Professional Documents
Culture Documents
Technical Report
Technical University of Denmark Dept. of Mechanical Engineering Coastal, Maritime and Structural Engineering
Executive Summary
The occurrence of the parametric roll phenomenon is related to the hull forms that experience large volumetric changes in the submerged portion during the wave passage, predominantly the ships with large bow and stern flares and fine underwater hull (such as containerships). Parametric roll behavior may lead to large roll angles experienced by the ship typically in longitudinal waves. The phenomenon may induce costly ship operation problems; in the last decades serious accidents of parametric rolling were reported that have resulted in loss and damage of cargo (France et. al., 2003; Hua et al., 2006). Due to its practical importance and modeling challenges the phenomenon has been an attractive research subject during the last years.
The extreme state of APL China as she reaches the Port of Seattle in November, 1998
The goal of this study was to improve the ship parametric roll model and to reflect on the possibilities for practical application. In this study the roll motion of a ship has been modeled by a 2 DOF hydrodynamic procedure, where roll and surge motion are coupled through the frequency of encounter. The speed variation due to e.g. added resistance has been accounted for in the model by the surge velocity. A time domain routine for the calculation of roll restoring righting arm (its time-variation being the dominant parametric excitation in the system) based on exact underwater hull geometry has been implemented in the model. The ship roll motion due to a regular wave critical for parametric roll occurrence has been simulated, as well as the ship roll response in a severe stochastic sea. It has been shown that the instantaneous restoring characteristics calculated via the detailed hull geometry are important for time simulation of roll. For the practical application, the hydrodynamic model needs to be relatively simple. The model linked to a probabilistic tool will return the probability of an extreme event, based on the available wave information on the route. The importance of restoring lever
calculation method in parametric roll probabilistic assessment has been considered in the study. The applicability of FORM (The First Order Reliability Method) to the prediction of the probability of parametric roll has been investigated and predictions have been compared to the results by Monte Carlo method. The satisfactory agreement between the results by the two methods has been found probabilities of the extreme roll event remain of the same order of magnitude. FORM requires by far the shortest computational time and therefore may be very beneficial for the use in on-board decision support systems where the response is to be simulated in real time. The present method for parametric roll prediction represents an efficient tool that can be used in early ship design stage where the accuracy in the deterministic method is important in order to satisfy given load boundaries. When several iterations are needed it is quite easy to step back to the hull form and revise it. Regarding the probabilistic assessment of roll in stochastic sea, recommendations are given for the feasible approach.
Contents
Executive Summary.2 1 Introduction5 2 Hydrodynamic Model.
2.1 2.2 2.3 2.4
..7
Literature Overview..7 1-DOF Roll Model8 2-DOF Roll Model...10 Calculation of the Roll Restoring Righting Lever... .11 2.4.1 Approximation of GZ curve. 11 2.4.2 Instantaneous GZ curve.....12 2.5 Effective Wave Concept.. ..13 2.5.1 Irregular Sea Application...14
3 Probabilistic Assessment..15
3.1 First Order Reliability Method Introduction.15 3.2 Design Point and Reliability Index..15 3.3 Mean Out-Crossing Rates and exceedance probabilities.17
4 Numerical Results...18
4.1 4.2 4.3 4.4 4.5 4.6 Vessel Characteristics...18 1-DOF Method Used by ABS...18 2-DOF Method Regular Wave Analysis....20 2-DOF Method Irregular Sea Analysis...26 Reliability Analysis Influence of GZ Calculation Method.29 Reliability Analysis Influence of Variable Ship Speed..33
Chapter 1 Introduction
Parametric roll phenomenon represents sudden increase in roll amplitude in longitudinal sea caused by the parametric roll resonance (principal or low cycle resonance i.e. excitation wave frequency equal to twice the roll natural frequency). Parametric excitation comes from the time-varying roll restoring characteristics in waves that may be quite large for fine and flared hull forms. Motivated by the loss of cargo due to the parametric excitation, the International Maritime Organization (IMO) issued lately an operational guidance to the master to prevent dangerous situations (IMO, 2007). This will certainly help the ship masters raise the awareness about parametric roll phenomenon. Masters who do not identify the
right conditions for parametric roll have been in for a huge surprise, with consequent loss of cargo and danger to the ship and crew, says Knut A. Dhlie, DNVs business director for container ships, at the DNV Greater China Technical Committee meeting. (The source: DNV Container Ship Update No. 1, 2006)
Current IMO intact stability code has been mainly associated with the hull forms operated in 1960s. The need to synchronize stability criteria related to novel ship designs with the latest achievements in the research of dynamic intact stability physical mechanisms has been indicated. The IMO working group proposed the framework for the development of the new criteria on the meetings in 2007(SLF 50/WP.2, 2007). A revision of the intact stability criteria has been discussed in e.g. the document submitted to SLF 51 (IMO, SLF/4/3, 2008). The new criteria should include roll restoring arm variation problems such as parametric excitation. The state of the art in methodology development and regulations in assessment of ship intact stability can be found in Francescutto (2007). A very detailed explanation of the physics behind parametric roll has been given in the ABS assessment of parametric roll in the container ship design (Shin et al., 2004) where the importance of the GZ stability curve variation for parametric roll onset has been emphasised. The stability of a ship is reduced on a wave crest. On a wave trough, the restoring moment is increased over its magnitude in still water and the ship rolls to the other side with an increasing roll angle with time, passing through the vertical upright position. This time variation in ship stability might give rise to a parametric roll excitation that is only limited by the ship non-linear restoring characteristic and the roll damping. The parametric excitation can be further diminished if the critical wave elevation dies out in the train of irregular waves. In this study, the influence of GZ calculation on the simulated roll response has been investigated. An overview of several efficient methods for roll prediction will be given. The present 2-DOF method, based on roll and surge coupled dynamically, is well suited for roll predictions in longitudinal long-crested sea. In the present study the container
ship in head sea condition has been studied. Results using different roll models have been compared. Influence of the GZ calculation method as well as variable ship speed on roll predictions in stochastic sea has been shown accounting for a number of critical sea states. Probability of extreme event has been discussed, aiming on the practical application.
1986) or lately in the procedure by Bulian et al. (2006). Using a 6-DOF prediction procedure De Kat and Paulling (1989) recognize the significance of the critical wave as part of the equivalent wave system especially in hydrostatic calculations.
