Sie sind auf Seite 1von 12

Proceedings of the 2012 9th International Pipeline Conference IPC2012 September 24-28, 2012, Calgary, Alberta, Canada

IPC2012-90480

HOW THE UNCERTAINTIES OF ILI DATA AFFECT PIPELINE CRACK ASSESSMENT


Christoph Jger NDT Systems & Services AG Stutensee, Germany Alfred Barbian NDT Systems & Services AG Stutensee, Germany ABSTRACT Crack inspection of pipelines using liquid-coupled ultrasonic technology has established itself as an accepted ILI (in-line inspection) technology that is routinely applied by many operators. The inspection is based on the pulse-echo technique using 45 shear waves which provide very good detection sensitivity. While length sizing of cracks is reasonably precise, the accuracy of (amplitude-based) depth sizing is limited to some extent. Apart from the crack size, the methods applied for crack assessment have to take into account the geometry of the actual pipe, the relevant material properties and the loading conditions. The work presented is based on the FAD method (Failure Assessment Diagram) being the most common technique for assessment of crack-like defects in pipes. It is shown how the assessment results depend on the quantities determined by ILI (e.g. crack size, wall thickness) and their associated measurement uncertainties. Using a deterministic approach the sensitivity with regard to the different parameters can be compared quantitatively. Probabilistic crack assessment based on the combination of the FAD method with Monte Carlo simulations is well suited for taking into account the statistical distribution of the measuring tolerances. It allows, for example, the calculation of statistical properties such as the probability of failure as will be demonstrated for realistic defect situations. INTRODUCTION Crack-like defects are one of the threats pipeline operators have to consider. Whereas the assessment of metal loss features based on methods like, for example, Herbert Willems NDT Systems & Services AG Stutensee, Germany Neb Uzelac NDT Systems & Services (America) Inc. Houston, Texas, USA

B31G or RSTRENG is routinely applied and rather straightforward, crack assessment is a more complex matter. This is due to the more complicated interaction involving the material properties, the stress state and the crack geometry. The information on the latter is usually obtained from the results of an in-line inspection. Apart from the length and depth data the respective measuring tolerances need to be taken into account for the assessment, i.e. for the calculation of critical pressures, in order to be on the safe side. In [1] it was elaborated how the critical pressure depends on variation of the input parameters and their tolerances assuming a specific crack configuration. The objective of the current paper is to generalize this approach by extending the application range with regard to wall thicknesses and diameters relevant for most pipelines and to discuss the dependency of assessment results on the measuring tolerances of ILI data. The results presented allow for a straightforward estimation of critical pressure for a variety of crack situations in pipelines. Additionally it is demonstrated how the sizing accuracy of the ILI data affects the assessment results in terms of probability of failure (PoF). CRACKS IN PIPELINES Cracks or crack-like features in pipelines can occur in various ways. Depending on their origin cracks may be classified as manufacturing or operationally related anomalies. Examples for manufacturing related features are cold or hot cracks in (D)SAW pipe, hook cracks in ERW pipe or lack-of-fusion in welds. Also sloping or surface-breaking laminations are assessed as crack-like defects [2]. Operationally related cracks are, e.g., fatigue

Copyright 2012 by ASME

cracks (often starting at weld irregularities or in combination with already existing cracks), stress corrosion cracking (SCC) or hydrogen induced cracking (HIC) associated with certain media. Due to the typical stress situation in pipelines with the hoop stress being the largest stress component, cracks are normally oriented in axial direction. However, additional stresses can also lead to a circumferential crack orientation. Whether new cracks will be generated or existing cracks will grow depends on the local stress conditions. Cracks will eventually lead to pipeline failure once a critical crack size is reached. Whether failure will occur by leak or rupture depends on the actual situation, see e.g. [3]. IN-LINE CRACK INSPECTION In order to prevent crack-related pipeline failure in-line inspection (ILI) can be applied to detect any crack-like feature long before it may reach a critical size. Ultrasonic methods using 45 shear waves have proven very successful for the in-line detection of cracks as small as 30 mm (1.2 in.) in length and 1 mm (0.04 in.) in depth [4, 5]. In the last 15 years, ILI tools for crack inspection have become an important part of the pipeline integrity management program for many operators worldwide. Ultrasonic techniques for crack inspection used in the ILI tools are mainly based on the reflection amplitudes obtained from crack-like reflectors. The detection sensitivity is very good as even small reflectors produce high reflection amplitudes. However, the depth sizing capabilities of amplitude based techniques are restricted as apart from defect depth the reflection amplitude depends on many other influences. Even though additional information such as multiple detection and certain signal characteristics may be used to improve depth estimation, the reliability of amplitude based depth sizing remains limited. This limitation is a well-known fact in ultrasonic testing and has led in the past to the development of more sophisticated sizing techniques such as TOFD (Time-of-Flight Diffraction Technique) or SAFT (Synthetic Aperture Focusing Technique) which are using amplitude independent time-of-flight information for depth sizing [6, 7]. However, these are rather demanding analysis techniques that would be difficult to implement in (free swimming) ILI tools for the time being. Typically, crack inspection data are graded using depth bands, e.g. [5]: < 1 mm, 1 mm 2 mm, 2 mm 4 mm and > 4 mm (or in inches < 0.04 in., 0.04 in. 0.08 in., 0.08 in. 0.16 in., > 0.16 in.). For depths > 4 mm (0.16 in.), the reflection amplitude shows a saturation behavior (see Fig. 1). The actual value of the saturation depth depends in particular on the sensor characteristics. Based on practical experience, the typical tolerance for the depth sizing may be approx. 1.0 mm to 1.5 mm (0.04 in. to 0.06 in.) for depths below the saturation level.

