1986
t\RCH1EF
ROLL RESPONSE OF A SHIP UNDER THE ACTION OF A SUDDEN EXCITATION
by A. Yucel Odaba0. and J. Vince
Lab. v. Scheepsbouwkunde
Technische Hogeschool
'Delft
SUMMARY
This paper examines the
impOrtanee
of roll damping on the gust responsE of a ship. As 4 Mathematical model anad
hoc roll equationis
used and the. action of-the wind gustis
represented by means of astep
functict, A solution ofthe
equation of motion is developed by using the asymptotic. method of Bogoliuhov and MitropolSki[3]. Computational results on theinfluence of roll damping, initial conditions and the gust magnitude are given
for a
stern trawler. Results cleatly indicates the importance of toll damping. Inthe
light of this investigation it is suggested thatin determining the gust response of
a
shipin 4
weathercriterion
the_INTRODUCTION
Jr. the assessment of intact ship stability by using various weather criteria the final
-step
is the determination of the maXimut roll angle under the action of a sudden wind gust. Which is assumed to Lake place when the ship reaches a maximum angle of roll towards the wind.In the estimation of tht gust response the effect of
roll
damping is neglected and the solution is reduced to an energy balance problem which is determined graphically (see. Fig,1).- Apart from the differences
in
the evaluation Of the gust exciting moment, various 'national criteria also differ in their waysan the
determination of the roll angles towards the wind,of,
Kefs,[12]. Therefore the initial angle at which the wind gust starts to act differs betWeen different national'criteria,
The aim of this work is to study the effects of the initial angle. variations and roll damping on the
Loll
responEe of a ship under the sudden action of a wind gust which iS represented by a step function To represent the roll motion dynamics an ad hoc equation with non-Linear damping and restoring moment is used, The method of solution isa
modified version of the asymptotic method of Krylov and Bogoliubov[3] which is presented in the Appendix, In oroler to make the explitir scittion simple the restoring moment in the examples is aNpressed as.
a sum of a linear and a cubic term although the method can comfortably' tccommcdste any kind of non-linearity.
The resultsindicate that the
linear damping coefficient has
a pronounced
effect on the maximum r.:,11 angle espeaially when the initial angl. (at which the gust starts to act) is large. Numerical experimentation has indicated that the difference between the computed maximum
response and
the
value obtained by area balancing increases with incTeasing gustmoment, It' is therefore advisable rict to neglect dynamic. effec'ts
In
such calcula,:icris,2 PROBLEM FORMULATION AND METHOD OF SOLUTION
Following the common assumption in uazious national weather criteria let us consider a ship which under the action of waves heeled to an angle 00 (see Fig.1) and than subjected to a sudden wind gust with the wave action ceasing to exist*. If the restoring moment of the ship
is expressed as a function of the angle heel, i.e.
M ° g(0)
the usual procedure reveals that the settling heel eagle (i. . the angle that will be approached as the time tends: to infinity) is deter-mined trom the solution of the equation
g(0s) Mg (1)
where M is the gust heeling moment and
0s is the settling angle.
Determination of step response to a, sudden gust is determined from the principle of balance of energy, excluding the damping effects:
Om
j[g(0)-M
] do = 0where Om is the maximum roll angle following the action of the wind gust.
' To reduce the degree- of Complication let
us assume that the righting moment be expressed as
g(0)
Bcp3
(3)(2)
* The reader can assess for himself the logic in the Sequence of events and their relationship with reality,
I.
then the settling and the maximum
responee
angles are obtained from the solution of the following equations3 B s -Mg- .= 0 .13 4 A 2 _A 2 B ,L4
u
A-
6 4' mm
g m
2 .o 'fo 't *O1 =If, however, one would like to consider the ship's reaction as a dynaMic system it would be more appropriate to study the response of a ship by trying to solve the equation of motion to a step function excitation. To a first approximation rolling motion of a ship can be described by the equation** (where the restoring moment is simplified):
where I is the virtual mass moment of inertia, N1 and N2 are linear
0
and non-linear damping coefficients to be derived from forced rolling tests, U(t) is the Heaviside step function, and
(0)
denotes the differen-tiation with respect to time,Dividing both sides of the equation with I one obtains 0
02); 0;20 to3
wu(t)
where 2y111/I0, co2A/I,
c=B/Io
and W=Mg/Io.. .
