Novembre- '1986 'ECOND0,141:!ERT-NON,LINEAR.EFFECTS", -IN ,n*FHOE sy;TEnns , y.
:Scheeijsbou4,0*
Tealini,itheliogoschroolVALE
SECOND ORDER NON-LINEAR EFFECTS
IN MARINE SYSTEMS by A. PITTALUGA M. DOGLIANI Technical Bullettin N. 98 Genova, November 1986-Tr: The,, traditional approach to the study of: theil--.behaviour of ships and'offshore systems at sea is based on the hypothesis that
"- the Isystetii _is- linear. Even if this approach - it::.-SatiSfaCtory for a great , number of practical applications, it is considerably limited
With respect to the study of ,certain. phenomena because of their
. non-linearity.
Examples,: are the capsizing ships or
Compliant, offshore, structures:
t ;
The. definition of ,a phenomena aS noti linear: is iunsatisfá.ctory because, it makes not clear which type it belongs to r
Therefore; it is _necessary to distinguish the _different kinds of :non-linear ,.systems and the casuistry they can model.
This paper deals-, with ,Volterra system,:-,in particular second _order systems, Which are defined by an -input/output function of the
kind'. -'' .-N..),.. .(t)ac (-tZii ,-41
f
vista) .-Where x(t) and y(t) are the ,input and output respectively of the
system.-- _
slow motions of
":-cirldra.
The theoretical characteristics of these, systems, the technique
by 7.1which it is possible to model a non-linear ' differential system by
using a quadratic model and the charactetistitS17 Of the response of such a; model- subjected to stochastic excitation are 'described.
Finally: the possibility of applying, this .technique to typical,
I - INTRODUCTION
It almost ,always happens., in technolo.gy. that the appearance of an .innovative .method is followed by -a period of enthusiasm.
during which more and more fields of application are found, a
period. of ...reflection during which people go on finding problems to
which .. the new .,,method,..-CANNO'T be applied; and by . a :re-analysis phase \-during _-which the .old fields Of study are resumed or further extensions of the new method_ are ..'StUdied. .
The study of the behaviour of ships and. Marine. systems at
.
sea has not overcame this taw; the appearance,. in the sixties, of
methods ..-based on respOnse-,. operators and spectral' density function was 'followed by a -flourishing ".development of aPplications and; at 7
the end. of the seventies, by a growing interest in non-linear
phenomena, those , to which method .CANNOT be _applied because of
the no longer 'vali&-hypotesis of linearity on which the 'method is
based.
As M. St'. -D,enis ,has -pointed. ,out, dealing. with non-linear
phenomena means, however,- acting like a monkey which divides the uniyerse, into bananas , and .non`-bananas: -it does not produce any _progress - about knowledge' of ,non bananas .' In .fact, very different
kinds of non-linearities exist; for example the effect of
-a-non-linear ,,damping over an otherwise-_ linear sistem, the impulsive effects' of an istanta.neous change of 'state; the Mathieu unstable
osCillations of a parametric excited system.'"
If we want to pass from negative knowledge (What does not
happen) to positive knowledge, we have to model every type of
phenomena by Using an appropriate non-linear systems which has its .ovin: main.- properties.- ,In this report, we 'deal with the Volterra
systems whose response_ y(t), to a continuo's -input x(t) is continuous
and may be , rappresented- by a convergent Volterra series of, the kind:
1(t).
J
where the termSjii(t,...), known as Volterra kernels, depend on the properties of 'the system and not on the excitation. The assumption of steadiness (the assumption that the Volterra .kernels do not vary
with respect to time) is implicit. The extension to non-steady
The above-mentioned written definition is generally applied to systems provided with state differential equations of the analytic
kind. Furthermore it is meaningful only if the 'series converges
quite rapidly .because the computational effort connected with the
utilization of a slowly convergent series is completely. beyond of the
capabilities of the present generation of computers.
When it can be applied, however, it retains the elegance and incisivity of the linear method allowing a detailed analysis of the problem.
In this work, we have restricted the "analysis to the steady
second order Volterra systems, the ones represented by the two first terms of the series, and we refer to systems of one degree of freedom only, in order to simplify the formulation.
In the following, the steady systems of one degree of freedom are defined as "Volterra systems".
In chapter 3, Volterra systems of the second order and their
properties are defined and. studied; in chapters 4 and 6 the case of
stochastic excitation is treated and, in chapter 7; we deal with
some pratical applications to marine systems.
In chapter 2 some preliminary recalls of the theory of linear systems are made in order to introduce the reader to the subject and just to show the analogies between the different approaches to the linear systems and the Volterra systems of superior order.
Furthermore, in chapter 5 an illustration of a very simple
physical system is given.
2 - RECALLS ON LINEAR SYSTEMS
A linear system is characterized by the' equation:
1.(t)=ja.h (t) (t. v. (2.1)
where y( t ) is the response of the system
x(t) is the excitation to which the _system is submitted
h(t) is the impulse response function which entirely
characterizes the system. 3
The Fourier transform of hit) is called the response operator
ON.
.. Out
14,1,4 h(4)c d
and it characterizes the response of the system subm-itted, to a sinusoidal excitation.
By performing experiments with pure sinusoidal excitation it
is then possible to characterize the system using its response
operator and by means of a Fourier antitransform it is possible to
find the impulse response function h(t).
Obviously following equation holds true:
= (-1 (",) (,-,) (2.2)
where capital letters denote the Fourier transforms of the relating
functions.
If the excitation is deterministic, equations (2.1) and (2.2) entirely solve the system.
(
If the excitation is a stochastic process, the response is
stochastic too and the equation (2.1) makes it possible to find the
characteristics of the response, among which:
the mean:
SD
(t). E[(t)]
h(z) E (t t)] dand the autocorrelation function:
t,,, ta) E r te.2)] h (.4) h (co (t4 --C2) CI d r,,
the last one allows the Fourier transform:
C° itj4L4 - i (.02 ka
FR% (W4 ,44-12)--=
if
r (t4 -2 e. ed '
t)
ci...ta,
-co )
If, as well as being a stochastic process the excitation is also stationary, the mean is constant
E Dc. to] Y-t
-and the au t oco rrel a vi o fUnc.:tfonw depe rid
7 :
In this ,case obviously,. the autoc.orretation symmetr.fc.-.