Length sizing of crack-like features is usually quite good. Here, the typical specification is 10 mm (0.4 in.) for lengths < 100 mm (4 in.) and 10 % otherwise. In addition to crack length and crack depth, the crack inspection tools referred to in this paper provide ultrasonic wall thickness data with a local resolution of approx. 3 mm (axial) x 80 mm (circumferential). The accuracy of the wall thickness measurement is approx. 0.4 mm (0.016 in.).

Figure 1: Amplitude response for crack-like reflectors and grading bands (schematically)

ASSESSMENT OF CRACKS General The objective of crack assessment of pressurized components is to calculate a maximum allowable operating pressure (MAOP) or critical pressure which the component can safely sustain in the presence of a crack. Several input data are required in order to perform an assessment: Pipe characteristics (diameter, wall thickness) Crack geometry (length, depth, orientation) Material properties (minimum specified yield strength SMYS, ultimate tensile strength UTS, toughness) Loading conditions (e.g. internal pressure) Based on these data the critical pressure can be calculated using available standards. The Failure Assessment Diagram (FAD) A well-established technique for the analysis of pressurized components containing a crack-like flaw is the Failure Assessment Diagram method described, e.g., in the standards API 579 [2] and BS 7910 [8]. A comprehensive explanation of the FAD method and its fracture mechanical background is given by Anderson [9]. A typical FAD is depicted in Fig. 2. For each crack-like defect detected, e.g., by a crack inspection tool, a so-called assessment point is calculated and plotted in the FAD. The coordinates (Lr, Kr) of this point depend on the geometry and the material properties of the considered component, the geometry of the crack and the relevant loading condition. The assessment curve (also called limit curve) separates the acceptable region of

Copyright 2012 by ASME

the FAD below this curve from the unacceptable area on and above the curve for which failure is predicted. The failure mode depends on the location of the assessment point: For small Lr and large Kr failure occurs by brittle fracture. At the other extreme (large Lr and small Kr), the failure mechanism is ductile overload causing burst by plastic deformation. In between these limits fracture is preceded by plastic deformation.
1.2 brittle fracture
Toughness ratio Kr

respective formulas are based on idealized flaw geometries (see e.g. [9]). As an example, surfacebreaking flaws are assessed as semi-elliptical flaws with length l and depth d (Fig. 3).

1.0

Figure 3: Idealized crack geometry


fracture + plastic deformation

0.8 0.6
0.4

SENSITIVITY ANALYSIS Background The mathematical expressions for the reference stress ref and for the stress intensity factor KI can become quite complex, depending on the flaw and the component geometries as well as on the loading condition. Since the FAD coordinates Lr and Kr are directly proportional to ref and KI, it therefore is not readily foreseeable how and how strong the assessment output (coordinates of assessment point, predicted critical pressure) is influenced by an input parameter and its inherent uncertainty. An example of how the assessment point responds to the variation of the input parameters is depicted in Fig. 4.
1.2 1 0.8 0.6 P=10MPa, length=25500mm, depth=4mm P=115MPa, length=100mm, depth=4mm

0.2 0.0 0.0 0.2

assessment point

plastic collaps 1.2

Figure 2: FAD with assessment point

0.4 0.6 0.8 Load ratio Lr

1.0

The load ratio Lr (horizontal FAD coordinate) is defined in Eq. 1 where ref is the so-called reference stress which is proportional to the primary membrane stress (e.g. the hoop stress due to internal pressure) and y is the material yield strength. Lr = ref / y (1)

The toughness ratio Kr (vertical FAD coordinate) is defined in Eq. 2. There, the stress intensity factor KI characterizes the local stresses in the vicinity of the crack tip. The fracture toughness Kmat is a material property describing the ability of a material to resist fracture when a crack is present. KI depends on primary loads (e.g. hoop stress due to internal pressure) as well as on residual stresses (e.g. residual stresses resulting from welding). Kr = KI / Kmat (2)

P=10MPa, length=100mm, depth=16.5mm

Kr

0.4
0.2 0 0 0.2 0.4

The failure assessment curve separates the FAD into an acceptable and an unacceptable region. The Level 2A failure assessment curve of BS 7910 which is equivalent to the API 579 Level 2 assessment curve is given by Eq. 3. The material dependent plastic collapse cut-off is Lr(max) = (y + u)/2y with y and u being the material yield strength and tensile strength, respectively. Kr = (1 - 0.14Lr ) (0.3 + 0.7 exp{-0.65Lr }) for Lr Lr(max) (3) Kr = 0 for Lr > Lr(max) Both the reference stress ref and the stress intensity factor KI depend on the component geometry, on the shape (length, depth), location (internal, external, embedded), and orientation (axial, circumferential) of the considered flaw as well as on the loading condition (internal pressure, bending stress, residual stress). The
2 6

Figure 4: FAD with assessment points for varying crack depth, crack length or pressure (D=20 in., t=0.5 in., steel grade X52)

0.6 Lr

0.8

1.2

The variation of crack depth from 1 mm (0.04 in.) to 6.5 mm (0.26 in.) in steps of 0.5 mm (0.02 in.) is indicated by the blue dots in Fig. 4. Here, a 100 mm (4 in.) long axial crack at a pressure of 10 MPa is assumed. At a depth transition from 6.0 mm (0.24 in.) to 6.5 mm (0.26 in.) the assessment point crosses the limit curve and, thus, failure is predicted. The green curve (diamonds) shows the effect of a variation of crack length from 25 mm (1 in.) up to 500 mm (20 in.) assuming a constant crack depth of 4 mm (0.16 in.). Independent of the length, the assessment points stay well below the limit curve. Thus the crack length has little effect in this situation.

