Академический Документы
Профессиональный Документы
Культура Документы
{.}r . finE
"
lmnct$sr friovac:
H.SUANIA
$iIf${Ar.Y
the n*cessrt,y o{
expem"ses
ot
am.d
specifi*
s&mrees
of ernor
j.n
slt
l"fue
rai
teg:hntcal areatj.on
&'"s iiL [h* r.taval d*rrlaire it is nelt al-ways easy to observe the neanity, are
equ;lil"y used the n'sca-18 ph],s:.sal mod&Ls" and the "aumetricatr models", hotfi as
si.*lp-li..ti.*at,i*ns ot ttrhe reali.tv, umderLi_ni.$ry " certaim properties of the rea} or
neql"ei:t,in$ o'i:frers, trnetrovarat for ttle pr.oblecil under consideratien.
I[fie s#a {e $hv.si"ca]. modeS.s perrnit the di"rect observati.on of tbe phenemema the
s!:.ip *'rvsteil]" is suhrnit,ted"to" haviuq a repli.eative or deseri.pti.ve eharacter, whi.le
Ltne nmmer:ilaal relei,i$el.s perm"tt, Lhe es{imatiom o{ the syste*ltos e$oLution im time.domain,
ltavinq a pnedictive cha::aryLer.
8esi.eles ptlr* ph.oysieal and numerj"eal models are also used hybrj.d modets, i.e.
nm.meri.cal" ruodels, based on rnathematieal descript,i"on of the physi.eaL phenomsrra, in
whieh are i.nt,r*duced experimentally measured data (as coefficients of equatio*s).
ilonsequ*m,t,ly, i.t inslde eaeh type of model is raiscd the problern of cliecki.nE
ttre pr'eclsaon {r:.t nroitr*.i-lialE} and uncertai.ntv anaLysis, betweeu the types o{ nrorl*}s
c:f t-he same clegr*e, aJ"t beinq submi"tted to the abjestive reaLity, is raised the
$r*b{em of, "parttal. validatisa"" whi}e rel"ated to this reality is raised ttre prmb}.enr
of "g-Lof;a.[ va.l-idationoo.These procedrares of parti.atr- and stroba.l" rra]idati n offer
Con
ceptual
Partiat validation
Numerical
Physi ca I
model
scate
-_..j!
FI6URE
1a
odel
-;;* - --,r..
modet
Different kinds
their
FU
LL SCAL E REALITY
( REAt S HIP )
6EOMETRICAL MODEL
STR U CTURAL
CONCEPTUAL MODEL
PHYSICAL STALE
MODE L
REPRESE NTATION
P
ERRO RS
A
R
TESTIN6 ERRORS
(MEASUREMENT
L
T
I
MODEL IESIS
DISCRETIZATION
E
ANALYTICAL
SOLUTION
RRORS
T
I
vl
DISCRETIZED
rl
FI
RRORS
COMPUIER
ERRORS
TEST
RESULT
iltl
CODE
A
E
POST-PROCESSING
EI
RI
PRO6 RAMMIN6
MATHEMATITAL MODEL
0i
DATA ANALYSIS
T
E
RROR S
STALE
TF
FINAL
Ni
ECTS
EXPERIMENTAL
RESU LTS
C
0
M
NUMERI CAL
INACCURACIES
ERRORS
s
ANALYTICAL RESULTS
DATA ANALYSIS
E
OMPARISONS
RRORS
WITFI:
OIHERNUMERICAL
RESULTS
EX
FU
LL SCALE FINAL
RESULTS
PERIMENI,\L
RESULTS
NUMERICAL ( C.F.
RE SU
D.