(2.2.1)
where n is the roll natural frequency of the hull and is the actual to critical damping ratio. Instead of the roll natural frequency in the restoring term, frequency related to the mean metacentric height has been used in Shin et al. (2004). If the restoring function f is assumed linear with the heel angle and harmonically oscillating around the mean metacentric height, it takes the form f ( , t ) = (1 + cos( e t ) , where = (GMmax GMmin) /(GMmax +GMmin) which corresponds to the equation given in Shin et al. (2004). After the frequencies are normalized by the encounter frequency and the roll expressed as a function of the new variable x and the damping coefficient such as ( ) = x( ) exp( ) , where = e t , Mathieu type of equation is obtained (ABS, 2004):
(2.2.2)
p = m + 2 e and 2 GM m m = , 0.8B 2 GM a a = , 0 .8 B
q= a e
(2.2.3)
(2.2.4) (2.2.5)
The areas of bounded or unbounded solutions of Mathieu type of equation, depending on the combination of p and q, are given by Ince-Strutt diagram in Fig. 2.2.1. Shaded areas represent stable zones i.e. solution for roll is bounded. If p=0.25 roll natural frequency is about twice the excitation wave encounter frequency or if p=1 roll frequency is isochronous with the excitation wave.
The frequency susceptibility condition is formulated so that the Mathieu parameter p is located in the unstable zone (bounded by the lines approximated to the first instability zone in the Ince-Strutt diagram): 1 q q2 q3 1 q4 1 q + p + (2.2.6) 4 2 8 32 3 128 4 2
n 2 < 0.5qk1 k 2 1 k 3 e
(2.2.7)
where k1 = 1 0.1875q 2
k 2 = 1.002 p + 0.16q + 0.759 q 2 16 + q 4 + 352q 2 + 1024 p k3 = 16q According to ABS guidelines, roll severity has to be checked by the recommended simplified numerical procedure (Eq. 2.2.1) if both frequency and damping inequalities are satisfied (Eq. 2.2.6, Eq. 2.2.7). In the following numerical study roll has been simulated using Eq. 2.2.1 where the GZ curve is approximated by the non-linear 5th order function of the heel angle and wave crest position calculated by recommendation from the guideline, i.e. xc = Vt L pp floor (Vt / Lw ) where floor denotes an integer (2.2.8)
number less or equal to the function argument. The damping coefficient is understood as equivalent linear damping coefficient i.e. using averaging method over one roll cycle 2 (Roberts, 1982) ( 0 ) = 1 + 4 /(3 ) 2 0 + 3 / 8 3 0 (see Eq. 2.3.1), where 0 is the roll amplitude. The above method represents an efficient screening tool convenient in early design stage. Dealing with the ship forms that are greatly susceptible to parametric roll the
method may not be suitable for quantitative study of roll and a detailed procedure with accurate description of the time varying ship restoring characteristics is needed.
where rx is the roll radius of gyration and g the acceleration of gravity. The roll frequency is given by the metacentric height GMsw in still water:
gGM sw rx
3
(2.3.2)
(2.3.3)
The added mass of water in surge is thus taken to be 5 per cent of the mass displacement and the surge force Fx is calculated from the incident linear wave pressure p, hence, radiation and diffraction effects are ignored. p( X , z , t ) = ge kz h( X , t ) (2.3.4)
as
Fx =
0 p( X , z, t ) p( X , z , t ) dV = B( x, z ) dzdx x x V L T
(2.3.5)
The linear wave elevation is represented by h( X , t ) = a cos(t kX ) , where a is the wave amplitude, k is the wave number, is the angular wave frequency. The derivative of the incident wave pressure p( X , z , t ) is with respect to the longitudinal x-coordinate system, measured relative to the ship. This coordinate is related to global X-coordinate through
10
& X (x, t ) = (x + (V + u )t )
(2.3.6)
The vertical integration in surge force (Eq. (2.3.5)) is from the draught T of the vessel to the mean water level and B(x,z) is the breadth variation. The second term on the right hand side in Eq. (2.3.3) is an attempt to model the action of the captain to maintain the constant speed V in waves. Though it is very simplistic, the model given by Eqs. (2.3.1) (2.3.3) is well-suited to illustrate the proposed stochastic procedure as it can model parametric rolling.
where the wave crest position xc is measured relative to the aft end of the vessel. Similarly, the GZ curve in still water is fitted by GZ sw ( ) = A0 sin + A1 + A3 3 + A5 5 (2.4.1.2)
The coefficients (A, C, D) in Eq. (2.4.1.1) and Eq. (2.4.1.2) are found by the least square method.
11
In a stochastic seaway the following approximation of the instantaneous value of the righting arm GZ(t) is then applied:
GZ ( , t ) = GZ sw ( ) +
h(t ) ( GZ ( , xc (t ) ) GZ sw ( ) ) 0.05 L
(2.4.1.3)
The instantaneous wave height h(t) along the length of the vessel and the position of the crest xc are determined by an equivalent wave procedure somewhat similar to the one used by Kroeger (1986): 2 a (t ) = Le 2 x 2 H ( X ( x, t ) , t ) cos Le dx; b(t ) = Le 0 X ( x, t ) = ( x + Vt ) cos
Le Le
H ( X ( x, t ) , t ) sin
0
2 x dx Le
h(t ) = 2 a 2 (t ) + b 2 (t ) Le 2a (t ) arccos 2 h(t ) xc (t ) = L Le arccos 2a (t ) e 2 h(t ) if b(t ) > 0 if b(t ) < 0 (2.4.1.4)
2.4.2 Instantaneous GZ Curve In the present method, the ship has been statically balanced on the free surface, taking into account the roll angle and the variable wave encounter frequency due to surge. The procedure has been explained in details in Vidic-Perunovic and Jensen (2009). At each time instant sectional area and centre of buoyancy have been determined for the ship balanced on the new draught and trim angle during the wave passage and for the given roll angle. The hull geometry has been described for the bare hull, without containers and superstructure.
