INTERNATIONAL MARITIME ORGANIZATION

4 ALBERT EMBANKMENT
LONDON SE1 7SR E
Telephone: 020 7735 7611
Fax: 020 7587 3210
IMO

Ref. T1/2.04 MSC.1/Circ.1227

11 January 2007

EXPLANATORY NOTES TO THE INTERIM GUIDELINES FOR ALTERNATIVE

ASSESSMENT OF THE WEATHER CRITERION

1 The Maritime Safety Committee, at its eighty-second session (29 November to

8 December 2006), approved the Explanatory Notes to the Interim Guidelines for alternative
assessment of the weather criterion, set out in the annex, aiming at providing the industry with
alternative means (in particular, model experiments) for the assessment of the severe wind and
rolling criterion (weather criterion), as contained in the Code on Intact Stability for all Types of
Ships Covered by IMO Instruments (resolution A.749(18)).

2 Member Governments are invited to bring the annexed Explanatory Notes to the Interim
Guidelines to the attention of interested parties as they deem appropriate.

***

I:\CIRC\MSC\01\[Link]
 MSC.1/Circ.1227

ANNEX

EXPLANATORY NOTES TO THE INTERIM GUIDELINES FOR THE ALTERNATIVE

ASSESSMENT OF THE WEATHER CRITERION

1 Introduction

These explanatory notes provide an example of the alternative assessment of severe wind and rolling
criterion (weather criterion) based on a series of model tests following the Interim Guidelines for the
alternative assessment of the weather criterion contained in MSC.1/Circ.1200, for better
understanding of the alternative procedures. Here the weather criterion specified in paragraph 3.2 of
the Code∗ is referred as “standard weather criterion”, whereas the Interim Guidelines
(MSC.1/Circ.1200) are referred as “Guidelines”.

2 The tested ship

The principal particulars, general arrangement and GZ curve of the RoPax ferry used in this example
are shown in table 2.1, figure 2.1 and figure 2.2 respectively.

Table 2.1 Principal particulars

Length between perpendiculars: Lpp [m] 170.0 Area of bilge keels: Abk [m^2] 61.32
Breadth: B [m] 25.0 Vertical centre of gravity: KG [m] 10.63
Depth: D [m] 14.8 Metacentric height: GoM [m] 1.41
draft: d [m] 6.6 Flooding angle: φf [degrees] 39.5
Displacement: W [tonnes] 14,983 Rolling period: Tr [sec] 17.90
Blockage coefficient: Cb 0.521 Lateral projected area: AL [m^2] 3,433.0
B/d 3.79 Height to centre of AL above WL: Hc [m] 9.71

Figure 2.1 General arrangement

1.4
1.2
1
G Z [m ]

0.8
0.6
0.4
0.2
0
0 20 40 60 80
A ngle of H eel[deg]

Figure 2.2 GZ curve

∗
Throughout these Explanatory Notes, the reference to the Code means the reference to the Code on Intact Stability
for all Types of Ships Covered by IMO Instruments (resolution A.749(18)), as amended.

I:\CIRC\MSC\01\[Link]
MSC.1/Circ.1227
ANNEX
Page 2

3 The determination of the wind heeling lever lw1

3.1 Model set-up

3.1.1 Ship model used for wind tests

The model for the wind test was built following paragraph 1.2.1 of the Guidelines. The length (Lpp)
of the model was 1.5 m (scale: 1/113). The lateral projected area in upright condition was 0.267 m2.
Compared to the cross section of the wind tunnel (3 m in breadth and 2 m in height), the blockage
ratio was 4.5%.

3.1.2 Ship model used for drifting tests

The model for the drifting test was built following paragraph 1.2.2 of the Guidelines with bilge keels
of greater than 10 mm in breadth. The length of the model was 2 m (scale: 1/85).

3.2 Wind tests

3.2.1 The arrangement for the wind tunnel tests is shown in figure 3.1. The connection between
the model and load cell had a rotating device for testing the model in heeled conditions. In heeled
conditions the height of the model was adjusted by the adjusting plate to keep the displacement
constant when floating freely. The change of trim due to heel was neglected.

3.2.2 In order to keep the blockage ratio less than 5%, the floor plate was set to the same level of
the floor of the tunnel. The gap between the model and the floor plate was kept within
approximately 3 mm and covered by soft sheets for avoiding the effect of downflow through
the gap∗.

Figure 3.1 Arrangement for wind tunnel tests

3.2.3 The vertical distribution of wind speed is shown in figure 3.2. For the test arrangement
(figure 3.1), the height of the ship model from the floor was approximately 19 cm in upright
condition. This means that the lower half of the model is placed in the boundary layer. The
distributions of wind speed in the lateral and longitudinal directions were almost uniform (deviation
less than 1%) around the model.

∗
In order to simplify the execution of the experiments and to avoid the need for the building of appropriate floor
plates for each heeling angle, the gap could be filled by water. However, in this case, if any buoyancy effect occurs
in the model due to the particular setup, then it is to be properly accounted for in the subsequent analysis of the data.

