Академический Документы
Профессиональный Документы
Культура Документы
Proceedings
of OMAE02
Proceedings of OMAE 02
Conference
on Offshore
Mechanics
21 International
21st International
Conference
on Offshore Mechanics
and Arctic
Engineering
and
Artic
Engineering
June 2328, 2002, Oslo, Norway
June 23-28, 2002,Oslo, Norway
st
OMAE2002-28093
PROBABILISTIC INSPECTION OPTIMIZATION OF FREE-SPAN SURVEYS
FOR SUBSEA GAS PIPELINES
Jens P. Tronskar
Det Norske Veritas Pte Ltd
Singapore
Gudfinnur Sigurdsson
Olav Fyrileiv
Olav Forli
Det Norske Veritas AS
HVIK, Norway
Joseph H. Kiefer
Conoco Inc.
Houston Texas, USA
Colin Lewis
Gulf Resources
Indonesia
ABSTRACT
Probabilistic methods have been used to develop the basis for free-span inspection of a gas pipeline system in the South China Sea.
The objective of the probabilistic analysis was to study the probability of fatigue failure associated with postulated planar flaws in the
HAZ of repair welds performed on some of the girth welds. The impact of flaws on the fatigue life under different free-span
conditions were studied.
Conventional free-span analysis involves computation of allowable free-span lengths based on onset of in-line vibrations and does not
normally consider fatigue crack growth. To consider the effect of the weld flaws on the failure probability a combined probabilistic
fatigue and fracture model is required. For the particular pipelines analysed automatic ultrasonic testing (AUT) was used replacing the
conventional radiography of the girth welds. Conservatism in the free-span assessment can then be significantly reduced by taking into
account detailed flaw sizing information from the AUT. The inspection records provide distribution of flaw height, length and position.
Combined with information on current distribution, material strength and fracture toughness distribution, a detailed probabilistic
fatigue crack growth and unstable fracture assessment can be conducted as per the Det Norske Veritas (DNV) 2000 Rules for
Submarine Pipeline Systems [1] using the response models of the DNV Guideline 14 for free-span analyses [2]. The objective of this
analysis is to estimate the critical free-span lengths and the time for fatigue cracks to penetrate the pipe wall.
KEYWORDS Subsea pipelines, Free spans, Girth welds, Probabilistic analyses, Fatigue crack growth, Unstable fracture
Conventional free-span analysis is based on allowable freespan lengths calculated conservatively from a criterion
involving onset of in-line vibrations and does not normally
consider fatigue crack growth. To consider the effect of the
weld flaws on the failure probability a combined probabilistic
fatigue crack growth and fracture model is required.
INTRODUCTION
Reliability methods have been found efficient for optimisation
of in-service inspection of pipelines in the North Sea and also
recently for optimisation of free-span inspection of gas
pipelines in South East Asia. The pre-requisites to apply the
methods are that statistical distributions pertaining to flaw size,
material strength and fracture toughness as well as current and
wave distribution are available.
1
1
Copyright
2002
2002 by
by ASME
ASME
Copyright
Proceedings of OMAE02
21st International Conference on Offshore Mechanics and Arctic Engineering
June 2328, 2002, Oslo, Norway
OMAE2002 PAPER NO. 28093
The present paper is divided into five main sections. The first
section explains the probabilistic fatigue and fracture models.
The second section details the wave and current induced load
and response analysis methods and presents the results in
terms of calculated axial stress ranges for the case study
analyses of 22" and 14" OD pipelines. The third section
presents the materials and flaw size data for the case study
analyses. The fourth and fifth sections present the results of the
probabilistic and the overall conclusions drawn based on the
experience from the case study.
