Content uploaded by Marios-Anestis Koimtzoglou
Author content
All content in this area was uploaded by Marios-Anestis Koimtzoglou on May 31, 2023
Content may be subject to copyright.
Sustainable Development and Innovations in Marine Technologies –Ergin & Guedes Soares (eds)
© 2023 copyright the Author(s), ISBN 978-1-032-41618-2
Assessing the risk during mustering in large passenger vessels: A digital
tool for real time decision support
A. Koimtzoglou, N. Themelis, N.P. Ventikos, K. Louzis & M. Koimtzoglou
School of Naval Architecture and Marine Engineering, National Technical University of Athens, Greece
K. Giannakis, P. Panagiotidis & S. Moustogiannis
Konnekt-able Technologies Ltd, Greece
M. Ramiro & J. Peña
Advantic Sistemas y Servicios S.L., Spain
D. Gomez & J. Ruiz
Atos Research and Innovation, Spain
A. Gardel
Universidad De Alcalá, Spain
P. Kavassalis, N. Traintafyllou & K. Ksystra
University of Aegean, Greece
ABSTRACT: PALAEMON is an EU project that aims to enhance the evacuation systems on passenger
ships by developing an intelligent mass evacuation ecosystem through the combination of smart devices and
ICT-based tools. Part of the PALAEMON ecosystem is the Smart Risk Assessment Platform (SRAP),
a digital risk assessment tool. SRAP consists of three modules that correspond to the main evacuation phases:
Situation assessment, Mustering process and Pre-abandonment assessment. This paper focuses on providing
a high-level presentation of the concept of the mustering assessment module. In particular, SRAP’s mustering
assessment model implements a Bayesian Network, that aims at increasing the situation awareness of the
Bridge Command Team and assisting the proper monitoring during the mustering process. The module’s
scope is to evaluate risk of delay for areas of the ship. The applied methodology for the development of the
Bayesian Network model is analysed, including the identification of the most important risk factors.
1 INTRODUCTION
Evacuation on passenger ships is a very complex
and challenging process, mainly due to the fact that
it is conducted under adverse conditions (e.g., ship
motions, blackout, etc.) and to the size of the crowd
that has to be managed. In recent years, an increase
in the capacity of the passenger ships is observed,
creating more issues to the overall evacuation pro-
cess. One of the most critical and difficult to manage
phases of the evacuation process is the mustering of
the passengers. A successful evacuation procedure is
based on the efficiency of the mustering process,
which has to be conducted timely and harmlessly.
PALAEMON is an EU-funded research project that
aims to enhance the current evacuation systems on
passenger ships by developing an intelligent mass
evacuation ecosystem through the combination of
smart devices and Information and Communication
(ICT) based tools that will assist the crew members
and the passengers during this process.
Smart Risk Assessment Platform (SRAP) collects
data from the other PALAEMON components to
enhance the situation awareness of crew members.
SRAP consists of three modules, that correspond to
the main evacuation phases: a) Situation assessment,
b) Mustering process assessment, and c) Pre-
abandonment assessment.
This paper analyses only the mustering assess-
ment module, that utilises data obtained from
PALAEMON smart devices and other ICT-tools to
evaluate the risk of mustering delay for the areas of
the vessels, divided in Main Vertical Zone (MVZ)
and the deck under investigation. To achieve this,
a Bayesian Network (BN) is implemented, that aims
at increasing the situation awareness of the Bridge
DOI: 10.1201/9781003358961-34
269
Command Team and assisting the proper monitoring
of the mustering process. The way this tool is inte-
grated in the PALAEMON ecosystem is an aspect of
major importance, together with the way data is
visually represented to be comprehensible and intui-
tive for crew members, in order to assist them in
identifying the areas of the vessel that need to take
mitigative and/or corrective actions during the mus-
tering stage.
