ABS TRACT
This
paper
presents a method for probabilistic
stabil.ty design of submarine pipelines related to SIS
design criteria.
The procedure as such may also cover
[JIS
criteria
by
including
strain
response.
Theprobability
distribution
of
accumulated
long-term
response is evaluated by
integration
of
conditional
short-term
distributions
over
the environment
para-meter range.
Theexceedance
probability
for given
displacement
levels
is determined from the resulting
cumulative distribution function, which typically
has
a-
slowly
decreasing upper tail.
The method is
illu-strated by application to a specific pipeline
system.
Varying the submerged pipe weight will affect both the
shape and location of the density function of
accumu-lated
displacements significantly. Sensitivity of the
response distribution to the main
sources
of
uncer-tainty is investigated.
The relative contributions to
the variability
of
long
term
displacement
arising
respectively
from
short-
and
long-term statistical
uncertainties
are
of
particular
concern.
Modeluncertainties are represented by increasing the single
sea State response variance based on
experience
from
simulation studies.
1. rNTROIIJCTIQN
Present
pipeline
stability
design is normally
based on a simple stability
check
requiring
balance
between
hydrodynamic forces and pipe-soil resistance.
Such
a
method
has
several
shortcomings:
Static
stability
calculations
are
based
on
a traditional
Munson
hydrodynamic
force
formulation
with
force
coefficients
based
on
steady flow conditions (DnV,
197G). Pipe soil resitance is modelled
by asimple
friction
term.
Hence,
the
physical mechanisms of
these forces
are
not
properly
represented
bythe
mathematical
models.
Consequently,
significant
uncertainties are attached to such a design procedure.
Recently
methods
and
tools
for
more
sophisticated response analyses
have
been
developed
(Wolfram et al., Hoithe et al, Verley et al, Wagner et
al, Slaattelid et al, Tryggestad et al, Labrakos et al,
Presented at the Eghrh International Conference on
Offshore Mechanics and Arctic Engineering
The Hague - Mach 19-23, 1989
March 19-23
OMAE Europe '89
The Hague
TECHNISCHE UNIVERSITEIT
Laboratorium voor
ScheepshydrOmeChaflI
Archief
Mekelweg 2, 2628 CD Deift
015- 786873 - Fax: 015 - 781838
PROBABILISTIC STABILITY DESIGN OF A SUBMARINE
PIPELINE SYSTEM
T. Sotberg and B. J. L&ra
SINTEF Division of Structural Engineering
R. L. P. Verley
STATO I L
Trondheim, Norway
Myrhaug
et al,
all
1967).
Realistic simulation of
pipeline
response
can
be
performed
accordingly,
accounting
for
such
effects
as
directional
wave
spreading,
boundary
layer
reductions,
accurate
description
of
hydrodyriamic
forces,
and
soil
resistance dependence on
pipe
penetration
into
the
soil.
Since
each
simulation is rather costly, the
number of governing parameters has been reduced
by
a
factor of 3 by utilization of dimensional analysis.
Aresponse data base was established fro. numerous
indi-vidual
simulations
(Lambrakos et al, 1987), see also
the Appendix of this paper.
Byinterpolation,
the
movement
and
strain
response can be predicted for a
given
set
of
para.eters
defining
a
short-term
stationary sea state.
A
number of the parameters essential for
deter-mination of long-term
pipeline
response
are
random
quantities.
Information
about
their probabilistic
properties is also available, and should
be
properly
represented
byany
rational design method. In this
paper, focus is placed on a probabilistic
method
for
lateral
stability
analysis
of
pipeline
systems,
corresponding
to
SLS(Serviceability
Limit
State)
design
criteria.
The
objective of such a method is
twofold:
Byrpecification
of
target
exceedance
probabilities
for
given lateral displacement limits,
required pipe weights can be evaluated.
Additionaly,
the
method
maybe
used as a tool in the prct-ess of
developing
newcodes.
Both
calibration
against
existing
codes and derivation of partial coefficients
for a specified
overall target
reliability
can
be
facilitated.
Probabilistic methods for different limit states
than
those
considered
here
have
been
outlined by
Larsen et al (1986).
