Pipeline Technology Conference, Ostend, 12-14 October 2009

Paper no: Ostend2009-006

Inherent safety factors in the

API girth weld defect tolerance
analysis – Part II
by G Wilkowski1, D J Shim1, F W Brust1, P Krishnaswamy1 ,
and D L Rudland2
1 Engineering Mechanics Corporation of Columbus (Emc2), Columbus, OH,
USA
2 US Nuclear Regulatory Commission, Rockville, MD, USA

This paper is © copyright of the organizers of the Pipeline Technology Conference held in Ostend, Belgium, on 12-14 October, 2009, and of
the author(s). No copying (either electronic or otherwise), transmission electronically, or reproduction in any way is permitted without the
specific permission of the copyright holders.
 “Inherent Safety Factors in the
API Girth Weld Defect Tolerance Analysis – Part II”
by

Wilkowski, G., Shim, D. J., Brust F. W., and Krishnaswamy, P

Engineering Mechanics Corporation of Columbus (Emc2)
3518 Riverside Drive – Suite 202
Columbus, OH 43221 (USA)
[email protected]

and
Rudland, D. L.
U.S. Nuclear Regulatory Commission
21 Church Street
Rockville, MD 20850 (USA)

ABSTRACT

The 2007 API 1104 standard for pipeline construction includes a new Appendix A for “Alternative
Acceptance Standards for Girth Welds”. Conservatism exists from several factors. The largest
conservatism is from a new empirical limit-load solution. This conservatism came from two factors. One
was not screening out the pipe test data that were conducted at conditions so the material behaved with a
low toughness. The second factor was recognizing the differences in elastically calculated bending
stresses and the fully plastic bending stress definitions. Properly screening the data and accounting for
using the more common engineering elastic bending stress definition shows a margin of 4/π to (4/π)2 on
failure stress, which can be significant in terms of flaw sizes. As a result, the new API 1104 Appendix A
(as well as CSA Z662) limit-load solution is so conservative, that both elastic-plastic fracture mechanics
(EPFM) approaches in Options 1 and 2 are not needed.
An additional conservatism of about 1.1 is due to an increase in the effective flow stress for pipes
under pressure and bending. This is a result that all the pipe tests conducted in the limit-load assessment
were conducted under pure bending and no internal pressure. As is now becoming known for strain-
based criteria, hoop stress can have an effect on reducing the stain at failure, but also increases the
effective flow stress of the material resulting in a higher failure stress. Pipes with internal pressure will
not ovalize as much as thin-walled pips under bending, which also increases the moment-carrying
capacity. Hence for pipe under pressure and bending, the inherent margins in the API analysis are 1.40 to
1.86 on failure stress.
Finally, the API Option 1 procedure was based on larger-diameter pipe results (i.e., ~30-inch
diameter), so they will be even more conservatism for smaller diameter pipes. Stress-based elastic-plastic
fracture mechanics failure criteria in other industries have shown to be a function of the pipe diameter
since the 1980’s.

INTRODUCTION

The 2007 Edition of the API 1104 standard for pipeline construction included a new Appendix A for
“Alternative Acceptance Standards for Girth Welds” [1]. The approach in this new Appendix A replaced
criteria that were developed in the late 1970’s [2]. The new criteria include a limit-load solution,

1
simplified graphs for elastic-plastic fracture toughness considerations (Option 1), and a much more
detailed failure-assessment-diagram (FAD) approach for the elastic-plastic fracture analyses of girth weld
defects (Option 2). The same limit-load equation used in the 2007 API 1104 Appendix A criterion was
also used in the Canadian standard for oil/gas pipeline weld evaluations in CSA Z662 [3]. These criteria
are stress-based criteria, not strain-based criteria. In 1979, a paper by Wilkowski and Eiber was presented
at the American Welding Society Pipeline II Conference with the same title as this paper [4], which noted
the need for a proper limit-load equation to be implemented with appropriate elastic-plastic fracture
analyses. Hence this paper, although 30 years later, is called Part II. The originally suggested
improvements are still needed.
In a recent DOE/PRCI report [5], the API limit-load criterion published in the 2007 API 1104
Appendix A was carefully examined. This work involved a follow-up of technical revisions to a draft
PRCI/DOT report [6] that was originally cited as a reference to the 2007 API 1104 Appendix A, but had a
number of technical errors that required later revision to that report [7]. In Reference [6], a limit-load
analysis was originally applied to surface cracks in girth welds in curved plate tests and past pipe tests.
That limit-load solution was based on empirically bounding all the full-scale line-pipe girth-weld defect
fracture tests that could be found, and having a correction factor developed relative to the existing Miller
limit-load solution [8], which was just another way of rewriting the original Net-Section-Collapse
solution developed by Broek in References 9 and 10. There are some very subtle but important aspects
between the Miller and Broek ways of presenting essentially the same failure criterion that affected the
conservatism in the 2007 API 1104 Appendix A criterion, so it is important to explain the fundamental
aspects.