As ap- consequence:, the regpOnSe process and the
foIIowing . equations%, hold: Its may _ -ti-ansorin- -r-3:(
an:cif-the fbllowing equation, :bolas
-(w)
:where: the . symbol ( .)* means the complex and - the .:EFourier t-r a n stotinv - which is also
symmetric, R
In the case' of stationary prOcesses it is possible to define an
aiitoiddVa ,z,
_
. i
proved_ that, if. R
(c4
) , then:, -, [ )4} *4(t-a ' " :.1.1 7 It evident- that:_ 24"4-1,ii; 2) , function is
1t is..then.: pOsSible:'
- 'autocoVariariCe. function
-where4, (co) is ,-The .:funCtiOn,,., Which- MeanS. that the 1:2x(w) contains an impulsive term at zero frequerice:
the equation:
_
Cu.+) ex (6J5. -4114 Co)
to the ,rf',ict
that t(o)
6 2 -is the ' variance ' Of: theprocess an is symmetric, a: ,:power _spectral, density
-
-function is usually written as follows:. .
Aai-
- Jr:
' - defined- lin
integral is the va,t-frice of the proCess.
;
SECOND ORDER. VO.L.TERIZA- SYSTEMS-,, 5111314ITTED TO DETERMINISTIC
EXCITAT.10N::.
-.
,We define second .order systems as the ones characterized b
(t).al oz-),)e (t-
r
- 11(0 \
where -the- terms n----:.(r)
Which -identify ,the system..
.,defihe..,the',Fourier- 7translorni...of the
6
.11,<
.stioh: a -way that its
4 tz)
an.d h (t:1,;) are the .:Volterra _kernels
-
-It is easy to .see that the kernel h(2)(t.1-1t-2) is -.symmetric,
because by replacing h(;) With
, , the equation .._ (3.1)These kernels may either be obtained in an analytic way,
from the differential. state equations of the system. or in an
experimental one, by means of adequate tests.
There are several analityc methods; an exhaustive study has been made by W. J. Rugh /1/ in chapter 3. An example of this type
of method, based upon a perturbation technique, is reported in
appendix 1.
There are also many methods with regards to the experimental
way; they are dealt in Rugh's book /1/, chapter 7. A method based on bimodal excitation is given in appendix 2.
The system's representation in the frequencies domain is a
little more complicated. It is possible to prove /1/ that the
following equation holds true:
co
>(w) =Law)X ( G) 4 4
j
1( - 9-, 4) X e) x (9) de-ao
where capital letters indicate the Fourier transforms of the functions represented by small letters.
This equation is given to show a peculiarity of the Volterra
systems, the "pollution" of frequencies.
Unlike the linear systems, the harmonic component of the response at w frequency does not only depends on the corresponding
exciting frequency but also on the non-linear contributions of all
the other excitant frequencies.
In particular, if the excitation is a zero mean, X(0) = 0, then generally the response mean is not zero:
00
Y(e)=
2Zttt
Yele)x(-1:9x(e)-00
Before going on to examin stochastic excitation, it would be
appropriate to underline a peculiarity of all the Volterra systems:
the uniqueness of the response.
Given an excitation x(t), one and only one response y(t) of
the system exists and this is limited if the kernels are integrable
(hypothesis implicitly made in this paper).
Therefore it jS better to test the. system'
,.creating a Volterra
or example, the ,parametric
As a consequence the.,Volterca systems-. are stable _arid do. not
present; bifurcation phenomena
where the non-linear terms are supposed to be small, may be
succesfully modelled by ;means of a, _second order Volterra system, only if the excitation is far from the instability Mathieu limits.
When the excitation x(t) - is a, stochastic process, ,the response
y(t ) is _obviously stochastic too. ,
'- Thus it is possiblet o define the
...:-.. _i.
n-th order autocorrelation-_functions of the response as
Atj (t) + c () 4_ b4 (4)
(t
- ,
- ,
SECOND ORDER VOLTERRA SYSTEMS SUBJECTED
EXCITATION
Eg(,) 1(4 n)J (4.1)
.1-,Let,';s consider, -::forti the ,sake of semplicity the 'pure, _quadratic
, ,. system,:... ,,... -00 , , .,,. - , - f:-..4.0
1r _
g-If.
,-- ticcsral lk._(t,,c,a ) le c_ .-i.2.) ,A,c,t4t2. {4.2)
. .
. T.he:;extenstah Of the. bRiorriiou s. systems w be ,;''developed later
By Substituting the equation (4,.2) in the equation .1) we
. obtain':
14(CI,r0rx
7 excited -before (C4,ct )11(c3A).--,t2)
rmil!c.t.i.('-rcY.-_- C ce;, TQ TOCKAST IC on. 7It is to be noted that the n-th order autocorrelation function
of the response depends on the excitation's autocorrelation function having order 2n.
For instance:
00
(Z)
'IA 1-i(t) ri) Ck- - ri ri rz.
-co
00
a) (.1)
el)a (r,,r, ) (Eli; )
r9
-CO
If, in addition, the excitation is stationary, so is the
response. As a consequence, the n-th order autocorrelation function
depends on n-1 time intervals. In particular, the response mean is constant:
14
j.0
0:1 (rAs -C2) r-y"Irt- 1-4)otri
dtLFurthermore, if the excitation is also zero mean gaussian,
c4, (2), (I) (t)
rif L4.11 r4) rit e-4)4"-r< tt) (*L- b4)4
cz.)
4.r% t-4) Lt2.-.t0 and
00
rit)(t, to =if h tr,ri) h(r r )r(2)(e. ) r-(23(
r)d-r
4 *1 31 4 '4 4 al A 00 . . 4-11 f (CA ( r3f r4) r2)(t., - 1/4'2 r;:1)(r-i-t i .c4 3- 4-24c, 00 , ) (t3, rX2)(t'i- tz *C4 et)(44-tz-t
by using the symmetry of the kernels we obtain:
2 ifcc,
f h (ti, 2-2 ) h(r3,z4 ) r,n-e4 -ti 4. ti-t 2) vintl-rz 4 f-1- t-0 cir4
ri
where the second term on the right is the autocovariance function of
the response.
