An Upheaval Buckling Limit State Function For Onshore Natural Gas Pipelines
An Upheaval Buckling Limit State Function For Onshore Natural Gas Pipelines
An Upheaval Buckling Limit State Function For Onshore Natural Gas Pipelines
net/publication/267598877
An Upheaval Buckling Limit State Function for Onshore Natural Gas Pipelines
CITATIONS READS
2 894
4 authors, including:
Joe Zhou
TransCanada PipeLines Limited
83 PUBLICATIONS 387 CITATIONS
SEE PROFILE
Some of the authors of this publication are also working on these related projects:
All content following this page was uploaded by Wenxing Zhou on 26 January 2015.
IPC2008-64687
ABSTRACT
The reliability-based design and assessment (RBDA) INTRODUCTION
methodology has gained increasing acceptance in the pipeline Trenching and backfilling of pipelines which operate at
industry, largely due to a multi-year PRCI program aimed at elevated temperatures induces large compressive axial forces.
establishing RBDA as a viable alternative for the design and The axial forces cause Euler buckling susceptibility. The
assessment of onshore natural gas pipelines. A key limit state susceptibility is increased at vertical overbend imperfections
of buried pipelines that operate at elevated temperatures is which may result from either the topography of the pipeline
upheaval buckling. The elevated temperatures generate large route or the trenching operation. If the weight of the pipeline
compressive axial forces that can cause Euler buckling combined with the resistance of the trench backfill are
susceptibility. The tendency to buckle is increased at vertical insufficient to hold the pipeline in place then the pipeline tends
imperfections (i.e. a series of cold formed bends) that primarily to displace vertically, resulting in upheaval buckling. Upheaval
occur due to topography. Upheaval buckling in itself is not an buckling has occurred in a number of land pipelines[1].
ultimate limit state but can lead to high strains, local buckling, Upheaval buckling may also interact with frost heave
high cycle fatigue, expensive remediation measures, and even within Artic land pipelines[1,2]]; this phenomenon is excluded
loss of pressure integrity. from the current work.
The critical forces at which upheaval buckling occurs for Upheaval buckling has also occurred in a number of
typical hill-crest type imperfections present in onshore elevated temperature subsea pipelines[3,4]. A major research
pipelines cannot be readily predicted using analytical methods. program was undertaken to develop design methodologies to
A parametric study is therefore undertaken using non-linear prevent upheaval buckling of subsea pipelines[5]. Significant
finite element analyses to generate a matrix of upheaval design and construction effort is expended to mitigate upheaval
buckling responses. The critical force for the onset of upheaval buckling of buried, elevated temperature subsea pipelines.
buckling is then developed using a series of empirical Although, upheaval buckling in itself is not an ultimate
relationships to capture the influences of all key parameters. limit state, it can lead to high strains, local buckling, high cycle
An upheaval buckling limit state function is subsequently fatigue, expensive remediation measures, and even loss of
developed by comparing the critical buckling force with pressure integrity. Hence upheaval buckling must be included
applied compressive force, which is a function of operating in any reliability-based design and assessment methodology.
pressure and temperature differential between the operating and
tie-in conditions. The limit state function can be readily
implemented in a reliability analysis framework to calculate the
pipeline failure probability due to upheaval buckling.
4
Figure 1 Idealisation of a Buried Pipeline Passing
Over a Hill Crest (exaggerated vertical scale)
3
Download
The pipe passing over the hill crest is idealised to be an arc Dynamic
download
of radius R subtending an angle θ. The side slopes are
2
therefore at an angle of θ/2 to the horizontal. The analysis is
performed for angle θ of 3°, 6° and 12°. It is assumed that the
1
bends would be fabricated using cold formed bends such that Static
the as- installed pipeline (except the bends themselves) would download
be stress-free. The FE model is therefore defined as stress-free 0
0 1 2 3 4 5
with residual stresses in the bends neglected. Vertical Displacement
Following a sensitivity analyses, the length of the straight
pipe between the bend at the imperfection crown and concave Figure 2 Download Response
curve is set at 50 m to avoid interaction between the two bends. The static download or overburden is simply the weight of
The model extends for 100 m beyond the lower bend. soil directly above the pipe:
The pipeline is modeled using ABAQUS elastic-plastic
Ws = γ H Dc (1)
pipe elements PIPE31. Symmetry boundary conditions are
applied to the model at the crest location. The element length where γ =
bulk unit weight of backfill;
is 0.1 m within 20 m of the hill crest and within the lower bend. H =
cover height TOP;
The element length is increased to 0.25 m between these two Dc =
outside diameter of pipe (including
zones and for 20 m beyond the bottom bend. The remaining 80 coatings).
