Catalog No.

L51749e

CONTINUED VALIDATION OF RSTRENG

Contract PR-218-9304

Prepared for the

Line Pipe Research Supervisory Committee
Pipeline Research Committee

of
Pipeline Research Council International, Inc.

Prepared by the following Research Agencies:

Kiefner & Associates, Inc.

Authors:
John F. Kiefner
Patrick H. Vieth
Itta Roytman

Publication Date:
December 20, 1996
“This report is furnished to Pipeline Research Council International, Inc. (PRCI) under
the terms of PRCI PR-218-9304, between PRCI and Kiefner & Associates, Inc. The
contents of this report are published as received from Kiefner & Associates, Inc.. The
opinions, findings, and conclusions expressed in the report are those of the authors and
not necessarily those of PRCI, its member companies, or their representatives.
Publication and dissemination of this report by PRCI should not be considered an
endorsement by PRCI or Kiefner & Associates, Inc., or the accuracy or validity of any
opinions, findings, or conclusions expressed herein.

In publishing this report, PRCI makes no warranty or representation, expressed or

implied, with respect to the accuracy, completeness, usefulness, or fitness for purpose of
the information contained herein, or that the use of any information, method, process, or
apparatus disclosed in this report may not infringe on privately owned rights. PRCI
assumes no liability with respect to the use of , or for damages resulting from the use of,
any information, method, process, or apparatus disclosed in this report.

The text of this publication, or any part thereof, may not be reproduced or transmitted in
any form by any means, electronic or mechanical, including photocopying, recording,
storage in an information retrieval system, or otherwise, without the prior, written
approval of PRCI.”

Pipeline Research Council International Catalog No. L51749e

Price: $345

Copyright, 1996
All Rights Reserved by Pipeline Research Council International, Inc.

PRCI Reports are Published by Technical Toolboxes, Inc.

3801 Kirby Drive, Suite 340

Houston, Texas 77098
Tel: 713-630-0505
Fax: 713-630-0560
Email: info@ttoolboxes.com
 Pipeline Research Council International, Inc.

J. P. Lucido, ANR Pipeline Company (Chairman)

Manager, Consulting Services Dept., Saudi Aramco
P. S. Anderson, Foothills Pipe Lines Ltd.
B. L. Browning, Exploration and Production Technology
R. B. Dun, Gas Tranmission Corporation
P. J. Dusek, Natural Gas Pipeline Company of America
G. Good, Transportadora de Gas de1 Norte
E. Herloe, Statoil
D. L. Johnson, Enron Corp
G. E. H. Joosten, N. V. Nederlandse Gasunie
R. B. Maas, Westcoast Energy Inc.
H. A. Madariaga, Southern California Gas Company
J. K. McDonald, AGL Pipelines Limited
M. L. McGonagill, Tenneco Energy
S. V. Nanney, PanEnergy Corporation
S. Nunez, El Paso Natural Gas Company
D. M. Nunn, TransGas Ltd.
C. W. Petersen, Exxon Production Research Company
D. E. Reid, TransCanada PipeLines, Ltd.
N. Schultz, Transcontinental Gas Pipe Line Corp.
P. R. Smullen, Shell Development Company
Y. Sone, Osaka Gas Company, Ltd.
P. M. Sorensen, Dansk Olie og Naturgas A/S
B. C. Sosinski, Consumers Power Company
M. Tallantyre, British Gas plc
E. E. Thomas, Southern Natural Gas Company
R. J. Turner, NOVA Gas Transmission Ltd.
J. Vainikka, GASUM Oy
D. B. L. Walker, BP Exploration
G. L. Walker, Pacific Gas Transmission Company
T. D. Willke, Gas Research Institute
K. F. Wrenn, Jr., Columbia Gas Transmission Corp.
M. L. Yoho, CNG Transmission Corp.
G. J. Bart, PRC International Staff
A. G. Cotterman, PRC International Staff
B. Dutton, PRC International Staff
 LINE PIPE RESEARCH SUPERVISORY COMMITTEE
Chairman

H. M. Crump, Enron Corp

Members

W. E. Amend, Southern California Gas Company

J. L. Barger, CNG Transmission Corporation
D. Batte, British Gas plc
D. W. Bodkins, Columbia Gas Transmission Corp.
C. Bonar, Gas Transmission Corporation
R. R. Bryant, Union Gas Limited
E. B. Clark, Columbia Gas Transmission Corp.
J. R. Coke, ARCO Exploration and Production
S. P. Cox, Saudi Aramco
B. S. Delanty, TransCanada PipeLines, Ltd.
W. J. DeVries, Consumers Power Co.
J. R. Ellwood, Foothills Pipe Lines Ltd.
R. E. Graham, TransGas Ltd.
W. J. Harris, Pacific Gas and Electric Company
W. E. Holmes, East Australian Pipeline Ltd.
C. D. Howard, Natural Gas Pipeline Co. of America
L. J. Jaskowiec, Great Lakes Gas Transmission Company
C. Juhl, Dansk Olie og Naturgas A/S
A. Korpela., Gasum Oy
V. B. Lawson, Westcoast Energy Inc.
K. Leewis, Gas Research Institute
A. C. Madsen, BP Exploration (Alaska) Inc.
A. J. Maghes, PanEnergy Corp.
H. MacPherson, Transportadora de Gas de1 Norte
B. D. Metzger, El Paso Natural Gas Company
W. G. Morris, PanEnergy Corporation
Y. Nishikawa, Osaka Gas Company, Ltd.
A. Noklebye, Statoil
T. R. Odom, Texas Gas Transmission Corporation
J. R. O’Donnell, Exxon Production Research Co.
G. G. Perkins, Shell Development Company
J. Rabinowitz, Algonquin Gas Transmission Co.
W. E. Ritchie, El Paso Natural Gas Company
A. B. Rothwell, NOVA Gas Transmission Ltd.
R. W. Scrivner, Transcontinental Gas Pipe Line Corp.
W. Sloterdijk, N. V. Nederlandse Gasunie
E. L. Smith, Natural Gas Pipeline Company of America
G. L. Smith, ANR Pipeline Company
R. Sutherby, NOVA Gas Transmission Ltd.
G. Vervake, Tenneco Energy
G. J. Bart, PRC International Staff
 EXECUTIVE SUMMARY

This report presents the results of Project PR 218-9304

sponsored by the A.G.A./PRCI Line Pipe Research Committee. The
purpose of this project was to obtain and analyze new information
on the behavior of corroded pipe that would address the validity
of RSTRENG. RSTRENG is the PC software which the pipeline
industry intends to use to evaluate the remaining strength of
corroded pipe. RSTRENG is an enhancement of the existing ASME
B31G criterion that the industry presently relies upon. RSTRENG
was developed on a prior PRCI-sponsored project, PR 3-805 and its
validity already has been demonstrated to a significant degree on
the basis of 86 burst tests of corroded pipe and pipe samples
containing corrosion-simulating flaws. Nevertheless, a
continuing validation effort is justified because of some
postulated limitations of the methodology and because of the
critical importance of the adequacy of the entire concept of
declaring a corroded pipeline to be serviceable by assessing its
remaining strength.
Contained herein is an analysis of the results of 129
new tests involving corroded pipe or pipe samples containing
corrosion-simulating defects. All of these results either have
been or will be added to the A.G.A./PRCI Database of Corroded Pipe
tests that consisted of 86 results at the time that Project PR 3-
805 (to develop RSTRENG) was completed. The latest results
provide both qualitative and quantitative validation of the
RSTRENG methodology. The qualitative validation arises from the
fact that the results help to more clearly define the
applicability and limitations of RSTRENG. The quantitative
validation arises from comparisons, where applicable, between the
actual failure pressures of the specimens and those that were
predicted using RSTRENG.
For reasons explained in the report, 47 of the 129 new
results do not lend themselves to RSTRENG validity comparisons.
However, the remaining 82 results when combined with the original

i
86 results in the database, yield ratios of predicted-to-actual
failure pressures which enhance the credibility of RSTRENG.

i i
 TABLE OF CONTENTS
Page

EXECUTIVE SUMMARY. . . . . . . . . . . . . . . . . . . . . . i

INTRODUCTION . . . . . . . . . . . . . . . . . . . . . . . . . 1

ASSESSMENT OF TESTS ON PIPES CONTAINING CORROSION-SIMULATING

FLAW. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3

Analysis of Burst Tests on Long Defects and Spirally

Oriented Defects . . . . . . . . . . . . . . . . . . . . . 3

The Effects of Spiral Orientation . . . . . . . . . . . . .5

Effect of Very Long Defect Length . . . . . . . . . . . . . .7

Effects of Flaws Lying on a Single Axial Line Separated

by Sound Pipe (Type II Interaction) . . . . . . . . . 8

Effects of Axially Oriented Pairs of Flaws Offset

Circumferentially but Overlapping Longitudinally
(Type I Interaction) . . . . . . . . . . . . . . . 10

Behavior of Patches of Metal Loss Having Appreciable

Circumferentially as Well as Longitudinal Extent . . . . . . . . . . . 12

Discussion of Index Nos. 93-105 and 118-124 . . . . 14

Analysis of Burst Tests on Long Defects and Various

Combinations of Possibly Interacting Defects . . . . . . 15

Effect of Defect Length (Index Nos. 125 through 129) . . . 16

Effects of Interaction of Long, Narrow Defects

(Type I and Type II Interactions) . . . . . . . . . 17

Effects of Interactions of Pits . . . . . . . . . . 21

Effects of Interaction of Areas of Missing Metal

(Type II Interaction) . . . . . . . . . . . . . . . 25

Behavior of Short, Deep Defects Lying Within Longer

Shallower Defects (Type III Interaction) . . . . . 29

Discussion of the British Gas Tests . . . . . . . . 31

Analysis of Full-Scale Tests Involving Combined Pressure and

Bending Loads on Corroded Pipe . . . . . . . . . . . . . 32

Results. . . . . . . . . . . . . . . . . . . . .32

i i i
 D i s c u s s i o n. . . . . . . . . . . . . . . . . . . . 34

Experimental and Analytical Studies of the Behavior of

Isolated and Closely Spaced Corrosion Pits in Pressurized
P i p e . . . . . . . . . . . . . . . . . . . . . . . . . 36
Summary of the Results . . . . . . . . . . . . . . 38

Circumferentially Arrayed Pits . . . . . . . . . .38

Longitudinally Arrayed Pits . . . . . . . . . . . .39

Spirally Arrayed Pits . . . . . . . . . . . . . . .4 0

Behavior of Pits Within a Corroded Region

(Type III Interaction) . . . . . . . . . . . . . .41

Discussion . . . . . . . . . . . . . . . . . . . . 42

ASSESSMENT OF BURST TESTS, H Y D R O S T A T I C T E S T F A I L U R E S A N D

SERVICE FAILURES INVOLVING CORRODED LINE PIPE. . . . . . . .42

Analysis of Burst Tests on Samples of Corroded

24-inch O.D. by 0.312-inch WT. Grade B Seamless
Line Pipe and 12.75-inch O.D. by 0.188-inch WT.
Grade X52 ERW Line Pipe. . . . . . . . . . . . . . . . 42

Burst Test of a Corroded Piece of 10.75-(Inch) O.D.

by 0.25-Inch WT. API 5L Grade X46 Seamless Line
P i p e. . . . . . . . . . . . . . . . . . . . . . . . . . 48

" B e f o r e " Measurements of Wall Thickness . . . . . . 49

RSTRENG2 Calculation . . . . . . . . . . . . . . . 50

Pressure Test . . . . . . . . . . . . . . . . . . . 50

" A f t e r " Measurements of Bulging and

Wall Thinning . . . . . . . . . . . . . . . . . . . 51

Analysis of Service Failure of Corroded Pipe. . . . . . 54

CORROSION-CAUSED METAL-LOSS ANOMALIES FOR WHICH RSTRENG

ANALYSES ARE NOT APPROPRIATE . . . . . . . . . . . . . . . . 5 7

Analysis of Selective Corrosion in a Low-Frequency

E R W S e a m. . . . . . . . . . . . . . . . . . . . . . . . 57

Analysis of Corrosion in a Furnace Lap-Welded Material . . . . . . 5 8

D i s c u s s i o n. . . . . . . . . . . . . . . . . . . . . . . 59

iv
QUALITATIVE AND QUANTITATIVE VALIDATION OF RSTRENG . . . . . 59

Qualitative Findings Regarding the Use of RSTRENG . . . . . . . .60

Quantitative Findings Regarding the Validity of RSTRENG. . . . . . . . .65

REFERENCES . . . . . . . . . . . . . . . . . . . . . . . . . 69

APPENDIX A - METHODS FOR DEALING WITH THE

INTERACTIONS OF CLOSELY SPACED
AREAS OF METAL LOSS . . . . . . . . . . . . . A-1

APPENDIX B - RSTRENG2 CALCULATIONS FOR THE ANOMALIES . . . . B-1

LIST OF TABLES

T a b l e 1. Burst Tests of Pipes Containing Long Axially

and Spirally Oriented Corrosion-Simulating
Slots and Patches . . . . . . . . . . . . . . . . . . . . . . . . . . 71

Table 2. Comparisons Between RSTRENG Predictions and Experiments

by British Gas . . . . . . . . . . . . . . . . . . . . . . . . .73

Table 3. Experiments on 48-Inch Pipe Containing Corrosion-

Simulating Machined Areas Subjected to Pressure and
Bending Moment . . . . . . . . . . . . . . . . . . . 75

Table 4. Results of Tests Involving Circumferential and

Longitudinal Arrays of Pits . . . . . . . . . . . . . . . . . . . . . . 76

Table 5. Results of Tests Involving a Small Pit Within

a Larger Corroded Area . . . . . . . . . . . . . . 78

Table 6. Description of Materials Involved in Corroded

Pipe Burst Tests at a Pipeline Operator's
Facility in 1970 . . . . . . . . . . . . . . . . . 79

Table 7. Results of Tests of Corroded Samples Compared

to RSTRENG Predictions . . . . . . . . . . . . . . 80

Table 8. List of 168 Results of Tests of Corroded Pipe

and Pipe Specimens Containing Corrosion-
Simulating Defects and Comparisons to
RSTRENG Predictions . . . . . . . . . . . . . . . .81

Table 9. List of 82 Ruptures of Corroded Pipe and Pipe

Specimens Containing Corrosion-Simulating Defects
a n d C o m p a r i s o n s t o R S T R E N G P r e d i c t i o n s . . . . . . 85

Table 10. List of 86 Leaks of Corroded Pipe and Pipe

Specimens Containing Corrosion-Simulating Defects
a n d C o m p a r i s o n s t o R S T R E N G P r e d i c t i o n s . . . . . . 87

V
 LIST OF FIGURES
Figure 1. Index Nos. 128-133, Long, Narrow Defects
and Possible Interactions . . . . . . . . . . . . . . . . . . . 89

Figure 2. Index Nos. 134-142, Arrays of Pits . . . . . . . . . . . . . . . . . 90

Figure 3. Index Nos. 143-151, Patches of Metal Loss. . . . . . . . . . . . . 91

Figure 4. Index Nos. 152-157, Compound Defects . . . . . . . . . . . . . . . . . . 92

Figure 5. General Shape of Individual Electrochemically

Machined Defects . . . . . . . . . . . . . . . . . . . . . . . 93

Figure 6. Typical Shape of Pit Within Corroded Region . . . . . . . . . . . . . 94

Figure 7. Circumferential, Longitudinal, and Spiral

Arrays of Pits . . . . . . . . . . . . . . . . . . . . . . . . . 95

Figure 8. Profiles of Anomalies 1-1 and 1-2

from Pipe Number 1 . . . . . . . . . . . . . . . . . . . . . 96

Figure 9. Plan and Profile Views of Anomaly 1-3

From Pipe Number 1. . . . . . . . . . . . . . . . . . . 97

Figure 10. Profiles of Anomalies 3-1 and 3-2 from

Pipe Number 3 . . . . . . . . . . . . . . . . . . 98

Figure 11. Plan and Profile Views of Anomaly 3-3

From Pipe Number 3 . . . . . . . . . . . . . . . 99

Figure 12. Anomalies From Pipe Number 3A: First Leak

(2A-1, TOP), Second Leak (3A-2), Bottom) . . . . . . . . 100

Figure 13. Anomaly That Ruptured Pipe Number 3A (3A-3B) . . . . . . . . . . 101

Figure 14. Profiles of Anomalies From Pipe Number 4. . . . . . . . . . . . . 102

Figure 15. Anomaly 5-1 From Pipe Number 5. . . . . . . . . . . . . . .103

Figure 16. Anomaly 5-2 From Pipe Number 5. . . . . . . . . . . . . . . 104

Figure 17. Photograph of Index 214 Before Test. . . . . . . . . . . . . . . 105

Figure 18. Contours of Pit Depths "Before" Pressure

Test, Index 214 . . . . . . . . . . . . . . . . . 106

Figure 19. RESTRENG Calculation for Index 214 Using

"Before" Measurements . . . . . . . . . . . . . . . . . . . . . 107

Figure 20. Pressure and Volume as a Function of Time. . . . . . . . . . 109

v i
Figure 21. D i a m e t r i c R e a d i n g s o f C o r r o d e d S a m p l e . . . . . . . . 110

Figure 22. Contours of Pit Depths "After" Pressure

T e s t , Index 214. . . . . . . . . . . . . . . . . 111

Figure 23. RSTRENG Calculation for Index 214 Using

"After" Measurements . . . . . . . . . . . . . . . . . . . . . 112

Figure 24. F r a c t u r e P a t h a n d O r i g i n L o c a t i o n . . . . . . . . . . . . . . 114

Figure 25. Origin of the Rupture . . . . . . . . . . . . . . . . . . 115

Figure 26. C o n t o u r s o f R e m a i n i n g T h i c k n e s s . . . . . . . . . . . . 116

Figure 27. RSTRENG2 Prediction of Failure Pressure. . . . . . . . . . . 117

Figure 28. M e t a l l o g r a p h i c S e c t i o n a t O r i g i n . . . . . . . . . . . . . 118

Figure 29. Photograph of Origin of Rupture . . . . . . . . . . . . . . . . 119

Figure 30. Photomacrograph of Cross Section Through

the Deepest Pit at the Origin . . . . . . . . . . . . . . . . 120

Figure 31. Photomacrograph of Cross Section Through Intact

Portion of Weld Showing Some Selective Corrosion . . . . . . . 120

Figure 32. RSTRENG Calculation for Selective Corrosion

Defect in ERW Seam . . . . . . . . . . . . . . . 121

Figure 33. Overall Pitted Area . . . . . . . . . . . . . . . 122

Figure 34. Origin . . . . . . . . . . . . . . . . . . . . . 122

Figure 35. RSTRENG Calculation for Pits Located in Bondline

Region of Lap-Welded Pipe . . . . . . . . . . . . . . . . . 123

v i i
THIS PAGE IS INTENTIONALLY BLANK
 FINAL REPORT
on
CONTINUED VALIDATION OF RSTRENG
by
J. F. Kiefner, P. H. Vieth, and I. Roytman

INTRODUCTION

This report presents the results of Project PR 218-9304

sponsored by the A.G.A./PRCI Line Pipe Research Committee. The
purpose of this project was to obtain and analyze new information
on the behavior of corroded pipe that would address the validity
of RSTRENG. RSTRENG is the PC software which the pipeline
industry intends to use to evaluate the remaining strength of
corroded pipe. RSTRENG is an enhancement of the existing ASME
B31G criterion(1)* that the industry presently relies upon.
RSTRENG was developed on a prior PRCI-sponsored project, PR 3-805
and its validity already has been demonstrated to a significant
degree on the basis of 86 burst tests of corroded pipe and pipe
samples containing corrosion-simulating flaws. Nevertheless, a
continuing validation effort is justified because of some
postulated limitations of the methodology and because of the
critical importance of the adequacy of the entire concept of
declaring a corroded pipeline to be serviceable by assessing its
remaining strength.
The original criterion for assessing the remaining
strength of corroded pipe was an outgrowth of extensive work on
the effects of longitudinal cracks in pressurized pipes(2,3). This

*Numbers in parenthesis refer to References on Page 69.