Lambrakos (1982)
used
a
wave-by-wave
approach
to
calculate
a
probabilistic
cumulative movement for pipelines. The
statistics
of
the
displacements
for
individual waves are combined
with the wave statistics.
Chao
(1988)
has
developed
a
procedure
for
calculating
the
long
term
cumulative displacement
response of
the
pipeline
by
assuming
a
Poisson
distribution of storms over the lifetime and applying
a Monte Carlo simulation technique to derive the
long
derived from the ]oint density of the same two vari-ables In our application st is approximated by a
lognormal distribution. The criterion adopted for identification of physica) unrealistic sea states with breaking waves is T I 3.2/H, Sea states outside ths5 region arc left out, and the joint density of R, and T,, has been renormalized.
The marginal density of significant wave height, H,, is given by a lognormal distribution for H, truncated at an upper limit, n. For larger H, a
Weibull distribution LS employed (Haver et al, 3986) Wave directionality d&ta are
given by a
two-dimensional table specifying the relative occurrence of waves as a function of Ha and mean wave propagation direction, 8, an in Table 1. Spreading of wave ener-gy around the mean direction is described by applica-tion of a cosine wave spreading tuncapplica-tion. The single sea state directional wave spectrum is expressed as follows
S (w,9) S (w)-g(9,8
n q w (2)
where S (w) is the one-dimensional wave energy spectrum (Jonswap type with y 3.3), and )8,8,) is
a cosine spreading function:
i(8,B (
= Ccos(e -8)
w w (3)
where C is a normalized factor.
The pipeline response in a single sea state is
described by a set of nc.n-dimensional parameters as outlined in the Appendix. The following 3 parameters describe the bottom flow velocity statistics and determine the wave and current loading on the
pipe-line:
Significant velocity normal to the pipeline: U,
Zero up-crossing period of flow velocity : T,,
Current velocity normal to the pipeline :
The last of these is assumed to be determin-istic, while the first are random variables. Their
joint density is determined by finding the (U,, T,,)
combinations corresponding to the (Rs, Tz) pairs, weighting over 8, in each case. To achieve this, a transformation of the directional spectrum into a normal velocity spectrum is required for fixed values of (H, , T,, ). Some of the steps in this process are (Sotberg et al, 1988):
The one-dimensional water particle velocity spectrum at the pipe level is calculated by multiplying S (w)
with a depth-attenuation factor which scales ith frequency.
The normal velocity spectrum for a given 8, is found
by applying a reduction factor, also accounting for wave spreading. A weighting over different 8,
is
subsequently performed.
The significant normal velocity is derived by integration of the corresponding spectral density. The zero up-crossing period is expressed through
moments of the same spectral density function
4. LONG TERM PIPELTNE RESPONSE
The long term distribution of accumulated pipe-line displacement for a reference period corresponding to the duration of a single sea state (3 hours), is
calculated as a weighted sus of the conditional single
sea state distributions. The weight function is the
long tCrm density function in the form given in the previous chapter.
.-jf(y(z)f(z)dz (4)
where
f )YJz) - conditional distribution of accumulated lateral displacement Y given bottom velocity
Statistics Z
f(z) - probabilistic density of bottom velocity data n.
The conditional statfstical parameters required
fr specification
o f(YIz) are evaluated by inter-polating between points in a data base. Each point is determined from simulations as described in Section 2.1. Given specific values for the set of dimension-less parameters defined in the Appendix, sufficient second-moment statistics for specification of the conditional density function are hence provided (Sotberg et al 1988).The integral in Eq. (4) is discretized and
cal-culated numerically as a weighted sum of conditional probability density functions for a number of
discre-tized values of the environmental data vector z. The long term distribution of lateral
displace-ment is hence evaluated from a Gaussian distribution
of short-term response in a single sea state and a long term distribution of sea states as outlined. lt is believed that the long-term distribution of wave
height is the critical factor with respect to the
upper tail in the long term response distribution, and thus also for the calculated probability of failure.
The long term displacement probability density will have a peak around zero, with an expectation also
located close to that peak. The distribution will be near symmetric around the mean value. The properties of this long term distribution are studied ir. more
detail in connection with the numerical examples in Chapter 6, where some typical plots are given.
