You are on page 1of 8

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

IPC2012-90236
PREDICTING THE FAILURE PRESSURE OF SCC FLAWS IN GAS TRANSMISSION PIPELINES
Raymond R. Fessler BIZTEK Consulting, Inc. 820 Roslyn Terrace Evanston, IL 60201 847-733-7410 BIZTEKrrf@aol.com David Batte Macaw Engineering, Ltd Floor 2, Q5 Quorum Business Park Newcastle upon Tyne NE12 8BS United Kingdom 4401159234503 david.batte@talk21.com Gabriela Rosca TransCanada st 450 1 Street S.W. Calgary, AB Canada T2P 5H1 403-920-2929 gabriela_rosca@transcanada.com Greg Van Boven Spectra Energy 1055 W. Georgia Street Vancouver, BC Canada V6E 3P3 604-691-5027 GVanBoven@spectraenergy.com

Gary Vervake Spectra Energy 5400 Westheimer Court Houston, TX 77056 281-543-8598 GTVervake@spectraenergy.com ABSTRACT
An important requirement for the management of stresscorrosion cracking (SCC) in natural gas transmission pipelines is the ability to predict accurately the burst failure pressure of flaws that have been discovered, particularly those found by crack detection in-line inspection (ILI). ASME B31.8S contains guidance for categorization of SCC based on predicted failure pressure for the cracks. Assessment of the segments is based on the severity category of SCC. As part of a Joint Industry Project (JIP) addressing the management of SCC in gas transmission pipelines, eight operators have assembled information relating to 85 in-service failures, hundreds of hydrostatic test failures, and dozens of pipe burst tests in which failure was due to SCC. Within the database are a wide range of pipe grades and sizes. Failures are due to both high pH and near-neutral pH SCC, and the flaws that initiated failure range from simple thumbnails to complex groups of cracks in a three-dimensional cluster. This paper presents some of the results from a comprehensive comparative study of the failure pressure predictions obtained using API 579 Level II, ln-secant, CorLAS and PAFFC methods for around 40 of the best-characterized datasets within the above database. From the results obtained, the sensitivities of the calculations to the calculation method used and to the input data, such as flaw profile, are examined. The

Sergio Limon Williams Gas Pipeline-West 295 Chipeta Way Salt Lake City, UT 84108 801-584-6787 sergio.limon@williams.com
results provide useful guidance to all those involved in predicting failure pressures as part of their threat management activities. Keywords: stress-corrosion cracking, fracture, failure pressure, pipeline

BACKGROUND
An important requirement for the integrity management of SCC in gas transmission pipelines is the ability to predict accurately the failure pressure of any flaws that are discovered. This is particularly the case for flaws discovered by ILI, but also for those discovered by excavations. The assessment and categorization of crack severity in accordance with the guidance in ASME B31.8S [1] is based on the predicted failure pressure derived from flaw dimensions. Several fracture-mechanics-based methods for predicting the failure pressure of axially oriented crack-like flaws in pipes have been developed and widely used for many years. While most of these methods have been extensively validated by comparison with burst tests on pipes containing artificial flaws, their ability to predict the failure behavior of complex-shaped flaws such as SCC has not been explored to any great extent. During the course of a JIP on SCC Management, the opportunity has been taken by the participating operators to assemble all the information available from well-documented in-

Copyright 2012 by ASME

service failures, hydrostatic test failures and burst tests on pipes removed from service, and to use this as the basis for a systematic examination of the abilities of the various methods to predict failure.

FAILURE PREDICTION METHODS


Several fracture-mechanics-based methods have been developed and are widely used within the pipeline industry. The earliest methods [2,3] dating back to the 1960s, the Battelle surface flaw equations, were based initially on flow-stressdependent failure and were applicable to materials with reasonably good toughness; they were subsequently modified and extended to include toughness-dependent failure. These equations, termed the NG-18 or ln-secant method, were validated against a database of burst tests on a variety of pipes containing surface-breaking axial flaws. The analytical method has recently been modified [4] to better represent the behavior of short flaws. In recent years, several alternative analytical methods have been developed, utilizing new advances in the understanding of fracture mechanics and materials behavior. The Pipeline Axial Flaw Failure Criterion (PAFFC, [5,6]) and CorLAS [7,8] consider both fracture-dependent and flow-stress-dependent failure. PAFFC includes consideration of stable ductile tearing prior to final failure, and CorLAS includes the option for an iterative calculation of effective flaw size based on actual measurements of flaw depth and length to find the lowest predicted failure pressure. Both methods are available as proprietary software. The Failure Assessment Diagram (FAD) approach considers toughness-dependent and flow-stress-dependent behavior concurrently, combining the load ratio and fracture ratio to determine whether a flaw is acceptable or unacceptable; again, stable ductile tearing prior to unstable failure is incorporated. The FAD approach has been incorporated in API 579 [9] and BS 7910 [10]. The derivation, validation and general application of these methods has been reported in detail elsewhere [11,12] and is not discussed further here.

