You are on page 1of 11

Pipeline Technology Conference, Ostend, 12-14 October 2009

Paper no: Ostend2009-009

Coupled numerical model of


gas-decompression and unstable
ductile crack propagation in
high-pressure gas pipelines
by K Misawa, Y Imai, and S Aihara

The University of Tokyo, Graduate School of Engineering, Bunkyo, Tokyo, Japan

W E DEVELOPED A new simulation model which simultaneously solves both crack propagation and
gas decompression behaviors of unstable ductile crack propagation in high-pressure gas pipelines.
The model also takes account of the backfill effect. Calculation results indicate that interaction between
crack propagation and gas decompression is significant. Influences of gas properties, backfill depth, yield
strength and diameter to thickness ratio on required toughness for crack arrest are estimated by the
present model and their results are discussed.

N ATURAL GAS HAS been becoming more important as world’s primary energy source in the 21st century [1].
High-pressure natural gas pipelines will increase worldwide. Moreover, new types of gas pipelines, i.e.
hydrogen gas and CO2 pipelines, will be needed for realizing sustainable energy society [2],[3].

Prevention of unstable ductile fracture in high-pressure gas pipelines is one of the most important issues in
pipeline integrity. More reliable and more accurate methodology for evaluation and prevention of this type of
fracture are necessary, especially for the new types of pipelines.

So called Battelle Two Curve approach, referred to as BTC hereafter, has been widely used for evaluating unstable
ductile fracture for natural gas pipelines [4]. The BTC predicts stability / instability of ductile crack propagation by
comparing crack resistance curve, so-called J-curve, and gas decompression curve. If the two curves intersect with
each other, a crack is predicted to propagate indefinitely under balanced crack velocity and decompression velocity.
The BTC is simple and widely used for determining required toughness of pipeline materials. However, this
approach cannot predict a process of transient crack propagation and arrest behaviors. Moreover, insufficient
accuracy has been pointed out for predicting the fracture behaviors of high-pressure pipelines with high-strength
steel pipes.

The Japanese HLP approach [5] can simulate transient behaviors of crack propagation and arrest of pipelines. The
model is based on the BTC but estimates crack velocity change from difference between the present crack velocity
and gas decompression velocity of the pressure at crack-tip. By recent modifications, the HLP approach predicts
crack stability / instability for wide ranges of pipe size and strength [6].

Both BTC and HLP approaches are fundamentally based on the two curves, crack resistance curve and gas
decompression curve, and do not take account of the interaction between crack propagation and gas
decompression behaviors. However in the actual pipe fractures, these two factors affect each other: a different
approach should be necessary to account for the interaction. This may be more important especially for new types
of gas pipeline like hydrogen and CO2 pipelines because the above approaches are heavily dependent on the
existing experimental data of full-scale burst tests of natural gas pipelines.

So far, full numerical models based on finite-element method have been developed [7],[8]. Most of these models

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

calculate pipe deformation and fracture behaviors under a given gas pressure distribution from analytic formulas.
There are few which simulate crack propagation and arrest considering the interaction between crack propagation
and gas decompression behaviors [9].

As will be shown in the present paper, interaction between crack propagation and gas decompression cannot be
ignored especially at crack decelerating period before crack arrest. Therefore, incorporation of the interaction
should be necessary for better accuracy of the prediction.

The authors have developed a new model which takes account of the fracture and gas interaction. Model
formulation and some example calculations are presented in the present paper and their results are discussed.

Modelling gas decompression and unstable crack propagation


Gas decompression model

The simplest form of equation for gas decompression during fracture of pipe is expressed as [4],

2
  1  x   1
p  p0 1    1 (1)
   1  a0 t  

where, p is pressure at x and t, p0 is initial pressure,  is specific heat ratio and a0 is sound velocity at initial
pressure. Gas is assumed as perfect gas. It is also assumed that the pipe end, x=0, is full bore open at time t=0 and
the gas flows one-dimensionally in an isentropic manner, i.e. no heat exchange or no friction between the pipe
wall.

In a pipe with a propagating crack, however, the gas escapes from an opened crack, Fig.1. By considering this gas
escape from the crack, mass conservation law is expressed as,

 
(  A)  (  u A)  m  0 (2)
t x

where,  is gas density, A is sectional area of the pipe, u is gas velocity and m
 is mass of escaping gas per unit pipe
length and time. If the gas is assumed to escape from the opened crack under choking condition, i.e. the gas
escaping with sound velocity, then, m  can be expressed as,

 1
 ( x, t ) p  2   1
m     (3)
RT   1

where,  (x,t) is crack opening displacement, R is gas constant and T is temperature.

