Leak Versus Rupture Considerations for Steel Low-stress Pipelines

GRI-00/0232

Topi cal Report

Leak Leak versus Ruptur e Consid Consid erations erations for Steel teel Lo w-Stress w-Stress Pipelines

Prepared by: B. N. Leis O. C. Chang T. A. Bubenik Battelle 505 King Avenue Columbus, Ohio 43201

for Gas Research Institute 1700 1700 S Moun Moun t Prosp ect Ave Des Plaines, Illinois 60018-1804

Cont ract No. 5000-270-81 5000-270-8194 94 January 2001 2001

Legal Notice This report was prepared by Battelle as an account of contracted work sponsored by the Gas Research Institute (GRI). Neither GRI, members of GRI, Battelle, officers, trustees, or staff of Battelle, nor any person acting on their behalf: Makes any warranty or representation, expressed or implied, with respect to the accuracy, completeness, or usefulness of the information contained in this report, or that the use of any apparatus, methods, or process disclosed in this report may not infringe upon privately owned rights; or  Assumes any liability liability with respect respect to the the use of, or for damages damages resulting resulting from from the use of, any information, apparatus, method, or process disclosed in this report.

ii

Form Approved OMB No. 0704-0188

REPORT DOCUMENTATION PAGE

Public reporting burden for this collection of information is estimated to average 1 hour per response, including the time for reviewing instructions, searching existing data sources, gathering and maintaining the data needed, and completing completing and reviewing the collection of information. Send comments regarding this burden estimate or any other aspect of this collection of information including suggestions for reducing this burden to Washington Headquarters Services, Services, Directorate for Information Information Operations and Reports, 1215 Jefferson Davis Highway, Suite 1204, Arlington, VA 22202-4302, and to the Office of Management and Budget, Paperwork Reduction Reduction Project (0704-0188), Washington, D.C. 20503.

1. AGENCY USE ONLY (Leave blank)

2. REPORT DATE

3. REPORT TYPE AND DATES COVERED

January 2001

draft topical - March 2000 to January 2001

4. TITLE AND SUBTITLE

5. FUNDING NUMBERS

Leak versus Rupture Considerations for Steel Low-Stress Pipelines

GRI contract no. 5000-270-8194

6. AUTHOR(S)

B. N. Leis, O. C. Chang and T. A. Bubenik 7. PERFORMING ORGANIZATION ORGANIZATION NAME(S) AND ADDRESS(ES) ADDRESS(ES)

8. PERFORMING ORGANIZATION ORGANIZATION REPORT NUMBER

Battelle 505 King Avenue Columbus, Ohio 43201-2693 9. SPONSORING/M ONITORING AGENCY NAME(S) AND ADDRESS(ES)

10. SPONSORING/MONITORING  AGENCY REPORT REPORT NUMBER

Gas Research Institute 1700 S Mount Prospect Ave Des Plaines, Illinois 60018-1804

GRI-00/0232

11. SUP PLEMENTARY NOTES

12a. DISTRIBUTION/AVAILABILITY DISTRIBUTION/AVAILABILITY STATEMENT

12b. DISTRIBUTION CODE

13. ABSTRACT (Maximum 200 words)

Pipeline codes and regulations worldwide have less stringent requirements for low-wall low-wall stress pipelines. Factors underlying this include a reduced exposure zone as pipeline pi peline pressure decreases, and the expectation that leaks occur for lower wall stress. This report evaluates leak versus rupture as a function of wall stress with a focus on pipelines operating between 20 to 40  percent of the specified minimum yield stress. Potential threats to the integrity of low-stress pipelines are identified and worst-case scenarios for these threats evaluated to determine leak versus rupture rupture as a function of wall stress. The evaluation is  based on an assessment of existing full-scale test data, incident experience in the United States, and mechanics and fracture calculations. For corrosion, these necessarily conservative thresholds were: 35 percent of S MYS based on full-scale testing, 35 percent of SMYS for the OPS incident database, and 30 percent of SMYS for mechanics modeling. Were other than worst-case circumstances evaluated, higher thresholds would be obtained. On this basis the leak to rupture transition transition for corrosion defects can be taken as 30 percent of SMYS. SMYS. The threshold for for delayed mechanical damage was likewise evaluated. Fullscale test data indicated this threshold was above that identified for rupture due to corrosion, whereas the steels represented in reportable incident database indicated a threshold on order of 25 percent percent of SMYS. Analysis indicates that delayed mechanical damage incidents reflect the combination of several unlikely circumstances. For this reason the mechanics-based mechanics-based threshold is formulated in probabilistic terms. As yet this work is not complete. Thus, at present, the leak to rupture transition for delayed mechanical damage is taken as 25 percent of SMYS. 14. SUBJECT TERMS

15. NUMBER OF PAGES

low-wall-stress pipelines, safety, integrity, leak versus rupture 16. PRICE CODE 17. SECURITY CLASSIFICATION CLASSIFICATION OF REPORT

Unclassified 

18. SECURITY CLASSIFICATION OF THIS PAGE

19. SECURITY CLASSIFICATION OF ABSTRACT

Unclassified 

20. LIMITATION OF ABSTRACT

Unclassified 

N 7540-01-280-5500

Standard Form 298 (Rev.2-89)

iii

Executive Summary This report evaluates leak versus rupture as a function of wall stress, with a focus on naturalgas pipelines operating through high-consequence areas. The expectation that leaks occur in lieu of ruptures in low-wall-stress systems has been evaluated in terms of operations and maintenance, as opposed to design and construction, because the report addresses concerns with existing pipelines. Quantifying leak versus rupture as a function of wall-stress (pressure) level requires use of line-pipe properties, as well as consideration of pipeline geometries and operating conditions. These are defined for purposes of this evaluation. Thereafter, the threshold stress stress for the transition from leak to rupture was evaluated primarily with reference to results of full-scale testing and field failure experience, and what mechanics analysis and numerical failure models  predict. While the process used to evaluate the stress for the transition from from leak to rupture is comparable for prescriptive- and performance-based integrity management p lans, the results in this report only address prescriptive plans. Accordingly, this transition is evaluated with a conservative basis made necessary to encompass the breadth of federally regulated low-wall stress pipelines. The lower-bound threshold for the transition from leak to rupture in the low-wall-stress  pipeline system was evaluated with reference to full-scale test data, incident data, and mechanics and fracture analysis. For corrosion, these necessarily conservative thresholds thresholds were: 35 percent of SMYS based on full-scale testing, 35 percent of SMYS for the OPS incident database, and, 30 percent of SMYS for mechanics modeling. Were consideration given to other than the worst-case circumstances needed to reflect the  breadth of conditions covered by federally regulated pipelines, higher thresholds would be obtained. Given the results generated, the leak to rupture transition for for corrosion defects in the low-wall-stress pipeline system can be taken as 30 percent of SMYS, a value that is conservative in comparison with in-service incidents. Thresholds for the transition from leak to rupture also were evaluated for immediate as well as delayed mechanical damage incidents with reference to full-scale test data, incident data, and mechanics and fracture analysis. Full-scale test data indicated this threshold was in in excess of 30 percent of SMYS, the lowest threshold identified for rupture due to corrosion, whereas the steels represented in reportable incidents possess toughness indicated a threshold on order of 25 percent of SMYS. Analysis indicated that rupture due to delayed mechanical damage required the coincidence of many unlikely circumstances. circumstances. Probabilistic calculations best indicate the coupled likelihood of such events, and are currently being done. In the absence of results from this continuing work, the threshold for rupture due to mechanical damage must be taken as the lesser of the above-cited abo ve-cited results, that is 25 percent of SMYS. The main conclusion is that thresholds for the transition from leak to rupture are consistent with the current regulatory provisions for low-wall-stress pipelines.

iv

Table of Contents Page Background................................................................................................................................1 Introduction................................................................................................................................2 Leak versus Rupture in Gas Pipelines ....................................................... ................................4 Pipeline Safety Threats ......................................................................................................... .....6 Approach..................................................................................................................................10 Property Trends with Vintage and Grade ................................................................................11 Typical Pipeline Geometries....................................................................................................16 Pipeline Service Conditions.....................................................................................................17 Leak versus Rupture Conditions..............................................................................................18 for the Two Typical Gas Pipelines...........................................................................................18 Results and Discussion ......................................................................................................... ...22 Code and Regulatory Implications.......................................................................................22 Corrosion and Metal Loss....................................................................................................22 Full-Scale Static Testing..................................................................................................22 Reportable In-Service Incidents.......................................................................................23 Mechanics Prediction.......................................................................................................24 Stress-Corrosion Cracking and Fatigue ...............................................................................24 Mechanical Damage.............................................................................................................25 Full-Scale Static Testing..................................................................................................25 Reportable In-Service Incidents.......................................................................................27 Cyclic-Stable Tearing and Failure due to Delayed Mechanical Damage........................30 Summary and Conclusions ..................................................... .................................................32 References................................................................................................................................33

v

Background Improvements to federal pipeline design, construction, operation, and maintenance regulations focus attention on enhanced pipeline safety in high-consequence areas (HCA’s). A key consideration in formulating provisions to ensure safety in HCA’s is the v olume and rate of natural gas released if the line should fail for any reason. Leaks involve slow, controlled releases with limited volume, whereas ruptures involve larger, higher-rate releases. For this reason, much attention is centered on whether a natural-gas pipeline leaks or ruptures when an incident occurs. This report evaluates leak versus rupture in reference to low-wall-stress  pipelines, to aid in formulating a rational basis for the pending prescriptive changes.

