You are on page 1of 6

International Pipeline Conference Volum e I

ASM E 1998

IPC1998-2033

FATIGUE CURVES FOR DAMAGE CALCULATIONS FOR A DENTED


AND OVALLED SECTION OF THE TRANSALASKA PIPELINE SYSTEM

Glen R. Stevick
Principal Mechanical Engineer
Berkeley Engineering And Research, Inc.

James D. Hart
Principal
SSD, Inc.

Bill Flanders
Senior Engineer
Alyeska Pipeline Service Company

ABSTRACT This paper describes the design and decision S-N relationships
that were developed during the Thompson Pass investigation to
During the summer o f 1996, the TransAlaska Pipeline System (TAPS) estimate fatigue damage. These curves are a combination of the AWS
experienced vibrations near Thompson Pass, about 25 miles north of A fatigue curve and an adjusted version o f the ASME Section VIII,
Valdez, Alaska. The vibrations could, on occasion, be detected by Division 2 design fatigue curve. These curves were used together
residents living near the pipeline at the bottom of the pass. Close to with a multi-axial fatigue model to compute fatigue damage due to
the source of the phenomena, small bushes could be seen moving in pipe stresses caused by pressure pulse cycling as well as that due to
response to the seismic shocks and a noise similar to a mortar firing operational startup and shutdown cycles.
in the distance could be heard. Alyeska Pipeline Service Company
initiated an extensive investigation and quickly determined that the INTRODUCTION
seismic shocks were a result of pressure pulses originating near the
slackline-packline interface. This only occurred when the slackline- The calculation of fatigue damage is an inexact science and, for new
packline interface was positioned near a terraced portion of the designs conservative assumptions are made in order to ensure that the
pipeline topography on the downstream side o f the pass. This probability of a crack initiating in a component during its service life
knowledge allowed Alyeska Pipeline to control the pulsations by back is extremely small. This is done, among other things, by using design
pressuring the pipeline to move the slackline-packline interface well S-N curves that incorporate a conservative margin of safety. Our
above the terrace location. concern for assessing the state of the pipe at Thompson Pass was
whether there was a significant probability that a crack had or would
One key aspect of the project was an extensive analytical investigation initiate, not whether there was a minute probability. The purpose of
of a dented and ovalled section of the pipeline near the origin of the the assessment was to guide the decision making process regarding
pressure pulses. The main concern at this location was that, as each short and long term actions needed to ensure the integrity of the
pressure pulse passes, the ovalled and dented pipe section tends to pipeline. If the cumulative fatigue damage at a dent location equals the
reround causing the pipe wall to flex a small amount. Since the damage that would be allowed for a new design (typically a
pulses occurred frequently under certain flow conditions, there was a cumulative damage of 1.0, based on a design S-N curve), this does
concern for possible fatigue damage to the pipe steel. The locations of not mean that the pipeline is on the verge o f failure, or even on the
maximum stress range were estimated to be near the 6:00 position on verge of crack initiation. For this reason, fatigue damage calculations
the outside surface of the pipe at the periphery of the dents in the X65 were made using two different S-N curves, namely a design curve
base metal (Stations 40959+28 and 40959+40 in Alyeska and a decision curve. The differences between these curves can be
terminology). explained as follows.

Copyright 1998 by ASME

Downloaded From: http://proceedings.asmedigitalcollection.asme.org/ on 09/12/2017 Terms of Use: http://www.asme.org/about-asme/terms-of-use