 2

work was carried out in the 1960s under sponsorship of

A.G.A./PRCI, and it led to the development of the NG-18 surface
flaw equation. Subsequently, an effort was begun under the
sponsorship of Texas Eastern Transmission Corporation which
continued under A.G.A./PRCI sponsorship which involved adapting
the surface flaw equation to predicting the remaining strength of
corroded pipe(4). The result of this work was the ASME B31G
methodology for evaluating corrosion-caused metal loss in
pipelines.
In the late 1980s, the excess conservatism of the B31G
criterion became an important issue. Using the B31G criterion,
pipeline operators discovered that often pipe was being removed
that still had more than adequate remaining strength. This
concern led to Project PR 3-805, the main output of which was
RSTRENG(5). The RSTRENG approach involves an improvement of the
B31G methodology while relying on the same basic theoretical
concept.
Since the initial release of the RSTRENG software, work
on evaluating the behavior of corroded pipe has continued along
several lines. Some of the continuing work has produced
pipe(6-11)
additional experimental data on the behavior of corroded .
This report addresses the significance of these additional data
with respect to he RSTRENG methodology. In addition, we have
continued to collect and analyze data from corrosion-caused
service and hydrostatic test failures to add to the A.G.A./PRCI
"Database of Corroded Pipe Tests".(12) These new data are analyzed
herein and their impact on the RSTRENG methodology is assessed.
For all calculations, the new RSTRENG2 ( 1 3 ) has been used.
 3

ASSESSMENT OF TESTS ON PIPES

CONTAINING CORROSION-SIMULATING FLAWS

Since the PR 3-805 work was completed and the 86-test

database was established several researchers have published
important test results on the behavior of corroded pipe. The
issues addressed include:
The behavior of very long flaws
The interaction of adjacent corroded regions
The behavior and interactions of isolated pits
The effects of diagonal orientation of corrosion
The effects of patches of metal loss having both
longitudinal and circumferential dimensions of
significance
The effects of large axial stresses on the behavior of
corroded pipe.
The results of each of these investigations are discussed and
analyzed below. Because interaction of adjacent flaws is a major
issue addressed in the following analyses, it may be useful for
the reader to review the definitions and methods of analysis we
intend to use to characterize interaction of multiple corrosion
pits, patches, and slots. These definitions and methods are
presented in Appendix A. In particular, it may be useful to
refer to Figure A-l.

Analysis of Burst Tests on Long Defects

and Spirally Oriented Defects

A series of experiments on long axially oriented and

spirally oriented corrosion-simulating grooves was carried out by
NOVA Corporation.(6) Twelve burst tests were conducted on samples
of 20-inch O.D. by 0.250-inch wall thickness X60 pipe containing
machined slots of various sizes and orientations. All of the
slots were machined to a single depth, 40 percent of the wall
 4

thickness, and all slots were the same width, 1 inch.

Subsequently six additional burst tests were conducted by Nova
involving patches of corrosion-simulating metal loss having
significant circumferential as well as axial extent. While these
latter tests were done after the publication of Reference 6, the
data were made available by Nova for analysis on this project.
It is noted that all of the experiments were pressure vessel
burst tests, so the axial stress was one-half that of the hoop
stress. This type of loading is used in the vast majority of
tests on corroded pipe, and it is much like the loading
experienced by most buried pipelines. However, it must be
recognized that the results of pressure vessel tests, especially
in cases such as some of these where flaws are not exclusively
longitudinally oriented, may not be representative of the
behaviors one could expect in the presence of extreme axial
loads.
The data from all of Nova's experiments are reproduced
in this report in Table 1. Because they are included in the
A.G.A./PRCI Database of Corroded Pipe Tests, they have been
a s s i g n e d "Index Numbers". The results of these tests have been
analyzed by means of RSTRENG and the comparisons are discussed
below. The RSTRENG calculations are presented in Appendix B.
The series of tests presented in Table 1 involved five
experiments with spirally oriented grooves (Index Nos. 93, 94,
95, 96, and 102), two experiments with single, longitudinally
oriented grooves (Index Nos. 97 and 98), three experiments with
longitudinally oriented grooves on a single axial line separated
by undamaged pipe (Index Nos. 99, 100, and 101), two experiments
with longitudinally oriented grooves offset circumferentially but
o v e r l a p p i n g l o n g i t u d i n a l l y ( I n d e x N o s . 103 and 104) and six
experiments on patches of corrosion (Index Nos. 119-124). Two
addition vessels, Index Nos. 105 and 118, were tested to failure
 5

with no defects to assess the ultimate burst pressure of the pipe

material.
The interaction of flaws designated Index Nos. 99, 100,
and 101 is defined in Figure A-1 of Appendix A as Type II
interaction. The interaction of flaws designated Index Nos. 103
and 104 is defined in Figure A-1 of Appendix A as Type I
interaction.

The Effects of Spiral Orientation

This series of tests is extremely valuable because it

is unique. No other data on spirally oriented flaws existed in
the A.G.A. PRCI Database of Corroded Pipe Tests prior to these.
As one can see from the date in Table 1, these spiral grooves
were oriented at angles ranging from 20 to 45 degrees from the
circumferential direction. I n s p i t e o f t h e i r v e r y l o n g l e n g t h
along the axes of the grooves, these defects failed at relatively
high pressure levels (80 to 100 percent of the burst pressure of
the defect-free specimen). All five produced ruptures along the
spiral grooves, but the pressure levels at failure are much
greater than those of axial flaws of similar length and greater
than one would expect on the basis of the "projected" length.
The latter is the length that the defect projects along the axis,
that is, its actual length multiplied by the cosine of the angle
between the defect and the axis of the pipe (or the sine of the
angle between the defect and the circumferential direction as it
is defined in Table 1). Note, for example, that even though the
flaws of Index Nos. 93, 94, and 95 have 15-inch-projected
lengths, they exhibited considerably higher failure pressures
than that of the 15-inch-long longitudinally oriented flaw
(Index 97).
The failure pressures of the spirally oriented flaws
were ordered with the angle from the axial direction. The
 6

greater the angle from the axis (or the less the angle as defined
in Table 1) the greater was the failure pressure. The authors of
Reference 6 propose the consideration and further verification of
a "spiral angle factor" or SAF that they evaluated empirically
from the data as follows:

Angle of Defect from

Circumferential
Direction, degrees SAF
20 1.3
30 1.2
45 1.1
90 1.0

They suggest that estimates of remaining strength based on the

projected length be multiplied by the SAF to get more accurate
(less conservative) estimates.
Another finding from these data was that the pair of
spiral grooves (Index 102) did not interact. This pair consisted
of two 37.41-inch grooves oriented at 20 degrees to the
circumferential direction, separated axially by 12.8 inches. One
of the pair failed at a pressure level expected if it had been
the only defect present.
The implications of these data for RSTRENG are examined
as follows. First, in the past it has been recommended that the
projected length of nonaxial flaws be used to evaluate the
remaining strength of a pressurized pipe containing such flaws.
These data provide insight into this recommendation.
 7

RSTRENG-Predicted Levels for Comparison

Angle from
Circumfer-
ential Actual Projected Ratio, Projected Ratio,
Direction, Failure Length Predicted to Length x Predicted to
Index No. degrees SAF Pressure, psig Alone, psig Actual SAF, psig Actual
93 20 1.3 2110 1241 0.59 1613 0.76
94 30 1.2 2008 1241 0.62 1489 0.74
95 45 1.1 1790 1241 0.69 1365 0.76
102 20 1.3 2211 1271 0.57 1652 0.75
97 90 1.0 1631 1283 0.79 1283 0.79

Obviously, the use of the projected length results in

overconservative predictions: the flawed pipes exhibited much
higher failure pressures than predicted. When the predictions
are adjusted by the SAF, m o r e r e a l i s t i c e s t i m a t e s a r e o b t a i n e d .
Note that Index 97, a longitudinal, 15-inch-long flaw, is
included in the above list to illustrate the tendency for RSTRENG
to give conservative predictions in general for this pipe
material. The estimates for the spiral defects based on the SAFs
seem to be in line with that of the unadjusted estimate for the
axially oriented defects. Thus, the concept of the SAF appears
valid and deserves further verification. In the meantime,
RSTRENG users, if they wish, can safely continue to evaluate
spirally oriented corrosion on the basis of its projected length.
In such a case, the answers derived are likely to embody excess
conservatism.

Effect of Very Long Defect Length

The tests designated Index Nos. 97 and 98 in Table 1

provide valuable insight into the behaviors of very long
longitudinal defects. The flaws in each of these tests
represented lengths of 0.75 pipe diameter and 2.0 pipe diameters,
respectively. The shorter of the two, Index 97, exhibited a
 8

failure pressure of 1,631 psig. The longer one, Index 98,

exhibited a failure pressure of 1,674 psig. Thus, the effect of
extra length (beyond one diameter) appears to be negligible. At
least, it is small enough to be overshadowed by scatter in the
data.
From the standpoint of flaw length, RSTRENG predicts
decreasing strength with increasing flaw length, but the trend
asymptotically approaches failure in the net section, that is, in
the limit it becomes proportional to 1 - d/t where d is the depth
of the defect and t is the nominal will thickness of the pipe.
The RSTRENG-predicted failure pressures for Index Nos. 97 and 98
are 1,283 and 1,194, respectively. The corresponding ratios of
predicted-to-actual failure pressures are 0.79 and 0.73.
Clearly, RSTRENG provided adequately conservative values
irrespective of length in this case. For practical purposes,
however, the results of Index Nos. 97 and 98 suggest that when
considering a piece of pipe with a very long corroded area, one
does not have to consider more than about l-diameter length as
long as that region contains the deepest pitting.

Effects of Flaws Lying on a Single Axial

Line Separated by Sound Pipe (Type II Interaction)

The results of tests designated Index Nos. 99, 100, and

101 provide insight into the behaviors of Type II interaction as
defined in Figure A-1 of Appendix A. The defects in these tests
were pairs of 6-inch-long axially oriented grooves lying on a
single axial line, separated by amounts of sound pipe of 3, 6,
and 12 inches. As shown in Table 1, the results of the tests
strongly suggest that little or no interaction took place. A l l
three tests failed at the same pressure level. In the cases of
the pairs separated by 3 and 6 inches, the ruptures ran through
both flaws, whereas in the case of the pair separated by
 9

12 inches, the rupture involved only one of the pair. Clearly,

in the latter experiment single-flaw failure behavior is
indicated by the one which failed. The other two results suggest
possibly slight interaction for the pairs separated by 3 and 6
inches, but the interaction, if any, occurs only at a near-
failure pressure level and does not appear to significantly alter
the single-flaw failure pressure. These results would seem to
suggest that pairs of Type II flaws must be closer together than
half a flaw length to interact significantly. These results seem
to be consistent with those presented in Figure A-3 in
Appendix A.
In the past, the method for using RSTRENG for Type II
interaction has been that suggested in Appendix A, namely, to
consider the flaw length to be the total length including the
lengths of both of the individual flaws and the length of the
separation between them. If the predicted failure pressure
derived in this manner is lower than that of the largest of the
two individual flaws, interaction is predicted and the failure
pressure based upon the total length becomes the predicted
failure pressure. As applied to the defects designated Index
Nos. 99, 100, and 101, t h e r e s u l t s w e r e a s f o l l o w s .

RSTRENG-
Actual Failure Predicted Pressure, Ratio of Predicted
Index No. Description Pressure, psig Psig to Actual

99 Two 6-inch flaws 1892 1375 0.73 treated as

separated by 3 single 6 inch
inches
1334 0.71 treated as
Type II with
tapered ends
1375 0.73 treated as
(no interaction) Type II with
square ends
 10

In the above table, only Index 99 (the pair separated

by 3 inches) is considered. As the following analysis will show,
RSTRENG would predict no interaction for the other two pairs.
First, the RSTRENG prediction for the single 6-inch flaw was
calculated. Next, the total 15-inch flaw was considered and the
3-inch separation was treated with values of flaw depth of 0.106
inch at the 6 and 9-inch locations and values of zero depth at
the 7 and 8-inch locations. O n t h i s b a s i s , RSTRENG treats the
flaw as one that tapers linearly from full depth at 6 and 9-inch
locations to full wall thickness at the 7 and 8-inch locations.
For this configuration, the predicted failure pressure is 1,334,
3 percent below that of the single 6-inch flaw. Hence, slight
interaction is predicted. Then, the input was changed by
inserting flaw depths of zero at the 6.01-inch and 8.99-inch
locations, essentially forcing RSTRENG to consider the flaws as
being square-ended (which they were). In this case, no
interaction is predicted, and the single flaw failure pressure of
1,375 psig is predicted. Clearly, the other two pairs of defects
are too widely separated to be predicted by RSTRENG to interact.
As noted before, the RSTRENG predictions for this pipe material
are quite conservative in general for reasons that may be
material related.

Effects of Axially Oriented Pairs of Flaws Offset

Circumferentially but Overlapping Longitudinally
(Type I Interaction)

Two examples of Type I Interaction appear in the

experiments summarized in Table 1. These were Index Nos. 103 and
104. Both were comprised of a pair of 15-inch single grooves
overlapping longitudinally to give a projected length of 20
inches. The two constituting Index 103 were separated
circumferentially by two pipe wall thicknesses, and the two
constituting Index 104 were separated circumferentially by four
 11

wall thickness. The results of the tests suggested that at least

one of the pairs of defects (Index 103) exhibited interaction.
In the case of Index 103, its failure pressure was 1,602 psig,
97.9 percent of that of Index 97, the single 15-inch defect. In
addition, the failure ran through all of one 15-inch flaw and the
part of the other flaw projecting beyond the end of the one which
was completely split. The jump in the fracture from one defect
to the other was abrupt.
In the case of Index 104, the degree of interaction, if
any, is less apparent. Its failure pressure, 1,529 psig, was
only 93.4 percent of the 1,637 psig level exhibited by the 15-
inch single flaw of Index 97. However, t h e f l o w s t r e s s l e v e l o f
the Index 104 material was only 96.6 percent as high as that of
the Index 97 material and its flaw depth was about 1.7 percent
greater. Hence, interaction or the lack thereof is not provable
on the basis of the pressure comparison. At least some
interaction seems to be more clearly indicated by the fact that
the rupture ran through one of the grooves and jumped to the
overlapping portion of the other.
In the past, it has been recommended that the projected
length of Type I defects be used to predict the failure pressure
via RSTRENG for offsets less than 6t (circumferential separation
of less than 6 times the wall thickness). In the case of the
flaws on Index Nos. 103 and 104, the following comparisons were
made.

RSTRENG-
Actual Failure Predicted Pressure, Ratio of Predicted
Index No. Description Pressure, psig psig to Actual

103 Type I Interaction, 1602 1287 0.80 treated as

two 15-inch flaws single 15 inch
separated by 2t
1260 0.79 treated as
single 20 inch
 12

RSTRENG-
Actual Failure Predicted Pressure, Ratio of Predicted
Index No. Description Pressure, psig psig to Actual

104 Type I Interaction, 1529 1233 0.77 treated as

two B-inch flaws single 15 inch
separated by 4t
1174 0.73 treated as
single 20 inch

Treating the pairs as single 15-inch grooves or as a single 20-

inch groove makes little difference in the RSTRENG predictions.
However, this might not be universally true for shorter,
overlapping flaws. I n f a c t , the data shown in Figure A-2 show
that it is not true; interaction becomes significant for closely
spaced, overlapping short flaws. This suggests that the practice
of using the projected length in RSTRENG for Type I interaction
should be continued for offsets of less than 6t.

Behavior of Patches of Metal Loss

Having Appreciable Circumferentially
as Well as Longitudinal Extent

After the publication of Reference 6, Nova conducted

further burst tests on pipes containing corrosion-simulating
metal-loss defects. They contributed the results to the
A.G.A./PRCI Database of Corroded Pipe Tests. The experiments were
designated Index Nos. 118 through 124, and they are listed in
Table 1. All of these experiments involved the same 20-inch O-D.
by 0.250-inch wall thickness X60 pipe material used for the
experiments designated Index Nos. 93 through 105. Index No. 118
was a defect-free burst test, but the others contained various
kinds of metal-loss defects. Index Nos. 119, 120, and 124
contained long, narrow slots similar to but with depths differing
from that of Index 98. The rest (Index Nos. 121 through 123)
involved patches of metal loss in which the circumferential
 13

extent of the metal loss was appreciable. In two cases, Index

Nos. 121 and 122, t h e m e t a l l o s s w a s i n t h e f o r m o f c o m p l e t e
circumferential bands.
When tested to failure in the form of end-capped
pressure vessels, all of the flaws which failed did so as axial
ruptures. Under this type of loading, it appears that the
circumferential extents of the defects played little or no role
in the failure behavior. One way to verify this is to analyze
and compare the results on the basis of RSTRENG.

RSTRENG
Description of Flaw Which Actual Failure Predicted Pressure, Ratio of Predicted
Index No. Failed Pressure, psig psig to Actual

119 Single, long narrow patch 1160 890 0.77

120 Single, long narrow patch 1711 1225 0.72
121 Circumferential band, 4.1 1813 1346 0.74
inches in axial extent
122 Circumferential baud, 8.2 1422 1195 0.84
inches in axial extent
123 Square patch 1226 1043 0.85
124 Single, long narrow patch 1218 957 0.79

W h i l e i t i s t r u e that one circumferential band defect

(Index 122) and one patch (Index 123) appeared to have failed at
pressure levels closer to the RSTRENG-predicted pressures, the
other circumferential band exhibited a predicted-to-actual
failure pressure ratio much the same as those exhibited by the
long, narrow patches (Index Nos. 97, 98, 119, 120, and 124). It
seems reasonable to believe that the scatter in the results is
the result of factors other than the circumferential extent of
the defect.
 14

Discussion of Index Nos. 93-105 and 118-124

The results of these tests provide valuable insight

into the behaviors of some special flaw geometries and their
treatment by means of RSTRENG. In summary, it appears that a
conservative strategy would consist of the following.
• One should continue to use the projected length of
spirally oriented flaws in the analysis. However, the
spiral angle factor if further validated for other
materials and flaw geometries has the potential for
greatly improving the accuracy of the analysis.
• For very long corroded areas, one can limit an RSTRENG
analysis to one diameter of length or a length of 20
inches, whichever is greater, so long as the deepest
pitting is included in the region analyzed.
• F o r T y p e I I i n t e r a c t i o n , RSTRENG can be used to analyze
the individual flaws and the overall combination. The
lowest value resulting from the various configurations
should be used.
• For Type I interaction, one should continue to use the
projected length of the combined flaws in the RSTRENG
analysis whenever the circumferential separation is
less than 6t. When the separation is greater than or
equal to 6t, the flaws may be treated separately.
• The axial extent and maximum depth of patches of
corrosion appear to remain the parameters of importance
in assessing remaining strength. Circumferential
extent appears to play no significant role.
 15

Analysis of Burst Tests on Long Defects and

Various Combinations of Possibly Interactins Defects

An extremely valuable set of data involving burst tests

of pipes containing simulated corrosion defects was presented by
British Gas researchers(') in 1992. The test matrix was developed
by pressurizing specimens of 24-inch O.D. by 0.486-inch wall
thickness X52 line pipe. The corrosion-simulating defects
included long axially oriented slots, pairs of such slots located
in close proximity both end-to-end and side-by-side, arrays of
pits, and patches of reduced wall thickness, some plain and some
containing pits. Patches, as explained in the discussions of
Nova's tests, are areas of metal loss having appreciable
circumferential as well as longitudinal extent. The data are
presented in Table 2 and they are analyzed below with the aid of
Figures 1 through 4. The main thrust of the analysis is to show
the impact of these results on the validity of RSTRENG.
The results of tests on 33 defects are summarized in
Table 2. To facilitate their inclusion into the A.G.A./PRCI
Database of Corroded Pipe Tests, each has been assigned an index
number ranging from 125 to 157. Index Nos. 125 through 129
represent a series of single-slot defects designed to evaluate
the effects of flaw length. The slots were narrow (0.006-inch
wide) and thus cracklike rather than pitlike. In this respect,
the results would tend to give lower-bound failure pressures for
the corresponding amount of pitlike corrosion. Index Nos. 130
through 133 (See Figure 1) address the interactions of closely
spaced slots. Index Nos. 134 through 142 (see Figure 2) address
the interaction of closely spaced round pits. Index Nos. 143
through 151 (see Figure 3) address the behavior of patches of
missing metal, their interactions with each other, and their
interactions with round pits. Finally, Index Nos. 152 through
 16

157 (see Figure 4) address short flaws or pits located within

longer flaws or areas of reduced wall thickness.
Each defect was analyzed by means of RSTRENG, in some
cases in more than one configuration (e.g., as individual short
defects, as an interacting pair, or as a continuous long defect).
Because the tests involved only two different joints of pipe,
only two values of flow stress were applicable. In the case of
the Index 125 test, the flow stress was 65,400 psi. In all other
cases, it was 64,900 psi. The failure pressures determined in
the tests, the RSTRENG-predicted failure pressures, and the
ratios of actual-to-predicted failure pressures are presented in
Table 2. The RSTRENG calculations are presented in Appendix B by
index number.

