~

@!B
. ..

..l

.! ,,c,,e

.,

%,, ,.,...~

THE SOCIETY OF NAvAL ARCHITECTS AND MARINE ENGINEERS 601P..o.I. Avenue, uite00, S 4 Jerseyity, Jersey7306 C New 0 ,,,.,0,,.,,., ,, ,. SW=,.,., S/W,m,,.., , M.,,,. W&M,,, ....1.,.., ..,.,, .,,,,,,., ,,,, .,,.,,.,, 0,,.,., 6.,,

Reliability Analysis of Offshore Structures

C. PafiOu and M. Shinozuka, Columbia University, New York, New York Y.-N. Chen, American Bureau of Shipping, Paramus, New Jersey

ABSTRACT

.kfthough these documents deal primarily with building structures and nuclear power plants, respectively, they address themselves to some of the basic issues of etr.ct .ral it egrity assessment and, as such, repzesmt the-art in the probabilistic engineering systems. The offshore and ship-building significant deavor. industry has also made the state-of.

The purposeof khk studyisto developa generaf the method forestirn.atinz system reliability of offshore structures with the ~d of the full distribution method
and to introduce a probabilistic definition of structural redundancy. Structures ?... treated as systems of structural components. Failure of any umber of the% mmp.mnts results in a redi.stri hution of cbe internaf or/and exter..1 forces. The probability of str.cturaf failure is then evaluated by examining a limited number of significant zequence.s of member failures that produce collapse of the structme. The structure examined is m indeterminate deep offshore truss under fully developed sea conditions. Two different types of material behavior are mrmidemd to characterize the type of failme of the oxnpormntq ductile and brittle beb aviom. Them reliability amdysia and m. dundmcy definition wifl form an important malyticd be sis for fuxther investigation of offshore structuralintegrity.

design and ane.ly~is of complex

progress in the same area of e gineerin g en. symposi@workshop retitled

1 fact, a mcmt

rDesigr., Inspection 1984) organized

and Redundant.yz

(Faulkner et d., Board, N-

by the Marim

Structures

tiomd Research ComIcil and held in November cussed the subject

1NTRDDUCTION

1983, dis.

matters indicated of reliability

i the title within analysis and design,

the gemral

framework

and he..isan indication ofthk industry. recognition

Recent advamms methodology in the prohabfitic safety analysis

thats,tmx.tum.l reliability area crucial issues igred,et i design pmxedm.s.

While these advmces and efforts are impressive, there are still a number of important questions that need to methodology Typically,

and probability-based

design procedures for

structures and stmct.ml Iicatim w,.

systems have resulted in the pubreports and pa-

of a large nmnber of technicaf S.me .f th.,. m. i.te.d.d ..d.s f..

use in dewkwing in providkg a

be answered effectively before a probabilistic

Pr.h.bility-b=.d theoretical

d=iw

and mm.

cm tmly respond to the needs of the id.stry,

basis for a risk a.ssesmnent procedures

guide.

the following items, whkb are all heavily interrelated, ap pear t. be in need of immediate Practiti.n.rs atkntion on the p-at of

Typical of the former is . NBS pblimtirm

by B, Elling-

wcmd et al. (1980) while belonging prominently to the latter is am NRC document hy J.W. H,ckman et al. (1983).

w well as researchers

reliability estimation A. Estimation of system reliability;

143

Pr-dures

must be d-bed
systems.

for .fih.r.

.,

water depth, z = vertical coordinate

axis, positive in the

Structures as structural B. Full distribution

upward direction and measured from the mean sea surface reliability .Ievatim and g = gravity accelemtion. Note that

methods

for improved

analysis of offshore m ship structures. W2 = kg tanh[kdl C. Load combination analyses wbicb load combinations For deep water, i.e., d + mdce to , S*,(W) and Us = kg (4) w of the water particle = uSnn(w) e+ (3) co, Eqs. 1 and 2 respectively (2)

are to be considered in the design, what in the appro.

for eachcombine Priate level of a target safety index $ion, tc. e D. Effects ofstructural redundancy reliability on performance damage-tolerant
nmst be implemented. or fail-stfe design c.ncep~

5imilarly, the horizontal component The method stmctures present study, however, the system Primarily reliability develops a acceleration

has the npectrd density function

for estimating

of offshore method. S&i(w) = w2Stti

, (w) = wSqq(w)e~
(5)

with the aid of the full distribution definition of structural

Also, a probabilistic is itrodumd redundancy

redundancy

For the purpose

of this study, we asmme

that the

in this study. These reliability analysis and definition will form an importmt of the questions am.lytical surrounding

offshore tower is excited by wavm under fully developed sea conditions for which the Pierson-Moskowitz spectrum

beais for the investigation load combination analy~.

of the followimg form is used (Pierson & Moskowitz., 1964);

EXPECTED AND

MAXIMUM

WATER PARTICLE VELOCITY

where the parameters

a and p are as.mrmd to be a =

ACCELERATION

0.00g1 and P = 0.74 in the nrnerical analysis that follows. The quantity W in Eq. 6 indicata a windspeed represenThe spectral is de-

