Proceedings of the ASME 2011 30th International Conference on Ocean, Offshore and Arctic Engineering

OMAE2011
June 19-24, 2011, Rotterdam, The Netherlands

OMAE2011-49

METOCEAN DESIGN CRITERIA FOR PIPELINE ON-BOTTOM STABILITY

Richard Gibson
BP Exploration
Chertsey Road Sunbury on Thames
TW16 7LN
UK
richard.gibson@uk.bp.com

ABSTRACT criteria must be established that incorporate the joint occurrence

This paper is concerned with a response based method for of the various environmental parameters; in particular, the design
determining metocean design criteria for offshore pipelines. The wave height, wave period, current speed and their associated
method determines a set of metocean parameters that are consis- directions.
tent with the extreme response of the pipeline, and hence, incor-
porates the dependence between them implicitly. However, there A response based method based upon [1] is proposed and is
are a number of challenges in its application. Firstly, the load- outlined in the following section. The paper then continues with
ing on a pipeline is dependent on the previous wave cycle, and details of the development of the method, including a model
hence, the drag and inertia coefficients vary within a sea-state. for the short term distribution of the response in terms of the
Secondly, along many pipeline routes the waves are depth lim- parameters of a sea-state. The method is then applied to a site
ited and the short-term distribution of wave induced velocity and in the Mediterranean Sea and the results are compared with
pipeline response can be difficult to define. These challenges are those obtained from the first order reliability method (FORM).
overcome through a number of approaches that include a para- The sensitivity of the results to assumptions about the diameter
metric representation of the distribution of the response and the of the pipeline and the coefficient of friction is investigated.
application of multivariate extreme value analysis. Furthermore, Finally, the paper concludes with a discussion of possible future
the sensitivity of the method to assumptions about the pipeline developments and a summary of the work.
design is examined, and the problems with using the combined
wave and current induced velocity as a proxy for the response are
discussed. The method is applied to a site in the Mediterranean
Sea and the results are compared to those from the application BACKGROUND
of the first order reliability method. At its simplest, the absolute stability of a pipeline is a func-
Keywords: Pipeline, Stability, Metocean, Design Criteria, tion of the horizontal force, the lift force and the sea-bed friction
Waves, Currents. (horizontal resistance)

INTRODUCTION (W − Fz )µ
γ= , (1)
The design of offshore pipelines involves an assessment Fy
of their on-bottom stability subject to loading from waves and
currents. In order to make this assessment, metocean design where γ is the safety factor, W the submerged weight of the

1 Copyright © 2011 by ASME

Downloaded From: http://proceedings.asmedigitalcollection.asme.org/ on 08/30/2017 Terms of Use: http://www.asme.org/about-asme/terms-of-use

 pipeline, µ the coefficient of friction, Fy the horizontal force and 20 7

Fz the lift force. This study is solely interested in pipelines that in 18

the extremes are dominated by drag loads, and hence, the forces 6

can be calculated using the drag term in Morison’s equation as 16

Fi = 0.5Ci ρDU 2 , where Ci is the drag or lift coefficient, ρ the 14

5

density of water, D the pipeline diameter and U the velocity.

However, many experimental and theoretical studies have shown 12

Wave Height (m)

4
that the drag and lift coefficients are a function of time and 10
depend upon the wake that formed during the previous wave
3
cycle [2]. In this study this will be modeled following the 8

approach described in [3], in which the drag and lift coefficients 6

2
are a function of the Keulegan-Carpenter number, KC, and the
ratio of steady to oscillatory velocity, M = Uc /Uw . 4

1
2
The safety factor expressed in equation 1 is a function of the
velocity squared (assuming inertial loading is unimportant), and 0
0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8 2
0

hence, it is tempting to consider the velocity as an appropriate Current Velocity (m/s)

proxy for the response itself. However, the variation in the drag Figure 1. CONTOURS OF VELOCITY (M/S) AS A FUNCTION OF
and lift coefficients with KC and M means that metocean pa- WAVE HEIGHT AND CURRENT SPEED ASSUMING CONSTANT WAVE
rameters consistent with a constant velocity do not coincide with STEEPNESS.
those for constant pipeline response. This is shown in figures 1
and 2 in which contours of total near-bed velocity U = Uw +Uc
and of the pipeline weight required for a safety factor of one 20 90

18 80

Fy 16
W = + Fz , (2) 70
µ 14
60

12
Wave Height (m)

are shown for a range of wave heights H and current speeds 50

Uc . Therefore, in this study the response that is considered 10

most appropriate is that provided by equation 2; however, 8

40

comparisons will be made with using the velocity as a proxy for 30

this response. Clearly, equation 2 is not a precise model for the 6

response either, as it does not take into account the embedment 4

20

of a pipeline due to oscillating wave loads, and hence, under-

estimates the wave induced increase in soil resistance (that is, 2 10

overestimates the effect of waves relative to currents). 0 0

0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8 2
Current Velocity (m/s)