subsequently subjected to fatigue loading. They obtained reasonably similar predictions using all three methods but found that, since that failures were of clusters of cracks rather than individual flaws, the choice of crack interaction rules had a significant influence on the results. They concluded that, for their cluster configurations, the CEPA interaction rules [15] were more consistent than the API 579 rules, which tended to be somewhat conservative. Rothwell and Coote [11] systematically applied NG-18 (in its original form), CorLAS, PAFFC and API 579 to a set of 22 in-service failures, hydrostatic test failures and pipe burst tests containing SCC from gas and oil pipelines. The sensitivity of the resulting predictions to the assumptions made about input data such as flow stress, toughness and flaw profile were examined. Rothwell and Coote concluded that, while all the methods considered could perform very well for results from wellcharacterized flaws of regular profile, the performance for flaws with irregular profiles was much more variable, with wide scatter between observed and predicted failure pressures. They concluded that the poor correlations arose largely because of the difficulty of representing real crack profiles (where they were known) in a consistent way within the models, most of which use a rectangular or elliptical shape. The exception was CorLAS, which was able to utilize the crack profile.

PRELIMINARY REVIEW OF DATA


Most of the participants in the JIP on SCC Management have experienced in-service or hydrostatic test failures due to SCC, and in many instances these have been investigated in depth by the companies concerned. Several participants have also conducted burst tests on pipe sections that contain SCC. Table 1 summarizes the information that was obtained from 11 sources for the study. TABLE1. SUMMARY OF INFORMATION AVAILABLE FOR FAILURE PREDICTION
Number & Type 5 failures 9 hydro I failure 9 hydro 1 failure 2 failures 1 hydro 12 burst 1 hydro 5 burst 1 hydro 2 burst 6 failures 13 hydro 3 hydro 5 burst 4 hydro a. b. Pipe Diameter, In. 8,36,42 8,20 24 24 14 26 26 16 30 30 36 24 11,36, 42 20-42 34 34,36 18 Grade X65 X52 X60 X60 X45 X52 X52 X52 X52 X52 X60 X50 X52, X65 X52X70 X52 X52, X65 X52 Failure stress, % SMYS 60-77 67-109 62 70-100 49 39,40 75 94-122 109 106 124,127 72-77 80-110 97-110 128-143 60-84 Defect depth, % wt 65-80 35-78 88 53-79 >90 80,90 60 33-72 50 36 66,50 73-82 28-67 70-100 29-49 65-94 Type of SCCa N N N N N b b b H H N H N N H N N

APPLICATION OF METHODS TO SCC FLAWS


Most of the validations of failure prediction methods have involved sharp, uniform cracks and crack-like features, both artificial (machined and spark-eroded notches, etc.) and natural (fatigue). Only in a few cases have stress-corrosion cracks been considered. The earliest examination of SCC failure predictions is reported in the NEB Report on SCC in Canadian oil and gas pipelines [13]. This included a comparative study of the applicability of NG-18, PAFFC and CorLAS for a set of burst tests on 14 SCCcontaining pipes removed from service. CorLAS and PAFFC resulted in predictions that usually were slightly conservative, with CorLAS giving slightly less scatter than PAFFC. Predictions obtained using the NG-18 method showed more conservatism and more scatter. Katz et al [14] compared the API 579 Level III method with the NG-18 method for predicting the failure pressures of several pipe burst tests on stress-corrosion cracks removed from service. They concluded that the API 579 Level III method was more accurate than the NG-18 method, which tended to be conservative for long cracks but not conservative for short cracks, despite having recently been modified [4] to include a stressmodifying factor for short cracks. Kariyawasam et al [15] used CorLAS, API 579 and BS 7910 to predict the failure pressures of several pipe burst tests on samples containing SCC that had been removed from service and

A B C D E F G H J

K L

N=NN pH, H=High pH Extensive corrosion in the cracks prevented determining whether they were intergranular (high pH) or transgranular (NN pH).

