E

4 ALBERT EMBANKMENT
LONDON SE1 7SR
Telephone: +44 (0)20 7735 7611 Fax: +44 (0)20 7587 3210

MSC.1/Circ.1652
26 April 2023

EXPLANATORY NOTES TO THE INTERIM GUIDELINES ON THE SECOND

GENERATION INTACT STABILITY CRITERIA

1 The Maritime Safety Committee, at its 102nd session (4 to 11 November 2020)

approved the Interim guidelines on the second generation intact stability criteria
(MSC.1/Circ.1627) (Interim Guidelines). In approving the Interim Guidelines, the Committee
recognized the necessity of developing associated explanatory notes to ensure uniform
interpretation and application.

2 To this end, the Committee, at its 105th session (20 to 29 April 2022), approved the
Explanatory notes to the Interim guidelines on second generation intact stability criteria
(Explanatory Notes), set out in the annex, as prepared by the Sub-Committee on Ship Design
and Construction, at its eighth session (17 to 21 January 2022).

3 The Explanatory Notes are intended to provide Administrations and the shipping
industry with specific guidance to assist in the uniform interpretation and application of the
Interim Guidelines.

4 Member Governments are invited to bring the annexed Explanatory Notes to the
attention of all parties concerned, in particular shipbuilders, shipmasters, shipowners,
ship operators and shipping companies, and recount their experience gained through their use
to the Organization.

***

I:\CIRC\MSC\1\MSC.1-Circ.1652.docx
 MSC.1/Circ.1652
Annex, page 1

ANNEX

EXPLANATORY NOTES TO THE INTERIM GUIDELINES ON THE SECOND

GENERATION INTACT STABILITY CRITERIA

CONTENTS

Part A – INTRODUCTION TO THE EXPLANATORY NOTES TO THE INTERIM

GUIDELINES ON THE SECOND GENERATION INTACT STABILITY
CRITERIA

Part B – GUIDANCE ON INDIVIDUAL SECTIONS OF THE INTERIM GUIDELINES

Chapter 1 General

1.1 Introduction

1.2 Definitions

1.3 Nomenclature

Chapter 2 Guidelines on vulnerability criteria

2.1 Preface

2.2 Assessment of ship vulnerability to the dead ship condition failure mode

2.3 Assessment of ship vulnerability to the excessive acceleration failure mode

2.4 Assessment of ship vulnerability to the pure loss of stability failure mode

2.5 Assessment of ship vulnerability to the parametric rolling failure mode

2.6 Assessment of ship vulnerability to the surf-riding/broaching failure mode

2.7 Parameters common to stability failure mode assessments

Chapter 3 Guidelines for direct stability failure assessment

3.1 Objective

3.2 Requirements

3.3 Requirements for a method that adequately predicts ship motions

3.4 Requirements for validation of software for numerical simulation of ship

motions

3.5 Procedures for direct stability assessment

Chapter 4 Guidelines for operational measures

4.1 General principles

I:\CIRC\MSC\1\MSC.1-Circ.1652.docx
MSC.1/Circ.1652
Annex, page 2

4.2 Stability failures

4.3 Operational measures

4.4 Acceptance of operational measures

4.5 Preparation procedures

4.6 Application

Appendix 1 Physical description of the stability failure modes addressed by second

generation intact stability criteria

1 Physical background of stability failure related to the dead ship condition

1.1 Modelling in the level 1 criterion for the dead ship condition

2 Physical background of stability failure related to excessive accelerations

2.1 Accelerations caused by ship motions
2.2 Synchronous resonance in ship motions

3 Physical background and scenario of pure loss of stability

3.1 Righting lever variation in waves

4 Physical background of parametric rolling

4.1 Development of parametric rolling
4.2 Frequency characteristics of parametric rolling
4.3 Influence of roll damping
4.4 Influence of speed and wave direction

5 Physical background of surf-riding and broaching

5.1 General description of surf-riding/broaching failure mode
5.2 Description of surf-riding equilibrium
5.3 Stability of surf-riding equilibrium
5.4 Transition from surging motion to surf-riding

Appendix 2 Examples of assessments using vulnerability criteria according to

second generation intact stability criteria

1 Example input data set

2 Example of assessment of ship vulnerability to the dead ship condition failure

mode

3 Example of assessment of ship vulnerability to the excessive acceleration

failure mode

4 Example of assessment of ship vulnerability to the pure loss of stability failure

mode

5 Example of assessment of ship vulnerability to parametric rolling

6 Example of assessment of ship vulnerability to surf-riding/broaching

I:\CIRC\MSC\1\MSC.1-Circ.1652.docx
 MSC.1/Circ.1652
Annex, page 3

Appendix 3 Elements for numerical modelling of roll motion in the vulnerability

criteria of the second generation intact stability criteria

1 Equation of motion

2 Equation of motion with respect to parametric rolling

2.1 Roll restoring
2.2 Evaluation of metacentric height and righting lever curve in
longitudinal waves
2.3 Assessment of the equation of motion in terms of inertia

3 Information regarding level 2 vulnerability criterion for the dead ship condition
3.1 Roll modelling in beam waves and wind
3.2 Roll damping, natural roll frequency and the effective wave slope
function

4 Calculation of maximum roll angle for assessment of parametric rolling in

check 2 of level 2 of the vulnerability criteria

5 Supplementing information on calculation for checking vulnerability to

parametric rolling

6 Discussion on the relationship between level 1 and 2 vulnerability criteria for

parametric rolling

7 Method to establish equivalence between regular and irregular waves as

provided in vulnerability criteria, level 2 for both pure loss of stability and
parametric roll stability failure modes (Grim's effective wave approach) to
assess the change of stability in longitudinal irregular waves

8 Determination of roll moment due to waves and effective wave slope function
8.1 Methods for evaluating the effective wave slope function
8.2 Standard methodology for the estimation of the effective wave slope
function
8.3 Direct calculation of the Froude-Krylov roll moment used in the
level 2 criterion for excessive acceleration failure mode

9 Estimation of roll damping

9.1 General
9.2 The simplified Ikeda Method
9.3 Equivalent linear roll damping coefficients for the dead ship and
excessive acceleration failure modes

10 Methods for determining wave cases for vulnerability criteria

Appendix 4 Theoretical background, validation, and application examples to

guidelines on direct stability assessment

1 Introduction

2 Validation of numerical methods for simulation of ship motions

2.1 Qualitative validation: Backbone curve
2.2 Qualitative validation: Response curve
2.3 Qualitative validation: Change of stability in waves

I:\CIRC\MSC\1\MSC.1-Circ.1652.docx
MSC.1/Circ.1652
Annex, page 4

2.4 Qualitative validation: Principal parametric resonance

2.5 Quantitative validation requirements
2.6 Validation example for dead ship condition stability failure mode
2.7 Validation example of linear method for excessive acceleration
failure mode

3 Direct counting
3.1 Introduction
3.2 Definition and characteristics of Poisson process
3.3 Definition of failure rate from sample data using exponential
distribution
3.4 Definition of failure rate from sample data from analysis of
probability of failure
3.5 Definition of failure rate from sample data using binomial distribution
3.6 Long-term statistics
3.7 Cautions in numerical simulations or model tests
3.8 Decorrelation
3.9 Direct counting procedures
3.10 Application examples of direct counting method
4 Direct stability assessment
4.1 Ships and loading conditions used in examples
4.2 Examples of full probabilistic direct stability assessment
4.3 Database for development of direct stability assessment procedures
for parametric rolling, pure loss of stability and excessive
acceleration failure modes
4.4 Design situations
4.5 Deterministic direct stability assessment
4.6 Definition of standard and thresholds
4.7 Example procedure for probabilistic assessment in design situations
4.8 Application examples of probabilistic assessment in design
situations

5 Statistical extrapolation methods

5.1 Extrapolation of failure rate over wave height
5.2 Critical wave method for surf-riding/broaching failure mode
5.3 Split-time/motion perturbation method (MPM)
5.4 Envelope peaks over threshold (EPOT) for pure loss of stability and
dead ship condition
5.5 Application of MPM and EPOT methods to full probabilistic
assessment
5.6 Linear superposition method for excessive acceleration failure
mode

6 Roll damping

7 Application examples of verification of failure modes

7.1 Parametric roll
7.2 Pure loss of stability
7.3 Surf-riding/broaching
7.4 Synchronous roll

I:\CIRC\MSC\1\MSC.1-Circ.1652.docx
 MSC.1/Circ.1652
Annex, page 5

Appendix 5 Theoretical background, validation, and application examples to

guidelines on operational measures

1 Background information
1.1 Ships and loading conditions used in background studies
1.2 Preparation of operational measures
1.3 Preparation of operational guidance in design phase
1.4 Probabilistic operational guidance
1.5 Deterministic operational guidance
1.6 Definition of thresholds
1.7 Simplified operational guidance
1.8 When operational measures are not suitable
1.9 Influence of propulsion, steering and seakeeping

2 Application examples
2.1 Operational guidance based on DSA for parametric and
synchronous roll
2.2 Operational guidance for parametric roll
2.3 DSA-based operational limitations related to areas or routes and
season
2.4 DSA-based operational limitations related to maximum significant
wave height
2.5 Simplified operational guidance for surf-riding/broaching failure
mode
2.6 Simplified operational guidance from level 2 vulnerability
assessment for parametric roll
2.7 Level 2-based operational limitations related to maximum wave
height

3 Wave cases for preparation of operational limitations using level 1 and level 2
vulnerability assessment

4 Supplementing information on preparation of simplified operational guidance

for surf-riding/broaching failure mode

Appendix 6 Application examples of treatment of loading conditions

1.1 Cruise ship

1.2 1,700 TEU container ship

I:\CIRC\MSC\1\MSC.1-Circ.1652.docx
MSC.1/Circ.1652
Annex, page 6

PART A

INTRODUCTION TO THE EXPLANATORY NOTES TO THE INTERIM GUIDELINES ON

THE SECOND GENERATION INTACT STABILITY CRITERIA (MSC.1/CIRC.1627)

1 The Maritime Safety Committee (MSC), at its 102nd session (4 to 11 November 2020),
approved the Interim guidelines on the second generation intact stability criteria
(MSC.1/Circ.1627), hereinafter referred to as "Interim Guidelines". These Explanatory Notes
to the Interim Guidelines are intended as a support in the application of the Interim Guidelines
by providing further clarifications and explanations to the elements therein.

2 These Explanatory Notes should be consulted for an improved understanding and

uniform application of the Interim Guidelines. The paragraph numbers used in part B directly
correspond to the numbering in the Interim Guidelines.

3 Because the approach taken in the Interim Guidelines is new for many Administrations
and the industry, these Explanatory Notes have been developed to assist the users. Some of
the concepts employed in the Interim Guidelines are relatively easy to grasp, others may
require more consideration, explanation and in-depth discussion. In view of this, the structure
of the Explanatory Notes follows that of the Interim Guidelines and provides comments and
explanations for those paragraphs of the Interim Guidelines that have been identified as most
benefiting from additional discussion, clarification or explanation.

4 The appendices provide additional information that is pertinent to the Interim

Guidelines. Beginning with the description of the stability failure modes addressed by the
Interim Guidelines in appendix 1, examples are provided of how the Interim Guidelines may
be applied and the process by which assessments of ship's vulnerability are made in
appendix 2. Appendix 3 is devoted to providing detailed explanation of the process by which
certain elements that are essential to proper assessment of the vulnerability criteria are
performed and the considerations that the user should take into account when performing the
calculations associated with those elements.

5 Because of their complexity and the considerations that should be employed in

performing a direct stability assessment, appendix 4 is devoted to this subject. This appendix
provides the background of direct stability assessment, gives detailed description of the
validation of the computer software to be used for direct stability assessment, and presents
examples of the application of direct stability assessment.

6 In a similar manner, appendix 5 presents the background of operational measures,

gives a description of the validation of the preparation of operational measures, and presents
examples of the preparation of operational measures, based on different data sources and
purposes.

7 The resources and effort required for a direct stability assessment and the preparation
of operational measures require that some economy be achieved by limiting the number of
loading conditions for a ship for which these activities will be performed. Therefore, the choice
of the loading conditions to be used for these purposes should be carefully considered. For
this reason, appendix 6 is prepared to aid the users in the selection of loading conditions.

8 Finally, the scientific community that has supported the development of the Interim
Guidelines published its research outcomes associated with the second generation intact
stability criteria in major scientific journals, conferences and book chapters. Some of these
publications discuss and provide further reading related to the underlying basis for the second

I:\CIRC\MSC\1\MSC.1-Circ.1652.docx
 MSC.1/Circ.1652
Annex, page 7

generation intact stability criteria. The contents of the external references included in these
Explanatory Notes are not to be considered as part of, or as an extension of, the text of these
Explanatory Notes. These are included only for respecting the copyrights or investigating the
background of the criteria. It is also noteworthy to mention here that several other contributions
to other literature were also relevant to the second generation intact stability criteria.1

1
Including, but not limited to:
Spyrou, K. J., Belenky, V. L., Katayama, T., Bačkalov, I., and Francescutto, A., eds., Contemporary Ideas
on Ship Stability: From Dynamics to Criteria, Springer, 2023, ISBN: 978-3-031-16328-9; and
Belenky, V., Spyrou, K., van Walree F., Neves, M.A.S., and Umeda, N., eds., Contemporary Ideas on Ship
Stability. Risk of Capsizing, Springer, 2019, ISBN: 978-3-030-00514-6.

I:\CIRC\MSC\1\MSC.1-Circ.1652.docx
MSC.1/Circ.1652
Annex, page 8

PART B

GUIDANCE ON INDIVIDUAL SECTIONS OF THE INTERIM GUIDELINES2

1 GENERAL

1.1 Introduction

1.1.1 Purpose

1.1.1.1 Motivation for and modes of stability failure covered by the second generation intact
stability criteria

The motivation for the development of performance-based criteria for intact stability is derived
from the appearance of new types of vessels and modes of operation. This is discussed, in
part, in the preamble to the 2008 Intact Stability Code, as adopted by resolution MSC.267(85):

"…in view of a wide variety of types, sizes of ships and their operating and
environmental conditions, problems of safety against accidents related to stability
have generally not been solved. In particular, the safety of a ship in a seaway involves
complex hydrodynamic phenomena which up to now have not been fully investigated
and understood. Motion of ships in a seaway should be treated as a dynamical system
and relationships between ship and environmental conditions such as wave and wind
excitations are recognized as extremely important elements. Based on hydrodynamic
aspects and stability analysis of a ship in a seaway, stability criteria development
poses complex problems that require further research."

Part A of the 2008 Intact Stability Code explains further: "It is recognized by the Organization
that performance-oriented criteria for the identified phenomena listed in this section (righting
lever variation, resonant roll in dead ship condition, and broaching and other
manoeuvring-related phenomena) need to be developed and implemented to ensure a uniform
international level of safety." Responding to this need, the Organization pursued the
development of second generation intact stability criteria for the identified phenomena as well
as the hazard associated with excessive accelerations.

1.1.1.2 As the framework for the development explains, the intact stability regulations
contained in the 2008 Intact Stability Code provide deterministic criteria associated with
physical typologies and sizes of ships operated in the 1960s. Ships that have typologies or
sizes outside the scope of those for which those regulations were developed, as well as some
ships that fully satisfy the 2008 Intact Stability Code, may be susceptible to different modes of
stability failures. These ships may have non-conventional hull geometry, modes of operation,
or loading conditions neither envisioned by nor within the applicability limits of the first
generation criteria. As a result, second generation intact stability criteria are considered as a
complement to the first generation criteria.

The aim of the second generation intact stability criteria is to establish recommendations for
ship design, applicable to all types of ships vulnerable to major dynamic modes of stability
failures.

2
This part was prepared to realize paragraph-to-paragraph comparisons with the Interim Guidelines.
Thus, the section and paragraph numbers are based on the Interim Guidelines.

I:\CIRC\MSC\1\MSC.1-Circ.1652.docx
 MSC.1/Circ.1652
Annex, page 9

1.1.2 Framework

1.1.2.1 The second generation intact stability criteria involve stability assessment using
methods aligned with physics of the phenomena under consideration. To facilitate this,
definitions, terminology and nomenclature specific to the second generation intact
stability criteria were developed.

1.1.3 Application logic

1.1.3.1 At least one design assessment option (level 1 or level 2 vulnerability criteria or direct
stability assessment) or operational measures should be applied to each considered
loading condition. The loading conditions to which the design assessment or
operational measures are applied should be based on actual loading during standard
operation, depending on the actual layout of cargo hold, the purpose of voyages and
the expected weather, and could include special loading for delivery voyages,
additional securing of fishing net, heavy weather ballast adjustment and so on.

1.1.4 Testing

1.1.4.1 With reference to item .3 that addresses the level 2 vulnerability criterion for the pure
loss of stability failure mode, it is noted that the dynamics of water on deck may significantly
affect the stability in waves for ships with low freeboard. The vulnerability criteria do not take
into account the dynamics of water on deck; therefore this may lead to very conservative
results.

1.1.5 Feedback

1.1.5.1 The feedback solicited in the Interim Guidelines can be separated into three types:

.1 the results of application of the Interim Guidelines;

.2 comments on the Interim Guidelines or Explanatory Notes, including

requests for increased clarity and explanation; and

.3 suggestions of alternatives or improvements/refinements of the criteria

contained in the Interim Guidelines.

Submissions of feedback on any of these types should be submitted to the Organization by

electronic means to [email protected] as directed by the following:

Feedback on results (.1 above) should be formatted similar to that provided by the examples
given in appendix 2 to the Explanatory Notes.

Feedback on the Interim Guidelines or the Explanatory Notes that comment on or request
clarity of the Interim Guidelines or Explanatory Notes should identify the element of the Interim
Guidelines or Explanatory Notes for which comment is given and the reason for which it is
given.

Feedback of a suggestion of an alternative or an improvement should include an analysis of

the basis for the suggestion and a comparison with the criterion given in the Interim Guidelines
that provides a comprehensive assessment of the basis for the comparison including all
intermediate results that compare the criteria on a common basis. Examples of such
comparison may be discovered in relevant publications.

I:\CIRC\MSC\1\MSC.1-Circ.1652.docx
MSC.1/Circ.1652
Annex, page 10

1.1.6 Relationship with mandatory criteria

1.1.6.1 The second generation intact stability criteria establish minimum recommendations
for ship design, applicable to all types of ships and major dynamic modes of stability failures.
Verification of the ship vulnerability to the major dynamic modes of stability failures is regarded
as a complementary assessment and shall not be used as an alternative to the 2008 Intact
Stability Code criteria.

1.1.7 Notes for application

1.1.7.1 A ship with an extended low weather deck means a vessel which is engaged primarily
in the transport of stores, materials and equipment to offshore installations and designed with
accommodation and bridge erections in the forward part of the vessel and an exposed cargo
deck in the after part for the handling of cargo at sea.

1.2 Definitions

1.2.1 The term "environmental condition" that is used in the Interim Guidelines is to be
intended as synonymous with "sea state", unless additional information is necessary to
characterize the environment for specific applications.

1.3 Nomenclature

1.3.2 General ship characteristics

"Ak = total overall area of the bilge keels (no other appendages) (m 2);"

The total overall area of the bilge keels means the sum of the projected area of the bilge keel
normal to the hull on both the port and starboard sides. Only for the dead ship stability level 1
criterion, the projected area of a bar keel, if fitted, can be added.

1.3.4 Loading condition characteristics

"GM = metacentric height of the loading condition in calm water (m), with or
without correction for free surface effect, as required;"

The free surface effect should be taken into account for calculating GM and righting lever in
calm water and in waves, except for the excessive acceleration failure mode.

"Tr = linear natural roll period in calm water (s)"

The free surface effect should be taken into account for calculating Tr, except for the excessive
acceleration failure mode.

2 Guidelines on vulnerability criteria

2.1 Preface

Vulnerability criteria are intended to distinguish between vulnerable and non-vulnerable

loading conditions of a ship; a ship's loading condition that satisfies the standard of any
assessment level of a stability failure mode is considered to have an acceptable level of safety
with reference to that stability failure mode.

I:\CIRC\MSC\1\MSC.1-Circ.1652.docx
 MSC.1/Circ.1652
Annex, page 11

2.2 Assessment of ship vulnerability to the dead ship condition failure mode

2.2.1 Application

2.2.1.4 The application of stability limit information is additional to the stability limits that are
obtained from application of the 2008 Intact Stability Code, part A. One should keep in mind
that resonance effects for these failure modes may create areas showing vulnerability within
the allowed GM limits obtained from applying the criteria in part A. It is therefore recommended
to perform these second generation stability calculations for different GMs for each draught
and trim to reveal these vulnerable areas.

2.2.2 Level 1 vulnerability criterion for the dead ship condition

2.2.2.1 For the purpose of level 1 vulnerability assessment for dead ship stability failure, the
method of assessment included in the "Severe wind and rolling criterion (weather criterion)" in
section 2.3 of part A of the 2008 Intact Stability Code is used, but the steepness factor s in
table 2.3.4-4 of section 2.3 of part A is substituted with the steepness factor s specified in
table 4.5.1 of the Interim guidelines for alternative assessment of the weather criterion
(MSC.1/Circ.1200) (see figure 2.2.1).

2.2.2.4

"φ1 = angle of roll to windward due to wave action (deg)."

Refer to the Explanatory Notes to the International Code on Intact Stability, 2008 (2008 Intact
Stability Code) (MSC.1/Circ.1281), section 3.5.4.

Table 2.2.2.4-4 of the Interim Guidelines is imported from MSC.1/Circ.1200 as shown in figure
2.2.1.

Figure 2.2.1 Explanation to table 2.2.2.4-4 of the Interim Guidelines

I:\CIRC\MSC\1\MSC.1-Circ.1652.docx
MSC.1/Circ.1652
Annex, page 12

2.2.3 Level 2 vulnerability criterion for the dead ship condition

2.2.3.1 The objective of the level 2 vulnerability assessment methodology is to provide a

simplified conservative probabilistic measure of vulnerability of the ship, in the considered
loading condition, to the dead ship stability failure mode.

Several assumptions are required to provide a suitable tool for level 2 vulnerability assessment.
Some of such assumptions are described in the relevant sections of these Explanatory Notes.
However, the fundamental set of underlying assumptions is as follows:

.1 the ship is assumed to be in dead ship condition in irregular waves and gusty
wind for a specified exposure time;

.2 wind and waves are assumed to blow/propagate in the same direction and
the water depth is sufficiently large (water depth is larger than half the
wavelength) to allow an assumption of infinite water depth;

.3 the ship is assumed to remain beam to wind and waves;

.4 the wind state is characterized by a mean wind speed and a gustiness

spectrum;

.5 the sea state is characterized by a wave elevation spectrum and waves are
assumed to be long-crested; and

.6 the roll motion of the vessel can be modelled as a one-degree of freedom

(1-DOF) system.

The measure, indicated as C, is a long-term probability index ranging from 0.0 (good) to 1.0
(bad), which is obtained as a weighted average of short-term indices CS.

2.2.3.2 The weighting factor Wi should be obtained from the wave scatter table such as
table 2.7.2.1.2 of the Interim Guidelines as the ratio of the value described in the cell to the
sum of the values in all cells, which represents the relative frequency of occurrence of
short-term environmental condition specified with the significant wave height and the mean
zero-crossing wave period.

The total number of short-term environmental conditions corresponds to the number of cells
with non-zero frequency of occurrences in the wave scatter table such as table 2.7.2.1.2 of the
Interim Guidelines because the short-term environmental conditions can be specified with the
significant wave height and mean zero-crossing wave period.

2.2.3.2.1 The short-term dead ship failure index, CS, depends on the characteristics of the
ship in the considered loading condition and the short-term environmental conditions of a given
duration. This is obtained by a simplified calculation methodology which takes into account
characteristics of roll in the considered short-term environmental condition.

For ships which are not port/starboard symmetric, the short-term dead ship failure index CS is
to be determined as the average between the index calculated for wind and waves coming
from both port and starboard sides.

A conceptual scheme of the assumed underlying simplified physical modelling of the

phenomenon is shown in figure 1.1 of appendix 1. The overview of the logic of the short-term
modelling is as follows:

I:\CIRC\MSC\1\MSC.1-Circ.1652.docx
 MSC.1/Circ.1652
Annex, page 13

.1 ship characteristic parameters are assumed to be available (displacement,

righting lever, roll damping, windage area characteristics, etc.);

.2 the environmental conditions are specified in terms of wind and waves;

.3 the wind state is provided as a mean wind speed and a wind gustiness
spectrum;

.4 the mean heeling moment due to wind is determined starting from the mean
wind speed;

.5 the spectrum of the roll moment due to the wind gust is determined starting
from the wind gustiness spectrum;

.6 the sea state is provided in terms of a sea elevation spectrum, from which a
wave slope spectrum is directly determined;

.7 the spectrum of roll moment due to the waves is determined starting from the
wave slope spectrum;

.8 the total spectrum of roll moment is assumed to be given by the sum of the
spectrum of roll moment due to wind gustiness and the spectrum of roll
moment due to waves; and

.9 the dynamics of roll is then assumed to be modelled by means of a

linear 1-DOF equation of motion, which takes into account the ship
characteristics (roll restoring, roll natural period, roll damping), the mean
heeling moment due to the mean wind speed, and the irregular roll moment
due to the combined effect of gust and waves.

In this assessment, heeling angles to the leeward side are implicitly assumed to be positive,
whereas heeling angles to the windward side are assumed to be negative.

"Texp = exposure time, to be taken as equal to 3600 s;"

For the purpose of this assessment, the ship is conventionally assumed to be exposed to each
short-term environmental condition for one hour (3,600s).

2.2.3.2.2 A set of standard environmental conditions is assumed that refers to both the
short- and the long-term. The short-term characterization is given in terms of mean wind speed,
spectrum of wind gust and spectrum of sea elevation. The long-term characterization is given
in terms of a wave scatter diagram. A deterministic relation is assumed between the mean
wind speed and the significant wave height. The wave conditions are described in section 2.7.2
of the Interim Guidelines. The mean wind speed is considered to be correlated only with the
significant wave height.

The wind is assumed to fluctuate around the mean wind velocity. The total wind speed is given
by the sum of the mean wind speed UW (m/s) and the gust fluctuation speed v (m/s).

2.2.3.2.3 Once the roll motion equation is set up, it needs to be solved in order to provide
the necessary information for the estimation of the short-term failure index CS. To this end, a
simplified approximate methodology is used in order to obtain an estimation of the short-term
roll motion resulting from the 1-DOF modelling. From this information, the short-term failure
index CS can be determined.

I:\CIRC\MSC\1\MSC.1-Circ.1652.docx
MSC.1/Circ.1652
Annex, page 14

𝑆𝑣 (𝜔): The spectrum of the gust, Sv ((m/s)2/(rad/s)), is of the Davenport type, and it depends
on the mean wind speed UW (m/s). This represents a distribution of power of the fluctuating
wind velocity around the mean wind velocity in the range of the fluctuation circular frequency
𝜔 (rad/s). In addition, the coefficient, K, is a typical frictional drag coefficient of sea surfaces.
𝜇𝑒 : For explanation of the equivalent linear roll damping coefficient, μe, see
paragraph 9.3.5 of appendix 3 for details.

2.2.3.2.4 For further details and considerations for the assessment of the estimated
effective wave slope function, refer to appendix 3, section 8.

2.3 Assessment of ship vulnerability to the excessive acceleration failure mode

2.3.1 Application

2.3.1.2 The location where passengers or crew unprotected by a safety device such as seat
belts and harness may be present refers to spaces, such as the navigation bridge, and other
spaces designated for use by passengers that may normally be occupied in service.
Locations where passengers or crew may occasionally be present, and typically not in heavy
weather, should not be considered (e.g. areas to which access is provided by vertical means
of access only).

2.3.2 Level 1 vulnerability criterion for the excessive acceleration failure mode

2.3.2.1
4π2 hr
"φ⋅ kL ⋅ (g + ) ≤ REA1 "
T2r
The natural roll period, Tr, should be evaluated following paragraph 2.7.1.2 of the Interim
Guidelines.

"φ = characteristic roll amplitude (rad) = 4.43 r s / δφ0.5;"

The characteristic roll amplitude is determined under the assumption that the roll variance is
predominantly determined by frequencies close to the natural roll frequency.

"δφ = non-dimensional logarithmic decrement of roll decay;"

The non-dimensional logarithmic decrement coefficient  should be calculated as

δφ = 0.5⋅ π ⋅ RPR
where RPR is determined according to paragraph 2.5.2.1 of the Interim Guidelines.

2.3.3 Level 2 vulnerability criterion for the excessive acceleration failure mode

2.3.3.2
"C = ∑Ni=1 Wi CS,i "
This measure, indicated as C, is a long-term probability index ranging from 0.0 (good) to 1.0
(bad), which is obtained as a weighted average of short-term indices CS,i.

"𝑊𝑖 = weighting factor for the short-term environmental condition, as specified

in 2.7.2 of the Interim Guidelines;"

I:\CIRC\MSC\1\MSC.1-Circ.1652.docx
 MSC.1/Circ.1652
Annex, page 15

The weighting factor 𝑊𝑖 should be obtained from the wave scatter table such as table 2.7.2.1.2
of the Interim Guidelines as the ratio of the value described in the cell to the sum of the values
in all cells, which represents the relative frequency of occurrence of short-term environmental
condition specified with the significant wave height and the mean zero-crossing wave period.

"N = total number of short-term environmental conditions, according to 2.7.2

of the Interim Guidelines;"

The total number of short-term environmental conditions corresponds to the number of cells
with non-zero frequency of occurrences in the wave scatter table such as table 2.7.2.1.2 of the
Interim Guidelines because the short-term environmental conditions can be specified with the
significant wave height and mean zero-crossing wave period.

2.3.3.2.2 For a more realistic assessment of excessive accelerations, the wave spreading is
taken into account by multiplication with the reduction factor 3/4 (= 0.75).

"N = number of intervals of wave frequency in the evaluation range, not to be

taken less than 100;"

This number is different from the total number of short-term environmental conditions in
paragraph 2.3.3.2 of the Interim Guidelines.

"a, b = cosine and sine components, respectively, of the Froude-Krylov roll

moment in regular beam waves of unit amplitude ( kN  m / m ), calculated
directly or using an appropriate approximation;"

See section 8.3 of appendix 3.

"Be = equivalent linear roll damping factor (kN m s), with Be = 2JT,roll μe where
e (1/s) is the equivalent linear roll damping coefficient;"
See section 9.3 of appendix 3.

2.4 Assessment of ship vulnerability to the pure loss of stability failure mode

2.4.1 Application

2.4.1.1 The assessment of vulnerability to pure loss of stability is carried out only for ships
with relatively high service speed. If the speed is low, the duration over which the ship is
exposed to decreased stability (reduced righting lever) is generally too short for a large heel
angle to develop.

2.4.2 Level 1 vulnerability criterion for the pure loss of stability failure mode

2.4.2.1 The level 1 criterion is a simplified version of the level 2 criteria. While the level 2
criteria evaluate the vulnerability using the righting lever curve in waves, the level 1 criterion
assesses on the metacentric height, including free surface correction, in waves.
∇ D -∇
"GMmin ≥ RPLA and A ≥ 1.0"
W (D-d)

Paragraph 2.4.2 should be used for hull forms that do not feature significant tumblehome,
which is represented when the following guideline is met:

I:\CIRC\MSC\1\MSC.1-Circ.1652.docx
MSC.1/Circ.1652
Annex, page 16

∇D -∇
" < 1.0"
AW (D - d)
For such vessels, level 2 vulnerability assessment, direct stability assessment, or development
of operational measures may be applied.

"GMmin = minimum value of the metacentric height (m) calculated as provided in

2.4.2.2 of the Interim Guidelines;"

The level 1 assessment of the vulnerability to pure loss of stability is performed using a
conservative approximation of the smallest GM value during the wave pass.

2.4.2.2 The calculation of GM in waves should be based on the simplified formula for GMmin
that requires only hydrostatic data.
ITL
"GMmin = KB + ∇
- KG"
Use of this simplified formula may be recommended for relatively full ships, where significant
variation of stability while the wave overtakes the ship is not expected. Usually, the simplified
formula produces a more conservative result.

"ITL = transverse moment of inertia of the waterplane at the draught dL (m4);"

A conservative, simplified formula of GMmin on a wave evaluates the second moment of the
area of the waterplane at an artificial low draught to calculate a GMmin. If the hydrostatic data
available for the ship includes data for this low draught, this calculation should be able to be
performed without using stability software to calculate the GMmin.

2.4.3 Level 2 vulnerability criteria for the pure loss of stability failure mode

Although the initial stability represented by GM used in the level 1 criterion explains an onset
of stability failure, the final occurrence of stability failure itself should be judged using the
righting lever GZ as the level 2 criteria. Thus, the level 2 criteria are normally expected to be
less conservative than the level 1 criterion. On the other hand, the calculation of GZ in waves
requires more computational effort than that of GM.

2.4.3.1 The level 2 assessment requires evaluation of two criteria in the condition under the
reduced righting lever due to a wave.
" max(CR1 ,CR2 ) ≤ RPL0 "
The ship is considered not to be vulnerable if the largest value among CR1 and CR2 does not
exceed the standard (i.e. 0.06).

2.4.3.2

"𝑊𝑖 = weighting factor for the short-term environmental condition as specified

in 2.4.3.2.2 of the Interim Guidelines;"

The weighting factor 𝑊𝑖 should be obtained from the wave scatter table such as table 2.7.2.1.2
of the Interim Guidelines as the ratio of the value described in the cell to the sum of the values
in all cells, which represents the relative frequency of occurrence of short-term environmental
condition specified with the significant wave height and the mean zero-crossing wave period.
2.4.3.2.1 The level 2 vulnerability assessment for pure loss of stability is based on the
calculation of weighted criteria CR1 and CR2 on the basis of criteria C1i and C2i determined for

I:\CIRC\MSC\1\MSC.1-Circ.1652.docx
 MSC.1/Circ.1652
Annex, page 17

each calculation wave as defined in 2.4.3.2.2 and 2.4.3.2.3 of the Interim Guidelines. In order
to reduce the calculation effort, criteria C1 and C2 are pre-computed for the set of 11 waves
defined in 2.4.3.2.1 (where the case i = 0 corresponds to calm water) of the Interim Guidelines.
Criteria C1i and C2i are then linearly interpolated from the pre-calculated data, as indicated
in 2.4.3.2.2 of the Interim Guidelines.

The software used for calculation of the GZ curves in waves should be capable of balancing
the ship in sinkage and trim for each position of the wave crest and each heel angle. The free
surface correction, corresponding to the loading condition under consideration, has to be
applied.

Calculations are carried out for 11 locations of the waves crest for each wave height: with the
wave crest located at the amidships and 0.1L, 0.2L, 0.3L, 0.4L and 0.5L forward, and 0.1L,
0.2L, 0.3L, 0.4L and 0.5L aft thereof.

2.4.3.2.2 Refer to section 2.7.2 of these Explanatory notes for further details.

2.4.3.2.3 See appendix 3, section 7.

2.4.3.3 Criterion 1

Criterion 1 is an evaluation of the angle of vanishing stability, φV. If φV is less than a standard
value (30 degrees), the loading condition is considered to be vulnerable in the wave case under
consideration.

For each GZ curve, the angle of vanishing stability has to be evaluated. The smallest value of
the angle of vanishing stability Vmin for each wave height is used for further calculations.

2.4.3.4 Criterion 2

Criterion 2 is an evaluation of the stable heel angle under the action of a heeling lever, φS, due
to an external heeling force which is related to the Froude number. If φS is greater than a
standard value (15 degrees for passenger ships and 25 degrees for other ships), the loading
condition is considered to be vulnerable in the wave case under consideration.
1 φsw > KPL2
"C2i = { "
0 otherwise
For each wave height i, the heel angle sw is the maximum static heel angle in a wave pass.
Hi
"lPL2 =8 ( ) dFn2 (m); "
λ
For each wave height Hi, an arm of a heeling moment lPL2i is computed as given in 2.4.3.4 of
the Interim Guidelines.

For each value of the arm of a heeling moment lPL2i and each position of the wave crest, a
static angle S is computed; if the computation is not possible (lPL2i exceeds maximum of the
GZ curve), the conventional value of 180 degrees is used.

2.5 Assessment of ship vulnerability to the parametric rolling failure mode

2.5.2 Level 1 vulnerability criterion for the parametric rolling failure mode

The background of the vulnerability level 1 criterion is available in section 6 of appendix 3.

The definition of sharp bilge is provided in paragraph 5.2 of appendix 3.

I:\CIRC\MSC\1\MSC.1-Circ.1652.docx
MSC.1/Circ.1652
Annex, page 18

2.5.3 Level 2 vulnerability criteria for the parametric rolling failure mode

2.5.3.1 C1 is also referred to as the "first check" and C2 is also referred to as the "second
check". For the background of the level 2 criteria, refer to section 6 of appendix 3. The standard
for the second check of the level 2 criterion was determined with the reports of a major large
heel incident of a containership. The containership encountered a severe storm in the North
Pacific which resulted in rolling of up to 40 degrees port and starboard in bow seas and the
loss overboard or destruction through collapse of more than 800 deck stowed containers.
The model experiment and numerical simulation revealed that the cause of the accident was
parametric rolling in head seas without the forward velocity. The sample calculation result used
to determine the standard of the level 2 criterion is provided in section 5.4 of appendix 2 as an
application example.3

2.5.3.2.1 For further details and considerations for the assessment of GM(Hi, λi), refer to
appendix 3, paragraphs 7.4 and 7.5.

2.5.3.2.3 Details regarding the procedure for the determination of wave cases in
table 2.5.3.2.3 of the Interim Guidelines are reported in section 10 of appendix 3.

2.5.3.3 The factors Ki in table 2.5.3.3 were derived from an original idea where the ship speed
was considered equal to the service speed and the heading of the vessel with respect to the
waves was sampled from longitudinal to beam waves, leading to an orthogonal projection of
ship service speed on the wave direction Vs ∙Ki with Ki = cos((i - 1) π/2N) i=1,2,…,N. However,
in the final version of the criterion in the Interim Guidelines, for realizing a simple and
conservative estimation, the waves are considered to be purely longitudinal and the forward
ship speed used in the calculations is varied on the basis of factors Ki in table 2.5.3.3. The
amplitude of restoring variation due to waves decreases when the ship heading is deviated
from the condition of longitudinal waves and the roll damping decreases when the ship speed
is smaller than the service speed. Thus, calculating the restoring variation in purely longitudinal
waves and the roll damping at a forward speed corresponding to Vs ∙Ki realizes a conservative
estimation. When the value of N is sufficiently large, the C2 index is expected to converge to a
certain value. Fig. 2.5.3.3 indicates the case of the C11 class containership shown in
appendix 2. In this case, N = 12 is sufficient. The formula in the Interim Guidelines uses N = 12.

The number of speeds that were sampled, 2N+1

3
Note: The information of the accident is taken from the document SLF45/6/7 "Head-sea parametric rolling
and its influence on container lashing systems" submitted by the United States in 2002.

I:\CIRC\MSC\1\MSC.1-Circ.1652.docx
 MSC.1/Circ.1652
Annex, page 19

Fig. 2.5.3.3 The relationship between the C2 index and the number of wave heading for
the C11 class containership

2.5.3.3.1 The equation of motion takes into account forces acting on the ship. The simplest
mathematical model that is capable of evaluating the maximum roll angle includes four
moments:

.1 inertia, including added inertia (or added mass) as a part of hydrodynamic

forces;

.2 roll damping, which expresses energy loss from roll motions in creating
waves, vortices and skin friction;

.3 roll restoring (stiffness) is determined from calculation of GZ in waves as

specified in 2.5.3.4.1 of the Interim Guidelines; and

.4 transverse wave forces are absent for a ship in exact following or head
long-crested seas.

Proper estimation of these four components is explained in greater detail in appendix 3,

sections 1 and 2.

2.5.3.4 The maximum roll angle to which reference is made in paragraph 2.5.3.4 of the Interim
Guidelines corresponds to the maximum roll angle when roll motion has reached a steady
state. The parametric roll response includes the initial transition from the initial conditions to a
steady state in which roll amplitudes are similar. Different criteria for "similarity" can be used:
relative (the difference is less than 3 - 5%) or absolute (less than one degree). Following these
criteria, the steady state portion of the response can be extracted (see figure 2.5.3.4.1) and
the resultant maximum roll angle can be found as an average of steady state roll amplitudes.

φ, deg
40

20
t, s
460 480 500 520 540
-20

-40

Figure 2.5.3.4.1 – Steady state portion of the roll motion

in parametric resonance conditions

The steady state parametric rolling is not the only possible type of roll response. If parametric
rolling is not possible for the given wave conditions, the response could be represented by
decaying roll oscillations – as shown in figure 2.5.3.4.2. The response is not expected to look
like a decaying sine function because of both the parametric excitation and non-linearity of the
equation of motion. In some cases, the roll angle may initially increase and then damp out.

I:\CIRC\MSC\1\MSC.1-Circ.1652.docx
MSC.1/Circ.1652
Annex, page 20

φ, deg
5

t, s
0 100 200 300 400 500

-5

Figure 2.5.3.4.2 – Roll response in absence of parametric rolling

Another possible response may include "capsizing" (see figure 2.5.3.4.3) if the GZ curve was
computed for the entire range of 180 degrees (like in figure 2.1 of appendix 3). If the GZ curve
is not computed for the full range, the calculation should be explicitly stopped once the roll
angle exceeds the cut-off roll angle.
φ, deg
200

100
t, s

0 100 200 300 400 500

-100

Figure 2.5.3.4.3 – Roll response with parametric roll and capsizing

2.5.3.4.1 For further details and considerations for the assessment of the roll angle, refer to
section 4 of appendix 3.

2.5.3.4.2 Refer to section 7 of appendix 3 for further details.

2.6 Assessment of ship vulnerability to the surf-riding/broaching failure mode

2.6.2 Level 1 vulnerability criteria for the surf-riding/broaching failure mode

2.6.2.1 The criterion and the standard for the Froude number in the level 1 criterion were
adopted as a part of the MSC/Circ.707 in 1995 and then superseded by MSC.1/Circ.1228.
This guidance concludes that, under the appropriate wave conditions, surf-riding may occur
when the ship speed is higher than:
1.8√L
VS ≥
cos(180∘ - β)
where VS is the speed of the ship in knots and  is the angle of wave encounter in degrees.
Assuming following seas  =  and transforming the formulae to be based on the Froude
number yields:
1.8 ⋅ 0.5144
Fn ≥ = 0.296 ≈ 0.3
√g
This is regarded as the lower limit of threshold for surf-riding under any initial condition for
conventional ships in the worst waves. It is compared with examples of such thresholds
calculated for some sample ships with the Melnikov analysis, which was used in the level 2
criterion, as shown in figure 2.6.1. This result also indicates that the worst wavelength for
surf-riding is comparable to the ship length. This means that longer ships require longer waves

I:\CIRC\MSC\1\MSC.1-Circ.1652.docx
 MSC.1/Circ.1652
Annex, page 21

for surf-riding but the lengths of steep ocean waves have certain limits so that a small possibility
for surf-riding exists for very long ships. Thus, the level 1 vulnerability criteria also require a
threshold of ship length, i.e. 200 m.

Systematic calculation results of the stability failure due to broaching by using the critical wave
method, which is specified in section 5.2 of appendix 4, and geometrical scaled ship hulls
shown in figures 2.6.2 and 2.6.3 indicate that, if the ship length exceeds 200 m, the danger of
broaching could be regarded as very small.

Figure 2.6.1 – Thresholds of surf-riding under any initial condition for some sample
ships in the wave steepness (H/λ) of 1/10 with different wave length to ship length ratio
(λ/L) compared with the nominal Froude numbers (Fn) of 0.3

Figure 2.6.2 – Broaching-induced stability failure probability per hour for geometrically
scaled ro-ro ship hulls with different sizes in the North Atlantic

I:\CIRC\MSC\1\MSC.1-Circ.1652.docx
MSC.1/Circ.1652
Annex, page 22

Figure 2.6.3 – Broaching-induced stability failure probability per hour for geometrically
scaled fishing vessel hulls, which is used in section 6 of appendix 2, with different
sizes in the North Atlantic

2.6.3 Level 2 vulnerability criterion for the surf-riding/broaching failure mode

2.6.3.2

"W2(Hs, Tz) = weighting factor of short-term sea state specified in 2.7.2.1 as a

function of the significant wave height, HS, and the zero-crossing
wave period, TZ, in which W2(Hs, Tz) is equal to the number of
occurrences of the combination divided by the total number of
occurrences in the table, and it corresponds to the factor Wi
specified in 2.7.2 of the Interim Guidelines;"

The weighting factor W2 should be obtained from the wave scatter table such as table 2.7.2.1.2
of the Interim Guidelines as the ratio of the value described in the cell to the sum of the values
in all cells, which represents the relative frequency of occurrence of short-term environmental
condition specified with the significant wave height and the mean zero-crossing wave period.

2.6.3.2.1 The calculation formula of, Wij , in paragraph 2.6.3.3 of the Interim Guidelines is the
joint probability density function of local wave steepness and local wavelength under the
stationary wave state with Bretschneider wave spectrum.

The envelope of an irregular wave time history is described with the slowly varying amplitude
and phase. Then the joint probability density of the envelope amplitude, phase and their time
derivatives can be calculated as the Gaussian process. If the wave spectrum is reasonably
narrow, the envelope amplitude can be regarded as one half of the local wave height and the
time derivative of the envelope phase can be related to the local wave period because of the
wave dispersion relation. Thus, the joint probability density function of the local wave height
and the local wave period can be obtained by transforming the envelope amplitude, phase,
and their time derivatives.

2.6.3.2.2 Further details on the assessment of the regular wave are provided in section 5.4
of appendix 1.

2.6.3.2.3 Because the level 1 criterion is a simple assessment of Froude number, no

resistance or thrust data is needed. However, the level 2 vulnerability criteria require the use
of reliable estimates of resistance and thrust. Before a ship is completed and sea trials are
performed, this data comes from model tests or other resistance estimation techniques.
With respect to model tests, not all ship designs are tested to the high Froude numbers that

I:\CIRC\MSC\1\MSC.1-Circ.1652.docx
 MSC.1/Circ.1652
Annex, page 23

correspond to the speeds at which surf-riding could be expected to occur. Therefore, an

accurate procedure is required for prediction of the resistance at high Froude numbers, based
on measured resistance data only extending to modest Froude numbers.

The calm water resistance, R(u), can be estimated either by using results of geometrically
scaled model tests and the standard scaling law or by using a numerical method with the
agreement of the approval authority. The ship's resistance should be estimated to a ship speed
up to 20% over the maximum service speed or the phase velocity (celerity) of the fastest wave
under consideration.

The calm water resistance curve, R(u), is constructed based on the available resistance data
using a polynomial approximation which may, but need not, include terms up to the 5th power:
5

R(u) ≈ ∑ ri ui = r1 u + r2 u2 + r3 u3 + r4 u4 + r5 u5
i=1

where, u speed of the ship (m/s) in calm water and r1, r2, r3, r4, r5 approximation coefficients
for the calm water resistance.

The polynomial fit should be appropriate to ensure the resistance is continuously increasing
as a function of speed in the appropriate range. The polynomial fit to approximate resistance
curve requires caution. Available data points may not extend to the phase velocity (celerity) of
the fastest wave under consideration. If this is the case, the following condition should be
verified for all values of ship speeds, u, up to the phase velocity of the fastest wave in
consideration, umax:
r1 + 2r2 u + 3r3 u2 + 4r4 u3 + 5r5 u4 > 0
where,
3gL
umax =
2

2.6.3.2.4 The propeller thrust should be estimated using geometrically scaled model tests
and standard scaling law or using a numerical method that is agreeable to the approval
authority. The propeller advance ratio range should cover the whole positive range of propeller
thrust coefficient.

The thrust deduction, tp, should be evaluated using geometrically scaled model tests and
standard scaling law. In absence of ship-specific model test data, the following approximations
can be made:
tp = 0.1, for single screw ships; and
DP
tp = 0.325⋅CB - 0.1185 , for twin screw ships;
√B⋅d

The wake fraction, wP, should be evaluated using geometrically scaled model tests and
standard scaling law. In absence of ship-specific model test data, a conservative assumption
of the wake fraction wP as 0.1 can be made.

Alternative methods can be used with the agreement of the approval authority.

For ships using propulsor(s) other than open propeller(s) or are using unconventional
propulsion arrangement, the propulsor thrust may be evaluated with a method appropriate to
the propulsor used and with the agreement of the approval authority.

I:\CIRC\MSC\1\MSC.1-Circ.1652.docx
MSC.1/Circ.1652
Annex, page 24

Podded propulsion: With podded propulsion, the underwater body of the unit affects the
propeller thrust considerably, and therefore the KT curve of the whole pod unit in open water
should be used. Otherwise, the same method for the conventional propeller can be used.

2.6.3.2.5 The critical number of revolutions of the propulsor corresponding to the surf-riding
threshold, ncr (rj, si), (rps) in paragraph 2.6.3.4.6, can be calculated by solving the equation by
using an appropriate method. In the case that the resistance in calm water is approximated
with the 5th power polynomial, the critical number of revolutions, ncr, corresponding to the
second surf-riding threshold may be calculated as follows for each wave case:

-m1 +√m21 -4m0 m2

ncr =
2m0
where
τ τ1 ci
m2 =-2π f0 ; m1 =2π fij
+8a0
ij

τ2 c2i -R(ci ) 64 1024

m0 =2π +8a1 -4πa2 + a3 -12πa4 + a
fij 3 15 5
The coefficients a1, a2, a3, a4 and a5 do not have indexes of the wave cases in order to keep the
formulae simple; however, ci and ki depend on the wave length and fij depends on both
wavelength and height. The complete amplitude for the wave force with the wave length, i,
and wave height, Hij, is calculated by formulae from paragraph 2.6.3.4.5.

Coefficients ri that are not used in the fitting of the calm water resistance curve shall be taken
equal to 0.

2.7 Parameters common to stability failure mode assessments

2.7.1 Inertial properties of a ship and natural period of roll motion

2.7.1.1 The factor 1000 appearing in the formula for JT,roll in paragraph 2.7.1.1 of the Interim
Guidelines is related to the fact that JT,roll is measured in t·m2.

2.7.1.2 For ships carrying containers, the following formula can be used alternatively for the
natural roll period:
Tr = 2π√(Ixx + A44 )/(mgGM)
where,
Nc
B 2 2 2 2
(b2i + h2i )mi
Ixx + A44 = 1.1m0 ( ) + 1.1m0 HSH + 1.1 ∑ {mi (yi + zi - zT ) +
( ) }
3 12
i=1

where,
m is the mass of ship in kilograms, HSH is the distance from the centre of mass of ship to the
centre of mass of the ship without containers on deck in metres, mi is the mass of each
container loaded on deck in kilograms, yi and zi in metres, are the transverse and vertical
coordinates of the container centre of mass, respectively, bi and hi in metres, are the breadth
and height of the container, respectively, zT is the vertical height of the ship centre of gravity in
metres, Nc is the number of containers and m0 is the mass of the ship without containers on
deck in kilograms.

I:\CIRC\MSC\1\MSC.1-Circ.1652.docx
 MSC.1/Circ.1652
Annex, page 25

2.7.1.3 For all ships, the following formula can be alternatively used for Ixx + A44:
Ixx + A44 = mK2
where,
K 2 Hsp Hs 2
( ) = 0.125 {Cu ∙ Cb + 1.10Cu (1 - Cb ) ( - 2.20) + ( ) }
B d B
Au
cu =
Lu B
A'
Hsp =D+ ( )
Lpp
cu = upper deck area coefficient
Au = projected area of upper deck
Lu = overall length of upper deck
Hsp = effective depth
A' = A + Ac
A' = lateral projected area of the forecastle and deck house (A) and
on-deck cargoes (Ac) .

2.7.2 Environmental data

2.7.2.1 This wave scatter table is taken from IACS Recommendation No.34
(Corr.1 Nov. 2001). It is based on data collected from the 1960s through the 1980s; it was
revised slightly in 2001. Wave data collected since 2001 is not represented in this table.

The sum of all numbers of occurrences for the given range of Tz and Hs out is 100,000.
Therefore, the number of occurrences for one cell divided by 100,000 is the likelihood or
probability of these combinations of Tz and Hs to be a short-term environment for a ship in a
given loading condition.

While the table does not explicitly state, the number of occurrences out of 100,000 presented
in each combination or cell of Tz and Hs reflect occurrences that are within ±0.5s of the given
Tz and within ±0.5m of the given Hs. Hence, the values of Tz and Hs for a given cell are the
centre values, respectively, of Tz and Hs, but not necessarily the mean values. The relationship
between peak wave period and mean zero-crossing wave period can be calculated for
Bretschneider wave energy spectrum as Tz = 0.710 Tp.

3 Guidelines for direct stability failure assessment

3.1 Objective

3.1.2 The criteria and assessment procedures are detailed in these explanatory notes; the
background of the standard of 2.6·10-3 stability failures per ship per year is explained in
paragraph 4.6.3, section 4.7 of appendix 4.

3.1.4 The dynamics of water on deck may significantly contribute to stability in waves for
ships with an extended low weather deck. Since modelling of the effect of water on deck is
difficult even with numerical methods that are employed in the direct stability assessment, it is
not addressed in the present guidelines. However, if the effect of water on deck is ignored in
numerical simulations, the results may be too conservative for the dead ship condition and
pure loss of stability failure modes.

I:\CIRC\MSC\1\MSC.1-Circ.1652.docx
MSC.1/Circ.1652
Annex, page 26

3.2 Requirements

3.2.1 The present guidelines consider the occurrence of excessive roll angles or excessive
rigid-body accelerations as intact stability failures. The former may result in capsizing while
both may impair normal operation of the ship and could be dangerous to crew, passengers,
cargo or ship equipment. As excessive roll angles, the following limits are considered: 40
degrees, exceedance of which may lead to malfunctioning of the ship's engine or to a major
shift of cargo, and thus static heel angle, which reduces freeboard; angle of vanishing stability
in calm water, exceedance of which is associated with capsizing; and angle of submergence
of unprotected openings in calm water, exceedance of which is associated with possible water
ingress in internal compartment, and thus significant change in ship stability characteristics.
In the direct stability assessment, the smallest of these three values should be used.
As excessive lateral acceleration, the value of 9.81 m/s2 is employed. These definitions of
stability failures imply rather severe consequences. Depending on the particular case, stricter
requirements may be applied if necessary.

3.2.2 Direct stability assessment should be performed with empty anti-roll tanks and
retracted anti-roll fins. If anti-roll fins are not retractable, they should be considered as fixed in
neutral position.

3.3 Requirements for a method that adequately predicts ship motions

3.3.2 General requirements

3.3.2.1 Modelling of waves

3.3.2.1.1 Several methods are available in engineering practice for modelling of irregular
waves: they include, for example, computing the wave elevation as a sum of harmonic
components with random phases and band-pass filtering the pseudo-white noise generated
by maximum-length sequence.

3.3.2.1.2 Modelling irregular waves as a finite sum of harmonic components with random
phases is a simple method, which computes the time history of wave elevation 𝜁(𝑡) as
𝜁 (𝑡) = ∑𝑀
𝑖=1 𝑎𝑖 𝑐𝑜𝑠 (𝜔𝑖 𝑡 + 𝜖𝑖 )
1/2
where ai = {2SZZ (ωi ) D(μi ) Δωi Δμi } are amplitudes, ωi frequencies, μi directions and ϵi
phases of harmonic components, SZZ is the wave energy spectrum and 𝐷 is the wave energy
angular spreading function. The phases of harmonic components 𝜖𝑖 are randomly selected in
the interval [0,2𝜋); their frequencies ωi , directions 𝜇𝑖 and amplitudes 𝑎𝑖 may also be randomly
varied. To generate random values, pseudo-random number generators can be used. In this
method, if more than one of ratios between different two component frequencies are rational,
the time history has a self-repetition period. If so, the duration of each simulation should be
limited, to avoid self-repetition effects, using the following requirements:

.1 absence of self-repetition of waves, which can be checked by computing the

autocovariance function of the free surface elevation using the wave energy
spectrum,
R(t) = ∑M
i=1 SZZ (ωi ) D(μi ) cos(ωei t) Δωei Δμi

where ωei = ωi - ω2i g-1 VS cos(μi ) is the encounter frequency. Until the peaks
of the autocovariance function do not increase with increasing simulation
time 𝑡, the duration of the simulation is still sufficiently small to ensure the
absence of self-repetition of waves; and

I:\CIRC\MSC\1\MSC.1-Circ.1652.docx
 MSC.1/Circ.1652
Annex, page 27

.2 absence of self-repetition of roll motion, for which the absence of

self-repetition of waves may not be sufficient due to narrow-banded nature
of roll excitation. To verify the absence of self-repetition effects of roll motion,
quantile plots can be used (see section 3.7 of appendix 4 for details).

3.3.2.1 Modelling of roll damping: avoiding duplication

3.3.2.2.2 Regarding item .2 addressing CFD computations, refer to the ITTC recommended
guidelines 7.5-03-02-03 (issued in 2014 or amended).

3.3.3 Requirements for particular stability failure modes

3.3.3.5 For excessive acceleration stability failure mode, the minimum requirements to the
resolved degrees of freedom do not include the sway motion. Since sway contributes to lateral
acceleration, special consideration should be given to accurate (or, at least, conservative)
reproduction of lateral acceleration if sway motion is not explicitly modelled.

In these cases, a proof should be provided that the error in the lateral acceleration is either
conservative or, if non-conservative, does not exceed 10% or 0.05g, whichever is larger.

3.4 Requirements for validation of software for numerical simulation of ship

motions

3.4.2 Qualitative validation requirements

Examples for the backbone curve and the roll response curve are shown in figures 3.4.1
and 3.4.2.

Figure 3.4.1 – Left: backbone curve (roll amplitude vs. roll frequency from roll decay
simulation); Right: righting lever curve (GZ) and derivative of GZ with respect to heel
angle vs. heel angle

Figure 3.4.2 – Left: roll amplitude in regular waves vs. wave frequency (one line per
wave height) and line through peak roll response (solid black line with circles); Right:
righting lever curve (GZ) and derivative of GZ with respect to heel angle vs. heel angle

I:\CIRC\MSC\1\MSC.1-Circ.1652.docx
MSC.1/Circ.1652
Annex, page 28

With reference to the change of stability in waves, see example in section 2.3 of appendix 4.
The background for this phenomenon is presented in chapter 3 of appendix 1.

With respect to the response curve for parametric roll in regular waves, additional information
is presented in section 2.4 of appendix 4. The background for this phenomenon is presented
in chapter 4 of appendix 1.

With reference to surf-riding equilibrium, the background for these phenomena is presented in
chapter 5 of appendix 1.

3.4.3 Quantitative validation requirements

3.5.5.3 Examples for the roll response curves are shown in figures 3.4.3 and 3.4.4.

With reference to the change of stability in waves, see example in section 2.3 of appendix 4.
The background for this phenomenon is presented in chapter 3 of appendix 1.

With respect to the response curve for parametric rolling in regular waves, additional
information is presented in section 2.5 of appendix 4.

Figure 3.4.3 – Computed and measured amplitude of roll in regular head waves

Figure 3.4.4 – Measured (EXP.) and computed with three numerical methods (NM1,
NM2, NM3) double amplitude of roll in regular beam waves

3.5 Procedures for direct stability assessment

3.5.2 Verification of failure modes

3.5.2.1 The physical background of stability failure modes is explained in appendix 1.

I:\CIRC\MSC\1\MSC.1-Circ.1652.docx
 MSC.1/Circ.1652
Annex, page 29

Verification of a failure mode in a direct stability assessment means identification of the

observed failure with one of the five modes addressed. Unambiguous identification of stability
failure mode may be difficult. Identification is needed to exclude an irrelevant failure from the
count, e.g. case of heel caused by broaching from assessment addressing parametric roll.

Relevant definitions are given below:

• Local roll period is the zero-crossing period of the roll motion, containing the
stability failure event.

• Local heave period is the zero-crossing period of the heave motion, containing
the stability failure event.

• Local pitch period is the zero-crossing period of the pitch motion, containing the
stability failure event.

• Local wave encounter period is the zero-crossing period of the wave elevation at
the centre of gravity of the ship, containing the stability failure event.

• Natural roll period is defined in calm water for the maximum roll amplitude during
the roll period containing the stability failure. Due to the non-linearity of roll motion,
the wave encounter frequency corresponding to the largest roll response may
significantly differ from the linear natural roll frequency, as shown in figures 2.1.1,
2.2.3 and 4.3.1 of appendix 4.

Chapter 7 of appendix 4 shows application examples of the verification of failure mode.

3.5.2.2 The background for verification of the pure loss of stability failure mode is presented
in chapter 3 of appendix 1. An example of a verification of the failure mode is presented in
appendix 4, section 7.2.

3.5.2.3 The background for verification of the parametric roll stability failure mode is
presented in chapter 4 of appendix 1. Particular attention is directed to figure 4.2 of appendix 1.
Examples of verifications of the failure mode are presented in appendix 4 and section 7.1.

3.5.2.4 The background for verification of the surf-riding/broaching stability failure mode is
presented in chapter 5 of appendix 1. An example of a verification of the failure mode is
presented in appendix 4, section 7.3.

3.5.2.5 The background for verification of the dead ship condition and the excessive
acceleration stability failure modes are presented in chapters 1 and 2, respectively, of appendix
1. An example of a verification of the failure mode is presented in appendix 4, section 7.4.

3.5.3 Environmental and sailing conditions

3.5.3.1 General approaches for selection of environmental and sailing conditions

3.5.3.1.3 For short-crested irregular waves, wave energy is spread with respect to the mean
wave direction. The cosine-squared wave energy spreading is given by
2 2
𝐷(μ’ ,μ)= ( ) {max(0, cos(μ’ – μ))}
π

I:\CIRC\MSC\1\MSC.1-Circ.1652.docx
MSC.1/Circ.1652
Annex, page 30

where μ’ is the wave direction and μ is the mean wave direction. For long-crested irregular
waves, D(μ’ ,μ) = 1, when μ’ = μ and 0 otherwise.

Since short-crested waves provide more realistic representation of sea state compared to
long-crested waves, this model can be used both in numerical simulations and model tests for
any stability failure mode. At the same time, the use of long-crested waves in numerical
simulations or model tests may be less expensive and is, in many cases, more conservative.
In such cases, the long-crested wave model can be used if it is more practicable for the
particular stability failure mode and numerical simulation method or model test.

If assessment employs irregular short-crested waves, the numerical method used should be
validated in model experiments in irregular short-crested waves or using a sufficient number
of wave directions in irregular long-crested waves.

3.5.3.2 Full probabilistic assessment

3.5.5.3.1 The estimate of the mean long-term rate of stability failures is a random variable
subject to uncertainty. To take this into account, the employed practical criterion is
the upper boundary 𝑟̄U of the 95%-confidence interval of the average “long-term”
stability failure rate, paragraph 3.3.5 of appendix 4. This criterion can be
conservatively calculated (for each addressed loading condition) as a weighted
average, over relevant sea states (Hs, Tz) and sailing conditions (vs, ), of the upper
boundary rU(Hs,Tz,,vs) of the 95%-confidence interval of the “short-term” stability
failure rate:

r̄ U = ∑ ∑ ∑ fs (Hs ,Tz )⋅fμ (μ)⋅fv (vs )⋅rU (Hs ,Tz ,μ,vs )ΔHs ΔTz Δvs Δμ
s μ vs

where fs(Hs, Tz) is the probability density of sea states, equal to wi / (HsTz), where Wi are the
weights provided in the scatter table, Hs, Tz are ranges of significant wave height and
zero-crossing wave period of cells in the scatter table, f() is the probability density of wave
directions, and fv(vs) is the probability density of forward speeds; note that ∑μ fμ (μ)Δμ =1,
∑vs fv (vs )Δvs = 1 and ∑s fs (Hs ,Tz ) Δh Hs ΔTz = 1.

In the assessment, the resolution of significant wave heights Hs should be at least 1.0 m, the
zero-crossing wave periods Tz should be at least 1.0 s and the mean wave directions 
should be at least 15 degrees. The range of ship forward speeds from zero to full service speed
should be divided into at least 6 intervals vs.

The upper boundary rU(Hs, Tz, , vs) of the 95%-confidence interval of the “short-term” stability
failure rate is obtained for each combination (Hs, Tz,  vs), by either direct counting, an
extrapolation of rU(Hs, Tz,  ,vs) over the significant wave height Hs or another statistical
extrapolation method.

The stability failure rate, evaluated with direct stability assessment, is the statistical estimate.
Statistical estimates have a natural variability, i.e. different numerical simulations or model
tests may produce different values due to random reasons. This variability is characterized by
a confidence interval, as shown in figure 3.5.1. The confidence interval contains a true value
of the failure rate with the confidence probability that is usually taken as 95%. The upper
boundary rU of the 95%-confidence interval of stability failure rate is used in direct stability
assessment in section 3.5.3.2 of the Interim Guidelines. This boundary is obtained for each
combination (Hs, Tz,  vs) by either direct counting, extrapolation of failure rate over significant
wave height or another statistical extrapolation method.

I:\CIRC\MSC\1\MSC.1-Circ.1652.docx
 MSC.1/Circ.1652
Annex, page 31

pdf(x)

Confidence
probability

x
Lower Confidence interval Upper
boundary boundary

Figure 3.5.1 – Confidence interval of normally distributed estimate

3.5.3.2.2 See paragraph 4.6.4 of appendix 4.

3.5.3.3 Assessment in design situations using probabilistic criteria

3.5.5.3.1 Whereas the full probabilistic assessment requires evaluation of stability failure rate
for a large (of the order of magnitude 104) number of combinations of sea state
parameters (Hs, Tz) and sailing conditions (vs, ), the idea of the direct stability
assessment in design situations, see section 4.4 of appendix 4, is to reduce the
evaluation of stability failure rate to few (up to 20) combinations (Hs, Tz, vs, ),
referred to as design situations; the employed sea states are rather steep to
increase the stability failure rate and thus minimize the total simulation or testing
time required for the assessment. The design situations are specific for stability
failure modes.

3.5.3.3.2 Example procedure for probabilistic assessment in design situations is described

in section 4.7.10 of appendix 4.

3.5.3.3.3 The threshold  in paragraph 3.5.3.3.2 is equal to 1.389∙10-4 1/s

(= 1 / (3600s × 2)) because it represents one stability failure every two hours.

3.5.3.3.4 Explanations to table 3.5.3.3.4 of the Interim Guidelines – Design situations for
each stability failure mode:

.1 range of zero-crossing wave periods Tz is specified for the dead ship

condition and excessive acceleration stability failure modes in terms of its
ratio, 0.7 to 1.3, to the natural roll period Tr. To define the natural roll period,
free roll decay simulations or tests should be performed before the
assessment;

.2 for pure loss of stability and surf-riding/broaching stability failure modes, the
appropriate assumption for the forward speed should be the maximum
attainable speed in waves considering added resistance due to wind and
waves, propulsion system and engine diagram; however, as a conservative

I:\CIRC\MSC\1\MSC.1-Circ.1652.docx
MSC.1/Circ.1652
Annex, page 32

simplification, the assessment is performed at maximum nominal service

speed, i.e. at the forward speed that is attained in waves at the propeller
rotation speed equal to the propeller rotation speed at the maximum ship
service speed in calm water;

.3 for pure loss of stability and surf-riding/broaching stability failure modes, the
wave period is given in terms of peak wave period corresponding to specified
wave lengths (1.0L for pure loss of stability and from 1.0L to 1.5L for
surf-riding/broaching); peak wave period corresponding to wave length  is
defined as Tp = (2πλ/g)1/2 and the mean zero-crossing wave period can be
calculated for Bretschneider wave energy spectrum as Tz = 0.710 Tp; and

.4 for parametric roll stability failure mode, assessment is performed in head

and following waves at zero forward speed; in model tests, soft spring
arrangement can be used to keep ship’s position and heading at zero speed.

3.5.3.3.5 One significant wave height Hs is used per mean zero-crossing wave period Tz.
This wave height corresponds to the probability density of sea state fs = 10-5 1/(m∙s).
The probability densities of sea states are defined as fs(Hs, Tz) = Wi / (HsTz), where Wi are the
weights provided in the scatter table and Hs, Tz are the ranges of significant wave height and
zero-crossing wave period of cells in the scatter table. To define the probability density of a
sea state with significant wave height Hs between significant wave heights Hs1 and Hs2,
corresponding to centres of the ranges in the scatter table, linear logarithmic interpolation can
be used:
𝛾 1−𝛾
𝑓𝑠 = 𝑓𝑠1 𝑓𝑠2 with γ=(Hs2 – H)/(Hs2 – Hs1 ) (similarly for interpolation between wave periods).

3.5.4 Direct counting procedure

3.5.4.1 The proposed example procedures are based on simulations of ship motions in
multiple independent realizations of the same irregular seaway. The seaway is modelled as a
sum of harmonic components. Independent realizations are produced by random variation of
the phases and, possibly, frequencies, directions and amplitudes of the components.
To generate random values, pseudo-random number generators are used.

3.5.4.2 The background and application examples of direct counting procedures to estimate
the upper boundary of the 95%-confidence interval of stability failure rate are described in
sections 3.3, 3.4 and 3.5 of appendix 4.

3.5.4.3 Since the available direct counting procedures assume that the occurrence of stability
failure can be described as a Poisson process, they should prevent self-repetition effects,
transient hydrodynamic effects at the beginning of simulations and autocorrelation of large roll
motions. To neutralize the effect of self-repetition, duration of each simulation is limited.
The effect of autocorrelation of big roll motions can be neutralized by stopping or ignoring a
simulation after an encountered stability failure. The effect of transient hydrodynamic effects
at the beginning of simulations can be neutralized by either switching off the counter of stability
failures and simulation timer during initial transients or by random variation of initial conditions
for each realization.

3.5.4.4 Regarding items .1 and .2:

.1 Numerical simulations or model tests are carried out for arbitrary

(e.g. constant) simulation time, which is limited by the maximum duration or
first stability failure, whichever happens earlier. After each simulation, the

I:\CIRC\MSC\1\MSC.1-Circ.1652.docx
 MSC.1/Circ.1652
Annex, page 33

number of stability failures encountered in the simulation N (1 or 0) and

duration of simulation t (time to failure if realization ended with a stability
failure or full duration of simulation otherwise) are recorded, and

.1 N* is calculated as N before the last simulation plus one;

.2 total number of failures N is increased by N;
.3 total simulation time tt is increased by t;
.4 maximum likelihood estimate of failure rate is updated as 𝑟̂ = 𝑁/𝑡t;
.5 conservative estimate of maximum likelihood estimate of the failure
rate is calculated as r̂ * =N* /tt; and
.6 upper and lower boundaries of 95%-confidence interval of the failure
rate are updated as rU =0.5χ21-0.05/2,2N* r̂ * /N* and
rL =0.5χ20.05/2,2N r̂ /N , respectively, where χ2p,f denotes the p∙100%-
quantile of χ2-distribution with f degrees of freedom (details can be
found in section 3.3 of appendix 4).

.2 Numerical simulations are carried out with a common predefined maximum

exposure time texp for each single realization. If a stability failure is
encountered in a simulation, further time history does not need to be
considered, and the simulation can also be stopped before texp for that
realization. After each simulation, the total number of carried out realizations,
M , and the total number of realizations during which at least one stability
failure has been observed, N , are recorded. The increase of the counter N
from the previous realization is N , with N = 1 if at least one stability
failure occurred during the present simulation and N = 0 otherwise. The
maximum likelihood estimate of the probability of at least one stability failure
in an exposure time texp is calculated as p = N / M and the corresponding
*

failure rate is calculated as 𝑟 ∗ = − 𝑙𝑛(1 − 𝑝∗ ) /𝑡𝑒𝑥𝑝 . The lower and upper

boundaries of the 95%-confidence interval for the probability of at least one
stability failure in an exposure time texp are then calculated. The lower
boundary is calculated as 𝑝𝐿 = 𝜈1 ⋅ 𝐹𝜈1 ,𝜈2;0.05/2 /(𝜈2 + 𝜈1 ⋅ 𝐹𝜈1,𝜈2 ;0.05/2 ) with
𝜈1 = 2 ⋅ 𝑁 and 𝜈2 = 2 ⋅ (M − N + 1) , for N  0 ; in case N = 0 , then pL = 0 .
The upper boundary is calculated as
𝑝𝑈 = 𝜈1 ⋅ 𝐹𝜈1 ,𝜈2;1−0.05/2 /(𝜈2 + 𝜈1 ⋅ 𝐹𝜈1 ,𝜈2;1−0.05/2 ) with 𝜈1 = 2 ⋅ (N + 1) and
𝜈2 = 2 ⋅ (𝑀 − 𝑁), for 𝑁 < 𝑀; in case 𝑁 = 𝑀, then 𝑝𝑈 = 1. The corresponding
lower and upper boundaries of the 95%-confidence interval for the failure
rate are finally calculated as 𝑟𝐿 = − 𝑙𝑛(1 − 𝑝𝐿 ) /𝑡𝑒𝑥𝑝 and
( )
𝑟𝑈 = − 𝑙𝑛 1 − 𝑝𝑈 /𝑡𝑒𝑥𝑝 , respectively. The symbol 𝐹𝜈1 ,𝜈2;𝑥 (with 𝑥 = 0.05/2 or
x = 1 − 0.05 / 2 ) indicates the inverse cumulative F-distribution with  1 and
 2 degrees of freedom, calculated at the specified value x . The same
approach can be used in case model tests are carried out instead of
numerical simulations.

I:\CIRC\MSC\1\MSC.1-Circ.1652.docx
MSC.1/Circ.1652
Annex, page 34

3.5.4.5 Full probabilistic direct stability assessment requires evaluation of the upper boundary
rU(Hs, Tz, , vs) of the 95%-confidence interval of the “short-term" stability failure rate for each
relevant combination of significant wave height Hs, mean zero-crossing wave period Tz, ship
forward speed vs, and mean wave direction  with respect to the ship heading.
For non-vulnerable ships and loading conditions, for combinations of wave periods and
directions away from roll resonance and for moderate significant wave heights, the stability
failure rate may be very low, so that application of direct counting is impractical.
Therefore, direct counting should be used for such combinations (Hs, Tz,  vs) for which this is
affordable, otherwise statistical extrapolation can be used, section 3.5.5.

Direct stability assessment in design situations does not require statistical extrapolation, thus
the use of direct counting is sufficient. However, statistical extrapolation still can be used when
this seems practicable.

Preparation of operational guidance using only direct counting is, in principle, possible, but
may be rather time-consuming. Use of statistical extrapolation may significantly accelerate the
preparation of operational guidance: in the same way as for the full probabilistic direct stability
assessment, direct counting is used to define the upper boundary of 95%-confidence interval
of stability failure rate for such combinations (Hs, Tz,  vs) for which this is affordable, otherwise
statistical extrapolation is used.

3.5.5 Extrapolation procedures

3.5.5.3 Extrapolation over wave height

3.5.5.3.1 Extrapolation of the mean time to stability failure or mean rate of stability failures
over significant wave height significantly reduces required simulation time since numerical
simulations or model tests are conducted at greater significant wave heights than those
required in the assessment, where stability failure rate is greater, and results are extrapolated
to lower significant wave heights. This method is especially useful for the full probabilistic direct
stability assessment and probabilistic operational measures, which require evaluation of
stability failure rate for all possible combinations of sailing conditions (vs, ) and sea states
(Hs, Tz): direct counting is used to define stability failure rate for such combinations (vs, , Hs,
Tz) where this is affordable, and extrapolation over significant wave height is applied to these
results to define the stability failure rate for the remaining combinations.

3.5.5.3.2 The extrapolation can be performed either in the form

ln T = A + B / H2s
or, equivalently, in the form
ln r = A + B / H2s
since 𝑟 = 1/𝑇; here T (s) is the mean time to stability failure, r (1/s) is the stability failure rate,
and A and B are coefficients which do not depend on the significant wave height but depend
on the ship loading condition, forward speed, zero-crossing wave period and mean wave
direction. Coefficients A and B can be obtained by, for example, linear regression of ln(r), or
ln(T), with respect to 1/Hs2. Information regarding this extrapolation procedure is provided in
section 5.1 of appendix 4, which contains validation and application examples of this procedure
to parametric and synchronous roll and pure loss of stability.

3.5.5.3.3 If rk are the maximum likelihood estimates of the stability failure rate obtained by
direct counting at significant wave heights Hsk, k = 1, …, K, and Nk are the numbers of stability
failures that were encountered to estimate rk, use at least three values of rk for extrapolation,

I:\CIRC\MSC\1\MSC.1-Circ.1652.docx
 MSC.1/Circ.1652
Annex, page 35

which were obtained for a range of significant wave heights not less than 2 m. Each of these
values should not exceed 5% of the reciprocal natural roll period of the ship, i.e. rk < 0.05 / Tr
(or Tk > 20 Tr). The values used should be checked for outliers and possible non-conservative
extrapolation; if necessary, points used for extrapolation should be added or removed.
2
The extrapolated stability failure rate re is defined by linear extrapolation of ln 𝑟̂𝑘 over 1/𝐻s𝑘 as
ln re = ∑Kk=1 bk ln r̂ k
The coefficients bk can be obtained by several methods, e.g. least-squares method, see
section 5.1.4 of appendix 4. Note that in any case, ∑Kk=1 bk = 1.

Details for the calculation of the upper boundary of the 95%-confidence interval of the
extrapolated stability failure rate, reU, are provided in section 5.1.3 of appendix 4.

3.5.5.4 Other extrapolation procedures

3.5.5.4.1For more details on these other extrapolation procedures, refer to appendix 4.

.1 Theoretical background, example and statistical validation for the envelope

peaks over threshold (EPOT) extrapolation procedure is described in section
5.4 of appendix 4. An applicable failure mode includes roll in the dead ship
condition and in pure loss of stability. Applicability may be extended to the
parametric roll failure mode after adjustment is made of the de-clustering
procedure and a proper statistical validation.

.2 Theoretical background and example for the Split-time/motion perturbation

method (MPM) extrapolation procedure is described in section 5.3 of
appendix 4. The Split-time/MPM can be applied for excessive acceleration
failure mode. The metric for excessive acceleration failure mode is
formulated on condition of exceedance of the target lateral acceleration.

.3 The critical wave method for surf-riding and broaching is described in section
5.2 of appendix 4.

4 Guidelines for operational measures

4.1 General principles

4.1.1 These Guidelines consider the following operational measures, see paragraph 4.3.1
of the Interim Guidelines: operational limitations related to areas or routes and season;
operational limitations related to maximum significant wave height; and operational guidance.
The applicability of these measures for each of the five stability failure modes is explained in
table 4.1.1.

Table 4.1.1 Applicability of the operational measures for five stability failure modes
Stability failure mode Operational measures
Operational limitations Operational limitations Operational
related to areas or related to maximum guidance
routes and season significant wave height
Dead ship condition applicable not applicable not applicable
Excessive acceleration applicable applicable applicable
Pure loss of stability applicable applicable applicable
Parametric rolling applicable applicable applicable
Surf-riding/broaching applicable applicable applicable

I:\CIRC\MSC\1\MSC.1-Circ.1652.docx
MSC.1/Circ.1652
Annex, page 36

For four stability failure modes (excessive acceleration, pure loss of stability, parametric rolling
and surf-riding/broaching), any operational measure can be used. For the dead ship condition
failure mode, operational limitations related to maximum significant wave height and
operational guidance are not applicable since this stability failure mode assumes inoperable
main propulsion plant and auxiliaries. Such ships cannot avoid heavy weather and, once in
heavy weather, cannot control speed and course to follow operational recommendations.