The desired return period response, X, (either weight W

or total velocity U), has been calculated from a metocean time Figure 2. CONTOURS OF PIPELINE RESPONSE (kN) AS A FUNC-
series of significant wave height, spectral peak wave period, TION OF WAVE HEIGHT AND CURRENT SPEED ASSUMING CON-
mean wave direction, current speed and current direction. This STANT WAVE STEEPNESS.
has been achieved by applying a peak-over-threshold, storm
based analysis, based upon [1], the steps of which are outlined
below.
3. The long term distribution of the response, p(Xmp ), where
p denotes the probability density function, has been deter-
1. Independent ‘storm’ events have been identified by identify- mined by fitting an extreme value distribution to the most
ing the peaks of X in a time series of metocean parameters. probable maximum value of the response associated with
2. The most probable maximum response Xmp associated with each peak event.
each event has been calculated. 4. The distribution of the response,

2 Copyright © 2011 by ASME

Downloaded From: http://proceedings.asmedigitalcollection.asme.org/ on 08/30/2017 Terms of Use: http://www.asme.org/about-asme/terms-of-use

 The short term distribution of pipeline response has been
determined through the random time domain simulation of sea-
Z states specified by their significant wave height, peak wave pe-
P(X) = P(X|Xmp )p(Xmp )dXmp , (3) riod and current speed. Each sea-state has been simulated for 18
hours in order to ensure stable statistics at the three hour level.
Individual waves have been separated by their zero-upcrossings
where P is the probability of exceedance, incorporating in order to determine their height H and period T . The breaking
both the long-term, p(Xmp ), and the short-term variability, criteria [5]
P(X|Xmp ), has been calculated and the return period value
of interest determined.

In equation 3 the asymptotic result (distribution of maxima Hb

= 0.142 tanh 2πd/λ , (5)
in a narrow banded process) λ

where λ is the wave length and d the water depth, has

been applied in order to reduce the wave height associated with
P(X|Xmp ) ≈ exp[− exp(− ln N(([X − γ]/[Xmp − γ])β − 1))] ,(4) breaking waves H > Hb to H = Hb . The near bed velocity U
and pipeline response W have been calculated for each wave
can be used if the short-term variability is described by a and both the distribution and the most probable maximum value
Weibull distribution P(X) = exp [−((X − γ)/α)β ], where α is the determined. The near bed velocity has been calculated using
scale, β is the shape, γ the location, and where N is the number both linear and Stream function [6] wave theories, as the latter
of extremes in an event. is believed to be an appropriate model for near bed velocity in
shallow and intermediate water depths. To what extent this is
correct and the water depth range over which this is the case is
not clear, and is one of the subjects of the LOWISH JIP.
METHOD
This model includes wave-current interaction throughout
In this section details of the response based approach for
most of the water column, but critically not in the near-bed
determining metocean design conditions are presented along
region where the wave boundary layer can be important. The
with some of the difficulties experienced in its application.
effect of wave induced oscillatory motions (nearly horizontal
near to the bed) is to increase the turbulence within the boundary
In order to apply the method it is necessary to determine the
layer and enhance the apparent roughness of the sea bed. This
short term distribution of the response within a sea-state. This is
can lead to a reduction in the current velocity and if the waves
required in order to identify peak events in step one, to calculate
and currents are not aligned then it can also alter the angle [7].
the most probable maximum value associated with an event in
In many cases, for larger diameter pipes, it is likely that the
step two, and to define the shape and location parameters in step
pipe centreline will be close to the edge of the wave boundary
four.
layer. However, most pipelines will have a degree of embedment
or are trenched, and therefore, boundary layer effects should
It would be appealing to calculate the short term distribution
be considered. This could be modeled using the approach
of response using a NewWave formulation, similar to that
described in [8]. However, to do so accurately the depth of
applied in [4], in which the shape of the wave associated with
embedment/trenching must be known and in this study it has
most probable maximum response is calculated for a desired
not been considered. In terms of determining metocean design
probability of exceedance. However, the occurrence of wave
criteria this (and most forms of) inaccuracy in the response
breaking, and the belief that in shallow or intermediate water
function leads to conservatism in the final results, as more than
depths a regular wave theory provides a better representation of
one point on the return period response surface is reported.
the near-bed velocities, means that two less elegant approaches
have been adopted: the brute force approach of using time
domain simulations, has been used for the calculation of the
most probable maximum response associated with a ’storm’
event; a parameterisation of time domain simulations, has been
used for identifying peak events and the integration of the short
and long term distributions.

