ARTICLE IN PRESS

Ocean Engineering 36 (2009) 330–338

Contents lists available at ScienceDirect

Ocean Engineering
journal homepage: www.elsevier.com/locate/oceaneng

Optimum selection of design parameters for transportation

of offshore structures
J.L. Alamilla, D. Campos , C. Ortega, A. Soriano, J.L. Morales
Mexican Institute of Petroleum, Eje central Lázaro Cárdenas 152, Col. San Bartolo Atepehuacan, Del. Gustavo A. Madero, 07730 México DF, México

a r t i c l e in f o a b s t r a c t

Article history: A probabilistic framework is presented to select the design significant wave height and design
Received 9 May 2008 transverse rotation for typical barges used in the Gulf of Mexico for marine transportation of structural
Accepted 30 December 2008 elements and/or systems. The selection of design transverse rotation is based on optimization procedure
Available online 20 January 2009
that trade-off between the performance of the barge subjected to a meteorological-oceanographic
Keywords: (metocean) hazard along the route and losses by structural damage. For this purpose probabilistic
Expected losses models to estimate the metocean hazard for marine transportation are shown. Afterwards, the design
Metocean hazard rotation is linked to the design of significant wave height and to the return period associated with such
Marine structure transportation wave conditions. The formulation is applied to an offshore transportation route in the Gulf of Mexico.
Significant wave heights
& 2009 Elsevier Ltd. All rights reserved.
Barge response
Exceedance probability
Design parameters

1. Introduction characteristics (marine securing) for the structure fastening to

The design of offshore structural systems considers two
phases. The first one is related to structural performance
subjected to loads and possible deterioration conditions during
the structure lifetime. The second considers the works required to
transport the structural systems from the fabrication yard to the
site of their placement and during the installation itself. The
purpose of the design for transportation, in this last phase, is to
avoid loss of stability of the barge-structure system as well as the
possible damage in joints and elements of the structural system
transported. As shown in Fig. 1, structural systems are placed on
the barge, and this is towed by a small vessel called tug boat.
The transportation starts from the fabrication yard and ends at the
installation site or vice versa. During this phase, the barge-
structure complex is subjected to dynamic loads, originated by
possible oceanographic and meteorological conditions prevailing
along the transportation route. During the design phase, the
barge-structure system is subjected to an ergodic and stationary
sea state, represented by spectral density of the wave in multiple
directions. The spectral density is characterized by a spectral peak
period and a significant wave height that is associated with a
given recurrence period. As a result of the dynamic analysis, the
stress state in structural elements and joints is checked not to
exceed the allowable stress. In the opposite case, the structural
system is strengthened. Together with this, mooring system
 Corresponding author.
E-mail address: dcampos@imp.mx (D. Campos).
the barge are determined. Mooring system must be able to stand
the stress state it is subjected to during the travel.
In Mexico, two reference codes in use are combined to make
the design for marine structure transportation, NRF-003-PEMEX-
2000 (PEMEX, 2000) and NRF-041-PEMEX-2003 (PEMEX, 2003).
The first one recommends a significant wave height with its
associated meteorological and oceanographic parameters, valid
for manoeuvres performed on typical routes within the Gulf of
Mexico and out of the hurricane season. This code mentions that
for manoeuvres performed during hurricane season, a reliable
forecast of environmental conditions should be available in order
to ensure sea states favorable for transportation. For other routes
and seasons, one should refer to NRF-041-PEMEX-2003 code
(PEMEX, 2003). This code recommends extreme environmental
conditions with the 10-year return period, without providing any
parameter, in the most exposed parts of the routes and furnishes
prescribed values of movements dependent upon transport
season, load magnitude and ship dimensions.
The American Petroleum Institute (API, 2000) recommends a
probability of environmental conditions from 1% to 5% to be
exceeded during transportation. Besides mentioned that must be
taken into account the barge-structure transition time (duration
of transportation), the accessibility to safety zones in the coast,
the climate and the characteristic of the barge. Similar recom-
mendations are given in references (LOC, 2003; NDI, 2002, 2005;
IMO, 1998).
A rigorous formulation to obtain in optimum manner the
design parameters for transportation of offshore structures is
presented. This formulation takes into account the possible

