Predictive and reliability o o_sulaiman _revision 2_

Jurnal Mekanikal
June 2011, No. 32, 38- 60

PREDICTIVE AND RELIABILITY BASED COLLISION RISK
AVERSION MODEL FOR INLAND WATERWAYS: THE CASE
FREQUENCY ESTIMATION FOR MALAYSIAN LANGAT RIVER

O.O. Sulaiman*1, Ab.Saman Abd. Kader2, A.H. Saharuddin1 and
Adi Maimun Abdul Malik2
1
Faculty of Maritime Studies and Marine Science, Universiti Malaysia Terengganu
21030 Kuala Terengganu, Terengganu, Malaysia
2
Marine Technology Department,Universiti Teknologi Malaysia,
81310 Skudai, Johor, Malaysia

ABSTRACT

Collisions of commercial ships cover the largest part of accidents scenario in waterways.
Waterways accidents expose vessel owners and operators, as well as the public to risk.
They attract possibility of losses such as vessel cargo damage, injuries, loss of life,
environmental damage, and obstruction of waterways. Collision risk is a product of the
probability of the physical event its occurrence as well as losses of various nature
including economic losses. Environmental problem and need for system reliability call for
innovative methods and tools to assess and analyze extreme operational, accidental and
catastrophic scenarios as well as accounting for the human element, and integrate these
into a design environments part of design objectives. This paper discusses modeling of
waterways collision risk frequency in waterways. The analysis consider mainly the
waterways dimensions and other related variables of risk factors like operator skill, vessel
characteristics, traffic characteristics, topographic, environmental difficulty of the transit,
and quality of operator's information in transit which are required for decision support
related to efficient, reliable and sustainable waterways developments. 5.3 accidents in 10,
000 years is observed for Langat River, this considered acceptable in maritime and
offshore industry, but for a channel using less number of expected traffic, it could be
considered high. Providing safety facilities like traffic separation, vessel traffic
management could restore maximize sustainable use of the channel.

Keywords: Collision, risk, reliability, frequency, inland waterways, environmental
prevention

1.0 INTRODUCTION

Collision in waterways falls under high consequence incidents, collision data may be
imperfect or inconstant, making it difficult to account for dynamic issues associated with
vessels and waterways requirement. Accounting for these lapses necessitated need to base
collision analysis on hybrid use of deterministic, probabilistic or simulation methods
depending on the availability of a data. Developing sustainable inland water transportation
(IWT) requires transit risk analyses of waterways components and relationship between
factors such as environmental conditions, vessel characteristics, operators' information

*
Corresponding author : o.sulaiman@umt.edu.my

38

Jurnal Mekanikal, June 2011

about the waterway, as well as the incidence of groundings and collisions, using available
data. Whatever information is available is useful for risk and reliability based decision
work of accidents rate of occurrence, consequence and mitigation [1, 7]. Risk and
reliability based design entails the systematic integration of risk analysis in the design
process targeting system risk prevention, reduction that meet high level goal and leave
allowance for integrated components of the system including environment that will
facilitate and support a holistic approach for reliable and sustainable waterways appropriate
and require trade-offs and advance decision-making leading to optimal design solutions.
Frequency estimation work on channel lead to fundamental sustainable model of
transit risk that include factors such as traffic type and density, navigational aid
configuration, channel design and waterway configuration and classification. For cases
where there are insufficient historical record to support their inclusion, more
comprehensive models of transit risk will have to rely on integral use of hybrid of
deterministic, probabilistic, stochastic method whose result could further be simulated or
employ expert judgment to optimize deduced result [2]. Risk based collision model are
derivative for improvement of maritime accident data collection, preservation and limit
acceptability using information relating to the following:

i. Ports for entering incidents, traffic characteristics, frequency of accident, "barge
train" movements as well as individual barges
ii. Vessel characteristics, record data on actual draft and trim, presence and use of
tugs, presence of pilots.
iii. Environmental condition, wind speed and direction, visibility, water level, current
speed and direction, Tide Forecast Error, Real-time Environmental Information
etc.
iv. Types of cargo and vessel movements.
v. Operator skill, quality of operator's information.
vi. Uncertainty in surveys/charts, geographical distribution of transit, topographic
difficulty of the transit, improve temporal resolution (transits by day or hour),
eliminate/correct erroneous and duplicate entries (e.g. location information).

This paper describes frequency analysis of risk based model, where accident frequency are
determined and matched with waterway variables and parameter. The result hopes to
contribute to decision support for development and regulation of inland water
transportation.

2.0 BACKGROUND

The study area is Langat River, 220m long navigable inland water that has been under
utilized. Personal communication and river cruise survey revealed that collision remain the
main threat of the waterways despite less traffic in the waterways. This make the case to
establish risk and reliability based model for collision aversion for sustainable development
of the waterways a necessity. Data related to historical accidents, transits, and
environmental conditions were collected. Accident data are quite few, this is inherits to
most water ways and that make probabilistic methods the best preliminary method to
analyze the risk which can be optimized through expert rating and simulation methods as
required.

39


Figure 1: Langat map

Barge and tug of capacity 5000T and 2000T are currently plying this waterway at
draft of 9 and 15m respectively. Collisions (including contact between two vessels and
between a vessel and a fixed structure), causes of collision linked to navigation system
failure, mechanical failure and vessel motion failure are considered in this work towards
design of safe and reliable the river for transportation. Safety associated with small craft is
not taken into account. The next section describes the relevant information relating to
channel, vessel and environment employed in the risk process. Lack of information about
the distribution of transits during the year, the joint distribution of ship size, flag particular,
environmental conditions become main derivative from probabilistic estimation. In total
risk management system of various methods is used according to result expectation and
performance contribution. The study use Langat River to a case study to test the model,
because it is a big River with big potential that is underutilized. The testing of the model on
Langat could help decision support for its development and regulation in future [3, 7]. The
model described is suitable for preventive safety reliability decision for new water way
development. When it is safe the environment is preserved and protected.