BENCHMARKSI
LTS
FIGURE 1b
calculations);
- problens of eonvergetrce for large aud complex equation systens, assoeiated to the
matrixes of added mass and darping coeffieie[ts i
- irregular frequeacies on eonrputations of these hydrodFramic quantities ;
- the number of "strip" under consideration and the number of polnts, describing
the shape ot each section (2I)) aud respeetively, the arnber and distribution of
the panels. over the hulf (3DI ;
b) Phy-s"_i-c_il gmpr"qr.*gsgsg-e!_A1!- hv :
- the effeets of the linear poteotial floy {the systen of non-statiotrary flaves,
generated by the hull's oseillation, interaction vith the stationary }ocal
potential around the ship, the three-dineasionality of the flor, etc.) i
- the etfects of the non-linear potential flov (the linearlzation of the condition
on the free surfaee, of the eondition on the hull aad the exciting foree
expressions) i
- vlseous effeets i
- the uncertainties in the rodellXng of the rave speetrum, sarpling choice of the
frequency domain of the irregular yaves and unproper rodelling of the vaves
propagation.
The existenee of a such large range
fig.lb, three cateqCIries of actions :
of
sourees
r) ghe"-c}l!.s -a{_ :
- the correetress ot the eontiuuous aad discretizeil rathenatical model trm the
point of vlev of the boundrry and initial condltions ;
the
eoherenee and eoaslstsncy of the eonp[ter code tthe renoval of the progranming
errors and the eorreet pertornance of the eonaectiona betree$ rodulee) i
lI )
cq"r3Bfl[]-i;q&g--[9- :
and 1800).
navlgates
?his vaY of, representatlon is also adequate for the irplerentation on the
board computers aeelating the captain to selCct the best shipplng roui" teipert
navigatiou systen), as it shovs very clearly the tendency of tnese eurvos depending
on the heading angle and shlp Bpeed.
For illustration in fig.?a, b, c is represented the evolution of the pitehing
angle'lh"-and-in fig.3a, b, e the evolution of the relative notior
ihe tore
perpendieular (the signless eurves belng the Ronanian results aad tbose, at
rarked sith
signs the Russian results).
The analysis of thase diagrams, the qualitative and quantitative co*parison
of the values and the calculatlon of tha systeaatic error pertit ur to foraulate
the
tolloring concluslo&f :
a) tor the heading angles of practical interest, respactively B'=lg0c-135" (doninant
head vaves) and p-o"-45o (dominant tolloving yaves) the results are eategorically
in the sane domaias, that reans that the partial nutual valldation of the tno
aspect, vith rore larked peaks, rshile the Russian curveg are .,snoother*.
ars dre to the tact that rhlle the tomanian eorputer prograir
yorked vith unidirectional ISSC irregular vave*, having the vholeenergi
eoneentrated on their directisn of propegation, the f,ussian cmputer prograr
rorked vith multidireetlonal vaves, having an energy spreading funetion of ttre
Theae differences
validation
In aeeordanee vith the general validation sehene in fig.lb and vith the
assertions in the previous ehapter, it results that the globat valiAatlon procedure
ean be perforned only in relation rlth the absolute obiective reatitf, in the
measure in vhieb it is knoyn and observable
The internediate stage of the nodel tests, affected by nany sources of error,
permit in its turn only a partial validation either of the uuneiieat results
or of
its orn experimental results in relation to the full scale measurenents.on the other
hand, even the results obtained by full scale experinents are *ffected by errors,
introdueing'a eertain degree of uncertalnty.consequently, the global vitidation
representc tho direct colparison of the tman yalueo, possibly true ln their om
confidence interval, obtalned in different vays as the fereat sourees of error
belong to the full seale experiments, thea are consldered referential, being also,
qualltatlvely and quantitativoly, tbe ',nearest'. to the objective reality.
tretluency domai.n tor pitching in regular head raves and applying the seeond relation
l,lrener-ttruci.n. hy means of a Fourier B&( analyzer lras determined the demanded wave
eneounter speetrum, dividing the pitching speetrum measured at the ship (the real
pitehinu) by the square of the pitehinc response function (measured on the model and
considered valid at the ship).?he calculatisas led to a signlficant height of the
expression.
Uader these eonditions, the four series of results nere reBresented in one and
the same sytrthesis diagram (fig.4) in vhich, besides the tvo curves, obtalned
numerically (CFD) ras also represented the absolute experimental eurve, reeorded at
the full scale, respeetively the senl-experinental curve, obtained in the end of the
year l98B by model tests and transposed to the real ship, each of then vith its
confidence interval.