Copyright 2012 by ASME

The black curve (crosses) is obtained for a pressure variation from 1 MPa (145 psi) to 15 MPa (2175 psi) in steps of 1 MPa at constant crack length and depth. As the coordinates Lr and Kr of the assessment point are proportional to the hoop stress the assessment point moves on a straight line starting at the origin of the FAD (as long as residual stresses can be neglected). The slope of this line depends on the geometry of the considered crack. In the present case, the critical pressure at which the assessment point reaches the limit curve is 12.85 MPa (1860 psi). Description of Defect Scenario In the following sections, it is analyzed how the results of crack assessment depend on the input parameters that are determined by means of ILI (i.e. crack depth d, crack length l and wall thickness t) and their respective sizing tolerances. The assessment also requires knowledge of the pipe material properties (yield strength, tensile strength, and toughness) for which specified minimum 1 values are usually considered . In order to investigate how the uncertainties of ILI data affect the results of pipeline crack assessment, an axial surface-breaking crack located in the base material away from welds is considered. Then the relevant load is only the primary (hoop) stress resulting from internal pressure and residual stresses can be neglected. Furthermore, it is assumed that the considered crack is isolated, i.e. located away from other anomalies, such that interaction effects are not relevant. For this sensitivity analysis, pipe diameters D from 10 in. to 40 in. and wall thickness values t from 0.25 in. up to 1 in. were considered. In all cases, a steel grade of API 5L X52 [10] was assumed (see Table 1).
Table 1: Input parameters used for the calculations Category Material properties Parameter SMYS y UTS u fracture toughness Kmat (converted from Charpy energy of 27 J) nominal diameter D Value 358 MPa (52000 psi) 455 MPa (66000 psi) 70 MPam (63.7 ksiin.) 273.1 mm (10 in.) 508.0 mm (20 in.) 762.0 mm (30 in.) 1016.0 mm (40 in.) 6.4 mm (0.25 in.) 12.7 mm (0.50 in.) 19.1 mm (0.75 in.) 25.4 mm (1.00 in.)

Newman & Raju (1982) solution for semi-elliptical internal axial surface flaws [11]. Comparison External Cracks vs. Internal Cracks In Fig. 5 the calculated critical pressure for an internal crack with d = 3 mm (0.12 in.) and l = 60 mm (2.4 in.) is compared to the results for an external crack of the same size. For D/t > 40 the critical pressure is virtually identical. For smaller D/t-values the critical pressure is higher for external cracks. However, the difference is less than 5 % for D/t > 12 which is valid for the considered cases. Therefore, all the calculations in the following were performed for internal cracks meaning that the results would be slightly conservative when applied to external cracks in the D/t range from approx. 10 to 40.
1.1 t=12.7mm t=19.1mm t=25.4mm

Pcrit (int) / Pcrit (ext)

1.05 1

0.95

0.9

0.85 0.8 0 10 20 30 40

Figure 5: Comparison of assessment results for internal and external cracks for different wall thicknesses

50 D/t

60

70

80

90

100

Normalized Pressure In the scope of the sensitivity analysis, a large variety of pipe geometries is considered resulting in a wide range of the design pressure Pd defined according to Eq. (4) where Fd is the design factor here assumed to be 0.72 and t is the nominal wall thickness. Pd = Fd 2t SMYS / D (4)

Pipeline geometry

Dividing the calculated critical pressure Pcrit by Pd the result of the FAD analysis can be expressed by the dimensionless quantity Pcn = Pcrit/Pd defining a normalized critical pressure Pcn. Pcn corresponds to the safety factor against failure for an internal pressure equal to Pd. Deterministic Analysis In the following sections it is deterministically analyzed how the critical pressure changes when varying a single input parameter. In this context "deterministic" means that the input parameters are used as given without taking into account any variations related to their statistical distributions. In contrast, the probabilistic analysis (later in the paper) includes the tolerances of the input parameters according to their respective distributions thus enabling the calculation of statistical properties such as PoF.

nominal wall thickness t

For the following calculations, the relevant reference stress equation according to BS 7910 was used and the stress intensity factor was computed according to the
1 Especially for LF-ERW pipe a lack of knowledge about the toughness in the weld area can be a problem for crack assessment.

Copyright 2012 by ASME

Note: In the figures shown below, colors represent the wall thickness t and symbols indicate the pipeline diameter D according to the definition given in Fig. 6.
t: 6.4mm (0.25"); D: 273.1mm (10") t: 6.4mm (0.25"); D: 508.0mm (20") t: 6.4mm (0.25"); D: 762.0mm (30") t: 6.4mm (0.25"); D: 1016.0mm (40") t: 12.7mm (0.50"); D: 273.1mm (10") t: 12.7mm (0.50"); D: 508.0mm (20") t: 12.7mm (0.50"); D: 762.0mm (30") t: 12.7mm (0.50"); D: 1016.0mm (40") t: 19.1mm (0.75"); D: 273.1mm (10") t: 19.1mm (0.75"); D: 508.0mm (20") t: 19.1mm (0.75"); D: 762.0mm (30")

1.4
1.2 1
Pcn

crack depth = 2 mm

0.8 0.6

t: 19.1mm (0.75"); D: 1016.0mm (40") t: 25.4mm (1.00"); D: 273.1mm (10") t: 25.4mm (1.00"); D: 508.8mm (20") t: 25.4mm (1.00"); D: 762.0mm (30") t: 25.4mm (1.00"); D: 1016.0mm (40")

0.4
0.2 0
1.4 1.2
1
Pcn

200

400 600 Crack length [mm]


crack depth = 4 mm

800

1000

Pcrit / Pd

Figure 6: Definition of colors and symbols used for labeling the 1.2 wall thicknesses and the pipe diameters
0.7