A compact, second-moment statistical characteri-zation of the long term displacement is provided by
the expectation, E)y ), and the variance, o2 - These
may simply be calculated from the density function in
Eq. (4). This calculation requires a very high
reso-lution in the numerical integration, however, due to
irregularities of the integraiìd. An alternative
approach, which also gives some more insight, is to compute these quantities via their shcrt-term
counter-parts:
EEl',]
= f E[y)zjf(r)dr (5) z and 02 = J E[y(z]2f(z)dz + /02 tz 1m)' y1 - E[YL]S (6)where E[ylz], are the conditional expectation and
variance, respectively. The first arid last terms in Eq. (6) are the contributions from the long term wave environmental variation and the second term is contribution from the short term response variances.
For a relatively light pipeline with a large number of displacements throughout its lifetime, the first and last terms in Eq. (6) are expected to dominate. For a heavy pipeline only moving for the
are expected to be increasingly important, and the second term will be significant.
The second-moment Statistics of accumulated
displacements for long reference periods (e.g. 1, 5, 20 or 100 years), can readily be derived fromthose ir
Eqs (5) and (6) by assuming indpendence between the response in each sea-state. Denoting the total number ct sea states
itt
the pend L N the expressions becomeE[Y.j NT.E[YLJ
(7)
ou =ì
'T
TYL
If desired, the distrubutiøn function of long
term accumulated response can be evaloatd by convo-lutir,g that in Eq. (4) a number of (N1 -1) times,
)Starisberg, 1986).
5. RELIABILITY ANALYSIS
For sand soil, the SIS design method should allow limited pipeline movements. Criteria for accumulated displacements will in general be site
specific, depending e.g. on distance to neighbouring structures and width of the pipeline corridor. The
corresponding serviceability limit state function is
expressed as g(z)
T
-where Y is a deterministic displacement limit, and Y is the actual random lateral displacement. The vector
z contains basic random variables describing
uncertainties in long term wave environment, sea state description (realization of wave time series),
hydro-dynamic load and soil resistance modelling. For a ULS (Ultimate Limit State) design criterion, the limit State function would involve strains as well as displacements.
The probability of failure P1 is the
probabi-lity that the vector z has a value for which g(z) 0:
Pf - P(g(z( 0)
= - Y
0) (9)
The reliability is very often expressed through the reliability index 8R defined as
= (Pe)
(10)
where
(.)
designates the cumulative standard normaldistribution function.
Implicitly, both reliability measures are re-ferred to a specific time period, since the statis-tical properties of the variable Y will be completely different for a single sea state, Eq. (4), as compared
to a reference duration of one or more years. The relation between probability of exceedence P,, and the so-called return period, R5 is ex-pressed as
p
ex
p
where N = 2920 is the annual number of sea states of duration 3 hours; N5 is the number f sea states for
the given reference period. Putting N5 = 1, we get:
p = ex R U p (8) (12)
w
and with a return period of e.g. R9 100 years, the corresponding probability becomes P,,
0.34 10.
Based on Eqs. (li) nd 12), the transition from single sea state probabilities to those for a one-year reference becomes:N.
P (1 year) 1 - (I - p (3 hrsj) - (13)
ex ex
row with N5 U 2920. Implicit in Eq. (13) is a specific choice of R9. All of Eqs. (11), (12) and (13) are also based on the assumption of independence between excursions from sea-state to sea-state.
These probabilities can be employed in two ways. For a given pipeline, the exceedence probability for a
specific T can be found. Conversely, by specifying a
target P1 in advance, calculation of the pipe weight
required to satisfy the displacement restriction can be performed. Alternatively, given P1 and pipe
weight, a corresponding Y1 will result.
6. EXAMPLES OF APPLICATION
The methods for probabilistic stability analysis are illustrated through some examples for an actual pipeline system. Determination of the submerged pipe weight W, to meet the specified criteria is the main design task. Sensitivity of the probability of failure to variation of submerged pìpe weight is examined. Also, effects of model uncertainties and sea state statistical uncertainties are touched upon.