The wave analysis petformed in the present study uses the assumption of the .m.aU amplitnde (Airy) theory imand the Then,

tative of a fully developed

sea condition.

density function for W = 1160 in/aec (29.46 m/see)

plying that the fluid is imiscid, ratio of wave amplitude

incompressible

picted in Fig. 1. 1 the present analysis, W is treated ~ . random variable governed by a log-normal function. It follows Leadbetter, from rmdom process theory (Cram.4r k distribution

t. wave length is small.

it can be shown that the power. spectral density function S,, of tbe horizontal component C of the water particle

velocity is related to the power spectral density function .%(w) of the water sufiace ~vfition 7(t) thrOugh

1958) that the expected

maxirrmm value of

l~(t)l fi O S t S T is given in e.pproxinmticm by

s(w)=w,.sn.(w)[-h[(+:~:l(+~l]
where u = circular frequency,
k = wave number,

(,) K, = {2 ln[z~+

d =

(0)2!]+

Y
.,/2 ln[2fi+ (0)T1

(8)

L-

and 7 = 0.5772 (Eulers constant) (9)

values of them POW.. spectral density functions in the form of a negative exponential water depth -z icre U.S.

deczease

function +(z) as the

In Eq.

8, T is the duration

of the storm,

U+(0) is the $+(,) = ..P[*] (19)

expected rate of zero crossing from below by the velocity

pr.c=s

w(t) and fi given by

Furthermore, integrations merically

in evaluating u;, o< and cr., the respective

i+(o)

= & .:

[10)

(Eqs. 11, 12 and 16) will be carried out nu= 24 md/sec. That w. covers den-

up to u = u.

where a, and c; arethestandard deviation ofO(t)and respective ~(t)and are obtained from integrating their power spectral demity functicmq

the frequency range over which the power spectral sity Sn. (u) is significant u= 0.289 radfsec

can be seen from the fact that of W = 1160

for a mean windspeed

.:=

J-so(w,&

infsec (29.46 m/see). (11)

Note here, that the value of w. was

take to be extremely large but tbe same numerical results and .:= J-S,(W)* (12) WIND. INDUCED Similarly, WAVE FORCES would be obtained using a smaller w.lue (i.e., 6 rad/see)

ii,..,.

+ E[ma.lti(t)l

in

O s ts

T] = K,r,

(13)

While the structure we deal with i thin study is a fixed cdfsbore truss as sbmvn in FLg, 3, we will first con-

with
sider vertically standhg offshore piles in order to ewdate wave forces on truss structures, estimated by Mor-

KU = ~2 ln[2ti+(0)Z] +

7 ~2 ln[2u+ (0)T]

(14)

the effect of wind-indmed

The wave force m an offshore pile is .ally and from the well-known empirical formula mgge.ted

ti+(o) = :.:
where U+(0) is the expected

(15) rison et al. (1950):

rate of zero crossing from u(t) and

below by the acceleration

proce~

f(.)
(16)

= cd

~v(z)

+ ~CDPWZ)l~(Z)l

(20)

0;=

~-sv(w,*

in w~,ch t = time, f(z) = horizotd the pile at water depth z, p = mm

force funit len@h of demity cd tbe water, are respectively the

with s.(w) In order = os,(w) to avoid .due

== Aoq(w)e%
analytical

(17)

D = pilediameter,

and C,,. and Co

inertial and drag coefiiciemts, complications,

In the nwnerical

analysis

that follows, Cm = 1,5 and CD = 1.0 are assumed. the dependence of the power spectral density functions lrI the dymrnic s,,(u), S,, (w) .md Su. (w) m z is simplified. cornpliihed by using when the velocity ~, = (:p),/~ .+ (18) and acceleration of the stmctmd m.. Thm is Wanalysis of pile response, the inter.

action betweenvczve and structure shouldbe considered

tion are of tbe same order of magnitude wat.. part ides. It is geuerally accepted

as that of the that tbe effect

in the factor exp(2ua z/g) in Eqs. 3, 5 and 16. Hence, the

145

..

of tK,. interaction

can be

incorporated into t be Morri-

direction thereof sdivided i intotwo equalcomponents, and eachisconsidered esan external force acting n each o end (. node of the truss structure to be analyzed) of the
member. Tbe structure is then subjected t. tbcme external forces resulting from the distributed forces actin g.. its members; all of

mn equation by .,ing tbe instantaneous relative velocity and acceleration between tbe structure and water particles. In tbe present study, bowmwr, this effect is di.we-

garded and the structural analysis is performed in a q.asistatic fashion. In a recent work (Paliou, C. et al., 1986)

F; represents the mm of these forces acting

this approach is extended to dynamic response evaluation, withcmt disregarding this effect, md incldig a fatigue

at mode i (Fig. 3). As mentioned in the previous section, tbe windspeed W that represents a fully developed wa stat. is =umed

am.lysis and t be effect of inspection. In order to circumvent undue analytical difficulty

to be a random variable governed by a log-rmmmd d,stribtion (Ymg, 1978) 1 Iog(clw*) .& Pi j

and -& the same time to be on the conservat iv. side, 0,,,.. ezp(~z) and ii,,.. ezp(~z), evaluated in the pre-

fw(.) ceding ae-ction, are used in Eq. a mmervativ. approximation and i(t) 20. Thb is obviously

2 loge W=-W-j[

1 } (22)

since in actuality the maxwill not usually occur at

in which w is meammd

i in/see,

p>

and ok

represent

imwn values of J(t)

the expected value and standard deviation of log YIm = log C, W, where YM is the annual expected maximum

the same time instant, but also bearing in mind that the main mbj ect of this paper is the development of a reli-

wave height and

abilityanalysis procedure, this approximation of the load configuration serves as an illustration. Hence,
For tbe NorthSea,theuseof& (23)

f(z) = CmP ~

%.a. e~p (~.)