The rest of the paper is structured as follows: Sec-
tion 2 starts with a description of the role of smart
devices in emergency situations and maritime evacu-
ation and continues by presenting briefly their char-
acteristics and challenges of using them. Section 3 is
dedicated to the description of the PALAEMON
system, its smart devices (i.e., smart bracelets, smart
cameras) as well as the Passenger Mustering and
Evacuation Process Automation System (PaMEAS),
a multi-component system combining the functional-
ity of an indoor positioning system with the capacity
of providing, to passengers and crew, instructions
that will help them reach the designated muster sta-
tions in an ordered and timely manner. Section 4 out-
lines the methodology applied to develop the BN
model for the mustering module, presents the critical
factors for the mustering assessment and the BN
model that was developed. In addition, a concise
description of the testing of the of the output in two
indicative case studies is presented in Section 5. Sec-
tion 6 describes how the results and information will
be visualized to support the management of muster-
ing and evacuation. The paper concludes with
insights regarding the mustering assessment model
and the future work that is to be performed towards
the finalisation of SRAP.
2 BACKGROUND
The evacuation of a ship is a very demanding pro-
cess, especially for the people that have to make
decisions regarding the safety of everyone onboard.
One of the most challenging aspects of the evacu-
ation is the safe and timely mustering of passengers
and their effective management, from the point that
the General Alarm sounds. All passengers need to be
alerted, proceed towards the closest/safest muster
stations in an orderly manner, assemble, be counted,
provided with life-jackets and be briefed regarding
the situation. It is a high risk and time-consuming
process, that needs to be effectively monitored by
the crew members (Koimtzoglou et al. 2021).
Consequently, it is critical to apply and develop
innovative methods to enhance the awareness of
crew members that have to make decisions for all
the people onboard and monitor the mustering pro-
cedure. There are several approaches trying to deal
with these challenges, but the integration of smart
devices with innovative methods, besides the exist-
ing ship legacy systems, is currently the way for-
ward. In such a context, Liu et al. (2022) propose
a new evacuation mechanism that considers the spa-
tial areas of the ship together with the people’s
behaviour. Their goal was to generate a model of the
evacuation capacity and provide an emergency deci-
sion-making system for the evacuation of crew
members and passengers. Moreover, Yue et al.
(2022) propose a general model and framework to
simulate the entire process of cruise ship evacuation.
Different quality indicators allow to compare differ-
ent evacuation mechanisms.
2.1 The role of smart devices in emergency
situations and maritime evacuation
Since it is not always possible to detect and prevent
some emergency situations, passenger ships can take
steps to mitigate risks against their passengers and per-
sonnel, to ensure their safety, by deploying a properly
designed emergency evacuation plan. In addition, in
order to minimise evacuation-related casualties in
amaritimeemergencyevacuation,smarttechnologies
can be used to share or relay information in real-time
among crew and evacuees. Technology is beginning to
dominate many aspects of the emergency evacuation
management situations. The role of technology in
emergency management is to connect, inform and
ultimately save the lives of those impacted by an emer-
gency. It can assist organisations to analyse, track and
study emergency preparedness and develop better solu-
tions for the people’ssafety.
2.2 Characteristics of the smart devices
Digital world and its capabilities can be utilized from the
passenger ship industry to enhance the evacuation pro-
cess and improve safety onboard. The deployment of an
automated emergency evacuation solution can drastic-
ally improve emergency communications. In case of
ashipevacuationincident,the most important task is to
ensure that all passengers and crew are accounted for
and marked safe. Digital emergency evacuation solu-
tions improve the emergency response drastically. Real-
time updates can provide critical knowledge to an inci-
dent commander and real-time access to ship systems
data holds the potential to improve safety and response
capabilities (e.g. a list of the missing people can be
broadcasted to the smart devices associated with the
emergency evacuation). This real-time information may
provide automatic updates on who is still missing or
unaccounted for and pinpoint their last known location.
2.3 Challenges of using smart devices in
emergencies
A primary challenge in emergency evacuation situ-
ation is the communication. From an organisation
point of view, the ability to articulate a situation to
emergency responders is of the utmost importance
when an incident occurs. Localisation of passengers
and crew in real-time, as well as incident context
information will assist to properly assess the risks
270
and make the right decisions. Having a clear emer-
gency action plan and smart digital tools that the
crew members are well informed of and well pre-
pared to use will limit panic and ensure that passen-
gers and crew members will arrive safely and as
quickly as possible to the mustering stations in case
of an emergency. The use of such technologies can
only further strengthen emergency ship evacuation
process.