3.0 BASELINE DATA

Vessel movement, port call consists of two transits in Langat River: one into and one out of
the port. Safe transit data consider the same barge type and size for risk analysis are
considered. The required radius of curvature at bends for 5000 DWT, Towed barge Length
= Barge Length + Tug Length + Tow Line, R> (4-6) length of barge train to meet the
navigation requirement (PIANC, 2007). Water level Mean, water level = 40cm seasonal
variation, Existing coastal environmental current. Coastal current, Average Speed in spring
tide 0.4 -1.2 m/s, Avg. Speed in Neap 0.2 - 1.0 m/s is considered part of environmental
parameters [4, 10].
Table 2: River Langat tributary
Channel Parameters
Width Depth
Maneuvering lane Draught
Vessel clearance Trim
Bank suction Squat
Wind effect Exposure allowance
Current effect mentFresh water adjust
Channel with bends Allowance maneuvering
Navigation aids Overdepth allowance
Pilot Depth transition
Tugs Tidal allowance

40


Table 3: River width and depth parameters

Design Approach channel
parameter
Straight Bend
98m 120m
3-6m 3-6m
Side slope 10H:1V 10H:1V
Estuarine 135.7km North (44.2km) South (9.9km)

Maneuvering lane
Bank clearance

Bank clearance
Allowance

Allowance

Channel width : One way Traffic
Straight channel = 98m, Bend = 120m, Depth= 6m

Figure 2: Channel width parameter Figure 3: Channel straightening and alignment

Table 4: Vessel requirement: Barge parameter

Barge parameter
2000 tons 5000 tons
Length (m) 67.3 76.2
Beam (m) 18.3 21.3
Depth (m) 3.7 4.9
Draft (m) 2.9 4.0

Table 5: Vessel requirement: Tug parameter

Tugs parameter
2000 tons 5000 tons
Length (m) 23.8 23.8
Beam (m) 7.8 7.8
Depth (m) 3.5 3.5
Draft (m) 2.8 2.8
Horse Power (hp) 1200 1200

3.1 Data Collection Limitation
Limitations in data collection poised hybrid combinatory use of historical, first principle, or
deterministic and stochastic analysis, future data collection effort can open opportunity for
improvement in validation analysis as well as understanding of accident risk. In this case
the data is good enough data to model a predictive and state space analysis model of
frequency of occurrence in the channel. Major data problems are as follows [12]

41


i. Vessel Casualty Data: Inherent problem with causality data have missing entries,
duplicate entries, and inaccuracies.
ii. Environmental Data: Limitations are associated with potential change in real-time
oceanographic data systems.
iii. Port-Specific Data: information about safe transits counts categorization by flag,
vessel type, vessel size, with tug escort and piloting information, taken at hourly by
authority.
iv. Surveys and Chart Data: it is important to compare conventional cartographic
uncertainty and with new technology to cover additional uncertainties.

4.0 SAFETY AND ENVIRONMENTAL RISK FOR IWT

Risk and reliability based model aim to develop innovative methods and tools to assess
operational, accidental and catastrophic scenarios. It requires accounting for the human
element, and integrates them as required into the design environment. Risk based design
entails the systematic integration of risk analysis in the design process. It target safety and
environment risk prevention and reduction as a design objective. To pursue this activity
effectively, an integrated design environment to facilitate and support a holistic risk
approach to ship and channel design is needed. Total risk approaches enable appropriate
trade off for advanced sustainable decision making. Waterways accident falls under
scenario of collision, fire and explosion, flooding, grounding.
Risk based design entails the systematic risk analysis in the design process
targeting risk preventive reduction. It facilitates support for total risk approach to ship and
waterways design. Integrated risk based system design requires the availability of tools to
predict the safety, performance and system components as well as integration and
hybridisation of safety element and system lifecycle phases. Therefore, it becomes
imperative to develop, refine, verify, validate reliable model through effective methods and
tools. The risk process begins with definition of risk which stands for the measure of the
frequency and severity of consequence of an unwanted event (damage, energy, oil spill).
Frequency at which potential undesirable event occurs is expressed as events per unit time,
often per year. The frequency can be determined from historical data. However, it is quite
inherent that event that don’t happen often attract severe consequence and such event are
better analyzed through risk based and reliability model. Figure 3.2 shows main
components of risk based design for IWT. Risk is defined as product of probability of event
occurrence and its consequence.

Risk (R) = Probability (P) x Consequence (C) (1)

Incidents are unwanted events that may or may not result to accidents. Necessary
measures should be taken according to magnitude of event and required speed of response
should be given. Accidents are unwanted events that have either immediate or delayed
consequences. Immediate consequences variables include injuries, loss of life, property
damage, and persons in peril. Point form consequences variables could result to further loss
of life, environmental damage and financial costs. The earlier stage of the process involves
finding the cause of risk, level of impact, destination and putting a barrier by all mean in the
pathway. Risk work process targets the following:

i Cause of risk and risk assessment, this involve system description, identifying the
risk associated with the system, assessing them and organising them in degree or
matrix. IWT risk can be as a result of the following: (i) Root cause, (ii) Immediate
cause, (iii) Situation causal factor, (iv) Organization causal factor.
ii Risk analysis and reduction process, this involve analytic work through deterministic
and probabilistic method that strengthen can reliability in system. Reduction process

42


that targets initial risk reduction at design stage, risk reduction after design in
operation and separate analysis for residual risk for uncertainty as well as human
reliability factor.

Uncertainty risk in complex systems can have its roots in a number of factors
ranging from performance, new technology usage, human error as well as organizational
cultures. They may support risk taking, or fail to sufficiently encourage risk aversion. To
deal with difficulties of uncertainty risk migration in marine system dynamic, risk analysis
models can be used to capture the system complex issues, as well as the patterns of risk
migration. Historical analyses of system performance are important to establish system
performance benchmarks that can identify patterns of triggering events, this may require
long periods of time to develop and detect. Assessments of the role of human and
organizational error, and its impact on levels of risk in the system, are critical in distributed,
large scale dynamic systems like IWT couple with associated limited physical oversight.
Effective risk assessments and analysis required three parts highlighted in the relation
below.

Risk modeling = Framework + Models + Process (2)

Reliability based verification and validation of system in risk analysis should be
followed with creation of database and identification of novel technologies required for
implementation of sustainable system.