The procedure to determine instantaneous righting arm is described in the following. The equilibrium equations between the gravity and the buoyancy forces read:
A( x ) dx = 0
0
(2.4.2.1)
And
12
xCOG
xA( x) dx = 0
0
(2.4.2.2)
where A(x) is the sectional area.. The solution procedure starts with an initial guess of the draught and trim. In the first iteration the draught T has been assumed as the design draught T0 and trim as zero angle 0. The draught at each section is calculated by:
T ( x) = T x tan + h( x, t )
(2.4.2.3)
After equilibrium is obtained, the GZ righting arm is calculated as function of time and heeling angle by:
GZ (t , ) =
Mb M g g
(2.4.2.4)
where (Mb-Mg) is the roll restoring moment and the mass displacement . The restoring moment from buoyancy is calculated as:
M b = g rb ( x) A( x)dx
0
(2.4.2.5)
M G = grCOG
(2.4.2.6)
The positions of buoyancy and gravity centre in local coordinate system are described by rb and rCOG. Here the origin of the local coordinate system is fixed to the ship keel and aft perpendicular. This procedure for finding the ship equilibrium position is adapted from Jensen (2001).
13
The methods used in the numerical examples in the following sections are labelled in the following way: Method 1 stands for the exact instantaneous GZ calculation as given in Vidic-Perunovic and Jensen (2009), method 2 denotes the approximation of the GZ curve as done by Jensen (2007) and method 3 represents the calculated GZ curve on the effective Grim wave as done by Bulian et al. (2006).
2.5.1 Irregular Sea Application The JONSWAP wave spectrum has been used for the irregular sea representation in method 1 and method 2: 5 4 g 2 exp r S ( ) = (2.5.1.1) 4 5 (2 ) 4 0 where ( 0 )2 r = exp (2.5.1.2) 2 2 2 0
0.07 0.09
for for
0 > 0
0 is the peak frequency and is the ratio of the maximum spectral density to that of the
corresponding Pierson-Moskowitz spectrum, here taken to be =3.3. In method 3 the irregular sea surface has been represented by the effective wave spectrum (Grim, 1961) as used in the work by Bulian et al. (2006). The spectrum has been determined from the sea spectrum using the following transfer function for deep water application and in head sea condition: kL kL sin 2 f ( ) = 2 kL 2 2
(2.5.1.3)
where L is the length of the ship and k is the effective wave number. The effective wave spectrum Seff() is thus obtained
S eff ( ) = f 2 ( ) S ( )
(2.5.1.4)
14
H ( X , t ) = ( ui ci ( X , t ) + ui ci ( X , t ) )
i =1
(3.2.1)
where the variables ui , ui are uncorrelated, standard normal distributed variables to be determined by the stochastic procedure and with the deterministic coefficients given by
15
= S (i )di
2 i
where i , ki = i2 / g are the n discrete frequencies and wave numbers applied. Furthermore, S() is the wave spectrum and d i the increment between the discrete frequencies. It is easily seen that the expected value E[H2] = S()d, thus the wave energy in the stationary sea is preserved. Short-crested waves could be incorporated, if needed, but require more unknown variables ui , ui . From the wave elevation, Eq. (3.2.1) Eq. (3.2.2), and the associated wave kinematics, any non-linear wave-induced response (t) of a marine structure can in principle be determined by a time domain analysis using a proper hydrodynamic model:
(3.2.3)
Each of these realisations represents the response for a possible wave scenario. The realisation which exceeds a given threshold 0 at time t=t0 with the highest probability is sought. This problem can be formulated as a limit state problem, well-known within time-invariant reliability theory (Der Kiureghian, 2000): g (u1 , u1 , u2 , u2 ,..., un , un )
0 (t0 u1 , u1 , u2 , u2 ,..., un , un ) = 0
(3.2.4)
The integration in Eq. (3.2.4) must cover a sufficient time period {0, t0} to avoid any influence on (t0) of the initial conditions at t=0, i.e. to be longer than the memory in the system. Proper values of t0 would usually be 1-3 minutes, depending on the damping in the system. Hence, to avoid repetition in the wave system and for accurate representation of typical wave spectra n = 15-50 would be needed. An approximate solution can be obtained by use of the First-Order Reliability Method (FORM). The limit state surface g is given in terms of the uncorrelated standard normal distributed variables {ui , ui } , and hence determination of the design point {ui , ui } , defined as the point on the failure surface g=0 with the shortest distance to the origin, is rather straightforward, see e.g. Jensen (2007). A linearization around this point replaces Eq. (3.2.4) with a hyperplane in 2n space. The distance FORM
FORM = min
(u
i =1
2 i
+ ui2 )
(3.2.5)
from the hyperplane to the origin is denoted the (FORM) reliability index. The calculation of the design point {ui , ui } and the associated value of FORM can be performed by standard reliability codes (e.g. Det Norske Veritas, 2003). Alternatively, standard optimisation codes using Eq. (3.2.5) as the objective function and Eq. (3.2.4) as the constraint can be applied.
16
(3.2.6)
can be considered as a design wave or a critical wave episode. It is the wave scenario with the highest probability of occurrence that leads to the exceedance of the specified response level 0.
{u , u }
i i
Z1 = {ui }
Z 2 = {ui }
{u , u }
i i
FORM = min
(u
i =1
2 i
+ ui2 )
(0 ) =
1 2 FORM
1 2 FORM 2
(u
i =1
2 i
+ ui2 ) i2
(3.3.1)
Thus, the mean out-crossing rate is expressed analytically in terms of the design point and the reliability index. For linear processes it reduces to the standard Rayleigh distribution. Finally, on the assumption of statistically independent peaks and, hence, a Poisson distributed process, the probability of exceedance of the level 0 in a given time T can be calculated from the mean out-crossing rate (0 ) : P max > 0 = 1 e (0 )T T (3.3.2)
The FORM is significantly faster than direct Monte Carlo simulations, but most often very accurate. In Jensen and Pedersen, (2006) also dealing with parametric rolling of ships in head sea the FORM approach was found to be two orders of magnitude faster than direct simulation for realistic exceedance levels and with results deviating less than 0.1 in the reliability index.
