r
nl rod u ct ion
A ship at rest in a calm sea ¡s an archetypal
example of a dynamical system in stable equilibrium. The system will return to this position, dynamically,
given some small perturbation due to the combined
effects of (positive) damping and restoring force. For small-amplitudes the comparison with the free response of a simple spring-mass-damper mechanical oscillator is clear. A further analogy extends to consider the motion of a ball rolling on a locally parabolic track, where the shape of the track represents the underlying potential energy of the system. Trajectories in the phase portrait spiral towards the equilibrium position, and all possible initial conditions cause transients which decay on lo
this point attractor [1].
The static stability of a floating vessel is typically limited by the angle of vanishing stability resulting from the softening spring characteristic of the GZ curve. For a symmetric hull geometry these
laránle unstable equilibria are represented by
saddle points in the phase portrait [2J. The unstable equilibria give rise to sparatrices which now divide
the plane of initial conditioThT domains of
attraction. Transient motion may no loh7 be captured by the attractor and initial conditions lying
outside of the domain of attraction or basin boundary
will lead to unstable, growing oscillations. This is a familiar scenario from ship dynamics where a floating Vessel inclined to a large angle will lose its stability
resulting in a rapid roll to an inverted position [3]. This is properly viewed in the two-dimensional phase
Portrait of displacement and velocity so that it can be Seen that ¡t is the combination of roll angle and roll Velocity that may
Abs tract
This paper develops an ad hoc criterion for stab/e rolling motion of a floating vessel ¡ri
regular waves under slowly changing conditions. The total energy associated with the angle of
vanishing stability ¡s compared with the total energy of the harmonic response. Incorporation of a factor of safety enables the determination of critical forcing parameters which delineate 'safe regions where stable rolling motion persists. An alternative measure of stability is also presented
which is applied to an idealised version of a biased floating vessel and the results are compared with
the dynamic instabilities obtained from numerical integration in previous studies.
Assistant Professor Scho0i of Engineering Duke University
Durham, NC 27706, USA.
A SIMPLIFIED LOWER-BOUND CRITERION
FOR STABLE ROLLING MOTION
Lawrence N. Virgin1
45
-In this unforced (autonomous) case the effects Th rapidly applied loads, for example modeled by
impulse, pulse on step inputs, can be assessed. This has obvious relation lo ocean wave slam loading. ThIs
paper is restricted to the consideration of periodic excitation which might anise for example as a regular wave train from a distant storm. Furthermore, the
forcing parameters are assumed to change slowly so
that transient effects are negligible [4j.
Returning to the analogy of the ball rolling on
the potential energy surface, the angle of vanishing
stability is manifested as two 'hilltops' on either side of the stable 'well' at the origin. Under the action of
harmonic excitation, the steady-state forced response
will depend on the magnitude of the forcing amplitude
and frequency, including the familiar resonance feature [51. This is akin to oscillating the track on which the ball is rolling; a horizontal projection of
this nonautonomous behavior now occupying a
three-dimensional phase space with lime as the third axis. It
is clear that under certain circumstances, especially proximity to resonance, the forcing may be such that the ball will 'escape' from the potential well
corresponding to a capsize situation.
For large-amplitude motion the nonlinear restoring force causes a variety of complex instability
phenomena preceding escape [6]. The total energy of
the response increases with the amplitude and this is
used as a measure of proximity lo capsize, i.e. the
total energy (kinetic and potential) of the response is
required to be somewhat less than the total energy
associated with the angle of vanishing stability. A safety factor is incorporated to ensure that this criterion is satisfied in terms of the forcing
parameters. Previous studies have indicated where
typical ordinary differential equations lose their stability leading to escape [6,71. The present paper outlines a simplified ad hoc criterion based on
considerations of limiting the tota energy of a system, and deriving conditions under which stable rolling motion persists.
ree ro ¡ng motion
n order to convey the underlying principle of the new criterion consider the free roll motion of a
floating body described by
O=f(O,O).
(1)Initially damping is ignored so that the ordinary differential equation of motion is of the form
OV'(0)=O
(2)where a prime denotes
differentiation with respect to
the roll angle O and a dot denotes differentiation with respect lo time t.
A first approximation to a typical restoring force (GZ cwve)
experienced by a body inclined to an angle O is
V'(0)=O-a&
(3)This type of system is often referred to as a softening
spring charijc.
lt can be seenthat equilibrium
points correspond to stationary values of the underlying potential energy, i.e.
OeO,0e±
(4)The saddle pointsassociated with the second two
equ)bra of equation (4) correspond to
teã?igle of
vanishing stability.
Returning to the analogy of the rolling ball, the
underlying potential energy curve is given by
V(9)=-O2-O4+c
(5)and trajectories in the phase portrait are given by contours of constant total energy, i.e.
