Available online at www.sciencedirect.

com

Procedia - Social and Behavioral Sciences 53 (2012) 178 – 188

SIIV - 5th International Congress - Sustainability of Road Infrastructures

Simulation of People Evacuation in the Event of a Road Tunnel Fire

Ciro Caliendoa*, Paolo Ciambellib, Maria Luisa De Guglielmoa, Maria Grazia Meob,
Paola Russob
a
University of Salerno, Department of Civil Engineering, Via Ponte Don Melillo 83030 Fisciano (Salerno) Italy
b
University of Salerno, Department of Industrial Engineering, Via Ponte Don Melillo 83030 Fisciano (Salerno) Italy

Abstract

This paper presents basic findings obtained by means of a simulation model for representing an evacuation process from a
curved bi-directional Italian road tunnel in the event of a fire. Simulations were carried out by using the STEPS people
evacuation model associated with a CFD model simulating the fire. It was found that the evacuation time is primarily
influenced by the walking time and to a less extent by the pre-movement time in all scenarios studied. The presence of an
alarm system reduced the evacuation time for most the tunnel users, allowing them to reach the tunnel exit in safety.

© 2012
© 2012 The
TheAuthors.
Authors. Published
Published by
by Elsevier
Elsevier Ltd.
Ltd. Selection and/or peer-review under responsibility of SIIV2012
Selection Committee
Scientific and/or peer-review under
Open access responsibility
under CC BY-NC-NDof SIIV2012
license. Scientific Committee
Keywords: Evacuation model; CFD model; Road tunnel; HGV Fire.

1. Introduction

The severe consequences to people, structures, traffic and environment occurred as a result of the disastrous
fire accidents in Alpine tunnels (Mont Blanc tunnel, 1999; Tauern tunnel, 1999; St. Gotthard tunnel, 2001) have
shown the necessity of improving the level of safety in the existing tunnels.
For this aim the Directive 2004/54/EC [1] has recommended minimum safety requirements for all road tunnels
(with length >500 m) of the trans-European road network (TERN). The Directive has the prime objective of
preventing critical events that endanger human life, environment, tunnel structures and equipment; the second is
of reducing the consequences of such events by enabling people to rescue themselves, ensuring efficient action by
emergency services, protecting the environment and limiting structure damage.
The evacuation process from a tunnel under emergency conditions is a complex phenomenon, which involves
considering different factors such as both the physical characteristics of the tunnel and human behaviour. But
while the first type of factor is deterministic and consequently easy to insert in an evacuation model, the variables

* Corresponding author. Tel.: +39-089-944140; fax: +39-089-964945.

E-mail address: [email protected]

1877-0428 © 2012 The Authors. Published by Elsevier Ltd. Selection and/or peer-review under responsibility of SIIV2012 Scientific Committee
Open access under CC BY-NC-ND license. doi:10.1016/j.sbspro.2012.09.871
 Ciro Caliendo et al. / Procedia - Social and Behavioral Sciences 53 (2012) 178 – 188 179