Copyright 2012 by ASME

Overall, 85 test cases are listed: 15 in-service failures, 46 hydrostatic test failures and 24 pipe burst tests. The pipe materials range from 8-inch to 42-inch diameter and grades X45 to X70, with the majority towards the middle of these ranges. Both high pH and near-neutral pH SCC are included, with maximum crack depths ranging from around 30% to over 90% of the wall thickness. Cracking generally has a high aspect ratio (surface length/depth), from around 10 to over 500, with the highest ratios being for cracks that had initiated at seam weld toes. Within the database, the failure-initiating SCC flaws have widely differing crack profiles from a single flaw closely resembling an elliptical profile to flaws that consist of two or more separate individual cracks; these cracks may be on a single radial-axial plane, particularly if they are at the toe of a seam weld, or they may be on multiple radial-axial planes within a three-dimensional cluster. The overall range of maximum crack depths and corresponding failure pressures (normalized as failure stress, % SMYS) is illustrated in Fig. 1. The in-service failures have the lowest failure stresses and (generally) deepest cracks. The hydrostatic test failure pressures are between 110% and 70% SMYS (30-70% depth), while the burst test pressures are generally higher, and some appear similar to the failure stress of flaw-free pipe. Overall there is, as expected, considerable scatter in the relationship between maximum flaw depth and failure stress, resulting largely from differences in flaw profile; there is no clear evidence that high pH and near-neutral pH SCC behave SCC FAILURE DATA differently.

API 579 Level II, using the measured toughness and the maximum depth over the total length. It is recognized that Level II was not intended for predicting failure pressures; it only provides the limiting pressure for fitness for service. Level III is intended for predicting failure pressures, but it requires a JR curve, which was not available, so Level II was used to see if it could be substituted. Space and time do not allow for a complete description of every case. Only a few representative examples are given in this paper. Complete listings of raw data, crack profiles, and material properties are published in the final report on this project [17] so that interested researchers can have access to the data.

RESULTS
There was one case where the crack profile was relatively smooth and not too different from a semi-ellipse, as shown in Figure 2, where the actual flaw profile is shown along with semiellipses that have the same area and either the same length or depth. The predictions for the failure pressure of that flaw are summarized in Table 2. For that case, the predictions from CorLAS and ln-sec were in excellent agreement with each other and with the actual failure pressure. The other two methods may have been influenced excessively by the shallow tails on the crack.

160 140 120 100 80 60 40 20 0 0

Failure stress, % SMYS

In service failures Hydrotests Burst tests

20

40

60

80

100

Defect depth, % wall thickness


FIGURE 1. RELATIONSHIP BETWEEN FAILURE STRESS AND MAXIMUM CRACK DEPTH FOR SCC-INITIATED FAILURES

FIGURE 2. PROFILE OF A SINGLE SIMPLE STRESSCORROSION CRACK IN A 36-INCH DIAMETER, 0.39 INCH WALL THICKNESS, X60 PIPE TABLE 2. PREDICTED FAILURE PRESSURES FOR THE FLAW ILLUSTRATED IN FIGURE 2. CALCULATED BY VARIOUS METHODS

COMPARATAIVE ANALYSIS
Detailed flaw profiles and material properties were used for the predictions of failure pressure. The following methods were used in the calculations: Ln-secant (NG-18) method, incorporating the most recent modification. Various choices of calculating the flow stress were explored. PAFFC CorLAS, using the actual material properties and the actual defect profile. The toughness was determined from testing, and the strain hardening was estimated using Option 2 correlation for pipeline steels. The J fracture toughness was estimated as 12 times CVN/Ac. The lower of the failure pressures using toughness and strength-based approaches was chosen.