3 Copyright © 2011 by ASME

Downloaded From: http://proceedings.asmedigitalcollection.asme.org/ on 08/30/2017 Terms of Use: http://www.asme.org/about-asme/terms-of-use

 Parameterisation
The short term distribution of both linear velocity and
pipeline response has been expressed in terms of the significant Htr ωtr
Utr = (8)
wave height, Hs, spectral peak wave period T p, current speed 2 sinh ktr d
√
and water depth, by fitting composite Weibull distributions α1 = Us / 2
to simulations of approximately 200 sea-states in the range: β1 = 2.0
5 ≤ Hs ≤ 20m, 6 ≤ T p ≤ 20s, 0 ≤ Uc ≤ 2.0m/s and 5 ≤ d ≤ 35m.
γ1 = Uc
Utr −Uc
α2 =
[− log(P1 (Utr ))]1/β2
Velocity. The short-term distribution of linear near-bed
β2 = 2.0 + 1.6S1 + 3.6Ur
velocity has been determined parametrically in a manner similar
to that of [9] and comprises two distributions: a Weibull γ2 = Uc ,
P1 (X|α1 , β1 , γ1 ) in the region below a transitional velocity Utr
and by a different Weibull P2 (X|α2 , β2 , γ2 ) above the transi- where Us is the significant near bed velocity (below the
tion. There are therefore seven parameters in total (three for transitional value the distribution is assumed to be Rayleigh).
each distribution and the transitional velocity); however, as
continuity of probability density is maintained by ensuring that
P1 (Utr ) = P2 (Utr ) and as the threshold is determined by the
7
current Uc = γ1 = γ2 only four of these are free parameters. Fur- No breaking
Breaking
thermore, the assumption of linear theory below the transitional Fit

velocity defines α1 and β1 , and therefore, this leaves only two 6

key values: the velocity associated with the breaking limit (the
transitional velocity); and the shape of the velocity distribution 5

above the limit.

Velocity (m/s)

The transitional velocity can be defined in terms of a transi-

tional wave height, using a local wave frequency ωtr = 2π/Ttr , 3

where Ttr = 1.1 × T1 = 1.1 × 2πm0 /m1 with m0 and m1 the

zeroth and first spectral moments respectively. 2

2
ωtr = gktr tanh ktr d (6) 0
10
−4 −3
10
−2
10
−1
10
0
10
0.892 Probability of Exceedance
Htr = tanh ktr d .
ktr Figure 3. DISTRIBUTION OF NEAR-BED VELOCITY: Hs = 10m,
Tp = 12s, U = 0m/s, d = 20m.
The shape of the distribution above the transitional veloc-
ity is defined in terms of the significant wave steepness and the
Ursell number (in a similar vein to the parameters of second-
order crest elevation given in [10])
Weight. Once a pipeline diameter and coefficient of
friction are assumed the pipeline weight required for a safety
2πHs factor of one is a function of wave height, wave period and
S1 = (7) current speed W = f (H, T,Uc ).
gT12
Hs
Ur = 2 3 . The parameterisation of the short term distribution of the
k1 d response follows from that of the velocity and is defined by
a Weibull P3 (X|α3 , β3 , γ3 ) in the region below a transitional
The seven parameters are defined as follows and an example response Wtr and by a different Weibull P4 (X|α4 , β4 , γ4 ) above
of the fit is shown in figure 3. the transition.

4 Copyright © 2011 by ASME

Downloaded From: http://proceedings.asmedigitalcollection.asme.org/ on 08/30/2017 Terms of Use: http://www.asme.org/about-asme/terms-of-use

 The lower threshold γ3 = γ4 = W0 is the response associated 120
No breaking
with the current, and hence, W0 = f (0, 0,Uc ). The transitional Breaking
Fit
response Wtr is related to the transitional velocity, and hence, 100
Wtr = f (Htr , Ttr ,Uc ).