related to human behaviour present difficulties in their exact definition due to their intrinsic randomness. The
literature presents some studies concerning the influence of the variables related to human factors in road tunnel
evacuations. These works [2-5] consider that the time required to escape (evacuation time) may be generally
divided into three phases, namely the detection time, the reaction time and the travel time. With reference to these
times several human behaviours could be observed during emergency. In addition the interactions between tunnel
occupants are also a crucial factor in evacuation process. In fact the user’s behaviour is also affected by the others
regarding the decision to leave the vehicle and how to escape. In general two main types of interaction groups are
formed: “emerging groups” and “established groups”. The former group type can naturally arise and dissolve
during emergency; while the latter, which is formed by family or friends, is characterized for staying together and
ensuring that each member evacuates under safe conditions. In this last case the evacuation time is the same for
the whole group, but walking speed is that of the slowest user.
In the light of the above considerations, the people evacuation process has been considered in this paper to
deserve a more detailed investigation. In this respect, simulations tools for the prediction of pedestrian
movements appear to provide a useful aid for designers and researchers. Nevertheless these simulation tools have
been hitherto used prevalently for modelling pedestrian flows in evacuation processes from public or private
buildings, whilst they have been performed only few times for application to a road transportation system and
more especially in tunnels in the event of a fire.
In the literature several simulation models with associated software are nowadays available as a tool for
investigating people safety in the case of an emergency in a tunnel. In general these evacuation models simulate
factors belonging to two main categories. The former category is related to physical features such as fire
spreading more especially in tunnels, vehicles being involved in the accident, and tunnel geometry. The latter is
related above all to human behaviour such as the time to react and leave the vehicle, while moving in a particular
direction, choosing both the walking speed and the exit door. In other words, the simulations of people
evacuations are generally carried out on the basis of a software using as input the results obtained from a
compatible software that models the spread of fire, smoke, and toxic gases, as well as the presence of obstacles
and the characteristics of the tunnel [6-11]. However, it is to be said that the majority of these studies do not take
sufficient account of the fact that the interaction between fire and evacuation is affected by the vehicles queuing
in the tunnel. This acts, in contrast with the condition of isolated vehicles, as an obstacle on the spreading of the
fire. In fact the queues of vehicles cause both higher temperatures and a reduced longitudinal velocity in air flow,
as well as reducing more rapidly the visibility distance in tunnel thus influencing people evacuation negatively.
As a result, there are at least three main reasons for justifying this paper. The first is motivated by the need to
investigate more efficiently the evacuation process due to fire in a tunnel. The second, is the setting up of a
simulation model for representing with sufficient realism the evacuation of people from a tunnel under emergency
conditions. Finally, the understanding of how the effects of fire may spread from the burning vehicle to other
areas of the tunnel taking into account the presence of queuing vehicles and how this interferes with evacuation.
This is the context wherein the present work is set. The paper sets out to study an evacuation process from a
curved bi-directional road tunnel in which a fire source due to a heavy goods vehicle (HGV) is considered. For
this aim the worst fire scenario was investigated, which was identified to be when the HGV was located in the
middle of the tunnel length. In this position of the HGV fire, it is assumed that other vehicles queue and come to a
halt without overtaking the burning vehicle. In addition people are prevented from escaping with their own
vehicles, but they have to leave their cars and walk towards a safe place. Given that along the length of the
investigated tunnel there are no emergency exits for the tunnel occupants, the users have to walk on sidewalks in
the direction of the exit portals (Portals A or B if they are upstream or downstream of the fire, respectively).
However the worst condition for people’s safety is more especially studied. This is for the zone downstream of
the fire (from the middle of the tunnel and exit B) because, due to the longitudinal ventilation direction of jet-
fans, tunnel occupants might be here exposed to hot gases and smoke with the danger of burning or suffocation.
For this purpose the simulation tool labelled STEPS for the prediction of pedestrian movement under emergency
180 Ciro Caliendo et al. / Procedia - Social and Behavioral Sciences 53 (2012) 178 – 188

conditions was used. This evacuation model was associated with a CFD model [12] for simulating fire. In other
words simulation of people evacuation was carried out by using as input in STEPS the results obtained from the
CFD software that modelled fire, smoke, toxic gases and visibility in the tunnel investigated.
In the light of the above considerations, the present paper is organised as follows: the next section deals with
the evacuation model that has been used and its calibration; then the results of simulations are presented and
discussed, and an evaluation of people safety is made. Finally, the conclusions and directions of future research
are reported.