3 THE PALAEMON ECOSYSTEM
Thanks to the digitalization and the usage of ICT
technologies, the maritime sector can benefit of not
only enhanced ship-shore communications, but also
increase the situation awareness of onboard deci-
sion-makers (e.g., Master, Bridge Command Team,
etc.) after an incident occurs. At the same time, the
knowledge base during all this process brings
a remarkable reduction of (potential) human errors,
thus leading to an overall safety level boost.
When it comes to deal with safety, PALAEMON
introduces a plethora of innovative technologies that
complements the existing ones. Combined with all
the off-the-shelf information gathered by shipboard
legacy systems (e.g., fire/smoke detectors, propul-
sion/steering system monitoring, etc.), the project
brings about an intelligence layer of services that,
altogether, give rise to a so-called smart evacuation
management system. Figure 1 presents the architec-
ture overview of the PALAEMON system, in particu-
lar it displays (from left to right) the workflow and
the information flows from its origin (i.e., Data
Sources) to the final stage, where the different ser-
vices consume the data and generate knowledge atop.
Briefly, the PALAEMON platform introduces an
open-source framework that offers a scalable, reli-
able and secure information pipeline, responsible for
guaranteeing a real-time streaming from origin to
destination, throughout all the evacuation phases.
The system will cater SRAP with all the information
coming from different sources. Besides, thanks to
the utilization of one of the timeliest event streaming
technologies (i.e., Apache Kafka), the communica-
tion across components can be straightforwardly
defined.
3.1 Smart Bracelets
Smart Bracelet (SB) is an Ingress Protection (IP)
IP68 rechargeable IoT device (wearable) that oper-
ates as a beacon with bidirectional communication
and sensing capabilities. It provides a cost-effective
two-fold digital solution. On one hand, it offers
monitoring of health indicators (i.e., heart rate,
SpO
2
), fall detection and an emergency button that
users can use to trigger assistance request to the
crew. On the other hand, it supports location tracking
when connected to PaMEAS localisation network
infrastructure. The SB provides ship digital system
with passengers/crew health and location informa-
tion in real-time. SB can display relevant informa-
tion provided from PaMEAS to support mustering
and evacuation processes on its screen (e.g., text,
signs, etc.). The SB electronics implements WiFi,
Bluetooth, RFID and GPS features.
3.2 Smart Cameras
In order to use data related with congestion by SRAP,
video feeds from cameras are utilised. Within PALAE-
MON, novel AI Smart Cameras (SC) are deployed,
gathering visual information about people in the sur-
veillance areas. SC nodes process video information in
embedded devices at the edge. Legacy Closed-Circuit
Tel evi sio n (CC TV) sy ste ms co uld b e a lso u sed e xec u t-
ing the AI algorithms in the central video server. SC
infrastructure monitors people’sactionsandprovides
SRAP with relevant information.
Each SC node has specific parameters that can be
modified remotely. These parameters can be set from
the central dashboard to correctly orchestrate the
evacuation and rescue of people. The quantity and
sampling of information sent by SC nodes depends
on the current status of the alert level. The main
information sent by each SC is the number of people
detected in the monitored area.
If the alert level is “evacuation emergency”,
SC detects the following situations: 1) Empty
space (i.e., area without people), 2) Misguided
people, 3) Many people running, and 4) Congestion/
trapped people.
3.3 PaMEAS
PaMEAS is a multi-component system combining
the functionality of an indoor positioning system
with the capacity of providing, to passengers and
crew, instructions in an effort to perform an ordered
and timely evacuation. Specifically, it permits real-
time tracking and monitoring of the position of pas-
sengers and crew and automatically launches, in the
case of an emergency, an evacuation “messaging”
policy.
Figure 1. Overview of PALAEMON ecosystem.
271
This policy is based on pre-established plans and
procedures but also receives real-time input from
several PALAEMON components and the ship’s
Command Centre (i.e. Bridge). It is expressed as
a set of rules that automatically compose the mes-
sages delivered to passengers depending on the
evacuation transition phase, the passenger identity
profile and health-status information, their location
and the availability of evacuation routes; to crew as
well, alerting them on eventual incidents (e.g., pas-
sengers locked in a cabin) during the evacuation pro-
cess which require their intervention.