0029-8018/$ - see front matter & 2009 Elsevier Ltd. All rights reserved.
doi:10.1016/j.oceaneng.2008.12.008
 ARTICLE IN PRESS
J.L. Alamilla et al. / Ocean Engineering 36 (2009) 330–338
Fig. 1. Structural systems transported by a barge.
systematic reconstructions of the structure, the associated
economical losses and the structural performance due to the
corresponding metocean hazard are evaluated. A general metho-
dology is presented to estimate the sea state and the dynamical
response of barge-structure, with the corresponding returns of
periods recommended to the design of transportation of marine
structures. Here the sea state is characterized by a spectral
density, specified in terms of a spectral peak period and a
significant wave height of a given recurrence period. The
dynamical response is specified by the transverse rotation (roll)
of the barge. The annual exceedance rate of significant wave
height is estimated here for a specific case; it is associated with a
known climate-time-route window and applied to every barge.
However the methodology can be applied to other cases where
the characteristics of the set of barges and the transportation
routes are known. To achieve it, a probabilistic model is developed
to estimate the metocean hazard along the transportation route
and its influence on the dynamic response of barges used for
transportation. The ideas of seismic hazard studies (Esteva, 1976,
1968; Cornell, 1968) are extended to the evaluation of the
metocean hazard for marine transportation of given duration.
The formulations provide a logic form to evaluate the metocean
hazard of the marine route and the vulnerability of barges in these
routes. To take into account the transition time of the barge, a
formulation is presented to estimate the exceedance probabilities
of significant wave heights associated with transport of given time
duration, based on the first-passage problem (Ditlevsen and
Madsen, 2007; Madsen et al., 1986).
With the purpose of evaluating the design requirements
specified in the established codes (PEMEX, 2000, 2003; API,
2000), the formulations are applied to real sea states and real
structures transporting in a region of the Gulf of Mexico. This
evaluation consists of determining in optimum manner the barge
rotation for a transportation route associated with a given time
interval (5 and 10 days). Afterwards, the return period associated
with such rotation and the corresponding significant height wave
are determined. Finally, the estimated values are compared with
the corresponding values specified in the codes (PEMEX, 2000,
2003).
2. Theoretical framework for the optimum selection of design
parameters
In this section a probabilistic formulation to select the design
transverse rotation for typical barges used in the Gulf of Mexico
for marine transportation of structural systems is described. Here
the characteristics of the barges are given; however the formula-
331
tion can be used to select the optimum characteristics of the
barge. The optimum design rotation j*D corresponds to the
transverse design rotation that minimizes the economical losses
and satisfies the safety conditions of the barge-structure system
among a set of the possible design rotations jD values. To the end,
the value of the design rotation to be specified in codes should be
the optimum design rotation.
During the route, the performance of the barge-structure
system, characterized by its stability and integrity, can be affected
by one or more sea states. In general the transportation results can
be synthesized in the following two possible scenarios:
(1) The barge survives; however the transported structural
system can present certain level of damage. This scenario
include the extreme cases: (a) the structural system does not
have any damage, and (b) the structural system is completely
damaged.
(2) The barge-structure system fails by instability.
In the following, these scenarios are taken into account in a
probabilistic formulation to establish a trade-off between eco-
nomical losses and barge-structures system safety. To select the
design transverse rotation, the total economical losses CT are
minimized, which include the losses by damages in the trans-
ported structure and the total losses of the barge-structure
system. An important point to consider is the systematic
restitution in case the structure is damaged during the voyage.
The expected value of CT is obtained as follows:
X
1
E½C T ðjD Þ ¼ C 0 ðjD Þ þ C D ðjD Þ ð1  pF T Þpn1 gðn1Þt F
FT e
n¼1
X
1
þ ðC 0 ðjD Þ þ C F Þ pnFT egðn1ÞtF (1)
n¼1
The linear operator E[  ] denotes expectancy. The first term of
right side of this equation denotes the expected initial cost C0(jD),
which includes the costs of mooring system designed to fix the
structure onto the barge, the structural evaluation, the construc-
tion and the installation of the mentioned mooring system. The
second term denotes the expected value of possible structural
damages occurred during the transportation, given that the barge-
structure system survived during the route; that is to say the
barge does not fail by instability at the nth shipping out to sea. The
second and third terms take into account the possible structural
reconstructions and their systematic shipping out to sea. In
Eq. (1), pF T denotes the failure probability of instability of barge-
structure system associated to the marine transportation route.
This instability is associated to the barge rolling over. The term
ð1  pF T Þpn1
FT denotes the survival probability at the nth shipping
out to sea, given that in previous n1 shipping out to sea, the
barge failed by instability. The term pnFT denotes the instability
failure probability at the nth shipping out to sea, given that in
previous n1 shipping out to sea, the barge failed by instability.
The capitalization function egðn1ÞtF takes into account the future
costs expressed in present value, with the discount rate g
corrected by inflation (Rosenblueth, 1987), and tF is the time
spent in the new planning and structural reconstruction. For the
first shipping out to sea (n ¼ 1), the capitalization function is
equal to one.
On the other hand, CF denotes the expected total loss of the
barge-structure system and includes the cost by structural design