4.1 Risk Framework
Risk framework provides system description, risk identification, criticality, ranking,
impact, possible mitigation and high level objective to provide system with what will make
it reliable. The framework development involves risk identification which requires
developing understanding the manner in which accidents, their initiating events and their
consequences occur. This includes assessment of representation of system and all linkage
associated risk related to system functionality and regulatory impact (See Figure 4 a and b)

(a) (b)

Figure 4: IMO Risk framework

Risk framework should be developed to provide effective and sound risk
assessment and analysis. The process requires accuracy, balance, and information that meet
high scientific standards of measurement. The information should meet requirement to get
the science right and getting the right science. The process requires targeting interest of
stakeholder including members of the port and waterway community, public officials,

43


regulators and scientists. Transparency and community participation helps ask the right
questions of the science and remain important input to the risk process, it help checks the
plausibility of assumptions and ensures that synthesis is both balanced and informative.
Employment of quantitative analysis with required insertion of scientific and natural
requirements provide analytical process to estimate risk levels, and evaluating whether
various measures for risk are reduction are effective.

4.2 Safety and Environmental Risk and Reliability Model (SERM)
There is various risk and reliability tools available for risk based methods that fall under
quantitative and qualitative analysis. Figure 5 show the analysis risk model flowchart
choice of best methods for reliability objective depends on data availability, system type
and purpose. However employment of hybrid of methods of selected tool can always give
the best of what is expect of system reliability and reduced risk.

(a) (b)

Figure 5: IMO Risk framework

4.3 SERM Process
SERM intend to address risks over the entire life of the complex system like IWT system
where the risks are high or the potential for risk reduction is greatest. SERM address
quantitatively, accident frequency and consequence of IWT. Other risk and reliability
components include human reliability assessment which is recommended to be carried out
separately as part of integrated risk process. Other waterways and vessel requirement
factors that are considered in SERM model are: (i) Construction (ii), Towing operations
and abandonment of ship, (iii) Installation, hook-up and commissioning, and (iv)
Development and major modifications
Integrated risk based method combined various technique as required in a process.
Table 2 shows available risk based design for techniques. This can be applied for each level
of risk. Each level can be complimented by applying causal analysis (system linkage),

44


expert analysis (expert rating), and organizational analysis (Community participation) in
the risk process. Figure 7 shows stakes holder that should be considered in risk process.
From Figure 2, the method use is risk analysis that involves frequency analysis where the
system is modeled with hybrid of deterministic, probabilistic and stochastic process.
Technically, the process of risk and reliability study involves the following four areas: (i)
System definition of high goal objective, (ii) Qualitative hazard identification and
assessment, (iii), Quantitative hazard frequency and consequence analysis, (iv) Risk
acceptability, sustainability and evaluation.

Table 6: Risk based design techniques

Process Suitable techniques
HAZID HAZOP, What if analysis, FMEA, FMECA
Risk analysis Frequency, consequence, FTA, ETA
Risk evaluation Influence diagram, decision analysis
Regulatory, economic, environmental, function elements
Risk control option
matching and iteration
Cost benefit analysis ICAF, Net Benefit
Human reliability Simulation/ probabilistic
Uncertainty Simulation/probabilistic
Risk monitoring Simulation/ probabilistic

The process of risk work can further be broken down into the following elements:

i. Definition and problem identification
ii. Hazard and consequence identification
iii. Analysing the likelihood’s of occurrence
iv. Analyzing consequences
v. Evaluation of uncertainty
vi. Risk control option (RCO) and risk control measure (RCM
vii. Sustainability of (cost safety, environment, injury, fatality, damage to structure,
environment) and risk acceptability criteria
viii. Reliability based model verification and validation: statistical software,
triangulation, iteration.
ix. Recommendation for implementation: Implement, establishing performance
standards to verify that the arrangements are working satisfactorily and continuous
monitoring, reviewing and auditing the arrangements

Employment of these benefit provide a rational. Formal environmental protection
structure and process for decision support guidance and monitoring about safety issues.
The scope of sustainable risk based design under consideration involves stochastic,
analytic and predictive process work leading to avoidance the harms in waterways. Figure
8 shows block diagram of SERM components for IWT. Safety and Environmental Risk and
Reliability Model (SERM) for IWT required having clear definition of the following
issues:

45


i. Personnel and attendance
ii. Identify activities
iii. Vessel accidents including passing vessel accident, crossing , random
iv. Vessel location and waterway geography on station and in transit to shore.
v. Impairment of safety functions through determination of likelihood of loss of key
safety functions lifeboats, propulsion temporary refuge being made ineffectiveness
by an accident.
vi. Risk of fatalities, hazard or loss of life through measure of harm to people and
sickness.
vii. Property damage through estimation of the cost of clean-up and property
replacement.
viii. Business interruption through estimation of cost of delays in production.
ix. Environmental pollution may be measured as quantities of oil spilled onto the
shore, or as likelihood’s of defined categories of environmental impact or damage
to infrastructures.

Allowance should be made to introduce new issue defining the boundary in the
port from time to time. The choice of appropriate types of risk tool required for the model
depend on the objectives, criteria and parameter that are to be used. Many offshore risk
based design model consider loss of life or impairment of safety functions. There is also
much focus on comprehensive evaluation of acceptability and cost benefit that address all
the risk components. Figure 9 shows the risk and reliability model combined process
diagram. The analysis is a purely technical risk analysis. When the frequencies and
consequences of each modelled event have been estimated, they can be combined to form
measures of overall risk including damage, loss of life or propulsion, oil spill. Various
forms of risk presentation may be used. Risk to life is often expressed in two
complementary forms. The risk experienced by an individual person and societal risk. The
risk experienced by the whole group of people exposed to the hazard (damage or oil spill).
Accident and incident are required to be prevented not to happen at all. The
consequence of no safety is a result of compromise to safety leading to unforgettable loses
and environmental catastrophic. Past engineering work has involved dealing with accident
issues in reactive manner. System failure and unbearable environmental problem call for
new proactive ways that account for equity requirement for human, technology and
environment interaction. The whole risk assessment and analysis process starts with system
description, functionality and regulatory determination and this is followed by analysis of:
(i)Fact gathering for understanding of contribution factor (ii), Fact analysis of check
consistency of accident history, (iii) Conclusion drawing about causation and contributing
factor, (iv) Countermeasure and recommendation for prevention of accident