DEFINING THOMPSON PASS "DESIGN FATIGUE CURVE At Thompson Pass, the vast majority of the applied stress cycles were
in the high cycle region. The calculated remaining lives of the subject
To obtain a design level curve, experimental results are used, giving dents were highly dependent on the level of the endurance limit
the measured number o f cycles to failure at various stress levels. There (allowable stress amplitude for an infinite number of cycles) in the
is always a great deal o f scatter in the results, and for any given TAPS design fatigue curve. The AWS A-curve endurance limit, 25
number o f cycles, a substantial factor of safety is applied to obtain the ksi range, was used in the TAPS curve.
allowable design stress. As will be described later, we developed a
design curve for the TAPS base metal utilizing the American Welding Supporting the use of the AWS A-curve in the high cycle region, an
Society (AWS) A-Curve (D l.l 1996) for high cycle stress levels, an almost identical endurance limit can be determined by adjusting the
extension of the A-curve for midrange stress levels and a Bi-axial ASME endurance limit for similar surface roughness, material
fatigue curve proposed in the literature for the low cycle life portion. ultimate strength and design factors. The ASME endurance limit for
The resulting composite fatigue curve is shown in Figure 1 with the an 85 ksi ultimate strength material is 13.5 ksi. Removing the design
American Society of Mechanical Engineers (ASME) fatigue curve factor of 2.0 on stress gives a mean data estimate of 27 ksi amplitude.
(ASME, 1995) and Bi-axial design fatigue curves proposed in the Adjusting for surface roughness by dividing by 0.9 for fine
literature (Rahka, 1993). Note that above 10,000 cycles, the curve is ground/commercial polish and multiplying by 0.63 for hot rolled parts
the same as the AWS A-curve. (Juvinall, 1967) gives an adjusted mean data estimate of 18.9 ksi.
Applying a safety factor of 1.5, the multiplicative equivalent of two
The AWS A-curve is based on fatigue tests of rolled steel beams standard deviations for the AWS A-Curve data, gives an allowable
(NCHRP, 1982). The residual stresses from the rolling process are of stress amplitude of 12.6 ksi. This is almost identical to the AWS A-
yield magnitude and cause the fatigue specimens to experience curve fatigue limit of 25 ksi range.
maximum mean stress effects. In addition, the rough as-rolled surface
would typically have more surface flaws than rolled plate of the type At the low cycle end of the design fatigue curve, an allowable strain
used for the TAPS pipe. Thus, the A-curve should be somewhat amplitude of 1 percent (300 ksi equivalent stress) was chosen for 1
conservative for the TAPS pipe. All o f the AWS fatigue curves are cycle to failure based on the proposed bi-axial fatigue design curve
obtained by applying a design factor corresponding to two standard and high bi-axial fatigue data given and described in (ASME, 1995).
deviations on stress to the mean o f the experimental results (the stress The cited bi-axial fatigue data from (Ives, 1966) and (Rahka, 1992)
at crack initiation). For the data the AWS A-curve is based on, see clearly show a significant drop in fatigue resistance with increasing bi-
Figure 2, this is approximately equal to a factor of 1.5 on stress. axiality. To fill in the mid-region, the AWS A-curve was extended
(using its original slope on a log-log plot) from its low-cycle end point
Comparison with the ASME S-N curve for pipe and pressure vessel at 10,000 cycles (60 ksi range) to 1000 cycles with a corresponding
materials provides additional support for use of the AWS A-Curve for allowable stress range of 88 ksi. The points determined at 1 and 1000
the high cycle portion of the TAPS design fatigue curve. Fatigue data cycles were then connected with a straight line on a log-log plot to
for the ASME curves were initially generated using small, smooth, complete the Thompson Pass (TP) design fatigue curve.
unwelded specimens tested in air under strain controlled conditions.
Equivalent stress amplitude data was obtained by multiplying the strain DEFINING THOMPSON PASS DECISION FATIGUE CURVE
range data by one-half Youngs modulus. Mean fatigue curves were
then obtained by drawing best fit curves through the equivalent stress It can not be emphasized too strongly that a calculated cumulative
amplitude data (Cooper, 1992). Mean stress effects were conservatively damage of 1.0 for a design S-N curve does not mean that there is
accounted for using a modified Goodman Diagram approach for stress cause for alarm, or that an immediate repair is needed. For the
amplitude values below the material yield strength (ASME, 1969). The purposes o f such short term decisions, it is more reasonable to base the
ASME fatigue design curves were determined by applying a factor 2.0 cumulative damage calculations on an S-N curve with a smaller factor
on stress or 20 on life, whichever was more conservative. The factor of of safety.
2.0 on stress governs in the high cycle region and 20 on life governs in
the low cycle region. These factors are not safety factors. The factor If we assume that the factor o f safety o f 1.5 in the design fatigue curve
of 20 was originally based on engineering judgement taking into account corresponds to the mean value minus two standard deviations, one
that the base test data was determined using small, smooth, unwelded standard deviation is equal to 0.167 o f the mean (1.0 - 2 (0.167) =
specimens. The factor of 20 was made up of the following sub-factors: 0.667, or a 1.5 safety factor). A more reasonable factor o f safety for a
decision level S-N curve is the mean minus one standard deviation.
Data Scatter (minimum to mean) 2.0 If the calculated cumulative fatigue damage using this curve were to
Size Effect 2.5 exceed 1.0 at some location in the pipeline, this would mean that there
Surface finish, atmospheric, etc. 4.0 is a significant probability of crack initiation at that location (about
one chance in six), again with a smaller probability of a crack large
The factor of 2.0 on stress was originally determined to provide an enough to cause a leak. In this case, we believe there would be cause
equivalent margin in the high cycle region. The fatigue data currently for more seri ous short term concern, and an immediate repair would
being utilized for the ASME Curves are based on larger specimens. probably be advisable.

