ROAD TUNNEL FIRE SAFETY - TU Delft Repositories

Master of Science Thesis

Department of Building Engineering

TU Delft

ROAD TUNNEL FIRE SAFETY:

“Determining the effect of the performance of technological systems

for fire detection on fire detection time and on the total evacuation time”

MSc student:

Marina Fragkopoulou

MSc Thesis Committee:

Prof. Ir. Rob Nijsse (TU Delft)

Prof. Dr. Ir. Serge Hoogendoorn (TU Delft)

Dr. Ir. Roel Schipper (TU Delft)

Ir. Dave Hensen (Deerns Nederland)

Ir. Sven Louwers (Deerns Nederland)

1

ABSTRACT

After a series of tragic accidents related to fires in a number of road tunnels, the global community was

alerted and, thus, new safety regulations and standards were introduced both in an international and a

national level. For that reason, many European countries are confronted with the need to refurbish the

existing tunnel infrastructure that could now be considered unsafe. Nevertheless, the actions and

consequences involved in the renovation of a road tunnel are, in many ways, costly and unpleasant for

a significant number of stakeholders concerned.

For that reason, this research study intends to investigate on alternative solutions that would ensure

the same level of safety that the regulations suggest. As a result, this project will determine whether a

more technologically advanced detection system that offers a timely response to a fire incident could

allow for a compromise in the safety objectives set by the current standards. The aim of this research

study, thus, is to demonstrate that an efficient and well integrated use of the appropriate technological

systems can increase the levels of safety in a tunnel. The main objective is to investigate on the effect

of a technologically advanced detection system on the detection time and, thus on the total evacuation

process.

In order to achieve the aforementioned objectives, simulation models are used in order to model the

performance of different fire detection technologies as well as the evacuation of the pedestrians. As a

result, the performance of the fire detection systems is assessed and their properties will be

determined. This way, the response time of the technological systems in the tunnel will be estimated

and the total evacuation time will be more explicitly specified.

The results of the research study indicated that the Linear Heat Detection system is a reliable system

for fire detection in tunnel applications resulting in a low detection time and accurate detection of the

fire location. It was concluded that the FDS software can be considered a reliable tool for modeling the

Linear Heat Detection technology provided that the proper validation process is followed.

Nevertheless, the modeling of the evacuation process with pedestrian simulation software still needs

to be researched and developed in order to produce reliable results.

2

TABLE OF CONTENTS

ABSTRACT ................................................................................................................................................. 1

TABLE OF CONTENTS ................................................................................................................................ 2

LIST OF FIGURES ....................................................................................................................................... 5

LIST OF TABLES ......................................................................................................................................... 9

PREFACE .................................................................................................................................................10

CHAPTER 1: General Introduction ..........................................................................................................11

1.1 INTRODUCTION ..........................................................................................................................11

1.2 RESEARCH STUDY .......................................................................................................................13

1.3 RESEARCH APPROACH AND STRATEGY ......................................................................................16

1.4 RESEARCH RESULTS ....................................................................................................................19

1.5 CONCLUSION AND DISCUSSION .................................................................................................22

1.6 LAYOUT OF THE THESIS ..............................................................................................................22

CHAPTER 2: Literature review ................................................................................................................24

2.1 INTRODUCTION ..........................................................................................................................24

2.2 REGULATIONS AND STANDARDS ................................................................................................24

2.3 STATE-OF-THE-ART .....................................................................................................................25

2.4 FOCUSED LITERATURE RESEARCH ..............................................................................................29

CHAPTER 3: Tunnel operation ................................................................................................................35

3.1 NORMAL OPERATION .................................................................................................................35

3.1.1 CASE STUDY: HUBERTUSTUNNEL .......................................................................................35

3.1.2 TUNNEL DESIGN CONSIDERATIONS ...................................................................................37

3.1.3 VENTILATION SYSTEM ........................................................................................................42

3.1.3.1 NATURAL VENTILATION ..................................................................................................42

3.1.3.2 MECHANICAL VENTILATION ...........................................................................................43

3.1.4 AIR QUALITY .......................................................................................................................44

3.1.5 TENABLE ENVIRONMENT ...................................................................................................47

3.2 EMERGENCY OPERATION ...........................................................................................................49


3.2.2 EGRESS EXITS ......................................................................................................................52

3.2.3 FIRE SUPPRESSION SYSTEMS ..............................................................................................53

CHAPTER 4: Computational Fluid Dynamics ..........................................................................................54

4.1 INTRODUCTION ..........................................................................................................................54

4.2 THE BASICS OF CFD .....................................................................................................................54

4.3 FIRE DYNAMICS SIMULATOR (FDS) SOFTWARE .........................................................................56

3 4.4 VERIFICATION AND VALIDATION OF SIMULATION MODELS ......................................................58

CHAPTER 5: Fire detection .....................................................................................................................71

5.1 INTRODUCTION ..........................................................................................................................71

5.2 DECISIONS RELATED TO FIRE DETECTION SYSTEMS ...................................................................71

5.2.1 LINEAR HEAT DETECTION (LHD) .........................................................................................72

5.2.2 MULTIPLE GAS DETECTION (MGD) .....................................................................................83

5.2.3 CCTV DETECTION ................................................................................................................84

5.2.4 CONCLUSIONS ....................................................................................................................86

CHAPTER 6: Fire Simulations ..................................................................................................................87

6.1 INTRODUCTION ..........................................................................................................................87

6.2 GENERAL INPUT ..........................................................................................................................87

6.2.1 SIMULATION TIME DURATION ...........................................................................................88

6.2.2 CHEMICAL REACTION FOR FIRE ..........................................................................................89

6.2.3 BOUNDARY CONDITIONS AT THE PORTALS .......................................................................89

6.2.4 FIRE LOCATION ...................................................................................................................91

6.2.5 FIRE SIZE .............................................................................................................................92


6.2.7 GEOMETRICAL MODEL .......................................................................................................98

6.2.8 MESH ..................................................................................................................................99

6.3 FIRE SIMULATIONS ...................................................................................................................100

6.3.1 PASSENGER CAR FIRE .......................................................................................................100

6.3.2 HGV FIRE SIMULATIONS ...................................................................................................119

6.4 CONCLUSIONS ..........................................................................................................................134

CHAPTER 7: Evacuation Simulations ....................................................................................................136

CHAPTER 8: Conclusions .......................................................................................................................155

REFERENCES .........................................................................................................................................159

4

5

LIST OF FIGURES

Fig.1.1 Urban and rural world population, 1950-2050[1] ........................................................................................ 11

Fig.1.3 Research Methodology Diagram ................................................................................................................ 17

Fig.1.4 Normal operation of tunnels - results .......................................................................................................... 19

Fig.1.5 Fire detection technologies - results .......................................................................................................... 19

Fig.1.6 Human behavior in emergencies - results ................................................................................................. 20

Fig.1.7 CFD simulations results ............................................................................................................................. 20

Fig.1.8 Evacuation simulations results .................................................................................................................. 21

Fig.2.1 Tunnel hazard development[16] ................................................................................................................ 32

Fig.3.1 Map of Hubertus tunnel, The Hague (source: www.maps.google.com) ..................................................... 36

Fig.3.2 Wind map of the Netherlands (source:www.knmi.nl) ................................................................................. 37

Fig.3.3 Standard cross-sections for road tunnels .................................................................................................. 40

Fig.3.4-3.5 Cross section of Hubertus tunnel in The Hague (source1: www.siemens.com, source

2:

www.denhaagfm.nl) ................................................................................................................................................ 41

Fig.3.6 A2 tunnel in Utrecht (source1: www.schlijper.nl, source

2: www.artajasa.asia) ........................................... 41

Fig.3.7 Longitudinal ventilation with central fan ...................................................................................................... 43

Fig.3.8 Longitudinal ventilation with jet fans ........................................................................................................... 43

Fig.3.9 Fully transverse ventilation system ............................................................................................................. 43

Fig.3.10 Supply semi-transverse ventilation system ............................................................................................... 43

Fig.3.11 Supply semi-transverse ventilation system ............................................................................................... 43

Fig.4.1 Validation and verification strategy ............................................................................................................ 58

Fig.4.2 Temperature-time graphs for grid sensitivity study .................................................................................... 61

Fig.4.3 Runehamar tunnel longitudinal section ...................................................................................................... 64

Fig.4.4 Mobile fan units at the east entrance of the Runehamar tunnel ................................................................. 64

Fig.4.5 The measurement station 458 m downstream of the fire ........................................................................... 65

Fig.4.6 HRR-time (min) graph for Runehamar fire test T1 ..................................................................................... 67

Fig.4.7 Comparison between simulation and Runehamar test results ................................................................... 67

Fig.4.8 Quadratic fire growth curves based on NFPA204 (2007) .......................................................................... 68

Fig.4.9 Temperature-time graphs for comparison of results of simulation with Runehamar test fire ..................... 70

Fig.5.1 Alarm thresholds for Linear Heat Detection (LHD) technology .................................................................. 73

Fig.5.2 FibroLaser sensor cable type (Type 1) ..................................................................................................... 74

Fig.5.3 FibroLaser sensor cable type (Type 2) ...................................................................................................... 74

Fig.5.4 FibroLaser controller unit ........................................................................................................................... 75

Fig.5.5 Alternative configurations for the FibroLaser controllers and sensor cable................................................ 75