Most risk based methods define risk as:

Risk = Probability (Pa) x Consequence (Ca) (3)

or in a more elaborate expression risk can be defined as:

Risk = Threat x Vulnerability x {direct (short-term) consequences + (broad) (4)
Consequences}

46


In risk analysis, serenity and probability of adverse consequence hazard are deal
with through systematic process that quantitatively measure , perceive risk and value of
system using input from all concerned waterway users and experts.
Risk can also be expressed as:

Risk = Hazard x Exposure (5)

Where hazard is anything that can cause harm (e.g. chemicals, electricity, Natural
disasters), while exposure is an estimate on probability that certain toxicity will be realized.
Severity may be measured by No. of people affected, monetary loss, equipment downtime
and area affected by nature of credible accident. Risk management is the evaluation of
alternative risk reduction measures and the implementation of those that appear cost
effective where:

Zero discharge or negative damage = Zero risk (6)

The risk and reliability model subsystem in this thesis focus on the following identified
four risks assessment and analysis application areas that cover hybrid use of technique
ranging from qualitative to qualitative analysis (John, 2000): (i) Failure Modes
Identification Qualitative Approaches, (ii) Index Prioritisation Approaches, (iii) Portfolio
Risk Assessment Approaches, and (iv) Detailed Quantitative Risk Assessment
Approaches.

5.0 COLLISIONS RISK MODELLING

Collision in waterways is considered low frequency and high consequence events that have
associative uncertainty characteristics / component of dynamic and complex physical
system. This makes risk and reliability analysis the modest methods to deal with
uncertainties that comes with complex systems. Employment of hybrid deterministic,
probabilistic and stochastic method can help break the barriers associated with transit
numbers data and other limitation. Conventionally, risk analysis work often deal with
accident occurrence, while the consequence is rarely investigated, addressing frequency
and consequence analyze can give clear cuts for reliable objectives. Risk and reliability
based design can be model by conducting the analysis of following elements of risk process
[13, 15]:

i. Risk identification
ii. Risk analyses
iii. Damage estimation
iv. Priotization of risk level
v. Mitigation
vi. Repriotization of exposure category: mitigate risk or consequence of events that
meet ALARP principle.
vii. Reassess high risk events for monitoring and control plans.
viii. Recommendation, implementation, continuous monitoring and improvement.

Collision is likely to be caused by the following factors shown in Figure 7 derived
from fault three analyses from RELEX software. The relex software is based on fault three
analysis where consequence of causal events are add up through logic gate to give
minimum cut set probability that trigger the event. It is more effective for subsystem
analysis.

47


P (collision) = P (propulsion failure) + P (loss of navigation failure) + (7)
P (Loss of vessel motion)

There is also causes are mostly as a result of causes from external sources like
small craft, are cause of cause, cause from other uncertainty including human error may
attract separate subsystem analysis.

5.1 Collision Data
Collision data are drawn from relevant marine administrator; there is expectation that most
data gaps can be covered by the probability estimations. The Langat River work model risk
through systemic analysis procedures for sustainable inland waterways transportation. It
determine the probability of failure or occurrence, risk ranking, damage estimation, high
risk to life safety, cost benefit analyze, sustainability and acceptability criteria [5, 14]. The
study analyze causal accidental relating to navigational, mechanical failure and human
error and ignored those identified as intentional for barge and tugs of 5000T and 2000T
having respective drift of draft greater than 9 to 15m. Table 7, 8 and 9 shows some of the
annual traffic summary, collision and the consequences on Langat. Seasonal trends can be
stochastically modeled from probabilistic result, environmental condition and traffic
volume fluctuation is also considered negligible. For visibility, navigation is considered to
be more risky at night than day time, the analysis follow generic assumption for evenly safe
distribution evenly during day and night.

Figure 6: Collision contributing factors Figure 7: Tugs puling barge in Langat

A critical review of risk assessment methodologies applicable to marine systems
reiterate that the absence of data should not be used as an excuse for not taking an
advantage of the added knowledge that risk assessment can provide on complex systems
[6]. Approximation of the risks associated with the system can provide a definition of data
requirements. The treatment of uncertainty in the analysis is important, and the limitations
of the analysis must be understood. However, data management system and better approach
can always accommodate little data or no data. Table 6 shown models that have been used
design of system based on risks in marine industry.
IMO and Sirkar et al (1997) methods lack assessment of the likelihood of the event,
likewise other model lack employment of stochastic method whose result could cover
uncertainties associated with dynamic components of channel and ship failure from causal
factors like navigational equipment, training and traffic control [14]. Therefore,
combination of stochastic, statistical and reliability method based on combination of
probabilistic, goal based, formal safety assessment, deterministic methods and fuzzy
method using historical data of waterways, vessel environmental, first principle
deterministic and traffic data can deliver best outcome for predictive, sustainable, efficient
and reliable model for complex and dynamic system like inland water transportation. The
general hypothesis behind assessing physical risk model is that the probability of an

48


accident on a particular transit depends on a set of risk variables require for analysis of
prospective reliable design. Figure 8 shows traffic data utilized in the model. Most of the
method above used historical data, the novel method in this paper used limited data of
traffic used to model the physics of the system, the transfer function and stochastically
project accident frequency. The projection is generic and can be used for any waterways
and it consider random collision not which is not considered by previous model.

