## Are you sure?

This action might not be possible to undo. Are you sure you want to continue?

**Hull-girder reliability of new generation oil tankers
**

Josˇ ko Parunov

a,Ã

, Ivo Senjanovic´

a

, Carlos Guedes Soares

b

a

University of Zagreb, Faculty of Mechanical Engineering and Naval Architecture, Department of Naval

Architecture and Ocean Engineering, Ivana Lucˇic´a 5, 10000 Zagreb, Croatia

b

Unit of Marine Technology and Engineering, Technical University of Lisbon, Instituto Superior Tecnico,

Av. Rovisco Pais, 1049-001 Lisboa, Portugal

Received 1 June 2005; received in revised form 7 March 2007; accepted 20 March 2007

Abstract

A series of new generation oil tankers is presently under construction. These ships differ from

traditional oil tankers by their unusual form and therefore direct hydrodynamic analysis is used to

determine design vertical wave bending moments instead of adopting IACS rule values. The purpose

of the paper is to quantify changes in hull-girder reliability resulting from the new design features. To

achieve this, ﬁrst-order reliability analysis is carried out with respect to ultimate collapse bending

moment of the midship cross section of a new generation oil tanker and of a conventional ‘‘rule’’

designed oil tanker. The stochastic model of wave-induced bending moment is derived from direct

hydrodynamic analysis performed according to IACS Recommendation No. 34 Standard Wave

Data, Rev 1, 2000. The probability distribution of the still water bending moment is assumed based

on the data from loading manuals. The model uncertainties of linear wave loads, non-linearity of the

response as well as load combination factors are included in the reliability formulation. The

reliability analysis is performed for three relevant loading conditions: full load, ballast and partial

load and for two states of the hull: the ‘‘as-built’’ hull and ‘‘corroded’’ hull according to anticipated

20-year corrosion. One of the most interesting conclusions from the study is that the annual hull-

girder reliability of new generation oil tanker is increased considerably compared to the conventional

oil tanker. Sensitivity and parametric studies are performed with respect to random variables

representing modelling uncertainties. The results of a sensitivity study enable sorting of pertinent

variables according to their relative importance, while parametric study is used to quantify changes in

the reliability indices for moderate variation of input parameters. Furthermore, some other results

and discussions are presented pointing out the beneﬁts of introducing the ship reliability methods in

ARTICLE IN PRESS

www.elsevier.com/locate/marstruc

0951-8339/$ - see front matter r 2007 Elsevier Ltd. All rights reserved.

doi:10.1016/j.marstruc.2007.03.002

Ã

Corresponding author. Tel.: +385 1 6168 226; fax: +385 1 6156 940.

E-mail address: jparunov@fsb.hr (J. Parunov).

design practice, especially if this refers to new designs.

r 2007 Elsevier Ltd. All rights reserved.

Keywords: Tanker structures; Longitudinal strength; Structural reliability; Stochastic model of loads;

Uncertainties

1. Introduction

Oil tankers are being dimensioned on the basis of the rules of classiﬁcation societies.

Ship rules rely strongly on past experience and good engineering judgment, partially

supported by theoretical analyses and model tests. These types of structural design rules

are deterministic rules, where all the uncertainties of the pertinent design parameters are

compensated by a single safety factor [1]. Experience with new ships in service has shown

that the structures designed in accordance with those deterministic ship rules are fairly

safe, i.e. structural damages in new ships are rare, especially ultimate failures of the hull-

girder. It is also well known, however, that such rules are hardly applicable for the new

types of structures or new methods for load or strength evaluation [2].

Nowadays, classiﬁcation societies have been developing semi-probabilistic rules, based

on partial safety factors derived from the reliability formulations (PSFs) [3]. PSFs reﬂect

uncertainties in each of the pertinent variables instead of using only one global safety

factor. Ideally, PSFs should be calibrated in such a way as to produce more uniform safety

level when a large number of conventional ships is considered [1]. However, similarly to

their deterministic predecessors, rules based on PSFs are not readily applicable to

individual designs where there is not enough accumulated experience.

Typical examples of such innovative design without feedback from past experience are

oil product tankers of new generation, which are presently under construction. The new

generation oil tanker is a novel design, characterized by unusually low length-to-beam

ratio. One of the most interesting features of this ship design is that the design vertical

wave bending moments are calculated by direct hydrodynamic and statistical analysis. This

approach is different from common practice, where design values of this load component

are determined by IACS UR S11 formulae [4]. Since new generation oil product tanker

represents a new design and also the advanced method for load evaluation is employed,

obviously the present rules for the construction of ordinary merchant ships may not

provide good means to assess their safety. The main problem is the fact that the safety

margin between load and resistance in traditional ship rules is not explicitly quantiﬁed [2].

The reliability methods represent convenient means to quantify safety margin between

load and resistance in novel ship designs. These methods are intended as a tool that

rationally takes into account uncertainties in loading, response and structural strength.

The basic idea of reliability methods is to represent pertinent variables as random

variables, each with an associated distribution type and parameters. After that, using

suitable procedures, the probability of structural failure (or its counterpart, the safety

index) may be calculated.

Hull-girder reliabilities of the new generation oil tanker and the conventional double-

hull oil tanker are compared in the present paper. As the new generation oil tanker is

designed based on vertical wave bending moments from direct hydrodynamic analysis, it is

particularly interesting to assess the consequences of such approach on ship safety. This

ARTICLE IN PRESS

J. Parunov et al. / Marine Structures 20 (2007) 49–70 50

question is nowadays very important, since there is a tendency to introduce direct design

methods in ship structural design [5]. The inﬂuence of other structural features of new

generation oil tankers, as unusually large ship breadth, that are not evident by pure

deterministic analysis, may also be identiﬁed by reliability assessment.

In the present study, the reliability formulation and assumptions similar to those

originally developed within the SHIPREL project, have been considered [6]. The basis of

this reliability formulation is the ultimate collapse bending moment of the midship cross

section for one year of operation.

2. Description of analyzed ships

2.1. The new generation oil product tanker

The series of new generation oil tankers, ordered by one of the leading shipping

companies, are presently under construction. The main particulars of the ship are reported

in Table 1.

The speciﬁc characteristics of this tanker are lower draught and higher speed compared

to the traditional, medium-range oil tanker. The breadth of the new product tanker is

much larger than of ordinary tanker of similar size (40 m compared to 32.2 m) resulting in

30% increase of cargo capacity. Lower draught would enable easier approach to ports and

reduced risk of grounding, and this should be a step forward in the protection of the

environment. Another advantage of the new generation product tanker is that the ship is

equipped with two independent propulsion systems in two separate engine rooms, twin

propeller shafts and two rudders. Since the machinery system failure is one of the major

causes of tanker accidents, this tanker is obviously much safer than the conventional

design.

Since the hull form of the new generation product tanker deviates signiﬁcantly from the

traditional oil tanker, the rules for the construction of merchant steel ships may not be

directly applicable. Rather, both the ship owner and the Classiﬁcation Society require

some additional analyses based on the ﬁrst principles. One of the most important

additional analyses is the computation of design vertical wave bending moments by direct

hydrodynamic analysis.