REVIEW OF LIMIT-LOAD SOLUTIONS FOR CIRCUMFERENTIAL SURFACE CRACKS IN

PIPES

In the original Net-Section-Collapse (NSC) analysis by Broek [9], failure was assumed to occur then
the whole cross-section of the pipe becomes fully plastic. The fully plastic stress was termed the flow
stress, which in fracture mechanics analyses is typically taken as the average of the yield and ultimate
strengths, i.e.,
σf = (σy +σu)/2 (1)

The fully plastic NSC concept is illustrated in Figure 1 for a circumferentially cracked pipe, where
even the area under the surface flaw reached the same idealized flow stress value at failure as the rest of
the pipe cross-section. One can then solve for the bending moment needed to reach the fully plastic NSC
conditions by calculating the neutral axis where the summation of the forces above and below the neutral
axis are equal. The neutral axis solution can be put in an integral form equation for an arbitrary crack
shape [11], or a constant depth surface crack can be assumed to get the following simplified expression
that Broek first proposed.

Mnsc = 4 σfRm2t {sinβ – (a/t)[(sinθ)]/2} (2)

Where;
Mnsc = moment to reach NSC failure conditions
Rm = pipe mean radius
t = pipe thickness
β = angle from the neutral axis to the bottom of the pipe
a = depth of constant depth surface flaw
θ = half of the angle of the circumferential crack

2
 For the constant-depth circumferential surface crack, β is simply given below for the case of pure
bending in Equation (3a) and combined bending and axial tension induced from internal pressure (endcap
loading stress) in Equation (3b). The end-cap-induced axial stress from pressure in the second term in
Equation (3b) could be replaced by any other axial membrane stress contributions.

β = [π – (a/t)θ]/2 (3a)

β = [π – (a/t)θ]/2 – πRi2p/(4Rmtσf) (3b)

Where;
p = pressure
Ri = inside radius of the pipe

From the bending moment, one can then determine the bending stress at failure using the elastic
definition of bending stress, i.e., σ = MnscRm/πRm3t, see Equation 4 when the Net-Section-Collapse
moment equation is put in terms of elastic bending stress for a thin-walled pipe. This is the aspect that
starts to cause difficulties.
σnsc = (4σf /π){sinβ - (a/t) [sinθ]/2} (4)

In looking at graphs of bending stress at failure normalized by the flow stress of the material versus
the flaw length, the failure stress is defined in Equations 1 thru 4 and is depicted in Figure 1. Note in
Figure 2 that as the crack length approaches zero, the bending stress at failure approached a value of 1.27
or 4/π times the flow stress. For pipes with low enough R/t so they do not fail by local buckling, this
result of an unflawed pipe approaching the fully plastic failure stress was found to be reasonable when
compared to unflawed pipe test data [12].
The NSC equations and flow stress definition have been validated by hundreds of pipe tests on
tougher nuclear piping steels [12, 13, 14] where over 300 circumferentially cracked pipe tests were
examined. The pipes were screened to ensure the toughness was high enough that limit-load should be
satisfied, and the statistical relationship of the actual material strengths (yield and ultimate) at the pipe test
temperature were related to the flow stress calculated for the qualifying pipe tests [15]. Hence, the NSC
analysis and its limits were well qualified in the development of the ASME Boiler and Pressure Vessel
Section XI pipe circumferential flaw evaluation criteria [16,17] and other standards [18, 19, 20].

Figure 1 Illustration of Net-Section-Collapse assumption of flow stress being reached in the entire
pipe cross-section

3
 1.4

Predicted bending failure stress/ Flow stress .

NSC using elastic bending stress

1.2

0.8

0.6
NSC a/t = 0.2

NSC a/t = 0.4

0.4
NSC a/t = 0.6

0.2 NSC a/t = 0.8

NSC a/t = 1.0

0
0 0.1 0.2 0.3 0.4 0.5

Normalized crack length, θ/π

Figure 2 Failure assessment graphs for Original Net-Section-Collapse failure stress for a
circumferentially cracked pipe under pure bending