2. STEPS evacuation model

In this work, the STEPS evacuation model by Mott MacDonald was used to evaluate the evacuation time. It
has so far been applied to various types of infrastructure (e.g. airports, exhibition centres, offices, railway stations,
shopping malls, stadia, tunnels) for simulating people movement and behaviour in various evacuation conditions
[13]. STEPS employs a modern agent-based approach which predicts the movement of discrete individual (virtual
people) through three-dimensional space. In agent–based modeling the agents are capable of interacting with the
environment and/or other agents following a list of rules that guide their movement. The geometry of the space
under consideration is divided into two basic entities: planes and paths. Planes are used to model any surface on
which people move (e.g. floors, platforms and stairs). Paths are geometric entities allowing only unidirectional
flow in single file which are used for the representation of movements devices. Connections between the planes
and paths are provided by exits. The flow rate through an exit is a function of its width and capacity. Planes are
divided into discrete entities using Cartesian meshes of chosen size. This mesh serves both as the basis for the
free-space movement algorithm and for solving the inter-agent collision problem in the case of people crowds.
The mesh size is chosen in general to represent the area occupied by a single person and can be set to different
sizes in different locations.
To each agent moving through the domain under consideration are assigned certain basic attributes: free
walking speed; awareness of the modelled environment; patience required to make queuing choices; association
with other members of a family or friends group; and pre-movement time in the event of an evacuation. Such
attributes can be assigned both deterministically and by distribution laws. Randomness is built into the solution
procedure both at the level of resolving equally-balanced decisions and in the movement algorithm, so that an
ensemble average of several simulations should be used [14].
Moreover, STEPS allows the import of data (e.g. temperatures of hot gases, smoke, toxic gases, etc.) from fire
simulation models. In this paper, the CFX by ANSYS [12] was used. Details of the CFD simulations were found
elsewhere [11]. The effects on the occupants’ movements can be calculated according to some values established
in the literature. For example smoke data, imported in terms of smoke extinction coefficient K, can be used to
affect walking speeds according to the Jin and Yamada [15] empirical relationships, also reported by PIARC [4]
for tunnel fire scenarios. In the presence of smoke the walking speed can be reduced because of both smoke
concentration and irritation, from 1.3 m/s (if K< 0.2 m-1) down to 0.3 m/s, as a function of smoke concentrations
and properties (i.e. if K> 1m-1 for non-irritating smoke or if K> 0.5 m-1 for irritating smoke). In fact, the walking
speed in non-irritant smoke decreases gradually as the smoke extinction coefficient increases, whereas in irritant
smoke the speed decreases rapidly due to both the physical obscuration effects and the physiological irritation
effect: as people cannot keep their eyes open for a long time without heavily running tears, they inevitably walk
zigzag or step by step along the side wall.

3. Validation of the evacuation model

The STEP model is here validated by means of a comparison with data of evacuation tests performed in one of
the tubes of the Benelux tunnel in Rotterdam (Netherlands) [16]. The tunnel tube is one-way, more than 1 km
 Ciro Caliendo et al. / Procedia - Social and Behavioral Sciences 53 (2012) 178 – 188 181

long, almost 10 m wide with driving lanes of 3.5 m width and an escape strip of 1.5 m next to either wall, and
also with 12 emergency exits. In the tests a truck slowed down and stopped in the middle of the tunnel, blocking
about 40 cars in both driving lanes. Some time after the truck had stopped, an alert hazard announcement was first
made via the loudspeakers in the tunnel, then the drivers (the only person in each car) were told to evacuate the
tunnel. The tests differ in the position of the cars in the tunnel with respect to the HGV and the emergency exit.

Figure 1. Comparison between evacuation time from evacuation test and modelling

We carried out simulations of the evacuation process of the reported tests, and performed a sensitivity analysis of
STEPS numerical parameters (time step and grid size). As input data of the model (i.e. walking speed, pre-
movement time) the distribution function of parameters was used as obtained from tests.
An example of results obtained is shown in Figure 1. A good prediction of test data was obtained: the evacuation
time was overestimated by the model, differences of about 15 s between test data and simulations were observed.
These differences were assumed in this work to be within acceptable tolerance levels.

4. Case of study

4.1. Tunnel description

The investigated tunnel is located in the South of Italy along a rural road (S.S.145) which connects the cities of
Castellamare di Stabia and Sorrento. The tunnel, labelled Varano, is a curved bi-directional road tunnel, 1.2 km
long, 10.5 m wide and 5.5 m high, with an almost uniform upward slope of 2%. At present, it has neither
emergency exits for people nor alarm systems triggering evacuation in the event of a fire accident. However, it is
equipped with two lay-bys in order to provide a safe place for parking broken down vehicles and, hence, prevent
tailbacks. The tunnel is going to be improved in compliance with the Directive 2004/54/EC [1], with a
longitudinal ventilation system that will be able to provide efficient forced airflow in the ascending direction,
both in ordinary traffic conditions and in the case of fire emergency, when a linear heat detection system will
activate all fans. These future features of the tunnel were taken into account in the simulations. More especially
the ventilation system is assumed to be constituted by eight pairs of axial jet fans with a direction of air flow from
the Portal A to B. The tunnel has a rectangular cross section with two lanes (Figure 2).
182 Ciro Caliendo et al. / Procedia - Social and Behavioral Sciences 53 (2012) 178 – 188