Fig.5.6 A2 tunnel in Utrecht ................................................................................................................................... 77

Fig.5.7 FibroLaser system configuration in the A2 tunnel tests by Siemens .......................................................... 77

Fig.5.8 HRR-time graph for the simulation of the tests by Siemens ...................................................................... 78

Fig.5.9 Temperature profile graph for comparison between simulation results and Siemens test result at

t=1min25sec ........................................................................................................................................................... 80


t=1min40sec ........................................................................................................................................................... 81


t=1min52sec ........................................................................................................................................................... 81

file:///C:/Users/User/Desktop/MSc%20Thesis%20Draft%20-%20Marina%20Fragkopoulou.docx%23_Toc443141184









































6 Fig.5.12 Temperature profile graph for comparison between simulation results and Siemens test result at

t=2min00sec ........................................................................................................................................................... 82

Fig.5.13 Principle of the electronic nose on which the MGD technology is based ................................................ 83

Fig.6.1 HRR-time graphs for a HGV fire and a passenger car fire respectively (based on test data) .................... 88

Fig.6.2 Alternative fire locations studied ................................................................................................................ 91

Table 6.1 HRR for typical vehicles according to experimental data ....................................................................... 93

Fig.6.3 Approximate HRR data from EUREKA EU499 bus fire test (solid line) ..................................................... 94

Fig.6.4 HRR-time curve for passenger car fire ...................................................................................................... 95

Fig.6.5 HRR-time curve for HGV fire ..................................................................................................................... 95

Fig.6.6 Jet fan sizes for unidirectional and reversible models from NOVENCO .................................................... 97

Fig.6.7 Specifications for jet fan models by NOVENCO ........................................................................................ 97

Fig.6.8 Configuration of jet fans in the simulation model ....................................................................................... 98

Fig.6.9 Geometrical models of cut-and-cover and bored tunnel typologies ........................................................... 98

Fig.6.10 Part of the tunnel where the finer grid is applied ...................................................................................... 99

Fig.6.11 Alignment of the meshes with different cell sizes in the model ................................................................ 99

Fig.6.12 Smoke propagation for the basic scenario of the cut-and-cover tunnel typology (Smokeview software)

............................................................................................................................................................................. 100

Fig.6.13 Detection point and detection time for basic scenario of cut-and-cover tunnel (PyroSim software) ....... 101

Fig.6.14 Temperature distribution (oC) for the basic scenario of the cut-and-cover tunnel typology (Smokeview

software) ............................................................................................................................................................... 101

Fig.6.15 Visibility conditions (soot visibility in m) for the basic scenario of the cut-and-cover tunnel typology

(Smokeview Software) .......................................................................................................................................... 102

Fig.6.16 Smoke propagation for the higher cross-section scenario of the cut-and-cover tunnel typology

(Smokeview software) .......................................................................................................................................... 103

Fig.6.17 Detection point and detection time for higher cross-section scenario of cut-and-cover tunnel (PyroSim

software) ............................................................................................................................................................... 104

Fig.6.18 Temperature distribution (oC) for the higher cross-section scenario of the cut-and-cover tunnel typology


Fig.6.19 Visibility conditions (soot visibility in m) for the higher cross-section scenario of the cut-and-cover tunnel

typology (Smokeview Software) ........................................................................................................................... 105

Fig.6.20 Smoke propagation for the alternative ventilation scenario of the cut-and-cover tunnel typology


Fig.6.21 Detection point and detection time for alternative ventilation scenario of cut-and-cover tunnel (PyroSim

software) ............................................................................................................................................................... 107

Fig.6.22 Temperature distribution (oC) for the alternative ventilation scenario of the cut-and-cover tunnel typology


Fig.6.23 Visibility conditions (soot visibility in m) around the ventilators (Smokeview Software) ......................... 108

Fig.6.24 Visibility conditions (soot visibility in m) for the alternative ventilation scenario of the cut-and-cover tunnel


Fig.6.25 Modeling of traffic obstructions in the cut-and-cover tunnel typology (Smokeview software) ................ 109

Fig.6.26 Smoke propagation in the presence of traffic obstructions in the cut-and-cover tunnel typology


Fig.6.27 Detection point and detection time in the presence of traffic obstructions in the cut-and-cover tunnel

(PyroSim software) ............................................................................................................................................... 110

Fig.6.28 Temperature distribution (oC) in the presence of traffic obstructions in the cut-and-cover tunnel typology


Fig.6.29 Visibility conditions (soot visibility in m) in the presence of traffic obstructions in the cut-and-cover tunnel


















































7 Fig.6.30 Visibility conditions (soot visibility in m) in the presence of traffic obstructions in the cut-and-cover tunnel


Fig.6.31 Smoke propagation for the basic scenario of the bored tunnel typology (Smokeview software) ........... 113

Fig.6.32 Detection point and detection time for the basic scenario of the bored tunnel typology (PyroSim

software) ............................................................................................................................................................... 113

Fig.6.33 Temperature distribution (oC) for the basic scenario of the bored tunnel typology (Smokeview software)

............................................................................................................................................................................. 114

Fig.6.34 Visibility conditions (soot visibility in m) for the basic scenario of the bored tunnel typology (Smokeview

Software) .............................................................................................................................................................. 115

Fig.6.35 Smoke propagation in the presence of traffic obstructions in the bored tunnel typology (Smokeview

software) ............................................................................................................................................................... 116

Fig.6.36 Detection point and detection time in the presence of traffic obstructions for the bored tunnel typology


Fig.6.37 Temperature distribution (oC) in the presence of traffic obstructions for the bored tunnel typology


Fig.6.38 Visibility conditions (soot visibility in m) in the presence of traffic obstructions for the bored tunnel



............................................................................................................................................................................. 119

Fig.6.40 Detection point and detection time for the basic scenario of the cut-and-cover tunnel typology (PyroSim

software) ............................................................................................................................................................... 120

Fig.6.41 Temperature distribution (oC) of the basic scenario for the cut-and-cover tunnel typology (Smokeview

software) ............................................................................................................................................................... 120

Fig.6.42 Visibility conditions (soot visibility in m) of the basic scenario for the cut-and-cover tunnel typology

(Smokeview Software) .......................................................................................................................................... 121

Fig.6.43 Smoke propagation for the higher cross-section scenario of the cut-and-cover tunnel typology


Fig.6.44 Detection point and detection time for the higher cross-section scenario of the cut-and-cover tunnel

typology (PyroSim software) ................................................................................................................................. 122

Fig.6.45 Temperature distribution (oC) of the higher cross-section scenario for the cut-and-cover tunnel typology


Fig.6.46 Visibility conditions (soot visibility in m) of the higher cross-section scenario for the cut-and-cover tunnel


Fig.6.47 Smoke propagation for the alternative ventilation scenario of the cut-and-cover tunnel typology


Fig.6.48 Detection point and detection time for the alternative ventilation scenario of the cut-and-cover tunnel


Fig.6.49 Temperature distribution (oC) of the alternative ventilation scenario for the cut-and-cover tunnel typology


Fig.6.50 Visibility conditions (soot visibility in m) of the alternative ventilation scenario for the cut-and-cover tunnel


Fig.6.51 Smoke propagation in the presence of traffic obstructions in the cut-and-cover tunnel typology


Fig.6.52 Detection point and detection time in the presence of the traffic obstructions in the cut-and-cover tunnel


Fig.6.53 Temperature distribution (oC) in the presence of traffic obstructions for the cut-and-cover tunnel typology


Fig.6.54 Visibility conditions (soot visibility in m) in the presence of traffic obstructions in the cut-and-cover tunnel


Fig.6.55 Visibility conditions (soot visibility in m) in the presence of traffic obstructions and turbulence ............. 129

Fig.6.56 Smoke propagation for the basic scenario of the bored tunnel typology (Smokeview software) .......... 130




















































8 Fig.6.57 Detection point and detection time for the basic scenario of the bored tunnel typology (PyroSim

software) ............................................................................................................................................................... 130

Fig.6.58 Temperature distribution (oC) for the basic scenario of the bored tunnel typology (Smokeview software)

............................................................................................................................................................................. 131

Fig.6.59 Visibility conditions (soot visibility in m) for the basic scenario of the bored tunnel typology (Smokeview

Software) .............................................................................................................................................................. 132

Fig.6.60 Smoke propagation in the presence of traffic obstructions in the bored tunnel typology (Smokeview

software) ............................................................................................................................................................... 132

Fig.6.61 Detection point and detection time in the presence of traffic obstructions for the bored tunnel typology


Fig.6.62 Temperature distribution (oC) in the presence of traffic obstructions in the bored tunnel typology


Fig.6.63 Visibility conditions (soot visibility in m) in the presence of traffic obstructions in the bored tunnel


Fig.7.1 Visualization of the evacuation simulation model (Smokeview software) ................................................ 140

Fig.7.2 Agent viewpoint upstream of the fire at t=85 sec in the basic scenario for passenger car fire ................. 142

Fig.7.3 Agent viewpoint downstream of the fire at t=85 sec in the basic scenario for passenger car fire ............ 142

Fig.7.4 Agent viewpoint downstream of the fire at t=110 sec in the basic scenario for passenger car fire .......... 143

Fig.7.5 Agent viewpoint downstream of the fire at t=130 sec in the basic scenario for passenger car fire .......... 143

Fig.7.6 Temperature distribution (oC) in the proximity of the fire at t=115 sec for a passenger car fire ............... 144