Downloaded From: http://proceedings.asmedigitalcollection.asme.org/ on 09/12/2017 Terms of Use: http://www.asme.org/about-asme/terms-of-use


We have thus calculated cumulative damage values using two S-N In actual application, the Von Mises mean and alternating stresses are
curves, one at a design level and one at a decision level. The determined from finite element analysis results for the application of
decision level fatigue curve is obtained by multiplying the design each pressure pulse level or bin used to describe the assumed
level stresses by 1.25 (multiply by 1.5 to get the mean, then by 0.833 histogram of pressure transients experienced by the subject dent. The
to determine the mean minus one standard deviation). The design TP design fatigue curve is then used to determine the number of
and decision Thompson Pass (TP) S-N curves are both shown in cycles to initiate a crack at the surface for each pressure pulse level.
Figure 3. Once the stress ranges and the number o f stress cycles for each level is
known, the calculation remaining fatigue life involves a
CALCULATING DAMAGE straightforward application of Miners Rule.

With the TP design fatigue curve defined, fatigue damage was It should be noted that the maximum cyclic stresses associated with a
calculated using Miners Rule. This rule states that the cumulative constrained pipe dent subjected to a positive pressure transient are
damage is equal to the sum o f the number of cycles at each stress level bending stresses (positive on the outer surface) and they are located at
divided by die number of cycles to cause failure at the same stress the periphery of the dent The resulting fatigue damage just below the
level. The equation is: outer pipe surface is significantly less than at the surface. Thus, zero
remaining fatigue life as defined using the above method indicates
crack initiation at the surface, not a through wall crack or leak.

where D is damage (cumulative damage ratio), N is the number of DISCUSSION OF RESULTS