The distribution of response below the transition can be 80

described by a constant shape parameter β = 1.3. Whilst a

Response
constant shape parameter is clearly appropriate for the linear 60

velocity (ie β = 2) the same is not obviously the case for

the response; however, the simulations suggest that the shape 40
parameter is typically in the range 1.2 ≤ β ≤ 1.4 for realistic
sea-states.
20

The scale parameter associated with the distribution below

the transition is constrained by the requirement that the prob- 0
−4 −3 −2 −1 0

ability of exceedance associated with the transitional velocity 10 10 10

Probability of Exceedance
10 10

should be equal to the probability of exceedance associated with

the transitional response. Figure 4. DISTRIBUTION OF PIPELINE RESPONSE: Hs = 10m,
Tp = 12s, U = 1m/s, d = 20m.
Finally, the shape of the distribution of the response above
the transition has been defined in terms of the steepness S1 and
the Ursel number Ur , with the scale constrained by the continuity by [11] that the extrapolation of a process in which different
of probability of exceedance at the transition P3 (Wtr ) = P4 (Wtr ). physical processes are important in different parts of the distri-
bution can lead to a bias in the extreme return period estimates.
The seven parameters are defined as follows and an example This has been investigated by simulating 200 years of data from
of the fit to the data is shown in figure 4. an environment with the following parameters: significant wave
height has been randomly sampled from a generalised Pareto
distribution with scale σ = 2.0, shape ε = −0.1 and threshold
γ = 5.0, which with 10 events a year gives a one hundred year
Wtr − γ3 Hs = 15m; the peak period has beenp randomly sampled from a
α3 = (9) normal distribution with mean µ = (Hs/0.06) and standard
[(Utr − γ1 )/α1 ]β1 /β3
β3 = 1.3 deviation σ = 0.1µ; the current speed has been randomly
sampled from a Weibull distribution with scale α = 0.25, shape
γ3 = W0 β = 1.0 and threshold γ = 0.0 which gives a one hundred year
Wtr −W0 return period value of 1.7 m/s.
α4 =
[− log(P3 (Wtr ))]1/β4
β4 = 1.3 + 1.7S1 + 4.0Ur Figures 5 and 6 compare bootstrap estimates of the one
γ4 = W0 . hundred year return period value from 10 years of simulated
data with the estimate from the full 200 years of data for water
depths of 30m and 20m respectively. It is clear that in the deeper
water the estimate from 200 years of data is close to the mean
Extrapolation of the 10 year samples (its the 65th percentile). In contrast,
Equations 8 and 9 have been used to identify peak events in shallower water there is considerable bias and the 200 year
in the metocean time series. Whilst it would be possible to use estimate is approximately the 95th percentile. However, owing
the same distributions to calculate the most probable maximum to wave breaking the upper limit is constrained and numerically
value associated with each event, it is more accurate and the results are very similar (5.75 m/s compared to 5.45 m/s) -
computationally not too onerous, to simulate each event and fortunately, and perhaps not surprisingly, the problem is well
calculate the most probable maximum value empirically. behaved. Finally, figures 7 and 8 compare estimates of the
most probable maximum velocity and response, calculated
Once this has been completed an extreme value distribution using the short term distributions given in equations 8 and 9,
has been fit to the most probable maximum values and the long with those calculated empirically. The results indicate that the
term distribution estimated. However, it has been pointed out parameterisation works well for velocity, but is positively biased

5 Copyright © 2011 by ASME

Downloaded From: http://proceedings.asmedigitalcollection.asme.org/ on 08/30/2017 Terms of Use: http://www.asme.org/about-asme/terms-of-use

 1 7

0.9
6
0.8

0.7 5

0.6

U Parametric Fit
4
Probability

0.5

0.4 3

0.3
2

0.2

1
0.1

0
40 60 80 100 120 140 160 180 200 220 0
Response 0 1 2 3 4 5 6 7
U Empirical

Figure 5. DISTRIBUTION OF 100 YEAR RETURN PERIOD ESTI-

Figure 7. COMPARISON BETWEEN THE MOST PROBABLE MAXI-
MATES OF RESPONSE FROM 10 YEARS OF SIMULATED DATA COM-
MUM VELOCITY U (m/s) CALCULATED EMPIRICALLY WITH THAT
PARED TO THE ESTIMATE FROM 200 YEARS OF SIMULATED DATA
CALCULATED USING THE PARAMETRIC DISTRIBUTION DESCRIBED
(DASH-DOT BLACK LINE) FOR A WATER DEPTH OF 30m.
BY EQUATION 8.

70
1

0.9 60

0.8
50

0.7
Response Parametric Fit

40
0.6
Probability

0.5
30

0.4

20
0.3

0.2 10

0.1
0
0 10 20 30 40 50 60 70
0 Response Empirical
64 66 68 70 72 74 76 78 80 82
Response

Figure 8. COMPARISON BETWEEN THE MOST PROBABLE MAXI-

Figure 6. DISTRIBUTION OF 100 YEAR RETURN PERIOD ESTI- MUM VELOCITY U (m/s) CALCULATED EMPIRICALLY WITH THAT
MATES OF RESPONSE FROM 10 YEARS OF SIMULATED DATA COM- CALCULATED USING THE PARAMETRIC DISTRIBUTION DESCRIBED
PARED TO THE ESTIMATE FROM 200 YEARS OF SIMULATED DATA BY EQUATION 9.
(DASH-DOT BLACK LINE) FOR A WATER DEPTH OF 20m.

