You are on page 1of 19

Pipeline Technology Conference, Ostend, 12-14 October 2009

Paper no: Ostend2009-008

Fracture arrestability of high


pressure gas transmission
pipelines by high strength
linepipes
by Hiroyuki Makino1, Izumi Takeuchi2, and Ryota Higuchi1
1 Corporate R&D Laboratories, Sumitomo Metal Industries Ltd, Amagasaki,
Hyogo, Japan
2 Tokyo Head Office, Sumitomo Metal Industries Ltd, Tokyo, Japan

U NCERTAINTIES OF KEY ITEMS which influence the arrest toughness prediction are investigated
because it is difficult to evaluate properly the crack arrestability of high strength line pipes by
existing models. Based on the investigation, Sumitomo model, which is HLP method with newly
developed crack velocity curve, is proposed as a new predicting model of fracture propagation and
arrest behavior in high pressure gas transmission pipeline. It is confirmed that accuracy of Sumitomo
model is improved from that of existing models.

Based on prediction by Sumitomo model, X100 full-scale fracture propagation test has been planned and
conducted. X100 grade pipes with 36 inch diameter and 19.05 mm wall thickness are pressurized to
22.07 MPa corresponding to 76.8 % SMYS (Specified Minimum Yield Strength). Different from past full-
scale fracture propagation tests, all pipes used in this test had almost same toughness to prove the effect
of toughness arrangement to arrest length of crack propagation. The test has been successful in
demonstrating the arrest of propagating ductile fracture within a short distance on both side. The test
has proved that X100 pipeline has a chance to have enough intrinsic crack arrestability even at high
pressure of over 20 MPa.

T HE WORLD ENERGY demand is continuously increasing. The environment preservation has


become a more important and urgent issue for our life in twenty first century. In this context,
energy efficiency and less CO2 emission is getting more focused than ever before. Natural gas, which is
known as clean energy due to less CO2 emission and less pollutant, should increase its share in primary
energy in the world. The large reserve of natural gas is generally far from the market. Therefore, the
efficiency of natural gas transportation is one of the key factors to increase the usage of natural gas.

High pressure gas transmission pipeline by high strength line pipes offers gas companies the opportunity
to reduce total cost of supplying gas to market. In response to this market demand, steel makers have
successfully developed ultra high strength line pipes such as X100 or X120 [1]. By improvement of steel
quality the toughness of such pipes also has been improved. High toughness ultra high strength line
pipes are now available in commercial basis.

Before the application of such high strength pipes for gas transmission pipelines, their structural
reliability should be clarified. Gas transmission pipelines must be designed to ensure the crack arrest
together with prevention of crack initiation. Therefore, toughness requirement for arrest of ductile

This paper is © copyright of the organizers of the Pipeline Technology Conference held in Ostend, Belgium, on 12-14 October, 2009, and of
the author(s). No copying (either electronic or otherwise), transmission electronically, or reproduction in any way is permitted without the
specific permission of the copyright holders.
2 Pipeline Technology Conference, Ostend, 12-14 October 2009

fracture, which is a unique fracture mode in high pressure gas pipelines, should be investigated. The
prediction of arrest energy is accepted widely in case of moderate pressure pipelines by conventional
grade line pipes. However, recent works indicate that for these high strength pipes applicability of
existing model is questionable and only way to overcome this problem is to derive full-scale experimental
evidence. And, it is thought that intrinsic arrest of these high strength pipes is hard to achieve.

Analysis of the arrest toughness


Predicting methods for the arrest toughness

Traditionally, pipelines are designed such that the toughness in pipe body is sufficient to intrinsically
arrest a propagating crack. Therefore, determination of the required toughness for the arrest of
propagating ductile fracture has been the focus of much research over the past thirty years. Many models
and formulae have been developed for determination of the required toughness. Among these models,
the Battelle Two Curve (BTC) method [2] and the Battelle simplified equation [3] are the most used
predicting methods.

In order to predict the toughness requirement, the BTC method takes a comparison of curves expressing
variation of crack velocity and of gas decompression velocity with pressure. If no intersection exists
between the crack and the gas decompression velocity curves, gas decompression velocity exceed crack
velocity for all pressure levels, the pressure at crack front will decreases below the arrest pressure and the
crack will arrests. On the other hand, if an intersection exists between these two curves, the pressure
level where crack and gas decompression run together at equal velocity exists, no further decrease of the
pressure at crack front is possible and the crack will continue to propagate. Thus, the tangent condition
between these two curves represents the boundary between arrest and propagation, and the
corresponding toughness level is referred to as the arrest toughness by this BTC method.

The Battelle simplified equation, expressing the arrest toughness in terms of hoop stress, diameter and
wall thickness, was developed on the basis of a statistical fit to calculated results from the BTC method.
Because of ease for application, this Battelle simplified equation also has been widely used, but there is
not a wide application of its background. This equation was originally presented [3] as having been
developed for an initial acoustic velocity of 396 m/s, the specific heat ratio of 1.4, yield strength from
414 to 552 MPa, hoop stress from 60 to 80 % of SMYS and a (Dt/2) product from 1290 to 12900 mm2.

The BTC method successfully accounted for a considerable quantity of full-scale experimental results on
lower strength pipes (X70 and lower) and the Battelle simplified equation have been adopted in several
national standards for pipeline systems. However, as the strength of pipe increases, direct application of
these models no longer correctly predicts the required toughness. In order to extend their applicability,
correction factors have been implemented to align the predicted toughness with results from full-scale
experimental tests on higher strength pipes. One of the highest correction factors proposed for ultra
high strength X100 pipe is 1.4 [4] for the Battelle simplified equation and 1.7 [5] for the BTC method.

The High-strength Line Pipe (HLP) Committee organized by the Iron and Steel Institute of Japan
pointed out that the fracture arrest energy is changed by the length of pipe. And, they have developed [6]
a simulation model for the propagating ductile fracture in pipelines from a series of seven times X70 full-
scale fracture propagation tests (ISIJ-HLP tests). This HLP method is a dynamic variant of the BTC
method and be able to calculate instantaneous crack velocity and propagating distance [7-8].