.Zp (&z)

+
(!31)

= 2.842 and c&,= 0.1

, + : CDPD&=

T was sugge.tednYang and Fre.denthd(1977).be deni sity function of W with these parameter values is depicted

The force perit Iegtb evahated Eq. 21 is based by on tbe Morrison formula for vertically standing members. Even wbm a member is inclined with respect to tbe ver-

in Fig. 2. The expected ml. of AW and standard deviation uw

of W can then be evaluated from

tical direction by a small angle 0, we assume that Eq. 21 pw = c.p(# cm still be used with %..s of u,,.., and ;,,,.=, indicates that i,,,.. cca O and w,,,., cos.9 in place This asmmptim basically md .W=,wn (z5)

+ :)

(24)

respectively.

cm.9 and %...

con # are assumed to

produce (in approximation)

forces per nit Imgth perpm. where 110 y

dimlar to the member axis. The effect of the cmnpomnte p = (/l& Iogcl). of i,,... and win.. in the direction parallel to the member Tbe horizontal com.=.; ponent of the re,.lt~t force acting almg for.. derived from the dktributed 110 .7

(26)

axis is disregarded in epproxinw.tim.

(27)

the member and in the perpmdimlar

The reliability analysis the proceeds as follows:

146

(a) Working in the range of windspeed from a mea ml.,

cormptation, however, s i quite involved since, inprinconsider all the possible sequences of ciple, rn.st we member failures that lead to cokp,e of thc strct ure In the present study, a strmtur.1 ered b have occurred flection collapse ie comid-

@w of N .6 standard deviations

- 60W ) toa mean + 6UW ), the probability

due ofW + 6 standard deviations (W range divided is

into 200 intervals and the

ofwidmpeed ieachintervsd isevaluated.

(Eq. 6), the (b)UsingthePierson.Moekowitz spectr.xn expected maximum wave particle velocity and acceleration (%,.. and ii,,...) which corresponds to each

when excessive structu~ ai de-

materialize after failure of a number of memstiffness matrix becomes at

bers or the corresponding least near singular.

windspeed is computed. (c) Having computed tbe values of the expected maxi.

(f) At this point, the notion of simultaneous

failure of It

members used in this study shall be made clear. is acknowledged

mum water particle velocity and acceleration, the distribution load exerted by the waves on each member is computed by the Morrison equation (Eq, 21). Thus, F; of the wave force acting

that if tbe intensity of the load in-

cmzses from zero to a certain level, the the probability of simultamous failure of two or more members

the horizontal mmponets

will be zero if the strengths of the members are random. However, for ewe of the probabbtic it is -umed ..s throughout analysis,

cm each ode of tbe structure can be calculated. (d) TWO cases are examined herein. Case I - Ductile Be-

this study that sirmdtane-

h.vioc

Whenever a number of members fgils either in or buckling, the effect of those

failures can take place in those members whose are less than the intern al forces resulting

tension, compression

strengths

members will be rephced

by pairs of extemaf forces

from this level of load intensity. (g) I. the evaluation of this mnditiormf probability men.

acting in the direction of the axes of these members .~d to tbe member forces due to the same Ioad,g condition of tbe intact structure Wbenewr (see Fig, 4.), Case

tinned in (e) above, the material strength .wcb as cry,, a~. or UBk is assumed to be a random variable g...

11- Bnttle Behatior

a number of memor bwkling,

.med by a normaldist rib.tion

k.si (248,04 MP.)

with mean val.-

36

bers faifs either in tension, mmprmsion O C&raf

foroyt,c=. and x2 EIk/(L~A~)

for

force will be assumed to a@, in t,he dirm.

OBb, with variousdues ofth...effi.ient of,=i.bbn. (h) hI order to evahte the conditimal described probability of fail-

members asisconsidered in ticm of theaxesofthese CW

I (see Fig, 4b). The external forces F; (due to The stresses i

ure Pf (w), the pmcedum

i tbe next sec-

waves) remain to act at M the odes.

tion is used. At this point, it should be noted that tbe brmch and bondig operatims appearing originally

the members which are still intact nmst therefore be rc-eval. ated under these loading c.mditions. proces. of r~eval.ati.n, ID this

in the work of Murotw certain approximations tain circumstances.

and Okada (1981) contained that may not be valid in cerof this paper to of this

those members wbic3 have to the mnztrmtion of the

failed do mot mtribte stiff mess matrix.

