SSC-301

PROBABILISTIC STRUCTURAL
ANALYSIS OF SHIP HULL
LONGITUDINAL
STRENGTH

Thisdocument
hasbeenapproved
forpublic
release
endsalqits
distribution
isunlimited.

SHIP STRUCTURE
1981

COMMITTEE

SHIP STRUCTLTW COWWITTEE

l%. SHIP STRLVCUSE COILMITTEE

is constit.ced co prosecute a research
p..gram co imPr.ve che hU1l .tr..t.re, of ship. ..d .ch= =.fne .C~U.tUres by
an extension of knowledge pertaining to design, materials and methods of
Con,trctio.
R4DM H. H. BELL (Chaizwwd
Chicf, Office af !Yezdcmt
Marine Safety
Us. Coost Gluzld

,Nr.J. GROSS
~P.*Y As.ist@
Administrator
,orConrre
retal Development
,YuritinEAdministration

<Mr.P. ,v.PALER)40
Deputy Director, HulL Gmlq
,Nava
1 Sea SyaternsCornnmd

MF. P. McDONALD
Chief, Branch of Offshore
Field @ezntimts
u.S. Geological Survey

Mr. W. N. RANNAN
Vice ?rwsident
Axetican Bmati of Ship@ng

Mr. C. J. WRI!O?S!WI!
Engi*er Ofploer
,% lit~
Sealift Conrncmd

CUR T. L

ROEW50W,

CI.
s. Coast Gumi

(Secx?tm )

SKIP STICUCTURS.SUECONMIHEE
The SHIP STSUCTURS SUSCOIU-IITTEE
acts for the Ship Structure Committee
m technical matters by providio~ techr.ical.o.rdir,atioofor the determination
of goals and objectives of the program, and by evaluating and interpreting
che resul,w in terms of structural desi~, construciton and qeration .
U.S. CUAST GUARU

MILITARY SEALI~

CAST R. L. BRO&7i
CDR J. C CARD
CUR J. A. SA?LC4L,JR.
CDR V. M. SIMFSON, JR.

Mr.
w.
Mr.
Mr.

G.
T.
A.
D.

CUMMANO

ASSE
w. CBASMAN
B. STAVOVY
STEIN

~F.lCAN

BUSEAU UF SNIPPING

NAVAL SEA SYSTEMS COM!ANU

Mr. R. 6.Cmu
MIS.J. B. OtBRISE
&. k.C. SANDBfXG
Wr. R. F. SWANN
LCDR D. W. WRIDDLW

U.S. GEOLOGICAL SURPSY

~. R. J. GIANG3RELLI
A&. J. 8. LXEWRI

Dr. D. LIU
L@. I. L. STE.W
MAIuTIMS AUIIIN1STRATION
m.
Dr.
W.
Mr.

INTSENATIUNAl SHIP STRUCTURES CONCRBSS

W.

NATIONAL KADSUY OF SCISNCES

SHIP SSSSASCR COWTTSE
Mr. A. D. SAFF - Liaison
h?. R. V. RUMKS - Liaison

N. 0. SA.WSR
!/.M. M4CLEAN
F. SEIROLD
M. W. TOW

S. G. STIAN.ST3V
- Liaison

AMTRICAN IRON & STSEL INST1WTE

Mr. R. E. ,STERNE- Liai,o
SYATEuNIVES.SITTUF NEW YUFW MARITIMS CULLEGE

THS SOCIETY OF NAVAL ASCEITSCTS

& .WINS SNGINSESS

Dr. !..
R. PORTSR - Liaison

Mr. N. 0. :WWRR

- Liaise

D S

WLDING

cOuNcxL

LC3R R. G. VORIWAN

W.

ksssmm

,?.V. XGOPMAN - Liai$on

3?. :;.
-B. ,CX -

Liaison

U S

cUAST WARD

ACADEMY

NAVAL ACADEXY

- Liaison

Member
Agencies:
United
States
Coast
Guard
Naval
SeaSystems
Command
Military
Sealift
Command
Maritime
Administration
United
States
Geological
Survey
Ameriwn
Bureau
ofSipping

Address
Correspondence
to:
Secretary,
Ship
Structure
Committee
U.S.Coast
Guard
Headquarters,
(G-M/TP
13)
Washington,
D.C.20593

structure
Committee
An Interagency
Advisory
Committee

Dedicated
to Improving
theStructure
ofShips

SR-1241
March1981
Uncertainties
areunavoidable
inanyengineering
design.Limitation
onthecontrol
ofmaterial
properties,
milltolerances
inplateandextruded
shapethickness,
timedependent
effects
suchasdeterioration
duetocorrosion,
cracking,
wearandtearareonlysomeofthefactors
that
contribute
totheuncertainties
associated
withtheactual
strength
of a ships
hull.Shipdesigners
andnavalarchitectsusually
treattheseitemsina qualitative
senseas
veryfewattempts
havebeenmadetoquantify
them.
Basedon previous
experience,
thequalitative
assessment
of theuncertainties
doesnotlenditself
to
systematic
improvement
of design
procedures.
Therefore,
theShipStructure
Committee
initiated
thisproject
to
develop
a computer
program
to analyze
theuncertainties
associated
withshiphullstrength.
Thedevelopment
of
theprogram
anditscontents
arepresented.

&/Q
RearAdmiral,
U.S.CoastGuard
Chairman,
ShipStructure
Committee

...

Technical
Report
Documentation
Page
1. ReportNo.

3. Recipients CotologNo.

2, GovernmentAccession~,>~

SSC-301
4. Title and
Subt#tl=

5. ReportDate

PROBABILISTIC
STRUCTURAL-ANALYSIS
OFSHIPHULL
LONGITUDINAL
STRENGTH

DECEMBER
1980
6. PctformingOrganizationCode

t\

8. Pbr{ormingOrgonizotionReportND,

7. Authorts)

J.C.Daidola
andN.S.Basar
9. PerformvrlgOrgonizottonNgmeand Address

10. WorkUn,t No. (T RAIS)

M. Rosenblatt
& Son,Inc.
350Broadway
NewYork,NY 10013

Il.
Controctor GronfNQ.
D07-CG-61908-A
Is.
Type of Repo,t ondPeriod Covered

.
12, SponsoringAgencyNamemid Addres~

U.S.CoastGuard
Office
ofMerchant
Marine
Safety
Washington,
D.C.20593

14. SponsoringAgencyCode

11

G-M

15. SupplementaryNotes

SHIPSTRUCTURE
COMMITTEE
Project
SR-1241
16. Abs.rroct

Existing
probabilistic
structural
design
methods
arereviewed,
their
applicability
toshiphullstructural
design
considered
andthemostpromising
probabilistic
analysis
techniques
areidentified.
Thecurrent
stateofknowledge
concerning
structural
modesof failure
andloaddistribution
isconsidered
withrespectto
itsimpact
onprobabilistic
structural
analyses.
Theemphasis
ison longitudinal
strength
considerations.
Fact:rs
influencing
strength,
tntermsofuncertainties
inshipstrength
distribution,
arerevfewed.
Different
methods
areproposed
toobtain
coefficients
of variation
forvarious
typesofdataontheuncertainties.
Samplecalculations
areperformed
fora number
ofshipsusinganapproximateprobabilistic
method
andy~elding
safety
margins
foreach.Thismethod
requires
thqtonlythecoefficients
ofvariation
ofthestrength
andloadbe
known.
A computer
program
isdeveloped
toperform
thiscalculation
forany
shipsubjected
toanyloadormodeoffailure.
17. Key Words