and construction, economical losses by deferred production,
barge, human injuries and human losses. In the last term of
Eq. (1), the cost C0(jD) is added due to the reconstructions
and shipping out to sea. The expected economical loss CD(jD)
corresponds to the expected structural damages and it is
 ARTICLE IN PRESS
332 J.L. Alamilla et al. / Ocean Engineering 36 (2009) 330–338
estimated as follows:
Z jF Z 1
C D ðjD Þ ¼ C D0 zðjD ; yÞf jjHðtE Þ ðyjxÞf HðtE Þ ðxÞ dx dy
0 0
In this equation, C D0 is the total cost of structural damage and
the damage function z(jD,j) is a function of the design rotation
jD and the maximum barge rotation j experienced during the
transportation. Here, f jjHðtE Þ ðÞ is the probability density function
of the barge rotation conditioned to given height waves and fH(  )
is the probability density function of the height waves associated
to the exposure time tE during the marine transportation route. A
limit of the integral of Eq. (2) is jF, because the structural
damages are conditioned to the barge survival. Note that for
jojD, the expected damage on the structure must be null,
meanwhile for jXjD, it is expected a certain level of damage.
Accordingly, the Eq. (2) can be represented as
Z jF Z 1
C D ðjD Þ ¼ C D0 zðjD ; yÞf jjHðtE Þ ðyjxÞf HðtE Þ ðxÞ dx dy (3)
jD 0
Eq. (1) can also be expressed as
E½C T ðjD Þ ¼ C 0 ðjD Þ þ ½C D ðjD Þð1  pF T Þ
1 (1986) and it was considered a time deferred production of two
þ ðC 0 ðjD Þ þ C F ÞpF T 
1  pF T expðgt F Þ
If the economical losses are expressed as a fraction of the total
failure cost: DT(jD) ¼ E[CT(jD)]/CF, D0(jD) ¼ C0(jD)/CF, DD(jD) ¼
CD(jD)/CF, then Eq. (4) can be rewritten as follows:
DT ðjD Þ ¼ D0 ðjD Þ þ ½DD ðjD Þð1  pF T Þ þ ðD0 ðjD Þ þ 1ÞpF T 
1
 (5)
1  pF T expðgt F Þ
An immediate consequence is that D051 and DD(jD)o1. If
pF T 51, then the Eq. (5) can be simplified and rewritten as
DT ðjD Þ  D0 ðjD Þ þ DD ðjD Þð1  pF T Þ þ ðD0 ðjD Þ þ 1ÞpF T
In this equation it can be noted that DD ðjD Þ  DD ðjD Þð1  pF T Þ,
which is similar to the Rosenblueth (1976) formulation.
On the other hand, if the occurrence process of the maximum
significant height waves is characterized by a Poisson process,
then the barge probability exceeded a given rotation during the
exposure time tE is exponential with parameter v(j)tE, where v(  )
is the annual exceedance rate barge rotation. This way, jF is a
rotation threshold, the system fail when the barge rotation above
this value and the system is completely stable when it is below
this threshold. This means that probability failure of system is
associated to an exposure time and can be obtained as
pF T ¼ 1  exp½nðjF Þt E 
If the probability density function f jF ðÞ characterizes the
uncertainties of the barge rotation instability, then pF T is obtained
as
Z 1
pF T ¼ 1  exp½nðxÞt E f jF ðxÞ dx (8)
0
2.1. Damage functions
The losses associated to the structural damage can be
expressed as the sum of the contribution of the loss of each
structural element (Eq. (9)). The economical losses are repre-
sented as the product of the mean total losses DDi of the ith
structural element by its corresponding damage function zi(  ),
expressed in terms of the rotations j and jD. If the barge is
subjected to a sea state, then it is to be expected a certain level of
damage in case that maximum rotation j is greater than the
design rotation jD.
X
(2) DD ðjD ; jÞ ¼ DDi zi ðjD ; jÞ (9)
i
From empirical information of the performance of typical
transported structures, it is possible to represent, in similar form
to Eq. (9), the expected total damage losses DD ðjD ; jÞ ¼
DD0 zðjD ; jÞ as the product of the normalized expect total cost
DD0 ¼ ðC D =C F Þ and the damage function z (Eq. (10)), 0pzp1, with
parameters k ¼ 0.0001 and e ¼ 2.8. These parameters were
obtained from a few statistics information of barge performance
at the Gulf of Mexico and mainly from subjective information. The
damage function given by Eq. (10) considers that there are not any
damage for jojD. Also the damage is practically null for j values
slightly greater than jD. This effect is controlled by the parameter e.
Here, CF value was estimated regarding the main costs as the
offshore structure cost, the deferred production and the barge
cost. For the analysis a typical offshore jacket platform of eight
legs was considered. From an analysis it was estimated that the
cost of offshore structure takes 40% of the deferred production
cost and the barge cost 8%, which represent a small fraction of
total cost. The deferred cost was estimated according to Stahl
(4)
years and structure service time of twenty years. Accordingly,
DD0  0:95.
(
0 jojD
zðjD ; jÞ ¼ (10)
1  exp½kðj  jD Þ  jD pjojF
According to empirical information, the initial losses can be
represented by Eq. (11). The parameter m0 ¼ 0.005 denotes the
normalized expected loss, which considers the structural raising
and placing cost. These are independent on jD. The term m1 jaD1 ,
with parameters m1 ¼ 0.0001 and a1 ¼ 1.3, denotes the normal-
ized mean cost of material and placing of mooring system. The
(6)
variation of these losses is exponential because the greater jD,
the difficulties in placing the mooring systems grows and more
time is needed. Whereas the term m2 ðjD  jE Þa2 with parameters
m2 ¼ 0.0001, a2 ¼ 1.8 and aE ¼ 251 considers the structural
strengthening as a function of aD. From engineering designs using
the NRF-041-PEMEX-2003 code with a design transverse rotation
response of 251, only few structures have been strengthened.
Here this rotation was taken as a minimum rotation threshold.
The variation of these losses is exponential because the greater aD,
the more strengthening and time is needed.
(
m0 þ m1 jaD1 jD pjE
D0 ðjD ; jE Þ ¼ (11)
m0 þ m1 jaD1 þ m2 ðjD  jE Þa2 jD 4jE
(7)
2.2. Metocean hazard
In order to apply the above formulation, it is necessary to
characterize the metocean hazard along the route. In this section,
three probabilistic models to estimate the metocean hazard for
marine transportation are shown. The first model allows estimat-
ing the annual exceedance probability for significant wave heights
for a given route. The second model allows estimating this
probability for the time period that the barge is exposed to sea