Fig.7.7 Smoke propagation and toxicity (FED) in the proximity of the fire at t=115 sec for a passenger car fire . 144

Fig.7.8 Agent viewpoint upstream of the fire at t=85 sec in the higher cross-section scenario for passenger car

fire ........................................................................................................................................................................ 145

Fig.7.9 Agent viewpoint downstream of the fire at t=85 sec in the higher cross-section scenario for passenger car

fire ........................................................................................................................................................................ 146

Fig.7.10 Temperature distribution (oC) in the proximity of the fire at t=155 sec for the higher cross-section

scenario for a passenger car fire .......................................................................................................................... 146

Fig.7.11 Smoke propagation and toxicity (FED) in the proximity of the fire at t=155 sec for the higher cross-

section scenario for a passenger car fire .............................................................................................................. 147

Fig.7.12 Agent viewpoint upstream of the fire at t=43 sec (point of fire detection) in the basic scenario for HGV

fire ........................................................................................................................................................................ 148

Fig.7.13 Agent viewpoint downstream of the fire at t=43 sec (point of fire detection) in the basic scenario for HGV

fire ........................................................................................................................................................................ 148

Fig.7.14 Agent viewpoint upstream of the fire at t=115 sec in the basic scenario for HGV fire............................ 149

Fig.7.15 Agent viewpoint downstream of the fire at t=115 sec in the basic scenario for HGV fire ....................... 149

Fig.7.16 Temperature distribution (oC) in the proximity of the fire at t=43 sec (fire detection point) for the basic

scenario for a passenger car fire .......................................................................................................................... 150

Fig.7.17 Smoke propagation and toxicity (FED) in the proximity of the fire at t=43 sec (fire detection point) for the

basic scenario for a HGV fire ................................................................................................................................ 150

Fig.7.18 Agent viewpoint upstream of the fire at t=55 sec (point of fire detection) in the higher cross-section

scenario for HGV fire ............................................................................................................................................ 151

Fig.7.19 Agent viewpoint downstream of the fire at t=55 sec (point of fire detection) in the higher cross-section

scenario for HGV fire ............................................................................................................................................ 152

Fig.7.20 Temperature distribution (oC) in the proximity of the fire at t=55 sec (fire detection point) for the higher

cross-section scenario for a HGV fire ................................................................................................................... 152

Fig.7.21 Smoke propagation and toxicity (FED) in the proximity of the fire at t=55 sec (fire detection point) for the

higher cross-section scenario for a HGV fire ........................................................................................................ 153

Fig.7.22 Smoke propagation and toxicity (FED) in the proximity of the fire at t=140 sec for the higher cross-

section scenario for a HGV fire ............................................................................................................................. 153


















































9

LIST OF TABLES

Table 2.1 Summary of recommended walking speed ............................................................................................ 33

Table 3.1 Traffic data for Hubertus tunnel vehicles/year ....................................................................................... 36

Table 3.2 CO concentration by traffic situation for the years 1995 and 2010 ........................................................ 45

Table 3.3 Visibility conditions by traffic situation .................................................................................................... 45

Table 3.4 Threshold concentrations for CO and NO2 according to various regulations ........................................ 45

Table 3.5 Alarm thresholds for various contaminants for multi-gas sensor PHD6 by Honeywell ........................... 46

Table 3.6 Exposure time and Incapacitation limits................................................................................................. 48

Table 4.1 Results of the model for fast fire development based on quadratic fire curves ...................................... 69

Table 7.1 Unimpeded walking speeds and body dimensions in FDS+Evac software .......................................... 139

Table 8.1 Overview of the results regarding the detection time in the various fire simulation scenarios ............. 155











10

PREFACE

The present report is the outcome of my MSc Thesis which was the final step for completing my

Master in Civil Engineering with specialization on Building Engineering and, more specifically, in

Building Technology and Physics. The topic of this research study was suggested by Mr. Antoon van

Rooijen during an interview in Deerns Nederland B.V. So to him, first of all, I would like to express my

gratitude for introducing to the interesting topic of fire safety in tunnels and for offering me an

internship in Deerns. Moreover, I would like to express my appreciation and gratitude to the people

representing Deerns who gave me the chance to attend a conference on Tunnel Fire Safety and at the

same time make a publication regarding my research in the COB technical magazine. They offered me

a valuable experience that had a major impact on my study.

Following, I would like to thank all the members of my committee for their constant support and

guidance throughout the research study: the committee chairman Prof. Ir. Rob Nijsse for his

contribution to the thesis content and for his help with organizing the regular meetings; Prof. Ir. Serge

Hoogendoorn for sharing his valuable knowledge and references and for referring me to

acknowledged professionals specialized on the research topic; Ir. Roel Schipper for his constant

presence offering guidance and advice and for helping with organizational matters; my supervisor in

Deerns Ir. Dave Hensen for sharing his deep and thorough knowledge on fire safety issues and for

guiding me through the details and the most scientific matters of my research study; finally, my special

thanks to my daily supervisor in Deerns, Ir. Sven Louwers, who was more of a coach and a mentor

and who believed in me throughout the research and helped me tackle many difficulties on the way.

I would, also, like to express my sincere gratitude to Mr. Leo Knies working in Siemens for sharing his

valuable knowledge and time in order to provide data for a fire detection technology. Without his help, I

am convinced that this research could not be accomplished. Furthermore, I would like to thank Mr.

Gerard Slootjes from Gemeente Den Haag for agreeing to share the results of wind measurements

conducted in Hubertustunnel that significantly raised the quality of my results.

Throughout this period, many people be it professionals or university teachers, contributed with their

knowledge to this research study and their help is deeply appreciated. Specifically, I would like to

thank Dr. Ir. Winnie Daamen and Drs. Erica Kinkel for their very interesting interviews on the topic of

human behavior during emergencies.

Additionally, I would like to sincerely thank all the people working in Deerns for being always willing to

offer their help and support in every way and for creating a friendly and warm working environment for

nine months. I would like to specially thank Mr. Rob Kisjes for his guidance and advice on my research

study, Mr. Arjan Pleysier for his advice and knowledge on the issue of wind implementation in the

research study, Mr. Goffe Schat and Mr. Henk Liberg my neighbors in the office who offered me tips

and advice and made my days in the office more interesting.

Finally, I would like to warmly thank all those people that stood by me through this challenging period.

First of all, I thank my family for their understanding, support and long distance motivation and love.

Special thanks to Konstantinos Labropoulos that without his help and contribution I would not be able

to complete my studies. Then, my friends in Delft and in Greece and my roommates for dealing with

my moody days and for keeping me company during the working days and nights: Vasilis, George,

Luis, Elli, Eftichia, Stavroula, Vasilis, Alexis, Manos, Diamantis, Sifis, Pantelis, Tilbe thank you all!

Marina Fragkopoulou

Delft, February 2016

11

CHAPTER 1: General Introduction

This introductory chapter aims to familiarize the reader with the topic and illustrate the importance of the research study. First, the significance of road tunnels is explained. Secondly, the challenges involved in the fire safety design of a road tunnel are shortly described. This way, the reader will gain an insight on the challenging task of proper fire detection in road tunnels.

1.1 INTRODUCTION

why are tunnels important?

In the era of an ever increasing urban development and population growth, the exploitation of the

underground space has become an important subject [2]. As a result of a significant augmentation in

traffic densities, highways and other types of transport links are expected to be brought to a breaking

point of their capacity in the following 20 to 30 years [3]. For that reason, there is a tendency towards

the development of tunnels in order to exploit space more efficiently and, at the same time, decrease

transportation time. More specifically, this global tendency results in a significant number of tunneling

activities realized worldwide in a time frame of 10 to 15 years [3].

As mentioned above, the continuous urban development and the shift of a significant part of the global

population to the cities both result in the need for more underground infrastructure. In that case, it is

interesting to consider in more detail, the percentage of the world’s population that lives in the cities.

Namely, in a global scale, 54 per cent of the population resides in urban areas in the year 2014. In

order to make a comparison, it is interesting to note that in 1950, 30 per cent of the global population

used to live in cities whereas, by 2050, 66 per cent of the world’s population is expected to be urban

[1]. The tendency of the urbanization can be clearly seen in the next graph.

It is true that the levels of

urbanization, largely, depend on

each region. Namely, in 2014, 73

per cent of Europe’s population is

residing in urban areas and this

figure is expected to reach 80 per

cent by 2050.

Fig.1.1 Urban and rural world population, 1950-2050[1]

Fig.1.2 Urban and rural population as proportion of total population[1]

12 Admittedly, tunnels and underground transport facilities are important transportation channels, not only

in terms of shorter journey times but increasingly out of consideration for the local population and the

environment. At the present time, in a worldwide scale, a number of several thousand tunnels existing

in the road networks can be reported [4]. In Europe alone, the transportation tunnels cover an overall

operational length of more than 15,000 km [3].

what about safety?

Generally speaking, important underground transport links are expected to be available without any

restrictions and to operate smoothly around the clock. Any interruptions due to accidents, technical

malfunctions or maintenance work could instantly result in traffic jams and delays. That is to say that

high levels of safety and reliable availability are, especially, important for that type of infrastructure.

This particularly refers to fire incidents in tunnels which constitute a major hazard to human life [3].