18. DistributionStatement

Document
isavailable
totheU.S.Public
through.
theNational
Technical
Information
, Service,
Springfield,
VA22161

Longitudinal
strength
Probabilistic
design
Hullgirder
failure
Coefficients
ofvariation
19. SecurityClassif. (oI this report)

UNCLASSIFIED
L
FormDOTF 1700.7(8-72)

29. SecurityClassil. (of this page)

UNCLASSIFIED
1
1
Reproduction
of cotnpl.eted
pageuthorized
*.
-%LtL

21. No. of Pages

22. Price

88
I

METt71C
CONVERSION
FACTORS
Approximate
Conversions
10MetricMeasures
Symbol

ApproximateConversionsf
Symbol

WhenYOIIknow

Muhiplyby

To find

WhsaYouKnow

Multi

SVmbal

lEU
lEfJGTH
in
h
Vd
mi

inches
lee!
yards
mites

2.5
30

0.9
1.6

c.mtmmtcrs
centimeters

<Ill
cm

kilametexs

lull

meters

nm
cm
m
m
km

millirrwlws
ceniimwwrs
nwmls
nwters
kitmmeters

AR

AREA
inz

mi2

squnta
inches
suuarm
icfet
s@mewards
square
mites
acres

6.5
0.02
0.9
2.6
0.4

z.quara
cemnmciers cmz
~2
Squmo
meters
~z
sqmro
nwcrs
sqmmkilmuet{:rs km2
ht+c
I;),*S
ha

cm
~2
km
hO

square
c*nlineters
0.
square
meters
1.
squara
kikmcters
0.
hect~es
[10,0001112} 2.

MASS[weight]

&
y

Winces

pumds
S!WMI
twta
(2000ibl

20
0.45
0.9

Omnt

kitoqranw
40+1,%0
s

u
kg
t

9
kg
i

gmrm

kiIogrwrts
mnnns
(10+0kfl)

VOLUME
tsp

Tbp
IIor!
c
pt
qt

gal
,,3
Vda

teaapams
mbles~s
ftuidcNnc*a
cups
pints
quam
gallmls
cubichot
cubicyard~

5
15
30
0,24
0.47
0.95
3.0
0.02

0,1s

Fahrenheit
lcmpwatum

5/9Ialier
subtract
ino
321

MASS(
0.

2.2
1.

VOL
nlilJi#itacs

miliililefs.
mltli~ilers
kite!s
Iiqers
tilers.
tiler2cb#c
nwiers
cubicnelnrs

ml
mt
mt
I
1
1

ml
I
I
L
~3

Celsius
lcr@ratur*

milliliters
Iiwm
liters
liters
cubic
meters
cubic
rrmaers

;3
ml

0.0
2,
1.
0.
35
1.

TEMPER
c

0.0
0.4
3.3
3.
t3.6

9/5
Cataius
Wllp4tathlre add

.1u,t 2.54
Iwlcl!vl.
f,.
11!.2,
*.=1COllwwblh,ns
,WJ
InLno
&M.3,11w!
Kllh+sr
SLwNW M!ic.
Putll.
2ni4
U,uw
dWuwhts
a,.+
M,-M8us,
Fnco
s2.25,
SOCmdutI
MO.
Cl
3.T0205.

F
-40
0
}t ,1
-20
-40
C

32
80
40
, ,I #
20
0

TABLEOF CONTENTS

Page
INTRODUCTION
..... ..... .. ... .. ... ... .. .. .. .... .. ...... ...

Secticn
2.

STATEMENT
ANDOBJECTIVES. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

fj~~t;~~ 3.

PROBABILISTIC
APPROACH
TO STRUCTURAL
DESIGN
............. .

3.1
3.2

General .** . .
. . . . . . . . . . . . . . . . . . . . . . . . . .
Probabilistic Methods
..... ..... ...................

3
4

3.2.1
Quantitative
Measure
of Performance
......
3.2.2 Classical Approach... ~.. . ~. . . . . . ..- ..-..,

Section

1.

Strength Statistics. ., . . . . . . . . . . . . . . . . . . . . . . . . . . . .

Factors

General. . . . . . . . . . . . . . . . . . . . . . . . . . ., . . , .
Strength Equation. . . . . . . . . . . . . . . . . . . . . . . .
Strength Distribut ions. . . . . . . . . . . . . . . . . . .
T?meDependentStrengths. . . . . . . . . . . . . . . . .

8
6
9
13

LoadStatistics . . . . . . . . . . . . . . . . . . . . . . . . . . . . , . . . . .

15

General. . . . . . . . . . . . . . . . . . . . . . . P, . . . , . . . r .
Equationsand Distribut ions. . . . . . . . . . . . . .

15
16

3.4.1
3.4.2
Sect7cm4.

2
7

3.3*I
3.3.2
3.3.3
3.3.4
3.4

Safety Index Approach

. .......... ...... ...
Strength Reductionand LoadMagnification

3.2.3
3.2.4

3.3

MODES
OFHULLFAILURE
. . . . . . . . . . . . ,,. . . . . . . . . . . . . . . . . . . . . .

Jtl

General. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Modesof FailureoF the Hull Girder. . . . . . . . . . . . . . .
Conclusion.. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

18
lb
20

Section 5.

LOADINGS
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

21

sectian 6.

PROBA!31LISTIC
STRUCTURAL
ANALYSIS
OF SHIPHULL
. . . . . . . - . . . . . . . . . . . . . . . . . . . . . . , . .
LONGITUDINAL
STRENGTH

23

General. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

23

. . . . . . . . .* .+ . . . . . . . . . c. . . r .
Methodology
Strength and LoadDistributions .F. +r..r.
Strength equations. . . . . . . . . . . . . . . . . . . . . . .
Time DependentS:rength Analyses. . . . . . . . .