The reason for this requirement is that the application of Rule formulae is limited to

ordinary hull forms with the ratio of ship length and beam higher than 5. From the main

particulars of the new product tanker, it appears that the ratio Lpp/B reads 4.4 being much

less than 5. Therefore, IACS UR S11 for ship longitudinal strength is not applicable. The

performed hydrodynamic and statistical analysis is described in detail by Parunov et al. [7].

ARTICLE IN PRESS

Table 1

Main characteristics of new generation oil product tanker

Length between perpendiculars Lpp 175.5 m

Moulded breadth B 40 m

Moulded depth D 17.9 m

Scantling draught T 13.0 m

Deadweight DWT 65200 dwt

J. Parunov et al. / Marine Structures 20 (2007) 49–70 51

The main conclusion of that study is that the design vertical wave bending moments

calculated by direct analysis are 20% higher than the corresponding values from IACS UR

S11. The principal difference between conventional tanker and new generation tanker is

that wave bending moments from direct calculations are adopted as design values for the

latter while the former is designed using rule values. Consequently, the hull structure of the

new generation tanker is more robust and safer than the structure of other tankers. The

concern of this study is to quantify this increase in robustness by means of reliability

analysis.

In the design check of the new generation tanker, conventional ship rules have been

employed without considering the ultimate bending capacity as a design criterion. Initially,

the midship section modulus at the strength deck is optimized to meet IACS UR S11 for

longitudinal strength. During the project development process, the section modulus was

increased in order to meet IACS UR S11 with the ‘‘rule’’ wave bending moment replaced

by the results of the direct hydrodynamic analysis. Therefore, the use of wave bending

moments from direct calculations in ship design is the only reason why the longitudinal

strength of this vessel is above the minimum rule requirements.

2.2. The ‘‘rule’’ oil tanker

The ‘‘rule’’ ship analyzed in the present study is an existing large double-hull tanker fully

satisfying the contemporary rules for design and construction of steel ships including IACS

UR S11. The particulars of the ‘‘rule’’ tanker are presented in Table 2.

Direct hydrodynamic analysis for determination of extreme vertical wave bending

moments is performed for this ship using the same assumptions as for the new generation

oil product tanker. Details of the analysis are presented in [7], where the overestimation of

the rule wave bending moments by results of direct analysis is found to be even 30%. The

overestimation of the rule wave bending moment is larger than in the case of new

generation tanker, indicating that the rule wave bending moments may not ensure

consistent safety level through different ships. This inconsistency is another motivation to

perform reliability analysis, as rational tool to deal with this problem.

3. Reliability formulation

When reliability methods are applied to assess structural safety, load effect components

and structural strength components are considered as random variables. Demand on the

structure and structural capacity are related through a mathematical expression, known as

the limit state equation. That expression deﬁnes if the structures fulﬁll their intended

ARTICLE IN PRESS

Table 2

Main characteristics of ‘‘rule’’ tanker

Length between perpendiculars Lpp 270 m

Moulded breadth B 48.2 m

Moulded depth D 23.0 m

Scantling draught T 17.1 m

Deadweight DWT 166300 dwt

J. Parunov et al. / Marine Structures 20 (2007) 49–70 52

purpose regarding a particular failure criterion or not. When considering longitudinal

strength of an oil tanker, it is usual practice to assume that ship hull behaves globally as a

beam under transverse load subjected to the still water and wave-induced load effects.

Among the load effects induced, the vertical bending moments at midship are of primary

importance. The ultimate bending moment is considered nowadays as the most realistic

limit state condition concerning longitudinal strength of the ship hull [3]. The ultimate

bending moment takes into account the load–deﬂection characteristics of the stiffened

panels composing the ship hull, including their post-buckling behavior.

The limit-state equation with respect to hull girder ultimate failure under vertical

bending moments, considered in the present study, reads:

^ w

u

M

u

À

^

M

sw

À c^ w

w

^ w

nl

^

M

w

o0, (1)

where M

u

is the deterministic ultimate hull-girder bending moment;

^

M

sw

the random

variable extreme still-water bending moment;

^

M

w

the random variable extreme vertical

wave bending moment; c the load combination factor between still water loads and wave

loads; ^ w

u

; ^ w

w

; ^ w

nl

the random variables representing modelling uncertainty of ultimate

strength, linear wave load and non-linearity of wave load.

The reliability analysis according to limit-state equation (1) is performed separately for

two independent failure modes—sagging and hogging. The hull-girder reliability in each of

the two failure modes is calculated for three elementary loading conditions—full load

condition (FL), ship in ballast (BL) and partial loading condition (PL). For rational

reliability assessment, the percentage of time that a ship spends in each of these loading

conditions has to be estimated. This estimate is made based on the statistical analysis of

load duration data for tankers performed by Guedes Soares [8], which is presented in

Table 3.

Annual safety indices b

FL

; b

PL

; b

BL

, and associated failure probabilities P

f ;FL

; P

f ;PL

;

P

f ;BL

are calculated for each of the elementary loading conditions separately for failures in

sagging and hogging. The global annual safety index for each of two failure modes b

t

is

then determined as

b

t

¼ ÀF

À1

ðP

f ;FL

þ P

f ;PL

þ P

f ;BL

Þ. (2)

Furthermore, the reliability assessment is performed for ‘‘as-built’’ ships as well as for

‘‘corroded’’ hulls, assuming 20-year corrosion according to ship rules [9]. The summing of

failure probabilities in Eq. (2) is possible if the assumption is adopted that the operating

conditions are statistically independent. However, even if the correlation of failures in

different loading conditions exists, Eq. (2) is still a good approximation if one of the three

loading cases dominates. It is well known that this is exactly what happens for oil tankers,

where full load dominates for failure in sagging and ballast in hogging.

ARTICLE IN PRESS

Table 3

Operational proﬁle adopted for tankers

Load cond. Harbour Full load Ballast load Partial load

Percentage of spent time 15% 35% 35% 15%

Voyage duration (days) 23.5 23.5 2.0

J. Parunov et al. / Marine Structures 20 (2007) 49–70 53

The failure probabilities are calculated using the PROBAN program, which is part of

the SESAM package for structural, hydrodynamic and reliability analysis [10]. In this

study, the probability of an event, given by limit state equation (1) is calculated using

FORM (ﬁrst order reliability method) implemented in PROBAN.

The sensitivity analysis is an important part of the reliability assessment since it enables

identiﬁcation of parameters that have the most important impact on safety indices. The

sensitivity factors used in this study are those deﬁned as

a

i

¼

@G

@y

i

y

Ã

i

ﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃ

P

i

@G

@y

i

2

y

Ã

i

s , (3)

where G(.) is the failure hyper plane in the reduced space of the standard normal variables

y

i

. Variables y

Ã

i

are the coordinates of the point in the reduced space closest to the origin.

After transformation of y

Ã

i

in original space, the most probable failure point x

Ã

i

is obtained.

The sensitivity factors calculated by PROBAN are further normalized to 100% sum.

4. Uncertainty model of ultimate vertical bending moment

The ultimate vertical bending moments of the example ships are calculated using

progressive collapse procedure proposed by Bureau Veritas [9]. Calculations are performed

for two states of the hull:

‘‘as-built’’ ship with ‘‘gross’’ thickness of structural elements, i.e. thickness of structural

elements as they are built-in in a new ship.