6.1 Long terni environment and pipeline dafa
The statistical parameters defining the condi-tional density of T9 on H, in Eq. (1) and the marginal density of H, , are identical to those specified by Haver et al (1986). wave directionality is accounted for by the distribution specified in Table 1. Spreading of wave energy is represented by a fixed exponent of n = 4 in Eq. (3). A constant effective
current velocity V = 0.57 m/s is included. The
boundary layer reduction is accounted for by the method of Slaatelid et al (1987). The current is assumed to be perpendicular to the pipeline with no distribution over direction. The water depth is 80 m, and some of the remaining key parameters are:
Saturated soil density p, = 2000
kg/&
Outer steel pipe diameter D, = 0.97 m
Steel thickness t = 23.5 mm
Natural gas density = 94.7 kg/r& The probabilistic design calculation was first performed by inclusion of statistical response varia-tion only due to differences between wave realira-tiens, in addition to long term wave environment statistics.
The effect of model uncertainties was
subse-quently taken into accout by increasing the standard
deviation in the single sea state response based on results from earlier sensitivity studies (Sotberg and
Remseth, 1986, a and b). This approach was adopted for both model uncertainties related to the hydrodynamic force and soil resistance modelling.
The basis for determination of the effect from
model uncertainties on the response variance is rather limited. Reliability analyses can, however, give valuable qualitative information on the sensitivity of the stability design process to these uncertainties.
6.2 Lesigr ezairpcs
The design crLterion ddopted is a
ateraJ
dis-p1&;.mrt Lmit Y r 2) m for free pipeline scuon
awiy from obstac1e or other structure
Cortespondin to the SF.S Condition. The correpondjn
target prob.3-bility of exceedence, for t reference period oÍ coe
year, is evaluated for R 100 years, ie. P 102 This desLgn check ensures safe operation
cf the
pipe-i pipe-inc.
The calculated long term probabiltty
density and the cuiulattve distribution functLon are given in
Figure 2 for a subnerged pipe weight equa'
to 4300 N/rn, whLch gives n Outer pipe diameter O
1.15 n.
The tai's of both functions have
been magnified due to
sa11ness o function va'ues.
It is observed that the distribution
is a1ost
sysmetric around a eean value very cloie to zero for this case, and the characteristic
upper tail for large displacement levels is illustrated. From the
distri-bution function the annual probability of exceeding the target displacement
level Y 20 s, is found to be 0.0082.
The corresponding reliability index is 2.40. Even a lateral displacement equal to ¿bout 18.5 in would satisfy the target probability
of exceedence
equal to 10-2 within a single year.
Increasing the submerged pipe weight to 5000 N/s
alters the tail in the distribution function as
seen from Figure 3. The tail is moved to a lower
displace-ment level and the shape has also been changed. Figure
4 shows the results for a submerged pipe weight reduced to 3000 N/in, which makes the upper tail shift to displacements of about 70
meters for the same probability levels.
The analyses reported above are all performed by including the statistical uncertainty due to differ-ences in wave realization. As a next step the uncer-tainty on the response due to uncertainties in the soil model is also taken into account. This is done by increasing the standard deviation of the response by 50 overall and somewhat more for the heavier pipeline cases or the lower displacement levels, where the sensitivity has been found to be largest, (Sotberg and Remseth, 1986, a and b).
Figure 5 gives a plot of the tail of the
distri-bution function similar to the one given in Figure 2, but now with soil model uncertainties included. The submerged pipe weight is still W! = 4300 N/rn. s
ex-pected, the tail in the distribution is moved to a
higher displacement level and the distribution is also
smoother, due to increased spreading of single sea state response. The probability of exceedence of the target displacement level has increased to P1 = 0.013. The expectation value of long term response is unaltered, but the long term variance has increased
due to larger short term response variances.
The effect of also including hydrodynamic force
model uncertainties is investigated by multiplying the response standard deviation by a factor of 2.5. This
is a relatively rough and ad hoc representation, but it should reflect a subjective, engineering type of
feel for the influence from model uncertainties. Plots
based on these calculations are given in Figure 6 for a pipe weight W, 4300 N/in. The probability of
ex-ceedence of the target displacement is as high as 0.13
which corresponds to a reliability index = 1.13. A lateral displacement of about 37 in corresponds to the target reliability index = 2.33 (P1
= 10-2).