24
24
25
26

4.1
4.2
4.3

:::

i)evel~prnent
of a probabilistic Structural

~n;l{sis
. .
6.2.2
$.2.3

-v-

TABLEOFCONTENTS
(cont.)
Page

6.3 Application
of Probabilistic
Structural
....***
..**...
.................*..*
......26
Analysis
Methodology
........ ........................**----.....26
6.3.1 General
..*.*...
..........................27
6.3.2 Method
ofApproach

* ... ..
INHULLSTRENGTH
Section
7. UNCERTAINTIES

........................29

.....*.
...............................................29
7.1 General
........................ .............29
7.2 Objective
Uncertainties

7.2.1
7.2.2
7.2.3
7.2.4

.........*.*..
...............................29
General
Formsof Existing
Data*.........................--.
Determination
ofCoefficients
of Variation
(COVS
)....;
COVSfromLiterature
Survey
........................32

Uncertainties
.... .........................---......45
7.3 Subjective
....*.**
.**.....
.....*..
*......
..................45
7.4 Conclusions
,
....... .*............ ................... ...46
CALCULATIONS
Section
8. SAMPLE

.................*..*
.............. *.....
............
8.1 General
..... ..............-.- .............-....%
8.2 Computer
Algorithm
.............*
...........................50
8.3 Analysis
andResults

.........***..
............................
-*............55
Section
9. CONCLUSIONS
........... . .......57
.. ......... * ..............* * ...........
...68
Section
11. ACKNOWLEDGEMENT
REFERENCES
..... ........ ............... . -............ .........59

..... .. *.................
Section
10. RECOMMENDATIONS

APPENDICES.
A. DERIVATION
OFSTRENGTH
COVEQUATIONS
...............................64
ANDDOCUMENTATION
FORCOMPUTER
PROGRAM.
....................67
B. LISTING

-vi-

LIST OFF,IGURES

NO.
.
1

PAGE
Curw!ative Long.TerrnDistribution of AverageBending
Moments
. .. . . .. .. ... .. .... ..... ... . .. .. .. .. . . .. .. . ...

22

Probability of Failure versus Safety Index. . . . . . . . . .

28

COVfor Depthof Stiffener Web.. . . . . . . . . . . . . . . . . . . . .

34

COVfor Breadthof Stiffener Flange...............:.

35

COVfor Breadth and Lengthof Plate . . . . . . . . . . . . . . . . .

36

ApproximateProbabilistic MethodAlgorithm. . . . . . . .

48

ApproximateMidship Section for UNIVERSE

IRELAND..

52

Reliability

54

Analysis . . . . . . . . . . . . . . . . .* . . . . . . . . . ,

-vii-

,,

.. ,

LIST OFTABLES
PAGE

NO.
.-

CarbonSteel Plates Data. . . . . . . . - - . . . . . . - . -. - , - , . . . .

33

As-Rolled Plate Data... . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

33

Data for Carbon-SteelWide-FlangeShapes

.. .. .. .. .. .. .

33

Flange BreadthUncertainty. . . . . . . . . . . . . . . . . . . . . . . . . . .

37

COVof Depth. . . . . . . ..**.*.

38

..*.**.
COVof Beam

. . . . . . . . . . . . . . .......

. . ...** . . . . . . . . . . . . . . -------

-.*-
. ..-..

39

Uncertainty--Depth Of Ship.. . . . . . . . . . . . . . . . . . . . . . . . . .

40

Uncertaint-y--Beamof Ship. . . . . . . . . . . . . . . . . . . . . . . . . . . .

41

Uncertainty--Thickness (Receipt Inspection). . . . . . . . . .

41

Uncertainty--Thickness (Undercut). . . . . . ..- -.- . . . . . . . .

42

Objective Uncertainties. . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

43

Objective Uncertainties. . .. - . . . . . . . . .- - -s...---

---

43

Subjective Uncertainties. . . . . . . . . . . . . . . . . . . . . . . . . --.

44

Apptoximate
Probabilistic Method.. . . . . . . . . . . . . . . . . .

47

~UNIVERSE
IRELAND
Characteristics . . . . . . . . . . . . . . . . . . .

51

uNIVERSE
lRELAb!D
Structural Variables. . . . . . . . . . . . . .

53

IIJNIVERSE
IRELAND
Uncertainty COVES. . . . . . . . . . . . . . . . .

53

St,rengthCOVEquationTerms. . . .. -------

-v{<i.

. .. .... .. .. .. .

53

NOMENCLATUR&

Area of Flarlge~
Area of Webs
Beamof Ship
Depthof Ship
MeanSquareValue
Density Functionof Lbad
Distribution Function of Load
Density Functionof Strength
Distribution Function of Strength
Density Function of Load
Distribution FunctionofLoad
Height of Static Wavethat Yields Average Irregular WaveLongitudinal
BenrJingMoment
Strength Factor or Parameter
Parameter
Lengthof S+ip
Meanof Margin of Safety
Margin of Safety,(S-Z)
Meanof Loacl
Heanof Strength

Meanof Still

WaterBending
Moment;
(mO- Deterministic SWBM)

Deckor BottomSection Modulus;orNumberof Data Points

Probability of Failure
Probability
Probability Density Function

-ix-

s/z

Reliability

R=

Rate of Corrosion

Failure GoverningStrength

NominalStrength Under Idealized and StandardTest Conditions

SM

RequiredSection Modulusof Ship Hull

Fay

Tensile Yield S,tres,sof Material

Thickness

tc

(For CorrosionAllowafice)

to

(Original)

tf

(Of Flange)

(Of Web)

$1

(For Limiting Stress}

Tn, ~JMeans of tn and t=.)

m1

Tolerance

v~

Coefficient of Variation (COV)of Strength, (w,)

Vz

Coefficient ~f Variation (COV)of Load, Fz/mz )

Vx

Coefficient of Variation of x

VXa

Coefficient of Variation of Objective Uncertainties of x

vXs

Coefficient of Variation of Subject ve Uncertainties of x

.%

Meanof Variable x

Failure GoverningLoad

Coefficient of Variation

*y

#yn

Density Functionof the ExtremeWaveBendingMoment

Distribution Functionof the ExtremeWaveBendingMoment

4zn

Density Functionof ExtremeLoadCunposed

of WaveBendingand Still
Water Bending

4 Zr,

Distribution Functionof the ExtremeLoadof WaveBendingand Still

Water Bending

-x-

Random
Variable RepresentingConstituerit parts. of Strength
Central Safety Factor,(mS/mz)
Density Function of Load
Distribution Function of Load
T

Time
Coefficient of Variation of Strength
Variance of Margin of Safety
Average Failure Stress
Variance of Load
Variance of Strength
Variance of Still

Water BendingMomentor ~enetalvariance

Average Failure Stress of Hul.l,Material

Safety lndex,(mN/~M)
Correlakinn Coefficient
StandardTabulated NormalFunction
.,.