Copyright 2012 by ASME

For the ln-sec method, the predictions based on a semiellipse with the same depth as the flaw were inferior to the predictions based on the length being the same. This trend was observed for most of the flaws that were studied. Considering all of the data, when the original depth was used, the predicted failure pressure based on the original depth was 30 to 38 % SMYS lower than the actual failure pressure with a standard deviation of 20 to 25% SMYS. When the original length was used along with flow stresses of the average of the yield strength and the ultimate tensile strength, 10,000 psi plus the yield strength, or the ultimate tensile strength, the predicted failure pressures were 11 to 20% SMYS lower than actual with a standard deviation of 16 to 18% SMYS. In addition, using the average of the yield strength plus the ultimate tensile strength or the yield strength plus 10,000 psi, or the ultimate tensile strength for the flow stress gave better results than were obtained using other values for the flow stress. Since there was little difference among those three methods, the average of the yield strength and ultimate tensile strength is used for the remainder of this paper. For clusters of small cracks, it is necessary to decide which ones interact such that they behave like a single crack. As stated above, the CEPA guidelines have been found to be better than most of the others. However, even those guidelines can lead to significant errors. Figure 3 shows the profile that was obtained from a 20-inch diameter, 0.252 inch wall thickness, X60 pipe. Although the profile looks relatively smooth, the fracture origin consisted of a number of small disconnected cracks that appeared to act as a single crack. The outside surface and fracture origin of that pipe are shown in Fig. 4.

Outside surface

Fracture surface at origin FIGURE 4. OUTSIDE SURFACE AND FRACTURE ORIGIN CORRESPONDING TO THE PROFILE IN FIGURE 3 TABLE 3. PREDICTED FAILURE PRESSURES FOR THE FLAW ILLUSTRATED IN FIGURE 3 CALCULATED BY VARIOUS METHODS
PFP, psig API 579 CorLAS PAFFC ln-sec 816 1359 961 1114 Actual PFP, psig 756 213 611 458 Actual PFP, % 48 14 39 29

FIGURE 3. PROFILE OF THE ORIGIN OF A SCC FAILURE IN A 20-INCH DIAMETER, 0.252 WALL THICKNESS, X60 PIPE The predicted failure pressures for the flaw are given in Table 3, where it can be seen that all of the methods predict failure pressures that are much lower than the actual failure pressure. There are at least two possible reasons for the discrepancies between the predicted and actual failure pressures. The metal between the cracks probably imparts greater strength than would be expected, and shielding of the cracks by neighboring cracks would be expected to lower their effect on strength of the pipe. This trend was very common among the cases that were studied. A similar effect was observed when the small cracks did not overlap. An example of this is shown in Fig. 5, where the origin appeared to consist of a number of interacting disconnected cracks.

FIGURE 5. PROFILE OF DISCONNECTED, NON OVERLAPPING STRESS-CORROSION CRACKS THAT FORM THE ORIGIN OF A FAILURE IN A 20-INCH DIAMETER, 0.252 WALL THICKNESS, X60 PIPE

Copyright 2012 by ASME

Outside surface Outside surface

Fracture surface at origin Fracture surface at origin FIGURE 6. OUTSIDE SURFACE AND FRACTURE ORIGIN CORRESPONDING TO THE PROFILE IN FIGURE 5 The predicted failure pressures corresponding to Fig. 5 and Fig. 6 are given in Table 4. TABLE 4. PREDICTED FAILURE PRESSURES FOR THE FLAW ILLUSTRATED IN FIGURE 6 CALCULATED BY VARIOUS METHODS
PFP, psig API 579 CorLas PAFFC ln-sec 876 1377 753 1489 Actual PFP, psig 721 220 844 108 Actual PFP, % 45 14 53 7

FIGURE 8. OUTSIDE SURFACE AND FRACTURE ORIGIN CORRESPONDING TO THE PROFILE IN FIGURE 7 TABLE 5. PREDICTED FAILURE PRESSURES FOR THE FLAW ILLUSTRATED IN FIGURE 7 CALCULATED BY VARIOUS METHODS
PFP, psig API 579 CorLas PAFFC ln-sec 841 1240 894 1016 Actual PFP, psig 704 305 651 529 Actual PFP, % 46 20 42 34

While most of the fracture origins were in the body of the pipe, there were some almost exclusively at the toe of the DSAW seam weld on tape-coated pipe. Figures 9 and 10 show an example of that.