1

Introduction Energy pipelines play a vital role in the safe and efficient transport of energy resources from the supply basin to market, covering distances often in excess of a thousand miles. This report deals with the portion of the energy pipeline system that moves natural gas – focusing on the  portion of the natural gas system that operates at lower stresses. Lower stress for present purposes is defined with reference to federal regulatory provisions governing natural-gas pipelines in the US, which is CFR Title 49, Part 192. What Part 192 considers low versus high stress can be inferred from analysis of paragraph s that distinguish design, construction, operation, and maintenance practices in terms of stress level, as follows. With reference to §192.5, the regulations set the maximum hoop stress in the wall thickness at 72 percent of the specified minimum yield strength (stress)1. Class location provides for thicker pipe as the perceived risk increases in terms of increasing population density. In the worst case, the class-location concept sets the minimum allowable working stress at 40 percent of the specified minimum yield stress (SMYS). Several other sections of the regulations include requirements associated with stress levels greater than 20 percent, 30 percent, or 40  percent. Many of these paragraphs involve operation above 40 percent, reflecting conditions dealt with by the class-location concept. The 30 percent bound involves up-rating, pressure testing, unique construction features, and regional differentiation of distribution versus transmission piping. The lowest bound on stress considered in Part 192 is 20 percent, which involves concerns for secondary stresses due to external loading or pipe shape (dents), as well as quality-control (QC) testing involving details such as welds. It follows that “low stress”  pipelines can be reasonably defined as pipelines operating at stresses between 20 and 40  percent of SMYS. Clearly, it is reasonable to expect that failure of low-wall-stress pipelines will result in a leak  because the stress driving fracture in such cases is much lower as compared to that in crosscountry transmission pipelines. The consequences associated with a leak likewise are reasonably expected to be less severe than where failure involves a rupture. This is because leaks lead to limited volume, slow releases, as compared to longer duration, larger volume releases from a rupture. The expected difference in consequences between pipelines operating at higher stress and those running with a lower wall stress is recognized in the regulations through provisions for the lower stress pipelines that are less stringent as compared to pipelines governed by class-location provisions. For this reason the design, construction, operation, and maintenance of such pipelines should ensure that if a failure occurs in such pipelines it is a leak, not a rupture. Likewise, the primary safety concern for such systems is ensuring that those provisions limit the likelihood of failure, and lead to a controlled limited-volume release if failure occurs. Recognizing this, the present report quantifies conditions for which leaks occur in lieu of ruptures as a function of decreasing wall stress. Because this report deals with existing natural-gas pipelines through high-consequence areas, the expectation that leaks occur in lieu of ruptures in low-wall-stress systems is evaluated in 1

Excluding lines operated to 80 percent under a grandfather clause.

2

terms of factors involved with operations and maintenance, as opposed to factors considered in design and construction. The circumstances that underlie this evaluation are presented in terms of line-pipe properties as a function of vintage, pipe geometry, and pipeline operating conditions. The next several sections outline the approach taken and define the scope.

3

Leak versus Rupture in Gas Pipelines Pipeline failures initiate when the wall of the pipe is breached by a crack or other defect. The opening that forms in natural gas pipelines either leaks or ruptures, depending on the length and type of defect, the line-pipe steel, the pipe geometry, and the pressure, temperature and composition of the gas. A leak occurs when the opening that forms is stable, that is, the length of the defect does not increase through unstable2 crack extension when the wall is breached. Because the length does not increase beyond that when the crack breaks through the wall, the leak-path is tight, and the area through which the gas can pass is small. The limited length, tight crack path, and limited area of the breach are characteristics of leaks. These traits restrict the gas flow associated with a leak, so that a small volume and low release-rate are also characteristics of a leak. A rupture occurs when the length of the defect exceeds a critical value, which leads to unstable extension of the breach. In some cases the unstable extension leads to a breach that is not much longer than that for a leak. Even so, the slit formed as the unstable extension occurs typically leads to local bulging, which increases the area of the slit and so also increases the volume and rate of flow. Occasionally the unstable propagation extends the length of the  breach to the order of the pipe diameter. But, more typically, unstable propagation occurs over several diameters, often to a length on the order of a pipe joint, and usually leads to a full-bore opening. For this reason, unstable growth of the breach that occurs with a rupture leads to  potentially much larger openings, and a correspondingly larger volume and rate of gas release. Rupture can be predicted as a function of the pipeline diameter and wall thickness, the  pressure, temperature and composition of the gas, the inherent fracture toughness of the line pipe steel, and the length of the defect when the wall is breached. These predictions are based on models that are validated by full-scale testing(e.g., see 1,2), and by such testing proven accurate for the range of conditions they are applied to. These predictions indicate that if the flaw is long enough, or the inherent line pipe toughness is low enough, the defect will be unstable when the pipe wall is breached and rupture will occur. How far it propagates unstably depends on the pipe geometry, the driving pressure in the pipeline, and the properties of the gas. Short ruptures occur when the energy driving crack growth is released, becoming too small to sustain it. This happens because, once the wall is breached and the defect begins to grow (rupture), the loading changes ahead of the crack tip. If the crack grows faster than the rate at which the  pipeline decompresses, growth continues. If the reverse occurs, the decompression wave  passes the crack tip, releasing the hoop stress that drives cracking before the crack arrives – so arrest occurs. Higher toughness steels dissipate more energy than less tough steels. For this reason tougher steels slow the speed of propagation. Slowing the fracture speed allows decompression wave 2

Unstable here has a technical meaning associated with effectively instantaneous axial crack extension, which is referred to as running fracture. Unstable extension beyond the initial breach occurs at speeds up to ~1300 feet-second-1. The term stable has a technical meaning that refers to a breach that forms under quasi-static conditions, which does not extend beyond the length associated with the leak.

4

to reach the crack tip in less time. As time equates to distance propagated, tough steels arrest fracture sooner, meaning a shorter fracture length and the possibility of a leak.  Lower pressure also means less energy is available to drive propagation, and so the likelihood for a long rupture is reduced. Lower pressures and reasonable toughness increase the likelihood for a stable breach, which translates to an increased likelihood that a leak will occur, or that arrest will be almost immediate. This technical basis underlies the expectation that lower-stress lines will leak rather than rupture.

5

Pipeline Safety Threats Because safety is the focus, leak versus rupture have been quantified in terms of the various threats to pipeline safety. Companion projects(e.g., see 3) have evaluated these threats based on analysis of the Office of Pipeline Safety (OPS) reportable3 incident database in term of the frequency of occurrence of a particular incident type. When evaluated in that format, these results point to incidents associated with outside forces and corrosion as the primary threats in cases where operational pressure and maintenance control failure. For this reason, analysis of leak versus rupture began in terms of these threats. Continuing growth of population centers means the proportion of mileage in the higherdensity-population classes will increase(4). This points to a continuing dominance of incidents associated with outside forces category unless action is taken to otherwise prevent such incidents. Of particular concern are incidents due to outside-forces contact with the pipeline by other than the pipeline owner or his contractors, which leads to third-party damage. Risk exposure to outside forces and other threats depends on both likelihood and consequences, which both increase as population density increases. The potential significance of outside forces and corrosion can be simply evaluated in this context by way of the classlocation category in the OPS database. Risk referenced to the data available in the OPS database is defined as the occurrence of an incident multiplied by the reported cost. The resultant risk has been evaluated on a per-mile basis to remove the bias associated with the significantly different mileages operating in each class location. This has been done based on data(5) gathered in a survey representing about 50,000 miles of pipeline operated by six transmission companies, a portion of which operates at less than 30 percent of SMYS. These data are presented in Figure 1. Because the data are limited for low-wall-stress pipelines, the incident database has been evaluated without regard to stress. Figure 1 presents the resulting distribution of mileage grouped by geographic region, and further subdivided by Class Location. This distribution of mileage is taken herein as representative of the distribution of mileage in the US. Figure 1 further splits the mileage in terms of population density in the counties through which the pipeline passes. Two categories are included. One represents higher-density areas, such as cities, while the second represents light- to medium-density areas, such as towns and small villages. The mileage in higherdensity areas represents about 20 percent of that in the light- to medium-density areas for this database. The results in Figure 1 indicate that the major difference between the more highly populated areas and the light- to medium-density population areas lies in the Class 2 and Class 3 pipeline

3

Reportable incidents involve three types of event. The first involves release of gas from a pipeline or of liquified natural gas (LNG) or gas from an LNG facility and i) a fatality or personal injury necessitating in-hospitalization; or ii) estimated property damage, i ncluding costs of gas lost by the operator or other, or both of $50,000 or more. The second involves an emergency shutdown of an LNG facility. The third involves an event judged significant by the operator, even though it did not fit 1) or 2) above.