Mm2il

Initial
Lateral
Deflectio+fi-of.Plating

-xi

SECTION1.0
INTRODUCTION

The conventional methodsof performing longitudinal structure designs

of ships makeuse of accumulatedexperience from previously built ships of
similar size and function. The accumulatedexperience is mostly expressedin
the form of semi-empirical formulas contained in classification society rules
and designspecifications. The designs resulting from this approachare uncertain
as to the degree of structural adequacythey afford even thoughthe ship designs
basedon these approacheshave given acceptable service, The uncertainty stems
from the assumptionsmaderegarding parametersaffecting the environmentand
the szt-rq-gthof the ship. Manyyears of design experience have shownthat by
using appropriate empirical marginsfor strength over expected load, the unknowns
can be accountedfor and ships with acceptable risk or probability of failure
Iev&ls designed.
With the advent of newship types, and the resultant lack of accumulated
experienceII on ve$se~sof similar size andfunction?
it has becomea professional
responsibility to look into amore,scientific, or rational, approach
to longitudinal strength design of ship hulls, in this context, various investigators in the ship research communityhave adoptedprobabilistic structural
anaiysisproceduresfrom mechanicaland civil engineering. II-I the probabilistic
approach since the quantitative values of manyof the factors affecting the
strength ;f the structure and the magnitudeof the load are statistically
determined, the resulting measureof the adequacyof the design is aIso
5tstistical in nature,
In the study presented in this report, various facets of probabilistic
sti-ucti~ral design were investigated with emphasison applicability to ships,
Section 2.0 gives a statement concerningthe detailed objectives of the
study. in Ssction 3.0, probabilistic structural analysis IS reviewedfrom L:general
standpoint and its applicability to ships is noted, Section 4,0 discusse~ the
possible structural modesof failure of a ship that pertain to lofigitudinal
strcn~thq The present situation with information on ship loads as ~li~y relate
to structural design is discussedin Section 5.0, and the probabilistic
structural analysis proceduresthat showpromisefor ship applications are
presented in Section 6.o. In Section 7.0, the invest~gations~analyses and
collected information performedand obta;ned as part of this study n the area
of the uncertainties of hull strength with respect to the statistical description
of -thertrength are presented. Section 8.o gives samplecalculations for
different ships using a probabilistic structural analysis prccedureembodied
in a computerprogramincluded in the Appendix. Sections 9,0 and !0 O
give the conclusionsand recommendations
respectively arrived at as a result of
these ~tudie,~.
The references cited in the report are listed in Section 11.0

-1-

SECTION
2.0
STATEMENT
OFOBJECTIVES
The objectives of this study were modified by the Ship Structure Committee during the courseof the project to be commensurate
with what was found
to be available and possible within the rather small funding allocated.
The final objectives can be stated as follows:
o Survey the existing literature on reliability analysis and probabilistic design methodsin structures. Comment
on the applicability to ships.
, or use an existing method,for the formulation
o Developa method
of strength in terms of the meansand variances of its uncertainties. Althougha mathematicaldistribution of strength is not required, observations are to be madewith respect to the impactof
using only meansand variances.
o Relate the existing bending
mcrnentdistributions calculated from
existing data to the developedstrength distributions using an existing
methodfor structural reliability analysis. Use available statistical strength parametermeansand variances and makeassumptions
for any strength or load parametersfor which no statistical data
are available.
o Developa FORTRAN
IV computerproqramto perform the aboveprocedure with the objective of determining the safety level of a given
ship subjected to a given load.
n AppIy the developedmmputer?zedproceduri tc actual ships,
o On the basis of ob,tainedresults , suggestfurther research to
develop suitable longitudinal strength criteria for future designs.

-2-

SECTION
3.0
PROBABILISTIC
APPROACH
TO STRUCTURAL
DESIGN

3.1

General

The objectives tif this study include the analysis of uncertainties

associated with ship hull strength and the developmentof expressionsfQr
Suchanalyses require the adoption of a probabilistic
structural reliability.
structural design approachsince a purely deterministic approachcannot yield
the desired information.
In the deterministic designof structures, the strength of the structure
is always increasedabove that whichwouldjust survive the greatest expected
luad by an empirical margin. The ratio of the latter to the former strength
is usually termed the factor of safety. It accountsfor all the unknownsin
the load and strength and yields a structure that shouldhavean acceptable
performancebasedon past experiences.
The fundamentalaims of a probabilistic approachare to moreclearly
and rationally define the necessarymargin, or factor of safety, and obtain a
quantitative measureof performancethrougha rational rather than empirical
analysis. The measureof performanceis usually called the probability of
failure or reliability.
With suchaims, it is not necessarythat a probabilistic analysis be exhaustive in that rationalization of evenonly one of the
unknownsin the factor of safety will put it on a sounderfooting. In this
vein, the-ultimate result of improvedprobabilistic analysis procedures,as
far as designers are concerned,will probably be rational factors of safety
basedon desired quantitative levels of performance. The probabilistic
analysis itself need not be executedby the designers, although this could
be possible.
A completeprobabilistic structural analysis wouldproceed in the
following manner[9]*:
o Conductan analysis of failure modes.effects. and criticality.
Identify-all significant failure rmdes-of the structure.
List the causeof these failure modes.
Identify all parameterscontributing to these causes.
Determinethe criticality of all siginficant failure modes
to the successof structures.
List the n~st.ctiitical failure modesin order of priority.
- Formulate the relationshi~ betweenthe critical .~aramete;sand the failure-governing cri~eria involved.
0 Determine the failure-governing load function.
0 Determine the failure-governing load distribution.
0 Determine the failure-governing strength function.
a Determin&the failure-governing skrength distribution,
o Calculate the probability of failure or reliability associatedwith
*

Numbersin brackets iqdicate similarly numberedreferences in Section 11.0.

-3-

the failure-governing
loadandstrength
distribution
for each
critical failure mode.
An upper boundcf.the total probability of failure or a lower
boundof the reliability will be the sum of the individual
probabilities of each of the critical failure modesunder the
assumptionthat these modesare mutually exclusive events.
Becauseof the difficulty associated with the determination of
the fiji]ul-e-governing
load and strength functions and di~tributions,a number
or probabilistic approachesor methodshave evolved. They differ fundamentally
in tile two primary aims of any probabilistic analysis as mentionedabove:
0 Quantitative measureof performance
0 Rational quantification of load and strength
Actually, not all the approachesare necessarily probabilistic in
the mathematicalsense in that for some
, probability densities and distributions
are not needed,and the output is not a probability.
Thesemethodsmaybe groupedas follows:
Classical probabilistic approach
0 Safety index approach
0 Strength reduction and load magnification factors approach
The presentation in this section is divided into three groups. The
first groupdiscussesthe general approachused in obtaining the quantitative
measureof performanceof a structure given the load and strength statistics.
T!-Icnext groupseach deal with details of the strength and load formulations
respectively, in a general sense. More specific mentionof these considerations,
as applicable to ships, is given in Sections 5.0 thru 7.0,respectively for
longitudinal stre;gth, and for uncertainties in-the ;trength-nfthe ship shull.
The literature contains abundantsourcesof probabilistic structural
ariayses Most of the work has beendone in the areas of civil and mechanical
e~g neer ng but has morerecently spread to naval architecture.
Probabilistic design conceptsfor structures were first proposedin
Since then, several investigators have presented
the ~.sm in 1947 [1].
further considerations for applications
in civil en~ineerinq.
. .
- . References[2]
thru [6], mechanicalengineering, references [7] thru [9], and more recentiyin naval architecture, reference [10].
Within the frameworkof the present study, a brief review of the
numerousmethodsas cited was performedto identify the oneswhich wouldseem
appropriate for future consideration in probabilistic structural analyses
of ships from the standpoint of design.
3,Z
3,2.1

Probabilistic Methods
Quantitative Measureof Performance

As previously menticned, the existing probabilistic structural

analysis methodsdiffer in the output measureof performanceof the structure

-4

being consic!el-ed.
Thosemethodsthat are moreprobabilistic in the mathematical
Their measureof performanceis
in terms of a probability defining failure or reliability.