Effect of Defect Length (Index Nos. 125 Throuqh 129)

In a confirmation of the previously discussed results

presented by Nova, the set of five uniform-depth defects (all 40
percent through the wall) represented by Index No. 125 through
129 shows that failure pressure tends to become independent of
defect length as uniform-depth defects become very long. The
failure pressure levels of the defects in this series decreased
with increasing length to a length of 24 inches. The 120-inch-
long defect actually exhibited a higher failure pressure than the
24-inch defect, probably as a result of variations in parameters
such as flow stress and remaining thickness. However, it is
clear that beyond a length of 24 inches, the effect of flaw
length becomes insignificant. As shown in Table 2, RSTRENG
predictions were made for Index Nos. 125, 126, 127, and 129.
None was necessary for Index 128 since its calculation is
identical to that of Index 127.
 17

It is significant that RSTRENG provided adequate

estimates of the failure pressures. Confidence in the use of
RSTRENG for long corroded areas is enhanced by these results.
Also, because the effect of length tends to wash out beyond a
certain level, those measuring corroded pipe in the field can
safely limit the length over which measurements must be made.
The critical requirement of such measurements would seem to be to
find the point of least remaining wall thickness. Then, it
appears that taking measurements up to one diameter in each
direction is sufficient to characterize the remaining strength
via RSTRENG. In fact, it is probably acceptable to limit
consideration to the one-diameter region centered on the deepest
point.

Effects of Interaction of Long, Narrow Defects

(Type I and Type II Interactions)

Potential interactions between long, narrow uniform-

depth defects (all 40 percent through the wall) located in close
proximity were investigated through the configurations shown in
Figure 1. Note that Index Nos. 128 and 129 were single defects
that were part of the family described above. Index Nos. 130 and
131 are pairs of 40-percent-through-the-wall, 6-inch slots
located on the same axial line separated by distances of I.5
inches and 3 inches, respectively. This type of interaction is
l
known as " Type-II" interaction. Index Nos. 132 and 133 were
pairs of 40-percent-through-the-wall 12-inch slots. One pair
(Index 132) consisted of two slots offset circumferentially by
1/2 inch. Longitudinally their projected areas met end to end.
This type of interaction is known as Type-I interaction. The
other pair consisted of two slots offset circumferentially by 1/2
inch, but longitudinally they overlapped 100 percent.
 18

The criteria for whether or not interaction of these

pairs occurred are as follows:
• For Index Nos. 130 and 131, interaction would be
indicated by a failure pressure below that of
Index 129, the single 6-inch flaw
• For Index Nos. 132 and 133, interaction would be
indicated by a failure pressure below that of
Index 128, the single 12-inch flaw.
It should be noted that interaction could also be
indicated if the mode of failure changed to rupture for the pairs
from that of a leak for the single flaw or if strain measurements
indicated gross yielding occurring between the two defects in the
pair when it is not yet occurring elsewhere in the pipe. Both
single defects (Index Nos. 128 and 129) failed as ruptures,
however, so no conclusions could be drawn from the modes of
failure. And, no strain measurements were reported in
Reference 7.
The results of the tests of the defects shown in
Figure 1 are presented in Table 2. For convenience, the results
are repeated below.

Actual Failure
Index No. Description Status Pressure, psig Interaction
128 Single 12 inch Ruptured 2393 __-
129 Single 6 inch Ruptured 2683 ---

130 Double 6 inch, 1.5 inch end- No failure > 2683 No

to-end separation
131 Double 6 inch, 3 inch end-to- No failure > 2683 No
end separation
132 Double 12 inch, no end-to- Ruptured 2233 Yes
end separation, 1/2-inch
circumferential offset
 19

Actual Failure
Index No. Description Status Pressure, psig Interaction
133 Double 12 inch, side by side, Ruptured 2503 No
1/2-inch circumferential
separation

It is noted that Index 132 failed first and that portion of the
vessel was cut off. Index 128 failed next at a pressure level
nearly seven percent higher, and it was cut out. Finally,
Index I29 failed at a pressure level 10 percent above that of
Index 128. Thus, the influence of prior cycles can be discounted
for each of these results. Index 133 was cut from the specimen
and rewelded as a partial cylindrical segment into another
specimen. It was then tested to failure by itself.
From the standpoint of interaction, Index Nos. 130 and
131 showed no tendency to interact. Both pairs of 6-inch defects
survived the pressure level at which the single 6-inch defect
failed. The pair of defects that comprise Index 132 appear to
have interacted; the failure pressure level of 2,233 psig is only
93 percent of that of a single 12-inch defect, Index 128. The
degree of interaction is suggested by comparing the Index 132
result to that of Index 126, a single 24-inch defect. The latter
failed at a pressure level of 2,030 psig in a material with a
slightly higher flow stress. Thus, the 24-inch-long pair of
offset 12-inch slots failed at a pressure level 10 percent above
that of the single 24-inch defect. Index 133, t h e s i d e - b y - s i d e
pair of 12-inch slots did not interact; their failure pressure
was five percent above that observed for the single 12-inch
defect.
The implications of the behaviors of the defects in
Figure 1 with respect to the applicability of RSTRENG is also
examined in Table 2, but for convenience the relevant comparisons
are shown below.
 20

RSTRENG
Actual Failure Predicted Ratio of Predicted
Index No. Description Pressure, psig Pressure, psig to Actual
128 Single 12 inch 2393 2173 0.91
129 Single 6 inch 2683 2515 0.94
130 Double 6 inch, 1.5-inch end- > 2683 2515 <0.94treated as
to-end separation single 6 inch
2232 <0.83 treated as
two 6-inch defects
separated by 1.5
inches

131 Double 6 inch, 3-inch end-to- > 2683 2515 <0.94 treated as
end separation single 6 inch
2413 <0.90 treated as
two 6-inch defects
separated by 3
inches
132 Double 12 inch, no end-to-end 2232 1995 0.89 treated as
separation, 1/2-inch single 24 inch
circumferential separation
2173 0.97 treated as
single 12 inch
133 Double 12 inch, side-by-side 2509 2173 0.87 treated as
1/2-inch circumference single 12 inch
separation

As suggested previously, the pairs of defects in Index

Nos. 130 and 131 appeared not to interact. As such, it would
seem to make sense to apply RSTRENG to each single 6-inch defect.
Indeed, such a prediction gives a lower-bound estimate as shown
above. Both pairs were also analyzed using the total lengths of
13.5 and 15 inches with 1.5-inch and 3-inch regions of full-wall
thickness in the center. The RSTRENG results are as follows.
The predicted pressure for Index 130 is less than that of
Index 131 and both predicted levels are less than that of a
single 6-inch flaw. Thus, RSTRENG would seem to be modelling
 21

some interaction that decreases with increasing spacing between

the flaws. Since no interaction appeared to have occurred, the
RSTRENG results for the combined pairs may be excessively
conservative.
Two RSTRENG calculations were made for the Index 132
defects. One was made for a single 24-inch flaw, and one was
made for a single 12-inch flaw. It turns out that both
predictions were conservative as shown above. However, since it
did appear that some interaction took place between the pairs, it
would seem prudent to apply the analysis to the 24-inch length as
though the flaw were continuous.
In the case of Index 133, no interaction occurred, and
it seems rational to treat the defect as a single 12-inch flaw.
Indeed, the RSTRENG calculation was quite conservative in this
case because for some unknown reason, the failure pressure of the
Index 133 flaw was much higher than the two single 12-inch flaws
(Index Nos. 127 and 128). This may have had something to do with
its being cut out and welded into another vessel.

Effects of Interactions of Pits

Potential interactions between pits were explored by

means of the arrays of pits shown in Figure 2. The arrays were
made up of single cylindrical pits 1.5 inches in diameter and
0.291 inch deep (60 percent through the wall thickness). The
criterion for interaction is whether or not any array exhibits a
failure pressure less than that of a single pit. Because some of
the pits were arrayed diagonally and circumferentially, it is
essential to remember that these were tests of pressure vessels.
Hence, the longitudinal stress applied to the defects, while
being much like that experienced in a normal buried pipeline, was
not nearly as important as the hoop stress. Therefore, one
cannot necessarily extrapolate the results of the tests to cases
 22

where a pipeline might be subjected to severe longitudinal

stress.
The results of the bursts tests on the pitted sample
shown in Figure 2 are summarized in Table 2. For convenience,
they are repeated as follows.

Actual Failure
Index No. Description Status Pressure, psig Interaction

134 3 pits separated No failure >3176 No

circumferentially by 1.5 inch
135 Single pit No failure >3176
136 3 pits arrayed Leaked 3176 No
circumferentially touching on
another
137 3 pits arrayed diagonally, Leaked 2944 Yes
touching one another
138 3 pits arrayed No failure
circumferentially, separated
by 1/2 inch
139 3 pits arrayed diagonally, Leaked 3016 Yes
separated by 1/2 inch
140 3 pits arrayed longitudinally, Cut and retested 3248 Undetermined
separated by 1/2 inch
141 3 pits arrayed longitudinally, Cut and retested 3451 No
separated by 1.5 inch
142 3 pits arrayed longitudinally, Ruptured 2726 Yes
touching one another

Index 142 failed first and was cut out. Index Nos. 140
and 141 were cut and retested. Index 137 failed second and was
patched. Its failure pressure was seven percent above that of
Index 142. Index 139 failed third at a pressure level 2.3
percent above that of Index 137. Finally, Index 136 failed at a
pressure level five percent above that of Index 139. A l l
failures except that of Index 142 were leaks. No pressure
 23

reversals occurred in this series and the effects of prior cycles

on the Index 136 and Index 137 results can probably be
discounted. In the case of Index 139, which failed at a level
only 2.3 percent above that of the previous cycle, the effect of
the previous cycle, if any, was probably not very significant.
One difficulty with this series of tests is that the
failure pressure level of a single pit was not clearly
established. The authors of Reference 7 suggest that the Index
Nos. 136, 137, and 142 (all comprised of touching pits) show
interaction, that Index Nos. 138, 139, and 140 show slight
interaction, and that Index Nos. 134 and 141 show no interaction.
Our assessment is that the results of tests in Index Nos. 137,
139, and 142 do indeed suggest interaction. For the rest, it is
our opinion that interaction is not clearly evident. For one
thing, the cutting, rewelding, and retesting of Index Nos. 140
and 141 raises significant questions about the validity of their
results. In the case of Index 136, we find the case for
interaction to be unconvincing and consider its result to be a
one-pit failure pressure. There is no reason to believe that
interaction occurred in this array as will be seen in a
subsequent discussion of other work on arrays of pits.
Therefore, if we assume the Index 136 result to be that of a
single pit, then the arrays of Index Nos. 137, 139, and 141, all
of which had lower failure pressures, clearly exhibited
interaction. By the same criterion, none of the others exhibited
interaction; however, interpretations of the results of Index
Nos. 140 and 141 are still obscured by the cutting and welding
into another specimen.
The implications of the behaviors of the arrays of pits
with respect to the applicability of RSTRENG is examined in
Table 2. For convenience, the relevant comparisons are shown
below.
 24

RSTRENG
Actual Failure Predicted Ratio of Predicted to
Index No. Description Pressure, psig Pressure, psig Actual
134 Circumferential array >3176 2849 <0.90 Treated as single
separated by 1.5 inch pit
135 Single pit >3176 2849 <0.90 Treated as single
pit
136 Circumferential array, 3176 2849 0.90 Treated as single pit
touching one another
137 Diagonal array (45 2944 2475 0.84 Based on projected
degrees to axis) touching length
one another
2723 0.92 Based on spiral
angle factor of
Reference 6
2849 0.97 Treated as single pit
138 Circumferential array >3176 2849 <0.90 Treated as single
separated by 1/2 inch pit
139 Diagonal array (45 3016 2337 0.77 Based on projected
degrees to axis) length
separated by 1/2 inch
2571 0.85 Based on spiral
angle factor of
Reference 6
2849 0.94 Treated as single pit
140 Longitudinal array 3248 2561 0.79 Treated as 3 pits
separated by 1/2 inch separated by 1/2 inch
2849 0.88 Treated as single pit
141 Longitudinal array 3451 2623 0.76 Treated as 3 pits
separated by 1/5 inch separated by 1.5 inch
2849 0.83 Treated as single pit
142 Longitudinal array 2726 2520 0.92 as 3 touching pits
touching one another

The applicability of RSTRENG suggested by these

comparisons is as follows:
 25

• Obviously, single pits should be treated as single

pits. In this case, ten increments were used to define
the pits. This degree of precision may not have been
necessary.
• For pressure vessel and normal buried pipeline
loadings, the deepest, longest pit in a circumferential
array is the one to analyze.
• For diagonally arrayed pits, the treatment of the
projected length always gives a conservative answer,
probably too conservative. Applying the spiral-angle
factor suggested in Reference 6 gives a better answer.
In the case of these 45-degree orientations, the
spiral-angle factor is 1.1. The RSTRENG value for the
projected length should be multiplied by it to get the
predicted failure pressure. Finally, even if one uses
RSTRENG to analyze only the deepest, longest pit on a
45-degree angle, the predicted pressure would probably
still be acceptable
• For longitudinal arrays, whenever the pits are touching
or separated by less than one wall thickness, prudence
dictates analyzing the entire defect. Isolated pits
separated from one another by more than one wall
thickness can be analyzed independently.

Effects of Interaction of Areas of

Missing Metal (Type II Interaction)

Potential interactions between patches of corrosion

comprised of many adjacent pits or continuous metal loss over
wide areas were investigated through the configurations shown in
Figure 3. The defects in Figure 3 are comprised of pairs of
standard units, one being a 6-inch-long, 4.5-inch-wide patch of
metal loss having a maximum depth of 0.194 inch (40 percent of
 26

the wall thickness) and the other being the 1.5-inch-diameter

cylindrical pit of 0.291-inch depth (60 percent of the wall
thickness) used in the previously described series of tests.
The results of the burst tests on these flaws are
summarized in Table 2, and they are listed again below for
convenience.

Actual Failure
Index No. Description status Pressure, psig Interaction
143 Two patches separated Cut and retested 2654 Yes
longitudinally by 1.5 inch
144 Two patches touching Ruptured 2567 Yes
longitudinally
145 Two patches separated No failure >2567 Undetermined
longitudinally by 3 inches
146 Two patches touching Cut and retested 3118 No
circumferentially
147 Two patches separated Rupture 2944 No
circumferentially by 1.5 inch
148 Single patch No failure >2944 ---
149 Patch and pit separated No failure >2944 No
longitudinally by 1.5 inch
150 Patch and pit separated Rupture 2770 Yes
longitudinally by 1/2 inch
151 Patch and pit touching Rupture 2596 Yes
longitudinally

Index 144 failed first so Index Nos. 143 and 146 were
cut, rewelded as a segments into another vessel, and retested.
Index 145 was not tested to failure. Index 150 failed second at
a pressure level 7.3 percent above that of Index 144. Thus, its
behavior was probably not influenced by the previous cycle of
pressure. Index 151 failed third but at a pressure level below
that of Index 150. Index 151 failed as 6.3 percent pressure
 27

reversal; hence, its failure pressure clearly was influenced by

one or both previous cycles. Index 147 failed fourth at a
pressure level 5.9 percent above the highest previous cycle.
Index Nos. 148 and 149 survived the pressure level at which
Index 147 failed but were not further tested.
As in the case of the pits, the failure pressure of the
single patch, Index 148, was not established. However, it is our
opinion that the failure pressure of Index 147 establishes the
single patch failure pressure. We doubt that the two
circumferentially separated patches interacted. On this basis,
we postulate that Index Nos. 143, 144, 150, and 151 interacted
significantly. In the case of Index 143, we believe this to be
so even though it was cut and rewelded because its failure
pressure is quite a bit below that of the single patch. We
believe that the pairs of defects of Index Nos. 146 and 149 did
not interact. Whether or not interaction occurred for the pair
at Index 145 cannot be determined because it was not tested to a
high enough pressure level.
The implications of the behavior of these patch-and-pit
combinations to the applicability of RSTRENG is assessed below.
The data of Table 2 are repeated here for convenience.

RSTRENG
Actual Failure Predicted Pressure, Ratio Predicted to
Index No. Description Pressure, psig psig Actual
143 Two patches separated 2654 2515 0.95 Treated as
longitudinally by 1.5 inch single 6 inch
2232 0.84 Treated as
two 6-inch defects
separated 1.5
inches
longitudinally
144 Two patches touching 2567 2274 0.89 Treated as
longitudinally touching 6-inch
defects
 28

RSTRENG
Actual Failure Predicted Pressure, Ratio Predicted to
Index No. Description Pressure, psig psig Actual
145 Two patches separated >2567 2515 <0.98 Treated as
longitudinally by 3 inches single 6 inch
146 Two patches touching 3118 2515 0.81 Treated as
circumferentially single 6 inch
147 Two patches separated 2944 2515 0.85 Treated as
circumferentially by 1.5 single 6 inch
inch
148 Single 6-inch patch >2944 2515 <0.85 Treated as
single 6 inch

149 Patch plus pit separated by >2944 2370 <0.81 Treated as

1.5 inch patch and pit
separated by 1.5
inch
150 Patch plus pit separated by 2770 2371 0.86 Treated as
1/2 inch patch and pit
separated by 1/2
inch
151 Patch plus pit touching 2770(a) 2343 0.85 Treated as
touching patch and
pit

(a) Considering the first cycle which the defect survived before failing at 2,596 psig as a 6.3 percent
pressure reversal.

The applicability of RSTRENG suggested by these

comparisons is as follows:
• Patches or patches and pits separated longitudinally
should be assumed to interact if the separation
distance is less than 1.5 inch. In such a case,
RSTRENG should be used over the whole length taking
into account the areas where there is a return to full
wall thickness.
• For patches or patches and pits separated or touching
circumferentially where the lengths of the individually
flaws do not overlap to form a projected length longer
 29

than any individual length, each anomaly should be

considered separately on the basis of its projected
profile of missing metal.