Proceedings of OMAE 02
21st International Conference on Offshore Mechanics and Arctic Engineering
June 2328, 2002, Oslo, Norway
OMAE2002 PAPER NO. 28093
a Y m a m
i =1
0
da
dN
dc
dN
= C A ( K A )m ;
= C C ( K C ) m ;
K A > K th
K C > K th
; c (t ini ) = c
(2)
; a(t ini ) = a 0
[
[
(3)
]
]
Proceedings of OMAE02
21st International Conference on Offshore Mechanics and Arctic Engineering
June 2328, 2002, Oslo, Norway
OMAE2002 PAPER NO. 28093
i.e.:
Lr , a =
ref =
Pb + 3Pm "
+
N
(16)
p "
2
2
2
{Pb + 3Pm "} + 9 Pm (1 " ) + 4( B )
N
p "
N = 3(1 " ) 2 + 4(
)
B
0.5
Where
2a
B
"=
B
(1 + )
c
(17)
(15)
The failure probability, i.e., the probability that the crack size
exceeds a critical crack size within the time period t (or N) is
then
(11)
PF = P( M 0)
K r = (1 014
. Lr 2 ) 0.3 + 0.7 exp( 0.65Lr 6 )}
ref
y
(18)
(14)
t 2002 by ASME
4
Proceedings of OMAE 02
21st International Conference on Offshore Mechanics and Arctic Engineering
June 2328, 2002, Oslo, Norway
OMAE2002 PAPER NO. 28093
Kr ,a
I
K
= I + =
+
mat
K mat
(19)
where :
KI
G1 (t )= K r (t ) K ( K I , K mat , )
K (t )
= K r (t ) I
(t )
K mat
(20)
VR =
The failure event for the limit state defined by equation (20) is
a first passage failure probability of the stochastic process
down through zero. The annual extreme load effect is assumed
identically distributed and taken as independent between the
years. The approximate combined annual fatigue crack growth
and fracture probability at time T-1 to T can thus be calculated
as:
Pf = Pf {G 1 (T ) 0 (G2 (T ) 0 G 2 (T 1) > 0)
Uc + U w
f0 D
(22)
Uw
fw D
(23)
(21)
=
Uc
Uc + U w
(24)
Proceedings of OMAE02
21st International Conference on Offshore Mechanics and Arctic Engineering
June 2328, 2002, Oslo, Norway
OMAE2002 PAPER NO. 28093
KS =
4 me T
w D2
(25)
R I ,1 = 0
R I , 2 = 0
0.18
3.2
In-line response
(26)
S in = 2 A in (A Y / D) ,in S
Response models
Force models
0.16
0.14
Ks,d=0.00
0.12
Ks,d=0.25
0.10
Ks,d=0.50
0.08
Ks,d=0.75
0.06
Ks,d=1.00
0.04
Ks,d=1.25
0.02
Ks,d=1.50
0.00
0.0
0.5
1.0
1.5
2.0
2.5
3.0
3.5
Reduced Velocity VR,d (=V R/f)
4.0
4.5
5.0
Cross-flow response
Proceedings of OMAE 02
21st International Conference on Offshore Mechanics and Arctic Engineering
June 2328, 2002, Oslo, Norway
OMAE2002 PAPER NO. 28093
(27)
S cr = 2 A cr (A Z / D) R k S
1.5
1.4
Cross-Flow VIVAmplitude (A Z/D)
1.3
=0.8-1.0; all KC
1.2
1.1
1
0.9
=0.0-0.8;
KC>30
=
0.8
0.7
=0.0-0.8;
KC<10
=
0.6
0.5
0.4
0.3
0.2
VRcr,onset
0.1
0
0
5
6
7
8
9
10
Reduced Velocity VR,d (=V R/f)
11
12
13
14
15
16
experimental laboratory test data and a limited number of fullscale test. The curves in Figure 2 embody to a large extent all
available test result.
3.4
1600000
(28)
22.5 m span
1400000
25.0 m span
27.5 m span
1200000
30.0 m span
no of cycles
1000000
35.0 m span
40.0 m span
800000
600000
400000
200000
0
0
10
20
30
40
50
60
70
80
Proceedings of OMAE02
21st International Conference on Offshore Mechanics and Arctic Engineering
June 2328, 2002, Oslo, Norway
OMAE2002 PAPER NO. 28093
Note however, that this is valid only for the in-line vibrations.
The cross-flow vibrations will, however, become critical for
such span lengths.