To determine the roll response to a gust action equation (7) needs to
, be solved with the initial congtions
) 410, ;(D) = 0 (8)
(.2y
(41-+ N 02-)
0 +
4
863
= M U(t)_
** See Ref [4) for a discussion on
the
equation of motion.(6)
Following the modified asymptotic method of Krylav-Bogoliubov, Murty(51, the solution is expressed in
rhe
following form:4
t) + k a e (a,W) + 0(e2)
0( 0-6
7
where a--=a(t) and 11)=*(t) varies with time, k Lakes values 41 depending on
the initial and settling conditions, and
Ne2)
indicates that the terms neglected are 'at least in second order in e.In accordance with the procedure outlined in the Appendix the non-linear part of the equation is
f(094)
=Y10 0 - 0
and by virtue of (4..8) and (A.10)3 1
. f Lf1
+s
--k
20s (34YT )4.e. {(62PY-0 c=0 -s
( 9 )
-y)-3] (a/2)
=IP 2
+ k[A1-3(1+yyl)1(a/2)3 1+e t-kclys[34112+IMe/2)
[3(14-1Y)tAY1](a/2)31+e2*0 (a/2)212y1 (A-y).-.3]:
e-405012)2[3+2yi(X+10l+e34ifyi(X+*)-11(a/2)3 -31,b 3 ke (1+i1(A+y))(e/2) f(0,,O) = 0: (11) (10)
I.
I.
1
F (a) = - ---4cO a2
(34-2yY1 )
o 2 s
F1(a) - 2[-s,-( _ )-3)(4/2)-4-kfty -3(14.1,Y
Ma/2)3
F2(a) = 45(an)2[2y ( -y)-3]
F3(a) = (a/2)3[Y1(X-/)-11 (12)
G1(a) = -k0:[3.411(.+Y)](4/2)43(.1+YY )tAy ](a/2)3
G (a) = -0s(612)[3+2y (X-py)]
G (4) = Ic(a/2)3(1+Y10.+Y)]
22
-Where A, = (y- to ) , and for n:!..4 On(a) =
Fu(e) =
O.Consequently, equations (A,12) and (6,13) become
yaAl-(y+.2A)A1-1-ya2Bi-2X01=20:(yi(A-y),3](e/2)
2k[Ay 3(1+y,(1)](a/2)3. (13)
-(2A)A3.--ya2W.,2AaB11 =-202[34-Ys (X+Y)](a/2)
- 21([3(1+YY1)t)0(1](a/2)3 '(14)
'Solving for Al(a) and
B1(a) one
obtainsA1 (a) = =-Y 0 (a/2)-(kJw2)13y+), (w 21(2)](012)3 (15)
s
2
B ( ) = s(3,iry1)/2X-(kA/2w2)(3+2YY )(a/2)2 .
By using these results one can construct a first approximation as
da .
dt * (a),
A,
Which after substitution takes the form
da_3
2 dt ='- pia - kp2ad*
'
cl7 = P3 - 191a-where 2 P1 = 1""14)s/2 p c[3y+y(w2+2y2)]/(81.02)9
, 2 1 = l+E0-2 (3+YY1)/2A, P4 a.(3+2yy )/(8w2). s a°P1i 2 2P1 2[(P1*kP2ao)e
-42ao] a(t) -(17)Donating
a(0)=a0
and *(0)=*o, the solutions for the first order approxi-mations are obtained as(18) 2 131+192a P3t P4 kn [ 2 (19)
*(t) =
11)0 2P2 131+kP2a
0Having obtained a(t) and *(t), the roll angle and the roll angular velocity are expressed as
1 *( 1
0(t) = Os +
le(t)et)
+Tka(t)e-*(t)
(20a)The initial conditions (8) demands that
a(0)e")(°).+ ka(0)e-*(0) =
2(00,0s)
- (21)
.(A-Y)a(0)e0j-k(X+y)a(0)e7*(°) = 0
whence
a(0)e*(°)
(00)(x+y)/A,
ka(0)a-0(0)
=0 -0a)(X-Y)/X.
Eliminating the exponential factor, a(o) is obtained as
[ (0)]2 = -k[(00-013)w/X]2.