17
Length L 202 m
Breadth B 32.2 m
Draught D 10.1 m
Mass displacement
34000 t
GMsw
0.012
0.40
0.42
0.83 m
18
Case 1 0.6 0.4 Roll [rad] 0.2 0 -0.2 -0.4 -0.6 0 200 400 t [s] Case 2 0.6 0.4 Roll [rad] 0.2 0 -0.2 -0.4 0 200 400 t [s] Case 3 600 800 1000 600 800 1000
0.6 0.4 Roll [rad] 0.2 0 -0.2 -0.4 0 200 400 t [s] 600 800 1000
Fig. 4.1.2: Simulated roll in a regular head wave (wave and ship condition given in Table 4.1.2) Table 4.1.2: ABS Susceptibility Criteria Wave length Lw Wave frequency w Encounter frequency e [m] [rad/s] [rad/s] [rad/s] [rad/s] [m/s] Case 1 Lpp 0.552 0.661 0.274 0.210 3.510 0.144 0.151 0.198 0.300 0.100 0.048 0.095 [rad] yes 0.55 Case 2 0.95Lpp 0.567 0.690 0.274 0.210 3.770 0.144 0.137 0.202 0.296 0.092 0.046 0.086 yes 0.34 Case 3 Lpp 0.552 0.661 0.274 0.210 3.510 0.330 0.062 0.198 0.300 0.100 0.110 0.094 no -
m a
0.5qk1 k 2 1 k 3
Parametric roll Roll amplitude
19
20
40
60
80
100
120
140
160
180
x [m]
Figure 4.3.1: Regular wave of the length equal to Lpp and the wave height 5% of Lpp (upper figure) and its position along the ship (lower figure).
3 2.5 2 1.5
GZ [m]
1 0.5 0 -0.5 -1
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
Figure 4.3.2: GZ curve in still water (marked by blue and black plus signs). GZ curve for wave crest amidships (marked by circles) and for wave through amidships (marked by triangle signs).
20
In addition, in Table 4.3.1 coefficients in Eq. (2.4.1.1) and Eq. (2.4.1.2) are given for the GZ curve approximated by the 5th order polynomial function.
Table 4.3.1 Coefficients in the analytical approximations for the GZ curves.
A0 -6.286
AI 0.757
A3 3.666
A5 -1.212
The peak in GZ curve for increasing ship speeds decreases because the condition when the wave trough is positioned amidships is not as stable as for zero speed since the encounter wave period becomes smaller due to the speed effect. For the same reason, the wave crest positioned amidships in the case with forward speed represents a more stable condition comparing to the case Lpp = Lw when Vconst=0 m/s. The condition Lpp = Lw is additionally violated continuously in case of oscillating ship speed compared to the constant speed case and therefore the peaks in GZ curves differ somewhat in these two cases for each mean ship speed in analysis (Fig.4.3.4).
1.5
0.5
GZ [m]
-0.5
-1
-1.5 100
110
120
130
140
150
160
170
180
time [s]
Figure 4.3.3: Dynamic stability curves calculated for different constant ship speeds.
V=Vconst=2 m/s V=Vinst (V0=2 m/s)
GZ [m]
GZ [m]
time [s]
Figure 4.3.4: Dynamic stability curves calculated for different mean ship speeds, for constant and instantaneous ship speed.
21
Surge velocity is presented in Fig. 4.3.5 for different mean speeds of the ship. The variation in surge velocity increases with the mean ship speed. The same tendency is present for the long period drift in speed it is larger for larger speeds. Consequently, roll angle becomes smaller for higher mean ship speeds since the variation in GZ curve diminishes with speed. Fig. 4.3.6 represents roll angle simulation for different speeds in regular wave used above. Parametric roll was triggered in this analyzed case. The deviation in roll response, when accounted for the surge velocity, from the case with constant speed is larger for the larger mean speed of the vessel (Fig. 4.3.7).
0.1
-0.1
-0.2 100
110
120
130
140
150
160
170
180
time [s]
Figure 4.3.5: Surge velocity variation for different mean ship speeds.
V=Vconst=0 m/s V=Vconst=2 m/s V=Vconst=4 m/s
110
120
130
140
150
160
170
180
time [s]
Figure 4.3.6: Roll angle calculated for different constant ship speeds
22
0.5
-0.5 150 152 154 156 158 160 162 164 166 168 170
time [s]
0.5
-0.5 150 152 154 156 158 160 162 164 166 168 170
time [s]
Figure 4.3.7: Roll angle calculated for different constant and instantaneous ship speeds
The roll response to the described excitation regular wave is given in Figs. 4.3.8-4.3.11. The roll has been simulated over 1000s. In order to demonstrate the speed effect two different ship speeds have been used in the calculation, V=2 m/s was kept as the constant ship speed in Fig. 4.3.8 whereas in Fig. 4.3.9 the surge velocity has been accounted for with similar results for V=4m/s in Fig. 4.3.10 and Fig. 4.3.11, respectively. The responses are simulated using the present method (method 1) i.e. exact GZ curve calculation, method 2 where GZ has been fitted to the calculation at one wave length by the fifth order polynomial (see Table 4.3.1) and method 3 where the GZ is calculated based on the effective wave concept. Predictions by method 1 and method 2 are similar since the excitation wavelength is the one used to approximate the GZ curves in method 2. Method 3 differs slightly from the two previous procedures and estimates the largest roll angles. This is because the effective wave will implicitly always have the crest positioned amidships which will lead to reduced stability as compared to the general case. The speed variation has clearly a damping effect on parametric roll motion (Fig. 4.3.9 and Fig. 4.3.11).
23
V=Vconst; V0=2m/s
method 1 (present method) method 2 (Jensen 2007) method 3 (Bulian et al. 2006)
100
200
300
400
700
800
900
1000
V=Vconst; V0=2m/s
Roll angle [rad] 0.5 0 -0.5 -1 400
450
500
600
650
700
Figure 4.3.8: Roll response prediction in critical regular wave with the amplitude 5%Lpp by three different methods. The ship speed is kept constant. The response is zoomed in the lower figure.
method 1 (present method) method 2 (Jensen 2007) method 3 (Bulian et al. 2006)
V=Vinst; V0=2m/s
100
200
300
400
700
800
900
1000
V=Vinst; V0=2m/s
-1 400
450
500
600
650
700
Figure 4.3.9: Roll response prediction in critical regular wave with the amplitude 5%Lpp by three different methods. The ship speed is variable. The response is zoomed in the lower figure.
24
V=Vconst; V0=4m/s
method 1 (present method) method 2 (Jensen 2007) method 3 (Bulian et al. 2006)
100
200
300
400
700
800
900
1000
V=Vconst; V0=4m/s
-1 400
450
500
600
650
700
Figure 4.3.10: Roll response prediction in critical regular wave with the amplitude 5%Lpp by three different methods. The ship speed is kept constant. The response is zoomed in the lower figure.