4.
4.1
Materials strength
Material
22" OD
YS
STDV
TS
STDV
(MPa)
(MPa)
(MPa)
(MPa)
544
13
643
11
544
647
Cumulative Probability
X70 pipe
14" OD
X65 pipe
Repair weld
497*
28
630*
30
1
0.9
0.8
0.7
0.6
0.5
0.4
0.3
0.2
0.1
0
0.00
Weibull fit
0.10
0.20
0.30
0.40
0.50
0.60
CTOD, (mm)
AWS 8010-G
4.2
Data
2002 by ASME
8
Proceedings of OMAE 02
21st International Conference on Offshore Mechanics and Arctic Engineering
June 2328, 2002, Oslo, Norway
OMAE2002 PAPER NO. 28093
Cumulative probability
Cumulative Probability
0.9
0.8
0.7
0.6
0.5
0.4
ECP (AUT)
Lognorm (AUT)
ECP (MET)
Lognorm (MET)
0.3
0.2
0.1
0
0
0.5
1.5
2.5
3.5
0.9
0.8
0.7
0.6
0.5
0.4
ECP
F(a/c) Weibull
0.3
0.2
0.1
0
0
Metallography
46
26
2.18
0.96
0.23
0.04
0.97
0.50
0.50
0.28
1.
2.
3.
4.
5.
Cumulative probability
ECP
F(2c) Lognorm
1.5
6.
1
0.9
0.8
0.7
0.6
0.5
0.4
0.3
0.2
0.1
0
Aspect ratio,
a/c
1
0.5
ht 2002 by ASME
9
Proceedings of OMAE02
21st International Conference on Offshore Mechanics and Arctic Engineering
June 2328, 2002, Oslo, Norway
OMAE2002 PAPER NO. 28093
if the span is present for more than 3 years. This implies that
to assure that the annual fatigue failure probability of the
pipeline does not exceed 1.0E-4, inspections of the pipeline
should be conducted at intervals not exceeding 3 years.
1.0E-03
1.0E-04
1.0E-05
1.0E-06
5.
1.0E-02
1.0E-03
1.0E-04
1.0E-05
1.0E-06
0
Time (years)
ght
10 2002 by ASME
10
Proceedings of OMAE 02
21st International Conference on Offshore Mechanics and Arctic Engineering
June 2328, 2002, Oslo, Norway
OMAE2002 PAPER NO. 28093
6.
CONCLUSIONS
ACKNOWLEDGEMENT
7.
AUT girth weld inspection can provide a basis for fitnessfor-service assessment in general and specifically for more
detailed free-span analysis.
8.
REFERENCES
[1]
[2]
[3]
[4]
[5]
[6]
[7]
[8]
[9]
Copyright
2002 by ASME
11
11
Proceedings of OMAE02
21st International Conference on Offshore Mechanics and Arctic Engineering
June 2328, 2002, Oslo, Norway
OMAE2002 PAPER NO. 28093
Table 3 Parameters and distributions applied for the probabilistic fatigue and unstable fracture analysis.
Parameter
Comments
The initial crack size is assumed to be semi-elliptical. The initial crack depth is modelled as log-normal
distributed, fitted from observed/measured defects, with a mean value, E[], and standard deviation, StD[],
given as:
E [a0 ] = 0.97 mm ;
StD[a0 ] = 0.50 mm ;
The initial crack aspect ratio is modelled as Weibull distributed, fitted from observed/measured defects, with
the following values,
22" OD:
E [a 0 / c0 ]= 0.47 ;
14" OD:
E [a 0 / c0 ]= 0.15 ;
Material parameter
m
Material parameter
C
Stress
intensity
threshold
E[Log10C ]= 12.232;
StD[Log10C ]= 0.1713
The threshold level for the crack growth rates is set equal to.
K th = 63 N/mm 3/2
Geometry function
Y
Youngs
E
Modulus,
The geometry function, Ycalc, is calculated using the empirical formulae for a finite plate subjected to
membrane and bending loading proposed by Newman and Raju (1981). The uncertainty in the geometry
function is included by multiplying Ycalc with variable Ymodel.. Ymodel is modelled as an unbiased Normally
distributed variable with CoV equal to 0.1.