Behavior of Short, Deep Defects Lying Within

Longer, Shallower Defects (Type III Interaction)

The behavior of compound defects where a shorter,

deeper defect exists within a longer, shallower one were examined
on the basis of the defects shown in Figure 4. This type of
interaction is referred to as "Type-III" interaction. Three of
these were patches, one plain patch (6 inches long by 0.220 inch
deep) and two with pits. The test vessel also contained three
long, narrow defects of the type discussed earlier. One of these
was a 3-inch-long uniform depth slot (0.194-inch) and the other
two were compound slots comprised of a 3-inch-long, 0.194-inch-
deep slot centered in a longer, shallower slot. The results of
the tests on these compound flaws are presented in Table 2 and
are listed again below for convenience.
The occurrences of pressure reversals at the Index 153
and Index 154 anomalies are not too surprising, cycles of
pressure to near-failure levels have been shown to cause such
occurrences.(14) The size of the reversal in the case of
Index 154 (39 percent) is alarmingly large. In two cycles, one
from zero to 2,857 psig and a second from zero to 2,828 psig,
this initially 60-percent-through defect grew to become a leak at
1,740 psig on the next cycle. One possibility, of course, is
that the leak actually developed on the second cycle but was not
noticed because another flaw failed. However, a s r e c o g n i z e d a n d
demonstrated in Reference 14, large pressure reversals resulting
in leaks are possible. This possibility would seem to have
implications for pipeline operators using RSTRENG (or any other
method of evaluation). However, it should be remembered that
 30

such a defect will not pass if its failure pressure as predicted

by RSTRENG is less than 100 percent of SMYS. So at the MAOP (72
percent of SMYS), the defect is not at or anywhere near its
failure pressure. Such a defect would not be expected to fail at
72 percent of SMYS unless or until it was pressured to a near-
failure pressure level (i.e., 100 percent of SMYS). Therefore,
we do not believe that any special warning needs to be raised
regarding the effect of pressure reversals on RSTRENG
calculations.

Actual Failure
Index No. Description Pressure, psig Status

152 Pit centered in patch 2857 Ruptured

153 Plain patch 2828 Ruptured as 1 percent
pressure reversal
154 Two pits within patch 1740 Leaked as 39 percent
pressure reversal
155 3-inch slot >3031 No failure
156 Deeper 3-inch slot within >3031 No failure
shallower 12-inch slot
157 Deeper 3-inch slot within 3031 Ruptured
shallower 24-inch slot

I f t h e p r e s s u r e r e v e r s a l s are discounted, RSTRENG gave

reliable predictions of the results of the first cycle of
pressurization as shown below and in Table 2. In all cases,
RSTRENG is applied to the worst-case projected profile.

RSTRENG
Actual Failure Predicted Pressure, Ratio of Predicted
Index No. Description Pressure, psig psig to Actual
152 Pit centered in patch 2857 2343 0.82
153 Plain patch 2857(a) 2426 0.85
 31

RSTRENG
Actual Failure Predicted Pressure, Ratio of Predicted
Index No. Description Pressure, psig psig to Actual
154 Two pits within patch 2857(a) 2306 0.81
155 3-inch slot >3031 2785 <0.92

156 Deeper 3-inch slot within >3031 2665 <0.88

shallower 12/inch slot
157 Deeper 3-inch slot within 3031 2665 0.88
shallower 24-inch slot

(a) Pressure reversal discounted.

Discussion of the British Gas Tests

The results of these tests provide additional insight

into the behaviors of long flaws, isolated and interacting pits,
patches of corrosion, and deep defects within shallow defects.
The results have also refocussed attention on pressure reversals
especially as they apply to deep corrosion pits. As far as the
use and validity of RSTRENG are concerned, the following points
are suggested by these results.
• For very long corroded areas, an RSTRENG analysis can
be confined to a one-diameter length or a length of 20
inches, whichever is greater, so long as the deepest
pit is included in the region analyzed.
• F o r T y p e I I i n t e r a c t i o n , RSTRENG can be used to analyze
the individual flaws and the overall combination. The
lowest value resulting from the various configurations
can be used.
• For Type I interaction, one can always safely consider
the projected length of the combined flaws in an
RSTRENG analysis when the separation is less than 6t.
 32

• Single pits separated by a wall thickness or more of

sound pipe do not interact s i g n i f i c a n t l y .
• The axial extent and maximum depth of patches of
corrosion appear to remain the parameters of importance
in assessing remaining strength. Circumferential
extent appears to play no significant role.
• For Type III interaction, a flaw within a flaw, RSTRENG
provides adequate predictions based on the worst-case
projected area.
• As was known previously, pressure reversals are a
factor in the behavior of deep pits. This does not
negate the validity of an RSTRENG calculation, however.
Any pit that is accepted via an RSTRENG calculation is
not likely to be affected by a large pressure reversal
unless it has recently been subjected to a high-
pressure hydrostatic test (i.e., at or near 100 percent
of SMYS).

Analysis of Full-Scale Tests Involving

Combined Pressure and Bending Loads on Corroded Pipe

Results

A r e c e n t paper by Kim(8) p r o v i d e s a u n i q u e s e t o f d a t a
on the behavior of corroded pipe. The results of tests involving
two extremes of longitudinal stress in conjunction with internal
pressure are presented. The results and their meaning with
regard to the use of RSTRENG are discussed herein.
Five experiments were carried out at Southwest Research
Institute on samples of 48-inch O.D. by 0.480-inch wall thickness
API 5L Grade X65 line pipe. Metal-loss-simulating corrosion was
created in each specimen by machining away 25 to 50 percent of
the wall thickness over rectangular areas of various sizes. Two
 33

identical areas of metal loss were created on each specimen, 180

degrees around the circumference from one another. In this
manner, one was subjected to compressive longitudinal stress and
the other to tensile longitudinal stress when the specimen was
subjected to a bending moment which placed both areas at extreme
distances from the neutral axis. The specimens were fabricated
with end caps and tested to failure with various combinations of
applied bending moment and internal pressure. The specific
metal-loss dimensions and loading conditions are described in
Table 3. Because the data have been added to the A.G.A./PRCI
Database of corroded pipe tests, the tests are numbered Index
No. 158-162. The RSTRENG calculations for these tests are
presented in Appendix B.
The test designated Index 158 involved three steps:
pressurization to 950 psig with no bending moment, bending until
an axial strain level of 0.4 percent was reached on the tension
side of the neutral axis, and pressurization to failure. The
specimen failed by axial fracture of the metal-loss area on the
compression side. The pressure level at failure was 1,480 psig.
The corroded area on the compressive side was observed to have
bulged outward significantly prior to failure as though a buckle
may have been forming.
The test designated Index 159 was pressurized to a
level of 950 psig with no bending applied. Bending was then
applied in an increasing amount until a failure occurred. The
mode of failure was a circumferential fracture on the tension
side but a buckle had begun to appear on the compressive side.
The test designated Index 160 involved pressurization
to 950 psig followed by application of bending to an axial strain
level of 0.2 percent on the tension side of the neutral axis.
The internal pressure was then increased, and failure occurred
when the pressure reached 980 psig. The mode of failure was an
axial rupture on the compression side of the neutral axis. As
 34

with Index 158, a buckle appeared to be forming prior to the

rupture.
In the test designated as Index 161, the specimen was
pressurized first to 800 psig. At this level, circumferential
strain became quite high and pressurizing was stopped. Bending
was then applied until an axial tensile strain of 0.2 percent was
obtained. Pressurization was resumed while this bending load was
held constant, and the specimen failed at 840 psig exhibiting an
axial rupture on the compressive side. The corroded area of the
compressive side appeared to have started to buckle prior to
failure.
The specimen designated as Index 162 was identical to
that designated as Index 159, and it was loaded in the same
manner. The results were also similar. Under increasing bending
displacement at a constant internal pressure of 950 psig, a
circumferential rupture occurred in the metal-loss area on the
tension side. Prior to the rupture, a buckle was beginning to
develop on the compression side.

Discussion

Of immediate interest are the burst pressures RSTRENG

predicts for these specimens. As seen in Table 3, the burst
pressures in Index Nos. 158, 160, and 161 are reasonably close to
the RSTRENG-predicted values. These are the specimens that
failed by axial ruptures at the metal-loss areas on the
compression side of the neutral axis. For these cases, the
RSTRENG burst-pressure predictions were conservative; the RSTRENG
safe pressures would have indeed been safe. For Index Nos. 159
and 162, however, t h e f a i l u r e s o c c u r r e d a t p r e s s u r e l e v e l s
significantly below the RSTRENG-predicted burst pressures because
the axial stress and circumferential extent of the metal loss
controlled the outcome of the tests.
 35

The above-described results reveal one of the

limitations on RSTRENG and on the B31G methodology in general.
The methodology considers only the effects of the longitudinal
extent and depth of penetration of corrosion through the wall
thickness on the pressure-carrying capacity of a pipeline not
subjected to extreme axial or bending loads. When large axial or
bending loads are applied and when the circumferential extent of
the metal loss is significant, especially when it greatly exceeds
the axial extent of the metal loss, the failure condition and/or
mode of failure may not be predictable by means of the B31G
methodology alone.
How serious is this shortcoming of the B31G
methodology? It is our opinion that this shortcoming must be
kept in mind, but that can be expected to have little or no
impact on the general applicability of the methodology. Consider
in particular that the failures of the specimens designated
Index 159 and Index 162 were preceded by the application of
bending moments large enough to cause near-collapse conditions in
the metal-loss area on the compression side of the neutral axis.
Such conditions are accompanied by very large deflections that
can be expected to occur only rarely. Circumstances which come
to mind would be a pipeline in a landslide or a subsidence area,
a pipeline subjected to extreme compressive load which then
becomes able to move as when excavated or when the soil
liquifies, or a pipeline which is excavated in a manner that
results in a large unsupported span.
A second factor to consider with respect to Index 159
and Index 162 is the ratio of circumferential to axial extent of
the defect (30 inches to 6 inches). Suppose the ratio had been
30 inches to 12 inches. In this case, the results may well have
been the same, but the burst pressure predicted by RSTRENG would
have been 972 psig. This overestimates the actual pressure level
at failure by only about 2 percent. This calculation suggests
 36

that the observed behavior (i.e., circumferential rupture) was

strongly influenced by the predominantly circumferential nature
of the defect. The point is that the special behavior observed
in the tests designated as Index Nos. 159 and 162 not only
requires extreme flexural deflections, it also can be expected to
occur only when the defect is vastly larger in the
circumferential direction than it is in the longitudinal
direction. When this is not the case, it is likely that the
metal loss will not pass the B31G criterion even if its
circumferential extent governs its behavior. It is our opinion
that the rare cases where the B31G methodology may not apply as
typified by Index Nos. 159 and 162 are easily recognizable in
terms of both observable metal-loss geometry and the special
loading conditions that are not likely to go unnoticed.
Finally, it is worth noting that in the cases of Index
Nos. 158, 160, and 161 the RSTRENG methodology works in spite of
the fact that the failures took place under the influence of
near-collapse axial compressive stresses. The longitudinal
compressive stress is significant in the sense that all three
failures initiated on the compressive side even though the hoop
stress on the defect was the same on the tensile stress side.
The effect is likely the result of yielding beginning earlier
compressive side than on the tensile side. However, the effect
a p p a r e n t l y w a s n o t g r e a t enough to suggest that RSTRENG should
not be used.

Experimental and Analytical Studies

of the Behavior of Isolated and Closely
Spaced Corrosion Pits in Pressurized Pipe

Some timely and useful work on the behavior of

corrosion pits has recently been published by Chouchaoui and Pick
of the University of Waterloo(9-11). The authors have conducted
burst tests on pipes containing electrochemically machined pits.
 37

The results of the first two references shed some light on the
degree of interaction that can be expected from arrays of single
pits in close proximity. Circumferential, longitudinal, and 45-
degree-spiral arrays were addressed. Burst pressures from the
experiments are compared with predicted burst pressures based on
the B31G criterion, RSTRENG, and elastic-plastic finite-element
analyses. Generally, the results tend to confirm the
conservativeness of the B31G criterion and the improvement
afforded by RSTRENG. Some well-known deficiencies of RSTRENG
(i.e., its inability to deal with pit interaction in all cases,
its inability to deal directly with orientations other than
longitudinal) are reiterated. In addition, a new flag is raised
with respect to RSTRENG's ability to account for different axial
stress states. In this work, two different carefully controlled
axial stress states clearly produced different results for
identical pit geometries. The RSTRENG predictions were still
adequate, but it is suggested that larger compressive axial
loading might produce burst pressures below the RSTRENG-predicted
levels. The authors also show that elastic-plastic finite-
element analyses can be used to give good estimates of burst
pressures although this method may require accounting for
anisotropic behavior when different axial stress states are
examined.
The third reference deals with the behavior of a
shorter, deeper pit located within a longer, shallower corroded
area. In this study, burst tests of pipes with electrochemically
machined defects were again used to generate experimental results
and both RSTRENG and finite-element analyses were used to predict
burst pressures.
 38

Summary of the Results

All of the burst tests were performed on 12.75-inch

(324-mm) O.D. by 0.250-inch (6.35-mm) wall thickness X46 pipe.
The electrochemically machined defects were shaped generally as
shown in Figure 5 (single pits) or Figure 6 (pit within a
corroded region). The results of the burst tests on
circumferential and longitudinal arrays are presented in Table 4,
and the results of the burst tests on pits within corroded
regions are presented in Table 5. The results of the spirally
arrayed pits were not analyzed in detail because the pit sizes
and material properties were not presented in the available
references.
Most, but not all of the arrays of pits produced
results with which RSTRENG comparisons could be made. A few of
the arrays where the spacing between pits was too great for
interaction to occur were not analyzed. For those cases which
were analyzed, index numbers were created in order to place the
comparisons in the A.G.A./PRCI Database of Corroded Pipe Tests.
The appropriate RSTRENG calculations are shown in Appendix B.

Circumferentially Arrayed Pits

The layouts of the circumferentially arrayed pits are

shown in Figure 7a, and the results of the tests are shown in the
top portion of Table 3. The Waterloo test numbers (e.g., S1cc)
denote single (S) pits, sequence (1), orientation (c for
circumferential), and end loading condition (c for closed ended
or o for open ended). The axial load in the closed-ended tests
comes from pressure acting on the end caps. The open-ended tests
were conducted in a special fixture which prevents any axial
loading. Thus, the "c" and "o" tests represent two distinctly
different biaxial stress states.
 39

The effect of the stress state on a single pit is

illustrated by the comparison between Tests S1cc and S1co (Index
Nos. 163 and 167 in Table 4). Note that the single pit of
Test S1co, the relatively shallower pit in a relatively stronger
material, exhibited a lower failure pressure than the single pit
in Test S1cc. Note also that RSTRENG substantially
underestimated the strength of the pit in Test S1cc but gave a
slight overestimate of the strength of the pit in Test S1co.
These results and others in the circumferential series reveal a
trend toward higher failure pressures for the isolated pits in
the case of closed-ended specimens. The authors of Reference 9
point out that the increased strengths of the closed-ended cases
are probably the result of delayed yielding caused by the biaxial
stress state.
The solid-colored circles in Figure 7a denote the
locations of the leaks. In two cases (Tests S2cc and S4co), the
failures occurred in locations other than the test pits, so the
failure pressure of the most critical array is not known.
However, it is clear from the six results where failures did
occur in the test pits, that circumferential separation is not a
significant factor under the loading conditions applied.
Therefore, it is correct to apply RSTRENG as was done in this
series to the worst individual pit. I n g e n e r a l , RSTRENG gave
reasonable predictions of the results even though it cannot
adequately account for the biaxial stress state. The question
arises, of course, as to how far off RSTRENG might be if the same
tests were conducted with a significant axial compressive load.

Longitudinally Arrayed Pits

The layouts of the longitudinally arrayed pits are

shown in Figure 7b, and the results of the tests are shown in the
bottom portion of Table 3. Without having seen the results, we
 40

would have been inclined to believe that longitudinally oriented

pits such as these would not interact if separated by as much as
one wall thickness (1wt) and that they would interact if
touching. The results of Tests S21c and S31c suggest that
interaction occurs at a spacing of 1wt and the results of
Tests S41c and S31o suggest that interaction does not occur at a
spacing of 2 wall thicknesses. These conclusions are based more
on the modes of failure than on the failure pressures because one
cannot readily distinguish the order of failure pressures from
possible scatter. The fact that the three-pit, one-wall-
thickness-spaced arrays of Tests S21c and S31c failed as ruptures
is strong evidence that interaction occurred. In contrast, the
fact that leaks occurred in single pits in Tests S41c and S31o
strongly suggests that the interaction of the 2 wt-spaced arrays
was negligible.
RSTRENG appeared to give adequate predictions of the
failure pressures as long as it was applied to the whole array in
the cases of the 1wt-spaced arrays. As in the case of the
circumferentially arrayed pits, the effect of differences in end
loading is evident in the longitudinally arrayed single pits. It
appears that there is no excess conservatism when RSTRENG is
applied to the cases where the end loading is zero. Again, this
raises a question regarding the adequacy of RSTRENG if a
substantial axial compressive load where to exist.

Spirally Arrayed Pits

As noted above, the critical dimensions and material

properties necessary for the analysis of the spirally arrayed
pits described in Reference 10 were not available at the time of
this report. However, one valuable conclusion is drawn by the
authors of Reference 10, and it can be verified from Figure 7c
which was taken from Reference 10. In particular, it can be seen
 41

from Figure 7c that all of the failures in the spirally arrayed

pits involved the leaking of the single largest pit.
Interaction, at least for the loading cases shown, was prevented
by the curvature of the pipe. Hence, it is appropriate to apply
RSTRENG to the dimensions of the most severe axial profile across
a 45-degree-spiral array rather than to project the entire array
onto the axial plane.
The results of the tests of spirally arrayed pits
seemed to be less sensitive to the state of axial stress. This
fact can be confirmed from Figure 3 of Reference 10. The reason
for this is not readily apparent to us.

Behavior of Pits Within a Corroded

Region (Type III Interaction)

The results of the Waterloo tests on the complex pit

shapes shown in Figure 6 are presented in Table 5. Note that
RSTRENG cannot distinguish any difference between Defect 1 and
Defect 2 in Figure 6 if the individual pit dimensions are the
same. In fact, the RSTRENG answer for Defect 3 would be the same
or almost the same as for Defects 1 and 2.
Two important conclusions regarding RSTRENG can be
drawn from the results shown in Table 5. First, the RSTRENG
predictions are quite adequate in all six cases. Secondly, the
effect of end loading in this series is not very apparent if it
is apparent at all. Consequently, w e b e l i e v e t h a t t h e e f f e c t o f
end loading may have less influence on the behavior of long
corroded areas than it appears to have on isolated, single pits.
This seems to have been the case with the extreme compressive
stresses in the previously described bending tests also.
 42

Discussion

The Waterloo experiments provide valuable insight into

the behavior of corrosion pits, especially that of isolated
single In particular, they have shown that the biaxial
pits.
stress state has a significant effect on single pits. It is less
clear whether this effect extends to longer, more axially
oriented metal-loss anomalies.
From the results, it appears that the worst single pit
of a circumferential array should be analyzed with RSTRENG at
least for cases not involving extreme axial loads. The
circumferential extent is apparently not significant when the
only load is from pressure. For longitudinal arrays of pits, the
results indicate that RSTRENG should be applied to all pits
having a separation of 1 wt or less. The experimental results
also suggest that the most severe axial profile across a spiral
array rather than the projected profile of the entire array gives
the most appropriate dimensions for analyzing the effect of the
anomaly, a t l e a s t i n t h e a b s e n c e o f e x t r e m e a x i a l l o a d i n g s . The
analytical work (finite-element analyses) done by the Waterloo
team undoubtedly adds to the understanding of pit interaction.
However, assessment of the analytical work is beyond the scope of
Project PR 218-9304.