Coupled with that, tunnel fires often, also, result in costly damage to surrounding infrastructure [5].

challenge 1: tunnel fire accidents

Although, accidents related to tunnel fires appear to occur in a less frequent basis than open road

accidents, there is no doubt that their impact can be much more significant [6]. Because of the

confined space, limited escape facilities and difficulties encountered by intervention forces in gaining

access to the tunnel fire impose the need for extensive safety arrangements which must be

complementary and mutually coordinated. The rising traffic densities and the growing demand for

underground communication links result in a higher probability of accidents, injuries and damage.

Added to this, there are other factors which increase the potential hazards of traffic tunnels:

The increasing length of modern tunnels

The transportation of hazardous materials

Two-way traffic (with undivided carriageways)

Higher fire loads due to growing traffic volumes and higher loading capacities

challenge 2: particularity of each tunnel case

What, also, constitutes a challenge is the fact that each tunnel has its unique characteristics and

requires to be treated as an individual case. For that reason, it is considered complicated for the tunnel

designer to identify a specific methodology in order to provide the required degree of safety [7]. At this

moment, the current standards and guidelines are being used for a prescriptive approach in decision-

making with regard to tunnel fire safety. It is true, though, that the background of the choices involved

in the establishment of those standards is, rarely, known by practicing engineers. Furthermore, by

using a completely prescriptive approach, the designers, users and operators of tunnels are not

effectively aware of the risks related to the three aspects mentioned above [6]. Under those

circumstances, a shift towards performance-based decision-making is being discussed associated

with risk and type of tunnel [5].

The question that rises is whether the existing standards can provide an efficient solution to the

complicated and ever changing issue of tunnel fire safety. In addition, a performance-based approach

requires significant knowledge and experience together with specialized tools to account for a proper

design for each individual tunnel.

challenge 3: human factor

In the confined environment of a tunnel, in case of a fire incident, a human life is threatened in various

different ways (inhaling of products of combustion and exposure to extremely high temperatures and

heat fluxes). Moreover, the people involved in the incident should leave their vehicles and self-

evacuate from the nearest exit and access the specially designated egress routes. Nevertheless, self-

13 evacuation is often obstructed by crashed vehicles or blocked exits because of traffic jams, by poor

visibility conditions because of the presence of smoke, by a collapse of the structure or an explosion or

by a power failure. Added to all the above, is the unpredictable human factor. People react differently

under stress and a careful consideration of their reactions should be made based on disaster

psychology [8]. As a result, in order to ensure a safe environment for self-evacuation, there is the need

to maintain tolerable temperatures, adequate visibility and air quality levels inside the tunnel [9].

challenge 4: coordinating the technological systems

With the intention to provide the prerequisites related to fire safety and to best respond to the

challenges involved in the safety design of a road tunnel, a number of different safety systems are

developed and installed [6]. In the fire safety design of a road tunnel, apart from the proper choice of

fire safety measures, a designer has to face the challenge of combining the systems in such a way

that is time and resources efficient and, at the same time, provides the required levels of safety and

ensures the compliance with the safety regulations and standards.

Resulting from the challenges identified and mentioned above, the focus of the research will be set on

the characteristics of fire detection technologies, the human factor and its behavior under emergency

situations and, finally, the parameters that affect the pedestrian evacuation. Regarding the fire

detection technologies, despite the apparent importance of the detection systems, standards in a

worldwide scale regarding the engineering and performance requirements were not defined [10].

Nevertheless, there is a significant number of suppliers and tunnel designers that are highly interested

in the performance and response of the fire detectors. On the other hand, the human factor during

emergency situations is a widely investigated topic that interests the academic community. Finally, the

pedestrian evacuation is of interest both for the academic research but , at the same time, for many

stakeholders that are involved in the tunnel fire safety design.

For all the aforementioned reasons, the goal of this research study would be to enhance the

knowledge on the broad area of road tunnel fire safety. With a focus on the safe evacuation of a

tunnel, the present work will investigate on the aspects that really affect the response time in the event

of a fire emergency.

1.2 RESEARCH STUDY

research questions

All in all, after conducting a preliminary research on the topic of tunnel fire safety and through an

inductive thinking process, the general idea of what this thesis would be about was formulated. The

absence of knowledge on the background of the current regulations and standards and the literature

gap related to the performance requirements of the detection systems were the main motives that

stirred my personal interest about the topic. As a result from all the above, the research topic can be summed up to the following main proposal:

Determining the effect of the performance of technological systems for fire detection on fire detection

time and on the total evacuation time

From the initial background research and the analysis of the main proposal mentioned above, the

following research questions are proposed.

What where the criteria for setting the common best practice and what do they mean for the

fire safety design of a road tunnel?

14

What are the properties of state-of-the-art fire detectors regarding their response times?

What are the conditions in the tunnel that would trigger a fire alarm as far as temperatures and

smoke are concerned?

What is the impact of an early fire detection on the safety design of the tunnel as described by

the current regulations?

What does that mean for the need to refurbish a road tunnel?

research aims and objectives

The study intends to investigate on the reasoning behind the current regulations, research on the

properties of the detection systems and assess whether an improved detection system could eliminate

the need for the refurbishment of a road tunnel.

The aim of this research is to demonstrate that an efficient and well integrated use of the appropriate

technological systems can increase the levels of safety in a tunnel. In a broader sense, on the one

hand, the aim is to cast a critical look into the existing regulations which impose the need for costly

tunnel refurbishments in most of European countries. On the other hand, this research aims to

examine whether a set of performance requirements for the technological systems of a tunnel can

compensate for the current more robust requirements and standards.

It is important to mention, at this point, that the aim of this research study could not be accomplished

without a thorough and comprehensive understanding of the normal operation of a road tunnel. For

that reason, the overall and initial aim of this research study would be to investigate into the elements

that comprise a complete and safe tunnel concept in order to, subsequently, determine the operation

of the tunnel systems in case of an emergency.

More specifically, the main objective of this research study is to investigate on the effect of a

technologically advanced detection system on the detection time and, thus on the total evacuation

process. In order to achieve this, the following sub-objectives should be set:

Identify the factors that were determinant for the set of the current regulations

Investigate the performance attributes of current fire detection technologies for roadway tunnel

protection

Determine the conditions needed to activate the detectors (temperature, visibility, smoke and

gases)

Assess the impact of a reduced detection time on the total evacuation time

Assess the impact of a reduced evacuation time on the application of the regulations

The personal study goals to be achieved throughout this research study are several and directly

related. In the first place, I am presented with the opportunity to combine the knowledge of the two

specializations I gained during my studies so far. On the one hand, the knowledge in Civil Engineering

I obtained during by Bachelor studies (specialization in Transportation and Planning) and, on the other

hand, the knowledge in Building Engineering (specialization in Building Physics) which I gained during

my MSc in TU Delft. Moreover, gaining personal experience on safety design and simulations of

15 airflows and egress in the demanding case of a road tunnel will significantly enhance my knowledge

and my competences.

In my opinion, what is particularly interesting is the fact that this topic is of great importance to the

global community. Many researchers and professionals are especially interested on filling the gaps in

knowledge in the urgent matter of tunnel safety. That became even more apparent by the responses I

got during my preliminary research by distinguished people of the international community. For that

reason, I strongly believe that I could be benefited by their expertise and experience.

What is more, two different types of simulation software are going to be used which will familiarize me

with the principles of simulating and assessing the reliability of simulation results. Coupled with that,

the opportunity is given to learn in depth the software used and be able to use it later in my work. Last

but not least, a critical approach on results and existing research, on the whole process of conducting

the investigation and on the information to include will enhance my perception and insight on scientific

matters.

16

1.3 RESEARCH APPROACH AND STRATEGY

At this point, it is considered important to describe the approach that is going to be followed throughout

this research study. This approach will help define a framework regarding the collection of information,

the data interpretation and their analysis. According to the content of the research study, the research

method to be followed will be defined in order to answer the research questions.

research approach

The investigation of the detection technologies requires a quantitative approach. The performance of

the fire detectors will be assessed based on certain parameters which are related either to the

specifications of the detectors or to the fire characteristics. These parameters include the response

time of the fire detection technologies, the set-points in which the detectors are activated, the different

fire scenarios and the fire development. The quantitative results will be derived from the simulations of

the fire detection operation.

The investigation of the evacuation process and the human behavior requires a more qualitative

approach. That is because of the unpredictability of the human behavior and the lack of experimental

data. This means that the relevant literature research will be studied in order to form qualitative

conclusions and assumptions regarding the human response in fire emergency and suggest design

solutions that have been proved to stimulate the human reactions.

research strategy

In order to retrieve the data mentioned above and, then, answer the research questions, a research

strategy had to be formulated. In the context of this research study, the strategy comprises of the

following three parts:

On the one hand, an extended literature survey will be conducted in order to gain an overview

of the problem. More specifically, a comprehensive insight into the basic concept of the tunnel

operation and safety measures should be gained by taking into account the structural design,

the egress routes, the ventilation system, the detection system and the fire-fighting system. Research will focus on the detection technologies and on the factors that strongly affect the

evacuation in a tunnel. On the same context, a thorough study of the current regulations and

standards will be carried out especially regarding the emergency exits and provisions. The

regulations will provide the point of reference for comparing the data collected or obtained

through numerical simulations.