states during the specified route. Finally, the last model allows
estimating the exceedance rate for structural response of the
barge employed. The structural response refers to transverse
rotation of the barge (roll movement) that is measured in degrees.
The developed formulations are supported in the knowledge
that we have oceanographic information about the marine route.
In this work this information was provided by Oceanweather Inc.
 ARTICLE IN PRESS
J.L. Alamilla et al. / Ocean Engineering 36 (2009) 330–338
(2006), and consists of significant wave height data with its
corresponding return periods and spectral peak period, and
corresponds exclusively to winter storms.
2.2.1. Exceedance probabilities of significant wave heights for given
marine transportation routes
Here, the metocean hazard associated with a marine transporta-
tion route is specified in terms of the significant wave height. This
hazard is represented by the exceedance rate x(h) of significant
wave heights h. Exceedance rate describes the mean annual number
of significant wave heights exceeding a given height. From this rate,
the exceedance probabilities are determined.
Oceanweather Inc. (2006) provided values of return periods for
a set of coordinates xk, {k ¼ 1,yy, n} specified by their latitude
and longitude as shown in Fig. 2. Based on the information
available, a given metocean hazard corresponds to each one of the
coordinates. As shown in Figs. 3 and 4, this hazard is characterized
by the exceedance rate of significant wave heights, as well as by
the spectral peak period (Tpeak) of each wave height.
If F Hk ðÞ is the probability at the coordinate k, the wave height is
Hkph, where Hk is the random variable of significant wave
heights, then the exceedance probability for Hk4h0 value can be
specified as
1  F Hk ðhÞ ¼ P½Hk 4h0 P½Hk 4hjh0 
According to this equation, the mean annual number of events
exceeding a given significant wave height Hk ¼ h, at the
coordinate xk of the route can be related with the distribution
function F Hk ðÞ, as follows:
xk ðhÞ ¼ x0k ð1  F Hk ðhÞÞ
where xk(h) is the annual mean number of exceedance of
significant height wave h, and x0k ¼ xk ðh0 Þ, the total annual mean
number of exceedance of the considered minimum height h0.
Height waves minor to h0 occur typically during the transporta-
tion route, then the design parameters must be above h0.
If it is assumed that barge travels at specified speed and that
metocean hazard in a region of the size ak is represented by
the exceedance rate corresponding to the coordinate xk, then the
hazard will be changing along the transportation route. That is to
say, during the transportation, the barge may be subjected to
different wave conditions that depend on a great extent on the
characteristics of each one of the route points. Hence, the
probability of the barge, traveling with velocity vk and being
exposed to metocean hazard of a region ak, is given by
p0k ¼ ðak =vk Þ=t E , where
X
n
ak
tE ¼
k¼1
vk
As shown in Fig. 2, these regions are associated with an area of
influence. The measurements of point k will be associated with an
coordinate Sea
k-1
k route can be expressed as
k+1
n xðhÞ
Coast
Fig. 2. Coordinate system of the marine transportation route.
333
area of influence ak ¼ (lk1,k+lk,k+1)/2, where lk1,k expresses the
distance between the coordinates k1 and k, while lk,k+1 expresses
the distance between the coordinates k and k+1, that is, for
the whole k ¼ 1,y, n. Accordingly, the mean number of waves
exceeding a given wave height, during the transportation can be
obtained as
X
n
xðhÞ ¼ p0k x0k ð1  F Hk ðhÞÞ (15)
k¼1
If the barge is always at the kth region, then Eq. (15) will be
x(h) ¼ xk(h). The equation is an average of the hazard during the
route taking into account the exposure time to each hazard. In
accordance with the above Eq. (5), the probability that a
significant wave height is exceeded along the transportation
route is obtained as
1X
n
P½H4h ¼ p0k x0k ½1  F Hk ðhÞ (16)
x0 k¼1
where the mean annual number of waves x0, is associated with a
given route and is obtained as
X
n
x0 ¼ p0k x0k (17)
k¼1
(12)
2.2.2. Exceedance probabilities of significant wave heights for given
marine transportation routes and transport of given time duration
The exceedance rate of significant wave height or its corre-
sponding exceedance probability is associated to the annual
window of winter storms. This probability is related to the case
(13)
that a barge always is navigating. However the barge only will be
subjected to one or more sea states during a short time interval. It
is important to estimate the exceedance probability that during
the time-route the barge is exposed to different sea states.
Here, the purpose is to determine the probability distribution
function that the wave of H ¼ h height is exceeded within a
given time period. This time period corresponds to transport
duration tE (5 or 10 days). Accordingly, it is necessary to determine
P[H(tE)4h]. This probability involves possible significant wave
heights along the route as well as the occurrence of these waves in
(0,tE].
The probability that occurs one or more wave heights H4h in
(0,tE] can be expressed as
" #
occurs one or more wave
P½Hðt E Þ4h ¼ P P½H4h
heights H4h; in ð0; t E 
¼ ð1  P½Nðt E Þ ¼ 0jH4hÞP½H4h (18)
(14)
where P[H4h] ¼ x(h)/x0 is the probability that the significant
wave heights are greater than h. P[N(tE) ¼ 0|H4h] is the prob-
ability that does not occur any event in the interval (0,tE]. Note
that the probability given by Eq. (6), is considered as the first-
passage problem (Ditlevsen and Madsen, 2007; Madsen et al.,
1986). Since the process of occurrence of significant wave heights
can be represented by a Poisson process, the exceedance
probability distribution of significant wave heights during a given
P½Hðt E Þ4h ¼ ½1  expðxðhÞt E Þ (19)
x0
Note that this probability is a function of the exceedance rate
x(h), which takes into account the exposure time at a given region.
This means that this probability is related to the barge location of
the barge. However, it does not consider spatial dependence of
two height waves at two regions separated at a given distance.
According to the last equation the mean number of wave heights
 ARTICLE IN PRESS