‘‘corroded’’ state of hull, which is the anticipated state after 20 years of ship service. For

the ‘‘corroded’’ state, the ultimate strength calculation is performed with ‘‘net’’

thickness, i.e. ‘‘gross’’ thickness reduced by the effect of corrosion. The thickness

reduction due to corrosion is taken according to BV rules [9], where corrosion reduction

of plates and longitudinals is between 1 and 2.5 mm, depending on the location of the

structural element. For comparison, the new common structural rules (CSR) [3] for oil

tankers propose corrosion reduction of plates and longitudinals for ultimate strength

calculation between 1.25 and 2 mm depending on the location of the structural element.

Therefore, the differences between the two approaches for corrosion deduction are

small with almost negligible inﬂuence on the ultimate bending moment capacity, which

is also conﬁrmed in practice. It is also worth noticing that applied CSR corrosion

deduction is not ‘‘full’’ corrosion for local strength assessment, but reduced corrosion

proposed by the Rules speciﬁcally for global strength assessment.

The results of the ultimate bending moment calculations of ‘‘as-built’’ and ‘‘corroded’’

hulls are presented in Tables 4 and 5, respectively.

The calculated moment is assumed to be the expected value of the ultimate bending

moment. Therefore, the mean value of the random variable ^ w

u

is taken equal to unity. This

assumption is based on calibration by experiments of identical method for ultimate

strength calculation [11]. In most recent reliability studies, ^ w

u

is assumed to follow a log-

normal distribution with COV ¼ 0.15, which is adopted also in the present study. This

ARTICLE IN PRESS

J. Parunov et al. / Marine Structures 20 (2007) 49–70 54

COV takes into account both the uncertainty in the yield strength and the model

uncertainty of the method to assess the ultimate capacity of the mid-ship section, since

both variables contribute to the ultimate moment. The coefﬁcient of variation of the yield

strength of the steel normally range from 8% to 10%, while the additional model

uncertainty is assessed by ‘‘engineering judgement’’, bringing the overall coefﬁcient of

variation to 0.15 [12].

5. Stochastic model of still-water bending moment

A Gaussian distribution is used as the stochastic model of still-water bending moment

for one voyage. The parameters of distribution were calculated from the loading manual,

according to the method proposed by Guedes Soares and Dogliani [13]. For each of the

elementary loading conditions (full load (FL), ballast (BL), and partial loading (PL)) the

mean values and standard deviations were calculated separately for departure and arrival

conditions. The mean values and standard deviations of the resulting normal distributions

were then obtained as average values for departure and arrival.

When the mean value m

sw

and the standard deviation s

sw

of the normal distribution are

known, the extreme value distribution for a given time period T

C

may be approximated

using Gumbel distribution with the following parameters [12]:

x

Ã

e

¼ F

À1

sw

1 À

1

n

sw

; a ¼

1 À F

sw

f

sw

, (4)

where n

sw

is the number of occurrences of a particular load condition in the reference

period T

C

(1 year). F

sw

is the cumulative probability distribution, F

À1

sw

its inverse while f

sw

is the probability density function of normal distribution with parameters m

sw

and s

sw

.

According to the operating scenario presented in Table 3, n

sw

¼ 5:4 for full load and

ballast, while n

sw

¼ 27:4 for the partial loading condition. The mean value m

se

and

standard deviation s

se

of the Gumbel distribution are then given as

m

e

¼ x

Ã

e

þ a Á 0:5772, (5)

ARTICLE IN PRESS

Table 5

Ultimate bending moments of ‘‘corroded’’ ships, MNm

Sag Hog

New generation tanker 3911 6075

‘‘Rule’’ tanker 11150 13118

Table 4

Ultimate bending moments of ‘‘as-built’’ ships, MNm

Sag Hog

New generation tanker 4812 7238

‘‘Rule’’ tanker 12702 14832

J. Parunov et al. / Marine Structures 20 (2007) 49–70 55

s

e

¼

p

ﬃﬃﬃ

6

p a. (6)

Table 6 presents the parameters of the stochastic model of the still-water bending

moment for one voyage and one year calculated according to the described procedure.

Surprisingly large differences in variability of still-water bending moments in ballast

loading condition for the two ships may be noticed in Table 6. Still-water bending

moments for the new generation tanker change considerably during voyages. Thus, for

ballast loading condition the still-water bending moment at departure reads 1780 MNm,

while at arrival its value is only 1300 MNm. The variability of still-water bending moment

for ‘‘rule’’ tanker is signiﬁcantly lower. For ballast condition, the SWBM at departure and

arrival reads 3771 and 3609 MNm, respectively. Large variability of SWBM in ballast for

double-hull tankers is documented also by Guedes Soares and Dogliani [13] where it is

noted that the SWBM may even change sign during the voyage. Obviously, the presented

data of SWBM may not be considered as generally representative for double-hull tankers,

since signiﬁcant differences between individual ships may be found. This conclusion

justiﬁes the methodology of determination of parameters of the SWBM statistical

distribution using individual ship’s loading manual rather than some general statistical

models.

Most of the previous reliability studies indicate that variability of SWBM is generally

not so important as the variability of wave loads and structural strength [14]. For that

reason, small model uncertainty of SWBM from loading manual compared to in-service

data is neglected in the present study.

6. Stochastic model of vertical wave bending moment

6.1. Random variable extreme vertical wave bending moment

^

M

w

For both ships, the evaluation of wave loading was carried out by means of linear strip

theory program WAVESHIP, a part of SESAM package [15]. The long-term distribution

of vertical wave bending moment is calculated by POSTRESP program, which is also part

of the SESAM package [16]. The POSTRESP uses standard long-term prediction

method that considers the total lifetime response history as a series of short-term episodes.

ARTICLE IN PRESS

Table 6

Parameters of still-water bending moment distributions, MNm (sagging SWBM is negative)

Ship L.C. One voyage One year

m

sw

s

sw

m

se

s

se

New generation tanker FL À615 310 À1016 276

PL 133 286 722 167

BL 1177 512 1841 455

‘‘Rule’’ tanker FL À1968 569 À2706 506

PL 384 957 2352 560

BL 3829 87 3943 78

J. Parunov et al. / Marine Structures 20 (2007) 49–70 56

Short-term responses are combined by a procedure that takes into account the relative

amount of exposure to the various levels of sea severity.

The computed long-term distribution is highly dependant on the assumptions used in the

analysis. The choice of wave scatter diagram containing the probabilities of occurrence of

various short-term sea states is the most important parameter in the long-term prediction

of extreme wave loads. The characteristic values of vertical wave bending moments from

the long-term distributions may differ up to 50% when different wave climate descriptions

are used [17]. Assumptions of heavy weather manoeuvering, such as ship speed reduction

and course changes, could also have large impact on tails of long-term distributions [18].

To standardize the procedure for computation of extreme wave loads, IACS has issued

Recommendation Note No. 34 as guidance for statistical analysis [19]. The basic

assumptions proposed by IACS for calculation of long-term distribution of wave bending

moments are:

The IACS North Atlantic scatter diagram should be used. This scatter diagram covers

areas 8, 9, 15 and 16, as deﬁned in Global Wave Statistics (GWS) [20]. The data from

the GWS are further modiﬁed by IACS in order to take into account the limited wave

steepness more properly [21].