For transverse yaves do not exist experimontal points for pitching, being,
effectively recorded only rolling, rhat means that the disturbing vave Yas a full
developed sea, being more like a unidirectional vave thar like a rnultidireetional
one and produeing very reduced or non-existent seeondary responses.Arryvay, for the
four heading angles for vhich exists the possibility of comparison can be remarked
a good grouping of the mean values, especially in head vaves.The analysis of these
values anil of the confidenee intervals, both ln relative and absolute setrse, led to
the validation table 3, aceording to vhieh the global validation can be considered
as satisfactory, both the tendencies of evolutlon of the caleulated values related
to the measured ones and the existenee domains of the solutions, i.e. the confidence
eoneordanee.
4. Conclusions
a)
b)
c)
The complexity of the phenomena and proeesses to rhich the ship "systen' is
exposed in irregular vaves inposes its research by means of different klnds of
models-eonceptual, physical, aumerieal (CfD), between rhich nust be performed
partial and global validation ;
Before their perfornanee for eaeh model is neeessary a serious analysis of the
sources of errors, their quantification, the determination of the sensitivity
to errors and the determination of the final eonfidenee intervals, resultinq from
"global validations" i
d) The numerical and experinental results, obtained for a container ship of 8000tdv,
indicate that the numerical nodels rhich vere used are partlally and globally
validated ln a satisfactory Yay ;
e) The insuffi.clency of the data, obtained at full seale, re$arding the behaviour
of the ship "system" in irregular vaves imposes in the future the focusing
of the attention on the performanee of seakeeping tests as eomplete anil precise
as possible.
5. Ieferenees
lll
"Predicting the
tests or computations ?
G.VAN OORTIIIERSSBN,
CFD
and
CAD
in
hydrodynamic performarce
in ship design
pp. 233-245
121 I.NOVAC, "Cercetari asupra
t3l
"Report
Proceedings
of the 19th
TABLE
DENOMINATION
LENGTH BETWEEN
PERPENDICULARS
CONF.
PROJ
E CT
BREADTH
H EI
6HT
DRAUGTH ( FORE
DRAU6TH ( AFT
ON BOARD
6, 318
DRAU6TH (MEDiUM
MEASURED
3,852
BISPLACEM ENT
79
TORRETTED METACENTRIC
45,4
1,7 66
H EI GHT
TRANSVERSE 6YRADIUS IN
ROLL DIRECTION
0,410
LONGITUDINAL 6YRADIUS IN
PITC
DIRECTIO N
0,253
SHIP
MEASURED
ON BOARD
SPEED
TABLE
ISSC
C
WAVE SPECTRUM
SI6NIFICANT
ODAL
PERIOD
DIMENSIONLESS
MEAN
SYMBOL
WAVE
HEl6HT
M
PERIOD
DIMENSIONLESS
MEAN WAVE LENGTH
1t3
Z=TE
Y'
908
MU
508
lr,4
6,3
8,9
1...1,5
1,5.. .2
>2
..
>0,5
0,25
70
0,25
.0,5
WAVE
DESIRIPTION
SHORT
EDIUM
LO NG
TABLE 3
-------_8ELATIVE
Item
sH tP- WAVE
Pitch measured on
lship triats): \;
board
3,63o.
oh
Uncertainty interva[ :
It
Uncertainly interval:
Pith f rom
C.F.
resutts, L
resu Its
z!