The literature research will include recent scientific papers that deal with the developments in

fire detection technologies and types of heat detection and the way to model their

performance. Moreover, scientific papers regarding the parameters that affect the human

behavior during evacuation will be studied. Through this knowledge, the input parameters for

the simulation models will be better defined and a higher level of reliability of the results will be

achieved. In addition, it has already been mentioned that the topic of tunnel fire safety is on

the spotlight the recent years. For that reason, many conferences are held dealing with the

most important aspects of tunnel fire safety. The proceedings, reviews and presentations of

these conferences will be studied in depth for two reasons. Firstly, to identify the issues that

gather most of the interest and for which more research is required. Secondly, to gain the most

recent knowledge and determine the trends in tunnel fire safety solutions.

On the other hand, experiment in the form of numerical simulations will be conducted in order

to obtain data regarding the fire detection and the evacuation process. More specifically,

computational fluid dynamics (CFD) models are going to be used in order to simulate the

17

conditions in the tunnel during a fire incident. Moreover, evacuation models will be used to

simulate the pedestrian flows and choice of evacuation routes.

The CFD models will provide information regarding the response time of the fire detectors

which depends on the physical characteristics of the fire and the conditions inside the tunnel.

The aim is to reduce the detection time in the scale of between one to two minutes compared

to the current standards so that the evacuation process can start earlier. The evacuation

models previously mentioned will help to determine the importance of an early start of the

evacuation procedure on the chances to survive the fire.

Complementary to the aforementioned investigation methods, discussions and interviews with

professionals and people involved in the tunnel infrastructure will be carried out. This way,

valuable knowledge can be obtained by their expertise and experience. Moreover, a better

insight into the realistic aspects of the problem of tunnel fire safety can be gained. As a result,

the conclusions will correspond more accurately to the reality.

To sum up, the research methodology is presented schematically in the following graph.

alternatives

As was already explained, the core of the research study will be the investigation of the detection

technologies using simulation models. This research method was considered suitable for the specific

research objectives since it can simulate different fire scenarios and compare the response time of the

various detection technologies. It is a method that is financially affordable and ethically acceptable.

Nevertheless, the use of simulation models involves two main issues. On the one hand, the

simulations can result in high computational time, limiting the number of scenarios that is feasible to be

examined within the time frame of the research study. In order to resolve this issue, a very limited time

scale was chosen for the simulations that was the most relevant for this research. More specifically,

only the first 3 minutes of the fire incident will be modelled since these are the ones that are related to

Fig.1.3 Research Methodology Diagram

18 the detection activation. In addition, the most representative fire scenarios were chosen so that the

number of required simulations could be reduced. Those two actions are expected to result in lower

computational cost. On the other hand, designers and engineers should make very considerate use of

such models in order to ensure the reliability of the results. In order to solve this second issue, a

validation of the models that are going to be used has been planned and will be described in a

preceding section.

As an alternative to the chosen research strategy, experiments in the form of full-scale tests could be

conducted. Those tests could be performed in tunnels where the detection technologies could be

installed. Although this method would be more accurate and the reliability of the results would not be

questioned, there are several drawbacks involved. First, those tests would require a lot of resources

because of their large scale. They are costly to perform and require appropriate equipment and safety

provisions. In addition, they would be time consuming and require trained staff. Last but not least, in

the case that people would be needed to participate, ethical issues would arise regarding the exposure

of people to a highly toxic environment.

All in all, the simulations were considered as the best choice since taking into consideration the time

frame and the resources of the present research study. Moreover, the experience and scientific

background of the researcher and the supervisors and partners allows for the appropriate use of the

simulation models.

19

1.4 RESEARCH RESULTS

As it has been, already, mentioned in the previous section as a research strategy, the data will, mainly,

be collected by three sources: an extended literature survey, numerical simulations and interviews with

professionals and academics. The results that will be retrieved by these sources will serve different

purposes that will be described in the following paragraphs.

literature survey results

The results derived from the literature research will, mainly, regard the input for the simulation models

as well as some recommendations for the integration of the different technological systems. More

specifically, the research will focus on the normal operation of the road tunnels, the detection

technologies and the characteristics of human behavior in fire emergencies. Each of these categories

includes different results that are going to be used and presented in different ways.

Normal operation of road tunnels: The results in this category, mainly, concern the operation

of the ventilation system on the tunnels, the safety signs and the general design of the tunnel.

As shown on the diagram, the results from the

research will be presented in sketches and text

in order to help understand and define the

normal operation of a road tunnel. In addition,

the literature study will provide information on

how to build the models that are going to be

used for the simulations. Namely, the geometry

of the tunnel will be decided based on the

results from the literature research. In addition,

the number of ventilators and their distances in

the longitudinal section will be defined as

well as their specifications and performance

requirements.

Fire detection technologies: The results from the literature research in this category, mainly,

concern the specifications of use of the various detector types and tests that were performed

on fire detectors by suppliers.

The results from the research will be presented in

tables, text that will be explaining the basic principles

of the detectors’ operation and sketches indicating the

location and configuration of the detectors. Once

again, the information provided by the literature

research will help to build the simulation model by

implementing correctly the detection technologies

before performing any simulations.

Human behavior in fire emergencies: The

results from the literature research in this category,

mainly, concern the reactions of people during an

emergency situation in road tunnels and the parameters that affect the decision to evacuate.

Moreover, the results will include design considerations regarding the safety signs and exits

and their location in the tunnel as well as the voice alarm message type.

sketches and text

ventilation system

configuration

safety signs and exits location

design of the tunnel

tables, text and sketches

response time of fire

detectors

installation guidelines of fire detectors

real fire tests on fire

detectors

Fig.1.4 Normal operation of tunnels - results

Fig.1.5 Fire detection technologies - results

20

CFD SIMULATIONS

RESULTS

Fire curve

(HRR - time, temperature - time)

Visualisations

(smoke development, smoke layer thickness, visibility conditions)

Excel Sheets

(HRR - time, temperature - time,

detector activation time)

As shown on the diagram on the right,

the results from the research will be

presented in text and some sketches

indicating the location of the emergency

exits. The exits are going to be designed

in the simulation model according to the

guidelines that are going to be obtained

through the research.

numerical simulations results

On the context of this research study, two types of software are going to be used. As mentioned

before, simulations of fire dynamics and evacuation flows need to be provided. Using the results of

both types of simulations, the correlation between the response of the technological systems of the

tunnel and the evacuation time will be examined.

Computational Fluid Dynamics Simulations: The results drawn by these simulations will regard

the conditions and time needed for the integrated system to respond. More specifically,

information will be provided about the fire development in the tunnel, the response times of the

different detector technologies and the smoke development related to visibility.

As shown in the above graph, the results are going to be presented in a way that enables

observation and comparison. As was, already, mentioned, the aim of this research study is to

determine different scenarios and define which detection technology would be more

appropriate for each scenario. For that reason, the results will be presented using

visualizations, like graphs and images, in order to better observe the differences between the

various cases.

text and sketches

human response in

fire emergencies

impact and configuration

of signs

configuration of emergency

exits

voice alarm message

Fig.1.6 Human behavior in emergencies - results

Fig.1.7 CFD simulations results

21

Pedestrian Evacuation Simulations: Evacuation simulation models can be built and results

about the quality and the safety levels of the tunnel can be shown by assessing normal

operation and conducting emergency evacuation scenarios. The behavioral characteristics of

the evacuees are determined by factors such as the walking speed and the list of activities

performed. After creating the model, the pedestrian flow can be simulated by running an

experiment and the outputs are easy to examine. Finally, measurements can be conducted

regarding the walking times and pedestrian densities in the tunnel.

As shown in the previous graph, the results are going to be presented in tables and Excel

sheets in order to offer a better overview of the different simulation scenarios performed.

Added to these, it would be interesting to present certain visualizations that better depict the

flows of people in case of a fire emergency in a tunnel.

validation of the simulation models

Apart from the general analysis of the qualitative and quantitative results that were previously

discussed, it is very important to ensure a proper validation. As was described in the previous

sections, in order to achieve the desired outcome, computational models will be used. In that case, an

assessment of the accuracy and reliability of those models should be carried out in order for the model

to become accepted as a part of the fire safety decision making [11]. The question that arises, thus, is

how confidence in modeling and simulation can be critically assessed [12].

There is a number of test programs regarding tunnel fire scenarios, already conducted and for which

the results are available. The methodology that is going to be used for the assessment of the value of

the simulation models will be that of comparison with the results of experiments. As a result, using the

inputs and results of the aforementioned tests, a comparison between the CFD models used and the

results of the tests will be conducted. Moreover, as was previously mentioned, during the literature

research and the interviews, real fire tests of fire detectors will be obtained. By simulating these fire

tests in the CFD Software and comparing the output with the results of the real tests, a validation of

the simulation of the detectors’ algorithm can be achieved. This could be adequate proof that the

detectors were accurately simulated and that the results can be considered reliable.

interviews results

The results that will be derived from the interviews, mainly, regard the assumptions made but, also,

some technical specifications for the systems of tunnel. The results of the interviews will be found

throughout the research study in order to support certain decisions. They will be presented as text

since they have a more explanatory purpose.