4.1.2 The aim of operational measures is to provide the same level of safety, quantified by
the upper boundary r̄ U of the 95%-confidence interval of the average "long-term" stability
failure rate, as the level provided by the criteria, procedures and standards provided by the
guidelines for vulnerability criteria in chapter 2 of the Interim Guidelines or the direct stability
assessment in chapter 3; for

.1 operational limitations related to areas or routes and season, this means that
the ship is subject to the design assessment procedures in chapter 2 or
chapter 3 with a different wave scatter table and corresponding wind
statistics, which correspond to a specified area or route during a specified
season;

.2 operational limitations related to maximum significant wave height, this

means that the ship is subject to design assessment procedures in chapter 2
of the Interim Guidelines or chapter 3 of the Interim Guidelines with a wave
scatter table which is considered up to a specified significant wave height;
wind statistics are correspondingly modified; and

.3 operational guidance, this means that acceptable combinations of ship

forward speed and wave height, period and direction with respect to ship
heading are defined so that the level of safety, provided by operation in these
combinations, is the same as provided by the design assessment procedures
in chapter 2 of the Interim Guidelines or chapter 3 of the Interim Guidelines.

4.2 Stability failures

4.2.1 For ships with an extended low weather deck, motions in waves may lead to
accumulation of water on deck. Dynamics of water on deck may significantly contribute to
stability in waves for such ships. Since modelling of the effect of water on deck is difficult even
with numerical methods that are employed in the direct stability assessment, it is not addressed
in the present guidelines. However, if the effect of water on deck is ignored in numerical
simulations, then the results may not be sufficiently conservative for the dead ship condition
and pure loss of stability failure modes.

4.3 Operational measures

4.3.2 The amount of information, preparation and planning significantly differs between
operational measures:

.1 operational limitations related to areas or routes and season are based on

design assessment combined with wave scatter table and corresponding
wind statistics that correspond to a specified area or route during a specified
season, therefore the application of these operational limitations does not
require weather data during the operation of the ship or any specific
information and planning; however, actual data should be always available
for the loading condition to ensure that the parameters of the loading
condition are within the applicability range of the operational limitations.
An estimate of the actual draught, trim, displacement, metacentric height,

I:\CIRC\MSC\1\MSC.1-Circ.1652.docx
 MSC.1/Circ.1652
Annex, page 37

longitudinal position of the centre of gravity and natural roll period should be
available at departure from the port and during the voyage, together with the
estimated change of these parameters until the arrival in the next port;

.2 operational limitations related to maximum significant wave height are based

on design assessment combined with a wave scatter table that is considered
up to a specified significant wave height and correspondingly modified wind
statistics. Therefore, application of these operational limitations requires a
weather forecast, containing at least the significant wave height, and an
actual passage plan which accounts for the most recent weather forecast
and actual loading condition. To ensure that the ship does not encounter a
situation when no acceptable sailing conditions are available, the passage
plan should be available for the next three days to allow for enough time for
routing adjustment. An estimate of the actual draught, trim, displacement,
metacentric height, longitudinal position of the centre of gravity and natural
roll period should be available at departure from the port and during the
voyage, together with the estimated change of these parameters until the
arrival in the next port; and

.3 operational guidance identifies for each sea state acceptable and

unacceptable combinations of ship forward speed and wave direction with
respect to ship heading, therefore it requires detailed forecast information
about wave energy spectrum (which contains, at least, significant wave
height, mean zero-crossing period and mean direction of wind sea and swell)
and wind characteristics, together with means for indicating unacceptable
combinations of ship speed and heading relative to mean wave direction.
To ensure that the ship does not encounter a situation when no acceptable
sailing conditions are available, the passage plan should be available for the
next three days to allow for enough time for routing adjustment. An estimate
of the actual draught, trim, displacement, metacentric height, longitudinal
position of the centre of gravity and natural roll period should be available at
departure from the port and during the voyage, together with the estimated
change of these parameters until the arrival in the next port.

4.4 Acceptance of operational measures

4.4.3 Operational measures can reduce the stability failure rate to any low level by
application of sufficiently strict operational measures. However, if operational limitations related
to maximum significant wave height or operational guidance excludes too many sailing
conditions in too many sea states, including moderate sea states, as not acceptable for some
loading condition, such loading condition cannot be considered as sufficiently safe in routine
practical operation even when operational measures are provided. Therefore, a loading
condition cannot be considered as acceptable if the ratio of the total duration of all situations
which should be avoided to the total operational time is greater than 0.2, section 1.8 of
appendix 5. This restriction relates to operational limitations related to maximum significant
wave height and operational guidance since operational limitations related to areas or routes
and season do not impose any restrictions on sailing conditions or sea states.

4.4.5 Acceptable sailing conditions may be unattainable in some sea states due to limits of
propulsion and steering systems of the ship or undesirable due to other problems, e.g.
excessive vertical motions, accelerations and slamming. Neglecting this contradiction may
lead to misleading guidance or even put the ship in danger if in some sea state all acceptable
sailing conditions are unattainable, dangerous or unfeasible.

I:\CIRC\MSC\1\MSC.1-Circ.1652.docx
MSC.1/Circ.1652
Annex, page 38

For example, for parametric roll in bow waves, roll motions may decrease with increasing
forward speed, but high speeds could be either unattainable or could lead to excessive vertical
motions or loads. Therefore, if practicable, it is suggested for wave directions from head waves
to 60 degree off-bow to define the maximum attainable forward speed taking into account the
added resistance in seaway and the ship's propulsion system, as well as maximum suitable
forward speed from the point of view of absolute and relative motions, vertical accelerations
and slamming.

The maximum attainable and maximum suitable forward speeds in bow waves can be defined
from model tests. In cases where this is not practicable, proven conservative estimates can be
used. A conservative alternative for parametric roll in bow waves is to evaluate roll motions at
zero forward speed.

In sea states where heading into seaway is necessary to avoid excessive lateral accelerations,
the ability of the ship to keep a sufficient forward speed in head waves is required.
The sufficient forward speed to keep heading into bow waves can be defined from model tests;
in cases where this is not practicable, proven conservative estimates can be used. As a
conservative assumption, 20% of the full service speed in calm water could be considered as
a general indication.

4.5 Preparation procedures

4.5.1 Operational limitations related to areas or routes and season

4.5.1.3 Preparation of operational limitations using some level 1 and level 2 vulnerability
assessment procedures requires definition of regular wave cases based on a modified wave
scatter table. Definition of such wave cases is detailed in section 10 of appendix 3.

4.5.2 Operational limitations related to maximum significant wave height

4.5.2.3 Preparation of operational limitations using some level 1 and level 2 vulnerability
assessment procedures requires definition of regular wave cases based on a modified wave
scatter table. Such wave cases are defined in the same way as for operational limitations
related to areas or routes and season, see details in section 10 of appendix 3.

4.5.3 General principles of preparation of operational guidance

4.5.3.2 Operational guidance can be prepared using any of three equivalent approaches
which are recommended for the preparation of operational guidance. The theoretical
background and application examples are included in sections 1 and 2, respectively, of
appendix 5 of these explanatory notes.
4.5.3.3 Operational guidance should clearly indicate acceptable and unacceptable sailing
conditions for each relevant sea state. A possible form of presentation of operational guidance
is a polar diagram, see an example in figure 4.5.1 and more examples in section 2
of appendix 5.

I:\CIRC\MSC\1\MSC.1-Circ.1652.docx
 MSC.1/Circ.1652
Annex, page 39

Figure 4.5.1 – Unacceptable sailing situations (combinations of ship forward speed

and wave direction with respect to ship heading, red) for 8400 TEU container ship in
axes ship speed (knots, radial coordinate) - mean wave direction (circumferential
coordinate) for significant wave height 7.0 m.

4.5.3.4 Other forms of operational guidance, different from polar diagrams, can be used; in
any case, for presentation of operational guidance it is recommended that:

.1 operational guidance is available for all loading conditions subject to

operational guidance;

.2 acceptable and unacceptable sailing conditions (combinations of mean wave

direction and ship forward speed) are easily and clearly identifiable for each
relevant sea state;

.3 resolution of mean wave directions is at least 15 degrees and the range of

ship forward speeds extends from zero to full service speed in at least six
intervals; and

.4 sea states cover all non-zero entries of the applicable wave scatter table with
discretization intervals not exceeding 1.0 m for significant wave heights and
1.0 s for zero-crossing wave periods.

4.5.4 Probabilistic operational guidance

4.5.4.1 Operational guidance indicates unacceptable sailing conditions, i.e. combinations

(vs, ) that should be avoided, for each range of sea states (Hs, Tz) in the relevant wave scatter
table. A probabilistic operational guidance uses the upper boundary rU of the 95%-confidence
interval of the "short-term" stability failure rate rU(Hs, Tz, , vs) as the criterion to distinguish
between acceptable and unacceptable sailing conditions.

4.5.4.2 See paragraph 1.6.2 of appendix 5. Additional levels may be added for augmented
guidance; levels corresponding to lower failure rates may serve as a warning. Additional levels,
corresponding to larger failure rates, may be helpful for finding the best way of exiting the
unacceptable sailing condition, in case the guidance is used in real time.

I:\CIRC\MSC\1\MSC.1-Circ.1652.docx
MSC.1/Circ.1652
Annex, page 40

4.5.4.3 The upper boundary rU(Hs, Tz , v0) of the 95%-confidence interval of the stability
failure rate should be defined for combinations (Hs, Tz, , v0) with a resolution of at least 1.0 m
for significant wave heights, 1.0 s for zero-crossing wave periods and 15 degrees for mean
wave headings; the range of ship forward speeds from zero to full service speed should be
divided into at least six intervals. This boundary can be obtained by either direct counting,
extrapolation of rU over significant wave height or another statistical extrapolation method:

.1 direct counting, see paragraph 3.5.4.4 of the Interim Guidelines and

corresponding explanatory notes, is time-consuming when stability failure
rate is low. If direct counting procedure in paragraph 3.5.4.4.1 is used,
computational time can be reduced using acceptance check during
simulation following example procedure in paragraph 4.7.10 of appendix 4
with an acceptance threshold  = 10-6 1/s. Numerical simulations should be
carried out for each combination (Hs, Tz), , v0) in multiple independent
realizations of the sea state according to recommendations in the
explanatory text to paragraph 3.5.4.3 of the Interim Guidelines; and

.2 combination of direct counting with statistical extrapolation may significantly

accelerate the preparation of operational guidance: direct counting is used
to define the upper boundary of 95%-confidence interval of stability failure
rate rU for such combinations (Hs, Tz, v0, ) for which it is affordable, otherwise
statistical extrapolation is used.

4.5.4.4 If a certain situation is assessed as unacceptable, all situations with greater Hs and
the same vs,  and Tz are considered unacceptable, whereas if a certain situation is found
acceptable, all situations with lower hs and the same vs,  and Tz are considered acceptable.

4.5.6 Simplified operational guidance

4.5.6.2 Regarding items .3 and .4:

.3 caution should be exercised, as a mathematical model of roll motions with 1

degree of freedom may not be always conservative in prediction of
parametric roll occurrences for all speed and headings. (See sections 1.7
and 2.6 of appendix 5); and

.4 refer to section 2.5 of appendix 5.

4.6 Application

4.6.3 For seaway consisting of two wave systems (e.g. wind sea and swell) with significant
wave heights hs1 and hs2, zero-crossing wave periods Tz1 and Tz2 and mean wave directions 1
and 2, the ship responses can be approximated by theoretical modelling of wave systems and
overlapping their effects, section 1.3 of appendix 5.

Ship responses to these two wave systems are defined separately, using pre-computed
databases of ship responses to wave systems with parameters (Hs1, Tz1, 1) and (Hs2, Tz2, 2)
with theoretical wave energy spectra: upper boundary rU1 and rU2 of the 95%-confidence
interval of the "short-term" stability failure rate when using probabilistic operational guidance;
mean three-hour maximum roll 3h1 and 3h2 or lateral acceleration ay3h1 and ay3h2 amplitude
when using deterministic operational guidance; and unacceptable sailing conditions (vs, ) for
each of these wave systems when using simplified operational guidance.

I:\CIRC\MSC\1\MSC.1-Circ.1652.docx
 MSC.1/Circ.1652
Annex, page 41

Unacceptable sailing conditions (vs, ) are defined for the combination of these wave systems
as those for which rU1 + rU2 > 10-6 s-1 for probabilistic operational guidance;  (3h1 + 3h2) > xlim or
 (ay3h1 + ay3h2) > xlim for deterministic operational guidance ( = 2 is the scaling factor and xlim is
the corresponding stability failure threshold, paragraph 3.2.1 of the Interim Guidelines for direct
stability assessment); and sailing conditions which are unacceptable in any of these two wave
systems for simplified operational guidance.

4.6.4 The master of a vessel operating under operational limitations related to maximum
significant wave height or operational guidance should always have an actual passage plan
which accounts for the most recent weather forecast and actual loading condition. To ensure
that the ship does not encounter a situation when all acceptable sailing conditions are
unattainable or unfeasible, the passage plan should be always available for the next three days
to allow for enough time to avoid a storm. As accurate as practicable data should be available
for the loading condition at departure from the port, including draught, trim, displacement,
metacentric height, longitudinal position of the centre of gravity, and natural roll period.
An estimate of the current state of these parameters should be always available during the
voyage together with the estimated change of these parameters for the time until the arrival
in the next port.

I:\CIRC\MSC\1\MSC.1-Circ.1652.docx
MSC.1/Circ.1652
Annex, page 42

APPENDIX 1

PHYSICAL DESCRIPTION OF THE STABILITY FAILURE MODES ADDRESSED BY

SECOND GENERATION INTACT STABILITY CRITERIA

1 Physical background of stability failure related to the dead ship condition

1.1 Modelling in the level 1 criterion for the dead ship condition

1.1.1 Dead ship condition was the first mode of stability failure addressed by the
physics-based severe wind-and-roll criterion, also known as the "weather criterion", which was
adopted by IMO in 1985 (resolution A.562(14)) and is now embodied in section 2.3 of part A
of the 2008 Intact Stability Code. The scenario of the weather criterion is shown in figure 1.1.
This scenario assumes that a ship has lost its power and has turned into beam seas, where it
is rolling under the action of waves as well as heeling and drifting under the action of wind.
Drift-related heel is a result of action of a pair of forces: wind aerodynamic force and
hydrodynamic reaction caused by transverse motion of the ship.

A ship lost its power and has

Wind turned into beam seas; she is
rolling in waves and drifts
under the wind
Drift Drift reaction

Wind gust A sudden (and long) wind

gust has occurred when
the ship rolled windward
Drift Drift reaction

The ship starts to roll back

Wind gust under combined wave and
wind action. Velocity of drift
and drift reaction start to
Drift Drift reaction increase

Wind gust The ship continues to roll

leeward while drift velocity and
drift reaction continue to
Drift increase, providing additional
Drift reaction heeling moment

Wind gust The ship has reached a

maximum roll angle on
the leeward side
Drift
Drift reaction
This is most likely the
instant for stability failure

Figure 1.1 Scenario of stability failure in dead ship conditions

I:CIRC\MSC\1\MSC.1-Circ.1652.docx
 MSC.1/Circ.1652
Annex, page 43

1.1.2 Next, a sudden and long gust of wind occurs. The worst possible instant for this is
when the ship is rolled at the maximum windward angle; in this case, action of wind is added
to the action of waves. The strengthening wind increases drift velocity, and this leads to an
increase of the hydrodynamic drift reaction. The increase of the drift velocity leads to
the increase of the hydrodynamic reaction and, therefore, to the increase of the heeling
moment by the pair of aerodynamic and hydrodynamic forces. The gust is assumed to last
long enough so the ship can roll to the other side completely; the achieved leeward roll angle
is the base of the criterion. If it is too large, or some openings may be flooded, the stability of
the ship is considered insufficient.

2 Physical background of stability failure related to excessive accelerations

2.1 Accelerations caused by ship motions

2.1.1 When a ship is rolling, the objects in higher locations travel longer distances. A period
of roll motions is the same for all the locations on board the ship. To cover longer distance
during the same time, the linear velocity must be larger. As the velocity changes its direction
every half a period, larger linear velocity leads to larger linear accelerations. Large linear
acceleration means larger inertial force (see figure 2.1).

2.1.2 Inertial forces acting in a horizontal plane are more dangerous for a human than
vertical inertial forces. The vertical inertia forces cause brief overloading, while horizontal
inertial forces cause humans to lose their balance, fall or even be thrown against walls,
bulkheads or and other structures. Large accelerations are mostly caused by roll motions so
they have predominantly lateral direction.

2.1.3 If the GM value is larger, the period of roll motion is smaller. Thus, for the same roll
amplitude the changes of linear velocity occur faster, so accelerations are larger.

Inertial forces at
Longer the location
distance
traveled
Large
Shorter
distance
traveled Small

Figure 2.1 Scenario of stability failure related to excessive accelerations

2.2 Synchronous resonance in ship motions

2.2.1 A large angle of roll may be caused by different physical mechanisms. Some of them
are already included as a part of vulnerability assessment of the second generation of IMO
stability criteria: pure loss of stability, parametric rolling and broaching. Among these
phenomena, parametric rolling is known to cause excessive accelerations.
However, synchronous resonance is not covered by other vulnerability criteria.

2.2.2 Synchronous resonance is a phenomenon of amplification of motion response when

the natural frequency of the ship motion is close to the frequency of the wave excitation.

I:\CIRC\MSC\1\MSC.1-Circ.1652.docx
MSC.1/Circ.1652
Annex, page 44

2.2.3 The frequency of wave excitation depends on wave frequency, ship heading relative
to waves and ship speed. When a ship sails against the waves (between head and beam wave
encounter angles) the frequency of encounter is higher than the frequency of waves. This
effect is the strongest in head waves, weakens in bow quartering seas and completely
disappears in beam seas. When a ship sails in the same direction as the waves, the frequency
of encounter decreases. This effect is the strongest in following seas, weakens in stern
quartering seas and completely disappears in beam seas. Higher speed increases this effect.

2.2.4 The motion amplification effect of the synchronous resonance is the strongest when
the encounter frequency is close to natural roll frequency (see figure 2.2). An increase of
the amplitude of excitation (angle of wave slope) leads to an increase of resonance effect at
all frequencies; however, the strongest increase is around the natural roll frequency
(see figure 2.2a).

2.2.5 An increase of roll damping leads to decrease of motion amplitude; the effect is
noticeable around natural frequency (see figure 2.2b). Thus, an increase of roll damping helps
to mitigate the effects of the synchronous resonance.

a) b)

30 Roll resonance for three different 30 Roll resonance for three different
values of wave steepness values of roll damping
Roll amplitude, deg

20 20

10 10

0 0.2 0.4 0.6 0.8 1 1.2 1.4 0 0.2 0.4 0.6 0.8 1 1.2 1.4
Encounter frequency, rad/s Encounter frequency, rad/s

Figure 2.2 Synchronous roll resonance: a) influence of wave slope and

b) roll damping

3 Physical background and scenario of pure loss of stability

3.1 Righting lever variation in waves

3.1.1 When a ship is under way through longitudinal waves, the submerged part of the hull
changes. These changes may become significant if the length of the wave is comparable to
the length of the ship. As a first example, one may observe the changes that occur when
the trough of a wave is located amidships (see figure 3.1). For many ships, the upper part of
the bow section is usually wide, due to bow flare. Bow flare provides protection from spray and
green water shipping, and provides opportunity for deck cargo stowage. The bow flare makes
the waterplane larger, if the upper part of the bow section becomes partially submerged.
The upper part of the aft section of the hull may be wider. Cargo stowage considerations often
mandate wide afterbodies. Therefore, the after part of the waterplane also increases, once
the upper part of the aft section becomes submerged. Unlike the bow and aft sections,
the midship section of most ships is often nearly wall-sided (see figure 3.1a). These
characteristics mean that very little change occurs in the waterplane width with variations in
draught. When the wave trough is amidships, the draught at the midship section is low, but as
the hull is wall-sided in this region, there is little waterplane change. As a result, when the wave
trough is located around the midship section, the overall waterplane area is increased
(see figure 3.1b).

I:\CIRC\MSC\1\MSC.1-Circ.1652.docx
 MSC.1/Circ.1652
Annex, page 45

a)

Wave trough amidships

b)

Calm water

Figure 3.1 Changes in hull geometry when a wave trough is amidships:

a) 3D view and b) waterplane

3.1.2 When the wave crest is located near amidships, the situation changes dramatically
(figure 3.2). The underwater part of the bow section is usually quite narrow, especially around
the waterline. Even for a bulbous bow, it is still narrower than for the section with bow flare.
The reason for this is the consideration of resistance. The faster the ship is, the finer its
underwater bow section tends to be. If the wave crest is amidships and the wave has a length
similar to a ship length, the wave trough is located around the bow section. This makes the
draught at the bow quite shallow. As a result, the waterplane becomes very narrow in this
region. The underwater part of the aft section is also very narrow. The main design
consideration is to provide the propulsor with enough water to efficiently power the ship.
Consideration of energy efficiency impels a designer towards a buttock flow stern design.
When the wave crest is located amidships, another wave trough is located near the aft section.
The draught at the stern becomes smaller, which tends to make the waterplane very narrow in
the aft part (see figure 3.2a). As mentioned previously, the midship section is typically more
wall-sided, so it does not significantly affect the waterplane. Figure 3.2b shows the effect of
the wave crest amidships, where the overall waterplane is reduced in area.

a)
b)

Calm water
Wave crest amidships

Figure 3.2 Changes in hull geometry when a wave crest is amidships:

a) 3D view and b) waterplane

3.1.3. The waterplane area has a significant effect on ship stability, which is well known from
ship hydrostatics. If the waterplane area is reduced, then the righting lever (GZ) curve is
reduced as well (see figure 3.3). The change of stability in waves, as examined above, is
the physical basis for the stability failure mode known as pure loss of stability. The dynamics
of pure loss of stability are different from that of parametric rolling, but are also closely related
to the severity and duration of waterplane changes. A possible scenario for the development
of a stability failure caused by pure loss of stability is shown in figure 3.4.

I:\CIRC\MSC\1\MSC.1-Circ.1652.docx
MSC.1/Circ.1652
Annex, page 46

Wave trough amidships

GZ

Calm water

Calm water
GZ

Calm water
Wave crest amidships

Figure 3.3 Stability, represented by righting lever curves, corresponding to waterplane

changes with the midship located on the wave trough (top)
and the wave crest (bottom)
Typical changes of
Ship is under way in following waves. A large longitudinal
stability caused by
wave is approaching the ship from the stern
relatively small waves

The large wave is overtaking the ship. If the time of Large decrease of the
exposure to the crest of the large wave is sufficiently long, instantaneous GZ
a stability failure may occur curve, caused by the
crest of a large wave

Typical changes of
The large wave has passed the ship. The stability caused by
ship has regained its stability relatively small waves

Figure 3.4 A possible scenario for the development of pure loss of stability

3.1.4 The case diagrammed in figure 3.4 shows a large wave approaching the ship from
the stern while the ship is under way with relatively high speed in following seas. If the celerity
(speed) of the large wave is just slightly above the ship speed, the duration needed for the large
wave to pass or overtake the ship may be long ("long" here means at least an order of
magnitude greater than the natural roll period). Once the crest of the large wave is near
the midship section of the ship, the righting lever may be significantly decreased. Further,
because of the significant duration that this condition may exist, a large heel angle may develop,
which could lead to capsize. Once the large wave overtakes the ship, the righting lever is
restored and the ship will return to the upright position, if a significant heel did not already occur.

I:\CIRC\MSC\1\MSC.1-Circ.1652.docx
 MSC.1/Circ.1652
Annex, page 47

3.1.5 A significantly reduced righting lever curve due to a relative wave profile does not
necessarily result in a total or partial stability failure. An external heel moment is required for
stability failure to occur. If no external heel moment exists, the upright condition will be retained
except for cases in which the metacentric height in waves becomes negative, which allows an
angle of loll. As the external moments to be relevant, wave exciting roll moment due to oblique
wave heading and heel moment induced by a hydrodynamic force due to ship manoeuvring
motions are candidates. Several existing model experiments using freely running ship models
in stern-quartering waves indicate that coupling with manoeuvring motion is essential to
explaining the forward speed effect on the stability failures of pure loss of stability. An example
is shown in figure 3.5. If we take account of only surge-roll coupled motion, the pure loss of
stability does not drastically depend on the forward speed. If we take account of also coupling
with manoeuvring motions, however, the drastic forward speed effect on pure loss of stability
can be explained. Thus, the heel moment induced by the hydrodynamic force due to ship
manoeuvring motions should be included.

35
Maximun roll angle(degrees)

30

25

20
Exp 4DOF 2DOF
15

10

0
0. 25 0. 3 0. 35 0. 4
Froude number

Figure 3.5 Forward speed effect on pure loss of stability in irregular

stern-quartering waves for a containership model in a model basin
(here, 2 DOF and 4 DOF mean surge-roll and surge-sway-yaw-roll
coupled simulation models, respectively) 4

4 Physical background of parametric rolling

4.1 Development of parametric rolling

4.1.1 Parametric rolling (a shortening of the formal term "parametric roll resonance") is
a dynamic stability phenomenon in which an amplification of roll motion is caused by periodic
variation of transverse stability in waves. The phenomenon of parametric rolling is
predominantly observed in head, following, bow and stern-quartering seas when the ship's
encounter frequency is approximately twice that of the ship’s roll natural frequency and
the ship's roll damping is insufficient to dissipate additional energy (accumulated because of
parametric resonance).

4.1.2 Figure 4.1 shows the process by which parametric rolling develops. If the ship rolls
while in the wave trough, increased stability (i.e. righting lever) provides stronger restoring, or
restoring moment. As the ship returns to the upright position, its roll motion rate is increased,
since there was an additional restoring from the increased stability. If at that time, however,
the ship has the wave crest at midship, the stability is decreased and the ship will roll further
to the opposite side because of the greater roll motion rate and less resistance to heeling.
Then, if the wave trough reaches the midship section when the ship reaches its maximum

4
Kubo. H., Umeda. N., Yamane. K., Matsuda. A., 2012, Pure Loss of Stability in Astern Seas: Is it Really
Pure?, Proceedings of the 6th Asia Pacific Workshop on Marine Hydrodynamics, Johor, pp. 307-312.

I:\CIRC\MSC\1\MSC.1-Circ.1652.docx
MSC.1/Circ.1652
Annex, page 48

amplitude roll, stability increases again and the cycle starts again. Note that there was one half
of the roll cycle associated with the passing of an entire wave. So, there are two waves that
pass during each roll period. That means the roll period is generally equivalent to twice that of
the wave period, as diagrammed in figure 4.2.

Stability is Stability is increased

Strong restoring
increased Stability is decreased, again; strong restoring
increases rolling
with strong ship rolls further cycle is repeated
(angular velocity)
restoring

Figure 4.1 Development of parametric roll resonance

Roll, deg Phase of wave

20

0 5 10 15 20 25 30

-20
Wave 1 Wave 2 Wave 3 Wave 4

-40
Roll period 1 Roll period 2

Figure 4.2 Time histories plots of parametric roll resonance

I:\CIRC\MSC\1\MSC.1-Circ.1652.docx
 MSC.1/Circ.1652
Annex, page 49

4.2 Frequency characteristics of parametric rolling

4.2.1 Parametric rolling is a resonance phenomenon and, similar to roll resonance in beam
waves (see figure 4.3a), parametric rolling has a limited frequency range (see figure 4.3b). The
principal difference between the two phenomena is that the span of the frequency range for
parametric rolling depends on the magnitude of stability change, while the frequency range for
roll resonance depends on wave height (see figure 4.3c). Also, if the beam waves are far from
the resonance frequency, the ship only rolls with very small amplitude. Parametric rolling does
not exist (the amplitude is equal to zero) outside of the frequency range.

a) b)
30 Roll resonance in beam seas
for a wave of three different 30 Parametric roll
Roll amplitude, deg

Roll amplitude, deg

values of wave steepness resonance for wave of
three different heights
20 20

10 10
Wave
Wave
frequency, 1/s
frequency 1/s
0 0.2 0.4 0.6 0.8 1 1.2 1.4 0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6
c)
stability change

Wave height 3
Magnitude of

Wave height 2

Wave height 1 Wave

frequency 1/s
0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6

Figure 4.3 a) roll resonance in beam seas; b) parametric roll resonance;

c) frequency range of parametric roll resonance

4.3 Influence of roll damping

4.3.1 When a ship rolls in calm water after being disturbed, the roll amplitudes decrease
successively due to roll damping (see figure 4.4). A rolling ship generates waves and eddies,
and experiences frictional drag. All of these processes contribute to roll damping. Roll damping
may play a critical role in the development of parametric roll resonance. If the "loss" of energy
per cycle caused by damping is more than the energy "gain" caused by the changing stability
in longitudinal seas, the roll angles will not increase and the parametric resonance will not
develop. Once the energy "gain" per cycle is more than the energy "loss" due to damping, the
amplitude of the parametric rolling starts to grow.

4.3.2 There is then a roll damping threshold for parametric roll resonance. If the roll damping
moment is higher than the threshold, then parametric roll resonance is not possible. If the roll
damping moment is below the threshold, then the parametric roll resonance can take place.
During the parametric roll resonance, the combination of harder restoring due to the increased
stability on the wave trough and larger achieved roll angles due to the decreased stability on
the wave crest, which occur generally equivalent to twice during the roll period, makes the roll
angle grow significantly. The only other condition that has to be met is that the energy loss due
to roll damping is not large enough to completely consume the increase of energy caused by
parametric roll resonance – the roll damping is below the threshold value.

I:\CIRC\MSC\1\MSC.1-Circ.1652.docx
MSC.1/Circ.1652
Annex, page 50

Roll Decreasing roll amplitude

during one roll period
Roll disturbance

time

Figure 4.4 Successively decreasing roll amplitudes due to roll damping in calm water

4.4 Influence of speed and wave direction

4.4.1 The frequency of encounter with waves changes when a ship is in motion. When
a ship is sailing in following or stern-quartering seas, the direction of waves and the ship
heading are similar (see figure 4.5a). As a result, the relative speed is small and a ship
encounters fewer waves during the same time period (compared to a zero-speed case).
The encounter period is increased (and the encounter frequency is decreased) in following or
stern-quartering waves.

4.4.2 When a ship is sailing in head or bow-quartering seas, the direction of waves and
the ship heading are opposite (see figure 4.5b). As a result, the relative speed is large and
a ship encounters more waves during the same time (compared with the zero-speed case).
The encounter period is decreased (and the encounter frequency is increased) in head or
bow-quartering waves.

4.4.3 The inception of parametric rolling depends on the frequency of encounter being in
the frequency range where the parametric rolling is possible (see figure 4.3c). Therefore,
the development of parametric rolling depends on speed and heading.

a)

b)

Figure 4.5 Influence of speed and wave direction on parametric resonance:

a) following and stern-quartering seas: the encounter period is longer than
the wave period; and b) head and bow-quartering seas: the encounter
period is shorter than the wave period

I:\CIRC\MSC\1\MSC.1-Circ.1652.docx
 MSC.1/Circ.1652
Annex, page 51

5 Physical background of surf-riding and broaching

5.1 General description of surf-riding/broaching failure mode

5.1.1 Broaching (a shortening of "broaching-to") is a violent uncontrollable turn that occurs

despite maximum steering efforts to maintain course. As with any other sharp turn event,
broaching is accompanied by a large heel angle, which has the potential effect of a partial or
total stability failure. Broaching is usually preceded by surf-riding, which occurs when a wave,
approaching from the stern, "captures" a ship and accelerates the ship to the speed of
the wave (i.e. the wave celerity). Surf-riding is a single wave event in which the wave profile
does not vary relative to the ship. Because most ships are directionally unstable in
the surf-riding condition, this manoeuvring yaw instability could lead to an uncontrollable turn –
termed "broaching."

5.1.2 Because surf-riding usually precedes broaching, the likelihood of surf-riding

occurrence can be used to formulate vulnerability criteria for broaching. In order for surf-riding
to occur, several conditions need to be satisfied:

.1 the wavelength should be comparable to the ship length or larger;

.2 the wave should be sufficiently steep to produce sufficient wave surfing force;
and

.3 the ship speed and course should be comparable to the wave celerity and
direction, respectively.

5.1.3 When a ship proceeds in following waves, three main forces act in the axial direction.
Thrust is the force produced by the ship's propulsor to propel the ship forward. Resistance
(or drag) is the force that opposes the forward ship motion. The surging wave force is the force
imparted by a wave to either push the ship forward or back depending on whether the ship is
on the face or back of a wave, respectively. These forces are represented in figure 5.1.

Thrust Resistance

Forward ship motion

Wave Surging Force

Figure 5.1 Forces acting on a ship in following waves

5.1.4 When a surging wave force is present, three conditions are possible in periodic waves:

.1 Surging motion. This condition occurs when the wave surge force is
insufficient to overcome the difference between the thrust of the ship's
propulsor and the resistance of the ship at the wave celerity. In this case,
the ship oscillates from increasing speed when on the front side of the wave
to decreasing speed when on the back side of the wave – an oscillatory
motion.

I:\CIRC\MSC\1\MSC.1-Circ.1652.docx
MSC.1/Circ.1652
Annex, page 52

5.1.5 The other two conditions involve the two ship speed thresholds that can cause
surf-riding and that are directly related to the thrust the ship's propulsor delivers to maintain
a given speed.

.1 Surf-riding under certain initial condition (first threshold of surf-riding). This is

the situation that the ship shall run with the wave phase speed only for certain
combinations of initial ship speed and position along the wave, and in
particular when the ship speed is sufficiently close to the wave celerity. Under
this condition, the forward surge force of the wave at a particular point on the
wave exceeds the difference between the thrust of the ship's propulsor and
the resistance of the ship at the wave celerity. In this case, surf-riding could
occur if the ship is sufficiently accelerated by an instantaneous external force.

.2 Surf-riding under any initial condition (second threshold of surf-riding). This

is the situation that the ship is forced to run with the wave irrespective of the
initial ship speed. Under this condition, the ship cannot be overtaken by
waves so that the periodic surging motion beyond one wavelength cannot
exist.

5.1.6 To explain these three conditions more fully and since surf-riding occurs when
the ship speed is equal to the wave celerity, locating the position of reference on the wave
crest allows a convenient way to understand surf-riding. In this view, when the ship surf-rides,
it appears to remain stationary because the reference position moves with the wave.

5.1.7 In the case of surging motion, the thrust delivered by the ship's propulsor is not
sufficient to propel the ship to a speed equal to the wave celerity in calm water, which is
depicted in figure 5.2a. Figure 5.2b shows the difference between the thrust and resistance in
calm water for the ship located at different positions on a wave; this difference is negative when
the resistance is greater than the thrust. Because there is no position on the wave in which
the thrust – resistance difference is fully compensated by wave force, the only motion occurring
is surging forward and backward depending on the ship's position on the wave.

a) b)
Force Force Wave Force (with
opposite sign)
The difference between
thrust and resistance at
wave celerity
Wave Phase
G, m
0 100 200
Self-propulsion
point

Set speed Wave Difference between

celerity thrust and resistance
Figure 5.2 a) resistance and propulsion showing self-propulsion point and
thrust-resistance difference, and b) wave forces and balance between
thrust and resistance shown for different ship positions on a wave

5.1.8 The mechanics of surging can be illustrated using the curves of thrust and resistance
as shown in figure 5.3. When the ship is on the back side of the wave, the surging force pushes
the ship backwards, which causes the instantaneous speed to decrease and the resistance to
become less than the thrust. This difference is directed forward, against the surging force.
When the ship is on the face of the wave, the surging force pushes the ship forward causing
the instantaneous speed to increase and the resistance to exceed the thrust. As the wave
passes the ship, these two conditions recur.

I:\CIRC\MSC\1\MSC.1-Circ.1652.docx
 MSC.1/Circ.1652
Annex, page 53

R<T R>T

Backward Forward
Self-propulsion point

Figure 5.3 Small surging motions around self-propulsion point

5.2 Description of surf-riding equilibrium

5.2.1 The value of the wave force depends on the location of the ship on the wave as well
as the wave height and wavelength. The face of the wave pushes a ship forward – hence,
the forward wave or surge force; while the back slope does the opposite. Indeed, there are
neutral points near the wave crest and wave trough. If the wave has appropriate length and
height, the surge force is sufficient to offset the negative difference between the thrust and
resistance. This creates two points of equilibrium as shown in figure 5.4. This figure
superimposes the surge force with the difference between thrust and resistance (the horizontal
line below the abscissa) and shows the intersections with the wave force curve to mark the two
points of equilibrium on the wave. The stable equilibrium marks the first threshold of surf-riding.

Force Positions on the wave

Wave Force (with
opposite sign)

Wave Phase  , m
G
0 100 200
Difference between
thrust and resistance
Unstable
Equilibrium Stable
Equilibrium

Figure 5.4 Wave forces and balance between thrust and resistance for different
positions of a ship on a wave showing the first threshold of surf-riding

5.3 Stability of surf-riding equilibrium

5.3.1 Figure 5.5 provides an example of the two points of equilibrium referred to as stable
and unstable. If a ship is considered to be surf-riding in which midship is located about 70 m
forward of the wave crest (marked as stable equilibrium near wave trough in figure 5.4 and
figure 5.5), the ship speed will be equal to the wave celerity. If the ship is disturbed from this
location forward and toward the wave trough, the surge force decreases. Therefore,
the difference between thrust and resistance will cause a decrease in the instantaneous ship
speed and the wave will start to overtake the ship. As the ship moves back on the wave face
toward the wave crest, the wave surge force increases and pushes the ship back to the stable
equilibrium.

I:\CIRC\MSC\1\MSC.1-Circ.1652.docx
MSC.1/Circ.1652
Annex, page 54

F, kN
1000 Resistance

G, m Thrust Wave force

0 100 200

Stable
Equilibrium

F, kN Resistance
Sum
1000
Thrust Wave force
G, m
0 100 200

Disturbed
position

Figure 5.5 A disturbance forward from the stable equilibrium

5.3.2 Conversely to the case shown by figure 5.5, figure 5.6 considers the ship to be
disturbed from the equilibrium backwards and towards the wave crest. In this case, the wave
force becomes larger than the difference between thrust and resistance. Thus, the ship speed
will increase and move on the wave forward to the surf-riding equilibrium (trough). Therefore,
in either case (i.e. a disturbance forward or backward), the ship will tend to move toward
the equilibrium near the wave trough, which makes this equilibrium stable.

F, kN Resistance
1000
Thrust Wave force
G, m
0 100 200

Stable
Equilibrium
Sum Resistance
F, kN
1000
Thrust Wave force
G, m
0 100 200

Disturbed
position

Figure 5.6 A disturbance backwards from the stable equilibrium

5.3.3 If a ship is now considered to be surf-riding with the midship located about 30 m
forward of the wave crest (marked as unstable equilibria near wave crest in figure 5.4), the ship
speed will be equal to the wave celerity. If the ship is disturbed from this location forward
(towards the wave trough as shown in figure 5.7), the wave force increases and will cause
the ship speed to increase and move the ship further forward on the wave until it arrives at
the stable equilibrium near the wave trough.

I:\CIRC\MSC\1\MSC.1-Circ.1652.docx
 MSC.1/Circ.1652
Annex, page 55

F, kN Resistance
1000
Thrust Wave force
100 G, m
200

Unstable
Equilibrium

F, kN Resistance
1000 Sum
Thrust Wave force
100 G, m
200

Disturbed
position

Figure 5.7 Disturbance forward from the unstable equilibrium

5.3.4 Conversely, if the ship is disturbed from this location backward, towards the wave
crest as shown in figure 5.8, the wave force decreases and the instantaneous speed also starts
to decrease. In this case, difference between thrust and resistance will cause a decrease in
the instantaneous ship speed which causes the wave to start to overtake the ship. There are
several scenarios that consider what may happen next, but in no case does the ship return
back to this equilibrium, which makes the equilibrium near the wave crest unstable.

5.3.5 If there is no surf-riding equilibrium, surf-riding is not possible and the ship will simply
surge. That means that all the combinations of instantaneous speed and position on the wave
lead to the same outcome. However, once points of equilibrium appear at certain positions on
the wave, not all the combinations of the wave position and instantaneous speed lead to
the same response. If a ship is "placed" exactly at the location of the stable equilibrium near
the wave trough and accelerated to the wave celerity, the ship will surf-ride. Any small
disturbance from this position will return the ship back to equilibrium. If a ship is placed at
the unstable equilibrium near the wave crest, accelerated to the wave celerity and then
disturbed towards the wave trough, it will end up at the stable surf-riding equilibrium as well.
Thus, there is a set of combinations of wave positions and instantaneous speeds that will lead
to surf-riding. One can say that these combinations form a "domain of attraction to surf-riding
equilibrium." Outside of this domain, two options are possible: surging or surf-riding. So, in
principle, once outside of the attraction domain, the ship either continues to surge or is
attracted to surf-riding equilibrium on some other wave.

I:\CIRC\MSC\1\MSC.1-Circ.1652.docx
MSC.1/Circ.1652
Annex, page 56

F, kN 1000 Resistance

100 G, m Thrust Wave force

200

Unstable
Equilibrium

Sum Resistance
F, kN
1000
Thrust Wave force
100 G, m
200
Disturbed
position

Figure 5.8 Disturbance backward from the unstable equilibrium

5.4 Transition from surging motion to surf-riding

5.4.1 When the energy/work balance of the wave surging force and the difference between
thrust and resistance is considered, the latter disperses the kinetic energy obtained from
the wave. When these two works are balanced, the ship's response is surging motion.
However, if a wave provides the ship with more kinetic energy than the difference between
thrust and resistance can disperse, this excessive kinetic energy eventually leads to
acceleration and to attraction to the surf-riding equilibrium. The surf-riding becomes a new
energy balance between the works of wave surging force and the difference between thrust
and resistance. The ship surf-rides because of the excessive kinetic energy imparted to
the ship.

5.4.2 The face of the wave provides more chances for surf-riding because the wave surging
force is directed forward. If the ship is on the back side of the wave, the wave surging force is
directed backward but a surging energy balance still may occur because both surging and
surf-riding may co-exist for the same speed setting and wave parameters. If the initial kinetic
energy level can be dispersed by the difference between thrust and resistance, surging will
occur; if not, surf-riding will occur. If the wave parameters are such that the wave adds too
much kinetic energy (steep waves) to ship motions that it cannot be dispersed by the difference
between thrust and resistance, then surging motions are no longer possible. Even when
the ship starts with low initial kinetic energy level on the back slope of the wave with the ship's
propulsor delivering a set thrust, each sequential wave will add a bit of kinetic energy that
cannot be dispersed; then inevitably surf-riding will occur as the ship moves towards stable
equilibrium. This is referred to as the "surf-riding under any initial condition" of surf-riding which
is the basis used for the surf-riding vulnerability criteria. The "surf-riding under certain initial
condition" is not used for these criteria.

5.4.3 For a particular wave, the thrust at the surf-riding under any initial condition identifies
the critical setting of the ship's propulsor in the vulnerability criterion for which surf-riding
becomes inevitable. For example, consider energy balance during the time that one wave
overtakes the ship. The Melnikov analysis or the systematic phase plane analysis can be used
to identify the surf-riding under any initial condition. The level 2 vulnerability criterion directly
uses the Melnikov analysis for many possible combinations of wave height and wavelength.

I:\CIRC\MSC\1\MSC.1-Circ.1652.docx
 MSC.1/Circ.1652
Annex, page 57

The level 1 vulnerability criteria are empirical estimates based on many calculated results of
such analysis assuming a wave steepness of 0.1, which is widely accepted as the practical
limit of stable gravity deep water waves.

I:\CIRC\MSC\1\MSC.1-Circ.1652.docx
MSC.1/Circ.1652
Annex, page 58

APPENDIX 2

EXAMPLES OF ASSESSMENTS USING VULNERABILITY CRITERIA ACCORDING TO

THE SECOND GENERATION INTACT STABILITY CRITERIA

1 Example input data set

1.1 The data for a ship's loading condition that is needed to complete the assessments
contained in the Second generation intact stability criteria varies and depends upon which
vulnerability criteria and which levels of each are to be assessed. Table 1.1 below indicates
what data for the ship and the respective loading condition is needed for the particular
vulnerability assessment to be performed.

Table 1.1 – Ship and loading condition data needed for vulnerability assessments
associated with each stability failure mode and level that is additional to that needed
to demonstrate compliance with section 2.2 of part A of the 2008 Intact Stability Code

Stability Failure Mode Pure Loss of

Acceleration
Surf-riding/
Parametric
Dead Ship

Excessive
Broaching
Condition

Stability

Rolling
Level

Level

Level

Level

Level

Level

Level

Level

Level

Level
1

2
Ship data:
Hull form (offsets) X X X X X X
Hydrostatic data X X X
Bilge keel data X X X X X X
Hull and Cargo forms above
X X
waterline

Loading condition data:

Righting lever curve, calm
X X
water
Righting lever curve, in
X X
waves
Roll radius of gyration X X X
Vertical centre of gravity X X X X X X X X

Propulsion data:
Operational speed X X
Calm-water resistance X
Propulsor thrust X

Specific examples of input values for and outcomes of assessments using vulnerability criteria
for both levels 1 and 2 and for each stability failure mode is presented in sections 2 through 6.

I:\SDC\8\SDC 8-18-Add.2.docx
 MSC.1/Circ.1652
Annex, page 59

1.2 Input Data

One subject ship is the C11 class post-Panamax containership. The

input was based on the data made available to the Intact Stability
Correspondence Group (ISCG) presented by ITTC and available at
(http://www.naoe.eng.osaka-u.ac.jp/imo/ssdp1.htm). The hull form is in
figure 1.1, while principal dimensions, basic hydrostatic table, and
relevant input parameters are in table 1.2. Note that the KG value
adjusted to account for the free-surface correction, which results in an
upright GM = 1.40 m. The GZ curve is shown in figure 1.2.

30

25

20

15

10

0
-20 -15 -10 -5 0 5 10 15 20

Figure 1.1 Hull form of C11 class containership

1.5 GZ, m

0.5

Heel angle, deg.

10 20 30 40 50 60 70

-0.5

Figure 1.2 GZ curve computed for a KG corresponding to a GM = 1.4 m

I:\CIRC\MSC\1\MSC.1-Circ.1652.docx
MSC.1/Circ.1652
Annex, page 60

Table 1.2 Principal dimensions, basic hydrostatic data and other relevant input
parameters
Length, bp, m 262.00 Bilge keel length ratio (lBK/Lbp) 0.2921
Beam, m 40.00 Bilge keel height ratio (hBK/B) 0.010
Draught amidships, m 11.50 Down flooding angle, ° 50
Trim, ° 0.0 Lateral windage area, m 2 7,887
KG, m 18.976 Height of centroid above WL, m 14.73
Volumetric displacement, m3 67,504 Wind heeling moment coefficient 1.17
Block coefficient 0.56 Natural roll period, s 30.35
Midship section coefficient 0.959 Natural frequency, s-1 0.21
GM, m 1.40 Number of propellers 1
Diameter of propeller, m 8.4 Propeller expanded area ratio 0.590
Propeller pitch ratio 0.743 Propeller number of blades 6
Location of excessive acceleration x = -50 m y = 20 m z = 40 m
assessment

2 Example of assessment of ship vulnerability to the dead ship condition failure

mode

2.1 The subject ship used here is the C11 class containership.

2.2 The subject ship used in this section (section 2) is slightly different from the one in the
sections 4 and 5. There are two differences: First, the natural roll frequency is normally
computed by approximation formula from the 2008 Intact Stability Code. In this section, direct
calculation was used. As a result, the period of roll used here is 30 s (table 1.2) while 23 s was
used in the 2008 Intact Stability Code. The second difference is the GM value adjustment
priority.5

2.3 A level 1 vulnerability assessment in dead ship condition failure mode is described in
section 2.2.2 of the Interim Guidelines. This assessment is essentially the weather criterion
with the table of roll periods extended to 30 s. With this change in mind, one gets the following
results:

• Area a = 0.0428
• Area b = 0.389
• Angle of heel under action of steady wind = 4.90 degrees
• Angle of roll to windward due to wave action = 13.46 degrees
• effective slope coefficient r =1.12

These results clearly indicate that the subject ship is not vulnerable to a stability failure in the
dead ship condition.

2.4 A level 2 vulnerability assessment for stability failure in the dead ship condition is
described in section 2.2.3 of the Interim Guidelines. The effective wave slope function is
calculated with the standard methodology described in paragraph 2.2.3.2.5 of the Interim
Guidelines and section 8.2 of the appendix 3. The obtained value is 0.723, which is compared
with the direct Froude-Krylov calculation, which is based on paragraph 8.1.1.4 of appendix 3,
and the weather criterion as shown in figure 2.1. The roll damping coefficient was calculated
by the methods of 9.2 of appendix 3 as the effective linear roll damping coefficient as a function

5
There could be some difference between GM computed from the moment of inertia of the waterplane area
and the value evaluated from the GZ curve. The GZ curves were adjusted so as to match the GM obtained
from waterplane calculations. In this section, the KG was adjusted to produce the desirable GM as assessed
by the GZ curve.

I:\CIRC\MSC\1\MSC.1-Circ.1652.docx
 MSC.1/Circ.1652
Annex, page 61

of the roll amplitude and then linear and cubic damping coefficients by the method of
paragraph 9.3.7 of appendix 3 are used for the stochastic linearization specified in
paragraph 9.3.5 of appendix 3.

Figure 2.1 Effective wave coefficient function for the C11 class containership

2.5 The product of the short-term failure index, which can be calculated with
paragraph 2.2.3.2.1 of the Interim Guidelines, and the weighted factor of short-term environmental
condition, which is available in section 2.7.2 of the Interim Guidelines, for each short-term
environmental condition, WiCs,i, is shown in table 2.1 for checking the calculation results.

Table 2.1 The value of 𝑾𝒊 𝑪ｓ𝒊 as the function of the significant wave height
and mean zero crossing wave period
Tz(s)
1.5 2.5 3.5 4.5 5.5 6.5 7.5 8.5 9.5 10.5 11.5 12.5 13.5 14.5 15.5 16.5 17.5 18.5
0.5 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
1.5 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
2.5 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
3.5 0 0 0 1.95E-20 0 0 0 0 0 0 0 0 0 0 0 0 0 0
4.5 0 0 0 0 1.26E-16 0 0 0 0 0 0 0 0 0 0 0 0 0
5.5 0 0 0 0 2.53E-12 8.83E-15 5.53E-19 0 0 0 0 0 0 0 0 0 0 0
6.5 0 0 0 0 3.7E-10 2.9E-11 9.92E-14 3.3E-17 0 0 0 0 0 0 0 0 0 0
7.5 0 0 0 0 0 2.53E-09 1.05E-10 5.68E-13 8.87E-16 7.81E-19 0 0 0 0 0 0 0 0
Hs(m) 8.5 0 0 0 0 0 3.69E-08 9.07E-09 3.43E-10 4.01E-12 3.3E-14 5.5E-16 3.35E-17 4.57E-18 6.15E-19 5.6E-20 2.44E-21 0 0
9.5 0 0 0 0 0 1.52E-07 1.03E-07 1.5E-08 6.88E-10 2.01E-11 8.49E-13 8.04E-14 1.23E-14 1.76E-15 1.76E-16 1.11E-17 4.4E-19 1.43E-20
10.5 0 0 0 0 0 0 3.7E-07 1.39E-07 1.71E-08 1.22E-09 9.83E-11 1.28E-11 2.21E-12 3.51E-13 4.16E-14 3.41E-15 1.88E-16 1.15E-17
11.5 0 0 0 0 0 0 5.74E-07 5.23E-07 1.37E-07 1.96E-08 2.66E-09 4.63E-10 9.28E-11 1.66E-11 2.3E-12 2.13E-13 1.65E-14 1.96E-15
12.5 0 0 0 0 0 0 6.03E-07 1.04E-06 4.96E-07 1.26E-07 2.61E-08 5.94E-09 1.38E-09 2.81E-10 4.44E-11 4.89E-12 3.9E-13 0
13.5 0 0 0 0 0 0 0 1.28E-06 1.16E-06 4.97E-07 1.58E-07 4.72E-08 1.33E-08 3.04E-09 5.66E-10 6.3E-11 1.12E-11 0
14.5 0 0 0 0 0 0 0 8.36E-07 1.25E-06 9.48E-07 4.02E-07 1.49E-07 4.84E-08 1.23E-08 2.41E-09 3.43E-10 0 0
15.5 0 0 0 0 0 0 0 0 7.19E-07 1.18E-06 6.59E-07 3.31E-07 1.15E-07 3.47E-08 5.64E-09 2.6E-09 0 0
16.5 0 0 0 0 0 0 0 0 0 6.48E-07 6.52E-07 3.27E-07 1.74E-07 4.69E-08 2.45E-08 0 0 0

2.6 The long-term probability index is obtained by summing up the above values following
paragraph 2.2.3.2 of the Interim Guidelines. The result of vulnerability assessment is as
follows:

C = 0.0000161 < RDS0 = 0.06 (2.1)

The result of level 2 vulnerability assessment is consistent with level 1 criterion and has
indicated no vulnerability of the considered ship to stability failure in the dead ship condition.

I:\CIRC\MSC\1\MSC.1-Circ.1652.docx
MSC.1/Circ.1652
Annex, page 62

3 Example of assessment of ship vulnerability to the excessive acceleration

failure mode

3.1 The subject ship is the C11 class container ship and details are shown in section 1.2.
Additional data of the assessment condition required for vulnerability assessment to the excessive
acceleration failure mode are shown in table 3.1. The considered location is the navigational deck,
based on paragraph 2.3.1.2.

Table 3.1 Assessment condition of C11 class container ship

ℎ𝑘 (m) Height of the navigational deck above the keel 48.72
𝑥(m) Longitudinal distance of the location where passenger or crew may 177.41
be present from the aft perpendicular
𝐺𝑀(m) Metacentric height without free-surface correction 8.00
𝐾𝐺(m) Height of the centre of gravity above the keel 12.75

3.2 Example of level 1 vulnerability assessment to the excessive acceleration

3.2.1 Calculation for Level 1 based on section 2.3.2 of the Interim Guidelines

.1 Factor 𝐾𝐿 , taking into account simultaneous action of roll, yaw and pitch motions:

𝑥 > 0.65𝐿, 𝐾𝐿 = 0.527 + 0.727𝑥/𝐿 = 1.019 (3.1)

.2 Natural roll period 𝑇𝑟 based on paragraph 2.7.1.2 of the Interim Guidelines:

𝑇𝑟 = 2𝐶𝐵⁄√𝐺𝑀 = 9.63 (𝑠) (3.2)

𝐶 = 0.373 + 0.023(𝐵⁄𝑑 ) − 0.043(𝐿⁄100) = 0.340 (3.3)

.3 Effective wave slope coefficient 𝑟:

𝐾1 +𝐾2 +𝑂𝐺∙𝐹
𝑟= 𝐵2 𝐶 𝑑
= 0.689 (3.4)
− 𝐵 −𝑂𝐺
12𝐶𝐵 𝑑 2

𝐵̃ = 2𝜋 2 𝐵⁄(𝑔𝑇𝑟2 ) = 0.869 (3.5)

̃
𝑇 = 4𝜋 2 𝐶𝐵 𝑑 ⁄(𝑔𝑇𝑟2 ) = 0.280 (3.6)
𝛽 = 𝑠𝑖𝑛(𝐵̃)⁄𝐵̃ = 0.879 (3.7)
𝜏 = 𝑒𝑥𝑝(−𝑇̃)⁄𝑇̃ = 2.698 (3.8)
𝐹 = 𝛽(𝜏 − 1⁄𝑇̃) = −0.766 (3.9)
𝐾1 = 𝑔𝛽𝑇𝑟2 (𝜏 + 𝜏𝑇̃ − 1⁄𝑇̃)⁄(4𝜋 2 ) = −2.356 (3.10)
𝐾2 = 𝑔𝛽𝑇𝑟2 (𝛽 − 𝑐𝑜𝑠 𝐵̃)⁄(4𝜋 2 ) = 14.479 (3.11)

.4 Wave steepness 𝑠:

Wave steepness 𝑠 is determined according to the natural roll period 𝑇𝑟 .

𝑠 = 0.082 (3.12)

.5 Non-dimensional logarithmic decrement of roll decay 𝛿𝜑 based on paragraph

9.3.1 of appendix 3 and paragraph 2.5.2.1 of the Interim Guidelines:

𝛿𝜑 = 0.5𝜋𝑅𝑃𝑅 = 0.651 (3.13)

𝑅𝑃𝑅 = 1.87, if the ship has a sharp bilge; otherwise,

= 0.17 + 0.425(100𝐴𝑘 ⁄(𝐿𝐵)), if 𝐶𝑚 ≥ 0.96,
= 0.17 + (10.625𝐶𝑚 − 9.755)(100𝐴𝑘 ⁄(𝐿𝐵)), if 0.94 < 𝐶𝑚 < 0.96

I:\CIRC\MSC\1\MSC.1-Circ.1652.docx
 MSC.1/Circ.1652
Annex, page 63

= 0.17 + 0.2125(100𝐴𝑘 ⁄(𝐿𝐵)), if 𝐶𝑚 ≤ 0.94, and

(100𝐴𝑘 ⁄(𝐿𝐵)) should not exceed 4
where 𝐶𝑚 ≥ 0.96,

𝑅𝑃𝑅 = 0.17 + 0.425(100𝐴𝑘 ⁄(𝐿𝐵)) = 0.414 (3.14)

.6 Roll amplitude 𝜑 :

𝜑 = 4.43𝑟𝑠 ⁄𝛿𝜑0.5 = 0.311(𝑟𝑎𝑑) (3.15)

.7 Lateral acceleration:

𝜑𝐾𝐿 (𝑔 + 4𝜋 2 ℎ𝑟 𝑇𝑟2 ) = 8.048 (𝑚⁄𝑠 2 ) > 𝑅𝐸𝐴1 = 4.64 (𝑚⁄𝑠 2 ) (3.16)

Height above the roll axis to the navigational deck

ℎ𝑟 = ℎ𝑘 − (𝐾𝐺 + 𝑑)⁄2 = 36.59(𝑚) (3.17)

Thus, this loading condition is possibly vulnerable to excessive accelerations.

3.3 Example of level 2 vulnerability assessment to the excessive accelerations

3.3.1 In level 2 vulnerability assessment for the excessive acceleration, simplified 1 degree of
freedom model is adopted for the amplitude of roll. For the calculation of the roll response, the
equivalent linear roll damping coefficient can be defined at the 15 degree of roll amplitude, which
is based on paragraph .9.3.8 of appendix 3, figure 3.1 shows the non-dimensional roll damping
coefficient 𝐵̂44 at the 15 degree of roll amplitude calculated by Ikeda’s simplified formula, which is
based on section 9.2 and paragraph 9.3.2 of appendix 3.

0.04
＾
B44

0.03

0.02

0.01

0
0 0.5 1 1.5 2
ω
̂ 𝟒𝟒 (𝝋𝒂 =15deg.)
Figure 3.1 Non-dimensional roll damping coefficient 𝑩

3.3.2 Froude-Krylov roll moment is calculated using the effective wave slope coefficient, based
on section 8.3 of appendix 3, as shown in figure 3.2.

I:\CIRC\MSC\1\MSC.1-Circ.1652.docx
MSC.1/Circ.1652
Annex, page 64

effective wave slope coeffirient

1

0.8

0.6

0.4

0.2

0
0 10 20 30 40
wavelength to ship breadth ratio
Figure 3.2 Effective wave slope coefficient

3.3.3 The lateral acceleration is calculated taking into account roll motion and vertical
acceleration. The standard deviation of lateral acceleration at zero speed and in beam seaway 𝜎𝐿𝐴𝑖
is obtained using the frequency spectrum of the seaway and reduction factor 0.75 for influence of
short-crestedness, based on paragraph 2.3.3.2.2 of the Interim Guidelines. The considered wave
scattering diagram is table 2.7.2.1.2 of the Interim Guidelines. Table 3.2 shows the standard
deviation of lateral acceleration 𝜎𝐿𝐴𝑖 for each wave condition.

Table 3.2 Standard deviation of lateral acceleration 𝝈𝑳𝑨𝒊

σLAi average zero-crossing period Tz (s)
Hs(m) 1.5 2.5 3.5 4.5 5.5 6.5 7.5 8.5 9.5 10.5 11.5 12.5 13.5 14.5 15.5 16.5 17.5 18.5
0.5 0.0005 0.0145 0.0528 0.1081 0.1851 0.2333 0.2395 0.2229 0.1987 0.1742 0.1520 0.1327 0.1164 0.1026 0.0910 0.0811 0.0726 0.0654
1.5 0.0016 0.0434 0.1585 0.3243 0.5553 0.6999 0.7184 0.6686 0.5961 0.5225 0.4559 0.3982 0.3493 0.3078 0.2729 0.2432 0.2179 0.1962
2.5 0.0027 0.0724 0.2642 0.5406 0.9255 1.1666 1.1975 1.1145 0.9935 0.8709 0.7598 0.6637 0.5821 0.5130 0.4548 0.4053 0.3632 0.3270
3.5 0.0038 0.1013 0.3699 0.7569 1.2958 1.6331 1.6763 1.5601 1.3910 1.2194 1.0640 0.9292 0.8149 0.7183 0.6366 0.5674 0.5083 0.4578
4.5 0.0049 0.1303 0.4756 0.9731 1.6658 2.0998 2.1552 2.0057 1.7883 1.5675 1.3678 1.1946 1.0479 0.9235 0.8185 0.7295 0.6536 0.5886
5.5 0.0060 0.1592 0.5812 1.1895 2.0362 2.5663 2.6342 2.4515 2.1856 1.9160 1.6715 1.4601 1.2806 1.1287 1.0005 0.8916 0.7989 0.7194
6.5 0.0071 0.1882 0.6869 1.4057 2.4062 3.0330 3.1130 2.8972 2.5830 2.2643 1.9756 1.7257 1.5133 1.3338 1.1824 1.0536 0.9441 0.8501
7.5 0.0082 0.2171 0.7926 1.6217 2.7765 3.5000 3.5917 3.3437 2.9804 2.6127 2.2795 1.9912 1.7461 1.5392 1.3642 1.2157 1.0895 0.9809
8.5 0.0093 0.2461 0.8983 1.8382 3.1467 3.9661 4.0706 3.7881 3.3779 2.9609 2.5834 2.2568 1.9791 1.7444 1.5463 1.3780 1.2345 1.1118
9.5 0.0104 0.2750 1.0040 2.0543 3.5171 4.4328 4.5497 4.2344 3.7749 3.3091 2.8874 2.5221 2.2118 1.9496 1.7280 1.5401 1.3799 1.2426
10.5 0.0115 0.3040 1.1095 2.2707 3.8872 4.8990 5.0289 4.6797 4.1725 3.6579 3.1906 2.7877 2.4446 2.1548 1.9100 1.7021 1.5251 1.3733
11.5 0.0126 0.3330 1.2153 2.4870 4.2568 5.3656 5.5082 5.1254 4.5706 4.0062 3.4957 3.0532 2.6775 2.3601 2.0919 1.8644 1.6703 1.5040
12.5 0.0137 0.3619 1.3210 2.7031 4.6271 5.8327 5.9867 5.5714 4.9679 4.3543 3.7987 3.3181 2.9103 2.5653 2.2738 2.0263 1.8155 1.6349
13.5 0.0148 0.3909 1.4265 2.9194 4.9980 6.2992 6.4653 6.0175 5.3647 4.7032 4.1037 3.5847 3.1431 2.7706 2.4556 2.1886 1.9609 1.7658
14.5 0.0159 0.4198 1.5323 3.1356 5.3675 6.7654 6.9448 6.4630 5.7619 5.0507 4.4068 3.8497 3.3764 2.9757 2.6374 2.3505 2.1062 1.8966
15.5 0.0170 0.4488 1.6380 3.3526 5.7385 7.2326 7.4236 6.9087 6.1595 5.3991 4.7106 4.1146 3.6083 3.1812 2.8194 2.5128 2.2514 2.0273
16.5 0.0181 0.4777 1.7438 3.5679 6.1082 7.6987 7.9025 7.3546 6.5574 5.7480 5.0150 4.3806 3.8419 3.3867 3.0013 2.6749 2.3967 2.1580

3.3.4 The short-term excessive acceleration failure index 𝐶𝑠,𝑖 , which is the probability that the
ship will exceed a specified lateral acceleration, is calculated using the standard deviation of lateral
acceleration 𝜎𝐿𝐴𝑖 , based on paragraph 2.3.3.2.1 of the Interim Guidelines. The long-term probability
index 𝐶 is calculated as a weighted average by weighting factor for the short-term condition 𝑊𝑖 ,
and 𝐶𝑠,𝑖 , based on paragraph 2.3.3.2 of the Interim Guidelines. Table 3.3 shows the value of 𝑊𝑖 𝐶𝑠,𝑖
for each wave condition.

Table 3.3 The value of 𝑾𝒊 𝑪𝒔,𝒊 for each wave condition

W iCs,i average zero-crossing period Tz (s)
Hs(m) 1.5 2.5 3.5 4.5 5.5 6.5 7.5 8.5 9.5 10.5 11.5 12.5 13.5 14.5 15.5 16.5 17.5 18.5
0.5 0.00E+00 0.00E+00 0.00E+00 0.00E+00 0.00E+00 0.00E+00 0.00E+00 0.00E+00 0.00E+00 0.00E+00 0.00E+00 0.00E+00 0.00E+00 0.00E+00 0.00E+00 0.00E+00 0.00E+00 0.00E+00
1.5 0.00E+00 0.00E+00 0.00E+00 7.21E-203 1.67E-70 1.09E-44 2.50E-42 9.93E-49 3.67E-61 2.04E-79 4.60E-104 5.22E-136 2.36E-176 2.42E-226 1.96E-287 0.00E+00 0.00E+00 0.00E+00
2.5 0.00E+00 0.00E+00 0.00E+00 6.97E-77 7.91E-28 9.48E-18 1.65E-16 1.10E-18 3.27E-23 5.78E-30 4.11E-39 5.87E-51 7.03E-66 2.59E-84 9.69E-107 1.19E-133 0.00E+00 0.00E+00
3.5 0.00E+00 0.00E+00 0.00E+00 6.64E-43 1.25E-16 1.02E-10 1.18E-09 1.47E-10 8.05E-13 2.48E-16 3.80E-21 2.12E-27 2.86E-35 5.71E-45 9.67E-57 7.35E-71 1.38E-87 0.00E+00
4.5 0.00E+00 0.00E+00 0.00E+00 0.00E+00 1.77E-12 3.57E-08 4.29E-07 2.10E-07 1.13E-08 8.41E-11 8.59E-14 1.04E-17 1.20E-22 1.00E-28 4.46E-36 7.02E-45 2.43E-55 0.00E+00
5.5 0.00E+00 0.00E+00 0.00E+00 0.00E+00 9.10E-11 3.42E-07 4.85E-06 5.34E-06 1.00E-06 4.07E-08 3.74E-10 7.33E-13 2.72E-16 1.62E-20 1.28E-25 1.09E-31 7.20E-39 4.13E-47
6.5 0.00E+00 0.00E+00 0.00E+00 0.00E+00 4.92E-10 6.74E-07 1.17E-05 2.24E-05 9.28E-06 1.07E-06 3.65E-08 3.72E-10 1.06E-12 7.62E-16 1.23E-19 3.78E-24 1.80E-29 1.22E-35
7.5 0.00E+00 0.00E+00 0.00E+00 0.00E+00 0.00E+00 5.90E-07 1.25E-05 3.64E-05 2.64E-05 6.10E-06 4.99E-07 1.48E-08 1.57E-10 5.55E-13 6.03E-16 1.83E-19 1.47E-23 1.92E-28
8.5 0.00E+00 0.00E+00 0.00E+00 0.00E+00 0.00E+00 3.29E-07 8.44E-06 3.43E-05 3.77E-05 1.45E-05 2.20E-06 1.38E-07 3.58E-09 3.76E-11 1.52E-13 2.17E-16 9.75E-20 1.24E-23
9.5 0.00E+00 0.00E+00 0.00E+00 0.00E+00 0.00E+00 1.73E-07 4.21E-06 2.27E-05 3.48E-05 1.98E-05 4.74E-06 5.15E-07 2.59E-08 5.95E-10 6.12E-12 2.63E-14 4.23E-17 2.91E-20
10.5 0.00E+00 0.00E+00 0.00E+00 0.00E+00 0.00E+00 0.00E+00 1.79E-06 1.19E-05 2.39E-05 1.85E-05 6.36E-06 1.05E-06 8.70E-08 3.60E-09 7.47E-11 7.36E-13 3.11E-15 8.31E-18
11.5 0.00E+00 0.00E+00 0.00E+00 0.00E+00 0.00E+00 0.00E+00 6.14E-07 5.29E-06 1.33E-05 1.33E-05 6.11E-06 1.42E-06 1.73E-07 1.13E-08 4.02E-10 6.80E-12 6.47E-14 5.79E-16
12.5 0.00E+00 0.00E+00 0.00E+00 0.00E+00 0.00E+00 0.00E+00 2.61E-07 2.12E-06 6.26E-06 7.83E-06 4.56E-06 1.39E-06 2.32E-07 2.20E-08 1.18E-09 3.26E-11 4.58E-13 0.00E+00
13.5 0.00E+00 0.00E+00 0.00E+00 0.00E+00 0.00E+00 0.00E+00 0.00E+00 7.94E-07 2.63E-06 3.97E-06 2.87E-06 1.09E-06 2.38E-07 3.03E-08 2.40E-09 8.67E-11 3.67E-12 0.00E+00
14.5 0.00E+00 0.00E+00 0.00E+00 0.00E+00 0.00E+00 0.00E+00 0.00E+00 3.16E-07 9.39E-07 1.82E-06 1.51E-06 7.00E-07 1.91E-07 3.06E-08 2.97E-09 1.65E-10 0.00E+00 0.00E+00
15.5 0.00E+00 0.00E+00 0.00E+00 0.00E+00 0.00E+00 0.00E+00 0.00E+00 0.00E+00 2.81E-07 7.68E-07 6.86E-07 4.08E-07 1.24E-07 2.58E-08 2.35E-09 4.90E-10 0.00E+00 0.00E+00
16.5 0.00E+00 0.00E+00 0.00E+00 0.00E+00 0.00E+00 0.00E+00 0.00E+00 0.00E+00 0.00E+00 2.33E-07 2.95E-07 1.63E-07 7.67E-08 1.51E-08 4.79E-09 0.00E+00 0.00E+00 0.00E+00

3.3.5 Based on paragraph 2.3.3.1 of the Interim Guidelines, the result of level 2 vulnerability
assessment to excessive acceleration is as follows:

I:\CIRC\MSC\1\MSC.1-Circ.1652.docx
 MSC.1/Circ.1652
Annex, page 65

𝐶 = ∑𝑁
𝑖=1 𝑊𝑖 𝐶𝑠,𝑖 = 0.00047 > 𝑅𝐸𝐴2 = 0.00039 (3.18)

Thus, this loading condition is possibly vulnerable to excessive accelerations.

4 Example of assessment of ship vulnerability to the pure loss of stability failure

mode

4.1 As an example, the following data of a containership at the full load condition are used:

LBP 262.0 m ship length between perpendiculars

L 262.0m ship length defined in 2008 Intact Stability Code
B 40.0 m moulded ship breadth
D 24.45 m moulded ship depth
d 11.5 m moulded mean ship draught
 0m initial trim
LCG 125.52 m longitudinal centre of gravity from aft perpendicular
CB 0.559 block coefficient
GM 1.965 m metacentric height with free surface correction
VCG 18.4 m vertical centre of gravity above baseline
KB 6.54 m vertical centre of buoyancy above baseline
Vs 12.165 m/s ship service speed

4.2 Level 1 based on section 2.4.2 of the Interim Guidelines

4.2.1 Firstly, the Froude number, Fn, is calculated.

Fn = Vs/√𝑔𝐿 = 12.165/√9.81 × 262 = 0.240 (4.1)

Thus, the criteria for pure loss of stability should be applied to the ship.

4.2.2 Secondly, the lower draught is calculated:

dL = Min{d - 0.25dfull, LSw/2} = Min{0.75 x 11.5, 262 x 0.0334/2} =

Min{8.625, 4.3754} = 4.3754 m (4.2)

dL = d - dL = 11.5 - 4.3754 = 7.125 m (4.3)

so that, according to the hydrostatic data as shown in Figure4.1, the relevant IL is 665,500m4.

4.2.3 Thirdly, GMmin is calculated as follows:

GMmin = KB + (IL/V) – VCG = KB + (IL/(CB x LBP x B x d) - VCG =

6.54 + (665,500/(0.559 x 262 x 40 x 11.5) - 18.4 = -1.982 m (4.4)

Therefore, GMmin( = -1.982) < RLA( = 0.05) so that the ship is judged as possibly vulnerable to
pure loss of stability.

I:\CIRC\MSC\1\MSC.1-Circ.1652.docx
MSC.1/Circ.1652
Annex, page 66

Figure 4.1 Relationship between draught and the moment of inertia of the waterplane
for the sample ship

4.3 Level 2 based on section 2.4.3 of the Interim Guidelines

4.3.1 Firstly, GZ curves for the ship with the wave steepness ranging from 0 to 0.1 are
calculated. Examples are shown in figure 4.2.

4.3.2 Secondly, criteria 1 and 2 are applied to the ship. The results are as follows:

CR1 = 0.000, where the critical wave steepness for the angle of vanishing stability
of 30 degrees is 0.07291.

CR2 = 0.003821, where the critical wave steepness for the angle of heel of 25 degrees
is 0.03941.

Thus: Max{CR1,CR2} = 0.003821. (4.5)

Since this is smaller than 0.06, the ship is not judged as vulnerable to pure loss of stability.

4.3.3 Figure 4.2 shows the GZ curves in waves at wave crest amidships under different
wave steepness. In case of a wave steepness of 0.04, the GM in waves is almost zero. On the
other hand, the level 1 criterion uses the wave steepness of 0.0334. Thus, this figure indicates
the possibility of negative GM for the ship exists as the level 1 criterion suggests with some
margin. The level 2 criterion explains that, if the ship meets a certain wave crest, significant
heel could occur under the riding on crest situation where the ship is slowly overtaken by a
wave crest. However, the level 2 criterion also indicates that the ship rarely meets such waves,
even in the North Atlantic. This is a physical explanation based on the vulnerability criteria for
the pure loss of stability failure of the ship.

I:\CIRC\MSC\1\MSC.1-Circ.1652.docx
 MSC.1/Circ.1652
Annex, page 67

GZ in waves (m)
2 wave steepness
of 0
1.5
wave steepness
1
of 0.04
0.5
wave steepness
0 of 0.07

-0.5 heeling level for

0 10 20 30 40 50 wave steepness
heel angle (degrees) of 0.04

Figure 4.2 GZ curves for the ship in waves at wave crest amidships

5 Example of assessment of ship vulnerability to parametric rolling

5.1 As an example, the following data of a containership at the full load condition are
used:

LBP 262.0 m ship length between perpendiculars

Lf 262.0 m ship length defined in 2008 Intact Stability Code
B 40.0 m moulded ship breadth
D 24.45 m moulded ship depth
d 12.34 m moulded mean ship draught
 0m trim
LCG 124.7 m longitudinal centre of gravity from aft perpendicular
CB 0.576 block coefficient
Cm 0.962 midship section coefficient
GM 1.965 m metacentric height with free surface correction
T 25.7 s ship natural roll period
VCG 18.37 m vertical centre of gravity above baseline
lBK/LBP 0.292 bilge keel length normalized with ship length between
perpendiculars
bBK/B 0.0100 bilge keel width normalized with moulded ship breadth
Vs 12.861 m/s ship service speed

I:\CIRC\MSC\1\MSC.1-Circ.1652.docx
MSC.1/Circ.1652
Annex, page 68

Moment of inertia of waterplane

(m4)
1500000
1000000
500000
0
0.00 4.00 8.00 12.00 16.00 20.00 24.00
draught (metres)

Figure 5.1 The moment of inertia of the waterplane vs. draught for the subject
ship

5.2 Level 1 based on section 2.5.2 of the Interim Guidelines

5.2.1 Firstly, the bilge keel area ratio, 100AK/(LBPB), is calculated:

100AK/(LBPB) = 100 x (lBK/LBP) x (bBK/B) x 2 = 100 x 0.292 x 0.01 x 2 = 0.584 (5.1)

5.2.2 Secondly, considering Cm = 0.962 at the full load condition, the standard, RPR, is
calculated:

RPR = 0.17+ 0.425(100AK/(LBPB)) =

0.17 + 0.425 x 0.584 = 0.4182 (5.2)

5.2.3 Thirdly, the lower draught is calculated:

dL = Min{d - 0.25dfull, LSw/2} = Min{0.75 x 12.34, 262 x 0.0167/2} =

= Min{9.255, 2.188} =2.188 m (5.3)

dL = d - dL = 12.34 - 2.188 = 10.152 m (5.4)

so that, according to the hydrostatic data, the relevant IL is 847,948 m4.

5.2.4 Fourthly, the higher draught is calculated:

dH = Min{D - d, LSw/2} = Min{24.45 - 12.34, 262 x 0.0167/2} =

= Min{12.11, 2.188} = 2.188 m (5.5)

dH = d + dH = 12.34 + 2.188 = 14.528 m (5.6)

so that, according to the hydrostatic data shown in figure 5.1, the relevant IH is 1,106,866 m4.

5.2.5 Fifthly, GM/GM is calculated as follows:

GM/GM = (IH - IL)/(2V)/GM = (IH - IL)/(2 x CB x LBP x B x d)/GM =

(1106866 - 847948)/(2 x 0.576 x 262 x 40 x 12.34)/1.965 = 0.8844 (5.7)

I:\CIRC\MSC\1\MSC.1-Circ.1652.docx
 MSC.1/Circ.1652
Annex, page 69

Therefore, GM/GM( = 0.8844) > RPR( = 0.4182) so that the ship is judged as possibly
vulnerable to parametric roll. The reason of this judgement is explained in paragraph 6.1 of
appendix 3.

5.3 First check of level 2 based on section 2.5.3 of the Interim Guidelines

5.3.1 Firstly, GM values in longitudinal waves are calculated for the 16 wave cases specified
in table 2.5.3.2.3 of the Interim Guidelines. As an example, the GM variation for the wave case
no.9, in which the wave height is 3.625m and the wavelength is 243.705m, are shown in
figure 5.2. In this case, the maximum GM, GMmax, is 3.306m: the minimum GM, GMmin,
is 0.807m. Thus, GM = (GMmax + GMmin)/2 = 2.057m: GM = (GMmax - GMmin)/2 = 1.250m.
Therefore, GM/GM = 0.6077.

Figure 5.2 GM variation in longitudinal waves for the wave case no. 9

On the other hand, RPR was already calculated in paragraph 5.2.2 as 0.4182: VPR is calculated
by the formula in paragraph 2.5.3.2.2 of the Interim Guidelines so that VPR = 0.1044m/s. Thus,
the requirement in paragraph 2.5.3.2.1 of the Interim Guidelines is not satisfied because GM
= 2.057m > 0 but GM/GM ( = 0.6077) > RPR ( = 0.4182). The requirement in
paragraph 2.5.3.2.2 of the Interim Guidelines is also not satisfied because VPR ( = 0.1044m/s)
< Vs ( = 12.861 m/s). Therefore, the C value defined in paragraph 2.5.3.2 of the Interim
Guidelines shall be 0. Repeating the same calculations for other wave cases, the dangerous
wave cases among 16 are identified as shown in table 5.1.
Table 5.1 Judgement for wave cases
wave case W λ(m) H(m) GM/GM VPR(m/s) Ci Wi C i
number
1 0.000013 22.574 0.35 0.0169 4.189 0 0
2 0.00165 37.316 0.495 0.0322 4.747 0 0
3 0.0209 55.734 0.857 0.0774 5.030 0 0
4 0.0928 77.857 1.295 0.1210 5.038 0 0
5 0.199 103.655 1.732 0.2410 4.630 0 0
6 0.249 133.139 2.205 0.1908 4.102 0 0
7 0.209 166.309 2.697 0.4336 3.215 1 0.209
8 0.129 203.164 3.176 0.5860 1.859 1 0.129
9 0.0625 243.705 3.625 0.6077 0.104 1 0.0625
10 0.0248 287.931 4.04 0.5793 1.852 1 0.0248
11 0.00837 335.843 4.421 0.5414 3.964 1 0.00837
12 0.00247 387.44 4.769 0.5014 6.272 1 0.00247
13 0.000658 442.723 5.097 0.4632 8.852 1 0.000658
14 0.000158 501.691 5.37 0.4130 11.757 0 0
15 0.000034 564.345 5.621 0.3594 14.963 0 0
16 0.000007 630.684 5.95 0.3157 18.398 0 0

I:\CIRC\MSC\1\MSC.1-Circ.1652.docx
MSC.1/Circ.1652
Annex, page 70

Then, by summing up WiCi, the final index, C1, is calculated as follows:

C1 = 0.4368. (5.8)

Since C1 is larger than the standard of 0.06, the ship is judged as possibly vulnerable to
parametric roll.