Variation 200 Crack Length 600 of 0 400 800 1000 In order to investigate the influence of crack length on Crack length [mm] critical pressure, the crack length was varied from 20 mm to 1000 mm (0.8 in. to 40 in.) while keeping all other input parameters constant. The resulting normalized critical pressure Pcn is shown in Fig. 7 as a function of crack length for cracks with a fixed depth of 2 mm (0.08 in.) and 4 mm (0.16 in.) and in Fig. 8 for cracks with a constant relative depth of 20 % and 50 % with respect to the wall thickness t. From Fig. 7 it is obvious that for a constant absolute crack depth Pcn increases with increasing wall thickness t (decreasing relative crack depth d/t). The wall thickness dependency of Pcn is less pronounced if the relative crack depth d/t is considered (Fig. 8). For constant d/t, however, Pcn still depends on the wall thickness t and decreases now with increasing t (increasing absolute crack depth d). This is due to the fact that the stress intensity factor KI is proportional to d (see, e.g. [9]). From Figs. 7 and 8 it is also obvious that the critical pressure as a function of crack length does not significantly depend on the pipe diameter as curves of the same color (same t but different D) show very similar behavior. The sensitivity of the assessment outcome Pcn with respect to crack length may be defined as the slope of the respective curves shown in Fig. 8. The sensitivity curves calculated according to this definition are depicted in Fig. 9. As an example, a slope of -0.002/mm is obtained for a 50 % deep and 100 mm (4 in.) long crack. This means that a change in length of 1 mm leads to a change of 0.002 (0.2 %) in Pcn. The negative sign indicates that an increase in crack length corresponds to a decrease in the critical pressure. Assuming a typical ILI sizing accuracy of 10 mm (0.4 in.) for the length measurement leads to an uncertainty of approx. 0.02 (i.e. 2 %) in the calculation of the critical pressure for this example.
0.2

0.8

0.6 0.4
0.2 0 200 400 600 Crack length [mm] 800 1000

Figure 7: Variation of critical pressure Pcn as a function of crack length for fixed values of absolute crack depth
1.4
1.2 crack depth = 20 %

1
Pcn

0.8 0.6

0.4
0.2 0 200 400 600 Crack length [mm] 800 1000

1.4
1.2 1
Pcn

crack depth = 50 %

0.8 0.6 0.4


0.2

200

Figure 8: Variation of critical pressure Pcn as a function of crack length for fixed values of relative crack depth

400 600 Crack length [mm]

800

1000

Copyright 2012 by ASME

0
Slope (Pcn vs. length) [1/mm]

-0.002 -0.004 -0.006 -0.008 -0.01 0


0

thickness of 19.1 mm (0.75 in.). As the sensitivity of the normalized critical pressure Pcn with respect to crack length is in good approximation independently of wall thickness and diameter (see Fig. 9) the behavior observable from Fig. 10 is also representative for pipes with other geometries: The sensitivity of Pcn with respect to crack length decreases with increasing crack length, for cracks longer than approx. 400 mm (16 in.) the critical pressure is virtually independent on crack length. For cracks shorter than approx. 400 mm (16 in.) the sensitivity of Pcn with respect to crack length increases with increasing crack depth.

crack depth = 20 % 200 400 600 Crack length [mm] 800 1000

Slope (Pcn vs. length) [1/mm]

-0.002 -0.004
-0.006 -0.008 -0.01 0 200 crack depth = 50 % 400 600 Crack length [mm] 800 1000

Variation of Crack Depth In order to investigate the influence of crack depth d on critical pressure, the depth was varied from 2 % to 80 % of the wall thickness while keeping the other input parameters constant. In Fig. 11 the resulting normalized critical pressure Pcn is shown as a function of absolute depth for all diameters and wall thickness values indicated in Table 1. The graphs on the top and bottom refer to cracks with a length of 100 mm (4 in.) and 400 mm (16 in.), respectively.
1.4
1.2

Figure 9: Sensitivity of Pcn with respect to crack length


0.000

crack length = 100 mm

Slope (Pcn vs length) [1/mm]

-0.002
Pcn

0.8

-0.004 -0.006

0.6
0.4

depth=10% depth=20%
0.2
0

-0.008 -0.010 0 200 400 600 Crack length [mm]

depth=30%
depth=50%

8 10 12 14 Crack depth [mm]

16

18

20

22

800

1000
1.4
1.2

Figure 10: Sensitivity of Pcn with respect to crack length for different crack depths; calculated for D = 508 mm (20 in.) and t = 19.1 mm (0.75 in.)

crack length = 400 mm

The curves shown in Fig. 9 are almost congruent for all considered values of diameter D and wall thickness t. The sensitivity of Pcn with respect to crack length is thus largely independent on the considered diameters and wall thicknesses and depends, for short cracks, mainly on the crack depth. Figure 10 shows how the sensitivity of Pcn with respect to crack length depends on the relative crack depth as calculated for a 20 inch pipe with a wall

Pcn

0.8

0.6
0.4

0.2
0

Figure 11: Variation of critical pressure Pcn as a function of absolute crack depth for fixed values of crack length

8 10 12 14 Crack depth [mm]