Only ship speed equal to zero is to be taken into account.

The well-known two-parameter Pierson–Moskowitz spectrum (ITTC spectrum) is

recommended.

Short-crested waves with the wave energy spreading function proportional to cos

2

ðWÞ

have to be used.

All heading angles should have equal probability of occurrence and maximally 301

spacing between headings should be applied.

Proper corrections for non-linear effects are to be applied.

A Weibull 2-parameter model is usually used to approximate the long-term probability

distribution of vertical wave bending moment computed by the above described procedure:

Fðx

p

Þ ¼ 1 À e

À

xp

y

ð Þ

l

À Á

, (7)

where y and l are the Weibull scale parameter and the shape parameter, respectively. It is

well known that the Weibull distribution is an excellent approximation of the amplitude of

various ship responses in waves. Fðx

p

Þ in Eq. (7) represents the probability that the

amplitude of the response variable is less than x

p

in one randomly chosen cycle. The

probability that the response amplitude remains less than a given value x

e

over a longer

time period, e.g. 1 voyage, 1 year or 20 years, is given by the Gumbel law

Fðx

e

Þ ¼ e

Àe

À

x

e

Àx

Ã

e

a

À Á

(8)

whose parameters x

Ã

e

and a are derived from the parameters of the Weibull distribution (7)

by the following relationships:

a ¼

y

l

ðln nÞ

ð1ÀlÞ=l

, (9)

x

Ã

e

¼ yðln nÞ

1=l

. (10)

ARTICLE IN PRESS

J. Parunov et al. / Marine Structures 20 (2007) 49–70 57

Symbol n in (9) and (10) represents the number of response cycles in a given long-term

period, while x

Ã

e

calculated by (10) represents the most probable extreme value in n cycles.

The mean value and the standard deviation of the Gumbel distribution are given by

Eqs. (5) and (6), respectively. The Gumbel distribution obtained by this procedure is

actually the inherent uncertainty of extreme vertical wave bending moment, as represented

by random variable

^

M

w

in Eq. (1). The coefﬁcient of variation of this uncertainty is usually

in the range of 6–10%.

The long-term distributions of vertical wave bending moments calculated for three

loading conditions are shown in Figs. 1 and 2 for the new generation oil tanker and the

‘‘rule’’ tanker, respectively. The linear IACS UR S11 vertical wave bending moments are

also included in Figs. 1 and 2 as horizontal lines. Table 7 describes the stochastic models of

the vertical wave-induced bending moment for application in hull-girder reliability

assessment.

It is interesting to notice from Figs. 1 and 2 that the relative magnitudes of VWBM are

different for FL, BL and PL as a consequence of differences in hull form and mass

distribution. It is also obvious that for both ships the partial loading condition leads to the

largest VWBM comparing to other loading conditions. This phenomena may be explained

by the fact that partial loading conditions considered in the present study are those

resulting in the largest still-water bending moments and shear forces among all loading

conditions. Consequently, such loading conditions result in the largest wave-induced loads,

as well. It should be kept in mind, however, that other non-homogenous partial loading

conditions could result in different values of VWBM. Another aspect of the problem is the

sensitivity of VWBM to the draught-to-length ratio (T/L) [22]. Due to the Smith correction

factor, namely, the loading conditions with lower draught, as ballast conditions, could

have larger wave pressure exerted on the ship bottom that would tend to increase VWBM.

ARTICLE IN PRESS

0

500

1000

1500

2000

2500

1 2 4 5 6 8 3 7

-log 10 (Prob)

V

e

r

t

i

c

a

l

W

a

v

e

B

e

n

d

i

n

g

M

o

m

e

n

t

[

M

N

m

]

FL

BL

PL

IACS

Fig. 1. Long-term distributions of VWBM of new generation oil product tanker.

J. Parunov et al. / Marine Structures 20 (2007) 49–70 58

This is another reason why it is difﬁcult to draw general conclusions and explanations

about relative magnitudes of VWBM for different loading conditions, since different

details could inﬂuence the results.

Another important conclusion that may be derived from Figs. 2 and 3 is the large

exceedance of the IACS UR S11 vertical wave bending moments by results of

hydrodynamic analysis. For full load condition and exceeding probability of 10

À8

, the

results of long-term predictions overestimate the rule VWBM by 30% and 20% for ‘‘rule’’

tanker and ‘‘new generation tanker’’, respectively. These overestimations could be reduced

if 3D method was applied instead of strip theory since it has been conﬁrmed by several

comparative studies that strip theory leads to higher VWBM compared to 3D methods,

e.g. [23,24]. However, the uncertainty model of wave loads used in the present study is

ARTICLE IN PRESS

0

1000

2000

3000

4000

5000

6000

7000

8000

9000

1 2 4 5 6 3 7

-log 10 (Prob)

V

e

r

t

i

c

a

l

W

a

v

e

B

e

n

d

i

n

g

M

o

m

e

n

t

[

M

N

m

]

FL

BL

PL

IACS

8

Fig. 2. Long-term distributions of VWBM of ‘‘rule’’ oil tanker.

Table 7

Stochastic model of vertical wave-induced bending moment (MNm)

Ship L.C. Weibull parameters Gumbel moments (one year)

y l n m

e

s

e

New generation tanker FL 114.4 0.98 1.13 Á 10

6

1734 156.6

PL 117.7 0.97 4.85 Á 10

5

1730 166.6

BL 106.6 0.96 1.16 Á 10

6

1717 157.1

‘‘Rule’’ tanker FL 383.9 0.97 1.045 Á 10

6

6019 550.6

PL 382.7 0.96 4.40 Á 10

5

5790 568.9

BL 350.7 0.96 1.045 Á 10

6

5659 522.8

J. Parunov et al. / Marine Structures 20 (2007) 49–70 59

calibrated speciﬁcally for the strip theory, while different uncertainty models would be

more suitable for the 3D method. In this way, the results of reliability assessment are

actually independent on the choice of seakeeping method for wave load assessment.

6.2. Uncertainty of wave load evaluated under linear assumptions ^ w

w

Simpliﬁcations, assumptions and inaccuracies of the linear engineering models used to

predict extreme wave loads on ship hulls are taken into account by the modelling

uncertainty ^ w

w

, that appears in Eq. (1).

Direct consequences of the modelling uncertainty are large discrepancies in long-term

predictions of extreme wave bending moments performed by various researchers or

institutions for the same ship, i.e. the same physical reality. The obvious example of these

discrepancies is the comparative study of long-term distributions of vertical wave bending

moments performed by leading classiﬁcation societies for the containership with no

forward speed [4]. The maximum and minimum predicted lifetime vertical wave bending

moments differed by as much as 80%. These differences could have a signiﬁcant impact on

ship design practice, since direct calculation methods are obviously not sufﬁciently

standardized to assure a consistent level of ship structural integrity.

For the present study, ^ w

w

is assumed to be a normally distributed random variable with

the mean value equal to 0.9 and coefﬁcient of variation equal to 0.15. This statistical model

is mainly based on the benchmarking study for the program WAVESHIP [25]. In that

study, the modelling error parameter f, deﬁned as the ratio of the lifetime bending

moment based on the measured transfer functions and the lifetime bending moment based

on the calculated transfer functions, is determined for a number of ships and shown in

Fig. 3. The modelling error may be simulated by a normal distribution with the mean value

ARTICLE IN PRESS

0

1

2

3

4

5

6

7

1.5 0.9 0.8 0.7 0.6 0.5 1.4 1.3 1.2 1 1.1

modelling error

n

u

m

b

e

r

o

f

o

b

s

e

r

v

a

t

i

o

n

s

Experimental data

Normal distribution

Fig. 3. Variability of the modelling error parameter.