5.4 Second check of level 2 based on section 2.5.3 of the Interim Guidelines

5.4.1 First, the maximum roll angles should be calculated for the effective waves having the
different effective wave height hj =0.01*j (j = 1 …10) by calculating the time series. As a case
example, the results for the Froude number of 0.0 in head waves are shown in figure 5.3.

wave steepness= 0.01 wave steepness= 0.01

60 60

40 40

roll angle [deg.]

roll angle [deg.]

20 20

0 0
0 100 200 300 400 500 0 0.2 0.4 0.6 0.8 1.0
-20 -20

-40 -40

-60 -60
time [sec.] relative ship position normalised by wavelength

wave steepness= 0.02

60

40
roll angle [deg.]

20

0
0 0.2 0.4 0.6 0.8 1.0
-20

-40

-60
relative ship position normalised by wavelength

wave steepness= 0.03 wave steepness= 0.03

60 60

40 40
roll angle [deg.]
roll angle [deg.]

20 20

0 0
0 100 200 300 400 500 0 0.2 0.4 0.6 0.8 1.0
-20 -20

-40 -40

-60
-60
relative ship position normalised by wavelength
time [sec.]

I:\CIRC\MSC\1\MSC.1-Circ.1652.docx
 MSC.1/Circ.1652
Annex, page 71

wave steepness= 0.04

wave steepness= 0.04
60
60
40
roll angle [deg.]

40

amplitude [deg.]
20
20
0 0
0 50 100 150 200 250 0 0.2 0.4 0.6 0.8 1.0
-20 -20
-40 -40

-60 -60
time [sec.] relative ship position normalised by wavelength

wave steepness= 0.05 wave steepness= 0.05

60 60

40 40

roll angle [deg.]

roll angle [deg.]

20
20
0
0 0 0.2 0.4 0.6 0.8 1.0
0 50 100 150 200 250 -20
-20
-40
-40
-60
-60 relative ship position normalised by wavelength
time [sec.]

wave steepness= 0.06 wave steepness= 0.06

60 60

40 40
roll angle [deg.]
roll angle [deg.]

20
20
0
0 0 0.2 0.4 0.6 0.8 1.0
0 200 400 600 800 -20
-20
-40
-40
-60
-60 relative ship position normalised by wavelength
time [sec.]

wave steepness= 0.07 wave steepness= 0.07

60 60

40 40
roll angle [deg.]
roll angle [deg.]

20
20
0
0 0 0.2 0.4 0.6 0.8 1.0
0 100 200 300 400 500 600 700 -20
-20
-40
-40
-60
-60 relative ship position normalised by wavelength
time [sec.]

I:\CIRC\MSC\1\MSC.1-Circ.1652.docx
MSC.1/Circ.1652
Annex, page 72

wave steepness= 0.08

180

120
roll angle [deg.]

60

0
0 100 200 300 400 500
-60

-120

-180
time [sec.]

wave steepness= 0.09 wave steepness= 0.1

180 180
120 120
roll angle [deg.]

roll angle [deg.]

60 60

0 0
0 50 100 150 200 0 10 20 30 40 50 60
-60 -60

-120 -120

-180 -180
time [sec.] time [sec.]

Figure 5.3 Time series of roll angle from the initial state (left) and steady states of roll
angle during two encounter wave cycles (right) as functions of the relative longitudinal
ship position to a wave trough, which is normalized with the wavelength.

5.4.2 The time series from the initial states include some transient behaviours so that the
steady states should be identified by checking the convergence towards periodic states.
In figure 5.3 it should be underlined that the cases for the wave steepness ranging
from 0.05 to 0.07 show overshoot responses so that the maximum roll angle in the overall time
series is not the maximum angle to be used for the calculation of the C2 index. While the
transient response depends on the initial wave phase, the steady-state does not. In the cases
for the wave steepness ranging from 0.01 to 0.07, each steady-state includes two lines within
one encounter wave cycle. This means that the encounter wave cycle is exactly half the roll
period so that this steady roll motion can be regarded as the parametric rolling that the level 1
criterion deals with. In the case of the wave steepness of 0.08, four lines exist within one
encounter wave cycle so that this periodic motion is not relevant to the phenomenon to be
evaluated here. In other words, one roll occurs during four encounter waves. In the cases for
the wave steepness ranging from 0.09 to 0.1, the roll angle exceeds 180 degrees so that this
transient behaviour is also not the phenomenon to be evaluated here.

5.4.3 The relationship between the wave steepness and maximum roll angle during the
steady motion can be obtained as shown in figure 5.4 by using the results shown in figure 5.3.
In this case, the critical wave steepness under which the maximum roll angle is equal to the
threshold of 25 degrees can be calculated by using the linear interpolation as 0.00948. It is
noteworthy here that the maximum roll angle under the wave steepness of 0.07 is smaller than
that of 0.06. Not only the amplitude of restoring moment variation but also the mean of the
restoring moment variation increases with the wave steepness. Thus, the condition for
parametric rolling could be violated with larger wave steepness often in regular waves but not
so often in irregular waves because of its spectrum of the incident waves. Thus, the
requirement in paragraph 7.5 of appendix 3 is indispensable for excluding the drawback due
to the approximation using the regular waves.

I:\CIRC\MSC\1\MSC.1-Circ.1652.docx
 MSC.1/Circ.1652
Annex, page 73

Fn = 0.0, head waves

60

maximum roll angle [deg.]

50

40

30

20

10

0
0 0.01 0.02 0.03 0.04 0.05 0.06 0.07
wave steepness

Figure 5.4 Relationship between the wave steepness and the maximum roll angle for the
Froude number of 0.0 in head waves.

5.4.4 Repeating the same procedure for different ship speeds and heading, the critical wave
steepness can be determined as shown in table 5.2. Here, if the largest value of the maximum
roll angle cannot be determined up to the wave steepness of 0.1, the critical wave steepness
is at least larger than 0.1. On the other hand, the 1/3 largest steepness of the effective wave
can be calculated from the wave scatter table as shown in table 5.3. Here, the shaded cells
indicate the effective wave steepnesses are larger than the critical wave steepness for the
Froude number of 0.0 in head waves so that the index Cs,i for the relevant cell should be 1.
Thus, the C2 (Fn = 0.0, h) value for the Froude number of 0.0 in head waves can be calculated
as 0.00948.

I:\CIRC\MSC\1\MSC.1-Circ.1652.docx
MSC.1/Circ.1652
Annex, page 74

Table 5.2 Critical wave steepness

Fn / heading Critical wave steepness
0.254 / head > 0.1000
0.252 / head > 0.1000
0.245 / head > 0.1000
0.234 / head > 0.1000
0.220 / head > 0.1000
0.201 / head > 0.1000
0.179 / head > 0.1000
0.154 / head 0.05760
0.127 / head 0.04457
0.097 / head 0.03367
0.066 / head 0.02357
0.033 / head 0.01451
0.000 / head 0.00948
0.000 / follow 0.00924
0.033 / follow 0.01720
0.066 / follow 0.03589
0.097 / follow > 0.1000
0.127 / follow > 0.1000
0.154 / follow > 0.1000
0.179 / follow 0.09294
0.201 / follow > 0.1000
0.220 / follow > 0.1000
0.234 / follow > 0.1000
0.245 / follow > 0.1000
0.252 / follow > 0.1000
0.254 / follow > 0.1000

Table 5.3 The 1/3 largest steepness of the effective wave

Effective
wave averaging zero-crossing period Tz(s)
steepness
Hs(m) 1.5 2.5 3.5 4.5 5.5 6.5 7.5 8.5 9.5 10.5 11.5 12.5 13.5 14.5 15.5 16.5 17.5 18.5

0.5 1.40E-09 2.74E-05 1.06E-04 2.01E-04 3.83E-04 7.33E-04 1.07E-03 1.29E-03 1.38E-03 1.38E-03 1.33E-03 1.26E-03 1.17E-03 1.08E-03 9.91E-04 9.09E-04 8.34E-04 7.65E-04

1.5 4.19E-09 8.22E-05 3.17E-04 6.04E-04 1.15E-03 2.20E-03 3.21E-03 3.86E-03 4.13E-03 4.14E-03 4.00E-03 3.77E-03 3.51E-03 3.24E-03 2.97E-03 2.73E-03 2.50E-03 2.30E-03

2.5 6.99E-09 1.37E-04 5.28E-04 1.01E-03 1.92E-03 3.66E-03 5.35E-03 6.43E-03 6.88E-03 6.90E-03 6.66E-03 6.28E-03 5.84E-03 5.39E-03 4.96E-03 4.55E-03 4.17E-03 3.83E-03

3.5 9.79E-09 1.92E-04 7.39E-04 1.41E-03 2.68E-03 5.13E-03 7.50E-03 9.00E-03 9.64E-03 9.66E-03 9.33E-03 8.79E-03 8.18E-03 7.55E-03 6.94E-03 6.36E-03 5.84E-03 5.36E-03

4.5 1.26E-08 2.47E-04 9.51E-04 1.81E-03 3.45E-03 6.60E-03 9.64E-03 1.16E-02 1.24E-02 1.24E-02 1.20E-02 1.13E-02 1.05E-02 9.71E-03 8.92E-03 8.18E-03 7.50E-03 6.89E-03

5.5 1.54E-08 3.01E-04 1.16E-03 2.22E-03 4.21E-03 8.06E-03 1.18E-02 1.41E-02 1.51E-02 1.52E-02 1.47E-02 1.38E-02 1.29E-02 1.19E-02 1.09E-02 1.00E-02 9.17E-03 8.42E-03

6.5 1.82E-08 3.56E-04 1.37E-03 2.62E-03 4.98E-03 9.53E-03 1.39E-02 1.67E-02 1.79E-02 1.79E-02 1.73E-02 1.63E-02 1.52E-02 1.40E-02 1.29E-02 1.18E-02 1.08E-02 9.95E-03

7.5 2.10E-08 4.11E-04 1.58E-03 3.02E-03 5.75E-03 1.10E-02 1.61E-02 1.93E-02 2.06E-02 2.07E-02 2.00E-02 1.88E-02 1.75E-02 1.62E-02 1.49E-02 1.36E-02 1.25E-02 1.15E-02

8.5 2.38E-08 4.66E-04 1.80E-03 3.42E-03 6.51E-03 1.25E-02 1.82E-02 2.18E-02 2.34E-02 2.35E-02 2.26E-02 2.14E-02 1.99E-02 1.83E-02 1.69E-02 1.55E-02 1.42E-02 1.30E-02

9.5 2.66E-08 5.21E-04 2.01E-03 3.83E-03 7.28E-03 1.39E-02 2.03E-02 2.44E-02 2.62E-02 2.62E-02 2.53E-02 2.39E-02 2.22E-02 2.05E-02 1.88E-02 1.73E-02 1.58E-02 1.45E-02

10.5 2.94E-08 5.76E-04 2.22E-03 4.23E-03 8.04E-03 1.54E-02 2.25E-02 2.70E-02 2.89E-02 2.90E-02 2.80E-02 2.64E-02 2.45E-02 2.26E-02 2.08E-02 1.91E-02 1.75E-02 1.61E-02

11.5 3.22E-08 6.30E-04 2.43E-03 4.63E-03 8.81E-03 1.69E-02 2.46E-02 2.96E-02 3.17E-02 3.18E-02 3.06E-02 2.89E-02 2.69E-02 2.48E-02 2.28E-02 2.09E-02 1.92E-02 1.76E-02

12.5 3.49E-08 6.85E-04 2.64E-03 5.04E-03 9.58E-03 1.83E-02 2.68E-02 3.21E-02 3.44E-02 3.45E-02 3.33E-02 3.14E-02 2.92E-02 2.70E-02 2.48E-02 2.27E-02 2.08E-02 1.91E-02

13.5 3.77E-08 7.40E-04 2.85E-03 5.44E-03 1.03E-02 1.98E-02 2.89E-02 3.47E-02 3.72E-02 3.73E-02 3.60E-02 3.39E-02 3.15E-02 2.91E-02 2.68E-02 2.46E-02 2.25E-02 2.07E-02

14.5 4.05E-08 7.95E-04 3.06E-03 5.84E-03 1.11E-02 2.13E-02 3.11E-02 3.73E-02 3.99E-02 4.00E-02 3.86E-02 3.64E-02 3.39E-02 3.13E-02 2.87E-02 2.64E-02 2.42E-02 2.22E-02

15.5 4.33E-08 8.50E-04 3.27E-03 6.24E-03 1.19E-02 2.27E-02 3.32E-02 3.98E-02 4.27E-02 4.28E-02 4.13E-02 3.89E-02 3.62E-02 3.34E-02 3.07E-02 2.82E-02 2.59E-02 2.37E-02

16.5 4.61E-08 9.04E-04 3.49E-03 6.65E-03 1.26E-02 2.42E-02 3.53E-02 4.24E-02 4.54E-02 4.56E-02 4.40E-02 4.15E-02 3.86E-02 3.56E-02 3.27E-02 3.00E-02 2.75E-02 2.53E-02

I:\CIRC\MSC\1\MSC.1-Circ.1652.docx
 MSC.1/Circ.1652
Annex, page 75

5.4.5 Repeating the same procedure for different ship speeds and heading, the C2 (Fn, )
values can be determined as table 5.4.

Table 5.4 C2 (Fn, ) values for different ship speeds and heading

Fn / heading C2(Fni, )
0.254 / head 0.00000
0.252 / head 0.00000
0.245 / head 0.00000
0.234 / head 0.00000
0.220 / head 0.00000
0.201 / head 0.00000
0.179 / head 0.00000
0.154 / head 0.00000
0.127 / head 0.00000
0.097 / head 0.00039
0.066 / head 0.01016
0.033 / head 0.15082
0.000 / head 0.39110
0.000 / follow 0.40224
0.033 / follow 0.08247
0.066 / follow 0.00018
0.097 / follow 0.00000
0.127 / follow 0.00000
0.154 / follow 0.00000
0.179 / follow 0.00000
0.201 / follow 0.00000
0.220 / follow 0.00000
0.234 / follow 0.00000
0.245 / follow 0.00000
0.252 / follow 0.00000
0.254 / follow 0.00000

5.4.6 By using the formula in paragraph 2.5.3.3 of the Interim Guidelines, the final value of
C2 is obtained as follows:

C2 = 0.02563 (5.9)

Since C2 > 0.025, the ship is judged as vulnerable to parametric roll.

5.5 Level 1 as shown in section 5.2 and the first check of level 2 as shown in table 5.1
indicate that parametric roll could occur for the ship under the wave steepness of about 0.009
or larger. However, these do not provide any information about the magnitude. Table 5.2 shows
that the parametric roll amplitude could be more than 25 degrees, which could induce cargo
damage and so on, only when the ship is at zero or very low forward velocity. It also indicates
the ship could often meet such sea states in the North Atlantic. Therefore, the level 2 criterion
judges the ship as vulnerable to parametric roll. At the same time, table 5.2 also suggests that
a ship can be operated without vulnerability to parametric rolling, provided that the ship service
speed is maintained in the specified sea states.

I:\CIRC\MSC\1\MSC.1-Circ.1652.docx
MSC.1/Circ.1652
Annex, page 76

6 Example of assessment of ship vulnerability to surf-riding/broaching

6.1 As an example, the following data of a fishing vessel is used:

LBP 34.5 m
B 7.6 m
d 2.65 m
trim 0.30 m
LCB 1.31 m aft from the midship
CB 0.597
Service Fn 0.40
Dp 2.60 m
tp 0.142
wp 0.156
MX/M 0.0667
r0 0
r1 -4273.53 [Ns/m]
r2 7491.11 [Ns2/m2]
r3 -2668.12 [Ns3/m3]
r4 408.20 [Ns4/m4]
r5 -17.005 [Ns5/m5]
0 0.2244
1 -0.2283
2 -0.1373

The local breadth B (m), local draught d (m) and sectional area A (m2) are provided below as
a function of the longitudinal section position (m) measured from the midship as shown in
table 6.1:

Table 6.1 Sectional areas, local breadth and local draught of the fishing vessel

x= -21.65 A= 0.00E+00 B= 0 d= 0
x= -20.55 A= 1.85E+00 B= 7.1555 d= 0.448703
x= -19.45 A= 2.61E+00 B= 7.4162 d= 0.539137
x= -18.35 A= 3.28E+00 B= 7.57 d= 0.629572
x= -17.25 A= 3.88E+00 B= 7.57 d= 0.715006
x= -15.53 A= 5.40E+00 B= 7.5857 d= 1.275049
x= -13.8 A= 7.82E+00 B= 7.58 d= 3.255005
x= -12.075 A= 1.04E+01 B= 7.5851 d= 3.175004
x= -10.35 A= 1.28E+01 B= 7.58 d= 3.105004
x= -6.9 A= 1.61E+01 B= 7.57 d= 2.955002
x= -3.45 A= 1.76E+01 B= 7.59 d= 2.800001
x= 0 A= 1.71E+01 B= 7.59 d= 2.65
x= 3.45 A= 1.58E+01 B= 7.6 d= 2.499999
x= 6.9 A= 1.38E+01 B= 7.4916 d= 2.354998
x= 10.35 A= 1.04E+01 B= 6.4367 d= 2.199996
x= 12.075 A= 8.00E+00 B= 5.3209 d= 2.134996
x= 13.8 A= 5.24E+00 B= 3.672 d= 2.044995
x= 15.525 A= 2.86E+00 B= 1.9245 d= 1.959995
x= 17.25 A= 1.17E+00 B= 0.91 d= 1.794994

I:\CIRC\MSC\1\MSC.1-Circ.1652.docx
 MSC.1/Circ.1652
Annex, page 77

The fitting qualities of resistance and thrust coefficient, based on paragraphs 2.6.3.4.6
and 2.6.3.4.4 of the Interim Guidelines, are shown in figures 6.1 and 6.2, respectively.
800000
700000
sample point
600000
500000 approximation
400000
RT (N)
300000
200000
100000
0
-100000 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7

-200000
-300000
Fn

Figure 6.1 Ship resistance curve and its approximation using

the fifth power polynomial

0.25

0.2 sample point

approximation
0.15

0.1
KT

0.05

0
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8
-0.05

-0.1
J

Figure 6.2 Propeller thrust coefficient curve and its approximation using the
second power polynomial

6.2 This ship is judged as possibly vulnerable in the level 1 check, based on section 2.6.2
of the Interim Guidelines, because its service Froude number is larger than 0.3 and the ship
length is smaller than 200 m.

6.3 Thus, it is necessary to check the vulnerability with the level 2. Firstly, the critical
Froude number for surf-riding needs to be calculated under any initial condition using the
Melnikov analysis, based on paragraph 2.6.3.4.6 of the Interim Guidelines. Examples of the
calculated critical Froude numbers for typical local waves are as shown in table 6.2.

I:\CIRC\MSC\1\MSC.1-Circ.1652.docx
MSC.1/Circ.1652
Annex, page 78

Table 6.2 Example of critical Froude numbers for surf-riding under some wave cases

Critical Froude number with linear

wave celerity

λ/L=1.250, 0.3296
H/λ=0.0504
λ/L=1.500, 0.3563
H/λ=0.0396
λ/L=1.500, 0.3428
H/λ=0.0504
λ/L=1.500, 0.3325
H/λ=0.0600
λ/L=1.750, 0.3577
H/λ=0.0504

6.4 Secondly, the probability index C values are calculated as a function of the nominal
Froude number, Fn, based on paragraph 2.6.3.2 of the Interim Guidelines, as shown in
table 6.3.

Table 6.3 Probability index C values for some nominal Froude number

C with linear wave

Fn
celerity
0.30 7.88E-04
0.35 2.31E-02
0.40 5.91E-02
0.45 8.77E-02
0.50 9.19E-02

This means that this subject ship is judged as vulnerable for surf-riding/broaching stability
failure in the level 2 check because the C value at designed Froude number of 0.4 is larger
than the standard of 0.005.

I:\CIRC\MSC\1\MSC.1-Circ.1652.docx
 MSC.1/Circ.1652
Annex, page 79

APPENDIX 3

ELEMENTS FOR NUMERICAL MODELLING OF ROLL MOTION IN THE VULNERABILITY

CRITERIA OF THE SECOND GENERATION INTACT STABILITY CRITERIA

1 Equation of motion

1.1 This section on the equation of motion is provided because it is used or referenced in
more than one of the stability failure mode vulnerability criteria. For level 2 of the vulnerability
criteria for parametric roll, the user is expected to solve this equation numerically. To do so,
the elements of the equation, especially the coefficients of the terms require some discussion
to provide the user with appropriate assessment. Other failure modes use these elements in
their assessments.

1.2 The equation of motion takes into account forces acting on the ship. The simplest
mathematical model that is capable of evaluating the maximum roll angle includes four
moments:

.1 inertia, including added inertia (or added mass) as a part of hydrodynamic

forces;

.2 roll damping, which expresses energy loss from roll motions in creating
waves, vortices and skin friction;

.3 roll restoring (stiffness) is modelled with the GZ curve with relative wave
elevation taken into account only for parametric rolling and pure loss of
stability failure modes; and

.4 transverse wave forces are absent for a ship in exact following or head
long-crested seas.

1.3 The roll inertia of a ship as a solid body is measured by the transverse moment of
inertia Ixx. Inertial forces are proportional to accelerations. There are also hydrodynamic forces
acting on a ship, subject to accelerated motion that are also proportional to the accelerations.
These hydrodynamic forces are usually expressed as an additional mass or a moment of
inertia and referred to as "added mass" in roll A44. The roll inertia is expressed as:

𝑀𝐼𝑁 = (𝐼𝑥𝑥 + 𝐴44 ) ⋅ 𝑊𝜑 (1.1)

where Wφ is the angular acceleration in roll; calculation of the moment of inertia as well as
added inertia is provided in section 2.7.1 of the Interim Guidelines.

1.4 Damping of roll motions is essentially a transfer of kinetic energy of a moving ship to
the environment. It is a complex process because this energy transfer occurs through different
physical phenomena. Skin friction causes the layers of water nearest to the hull to move.
The moving surface of the hull leads to formation of vortices; the kinetic energy of the water
moving in those vortices is taken from the ship. Due to its motion, the ship also makes waves
on water surface that also dissipate energy. The complexity of these physical phenomena is
the reason why a model test is the most reliable source of information on roll damping.
However, recent developments in computational fluid dynamics (CFD) hold good promise for
the availability of this computational method in the future. In the absence of ship-specific or
prototype data, the simplified Ikeda method can be applied (see section 9.2). A moment of roll
damping is presented in the following form:

I:CIRC\MSC\1\MSC.1-Circ.1652.docx
MSC.1/Circ.1652
Annex, page 80

𝑀𝐷 = (𝐼𝑥𝑥 + 𝐴44 ) ⋅ (𝛿0 𝑉𝜑 + 𝛿1 𝑉𝜑 |𝑉𝜑 | + 𝛿2 𝑉𝜑3 ) (1.2)

where 0, 1 and 2 are coefficients computed with the simplified Ikeda method and Vφ is
the angular velocity of roll motions.

The simplified Ikeda method contains some empirical elements and, for this reason, the range
of its applicability should be observed.

1.5 Alternatively, numerical simulations can be used for the estimation of roll damping,
based on the solution of viscous hydrodynamic equations with the guidance of CFD for roll
damping. In this case, validation of simulations should be performed for selected loading
conditions to the satisfaction of Administration. Validation should be performed in comparison
with model tests performed according to the procedures in MSC.1/Circ.1200 or alternative test
procedures.

2 Equation of motion with respect to parametric rolling

2.1 Roll restoring

2.1.1 A proper representation of roll restoring is very important for the correct representation
of parametric rolling. The variation of stability in waves is a primary mechanism of development
of parametric rolling (an explanation is provided in chapter 4 of appendix 1). The calculation of
the instantaneous roll restoring, while straightforward if disturbance due to the ship hull is
ignored, may be too complex for the level 2 vulnerability check. Hence, a quasi-static approach
can be used instead.

2.2 Evaluation of metacentric height and righting lever curve in longitudinal waves

2.2.1 The quasi-static approach means that the GZ curve for the ship on a wave is
calculated using the hydrostatic algorithm (in which forces and moments are balanced in heave
and pitch), but the waterplane is not flat – it is determined from the intersection of a wave and
the hull surface. Known also as "wave-pass" calculations, the capability for this calculation is
provided by a number of commercially available hydrostatic software packages. For the
assessment of parametric rolling, calculation of the GZ curve up to 180 degrees is
recommended if the superstructures including down-flooding opening are properly modelled;
it sets a natural maximum and prevents the numerical solution from growing too large and
causing a numerical error. Alternatively, a maximum cut-off angle can be used. Exceedance
of the cut-off roll angle stops the calculation. Practical value of the cut-off roll angle may be set
equal to the largest available angle in the GZ curve calculation and is expected to be around 50
to 90 degrees. Figures 2.1a and 2.1b show the GZ variation in waves as a series of curves.
Each curve is calculated for a particular position of the wave crest relative to the midship which
results in a surface shown in figure 2.1c. For the intermediate values of heel angle and of the
wave crest position, a bilinear or bi-cubic spline interpolation can be used.

I:\CIRC\MSC\1\MSC.1-Circ.1652.docx
 MSC.1/Circ.1652
Annex, page 81

a)
1.5 GZ, m c)

1 GZ

0.5

Heel, deg
0 10 20 30 40 50

-0.5

b)
GZ, m
2

Heel, deg
0 20 40 80 100 120 140 160

-2
Angle of heel
Position of the wave
-4 crest relative to the
midship section
-6

-8

Figure 2.1 The GZ curve in waves (steepness 0.02, C11 class containership, full load):
a) positive range, b) full range and c) as a surface

2.2.2 The definition of wave crest position is illustrated in figure 2.2. The position of
the wave crest is a function of time:

𝑋𝐶 (𝑡) = 0.5𝜆 𝑠𝑖𝑛( 𝜔𝑒 𝑡) (2.1)

where,
𝜔2
e = the wave frequency of encounter = 𝜔𝑒 = 𝜔 − 𝑔
𝑉𝑆 𝑐𝑜𝑠 𝜓, and
𝜓 = is the relative wave heading (0 degrees – following waves, 180
degrees – head waves).

Thus, the value of the GZ curve in waves can be presented as a function of time and angle of
heel, φ:

𝐺𝑍 = 𝐺𝑍(𝑡, 𝜑) (2.2)
30

20 XC

10

-100 -50 0 50 100

Position of the wave crest relative to midship, m

Figure 2.2 Definition of the position of the wave crest relative to the midship section

I:\CIRC\MSC\1\MSC.1-Circ.1652.docx
MSC.1/Circ.1652
Annex, page 82

2.2.3 For the symmetric GZ curve, the total restoring moment is expressed as:

𝑀𝑅 = sign(𝜑) ⋅ 𝜌𝛻𝑔 ⋅ 𝐺𝑍(𝑡, |𝜑|) (2.3)

1 𝜑≥0
sign(𝜑) = {
−1 𝜑 < 0

For the asymmetric GZ curve, the calculations are to be done for starboard and portside
separately and the total restoring moment is expressed as:

𝑀𝑅 = 𝜌𝛻𝑔 ⋅ 𝐺𝑍𝐵 (𝑡, 𝜑) (2.4)

𝐺𝑍 (𝑡, 𝜑) 𝜑 ≥ 0
𝐺𝑍𝐵 (𝑡, 𝜑) = { 𝑆
𝐺𝑍𝑃 (𝑡, 𝜑) 𝜑 < 0

Where GZB(t,) = complete righting curve (m);

GZS(t,) = righting curve for starboard (m); and

GZP(t,) = righting curve for port side (m).

Restoring action of the righting lever curve is assumed to be described by a positive value
while rolling to the starboard and a negative value while rolling to the port side.

2.3 Assessment of the equation of motion in terms of inertia

2.3.1 Following Newton's second law, the equation of roll motion is expressed as the inertial
force equal to the sum of all other forces. Since the ship is in longitudinal waves, there is
negligible or no direct forcing that comes from the waves:

𝑀𝐼𝑁 = −𝑀𝐷 − 𝑀𝑅 (2.5)

2.3.2 In the equation in the previous paragraph, the negative sign is inserted because both
damping and restoring forces are directed against the roll motion or the rate of motion.
The equation of roll motion can be rewritten with each force as a function of motion parameters
or time:

𝑀𝐼𝑁 (𝑊𝜑 = 𝜑̈ ) + 𝑀𝐷 (𝑉𝜑 = 𝜑̇ ) + 𝑀𝑅 (𝑡, 𝜑) = 0 (2.6)

This equation relates the roll motion with the roll rate and the angular roll acceleration. These
quantities are related through differentiation: the angular velocity is a derivative of roll and
the angular acceleration is a derivative of angular velocity. Thus, this equation is a differential
equation. The solution of this differential equation is a time history of roll motions, similar to
that shown in figure 2.3, which shows parametric rolling. As the ship is sailing in longitudinal
waves, there is no forcing in the transversal plane, so the observed rolling motion in periodic
waves is usually a result of parametric resonance if the roll period is twice the encounter wave
period.

I:\CIRC\MSC\1\MSC.1-Circ.1652.docx
 MSC.1/Circ.1652
Annex, page 83

Figure 2.3 Time history of parametric rolling from the initial state (left) and steady
states of roll angle during two encounter wave cycles (right) as functions of the
relative longitudinal ship position to a wave trough, which is normalized with the
wavelength.

2.3.3 The equation in the previous paragraph can be solved with an appropriate numerical
method. For this purpose, the equation is presented in a form of a vector-valued equation:

𝜑̇ 𝑉𝜑
( ̇ )=( 1 ) (2.7)
𝑉𝜑 (−𝑀𝐷 (𝑉𝜑 ) − 𝑀𝑅 (𝑡, 𝜑))
𝐼𝑋 +𝐴44

If 𝑉𝜑 and 𝜑 at the specified time, t, the 𝑉𝜑 and 𝜑 at the next time step can be determined. Thus,
the time series starting from an initial state can be determined.

2.3.4 The vector-valued equation in paragraph 2.3.3 can be integrated with a reliable
algorithm such as the fourth-order Runge-Kutta method with the initial conditions for the
calculations, i.e. values of roll angle and roll rate at the beginning (or at time step t = 0). The
solution, as illustrated in figure 2.3, was computed with assumed initial conditions (φ = 5
degrees and Vφ = 0 degrees/s). While the calculation can assume zero for both φ and Vφ,
the development of parametric rolling may not occur until a much longer duration is calculated.

2.3.5 To complete the inputs necessary for the calculation, two more parameters are
needed: the time step t and the total simulation time Tsim. The time step t should be
sufficiently small to achieve numerical convergence, and the total simulation time Tsim should
be sufficiently long to achieve a steady state or a capsize.

3 Information regarding level 2 vulnerability criterion for the dead ship condition

3.1 Roll modelling in beam waves and wind

3.1.1 Roll in the dead ship condition is represented by the uncoupled roll model in beam
waves and wind. Roll motion is due to the forcing of wind and waves. The wave exciting
moment is proportional to the effective wave slope, which is the product of the wave slope and
the effective wave slope coefficient at each wave frequency. The wind moment is due to the
time-independent mean wind speed and the time-varying component of the wind velocity, i.e.
the gust. The time dependent wind exciting moment is related to the time-varying component
of the wind velocity. Since the relative roll angle is responsible for vanishing stability and down
flooding in beam waves, the Interim Guidelines use the effective relative roll angle in its
paragraph 2.2.3.2.3.

I:\CIRC\MSC\1\MSC.1-Circ.1652.docx
MSC.1/Circ.1652
Annex, page 84

3.1.2 The effective wave slope angle can be defined by the product of the wave slope
angle and the effective wave slope function. The wave slope angle can be calculated by the
product of the wave number and the wave elevation. Thus, the wave slope spectrum shall be
obtained by the product of the wave energy spectrum and the square of the wave number as
shown in paragraph 2.2.3.2.3 of the Interim Guidelines.

3.1.3 The wind velocity consists of the constant component and the time-varying component.
The wind-induced roll moment is proportional to the square of the wind velocity. Since the
time-varying component is usually much smaller than the constant component, the time-
varying roll moment is proportional to the product of the constant wind velocity and the time-
varying component of the wind velocity. Thus, the wind loading spectrum can be obtained by
the product of the wind velocity power spectrum and the square of the mean wind velocity as
shown in paragraph 2.2.3.2.3 of the Interim Guidelines.

3.2 Roll damping, natural roll frequency and the effective wave slope function

3.2.1 Paragraph 2.2.3.2.3 of the Interim Guidelines requires the stochastic linearization
method for equivalent roll damping, which is described in paragraph 9.3.5. In this method, it is
necessary to calculate the standard deviation of the roll angular velocity. Since the roll angular
velocity is the derivative of the roll angle with respect to time, the power spectrum of the roll
angular velocity is the product of the roll power spectrum and the square of the circular
frequency. Thus, the standard deviation of the roll angular velocity can be evaluated as the
area of power spectrum of the roll angular velocity.

3.2.2 The spectrum of roll angular velocity, 𝑆𝑑𝜙 (𝜔) ((rad/s)2/(rad/s)), which is necessary
for the determination of the equivalent linear roll damping coefficient 𝜇𝑒 (1/s), is calculated as
follows:
𝑆 (𝜔)
𝑆𝑑𝜙 (𝜔) = 𝜔2 ∙ 𝐻2 (𝜔) ∙ (𝜌𝑔∇𝐺𝑀)
𝑀
2 (3.1)

where
𝑆𝑀 (𝜔) = 𝑆𝑀𝑤𝑎𝑣𝑒𝑠 (𝜔) + 𝑆𝛿𝑀𝑤𝑖𝑛𝑑,𝑡𝑜𝑡 (𝜔) (3.2)
𝑆𝑀𝑤𝑎𝑣𝑒𝑠 (𝜔) = (𝜌𝑔∇𝐺𝑀𝑟𝑒𝑠 )2 ∙ 𝑟 2 (𝜔) ∙ 𝑆𝛼𝛼 (𝜔) (3.3)

In equation (3.1), 𝑆𝑀 (𝜔) ((N m)2/(rad/s)) is the spectrum of total roll moment due to waves and
wind gustiness, while the other quantities appearing in equation (3.1) are defined in
section 2.2.3 of the Interim Guidelines.

3.2.3 The natural roll frequency can be determined mainly with the slope of the righting
arm with respect to the heel angle. In case of the ship roll motion around the upright condition,
the slope is the metacentric height. In case of the ship with beam wind, it should be corrected
with the heel angle due to the constant wind velocity as shown in the paragraph 2.2.3.2.3 of
the Interim Guidelines.

3.2.4 The effective wave slope function for a conventional mono-hull vessel can be
evaluated as described in section 8.2. For multi-hull vessels or ships having very short natural
roll periods, alternative methods described in section 8.1 are available.

I:\CIRC\MSC\1\MSC.1-Circ.1652.docx
 MSC.1/Circ.1652
Annex, page 85

4 Calculation of maximum roll angle for assessment of parametric rolling in check

2 of level 2 of the vulnerability criteria

4.1 The numerical simulation for the parametric roll response should start with the initial
condition specified in paragraph 2.3.4. The simulated response could resolve to either the
upright condition, a periodic steady-state, a non-periodic oscillatory behaviour or the
exceedance of 180 degrees. If parametric rolling is not possible for the given wave conditions,
the response is represented by decaying roll oscillations, as shown in figure 4.1. In this case,
the maximum roll angle occurs during the transient stage and it could depend on the initial
wave phase. Thus, it should be ignored. Thus, the maximum roll angle here should be regarded
as zero. When the wave steepness is zero, a simple roll decay curve is obtained so that the
maximum roll angle should be regarded as zero in place of the initial roll angle of 5 degrees.

wave steepness= 0.02

60

40
roll angle [deg.]

20

0
0 200 400 600 800 1000
-20

-40

-60
time [sec.]

Figure 4.1 An example of roll response in absence of parametric rolling

4.2 A typical example of the simulated periodic response is shown in figure 4.2. In this
case, the maxim transient roll angle is about 47.3 degrees but the steady-state amplitude is
about 43.6 degrees. Since the transient behaviour depends on the initial wave phase, the
maximum roll angle should be determined only from the steady-state. The steady-state can be
identified if the difference of the roll angle at every two wavelengths converges to a negligibly
small value. If the steady-state roll angles during the two encounter wave cycle are plotted as
a function of the relative longitudinal ship position to a wave trough, in other words, the phase
difference to waves, figure 4.3 can be drawn. Here, the two lines can be found. This means
that one roll cycle occurs when the ship meets two encounter wave cycles. Thus, this rolling
behaviour is the parametric rolling that the level 1 criterion deals with.
wave steepness= 0.06
60

40
roll angle [deg.]

20

0
0 200 400 600 800
-20

-40

-60
time [sec.]

Figure 4.2 An example of periodic roll response with the overshoot

I:\CIRC\MSC\1\MSC.1-Circ.1652.docx
MSC.1/Circ.1652
Annex, page 86

wave steepness= 0.06

60

40
roll angle [deg.]
20

0
0 0.2 0.4 0.6 0.8 1.0
-20

-40

-60
relative ship position normalised by wavelength

Figure 4.3 The periodic steady-state roll during one wavelength

4.3 If we slightly increase the wave steepness in the above case, the symmetry of steady
roll motion breaks as shown in figure 4.4. The absolute value of the maximum roll angle is
smaller than that of the minimum roll angle. This is one of the typical non-linear phenomena.
In the case of asymmetric roll, the average of two absolute values should be used as the
maximum roll angle to be used for the calculation of the C2 value.

Figure 4.4 An example of asymmetric periodic roll response

4.4 Another type of periodic steady roll response is shown in figure 4.5. This roll cycle
occurs within one wave encounter cycle so that this is not the parametric roll that the level 1
criterion deals with. For avoiding the inconsistency between the two levels, this response
should be dealt with outside the framework of the multi-layered criteria structure.

wave steepness= 0.06

60

40
roll angle [deg.]

20

0
0 0.2 0.4 0.6 0.8 1.0
-20

-40

-60
relative ship position normarised by wavelength

I:\CIRC\MSC\1\MSC.1-Circ.1652.docx
 MSC.1/Circ.1652
Annex, page 87

Figure 4.5 An example of another type of period roll response

4.5 More complicated response sometimes occurs as shown in figure 4.6. In this case,
the roll motion seems to consist of one larger roll cycle together with two smaller roll cycles.
However, it is not a simple repeat. If we draw a phase portrait as figure 4.7, where the abscissa
is the roll angle and the ordinate is the roll angular velocity, it can be confirmed this is not a
periodic motion. This response should be dealt with outside the framework of the multi-layered
criteria structure.

wave steepness= 0.06

60

40
roll angle [deg.]

20

0
0 200 400 600 800 1000 1200
-20

-40

-60
time [sec.]

Figure 4.6 An example of non-periodic roll response

Figure 4.7 The phase trajectory of non-periodic roll response

4.6 When the calculated roll angle leaves the entire range of the given GZ curve,
such as ±180 degrees, the calculation should be explicitly stopped as the occurrence of
capsizing. The obtained roll behaviour is just transient so that the maximum roll angle cannot
be determined. Its example is shown in figure 4.8. The periodic parametric rolling occurs with
the wave steepness of 0.08 and the capsizing occurs if the wave steepness is slightly
increased.

I:\CIRC\MSC\1\MSC.1-Circ.1652.docx
MSC.1/Circ.1652
Annex, page 88

wave steepness= 0.09

180

120
roll angle [deg.]

60

0
0 20 40 60 80
-60

-120

-180
time [sec.]

Figure 4.8 Examples of roll response with parametric rolling and capsizing

5 Supplementing information on calculation for checking vulnerability to

parametric rolling

5.1 For the calculation of the bilge keel area to be used in paragraph 2.5.2.1 of the Interim
Guidelines, the fin stabilizers may be regarded as a kind of bilge keel if they are non-retractable
and calculated as a part of total bilge keel area. On the other hand, the centre skeg should not
be included unless the midship section coefficient is less than 0.9. This is because existing
model experiments indicate that the increase of the roll damping due to a centre skeg is almost
zero for the case of small rise of floor. This may be because the pressure created by the skeg
on the bottom works as a negative roll damping.

5.2 The "sharp bilge" in paragraph 2.5.2.1 of the Interim Guidelines means that bilge
radius is smaller than 1% of the ship's breadth and the angle between piece-wise lines
representing the bilge is smaller than 120° as shown in figure 5.1.

I:\CIRC\MSC\1\MSC.1-Circ.1652.docx
 MSC.1/Circ.1652
Annex, page 89

water line

angle between piece-

wise lines representing
the bilge

Figure 5.1 Definition of the angle between piece-wise lines representing the bilge

a. The value of sw in paragraph 2.5.2.2 of the Interim Guidelines can be corrected if

operational limitations are implemented. In this case, a wave scatter table should be provided
that is consistent with the implemented operational limitations, in accordance with section 2.7.2
of the Interim Guidelines, and sW can be determined as specified in section 10.1.

5.4 For the calculation in paragraph 2.5.3.2.1 of the Interim Guidelines, a sinusoidal wave
should be used without hydrodynamic disturbance due to the ship. The water pressure due to
the wave may include the effect of wave particle velocity assuming that water depth is larger
than the wavelength or a hydrostatic pressure profile starting from the actual free surface
should be used.

6 Discussion on the relationship between level 1 and 2 vulnerability criteria for

parametric rolling

6.1 The vulnerability level 1 criterion and the first check of level 2 criteria represent a
simplification of the second check of level 2 criteria. By simplifying equation (2.7) and
examining steady solutions, the occurrence condition for parametric roll is obtained as follows:

𝛿𝐺𝑀1 2𝛿0
> (6.1)
𝐺𝑀 𝜔𝑟

The right-hand side of this equation indicates the normalized roll damping coefficient multiplied
by 2. These vulnerability criteria apply the empirical formulae for the equivalent roll damping
as the function of bilge keel area and the midship coefficient, which was developed using
Ikeda's simplified estimation method. This outcome is used for the vulnerability level 1 criterion
and the first check of the level 2 criteria. Therefore, these requirements are a simplification of
the second check of the level 2 criteria.

6.2 In the level 1 criterion a wavelength equal to the ship length is considered as a typical
worst-case condition. In the level 2 first check criterion, wavelengths and corresponding wave
steepnesses of the calculation "wave cases" are determined from the considered wave scatter
table. The steepness used in the level 1 criterion is defined as the maximum wave steepness
among those in the wave cases of the level 2 first check criterion. The procedure for the
determination of the wave cases from the relevant wave scatter table is described in section
10 of this appendix. For unrestricted service, the relevant wave scatter table is table 2.7.2.1.2
of the Interim Guidelines. The wave scatter diagram can be different in case operational
limitations are implemented. In that case, wave cases for the level 2 first check criterion and
the steepness in the level 1 criterion are modified following the procedure in section 10.

I:\CIRC\MSC\1\MSC.1-Circ.1652.docx
MSC.1/Circ.1652
Annex, page 90

6.3 The difference between the level 1 criterion and the first check of the level 2 criteria
are as follows. While the level 2 uses 16 different reference waves for calculating
the metacentric height variation, the level 1 uses the wave having the worst wavelength, i.e.
the wavelength to ship length ratio of 1 and a wave steepness which is the maximum among
those of the wave cases used for level 2 first check. Thus, the level 1 criterion is expected to
be more conservative than the first check of the level 2 criteria.
6.4 While both the level 1 criterion and the first check of the level 2 criteria judge the
danger of parametric roll with its occurrence condition, the second check of level 2 criterion
judges with the magnitude of parametric roll. This means that, if parametric roll occurs with
small amplitude, the second check of level 2 criteria could conclude that the ship is not
vulnerable to stability failure due to parametric roll. Thus, the second check of level 2 criteria
is less conservative in principle.

7 Method to establish equivalence between regular and irregular waves as

provided in vulnerability criteria, level 2 for both pure loss of stability and
parametric roll stability failure modes (Grim's effective wave approach) to
assess the change of stability in longitudinal irregular waves
7.1 Although actual ocean waves are irregular, it is practical to simplify the irregular waves
as a regular wave. Since the restoring variation due to waves has the non-linear relationship
with the incident waves, the response amplitude operator approach used for the dead ship
condition and excessive acceleration stability failure modes cannot be applied to the pure loss
and parametric rolling failure modes. To overcome this situation, one traditional approach was
developed in Germany, which is known as Grim's effective wave (see figure 7.1).

LBP
heff

Figure 7.1 A concept of Grim's effective wave

7.2 This approach is based on the concept that an irregular wave spatial profile around
the ship can be substituted by a longitudinal regular wave with the wavelength equal to the
ship length and with the wave crest or trough situated at the longitudinal centre of gravity. If we
calculate the righting arm GZ with this regular wave, the relationship between GZ and its wave
height can be regarded as non-linear but non-frequency-dependence so that stochastic
estimation of GZ in irregular waves is possible. A comparison of the application of effective
wave concept with direct stability calculations has well verified this approach as shown in
figure 7.2.

Figure 7.2 Verification of Grim's effective wave by using a direct calculation of the GZ
in longitudinal waves with the wave crest amidships and a heel angle of 30 degrees and
the wave height of 5% L for a trawler6

6
Umeda, N. and Yamakoshi, Y. Probability of Ship Capsizing due to Pure Loss of Stability in Quartering Seas.
Naval Architecture and Ocean Engineering, The Society of Naval Architects of Japan, Vol. 30, pp. 73–85, 1994.

I:\CIRC\MSC\1\MSC.1-Circ.1652.docx
 MSC.1/Circ.1652
Annex, page 91

7.3 In the level 2 vulnerability criteria for pure loss of stability, for determining the indices
for each Hi in paragraph 2.4.3.2.2 of the Interim Guidelines, the relationship between h in
paragraph 2.4.3.2.1 of the Interim Guidelines and the index should be obtained by calculation.
Here, if the index has a peak at the certain wave height hp, the peak value of the index should
be used when the wave height is larger than hp. When local peaks are present, this
conservative approach is used with reference to each local peak up to a wave height where
direct calculations are more conservative.
7.4 For the level 2 vulnerability criteria for pure loss of stability, the 3% largest effective
wave height, Hi, in paragraph 2.4.3.2.3 of the Interim Guidelines should be calculated as
follows:

𝐻𝑖 = 5.9725 √𝑚0 (7.1)

2 2
𝜔2 𝐿 𝜔2 𝐿 𝜔2 𝐿 𝜔2 𝐿
𝜔𝐿 𝑠𝑖𝑛( ) 3𝜔 𝑠𝑖𝑛( )
𝑔 2𝑔 𝑔 2𝑔
𝑚0 = ∫0.01𝜔 { 2 } 𝑆𝑍𝑍 (𝜔) 𝑑𝜔 + ∫𝜔 𝐿 { 2 } 𝑆𝑍𝑍 (𝜔)𝑑𝜔]
𝐿 𝜔2 𝐿 𝐿 𝜔2 𝐿
𝜋 2 −( ) 𝜋 2−( )
2𝑔 2𝑔

(7.2)
where

2𝑔𝜋
𝜔𝐿 = √ (7.3)
𝐿

If Hi > 0.1L, Hi should be set as Hi =0.1L.

7.5 In the level 2 vulnerability criteria for parametric rolling, for determining the maximum
roll angle, for each Hri in paragraph 2.5.3.4.2 of the Interim Guidelines, the relationship
between hj in paragraph 2.5.3.4 of the Interim Guidelines and the maximum roll angle should
be obtained by calculation. Here, if the maximum roll angle has a peak at the certain wave
height hp, the peak value of the maximum roll angle should be used when the wave height is
larger than hp. When local peaks are present, this conservative approach is used with reference
to each local peak up to a wave height where direct calculations are more conservative.
7.6 For the level 2 vulnerability criteria for parametric rolling, the representative wave
height, Hri, in paragraph 2.5.3.4.2 of the Interim Guidelines, which corresponds to the 1/3
largest effective wave height, should be calculated as follows:

𝐻𝑟𝑖 = 4.0043√𝑚0 (7.4)

If Hri > 0.1L, Hri should be set as Hri =0.1L.

8 Determination of roll moment due to waves and effective wave slope function

8.1 Methods for evaluating the effective wave slope function

8.1.1 The use of uncoupled roll model for beam waves can be justified from the viewpoint
of coupled ship dynamics but only when the wavelength is sufficiently longer than the ship
breadth and the wave exciting roll moment is calculated by integrating incident wave pressure
on its own as described in paragraph 3.5.4 of the explanatory notes to the 2008 Intact Stability
Code, MSC.1/Circ. 1281. The amplitude of wave exciting roll moment is conveniently
expressed as a linear slope of a wave across the transverse sections of the ship. There are a
number of methods that can be applied to determine this function, r(ω):

I:\CIRC\MSC\1\MSC.1-Circ.1652.docx
MSC.1/Circ.1652
Annex, page 92

.1 a linear strip theory of coupled sway-yaw-roll motions with the non-linear roll
damping in regular beam waves, in which the amplitude of roll motion under
the synchronous condition, 𝜑𝑚 , in radians, can be determined. By using the
roll damping moment, the amplitude of the wave exciting roll moment for the
uncoupled roll motion can be calculated. Then the effective wave slope
coefficient can be determined by equation (8.1.1).
2 +𝑐∙𝜑 3 )
2∙(𝑎∙𝜑𝑚 +𝑏∙𝜑𝑚 𝑚
𝑟(𝜔𝑟 ) = 𝜋 2∙𝑠
(8.1.1)

where a, b and c are the Froude extinction coefficients defined in paragraph 9.3.4
of appendix 3. The obtained value of r can be used for all frequencies, in which
the ratio of wavelength to ship breadth is larger than 0.5: r(ω) should be
assumed to be 0 for ratios of wavelength to the ship moulded breadth less than
0.5. Alternatively, the above procedure can be applied to the case where the
synchronous roll amplitude of coupled motions can be determined with a model
test in regular beam waves according to MSC.1/Circ.1200.

.2 a station simplified method in which the sectional area curve along the ship
length is kept constant, but the geometry of each station is assumed to be
rectangular. This allows the calculation of the distribution of incident wave
pressure caused by the wave elevation across the transverse section to be
performed analytically. The value of r(ω) should be evaluated at the natural
roll frequency for the wavelength to ship's breadth ratio of 0.5 or larger; r(ω)
should be assumed to be 0 for ratios of wavelength to the ship moulded
breadth less than 0.5.

.3 a direct pressure integration of the incident wave pressure over the

submerged hull surface up to the level of the incident wave profile.

.4 a direct pressure integration method of the incident wave pressure over the
submerged hull surface up to the level of mean waterline without the ship
motions.

8.1.2 For the latter two methods, r(ω) should be calculated using the following equation:

𝑀
𝐹𝐾 (𝜔)
𝑟(𝜔) = 𝑊⋅𝐺𝑀⋅𝛼 (8.1.2)
0 (𝜔)

where

MFK = the amplitude of the Froude-Krylov roll moment, calculated by

integrating wave pressures on the ship hull;

W = the weight of displacement;

GM = the metacentric height without free surface correction; and

0 = is the amplitude of the angle of wave slope at the centreline.

8.1.3 Since the vulnerability criteria is required to be applied with minimal computational
efforts, the method described in 8.1.1.2 is recommended as standard method. It requires only
sectional breadth, sectional draught and sectional area as well as the longitudinal position of
the transverse sections and the vertical height of ship gravitational centre.

I:\CIRC\MSC\1\MSC.1-Circ.1652.docx
 MSC.1/Circ.1652
Annex, page 93

8.1.4 The standard method for the estimation of the effective wave slope is applicable only
to standard mono-hull vessels. For a ship which does not fall in this category, other prediction
methods described in 8.1.1.1, 8.1.1.3 and 8.1.1.4 should be applied. For a ship having very
short natural roll period, the method described in 8.1.1.1 is desirable.

8.2 Standard methodology for the estimation of the effective wave slope function

8.2.1 The standard methodology for the estimation of the wave slope function is based on
the following assumptions and approximations:

.1 The underwater part of each transverse section of the ship is substituted by

an "equivalent underwater section" having, in general, the same breadth at
waterline and the same underwater area of the original section; however:

.1 sections having zero breadth at waterline, such as those in

the region of the bulbous bow, are neglected; and

.2 the draught of the "equivalent underwater section" is limited to

the ship sectional draught.

.2 The effective wave slope coefficient for each wave frequency is determined
by using the "equivalent underwater sections" considering only
the undisturbed linear wave pressure.

.3 For each section, a formula is applied which is exact for rectangles.

8.2.2 For each longitudinal position 𝑥 along the vessel, the draught 𝑇𝑒𝑞 (𝑥 ) (m),
the breadth 𝐵𝑒𝑞 (𝑥 ) (m) and the underwater sectional area 𝐴𝑒𝑞 (𝑥 ) (m2) of the "equivalent
vessel" are to be calculated as follows:

𝐴𝑒𝑞 (𝑥) = 𝐴(𝑥)

𝐴(𝑥) 𝐵𝑒𝑞 (𝑥) = 𝐵(𝑥)
if 𝐵(𝑥)
≤ 𝑇(𝑥) then
𝐴(𝑥)
( )
{𝑇𝑒𝑞 𝑥 = 𝐵(𝑥)
if 𝐴(𝑥) > 0 and 𝐵(𝑥) > 0:

𝑇𝑒𝑞 (𝑥) = 𝑇(𝑥) (8.2.1

𝐴(𝑥) )
if > 𝑇(𝑥) then {𝐵𝑒𝑞 (𝑥) = 𝐵(𝑥)
𝐵(𝑥)
{ 𝐴𝑒𝑞 (𝑥) = 𝐵𝑒𝑞 (𝑥) ⋅ 𝑇𝑒𝑞 (𝑥)

𝐴𝑒𝑞 (𝑥) = 0
otherwise: {𝐵𝑒𝑞 (𝑥) = 0
{ 𝑇𝑒𝑞 (𝑥) = 0

where 𝐴(𝑥 ) , 𝐵(𝑥 ) and 𝑇(𝑥 ) are, respectively, the underwater sectional area,
the sectional breadth at waterline and the sectional draught of the ship.

8.2.3 The underwater volume 𝛻𝑒𝑞 (m3), the transverse metacentric radius 𝐵𝑀𝑇,𝑒𝑞 (m),
the vertical position of the centre of buoyancy 𝐾𝐵𝑒𝑞 (m) and the vertical position of centre of
gravity 𝐾𝐺𝑒𝑞 (m) of the "equivalent vessel" are to be calculated as follows:

I:\CIRC\MSC\1\MSC.1-Circ.1652.docx
MSC.1/Circ.1652
Annex, page 94
𝑥𝐹𝐸
𝛻𝑒𝑞 = ∫ 𝐴𝑒𝑞 (𝑥)𝑑𝑥
𝑥𝐴𝐸
𝑥𝐹𝐸
1 1
𝐵𝑀𝑇,𝑒𝑞 = ⋅∫ ⋅ 𝐵3𝑒𝑞 (𝑥)𝑑𝑥
𝛻𝑒𝑞 𝑥𝐴𝐸 12
𝑥𝐹𝐸
(8.2.2)
1 −𝑇𝑒𝑞 (𝑥)
𝐾𝐵𝑒𝑞 = 𝑇 + ⋅∫ ⋅ 𝐴𝑒𝑞 (𝑥)𝑑𝑥
𝛻𝑒𝑞 𝑥𝐴𝐸 2

{𝐾𝐺𝑒𝑞 = 𝐾𝐵𝑒𝑞 + 𝐵𝑀𝑇,𝑒𝑞 − 𝐺𝑀

where 𝐺𝑀 is the upright metacentric height (m), 𝑇 is the draught amidships (m), and xAE and
xFE correspond to the longitudinal coordinates of the aft end and forward end of the ship,
respectively. The vertical positions 𝐾𝐵𝑒𝑞 and 𝐾𝐺𝑒𝑞 are defined with respect to the base line of
the ship.
𝑥𝐹𝐸
∫𝑥 𝐶(𝑥)𝑑𝑥
𝐴𝐸
𝑟(𝜔) = | |
𝛻𝑒𝑞 ⋅ 𝐺𝑀
where
0 if 𝐴𝑒𝑞 (𝑥) = 0 and 𝐵𝑒𝑞 (𝑥) = 0
𝐶(𝑥) = {
𝐴𝑒𝑞 (𝑥) ⋅ [𝐾1 (𝑥) + 𝐾2 (𝑥) + 𝐹1 (𝑥) ⋅ 𝑂𝐺𝑒𝑞 ] otherwise
and
𝑘𝑤 = 𝜔2 /𝑔 ; 𝑂𝐺𝑒𝑞 = 𝐾𝐺𝑒𝑞 − 𝑇
𝐵𝑒𝑞 (𝑥)
𝑠𝑖𝑛 (𝑘𝑤 ⋅ ) (1 + 𝑘𝑤 ⋅ 𝑇𝑒𝑞 (𝑥)) ⋅ 𝑒−𝑘𝑤⋅𝑇𝑒𝑞 (𝑥) − 1
2
𝐾1 (𝑥) = ⋅
𝑘𝑤 ⋅ 𝐵𝑒𝑞 (𝑥) 𝑘2𝑤 ⋅ 𝑇𝑒𝑞 (8.2.3)
( )
2
𝐵𝑒𝑞 (𝑥)
𝑒−𝑘𝑤 ⋅𝑇𝑒𝑞 (𝑥) 𝐵𝑒𝑞 (𝑥) 𝑠𝑖𝑛 (𝑘𝑤 ⋅ )
2
𝐾2 (𝑥) = − ⋅ [𝑐𝑜𝑠 (𝑘𝑤 ⋅ )− ]
𝑘2𝑤 ⋅ 𝑇𝑒𝑞 (𝑥) 2 𝑘𝑤 ⋅ 𝐵𝑒𝑞 (𝑥)
( )
2
𝐵𝑒𝑞 (𝑥)
1 − 𝑒−𝑘𝑤 ⋅𝑇𝑒𝑞 (𝑥) 𝑠𝑖𝑛 (𝑘𝑤 ⋅ 2 )
𝐹1 (𝑥) = − ⋅
𝑘𝑤 ⋅ 𝑇𝑒𝑞 (𝑥) 𝑘𝑤 ⋅ 𝐵𝑒𝑞 (𝑥)
( )
{ 2

8.3 Direct calculation of the Froude-Krylov roll moment used in the level 2 criterion
for excessive acceleration failure mode

8.3.1 The parameters a and b in paragraph 2.3.3.2.2 of the Interim Guidelines are the cosine
and sine components, respectively, of the Froude-Krylov roll moment in regular beam waves
of unit amplitude (kN·m/m).

8.3.2 The parameters a and b can be calculated by integration of the undisturbed linear
wave pressure over the mean wetted hull surface of the ship, SH, with the following formulae:
1 1
𝑎= 𝜌𝑔 ∬𝑆 𝑒 𝑘𝑤𝑧 𝑐𝑜𝑠(𝑘𝑤 𝑦) 𝑛4 𝑑𝑆 , 𝑏 = − 𝜌𝑔 ∬𝑆 𝑒 𝑘𝑤𝑧 𝑠𝑖𝑛(𝑘𝑤 𝑦) 𝑛4 𝑑𝑆 (8.3.1)
1000 𝐻 1000 𝐻

I:\CIRC\MSC\1\MSC.1-Circ.1652.docx
 MSC.1/Circ.1652
Annex, page 95

where g is the acceleration due to gravity (m/s2), ρ is the density of sea water (kg/m 3), kw=ω2/g
is the wave number (rad/m), (x, y, z) are the coordinates of a generic point over SH (m3), with
the x-axis directed longitudinally from stern to bow, the y-axis directed transversally from
starboard to port side and the z-axis directed upwards. The quantity n4 (m) is defined as follows:

𝑛4 = (𝑧 − 𝑧𝐺 ) ⋅ 𝑛𝑦 − (𝑦 − 𝑦𝐺 ) ⋅ 𝑛𝑧 (8.3.2)

where yG (m) and zG (m) are, respectively, the transversal and vertical coordinates of the centre
of gravity (without correction for free surface), ny (-) and nz (-) are, respectively, the transversal
and vertical components of the unit normal vector to the hull pointing towards the fluid.

8.3.3 For laterally symmetric ship hulls, the parameters a and b can also be determined as
follows:
1 𝜔2
𝑎 = 0, 𝑏 = 1000 𝑚 ⋅ 𝑔 ⋅ 𝐺𝑀 ⋅ 𝑟 ⋅ 𝑔
(8.3.3)

where m (kg) is the ship mass, GM (m) is the metacentric height, and r (-) is the effective wave
slope function calculated according to the linear Froude-Krylov approach.

9 Estimation of roll damping

9.1 General

9.1.1 There are a number of ways by which roll damping data can be obtained for the
purpose of determining damping coefficients for use in the level 2 vulnerability criteria
assessments. These methods include, but are not limited to, model experiments, the ITTC
recommended procedure (7.5-02-07-04.5) approved in 2021 or amended, and the simplified
Ikeda method. The procedure for determining roll damping coefficients is presented in the
following sections.

9.2 The simplified Ikeda method 7

9.2.1 General considerations

9.2.1.1 The equivalent linear roll damping coefficient, B44( 𝜑𝑎 ) as a function of the roll
amplitude, 𝜑𝑎 , can be obtained by the following prediction method. Here, the roll damping
coefficient B44 and the circular roll frequency 𝜔 = 2𝜋 ⁄𝑇𝑟 are non-dimensionalized using the
following equations:

𝐵44 𝐵 𝐵
𝐵̂44 = 𝜌𝛻𝐵 2 √2𝑔 𝜔
̂ = 𝜔√2𝑔 (9.2.1)

The prediction method used here separates the roll damping into five damping components:
the skin-friction damping component, BF, the wave-making damping component, BW, the
eddy-making damping, BE, the bilge keel damping component, BBK, and the lift component, BL,
which is added for forward speeds. 8 This allows the roll damping coefficient B44 to be

7
Kawahara, Y., Maekawa, K., Ikeda, Y. A Simple Prediction Formula of Roll Damping of Conventional Cargo
Ships on the Basis of Ikeda's Method and Its Limitation. Chapter 26 of Contemporary Ideas on Ship Stability
and Capsizing in Waves, Neves, M.A.S., Belenky V.L., de Kat, J.O., Spyrou, K. and Umeda, N., eds.,
Springer, ISBN 978-94-007-1481-6, pp. 465-486, 2011.
8
If the simplified Ikeda method is used for individual components as opposed to the total roll damping
coefficient, due caution should be exercised and validation should be carried out on the basis of experimental
data.

I:\CIRC\MSC\1\MSC.1-Circ.1652.docx
MSC.1/Circ.1652
Annex, page 96

expressed, in dimensional form, as follows:

𝐵44 = 𝐵𝐹 + 𝐵𝑊 + 𝐵𝐸 + 𝐵𝐿 + 𝐵𝐵𝐾 (9.2.2).

9.2.1.2 For each component, a prediction formula was developed based on hydrodynamics
together with the parameters adjusted with systematic model experimental results and
validated with many merchant ships. As a result, the obtained formula uses hull geometry
offset and bilge keel details. Then, for simplicity, regression analyses for the systematically
calculated results of the established prediction method were executed. As a result, a simplified
prediction method, which only uses the ship principal particulars and the bilge keel dimensions,
was obtained. Users are cautioned that all digits of coefficients in the formulae must be used.

9.2.1.3 The formulae for the respective components in 9.2.2, 9.2.3, 9.2.4, 9.2.5, or 9.2.6 are
applicable for ranges of certain vessel parameters, which derive from the regression analysis
on which they are based. Because of the nature of the regression analysis, if a ship's
parameter exists outside its applicable range, the parameter value should be kept at the
corresponding maximum or minimum limit value for the use of the associated formula.

The parameters CB, B/d, OG/d, and Cm have the associated application limits for the BW, BE,
and BBK damping components:

0.5 ≤ CB ≤ 0.85; 2.5 ≤ B/d ≤ 4.5; -1.5 ≤ OG/d ≤ 0.2; and 0.9 ≤ Cm ≤ 0.99.
(9.2
.3)

̂ ≤ 1.0 is applicable only for the BW damping component.

𝜔

For the BBK damping component, additional application limits apply:

0.01 ≤ bBK/B ≤ 0.06 and 0.05 ≤ lBK/LBP ≤ 0.4,

where bBK and lBK indicate the width and length of each bilge keel, respectively.

There are no application limits for both the BF and the BL damping components. For the
application of the method, the length between perpendiculars, LBP , should be taken equal to
L.

9.2.2 Skin-friction damping component

9.2.2.1 The skin-friction damping component is given by the following equation:

4
𝐵𝐹 = 3𝜋 𝜌𝑠𝑓 𝑟𝑓3 𝜑𝑎 𝜔𝑐𝑓 (9.2.4)
where cf is the frictional coefficient, rf is the average radius from the axis of rolling, sf is the
wetted surface area, and the roll amplitude, 𝜑𝑎 , is expressed in radians. These parameters are
given as follows:
1.328 √𝑇𝑟 𝜈
𝑐𝑓 = ( ) (9.2.5)
√3.22 𝑟𝑓 𝜑𝑎

(0.887+0.145CB )⋅(1.7d+CB B)-2⋅OG

rf = π
(9.2.6)

sf =LBP (1.75d+CB B) (9.2.7)

I:\CIRC\MSC\1\MSC.1-Circ.1652.docx
 MSC.1/Circ.1652
Annex, page 97

where 𝜑𝑎 denotes roll amplitude (radians),  is the coefficient of dynamic viscosity, OG is the
distance from calm water surface to the centre of gravity (downward direction is positive), LBP
is the ship length between perpendiculars, and d is the draught. In the present method,
therefore, OG = d - KG.

9.2.3 Wave-making damping component

9.2.3.1 The non-dimensionalized wave-making damping component is given by the following

equation:

𝐴 −1
𝐵̂𝑊 = 𝜔̂1 ⋅ 𝑒𝑥𝑝 (1.44 𝐴2 (𝑙𝑜𝑔( 𝜔
̂) − 𝐴3 )2 ) (9.2.8)
where

𝑥1 = 𝐵⁄𝑑 ; 𝑥2 = 𝐶𝑏 ; 𝑥3 = 𝐶𝑚 ; 𝑥4 = 1 − 𝑂𝐺 ⁄𝑑 ;
𝐴1 = 𝐴𝐴1 ⋅ ∑3𝑖=1 ∑4𝑗=1 ∑5𝑘=1 𝑄1𝑗+4(𝑖−1),𝑘 𝑥15−𝑘 𝑥24−𝑗 𝑥43−𝑖
𝐴𝐴1 = 1.0 + (1 − 𝑥4 ) ⋅ ∑2𝑖=1 ∑3𝑗=1 ∑5𝑘=1 𝑄1𝑗+3(𝑖−1)+12,𝑘 𝑥15−𝑘 𝑥23−𝑗 𝑥32−𝑖
𝐴2 = ∑5𝑖=1 𝑄2𝑖 𝑥45−𝑖 ;

Table 9.1 presents the factors of Q1 that includes two indices: the first index refers to the row
number and the second index refers to the column number.

Table 9.1 – Factors Q1 and Q2

Factor Q1
column→ 1 2 3 4 5
row↓
1 0.00000 0.00000 0.00000 0.00000 0.00000
2 0.00000 -0.002222 0.040871 -0.286866 0.599424
3 0.00000 0.010185 -0.161176 0.904989 -1.641389
4 0.00000 -0.015422 0.220371 -1.084987 1.834167
5 -0.0628667 0.4989259 0.52735 -10.7918672 16.616327
6 0.1140667 -0.8108963 -2.2186833 25.1269741 -37.7729778
7 -0.0589333 0.2639704 3.1949667 -21.8126569 31.4113508
8 0.0107667 0.0018704 -1.2494083 6.9427931 -10.2018992
9 0.00000 0.192207 -2.787462 12.507855 -14.764856
10 0.00000 -0.350563 5.222348 -23.974852 29.007851
11 0.00000 0.237096 -3.535062 16.368376 -20.539908
12 0.00000 -0.067119 0.966362 -4.407535 5.894703
13 0.00000 17.945 -166.294 489.799 -493.142
14 0.00000 -25.507 236.275 -698.683 701.494
15 0.00000 9.077 -84.332 249.983 -250.787
16 0.00000 -16.872 156.399 -460.689 463.848
17 0.00000 24.015 -222.507 658.027 -660.665
18 0.00000 -8.56 79.549 -235.827 236.579
Factor Q2
Index 1 2 3 4 5
Q2 0.00000 -1.402 7.189 -10.993 9.45

𝐴3 = 𝐴𝐴3 + ∑7𝑖=1 ∑7𝑗=1 𝑄3𝑖,𝑗 𝑥27−𝑗 𝑥47−𝑖 ; and

4 2 4

𝐴𝐴3 = ∑ 𝑄41,𝑖 𝑥14−𝑖 ⋅ ∑ ∑ 𝑄4𝑗+1,𝑘 𝑥24−𝑘 𝑥42−𝑗

𝑖=1 𝑗=1 𝑘=1
10−𝑖
9 4 3

⋅ (∑ 𝑄5𝑖 (𝑥4 − ∑ 𝑄44,𝑗 𝑥14−𝑗 ) + ∑ 𝑄5𝑖+9 𝑥13−𝑖 )

𝑖=1 𝑗=1 𝑖=1

I:\CIRC\MSC\1\MSC.1-Circ.1652.docx
MSC.1/Circ.1652
Annex, page 98

Table 9.2 presents the factors of Q3 that includes two indices: the first index, i, refers to the row
number and the second index, j, refers to the column number.
Table 9.2 – Factors Q3

Factor Q3
column→ 1 2 3 4 5 6 7
row↓

1 -7686.0287 30131.5678 -49048.9664 42480.7709 -20665.147 5355.2035 -577.8827

2 61639.9103 -241201.0598 392579.5937 -340629.4699 166348.6917 -43358.7938 4714.7918
3 -130677.4903 507996.2604 -826728.7127 722677.104 -358360.7392 95501.4948 -10682.8619
4 -110034.6584 446051.22 -724186.4643 599411.9264 -264294.7189 58039.7328 -4774.6414
5 709672.0656 -2803850.2395 4553780.5017 -3888378.9905 1839829.259 -457313.6939 46600.823
6 -822735.9289 3238899.7308 -5256636.5472 4500543.147 -2143487.3508 538548.1194 -55751.1528
7 299122.8727 -1175773.1606 1907356.1357 -1634256.8172 780020.9393 -196679.7143 20467.0904

Table 9.3 presents the factors of Q4 that includes two indices: the first index refers to the row
number and the second index refers to the column number. Table 9.3 also presents the factors
of Q5 that includes one index ranging from 1 to 12 and is obtained from the cell below the index
in table 9.3.

Table 9.3 – Factors Q4 and Q5

Factor Q4
column→
row↓
1 2 3 4
1 -0.3767 3.39 -10.356 11.588
2 -17.102 41.495 -33.234 8.8007
3 36.566 -89.203 71.8 -18.108
4 0 -0.0727 0.7 -1.2818
Factor Q5
Index 1 2 3 4
Q5 -1.05584 12.688 -63.70534 172.84571
Index 5 6 7 8
Q5 -274.05701 257.68705 -141.40915 44.13177
Index 9 10 11 12
Q5 -7.1654 -0.0495 0.4518 -0.61655

9.2.4 Eddy-making damping component

9.2.4.1 The non-dimensionalized eddy-making damping component is given by the following

equation:
4𝜔
̂ 𝜑𝑎
𝐵̂𝐸 = 𝐶 (9.2.9)
3𝜋𝑥2 ⋅𝑥1 3 𝑅

Where,
roll amplitude, 𝜑𝑎 , is expressed in radians;

𝑥1 = 𝐵⁄𝑑 ; 𝑥2 = 𝐶𝐵 ; 𝑥3 = 𝐶𝑚 ;
𝐶𝑅 = 𝐴𝐸 ⋅ 𝑒𝑥𝑝 (𝐵𝐸1 + 𝐵𝐸2 ⋅ 𝑥3 𝐵𝐸3 );

𝐴𝐸 = (−0.0182𝑥2 + 0.0155) ⋅ (𝑥1 − 1.8)3 + ∑5𝑖=1 𝑄61,𝑖 𝑥25−𝑖 ;

𝑂𝐺 𝑂𝐺
𝐵𝐸1 = (−0.2𝑥1 + 1.6) ⋅ (3.98𝑥2 − 5.1525) ( ∑3𝑖=1 𝑄62,𝑖 𝑥23−𝑖 +
𝑑 𝑑
∑2𝑖=1 𝑄62,𝑖+3 𝑥22−𝑖 );

𝑂𝐺 𝑂𝐺
𝐵𝐸2 = (0.25 𝑑
+ 0.95) ⋅ 𝑑
+ ∑5𝑖=1 𝑄63,𝑖 𝑥25−𝑖 ; and

I:\CIRC\MSC\1\MSC.1-Circ.1652.docx
 MSC.1/Circ.1652
Annex, page 99

𝐵𝐸3 = (46.5 − 15𝑥1 ) ⋅ 𝑥2 + 11.2𝑥1 − 28.6;

Table 9.4 presents the factors of Q6 that includes two indices: the first index refers to the row
number and the second index refers to the column number.
Table 9.4 – Factors Q6

Factor Q6
column→
row↓
1 2 3 4 5
1 -79.414 215.695 -215.883 93.894 -14.848
2 0.9717 -1.55 0.723 0.04567 0.9408
3 0 -219.2 443.7 -283.3 59.6

9.2.5 Bilge keel damping component

9.2.5.1 The non-dimensionalized bilge keel damping component is given by the following
equation:

𝐵̂𝐵𝐾 = 𝐴𝐵𝐾 ⋅ 𝜔
̂ ⋅ 𝑒𝑥𝑝(𝐵𝐵𝐾1 + 𝐵𝐵𝐾2 ⋅ 𝑥3 𝐵𝐵𝐾3 ) (9.2.10)

where
𝑥1 = 𝐵⁄𝑑 ; 𝑥2 = 𝐶𝐵 ; 𝑥3 = 𝐶𝑚 ;
𝑏𝐵𝐾 𝑙
𝑥6 = 𝜑𝑎 (𝑑𝑒𝑔) ; 𝑥7 = 𝐵
; 𝑥8 = 𝐿𝐵𝐾 ;
𝐵𝑃

𝐴𝐵𝐾 = 𝑓1 ⋅ 𝑓2 ⋅ 𝑓3;

𝑓1 = (𝑥1 − 2.83)2 ∑3𝑖=1 𝑄71,𝑖 𝑥23−𝑖 + ∑3𝑖=1 𝑄72,𝑖 𝑥23−𝑖 ;

𝑓2 = ∑3𝑖=1 𝑄73,𝑖 𝑥63−𝑖 ;

𝑓3 = ∑2𝑖=1 ∑3𝑗=1 𝑄73+𝑖,𝑗 𝑥73−𝑗 𝑥83−𝑖 ;

3
𝑂𝐺
𝐵𝐵𝐾1 = ⋅ (5𝑥7 + 0.3𝑥1 − 0.2𝑥8 + ∑ 𝑄76,𝑖 𝑥63−𝑖 )
𝑑
𝑖=1
𝑂𝐺 3−𝑖
𝐵𝐵𝐾2 = −15𝑥7 + 1.2𝑥2 − 0.1𝑥1 + ∑3𝑖=1 𝑄77,𝑖 (
𝑑
) ; and

𝑂𝐺
𝐵𝐵𝐾3 = 2.5 𝑑
+ 15.75.

Table 9.5 presents the factors of Q7 that includes two indices: the first index refers to the row
number and the second index refers to the column number.

Table 9.5 – Factors Q7

Factor Q7
column→
row↓
1 2 3
1 0 -0.3651 0.3907
2 0 -2.21 2.632
3 0.00255 0.122 0.4794
4 -0.8913 -0.0733 0
5 5.2857 -0.01185 0.00189
6 0.00125 -0.0425 -1.86
7 -0.0657 0.0586 1.6164

I:\CIRC\MSC\1\MSC.1-Circ.1652.docx
MSC.1/Circ.1652
Annex, page 100

9.2.6 Lift damping component 9

9.2.6.1. The non-dimensionalized lift damping component is given by the following equation:

𝑆 𝑈𝐾 𝑙 𝑙 𝑂𝐺 𝑂𝐺 𝐵 2
𝐵̂𝐿 = 𝐿 2𝛻𝐵𝑛 20 𝑅 (1 − 1.4 𝑙 + 0.7 𝑙 𝑙 ) √2𝑔 (9.2.11)
𝑅 0 𝑅

where
2𝜋𝑑 𝐵
𝐾𝑛 = 𝐿𝐵𝑃
+ 𝜅 (4.1 𝐿 − 0.045);
𝐵𝑃

𝑆𝐿 = 𝐿𝐵𝑃 𝑑, 𝑙0 = 0.3𝑑 , 𝑙𝑅 = 0.5𝑑, 𝑈 = 𝐹𝑛 √𝐿𝐵𝑃 𝑔; and

0 𝐶𝑚 ≤ 0.92
𝜅 = {0.1 0.92 < 𝐶𝑚 ≤ 0.97
0.3 0.97 < 𝐶𝑚

9.3 Equivalent linear roll damping coefficients for the dead ship and excessive
acceleration failure modes

9.3.1 In level 1 vulnerability assessment for the excessive acceleration stability failure mode,
the non-dimensional logarithmic decrement of roll decay δ is calculated as

δφ = 0.5RPR (9.3.1)

where RPR is determined according to paragraph 2.5.2.1 of the Interim Guidelines.

9.3.2 In the level 2 vulnerability assessments for the excessive acceleration stability
failure mode and the dead ship stability failure modes, the linear, quadratic and cubic roll
damping coefficients can be determined from the equivalent linear roll damping coefficients
B44(𝜑𝑎 ) (Nm/(rad/s)), which can be obtained with the simplified Ikeda method, for some different
roll amplitude by a least square method or equivalent.
2𝜋2 𝐵44 (𝜑𝑎 ) 𝛿0 4 3
= + 3⋅𝜋 ⋅ 𝛿1 ⋅ 𝜔𝑟 ⋅ 𝜑𝑎 + 8 ⋅ 𝛿2 ⋅ 𝜔𝑟2 ⋅ 𝜑𝑎 2 (9.3.2)
𝑊𝐺𝑀𝑇𝑟2 2

where 𝑊 = 𝜌 ⋅ 𝑔 ⋅ 𝛻(N). Here, one or more roll damping coefficients can be set a priori to zero,
provided that the final fitting is sufficiently accurate.

9.3.3 Alternatively, these roll damping coefficients, 0, 1 and 2, can be determined with roll
decay test of the scaled ship model in calm water. The model is initially inclined up to a certain
heel angle. This initial angle should be larger than about 25°. If the mean roll angle between
the initial angle and the next peak angle is smaller than 20°, the initial angle should be
increased to obtain a mean angle of 20° or over. When the initial roll angle is given to the
model, additional sinkage and trim should be minimum. The model should be released from
an initial angle with zero roll angular velocity. During this test, no disturbance including waves
propagating in the longitudinal direction of the basin and reflected by its end should be given
to the model. At least four tests with different initial angles are required. If the roll damping is
very large, the number of tests should be increased to obtain a sufficient number of peaks of
the roll angle. Recording of the roll time history should start before the release of the model to
confirm that no angular velocity is given when releasing. Recording should continue until the

9
Ikeda, Y. Prediction Methods of Roll Damping of Ships and Their Application to Determine Optimum
Stabilization Devices. Marine Technology, 41(02), 89-93, 2004.