cycles at stress level i and Nf is the number o f cycles necessary to
cause failure at the same stress level. Crack initiation is assumed to For the purposes o f decision making on long term actions, a design
occur when D=I.O. level S-N curve is clearly appropriate. If the calculated cumulative
fatigue damage using this curve were to exceed 1.0 at some location in
The TP design fatigue curve was used to determine Nf for stress level the pipeline, this would mean that there is a very small probability of
i. The choice of a base metal curve is appropriate because the critical crack initiation at that location, with a still smaller probability of a
dent is approximately 6 feet away from the nearest girth weld, and the crack large enough to cause a leak. If one were to calculate a
highly stressed region of the dent is near the 6:00 position of the pipe cumulative fatigue damage o f 1.0 for any dent, this would indicate
cross section, approximately 90 degrees from the longitudinal seam that the pipe no longer satisfies typical design criteria at the dent, not
that it is on the verge of crack initiation. The authors believe the
weld.
proper decision in this case would be to schedule a repair of the pipe
The stresses at the critical dents are bi-axial in the longitudinal and within a reasonable time period and, until that time, to monitor the
hoop directions (the radial stress is not significant). The Von Mises pipe with leak detection sensors. As described above, cumulative
method o f combining stresses as described in (Shigley, 1983) is used fatigue damage values were calculated using two different forms of
to determine the equivalent mean and alternating stresses. First the the TP-Curves, as follows.
stress state at the minimum and maximum applied load are
determined. Then the alternating and mean stresses are determined for (1) A design curve, which indicates whether the pipe at the critical
each direction in the coordinate system being used, axial (x direction) dent location would satisfy the criteria for new pipeline design.
and circumferential (y direction) in this case: This is based essentially on mean values minus two standard
deviations. With this curve, there is only a small probability that
(2a) a cumulative damage of 1.0 will correspond to actual crack
2 2 initiation. (For a design value o f exactly the mean minus two
standard deviations the theoretical probability of crack initiation
_ G y \Jr &y\ _ &yi ~ &y\ (2b) is 2.5%, or 1 in 40). The probability that the pipe will leak is less
Gy* 2 2 than this, since the crack must first propagate through the pipe
wall. Crack growth assessments were not performed for the
dented region, however, the majority of the stress cycling is
where the subscripts m and a correspond to mean and alternating through the thickness bending o f the pipe wall rather than
stress for the particular direction and the subscripts 1 and 2 refer to membrane, therefore it is believed that crack propagation through
minimum and maximum load conditions. The Von Mises mean and the pipe wall would be a slow process.
alternating stresses are then determined from the mean and alternating (2) A decision curve, which is better suited for deciding on actions
component stresses: to be taken in the short term, in particular whether it would be
prudent to undertake an immediate repair of the pipe. With this
_ / 2 i 2 (3a) curve, there is roughly a 16% probability (about a 1 in 6 chance)
O m ^ JOim~&xmOyn O ym
that a cumulative damage of 1.0 will correspond to actual crack
_ / 2 4_ initiation, with, again, a smaller probability that the pipe will
O a *yO m ~ 0 to 0 ya O ya (3b)
leak.

Downloaded From: http://proceedings.asmedigitalcollection.asme.org/ on 09/12/2017 Terms of Use: http://www.asme.org/about-asme/terms-of-use


The results of the overall analysis are discussed in more detail in A technical peer review by the Joint Pipeline Office, a joint
accompanying papers (Baskurt, 1998 and Hart, 1998). These papers federal/state oversight agency, confirmed the analysis performed by
describe the work done in determining a history of the pressure Alyeska and supported Alyeskas conclusions.
transients, size o f the dents, soil stiffness and the finite element model
of the dents. A brief summary of the results are discussed here to Although the pipe was deemed to be stable and safe from the integrity
indicate the effects of using the design and decision level fatigue analysis, Alyeska elected to perform an inspection of the ovaled and
curves in practice. dented area at pipeline milepost 776 to verify the fatigue analysis and
to confirm, by testing, some of the key assumptions made in the
The fatigue analyses using the design level S-N curve were performed computer model regarding pipe support and geometry. The site was
and considered first to determine whether or not the pipe still satisfied excavated in July 1997. The excavation also allowed investigation of
the type of design criteria that might be used for new pipeline design. the cause of dent and allowed Alyeska to make repairs if necessary.
In other words, was there any remaining life based on a design The pipe was examined by wet fluorescent magnetic particle methods
analysis with its inherent safety factors and conservatism. The that could reveal microcracks on the outside surface of the pipe and
decision level S-N curve analyses were then performed to determine also by non-destructive ultrasonic methods that examine the interior of
what short term measures needed to be taken to ensure the structural the pipe wall for crack features. No evidence of fatigue damage or
integrity o f the pipe. cracking was found, confirming the main conclusions of our analyses.