J. Parunov et al. / Marine Structures 20 (2007) 49–70 60

of 0.9 and the coefﬁcient of variation of 10%. The coefﬁcient of variation of ^ w

w

is ﬁnally

increased to 0.15, accounting for other uncertainties of linear engineering models.

6.3. Uncertainty in non-linear effects ^ w

nl

The most important source of non-linear behavior in wave loads are the non-vertical

ship sides. The main consequence of this non-linearity is the difference between sagging

and hogging vertical wave bending moments. Using linear theory, it is not possible to

distinguish the sagging and hogging bending moments since only a single linear bending

moment is obtained. Usually, the sagging bending moment is higher, while the hogging

bending moment is lower than the linear prediction. The non-linear hydrodynamic

methods, such as quadratic strip theory, are capable of predicting differences in sagging

and hogging bending moments [26]. Despite the existence of such methods, the linear strip

theory is still the most popular tool for seakeeping computations. Non-linearity, especially

the difference in sagging and hogging wave bending moments is approximately taken into

account by introducing non-linear correction factors F

S

and F

H

for sagging and hogging,

respectively. In that way, the non-linear bending moments in sagging M

S

and hogging M

H

are obtained as M

S

¼ F

S

M

L

and M

H

¼ F

H

M

L

, where M

L

is the vertical wave bending

moment computed by linear analysis. In this work, the non-linear correction is assumed to

be the same as in the IACS UR S11, which means that the ratio between the wave bending

moments in sagging and hogging depends only on the block coefﬁcient C

b

. It is further

assumed that the linear value of vertical wave bending moment is the mean value between

sagging and hogging wave bending moments. By applying these assumptions, one obtains

the correction factors of wave bending moments in sagging and hogging as

F

S

¼

M

S

M

L

¼

2R

1 þ R

; F

H

¼

M

H

M

L

¼

2

1 þ R

, (11)

where R represents the ratio of the vertical wave bending moments in sagging and hogging

from IACS UR S11:

M

S

M

H

¼ R ¼

C

B

þ 0:7

1:73 Á C

B

. (12)

For the reliability assessment in the present study, the uncertainty of non-linear effects

^ w

nl

is assumed to be a normally distributed variable with mean value equal to non-linear

correction factors calculated by Eqs. (11) and (12). The coefﬁcient of variation of this

uncertainty is assumed to be 0.15 [12,14].

7. Combination between still water and wave bending moment

The load combination factor c that appears in Eq. (1) is used to account for the fact that

the extreme values of wave loads and still-water loads do not occur at the same time.

Accordingly, the total extreme vertical moment must be somewhat reduced. The load

combination factors are usually calculated using the Ferry-Borges and Castanheta load

combination method based on an assumed operating scenario [27]. In the present analysis,

the load combination factors used for the reliability assessment of the oil tankers were

those calculated by Guedes Soares and Teixeira [12]. These load combination factors are

given in Table 8.

ARTICLE IN PRESS

J. Parunov et al. / Marine Structures 20 (2007) 49–70 61

It should be noted that these values do not change signiﬁcantly for different amplitudes

of the load variables within the same operational proﬁle and outcomes of various load

combination studies for oil tankers lead to similar results [28].

8. Results of reliability analyses

The annual safety indices are calculated by the PROBAN program using the FORM

method [10]. Safety indices b are calculated for two different ship states (‘‘as-built’’ ship

and ‘‘corroded’’ ship according to 20-year corrosion from ship rules), for three different

loading conditions (full load, ballast and partial load) and for two separate failure modes

(hogging and sagging). The summary of the stochastic model adopted is shown in Table 9

where the notation adopted in Eq. (1) is used.

8.1. Results of reliability analyses of ‘‘as-built’’ ships

Results of the reliability calculations for ‘‘as-built’’ state of hulls are presented in Table 10

together with the associated global annual failure probabilities P

ft

calculated by Eq. (2).

As may be seen from Table 10, the full load condition dominates for the sagging failure

mode for both ships. For the hogging failure mode, ballast is the most important loading

condition. Partial loading condition is of less importance, although it may generate some

extreme wave and still-water load effects, as elaborated in 6.1. The reason is that tankers

generally spend less time in partial loading conditions compared to ballast and full load.

Consequently, the number of cycles of wave loads and the load combination factors are

reduced for partial loading condition. Such considerations are not used by conventional

deterministic approach, which points out the advantage of using the reliability methods.

From Table 10, one may clearly see the beneﬁts of using direct hydrodynamic methods

in design of an oil tanker. The hull girder reliability of a new generation tanker is

signiﬁcantly higher than the reliability of the ‘‘rule’’ tanker for both sagging and hogging

failure modes.

Although the calculated reliability index of 2.25 for ‘‘as-built’’ ‘‘rule’’ tanker in sagging

looks rather low, other researchers obtained similar results. For example, the average

annual reliability index of three ‘‘as-built’’ large tankers in sagging reads 2.21 according to

the analysis presented by Teixeira and Guedes Soares [28]. The same authors calculated the

average value of reliability indices for ‘‘as-built’’ bulk-carriers in sagging as 2.13 [12]. This

leads to the conclusion that the reliability indices calculated in the present study are

comparable to other similar analyses.

ARTICLE IN PRESS

Table 8

Load combination factors

Load condition c

FL 0.92

BL 0.91

PL 0.80

J. Parunov et al. / Marine Structures 20 (2007) 49–70 62

8.2. Results of reliability analyses of ‘‘corroded’’ ships

The results of the reliability calculations for ‘‘corroded’’ hulls are presented in Table 11

together with the associated global annual failure probabilities P

ft

calculated by Eq. (2).

ARTICLE IN PRESS

Table 9

Summary of stochastic model adopted (sagging SWBM is negative)

Variable Distribution Mean COV

^

M

sw

(MNm)

Gumbel New generation tanker FL À1016 0.27

PL 722 0.23

BL 1841 0.25

‘‘Rule’’ tanker FL À2706 0.19

PL 2352 0.24

BL 3943 0.02

^

M

w

(MNm)