This policy will be: a) communicated directly to
passengers’personal devices (e.g., mobile phones,
smart bracelets, smartwatches, etc.) via easy to read
and comprehend text messages, but also via audio
and visual data or video transmission and warnings,
whenever this is needed and, b) projected to the
physical ship space, by using the appropriate signage
indicating the suggested evacuation paths.
PaMEAS essential objective is to optimize the
evacuation process via an “augmented”, technology-
aided evacuation process redesign. PaMEAS evacu-
ation instructions will be forwarded to passengers’
and crew members devices through PaMEAS-N.
PaMEAS-IoT will complete the operation by activat-
ing specific IoT signage equipment, such as LED
emergency lighting, that will signal to passengers the
suggested routes of escape.
PaMEAS provides the means of execution of the
Standard Evacuation Functions: a) tracking the status
and location of passengers, b) marking the Evacuation
Path, c) engaging the messaging process and preparing
the passengers and crew for evacuation, d) providing
directions to and through the Evacuation Paths, and, e)
tracking the status and location of resources (crew sup-
port, etc.) and reassessing response plans. It also per-
forms Incident Management Functions, i.e., incident
the real-time management of incidents, such as
apassenger’sinjurycase,iftheyoccurduringthe
evacuation process. PaMEAS establishes a spanning
service layer that covers the whole evacuation process
from the activation of the Evacuation Plan until the use
of the Mass Evacuation Vessel (MEV) and the subse-
quent abandonment of the ship.
4 MUSTERING ASSESSMENT MODEL
This section is dedicated to the description of the Mus-
tering Assessment model which is divided into two sep-
arate BNs. BNs are a Directed Acyclic Graphs (DAGs)
that include random variables (nodes) connected with
edges that represent probabilistic and causal relationship
between them (Jensen & Nielsen 2007). The first BN is
utilised for estimating the status of each passenger indi-
vidually, whereas the second for estimating the risk of
delay for the specificareaofthevessel.
The mustering assessment model is activated fol-
lowing the sounding of the GA and aims at quantify-
ing the mustering effectiveness, in terms of delays
and whether passengers are at risk as they move
towards the muster stations. Its main functionality is
to assist crew members to identify which areas of the
ship mitigative and/or corrective actions may be
needed during the mustering process.
The model consists of a series of nodes that
derived from a process similar to the one described
thoroughly by Ventikos et al. (2021). Computation-
ally, BN exploits Bayes’theorem for conditional
probabilities, which is used for updating beliefs
given the observation of evidence. After identifying
the variables that are important for the mustering
procedure and converting them into nodes in
a specific structure, a quantification of the strength
of the relationship between them by specific Condi-
tional Probability Tables (CPTs) had to be
developed.
The CPTs of the mustering assessment model
were based on two series of interviews with experts
and on the results of two PALAEMON stakeholders’
workshops. During the second workshop attendants
were asked to rank a series of factors for the muster-
ing assessment from the most to the least important.
The importance of each risk factor has been reflected
in the CPTs of the model. The factors were listed in
the following order: Blocked evacuation routes,
Number of trapped passengers, Condition of Muster
Stations (open/blocked), Presence of crew members
to guide/assist passengers, Congested evacuation
routes.
4.1 Literature review
BNs have proven to be useful tools for analysing the
risk during evacuation in the maritime domain
(Eleye-Datubo et al. 2006). DNV (2019) investigated
the possibilities and limitations of implementing
risk-based methods for the Goal-Based Standards
and Safety-Level Approach (GBS-SLA) compliant
to IMO instruments for life-saving appliances. The
study mainly focused on risk assessment by applying
aquantitative risk-based method. The probability of
human losses due to each hazard is quantified by
separate BN sub-models based on publicly available
information. During evacuation and abandonment,
the most important risk factors that affect the safety
of the persons on board are the following: capsizing,
faulty evaluation of situation, failure to muster, acci-
dent during the evacuation process, missing life-
saving means and accident during waiting for
rescue.