Using the design level S-N curve, cumulative fatigue damage fractions REFERENCES
exceeding 1.0 were calculated only for the cases with the most
conservative assumptions o f pressure transient history and soil ASME, 1969, Criteria o f the ASME Boiler and Pressure Vessel
stiffness. For what were considered reasonable assumptions, the Code for Design by Analysis in Section III and Section VIII, Div. 2.
cumulative fatigue damage fraction never exceeded a value of 0.6 for ASME, 1995, Boiler and Pressure Vessel Code, Section VIII,
the design level S-N curve. Many variations of transient pressure Division 2 - Alternative Rules.
levels and soil stiffness were considered to determine their Baskurt, U. J., Bradshaw, P. M., and Hart, J. D., 1998, Slackline
significance on the fatigue life o f the pipe. Testing of the Trans Alaska Pipeline System at Thompson Pass,
International Pipeline Conference, Calgary, Alberta Canada.
For the decision level S-N curve, the calculated damage values were Cooper, W. E., 1992, The Initial Scope and Intent o f the Section
approximately one fifth of those for the design S-N curve. The III Fatigue Design Procedures, PVRC Workshop on Cyclic Life and
calculated damage factions were no more than 0.21 for any of the Environmental Effects in Nuclear Power Plants, Clearwater, Florida.
cases considered reasonable in terms of assumptions. D l.1-1996 Structural Welding Code - Steel, American Welding
Society, 550 NW LeJeune Road Miami, FL 33126.
A similar fatigue analysis was performed for the girth weld (Station Hart, J. D., Powell, G. H., Maple, J. A., Stevick, G. R., Norton,
40959+34) between the two relatively large dents at Thompson Pass. D., 1998, Fatigue Damage Calculations For a Dented and Ovalled
Fatigue design and decision level curves were assembled in a similar Section of the Trans-Alaska Pipeline System at Thompson Pass,
manner based on the AWS X-Curve (same as the AWS Cl curve) International Pipeline Conference, Calgary, Alberta Canada.
which is recommended for as-welded, full penetration butt splices and Ives, K. D., Kooistra, L. F., Tucker, J. T., 1966, Equibiaxial
groove welds. Based on the proximity to the slackline interface (i.e., Low-Cycle Fatigue Properties of Typical Pressure-Vessel Steels,
near the origin o f the pressure pulses) this weld is the most likely to be Journal of Basic Engineering.
subjected to a relatively high level of stress cycling from the pulses. It Juvinall, R. C., 1967, Engineering Considerations o f Stress,
is postulated that, with all other factors being equal, if this weld has Strain and Strength, McGraw-Hill.
little or no fatigue damage or crack growth, then welds further NCHRP, 1982, Evaluation of Fatigue Tests and Design Criteria
downstream would have sustained little or no fatigue damage or crack on Welded Details, National Cooperative Highway Research
growth. The calculated cumulative damage fraction for the girth weld Program Report 286, Project 12-15(5).
never exceeded 0.22, thus the dents govern the fatigue life of the pipe Rahka, K., 1992, Design Limits for Low Cycle Fatigue of
at Thompson Pass. Notched Components Baltica II, International Symposium on Life
and Performance of High Temperature Materials and Structures,
With an estimated fatigue damage fraction no greater than 0.6, the Tallinn, Estonia.
static strength of the pipe is essentially unimpaired. This means that if Rahka, K., 1993, Review of Strain State Effects on Low Cycle
the pressure pulses were stopped, as was accomplished in early 1997, Fatigue of Notched Components PVP, Vol 263, High Pressure -
the pipe at the critical dent locations is essentially undamaged for Codes Analysis and Applications, ASME.
future operating conditions. Shigley, J.E., and Mitchell, L.D., 1983, "Mechanical
Engineering Design ", 4th Edition, McGraw Hill Book Company.
This information was used to make recommendations to Alyeska,
namely that there did not appear to be any reason, based on technical
grounds, for excavating and repairing the pipe, since there is nothing
to repair.

Downloaded From: http://proceedings.asmedigitalcollection.asme.org/ on 09/12/2017 Terms of Use: http://www.asme.org/about-asme/terms-of-use


S T R A IN A M P L IT U T D E (% )

F IG U R E 1. Thom pson Pass fa tig u e cu rve shown w ith u n iax ia l and


b ia x ia l fa tig u e d a ta and cu rve fits from R e fe re n c e [3].


CL

111
o
z
<
cc
C
O
CO
111
IT
I
CO
cc
C
O

F IG U R E 2. F a tig u e d a ta , from R e fe re n c e [4], used as th e basis


fo r the A W S-A design fa tig u e cu rve [1].

Downloaded From: http://proceedings.asmedigitalcollection.asme.org/ on 09/12/2017 Terms of Use: http://www.asme.org/about-asme/terms-of-use


Figure 3 The "Design" AW S-A and "Decision" S-N Fatigue Curves

Downloaded From: http://proceedings.asmedigitalcollection.asme.org/ on 09/12/2017 Terms of Use: http://www.asme.org/about-asme/terms-of-use

You might also like