There were a number of cases where two neighboring cracks were so close together that the CEPA guidelines said they should be treated as a single flaw longer than either one. An example of this is shown in Fig. 8. The predicted failure pressures are given in Table 5.

FIGURE 9. PROFILE OF A SINGLE CRACK THAT RECENTLY WAS FORMED BY THE COALESCENSE OF TWO SMALLER CRACKS

FIGURE 7. PROFILE OF TWO SEPARATED CRACKS THAT MIGHT INTERACT TO BE EQUIVALENT TO ONE LARGER CRACK

Copyright 2012 by ASME

similar and comparatively close to the one-to-one line. This relationship is shown quantitatively in Table 7.

Outside surface

Fracture surface at origin FIGURE10. OUTSIDE SURFACE AND FRACTURE ORIGIN CORRESPONDING TO THE PROFILE IN FIGURE 9 Figure 10 (top) shows some secondary cracks at the toe of the seam weld near the origin of the fracture. Figure 10 (bottom) shows the fracture surface at the origin. The predicted failure pressures are given in Table 6. TABLE 6. PREDICTED FAILURE PRESSURES FOR THE FLAW ILLUSTRATED IN FIGURE 9 CALCULATED BY VARIOUS METHODS
PFP, psig API 579 CorLas PAFFC ln-sec 978 983 538 785 Actual PFP, psig 22 17 462 215 Actual PFP, % 2.2 1.7 46 22

FIGURE 12. UNITY PLOT FOR CorLAS RESULTS

The predictions for all cases where full crack profiles were available are shown in Figure 11 where the solid line indicates a one-to-one relationship and the dashed line is 10% above the solid line. Most of the predictions are conservative. The few that are not are within 10% of the one-to-one relationship.

FIGURE 13. UNITY PLOT FOR LN-SEC RESULTS

FIGURE 11. UNITY PLOT FOR ALL OF THE DATA FOR WHICH CRACK PROFILES WERE AVAILABLE Unity plots for the individual methods are shown in Figures 12 to 15. It can be seen that the CorLAS and ln-sec results are

Copyright 2012 by ASME

FIGURE 14. UNITY PLOT FOR API 579 RESULTS

underpin the application of SCC threat management based on ILI and SCC direct assessment. Although the predictive models have been shown to be relatively accurate for single, simple crack profiles, such as those produced by machining or fatigue, the average error and the scatter in it are much greater for stress-corrosion cracks, which usually have irregular profiles and occur in clusters or colonies. Both of those features could conceivably contribute to the problem. The ligaments between the deeper parts of the cracks might impart more strength than expected, although there were no obvious correlations between the nature of the crack profiles and the errors in the predicted failure pressures. Perhaps moe significant is the likelihood that neighboring cracks would be expected to shield the largest crack to some extent. The semi ellipses that were used in the ln-sec calculations had the same area as the actual crack and the same total length. The depths of the semi ellipses typically were 50-75% of the maximum crack depth. Even so, the predicted failure pressures were usually too low. If the defects were circumscribed by a semi ellipse with length and depth equal to the maximum length and maximum depth of the defect, as required by BS 7910, the predicted failure pressures would have been still lower. The results of this study are consistent with reports in the literature that, with the possible exception of API 579 Level III (which could not be included for lack of JR curves), CorLAS predictions usually were the most accurate. However, the relative success of ln-sec and the relatively poor performance of PAFFC were inconsistent with previously published studies. The improvement in ln-sec may be due to the recent revisions in the code and to the choice of a flow stress based on actual tensile properties. Notwithstanding the uncertainties in the predictions, it is encouraging to note that almost all the failure predictions are conservative. However, this gives rise to concerns when the predictions are used to assess the severity of ILI-discovered defects and determine the manner of their mitigation. Because the calculations are usually conservative, and in some cases very conservative, they can lead to over-predictions of defect severity and a disproportionate response by the operator. Despite the advances that have been made during the present study, there is clearly a need for further improvements in the accuracy of failure pressure predictions, particularly for complex flaws.