In addition, according to the simulation scenarios
conducted by Sarshar et al. (2013a) for fire incidents,
“Passenger Condition”and “Evacuation Time”
seemed to be critical factors that affect the probabil-
ity of human losses. The aforementioned model was
enhanced with dynamic elements for the prediction
of the probability of panic and evacuation time.
Panic contributes to a significant increase of required
evacuation time and of the probability that passen-
gers would fail to reach the embarkation stations
272
(Sarshar et al. 2013c). Sarshar et al. (2013b) also
introduced a Dynamic Bayesian Network (DBN) for
congestion modelling and analysis during a ship
evacuation after a fire incident.
4.2 Risk factors
The mustering model aims at evaluating the effect-
iveness of the mustering process with respect to time
and in terms of the below described aspects:
Potential for individual injuries: A main param-
eter that affects the status of a passenger while
attempting to arrive at a muster station is related to
the potential for experiencing an injury due to the
presence of hazards (e.g., smoke, high temperatures,
presence of water, effects of an explosion, trips and
falls due to heeling, etc.). The probability that pas-
sengers may be injured is assumed as the combin-
ation of the presence of hazards and an alarm
indicating whether a passenger has fallen or not.
Efficient movement of individuals: The ability to
move towards the assigned muster station is an indi-
cation that the passenger is in a position where he/
she follows the cues and the correct evacuation
route. The factors that affect this ability are the pas-
senger speed, followed by O
2
saturation and heart
rate. In particular, low speeds combined with low O
2
saturation may be an indication of passengers that
are immobilized or trapped and thus, are in peril.
Additionally, according to Reaves & Angosta (2021)
heart rate above a specific limit may indicate that
passengers are in panic (fight or flight), hence their
ability to move efficiently is reduced.
Individual Status: Medium movement efficiency
combined with medium potential for injuries leads to
adelayedmovementofthepassengertowardsthe
muster station. In case of low movement efficiency and/
or presence of hazards it is assumed that assistance by
the crew will be required. The potential of injuries is
considered more critical compared to movement
efficiency.
Group performance: With reference to a specific
deck at a MVZ, the group performance index depends
on the individual’sstatus,aswellasthedistanceofthe
deck from the muster station and the existence of con-
gestion in critical locations on the deck. The distance of
the deck from the muster station implies that the longer
the location of the group of the passengers from the safe
area, the higher the time needed to reach it and thus the
higher the exposure to hazards. On the other hand, con-
gestion in critical locations throughout the evacuation
routes, such as stairways, corridors or exits will increase
the exposure to hazards due to the delay in arriving at
the muster station, while the occurrence of panic, dis-
ordering and even injuries is also possible.
Status of supporting systems: The proper (as
designed) operation of ship systems supporting the mus-
tering process is a contributing factor for ensuring that
the passengers receive the right guidance while moving
to the muster stations. Such systems are the Public
Announcement (PA) system, the wayfinding system
including low-lights and exit signs and the presence of
crew members in their assigned location aiming at pro-
viding guidance and help. Therefore, the malfunction or
inoperability of these systems will result in the delay of
the mustering.
Status of escape routes: The occurrence of con-
gestion or the existence of a blockage that may result
in the unavailability of an exit or inaccessibility of
a specific path will lead to the increase of the
required time for the passengers to arrive at their
assigned muster station.
Ava i l abil i t y o f must e r s tatio n s : T here i s a p o ssibi l i ty
that muster stations will be inaccessible (or not safe to
approach or stay) due to the presence of hazards in its
vicinity (e.g., fire in the deck below). Such events will
result in some escape routes being less safe to follow,
leading as a consequence to a delayed duration of the
mustering, as passengers shall use alternative escape
routes, spending time to their identification and/or
forced to cover a longer distance than expected.
Passenger flow on area: The flow of the passen-
gers on the vessel area under examination relates to
the individual status of every passenger located in
that area. The aggregation of all the individual status
indices in the specific area provides the basis for the
CPT of the node “Passenger flow on area”according
to the following rationale:
•The probability of the “Normal”state of the pas-
senger flow node is equal to the percentage of the
passengers whose “Free movement”state in the
individual status node has higher probability
compared to the two other states.
•The probability of the “Delayed”state of the pas-
senger flow node is equal to the percentage of the
passengers whose “Movement delayed”state in
the individual status node has higher probability
compared to the two other states.