Short Term Variability

Once the most probable maximum value for a particular
return period has been determined the calculation of the actual
for the response. However, the latter is considered suitable for return period value requires the integration of the short term
identifying peak events and as a conservative estimate of the distribution over the long term distribution (equation 3). The
shape of the distribution. simplest assumption is that the short term distribution has
zero location, γ = 0.0, and a shape parameter associated with

6 Copyright © 2011 by ASME

Downloaded From: http://proceedings.asmedigitalcollection.asme.org/ on 08/30/2017 Terms of Use: http://www.asme.org/about-asme/terms-of-use

 1 Exponential and generalised Pareto distributions to the most
0.9
probable maximum response associated with storm events. The
results of the likelihood ratio test are shown in figure 11 and
0.8
indicate that at many thresholds there is little to choose between
0.7 the two models. The Exponential fit at a threshold of 8kN is
shown in figure 12 and is selected as the most appropriate value.
0.6
This gives a 100 year most probable maximum response of 27.4
0.5 kN and when the short term variability is included the estimate
P

is 30kN.
0.4

0.3

Hs:3.5m Tp:6s U :0 m/s

0.2 c
Hs:1.6m Tp:6s Uc:0.75 m/s
Estimates of the 100 year return period velocity calculated
0.1 Hs:0.9m Tp:6s Uc:1.5 m/s using both linear and Stream function theories, and also the
Hs:2.3m Tp:6s U :0 m/s
c
response calculated using linear theory, have been made in the
0
2.4 2.5 2.6 2.7 2.8 2.9 same way and are shown in table 1. The results indicate that in
Maximum velocity (m/s)
this water depth the choice of wave model has very little affect
Figure 9. THE DISTRIBUTION OF MAXIMUM VELOCITY IN A THREE
on either the velocity or the response. The contours associated
HOUR SEA-STATE (WATER DEPTH 5M). THE THICK SOLID LINE IS
with the 100 year return period velocity and response are shown
THE SOLUTION THAT DOESN’T CONSIDER WAVE BREAKING.
in figure 14 as a function of current and wave height, althought
they could be expressed in terms of spectral parameters instead.
The sharp right hand edge to the contour of constant response is
non-breaking waves, which for velocity would be β = 2 and particularly noticeable and is due to a reduction in the drag and
for response β = 1.3. However, if wave breaking is impor- lift coefficients with increasing current, which can also be seen
tant the former can be non-conservative and the latter overly in figure 2. All combinations of parameters on the contour of
conservative (particularly, as if wave breaking is important the return period response are a suitable set of conditions, however,
long term distribution tends to be very flat). Figure 9 shows the only a few are typically provided: the return period wave height
distribution of maximum velocity in three hour sea-states that all with the associated current and period; the return period current
have the same most probable maximum value, but have different with associated wave height and period; and a selection of cases
significant wave height, peak wave period and current. In order for different wave periods or wave steepness. By selecting
to account for the actual location (owing to current) and shape an appropriate range of conditions any mismatch between the
(varying with wave breaking) of the short term distribution, the actual structural response and the assumed response function
joint distribution of current with response in the extremes must will lead to an inefficient design rather than an non-conservative
be determined. This can be achieved by applying the method design. It is clear that using velocity as a proxy for the response
of [12] and [13], which involves a regression in Gumbel space. would give a rather different set of conditions from those using
the absolute stability, and hence, would lead to a more inefficient
design.

APPLICATION
The method has been applied in order to derive metocean
Figure 13 shows the variation in the most probable maxi-
criteria for pipelines located in the Eastern Mediterranean in
mum response for different assumptions about the diameter of
water depths from 5m to 700m. The metocean dataset is of
the pipe and the coefficient of friction. A doubling of the friction
9.3 years duration spanning from 1999 to 2009. It comprises
approximately halves the response, as the Fy /µ term in equation
hindcast waves and modeled currents, which have both been
2 dominates. A halving of the diameter reduces the response
compared to local measurements. The results from two water
by a factor of three, as it reduces the loading both directly
depths, 35m and 10m, are presented in this section.
(reducing the diameter itself) and indirectly (increases the
Keulegan-Carpenter number, and hence, reduces the drag and
lift coefficients). What is more important is the effect on the con-
Water Depth: 35m tours of the 100 year return period response. These are shown in
Figure 10 shows estimates of the 100 year return period figure 14, which indicates that whilst in this case the results are
pipeline response calculated using Stream function wave theory insensitive to the diameter, the coefficient of friction is important.
for a water depth of 35m from maximum likelihood fits of