Analyzed results of the arrest toughness

Arrest toughness for the full-scale fracture propagation tests using X100 or higher strength pipes [4-5, 9-
13] are analyzed by above mentioned three predicting methods. Analyzed results are summarized in Fig.
Pipeline Technology Conference, Ostend, 12-14 October 2009 3

1 together with the ISIJ-HLP tests. The ISIJ-HLP tests on X70 grade 48” diameter 18.3 mm wall
thickness pipes consisted of five tests pressurized with air and two tests with a rich natural gas. Six tests
were pressurized to 11.6 MPa corresponding to 80 % SMYS and one rich natural gas test was pressurized
to 10.45 MPa corresponding to 72 % SMYS.

As shown in Fig. 1(b), except for the test results of DemoPipe Project, the BTC method seemed to
explain successfully the upper limit of propagating points with a correction factor “1.7”. However, this
BTC method underestimates the arrest toughness of both the DemoPipe tests in high degree. In this
figure, no more trend line dividing arresting points from propagating points can be drawn in order to
derive a new correction factor. The HLP method is also in a similar situation. As shown in Fig. 1(c),
except for the test results of DemoPipe Project, the HLP method seemed to divide successfully arresting
points from propagating points with minimum errors. However, this method also can’t explain test
results of the DemoPipe Project. Especially in the case of 2nd DemoPipe test, this method also
underestimates the arrest toughness in high degree.

Concerning the Battelle simplified equation, this method fails to predict correctly even the lower
strength X70 test results. This is, in some sense, a reasonable result because the ISIJ-HLP X70 tests
include two tests pressurized with rich natural gas while the Battelle simplified equation considers the
medium as a single-phase gas. It is interesting that the necessary correction factor for the Battelle
simplified equation is smaller than the necessary correction factor for the BTC method if only the single-
phase gas fracture propagation tests are concerned. However, even when pure methane or lean natural
gas is the conveyed gas, the applicability of the Battelle simplified equation to X100 or higher strength
pipes is questionable because this equation considers the medium as not only a single-phase gas but an
“ideal gas”.

Uncertainties of key items which influence the accuracy of arrest


toughness prediction
Uncertainty of the gas decompression curve

When a pipeline ruptures, gas can escape from the full area of pipe in a process that is essentially
isentropic. The decompression disturbance travels down the pipeline away from the rupture initiation
site, with each pressure level P propagating at a fixed speed w, given by the difference between local
acoustic velocity and corresponding fluid outflow velocity. The curve of P vs. w is called a “gas
decompression curve”. Gas decompression curve can be calculated on the basis of a one-dimensional
isentropic homogeneous equilibrium approach [2, 14].

Example of comparison of predicted vs. measured gas decompressions are shown in Fig. 2. This is the
case of Tokyo Gas fracture propagation test on X80 [15]. As shown in this figure, experimental data
support relatively high accuracy of the prediction of gas decompression curve by above mentioned
approach. Even in the case of higher initial pressures, where the X100 full-scale fracture propagation tests
were carried out, relatively good agreements between predicted and measured gas decompressions are
also reported [12].

In more details, predicted gas decompression curve usually gives slightly non-conservative side estimation
compared to experimental data, and different program code also changes slightly the predicted result.
These differences in gas decompressions change the calculated arrest toughness. This effect is not
negligible but seems to be relatively small as shown in Table 1, where the comparison of calculated BTC
arrest energies using different program code or using extrapolated experimental data is shown.
4 Pipeline Technology Conference, Ostend, 12-14 October 2009

Uncertainty of the crack velocity prediction

The curve expressing variation of crack velocity with pressure is called a “crack velocity curve”. The crack
velocity curve used in the BTC method, for soil backfill, can be stated as following equations.
1/ 6
 flow  Pd 
(1) Vc  0.275     1
R  Pa 

4 t   10 3      R 
(2) Pa     flow  cos 1 exp 
3.33   D  24  Dt / 2   2 
 flow 

(3)  flow   y  68.95

(4) R  C v / Ac
where,
Vc is the crack velocity, in m/sec
σflow is the flow stress of material, in MPa
σy is the yield stress of material, in MPa
R is the material’s resistance to fracture, in Joules/mm2
Pd is the decompressed pressure at crack tip, in MPa
Pa is the arrest pressure, in MPa
t is the wall thickness, in mm
D is the pipe diameter, in mm
E is the elastic modulus of steel, in MPa
Cv is the full-size CVN energy, in joules
Ac is the fracture area of CVN specimen, in mm2

On the other hand, the crack velocity curve used in the HLP method, for soil backfill, can be stated as following
equations.
0.393
 flow  Pd 
(5) Vc  0.670     1
R  Pa 

t   3.81  107  R 
(6) Pa  0.382    flow  cos 1 exp 
D  Dt   flow
2 
 

(7)  flow   y   T  / 2

(8) R  D p est  / Ap

D p est   3.29  t 1.5  C v


0.544
(9)
where

σ'flow is the flow stress of material, in MPa


σT is the tensile stress of material, in MPa
R' is the material’s resistance to fracture, in Joules/mm2
Dp(est) is the estimated value for total energy of pre-cracked DWTT, in Joules
Ap is the fracture area of pre-cracked DWTT specimen, in mm2

An approach based on DWTT energy and recalibration of constant and exponent in the crack velocity equation for
direct application of pre-cracked DWTT energy is one of the features of the HLP method. Equation (9) is the
correlation between CVN energy and pre-cracked DWTT energy, which was established by HLP Committee based
on test results from varied pipes of X60 to X100 in grade and 10 mm to 32 mm in wall thickness [8]. In the BTC
Pipeline Technology Conference, Ostend, 12-14 October 2009 5

method, CVN energy per unit area is used as the material’s resistance. Test results for “conventional design
conditions” supported this approximation of the material’s resistance expressed in term of CVN energy per unit
area. However, it has become clear that this CVN-based model fails to predict correctly the arrest toughness in
applications with higher toughness materials, and it is known that there is an empirically found upper limit [16-18]
of direct applicability of CVN energy calculated from this BTC method.