Table 6 : Previous risk work
Model Application Drawback
Brown et al (1996) Environmental Performance of
Tankers
Consequences of collisions and Difficulties on
Sirkar et al (1997) groundings quantifying consequence
metrics
Brown and Hybrid use of risk assessment, Oil spill assessment
Amrozowicz probabilistic simulation and a spill limited to use of fault
consequence assessment model three

Sirkar et al (1997) Monte Carlo technique to estimate Lack of cost data
damage and spill cost analysis for
environmental damage
IMO (IMO 13F Pollution prevention index from Lack (Sirkar et al, 1997)).
1995) probability distributions damage and rational
oil spill.
Research Council Alternative rational approach to Lack employment of
Committee(1999) measuring impact of oil spills stochastic probabilistic
methods
Prince William The most complete risk assessment Lack of logical risk
Sound,Alaska, assessment framework
(PWS (1996) (NRC,1998))
Volpe National Accident probabilities using Lack employment of
Transportation statistics and expert opinion. stochastic methods
Center (1997)).
Puget Sound Area Simulation or on expert opinion for Clean up cost and
(USCG (1999). cost benefit analysis environmental damage
omission

Table 7: Tug boat and vessel activities along river for 2008
Jetty 3 nos.
Daily 9 times.
Weekly 63 times.
Monthly 252 times.
Annually 3024 times.

Table 8: Vessel traffic
Total number of barge Time Traffic
12 Every day (24 hrs.)
6 (every 4 hrs) Incoming
6 (every 4 hrs) outgoing

49


Table 9: Common to traffic

All Speed 2 – 3 knots
Traffic All single way traffic
Lay -bys Proposed four locations for Lay-bys

5.2 Traffic Frequency Estimation Modeling
Traffic density of meeting ship
Nm
Traffic density of meeting ship: ρ = hips/݉ଶ (8)
ν .τ .W
Where Nm is number of ships frequenting the channel, v is speed of the ship, T= time of
traffic activities per annum and W is width of the channel.

Figure 8: 5000 barge data and Langat waterway

5.3 Analysis of Present Situation
Traffic situation: Below are representation of various collision situations for head- on,
overtaking and crossing (angle) collision scenario (see Figure 9).
Where: B1 = mean beam of meeting ship (m), V1 = mean speed of meeting ship (knots), B2
= beam of subject ship (m), V2 = speed of subject ship (knots), Nm = arrival frequency of
meeting ships (ship/time), D= relative sailing distance.

Expected number of collision Ni= 9.6.B.D.ρ s 1/passage. (9)

(a) overtaking (b) passing cases (c) Random
Figure 9 Collision situations

50


Table 10 and 11 show relevant data from previous analysis used for approximation.

Table 10: Expression for collision situation [8, 11]
Expression Head – on Overtaking Random
( B1 + B2 ) (V1 + V2 N 4
Basic 4 x B X D X ρS
W . V ⋅ V . D. Nm
1 2
Ni = τ .V . ( π L+2*B)

( B1 + B2 ) V1 + V2
Standardised 4 x B X D X ρS . V ⋅V . D. Nm 9.6.D. ρn B
W 1 2
Relative 1 1 2.4

Approximations: L=6B, D=W, Ni= Pi (10)

Necessary period for ship to pass the fairway T=D/v = 3000/3 = 1000 sec (11)

Table 11: Failure per nautical mile and failure per passage for different waterways [8]

Fairway µ c (failure per nautical Pc(failure per passage
mile or hour) or encounter)
UK 2.5 x10-5 1.x10-4
US 1.5 x10-5 1.4.x10-5
Japan 3.0 x10-5 1.3.x10-5

Therefore average Pc and µ c = 2.5 x 10-5 for random (12)

Probability of loosing navigation control within the fairway

Pc = µ c ⋅T failure / passage (13)

Probability of collision Pa= (Pi. Pc collision / passage) (14)

Collision per annual (Na) = Pa. Nm Collision per year (15)

In the frequency analysis, the annual frequency of each failure case is estimated.
Separate frequencies are estimated for each operating phase as required. In modelling the
development, consequences and impact of the events, each failure case is split into various
possible outcomes. the outcomes are the end events on an event tree or chain of event trees.
Each outcome probability is estimated by combining the probabilities for appropriate
branches of the event tree. The outcome frequency (Fo ) is then:

FO = Fe ∏ Pb (16)

51


Where, Fe is failure frequency, Pb probability of one segment. Not all possible outcomes are
modeled. Representative scenarios are selected for modeling, and the scenario frequency is
taken as:

FS = ∑F
outcomes
O (17)

Failure per nautical mile and failure per passage can be selected from previous
representative work. Necessary period for ship to pass the fairway T=D/v = 3000/3 = 1000
sec. The result of accident frequency (Fa) can be compare with acceptability criteria for
maritime industry. If it is two high the system could be recommended to implement TSS. If
the result is high TSS can be model to see possible reduction due to its
implementation.Table 12 shows frequency risk acceptability criteria for maritime and
offshore industry.

Table 12: Frequency acceptability criteria

Frequency classes Quantification
Very unlikely once per 1000 year or more likely
Remote once per 100- 1000 year
Occasional once per 10- 100 year
Probable once per 1- 10 years
Frequent more often than once per year

5.4 Frequency Analysis Result
This result indicates that the collision in Langat is not risk on ALARP graph. Accident per
year of 5.3E-5 is observed for current 3 number of vessel operating at speed of 3 knot. But
physical observation revealed that there is significant and exception increase in collision
that needs to be address for a channel with less traffic density. It is also observed from the
plot of frequency Vs speed that when traffic density is changing traffic density of 5 and 6
and speed up to 5 considered to be cause high risk of accident frequency in the waterway
(See Figure 10).

Fa vs V Fa vs V
V=+1 D=W=3B B=21.3 M=+500 Nm=x
V=+1 D=W=3B B=21.3 M=+500 Nm=x
1.0E-05
1.1E-03
Nm2 9.5E-06

9.0E-06
1.0E-03 Nm3
8.5E-06

9.0E-04 Nm4 8.0E-06

7.5E-06
Fa :-Expected number of collIsion per year

Nm5
8.0E-04 7.0E-06

Nm6 6.5E-06
7.0E-04
6.0E-06

5.5E-06
6.0E-04 Nm2
5.0E-06
Nm3
5.0E-04 4.5E-06
Nm4
4.0E-06
4.0E-04 3.5E-06 Nm5

3.0E-06 Nm6
3.0E-04
2.5E-06

2.0E-06
2.0E-04
1.5E-06

1.0E-04 1.0E-06

5.0E-07

0.0E+00 0.0E+00
1 4 7 10 13 16 19 22 25 28 31 34 37 40 43 46 49 1 4 7 10 13 16 19 22 25 28 31 34 37 40 43 46 49
V :-Velocity
V :-Velocity
(a) (b)
Figure 10: Accident frequency Vs at changing number of ship