EVACUATION SIMULATIONS

RESULTS

Tables

(evacuation time for different scenarios)

Visualisations

(way-finding: exit choice, flows and bottlenecks)

Excel Sheets

(simulations scenarios and their

characteristics)

Fig.1.8 Evacuation simulations results

22

1.5 CONCLUSION AND DISCUSSION

The final conclusion of this project will have two aspects:

On the one hand, the performance of the fire detection systems will be assessed and their properties

will be determined. This way, the response time of the technological systems in the tunnel will be

estimated and the total evacuation time will be more explicitly specified. More specifically, with the

information provided by the simulations, the conclusions drawn will relate to the conditions needed to

activate the detectors and the response time of the detection system. The final aim is the timely

activation and coordination of all the technological systems involved in the evacuation process.

On the other hand, either a validation or skepticism will be introduced regarding the current regulations

and standards. As a result, a critical assessment will be presented regarding the need to intervene on

the existing tunnels in order to comply with the regulations. Together with the detection times that will

be determined, the total time needed for evacuation will be investigated for different scenarios.

Finally, in order to draw relevant conclusions, a comparison of the results of the simulations of the

aforementioned scenarios will be conducted. The main criteria that will be used for the comparison will

be the total evacuation time and the distance between the emergency exits in relation to the one

determined by the standards. The overall objective is to conclude on the effect of the detection

systems on the total evacuation time and on the tolerance regarding the compliance with the

regulations.

The outcome of this research study will provide useful information for the decision-making during the

fire safety design of a road tunnel. In the overall shift towards performance-based design, this study

could prove that with the right tools and expertise, each individual case can be separately and, yet,

successfully addressed. Nevertheless, it would be, also, important to highlight all the issues and

problems involved in such a process in order for a designer to be aware of the advantages and

drawbacks before aiming for a performance-based approach.

In addition, this study will draw important conclusions and make recommendations for practice

regarding the various detection technologies. That will help a number of stakeholders to suggest

solutions that will result in safer tunnels. For example, suppliers of detection technologies and

consultancy firms could offer guidelines based on this research study on the appropriateness of each

technology for each tunnel.

Last but not least, part of the conclusion would be to make recommendations for future research.

Through a preliminary research, it was already realized that the area of tunnel fire safety is a very

complex one involving many components and many issues. For that reason, not all the literature gaps

could be filled in by this research study and many interesting issues that need to be further examined

will arise.

1.6 LAYOUT OF THE THESIS

In this research study, various topics of different research fields are addressed. For that reason, a

clearly structured layout should be chosen so that the reader can follow the logical continuity of the

information. As a result, each chapter consists of two parts: the first part contains the theory behind

the choices and the tasks performed within the research study; the second part contains a description

of the input parameters used in every task together with some basic conclusions.

In addition, there is another general distinction between the chapters of the thesis. Chapters 2 and 3

contain more theoretical information regarding the review of the existing literature, the current

practices in the topic of interest and information on the operation of road tunnels.

23 More specifically, the thesis is comprised of the following chapters:

CHAPTER 1: A general introduction presenting the area of interest of this research study, the challenges

related to the fire safety design of tunnels and the main goal of this work. The formulation of the

research problem and research objectives together with a description of the research approach and

the strategy to be followed was presented.

CHAPTER 2: A review of the existing literature that is related to the chosen topic. Any literature gaps on the

area of fire safety of tunnels will be determined in order for this research study to deal with them and,

as a result, add to the general knowledge.

CHAPTER 3: A description of the types and operation of road tunnels. Information on both the normal and

emergency operation of the road tunnels together with the systems involved will be presented. The

choice of the road tunnels to be simulated will be indicated and motivated.

CHAPTER 4: An insight into the basics of fire detection: the role of fire detection in a road tunnel, the types

of fire detection available in the market and their working principles and performance. In addition, the

choice of specific types of fire detectors to be simulated will be justified.

CHAPTER 5: A description of the basics of Computational Fluid Dynamics (CFD). The basic principles of

the software used will be mentioned in order to better understand the input and output of the

simulations. The crucial issue of validation of the numerical models will be dealt with and the validation

cases will be presented.

CHAPTER 6: A detailed description of the simulation scenarios, the relevant input and the respective

results. The output of the simulations is going to be presented and shortly analyzed. The most

important results are going to be shown by means of visual output and tables.

CHAPTER 7: A description of the evacuation simulations and the input used. The output of the simulations

is, also, going to be presented and shortly analyzed. The most important results are going to be shown

by means of visual output and tables.

CHAPTER 8: A formulation of conclusions regarding the outcome of the research. A description of the

connection of the conclusions to the research objectives is going to be presented. The general impact

of the conclusions is, also, going to be discussed. Finally, recommendations for future research and

recommendations for practice will be suggested

24

CHAPTER 2: Literature review The purpose of this chapter is to present the literature that was studied in order to help formulate the topic of the research study and define the focus. At the same time, it aims to present the relevant research already performed in the field of tunnel safety. The basic literature studied was relevant with the regulations for tunnel safety, the normal and emergency operation and systems of tunnels and the most important aspects of human behavior before and during evacuation.

2.1 INTRODUCTION

At this point, the starting points and the basis on which the research study will be further

developed need to be described. In order to comprehend the complicated subject of tunnel fire

safety, references to the current regulations, the state of development of the fire protection

systems and the evacuation parameters need to be made. As a result, a review of the existing

research that is relevant to the most important aspects of fire safety design in road tunnels was

carried out and will be presented in this chapter.

In addition, the topic of tunnel fire safety is a multi-dimensional topic with many aspects that are

interesting to be researched upon. For that reason it was considered necessary to assess the

existing research and determine the literature gaps in order to narrow down the focus of this

study. Nevertheless, because of the special importance and challenges involved in the tunnel fire

safety design, a significant amount of scientific research is available in the form of scientific

papers, books, handbooks and conference reports. In the context of this research study, the most

relevant ones that were, also, considered most important and complete are going to be

mentioned. This way, the following literature review will offer to the reader an overview of the

most important material that contains all the information related to a wide range of aspects related

to tunnel safety.

2.2 REGULATIONS AND STANDARDS

As has been previously noted, safety is one of the most significant aspects to be considered in

the design of a road tunnel system [13]. The challenge for the tunnel designer is to ensure an

optimal guaranteed level of safety while, at the same time, maintaining the construction and

operation costs at a reasonable level [14]. According to statistics of four countries with developed

tunnel infrastructure (France, Germany, Switzerland and Italy) the chance of a fire related

accident in a road tunnel is relatively small compared to that of an accident in an open road.

Namely, according to French statistics, for every 100,000,000 cars that cross the tunnel, one or

two fires are expected to occur per kilometer of the tunnel. Likewise, for every 100,000,000

Heavy Goods Vehicles (HVG) crossing the tunnel, approximately eight fires are expected to occur

per kilometer of the tunnel [15]. However small the chance of a fire incident might appear at first

glance, the high traffic densities in road tunnels together with the significantly large tunnel length

in Europe, can offer the ground to reconsider the probability of such an incident.

In fact, major fire related accidents with numerous casualties in many European tunnels have

triggered the public awareness regarding the consequences of tunnel fires [16]. Some of these

events that have been widely researched upon and are repeatedly mentioned in literature related

25 to tunnel safety are the Mont Blanc Tunnel Fire in 1999 [17], the Tauern tunnel fire in 1999 [18]

and the St. Gotthard fire in 2001[6] [4].

The impact of these incidents was, by all means, decisive for the global community and,

consequently, led to the revision of the existing regulations and standards and the introduction of

new sets of requirements with the objective to minimize the risk of casualties in tunnels. For

instance, the European Parliament and the Council of the European Union composed in 2004 the

Directive 2004/54/EC setting the minimum requirements for tunnels in the Trans-European Road

Network [19].

Another pivotal reason for revising the existing regulations and that is, also, stated in the Directive

2004/54/EC was that the design of the European tunnels that were put into operation some years

ago corresponds to obsolete technical possibilities and different transport conditions from those at

present. Consequently, taking into account the rapid developments related to technical solutions,

national legislators, currently, permit the use of a more performance-based design approach as

an alternative to the more traditional prescriptive approach. That is the case, for instance, in the

United States regulation (NFPA502, 2011) which suggests the use of a risk based approach. All

in all, the main concept of the performance-based approach is that a certain level of safety should

be provided, with the possibility to use every technological solution available [20].

Furthermore, a point of focus will be the maximum allowable distance between the egress exits.

According to the Directive 2004/54/EC, in the tunnels where emergency exits are needed, the

distance between two consecutive emergency exits should not exceed 500 meters. On the other

hand, according to NFPA 502, the spacing between emergency doors should not exceed 300

meters. It is stated, though, that this spacing should be justified by calculations and an

engineering analysis. More specifically, for a uni-directional tunnel equipped with longitudinal

ventilation, these distances will highly depend on the fire detection system and its timely

response. This way, the fire will be detected as soon as possible and the emergency ventilation

can be activated in order to control the smoke [4]. As a result, it can be concluded that the

spacing between the emergency exits highly depends on the properties of the fire detection

system.

What is, therefore, important, is for the European countries to comply to those revised regulations

in order for a uniform and high level of tunnel safety to be provided to all the European citizens

[19]. As a consequence, the need to refurbish the older tunnels already in operation rises.

Nevertheless, disrupting the operation of a road tunnel has significant consequences, both

economic and social. Under those circumstances, a careful consideration of the risks involved in

each individual tunnel together with the efficiency of the interventions available should be taken

into account in order to plan a strategy that is both economically feasible and results in a safe

tunnel operation. Accordingly, the safety planning during the design, construction and operating

phase of a road tunnel should aim to achieve a balance between optimal safety levels and

reasonable construction and operational costs [21].