Gumbel New generation tanker FL 1734 0.09

PL 1730 0.10

BL 1717 0.09

‘‘Rule’’ tanker FL 6019 0.09

PL 5790 0.10

BL 5659 0.09

M

u

(MNm) Deterministic New generation tanker ‘‘As-built’’ Sag 4812

Hog 7238

‘‘Corroded’’ Sag 3911

Hog 6075

‘‘Rule’’ tanker ‘‘As-built’’ Sag 12702

Hog 14832

‘‘Corroded’’ Sag 11150

Hog 13118

^ w

w

Gaussian 0.9 0.15

^ w

nl

Gaussian Sag 1.03 0.15

Hog 0.97 0.15

^ w

u

Log-normal 1 0.15

c Deterministic FL 0.92

PL 0.80

BL 0.91

Table 10

Safety indices in sagging and hogging for ‘‘as-built’’ ships

Ship Cnd. b

FL

b

BL

b

PL

P

ft

b

t

New generation tanker Sag 2.97 6.01 5.63 1.49E–03 2.97

Hog 6.89 3.50 5.90 2.33E–04 3.50

‘‘Rule’’ tanker Sag 2.25 5.11 4.86 1.22E–02 2.25

Hog 5.10 2.88 3.81 2.06E–03 2.87

J. Parunov et al. / Marine Structures 20 (2007) 49–70 63

Some interesting conclusions may be drawn from the results for ‘‘corroded’’ hulls that

are presented in Table 11. The annual failure probability of ‘‘corroded’’ ‘‘rule’’ tanker in

sagging is about 2.5 times higher than the corresponding value of the new generation

tanker. It may be noted that the ratio of failure probabilities of ‘‘rule’’ tanker and new

generation tanker is much reduced compared to ‘‘as-built’’ ships, where this ratio reads

more than 8. The relatively large breadth of the new generation tanker may explain this

phenomenon, since the main deck has signiﬁcant impact on the strength characteristics of a

hull-girder. When such large deck is corroded, the hull-girder properties are rapidly

reduced as well as the reliability index with respect to the ultimate bending moment. These

results emphasize the importance of the corrosion-protection system and its inﬂuence on

the safety of new generation tanker.

Despite the effect described above, it may be noted by comparing Tables 10 and 11, that

the reliability of ‘‘corroded’’ new generation tanker in sagging ðb

t

¼ 2:06Þ is only slightly

lower than the reliability of ‘‘as-built’’ ‘‘rule’’ double-hull tanker ðb

t

¼ 2:25Þ. This is also

one illustrative example how, within the framework of reliability-based methods, relative

increase of safety of new generation oil tanker may be described compared to the

traditional design.

9. Sensitivity analysis and parametric study

9.1. Sensitivity analysis

Sensitivity analyses are performed for cases that resulted in the lowest safety indices, i.e.

full load in sagging and ballast load in hogging for ‘‘corroded’’ hulls. The sensitivity factors

a

i

are calculated by PROBAN according to Eq. (3) and further normalized to 100% sum.

These normalized sensitivity factors are presented in Tables 12 and 13. In the same tables,

the coordinates of the design point x

Ã

i

, representing the most probable combination of

random variables in the case of failure are also included.

By analyzing the results in Tables 12 and 13, it is obvious that the uncertainty of

ultimate longitudinal strength is generally the most important random variable. However,

for the new generation oil tanker in ballast condition, the importance of the vertical still-

water bending moment overcomes that of the ultimate longitudinal strength. It is also

obvious that for new generation oil tanker the importance of the wave-induced load

variables (

^

M

w

, ^ w

w

and ^ w

nl

) decreases from full load to ballast. Similar conclusions were also

obtained by Guedes Soares and Teixeira [12]. For the ‘‘rule’’ oil tanker, however, the same

conclusions are not valid as the importance of the still-water bending moment is quite

ARTICLE IN PRESS

Table 11

Safety indices in sagging and hogging for ‘‘corroded’’ ships (20-year corrosion from ship rules)

Ship Cnd. b

FL

b

BL

b

PL

P

ft

b

t

New generation tanker Sag 2.06 5.46 4.99 1.97E–02 2.06

Hog 6.34 2.81 5.17 2.48E–03 2.81

‘‘Rule’’ tanker Sag 1.65 4.76 4.46 4.95E–02 1.65

Hog 4.72 2.26 3.27 1.25E–02 2.24

J. Parunov et al. / Marine Structures 20 (2007) 49–70 64

marginal for both failure modes. Also, the importance of wave-induced load variables

decreases only slightly from full load to ballast compared to the new generation tanker

case.

This example is another evidence of the advantage of reliability methods compared to

deterministic or even semi-probabilistic methods. Partial safety factors in semi-

probabilistic rules should reﬂect the relative importance of pertinent variables. However,

it is obvious from the presented analysis that the relative importance of wave-induced and

still-water loads may vary considerably between individual ships of the same type.

Therefore, only direct reliability analysis performed for each individual design may provide

sufﬁciently accurate information about the importance of the uncertainty of the pertinent

variables.

9.2. Parametric study

To estimate the mean values and COVs of modelling uncertainties in a rational manner,

large amount of collected full-scale measurement data from actual ships would be

required. Since this would be quite an expensive and time-consuming request, the

researchers are faced with only a limited quantity of available data. Consequently, the

parameters of modelling uncertainties ^ w

w

, ^ w

nl

and ^ w

u

speciﬁed in Table 9 represent ‘‘the best

estimates’’ that might be changed if more data for comparison of full-scale measurements

and numerical computations were collected. To explore the implications of these potential

variations of parameters on the calculated safety indices, a parametric study is performed.

This is done in such a way that COVs and the mean values of modelling uncertainties are

varied within reasonable range of their values. As in the sensitivity analysis, the parametric

study is performed for ships in ‘‘corroded’’ states, for the sagging failure mode in full load

condition and for the hogging failure mode in ballast condition. Only one of the

ARTICLE IN PRESS

Table 13

Sensitivity factors and coordinates of design points for ballast load in hogging failure mode (‘‘corroded’’ hulls)

Ship ^ w

u

^ w

w

^ w

nl

^

M

w

^

M

sw

New generation tanker a

i

; % 35.1 3.6 3.6 1.6 56.0

x

Ã

i

0.77 0.97 1.05 1745 3066

‘‘Rule’’ tanker a

i

; % 57.0 16.5 16.5 9.9 0.1

x

Ã

i

0.77 1.02 1.10 5954 3936

Table 12

Sensitivity factors and coordinates of design points for full load in sagging failure mode (‘‘corroded’’ hulls)

Ship ^ w

u

^ w

w

^ w

nl

^

M

w

^

M

sw

New generation tanker a

i

; % 43.2 12.5 12.5 6.1 25.7

x

Ã

i

0.81 1.00 1.14 1786 1285

‘‘Rule’’ tanker a

i

; % 46.6 18.6 18.6 9.3 6.8

x

Ã

i

0.84 0.99 1.14 6202 2835

J. Parunov et al. / Marine Structures 20 (2007) 49–70 65

parameters is varied in each reliability analysis, while all the others retain their values as

speciﬁed in Table 9.

9.2.1. Variation of COVs

‘‘The best estimate’’ of COVs of ^ w

w

, ^ w

nl

and ^ w

u

reads 0.15 according to Table 9. In the

parametric study, the safety indices are calculated also for COVs of 0.1 and 0.2. The results

of the parametric study are presented in Tables 14 and 15 for the new generation tanker

and the ‘‘rule’’ tanker, respectively.

As may be seen from Tables 14 and 15, the COV of ^ w

u

has the most important inﬂuence

on the safety indices in all cases. The inﬂuence of COV of wave load uncertainties ^ w

w

and

^ w

nl

is much less important. Furthermore, it may be noted that the same value of safety

index is obtained if COV of ^ w

w

or ^ w

nl

are varied for the same amount. These results are fully

consistent with the sensitivity factors speciﬁed in Tables 11 and 12: the sensitivity factor for

^ w

u

is the largest in all cases, while the sensitivity factors of ^ w

w

and ^ w

nl

are exactly the same

for both ships in all loading conditions and failure modes.