Figure 2. Cross section of the tunnel (median in two continuous strips marked on the pavement).

The annual average daily traffic (AADT) is more than 10,000 vehicles per day for each traffic direction, with a
percentage of heavy vehicles slightly less than 5%. The speed limit for vehicles is 50 km/h.

4.2. Scenario

An emergency scenario under congested traffic was assumed as follows: a HGV, driving in the upward direction,
comes to a halt and catches fire (HRR peak of 50 MW) at the tunnel centre, causing a queue on each fire side.
A number of 78 cars (carrying two persons per car) is supposed to queue up and stop, maintaining a minimum
distance from the vehicle in front of about 2 m, up to the tunnel downstream portal, which is reached by the
queue after 5 min from the fire start.
People caught up in the traffic jam downstream the HGV fire can be considered as being unable to leave the
tunnel by car. Since along the length of this tunnel there are no emergency exits, people evacuation is assumed to
occur along the tunnel sidewalk in the direction of the exit portals. The evacuation process, however, is worse for
people who are in the zone downstream of the fire because, due to the longitudinal ventilation direction of jet-
fans, they might be exposed to hot gases and smoke with the danger of burning or suffocation. Therefore, in the
following only this evacuation route is considered (from the middle of the tunnel length to portal B). The
maximum distance to walk to reach exit B is 590 m. In Figure 3, an example of STEPS snapshots for the
simulated scenario is reported during the first minutes of the people’s escaping.
 Ciro Caliendo et al. / Procedia - Social and Behavioral Sciences 53 (2012) 178 – 188 183

Figure 3. STEPS snapshots: a) near the burning HGV; b) along the downstream tunnel section completely full of vehicles.

4.3. Simulation Settings

Egress simulations by STEPS were carried out for 6 scenarios, differing in behavioural features, characteristic
times and dependence on fire-induced smoke (see Table 1).
After the start of the fire, three time occasions were considered for each person: i) the time when the car is still in
motion; ii) the pre-movement time (the time for detecting, alerting, reacting and leaving the vehicle); iii) the
walking time required to reach a safe place. The times during which the cars are still in motion fall in the range 1-
309 s, with cars gradually stopping every 4 s from the one closest to the HGV up to the farthest one. The pre-
movement time and walking speed were assumed to be random variables. The pre-movement time was firstly
simulated for the Scenarios 1-4 (Table 1) by assuming for each person a normal distribution with: mean=120 s,
σ= 30 s, minimum= 30 s, maximum= 210 s [10]. Then for Scenarios 5-6 (Table 1), a fire alarm system was
assumed to be activated by the heat detection system 2 min after the fire starts, according to CFD fire simulation
184 Ciro Caliendo et al. / Procedia - Social and Behavioral Sciences 53 (2012) 178 – 188