Youngs modulus is modelled as fixed with value 2.07 105 MPa
Copyright
2002 by ASME
12
12
Proceedings of OMAE 02
21st International Conference on Offshore Mechanics and Arctic Engineering
June 2328, 2002, Oslo, Norway
OMAE2002 PAPER NO. 28093
Parameter
Comments
Stress
Concentration
Factor, SCF
The SCF due to eccentricity is modelled as Normal distributed with mean value 1.0 and a CoV equal to 0.1.
CTOD
Fatigue loading
The long-term stress range distribution is calculated from the stress range response from in-line VIV for the
addressed free span. It is conservatively assumed that the obtained maximum stress range distribution for a
span occurs at the location of the weld. It is further assumed that the weld defect considered is located at a 3
or 9 oclock position, thereby being exposed to the maximum occurring stress range distribution.
The occurring stress range distribution is fitted to a Weibull distribution. Dependent on the span length
considered, the following parameters apply,
x
F ( x)) = 1 exp
22" OD:
# cycles/year
40 m
4.9 MPa
0.8
3.6 E5
45 m
11.5 MPa
1.2
2.0 E6
50 m
13.4 MPa
1.5
2.4 E6
# cycles/year
22.5 m
1.15
0.84
0.59E1
25.0 m
2.24
0.74
1.2E4
27.5 m
2.98
0.76
4.5E5
5.68
0.98
2.30E6
Span length
14" OD
Span length
30.0 m
Uncertainty on
Weibull Scale
Parameter
An un-biased uncertainty is introduced on the Weibull scale parameter to incorporate uncertainty related to
the stress modelling. The uncertainty factor on the scale parameter is unbiased with a CoV equal to 0.2. The
uncertainty in the Weibull scale parameter is included to allow for uncertainty in the stress ranges established
from the VIV models.
Copyright
2002 by ASME
13
13
Proceedings of OMAE02
21st International Conference on Offshore Mechanics and Arctic Engineering
June 2328, 2002, Oslo, Norway
OMAE2002 PAPER NO. 28093
Annual dynamic
Peak Stress
The distribution of the annual maximum peak stress due to in-line VIV stress as obtained from the Fat-Free
assessment. The dynamic peak stress distribution is modelled as Gumbel distributed with the following
parameters,
Span length
Static
environmental
stress
Mean Value
CoV
22.5 m
14.4
0.17
25 m
20.8
0.17
27.5 m
30.8
0.17
30.0 m
49.2
0.18
The in-line static environmental stress introduced due to drag forces on the pipeline are modelled as Normal
distributed with the following values;
Span length
Mean Value
CoV
22.5 m
6.6
0.1
25.0 m
8.1
0.1
27.5 m
9.8
0.1
30.0 m
11.7
0.1
Pre-tension stress,
pressure induced
stress and
temperature
induced stress
The axial stress level in the pipeline due to the lay-tension, pressure induced stress from maximum operating
pressure and temperature induced stress is obtained from the Fat-free assessment, and modelled as Normal
distributed with mean value 80.2 MPa and CoV equal to 0.1.
Crack size
In the case of the combined event of fatigue crack growth and final fracture, the crack size is assumed equal
to the size at the end of the year under consideration.
Yield stress
The yield stress, is modelled as Normally distributed with mean value and CoV given by ;
[ ]
E S y = 497 MPa ;
[ ]
CoV S y = 0.04
Uncertainty in FAC
No uncertainty is introduced in the modelling of the failure assessment curve. However, although the
experimental data used to establish the FAC exhibit a wide scatter the general R6 FAC diagram is considered
conservative as it was chosen deliberately by Ainsworth [6] to represent a lower bound for all fracture
assessment curves obtained for many different materials.
Plasticity
correction factor
The plasticity correction factor is modelled according to BSI 7910: 1999 [3], where the residual stresses are
conservatively assumed equal to the yield stress.
yright
2002 by ASME
14
14