The equation of motion for one-dimensional flow is expressed as,

u u 1  p
u  0 (4)
t  x  t

Equation of state for perfect gas is expressed as,

 1
T  p 
  (5)
T0  p0 
Pipeline Technology Conference, Ostend, 12-14 October 2009 3

where, T0 is initial temperature of gas.


Gas flow is assumed isentropic and then,

1
  p 
  (6)
 0  p0 

where,  is initial gas density.

Equations (2), (4), (5) and (6), together with Eq.(3) can be solved numerically if crack opening displacement, , is
given as a function of x and t. In the present model,  is given from the pipe deformation and fracture model,
which is explained below. It is also assumed that the crack is initiated at x=0 and t=0, and the gas begin to escape
from the opened crack. At the same time, the gas is assumed to escape at x=0 in the axial direction because the
crack can be assumed to fully open after short period from crack initiation.

Pipe deformation and fracture model

The crack propagation model is based on the model by Freund et al [10]. For a short time increment, crack velocity
can be regarded as constant and pipe deformation and fracture process are treated as steady state. We assume that
the pipe maintains its cross-section as circular and its radius changes from initial radius, a, to a .  , the radius
change, is a function of , axial coordinate with origin at crack-tip.

The deformation field is expressed as,

 2  2
u  a ( )   cos    a ( ) (7)
 2  8

v  (sin    ) ( ) (8)

w  (1  cos  ) ( ) (9)

where, u, v and w are displacement of the pipe wall in the axial, circumferential and radial direction, see Fig.2.
From the above assumption, it is understood that principal deformations are circumferential and axial bending.

Following Freund et al. [10], plastic zone ahead of the crack-tip is treated as cohesive zone with its length . Within
this zone, cohesive stress, ch is applied along the crack face. Value of ch is assumed equal to yield stress of the
material considering its increase by high strain-rate.

The principle of virtual work gives us a governing equation of the pipe deformation,

 2 2  1 2
a 2    N (  )  M (  )  a p ( )   p h V 2 r  ( )   ch  0 (10)
 8   a 3

where, p is pressure as a function of , p is density of pipe material, h is pipe wall thickness, V is crack velocity and
ch is non-zero only in the cohesive zone. N is axial bending stress as a function of axial strain, . M is
circumferential bending moment as a function of circumferential bending strain, .  depends on '' and , but
assumed to depend on '' only for simplicity.  is equal to   / r 2 . From these assumptions, Eq.(10) is
expressed as a fourth order ordinary differential equation of . Note that  ( x, t )  2  in Eq.(3). Functions N
and M are determined by assuming a power-law hardening of the pipe material.

To solve Eq.(10), boundary conditions at the tip of the cohesive zone should be given:  ( )   ( )  0 . In
4 Pipeline Technology Conference, Ostend, 12-14 October 2009

addition, by assuming that  ( ) is constant within the cohesive zone,  ( )   t /   2 , where  t is crack-
tip opening displacement,  t  K R / E  Y ,   ( K R /  Y ) /  , where  Y is yield stress of the pipe material,
2 2

and  ( )  0 .

For determining crack velocity and crack opening profile under given pressure distribution, we have to introduce
energy balance. Crack driving force in dynamic fracture is expressed as,

K R (V ) 2 1  dW dU dEk 
G      (11)
E h  da da da 

where, K R (V ) is crack propagation resistance as a function of crack velocity, W is work done by internal pressure,
U is strain energy of the pipe and Ek is kinetic energy of the pipe. It should be noted that Ek should include kinetic
energy of backfill soil in case of a buried pipe.

In addition, crack propagation resistance should be assumed. From a measurement of plastic strain distributions
near the propagating crack of full-scale burst tests [11],[12], dependency of crack resistance on crack velocity is
assumed as,

 V  
m

K R (V )  K R 0 1  c1    (12)
  V0  

where, KR0 is crack resistance at V=0, c1 and m are experimentally determined constants and V0 is reference crack
velocity.

Iterative procedure is performed to obtain crack velocity and crack opening profile. Under assumed value of crack
velocity, Eq.(10), with boundary conditions determined from Eq.(12), is solved. Using the crack velocity and
obtained crack opening profile, Eq.(11) is evaluated. This procedure is repeated until convergence is reached.