m of pipeline is modeled with elements of 2 .0 m length. The static soil download is normalized to pipe diameter i.e.
To allow the short model to incorporate axial pipeline the static download is expressed by γH. Two values of γH are
feed-in to the buckle, a spring is attached to the end of the modeled – 12 and 5 kN/m2
model to represent the axial stiffness and resistance of the The dynamic component is modeled using the pipe-soil
pipeline. The spring modeling is presented in Annex A. An response models taken from the American Lifelines Alliance
end force is applied to the end of the model to balance the force “Guideline for the Design of Buried Steel Pipe”[6]. The uplift
generated by the application of internal pressure and force is given by:
temperature.
Qu = N cv c Dc + N qv γ H Dc (2)
Pipeline Material
Three linepipe steel grades, X52, X65 and X80, are used to where Ncv = vertical uplift factor for clay (0 for c = 0);
determine the effects of the yield strength on the upheaval Nqv = vertical uplift factor for sand (0 for φ = 0°);
buckling behaviour. The yield strength is defined by the API c = backfill soil shear strength.
specified minimum yield strength (SMYS). The stress-strain
curves are modeled using Ramberg-Osgood parameters for the Ncv is given by:
plastic response and standard elastic parameters for the elastic
H
response. The effect of hardening is examined by modifying N cv = 2* ≤ 10 (3)
the post-yield response of the X65 curve. Dc
Nqv is given by:
-3000
using the ABAQUS pipe-soil interaction element PSI34.
-2000
Analysis Matrix
The analysis considers a matrix of 252 cases. The pipe -1000
parameters are given in Table 1. For the sensitivity analyses,
only the parameter of interest is varied. 0
0.0 0.5 1.0 1.5 2.0
Parameter Base Case Sensitivities Crown uplift (m)
Grade X65 X52, X80 Figure 3 Typical Effective and Applied Forces vs.
Uplift Response
Hardening Good Low
As the model is initially loaded the force throughout the
Hoop Stress 60% of SMYS 80% of SMYS model is constant i.e. the effective force is equal to the applied
force. However, as the crown of the buckle begins to lift,
Bend Radius 38D 50D, 20D
Figure 3 indicates that the effective force does not increase at
Table 1 Summary of Pipe Parameter for Analysis the same rate as the remote applied force. This difference
Matrix between the two forces is due to axial feed-in to the developing
For each pipe parameter case a set of 18 analyses is buckle. The figure indicates that the effective force reaches a
performed given by: critical point at which the force is a maximum compressive
• Three bend angles θ of 3°, 6° and 12° value. The corresponding applied force at the critical point is
(corresponding to slope angles of 1.5°, 3° and 6°) also indicated in Figure 3. From the critical point the uplift
• Two static soil downloads (γH) of 12 and 5 kN/m3 increases fairly rapidly with increasing temperature or applied
• Three dynamic soil downloads of 6, 2 and 9 kN/m3 force, especially for the lower imperfection angles.
• Soil mobilisation displacement of 10 mm Large vertical movements do not necessarily lead to an
In addition a number of cases are also repeated with the ultimate limit state. Ultimate limit states are likely to include
increased soil mobilization displacement of 40 mm. local buckling, fatigue and weld fracture. A preliminary check
on local buckling is undertaken by comparing the compressive
(assumed to be 0.90);
-4000
αgw = girth weld factor (taken as 0.60 for D/t =
Force (kN)
-3000
angles, a small increase in applied force (or temperature) above
the previously defined critical point leads to a fairly rapid
-2000 increase in tensile total axial strain. Once more this confirms
that the previous definition of critical point is both robust and
-1000
appropriate.