7 Copyright © 2011 by ASME

Downloaded From: http://proceedings.asmedigitalcollection.asme.org/ on 08/30/2017 Terms of Use: http://www.asme.org/about-asme/terms-of-use

 35 22
Exponential Exponential σ = 4.0
Generalised Pareto Data
20
100 year return period MPM response (kN)

30

Most Probable Maximum Response (kN)

18

25
16

14
20

12

15
10

10 8 −2 −1 0
0 2 4 6 8 10 12 14 16 18 20 10 10 10
Threshold (kN) Probability of Exceedance

Figure 10. ESTIMATES OF THE 100 YEAR RETURN PERIOD MOST Figure 12. EXPONENTIAL FIT TO THE DATA IN 35M WATER DEPTH
PROBABLE MAXIMUM RESPONSE FROM EXPONENTIAL AND GEN- FOR A THRESHOLD OF 8kN.
ERALISED PARETO FITS TO THE DATA IN 35M WATER DEPTH.
14
D = 0.5m, µ = 0.4
1
D = 1.0m, µ = 0.8
12
0.9

0.8
10

0.7
Likelihood Ratio Test

8
0.6
Value

0.5 6

0.4
4
0.3

0.2 2

0.1
0
0 5 10 15 20 25
0 D = 1.0m, µ = 0.4
0 2 4 6 8 10 12 14 16 18 20
Threshold (kN)
Figure 13. SCATTER PLOT OF MPM RESPONSE IN STORMS SHOW-
Figure 11. LIKELIHOOD RATIO TEST FOR THE EXPONENTIAL AND ING THE EFFECT OF VARYING THE PIPELINE DIAMETER AND CO-
GENERALISED PARETO FITS TO THE DATA IN 35M WATER DEPTH. EFFICIENT OF FRICTION.

Comparison with the results of the FORM. Weibull distribution with scale α = 0.022 + 0.033Hs and shape
The inverse first order reliability method has been ap- β = 2.05 − 0.1Hs.
plied, following [14]. The distribution of significant
wave height is distributed according to the Exponential The FORM contours are shown in figures 15 and 16 in the
P(Hs > X) = exp [−(X/0.65)]. The distribution of peak plane of mean current and mean peak period respectively. The
wave period conditional on the significant wave height is surface associated with the 100 year return period condition has
distributed according to a Log-Normal distribution with the been searched for the point of maximum response calculated
mean and variance given as µ = 1.11 + 0.72 log[Hs + 2.59] and using the distributions of the response described by equations 8
σ2 = 0.0065 + 0.1 exp [−1.2Hs0.77 ]. Finally, the current condi- and 9. This is the design point, which has been used as the input
tional on the significant wave height is distributed according to a to a random time domain simulation in order to determine a

8 Copyright © 2011 by ASME

Downloaded From: http://proceedings.asmedigitalcollection.asme.org/ on 08/30/2017 Terms of Use: http://www.asme.org/about-asme/terms-of-use

 2 8
U
W
1.8
W µ=0.8 7
W D=0.5m
1.6
6
1.4

1.2 5
Uc (m/s)

Hs (m)
1 4

0.8
3
0.6

2
0.4

0.2 1

0
11 11.5 12 12.5 13 13.5 14 14.5 15 0
Hs (m) 0 5 10 15 20 25
Tp (s)

Figure 14. CONTOURS OF CONSTANT RETURN PERIOD VELOCITY

Figure 15. ENVIRONMENTAL CONTOURS OF ISO PROBABILITY
OR RESPONSE AS A FUNCTION OF SIGNIFICANT WAVE HEIGHT
DENSITY CALCULATED USING THE FORM: MEAN CURRENT.
(CONSTANT WAVE STEEPNESS) AND CURRENT SPEED FOR 35M
WATER DEPTH.
0.7

Table 1. 100 YEAR RETURN PERIOD ESTIMATES OF MOST PROB-

0.6
ABLE MAXIMUM (MPM) AND MAXIMUM, VELOCITY AND RESPONSE
FOR 35M WATER DEPTH.
0.5

0.4
Uw (m/s)

Parameter Wave Theory MPM Max Ratio Max/MPM 0.3

U (m/s) Linear 2.8 3.0 1.071

0.2
U (m/s) Stream 2.8 3.0 1.071
W (kN) Linear 28.2 30.7 1.089 0.1

W (kN) Stream 27.4 30.0 1.095

0
0 1 2 3 4 5 6 7 8
Hs (m)