D. romanian
C.E
Uncerlainty interva[
11
Re la
tive bias
-:q/[t r
Bias
Re [a
error:
Bias
17
18
19
20
100 .r"
tli - rUF
13
16
rpop
(tI-n)/rLIx
15
: !Bo/o
- \E
tY,
12
1L
7 oh
tpor:
l\!-:[Ftt:|Gx
Bias error
0,1500
3,1+2o
!0,1370
0,120
3,2
oA
0,27
2,97o
3,220
r 0,057
3,10
r0,0640
3,570
t0,1240 r 0,1434
3,45o
!0,213
!0,21+2o
3,530
2,9 60
3,380
0,2820
0,2370
0,27 00
0,29o
0,230
0, 35o
8,49"/o
7,1+2oh
9,8
0,49o
0,19
13,510/o
5,9 oA
!0,2730 I 0,2530
0,305
0o
3,05 o
3,620
3,92"
450
o/o
0,25
, (ef- nf
/rL, x
6,920/"
0
0,40
4,g7oA 11,33'h
0,09
6,67 "
0
,160
3,0 4"/o
4,73oA
0,20
0,0 5o
0,120
3,85o4
5,52Vo
1,64oA
3/+8Y"
0, 070
0,110
1,g3oh
3,12o/o
4,73"h
5,62o/"
0,09o
0,0 go
0,09o
0,07o
2,09% 2,55"A
3,04oA
2,07 "/"
0,15
0,14
0,19
100./.
ttI-tt?l
0,19
iiElft,7,},',.'i'oi ,"
Bias ePror, I l!- ,tlt
l:n{ -
ger/s'
900
1oo.h
error: tr{-rtil
Retative bias
3,130
r0,0730 !0,0620
3,90o
rt?e.
10
4./"
D. russian
Uncentainty interval
Pith f rom
135 0
1800
DENoMtNATtoNNTtoNS
100 "/.
a
J
a
J
sE,
4 Fd.
o oE
E t/tl
tr,
G
ro
+L
.g
E
IE
L
.o
L
o
E
<u
=
o
@
(U
E
L
l-l
0l
o
EL
o
c
o
L
E
+
'a :
+
+ti
o
+o,
+o
o,
-,o
o
(U
.E
z.
-t
=
fo
o
tA
c;
:o- iI
+U
TL
t-;
==o
ci
TL
L;
I
I
t
I
I
,rl
:
I
UJI
ci
qr
LJI
,.1
Ett
,l
r-I
=t
el J
q
E I
a
5
(5
t*L
Itr
lr
I ta lz'
IB
IIrr.r
lr
kl=
lil olo
{lr__.r
aIE
HIE
c I , z.ltt
.l ;lE
colo
kt
---.r
ls
ld
\$
l{J
'tr
& =
aI
l+l
*\ \'\57
\-\
\
=
O
E =
E
q2
=
&
_o
a
a =
(U
co
E
U
1.
a o
t..\.
c) F
o-
$l$
6
lfJ
F
J
lr,
o
o
=
o
q
o
(U
I,
(U
a_
(J
.Ef,
t,r
=
E
.r-ol
\<
=
ol
=l
<I
zl
u-r
ol
E
ul t! =\
F
u-l
as:
sl
rnl =
o.I
*'.t
r\
-s
el
ol
e. l
el
rr-r
d"
IJ
F
=
t!
-l
<l
lrl
F
J
lrj
EI
ol
TJ
z.
lrl
s
o
rL G
lrl
z. F
o
(J z.
o
G, =
o
E g
I
&.
trl E,
vl F
-<
6
L,.)
nr
s\o
o\
I saalbap
*-z/,4F6
T
.[
>seJUqP j]
LJ
z.
cl
lrJ
o
r!
z.
o
iJ
u-
-o
o\
6
o\
z"
lrJ
!"..l
Rf
frta
ilE
ss
| \ri
rfl
tfl
)' -a,eal;
7=
o;o,'b - q27'
4+s
l/" - /ot
Rr rb-Z&t
w
tkN)
/KN)
l,*,
fr,'a
90
--
oL<i<o,sl; f= 6,0s
.45
'Pl-d&^ees]
Fg ica.
Cr-rxtrrro'r!iP4"
lw*
ot80
laf
90
b-
="otlaeg'-1fl
O,25L.i .o,SL; i. 5,.3o O"r*1, Vu"),3ii!-
l?,o,
Fig'
to
*1a'YlJ.u1
fi7.36'
-lt- to#,-""1
ie
vmox
-44rn
/.t'>
l,;:;i
P'va
16
-rL&s*ul
fg. zc'
-Aclegraas]
fig. Ec.