52


Figure 11 shows accident frequency at changing width and beam of the channel. Risk is
acceptable for accident per 10, 000 year, if proposed maintenance of channel improvement
plan is implemented. Beam and wide have linear relationship (3B=W).
Fa vs B Fa vs W
1.5E-05
2.0E-04 1.4E-05 v=10
1.9E-04
1.3E-05 v=20
1.8E-04
1.7E-04 1.2E-05 v=30
1.6E-04
1.1E-05


1.5E-04 v=40
1.4E-04 1.0E-05
v=50
1.3E-04
9.0E-06
1.2E-04
1.1E-04 8.0E-06
1.0E-04
7.0E-06
9.0E-05
8.0E-05 6.0E-06
7.0E-05
5.0E-06
6.0E-05
5.0E-05 4.0E-06
4.0E-05
3.0E-06
3.0E-05
2.0E-05 2.0E-06
B=16.8+0.5
1.0E-05
1.0E-06
0.0E+00
17 18 20 21 23 24 26 27 29 30 32 33 35 36 38 39 41 0.0E+00
B:- Beam

W :- Width

(a) (b)

Figure 11: Accident frequency Vs beam and width of the channel

The maximum speed is round 10 knot for width of 64m and probability of 1/1000 years,
other speed above this are intolerable. As width of the channel decrease there is higher risk
-> Accident frequency probability increase. The maximum width considered for Langat
River is 64; this width is considered too small and risky for the channel for accident per
1000 years. Different speed should be advised to ship for such situation. Width of channel
can change as a result of erosion. Increasing channel width to 250m could allow speed of
20 knot at acceptable Fa (Na) of 1x10E-4. Ship operating at Langat at 3 knot at River
Langat, is considered not high risk for accident per 100, 000 years. The regression equation
for the trend is represented by y is 2E-08x + 1E-05 @ R² is 1. Similar trend is observe for
Figure 12b, the beam and width are related according to PIANC W=3B AND L=6B. Table
14 shows regression equations for the frequency analysis.
Figure 12(a) and (b) shows cross plotting of the channel variable, both plots are the
same; the defense is that Figure 12(b) is logged because of large number shows the risk
level for all channel parameters variables (speed, width, number of ships, and beam of ship).
It is observed that the maximum of ship can up to 4, at the point where speed and Number
of ship curves meet, provided all channel and vessel safety parameters are in place.
Combin ed graph

1
13
15
17
19
21
23
25
27
29
31
33
35
37
39
41
43
45
47
49
1
3
5
7
9
11

V change
0.1 W change
NM change
B change
0.01
Accident frequency

0.001

0.0001

0.00001

0.000001

0.0000001
Speed

(a)

53


Fa vs V

1.0E+00
1 4 7 10 13 16 19 22 25 28 31 34 37 40 43 46 49

1.0E-01

1.0E-02

1.0E-03

1.0E-04

1.0E-05

1.0E-06 B change (16.8+.5)
W change (50.4+1.5)
Nm (+1)
V (+1)
1.0E-07
V :-Velocity

(b)

Figure 12: Cross plotting of channel variables (speed, width, number of ships, and beam of
ship)

Table 13: Regression equation for Frequency analysis
Fa @Nm changing y = 2E-05e-0.11x R² = 0.826 Exponential
Speed
Fa @V y = 2E-05e-0.11xR² = 0.826 R² = 1 Square
Fa W y = 2E-08x + 1E-05 R² = 1 Square
Fa B y = 9E-07x + 0.000 R² = 0.999 Linear

6.0 UNCERTAINTY AND SYSTEM COMPLEXITY ANALYSIS

6.1 Subsystem Level Analysis
For total risk work the following analysis could perform separately as part of subsystem
risk level analysis include (i) power transmission (loss of propulsion), (ii) navigation (loss
of mooring function and (iii) human reliability, subsystem level analysis can be facilitated
by using frequency calculation through Fault Tree Analysis (FTA) modeling involve top
down differentiation of event to branches of member that cause them or participated in the
causal chain action and reaction. While consequence calculation can be done by using
Event Tree Analysis (ETA), where probability is assigned to causal factor leading to
certain event in the event tree structure.

6.2 Channel Complexity Analysis
Channel complexity that could be addressed in the risk and reliability work are visibility
weather, squat, bridge, river bent and human reliability. Figure 19 show channel
complexity for Langat. Poor visibility and the number of bend may increase in the risk of
and collisions. A model extracted from Dover waterway studies concluded with the
following:

Fog Collision Risk Index (FCRI) = ( P1+ VI1+ P2+ VI2+ P3 . VI3 ) (18)

54


Where: = Probability of collision per million encounters, = Fraction of time that the
visibility is in the range k, K = Visibility range: clear (>4km), Mist/Fog (200m- 4km),
Tick/dense (less than 200m). Empirically derived means to determine the relationship
between accident risk, channel complexity parameters and VTS is given by equation :

R = -0.37231 - 35297C + 16.3277N + 0.2285L -0.0004W + 0.01212H + 0.0004M (19)

For predicted VTS consequence of 100000 transit, C = 1 for an open approach area
and 0 otherwise, N = 1 for a constricted waterway and 0 otherwise, L = length of the traffic
route in statute miles, W = average waterway/channel width in yards, H = sum of total
degrees of course changes along the traffic route, M = number of vessels in the waterway
divided by L.
Barge movement creates very low wave height and thus will have insignificant
impact on river bank erosion and generation of squat event. Speed limit can be imposed by
authorities for wave height and loading complexity. Human reliability analysis is also
important to be incorporated in the channel; complexity risk work, this can be done using
questionnaire analysis or the technique of human error rate prediction THERP probabilistic
relation
m
PEA = HEPEA ∑ k =1
FPS k ⋅Wk ⋅ + C (20)

Where: PEA = Probability of error for specific action,HEPEA = Nominal operator error
probability for specific error, PSFK = numerical value of kth performance sapping factor,
WK= weight of PSFK (constant), m=number of PSF, C= Constant.

