Advanced Evacuation Analysis - Testing The Ground On Ships
Advanced Evacuation Analysis - Testing The Ground On Ships
Advanced Evacuation Analysis - Testing The Ground On Ships
ABSTRACT
This paper focuses on testing the applicability of the International Maritime
Organisation (IMO) current guidelines for advanced evacuation analysis1 to a large
passenger ship and examines critically and exhaustively the sensitivity of the
parameters involved. Simulation results are presented and discussed, leading to clear
indications of how to ensure confidence in the information provided from such a
limited-scope analysis.
INTRODUCTION
Recent well-published disasters of Ro-Ro/passenger ships together with trends of
largely increased capacity of passenger carrying ships have brought the issue of
effective passenger evacuation, being the last line of defence, in an emergency to the
centre of attention of the maritime industry worldwide. With passenger numbers now
ranging up to 6,000 on a single large cruise liner, with ships often trading in pristine
environmental areas and with rapidly growing consciousness for safety and
environmental protection among ship operators, assurance of both these issues at the
highest of levels have become the main targets for technological innovation in the
maritime industry as well as key factors for gaining and sustaining competitive
advantage.
In response to emerging needs, the 1995 International Conference on the Safety of Life
at Sea (SOLAS 95) addressed this issue specifically by the adoption of a new
regulation SOLAS II-2/28-1.3, where it is stated that escape routes onboard Ro-Ro
ferries shall be evaluated by a suitable evacuation analysis. In view of the above, in
January 1999, the IMO decided to develop Interim Guidelines for the execution of the
evacuation analysis. Following this, a Working Group within IMO was set up to study
the practicality of these guidelines, to monitor the evolution of passenger ship
evacuation simulation tools and to oversee the development of suitable rules and
regulations and of procedures and systems for existing and new ships, which recently
led to revised Interim Guidelines for passenger evacuation that address both simplified
and advanced analyses1.
Considering the wider environment, it is therefore rather disappointing but
understandable that the IMO Interim Guidelines for an advanced evacuation analysis
were influenced considerably by the preceding guidelines and thinking of the
-1-
simplified evacuation analysis, to the extent that the usefulness of such an analysis it is
at best questionable. There is no surprise here really considering that the maritime
community is rushing yet again to provide solutions to a problem that still lacks
rigorous definition and in depth understanding. Moreover, a stochastic treatment of
the problem necessitates that some basic rules must be adhered to if the results are to
be representative and hence useful. Yet, we advanced into guidelines before setting
and observing the rules! Even though it is early days still, there is no time soon
enough in attempting to fill these big gaps. This constitutes the aim of this paper.
pre-evacuation activities. In addition, in-situ response time or any change in the state
of a moving agent through intervention of e.g., crew ought to be considered.
Response (awareness) time is certainly a random variable hence it has to be sampled
for various distributions in order to evaluate its effect on evacuability
Environment (env)
Geometry
Topology
Semantics
Distribution (d)
Spatial location
of people
Evacuation plan
Crew Functionality
(XML) and a range of supporting tools allows the program to be operated as well for
any main stream software application. Evi's continuous space, discrete decision
pedestrian motion model can flexibly adapt to the complexities of any ship's geometry
and topology. In Evi 3.0, agents are considered as a vehicular transport system
capable of carrying information, interacting with other agents and autonomously
travelling around the ship environment. By programming individual agents to perform
certain tasks with elements called Objectives, it is possible to reproduce any
evacuation procedure by incorporating the actions specified in e.g. the vessels' muster
list. Crew responsibilities such as, checking cabins, alerting passengers who are not
aware of the emergency, searching for lost passengers and controlling stairways, are
straightforwardly incorporated by means of Objectives and the underlying message
structure (allowing agents to interact with and influence other agents). More complex
tasks can also be defined by scheduling of the Objectives. As evacuation procedures
may differ between different ship types and operators, Objectives provide a
fundamental technique for accurately defining appropriate responses to any emergency
scenario. The system utilises a modern graphical user interface within a virtual reality
environment.
that E+L starts before all the agents reach the assembly station). This criterion reflects
the required survivability (time-based) criterion for progressive flooding, one of the
most prevalent causes of Ro-Ro ship losses. It is also congruent with the 60-minute
structural fire integrity criterion of any independent main vertical zone within the
ships structure. For the calculation of T, a safety margin of D=200s and 600s, for day
and night case, respectively, has to be added in order to account for model omissions,
assumptions and the limited number and nature of the evaluated benchmark scenarios.