The advantages of the DWTT-based approach in comparison with the CVN-based approach are the followings; [19]

 Development of a larger plastic zone due to longer ligament of DWTT specimen, and same thickness of
DWTT specimen as that of pipe, allow the fracture appearance to reflect more accurately the actual pipe
behavior.
 Longer ligament of DWTT specimen diminishes the influence of increasing proportion of crack
initiation energy in total energy absorption of tougher materials and this influence could be effectively
removed by modification of DWTT notch.
 Because the thickness of DWTT specimen is the same as that of pipe, this diminishes the difficulties
associated with “nonlinear wall thickness effect on absorbed energy”.

Figure 3 shows comparison of crack velocity curves predicted by Battelle equation (Eq. (1)) and HLP equation (Eq.
(5)). Figure 3(a) is the case of 2nd ECSC test’s initiation pipe and Fig. 3(b) is the case of 2nd DemoPipe test’s
initiation pipe. As shown in this figure, there is a big difference between these two predictions. On the other hand,
experimentally measured maximum crack velocity of 2nd ECSC Test is about 310 m/s and that of 2nd DemoPipe
Test is about 260 m/s. Therefore, it is obvious that the Battelle equation underestimates the crack velocity in case
of these high strength high toughness pipes, while the HLP equation predicts relatively reasonable values.
Comparison of predicted vs. measured maximum crack velocities is shown in Fig. 4. In this context, predicted
maximum crack velocity is the predicted crack velocity at initial pressure. As shown in this figure, in case of the
Battelle equation, the degree of underestimation in crack velocity is considerable and the ratios of the measured
values to the predicted values are quite high for all cases.

Figure 5 shows tendency line in accuracy of the HLP’s crack velocity equation. Y-axis of this figure is the ratio of
measured maximum crack velocity to predicted maximum crack velocity. In case of the HLP equation, accuracy of
crack velocity prediction has dependency to both pipe diameter and wall thickness. These dependencies are well
supported by the additional analyzed data of E.N.I. sponsored CSM X80 full-scale fracture propagation test [20],
where the 56” diameter 26 mm wall thickness pipes were pressurized to 16.1 MPa corresponding to 80 % SMYS.
As shown in Fig. 5(c), when the parameter of (D1/2 × t1/4) is selected, scatter becomes narrower and there is a
strong dependency of the accuracy of crack velocity prediction to this parameter. This tendency line well explains
that the HLP’s crack velocity equation needs no correction factor in the rage where this equation was originally
calibrated. The HLP’s crack velocity equation was originally calibrated based on the full-scale data of 48” diameter
18.3 mm wall thickness pipes. This tendency line supports the idea [21] that the constant in crack velocity equation
may increases as the pipe diameter increases.

Figure 6 shows tendency line in accuracy of the Battelle’s crack velocity equation. In case of the Battelle equation,
accuracy of crack velocity prediction has dependency to both pipe diameter and CVN energy of initiation pipe.
Wide scatter is observed in these dependencies. However, when the parameter of (D1/2 × CVN(initiation)1/4) is
selected, scatter becomes narrower and there is a strong dependency of the accuracy of crack velocity prediction to
this parameter. This implies that the Battelle’s crack velocity equation underestimates the effect of pipe diameter
and overestimates material’s resistance of higher toughness pipes at the same time. As shown in Fig. 6(c), this
tendency line well explains that the Battelle’s crack velocity equation needs no correction factor in the range where
this equation was originally calibrated. The Battelle’s crack velocity equation was originally calibrated based on the
full-scale data of 30” diameter 0.375” wall thickness pipes. The range of CVN energy seems to be mainly between
20 to 50 Joules equivalent for full-size test specimen judging from figures in the original paper [3].

Uncertainty of the arrest pressure prediction

In west side of 1st DemoPipe test and east side of 2nd DemoPipe test, steady state crack propagation was observed.
Therefore, it is expected that the crack velocity curve is located as almost tangent condition with gas decompression
curve at these conditions. However, predicted crack velocity curves of pipes are located apart from gas
decompression curves in both predictions as shown in Fig. 7. It implies that both predictions overestimate the
arrest pressure of these pipes. Especially in the case of 2nd DemoPipe test, these overestimations are in high degree
6 Pipeline Technology Conference, Ostend, 12-14 October 2009

for both predictions. This 2nd DemoPipe test seems to have no specialty in test parameters but have the lowest
ratio of pipe diameter to wall thickness (D/t) among the full-scale fracture propagation tests analyzed.

On the other hand, accuracy of the arrest toughness prediction by HLP method seems to have a dependency to the
D/t ratio as shown in Fig. 8(b). This dependency is not clear in the case of BTC method. As discussed above,
accuracy of gas decompression curve prediction is relatively high and the accuracy of crack velocity prediction is
also relatively high in case of the HLP method. Therefore, there is a high possibility that this dependency in the
accuracy of arrest toughness prediction to the D/t ratio may directly connects to the degree of overestimation of
the arrest pressure. It means that as the D/t ratio of pipe decreases, the arrest pressure may decreases more than the
existing models predict. There seems to be a possibility that accuracy of calculated equivalent crack length in arrest
pressure equation has a dependency to D/t ratio, and further investigation is required in order to clarify this point.

Development of a new predicting model


Sumitomo's crack velocity curve

Based on the investigation on uncertainties of key items which influence the arrest toughness prediction in the
previous section, following new equations for the crack velocity curve, named Sumitomo's crack velocity curve, are
developed.

 flow  Pd 
(10) Vc       1
R   Pa 
t   4.57  10 7  R  
(11) Pa    0.380    flow  cos 1 exp 
D  Dt   flow 
 2

1/ 4
(12)  Dt 
  0.670   
 D0 t 0 
5/2 1 / 2
(13) D t
  0.393      
 D0   t0 
3.42
(14)  3
 t/D 
3.22  0.20   
 t 0 / D0 
(15) D0  1219.2 ( mm)

(16) t0  18.3 (mm)

The parameters of α, β and γ in these equations are characteristically adopted as a function of OD and WT and
showing best fitting curves for the experimental data analyzed. Experimental data indicate followings;

 Experimental data support the approximation of material resistance, in crack velocity curve, expressed in
term of DWTT energy per unit area rather than CVN energy per unit area.
 As the pipe size (D and/or t) increases, maximum crack velocity increases more than the Battelle's or the
HLP's equation predicts. It means that the constant in crack velocity equation may increases as the pipe
size increases.
 As the D/t ratio of pipe decreases, crack velocity curve shifts right-lower side (near the gas decompression
curve) more than the Battelle's or the HLP's equation predicts. It means that the arrest pressure may
decreases and the exponential in crack velocity equation may decreases as the D/t ratio of pipe decreases.