V=Vinst; V0=4m/s
method 1 (present method) method 2 (Jensen 2007) method 3 (Bulian et al. 2006)
100
200
300
400
700
800
900
1000
V=Vinst; V0=4m/s
-1 400
450
500
600
650
700
Figure 4.3.11: Roll response prediction in critical regular wave with the amplitude 5%Lpp by three different methods. The ship speed is kept constant. The response is zoomed in the lower figure.
25
0.5
0.6
0.7
0.8
0.9
1.1
1.2
1.3
1.4
Figure 15: JONSWAP and effective wave spectrum normalized by the peak value, Hs=7m, Tz=7s
26
0.5 0 -0.5 -1
200
400
600
800
1000
1200
1400
1600
1800
time [s]
1
200
400
600
800
1000
1200
1400
1600
1800
time [s]
1
0.5 0 -0.5 -1
200
400
600
800
1000
1200
1400
1600
1800
time [s]
Figure 4.4.2: Simulated roll response by method 1, method 2 and method 3 in irregular sea Hs=14m, Tz=9.5s and the constant ship speed V=3m/s.
27
0.5 0 -0.5 -1
200
400
600
800
1000
1200
1400
1600
1800
time [s]
1
200
400
600
800
1000
1200
1400
1600
1800
time [s]
1
0.5 0 -0.5 -1
200
400
600
800
1000
1200
1400
1600
1800
time [s]
Figure 4.4.3: Simulated roll response by method 1, method 2 and method 3 in irregular sea Hs=14m, Tz=9.5s and the ship speed V=3m/s accounting for the variable speed effect.
28
29
50
150
50
150
0.5
Roll [rad]
-0.5
50
150
Fig. 4.5.1: Critical wave episode amidships (a). Corresponding GZ righting arm (b). Most probable roll when 0=0.5rad, head sea condition. Results have been obtained by use of method 1.
10 Wave elevation [m] 5 0 -5 -10 160 Hs=9m Tz=11.7s V=3m/s (a) 1 GZ righting arm [m] (b) 0.5 0 -0.5 -1 160
165
180
165
180
0.5
Roll [rad]
-0.5 160
165
180
30
1 0.8 0.6
method 1 method 2
0.4 0.2 0 -0.2 -0.4 -0.6 -0.8 -1 0 20 40 60 80 100 120 140 160 180
time t [s]
Fig. 4.5.3: GZ righting arm yielding max roll 0=0.5 rad by exact calculation (full line, method 1) and fitted by 5th order polynomial plus the sinus term (dashed line, method 2).
As the result of the probability analysis, safety index by the three different methods for GZ calculation has been compared in Fig. 4.5.4 for maximum conditioned roll angle taking values from 0.3-0.5 rad. Method 1 is given by circles, method 2 is given by stars and method 3 given by diamonds accounts for GZ evaluated at the effective Grim wave with the crest fixed amidships at all times (Grim, 1961; Bulian et al., 2006). Results have been generated for four different significant wave heights (Hs=7m, 9m, 11m and 14m), zero up-crossing period Tz=11.7s and constant ship speed V=3m/s. Expectedly, the shortest distance to the failure surface has been obtained by the method 3 by which the smallest values for reliability index have been calculated, whereas predictions by use of method 1 and method 2 are very close for different roll angles. The accurate GZ calculation represents the least conservative method of the three, but the calculation time is much longer than utilizing method 2. Further, results for the reliability index by use of the method 2 have been compared to the predictions by Monte Carlo analysis in Table 4.5.1, for the sea state Hs=14m, Tz=11.7s and the ship speed V=3m/s. Good agreement is clearly seen from the results.
31
6 Safety index 4 2
6 4 2 0 0.2
Hs=9m Tz=11.7s
0 0.2
0.3
0.5
0.6
0.3
0.5
0.6
6 4 2 0 0.2
Hs=14m Tz=11.7s
0.3
0.5
0.6
0.3
0.5
0.6
Fig. 4.5.4: Safety index calculated for head sea in different sea states by exact GZ calculation (circles, method 1), GZ fitted by 5th order polynomial plus the sinus term (stars, method 2), GZ by use of the fixed Grim effective wave (diamonds, method 3).
Table 4.5.1: Reliability index by Method 2 and Monte Carlo analysis for the sea state Hs=14m Tz=11.7s and the ship speed V=3m/s Reliability index Reliability index by Monte Carlo analysis Roll [rad] by Method 2 90% confidence interval (by Method 2) 0.3 0.4 0.5 2.4673 2.8285 3.3041 2.4099 - 2.5743 2.7194 - 2.9719 3.2247 - 3.6202
32
In the following results are presented in order to demonstrate that the model of the ship is able to predict parametric roll and after that, safety indices have been calculated and compared using different structural reliability methods for large roll angle events. Using the hydrodynamic model, the response has been simulated from t=0 till t=180s, for number of frequencies n=25. In Fig. 4.6.3 the results of the simulation for the design point corresponding to the maximum roll angle 0.5 rad are shown. The sea state is taken to be Hs=12m and Tz=11.7s, and the constant ship speed as V=10 m/s. The zero-crossing period is chosen such that parametric roll can be expected due to occurrence of encounter frequencies in the range of twice the roll frequency. The influence of the surge velocity on the ship speed has been presented in Fig. 4.6.3 (lower to the right). The surge acceleration is given in the upper right figure. The critical wave episode gets the nonlinear shape due to the time variation in the velocity (upper left). From the ship it looks as if the ship spends more time on the crest than on the through of the wave. The associated most probable roll response is given in Fig. 4.6.3 (lower left). Approximately around the time t=175s the wave crest is located amidships.
33
At this point the ship passes through the vertical upright position i.e. the roll angle is close to zero. The short period speed variation is zero and the acceleration is at its maximum when the ship has climbed the wave crest. The calculation of the safety index has been conducted for the sea state with Hs=12m and the two zero up-crossing periods Tz=11s and Tz=11.7s. Safety indices have been calculated by use of FORM and compared to the results obtained by direct Monte Carlo simulations. Different ship speeds are taken in calculation and roll angles between 0.3-0.5 rad have been considered as they may be relevant in on-board decision support systems. In Fig. 4.6.4 and Fig. 4.6.5 results for the safety index as function of roll angle are shown in case of constant and variable ship speed ranging from 5 10 m/s.