6.3 Reliability Based Validation
Reliability analysis is designed to cater for uncertainty and to provide confident on the
model. It is important for this to be carried out separately. Reliability work could include
projection for accident rate for certain number of year the following techniques:

(1) Accident mean, variance and standard deviation from normal distribution

For 10 years =>Mean ( µ ) = 10 x Na (21)

Variance ( σ ) = 10 x Na x (1-Na), Standard deviation = σ , Z = (X- µ ) / σ (22)

(2) Stochiatic process using poison distribution, Year for system to fail from binomial,
mean time to failure and poison distribution, or determination of exact period for next
accident using binomial function. Ship collisions are rare and independent random event in
time. The event can be considered as poison events where time to first occurrence is
exponentially distributed.

Fr ( N/γ , T ) = eγ. T( ) γ , ( γ.T )N). N! (23)

Binomial distribution – for event that occurs with constant probability P on each trail, the
likelihood of observing k event in N trail is binomial distribution.

Ν
L(K/N,P)= ( ) p k (1-P)N.K (24)
Ρ

55


Where average number of occurrence is NP. (3) Comparing the model behaviour apply to
other rivers of relative profile and vessel particular. (4). Triangulating analysis of sum of
probability of failure from subsystem level failure analysis. And (5) Implementation of
TSS is one of the remedies for collision risk observed and predicted in Langat; this can be
done through integration of normal distribution along width of the waterways and
subsequent implementation frequency model. And the differences in the result can reflect
improvement derived from implementation of TSS.

1
1 − 12
e 2 (x − )
(x) =
µ 2π µ (25)

1
1 − 12
e 2 (x − )
(x) =
µ 2π µ (26)

(3) Safety level and cost sustainability analysis. Figure 13 shows the best accident
frequency that is acceptable,. Ct is is the total cost, Co is the cost of damage, and Cc is the
cost of repair.
Co Cc & Ct vs Fa
250000000

200000000

150000000
ot
Cs

Co
Cc
100000000
Ct

50000000

0
3. -05

3. -05

3. -05

4. -05

4. -05

5. -05

5. -05

6. -05

6. -05

7. -05

7. -05

8. -05

8. -05

9. -05

1. -05

1. -04

1. -04

1. -04

1. -04

1. -04

1. -04

1. -04

1. -04

1. -04

4
-0
E

E

E

E

E

E

E

E

E

E

E

E

E

E

E

E

E

E

E

E

E

E

E

E

E
80

14

51

89

29

71

16

62

10

60

13

67

23

81

41

00

07

13

20

27

34

42

49

57

65
2.

Figure 13: Risk cost benefit analysis

6.4 Validation Result
Validation and reliability analysis of the model yield the following result. Figure 14 shows
accident frequency residual plot from Minitab is shown with good fitness. Figure 15:
Shows accident consequence validations, accident consequence good to fit to the method,
residual graph of Cumulative Density Function (CDF) profile tracing infinity. In this
analysis Frequency is refer to as Fa or Na.

Figure 14: Accident frequency residual plot Figure 15: Accident consequence validation

56


Figure 16 shows residual histograms distribution diagram for accident frequency, skewed
to low risk area, outlier can be removed.

Figure 16: Residual histograms distribution diagram for accident frequency

Figure 17(a) Shows Log normal plots Accident frequency (Na), distribution shows a good
to fit. Curve Figure 17(b) also show a very good curve fit for the model.

Lognorm base eProbability Plot for Na
al
ML Estim - 95%CI
ates
Probability Plot of FA_NMCHANGE
Normal
99 99
M Estim
L ates Mean 0.00005357
Location -5.86094 StDev 0.00001331
95 95 N 50
Scale 0.764552 AD 0.821
90 90
P-Value 0.032

80 Goodness of Fit 80
AD* 2.18 70
70
Percent
Percent

60
60
50
50
40
40
30
30
20
20
10
10
5
5

1
1 0.00002 0.00003 0.00004 0.00005 0.00006 0.00007 0.00008 0.00009
F MCH G
A_N AN E
1.00E-08 1.00E-07 1.00E-06 1.00E-05 1.00E-04 1.00E-03
Data

(a) (b)

Figure 17 : Log normal plot Accident frequency (Na)

Figure 18 shows process reliability capability, the fitting of the curve revealed the
reliability of the frequency model.

57


Process Capability of FA
Run Chart of FA
LSL USL 0.00020
Process Data Within
LSL 3e-007 Overall
Target *
Potential (Within) Capability
0.00015
USL 0.0002
Sample Mean 6.73842e-006 Cp 9.67
Sample N 50 CPL 0.62
CPU 18.72 0.00010

FA
StDev(Within) 3.4414e-006
StDev(Overall) 2.7649e-005 Cpk 0.62
Overall Capability
Pp 1.20
0.00005
PPL 0.08
PPU 2.33
Ppk 0.08
Cpm * 0.00000
1 5 10 15 20 25 30 35 40 45 50
04 000 004 008 012 016 020 Observation
00 0 0 0 0 0 0
.0 0.0 0.0 0.0 0.0 0.0 0.0
-0 Number of runs about median: 2 Number of runs up or down: 1
Expected number of runs: 26.0 Expected number of runs: 33.0
Observed Performance Exp. Within Performance Exp. Overall Performance
PPM < LSL 100000.00 PPM < LSL 30681.49 PPM < LSL 407933.86 Longest run about median: 25 Longest run up or down: 49
PPM > USL 0.00 PPM > USL 0.00 PPM > USL 0.00 Approx P-Value for Clustering: 0.000 Approx P-Value for Trends: 0.000
PPM Total 100000.00 PPM Total 30681.49 PPM Total 407933.86 Approx P-Value for Mixtures: 1.000 Approx P-Value for Oscillation: 1.000

(a) (b)
Figure 18: Process capability

Figure 19 shows the matrix plot for the model, the safe areas for the variable workability
are shown in the matrix plot.