0
0.0% -0.1% -0.2% -0.3% -0.4% -0.5% -0.6% -0.7% -0.8% -0.9% -1.0% INFLUENCE OF KEY PARAMETERS AND
Peak Mechanical Axial Strain (%)
DEVELOPMENT OF EMPIRICAL MODEL
3° - applied force 6° - applied force
12° - applied force critical applied force
Murphy-Langner buckling strain OS-F101 buckling strain - no pressure
Methodology
OS-F101 buckling strain - ½ pressure An initial attempt has been made to develop an analytical
Figure 4 Typical Applied Force vs. Compressive Axial solution for the upheaval buckling response of the stress-free
Strain Response pipe overlying a hill shaped imperfection, as such a solution
would greatly assist the development of the limit state function.
However, the problem proved to be intractable.
1
The mechanical strain is given by the total axial strain minus the thermal Hence, the approach taken is to develop best fit prediction
strain. The difference is small: at 100 °C the absolute increase in strain from equations to the FE results for critical effective and applied
total to mechanical strain is 0.117%.
0.6285 0.98
1
Peff _ base = − 99.89 ⋅ − 3.297 ⋅W 0.5771
sin (θ / 2)
0.96
tot
0.94
(9)
0.92
FE Data
where Peff_base =
critical effective force (kN);
0.90 Best Fit
θ =
total imperfection angle;
total download (pipe weight + soil
Wtot = 0.88
0 -1000 -2000 -3000 -4000 -5000
download) (kN/m).
Effective Force (kN) - hs = 60
The critical effective forces from the FE results for three
bend angles and six combinations of static and dynamic Figure 7 Hoop Stress Correction Factor – Best Fit
download are plotted in Figure 6 as individual marker points, Curve
along with predictions from the best fit equation which are
plotted as traces. Material Yield Strength Correction Factor
-5000
Increasing the material grade increases the critical effective
-4500
force, although the level of increase is less for
-4000
-3500
imperfection/download combinations which generally have a
Effective Force (kN)
-3000
lower critical effective force
-2500
A yield strength correction factor should be applied to the
-2000 s1d3, - 16.3 kN/m
effective force for pipeline yield strengths other than 448 MPa
-1500
s1d1, - 13.3 kN/m (grade X65). The calculated correction factors from the FE
s1d2, - 9.3 kN/m
-1000 s2d3 - 12.0 kN/m
analyses are plotted in Figure 8 for yield strengths
-500 0 5 10 15 20 25 30 35
s2d1 - 9.0 kN/m
40 45 corresponding to X52 and X80 material grades. The figure
s2d2 - 5.0 kN/m
0 also includes the best fit equation for the yield strength
1/ SIN(Bend Angle /2) correction factor which is a function of the base case effective
Figure 6 Comparison of FE and Predicted Critical force and of the yield strength, and takes the form:
Applied Forces
The mean and the standard deviation of the errors between
the eighteen FE results and the best fit equation predictions are
X52 - data
1.00 Fdia (13)
X80 - data 24
X52 - best fit
0.95 where Do is in inches.
X80 - best fit
ratio and therefore a correction factor should be applied to the Diameter (inch)
effective force for Do /t ratios other than Do /t = 60. Figure 10 Diameter Correction Factor– Best Fit Curve
The calculated correction factors from the FE analyses are
plotted in Figure 9 for Do /t of 40 and 90. The figure also Final Effective Force
includes the best fit equation for the D/t correction factor which The final value of the critical effective force is then given by:
is purely a function of the D/t ratio and is of the form:
Peff = Peff _ base ⋅ Fhs ⋅ F ys ⋅ FDt ⋅ Fdia (14)
= 1 − 0.3710 ⋅ ln o
D /t
FDt
60
(12)
Applied Force
where Do = pipe outside diameter; The ratio of critical applied to critical effective forces is a
t = pipe wall thickness. function of the effective force, diameter and D/t ratio. The
critical applied force is given by:
1.3 Calculated Factor - D/t = 40
Calculated Factor - D/t = 60
1.2
Calculated Factor - D/t = 90
− 661.4 − 53.71⋅ f
1.1 1 - 0.3710*ln(D/t/60)
Papp = 0.8593 ⋅ exp 0.55 1 .33
⋅ Peff
Do / t 24
⋅ ⋅
F(D/t)
1.0 Peff
60 Do
0.9
0.8 (15)
0.7 where f = pipe-soil axial resistance (kN/m).