results. The time to react and leave a car after the signal alert was assumed equal to 0.5 min, in agreement with
PIARC [4]. Therefore, simulations were carried out by setting the pre-movement time equal to 0.5 min for the
people in the cars stopping after the alarm, and 2.5 min for people in the 30 cars stopping before the alarm. All
occupants’ initial positions are assumed to be close to the vehicles, hence, the pre-movement time includes the
time required to leave the car. In all the scenarios the tunnel sidewalk is defined as the evacuation walking plane
to which all the car planes, corresponding to the starting point of evacuation, are connected by exits. Occupant
load of 2 people (unimpeded adults) per car and standard dimensions of cars (4.5 x 2m 2) were assumed. Grid size
equal to 0.5 m and time step of 0.1 s were used. With reference to the walking speed, according to PIARC [4] this
lies in the range 0.5-1.5 m/s. In this work, an unimpeded walking speed was assumed for each person according
to a normal distribution [10] with: mean= 1.25 m/s, σ= 0.3 m/s, minimum= 0.95 m/s, maximum= 1.55 m/s.
For each evacuation scenario 35 runs were carried out. This number of runs was identified as that required to
obtain a satisfactory accuracy (a significance level of 0.05) of resulting normal distributions.
The effect of behavioural features (e.g. unrelated people or “family” groups for two occupants each car) was also
assessed. The behaviour of “family” groups considers, as has been aforementioned, that people stay together and
ensure that each member has been evacuated safely, so as a result their walking speed corresponds to the slowest
user while response times and evacuation routes is the same for the whole group.
Finally, the effect of the fire-induced reduced visibility at eye level on walking speed was investigated by
importing smoke data from CFD simulations. Fire simulations were carried out by means of the CFD software
CFX by ANSYS [12]. Details of simulation software and settings were found in a previous work [11]. In the
simulations the longitudinal slope of 2% of the tunnel was considered, the effects of forced ventilation were taken
into account, with the effect of the queuing of vehicles on the fire development and the spreading of smoke and
hot gases. In particular, CFD results in terms of smoke concentration at eye level along the egress path were here
used to calculate the extinction coefficient (K) by Eq 1-2, assuming for HGV: smoke yield ysa 0.025 kg/kg; mass
optical density Dma 100 m2/kg [17]. Light obscuration due to smoke is expressed by the extinction coefficient K
(m-1) or of optical density D (m-1) in Eq. 1, as a measure of the attenuation in intensity I of a light beam passing
through a distance L (m) of smoke:
I
e  KL 10 DL (1)
I0
The optical density D can be expressed in terms of smoke properties:
Cs
D Dm (2)
ys
where Dm is the mass optical density (m2/kg), Cs is the smoke concentration (kg/m3) and ys the smoke yield
(kg/kg). During the evacuation of each person STEPS uses the imported K values at the corresponding position
along the egress pathway to evaluate the walking speed according to Jin & Yamada relationships [15].
The main conditions of simulated scenarios are summarized in Table 1.

Table 1. Summary of the evacuation conditions for the simulated scenarios

Scenario Family Pre-movement time Maximum unimpeded walking speed

Smoke
[s] [m/s]
1 NO mean= 120; σ= 30; min= 30; max= 210 mean= 1.25; σ= 0.3; min= 0.95; max= 1.55 NO
2 YES mean= 120; σ= 30; min= 30; max= 210 mean= 1.25; σ= 0.3; min= 0.95; max= 1.55 NO
3 YES mean= 120; σ= 30; min= 30; max= 210 mean= 1.25; σ= 0.3; min= 0.95; max= 1.55 Non-irritant
4 YES mean= 120; σ= 30; min= 30; max= 210 mean= 1.25; σ= 0.3; min= 0.95; max= 1.55 Irritant
5 YES Related to alarm system and car position mean= 1.25; σ= 0.3; min= 0.95; max= 1.55 Non-irritant
6 YES Related to alarm system and car position mean= 1.25; σ= 0.3; min= 0.95; max= 1.55 Irritant
 Ciro Caliendo et al. / Procedia - Social and Behavioral Sciences 53 (2012) 178 – 188 185

5. Results

5.1. Evacuation times

In Table 2, for each scenario investigated mean data gathered over runs in terms of characteristic times (i.e. the
time when car is in motion, the pre-movement time, the walking time and the evacuation time) and of average
walking speed are reported. Statistical treatment analysis of data over all the 156 users involved provides
estimation of mean, standard deviation, minimum and maximum.