After the Broek NSC collapse solution was well established and accepted in international codes and
standards, Miller wrote a paper comparing different limit-load solutions for cracks in structures [8]. In
the work by Miller, he used the NSC equation, but preferred that in the graph of failure stress/flow stress
versus crack length, as in Figure 2, that the solution approach a value of 1.0 rather than 4/π, which would
be the same trend from pure tension loading on a cracked pipe or flat plate. Hence, Miller effectively
used the fully plastic definition of the bending stress (σbending = MRm/4Rm3t) rather than the elastic solution
(σbending = MRm/πRm3t) used in all common engineering stress analyses. The Miller equation differs from
the Net-Section-Collapse equation by a factor of 4/π when talking in terms of “bending stress” without
qualifying the definition of the bending stress. This is a subtle but important point as will be shown in
the following comparisons to line-pipe experimental data.
The Wang modification [6] of the Miller equation has a function that gives a lower predicted failure
stress as the crack size increases. This modification was a curve-fit correction to experimental results that
will be examine further later in this paper. The results is that the Wang modification becomes lower than
the Miller equation by another factor of π/4 for longer cracks. The Wang modification on the Miller
solution is given in the first term in Equation 5a. Hence, the Wang equation is lower than the Miller
equation by π/4 (79%) if the fully plastic bending stress definition is used in the Wang equation. If the
elastic bending stress is used in the Wang equation, then it is lower by an additional factor of π/4 or
(π/4)2 (a total of 62%) from the NSC predicted failure stress.

σ L ⎡π 2.5 ⎤ ⎡ ⎛ ( a / t )(θ / π ) ⎞ ( a / t ) sin (θ ) ⎤ if ( a / t )(θ / π ) < 0.05 (5a)

= ⎢ + 385[(0.05 − ( a / t )(θ / π )] ⎥ ⎢cos⎜ ⎟− ⎥
σ f ⎣4 ⎦⎣ ⎝ 2 ⎠ 2 ⎦

4
 σ L π ⎡ ⎛ (a / t )(θ / π )π ⎞ (a / t ) sin ((θ / π )π ) ⎤
= cos⎜ ⎟− if ( a / t )(θ / π ) ≥ 0.05 (5b)
σ f 4 ⎢⎣ ⎝ 2 ⎠ 2 ⎥
⎦

A graphical comparison of the NSC, Miller, and Wang limit-load solutions for circumferential
surface-cracked pipes is illustrated in Figure 3 using failure stress/flow stress versus the normalized crack
length (θ/π) with curves for different normalized surface crack depths. Again, the only difference
between the NSC and Miller solutions is the use of elastic or fully plastic stress, whereas the Wang
solution uses Equation 5a as an additional modification on the Miller fully-plastic solution. One can see
that if the subtle difference in the definition of the bending stress is not accounted for, the failure stress
predictions could be significantly different.

1.4
NSC using elastic bending stress
Predicted failure stress/ Flow stress.

1.2
Miller using fully plastic
bending stress
1

0.8

0.6

Wang modification with fully plastic bending stress

0.4

NSC a/t = 0.1 NSC a/t = 0.3 NSC a/t = 0.5

0.2 Miller a/t = 0.1 Miller a/t = 0.3 Miller a/t = 0.5
Wang a/t = 0.1 Wang a/t = 0.3 Wang a/t = 0.5

0
0 0.1 0.2 0.3 0.4 0.5

Normalized crack length, θ/π

Figure 3 Comparison of NSC, Miller, and Wang limit-load solutions for circumferential surface
cracks in pipes under pure bending

The Wang Limit-Load Solution

In a recently completed DOT/PRCI project on development of girth weld defect tolerance

approaches, a limit-load solution was implemented from work by Wang [6]. As noted earlier, this same
limit-load equation was implemented into the 2007 API 1104 Appendix A as well as CSA Z662 [1,3]. In
the earlier draft version of the DOT/PRCI report [6], this limit-load solution was applied to mini-wide
plate tests. It was subsequently noticed that there was a large difference between the limit-load equations
validated for mini-wide-plate tests versus that used for pipes in bending, see Figure 4. Forty mini-wide
plate tests were compared to the flat-plate limit-load solution in Reference [7] with the ratio of

5
experimental/predicted stress being 1.07 with a coefficient of variance (standard deviation/mean) of 3
percent. This amount of error is typically seen in predictions of good limit-load solutions [14]. However,
when using the Wang-modified/Miller limit-load equation, the experimental/predicted failure stress ratio
was 1.43 which was a factor of 1.33 more conservative, and well outside of normal predictions. Since
there are other safety factors used in the API code, it was desirable to eliminate this degree of
conservatism.
The original technical basis of the Wang modified limit-load equation came from examining a
database of full-scale linepipe fracture tests with girthweld defects. The pipe tests were under pure
bending (no internal pressure), and the experimental stress was the elastically calculated bending stress
not the fully plastic definition. The original database was compiled by Rosenfeld [21] for PRCI;
additional data were input as part of the efforts in [7]. The final data set included;
• Glover data from Welding Institute of Canada [22,23],
• Erdogan [24],
• Worswick, Pick and Glover [25*],
• Hopkins EPRG tests [26],
• Wilkowski original API validation tests [27], and
• Wilkowski repair groove pipe tests (small and large-diameter pipe tests) [28].