16

18

20

22

Copyright 2012 by ASME

1.4
1.2 1
Pcn

crack length = 100 mm

0.8

0.6 0.4
0.2 0 0
1.4 1.2
1

10

20

30 40 50 Crack depth [%]


crack length = 400 mm

60

70

80

Pcn

0.8
0.6

0.4 0.2
0 0 10 20 30 40 50 Crack depth [%] 60 70 80

Figure 12: Variation of critical pressure Pcn as a function of relative crack depth for fixed values of crack length
0
Slope (Pcn vs. depth) [1/mm]

crack length = 100 mm

-0.05
-0.1 -0.15

When plotted versus the relative crack depth, all curves for a constant crack length show a similar overall behavior (see Fig. 12): with increasing depth the critical pressure steadily decreases. This behavior is different to the influence of the crack length, where the critical pressure converges towards a depth-specific constant with increasing crack length. The observed variation of Pcn as a function of depth does not significantly depend on the pipe diameter as curves of same color (same t but different D) show a very similar behavior. From the lower image in Fig. 12 it can be deduced that cracks shallower than 15 % have a critical pressure above the pipeline design pressure Pd (Pcn > 1) and, thus, do not pose an immediate threat to pipeline integrity. In Fig. 13 the slope of the corresponding curves of Fig. 11 is depicted versus the relative crack depth. Similar to length this slope is now a measure of the sensitivity of Pcn with respect to variations in the absolute crack depth. Since given in the unit of 1/mm, the curves in Fig. 13 directly correspond to the uncertainty in Pcn associated with a crack depth sizing accuracy of 1 mm (0.04 in.). Assuming, for example, a 50 % deep crack with a length of 100 mm (4 in.) and a wall thickness t = 6.4 mm (0.25 in.) the uncertainty in Pcrit/Pd due to an uncertainty in depth of 1 mm amounts to approx. 0.22 (22 %). Likewise, using a wall thickness of t = 12.7 mm (0.5 in.) one obtains 0.10 (10 %). How strong Pcn is affected by a change in crack depth depends mainly on the actual wall thickness but also on the crack length and depth itself (see Fig. 13): The depth dependency of Pcn increases with decreasing wall thickness. Furthermore, for the considered examples the depth dependency of Pcn increases with crack length for cracks up to a depth of approx. 60 70 %. Variation of Wall Thickness Besides the crack length and the crack depth, the wall thickness t in the vicinity of the flaw is an essential input parameter required for the assessment. Figure 14 shows how the critical pressure of cracks with different depth is affected when the wall thickness is varied between -15 % and +15 % around the nominal value of 6.4 mm (0.25 in.), 12.7 mm (0.5 in.) and 19.1 mm (0.75 in.). The calculations were performed for crack depths of 20 %, 30 % and 50 % of the nominal wall thickness tnom assuming a crack length of 100 mm (4 in.), and a pipe diameter of 508 mm (20 in.). The displayed critical pressure Pcn is normalized with regard to the design pressure for the respective nominal wall thickness. In this figure, colors denote wall thickness according to the definitions in Fig. 6, symbols indicate crack depth d (square: d = 20 %, circle: d = 30 %, triangle: d = 50 %).

-0.2 -0.25
-0.3 0 10 20 30 40 50 Crack depth [%] 60 70 80

0
Slope (Pcn vs. depth) [1/mm]

crack length = 400 mm

-0.05 -0.1 -0.15


-0.2 -0.25 -0.3

10

20

Figure 13: Sensitivity of Pcn with respect to crack depth for lengths of 100 mm (4 in., top) and 400 mm (16 in., bottom)

30 40 50 Crack depth [%]

60

70

80

Copyright 2012 by ASME

1.5

1.3
1.1
Pcn

9 % for t = 6.4 mm (0.25 in.), respectively 1.6 % for t = 25.4 mm (1 in.). Influence of Steel Grade In the preceding part, the sensitivity of the normalized critical pressure Pcn was investigated with respect to crack length, crack depth and wall thickness assuming steel grade API 5L X52. In this section, the impact of the type of pipe steel (see Table 2) on the assessment is examined. Here, for all the steels grades considered the same value of 70 MPam is assumed for the fracture toughness Kmat (minimum requirement for Charpy energy according to API 5L). Although the critical pressure for a fixed crack size obviously increases with increasing tensile strength and, thus, depends on the steel grade, the sensitivity of Pcn with respect to variations in the crack length and crack depth is virtually independent of the steel grade as can be seen from Figs. 16 and 17.
0

0.9
0.7

0.5
0.3 4 6 8 10 12 14 16 Wall thickness [mm] 18 20 22

Figure 14: Variation of critical pressure Pcn as a function of a wall thickness variation of 15% as calculated for nominal wall thicknesses of 6.4 mm (0.25 in.), 12.7 mm (0.5 in.) and 19.1 mm (0.75 in.); symbols refer to crack depth (see text)
0.25 d=3mm, l=100mm d=20%, l=100mm d=30%, l=100mm d=50%, l=100mm d=3mm, l=400mm d=20%, l=400mm d=30%, l=400mm d=50%, l=400mm

Slope (Pcn vs. t) [1/mm]

0.20 0.15
0.10 0.05 0.00 6 8 10 12 14 16 18 20 22 24 26 Wall thickness [mm]

Slope (Pcn vs. length) [1/mm]

-0.001 -0.002
-0.003 -0.004

X42 X52 X60

-0.005
-0.006 -0.007

Figure 15: Sensitivity of Pcn with respect to wall thickness t

200

From Fig. 14 it can be seen that the critical pressure varies linearly with wall thickness for the considered wall thickness variations. The slope of the curves indicated in Fig. 14 is virtually independent of crack depth but significantly decreases with increasing nominal wall thickness. This is illustrated in Fig. 15 depicting the slope of the respective curves shown in Fig. 14. In addition, Fig. 15 also indicates the results for wall thicknesses up to 25.4 mm (1 in.) as well as for crack lengths up to 400 mm (16 in.). As can be seen, the slope is also independent on the crack length. Additionally, it is found that the results are almost the same for diameters from 10 in up to 40 in. From Fig. 15 one can readily see how the uncertainty of wall thickness data is related to the critical pressure. At a nominal wall thickness of e.g. 6.4 mm (0.25 in.), a variation of 1 mm (0.04 in.) leads to change of approx. 0.22 (22 %) in Pcn irrespectively of the actual crack dimensions. Likewise, at a nominal wall thickness of 25.4 mm (1 in.) a variation of 1 mm (0.04 in.) yields a change in Pcn of approx. 0.04 (4 %). Assuming now that the local wall thickness can be measured with an accuracy of 0.4 mm (0.016 in.), as achievable by ultrasonic ILI, one obtains an uncertainty in Pcn of approx.