6

mileage. Very little mileage lies in Class 4 Locations for this survey based on almost 50,000 miles of gas-transmission pipeline. These trends can be rationally explained by noting that: initial routing in design kept gas-transmission pipelines away from what were then the urban areas, and, urbanization is slowly encroaching on segments of pipeline right of way (RoW), which should lead to gradual increases first in Class 2 mileage, then Class 3 mileage, and eventually Class 4 mileage. •

•

Figure 1 indicates that population encroachment has resulted in increases in Class 2 and Class 3 mileage, but as yet population has not sufficiently concentrated around the pipeline RoW to lead to much Class 4 mileage. As a recent National Research Council study(4) notes, a steady increase in the proportion of mileage in the higher density population classes is inevitable. Because there is very little mileage in Class 4 locations, normalizing by mileage in these locations also could lead to an erroneous bias in the results. Accordingly, data for Classes 3 and 4 have been combined for subsequent analysis involving the pipeline mileage. Risk referenced to the data available in the OPS database and defined as the occurrence of an incident multiplied by the reported cost in the database is shown in Figure 2. This incident database categorizes incidents based on the four categories used on the OPS incident reporting form. Figure 2 indicates the significant risk associated with outside-force incidents occurs in areas of higher population density. This incident category comprises acts of God, which typically result in widespread damage involving multiple pipe joints, and acts of man, which involve mechanical damage to the pipeline. These results show that the risk of outside force incidents, which are dominated by mechanical damage due to third-party contact, is five times that of any other when evaluated in terms of monetary risk. When evaluated on a per-mile

Figure 1 Distribut ion in m ileage for com panies comprisi ng the ~50,000 miles of US mainline as-transmission i elines

7

 basis, the data show that the risk is greatest by far in Class 3 and 4 Locations, a situation that is not expected to change since urban areas continue to expand (4). This format orders threat significance from most to least important category as outside forces, material and construction defects, corrosion, and other. In contrast, threat significance ranked by frequency reverses the order of the corrosion and material and construction categories, and emphasizes Class 1  Locations. Based on Figure 2 and the incident data reported in Reference 3, mechanical damage and third-party contact comprise the greatest threat to natural-gas pipeline safety, including low-wall-stress systems. Consideration also has to be given to active material and construction defects and corrosion.

Figure 2 Monetary risk p er year, by inci dent category and class loc ation, normalized by mil eage Kiefner et al(6) have suggested an alternative to the four categories used in Figure 2 in terms of 22 root-cause categories. These 22 root causes could be grouped as: external forces and encroachments, which subdivides the OPS outside-forces category, several forms of environmentally driven processes that currently comprise the OPS corrosion category, welds and materials related defects, which would fall into the OPS material and or construction category, other, which remains a catchall category including what are termed miscellaneous and unknown, and, equipment and operations, which breaks down incidents that typically fell into the other category or material and construction categories on the current OPS form. • •

•

•

•

When the OPS incident data are re-categorized into the 22 root-cause categories suggested by Kiefner, root causes involved with external forces and encroachments remain the dominant risk, with a focus as anticipated in Class 3 and 4 locations. External and internal corrosion likewise show up as significant root causes, as do root-causes associated with operations and material and construction. The remaining root causes appear inconsequential relative to

8

external forces and encroachments, external and internal corrosion, operations, and material and construction. The net effect of the breakdown into smaller groups is to spread the apparent significance of one category across many root-causes, which occurs because there are 22 bins to categorize into as opposed to four. But, regardless of how the causes are evaluated, the conclusions reached above in reference to Figure 2 remain the same. That is, mechanical damage and third-party contact comprise the greatest threat to natural-gas pipeline safety, including low-wall-stress systems. Consideration also has to be given to deterioration by corrosion, as well as active material and construction defects. For low-wall-stress pipelines made from line pipe that was either mill-tested and/or cold expanded, deleterious material defects should have been culled from the pipeline. Likewise, for low-wall-stress pipelines subjected to regulatory-based pre-service hydrotesting, or subsequent re-testing, all remaining material defects and construction defects that threaten safety at much reduced allowable operating stress should have been removed prior to service . Accordingly, concern for safety hereafter focuses specifically on corrosion and postconstruction mechanical damage, for both immediate and delayed mechanical–damage incidents.

9

 Approach Whether a leak or rupture occurs has been quantified as a function of wall stress, for a range of  parameters including line pipe properties and defect type and size. Several approaches have  been pursued to establish the dependence of failure behavior on pipeline service, pipe geometry, and line pipe properties, to establish credibility of the results, which included: what is inferred from code or regulatory provisions world-wide, what the results of full-scale testing and field experience show, and, what mechanics analysis and numerical failure models predict. • • •

These approaches were implemented to evaluate the effects of: operating conditions and failure scenarios,  blunt metal-loss defects such as corrosion, and, sharp crack-like defects as occur in the wake of mechanical damage and/or grow in service due by fatigue, stress-corrosion cracking, etc. • • •

Where the approach evaluated leak versus rupture associated with mechanical damage and corrosion through analysis and model predictions, use was made of pipeline geometries and  properties characteristic of the US pipeline system. Archival data have been analyzed to establish trends as a function of line-pipe vintage and grade. Likewise, the OPS incident database has been evaluated to identify commonly occurring pipeline geometries and grades. On this basis, a set of “generic pipeline geometries” has been identified as the focus for this investigation. Results characterizing properties as a function of vintage and grade are presented first, followed by generic geometries and the pipeline’s service conditions.

10

Property Trends with Vintage and Grade Research over the past 30 years has led to significant advances that now lead to consistent commercial production of higher-strength steels that also have high fracture toughness and are weldable – the necessary traits of modern line pipe steels. These desirable properties derive  primarily from the use microalloying, controlled rolling, and controlled accelerated cooling to  produce fine grain size in the steel plate. This approach to develop higher strength grades of line-pipe steel is necessary because the traditional approach based on increasing carbon and other alloying additions led to problems with hydrogen-embrittlement and decreased toughness for grades beyond about X60. The economic push underlying the desire to use grades with strength beyond X60 began the evolution of modern line-pipe steel. This began in the 60s with the introduction of highstrength low-alloy steel (HSLA) making practices, to avoid the potential toughness and cracking problems that would be associated with the traditional strengthening mechanisms in applications to higher-strength grades. The HSLA practices of the 60s transitioned through a range of thermal-mechanical processing (TMP), including heavy rolling practices, beginning in the 70s. Continued improvement in dynamic-ductile fracture resistance led to further evolution, beginning in the 80s, which culminated in today’s thermal-mechanical controlled  processed (TMCP) line-pipe steels. This class of steels makes use of a range of finishing  practices, including accelerated cooling and controlled rolling. It is noteworthy that steel- and  pipe-making practices are not standardized, and it is still possible to purchase line pipe today made with steel processed as it was up through the 60s. A companion report indicates the line pipe properties of concern in assessing whether a leak or rupture occurs are those that characterize the two common limit states for pipeline failure – (7)  plastic collapse and fracture . Plastic collapse is characterized by the mechanical properties of the line pipe. The simplest way to establish the mechanical property of function of properties controlling plastic collapse is to correlate failure of defect-free line pipe with the related mechanical properties. Reference 8 assembled such data and discussed a range of stressing conditions including stress  biaxiality. Figure 3 derived from those data shows that the ultimate tensile strength (UTS), a  parameter that historically has been measured within the pipeline industry, correlates very well with plastic-collapse controlled failures in line pipe. The UTS is the largest stress measured in a standard tension test of the line-pipe steel. A similar conclusion has recently been drawn by the European Pipeline Research Group (EPRG). It follows that approaches accounting for the effect of defects on pipeline fitness for service should use the UTS as the bounding value for defect-free pipelines, as opposed to other properties such as the actual or specified-minimum yield stress. In addition to the UTS, it is appropriate to determine the yield stress of the steel, as this value determines that the line pipe remains elastic in service. It likewise is useful to characterize the variation in the ratio of yield stress to the UTS, denoted as Y/T, as this also is an important  parameter in satisfying requirements in Part 192. Figure 4 presents the trend in Y/T as a

11

Figure 3 Illustrating t he viability of UTS as a failur e crit erion for plastic c ollapse in steel pipelines