Backfill Model

Backfill has a pronounced effect on crack propagation through added mass effect and possibly through flow
resistance of soil. In the BTC and HLP approach, backfill effect is incorporated by changing a factor according to
the backfill and non-backfill condition in the crack resistance curve [4][5]. Inoue modeled the influence of backfill
based on a dimensional analysis [13]. In the present model, backfill effect is modeled through the added mass
 , then the soil
effect. Soil is assumed incompressible. If the pipe wall moves in the radial direction with velocity w
at distance r from pipe center moves by ( a / r ) w  in the radial direction. Then kinetic energy of the soil in a
sector d may be expressed as,

2
a  hb 1 a 
Ek ( soil )    soil (r d )  w  dr (13)
a 2 r 

,
where, hb is backfill depth. By equating the above value with equivalent mass which moves at constant velocity w
(1 / 2)  soil (a d ) heq w 2 , we obtain,

 h 
heq  a log1  b  . (14)
 a

heq may be called “equivalent backfill depth”. For incorporating the backfill effect,  p h in Eq.(10) should be
replaced by (  p h   soil heq ) .
Pipeline Technology Conference, Ostend, 12-14 October 2009 5

Coupled Model

If pressure distribution is given, crack deformation and crack velocity can be solved with the Pipe Deformation and
Fracture Model, 2.3. On the other hand, with the Gas Decompression Model, pressure distribution at the next
time step can be obtained, 2.1. Thus, solving the both models by turns, we can solve the behaviors of gas
decompression and dynamic crack propagation, coupling the two models.

A crack can be arrested possibly by two ways: arrest by insufficient crack driving force and by crack deviation from
axial to circumferential direction leading to ring-off. In the reported full-scale burst tests, cracks were arrested by
crack deviation more often than by axial crack arrest [6],[14],[15],[16]. Therefore, condition for crack deviation
rather than axial crack arrest should be sought. To date, there have been no exact models to predict crack deviation
in pipes. However, stability of crack propagation direction in plane problem was discussed in [17]. It was predicted
that crack propagation direction becomes unstable and the crack tends to deviate from the initial crack direction if
the applied stress perpendicular to the crack is smaller than that parallel to the crack. This condition might be
applied to a crack in a pipe. Namely, it might be predicted the crack deviates if hoop stress becomes less than axial
stress at crack-tip section. Initial value of this ratio is equal to 2 if pipe ends are closed. If we assume that
decompression wave does not reach the pipe ends and axial stress does not change until crack arrest, the ratio of
hoop stress to axial stress might be equated to twice the ratio of pressure at crack-tip to initial pressure. This means
that the crack propagation direction is predicted to become unstable if the pressure at crack-tip becomes less than
half the initial pressure. Limited full-scale burst test data [6], [16] indicate that cracks deviated when the ratio of
pressure at crack-tip to initial pressure were around 0.4 to 0.45. Then, the authors set a criterion for crack
deviation leading to crack arrest by ring-off as the ratio of crack-tip pressure to initial pressure less than a constant
value, e.g. 0.4.

Results of numerical simulation


Figure 5 shows a comparison of gas decompression behavior between full-scale air burst test [14] and calculation by
the present model [18]. Measured crack velocity data were used in the calculation. Also, crack opening profile was
assumed. Close agreement is recognized between the measurement and calculation. Also is shown a calculation by
the analytic formula, Eq.(1). It is clear that the prediction by Eq.(1) is insufficient for gas decompression behaviors
in the pipe with a propagating crack. The present model can predict pressure distribution behind as well as in front
of the crack-tip. It should be noted that Eq.(1) is insufficient for predicting pressure at and in front of the crack-tip
with crack decelerating regime.

Figure 6 shows an influence of gas property on crack propagation behaviors. Comparison was made between pure
methane and hydrogen gas. The crack begins to decelerate just after crack initiation and tends to saturate to a
constant velocity. The saturated crack velocity is lower in hydrogen gas than in methane gas. This is primarily due
to earlier gas decompression in hydrogen gas due to its higher sound velocity. This result strongly suggests that
unstable ductile crack propagation is more easily arrested in hydrogen pipelines than in methane gas pipelines.
This behavior agrees with the experiments conducted by part of the authors [11].

Figure 7 shows an influence of backfill on crack propagation behaviors. Initial pressure is 12MPa. While the value
of K R 0 is assumed constant, 500MPa m , for 10m from the pipe center, it is continuously increased with
gradient, 10MPa m / 1m . After reaching maximum value, crack velocity decreases with increasing crack length.
This is because energy is balanced at lower crack velocity with higher crack resistance. Influence of backfill is
significant: crack velocity is decreased with deeper backfill. However, influence of backfill depth tends to saturate.