The reliability index 8 is shown as a
func-tion of submerged pipe weight by the solid lines in Figure 7. Also, the lateral pipe displacement,
id.,' as a function of the saine for 8R1 = 2.33 us depicted
by dashed lines. The cuives marked a) represent
results obtained by only accounting for otatistical
variation due to different wave realizations. Curves b) and cl correspond to inclusion of soil model uncertainties and bot.h types of model uncertainties respectively.
From this figure, the design weight
correspond-ing to the target probability can be read
from the
horizontal line marked by . It is seen that taking
all the uncertainties into account would increase the design weight from ¿bout 4200 N/rn, with only realization uncertainties included to above 6000 N/rn. However, curve ci represents an
upper estimate of total load effect uncertainties corresponding to the factor 2.5.
The second-moment statistics of long-term
accu-mulated displacements as computed from Eqs. (5) and
(6) are displayed in Figure 8 as a function of sub-merged pipe weight. Both expected value and
standard deviation decrease for increasing weight, coirespond-ing to improved stability. By including model uncer-tainties for a fixed W, , the standard deviation is
succesively augmented as illustrated by curves b) and cl.
The relative importance of different contribu-tions to the long term standard deviation, ci1 , can be
assessed by computing the ratio of the second term in term in Eq. (6) to the value of the complete expres-sion. This ratio is depicted in Figure 9, which demonstrates that the variance of short response becomes increasingly important for increasing pipe weights. This also implies that choice of
distri-bution function for accumulated short term response
will be crucial for the tail of the long term ::esponse distribution of heavier pipes.
Sensitivity of the estimated to variation in
short term standard deviation, o , is studied by recomputing the safety index for increasing values of
the latter. Denoting the reference values by 8 and
o , respectively, the curves in
Figure 10 are obtained. is the calculated reliability index with the short term standard deviation equal
to zero, i.e.
whereas o1 is the reference value of the short term standard deviation of displacement response when
only wave realization uncertainty is included. The
dashed line gives the ratio between the short term
contribution (from sïngle sea state response) and the total long term variance as a function of the short
term variance. The results are valid for a fixed pipe
weight, W, = 4300 N/s.
The results indicate that by increasing the
standard deviation of response from zero, through the reference value and up to 2.5 times the reference the
reliabiLity index gradually decreases relative to BAs (=2.46) to half that value. The ratio between the
short term contribution and total long term variance similarly is a strongly increasing function o! the
short term variance.
The sensitivity factor with respect to model uncertainties, s, can generally be evaluated
from Figure 10 by the relation:
= 8R0
- e2)
The variation of a as a function of the short term variance is illustrated in Table 2. The
sensitivity to the short term response is increased from 16 per cent when only realization uncertainty is
included, to about 70 per
cent when
all model uncertainties (high estimate) are taken into account.This result indicates that for reasonable values of model uncertainties the long term wave environmental
n = /
1distrbntion is as important with respect t.')stiahility desi.;n .' iubmarin.' pipelines s th. us':ert.utnty in Ei 'Id'..O
iLort
trrnt epririse 7. DISCUSSIONA net hod or nrr..b nbili .t srah ¡1 t design
if
auibr.rrne pipe1ne y:;t,'mn his been out1ind. F.mph.siia
was pst ori thr' FLS design critericri .iri'i pp$ ori sand soil.. The pro':edire was illustrated by
.ippii:ation to
e specific pipeline
system.
'arying the sLibmorgedpipe
weight affects both the stirpe .ini locationof the probability
density function of
ac-cumulateddisplace-ments scgniii':antly. Displaement
corresponding to
one-year
targetexcedsnce
probability of p 10-2ranges from 12 m for a wei.;ht of 5300 N/n
to 7 ii for 3000 N/n.
Model uncertainties are
nc1uUed by inereasin; the single icea state response vari.snce.
Fr
.t finedpipe weight of 4300 N/rn and with p
1O, the
accumulated displacement var ies from li n for no model
uncertainties,
through
22 ni with soiltoiiStiflCC
models uncertainties,ti)
37 n by als's accountingrot
additronalhydrodyne,miç force
modeluncertainties.