Figure 9 shows comparison of predicted and measured maximum crack velocity in case of Sumitomo's crack
velocity equation. As compare this figure with Fig. 4, accuracy of the Sumitomo's crack velocity equation is
considerably higher than that of the Battelle's or the HLP's equation.
Pipeline Technology Conference, Ostend, 12-14 October 2009 7

Sumitomo model

Analyzed results of arrest toughness in tests by Sumitomo model (HLP method with Sumitomo's crack
velocity curve) are summarized in Fig. 10. As compare this figure with Fig. 1, it is well confirmed that
accuracy of Sumitomo model is improved from that of existing models.

Discussion
Effect of toughness arrangement on crack propagation

Traditionally, increasing toughness arrangement is used in full-scale fracture propagation tests. However, since the
propagating ductile fracture is an interactive phenomenon between gas decompression and crack propagation, the
arrest toughness in test depends on pipe toughness arrangement, in other words the crack propagation history. In
particular, toughness of initiation pipe may control maximum crack velocity and may have a strong effect on total
crack propagation length. Figure 11 shows the effect of increasing toughness arrangement on crack propagation,
which is simulated by Sumitomo model. In Fig. 11(e), four cases of simulated crack propagation for different pipe
toughness arrangement are compared. As shown in this figure, as the gradient of increasing toughness arrangement
becomes steeper, crack may propagate longer and the arrest toughness in test tends to increase. This effect always
exists in this kind of traditional style test. However, in a real pipeline, there is no crack initiation pipe with
toughness far lower from average value. So, when we assess the crack arrestability of a pipeline, we have to consider
the difference between the increasing toughness arrangement in test and the even toughness arrangement in a real
pipeline.

Increasing toughness arrangement used to have an effective aspect that only one test can determine approximately
the required toughness level at interested operating condition of a pipeline. And this testing method seems to have
worked well in case of the tests for conventional operating conditions using conventional grade pipes, where the
required toughness level lies in the middle of available values. However, in case of X100 or higher grade pipes, the
required toughness level may lie near the upper borderline of available values and harmful aspect of the increasing
toughness arrangement comes out. Higher crack velocity produced by unrealistically lower toughness initiation
pipe is harmful to assess the crack arrestability of a pipeline. This effect may mislead into underestimating the crack
arrestability of a real pipeline with even toughness arrangement. Therefore, for the full-scale crack propagation
tests, in order to assess properly the crack arrestability of a pipeline, the even toughness arrangement as same as a
real pipeline instead of the increasing toughness arrangement and the judgment on the basis of total crack
propagation length is recommended.

Intrinsic crack arrestability of X100 pipeline

As shown in Fig. 11, in 1st DemoPipe test, crack propagated through the pipe with highest Charpy energy 355 J.
However, this test result does not mean that X100 pipeline with 355 J Charpy energy cannot arrest a propagating
ductile fracture in a short distance. As shown in Fig. 11(e), Sumitomo model predicts that X100 pipeline with 280 J
Charpy energy has enough intrinsic crack arrestability to arrest a propagating ductile fracture in a short distance so
far as the even toughness arrangement is basically realized in a real pipeline. This is the case of testing pressure
corresponding to 80 % SMYS, and the case of more moderate design factor is shown in Fig. 12. This is the case of
2nd DemoPipe test, and testing pressure corresponds to 75 % SMYS. In this case, Sumitomo model predicts that
X100 pipeline with 240 J Charpy energy has enough intrinsic crack arrestability to arrest a propagating ductile
fracture in a short distance, as shown in Fig. 12(e).

X100 full scale fracture propagation


test with even toughness arrangement
Objectives of the test

Based on past full-scale fracture propagation tests, it is thought that intrinsic arrest of high strength X100
8 Pipeline Technology Conference, Ostend, 12-14 October 2009

pipe is hard to achieve. However, prediction by Sumitomo model indicates that X100 pipeline has a
chance to have enough intrinsic crack arrestability. In order to verify this indication, X100 full-scale
fracture propagation test with even toughness arrangement has been planned and conducted. Objectives
of this test are the followings;

 To prove the crack arrestability against propagating ductile fracture in X100 line pipe.
 To prove the effect of toughness arrangement to arrest length of propagating ductile fracture.
 To confirm the validity of Sumitomo model for predicting propagation and arrest behavior in
high pressure gas transmission pipeline.

Test conditions

Test conditions of the X100 full-scale fracture propagation test are summarized in Table 2. Pipe size of 36 inch OD
× 19.05 mm WT is selected and target testing pressure is 22.0 MPa corresponding to 76.6 % SMYS. Russian gas,
which is classified into lean natural gas, is selected as pressurizing medium. Most remarkable point in this test is to
adopt the even toughness arrangement.

Test line consists of nine pipes for total length of about 90 m. In order to allow the decompression of gas during
the test to occur as in a real pipeline, two reservoirs of 36 inch OD and about 110 m length each are placed at both
ends of the test section, thus giving a total length of about 310 m. A scheme of test line is shown in Fig. 13 [22].
The line position is held by two concrete blocks at both test section ends which allow to reproduce the constraint
of a pipeline of infinite length. Two additional concrete blocks (not reported in Fig. 13) are placed at the end of
each reservoir in order to prevent any lateral and longitudinal movement of the whole line. The test line is entirely
buried under 1.2 m of soil minimum; measurements taken after burying indicates values ranging from 1.2 m to
1.4m.

Properties of X100 pipes for the test

Properties of test pipes used in the full-scale fracture propagation test are shown in Fig. 14. The material strength is
well above the minimum required yield value for the API 5L X100 grade. All the pipes exhibit tensile properties
with a marked uniformity in terms of yield and tensile strength as well as yield to tensile ratio. The average Charpy
absorbed energy value is 283 J and the maximum scatter value is 7 J (2.5%). The average press-notch and pre-
cracked DWTT values are respectively 11002 J and 6956 J and the maximum scatters are 412 J (3.7%) and 417 J
(6.0%). Therefore, it is evident that the set of pipes involved in the test exhibits very similar toughness properties.