ASSESSMENT OF BURST TESTS, HYDROSTATIC TEST FAILURES

AND SERVICE FAILURES INVOLVING CORRODED LINE PIPE

Analysis of Burst Tests on Samples of Corroded 24-inch O.D.

by 0.312-inch WT. Grade B Seamless Line Pipe and 12.75-inch O.D.
by 0.188-inch WT. Grade X52 ERW Line Pipe

Recently, a series of a formerly proprietary reports of

burst tests was made available for analysis on Project
 43

PR 218-9304. The reports involved tests conducted nearly

25 years ago on samples of 24-inch OD and 12.75-inch OD line pipe
which had been removed from an operating main pipeline and an
operating gathering line, respectively. The pipeline operator

who conducted the tests also donated several samples of the

24-inch OD pipe to the Pipeline Research Committee (PRC) in 1971
along with two of the results obtained in the operator's own test
facility. With PRC funding, B a t t e l l e r e s e a r c h e r s c o n d u c t e d t e s t s
on this pipe in 1971, obtaining four test results. These four

results along with the two provided by the operator were part of
the original data base of 47 corroded pipe burst tests used to
validate the original B31G equation (Index Nos. 39 through 44).
The scope of the operator's own test program was not known at the
time, however, because the data were not released until they were
recently given over to Project PR 218-9307. For several reasons
these data proved to be interesting and useful. They are
analyzed below.
The work was carried out in three phases, and a
separate report was made for each phase. In discussing the
results, however, it is more efficient to discuss the individual
tests without reference to the various phases of the work. Two
types of pipe material were involved, 24-inch OD by 0.312-inch wt
Grade B seamless line pipe and 12.75-inch OD by 0.188-inch wt
Grade X52 ERW line pipe. The descriptions of these materials are
summarized in Table 6. The materials listed in Table 6 include
those donated to the NG-18 project by the pipeline operator
(Index Numbers 38 through 42).
Two things stand out in Table 6. First, the wall
thicknesses of the 24-inch-OD seamless material varied widely.
The measurements, b a s e d o n a v e r a g e s o f 6 t o 8 v a l u e s t a k e n a t
both ends of each sample ranged from 0.307-inch to 0.444-inch for
 44

the nominally 0.312-inch-thick material. Seamless pipe thickness

is known to be more variable than pipe made from skelp, but these
values ranged more widely than one might expect. Secondly, the
yield and ultimate strengths as determined by flattened
transverse tensile specimens were unusually high for a Grade B
material (35,000 psi minimum yield strength and 60,000 psi
minimum ultimate strength). When tested as pressure vessels,
however, the 24-inch-OD samples tended to exhibit yielding at
hoop stress levels ranging from 35,800 psi to 49,250 psi. For
those samples tested in the NG-18 program (Index Numbers 39
through 42), the pressure vessel yield strengths were determined
at 0.2 percent offset circumferential strain based on pressure-
volume plots. The pressure yield strengths of the 24-inch-OD
samples were lower in every case than the values determined by
flattened tensile tests. These results are contrary to
expectations; pressure vessel yield strengths are generally 7 or
8 percent higher than uniaxial tensile tests for line pipe
materials. A possible explanation for the relatively low
pressure yield strengths may be the inherent variations in the
wall thickness in combination with the large amounts of
corrosion-caused metal loss. These circumstances would tend to
result in a significant amount of thin material being stressed to
its yield strength at a lower pressure level. Thus, the unusual
properties of this material may or may not be related to the fact
that it was manufactured in 1930 or earlier when the manufacture
of large-diameter seamless pipe was just getting under way.
Whatever the explanation, the anomalous behavior may have had
some bearing on the subsequent comparisons of the results to
RSTRENG predictions as will be discussed.
Prior to testing the corroded samples, wax negative
molds and positive plaster casts of the corrosion pits were made
 45

to preserve the pretest geometry of the corrosion. At the time

of these tests similar casts were being made of corroded pipe
samples being tested as part of the NG-18 program. After the
tests, the pretest geometry of each failure location could then
be reviewed in detail without any influence from distortion or
yielding that might have occurred during the pressure tests.
The pressure tests of the samples provided 22 leaks and
ruptures of which the 18 shown in Table 7 were analyzed and
compared to RSTRENG predictions. Those not analyzed were left
out either because the reports did not provide their remaining
thickness profiles, or in one case (Pipe Number 2) because the
first failure occurred at a prior repair patch rendering the rest
of the corroded portion of the sample useless. The profiles used
for the analyses of each are shown in Figures 8 through 16. The
profiles of the anomalies on Samples 1, 3, and 5 (Figures 8
through 11, and 15 and 16) are based on measurements taken after
the failures since plaster casts were not available for these
locations. The profiles for Samples 3A and 4 (Figures 12 through
14) are based on plaster casts.
It is noted that those of Sample 4 shown in Figure 14
exhibit "before" and "after" profiles. The " a f t e r " profile
represents outward bulging of the corrosion weakened areas. This
phenomenon could have a significant effect on an RSTRENG
evaluation of an anomaly because any bulging would tend to
conceal the true depth of the pits. The reports from which these
data were taken noted the fact that the bulging occurred and that
dimpling of the ID surface beneath the pits took place prior to
failure of the pits. It was not clear from the reports how the
" a f t e r " measurements were made for Failures 2 and 5 which were
supposed to have occurred as ruptures. Perhaps these
measurements were made after only the first cycle which resulted
 46

in a leak of Area D-8a. In any case, n o q u a n t i t a t i v e a n a l y s i s o f

these examples of "bulging" was undertaken by the author of the
original report or as a part of Project PR 218-9304. However,
these examples called attention to the phenomenon and led to a
more quantitative assessment in a later test.
The RSTRENG predictions of failure pressures based on
the profiles shown in Figures 8 through 16 are compared to the
actual failure pressures in Table 7 on the basis of the
parameter, Pp/Pa. It is noted that Anomaly 5-1, in Figure 15 was
adjacent to an old, existing puddle weld. However, the failure
seemed to have taken place in a subsequently corroded area. So
the result was included in the analysis.
The comparisons between RSTRENG-predicted failure
pressures and actual failure pressures exhibit a pattern that is
not unexpected relative to past test results on corroded pipe.
The predictions tend to underestimate the failure pressures for
the leaks. The Pp/Pa ratios for 11 leaks ranged from 0.926 to
1.430 with 9 of the 11 being 1.0 or more. This trend is somewhat
more toward the nonconservative side than the overall trend of
the data for all leaks (85 cases) shown in Table 10 (Ratio of
Case 1 to Actual), but the highest Pp/Pa value in Table 7 (1.430)
is not the highest in the database (the ratio for Index 9 was
1.448).
The Pp/Pa predictions in Table 7 for the 7 ruptures
range from 0.773 to 1.055. These data lie within the overall
range for 83 ruptures shown in Table 9 (they ranged from 0.620
for Index 40 to 1.058 for Index 48). The low value of Pp/Pa for
Test 5-2 in Table 7 (0.773) may be the result of the profile not
accounting for the high levels of remaining thickness between
pits (only the depths of the pits were provided in Figure 16).
 47

Aside from this Pp/Pa value, the next lowest value was 0.964 for
Test 3-3.
The Pp/Pa values for the 24-inch OD material seem to be
on the high side even though they are not outside the range of
the overall database. To this one may add the fact that one of
the PRCI-sponsored tests on this material (Index 41) exhibited a
leak with a Pp/Pa ratio of 1.397. The d/t values of the 11 leaks
in Table 7 ranged from 0.52 to 0.88 with only two of the
anomalies involving d/t values in excess of 0.80, so the
relatively high Pp/Pa ratios are not explainable in terms of
excessive depth of the pits.
Another possible reason for the Pp/Pa trend might be
related to the fact that the pressure yield strength of the
material was less than the flattened tensile yield strength.
Flattened tensile yield strength is used and should be used to
calculate the flow stress of the material for RSTRENG
calculations. To do anything else would be both inconvenient and
inconsistent. Nevertheless, as the results presented in the
three reports and the NG-18 result for Index 41 showed, the leaks
occurred fairly consistently at about the same pressure levels as
yielding was taking place. If the unusually low pressure yield
strength of the material is a real material property (and this
was not proven), the relatively high Pp/Pa ratios could be the
result of an abnormally low "effective" flow stress. Alternative
Pp/Pa ratios calculated on the basis of such an "effective" flow
stress would tend to fall more nearly in the middle of the
database.
Another significant result from among the data of
Table 7 is the 17 percent pressure reversal exhibited by
Test 3A-2. As explained previously in conjunction with the
 48

Index 154 anomaly (British Gas test of a machined compound pit

described on Page 29), the failure of a pit by leaking as a large
pressure reversal is not all that unusual. It does not negate
the value of RSTRENG as a means of evaluating the remaining
strength of corroded pipe, because defects which are found by
RSTRENG to be anywhere near their failure pressure at the MAOP
would be removed or repaired. In addition, under the normal
circumstances of the use of RSTRENG, the prior cycle of pressure
which causes the pressure reversal does not exist.
One final useful piece of information found in these
three reports was that the pipeline operator was able to make
repeated test cycles in a number of cases because of successful
deposited-weld-metal repairs to some of the test leaks. The

reports note that the technique used involved SMAW weld deposits
with low hydrogen electrodes. These repairs were found to
withstand subsequent pressure cycles whereas old, existing puddle
welds made in the field with non low-hydrogen electrodes tended
to cause early leaks and had to be removed or repaired with low-
hydrogen electrodes.
In general, the results presented in these three
reports provided additional validation of the B31G and RSTRENG
evaluation concepts. As shown in Table 7, index numbers were
assigned to those results not already having an index number, and
the data are included in the overall database.

Burst Test of a Corroded Piece of 10.75-(Inch) O.D.

by 0.25-Inch WT. API 5L Grade X46 Seamless Line Pipe

The sample of 10.75-inch O.D. pipe shown in Figure 17

was subjected to detailed measurements and burst testing to add
to our knowledge of the behavior of corroded pipe and to provide
 49

further data for the validation of RSTRENG. As has been noted

many of the past corroded pipe experiments have omitted the
comparison of "before" and "after" wall thickness measurements
which would indicate how much thinning and/or outward bulging
takes place prior to the failure event. The issue is important
because "before" measurements are always used to evaluate
corrosion in a live pipeline for repair or replacement; whereas
in the examination of service or test failures, typically only
"after" measurements are available. Also, if outward bulging has
occurred at an anomaly in an in-service pipeline, as suggested by
the results of the tests described in the previous section, the
measurements taken from the O.D. surface may not accurately
reflect the pit depths.

"Before" Measurements of Wall Thickness

A detailed contour map of pit depths was made as shown

in Figure 18. These pit depths were measured from a straight
edge positioned in the necessary locations parallel to the axis
of the pipe, resting on non-corroded portions of the O.D.
surface. Because the piece of pipe was short, it was possible to
check some remaining thicknesses with a 6-inch-throat micrometer.
This technique indicated that the pit depths measurements were
acceptably accurate. In addition, a straightedge was laid
parallel to the axis on the inside surface of the pipe beneath
the deepest pits. This confirmed that the wall thickness at the
pits was not already bulged outward, and it suggested that the
pressure levels this piece of pipe had experienced in service
(-1200 psig) were not near its failure pressure level.
The tensile properties of the pipe material were
determined from a test of a transverse flattened specimen. The
 50

yield strength at 0.5 percent total strain was found to be

49,000 psi and ultimate tensile strength was 77,000 psi Hence,
the flow stress of the material for calculation purposes was
59,000 psi. Transverse flattened Charpy V-notch specimens
(1/2-size) revealed an 85 percent shear area transition
temperature of 90°F and an upper shelf energy of 14 ft.lb.
Corrected for thickness effects, the full-scale fracture
propagation transition temperature of the pipe material is 86°F.
The full-size Charpy specimen equivalent upper shelf energy is
28 ft.lb.

The pit depth measurements along the heavy dashed line

in Figure 18 over an axial distance of 8.75 inches were used to
make a remaining strength calculation. The RSTRENG2 output is
shown in Figure 19. The predicted burst pressure is 1703 psig
corresponding to 79.6 percent of SMYS.

Pressure Test

End caps were welded on the specimen and a pressure

test was carried out at Stress Engineering Services, Inc. in
Houston, Texas. As shown in Figure 20 the pipe developed a leak
at a pressure level of 1715 psig (it did not rupture as might be
suggested by the "Burst Pressure" terminology in Figure 20). An
attempt to cause a rupture by continued pumping was unsuccessful
as the growing leak overcame the capacity of the pump. The pipe
specimen did not undergo yielding prior to the attaining of the
1715-psig pressure level. This was confirmed by comparing the
" b e f o r e " a n d " a f t e r " diameter measurements presented in Figure 21
 51

which show negligible changes in diameters at various locations.

Since this pressure level is only 71 percent of SMYS, the absence
of yielding is not surprising. The failure pressure of 1715 psig
compares favorably to the RSTRENG2-predicted pressure of
1703 psig.

"After" Measurements of Bulging and Wall Thinning

After the burst test the segment containing the leak

was cut from the specimen. The fact that the mode of failure was
a leak rather than a rupture was fortuitous. It provided an
opportunity to assess the effect of outward bulging. A new set
of pit depth measurements was made, and as the "after" contour
map is shown in Figure 22. The differences between the pit depth
contours of Figure 18 and Figure 22 are subtle but significant.
Note that the 0.200-inch contours which exist in Figure 18
between the 5- and 7-inch lines are absent from Figure 22. The
outward bulging in the vicinity of the leak has "taken away" pit
depth. In fact, of course the inside surface is no longer flat;
it has been pushed outward about 0.040 inch at the leak and as
much as 0.020-inch at many locations within about l-inch of the
leak. The net wall thickness has remained the same as was
verified by micrometer measurements. The actual necking was
confined to a region so close to the leak that it could not be
detected with a ball-end micrometer.
The effect of such outward bulging could be important
in the following situation. If this particular pipe had been
operated at a pressure near 1700 psig instead of 1200 psig as it
was, m u c h o f t h e o b s e r v e d b u l g i n g c o u l d h a v e t a k e n p l a c e p r i o r t o
the development of the leak. Under such a circumstance, if
anyone had made an attempt to measure the corrosion while the
 52

pipe was still in the pipeline, an erroneous set of pit depths

would have been obtained. To illustrate the difference, the
"after-bulging" profile of pit depths was used to make second
RSTRENG2 calculation as shown in Figure 23. The RSTRENG2-

calculated burst pressure based on the as-bulged pit depths is

1854 psig. This value is 9 percent higher than the level
calculated using the pre-bulging pit depths and 8 percent higher
than the pressure at which the leak occurred.
The differences in the pit depths before and after
bulging were as follows:

Positions Along Pit Depth

Pit Depth After Different
Axes of Figures Before Bulging
18 and 22 Bulging
3.6 60 60 0
3.25 90 80 10
3.5 170 150 20
3.6 180 160 20
4 150 150 0
4.1 100 120 -20
4.25 160 140 20
4.5 190 160 30
4.75 200 160 40
5 180 160 20
(a)
5.25 200 170 30
(a)
5.50 200 180 20
5.15(a) 160 150 10
6 100 70 30
6.25 110 80 30
6.5 110 170 0
(a)
Location of leak

The maximum measured bulging did not occur right at the leak but
at a nearby deep pit.
 53

The concern raised by this result is that any near-

failure pit is likely to have experienced similar outward
bulging. As a result the pit depth measurements one might make
in such a case would not accurately reflect the severity. Does
this result in an unsafe situation? Probably not, because if
this one example is typical the margin of error is small enough
that the anomaly would be categorized as a repair or cut out
anyway. Assume, for example, that this pipe had been in
operation at an MAOP of 1700 psig. In such a case, if it had
been found in the as-bulged situation reflected by the pit-depth
contours of Figure 22, its calculated burst pressure of 1854 psig
would still put it in the category of an unacceptable anomaly.
Its apparent margin of safety would be too small to justify
leaving it unrepaired.
Another question raised by this test concerns whether
or not the bulging could have been detected without access to the
inside of the pipe. Given the irregular surface of the pitted
area it seems unlikely that ultrasonic thickness readings could
have been made in the vicinity of the leak. Thus, i t i s
reasonable to assume that the bulging would be undetectable.
When considered in terms of internal corrosion the term
bulging is probably inappropriate. It seems likely, though no
observations of such behavior were made, that the corresponding
phenomenon with internal corrosion would be inward dimpling. The
bulging/dimpling phenomenon probably represents an attempt by the
material as the stress level approaches the ultimate strength
level to realign itself with membrane stress in the surrounding
thicker material. In any case this one test result calls
attention to and provides some quantitative measurements of a
phenomenon which can be expected to occur in a region of metal
loss which is pressurized to a near failure level.
 54

Analysis of Service Failure of Corroded Pipe

Presented in this section is an assessment of the

failure of a piece of 26-inch O.D. by 0.281-inch W.T. API 5L
Grade X52 pipe with a flash-welded seam. This pipe ruptured in
natural gas service from external corrosion at a pressure level
of 779 psig corresponding to 69.3 percent of SMYS. The fracture
path and origin location are shown in Figure 24. The flash-welded
seam of the pipe was not involved and no selective corrosion of
the seam was noted. Since the rupture destroyed any evidence of
what was coated (even though no fire occurred), it is unclear how
much of the pipe was coated. The absence of pitting outside the
localized origin region suggests that not much of the pipe had
been exposed. The origin and the grid used for remaining
thickness measurements is shown in Figure 25.
The contour map of remaining thickness measurements is
shown in Figure 26. Note that the origin was only 2-inches in
length; definite regions of propagating shear fracture existed
outside of this region. However, it is reasonable to believe
that a much larger portion of the metal loss influenced the
occurrence of the failure. This issue is discussed below.
The tensile properties of the pipe were determined by
means of a transverse flattened tensile specimen. The yield
strength at 0.5 percent total strain was 62,000 psi and the
ultimate tensile strength was 89,000 psi. The 85 percent shear
area transition temperature of the material as determined by
means of transverse flattened 0.57-size Charpy V-notch specimens
was 120°F. When corrected for thickness effects one finds that
the full-scale fracture propagation transition temperature of the
material is 114°F. The upper shelf energy based on the 0.57-size
 55

specimens was 19 ft.lb. which corresponds to a full-size Charpy

upper-shelf energy of 33 ft.lb.
The RSTRENG2 calculation for the remaining thickness
profile along the rupture is shown in Figure 27. The average
actual uncorroded thickness was 0.281-inch the same as the
nominal value. The actual yield strength of 62,000 psi was used
for SMYS in the calculation to give a flow stress of 72,000 psi.
The resulting predicted failure pressure was 786 psig which
compares favorably with the observed pressure at the time and
location of failure of 779 psig.
This failure was of particular interest for two
reasons. First, the failure occurred as a rupture even through
the fracture origin is clearly confined to a 2-inch region. A
simple calculation using the Folias factor for a 2-inch-long
through-wall crack in this material (M=1.3425) and the flow
stress of 72,000 psi suggests that a 2-inch crack would not be
expected to cause a ductile rupture at a pressure level of
779 psig. (The predicted rupture pressure is 1159 psig.)
Second, because the origin region contained some necking the
question of how to determine the true remaining thickness for use
in RSTRENG2 arises.
With respect to the issue of the 2-inch origin it is
obvious that the surrounding pipe at both ends of the origin,
having been thinned by metal loss, has influenced the behavior.
T h e e x t e n t o f t h e influence is suggested by the RSTRENG2
effective length (10-inches), and the physical reality is that
t h e 2 - i n c h o r i g i n region of 100 mils remaining thickness was
centered within a lo-inch region having a remaining thickness of
150 mils. Repeating the 2-inch through-wall flaw calculation
using a 0.150-inch wall thickness instead of 0.281-inch thickness
(remembering that the hoop stress is 63,513 psi because of the
 56

reduced wall thickness) one finds that the Folias Factor is

M=1.64 and that the effective stress level is more than
sufficient to cause the pipe to rupture. In fact, the situation

of this metal loss region is that of "Type III" interaction

described in Appendix A of this document, and the result confirms
that RSTRENG2 provides valid calculations of remaining strength
for Type III flaws.
To assess the " t r u e " w a l l t h i c k n e s s a t t h e o r i g i n a
metallographic section was made at Location A-A shown in
Figure 26. This section is shown in Figure 27. The least

thickness in the "necked" r e g i o n i s j u s t o v e r 5 0 m i l s . I t

appears that the true wall thickness could have been as thin as
about 70 mils. To be certain that we were not including any
necking however, we took measurements no closer than 0.2-inch
from the edge of the fracture. Thus, our minimum remaining
thickness readings were 100 mils. With these values, the
RSTRENG2 calculation still gave a reasonable result. It appears
that we could have used a remaining thickness as close as
0.10-inch from the edge (about 70 mils) without getting into the
seriously necked region. In any case our use of 100 mils seems
to be conservative. It is not clear based on this one case that
measurements should always be taken at a distance of 0.2-inch.
It is possible that specific metal loss geometries could exhibit
necking over a larger area. When the question arises in
conjunction with a particular failure, it should be possible to
resolve the issue by means of a metallographic section as was
done in this case.
 57

CORROSION-CAUSED METAL-LOSS ANOMALIES FOR

WHICH RSTRENG ANALYSES ARE NOT APPROPRIATE

An important consideration with respect to evaluating

corroded pipe is the recognition that RSTRENG and other analysis
methods including finite-element models are based upon the
assumption that the material will behave in a reasonable ductile
manner. When a metal-loss anomaly exists in a material that will
behave in a brittle manner, the various analysis methods
including RSTRENG may not give conservative predictions of
failure pressure and should not be relied upon in those
instances. Two examples involving past service failures
illustrate this type of behavior. One involved selective
corrosion of a low-frequencyERW seam and the other involved
corrosion in a piece of furnace lap-welded pipe. These cases are
described below.