2.3 STATE-OF-THE-ART

At this point, it is considered necessary to further address the issue of fire safety in road tunnels

in order to specifically define the parameters that are of importance throughout a fire safety

design. Generally speaking, in order to provide the required levels of safety in a road tunnel,

particular structural, technical and organizational measures should be applied. The primary

26 objective of these measures would be to prevent critical incidents that might compromise human

lives, the tunnel installations and the environment. The secondary objective would be to reduce

the impact of critical events (accidents and fires). [21] In fact, to provide all the prerequisites

(according to the Directive 2004/54/EC [19], NFPA502 [22] and PIARC [21]) related to:

Safe self-evacuation of people involved in the critical incident

Safe and efficient rescue operations by the emergency services

Minimal effects on the environment and

Minimal material damage

However, there are various challenges included in the fire safety design of a road tunnel. That is

because, in the confined environment of a tunnel, the development of excessive heat and smoke

together with the difficulties of fire suppression can consist a threat to human life in various

different ways such as the inhaling of products of combustion (carbon monoxide, carbon dioxide)

and the exposure to extremely high temperatures (up to 1350oC) and heat fluxes (even more than

300kW/m2). Moreover, in case of a fire incident inside a tunnel, the people involved should leave

their vehicles and self-evacuate from the nearest exit and access the specially designated egress

routes. Nevertheless, self-evacuation is often obstructed by crashed vehicles or blocked exits

because of traffic jams, by poor visibility conditions because of the presence of smoke, by a

collapse of the structure or an explosion or by a power failure. Added to all the above, is the

unpredictable human factor. People react differently under stress and a careful consideration of

their reactions should be made based on disaster psychology [8]. As a result, in order to ensure a

safe environment for self-evacuation, there is the need to maintain tolerable temperatures,

adequate visibility and air quality levels inside the tunnel [9].

With the intention to provide the prerequisites related to fire safety as described previously and to

best respond to the challenges involved in the safety design of a road tunnel, a number of

different safety systems are developed and installed [6]. More specifically, these systems relate

to:

The initiation phase and development phase of fire spread

The development and spread of smoke and other hazardous gases

The detection and activation of alarm

The intervention of the emergency services and the evacuation of the people and, finally

The repair and retrofit phase

All the different phases mentioned above require different fire protection measures which respond

accordingly to the type of emergency. The complexity of designing an efficient interaction and

coordination of these systems lies to the fact that the time scales related to the response time of

active protection systems differs from the structural response time for maintaining the structural

integrity and the reaction time of the occupants during self-evacuation [23].

In order to deal with all the aforementioned topics involved in tunnel fire safety, a lot of research

has been already performed in various related fields. It was necessary, thus, to gather this

knowledge and present it in a concentrated and comparative way so that the tunnel designer

could have an overview of the results and conclusions. A very important step towards this

direction was the NCHRP Synthesis 415 – Design Fires in Road Tunnels [4] which offers in depth

information about the current state of the practice of fire safety in road tunnels. The focus is on

fire dynamics in tunnels and the coordination of the fire systems and operations. The information

included in the NCHRP Synthesis 415 is the result of a literature review, a survey conducted by

U.S. and international transportation agencies and tunnel owners, various reports of past

27 experience with tunnel fire safety systems (ventilation systems, fire protection and fire detection

systems).

In addition, the NCHRP Synthesis 415 offers extensive information of conducted tunnel safety

projects, fire tests in tunnels and comparisons of national and international requirements and

guidelines for tunnel fire safety. All the aforementioned together with a detailed description of past

fire incidents in tunnels are presented in extensive appendices.

Based on the reviewed literature and the outcomes of the surveys, the following general

observations can be derived:

Tunnels are risky and challenging environments that cannot be absolutely safe. The goal

is to prevent tunnel fires from happening

The probability of a fire incident in a tunnel according to surveys is once or twice a year.

Most of these incidents are small scale ones and include small fires.

The tunnel fires that involve HGV or fuel tankers, although more rare, are considered to

have severe consequences on the tunnel structure, the occupants and the economy in

general

Catastrophic tunnel fires are less common than open road fires. It is estimated that that

less than 200 individuals were killed in road tunnels in total in a worldwide scale.

Moreover, it is recorded that less than 20 tunnels worldwide have been structurally

damaged as a result of a fire incident

In addition to the literature and surveys that were reviewed in the context of the NCHRP

Synthesis 415, many past fire incidents in tunnels involving HGV were also analyzed. The

conclusions drawn by this analysis were the following:

Fires in tunnels have a much faster development than it is expected. A number of

recorded tunnel fires resulted in fire curves with very fast development in the first 5 to 10

minutes. At the same time, there is a steep temperature gradient and the heat and smoke

emission is significant.

Temperatures that exceed 1000oC can be reached

The volume of the smoke produced is higher than expected in the first stages of the fire

The NCHRP Synthesis 415 was published in 2011 and provides an overview of the research and

practices until then. As relevant research and practices progress further, new knowledge and

material will be added in the current edition.

Overall, the NCHRP Synthesis 415 reviews the existing knowledge on fire safety in tunnels and

identifies the gaps in research. In addition, gaps in the national and international regulations and

standards were determined. A summary of those gaps as mentioned in the NCHRP Synthesis

415 is the following:

The standards and regulations need to better consider the various systems of a tunnel

and their interactivity

The current technical innovations should be taken under consideration in order to reach

more ambitious safety objectives

Human behavior of the users and operators of the tunnel should be more accurately

assessed and predicted

28 Furthermore, knowledge gaps were identified in many categories involved in tunnel fire safety. In

order to define the specific topic of this research study, the gaps that were more important were

determined. These gaps, mainly, considered two main categories that are described as follows:

Training and education:

First, it is important to find ways to train the tunnel operators and the people that first

respond to a fire incident.

Secondly, it is necessary to gain more insight into the human behavior of the people

involved in fire incidents in tunnels. This way tunnel emergency management can be

better organized and, as a result, more effective in saving the lives of the tunnel users.

Physics, numerical modeling, testing:

First, the verification of numerical modeling through full-scale fire tests is very important

with main interest the measuring of smoke and pollutants production rate.

In addition, further research is needed in order to evaluate the current developments of

numerical fire and evacuation simulations.

All in all, the research should aim to determine the design parameters for the numerical

simulations (fire and evacuation) and develop standard methods for the assessment and

validation of the results of the numerical simulations.

Specifications, regulations and technology:

First, fire detection based on CCTV should be further developed and adjusted for tunnel

applications. Moreover, research on the development of the ventilation system and its

integration with the other tunnel systems should be carried out.

Secondly, specifications are needed for the tunnel fire safety devices. More specifically,

robust, reliable and maintainable devices and systems should be commercially available.

These devices should be specifically designed for the tunnel environment taking into

consideration parameters like toxic substances, chemicals, dirt and other pollutants that

are present in a tunnel.

After the preliminary literature review described above, the focus had to be narrowed down in

order to formulate a research topic suitable for a MSc Thesis. As a result, the topics related to

tunnel fire safety that were more relevant to my academic background and, at the same time,

were of both academic and commercial interest were determined. Inspired by the current

practices in tunnel fire safety and the gaps in literature, the points of interest determined were the

following:

Taking into account the latest developments in technology in order to achieve a timely

response in case of a fire emergency

Defining specifications for the tunnel fire safety devices and coordination of the systems

Gaining more understanding regarding the human behavior during an evacuation in case

of a fire incident

Determining how the smoke and its toxic composition affects the health of the evacuees

and the total evacuation process

Exploring ways for validation of the results of CFD numerical simulations and determining

the design parameters

All in all, it was considered necessary to further research into the state of development of the

above topics and into the knowledge already included in the existing literature. Since the NCHRP

Synthesis 415 presents an overview of the research until 2011, most of the additional sources

studied are dated after 2011.

29

2.4 FOCUSED LITERATURE RESEARCH

In the following section, a more elaborate literature review regarding the selected fields of interest

is going to be presented. More specifically, these fields were fire detection and evacuation in road

tunnels.

fire detection

From the literature review described above, it was derived that the fire detection system plays a

very decisive role to the total fire safety design of a tunnel. Taking into account, thus, the

importance of the fire detection system, extensive research on the latest developments on the fire

protection technologies was carried out not only in literature but also in the market.

Most current research has mainly focused on the emergency ventilation operation and the smoke

control in tunnels [24] [25] [26, 27]. Limited recent studies [10, 28] and test programs were

conducted with the aim to investigate the effect of airflow on the performance of fire detection

systems in tunnels [29]. This was recognized, hence, as an issue to be investigated in Phase I of

the International Road Tunnel Fire Detection Research Project [30]. Following, Phase II of the

International Road Tunnel Fire Detection Research Project included full-scale ventilated fire tests

in a laboratory tunnel. In these tests, nine fire detection systems standing for five different types

of fire detection technologies were evaluated [29].

In the International Road Tunnel Fire Detection Research Project various recommendations for

future research on fire detection in tunnels were derived. These recommendations are the

following:

A number of fire scenarios were used to investigate the performance of detection

technologies. Still, further research is required to define standard tests that will be used to

evaluate the performance of fire detection systems in tunnels.

Research on the impact of transverse and semi-transverse ventilation on the

performance of fire detection technologies should be carried out.