I:\CIRC\MSC\01\[Link]
 MSC.1/Circ.1227
ANNEX
Page 3

30

25

Height (cm)
20

15

10

0
0.0 0.5 1.0 1.5
Normalized wind speed

Figure 3.2 Vertical distribution of wind speed

3.2.4 The wind speed was varied from 5 m/s to 15 m/s in upright condition and confirmed that the
drag coefficient is almost constant in this speed range. For the full tests a wind speed of 10 m/s was
used, corresponding to a Reynolds’ number of 1.52 × 105, as defined by the following equation:

U∞ B
Re = (N-3.1)
ν

where U ∞ is the uniform wind speed outside the boundary layer, B is the breadth of the model
and ν is the kinematic viscosity coefficient of air.

3.2.5 The horizontal force Fwind, the heeling moment M and the lift force L were measured by the
load cell. The heeling moment M was converted to the one with respect to point O, defined as Mwind,
by the following equation:

M wind = M − Fwind l cos φ + L ⋅ l sin φ (N-3.2)

where l is the distance from the centre of the load cell to point O. The point O is defined as the cross
point of the centreline of the ship and waterline in upright condition.

3.3 Drifting tests

3.3.1 The drifting test was carried out in a basin with dimensions of 50 m in length, 8 m in breadth
and 4.5 m in depth. The set-up is shown in figure 3.3. To connect the model, the load cell and the
heaving rod the same connection setup as in the wind tunnel tests was used. The horizontal
force Fwater and the heeling moment M were measured by the load cell. They were determined as
time averages in stationary condition after the transient phase, which appears when the carriage starts
to run. The heeling moment with respect to point O, defined as Mwater, was calculated similarly to
the wind tunnel tests. However, the last term in equation (N-3.2) is not necessary since the vertical
force is globally zero due to heave-free arrangement.

I:\CIRC\MSC\01\[Link]
MSC.1/Circ.1227
ANNEX
Page 4

hea
ve

d
rift

load cell

ship model

Figure 3.3 The arrangement for drifting test

3.3.2 The drifting speed should be determined to make the drifting force equal to Fwind as
defined by equation (1.2) of the Guidelines in model scale. The wind speed should be assumed to
be 26 m/s in ship scale. However, in this experiment, the speed was varied to cover the expected
range of the wind drag coefficient, CD, since the wind tunnel tests were carried out after the
drifting tests.

3.4 Results of wind tests

3.4.1 The measured drag coefficient (CD), lift coefficient (CL) and heeling moment coefficient (CM)
are shown in figure 3.4. They are non-dimensionalized by the following equations:

 CD   Fwind   1 
 C  =  L   2 ρ airU AL 
2
(N-3.3)
 L    

1 2 AL
2 
CM = M wind  ρ U  (N-3.4)
L pp 
air
2

3.4.2 In the figure the angle of heel is defined as positive when the ship heels to lee side (refer to
figure 3.1). The broken line is the heeling moment coefficient of the standard weather criterion,
calculated from equation (N-3.5), which is derived from the equation in paragraph [Link] of the
Code. However, in order to be compared with the test results, Z is replaced by the height of the
centre of the lateral projected area above waterline, i.e. Hc in table 2.1.

M wind = P ⋅ A ⋅ Z [N-m] (N-3.5)

I:\CIRC\MSC\01\[Link]
 MSC.1/Circ.1227
ANNEX
Page 5

CD
CL 1.25
CM
CM (standard criterion)
1

0.75

CD, CL, CM
0.5

0.25

0
-35 -30 -25 -20 -15 -10 -5 0 5 10 15 20 25 30 35
Angle of heel (degrees)

Figure 3.4 Measured drag coefficient (CD), lift coefficient (CL)

and heeling moment coefficient (CM)

3.4.3 It is characteristic in the figure that all the quantities (CD, CL and CM) vary significantly with
heel angle. As for the heeling moment, it is smaller than the standard criterion and further reduces
when the ship heels, especially to lee side. The lift force is not so small and close to the drag force
when the heeling angle is -5° (weather side). However, the adjustment of the vertical position of
the model is not necessary since the lift force is 0.7% of the displacement of the ship in a wind
speed of 26 m/s.

3.4.4 For comparing the test results with Z in equation (N-3.5), the measured heeling moment was
converted to the height of the centre of wind force above waterline, lwind, by the following equation:

lwind = Mwind / Fwind (N-3.6)

3.4.5 The result is shown in figure 3.5. It can be observed that the centre of wind force is also a
function of heel angle.