5?n5e, generally, are of the classical type.

The other methodshaveevolved primarily due to the difficulties

associated with executing a fully probabilistic procedure. Their measureof
performanceis not a probability at all, instead, it is a numberindicating
either a margin of safety or reduction and magnification factors for strength
and load, respectively. Thesenumbersdo not havea physical siqnificande
like probabil-ity of failure or reliability,
but they can be comp~redto each
other for previous successfuland unsuccessfuldesigns to obtain lim ting values.
3.Z.2

Classical Approach

The one common

point in all probabilistic structural ana vsis vrocedur-esis the definition of the probability of failure and reliability.
1
f
the failure-governing load is Z and the failure-governing strength S, then the
probability of failure, Pf, is qiven by all probabilities that the failure~overning ~oadexceedsthe-fail~re-governi ngstrength:
(1)

Pf = P (Z>s)

The probability df fai ure is also called the unreliability,

the reliabil ity, R, becomes:

while

(2)

R = I-Pf = P (S>Z)

Equation (1) is presented in muchof the literature. for exam~le

in [10], a~ directly applicable to ships in the following manner:
Pf = P [s<2] =P [~1] = P [Q<l]
= P [(S-Z)<O] = P [M<O]

(3).

The terms Q and M of Equation (3) are functions of two randomvariables:

the stren~th, S, and the load, i?, and themselvesrandomvariables whose
probability must be determinedby joint probability densi,tyand distribution
functions. However,there seemsto be a universal a reementto qon~ider
the load and strength statistti@ly independentso t i at thestatistics
af ElandQ can
b directly determinedfrom thosd~f S and Z. This assumptionappears to be
reasonable for most strength considerations as long as the effects on the
structure of being in an aqueousenvironmentwith wavesfor a long period of
timcare accountedfor in the strength. If ~ [z,) and@&}are the probability
density and distribution functions of the load, respectively, and fs (s) and F* (s)
those of strength, then it can be shownthat the density and distribution
functions of Q are, [10]:
Q (d

= ~~(j(z)

FQ (q) o~~(d

fs (qz) z dz

(4)

FS (qz) dz

(5)

5-

and the prcbabil ity of failure becomes:

Pf=e

~+(dF~ (z) dz

= ]-j~(Z)

fs (Z) dz

(6)

(7)

Equations (6) and (7) are rather simple and could easily be
evaluated provided the derlsity arid distribution functions of load and
strength are known. This is where the crux of the matter lies and will
The methodsthat makeuse
bradiscussed later in Sections 3.3and 3.1}.
of Equations (6)and (7)varySignifi=rrtly
incornp~exity
andeffort
required for execution.
Equation (7) can be evaluated for each modeof failure and,
as nated previously, the sumof all probabilities of failure for all modes
will give an upper bound. To do better would require the joint probability
de~sity function of strength in the various failure modeswhich would be
atbest very difficult to obtain. A lower boundorI the probability of
failure can.be determinedby assumingthat the modesof failure are perfectly
correlated.

3.2.3

Safety~
.

The difficulty in obtaining load and strength density and distribution functions has led investigators to develop approacheswhich miniin!ze the effort required. For instance, in,the area of ships, [13] contains
an approximate semi-probabilistic design methodwhich wasmotivated, among
other things, by the lack of data on loads and strength and by the controThe method
versial status of forms of load and strength distributions.
requires that orIly the meansand variances of the load and strength be known.
This approximateapproachconsiders the margin o safety II
5
of Equation (3) as a randomvariable
withmeanw andvariance
OH.

f = P ~M<O]= P [~w<~] CM = P [WY] = FG (-7)

(8)

By using the error distribution of M, [161, discussed in moredetail

Section 3.3, the meanand the variance of H can be written:
(9)
0H2 = ~z

S+i!

~2

(10}

where:
riis, 0s= meanand variance, respectively of strength.

rn~: q = meanand variance of total load.

The following results are obtained by algebraic processes:

in

~j=, .-.-
1+~vsz+vzz+~%$
l-+f52

..8

1/2

(12)

(13]

s
ski .
gN

(,1,)

/u
Y = safety index = mMM

Vkre:

central safety factor = m$/mZ

OS
coefficient
of
variation
(COV)of
strength
=

s =
s
SH = required section modulusof the ship hull

z= COVof load = rjmZ

=required section mod~lusof s-hiphull

91

aN = average of failure stress of hull material

mn = meanof the margin of safety
2 --= variance of the marginof safety
h
FromEquation (8),it can be seen that each vaiue of the safety
indexY is associated with someprobabil ity of failure. However,Equation (8)
cannot be evaluated since the distribution function F is not known. If
erioug~information were a~ailable to detsrrnineF~, th% Equations (6I and (7)
fif the ciassical approachcould be used directly.
FromEquations (11)
through (14),
itcanbe seenthattheinputs
neededtoobtain
a hulldesign
, meanof the bendingmoment,and
strength
arethestrength
andloadCOVS
the safety index Y= The am~untof computationis insignificant.
The safety index Y is a sinqle numberthat must.be obtained
on Yhe bas,isof manyteciinf:cal~~actdrs. l.t has previouslybee~

proposed
[13]todetermine
thisvaluefromexisting
designs
totakeinto
if the probability
Inaddition,
account
thevastaccumulated
experience.
of failure associated with past designs is socially acceptable, then this
aspect is also considered.
3.2.4

~h

Reductionand LoadIlagnification Factors

This method, discussed in [5,62,63],

is similar to the approximate
inethoddescribed above in that only meansand variances of the load and
strength are used to obtain relative and semi-probabilistic measuresof the
structures performance. In this case, the measuresof performancearethe
strength reduction and load magnificatioti factors.