E= +V
(0)
-4
+c
(6)where the constantdepends on the initial
conditions. The value of the total energy at the saddle points is
E=
i (
+ i2
-where the arbitrary constant c is taken as zero, giving
E=
4 (8)
The restoring force of equation (3) is shown plotted in
figure i(a) together with the underlying potential
energy function of equation (5) in figure 1(b). The phase trajectories for this conservative System
describe ellipses as given by equation
(6) and are illustrated in figure 1(c).
The safety factor p can be incorporated into the
analysis to give a limiting total energy value of
E=--4a
(9)-
46
-V 1.0 O 1.0 FFigure 1. (a) Restoring force; (b) potential
energy;
and (c) phase portraitfor a softening spring. The ad hoc boundary for stable rolling is shown for p=O.8.
(cr=l.0, ¡3=0.0)
and this is shown in figure 1(c) for p = 0.8. In other words, the motion of
the vessel should remain at least
within the shaded region for stabìlity.
Forced rolling motion (Linear)
Harmonically excited roll motion can be approximately described by
where the forcing amplitude F andfrequency u can be
thought of as
sinusoidally perturbing the potential
energy function.
For relatively small amplitude motion the nonlinearity in the restoring force is negligible, so that the roll motion can be described by
and the steady-state
response is given by
9=Asin(wt-,)
(12) where(+L
(7) -LO 1.0 eand
wj3
tanq=
- ('y.
The familiar amplification of the response near
resonance (w = 1) can be seen.
There are a number of ways in which this response can be compared with the static stability features of the underlying GZ curve [4]. Firstly, the
maximum total energy of the response can be used as a 'measure of proximityto the total energy associated
with the angle of vanishing stability. Later, the
amplitude and maximumvelocity of the forced response is used as a measure of proximity to the amplitude and maximum velocity associated with the ingle of vanishing stability.
Maximum total energy bound
The total energy associated with the forced response is
where O is given by equation (12). Therefore, for
a
given F and f3 the total energy will depend primarily on w, i.e.
E=-(A o)2cos2(ot) +-A2sin2(ut)
(14)
which, for low damping, is a maximum at
resonance
(w = 1), giving a maximum total energy of
E4A2
.1 F2
2p
(15)For example, for a damping coefficient of
f3 = 0.1, nonlinearity parameter ofCZ = 1, and a safety factor of p = 0.8, a forcing amplitude of F = 0.0632 would result in an oscillation whose maximum total energy
is equal to a prescribed proportion of the total energy
associated with the angle of vanishing stability, i.e.
.L F2<P
2 p 4cx (16)
A forcing frequency less than or greater than the resonant value will result in an oscillation with less
total energy and therefore would require a greater level of forcing to violate the condition of
equation (16). With co=0.8 Fc=0.233 and with w=1.2, Fc0 .2405.
Figure 2 shows the maximum total energy of the forced linear response (equation (14)) for ¡3
= 0.1 and F = 0.05 , 0.07. Also plotted on this diagram is the total energy associated with the angle of
Vanishing stability (equation (9)). This illustrates
that the limiting condition is reached between F=0.05 and F0.07, thus confirming the value of Fc = 0.0632 for equation
(16). The relative contributions of the displacement and velocity terms in equation (14)
depend on w but the light damping present in the System results
in the symmetric nature of the curves.
(13)
47
-F0.07
F=0.05
However as the amplitude of the motion grows
then tile_effect of the noni
rriing force wHI
grow. In this case the response must be based upon the
sotutions to equation (10) rather than the assumed form of equation (11). The criterion of equation (16)
can be summarized as
Energy (response) < p Energy (instability)
The next section obtains expressions for the total energy of the response based on nonlinear (i.e. larger
amplitude) motion.
Forced rolling motion (Nonlinear)
The use of small oscillation theory which leads to the linear response of equation (12) results in amplitudes which would simply grow indefinitely with no instability. For consistency the forced response
must be based on the same equation as the unforcedor
free response with the addition of external excitation: i.e. equation (10) must be solved to obtain the
nonlinear response which can then be compared with
the approach to instability. An approximate solution can be obtained to equation (10) using various methods [8]. Here, a harmonicbalance approach is
used whereby a solution is assumed of the form
0=asinot+bcosot
(17)which can then be substituted into equation (10) and
after equating sine nd cosine coefficients leads to the
following algebraic equations
-ao)2- b o
+a-3-a
--a b2r0
(18)
which can be rearranged to give
F2 =A2[ (i
- o - A2)2+ (o)2]
(19)where A is the amplitude of the response, i.e.,
A = Y a2 + b2 (20)
Given F and ¡3, equation (19) can be solved for
A in terms of w. Figure 3 shows typical amplitude
response diagrams with F=0.2 and F=0.3, and ¡3=0.6.