30 40 50 60 70 80 90 100
A number of formulations exist for determining pipe-soil
D/t Ratio
axial resistance and any of these can be used in combination
Figure 9 D/t Correction Factor– Best Fit Curve with Equation (15). For example, the pipe-soil axial resistance
can be calculated using the following equation:
Error Summary
ST = − Eα T (T2 − T1 )π (D0 − t ) t (19)
The error bands between the FE results and the empirical
equation predictions are summarized in Figure 11. Error bands π (D0 − 2t )2
are given for both critical effective force at the buckle and S P = −(1 − 2v ) p (20)
applied force remote from the buckle, for the base forces and 4
for the resulting forces when each of the correction factors is where ST = temperature-induced force;
applied in isolation. The mean errors are indicated by the Sp = pressure-induced force;
square markers and the error bands show the ±2s.d. limits. The E = Young’s modulus;
maximum actual positive and negative errors are plotted as aT = thermal coefficient of expansion for steel;
diamonds. T2 = operating temperature;
T1 = tie-in temperature;
14
12
Effective Forces v = Poisson’s ratio;
10
Applied Forces
p = operating pressure.
8 The model error B associated with Equation (15) is
6
characterized by comparing the buckling forces predicted from
4
Error (%)
2
FEA (considered as “true” buckling force) and the equation for
0 the 252 parametric cases considered in the study. The model
-2 error is characterzied by the ratios between the predictions from
-4
FEA and Equation (15). Statistical analyses indicate that the
-6
-8
model follows a normal distribution with a mean of 1.0 and a
-10 standard deviation of 0.04. Note that the model error is
Base Hoop Stress Grade D/t Diameter
predicated on the fact that Equation (16) is used to calculate the
pipe-soil axial resistance with a fixed set of values.for µ (0.5),
Figure 11 Buckling Force Function Error Summary H (0.7 m), and φ (30 degrees). Using a different set of values
for µ, H, and φ in Equation (16) or an alternative method to
A further check on the robustness of the empirical model calculate the pipe-soil axial resistance will likely result in
would be to determine the error between FE and predicted different model error characteristics.
critical forces for a reasonably large number of cases where
more than one parameter is simultaneously varied from the CONCLUSIONS
base values. Unfortunately this exercise was not carried out Generating an analytical model of the upheaval buckling
during the course of the work due to time and resource response of an offshore pipeline which is installed stress free
constraints. over a hill shaped imperfection appears to be intractable and
hence FE analysis is required.
UPHEAVAL BUCKLING LIMIT STATE FUNCTION A limit state function to describe upheaval buckling has
Once the empirical equation for predicting the critical been developed by comparing the critical buckling force with
buckling force is developed, the upheaval buckling limit state applied compressive force. The critical buckling force is
function is established by comparing the critical force with the calculated using a set of empirical equations best fit to a matrix
applied compressive force due to pressure and temperature. of FE results. The applicability of the limit state model is
The limit state function is expressed as follows: limited to the range of variables considered in the FE matrix.
g = −( BPapp − S ) (17) More specifically, the limit state function is applicable to
pipeline diameters between 406.4 and 1066.8 mm (16 to 42
inches), diameter-over-wall thickness ratios between 40 and 80,
steel grades between 358 and 551 MPa (X52 to X80), hoop
ACKNOWLEDGMENTS
The authors would like to thank Pipeline Research Council
International (PRCI) for sponsoring the work and permitting
publication of this paper.
P0 − PM
Ls = (A5)
PM f
PB
Combining Equations A4 and A5 yields the spring
displacement in terms of the spring force:
X
6000
Axial Force in Spring [kN]
4000
2000
0
-0.4 -0.2 0 0.2 0.4
-2000
-4000
-6000
-8000
End Displacement [m]