I:\CIRC\MSC\1\MSC.1-Circ.1652.docx
 MSC.1/Circ.1652
Annex, page 101

model has reached rolling angles smaller than 0.5°. This eventually requires that the length of
the basin should be sufficiently large.

9.3.4 Assuming that the absolute values of measured consecutive extremes (one
maximum and following minimum or vice versa) of roll angle during roll decay are φ1, φ2 , ….
(radians), the mean roll angle 𝜑𝑚𝑖 = 𝜑𝑖+𝜑2 𝑖+1 and the decrement Δ𝜑 = 𝜑𝑖 − 𝜑𝑖+1 are calculated.
Then Froude's extinction coefficients, a, b and c are determined by fitting these data with
equation (9.3.3).
2 3
Δ𝜑 = 𝑎𝜑𝑚 + 𝑏𝜑𝑚 + 𝑐𝜑𝑚 (9.3.3)

Here, one or more extinction coefficients can be set a priori to zero, provided that the final
fitting is sufficiently accurate. Then, the roll damping coefficients, by using the energy
conservation law, are calculated as follows:

𝛿0
2
= 2𝑎
𝑇𝑟
(9.3.4)
3𝑏
𝛿1 = (9.3.5)
4
4𝑐
𝛿2 = 3𝜋2 𝑇𝑟 (9.3.6)

9.3.5 The evaluation of ship motions for vulnerability assessment for dead ship condition
and excessive acceleration requires linearized roll damping coefficient e (1/s), using statistical
linearization. The linearized roll damping coefficient e can be calculated from the following
algebraic equation:

𝛿0 2 3 2 ( )
𝜇𝑒 = + √ ⋅ 𝛿1 ⋅ 𝜎𝑑𝜑 (𝜇𝑒 ) + ⋅ 𝛿2 ⋅ 𝜎𝑑𝜑 𝜇𝑒 (9.3.7)
2 𝜋 2

where 𝑀𝐷 = 2𝜇𝑒 (𝐼𝑥𝑥 + 𝐴44 )𝜑̇ and 𝜎𝑑𝜑 is the standard deviation of roll angular velocity.
As both parts of the equation contain the unknown damping coefficient e, the equation
is solved numerically using any appropriate iterative algorithm.

9.3.6 The evaluation of ship motion for the vulnerability assessment for parametric rolling
could be based on the following procedure:

.1 The roll motion in calm water can be modelled as follows:

(𝐼𝑥𝑥 + 𝐴44 )𝜑̈ + (𝐼𝑥𝑥 + 𝐴44 )(𝛿0 𝜑̇ + 𝛿1 𝜑̇ 2 + 𝛿2 𝜑̇ 3 ) + 𝑊𝐺𝑀𝜑 = 0 (9.3.8)

where

𝜑̈ = roll angular acceleration;

𝜑̇ = roll angular velocity; and

W = ship weight.

.2 If the equivalent linear damping coefficient, ( 𝐵44 (𝜑𝑎 ) , is introduced, the

following is obtained:

(𝐼𝑥𝑥 + 𝐴44 )𝜑̈ + 𝐵44 (𝜑𝑎 )𝜑̇ + 𝑊𝐺𝑀𝜑 = 0 (9.3.9)

I:\CIRC\MSC\1\MSC.1-Circ.1652.docx
MSC.1/Circ.1652
Annex, page 102

then,

𝜑̈ + 2𝛼𝑒 𝜑̇ + 𝜔𝑟2 𝜑 = 0 (9.3.10)

where,
𝐵44 (𝜑𝑎)
2𝛼𝑒 = 𝐼𝑥𝑥+𝐽𝑥𝑥
; and
𝑊𝐺𝑀
𝜔𝑟 = √ .
𝐼𝑥𝑥+𝐽𝑥𝑥

.3 On the other hand, the solution of equation (9.3.10) is given by 𝜑 =

𝜑𝑎 𝑒 −𝛼𝑒 𝑡 𝑐𝑜𝑠(𝜔𝑟 𝑡 − 𝜀) and the extinction curve is given by
3 2
𝛥𝜑 = 𝑎𝜑𝑚 + 𝑐𝜑𝑚 = (𝑎 + 𝑐𝜑𝑚 )𝜑𝑚 = 𝑎𝑒 𝜑𝑚 (9.3.11).

.4 Thus,
𝛼𝑒 𝑇𝑟 𝛼𝑒 𝜋 𝐵 (𝜑 ) 𝜋
𝑎𝑒 = = = 2(𝐼44 +𝐴𝑎 (9.3.12).
2 𝜔𝑟 𝑥𝑥 44 ) 𝜔𝑟

9.3.7 Using the above relationship, a procedure to determine linear and cubic
damping coefficients is as follows:

.1 First, B44 is obtained with the roll amplitude, 𝜑𝑎 , of 1 degree using the
simplified Ikeda method. Using equation (9.3.12) and assuming a = ae, the
value of a is obtained.

.2 Then, B44 is obtained with the roll amplitude of 25 degrees using the simplified
Ikeda method. Using equation (9.3.12), the value of ae is obtained.

.3 Then, c is determined with the following equation and the value of a

determined at the step .1:
2
𝑎𝑒 = 𝑎 + 𝑐𝜑𝑚 (9.3.13)

where 𝜑𝑚 corresponds to 25 degrees.

.4 Using equations (9.3.4) and (9.3.6), linear and cubic roll damping coefficients
can be calculated as follows:
1 𝜔𝜑
2
𝛿0 = 𝜋
𝑎 (9.3.14)

4𝑐 2𝜋
𝛿3 = 3𝜋2 (𝜔 ) (9.3.15)
𝜑

9.3.8 For the level 2 criterion for the excessive acceleration failure mode, the following two
alternative methods can be used to estimate the roll damping. In the stochastic linearization
method for equivalent roll damping, the roll amplitude can be alternatively approximated as
follows:

𝜑𝑎 = 0.3𝑇𝑟 𝜎𝑑𝜑 (9.3.16)

Alternatively, the equivalent linear roll damping coefficient can be defined at 15 degrees of roll
amplitude.

I:\CIRC\MSC\1\MSC.1-Circ.1652.docx
 MSC.1/Circ.1652
Annex, page 103

10 Methods for determining wave cases for vulnerability criteria

10.1 The selection of the waves to be used in vulnerability criteria is based on the following:

.1 A wave scatter table is selected (the basis for unrestricted service is

considered to be IACS Recommendation No.34 (Corr.1 Nov. 2001);
.2 For each spectral period reported in the wave scatter table (typically the zero
crossing period) the "reference significant wave height" is selected as the
conditional mean significant wave height. This provides a "reference
significant wave height" for each spectral period of the wave scatter diagram.
Eventually we obtain a series of "reference environmental conditions"
(𝑇𝑟𝑒𝑓,𝑖 , 𝐻1/3,𝑟𝑒𝑓,𝑖 = 𝐻1/3,𝑟𝑒𝑓 (𝑇𝑟𝑒𝑓,𝑖 )) with 𝑖 = 1, . . . , 𝑁 and 𝑁 being the number
of periods in the wave scatter diagram and 𝑇𝑟𝑒𝑓 being the mean period of the
spectrum. Each environmental condition has an associated probability 𝑊𝑖 =
𝑃𝑟{(𝑇𝑟𝑒𝑓,𝑖 , 𝐻1/3,𝑟𝑒𝑓,𝑖 )} which is obtained from the wave scatter table as the
probability associated to the spectral period 𝑇𝑟𝑒𝑓,𝑖 .

.3 The set of 𝑁 "calculation waves" ("wave cases") are selected separately for
parametric roll and pure loss of stability starting from the obtained set of
"reference environmental conditions". The following equivalence is used (in
the formulae, 𝑔 is the acceleration due to gravity equal to 9.81m/s2):

For parametric roll:

2
𝑔⋅𝑇𝑟𝑒𝑓,𝑖
Wavelength: 𝜆𝑖 =
2𝜋
Wave height: 𝐻𝑖 = 𝑘𝑃𝑅 ⋅ 𝐻1/3,𝑟𝑒𝑓,𝑖 (10.1.1)
Weighting factor: 𝑊𝑖
{ with 𝑘𝑃𝑅 = 0.70

For pure loss of stability:

2
𝑔⋅𝑇𝑟𝑒𝑓,𝑖
Wavelength: 𝜆𝑖 = 2𝜋
Wave height: 𝐻𝑖 = 𝑘𝑃𝐿 ⋅ 𝐻1/3,𝑟𝑒𝑓,𝑖 (10.1.2)
Weighting factor: 𝑊𝑖
{ with 𝑘𝑃𝑅 = 1.40

.1 The first check of level 2 vulnerability calculation for parametric rolling are
carried out using waves defined in equation (10.1.1).

.2 Level 1 vulnerability calculations are carried out using the following

conservative value for the wave steepness parameter:

For parametric roll:

𝐻1/3,𝑟𝑒𝑓,𝑖
𝑠𝑤 = 𝑚𝑎𝑥 (𝑘𝑃𝑅 ⋅ ) 𝑖 = 1, . . . , 𝑁 (10.1.3)
𝜆𝑖
For pure loss of stability:
𝐻1/3,𝑟𝑒𝑓,𝑖
𝑠𝑤 = 𝑚𝑎𝑥 (𝑘𝑃𝐿 ⋅ 𝜆 ) 𝑖 = 1, . . . , 𝑁 (10.1.4)
𝑖

10.2 Application example based on reference wave scatter table

10.2.1 An example application of the procedure is reported in the following. The standard

I:\CIRC\MSC\1\MSC.1-Circ.1652.docx
MSC.1/Circ.1652
Annex, page 104

reference environmental conditions in paragraph 2.7.2 of the Interim Guidelines are used, and
the corresponding wave scatter table is table 2.7.2.1.2, which refers to IACS Recommendation
No.34 (Corr. Nov.2001). Assuming a Bretschneider sea spectrum, the zero-crossing period,
𝑇𝑍 , presented in table 2.7.2.1.2 can be converted to the mean period, 𝑇𝑚𝑒𝑎𝑛 , as follows:

Bretschneider sea spectrum:

(10.2.1))
𝑇𝑚𝑒𝑎𝑛 = 1.0864 ⋅ 𝑇𝑧

If the sea spectrum to be used has a spectral shape that is different from the shape of the
Bretschneider spectrum, then the equation 10.2.1 should be modified appropriately.

10.2.2 The reference period to be used for the determination of the wavelength in the "wave
cases" is assumed to be the mean spectral period defined, starting from the zero-crossing
period provided by the wave scatter table (table 2.7.2.1.2).

If 𝑃𝑖𝑗 is defined to be the probability associated with the sea state (𝑇𝑧,𝑖 , 𝐻1/3,𝑗 ) characterized by
period 𝑇𝑧,𝑖 and significant wave height 𝐻1/3,𝑗 the conditional average significant wave height
𝐸{𝐻1/3 |𝑇𝑧,𝑖 } is determined as follows:
1
𝐸{𝐻1/3 |𝑇𝑧,𝑖 } = ∑𝑀𝑗=1 𝑃𝑖𝑗 ⋅ 𝐻1/3,𝑗
𝑊𝑖
where
𝑀 (10.2.2)
𝑊𝑖 = ∑ 𝑃𝑖𝑗
𝑗=1

Then, the reference significant wave height is assumed to be equal to the conditional average
significant wave height obtained in equation (10.2.2), i.e.:

𝐻1/3,𝑟𝑒𝑓,𝑖 = 𝐸{𝐻1/3 |𝑇𝑧,𝑖 } (10.2.3)

The wavelength 𝜆𝑖 is determined as:

2 2
𝑔⋅𝑇𝑟𝑒𝑓,𝑖 𝑔⋅𝑇𝑚𝑒𝑎𝑛,𝑖 𝑔⋅(1.0864⋅𝑇𝑧,𝑖)2 (10.2.4)
𝜆𝑖 = 2⋅𝜋
= 2⋅𝜋
= 2⋅𝜋

10.2.3 The conditional average significant wave height as a function of the zero-crossing
period is presented in table 10.1.

Table 10.1 Conditional average significant wave height

as a function of the zero-crossing period

Conditional average
Zero crossing significant
period 𝑇𝑧 [s] wave height
𝐸{𝐻1/3 |𝑇𝑧 } [m]
3.5 0.500
4.5 0.707
5.5 1.225
6.5 1.850
7.5 2.474
8.5 3.150
9.5 3.852

I:\CIRC\MSC\1\MSC.1-Circ.1652.docx
 MSC.1/Circ.1652
Annex, page 105

Conditional average
Zero crossing significant
period 𝑇𝑧 [s] wave height
𝐸{𝐻1/3 |𝑇𝑧 } [m]
10.5 4.537
11.5 5.179
12.5 5.771
13.5 6.315
14.5 6.813
15.5 7.281
16.5 7.671
17.5 8.029
18.5 8.500

10.2.4 On the basis of the reference wave scatter table and of the results in table 10.1:

.1 The wave steepness sW in paragraph 2.4.2.2 of the Interim Guidelines can

be determined using (10.1.2), (10.1.4) and (10.2. 4);

.2 The wave steepness sW in paragraph 2.5.2.2 of the Interim Guidelines can

be determined using (10.1.1), (10.1.3) and (10.2.4);

.3 The wave cases reported in table 2.5.3.2.3 of the Interim Guidelines can be
determined using (10.1.1) and (10.2.4).

I:\CIRC\MSC\1\MSC.1-Circ.1652.docx
MSC.1/Circ.1652
Annex, page 106

APPENDIX 4

THEORETICAL BACKGROUND, VALIDATION, AND APPLICATION EXAMPLES FOR

GUIDELINES ON DIRECT STABILITY ASSESSMENT

1 Introduction

1.1 In a probabilistic direct stability assessment and probabilistic operational measures,

probability of stability failure is used directly as a safety measure (criterion); therefore, these
options require counting of stability failures, which is described below in sections concerning
direct counting.

1.2 Direct counting means that stability failures need to be encountered in simulations,
which leads to a problem of rarity since stability failures are very rare for relevant ships and
loading conditions; therefore, very long simulations are required. Besides, accurate estimation
of the stability failure probability requires encounter of a sufficiently large number of stability
failures, which further increases the required simulation time.

1.3 At the same time, direct stability assessment should enable the most accurate
assessment within second generation intact stability criteria (SGISC), taking into account as
much relevant physics as possible and in the most accurate way. This means that the
simulation tools employed are rather slow and require much more computational time than
tools used in level 1 and level 2 vulnerability assessment.

1.4 Therefore, some methods to accelerate assessment are required in the probabilistic
procedures used. Several such methods are used in the Interim Guidelines: assessment in
design situations, statistical extrapolation and deterministic assessment.

2 Validation of numerical methods for simulation of ship motions

2.1 Qualitative validation: Backbone curve

2.1.1 Qualitative validation requirements are summarized in table 3.4.2 of the Interim
Guidelines. A demonstration of consistency between the calculated roll backbone curve and
the GZ curve in calm water is required for software where the hydrostatic and Froude-Krylov
forces are calculated with a body-exact formulation. The roll backbone curve is a dependency
of the roll frequency in calm water on the initial roll amplitude.

2.1.2 To serve the purpose of the qualitative validation, the roll backbone curve provided
by the calculation code should be computed in the same software configuration that is used
for the direct stability assessment. Figure 2.1.1 illustrates calculations of the backbone curve.
The backbone curve, computed with a potential flow code, suggests an initial increase of GM
as the backbone curve shows an initial hardening, which is confirmed by the local GM
(derivative of GZ) shown in figure 2.1.2.

I:\CIRC\MSC\1\MSC.1-Circ.1652.docx
 MSC.1/Circ.1652
Annex, page 107

, deg

60

50

40

30

20

10

GZ, m , 1/s

1.5 1.0 0.5 0 0.05 0.1 0.15 0.2 0.25
Figure 2.1.1: The roll backbone curve and the GZ curve

2.1.3 As expected, the backbone curve “bends” towards zero with the softening of the
dynamical system around 40° of the initial amplitude. An unstable equilibrium at the angle of
vanishing stability slows down the dynamical system causing the roll period to grow.

2.1.4 The calculations of the backbone curve were done by series of simulations in calm
water. All initial conditions except for the roll angle are set to zero. The initial roll angle is set
close to the angle of vanishing stability. In total, 4 to 6 simulations are recommended with
different initial roll angles. The difference between roll angles can be taken as 0.1 º. In general,
it may not always be possible to perform roll decays starting very close to the angle of vanishing
stability. In such cases, the roll decay should start from large enough initial angles.

GM, m
1.6

1.4

1.2

1.0

0.8

0.6 , deg
0 5 10 15 20 25 30 35 40 45
Figure 2.1.2: local GM (derivative of GZ) vs. heel angle

I:\CIRC\MSC\1\MSC.1-Circ.1652.docx
MSC.1/Circ.1652
Annex, page 108

2.1.5 A calculation of the natural frequency is performed on a half-oscillation basis:

𝜔𝜑 (𝐴𝑖 ) = 𝜋/(𝑡𝑖+1 − 𝑡𝑖 ), where an amplitude Ai occurs between the zero-crossings at ti+1 and ti,
shown in figure 2.1.3.

A0 , deg

40

20 Ai
250 time, t, s
0 50 100 150 200
-20

-40
t0 ti ti+1
-60

Figure 2.1.3: Roll motion computed with potential flow code

2.1.6 The largest amplitude is the initial roll angle, A0, as shown in figure 2.1.3. The lowest
frequency on the backbone curve is computed with a quarter of an oscillation, 𝜔𝜑 (𝐴0 ) =
0.5𝜋/𝑡0 .

2.2 Qualitative validation: Response curve

2.2.1 The qualitative validation from table 3.4.2 of the Interim Guidelines contains a
requirement for the demonstration of consistency between the calculated roll backbone curve
and the roll response curve. The roll response curve is a dependence of roll amplitude in
regular waves on the frequency of those waves.

2.2.2 Calculation of the roll response curve was done by a series of simulations in regular
waves with frequencies in a range from 0.7 to 1.2 of the roll natural frequency. Two sets of
simulations were performed: one set from low frequency to high frequency and the other set
from high frequency to low.

2.2.3 Each simulation was carried out until a steady state was achieved. Duration of the
simulation was at least 30 roll periods. A criterion of a steady state was the observation of a
periodic solution where the difference between the successive amplitudes was within 1%. The
reported amplitude was computed as an average of five successive amplitudes satisfying this
steady state requirement.

2.2.4 The steady state initial conditions were computed for a time instant of a wave
zero-crossing during one of five roll oscillations where amplitudes satisfy the steady state
requirement of 2.2.3. Initial conditions for the first simulation were set to zero; steady state
initial conditions from previous simulations were used for the second and further simulations.

2.2.5 The consistency between the roll response curve and the backbone curve is
observed when the response curves bend together with the backbone curve as can be seen
in the example described in paragraphs 2.2.6-2.2.10.

I:\CIRC\MSC\1\MSC.1-Circ.1652.docx
 MSC.1/Circ.1652
Annex, page 109

2.2.6 Example calculations10 are presented for a fishing vessel, for which a panel model is
shown in figure 2.2.1 (without a forecastle), the principal characteristics are given in table 2.2.1,
and the GZ curve, used for the single degree of freedom (1-DOF) response curve calculation,
is shown in figure 2.2.2.

Table 2.2.1 Principal dimensions of a fishing vessel

Length BP, m 22.00
Breadth moulded, m 6.62
Draught amidships, m 2.70
KG, m 2.78
Displacement, t 197.00
GM, m 0.22

Figure 2.2.1
Hull geometry
of a fishing
vessel

GZ, m

0.2

 deg
0 20 40 60

-0.2

Figure 2.2.2: Righting arm (GZ) curve of fishing vessel for response curve calculation

2.2.7 Calculations were carried out with a 3D potential flow code and consisted of a series
of simulations in regular beam waves. After the initial transient, positive and negative peaks
were measured, shown in figure 2.2.3 (⊡ for positive peaks and + for negative peaks).
Simulation started from high and then from low wave frequencies. Initial conditions were
assigned using the steady state response, achieved for the previous wave frequency.

2.2.8 The roll response amplitudes computed by the 3D potential flow code are located on
both sides of the backbone curve, showing an expected "bending pattern". The consistency
between backbone curve and roll response curve has been demonstrated.

10
Shin, Y. S, Belenky, V. L., Lin, W. M., Weems, K. M. and Engle, A. H. Nonlinear time domain simulation
technology for seakeeping and wave-load analysis for modern ship design. SNAME Trans. Vol. 111, pp.
557-578, 2003.

I:\CIRC\MSC\1\MSC.1-Circ.1652.docx
MSC.1/Circ.1652
Annex, page 110

0.8 Amplitude, rad

Expected shape of Potential flow code:

the response curve Positive peak
0.6 based on a single Negative peak
B DOF roll equation
Backbone curve

Integration of roll equation

Positive peak
Negative peak
0.4
A

0.2

0. 1 1.2 1.4 1.6

8 Circular frequency of wave, ωW, rad/s
Figure 2.2.3:Response curve of roll, based on 3D potential flow simulations; circular
frequency range from 0.9 to 1.18 rad/s, wave amplitude 0.4 m

2.2.9 To reveal expected shape of the roll response curve, 3D potential flow simulations
were supplemented with a direct numerical integration of the approximate differential equation
for roll, using the actual GZ curve and approximate Froude-Krylov wave excitation:
2
𝜑̈ + 𝛿0 𝜑̇ + 𝑓(𝜑) = 𝛼𝑚 𝜔𝜑0 sin (𝜔𝑊 𝑡) (2.2.1)

where 0 is the natural roll frequency,  is the roll damping coefficient, m is the effective
amplitude of the wave slope, W is the wave frequency, and 𝑓 (𝜑) is the stiffness function,
expressed through the GZ curve as 𝑓 (𝜑) = 𝑚g𝐺𝑍(𝜑)/(𝐼𝑥 + 𝐴44 ) , where m is mass
displacement, Ix the transverse moment of inertia of ship mass and A44 is added mass in roll.
Numerical integration of the ordinary differential equation was done in the same way as in 3D
potential flow simulations: the steady state condition of the previous frequencies was used as
initial conditions for the next wave frequency. The results are shown in figure 2.2.3 ( for
positive and  for negative peaks).

2.2.10 As expected, the bending of the backbone curve causes a response curve to fold,
forming a range of frequencies where two stable steady state responses are possible;
identified by points A and B in figure 2.2.3. The observation of the folding of the 3D potential
flow response demonstrates consistency between the backbone curve and the response curve
that satisfies the requirement in line 2 of table 3.4.2 of the Interim Guidelines. However, the
regions of frequency with multiple steady states are not always present, and the response
curve may still bend and show only one single steady state for each forcing frequency.

2.3 Qualitative validation: Change of stability in waves

2.3.1 The qualitative validation, table 3.4.2 of the Interim Guidelines, requires
demonstration of a capability to reproduce the wave pass effect. The objective is to verify that
the stability decreases when the wave crest is located near the midship sections. An example

I:\CIRC\MSC\1\MSC.1-Circ.1652.docx
 MSC.1/Circ.1652
Annex, page 111

of a possible procedure can be found in the literature. 11 Verification of the variation of

instantaneous stability in waves with a numerical simulation is carried out as follows:

.1 Set the wave length equal to the ship length.

.2 Set the simulation in following waves at a ship forward speed equal to the
wave celerity.

.3 Set an initial position of a ship to have a wave trough near the midship
section. The ship is expected to remain stationary relative to the wave.

.4 Apply a constant external heeling moment.

.5 Record the ship motions until the transition is completed and an equilibrium
state is achieved; then, extract the equilibrium heel angle.

.6 Repeat the procedure for different values of the external heeling moment.

.7 Repeat the procedure for an initial position of the ship corresponding to a

location of the wave crest near the midship section.

2.3.2 Example calculations were carried out for the ONR tumblehome top configuration;12
principal characteristics are given in table 2.3.1. The wave length is equal to the ship length
(154 m) and the wave height is 6 m. Two wave positions were used for qualitative validation,
with a wave crest midships and wave trough midships.

Table 2.3.1 Principal dimensions of ONR tumblehome top configuration

Length BP, m 154.00

Breadth moulded, m 18.00
Draught midships, m 5.50
KG, m 8.32
Displacement, t 8,675.60
GM, m 1.50

2.3.3 Figure 2.3.1 shows the achieved heel angles under the external heeling moment
(heeling lever curves), computed with a 3D potential flow code for the wave crest and the wave
trough located midships. The figure demonstrates the change of stability in waves: the
achieved heel angle under the same heeling moment is significantly larger when wave crest is
located near midships, compared to when the wave trough is located near midships. The figure
also shows that the stability on a wave trough is better, while stability on a wave crest is worse
in comparison with calm water.

11
Belenky, V. and Weems, K.M. Probabilistic Qualities of Stability Change in Waves. Proc. 10th Intl. Ship
Stability Workshop, Daejeon, Korea, pp. 95-108, 2008.
12
Bishop, R. C., Belknap, W., Turner, C., Simon, B. and Kim, J. H. Parametric Investigation on the Influence
of GM, Roll Damping, and Above-Water Form on the Roll Response of Model 5613. Report NSWCCD-50-
TR-2005/027, 2005.

I:\CIRC\MSC\1\MSC.1-Circ.1652.docx
MSC.1/Circ.1652
Annex, page 112

0.4 GZ, m
Wave trough
0.3 amidships Figure 2.3.1:
Calm water GZ curves from
0.2 both static
calculations in
Wave crest waves (solid lines)
0.1 amidships
Heel, φ, and heeling lever
deg curves (circles)
0 10 20 30 40 50 60 70 computed with a
3D potential flow
-0.1 code

-0.2

2.4 Qualitative validation: Principal parametric roll resonance

2.4.1 The qualitative validation, as shown in table 3.4.2 of the Interim Guidelines, requires
demonstration of a capability to reproduce a principal parametric roll resonance. The objective
is to observe an increase and stabilization of amplitudes in following or head waves at an
encounter frequency that is about twice the natural roll frequency (which is termed principal
parametric roll resonance).

2.4.2 Figure 2.4.1 shows a time history from 3D potential flow code simulations in following
waves for the C11 class container ship at zero forward speed. The draught is 12.7 m, the KG
is 19.0 m, the GM is 1.29 m and the natural roll frequency is 0.199 rad/s. The wave height
was 2 m and the wave frequency was 0.42 rad/s (equal to the encounter frequency at zero
forward speed). Figure 2.4.2 shows the entire frequency range of the principle parametric
resonance computed for these conditions.

40 Roll, φ, deg

20
time, t, s
0 5 10 15 20 25 30 35
0 0 0 0 0 0 0
-20

-40
Wave elevation at CG, m
1

0.5
time, t, s
0 5 10 15 20 25 30 35
-0.5 0 0 0 0 0 0 0

-1

Figure 2.4.1: Time history of the principle parametric resonance computed with a 3D
potential flow code for the C11 class container ship with a natural roll period of 0.199
rad/s at zero forward speed, a wave height of 2 m and a wave frequency of 0.42 rad/s

I:\CIRC\MSC\1\MSC.1-Circ.1652.docx
 MSC.1/Circ.1652
Annex, page 113

40 Maximum roll angle, φmax, deg Figure 2.4.2:

The frequency range
30 of the principle
parametric resonance
computed with 3D
20
potential flow code
for the C11-class
10 container ship with a
natural roll period of
0.3 0.35 0.4 0.45 0.5 0.199 rad/s at zero
Wave frequency, ωW, rad/s forward speed and a
wave height of 2 m

2.4.3 Since there is no direct roll excitation in following waves, parametric resonance is the
only reason for increasing roll amplitude; a non-zero amplitude response is grouped around a
wave encounter frequency that is twice the roll frequency. Thus, it is the principle parametric
resonance.

2.5 Quantitative validation requirements

2.5.1 Indicative requirements and acceptance criteria for the quantitative validation of
numerical methods for the simulation of ship motions for direct stability assessment are
summarized in table 3.4.3 of the Interim Guidelines; rows 1, 2 and 4 are relevant for the
parametric roll stability failure mode.

2.5.2 Row 1 of table 3.4.3 of the Interim Guidelines contains requirements for the response
curve of parametric roll: the maximum (over encounter frequency) roll amplitude should not be
underpredicted by more than 10% if the amplitude is less than the angle of the maximum GZ
and 20% otherwise. At the same time, an underprediction of less than two degrees can be
disregarded. A comparison of 3D potential flow code results with a model test on parametric
roll13 is shown in figure 2.5.1. The model test included all six DOF and was run at a fixed speed
of 10 knots in a wave with full-scale height of 8.4 m and period of 14.0 s. The difference
between the measured and the computed parametric roll amplitude was within one degree.
However, the comparison is available for one frequency and one can accept partial satisfaction
of the requirements in the first row.

13
France, W. N., Levadou, M., Treakle, T.M., Paulling, J.R., Michel, R W.K and Moore, C. An Investigation of
Head-Sea Parametric Rolling and its Influence on Container Lashing Systems, Marine Technology, 40(1):
1-19, 2003.

I:\CIRC\MSC\1\MSC.1-Circ.1652.docx
MSC.1/Circ.1652
Annex, page 114

40

20
Roll (Deg.)

-20

350 400 450 500 550 600 650

6
4
Pitch (Deg.)

-2

-4 LAMP-2 with Tuned Roll Damping

Experiment
350 400 450 500 550 600 650
Time (s)
(s) series of parametric rolling between the
Figure 2.5.1:A comparison in a time
experiment and a 3D potential flow code

2.5.3 Row 4 of table 3.4.3 of the Interim Guidelines contains requirements for the variance
testing for parametric roll. The objective is to demonstrate a correct, in terms of statistics,
modelling of roll response in irregular waves. An example using the C11 class post-Panamax
containership, whose details are available in appendix 2, is shown in figures 2.5.2 and 2.5.3. 14
Here, the model experiment was conducted at a towing tank: the ship model was towed by a
towing carriage using soft elastic ropes. The measuring time durations of the experiments
were 4,200, 2,400 and 1,200 s in full scale for the Froude numbers of 0.0, 0.05, 0.1,
respectively. The numerical model used here is a heave-roll-pitch coupled model. The
non-linear Froude-Krylov forces are directly calculated by integrating the wave pressure up to
an irregular wave surface profile. The 2D hydrodynamic forces used for the radiation and
diffraction forces are calculated for the submerged hull by the integral equation method with
an instantaneous roll angle taken into account. The radiation forces in roll are calculated for
the natural roll frequency and those in vertical motion (heave and pitch) are done for the peak
of mean wave frequency. The linear and quadratic roll damping coefficients are used in the
mathematical model: they are determined from an experimental result of roll decay tests. The
numbers of realizations for both the experiment and the simulation are 10 for the Froude
number of 0.0 and 0.05 and are 5 for the Froude number of 0.1. In all cases, the confidence
intervals of the ensemble average of variance of roll angle from the simulation are overlapped
with those from the experiment or more conservative than the experiment. Thus, this numerical
code complies with the row 4 of table 3.4.3 of the Interim Guidelines.

14
Hashimoto, H. and Umeda, N. (2019) "Prediction of Parametric Rolling in Irregular Head Waves" Chapter 16
of Contemporary Ideas on Ship Stability. Risk of Capsizing, Belenky, V., Spyrou, K., van Walree F., Neves,
M.A.S., and Umeda, N. eds., Springer, ISBN 978-3-030-00514-6, pp. 275-289.

I:\CIRC\MSC\1\MSC.1-Circ.1652.docx
 20

roll a
10
0
0 0.025 0.05 0.075 0.1
MSC.1/Circ.1652
Annex, page 115

Exp. Cal. Average

Figure 2.5.2: The variance of the parametric roll angle in irregular head waves
for several Froude numbers with T01 = 9.99s and Hs = 7.82m

Figure 2.5.3: Variance of parametric roll angle in irregular head waves for several
significant wave heights with T01 = 9.99s and Fn = 0.0

2.5.4 Row 5 of table 3.4.3 of the Interim Guidelines contains requirements for conditions
for surf-riding/broaching. For broaching, figure 5.2.2 in appendix 4 is an example of the relevant
validation using the ONR flare topside vessel. The differences between the model experiment
and the coupled surge-sway-yaw-roll simulation are 15% for the critical wave steepness at the
nominal Froude number of 0.4 and 1% for the speed setting at the wave steepness of 0.05.
Thus, this numerical model is quantitatively validated under these conditions.

2.6 Validation example for dead ship condition stability failure mode

2.6.1 An example of validation of numerical method is provided for a large cruise ship;
table 2.6.1 shows its principal particulars and figure 2.6.1 shows its GZ curve.

I:\CIRC\MSC\1\MSC.1-Circ.1652.docx
MSC.1/Circ.1652
Annex, page 116

Table 2.6.1: Principal particulars of a cruise ship

Length between perpendiculars 246.00 m
Breath 37.50 m
Draught 8.21 m
Metacentric height (GM) 2.50 m
Natural roll period 23.55 s

GM · φ
GZ
GZ

heel angle (degrees)

heel

Figure 2.6.1: GZ curve of cruise ship

2.6.2 A coupled sway-heave-roll-pitch model was applied in long-crested irregular beam

waves. Radiation and diffraction forces and moment were calculated using strip theory, and
the roll damping moment was estimated from the roll decay test for a scaled model.

GM = 2.5m

Uw (m/s)

Figure 2.6.2: The variance of the roll angle of the cruise ship in beam wind and waves

2.6.3 For the validation, model experiments using a scaled model were used. The sway
motion was softly constrained, and the yaw motion was fixed by a cable system. A comparison
of the roll variance in figure 2.6.2 indicates that the 95% confidence intervals of the simulation
and the experiment are overlapped so that the numerical code explains the model
experiments.

I:\CIRC\MSC\1\MSC.1-Circ.1652.docx
 MSC.1/Circ.1652
Annex, page 117

2.7 Validation example of linear method for excessive acceleration failure mode

2.7.1 A ship motion prediction based on linear potential hydrodynamics, otherwise referred
to as the ”linear superposition method” in the following, can be used for direct stability
assessment for the excessive acceleration failure mode, if it is quantitatively validated based
on paragraph 3.4.3 of the Interim Guidelines for a ship and its loading conditions similar to
those to be assessed. This is because excessive accelerations can be assessed with sufficient
accuracy even with a linear theory if the effect of nonlinear damping can be approximated by
using stochastically equivalent linearization (paragraph 9.3.5 of appendix 3) or by using
linearized damping at 15º (paragraph 9.3.8 of appendix 3).

2.7.2 To validate this method, numerical results are compared with model experiments 15
conducted according to the ITTC recommended procedure 7.5-02-07-04 for intact stability
model tests. An example of the validation is provided for a container ship of which the principal
particulars are shown in table 2.7.1. First, the numerical code for the Response Amplitude
Operators is validated. Here, the Salvesen-Tuck-Faltinsen method based on strip theory is
used. The ship motions were calculated with five degrees of freedom; hydrodynamic forces
were calculated by two-dimensional source distribution method. The roll damping coefficient
was estimated from a roll decay test. The comparison of lateral acceleration in regular beam
waves between numerical and experimental results in figure 2.7.1 shows a good agreement.

Table 2.7.1: Principal particulars of the container ship

Length between perpendiculars 320.00 m
Breadth 42.80 m
Draught at aft perpendicular 9.073 m
Draught at midship section 8.08 m
Draught at forward perpendicular 7.083 m
Metacentric height, GM 8.54 m
Height of bilge keel 0.60 m
Length of bilge keel 96.00 m
Longitudinal distance of bridge from aft perpendicular 77.70 m
Height of bridge from baseline 50.00 m

15
Kuroda, T., Hara, S., Houtani, H. and Ota, D. Direct Stability Assessment for excessive acceleration failure
mode and validation by model test. Ocean Engineering, Vol.187, 106137, 2019.

I:\CIRC\MSC\1\MSC.1-Circ.1652.docx
MSC.1/Circ.1652
Annex, page 118

8
Experimental result
Lateral Acceleration[m/s ]
2

Numerical result
6
Figure 2.7.1:
A comparison of the
4 lateral acceleration in
regular beam waves
between numerical and
2 experimental results

0
0 1 2 3
wave length/ship length

2.7.3 Second, the linear superposition method is validated. The mean of one-third largest
amplitudes of lateral acceleration in short-crested irregular beam waves is compared between
level 2 vulnerability criterion, linear superposition method and experiments in figure 2.7.2 at
significant wave height 5.5 m and mean zero-crossing wave period 9.5 s. Short-crested
irregular waves are modelled where the frequency spectrum is the ITTC recommended
unlimited fetch spectrum (1978) and the wave energy spreading is the fourth power of the
cosine function. Seven model test runs of three-hours duration (in full scale) were carried out,
with randomly varied phases and directions of harmonic wave components discretizing the
wave energy spectrum. The comparison shows that the linear superposition method is close
to the experiments and provides conservative estimates; further, the level 2 vulnerability
criterion is consistent with respect to the linear superposition method.

Figure 2.7.2:
Mean amplitudes of highest one−third

3.5
A comparison of
the mean of the
one-third largest
3.4 lateral
of lateral acceleration[m/s ]
2

Level 2 vulnerability criterion

acceleration
amplitudes in an
3.2 The superposition method irregular short-
crested beam
Experimental result and
waves between
3 its 95％ confidence interval the level 2
vulnerability
criterion, linear
superposition
2.8 method and
experiment

2.7.4 These results indicate that the linear superposition method with transfer functions of
ship motions obtained with frequency-domain calculations is feasible for the direct stability
assessment for the excessive acceleration failure mode.

I:\CIRC\MSC\1\MSC.1-Circ.1652.docx
 MSC.1/Circ.1652
Annex, page 119

3 Direct counting

3.1 Introduction

3.1.1 Direct counting is counting of the number of stability failures per given exposure time.
Direct counting is used in the probabilistic design stability assessment in design situations as
well as in the full probabilistic direct stability assessment and in operational measures. It can
be used in combination with statistical extrapolation, which itself uses direct counting.

3.1.2 Counting the number of stability failures per given exposure time assumes a
stationary Poisson process. The stationarity assumption is justified since both the design
assessment and operational measures consider ensemble statistics over a large number of
ships, each of which operates in stationary conditions for unlimited time. The Poisson process
assumption requires, in addition, that stability failures happen independently, i.e. the
occurrence of one failure does not affect the probability of occurrence of a second failure. The
validity of this assumption is based on two heuristic considerations: clumping heuristic
(although large roll motions tend to appear in groups, the occurrence of such groups may be
independent), and rarity heuristic (rare events tend to be independent).

3.1.3 To ensure that numerical simulations or model tests also satisfy the requirements of
a stationary Poisson process, special procedures are required that are considered below.

3.2 Definition and characteristics of Poisson process

3.2.1 A counting process is defined as a stochastic process, when a random variable N

counts the number of events (stability failures for the Interim Guidelines) in a time interval
from 0 to t. For a stationary Poisson process, the failures are independent (the numbers of
failures in non-overlapping time intervals are independent) and stationary (the number of
failures depends only on the length of a time interval and not on its location in time).

3.2.2 There are several equivalent definitions of a Poisson process; the following 16 is used
here: Poisson process with a constant rate r > 0 is a counting process where the number of
events N(t) in a time interval of length t satisfies the Poisson distribution with the mean rt, i.e.

p{N(t) = k} = (rt)ke−rt/k! for k = 0, 1, ... (3.2.1)

3.2.3 The probability mass function of Poisson distribution, f(k) = p{N(t) = k}, eq. (3.2.1), is
equal to the probability that the number of events during a time interval t is equal to k.

3.2.4 The useful properties of a Poisson process include:

.1 the superposition property: sum of independent Poisson processes N1, …,

Nk, i.e. N1 +∙∙∙+ Nk, is a Poisson process with the rate r1 +∙∙∙+ rk. This means
that failure rates can be found separately for different stability failure modes
and summed to give the total stability failure rate; conversely, if the sum of
two independent random variables is Poisson distributed, so are each of
these two variables;

.2 the random split property: if each event in a Poisson process N(t) with rate r
is randomly tagged as either process N1(t), with probability p, or N2(t), with
probability 1 - p, then the two resulting processes N1(t) and N2(t) are
independent Poisson processes with rates rp and r(1 - p), respectively;

16
Ross, Sheldon M. Stochastic Processes. Wiley, 1996.

I:\CIRC\MSC\1\MSC.1-Circ.1652.docx
MSC.1/Circ.1652
Annex, page 120

.3 the thinning property: if each event of a Poisson process with rate r is

randomly marked, with probability p, then the marked process is a Poisson
process with rate rp;

.4 the mean of a Poisson process, equal to the mean number of events per
interval t (i.e. the rate is equal to the expected number of events per unit
time), is given by:
∞
𝐸{𝑁(𝑡)} = ∫0 𝑡𝑓(𝑡)𝑑𝑡 = 𝑟𝑡 (3.2.2)

.5 the variance of the Poisson process is equal to the mean, Var{N(t)} = rt.

3.2.5 A special case of eq. (3.2.1) is k = 0, which provides the probability p that no stability
failures occur from time 0 to time t:

p  p{N(t) = 0} = e−rt (3.2.3)

3.2.6 From eq. (3.2.3), the probability p* can be defined as:

(3.2.4)
p*  p{N(t) > 0} = 1 − p{N(t) = 0} = 1 − p = 1 − e−rt
which means that at least one stability failure occurs during the time interval t, i.e. that k > 0
("probability of stability failure during time t").

3.2.7 The linearization of eq. (3.2.4) with respect to rt leads to a popular approximation
(note, however, that this approximation is valid only for small values of product rt):

p*  rt (3.2.5)

3.2.8 A Poisson process can also be seen as a sequence of time intervals T1 (from t = 0 to
the first failure), T2 (between the first and second failures), etc. These time intervals are also
random variables. The probability that the time until the first failure exceeds t, i.e. p{T1 > t}, is
the same as the probability that no failures occur before time t, i.e. p{N(t) = 0} = e−rt, eq. (3.2.3).
Therefore, p{T1 > t} = e−rt, which means that T1 is an exponentially distributed random variable.
Similarly, all time intervals Ti are exponentially distributed random variables with the rate r.

3.2.9 Therefore, a Poisson process can also be defined as a counting process in which the
time intervals between events are independent random variables, which are exponentially
distributed with rate r (note that this definition automatically ensures independent and
stationary increments):

p{T>t} = e−rt for t > 0 and 0 otherwise (3.2.6)

3.2.10 An important property of the exponential distribution is its memoryless property: if a

failure has not occurred until time t, the distribution of the remaining waiting time is the same
as the distribution of the original waiting time, i.e. the remaining waiting time has no memory
of the previous waiting time. The exponential distribution is the only continuous distribution
with this property: if the time intervals between arrivals are not exponential, the process will
not be a Poisson process.

I:\CIRC\MSC\1\MSC.1-Circ.1652.docx
 MSC.1/Circ.1652
Annex, page 121

3.2.11 Other useful properties of the exponential distribution include:

.1 corresponding to sum of Poisson processes, if T1,…,Tk are independent

exponentially distributed random variables with rates r1,…,rk, then
min(T1,…,Tk) is exponentially distributed with the rate r1+∙∙∙+rk, and the index
of the variable that achieves the minimum is distributed according to the law
p{i | Ti = min(T1,…,Tk)} = ri/(r1 +∙∙∙+ rk);

.2 according to eq. (3.2.4), cumulative distribution function of time to failure is

F(t)  p{T < t} = 1 − p{T > t} = 1 − e−rt for t > 0 and 0 otherwise (3.2.7)

.3 probability density function of exponential distribution, i.e. of time intervals

between events in a Poisson process, is

f(t) = dF(t)/dt = re−rt for t > 0 and 0 otherwise (3.2.8)

.4 the mean of exponentially distributed random variable T (i.e. mean time

between stability failures) is
∞
𝐸{𝑇} ≡ 𝑇̄ = ∫0 𝑡𝑟𝑒 −𝑟𝑡 𝑑𝑡 = 1/𝑟 , and (3.2.9)
∞
.5 the second moment 𝐸{𝑇 2 } = ∫0 𝑡 2 𝑟𝑒 −𝑟𝑡 𝑑𝑡 = 2/𝑟 2 , then the variance of the
time between failures is 𝑉𝑎𝑟{𝑇} ≡ 𝐸{𝑇 2 } − 𝐸 2 {𝑇} = 2𝑇̄ /𝑟 − 𝑇̄ 2 = 1/𝑟 2 = 𝑇̄ 2 ,
and the standard deviation of time between failures is then equal to

T = (Var{T})1/2 = 1/r = T̄ . (3.2.10)

3.2.12 Since the rate r is the only parameter defining a Poisson process, any statistical
characteristic of the exponential distribution, including the variance and the standard deviation,
is known once the stability failure rate is known. To confirm eq. (3.2.10), figure 3.2.1 shows the
ratio 𝜎{𝑇}/𝑇̄ equal to 1 according to eq. (3.2.10), as a function of the number of counted
failures N, and figure 3.2.2 shows the estimate of the standard deviation {T} as a function of
the estimate of the mean time to failure 𝑇̄ after N = 200 counted failures.
(T,i), s
(T,i)/T

N T, s

Figure 3.2.1: The ratio of the estimate of Figure 3.2.2: The estimate of the standard
the standard deviation of the time to deviation of the time to failure vs. the
failure to the estimate of mean time to estimate of the mean time to failure
failure vs. number of failures from 200 simulated failures

3.3 Definition of failure rate from sample data using exponential distribution

3.3.1 Both the Poisson distribution and the corresponding exponential distribution are
defined by a single parameter, the stability failure rate r. To define it from a series of numerical
simulations or model tests, consider time intervals Ti between failures and define the sample
mean time to failure after N failures as:

I:\CIRC\MSC\1\MSC.1-Circ.1652.docx
MSC.1/Circ.1652
Annex, page 122

𝑇̂ = (1/𝑁) ∑𝑁
𝑖=1 𝑇𝑖 (3.3.1)

3.3.2 To estimate the stability failure rate:

.1 the joint probability density function Lr of all individual time intervals Ti is

defined: since Ti are independent, then eq. (3.2.8) leads to

𝐿𝑟 (𝑇1 , 𝑇2 ,∙∙∙, 𝑇𝑁 ; 𝑟) = ∏𝑁
𝑖=1 𝑟𝑒
−𝑟𝑇𝑖 (3.3.2);

.2 the most probable value of r, called the maximum likelihood estimate, is the
value that maximizes Lr (i.e. the value of r that is most probable for a given
measured data set T1, T2, ∙∙∙, TN.). To do this, it is more convenient to
maximize ln( 𝐿𝑟 ) = ∑𝑁𝑖=1(ln 𝑟 − 𝑟𝑇𝑖 ) , which is possible as a logarithm is a
monotonously increasing function, i.e. Lr and ln(Lr) have maximum at the
same r. Hence, the maximum of ln(Lr) can be defined from the condition
d ln(Lr)/dr = 0, i.e.

d ln 𝐿𝑟 /d𝑟 = ∑𝑁 𝑁 ̂
𝑖=1(1/𝑟 − 𝑇𝑖 ) = 𝑁/𝑟 − ∑𝑖=1 𝑇𝑖 = 𝑁/𝑟 − 𝑁𝑇 = 0 (3.3.3);

.3 from eq. (3.3.3), the maximum likelihood estimate of the stability failure rate
can be calculated simply as:

𝑟̂ = 1/𝑇̂. (3.3.4).

3.3.3 The total simulation time (or total model testing time) tt is defined as
𝑁
𝑡t = ∑ 𝑇𝑖 (3.3.5)
𝑖=1

3.3.4 Using eq. (3.3.1), eq. (3.3.4) and the definition (3.3.5), the maximum likelihood
estimate of the stability failure rate 𝑟̂ can be calculated simply as the total number of failures
divided by the total simulation (or model testing) time:

𝑟̂ = 𝑁/𝑡t (3.3.6)

3.3.5 Since the estimate of the stability failure rate is a random variable, it varies between
different series of simulations or model tests and, thus, it can be defined only with uncertainty,
which decreases with an increasing duration of numerical simulations or model tests. To
account for this uncertainty, direct stability assessment and operational measures use as a
practical criterion the upper boundary of the 95%-confidence interval of the stability failure rate.
The (1 − )100%-confidence interval for the stability failure rate, where  is a small value (e.g.
for 95%-confidence interval,  = (1 − 95/100) = 0.05, is calculated as the confidence interval
of the rate parameter of an exponential distribution:17

2𝑁 1 2𝑁
2 < 2 < (3.3.7)
𝑟 𝑟̂ 𝜒𝛼/2,2𝑁
𝑟̂ 𝜒1−𝛼/2,2𝑁
2
3.3.6 Here, 𝜒𝑝,𝑓 denotes the p∙100%-quantile (corresponding to a lower tail area, equal to
the cumulative probability p) of the 2-distribution with f degrees of freedom, figure 3.3.1. The
2
function 𝜒𝑝,𝑓 is available in many software packages. For a small N, table 3.3.1 may be used,
providing 𝜒𝑝,𝑓 values for p = 1 – /2 and p = /2 (at  = 0.05) and f = 2N.
2

17
Ross, Sheldon M. Introduction to probability and statistics for engineers and scientists. 4th ed., Associated
Press, p. 267, 2009.

I:\CIRC\MSC\1\MSC.1-Circ.1652.docx
 MSC.1/Circ.1652
Annex, page 123

f
Figure 3.3.1:
Examples of the
functions 𝝌𝟐𝒑,𝒇 and
ꭓ2p,f , ꭓ21-p,f

𝝌𝟐𝟏−𝒑,𝒇 for f = 1, 5, 10
and 50 degrees of
freedom vs. the
cumulative
probability p

Table 3.3.1. Functions 𝝌𝟐𝟏−𝜶/𝟐,𝟐𝑵 and 𝝌𝟐𝜶/𝟐,𝟐𝑵 for  = 0.05 and N = 1, 2, …, 24

2 2 2 2 2 2
N 𝜒1−𝛼/2,2𝑁 𝜒𝛼/2,2𝑁 N 𝜒1−𝛼/2,2𝑁 𝜒𝛼/2,2𝑁 N 𝜒1−𝛼/2,2𝑁 𝜒𝛼/2,2𝑁

1 7.3778 0.0506 9 31.5264 8.2307 17 51.9660 19.8063

2 11.1433 0.4844 10 34.1696 9.5908 18 54.4373 21.3359
3 14.4494 1.2373 11 36.7807 10.9823 19 56.8955 22.8785
4 17.5345 2.1797 12 39.3641 12.4012 20 59.3417 24.4330
5 20.4832 3.2470 13 41.9232 13.8439 21 61.7768 25.9987
6 23.3367 4.4038 14 44.4608 15.3079 22 64.2015 27.5746
7 26.1189 5.6287 15 46.9792 16.7908 23 66.6165 29.1601
8 28.8454 6.9077 16 49.4804 18.2908 24 69.0226 30.7545

3.3.7 For a large N (N  25 is sufficient for practical purposes), the 2 distribution converges
2
to a normal distribution (which is also available in many software packages): 𝜒𝑝,2𝑁 → 2𝑁 +
2√𝑁 ⋅ ℵ0,1 (𝑝), where ℵ0,1 (𝑝) is the p∙100%-percentile of the standard normal distribution.

3.3.8 Solving eq. (3.3.7) with respect to r gives:

2 2
0.5𝑟̂ 𝜒𝛼/2,2𝑁 /𝑁 < 𝑟 < 0.5𝑟̂ 𝜒1−𝛼/2,2𝑁 /𝑁 (3.3.8)

3.3.9 From eq. (3.3.8), the upper rU and lower rL boundaries of a (1 − )100%-confidence
interval for the stability failure rate can be calculated as:
2
𝑟U = 0.5𝑟̂ 𝜒1−𝛼/2,2𝑁 /𝑁 (3.3.9)
2
𝑟L = 0.5𝑟̂ 𝜒𝛼/2,2𝑁 /𝑁 (3.3.10)

3.3.10 The upper rU and lower rL boundaries of a 95%-confidence interval for the stability
2 2
failure rate are then 𝑟U = 0.5𝑟̂ 𝜒0.975,2𝑁 /𝑁 and 𝑟L = 0.5𝑟̂ 𝜒0.025,2𝑁 /𝑁, respectively; the normal
distribution approximation, paragraph 3.3.7, gives 𝑟U = 𝑟̂ (1 + 1.96𝑁 −1/2 ) and
−1/2
𝑟L = 𝑟̂ (1 − 1.96𝑁 ) for a 95%-confidence interval.

3.3.11 Using these estimates, any other characteristic of the process can be calculated with
formulae from section 3.2. For example, the upper pU and lower pL boundaries of the
(1 − )100%-confidence interval of probability of stability failure during time interval t (i.e.
probability that at least one stability failure happens during time interval t) can be calculated
according to eq. (3.2.4) as:

I:\CIRC\MSC\1\MSC.1-Circ.1652.docx
MSC.1/Circ.1652
Annex, page 124

𝑝U = 1 − exp( − 𝑟𝑈 𝑡) (3.3.11)
𝑝L = 1 − exp( − 𝑟L 𝑡) (3.3.12)

3.4 Definition of failure rate from sample data from analysis of probability of failure

3.4.1 Alternatively, the probability 𝑝̂ that at least one failure happens within a given
exposure time texp can be defined directly from the results of M simulations, each of the same
duration texp. Denoting as N the total number of simulations in which at least one stability failure
was encountered, the maximum likelihood estimate of this probability can be calculated as:

p ̂  N/M (3.4.1)

3.4.2 Note that since eq. (3.4.1) requires only the number of such simulations in which at
least one stability failure was observed, the continuation of simulations after the first
encountered stability failure is not necessary. Accordingly, the actual total simulation time for
a simulation with a stability failure can be shorter than the reference exposure time texp.

3.4.3 The substitution of eq. (3.4.1) in eq. (3.2.4) leads to N/M = 1 – exp( - rtexp), and then
to the following estimate of the failure rate:

r = - ln(1 – N/M)/texp (3.4.2)

3.4.4 Three items should be noted:

.1 the estimate given by eq. (3.4.2) converges to the result given by eq. (3.3.6)
when M increases to infinity while the total simulation time tt is kept constant
(i.e. when texp = tt / M → 0). Thus, eq. (3.3.6) provides better accuracy than
eq. (3.4.2) for the same total simulation time;

.2 eq. (3.4.2) cannot be used if every simulation contains at least one failure (i.e.
M = N). To avoid this issue and, besides, improve accuracy of the estimate
(3.4.2), the results can be post-processed by separating the simulations into
intervals shorter than texp but longer than the decorrelation time described in
section 3.8 of this appendix. The length of these shorter intervals should be
the same for all simulations; this length should be used instead of texp in
eq. (3.4.2). If such post-processing does not resolve this issue, post-
processing of the results can be performed using the procedures described
in sections 3.3 and 3.5; and

.3 the treatment of stability failures that occur during initial transients while
keeping the exposure time texp constant requires special care when using the
approach based on eq. (3.4.2).

3.4.5 For the determination of the confidence interval for the probability 𝑝̂ that at least one
failure happens within the specified exposure time texp, the Clopper-Pearson method can be
used, according to which the lower and upper boundaries of a (1 - )∙100%-confidence interval
for p* can be found as:
.1 the lower boundary of the (1 - )∙100%-confidence interval:
𝜈1 ⋅𝐹𝜈1,𝜈2;𝛼/2
pL = 0 for N = 0 or 𝑝𝐿 = otherwise (3.4.3)
𝜈 +𝜈 ⋅𝐹
2 1 𝜈1 ,𝜈2 ;𝛼/2
where 1 = 2N and 2 = 2(M - N + 1); and

.2 the upper boundary of the (1 - )∙100%-confidence interval:

I:\CIRC\MSC\1\MSC.1-Circ.1652.docx
 MSC.1/Circ.1652
Annex, page 125

𝜈1⋅𝐹𝜈1,𝜈2;1−𝛼/2
pU = 1 for N = M or 𝑝𝑈 = otherwise (3.4.4)
𝜈2 +𝜈1 ⋅𝐹𝜈1,𝜈2;1−𝛼/2
where 1 = 2(N + 1) and 2 = 2(M - N).

3.4.6 In equations (3.4.3) and (3.4.4), 𝐹𝜈1 ,𝜈2;𝑥 (with x = / 2 or 1 - / 2) is the inverse
cumulative F-distribution with 1 and 2 degrees of freedom, calculated at a value x. This
function is available in many software packages.

3.4.7 The upper rU and lower rL boundaries of the (1 − )100%-confidence interval of the
failure rate estimate can be calculated from the assumed relation p =1 - exp(-r·texp), as

rU = - ln(1 - pU)/texp (3.4.5)

rL = - ln(1 - pL)/texp (3.4.6)

3.4.8 A 95%-confidence interval corresponds to a significance level  = 0.05.

3.4.9 The value of the maximum exposure time texp is specified by the user and it may
depend on the considered conditions.

3.5 Definition of failure rate from sample data using binomial distribution

3.5.1 This procedure is based on the assumption of a binomial distribution of the failure rate
estimate.18 This distribution is equivalent to the Poisson process assumption for failure events
and the exponential distribution of the time before and between failures, which is essential for
a direct counting procedure.

3.5.2 The binomial distribution describes the probability that there will be NaU independent
up-crossing events of a level a or down-crossing events of a level –a, associated with stability
failure, out of a total 𝑁𝑎 = ∑𝑁𝑟 𝑘=1 𝑁𝑘 instances of observation of roll motion or lateral
acceleration. The data set of observation consists of Nr records. Each record contains Nk
observations, k = 1,…Nr; i.e. the records may contain different numbers of observations and
may be of different durations.

3.5.3 The first up-crossing (or down-crossing) after the initial transition time is an
independent event. The next independent up-crossing (or down-crossing) is counted only after
decorrelation time Tdc has passed. The total number of independent up-crossings and
down-crossings is 𝑁𝑎𝑈 = ∑𝑁𝑟
𝑘=1 𝑁𝑈𝑘 , where NUk is the number of independent up-crossings and
down-crossings observed during the k-th record.

3.5.4 The failure rate is estimated as 𝑟̂ = 𝑁𝑎𝑈 Δ𝑡/𝑇𝑎 , where Δt is the time increment used in
the simulation, 𝑇𝑎 = ∑𝑁𝑟 𝑘=1(𝑁𝑘 Δ𝑡 − 𝑇ramp ) is the total time of all records with a ramp time
Tramp excluded to account for initial transients.

3.5.5 The number of independent up-crossings or down-crossings 𝑁𝑎𝑈 is a random variable

with a binomial distribution. The binomial distribution has only one parameter, the probability
that the event will occur at any particular instant of time. This probability can be estimated as
𝑝̂ = 𝑁𝑎𝑈 Δ𝑡/𝑇𝑎 .

18
Leadbetter, M.R., Rychlik, I. and Stambaugh, K. Estimating Dynamic Stability Event Probabilities from
Simulation and Wave Modeling Methods. Chapter 22 of Contemporary Ideas on Ship Stability. Risk of
Capsizing, Belenky, V., Spyrou, K., van Walree F., Neves, M.A.S., and Umeda, N., eds., Springer,
ISBN 978-3-030-00514-6, pp. 381-391, 2019.

I:\CIRC\MSC\1\MSC.1-Circ.1652.docx
MSC.1/Circ.1652
Annex, page 126

3.5.6 The variance of a variable NaU that satisfies a binomial distribution can be estimated
as 𝑉𝑁𝑈 = 𝑇𝑎 𝑝̂ (1 − 𝑝̂ ))/Δ𝑡 ≈ 𝑁𝑎𝑈 since t and 𝑝̂ are small values. Since 𝛥𝑡 is small, also p̂
̂
is small, therefore 𝑉̂𝑁𝑈 can be determined as 𝑉̂𝑁𝑈 = 𝑁𝑎𝑈 .

𝑄𝑁 (0.5⋅(1+𝑃𝛽 )) 𝑁𝑎𝑈
3.5.7 𝑟𝑈,𝐿 = 𝑟̂ ⋅ (1 ± ) with 𝑟̂ = (3.5.1)
√𝑁𝑎𝑈 𝑇𝑎
where QN is the quantile of the standard normal distribution and Pß is the accepted confidence
probability. For Pß = 0.95, 𝑄𝑁 (0.5(1 + 𝑃𝛽 )) = 1.96.

3.5.8 For the total number of failures NaU = 25 and less, it is recommended to use a binomial
distribution directly for the evaluation of the boundaries of confidence interval. Using a normal
approximation for the binomial distribution may be inaccurate for a small number of events due
to the practical limits of applicability of the Central Limit Theorem:

𝑟U = 𝑄𝐵 (1 − 𝛼/2, 𝑇𝑎 /Δ𝑡, 𝑝̂ )⁄𝑇𝑎 (3.5.2)

𝑟𝐿 = 𝑄𝐵 (𝛼/2, 𝑇𝑎 /Δ𝑡, 𝑝̂ )⁄𝑇𝑎
(3.5.3)

where 𝑄𝐵 (𝑃, 𝑁, 𝑝) is a quantile of the binomial distribution for probability P, the number of trials
(time steps) N, and parameter p.

3.6 Long-term statistics

3.6.1 The long-term operation can be considered as a sum of stationary Poisson processes,
each of which has a constant rate ri and occurs in a stationary situation which is encountered
with a probability pi. Then, applying the sum property and tagging property of a stationary
Poisson process, the long-term operation can be considered to be a stationary Poisson
process with a constant rate 𝑟̄ , which is equal to:

𝑟̄ = ∑𝑖 𝑟𝑖 𝑝𝑖 (3.6.1)

3.6.2 All the properties of a Poisson process remain applicable by using the mean rate 𝑟̄ ,
for example:

.1 the distribution of the number of failures in a time interval from t to t + , i.e.

the probability that k failures occur in time interval from t to t + , is equal to:

𝑝{𝑁(𝑡 + 𝜏) − 𝑁(𝑡) = 𝑘} = (𝑟̄ 𝜏)𝑘 exp( − 𝑟̄ 𝜏)/𝑘! (3.6.2)

.2 the probability that no failures occur from time t to time t +  is:

𝑝 ≡ 𝑝{𝑁(𝑡 + 𝜏) − 𝑁(𝑡) = 0} = exp( − 𝑟̄ 𝜏) (3.6.3)

.3 the probability that at least one failure happens from time t to time t + 
("probability of failure during time ") is:

𝑝∗ ≡ 𝑝{𝑁(𝑡 + 𝜏) − 𝑁(𝑡) > 0} = 1 − exp( − 𝑟̄ 𝜏) (3.6.4)

3.6.3 Therefore, the problem reduces to the definition of a constant rate ri in each stationary
situation from either the numerical simulations or model tests in such a way that ensures that
the process is a stationary Poisson process.

I:\CIRC\MSC\1\MSC.1-Circ.1652.docx
 MSC.1/Circ.1652
Annex, page 127

3.7 Cautions in numerical simulations or model tests

3.7.1 Numerical simulations (or model tests) and the counting procedure used should
ensure the stationarity of the process and the independence of the counted stability failures.
The following straightforward procedure may seem suitable: run a single simulation of roll
motion of sufficient duration and, after each encountered stability failure, increase the number
N of failures by 1. Then, increase the total simulation time tt by the simulation time from the
previous failure and update the estimate of the failure rate 𝑟̂ = 𝑁/𝑡𝑡 , eq. (3.3.6), and the
estimate of the upper boundary of its (1 − )100%-confidence interval 𝑟U = 0.5𝑟̂ 𝜒1−𝛼/2,2𝑁
2
/𝑁,
eq. (3.3.9). Once rU is less than the acceptance standard, the simulations can be stopped and
the loading condition can be considered acceptable.

3.7.2 However, the collection of sufficient statistics in a single simulation run of sufficient
duration may be impossible in practice since the duration of numerical simulations or model
tests is limited because:

.1 the duration of model tests is limited by tank size and wave reflection effects;

.2 the duration of numerical simulations and model tests is limited by

self-repetition effects if the sea state is modelled as a finite sum of harmonic
components:

𝜁(𝑡) = ∑𝑀
𝑖=1 𝑎𝑖 cos( 𝜔𝑖 𝑡 + 𝜀𝑖 ) (3.7.1)

where ai = {2SZZ(i)D(i)ii}1/2 are amplitudes, i are frequencies, i are

headings and i are phases of harmonic components, randomly selected in
the interval [0,2], SZZ is the wave energy spectrum and D is the wave energy
angular spreading function; and

.3 for resonance phenomena, such as parametric roll and synchronous roll, the
bandwidth of significant roll response is narrow. This narrow bandwidth of
encounter frequencies may lead to a self-repetition effect and may further
limit the duration of a roll numerical simulation.

3.7.3 A solution is to generate multiple independent realizations of the same sea state, by
a random variation of phases i for each realization, and to simulate ship motions in each such
realization for a limited time. The frequencies i, headings i and perhaps also amplitudes ai
of components can also be randomly varied. To generate random values, pseudo-random
number generators can be used.

3.7.4 For a set of simulations in multiple realizations of the same sea state, eq. (3.3.6) is
still applicable; in this equation, N and tt should be taken as the total number of stability failures
and the total simulation time, respectively, over all realizations.

3.7.5 Transient hydrodynamic effects that occur at the beginning of each simulation lead to the
violation of the stationarity requirement and thus reduce the accuracy of the failure rate estimation.
These effects may also lead to numerical instabilities in time-domain numerical simulation codes.
To address this problem, the following possibilities (or their combination) can be used:

.1 Exclude a certain period of time after the start of each simulation from
post-processing. For this purpose, 50 roll periods is suggested for stability
failure modes dominated by roll motion. The initial transients are excluded
from tt (section 3.3), texp (section 3.4) or Ta (section 3.5). If a stability failure
occurs during the initial transient, it is not counted in the number of failures N
and simulation timer is set again to zero after such stability failure.

I:\CIRC\MSC\1\MSC.1-Circ.1652.docx
MSC.1/Circ.1652
Annex, page 128

.2 Smoothly (e.g. linearly) increase the wave (or forces acting on the ship) from
zero to the nominal value during a specified ramp time, the duration of which
for stability failure modes dominated by roll is recommended to be set equal
to 10 roll periods. This time is excluded from the post-processing.

.3 Randomly vary the initial conditions for each realization.

3.7.6 The independence of stability failures may also be violated by the autocorrelation of
large roll motions since large roll amplitudes, caused by resonance, tend to appear in groups.
If the direct counting approach in section 3.4 is used, this effect is automatically neutralized
since occurrence of the second, etc. stability failures in the same simulation does not change
the result. If the direct counting approaches in sections 3.3 or 3.5 are used, neglecting this
effect may lead to overestimation of the stability failure rate estimate and underestimation of
the size of the confidence interval. To neutralize this effect, a simple solution is to stop a
simulation if a stability failure is encountered. Another possibility is to switch off the simulation
timer tt and stability failure counter N after an encountered failure for the decorrelation time; an
example procedure for that approach is provided in section 3.8.

3.7.7 Stopping a simulation after an encountered failure and switching off the
post-processing for the decorrelation time have a similar amount of unused data. In the former
method, this is due to transient effects that occur at the start of a new simulation. In the latter
method, this is due to the decay of the autocorrelation function of roll motion. However, the
former method is simpler. Another benefit of the former method is that restarting also handles
self-repetition effects because, for relevant durations of simulations and stability failure rates,
a stability failure is not encountered in each simulation. Therefore, after the first encountered
failure, there is no benefit in waiting for a second failure in the same simulation.

3.7.8 After removing portions of the time histories of roll motion affected by self-repetition
effects, initial transients and autocorrelation of stability failures, the remaining pieces represent
a single stationary Poisson process: the removed pieces can be disregarded due to the
memoryless property, the durations of the remaining pieces may be arbitrary (in particular,
equal), and whether a failure was encountered in each simulation or not, figure 3.7.1, can be
disregarded. This means that eq. (3.3.6) can be used for the maximum likelihood estimate of
the failure rate, with N and tt representing sums over all remaining pieces of the simulations.
Similarly, a sample mean time to failure is 𝑇̂ = 𝑡t /𝑁, and all formulae from sections 3.2, 3.3,
3.4 and 3.5 can be directly applied.

3.7.9 To define the maximum duration of simulations to avoid self-repetition effects, two
requirements should be checked: first, there should be an absence of self-repetition of waves
and, second, an absence of self-repetition of roll motion, for which the absence of
self-repetition of waves may not be sufficient due to a narrow-banded roll response.
φ, degrees

T,s

I:\CIRC\MSC\1\MSC.1-Circ.1652.docx
 MSC.1/Circ.1652
Annex, page 129
φ degrees

T,s
φ, degrees

T,s

(a)
(b)
Figure 3.7.1: A roll motion in multiple realizations of a sea state (a) and the resulting
Poisson process (b)

3.7.10 To check for the absence of the self-repetition of waves, the autocovariance function
𝑅w (𝜏) of wave elevation is computed as
∞ T1 𝑁 T2
𝑅w (𝜏) = ∫ 𝑆𝑍𝑍 (𝜔)cos (𝜔𝜏)𝑑𝜔 ≈ ∑ 𝑆𝑍𝑍𝑖 cos(𝜔w𝑖 𝜏) (3.7.2)
0 𝑖=1
where τ is time lag.

3.7.11 The autocorrelation function 𝑟w (𝜏) is normalized by the variance or 𝑅w (𝜏 = 0):

𝑟𝑤 (𝜏) = 𝑅𝑤 (𝜏)/𝑅𝑤 (0) (3.7.3)

3.7.12 The autocorrelation function, computed using eq. (3.7.2) for long-crested waves and
using a discretization with 210 frequencies for 30 minutes of the target record length, is shown

I:\CIRC\MSC\1\MSC.1-Circ.1652.docx
MSC.1/Circ.1652
Annex, page 130

in figure 3.7.2. No increases of the autocorrelation function are observed; thus, no

self-repetition effect should be expected.

1
rw()

0.5

-0.5

-1
0 10 20 30 40 , s
rw ()
1

-1
0 200 400 600 800 1000 1200 1400 1600 1800 , s
Figure 3.7.2: The autocorrelation function for 210 frequencies

3.7.13 The described check for the self-repetition effect was performed for a fixed location.
If a ship is moving forward in head or bow-quartering waves, the rate of wave encounter is
greater compared to the fixed location, so the actual check of the self-repetition effect is done
for each speed and heading combination, using encounter frequency 𝜔e𝑖 , instead of the wave
frequency:
2 ⁄
𝜔e𝑖 = 𝜔w𝑖 − 𝑘w𝑖 𝑣s cos𝜇 : 𝑘𝑤𝑖 = 𝜔𝑤𝑖 𝑔 (3.7.4)

3.7.14 The autocovariance and the autocorrelation function were computed as follows:

𝑅we (𝜏) ≈ ∑𝑁
𝑖=1 𝑆𝑤𝑒𝑖 cos(𝜔e𝑖 𝜏)∆𝜔𝑤𝑒𝑖 , 𝑟we (𝜏) = 𝑅we (𝜏)/𝑅we (0) (3.7.5)

where Swei is the spectrum of the encounter waves.

3.7.15 Figure 3.7.3 shows the autocorrelation function, computed for forward speed
of 20 knots in head seas. The growth of the autocorrelation function indicates a self-repetition
effect.

I:\CIRC\MSC\1\MSC.1-Circ.1652.docx
 MSC.1/Circ.1652
Annex, page 131

1
rwe()

-1
0 200 400 600 800 1000 1200 1400 1600 1800 , s
Figure 3.7.3: The autocorrelation function for 210 frequencies at 20 knots in head seas

3.7.16 Figure 3.7.3 indicates the beginning of self-repetition effects at about 650 s. However,
for massive computations, such as those required in a direct stability assessment, the search
for the self-repetition effect needs to be done automatically. For this purpose a peak-based
envelope of the autocorrelation function is used, see figure 3.7.4.

0.04
rwe()

0.02

-0.02

60 61 62 63 64 65 66 67 68 69 , s
10 0 0 0 0 0 0 0 0 0
rwe()

0.5

0.5

-1
0 100 200 300 400 500 600 700 800 900 , s

Figure 3.7.4: Use of an envelope to detect the inception of a self-repetition effect

3.7.17 A natural criterion for a time of inception of the self-repetition effect is when the
envelope starts to increase. However, implementation of this criterion may encounter a
problem due to small amplitude oscillations which may be observed on the envelope; for an
example, see the zoomed-in portion of the autocorrelation plot in figure 3.7.4. Averaging over
every three values seems to remediate this problem, figure 3.7.5.

I:\CIRC\MSC\1\MSC.1-Circ.1652.docx
MSC.1/Circ.1652
Annex, page 132

0.04
rwe()

0.02

-0.02

600 610 620 630 640 650 660 670 680 690 , s
Figure 3.7.5: On the detection of the inception of a self-repetition effect

3.7.18 To prevent the criterion of self-repetition from being too sensitive, a threshold of 0.005
is introduced. The search for an increase in the averaged envelope starts only when it exceeds
the threshold. The times where the self-repetition effect was detected are shown in table 3.7.1.
For the initial frequency discretization with 210 frequencies and for these combinations of
speed and heading, the wave model (3.7.1) is statistically valid up to the time durations in
table 3.7.1. To increase the time duration of the validity of a model in head and bow-quartering
seas, the number of frequencies needs to be increased to 560; this provides 30 minutes of
validity for all combinations of speeds and headings, see table 3.7.2.

Table 3.7.1: Time until self-repetition Table 3.7.2: Time until self-repetition
occurs for 210 frequencies for a peak wave occurs for 560 frequencies for a peak wave
period of 14 s period of 14 s
Speed, kn Speed, kn
Heading, ° 5 10 15 20 Heading, ° 5 10 15 20
105 1752 1570 1427 1307 105 1920 1920 1920 1920
120 1579 1318 1131 988.9 120 1920 1920 1920 1920
135 1465 1159 962.8 824.6 135 1920 1920 1920 1920
150 1379 1060 861.3 727.8 150 1920 1920 1920 1920
165 1335 1005 808.6 677.5 165 1920 1920 1920 1847
180 1318 988.9 791.1 659.3 180 1920 1920 1920 1802

3.7.19 To check for the self-repetition of roll motion, an ensemble-average autocorrelation

function of roll motion can be computed; however, such computations are time-consuming and
prone to numerical errors. A simpler check is to use quantile plots of time to failure to verify
whether the distribution of time to failure deviates from the exponential distribution.
To investigate self-repetition effects, parametric roll resonance in head waves and
synchronous roll resonance in beam waves were simulated for a systematically varied
significant wave height in two types of simulations: in one, denoted “limited”, the simulation
time was limited to three hours (while simulations were stopped after the first failure in any
case); and in the other denoted “unlimited”, the simulations were run always until the first failure
occurred.

3.7.20 To compare the distribution of time to failure with the exponential distribution, quantile
diagrams (QQ diagrams) like those shown in figure 3.7.6 were used. This figure shows QQ
diagrams derived from "limited" and "unlimited" simulations. Since the cumulative distribution
function of an exponentially distributed time to failure is F(t) = 1 − e−rt for t > 0, the ratio 𝑇𝑖 /𝑇̄
should be equal to −ln(1 − Fi) (see the blue dashed lines in figure 3.7.6). For comparison, the
cumulative distribution function Fi was calculated from the sample data as i / (N + 1), where i
is the index of a stability failure when stability failures are sorted in ascending order of Ti.
Figure 3.7.6 shows that the "unlimited" simulations overestimate the time to failure compared

I:\CIRC\MSC\1\MSC.1-Circ.1652.docx
 MSC.1/Circ.1652
Annex, page 133

to the exponential distribution and compared to the 'limited' simulations for cases with a large
time to failure. This can be explained by the presence of self-repetition effects: the same
ʺuncriticalʺ realization (i.e. a realization where a stability failure does not happen soon) repeats
itself again and again (since the repetition is not exact, failure may eventually occur but much
later than it should).
“

-ln(1-F)
-ln(1-F)

synchronous roll (dead ship), T = 28.9h parametric roll, T = 23.5h

̂
𝑇𝑖 /𝑇 ̂𝑇
𝑇𝑇𝑖𝑖//𝑇̂

Synchronous roll (dead ship) T = 7.9h

-ln(1-F)

-ln(1-F)

parametric roll, T = 4.1h

̂
𝑇𝑖 /𝑇 ̂
𝑇𝑖 /𝑇
synchronous roll (dead ship, T = 1.1h parametric roll, T = 33.9 min
-ln(1-F)

-ln(1-F)

Figure 3.7.6: Quantile diagrams from 'limited' (⚫) and 'unlimited' () simulations for
synchronous roll (left) and parametric roll (right) resonance cases

3.7.21 This means the simulations that have a long duration lead to a deviation from the
Poisson process, which means that the notion of failure rate and the formulae from this section
are not applicable. Moreover, using these formulae despite this (i.e. assuming that the process
were a Poisson process) would lead to an underestimation of the failure rate, i.e. a
non-conservative error which should be avoided. Thus, the maximum duration of simulations
should be limited. According to the presented results, when at least 1.9∙10 4 frequencies are
used, simulations up to three hours are possible. In general, quantile plots can be used to
verify absence of self-repetition effects.

3.7.22 To check whether the described measures ensure applicability of the Poisson process
assumption, the 2 goodness-of-fit test was applied to several cases of parametric roll
resonance and synchronous roll resonance in head and beam waves, respectively, at
systematically varied significant wave heights. Only "limited" simulations of a three-hour
duration (or until failure if it happened) were used. Random realizations of the same sea state
were repeated until about 103 failures were encountered in each sea state. The observed times
to failure were compared with the exponential distribution, which was defined using the
maximum likelihood estimate for the failure rate 𝑟̂ = 1/𝑇̂, eq. (3.3.4).

3.7.23 The full range t  0 of the time to failure was subdivided into a number of k  5 intervals
of equal probability F = (1 − e−rt) = 1/k; the number of intervals was systematically increased
up to a maximum k = N/5. The number Oi of the observed times to failure within each interval i

I:\CIRC\MSC\1\MSC.1-Circ.1652.docx
MSC.1/Circ.1652
Annex, page 134

was counted, and the expected number Ei was calculated, according to the assumed
exponential distribution, as N / k. Subsequently:

.1 the test statistic was calculated as

𝑥 = ∑𝑘𝑖=1(𝑂𝑖 − 𝐸𝑖 )2 /𝐸𝑖 (3.7.6)

.2 the critical value of the test statistic at the significance level  = 0.05 was
2
defined as 𝑐5% = 𝜒1−𝛼,𝑓 , i.e. the value of the 2 distribution, at the cumulative
probability 1 -  = 0.95 and the number of degrees of freedom f = k − p − 1,
where p = 1 is the number of parameters of the assumed distribution
estimated from the sample data; and

.3 the results were presented (see figure 3.7.7) as the ratio x / c5% depending
on the number of intervals k: when x / c5% < 1, the null hypothesis that the
data follow the assumed distribution cannot be rejected at the significance
level of 5%.
x / c5%

𝑇̂
𝑇̂
x / c5%
x / c5%

𝑇̂ 𝑇̂

𝑇̂
x / c5%

x / c5%

𝑇̂

Figure 3.7.7: The ratio of the 2-test statistic to the critical value 𝝌𝟐𝟏−𝜶,𝒌−𝟐 for the
significance level  = 5% vs. the number of intervals k of the time to failure for
synchronous roll resonance (left) and parametric roll resonance (right) for several
values of the sample mean time to failure

3.7.24 Figure 3.7.8 shows the ratio x / c5% for k = 200 as a function of the sample mean time
to failure. For synchronous roll resonance, the Poisson process model is acceptable (at the
5% significance level) in all studied cases. On the other hand, for parametric roll resonance,
the results disagree with the Poisson process assumption: marginally at 𝑇̂ ≈ 2 hours and
increasingly greater for 𝑇̂ decreasing below two hours. However, the 2 test is considered as
very strict when the amount of data is large.