Hs=12m Tz=11.7s V0=10m/s 0.5 Wave elevation [m] 5 0 -5 100 Vconst Vinst 120 140 t [s] 160 180 Surge Acc[m/s 2] H(L/2,t) Vinst
-0.5 100
120
140 t [s]
160
180
1 Roll angle [rad] 0.5 0 -0.5 -1 100 Vconst Vinst Ship Velocity [m/s]
12
10
120
140 t [s]
160
180
Figure 4.6.3. Time domain variation of exciting critical wave episode shown at amidships (upper left figure) conditioned on a roll angle of 0.5 rad at t0=180 s, surge acceleration (upper right figure), roll angle (lower left figure), ship velocity including surge effect (lower right figure).
34
6 4
V=5 m/s
6 4 2
V=7 m/s
6 4
V=8 m/s
6 4
V=9 m/s
6 4
V=10 m/s
Figure 4.6.4. : Reliability index by use of FORM and Monte Carlo method, for sea state Hs=12m Tz=11s and roll angles 0.3-0.5.
Vconst FORM Vinst FORM Vconst MC 6 Vinst MC 4
6 4
V=5 m/s
V=7 m/s
6 4
V=8 m/s
V=10 m/s
Figure 4.6.5. : Reliability index by use of FORM and Monte Carlo method, for sea state Hs=12m Tz=11.7s and roll angles 0.3-0.5.
35
In principle, the following two situations have been experienced during utilization of FORM analysis implemented in PROBAN: The solution is wrong i.e. safety index is found much larger than the one predicted by Monte Carlo method. Safety index by FORM is generally lower than the prediction by Monte Carlo method. The probability of the event is still of the correct order of magnitude. This behavior is more pronounced with the increase of non-linearity in the response and hence, in the limit state surface.
The latter situation is present in the largest number of cases meaning that the accuracy of FORM generally is satisfactory. Similar findings about FORM conservatism have been recently addressed in the work by Yang et al. (2006). An attempt to improve the searching algorithm has been suggested in Wang and Grandhi (1996). The probability level by FORM and Monte Carlo with 90% confidence interval is given in Table 4.6.1 and Table 4.6.2, for the range of ship speeds, for two sea states and three different maximum roll angles where the surge effect has not been included in the model i.e. the ship speed is kept constant at all times. The analogous results have been given in Table 4.6.3 and Table 4.6.4 for the case of instantaneous ship speed, i.e. when the surge has been accounted for in the hydrodynamic model. The comparison shows generally good agreement in order of magnitude, especially in case of constant ship speed, with some conservatism, as already mentioned. Second order reliability method (SORM) is in general more accurate to use since largely curved failure surfaces can be better approximated. The comparison with the SORM analysis for one speed V=6 m/s and the sea state Hs=12m, Tz=11.7s (Fig.4.6.6) shows very good agreement with the other methods in case of constant ship speed and, as expected, better agreement between the SORM and the predictions by the direct simulation method.
7
Vconst FORM Vinst FORM Vconst SORM Vinst SORM Vconst MC Vinst MC
3 2 1 0 0.2
0.3
0.4
0.5
0.6
0 [rad]
Figure 4.6.6. Reliability index by FORM, SORM and Monte Carlo procedure, for a specific speed and the sea state, and for constant and non-constant ship speed.
36
Table 4.6.1. Probability of the response exceedance by FORM and Monte Carlo procedure with 90% confidence interval, sea state Hs=12m Tz=11.7 s, case with constant speed Vconst. (Vconst) Hs=12 m Tz=11.7 s Roll =0.3 rad V0 [m/s] 5 6 7 8 9 10 5 6 7 8 9 10 5 6 7 8 9 10 Monte Carlo Probability 90% confidence interval 3.51410-2 4.146 10-2 3.74210-2 4.591 10-2 3.94710-2 4.613 10-2 3.38910-2 4.086 10-2 2.243 10-2 2.757 10-2 2.80910-2 3.271 10-2 5.30910 7.24610 6.626 10-3 9.574 10-3 7.790 10-3 1.010 10-2 8.477 10-3 1.112 10-2 4.909 10-3 7.491 10-3 8.042 10-3 1.062 10-2 1.472 10 6.306 10 4.796 10-4 1.187 10-3 7.027 10-4 1.520 10-3 1.119 10-3 2.103 10-3 8.860 10-4 1.781 10-3 7.027 10-4 1.520 10-3
-4 -4 -3 -3
Table 4.6.2. Probability of the response exceedance by FORM and Monte Carlo procedure with 90% confidence interval,, sea state Hs=12 m Tz=11 s, case with constant speed Vconst. (Vconst) Hs=12 m Tz=11 s Roll =0.3 rad V0 [m/s] 5 6 7 8 9 10 5 6 7 8 9 10 5 6 7 8 9 10 Monte Carlo Probability 90% confidence interval 7.55110-2 8.276 10-2 5.62310-2 6.257 10-2 4.46110-2 5.032 10-2 6.48110-2 7.159 10-2 4.532 10-2 5.108 10-2 4.90210-2 5.498 10-2 1.90710-2 2.29310-2 1.249 10-2 1.565 10-2 1.198 10-2 1.509 10-2 8.477 10-3 1.112 10-2 1.907 10-2 2.293 10-2 1.767 10-2 2.139 10-2 1.514 10-3 2.753 10-3 1.119 10-3 2.214 10-3 1.457 10-3 2.677 10-3 6.557 10-3 8.910 10-3 4.719 10-3 6.747 10-3 3.391 10-3 5.142 10-3 FORM Probability 6.013 10-2 3.691 10-2 2.741 10-2 1.089 10-2 6.942 10-3 6.054 10-3 3.143 10-2 1.556 10-2 1.647 10-2 7.922 10-3 4.599 10-3 3.954 10-3 6.201 10-3 1.608 10-3 5.441 10-3 3.474 10-3 1.890 10-3 1.684 10-3
FORM Probability 2.917 10-2 2.465 10-2 3.606 10-2 2.198 10-2 1.747 10-2 8.432 10-3 1.073 10 9.285 10-3 1.622 10-2 8.568 10-3 8.207 10-3 3.103 10-3 1.059 10 1.165 10-3 2.628 10-3 1.181 10-3 1.775 10-3 3.279 10-4
-3 -2
37