Figure 8 show a change of pressure at crack-tip for the same conditions as Fig.7. It is noted that the crack is forced
to propagate at constant velocity, 200m/s, for initial 5m and then crack velocity is calculated by Eq.(11). After 5m
crack propagation, pressure decreased to 60% of the initial pressure and continues to decrease. Pressure decrease is
more pronounced with deeper backfill. Again, the influence of backfill depth tends to saturate. From Fig.8, we can
determine K R 0 value at crack arrest based on the crack deviation criterion explained in 2.4. This value is regarded
as required toughness for crack arrest. Result is shown in Fig.9. In the figure, specified minimum yield strength,
SMYS, is also varied, while design factor, initial hoop stress by SMYS, is kept constant at 0.71. It is predicted that a
6 Pipeline Technology Conference, Ostend, 12-14 October 2009

crack can be arrested with about 60% toughness for 1m deep backfill as compared with no-backfill condition.
Influence of SMYS is also significant: required toughness increases with increasing SMYS.

Figure 10 shows influence of pipe diameter to thickness ratio on required toughness based on the same
assumptions as in Figs.7 through 9. Again, the influence of backfill is significant. It should be noted that initial
pressure is inversely proportional to diameter to thickness ratio. Influence of diameter to thickness ratio has no
significant influence on required toughness for crack arrest.

Because above results are only trial and depend on many factors including crack resistance dependence on crack
velocity, Eq.(12). Additional calculations are necessary for more detailed discussions.

Conclusions
The authors developed a new model to simulate dynamic crack propagation and gas decompression behaviors in
high-pressure gas pipelines, coupling the two phenomena and some results of trial calculations are shown.
Obtained results are as follows:

 Interaction between crack propagation and gas decompression is significant especially at crack decelerating
stage. By taking account of the gas escape from an opened crack, pressure distribution ahead and behind the
crack-tip can be accurately calculated.
 By coupling pipe deformation and fracture model with gas decompression model, together with crack
resistance curve as a function of crack propagation velocity, history of crack propagation can be simulated. In
the model, influence of backfill is incorporated by considering kinetic energy of soil.
 Influence of backfill, yield strength, diameter to thickness ratio on crack resistance toughness for crack arrest is
estimated. A criterion for crack deviation leading to crack arrest by ring-off is assumed based on the ratio of
hoop stress to axial stress at crack-tip. It has been shown that backfill effect is significant but it tends to
saturate with backfill depth.

Acknowledgement
Part of the present study was conducted by “Research for establishing reliability analysis methodology of hydrogen
gas transmission pipelines”, 2008-2009fy, funded by NEDO, Japan. The authors express their sincere
acknowledgment to NEDO.

References
1. Key world energy statistics, Int. Energy Agency, 2008.
2. W.C.Leighty, M.Hirata, K.O’Hashi, H.Asahi, J.Benoit, G.Keith, “Large renewables – hydrogen energy
systems: gathering and transmission piplines for windpower and other diffuse, dispersed sources”, for the
presentation at World Gas Conference 2003, Tokyo, 1-5 June 2003., Int. Gas Union.
3. A.Cosham, R.J.Eiber, “Fracture control in carbon dioxide pipelines – the effect of impurities”, Proc. Of
IPC2008, 7th Int. Pipeline Conf., IPC2008-64346, Sept 29-Oct 3, 2008, Calgary Alberta, Canada, ASME.
4. R.J.Eiber, T.A.Bubenik, W.A.Maxey, “Fracture control technology for natural gas pipelines”, PR-3-9113,
Pipeline Research Council International Catalog No. L51691.
5. T.Inoue, H.Makino, S.Endo, T.Kubo, T.Matsumoto, “Simulation method for shear fracture propagation
in natural gas transmission pipelines”, Proc. 13th Int. Offshore and Polar Eng. Conf., Honolulu, Hawaii,
USA, May 25030, 2003, 121-128.
6. H.Makino, I.Takeuchi, R.Higuchi, “Fracture propagation and arrest in high-pressure gas transmission
pipeline by ultra high strength lilepipes”, Proc. Of IPC2008, 7th Int. Pipeline Conf., Sept 29-Oct 3, 2008,
Calgary Alberta, Canada, IPC2008-64078, ASME.
7. P.E.O’Donoghue, M.F.Kanninen, C.P.Leung, G.Demofonti, S.Venzi, “The development and validation of
a dynamic fracture propagation model for gas transmission pipelines”, Int. J. Pres. Ves. & Piping, vol.70
(1997), 11-25.
8. X.C.You, Z.Zhuang, C.Y.Huo, C.J.Zhuang, Y.R.FEng, “Crack arrest in rupturing steel gas pipelines”, Int.
Pipeline Technology Conference, Ostend, 12-14 October 2009 7