Execution of the test

The fracture propagation test was executed on 12th March, 2008 within the full-scale burst test site of Centro
Sviluppo Materiali (CSM) located in Sardinia, Italy. The test line was pressurized with lean gas at the target
pressure (22.07 MPa actual value). Averaged chemical composition of used lean gas is shown in Table 3. Measured
test temperature (averaged gas temperature) was 9.7 °C. The burst was caused using an explosive charge 500 mm
long, located on the upper generatrix of the initiation pipe. The explosion creates a through thickness cut as long
as to guarantee the “break condition” on pipe. Sequential photos taken at the moment of the burst are shown in
Fig. 15. Escaping gas did not catch fire due to the strongly blowing wind.

In this test, fracture was regularly injected and propagated on the upper side at a very high speed along both
longitudinal directions. After leaving the initiation pipe, crack was rapidly slowed down and arrested within the
first pipe on both side. Test section after the burst is shown in Fig. 16. Experimental evidences revealed a very low
presence of separations and that fracture propagated in a fully ductile manner, as shown in Fig. 17.

Measured gas decompression behavior

During the fracture propagation test, pressure transducers has been used to record the pressure trend along the
line. They were installed at 90° with respect to the upper generatrix. Example of measured gas decompression curve
compared with predicted curve is shown in Fig. 18. Predicted gas decompression curve gives slightly non-
conservative side estimation compared to the experimental data. This tendency is as same as past experimental data
Pipeline Technology Conference, Ostend, 12-14 October 2009 9

of ordinary initial pressure. However, as shown in Fig. 18, accuracy of the predicted gas decompression curve is
relatively high even in the case of higher initial pressure of above 20 MPa.

Measured crack propagation and arrest behavior

Timing wires give information on fracture position as a function of time and are used to calculate crack velocity.
Crack velocity measured in this fracture propagation test is depicted in Fig.19. It shows that the fracture was
launched at high speed of about 280 m/s but left the initiation pipe at relatively low speed of about 150 to 180
m/s. Then the fracture entered in the pipes next to the initiation pipe, and continued to decelerate until arrest
occurred after 7.8 m on the east side and 10.1 m on the west side with respect to the initiation point. The test has
been successful in demonstrating the arrest of propagating ductile fracture within a short distance on both side and
proved that arrest length of propagating ductile fracture is influenced by pipe toughness arrangement.

Crack propagation and arrest behavior predicted by Sumitomo model is also shown in Fig.19 to be compared with
measured one. This predicted result coincides very closely with the test result and it has been proved that fracture
length of this test could be accurately predicted by Sumitomo model. And, the test has proved that X100 pipeline
has a chance to have enough intrinsic crack arrestability even at high pressure of over 20 MPa.

Conclusions

Uncertainties of key items which influence the arrest toughness prediction are investigated. Based on the
investigation, new equations for the crack velocity curve (Sumitomo's crack velocity curve) are developed.
And, Sumitomo model (HLP method with Sumitomo's crack velocity curve) is proposed as a new
predicting model of fracture propagation and arrest behavior in high pressure gas transmission pipeline.
It is confirmed that accuracy of Sumitomo model is improved from that of existing models.

Based on prediction by Sumitomo model, X100 full-scale fracture propagation test with even toughness
arrangement has been planned and conducted. The test has been successful in demonstrating the arrest
of propagating ductile fracture within a short distance on both side and proved that arrest length of
propagating ductile fracture is influenced by pipe toughness arrangement. The test has also proved that
X100 pipeline has a chance to have enough intrinsic crack arrestability even at high pressure of over 20
MPa.

Acknowledgement
The full-scale fracture propagation test reported in this paper was accomplished by a cooperative effort of CSM.
Sumitomo gratefully acknowledges all CSM members for executing the test. BP, Exxon Mobil, Eni Div. Gas &
Power and Saipem Energy Service are acknowledged for attending the test. Thanks are also given to Marubeni
Itochu Steel and Sumitomo Corporation for their collaboration.

References
1. S. Okaguchi, M. Hamada, H. Makino, A.Yamamoto, N. Takahashi and I. Takeuchi, 2005, Production and
Development of X100 and X120 Grade Line Pipes, Seminar Forum of X100/X120 Grade High Performance Pipe
Steels, Beijing, China.
2. R. J. Eiber, T. A. Bubenik and W. A. Maxey, 1993, “Final Report on Fracture Control Technology for Natural Gas
Pipelines”, BATTELLE, Columbus.
3. W. A. Maxey, 1974, “Fracture Initiation, Propagation and Arrest”, 5th Symposium on Line Pipe Research, AGA,
Houston, J.
4. G. Demofonti, G. Mannucci, C. M. Spinelli, L. Barsanti and H. G. Hillenbrand, 2000, “Large Diameter X100 Gas
Linepipes: Fracture Propagation Evaluation by Full-Scale Burst Test”, 3rd International Pipeline Technology
Conference, Belgium, Volume I, Pp.509-520.
5. L. Barsanti, G. Mannucci, H .G. Hillenbrand, G. Demofonti and D. Harris, 2002, “Possible Use of New Materials for
10 Pipeline Technology Conference, Ostend, 12-14 October 2009