400 600 Crack length [mm]

800

1000

Figure 16: Sensitivity of Pcn with respect to crack length for different steel grades, calculated for crack depth d = 50 %, D = 508 mm (20 in.), t = 12.7 mm (0.5 in.)
0
t=6.4mm, X42 t=6.4mm, X52 t=6.4mm, X60

Slope (Pcn vs. depth) [1/mm]

-0.05 -0.1 -0.15

t=12.7mm, X42 t=12.7mm, X52 t=12.7mm, X60 t=19.1mm, X42

-0.2 -0.25
-0.3 10 20 30 40 50 Crack depth [%] 60 70

t=19.1mm, X52 t=19.1mm, X60


t=25.4mm, X42 t=25.4mm, X52 t=25.4mm, X60

Figure 17: Sensitivity of Pcn with respect to crack depth for different steel grades, calculated for l = 100 mm (4 in.), D = 508 mm (20 in.)

Copyright 2012 by ASME

Table 2: Steel grades (data taken from [10])


Steel grade (API 5L) X42 X52 X60 SMYS 289 MPa (42 ksi) 358 MPa (52 ksi) 413 MPa (60 ksi) UTS 413 MPa (60 ksi) 455 MPa (66 ksi) 520 MPa (75.4 ksi)

IMPACT OF ILI SIZING TOLERANCES ON THE CRITICAL PRESSURE Crack assessment of pipelines is usually based on data provided by appropriate in-line inspection tools. Reliable assessment results have to rely on precise sizing information. In order to be on the safe side the sizing tolerances as specified by the tool provider have to be taken into account. Based on the results presented before it is straightforward to see how the assessment results (critical pressure) are affected by the ILI measuring tolerances as specified for crack depth, crack length, and wall thickness. For the example calculations below, the following sizing tolerances are considered:

wall thickness: 0.4 mm (0.016 in.) length: 10 mm (0.4 in.) for l 100 mm (4 in.); 10% otherwise depth: 1.0 mm (0.04 in) As an example, a pipe with diameter D = 20 in. is considered. The calculations are performed for nominal wall thicknesses of t = 12.7 mm (0.5 in.) and t = 6.4 mm (0.25 in.). The crack sizes are assumed to be: Length: 50 mm (2 in.), 100 mm (4 in.) and 400 mm (16 in.) Depth: 16 %, 31 % and 47 % (corresponding to 1 mm, 2 mm and 3 mm for t = 6.4 mm (0.25 in.), and 2 mm, 4 mm and 6 mm for t = 12.7 mm (0.5 in.), respectively). Table 3 shows the resulting uncertainties in the critical pressure Pcrit in %. Also included in Table 3 are the individual contributions to the uncertainty in Pcrit due to the tolerance of the respective input parameters (length, depth and wall thickness).

Table 3: Uncertainty in critical pressure Pcrit resulting from ILI sizing tolerances for crack depth, crack length, and wall thickness
t = 12.7 mm (0.50 in.) Crack size depth [%] 16 31 47 16 31 47 16 31 47 length [mm] 50 50 50 100 100 100 400 400 400 Uncertainty [%] in Pcrit due to uncertainty in depth, length, wall thickness 12 13 17 13 16 20 17 21 28 depth 7 7 8 9 10 12 13 16 21 length 1 2 5 1 1 3 0 1 1 t = 6.4 mm (0.25 in.) Uncertainty [%] in Pcrit due to uncertainty in depth 11 15 19 15 21 27 20 29 40 length 1 3 5 1 1 3 0 1 1 wall thickness 7 8 10 7 9 11 7 9 13

wall depth, length, thickness wall thickness 3 4 4 4 4 5 4 5 6 20 27 34 23 32 40 27 38 54

For a specific wall thickness the uncertainty in the assessment result increases with increasing crack depth and with increasing crack length. The uncertainty in the critical pressure is clearly governed by the depth tolerance while the contribution due to the length tolerance is almost negligible, in particular for cracks longer than approx. 400 mm (16 in.). The contribution of the wall thickness itself is relatively small when compared to the depth contribution. This is in particular due to the good accuracy of the wall thickness measurement.

The sensitivity analysis has shown that the influence on the critical pressure due to the uncertainties in crack depth and wall thickness increases with decreasing wall thickness (Figs. 13, 15). Thus, for pipes with smaller wall thickness higher uncertainties are expected while in pipes with thicker wall the uncertainties in the critical pressure will be lower. This can be clearly seen from the example in Table 3 where for a wall thickness of 6.4 mm (0.25 in.) the uncertainties in Pcrit due to the sizing tolerances for crack depth and wall thickness are approx. twice as high as the