An additional difficulty was that the database of these pipe tests was not first screened by Wang to
use only limit-load failures. Consequently, the initial assessments included even brittle fracture girth
weld tests. In examining the entire linepipe girth weld database, it can be seen that all but one pipe test
datapoint with a CTOD value down to 0.001 inch (0.025 mm) were bounded by the Wang empirical limit-
load correction, see Figure 5. When properly screening out the non-limit-load failures as illustrated in
Figure 6, it was shown that there was a safety factor of at least 4/π on the failure stress, see Figure 7.
However, even the results in Figure 7 are not consistent with limit-load solutions in other
international circumferentially surface-cracked pipe defect acceptance criteria [16,17]. This aspect was
examined further by using the full-scale nuclear pipe test data in CIRCUMCK [29] which contained over
300 full-scale pipe tests. The data in that database were selected for only circumferential surface cracks
in pipes in pure bending. Those tests were then further screened so that they should be limit-load failures
and not EPFM or brittle factures. The screening was done using the Dimensionless Plastic-Zone
Parameter (DPZP) Screening criterion from Reference [30], where limit-load conditions are satisfied from
Equation (6). This criterion also accounts for constraint effects on the toughness from high-constraint
specimen data from CT or bend specimens.

(2JiE)/(π2Dσf) > 0.2 (6)

Where,
Ji = toughness at crack initiation using the J-integral fracture mechanics parameter
E = elastic modulus
D = pipe diameter
σf = flow stress = (σy + σu)/2

The 64 full-scale nuclear pipe fracture tests with circumferential surface cracks are compared to the
NSC analysis in Figure 8. This figure shows that the average experimental/NSC predicted failure stress
was 1.105 which means that the NSC analyses for all these tests was 10.5 percent conservative. The
standard deviation was 9.6% of the average value, which is a reasonably tight scatterband. Furthermore,
the mean minus one standard deviation values was then 1.009, showing that the NSC analyses for this
data set was slightly conservative but still reasonably accurate.

*
Note, the official PRCI report for Reference 25 misspelled the names of Worswick and Pick as “Worlock and
Piok” on the report cover page.

6
 If the Wang-modified limit-load equation was used with the elastic bending stresses, then that same
dataset would have an average experimental/predicted maximum load of 1.79, which is clearly overly
conservative.
As a result of the conservatism in the Wang limit-load solution used in the 2007 API 1104 Appendix
A criteria, in reality the simplified EPFM approach in Option 1 graphs were not needed, and the FAD
curve approach in Option 2 was also not needed, see Figure 5.

Limit-load solutions for semi-elliptical

surface flaws in flat plate in tension
1.0
Failure stress/flow stress

0.8

0.6

Limit-load solutions for

circumferential surface flaws
0.4 in pipes under bending

a/t=0.2 in Plate a/t=0.2 in pipe

0.2 a/t=0.4 in Plate a/t=0.4 in pipe
a/t=0.6 in Plate a/t=0.6 in pipe
a/t=0.8 in Plate a/t=0.8 in pipe
0.0
0% 10% 20% 30% 40% 50%

Crack length, % of circumference

Figure 4 Comparison of proper limit-load solution for a mini-wide plate test, and Wang limit-load
solution for a circumferential surface crack in a pipe

7
 1.6
Using unpressurized pipe bending test
Eliminated a/t > 0.58 or θ/π > 0.125
1.4 20" < pipe diameter < 42"

Net-Section-Collapse analysis

1.2 Wang-modification
to NSC limit-load
Miller 4/π higher relative to Wang-
1.0
Miller

modification for larger cracks

CTOD = 0.005"
/ σL

0.8 Wang-modification
CTOD = 0.002"
to Miller limit-load
Exp

CTOD = 0.001"
σL

0.6 CTOD = 0.001"

CTOD = 0.002"
CTOD = 0.003"
CTOD = 0.004" These data meet Wang-modified/Miller limit-load criteria
0.4
CTOD = 0.005"
CTOD = 0.009"
0.2 CTOD = 0.022" These data meet new limit-load screening criteria
CTOD > 0.050"

0.0
0.00 0.01 0.02 0.03 0.04 0.05 0.06 0.07 0.08 0.09 0.10
( a/t) (θ/π)
Figure 5 Comparison of all available full-scale linepipe girth weld defect fracture experimental
data to various limit-load analyses