The sign of k can now be determined from the requirement that a(0) must be real. Since for the rolling motion A2<0, k
needs to be +1. Hence
ao = a(0) =
(0 -0
o s)w/a A=ia (22)Equations can further be handled to yield (with k=+1)
1 [e
*(0)
a-*(0)] = (d,0 s )/a =7
[ego)
e-11)(0)] y(00-05)/(Aao) = -iy/windicating that *(t) is
a
purely imaginaryfunction
of time, i.e. tp(t)=16(t),,such thatcos 6(0) = a/w , sin
e(o)
= - y w resulting in6(0) = - tan-1 (y/a)
and
- -1 P1.41)2
2.
e(t) = pt - tan
(y/a) t r
Ln [- -2]P2 p +p a 1 .2 o where 2 P; = a 4' 08(34-YY1)/(2a) and. pi = ea(3+2yy1)/(8w2).
Consequently, the first order approximate solution
= Os
+ a(t) cos 6(t)where a(t) and e(t) are defined by equations (18) and (24) respectively.
Further improvement on the solution can be obtained by computing the function u1(a,0) which, following the procedure outlined in the Appendix, is obtained as follows: O 20 a2 s
u14
(a
6) = s 3+Y Y)a + -f(a2 2 -82n2-) cos 28 + 2 -2co (Z2+11 where +0 3 ' ) + 2a2 (E3+n3(3+2y Y)/4, 82= -y1a/2, E2 = w2_4a2
'{(a .34-83n3)
cos 36
(c33-933)
sin 381 (26)a3
= (1+yy )/8 8a/8
E w2.4.2 .2
The improved solution can thus be expressed as
(1)(t) = (1)
s
+ a(t) cose(t)'+
eu1(a,e)
-which will be used to compute the roll response of a ship to a step function excitation under varying initial conditions, damping charac-teristics and excitation magnitude.
3 APPLICATION OF THE METHOD
To analyse the effect of roll damping on the peak roll response of a ship under the action of a sudden wind gust, the method described in the previous part has been applied for the calculation of gust response of a stern trawler. Principal particulars of the ship and the
coefficients of the equation of motion are given in Tables 1 and 2 respeCtively. Computations have been carried out up to first order accuracy, including the correction function su1(a,0). Variation of the functions a(t), 0(t) and eu1(a,0) of the solution
s + a(t) cos8(t) + 00-1(4iP)
with time for a sample calculation CMg=4965, 00=-0.4) are illustrated in Fig.2.
To make calculations indicative of the effects of the variation of roll damping and of gust magnitude four values of Ni, N2 and M have been used while keeping the rest of the coefficients constant. Computations indicated that for this particular ship and the loading condition
variation of N2 had only a marginal effect. Variation of N1, however, indicated the strong dependence of the peak roll response on the value of roll damping moment. Figure 3 illustrates these results for M =198.6.
In this figure 0101 indicates the fesponse angle obtained by the energy balance method (see Fig.1). An examination of the response curves and the comparison of peak response values with corresponding 0102 shows that ,the maximum difference reaches to 0.193 radians (-11 degrees) sufficient
to make a ship stable or unstable in a weather criterion evaluation. An important point to notice here is the increase in the role of roll
damping with increasing initial roll angle.
Computations carried out with different wind gust moment magnitude confirmed the above trend and indicate
a
difference of .5to
10 degfees it the peak response Values compared. to the corresponding Om values. Furthermore the differences increased with increasing gust moment magnitude.4 CONCLUDING REMARKS
From the limited study presented here one can clearly notice the impor-tant effect of roll damping- on the peak response of a .ship to the sudden action of a. wind gust. Since one Li interested in the response peak gust duration has not been included in the study. If however gust duration is less than the.half of the natural roll period the over prediction by the energy.balance Method will become more pronounced.
In the light of this investigation it is suggested that if the gust response has to be included in a-weather criterion the tole of damping
should be taken into account.
ACKNOWLEDGEMENT.
The authors express their thanks to the Council of the British Ship Research Association for permission to publish this paper and to the Department of Trade (UK), Marine Division as the financial sponsor of the work on which it is based.
REFERENCES
Intact Stability, Including Analysis of Intact Stability -Casualty Records - Weather Criterion. IMCO Document STAB
27/5/4, Submitted by Japan,
1982.