Figure 4 Y/T as a function of actual yield stress 12

function of grade for a wide variety of steels. This trend can be used to estimate the value of UTS for a given grade of steel, in cases where the UTS has not been determined. Fracture is characterized in terms of fracture toughness, measured a ccording to established standards whose roots lie in a technology known as fracture mechanics. The development of this technology is quite recent compared to the origin of the pipeline industry. Thus, concern for fracture resistance developed in the pipeline industry years before the birth of modern fracture theories. For this reason, the pipeline industry adopted the then available Charpy-vee notch (CVN) impact test as their measure of apparent toughness. The energy measured in the CVN test remains the industry’s standard measure of fracture resistance today. Use of the modern fracture theories in pipeline applications has been facilitated through correlations  between CVN energy and parameters valid within fracture-mechanics theories. Such correlations, which have been developed and demonstrated valid for a variety of line-pipe steels, can be found in the literature(e.g., see 9). As may be expected from the discussion on the evolution of steels, archival Battelle data indicate that apparent toughness characterized by CVN energy changed very little until the early 70s. This is evident in Figure 5a, which presents results representing in excess of 600  joints of line pipe reflecting steel produced from about the 30s through the 80s. It is clear from Figure 5a that toughness began to increase significantly, although somewhat inconsistently starting in the early 70s. Figure 5b, based on archival data for more than 300 pipe joints dominated by more recently developed steels, indicates a weak dependence of apparent-toughness on grade. However, this trend reflects a casual relationship, as changes that affected improvements in SMYS went hand in hand with changes that positively effected toughness. Because line-pipe steels have evolved as discussed earlier, there can be significant differences in the values of the properties of that control whether a leak or rupture occurs. For this reason, it is necessary to evaluate leak versus rupture under conditions that represent the line-pipe used in the low-wall-stress pipeline system. The trends in Figures 3 through 5, coupled with changes in steel chemistry, finishing, and rolling practices that affected changes in strengthening mechanisms and cleanliness can be identified by decade, to define combinations of grade and toughness typical of pipelines constructed in that era. Table 1 summarizes such combinations, which are considered representative of the line pipe in the ground in the US.

Table 1 Summary of property combin ations by decade Decade

50s

60s

70s

80s

90s

Grade CVN, ft-lb

X42/X52 20

X52/X60 30

X60/X65 40

X65/X70 65

X75 80

Table 1 suggests that Grade X52 line-pipe be coupled with a typical plateau toughness of 30 ftlb, to represent much of the pipeline-miles constructed through the late 60s. This combination

13

a) as a funct ion of year produ ced

b) weak apparent tie between toughness and yield stress

Figure 5 Some trends in CVN plateau energy

14

could be paired with Grade X42 coupled with a typical plateau toughness of 20 ft-lb to reasonably represent much of the construction through the late 50s. Consider now pipeline geometries to tie to these properties, to define generic pipelines, which then will be evaluated to quantify leak versus rupture behavior.

15

Typical Pipeline Geometries As this reports focuses on safety considerations, the pipeline geometries considered are biased toward circumstances that have posed problems in past. Accordingly, pipeline geometries considered representative have been identified with reference to the OPS incident database, which trends incidents as a function of many parameters, including pipe diameter and wall thickness. Typical geometries have been identified that reflect much different values of pipe diameter to wall thickness, denoted hereafter as D/t. The OPS database and results archived at Battelle for non-reportable incidents occurring between 1970 and 1984, when the reporting requirements changed, have been used for this purpose. These results suggest the use of a 30-inch diameter  pipeline with 0.281 inch-thick wall to represent a high D/t case (D/t = 106) and a 16-inch diameter pipeline with 0.250 inch-thick wall to represent the a low D/t case (D/t = 64). Results in the literature(e.g., 10) indicate that diameter by itself has a second-order effect on critical flaw length, all else being equal Likewise, these results indicate that wall thickness has a second-order effect on the critical-flaw length, all else being equal. It follows that results for the two geometries selected are suffice to address the range of line-pipe geometries in service today.

16

Pipeline Servic e Conditions  Natural-gas transmission pipelines in the US have typically operated at near-capacity, experiencing relatively small and infrequent pressure cycles. Because of their regulatory-based minimum burial depth of 36 inches, thermal cycling likewise is neither frequent nor severe. In contrast, lines operated at lower stresses tend to experience more frequent pressure cycles, although their magnitude may be smaller because MAOP is much less. It follows that possible in-service growth due to pressure cycling should be addressed. To this end, daily demandinduced cycling from a maximum operating pressure (MOP), taken as 30 percent of SMYS, to 70 percent of MOP has been evaluated. Given the focus on low-wall-stress pipelines, and the typical service of lower-stress lines as links between higher pressure sources and city gates, there is limited need for added compression. Accordingly, the average gas temperature along a typical natural-gas transmission line can be adopted as representative, which for present purposes is taken as 60 F. It is somewhat more difficult to represent the typical coating type and condition, and conditions along the RoW. Experience indicates that soil and moisture conditions can vary along a RoW for a given pipeline, which effectively precludes identifying a typical state. Coating condition likewise can vary significantly along a pipeline, due to differences in soil stresses affected by the variation in soil and moisture, which again precludes identifying typical circumstances for  purposes of this report. Fortunately, these difficulties can be circumvented through use of degradation rates that conservatively represent corrosion, stress corrosion, or other such mechanisms. Kinetics that reflect in-service failures4 , or were derived from accelerated testing that reflects worst-case field conditions, have been used for present purposes, as discussed when they are utilized .

4

Early in-service failures reflect the combination of worst-case circumstances associated with the operating environment, service conditions, and pipeline material, construction, and coating quality. Kinetics under such circumstances tend toward the upper-bound of what occurs elsewhere on the same and other similar systems.

17

Leak versus Rupture Condit ions for the Two Typi cal Gas Pipelines Whether or not a leak or rupture occurs can be determined using Battelle’s model for pipeline fracture initiation and instability(11). This model has been extensively validated by close comparison of predictions of the failure pressure for a wide range of full-scale and laboratory testing(e.g., 12,13), as well as accurate, blind predictions of actual field incidents and hydrotest failures(e.g., see 14, p135-9). It has also been proven valid in its ability to predict flaw growth  behavior under such conditions(e.g., 15). Whether leak or rupture occurs has been determined for the two “typical” combinations of  pipeline geometry and property identified above, for operation at 30 percent SMYS. The results generated using Battelle’s model are shown in Figure 6 on coordinates of normalized failure pressure on the y-axis, and critical flaw depth on the x-axis. As the format of the charts in Figure 6 is unusual, some discussion of their content is appropriate before continuing. Referring to Figure 6, the red and black contours on this chart represent fracture initiation boundaries as a function of flaw depth normalized (i.e., divided)  by the wall thickness. Thus, a normalized flaw depth equal to 0.9 represents a flaw that is 90  percent through the wall. The contours on this chart start at a normalized depth of 0.9, which is represented by the lowest trend. Each trend above that for 90 percent represents another step reduction of depth equal to one-tenth of the wall thickness. Counting these red and black contours indicates that Figure 6a has ten in total, which thus represent depths from 90 percent through 10 percent of the wall thickness. In contrast, Figure 6a has only nine, which thus represent depths from 90 percent through 20 percent of the wall thickness. With the coordinate system used in Figure 6, a value of 0.72 on the y-axis corresponds to a line  pressure causing a wall stress equal to 72 percent of SMYS. Likewise, a value of 0.60 corresponds to Class 2, while values of 0.5 and 0.4 respectively reflect classes 3 and 4. Flaw sizes causing failure at any one of these values are found by tracing a horizontal line across the chart. Where that line intersects each of the depth contours defines the depth for failure at that normalized pressure. The corresponding critical length is found by dropping a vertical line from that intersection to the y-axis. Results of the calculations presented in Figure 6 represent two different pipelines. Figure 6a represents the situation for the earlier-introduced “typical” 16-inch diameter pipeline (0.250inch thick wall made of X42 grade line pipe with a 20 ft-lb CVN energy). Figure 6b shows the corresponding results for the “typical” 30-inch diameter pipeline (0.28 1-inch thick wall X52 grade line pipe with a 30 ft-lb CVN energy). The solid blue line in these figures corresponds to fracture propagation or crack instability. Thus, this line is the boundary between leak and rupture, as a function of normalized pressure (wall stress) and defect length when the wall is  breached. Combinations of defect length and depth that lie above this boundary fail as ruptures, whereas those below it fail as leaks. The green trend signals the onset of stable tearing, which eventually culminates in rupture if the loading is sustained at the corresponding wall-stress (equally pressure).