1.6
Blunt flaw tests with a/t > 0.5

1.4

1.2
Miller 4/ π higher than Wang
modification for larger cracks
1.0
Miller
/ σL

Know n ductile
0.8 EPFM failures
Wang modification
Exp
σL

0.6 Erdogan data

Glover data (some ductile some brittle)
Hopkins data
0.4 Known brittle initiation failures
Wilkow ski - 4'' dia repair groove data
(Tests at -35, -55, -60, -95C)
Wilkow ski - 6'' dia repair groove data

0.2 Wilkow ski - 30" dia repair groove data

Wilkow ski - API data
Worsw ick data
0.0
0.00 0.01 0.02 0.03 0.04 0.05 0.06 0.07 0.08 0.09 0.10 0.11
(a/t )(θ/π)
Figure 6 Illustration of invalid limit-load data points from Hopkins and Wilkowski data (Note;
virtually all the Glover data were not valid limit-load cases, as well as some of the
Worswick tests.)

8
 1.6
Using unpressurized pipe bending test data -
Eliminating brittle and EPFM failures,
1.4 Eliminated a/t > 0.5 or beta > 0.125,
NSC limit-load
1.2 Wang-modification
to NSC limit-load
Miller 4/π higher relative to Wang-
1.0
Miller

modification for larger cracks

/ σL

0.8
Wang-modification
Exp

to Miller limit-load
σL

0.6
Erdogan data
Hopkins valid data
0.4
Wilkow ski - 6'' dia repair groove data
Wilkow ski - 30" dia repair groove data
0.2 Glover CTOD = 0.009 inch
Worsw ick data

0.0
0.00 0.01 0.02 0.03 0.04 0.05 0.06 0.07 0.08 0.09 0.10 0.11
( a/t )(θ/π )
Figure 7 Comparison of valid limit-load test data to various limit-load equations

1.6
Experimental/NSC predicted maximum load

1.4

1.2
Average

0.8

Average minus 1 Std Dev

0.6

0.4

0.2

0
0 10 20 30 40 50 60 70 80 90 100

Crack length, percent of circumference

Figure 8 Comparison of experimental/predicted failure stress as a function of crack length from

64 circumferentially surface-cracked nuclear pipe experiments screened to include only
valid limit-load failures

9
Effect of Internal Pressure

There are additional contribution to increase the conservatism for pipes under pressure and bending
loads. This can be an important consideration since all the limit-load and EPFM corrections in API 1104
Appendix A and CSA Z662 were based on only pipe tests under pure bending. Two aspects can come
into play with combined pressure and bending loads. The first is that pipe under high bending loads
without internal pressure will ovalize. This ovalization reduces the pipe’s moment of inertia and lowers
the moment-carrying capacity. The magnitude of the reduction of the load-carrying capacity is sensitive
to the pipe R/t ratio, strength, and applied loads. Empirical corrections were suggested in Ref [12], and is
shown in Figure 9. Internal pressure will have the effect of forcing the pipe to maintain a circular cross-
section (a basic assumption in all the limit-load solutions), so the moment-carrying capacity will increase
for thinner pipes (as in the case of most linepipe). Of course, any axial stress induced from the pressure
will cause the pipe cross-section to reach plastic conditions earlier than under just pure bending. Long-
distance buried pipelines will not have the same longitudinal stress from pressure as endcapped shorter
sections of pipe runs, so this is an advantageous aspect.
Another interesting consideration for the effects of hoop stress on stress-based limit-load criteria
comes from on-going research efforts on strain-based criteria. In the strain-based criteria development, it
is generally agreed upon that the hoop stress reduces the strain capacity of flawed pipes [5]. However, as
pointed out in Reference [5], the hoop stress raises the load capacity of flawed pipes for stress-based
analysis. This comes about from an increase in the effective flow stress of the material under biaxial
loading, where the von Mises yield surface criterion can be used to determine the increased flow stress.
Finite element (FE) analysis illustrating this effect is shown in Figure 10. In this example, FE analyses
were conducted with and without hoop stress in the pipes, which showed an increase in the maximum
load-capacity. A simple von Mises correction was also made for the hoop stress effect to the non-hoop
stress pipe case, and that agreed excellently with the FE results with hoop stress. From this work, it can
be shown that for pipes with hoop stress loading of 60% to 80% SMYS, the flow stress will be raised by
about 10 to 15 percent, which increases the axial load-carrying capacity by that amount.
To further assess the significance of combined pressure and bending loads, the few nuclear pipe tests
with circumferential surface cracks, under pressure and bending loads, and with sufficient toughness to
reach limit-load by Equation 3 were added to the earlier graph of just pure bending tests. Although there
are not nearly as many pipe tests, the average value of the experimental/predicted maximum failure stress
(bending and axial stress contributions included) were larger than just pure bending tests by ~20 percent,
see Figure 11. This 20-percent factor includes both the hoop stress effects on pipe ovalization and the
increased flow stress under combined loading.