It is the p.rpme

offer a more detailed and accurate presetatim W, the conditioned probis comp.t ed. Thk

(e) For a number of wi.dep.eds

method, thm improving tbe origin d developmmt. (i) FIndlY, the unconditional probability of fail.,.

ability of failure of the etmctnze

Pr or

147

the structure isevaluateds a

where SG~,,~) = probabi~iiy that ~emher 1 fails after mem. ber i fails first and member k fails second. G~,kl ~ ~iven the fajlure .f ~embers i and k, the

while all other memprobability thatrnernber 1fails RELIABILITY TURES (W), Similar definitions apply to SG~ SG~,k@), ,tc, ANALYSIS OF REDUNDANT STRUCbers (except members i, k, 1) survive.

The probability oftheevent G:) thatrnemheri will

fail first, while no other members have failed is

At this point, it should be noted that the probability F~~ is .On(ltiond due to the fact that member k b= .1-

readysurvived f) U (i.e., stress ofmember k intb.int~t atru.tre). orderto compute,in 3ener.1 In the prob.~,hky~f,,i,,...,il (i.e., ~suming a sequential failure of
where F(o) = probability F.(o) = probability of failure of member i of s.rvivd of member j (30) (31)

members i,+ i,used =k

- i.)thefollowing expression is

p(.2<...l_<f.

<...<.m)
for condition (.) for mditior(@)

n = number of members in the structure The probability SG~)

(32)

~.[,1....,kl . ~pi.1 ~ ~ @fl,...,l.l _

of the event that member k

fails after member i h= failed first, while .11 tbe remaining members survive is where

for coalition (c)

Condition (a) : ~fJ18...:t) ~;xs,;.1

where G:) = ~:) fi F;(.)

u: ,-)

= maz{lu~[,,..,.,),

> i)} (,,~)l < {IT, 10: -)1 :

Condition
(34) ;=, i#i,b
k

(b) :

(..,....,8) C:,,) o) >

Ccmditim (c) : {u: n-1 ,,1 [G..., < ~) i,) u,

with for all 1 = 1,..., F~l = probability that member k will fail under reand d,stributim of the load immediately after tbe event }f,,..., ).) = ~,ob{cf . .....l > ~y} (37) which

that member i and only member i has failed. and ~~(,1 = probahifi~y that member k will survive ..der redistribution of the lm.d immediately after tbe event Consider now a set of I+ members (r,, rz, . . . , n,) are identified as combination sequence (out of . possible

p and whom failures in any

k,! sequences) will produce

that member i and only member i has failed. Frthemnore

system failure. Assme further that there are a t.atd of m scb combinations of mernbem. Let rP,i (j = 1,2, ..., k,) q when

identify the member which fails j-th in seqnmce combi atim p is considered.

For example, supposethere

isa truss consisting of.=4 members;all these members

148

areidentifieda member ID number 1,2,3 and 4. Smp. by po,ef.rhb.,, thefailure tbak oftwomembers will produce
collapsein this truss. This tn.... that there are ,C, = 6

all powible mns.rning.

failure modes,

and thus is .SU ally t..

tirne-

An alternative upper bound is evaluated hy

P,,, = ~
cmnb,nati.ns of rnembem whose failure will result in ..1. lapse. Tbe first combination

Prob{EFSP,}

(43)

consisting of the two memID number p=l,

where ~

signifies mmnmatiom over the selected dorni ant

bers 1 and 2 is identified by combination and nince this combination The second combination of the two members

modes of fail . . . We assume that there are m of them. The lower bound of failme probability puted as can be mm-

involves two members, k, = 2. is identified by p=2, consisting

1 and 3 and for this mmbin ation,

PrL = maz(P,,)Prob{EFSpq} where the mmim.m

q that me examined, lar failure rnecbmism member failures.

(44)

k% = 2, and S. on. Al.., k,! = 2! = 2 sequmces ers

foreachcombination, there are

-according to which the two membFor instance, if mm. is taken over .11 the pairs of p .md and ewh pair repmsmts and a particular sequetid a particupath of

of that mrnbimmion can fail. ID umber p=l, q=l

bination

identifies the sequence of

failure 1 b 2, while ifq=2, r,, I = 2 ad

the sequence 2 + 1. Therefore, Followirg Mumtsu md Okada (1981), operations, three steps,

example. mm = 1 inthis forthe evmt that member i fsils folof (j-l) other members, which comist of branching upper and lower am comsid.

EF~] Writing lowing the fail..

bound adjustments

and bodin~

operations

reed, i order to evaluate Ptu and PfG. Step 1 : Brmmhig Using the notation itrod.ced above, we note that mant Modes A combimatim of members and their particlm failOperations - Selection of Dmni-

Pmb{EFSP,}

(.l,,,.% ....<...,,,.,l) = SG,,,,,

of a red.dmt

(39)
me sequence, or a pair of p and q, are selected m that

The failure probability now be written as

,

structure can

it yields a faihne mechanism with the l.srgest [stmct.ral) failure probability among pmaihle pairs of p and q. such that To

k,!