Table 2. Statistical treatment over all 156 people of individual means over runs of characteristic times and average walking speed

Scenario Scenario Scenario Scenario Scenario Scenario

1 2 3 4 5 6
car in motion
time Deterministic values in the range 1-309 s depending on car position along the tunnel
[s]
Mean 119 120 120 119
pre-movement Std.Dev. fixed values fixed values
13.4 6.1 5.8 6.1
time in the range in the range
[s] Minimum 90 105 107 106 30-149 s 30-149 s
Maximum 143 137 139 135
Mean 216 250 264 420 241 305
walking time Std.Dev. 125.5 134.9 145.7 252.5 142.3 241.5
[s] Minimum 2 13 19 19 4 4
Maximum 451 487 540 899 512 873
Mean 490 524 539 695 449 514
evacuation Std.Dev. 42.6 45.4 55.5 162.0 86.5 194.9
time
[s] Minimum 413 442 441 447 342 342
Maximum 590 602 657 1021 662 1023
Mean 1.2 1.1 1.00 0.7 1.1 1.0
average Std.Dev. 0.06 0.09 0.07 0.06 0.03 0.18
walking speed
[m/s] Minimum 1.1 0.7 0.5 0.6 1.0 0.6
Maximum 1.4 1.2 1.1 0.9 1.2 1.2

Such results give a first overall indication of the assessed features:

i) unrelated people (Scenario 1) are able to reach the tunnel exit in a shorter time when compared to family
groups (Scenario 2): the estimated mean values of total evacuation time are 490 s and 524 s, respectively;
ii) the presence of smoke affects the egress process by reducing the walking speed, to a greater extent in the case
of irritant smoke: the mean of the total evacuation time is 539 s in the case of non-irritant smoke (Scenario 3) and
695 s in the presence of irritating smoke (Scenario 4), respectively, in comparison with the value of 524 s in the
absence of smoke (Scenario 2); the average walking speed is estimated to be reduced from 1.1 m/s with no smoke
to 1.0 m/s with non-irritant smoke, down to 0.7 m/s in the presence of irritating smoke;
iii) the presence of a fire alarm system, for triggering the evacuation process, shortens the egress process in
comparison with scenarios with stochastic pre-movement times: the mean evacuation time is 449 s in the case of
non-irritant smoke (Scenario 5) and 514 s in the presence of irritating smoke (Scenario 6) in contrast to 539 s
(Scenario 3) and 695 s (Scenario 4), respectively.
In all the scenarios investigated the evacuation time is primarily influenced by the walking time and to a less
extent by the pre-movement time.
186 Ciro Caliendo et al. / Procedia - Social and Behavioral Sciences 53 (2012) 178 – 188

In Figure 4, the distribution function of evacuation times are reported for all scenarios. In the absence of smoke
(Scenario 2) almost all the people leave the tunnel within 10 min, whereas the presence of non-irritating smoke
(Scenario 3) delays total tunnel evacuation by about 1 min. On the other hand, irritant smoke (Scenario 4) causes
a more significant reduction in walking speed, so that all tunnel occupants can be estimated to exit the tunnel
approximately 17 min after the fire start in contrast to non-irritant smoke where all users can exit the tunnel
within 11 min. Similar total evacuation time values are observed in the presence of an alarm system (17 min and
11 min respectively for Scenario 5 and 6) although some people start to evacuate earlier.

Figure 4. Function of cumulative distribution of the evacuation time for simulated scenarios

5.2. Acceptable safety criteria for tunnel user survival in the event of fire

In the event of a tunnel fire several factors could affect tenable conditions for people, and dangerous effects could
be due to hot and toxic combustion gases as well as to radiation [4,18]. CFD results in terms of temperature,
radiant heat flux and toxic gases concentrations predicted along the egress path were here used, in order to
evaluate the risk of impaired escape, incapacitation or even death due to asphyxiation, hyperthermia or severe
burns. These values with STEPS results in terms of people position over time allows the determination of ASET
(available safe escape time). These values refer to the person closest to the HGV when the fire started, and so
exposed to the most severe fire-induced conditions and also needing to walk the maximum distance to the tunnel
exit portal.
According to the asphyxiation gas model by Purser [19], dangerous conditions due to toxic concentrations were
never achieved. On the other hand, the severe hazard due to convective heat exposure and above all to radiant
heat may endanger and impair escape before the people reached the tunnel exit portal. The fractional dose of total
heat acquired during exposure can be calculated using Equation 3, with the tenable limit predicted when the FED
for heat equals 1:
t2
§ 1 1 ·
FED ¦¨¨© t
t1 Irad
 ¸'t
t Iconv ¸¹
(3)