High Pressure Linepipe Construction: An Opening on X100 Grade Steel”, 4th International Pipeline Conference,
Calgary, Pp.287-298.
6. H. Makino, T. Kubo, T. Shiwaku, S. Endo, T. Inoue, Y. Kawaguchi, Y. Matsumoto and S. Machida, 2001, “Prediction
for Crack Propagation and Arrest of Shear Fracture in Ultra-high Pressure Natural Gas Pipelines”, ISIJ International,
41, p.381.
7. H. Makino, Y. Kawaguchi, Y. Matsumoto, S. Takagi, S. Yoshimura, 2002, “Interactive Simulation of Gas
Decompression and Crack Propagation in Natural Gas Transmission Pipelines”, 4th International Symposium on
Computational Technologies for Fluid/Thermal/ Chemical Systems with Industrial Applications, ASME, Vancouver.
8. H. Makino, T. Inoue, S. Endo, T. Kubo and T. Matsumoto, 2002, “Simulation Method for Crack Propagation and
Arrest of Shear Fracture in Natural Gas Transmission Pipelines”, International Conference on the Application and
Evaluation of High-Grade Linepipes in Hostile Environments, Yokohama.
9. S. D. Papka, J. H. Stevens, M. L. Macia, D. P. Fairchild and C. W. Petersen, 2003, “Full-Size Testing and Analysis of
X120 Linepipe, Symposium on High Performance Materials in Offshore Industry”, ISOPE, Honolulu.
10. S. D. Papka, J. H. Stevens, M. L. Macia, D. P. Fairchild and C. W. Petersen, 2004, “Full-scale burst, crack-arrest testing
vets X120 line pipe”, Oil & Gas Journal, Feb.23.
11. JIP on Fracture Control in High Strength (X100) Steel Pipelines: Data Report for Test2, ADVANTICA
TECHNOLOGY, Report Number R4852, 2002.
12. R. M. Andrews and A. D. Batte, 2003, “Developments in Fracture Control Technology for Gas Pipelines Utilizing
High Strength Steels”, IGU 22nd World Gas Conference, Tokyo.
13. R. M. Andrews, A. D. Batte, B. J. Lowesmith and N. A. Millwood, 2003, “The fracture Arrest Behaviour of 914mm
Diameter X100 Grade Steel Pipelines”, The 14th EPRG/PRCI Joint Technical Meeting on Pipeline Research, Berlin.
14. G. Demofonti, G. Mannucci, M. Di Biagio, H. G. Hillenbrand and D. Harris, 2004, “Fracture propagation resistance
evaluation of X100 TMCP steel pipes for high-pressure gas transportation pipeline using full-scale burst tests”, 4th
Internal Conference on Pipeline Technology, Ostend, Belgium.
15. D. G. Jones and D. W. Gough, 1981, “Rich Gas Decompression Behavior in Pipelines”, AGA-EPRG Linepipe
Research Seminar IV, BRITISH GAS, Duisburg.
16. S. Kawaguchi, N. Hagiwara, T. Masuda, C. Christensen, H. P. Nielsen, P. B. Ludwigsen, T. Inoue, D. L. Rudland and
G. M. Wilkowski, 2004, ”Application of X80 in Japan: fracture control”, 4th Internal Conference on Pipeline
Technology, Ostend, Belgium.
17. R. Eiber, L. Carlson, B. Leis, 2000, “Fracture Control Requirements for Gas Transmission Pipelines”, 3rd
International Pipeline Technology, Volume I, Belgium, pp. 437-453.
18. B. N. Leis, 2000, “Predicting Fracture Arrest Based on a Relationship Between Charpy Vee-Notch Toughness and
Dynamic Crack-Propagation Resistance”, 3rd International Pipeline Technology, Volume I, Belgium, pp. 407-420.
19. C. R. Killmore, F. J. Barbaro, J. G. Williams, A. B. Rothwell, 1997, “Limitations of The Charpy Test for Specifying
Fracture Propagation Resistance”, International Seminar on Fracture Control in Gas Pipelines, WTIA/APIA/CRC-
MWJ, Sydney, 4.
20. A. B. Rothwell, 2000, “Fracture Propagation Control for Gas Pipelines – Past, Present and Future”, 3rd International
Pipeline Technology, Volume I, Belgium, pp.387-405.
21. G. Mannucci, G. Demofonti, L. Barsanti, C. M. Spinelli, 2000, “High Pressure Long Distance Transmission Gas
Lines: Structural Integrity of High Strength Steel Grade Pipeline”, 3rd International Pipeline Technology, Brugge,
Belgium, Volume I, pp.303-314.
22. T. Akiyama, 1998, “Influence of the diameter for running shear fracture propagation of high pressure gas pipeline,
PVP-Vol.371, High Pressure Technology, pp.131-136.
23. A. Meleddu, A. Lucci, M. Di. Biagio, August 2008, "Ductile Fracture Propagation Full Scale Burst Test on Sumitomo
X100, 36in OD, 19.05mm Wall Thickness", CSM Report No. 13639R.
Pipeline Technology Conference, Ostend, 12-14 October 2009 11

500 500
(a) Battelle simplified equation (b) Battelle Two Curve method
CVN energy of pipe (J)

CVN energy of pipe (J)


400 1 : 1.4 400
1 : 1.7
300 1:1 300 1:1

200 200

100 100

0 0
0 100 200 300 400 0 100 200 300 400
Predicted arrest energy (J) Predicted arrest energy (J)
500
(c) HLP method : ISIJ-HLP X70 Tests Arrest
CVN energy of pipe (J)

400
: ISIJ-HLP X70 Tests Propagate
: ECSC Program on X100 Arrest
300 1:1
: ECSC Program on X100 Propagate
: ExxonMobil X120 Test Propagate
200
: Advantica's JIP on X100 Arrest
: Advantica's JIP on X100 Propagate
100
: DemoPipe Project on X100 Arrest
: DemoPipe Project on X100 Propagate
0
0 100 200 300 400
Predicted arrest energy (J)

Figure 1 Actual Charpy energy vs. predicted arrest energy -1

Figure 2 Example of comparison of predicted vs. measured gas decompressions


(Tokyo Gas burst tests on X80, figure from reference 15)
12 Pipeline Technology Conference, Ostend, 12-14 October 2009

Table 1 Comparison of calculated BTC arrest energies using different program code or using extrapolated experimental data

1st 2nd ExxonMobil Advantica Advantica 1st 2nd


ECSC ECSC X120 JIP No.1 JIP No.2 DemoPipe DemoPipe
SUMITOMO (HLP original) 169 J 152 J 168 J 135 J 158 J 153 J 126 J
CSM (GASMISC) 176 J 154 J 148 J 120 J
ExxonMobil 165 J
Advantica (GASDECOM) 128 J 155 J
Advantica (FRACPROP) 120 J 145 J
Advantica (using extrapolated
150 J 165 J
experimental data)