P,=

Prob{ IJ U ,=1 ,=,

EFS,,}

(40)

this end, first identify member r,,l

The upper bound of the failure probability be can

Prob{EFj:!,

} = G:]

= maz[,,c,.,]Prob{

EF/,]}

(45)

computed either by

wbem Prob{ EF/,l ]} = G!) is theprobability

that mem-

.
P,u = ~

Prob{EF/)}

(41)

ber i, and CAY member il will fail under the prescribed loading cond,tim. Hence, the probability that .11 other

,= , or by _ Pf,, = ~
,=,

,,, ~ Prob{EFSP,}
,=,

(42)

members will remain intact mmd be take into consideratim in evaluating this probability. Set I., consists of

These two formulations st dy, however, becawe

are not used in the present

for those rnernber. which Prob{EF/,l))

= Gf]

are larger

the first is too conservat iv. with.

than a certainrescribed p value.

Then, identify .,,2 s.cb that

out discriminating betweenreddantand non- red.ndant strctres the secondrequires and en.rneratio of

Fr.b{EFj::,}

= Gfi,;lJ = maz[i,=,.,lProb{

EFf:)} (46)

lowed.

First, consider members ik,,, in tbe set I.,n, exof (=

cePt r,,,, ko,, ad examine if k,,, members consisting ,,,,,, (= r,,,d,,), rP,d2 (= rp,fgw)j , rp,k-,.$, , rp,dtb,.-,) represent

where Pr.b{EF\~

]} = G:{]

i, the probability of failure

of member i, and only mernb w i, after tbe redistribution of itemal forces immediately Set 1., mbseq.ent to the failure

,P<<,f,[k, r.

<-l))>,P,,k.<,

a fail-

comb, ation p in a particular

,eq.en.e

q, where associated

of member r,,,.

m.sists those members i, for of

k,,, > k,,.

Tbe strctral failure probability

whichProb{EFj,2]}
Proceeding r. occurs

.I, greater than a prescribed value.

with th,s particulw only when

pair of p and q will be evaluated

similmly and examining if structural fail-

at tbe end of each member failure, the pm-

Prob{EFr,,,,,,,i
tic.lar combination p and particular sequence q which together with structural ~h~re ib,, c 1,,,,

, , } > 10-P,L

(52)

produce structural failure will be idmtified 1 mt~ I.,,,,,...,

but ik,, # r,,,, i,,.

If I@

W is satisfied,

L, ,. Then, the corresponding . in k,,

then the combination failure mode.

p and q repreaer.ta a dominant

failure probability

Tbe above procedure is repeated until all tbe possible

Prob{EISP,,,}

= Prob ~ {EF\~j,,,} j=,

(47)
paim of P md q =e exhaus~d. Not% b~ever, that U. 52

Step 2:

Adjustments

of Upper and Lower Bounds

limits to a minimum tbe number of failure modes for which tbe stmctrrd Computer failure probabilities are to be computed. to implement

Define PfL and Pfu (1) es

codes have been developed procedures

J?.

= Pr. (1) = PrA{EF%+~,

(48) tbe analytical indicated above and numerical

and Pf (1) = Pf. Then, find tbe second combination (49)

examples have been worked ot using tbe structure sbmvn in Fig. 3 Tbe upper and lower bounds evalmted with the aid of tbe procedure described above are still conditional to

p and q with

the system failure probability Pf,, (1) and Pfr (1) so that

Prob{EFSP,, Q-) and update

a specific wind speed and corresponding and be..

Ioad,ng cond,tion,

they should actually be denoted by Pfu (w) and

Pf (2) = Pf (1)+

Pf. (2)

(50)

Pffi (w), respectively,

1. the present study, we use Pfu (w) for Pf (w) in ap-

where PfL (2) = Prob{EFSp,,f,,}

and

proximation,
Pf. = ffL (2) (51)

since less complicated

nume,icd

ex~ple.

carried out ind,.ated m timesunparticularly

that these bounds

are quite close where tbe

ifPf= (2) z Pffi (1). Repeat the pmcedre

in the range of high windspeeds

til the dominant modes have been examined. Then all pr.(m)becOmestheupperbOund Oftbefailure probability oftbesystem. Step3: BoundingOperations To findp and q, the following procedure isfol-

condition d probability

values become more cmcial.

DEF1N1TION

OF STRUCTURAL

REDUNDANCY

There are a mber

of definitions for structural

re-

150

dundancy

ranging from that implied

by tbewell-known

tbefailure ofmember 1,twenty(20) out ofone hundred structures suffer will from tbe failure of member 2, thirty
(30) out of one hundred structures will suffer kom the failure of member 3, but all will survive. After the redistri-

degree structural of indeterminacy instructural analysis tothose listed below, assuggestedy Lloydand Claws.. b (1984)

bution ofloads intbeae structures which.wffemd fromthe RedundantFactor= intacttrength s intacttrength-damaged s e.trength failure ofo.oftheir embers and survived, it is assumed m
for illustrative purposes that the probability of survival of

those structures has been found to be 0.2, 0.8 and 0.1 for

Reserve Resistance Factor n environmental loadatcoUapse (undamaged) design environmental load

atmctures without members 1 or 2 or 3, respectively, i.e

1 G~] G!) = 0.8 Residual esistant R Factor D emvironrnental loadatcollapse (damaged) environmental loadatcollapse (undamaged)
1 G~l G!) = 0.1

(54)

The frequency interpretation

of Eqs. 54 now states that

two (2) out of the ten (10) structures which survived after

In the present study, however, attemptisnwade an

Tbe todefine heredundancy a probabilistic t by measure. definition introduced here uses, as a mesa.re of redun.

the first failure of member 1 will again survive the redistribution of the loads, etc.