I:\CIRC\MSC\1\MSC.1-Circ.1652.docx
 MSC.1/Circ.1652
Annex, page 135

parametric roll (PR)

Synchronous roll (DS)

Figure 3.7.8: The ratio x/c5% for synchronous roll resonance () and parametric roll
resonance () vs. the sample mean time to failure at k = 200

3.8 Decorrelation

3.8.1 The independence of stability failures may also be violated by the autocorrelation of
large roll motions since large roll amplitudes, caused by resonance, tend to appear in groups.
Neglecting this effect may lead to an overestimation of the stability failure rate estimate and an
underestimation of the size of the confidence interval. One of the possibilities to neutralize this
effect is to switch off the simulation timer tt and the stability failure counter N after an
encountered failure until the envelope of the autocorrelation function of roll motion reduces to
a specified level; here, an example procedure for that possibility is provided.

3.8.2 A practical way to judge if the events related to a stochastic process are independent
is to compare the time between the events with a decorrelation time of the process. The
decorrelation time is a duration when the autocorrelation function value becomes insignificant,
so that two-time sections of the process are considered uncorrelated. Uncorrelated values are
assumed to be independent. This is an assumption since the correlation only indicates a
dependence in terms of the second joint moment of the distribution. For example, for a joint
PDF of random variables x and y with their respective mean values Ex and Ey, the second joint
∞
moment is a covariance, which is defined as ∫−∞(𝑥 − 𝐸𝑥 )( 𝑦 − 𝐸𝑦 )𝑝𝑑𝑓(𝑥, 𝑦)𝑑𝑥𝑑𝑦 . The
decorrelation time can be used as an approximate indicator of the independence of events
related to a stochastic process.

3.8.3 The autocovariance function (autocorrelation function is a normalized autocovariance

function) of a single realization of an ensemble is estimated as:

1 𝑁−𝑖
𝑅̂𝜑 (𝜏𝑖 ) = ∑ (𝜑𝑗 − 𝐸̂𝑎𝜑 )(𝜑𝑖+𝑗 − 𝐸̂𝑎𝜑 ) (3.8.1)
𝑁 𝑗=1

I:\CIRC\MSC\1\MSC.1-Circ.1652.docx
MSC.1/Circ.1652
Annex, page 136

where N is the number of a point in the realization, 𝜏𝑖 is i-th time lag and 𝐸̂𝑎𝜑 is a mean estimate
of the ensemble:

𝐸̂𝑎𝜑 = ∑𝑁𝑟 ̂
𝑘=1 𝑊𝑘 𝐸𝜑𝑘
(3.8.2)
where Nr is the number of realizations in an ensemble and Wk is a statistical weight of the k-th
realization, 𝑊𝑘 = 𝑁𝑘 / ∑𝑁𝑟
𝑛=1 𝑁𝑛 , where Nk is a number of data points in the k-th realization.

3.8.4 𝐸̂𝜑𝑘 is a mean estimate of k-th realization:

𝑁𝑘
1
𝐸̂𝜑𝑘 = ∑ 𝜑𝑖 (3.8.3)
𝑁𝑘
𝑖=1

3.8.5 Then, the ensemble estimate of the autocovariance function is expressed as:

𝑅̂𝑎𝜑 (𝜏𝑖 ) = ∑𝑁𝑟 ̂

𝑘=1 𝑊𝑘 𝑅𝜑𝑘 (𝜏𝑖 ) (3.8.4)

𝑅̂𝜑𝑘 (𝜏𝑖 ) is the autocovariance estimate of the k-th realization. An estimate of the
autocorrelation function is obtained by normalizing the estimate of the autocovariance function
by its zero-term, which is an ensemble variance estimate in this context:

𝑟̂𝑎𝜑 (𝜏𝑖 ) = 𝑅̂𝑎𝜑 (𝜏𝑖 )/𝑅̂𝑎𝜑 (0) (3.8.5)

3.8.6 The ensemble estimate of the autocorrelation function is shown in figure 3.8.1. The
decorrelation time is defined here as the time lag for the reduction of the autocorrelation function
below 0.05. Since the estimate of autocorrelation function has an oscillatory character, it makes
sense to use its envelope rather than the autocorrelation function itself; see figure 3.8.1.

3.8.7 The decorrelation time, evaluated as the first intersection of the envelope with the
threshold level 0.05, was determined to be 759.7 s. This is a rather large value in comparison
to the wave elevation or synchronous roll where the decorrelation time is usually around
one minute. Large autocorrelation times for parametric roll is generally expected since
autocorrelation function of parametric roll decays for a long time.19

Autocorrelation function, (-)

1
Envelope

0.5
Decorrelation time

0.05
0

-0.5

Time lag, s
-1
0 100 200 300 400 500 600 700 800 900 1000

19
Belenky, V. and Weems, K.M. Probabilistic Properties of Parametric Roll. Chapter 6 of Parametric
Resonance in Dynamical Systems, Fossen, T. I. and Nijmeijer, H., eds., Springer, ISBN 978-1-4614-10423-
0, pp. 129-146, 2012.

I:\CIRC\MSC\1\MSC.1-Circ.1652.docx
 MSC.1/Circ.1652
Annex, page 137

Figure 3.8.1: The ensemble estimate of the autocorrelation function for a wave
heading of 1°, a speed of 5 knots, a significant wave height of 3.5 m, and a mean
zero-crossing wave period of 8.5 s
3.8.8 Sometimes, the estimate of autocorrelation function for parametric roll resonance
does not reach the threshold level with increasing time lag. One possibility for this is that the
envelope has at least one minimum and the lowest minimum is still above the threshold level,
which is a result of insufficient data for the estimation of autocorrelation function to the
threshold level. The time lag corresponding to the global minimum of the envelope then can
be used as a decorrelation time. Another possibility is that the envelope monotonically
decreases but does not reach the threshold level, which is likely caused by an insufficient
length of the estimate of the autocorrelation function. Since it may be difficult to estimate the
autocorrelation function beyond one half of the record length, the conservative solution is to
take the decorrelation time as equal to the length of the record.
3.9 Direct counting procedures
3.9.1 The proposed procedures are based on simulations of ship motions in multiple
independent realizations of the same irregular seaway and provide the estimate of the upper
boundary rU of the 95%-confidence interval of the rate of stability failures. The procedures can
prevent self-repetition effects, transient hydrodynamic effects at the beginning of simulations
and autocorrelation of large roll motions.
3.9.2 The sea state is modelled as a sum of harmonic components, see eq. (3.7.1). The
phases and, possibly, frequencies, directions and amplitudes of wave components are
randomly varied for each realization.
3.9.3 To neutralize the effect of self-repetition, the duration of each simulation is limited. If
the direct counting approaches described in sections 3.3 or 3.5 are used to define rU, the
duration of each individual simulation can be arbitrary (in particular, constant). If the approach
described in section 3.4 is used, each simulation should have a common exposure time texp.
3.9.4 With the approaches described in sections 3.3 and 3.5, the effect of the
autocorrelation of large roll motions can be neutralized either by stopping a simulation after the
first encountered stability failure or by switching off the simulation timer tt and the stability failure
counter N after encountered failure until the envelope of the autocorrelation function of roll
motion reduces to a specified level. When the approach described in section 3.4 is used, this
effect is automatically neutralized since occurrence of the second, etc. stability failures in the
same simulation does not change the result (which means that a simulation can be stopped
after the first encountered stability failure as well).
3.9.5 The effect of transient hydrodynamic effects at the start of simulations can be
neutralized by switching off the counter of stability failures and the simulation timer during initial
transients or by random variation of initial conditions for each realization, see paragraph 3.7.5.
3.9.6 Numerical simulations (or model tests) are performed for arbitrary (sections 3.3
and 3.5) or constant (section 3.4) simulation time. If a stability failure is encountered in a
simulation, further time history is not considered (thus, the simulation may be stopped). After
each simulation, the number of the realization M, the number of stability failures encountered
in the simulation N (1 or 0) and duration of simulation t are recorded; the total number of
failures N is increased by N and the total simulation time tt is increased by t.

3.9.7 Based on these parameters, rU is updated after each simulation as follows:

.1 when section 3.3 is used: N* is calculated as N + 1, the maximum likelihood

estimate of the failure rate is calculated as 𝑟̂ = 𝑁/𝑡t and its conservative

I:\CIRC\MSC\1\MSC.1-Circ.1652.docx
MSC.1/Circ.1652
Annex, page 138

estimate is calculated as 𝑟̂ ∗ = 𝑁 ∗ /𝑡t . A conservative estimate of the upper

boundary of the 95%-confidence interval of failure rate is calculated as
2 ∗ ∗
𝑟U = 0.5𝜒1−0.05/2,2𝑁 ∗ 𝑟̂ /𝑁 , and the lower boundary is calculated as 𝑟L =
2
0.5𝜒0.05/2,2𝑁 𝑟̂ /𝑁;

.2 when section 3.4 is used: the upper boundary of the 95%-confidence interval
of failure probability for an exposure time texp is 𝑝U = 𝜈1 𝐹𝜈1 ,𝜈2 ;1−0.05/2 /(𝜈2 +
𝜈1 𝐹𝜈1 ,𝜈2;1−0.05/2 ) with 𝜈1 = 2(𝑁 + 1) and 𝜈2 = 2(𝑀 − 𝑁 ), for 𝑁 < 𝑀; if 𝑁 = 𝑀,
then 𝑝U = 1. The lower boundary is 𝑝L = 𝜈1 𝐹𝜈1,𝜈2 ;0.05/2 /(𝜈2 + 𝜈1 𝐹𝜈1 ,𝜈2;0.05/2 )
with 𝜈1 = 2𝑁 and 𝜈2 = 2(𝑀 − 𝑁 + 1) , for 𝑁 > 0; if 𝑁 = 0, then 𝑝L = 0. The
corresponding lower and upper boundaries of the
95%-confidence interval of the failure rate are calculated as 𝑟L =
− ln(1 − 𝑝L ) /𝑡exp and 𝑟U = − ln(1 − 𝑝U ) /𝑡exp , respectively; or

.3 when section 3.5 is used: the decorrelation time Tdc.is computed as described
in subsection 3.8. The first failure event in each record after the initial
transient is always counted; other failure events are counted only if time
between them exceeds the decorrelation time. The estimate of the failure
rate is computed with the formula in paragraph 3.5.5 and the boundaries of
the confidence interval are evaluated as described in paragraph 3.5.7.

3.10 Application examples of direct counting method

3.10.1 General

3.10.1 Application examples are based on modelling of ship motions in multiple independent
realizations of an irregular seaway and counting the number of stability failures to provide the
estimates of the boundaries of the 95%-confidence interval of the stability failure rate for the
full probabilistic direct stability assessment.

3.10.2 Application example based on approach in section 3.3

3.10.2.1 Numerical simulations (or model tests) are performed for an arbitrary duration. If a
stability failure is encountered in a simulation, the simulation is stopped. After each simulation,
the number of the realization M, the number of encountered stability failures N (1 or 0) and
the duration of simulation t (the time to failure if the realization ended with a stability failure
or the full duration of simulation otherwise) are recorded. Based on these parameters, the total
number of failures N and total simulation time tt are increased by N and t, respectively. For
a Poisson process, all simulations in which a stability failure did not occur can be treated as
one combined simulation; one stability failure is conservatively assumed at the end of the last
simulation if it did not end with a stability failure (for this purpose, N* is set to N+1).

3.10.2.2 The maximum likelihood estimate of the rate of failure is updated as 𝑟̂ = 𝑁/𝑡𝑡 , its
conservative estimate is updated as 𝑟̂ ∗ = 𝑁 ∗ /𝑡𝑡 , a conservative estimate of both the upper
2 ∗ ∗
boundary of the 95%-confidence interval of failure rate is updated as 𝑟U = 0.5𝜒1−0.05/2,2𝑁 ∗ 𝑟̂ /𝑁
2
and the lower boundary is updated as 𝑟L = 0.5𝜒0.05/2,2𝑁 𝑟̂ /𝑁.

3.10.2.3 Table 3.10.1 shows an example of results according to this procedure for parametric
roll resonance in a head seaway with a probability density fs = 10-5 (ms)-1 in the North Atlantic
wave climate of a 1700 TEU container vessel with a GM of 1.8 m using simulations of constant
one-hour duration; figure 3.10.1 shows rU and rL depending on the number of simulations M.

I:\CIRC\MSC\1\MSC.1-Circ.1652.docx
 MSC.1/Circ.1652
Annex, page 139

Table 3.10.1: Example results of direct counting procedure according to section 3.3
.
M ΔN Δt,s N* N tt,s 𝑟̂ ,1/s rU,1/s rL,1/s M ΔN Δt,s N* N tt,s 𝑟̂ ,1/s rU,1/s rL,1/s
1 1 1682.0 1 1 1682.0 5.945e-4 2.193e-3 1.505e-5 21 1 2235.5 7 7 60990.5 1.148e-4 2.141e-4 4.614e-5
2 0 3600.0 2 1 5282.0 1.893e-4 1.055e-3 4.793e-6 22 1 3505.0 8 8 64495.5 1.240e-4 2.236e-4 5.355e-5
3 1 2270.0 2 2 7552.0 2.648e-4 7.378e-4 3.207e-5 23 1 261.5 9 9 64757.0 1.390e-4 2.434e-4 6.355e-5
4 0 3600.0 3 2 11152.0 1.793e-4 6.478e-4 2.172e-5 24 1 1969.5 10 10 66726.5 1.499e-4 2.560e-4 7.187e-5
5 0 3600.0 3 2 14752.0 1.356e-4 4.897e-4 1.642e-5 25 0 3600.0 11 10 70326.5 1.422e-4 2.615e-4 6.819e-5
6 1 1025.5 3 3 15777.5 1.901e-4 4.579e-4 3.921e-5 26 0 3600.0 11 10 73926.5 1.353e-4 2.488e-4 6.487e-5
7 1 2129.5 4 4 17907.0 2.234e-4 4.896e-4 6.086e-5 27 1 1275.0 11 11 75201.5 1.463e-4 2.445e-4 7.302e-5
8 1 1111.5 5 5 19018.5 2.629e-4 5.385e-4 8.536e-5 28 0 3600.0 12 11 78801.5 1.396e-4 2.498e-4 6.968e-5
9 0 3600.0 6 5 22618.5 2.211e-4 5.159e-4 7.178e-5 29 1 2710.5 12 12 81512.0 1.472e-4 2.415e-4 7.607e-5
10 0 3600.0 6 5 26218.5 1.907e-4 4.450e-4 6.192e-5 30 1 1919.0 13 13 83431.0 1.558e-4 2.512e-4 8.297e-5
11 0 3600.0 6 5 29818.5 1.677e-4 3.913e-4 5.445e-5 31 1 3445.0 14 14 86876.0 1.611e-4 2.559e-4 8.810e-5
12 0 3600.0 6 5 33418.5 1.496e-4 3.492e-4 4.858e-5 32 0 3600.0 15 14 90476.0 1.547e-4 2.596e-4 8.460e-5
13 0 3600.0 6 5 37018.5 1.351e-4 3.152e-4 4.386e-5 33 1 1884.5 15 15 92360.5 1.624e-4 2.543e-4 9.090e-5
14 0 3600.0 6 5 40618.5 1.231e-4 2.873e-4 3.997e-5 34 0 3600.0 16 15 95960.5 1.563e-4 2.578e-4 8.749e-5
15 0 3600.0 6 5 44218.5 1.131e-4 2.639e-4 3.672e-5 35 1 634.0 16 16 96594.5 1.656e-4 2.561e-4 9.468e-5
16 1 136.5 6 6 44355.0 1.353e-4 2.631e-4 4.964e-5 36 0 3600.0 17 16 100194.5 1.597e-4 2.593e-4 9.128e-5
17 0 3600.0 7 6 47955.0 1.251e-4 2.723e-4 4.592e-5 37 0 3600.0 17 16 103794.5 1.542e-4 2.503e-4 8.811e-5
18 0 3600.0 7 6 51555.0 1.164e-4 2.533e-4 4.271e-5 38 0 3600.0 17 16 107394.5 1.490e-4 2.419e-4 8.516e-5
19 0 3600.0 7 6 55155.0 1.088e-4 2.368e-4 3.992e-5 39 0 3600.0 17 16 110994.5 1.442e-4 2.341e-4 8.239e-5
20 0 3600.0 7 6 58755.0 1.021e-4 2.223e-4 3.748e-5 40 1 1801.5 17 17 112796.0 1.507e-4 2.304e-4 8.780e-5

rU
Figure 3.10.1
An example of
results of the direct
counting procedure
rL according to
section 3.3.

4 Direct stability assessment

4.1 Ships and loading conditions used in examples

4.1.1 Five ships were used as examples: a cruise vessel, a RoPax vessel, and three
container ships of 1700, 8400 and 14000 TEU capacity. For each ship, five loading conditions
were selected: three loading conditions with small GM values and two loading conditions with
large GM values. Table 4.1.1 shows the length between perpendiculars, waterline breadth,
draught, metacentric height and linear natural roll period for the studied ships and loading
conditions (the pitch and yaw radii of inertia, ryy =√𝐼𝑦𝑦 /𝑚 and rzz =√𝐼𝑧𝑧 /𝑚 , respectively, were
set equal to LBP / 4).

Table 4.1.1: Ships and loading conditions used in examples

Ship LBP, m Bwl,m Loading LC01 LC02 LC03 LC04 LC05

condition:

Cruise 230.9 32.2 d, m 6.9

vessel GM, m 1.5 2.0 2.5 3.25 3.75
Tr, s 19.8 17.2 15.4 13.6 12.6
RoPax 175.0 29.5 d, m 5.5
vessel GM, m 3.7 4.5 5.2 5.9 6.6

I:\CIRC\MSC\1\MSC.1-Circ.1652.docx
MSC.1/Circ.1652
Annex, page 140

Tr, s 11.8 10.7 9.8 9.4 9.0

1700 TEU 159.6 28.1 d, m 9.5 5.5
container GM, m 0.5 1.2 1.9 5.75 6.75
ship Tr, s 29.3 19.4 15.4 8.8 8.2
8400 TEU 317.2 43.2 d, m 13.93 14.44 14.48 11.36
container GM, m 0.89 1.26 2.01 5.0 6.93
ship Tr, s 36.7 31.3 25.7 15.4 13.2
14000 TEU 349.5 51.2 d, m 14.5 8.5
container GM, m 1.0 2.0 3.0 9.0 12.0
ship Tr, s 38.8 27.6 22.6 13.0 11.4

4.2 Examples of full probabilistic direct stability assessment

4.2.1 For each ship and each loading condition, a full probabilistic direct stability
assessment was performed using numerical simulations of ship motions in waves to provide a
database for the development of simplified procedures. Simulations were performed at six
forward speeds, presented in table 4.2.1, for the mean zero-crossing wave periods Tz and
significant wave heights Hs covering all entries in the North Atlantic wave scatter table, IACS
Recommendation No.34 (Corr.1 Nov. 2001) (table 2.7.2.1.2 of the Interim Guidelines), and for
mean wave directions  from 0 to 180º with an increment of 10º.

Table 4.2.1: Non-dimensional forward speeds used in analysis

Ship LBP, Froude numbers
m

Cruise vessel 230.9 0.0 0.0454 0.0908 0.1362 0.1816 0.2270

RoPax vessel 175.0 0.0 0.0546 0.1093 0.1639 0.2185 0.2732
1700 container vessel 159.6 0.0 0.0481 0.0962 0.1443 0.1924 0.2405
(CV)
8400 container vessel 317.2 0.0 0.0452 0.0904 0.1356 0.1808 0.2259
(CV)
14000 container vessel 349.5 0.0 0.0427 0.0854 0.1281 0.1708 0.2135
(CV)

4.2.2 For each combination of sea state (Hs, Tz) and sailing condition (v0, ), numerical
simulations of ship motions in realizations of the same sea state were performed by using a
random variation of frequencies, directions and phases of the wave components comprising
each sea state. Each simulation was conducted for a simulation time of two hours or until the
first stability failure occurred (a stability failure was defined as the exceedance of a roll angle
of 40 º or a lateral acceleration of one g). After this, simulations were repeated in other
realizations until N = 200 stability failures were encountered.

4.2.3 Direct counting was used for each "short-term" situation (Hs, Tz, v0, ), for which the
total simulation time of 3.4106 hours was sufficient to encounter 200 stability failures. The
maximum likelihood estimate of stability failure rate was calculated as r = N/tt, eq. (3.3.6), and
2
the upper boundary of its 95%-confidence interval was calculated as 𝑟U = 0.5𝑟𝜒1−0.05/2,2𝑁 /𝑁,
eq. (3.3.9); otherwise the extrapolation of the stability failure rate over the significant wave
height was used.

I:\CIRC\MSC\1\MSC.1-Circ.1652.docx
 MSC.1/Circ.1652
Annex, page 141

4.2.4 A conservative estimate of the upper boundary 𝑟̄U of the 95%-confidence interval of
the average "long-term" stability failure rate was calculated according to the explanatory note
to paragraph 3.5.3.2.1 of the Interim Guidelines as a weighted average over the sea states
(Hs, Tz) and the sailing conditions (v0, ), of the upper boundaries rU of the 95%-confidence
intervals of the "short-term" stability failure rate:

𝑟̄ U = ∑𝑠 ∑𝜇 ∑𝑣s 𝑓s (𝐻s , 𝑇z ) ⋅ 𝑓𝜇 (𝜇) ⋅ 𝑓𝑣 (𝑣𝑠 ) ⋅ 𝑟U (𝐻s , 𝑇z , 𝜇, 𝑣s )Δ𝐻s Δ𝑇z Δ𝑣s Δ𝜇 (4.2.1)

4.2.5 In the calculations performed according to eq. (4.2.1), the mean wave directions 
and ship forward speeds vs were assumed uniformly distributed between 0 and 360º and
between zero and the maximum service speed, respectively.

4.2.6 Table 4.2.2 shows the "long-term" weighted average 𝑟̄U of the upper boundaries of
the 95%-confidence intervals of the "short-term" stability failure rate due to parametric and
synchronous roll (differentiation between parametric and synchronous roll is described in
section 4.4) for all tested ships and loading conditions; the values of 𝑟̄U not satisfying the
standard of 2.6∙10-8 1/s are indicated with bold font.

Table 4.2.2: The "long-term" weighted average 𝒓̄ 𝐔 , 1/s, of the upper boundaries of
the 95%-confidence intervals of the "short-term" stability failure rate for all tested
ships and loading conditions
Parametric roll Synchronous roll
Ship
LC01 LC02 LC03 LC04 LC05 LC01 LC02 LC03 LC04 LC05

Cruise 1.67e-06 4.74e-08 3.90e-09 5.28e-10 2.41e-10 4.59e-07 8.01e-09 6.08e-10 1.86e-10 1.23e-10
vessel
RoPax
5.47e-10 1.56e-11 1.51e-11 1.15e-11 9.49e-12 5.67e-10 9.90e-12 3.38e-11 4.83e-11 7.01e-11
vessel
1700 TEU 3.65e-05 9.51e-07 7.50e-09 3.53e-15 2.01e-15 1.41e-05 4.09e-07 3.86e-09 1.36e-12 1.54e-12
CV
8400 TEU
4.01e-06 2.44e-06 2.57e-07 9.30e-11 7.31e-13 1.64e-06 9.79e-07 1.17e-07 2.01e-07 7.24e-09
CV
14000 TEU 6.47e-05 2.00e-05 1.44e-06 7.95e-19 4.52e-19 5.50e-05 1.49e-05 9.07e-07 3.07e-16 3.46e-16
CV

4.3 Database for the development of direct stability assessment procedures for the
parametric roll, pure loss of stability and excessive acceleration failure modes

4.3.1 The example ships demonstrated stability failures due to principal parametric roll
resonance in bow waves, principal and fundamental parametric resonance in stern waves
(fundamental resonance occurs at an encounter frequency that is about the same as the
natural roll frequency) and synchronous roll in beam waves (which is relevant for the dead ship
and excessive acceleration stability failure modes). Some of the loading conditions indicated
large heel angles in following waves at high forward speeds, although their maximum speeds,
while sufficient for vulnerability to the pure loss of stability failure mode, were not sufficiently
high for strong pure loss of stability failures. Surf-riding/broaching was not found to be relevant
for any of the ships.

4.3.2 To develop and validate procedures for a direct assessment, including an

extrapolation of the stability failure rate over the wave height and probabilistic and deterministic
assessment in design situations, stability failures that were identified in simulations were
separated with respect to the stability failure modes.

I:\CIRC\MSC\1\MSC.1-Circ.1652.docx
MSC.1/Circ.1652
Annex, page 142

4.3.3 Parametric roll (specifically, principal parametric roll resonance) in bow waves was
detected in mean wave directions from 180 (head up) to about 70º off-bow. Nevertheless, in
all cases where principal parametric roll resonance in bow waves occurred, head waves led to
the largest roll motions, figure 4.3.1 (top left and top middle plots). Therefore, for parametric
roll in bow waves, an assessment in head waves was expected to detect the worst situations
and, moreover, include most relevant stability failure events. To select relevant simulation
results from the full database for validation and calibration of simplified methods for parametric
roll resonance in bow waves, three sets of reference data were generated: for wave directions
from 170 to 180, from 160 to 180, and from 150 to 180.

Fr
Fn Fr
Fn Fr
Fn
T, s

T, s

s
max
max, , max,
max, max,
max,

Fn
Fr Fr
Fn Fn
Fr
T, s

T, s

max
max, ,
max
max, , max,
max,

Figure 4.3.1:
Figure 4.3.1Colour plots
Colour of the
plots of mean three-hour
the mean maximum
three-hour roll amplitude
maximum vs. the mean
roll amplitude vs. wave
the
period (s, radial coordinate) and the mean wave direction (circumferential coordinate, 0, 90 and
mean wave period (s, radial coordinate) and the mean wave direction (circumferential
180º correspond to following waves, waves from port side and head waves, respectively) for
coordinate,
parametric roll0, resonance
90 and 180atdegrees
low GM =correspond
0.89 m (left) to
and following
medium GM waves,
= 2.01waves from and
m (middle) portfor
side
synchronous roll at high GM = 6.93 m (right) at low (top) and high (bottom) speeds for m
and head waves, respectively) for parametric roll resonance at low GM = 0.89 a
(left) and container
8400 TEU medium GM = table
ship, 2.01 m (middle)
4.1.1. and and
The black for synchronous roll at high
blue lines correspond GM
to the = 6.93
ratio m
of the
(right)
peak waveat low (top) and
encounter highto(bottom)
period the linearspeeds
naturalfor a 8400
roll periodTEU container
of 2:1 and 1:1, ship, table 4.1.1.
respectively, and
The black lines
the green and blue lines
bound thecorrespond
region with to the ratio of
a projected thelength
wave peak wave μ betweenperiod
λ / cosencounter 0.5LBP
andthe
to 2.0L BP.
linear natural roll period of 2:1 and 1:1, respectively, and the green lines bound
the region with a projected wave length λ / cos μ between 0.5LBP and 2.0LBP.
I:\CIRC\MSC\1\MSC.1-Circ.1652.docx
 MSC.1/Circ.1652
Annex, page 143

4.3.4 Parametric roll resonance in stern waves was detected in wave directions from
following (0°) up to about 80 off-stern. Unlike for parametric roll in bow waves, for which head
waves always represent the worst case, following waves were not always worst (over all stern
wave directions) for parametric roll in stern waves. Moreover, for some loading conditions at
certain forward speeds, parametric roll did not occur in following waves while it did occur in
stern-quartering waves, figure 4.3.1 (bottom left and middle).20 This means that for some ships
in some loading conditions, an assessment in following waves may not detect the possibility
of severe parametric roll in stern waves.
4.3.5 This is inconvenient since the need to address parametric roll in stern-quartering
waves in a simplified assessment means that the number of required design situations will be
significantly increased to include all wave directions from following (0) to 90 off-stern; moreover,
this means a significant increase in the number of model tests and much more advanced model
testing facilities will be required.
4.3.6 To check whether addressing parametric roll specifically in stern-quartering waves is
essential for a direct stability assessment, the results of the full assessment are plotted in
figure 4.3.2 as follows: the y-axis corresponds to the total stability failure rate over all wave
directions, whereas the x-axis corresponds to the sum of stability failure rates over parametric
roll in bow and stern waves (i.e. the sectors from 150 to 180 and 0 to 30, respectively) and
synchronous roll in beam waves (i.e. the sector from 60 to 120) for all ships and loading
conditions (differentiated by symbol type and colour) and forward speeds. Thus, the x-axis
variable neglects parametric roll in stern-quartering waves, which is included in the y-axis
variable.
4.3.7 Since the dependency in figure 4.3.2 is monotonous and sharp, the contributions from
parametric roll resonance in stern-quartering waves do not need to be additionally addressed
in a simplified direct stability assessment (unlike in operational measures): taking into account
parametric roll resonance in following waves is sufficient to represent the contributions of
parametric roll resonance in all stern wave directions. The reason is that parametric roll
resonance in stern-quartering waves becomes important with increasing forward speed (when
parametric roll decreases) whereas much larger contributions occur in following waves at low
forward speeds. Another contribution to parametric roll resonance in stern-quartering waves
comes from fundamental resonance, which is, however, weaker than the principal resonance.
w

r,
Figure 4.3.2: Total stability failure rate in all wave directions vs. the sum of the stability
failure rates due to parametric roll in bow and stern waves and synchronous roll in
beam waves; symbol type and colour differentiate ships and loading conditions

20
Shigunov, V. el Moctar, O., and Rathje, H. Conditions of parametric rolling, Proc. 10th Int. Conf. on Stability
of Ships and Ocean Vehicles, 2009.

I:\CIRC\MSC\1\MSC.1-Circ.1652.docx
MSC.1/Circ.1652
Annex, page 144

4.3.8 Therefore, for the development and validation of simplified procedures for parametric
roll resonance in stern wave directions, three comparative sets of data were generated from
the full database of assessment results, corresponding to wave directions from 0 to 10,
0 to 20 and 0 to 30º.

4.3.9 For synchronous roll in beam waves, the relevant wave directions were found from
about 40° off-bow to about 40° off-stern, depending on the forward speed, figure 4.3.1 (top
right and bottom right). However, at low forward speeds, the wave directions close to the beam
are sufficient to assess synchronous roll. Therefore, to select relevant cases for validation from
the full database of assessment results, three comparative sets of reference data were
generated, for wave directions from 80 to 100, 70 to 110 and 60 to 120.

4.3.10 Reference data for the pure loss of stability failure mode were also generated,
although this stability failure mode was especially difficult to identify since none of the selected
ships was expected to undergo severe pure loss of stability due to low maximum speeds
(although in the region of vulnerability). Three simple conditions were used: following waves;
encounter period (corresponding to peak wave period) exceeding 30 s; and a wave length,
corresponding to the peak wave period, close to the ship length.

4.4 Design situations

4.4.1 The full probabilistic direct stability assessment requires averaging of the stability
failure rate over all sea states of a relevant wave environment (which is quantified in a wave
scatter table) and all relevant sailing conditions and thus a large computational time. It can be
proposed to reduce the assessment to few combinations of sea state parameters (wave height
and period) and sailing conditions (ship forward speed and relative wave direction), referred to
as design situations. The idea is that a simplified safety criterion S can be used for norming if
the dependency of the actual criterion W (average stability failure rate) on such a criterion (a)
is monotonous and (b) shows little scatter between different ships, loading conditions and
forward speeds. The standard for this simplified criterion (further referred to as threshold to
differentiate it from the standard used for the actual criterion) can be defined using a sufficient
number of representative case studies, figure 4.4.1. Note that the exact dependency W(S) does
not matter in the practical approval and is not required, as long as it is proven that this
dependency satisfies conditions (a) and (b).

Figure 4.4.1:
The idea of the simplified safety
criterion S; W is the "true" safety
measure, e.g. a mean long-term
probability of stability failure
S

4.4.2 To verify conditions (a) and (b) in paragraph 4.4.1, the average rate of stability failures
was computed using the results of the full probabilistic assessment as given in eq. (4.2.1).
However, different forward speeds were applied and evaluated separately because the
selection of a suitable speed to be used in design situations was one of the tasks of this
investigation:

𝑊 = 𝑟̄ (ship, LC, 𝑣0 ) = ∑𝑠 ∑𝜇 𝑓s (𝐻s , 𝑇z ) ⋅ 𝑓𝜇 (𝜇) ⋅ 𝑟(𝐻s , 𝑇z , 𝜇, 𝑣0 )Δ𝐻s Δ𝑇z Δ𝜇 (4.4.1)

I:\CIRC\MSC\1\MSC.1-Circ.1652.docx
 MSC.1/Circ.1652
Annex, page 145

4.4.3 As the first step, the wave directions for design situations were assumed as 180° for
parametric roll in bow waves, 0° for parametric roll in stern waves, 90° for synchronous roll in
beam waves and 0° for pure loss of stability.

4.4.4 The second step was the selection of wave height in order to use one significant wave
height per wave period. Several approaches to the selection of sea states for design situations
were compared in documents SDC 4/5/8 and SDC 4/INF.8, including sea states according to
the steepness table from MSC.1/Circ.1200, sea states along constant steepness lines
𝐻𝑠 = const ⋅ 𝑇𝑧2 , along lines of constant density of sea state occurrence probability, and along
lines of constant normed and not normed quantiles of sea state occurrence probability. The
results presented in documents SDC 4/5/8 and SDC 4/INF.8, confirmed here, indicate that
from these options, sea states selected along the lines of constant density of sea state
occurrence probability, figure 4.4.2, provide the best correlation between W and S. Therefore,
results are shown here only for such design sea states. The lines of constant density of sea
state occurrence probability were defined using a logarithmic interpolation of probabilities (see
the explanatory note to paragraph 3.5.3.3.5 of the Interim Guidelines).

Figure 4.4.2:
Contours of constant
density of the
Hs
,

probability of the
occurrence of a
particular sea state

Tz ,
4.4.5 As the simplified criterion S in these sea states, the maximum (over design sea states)
stability failure rate r was used.

4.4.6 Figures 4.4.3 to 4.4.6 show the "long-term" average stability failure rate W vs. maximum
(over design sea states) failure rate in design sea states with probability densities 10-7, 10-6,
10-5 and 10-4(ms)-1 for all failure modes, in which each point corresponds to one ship and
combination of loading condition and forward speed. Sharp monotonic dependencies in
figure 4.3.2, concerning selection of wave directions for design situations, and in figure 4.4.3
to 4.4.6, at fs = 10-4 (ms)-1 and less, concerning the selection of wave heights for design
situations, indicate that the accuracy of the simplified criterion S is satisfactory and improves with
increasing wave steepness. Note that the required model testing or numerical simulation time
quickly reduces with increasing wave height. Therefore, it is more efficient to use design sea
states of greater steepness; however, sea states of too large steepness may be difficult to realize
in model tests or numerical simulations.

I:\CIRC\MSC\1\MSC.1-Circ.1652.docx
MSC.1/Circ.1652
Annex, page 146

fs = 1.0e-4
fs = 1.0e-6
fs = 1.0e-7

fs = 1.0e-5

Figure 4.4.3: The W(S) for design situations for parametric roll in bow waves: mean long-
term stability failure rate W(ship,LC,v0), 1/s, y-axis, in wave directions from 150° to 180°
vs. simplified criterion, 1/s, x-axis – short-term stability failure rate in head waves,
maximum over design sea states along lines with sea state probability density fs of (top
left to bottom right) 10-7, 10-6, 10-5 and 10-4 (ms)-1

I:\CIRC\MSC\1\MSC.1-Circ.1652.docx
 MSC.1/Circ.1652
Annex, page 147

S
fs = 1.0e-4
fs = 1.0e-6

, ,
W W
S
S
fs = 1.0e-7

fs = 1.0e-5

, ,
W W
Figure 4.4.4: The W(S) for design situations for parametric roll in stern waves: mean
long-term stability failure rate W(ship,LC,v0), 1/s, y-axis, in wave directions 0° to 30° vs.
simplified criterion, 1/s, x-axis – short-term stability failure rate in following waves,
maximum over design sea states along lines with sea state probability density fs of (top
left to bottom right) 10-7, 10-6, 10-5 and 10-4 (ms)-1

I:\CIRC\MSC\1\MSC.1-Circ.1652.docx
MSC.1/Circ.1652
Annex, page 148

S
S
fs = 1.0e-6

fs = 1.0e-4

,
W W
S
S
fs = 1.0e-7

fs = 1.0e-5

, ,
W W

Figure 4.4.5: W(S) for synchronous roll in beam waves: mean long-term stability failure
rate W(ship,LC,v0), 1/s, y-axis, in wave directions from 60° to 120° vs. simplified criterion,
1/s, x-axis – short-term stability failure rate at  = 90°, maximum over design sea states
along lines with sea state probability density fs equal to (from top left to bottom right)
10-7, 10-6, 10-5 and 10-4 (ms)-1

I:\CIRC\MSC\1\MSC.1-Circ.1652.docx
 MSC.1/Circ.1652
Annex, page 149

fs = 1.0e-7 fs = 1.0e-6

W
W

,
,

S S

fs = 1.0e-5 fs = 1.0e-4

W
W

,
,

S S

Figure 4.4.6: The W(S) for pure loss of stability: mean long-term stability failure rate
W(ship,LC,v0), 1/s, y-axis, vs. simplified criterion, 1/s, x-axis – short-term stability failure
rate in following waves, maximum over design sea states with occurrence probability
density fs of (top left to bottom right) 10-7, 10-6, 10-5 and 10-4 (ms)-1

4.4.7 Results presented above allow a reduction in the number of assessment situations
due to using one wave direction per failure mode (which is a reduction factor of about 19) and
one wave height per wave period (which is a reduction factor of several orders of magnitude
since assessment at low wave heights requires very long simulations, if feasible at all). Another
reduction possibility is the selection of a suitable forward speed: if, for example, only one speed
needs to be used per failure mode, this will lead to a reduction of the number of assessment
situations by about one order of magnitude (for some stability failure modes) and will allow
significant simplifications in numerical simulations or the set-up of model tests.

4.4.8 For the dead ship condition and excessive accelerations stability failure modes, only
zero forward speed is applied in the full assessment anyway. For the pure loss of stability
failure mode, the rate of stability failures increases with increasing speed and, therefore, the
maximum speed should be used. To select the forward speed for design situations for the
parametric roll stability failure mode, figure 4.4.7 (left) shows the failure rate (maximum over
wave periods) due to parametric roll in head waves along the line fs = 10-5 (ms)-1 as a function
of Froude number. Each plot corresponds to one ship and each line corresponds to one loading
condition.

LBP
LBP GM
LBP LBP GM
LBP GM
LBP GM
LBP GM

LBP GM
LBP GM
LBP GM
Fn Fn

I:\CIRC\MSC\1\MSC.1-Circ.1652.docx
MSC.1/Circ.1652
Annex, page 150

LBP GM
LBP GM LBP GM
LBP GM LBP GM
LBP GM

Fn Fn

10-1 10-510-1 10-3

Lpp=159.6 m, GM=0.5 m
-6
10
Lpp=159.6 m, GM=1.2 m 10-5
10-2
10-2 Lpp=159.6 m, GM=1.9 m 10-7
Lpp=159.6 m, GM=5.8 m 10-7
Failure Rate, 1/s

Failure Rate, 1/s

Failure Rate, 1/s

Failure Rate, 1/s

-8
10 -3
Lpp=159.6 m, GM=6.8 m 10

10-3 10-9 10-9

-4
10
-10 Lpp=175.0 m, GM=3.7 m
LBP GM 10
LLpp=175.0 GM 10-11
BP m, GM=0.5
Lpp=159.6 m, GM=5.2 m
m
10-4
LBP GM
10-11 LLpp=175.0
BP GM
m, GM=1.2
GM=5.9 m
m
LBP GM 10-5 Lpp=159.6
LLpp=175.0
m,
GM 10-13
BP m, GM=1.9
GM=6.6 m
m
LBP GM 10
-12
Lpp=159.6 m,
LBP GM Fr Fn FrFn
Fr
10-5 10-1310-6 10-15
0 0.02 0.04 0.06 0.08 0.1 0.12 0.14 0.16 0.18 0.2 0.22 0.24 0 0 0.02
0.02 0.04
0.04 0.06
0.06 0.08 0.1
0.1 0.12
0.12 0.14
0.14 0.16
0.16 0.180.18 0.2 0.20.22 0.22
0.24 0 0.02 0.04 0
10-2 10-210-2 10-2
LBP GM Lpp=317.2 m, GM=0.9 m
10-3
LBP GM Lpp=317.2 m, GM=1.3 m 10
-3

LBP GM 10-4 Lpp=317.2 m, GM=2.0 m

10-3 10-4
10
-4
LBP GM Lpp=317.2 m, GM=5.0 m
LBP GM Lpp=317.2 m, GM=6.9 m
Failure Rate, 1/s

Failure Rate, 1/s

Failure Rate, 1/s

Failure Rate, 1/s

-5
-6 10
10
10-5
10-4 10-6
10-6 -8
10 -7
10
Lpp=230.9 m, GM=1.5 m Lpp=230.9 m, GM=1.5 m
-7
10 Lpp=230.9 m, GM=2.0 m -5
10 LBP GM Lpp=230.9 m, GM=2.0 m 10-8
10-10
Lpp=317
Lpp=230.9 m, GM=2.5 m
10-8
LBP GM Lpp=230.9 m, GM=2.5 m -9 Lpp=317
Lpp=230.9 m, GM=3.3 m 10
Fr Fnm LBP GM FrFn
Fr Lpp=317
Lpp=230.9 m, GM=3.8
10-9 10-1210-6 10-10
0 0.02 0.04 0.06 0.08 0.1 0.12 0.14 0.16 0.18 0.2 0 0 0.020.02
0.04 0.04
0.06 0.06
0.08 0.1
0.08 0.120.1 0.140.12
0.16 0.14
0.18 0.16
0.2 0.22
0.18 0.240.2 0 0.02 0.04 0.0
10-1 10-1

-2
10
10-2
10-3
Failure Rate, 1/s

Failure Rate, 1/s

-4
10 -3
10

10-5
-4
-6 10
10
LLpp=349.5
BP GM
m, GM=1.0 m LLpp=349.5 m, GM
GM=1.0 m
10-7 BP
LLpp=349.5
BP GM
m, GM=2.0 m 10-5 L
Lpp=349.5
BP m, GM
GM=2.0 m
10
-8 LLpp=349.5
BP GM
m, GM=3.0 m LLpp=349.5
BP m, GM
GM=3.0 m
Fn
Fr Fr
Fn
10-9 10-6
0 0.02 0.04 0.06 0.08 0.1 0.12 0.14 0.16 0.18 0.2 0.22 0 0.02 0.04 0.06 0.08 0.1 0.12 0.14 0.16 0.18 0.2 0.22

Figure 4.4.7: Maximum (over all wave periods) short-term stability failure rate, 1/s, at
a wave height corresponding to sea state probability density fs = 10-5 (ms)-1 (y-axis) vs.
Froude number (x-axis) in head (left) and following (right) waves for (from top to
bottom) the 1700 TEU container ship, the RoPax vessel, the cruise vessel and the 8400
and 14000 TEU container ships; one line corresponds to one loading condition

4.4.9 The results show that for all loading conditions with a high failure rate, the failure rate
due to parametric roll decreases with increasing forward speed, which is due to broadening of
the encounter wave spectrum with increasing forward speed in bow waves and increasing roll
damping with increasing forward speed (according to operational experience, parametric roll
accidents in bow waves have usually happened at low forward speed). For the RoPax vessel
in all loading conditions and the cruise vessel in two loading conditions with the largest GM,
the stability failure rate increases with increasing forward speed; however, the stability failure
rate for these cases is very small anyway. Therefore, it seems appropriate to use only zero

I:\CIRC\MSC\1\MSC.1-Circ.1652.docx
 MSC.1/Circ.1652
Annex, page 151

forward speed in design situations for parametric roll in bow waves (if zero speed is difficult to
implement in simulations or model tests, as low as practicable forward speed can be applied
instead).

4.4.10 For parametric roll in stern waves, figure 4.4.7 (right) shows a more complex
dependency of the failure rate on the Froude number in design sea states in following waves.
This is due to more complex relationship between the wave frequency and the encounter
frequency in stern waves and thus a more complex behaviour of the encounter wave spectrum.
However, in all cases with a large stability failure rate, a simplified assessment only at zero
forward speed will either not introduce any non-conservative error or will be conservative, thus
zero (or as low as practicable) forward speed seems appropriate also for parametric roll in
following waves.

4.4.11 Zero forward speed in head or following waves may not be realistic in severe seas
because of the difficulties for a ship (with a usual steering system) to keep course. However,
this assumption is acceptable as a practical simplification for the roll motion assessment
procedure (which, however, will require some adjustment of the test set-up).

4.5 Deterministic direct stability assessment

4.5.1 A difficulty of a probabilistic assessment is the need to encounter stability failures in

simulations or model tests, which may require long durations of simulations or model tests for
relevant cases. An appealing idea is to combine design situations with non-probabilistic
criteria, e.g. mean maximum roll amplitude per given exposure time, mean roll amplitude. The
idea is still the same as shown in figure 4.4.1: as long as a simplified criterion S is monotonically
related to the true safety measure W (average safety failure rate) and the scatter between
ships, loading conditions and forward speeds is small, the simplified criterion can be directly
used for norming and its acceptance threshold can be defined directly using the results of a
deterministic assessment for a sufficient number of representative sample cases.

4.5.2 In documents SDC 4/5/8 and SDC 4/INF.8, this method was verified for roll in beam
seas to address dead ship condition and excessive acceleration stability failures. Two
deterministic criteria (the mean roll amplitude and the mean three-hour maximum roll
amplitude) were tested for different ships, loading conditions and forward speeds in irregular
beam seas. The latter criterion has shown significantly better results than the former one,
therefore, it was used here in combination with design situations to develop the deterministic
assessment.

4.5.3 The same ships and loading conditions were used as listed in table 4.1.1. In the first
step, different forward speeds were evaluated separately. For each failure mode, one wave
direction was used for design situations: for parametric roll in bow waves, 180°; for parametric
roll in stern waves, 0°; for synchronous roll in beam waves, 90° (associated with both the
dead-ship condition and excessive acceleration failure modes); and for pure loss of stability, 0°.
As in the previous section, the sea states selected along the lines of constant density of seaway
occurrence probability, fs, figure 4.4.2, were used as design sea states.

4.5.4 As the simplified criterion S, the maximum (over design sea states) of the mean
three-hour maximum roll amplitude was used. To define it, three-hour numerical simulations were
performed in 50 realizations of each sea state by a random variation of frequencies, directions and
phases of the components of the modelled seaway. The maximum roll amplitude was defined from
each simulation and averaged over 50 maxima. An evaluation of the three-hour maximum roll
amplitude is impossible in cases with capsizing, since the roll amplitude is not defined.
To distinguish such cases in plots, the mean three-hour maximum roll amplitude is shown for these
cases as 60° for ease of identification since the mean three-hour maximum roll amplitude never
achieved 60° in simulations where capsizing did not occur for the considered ships.

I:\CIRC\MSC\1\MSC.1-Circ.1652.docx
MSC.1/Circ.1652
Annex, page 152

4.5.5 Figures 4.5.1 to 4.5.4 show the average "long-term" stability failure rate W vs. mean
three-hour maximum roll amplitude for parametric roll in bow (figure 4.5.1) and stern
(figure 4.5.2) waves, synchronous roll in beam waves (figure 4.5.3) and pure loss of stability
(figure 4.5.4). The mean three-hour maximum roll amplitude is defined as the maximum over
all wave periods in design sea states with the occurrence probability density fs = 10-7, 10-6, 10- 5
and 10-4 (ms)-1 for wave directions 180° (for parametric roll in bow waves), 0° (parametric roll
in stern waves) and 90° (synchronous roll in beam waves), and for combined conditions of 0°
wave direction, encounter peak wave period more than 30 s and wave length equal to ship
length (for pure loss of stability). Each point corresponds to one ship in one loading condition
at one forward speed.

Figure 4.5.1:Parametric roll in bow waves: the long-term average failure rate
W(ship,LC,v0), 1/s, y-axis, in wave directions from 150° to 180° vs. the deterministic
criterion, x-axis – mean three-hour maximum roll amplitude in head waves, maximum
over design sea states along lines with sea state occurrence probability density fs (from
top left to bottom right) 10-7, 10-6, 10-5 and 10-4 (ms)-1; each point means one ship
(different symbols), one loading condition and one forward speed; points with mean
three-hour maximum roll amplitude 60o indicate capsizing

I:\CIRC\MSC\1\MSC.1-Circ.1652.docx
 MSC.1/Circ.1652
Annex, page 153

Figure 4.5.2: Parametric roll in stern waves: long-term average stability failure rate
W(ship,LC,v0), 1/s, y-axis, in wave directions from 0 to 30º vs. deterministic criterion,
x-axis – mean three-hour maximum roll amplitude in following waves, maximum over
design sea states along lines with sea state occurrence probability density fs (from top
to bottom) 10-7, 10-6, 10-5 and 10-4 (ms)-1; each point means one ship (different symbols),
one loading condition and one forward speed; points with a mean three-hour maximum
roll amplitude of 60o indicate capsizing

Figure 4.5.3: Synchronous roll in beam waves: long-term average stability failure rate
W(ship,LC,v0), 1/s, y-axis, in wave directions from 60 to 120º vs. deterministic criterion,
x-axis – mean three-hour maximum roll amplitude at  = 90o, maximum over design sea
states along lines with sea state occurrence probability density fs (from top left to
bottom right) 10-7, 10-6, 10-5 and 10-4 (ms)-1; each point means one ship (different
symbols), one loading condition and one forward speed; points with a mean three-hour
maximum roll amplitude of 60o indicate capsizing

I:\CIRC\MSC\1\MSC.1-Circ.1652.docx
MSC.1/Circ.1652
Annex, page 154

Figure 4.5.4: Pure loss of stability in following waves: long-term average stability failure
rate W(ship,LC,v0), 1/s, y-axis, vs. deterministic criterion, x-axis – mean 3h maximum roll
amplitude in following waves, maximum over design sea states along lines with sea
state occurrence probability density fs (top left to bottom right) 10-7, 10-6, 10-5 and
10-4 (ms)-1; each point means one ship (different symbols), one loading condition and
one forward speed; points with a mean three-hour maximum roll amplitude equal to 60o
indicate cases with capsizing

4.5.6 The correlation between the average long-term stability failure rate and the mean
three-hour maximum roll amplitude in design sea states is very poor, especially in cases with
small roll motions. Although increasing roll motions significantly improve this correlation, they
also lead to capsizings, which make the evaluation of the deterministic criterion impossible.

4.5.7 To select forward speeds to be used in the design situations, the mean three-hour
maximum roll amplitude in head and following (for parametric roll), beam (synchronous roll)
and following (pure loss of stability) waves in sea states with a probability density 10 -5 (ms)-1
is plotted vs. forward speed in figures 4.5.5 to 4.5.8. The results are similar to the speed
dependency of the probabilistic criterion: for parametric roll in head waves and for synchronous
roll, decreasing forward speed maximizes the three-hour maximum roll amplitude, whereas for
pure loss of stability, the greatest roll responses correspond to the maximum forward speed.
For parametric roll in following waves, the maximum roll may either decrease or increase with
increasing forward speed; however, for the most critical loading conditions, low forward speeds
lead to the maximum roll response. Therefore, similar recommendations are given for the
selection of forward speed as those given for the probabilistic direct stability assessment in
design situations.

I:\CIRC\MSC\1\MSC.1-Circ.1652.docx
 MSC.1/Circ.1652
Annex, page 155

Vs, m/s Vs

Vs

Vs Vs

Vs

Vs

Figure 4.5.5: The mean 3h maximum roll Figure 4.5.6: The mean 3h maximum roll
amplitude due to parametric roll in head amplitude due to parametric roll in
waves in sea states with fs=10-5 (ms)-1 vs. following waves in sea states with fs=10-5
forward speed. each plot corresponds to (ms)-1 vs. forward speed. each plot
one ship; different symbols correspond corresponds to one ship; different
to different loading conditions. symbols correspond to different loading
conditions.

I:\CIRC\MSC\1\MSC.1-Circ.1652.docx
MSC.1/Circ.1652
Annex, page 156

Vs Vs

Vs Vs

Figure 4.5.7:The mean 3h maximum roll amplitude due to synchronous roll in beam
waves in sea states with fs=10-5 (ms)-1 vs. forward speed; each plot corresponds to
one ship; different symbols correspond to different loading conditions

Vs Vs

Figure 4.5.8: The mean 3h maximum roll amplitude due to pure loss of stability in
following waves in sea states with fs=10-5 (ms)-1 vs. forward speed. each plot
corresponds to one ship; different symbols correspond to different loading
conditions.

4.6 Definition of standard and thresholds

4.6.1 To distinguish acceptable from not acceptable loading conditions, an acceptance

standard should be defined for the average long-term stability failure rate W, as well as
consistent acceptance thresholds for the criteria used in the simplified assessment procedures
(i.e. for the mean stability failure rate r and the mean 3h maximum roll amplitude 3h in design
situations) for all stability failure modes.

4.6.2 Consider the relationship between the standard for the "long-term" average stability
failure rate in the full probabilistic direct stability assessment and the average stability failure
rate in the real operation (i.e. the actual safety level). This relationship is uncertain, whereas
the average stability failure rate in the full probabilistic direct stability assessment may differ
from the average failure rate in real operation by a few orders of magnitude due to several
factors:

I:\CIRC\MSC\1\MSC.1-Circ.1652.docx
 MSC.1/Circ.1652
Annex, page 157

.1 the full probabilistic direct stability assessment is conducted in a rather

severe North Atlantic wave environment while the mean safety level relates
to the worldwide operation;

.2 ship routeing and heavy-weather avoidance are not considered;

.3 the assessment is performed separately for each loading condition, thus the
worst loading condition (which may never occur in practice) has a
dominating effect on the results; and

.4 unsafe forward speeds and courses, avoided in reality in heavy weather,

produce dominating (by few orders of magnitude) contributions to the
long-term stability failure rate. For example, principal parametric roll
resonance in following waves, especially at low speeds, provides dominating
contributions to the failure rate for loading conditions with a low GM,
figure 4.6.1, whereas in reality, such situations are avoided (because of the
possibility of stern slamming, low freeboard) or impossible (because of an
inability to maintain course).
w

w
,

r, r,

Figure 4.6.1: Contributions to the average long-term stability failure rate W (1/s,
y-axis) from principal parametric roll resonance in bow (left) and stern (right) waves
(1/s, x-axis); symbol types and colours differentiate ships and loading conditions from
table 4.1.1

4.6.3 To estimate the standard for the long-term average stability failure rate for the full
probabilistic direct stability assessment, data from FSA studies for container vessels
(MSC 83/INF.8), LNG carriers (MSC 83/INF.3), crude oil tankers (MEPC 58/INF.2), cruise
ships (MSC 85/INF.2), RoPax vessels (MSC 85/INF.3) and general cargo vessels
(MSC 88/INF.8) can be considered. According to the aforementioned FSA studies, losses due
to foundering are reported for container ships and general cargo vessels (9.7810-4
and 5.1010-3 losses per ship per year, respectively). Since the second generation intact
stability criteria address not only total losses but also other stability failures, another relevant
figure is the average frequency of accidents due to heavy weather, which is reported for
container ships and LNG carriers as 2.6410-3 and 3.2010-3 accidents per ship per year,
respectively. The lower of these figures is used here to define the standard, which is 2.6410-3
stability failures per ship per year.

4.6.4 The value 2.6410-3 stability failures per ship per year corresponds to the mean time
to failure of 378.8 years for one ship. To relate this number to direct stability assessment
results, a number of considerations should be taken into account:

.1 the direct stability assessment is performed for each loading condition, even for
those in which the ship may rarely sail in reality; to consider this, an assumed
factor of 0.1 is applied to the time to stability failure obtained above;

I:\CIRC\MSC\1\MSC.1-Circ.1652.docx
MSC.1/Circ.1652
Annex, page 158

.2 to account for time in port and maintenance, time in port is assumed to

be 20% of the total design life; therefore, a factor of 0.8 is applied;

.3 the direct stability assessment is performed for the severe North Atlantic
wave environment, whereas the results of the FSAs relate to worldwide
service; to consider a reduced time in such a severe wave environment in
reality, an assumed reduction factor of 0.2 is applied;

.4 the direct stability assessment assumes that a ship randomly encounters

sea states according to their occurrence frequencies in the wave scatter
table; however, in reality, ships use weather routeing and heavy-weather
avoidance; to account for this, a reduction factor of 0.2 is applied; and

.5 when these factors are applied to the time to stability failure of 378.8 years,
a stability failure per ship of 1.2 years is obtained, which means that the
standard for the average long-term stability failure rate (that should be
ensured by a direct stability assessment and operational measures)
2.610-8 1/s.

4.6.5 A similar study on defining standards for the vertical bending moment 21 shows that
setting a standard using results of a numerical analysis as once per the design life in the North
Atlantic wave environment leads to too conservative results compared to the existing,
sufficiently safe, fleet and rules of classification societies. To harmonize the results of a direct
assessment with classification rules, a "routeing factor" of 0.85 was proposed, which should
be applied as a multiplication factor to the wave heights. For comparison, the present results
of the full probabilistic direct stability assessment were re-evaluated with 0.85-scaled wave
heights. This leads to the standard for the average long-term stability failure rate
of 1.410-8 1/s, which is close to the estimate given in paragraph 4.6.4.5.

4.6.6 To define the threshold for the short-term stability failure rate r in design situations
based on this standard, a number of considerations should be noted:

.1 the results of a direct stability assessment with respect to the dead ship
condition stability failure mode in design situations were compared with the
results of an assessment of the Weather Criterion of the 2008 Intact Stability
Code for reference; this led to the upper and lower estimates for the design-
situation threshold for r shown in table 4.6.1;

.2 the threshold for the stability failure rate in design situations should not be
too high (otherwise it will be difficult to minimize the influence of initial
conditions) or too low (otherwise testing or simulation time will be too large);
thus, an appropriate threshold should be between about 10 -4 1/s
and 10-3 1/s; and

.3 according to paragraph 3.7.24, parametric roll increasingly deviates from the

Poisson process assumption for a failure rate exceeding about one failure
in two hours, i.e. 1.4∙10-4 1/s.

21
Derbanne, Q., Storhaug, G., Shigunov, V., Xie, G., and Zheng, G. Rule formulation of vertical hull girder
wave loads based on direct computation, Proc. PRADS 2016, 4-8 September, Copenhagen, 2016.

I:\CIRC\MSC\1\MSC.1-Circ.1652.docx
 MSC.1/Circ.1652
Annex, page 159

Table 4.6.1: Estimates of the lower and upper bounds for the design-situation threshold r
from a comparison with the 2008 IS Code Weather Criterion
fs, (ms)-1 10-2 10-3 10-4 10-5 10-6 10-7
-34 -15 -10 -8
lower 1.810 1.010 2.810 7.510 1.810-6 1.410-5
upper 1.710-9 2.810-5 4.710-4 1.710-3 3.510-3 5.610-3

4.6.7 Figure 4.6.2 plots the dependencies of the average long-term stability failure rate W
on the short-term stability failure rate in design situations r for parametric roll and synchronous
roll and pure loss of stability from figure 4.4.3 to 4.4.6 for fs = 10-2 to 10-7 (ms)-1 together with
the bounds for the W-standard and the r-threshold according to the considerations given in
paragraph 4.6.4. The bounds for the W-standard are transferred into the bounds for the
r-threshold and the other way around using these dependencies. In figure 4.6.2, colours of the
bounding rectangles correspond to the FSA studies (red), the comparison with the Weather
Criterion (green) and the considerations given in paragraph 4.6.6.2 (blue). The overlapping
areas, indicated with arrows in figure 4.6.2 and shown in increased resolution in figure 4.6.3,
indicate the possibility of a non-contradicting combination of these bounds if the direct stability
assessment in design situations is performed in design sea states with a probability density
fs = 10-7 to 10-5 (ms)-1. For greater or lower values of fs, the areas corresponding to various
estimates do not overlap. However, for design sea states with the probability density
of 10-4 (ms)-1, the only limitation is a long simulation time, which is not a crucial problem for
some numerical methods.

Figure 4.6.2: Combined dependencies of the long-term average stability failure rate W
on the short-term design-situation stability failure rate r for all stability failure modes
from figure 4.4.3 to 4.4.6 for fs (from top left to bottom right) of 10-2 to 10-7 (ms)-1 together
with bounds according to paragraph 4.4.6; the arrows indicate overlapping areas where
a non-contradicting combination of bounds is possible

I:\CIRC\MSC\1\MSC.1-Circ.1652.docx
MSC.1/Circ.1652
Annex, page 160

Figure 4.6.3: The definition of the standard and the threshold (increased resolution plots
from figure 4.6.2); the thick line rectangle indicates overlapping area

4.6.8 According to the analysis results shown in figure 4.6.3, the value of 2.610-8 1/s seems
suitable as a conservative estimate for the standard for the long-term average stability failure
rate W in the full probabilistic direct stability assessment.

4.6.9 For the threshold for the mean short-term stability failure rate r in design situations,
the following values seem suitable: one stability failure in 20h in design sea states with
fs = 10-4 (ms)-1; one stability failure in two hours in design sea states with fs = 10-5 (ms)-1; and
one stability failure in 40 minutes in design sea states with fs = 10-6 (ms)-1. However, one
stability failure in 40 minutes contradicts the requirement given in paragraph 4.6.6.3.
Therefore, it seems suitable to use as design sea states those with the probability density of
occurrence between fs = 10-4 (ms)-1 (with the short-term threshold for the mean stability failure
rate r equal to one in 20h) and 10-5 (ms)-1 (with the threshold of one stability failure in
two hours). The latter seems more practicable since it means that there is less simulation time
required for the direct stability assessment.

4.6.10 Table 4.6.2 shows the significant wave height vs. the mean zero-crossing wave period
at fs = 10-5 for unrestricted service, i.e. wave scatter table from IACS Recommendation No.34
(Corr.1 Nov. 2001) (also presented in the Interim Guidelines, table 2.7.2.1.2).

Table 4.6.2: Significant wave height depending on the mean zero-crossing wave period for sea
states with an occurrence probability density of 10-5 (ms)-1 according to IACS
Recommendation No.34 (Corr.1 Nov. 2001)

Tz, s 4.5 5.5 6.5 7.5 8.5 9.5 10.5 11.5 12.5 13.5 14.5 15.5 16.5
Hs, m 2.8 5.5 8.2 10.6 12.5 13.8 14.6 15.1 15.1 14.8 14.1 12.9 10.9

4.6.11 To define the threshold for the mean three-hour maximum roll amplitude φ3h for the
deterministic direct stability assessment in design situations, a different approach was used:

.1 the threshold for φ3h was set to one half of the heel angle in the definition of
the stability failure to avoid capsizing in model tests or numerical
simulations;

I:\CIRC\MSC\1\MSC.1-Circ.1652.docx
 MSC.1/Circ.1652
Annex, page 161

.2 the maximum value of the average long-term stability failure rate W was
found over all ships, loading conditions and forward speeds satisfying the
chosen φ3h-threshold in design situations; and
.3 in this way, the maximum value of the average long-term stability failure rate
W becomes a function of the probability density fs defining the design sea
states in which the deterministic direct stability assessment is performed.
4.6.12 Figure 4.6.4 displays a plot of the the resulting average long-term stability failure rate
W (y-axis) as a function of the probability density fs defining design sea states (x-axis). To
satisfy the selected standard for the average long-term stability failure rate of 2.610-8 1/s
(shown as a horizontal dashed line), the design sea states with the probability density
fs = 710-5 (ms)-1 (circle) should be used. Table 4.6.3 shows the significant wave height vs. the
mean zero-crossing wave period for fs = 710-5 (ms)-1 for unrestricted service (IACS
Recommendation No.34 (Corr.1 Nov. 2001) wave scatter table and the Interim Guidelines,
table 2.7.2.1.2).

Figure 4.6.4
Definition of
design sea states
w

for a deterministic
assessment

f
s
Table 4.6.3: The significant wave height vs. the mean zero-crossing wave period for a
sea state probability density of 710-5 (ms)-1 in the North Atlantic wave environment
Tz,s 4.5 5.5 6.5 7.5 8.5 9.5 10.5 11.5 12.5 13.5 14.5 15.5
Hs,m 2.0 4.4 6.9 9.1 10.9 12.1 12.8 13.1 13.0 12.5 11.3 9.0

4.6.13 The proposed standard for the average long-term stability failure rate W and the
thresholds for the mean short-term failure rate r and the mean three-hour maximum roll
amplitude φ3h in design situations can be further adjusted using a full probabilistic direct
stability assessment or a direct stability assessment in design situations for ships that have
experienced accidents in the loading conditions during those accidents and applying
figure 4.6.2 or 4.6.4 to scale the long-term standard into the short-term threshold or vice versa.
4.7 Example procedures for a probabilistic direct stability assessment in design
situations
4.7.1 A probabilistic direct stability assessment in design situations does not require
statistical extrapolation, therefore this example procedure uses direct counting. Since an
assessment in design situations (as well as probabilistic operational guidance) uses a short-
term acceptance standard, a more efficient acceptance check (i.e. check whether the upper
boundary of the 95%-confidence interval of the failure rate does not exceed a standard) can
be used with the direct counting procedure in paragraph 3.5.4.4.1 of the Interim Guidelines:
namely, a decision on acceptance can be made during a running simulation, see an example
procedure in paragraph 4.7.10. The unacceptance check is the same as that given in the
general direct counting procedure: the lower boundary of the 95%-confidence interval of the
failure rate is updated after each encountered failure, and a loading condition is considered as
unacceptable once this boundary exceeds the standard in at least one design situation.

I:\CIRC\MSC\1\MSC.1-Circ.1652.docx
MSC.1/Circ.1652
Annex, page 162

4.7.2 An example procedure using the direct counting approach in section 3.3 of this
appendix is used for the estimation of the failure rate and its confidence interval. The procedure
uses numerical simulations or model tests in multiple independent realizations of the same sea
state, which is assumed to be modelled as a sum of harmonic components, eq. (3.6.1). To
generate such realizations, the phases i are randomly selected in the interval [0, 2 ), whereas
the frequencies i and directions i can be either assumed constant in all simulations or
randomly selected (within ranges i and i, respectively) for each simulation. The
amplitudes ai can be assumed to be constant or randomly varied. To generate random values,
a pseudo-random number generator can be used.

4.7.3 To ensure that stability failures in numerical simulations or model tests satisfy the
assumptions of a Poisson process, the procedure neutralizes self-repetition effects (by limiting
duration of each simulation), transient hydrodynamic effects at the beginning of simulations
(by switching off the counter of stability failures and simulation timer during initial transients)
and autocorrelation of large roll motions (by stopping a simulation after the first encountered
stability failure). Apart from a limited duration and stopping after the first stability failure,
duration of each simulation is arbitrary.

4.7.4 For the acceptance check, the upper boundary rU of the 95%-confidence interval of
the failure rate should not exceed a standard ; combined with eq. (3.3.9) with  = 0.05 and
2
definition 𝛽1 (𝑁) = 0.5𝜒1−0.05/2,2𝑁 /𝑁, this provides the following condition:

𝑇̂ > 𝑇̂A ≡ 𝛽1 (𝑁)/𝜆 (4.7.1)

4.7.5 Eq. (4.7.1) is not convenient when the stability failure rate is low since it requires the
encounter of at least one stability failure while the time until the first failure may be too long for
acceptable loading conditions (and even for not acceptable ones away from resonance).
Obviously, continuing simulations in such cases is senseless since the absence of a failure
during a long time is a sign of an acceptable safety level. To address such cases, eq. (4.7.1)
can be used assuming (conservatively) that N = 1 and thus 𝑇̂ = 𝑡; this leads to the conclusion
that a simulation can be stopped (with the acceptance decision) when the simulation time
without failure achieves

t  tA  1(1)/ (4.7.2)

4.7.6 To extend this idea on the second and further stability failures, introduce in eq. (4.7.1)
the definition eq. (3.3.1), rewritten as 𝑇̂ = 𝑇̂𝑁−1 ⋅ (1 − 1/𝑁) + 𝑇𝑁 /𝑁, where 𝑇̂𝑁−1 is the sample
mean time to failure defined from the already encountered N − 1 stability failures and assume,
conservatively, that TN = t. The result is that a simulation can be stopped with the acceptance
decision in the considered design situation before the N-th stability failure is encountered when
the accumulated simulation time without failure after the N − 1-th stability failure achieves

𝑡 ≥ 𝑡A ≡ 𝛽1 (𝑁) ⋅ 𝑁/𝜆 − 𝑇̂𝑁−1 ⋅ (𝑁 − 1) (4.7.3)

4.7.7 When N = 1, eq. (4.7.3) automatically reduces to t  tA  1(1)/, i.e. eq. (4.7.2).
Therefore, only eq. (4.7.3) is required as an acceptance check for the first and further
simulations.

4.7.8 To save simulation time, it is possible to define as unacceptance, the check of the
requirement that the lower boundary rL of the 95%-confidence interval of the failure rate
exceeds the standard , which, combined with eq. (3.3.9) with =0.05 and definition 𝛽2 (𝑁) =
2
0.5𝜒0.05/2,2𝑁 /𝑁, leads to the following condition:

I:\CIRC\MSC\1\MSC.1-Circ.1652.docx
 MSC.1/Circ.1652
Annex, page 163

𝑇̂ < 𝑇̂F ≡ 𝛽2 (𝑁)/𝜆 (4.7.4)

4.7.9 To illustrate the progressive uncertainty reduction with an increasing number of

encountered failures, figure 4.7.1 shows 1 and 2 vs. N. A calculation of the function 𝜒𝑝,𝑓
2
,
required to define 1 and 2, is described in paragraph 3.3.6 of this appendix.

ß
1

Figure 4.7.1 Functions

1(N) and 2(N)

ß
2 N

4.7.10 The described example procedure for probabilistic assessment in design situations
can be summarized as follows:

.1 numerical simulations or model tests are carried out in multiple independent

realizations of the same sea state by a random variation of phases (and,
possibly, frequencies, directions and amplitudes) of harmonic waves that
discretize the wave energy spectrum. Simulations or tests can be carried out
for an arbitrary duration until the maximum duration or first stability failure;
the counter of stability failures and the simulation timer are switched off
during the initial transients;

.2 after each simulation or test, the total number of failures N is increased by

N (1 if ended with a stability failure and 0 otherwise), the total simulation
time tt is increased by the duration of simulation t and the sample mean
time to failure is updated as 𝑇̂ = 𝑡𝑡 /𝑁;

.3 if a simulation achieves time 𝑡A = 𝛽1 (𝑁) ⋅ 𝑁/𝜆 − 𝑇̂𝑁−1 ⋅ (𝑁 − 1) without a

2
stability failure with 𝛽1 (𝑁) = 0.5𝜒1−0.05/2,2𝑁 /𝑁, the loading condition can be
considered acceptable in the considered design situation without further
simulations; and

.4 if the sample mean time to failure after a stability failure satisfies the
condition 𝑇̂ < 𝑇̂F , where 𝑇̂F = 𝛽2 (𝑁)/𝜆 and 𝛽2 (𝑁) = 0.5𝜒0.05/2,2𝑁
2
/𝑁 , the
loading condition can be considered as not acceptable without further
simulations.
4.8 Application examples of probabilistic assessment in design situations

4.8.1 Parametric roll

4.8.1.1 The example uses a numerical simulation of ship motions in multiple independent
realizations of the sea state and a direct counting of stability failures. The sea state is modelled
as a sum of harmonic components that discretize the Bretschneider wave energy spectrum
and the cos2-wave energy angular spreading function into 19 directional ranges i and 103

I:\CIRC\MSC\1\MSC.1-Circ.1652.docx
MSC.1/Circ.1652
Annex, page 164

frequency ranges i per direction. For each realization, the wave energy spectrum was
discretized into harmonic components with equal amplitudes, the phases of components were
randomly selected in the interval [0, 2), and their frequencies and directions were randomly
selected within the ranges i and i, respectively.

4.8.1.2 To ensure that the stability failures in numerical simulations satisfy Poisson process
assumptions, self-repetition effects were avoided by limiting the simulation time in each
seaway realization to one hour, the transient hydrodynamic effects at the beginning of each
simulation were eliminated by switching off the counter of stability failures and simulation timer
during the initial 50 roll periods, and the autocorrelation of large roll motions was avoided by
stopping a simulation after the first encountered stability failure.

4.8.1.3 Assessment was performed for a 1700 TEU container ship in loading conditions with
GM = 1.7 m, 1.8 m, …, 2.2 m for the parametric roll stability failure mode, using as the criterion
the maximum (over all required design situations) of the upper boundary of the 95%-
confidence interval of the short-term stability failure rate. For acceptance, this criterion should
not exceed the threshold equal to one stability failure in two hours in design sea states with a
probability density of 10-5 (ms)-1. The design situations correspond to zero forward speed in
head and following mean wave directions. The mean zero-upcrossing wave periods from 4.5 s
to 16.5 s with an increment of 1.0 s were used. For each wave period, the significant wave
height was selected according to table 4.6.2.

4.8.1.4 Each simulation was run for one hour but stopped after the first stability failure or
when a sufficient simulation time for acceptance 𝑡A = 𝛽1 (𝑁) ⋅ 𝑁/𝜆 − 𝑇̂𝑁−1 ⋅ (𝑁 − 1) was
2
achieved with 𝛽1 (𝑁) = 0.5𝜒1−0.05/2,2𝑁 /𝑁. If the sufficient simulation time for acceptance tA was
achieved during a simulation, no further simulations in the considered design situation were
conducted and the loading condition was considered as acceptable in the considered design
situation. Otherwise, after the simulation, the number of failures N (1 or 0) in the simulation
and the duration of the simulation t (equal to the time to failure if the realization ended with a
stability failure or one hour if a stability failure was not encountered) were defined and the total
number of failures N and the total simulation time tt were increased by N and t, respectively.
The unacceptance condition was checked: if 𝑇̂ < 𝑇̂F , where 𝑇̂F = 𝛽2 (𝑁)/𝜆 , 𝛽2 (𝑁) =
2
0.5𝜒0.05/2,2𝑁 /𝑁 and 𝑇̂ = 𝑡t /𝑁 is the sample mean time to failure, no further simulations were
conducted for the considered loading condition, which was judged as unacceptable.

4.8.1.5 Tables 4.8.1 to 4.8.4 show results for loading conditions with GM = 1.8 m (not
acceptable) and 1.9 m (acceptable) together with the required simulation time ts until each
stability failure, cumulative simulation time over all wave periods tc until the decision. For
brevity, the results are shown not after each simulation, but only after stability failures. In these
tables, ʺAʺ and ʺFʺ denote acceptance and not acceptance, respectively.