10
Figure 9 Graph of ratio of experimental-to-NSC predicted failure stress as a function of radius-to-
thickness ratio for unpressurized circumferentially surface-cracked nuclear pipe tests
18,000

16,000

14,000

12,000
Axial Force, N

10,000

8,000

6,000

FE with no Internal Pressure

4,000
FE with Internal Pressure

2,000 Simple von Mises correction after yield on FE

no-pressure case axial forces

0
0 10 20 30 40 50 60 70 80 90 100

Displacement, mm

Figure 10 Illustration of hoop stress effect on axial load-carrying capacity for 36-inch diameter by
0.5-inch thick X100 pipe with hoop stress of 60% SMYS and circumferential flaw 3-mm
deep and 25-mm long

11
 Experimental/NSC predicted maximum load
1.6
Average minus 1 Std Dev
pressure & bending
1.4
Average - pressure & bending

1.2
Average - pure bending
1

0.8
Average
0.6
minus 1 Std Dev
pure bending

0.4

Pure bending pipe data

0.2
Pressure and bending pipe data

0
0 10 20 30 40 50 60 70 80 90 100

Crack length, percent of circumference

Figure 11 Effect of combined pressure and bending on NSC predicted failure stresses for
circumferentially surface-cracked nuclear pipe

Effect of Pipe Diameter

The 2007 API 1104 Appendix A elastic-plastic fracture mechanics (EPFM) analyses are presented in
a simplified manner in the Option 1 approach and in a more detailed FAD or Failure Assessment Diagram
approach in Option 2. These analyses assume the maximum load-carrying capacity of the flawed pipe
first comes from the limit-load solution, and if the toughness is not sufficient, the allowable load-carrying
capacity is reduced from the limit-load solution. As noted earlier in this paper, the limit-load solution in
the current API 1104 and CSA-662 standards are overly conservative, so that the EPFM factors will add
additional conservatism.
Compounding this effect is that the Option 1 approach came from analyses of the large-diameter pipe
tests, average pipe diameter was ~32-inches. It has been established for decades that the elastic-plastic
correction factor is also a strong function of the pipe diameter. That is, as the pipe diameter increases, it
is more likely that EPFM failure will occur. Conversely, smaller-diameter pipes are more likely to fail at
limit-load conditions. This happens because the plastic-zone developed from the crack location needs to
be large enough to go to the pipe neutral axis, which is easier to accomplish in small-diameter pipe. This
is basic assumption in the Net-Section-Collapse solution, see Figure 1. As an example, in the ASME
Boiler and Pressure Vessel Code, the pipe flaw evaluation criteria in Section XI have a simplified EPFM
correction factor on the limit-load solution, somewhat akin to the Option 1 approach in the 2007 API
1104 Appendix A criteria; however, there is a relative simple correction factor as a function of the pipe
diameter, where

Limit load/EPFM max load = f(pipe diameter, material toughness, material strength) (4)

12
which is termed a Z-factor. Many technical papers and projects have been conducted to establish the
appropriate Z-factor relationships for flaws in different types of welds as a function of pipe diameter
[14,31,32]. One approach to include pipe diameter, material toughness, and material strength in a simple
one-line equation was called the Simplified Dimensionless Plastic-Zone Parameter analysis [30]. Figure
12 shows this empirically derived relationship, where if the DPZP parameter of 2EJi/(π2Dσf2) is greater
than 0.2 then limit-load conditions will exist. A relatively simple equation exists from this work in terms
of using the J-integral fracture mechanics parameter, but that could easily be expressed in terms of the
CTOD parameter as well. One can also see from the data trends presented in Figure 5 that CTOD failure
loci can be established for the large-diameter linepipe tests in terms of CTOD, and then the diameter
effects can also be accounted from using the DPZP parameter approach. This would enable simplifying
the Option 1 and 2 approaches in Appendix A and obtaining criteria that are not as penalizing.

Figure 12 Simplified DPZP analysis for circumferential surface cracks in pipes

MAGNITUDE OF EFFECT ON FLAW SIZES

The effect of the magnitude of the difference in just the limit-load equation in the 2007 API 1104
Appendix A on the allowable flaw sizes is illustrated in Figure 13. In Figure 13(a), the Appendix A flaw
sizes are compared to the maximum limits in Section 9 of the 2007 API 1104 main body, while Figure
13(b) the smallest correction factor of 4/π was applied to the limit-load solution. The applied stress was
100% SMYS for different grades of pipes, and the CTOD value was greater than 0.25 mm (basically the
limit-load solution). As one can see, there is a huge shift in the calculated curves relative to the maximum
workmanship flaw sizes.