where tIrad and tIconv are the time to incapacitation due to radiant and convective heat according to the relevant
exposure-time curves, respectively.
The ASETs for people nearest the fire in the event of the various scenarios resulted as follows: in all the
scenarios, the time to incapacitation by total heat was predicted as being equal to a450 s (causing pains) and
equal to a510 s (producing severe burns) both in the case of non-irritant smoke and of irritant smoke. No
 Ciro Caliendo et al. / Procedia - Social and Behavioral Sciences 53 (2012) 178 – 188 187

significant differences in the ASETs in the presence of the alarm system were predicted for the assessed first
person, since his pre-movement time is approximately the same in the event of all four scenarios. The comparison
between the ASET and the required safety egress time (RSET) showed that lethal conditions may occur for the
tunnel users in the event of the hypothesized fire scenarios.
Indeed, in the presence of an alarm (Scenarios 5 and 6) 60 th percentiles of evacuation time are less than 450 s
(ASET for pains) which means that these people have been evacuated safely, while more than 70th percentiles are
less than ASET for severe burns, hence, these people during evacuation are incapacitated by pain due to heat.
On the contrary, longer evacuation times are needed without an alarm system, hence all the people suffered pain
due to the heat. Among them those that evacuated before severe health effects (3° burns) are much more for the
scenario 1 than for the other scenarios.

6. Conclusions

This research was motivated prevalently by the need to make an analysis of people evacuation process from a
curved bi-directional road tunnel in which a fire source due to heavy goods vehicle was considered. For this aim
the worst scenario was investigated, which was identified to be when the HGV was located in the middle of the
tunnel length. People were assumed to be prevented from escaping with their own vehicles, but had to abandon
their cars and walk along sidewalks in the direction of the exit portals. In particular the worst condition for
people’s safety was investigated which, due to the longitudinal ventilation of jet-fans, corresponds to the area
downstream of the fire. In order to achieve the objective, the STEPS software for simulating people’s evacuation
process was used in conjunction with the fire simulation CFX code. On the basis of these simulations the
conclusions given below may be drawn.
The evacuation time is primarily influenced by the walking time and to a less extent by the pre-movement time in
all the scenarios investigated. The mean of the total evacuation time was found to be 539 s in the case of non-
irritant smoke and 695 s in the presence of irritating smoke, in comparison with the value of 524 s, found in the
absence of smoke.
The presence of smoke affects the egress process by reducing the walking speed, to a greater extent in the case of
irritant smoke. The average walking speed was estimated to be reduced from 1.1 m/s with no smoke to 1.0 m/s
with non-irritant smoke, down to 0.7 m/s in the presence of irritating smoke.
The presence of an emergency ventilation system is able to partially control the evolution of smoke during the
fire. On the contrary, the presence of an alarm system can reduce the evacuation time for most tunnel users,
allowing them to reach the tunnel exit in safety. When an alarm system is present the mean evacuation time was
found to be 449 s in the case of non-irritant smoke and 514 s in the presence of irritating smoke, respectively.
The comparison between the available safe escape time (ASET) and the required safety egress time (RSET)
showed that lethal conditions may occur for the tunnel users in the event of the hypothesized fire scenarios.
Although the authors are confident that through carrying out an appropriate analysis they have simulated with
sufficient realism the people evacuation process from the tunnel in the event of a fire, there are still some points
of interest worth investigating. Further research is needed to allow a more definitive analysis of fire safety in
tunnels. This should include more fire scenarios (i.e. tanker burning) and user behaviour (e.g. that of disabled
people), as well as analysing in greater detail the effect of fire parameters on the escape process, in order to
assess the critical distance from the fire for a safe evacuation in all scenarios. Fast and reliable suppression
systems, and/or smoke control systems, should also be better investigated. Therefore, future studies should make
these additional developments possible.