FIGURE 15. UNITY PLOT FOR PFFAC RESULTS TABLE 7. AVERAGE DIFFERENCES BETWEEN ACTUAL AND PREDICED FAILURE PRESSURES Actual PFP, %SMYYS 42 17 34 17 Standard Deviation, %SMYS 19 14 30 18

CONCLUSIONS
A significant database of well-documented in-service failures, hydrostatic test failures and pipe burst tests has been used to explore the accuracy of failure prediction calculations for SCC. All the methods gave rise to conservative predictions; methods that simply used maximum flaw depth and overall flaw length were the most conservative, sometimes excessively so. Methods that focused on the part of the flaw with greatest depth, or that used an ellipse of equivalent area, were generally less inaccurate. The ability of CorLAS to undertake several iterative calculations with different flaw profiles and select the worst case usually resulted in the best overall performance.

API 579 CorLAS PAFFC ln-sec

ACKNOWLEDMENTS
The work described in this paper was part of a Joint Industry Project (JIP) with the generous support of the following companies: CenterPoint Energy, El Paso Pipeline Group, Kinder Morgan Energy Partners, NiSource Gas Transmission and Storage, Panhandle Energy, Spectra Energy, TransCanada Pipelines Limited, and Williams.

COMMENTS AND DISCUSSION


The assembly of 85 well-documented cases of failure due to SCC flaws has provided a substantial database that can be used to compare the effectiveness of the different failure-prediction methods, or hopefully to develop a better method for making those predictions. The present work represents a substantial step forward in determining the failure behavior of SCC, helping to

Copyright 2012 by ASME

REFERENCES
[1] ASME/ANSI B31.8S Gas Transmission and Piping Systems, 2004, Managing System Integrity of Gas Pipelines: Supplement ASME B31.8S, ASME. [2] Kiefner, J, 1969, Paper G presented at 4th Symposium on Line Pipe Research. [3] Kiefner, J, Maxey, W, Eiber, R ,and Duffy, A, 1973, Failure Stress Levels of Flaws in Pressurized Cylinders, ASTM STP536, Progress in Flaw Growth and Fracture Toughness Testing. ASTM, Philadelphia, p461-481. [4] Kiefner, J, October 6th, 2008 and October 13th, Modified Equation Aids Integrity Management, Oil and Gas Journal October 6th 2008 p78-82 and October 13th 2008 p64-66 [5] Leis, B, Brust, P, and Scott, P, 1991, Development and Validation of a Ductile Flaw Growth Analysis for Gas Transmission Line Pipe, PRCI Report Cat No L51643. [6] Leis, B and Ghadiali, N, 1994, Pipeline Axial Flaw Failure Criteria PAFFC, PRCI Report Cat No L 51720. [7] Jaske, C, Beavers, J, and Harle, B, 1996, Effect of Stress Corrosion Cracking on Integrity and Remaining Life of Gas Pipes, paper presented at NACE Corrosion 96. [8] Jaske, C and Beavers, J, Development and Evaluation of Improved Model for Engineering Critical Assessment of Pipelines, Paper IPC 2002-27027, 4th International Pipeline Conference, Calgary, 2002. [9] Fitness-for-service, API Recommended Practice API 579, 2000, API Publishing. [10] Guide on Methods for Assessing the Acceptability of Flaws in Fusion Welded Structures. BS 7910:2005, British Standards Institution, London. [11] Rothwell, A and Coote, R, 2009, A Critical Review of Assessment Methods for Axial Planar Flaws in Pipe, Paper No. 21009.052, Pipeline Technology Conference, Ostend. [12] Kiefner, J and Leewis, K, 2010, Guidance for the Assessment of Pipeline Defects, PRCI Report PR-218-05404. [13] Stress Corrosion Cracking in Canadian Oil and Gas Pipelines, 1996, National Energy Board Report MH-2-95. [14] Katz, D, Gao, M, Limon, S, and Krishnamurthy, R, 2005, Advances in Crack Assessment for Pipeline Integrity, Paper presented at 11th International Conference on Fracture, Turin. [15] Kariyawasam, S, Arumugam, U, Callar, G, Clarke, Huggar, A, Senf, P,, and Law, M, 2007, Stress Corrosion Crack Detection, Analysis and Assessment Improvements for Effective Integrity Management, paper presented at 16th APIA/PRCI/EPRG Biennial Conference on Pipeline Research, Canberra. [16] Stress Corrosion Cracking Recommended Practice, 2007, Second Edition. Canadian Energy Pipeline Association, Calgary. [17] Final Report on JIP, to be published by ASME.

Copyright 2012 by ASME

You might also like