By using the gaussian properties of the excitation it is then
possible to obtain every order of the autocorrelation functions of
the response. These normally do not satisfy the gaussian properties; for this reason the response distribution is not gaussian.
The above mentioned considerations may be made, of course
also in the frequency domain but the results is more difficult to obtain. Thus, we compute only the mean and the power spectrum of
the response, without entirelly justifying them.
Taking account of the equation:
420 i to., r
j
rx( t)= Qx(c...,,, ) e 4 4.'4
27-i and using the
C;-)
44, roi (4,2ra(c.,1 La-9a)
we obtain: freo
5--- jj Wttit2)1 gif R
jc,,i)jc.:. itm4r4e-ic.i2rago(G-si+coa) 4 CA-$2 4 Wi 41 ri cirL.
4: :2if R (c..44) So(446+(u2)/ZirmirL
e..-".".`" 'T*
ex x
iit d r., dire at., awl
-co
-4--ifix(")41 H ("41,Lua) g.Cc4,*(,,a)
27C in addition it is: 40 = [r;43)(x) -d Thus : ( r Gil 10 ao rjx(ctil) (4J4,-(lie)ci 27r -*0 020 ar, - 6,34 ( C-4 t-) -L (a")((o)o ifi v)(C,r2)''')(4-3,z4) -L4j 2, c 644, z 27t oo ; t 3-ez...z)eco2
-± j ZxCLot)a otc.-32 ci r., dra 4 r3c1 r4 cit . 2fr
- oo o
- t'o-hros -itu.e..
'
i'lit211Cw4) gli(-'2)Fcs,rs) t e ci ;vizi.j
V) ( r34-co -00 -00
cw2 r3
azt
/ 0°- i (Ls,. wi ic.)if Qfc(444) ett(ak) W(434,4-12) 44 (-CUL 1'641) 2 ir go (441 1-6.3) d CUE 470' :241t cx,Qic (0.34 ) - H ( (0.3,1 ) ( Cul 4-61 6.41) cl co 4 Remembering that: Q(c.0) = (-(4) H (44-1.1,Lua) H (4.):,100 - 1-4*(-(010... yr_ and calling _ (4.)4.CA-34 )
we reach the usual notation
gr
--(2)
(Cei) ek GY) (CO + gr) ,)41 0,3 +at)
kur)ar
-co
often 14 Wz) depends only on the difference c02- ca)-1 , that
means
(Q-21,w2)
H (-"'L
= (j-j2)
2
in this case:
V.:)(45) fc13,(442A (4J4a) I H(a" 1)11cAr using to the spectral density functions, we obtain:
5 (Ca) Z.1 S)e(1/(1)$/t (-""?1) 11("191244e4 r
-00
( 1 4.'Oki-4 (a"- ,1)12.- su( e-G.1)1H
11
u112?e%
In the case of two-terms systems, defined by equation (3.1), it is then possible to write:
where v' t) ..is the and y( t)
by .theequatIon_
r
i-.4)
-The .0:eyelol.. ..tnt is :difficult and .gives rise kind.: .7 (ei.01) rka (t-4 . in Which': .100i it) (t.4 z) . --(-Ci4) (4 ) (4 4 ) (4-,'z) -rr`14 particUlar:
4()
It) to,Li.(ret-,.,z44
(4) . . iyo 47r re, d..rA .i4 '()
ta,r1) ric (E-4- ,)Equations striiiliaf).-tO/ the previous ones' may be..;.ritten_ in the
frequency domain, where the cross "spectrum of the processes
y1 and
y appears.
It should be noted that the term exciting process ,.is.,-g-aussian, thus the linear
of the response are 'untorrelated.
EXAmpLE--Just
to illustrate the equations deVeloped in the ;previous.:chapters, let us consider the two prisms in ,fig's._ 1. and one . with al. rectangular basis, the Other with a. triangular one',.each having
mass M; unitary displacement in the 'initial. condition, 'IC, and draft .
(4f 4)
) 6- C§;:- ta) 4 1r
_zero if the_ :and -quadratic, terms
h. The prisms are .float frig. W....ith .the water :plane p-ay.411::61?- tO the
genaratrixes. The added -mass 'is... independent of both frequency and draft and is taken into '7acCrOi.itit in the ...Trials -A4,; the damping iS, L.. .
. ..., ... ,
_
. . . .
Let's (.--..?nsider the heave :metion.:---y(f) only and call- _x(t=)'. itS.
.--...7 .
. excitation.
is: The state equation' of the L system (rectangular basis prism)
and.. of -- the ..the NsYstern, (trtankulatbasis prism),
assumi.ng:
By making:_
we.robtain foi7. Sysfem
()
2.4-00 (4t) 4and for the N.Syste.rii-..
c) 4"Joi(j
and neglecting the the N system: - L-I6 C4) ( 2fic+e-6)1 (4 2. 4(45 42G-Jo g (4 ) 4, or higher. ,*(4) - k.E011. Ei(0-7act.)3 J r ' we obtain for
It is clear that, when , the N, system approaches ,the L system 'and that the solution of the latter coincides With the first
order solutiOn4.., Of the former Thus we -study the N -SYstern
only,-knowing that its first orde`r. .solution; cbiriCid-es.. with the 'exact'
-The firstorde.r Voiterr,A kernel is: from i,Vhich 4
()
t443( )t
r Thus: tic4r.)
( - 2-4)-where the Second orde-i.. VOiterra kernel i
(r1, r1)t (ryiiei
L -r) r)- de 4(2) ) , ( .
(A) (4) (4) (4)
-(ti)
rz -(r2)'4+ 2 cue...va s
Th-erefore, in the frequency domain:
(A) e II, Cr) e
dt
4-,* (4-$Ge.
A (.0 ' - I 4-4 r4 _ ( (AJ2 ezdr4 d n . (r4( ;).L- e (Al ' , (41 ' CO (4) , (.1 H (41-(02) II 041) H (4-)a)-1-1 ('-',1)-_-14 Cw 14-If how, the stochastic exciting process x(t) .-zero meangaussian one,its autocorrelation function
,
.where? A and T
a-rec.:nstants
whose - dim-enSi:onsare -ELT] and [Tj respectively:, also the linear component of the response ha-s zero rile:in% 'and the .-autoCorrelation function.
