assumptions based on a series
of sensitivity tests are made. Further work is required to obtain data to support further
predictions.

3) Other parameters such has pressure drop ratio, number of cycles and the hitting rate are either
taken from literature or design guidelines. Hence, when interpreting the failure frequency results
one must realise that the failure frequency may vary significantly from pipeline to pipeline.

4) In this report, only lognormal function is adopted to represent the actual distribution of the defects
in pipeline. However, there are other functions such as Weibull and exponential distributions,
which can also be used to fit the data. In addition, the crack length and depth should be examined
very carefully since the results are very sensitive to these inputs.

ACKNOWLEDGEMENTS

The authors would like to thank the NATURALHY project partners for providing data and the
European Union for kindly providing the financial support for this study.

REFERENCES

1. Gas Pipeline Incidents, 6th EGIG report 1970-2004, EGIG 05.R.0002. 2005.
2. Seng, O.L., C.Y. Wing, and G. Seet, The Elastic Analysis of a Dent on Pressurised Pipe.
International Journal of Pressure Vessels and Piping, 1989. 38: p. 369-383.
3. Orynyak, I.V. and Y.S. Yakovleva, Application of the crack compliance method to long axial
cracks in pipes with allowance for geometrical nonlinearity and shape imperfections (dents).
Engineering Fracture Mechanics, 2008. 75: p. 4052-4065.
4. Bai, Y. and R. Songb, Fracture assessment of dented pipes with cracks and reliability-based
calibration of safety factor. International Journal of Pressure Vessels and Piping, 1997. 74: p. 221-
229.
5. Lyons, C., et al. A methodology for the prediction of pipeline failure frequency due to external
interference. in International Pipeline Conference. 2008.
6. Guide to methods for assessing the acceptability of flaws in metallic structures (BS7910). 2005:
British Standards Institution.
7. Kiefner, J., Pipeline Incidents Caused by Mechanical Damage, in Mechanical damage technical
workshop. 2006.
8. Zhang, L., Adey, R, Predicting the Probability of Failure of Gas Pipelines Including Inspection
and Repair Procedures. in International Conference on Hydrogen Safety 2007. 2007

12