0.5 0.8 1.2
i 1.5
Figure 2. Variation of the maximum total energy with frequency for the linear response of a harmonically oscillating system. (/3=0.1).
E
max
0.2
0.4
0.2
The familiar softening spring effect can be clearly seen although with this relatively high level of damping the jump in resonance feature is not
observed. A
0.0
E 0.2 0.1Figure 3. Amplitude response diagram for the ful/y nonlinear system. (/3=0.6).
Again the total energy associated with the
response is given by equation (14) where now the (nonlinear) frequency-amplitude relationship is governed by equation (19). Figure 4 illustrates the total energy of the harmonic response for a=1, ]3=0.6 and u=0.8 through a complete cycle of motion. lt can
be seen that the maximum total energy occurs at
(o. = fl211 (21) 0.0
...r
0 2/
F=0.31 4 6t-
) radFigure 4. Variation of the total energy of the harmonic response through a complete cycle. (/3=0.6, a=I.0, a)=0.8 and F increments by 0.02 from 0.21 to 0.31). The curves in this figure start with F=0.21 and increment by 0.02 to the value of F=0.31 giving a maximum total energy greater than the E=0.2 obtained from equation (9). This is confirmed by the upper curve in figure 3 reaching amplitudes greater
than
E=A2, i.e.
A =0.632
corresponding to a safety parameter of p=O.8. In this way, the maximum total energy of the
nonlinear responsecan be obtained for different
values of e given F and ¡3. This is shown in figure 5 for
p=O.6 and p=0.8. lt can be seen that a larger forcing
amplitude F can be tolerated for p=0.8 than for p=O.6 for the total energy criterion to be met. The curves in figure 5 thus represent levels of forcing which ensure
that the magnitude of the response isconstrained to
remain somewhat below the loss of stability [9].
48
-/
P=0.8 0.40.2 0.6 1.0
So far no mention has beenmade of the mechanisms by which a system such as equation (10)
will typically lose its stability. This is a subject of current research. lt is important for the curves of figure 5 to be beiow the forcing levels at which these
nonlinear phenomena occur in order for the total
energy criterion to be lower bound and useful [10]. As mentioned earlier it is also possible to limit the
response by the amplitude and maximum velocity associated with the tota! energy of the angle of
vanishing stability. This approach has been successfully used by the author and is briefly
introduced here to show that such ad hoc criteria can
ensure stable rolling motion [4].
Amplitude and maximum velocitybound An alternative measure ofproximity to instability was developed in reference [4]. The separatrix joining the saddle points in figure 1 is
gé5equation (6) with
c=-1/4a. Rearranging equation (6) leads to= ± [
. e +
2(22) where p is again used as a safety margin.
From this alternative form for constant total energy it is possible to obtain the maximum velocity as
_+p
max-2aJ
and the maximum displacement
I ±(1
The alternative criterion for stable oscillations now requires that the maximum velocity and amplitude of
the forced response to be less than the maximum
velocity (equation(23)) and maximum displacement (equation (24)) respectivey for the undamped,
unforced conservative system.
Assuming the linear forced rolling model of equation (11) leads to the following condition to limit the amplitude
FC-I-1-(
-
I ±( l-p) fl(1 (25) and the maximum velocityFigure 5. Limiting values of the forcing parameters
(/3=0.6, a=1.0).
p0.6
0.2 0.8 1.4 0.3 0.2rwhich
are plotted in figure 6(a) for =0.1, p=O.8 and ct=1.0 (for the unforced system), cx=0.0 (for the
forced system). The response of the nonlinear equation
(10) leads to the condition limiting the amplitude as
0.8 F 0.6 0.4 0.2
F=
Fc=[
L 2aw2 0.5 1.0and limiting the maximum velocity as
-L
I p (( p \2 2\
+(f3o)
L2ao)2
82J
which are plotted in figure 6(b) for 3=0.1 p=O.8 and
cx=1 .0, now for both the unforced and forced systems. The effect of the softening spring can now be observed,
resulting in frequencies around w=0.75 producing oscillations with the greatest energy for fixed forcing
and damping. The two conditions of maximum
displacement and maximum velocityare both required
because of the effect of the frequency on the response.
This s related to the relative contributions of the two terms on the right hand side of equation (14).