0.150

0.125

0.100
lwind (m)

0.075

Centre of lateral 0.050

projected area
0.025

0.000
-35 -30 -25 -20 -15 -10 -5 0 5 10 15 20 25 30 35
Angle of heel (degrees)

Figure 3.5 Height of the centre of wind force above waterline (model scale)

I:\CIRC\MSC\01\[Link]
MSC.1/Circ.1227
ANNEX
Page 6

3.5 Results of Drifting Tests

3.5.1 In the same manner as equation (N-3.6), the measured heeling moment generated by the drift
motion, Mwater, was converted into the height of the centre of drift force above waterline, lwater. The
values normalized by the draft are shown in figure 3.6, where the angle of heel is positive when the
ship heels to the drift direction as shown in figure 3.3.

3.5.2 It can be observed that, in the examined case, the centre of drifting force is above half draft
(which is the assumption in the standard criterion) and is generally above the waterline. This
phenomenon appears when breadth/draft ratio is large, due to the pressure distribution on the bottom.

0.6 CD=0.5 CD=0.6

CD=0.7 CD=0.8
CD=0.9 CD=1.0
0.4 CD=1.1

0.2
lwater / draft

0.0
-20 -15 -10 -5 0 5 10 15 20 25 30 35
Angle of heel (degrees)
-0.2

-0.4
Half draft

-0.6

Figure 3.6 Height of the centre of drift force for assumed wind drag coefficients

3.6 Determination of lw1

3.6.1 The heeling moments evaluated by wind tests, Mwind, and drifting tests, Mwater, were
substituted into equation (N-3.7) (the same as equation (1.1) of the Guidelines) and the wind heeling
lever, lw1, was calculated as a function of heel angle as shown in figure 3.7 and figure 3.8.

M wind + M water
lw1 = (N-3.7)
∆

3.6.2 In figure 3.7, the heeling levers due to wind ( M wind ∆ ) and drift motion ( M water ∆ ) are also
included. In both figures, lw1 at angles greater than 30° is assumed to keep the same value as at 30°
(see paragraph 1.6 of the Guidelines). Figure 3.7 shows that, in the considered case, the wind
heeling lever estimated by using the complete procedure, i.e. by using wind and drift tests, is
sensibly smaller than that required by the standard weather criterion.

I:\CIRC\MSC\01\[Link]
 MSC.1/Circ.1227
ANNEX
Page 7

0.15
Standard criterion
0.125
Wind test
0.1 Drift test

lw1 [m]
Wind + Drift tests

0.025

-0.025

-0.05
-20 -10 0 10 20 30 40
Angle of heel (degrees)

Figure 3.7 Wind heeling lever, lw1, evaluated by the tests

0.5
GZ, lwl [m]

0
-20 -10 0 10 20 30 40 50

-0.5 GZ
Standard Criterion
Wind + Drift tests
-1
Angle of heel [degrees]

Figure 3.8 Wind heeling lever, lw1, compared with the GZ curve

4 The determination of the roll angle φ1

4.1 Model basin

The model basin used for roll decay tests and rolling motion tests in waves was the same used for the
drifting tests (50 m in length, 8 m in breadth and 4.5 m in depth). The overall length of the
model (2.14 m) was small enough compared to the breadth of the basin.

4.2 Model set-up

4.2.1 The model was the same used for the drifting tests (Lpp = 2 m, scale: 1/85). It was built up to
the upper vehicle deck, till which buoyancy is included in the stability calculation. The top was built
open, but water did not enter into the model in waves with the largest steepness.

4.2.2 The model was ballasted to the loading condition for the ship, as shown in table 2.1. To
ensure correct displacement and attitude, the colour of the model was changed between above and
below the load line. The GM as measured by an inclining test was 1.67 cm, corresponding to
an 0.7% error to the scaled value of the ship. The natural roll period was also measured to be 1.92 s,
corresponding to an 1.2% error.

I:\CIRC\MSC\01\[Link]
MSC.1/Circ.1227
ANNEX
Page 8

4.3 General experimental set-up

4.3.1 The roll motion was measured by means of an inclinometer, for which uniform frequency
response was confirmed. Soft and slack cables were attached from the carriage to the model for
power supply and signal recording. The whole ship motion was recorded by a video camera.

4.3.2 The wave elevation was measured by wave probes for the waves corresponding to the
minimum and the maximum frequency used in the tests. This was done without the model
at 3 locations along the length of the basin, spanning a length wider than the expected drift range of
the model, i.e. 7.1 m. The maximum variations among the three measuring probes with respect to
the required value of wave height and wave period were 4% and 1% respectively.

4.4 Direct measurement procedure

4.4.1 From the table of wave steepness (table 2.1 of the Guidelines) the assumed wave steepness, s,
for this ship is 0.0383 (1/26.1). Roll amplitude of the model was measured in beam waves with this
steepness. Moreover, the waves with s = 1/40 and 1/60 were also used for the alternative procedures
mentioned later. The maximum wave height was 27.8 cm, which is close to the maximum limit of
the wave generator.