13
 0.50
Applied stress = SMYS
0.45 CTOD>0.25mm

0.40 X52
X60
0.35 X70
0.30 X80
X100
a/t

0.25

0.20
API 1104
0.15
Workmanship Allowable
0.10 Flaw Sizes

0.05

0.00
0 0.025 0.05 0.075 0.1 0.125

Allowable length/Pipe Circumference

(a) Using recent 2007 API 1104 Appendix A limit-load equation
0.5
Applied stress = SMYS
0.45 CTOD>0.25mm
X52
0.4
X60
0.35 X70
0.3 X80
X100
a/t

0.25

0.2
API 1104
0.15 Workmanship
0.1 Acceptable
Flaw size
0.05

0
0 0.02 0.04 0.06 0.08 0.1 0.12 0.14

Allowable length/Pipe Circumference

(b) Using limit-load equation with 4/π correction
Figure 13 Differences in API workmanship allowable flaw sizes compared to API Appendix A
criteria and accounting for factor of 4/π in the limit-load solution

14
CONCLUSIONS

The following major conclusions may be drawn from the above work:
1. For large-diameter pipes under hoop stress and bending loads, the inherent margins in the API
analysis is a combination of the two conservatisms discussed in the paper and would range from a
safety factor of 1.40 to 1.86 on the axial failure stress.
2. For pipe being loaded under high bending stresses during construction with no internal pressure
(i.e., lowering pipe into a ditch, or offshore pipe laying), there is not increased margin due to the
effect of hoop stresses. .
3. For deep-water pipe laying, in the sag bend region of the pipeline, the external pressure can be
detrimental for; (i) stress-based criteria, and possibly (ii) beneficial for strain-based criteria.
4. The above efforts emphasize the need for consistency between stress-based and strain-based girth
weld defect acceptance criteria.
5. Finally, since the Option 1 approach in API 1104 is based on larger diameter pipes, it will add
additional conservatism to the above factors for pipes smaller than ~32 inches. Simplified EPFM
correction factors as a function of pipe diameter exist in other industries for more than 20 years
and should be adopted in the API criterion as well.

REFERENCES

[1] API Standard 1104, “Welding of Pipelines and Related Facilities,” 20th Edition July 2007
addendum.
[2] API Standard 1104, “Welding of Pipelines and Related Facilities,” 17th Edition, September 1988.
[3] CSA Z662-03 “Oil and Gas Pipeline Systems”, published by Canadian Standards Association.
[4] G. M. Wilkowski and R. J. Eiber, “Inherent Safety Factors in the API Girth Weld Defect Tolerance
Analysis,” presented at the American Welding Society Pipeline II Conference, Houston, Texas,
February 12-14, 1979.
[5] Wilkowski, G. Rudland, D., Shim, D. J., Brust, F. W., and Babu, S., “Advanced Integration of Multi-
Scale Mechanics and Welding Process Simulation in Weld Integrity Assessment,” DOE Award
Number: DE-FC36-04GO14040, published October 2008.
[6] Y. Y. Wang and others, “A Comprehensive Update in the Evaluation of Pipeline Weld Defects,”
U.S. DOT Agreement No. DTRS56-03-T-0008, PRCI Contact No. PR-276-04503, draft report
unpublished, May 2006.
[7] Wilkowski et al., “A Comprehensive Update in the Evaluation of Pipeline Weld Defects,” U.S. DOT
Agreement No. DTRS56-03-T-0008, PRCI Contact No. PR-276-04503, final report submitted for
approval Feb 2008.
[8] Miller, A., "Review of Limit Codes of Structure Containing Defects," International Journal of
Pressure Vessels and Piping, Vol. 32, 1988, pp. 191-327.
[9] Kanninen, M. F., Broek, D., Marschall, C. W., Rybicki, E. F., Sampath, S. G., Simonen, F. A., and
Wilkowski, G. M., “Mechanical Fracture Predictions for Sensitized Stainless Steel Piping with
Circumferential Cracks,” EPRI Report NP-192, September 1976.
[10] Kanninen, M. F., Zahoor, A., Wilkowski, G., Abou-Sayed, I., Marschall, C., Broek, D., Sampath, S.,
Rhee, H., and Ahmad, J., “Instability Predictions for Circumferentially Cracked Type 304 Stainless
Steel Pipes Under Dynamic Loadings,” EPRI Report NP-2347, April 1982.
[11] Rahman, S., and Wilkowski, G. M., “Net-Section-Collapse Analysis of Circumferentially Cracked
Pipes: Part I - Arbitrary-Shaped Cracks and Generalized Equations,” ASME PVP Conference, PVP
Vol. 350, pp 45-62, July 1997.