(2) '
r
(t)
--\ wherealsT the
mean:-and the autoc'orrefation fUnctiOn,
4:10(
r-1 Lt.
rid Or) )L1Itt
r)
A2- e 47 1'
r a ci
-4, 21
tr iI
--.7--.2 4 VI (r4;r2) 4 c2Thus, in the frequency domain:
A-(ax (G=5)
whereas the mean is: 1/4J
-In figures to 12, relating to the N system,, the following items are shown :
1
i
i (r'AS., tI) iv, (.64O;) 4quadratic component of the response has a hon zero
C's0
(c...,Tr
4-A
As stated before, due to the fact that the exciting process is
gaussian , the linear and quadratic components of the response are
uncoreelated_., and therefore the spectral density function of the
response is.
dre
- the first arid second order Volter.ra kernels -,
It is,- therefore, : possible to
distribution 99- (9) of the process' by
-(115)2
,the first .1 d second order_ response operators.
- the autocorrelat ion iunction =:an(.1 the spectrum o
process
= 0.05
4. 01 rad/§ec
9 sec
16
the_ ,aut0Correlat ion func,ticp and spectrum ,
respon_se
-the autocOrrelation 'function and spectrum of , the second ...order ,response
_ &Valuated-, analytically or numerically, with the following
,
given to the parameters:
e e'xciting
e %first order
values
-,RESPONSE -DISTRIBUTION IN A SECOND ORDER VOLTERRA .SYSTEM
The response being stationary, the nth order autocorrelation
-function, evaluated for zero value of all its variables, is equal to
the n-th moment of the probability distribution ,because of the steadiness of the_ response:
firid the i-nitial- . probability
Using a Gram-Charlier .series
Or which 0295 .m2
- In figures l3 to 15 the following items, 'obtained by ineans of
-a computer simulation, are shown:
a realisation of the exciting stochastic process its lineal'. response.
-- .
!where i,2. ;:)..4:3 .. ,,a4 -..., .... are -. the. cent ral ..moments,': which - are
. ea, sily, obtained
from '3 -,
rt.,z.ny 171A- , moments,. --_-:.fir- is the.., ...variance 'of the process- (t.) and
j
i is the normalized variable4 (I -5")//gt
To evaluate- -The distribution the extremes df
(E) in , the knowledge of the joint distribution of the
process and of its derivative is needed.
'
Let ug write the ,equation (3.1) as:
ix 039 cua- 4.-
V)("(,bq5.,from which
().r .c2)..
c(6.)cte- 4-) t1 Ck- -Gsz) lc(6-2) citGi Giqz(2.) where h means CZ) .11 C(,2)2 r-iPe The 1-th
process and of its derivative is defined.-1Dy
and m-th order cross
(elm) et41.0 El( 4) (t-t)Zj(t-41) 17 X.( 1574) (trz correlation function '4(
Let us consider, again fOr the sake -of simplicity, a pure
The (14-m)-th order moment. of the joint -distribution of y(t) and Y(t) may be evaluated as follows:
(gem)
'711 ,(°/.--/
W1 ft
\
-Knowledge of the moments allow's us to iexpres's the
charac-teristic function of_ the , joint -distribution by 'a Mc. Laurin series
and, from these, to _obtain the joint probability distribution /2/.
The joint distribution p(Y,9), being known, the probability that
(5,_'s,)-(t,,,,. ie.o..
quadraticsystiem:co
;00, .(/)
. Cl) 1.00
r)( (N-r4) 4.0 ' )...1C(414'176iji _i) )4N4ia- .1.2t444.4)1 a C - - - 41 r2{4Z4A4
-a given extreme (t.) will not be exceeded In the inIeryal
may be obtained as follows. .
First of'., all one must Compute the mean- number of times that
the process y(t) crosses the threshold u from below:
E [kir (1-4 =
where:
fe'if(1T Ica) c'o
More, if we assume that the crosses of the threshold are
independent (hypothesis correct when ), the distribution of
the crosses is a Poisson one and the probability, of not crossing the
threshold in 0, T is exP
' The probability that a certain extreme will not be exteedeed
---4
in 0,T is, therefore:
1)(1-) P`e (-T) e
where 1;y(9T) is the probability of being below the threshold_ at the beginning of the interval.
Therefore, the derivative of POT) is the probability
dis-tribution of the extremes of 9T in the interval o, T .
7 - MARINE SYSTEMS APPLICATION
A. field of application - of the quadratic Volterra systems in
marine engineering is that of low frequency resonant systems, all those whose oscillation periods are greater than 50-100 seconds; At
these frequencies there is,, very little hydrodynamic damping, and
the linear excitation, with the 'same frequency as sea waves, is
almost zero. The consequence is, that the response is dominated by the non-linear, low frequency excitation which may arise from the presence of quadratic terms in the state equations of the system. A typical example is given by the. slow oscillations of moored marine
systems, in which- the quadratic" term arises mainly from the
relative motion between the free profile. of the
sea and the hull
surface of the system, in addition to the quadratic terms- in the
PC'irii)cii :AT
1°1
equations relative to the velocity -potentials. These problems are
'dealt in references /3/ and /4/.
Another field of application is to be found in those systems
whose 'response has a non-zero mean even if subjected " to a zero
mean excitation. The easiest 'example is the - surface of the sea
itself; it is known that the crest of very steep waves is much , higher and narrower than the trough.
An example in the naval field is given by the wave bending moment which may present considerable differenCieS between hogging and sagging values. The reason is to be found in the flare as well as the non-symmetry of the wave profile and the non-linearities of the hydrodynamic potentials.
A quadratic model of this phenomenon has been proposed by
Jensen and Pedersen /5/ with satisfactory results.
. Another example' is the added resistance of ships. It is known, infact, that a ship's resistance_ is higher in a rough sea
than in calm water. This increase in the resistance is due to the
mean longitudinal wave force which is not .zero because of quadratic effects.
The application to "emisymmetric" phenomena, such as the
non-,linear roll of ships, requires the use of odd order Volterra
systems .(third or greater). This leads to a significant increase of the computational onus, therefore it is probably more convenient to use _other methods,- like- the pure perturbation technique /6/ or
numeric simulation in the -time 'domain.'