•The probability of the “Disrupted”state of the
passenger flow node is equal to the percentage of
the passengers whose “Assistance required”state
in the individual status node has higher probabil-
ity compared to the two other states.
Table 1. Data from smart devices utilised by SRAP.
PALAEMON-
Component Data Category Data Type
PaMEAS Passengers
& Crew
Location Geofence
Heart rate
Smart
Bracelets
Passengers
& Crew
Health Oxygen satu-
ration
Condition Fall detection
Congestion/
Smart Cameras Passengers
& Crew
Evacuation
routes
Blockage points
Presence of peo-
ple in a specific
area
273
4.3 Exploited data to SRAP
The successful operation of SRAP, among others, is
based on the data generated by the PALAEMON
devices mentioned in the previous section. SRAP
exploits and combines data taken from the ship’slegacy
systems (smoke detectors, flooding sensors, etc.), as well
as the output of other PALAEMON components. The
data utilised by SRAP is shown in Table 1.
4.4 Bayesian network
The Mustering Assessment model consists of two sep-
arate BN models. The first aims at assessing the indi-
vidual status of each passenger, whereas the second
aims at assessing the risk of delay for areas of the
vessel (Figure 3). A bottom-up approach is applied in
order to characterize the progress of the mustering
process for specificareasofthevessel,thatarediv-
ided per deck and MVZ as described in Section 6.
At the microscopic level, which reflects the risk to
individuals (i.e., each passenger or crew member),
an index for each person on the specific area is cal-
culated based on the individual data from other
PALAEMON components.
At the mesoscopic level, which reflects the
“Group performance”in the area under examination,
the indices calculated for the individuals are aggre-
gated and provided as input to the “Passenger flow
on deck”that subsequently influences the Group per-
formance node. The aggregation is based on the per-
centages of the passengers, which is directly fed to
the respective CPT of the “Passenger flow on area”.
As a result, SRAP will provide the Master a holistic
view regarding the progress of the mustering process.
5 INDICATIVE CASE STUDIES
This section presents two indicative case studies to
highlight the way that the mustering model will
operate. In particular, the first case study presents the
individual status node (top left part of Figure 2),
while the second case study presents the final out-
come of the model, i.e., the risk of delay for
a specific area of the ship (bottom right part of
Figure 2).
In the first case study a passenger is located on an
area of the ship where smoke is present due to a fire
accident. At the beginning the passenger moves effi-
ciently towards the muster station (i.e., 46% prob-
ability) however, due to the presence of smoke there
is an increased probability of 40% for delay due to
reduced visibility. When the O
2
saturation becomes
critical, along with the close to zero speed (indicat-
ing immobilisation), while the fall detection alarm
has been activated. Thus, the outcome is that the pas-
senger requires assistance with probability 54%.
Tables 2 and 3 provide the input data and the result
of the individual status node (presenting all three
states) for this case study.
The second case study presents the part of the mus-
tering model which evaluates the risk of delay for each
area of the ship. The input in the parent nodes is
shown on Table 4. At first, the risk of delay for an area
in a long distance from the Muster Stations is shown.
70% of the passengers in this area are moving without
any delay (i.e., Normal) whereas the remaining 30%
experience some delays (i.e., 30% delayed). Therefore,
the performance of people in this area, i.e., group per-
formance, is considered high (efficient), thus there is
Figure 2. The Mustering Assessment BN.
274
no congestion and all supporting systems are operating
properly. The outcome is that the risk of delay for this
area is low, i.e., 59% (Table 5).
On the contrary, for another area closer to the
Muster Stations (i.e., moderate deck distance) there is
amoderatecongestion.Moreover,thepassengerflow
states changed, having in this area half of the passen-
gers moving efficiently (i.e., Normal 50%) and the
other half experiencing some delays (i.e., 50%
Delayed). In that case the respective result for the risk
of delay increases from low to moderate and very close
to high (Table 5).