Iv=lccl

dancy, the probability

P; that the structure will ewntu-

Failr. of members
.Ily survive, given the failure of one or more (but simul. taneously) of ita members. Usins the notation introduced in the previous section, the following example attempts pretation. of tbe red.ndmcy It is supposed to give a frequency interdefined above. N nominally Swvivc #1 /i)o=m 10 / #2 \ #3

20

I Redistribution of loads

I
16

probability

that there are initially

Swvive:
identical but statistically different k-member
structures. For illustrativeurposes, N is assumed to be one budmd p

Ill
2 Thus,

= 21

makes clear that twenty-one strct.res members,

(21) out of the sixty (60)

(N=

100) and k equal to three (k=

3), Furthermore, it is

which survived tbe first failure of one of their will eventually ( i.e., after redistribution to tbe definition of

assumed that tbe pmb ab,lit ies of failure G ~) of members i (i=

s due loading 1,2,3) oftheintacttructure totheinitial

are

loads) survive. d.ced

according

intrc-

mnfigmtim

pmviowdy, tbe redundancy probability

P; is 21/60

= 0.35. G~) = 0,1, G$) = 0,2 and

G~] = 0.3

(53)
Explicitly written, Eqs. 53 and 54 lead to

i of Eq. 53 states Tbe frequency nterpretation thatt.

(10) out of me hundred (100) strmtmes will suffer from

151

NUMERICAL

EXAMPLE

The Pierson-Mmkowitz

spectrum

given in Eq.

6 is p =

used for wa surface elevation 0.74 and g = 386.4 in/=2

II with a = 0.00s1,

(9.81 m/sec2).

The spectral

density beyond w. = 24 rad/sec is disregarded as insignificant. The deep water assumption is used with d = 3W ft (91.5 m) in Eq. 1. This asmmptirm resulted in a simpler (55) expreseio for the dependence of the horizontal on z of the spectral der.cmnponet, of the water

sity fctios The numerator of the second term on the right-hand side of Eq. 55 is nothing else but the probability without of failure of si-

particle velocity and acceleration. ~tonn i F+

The duration T of each

7 and 13 is asmrned to be four (4) hours.

of the structure, mltmeoun fail..

including tbe probability

While such a duration should also be considered random, the easmption of T being equal to four hom does d

of more than one member in the intact

structures which, in this example, would produce coil.pne of the atmctms. in order to include this event and to make tbe definition of tbe redundancy should be defined that probability completely gener~> it

appear to be wmee.scmable. In employing amnw Cm = the Morrison formula in Eq, 9.61 21, we x 10

1,5, CD =

1.0 and p =

kipnsec2/i4

(1.028 x 10-8 N.see;/m4),

The qrmtity spec-

w- in tbe frequency at whick tbe P,emon-Moskowitz tm.1 density takm a maximum v.1.

and in given by Eq. in/see a l.g-

p:=l ~ Gf) i=,

lf + SF(0)

=1P;

(56) I& For example, w(29,46 m/see). = 0.289 rad/sec for W = llM function of W b=

The dwtribution

where SF to) is the mmation

of the probabilities

of the

nonnal distribution

as shown in E.+ 22 where the constant

first ofmom thanonememevents ofsimultaneous faihre

which lead, however, h mrvival of the structure. ber, Eq. 56, p} h defined M the non-redund~cy In

C, is giv.n h M
(0.0205 sec/m).

23 ~d is .wJ
We =sum.

fA o.o@J5212=C21 in
= 2.842 and uk =

that pi

probabil-

0.1. These correspond (29.46 m/see) 0.1. and Vw

to IJ.W= E[W] = W = 1160 in/see (coefficient of variation of W) a

ity, i.e., the probability

that the structure till event.~b

collapse given the first failure of one or more (but simultaneously) of its members.

The truss considered The probability Pi in the numerical as example dbshown in Fig. .umed in this study can be witten p,=

for the numerical

example

is

3 and its geometrical

and other ch arac-

teristics are listed in Table 1. The material strengths UY,,

/-p;wfw@),w o

(57)

OY. and UBk are assumed to be normally distributed random variables as mentioned in the previous section. AIso,

inwhichP;(w) isthecondit. i alnon.reddancyrobap bility ofthestructure fully evelopedeastate ina d s repre aented a windspeedw. by

their alues v arestatistically independentrommember to f

member

152

The conditional probability Pf(w) (upper bound)is evaluatedorseveral f windspeeds: = 621 ir,/sec W (15.77 ml...),894 in[sec(22.71 m/see),1160 in/see(29.46 ITI/see), 1425 in,fsec (36.2rn/see), 698 in/see(43.13 1 rn/8ec), 1963injsec 49.86 /aec), ( m th.scovering tberan8e from PW -4 Ow to #w + 6 OW. The values of P,(w)

cially important

in the methodology

is the fact that it mmpoents

treats tbe structure as a system of strctral with the .ndemtand,ng

that failure of them components of the internal (and

in my number results in redistribution possibly external) forces.