Assumptions include: crew ready at their duty stations immediately, passengers know
exactly where to go, following signage and crews instructions, presence of smoke,
heat and toxic fire products do not affect passenger/crew performance, family group
behaviour is not considered and effect of ship motion, heel and trim are not
considered.
Deck 3 - MES
Deck 7 - Liferafts
uncertainty in the parameters that drive the evacuation process (d, r(t) and s(ni)), it is
essential to quantify the uncertainty of the simulation results, which can be done by
assessing the sensitivity of the calculated assembly time to variations and/or
inaccuracy in the input parameters. An attempt to quantify this uncertainty is
presented in the next section for the sample ship on the basis of the IMO assumptions.
Table 1: Results of advanced evacuation analysis according to IMO guidelines1
Primary
Cases
Night
Day
Simulation
result
T95
1050.25s
1015.50s
Safety
margin
600s
200s
Travel
time
Evacuation
time
T = T95 +
TI = T +
1650.25s
1215.50s
2850.25s
2415.50s
2
(E + L )
3
SENSITIVITY ANALYSIS
Do changes in input parameter values lead to "reasonable" changes in output values,
both in magnitude and direction of change? Can the main variables be examined?
Does the evacuation simulation model predict average test results accurately?
Sensitivity analyses constitute the normal route to finding answers to these questions.
The calculated assembly time depends, on one hand, on the performance of the
evacuation model itself (the simulation tool), and on the other, on a number of
assumptions regarding the parameters comprising the input to the model and drive the
evacuation process. However, much has to be done to ensure that the ensuing
assumptions are appropriate and that whenever there is uncertainty, this can be
quantified. In view of the lack of appropriate and relevant experimental or full-scale
trials data, an exhaustive analysis of the sensitivity of evacuability to variations in the
input parameters is absolutely essential for two reasons: (i)-to quantify the uncertainty
in the results given the uncertainty in the input parameters with a view to asserting a
degree of confidence that can be placed upon the results, and (ii)-to assess the
robustness of the evacuation model (simulation tool) itself, with a view to evaluate
consistency and identify differences among the different approaches, so that
knowledge and understanding can be progressed further.
In the sensitivity analyses presented in this section, an attempt was made to quantify
the uncertainty associated with the assembly time calculated on the basis of the IMO
benchmarking night and day time scenarios (case 1 and 2) by estimating the robustness
of the 95 percentile of the calculated assembly time distribution (referred subsequently
as Assembly Time) to variations in the input parameters in relation to the values
implicit in the IMO guidelines. Thus, considering that evacuability is driven by four
principal parameters: env (environment the ships layout), d (spatial distribution of
people on board the ship), r(t) (response/awareness time of each person), and s(ni)
(walking speed of each person), the following has to be explained and emphasised:
env: For any comparisons to be meaningful we need to assume a time invariant
environment for evacuation simulations; thus the geometry, layout, topology and
semantics of the ships layout were kept unchanged.