4.4.2 The model was freely drifting for avoiding effects of guide system and guide ropes were
fitted to the model on the centreline at the stem and stern at a vertical height around the centre of
gravity. The periodic yaw motion was small, but the change of mean heading angle was controlled
by the ropes carefully by human hands to keep it within 15°. The carriage followed the mean drift
motion of the model. When the heading was corrected by the ropes with large force, the
corresponding part of the measured record was neglected in the analysis. The effect of correction
was negligible in small waves, however in high waves the correction was not easy in some cases.

4.4.3 The measured roll responses were subjected to Fourier analysis to extract the components at
encounter frequencies. The results are shown in figure 4.1 for all wave steepnesses. Due to the
nonlinearity of roll damping, the non-dimensional roll amplitudes are larger in smaller steepness.
On the other hand, due to the linearity of GZ curve (see figure 2.2) the peak frequencies
do not significantly change even in high waves. The roll peak amplitude at the required steepness,
i.e. s = 1/26.1, was φ1r = 27.6 ° from the experiments. According to formula (2.1) of the Guidelines,
the “angle of roll to windward due to wave action”, φ1 , is thus determined as φ1 = 0.7φ1r = 19.3 °.

7 S =1/60
R ollA m plitude / W ave slope

6 =1/40
=1/26.1
5
4
3
2
1
0
0.6 0.8 1.0 1.2 1.4
W ave frequency / N aturalrollfrequency

Figure 4.1 Roll amplitude in beam regular waves (left: in degrees, right: non-dimensional)

I:\CIRC\MSC\01\[Link]
 MSC.1/Circ.1227
ANNEX
Page 9

4.5 Alternative procedures

In this section, alternative procedure 1 (Three steps procedure) is addressed.

4.5.1 Roll decays test

[Link] In calm water the model was initially inclined up to heel angles larger than 25° and released
with zero roll angular velocity. Four tests with different initial angles were conducted. As an
example, the relation of φm (mean roll angle) and δφ (decrement of roll angle per half cycle) for a
test is shown in figure 4.2.

[Link] The measured nonlinear roll damping coefficient, N, as a function of roll amplitude is
shown in table 4.1. The dependence of N coefficient on roll amplitude is small since the linear
component (wave making damping) is small for this ship.
δφ (degrees)
6

0
0 5 10 15 20 25

φm (degrees)

Figure 4.2 An example of roll decay test

Table 4.1 Roll damping coefficient N

Roll Amplitude N [1/degrees]
10.0 0.0122
12.5 0.0117
15.0 0.0113
17.5 0.0111
20.0 0.0109
22.5 0.0108
25.0 0.0106

4.5.2 Effective wave slope coefficient r

To measure the effective wave slope coefficient r, two options are described in paragraph [Link] of
the Guidelines. Here the results of roll motion tests, mentioned in paragraph 4.4 above, have been
used for s = 1/60 and the value r = 0.759 was obtained.

I:\CIRC\MSC\01\[Link]
MSC.1/Circ.1227
ANNEX
Page 10

4.5.3 Determination of roll angle φ1

By iterative calculation as described in [Link] of the Guidelines (Third step), φ1r was determined
as 27.9° and thus φ1 = 0.7φ1r = 19.5 °, which is very close to the estimation of direct measurement
procedure (section 4.4 above). One reason for this agreement is the almost linear characteristics
shown by the GZ curve up to 20° (see figure 2.2).

5 The assessment of weather criterion

5.1 The comparison of the different assessments of the weather criterion using experimental
results is summarized in table 5.1. In the table all the possible combinations of the wind tests and the
drifting tests for estimating lw1 are included (see paragraph 1.5 of the Guidelines). As for φ1 , the
standard criterion (paragraph [Link] of the Code) and the result of the direct measurement procedure
are included. The results of the three steps procedure can be omitted here since the estimated φ1 was
almost equal to the one of direct measurement procedure for this ship. The PIT estimation is
reported in section 6 below. The last line of table 5.1 shows the critical values of KGs, in
which b/a=1. These last results are to be taken with some caution, since the effects of changing the
vertical centre of gravity on Tr and on the other quantities related to roll motion including φ1 have
been neglected. A considerable extension of the experimental tests would have been indeed required
to correctly evaluate the limiting KG curve.

Table 5.1 Assessment of weather criterion

5.2 Table 5.1 shows that the alternative assessment by model experiment can change the ratio of
areas, b/a, significantly with respect to the standard criterion, whereas the changes in the limiting
value of the vertical centre of gravity are more contained. For this ship, the increased φ1 obtained by
experiments makes b/a smaller, and the lw1 evaluated through all the combinations of the wind tests
and drifting tests, complete procedures and simplified procedures, tends to make b/a larger than the
standard weather criterion as contained in paragraph 3.2 of the Code. It has to be noted, however,
that the fluctuations are related to the large variation in the vertical centre of hydrodynamic pressure
in drift motion of this ship. More extensive confirmation are awaited from the experience gained
through the application of the Interim Guidelines.
I:\CIRC\MSC\01\[Link]
 MSC.1/Circ.1227
ANNEX
Page 11

6 Alternative procedure 2: Parameter identification technique (PIT)

6.1 Introduction

6.1.1 The PIT technique is a general methodology for the determination of the numerical values for
a certain number of parameters in a given analytical model, in such a way that the model can
represent the physical behaviour of the system under analysis in the given conditions. Although the
PIT technique is also suitable for the direct analysis of roll decays in calm water in order to obtain
the ship natural frequency and the damping parameters, the roll motion of a ship in beam sea is dealt
with in this document.