Table 4.8.1: Assessment results for GM = 1.8 m in head waves

Tz,s N Ti,s 𝑇̂,s tA,s A 𝑇̂𝐹 ,s F ts,s tc,s tc,hours
7.5 1 9.57e+2 9.57e+2 2.66e+4 - 1.82e+2 - 9.57e+2 9.57e+2 0.3
2 1.01e+4 5.54e+3 3.92e+4 - 8.72e+2 - 1.01e+4 1.11e+4 3.1
3 5.02e+2 3.86e+3 4.09e+4 - 1.48e+3 - 5.02e+2 1.16e+4 3.2
4 7.49e+2 3.08e+3 5.15e+4 - 1.96e+3 - 7.49e+2 1.23e+4 3.4
5 4.94e+3 3.46e+3 6.14e+4 - 2.34e+3 - 4.94e+3 1.73e+4 4.8
6 9.50e+2 3.04e+3 6.67e+4 - 2.64e+3 - 9.50e+2 1.82e+4 5.1
7 1.03e+4 4.08e+3 7.58e+4 - 2.89e+3 - 1.03e+4 2.85e+4 7.9
8 4.02e+2 3.62e+3 7.53e+4 - 3.11e+3 - 4.02e+2 2.89e+4 8.0
9 1.28e+3 3.36e+3 8.46e+4 - 3.29e+3 - 1.28e+3 3.02e+4 8.4
10 7.48e+3 3.77e+3 9.28e+4 - 3.45e+3 - 7.48e+3 3.77e+4 10.5
11 1.29e+3 3.54e+3 9.47e+4 - 3.59e+3 F 1.29e+3 3.90e+4 10.8

I:\CIRC\MSC\1\MSC.1-Circ.1652.docx
 MSC.1/Circ.1652
Annex, page 165

Table 4.8.2: Assessment results for GM = 1.8 m in following waves

Tz,s N Ti,s 𝑇̂,s tA,s A 𝑇̂𝐹 ,s F ts,s tc,s tc,hours
7.5 1 2.46e+3 2.46e+3 2.66e+4 - 1.82e+2 - 2.46e+3 2.46e+3 0.7
2 2.65e+3 2.56e+3 3.77e+4 - 8.72e+2 - 2.65e+3 5.11e+3 1.4
3 3.58e+3 2.90e+3 4.69e+4 - 1.48e+3 - 3.58e+3 8.70e+3 2.4
4 7.58e+2 2.36e+3 5.44e+4 - 1.96e+3 - 7.58e+2 9.45e+3 2.6
5 1.44e+3 2.18e+3 6.43e+4 - 2.34e+3 F 1.44e+3 1.09e+4 3.0

Table 4.8.3: Assessment results for GM = 1.9 m in head waves

Tz,s N Ti,s 𝑇̂,s tA,s A 𝑇̂𝐹 ,s F ts,s tc,s tc,hours
7.5 1 2.11e+2 2.11e+2 2.66e+4 - 1.82e+2 - 2.11e+2 2.11e+2 0.1
2 3.15e+4 1.59e+4 3.99e+4 - 8.72e+2 - 3.15e+4 3.17e+4 8.8
3 4.31e+4 2.49e+4 2.03e+4 A 1.48e+3 - 2.03e+4 5.20e+4 14.4
8.5 1 4.76e+4 4.76e+4 2.66e+4 A 1.82e+2 - 2.66e+4 7.86e+4 21.8
9.5 1 3.01e+4 3.01e+4 2.66e+4 A 1.82e+2 - 2.66e+4 1.05e+5 29.2
10.5 1 2.45e+4 2.45e+4 2.66e+4 - 1.82e+2 - 2.45e+4 1.30e+5 36.0
2 2.04e+4 2.25e+4 1.56e+4 A 8.72e+2 - 1.56e+4 1.45e+5 40.3
11.5 1 1.56e+4 1.56e+4 2.66e+4 - 1.82e+2 - 1.56e+4 1.61e+5 44.7
2 3.49e+4 2.53e+4 2.45e+4 A 8.72e+2 - 2.45e+4 1.85e+5 51.5
12.5 1 1.57e+4 1.57e+4 2.66e+4 - 1.82e+2 - 1.57e+4 2.01e+5 55.9
2 1.99e+4 1.78e+4 2.44e+4 - 8.72e+2 - 1.99e+4 2.21e+5 61.4
3 6.82e+4 3.46e+4 1.64e+4 A 1.48e+3 - 1.64e+4 2.37e+5 65.9
13.5 1 6.05e+3 6.05e+3 2.66e+4 - 1.82e+2 - 6.05e+3 2.43e+5 67.6
2 8.33e+4 4.47e+4 3.41e+4 A 8.72e+2 - 3.41e+4 2.78e+5 77.1

Table 4.8.4: Assessment results for GM = 1.9 m in following waves

Tz,s N Ti,s 𝑇̂,s tA,s A 𝑇̂𝐹 ,s F ts,s tc,s tc,hours
7.5 1 2.39e+3 2.39e+3 2.66e+4 - 1.82e+2 - 2.39e+3 2.39e+3 0.7
2 1.57e+5 7.96e+4 3.77e+4 A 8.72e+2 - 3.77e+4 4.01e+4 11.1
8.5 1 5.15e+3 5.15e+3 2.66e+4 - 1.82e+2 - 5.15e+3 4.53e+4 12.6
2 1.08e+3 3.11e+3 3.50e+4 - 8.72e+2 - 1.08e+3 4.63e+4 12.9
3 2.78e+4 1.13e+4 4.58e+4 - 1.48e+3 - 2.78e+4 7.41e+4 20.6
4 6.56e+3 1.01e+4 2.91e+4 - 1.96e+3 - 6.56e+3 8.07e+4 22.4
5 8.82e+4 2.58e+4 3.32e+4 A 2.34e+3 - 3.32e+4 1.14e+5 31.6
9.5 1 1.89e+4 1.89e+4 2.66e+4 - 1.82e+2 - 1.89e+4 1.33e+5 36.9
2 1.09e+3 9.98e+3 2.12e+4 - 8.72e+2 - 1.09e+3 1.34e+5 37.2
3 2.78e+3 7.58e+3 3.21e+4 - 1.48e+3 - 2.78e+3 1.37e+5 37.9
4 1.16e+4 8.59e+3 4.04e+4 - 1.96e+3 - 1.16e+4 1.48e+5 41.2
5 2.11e+4 1.11e+4 3.94e+4 - 2.34e+3 - 2.11e+4 1.69e+5 47.0
6 8.16e+3 1.06e+4 2.85e+4 - 2.64e+3 - 8.16e+3 1.78e+5 49.3
7 2.73e+3 9.49e+3 3.03e+4 - 2.89e+3 - 2.73e+3 1.80e+5 50.1
8 1.48e+4 1.02e+4 3.74e+4 - 3.11e+3 - 1.48e+4 1.95e+5 54.2
9 2.48e+4 1.18e+4 3.23e+4 - 3.29e+3 - 2.48e+4 2.20e+5 61.1
10 8.19e+3 1.14e+4 1.70e+4 - 3.45e+3 - 8.19e+3 2.28e+5 63.4
11 9.81e+3 1.13e+4 1.82e+4 - 3.59e+3 - 9.81e+3 2.38e+5 66.1
12 5.55e+4 1.50e+4 1.77e+4 A 3.72e+3 - 1.77e+4 2.56e+5 71.0
10.5 1 6.09e+4 6.09e+4 2.66e+4 A 1.82e+2 - 2.66e+4 2.82e+5 78.4
11.5 1 2.91e+3 2.91e+3 2.66e+4 - 1.82e+2 - 2.91e+3 2.85e+5 79.2
2 2.01e+4 1.15e+4 3.72e+4 - 8.72e+2 - 2.01e+4 3.05e+5 84.8
3 1.97e+3 8.34e+3 2.90e+4 - 1.48e+3 - 1.97e+3 3.07e+5 85.3
4 1.18e+4 9.21e+3 3.81e+4 - 1.96e+3 - 1.18e+4 3.19e+5 88.6
5 2.78e+4 1.29e+4 3.69e+4 - 2.34e+3 - 2.78e+4 3.47e+5 96.3
6 4.75e+3 1.16e+4 1.94e+4 - 2.64e+3 - 4.75e+3 3.51e+5 97.6

I:\CIRC\MSC\1\MSC.1-Circ.1652.docx
MSC.1/Circ.1652
Annex, page 166

7 4.78e+3 1.06e+4 2.47e+4 - 2.89e+3 - 4.78e+3 3.56e+5 99.0

8 1.57e+3 9.46e+3 2.97e+4 - 3.11e+3 - 1.57e+3 3.58e+5 99.4
9 2.96e+4 1.17e+4 3.78e+4 - 3.29e+3 - 2.96e+4 3.87e+5 107.6
10 1.15e+4 1.17e+4 1.77e+4 - 3.45e+3 - 1.15e+4 3.99e+5 110.8
11 1.07e+3 1.07e+4 1.57e+4 - 3.59e+3 - 1.07e+3 4.00e+5 111.1
12 1.81e+4 1.13e+4 2.39e+4 - 3.72e+3 - 1.81e+4 4.18e+5 116.1
13 2.14e+2 1.05e+4 1.50e+4 - 3.83e+3 - 2.14e+2 4.18e+5 116.2
14 8.84e+3 1.04e+4 2.39e+4 - 3.94e+3 - 8.84e+3 4.27e+5 118.6
15 1.97e+4 1.10e+4 2.41e+4 - 4.03e+3 - 1.97e+4 4.47e+5 124.1
16 2.24e+4 1.17e+4 1.34e+4 A 4.12e+3 - 1.34e+4 4.60e+5 127.8
Table 4.8.4 (cont’d): Assessment results for GM = 1.9 m in following waves
Tz,s N Ti,s 𝑇̂,s tA,s A 𝑇̂𝐹 ,s F ts,s tc,s tc,hours
12.5 1 5.91e+3 5.91e+3 2.66e+4 - 1.82e+2 - 5.91e+3 4.66e+5 129.5
2 1.42e+4 1.00e+4 3.42e+4 - 8.72e+2 - 1.42e+4 4.80e+5 133.4
3 1.14e+4 1.05e+4 3.19e+4 - 1.48e+3 - 1.14e+4 4.92e+5 136.6
4 4.34e+3 8.96e+3 3.16e+4 - 1.96e+3 - 4.34e+3 4.96e+5 137.8
5 8.40e+3 8.85e+3 3.79e+4 - 2.34e+3 - 8.40e+3 5.05e+5 140.1
6 4.77e+3 8.17e+3 3.98e+4 - 2.64e+3 - 4.77e+3 5.09e+5 141.5
7 9.33e+3 8.34e+3 4.50e+4 - 2.89e+3 - 9.33e+3 5.19e+5 144.1
8 4.89e+3 7.91e+3 4.55e+4 - 3.11e+3 - 4.89e+3 5.24e+5 145.4
9 1.03e+4 8.17e+3 5.02e+4 - 3.29e+3 - 1.03e+4 5.34e+5 148.3

10 1.85e+3 7.54e+3 4.95e+4 - 3.45e+3 - 1.85e+3 5.36e+5 148.8

11 1.44e+4 8.16e+3 5.70e+4 - 3.59e+3 - 1.44e+4 5.50e+5 152.8
12 9.41e+3 8.27e+3 5.19e+4 - 3.72e+3 - 9.41e+3 5.59e+5 155.4
13 8.08e+2 7.69e+3 5.17e+4 - 3.83e+3 - 8.08e+2 5.60e+5 155.6
14 2.15e+4 8.68e+3 6.00e+4 - 3.94e+3 - 2.15e+4 5.82e+5 161.6
15 1.18e+4 8.89e+3 4.76e+4 - 4.03e+3 - 1.18e+4 5.94e+5 164.9
16 3.54e+3 8.56e+3 4.48e+4 - 4.12e+3 - 3.54e+3 5.97e+5 165.9
17 1.49e+4 8.93e+3 5.02e+4 - 4.19e+3 - 1.49e+4 6.12e+5 170.0
18 2.81e+4 9.99e+3 4.42e+4 - 4.27e+3 - 2.81e+4 6.40e+5 177.8
19 1.23e+4 1.01e+4 2.50e+4 - 4.33e+3 - 1.23e+4 6.52e+5 181.2
20 1.68e+4 1.04e+4 2.15e+4 - 4.40e+3 - 1.68e+4 6.69e+5 185.9
21 7.11e+2 9.98e+3 1.35e+4 - 4.46e+3 - 7.11e+2 6.70e+5 186.1
22 6.27e+3 9.81e+3 2.15e+4 - 4.51e+3 - 6.27e+3 6.76e+5 187.8
23 3.55e+4 1.09e+4 2.40e+4 A 4.56e+3 - 2.40e+4 7.00e+5 194.5
13.5 1 1.18e+3 1.18e+3 2.66e+4 - 1.82e+2 - 1.18e+3 7.01e+5 194.8
2 7.55e+4 3.83e+4 3.89e+4 A 8.72e+2 - 3.89e+4 7.40e+5 205.6
14.5 1 1.34e+6 1.34e+6 2.66e+4 A 1.82e+2 - 2.66e+4 7.67e+5 213.0

4.8.1.6 Unacceptance requires much less simulation or testing time than acceptance since
the exceedance of the threshold in only one situation is sufficient. If simulations in head waves
are done first, unacceptance requires 87 minutes (if simulations in following waves are done
first, only 2.7 minutes) for GM = 1.7 m. For GM = 1.8 m, unacceptance requires 10.8h or 3.0h
simulation time, respectively. For acceptance, significantly more simulation or testing time is
required since all design situations must be considered and the failure rate is lower: for GM =
1.9 m, acceptance requires 290.1 hours of simulation time (which consist of 77.1 hours in head
waves and 213.0 hours in following); for GM = 2.0 m, 262.8 hours are required (69.3 hours in
head waves and 193.5 hours in following).

4.8.1.7 The decision and the total required simulation or testing time to make this decision
are summarized in table 4.8.5, separately for the assessment in head waves, in following
waves and in both head and following waves. For this ship, much less simulation or testing
time (bold figures) is required for the unacceptable loading conditions in following waves and

I:\CIRC\MSC\1\MSC.1-Circ.1652.docx
 MSC.1/Circ.1652
Annex, page 167

for acceptable loading conditions in head waves. Since the assessment for acceptable loading
conditions requires much more time than for unacceptable loading conditions, for this ship the
assessment in head waves requires significantly less time than in following waves.

Table 4.8.5: The decision and the required simulation or testing time if the assessment
is done in head, in following or both in head and following waves (results in the
second column are shown as "result for head waves + result for following
waves = result for both wave directions")
GM, m Decision (A=acceptable, Total simulation time, Total simulation
F=not acceptable) hours time, %

1.7 F+F=F 1.4 + 0.1 = 1.5 97 + 3 = 100

1.8 F+F=F 10.8 + 3.1 = 13.9 78 + 22 = 100
1.9 A +A=A 77.1 + 213.0 = 290.1 27 + 73 = 100
2.0 A+A=A 69.3 + 193.5 = 262.8 36 + 74 = 100

Total 158.6 + 409.6 = 568.2 28 + 72 = 100

4.8.2 Dead ship condition

4.8.2.1 An example of the direct stability assessment in design situations for the dead ship
condition is provided for a large cruise ship with the principal particulars and GZ curve shown
in table 2.5.1 and figure 2.5.1, respectively.
4.8.2.2 The results of both the level 1 and level 2 vulnerability assessment show that this ship
in the considered loading condition is not vulnerable to stability failure in the dead ship
condition with a sufficient margin:
.1 Level 1: b/a = 2.171 > 1.0, φ0 = 6.19º < 16º; and

.2 Level 2: C = 0.001057 < 0.06.

4.8.2.3 A coupled sway-heave-roll-pitch model was applied in changing wind and long-
crested irregular beam waves at zero forward speed, which considered the pitch effect in the
calculation of the restoring moment. Radiation and diffraction forces and moments were
calculated with a strip theory, and the roll damping moment was estimated from a roll decay
test. The mean wind velocity was provided using the formula in 2.2.3.2.2 of the Interim
Guidelines. The validation of the numerical method without wind and dynamic pitch effects is
described in section 2.6 of this appendix.

4.8.2.4 The probabilistic assessment in design situations was executed according to

section 3.5.3.3 of the Interim Guidelines, using the direct counting procedure specified in
paragraph 3.5.4. The sea states shown in table 3.5.3.3.5 of the Interim Guidelines were used.

4.8.2.5 Table 4.8.6 shows the results: “A” and “F” indicate acceptance and unacceptance,
respectively. In this calculation, the verification of the failure mode was applied: only if the ratio
of the instantaneous roll period to the instantaneous heave period or to the instantaneous
period of wind velocity fluctuations was between 0.8 and 1.2, the stability failure was judged
as a dead ship condition stability failure, which was addressed in this assessment example.

I:\CIRC\MSC\1\MSC.1-Circ.1652.docx
MSC.1/Circ.1652
Annex, page 168

Table 4.8.6: Results of the probabilistic direct stability assessment in design

situations for the dead ship condition stability failure mode for cruise ship

Tz (s) 4.5 5.5 6.5 7.5 8.5 9.5 10.5 11.5 12.5 13.5 14.5 15.5 16.5
Hs (m) 2.8 5.5 8.2 10.6 12.5 13.8 14.6 15.1 15.1 14.8 14.1 12.9 10.9
Uw (m/s) 12.0 18.9 24.6 29.2 32.6 34.8 36.2 37.0 37.0 36.5 35.3 33.3 29.8
Judgement A A A A A A A A A A A A A

4.8.2.6 The results indicate that, if verification of failure modes is used, the considered loading
condition is judged as acceptable with respect to the dead ship condition failure mode, which
is consistent with the vulnerability assessment. If the verification of failure modes is not used,
the effect of parametric roll in beam waves could appear.

5 Statistical extrapolation methods

5.1 Extrapolation of failure rate over wave height

5.1.1 Background

5.1.1.1 The extrapolation of the stability failure rate over the significant wave height22 is based
on the following suggestion: rare events happen, with some unknown probability, when a wave
or a wave group consisting of a certain (unknown) number of waves is encountered, which
exceeds a certain (also unknown) height. Denoting p, n and h as the unknown probability,
number of waves and wave height, respectively, the rate of such events can be approximated
as r p∙f(h,n : Hs), where f is the frequency of encountering rare wave groups with the required
characteristics. Since such wave groups are approximately Rayleigh-distributed,
𝑓 ∼ exp( − 2𝑛ℎ 2 /𝐻s2 ), and then:

ln 𝑟 = 𝐴 + 𝐵/𝐻s2 (5.1.1)

5.1.1.2 In eq. (5.1.1), r is the stability failure rate and the parameters A and B do not depend
on Hs but depend on the ship, the loading condition, forward speed and wave period and
direction. Since r = 1/T, where T is the mean time to failure, the extrapolation can also be
performed directly for the mean time to failure:

ln 𝑇 = 𝐴 + 𝐵/𝐻s2 (5.1.2)

5.1.1.3 Instead of defining parameters A and B explicitly in terms of unknown quantities p, n

and h, the idea of the extrapolation is to define these parameters empirically by direct counting
at greater significant wave heights; note, however, that at greater significant wave heights, the
events must remain sufficiently rare for this formulation to remain valid. It is also important to
emphasize that this method does not imply any assumptions about the relation between
exceedance rates of different reaction levels.

5.1.1.4 This method is especially useful for the full probabilistic direct stability assessment
and probabilistic operational measures, which are based on the evaluation of the stability
failure rate for all possible combinations of sailing conditions (vs, ) and sea states (Hs, Tz). First,
direct counting is used to define the stability failure rate for such combinations (vs, , Hs, Tz)
when resources allow a direct counting calculation. Beyond direct counting, the extrapolation
over the significant wave height, eq. (5.1.1), is applied to define the stability failure rate for the
remaining combinations.

22
Tonguć, E. and Söding, H. Computing capsizing frequencies of ships in seaway. Proc. 3rd Int. Conf. on
Stability of Ships and Ocean Vehicles, 1986.

I:\CIRC\MSC\1\MSC.1-Circ.1652.docx
 MSC.1/Circ.1652
Annex, page 169

5.1.1.5 The coefficients A and B can be obtained by, for example, linear regression of ln r, or
ln T, with respect to 1/Hs2, according to eq. (5.1.1) or (5.1.2), respectively. Data to be used for
the linear regression are those at greater significant wave heights, as described in
paragraph 5.1.1.3.

5.1.2 Validation

5.1.2.1 Document SDC 4/INF.8 shows that results obtained from direct counting can be used
for an extrapolation, eq. (5.1.1), if ln T > 6, i.e. ln r < -6, to avoid possible concave portions of
the dependency of ln T on 1/Hs2, i.e. a non-conservative extrapolation. In the Interim Guidelines,
this condition is formulated as a boundary: the values of the stability failure rate obtained by
direct counting can be used for an extrapolation, eq. (5.1.1), when they do not exceed 5% of
the reciprocal natural roll period of the ship, i.e. r < 0.05 / Tr (or T > 20 Tr).

5.1.2.2 In documents SDC 4/5/8 and SDC 4/INF.8, the extrapolation of the stability failure
rate over the significant wave height in eq. (5.1.1) was validated by comparison with direct
counting for synchronous roll in irregular short-crested beam waves (relevant for dead ship
condition and excessive acceleration stability failure modes).

5.1.2.3 Here, eq. (5.1.1) is validated for parametric roll in bow and stern waves, synchronous
roll in beam waves (associated with both the dead ship condition and excessive acceleration
stability failure modes) and pure loss of stability in stern waves for five ships in six loading
conditions each: in table 4.1.1, at six forward speeds; in table 4.2.1, in irregular short-crested
waves at 14 mean zero-crossing wave periods and systematically varied significant wave
heights. To quantify the accuracy of the extrapolation, several variants of extrapolation were
tested by varying the number of extrapolation points. For each variant, a specific number of
wave heights were selected ranging from 4 to 11 where the minimum wave height in each
variant is one for which results can be obtained by direct counting and for which ln T > 6.
Excluding this minimum significant wave height, the remaining wave heights were used to
perform the extrapolation using eq. (5.1.1) for each variant. The result of direct counting at the
minimum significant wave height was used to find the deviation between the extrapolated and
directly computed mean time to failure in each variant.

5.1.2.4 Figure 5.1.1 shows validation results as histograms of the ratio of the extrapolated to
directly computed estimate of the mean time to failure: y-axis corresponds to the number
(normed on 1) of cases in bins and x-axis shows the ratio of the extrapolated mean time to
failure Te to the directly estimated one T.

I:\CIRC\MSC\1\MSC.1-Circ.1652.docx
SDC 8/18/Add.4
Annex, page 170

Te/T Te/T Te/T

Te/T Te/T Te/T

I:\CIRC\MSC\1\MSC.1-Circ.1652.docx
 MSC.1/Circ.1652
Annex, page 171

Te/T Te/T Te/T

Te/T
Te/T Te/T

Figure 5.1.1 Histogram (number of cases normalized on 1) of the ratio Te/T and 95%-confidence interval of directly computed T (vertical lines)
for (from top to bottom) parametric roll in bow waves, parametric roll in stern waves, synchronous roll in beam waves, pure loss of stability
(bottom left and middle) and all cases together (bottom right). The symbols differentiate the number of points used for the extrapolation.

I:\CIRC\MSC\1\MSC.1-Circ.1652.docx
SDC 8/18/Add.4
Annex, page 172

5.1.2.5 To quantify the accuracy of an extrapolation, the percentage of such extrapolated

values that lie within the 95%-confidence interval of the directly computed estimate was
calculated, see table 5.1.1. The results show that the extrapolation, eq. (5.1.1), provides
sufficiently accurate results and thus is a useful practical tool to accelerate direct stability
assessment and the preparation of operational measures.

Table 5.1.1. Percentage of extrapolated values of the time to stability failure within a
95%-confidence interval of a directly computed estimate

Number of wave heights used for extrapolation 3 4 5 6 7 8 9 10

Parametric resonance in bow waves
Wave directions 150 to 180 degrees 79 83 85 84 83 81 78 81
Wave directions 160 to 180 degrees 79 82 84 82 81 79 77 79
Wave directions 170 to 180 degrees 78 82 83 81 80 78 77 76
Parametric resonance in stern waves
Wave directions 0 to 10 degrees 79 82 80 76 73 75 71 62
Wave directions 0 to 20 degrees 79 83 84 81 78 80 79 68
Wave directions 0 to 30 degrees 79 82 81 79 76 78 76 68
Synchronous resonance in beam waves
Wave directions 70 to 110 degrees 77 83 85 87 88 88 85 77
Wave directions 50 to 130 degrees 77 82 83 85 85 85 82 74
Wave directions 30 to 150 degrees 77 82 83 84 84 84 82 78
Pure loss in following waves 77 82 83 84 84 86 87 88
All above cases 77 81 82 83 82 81 79 75

5.1.3 Example procedure

5.1.3.1 Assume that direct counting has provided K maximum likelihood estimates 𝑟̂𝑘 of the
stability failure rate at significant wave heights Hsk, where k = 1, …, K; where each 𝑟̂𝑘 is defined
after encountering Nk stability failures (each Nk is not necessarily the same) so that the upper
boundaries of the 95%-confidence interval of stability failure rate obtained by direct counting
2
are 𝑟U,𝑘 = 𝑟̂𝑘 ⋅ 0.5𝜒1−𝛼/2,2𝑁 𝑘
/𝑁𝑘 , eq. (3.3.9), where  = 0.05.

2
5.1.3.2 A linear extrapolation of ln 𝑟̂𝑘 over 1/𝐻s𝑘 , eq. (5.1.1), provides the extrapolated failure
rate re:

ln 𝑟e = ∑𝐾
𝑘=1 𝑏𝑘 ln 𝑟̂𝑘 (5.1.3)

5.1.3.3 The coefficients bk, ∑𝐾

𝑘=1 𝑏𝑘 = 1, can be obtained by a least-squares method, see
paragraph 5.1.4.3.

5.1.3.4 To define the upper boundary reU of the 95%-confidence interval of the extrapolated
stability failure rate, the 2-distribution can be approximated by use of a normal distribution for
a large N (see paragraph 3.3.7) and apply a linear combination of normal distributions to obtain:
2 (5.1.4)
𝑟eU = 𝑟e ⋅ 0.5𝜒1−𝛼/2,2𝑁e
/𝑁e

5.1.3.5 In eq. (5.1.4),  = 0.05, and Ne is defined from:

1/𝑁e = ∑𝐾 2
𝑘=1 𝑏𝑘 /𝑁𝑘 (5.1.5)

5.1.3.6 When all Nk are the same, i.e. N1 = N2 = ∙∙∙ = Nk = N, then 𝑁e = 𝑁/ ∑𝐾 2

𝑘=1 𝑏𝑘 .

I:\CIRC\MSC\1\MSC.1-Circ.1652.docx
 MSC.1/Circ.1652
Annex, page 173

5.1.4 Application examples

5.1.4.1 Examples concern loading condition LC01 of the cruise vessel and the 14,000 TEU
container ship (see table 4.1.1) in irregular short-crested head and beam waves. The maximum
likelihood estimates 𝑟̂𝑘 of the stability failure rate, used for the extrapolation, were provided by
direct counting for N = 200 and, for comparison, 20 stability failures in 1.0 m ranges of significant
wave heights Hsk, k = 1, …, K. The upper boundaries of the 95%-confidence interval of the
2
stability failure rate obtained by direct counting were calculated as 𝑟U,𝑘 = 𝑟̂𝑘 ⋅ 0.5𝜒1−𝛼/2,2𝑁 /𝑁,
where  = 0.05. For extrapolation, the results for which rk < 0.05 / Tr were used (see paragraph
5.1.2.1) and the significant wave heights Hsk used for the extrapolation were not less than 2.0 m
(see paragraph 3.5.5.3.3 of Part B).

5.1.4.2 The extrapolated stability failure rate re was obtained by linear extrapolation of ln 𝑟̂𝑘
2
over 1/𝐻s𝑘 , eq. (5.1.2), using least-squares method as:
ln 𝑟e = ∑𝐾
𝑘=1 𝑏𝑘 ln 𝑟̂𝑘 (5.1.6)

5.1.4.3 For the least-squares extrapolation, the coefficients bk are calculated as:

bk = x/X + (1 – Kx/X)(xkX – X2) / (X2 – KX2) (5.1.7)

where 𝑋 = ∑𝐾 𝐾 2 2
𝑘=1 𝑥𝑘 , 𝑋2 = ∑𝑘=1 𝑥𝑘 and 𝑥𝑘 = 1/𝐻s𝑘 ; x = 1/Hs refers to the significant wave height
2

for which the extrapolation is performed.

5.1.4.4 The upper boundary reU of the 95%-confidence interval of the extrapolated stability
failure rate was calculated as:
2
𝑟eU = 𝑟e ⋅ 0.5𝜒1−𝛼/2,2𝑁e
/𝑁e (5.1.8)

where  = 0.05 and, for constant N=200 or N=20, Ne was calculated as 𝑁e = 𝑁/ ∑𝐾 2

𝑘=1 𝑏𝑘 .

5.1.4.5 Figure 5.1.2 shows the maximum likelihood estimates 𝑟̂𝑘 of the failure rate obtained
by direct counting ("suitable" points are those with rk less than 0.05 / Tr ; "not suitable" points are
those with rk > 0.05 / Tr.) and the extrapolated failure rate using three (K = 3) and all suitable
points, as well as the upper boundary of the 95%-confidence interval of stability failure rate,
obtained by both direct counting and extrapolation, vs. 1/Hs2. Note that there is a broadening
of the confidence interval of stability failure rate with a decreasing significant wave height due
to extrapolation.

I:\CIRC\MSC\1\MSC.1-Circ.1652.docx
MSC.1/Circ.1652
Annex, page 174

N=20 N=20
ln r, ln(1/s)

ln r, ln(1/s)
K=7
K=3
K=3 K=5
1/Hs2, 2
1/m 1/Hs2, 2
1/m
N=200 N=200

ln r, ln(1/s)
ln r, ln(1/s)

K=4 K=3
K=3 K=4
1/Hs2, 1/m2 1/Hs2, 1/m2
Figure 5.1.2 Examples of extrapolation of failure rate over significant wave height in
beam (left) and head (right) waves when direct counting uses 20 (top) and 200
(bottom) stability failures: maximum likelihood estimate of failure rate suitable ()
and not suitable () for extrapolation, extrapolated failure rate using three points
within Hs range 2.0 m (blue) and all suitable points (red) and corresponding upper
boundaries rU of 95%-confidence interval (dashed lines)

5.2 Critical wave method for surf-riding/broaching failure mode

5.2.1 Description of method

5.2.1.1 The critical wave method is one of the statistical extrapolation procedures for the
direct stability assessment for the surf-riding/broaching stability failure mode.

5.2.1.2 The critical wave method is a combination of a probabilistic "non-rare" procedure and
a deterministic "rare" procedure. The "non-rare" procedure can be regarded as the process of
the definition of the initial condition of the "rare" procedure. The "non-rare" procedure is either
an evaluation or an estimation of the probability of the encounter of a single large wave that is
characterized by the exceedance of values of parameters while the initial conditions belong to
a specified range. The "rare" procedure is the determination of the parameters of a single wave
and initial conditions that lead to stability failure.

5.2.1.3 For broaching associated with surf-riding, we may assume that broaching is a single
wave event, because surf-riding can be regarded as a single wave event. As well established
in non-linear dynamics, surf-riding in regular following waves has two different types: one type
occurs under any initial state of surge displacement and velocity if the wave and operational
conditions satisfy a critical condition and the other type occurs under the limited initial state of
surge displacement and velocity. The latter means that, if a ship is initially placed on a stable
surf-riding state for example, a ship keeps the surf-riding forever. Because of a
two-dimensional nature of the phase plane of dynamics, a self-propelled ship cannot enter the
initial state for the latter surf-riding without additional forcing other than assumed waves.
Therefore, if a ship keeps a specified propulsor thrust (a specified propeller revolutions for a
ship with propellers) with the initial propulsor thrust below the specified thrust, the ship cannot
experience the latter type of surf-riding so that the initial condition dependence of surf-riding in
regular waves can be excluded. In irregular waves, possibility of the former type of surf-riding
may exist but is negligibly small according to existing numerical investigations.

I:\CIRC\MSC\1\MSC.1-Circ.1652.docx
 MSC.1/Circ.1652
Annex, page 175

Further investigations also confirmed that broaching probability can be satisfactorily evaluated.
Thus, we may ignore the effect of initial conditions. This approach is adopted in the level 2
vulnerability criteria for surf-riding, section 2.6.3 of the guidelines.

5.2.1.4 First, the combinations of the wavelength and wave steepness leading to the first-type
surf-riding in regular following waves are determined using the Melnikov analysis, as adopted
in the level 2 vulnerability criteria (2.6.3.4.6 of the Interim Guidelines) or its equivalent, under
the specified nominal Froude number and the autopilot course from the wave direction.

5.2.1.5 Second, the numerical simulation based on a surge-sway-yaw-roll coupled model with
static heave, pitch and an autopilot or equivalent in regular stern-quartering waves should be
executed for varied wavelength to ship length ratio and wave steepness inside the region of
the first-type surf-riding. Here the initial conditions of the ship motions should be set to be a
periodic state under a small Froude number such as 0.1 and a small autopilot course from the
wave direction such as 0 degrees. The proportional gain for the autopilot should be set as a
practical but reasonably large value, such as 3, and the differential gain should be the minimum
for avoiding a directionally unstable phenomenon in calm water (e.g. continued overshooting
or zigzag). The integral gain, the non-linear elements and the band pass filter of the autopilot
may be excluded.

5.2.1.6 If the instant exists when both the yaw angle and yaw angular velocity increase over
time despite the maximum opposite application of rudder deflection, it can be identified as a
broaching event. Further, if a failure event (as defined in paragraph 3.2.1 of the Interim
Guidelines) occurs during this wave encounter, this combination of the wavelength and the
wave steepness should be regarded as a stability failure condition due to broaching in regular
waves.

5.2.1.7 Third, the joint probability density function of wavelength and wave height in stationary
irregular waves with a cell of a given significant wave height, Hs, and mean zero-crossing wave
period, Tz, is integrated within that cell of the stability failure condition due to broaching in
regular waves. The obtained value indicates the conditional probability of the stability failure
due to broaching per encounter wave in stationary irregular waves of the given significant wave
height and mean zero-crossing wave period under the specified nominal Froude number and
the autopilot course. The probability density function and the numerical integration method to
be used here can be found in paragraphs 2.6.3.2 and 2.6.3.3 of the Interim Guidelines.

5.2.1.8 Repeating the procedures presented in paragraphs 5.2.1.3 to 5.2.1.7 for various
significant wave heights and mean wave periods and integrating the product of the conditional
probability obtained with the joint probability density function of the significant wave height and
mean zero-crossing wave period, the average probability of stability failure due to broaching
per encountered wave is found, conditional on the specified nominal Froude number and the
autopilot course. The joint probability density function of the significant wave height and mean
zero-crossing wave period and the numerical integration method to be used here can be found
in 2.6.3.2 of the Interim Guidelines.

5.2.1.9 Repeating the above procedures for various autopilot courses and integrating the
product of the conditional probability, obtained above, with the probability density function of
the autopilot courses, the average conditional rate of stability failures due to broaching, 1/s, is
obtained at the specified nominal Froude number as follows:

r = – ln(1 – p) / Twe (5.2.1)

I:\CIRC\MSC\1\MSC.1-Circ.1652.docx
MSC.1/Circ.1652
Annex, page 176

where p is the average probability of stability failure due to broaching per encountered wave,
conditional on the specified nominal Froude number, and Twe is the mean zero-crossing wave
encounter period.

5.2.1.10 Since the critical wave method provides a conservative estimate of the probability of
stability failure due to surf-riding/broaching, eq. (5.2.1) can be used to estimate the upper
boundary of the 95%-confidence interval of the stability failure rate due to broaching. A further
conservative assumption used here is to use as the criterion the product of the stability failure
rate r obtained with eq. (5.2.1) with the probability of the specified ship speed and course.
The average of this product over the ship speeds and wave heading should not exceed the
standard 2.610-8 1/s specified in 3.5.3.2.2 of the Interim Guidelines.

5.2.2 Application example

5.2.2.1 Application example of the critical wave method for surf-riding/broaching failure mode
based on 5.2.1 concerns the ONR flare topside vessel shown in Figure 5.2.1. Since the ship
length is 154 m and the service Froude number is 0.4, this ship is regarded vulnerable to
broaching according to the level 1 vulnerability criterion.

Figure 5.2.1 Body plan of the ONR flare

topside vessel.

5.2.2.2 The tool used to calculate the broaching probability in the North Atlantic wave climate
is based on a combination of a surge-sway-yaw-roll simulation model with a proportional
autopilot in regular waves and stochastic wave theory. First, the simulation model estimates
the deterministic broaching zone as a function of wave steepness and wavelength for the given
propeller revolution and the autopilot course with respect to wave direction. Second, the
broaching probability in stationary sea states specified by the significant wave height, mean
wave period and wave energy spectrum is calculated by integrating the probability density
function of the local wave height and the wavelength within the deterministic broaching zone.
Finally, the average broaching probability per encountered wave in a specified sea area
represented by a wave scatter table is obtained by integrating the product of the broaching
probability per sea state with the occurrence probability of sea state.

5.2.2.3 The simulation model is based on a non-linear manoeuvring model with the
wave-induced forces and moments assuming low encounter frequency. The manoeuvring
forces and moments in calm water, including resistance and propeller thrust, are estimated
with circular motion captive model tests. The roll damping coefficient is estimated from roll
decay tests with a scaled ship model. The wave-induced forces and moments in
stern-quartering waves are calculated with a linear slender-body theory under the low
encounter frequency assumption as the sum of the Froude-Krylov components and
hydrodynamic lift due to wave particle velocity and corrected using captive model tests or
viscous-flow computational fluid dynamic methods. The interactions between the manoeuvring
forces and waves are ignored under the assumption that the wave steepness and the ship
motion normalized with respect to the forward velocity are not large. The stochastic wave
theory based on wave envelope theory proposed by Longuet-Higgins is used here, similar to
that used in the level 2 vulnerability criteria.

I:\CIRC\MSC\1\MSC.1-Circ.1652.docx
 MSC.1/Circ.1652
Annex, page 177

5.2.2.4 Broaching is defined as a phenomenon when the ship cannot keep straight course
despite the maximum opposite steering efforts. Based on this definition, the following criteria
are used (note that it does not include any quantitative value for course deviation from the
autopilot course):

.1 when the rudder deflection reaches its starboard limit, both the ship yaw
angular velocity and acceleration have the sign towards port; and

.2 when the rudder deflection reaches its port limit, both the ship yaw angular
velocity and acceleration have the sign towards starboard.

5.2.2.5 The broaching probability calculated with the described method is also compared with
that measured in model experiments in irregular waves, based on the ITTC recommended
procedure 7.5-02-07-04 for intact stability model tests. Examples of the comparison are shown
in figures 5.2.2 and 5.2.3.23 These results indicate that the used prediction procedure can be
applied for the direct stability assessment of the ONR flare topside vessel.

Fn

Figure 5.2.2 Comparison of measurement Figure 5.2.3 Comparison of broaching

and simulation in broaching region for probability between measurement and
ONR flare topside vessel in regular waves; simulation for ONR flare topside vessel
wavelength to ship length ratio is 1.25, in long-crested irregular waves;
autopilot course is 15 degrees from wave significant wave height is 9.65 m, mean
direction wave period 11.11 s, autopilot course
15 degrees from wave direction
5.2.2.6 In addition, example comparison of time series between the model experiment and
numerical simulation is shown in figure 5.2.4. Here, two broaching instances occurred, at
approximately 67 s and 97 s. The simulation reproduces these two instances and predicts
slightly conservatively the roll angle due to broaching.

23
Umeda, N., Usada, S., Mizumoto, K., Matsuda, A. Broaching probability for a ship in irregular
stern-quartering waves: theoretical prediction and experimental validation. Journal of Marine Science and
Technology, 21(1), pp. 23–37, 2016.

I:\CIRC\MSC\1\MSC.1-Circ.1652.docx
MSC.1/Circ.1652
Annex, page 178
Non-dimensional
relative surge

Figure 5.2.4. Comparison between model experiments and numerical calculations

5.2.3 Results and discussion

5.2.3.1 Using the described procedure, the probability of exceedance of a 40 degree roll
angle due to broaching associated with surf-riding in stationary sea states is obtained, which
depends upon the significant wave height Hs, the mean wave period T01, and the autopilot
course with respect to the wave direction at a nominal Froude number of 0.4. The results are
presented in tables 5.2.1 to 5.2.12. The wave energy spectrum follows the ITTC
recommendation (1974) and the wave scatter table, IACS Recommendation No.34 (Corr.1
Nov. 2001) (see the Interim Guidelines, Table 2.7.2.1.2). The Froude number (corresponding
to service speed) is 0.4; the rudder gain for the autopilot is 3.

Table 5.2.1. Probability of exceedance of a 40 degree roll angle due to broaching in

stationary sea states at an autopilot course of 3.75 degrees and a nominal Froude
number of 0.4

Hs (m)

Table 5.2.2. Probability of exceedance of a 40 degree roll angle due to broaching in

stationary sea states at an autopilot course of 7.5 degrees and a nominal Froude
number of 0.4
T01(sec)

I:\CIRC\MSC\1\MSC.1-Circ.1652.docx
 MSC.1/Circ.1652
Annex, page 179

Hs (m)

Table 5.2.3. Probability of exceedance of a 40 degree roll angle due to broaching in

stationary sea states at an autopilot course of 11.25 degrees and a nominal Froude
number of 0.4
T01(sec)

Hs (m)

Table 5.2.4. Probability of exceedance of a 40 degree roll angle due to broaching in

stationary sea states at an autopilot course of 15 degrees and a nominal Froude
number of 0.4
T01(sec)

Hs (m)

Table 5.2.5. Probability of exceedance of a 40 degree roll angle due to broaching in

stationary sea states at an autopilot course of 18.75 degrees and a nominal Froude
number of 0.4

I:\CIRC\MSC\1\MSC.1-Circ.1652.docx
MSC.1/Circ.1652
Annex, page 180

T01(sec)

Hs (m)

Table 5.2.6. Probability of exceedance of a 40 degree roll angle due to broaching in

stationary sea states at an autopilot course of 22.5 degrees and a nominal Froude
number of 0.4
T01(sec)

Hs (m)

Table 5.2.7. Probability of exceedance of a 40 degree roll angle due to broaching in

stationary sea states at an autopilot course of 26.25 degrees and a nominal Froude
number of 0.4
T01(sec)

Hs (m)

Table 5.2.8. Probability of exceedance of a 40 degree roll angle due to broaching in

stationary sea states at an autopilot course of 30 degrees and a nominal Froude
number of 0.4

I:\CIRC\MSC\1\MSC.1-Circ.1652.docx
 MSC.1/Circ.1652
Annex, page 181

T01(sec)

Hs (m)

Table 5.2.9. Probability of exceedance of a 40 degree roll angle due to broaching in

stationary sea states at an autopilot course of 33.75 degrees and a nominal Froude
number of 0.4
T01(sec)

Hs (m)

Table 5.2.10. Probability of exceedance of a 40 degree roll angle due to broaching in

stationary sea states at an autopilot course of 37.5 degrees and a nominal Froude number
of 0.4
T01(sec)

Hs (m)

Table 5.2.11. Probability of exceedance of a 40 degree roll angle due to broaching in

stationary sea states at an autopilot course of 41.25 degrees and a nominal Froude
number of 0.4
T01(sec)

I:\CIRC\MSC\1\MSC.1-Circ.1652.docx
MSC.1/Circ.1652
Annex, page 182

Hs (m)

Table 5.2.12. Probability of exceedance of a 40 degree roll angle due to broaching in

stationary sea states at an autopilot course of 45 degrees and a nominal Froude
number of 0.4
T01(sec)

Hs (m)

5.2.3.2 The effect of the autopilot course, as shown in Figure 5.2.5, shows that a broaching
danger exists in the autopilot course range from 10 to 30 degrees. For a uniform course
distribution, the broaching probability in the North Atlantic is 2.02∙10 -4 (this is a conditional
exceedance probability per an encountered wave of a 40-degree roll angle due to broaching
associated with surf-riding), which results in the failure rate of 3.2210-6 1/s. Since it is greater
than the standard of 2.610-8 1/s, this ship can be regarded as vulnerable for broaching if it is
operated without operational measures.

5.2.3.3 Since this ship is also judged as vulnerable to broaching failure by the vulnerability
level 1 and level 2 criteria, this result can be regarded as consistent.

5.2.3.4 An operational guidance can be developed using tables 5.2.1 to 5.2.12 by specifying
the acceptable stability failure probability for each stationary sea state. Figures 5.2.5 and 5.2.6
indicate that reducing the nominal forward speed and increasing the threshold of roll angle are
effective for decreasing the stability failure probability due to broaching.

I:\CIRC\MSC\1\MSC.1-Circ.1652.docx
 MSC.1/Circ.1652
Annex, page 183

Figure 5.2.5. exceedance probability of a 40 Figure 5.2.6. Probability per

degree roll angle per encountered wave due to encountered wave of broaching
broaching associated with surf-riding vs. associated with surf-riding vs.
wave heading for the ONR flare topside vessel nominal Froude number for ONR
in the North Atlantic wave climate flare topside vessel in the North
Atlantic wave climate

5.3 Split-time/motion perturbation method

5.3.1 Theoretical background for the application of extreme value theory

5.3.1.1 The split-time method, also known as the motion perturbation method (MPM), and
envelope peaks over threshold (EPOT) method are two extrapolation methods
(Interim Guidelines, paragraph 3.5.5.4.1) based on the application of statistical extreme value
theory. The text given in this section is meant to be used for familiarization with the principle
of extreme value theory and should be considered as guidance for its practical application.

5.3.1.2 Extreme value theory describes statistical properties of a largest value in a sample of
independent data points. It has been applied in engineering (e.g. flood management) and other
areas. The central idea of the extreme value theory is that the distribution of the largest value
in an independent data sample is different from the distribution of the sample. This idea can
be illustrated with the following example:

• take a number of roll angle records;

• estimate the autocorrelation function (refer to paragraph 3.8.3);

• evaluate the decorrelation time (paragraph 3.8.7);

• extract independent roll angle data points by taking the points that are no closer
to each other than the decorrelation time. This extraction makes several samples
of independent roll angle data points (These independent roll angle data points
have a distribution that can be estimated with a histogram, this distribution is
further referred as "an underlying distribution"); and

• select the largest data point from each independent roll angle data sample.
The outcome of this selection is the distribution of the largest data points
(i.e. values) from each sample that is different than the underlying distribution.

5.3.1.3 For a sample with n data points, the distribution of the largest value in a sample is:
𝑝𝑑𝑓1 (𝑦|𝑛) = 𝑝𝑑𝑓(𝑦)(𝑐𝑑𝑓(𝑦))𝑛−1
where 𝑝𝑑𝑓(𝑦) and 𝑐𝑑𝑓(𝑦) ) are the probability density function (PDF) and the cumulative
distribution function (CDF), respectively, of the underlying distribution. With an increase of the

I:\CIRC\MSC\1\MSC.1-Circ.1652.docx
MSC.1/Circ.1652
Annex, page 184

sample size, the distribution of the largest value tends to a universal limit that depends only on
a small set of parameters characterizing the distribution of the sample data. This limit
distribution is known as the Generalized Extreme Value Distribution (GEVD). The GEVD is
defined by three parameters, a shape parameter , a scale parameter  and location value 
and its CDF is:
𝑦−𝜇
exp (− exp (− 𝜎
)) for ξ = 0
𝑦−𝜇 −(1/ξ) 𝑦−𝜇
exp (− (1 + ξ ) ) for ξ ≠ 0 and ξ > −1
cdf𝐺𝐸𝑉 (𝑦) = 𝜎 𝜎 (5.3.1)
𝑦−𝜇
0 for ξ > 0 and ξ 𝜎
≤ −1
𝑦−𝜇
{ 1 for ξ < 0 and ξ 𝜎
≤ −1
The PDF of the GEVD is:
𝑦−𝜇 𝑦−𝜇
exp (− ) ∙ exp (− exp (− )) for ξ = 0
𝜎 𝜎
−(1+1/ξ) 𝑦 − 𝜇 −(1/ξ)
pdf𝐺𝐸𝑉 (𝑦) = (1 + ξ 𝑦 − 𝑢 ) ∙ exp (− (1 + ξ ) ) for ξ ≠ 0 and ξ
𝑦−𝜇
> −1
𝜎 𝜎 𝜎
{0 otherwise
(5.3.2)
The parameters ,  and  are found from data. Figures 5.3.1 and 5.3.2 illustrate this concept.
The PDF of the largest value 𝑝𝑑𝑓1 (𝑦|𝑛) is computed for an increasing volume of the sample
from 101 to 107. Figure 5.3.1 shows the convergence with the 𝑝𝑑𝑓(𝑦) standard normal
(where the mean value is 0 and standard deviation / variance is 1). Figure 5.3.2 uses the
Rayleigh distribution for the illustration. The GEVD shape parameter  is known to be zero for
both normal and Rayleigh distributions. The other GEVD parameters  and , have the
following relationship to the GEVD mean value 𝐸𝐺𝐸𝑉 and variance 𝑉𝐺𝐸𝑉 :
√6𝑉𝐺𝐸𝑉
𝜎= 𝜋
; 𝜇 = 𝐸𝐺𝐸𝑉 − 0.57721𝜎 (5.3.3)
The value 0.57721 is known as the Euler–Mascheroni constant. The fitting is carried out by
finding the mean value 𝐸1 and the variance 𝑉1 of the distribution of the largest value 𝑝𝑑𝑓1 (𝑦|𝑛)
and using these values instead of the 𝐸𝐺𝐸𝑉 and the 𝑉𝐺𝐸𝑉 , respectively:
∞ ∞
𝐸𝐺𝐸𝑉 = 𝐸1 = ∫−∞ 𝑦 ∙ 𝑝𝑑𝑓1 (𝑦|𝑛)𝑑𝑦 ; 𝑉𝐺𝐸𝑉 = 𝑉1 = ∫−∞(𝑦 − 𝐸1 )2 𝑝𝑑𝑓1 (𝑦|𝑛)𝑑𝑦 (5.3.4)

First-order statistic distribution for the largest instance of n samples

Generalized extreme value (GEV) distribution

2.5 pdf(y) Figure 5.3.1 Conversion of

n=107 the distribution of the
2 n=106
n=105
largest value in a sample to
n=104 a GEVD with an increasing
1.5
n=1000 number of samples, where
n=100 the standard normal
1 n=10
Underlying distribution is the underlying
distribution:
standard normal 0.5 distribution

y
-3 -2 -1 0 1 2 3 4 5 6 7

I:\CIRC\MSC\1\MSC.1-Circ.1652.docx
 MSC.1/Circ.1652
Annex, page 185

Generalized extreme value (GEV) distribution

First-order statistic distribution for the largest
instance of n samples
Figure 5.3.2 Conversion of
2.5 pdf(y) the distribution of the
n=107
n=106 largest value in a sample to
2 n=105 the GEVD with an
n=104 increasing number of
1.5 n=1000 samples where the
n=100
Rayleigh distribution is the
1 n=10
Underlying distribution:
underlying distribution
Rayleigh 𝒙 𝒙𝟐
0.5 𝑽𝒚
𝐞𝐱𝐩 (− 𝟐𝑽 ): with 𝑽𝒚 = 𝟏. 𝟎
𝒚

y
0 1 2 3 4 5 6 7

5.3.1.4 The statement in paragraph 5.3.1.3 is known as the first extreme value theorem or
the Fisher-Tippett-Gnedenko theorem. It is proven by expressing a distribution of the largest
value in a sample of independent identically distributed data points and showing that the limit
of that distribution depends only on a small set of parameters that characterize the distribution
of the sample data points.

5.3.1.5 Practical use of the GEVD involves collecting data, forming "blocks" from this data
that may be considered independent, selecting the largest values from each block, fitting the
GEVD with these block maxima data, and then estimating the shape ( ), scale () and location
() parameters.

5.3.1.6 In an effort to increase utilization of collected data (GEVD fitting uses only one data
point from each block of data), it was proven that the GEVD can be approximated above a
sufficiently large threshold u with a generalized Pareto distribution (GPD) as
𝑦−𝑢
1 − exp (− 𝜎
) for ξ = 0

cdf(𝑦) = 𝑦−𝑢 −1/ξ 𝑦−u (5.3.5)

1 − (1 + ξ 𝜎
) for ξ ≠ 0 and ξ 𝜎
> −1

{0 otherwise
1 𝑦−𝑢
exp (− ) for ξ = 0
𝜎 𝜎

pdf(𝑦) = 1 𝑦−𝑢 −(1+1/ξ) 𝑦−u (5.3.6)

(1 + ξ ) for ξ ≠ 0 and ξ > −1
𝜎 𝜎 𝜎

{0 otherwise

5.3.1.7 The statement in paragraph 5.3.1.6 is known as the second extreme value theorem
or the Pickands-Balkema-de Haan theorem. It is proven by consideration of the distribution in
paragraph 5.3.1.3 under the condition that the random variable y exceeds the threshold u.

5.3.1.8 The practical use of the GPD involves the collection of data, selecting independent
data points (de-clustering – using a similar procedure as that used in direct counting),
searching for the location of the threshold, and then estimation of the shape and scale
parameters from the de-clustered data points above the threshold.

I:\CIRC\MSC\1\MSC.1-Circ.1652.docx
MSC.1/Circ.1652
Annex, page 186

5.3.1.9 The estimation of the shape and scale parameters can be performed with several
methods. For example, one method (sometimes referred as the moment method) uses the
mean value and variance of the GPD. For a given threshold u, the mean value EGPD and the
variance VGPD are expressed as:
𝜎
𝐸𝐺𝑃𝐷 = 𝑢 + for 𝜉 < 1
1−𝜉

𝜎2 1
𝑉𝐺𝑃𝐷 = (1−𝜉)2(1−2𝜉)
for 𝜉 < 2 (5.3.7)

The mean value and the variance can be also estimated from the data directly. Then, the scale
and shape parameters can be found, treating the formulae in this paragraph as a system of
algebraic equations. Note that the mean value and variance of the GPD do not always exist,
∞ ∞
as the integrals ∫0 𝑦 ∙ pdf(𝑦) 𝑑𝑦 and ∫0 (𝑦 − 𝐸𝐺𝑃𝐷 )2 ∙ pdf(𝑦)𝑑𝑦 do not always converge for 𝜉 ≥
1 and 𝜉 ≥ 1/2, respectively.

5.3.1.10 Another method is based on the maximum likelihood estimator (MLE). The idea of the
MLE is to find such values of parameters that are "most likely" to fit the data. The data points
that have been observed are the facts. At the same time, they are instances of a random
variable. Because these particular values were observed, they are more likely to occur than
others. This means that the probability of observing these particular values reaches a
maximum when the correct parameters are used for distribution. Recall that a probability of
observation of a certain value is pdf(𝑦)𝑑𝑦, where the likelihood function can be expressed
through a product of these probabilities (as the data points were made independent through
de-clustering). For a given threshold u, having N observations of random variable y, the
likelihood is:

𝐿(𝜉, 𝜎) = ∏𝑁
𝑖=1 pdf(𝑦𝑖 ; 𝜉, 𝜎 ) 𝑖 = 1, … , 𝑁 (5.3.8)

The semicolon in the expression pdf(𝑦𝑖 ; 𝜉, 𝜎) means that a computation is carried out for the
value 𝑦𝑖 ; while the scale and shape parameters values are  and , respectively.
The parameters are estimated by maximizing the value of the likelihood function 𝐿(𝜉, 𝜎) .
In practice this is made easier by taking the natural logarithm of the likelihood function, as the
likelihoods usually are small values; also, products are transformed into sums, thus providing
additional convenience and referred to as the log likelihood function:

ℓ(𝜉, 𝜎) = ln(𝐿(𝜉, 𝜎)) = ∑𝑁

𝑖=1 pdf(𝑦𝑖 ; 𝜉, 𝜎 ) (5.3.9)

5.3.1.11 To complete the GPD fit, a threshold needs to be found. Choosing a correct threshold
is critical to ensuring the applicability of the GPD. If the threshold is too low, the fitted GPD is
not an approximation of the tail, because the conditions of the second extreme value theorem
have not been met. If the threshold is too high, too many data points will remain unused and
the result will be less accurate. The second extreme value theorem states that the GPD can
be used for an approximation of the tail of any distribution if the threshold is high enough.
That means that above a certain threshold the GPD approximation must be invariant to the
threshold. The simplest way to check this is to observe stabilization of the shape parameter
estimate over a number of thresholds. Other methods are based on finding such a threshold
that will minimize the difference between the observed data and fitted tail of the distribution.

5.3.1.12 Since the shape and scale parameters are estimated from data (which are random
numbers), the estimated parameters are also random numbers. Their distribution can be
approximated with a bivariate normal distribution, assuming that the deviations of the estimates
from true values are caused by many different random reasons. This assumption allows the

I:\CIRC\MSC\1\MSC.1-Circ.1652.docx
 MSC.1/Circ.1652
Annex, page 187

Central Limit Theorem to be applied; this theorem is a standard technique used for the
assessment of the uncertainty of estimation. The assumption of a normal distribution needs to
be treated with certain caution, as it supports both negative and positive values, while the scale
parameter is positive. To define the bivariate normal distribution, a covariance matrix of that
bivariate normal distribution needs to be estimated. The estimation of a covariance matrix can
be performed with the delta method, in which the minimization of the log likelihood function
from paragraph 5.3.1.10 is considered to be a deterministic function of random variables.
This function is linearized using a Taylor series expansion in which only the first-order terms
are retained. A linear function of normally distributed arguments yields the following covariance
matrix for the joint distribution of the scale  and shape  parameter estimates:

−1
𝜕2 ℓ(𝜉,𝜎) 𝜕2 ℓ(𝜉,𝜎)
𝑉𝜉 𝑟𝜉𝜎 √𝑉𝜉 𝑉𝜎 𝜕𝜉 2 𝜕𝜉𝜕𝜎
𝐶 (𝜉, 𝜎) = ( ) = − (𝜕2 ℓ(𝜉,𝜎) 𝜕2 ℓ(𝜉,𝜎)
) (5.3.10)
𝑟𝜉𝜎 √𝑉𝜉 𝑉𝜎 𝑉𝜎
𝜕𝜉𝜕𝜎 𝜕𝜎2

where 𝑉𝜉 , 𝑉𝜎 are the variance of the estimated shape and scale parameters respectively; and
𝑟𝜉𝜎 is the correlation coefficient of the estimated shape and scale parameters. The confidence
intervals for the shape and scale parameters can be evaluated using marginal normal
distributions for the shape and scale estimates. These confidence intervals may be needed
also for selecting the threshold u.

5.3.1.13 The shape parameter describes the type of tail: heavy, exponential or light, as shown
in figure 5.3.3. (There is no universally accepted definition of either a heavy or light tail.
Other sources may use heavy/light tail in a different context.) For example, the exponential tail
( = 0) describes the extreme values of a normal distribution. The heavy tail (   ) is above
the exponential tail while the light tail (  ) is below it. Because the exponential tail is the
smallest infinite tail, the light tail has a limit, which is its right boundary. The heavy tail is
unbounded.
PDF

Exponential tail  = 0

Heavy tail  > 0

Figure 5.3.3 Types of tail
Light tail  < 0
y
For  < 0, right boundary

5.3.1.14u=
Using the GPD to extrapolate the probability of exceedance yields the conditional
probability that the level of failure c has been exceeded if the threshold u has been exceeded:
𝑦−𝑢
exp (− ) for ξ̂ = 0
̂
𝜎
1
−̂
̂(𝑦 > 𝑐 ) = 1 − cdf(c) = (1 + ξ̂(𝑦−𝑢))
P ξ
for ξ̂ ≠ 0 and
ξ̂ (𝑦−𝑢)
> −1 (5.3.11)
̂
𝜎 𝜎̂
ξ̂ (𝑦−𝑢)
{0 for ξ̂ ≠ 0 and ≤ −1
̂
𝜎

The symbol "hat" ̂ indicates that the value is actually an estimate, i.e. a random quantity.
The extrapolated stability failure rate of c can be estimated as:
̂(𝑦 > 𝑐 )
𝑟̂ (𝑐) = 𝑟̂ (𝑢)P (5.3.12)
where 𝑟̂(𝑢) is a rate of exceedance of the threshold that is estimated with direct counting
(see section 3 of this appendix for more description).

I:\CIRC\MSC\1\MSC.1-Circ.1652.docx
MSC.1/Circ.1652
Annex, page 188

5.3.1.15 Since the extrapolated estimate is a random number, its uncertainty needs to be
assessed in a form of confidence interval. For this purpose, the boundary method may be used
where corresponding boundaries are multiplied. The confidence intervals of the estimated rate
of exceedance of the threshold 𝑟̂(𝑢) and the extrapolated estimate of the conditional probability
̂ (𝑦 > 𝑐) need to be computed for the confidence
of exceedance of the stability failure level P
probability √1 − 𝛼.
̂𝑈 (𝑦 > 𝑐; √1 − 𝛼)
𝑟̂𝑈 (𝑐; 1 − 𝛼) = 𝑟̂𝑈 (𝑢; √1 − 𝛼)P
̂𝐿 (𝑦 > 𝑐; √1 − 𝛼)
𝑟̂𝐿 (𝑐; 1 − 𝛼) = 𝑟̂𝐿 (𝑢; √1 − 𝛼)P (5.3.13)

The calculation of the boundaries of the confidence interval for direct counting is described in
section 3 of this appendix.

5.3.1.16 The calculation of the boundaries of the confidence interval for the extrapolated
estimate P ̂ 𝑈,𝐿 (𝑦 > 𝑐; √1 − 𝛼) can also be estimated with the boundary method using a
bivariate distribution of estimates of the scale  and shape  parameters, as discussed
in 5.3.1.12. The application of the boundary method for the extrapolated estimate is known to
produce a slightly conservative result. Alternatively, the distribution of the extrapolated
estimate can be computed by treating the formula in paragraph 5.3.1.14 as a deterministic
function of random arguments: the estimated scale and shape parameters bivariate
distribution. Other methods exist and are described in technical literature.

5.3.1.17 The formula in paragraph 5.3.1.14 yields a zero-value result if the following conditions
exist: ξ̂(𝑐 − 𝑢)/𝜎̂ ≤ −1; the right boundary has been encountered (see figure 5.3.1); and the
failure level is beyond this right boundary. Because estimates of the scale and shape
parameters are random quantities, the zero-value result is always possible. In the case of a
zero-value result, the upper boundary of the confidence interval still may not be zero even if
the estimate itself is zero. However, at the same time, if the upper boundary of the extrapolated
estimate is zero, the result of extrapolation is zero.

5.3.1.18 The confidence interval of the extrapolated estimate in paragraph 5.3.1.15 may be
large because the uncertainty is driven by two estimated parameters of the GPD distribution.
At the same time, the application of the GPD is completely data-driven, i.e. it is not specific for
any particular mode of failure. However, this universality may result in a large uncertainty and
increase the requirements for the number of data points of the sample.

5.3.2 Theoretical background for split-time / MPM method

5.3.2.1 The split-time method, also known as the motion perturbation method (MPM), is one
of the extrapolation methods mentioned in paragraph 3.5.5.4.1 of the Interim Guidelines.
The split-time method is intended for estimating the probability of complex and rare physical
phenomena in which the physics of the problem changes with the extreme response, such as
that caused by stability failure in dead ship condition or pure loss of stability in stern-quartering
and following waves. To account for changing physics and to reduce uncertainty, this
extrapolation is performed for a specially computed metric, rather than the roll angle itself.

5.3.2.2 A stability failure of a stability-compliant ship in severe conditions cannot be observed

during a simulation or set of simulations of reasonable length. Therefore, a special metric of
failure likelihood is introduced. For a failure through motion to starboard, this metric is
computed at the instant of up-crossing of an intermediate roll threshold us. The roll rate is
perturbed until the failure through motion to starboard is observed, shown in figure 5.3.4.
The difference between the roll rate at up-crossing 𝜑̇ 𝑈,𝑖 and the roll rate when failure is
observed 𝜑̇ 𝑆,𝑖 is the metric, as this difference indicates "the distance to failure":

I:\CIRC\MSC\1\MSC.1-Circ.1652.docx
 MSC.1/Circ.1652
Annex, page 189

𝑦𝑆,𝑖 = 𝑐 + 𝜑̇ 𝑈,𝑖 − 𝜑̇ 𝑆,𝑖 ; 𝑐 = 1 rad/s ; 𝑖 = 1, … , 𝑁𝑈 (5.3.14)

where 𝑁𝑈 is the number of up-crossings and c is a dimensional location constant, for use with
a failure threshold. The value 1 rad/s represents a level sufficiently large to be convenient in
the calculations. The particular value of the dimensional location constant does affect the result
of calculation.

, deg
40

Target
Instant of
angle
upcrossing
20 Figure 5.3.4 Motion
Intermediate threshold perturbation for
computing capsizing
t, s metric
690 700 710 720 730 740 750

-20
Decorrelation time Tdc

5.3.2.3 Computation of the metric values, defined in paragraph 5.3.2.2 over a number of
up-crossings creates a sample that can be used for extrapolation up to the level 𝑐 = 1 rad/s.
An estimate of a conditional probability of exceedance of the level c under the condition of
up-crossing of the intermediate threshold us: 𝑃̂ (𝑦𝑆 ≥ 𝑐| 𝜑 = 𝑢𝑠 ∩ 𝜑̇ > 0 ) is used for the
estimate of failure rate through motion to starboard r̂𝑆 as

r̂𝑠 = 𝑟̂𝑈 𝑃̂ (𝑦𝑆 ≥ 𝑐 | 𝜑 = 𝑢𝑠 ∩ 𝜑̇ > 0 ) (5.3.15)

where r̂𝑈 is an estimate of the up-crossing rate of the intermediate threshold 𝑢𝑠 , evaluated
using direct counting; and the symbol ∩ means "and".

5.3.2.4 For failure through motion to port, the metric is computed at the instant of
down-crossing of an intermediate roll threshold 𝑢𝑝 . The roll rate is perturbed until the failure is
observed. The difference between the roll rate at down-crossing 𝜑̇ 𝐷,𝑖 and the roll rate when
failure through motion to port is observed 𝜑̇ 𝑃,𝑖 is the metric:
𝑦𝑃,𝑖 = 𝑐 + |𝜑̇ 𝐷,𝑖 | − |𝜑̇ 𝑃,𝑖 | ; 𝑐 = 1 rad/s ; 𝑖 = 1, … , 𝑁𝐷 (5.3.16)
where 𝑁𝐷 is the number of down-crossings.

5.3.2.5 The computation of the metric values, defined in paragraph 5.3.2.4 over a number of
down-crossings creates a sample that can be used for an extrapolation up to the level 𝑐 = 1 rad/s.
An estimate of a conditional probability of exceedance of the level c under the condition of
down-crossing the intermediate threshold 𝑃̂ (|𝑦𝑃 | ≥ 𝑐 | 𝜑 = 𝑢𝑝 ∩ 𝜑̇ < 0) is used for the estimate
of the failure rate through motion to port r̂𝑃 as
𝑟̂𝑃 = 𝑟̂𝐷 𝑃̂(|𝑦𝑃 | ≥ 𝑐|𝜑 = 𝑢𝑝 ∩ 𝜑̇ < 0) (5.3.17)
where 𝑟̂𝐷 is an estimate of the down-crossing rate of the intermediate threshold 𝑢𝑝 , evaluated
using direct counting.

I:\CIRC\MSC\1\MSC.1-Circ.1652.docx
MSC.1/Circ.1652
Annex, page 190

5.3.2.6 The failure rate through motion to starboard or port is estimated as

𝑟̂𝐹 = 𝑟̂𝑆 + 𝑟̂𝑃 (5.3.18)

5.3.2.7 The metrics 𝑦𝑆 and 𝑦𝑃 are based on roll rate values at the instant of up-crossing or
down-crossing. The non-linearities of roll motion, associated with roll rate are usually weak;
distribution of roll rate is close to normal. Therefore, the distribution of roll rates at the instants
of up-crossing or down-crossing follows closely to the Rayleigh24 distribution and the tail is
exponential.25

5.3.2.8 The application of the split-time/motion perturbation method does not depend on any
information of non-linearity of roll motion because all the non-linear factors automatically are
accounted for during the calculation of the metric. The method can handle unusual shapes of
GZ curves and is capable of estimating the probability of capsizing.

5.3.3 Description of split-time/motion perturbation method extrapolation procedure

5.3.3.1 The input data for this extrapolation procedure is produced by a numerical simulation
carried out as required by section 3.3 of the Interim Guidelines. The input data are presented in
a form of a set of NR records, each of which contains Nj instantaneous roll angles {𝜑𝑖 }𝑗 , roll rates
{ 𝜑̇ 𝑖 }𝑗 (Note: The dot above the symbol means derivative with respect to time), motions and
velocities in other degrees of freedom, each of which is computed at time instances {𝑡𝑖 }𝑗 using
a constant time increment t. (An index inside the braces identifies a number of a value inside a
record, an index after the braces identifies a number of a record in a set of records, NR.)
{𝜑𝑖 }𝑗 ; 𝑖 = 1, … , 𝑁𝑗 ; 𝑗 = 1, … , 𝑁𝑅 . (5.3.19)
5.3.3.2 The decorrelation time for roll motion, Td, is evaluated as described in paragraphs 3.3
through 3.8 of appendix 4.
5.3.3.3 The intermediate threshold, us, for roll angles to starboard is established in order to
achieve an average of 7 to 10 up-crossings for each 30 minutes of simulation time.
The up-crossing of the threshold, us, is counted at each time when
{ 𝜑𝑖−1 }𝑗 < 𝑢𝑠 𝑎𝑛𝑑 { 𝜑𝑖 }𝑗 ≥ 𝑢𝑠 𝑓𝑜𝑟 𝑢𝑠 > 0 (5.3.20)
5.3.3.4 The intermediate threshold, up, for roll angles to port is established in order to achieve
an average of 7 to 10 down-crossings for each 30 minutes of simulation time.
The down-crossing of the threshold, up, is counted at each time when
{ 𝜑𝑖−1 }𝑗 ≥ 𝑢𝑝 𝑎𝑛𝑑 { 𝜑𝑖 }𝑗 < 𝑢𝑝 𝑓𝑜𝑟 𝑢𝑝 < 0 (5.3.21)
The additional steps of the procedure are described for roll angles to starboard only because
the computations associated with port side roll angles are analogous.

5.3.3.5 The time instances of the up-crossings with linear interpolation are calculated using
this equation:
𝑡𝑖 −𝑡𝑖−1
{ 𝑇𝑈,𝑘 }𝑗 = { 𝑡𝑖−1 + (𝑢𝑠 − 𝜑𝑖−1 )} 𝑖𝑓 ( 𝜑𝑖−1 − 𝑢𝑠 ) (𝜑𝑖 − 𝑢𝑠 ) < 0; 𝑘 = 1, … , 𝑁𝑈𝑗 (5.3.22)
𝜑𝑖 −𝜑𝑖−1 𝑗

5.3.3.6 The value of the roll rate at the instant of each up-crossing with linear interpolation is
calculated next:

24
Leadbetter, M.R., Lindgren, G. and Rootzén, H. Extremes and Related Properties of Random Sequences
and Processes. in: Springer Series in Statistics, Springer-Verlag, New York-Berlin, 1983.
25
Coles, S. An Introduction to Statistical Modeling of Extreme Values. in: Springer Series in Statistics,
Springer-Verlag London Ltd., London, 2001.

I:\CIRC\MSC\1\MSC.1-Circ.1652.docx
 MSC.1/Circ.1652
Annex, page 191

𝜑̇𝑖−𝜑̇𝑖−1
{ 𝜑̇ 0,𝑘 }𝑗 = {𝜑̇ 𝑖−1 + (𝑢𝑠 − 𝜑𝑖−1 )} 𝑖𝑓 ( 𝜑𝑖−1 − 𝑢𝑠 ) (𝜑𝑖 − 𝑢𝑠 ) < 0; 𝑘 = 1, … , 𝑁𝑈𝑗 (5.3.23)
𝑡𝑖−𝑡𝑖−1 𝑗

5.3.3.7 For each up-crossing and for all records, the evaluation of the MPM metric should be
performed:

.1 the unperturbed solution is calculated starting from { 𝑇𝑈,𝑘 }𝑗 with the set of
initial conditions { 𝑋0,𝑘 }𝑗 at the time instance { 𝑇𝑈,𝑘 }𝑗 . The duration of the
unperturbed solution is recommended to be the largest among 30 natural roll
periods, 1.5Td or 300 s. The unperturbed solution should be verified to
coincide with a portion of the original simulation following the up-crossing.

.2 The perturbed solution for 𝜑̇ 𝑝,𝑘 = 𝜑̇ 0,𝑘 + 𝑚Δ𝜑̇ , 𝑚 = 1, 2, … is calculated while

keeping all other initial conditions from the set { 𝑋0,𝑘 }𝑗 . The values for Δ𝜑̇ that
should be assumed are in the range of 0.001…0.0001 rad/s. The maximum
angle achieved in the perturbed solution during the decorrelation time Td
should be recorded {(𝜑𝑚𝑎𝑥,𝑚 )𝑘 } .
𝑗

.3 Once the {(𝜑𝑚𝑎𝑥,𝑚 )𝑘 } ≥ 400 (or whatever roll angle is taken to be critical)
𝑗
is observed for 𝜑̇ 𝑝,𝑘 = 𝜑̇ 0,𝑘 + 𝑚Δ𝜑̇ , the critical roll rate is computed as
𝜑̇ 𝐶,𝑘 = 𝜑̇ 0,𝑘 + (𝑚 − 1)Δ𝜑̇ . (5.3.24)

.4 The metric value for the k-th up-crossing at the j-th record is computed as
𝑦𝑘 = 𝑐 + 𝜑̇ 0,𝑘 − 𝜑̇ 𝐶,𝑘 ; 𝑐 = 1.0 𝑟𝑎𝑑/𝑠. (5.3.25)

5.3.3.8 For all the metric values computed for each record 𝑦𝑘 , a de-clustering is carried out
using the following steps:

.1 The cluster is defined as a group of the metric values, 𝑦𝑘 , corresponding to

up-crossings that have occurred in time instances that are closer to each
𝑘=𝑓
other than the decorrelation time Td: {𝑦𝑘 }𝑘=𝑏𝑞 , 𝑞 = 1, 2, …, where bq and fq are
𝑞
indices of the beginning and the end of q-th cluster.

.2 The de-clustered values of the metric are determined as the maximum value
𝑘=𝑓
within each cluster, 𝑦1𝑞 = max ({𝑦𝑘 }𝑘=𝑏𝑞 ).
𝑞

5.3.3.9 The de-clustered values of the metric, 𝑦1𝑞 , are considered independent, xn, and are
presented in a single record for fitting of an exponential tail for the distribution as explained in
paragraph 5.3.2.7.26
𝑥𝑛 = {𝑦1𝑞 }𝑗 ; 𝑛 = 1, … , 𝑁𝑥 (5.3.26)

26
Belenky, V., Weems, K., Pipiras, V. and Glotzer, D. (2018). Extreme-value properties of the split-time metric,
Proc. 13th Intl. Conf. on Stability of Ships and Ocean Vehicles STAB 2018, Kobe, Japan.

I:\CIRC\MSC\1\MSC.1-Circ.1652.docx
MSC.1/Circ.1652
Annex, page 192

5.3.3.10 The prediction error technique27 is available for automatically finding the beginning of
the distribution tail. Certain caution should be exercised while using automatic threshold finding
algorithms, in general, because the sample data sets are subjects to random variation.

5.3.3.11 To use the prediction error technique, the metric values, xn, are sorted in descending
order
𝑥1𝑛 = sort desc (𝑥𝑛 ) (5.3.27)

5.3.3.12 A starting index and a final index for the search for the beginning of the distribution
tail, should be established. The interval for automatic search of the threshold should contain
sufficiently large values for invoking extreme value properties. Also, there should be a sufficient
amount of these large values for statistical methods to be numerically stable. For the typical
data set, produced by MPM for moderate-to-high seas states, 2% largest values of the metric
usually meet this requirement. However, if the number of data points fall significantly below 40,
calculation of the error function may encounter difficulties. For the typical moderate-to-high
data set, it is not reasonable to expect that extreme value properties can be used for the data
below upper 20%. The range for automatic threshold search is initial set as
𝑘𝑏𝑒𝑔 = min(40, 0.02𝑁𝑥 ) ; 𝑘𝑓𝑖𝑛 = 0.2𝑁𝑥 (5.3.28)
The range may be adjusted if the data set significantly differs from recommendations in 5.3.3.3.

5.3.3.13 The distribution tail of the MPM metric is approximated with conditional exponential
distribution (exponential tail), see justification in paragraph 5.3.2.7. The pdf of the exponential
tail is expressed as (1/𝛽)exp(−𝑥/𝛽) , where the scale parameter  equals to the mean value.
An estimate for the exponential tail scale parameter, 𝛽̂𝑘 , for the tail beginning at 𝑥1𝑘 , is
calculated:
1
𝛽̂𝑘 = 𝑘 ∑𝑘𝑛=1(𝑥1𝑛 − 𝑥1𝑘 ) ; 𝑘 = 𝑘𝑏𝑒𝑔 , … , 𝑘𝑓𝑖𝑛 (5.3.29)
5.3.3.14 The determination of the beginning of the distribution tail is essentially finding how
large a value should be to use extreme properties for practical calculation. When these
properties can be used, approximated distribution tail fits well with the data. The prediction
error technique uses the difference in natural logarithms of quantiles (inverse function for
CDF). Averaged square of this difference lead to formulation of the error function for a
particular threshold.
The error function for each candidate beginning of the distribution tail, Γ̂(𝑘), is calculated:

1 𝑘+1 −1 𝑛 22 𝑘+1 −1 𝑛 2
Γ̂(𝑘) = 𝛽̂2 ∑𝑘𝑛=1 ( 𝑛 − 1) (𝑥1𝑛 − 𝑥1𝑘 + 𝛽̂𝑘 log ( )) + ∑𝑘𝑛=1 ( − 1) (log ( )) − 1
𝑘 𝑘+1 𝑘 𝑛 𝑘+1
(5.3.30)

5.3.3.15 The beginning of the distribution tail, w, is the value of the metric value that
corresponds to the index, k1, that corresponds to the minimum value of the error function, Γ̂(𝑘).
𝑘1 = 𝑘(Γ̂(𝑘) → 𝑚𝑖𝑛); 𝑤 = 𝑥1𝑘1 (5.3.31)
5.3.3.16 The values of the MPM metric are calculated at the instants of the up-crossing of the
intermediate threshold 𝑢𝑠 . Some of these values exceed the beginning of the distribution tail
for the metric. The estimate of a rate of exceedance of the beginning of the distribution tail is

27
Mager, J., Automatic threshold selection of the peaks over threshold method, Master's Thesis, Technische
Universitat Munchen, 2015.

I:\CIRC\MSC\1\MSC.1-Circ.1652.docx
 MSC.1/Circ.1652
Annex, page 193

used instead of the estimate of up-crossing of the intermediate threshold 𝑢𝑠 , in

paragraph 5.3.2.3 and, respectively:
𝑘1
𝑟̂𝑈 = 𝑁𝑅 𝑁
(5.3.32)
∆𝑡 ∑𝑗=1 𝑗

5.3.3.17 The extrapolated estimate of the failure rate through the starboard heel angles is
calculated using the exponential tail with the beginning at w (conditional exponential
distribution for the random variables exceeding w):
𝑟̂𝑆 = 𝑟̂𝑈 exp(−(c − 𝑤)/𝛽̂𝑘1 ) (5.3.33)
5.3.3.18 The estimate of scale parameter, 𝛽̂𝑘1 is essentially an estimate of a mean value
(see 5.3.3.13). Given a sample 𝑧𝑖 ; 𝑖 = 1 … 𝑁 , the standard deviation 𝜎̂𝐸 of a mean value
estimate 𝐸̂ = (1/𝑁) ∑𝑁𝑖=1 𝑧𝑖 can be found with the estimate of standard deviation
2 1/2
̂ = ((1/𝑁) ∑𝑁
σ ̂
𝑖=1(𝑧𝑖 − 𝐸 ) ) as 𝜎̂𝐸 = σ
̂/√𝑁. The estimate of the standard deviation, 𝜎̂𝛽 , of the
scale parameter estimate, 𝛽̂𝑘1 , is calculated:
1 2 ̂2
𝜎̂𝛽 = √∑𝑘1 2
𝑛=1(𝑥1𝑛 − 𝑤 ) − 𝑘1 𝛽𝑘 (5.3.34)
𝑘1

5.3.3.19 Then, the estimate of the standard deviation of the rate of up-crossing through the
beginning of the distribution tail, 𝑟̂𝑈 , is calculated as:
√𝑘1(1−𝑟̂𝑈 ∆𝑡) √𝑘1
𝜎̂𝑈 = 𝑁
𝑅 𝑁
≈ 𝑁𝑅 𝑁
(5.3.35)
∆𝑡 ∑𝑗=1 𝑗 ∆𝑡 ∑𝑗=1 𝑗

5.3.3.20 A half non-dimensional confidence interval, Kα1, is calculated using the standard
function for a normal standard quantile QN (i.e., a standard deviation equal to 1 with a mean
value equal to zero) with a confidence probability 1 -  :
1+√1−𝛼
𝐾𝛼1 = 𝑄𝑁 ( ) (5.3.36)
2

5.3.3.21 The lower and upper boundaries of the confidence interval of the extrapolated
estimate, 𝑟̂𝑆,𝑙𝑜𝑤 and 𝑟̂𝑆,𝑢𝑝 , respectively, are calculated as:

𝑟̂𝑆,𝑙𝑜𝑤 = (𝑟̂𝑈 − 𝐾𝛼1 𝜎̂𝑈 )exp (−(c − 𝑤)/(𝛾̂𝑘1 − 𝐾𝛼1 𝜎̂𝛽 )) (5.3.37)

𝑟̂𝑆,𝑢𝑝 = (𝑟̂𝑈 + 𝐾𝛼1 𝜎̂𝑈 )exp (−(c − 𝑤)/(𝛾̂𝑘1 + 𝐾𝛼1 𝜎̂𝛽 )) (5.3.38)
5.3.3.22 The procedure for the estimation of the extrapolated failure rate with port roll angles
differs only in the formula for the metric:
𝑦𝑃,𝑖 = 𝑐 + |𝜑̇ 𝐷,𝑖 | − |𝜑̇ 𝑃,𝑖 | ; 𝑐 = 1 rad/s ; 𝑖 = 1, … , 𝑁𝐷 ] (5.3.39)

5.3.3.23 The split-time/motion perturbation method can be applied for the excessive
acceleration failure mode. The metric for the excessive acceleration failure mode is formulated
on condition of exceedance of the target lateral acceleration.
5.3.4 Example of application

5.3.4.1 The extrapolation procedure is demonstrated for the ONR tumblehome topside ship.
Table 5.3.1 shows the principal dimensions and environmental parameters for this design.
To demonstrate the capabilities of the split-time method, capsizing event through the starboard
was taken as a target for extrapolation. The target of extrapolation makes difference only for
the metric calculation.

I:\CIRC\MSC\1\MSC.1-Circ.1652.docx
MSC.1/Circ.1652
Annex, page 194

Table 5.3.1 Principal dimensions and environmental parameters

Length BP, m 154 Significant wave height, m 9
Breadth, m 18 Mean zero-crossing period, s 14
Draught, m 5.5 Speed, knots 6
KG, m 7.5 Relative wave heading, deg 45

5.3.4.2 The extrapolation data sample consists of 86 half-hour records. The intermediate
threshold was chosen at 12 degrees, resulting in 569 up-crossing events over 86 records.
The de-clusterizing procedure produced 369 independent values of the metric, ranging
from 0.597 rad/s to 0.993 rad/s.

5.3.4.3 For the choice of the threshold with the prediction error criterion, the mean squares
prediction error function is shown in figure 5.3.5; other intermediate numerical results are given
in table 5.3.2, and the final extrapolation results are shown in figure 5.3.6.

Table 5.3.2 Intermediate results of fitting exponential tail

Secondary threshold, u, rad/s 0.77
Available points 30
Parameter 𝛽̂ rad/s 0.066
Variance of 𝛽̂, (rad/s)2 1.041e-4
Conditional extrapolated 0.03
estimate ̂𝑃 (𝑦 > c)
Rate exceedance of secondary 1.971e-4
threshold 𝑟̂𝑤 , s-1
Failure rate 𝑟̂ , s-1 5.936e-6

3 Value of mean squares prediction error function Γ̂(𝑘)

0.72 0.74 0.76 0.78 0.8 0.82 0.84 Threshold, u=xk rad/s

Figure 5.3.5 Mean squares prediction error function

I:\CIRC\MSC\1\MSC.1-Circ.1652.docx
 MSC.1/Circ.1652
Annex, page 195

10-4
Failure rate 𝑟̂ , s-1

10-5

10-6

10-7

Figure 5.3.6 Results of extrapolation for exceedance of 40 degrees

5.4 Envelope peaks over threshold method for pure loss of stability and dead ship
condition
5.4.1 Theoretical background

5.4.1.1 This section provides essential information concerning application and validation of
the envelope peaks over threshold (EPOT) method, one of the statistical extrapolation
methods included in the Interim Guidelines.