Analysis of Selective Corrosion

in a Low-Frequency ERW Seam

The low-frequency ERW seam failure occurred in gas

service at a pressure level of 708 psig. The pipe material
involved was 18-inch O.D. by 0.375-inch wall thickness API 5L
Grade A installed in 1948. The corrosion was internal. The seam
was oriented on the bottom of the pipe, the pipe was located in
the low point of a compressor station discharge loop, the loop
sat idle for a long time, and the corrosion developed as the
result of water accumulation and slightly sour conditions.
Copies of figures from the failure analysis report are presented
herein as Figures 29, 30, and 31 and show the nature of the
corrosion. Figure 32 shows the profile of the missing metal.
The yield strength of the parent material was found to be
41,900 psi. The strength of the weld seam was not measured.
 58

The material in the bondline of the ERW seam was

evaluated by means of Charpy V-notch specimens. Only a few tests
were conducted. It was learned that the bondline region
exhibited no shear area at a temperature of 80°F. Specimens
tested at 212°F did not fracture entirely in the bondline but
where the bondline did fail it appeared to exhibit at least
partly cleavage behavior. Also, as seen in Figure 29, the
service failure fracture surfaces clearly exhibit cleavage
fracture.
The RSTRENG calculation for the metal loss shown in
Figure 32 based on a parent-metal flow stress of 51,900 psig
results in a predicted failure pressure of 1,429 psig. This
greatly overestimates the failure pressure and illustrates the
point that RSTRENG should not be used where brittle failure of a
corrosion-caused anomaly may be expected.

Analysis of Corrosion in
a Furnace Lap-Welded Material

A corrosion-caused service failure occurred in a

brittle manner in a pipeline comprised of 22-inch O.D. by
0.312-inch wall thickness lap-welded material manufactured in
1931. The material was probably Grade A. In any case, its yield
strength was found to be 36,900 psi. The failure occurred at a
pressure level of 449 psig.
The external corrosion was comprised of a series of
pits which can be seen in the copies of the attached photographs
(Figures 33 and 34) from the original analysis report. From the
pattern on the surface of the pipe, it appears that the pits and
consequently the origin of failure lies in the lap-welded region.
No metallographic work was done, so this was never verified.
However, the fracture was quite brittle. No ductility was
observed in the ligament beneath any of the pits.
 59

The profile of the missing metal for Pit Group 1 (the

probable origin) can be seen on the RSTRENG plot (Figure 35).
The RSTRENG-predicted failure pressure of 1,057 psi greatly
exceeds the observed failure pressure of 440 psig.

Discussion

These two examples show why the current analytical

models for the failure pressure levels of corrosion-caused metal
loss including RSTRENG apply only to the normally expected
conditions of ductile behavior. One should never attempt to
a p p l y t h e m t o s e l e c t i v e c o r r o s i o n in ERW seams and should be very
careful about applying them to very old line-pipe materials that
are apt to be quite brittle at normal ambient temperatures.

QUALITATIVE AND QUANTITATIVE VALIDATION OF RSTRENG

This report presents an analysis of the results of 129

tests and, in a few cases, service failures involving corroded
pipe or pipe samples containing corrosion-simulating defects.
All of these results either have been or will be added to the
A.G.A./PRCI Database of Corroded Pipe tests that consisted of 86
results at the time that Project PR 3-805 (to develop RSTRENG)
was completed. The latest results provide both qualitative and
quantitative validation of the RSTRENG methodology. The
qualitative validation arises from the fact that the results help
to more clearly define the applicability and limitations of
RSTRENG. The quantitative validation arises from comparisons,
where applicable, between the actual failure pressures of the
specimens and those that were predicted using RSTRENG.
The test results and service failures were assigned
index numbers that continue the numbering system established by
 60

P r o j e c t PR 3-805. The types of tests or failures were as follows.

• Index Nos. 189, 191, 195, 214; Ductile-mode in-service

failures of corroded pipe. (Index 189 involved
internal corrosion.)
• Index Nos. 196, 197; Brittle-mode in-service failures
of corroded pipe. (Index 196 involved selective
corrosion in an ERW seam.)
• Index Nos. 188 and 190; H y d r o s t a t i c t e s t f a i l u r e s o f
corroded pipe.
• Index Nos. 87-92, 106-117, 192-194, 198-213, and 215
( t o t a l o f 3 8 ) ; Burst tests of corroded pipe samples
previously removed from service.
• Index Nos. 9 3 - 1 0 5 a n d 1 1 8 - 1 2 4 ( t o t a l 2 0 ) ; B u r s t t e s t s
of machined corrosion-simulating slots and patches some
oriented spirally and some involving multiple,
interacting defects.
• Index Nos. 1 2 5 - 1 5 7 ( t o t a l o f 3 3 ) ; B u r s t t e s t o f
machined corrosion-simulating slots, pits, and patches
some involving multiple, interacting defects.
• Index Nos. 1 5 8 - 1 6 2 ( t o t a l o f 5 ) ; Combined pressure and
bending tests of large diameter pipes with machined
patches of corrosion.
• Index Nos. 163-187 (total of 25); Pressure tests of
specimens containing various arrays of small pits.

Qualitative Findings Regarding the Use of RSTRENG

Several important issues regarding the use of RSTRENG

were resolved or partly resolved by the analysis of the 129 tests
and service failures.
First and foremost, for ductile-mode failures, the
service and hydrostatic test ruptures and burst tests of corroded
 61

pipe proved once again that RSTRENG provides reliable estimates

of failure pressures. In contrast, the two brittle-mode service
failures exhibited failure pressure levels well below the
pressure levels one would have predicted using the dimensions of
the corrosion in RSTRENG. As a consequence, users of RSTRENG are
forewarned that it should not be applied to selective corrosion
of an ERW seam or to corrosion-caused metal loss in any materials
that are apt to exhibit brittle fracture initiation behavior.
On the issue of flaw orientation, the newly available
results of burst tests on spirally oriented flaws show that
RSTRENG calculations based upon the axially projected length of
an off-axis flaw are overconservative. An empirically derived
spiral angle factor has been postulated as means of modifying the
RSTRENG prediction based on the projected length. Initial
results look promising, but more research is needed to establish
confidence in the spiral-angle-factor approach. In the meantime,
RSTRENG users can continue to use the projected length knowing
that the predicted safe operating pressures will be excessively
conservative but safe nevertheless.
On the matter of flaw length, new data derived from
tests by both Nova and British Gas clearly show that flaw length
beyond a one-diameter length has no significant impact on failure
pressure. Therefore, to evaluate corroded pipe, one needs to
consider only a one-diameter length or a length of 20 inches,
whichever is greater, provided, of course, that the length
considered contains the deepest part of the corrosion.
The issue of flaw interaction was resolved to some
extent. In cases of two corroded areas that lie on the same
axial line but are separated by sound pipe (Type II interaction).
RSTRENG can be used to evaluate both flaws independently and the
overall pair of areas taken as a single flaw. The lowest
predicted failure pressure can be used as a lower bound value for
remaining pressure-carrying capacity. However, the tests by
 62

British Gas as well as those presented in Appendix A suggest that

the interaction of two corroded areas will be almost
insignificant when the length of sound pipe between them exceeds
1 inch. Therefore, it is suggested that potential interaction be
ignored if the length of sound pipe separating two Type II
corroded areas exceeds 1 inch.
From the standpoint of interaction between two axial
flaws which are offset from one another around the circumference
but overlapping in terms of their lengths as projected onto the
longitudinal plane (Type I interaction), the experiments of Nova,
British Gas, and Appendix A point to very limited interaction if
the circumferential spacing exceeds two wall thicknesses. Until
more data are available, however, we suggest continued use of the
overall projected length (and area) to represent the two corroded
areas unless the circumferential separation equals or exceeds six
wall thicknesses. In the latter case, the flaws can be
considered separately.
The experiments of British Gas and the University of
Waterloo as well as the one described in Appendix A strongly
suggest that RSTRENG adequately predicts the failure pressures of
shorter, deeper pits within longer, shallower pits (Type III
interaction). The RSTRENG calculation in such a case should be
based on the worst-case profile through the defect.
The experiments by British Gas and the University of
Waterloo strongly indicate that individual corrosion pits do not
interact if they are separated by one wall thickness or more of
sound pipe.
The issue of pressure reversals surfaced in the British
Gas tests. One large reversal (39 percent) resulted in a leak in
a relatively deep pit. The phenomenon of pressure reversals is
well known but their occurrence is usually infrequent. They are
most likely to occur when a pipeline is subjected to cycles of
pressure from zero to near- f a i l u r e levels of any existing
 63

defects. The potential for pressure reversals is not relevant to

most RSTRENG evaluations because any pit that is accepted is not
a likely candidate for a pressure reversal except in the unusual
circumstance where it has just been subjected to or will
immediately be subjected to a high-pressure hydrostatic test
(i.e., to 100 percent of SMYS).
The issue of axial stress effects was addressed to some
extent by two different sets of tests one conducted at Southwest
Research Institute (SWRI) and the other conducted at the
University of Waterloo (U of W). The results of the SWRI tests
show that under conditions of extreme axial compression or
tension, the circumferential extent of a corrosion anomaly
becomes important. In the case of axial tension, the mode of
failure may change to circumferential as opposed to longitudinal.
In such a case, the RSTRENG-predicted burst pressure provides no
guidance, and an analysis that takes into account the axial
stress and the dimensions of the metal loss on a circumferential
cross section of the pipe is required. In the case of extreme
axial compression, the burst pressures in the SWRI experiments
appeared to be possibly a little lower than it might have been if
the axial load had been zero or tensile in nature. However, the
mode of failure remained that of axial fracture combined with
incipient buckling. Whether the failure pressure was affected by
the compressive stress itself or the incipient buckling was not
clear from the results.
In the U of W tests, t h e e f f e c t o f v a r i a t i o n s i n a x i a l
stress (zero to one-half of hoop stress) appeared to influence
the behavior of individual pits only. In the case of flaws with
appreciable axial extent, the variations appeared to have no
effect.
The bottom line on axial stress effects seems to be
that it will influence the behavior of corroded pipe if it
reaches an extreme state, that is, plastic deformation in tension
 64

or bending and incipient buckling or plastic deformation in

compression. To the extent these conditions might exist on a
pipeline, the operator would have to be concerned about more
extensive evaluations than using RSTRENG alone. It is our
belief, however, that RSTRENG remains the most valuable means of
evaluating the failure pressure of the pipe, and that it could be
used even under conditions of extreme axial load. The caveat
under extreme axial tension would be that the operator would also
have to consider the circumferential extent of the corrosion and
its potential behavior in the presence of axial tension or
bending.
Aside from the extreme axial loading conditions
described above, the results of tests on patches and
circumferential bands of metal loss clearly show that the
longitudinal extent and depth of the corrosion as used in RSTRENG
are the factors which govern the remaining strength. In the
absence of extreme axial loads, the circumferential extent is of
no consequence.
Finally, the implications of near-failure plastic-
strain distortions of corroded pipe with respect to corrosion
measurements and RSTRENG calculations have been mostly resolved.
Necking at the edge of a metal-loss-induced failure is apparently
confined to relatively narrow strips within 0.2-inch of the edges
of the failure. Remaining wall thickness measurements outside of
these strips are believed to accurately reflect the prefailure
wall thickness for the purposes of RSTRENG calculations. In any
case the development of a map of contours of remaining thickness
surrounding the origin can be used to infer the prefailure
thickness near the edges of the failure.
The other form of distortion, outward bulging at near-
failure pressure levels at an area of external corrosion, does
impact the accuracy of both measurements of pit depths and the
resulting RSTRENG calculations. If such bulging were to exist at
 65

an anomaly in an in-service pipeline, it would cause the apparent

pit depths to be less than the actual pit depths. This
phenomenon is probably not detectable by means of ultrasonics
because it is likely to exist in a region where the transducer
cannot be placed on a large enough flat spot to get an accurate
reading of the remaining thickness. In the one case of such
bulging which was examined herein the effect of the bulging was
shown to alter the RSTRENG-predicted failure pressure by
9 percent. However, it is believed that this phenomenon will not
lead to unsafe judgements of corroded areas. For such bulging to
occur, the pipe must be in a near-failure condition at the
maximum normal operating pressure. It can be expected that the
RSTRENG-predicted burst pressure based on the apparent (i.e.
incorrect) pit depths will still be too low to permit leaving the
area unrepaired.

Quantitative Findings Regarding the Validity of RSTRENG

As shown in Reference 5 (the report on Project

PR 3-805, the RSTRENG report), a database of 86 tests was
analyzed and used to validate RSTRENG. the results of that
validation effort are summarized below.

CASE 1 CASE 2 CASE 3 (a)

(RSTRENG) (0.85 dL_ (2/3 dL)
Ruptures
Mean 0.850 0.622 0.537
Std. Dev. 0.150 0.138 0.178
Quantity 41 41 41
Leaks
Mean 0.992 0.675 0.630
Std. Dev. 0.180 0.221 0.209
Quantity 45 45 49
(a) Includes results for L 2 /Dt >20. In such cases, t h e e x i s t i n g c r i t e r i o n
reverts to the remaining thickness approach.
 66

The basis for the above analyses is as follows. The

ratios of RSTRENG-predicted failure pressures to the actual
failure pressures were calculated. The "Case-1" predictions are
based on RSTRENG itself (effective length, effective area). The
"Case-2" predictions are based on the area calculated from
0.85 dL where d is the maximum pit depth and L is the overall
axial length of the corroded area. The "Case-3" predictions are
based on the standard B31G approach (parabolic area, 2/3 dL).
For 41 leaks, the mean value of the Case-l ratios was 0.85 and
the standard deviation was 0.15. Values for the other possible
comparisons are also shown. In the above analysis, the data were
divided into leaks and ruptures because it appeared that the
methodology worked better for ruptures than for leaks. Also, in
the above analysis, the Case-2 and Case-3 values were calculated
based upon flow-stress values of SMYS + 10,000 and 1.1 SMYS,
respectively. At the time these values were calculated, it was
concluded that they gave a reasonable credibility to the RSTRENG
methodology.
With the input of considerably more data that became
available after Reference 5 was issued, it is desirable to
reevaluate these same comparisons. Not all 129 of the new data
are used in this respect because certain of test results were
inappropriate. The omitted tests and the reasons for their
omission are as follows.

Total
Index No. Omitted Reason
93-96, 102, 137, 139 7 Spiral orientation
103, 104, 132, 150, 172- 14 Obvious interaction
181
105, 118 2 Defect-free burst tests
107 1 Fatigue crack in pit
caused premature leak
 67

Total
Index No. Omitted Reason
130, 131, 1 3 , 4 135, 138, 12 No failure
145, 148, 1 4 9 , 155, 156,
164, 170
140, 141, 143, 146 4 cut, removed, rewelded
159, 162 2 Circumferential failure
192-194 3 Actual failure pressures
>1.5 times predicted
values probably because
actual yield strength
was not measured and was
probably much higher
than assumed
196, 197 2 Brittle behavior
TOTAL 47

The remaining 168 results including the original 86 results are

listed in Tables 8, 9 and 10. The ratios of the predicted-to-
actual failure pressures are shown therein for Cases 1, 2, and 3
as defined previously. Note that in the new analyses, the Case-2
and Case-3 comparisons are based upon actual yield +10,000 and
1.1 times actual yield, respectively, if actual yield was
available so the new statistics for these two cases are not
readily comparable to the old ones. In any case, the means and
standard deviations for all of the combinations are as follows.
 68

CASE 1 CASE 2 CASE 3 (a)

(RSTRENG) (0.85 dL) (2/3 dL)
ALL, DATA (168 results)
Mean 0.935 0.825 0.784
Std. Dev. 0.168 0.187 0.217
RUPTURES (82 results)

Mean 0.867 0.804 0.737

Std. Dev. 0.127 0.154 0.206

LEAKS (86 results)

Mean 1.001 0.845 0.830

S t d . Dev. 0.175 0.213 0.217
(a) Includes results for L 2 /Dt >20.

The thing that appears to be borne out by the new data

is the previous observation that some of the early test data
involving leaks may have had inadequate documentation. The
poorest agreement between RSTRENG predictions and actual failure
pressures is still associated with old test results, notably
Index Nos. 6, 9, 11, 12, 16, 20, 32, 37, 41, and 8 of the 11
leaks from the 1970 tests performed by a pipeline operator (Index
Nos. 198, 199, 200, 201, 202, 203, 207, and 208). Only two new
leak cases (Index Nos. 111 and 113) produced results where
RSTRENG significantly overestimated the pressure at which the
leaks occurred, and in these latest two cases, the overestimates
were not nearly as bad as in some of the older tests.
 69

REFERENCES

(1) ANSI/ASME B31G - 1 9 8 4 M a n u a l f o r D e t e r m i n i n g t h e R e m a i n i n g

Strength of Corroded Pipelines, ASME, New York.

(2) M a x e y , W . A . , K i e f n e r , J . F . , E i b e r , R . J . , a n d D u f f y , A .
R "Ductile Fracture Initiation, Propagation, and Arrest in
C y l i n d r i c a l V e s s e l s " , Fracture Toughness, Proceedings of the
1971 National Symposium on Fracture Mechanics, Part II, ASTM
STP 514, American Society for Testing and Materials, pp 70-
81 (1972).

(3) Kiefner, J. F., Maxey, W. A., Eiber, R. J., and Duffy, A.

R "Failure Stress Levels of Flaws in Pressurized
Cylinders", Progress in Flaw Growth and Fracture Toughness
Testing, ASTM STP 536, American Society for Testing and
M a t e r i a l s , pp 461-481 (1973).