8 - CONCLUSIONS
In the present_ report the main properties Of second order Volterra systems are *summerized and -their applicability to the study
of some typical non-linear marine systems_ is discussed.
- In preparing this, _ our intention was not
to produce something original and we therefore apologize any readers to whom this method is already known_.
The application to the ma.rine field does not seem to have
been - Viidepread- the method -Der-haps .merits; his report seem's- to Us tO be an useful'. sy_rithe:.;iS:
"Arclii.tects; VoI. 121,.
,and therefor,e
. ,.., ... .., . . ., .
...h 1 s. c ont-1,..c.,,t.ion we - would like to,' merit ion '.tne, paper; by ,, A
,
Na'eSjs' ./-7,'',, 5imili'4,r,..i.in O:Ontent, ,publb_lished in; 'he Journal of Ship .--..Research iii'Decerriber,, -1.-985,-.3...fii1Z:..11 'reached:. Italy _while the.':'.preent,_..
.:. _ ..._ . 1" -, ,...,
- .
- " . .. ..._ , .
wOrk wa.s: -beirig-ptihted. . - ,
.
'...-,
/1/. Rugh:"'r,W.I., ."'Ncifr linea system: theory:. The.' Volterra/Wiener
app.r6ac'h--; J6lii-C;1-logkin'S; 1981' -/2/ Lin ".PrObabilistic theory
McGrad-H11I, 1967 y
structural dyridMicS'
....
/3/. Pinkster J.A.---;.--1-HUI)Sdians R.I-E,..kr, '.'The .low irk:pi:enc.)/ motions of
:
-semisubrnersible on waii-eS!' ,. ProCeedings -of-. BOSS 82,:'1.- MIT,
Cambridge, . Massachusetts, -082..
.
, . .
' /4/ RobertsJ.8.,
. . "Non' linear analysis of slow drift. oscillations of.. ., . .-:: moored vessels in randOM,,seast, I.-..- .Shlp Researdh.. VOL 25, N..'
'-. 2, 1981 ''.., :- - . - ...
/5/ 'Jensen J.J:, Pedersen, P.T., "Wave induced bending momehts, - ships - A quadratic theqry' -171.1e Royal Institution Of Nava.1._,
-/6/ .:Ferro; G. ,..,"..'Caret-ti. -.'..Metodi per '1' ahalisi 'del
CompartaMento-ti.Pn-. lineare delle nayi in Confirso";,;-R...I.NA., ,R,e7P:Ort N..
220, 1985
/7/ Naeas A., Statistical analysis of second. order 'response of 'Marine structures", J. Ship 'Research, Vol.. -29, N. 4, 1985
20
-L
Nick)]F
Dcto,(4)/
.Appendix 1
ANALYTIC DERIVATION OF SECOND ORDER VOLTERRA KERNELS BASED ON THE PERTURBATION METHOD
As is well known, the perturbation technique is based on the
assumption that the response y () can be developed in a Taylor series around a perturbative parameter
()
EA4,(t)
(A1.1)by putting this equation in the state equations of the system, and equalizing the terms containing the same power of E , recursive
equations may be obtained of the kind:
4 -4 Ct.)] (A1.2)
where L is a linear operator and FL is a finite (not necessarily
linear) function.
In many cases, if h(V) is the impulse response function of the
linear L system it is possible to write the first equation of (A1.2) as: L
[yi(o)r [:*(4)]
t.b) from which: (A1.3)v1(r);(4-r)dt
ti(r),cdr
-00 -00 and: L(0]
1[10C t k)3ti(r)0c(i-t)cie]
from which:xWz
In(t) r) (A1.4)It is almost always possible, by means of some rearrangement, to express 2(t) by a Volterra kernel series.
For instance, in the very simple case that:
I
aucip
FA CO a 1C (4) lii(4) ".: ICC+) h(e) ICti- -r)
de
21
00 by cha.--Ting the equation is obtained: Whereas It is clear that,. in variable Yz-41
The 'example- refers to t e system:
(t) (4.)
141) (ri.-ri)
of integ-ratiOn. the - following
if the amplitude o the excitation u(t) is -small; we may state.:
this case,
z,
z)(c4ira)
6 n(ti)t 1(4- r4)
22
Obviously the system is a second order one if the - kernels of
Order higher than the second which may arise from the (Al 4) are hegligeble, therefore the perturbed - series (al .1) converges at the
second term. (.k) E -rc
e?
c)
0(63)4- 62i ct) esiz ct) - tt) Ev4(t) 4-, -tJt(6)
("
with reference to we et: i1 .(.40 . ti Cr.%) ,(k-r, vi(ri, ,cC-'z'.., - 1-1) 621:ez 4-r-,EXPERMINtAL DERIVATION OF SECOND ORDER VOLTERRA- KERNELS BY
THE BIMODAL EXCITATION
_
_ This -meth<5d-,cOnsst's in subjetting.-- t esystem-to excit-ation,
of the kin'd:
:Appendix 2
i
C:2-5 0'11 ) 4.
The -fresponse i1I be:
_Pp
1.1 'Cc, 1 Ai (-Gs (C-E)1 4.-A ,
;1)
jtc.,,r1)1Ai
.44,(t-r-4)1c.'s4.21 cu2 (t--e.4)2 LA3ECua(V-r,111
A1A s[6.4 (k,r4)) Lai tcy,7.(e,-,..ra)3
4. AtiA Er-.3% - 1.4 dcid'!.%
by means of some rearrangements
3'('1/4 ) z (42$ ( 441 4() )4 4 UssLc(-44.1 4 C4/3), (24-60 Id ECC141' ":")
- ea
. .' C3 (A2.3)-yv(9. Ih'eT. e.:orripOhehit':-.,at,.!:.ttie. .fietlyehty cal -1.:ot it ,may:!...be written as:
74.2-
et-(c-ic, cu let); i(44tri-,ahta
-) ,s
''4)'
4a
The, last twe,(=iiite&a.,1s: re. equal 1-..leca-iise.: t
properties of h theiefore.:.