18

a) the 16-inch diameter pipeline

b) the 30-inch diameter pip eline Figure 6 Conditions for fracture init iation and crack instabil ity 19

The results in Figure 6a indicate instability in the “typical” 16 inch diameter pipeline occurs at flaw lengths greater than ~9 inches. Figure 6b indicates that the “typical” 30-inch diameter  pipeline the critical length increases to ~13 inches. Such flaws are quite long as compared to most sources of sharp defects, such as mechanical damage. The corresponding depth when these flaws become critical is in excess of 80 percent of the wall thickness. Combinations of this size are unusual in terms of metal-loss corrosion, which tends not to develop a consistent depth over the area of the corrosion patch. Very few circumstances create nearly through-wall defects with these rather large lengths. This indicates that operation at 30 percent SMYS is unlikely to initiate rupture, unless this contact causes deep axial gouging along the length of the pipe, over quite significant lengths. Note that the increased critical defect length between the two cases represented in Figure 6 is largely due to the increased toughness associated earlier in this report with the X52 steel as compared to that ascribed to the X42 steel. The results presented in Figure 6 indicate that 30 percent SMYS cannot be viewed as a threshold below which fracture initiation or propagation would not occur in the presence of sharp crack-like defects. However, it is clear that failure leading to a rupture appears very unlikely, because it requires a very long axially oriented defect (~ 9 to 13 inches), which also is 5 very deep (~80 percent of the wall) over its full length .  Lower wall stresses lead to the transition from leak to rupture at longer and deeper features. For example, for the 16 inchdiameter pipeline operating at 20 percent of SMYS, the transition from leak to rupture occurs for an axially oriented defect increases from a length of 9 inches to a length of 16 inches, and requires a depth in excess of 80 percent of the wall thickness over its full length. For the same  pipeline operating at 10 percent of SMYS, the transition from leak to rupture requires an axially oriented defect that is many feet in length, with a depth in excess of 90 percent of the wall thickness over its full length. For operation at a little less than 10 percent of SMYS, all features leak, regardless of their length or depth. Higher toughness line pipe leads to an increase in these threshold lengths and depths. In the unlikely event that rupture initiates, propagate versus arrest calculations have been made for the same circumstances to assess the extent of the rupture. These calculations were made using Battelle’s model for dynamic-ductile fracture, which was first developed in the 70s(16) and recently extended (17,18). This model also has been extensively validated by full-scale testing(e.g., 19), with its recent extension validated by successful blind predictions for full-scale running fracture experiments(e.g., 2). Calculations made with the Battelle dynamic-ductile fracture model predict the required arrest toughness, which by comparison with the toughness available from the line-pipe steel determines if arrest occurs following initiation. These results indicate the arrest toughness is about 6 ft-lb for the two typical pipeline geometries when operating at a wall stress of 30  percent SMYS. The required toughness is less at lower wall stresses. With reference to Figure 5a, this required arrest toughness is quite low in comparison to the historical plateau values. This indicates that for such steels the  fracture will arrest very quickly in the event that fracture initiation occurs. 5

These sizes don’t imply that all such features will fail – as they reflect a specific combination of wall stress and pipe properties, most importantly toughness. Lower operating pressure or higher toughness cause the transition form leak to rupture to shift to significantly longer and still deeper flaws.

20

Calculations are made later in this report to evaluate the likelihood of fracture initiation as a function of operating conditions.

21

Results and Discuss ion Code and Regulatory Implic ations Several natural-gas pipeline codes and regulations used worldwide were evaluated to identify  provisions that might distinguish between likely-leak versus rupture as a function of wall stress. Documents reviewed included US ASME B31.8-95, Title 49 CFR Part 192–98 US, Z662-96 Canada, AS 2885-97 Australia, HSE 825-96-UK, IGE TD/1-93 UK, ISO CD 1362396 World, EN 1594-Europe, NEN 3650-91 Netherlands, SNIP Russia, Din 2413-93-German, Algerian-91, French-70, and Norway-86. Many of these documents are patterned conceptually after the ASME B31.8 code. Neither those like the B31.8 code nor the others contain explicit stress-related leak versus rupture provisions, although most reflect less stringent provisions for lower-stress pipelines. It follows that pipeline design codes and regulations offer little to define a stress-based threshold below which failure will always result in a leak .

Corrosion and Metal Loss Corrosion and metal loss failures include internal as well as external corrosion. Such defects can be due to dissolution, the usual degradation process leading to corrosion. These features also can be due to other mechanisms, such as microbiologically influenced corrosion (MIC). Experience through full-scale testing and mechanics analysis indicate that the failure behavior of external metal loss is not too different from that for internal metal loss, all else being equal. This is because the failure behavior of corrosion and metal-loss features is controlled primarily  by the defect and line-pipe geometry, and the properties of the line pipe. Because the mechanism producing the metal loss is not a first-order factor, nor is its location inside or outside of the pipe, the transition from failure by leak versus rupture occurs is evaluated independent of these factors. The threshold wall stress for the transition from leak to rupture is evaluated for corrosion and metal loss in terms of available failure data and mechanics analysis. Available data generated in full-scale testing are considered first. Thereafter, the transition is evaluated in terms of the OPS incident database. Finally, mechanics analysis is used to determine the threshold wall stress for this transition.

Full-Scale Static Testing There is a large database of full-scale test results developed for metal loss and corrosion, which was generated initially to calibrate B31G(20). This database has expanded since – to first calibrate modified B31G(21), and then RSTRENG(22), and thereafter to validate other more recent, general metal-loss failure criteria(e.g., 8, 23). This database has been interrogated to identify a threshold wall stress for the transition from leak to rupture, or failure, for metal loss and corrosion defects. The available full-scale database represents primarily pressure-to-failure experiments, for pipes containing defects whose maximum depth exceeded 90 percent of the wall thickness, with lengths that ranged to more than 100 inches. These defects represent actual corrosion features as well as machined metal-loss features, which were evaluated in well over 100 experiments, in line pipe from Grade B through X65. For these data, the lowest failure stress was 35 percent

22

of SMYS. Failures were said to occur by plastic collapse, and involved both leaks and ruptures – but none failing below 35 percent of SMYS. Based on this database it can be stated that the threshold for failure of long and deep external corrosion and metal-loss features is 35 percent of SMYS. This is so for several grades of line  pipe, including earlier-vintage grades. It follows that the rupture threshold for corrosion is 35  percent of SMYS for these data. Because the failure behavior of internal corrosion is not too different from external corrosion, it likewise can be concluded that a similar threshold applies  for internal corrosion. Finally, because the failure behavior of corrosion and metal-loss features depends on the geometry of the features and not the mechanism producing the feature, this threshold applies for the usual dissolution-based corrosion, as well as other mechanisms such as microbiologically influenced corrosion.

Repor table In-Service Inci dents The OPS incident database for corrosion failures contains information about pipe geometry, SMYS, failure pressure, and an indication of leak versus rupture. This database also reports many other parameters that reflect the nature of the failure, such as the length of the rupture, and the consequences of the failure, such as the cost of the incident, the duration of uncontrolled release. Failure stress has been calculated from the reported failure pressure and  pipe geometry and compared to SMYS. Then the results were sorted on leak versus rupture. It became evident during the calculation of the wall stress at failure and the sorting process that there are gaps in the data reported. More importantly, it is clear that there are embedded errors in the data, as cases where the diameter is less than the wall thickness can be found, as well as situations where SMYS is wrong, or the calculated stress is an order of magnitude greater than the SMYS for any available steel grade. There apparently also is confusion regarding the technical definitions of and resulting differences between a leak and a rupture. The erroneous data entries coupled with the confusion in the technical definition of leak versus rupture lead to some incidents being reported as “ruptures” at quite low levels of SMYS. Incidents labeled as ruptures at low values of SMYS are tied to a small cost-consequence, a very short to zero-length rupture, or a short time to control the release, all of which are inconsistent with the technical definition of a rupture discussed earlier in this report. In spite of this inconsistency, the data have been evaluated to identify circumstances that underlie leaks as compared to ruptures. This has been done using incident cost as a filter for the just listed, technically-based traits of ruptures, excluding cases where cost of the incident was <$10,000. With this provision to identify events more consistent with technical definition of a rupture, sorting of the incident database points to a threshold of 35 percent SMYS for the transition  from leak to rupture.

23

Mechanics Prediction Analysis using the same extensively validated Battelle model for fracture initiation and instability(11) introduced earlier has been used in conjunction with a “strength of materials” type analysis to predict failure pressure associated with rupture of corrosion an d metal-loss defects. This analysis considered rupture due to a plastic-collapse “limit state”. However, the length and the depth when the defect becomes critical have been determined under the very conservative assumption that these corrosion and metal-loss features behave as sharp cracks, as follows. With reference to Figure 6, the maximum depth avoiding rupture is conservatively found to be ~0.75 for operation corresponding to 30 percent of SMYS. Thus, plastic collapse occurs through a remaining wall thickness, t, equal to 25 percent of the initial (design) value of the wall thickness. Thus, the thickness at the limit state, PLIMIT, is 25 percent of the initial thickness, tINITIAL. Using this boundary condition, the pressure below which ruptures no longer occur is evaluated, as follows. From Barlow’s equation the wall hoop stress, denoted , is equal to pressure, P, multiplied by the pipe radius, R, divided by the wall thickness, t. On that basis, the limiting pressure, PLIMIT, is given by PLIMIT = (UTS x tLIMIT )/ R , while PSMYS = (SMYS x tINITIAL )/ R, with tLIMIT 0.25 tINITIAL for the transition from leak to rupture. Based on the trends shown earlier in Figure 4, one can conservatively relate SMYS and UTS as 1.2 x SMYS ~ UTS for most grades so that P LIMIT / PSMYS = 1.2 x 0.25. Substituting into the above leads to 30 percent of SMYS as the threshold for the onset of rupture. Thus, this conservative analysis leads to a threshold stress of 30 percent of SMYS for corrosion. A probabilistic formulation would lead to a potentially much greater threshold stress. It follows that this conservative worst-case analysis leads to a threshold for rupture equal to 30 percent of SMYS , which is of the same magnitude indicated by analysis of the OPS corrosion-related incident data, which reflect the lower tail of a field population and so too reflect worst-case situations. In contrast, the result noted earlier in terms of full-scale testing was about double these values.