6.1.2 The general idea on which the PIT is based is that the given analytical model is assumed to
be able to predict the amplitude of roll motion of the ship in beam sea, and this model is
characterized by a general form with a certain number of free parameters. The free parameters
should be fixed in order to obtain the best agreement between available experimental data and
numerical predictions from the model. When such parameters are determined, the model is assumed
to be suitable for extrapolation. In the case of roll motion in beam sea, the model parameters is fit by
using the ship roll response data for a small steepness in order to predict the ship behaviour at a
larger steepness for which direct experiments cannot be carried out, or for which direct experiments
are not available.

6.1.3 The general equation assumed suitable for the modelling of roll motion in beam sea is,
according to the Guidelines, the following:


() ω 
φ&& + d φ& + ω0 ⋅ r (φ ) = ω0 ⋅ π ⋅ s ⋅ ξ   ⋅ cos(ω ⋅ t )
2 2

  ω0 
 &
() & &&
d φ = 2 µ ⋅ φ + β ⋅ φ φ + δ ⋅ φ
&3
 (N-6.1)
r (φ ) = φ + γ 3 ⋅ φ + γ 5 ⋅ φ
3 5

 2
 ω  ω ω 
ξ  ω  = α 0 + α1 ⋅ ω + α 2 ⋅  ω 
  0 0  0

6.1.4 Where the following parameters are in principle to be considered as free (units are reported
assuming the roll angle to be measured in radians):

– Damping coefficients: µ (linear damping (1/s)), β (quadratic damping (1/rad)),

δ (cubic damping (s/rad2));

– Natural frequency ω0 (rad/s);

– Nonlinear restoring coefficients: γ 3 (cubic term (nd)), γ 5 (quintic term (nd));

– Effective wave slope coefficients: α 0 (constant (nd)), α1 (linear term (nd)),

α 2 (quadratic term (nd)).

I:\CIRC\MSC\01\[Link]
MSC.1/Circ.1227
ANNEX
Page 12

6.1.5 The wave steepness s, as well as the forcing frequency ω (to be measured directly from the
roll time histories in order to account for Doppler effect if the drift speed is large), are given data
from experiments.

6.1.6 The total number of free parameters is, thus, in principle, equal to 9. Such a large number of
parameters can be effectively determined from experimental data only when the number of
experiments is large, i.e., at least two (but is better three) wave steepnesses leading to response
curves spanning a large range of rolling angles from the linear range (below, say, 10°) up to the
nonlinear range (say, at least 40°). In addition, experimental data should span a large range of
ω
frequencies from low to high frequency range (say, from about 0.8 or lower to about 1.2 or
ω0
higher). The necessity of spanning such a large domain is due to the fact that different parameters
have a different importance in different ranges.

6.1.7 While damping plays an important role mostly around the peak region, the effective wave
slope is better determined if the low frequencies region of the response curve is also available.
Linear terms in both damping and restoring are dominant in the region of small rolling amplitudes,
while the effects of nonlinear terms are noticeable only in the region of large rolling amplitudes.
The roll response curve tends to bend to the low frequency region when GZ is of the softening type,
and towards the high frequency region when GZ is of the hardening type. Both type of bending
could be noticeable when the righting lever is of the S-type.

6.1.8 The general use of the PIT in the framework of the experimental determination of the roll
angle φ1r (See the Guidelines) will likely to be similar to that of the Three steps procedure, i.e. as
follows:

.1 carry out experiments at a single steepness sexp smaller than the required one sreq ;

.2 determine model parameters in order to fit the experiments at sexp ;

.3 utilize the obtained parameters in order to predict the peak of the ship roll response
at sreq ;

6.1.9 Since only one steepness is likely to be available, the number of parameters should be
reduced in order to achieve convergence of the methodology without spurious effects on
undetectable parameters. A reduced model is then to be used.