(4) Kiefner, J. F., and Duffy, A. R., Summary of Research to

Determine the Strength of Corroded Areas in Line Pipe,
presented at a Public Hearing at the U.S. Department of
Transportation (July 20, 1971).

(5) Kiefner, J. F., and Vieth, P. H., "A Modified Criterion for
Evaluating the Remaining Strength of Corroded Pipe", Project
PR-3-805, Pipeline Research Committee, American Gas
Association, Catalog No. L51609.

(6) Coulson, K. E. W. and Worthingham, R. G., "Standard Damage-

Assessment Approach is Overly Conservative" (Part 1) and
"New Guidelines Promise More Accurate Damage Assessment"
(Part 2), Oil and Gas Journal (April 9 and April 16, 1990).

(7) Hopkins, P. and Jones, D. G., "A Study of the Behavior of

Long and Complex-Shaped Corrosion in Transmission Pipelines",
Offshore Mechanics and Arctic Engineering Symposium, ASME
(1992).

(8) Kim, H. O., "Model Simplifies Estimate of Bending Strength

in Corroded Pipe", Oil and Gas Journal, pp 54-58 (April 19,
1993).

(9) Chouchaoui, B. A., and Pick, R. J., "Behavior of

Circumferentially Aligned Corrosion Pits", accepted for
publication in Int. J. Pres. Ves. & Piping (1993).
 70

(10) Chouchaoui, B . A . a n d P i c k , R . J . , " I n t e r a c t i o n o f C l o s e l y

Spaced Corrosion Pits in Line Pipe", Offshore Mechanics and
Arctic Engineering Symposium, ASME (1993).

(11) Chouchaoui, B. A., and Pick, R. J., "Behavior of Isolated

Pits Within General Corrosion", submitted to Pipes &
Pipelines Internation (1993).

( 1 2 ) V i e t h , P . H . , a n d K i e f n e r , J . F . , Database of Corroded Pipe

Tests, Pipeline Research Committee, American Gas
Association, April 4, 1994.

(13) Vieth, P. H., and Kiefner, J. F., RSTRENG2 User's Manual,

Pipeline Research Committee, American Gas Association, March
31, 1993.

(14) Kiefner, J. F., Maxey, W. A., and Eiber, R. J., "A Study of
the Causes of Failures of Defects That Have Survived a Prior
Hydrostatic Test", NG-18 Report No. 111, American Gas
Association (November 3, 1980).
 TABLE 1. BURST TESTS OF PIPES CONTAINING LONG AXIAL AND
SPIRALLY ORIENTED CORROSION-SIMIULATING SLOTS AND PATCHES
Defect Length, inches
Projected or
Wall Thickness, Axial or Circumferential Yield Strength, Actual Failure Ratio to Burst
Index No. Description Patch Depth, inch inch Spiral Extent Extent psi Mode of Failure Pressure, psig Pressure(a)
93 20° from 0.106 0.263 43.9 15 62,200 Rupture along 2110 0.94
circumferential flaw
94 30° from 0.106 0.266 30 15 66,100 Rupture along 2008 0.90
circumferential flaw
95 45° from 0.103 0.258 21.2 15 64,700 Rupture along 1790 0.80
circumferential flaw
96 20° from 0.101 0.255 21.9 7.5 62,200 Rupture along 2298 1.03
circumferential flaw
97 Single, 0.103 0.261 15 15 64,400 Rupture along 1631 0.73
longitudinal flaw
98 Single, 0.101 0.262 40 40 62,000 Rupture along 1674 0.75
longitudinal flaw
99 Type II, 0.101 0.256 6 15 63,500 Rupture through 1892 0.84
longitudinal both flaws
100 Type II, 0.096 0.255 6 18 64,800 Rupture through 1892 0.84
longitudinal both flaws
101 Type II, 0.101 0.255 6 24 65,000 Rupture through 1892 0.84
longitudinal one flaw
102 Type I, 20° from 0.104 0.259 37.4 12.8 64,100 Rupture through 2211 0.99
circumferential both flaws
103 Type I, 0.106 0.263 15 20 64,600 Ruputre through 1602 0.72
longitudinal 2t all of one and
separation part of other
104 Type I, 0.102 0.254 15 20 61,900 Rupture through 1529 0.68
longitudinal 4t one flaw and part
separation of other
105 Defect-free burst 66,100 Rupture 2240 1.00
test
118 Defect-free burst 62,364 Rupture 2211 1.00
test
119 Single long. 0.134 0.252 39.4 1.0 62.350 Rupture 1160 0.52
narrow patch
 TABLE 1. (Concluded)

Defect Length, inches

Projected or
Wall Thickness, Axial or Circumferential Yield Strength, Actual Failure Ratio to Burst
Index No. Description Patch Depth, inch inch Spiral Extent Extent psi Mode of Failure Pressure, psig Pressure”’
120 Patch A 0.087 0.252 40.0 1.0 62,350 Rupture 1711 0.77
Patch B 0.130 0.252 6.0 1.0 62,350 No failure ___- __--
121 Circumferential Full
Band A 0.118 0.252 4.1 circumference 63,075 Rupture 1813 0.82
Circumferential Full
Band B 0.050 0.252 8.2 circumference 63,075 No failure ___- ____
122 Circumferential Full
Band A 0.128 0.252 4.1 circumference 63,075 No failure ____ ___-
Circumferential 0.252 8.2 Full
Band B 0.114 circumference 63,075 Rupture 1421 0.64
123 Patch A 0.134 0.252 8.1 8.1 62,350 Rupture 1226 0.55
Patch B 0.126 0.252 4.1 4.1 62,350 No failure ____ ____
124 Single long, 0.126 0.252 39.4 1.0 63,100 Rupture 1218 0.55
narrow patch 2
For Index Nos. 93-104, the comparison is to Index 105. For Index Nos. 119-124, the comparison is to Index 118.
 TABLE 2. COMPARISONS BETWEEN RSTRENG PREDICTIONS AND EXPERIMENTS BY BRITISH GAS'"
(ALL ON 24-INCH O.D. BY 0.486-INCH WALL THICKNESS X52 PIPE)

Defect as Referenced in
British Gas Paper Dimensions

Index Defect Length, L, Actual Failure RSTRENG Prediction, Ratio of Predicted

No. Defect Designation inches Defect Depth, d, inch Pressure, psig psig to Actual Description of RSTRENG Application

I25 Vessel 1-1 3048-mm slot 120 0.194 2103 1853 0.88
126 Vessel 2-1 610-mm slot 24 0.194 2030 1995 0.98
127 Vessel 2-2 305-mm slot 12 0.194 2248 2173 0.97
128 Vessel 2-3 305-mm slot 12 0.194 2393 2173 0.91 Apparently same as C in Figure 4
129 Vessel 2-4 152-mm slot 6 0.194 2683 2515 0.94 Apparently same as A in Figure 4

130 B Figure 4 (no failure) Type II(b) 0.194 >2683 2515 <0.94 Treated as single 6-inch defect
2232 <0.83 Treated as two 6-inch defects with 1.5-inch
separation
131 D Figure 4 (no failure) Type II 0.194 >2683 2413 <0.90 Treated as two 6-inch defects with 3-inch
separation
132 E Figure 4 Type I(c) 0.194 2233 1995 0.89 Treated as 24-inch continuous defect
2173 0.97 Treated as single 12-inch defect
133 F Figure 4 (cut and 12 0.194 2509 2173 0.87 Treated as single 12-inch defect
retested)

134 A Figure 5 (no failure) 1.5 0.291 >3176 2849 <0.90 Treated as single pit
135 B Figure 5 (no failure) 1.5 0.291 >3176 2849 <0.90 RSTRENG analysis same as Index 135
136 C Figure 5 1.5 0.291 3176 2849 0.90 RSTRENG analysis same as Index 134
137 D Figure 5 Diagonal 0.291 2944 2475 0.84 Based on projected length of 3.17 inches
I38 E Figure 5 (no failure) 1.5 0.291 (g) 2849 <0.97 RSTRENG analysis same as Index 134
139 F Figure 5 Diagonal 0.291 3016 2337 0.77 Based on projected length of 3.88 inches
140 G Figure 5 (cut and Type II 0.291 2849 0.88 RSTRENG analysis same as Index 134
retested) 2561 0.79 5.5-inch length with two 1/2-inch
separations

141 H Figure 5 (cut and Type II 0.291 3451 2849 0.83 RSTRENG analysis same as Index 130
retested) 2623 0.76 7.5-inch length with two 1-1/2-inch
separations
142 I Figure 5 4.5 0.291 2726 2520 0.92 Treated as three touching pits

143 A Figure 6 (cut and Type II 0.194 2654 2515 0.95 RSTRENG analysis same as Index 129
retested) 2232 0.84 RSTRENG analysis same as Index 130
 TABLE 2. (Concluded)

Defect as Referenced in
British Gas Paper Dimensions

Index Defect Length, L, Actual Failure RSTRENG Prediction, Ratio of Predicted

No. Defect Designation inches Defect Depth, d, inch Pressure, psig psig to Actual Description of RSTRENG Application

144 B Figure 6 12 0.194 2567 2274 0.89 Treated as two touching 6-inch defects
145 C Figure 6 (no failure) 6 0.194 >2567 2515 <0.98 RSTRENG analysis same as Index 129
146 D Figure 6 (cut and 6 0.194 3118 2515 0.81 RSTRENG analysis same as Index 129
retested)
147 E Figure 6 6 0.194 2944 2515 0.85 RSTRENG analysis same as Index 129

148 F Figure 6 (no failure) 6 0.194 >2944 2515 <0.85 RSTRENG analysis same as Index 129
149 G Figure 6 (no failure) Type II 0.194/0.291 >2944 2370 <0.81 Treated with 1.5-inch gap
150 H Figure 6 Type II 0.194/0.291 2770 2371 0.86 Treated with 1/2-inch gap
151 I Figure 6 (6.3% reversal) 7.5 0.194/0.291 2596(d) 2343 0.85(f) No gap, pit touches slot

152 A Figure 7 Type III(e) 0.194/0.292 2857 2343 0.82 Deep 2-inch area, 6-inch shallow area
4
153 B Figure 7 (1% reversal) 6 0.220 2828(d) 2426 0.85(f) Single 6-inch area lb
154 C Figure 7 (39% reversal) Type III 0.194/0.291 1740(d) 2306 0.81(g) Two deep pits within 6-inch shallow area
155 D Figure 7 (no failure) 3 0.194 >3031 2785 <0.92 Single 3-inch slot
156 E Figure 7 (no failure) Type III 0.047/0.194 >3031 2665 <0.88 Deep 3-inch slot within 12-inch shallow slot
157 F Figure 7 Type III 0.047/0.194 3031 2665 0.88 Deep 3-inch slot within 24-inch shallow slot

(a) Hopkins, P., Jones, D. G., “A Study of the Behavior of Long and Complex-Shaped Corrosion in transmission Pipelines”, International Offshore Mechanics and Arcric Engineering Conference, ASME
(1992).
(b) Type II interaction: Axially-oriented defects on same axial line separated by good pipe.

(c) Type I interaction: Axially-oriented defects separated circumferentially but at least partially overlapping along the axis.

(d) Pressure reversal.

(e) Type III interaction: One or more deep pits located within a long, shallow axially-oriented slot or corroded area.

(f) Ignores pressure reversal. Index 151 initially survived 2770 and Index Nos. 153 and 154 initially survived 2857 psig.

(g) Maximum pressure attained was probably 2,944 psig because of location between Index Nos. 137 and 139.
 TABLE 3. EXPERIMENTS ON 48-INCH PIPE CONTAINING CORROSION-SIMULATING MACHINED AREAS
SUBJECTED TO PRESSURE AND BENDING MOMENT
Corroded Area Dimensions
Internal Bending Predicted
Pressure Moment Failure
Circum- Level at at Pressure,
Index Axial, ferential, Failure Failure, (RSTRENG2) Mode of Mode of
No. inches inches d/t PF, psig ft-kips PR, psig Loading Failure P R /P F
158 18 12 0.25 1480 3769 1226 Constant Axial 0.83
flexural fracture,
displace- compression
ment, side
increasing
pressure
159 6 30 0.50 950 3532 1165 Constant Circumfer- 1.23
pressure, ential
increasing fracture,
flexural tension
displace- side
ment
160 18 12 0.50 980 3729 898 Constant Axial 0.92
flexural fracture,
displace- compression
ment, side
increasing
pressure
161 30 6 0.50 840 3831 843 Constant Axial 1.00
flexural fracture
displace- compression
ment, side
increasing
pressure
162 6 30 0.50 950 3155 1165 Constant Circum- 1.23
pressure, ferential
increasing fracture,
flexural tension
displace- side
ment
 TABLE 4. RESULTS OF TESTS INVOLVING CIRCUMFERENTIAL
AND LONGITUDINAL ARRAYS OF PITS
Yield Wall Length of Depth of RSTRENG Ratio of
Index Waterloo Strength, Thickness, Corrosion, L, Corrosion, d, Pressure at Mode of Location of Predicted Predicted to
No. NO.(a) ksi inch inch inch Failure, psig(a) Failure Failure”’ Pressure, psig Actual

163 S1cc 51.6 0.243 0.79 0.147 2734 Leak 7a (Set 1) 2157 0.79
(c) 2204 <0.79
164 S2cc 51.6 0.247 0.78 0.148 >2774 Rupture
165 S3cc 51.6 0.246 0.78 0.149 2795 7a (Set 1) 2190 0.78
166 S4cc 61.2 0.243 0.78 0.148 2819 7a (Set 2) 2495 0.89
167 S1co 55.4 0.252 0.79 0.127 2413 7a (Set 1) 2446 1.01
168 S2co 55.4 0.237 0.76 0.141 2352 7a (Set 1) 2250 0.96
169 S3co 54.1 0.248 0.78 0.141 2313 7a (Set 1) 2324 1.00
(c) 2304 <1.12
170 S4co 54.1 0.248 0.79 0.147 >2054
171 S11c 55.3 0.247 0.78 0.148 2554 7b (Set 2) 2336
172 S21c-2 54.8 0.248 1.56 0.149 >2191 7b (Set 1) 1988 <0.91 (2 pits
touching)
173 S21c-3 54.8 0.248 2.84 0.149 2191 Rupture 7b (Set 2) 1830 0.84 (3 pits
separated by
1wt)
174 S31c-3 55.3 0.246 2.84 0.149 2273 Rupture 7b (Set 1) 1825 0.80 (3 pits
separated by
1wt)
175 S31c-4 55.3 0.246 4.62 0.149 >2273 7b (Set 2) 1812 <0.80(4 pits
separated by
2wt)
176 S41c 54.8 0.243 0.79 0.146 2212 7b (Set 2) 2277 1.03
177 S11o 54.1 0.253 0.82 0.120 2293 7b (Set 1) 2414 1.05
 TABLE 4. (Continued)

Yield Wall Length of Depth of RSTRENG Ratio of

Index Waterloo Strength, Thickness, Corrosion, L, Corrosion, d, Pressure at Mode of Location of Predicted Predicted to
No. No.“’ ksi inch inch inch Failure, psig(a) Failure Failure(b) Pressure, psig Actual
178 S21o-2 54.1 0.252 1.56 0.146 2012 7b (Set 1) 2039 1.01 (2 pits
touching)
179 S21o-3 54.1 0.252 2.84 0.146 >2012 7b (Set 2) 1880 <0.93 (3 pits
separated by
1wt)
180 S31o 51.6 0.254 0.78 0.149 2152 7b (Set 2) 2279 1.06
181 S41o 51.6 0.250 0.78 0.146 2252 7b (Set 2) 2243 1.00

(a) Last letter in sequence, c or o, designates end loading condition, C refers to a closed-ended pressure vessel, o refers to an open-ended pipe with no axial load
(requires special test fixture).

(b) Refer to Figures 7a of 7b to see arrangement of pits.

(c) Failed at a location other than the pits of interest.

 TABLE 5. RESULTS OF TESTS INVOLVING A SMALL PIT WITHIN A LARGER CORRODED AREA
Lengths Depths
Waterloo Wall RSTRENG Ratio of
Index Test Defect Overall. Small Pit, Thickness, Yield Pressure at Mode of Predicted Predicted to
No. No.(a) No.(b) inches inch General, inch Pit, inch inch Strength, psi Failure, psi Failure Pressure, psig Actual
182 G1c 3 2.20 0.81 0.106 0.178 0.270 54.1 2393 Leak, pit 2153 0.90
183 G2c 1 0.80 0.098 0.174 0.261 55.3 2302 Leak, pit 1925 0.84
184 G3c 2 4.11 2.19 0.107 0.183 0.268 55.3 2126 Leak, pit 1700 0.80
185 G1o 2 2.20 0.87 0.108 0.183 0.267 58.4 2350 Leak, pit 2220 0.94
186 C2o 1 4.25 0.84 0.108 0.175 0.265 52.1 2081 Leak, pit 1798 0.86
187 G3o 2 4.19 2.18 0.095 0.166 0.259 58.4 2028 Leak, pit 1810 0.89

(a) Last letter in sequence, c or o, designates end loading condition, C refers to a closed-ended pressure vessel, o refers to an open-ended pipe with no axial load
(requires special test fixture).
(b) Refer to Figure 6 for description of defect.
 TABLE 6. DESCRIPTION OF MATERIALS INVOLVED IN CORRODED PIPE BURST TESTS AT A
PIPELINE OPERATOR'S FACILITY IN 1970.

Actual Apparent
Wall Length of Yield Ultimate Pressure
Pipe Diameter, Thickness, Specimen, Strength, Strength, Elongation Yield
Number Inches Inches Feet psi psi Percent Strength, psi
1 24 0.396 39 57,100 86,100 26 42,900
2 24 0.332 27 54,600 85,800 24 45,200
3 24 0.378 27 52,200 86,400 27 35,800
4 24 0.307 15 53,350 77,550 29 49,250

5 12.75 0.197 15 58,600

6 (a) 24 0.355 7 65,000 79,000 34 4
ul
(b)
39 24 .417 11 50,200 79,000 25 39,600

40 ( b ) 24 .410 10 46,800 81,300 28.5 45,100

41 ( b ) 24 .396 10 50,200 79,000 25
42 ( b ) 24 I .444 I 11 1 50,200 1 79,000 1 25 1 41,800 I

(a) Pipe Number 6 was not subjected to pressure testing.