Figure 16. ENVIRONMENTAL CONTOURS OF ISO PROBABILITY

more accurate estimate of the 100 year most probable maximum DENSITY CALCULATED USING THE FORM: MEAN SPECTRAL PEAK
response. Finally, the response has been increased by 9.5% WAVE PERIOD.
(table 1) in order to account of the short term variability. This
gives a response of 29.0 kN which is similar to that calculated
using the previous method. (no threshold), shown in figure 18, that the generalised Pareto
distribution is an appropriate model. This gives a 100 year
most probable maximum response of 37.0 kN. However, if
it is assumed that the short term distribution is described by
Water Depth: 10m a Weibull with a shape parameter β = 1.3 (ie non breaking)
Figure 17 shows estimates of the 100 year return period then the estimate of the maximum response is 49.3 kN - an
pipeline response calculated using Stream function wave theory increase of 33% over the most probable maximum value.
for a water depth of 10m from maximum likelihood fits of a However, in reality the shape parameter is much larger as many
generalised Pareto distribution to the most probable maximum of the largest events are associated with breaking waves. This is
response associated with storm events. It is clear from the fit clear in figure 18 from the flat shape to the long term distribution.

9 Copyright © 2011 by ASME

Downloaded From: http://proceedings.asmedigitalcollection.asme.org/ on 08/30/2017 Terms of Use: http://www.asme.org/about-asme/terms-of-use

 40 40
Generalised Pareto Generalised Pareto
Data
39
35

38

Most Probable Maximum Response (kN)

100 year return period response (kN)

30
37

25
36

35 20

34
15

33
10
32

5
31

30 0 −3 −2 −1 0
0 2 4 6 8 10 12 14 16 18 10 10 10 10
Threshold (kN) Probability of Exceedance

Figure 17. ESTIMATES OF THE 100 YEAR RETURN PERIOD MOST Figure 18. GENERALISED PARETO FIT TO THE DATA IN 10M WATER
PROBABLE MAXIMUM RESPONSE FROM GENERALISED PARETO DEPTH FOR A THRESHOLD OF 0 kN.
FITS TO THE DATA IN 10M WATER DEPTH.
0.45

0.4
In order to ascertain the correct shape of the short term
distribution the dependence between current and the response, 0.35

shown in figure 19, has been determined using the method 0.3
of multivariate extreme value analysis described in [12] (an
alternative approach would be to consider the dependence of 0.25
U (m/s)

the shape parameter itself). The details of the method are

c

0.2
not described here, but the resulting distribution of current
conditional on 100 year return period response is shown in 0.15
figure 20. This has mean value of 0.4 m/s, which together with
the most probable maximum response can be used to determine 0.1

the associated shape parameter as β = 2.3. If this is used in the 0.05

integration of short and long term response then the resulting
maximum response is 42.5 kN (a 15% increase on the most 0
0 5 10 15 20 25 30 35 40
probable maximum value). Response (kN)

Figure 19. SCATTER DIAGRAM OF PIPELINE RESPONSE AND CUR-

Estimates of the 100 year return period velocity calculated RENT.
using both linear and Stream function theories, and also the
response calculated using linear theory, have been made in the
same way and are shown in table 2. The results again indicate
that the choice of wave model has very little affect on either
the velocity or the response. Figure 21 shows the contours
of constant 100 year return period response and demonstrates
that the velocity is not an appropriate substitute. In contrast to
the results in a water depth of 35m, the plot also indicates that
assumptions about both the diameter of the pipe and coefficient
of friction are important.

10 Copyright © 2011 by ASME

Downloaded From: http://proceedings.asmedigitalcollection.asme.org/ on 08/30/2017 Terms of Use: http://www.asme.org/about-asme/terms-of-use

 0.8 Table 2. 100 YEAR RETURN PERIOD ESTIMATES OF MOST PROB-
ABLE MAXIMUM (MPM) AND MAXIMUM, VELOCITY AND RESPONSE
0.7 FOR 10M WATER DEPTH.