Stress-Corrosi on Cracking and Fatigu e Stress-corrosion cracking (SCC) and fatigue both could lead to failure at defects in pipelines  for certain combinations of right-of-way conditions and se rvice loadings. As both mechanisms  potentially lead to sharp cracks that promote ruptures rather than leaks, this section evaluates their possible influence on the transition threshold wall stress for service conditions that occur in low-wall-stress pipelines. Fatigue is considered first, then SCC. The literature indicates that the fatigue mechanism exhibits a threshold stress level below which the kinetics slow to the point they no longer are a practically significant concern(e.g., see 7). This is apparent in the endurance- or fatigue-limit found for constant-amplitude cycling of non pre-cracked specimens. It is likewise evident in the threshold for crack propagation apparent for pre-cracked specimens subjected to constant-amplitude cycling. From a pipeline  perspective, the threshold for fatigue is apparent in the very long lives of pre-flawed vessels cycled under conditions relevant to high-stress pipelines(e.g., see 24). It is also evident in  parametric sensitivity studies for high-stress pipelines(e.g., see 25). As cycling leading to fatigue failures under high-stress pipeline service conditions is relatively more severe than that for

24

low-wall-stress pipelines, the results of those studies are transferable conservatively to lowwall-stress pipelines. This transferability is validated by the absence of fatigue as the root cause of failures in gas pipelines, save for cases involving high-wall-stress transmission  pipelines where cracking had initiated along the long-seam weld prior to commissioning as a result of cyclic loading caused by improper rail shipment. The literature likewise indicates that SCC exhibits a threshold-stress below which the kinetics slow to rate so low it is no longer practically significant. Such thresholds are clearly defined for the high-pH SCC environment(e.g., see 26). Comparable thresholds can be anticipated for the low-pH SCC environment, as laboratory testing with this environment indicates that cracking does not nucleate or grow under similar loading conditions(27). The stress level in low-wallstress pipelines falls below thresholds known for pipeline steels and high-pH SCC environments(e.g., 26), indicating that such cracking is highly unlikely on these pipelines. The kinetics of the high-pH SCC known to be active on US pipelines also slow significantly as temperature decreases(e.g., 26). Thus, the already benign low-wall-stress pipeline situation  becomes even less a concern, as such pipelines run at relatively lower temperatures absent the significant work of compression needed in cross-country transmission pipelines to maintain high-volume flow.  It follows that neither fatigue nor SCC is a practical concern for low-wall-stress pipelines, so the effect of these mechanisms on the transition from leak to rupture is considered no further.

Mechanical Damage Mechanical damage defects include smooth dents, gouges, and dents in gouges, which can fail immediately when the damage is inflicted, or during re-rounding as the damaging implement moves away from the pipeline. If immediate failure does not occur, these defects might fail due to the effects of in-service loading, a situation termed a delayed failure. This section is concerned with leak versus rupture for both immediate and delayed mechanical–damage incidents. As with corrosion, this section considers full-scale data, incident data, and mechanics analysis to assess the transition from leak to rupture a s a function of wall stress.

Full-Scale Static Testing Figure 7a presents data for full-scale testing of simulated mechanical damage defects for  pressure to failure experiments reported by the EPRG(28). These data include leaks and ruptures for testing done for a wide range of pipe grades and toughness values, which are  biased toward steels used in early pipeline construction. Damage severity for smooth (plain) dents is measured along the x-axis defects in terms of dent depth normalized by pipe diameter. Damage severity for gouges is measured along the y-axis in terms gouge depth normalized by  pipe-wall thickness. Gouges in dents are represented by data that lay off either axis. Figure 7a groups the failure data expressed in terms of wall stress, as a fraction of SMYS. Data are presented for testing grouped in one of three wall-stress levels: at or below 30 percent of SMYS, between 30 to 50 percent of SMYS, and between 50 and 72 percent of SMYS. The figure also includes contours that represent a lower bound on the damage size corresponding to each of the three failure-stress groups. As expected, the figure shows that higher wall stress is required to fail less severe damage, which is evident in the relative

25

a) EPRG data – report does not dis tingu ish leak versus ruptu re

b) Battelle and Canmet data – stable means leak, unstable means rupture

Figure 7 Full-scale test results fo r simu lated mechanical-damage defects 26

frequency of tests causing failure for each of the three wall-stress groups. It is apparent from Figure 7a that dents as shallow as 2 percent of the diameter combined with gouges ~10 percent of the wall thickness fail between 30 percent and 50 percent of SMYS. These same data show that dents as shallow as 3.5 percent of the diameter combined with gouges ~10 percent of the wall thickness fail at pressures causing wall stress less than 30 percent of SMYS. These results indicate that rather small features fail for operation at or above 30 percent of SMYS. As was evident in regard to Figure 6, whether these data represent a leak or rupture depends on the length of the damage when the wall is breached . While the results presented in Figure 7a were not reported in a format that identified leak versus rupture, a significant portion of the full-scale experiments involving simulated mechanical damage does include this parameter. Such results are presented in Figure 7b, for testing done at Canmet and Battelle6. The format of Figure 7b is like that of Figure 7a, except for nominal differences in the scale of the axes. These results represent steels from grade X52 to X70 with full-size equivalent (FSE) CVN energies from 15 to 68 ft-lb. Dent lengths ranged up to 90 inches, with gouge lengths up to about 25 inches. As with Figure 7a, Figure 7b groups the failure pressure data expressed in terms of wall stress, as a fraction of SMYS. Data are presented for testing grouped in one of three wall-stress levels: at or below 30 percent of SMYS, between 30 and 72 percent of SMYS, and above 72 percent of SMYS. Consistent with expectations and Figure 7a, Figure 7b shows that a higher wall stress is required to fail less severe damage, which is evident in the relative frequency of tests causing failure for each of the three wall-stress groups. While quite shallow dents and gouges are again found to cause failure, Figure 7b shows that leaks occur for all testing done at a wall stress less than 30 percent of SMYS . More importantly, even quite large damage can be seen to survive stresss in excess of 72 percent of SMYS. This could be taken to mean that the trends shown in Figure 6 are quite conservative. Equally, it could be taken to mean that the larger defects in this database have been tested in steels more resistant to fracture initiation than those evaluated using the lower-bound properties considered in Figure 6. This means that the distribution of both line-pipe properties and service pressures can significantly affect the incidence of mechanical damage failures, and the threshold for leak versus rupture in the low-wall-stress  pipeline system.

Repor table In-Service Inci dents The process used to evaluate the OPS database for corrosion was repeated for incidents categorized as due to outside forces. As might be expected, the tabulated data for outsideforces incidents was prone to the same gaps and embedded errors noted in regard to analysis of the corrosion incidents. As noted there, cases where the diameter is less than the wall thickness can be found, as can situations where SMYS is clearly wrong. Likewise, cases can  be found where the stress at failure calculated from the database represents an order of magnitude greater than SMYS for any available line-pipe steel grade. There also was evidence of confusion regarding the technical definitions of and resulting differences between a leak and a rupture. For this reason, the database was filtered as it was for the corrosion data, excluding cases where cost of the incident was <$10,000 to reflect the technical inconsistencies such as a 6

Appendix E of Reference 29 evaluates the available full-scale database for simulated mechanical damage, and includes tabulations of the results presented in Figure 7b.

27

reported short or zero-length rupture, or a very short time to control the release. With this  provision to identify events consistent with technical definition of a rupture, the incident database points to a threshold of 25 percent SMYS for the transition from leak to rupture . The longest rupture evident in the OPS database for low-wall-stress pipelines (pressure less than that causing a wall stress 30 percent of SMYS) is on the order of 10 feet, with most less than 5 feet, indicating the rupture length was short and contained in all cases. Field failures archived at Battelle support the observations noted above, except for one rupture at about 26 percent of SMYS. Two keys to the occurrence of this rupture on a low-wall-stress  pipeline lie in the very significant length and axial orientation of the mechanical damage.  Damage that couples these two unique characteristics reflects unusual circumstances in comparison to the mechanical damage triggering immediate failures, which typically involve  puncture of the pipe wall. Puncture typically is due to contact normal to the pipe wall, which is inflicted typically by the tooth of an excavator. As the term puncture suggests, the tooth or other damage implement penetrates the pipe wall. Because the motion is into rather than along the wall, the puncture often has planar dimensions that mirror the damage implement, such as an excavator tooth. These teeth have planar dimensions that typically are smaller than the critical lengths calculated in regard to Figure 6. For this reason, immediate failure due to  puncture often is stable, that is does not rupture, even in relatively low toughness line pipe. In contrast to penetration, the damage leading to the above-cited archival failure ran along the  pipe, due to a scraping contact. While the length of such contact is theoretically controlled by the reach of the backhoe, or unlimited in regard to contact by a bulldozer blade, axial contact to a cylinder is inherently unstable as the normal force causes the contact to “ride around” the 7 8  pipe. This instability in axial tracking is evident in Figure 8 , which shows the scar created in the thin-film epoxy coating during scraping contact to generate acoustic signatures as part of a  project developing the capability to detect third-party contact in real time(30). This scrape remained axially stable for about 8 inches, prior to riding around this 24-inch diameter  pipeline. This type of instability occurred consistently, with its occurrence increasing as the normal-contact force increased. It follows that severe damage that is axial and traverses a significant distance along the pipe is unlikely based on our experience through many attempts to generate axial scraping contact . This work indicated that such contact is more prone to slide obliquely around the pipe. It also indicated that contact with a single corner on one tooth was required to inflict significant mechanical damage due to scraping contact. It follows that, while mechanical damage has caused incidents on low-wall-stress pipelines, the required long axial contact for a rupture of such damage is likely not common. Battelle’s archival data and the OPS database indicate that there is no evidence of ruptures with consequential effects at pressures developing a wall stress less than 25 percent of SMYS . Continuing work is evaluating the likelihood of incidents due to mechanical damage as a function of operating conditions, line pipe properties, and the nature of sliding contact along a  pipeline. 7