334 J.L. Alamilla et al. / Ocean Engineering 36 (2009) 330–338

100 100
100

Exceedence annual rate (ξ2)

Exceedence annual rate (ξ3)

Exceedence annual rate (ξ1)

10 10
10

1 1 1

0.1 0.1 0.1

0.01 0.01 0.01

1E-3 1E-3 1E-3

1E-4 1E-4 1E-4

10
11
12
13
14
15
16

10
11
12
13
14
15
16

10
11
12
13
14
15
16
3

6
7
8
9

6
7
8
9

6
7
8
9
Hs (m) Hs (m) Hs (m)
100 100 100
Exceedence annual rate (ξ4)

Exceedence annual rate (ξ6)

Exceedence annual rate (ξ5)
10 10 10

1 1 1

0.1 0.1 0.1

0.01 0.01 0.01

1E-3 1E-3 1E-3

1E-4 1E-4 1E-4

10
11
12
13
14
15
16

10
11
12
13
14
15
16

10
11
12
13
14
15
16
3

6
7
8
9

6
7
8
9

7
8
9
Hs (m) Hs (m) Hs (m)

100 100 100

Exceedence annual rate (ξ9)

Exceedence annual rate (ξ8)
Exceedence annual rate (ξ7)

10 10 10

1 1 1

0.1 0.1 0.1

0.01 0.01 0.01

1E-3 1E-3 1E-3

1E-4 1E-4 1E-4

10
11
12
13
14
15
16

10
11
12
13
14
15
16

10
11
12
13
14
15
16
3

7
8
9

7
8
9

7
8
9
Hs (m) Hs (m) Hs (m)

100 100
Exceedence annual rate (ξ10)

Exceedence anual rate (ξ11)

10 10

1 1

0.1 0.1

0.01 0.01

1E-3 1E-3

1E-4 1E-4
10
11
12
13
14
15
16

10
11
12
13
14
15
16
3

7
8
9

7
8
9

Hs (m) Hs (m)

Fig. 3. Annual exceedance rate of significant wave heights for several points located over the transportation route. The dots represents the estimates of Oceanweather Inc.
(2006) and the line is the fitted. The axes are in logarithmic scale.

associated to a given route of given time duration can be to determine exceedance rate of intensities is given herein. On
obtained as the basis of transportation barges with uncertain properties, for
each sea state characterized by a significant wave height h in
x0ðtE Þ ¼ x0 P½HðtE Þ4h0  a region k, a probability density function f Z k H ðÞ of rotation Z ¼ z
¼ x0 ð1  expðxðh0 Þt E ÞÞ (20) (roll) can be obtained. Accordingly, the probability density
function of the rotation associated with the ship’s route for a
given sea state, may be represented by Eq. (21). For each point
2.2.3. Exceedance probabilities of barge response along the route, there is a probability function of rotation
In order to relate the dynamic response of barges subjected conditioned to given values of significant wave height. The
to sea states with given return periods, a probabilistic formulation total probability function depends on the exposure time in
 ARTICLE IN PRESS

J.L. Alamilla et al. / Ocean Engineering 36 (2009) 330–338 335

17 analysis to quantify wave return periods. The return periods

of significant wave heights h given by Oceanweather Inc. (2006)
16 for each one of the n ¼ 11 coordinates were transformed to
exceedance rate values as shown in Fig. 3. Exceedance rates near
15 to 104 per year were obtained by means of extrapolation.
A practical form to represent the exceedance rate associated to
each point is by means of an appropriate continuous function.
14
Tpeak (s)

This function must be related with an extreme value distribution.

However as it is shown in Fig. 3, the Eq. (24) fitted adequately,
13 even for high values of Hk ¼ h and has the advantage to be simple.
Here x0k is the mean number of waves associated with the
12 minimum height threshold h0 of significant wave height, herein
considered as 3.0 m. The assumption of this value is discussed
11 subsequently. The parameter x1k describes the form of exceedance
rate. The parameters x0k and x1k were obtained by least-square fit.
10 The continuous line in Fig. 3 shows exceedance rates fitted to
6 7 8 9 10 11 12 13 14 15 16 values provided by Oceanweather Inc. (2006). With the purpose to
h (m) verify the significant statistical differences between the rates
given by Oceanweather Inc. (2006) and the corresponding rates
Fig. 4. Spectral peak period vs. significant wave heights corresponding to estimated by Eq. (24), the authors obtained confidence intervals
coordinates over the transporting route. of the quotient of the fitted value rate divided between the
corresponding rates given by Oceanweather Inc. (2006). This
each point. quotient was selected for the analysis because all rate values
(small and greater) can be compared in the same way. The sample
X
n
f ZjH ðzjhÞ ¼ f Zk jH ðzjhÞp0k (21) mean of this quotient must be one, with standard deviation of
k¼1 0.003. In all the cases, the mean of the quotient is within the range
of values between 0.998 and 1.002, which corresponds to a
From the probability density function fZ|H(  ) and the marginal
probability of 0.95. This small range indicates that the Eq. (24) can
probability density function of height waves, given by x1
0 dx(h)/
be used to represent adequately the exceedance rates.
dh, and applying the probability total theorem, the unconditioned
probability distribution of barge rotations can be expressed as xk ðhÞ ¼ x0k expðx1k ðh  h0 ÞÞ (24)
Z
1 1 dxðhÞ It is possible to observe that the maximum wave for a return
P½Z4z ¼  ð1  P½Z4zjH ¼ hÞ dh (22)
x0 0 dh period of 100 year, of winter storms, takes height values between
7 and 12 m. These height waves are agreeable with the observed
Substituting adequately Eqs. (21) and (15), and multiplying by
storm of the October 23 of 2007, in which occurred very height
x0, the Eq. (21) can be expressed in terms of the exceedance rate of
waves reaching the first deck of platform type jacket, and its
the response along the route of barges as follows:
elevation above the sea is 17.5 m. This can be seen in Fig. 5.
X
n Z 1 According to Eqs. (1) and (14), the probability distribution
dxk ðhÞ
nðzÞ ¼  p0k ð1  P½Z k pzjhÞ (23) function F Hk ðhÞ in every coordinate k is exponential, which is
k¼1 0 dh
obtained from Eqs. (24) and (13), and can be expressed as follows:

F Hk ðhÞ ¼ 1  expðx1k ðh  h0 ÞÞ (25)