0.6

0.5 Parameter Wave Theory MPM Max Ratio Max/MPM

U (m/s)

U (m/s) Linear 3.1 3.35 1.08

c

0.4

U (m/s) Stream 3.0 3.36 1.12

0.3
W (kN) Linear 39.0 43.5 1.12
0.2 W (kN) Stream 37.0 42.5 1.15

0.1
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
Probability of exceedance
with the application of the FORM. The non-conservatism (or
Figure 20. THE DISTRIBUTION OF CURRENT CONDITIONAL ON A inefficiency) of using the combined wave and current induced
RESPONSE W = 37 kN. velocity as a proxy for the response has been established and in
shallower water the importance of considering the shape of the
6 short term distribution of response highlighted.
U
W
W µ=0.8 The method described in this paper used a two-dimensional
5 W D=0.5m
model of the absolute stability of the pipeline. However, during
a sea-state the movement of the pipe can lead to penetration
4 into the soil that significantly increases its resistance to lateral
movement. This alters the relative importance of waves and
currents, and hence, the design conditions for the pipeline.
Uc (m/s)

3
Therefore, a more sophisticated response based model would
be one that was able to model the three-dimensional movement
2 of the pipeline. This would also enable the spatial variation
in wave induced forcing along the pipeline to be considered
1
along with boundary layer effects. However, for this to be
appropriate, enough details of the pipeline design, route, and
the soil properties would have to be available. This will be the
0
1 2 3 4 5 6 7 8 subject of future work.
Hs (m)

Figure 21. CONTOURS OF CONSTANT RETURN PERIOD VELOCITY

OR RESPONSE AS A FUNCTION OF SIGNIFICANT WAVE HEIGHT
(CONSTANT WAVE STEEPNESS) AND CURRENT SPEED FOR 10M
ACKNOWLEDGMENT
WATER DEPTH.
Thanks go to the following: Colin Grant and BP for support-
ing this work; Peter Tromans and Chris Swan for many interest-
ing discussions on response based methods and nonlinear waves;
CONCLUSIONS AND FUTURE WORK and Jason McConochie for suggestions and advice.
A method for determining metocean design criteria for the
assessment of the on-bottom stability of offshore pipelines has
been developed. The method is based upon using the pipeline REFERENCES
weight required for absolute stability as a suitable response. The [1] Tromans, P., and Vanderschuren, L., 1995. “Response
method requires the short term distribution of the response to be based design conditions in the North Sea: application of
determined and this has been described in terms of the param- a new method”. In OTC 7683, Offshore Technology Con-
eters of the sea-state. The method has been applied to a site in ference, Houston.
the Mediterranean and has been shown to be in good agreement [2] Sabag, S., Edge, B., and Soedigdo, I., 2000. “Wake ii

11 Copyright © 2011 by ASME

Downloaded From: http://proceedings.asmedigitalcollection.asme.org/ on 08/30/2017 Terms of Use: http://www.asme.org/about-asme/terms-of-use

 model for hydrodynamic forces on marine pipelines in-
cluding waves and currents.”. Ocean Engineering, 27,
pp. 1295–1319.
[3] DNV-RP-F109. On-bottom stability design of submarine
pipelines, 2007. Tech. rep., DNV.
[4] Tromans, P. S., and Vanderschuren, L., 2002. “The statis-
tics of wave crest elevation calculated to second order.”. In
Proceedings of Conference OMAE2002-28535.
[5] DNV-RP-C205. Environmental conditions and environ-
mental loads, 2007. Tech. rep., DNV.
[6] Dean, R. G., 1965. “Stream function representation of non-
linear ocean waves.”. Journal of Geophysical Research, 70,
pp. 4561–4572.
[7] Grant, W., and O.S., M., 1986. “The continental-shelf bot-
tom boundary layer”. Annual Review of Fluid Mechanics,
18, pp. 265–305.
[8] Holmedal, L., Myrhaug, D., and Rue, H., 2003. “The
sea bed boundary layer under random waves plus current”.
Continental Shelf Research, 23, pp. 717–750.
[9] Battjes, J., and Groenendijk, H., 2000. “Wave height distri-
butions on shallow foreshores”. Coastal Engineering, 40,
pp. 161–182.
[10] Forristall, G. Z., 2000. “Wave crest distributions: obser-
vations and second-order theory.”. Journal of Physical
Oceanography, 30, pp. 1931–1943.
[11] Tromans, P., 2010. Private communication. Tech. rep.
[12] Heffernan, J., and Tawn, J., 2004. “A conditional approach
for multivariate extreme values.”. Journal of the Royal Sta-
tistical Society B.
[13] Jonathan, P., Flynn, J., and Ewans, K., 2009. “Joint mod-
eling of wave spectral parameters for extreme sea states”.
In 11th International Workshop on Wave Hindcasting and
Forecasting.
[14] Winterstein, S., Ude, T., Cornell, C., Bjerager, P., and
Haver, S., 1993. “Environmental parameters for extreme
response: inverse form with omission factors”. In ICOS-
SAR, Innsbruck.

12 Copyright © 2011 by ASME

Downloaded From: http://proceedings.asmedigitalcollection.asme.org/ on 08/30/2017 Terms of Use: http://www.asme.org/about-asme/terms-of-use