Académique Documents
Professionnel Documents
Culture Documents
@!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.,,
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.
1 fact, a mcmt
and Redundant.yz
by the Marim
Structures
1983, dis.
the gemral
framework
and probability-based
Pr.h.bility-b=.d theoretical
d=iw
and mm.
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
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
143
Pr-dures
must be d-bed
systems.
for .fih.r.
.,
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)
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
for estimating
, (w) = wSqq(w)e~
(5)
redundancy
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
EXPECTED AND
MAXIMUM
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,
sea condition.
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
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
maxirrmm value of
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-
values of them POW.. spectral density functions in the form of a negative exponential water depth -z icre U.S.
deczease
In Eq.
8, T is the duration
of the storm,
pr.c=s
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
.:=
J-so(w,&
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 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
proce~
f(.)
(16)
= cd
~v(z)
+ ~CDPWZ)l~(Z)l
(20)
0;=
~-sv(w,*
== Aoq(w)e%
analytical
(17)
D = pilediameter,
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.
tion are of tbe same order of magnitude wat.. part ides. It is geuerally accepted
145
..
of tK,. interaction
can be
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)
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
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)
in which w is meammd
i in/see,
p>
and ok
represent
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-
abilityanalysis procedure, this approximation of the load configuration serves as an illustration. Hence,
For tbe NorthSea,theuseof& (23)
f(z) = CmP ~
+
(!31)
, + : 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-
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.
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)
(27)
146
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
windspeed is computed. (c) Having computed tbe values of the expected maxi.
failure of It
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
cmzses from zero to a certain level, the the probability of simultamous failure of two or more members
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-
h.vioc
failures can take place in those members whose are less than the intern al forces resulting
tension, compression
strengths
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...
36
for
OBb, with variousdues ofth...effi.ient of,=i.bbn. (h) hI order to evahte the conditimal described probability of fail-
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
and Okada (1981) contained that may not be valid in cerof this paper to of this
It is the p.rpme
offer a more detailed and accurate presetatim W, the conditioned probis comp.t ed. Thk
method, thm improving tbe origin d developmmt. (i) FIndlY, the unconditional probability of fail.,.
Pr or
147
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.
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)
- i.)thefollowing expression is
p(.2<...l_<f.
<...<.m)
for condition (.) for mditior(@)
(32)
fails after member i h= failed first, while .11 tbe remaining members survive is where
u: ,-)
= maz{lu~[,,..,.,),
Condition
(34) ;=, i#i,b
k
(b) :
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
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.
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
failure modes,
tirne-
P,,, = ~
cmnb,nati.ns of rnembem whose failure will result in ..1. lapse. Tbe first combination
Prob{EFSP,}
(43)
where ~
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-
(44)
bination
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.
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,}
(39)
me sequence, or a pair of p and q, are selected m that
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,=
EFS,,}
(40)
Prob{EFj:!,
} = G:]
= maz[,,c,.,]Prob{
EF/,]}
(45)
computed either by
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
= Gf]
are larger
out discriminating betweenreddantand non- red.ndant strctres the secondrequires and en.rneratio of
Fr.b{EFj::,}
= Gfi,;lJ = maz[i,=,.,lProb{
EFf:)} (46)
lowed.
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:{]
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-
,eq.en.e
q, where associated
of member r,,,.
whichProb{EFj,2]}
Proceeding r. occurs
Prob{EFr,,,,,,,i
tic.lar combination p and particular sequence q which together with structural ~h~re ib,, c 1,,,,
, , } > 10-P,L
(52)
If I@
W is satisfied,
failure probability
Prob{EISP,,,}
(47)
paim of P md q =e exhaus~d. Not% b~ever, that U. 52
Step 2:
Adjustments
limits to a minimum tbe number of failure modes for which tbe stmctrrd Computer failure probabilities are to be computed. to implement
J?.
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
Ioad,ng cond,tion,
Pf (2) = Pf (1)+
Pf. (2)
(50)
and
proximation,
Pf. = ffL (2) (51)
nume,icd
ex~ple.
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
DEF1N1TION
OF STRUCTURAL
REDUNDANCY
re-
150
dundancy
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
1 G~] G!) = 0.8 Residual esistant R Factor D emvironrnental loadatcollapse (damaged) environmental loadatcollapse (undamaged)
1 G~l G!) = 0.1
(54)
two (2) out of the ten (10) structures which survived after
the first failure of member 1 will again survive the redistribution of the loads, etc.
Iv=lccl
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
Swvive:
identical but statistically different k-member
structures. For illustrativeurposes, N is assumed to be one budmd p
Ill
2 Thus,
= 21
(N=
3), Furthermore, it is
which survived tbe first failure of one of their will eventually ( i.e., after redistribution to tbe definition of
according
intrc-
mnfigmtim
P; is 21/60
G~] = 0.3
(53)
Explicitly written, Eqs. 53 and 54 lead to
151
NUMERICAL
EXAMPLE
The Pierson-Mmkowitz
spectrum
given in Eq.
6 is p =
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-
While such a duration should also be considered random, the easmption of T being equal to four hom does d
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
w- in tbe frequency at whick tbe P,emon-Moskowitz tm.1 density takm a maximum v.1.
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
of the probabilities
of the
nonnal distribution
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-
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,=
example
is
/-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
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
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
I and II appe~s
components
of the water
particle
is sub-
Comtrcting
and acceleration
jected to during each storm are derived from the P&onMmkmvitz spectmrn for the sea srfaw elevation. III its
is negli.
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
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
imum vdms
56 is computed
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
is disregarded
X 102 forCase 1 and 0,93 x 101 for Case H and the mrrespodig wmond,tio al redundancy pmbabdity P;
attenatim Aim,
disregarding
of relative motion
and tbeoff-
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
strctrd
153
6. Morrison, J.R, et
al.,
1950.
The
Force Exerted
Petroleum !hmsc.ctions,
quite time-consuming
if all the possible sequences leading and if the strctre component.. not only beof structural
Reliability
As.
semment of a Redundant
Proceeding. of
The probability
analyais is important
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-
ad Engineering Mechan-
Colrnbia University.
1. Cmm&,
H. and Leadbetter,
of Geoph@el
Rm,mh,
of
Code Reqiremets
nical Report No. 2 under NSF Grant No. ENV-7508895, Tbe George Washtgt,on University, Waahig. ton, Dc.
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.
Risk
NfJREG/CR23C0,
5. Lloyd, J.R. and Cbnwson, W. C., 1984. Reserve F.esidwd forms, Strength of Pile Founded, 05hore
154
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 (@)
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
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 ;
L
05
10
1.5
u (r.ad/see)
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
0.366
PI (w)
casen
155
EL. 0< +
EL. 0,
-g-F
60 $5
,+
-t--
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,
,s
Behovi..r).
,,.,
,,
,,.,
r!-.
Case II
!@ !0-, ,.,
Cese 1 \
FIG. 6.
,..!) ,,. !,
Case11,F; = 0.93x 10-1 , @ r
.
wind Speed (in/se.)
157