3. Application case The exceedance rate corresponding to transportation route
obtained from Eq. (3) is shown in Fig. 6. With dot line the
In this work the transportation route is defined by 11 exceedance rate for the different coordinates along the route are
geographic coordinates corresponding to the transportation route
specified in PEMEX (2000) and to the points with available
measurements. The route is located in front of the coast of Mexico,
from Tampico to Cantarell oil field. Historically, in this zone of the
Gulf of Mexico, the intensities of winter storms with south
direction are greater than the intensities from hurricanes because
the Peninsula of Yucatan is a natural barrier against hurricanes.
Oceanweather estimated return periods for given significant
height waves for 11 coordinates along the route. For this purpose
Oceanweather worked with the main 80 winter storms which
occurred in the time interval between 1958 and 1999. Besides the
storms data, Oceanweather also took into account directions and
wind velocities estimated from the Quick Scatterometer satellite
and from databases of National Center for Environmental
Prediction and from National Center for Atmospheric Research.
It also took into account measurements of wind velocities,
currents and height waves from 42055 NOAA buoy, located in a
zone with a deep of 3380 m. For each winter storm event,
Oceanweather simulated wind fields over the ocean surface and
estimated other parameters besides significant height waves and
spectral peak periods. Finally, it performed an extreme value Fig. 5. Picture of winter storm in the Bay of Campeche on October 23, of 2007.
 ARTICLE IN PRESS

336 J.L. Alamilla et al. / Ocean Engineering 36 (2009) 330–338

shown. It is shown that the significant wave height for a return
period of 100 years is 10.5 m. For the recurrence period, the mean
wave height that predominant along the route is about 10 m. As
we expected, the mean rate associated with the route, Eq. (3), is
inside the rates related with every coordinate.
According to Eq. (8), Fig. 7 shows probabilities to exceed given
values of significant wave heights associated with 5- and 10-day
periods of transportation. The probabilities of exceeding great
wave heights while en route for 5 or 10 days are lower than the
corresponding probabilities associated with very long times. With
the aim of visualizing the evolution of these functions with the
time, exceedance probability functions for 1 and 100 years are
shown in Fig. 7. According to Eq. (8), when tE-N, the form of
exceedance probability function tends to the overall exceedance
probability.
On the other hand, a detailed analysis based on the informa-
tion provided by Oceanweather Inc. (2006), determined that the
natural logarithm of the spectral peak period T Peakk of the wave at
each coordinate k, is the linear logarithm function of the
significant wave height. Therefore, the values reported were used
to relate T Peakk with the significant wave height h, using the
100
10
Exceedence annual rate
following function:
T Peakk ðhÞ ¼ expðg0k þ g1k lnðhÞÞ (26)
where g0k and g1k are fitted parameters, associated with each
coordinate. Fig. 3 shows fitted functions which are seen to fit
adequately to the reported values. Dynamic characteristics of the
wave are observed to change along the marine transportation
route, resulting in different barge responses, even for a given
wave height. The lowest dispersion in the spectral peak period
corresponds approximately to a significant wave height of 9 m.
The variability in this period increases with the wave height. In
this paper, the International Ship Structures Congress (ISSC)
Spectrum is used. The ISSC Spectrum is a two parameters
spectrum (Chakrabarti, 1994). The two parameters for ISSC
Spectrum are significant wave height (h) and spectral peak period
(TPeak).
The barge-structure system response was obtained from a
dynamical analysis performed in frequency domain using the strip
theory to estimate the forces on the barge. In order to estimate the
roll response, it was considered that barge system is affected
mainly by the perpendicular waves to the barge course. In general
the marine route is located in front of coast and the directions of
wave propagation are perpendicular to this. However, rigorously it
should be taken into account the influence of all propagation wave
directions (Aalbers et al., 1996; Baxevani et al., 2000), but it is
beyond of this work.
The variability of dynamic responses of the barges will be the

function of both the sea state (significant wave height and spectral
1 peak period) and barge characteristics (transported mass and
geometric properties of the barge). However, the analysis of
dynamic response has shown that the transported mass does not
0.1
significantly influence the dynamic response variability since the
weight is controlled with water used as ballast in barge tanks. The
0.01 ballast is used to ensure equilibrium and stability conditions of
the barge.
In this paper, the probability density function of dynamic
1E-3 response of the barges associated with uncertain sea states of
given significant wave heights, Eq. (9), was obtained using Monte
Carlo simulation. To this purpose, eight systems of barges
1E-4
typically used in the Mexican oil industry were subjected to
3 4 5 6 7 8 9 10 11 12 13 141516
possible sea states that may affect barges along the route.
h (m)
Simulated values are shown in Fig. 8; as expected, the transverse
Fig. 6. Annual exceedance rate of significant wave height corresponding to rotation of the barge is seen to increase with the wave height,
coordinates over the transporting route. The axes are in logarithmic scale. so the greater is the significant wave height, the greater is
the response range. For significant wave heights above 9 m, the
response range increases significantly as a consequence of the
1
infinity 120
5 days 110
10 days 100
0.1
1 year
roll response (degrees)

100 years 90
P [H (t0) > h]

80
70
0.01
60
50
40
1E-3 30
20
10
1E-4 0
3 4 5 6 7 8 9 10 11 1213141516 0 2 4 6 8 10 12 14 16 18 20 22 24 26
h (m) h (m)