Table 4.6.3. Probability of the response exceedance by FORM and Monte Carlo procedure with 90% confidence interval, sea state Hs=12m Tz=11.7s, case with variable speed Vinst. Hs=12 m Tz=11.7 s (Vinst) Roll =0.3 rad V0 [m/s] 5 6 7 8 9 10 5 6 7 8 9 10 5 6 7 8 9 10 Monte Carlo Probability 90% confidence interval 1.83910-3 4.161 10-3 4.50410-3 7.829 10-3 4.37310-3 6.827 10-3 6.64910-3 8.795 10-3 7.422 10-3 9.912 10-3 4.05310-3 5.947 10-3 3.50210 9.83110 3.081 10-4 9.141 10-4 2.259 10-4 7.741 10-4 3.929 10-4 1.059 10-3 7.890 10-4 1.744 10-3 3.200 10-4 1.013 10-3 7.104 10 7.290 10 6.302 10-5 1.770 10-4 2.790 10-5 1.054 10-4 2.518 10-6 9.748 10-5 1.184 10-5 1.215 10-4 3.285 10-5 1.671 10-4
-6 -5 -4 -4
Table 4.6.4. Probability of the response exceedance By FORM and Monte Carlo procedure with 90% confidence interval,, sea state Hs=12 m Tz=11 s, case with variable speed Vinst. (Vinst) Hs=12 m Tz=11 s Roll =0.3 rad V0 [m/s] 5 6 7 8 9 10 5 6 7 8 9 10 5 6 7 8 9 10 Monte Carlo Probability 90% confidence interval 1.37510-2 1.63310-2 1.550 10-2 1.850 10-2 1.140 10-2 1.415 10-2 1.552 10-2 1.902 10-2 3.350 10-2 3.805 10-2 1.876 10-2 2.258 10-2 1.74710-3 2.75310-3 2.096 10-3 3.304 10-3 1.309 10-3 2.358 10-3 1.743 10-3 3.057 10-3 7.011 10-3 9.211 10-3 3.152 10-3 4.848 10-3 5.51010-5 3.61610-4 1.324 10-4 5.676 10-4 8.404 10-6 3.249 10-4 4.738 10-5 4.860 10-4 5.235 10-4 1.254 10-3 8.817 10-5 5.785 10-4 FORM Probability 1.034 10-5 1.523 10-2 3.326 10-2 1.857 10-2 1.709 10-2 9.428 10-3 1.671 10-7 5.442 10-3 1.717 10-2 8.202 10-3 6.605 10-3 2.991 10-3 4.471 10-9 8.521 10-4 4.246 10-3 1.833 10-3 8.698 10-4 3.815 10-4
FORM Probability 6.543 10-4 2.704 10-2 3.985 10-5 2.745 10-2 1.111 10-2 1.175 10-2 7.097 10 8.857 10-3 2.648 10-6 9.715 10-3 3.475 10-3 3.289 10-3 8.532 10 9.168 10-4 2.182 10-7 1.254 10-3 3.899 10-4 3.057 10-4
-6 -5
38
decision support system, based on the wave information to be encountered by the ship, the ship master can undertake adequate measures to prevent or reduce parametric roll phenomenon if necessary. Alternatively to decision support systems, roll event probability charts can be available showing which unfavorable combination of operation condition and the sea state to avoid. As a step towards the practical implementation, probability of the extreme roll event in stochastic sea has been calculated for different ship operational conditions and using different calculation methods. It has been shown in the present study that even though important in deterministic calculations, the exact calculation of the GZ curve may not be necessary in order to obtain the correct probability level and that an approximation of the GZ curve by a 5th order non-linear function with respect to the moving wave crest position in time should suffice. Further, it has been shown that, especially in case the ship speed is considered constant, approximate solution for probability of extreme roll obtained by FORM is in quite a good agreement with the direct Monte Carlo simulations. This conclusion is beneficial a much shorter computational time is required in order to obtain the final result retaining the correct level of accuracy.
40
References
American Bureau of Shipping (ABS). Guide for the Assessment of Parametric Roll Resonance in the Design of Container Carriers. September 2004. Bulian, G., Francescutto, A. and Lugni, C. Theoretical, numerical and experimental study on the problem of ergodicity and practical ergodicity with an application to parametric roll in longitudinal long crested irregular sea . Ocean Engineering, 2006, Vol. 33, pp 1007-1043. Chang, B.C. On parametric rolling of ships using a numerical simulation method. Ocean Engineering, 2008, Vol. 35, pp 447-457. De Kat, J.O., and Paulling, J.R. The Simulation of Ship Motions and Capsizing in Severe Seas. SNAME Transactions, 1989, Vol. 97, pp 139-168. Der Kiureghian, A. The geometry of Random Vibrations and Solutions by FORM and SORM. Probabilistic Engineering Mechanics, 2000, Vol. 15, pp 81-90. Det Norske Veritas. Parametric Rolling A Problem Solved? Container Ship Update No.1, 2006. Det Norske Veritas. Proban, General Purpose Probabilistic Analysis Program, Version 4.4, 2003. France, W.N., Levadou, M., Treakle, T.W., Paulling, J.R., Michel, R.K. and Moore, C. An Investigation of Head-Sea Parametric Rolling and Its Influence on Container Lashing Systems. Marine Technology, 2003, Vol. 40, No. 1, pp 1-19. Francescutto, A. Intact Stability of Ships Recent Developments and Trends. Proc. of the 10th International Symposium on Practical Design of Ships and Other Floating Structures PRADS, Vol. 1, Houston, 2007. Grim, O. Beitrag zu dem Problem der Sicherheit des Schiffes im Seegang. Schiff und Hafen, 1961, H. 6, pp 490-497. Hua, J., Palmquist, M., and Lindgren, G. An Analysis of the Parametric Roll Events Measured Onboard the PCTC AIDA. Proc. of the 9th International Stability of Ships and Ocean Vehicles 2006, pp 109-118. IMO, SLF/51/4/3 Sub-committee on Stability and Load Lines and on Fishing Vessels Safety, 51th Session. Revision of the Intact Stability Code. May 2008. IMO Revised Guidance to the Master for Avoiding Dangerous Situations in Adverse Weather and Sea Conditions. Resolution MSC.1/Circ.1228. 2007. IMO, SLF 50/WP.2. Revision of the Intact Stability Code Report of the Working Group (part I). IMO, London, 2007. 41