8

An alternative case involves adjacent teeth on a backhoe riding along the crown of the pipe, which is inherently stable. The use of teeth shields with rounded ends would ensure the unstable situation, while the rounded ends would limit damage of this nature. Preventing contact is even better. Figure 8 is an expanded view of part d of Figure C4 in Appendix C of Reference 29

28

9:00

 Approximate Clock Position Scraping direction

Point of Instability

12:00

Figure 8 Illustrating the unstable nature of axial scraping along the length o f a pipeline -- Scrape 5.01

Mechanics and Fractur e Predicti ons Mechanical damage occurs as smooth (plain) dents, gouges, and dents in gouges. Failure at a  plain smooth dent is virtually impossible – unless the depth is very large compared to either the affected length or circumference. In contrast, gouges or dents with gouges can fail immediately when the damage is inflicted, or during re-rounding as the damaging implement moves away from the pipeline. If immediate failure does not occur at these gouge or gouge-indent defects, they might fail due to the effects of in-service loading, a situation termed a delayed failure. This section first discusses the current formulation of a model of leak versus rupture for immediate failures, and then addresses delayed mechanical–damage incidents. Analysis indicates that failure at mechanical damage can occur immediately, or be delayed in time. Immediate failures can occur by plastic collapse, or be toughness controlled, whereas delayed failures will be toughness controlled (29). Because failure occurs at a lower stress when controlled by toughness as compared to plastic collapse, it is necessary to evaluate the  possibility of leak versus rupture at mechanical damage under toughness-controlled conditions. This is particularly the case when dealing with a toughness population such as that indicated in Figure 5. For this reason, the continuing work is formulating a model of the leak versus rupture threshold consistent with the approach and technology developed in Reference 29. The deterministic formulation discussed in Reference 29 is being embedded in a numerical  probabilistic evaluation of the likelihood of failure. To this end, the model underlying Figure 6 is being coupled with distributions of operating pressure to generate probability-density functions (PDFs) of critical flaw size as a function of mean line-pipe toughness, for the distribution of pipe sizes and grades representative of the low-wall-stress pipeline system. The effect of material properties will be used to assess their role in the likelihood for failure. Consideration also will be given to the distribution of potential damage sizes developed from the distribution of stable axial scaring lengths based on the results reported in Reference 29,

29

coupled with the distribution of excavator equipment sizes reported in the literature(31). As this work is just now beginning, results are not anticipated for some time.

Cyclic-Stable Tearing and Failure due to Delayed Mechanical Damage The several circumstances that are apparently necessary for mechanical damage to lead to a delayed rupture indicate that the transition from leak to rupture cannot be simply characterized in terms of a constant value of MOP. Rather, the chance of rupture depends on the MOP for the pipeline, the likelihood of immediate failure, and other equally uncertain factors, such as cyclic variation of MOP involving large swings approaching MOP. The mechanism for inservice growth of cracking that forms at mechanical damage defects is cyclic-stable tearing(29) (CST). Cyclic stable tearing can occur in spite of the fact that such cycling in low-wall-stress  pipelines does not promote fatigue. This is because CST depends on time at stress, whereas fatigue depends on the frequency and amplitude of the pressure cycles. Like both fatigue and SCC, the CST mechanism exhibits a threshold, which is expressed in terms of nonlinear fracture mechanics(29), that can be correlated with CVN energy(7, 9). CST is only active at crack-like defects, so this mechanism will not lead to crack growth in any pipeline absent a crack-like defect. The threshold for CST is typically well above that for fatigue crack growth, except in very lowtoughness steels, and in situations where the ductility is locally reduce d to a very low level due, for example, to cold work (29). Thus, where both crack-like defects are present and the toughness local to the defect is low, CST can promote in-service growth. Mechanical damage that involves sliding contact can affect significant cold work, and so can lead to locally reduced toughness. Where cold work due to sliding contact runs axially a distance greater than the critical length, CST can cause delayed failure at mechanical damage. Such situations require the low-likelihood occurrence of sliding contact over an axial distance greater than the critical length also be subjected to pressure cycling to a maximum stress that exceeds the local threshold for CST for that specific line-pipe steel. It follows that the archival failure at 26  percent of SMYS discussed above in regard to in-service incidents represents a very unlikely combination of circumstances. For that X42 pipeline cracking initiated occurred because of sliding mechanical damage that created an axial gouge whose length was in excess of 14 inches. In-service growth nucleated at this gouge and grew due to CST, due to infrequent  pressure cycles whose peak pressure approached MOP. Cyclic stable tearing driven by occasional pressure cycles whose peak pressure approaches MOP gradually deepens and lengthens cracks nucleated when the pipe re-rounds in the wake of the damage implement. Reference 29 indicates that if cracking does not initiate during contact, or re-rounding in the wake of contact, subsequent crack initiation (or growth) from the damage is unlikely by a fatigue or CST mechanism for the frequency and magnitude of pressure cycles typically experienced on cross-country gas-transmission pipelines9. Accordingly, the presence or absence of cracks immediately following outside-forces contact is a critical factor in the 9

This conclusion, based on historically typical operation, could change if gas pipelines experience large demand-induced pressure swings, as would occur if a pipeline was the prime energy source for distributed electrical-power generation operating in peak-shaving service.

30

subsequent fitness-for-service of the resultant mechanical damage. As evident in Figure 6, rupture requires long, axial damage that is consistently quite deep along its length. Damage that is axially long, and consistently deep, is unusual in comparison to the mechanical damage triggering immediate failures, because axial contact, due for example to backhoes, is  prone to slide obliquely around the pipe10. It follows that while sliding mechanical contact has caused damage leading to rupture on low-wall-stress pipelines, it is infrequent in practice and indicated analytically to be very unlikely. Moreover, the OPS database, as well as analysis done herein, indicates that such ruptures arrest quickly because of the low wall-stress, which limits their consequences. On this basis, delayed failure due to mechanical damage can be viewed as highly unlikely for operation of the low-wall-stress pipeline system. However, care must be taken to avoid contact with pipelines to ensure this remains the case. Likewise, where  plausible care should be taken to limit the frequency of large-amplitude pressure swings, as well as keep the peak pressure in such swings from exceeding historical values of MOP. Control of potential mechanical-damage incidents by avoiding immediate failure and limiting exposure to delayed failure of mechanical damage may be better achieved by measures other than operational controls. Avoiding immediate failures requires avoiding potential contact, whereas the potential for future delayed failures can be achieved by detecting contact when and where it occurs. Technology improvements on both fronts offer near-term potential to manage the extent of future sources for mechanical damage incidents(30, 32). In regard to control of delayed failures, direct-assessment technologies(33) that detect the effects of the coating damage done during mechanical damage offer one means of locating potential sites of damage, as do ILI tools under development to detect mechanical damage(34) for portions of the limited portion of the low-wall-stress pipeline system that are piggable, given that pigs will be developed for the diameters of concern.

10

Exceptions include contact due bulldozer and grader blades, ripper teeth, and some forms of ditchers. While dozer and grader contact are unexpected for pipelines buried at code depth, rippers and ditchers reach such depths.