=
11('''cz.,,T,)-4
stv. _
2)) s
I u.(cuir.4 -4- z) d Are .4'ZrA)rL) A 4 LA) t (co.. r: `"tr4) 4-:
-tbtCcLPl4cuDtJ Ctts' -r coll-r4 )}
d14-r
-This means that :
*.=
,
L):
e'-:Fo,ur.iii--,trahsfo,r-h.. of 1the kernel.liat&A7
24
e symmetry
(42.-5,
With the -exception of the finite number of
frequen-,cies(cof,1/4) for which (c,2/4-4)2) equalize one of the other frequencies
which -appear in the (A2.3) (*) , the term $2.(t)can be isolated by
mean of an harmonic analysis of the response signal of the system.
By completing IMOvw2)and CY44,(L). taking the .limit in maT- cases, or
g
by using other methods, it is then possible to compute h( a".1,t2.) By an analogous procedure, it is possible to demonstrate that
The term ii1)(t) may be obtained by analyzing the component at the
frequency W./ (or ci)2 ).
The disadvantage of this method is that it is impossible to
directly verify the order of the system. In other words, the first
and tsecond order terms may be affected by unrecognized higher
order .effects. If we are not sure that the system is of second
order, further enquiries will be necessary, see Rugh /1/.
(*) For instance if cOi =Oa , the component at the frequency coii-wzin
the equation (A2.3) covers the one at frequency 2G01 and
the y'12 as defined in the equation (A2.4), can not be
isolated.
t US -tal<ing.4: 0,-:1" ' _ ; . : ,
cons'idei-. the f.driction-s':
60 h(co c,r2-t.) ,..RerheMbeting that : Apjehthx 3 r-: -e. ( S.,0 (ra one mayobiain f. YE. ri.=7 26 "rt7r ( - - s-1) d one Obtains: .. ( _)ti 6,172_ z )clz I obtains: w , one -obi , ll (t) ' . . ) tl (7(4 .-) .2c.k - r.1) ...c
The-two -
:-.)-;C , formulAtions of fi,(t) -. 0Fq'',... . eiqui.v:alent... Thus F(L) may be obtained by their mean,. : obtaining', in "thi.s:,.:.
,
4y, .,a
: kernel - 1.4:rhi'C.Y; ;',i's ,symmetric and consistent .1.4.ithhe , adopted,::,..
. .simbold: .: .. .
-.
After this, 2.(t).:. ;4(0 - (4)
.c(t)
. and:1(4).:::_ffee[J°°Id(r)4 (Li-r)
)dc
- i (t)1,1( - 14(rs)
* S-0(rz-ti)jx( E-74)1e k-1-2) dr.( Girt
making:
kt4in.) in(r) (r4- Ll(t-2--c) aft- ((c',4) (rz) Z4- rt) +
(t), -G., PROOF 2 PROOF 3 (a) ti (c11u.2) We obtain: (4) H 4- h (eA) -(1) ( -
t
)cidr:
0-01* 6.;rcshi (wo 14-12
t)cir
CAJ 0(A-% 1)411- 0
Remembering that e`c-t'2'.cos(co.erf-Lsin(we) , we obtain:
( ) c
c to) z. a 2..c.AsCcor)seu (wo cir
(4- 1.16
I r
Z
C)
(c.oj't) az]
By applying the prostaferesi formulas and developing the
integrals, we have:
&Jo?. YG.,02 -6-#0 4 2
cow.11
(04.13-t-2)t 4-
I g2 J1 cJ
ri(Gui,uut) (4),,u...) +la ( cAN, (1.0 4- r4(4-4,t0a)
Where, using the saffie s-imbuts as in .proofs 1. and 2, it co ti(coliwz), (r) h
ti(r -r)
Q --CO o. -1.C4-11(ri-e) ti(.)) d-r., Vi(r,-tJ
fcb%36(r) A) (41) e 1-4 (G"(4) (4) ci41 (.-#2 -Cl)(4)dc
44 ")(04.0 CcaJz)C4
4AI c'2wi")
It 'rt) ="r,) r4 Czctri d Tz J.1 _ c4_944 (o/ )2,1f
6 (t,) (4) -00 (4) Cl)-
'4'2) 1,4 -( 6)4In- -the,_ same wa.y:
(L, Thus: c-t) ((04 c
H (w21
+ CA-4) 'jjeo -_ cotr z e dri ) _ t't (4) (CY4) 4(:L.4)
14(r2) (r4-rt) e ( (4-'4 +-1"/.) 't4 r4) e . driIi0i
ti(r4) -zt) e -28 4 t cbriddr
_ e c ez ' e ct riar G41 CO 14(ch-1,0- W(.16°2) W (C4,1)-(24d
By substituting the expression from proof 2 in the place
-of 1.4 (1) . we have:
(7)A L2 )
1.1 4"1 I CaJj) H c 4J1 C4j1) 4- A
"
&Pa calQ.e 11)( c.A) =
ft (
N -
fiN)vv, 1-1 (2)(6-'d ,14,2
) =FLCFI
2r)
) Where: 2. (-7/
- (44+0)414.4 eo...,44.02,..4,E(444441Z)21-1g144641 --L(too2 -W21) 24 4 f (AJac.J2.2] 49,2 - tua)2 Fi 42f c
(o-s,4.wt.) c44174.$1-( ca.1/4,2-G-hz (c4.0.21-(4.022 )4- 124,02G-hicu,,z(c.L2-24...1 4.,,t2 )44" 4 S4 0-'oel 0-24 Cuz (144i cJcs?)
Ps 2s (4203641a jz [an (4,J. wit)
4. Ca..94( Cia02- Gail) Cue, (4-11 (..4,03
Thi way, by relation of It t =1+5, r2 exchange
must be underlined that, F(w1)being equal to El (wziebi ,
e H(2) and kr) H (2) are invariable with respect to an
of c..)1 with ck.)2
s conclusion might have been reached. in a more general noting that t/(23("6--.1:60 e-144)-1Z-"le-c1-)zPr--2' is invariable in
to an exchange of indiCc6, l)eLause of the simmetry
--.- SECOND ORDER 'VOLTERRA KERNEL
FIRST. ORDER RESPONSE OPERATOR
(1)
Im- (43) IMMAGINART PART
,(1)
Re H (to) REAL
PART-. 33 (1) Im H (1) Re H ) rad/tec)
6(a) :SECOND ORPt44 RESPONSE cOPERAT,61,
(REAJ PART) ,
coif:rad/sec Re H(2)(011 . .4
4
P-4-4 4 w '4141
\N-q
4top.,,- \'.1&4W .4
0 to ,44I Prorr-___
...