30 30
(a) Predicted crack velocity curves for (b) Predicted crack velocity curves for
the 2nd ECSC Test's initiation pipe the 2nd DemoPipe Test's initiation pipe

20 Battelle (Eq.1)
Pressure (MPa)

Pressure (MPa)
20
Battelle (Eq.1)
HLP (Eq.5)
HLP (Eq.5)
10 10

Experimentally measured Experimentally measured


maximum crack velocity is abt. 310 m/s maximum crack velocity is abt. 260 m/s
0 0
0 100 200 300 400 0 100 200 300 400
Crack velocity (m/sec) Crack velocity (m/sec)

Figure 3 Comparison of crack velocity curves predicted by Battelle equation and HLP equation
maximum crack velocity (m/sec)

maximum crack velocity (m/sec)

500 500
(a) Prediction of the Battelle equation (b) Prediction of the HLP equation
Experimentally measured

Experimentally measured

400 400

300 300

200 200

100 100
1:1 1:1
0 0
0 100 200 300 400 500 0 100 200 300 400 500
Predicted maximum crack velocity (m/sec) Predicted maximum crack velocity (m/sec)

Figure 4 Comparison of predicted vs. measured maximum crack velocities -1


Pipeline Technology Conference, Ostend, 12-14 October 2009 13

crack velocity to the predicted value

crack velocity to the predicted value


2 2
Ratio of the measured maximum

Ratio of the measured maximum


(a) Dependency to the pipe diameter (b) Dependency to the wall thickness

1 1

: CSM X80 Test


0 0
800 1000 1200 1400 10 15 20 25
Pipe diameter: D (mm) Wall thickness, t (mm)
crack velocity to the predicted value

2
Ratio of the measured maximum

(c) Tendency line from the originally


calibrated data range

Data set where the


HLP equation was originally calibrated
0
50 60 70 80 90
D1/2×t1/4

Figure 5 Tendency line in accuracy of HLP’s crack velocity equation


crack velocity to the predicted value

crack velocity to the predicted value

3 3
Ratio of the measured maximum

Ratio of the measured maximum

(a) Dependency to the pipe diameter (b) Dependency to


the initiation pipe's toughness

2 2

1 1
800 1000 1200 1400 0 100 200
Pipe diameter: D (mm) CVN(initiation) (Joules)
crack velocity to the predicted value

3
Ratio of the measured maximum

(c) Tendency line from the originally


calibrated data range
2

Data range where the


Battelle equation was originally calibrated
0
60 80 100 120 140
1/2 1/4
D ×CVN(initiation)

Figure 6 Tendency line in accuracy of Battelle’s crack velocity equation


14 Pipeline Technology Conference, Ostend, 12-14 October 2009

30 30
(a) 1st DemoPipe Test (237J) (b) 2nd DemoPipe Test (223J)
Crack velocity curves for No.8797 pipe Crack velocity curves for No.8839 pipe
Pressure (MPa)

Pressure (MPa)
20 20 Battelle
Battelle HLP
HLP

10 10
Gas decompression curve Gas decompression curve

0 0
0 200 400 600 0 200 400 600
Crack or gas decompression velocity (m/sec) Crack or gas decompression velocity (m/sec)

Figure 7 Predicted crack velocity and gas decompression curves at conditions where steady state crack propagation was observed

4 4
(a) Battelle Two Curve method (b) HLP method
to predicted arrest energy

to predicted arrest energy


Ratio of actual CVN

Ratio of actual CVN


3 3
Maximum propagate line
2 2

1 1

0 0
40 50 60 70 80 40 50 60 70 80
D/t ratio D/t ratio

Figure 8 Tendency line in accuracy of the arrest toughness prediction

500
maximum crack velocity (m/sec)

Sumitomo's crack velocity equation Sumitomo model


Experimentally measured

500 1:1
CVN energy of pipe (J)

400
400
300
300
200
200

100 100
1:1
0 0
0 100 200 300 400 500 0 100 200 300 400 500
Predicted maximum crack velocity (m/sec) Predicted arrest energy (J)

Figure 9 Comparison of predicted vs. measured Figure 10 Actual Charpy energy vs.
maximum crack velocities -2 predicted arrest energy -2
Pipeline Technology Conference, Ostend, 12-14 October 2009 15

st
(a) Test result (1 DemoPipe Test : X100, 36"OD, 16mmWT, 0.80SMYS)

WEST EAST

Propagation distance : over45m over45m

(b) Pipe toughness arrangement (Charpy V(J)) in test


Pipe No. 8808 8795 8797 8786 8781 8783 8780 8799 8776
CVN(J) 291 249 237 215 193 228 223 258 355

(c) Simulation of crack propagation for the test condition


400 400
○ : Experimental data
Crack velocity

Crack velocity
300 300
(m/sec)

(m/sec)
200 200

100 100

0 0
50 40 30 20 10 0 10 20 30 40 50
Propagating distance (m) Propagating distance (m)

(d) Assumed pipe toughness arrangement (Charpy V(J))


Pipe No. 4W 3W 2W 1W INI 1E 2E 3E 4E Gradient
(1) 310 280 250 220 190 220 250 280 310 30J/Pipe
Arrange-

(2) 300 280 260 240 220 240 260 280 300 20J/Pipe
ment

(3) 290 280 270 260 250 260 270 280 290 10J/Pipe
(4) 280 280 280 280 280 280 280 280 280 0J/Pipe

(e) Simulation for the assumed pipe toughness arrangement


400 400
Pipe toughness arrangement

Crack velocity
Crack velocity

300 300
(m/sec)

(m/sec)
200 (1) 200
(1)
100 (2) (2) 100
(3) (4) (4) (3)
0 0
50 40 30 20 10 0 10 20 30 40 50
Propagating distance (m) Propagating distance (m)

Figure 11 Effect of increasing toughness arrangement on crack propagation -1


(simulated by Sumitomo model)
16 Pipeline Technology Conference, Ostend, 12-14 October 2009

nd
(a) Test result (2 DemoPipe Test : X100, 36"OD, 20mmWT, 0.75SMYS)