Intact Stability, Including Analysis of Intact Stability
. .Casualty Records,.- Weather Criterion. IMCO Document STAB
27/5/3, Submitted by USSR, 1982.
Asymptotic Methods in the Theory of Non-Linear Oscillations. BOGOLIUBOV, N.N. and MITROPOLSKY, Y.A. Hindustan Publ. Corp. Delhi, 1961.
Hydrodynamic Reaction to Large Amplitude Rolling Motion.
ODABASI, A.Y. Int. Shipbuilding Progress, Vol.28, p.74, 1981.
T[5] A Unified Krylov-Bogoliubov Method for Solving Second-Order Non-Linear Systems. MURTY, I.S.N. Int. j. Non-Linear Mech., VO1.6, p.45, 1971.
[6] On Perturbation Methods. Involving Expansions in Terms of a
Parameter. BELLMAN, R. Quart..
of Appl.
Math. Vo1,13, p.195,Table 1
Principal Particulars of the Sample Ship
* Displacement
mass
Table 2
Coefficients of the Equation of Motion
Ship Stern Trawler
Lpp (m) 56.85 B (m) 12.20 D (m) 6.8 T (m) 4.10 Trim (m0-aft) 1.20 CB 0.515 A (tonnes) 1567
Ship Stern Trawler
63555 0 Ni 6172 8 N2 10735 A 10545 1316 N. 198.6
FIG. 1
DEFINITION OF GUJASI-DYNAMIC ROLL ANGLE
0 8
r
0.4
0.7 I-O. 3 Ox 1ORI
, A', EU x10 20.6-0.2
. 0.5-0.10.4 -
0. 3
-- 0. 1/
0.2
--0.2
. 0. -O. 1- 0.4
./
= 63555.0
NI = 6172.0
=10735.0
A =10545.0
B=1316.84
Mg = 496.5
\ 00 = -0.4
I.
\
12 \
14 16/
--
0. 5FIG. 2
VARIATION OF THE SOLUTION
COMPONENTS WITH TIME
-e
AEU
/
18 20time
(secs)0-3
0
dians)
0.2
01
O3-0.1
-0.2
-0:3
-0.4
0diant)
0 .2 0.1 0.1 0 .2 12N1= 51440
N1 = 77160 time 20 (secs) time 20(secs)0-3
(radians)
0.2 0'1 0-02
0.3
0.4
03
(radians)
0.2 0-2 = 6172.0FIG. 3
EFFECTS OF ROLL DAMPING
AND INITIAL ANGLE
ON GUST RESPONSE
12 N1 = 9259-0time r
0 (secs) I time r 20 (secs' sfro(radians)
(radians)
orn-0-1
01376
-0-2
0.2386
-03
0.3380
-0.4
0.4389
01
-0-1
.0'3-0.4
03
0
'ad ions) 0.2 Mg = 397.204
0.3
0
(radians)
0.2
O!1 0.102
-03
_ - 0-4AMrLJMITI AI
t mni
P z_ern(radians)
,.jvg 198.6
297. 9 397. 2 496. 5-0.1"
0 .1376 0-15610 -1760 0 .1942
-0.2
0. 23860- 2564 0.2757 0-2949
-0.3
0-3380 0.3568
O. 3770 0.3950
-0-4
0.4389 0-4571
0 .4767 0 -4962
APPENDIX
THE METHOD OF SOLUTION
Let the equation of motion of a single-degree-of-freedom non-linear system subject to step function excitation be
X + 2yic + w2x + ef(x) = Mgt)
(A:1)
where x.=x(t) is the system response, y, w and W are positive
constants. e is a small parameter, an over dot denotes differentiation with
respect to time, U(t) is the Heaviside step function defined by
0 for t 0,
U(t) = (
1 for t O.
(A,2)
and f(x,k) is any analytical non-linear function of x and
As the time tends to
infinity
any stable motion will tend approach to stable position, i,e. x(t)-oc.s as t4co, where acceleration and velocityvanish, i.e. X4.0, 3140 as t-).,=. The value of xs can be determined from
2
xs + ef(xs, ) = W.