Fig. 7. Exceedance probability of significant wave height for routes of given time Fig. 8. Simulations of dynamical response of roll motion of barges as a function of
duration. The axes are in logarithmic scale. significant wave height for a given transportation route.
 ARTICLE IN PRESS
J.L. Alamilla et al. / Ocean Engineering 36 (2009) 330–338
variability in dynamic characteristics of the wave along the route,
specifically of the spectral peak period variability. In this work it is
assumed that barge fails due to instability as rotations value is
greater than 501. Above this value the linear analysis is considered
also unrealistic; even in Fig. 7 is shown responses greater than this
value.
In Fig. 8, the concentration of simulated values is observed to
change with the wave height, which makes difficult the use of an
analytical probability function, reported in the literature. There-
fore, an empirical model was used in this work to estimate the
probability of exceedance from simulated values as P[Z4z] ¼
1(i(z)/(M+1)), where i(z) is the number of simulated rotations,
smaller or equal to z and corresponding to a given wave height
value. M is the overall number of simulations associated with the
wave height in question.
According to Eq. (9), the probability density function fZ|H(  ) is a
function of the associated f Z k jH ðÞ in the coordinate k. It is
necessary to obtain explicitly the probability density functions
of f Z k jH ðÞ. This can be avoided by simulating rotations for given
wave heights. Here, to estimate fZ|H(  ), we obtain the coordinate k
by simulation, after this, and from the metocean characteristics in
the simulated coordinate, we obtain the maximum value of the
roll response of the barge. This response is obtained numerically
from a dynamic analysis of barge. This is performed a great
number of times and from the simulated rotation values for given
wave heights; the probability density function fZ|H(  ) is obtained.
According to this, the exceedance rate given by Eq. (10) is obtained
as follows:
Z 1
dxðhÞ
nðzÞ ¼  ð1  P½ZpzjhÞ dh (27)
0 dh
In this work h0 is the minimum significant wave height, which
is related to the hazard for the barge-structure system. The hazard
does not change significantly as wave height hph0. On contrary,
when h is greater than h0, hazard changes greatly. This can be seen
in Fig. 9, where the hazard associated to barge response of about
151 does not change. This hazard changes for responses greater
than this value. In Fig. 6, it is observed that the wave height
associated with the response of 151 is 3 m approximately. For this
reason, in this work the value h0 was taken to be 3 m. This value
plays an important role, because it is associated with the
exceedance probabilities for given wave heights.
Currently, the marine transportation design using the NRF-
003-PEMEX-2000 code (PEMEX, 2000) is made with a significant
100
10
Exceedence annual rate, ν
1 Damage
0.1
0.01
1E-3
1E-4
10 20 30 40 50 60 70 8090100
roll response (degrees)
Fig. 9. Annual exceedance rate of roll movement for a given transportation route.
The axes are in logarithmic scale.
337
wave height of 5.3 m, whereas the NRF-041-PEMEX-2003 code
(PEMEX, 2003) establishes a design with a minimum transverse
rotation response of 251, approximately corresponding to a
10-year return period. Fig. 6 shows that a significant wave height
chosen for a 10-year return period would correspond to the height
of 8.6 m and to the transverse rotation of 411 (see Fig. 9); this
means that the barge-structure system will become unstable or
that the reinforcement of structural and mooring systems would
be doubled with regard to the standards currently in force.
Professional experience with the use of 5.3 m height and 251,
specified in the current standards, has shown apparently
satisfactory behavior; however, the amount of reinforcement
for structural elements and the use of mooring structures are
considered to be high.
With the aim of establishing bases for the selection and review
of parameters to control the metocean hazard, we have developed
probabilistic models shown in the above sections and fed from
updated information provided by Oceanweather Inc. (2006). In
Fig. 10, it is shown the expected losses for a transition time of
5 days. The rotation that minimizes the total losses is between 201
and 231. For engineering applications the optimal design rotation
j*D can be taken as 231. According to Fig. 10, the designs for
values less than j*D may result in a certain level of damage. The
damage will increase as the design rotation decrease. An extreme
case is the scenario where the structure is not fixed to the barge;
the damages will be very high.
The total loss was obtained by means of Eq. (3). To evaluate
this loss the mean rate vtE ðjÞ exceeding a given rotation j during
the transition time tE was obtained according to Eq. (27), and it is
similar to the annual rate given by Eq. (26). However in the
particular case of vtE ðjÞ, the derivative of the probability
complement of Eq. (19) is substituted adequately in Eq. (26).
Z 1
dxðhÞ
ntE ðjÞ ¼  ð1  P½ZpjjhÞ½1 þ ðxðhÞt E  1Þ expðxðhÞt E Þ dh
0 dh
(28)
Finally the expected losses are obtained in equivalent form to
Eq. (3), as
Z jF
C D0 dnt0 ðyÞ
C D ð jD Þ ¼  zðjD ; yÞ dy (29)
x0ðtE Þ jD dy
On the other hand, in Fig. 10 it is observed that the losses
corresponding to instability failure do not contribute significantly
to the expected total loss and neither to the optimization. This is
0.15
Type of expected losses
Initial
Failure
Expected losses
0.10
Total
0.05
0.00
0 10 20 30 40 50
roll D, (degree)
Fig. 10. Expected losses vs. design roll parameter, for a given route of five days.