6 VISUALISATION
As previously described, the mustering of the passen-
gers is one of the most important parts of an evacu-
ation process when an incident occurs in the ship. As
the capacity of the passengers is increasing on large
passenger ships, an optimal mustering process is the
key to a successful evacuation. In order for the Master
of the ship to have a better assessment of the mustering
process, the existence of the proper visualisation is
needed. The proper visualisation is able to assist the
Master to have a general view of the whole ship and
the situation of all the passengers (and crew).
For this purpose, the PALAEMON Incident Man-
agement Module (PIMM) has been developed in order
to combine the PALAEMON components for the
scope of the evacuation process. The scope of the
PIMM is to be the Human Machine Interface (HMI)
that displays all the information generated (i.e. dash-
board) by the underlying PALAEMON devices and
systems (including SRAP). It is very important for the
components to exchange data in order to improve their
functionalities as it leads to a better whole solution for
the Master/Bridge in case of an incident.
SRAP’sfinal outcome (i.e., the level of risk of
delay) is calculated for areas of the ship that are div-
ided per MVZ and the specificshipdeckunderexam-
ination and is visualised with a colour scale on the
PIMM. So, for the Mustering process, a visualized dia-
gram of the ship has been created and separated into
parts that the ship is made up of as shown in Figure 3.
SRAP provides its main output to PIMM every
time that SRAP will be provided with revised input
data. The SRAP indicates the state of the parts of the
ship. The parts that are colorized in red are areas
where the risk of delay is high, the parts that are col-
orized in orange are areas that the risk of delay is
moderate, and the parts that are colorized in green
are areas where the risk of delay is low (Figure 3).
PaMEAS component implements an IT-enabled
functionality that can streamline and support in real-
time the evacuation process, by providing data to
PIMM, such as the total number of passengers, how
many of them are currently located in each muster
station, how many have been board in the Mass
Evacuation Vessels (MEVs). An indicative visualiza-
tion of PaMEAS on PIMM is shown in Figure 4.
Table 2. Parent nodes states for the 1
st
case study.
Node 1 2
Heart Rate Normal Normal
O2 Saturation Normal Critical
Passenger speed As expected Immobilised
Presence of hazards Smoke Smoke
Fall detection No Yes
Table 3. The individual status states for the 1
st
case study.
Node States 1 2
Assistance required 14% 54%
Individual status Movement delayed 40% 34%
Free movement 46% 12%
Table 4. Parent nodes states for the 2
nd
case study.
Node 1 2
Deck distance from
Muster station Long Moderate
Availability of
Muster stations Accessible Accessible
70% Normal 50% Normal
Passenger flow 30% Delayed 50% Normal
0% Disrupted 0% Disrupted
Congestion Low Moderate
Blockage No No
Status of PA Operational Operational
Status of wayfinding
System Operational Operational
Presence of crew Present Present
Table 5. The risk of delay states for the 1
st
case study.
Node States 1 2
Low 59% 29%
Risk of delay Moderate 27% 36%
High 14% 35%
Figure 3. Indicative figure of SRAP’smusteringprocess
assessment output on PALAEMON Dashboard (i.e., PIMM).
275
7 CONCLUSIONS
The difficulty in managing emergency situations, but
also their complexity, have shown that the utilization
of technology can provide an advantage in order to
perform the evacuation process and specifically the
mustering of the passengers onboard ships in a more
efficient way. PALAEMON is a project that aims to
develop a holistic system that will combine ICT
smart devices and Artificial Intelligence algorithms
to increase the efficiency of evacuation process.
SRAP combines data from various smart devices
to increase the situation awareness of the crew mem-
bers during all phases of the evacuation process. The
mustering model presented in this paper consists of
two BN models that evaluate the individual status of
each passenger as well as the risk of delay for the
ship areas, divided per MVZ and deck. The model
has been developed as a combination of two BN
models that describe the causal relationships between
the identified risk factors that have been converted
into nodes. Furthermore, the model combines various
data from different smart devices (i.e. SB and SC)
and ICT tools (i.e. PaMEAS) to achieve its goal.
The results from the indicative case studies high-
light that the structure of the model provides promis-
ing results, however more detailed scenarios should
be tested to finalise it. The visualisation of the model
includes all the information required by the Masters
and crew members, as it provides a holistic overview
of the ship and its areas during the mustering phase.