6.1.10 On the bases of a series of studies and on the experience gained in the past (see, e.g., [1][2]),
the following reduced model can be proposed when only one steepness is available:

1 steepness reduced model:

(N-6.2)
( )
φ&& + β ⋅ φ& φ& + ω02 ⋅ φ + γ 3 ⋅ φ 3 = ω02 ⋅ π ⋅ s ⋅ α 0 ⋅ cos (ω ⋅ t )

where the damping has been considered to be purely quadratic due to the fact that only one
amplitude response curve is available. The frequency dependence of the effective wave slope has
been dropped because we are mainly interested in this context in the ship response at peak, and so
the tails are of less (or none) importance for the final evaluation of φ1r (even if the low frequency tail
I:\CIRC\MSC\01\[Link]
 MSC.1/Circ.1227
ANNEX
Page 13

of the roll response is fundamental for the fitting of the value of α 0 ). As a note, the coefficient α 0
in the reduced model (N-6.2) corresponds to the effective wave slope “ r ” of the Three Steps
Procedure. A cubic nonlinear restoring term has been kept, but it can be removed if the GZ curve is
sufficiently linear in the expected response range, or if there is no evidence of bending from the
experimental response curve (provided the experimental peak is sufficiently large to allow the
identification of the possible nonlinear behaviour).

6.1.11 In the case where two response curves are available determined at two different steepnesses,
it is possible to introduce an additional linear damping term and an additional 5th degree restoring
term:

2 steepness reduced model:

(N-6.3)
( )
φ&& + 2 µ ⋅ φ& + β ⋅ φ& φ& + ω02 ⋅ φ + γ 3 ⋅ φ 3 + γ 5 ⋅ φ 5 = ω02 ⋅ π ⋅ s ⋅ α 0 ⋅ cos (ω ⋅ t )

6.1.12 Regarding the damping term in the previous reduced models, in general the quadratic
damping component seems to be more suitable for the analysis of hulls with bilge keels or with an
expected large vortex generation. On the other hand, the substitution of the quadratic term β ⋅ φ& φ&
with a cubic term δ ⋅ φ&3 could be more suitable for bare hulls.

6.1.13 The use of different nonlinear damping models, can lead to different results in the prediction
of the final rolling amplitude. For this reason, in the absence of sufficient evidence for the selection
of one nonlinear model versus the others, the use of the average of the two predicted peak rolling
amplitudes is recommended. A pure linear model, on the other hand, is almost always inadequate for
the representation of roll damping at zero speed.

6.2 General comments on PIT implementation

6.2.1 The PIT technique needs to be implemented in a suitable computer code, and it is not
amenable to hand calculations. A block diagram for the implementation of the PIT is reported in
figure 6.2. As it can be seen, the procedure is based on two main components:

.1 a differential equation solver used to determine the roll response predicted by the
model for different trial sets of parameters; and

.2 a suitable minimization algorithm used to achieve the optimum set of parameters by

minimizing the sum of the squared differences between experimental and predicted
roll amplitudes.

6.2.2 The differential equation solver could be basically of two types:

.1 Exact time domain solver: it numerically solves the general differential equation
(N-6.1) by using discrete time step algorithms (like the Runge-Kutta) for a certain
number of forcing periods, until the roll steady state is achieved. Finally, each time
history is analysed in order to get the steady state roll amplitude; and

I:\CIRC\MSC\01\[Link]
MSC.1/Circ.1227
ANNEX
Page 14

.2 Approximate frequency domain solver: it uses an analytical approximate solution of

the differential equation (N-6.1) in order to determine the nonlinear roll response
curve in frequency domain. Typically used analytical methods are the harmonic
balance technique, the multiple scale method and the averaging technique [3].

6.2.3 The two approaches have different pros and cons.

6.2.4 Time domain integration requires more computational time, but it solves the original
differential equation without approximations (apart from numerical accuracy). On the other hand, in
case of strong bending of the response curve, when multiple solutions are possible for the same
forcing frequency, then care must be taken in the numerical determination of the roll amplitude in
order to correctly deal with all the present solutions (see figure 6.1).

bare hull tests

40

bifurcation
region
Roll amplitude (degrees)

30

20

10

ω0
0
3.2 4.0 4.8 5.6
ω model scale (rad/s)

Figure 6.1 Example of experimental and numerically fitted nonlinear

response curve in the case of softening GZ

6.2.5 A typical numerical method that could be used for dealing with this problem is based on the
“frequency sweep” idea, where the forcing frequency is slowly changed in the time domain
integration from the highest value to the lowest one, and then vice-versa, in order to detect jumps due
the presence of bifurcations (see figure 6.1).

6.2.6 Analytical approaches are approximate solutions, and this is the biggest drawback.
However, the agreement between numerical simulations and analytical solutions is often surprisingly
good, and more than sufficient for practical applications. In addition, if the fitting of the
experimental data is based on an analytical method, and the same analytical method is used for the
extrapolation, i.e. a consistent methodology is used without mixing the analytical and the numerical
approach, good agreement is expected between numerical and analytical approaches. The analytical
methods are usually much faster than the direct time domain integration, and they are able to
determine multiple stable solutions in region where more than one solution is present, making the
dealing with this type of problem easier.

6.2.7 The differences in the final predicted roll peak φ1r between the application of the numerical
and of the analytical approach are expected to be below the usual experimental uncertainty (that
could be considered of the order of ± 2°).
I:\CIRC\MSC\01\[Link]
 MSC.1/Circ.1227
ANNEX
Page 15

6.2.8 The minimization algorithm could be any reliable minimization procedure

(e.g., Levenberg-Marquardt method, or any more advanced stochastic/deterministic method).