Jensen, J.J. Load and Global Response of Ships. Elsevier, Oxford, 2001. Jensen J.J. and Capul, J. Extreme Response Predictions for Jack-up Units in Second Order Stochastic Waves by FORM. Probabilistic Engineering Mechanics, 2006, Vol. 21, No. 4, pp 330-337. Jensen, J.J., Mansour, A.E. and Olsen, A.S. Estimation of Ship Motions using ClosedForm Expressions. Ocean Engineering, 2004, Vol. 31, pp 61-85. Jensen, J.J. Efficient Estimation of Extreme Non-linear Roll Motions using the Firstorder Reliability Method (FORM), J. Marine Science and Technology, 2007, Vol. 12 (4), pp 191-202. Jensen, JJ, Pedersen, PT, Vidic-Perunovic, J.Estimation of Parametric Roll in a Stochastic Seaway. IUTAM Symposium on Fluid-Structure Interaction in Ocean Engineering, 2007. Hamburg, Germany. Kerwin, J. E. Note on Rolling in Longitudinal Waves. Int. Shipbuild. Progress, 1955, 2(16) pp 597-614. Krger, S., Hinrichs, R. and Cramer, H. Performance Based Approaches for the Evaluation of Intact Stability Problems. Proc. PRADS2004, 2004, Travemnde, September, Germany. Kroeger, H.-P. Rollsimulation von Schiffen im Seegang. Schiffstechnik, 1986, Vol. 33, pp 187-216. Nayfeh, A. H. On the Undesirable Roll Characteristics of Ships in Regular Seas. J. Ship Research, 1988, 32 (2), pp 92-100. Neves, M. A. S. and Rodriquez, C. A. A Coupled Third Order Model of Roll Parametric Resonance. Proc. Maritime Transportation and Exploitation of Ocean and Coastal Resources, 2005. pp 243-253. Taylor and Francis, London. Oh, I., Nayfeh, A. H. and Mook, D. T. A Theoretical and experimental investigation of indirectly excited roll motion in ships. Phil. Trans. R. Soc. London, 2000. A, 358 (1771), pp 1853-1881. Paulling, J. R. and Rosenberg, R. M. On Unstable Ship Motions Resulting from Nonlinear Coupling. J. Ship Research, 1959, Vol. 3, No. 1, pp 36-46. Paulling, J. R. On Parametric Rolling of Ships. Proc. PRADS2007, 2007, Houston, USA. Rackwitz, R., and Fiessler, B. Structural Reliability Under Combined Random Load Sequence. J. Computers and Structures, 1978, Vol. 9, pp. 489-494.
42
Roberts, J. B. Effect of Parametric Excitation on Ship Rolling Motion in Random Waves. Jour. Of Ship Research, Dec 1982,Vol. 26, No. 4, pp. 246-253. Shin, Y.S., Belenky, V.L., Paulling, J.R.,Weems, K.M. and Lin, W.M. Criteria for Parametric Roll of Large Containerships in Head Seas. Transactions of SNAME, 2004, Vol.112, pp 14-47. Spanos, D. and Papanikolaou, A. SAFEDOR International Banchmark Study on Numerical Simulation Methods for the Prediction on Parametric Rolling of Ships in Waves. NTUA-SDL Report, April 2009, Rev. 04. Spyrou, K.J., Tigkas, I., Scanferla, G., Pallikaropoulos, N., and Themelis, N. Prediction potential of the parametric rolling behaviour of a post-panamax containership. Ocean Engineering, 2008, Vol. 35, pp 1235-1244. Thomsen, J. J. Vibrations and Stability, second edition, Springer Verlag, Berlin Heidelberg, 2003. Vidic-Perunovic, J, and Jensen, JJ. ''Parametric Roll due to Hull Instantaneous Volumetric Changes and Speed Variations''. Accepted for publication in Ocean Engineering, 2009. Umeda, N., and Yamakoshi, Y. Probability of ship capsizing due to pure loss of stability in quartering seas. Selected Papers from the Society of Naval Architects of Japan. Naval Architecture and Ocean Engineering, 2008, Vol. 30, pp 73-85. Wang, L., and Grandhi, R.V. Safety Index Calculation Using Intervening Variables for Structural Reliability Analysis . J. Computers and Structures, 1996, Vol. 59, pp. 563571. Yang, D., Gang, L., and Gengdong. C. Convergence analysis of first order reliability method using chaos theory. J. Computers and Structures, 2006, Vol. 84, No. 6, pp. 1139-1148.
43
Deliverables
Journal Articles: 1. Vidic-Perunovic, J, and Jensen, JJ, (2009). ''Parametric Roll due to Hull Instantaneous Volumetric Changes and Speed Variations''. Accepted for publication in Ocean Engineering. Conference Papers Published in Books of Proceedings: 1. Jensen, JJ, Pedersen, PT, Vidic-Perunovic, J, (2008).''Estimation of Parametric Roll in a Stochastic Seaway''. IUTAM Symposium on Fluid-Structure Interaction in Ocean Engineering. Hamburg, Germany. 2. Jensen, JJ, Vidic-Perunovic, J, Pedersen, PT, (2007). 'Influence of Surge Motion on the Probability of Parametric Roll in a Stationary Sea State''. Proc. of the 10th International Ship Stability Workshop. Hamburg, Germany. 3. Vidic-Perunovic, J, Rognebakke, O, Jensen, JJ, Pedersen, PT, (2008). 'Influence of Surge Motion on Extreme Roll Amplitudes''. Proc. of the 27th International Conference on OffshoreMechanics and Arctic Engineering - OMAE2008. Estoril, Portugal. 4. Galeazzi, R,Vidic-Perunovic, Blanke, M, J, Jensen, JJ, (2008). ''Stability Analysis of the Parametric Roll Resonance under Non-Constant Ship Speed''. Proc. of the 9th Biennial ASME Conference on Engineering System Design and Analysis- ESDA2008. Haifa, Israel. 5. Vidic-Perunovic, J, Jensen, JJ, (2009). ''Estimation of Parametric Rolling of Ships Comparison of Different Probabilistic Methods''. Proc. of MARSTRUCT 2009. Lisbon. Portugal 6. Vidic-Perunovic, J, (2009). ''Effect of Roll Restoring Lever Calculation on Parametric Roll Prediction in a Stationary Sea State''. Proc. of Proc. of 10th International Conference on Stability of Ships and Ocean Vehicles STAB 2009. Snkt. Petersburg, Russia.
44