-7-

The strength reduction factor. f., and load magnification factor.

s
can be defined as follows:

z
f. ~

= minimumstrength = S-K$ S
average strength
m~

=. 1-KSVS; $< 1

(15)

l%cto~$
givi~thenumberofstandard
deviations betweenthe averageand the
minimumstrengths and the maximum
loads,
respectively.
For a safe design, the minimumstrength mustexceedor equal
the maximum
load:

where:

%>= K

(17)

mf > m

s s-

z z

The values of acceptable strength reductions factors and load

magnification factors could be obtained from past designs in a similar
-fashionto the safety index of the previous section.
In [5], this approachhas beenextendedto fatigue for both the
cocstznt rangeand the randomloads.
Similarly to the safety index approach, the analyses required
to ex.ec~tethis methodare quite limited in extent and tiomplexi~y.
3.3

Strength Statistics

3.3.1General
It mustbe first stated that the strength of the hull girder
mayor maynot vary with time dependingon the failure modebeing considered. Time invariant strengths will include yielding and buckling.
Time variant strengths will include fracture, fatigue, Snd reducedstrengths
due to corrosion. For ships, time variant strengths will a so normally
include randomloadings of low or high cycles, and possibly thermal loadings.
This scenario shouldcover the most significant nmdesof hu 1 girder failure
which need to be addressed.
3.3.2

Strength Equation

The strength of a structure is principally described in two

different ways in the numerousprobabilistic structural design methodsto
be found In the literature.

s = f (E,, ~2, ----or:

S= k,k2k3----knS

-8-

En)

(18)
(19)

where:

Constituent parts of the strength which

= are assumedto be randomvariables

cl----en

s
i---

= Nominalstrength determinedunder idealized

and standard test conditions

= Strength factors to convert the nominal

strength to actual strength. (Thesefactors

are assumedto be random
variables)-

The Kfactors.account for physical variables suchas size, forming and manufactut-ing processes, surface finish, load, heat treatment, direct surface
environment, temperature, timez corrosion~ etcThe approachgiven by Equation (18) has beenused in ships,
but the actual examplesdevelopedhave beensuch that only the explicit
functional strength constituents, F, have beenconsideredas random
variables or uncertainties in the strength. As the probabilistic analyses
becomemorecomprehensive
and moreuncertainties becomeidentified, some
of these maynot appear as constituents in the strength equation, and
the approachdepicted in Equation (19) mayhave to be adoptedin addition
to that in Equation (18).

3.3.3

Strength Distributions

Equations (18) and (19) give qeneral expressionsfor the

strength, but since the strength is statistical in nature, the probability
densityand distribution function mustbe specified to completely characterize
it and allow the probability of failure to be evaluated by Equations (6) and

(7) .
The probabilistic structural analysis approachesfound in
the literature assumethat the strength distribution can be determinedin
one of the following ways:
0 Actual component
strength distribution determinedby
actual testing under the exact geometry, application,
andoperational
environment
inwhich the component
shall function.
0 Component
strength distribution synthesizedfrom the
knowndistributions of the constituent parts and
strength factors as given in Equations (18) and (19).
o An assumptionmadeas to what type of distribution the
strength will follow, i.e. normal, lognormal, Weibull, etc.
0 An assumptionmadethat all that can be determinedof
the strength is its COV.
The first
machinedesign and someof
This approachwould hardly
size of the structure, the

of the aboveapproachesis usedextensively in

the test equipmentrequired is described in [7].
seemrealistic for ships becauseof the large
implication of using the whole ship as a dis-

-9-

cardable test component,and the large data samplerequired for conclusive

of the ship structure could be tested
resu)ts. Whetheror not components
;~ndresults extrapolated to the whole ship appearsquestionable. In the
case of welded ship grillages under compressiveload [64]:
Further experimental evaluation of grillage strength also
has a key part to play but cannot be expectedto provide direct statistical
descriptions of grillage strength; large-scale tests of the type described
in the present paper are too expensive to carry out in sufficient n(~rirbers
and small-scale tests are statistically unrepresentative for the reasons
mentionedabove. It is suggestedthat the main role of further
grillage tests should,therefore~be to guide the developmentof improved
analysis methodsand to check the accuracyof suchmethodsand design
data with provision of empirical corrections wherenecessary.
The secondapproachrequires that the distributions of
the constituent parts and strength factors be known. It may, for example,
be necessarythat the distribution of the dimensionsof depth, beam,and
the area of flanges be known. Suchquantities are muchmoreamenable
to scrutiny in ships than the overall testing of the hull girder. AS
discussed in Section 7.0,however,not muchdata presently exist for many
of the variables, and consequentlythe distributions themselvescannot
This would seemto be a promisingarea in the future,
be identified.
if an effortismadetocollect
suchdata.
If the distribution of the constituent parts arid functions
are known,there are various methodsfor synthesizing their distribution
Reference [7] gives eight
to obtain the overall strength distribution.
metimds:

0
0
0
0
0
0
0

The algebra of normal function method

The changeof variable method
The momentgenerating function method
The Fourier transform, convolution, and inversion method
The Mellin transform, convolution and inversion method
The characteristic function method
The cumulative distribution function method
The MonteCarlo method

The MonteCarlo methodwill always give results even for complexfunctions

of non-identically distributed randomvariables although the length and
complexity of the computationswill reportedly be quite extensive and
possibly unrealistic.
The third approachrequires that assumptionsbe made
concerningthe distribution of the strength. Of course the samecould be
donewith the constituent parts and factors, and the secondmethodused
This approachseemsto be
to construct the strength distribution.
universal in Ihe literature for civil engineering and naval architecture.
it is natural that these two disciplines wouldmakegreater use of this
last approachbecauseof the size and complexity of the structure analyzed.

-1o-

This approachrequires the adoption of a distribution (such

as tha normal, lognormal, Weibull, etc.) and the specification of necessary
parametersof the distribution to obtain numerical values from tabulated
Je[lsity and distribution functions. The necessaryparamete~sareat ieast the
first and secondmomentsof the distribution, the meanand variance.
Most of the assumeddistributions in the literature on
structural analysis are the normal and the lognormaldistributions.
It
would seemnatural for investigators to makesuchassumptionssince experimental measurementsin science and engineering seemto approximate,
rather well, the normal law. However,the integrations of Equations (6)
ancl(7)for theprobability
of failure
involve important constituent
parts at thetaile~dof the distributions which can vary greatly depending
In r ference [~]}it is stated that for
on th~ assumeddistributions.
the probability offailure P 4 105, the calculated probability is sensitive to the assumeddistri %ution and the results can only be used
On the other hand for probabilities of failure Pf> 103 ,
relatively.
such problemswould not be too serious.
As reported in [15], the
ind.cate a current probability of failure
10-4 so that these approximationsmaynot
ships if the historical safety levels are

record of world ship catastrophes

for ships in the order of
be a problemin the case of
consideredadequate.