15
[12] Krishnaswamy, P., Scott, P., Mohan, R., Rahman, S., Choi, Y. H., Brust F., Kilinski, T., Ghadiali,
N., Marschall, C., and Wilkowski, G., “Fracture Behavior of Circumferential Short-Surface-Cracked
Pipe,” NUREG/CR-6298, November 1995.
[13] Brust, F., Scott, P., Rahman, S., Ghadiali, N., Kilinski, T., Francini, R., Marschall, C., Muira, N.,
Krishnaswamy, P., and Wilkowski, G., “Assessment of Short Through-Wall Circumferential Cracks
in Pipes - Experiments and Analyses,” NUREG/CR-6235, April 1995.
[14] Wilkowski, G. M., Olson, R. J., and Scott, P. M., “State-of-the-Art Report on Piping Fracture
Mechanics,” U.S. Nuclear Regulatory Commission report NUREG/CR-6540, BMI-2196, February
1998.
[15] G. Wilkowski, M. Fleming, D. Rudland, and P. Scott, “Statistical Determination of Flow Stress from
the CIRCUMCK Piping Fracture Database,” 1998 ASME PVP Conference.
[16] “Evaluation of Flaws in Austenitic Steel Piping,” (Technical basis document for ASME IWB-3640
analysis procedure), prepared by Section XI Task Group for Piping Flaw Evaluation, EPRI Report
NP-4690-SR, April 1986.
[17] “Evaluation of Flaws in Ferritic Piping,” EPRI Report NP-6045, prepared by Novetech Corporation,
October 1988.
[18] API RP 579, “Fitness-for-Service,” First Edition, January 2002.
[19] “Assessment of the Integrity of Structures Containing Defects”, R6 Revision 4, April 2001.
[20] RCCM, “Regles de Conception et de Construction des Materiels Mecaniques des Ilots Nucleaire
REP”, AFCEN, Paris, 1993.
[21] Rosenfeld, M. J., “Serviceability of Corroded Girth Welds,” PRCI Catalogue No. L51742, May
1996.
[22] Glover, A. G., “Effects of Real Defects on Girth Weld Fracture Behavior,” PRCI Catalogue No.
L51457e, June 1984.
[23] Glover, A. G., “Small and Full Scale Fracture of Thick Section Girth Weldments,” PRCI Catalogue
No. L51533e, March 1987.
[24] Erdogan, F., “Theoretical and Experimental Study on Fracture in Pipelines Containing
Circumferential Flaws,” DOT Report DOT-RC-82007, by Lehigh University, August 1982.
[25] Worolok, M. J., Piok, R. J., and Glover, A. G., “The Effect of Misalignment on the Fracture
Behavior of Girth Welds,” PRCI Catalogue No. L51489e, August 1985.
[26] Hopkins, P., Demofonti, G., Knauf G., and Denys, R., “An Experimental Appraisal of the
Significance of Defects in Pipeline Girth Welds,” Paper 20, Proceedings of the EPRG/NG18 Eighth
Biennial Joint Technical Meeting on Line Pipe Research, Paris, May 14-17, 1991.
[27] Wilkowski, G. and Eiber, R. “Evaluation of the Inherent Safety Factors in the Tentative API Girth
Weld Defect Tolerance Approach,” American Gas Association, Catalog No. L51385, June 1977.
[28] Wilkowski, G. and Eiber, R., “Evaluation of Tensile Failure of Girth Weld Repair Grooves in Pipe
Subjected to Offshore Laying Stresses,” ASME Journal of Energy Technology, March, 1981, pp. 48-
57.
[29] Scott, P. M. and Wilkowski, G. M., “Development and Application of a Database of Pipe Fracture
Experiments,” in Fatigue and Fracture - 1996 - Volume 1, PVP - Vol. 323, July 1996, pp. 13-26.
[30] Wilkowski, G., and Scott, P., “A Statistical Based Circumferentially Cracked Pipe Fracture
Mechanics Analysis for Design on Code Implementation,” Nuclear Engineering and Design, Vol.
III, 1989, pp. 173-187.
[31] Wilkowski, G. M., Scott, P. M., Krishnaswamy, P., and Choi, Y. H., “Proposed Modification of the
ASME Section XI Pipe Flaw Evaluation Criteria Based on New Surface-Cracked,” in Fatigue and
Fracture - 1996 - Volume 1, July 1996, pp. 51-64.
[32] G. Wilkowski, H. Xu, D.-J. Shim, and D. Rudland, “Determination of the Elastic-Plastic Fracture
Mechanics Z-factor for Alloy 82/182 Weld Metal Flaws for Use in the ASME Section XI Appendix
C Flaw Evaluation Procedures,” Proceedings of ASME-PVP 2007, paper PVP2007- 26733, July 22-
26, 2007.

16