Seond-rnoment statistics of long term
displace-ments depend heavily on pipe weight. HCdvy pipes will
move Only during extreme sea states, implying
rela-tively small expectation values and standard
devia-tions. Concurrently,
the single sea state response variance will contribute increasingly more to the
variance of the long term response.
In general, the probabilistic approach offers a versatile tool for pipeline stability analysis when a
number of the governing design parameters are random.
The relative influence from different sources of uncertainty can be quantified and treated in a uniform way. Improving the statistical models by
collection of data may reduce the uncertainties, resulting in more realistic and economic pipeline design. This
also applies to progress in quantification of model uncertainties, and possible refinement f the models as such.
In this process of advancing the general basis for prOhsbilittc pipeline design, a number of specific topics need further clarification. A non-exhaustive list may be as follows:
Identification of the distribution function for short-term pipeline displacement to a higher confidence level by incredsing number of simulations.
Inclusion of uncertainties in statistical para
meters defining the long-term wave environment.
Inclusion of statistical models for current
magnitude and direction, and joint statistics of current and waves.
More detailed study of model uncertainty effects on response. tmproved data would give closer bounds on the reliability index.
Furthermore, similar ptobabilistic analyses for the iLS limit state should be undertaken by also
con-sidering strain response. Failure functions
corre-sponding to total collapse and local buckling could then be defined. The methods discussed above should still be relevant also in that context.
ACKNQWLEEXEt4ENT
The authcrz wioh te cYpress their appreciation to Statoil and Esso Urge for permission to publish
this paper.
REFERENCES
Chao, J.C.: "Calculation of Long Term Cumulative Movements for Cubmarine Pipelines", Proc. Seventh
tnt. Syrrp. ori Offshore Mech. arid Artic Enqn.
Houston, Febr 7 - 12, 1988.
PnV Rules for Design, Construction and Inspection
of Submarine Pipelines and Pipeline Risers, 1976. Haver, 0. and Nyhus, K.A. "A Wave Climate
Description for Long Term Response Calculations",
Proc. Fifth Ent. Symp. on Offshore Mec-h, and Arctic Enqn, . Tokyo, April 13-18, 1986, Vol. IV pp. 27-35.
Holthe, K., Sotberg, T. and Chao, J.C.: "An
Efficient Computer Model for Predicting Submarine Pipeline Response to Waves and Current", Proc. of Nineteenth Offshore Technology Conference. Paper No. 5502, Houston, 1987.
Lambrakos,
KF.:
"Marine Pipeline DynamicResponse to Waves from Directional Wave Spectra",
Ocean Eniqineerinq, Vol 9, No 4, pp 385-405, 1982. Lambrakos,
KF.,
Remseth, S., Sotberg, T. andVerley, R.: "Generalized Response of Marine Pipelines",
Proc.
of Nineteenth OffshoreCon-ference, Paper No. 5507, Houston, 1987.
Larsen, E.U., Skjong, R. and Madsen, 11.0.:
"Assessment of Pipeline Reliability Under the Existence of Scour-Induced Free Spans", Proc. of Eighteenth Offshore Technology Conference, Paper No. 5343, Houston, 1986.
Myrhaucj, D.: "A Theoretical Model of Combined Wave and Current Boundary Layers Near a Rough Sea
Bottom", PrOC. 3rd. Int. Syap. on Offshore Mech. arid Arctic Eriqnq. , New Orleans, Feb. 12-16, 3984,
Vol. 3 pp. 559-568.
Slaattelid, 0.14., Myrhaug, D. and Lambrakos,
KF.: "North Sea Bottom Boundary Layer Study for Pipelines", Proc. of Nineteenth Offshore
Technology Conference, paper No. 5505, Houston, 1987.
Sotberg, T., Remseth, S.: "Pipe Response Sensitivity to the Wake Force Model", M.ARLNTEK Project Report SF 603455.60.01, The
SINTEF-Group, Trondheim, Norway, Dec 1986.
Sotberg, T., Remseth, S.: "Pipe Response
Sensitivity to the Pipe-Soil Interaction Models", MARINTEK Project Report SF 603455.60.02, The
SINTEF-Group, Trondheim, Norway, Dec 1986.