5.4.1.2 The stability failure modes for which the EPOT method is applicable include the dead
ship condition and pure loss of stability. The applicability of the EPOT method may be extended
to the parametric roll failure mode after adjustment of the de-clustering procedure and a proper
statistical validation. EPOT is, in principle, also applicable to the excessive acceleration
stability failure mode.

5.4.1.3 The basic idea of the peaks over threshold (POT) approach is to fit a generalized
Pareto distribution (GPD) to the observed data above a particular threshold value of the
response (e.g. above a certain roll angle). This idea is extended in this method as described
below to become the EPOT method.
The mathematical background of the method is the second extreme value theorem, which
states that the tail of an extreme value distribution can be approximated with a GPD above a
"large enough" value.28 See section 5.3.0 for the theoretical background for this method. A key
feature of the POT extrapolation is that it can capture the non-linearity of the large amplitude
response, such as that caused by the changes in the restoring moment in large roll angles and
in waves.

5.4.1.4 However, the standard POT method is only applicable to independent data. The roll
motions of a ship are not independent because they are correlated through the ship's inertia,
the wave excitation, and the "memory" in the hydrodynamic forces. Therefore, the application
of POT requires an extraction of independent points (or "de-clustering") from the roll motion
data time history. Fitting an envelope to the roll motion time history, Figure 5.4.1, is a
convenient way to de-cluster the data if the peaks of the envelope of the roll response are
sufficiently far from each other to provide the necessary independence. The use of an envelope
to de-cluster the roll motion provides the additional letter in the acronym of the method, so POT
becomes EPOT – envelope peaks over threshold. An additional check may be necessary for

28
Pickands, J. Statistical inference using extreme order statistics, The Annals of Statistics, Vol. 3, No.1,
pp. 119-131, 1975.

I:\CIRC\MSC\1\MSC.1-Circ.1652.docx
MSC.1/Circ.1652
Annex, page 196

application to parametric roll motion because the decorrelation time may be large. The time
duration between the peaks of the envelope should not be less than the decorrelation time,
which is estimated as described in section 3.8.

20 (t), deg Peak of the

envelope
10

0 t, s Figure 5.4.1 De-clustering using

210 220 230 240 250
envelope
-10
Envelope
-20

5.4.1.5 The uncertainty of the extrapolated estimate can be decreased by inclusion of physical
considerations in a data-driven model. In particular, this can be done by assuming the type of
the tail, based on the physics of non-linear roll motion as explained in
paragraphs 5.4.1.6 - 5.4.1.14

5.4.1.6 The GZ curve of most ships features both a maximum GZ value near the midpoint of
the range of stability and an angle of vanishing stability at which unstable equilibria appear.
These GZ curve characteristics lead to a heavy tail after the angle of the maximum GZ curve,
which switches to a light tail as the roll angle approaches and is close to the angle of vanishing
stability. Figure 5.4.229 shows such a probability distribution function (pdf) that is computed for
a dynamical system with piecewise linear restoring (the piecewise linear approximation of the
GZ curve allows a closed-form solution for the tail of the distribution of the peaks and the
instantaneous values of the roll angle).
100
PDF( max)
10
Piecewise linear response
1
0.1 Inflection point
0.01
10-3 Figure 5.4.2 PDFs of peaks of linear
-4
Linear response response and piecewise linear
10
10-5
restoring response
10-6
10-7
10-8

0.5 0.6 0.7 0.8 0.9 max, rad

5.4.1.7 The roll angles associated with dynamic stability failures, are usually located around
and beyond the angle of the maximum of the GZ curve. Therefore, the assumption of a heavy
tail appears appropriate for extrapolation problems associated with dynamical stability failures.
When the shape parameter indicates a heavy tail, ξ > 0, and threshold value 𝑢 = 𝜎⁄ξ, then the
GPD is equivalent to a Pareto distribution with scale 𝑦𝑚 = 𝑢 = 𝜎⁄ξ and shape 𝛾 = 1⁄ξ:

𝑦
𝛾
𝑢𝛾 𝑦 𝛾 𝑢 𝛾
𝑚
pdf(𝑦) = 𝛾 𝑦𝛾+1 = 𝛾 𝑦𝛾+1 cdf(𝑦) = 1 − ( 𝑦𝑚 ) = 1 − (𝑦) (5.4.1)

Note that the Pareto distribution is a natural logarithm of an exponential distribution.

29
Belenky, V., Glotzer, D., Pipiras, V. and Sapsis, T. Distribution tail structure and extreme value analysis of
constrained piecewise linear oscillators. Probabilistic Engineering Mechanics Vol. 57, pp. 1-13, 2019.

I:\CIRC\MSC\1\MSC.1-Circ.1652.docx
 MSC.1/Circ.1652
Annex, page 197

5.4.1.8 The conditional probability of exceedance of a target value y associated with dynamic
stability failure is expressed as:

P(Y > y | Y > u) = (u/y) = (y/u)-1/ (5.4.2)

5.4.1.9 In equation (5.4.2), the threshold u (paragraph 5.4.1.8) is determined from the
applicability considerations so that only one parameter needs to be fitted. Decreasing the
number of parameters from two (in case the GPD is used) to one decreases the statistical
uncertainty.

5.4.1.10 Caution should be exercised when approximating the tail for the excessive
acceleration failure mode in the following areas:

.1 When using the GPD, a negative value of the shape parameter may indicate
excessive uncertainty caused by insufficient data.30

.2 Using the Pareto distribution contains the assumption that the roll motions
are in the vicinity of roll angles at which the GZ curve is at its maximum value.

.3 An exponential tail approximation (that uses the prediction error

technique, see 5.3.3) may be used when roll angles are not expected to be
the vicinity of maximum GZ curve.

5.4.2 Description of the EPOT extrapolation procedure

5.4.2.1 The input data for this extrapolation procedure is produced by a numerical simulation
that is carried out as required in section 3.3 of the Interim Guidelines. The input data are
presented in a form of a set of NR records (referred to as an Ensemble), each of which contains
Nj instantaneous roll angles {𝜑𝑖 }𝑗 , computed at time instances {𝑡𝑖 }𝑗 using a constant time
increment t (a symbol given in braces identifies a record, an index inside the braces identifies
a number of a value inside a record, and an index after the braces identifies a number of a
record in a set of records):
{𝜑𝑖 }𝑗 ; 𝑖 = 1, … , 𝑁𝑗 ; 𝑗 = 1, … , 𝑁𝑅 . (5.4.3)
5.4.2.2 The ensemble-averaged mean value (𝐸̂𝜑𝑎 ) is estimated as:
𝑁 𝑁𝑗
𝑅 ∑
∑𝑗=1 { 𝜑𝑖 } 𝑗
𝐸̂𝜑𝑎 = 𝑖=1
𝑁 𝑅 𝑁
(5.4.4)
∑𝑗=1 𝑗

5.4.2.3 The instantaneous roll angles of a set of roll records ({φC,i}j) is centred relative to the
estimate of the ensemble-averaged mean value:
{ 𝜑𝐶,𝑖 }𝑗 = { 𝜑𝑖 }𝑗 − 𝐸̂𝜑𝑎 (5.4.5)

5.4.2.4 The zero-crossing time instances ({TZ,q}j) are found with linear interpolation, for each
record in the set, NZ,j :
𝑡𝑖−𝑡𝑖−1
{ 𝑇𝑍,𝑞 } = { 𝑡𝑖−1 − 𝜑𝐶𝑖−1 } 𝑖𝑓 𝜑𝐶𝑖−1 𝜑𝐶𝑖 < 0; 𝑞 = 1, … 𝑁𝑍,𝑗 (5.4.6)
𝑗 𝜑𝐶𝑖 −𝜑𝐶𝑖−1 𝑗

30
Pipiras, V. Pitfalls of data-driven peaks-over-threshold analysis: Perspectives from extreme ship motion.
Probabilistic Engineering Mechanics, Vol. 60 103053 doi.org/10.1016/j.probengmech.2020.103053, 2020.

I:\CIRC\MSC\1\MSC.1-Circ.1652.docx
MSC.1/Circ.1652
Annex, page 198

5.4.2.5 The values of an envelope are peaks that exist between two zero-crossing time
instances ({φe,q}j). These values are found as the maxima of the absolute values of the
instantaneous roll angles between two adjacent zero-crossing time instances:
{ 𝜑𝑒,𝑞 }𝑗 = { max (|𝜑𝐶𝑖 |)}𝑗 𝑖𝑓 { 𝑇𝑍,𝑞 < 𝑡𝑖 ≤ 𝑇𝑍,𝑞+1 }𝑗 ; 𝑞 = 1, … 𝑁𝑍,𝑗 − 1 (5.4.7)

5.4.2.6 The time instances, corresponding to the values of the envelope, ({te,q}j), are
recorded:
{ 𝑡𝑒,𝑞 } ; 𝑞 = 1, … 𝑁𝑍,𝑗 − 1 (5.4.8)
𝑗

5.4.2.7 The ensemble-averaged mean value of the envelope (𝐸̂𝜑,𝑒 ) is estimated as:
𝑁 𝑁𝑧𝑗−1
𝑅 ∑
∑𝑗=1 { 𝜑𝑒,𝑞}
𝑞=1 𝑗
𝐸̂𝜑,𝑒 = 𝑁𝑅 (5.4.9)
∑𝑗=1(𝑁𝑧𝑗 −1)

5.4.2.8 The time at which the envelope crosses (or intersects) the ensemble-averaged mean
value of the envelope, Êφ,e, is called an envelope mean-crossing time instance {Te,m}j. and each
record j contains certain number NE,j of such time instances. The values of these time instances
{Te,m}j are determined with linear interpolation for each of the NE,j time instances in each
record j :
𝑡 −𝑡
{ 𝑇𝑒,𝑚 }𝑗 = { 𝑡𝑒,𝑞−1 + 𝜑𝑒,𝑞 −𝜑𝑒,𝑞−1 (𝐸̂𝜑,𝑒 − 𝜑𝑒,𝑞−1 )} ; 𝑚 = 1, … , 𝑁𝐸,𝑗 (5.4.10)
𝑒,𝑞 𝑒,𝑞−1 𝑗

5.4.2.9 The values of the mean-crossing peaks of the envelope ({φp,m}j) are found as the
maxima of the envelope between two adjacent mean-crossing time instances, exceeding the
estimated mean of the envelope (values of the envelope below 𝐸̂𝜑,𝑒 are not considered)
{ 𝜑𝑝,𝑚 }𝑗 = { max (𝜑𝑒,𝑞 )}𝑗 𝑖𝑓 { 𝑇𝑒,𝑚 < 𝑇𝑒,𝑞 ≤ 𝑇𝑒,𝑚+1 and 𝜑𝑒,𝑞 > 𝐸̂𝜑,𝑒 }𝑗 ; 𝑚 = 1, … 𝑁𝐸,𝑗 − 1
(5.4.11)
5.4.2.10 The peaks of the envelope are independent. They are presented in a single record
and sorted in descending order to enable the use of the prediction error technique for threshold
selection:31
𝑁𝑅
𝑌𝑛 = sort desc { 𝜑𝑝,𝑚 }𝑗 ; 𝑛 = 1, … , 𝑁𝑌 ; 𝑁𝑌 = ∑𝑗=1 (𝑁𝐸,𝑗 − 1) (5.4.12)

5.4.2.11 To search for the threshold, a beginning index (kbeg) and a final index (kfin) should be
established. The interval for automatic search of the threshold should contain sufficiently large
values for invoking extreme value properties. Also, there should be sufficient amount of these
large values for statistical methods to be numerically stable. For the typical data set, produced
for EPOT in moderate-to-high seas states, the 2% largest envelope peak values usually meet
this requirement. However, if the number of data points fall significantly below 40, the
calculation of the error function may encounter difficulties. Experience has shown that, for the
typical moderate-to-high seas data set, extreme value properties should not be used for the
data below the upper 20%. The range for the automatic threshold search is initially set as:
𝑘𝑏𝑒𝑔 = min(40, 0,02𝑁𝑌 ) ; 𝑘𝑓𝑖𝑛 = 0.2𝑁𝑌 (5.4.13)

5.4.2.12 The Hill’s estimator for the shape parameter, k is computed for each threshold, k:
1
𝜉̂𝑘 = 𝑘 ∑𝑘𝑛=1 log(𝑌𝑛 /𝑌𝑘 ); 𝑘 = 𝑘𝑏𝑒𝑔 , … , 𝑘𝑓𝑖𝑛 (5.4.14)

31
Mager, J. Automatic threshold selection of the peaks over threshold method. Master's Thesis, Technische
Universitat Munchen, 2015.

I:\CIRC\MSC\1\MSC.1-Circ.1652.docx
 MSC.1/Circ.1652
Annex, page 199

Hill's estimator is a standard technique for the estimation of the shape parameter for the Pareto
distribution. Hill’s estimator is based on the relation between the Pareto and exponential
distributions (see paragraph 5.4.1.7). The estimate is essentially an average of the natural
logarithm of data above the threshold (this can be compared to equation (5.3.15) in
paragraph 5.3.3.14).

5.4.2.13 The determination of the beginning of the distribution tail is essentially finding the
location at which a large value should be in order to use extreme properties of the distribution
for practical calculation. When these properties can be used, the approximated distribution tail
fits well with the data. The prediction error technique uses the difference in natural logarithms
of quantiles (the inverse function for the cumulative function distribution (cfd). The averaged
square of this difference leads to formulation of the error function for a particular threshold.
The error function (Γ̂(𝑘)) is computed as:
2
1 (log(𝑌𝑛−1 /𝑌𝑘−1)+𝜉̂𝑘 log(𝑛/(𝑘+1))) 2 (log(𝑛/(𝑘+1)))2
Γ̂(𝑘) = 𝑘
2 ∑𝑛=1 + ∑𝑘𝑛=1 −1 (5.4.15)
𝜉̂𝑘 (∑𝑘 −2
𝑗=𝑛 𝑗 )
𝑘 (∑𝑘 −2
𝑗=𝑛 𝑗 )

5.4.2.14 The index, (k1), that corresponds to the minimum of the error function, (Γ̂(𝑘)), and the
threshold, (k), corresponding to this index are determined as:
𝑘1 = 𝑘(Γ̂(𝑘) → 𝑚𝑖𝑛); 𝑢 = 𝑌𝑘1 (5.4.16)
5.4.2.15 The estimate of a rate of exceedance (𝑟̂𝑢 ) over the threshold, u, is calculated as:
𝑘1
r̂𝑢 = 𝑁𝑅 𝑁
(5.4.17)
∆𝑡 ∑𝑗=1 𝑗

5.4.2.16 The extrapolated estimate of exceedance rate (𝑟̂𝑐 ) of the target value, c, is then
calculated as:
̂
𝑟̂𝑐 = 𝑟̂𝑢 (𝑐/𝑢)−1⁄𝜉𝑘1 (5.4.18)

5.4.2.17 The estimate of the standard deviation of rate (𝜎̂𝑢 ) of the exceedance, 𝑟̂𝑢 , is calculated
as:
√𝑘1(1−𝑟̂𝑢∆𝑡) √𝑘1
𝜎̂𝑢 = 𝑁𝑅 ≈ 𝑁𝑅 𝑁
(5.4.19)
∆𝑡 ∑𝑗=1 𝑁𝑗 ∆𝑡 ∑𝑗=1 𝑗

5.4.2.18 The estimate of the standard deviation (𝜎̂𝜉 ) of the shape parameter, 𝜉̂𝑘1 , is calculated
as a standard deviation for a mean value estimate (since Hill’s estimator is a mean value of
natural logarithms of data points, see paragraph 5.4.2.12):
𝜎̂𝜉 = 𝜉̂𝑘1 /√𝑘1 (5.4.20)
5.4.2.19 Because EPOT combines two estimates, each of which is computed with its
confidence interval, the final result should be presented with the confidence probability 1 − 𝛼 =
0.95 and, for this reason, each component estimate is computed with the confidence probability
√1 − 𝛼 as:
1+√1−𝛼
𝐾𝛼1 = 𝑄𝑁 ( 2
) (5.4.21)

5.4.2.20 The boundaries of the confidence interval of the extrapolated estimate, (𝑟̂𝑐,𝑙𝑜𝑤 and
𝑟̂𝑐,𝑢𝑝 , respectively) are calculated as:
̂
r̂𝑐,𝑙𝑜𝑤 = (𝑟̂𝑢 − 𝐾𝛼1 𝜎̂𝑢 )(𝑐/𝑢)−1⁄(𝜉𝑘1 −𝐾𝛼1𝜎̂𝜉) (5.4.22)
̂
𝑟̂𝑐,𝑢𝑝 = (𝑟̂𝑢 + 𝐾𝛼1 𝜎̂𝑢 )(𝑐/𝑢)−1⁄(𝜉𝑘1+𝐾𝛼1𝜎̂𝜉) (5.4.23)

I:\CIRC\MSC\1\MSC.1-Circ.1652.docx
MSC.1/Circ.1652
Annex, page 200

5.4.3 Description of the extrapolation procedure and an example of its application

5.4.3.1 The extrapolation procedure is demonstrated for the ONR tumblehome topside ship.
Table 5.3.1 shows the principal dimensions and environmental parameters of this ship design.

5.4.3.2 The extrapolation data sample consists of 86 half-hour records, produced with a
volume-based simulation tool. A portion of a sample record is shown in figure 5.4.3 and a
fragment of the envelope for the de-clustering procedure is shown in figure 5.4.4. The objective
is to estimate an exceedance rate for a roll angle of 40 degrees as given by the Interim
Guidelines.

20
Roll, deg

10

-10

-20 Time, s
0 50 100 150 200 250 300 350 400

Figure 5.4.3 An example of roll record

Roll φ, deg
20 Envelope Mean-crossing peaks of envelope, φe
Mean of the
envelope, Êφ,e
10

-10

Time, s
-20
40 60 80 100 120 140 160 180 200

Figure 5.4.4 A fragment of the envelope and the procedure of de-clustering with
mean-crossing peaks

5.4.3.3 The total number of data points after de-clustering is determined: 𝑁𝑌 = 2619 .
Per paragraph 5.4.2.11, the beginning and final index are evaluated as:
𝑘𝑏𝑒𝑔 = min(40, 0,02𝑁𝑌 ) = 40 ; 𝑘𝑓𝑖𝑛 = 0.2𝑁𝑌 = 522 (5.4.24)

I:\CIRC\MSC\1\MSC.1-Circ.1652.docx
 MSC.1/Circ.1652
Annex, page 201

5.4.3.4 The threshold u is found by an index that corresponds to a minimum of the mean
squared prediction error function that is computed per paragraph 5.4.2.13 and shown in
figure 5.4.5:

4 Value of error function Γ̂(𝑘)

12 13 14 15 16 17 Threshold, u = Yk , deg

Figure 5.4.5 The mean squares prediction error function

5.4.3.5 As described in paragraph 5.4.2.14, the index k1 corresponding to a minimum of the

error function Γ̂(𝑘) and the threshold are calculated as:
𝑘1 = 𝑘(Γ̂(𝑘) → 𝑚𝑖𝑛) = 446; 𝑢 = 𝑌𝑘1 = 17.370 (5.4.25)

5.4.3.6 Per paragraph 5.4.2.15, the estimate of a rate of exceedance (𝑟̂𝑢 ) over the threshold,
u, is calculated as:
𝑘1
𝑟̂𝑢 = 𝑁 𝑅 𝑁
= 4.99 ∙ 10−4 1/𝑠 (5.4.26)
∆𝑡 ∑𝑗=1 𝑗

5.4.3.7 Per paragraph 5.4.2.12, the Hill estimator for the shape parameter, k1 is computed
for the selected threshold:
1
𝜉̂𝑘1 = 𝑘1 ∑𝑘1
𝑛=1 log(𝑌𝑛 /𝑌𝑘1 ) = 0.161 (5.4.27)

5.4.3.8 Per paragraph 5.4.2.16, the extrapolated estimate of the exceedance rate (𝑟̂𝑐 ) of the
target value, 𝑐 = 400 , is then calculated as:
̂
𝑟̂𝑐 = 𝑟̂𝑢 (𝑐/𝑢)−1⁄𝜉𝑘1 = 2.76 ∙ 10−6 1/𝑠 (5.4.28)

5.4.3.9 Per paragraph 5.4.2.17, the estimate of the standard deviation of the rate (𝜎̂𝑢 ) of the
exceedance, 𝑟̂𝑢 , is calculated as:
√𝑘1(1−𝑟̂𝑢∆𝑡) √𝑘1
𝜎̂𝑢 = 𝑁
𝑅 𝑁
≈ 𝑁𝑅 𝑁
= 5.73 ∙ 10−5 1/𝑠 (5.4.29)
∆𝑡 ∑𝑗=1 𝑗 ∆𝑡 ∑𝑗=1 𝑗

I:\CIRC\MSC\1\MSC.1-Circ.1652.docx
MSC.1/Circ.1652
Annex, page 202

5.4.3.10 Per paragraph 5.4.2.18, the estimate of the standard deviation (𝜎̂𝜉 ) of the shape
parameter, 𝜉̂𝑘1 , is calculated as:
𝜉̂ 𝑘1
𝜎̂𝜉 = = 0.018 (5.4.30)
√𝑘1

5.4.3.11 As EPOT combines two estimates, each is computed with its confidence interval.
In order to present the final result with the confidence probability 1 − 𝛼1 = √ 1 − 𝛼 = 0.975,
each component estimate is computed with the confidence probability 1 − 𝛼1 = √ 1 − 𝛼 =
0.975. Per paragraph 5.4.2.19, a one-half non-dimensional confidence interval (K1) that uses
a standard function for a normal standard quantile QN (i.e. a standard deviation = 1, with a zero
mean) with confidence probability, 𝛼1 = 0.975 is calculated as:
1+√1−𝛼
𝐾𝛼1 = 𝑄𝑁 ( 2
) = 2.236 (5.4.31)

5.4.3.12 Per paragraph 5.4.2.20, the boundaries of the confidence interval of the extrapolated
estimate, (λ̂𝑐,𝑙𝑜𝑤 and λ̂𝑐,𝑢𝑝 , respectively) are calculated as:
̂
𝑟̂𝑐,𝑙𝑜𝑤 = (𝑟̂𝑢 − 𝐾𝛼1 𝜎̂𝑢 )(𝑐/𝑢)−1⁄(𝜉𝑘1−𝐾𝛼1𝜎̂𝜉) = 4.54 ∙ 10−7 1/𝑠 (5.4.32)
̂
𝑟̂𝑐,𝑢𝑝 = (𝑟̂𝑢 + 𝐾𝛼1 𝜎̂𝑢 )(𝑐/𝑢)−1⁄(𝜉𝑘1+𝐾𝛼1𝜎̂𝜉) = 9.00 ∙ 10−6 1/𝑠 (5.4.33)
The result of extrapolation for 𝑐 = 400 with its confidence interval is shown in figure 5.4.6.

10-4 Failure rate 𝑟̂ , s-1.

10-5

Figure 5.4.6 The results of

an extrapolation for a target
roll angle of 40 degrees
10-6

10-7

5.4.4 Statistical validation

5.4.4.1 Limited statistical validation of EPOT was carried out following the recommendations
of paragraph 3.5.6 of the Interim Guidelines. According to the recommendation in paragraph
3.5.6.3, a reduced order mathematical model (volume-based 3-DOF calculations) was applied.
This fast code creates very large samples of data in which large roll angles associated with
rare failures are observable. The observations estimate a "true value" from direct counting.

5.4.4.2 A series of validation data sets was computed for the ONR tumblehome configuration
with KG = 7.5 m, GM = 2.2 m. Simulations were performed with independent pseudo-random
realizations of a seaway with a Bretschneider spectrum at a significant wave height 9 m, modal
period 15 s and ship speed 6 knots. Table 5.4.1 shows other simulations parameters. Several
maximum roll angles defining stability failure angles are examined.

I:\CIRC\MSC\1\MSC.1-Circ.1652.docx
 MSC.1/Circ.1652
Annex, page 203

Table 5.4.1 "True value" calculations

Total number
Heading, Number of Largest Number of exceedances of
of 30-min
deg. targets target largest target
records
15 570,000 5 20 14
22.5 200,000 7 27.5 16
30 200,000 13 45 9
37.5 200,000 15 60 7
45 690,000 15 70 8
60 600,000 15 70 12
90 690,000 9 37.5 12
135 690,000 3 20 6

5.4.4.3 The extrapolation procedure was applied to a series of small subsets of this large
sample and the extrapolated estimates were compared with the "true value". Figure 5.4.7
shows an example comparison for a 45 degree heading (stern-quartering seas) and a target
roll value of 45 degrees. Fifty extrapolation estimates are carried out, each computed from 100
hours of data. The main index of performance is the passing rate, which indicates the
percentage of successful extrapolations. An extrapolation is considered successful if the
confidence interval of the extrapolated exceedance rate includes the "true value". The example
shown in figure 5.4.8 has 45 successful extrapolations, resulting in a passing rate of 90%.

10-5 Extrapolated estimate 1/s

"True"

10-6
value

10-7
Figure 5.4.7 example of
10-8 extrapolation validation
for a heading of
10-9 45 degrees and target
value of 45 degrees
10-10

10-11
Extrapolated case index
0 10 20 30 40 50

5.4.4.4 Three validation tiers32 are applied: extrapolation for one target value; extrapolation
for all target values, and extrapolation for all operational and environmental conditions.

5.4.4.5 The tier 1 validation is a set of comparisons of extrapolated estimates with the true
value, figure 5.4.7. The second tier considers all available target angles; the passing rates are
shown in figure 5.4.8. An acceptable passing rate for 50 extrapolation data sets is from 0.88
to 1 (paragraph 3.5.6.7 of the Interim Guidelines). This variation of the passing rate can be
explained by the natural variability of the statistical estimates. The extrapolations are
acceptable for all targets excluding 50 and 60 degrees, for which the passing rates fell to 0.86.
The average passing rate for the 45 degrees heading is 0.90, which is within the acceptable
range.

32
Smith, T.C. Validation Approach for Statistical Extrapolation. Chapter 34 of Contemporary Ideas on Ship
Stability. Risk of Capsizing, Belenky, V., Neves, M., Spyrou, K., Umeda, N., van Walree, F., eds., Springer,
ISBN 978-3-030-00514-6, pp. 573-589, 2019.

I:\CIRC\MSC\1\MSC.1-Circ.1652.docx
MSC.1/Circ.1652
Annex, page 204

Figure 5.4.8 Passing rate for

heading 45 degrees

5.4.4.6 The third tier of validation assesses the performance over all available conditions.
The passing rates are shown in figure 5.4.9. Two lines are shown: one corresponds to an
averaged passing rate over all target values, while the other corresponds to the smallest
passing rate value encountered among all the target values. For 45 degrees heading, the latter
corresponds to a minimum shown in figure 5.4.9. The extrapolation did not work for heading
135 degrees.

Figure 5.4.9 Passing rate for

all headings

5.4.4.7 A validation at the heading of 135 degrees has a very likely failure due to insufficient
data in the non-linear region. The average conservative distance (measure of practical
statistical uncertainty) does not exceed an order of magnitude in terms of the exceedance rate.
This performance seems to be sufficient to distinguish between realistic and distant chances
of dynamic stability failure.

5.4.4.8 Overall, the validity of EPOT, except for the 135 degree heading, can be characterized
as acceptable. The passing rate falls short of the required 0.88 for few cases, but not by much.
The average passing rates exceed 0.88 for all cases except for the heading of 135 degrees.

5.5 Application of MPM and EPOT methods to full probabilistic assessment

5.5.1 A simplified sample of full probabilistic direct stability assessment was carried out for
pure loss of stability failure mode using EPOT to estimate the exceedance rate of 40 degrees
roll angle on either side of the ship and split-time method (motion perturbation method, MPM)
to estimate capsizing rate to starboard only. Calculations were carried out for the principal
dimensions and environmental parameters given in table 5.3.2. Primary focus of this example
was feasibility of computational procedures rather than quantitative assessment.
Thus, computations were performed for 3 DOF: heave, roll and pitch – surge was not included
while required by the Interim Guidelines for the quantitative assessment.

5.5.2 Computations were carried out for a relatively coarse set of environmental and
operation conditions. Speed varied from 0 to 10 knots, assuming that high speeds will not be
practical in high sea states and lower sea states do not make a significant contribution towards
the final results. The headings were 45, 90, 135, 225, 270 and 315 degrees. For computational

I:\CIRC\MSC\1\MSC.1-Circ.1652.docx
 MSC.1/Circ.1652
Annex, page 205

speed, ship motions were evaluated with volume-based code, using body-non-linear
formulation for Froude-Krylov and hydrostatic forces. The waves were long-crested, with the
conditions defined in the wave scatter table from the Interim Guidelines There were total
of 40 hours (80 records of 30 minutes duration) of simulation time histories generated for each
combination of speed and heading.

5.5.3 For the consideration of computational speed and simplicity of this example, not all
cells of the scatter diagram were used for calculations; interpolation was used between cells.
Linear extrapolation was also used for the sea states with a mean zero-crossing period
of 12.5 s; this is a conservative approach as those sea states do not contribute much to the
final result. The calculations were not performed for the cells with zero statistical weight.
Finally, calculations were limited by the sea states with significant wave height 5.5 m and
above as no appreciable estimates were obtained for significant wave height 4.5 m and below,
because contribution of these estimates are expected to be negligible, see tables 5.5.2 to 5.5.5.

5.5.4 The long-term results in table 5.5.1 were computed assuming an equal probability for
all speeds and headings for all the weather (tables 5.5.2 to 5.5.5 show results per sea state).
These results show very large values both for exceedance rate and capsizing rate compared
to the standard rate 2.6∙10-8 (1/s) in the Interim Guidelines. Such large values could be
expected because the ONR tumblehome configuration is known for its vulnerability for pure
loss of stability.

Table 5.5.1 Assessment results for equal probability of speeds and headings
Estimate rate, 1/s
estimate of exceedance rate of 40 degree 1.157e-7
Upper boundary of exceedance rate of 40 degree 2.079e-7
estimate of capsizing rate 1.434e-8
Upper boundary of capsizing rate 4.081e-8

5.5.5 Note that the average failure rate over all sea states and sailing conditions is
dominated by a few sea states characterized by high steepness of waves, for which the
estimates can be also obtained by direct counting.

Table 5.5.2 Estimate of exceedance rate of 40 degree roll, 1/s (columns are mean
zero-crossing period in seconds, rows are significant wave heights in metres)

I:\CIRC\MSC\1\MSC.1-Circ.1652.docx
MSC.1/Circ.1652
Annex, page 206

Table 5.5.3. Estimate of upper boundary of exceedance rate of 40 degree

roll, 1/s (columns are mean zero-crossing period in seconds, rows are
significant wave heights in metres)

Table 5.5.4 Capsize rate estimate, 1/s (columns are mean zero-crossing
period in seconds, rows are significant wave heights in metres)

Table 5.5.5 Upper boundary of capsize rate estimate, 1/s (columns are
mean zero-crossing period in seconds, rows are significant wave heights
in metres)

5.6 Linear superposition method for excessive acceleration failure mode

5.6.1 Description of method

5.6.1.1 Linear superposition method is one of statistical extrapolation procedures that can be
used for the direct stability assessment for the excessive acceleration failure mode if roll
damping is modelled using the stochastically equivalent linearization technique.

5.6.1.2 With this method, the energy spectra of ship responses in irregular waves can be
obtained by linear superposition using a product of the Response Amplitude Operators (RAOs)
of ship responses with the wave energy spectrum.

5.6.1.3 The variance [𝑅𝐿 (𝜇)]2 of the lateral acceleration at a constant wave heading angle 
in irregular long-crested waves is obtained using sea wave elevation spectrum 𝑆𝑍𝑍 (𝜔) and
RAO of the lateral acceleration [𝐴(𝜔, 𝜇)] by the following equation:
∞
𝑅𝐿 (𝜇)2 = ∫0 𝑆𝑍𝑍 (𝜔) ⋅ [𝐴(𝜔, 𝜇)]2 𝑑𝜔 (5.6.1)

I:\CIRC\MSC\1\MSC.1-Circ.1652.docx
 MSC.1/Circ.1652
Annex, page 207

5.6.1.4 In short-crested irregular waves, this variance is calculated considering wave energy
spreading function D with respect to the mean wave direction.

𝜋/2 ∞
𝑅𝐿 (𝜇 )2 =∫ ∫ 𝑆𝑍𝑍 (𝜔) ⋅ 𝐷(𝜔, 𝜒) ⋅ [𝐴(𝜔, 𝜇 − 𝜒)]2 𝑑𝜔 𝑑𝜒 (5.6.2)
−𝜋/2 0

5.6.1.5 Repeating such calculations for all relevant significant wave heights and
zero-crossing wave periods for a specified wave heading and ship speed, the "short-term" and
average "long-term" failure probabilities per encounter wave are calculated using the
calculated variance of lateral acceleration in the same way as in level 2 vulnerability criterion
for the excessive acceleration stability failure mode.
5.6.1.7 The average "long-term" stability failure rate, 1/s, can be calculated by

𝑟̅ = − ln(1 − 𝑝) /𝑇e (5.6.3)

where p is the average "long-term" failure probability per encountered wave and Te is the mean
wave encounter period at a specified ship speed 𝑣0 , given by

𝑇𝑤
𝑇𝑒 =
2𝜋 (5.6.4)
(1 − 𝑔𝑇 𝑣0 cos 𝜒)
𝑤

where 𝑇𝑤 is the mean wave period. In the case of full probabilistic assessment, 𝑇𝑤 is mean
value of zero-crossing wave period in North Atlantic.

5.6.1.8 The numerical method is validated in section 2.7 of this appendix. To validate the
statistical extrapolation, table 5.6.1 compares the average "long-term" failure probability per
encounter wave calculated with level 2 vulnerability criterion, the linear superposition method
and time domain simulations. An example of validation was presented for a container ship,
which is the same as section 2.7, and the modelling of short-crested irregular waves are also
the same as in section 2.7. The time domain simulation tool used for comparison models ship
motions in six degrees of freedom; hydrodynamic forces are calculated using the
three-dimensional source distribution, considering memory effect. This tool showed sufficient
accuracy for direct stability assessment for excessive acceleration failure mode in comparison
with model tests. The average "long-term" failure probabilities per encounter wave calculated
by the linear superposition method and time domain simulations are close to each other and
consistent with respect to level 2 vulnerability criterion.

Table 5.6.1 Average "long-term" failure probabilities per encounter wave calculated
with level 2 vulnerability criterion, linear superposition method and time domain
simulations

Linear
Time domain
Method: Level 2 superposition
simulations
method
Average long-term
failure probabilities
5.35·10-5 2.02·10-5 1.68·10-5
per encounter
wave

I:\CIRC\MSC\1\MSC.1-Circ.1652.docx
MSC.1/Circ.1652
Annex, page 208

5.6.2 Application example

5.6.2.1 An example of application of the linear superposition method for the excessive
acceleration stability failure mode concerns a container ship with the principal particulars in
table 2.7.1.

5.6.2.2 The transfer functions of ship motions in waves were computed in frequency domain
using the Salvesen-Tuck-Faltinsen method, based on strip theory. Details of the numerical
method are given in section 2.7. The ship motions were calculated in five degrees of freedom;
and hydrodynamic forces were calculated by a two-dimensional source distribution method.
The roll damping coefficient was estimated from a roll decay model test. Short-crested irregular
waves were employed, where the wave energy spectrum was the ITTC recommended
unlimited fetch spectrum (1978) and the wave energy spreading was described by square of
cosine function. The significant wave height and mean wave period correspond to wave scatter
diagram of IACS Recommendation No.34 (Corr.1 Nov. 2001) (see the Interim Guidelines,
Table 2.7.2.1.2).

5.6.2.3 Table 5.6.2 presents the short-term failure probability of exceedance of the lateral
acceleration 9.81 m/s2 in irregular short-crested beam waves at zero ship speed, calculated
for significant wave heights and zero-crossing wave periods corresponding to the centres of
the cells of IACS Recommendation No.34 (Corr.1 Nov. 2001) (see the Interim Guidelines,
Table 2.7.2.1.2). The average "long-term" failure probability per encountered wave 𝑝 is
obtained by integrating the product of the short-term probability in a given sea state and the
occurrence frequency of the sea state over the possible sea states in the assumed water area.
The average "long-term" failure probability per encountered wave for this wave climate was
calculated as 8.73∙10-6.

Table 5.6.2 Short-term probability of exceedance of lateral acceleration

exceeding 9.81 m/s2 in irregular short-crested beam waves
Tz Tz : average zero up-crossing wave periods [s]
Hs
Hs : significant wave height [m]

5.6.2.4 Figure 5.6.1 shows the average "long-term" probabilities of exceedance of the lateral
acceleration 9.81 m/s2 per encountered wave at zero ship speed for various wave headings
from head to following waves. For uniform distribution of wave heading probabilities, the
average "long-term" probability of exceedance of lateral acceleration 9.81 m/s2 per
encountered wave 𝑝 is 2.33·10-6. The average "long-term" stability failure rate 𝑟 (1/s) is
obtained using the average "long-term" probability per encountered wave 𝑝 and mean wave
encounter period 𝑇𝑒 , as shown in eq. (5.6.3). Here, since the ship speed is zero, the expected
value of the zero up-crossing wave period in the North Atlantic is used for 𝑇𝑒 . The average
"long-term" stability failure rate is 2.63·10 -7, which is larger than the standard of 2.6·10-8 (1/s).
Therefore, the ship is judged as unsafe to the excessive acceleration failure mode.

I:\CIRC\MSC\1\MSC.1-Circ.1652.docx
 MSC.1/Circ.1652
Annex, page 209

average long−term probability

per encountered wave
−6
10

−8
10

−10
10

0 30 60 90 120 150 180

wave heading[deg.]
Figure 5.6.1 Average "long-term" probability of exceedance of lateral acceleration
9.81 m/s2 per encountered wave depending on wave heading

6 Roll Damping

6.1 Calibration of roll damping in simulation codes

6.1.1 Paragraph 3.3.2.2 of the Interim Guidelines contains requirements for modelling roll
damping. In that context, paragraph 3.3.2.2.1 indicates that roll decays may be used for the
calibration of roll damping.

6.1.2 Most simulation codes internally compute the wave component of roll damping.
Some simulation codes allow for directly computing roll damping contributions from different
appendages (e.g. rudders, fins, bilge keels). These models, however, do not account for all
effects associated with roll damping. Therefore, semi-empirical tuning damping parameters
are also present in simulation codes for allowing calibration of roll damping. The scope of these
semi-empirical tuning damping parameters is to account for those effects that are not explicitly
addressed by specific numerical models available in the simulation code. Combining specific
numerical models and tuning damping parameters with information from roll decay test data
allows reasonable numerical simulations outside of the range of test parameters.

6.1.3 A calibration process is therefore generally carried out. The calibration process
essentially corresponds to the modification of the tuning roll damping parameters in the
simulation code in such a way that the numerically simulated roll decay is representative of the
experimental roll decay.

6.1.4 Figure 6.1 shows an example comparison between experimental roll decays and
simulated roll decays after tuning. Numerical and experimental data in the example figure are
compared in terms of dimensionless linear equivalent roll damping coefficient e as a function
of the roll amplitude, for three different forward speeds.

I:\CIRC\MSC\1\MSC.1-Circ.1652.docx
MSC.1/Circ.1652
Annex, page 210

0.08
c roll damping coefficient as
fraction of critical (-) 15 kt

10 kt
0.06

0.04
5 kt

0.02

Roll amplitude (deg)

0 2 4 6 8 10 12 14 16 18 20
Figure 6.1 Example comparison of results from experimental roll decays (markers) and
numerical roll decays after calibration (solid lines), for three different speeds.

7 Application examples of verification of failure modes

7.1 Parametric roll

7.1.1 Section 3.5.2 of the Interim Guidelines contains guidance on verification of the mode
of failure. Per paragraph 3.5.2.1 of the Interim Guidelines, the objective is to examine whether
the stability failure corresponding to the expected mode has been observed, provided that the
numerical method has been validated for reproducing stability failure of the mode under
examination. The judging criteria for the parametric roll are detailed in paragraph 3.5.2.3 of the
Interim Guidelines. They suggest comparison of the period of roll motion to the local encounter
period of waves. The roll period is expected to be close to the natural roll period at the observed
amplitude see section 2.1 of this appendix. The roll period is expected to be close to twice the
local wave encounter period.

7.1.2 Figure 7.1.1 shows the time history of roll motion for 5 minutes, from 400 s to 700 s
(the initial transition is not shown). One can easily see 10 complete roll oscillations, making the
observed roll period about 30 s. The natural roll frequency, corresponding to the considered
loading condition with GM = 1.4 m, is about 0.21 rad/s; the corresponding natural roll
period 29.9 s is very close to the visual estimate from figure 7.1.1.

7.1.3 Figure 7.1.2 shows time history of wave elevation at the position of the centre of
gravity of the ship. Periods measured from this figure are essentially wave encounter periods.
There are 23 full oscillations (counting by zero crossings), producing an encounter period
about 13 s (actually, it is close to 13.9 s, calculated by the simulation code for this record).
The ratio between the observed roll period and observed wave encounter period is about 2.3.

I:\CIRC\MSC\1\MSC.1-Circ.1652.docx
 MSC.1/Circ.1652
Annex, page 211

40
Roll, deg

20

-20

-40
450 500 550 600 650 Time, s
Figure 7.1.1 Verification of parametric roll: roll motion for heading 1° (almost following),
speed 5 knots, significant wave height 3.5 m, mean zero-crossing period 8.5 s

4
Wave elevation at CG, m

-2

-4
450 500 550 600 650 Time, s
Figure 7.1.2 Verification of parametric roll: wave elevation at CG for heading 1°
(almost following), speed 5 knots, significant wave height 3.5 m, mean zero-crossing
period 8.5 s

7.1.4 This confirms that criteria from paragraph 3.5.2.3 of the Interim Guidelines are
satisfied and stability failure can be positively verified as parametric roll.

7.1.5 Figure 7.1.3 shows another example of time record of large heel due to parametric
roll from a model experiment of a 6,600 TEU container ship in short-crested irregular head
waves of significant wave height 0.221 m and mean wave period 1.32 s. At the time of
about 139 s, the roll amplitude exceeded 20 degrees, thus this is not a stability failure
according to the Interim Guidelines but a kind of large heel incident. The local roll period
including the time instance of large roll amplitude was 2.98 s, which was almost twice the local
pitch period of 1.47 s and close to the natural roll period of 3.2 s. Thus, the procedure described
in paragraph 3.5.2.3 of the Interim Guidelines suggests that this large heel incident can be
judged as the result of parametric roll.

I:\CIRC\MSC\1\MSC.1-Circ.1652.docx
MSC.1/Circ.1652
Annex, page 212

30

20
angle (degrees)

10
roll
0
pitch
100 110 120 130 140 150 160
-10

-20

-30
time (s)
Figure 7.1.3 Time record of roll motion due to parametric roll from model experiment
for 6,600 TEU container ship in short-crested irregular head waves

7.1.6 It is noted, however, that the analysis based on pitch motion as a proxy for the analysis
based on wave elevation, due to its indirect nature, may lead to inaccurate interpretation of the
occurring phenomenon. Therefore, it is advised, whenever possible, to base the analysis on
the comparison of local roll behaviour and local wave characteristics, in accordance with
paragraph 3.5.2.3 of the Interim Guidelines.

7.2 Pure loss of stability

7.2.1 An example of time record of stability failure due to pure loss of stability from a model
experiment of the C11-class container ship in irregular long-crested stern-quartering waves is
shown in figure 7.2.1. The significant wave height was 0.165 m, the mean wave period 1.295 s,
and the nominal Froude number 0.35 and the specified autopilot course is -30 degrees from
the wave direction. At the time of about 151 s, the roll angle exceeded 40 degrees so that it is
a stability failure. Here the local roll period including the time instance of the stability failure is
close to the local pitch period, which corresponds to the local wave encounter period. The local
pitch period is about 4.4 s, which is much larger than the natural roll period of about 2 s.
In addition, the stability failure occurred when the pitch up-crossed zero. This means a
bow-upward movement as the ship is overtaken by waves, thus this up-crossing means a wave
crest located amidship, when the metacentric height is smallest. Since the increase rate of the
pitch angle near the wave crest is much smaller, the ship model spends longer time at the
wave crest. Thus, the procedure described in paragraph 3.5.2.2 of the Interim Guidelines
suggests that this stability failure can be judged as pure loss of stability.

I:\CIRC\MSC\1\MSC.1-Circ.1652.docx
 MSC.1/Circ.1652
Annex, page 213

Figure 7.2.1 Time record of stability failure due to pure loss of stability from model
experiment for C11-class container ship in irregular long-crested stern-quartering
waves

7.2.2 It is noted, however, that the analysis based on pitch motion as a proxy for the analysis
based on wave elevation, due to its indirect nature, may lead to inaccurate interpretation of the
occurring phenomenon. Therefore, it is advised, whenever possible, to base the analysis on
the comparison of local roll behaviour and local wave characteristics, in accordance with
paragraph 3.5.2.2 of the Interim Guidelines.

7.3 Surf-riding/broaching

7.3.1 Figure 7.3.1 shows an example of a time record of stability failure due to broaching
associated with surf-riding from a model experiment for the ONR flare topside vessel in
irregular long-crested stern-quartering waves at significant wave height 0.207 m, mean wave
period 1.627 s, the specified auto pilot course -15 degrees from the wave direction and nominal
Froude number 0.44. At the time instance of about 19 s, the roll angle ( 𝜑 )
exceeded 30 degrees, which is not a stability failure according to the Interim Guidelines, but a
kind of large heel incident. The rudder deflection () reached the maximum angle to starboard
but the angular velocity and angular acceleration in yaw ( ) increased in the direction of port
turn. During the initial stage of course deviation, the pitch angle ( ) has an almost constant
negative value, which indicates that the ship is temporarily surf-ridden at the wave downslope.
Thus, the procedure described in paragraph 3.5.2.4 of the Interim Guidelines suggests that the
reason of large heel can be judged as broaching.

Figure 7.3.1 Time record of large heel due to broaching associated with surf-riding from
model experiment for ONR flare topside vessel in irregular long-crested stern-quartering
waves

I:\CIRC\MSC\1\MSC.1-Circ.1652.docx
MSC.1/Circ.1652
Annex, page 214

7.4 Synchronous roll

7.4.1 Figure 7.4.1 shows an example of analysis of the relationship between the local roll
period and local roll amplitude from a model experiment for a 246 m-long cruise ship in irregular
long-crested beam waves with significant wave height 0.128 m (full scale: 10.5 m) and mean
wave period 1.93 s at zero forward speed. The natural roll period of the model was 2.60 s.
The diagram indicates that the ratio of the local roll period to the local heave period, which is
equal to the local wave encounter period, is about 1 for larger local roll amplitudes. Thus, the
procedure described in paragraph 3.5.2.5 of the Interim Guidelines suggests that larger heel
angles can be judged as synchronous roll.

7.4.2 It is noted, however, that the analysis based on heave motion as a proxy for the analysis
based on wave elevation, due to its indirect nature, may lead to inaccurate interpretation of the
occurring phenomenon. Therefore, it is advised, whenever possible, to base the analysis on
the comparison of local roll behaviour and local wave characteristics, in accordance with
paragraph 3.5.2.5 of the Interim Guidelines.

Figure 7.4.1
Local roll
amplitude
(degrees) vs. ratio
of local roll
period Troll to local
heave period Theave
from model test
for 246 m-long
cruise ship in
irregular long-
crested beam
waves; yellow
Troll / Theave colour indicates
large occurrence
frequency

I:\CIRC\MSC\1\MSC.1-Circ.1652.docx
 MSC.1/Circ.1652
Annex, page 215

APPENDIX 5

Theoretical background, validation, and application examples for Guidelines on

operational measures

1 Background information

1.1 Ships and loading conditions used in background studies

1.1.1 Five ships were used in the studies: a cruise vessel, a RoPax vessel and three
container ships of 1700, 8400 and 14000 TEU capacity. For each ship, five loading conditions
were selected: three loading conditions with small GM values (relevant for parametric roll, pure
loss of stability and stability in dead ship condition) and two loading conditions with large GM
values (relevant for excessive accelerations).

1.1.2 Table 1.1.1 summarizes the parameters of ships and loading conditions, and
figure 1.1.1 shows examples of the calm-water righting lever curves for typical loading
conditions with low metacentric height.

Table 1.1.1 Ships and loading conditions used in study

Ship LBP, m Bwl, m LC: 01 02 03 04 05
draught,
6.9
Cruise Vessel 230.9 32.2 m
GM, m 1.5 2.0 2.5 3.25 3.75
draught,
5.5
RoPax Vessel 175.0 29.5 m
GM, m 3.7 4.5 5.2 5.9 6.6
draught,
1700 TEU 9.5 5.5
159.6 28.1 m
Container Ship
GM, m 0.5 1.2 1.9 5.75 6.75
draught,
8400 TEU 13.93 14.44 14.48 11.36
317.2 43.2 m
Container Ship
GM, m 0.89 1.26 2.01 5.0 6.93
draught,
14000 TEU 14.5 8.5
349.5 51.2 m
Container Ship
GM, m 1.0 2.0 3.0 9.0 12.0

Figure 1.1.1 Calm-water righting lever curves for typical loading conditions with low GM

I:\CIRC\MSC\1\MSC.1-Circ.1652.docx
MSC.1/Circ.1652
Annex, page 216

1.2 Preparation of operational measures

1.2.1 An important question is in what phase of ship life cycle operational measures should
be provided, namely in design phase, in port before departure or directly en route:

.1 pre-computation in the design stage allows most comprehensive numerical

tools and statistical procedures, qualified staff, dedicated hardware and a
detailed check by the administration; a drawback is that the computations
can be performed only for assumed input parameters, most importantly,
standard seaway spectra;

.2 pre-computation before departure allows, in principle, using comprehensive

numerical tools and statistical procedures together with qualified staff and
dedicated hardware and, in addition, most accurate data about loading
condition and the most actual weather forecast available. In principle,
operational measures can be checked by the administration together with
the weather forecast, but this requires a corresponding infrastructure.
The drawback is a possibility of unforeseen delays in ship operation; and
.3 computations (on board or onshore) during operation allows using the most
actual weather and loading condition data, but check by the administration
is not possible. Besides, this approach requires significantly simplified
numerical tools and statistical procedures, so that the advantage of more
accurate weather data is to some degree compensated by the reduced
accuracy of numerical tools and statistical procedures.
1.2.2 With any option, accurate weather forecast and corresponding operational measures
should be ready in a sufficient time before a storm, e.g. three days, to allow for route change
if safe operation in the foreseen storm is not possible. Note that operational limitations related
to areas or routes and season do not require a weather forecast since they are prepared for a
specified wave scatter table, whereas operational guidance and operational limitations related
to maximum significant wave height require a weather forecast.

1.3 Preparation of operational guidance in design phase

1.3.1 A drawback of the pre-computation of operational guidance in the design stage is that
it relies on assumed theoretical wave energy spectra and thus, deviation of real sea states
from this assumption may lead to erroneous operational recommendations, especially in the
cross sea when wind sea and swell have significantly different directions.
1.3.2 One relevant consideration is that the influence of swell is usually noticeable in small
to moderate sea states and relatively small in strong storms, which are dominated by wind sea.
Figure 1.3.1 compares theoretical relationship between wind speed and wave height of wind
sea (solid line) in with hindcast data for two locations in North Atlantic ( ,), showing that
influence of swell (indicated by the difference in wave height between the theoretical
relationship and hindcast data at a given wind speed) is noticeable at small wave heights but
becomes relatively insignificant in more severe storms. Figure 1.3.2, plotting the significant
wave height of swell vs. the significant wave height of wind sea according to the same data,
confirms this.

I:\CIRC\MSC\1\MSC.1-Circ.1652.docx
 MSC.1/Circ.1652
Annex, page 217
vw,m/s

Hs,m
Hs,m Hs,m
Figure 1.3.1 Theoretical relationship between Figure 1.3.2 Significant wave height of swell
wind speed (y- axis) and wave height (x axis), (y- axis) plotted vs. significant wave height of
solid line, vs. hindcast for North Atlantic wind sea (x axis) for hindcast data shown in
(,) figure 1.3.1

1.3.3 To verify this consideration, worldwide hindcast data from the ERA33 Interim database
were used to estimate the likelihood of severe cross sea. From data for one year
(about 30 million entries), seaways were selected for which the angle between wind sea and
swell was more than 80 degrees: for about 0.01% of all data, the heights of both wind sea and
swell were more than 4 m; for about 0.001% of data more than 5 m, and for 0.0001% of data
more than 6 m, i.e. the likelihood of a cross sea where both wind sea and swell are severe is
negligible.

1.3.4 To check whether numerical simulations using theoretical wave energy spectra can
be applied to approximate roll responses to complex measured wave energy spectra, several
cross sea situations were selected from the ERA Interim database. For these situations, two
questions were investigated: first, what influence the overlapping effect of wind sea and swell
has, i.e. how much the combined response to the two separate (wind sea and swell) wave
energy systems differs from the response to the total spectrum and, second, how large is the
effect of the approximation of the real wave energy spectrum with a theoretical spectrum.

1.3.5 To answer these questions, numerical simulations of ship motions in irregular waves
were performed for the selected situations for the following modes: first, for the measured wave
energy spectrum, including wind sea and swell; second, for separate wave energy spectra of
wind waves and swell, derived from the measured wave energy spectrum (the responses, i.e.
the rate of stability failures, to these two separated spectra were summed); and third, for
approximated wave energy spectra, separately for wind waves and swell. JONSWAP wave
energy spectrum with the peak enhancement factor 3.3 and cos2 wave energy spreading with
respect to the mean wave direction was used for approximation (the responses, i.e. the rate of
stability failures, to the separate theoretical spectra were summed). In the definition of the
separated wave energy spectra of wind sea and swell from the measurements, the significant
wave height, mean period and mean direction of the wave energy spectrum, wind sea
spectrum and swell spectrum were kept unchanged. The ship course was varied
from 0 to 360 degrees every 10 degrees. Numerical simulations were performed for 200
realizations of each seaway until the first exceedance of 40 degrees roll angle.

1.3.6 This comparison was performed for all ships and loading conditions listed in
table 1.1.1, at six forward speeds equally distributed between zero and full speed in calm water.
Here, results are shown for two situations, table 1.3.1, for 1700 TEU container ship in loading
condition LC01 and 14000 TEU container ship in loading conditions LC01 and LC02.

33
ERA is the abbreviation for ECMWF Reanalysis, where ECMWF is the abbreviation for European Centre for
Medium-Range Weather Forecasts.

I:\CIRC\MSC\1\MSC.1-Circ.1652.docx
MSC.1/Circ.1652
Annex, page 218

Table 1.3.1 Parameters of wave energy spectra for two situations

Situation A B
Significant wave height, m: total, wind sea, swell 9.8, 4.0, 8.9 8.7, 5.4, 6.8
Mean wave period, s: wind sea, swell 12.4, 10.9 9.0, 14.3
Mean wave propagation direction, deg: wind sea, swell, shift 60, 153, 93 213, 27, 186

1.3.7 The results in figure 1.3.3 (situation A) and figure 1.3.4 (situation B) show that the
separate simulations for wave energy spectra of wind sea and swell and summing the resulting
stability failure rates (compare the left and middle columns in figure 1.3.3 and 1.3.4) leads to
slightly non-conservative results, whereas modelling of wind sea and swell systems using a
theoretical spectrum leads to slightly conservative results (compare the middle and right
columns in figures 1.3.3 and 1.3.4). The total effect due to both separate treatment of wind sea
and swell and theoretical approximation of wave energy spectra for these wave systems is
slightly conservative in situation A and slightly non-conservative in situation B.

Fn Fn Fn

Fn
Fn Fn

Fn Fn Fn

Figure 1.3.3 Situation A: colour plot of stability failure rate, 1/s, vs. Fn number (radial
coordinate) and ship course (circumferential coordinate) for measured wave energy spectrum
(left), summed rate for separate wind waves and swell wave energy spectra (middle) and
summed rate for separate wind waves and swell wave energy spectra approximated with

I:\CIRC\MSC\1\MSC.1-Circ.1652.docx
 MSC.1/Circ.1652
Annex, page 219

Fn Fn Fn

Fn Fn Fn

Fn Fn Fn

Figure 1.3.4 Situation B: colour plot of stability failure rate, 1/s, vs. Fn number (radial
coordinate) and ship course (circumferential coordinate) for measured wave energy spectrum
(left), summed rate for separate wind waves and swell wave energy spectra (middle) and
summed rate for separate wind waves and swell wave energy spectra approximated with
JONSWAP spectrum with  = 3.3 and cos2 wave energy spreading (right) for 1700 TEU
container ship in LC01 (top) and 14000 TEU container ship in LC01 (middle) and LC02 (bottom)
JONSWAP spectrum with  = 3.3 and cos2 wave energy spreading (right) for 1700 TEU
container ship in LC01 (top) and 14000 TEU container ship in LC01 (middle) and LC02 (bottom)

1.3.8 In all considered cases, theoretical modelling of wave systems and overlapping their
effect by summing the failure rates corresponding to each of the systems leads to practically
acceptable recommendations for a ship's forward speed and course, thus production and
acceptance of operational guidance in design phase is considered an acceptable option.

I:\CIRC\MSC\1\MSC.1-Circ.1652.docx
MSC.1/Circ.1652
Annex, page 220

1.4 Probabilistic operational guidance

1.4.1 To prepare a reference database, numerical simulations of motions in waves were

conducted for each ship and each loading condition in table 1.1.1 at six forward speeds, equally
distributed from zero to full speed in calm water, for all sea states (significant wave heights Hs
and zero-crossing wave periods Tz) in the North Atlantic wave scatter table, IACS
Recommendation No.34 (Corr.1 Nov. 2001), and for wave directions  from 0 (following waves)
to 180 (head waves) deg. every 10 deg. For each sailing condition (combination of forward
speed and wave direction) and each sea state, numerical simulations were performed in
multiple independent realizations of the same sea state for 2 hours or until the first exceedance
event until 200 stability failures were encountered. Realizations of the same sea state were
generated by random variation of frequencies, directions and phases of wave components
composing the sea state.

1.4.2 Direct counting was used to define the time 𝑇𝑖 to each stability failure; the expected
time until stability failure was calculated by averaging over N=200 stability failures as

𝑇̂ = (1/𝑁) ∑𝑁
𝑖=1 𝑇𝑖 (1.4.1)

1.4.3 The maximum likelihood estimate 𝑟̂ of stability failure rate was calculated as

𝑟̂ = 1/𝑇̂ (1.4.2)

1.4.4 For such combinations of sailing condition and sea state where the total simulation
time of 3.4106 hours was not sufficient to encounter 200 stability failures, a statistical
extrapolation of stability failure rate over significant wave height from the sea states with
greater significant wave heights was used.

1.4.5 Operational measures identify unacceptable (those that should be avoided) sailing
conditions for each sea state in such a way that avoiding these conditions ensures the same
safety level as the safety level provided by design assessment standards. For operational
limitations, this requirement is straightforward since operational limitations represent design
assessment procedures with changed environmental conditions. For operational guidance, the
general principle is the same; however, to distinguish acceptable and unacceptable sailing
conditions, a "short-term" criterion is required that can be defined for each situation
(and compared with a corresponding "short-term" standard), whereas the safety level provided
by operational guidance should be calculated over all situations (and compared with a
"long-term" standard). For convenience, the "short-term" standard will be referred to as
threshold.

1.4.6 Therefore, the safety level provided by operational guidance was computed as a
function of a variable short-term acceptance threshold. Consistent with the discussion above,
the safety level provided by operational guidance was defined as the average stability failure
rate over all acceptable sailing conditions in all sea states (to investigate the dependency of
results on the forward speed, different forward speeds were addressed first treated separately).
The appropriate value of the short-term acceptance threshold was found from the requirement
that this safety level is equal to the acceptance standard in the design assessment
requirements.

∑s,𝜇 𝑟(𝐻s , 𝑇z , 𝜇, 𝑣0 )𝑓OG (𝐻s , 𝑇z , 𝜇)Δ𝐻s Δ𝑇z Δ𝜇

𝑤OG = (1.4.3)
∑s,𝜇 𝑓OG (𝐻s , 𝑇z , 𝜇)Δ𝐻s Δ𝑇z Δ𝜇

1.4.7 In eq. (1.4.3), wOG, 1/s, is the average stability failure rate conditional on operational
guidance, fOG(Hs,Tz,) is the conditional probability density of sea state with significant wave

I:\CIRC\MSC\1\MSC.1-Circ.1652.docx
 MSC.1/Circ.1652
Annex, page 221

height Hs, zero-crossing period Tz and mean direction , set to zero if a combination (Hs,Tz,)
is unacceptable and equal to fs(Hs, Tz)∙f() otherwise, and r(Hs, Tz, , v0) is the average
"short-term" stability failure rate, 1/s, in a specific sailing condition in a specific sea state.
For research purposes, the total stability failure rate as well as contributions from various
stability failure modes were calculated.

1.4.8 Two probabilistic "short-term" criteria were tested as candidates: the average stability
failure rate r and the product rfs. Figures 1.4.1(a-e) show the average stability failure rate wOG
(total and contributions from stability failure modes) vs. systematically varied short-term
threshold of rfs, and figures 1.4.2 (a-e) show corresponding dependencies for the
systematically varied short-term threshold of r.

Figure 1.4.1.a. Average rate of stability failures wOG, 1/s, for all failure modes (top left),
parametric roll in bow (top right) and stern (bottom left) waves and synchronous roll, relevant
for dead ship condition and excessive acceleration stability failures (bottom right) depending
on rfs-threshold, 1/(m·s2) for cruise ship; types and colours of lines differentiate loading
conditions, lines of the same type and colour correspond to various forward speeds for same
loading condition

I:\CIRC\MSC\1\MSC.1-Circ.1652.docx
MSC.1/Circ.1652
Annex, page 222

Figure 1.4.1.b. Same as Figure 1.4.1.a for 14000 TEU container ship

Figure 1.4.1.c. Same as Figure 1.4.1.a for 1700 TEU container ship

I:\CIRC\MSC\1\MSC.1-Circ.1652.docx
 MSC.1/Circ.1652
Annex, page 223

Figure 1.4.1.d. Same as Figure 1.4.1.a for 8400 TEU container ship

Figure 1.4.1.e. Same as figure 1.4.1.a for RoPax ship

I:\CIRC\MSC\1\MSC.1-Circ.1652.docx
MSC.1/Circ.1652
Annex, page 224

Figure 1.4.2.a. Average rate of stability failures wOG, 1/s, for all failure modes (top left),
parametric roll in bow (top right) and stern (bottom left) waves and synchronous roll,
relevant for dead ship condition and excessive acceleration failure modes (bottom
right) depending on r-threshold for cruise ship; types and colours of lines differentiate
loading conditions, lines of the same type and colour correspond to various forward
speeds for same loading condition

Figure 1.4.2.b. Same as in figure 1.4.2.a for 14000 TEU container ship

I:\CIRC\MSC\1\MSC.1-Circ.1652.docx
 MSC.1/Circ.1652
Annex, page 225

Figure 1.4.2.c. Same as in figure 1.4.2.a for 1700 TEU container ship

Figure 1.4.2.d. Same as in figure 1.4.2.a for 8400 TEU container ship

I:\CIRC\MSC\1\MSC.1-Circ.1652.docx
MSC.1/Circ.1652
Annex, page 226
wOG, 1/s

Figure 1.4.2.e. Same as in figure 1.4.2.a for RoPax ship

1.4.9 These results prove rfs as a suitable criterion to be used for operational guidance since
it leads to close dependencies of the long-term safety level wOG on the rfs-threshold for all ships,
loading conditions and forward speeds until saturation (when further relaxation of the threshold
does not change safety level anymore: at rfs of about 10-5 1/(m·s2), the long-term safety level
becomes saturated for all considered ships, loading conditions and forward speeds).

1.4.10 Using directly the stability failure rate r as a "short-term" criterion leads to some
spreading of the safety level (at the same value of the short-term threshold) between forward
speeds and loading conditions for the same ship and between ships, which means that using
r as the "short-term" criterion for operational guidance will lead to spreading of the safety level
provided by operational guidance between different ships and loading conditions.

1.4.11 An adequate "short-term" criterion should provide similar safety level for all ships and
all loading and sailing conditions, i.e. not allow unsafe sailing conditions while not imposing
unnecessary restrictions on safe sailing conditions. To check whether the proposed criteria
satisfy these requirements, figure 1.4.3 (left) shows the results as a histogram of the total
number of ships, loading conditions and forward speeds (normalized to 1) plotted against the
resulting safety level wOG for rfs-criterion, and figure 1.4.4 shows the corresponding results for
the r-criterion. Using the rfs-criterion effectively removes cases with insufficient safety level,
whereas cases that were safe enough without operational guidance are not influenced. As a
result, all cases influenced by operational guidance achieve very close safety level. Using r as
a criterion for operational guidance provides a similar, while slightly poorer, quality to using rfs.

I:\CIRC\MSC\1\MSC.1-Circ.1652.docx
 MSC.1/Circ.1652
Annex, page 227

rfS

Figure 1.4.3 Number of

ships, loading
conditions and forward
speeds normed on 1
having long-term
safety level wOG, 1/s (x-
axis) for various rfs-
threshold values
(indicated in plot)

wOG

r= e-13

Figure 1.4.4 Number of

ships, loading
conditions and forward
speeds normed on 1
having long-term safety
level wOG, 1/s (x- axis)
for various r-threshold
values (indicated in
plot)

wOG

1.5 Deterministic operational guidance

1.5.1 Here, it was investigated whether operational guidance based on a non-probabilistic

criterion is possible. Such operational guidance is simpler in production and acceptance than
a probabilistic one. A drawback is, however, that deterministic operational guidance does not
ensure consistent safety level across various ships, loading conditions and sailing conditions,
thus it is difficult to ensure consistency with direct stability assessment.

1.5.2 A large inaccuracy of a deterministic operational guidance must be compensated by

its excessive conservativeness (to keep a suitable safety level). Note, however, that an
excessive conservativeness of operational guidance is a smaller problem than excessive
conservativeness of direct stability assessment. Usually, operational practices are based on
more conservative requirements than design assumptions anyway.

1.5.3 The approach is based on the same idea as in the probabilistic operational guidance,
eq. (1.4.3), but instead of a probabilistic criterion (stability failure rate r or product rfs above), a
non-probabilistic criterion is used to differentiate between safe and unsafe sailing conditions.
Studies on the development of deterministic direct stability assessment showed that among
the compared non-probabilistic criteria, namely standard deviation of roll angle, average roll
amplitude, significant roll amplitude and three-hour maximum roll amplitude, the latter provides
the best results in direct stability assessment compared to the others; therefore, it was also
used as a criterion in the deterministic operational guidance.

1.5.4 To compute the expected maximum three-hour roll amplitude, numerical simulations
were carried out in 50 realizations of the same sea state; the realizations were generated by
random variation of frequencies, directions and phases of harmonic components modelling
seaway. A difficulty in the application of deterministic criteria is occurrence of capsizing in some

I:\CIRC\MSC\1\MSC.1-Circ.1652.docx
MSC.1/Circ.1652
Annex, page 228

realizations. In such cases, the maximum three-hour roll amplitude cannot be defined and thus,
the mean three-hour maximum roll amplitude cannot be calculated. To indicate such cases in
plots, mean -hour maximum roll amplitude is shown as 60 degrees in plots (since in situations
where capsizing did not happen, mean 3-hour maximum roll amplitude never achieved 60
degrees).

1.5.5 Figures 1.5.1 (a-e) show the average "long-term" stability failure rate wOG (total and
due to various stability failure modes) vs. the systematically varied threshold of the
mean three-hour maximum roll amplitude. The results indicate significant scatter of the
dependencies of wOG on the deterministic threshold between ships, loading conditions and
forward speeds; saturation happens at about 30 degrees of mean three-hour maximum roll
amplitude for all considered ships, loading conditions and forward speeds.

Figure 1.5.1.a. Average stability failure rate wOG, 1/s, for all failure modes (top left), parametric
roll in bow (top right) and stern (bottom left) waves and synchronous roll, relevant for dead
ship condition and excessive acceleration failure modes, (bottom right) vs. threshold of mean
three-hour maximum roll amplitude for cruise ship; line types and colours differentiate loading
conditions, lines of the same type and colour correspond to various forward speeds

I:\CIRC\MSC\1\MSC.1-Circ.1652.docx
 MSC.1/Circ.1652
Annex, page 229

Figure 1.5.1.b. Same as in figure 1.5.1.a for 14000 TEU container ship

Figure 1.5.1.c. Same as in figure 1.5.1.a for 1700 TEU container ship

I:\CIRC\MSC\1\MSC.1-Circ.1652.docx
MSC.1/Circ.1652
Annex, page 230

Figure 1.5.1.d. Same as in figure 1.5.1.a for 8400 TEU container ship

Figure 1.5.1.e. Same as in figure 1.5.1.a for RoPax ship

1.5.6 To check how deterministic operational guidance influences safety level of different
ships and loading conditions at different forward speeds, figure 1.5.2 (left) shows histogram of
the total number of ships, loading conditions and forward speeds normalized to 1, plotted
against the achieved safety level wOG for various values of the threshold for the mean
three-hour maximum roll. Note that the results for the threshold values of 40 and 60 degrees
are very similar (cases with 60 degrees maximum roll amplitude mean here such cases where
at least one capsize happened in 50 simulations of three hours duration each).

I:\CIRC\MSC\1\MSC.1-Circ.1652.docx
 MSC.1/Circ.1652
Annex, page 231

Figure 1.5.2 Total number

of ships, loading
conditions and forward
speeds (normalized to 1)
with long-term safety level
wOG, 1/s (x- axis) for
various values (indicated
in plot) of threshold of
mean 3-hour maximum
roll amplitude

wOG

1.5.7 The deterministic approach does not fully exclude cases with insufficient safety level:
in fact, strengthening the threshold from 60 to 25 degrees little influences the mean "long-term"
rate of stability failures at and below 10 -7 1/s. The safety level of all cases influenced by
operational guidance is broadly spread.

1.6 Definition of thresholds

1.6.1 To differentiate between acceptable and unacceptable sailing conditions

(combinations of forward speed and course) in each sea state, the "short-term" acceptance
thresholds for rfs, r and 3h are defined from the "long-term" standard (safety level) using the
dependencies of the average "long-term" stability failure rate over all acceptable sailing
conditions in all sea states on the short-term acceptance threshold described above.

1.6.2 Appendix 4 defines wOG=2.610-8 1/s as the required "long-term" standard (safety
level). To derive the short-term thresholds, figures 1.6.1 (a-c) and table 1.6.1 show the
"long-term" average stability failure rate wOG (averaged over all speeds) depending on rfs- (left),
r- (middle) and 3h- (right) thresholds; the results indicate 10-10 1/(m·s2) and 10-6 1/s as
appropriate thresholds for rfs and r, respectively.

I:\CIRC\MSC\1\MSC.1-Circ.1652.docx
MSC.1/Circ.1652
Annex, page 232

Figure 1.6.1.a. Average "long-term" stability failure rate wOG, 1/s, averaged over all speeds, vs.
"short-term" rfs-threshold, 1/(m·s2); each plot corresponds to one ship, each line corresponds
to one loading condition

I:\CIRC\MSC\1\MSC.1-Circ.1652.docx
 MSC.1/Circ.1652
Annex, page 233

Figure 1.6.1.b. Average "long-term" stability failure rate wOG, 1/s, averaged over all speeds, vs.
"short-term" r-threshold, 1/s; each plot corresponds to one ship, each line corresponds to one
loading condition

I:\CIRC\MSC\1\MSC.1-Circ.1652.docx
MSC.1/Circ.1652
Annex, page 234

Figure 1.6.1.c. Average "long-term" stability failure rate wOG, 1/s, averaged over all speeds, vs.
"short-term" 3h-threshold, degree; each plot corresponds to one ship, each line corresponds
to one loading condition

Table 1.6.1 Definition of short-term threshold for operational guidance

Ship Loading condition rfs, 1/(m·s2) r, 1/s 3h,o
Cruise vessel 01 1.310-10 4.710-6 -
02 4.210-10 9.710-5 -
1700 TEU container ship 01 6.510-11 1.710-6 20.7
02 1.110-10 5.910-6 29.7
8400 TEU container ship 01 1.210-10 5.210-6 20.6
02 1.810-10 6.110-6 23.6
03 5.410-10 1.610-5 34.8
04 2.710-10 2.510-5 33.6
CV-14000 container ship 01 6.310-11 1.710-6 7.7
02 7.110-11 2.110-6 17.9
03 1.710-10 5.410-6 -
Selection 10-10 10-6 s. text

I:\CIRC\MSC\1\MSC.1-Circ.1652.docx
 MSC.1/Circ.1652
Annex, page 235

1.6.3 To illustrate the difficulty of the definition of the required "short-term" threshold for the
mean 3-hour maximum roll amplitude 3h, figure 1.6.2 shows some typical dependencies of
the exceedance rate of certain roll amplitude on this amplitude. Such dependencies strongly
depend on the form of the righting lever curve, stability failure mode and wave height, period
and direction. These results indicate that the dependencies of the failure rate on roll amplitude
can be unpredictable, therefore, some sort of extrapolation of failure rate over roll amplitude is
in general impossible.
Rate, 1/s

Roll Angle, deg Roll Angle, deg

Rate, 1/s

Roll Angle, deg Roll Angle, deg

Figure 1.6.2. Exceedance rate (1/s, y- axis) of roll amplitudes shown along x- axis (degree)
together with righting-lever curves in calm water for (top left to bottom right) cruise vessel in
loading condition LC01 (parametric roll in head and following waves and synchronous roll in
beam waves), 1700 TEU container ship in LC01 (parametric roll in head and following waves)
and LC02 (synchronous roll in beam waves) and 8400 TEU container ship in LC01 (parametric
roll in head and following waves)

1.6.4 One pragmatic alternative is to find a simple empirical formula for the 3h-threshold
based on its relation with the safety level. Figure 1.6.1 (right) shows the "long-term" stability
failure rate wOG, averaged over all forward speeds, vs. short-term 3h-threshold for sample ships
and loading conditions, and table 1.6.1 shows the resulting threshold values. Figure 1.6.3,
comparing these values with the calm-water capsize heel angle, shows that the threshold can
be approximated as half of the calm-water capsize heel angle (generally, as half of the heel
angle defining stability failure). Although this definition appears not conservative in some cases,
note that the results of probabilistic assessment used to define thresholds are conservative
due to conservative extrapolation of stability failure rate over wave height. This does not matter
for the definition of r- and rfs-thresholds but influences the definition of 3h-threshold.

I:\CIRC\MSC\1\MSC.1-Circ.1652.docx
MSC.1/Circ.1652
Annex, page 236

Figure 1.6.3 3h-threshold vs.

calm-water capsize heel angle

1.6.5 Another way is to use a relation following from the Rayleigh distribution of roll
amplitudes, i.e. 3h sf, where sf defines stability failure, ={ln(T/Tr) / ln(T3h/Tr)}0.5 (but not
less than 1), T = fs/10-10, Tr is the natural roll period and T3h means three hours in seconds.
Figure 1.6.4 shows that this approximation is suitable for synchronous roll in beam waves and
conservative for parametric roll.

I:\CIRC\MSC\1\MSC.1-Circ.1652.docx
 MSC.1/Circ.1652
Annex, page 237

GM = 0.88m

GM = 1.95m

GM = 7.5m

GM = 3.797m

GM = 2.6m

GM = 0.67m

GM = 3.8m

GM = 2.68m

Figure 1.6.4 𝐥𝐧( − 𝐥𝐧 𝒓) vs. ln{3h –  )/} for parametric roll in stern-quartering (,) and bow
(,) waves and synchronous roll (,); blue dashed lines show Rayleigh distribution

I:\CIRC\MSC\1\MSC.1-Circ.1652.docx
MSC.1/Circ.1652
Annex, page 238

1.6.6 Figures 1.6.5 (a-c) show examples of mean 3-hour maximum roll amplitude, its double
value and maximum 15-hour roll amplitude, defined from 15-hour simulations for several
parametric and synchronous roll situations, vs. significant wave height; figures 1.6.6 (a-c) show
corresponding results using factor . The results indicate that doubling 3h produces slightly
more conservative results than using factor , and both provide the limiting significant wave
height 1 m to 2 m less than that leading to capsizing in three hours.

max.over all max.over all

Roll Amplitude, deg.

Roll Amplitude, deg.

2max.3h-av.

2max.3h-av.
EA
EA
cw max.not capsized
max.3h-av. max.not capsized
cw
max.3h-av.

Significant Wave Height, m max.over all

Significant Wave Height, m 2max.3h-av.

2max.3h-av. max.over all

Roll Amplitude, deg.

Roll Amplitude, deg.

max.not capsized cw max.not capsized

EA
EA
max.3h-av.
cw

max.3h-av.

Significant Wave Height, m Significant Wave Height, m

Figure 1.6.5.a. 3h defined excluding capsizing events (⚫), its double value (solid line),
maximum 15-hour roll amplitude taking () and not taking () into account capsizes and calm-
water capsize heel angle (horizontal dashed line) vs. significant wave height for parametric
roll in following waves for several ships and loading conditions
max.over all max.over all
Roll Amplitude, deg.
Roll Amplitude, deg.

2max.3h-av.

2max.3h-av.

cw max.not capsized

max.3h-av. max.not capsized

cw

max.3h-av.

Significant Wave Height, m max.over all

Significant Wave Height, m
2max.3h-av.
2max.3h-av.
max.over all
Roll Amplitude, deg.

Roll Amplitude, deg.

EA max.not capsized
cw max.not capsized
max.3h-av. EA
cw

max.3h-av.

180º

Significant Wave Height, m Significant Wave Height, m

Figure 1.6. 5. b. Same as in figure 1.6.5.a for parametric roll in head waves

I:\CIRC\MSC\1\MSC.1-Circ.1652.docx
 MSC.1/Circ.1652
Annex, page 239

max.over all max.over all

Roll Amplitude, deg.

2max.3h-av.

2max.3h-av.

EA EA
cw max.not capsized
max.not capsized
max.3h-av. cw

max.3h-av.

Significant Wave Height, m Significant Wave Height, m

2max.3h-av.
2max.3h-av.

max.over all
max.over all
Roll Amplitude, deg.

Roll Amplitude, deg.

cw max.not capsized
max.not capsized
cw EA
max.3h-av. max.3h-av.
EA

Significant Wave Height, m Significant Wave Height, m

Figure 1.6.5.c. Same as in figure 1.6.5.a for synchronous roll in beam waves (relevant for dead
ship condition and excessive acceleration failure modes)

max.over all max.over all

Roll Amplitude, deg.
Roll Amplitude, deg.

max.3h-av.

max.not capsized
cw max.3h-av.
max.not capsized
max.3h-av. cw

max.3h-av.

Significant Wave Height, m Significant Wave Height, m

max.over all
max.over all
Roll Amplitude, deg.

Roll Amplitude, deg.

max.3h-av.

max.3h-av.

max.not capsized cw

cw max.not capsized

max.3h-av.
max.3h-av.

Significant Wave Height, m Significant Wave Height, m

Figure 1.6.6.a. Mean three-hour maximum roll amplitude (⚫ ), its value multiplied with factor 𝜶
(solid line), maximum roll amplitude taking () and not taking () into account capsizes and
calm-water capsize heel angle (horizontal dashed line) vs. significant wave height for
parametric roll in following waves for several ships and loading conditions
mplitude, deg.

I:\CIRC\MSC\1\MSC.1-Circ.1652.docx
MSC.1/Circ.1652
Annex, page 240

max.over all max.over all

Roll Amplitude, deg.

max.3h-av.
EA
max.3h-av.
cw
max.not capsized
max.not capsized cw
max.3h-av.

max.3h-av.