Copyright 2012 by ASME

corresponding values for the thicker wall with t = 12.7 mm (0.5 in.). As the influence of the length sizing tolerance is negligible, the uncertainty in the critical pressure is primarily determined by the sizing tolerances for crack depth and wall thickness. Thus, the overall uncertainty in Pcrit is approx. inversely proportional to the wall thickness when assuming the same relative crack depth. As the sensitivity of the critical pressure with respect to the different input parameters is in good approximation independent on the pipeline diameter, the values in Tables 3 are also valid for other diameters. The following section focus on probabilistic crack assessment using the Monte Carlo method. This enables calculation of the probability of failure (PoF) for a given defect situation by taking into account the measuring tolerances of input parameters in terms of their statistical distributions. In particular, quantitative recommendations (e.g. pressure reduction) can be derived for a given defect scenario on how to reduce the PoF to a specific target value. PROBABILISTIC ANALYSIS Background In the previous sections it was shown how the variations of a single input parameter affect the outcome of the FAD analysis. Such deterministic calculations rely on a fixed set of input parameters (i.e. well-defined values for crack depth, crack length, fracture toughness, etc.). In practice, however, all input parameters are affected by specific uncertainties. A reliable deterministic assessment has thus to be based on conservative values for the input parameters. A different approach is the use of probabilistic fracture analysis methods which treat input values as probabilistic distribution functions rather than well defined deterministic quantities. This way, such methods can directly account for the uncertainties associated with different input parameters. A straightforward way of introducing probability into the FAD method is the use of the Monte Carlo simulation. The n input parameters p1...pn are described by means of n probability density functions and the FAD assessment is iteratively repeated, each time i with a set of parameters pi,1...pi,n randomly generated according to appropriate probability functions. Each Monte Carlo iteration i results in a single assessment point (Lr, Kr)i. The uncertainty in the input parameters leads to a scatter of these assessment points. The number of the assessment points lying outside the failure assessment curve (Nfailures) divided by the total number of N iterations gives the probability of failure (PoF): PoF = Nfailures / N. The knowledge of the PoF then allows an estimation of the risk associated with the failure of a technical component by using the definition risk = PoF x consequence of failure.

Example For the probabilistic assessment a 4 mm (0.16 in.) deep and 100 mm (4 in.) long crack in a pipeline with a diameter of 20 in. with a wall thickness of 12.7 mm (0.5 in.) is considered. A deterministic assessment of this reference crack with the material properties of Table 1 yields a critical pressure of 12.85 MPa (1860 psi) which is just below the pipeline design pressure of 12.89 MPa (1870 psi). For this reference crack Monte Carlo simulations were performed with different values for the depth and length sizing accuracies. For this probabilistic assessment it was assumed that the measurement errors are normally distributed and that sizing accuracies are given at a confidence level of 80 % (meaning that the actual value is within the specified accuracy with a probability of 80 %). To account for its statistical variation, the material fracture toughness was considered as being normally distributed with a mean value of 70 MPam and a standard deviation of 5 MPam. For the yield and tensile strength the specified minimum values according to the steel grade were used. The calculations were performed for an assessment pressure of 10 MPa. The relevant input parameters for the Monte Carlo simulations and the resulting PoF values are listed in Table 4. For each 6 analyzed case indicated in Table 4 N = 5 x 10 Monte Carlo iterations were calculated. Assuming a length sizing accuracy of 10 mm (0.4 in.) and a depth sizing accuracy of 1 mm (0.04 in.) yields a PoF of 7.98E-3 for the assessment pressure of 10 MPa (see case 2 of Table 4).
Table 4: Probability of failure for an assessment pressure of 10 MPa for different values of depth and length sizing accuracy case ID 1 2 3 4 5 6 sizing accuracy @ 80 % confidence level depth [mm] length [mm] 0.5 1.0 1.5 1.0 1.0 1.0 10 10 10 5 15 10 constant (70) normal distribution: mean = 70 std.dev = 5 Kmat [MPam] resulting PoF 2.41E-5 7.98E-3 4.78E-2 7.49E-3 8.73E-3 6.15E-3

The results of the Monte Carlo simulations corresponding to cases 1, 2, and 3 of Table 4 are shown on the left-hand side of Fig. 18 (top to bottom). With increasing depth sizing accuracy, the scatter of the assessment points reduces significantly leading to a considerable reduction in the PoF. As can also been seen from Table 4, the depth sizing accuracy has the strongest effect on the PoF (see case 1, 2 and 3) whereas variations in the length sizing accuracy

10

Copyright 2012 by ASME

(see case 2, 4 and 5) or variations in the fracture toughness (see case 2 and 6) do not significantly affect the assessment results.
assessment pressure 10 MPa POF = 2.41E-5 assessment pressure 8 MPa
POF = < 1E-6

7.1 MPa (1030 psi) to yield a PoF 10 . For a depth sizing accuracy of 1.0 mm (0.04 in.) a pressure reduction down to 8.7 MPa (1260 psi) would be sufficient to yield the same target PoF. Likewise, for a measuring accuracy of 0.5 mm (0.02 in.) a maximum pressure of 10.2 MPa -4 (1480 psi) would be tolerable to obtain a PoF 10 .
+/- 1.5 mm 1.0E+00 1.0E-01 1.0E-02 +/- 1.0 mm +/- 0.5 mm

-4

POF = 7.98E-3

POF = 4.70E-6

PoF

1.0E-03 1.0E-04

1.0E-05
6 7 8 9 10 Pressure [MPa] 11 12

POF = 4.78E-2

POF = 1.04E-3

Figure 19: PoF versus internal pressure for different values of depth sizing accuracy

Figure 18: FADs with results of Monte Carlo simulations for varying depth sizing accuracy (case 1 3 of Table 4); top: 0.5 mm; middle: 1.0 mm; bottom: 1.5 mm

The FADs on the left-hand side of Fig. 18 were calculated for an assessment pressure of 10 MPa (1450 psi). In the corresponding figures on the right-hand side the results of Monte Carlo simulations for a reduced internal pressure of 8 MPa (1160 psi) are depicted. Besides the pressure, all other input parameters were kept constant. As a consequence of the lower pressure the data points move towards the center of the FAD and the number of points outside the limit curve (i.e. the PoF) decreases. How much the PoF is lowered by a certain reduction in pressure depends on the considered situation (here: the sizing accuracy). The results of a more detailed analysis of pressure dependence of the PoF are visualized in Fig. 19. The three curves show the PoF of the considered reference crack (4 mm deep, 100 mm long) as a function of internal pressure for different values of depth sizing accuracy. The values for length sizing accuracy and material properties were kept constant and equal to the values for case 2 in Table 2. For a depth sizing accuracy of 1.5 mm (0.06 in.) the pressure would have to be reduced down to approx.