9.2.2. Variation of mean values

The parametric study of the mean values is performed using the following mean values

of modelling uncertainties:

mean value of ^ w

u

between 0.95 and 1.05 (‘‘the best estimate’’ in Table 9 is 1.0);

mean value of ^ w

w

between 0.85 and 0.95 (‘‘the best estimate’’ in Table 9 is 0.9);

mean value of ^ w

nl

equal to 1.14 in sagging and 1.12 in hogging. (‘‘the best estimates’’ in

Table 9 are 1.03 and 0.97 in sagging and hogging, respectively).

While the means of ^ w

u

and ^ w

w

are small variations of their ‘‘best estimates’’, the mean

values of non-linear correction factors ^ w

nl

are signiﬁcantly larger for both sagging and

ARTICLE IN PRESS

Table 14

Safety indices for various coefﬁcients of variation of modelling uncertainties (New generation tanker, ‘‘corroded’’

hull)

COV ð^ w

u

Þ COV ð^ w

w

Þ COV ð^ w

nl

Þ

0.1 0.2 0.1 0.2 0.1 0.2

b

FL

(Sag) 2.39 1.75 2.14 1.97 2.14 1.97

b

BL

(Hog) 3.15 2.46 2.84 2.77 2.84 2.77

Table 15

Safety indices for various coefﬁcients of variation of modelling uncertainties (‘‘Rule’’ tanker, ‘‘corroded’’ hull)

COV ð^ w

u

Þ COV ð^ w

w

Þ COV ð^ w

nl

Þ

0.1 0.2 0.1 0.2 0.1 0.2

b

FL

(Sag) 1.95 1.39 1.75 1.55 1.75 1.55

b

BL

(Hog) 2.75 1.85 2.38 2.13 2.38 2.13

J. Parunov et al. / Marine Structures 20 (2007) 49–70 66

hogging failure modes. This large increase of non-linear correction factors is based on

some recent studies showing that even full form tankers of medium size may have

signiﬁcant non-linear contributions to the total vertical wave bending moments [5]. This is

explained by the simultaneous effect of ship length and large bulbous bow. In such case,

both hogging and sagging vertical wave bending moments may be signiﬁcantly higher than

the linear prediction. This is contrary to the IACS UR S11, where difference between

sagging and hogging is almost negligible for full-form ships with large block coefﬁcients, as

oil tankers. For this reason, large variation of mean value of non-linear effects is included

in the parametric analysis. The results of the analysis are presented in Tables 16 and 17 for

new generation tanker and rule tanker, respectively.

From Tables 16 and 17 one may conclude that the variation of mean of non-linear

effects mð^ w

nl

Þ has the largest impact on the calculated safety indices. This is expected as the

range of variation of this parameter is the largest. However, it is interesting to notice that

in most cases relatively similar results are obtained with variation of the mean value of the

ultimate strength mð^ w

u

Þ, although its range of variability is much lower. This is another

evidence of the large sensitivity of the calculated reliability indices to the modelling

uncertainty of ultimate strength.

The results of the parametric study for COVs and mean values are summarized in Fig. 4.

It may be concluded from Fig. 4 that the largest variability has the safety index of the

‘‘rule’’ tanker in hogging, that varies +0.49/À0.45 around the value calculated with ‘‘the

best estimates’’ of parameters. The variability of other safety indices is somewhat lower.

Thus, the variability in sagging reads +0.33/À0.31 for the new generation tanker and

+0.3/À0.32 for the ‘‘rule’’ tanker. Obviously, the variability of safety indices is not

negligible and they emphasize the importance of proper uncertainty assessment of

pertinent variables for a credibility of a reliability study.

ARTICLE IN PRESS

Table 16

Safety indices for various mean values of modelling uncertainties (New generation tanker, ‘‘corroded’’ hull)

mð^ w

u

Þ mð^ w

w

Þ mð^ w

nl

Þ

0.95 1.05 0.85 0.95 1.12 1.14

b

FL

(Sag) 1.83 2.28 2.21 1.92 – 1.79

b

BL

(Hog) 2.61 3.01 2.89 2.73 2.60 –

Table 17

Safety indices for various mean values of modelling uncertainties (‘‘Rule’’ tanker, ‘‘corroded’’ hull)

mð^ w

u

Þ mð^ w

w

Þ mð^ w

nl

Þ

0.95 1.05 0.85 0.95 1.12 1.14

b

FL

(Sag) 1.42 1.88 1.83 1.48 – 1.33

b

BL

(Hog) 2.00 2.51 2.44 2.09 1.81 –

J. Parunov et al. / Marine Structures 20 (2007) 49–70 67

The results of the presented parametric study might be useful when comparing the

reliability analyses based on differing assumptions. The ranges of variability of the safety

indices calculated in the present work may be used to approximately estimate the

implications of different parameters of modelling uncertainties on the results of reliability

studies.

10. Conclusion

The new generation oil tanker, which is characterized by shallow draught and low

length-to-beam ratio, is designed based on the vertical wave bending moments obtained by

direct hydrodynamic and statistical analysis instead of using rule values. The aim of this

paper is to quantify the changes in notional hull-girder reliability as a result of such

innovative approach in ship design. For that purpose, a comparative hull-girder reliability

study of a new generation oil tanker and an existing conventional ‘‘rule’’ tanker is carried

out. The reliability assessment is performed for ‘‘as-built’’ state of ships and for

‘‘corroded’’ ships, assuming 20-year corrosion, as given by ship rules.

Results of the analysis of ‘‘as-built’’ ships show that the structural reliability of the new

generation oil tanker is much higher than the reliability of the ‘‘rule’’ oil tanker. For

‘‘corroded’’ hulls, the ratio of failure probabilities between the ‘‘rule’’ tanker and the

new tanker design is reduced, as a consequence of the unusually large breadth of the new

generation oil tanker. Nevertheless, the reliability of the ‘‘corroded’’ tanker of new

generation is almost as high as the reliability of the ‘‘as-built’’ ‘‘rule’’ tanker.

The sensitivity analysis shows that for various ships, even of the same type, the relative

impact on structural safety of still-water and wave-induced bending moments may be quite

different. While for the new generation tanker the calculated reliability index is very

sensitive to the still-water bending moment in ballast loading condition, the reliability of

the ‘‘rule’’ tanker is almost completely insensitive to this variable. The sensitivity analysis

ARTICLE IN PRESS

0

0.5

1

1.5

2

2.5

3

3.5

R

e

l

i

a

b

i

l

i

t

y

i

n

d

e

x

New generation tanker -sag

"Rule" tanker - sag

New generation tanker - hog

"Rule" tanker - hog

Fig. 4. Ranges of reliability indices calculated by parametric variation.

J. Parunov et al. / Marine Structures 20 (2007) 49–70 68

shows that the uncertainty of the ultimate bending moment capacity is generally the most

important variable in reliability assessment.

The parametric variation is performed for plausible ranges of COVs and mean values of

the modelling uncertainties. The results of the parametric study conﬁrm the results of the

sensitivity analysis regarding the huge importance of the uncertainty of ultimate strength.