(b) Index Number (refers to test made as part of NG18 project using pipe donated by the
pipeline operator).
TABLE 7. RESULTS OF TESTS ON CORRODED SAMPLES COMPARED TO RSTRENG PREDICTIONS

RSTRENG Mode of
Total Effective Wall Yield RSTRENG Failure
Length Length Thickness Strength PRED. Actual L= Leak INDEX
No. in. in. in. psi R=Rupture d/t NO.
Pp Pa Pp/Pa
1-1 3 3 .390 57,100 1974 1380 1.430 L .76 198
1-2 3.50 3.15 .390 57,100 1926 1460 1.319 L .52 199
1-3(a) 15 10 .395 57,100 1457 1470 0.991 R .77 43
3-1 5.5 3 .370 52,200 1185 1075 1.102 L .88 200
3-2 2.5 2.25 .370 52,200 1660 1215 1.366 L .88 201
(b)
3-3 13 9 .364 52,200 1220 1265 0.964 R .70 44
3A-1 3 3 .370 52,200 1775 1350 1.315 L .70 202
3A-2 3 3 .335 52,200 1493 1120 ( c ) 1.333 L .64 203
3A-3B 14 9 .370 52,200 1424 1435 0.992 R .59 204
4-1 3.4 3.2 .330 53,400 1050 1050 1.000 L .73 205
4-2 8 7.6 .330 53,400 1161 1100 1.055 R .64 206
4-3 2.6 2.6 .330 53,400 1442 1240 1.163 L .70 207
4-4 3 3 .330 53,400 1458 1260 1.157 L .73 2.08
4-5 4.5 4.35 .330 53,400 1325 1280 1.035 R .73 209
4-7 4.7 4.5 .330 53,400 1334 1350 0.988 L .61 210
4-9 1.8 1.8 .330 53,400 1626 1505 0.926 L .70 211
5-1 3.3 2.96 .197 58,570 1636 1660 0.986 R .66 212
5-2 15 6 .197 58,570 1399 1810 0.773 R .56 213
(a) This test was Index 43 of the original 47 NG-18 tests of corroded pipe.
(b) This test was Index 44 of the original 47 NG-18 tests of corroded pipe.
(c) 1 7 % p r e s s u r e r e v e r s a l f r o m 1 3 5 0 p s i g o n p r e v i o u s c y c l e .
FIGURE 1. INDEX NOS. 128-133, LONG, NARROW DEFECTS AND POSSIBLE INTERACTIONS
FIGURE 2. INDEX NOS. 134-142, ARRAYS OF PITS
FIGURE 3. INDEX NOS. 143-151, PATCHES OF METAL LOSS
FIGURE 4. INDEX NOS. 152-157, COMPOUND DEFECTS
 93

FIGURE 5. GENERAL SHAPE OF INDIVIDUAL ELECTROCHEMICALLY MACHINED

DEFECTS (COURTESY OF UNIVERSITY OF WATERLOO)
 94

FIGURE 6. TYPICAL SHAPE OF PIT WITHIN CORRODED REGION

(COURTESY OF UNIVERSITY OF WATERLOO)
FIGURE 7. CIRCUMFERENTIAL, LONGITUDINAL, AND SPIRAL ARRAYS OF PITS
(COURTESY OF UNIVERSITY OF WATERLOO)
 96

FIGURE 8. PROFILES OF ANOMALIES 1-1 AND 1-2 FROM PIPE NUMBER 1.

FIGURE 9. PLAN AND PROFILE VIEWS OF ANOMALY 1-3 FROM PIPE NUMBER 1.
98
FIGURE 11. PLAN AND PROFILE VIEWS OF ANOMALY 3-3 FROM PIPE NUMBER 3.
 100

F I G U R E 1 2 . ANOMALIES FROM PIPE NUMBER 3A: F I R S T L E A K

( 3 A - 1 , T O P ) ,S E C O N D L E A K ( 3 A - 2 , B O T T O M ) .
 ACTUAL PROFILE OF CORROSION AREA

FIGURE 13. ANOMALY THAT RUPTURED PIPE NUMBER 3A (3A-3B).

 CONTOUR OF CORROSION PITS AT POINTS OF FAILURE
MEASUREMENTS TAKEN FROM PLASTER MOLDS

FIGURE 14. PROFILES OF ANOMALIES FROM PIPE NUMBER 4.

 103

FIGURE 15. ANOMALY 5-1 FROM PIPE NUMBER 5.

 104

FIGURE 16. ANOMALY 5-2 FROM PIPE NUMBER 5.

 105

F I G U R E 1 7 . PHOTOGRAPH OF INDEX 214 BEFORE TEST

FIGURE 18. CONTOURS OF PIT DEPTHS “BEFORE” PRESSURE TEST, INDEX 214
 107

FIGURE 19. RESTRENG CALCULATION FOR INDEX 214 USING

"BEFORE" MEASUREMENTS
 108

FIGURE 19. (CONTINUED)

 110

Note:
(Note 1) The accuracy of the readings is dependent upon the location of the micrometer placement within
the corroded section of the pipe segment. This is especially true for those values measured 0° relative to
the corroded section.
(Note 2) See figure below for location of the measurements. Angles are relative to line drawn through
center of corroded region.

Burst Test on February 2, 1995 Stress Engineering Services, Inc. PN7187

FIGURE 21. DIAMETRIC READINGS OF CORRODED SAMPLE

FIGURE 22. CONTOURS OF PIT DEPTHS "AFTER" PRESSURE TEST, INDEX 214
 112

(*) If the calculated safe maximum pressure for the criteria that
the user follows (CASE 1, CASE 2, or CASE 3) is less than
the established pressure, remedial action must be taken.

FIGURE 23. RSTRENG CALCULATION FOR INDEX 214 USING

"AFTER" MEASUREMENTS
 113

FIGURE 23. (CONTINUED)

FIGURE 24. FRACTURE PATH AND ORIGIN LOCATION
 115

FIGURE 25. ORIGIN OF RUPTURE

 116

FIGURE 26. CONTOURS OF REMAINING THICKNESS

 117

Pressure, psi = 809

Safe Predicted
Maximum Burst Factor
Pressure Pressure of
psi (*) psi Safety
-------- -------- --------
CASE1 - E f f e c t i v e A r e a 475 786 0.97
CASE2 - 0.85 dL Area 479 793 0.98
CASE3 - B31G 317 525 0.65

Max. P i t D e p t h , i n c h = 0.181
Total Length, inch = 19.00 Max. Depth/Thickness = 0.64
Eff. Length, inch = 10.00 E f f . A r e a , inchA = 1.673
Start, inch = 7.00 Stop, inch = 17.00
(*) If the calculated safe maximum pressure for the criteria that
t h e u s e r f o l l o w s ( C A S E 1 , CASE 2, or CASE 3) is less than
the established pressure, remedial action must be taken.

-------------------- Lengths and Pit Depths--------------------

Length Depth Length Depth Length Depth
inch MIL inch MIL inch MIL
-------_ ----- -------- ----- -------- -----
0.00 0
0.50 31
3.50 81
7.00 131
10.00 181
14.50 181
17.00 131
18.30 81
18.60 31
19.00 0
- -

FIGURE 27. RSTRENG2 PREDICTION OF FAILURE PRESSURE

 119

OD surfaces
at
back-to-back

0.6X

FIGURE 29. PHOTOGRAPH OF ORIGIN OF RUPTURE (NOTE CHEVRONS

POINTING TO CENTRAL 4-INCH-LONG GROUP OF PITS)
 120

ID Surface

OD Surface

1.4x

F I G U R E 3 0 . PHOTOMACROGRAPH OF CROSS SECTION THROUGH THE DEEPEST

PIT AT THE ORIGIN

FIGURE 31. PHOTOMACROGRAPH OF CROSS SECTION THROUGH INTACT

PORTION OF WELD SHOWING SOME SELECTIVE CORROSION
 121

FIGURE 32. RSTRENG CALCULATION FOR SELECTIVE CORROSION DEFECT

IN ERW SEAM
 122

FIGURE 33. OVERALL PITTED AREA

FIGURE 34. ORIGIN

 123

FIGURE 35. RSTRENG CALCULATION FOR PITS LOCATED IN BONDLINE

REGION OF LAP-WELDED PIPE
THIS PAGE IS INTENTIONALLY BLANK
 APPENDIX A

METHODS FOR DEALING WITH THE INTERACTION

OF CLOSELY SPACED AREAS OF METAL LOSS
THIS PAGE IS INTENTIONALLY BLANK
 A-1

APPENDIX A

METHODS FOR DEALING WITH THE INTERACTION

OF CLOSELY SPACED AREAS OF METAL LOSS

No provision exists in the original B31G criterion for

considering the effects of interaction between closely spaced
corrosion anomalies. As a result, users of the criterion must
either choose to treat such areas as being continuous
(introducing excessive conservatism) or treat them as separate
defects based on an arbitrary criterion of separation distance
(possibly resulting in a nonconservative assessment).
Examples of the kinds of interaction commonly
encountered in corroded pipe are illustrated in Figure A-l.
Interactions between flaws of these three types were studied
b r i e f l y(A-1) under Pipeline Research Committee sponsorship about
the time the original B31G criterion was being developed.

Type I Defects

Type I interactions are those in which the flaws are

separated circumferentially, but their individual profiles
overlap when projected to a single plane through the wall
thickness forming a projected profile of length, L, as shown in
Figure A-l. The calculated remaining strength of the resulting
profile tends to underestimate the remaining strength because the
islands of full-thickness material between the individual flaws
prevent their acting as a single flaw. Experiments described in
Reference A-l and summarized in Figure A-2 were intended to
provide a basis for a more accurate method of calculating the
remaining strength of pipes containing Type I flaws.
 A-2

If one assumes that the relationship between the

circumferential spacings and the failure pressures varies
linearly, the most conservative relationship that can be used to
describe the results in Figure A-2 is
PI = 65 T + 620 (A-1)
where T is the circumferential spacing in units of full-wall
thickness. Thus, the failure pressure for T = 0 is 620 psig,
that of the case of the 8.4-inch-long Single flaw. As T becomes
1 , 2 , 3 , e t c . , the failure pressure by Equation (A-l)
successively increases such that
at T = 1, P, = 685
at T = 2, P, = 750
at T = 3 , 1 P = 815
at T = 4, P1 = 880
at T = 5, P, = 945
at T = 6, P, = 1010.
D a t a i n F i g u r e A - 2 e x i s t s f o r T = 1 , T = 2 , a n d T = a~. At T + 1,
the observed failure pressure was 695 psig. At T + 2, the
observed failure pressure was 750 psig. At T = 00 ( a s i n g l e 4 . 8 -
inch defect), the failure pressure was 990 psig. The
relationship suggested by Equation (14) arises from and agrees
reasonably well with the data. Because these data were derived
from a test of only one size of pipe, one material, and one set
of flaw geometries, extrapolation of these results to other pipe
size, materials, and flaws involve some risk. Nevertheless,
these results strongly suggest that at a separation distance of 6
wall thicknesses, Type I defects can be expected to exhibit
little or no interaction. The appropriate means of handling a
Type I situation would be to treat the defects as a single defect
if the circumferential spacing is less than 6T and to treat the
defects separately if the spacing is equal to or greater than 6T.
A better means of handling those spaced at less than 6T awaits
further experimental or analytical work to develop a well-
validated relationship.
 A-3

Type II Defects

Type II defects as shown in Figure A-l, are those which

lie on the same axial line but are separated by a length of full-
wall thickness pipe. Experiments involving Type II defects were
presented in Reference A-l and are summarized in Figure A-3. AS
Figure A-3 shows, a s o n e w o u l d l o g i c a l l y e x p e c t , t h e f a i l u r e
stress level increases with increasing spacing until there is no
interaction and the failure is based on the behavior of the
largest of the two flaws.
Also included in Figure A-3 is solid curve "predicting"
the relationship between failure stress and defect spacing. This
curve is based upon the NG-18 surface flaw equation as used in
Figure A-4. Essentially, the two defects are analyzed as though
they are one. The area of the combined flaw is taken as the sum
of the two areas of the individual flaws. The length of the
combined flaw is the sum of the lengths of the individual flaws
plus the separation distance.
Obviously, there is a separation distance beyond which
the defects should not be considered in the manner shown in
Figure A-4. That limit is the distance at which the predicted
failure stress for the combined defect exceeds that of the most
severe of the individual defects. For the data presented in
Figure A-3, the pair of defects separated by one defect length
(4.8 inches) did not interact. The figure also shows that the
predicted failure stress levels of the combined defects begin to
exceed that of the individual flaws at a spacing of about 3
inches. The actual degree of interaction in every case was less
than predicted (i.e., the actual failure stress exceeded the
predicted failure stress). In fact, the data suggest that
significant interaction occurs only when the spacing is 1 inch or
less. This result has led some operators to the conclusion that
corrosion anomalies do not interact at all unless they are
continuous. As a compromise until more data are available, it is
 A-4

suggested that corrosion anomalies separated along the axis of

the pipe by more than 1 inch be considered to act independently.
When separated by less than 1 inch, they can be analyzed as
outlined in Figure A-4.

Type III Defects

The geometric parameters of the Type III defect are

shown in Figure A-5. Also shown are the lengths and areas
related to seven candidate types of analysis with respect to
predicting the failure stress of the defect. These methods are
described below. All are based upon the NG-18 surface flaw
equation. Note that nothing in the analysis requires the deep
defect to be centered within the shallow defect. Hence, two
defects of unequal depths touching end-to-end may be treated in
this manner.

Method 1 - E x a c t A r e a . In this method as shown in

Figure A-5, the length is taken as L,, and the area is taken as
A = LA + L2 (dz - d,) . A , i s L,t.

Method 2 - S h o r t , D e e p D e f e c t O n l y . In this case, the

long defect is ignored. The defect length is taken as & and the
a r e a i s t a k e n a s A = &d,. A0 i s ht.

Method 3 - S h o r t . D e e p D e f e c t i n a R e d u c e d T h i c k n e s s
Pipe. For this method, the short, deep defect is considered to
exist in a pipe the thickness of which is equal to the net
thickness beneath the long defect (t-d1). The length of the
d e f e c t i s L, a n d t h e a r e a A i s &(d2 - d , ) . A, is L, (t - d,) .
B e c a u s e t h e p i p e i s a s s u m e d t o b e (t - d , ) i n t h i c k n e s s , t h e h o o p
stress must be calculated from the Barlow formula using (t - d,).
Hence,
PxR
S =
(t - d , )
 A-5

Method 4 - Equivalent Length, Total (Equivalent) Area.

In this method, the entire defect area is considered. That is
A = L 1 d 1 + L 2 (d 2 - d 1 ) a s i n M e t h o d 1 , b u t t h e l e n g t h i s d e f i n e d
as L = A / d 2 o r t h e t o t a l a r e a d i v i d e d b y t h e m a x i m u m d e p t h d 2 .
This method creates for analysis purposes, a single rectangular
defect with an area and maximum depth equal to that of the actual
defect. Aº is Lt.

Method 5 - Long, Shallow Defect Only. In this case,

the existence of the deep defect is ignored. Length is taken as
L1 and area is A = L1d1. i s L1t.

Method 6 - E f f e c t i v e A r e a . In this case, successive

trial calculations are used to find the minimum failure pressure
based on the effective area and length. This is, in fact, what
RSTRENG does so this could be called the "RSTRENG" method.

Method 7 - P a r a b o l i c A r e a . This is the method

currently embodied in the B31G criterion.

Comparisons of the Seven Methods for Type III Defects

Comparisons of five of the seven methods for predicting

the behavior of a specific case of this type of defect are shown
in Figure A-6. The figure represents a sensitivity study using
the above-described variations on length and area and the NG-18
surface flaw equation. The pipe material is 36-inch O.D. by
0.400-inch wall X60 (actual flow stress = 73,100 psi). Methods 2
and 5 are not included in these comparisons because upon brief
consideration it becomes obvious that neither of these methods is
appropriate. In the case of Method 2, the predicted failure
p r e s s u r e s w o u l d b e c o m e u n r e a l i s t i c a l l y h i g h a s t h e l e n g t h , &,
becomes very short. In the case of Method 5, the failure
pressure is independent of L2 and d2, c l e a r l y n o t a r e a l i s t i c
 A-6

situation. Hence, t h e f o l l o w i n g c o m p a r i s o n s i n v o l v e o n l y
Methods 1, 3, 4, 6, and 7.
The analyses are summarized in Figure A-6. The flaw
parameters are as follows. The overall flaw is 13 inches in
length. In one region its depth, d,, is 0.120 inch. Its depth
i n t h e o t h e r r e g i o n , d,, i s 0 . 3 0 0 i n c h . The length of the deeper
r e g i o n , &, v a r i e s f r o m 0 t o 6 . 5 i n c h e s . For the trivial case,
L 2 = 0, the failure pressure level of the 13-inch-long
rectangular flaw with a uniform depth of 0.120 inch is 1,203 psig
as predicted by both the exact area method and the effective-area
method. As L2 becomes nonzero, t h e f a i l u r e p r e s s u r e s p r e d i c t e d
by both Method 1, the exact-area method and Method 6, the
effective-area method, decrease. For L2 values in excess of 2
inches, the effective-area predictions fall below those of the
exact area method. The effective-area method predictions at
h i g h e r v a l u e s o f L, r e f l e c t t h e o b s e r v a b l e f a c t t h a t t h e b e h a v i o r
of the flaw becomes controlled to an increasing extent by the
shorter flaw. It is seen that the failure pressure predicted by
t h e e f f e c t i v e - a r e a m e t h o d f o r t h e s i t u a t i o n i n w h i c h L, = 6 . 5 i s
735 psig. This result merely reflects the fact that the minimum
failure pressure is based on the 6.5-inch-long 0.300-inch-deep
effective area, independent of the longer defect.
The one available test result for this type of flaw was
obtained on a 36-inch-diameter, 0.400-inch wall pipe material
with a flow stress level of 73,100 psi. The defect in the
specimen was 13 inches in length overall with a centrally located
shorter flaw of 6.5 inches in length. The depths, r e s p e c t i v e l y ,
were 0.120 inch and 0.300 inch. Its failure pressure level was
749 psig. The failure pressure predicted by the effective-area
method was 735 psig. Thus, the failure pressure of the test flaw
was adequately predicted by the effective-area method but was
overestimated by the exact-area method which gave a predicted
failure pressure of 910 psig.
 A-7

Note that the failure pressures predicted by Methods 3

and 4 fall considerably below those predicted by the effective-
are method indicating that they are excessively conservative.
Lastly, consider the predictions via Method 7
(parabolic or existing B31G method). This method is not capable
o f c o n s i d e r i n g t h e f a c t t h a t L 2 v a r i e s and must be based on L1 =
13, and d2 = 0.300. Method 7 yields a predicted failure pressure
(with no factor of safety) of 982 psig, 1.31 times the actual
failure pressure of test specimen with the compound flaw of 6.5
inches at 0.300-inch depth within the 13-inch-long, 0.120-inch-
deep flaw. The use of the effective-area method is clearly
preferable to the handling of this type of defect by means of the
old B31G criterion. In such a case, the existing B31G criterion
allows a defect of 0.300-inch depth to be no more than 1.97
inches in length or the 13-inch-long flaw to be operated at no
more than 595 psig. However, the B31G criterion really does not
permit one to make this calculation because Q is greater than
4.0.

REFERENCE

(A-1) Kiefner, J. F., "Fracture Initiation", in the 4th

Symposium on Line Pipe Research, Paper G, American Gas
Association Catalogue No. L30075 (November 18, 1969).
TYPE I (PLAN VIEW) TYPE II (PROFILE VIEW)

TYPE III (PROFILE VIEW)

FIGURE A-1. TYPES OF INTERACTION OF MULTIPLE CORROSION ANOMALIES

 A-9

FIGURE A-2. BEHAVIOR OF CIRCUMFERENTIALLY SPACED (TYPE I) DEFECTS

 A-10

FIGURE A-3. TYPE II LONGITUDINAL DEFECT BEHAVIOR: FULL-SCALE

EXPERIMENTAL RESULTS
 A-11

FIGURE A-4. PARAMETERS FOR LONGITUDINAL TYPE II DEFECTS (RECTANGULAR AREA)

 A-12

Method Defect Length Defect Area

1. Exact area L, A =L,d, + L2(d2-dl)

A,= L,t

2. Considering only the L2 A =L,d,

deeper flow A,= L,t

3. Considering deeper flow L2 A = L,(d,- d,)

in pipe of reduced A,= L2(t-d,)
wall thickness Hoop Stress= P x R/O-d,)

4. Equivalent area A/d, A = L,d, + L,(d,- d,)

A,= (A/d,)t

5. Ignoring deeper flaw LI A =L,d,

A,= L,t

6. Effective area Effective Effective Area

7. Parabolic area Ll A = 213 (L,d21

A, = L,t

FIGURE A-5. BASES FOR CALCULATIONS OF FAILURE PRESSURE FOR A TYPE III
 A-13

FIGURE A-6. COMPARISONS OF CALCULATIONS OF FAILURE PRESSURE

BY VARIOUS METHODS FOR A TYPE III FLAW
THIS PAGE IS INTENTIONALLY BLANK
 APPENDIX B

RSTRENG CALCULATIONS
THIS PAGE IS INTENTIONALLY BLANK