...,,,,
Iipp A
top.
' v1. I %S
/04*".
"
40,40 4.,
44 .401.-4--0.Z4,11,
4,,Alhas.,
4 a 0 °' 114-
4 1 .-44; Tj+1 s t 41 ff0W leftiel. Ags'zr4
4
.4p,wiw1 r
0.1Z.
..
0 . 1 4 44111111,A16...4,
4..1.
1 11, 0 . 3'.4'
.*
0.2Fig. 6(b) SECOND' ORDER RESPONSE OPERATOR
(REAL PART)
-40 -20 - 0t[sec]
.
Fig. AUTOCORRELATION:FUNt7ON OF THE FIRST ORDER
1'1
40
_
I--600
--0.050
11 - AUTOCORRELATION .FUNCTION OF THE SECOND ORDER RESPONSE_
0 . 10 0 . 20
POWER SPECTRUM OF THE SECOND ORDER RESPONSE
17,
itaisec
0
-6.
60--1 .1
1:8'00.7.?
Ti.
rjt
;
c,E1:3
10:JO 1500-:25
35O400 49.-00 50i.7i0 .550?BT sep
BTONput4I-Siudia--46iSeparatOri filtrO per acgue Oteo'Se_di:sentina
e:su-miLlratori di'-contenUto oleos° nel-Leacgue di:-Scarito daLle nevi
PT 69-- FEBBRAI0,1980 / 1116'", TTI
Un -archivio di- Coeffitienti idrodinamicii.di_pr_essione e-.di.forZa
per cilindri bscillanti in superfitie:tibera,.
- .41-file7of'hydrodYnaMic.'01ressure and force
coefficients-forcylinders
oscillating in a.free surface
ST 70 - LUGLIO 1986 / CAMPPTPLP,
EA1A5SIND,-M1lRCHESL._NALQINT1N-=s-7LinAlgoritMo,per'il calcolo.deLle superfici lavate deLle.nayi-petrolliere
(COW.)
An algorithm for washed surfaces computation in-,tariker-s-:-(C6de Oil
Washing)
BT 71 OTTOBRE 1900 / SPINELLI
Convenzione MARPOL '73, come emehdafa-:dal prOtocolL0,-78.
Interpretazione. _ . normatiya=contrOlji--RINR-Oer riLascib.
- di dichiarazioni di corrispondenzalaCie.horme deLla:con'venzione
_
BT 72 - DICEM3RE-1980 /PITTALUGP,
I:carichi
d'ondadi, progettO e La, Ltio correlazione,- - ,
Design Wave' Loads-and their correlation.
BT JS - OVEMBRE 1981 / FERRO
Metodi e Problemi neWanalisi
deWaffidabiLita'
navaLi . _
BT 73 7.DICEMBRE 1980 4HCARETTI.':
Confronto_ teorico-sperimentale. .degLiHoperatori, , . risposta-f-deLLa
pressione .tP-Onda tareha .
toMpariSor-rbe'tieieen'theoretical and..-experiMehtplesponseampLitude
pOergtorS'OnhuLCs...,
BT 74 - FEBBRAIO 1981 / PITTALUGP,
Una ProCedura per !..a :0_alutazione del.,La.robustezza.,trasverSale
navi per iL-t.r,atOdr:tb:.di. Merci alla
procedure for '.thet,--iransverse strendthassessment:of_targe
..carrier ships
Methods-and'probleMs in retiabjLj.t nalysis of ship structures
BT 76 - CUCEMBRE 1981 / BISAGNO, MARCAESI, VALENTIN
GIPSY - Un'post-procesSor-Oer
l'ahaLisia4ceLenentifiniti
GIPSY - A poStOrOcessOr'for finite eleMent_analysis_
_ _ .
BT 77::'--GENNAINA98242FFJJ
ApplicabiUita' deLle tecniche affidabilistiche
alla
progettazionenavaLe
Applicability of reliabiLity
concept_to ship,design. .
ST 78 - FEBBRAIO 1982 / SELVAGGI
ProbLemi'di progettaZione per na,yi
Liguefa-tti"--4-6asSa 'teMperatura'
strut ture
L
pc) L L
.1rEpc=r-11.t7.a.-Design problems for ships carryng Low temperature liquefied gases BT 79 - APRILE 1982 / SPINELLI
Convenzione MPRPOL 1973, come emendata dal protocoLlo 1978
- interpretazione della normativa controlLi RINP per il riLascio di dichiarazioni di corrispondenza aLle norme delta convenzione
Cseconda ediZione)
BT 80 - 14136I0 1962 / FERRO, ZILIOTTO
Applicazione di una procedura diretta per il calcoto dei carichi
d'onda per Le analisi di robustezza trasversale
AppLication of a direct procedure to the assessment of wave loads
for the transverse strength analysis of ships
BT 81 - PPRILE 1983 / MICILLO
Applicazione del procedimenii speciaLi di saldatura nelle costruzioni
navali
BT 82 - NOVEMBRE 1983 / MPRCHESI, ZILIOTTO
Comportamento post-critico di panneLli nervati-1confronti tra
risuLtati numerici e prove sperimentali 83 - DICEMBRE 1983 /-PLIMENTO
I materiali per _La costruZione degLi sCafi
Note sulle caratteristiche e Prove del materiali second° La normativa del RINA
Materials for huLL structures
Review of the properties and tests Of the materials accordingly.
with RINP requirements BT 84 - GENNPIO 1904 / FERRO
Advances in the calculation of the maxima of ship responses
Paper presented at the Euromech Colloquium 165, reliability theory
of structural engineerind systems, JUne 15-17, 1982, Engineering
Academy of Denmark - reprinted from Dialog 5-82
BT 86 - GENNPIO 1904 / FERRO, CERVETTO.
Reliability of marine structures under dynamic Loadings Paper presented at the International Workshop on stochastic Methods in structural mechanics, June 9-12, 1983, University of Pavia - reprinted from the proceedings_
BT 86 - GENNAIO 1984 / ROBIN], ZILIOTTO
Wave torsional moments in ships with -Large hatch openings
Paper presented at the VI Italian - Polish seminar - Genoa,
November 1983.