Figure 6.2 Block diagram for the PIT procedure

6.3 Application of PIT to experimental data

6.3.1 The same experimental data used in the Alternative Procedure 1 above have been used in the
application of the PIT. Scope of this application is to predict the roll response peak for the tested
required steepness s = 0.0383 by starting from available data at smaller steepnesses, i.e. 1/40
and 1/60. The following three calculations have been carried out:

.1 Calculation 1: prediction of φ1r by fitting of the model on the steepness s = 1/60;

I:\CIRC\MSC\01\[Link]
MSC.1/Circ.1227
ANNEX
Page 16

.2 Calculation 2: prediction of φ1r by fitting of the model on the steepness s = 1/40;

.3 Calculation 3: prediction of φ1r by fitting of the model on both the steepness

s = 1/40 and s = 1/60;

6.3.2 In the case of calculations 1 and 2, being only one steepness available, the reduced model
(N-6.2) has been used, and because of the linearity of the GZ curve and because of the absence of
any evident bending in the response curve it has been assumed that γ 3 = 0 .

6.3.3 In the case of calculation 3, being two steepnesses available, additional terms have been
added. Two different analytical model have then been used: the first model is exactly the same
as that used for calculation 1 and 2, whereas in the second model the linear damping
coefficient µ has been left free (see (N-6.3)). However, in both cases, the assumption of linear
restoring, i.e., γ 3 = 0 and γ 5 = 0 , has been kept.

6.3.4 In all cases the roll response curve has been determined through an analytical approximate
nonlinear frequency domain approach where the response curve is obtained by means of the
harmonic balance technique [3].

6.3.5 The used analytical models and the results obtained through the application of the PIT are
summarized in Table 6.1, while a global picture of the roll response curves is given from figure 6.3
to figure 6.6.

6.3.6 From the analysis of the reported exercise it seems that the PIT together with the proposed
analytical reduced models is able to reasonably predict the ship roll response curve at the largest
steepness by starting from the fitting of the roll response curve(s) experimentally obtained at lower
steepnesses. The pure quadratic damping model allows for the achievement of good predictions of
the experimental peak, probably thanks to the presence of bilge keels. In the case of linear+quadratic
damping model, a negative linear damping coefficient has been obtained, that is, of course,
physically meaningless. However, the equivalent linear damping in the range of tested angles as
given by the fitted model in Calculation 3-LQ is, of course, positive. The negative sign in the linear
damping coefficient is thus due to the fact that the equivalent linear damping obtained from the fitted
model in the range of tested rolling amplitudes better fits the experimental data according to the
minimization procedure. If a series of experiments had been carried out at smaller steepnesses with
subsequent fitting, it would have increased the linear damping coefficient, making it, probably,
positive. Bearing in mind the theoretical background of the PIT technique, negative linear damping
coefficients are often not a real practical problem, even if their presence usually indicates that
different types of analytical modelling for the damping function could lead to a better representation
of the real ship damping.

I:\CIRC\MSC\01\[Link]
 MSC.1/Circ.1227
ANNEX
Page 17

Table 6.1 Analytical models used in the fitting and fitted parameters (model scale)

Calculation 1 Calculation 2 Calculation 3-Q Calculation 3-LQ

Steepness used
1/60 1/40 1/60 and 1/40
in the fitting
φ&& + β ⋅ φ& ⋅ φ& + ω 02 ⋅ φ = φ&& + 2 µ ⋅ φ& + β ⋅ φ& φ& + ω 02 ⋅ φ =
Analytical model
ω 02 ⋅ π ⋅ s ⋅ α 0 ⋅ cos(ω ⋅ t ) = ω 02 ⋅ π ⋅ s ⋅ α 0 ⋅ cos (ω ⋅ t )
ω0 = 3.345rad / s
ω0 = 3.344rad / s ω0 = 3.348rad / s ω0 = 3.346rad / s µ
= −0.028
Fitted ω0
β = 0.520rad β = 0.518rad β = 0.519rad
−1 −1 −1

coefficients
β = 0.684rad
−1
α 0 = 0.873 α 0 = 0.857 α 0 = 0.864
α 0 = 0.833
Predicted value
in degrees of
φ1r for 28.3 28.1 28.2 27.0

s = 0.0383
Corresponding
value of 19.8 19.7 19.7 18.9
φ1 = 0.7 ⋅ φ1r
Experimentally
determined φ1 19.3
in degrees
Roll amplitude [degrees]

Figure 6.3 Response curves for Calculation 1

I:\CIRC\MSC\01\[Link]
MSC.1/Circ.1227
ANNEX
Page 18

Roll amplitude [degrees]

Figure 6.4 Response curves for Calculation 2

Roll amplitude [degrees]

Figure 6.5 Response curves for Calculation 3-Q

I:\CIRC\MSC\01\[Link]
 MSC.1/Circ.1227
ANNEX
Page 19

Roll amplitude [degrees]