338 J.L. Alamilla et al. / Ocean Engineering 36 (2009) 330–338
0.15
10 days, DD = 0.95
0
5 days, DD = 0.95
0
5 days, DD = 0.80
0
Expected losses
5 days, DD = 0.70
0.10 0
0.05
0.00
0 10 20 30
roll ϕD ,(degree)
Fig. 11. Expected total losses vs. design roll parameter, for a given routes of five
and ten days. Also it is shown the influence of D0 over the total losses for a route of
five days.
due to the barge instability failure probability pF T ¼ 8  10 is
too small. Here it is assumed that the barge instability occurs for
rotations greater than 501. To evaluate the influence of jF over the
optimum design rotation, the total losses as a function of jD were
estimated for several values of jF between 401 and 801. These
total losses were practically identical for all jD values. This
behavior is due to the high stability of the current barges used
for offshore transportation and the metocean hazard of the
analyzed marine route. As discussed above, the initial losses
change with jD in exponential form, and contribute significantly
to the expected total loss. However the influence of these initial
losses is less than the expected losses associated to structural
damage. In Fig. 10, it can be seen that the structural damage losses
decrease abruptly and it is practically zero around 201 design
rotation, which is a consequence of metocean hazard along the
route. Fig. 11 shows the expected damage losses for several values
of DD0 and it can be seen that it does not influence significantly on
the minimum total losses. This value only has influence on the
losses associated to small jD.
In general the expected losses represent a inferior threshold
because of the spatial dependence between the possible extreme
wave occurrence in a site and the height wave in the site where
the barge is located. In other words the extreme wave occurrence
in a zone does not imply extreme occurrence in another site.
However these occurrences are correlated, decreasing with the
distance. Here the transition time is relatively short and for this
reason we expect that occurrence of events to be strongly
correlated and it implies that the losses and the design rotation
are similar to the obtained in this work.
As shown in Fig. 7, the probabilities of exceedance are similar.
Fig. 11 shows that the losses for 5- and 10-day routes are similar
because the hazard associated to these transition times are similar as
shown in Fig. 7. Therefore, for engineering practice the design
ARTICLE IN PRESS
parameters for 5- and 10-day routes are equal. Accordingly for an
optimum design rotation of j*D ¼ 231 based on Figs. 5 and 8, its
respective significant wave heights HS is 4 m, which corresponds
to 10% of exceedance probability.
4. Conclusions
The problem to select the design parameters for transportation
of offshore structures is solved with an optimum trade-off among
consequences expressed in terms of economical losses, metocean
hazard and barge-structure performance. In particular the design
barge rotation is optimized and it is linked to the return period
and its corresponding significant wave height. To our knowledge
there does not exist a model to optimize the design parameters for
marine transportation. The methodology was applied to a marine
route with transition time of 5 and 10 days. The selected design
40 50 parameters were identical in both cases and slightly minor to the
design parameters specified in current codes.
References
Aalbers, A.B., Leenaars, C.E.J., Quadvlieg, F.H.H.A., 1996. Voyage acceleration
climate: a comprehensive statistical method for the evaluation of design
5
loads for offshore transport. OTC 8123, Offshore Technology Conference. In:
Proceedings, 6–9 May.
API (American Petroleum Institute), 2000. Recommended Practice for Planning,
Designing and Constructing Fixed Offshore Platforms – Working Stress Design.
21st ed., API RP 2A-WSD.
Baxevani, A., Podgorski, K., Rychlik, I., 2000. Distribution of responses experienced
by a vessel. Comkiss Scientific Report, September.
Chakrabarti, S.K., 1994. Hydrodynamics of Offshore Structures. Computational
Mechanics Publication.
Cornell, C.A., 1968. Engineering seismic risk analysis. Bulletin of the Seismological
Society of America 58, 1583–1606.
Ditlevsen, O., Madsen H.O., 2007. Structural Reliability Methods. Internet Edition
2.3.7. /http://www.mek.dtu.dk/staff/od/books.htmS.
Esteva, L., 1968. Bases for the formulation of seismic design decisions. Research
Report 182, Institute of Engineering, UNAM. Mexico (in Spanish).
Esteva, L., 1976. Seismicity. In: Lomnitz, C., Rosenblueth, E. (Eds.), Seismic Risk and
Engineering Decisions. Elsevier, Amsterdam, pp. 179–224 (Chapter 6).
IMO (International Maritime Organization), 1998. Guidelines for Safe Ocean
Towing. MSC/Circ.884 21. December.
LOC (London Offshore Consultants, Inc.), 2003. Guidelines for Marine Operation-
Barge Transportation. LOCH/GUIDELINES/R002 Rev. 0. May.
Madsen, H.O., Krenk, S., Lind, N.C., 1986. Methods of Structural Safety. Prentice-
Hall, Englewood Cliffs, NJ.
NDI (Noble Denton International Ltd.), 2002. General Guidelines for Transportation
and Installation of Steel Jackets. Report No: 0028/NDI Rev. 1. May.
NDI (Noble Denton International Ltd.), 2005. General Guidelines for Marine
Transportations. Report No: 0030/NDI Rev. 2. April.
Oceanweather Inc., 2006. Update of metocean design data for Zona Norte and
Sonda de Campeche. Final Report Submitted to Mexican Institute of Petroleum,
July.
PEMEX (Petróleos Mexicanos), 2000. Design and Requalification of Fixed Marine
Platform in the Bay of Campeche. NRF-003-PEMEX-2000. Mexico (in Spanish).
PEMEX (Petróleos Mexicanos), 2003. Load, Moors, Transport and Installation of
Platforms in the Sea. NRF-041-PEMEX-2003. Mexico (in Spanish).
Rosenblueth, E., 1976. Optimum design for infrequent disturbances. Journal of
Structural Division, ASCE 102 (ST9), 1807–1825.
Rosenblueth, E., 1987. What should we do with structural reliabilities. In:
Proceedings of 5th International Conference on Applications of Statistics
and Probability in Civil Engineering. University of Waterloo, Waterloo, Canada,
pp. 24–34.
Stahl, B., 1986. Reliability Engineering and Risk Analysis. In: McClelland, B., Reifel,
M.D. (Eds.), Planning and Design of Fixed Offshore Platforms. Van Nostrand
Reinhold, New York, pp. 59–98 (Chapter 5).