If the strength is assumedto be normally distributed

the probabil ity density and distribution functions are:

j+)

=T*
s

~ (s) JsJ.@
-z
where:

exp -1/2(~:)
As s?!! (.)

(20)
(21)

s = mea~of strength S
s = standard deviation of strength S
Y = standard tabulated normal function.
s

Consequently,under suchan assumption,theonly quantities that need to be

tietermined are the meansand variances of the strength- Then, the probability of failure given by Equations (6)and (7) can be evaluated
(provided the load distribution is known). The latter statement is not
trivial
since~in fact, the meansand variances of ships strength are not
easi ly determinable.,
The approach,in general, has been to expandthe strength
function in terms of its constituents in a Taylor Series about the means
of t}le C(]n$tituents:

s = f(E,,

E2,---- PEn)

-11-

f(r,

~f
+&(Ei
- Ti
) (~~)~t(22)
.

rz, ----, q. )

+ v2z-%i

<+

- ~i)2 (-?!.)
aEi2&;

+....

+ (Remainder)

in which tne derivatives are evaluatea at tne constituent means,K , v2,-1

-- G and the remainder consists of the higher derivatives.
n
If it is assumedthat the higher derivatives are small
Cr zero and that the coefficients of variation of the constitlJents are
small, in the order of 15 per cent or less [16], then Equation (22) can
be linear zed and the following obtained:
~ f(~], z2,----zn)
!!?s
~ (23)

is the correlation coefficient between :. and Ej.

Where &
These assumptionsmaymot turn out to be correct for all shiDs f~r all
It is indicated in [14] that the inclusion of ~onmodesof failure.
Iinearities
in the strength distribution causesvarious changesonly in
the predictions of long-term probability of failure.
Further, makingthe assumptionthat the constituent parts
independent, the corr&lation becomeszero and Equation (24)

are statistically
reduces to:

(25)
..

Equations (23) and (25) have been used in ship studies to

date. The assumptionof~erc cmrelation inherent in Equation (25) may
be reasonable for manyof the constituent parts. For example, in the
case of thestrength defined by Equation (27), the beam(B ) should have
no effect an the depth (D) and similarly both D and B should have no
effect on.plate thicknesses tf and tw On the other hand, as an example,
the strengths in different failure modesof the samepaqel maybe highly
cGri-elatec! [(16].

IfEquation
(25)iswritten
,intermsof a coefficient of
variation

(COV):

m&6:i*
Cov

g=$=
COVsof constituent parts
4
-12-

(26)

Equations (23), (25) , and (26) then gi,ve the strength parame~erls~lean,
variance,and CGVrespectively in terms of the meansand variances of
the constituent parts, ( E ). Thesemust be determinedfrom data or by
estimation as discussed in detail in Section 7.o. The definition of
the strength is then completeand the probability of failurd can then
be evaluated. The greatest amountof effort is neededin determining
the strength COV,and is only a fraction of that required by the first
~ViO approaches. Onewould,of course,havea lesser degreeof confidence
in tha results.
The fourth approachrequires only that the COVor the
man and variance of the strength be known. The procedureto obtain
datacanonlybe usedin the semi~
these was just given above. Ttiese
probabilistic methodsoutlined in Sections 3.2.3 and 3.2.4. This approach
req~ires the least computational effort to obtain its results.

3.3.4

Time DependentStrengths

In general, whenevera critical failure modeinvolves

a t?mevariant strength suchas it does in the cases of fracture, fatigue,
thermal effect , and corrosion, the variations with time must be accounted
for.
If the strength can be treated as a function of time, the general
probabilistic
procedures
presented
previously
can be utilized.
Mechanical reliability
for components
exposedto fatigue is
discussed in [81and [9]. The approachtherein is to use the form of
strength given by Equation (19) which would take care of sometime-dependent
effects through the K coefficients; this is implied but not stated.
djrectly

Fromthe standpoint of fatigue, the following problemsare

addrzssed in these references:
Q Fatigue under a fixed alternating load level, given
the cycles to failure distribution of the component.
0 Fatigue for a specified life given the broad band
strength and load distributions for that life.
Cumulative fatigue under sequential groupsof stresses,
each group having a specific numberof cycles and ths
samemaximum
and meanalternating stress levels.

fh~ approachestG solving these problemsare identical to those previously

discussed herein in that all analyses are performedat a given time in the
life of the componentand at a constant
loadlevel.
Reference [17] reports on studies conductedto
investigate time-varying structural probabilistic strengths in the jet
The basisof the general procedureproposedis a compuengine field.
tational
sequenceto determitm probability of failure w tin-wconsisting
of two phases: the first is a failure ~robability phaseand the second
a de~radation of strength phase. Thus, a probability of failure .calculaticn is made, followed by a strength degradation calculation reflecting
The
someoperation time. The sequencecan be repeated indefinitely.

-13-

crux of the procedure revolves around identifying a time-varying strength

degradation scenarifi. Several types are proposedbut the analyses reported
in that paper were of a preliminary nature. It is noted that additional
work was.in progress at that time.
During the course of the study presented herein, a potential scenario for corrosion of ship hulls was envisioned. If the
modeof failure under consideration is that of yielding during bending
of the hu?i girder as a free-free beam, it can easily be shownthat
the strength equation is:
s = f(D, tf,

B, tw, ~Y)
(27)

= N5y E D(@+l/3twD) s
Y
where:

= deck or bottomsection modulus

= tensile strength
Y
D = section depth
B

= section beam

= area of flanges

Aw w area of webs
f

= Af/2B = equivalent thickness of one flange

t w = Aw/2D=equivalent thickness of one web

if corrosion is introduced, then Af, Aw, tf arid tw
becomefunctions of time as the plating corrodes.
The plate thicknesses maythen be considereda function of time as follows:
(28)
where:

Plate thickness in time -

t(r)

to

= original thickness at T=O

R= = Rate of corrosion, also a randomvariable

The ~tren~t~, ~ou~~ then becomea function of time and the probability of
fa;lure c~uld be estimated at various times during the ships life using
the probabil istic theory previously presented. Alternatively3 the original
strength at time t=o ccil~d be multiplied by a factor kc, reflecting equation
(23), also a randomvariable, to accountfor a specific reduction in
~trengti~ at a certain time in the vessel life.

Another
approachto consider
theeffect
of corrosion
whi~!~ C50CSnot result in a time dependentstrength is to take the total
plate Lhicknessas the sumof the thickness required for limiting stresses,
-14-

pl us

t,

thickuess

for

by

Equation
~:

tc

allowance,

[62]:
(29)

=t+tc

t
Which

corrosion

(26)

yields:

(>)2

. ~:

7=22
(~)

(30)

6=

n
where:
= COV of

6t

the

plate

thickness

due

to

production

tolerances

n
6

c
As

pointed

strength
a Monte

out

COV due

in

[62],

the

fatigue-crack

for
a

tanker

both
the

gives
a
function
a

as

random
variables.
also
presented.

is

random

source

probabilistic
given
therein

of

logarithmic

then

a method

fatigue

quantitative

study
is
is built
and
is
rating
the
the

information

limited
through

service

with

[t

results
rel iabi

point

of

obtained
by
1 ity
of ship
[19].
view.
The

Section

3.2.4.,

random

fatigue.