J. Fracture, vol. 123 (2003), 1-14.


9. G.Mannucci, M.Di Biagio, G.Demofonti, A.Fonzo, P.Salvini, A.Edwards, “Crack arrestor design by finite
element analysis for an X100 gas transportation pipeline”, Pipeline Technology Conf. 2004, 4th Int.
Conf., 9-13 May 2004, Ostend, Belgium, Scientific Survey Ltd. UK.
10. L.B.Freund, D.M.Parks, J.R.Rice, “Running ductile fracture in a pressurized line pipe”, ASTM STP 590,
1976, 243-262.
11. S.Aihara, E.Ostby, H.I.Lange, K.Misawa, Y.Imai, C.Thaulow, “Burst tests for high-pressure hydrogen gas
line pipes”, Proc. Of IPC2008, 7th Int. Pipeline Conf., Sept 29-Oct 3, 2008, Calgary Alberta, Canada,
IPC2008-64166, ASME.
12. Y.Imai, “Measuring method of fast ductile fracture resistance of steel for high-pressure gas pipeline”,
Mater’s thesis, The University of Tokyo, May, 2008.
13. T.Inoue, “Crack propagation and arrest of structural steels and pipelines”, Doctoral thesis 2008:105,
Norwegian University of Science and Technology, May, 2008.
14. E.Sugie, H.Kaji, T.Taira, M.Ohashi, Y.Sugimoto, “Shear fracture arrestability of controlled rolled steel
X70 line pipe by full-scale burst test”, Transactions of the ASME, J. Energy Resources Technology, March
1984, Vol.106, 55-62.
15. R.M.Andrews, N.A.Milwood, A.D.Batte, B.J.Lowesmith, “The fracture arrest behaviour of 914mm
diameter X100 grade steel pipelines”, Proc. IPC 2004, Int. Pipeline Conf., October 4-8, 2004, Calgary,
Alberta, Canada, IPC04-0595, ASME.
16. S.Kawaguchi, K.Murai, Y.Hashimoto, N.Hagiwara, H.Yajima, M.Toyoda, “Full-scale burst tests of ultra-
high pressured rich-gas pipelines under buried and unburied conditions”, Proc. Of IPC2008, 7th Int.
Pipeline Conf., Sept 29-Oct 3, 2008, Calgary Alberta, Canada, IPC2008-64434, ASME.
17. B.Cotterell, J.R.Rice, “Slightly curved or kinked cracks”, Int. J. Fracture, Vol.16, No.2, April 1980, 155-
169.
18. K.Misawa, Y.Imai, S.Aihara, “Crack propagation behavior analyses of simulation model for ductile crack
propagation in high-pressure gas pipelines coupling gas-decompression and dynamic fracture”, Proc. 157th
meeting of the Iron and Steel Institute of Japan, CAMP-ISIJ, vol. 22 (2009)-402.
8 Pipeline Technology Conference, Ostend, 12-14 October 2009

m m dx

 ( x, t )

p  x, t 
x
0 L
(a) gas escaping from fully and partly opened crack (b) gas escaping from partly opened crack

Fig.1 Model of gas escape from a propagating crack.

w
v

a+

a


Fig.2 Assumption of pipe deformation



0 

cohesive zone

crack-tip
Fig.3 Function representing pipe radius change

w

Fig.4 Assumption of soil motion associated with pipe wall movement


Pipeline Technology Conference, Ostend, 12-14 October 2009 9

Fig.5 Comparison of gas decompression behavior between experiment and calculation.

Fig.6 Influence of gas property on crack propagation behaviors.

Fig.7 Influence of backfill on crack propagation velocity.


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

Fig.8 Influence of backfill on pressure at crack-tip.

Fig.9 Required toughness values for crack arrest in toughness gradient test.
Pipeline Technology Conference, Ostend, 12-14 October 2009 11

Fig.10 Influence of pipe diameter to thickness ratio on required toughness values.

You might also like