Significant Wave Height, m Significant Wave Height, m

max.over all
max.over all

Roll Amplitude, deg.

max.3h-av.
max.3h-av.
max.not capsized max.not capsized
cw

cw

max.3h-av.
max.3h-av.

Significant Wave Height, m Significant Wave Height, m

Figure 1.6.6.b. Same as in figure 1.6.6.a for parametric roll in head waves

max.over all max.over all

Roll Amplitude, deg.

Roll Amplitude, deg.

max.3h-av.
max.3h-av.
cw
max.not capsized
cw
max.not capsized
max.3h-av.
max.3h-av.

Significant Wave Height, m Significant Wave Height, m

max.over all
max.over all
Roll Amplitude, deg.

Roll Amplitude, deg.

max.3h-av.
max.not capsized
max.3h-av. cw
max.not capsized

cw

max.3h-av.

max.3h-av.

Significant Wave Height, m Significant Wave Height, m

Figure 1.6.6.c. Same as in figure 1.6.6.a for synchronous roll in beam waves (relevant for dead
ship condition and excessive acceleration failure modes)

1.6.7 Table 1.6.2 shows conservative and non-conservative errors, defined as percentage
of the number of situations with conservative or non-conservative errors from the total number
of situations, of deterministic operational guidance based on 23h-criterion vs. probabilistic
operational guidance based on r-criterion.

I:\CIRC\MSC\1\MSC.1-Circ.1652.docx
 MSC.1/Circ.1652
Annex, page 241

Table 1.6.2 Percentage of errors of deterministic operational guidance

Ship Cruise CV 1700 TEU CV 8400 TEU
LC LC01 LC01 LC02 LC01 LC03
Non-conservative 2.4 1.6 2.5 3.5 2.1
Conservative 4.1 9.9 5.6 1.6 0.0

1.7 Simplified operational guidance

1.7.1 Operational guidance provides detailed recommendations regarding ship's forward

speed and course and therefore, requires methods of the level corresponding to direct stability
assessment. However, sometimes simpler conservative recommendations for the forward
speed and course, provided by simpler means, such as level 1 or level 2 criteria, are sufficient
for the shipowner and acceptable for the administration:

.1 for pure loss and surf-riding/broaching stability failure modes, operational

limitation of the maximum acceptable significant wave height, defined with
level 2 vulnerability assessment can be combined with the forward speed
limit according to level 1 vulnerability criterion in following and
stern-quartering seaways at greater significant wave heights; and

.2 for excessive accelerations, where the level 2 vulnerability assessment is

performed at zero forward speed, forward speed effect can be added to
level 2-based operational limitations in a conservative way.

1.7.2 Check 2 of level 2 parametric roll criterion provides dependency of roll motion on the
forward speed; hence it is useful to check whether this dependency is sufficiently accurate for
a simplified operational guidance. Here, the sensitivity of this criterion to changes in forward
speed is compared with direct stability assessment.

1.7.3 According to an earlier version of the present criterion, a loading condition was
considered not vulnerable to parametric roll stability failure mode if 𝐶2 = (1/7){∑3𝑗=1 𝐶2𝐻 (Fnj ) +
𝐶2𝐻 (0) + ∑3𝑗=1 𝐶2𝐹 (Fn𝑗 )} < 0.06 , where 𝐶2𝐻 (Fn𝑗 ) and 𝐶2𝐹 (Fn𝑗 ) refer to sailing in head and
following waves, respectively, at a Froude number Fnj and are calculated for each of Fnj as a
sum over all N sea states of a scatter table as 𝐶2𝐻,𝐹 = ∑𝑁 𝑖=1 𝑤𝑖 𝑐𝑖 ; wi is the normed probability of
a sea state i, and ci=1 when roll amplitude exceeds 25 o and 0 otherwise.

1.7.4 To verify whether criteria 𝐶2𝐻,𝐹 can be used for forward speed recommendations, their
dependency on the forward speed for all sample ships in all loading conditions was compared
with the dependency on forward speed of the stability failure rate due to parametric roll
𝐻
obtained from direct stability assessment, separately in head (denoted as 𝑤𝑃𝑅 ) and following
𝐹
(𝑤𝑃𝑅 ) waves. In the numerical simulations, roll damping was defined from roll decay model
tests (for the three container ships and the cruise ship) and with the simplified Ikeda method
(for the RoPax ship).

1.7.5 For comparison, 40, 25 and 15 degrees heel angles were used to define stability
𝐻
failure. Figures 1.7.1 (a-c) show 𝑤𝑃𝑅 (left y axis) and 𝐶2𝐻 (right y axis) vs. Froude number
(x- axis) for 15 (left), 25 (middle) and 40 (right) degrees definitions for different loading
conditions (differentiated with lines of the same type: those with symbols refer to direct
𝐻
assessment result 𝑤𝑃𝑅 and those without symbols to check 2 of level 2 result 𝐶2𝐻 ) of sample
ships (each ship corresponds to one row); figures 1.7.2 (a-c) show corresponding results for
parametric roll in following waves.

I:\CIRC\MSC\1\MSC.1-Circ.1652.docx
MSC.1/Circ.1652
Annex, page 242

Fn Fn

Fn Fn

Fn

Figure 1.7.1.a. Parametric roll in head waves: direct assessment result 𝒘𝐇 𝐏𝐑 (left y- axis) and
check 2 of level 2 result 𝑪𝐇 𝟐 (right axis) vs. Fn (x- axis) for 15 degrees definition of stability
failure for all loading conditions; each line corresponds to one loading condition; black lines
with symbols refer to 𝒘𝐇 𝐏𝐑 , same type blue lines without symbols to 𝑪𝟐
𝐇

I:\CIRC\MSC\1\MSC.1-Circ.1652.docx
 MSC.1/Circ.1652
Annex, page 243

Fn Fn

Fn Fn

Fn

Figure 1.7.1.b. Same as in figure 1.7.1.a for 25 degrees definition of stability failure

I:\CIRC\MSC\1\MSC.1-Circ.1652.docx
MSC.1/Circ.1652
Annex, page 244

Fn Fn

Fn Fn

Fn

Figure 1.7.1.c. Same as in figure 1.7.1.a for 40 degrees definition of stability failure

I:\CIRC\MSC\1\MSC.1-Circ.1652.docx
 MSC.1/Circ.1652
Annex, page 245

Fn Fn

Fn Fn

Fn

Figure 1.7.2.a. Parametric roll in following waves: results of direct assessment 𝒘𝐅𝐏𝐑 (left y- axis)
and check 2 of level 2 𝑪𝐅𝟐 (right y-axis) vs. Froude number (x- axis) for 15 degrees definition of
stability failure for all loading conditions; each line corresponds to one loading condition:
black lines with symbols refer to 𝒘𝐅𝐏𝐑 , same type blue lines without symbols to 𝑪𝐅𝟐

I:\CIRC\MSC\1\MSC.1-Circ.1652.docx
MSC.1/Circ.1652
Annex, page 246

Fn Fn

Fn Fn

Fn

Figure 1.7.2.b. Same as in figure 1.7.2.a. for 25 degrees definition of stability failure

I:\CIRC\MSC\1\MSC.1-Circ.1652.docx
 MSC.1/Circ.1652
Annex, page 247

Fn Fn

Fn Fn

Fn

Figure 1.7.2.c. Same as in figure 1.7.2.a. for 40 degrees definition of stability failure

1.7.6 The results show that check 2 of level 2 parametric roll criterion produces in general
good results at low GM. However, with increasing GM, the agreement worsens: this criterion
indicates that large roll amplitudes move to higher forward speed or disappear, so that
parametric roll becomes not dangerous any more at low forward speeds, whereas direct
simulations indicate danger of parametric roll at low forward speeds (with the exception of
RoPax, for which failure rate due to parametric roll is always very small). The agreement
between check 2 of level 2 and direct simulation improves for 40 o heel angle as failure definition
instead of 25o and worsens for 15o.

1.7.7 To check the reason for this difference, figures 1.7.3 and 1.7.4 show failure rate due
to parametric roll in head and following waves together with roll amplitude according to check 2
of level 2 depending on Froude number for 8400 TEU container ship, for which the differences
between check 2 of level 2 and direct assessment in figures 1.7.1 and 1.7.2 are greatest, in
three loading conditions with the smallest GM values at three significant wave heights and
various mean wave periods. The figures show that the dependency of roll motion on forward
speed differs between check 2 of level 2 and direct simulations.

I:\CIRC\MSC\1\MSC.1-Circ.1652.docx
MSC.1/Circ.1652
Annex, page 248

Hs = 4m Hs = 8m Hs = 12m

Fn Fn Fn
Hs = 4m Hs = 8m Hs = 12m

Fn Fn Fn
Hs = 4m Hs = 8m Hs = 12m

Fn Fn Fn
Figure 1.7.3. Failure rate due to parametric roll in head waves (left y- axis, black lines with symbols) and roll amplitude from check 2 of
level 2 PR criterion (right y axis, blue lines without symbols) vs. Froude number (x- axis) for 8400 TEU container ship in three loading
conditions (rows) at significant wave heights (columns) 4, 8 and 12 m; different lines correspond to different wave periods

I:\CIRC\MSC\1\MSC.1-Circ.1652.docx
 MSC.1/Circ.1652
Annex, page 249

Hs = 4m Hs = 8m Hs = 12m

Fn Fn Fn
Hs = 8m Hs = 12m
Hs = 4m

Fn Fn Fn
Hs = 4m Hs = 8m Hs = 12m

Fn Fn Fn
Figure 1.7.4 Failure rate due to parametric roll in following waves (left y- axis, black lines with symbols) and roll amplitude from check 2
of level 2 PR criterion (right y axis, blue lines without symbols) vs. Froude number (x- axis) for 8400 TEU container ship in three loading
conditions (rows) at significant wave heights (columns) 4, 8 and 12 m; different lines correspond to different wave periods

I:\CIRC\MSC\1\MSC.1-Circ.1652.docx
MSC.1/Circ.1652
Annex, page 250

1.7.8 These results mean that, first, using check 2 of level 2 parametric roll criterion to
provide forward speed recommendations requires further validation and eventually
improvement of this criterion and, second, that direct stability assessment for parametric roll in
head waves can be conducted at zero (or as small as practicable) forward speed.

1.7.9 Model test results for the 8400 TEU container ship in figure 1.7.5 confirm that in
irregular waves, low forward speeds are more critical for parametric roll in head waves than
higher forward speeds, even when resonance condition suggests that higher forward speed
should be more critical (compare with figure 1.7.6 which concerns parametric resonance in
regular head waves for the same ship).

Figure 1.7.5 Measured () and computed () RMS of roll angle (y- axis) in irregular head waves
vs. Froude number (x- axis) and wave period

I:\CIRC\MSC\1\MSC.1-Circ.1652.docx
 MSC.1/Circ.1652
Annex, page 251

Figure 1.7.6 Measured (,) and computed (⚫) roll amplitude (y- axis) in regular head waves
vs. Froude number (x- axis) and wave period

1.8 When operational measures are not suitable

1.8.1 Operational measures can reduce the average stability failure rate to any specified
level; thus, any loading condition of any ship can be made "sufficiently safe" by application of
sufficiently strict operational measures. However, if too many sailing conditions in too many
sea states, especially in moderate sea states, should be excluded as unacceptable for some
loading condition, it cannot be considered as sufficiently safe in routine practical operation.
Therefore, if the total amount of acceptable sailing conditions becomes too small for some
loading condition, it should not be considered as acceptable even when operational measures
are provided. It follows from these considerations that a suitable criterion is desirable to
distinguish such loading conditions for which operational measures are a possible solution
from those which cannot be accepted even with operational measures. For this purpose, the
total duration of acceptable sailing conditions in all relevant sea states, according to
operational measures, as a percentage of the total operational life at sea is identified. Such a
percentage is frequently referred to as operability, O, defined as

O= 
acceptable
f s ( H s , Tz ) f  (  ) f v0 ( v0 ) H s Tz v0
(1.8.1)
H s ,Tz ,  ,v0

I:\CIRC\MSC\1\MSC.1-Circ.1652.docx
MSC.1/Circ.1652
Annex, page 252

where  is the wave heading, v0 is the ship speed, Hs is the significant wave height, Tz is the
mean zero-crossing period, and fs (( Hs, Tz ) f () fv0 (v0)) is the joint probability density function
of the sailing condition (, v0 ) and the sea state ( Hs, Tz ).

1.8.2 Similarly to other criteria, the threshold for operability can be defined from case
studies. The operability O, corresponding to the "short-term" threshold of stability failure rate
r = 1.0∙10-6 1/s for the five studied ships in all loading conditions is presented in table 1.8.1
(presented separately for different forward speeds, noting that RoRo vessel is not shown since
it has an operability=1.0 at all speeds). The results in table 1.8.1 suggest that a minimum
operability of 0.8 is an appropriate threshold because only one ship fails this requirement in
two loading conditions (bold values): one loading condition (LC02) fails this requirement at
zero forward speed and the other (LC01) also fails the weather criterion of the 2008 Intact
Stability Code. The average operability over all forward speeds in table 1.8.2 exceeds 0.8 for
all studied ships in all loading conditions. Moreover, the maximum significant wave height of
about 4.4 m, which corresponds to an operability=0.8 in the North Atlantic, figure 1.8.1,
confirms that this value is not too conservative. Therefore, the value 0.8 can be accepted as
an appropriate operability standard to eliminate loading conditions for which operational
guidance is not a suitable alternative.

Table 1.8.1 Operability (separately over forward speeds) corresponding to a short-term

threshold of the stability failure rate r=1.0∙10-6 1/s for all studied ships and loading
conditions
ship Fr LC01 LC02 LC03 LC04 LC05 ship Fr LC01 LC02 LC03 LC04 LC05
0.00 0.836 0.958 1.00 1.000 1.000 0.00 0.856 0.760 0.915 1.000 1.000
14000 TEU
1700 TEU

0.05 0.852 0.983 1.000 1.000 1.000 0.04 0.861 0.811 0.957 1.000 1.000
0.10 0.831 0.995 1.000 1.000 1.000 0.09 0.833 0.916 0.988 1.000 1.000
CV

CV

0.14 0.855 0.996 0.999 1.000 1.000 0.13 0.794 0.985 0.998 1.000 1.000
0.19 0.930 0.978 0.998 1.000 1.000 0.17 0.820 0.991 0.997 1.000 1.000
0.24 0.905 0.976 1.000 1.000 1.000 0.21 0.872 0.981 0.999 1.000 1.000
0.00 0.943 0.922 0.967 0.993 0.999 0.00 0.937 0.997 1.000 1.000 1.000
8400 TEU

0.05 0.946 0.946 0.991 0.993 0.999 0.05 0.970 0.998 1.000 1.000 1.000
Vessel
Cruise

0.09 0.956 0.991 0.999 0.996 1.000 0.09 0.994 0.999 1.000 1.000 1.000
CV

0.14 0.993 1.000 1.000 0.997 1.000 0.14 0.997 0.999 0.999 1.000 1.000
0.18 0.999 1.000 1.000 0.998 1.000 0.18 0.987 0.997 0.999 1.000 1.000
0.23 0.996 0.999 1.000 0.999 1.000

Table 1.8.2 Operability (average over all forward speeds) corresponding to a short-
term threshold of the stability failure rate r=1.0∙10-6 1/s for all studied ships and loading
conditions
Loading condition
Ship
LC01 LC02 LC03 LC04 LC05
Cruise Vessel 0.977 0.998 1.000 1.000 1.000
1700 TEU Container Ship 0.868 0.981 0.999 1.000 1.000
8400 TEU Container Ship 0.972 0.976 0.993 0.996 1.000
14000 TEU Container Ship 0.839 0.907 0.975 1.000 1.000

I:\CIRC\MSC\1\MSC.1-Circ.1652.docx
 MSC.1/Circ.1652
Annex, page 253

Figure 1.8.1. Maximum

significant wave height
vs. operability in North
Atlantic wave climate

1.9 Influence of propulsion, steering and seakeeping

1.9.1 So far, propulsion and steering abilities of a ship, as well as seakeeping problems
such as excessive vertical motions and accelerations and excessive loads at high forward
speed in bow waves, have not been considered in design assessment and operational
measures concerning dynamic stability. For some stability failure modes, this may lead to
non-conservative errors in design assessment or misleading operational recommendations:

.1 for pure loss of stability and surf-riding/broaching stability failures, which are
relevant in stern waves, consideration of propulsion and steering abilities
and seakeeping problems is not critical for dynamic stability;

.2 for dead ship condition stability failure mode, relevant only at zero forward
speed in beam seaway, such problems are also not critical;

.3 for excessive acceleration stability failures, forward speed in beam seaway

rather moderately influences roll motion (due to decreasing roll damping with
decreasing forward speed); this does not influence the design assessment
(which is performed at zero forward speed) but has a moderate influence on
operational guidance. A more important issue for operational guidance is the
course-keeping ability in bow seaways: if the ship is not able to avoid
excessive roll motions because it cannot steer into seaway, it is advisable to
consider this in the operational guidance; and

.4 for parametric roll in bow waves, neglecting propulsion, steering and

seakeeping abilities can lead to over-estimation of ship's safety in the design
assessment (due to contributions from safe but unattainable ship's speed
and course combinations) and to dangerous errors in terms of operational
guidance (when attainable ship's speed and course combinations in a storm
are unacceptable whereas all acceptable combinations are unattainable).

1.9.2 Figure 1.9.1 shows a colour plot of roll amplitude depending on forward speed and
course together with the line of maximum attainable speed (solid black line) and line of
maximum available steering effort (yellow dashed line) for the 8400 TEU container ship in three
loading conditions: in bow waves, majority of forward speeds that lead to small roll motions are
unattainable due to added resistance in seaway. Note that this observation is confirmed by
experience: all parametric roll accidents in bow waves happened at low forward speeds.

I:\CIRC\MSC\1\MSC.1-Circ.1652.docx
MSC.1/Circ.1652
Annex, page 254

T = 17.600s T = 17.600s T = 17.600s

Hs =10.5m Hs =10.5m Hs =10.5m

Figure 1.9.1 Colour plots of mean 3-hour maximum roll amplitude vs. forward speed (m/s,
radial coordinate) and wave direction (circumferential coordinate, head waves at top) for 8400
TEU container ship in loading conditions (from left to right) LC01, LC02 and LC03 together
with lines of maximum attainable speed (black solid) and maximum available steering effort
(yellow dashed)

1.9.3 To estimate the influence of propulsion ability on parametric roll in head waves,
average (over all significant wave heights and wave periods) rate of parametric roll stability
failures in head waves was calculated with and without considering maximum attainable speed
in head waves. In both cases, the forward speed was applied that minimizes the stability failure
rate, but in the calculations accounting for propulsion ability, the range of speeds was restricted
by the condition that the required engine power does not exceed the available power.
Figure 1.9.2 shows the result as the stability failure rate considering speed limit plotted vs. the
stability failure rate without considering speed limit.

Figure 1.9.2
Rate of
parametric roll
stability failures
in head waves
considering (y
axis) and not
considering (x
axis) attainable
forward speed
for three
r

container ships
(different
symbols) in
three loading
r conditions each

1.9.4 The results show that the stability failure rate increases by several orders of
magnitude if propulsion ability is considered. This means that it is advisable to consider
propulsion ability in operational guidance to prevent from misleading recommendations.
The attainable forward speed can be defined from model tests or numerical computations;
alternatively, simple empirical formulae can be established.

I:\CIRC\MSC\1\MSC.1-Circ.1652.docx
 MSC.1/Circ.1652
Annex, page 255

2 Application examples

2.1 Operational guidance based on DSA for parametric and synchronous roll

2.1.1 Operational guidance was prepared for all ships and loading conditions in table 1.1.1
by identifying unacceptable sailing conditions (v0, ) for each range of sea states (Hs, Tz) in the
North Atlantic wave scatter table. Unacceptable sailing conditions for probabilistic operational
guidance are those for which the upper boundary rU of the 95%-confidence interval of the
"short-term" stability failure rate exceeds standard 10 -6 s-1, while for deterministic operational
guidance, those for which the mean three-hour maximum roll amplitude 3h or lateral
acceleration amplitude ay3h exceed 0.5∙40o or 0.5∙g, respectively.

2.1.2 For each assumed situation (Hs, Tz, v0, ), numerical simulations of ship motions were
carried out in multiple independent sea state realizations produced by random variation of
phases, frequencies and directions of wave components discretizing the wave energy
spectrum. Transient hydrodynamic effects at the beginning of simulations were neutralized by
switching off the counter of stability failures and simulation timer during initial transients.

2.1.3 For probabilistic operational guidance, each simulation was conducted for 2 hours’
simulation time or until the first stability failure (exceedance of 40 degree roll angle or lateral
acceleration g) if a stability failure occurred earlier, after which new simulation was conducted
in another realization of the same sea state, until N = 200 stability failures were encountered.
After that, the upper boundary rU of the 95%-confidence interval of stability failure rate was
2
calculated as 𝑟U = 0.5𝑟𝜒1−0.05/2,2𝑁 /𝑁, where r = N/tt and tt is the total simulation time required
to encounter 200 stability failures. For such assumed situations (Hs, Tz, v0, ) where a total
simulation time of up to 3.4106 hours was not enough to encounter 200 stability failures,
extrapolation of stability failure rate over significant wave height was used (section 3.5.5.3 of
the Interim Guidelines). For deterministic operational guidance, five 3-hour simulations were
carried out for each assumed situation (Hs, Tz, v0, ); the mean 3-hour maximum roll amplitude
3h and mean 3-hour maximum lateral acceleration amplitude ay3h were calculated as average
values of the five 3-hour maxima from these simulations.

2.1.4 Table 2.1.1 shows the conservative estimate of the upper boundary of
the 95%-confidence interval of the average “long-term” stability failure rate 𝑟̄U , calculated,
according to the explanatory note to paragraph 3.5.3.2.1 of the Interim Guidelines, as the
"long-term" weighted average of the upper boundaries of the 95%-confidence intervals of the
"short-term" stability failure rate, for unrestricted service without operational guidance (i.e.
result of full probabilistic direct stability assessment) and with operational guidance. The
estimates of contributions to the total stability failure rate from principal parametric resonance
in bow and stern waves and synchronous resonance in beam waves are also shown.
Table 2.1.2 shows this upper boundary without and with operational guidance together with
operability resulting from the use of operational guidance. The minimum operability over
considered ships and loading conditions is 0.839, i.e. the ratio of the total duration of all
unacceptable situations to the total operational time (0.161) does not exceed 0.2, i.e. all loading
conditions can be considered as acceptable when operational guidance is used.

I:\CIRC\MSC\1\MSC.1-Circ.1652.docx
MSC.1/Circ.1652
Annex, page 256

Table 2.1.1 "Long-term" weighted average 𝒓̄ 𝐔 , 1/s, of upper boundaries of 95%-

confidence intervals of "short-term" stability failure rate for unrestricted service
without and with operational guidance: total (ALL) and due to principal parametric
resonance in bow (PRB) and stern (PRS) waves and synchronous resonance (SR) in
beam waves
Without operational guidance With operational guidance
LC
ALL PRB PRS SR ALL PRB PRS SR
01 2.214e-6 8.157e-7 6.407e-7 4.013e-7 6.238e-9 1.254e-9 1.857e-9 2.526e-9
02 5.706e-8 1.927e-8 2.220e-8 7.006e-9 1.219e-9 2.378e-10 4.218e-10 3.246e-10
Cruise

03 4.587e-9 3.979e-10 3.020e-9 5.323e-10 2.935e-10 3.409e-11 1.348e-10 7.222e-11

04 7.582e-10 1.020e-10 3.617e-10 1.642e-10 8.980e-11 1.259e-11 3.661e-11 2.622e-11
05 3.911e-10 8.124e-11 1.307e-10 1.091e-10 6.185e-11 1.065e-11 2.017e-11 1.940e-11
01 5.362e-5 3.802e-6 2.811e-5 1.230e-5 1.416e-8 1.030e-9 3.578e-9 7.741e-9
CV 1700

02 1.432e-6 2.798e-7 5.527e-7 3.585e-7 5.493e-9 9.476e-10 1.269e-9 2.569e-9

03 1.198e-8 3.048e-10 6.290e-9 3.390e-9 4.669e-10 3.027e-11 1.828e-10 1.907e-10
04 1.388e-12 1.363e-15 2.032e-15 1.378e-12 4.789e-13 9.539e-16 1.795e-15 4.721e-13
05 1.500e-12 1.748e-15 1.348e-16 1.490e-12 4.885e-13 1.145e-15 1.089e-16 4.829e-13
01 6.614e-6 8.859e-7 3.011e-6 1.592e-6 4.915e-9 4.709e-10 7.399e-10 2.388e-9
CV 8400

02 3.907e-6 8.559e-7 1.462e-6 9.285e-7 3.677e-9 3.715e-10 5.542e-10 2.127e-9

03 4.113e-7 1.162e-7 1.235e-7 1.088e-7 1.946e-9 3.250e-10 5.532e-10 8.926e-10
04 2.303e-7 8.462e-11 2.130e-11 2.294e-7 1.803e-9 8.954e-12 3.407e-12 1.740e-9
05 8.954e-9 9.143e-13 1.140e-14 8.943e-9 4.217e-10 3.395e-13 7.463e-15 4.193e-10
01 1.286e-4 6.742e-6 5.112e-5 4.936e-5 1.492e-8 1.236e-9 1.450e-9 1.067e-8
CV 14000

02 3.645e-5 8.235e-6 9.367e-6 1.315e-5 1.203e-8 4.983e-10 3.051e-9 6.746e-9

03 2.435e-6 1.015e-6 2.495e-7 7.956e-7 5.154e-9 1.030e-9 6.813e-10 2.603e-9
04 5.306e-16 0.0 0.0 5.306e-16 5.103e-16 0.0 0.0 5.103e-16
05 2.068e-15 0.0 0.0 2.068e-15 1.810e-15 0.0 0.0 1.810e-15
01 1.223e-9 5.589e-11 4.300e-10 5.082e-10 1.365e-10 7.638e-12 4.040e-11 6.208e-11
02 - - - - - - - -
RoPax

03 5.493e-11 8.643e-12 5.055e-12 3.311e-11 1.286e-11 1.991e-12 1.478e-12 7.394e-12

04 6.207e-11 8.255e-12 1.986e-12 4.578e-11 1.247e-11 1.692e-12 6.730e-13 8.774e-12
05 7.839e-11 7.538e-12 8.646e-13 6.495e-11 1.348e-11 1.606e-12 2.868e-13 1.057e-11

Table 2.1.2. "Long-term" weighted average 𝒓̄ 𝑼 , 1/s, of upper boundaries of 95%-

confidence intervals of "short-term" stability failure rate for unrestricted service
without and with probabilistic operational guidance (corresponding to short-term
threshold of stability failure rate rU =10-6 1/s) and operability due to use of operational
guidance
Loading condition
Ship
01 02 03 04 05
Average "long-term" stability failure rate 𝑟̄U , 1/s, without using operational guidance
Cruise Vessel 2.214e-6 5.706e-8 4.587e-9 7.582e-10 3.911e-10
1700 TEU Container Ship 5.362e-5 1.432e-6 1.198e-8 1.388e-12 1.500e-12
8400 TEU Container Ship 6.614e-6 3.907e-6 4.113e-7 2.303e-7 8.954e-9
14000 TEU Container Ship 1.286e-4 3.645e-5 2.435e-6 5.306e-16 2.068e-15
RoPax 1.223e-9 - 5.493e-11 6.207e-11 7.839e-11
"Long-term" weighted average 𝑟̄U , 1/s, due to using probabilistic operational guidance
Cruise Vessel 6.238e-9 1.219e-9 2.935e-10 8.980e-11 6.185e-11
1700 TEU Container Ship 1.416e-8 5.493e-9 4.669e-10 4.789e-13 4.885e-13
8400 TEU Container Ship 4.915e-9 3.677e-9 1.946e-9 1.803e-9 4.217e-10
14000 TEU Container Ship 1.492e-8 1.203e-8 5.154e-9 5.103e-16 1.810e-15
RoPax 1.365e-10 - 1.286e-11 1.247e-11 1.348e-11
Operability due to using probabilistic operational guidance
Cruise Vessel 0.977 0.998 1.000 1.000 1.000
1700 TEU Container Ship 0.868 0.981 0.999 1.000 1.000
8400 TEU Container Ship 0.972 0.976 0.993 0.996 1.000
14000 TEU Container Ship 0.839 0.907 0.975 1.000 1.000
RoPax 1.000 1.000 1.000 1.000 1.000

I:\CIRC\MSC\1\MSC.1-Circ.1652.docx
 MSC.1/Circ.1652
Annex, page 257

2.1.5 Figures 2.1.1(a and b) show examples of operational guidance in form of polar plots,
showing unacceptable sailing situations (red) in selected sea states for loading condition LC01
of 1700 TEU and 8400 TEU container ships.

Hs = 4m Hs = 5m Hs = 6m

Hs = 4m Hs = 5m Hs = 6m

Figure 2.1.1.a. Examples of probabilistic (top) and deterministic (bottom) operational

guidance in axes ship speed (knots, radial coordinate) - mean wave direction
(circumferential coordinate) vs. significant wave height (columns) for 1700 TEU
container ship

I:\CIRC\MSC\1\MSC.1-Circ.1652.docx
MSC.1/Circ.1652
Annex, page 258

Hs = 7m Hs = 8m Hs = 9m

Hs = 7m Hs = 8m Hs = 9m

Figure 2.1.1.b. Same as in figure 2.1.1.a for 8400 TEU container ship

I:\CIRC\MSC\1\MSC.1-Circ.1652.docx
 MSC.1/Circ.1652
Annex, page 259

2.2 Operational guidance for parametric roll

2.2.1 An example for development of operational guidance for a C11 class containers ship
is shown in figure 1.1 of appendix 2. Principal dimensions are given in table 1.2 of appendix 2,
assuming that GM = 1.4 m.

2.2.2. Examples of operational guidance in figure 2.1.1 show only acceptable and
non-acceptable combinations of speed and heading. Additional information on the proximity of
the condition to the boundary between acceptable and unacceptable sailing conditions may be
useful in operational decision-making. The additional information inside the zone of
unacceptable sailing conditions may be useful in assisting for leaving this zone; additional
information outside of this zone may serve as a warning.

2.2.3 As defined in paragraph 4.5.4.2 of the Interim guidelines, acceptable sailing

conditions are those for which the upper boundary of the estimate of failure rate is less than
10-6 s-1. Here, roll angle 40 degrees was used as the definition of stability failure, and the
grading of danger of stability failure is shown with additional thresholds larger than 40 degrees.
Additional thresholds of roll amplitudes less than 40 degrees are shown as a warning of
approaching danger of stability failure.

2.2.4 The example of an augmented guidance is shown in figure 2.2.1 where multiple roll
angle thresholds are used in addition to the threshold of 40 degrees. The thresholds 25, 30
and 35 degrees serve as waring and shown in yellow and orange colours; the thresholds 45,
and 50 degrees are included to assist in return to acceptable sailing condition, thus are shown
in different degrees of dark red colour.

Expected roll response

Warning

Unacceptable Figure 2.2.1.

Sample
operational
guidance in
form of polar
plot for C11-
class container
ship at
significant wave
height 9 m,
mean zero-
crossing wave
period 9.94 s,
draught 11.5 m
and GM=1.4 m

2.3 DSA-based operational limitations related to areas or routes and season

2.3.1 Examples of operational limitations related to areas or routes and season concern
ships and loading conditions shown in table 1.1.1 and sample operational routes and seasons
in table 2.3.1. Environmental conditions are specified by wave scatter tables according to IACS

I:\CIRC\MSC\1\MSC.1-Circ.1652.docx
MSC.1/Circ.1652
Annex, page 260

Recommendation No.34 (Corr.1 Nov. 2001) (area 1) and Global Wave Statistics (areas 2 to 6).

Table 2.3.1. Routes and seasons used in examples of operational limitations

1 North Atlantic, IACS Recommendation No.34 (Corr.1 Nov. 2001) (annual average)
2 Representative worldwide route for Panamax container ships trade* (annual average)
3 Representative east-bound route for post-Panamax container ships trade** (annual average)
4 Same as 3, summer
5 Sample European route*** (annual average)
6 Same as 5, summer
*
North Sea, Dover Strait, Biscay, North Atlantic, USA East Coast, Caribbean Sea, Panama Channel, West
coast of North America, North Pacific Ocean, Japan East Coast, South China Sea
**
North Sea, Dover Strait, Bay of Biscay, Gibraltar, Mediterranean Sea, Red Sea, North Indian Ocean, Strait
of Malacca, South China Sea, Japan
***
Baltic Sea, Skagerrak, North Sea, Dover Strait, Biscay

2.3.2 Operational limitations were produced, according to section 4.5.1, by application of

full probabilistic direct stability assessment (section 3.5.3.2 of Interim guidelines for direct
stability assessment) to each ship, loading condition and sample route, using numerical
simulations of ship motions at six forward speeds in waves of mean zero-crossing periods Tz
and significant heights Hs covering the relevant scatter tables every 1.0 s and 1.0 m,
respectively, and mean wave directions  from 0 to 180o every 10o.

2.3.3 Numerical simulations of ship motions were performed in realizations of the same sea
state by random variation of frequencies, directions and phases of wave components
discretizing the wave energy spectrum. Each simulation was conducted for the simulation time
of 2 hours or until the first stability failure (exceedance of 40 o roll angle or lateral acceleration g)
if it happened earlier, after which simulation was repeated in another realization, until N = 200
stability failures were encountered. For such combinations (Hs, Tz, v0, ) for which the total
simulation time of up to 3.4106 hours was enough to encounter 200 stability failures, direct
counting was used, namely, maximum likelihood estimate of stability failure rate was calculated
as r = N / tt, eq. (3.3.6), and the upper boundary of its 95%-confidence interval as 𝑟U =
2
0.5𝑟𝜒1−0.05/2,2𝑁 /𝑁, eq. (3.3.9); otherwise extrapolation of stability failure rate over significant
wave height was used.

2.3.4 A conservative estimate of the upper boundary 𝑟̄U of the 95%-confidence interval of
the average "long-term" stability failure rate was calculated as a weighted average over all
combinations (Hs, Tz, v0, ) of the upper boundaries rU of the 95%-confidence intervals of the
"short-term" stability failure rate, see explanatory note to paragraph 3.5.3.2.1 of the Interim
Guidelines, assuming uniform distributions of mean wave directions between 0 and 360 o and
ship forward speeds between zero and maximum service speed.

2.3.5 Table 2.3.2 shows the "long-term" weighted average 𝑟̄U of the upper boundaries of
the 95%-confidence intervals of the "short-term" stability failure rate for unrestricted operation
(area 1) and specific routes and seasons (2 to 6, table 2.3.1); bold numbers indicate
unacceptable loading conditions. For 64% of loading conditions, 𝑟̄U decreases compared to
unrestricted operation due to operational limitations, moreover, it decreases for 85% of
relevant loading conditions, i.e. those for which change of route may influence acceptance
decision. However, only one unacceptable loading condition of one ship changes to acceptable
due to change of route, whereas the result is opposite for another loading condition. Thus,
operational limitations related to areas or routes and season are less efficient than operational
guidance (which rendered as acceptable all considered loading conditions for all considered
ships, section 2.1), unless they are combined with other operational measures.

I:\CIRC\MSC\1\MSC.1-Circ.1652.docx
 MSC.1/Circ.1652
Annex, page 261

Table 2.3.2. "Long-term" weighted average 𝑟̄U of upper boundaries of 95%-confidence intervals
of "short-term" stability failure rate for areas or routes and seasons specified in table 2.2.1
Areas or routes and seasons
Ship LC
1 2 3 4 5 6

01 2.214e-6 1.549e-6 5.125e-7 6.940e-7 5.154e-7 4.243e-7

02 5.706e-8 1.114e-7 4.859e-8 6.461e-8 4.800e-8 2.702e-8
Cruise

03 4.587e-9 1.025e-8 4.066e-9 5.798e-9 4.051e-9 1.911e-9

04 7.582e-10 1.781e-9 5.266e-10 8.253e-10 5.456e-10 2.592e-10
05 3.911e-10 1.085e-9 3.276e-10 5.103e-10 3.388e-10 1.611e-10
01 5.362e-5 2.881e-5 1.117e-5 1.323e-5 1.118e-5 1.068e-5
CV 1700

02 1.432e-6 1.950e-6 1.076e-6 1.226e-6 1.058e-6 7.199e-7

03 1.198e-8 5.334e-8 3.686e-8 4.588e-8 3.567e-8 1.555e-8
04 1.388e-12 3.313e-11 2.290e-11 2.912e-11 2.140e-11 9.222e-12
05 1.500e-12 5.260e-11 4.137e-11 5.220e-11 3.864e-11 1.589e-11
01 6.614e-6 1.515e-6 3.156e-7 5.321e-7 3.306e-7 2.657e-7
CV 8400

02 3.907e-6 1.006e-6 2.104e-7 3.526e-7 2.204e-7 1.839e-7

03 4.113e-7 1.948e-7 4.685e-8 7.364e-8 4.865e-8 3.973e-8
04 2.303e-7 6.816e-8 1.383e-8 2.498e-8 1.484e-8 5.745e-9
05 8.954e-9 6.772e-9 1.550e-9 2.681e-9 1.662e-9 7.203e-10
01 1.286e-4 3.007e-5 6.692e-6 1.068e-5 6.922e-6 7.418e-6
CV 14000

02 3.645e-5 9.328e-6 2.044e-6 3.307e-6 2.117e-6 2.117e-6

03 2.435e-6 1.028e-6 2.541e-7 3.901e-7 2.618e-7 2.278e-7
04 5.306e-16 2.294e-20 5.613e-21 9.760e-21 6.113e-21 2.251e-21
05 2.068e-15 9.388e-18 2.676e-18 4.343e-18 2.840e-18 1.257e-18
01 1.223e-9 5.105e-9 2.119e-9 2.959e-9 2.093e-9 1.028e-9
02 - - - - - -
RoPax

03 5.493e-11 4.279e-10 1.937e-10 2.655e-10 1.886e-10 9.002e-11

04 6.207e-11 6.398e-10 3.178e-10 4.254e-10 3.056e-10 1.435e-10
05 7.839e-11 1.068e-9 5.942e-10 7.767e-10 5.647e-10 2.572e-10

2.4 DSA-based operational limitations related to maximum significant wave height

2.4.1 Operational limitations related to maximum significant wave height accept operation
in sea states up to a specified maximum significant wave height. Sample operational limitations
were developed for all ships and loading conditions in table 1.1.1 using the North Atlantic wave
scatter table according to IACS Recommendation No.34 (Corr.1 Nov. 2001) (also shown in
table 2.7.2.1.2), limited by a systematically varied maximum significant wave height with a step
1.0 m.

2.4.2 Operational limitations were produced, according to section 4.5.2 , by applying full
probabilistic direct stability assessment (section 3.5.3.2) for each specified maximum
significant wave height, using numerical simulations of ship motions at six forward speeds in
waves of mean zero-crossing periods Tz and significant wave heights Hs covering all entries in
the scatter table below the specified maximum significant wave height with steps 1.0 s
and 1.0 m, respectively, and mean wave directions from 0 to 180 o every 10o.

2.4.3 Numerical simulations of ship motions were performed in realizations of the same sea
state by random variation of frequencies, directions and phases of wave components
discretizing the wave energy spectrum. Each simulation was conducted for the simulation time
of 2 hours or until the first stability failure (exceedance of 40 o roll angle or lateral acceleration g)
if it happened earlier, after which simulation was repeated in another realization, until N = 200
stability failures were encountered. Transient hydrodynamic effects at the start of simulations
were neutralized by switching off the counter of stability failures and simulation timer during
initial transients. For such combinations (Hs, Tz, v0, ), for which the total simulation time
of 3.4106 hours was enough to encounter 200 stability failures, direct counting was used: the
maximum likelihood estimate of stability failure rate was calculated as r = N / tt, eq. (3.3.6), and

I:\CIRC\MSC\1\MSC.1-Circ.1652.docx
MSC.1/Circ.1652
Annex, page 262

2
upper boundary of its 95%-confidence interval as 𝑟U = 0.5𝑟𝜒1−0.05/2,2𝑁 /𝑁 , eq. (3.3.9);
otherwise extrapolation of failure rate over significant wave height was used.

2.4.4 A conservative estimate of the upper boundary 𝑟̄U of the 95%-confidence interval of
the average "long-term" stability failure rate was calculated as a weighted average, over all
combinations (Hs, Tz, v0, ), of the upper boundaries rU of the 95%-confidence intervals of the
"short-term" stability failure rate, see explanatory note to paragraph 3.5.3.2.1, assuming
uniform distributions of mean wave directions between 0 and 360 o and ship forward speeds v0
between zero and maximum service speed.

2.4.5 Figure 2.4.1 shows the "long-term" weighted average 𝑟̄U of the upper boundaries of
the 95%-confidence intervals of the "short-term" stability failure rate (total and separately due
to principal parametric resonance in bow and stern waves and synchronous roll in beam
waves) and operability vs. the maximum significant wave height. Table 2.4.1 shows the
maximum significant wave height at which the "long-term" weighted average of the upper
boundaries of the 95%-confidence intervals of the 'short-term' stability failure rate is equal to
the standard 2.6∙10-8 1/s and the operability corresponding to this wave height.

Figure 2.4.1 (below):

"Long-term" weighted average 𝒓̄ 𝐔 of upper boundaries of 95%-confidence intervals of
"short-term" stability failure rate (left y-axis): total (ALL, black solid lines) and due to
principal parametric resonance in bow (PRB, red dashed lines) and stern (PRS, red
dash-dot lines) waves and synchronous roll (SR, green dash-dot-dot lines) in beam
waves and operability (right y-axis, blue dashed line marked O) vs. maximum significant
wave height (x-axis) in North Atlantic wave climate

I:\CIRC\MSC\1\MSC.1-Circ.1652.docx
 MSC.1/Circ.1652
Annex, page 263

I:\CIRC\MSC\1\MSC.1-Circ.1652.docx
MSC.1/Circ.1652
Annex, page 264

Table 2.4.1. Maximum significant wave height at which "long-term" weighted average
𝒓̄ 𝐔 of upper boundaries of 95%-confidence intervals of "short-term" stability failure rate
does not exceed standard 2.6∙10-8 1/s (“unl.” stands for ʺunlimitedʺ, meaning that
limitation is not required) and corresponding operability (bold numbers indicates
values less than 0.8) in North Atlantic wave climate

Ship  Maximum significant wave height, m Corresponding operability

Loading conditions → 01 02 03 04 05 01 02 03 04 05
Cruise Vessel 5.266 10.763 unl. unl. unl. 0.883 0.999 1.0 1.0 1.0
1700 TEU CV 2.814 4.746 unl. unl. unl. 0.557 0.839 1.0 1.0 1.0
8400 TEU CV 5.009 5.111 6.796 9.576 unl. 0.850 0.858 0.951 0.994 1.0
14000 TEU CV 2.479 3.317 5.225 unl. unl. 0.476 0.639 0.876 1.0 1.0
RoPax unl. unl. unl. unl. unl. 1.0 1.0 1.0 1.0 1.0

2.4.6 Note that in the same wave climate, a probabilistic operational guidance allows
achieving significantly greater operability than operational limitations related to maximum
significant wave height, compare tables 2.1.2 and 2.4.1. Moreover, according to the results in
table 2.1.2, the ratio of the total duration of all unacceptable situations to the total operational
time does not exceed 0.2 for all considered loading conditions when operational guidance is
used (which means that all considered loading conditions are acceptable if operational
guidance is provided), whereas this ratio exceeds 0.2 (i.e. operability is less than 0.8) for three
of the same loading conditions when operational limitations related to maximum significant
wave height are used. This means that operational limitations related to maximum significant
wave height are less efficient than operational guidance for the considered example
applications.

2.5 Simplified operational guidance for surf-riding/broaching failure mode

2.5.1 The simplified operational guidance based on the methodology used for the
vulnerability criteria for the surf-riding/broaching failure mode is provided in
paragraph 4.5.6.2.4. The simplified operational guidance is based on indications from the
Operational Diagram for the Master in MSC.1/Circ./707, which was used for practical
application between 1996 and 2007 without actual problems reported. However, the simplified
operational guidance based on level 2 vulnerability criterion specified in paragraph 4.5.6.2.4.2
requires validation and application examples.

2.5.2 For this purpose, the probabilistic indices based on paragraph 4.5.6.2.4.2, CHT, equal
to the conditional probability of surf-riding when the ship meets a wave, were calculated for
several short-term sea states with different ship speeds and courses and compared with the
broaching failure probability obtained with the critical wave method, which is equal to the
conditional probability of large heel due to broaching associated with surf-riding when the ship
meets a wave. The subject ship is an offshore research vessel, which is similar to a vessel lost
in stern quartering waves at high forward speed; the sea states are based on those specified
as the design situations in paragraphs 3.5.3.3.4 and 3.5.3.3.5. The tested ship speed ranges
from 0.015 to 0.4 and the relative wave heading from 1 to 90 degrees.

2.5.3 Figures 2.5.1 and 2.5.2 indicate that the results of level 2 vulnerability criterion provide
reasonably conservative estimate of the danger due to broaching in comparison with direct
assessment for the considered sea states and sailing conditions, thus the simplified
operational guidance based on level 2 vulnerability criterion specified in paragraph 4.5.6.2.4.2
is effective to assist the ship master to avoid dangerous situations. Note that such simplified
operational guidance can be obtained as intermediate calculation data from the application of
level 2 vulnerability criterion, thus no additional computational effort is required for the
preparation of such simplified operational guidance.

I:\CIRC\MSC\1\MSC.1-Circ.1652.docx
 MSC.1/Circ.1652
Annex, page 265

Hs = 2.80, Tz = 2.66, Fn = 0.15 Hs = 2.80, Tz = 2.66, Fn = 0.20

Hs = 2.80, Tz = 2.66, Fn = 0.25 Hs = 2.80, Tz = 2.66, Fn = 0.30

Hs = 2.80, Tz = 2.66, Fn = 0.35 Hs = 2.80, Tz = 2.66, Fn = 0.40

Figure 2.5.1 Comparison of level 2 criterion with failure probability computed with critical
wave method in sea state with mean zero-crossing wave period 2.66 s (which corresponds to
wavelength 1.0L) and significant wave height 2.8 m from table 3.5.3.3

I:\CIRC\MSC\1\MSC.1-Circ.1652.docx
MSC.1/Circ.1652
Annex, page 266

Hs = 2.80, Tz = 3.22, Fn = 0.15 Hs = 2.80, Tz = 3.22, Fn = 0.20

Hs = 2.80, Tz = 3.22, Fn = 0.25 Hs = 2.80, Tz = 3.22, Fn = 0.30

Hs = 2.80, Tz = 3.22, Fn = 0.35 Hs = 2.80, Tz = 3.22, Fn = 0.40

Figure 2.5.2. Comparison of level 2 criterion with failure probability computed with critical
wave method in sea state with mean zero-crossing wave period 3.22 s (which corresponds to
wavelength 1.5L) and significant wave height 2.8 m from table 3.5.3.3.5

2.6 Simplified operational guidance from level 2 vulnerability assessment for

parametric roll

2.6.1 For the use of the simplified operational guidance for parametric rolling, which is
defined in the paragraph 4.5.6.2.3 of the Interim Guidelines, it is essential to confirm the
prediction accuracy of the method in the second check of the level 2 criterion for parametric
rolling. This is because the operational guidance should be used for actual irregular waves
while the method in the level 2 is based on the calculation in the equivalent regular waves.

I:\CIRC\MSC\1\MSC.1-Circ.1652.docx
 MSC.1/Circ.1652
Annex, page 267

Thus, the comparisons between the numerical results using the second check of the level 2
methodology and the existing experimental results in irregular waves are presented here.

2.6.2 The comparisons for the C11 class containership in head waves are shown in
figures 2.6. 1-2. The simulation results to be used in the simplified operational guidance
somewhat overestimate the experimental results34 in irregular waves.

Fn

Figure 2.6. 1 Comparison in the roll amplitude between the simulation to be used for
simplified guidance and the model experiment in irregular waves for the C11 class
containership in head seas with the significant wave height of 7.82 m and the mean
wave period of 9.99 s for different Froude numbers

Hs[m]

Figure 2.6.2 Comparison in the roll amplitude between the simulation to be used for
simplified guidance and the model experiment in irregular waves for the C11 class
containership in head seas with the mean wave period of 9.99 s and the Froude number
of 0.0 for different significant wave heights

2.6.3 The comparison for the 150m-long containership in following waves is shown in
figure 2.6.3. Here the roll angle of 80 degrees indicates capsizing. The reason for the frequent
occurrence of capsizing due to parametric rolling is small metacentric height, which is critical
to the 2008 Intact Stability Code and severest sea state in the oceans. In this case, also the

34
Hashimoto, H. and N. Umeda. Prediction of Parametric Rolling in Irregular Head Waves. Chapter 16 of
Contemporary Ideas on Ship Stability. Risk of Capsizing, Belenky, V., Spyrou, K., van Walree F., Neves,
M.A.S., and N. Umeda, eds., Springer, ISBN 978-3-030-00514-6, pp. 275-289, 2019.

I:\CIRC\MSC\1\MSC.1-Circ.1652.docx
MSC.1/Circ.1652
Annex, page 268

simulation results to be used in the simplified operational guidance somewhat overestimate

the experimental results35 in irregular waves.

Fn

Figure 2.6.3 Comparison in the roll amplitude between the simulation to be used for
simplified guidance and the model experiment in irregular waves for the 150m-long
containership having the metacentric height of 0.15m in following seas with the
significant wave height of 13.26 m and the mean wave period of 10.92 s for different
Froude numbers

2.6.4 The comparison for the 192m-long PCTC in head waves is shown in figure 2.6.4. In
this case, the simulation results to be used in the simplified operational guidance, which is
labelled as "simulation 1" do not always overestimate the experimental results 36 in irregular
waves. This is because the roll amplitude in regular waves does not always increase with the
increasing the wave height, which is due to the increase of the mean of GM with the wave
height. Roughly speaking, the amplitude of the GM variation is proportional to the wave height,
while the mean of GM variation is proportional to the square of the wave height. Thus,
sometimes the condition for parametric roll will not be met at a certain wave height or above.
On the other hand, the roll amplitude in irregular waves normally increases with the increasing
the wave height because of the spectrum of incident waves. To confirm this mechanism, the
simulation ignoring the mean of GM variation is also shown as "simulation 2". As a result, the
roll amplitude ignoring the mean of GM variation increases with the increasing wave height.
This is a drawback of the simplified approach using the effective regular waves. Thus, the
second check of the level 2 criterion requires to keep the maximum roll amplitude if the roll
amplitude decreases with the increasing wave height in paragraph 3.5 of appendix 3.
Therefore, the same procedure should be applied to the simplified operational guidance.

35
Umeda, N., M. Hamamoto, Y. Takaishi, Y. Chiba, A. Matsuda, W. Sera, S. Suzuki, K. J. Spyrou and K.
Watanabe. Model Experiments of Ship Capsize in Astern Seas. Journal of the Society of Naval Architects of
Japan, Vol.177, pp.207-217, Jun. 1995.
36
Umeda N., Hashimoto H., Tsukamoto I., Sogawa Y. Estimation of Parametric Roll in Random Seaways.
In: Fossen T., Nijmeijer H. (eds) Parametric Resonance in Dynamical Systems. Springer, New York, NY,
ISBN 978-1-4614-10423-0, pp: 45-59, 2012.

I:\CIRC\MSC\1\MSC.1-Circ.1652.docx
 MSC.1/Circ.1652
Annex, page 269

Hs[m]

Figure 2.6.4 Comparison in the roll amplitude between the simulation to be used for
simplified guidance and the model experiment in irregular waves for the 192m-long
PCTC in head seas with the mean wave period of 9.76 s and the Froude number of 0.0
for different significant wave heights

2.6.5 In conclusion, if the requirement in paragraph 3.5 of appendix 3 is also taken into
account, the calculation based on the second check of the level 2 criterion normally
overestimates the model experiment. Since the numerical model used for the full operational
guidance is required to be validated with model experiments, the simplified operational
guidance for parametric rolling is expected to provide conservative estimates for the danger of
parametric rolling in actual longitudinal seas. The remaining issue is the absence of the
requirement for heading in the simplified guidance. However, the magnitude of parametric
rolling in oblique waves can be reduced if the heading angle leaves from the longitudinal waves.
Therefore, the safety level realized with the simplified operational guidance for parametric
rolling is logically higher than that with full operational guidance.

2.7 Level 2-based operational limitations related to maximum wave height

2.7.1 A C11-class post-Panamax container ship was used as a sample ship; table 2.7.1
shows principal dimensions. This ship in the considered loading condition failed to pass
check 1 and check 2 of vulnerability assessment for parametric roll, table 2.7.2. Therefore,
operational measures were considered.

Table 2.7.1 Principal parameters of sample ship and loading condition

Length between perpendiculars, LBP 262.0 m
Breadth, B 40.0 m
Draught, d 11.5 m
Total projected area of bilge keels, ABK 30.6 m2
Service speed, Vs 23.6 knots
Metacentric height, GM 1.965 m
Natural roll period, Tr 25.1 s

Table 2.7.2 Results of level 2 vulnerability check for parametric roll for sample ship
Value Required value Judgement
C1 = 0.4368 RPRO = 0.06 fail
C2 = 0.02592 RPR1 = 0.025 fail

2.7.2 If level 2 vulnerability criterion for parametric roll is applied for the preparation of
operational limitations, the corresponding procedure of the vulnerability check and the value
of the standard should be used.

I:\CIRC\MSC\1\MSC.1-Circ.1652.docx
MSC.1/Circ.1652
Annex, page 270

2.7.3 Operational limitations related to maximum significant wave height are developed for
environmental conditions which are defined by cutting, at a specified significant wave height,
the wave scatter table for a specified area or a specified route during a specified season and
by corresponding modification of wind statistics.

2.7.4 The wave scatter table from level 2 vulnerability assessment for parametric roll was
used, paragraph 2.5.3.4.2. Sea state probabilities up to specified significant wave heights were
summed, as shown in table 2.7.3. Since the operability should be not less than 0.8 to apply
operational limitations related to maximum significant wave height, the maximum significant
wave height should exceed 5.0 m for the considered scatter table.

Table 2.7.3 Sum of sea state probabilities up to specified maximum significant wave
height
Specified maximum Hs, m Probability of occurrence Sum
1.0 0.030504 0.03050
2.0 0.225754 0.25626
3.0 0.238104 0.49436
4.0 0.191277 0.68564
5.0 0.132894 0.81853
6.0 0.083281 0.90181
7.0 0.048063 0.94988
8.0 0.025862 0.97574
9.0 0.013087 0.98883
10.0 0.006262 0.99509
11.0 0.002848 0.99794
12.0 0.001236 0.99917
13.0 0.000511 0.99968
14.0 0.000205 0.99989
15.0 0.000077 0.99997
16.0 0.000028 0.99999
17.0 0.000009 1.00000

2.7.5 Check 2 of level 2 vulnerability criteria for parametric roll was used for the preparation
of operational limitations. The results in table 2.7.4 indicate that the maximum operational
significant wave height is 10.0 m for the considered loading condition. Since it is greater
than 5.0 m, paragraph 2.7.4, operational limitations with the maximum operational significant
wave height of 10.0 m can be applied.

Table 2.7.4 Vulnerability check with cut wave scatter diagrams

Specified Hs for cut-off, m C2 Judgement
5.0 0.00858 pass
6.0 0.01509 pass
7.0 0.01892 pass
8.0 0.02194 pass
9.0 0.02391 pass
10.0 0.02496 pass
11.0 0.02551 fail
12.0 0.02575 fail
13.0 0.02585 fail
14.0 0.02590 fail
15.0 0.02591 fail
16.0 0.02592 fail
none 0.02592 fail

I:\CIRC\MSC\1\MSC.1-Circ.1652.docx
 MSC.1/Circ.1652
Annex, page 271

3 Wave cases for preparation of operational limitations using level 1 and level 2
vulnerability assessment

3.1 Wave cases and associated parameters for the preparation of operational limitations
using level 1 and level 2 vulnerability assessment criteria should be defined in accordance with
the procedure described in section 10 of appendix 3.

4 Supplementary information on preparation of simplified operational guidance

for surf-riding/broaching failure mode

4.1 For predicting surf-riding/broaching, accurate estimation of the wave-induced surge

force is indispensable. For level 2 vulnerability criterion for surf-riding/broaching in 2.6.3.4.5,
the Froude-Krylov component is considered. The Froude-Krylov component alone normally
overestimates the wave-induced surge force in situations relevant for surf-riding. This is
because the diffraction component, which is the effect of hull disturbance on incident waves is
not small. Thus, the diffraction effect may be considered for practical use.

4.2 The diffraction effect can be estimated with CFD or three-dimensional linear potential
flow codes; for simplified operational guidance, the following empirical formula for fij can be
used instead of that in paragraph 2.6.3.4.5 in MSC.1/Circ.1627 as an alternative to CFD or
equivalent:

𝑓𝑖𝑗 = 0.5𝜌𝑔𝜇𝑥 𝑘𝑖 𝐻𝑖𝑗 √𝐹𝑐𝑖2 + 𝐹𝑠𝑖2 (4.1)

4.3 In eq. (4.1), the coefficient x is defined using the following formulae based on
experimental data for two container ships, a car carrier, a RoRo ship, a fishing vessel and two
warships:

x = 1.46 Cb – 0.05 for Cm < 0.86

= (5.76 – 5.00 Cm) Cb – 0.05 for 0.86  Cm < 0.94 (4.2)
= 1.06 Cb – 0.05 for Cm  0.94

where Cb is the block coefficient and Cm is the midship section coefficient.

I:\CIRC\MSC\1\MSC.1-Circ.1652.docx
MSC.1/Circ.1652
Annex, page 272

APPENDIX 6

APPLICATION EXAMPLES OF TREATMENT OF LOADING CONDITIONS

Introduction

1 The application of vulnerability criteria should cover all loading conditions intended for
the ship's operation. In this respect, GM, draught and trim should be appropriately considered.
However, for the sake of simplicity, the application examples here cover only combinations of
GM and draught. Specifically, matrix calculations have been carried out, where vulnerability
criteria have been applied for each combination of GM and draught; and the roll period has
been estimated in accordance with paragraph 2.7.1.2.

2 Paragraphs 2.2.1.3, 2.3.1.3, 2.4.1.3, 2.5.1.2 and 2.6.1.2 allow the use of direct stability
assessment or operational measures as alternatives to the vulnerability criteria specified for
each particular failure mode. Nevertheless, for demonstration purposes, the application
examples stipulated here show only the results of probabilistic direct stability assessment and
probabilistic operational measures for the parametric rolling and pure loss of stability failure
modes. It is also noted that the verification of failure modes, according to section 3.5.2, has
not been carried out for the reported example applications. Therefore, reported direct stability
assessment results are conservative compared to those that would be obtained by applying
section 3.5.2. Furthermore, for each example ship, results of the direct stability assessment
are shown only for a typical draught. Examples of direct stability assessment for other loading
conditions can be found in section 4.2 of appendix 4.

3 Since direct stability assessment requires significant computational efforts, the user
may be guided by a sequential logic of application of the Interim Guidelines (section 1.1.3
therein). In this regard, direct stability assessment may be applied to the loading conditions
that are indicated to be potentially vulnerable according to the vulnerability criteria for the
relevant failure mode.

4 The numerical model used in the direct stability assessment should be validated
based on paragraph 3.4.1.2 and the identified failure mode in the direct stability assessment
should be the same as that used in the validation (see paragraph 3.5.2.1).

5 In these application examples, whenever the Weather Criterion is mentioned, the use
of MSC.1/Circ.1200 is not taken into account. Thus, for the actual application, the possibility
of its use could be considered in accordance with the provisions of part A, paragraphs 2.3.3
and 2.3.5 of the 2008 IS Code.

6 When comparing specific loading conditions with results from the assessment, GM
values corrected for free surface effects should be used for the dead ship condition, pure loss
of stability and parametric rolling failure modes (see paragraphs 2.2.1.7, 2.4.1.7 and 2.5.1.6,
respectively); whereas GM values not corrected for free surface effects should be used for
excessive acceleration failure mode (see paragraph 2.3.1.7).

7 In the matrix calculation shown in this appendix, red and blue colours indicate that the
ship is "possibly vulnerable" and "acceptable" to the failure modes, respectively.

I:\CIRC\MSC\1\MSC.1-Circ.1652.docx
 MSC.1/Circ.1652
Annex, page 273

1 Examples of the treatment of loading conditions

1.1 Cruise ship

1.1.1 An existing cruise vessel with the length between perpendiculars 230.9 m and
waterline breadth 32.2 m was used as an example.

1.1.2 The criteria from part A of the 2008 IS Code and the deterministic damage stability
requirements in the SOLAS (as amended by resolutions in effect as on 1 July 2004) result in
minimum GM dependencies on the draft shown in figure 1.1.1.
GM

Figure 1.1.1. Minimum GM curves vs. draft according to the 2008 IS Code part A criteria
and SOLAS damage stability requirements for cruise vessel

1.1.3 Figure 1.1.2 shows results of assessment with respect to level 1 and level 2
vulnerability criteria for parametric roll stability failure mode. Since it is sufficient to satisfy one
of these three assessment options, these criteria do not suggest additional recommendations
on the minimum GM for this vessel compared to the criteria from part A of the 2008 IS Code
and SOLAS damage stability requirements because the mandatory criteria shown in
figure 1.1.1 supersede those recommended in figure 1.1.2.

I:\CIRC\MSC\1\MSC.1-Circ.1652.docx
MSC.1/Circ.1652
Annex, page 274

GM
GM

Figure 1.1.2 Result of level 1

(top left – PR_L1), level 2 check 1
(top right – PR_L2_1) and level 2
check 2 (bottom left – PR_L2_1)
GM

assessment for the parametric roll

stability failure mode for cruise
vessel

1.1.4 Assessment with respect to level 1 and level 2 vulnerability criteria for the pure loss
of stability failure mode, figure 1.1.3, shows that the ship is vulnerable with respect to level 1
criterion in all combinations of draught and GM shown in figure 1.1.3. This failure mode does
not suggest additional recommendations on the minimum GM for this vessel compared to the
criteria from part A of the 2008 IS Code and SOLAS damage stability requirements.
GM
GM

Figure 1.1.3 Results of level 1 and level 2 assessment for pure loss of stability failure
mode for cruise vessel
1.1.5 Because the length of the vessel is greater than 200 m, an assessment with respect
to the level 1 criterion for surf-riding/broaching indicates that the ship is not vulnerable to this
stability failure mode in all loading conditions.
1.1.6 Since the natural roll period of the ship in the considered loading conditions is below
(or only marginally above) 20 s, there is no difference between the Weather Criterion from
part A of the 2008 IS Code and level 1 criterion for dead ship condition stability failure mode,

I:\CIRC\MSC\1\MSC.1-Circ.1652.docx
 MSC.1/Circ.1652
Annex, page 275

figure 1.1.4. Application of level 2 criterion for dead ship condition stability failure mode does
not suggest additional recommendations on the minimum GM value.

GM
GM

Figure 1.1.4. Results of level 1 and level 2 assessment for dead ship condition stability
failure mode for cruise vessel, DS_L1 and DS_L2, respectively
1.1.7 The application of vulnerability criteria for the excessive acceleration stability failure
mode leads to recommendations on the upper limit of GM. Figure 1.1.5 shows that level 2
allows significantly greater maximum GM values than level 1, which are also well above the
GM range relevant in practice for this vessel. Users should note that the GM values associated
with results of excessive acceleration criteria represent metacentric heights without correction
for free surface.
GM

GM

Figure 1.1.5. Results of level 1 and level 2 assessment for excessive acceleration
stability failure mode for cruise vessel

1.1.8 For the considered cruise vessel, the second-generation intact stability criteria do not
result in any additional recommendations relative to part A of the 2008 IS Code and the
damage stability requirements of SOLAS on the minimum GM values (for the considered,
practically relevant, range of draughts). Since the limiting criterion for the minimum GM values
is the Weather Criterion, figure 1.1.6, and that the assessment for the level 2 criterion for the
dead ship failure mode indicates lower minimum GM values, then the application of
MSC.1/Circ.1200 may be utilized to evaluate revised minimum GM values. The vulnerability
criteria for excessive accelerations impose additional limitations on the maximum GM values
(however, these limitations are above the GM values relevant in practice).

I:\CIRC\MSC\1\MSC.1-Circ.1652.docx
MSC.1/Circ.1652
Annex, page 276
GM

Figure 1.1.6. GM limits according to the 2008 IS Code part A, SOLAS damage stability
requirements and second-generation intact stability criteria (including maximum GM
according to excessive accelerations criterion) for cruise vessel

1.1.9 Six loading conditions at the practically most relevant draught 6.9 m and GM
values 1.5, 2.0, 2.5, 3.25 and 3.75 m, table 4.1.1 of chapter 4 of appendix 4, were assessed
using full probabilistic direct stability assessment. Figure 1.1.7 plots the conservative estimate
of the upper boundary 𝑟̄U of the 95%-confidence interval of the average "long-term" stability
failure rate, calculated as the weighted average of the upper boundaries of the 95%-confidence
intervals of the "short-term" stability failure rate (see explanatory note to paragraph 3.5.3.2.1
of the Interim Guidelines). The standard 2.6∙10-8 1/s for 𝑟̄U is satisfied for GM values
above 2.158 m. This is greater than the minimum required GM from chapter 2 of
MSC.1/Circ.1627 – vulnerability requirements (which indicates inconsistency between the
vulnerability assessment and direct stability assessment for the considered ship and draught).
However, the weather criterion requires greater minimum GM and, therefore, this consistency
does not suggest any additional recommendations for the minimum GM value.
rU

GM

GM
Figure 1.1.7. Conservative estimate of upper boundary 𝒓̄ 𝐔 of 95%-confidence interval
of average "long-term" stability failure rate vs. GM at draught 6.9 m for cruise vessel
in comparison with acceptance standard 2.6∙10 -8 1/s and resulting acceptable GM
range

I:\CIRC\MSC\1\MSC.1-Circ.1652.docx
 MSC.1/Circ.1652
Annex, page 277

1.1.10 Probabilistic operational guidance was prepared for the same loading conditions by
identifying unacceptable sailing conditions (v0,), i.e. sailing conditions for which the upper
boundary of 95%-confidence interval of "short-term" stability failure rate exceeds acceptance
standard 10-6 s-1, for each sea state (Hs,Tz) in the North Atlantic wave scatter table. Figure 1.1.8
shows "long-term" weighted average 𝑟̄U of upper boundaries of the 95%-confidence intervals
of the "short-term" stability failure rate with and without using operational guidance together
with the operability due to the use of operational guidance vs. GM. Since operability
exceeds 0.8, operational guidance is an acceptable option for all considered loading conditions.
rU

rU

rU

rU

GM,m
Figure 1.1.8. "Long-term" weighted average of upper boundaries of 95%-confidence
intervals of "short-term" stability failure rate with and without OG and operability vs.
GM at draught 6.9 m for cruise vessel

1.1.11 Examples of operational limitations related to areas or routes and season concern the
same loading conditions for sample operational routes and seasons from table 2.2.1, chapter 2
of appendix 5. Table 1.1.1 shows the upper boundary 𝑟̄U of the 95%-confidence interval of
average "long-term" stability failure rate for unrestricted operation (area 1) and specific routes
and seasons (areas 2 to 6); red colour indicates unacceptable loading conditions. The stability
failure rate generally decreases for considered sample routes and seasons compared to
unrestricted service but the reduction is insufficient to render the loading conditions that are
unacceptable for unrestricted operation acceptable for considered specific routes and
seasons.

Table 1.1.1. "Long-term" weighted average 𝒓̄ 𝐔 of upper boundaries of 95%-confidence

intervals of "short-term" stability failure rate for areas or routes and seasons specified
in table 2.2.1, chapter 2 of appendix 5
Areas or routes and seasons
LC
1 2 3 4 5 6
01 2.214e-6 1.549e-6 5.125e-7 6.940e-7 5.154e-7 4.243e-7
02 5.706e-8 1.114e-7 4.859e-8 6.461e-8 4.800e-8 2.702e-8
03 4.587e-9 1.025e-8 4.066e-9 5.798e-9 4.051e-9 1.911e-9
04 7.582e-10 1.781e-9 5.266e-10 8.253e-10 5.456e-10 2.592e-10
05 3.911e-10 1.085e-9 3.276e-10 5.103e-10 3.388e-10 1.611e-10

1.1.12 Operational limitations related to maximum significant wave height were developed
for loading conditions LC01, LC02 and LC03 for the North Atlantic wave scatter table, limited
by a systematically varied maximum significant wave height with a step 1.0 m. Figure 1.1.9
shows the "long-term" weighted average 𝑟̄U of the upper boundaries of the 95%-confidence
intervals of the "short-term" stability failure rate and operability vs. maximum significant wave

I:\CIRC\MSC\1\MSC.1-Circ.1652.docx
MSC.1/Circ.1652
Annex, page 278

height. Table 1.1.2 shows the significant wave height at which 𝑟̄U is equal to the
standard 2.6∙10-8 1/s, together with operability corresponding to this wave height. Note that
since operability exceeds 0.8, operational limitations related to maximum significant wave
height is an acceptable option for all considered loading conditions, and that in the same wave
climate, a probabilistic operational guidance allows achieving significantly greater operability
than operational limitations related to maximum significant wave height, i.e. operational
limitations related to maximum significant wave height are less efficient than operational
guidance.

rU

rU rU rU

Figure 1.1.9. "Long-term" weighted average 𝒓̄ 𝐔 of upper boundaries of

95%- confidence intervals of "short-term" stability failure rate (left y-axis, black solid
line) and operability (right y-axis, blue dashed line) vs. maximum significant wave
height (x-axis) in North Atlantic wave climate for cruise vessel

Table 1.1.2. Maximum significant wave height at which "long-term" weighted average
𝒓̄ 𝐔 of upper boundaries of 95%-confidence intervals of "short-term" stability failure
rate does not exceed standard 2.6∙10-8 1/s and corresponding operability in North
Atlantic wave climate for cruise vessel
Loading condition LC01 LC02 LC03
GM, m 1.5 2.0 2.5
Maximum significant wave height, m 5.266 10.763 unlimited
Corresponding operability 0.883 0.999 1.000

1.2 1700 TEU container ship

1.2.1 This example concerns a container ship with the length between
perpendiculars 159.6 m and waterline breadth 28.1 m.

1.2.2 Figure 1.2.1 shows minimum GM values vs. draught according to the criteria from
part A of the 2008 IS Code and damage stability requirements in SOLAS (as amended by
resolution MSC.216(82)).

I:\CIRC\MSC\1\MSC.1-Circ.1652.docx
 MSC.1/Circ.1652
Annex, page 279
GM

Figure 1.2.1. Minimum GM curves according to the 2008 IS Code part A criteria and
SOLAS damage stability requirements for a 1700 TEU container ship

1.2.3 Figure 1.2.2 shows results of vulnerability assessment for parametric roll stability
failure mode. Since it is sufficient to satisfy one of these three assessment options, they do not
impose additional limitations on the minimum GM value compared to the criteria from part A of
the 2008 IS Code and SOLAS damage stability requirements. Figure 1.2.3 shows assessment
results with respect to vulnerability criteria for pure loss of stability failure mode. Assessment
with respect to level 1 criterion for surf-riding/broaching indicates that since the operational
Froude number of the vessel is less than 0.3, the vessel is not vulnerable to this stability failure
mode in any of the considered loading conditions, therefore level 2 assessment was not
performed. The assessment with respect to vulnerability criteria for dead ship condition,
figure 1.2.4, shows that the ship is not vulnerable to this stability failure mode in all considered
combinations of draught and GM.
GM
GM

Figure 1.2.2 Results of level 1 (top left),

level 2 check 1 (top right) and level 2
check 2 (bottom left) vulnerability
GM

assessment for parametric roll stability

failure mode for 1700 TEU container
ship

I:\CIRC\MSC\1\MSC.1-Circ.1652.docx
MSC.1/Circ.1652
Annex, page 280

GM
GM

Figure 1.2.3. Results of assessment with respect to level 1 (left) and level 2 (right)
vulnerability criteria for pure loss of stability failure mode for 1700 TEU container ship

GM
GM

Figure 1.2.4 Results of application of level 1 (left) and level 2 (right) vulnerability
criteria for dead ship condition stability failure mode for 1700 TEU container ship

1.2.4 Application of vulnerability criteria for excessive acceleration stability failure mode
leads to recommendations on the upper limit of GM, shown in figure 1.2.5. Level 1 leads to a
very restrictive maximum GM limit, which is slightly lifted by applying level 2 assessment. Note
that the GM values associated with results of excessive acceleration criteria represent
metacentric heights without correction for free surface.
GM

GM

Figure 1.2.5. Results of level 1 and level 2 assessment for excessive acceleration
stability failure mode for 1700 TEU container ship

1.2.5 The vulnerability assessment according to the second-generation intact stability

criteria indicates for this ship additional limitations compared to the present requirements of
part A of the 2008 IS Code and damage stability requirements of SOLAS on minimum

I:\CIRC\MSC\1\MSC.1-Circ.1652.docx
 MSC.1/Circ.1652
Annex, page 281

acceptable GM values, as well as requirements on maximum acceptable GM values,

figure 1.2.6. For draughts greater than 7.9 m, level 2 of pure loss of stability vulnerability
assessment suggests increasing the minimum GM values compared to the damage stability
requirements, whereas the maximum GM values are suggested to be limited by the
vulnerability assessment for excessive acceleration stability failure mode. It is important to note
that the GM values associated with results of excessive acceleration criteria represent
metacentric heights without correction for free surface.
GM

Figure 1.2.6. GM limits according to the 2008 IS Code part A, SOLAS damage stability
requirements and vulnerability assessment of second-generation intact stability
criteria (including maximum GM from excessive accelerations criterion) for 1700 TEU
containership

1.2.6 For comparison, full probabilistic direct stability assessment was applied for three
loading conditions with GM = 0.5, 1.2 and 1.9 m at a typical loaded draught 9.5 m, table 4.1.1
of chapter 4 of appendix 4. Table 4.2.2 of chapter 4 of appendix 4 shows the resulting
"long-term" weighted average 𝑟̄U of the upper boundaries of the 95%-confidence intervals of
the "short-term" stability failure rate, which is plotted in figure 1.2.7. The standard 2.6∙10-8 1/s
for 𝑟̄U is satisfied for GM values greater than 1.788 m, which is larger than the required
minimum GM values from the Weather Criterion and damage stability requirements. The direct
stability assessment indicates the need to use a GM value higher than the one resulting from
the level 2 vulnerability assessment for parametric roll stability failure mode for the considered
ship and draught. Since the stability failure rate at small GM values is dominated by the
parametric roll stability failure mode, operational measures may be utilized in such loading
conditions (see the examples below).

I:\CIRC\MSC\1\MSC.1-Circ.1652.docx
MSC.1/Circ.1652
Annex, page 282
rU

GM

GM, m
Figure 1.2.7. Computed conservative estimate of upper boundary 𝒓̄ 𝐔 of 95%-
confidence interval of average "long-term" stability failure rate vs. GM at
draught 9.5 m for a 1700 TEU container ship compared with acceptance
standard 2.6∙10-8 1/s and resulting acceptable GM range

1.2.7 Operational guidance was developed for loading conditions with draught 9.5 m and
GM values 0.5, 1.2 and 1.9 m by identifying unacceptable sailing conditions (v0, ), i.e. those
for which the upper boundary of 95%-confidence interval of "short-term" stability failure rate
exceeds acceptance standard 10 -6 s-1, for each sea state (Hs, Tz) in the North Atlantic wave
scatter table. Figure 1.2.8 shows the resulting computed "long-term" weighted average 𝑟̄U of
the upper boundaries of the 95%-confidence intervals of the "short-term" stability failure rate
with and without operational guidance, together with the operability resulting from the use of
operational guidance, depending on GM. Since operability exceeds 0.8 for all considered
loading conditions, operational guidance is an acceptable option for all of them; note that the
upper boundary 𝑟̄U of the 95%-confidence interval of the average "long-term" stability failure
rate reduces, due to the use of operational guidance, below the standard of the full probabilistic
assessment 2.6∙10-8 1/s for all considered loading conditions.
rU

rU

rU

rU

GM,m
Figure 1.2.8. Computed "long-term" weighted average of upper boundaries of 95%-
confidence interval of "short-term" stability failure rate (with and without operational
guidance) and operability vs. GM at draught 9.5 m for 1700 TEU container ship

I:\CIRC\MSC\1\MSC.1-Circ.1652.docx
 MSC.1/Circ.1652
Annex, page 283

1.2.8 Examples of operational limitations related to areas or routes and season concern the
same three loading conditions for the sample operational routes and seasons in table 2.2.1,
chapter 2 of appendix 5. Table 1.2.1 shows the computed "long-term" weighted average 𝑟̄U of
the upper boundaries of the 95%-confidence intervals of the "short-term" stability failure rate
for unrestricted operation (area 1) and specific routes and seasons (areas 2 to 6); the red
colour indicates unacceptable loading conditions. Although the stability failure rate decreases
compared to unrestricted service, this reduction is insufficient to render loading conditions
LC01 and LC02 acceptable.

Table 1.2.1. "Long-term" weighted average 𝒓̄ 𝐔 of upper boundaries of 95%-confidence

intervals of "short-term" stability failure rate for 1700 TEU container ship; areas or
routes and seasons per table 2.2.1, chapter 2 of appendix 5
Areas or routes and seasons
LC GM, m
1 2 3 4 5 6
01 0.5 5.362e-5 2.881e-5 1.117e-5 1.323e-5 1.118e-5 1.068e-5
02 1.2 1.432e-6 1.950e-6 1.076e-6 1.226e-6 1.058e-6 7.199e-7
03 1.9 1.198e-8 5.334e-8 3.686e-8 4.588e-8 3.567e-8 1.555e-8

1.2.9 Operational limitations related to maximum significant wave height were developed
for the same three loading conditions for the North Atlantic wave scatter table, limited by a
systematically varied maximum significant wave height with a step 1.0 m. Figure 1.2.9 shows
the resulting computed "long-term" weighted average 𝑟̄U of the upper boundaries of the 95%-
confidence intervals of the "short-term" stability failure rate and operability depending on the
maximum significant wave height, and table 1.2.2 shows the significant wave height which
corresponds to 𝑟̄U matching the required standard 2.6∙10 -8 1/s, together with the operability
corresponding to this wave height. Note that due to the use of the operational limitations related
to maximum significant wave height, loading condition LC02 becomes acceptable (whereas
loading condition LC01 remains unacceptable), whereas using operational guidance renders
both loading conditions LC01 and LC02 acceptable, and that in the same wave climate,
probabilistic operational guidance provides significantly greater operability than operational
limitations related to maximum significant wave height, i.e. the latter are less efficient than
operational guidance.

rU

rU
rU

rU

Figure 1.2.9. Computed "long-term" weighted average 𝒓̄ 𝐔 of upper boundaries of 95%-

confidence intervals of "short-term" stability failure rate (left y-axis, black solid line)
and operability (right y-axis, blue dashed line) vs. maximum significant wave height
(x-axis) in North Atlantic wave climate for 1700 TEU container ship

Table 1.2.2. Maximum significant wave height at which "long-term" weighted average
𝒓̄ 𝐔 of upper boundaries of 95%-confidence intervals of "short-term" stability failure

I:\CIRC\MSC\1\MSC.1-Circ.1652.docx
MSC.1/Circ.1652
Annex, page 284

rate satisfies acceptance standard 2.6∙10-8 1/s and corresponding operability in North
Atlantic wave climate
Loading condition 01 02 03
Metacentric height, m 0.5 1.2 1.9
Maximum significant wave height, m 2.814 4.746 unlimited
Corresponding operability 0.557 0.839 1.0

___________

I:\CIRC\MSC\1\MSC.1-Circ.1652.docx