A static bias, due perhaps to a constant wind loading or shifted cargo has been shown to have a marked effect on the dynamics of a rolling ship. This effe,ct can be incorporated into the dynamic model with the addition of a constant term in the equation of motion. The criteria suggested in this paper can
Include this effect with relative ease. Bifurcational behavior
Considerable research has been conducted into
the loss of stability of forced mechanical oscillators
[1]. These inherently nonlinear phenomena can be 1.5 (4, w
i-F={I±(1-p)(1o2-
1±(1p)))2+(Ú))2)]2 (26)Figure 6. (a) Stability criterion based on the assumption of linear resonance. (ß=O.1, p=O.B). (b) nonlinear resonance.
2.0
(27)
(28)
49
-investigated in one of two ways. Firstly, an analytical method such as the Routh-Hurwitz criterion or Floquet theory [9] may be used to study stability characteristics of small perturbations about the steady-state solutions as described byequation (17).
For example, equation (10) has been shown to lose its
stability close to resonance in one of two generic ways:
a jump to resonance at a saddle-node bifurcation; or a
flip bifurcation leading to subharmonic oscillations initiating a sequence of period-doubling bifurcations
leading to chaos and escape 16]. For slowly changing conditions both of these events have been shown to
occur after the critical conditions of equations (27) and (28) have been reached, with a suitable choice for
p.
The second approach to studying the dynamic
stability of nonlinear oscillators involves extensive
numerical integration. As an example of an asymmetric mechanical oscillator consider the following equation:
O+3O+O-O2=Fsincot
(29)which has a potential of
V(8)=--9--+c.
(30)Solutions to this equation may 'escape over the lower
of the Iwo hilltops analogous to figure 1(c). Figure 7 shows instabilities obtained by the numerical integration of equation (29) which lead to escape or capsize, together with the proposed ad hoc criterion
suggested here [4]. For a given damping ratio and
safety factor, forcing parameters within the shaded region have been shown to lead to escaping
solutions[6J. The ad hoc criterion of equations (27) and (28) is also plotted indicating a conservative estimate of stable rolling motion. Instabilities such as a relatively small jump in amplitude may occur below
these forcing levels but do not lead to escape for this
system. For further details on these instabilities see reference [4]. F 0.3 0.2 0.1 - - Numerical bifurcations Approximate -criterion p=O.8 ß=0.1
Figure 7. A comparison between the maximum
velocity and amplitude bounds and the numerically determined bifurcations leading to escape (foran unsymmetrical system). (ß=O.1, p-OS).
!ConclusioflS
research into a possible new criterionThis paper describes the early stages offor antifyingthe stability of a rolli vessel in a regular and slowly
cfanging environment. Although the suggested method isbased on simplistic equations and takes no account of
transient motion it is a useful way of introducing important dynamic effects: in contrast to much of the existing stability criteria based on static concepts [11]. A particular advantage ofthe proposed method is the inclusion of a factor of safety which can be
adjusted to different circumstances and improved information. Further developments would include coupling with other degrees of freedom, improvements
n the modeling process, more realistic forcing conditions and incorporation of transient motion effects to take account ofwave conditions which are not slowly changing.
Acknowledgement
The author would like to express his
appreciation to Raymond Plaut of Virginia Polytechnic Institute and State University, and Michael Thompson of University College London for useful discussions.
References
Thompson, J.M.T. and Stewart, W.B. Nonlinear Dynamics and Chaos, 1986, Wiley:London.
Caldwell, J.B. and Yang,Y.S. Risk and reliability analysis applied to ship capsize : a preliminary study.
Proceedings. International Conference. The Safeship
Proiect : Ship Stability and Safety 1986, R.I.N.A.
50
-Kuo, C. and Odabasi, AY. Application of dynamic
systems approach to ship andocean vehicle stability.
Proceedinas of the International Conference on
Stability of Ships and Ocean Vehicles 1975,
Stia thclyde.
Virgin, L.N. Approximate criterion for capsize
based on deterministic dynamics. Dynamics and Stability of Systems, 1989, 4, 55-70.
Bishop, R.E.D. and Johnson, D.C. The Mechanics of
Vibration, 1960, Cambridge University Press.
Thompson, J.M.T. Chaotic phenomena triggering the escape from a potential well, Proceedings of the Royal
Society, A421, 1989,
195-225.Virgin, L.N. The nonlinear rotting response of
a
vessel including chaotic motions leading to capsize in
regular seas. Applied Ocean Research 1987, 9,
89-95.
Hayashi, C. Nonlinear oscillations in physical systems. McGraw-Hill:New York, 1964.
Virgin, L.N. and Cartee, L.A. A note on the escape
from a potential well. International Journal of NonLinear Mechanics submitted.
Welticome, J.F. An analytical study of the
mechanism of capsize. Proceedings of the International Conference on Stability of Ships and Ocean Vehicles, 1975, Strathclyde.
International Maritime Organisation. Intact stability criteria for passenger and cargo ships, 1987,London, IMO