Matrix Plot of FA vs V, W, B
M atrix Plot of V, W , B , F A, FA _NM CHAN GE
50 75 100 0. 0000 0.0001 0.0002
50 75 100
50
0.00020

25 V

0 120
0.00015
80
W
40

35
0.00010
FA

25
B
15
0.0002

0.0001
0.00005
FA

0.0000
0.00008

0.00006 0.00000
FA_NM CHANG E
0.00004
0 20 40 10 20 30
0 20 40 10 20 30 0.00004 0.00006 0.00008
V W B

(a) (b)
Figure 19: Matrix plot

Figure 20 a, b, and c shows the capability report for the model.

Capability Analysis for FA Capability Analysis for FA
Diagnostic Report Sum ary Report
m

I-M Chart (transform
R ed) Custom R
er equirements
Confirmthat the process is stable. Howcapable is the process?
2000 Upper Spec 0.0002
Individual Value

0 6 Target *
Low High Lower Spec 3e-007
1000
Z.Bench =1.11 Process Characterization

Mean 6.738E-06
0
Standard deviation 2.765E-05

Actual (overall) capability
Moving Range

100
Pp 0.54
Actual (overall) Capability
Are the data inside the limits? Ppk 0.41
50 Z.Bench 1.11
LSL USL %Out of spec 13.28
0 PPM(DPM O) 132809
1 6 11 16 21 26 31 36 41 46 Com ents
m

Conclusions
-- The defect rate is 13.28% w estimates the
, hich
Norm Plot (lam =-0.50)
ality bda percentage of parts fromthe process that are outside the
The points should beclose to the line. spec limits.
Norm Test
ality
(Anderson-Darling) Actual (overall) capability is w the customer experiences.
hat
Original Transformed

Results Fail Pass
P-value <0.005 0.078

-0.00004 0.00000 0.00004 0.00008 0.00012 0.00016 0.00020

(a) (b)

58


Capability Analysis for FA
Report Card
Check Status Description

Stability Stability is an important assumption of capability analysis. To determine whether your process is
! stable, examine the control charts on the Diagnostic Report. Investigate out-of-control points and
eliminate any special cause variation in your process before continuing with this analysis.

Number of You have 50 subgroups. For a capability analysis, this is usually enough to capture the different
Subgroups i sources of process variation when collected over a long enough period of time.

Normality The transformed data passed the normality test. As long as you have enough data, the capability
estimates should be reasonably accurate.

Amount The total number of observations is less than 100. You may not have enough data to obtain
of Data ! reasonably precise capability estimates. The precision of the estimates decreases as the number of
observations becomes smaller.

(c)

Figure 20: Log normal plot Accident frequency (Na)

7.0 CONCLUSIONS

Hybrid of deterministic, statistical, historical, probabilistic and stochastic method along
with channel and vessel profile baseline data has been used to model accident possibility in
waterway in order to meet condition for safe transits, and environmental conditions for
inland waterway. Factors such as vessel type and size, traffic density, speed and visibility
conditions are major risk factor of accidents the probabilistic method represent reliable
method to develop models for safety and environmental prevention and collision accident
risk aversion who precedence is could be short term (damage) or long term (impact of oil
outflow) environmental impact. Accident collision per number of year has been
determined for potential decision support for limit definition for number of ship, speed,
required width and beam of ship. Variables that affect accident rates have been simulated
for necessary limit acceptability purpose for the channel. Accident rate has increased
compare to previous year, a situation that required attention for solution. Advantage of
implementing of TSS in respect to beam requirement is also presented. Implications of
concept of uncertainty can help also on decision support relating to navigational aids and
transit regulations for poor visibility conditions as well has employment of improved
navigation systems, such as electronic charts, GPS receivers, and VTS, to mitigate causal
factors.

REFERENCES

1. Yacov T. Haimes. 1998. Risk Modeling, Assessment and Management. John Wiley &
Sons, INC. Canada pp. 159 - 187.
2. Amrozowicz, M.D. 1996. The Quantitative Risk of oil Tanker Groundings. Master’s
degree thesis, Ocean Engineering Department, Massachusetts Institute of Technology,
Cambridge, Massachusetts.
3. Department of Environment, Modeling and data integration in the study of sediment,
2000, Kuala Lumpur, Malaysia.
4. Kielland, P., Tubman, T., 1994. On estimating map model errors and GPS position
errors. Ottawa, Canada: Canadian Hydrographic Service.
5. DnV. 2001, Marine Risk Assessment, Her majesty stationary office United Kingdom.
6. Sirkar, J., Ameer, P., Brown, A., Goss, P., Michel, K., Nicastro, F. and Willis, W. 1997.
A Framework for Assessing the Environmental Performance of Tankers in Accidental
Groundings and Collisions. SNAME Transactions

59


7. David Vose, 1996. Risk Analysis – A Quantitative Guide. John Wiley & Sons, INC.
canada pp. 67-87.
8. Fujii Y. 1982. Recent Trends in Traffic Accidents in Japanese Waters. Journal of
Navigation. Vol35 (1), pp. 88- 102
9. Millward, A. 1990. A Preliminary Design Method for the Prediction of Squat in
Shallow Water. Marine Technology 27(1):10-19.
10. John X. Wang. 2000. What Every Engineer Should Know about Risk Engineering and
Management. Markel Deker Inc, Switzerland, pp. 112-128.
11. Lewison, Gr. G., 1978. The Risk Encounter Leading to Collision. Journal of
Navigation, Vol 31 (3), pp. 288- 109.
12. M. Moderras. 1993. What Every Engineer Should Know about Reliability and Risk
Analysis. MarkelDeker Inc, Switzerland, pp. 299-314.
13. DnV, BV, SSPA, 2002, Thematic Network for Safety Assessment of Waterborne
Transportation.
14. PV Varde, ASrividya, VVs Sanyasi Rao, Ashok Chauhan. 2006. Reliability, Safety,
and Hazard – Adavnced Informed Technology. Narosa Publishing House, India,
PP339.
15. John McGregor. 2004. (BV report), Pollution Prevention and Control.

60

Predictive and reliability o o_sulaiman _revision 2_

More Related Content

What's hot (18)

Similar to Predictive and reliability o o_sulaiman _revision 2_ (20)

More from Oladokun Sulaiman Olanrewaju (20)

Predictive and reliability o o_sulaiman _revision 2_