3.4

Load

has

Genera
As

istic

is

Of

sufficient

S-!1

noted

used

extended

1 istic

analysis

loads.
of

fatigue.
on the

amount
in

as

to

Section

2.0,

Of

hull

under
born

provide
on the

[5]

which

The strength
coefficients
are

regarded

Statistic

data

structure,

this

only
a standard
ity-control
led

clearly

therein

used
of

reference

study

which

maintenance

in

This
this

curve,

is

members

these
raridom
variables
in time
by factors
other

defects

be

procedure

that:

treating
well
qual

therefore,

basis

normal

in mind
information
of

previously

constant

ship
which
fab:-icat
ion
ope-

that
on

design-oriented

discussed
stress

in

range

and

design,
and

discussed

do not
include
details
the
load
is one of the

probabil

wave

the

unexpected

approach

of
using

structural

course

it

of
of

The

structure
law and

this
analysis
will
structures
merely

been

group

others

Statistics

3.4.1

structural
characteristics

and
the

satisfactory

should,

one
by

probabi

transverse

of

to within
a range
sound workmanship

put
inlo
conditions.

for

analysis
on
of strength

lack

on

from

joint.

of
ship
on Miners

and
of

vary

addressed

presented

during

approximation

llbeca~e
or

is

st i 11 water
found

considered

butt-welded

A sensitivity
The degradation

not

wil
been

and

evaluation
is based

1 inear

is

rate
has

[651 .

technique

longitudinal

to

only

this

at

the

subjected

represents

and

initiation

analyzing

corrosion

corrosion

members
to another
Carlo
simulation

In [19],
of

to

it will
be
mechanics

structural

in

concerning
two major

design.

the

objectives

the
load
distribution.
considerations
of any

discussed
that
must
The

here
from
the
be considered
1 iterature

-15-

on

of

this

study

However,
since
probabi
I istic
standpoint
for
appl

loads

does

of
i cation
not

in

address
this
with
respect
this

to

ships

the

fr,l

types

included

in

of

the

Section

situation

5.o

of

of

loads

water

due

Ships

owrl

Thermal

type

of

has

been

Equations

and

strength

and

stated
to

that

presented

3 .3.3

design

Lending

momerits

of

and

same

weight

and

description.
and

require
Hence,

distributions

th+

procedures

into
sti

that
of
11 water

effects

of

should

be

be

only

found

in

the

considered

11

on

water

load

must

since

this

hence,

wi 11

the

add

the

local

random

analyses

of

be

considered.

of

longitudinal

loadings,
load

strength,

such

toward

respect

those

found

ships

presented

ng

moments

in

to

[10]

as

that

increasing

and

probabil

bending
term,

moments
are
assumed
to fol
and an exponential
probability

(X)

been

the

wave

cons idered.
low

(l/k)

. (x/k)

JxfL(x)

L-l

in

probabi

this

1 istic

total

to

water

the.

overal

load
head,
load

and

c--(x/k)

dx = \-e-(x/k)
-T6-

probabi

proposed
1 istic

in

the

structural

bending
moments
and sti 11The ampl itudes
of
the wave

a Rayleigh
law in the

both
the
short-term
respectively,
funct
ions,

f, (x) =
FL

have

used

distribution
Iong. term.

in
Using

and the
long-term
wave
are
given
as follows:

x>O

>0

is tic

wave

information
Section
5.0.

due

distributions

been

therein;

bendi

distribution,
and densi ty

specific

have

water

Weibull
bution

to

stress.
With

1 iterature,

case

of

be

similar

1 iterature
sti

the

(n

should
bending

distributions.

ana Iys

effects

synthesizing

the
whole
and wave

This
is primarily
due to lack
of
as discussed
further
in
of
loads,
that
in any complete
here,
however,

the

apply

well.

types
out

include

these

the
same
al 1 that

would

to other
be pointed
total

for
In

safety.

appl icable
I t should
is,

directly.

vibration.

expressions

the

thermal

to

have

with
of

parts
combining

analyses

induced

margin

ships,

strength

ships

propelrer

deal

as

to

for

slamming,

and

distributions

slamming,
in

(7)

equations

the
constituent
procedure
for

frequency)

including

ity,

mathematical

respect

The
;tructur.~1

carry

strength

load

springing,

moments,

consist

buoyancy.

(row
:

and

through

load
their

the

of
The

(1)

the

With
distributions
emphasized.

hul 1 girder

Distributions

reliabil

for
for

cases

induced

springing,

failure,

and

frequency)

whipping,

of

expressions

most

the

train.

wave
(high

Equations
ity

expressions

to

to weight

wave

Quasi-static

probabil

ied

effects.

Oynamic

the

appl

[59] :

lowing

Calm

in

appraisal

is

report.
The

or

A qualitative

point
extensively.
to
loads
applied

(3?)
(32)

the
short
the
distri-

wi 11

i!.~ 2 for short term

= 1 for long term
k = &for short term
k = A for long term
E = meansquare value of L taken over a short period of time
a = expected valueof L taken
overa longperiod
of time
It should be pointed out that in [60] it is shownthat the
exponential Tawunderestimatesthe data measuredonboardan
Ore/Sulk/Oil carrier. Therein, it is concludedthat mathematicalmodels
basedon the normalor general Weibull distributions give excellent agreement with statistical data for the ship analyzed. Reference [61] shows
Ihat for two other ship%the Wibull distribution does not exactly fit
the data.
In Reference [12], order statistics
areusedreobtain
theextreme
wavebending
momentdensity
anddistribution
functions
using
equations (31) and (32)4(Y) =

Yn

These extreme functions become:

& (y/k) Ll-e - (y/k) [l-e-

(y/k) ]n-~ y>n

Q(YL p[yn<y] = [1-e- (y/k)k] n y>()

Y,~
where n is the numberof wave records considered.

(33)
(34)

bending momentisincorporated
firstas
ciaterministic
andthenasa normally
distributed
randomvariable. The
combinedstill-water and wave bendingmomentprobability density and
distribution func~ions in the deterministic case are:
.
.
The still-water

= 0, otherwise, ~ >Me
(36)
= O-otherwise, z >mo
where m is the deterministic bendingmoment.
o
The prohabil itydensity and distribution functions in the

and
9dl
mzm(z)
= -,
k uafi

- ~/@~-i,e-(@)J
o

.[~_ ~-whl]m-lj:. ,-~~%) d~v

where m and a are the meanand standard deviation of the stiil
momentrespectively.
-1 7

(m}
waterbending