SUMMARY & OUTLOOK Based on the FAD method it was examined how the results of crack assessment depend on the input parameters that are determined from ILI (i.e. crack depth, crack length, and wall thickness) and their associated measuring tolerances. By introducing a normalized critical pressure Pcn the results of the sensitivity analysis can be applied to a wide range of pipe geometries and crack sizes. The sensitivity of the calculated critical pressure with respect to variations in the considered input parameters can be summarized as follows: Sensitivity of Pcn with respect to crack length: o The sensitivity of Pcn with respect to crack length is in good approximation independent of the pipe geometry (diameter, wall thickness). o The sensitivity decreases with increasing crack length; for cracks longer than approx. 400 mm (16 in.) the assessment result is virtually independent on changes in crack length. Sensitivity of Pcn with respect to crack depth: o The sensitivity of Pcn with respect to crack depth is in good approximation independent on the pipeline diameter but depends significantly on the wall thickness. Besides the wall thickness, the sensitivity depends on the crack depth itself and on the crack length. o The uncertainty in the assessment result is largely dominated by the depth tolerances when

11

Copyright 2012 by ASME

assuming tolerances currently available with inline crack inspection. o For crack depths < 15% the critical pressure is above the design pressure for all cases considered (independent of crack length and pipe geometry) Sensitivity of Pcn with respect to wall thickness: o The sensitivity of Pcn with respect to variations in the wall thickness is in good approximation independent of crack geometry and pipe diameter. o Assuming an accuracy of 0.4 mm (0.02 in.) for the wall thickness measurement (available from UT inspections) an uncertainty in Pcn of approx. 9 % is obtained for a wall thickness of 6.4 mm (0.25 in.). For a wall thickness of 25.4 mm (1 in.) the uncertainty in Pcn reduces to 2%.

The sensitivity of Pcn with respect to variations in the crack length and crack depth is virtually independent on the steel grades considered. Furthermore, it was demonstrated how the FADmethod can be combined with Monte Carlo simulations in order to calculate the probability of failure assuming statistical distributions of input parameters. The results can be used to estimate the necessary pressure reduction in the presence of a crack in order to stay below a specified probability of failure. In this work only defects in the base material under static loads were considered assuming given material properties. In the presence of dynamic loading the impact of fatigue crack growth needs to be included in the assessment if future behavior of the component under consideration needs to be evaluated. Furthermore, cracklike defects are often located near or at welds. Then, the assessment also has to account for different material properties involved (weld material, heat affected zone, base material) as well as for residual stresses due to the welding process. Furthermore it should be mentioned that other assessment methods than the FAD approach may yield different results depending on the actual situation. A general drawback of the currently applied assessment methods is the restriction to simplified defect shapes. Here, the availability of crack profiles (from field verification data or possibly from ILI data directly [12]) in combination with advanced FE (finite element) techniques will allow for further improvements with regard to defect assessment. ACKNOWLEDGMENTS The calculations were performed using the software tool PVrisk developed at the Fraunhofer Institute for NonDestructive Testing (FHG IzfP) in Saarbrcken/Germany by Prof. D. D. Cioclov and Dr. J. Kurz.

REFERENCES [1] Jger, C., Willems, H., "On the Possibilities of Pipeline Crack Assessment Using ILI Results", Pipeline Pigging and Integrity Management Conference Europe, Prague, 2011. [2] American Petroleum Institute and American Society of Mechanical Engineers, "API 579-1/ASME FFS-1 Standard Fitness-for-Service", 2007 [3] Wilkowski, G., "Leak-Before-Break What Does it Really Mean?" Journal of PVT, 122 (3), 2000, pp. 267272. [4] Willems, H., Barbian, O.A., "The State of Inline Crack Inspection using Ultrasonics", 3R International, Oct./Nov. 1999 (38), pp. 734-738. [5] Willems, H., Barbian, O.A., "State of In-Line Weld Inspection in Pipelines using Ultrasonics" Rio Pipeline Conference & Exposition, Rio de Janeiro, 2007. [6] Silk, M.G., 1984, "The Use of Diffractionbased Time-of-flight Measurements to Locate and Size Defects", British Journal of Non-destructive Testing, May 1984, pp. 208-213. [7] Brekow, G., Boehm, R., Brackrock, D., Kreutzbruck, M., "Quantitative Defect Sizing on Components with Different Wall Thickness Using UT-SAFT", 17th World Conference on Nondestructive Testing, Shanghai, 2008. [8] British Standard Institution, "BS 7910:2005 Guide on methods for assessing the acceptability of flaws in metallic structures", 2005. [9] Anderson, T.L., "Fracture Mechanics Fundamentals and Applications", 3rd Edition, CRC Press, 2005. [10] American Petroleum Institute: Specification for Line Pipe. ANSI/API Specification 5L, 2007. [11] Newman, J.C. & Raju, I.S. (1982) in: Murakami, Y. (ed.), "Stress Intensity Factors Handbook", Pergamon Press, 1987. [12] Barkdull, L., Willems, H., "Post Assessment of Ultrasonic Crack Detection Inline Inspection Data", International Pipeline Conference IPC, Paper IPC200864357, Calgary, 2008.

12

Copyright 2012 by ASME

Das könnte Ihnen auch gefallen