-7-
d: The spatial distribution of people during the day time, will differ considerably
from that of the night time scenario. In the night time scenario, most people are
assumed to be in their cabins, spaces with more or less simple and repetitive
geometry and topology (cabins & corridors), hence the variability of the spatial
distribution is likely to be very limited, and so will be its effect on the calculated
assembly time. During the daytime scenario however, people usually occupy all
available accommodation spaces (public spaces, open decks, etc.) at different
numbers. The randomness of the spatial distribution of people in the day case
scenario, coupled with increased (with respect to cabin/corridors) complexity in
terms of geometry and topology of the corresponding spaces, will tend to increase
the degree of uncertainty in the calculated assembly time.
r(t): The response time of each person, is a random variable modelled with a
uniform distribution. The current IMO definition of the expected value and
variability of awareness time is arbitrary; hence the effect of inaccuracy in the
prediction of both parameters (mean and standard deviation) ought to be evaluated as
well as the impact of the choice of the probability distribution function (probability
model).
s(ni):The walking speed of each person, is a random variable modelled with a
uniform distribution, with a different expected value and variability for each
demographic band of the population. Similarly to response time, the impact of
choice of the probability distribution function and its parameters ought to be
evaluated. It is expected however, that the choice of probability distribution function
(normal, uniform, etc.) in this case will have little or no impact in the calculated
assembly time, as the speed achieved by each individual person is very much
affected (reduced) due to evacuation dynamics.
On the basis of the above reasoning, a number of simulation cases (Di, Rj, Sk agents
distribution, response time and speed of movement respectively) were carried out for
the same vessel evaluated in the IMO case study (env=const) in order to address the
above aspects. The results of these simulations are presented and discussed next. The
size of the sample distribution for estimating the assembly time comprises at least 20
runs, which has been found to be sufficient to ensure good statistical reliability.
-8-
R1: The response time modelled as a random quantity uniformly distributed is defined
by two parameters: mean and standard deviation1 given in the IMO guidelines.
Simulations were carried out for various response time distributions with constant
COV. The results demonstrate that whilst the total assembly time increases with
increasing response time (mean value), the individual travel times decrease, due to
reduced congestion and bottlenecks (see Figure 4) an hence increased individual travel
speed. The COV of the input response time distribution had a less significant effect,
as the variation of the resulting assembly time was of less than 5%.
Individual travel time(s), maximum
1400
1200
1000
800
600
400
200
0
0
200
400
600
Response Time / mean value (s)
1400
1200
1000
Day
Time
800
600
400
Night
Time
200
0
0
800
200
400
600
Response Time / mean value (s)
800
Figure 4: Impact of response time in the total assembly and individual travel time
(COVResponse=const)
In terms of the values implicit in the IMO guidelines, the results indicate that a
variation of 30% of the expected nominal (mean) value of the response time would
result in variations of 7% and 18% for the day and night time distribution,
respectively (response mean is 300s and 600s, respectively; see Figure 5-left).
1300
NIGHT CASE
100%
DAY CASE
70%
Response Time Stdev (% IMO input):
70%
130%
100%
40%
40%
70%
100%
70%
130%
100%
130%
Response time:
Normal distribution
Uniform distribution
1200
1100
1000
900
COV=0.17
Day Time
800
700
130%
100
200
300
400
500
600
700
-9-
S1: The walking speed is also modelled with a uniform distribution is defined by two
parameters: mean and standard deviation1 given in the IMO guidelines for specific
demographic bands. Simulations were carried out for variations of these values
(between 50% and +200%) with respect to the corresponding IMO speed band
values. The results, illustrated in Figure 6 indicate that a variation of 30% of the
expected nominal (mean) value of the walking speed bands would result in variations
of the assembly time of approximately +18% (if speed under predicted by 30%) and
8% (if speed over predicted by 30%) and this is consistent for both day and night
time IMO scenarios. The effect on achieved individual travel speed is even larger
(Figure 6-right). The effect of the spatial distribution of people is consistent with the
observations from case D1 in that for the night time scenario the assembly time is
lower than for the day case for all nominal speed values by approximately 16%.
2
Individual travel time (% of IMO results)
1.5
0.5
50%
100%
150%
200%
day case
night case
1.5
0.5
50%
100%
150%
200%
U(S,0.00)
U(S,0.15)
U(S,0.25)
U(S,0.35)
U(S,0.50)
2
1
0
0
0.5
1.5
U(S,0.00)
U(S,0.15)
U(S,0.25)
U(S,0.35)
U(S,0.50)
2
1
0
0
0.5
1.5
proportionally the travel time, which can be expected, as the actual speed is likely to
be largely affected by the dynamics of evacuation, congestion in particular.