Figure 6.6 Response curves for Calculation 3-LQ

6.3.7 In order to better explain this latter point, an additional calculation (Calculation 3-LQC) has
been carried out using experimental data for steepnesses s=1/60 and s=1/40 in the fitting procedure
together with a more flexible linear+quadratic+cubic model for the damping, keeping the linear
restoring assumption, i.e.:

φ&& + 2 µ ⋅ φ& + β ⋅ φ& φ& + δ ⋅ φ&3 + ω02 ⋅ φ = ω02 ⋅ π ⋅ s ⋅ α 0 ⋅ cos (ω ⋅ t ) (N-6.4)

6.3.8 The obtained parameters are as follows:

ω0 = 3.345rad / s
µ
= 0.013
ω0
β = 0.126rad −1
δ ⋅ ω0 = 0.929rad −2
α 0 = 0.844

6.3.9 It can be seen that now the negative linear damping has disappeared, and that the nonlinear
damping component is distributed among the quadratic and cubic term. Although this result is more
sound from a physical point of view, it is not necessarily the best one in terms of the predicted roll
peak at s = 0.0383. The predicted peak of the roll response is, indeed, φ1r = 26.6° leading to
φ1 = 18.6°. The reduction in the predicted roll peak is likely due to the introduction of the cubic
term. A summarizing plot is given in figure 6.7.

I:\CIRC\MSC\01\[Link]
MSC.1/Circ.1227
ANNEX
Page 20

Roll amplitude [degrees]

Figure 6.7 Response curves for Calculation 3-LQC

6.4 Final remarks

6.4.1 The PIT technique has successfully been applied to the experimental data used in the
previous sections for the application of the Three Steps Procedure.

6.4.2 It can be concluded that, for the ship under analysis, a pure quadratic model for damping,
together with a pure linear model for the restoring term is sufficient, for practical purposes, to predict
the roll peak φ1r at the steepness required by the alternative assessment of Weather Criterion.

6.4.3 It is however important to underline that for ships having significant nonlinear GZ curves,
it is necessary to introduce a nonlinear correction in the restoring term in order to account for the
bending of the response curve and the corresponding peak frequency shift. It is in addition noted,
from the experience gained from this exercise, that an additional test in the range of low forcing
frequencies (say ω = 0.75 ⋅ ω0 ) could help in the fitting of the effective wave slope, allowing to take
into account a frequency dependence of this coefficient. This latter frequency dependence could be
important when the bending of the response curve is significant.

6.4.4 As an additional note, it can be said that the application of different tentative models in the
PIT allows for an assessment of the likely level of uncertainty inherent in the extrapolation.

6.4.5 In the case under analysis, the level of uncertainty is of the order of ± 2°, however this figure
strongly depends on the actual analysed case.

I:\CIRC\MSC\01\[Link]
 MSC.1/Circ.1227
ANNEX
Page 21

6.4.6 The value of the effective wave slope obtained through the PIT (about 0.85 on average) is
slightly different from the value obtained through the application of the Three steps
procedure (r = 0.759). This difference can be readily explained by recalling that, in the Three steps
procedure, the damping is evaluated from the roll decays tests, while the effective wave slope is
evaluated from the roll tests in beam waves, using the previously obtained damping coefficient. In
the PIT approach, on the contrary, both the damping and the effective wave slope are determined
from the same experimental data in beam waves, for this reason the final outcomes could differ in
terms of single components. The final predictions of the angle φ1r given by the PIT technique and
by the Three steps procedure are however very close: the two alternative procedures can be then
considered, for this particular case, as equivalent from a practical point of view.

6.5 References

[1] Francescutto, A., Contento, G., “Bifurcations in Ship Rolling: Experimental Results and
Parameter Identification Technique”, Ocean Engineering, Vol. 26, 1999, pp. 1095-1123.

[2] Tzamtzis, S., Francescutto, A., Bulian, G. and Spyrou, K., “Development and testing of a
procedure for the alternative assessment of Weather Criterion on experimental basis”,
Technical Report, University of Trieste, Dept. Naval Architecture & Environmental
Engineering, 2005.

[3] Nayfeh, A.H., Mook, D.T., “Nonlinear Oscillations”, John Wiley & Sons, Inc., 1979.

[4] IMO Document, SLF 47/6/18, “Proposal of Guidelines for a standard model test procedure to
determine the steady wind heeling lever”, Submitted by Italy and Japan, 7 July 2004.

[5] IMO Document, SLF 47/6/19, “Proposal of Guidelines for model tests to determine the roll
angle for the weather criterion”, Submitted by Japan, 7 July 2004.

[6] IMO Document, SLF 48/4/15, “Comments on draft guidelines for alternative assessment of
weather criterion based on trial experiment results”, Submitted by Japan, 8 July 2005.

______________

I:\CIRC\MSC\01\[Link]