References

[1] Directive 2004/54/EC (2004). “Directive 2004/54/EC on minimum safety requirements for tunnels in the Trans-European Road
Network”. Official Journal of the European Union. L.167, Bruxelles, 30 April.
188 Ciro Caliendo et al. / Procedia - Social and Behavioral Sciences 53 (2012) 178 – 188

[2] Papaioannou, P. and Georgiou, G. (2008). “Human Behaviour in Tunnel Accidents and Incidents: End-users, Operators and Response
Teams”. UPTUN Workpackage 3 Human response D32, http://www.uptun.net.
[3] Persson, M. (2002). “Quantitative risk analysis procedure for the fire evacuation of a road tunnel”. Report 5096, Department of Fire
Safety Engineering, Lund University, Sweden ISSN 1402-3504.
[4] PIARC (1999).“Fire and smoke control in road tunnels”.PIARC Technical Committee C5 Road Tunnel, http://www.piarc.org.
[5] Purser, D.A. (2011).”Design behavioural scenarios for escape behaviour modelling in tunnels and underground complexes”. Proceedings
of Advanced Research Workshop on Evacuation and human behaviour in emergency situations. Santander, Spain.
[6] Yang, G., Peng, L., Zhang, J. and Yong-lin, A. (2006a).“Simulation of people’s evacuation in tunnel fire”. Journal of Central South
University of Technology 13 (3), 307-312.
[7] Yang, G., Peng, L., Peng, J., Zhang, J., Zhao, M. and Yong-lin, A. (2006b). “Cross passage interval of super long tunnel from view of
people’s safe escape”. Proceedings of International Symposium on Safety Science and Technology. Beijing, China.
[8] Liu Y., Apte V., Luong Y., Liu X., Yung D. (2007) “A Methodology for Assessment of Visibility during Road Tunnel Fires”. Journal of
Fire Protection Engineering 17, 65-79.
[9] Kretz, T., Mayer, G., Muehlberger, A. (2010). “Behaviour and perception-based pedestrian evacuation simulation”. Proceedings of
Conference on Pedestrian and Evacuation Dynamics 2010 (PED2010), Gaithersburg, Maryland, U.S.A.
[10] Ronchi, E., Colonna, P., Capote, J., Alvear, D., Berloco, N. and Cuesta, A. (2012). “The evaluation of different evacuation models for
assessing road tunnel safety analysis”. Tunnelling and Underground Space Technology 30, 74-84.
[11] Caliendo, C., Ciambelli, P., De Guglielmo, M.L., Meo, M.G. and Russo, P. (2012). “Numerical simulation of different HGV scenarios in
curved bi-directional road tunnels and safety evaluation”. Tunnelling and Underground Space Technology 31, 33-50.
[12] ANSYS (2010).“ANSYS CFX, Release 13.0”, Copyright© 2010 SAS IP, Inc.
[13] Mott MacDonald (2012).“STEPS Simulation of Transient Evacuation And Pedestrian movementS User Manual (5.0 Version)”. Mott
MacDonald Simulation Group, United Kingdom.
[14] Waterson, N.P. and Pellissier, E. (2012), “The STEPS Pedestrian Microsimulation Tool - A Technical Summary”.
http://www.steps.mottmac.com (March 2012).
[15] Jin, T. and Yamada, T. (1985). “Irritating Effects of Fire Smoke on Visibility”, Fire Science and Technology 5 (1), 79-89.
[16] Norén, A. and Winér, J. (2003). “Modelling Crowd Evacuation from Road and Train Tunnels - Data and design for faster evacuations”.
Report 5127, Department of Fire Safety Engineering, Lund University, Sweden.
[17] FIT-European Thematic Network Fire in Tunnels (2005).“Design Fire Scenarios”. Technical Report Part 1, A. Haack, STUVA.
[18] UPTUN (2008). “Target criteria”.UPTUN Workpackage 2 Fire development and mitigation measures D221. http://www.uptun.net.
[19] Purser, D.A. (2002). “Toxicity Assessment of Combustion Products”.SFPE Handbook of Fire Protection Engineering (3rd ed.), P. J.
DiNenno et al. (Eds.), National Fire Protection Association, Quincy, MA, pp. 2-83–2-171.