The'solution is sought in the following form
x(t) x +
e*
+ k --e-* +s 2 .2
(A.3)
l(a,0) + e u2(a,*) (A.4
where a(t) and
OW
are time varying amplitude andexponent, respectively, ank k can take values +1 or -1 depending on the conditions. The functions a(t) and *(t) are expressible as
da dt la
e41(4)
+ ek2(0.) !=1, = A + eB (a) + e.2B2(a) + . dt where .2 2 = w )Substitution of (A04) and (A05) into the terms
M+2r142x
yields R+.2y11+w2x=w2x0+lee/q-yaAi+(y+2X)A1-ya2B1+2XaB1l 2 1 u -ikee-/qyaA1,-(y-2X)Acya2B1-2iay+el(yar" 2 aa BMa* 2u 2u .,2Aya1 +2
D1 +
2)ty4 + w-u1} +
0(e.2)2 .
IP2
au
Where 0(e2) denotes terms of second and higher
order
powers of e.' The non-linear term can be expanded into a series in terms of e as2raf dx af clic]
.
ef(x,i) = E[f]
0(2)
(A7)e 'ax de
a
de c=0iC0fi = f0 = f(x + ee*
+ kae,
(A-y)ae* - x= (A.5)
(A.6)
ik(X+y)ae-*). (A.8)au
Y a aaSubstituting (A.7) and A.8) into (A.1) and equating the coefficients of E we obtain , 2 2 2-
aul
3u1
2 a u1aui
2 (ya) -3a2 aay a 2Aya isT-p- + X + 2Xy - +
w ul
alp
42
= - ya2
-.A2XaB1
+ iitel[yaAl (y-2X)
-
ya2B1 - 2XaB1] + f(xs,0) - fa.(A.9).
By using the definition Of thethe difference
fo-if(Xs,-0) can
be expandedas
fo - f(x8,0) = F(a) + E [E.
-We"
+ (a)e-1141n=1 n
and substitution of this expansion in (A.9) Yields
F (a) + elqyaAi-(y+2X)A1+ -'2XaB1 -2F (a)1+ike'llyaA!
1
(y-2X)A1-ya2B;72XaB12kG1(a)]- E [Ft(a)en4)+Gt(a)e 4=2
In order to avoid the appearance of secular terms in u1(a0P) the coefficients of e in (A.11) are equated to zero
ya.A! - y+2X)A1 + 1a2Bi - 2XaB1 = 21' (a)
1 IP) (A,10) 2 2 a2u
3u
i
23u
- Y2a 1 3u1 2 .1(ya) 2Xya - + 2Xy + w u =
.aa .2 Ba Da34, 1 r! tr.
ft
I
It
from which the unknown functions Al(a) and
B/(a)
are determined. The rest of the equation (A:11) yields a partial differential equation for the determination of the unknown function u/(a.,*)2 ,2 ,2
2 a u1 2 au1 ° ui 2 _ 1.02u
(ya
2 a aa -2Ala Ma* + 2Al alp +
aa.
= -F0(a) - E (F(e)e' + 011(4)e-"].
n=2
One may assume that u1(a.,0 can be expressed as
u1(a'V) = v
0(a) + 1.Iv
(a)en* + w (a)e-1141.
n=2
and the equations to yield the unknown functions vo(e), 7(a), %(a) are obtained by substituting the expansion (A.15) into (A.14).
vn(a)=Fn(a),(n=0,2,3,..),
- (ya)2v:;(a)-y(y+2nA)av;( ) +(n2A2+2nyX+w2
_ 2 " 2 2 2
(ya) wri(a)-y(y-2nA)aw;(4)+(n A =2,nyX+w )wl'a(a)=-Gn(a),(n=2,3,...);
Although the method presented may be thought of as a small patameter expansion technique and hence only
valid
for small values ofe,
this in fact does not pose any serious drawback. Following Beilmen[6] anOther .perturbation parameterv may
be defined asv.= c/(1+e)
which for all positive c takes values between 0 and 1, i.e. 0.1p<1.
Therefore all the previous analysis can be repeated with c replaced by p .since
2 3,
E =
mu+ .11
+ o(p3)which implies that the first order problem in p coincides with the first order problem in e (e being replaced with p) and differences starts to exist in higher order approximations, i.e.
gf(x, "o+p2+....)[(41,0 + p(df/dp)11.0 +