BT 87 - MPRZO 1984 / SPINELLI
Convenzione MARPOL 73, come emendata dal_prOtocoLLo 78
- Interpretazione deLla'normativa ContrOtli RINA per il riLascio di dichiarazioni di corrispondenza aLle norme della convenzione
Cterza edizione)
BT 88 - 11166I0 1984 / PPSINI
La saldatura subacquea -;*Stato dell'arte
BT 89 - GIUGNO 1984 / FERRO
Stochastic model's 'for Low-frequency, springing and impact Loads on ships
_
BT 90 7 0041410 190$ liTTPILU(11, BISFIGNO
_
-Mode rie tecnici7e= di ,-.#641-"i:gir deL cOMOOr tamektb, in mare':
BT 94'-''.,7.M1=166-10, 1985 CFAZZL/L0 . .
Pandra'inica sui fondamenti teorici deLle fliec-Canide deLLa frattura
.-BT 99,. ...I.LUGLIO 1989 :/ 1 .CERVETT m i'-'' FERRO. -,_ . . .
9f-f-idabiLilta!, e r idbi:idariz'a ;:ne L Le f orcd-ioni .' offshore
.5_,Nisteni- reL,iab-il: i tc,-: of offshore foundations
'BT -96 ..4 'APRILE 1986,- -I- CLBIENto . _.
, . -
-Niezi di sat_ vataggib 7'.Le' n-uove norme .-" de L-La convenzi_cne.:SOLPS,'_-. 1974. , .. . - C833
.
: , ,:-:-., . . . ,
-_61 97:14_,-Setterretire- 1986/ - Pi t taciog.a:. . ... ' . . ...
imiLiaritieS and'differenteS...in- t1:5ih,:WL--.e:d'itleamt,, theorie.... '''-',
-,. .
DT "98 . ..- NoveMbre : 1986-i Pitt'a louga -r- Ii3gLitik. 7..
. ,
54Ond ,di...ci .i- -146.64 n:146,4i-- effects , in marine- yatetri'S
.-., .. , .
, _ . . .., ...
--. .
-.Prospettive deLLa Or Oge't ezipfle aff-id6.61 I. is tica -.del. Le. sit rt..it tiire marine
. ._ .
HT 91 GENNRIO -1985 1-.604v Ei ZILIOTTO.
Addl. iCazione deLLaneccanica del_ La; f r at.tur a neLLe veri.fithe a f at ica.
-IC C3 FREE C3FRIt7.- 3 lEt 1 1'9' US (S.
RP 204 -::SENNAIO '1980 :S. FANO:WU-1d.. Za,1TTO '
i n un !7' p o df.E.1 1EE:1 1 e , e: d e.(4 1 i
.
:L ED C:1 :I Hi;cZ,.. a I: C: :L t-liyaff
,
RP 205 (-*) :FEBBRAIO 1980 .1 FPINr.f.IULLI,' .TEDESCHI. , ZILIOTTO
D et Er f7.1 1 ...D. t
t
12 t e eL, adE g.:I. el ene i 3.pondeaz -.
de.1..1 t C.T! a f.71 1",; h y d Li S.,
RP .SFTTEMBRE. 1986 /
Desrr -.-.seippl 12nn d a su I 1 e se:zmi
trasversaI car ena
RP 2-07 - 'SETTEMBREH1990 / FA.NC FULL t ZILlOTTO
'Alcune consideraZzIIrisu 1 ana.kisi-.. deiXe sOi 1 eda. tazion i:n serbatoi
1 i ndri c 'dri ontai da. 3e11 e '208 OTTOBRE71980. /. FERRO
-_,Aff.idabd,1 it a (7, d lie condscenz e e. pr ospettive.
-dei le cerc.nT
RP 209. - OTTO8RE 1980 1 FERRO
daratt eri c h e. s tat darlcc _d pannel 1 i
ner vatl
RP 210- -:.OTTOBRE:1980 .FERRO. FITTALUBA.
sampli +.idato' per e 1 ung ter ffy.i.ne ded , di si dshing.
RP 211 -*AGOSTO1'081. .FITTALUGA, ,SCI.ACCA, ZILIOTTO
A:itune. note. sufIe .w.'f...braz 1 eFSi.onai Oegl 1 bOri porta E.1 ica
.RP 212. 7-- AGOSTO 1981.: .1. FERRO, FITTALU6A
u.en ..d.(72 et:I E p :I or! iflet
sul cal cold, della ri spost
anavi-RP 2.13 -DICEMEIRE, 1981. 1-FERRO.
Ii 1u ey:n z.a deld a A...:4.-dhfaa band a -st. a Ii st r i buz i one d 1 p !.t. chi
un process° stOtasticz
RP 214 (*) - ,MAR20 1982. / ALBERTI.4 ZILLOTTO
Ahal i si del cdmport amentO Litt i.Arale: di navi porta.contenitrrl
Sug get te i on e
Rp', 215 - eIutNo 1982 -/ CAREtTI...
] Nen-1 i near., fr-cluency doma.i'n analysis. moti..on7... and lo.Ad
of
Ships i rirregui sr- waves
RP 216 - APR ILE 1983 CAZZULD
Un programia - 'ver ifi c a. gl °tale delIe temperature die:L LO
o--RP -217 7. AGOSTO 198'3 ALBERT BERRINO
spErlMentale di Un metodc cia ci:.:c:l.
del cdmportabeto
-.RP 21E1 ( te;'): ' - MA. RZO
It§.84...:.../i..:Fkkka-... - ,...-.::- - ,,.,. , - ----
-.-F(D Llncl-arri4n-Ea l's.'' -.:.c.) i 7 : 01- 6 c....6 d Li,- iE...,
structures ' . ..
221 MAGOP 1.9g5,,,, CASC1ATI,
'DIRE-TTOFiE RESPONSABILE .CARLO CASTELLI r..,CENITROSTAMPA R.I.NA. EDITOR-E- R NA_ VIA CORSICA 12 - - TEL. 53851 AUTOFIIZZAZIONE "TRIBUNALE JOIGENOVA -N° 27i732del.:10 aprile 1973