Sotberg, T., Lambrakos,
KF.,
Reaseth, S., Verley, R.L.P. , Wolfram, jr., W.R. : "StabilityDesign of Marine Pipelines", Proc. BOSS
Conference, Trondheim, June 1988.
Stansberg,
CT.:
"Estimation of Long TermPinelpe
Pro7ec Repor
APPENDIX
Calculation of lateral pipeline displacement
- a non-dimensional description
Genera I
The computer program PONDUS, (Holthe et al., 1987) has been employed for
calculation of lateral displacements for pipelines during a sea state.
The program includes itate-of-the-art models
for the fluid-pipe-soil interaction.
A generalized decriptìon of the physical problem
has several benefits. The pipeline
response is ex-pressed in terms of a few non-dimensional
parameters representing combinations of a larger number of physical quantities describing the pipeline and the ocan environment. The generalized respone data constitute a data base for transformation from basic
load parameters to response quantities in the probabi-listic design method outlined in this piper.
Nsn-dimessional response description
Scaling of the response is performed by employ-ing the equation of notion for the pipeline.
The equation of Sotion for pipeline Sections
with no boundary disturbances
from end constraints, or
possibly a free-free section is
m-YrF 4F -F
D I S (15)where s is the mass of the pipeline per unit length; y and t denote pipeline displacement and time,
respec-tively; F5 and F1 are the drag and inertia
hydro-dynamic forces per unit length, respectively and F5
is the soil resistance force per unit length. The external forces (F0, F1 Fs I for the moving
pipe are defined as follows
F = 1/lp D
C0()Iu - 1I(u
-D w lt lt [C-'w T'
M lt (CMl) J F t F . f(W -F ) +R' and the lift force L is
5 s L
FL = 1/2 0D C0))(u -Ft
Here, Fe is a remaining nonlinear soil resistance force depending on soil density and pipe penetration, (Wagner et aL, 1987) . The flow velocity u, which
consists of wave and current components, represents the effective flow velocity calculated by the Wake Force Model, (Verley et aL. 1987). The coefficients C5 (drag) and C, (lift) are functions of the distance,
s, the water part:cles travel from the instant of flow
reversal (making these coefficients tine dependent). Tryggestad, S folistad, T. Nilien, J. and
Gelanger, F.: Measuring Technique
lsd Field Data for Pipe.)ine Ctahility Studies', Proc.
of Nine teenth Orfshore Technology Conferencp
Paper No. 550G, Houston, 1987.
Verley, R. Lambrakos,
KF,
and Reed, K.'Prediction of Hydrodynamic Forces
on Seabed Pipelines' Proc. of Nineteenth
Offshore Technology Conference, Paper No. 5503, Houston, 1987.
Wagner, D., Murff, D., Brennodden, H. and Sveggen, O. 'Pipe-Soil Interactive Model', Proc.
if Nineteenth Offshore
Technoloìy ('inference, Paper No. 5504, Houston, 1987.
Wolfram Jr., W.R., Getz, JR. and Verley, R.L.P. PIPESTAB Project: Improved Design Baits for Sub-marine Pipeline Stability', Proc. of Ninetnth Offshore Tochnoloqy Conference. Paper No. 5501, HoustOn, 1987.
The coefficients C' (inertia) and f (soil friction) are assumed constant
The various quantities in the equation of motion are scaled .,,a follows:
j' y/D, t' = tíT, U' = u/U, s' = s/D (20)
where U, is the significant particle velocity normal to the pipe. substituting the scaled
quantities and collecting terms lead to the following dimensionless
equation of motion:
[2LKN
+ C j
----2C0(s')
(Ku' -
-rI{Ku'
- --i) +lt n M Ti 2f CL (s [Ku' - Ï_j2 + g.0 - f r lt M lt T
This dimensionless equation illustrates that the
relative pipeline displacement (y') depends on the
quantitites K, L, N (which are defined below), and u', s'. Analysis of the wake model equations and response
simulations has shown that u' and s' scale with the
parameters 1< and M (a parameter representing the steady current in the flow) - Although other forms of
the dimensionless equation are possible, equation (20) is quite convenient since the influence from the para-meter N is greatly reduced by the term 2LKN/m being small compared to C for most cases of interest.