Finally, future research includes the integrated val-
idation of the mustering module not only for the case
of one passenger, but for many passengers simultan-
eously affected by events taking place in different
decks and MVZs at the same time. Following the com-
pletion of this step SRAP shall be validated as
a whole, through a case study simulating all the phases
of the evacuation and implementing all three modules
(i.e., Situation assessment, Mustering assessment and
Pre-abandonment assessment) to be fully integrated in
PALAEMON system.
ACKNOWLEGDMENTS
The work presented was supported by the project
“PALAEMON: A holistic passenger ship evacuation
and rescue ecosystem”, funded by the EU’s Horizon
2020 research and innovation programme, EU.3.4. -
SOCIETAL CHALLENGES - Smart, Green and
Integrated Transport of the European Union under
the Topic MG-2-2-2018 - Marine Accident Response
(Grant Agreement No. 814962).
REFERENCES
DNV. 2019 BMVI study on safety model for life-saving
appliances Risk Model, p. 91. Available at: https://www.
bmvi.de/SharedDocs/DE/Anlage/WS/study-on-safety-
model-teil3.pdf?__blob=publicationFile.
Eleye-Datubo, A. G., Wall, A. Saajedi, A. Wang J. 2006. Enab-
ling a Powerful Marine and Offshore Decision-Support
Solution Through Bayesian Network Technique. Risk Ana-
lysis 26 (3): 695–721. https://doi.org/10/bvghhr.
Jensen, F.V. Nielsen, T.D. 2007. Bayesian Networks and
Decision Graphs. 2nd ed. Information Science and Stat-
istics. New York: Springer.
Koimtzoglou, A. Louzis, K. Michelis, A. Koimtzoglou, M.
A. 2021. Mass Evacuation of Large Passenger Ships: A
State-of-the-Art AnalysisSetting the foundations for the
Intelligent Evacuation Ecosystem PALAEMON. Proc.
Annual Conference of Marine Technology, Athens, Hel-
lenic Institute of Marine Technology. 137–150.
Liu, Z. Li, Y. Zhang, Z. Yu, W. 2022. A New Evacuation Acces-
sibility Analysis Approach Based on Spatial Information.
Reliability Engineering & System Safety,108395.
Reaves, C. Angosta, A.D. 2021. The relaxation response:
Influence on psychological and physiological responses
in patients with COPD, Applied Nursing Research, 57,
p. 151351. doi:10.1016/j.apnr.2020.151351.
Sarshar, P. Radianti, J. Granmo, O.C. Gonzalez, J. 2013a.
A Bayesian network model for evacuation time analysis
during a ship fire. Proc. 2013 IEEE Symposium on
Computational Intelligence in Dynamic and Uncertain
Environments (CIDUE), Singapore: IEEE, pp. 100–107.
doi:10.1109/CIDUE.2013.6595778.
Sarshar, P. Radianti, J. Granmo O.C. Gonzalez, J.J. 2013b.
AdynamicBayesianNetworkModelforPredictingCon-
gestion During a ship Fire Evacuation, Wor l d Cong r e s s on
Engineering and Computer Science 2013 Vol I WCECS
2013,23-25October2013,SanFrancisco,USA.
Sarshar, P. Radianti, J. Gonzalez, J.J. 2013c. Modelling
panic in ship fire evacuation using dynamic Bayesian
network. Proc. 2013 Third International Conference on
Innovative Computing Technology (INTECH), London,
United Kingdom: IEEE, pp. 301–307. doi:10.1109/
INTECH.2013.6653668.
Ventikos, N. Themelis, N. Louzis, K. Koimtzoglou, A.
Michelis, A. Koimtzoglou, M. A. 2021. Evaluating risk
during evacuation of large passenger ships: A smart risk
assessment platform for decision support. Proc. 6th
International Conference on Maritime Technology and
Engineering - MARTECH 2022.
Yue, Y, Gai, W. M. Deng, Y. F. 2022. Influence factors on
the passenger evacuation capacity of cruise ships: Mod-
elling and simulation of full-scale evacuation incorpor-
ating information dissemination. Process Safety and
Environmental Protection, 157, 466–483.
Figure 4. Indicative visualisation of PaMEAS on PIMM.
276