S2: Simulations were carried out with the IMO demographic distribution (7/7/16/10-%
of the population assigned different speeds) and compared with a corresponding case
in which the demographic distribution is equally distributed (10/10/10/10/10-% of the
population assigned different speeds). The results indicated that the distribution of the
speed bands among the population has no significant effect (less than 2%) in the total
assembly time. Equally, if male/female average speed is assigned to all speed bands
(demographics ignored), the calculated assembly time does not change by more than
5%. Moreover, assigning the speed to the IMO demographic distribution bands gives
conservative results, in the sense that this results in longer assembly time.
S4: Simulations to address the effect of nominal speed without taking into account the
demographic composition of the people were also undertaken. As can be observed in
Figure 8, a nominal speed of approximately 0.8 m/s would result in travel times similar
to those calculated from simulations with the IMO demographic distribution. This
value may reflect the average actual (achieved) walking speed of all people in the
simulation.
2000
1800
1.5
day case
0.5
1600
1400
1200
day case
1000
800
night case
600
400
200
0
0
0.5
1.5
0.5
1.5
However, other factors like demographic distribution and nominal speed values
appear to have significant impact on the calculated assembly time. It would appear
that due to the dynamics of the evacuation process (congestion), a global average
value of the actual speed would produce assembly times very close to those
predicted with a full demographic distribution resulting in various speed bands, as
suggested in the IMO guidelines.
The implemented evacuation model appears to be very robust to most of the input
parameters despite the implicit uncertainty. The obtained COV for a set of initial
conditions and evacuation dynamics is of the order of 1%-4%.
CONCLUDING REMARKS
Undoubtedly, advanced evacuation simulation tools used within an appropriate
evaluation framework (of the type suggested in the IMO guidelines), will allow for a
rigorous quantification and valid evaluation of the evacuation performance of a ship,
being it for design or operational purposes. In this respect, it has to be emphasised that
in the actual IMO guidelines, the evaluation of evacuability is currently made on the
basis of 60 minutes as the life safety criterion (40 minutes for assembly time). It is felt
that the adoption of a prescriptive (fixed) criterion is rather inconsistent with the
performance-based approach adopted in the evacuation analysis. Instead, such a
criterion should derive from life safety and/or performance criteria from a number of
ship-related critical scenarios associated with fire and/or large scale flooding of the
vessels hull, or both; these constitute the two principal hazards for which a ship may
need to be evacuated. On the basis of this study, the first systematic evaluation of the
IMO guidelines onboard a real ship of the type these guidelines are meant to apply, it
can be stated that there is still some ground to cover, aiming towards close form
expressions of evacuability as functions of the aforementioned variables if application
to other ship environments demonstrated tendency to shape functions that can be
standardised. Moreover, there is still considerable effort required to address sensitivity
of design and operation-related variables to populate data bases or in time create
knowledge bases to facilitate design for ease of evacuation or to address the risk of
passenger evacuation in an all embracing risk-based approach. These are some of the
issues of ongoing research at SSRC, results of which will be reported in the near
future.
REFERENCES
1. IMO (2002), Interim Guidelines for Evacuation Analysis of New and Existing
Passengers Ships, MSC/Circ.1033, June 2002.
2. Vassalos D., Kim H., Christiansen G., Majumder J. (2001), A Mesoscopic
Model for Passenger Evacuation in a Virtual Ship-Sea Environment and
Performance-Based Evaluation, Pedestrian and Evacuation Dynamics April
4-6, 2001 Duisburg.
3. NFPA (1995), SFPA Fire Protection Engineering Handbook, 2nd Edition, US
National Fire Protection Association, 1995
4. www.fseg.gre.ac.uk/exodus/
5. www.germanlloyd.org/aeneas/aeneas.html
- 12 -