It is also shown that different assumptions about non-linear effects may lead to signiﬁcant

changes in the calculated safety indices. The results of the presented parametric study

might be useful for comparison of reliability analyses that are based on different

parameters of modelling uncertainties.

References

[1] Mansour A, Wirsching P, White G, Ayyub B. Probability based ship design: implementation of design

guidelines. Ship Structure Committee. SSC-392, 1996.

[2] Faulkner D. An independent assessment of the sinking of the MV DERBYSHIRE. Trans SNAME

1998;106:59–103.

[3] ABS, DNV, Lloyd’s Register. Common Structural Rules for Double Hull Oil Tankers, 2006.

[4] Nitta A, et al. Basis of IACS uniﬁed longitudinal strength standard. Mar Struct 1992;5:1–21.

[5] Guedes Soares C. On the uncertainty in long-term predictions of wave induced loads on ships. Mar Struct

1999;12:171–82.

[6] Guedes Soares C, Dogliani M, Ostergaard C, Parmentier G, Pedersen PT. Reliability based ship structural

design. Trans SNAME 1996;104:357–89.

[7] Parunov J, Senjanovic´ I, Pavic´ evic´ M. Use of vertical wave bending moments from hydrodynamic analysis in

design of oil tankers. Trans RINA 2004;146:247–60.

[8] Guedes Soares C. Inﬂuence of human control on the probability distribution of maximum still-water load

effects in ships. Mar Struct 1990;3:319–39.

[9] Bureau Veritas. Rules for Classiﬁcation of Steel Ships 2003.

[10] Det Norske Veritas. PROBAN—General Purpose Probabilistic Analysis Program. SESAM User’s Manual:

Høvik, 2002.

[11] Gordo JM, Guedes Soares C. Approximate assessment of the ultimate longitudinal strength of the hull.

J Ship Res 1996;40(1):60–9.

[12] Guedes Soares C, Teixeira AP. Structural reliability of two bulk carrier designs. Mar Struct 2000;13:107–28.

[13] Guedes Soares C, Dogliani M. Probabilistic modelling of time-varying still-water load effects in tankers. Mar

Struct 2000;13:129–43.

[14] Casella G, Rizzuto E. Second-level reliability analysis of a double-hull oil tanker. Mar Struct 1998;11:373–99.

[15] Det Norske Veritas. WAVESHIP—wave loads on slender vessels. SESAM User’s Manual, Høvik, 1993.

[16] Det Norske Veritas. POSTRESP—interactive postprocessor for general response analysis. SESAM User’s

Manual, Høvik, 1993.

[17] Guedes Soares C, Trova˜ o MFS. Inﬂuence of wave climate modelling on the long term prediction of wave

induced responses of ship structures. In: Dynamics of marine vehicles and structures in waves. Amsterdam:

Elsevier Science Publishers; 1991. p. 1–10.

[18] Guedes Soares C. Effect of heavy weather maneuvering on the wave induced vertical bending moments in

ship structures. J Ship Res 1990;34(1):60–8.

[19] IACS Recommendation No. 34. Standard Wave Data. Rev 1, 2000.

[20] Hogben N, Dacunha NMC, Olliver GF. Global wave statistics. Feltham: British Maritime Technology Ltd.;

1986.

[21] Bitner-Gregersen EM, Cramer EH, Løseth R. Uncertainties of load characteristics and fatigue damage of

ship structures. Mar Struct 1995;8:97–117.

[22] Jensen JJ, Mansour AE. Estimation of ship long-term wave-induced bending moment using closed-form

expressions. Trans RINA 2002;144:41–55.

[23] Guedes Soares C. On the deﬁnition of rule requirements for wave induced vertical bending moments. Mar

Struct 1996;9:409–25.

[24] McTaggart M, et al. Motions and loads of a hydroelastic frigate model in severe seas. Trans SNAME

1997;105:427–54.

ARTICLE IN PRESS

J. Parunov et al. / Marine Structures 20 (2007) 49–70 69

[25] Parunov J, Senjanovic´ I. Incorporating model uncertainty in ship reliability analysis. Trans SNAME

2003;111:377–408.

[26] Jensen JJ. Load and global response of ships. Oxford: Elsevier Science Ltd; 2001.

[27] Guedes Soares C. Combination of primary load effects in ship structures. Probabilistic Engineering

Mechanics 1992;7:103–11.

[28] Teixeira AP, Guedes Soares C. Assessment of partial safety factors for tankers. In: Guedes Soares, Garbatov,

Fonseca, editors. Maritime transportation and exploitation of ocean and coastal resources, vol. II. London:

Taylor & Francis; 2005. p. 1601–11.

ARTICLE IN PRESS

J. Parunov et al. / Marine Structures 20 (2007) 49–70 70

- 7898_EL ASQ Certification1
- Sensitivity and Uncertainty Analysis Coupled With Automatic Calibration for a Distributed Watershed Model
- analisis multicriterio
- IT Manager
- GSA_01
- 00-152 Alpaca Enterprise
- Project Funding (1)
- 2SC5387
- TAT Initiatives
- Economic Modeling for Petroleum Projects
- 2009-04
- Basic Reliability Formulation Handbook2
- 6442_chap01
- ETCE_779_LECTURE_17.pdf
- Tugas 4_Comparing Alternative and Dealing with Uncertainty Ver 0.pptx
- Colloquium Presentation MIB
- Impact of Emergency Shutdown Devices on Relief System Sizing and Design
- folk 2009
- durability design.pdf
- TS-BSC-SW-0745-I9.0
- Sensitivity Analysis of a Dynamic Fleet Management Model Using Approximate Dynamic Programming
- Cromey Etal Aqua2002
- ANSYS Advantage Oil and Gas Special Issue 2014
- 2SK3320_datasheet_en_20140301.pdf
- When Maintenance is Not Maintenance
- 11312_11_Progressing Cavity Pumps, Downhole Pumps and Mudmotors
- Fleet Strategy - TL Foundations
- howtoimplement5s-160107121405
- Dossier 2005 Argile - Tome Safety evaluation of a geological repository
- Assessing Eigen Value Sensitivities

- 9fcfd50b78a2169366.pdf
- ANSYS Conference 2008 Paper- SeaTech.pdf
- 3deec51cdd342cb593.pdf
- Atmospheric Storage Tanks Venting Req API 2000 (6th ED 2009)
- RakMek_43_3_2010_3.pdf
- 3.Yankee Doodle
- ultimate strength
- [Scientific Proceedings Faculty of Mechanical Engineering] Preliminary Computational Fluid Dynamics (CFD) Simulation of EIIB Push Barge in Shallow Water
- Draft Survey Guidelines 2010.1
- SMK3522_Wk06.pdf
- Tanker Technology Limitations Double Hulls
- A749(18)E.pdf
- MSN17
- Ultimate strength of a square plate
- 1 REVIEW of TANKER SAFETY Papanikolaou Eliopoulou Hamann to Samiotis Goulielmos
- 04 MARIN Challenging Wind and Waves (1)
- TM_55-1905-223-SDC
- Ada 294491
- AuxHullInspector
- MARIN Challenging Wind and Waves (1)
- Maestro Tutorial
- Capability Description
- h 0545567
- ultimate strength

Close Dialog## Are you sure?

This action might not be possible to undo. Are you sure you want to continue?

Loading