at

Tbe external forces that act on the offshore structure stem from wind-induced under the assumption waves. These forces am evaluated of the small amplitude wave tbe.ry that the flow is irrotational

these wind velocities w for Cases I and II are plotted n i

Fig. 5.. The unconditional obtained Pf(w)fIv with tbe aid probability of Eq. of failure PF in then 28. The integmnd in Fig. 5b.

together with the assumption and inviscid. Tbe velocity horizontal

(w) for both @es

I and II appe~s

components

of the water

particle
is sub-

Comtrcting

such a plot, tbe values of Pf (w) other than AS

and acceleration

the offshore structure

those aheady computed are obtained by interpolation.

jected to during each storm are derived from the P&onMmkmvitz spectmrn for the sea srfaw elevation. III its

can be seen from the plots of P, (w) fw (w) m log-scale, the

contribution towards PF from lower windspeeds

is negli.

analytical form, the spectrum mtaim value representative l-t

a rnem windspeed

gible.

The final results indicate that PF = 0.24 x 10-$ 10-s to (Case II) for this truss, conditional on-red. dancy

of each storm, which is assumed to

(CaseI) and 0.88x

With probability in @. respect

four hours. Assuming further th-at the sea surface dis a G aumian random process, the expected maxof the horizontal water particle velocity and

the

e.ation

f; (w), the denominator

of the ne.ond term

imum vdms

56 is computed

at the same windspeed values an of PJ(w). The remits for

acceleration in their absolute values am .waluated with the

those used for the cornp.t.tim

theory. aid of random process

The atteation of these expected nmximm values For simpand a

P; (w) are shown for both Cams I and IIinTablez .nd am plotted in Fig.
6.. Fig. 6b displays the values of in computed u 0,99

along the water depth depends on the frequency. licity of analysis, this dependence

the integrand of Eq. 57, Fkm.lly, P;

is disregarded

X 102 forCase 1 and 0,93 x 101 for Case H and the mrrespodig wmond,tio al redundancy pmbabdity P;

form of negative expoentid the freqmmy cmnpoents,

attenatim Aim,

is med for .11 the effect

disregarding

is equal to 0.99 for Case I and 0.91 for Case 11.

of relative motion

between a water particle

and tbeoff-

structure, the Morrison eqatim shore CONCLUS1ONS

is used to mrnpte tak,ng

m the external forces acting on the truss embers,

into mmiderat ion the angle of icli ation of these rnemAn armly tical method, .merical pter codes are developed procedure and mmof bem with respect to tbe vertical direction and transform. ing these distributed forces into ccmcmtr.ted on the odes of the tress. The probability of structural failure k evaluated by seforces acting

to evaluate the prob-ablity

strctrd

failure, While the stmctre tbe methodology

considered is a fixed is general enough to Cru.

offshore strct.re, be extended

to other types of offshore ,tructurm,

exmninin g on] y a limited number of higher-pmbability

153

quencesof member ftilures thatproducecollapse of the

structute. The probability analysis involved could become

6. Morrison, J.R, et

al.,

1950.

The

Force Exerted

by Surface Waves m P,les~ AlME, 7. Mumts, Vol. 189.

Petroleum !hmsc.ctions,

quite time-consuming

if all the possible sequences leading and if the strctre component.. not only beof structural

to structural collapse were comider.d m.sists of a large umber of structml

T. and Okada, H., 1981. Strmtre,n

Reliability

As.

semment of a Redundant

Proceeding. of

The probability

analyais is important

the S+d ICOSSA R, llodheim,

8. Palio, C. and Shinozka,

Norway, pp. 215-329. ad

cause it will provide us with tbe probability

M., 1986. Reliability Techiad

failure but also because it will make it pmsible to define tbe redundancy probabilistically, ab,lity of structml of the members. REFERENCES for example, = tbe prob.

Durability No.

of Marine Stmctureg,

Report Depart-

1 under ASS Grant No. CU02396501,

survival give the first failme of one

rnmt of Civil En~ieering ics,

ad Engineering Mechan-

Colrnbia University.

9. Pierson, W.J., Jr. and Mcakowitx, L.A., 19S4. A

Proposed Spectral Form for Fully Developed Wind

1. Cmm&,

H. and Leadbetter,

M. R., 1958. Stationary

Sems Bawd on the Similarity Tbecmy of Kite.i,prodskii, Jomml

and Related Stochastic Procea.ee, John Wiley & Sons,

Inc., New York. 2. Ellingwood, B. et al., 19S0. Development for American of a N&

of Geoph@el

Rm,mh,

Vol. 69, No.

24, pp. 5181-5193. 10, YmE, J-N,, 1978. Statiatiml Fatigue

Analysis

of

Pmbability-B*ed tional Standard

Load Criterion A58; Building

Deep Offshore Strct.res

Under Storm Wavea,n Tech

Code Reqiremets

nical Report No. 2 under NSF Grant No. ENV-7508895, Tbe George Washtgt,on University, Waahig. ton, Dc.

for Minimum Structures, 3. Faulkner,

Design Loads in Buildings and Other NBS Special Publication 577.