Thus, for a given sea state and sar,d density,
the four dimensionless groups governing the pipeline displacement are: [IT K = (22)
L-
(23) D W s 1/2pDIJ2 V M c U N = (25)where T,, and V are zero up-crossing wave period and
steady current for the sea state, respectively; D and
W, are the pipeline outer diameter and submerged
weight per unit length; g and p, are the acceleration
of gravity and mass density of water, respectively.
The wave velocity U,, and the current V refer to the (flow) components normal to the pipeline.
The scaling parameters K, L, M, N can be interpreted as follows: K is a Keuleqan-Carpenter
number, L is a ratio between pipe weight and
hydrodynamic forces, M is a current to wave velocity
ratio, and N is a representative acceleration for the ea state.
(21)
.-. i-:.:i±:ñ--±±-'-i-'-'---»----
. .h--.--
-Er.-f,
Figure 4
Probability density and distribution function,
W= 3000 N/rn, realization uncertainty included
-2
I D 1 2510'
10 .08 .06 .04- .02-.00Figure
1Distribution of lateral displacement
xlO 4 .35 30-b. .20 .10 00 P( Y) 3 P(Y)J .08 4 .00
sia
Figure
alO .04 .03 .02 .01 .00Figu'-e
-T, 0 5 20 30 40 DISPLACEMENT Y (a)Probability density and
IV
4300 N/ni, soil model
10 510 .35 30. .20-10 00distribution
-4 20 30 40 DISPLACEMENT Y (a)function,
uncertainty included
Fure
1.0.. F(s)0.8.
06
0.4 0.2 0.0 -4Figure
-2 0 2 4 DISPLACEMENT Y (m) 2(a)Probability density
4300 N/n', realization
of lateral
5 10 15 20 25 30 35 DISPLACEMENT Y (mIpipe displacement
uncertainty included
-4 -1 .30 l-FIT .2') 10 t - .00 . 20 6 20 40 60 80DISPLACEMENT Y (m) DISPLACEMENT T (a)
Probability density and distribution function,
5V
4300 N/rn, ail model uncertainties included
-2 0 2 4 10 15 20 25 30 35
DISPLACEMENT T Im) DISPLACEMENT Y Im) 2(b)
Cumulative distribution function.
4300 N/rn
.20 - lo .00
Figure
35 30-.10 .00N
3 10 15 20 25 30 5 10 15 20 25DISPLACEMENT
(a)
DISPLACEMENT Y (m)Probability density and distribution function,
5000 N/rn. realization uncertainty included
30
s'a_4
35 30- 10-.00 .04 .03 02-.01 .00 60 80 100 120 DISPLACEMENT Y (a) 40 60 80 100 120 40 DISPLACEMENT Y (a)(al Realization uncertainty
b) Realization -
oiImode! uncertainty
cl Realization * ali model uncertainty
Figure 8
Standard deviation of the long term displacement
and long term expected displacement versus
Figure 9
Ratio between short term contribution and total
long term variance versus
0.5
0.00.0
0.5 1.0 1.5 2.0 2.5 °y '0YO ll R0ratio
Table
Wave directionality data
/
i
TabLe
2Sensïtivity factor as a function of the
short
term standard deviation of response
li 8
(deg)
(a)
0 4590
135 180225
270
315 1 4822433
1520 686 783 407 591 3013 2 37233780
2640
3731 58385410
3152
2242
3 3356822
750
26917332
5622
313222652
42377
159 90 12264730
4266
2918
2040
5 1214 27 9 4232365
2133
1592 1428 6583
7 3 171 1301970
756816
7 282 20
50 710388
292
377 8 122 0 0 16 355 136 100 1134 . 61 0 0 5 142 39 32 87 10 31 0 0 1 64 13 11 44 11 13 00
0 19 5 3 14 12 4 0 0 0 9 1 1 5 13 2 0 0 0 2 0 0 2 14 0 0 0 0 0 0 0 1ay/ayo
13fBo
10.975
0.16
1.5
0.91
0.31
2.5
0.46
0.74
Figure 7
Reliability index 13R versus pipe weight, and design
ligure 10
Sensitivity of 6R to short term variance
displacement for 13Rt
2.33 versus pipe weight
3R