WEST EAST

Propagation distance : abt.18.5m


Over44m
(Arrest at a crack arrestor)
(b) Pipe toughness arrangement (Charpy V(J)) in test
Pipe No. 8824 8826 8834 8831 8837 8835 8839 8836 8851
CVN(J) 267 240 252 247 211 206 223 249 257

(c) Simulation of crack propagation for the test condition


400 400
○ : Experimental data
Crack velocity

Crack velocity
300 300
(m/sec)

(m/sec)
200 200

100 100

0 0
50 40 30 20 10 0 10 20 30 40 50
Propagating distance (m) Propagating distance (m)

(d) Assumed pipe toughness arrangement (Charpy V(J))


Pipe No. 4W 3W 2W 1W INI 1E 2E 3E 4E Gradient
(1) 270 240 210 180 150 180 210 240 270 30J/Pipe
Arrange-

(2) 260 240 220 200 180 200 220 240 260 20J/Pipe
ment

(3) 250 240 230 220 210 220 230 240 250 10J/Pipe
(4) 240 240 240 240 240 240 240 240 240 0J/Pipe

(e) Simulation for the assumed pipe toughness arrangement


400 400
Pipe toughness arrangement

Crack velocity
Crack velocity

300 300
(m/sec)

(m/sec)
200 200
(1) (1)
100 100
(2) (3) (4) (4) (3) (2)
0 0
50 40 30 20 10 0 10 20 30 40 50
Propagating distance (m) Propagating distance (m)

Figure 12 Effect of increasing toughness arrangement on crack propagation -2


(simulated by Sumitomo model)

Table 2 Conditions for the X100 full-scale fracture propagation test

Remarks
Pipe Grade X100
Nominal Outer Diameter 36 inch
Nominal Wall Thickness 19.05 mm
Test Pressure Target: 22.0 MPa (0.766SMYS) Actual: 22.07 MPa (0.768SMYS)
Pressurizing Gas Russian Gas (99% Methane)
Backfill Depth 1.2 m to 1.4 m Local soil
Test Temperature Ambient (Not Controlled) Measured Temperature: 9.7 degC
Toughness of Pipes 278 J to 290 J Average: 283 J
Pipe Toughness Arrangement Even
Length of Test Line About 90 m
Pipeline Technology Conference, Ostend, 12-14 October 2009 17

WEST EAST

W4 W3 W2 W1 I E1 E2 E3 E4

Reservoir~ 110 m Test Line ~ 90 m Reservoir~ 110 m

Figure 13 Scheme of the test line

(a) Pipe Arrangement


WEST Inside
Marking
Inside
Marking
Inside
Marking
Inside
Marking
Inside
Marking
Inside
Marking
Inside
Marking
Inside
Marking
Inside
Marking
EAST
*Set inside marking of every pipe toward WEST

(b) Pipe data Initiator


Pipe No. 7I0056 7I0065 7I0069 7I00647I0059 7I0061 7I0063 7I0062 7I0072
Length(m) 10.16 10.14 10.16 10.14 8.60 10.15 10.14 10.13 10.11
YS(MPa) 806 823 798 802 823 840 832 780 808 *(round bar)
TS(MPa) 848 863 852 854 849 867 861 842 858 *(round bar)
Y/T ratio 0.950 0.954 0.937 0.939 0.969 0.969 0.966 0.926 0.942 *(round bar)
maxEL(%) 21.4 21.8 21.5 18.5 20.3 17.1 19.3 20.0 19.5 *(round bar)
CharpyV(J) 283 282 281 279 278 283 285 287 290 *(+10℃)
press-notchDWTT(J) 11208 10675 11028 10735 11062 11097 10856 10941 11414 *(+10℃)
pre-crackDWTT(J) 7182 6539 6676 7373 6967 6967 6984 7121 6796 *(+10℃)

Figure 14 Properties of test pipes for the X100 full-scale fracture propagation test

Table 3 Average gas chemical composition (mole %)

CH4 C2H6 C3H8 iC4H10 nC4H10 iC5H12 nC5H12 neoC5H12 C6H14 (+ ) N2 CO2 He
99.368 0.016 0.006 <0.001 <0.001 <0.002 <0.001 <0.001 <0.047 0.495 0.064 0.004

Figure 15 Sequential photographs just after the burst


18 Pipeline Technology Conference, Ostend, 12-14 October 2009

(b) Arrest on west side (on pipe W1)

(a) View from west to east side (c) Arrest on east side (on pipe E1)

Figure 16 Test section after the burst

Figure 17 Fracture appearance

(a) WEST Side (b) EAST Side


(No.3 Pressure Transducer) (No.13 Pressure Transducer)
20 20
Pressure (MPa)

Pressure (MPa)

10 10
: predicted curve : predicted curve
: measured value : measured value
0 0
0 100 200 300 400 500 600 0 100 200 300 400 500 600
Gas decompression velocity (m/sec) Gas decompression velocity (m/sec)

Figure 18 Comparison of predicted vs. measured gas decompression curve


Pipeline Technology Conference, Ostend, 12-14 October 2009 19

(a) Test result


(SMI-No.1-X100 Test, 22.07MPa(0.768SMYS), 9.7degC, Lean natural gas)
WEST EAST

Propagation distance : 10.1m 7.8m

(b) Pipe data (X100, 36"OD, 19.05mmWT)


Pipe No. 7I0056 7I0065 7I0069 7I0064 7I0059 7I0061 7I0063 7I0062 7I0072
Length(m) 10.16 10.14 10.16 10.14 8.60 10.15 10.14 10.13 10.11
YS(MPa) 806 823 798 802 823 840 832 780 808
TS(MPa) 848 863 852 854 849 867 861 842 858
CharpyV(J) 283 282 281 279 278 283 285 287 290

(c) Simulated vs. measured crack propagation behavior


400 400
WEST Side EAST Side
Crack velocity

Crack velocity
300 Simulated crack length : 8.8m Simulated crack length : 7.6m 300
(m/sec)

(m/sec)
Crack length in test : 10.1m Crack length in test : 7.8m
200 200

100 solid line : simulated result 100


circle : experimental data
0 0
50 40 30 20 10 0 10 20 30 40 50
Propagating distance (m) Propagating distance (m)

Figure 19 Predicted vs. measured crack propagation behavior

You might also like