11. Yang, J-N. ad Fzeudenthal, A. M., 1977. D., Shimmka, M,, Fiebmndt, R,R, and ity As-meat Fmnck, I. C,, Eds., 1984, The Bole of Design, hu-pecof Offihore Platforms

Reliabil-

in Seismic Re-

gions, NSF Technical Report, School of Engineering tim, and Redundoncg i Marine .%uctuml Reliabiland Applied ity, Committee on Marine Structures, Marine E%r,rd, versity, WashiEton National Research Council, Waebingtcm, DC. 4. Hkkmm, Guide; tic J.W. et al, 1983, PRA Procedmes of ProbabilisPower Plants, DC. Science, The George Washington Uni.

A Guide to tbe Performance Assessment for Nuclear

Risk

NfJREG/CR23C0,

F,md Report. and Plat-

5. Lloyd, J.R. and Cbnwson, W. C., 1984. Reserve F.esidwd forms, Strength of Pile Founded, 05hore

The Role of Design, Znspeetioti, and RedunStrtwtwml RdimWity, NRC, Wasb-

dancy in Mcrine ingtin,

DC, pp. 157-198,

154

TA BIJ3 1, Geometrical and Material Characteristics

ofstructural C.rnpanents

F
T
2 .7 4 5 0 7 8 9 10 11 12 19 14 [5 16 17 !8

,.h @
({,)

D;a,,wtw (i) 50,1 sol M.1 54.1 50.1 50.1 50.1 60.1 45.0 45,0 45.0 36.0 S6.o 38.0 36.0 .98.0 96,0 27.0

rhicke= (in) 2.0 2,0 2,0 2,0 2.0 2.0 2.0 2.0 1.0 1.0 1.0 0.75 0.75 0.75 0.75 0.75 0.75 1.25

Area (@)

Inertia Mom, (i.,)

105.5 80,4 70,3 45.2 ,05.5 m.4 70.3 45,2 120.0 151.7 151.7 121.2 121.2 103.3 103.3 78,6 78.8 EJ2.o

Modulus of Elasticity = 290W ksi UYC>UY, = 36 km

T
302.2 302.2 302.2 302.2 802.2 302.2 302.2 S02.2 138.2 198.2 198.2 83.1 83.1 8.?.1 83.1 83.1 83.1 101.1 87564 87554 8,554 87554 8,554 87554 875.5, 87554 934f,9 .33489 33469 129C6 12QG6 I2W3 ,2306 129W ,2X6 8401 ;

W E 1160in/,ec (Mean Windspeed)

L
05

10

1.5
u (r.ad/see)

Note : 1 ft= 0.3048 m ; 1 h = 25,4 m 1 ksi = 6,89 MPa.

FIG. I One-Sided Spectralensity D ofOcean Wave El.vation,

TABLE z. Conditional Probabilities and P;(w) Due Pf (w) to Various Win&peed Values

~ >

~3,

E h(k&c)
621.22 894.51 1159.78 1425.01 1698.30 1963.58 i=;(w) c&se 1

Windsped

PAW) ca8e 1

0591 .10-, 0.553 10-4 . 0,807.10-X 0.12.5 .10-1 0,141 0.624

0.366

A 900 1(CO T!OI 1200 1303 ILOI 15c0

P;(w) cue 11 0,244.104 0.171 .10-, 0.641 .10-1 0.335 0.997 1.000

PI (w)

ExpectedMaximum Storm WindspeedW (i/kec.)

casen

0.523.10S 0.359 10-4 . 0.420 10-3 . 0,216 0.217 0.587

0.144 10-0 . 0.947.107 0.517 10-4 . 0.419 10-2 . 0.141 0.624

m G. z Probability Density Function ofExpected Maxi. mum Storm W indspeed.

Not. : 1 in= 25.4 nun

155

EL. 0< +

EL. 0,

-g-F

60 $5

,+

-t--

EL. 45< 4 -4. 4 A

EL. 05,

3 ~ 3 .3 J.

+=5
12

2 EL.195,

2 2

2 2

EL.195,

1 w + 1209

EL.300,

1 L 120, .

EL.300,

PIG. 3 Intact Structure.

FIG... Loading ConditionSubsequentto Failure of Member 13 (CaseII Brittk Behmio. r).

FIG. 6. Condition Probability Failure of Pf(w)

,s

Windspeed. FIG. . . Lo&ding Condition Subsequent to Failure of

Member 13 (Case I - Ductile

Behovi..r).

,,.,
,,
,,.,

r!-.

Case II

Wind Speed (in/se.)

!@ !0-, ,.,

Cese 1 \

FIG. 6.

lntegrand,(w) fw (w) m. Windspeed. P

,,., ,..,. cam 1: P; = 0.99.10-2

,..!) ,,. !,
Case11,F; = 0.93x 10-1 , @ r

Wind Speed (in/see)

FIG.eb Integrand P; (W)fw (w) vs. Windspeed.

.
wind Speed (in/se.)

FIG.6. Conditional robability Failure (w) ,s, P of P; Windspeed.

157