31

Summary and Conclu sion s Leak versus rupture was evaluated as a function of stress with a focus on operations and maintenance issues in low-wall-stress natural-gas pipelines operating through highconsequence areas. Following some technical definitions and discussion of leak, rupture, and fracture arrest, the work scope was presented, as was the approach. Data needed to quantify leak versus rupture as a function of stress level are presented in terms of line-pipe properties as a function of  pipeline vintage, geometry, and operation. Thereafter, the threshold stress for the transition from leak to rupture was evaluated primarily with reference to results of full-scale testing and field failure experience, and what mechanics analysis and numerical failure models predict. While the process used to evaluate the stress for the transition from leak to rupture is comparable for prescriptive- and performance-based integrity management p lans, the results in this report only address prescriptive plans. Accordingly, this transition is evaluated with a conservative basis made necessary to encompass the breadth of federally regulated low-wall stress pipelines. The lower-bound threshold for the transition from leak to rupture in the low-wall-stress  pipeline system was evaluated with reference to full-scale test data, incident data, and mechanics and fracture analysis. For corrosion, these necessarily conservative thresholds designed to reflect the breadth of conditions covered by federally regulated pipelines were: 35 percent of SMYS based on full-scale testing. 35 percent of SMYS for the OPS incident database. 30 percent of SMYS for mechanics modeling. Were other than worst-case circumstances used in the evaluation, higher values would be obtained. Given the results generated, the leak to rupture transition for corrosion defects in the low-wall-stress pipeline system can be taken as 30 percent of SMYS, a value that is conservative in comparison with in-service incidents. • • •

Thresholds for the transition from leak to rupture also were evaluated for immediate as well as delayed mechanical damage incidents with reference to full-scale test data, incident data, and mechanics and fracture analysis. Full-scale test data indicated this threshold was in excess of 30 percent of SMYS, the lowest threshold identified for rupture due to corrosion, whereas the steels represented in reportable incidents possess toughness indicated a threshold on order of 25 percent of SMYS. Analysis indicated that rupture due to a delayed mechanical damage incident required the coincidence of several unlikely circumstances. These circumstances included survival of the initial contact leading to the possibility of a delayed failure, the creation of a long scar due to sliding contact that ran axially along the pipeline, and operation of the pipeline at a pressure greater than when damage was inflicted. Probabilistic calculations  best indicate the coupled likelihood of such events, and are currently being done. In the absence of results from this continuing work, the threshold for rupture due to mechanical damage must be taken as the lower of the above-cited results, that is 25 percent of SMYS. The main conclusion is that thresholds for the transition from leak to rupture are consistent with the current regulatory provisions for low-wall-stress pipelines. 32

References 1.

2. 3.

4. 5.

6. 7.

8.

9.

10.

11.

12.

Leis, B. N., Eiber, R. J., Carlson, L. E., and Gilroy-Scott, A., “Relationship Between Apparent Charpy Vee-Notch Toughness and the Corresponding Dynamic CrackPropagation Resistance” , International Pipeline Conference - 1998, ASME, Calgary, pp. 723-732. Eiber, R. J., Leis, B. N., Carlson, L. E., Horner, N., and Gilroy-Scott, A., “Full-Scale Tests Confirm Pipe Toughness”, Oil & Gas Journal, Nov.8, 1999, pp. 48-54. Selig, B., et al, “Natural Gas Transmission Pipeline Integrity: Prevention, Detection, and Repair Practices”, GRI-00/0193, December, 2000: see also Trench, C.J., “The Safety Performance of Natural Gas Transmission and Gathering Systems”, Allegro Report to GRI, GRI-00/0077, March, 2000. Anon., Pipelines and Public Safety, Special Report 219, Transportation Research Board, United States National Research Council, 1988 Eiber, R., McGehee, W., Hopkins, P., Smith, T., Diggory, I, Baldwink, T., and McHugh, D., “Valve Spacing Basis for Gas Transmission Pipelines”, PR 2249-9728, Jan 2000, PRCI. Kiefner, J. F., et al., “Analysis of Pipeline Incident Data – 1984 to 1998” final report to PRCI, PR-218-9801, 1999 Leis, B. N., and Bubenik, T. A., “Primer on Design to Avoid Failure in Steel Transmission Pipelines”, Battelle Final Report to GTI, GRI 00/0229, January 2001: further details can  be found in Timoshenko, S., and Woinowsky-Krieger, S., Theory of Plates and Shells, McGraw-Hill, 1959, for structural failure, and Broek, D., Elementary Engineering Fracture Mechanics, Noordhoff, 1974, for fracture issues. Leis, B. N., and Stephens, D. R., “An Alternative Approach to Assess the Integrity of Corroded Line Pipe – Part One: Current Status and Part Two: Alternative Criterion”, 7th International Conference on Offshore Pipelines and Polar Engineering, Vol. 3, pp 624 – 634 and pp 635 – 641, 1997, Leis, B. N., and Brust, F. W., “Ductile Fracture Properties of Selected Line Pipe Steels”, (NG-18 Report No. 183) A.G.A. Cat. No. L51604, January 1990: for recently updated correlations and a much expanded database see Leis, B. N., and Kilinski, T. J., “Mechanical and Fracture Properties Essential in Integrity Management of Steel Transmission Pipelines”, Draft Final Report, PRCI Report PR3-9606/9737, Dece mber, 2000 Leis, B. N., and Rudland, D. L., “Evaluation of the Benefits of Hydrotesting GasTransmission Pipelines”, Phase Two – Year One Interim Report, PR-3-9523: see also Leis, B. N., and Rudland, D. L., “Toward Alternatives to Hydrotesting to Ensure Safe Serviceability of Pipelines,” 11th PRCI/EPRG Biennial Joint Technical Meeting on Line Pipe Research, pp 28.1-20, Arlington, 1997. Leis, B. N., and Ghadiali, N., “Pipe Axial Flaw Failure Criteria—PAFFC: User's Manual and Software”, Release 1.0, NG-18 Report No. 211, AGA Cat. No. L51680, 1994, through Release 4.0, to appear January 2001 Brust, F. W., and Leis, B. N., "A Study of Primary Creep Crack Growth at Room Temperature", 5th International Conference on Numerical Methods in Fracture Mechanics,  pp 321-332, 1990

33

13. Brust, F. W., and Leis, B. N., "A New Model for Predicting Primary Creep Damage in Axially-Cracked Cylinders: Part I – Theory ", Engineering Fracture Mechanics, Vol. 43,  pp615-627, 1992: see also Brust, F. W., and Leis, B. N., "A New Model for Predicting Primary Creep Damage in Axially-Cracked Cylinders: Part II – Application ",  Engineering Fracture Mechanics, Vol. 43, pp 629-639, 1992 14. Anon., Report of the Public Inquiry Concerning Stress Corrosion Cracking on Canadian Oil and Gas Pipelines, Proceeding MH-2-95: National Energy Board of Canada,  November 1996 15. Leis, B. N., and Brust, F. W., "Validation of Room-Temperature Primary Creep Crack Growth Analysis for Surface-Cracked Pipes", Nuclear Engineering and Design, Vol. 142,  pp 69-75, 1993. 16. Maxey, W. A., 1974, “Fracture Initiation, Propagation, and Arrest”, 5th Symposium on Line Pipe Research, A.G.A. Catalog No. L30174, Paper J, 1979 17. Leis, B. N., “Relationship Between Apparent Charpy Vee-Notch Toughness and the Corresponding Dynamic Crack-Propagation Resistance”, Battelle report to Ro bert J. Eiber, Consultant, Inc., For Alliance Pipeline Co., June, 1997 18. Leis, B. N., and Eiber, R. J., “Fracture Propagation Control in Onshore Transmission Pipelines”, (Invited Paper) Proceedings, 1998 Meeting On Onshore Pipeline Technology, International Business Conferences/London (IBC), Istanbul Turkey, December 1998, pp. 2.1 – 2.35. 19. Maxey, W. A., 1974, “Fracture Propagation Studies”, 6th Symposium on Line Pipe Research, A.G.A. Catalog No. L30175, Paper J, 1974 20. Anon., ANSI/ASME B31G –1991, “Manual for Determining the Remaining Strength of Corroded Pipelines”, ASME, 1991 21. Kiefner, J. F., and Vieth, P. H., “Modified Criterion for Evaluating the Remaining Strength of Corroded Pipe”, Pipeline Research Supervisory Committee, A.G.A. Catalog  No. L51609, 1989 22. Vieth, P. H., and Kiefner, J. F., “Database of Corroded Pipe Tests”, Pipeline Research Supervisory Committee, A.G.A. Catalog No. L51689, 1994: see also Vieth, P. H., and Kiefner, J. F., “RSTRENG2 Users Manual”, Pipeline Research Supervisory Committee, A.G.A., 1993 23. Leis, B. N., and Stephens, D. R., “An Alternative Approach to Assess the Integrity of Corroded Line Pipe – Part One: Current Status and Part Two: Alternative Criterion”, 7th International Conference on Offshore Pipelines and Polar Engineering, Vol. 3, pp 624 – 634 and pp 635 – 641, 1997 24. Batte, A.D., Fu, B., Kirkwood, M.G., and Vu, D.; A New Method for Determining the Remaining Strength of Corroded Pipelines, 16th International Conference on Offshore Mechanics and Arctic Engineering, (OMAE 1997), American Society of Mechanical Engineers, Yokohama, Japan, 1997 25. Fowler, J. R., et al, “Cyclic Pressure Fatigue Life of Pipelines With Plain Dents, Dents With Gouges, and Dents With Welds,” Final Report on PRCI PR-201-927 and PR-2019324, 1994: see also Keating P. B. and Hoffmann, R. L., “Fatigue Behavior of Dented Petroleum Pipelines”, Texas A&M University Final Report to US Department of Transportation, 1997

34