Further research should be performed on the fire conditions in a tunnel that is

longitudinally ventilated in order to determine the impact of the airflow on smoke spread,

temperature distribution and flame spread in the tunnel. In relation to this research, the

performance of systems of Video Image Detection (VID) should be studied in the case

that the detectors are installed upstream the fire.

The Linear Heat Detection (LHD) systems should be tested in order to determine if they

are subject to false alarms in tunnels.

Further work is required in order to link CFD data to Video Smoke Detection and flame

detection regarding smoke characteristics (density, obscuration and visibility) and flame

characteristics.

Another set of fire detection experiments were performed in the Runehamar test tunnel in Norway

on March 2007 [27] in order to examine what kind of principle of fire detection is more suitable for

detecting a fire in an early stage in a tunnel.

A study that was published in 2004 worked on developing advanced algorithms of fire detection

for residential applications. The goal of the algorithms would be to result in fast fire detection

while, at the same time, eliminating the false alarms. Many parameters were measured and

studied and, finally, three algorithms have proven to achieve the above goal [31]:

30

Temperature rise and CO

CO and ionization detector

Temperature rise, CO and ionization detector

An interesting study would be to investigate whether these algorithms would prove to be effective

in tunnel applications.

What is more, despite the apparent importance of the detection systems, standards in a

worldwide scale regarding the engineering and performance requirements were not defined [10].

Recent studies that were carried out by the Fire Protection Research Foundation, indicated that

there is limited information regarding the performance of the latest fire detection technologies and

guidelines for their use [9]. As a matter of fact, a number of test programs were carried out in

Europe and Japan that were, mainly, focused on the performance of linear heat detectors and

optical frame detectors [32], [33], [34], [35], [36]. The detectors were evaluated with pool fires on

a constant heat release rate of 3MW. Tests were not performed for other fire scenarios or types

of heat detectors.

In addition, regulations hardly contain any values for the maximum acceptable detection time and

the level of accuracy for the fire location. In certain national guidelines, some values regarding

time for fire detection are mentioned. In general, the detection time should be between 1 to 2.5

minutes depending on the fire development and the existing ventilation conditions. The

requirements regarding fire detection time and alarm systems that are provided in national

guidelines are summarized below [4]:

An alarm should be triggered at maximum 60 sec after the ignition of a fire or when the

fire energy load reaches at maximum 5MW

The fire detection system should be capable of detecting small fires (with HRR of 1.5 – 5

MW) or detecting larger fires at an early stage

Fire detection should be possible even when the airflow speed in the tunnel is up to 6 m/s

A fire location should be determined with an accuracy of 20 to 50 meters

Resulting from all the above, it is concluded that many difficulties lie in choosing the most

appropriate detection system for a tunnel since there is limited technical information regarding

their performance, availability and reliability [37].

tunnel evacuation and human factor

The detection time of a fire is directly linked with the total evacuation time and, consequently, with

the chances of a successful evacuation. For that reason, during the literature review, it is

considered equally important to investigate on the parameters that affect the evacuation process.

More specifically, the factors to be considered during an evacuation can be divided into two

categories: physical characteristics and human behavior. As a result, the recent relevant literature

was studied regarding different aspects of human behavior [16, 38-40] like pre-evacuation time,

interactions between occupants and between occupants and fire [32], herding behavior and,

finally, exit choice. All the relevant information for the above factors can be retrieved from data

regarding actual accidents [15], experiments or drills [41].

First, it is important to identify the significance of the human factor as far as the design of the

escape provisions is concerned. As has been noted, tunnel environments are characterized by

the lack of natural light and reduced visibility. In most cases, they are experienced by the users

31 as uninviting, confining and, thus, stressful environments [16]. Apart from these effects on the

user’s psychology, tunnels are, also, unfamiliar to users and are not immediately accessible to

the emergency staff in order to provide help [20]. Under those circumstances, it is important to

realize that the evacuation of a tunnel is, certainly, a matter of self-evacuation. As a result, there

is a significant number of studies concerning the improvement of the means of egress in a road

tunnel as well as the enhancement of the knowledge about the human behavior in evacuation

situations [20].

An important study by John Leach [42] determined that during an emergency people are usually

divided in three main categories according to their reaction and actions. The first category

includes the leaders that are the most active group and they take the decisions and act. The

second category includes the followers that are more passive and they observe the leaders and

follow their actions. The last category includes the blockers that are, usually, the people that

create problems either because of their reactions under stress or because of their false decisions

like, for example, doing a U-turn to leave the tunnel. The herding behavior plays a significant role

in the total evacuation process. It can be said that the moment the first person leaves their car to

move towards an emergency exit is the moment the evacuation starts because others will follow.

[43]. Through an evacuation experiment [44], the impact of social influence on the evacuation

process and, more specifically, on the decision to leave the vehicle and on the choice of exit.

Despite the existing studies on human behavior in emergencies, there is still limited information

because there are no experiments that can be considered reliable enough for, mainly, two

reasons [43]:

There is an ethical issue involved in the experiments. People should not be exposed to

dangerous conditions like toxic smoke and cannot be subject to extremely stressful

situations.

The experiments cannot produce reliable results because they cannot reproduce the

exact situation. In most cases, the participants are aware that they are participating in an

experiment.

As a result, a lot of observations and conclusions regarding the human behavior depend on

former fire incidents in tunnels [6]. Some of the main observations during past incidents in tunnels

are the following:

People often tend to underestimate the danger and experience a denial phase

People are reluctant to abandon their property and they stay in their vehicles

The fast development of fire and smoke is usually underestimated

The self-evacuation process does not start immediately

Egress routes are often blocked by vehicles or other obstructions

Refuge areas might not be designed for the specific fire size and not have the necessary

fire endurance

People can be trapped in the designated refuge areas

In order to, finally, evaluate the impact of technological systems and egress provisions on the

reaction time of the users, it is important that the general evacuation process is, hence,

introduced. To begin with, the evacuation process includes all the steps starting from the ignition

of the fire until the safe evacuation of all the occupants in a safe area. The total evacuation time

can be divided in the following three phases [45]:

32

Awareness phase: The time required to become aware of the fire ignition [46]. It mainly

depends on the speed of the threat detection and identification, the determination of its

location and its magnitude and on the time needed to transmit the message to the

occupants in the most appropriate way. It is worth to note that the aspects mentioned are

all strongly dependent on the sensor systems installed in the tunnel [13].

Response phase: The time required to take the decision to act and start the self-

evacuation. According to experiments conducted in a Benelux tunnel, people, usually,

remain passive for the first 5 to 6 minutes before starting to evacuate [38]. It can be said,

thus, that the response time mainly depends on the particular user and the experiences

of each individual [13].

Movement phase: The phase during which the action is taken and the evacuation is

taking place [14]. The movement time, mainly, depends on the physical characteristics

and health condition of each individual. In addition, the layout of the tunnel egress exits

and the distance to the closest safe exit are determinant factors for the movement time.

In fact, the above factors are independent of the sensor type or any other equipment in

the tunnel but, instead, they depend on the design of the tunnel [13].

As can be derived from all the above, the total time needed for evacuation can be calculated as

the sum of the times needed for the three phases. A more accurate calculation of the total

evacuation time can be achieved by knowing the duration of the first two phases [14]. A

distinction should, now, be made between the available safe egress time (ASET) which depends

on the rate of fire development and the smoke spread and the required safe egress time (RSET)

which is, basically, a function of the occupants [16]. Regardless the importance of the evacuation

time in the fire safety design of a tunnel, limited research has been conducted related to these

phases. Based on previous experience and a number of evacuation exercises, it was noted that,

in general, 5 to 15 minutes are needed for an individual to decide to act. This implies that, after

that decision, there is only little time available for the self-evacuation [14].

Fig.2.1 Tunnel hazard development[16]

In case of a fire incident in a tunnel, the people involved cannot take the decision to evacuate due

to a lack of information. During a fire incident in a tunnel people should be provided with clear and

accurate information on the type of the emergency, the location of the fire and the actions they

need to take [47]. The voice alarm message, the signage and the lights can significantly affect the

level of information provided to the tunnel users. Although a spoken message could sometimes

be difficult to be perceived, it can be considered effective since it alerts the tunnel users and

urges them to look for any additional information [44]. Experiments and research have been

conducted in order to determine the influence of signs and sound on the behavior of the

33 evacuees [48], [49], [50]. Research by Margrethe Kobes [51], [52] has suggested that exits signs

that are placed lower are perceived more easily than those placed higher. This is because, on the

one hand, the smoke layer usually covers the signs that are placed higher and, on the other

hand, the evacuees look straight and stay low because of the smoke. It was also, suggested that

dynamic signage (for example, green flashing lights at the exits) can be more effective in being

noticed by the evacuees [53].

Finally, an important aspect to be considered that affects the evacuation time is the influence of

smoke on the walking speed. Jin, 2002 [54] has suggested that the walking behavior in smoke-

filled environments differentiates than that in smoke-free environments. Usually, the low visibility

level because of the smoke yield has a negative impact on the ability of the users to evacuate.

According to experiments, the walking speed of people in a smoke-filled environment is lower

than that in normal conditions. Finally, it was observed that under low visibility conditions,

evacuees tend to walk close to the walls for guidance [55], [56], [57]. Depending on the age and

health state of the evacuees, the walking speeds can vary between 1 to 1.6 m/s [4]. Feedback

from various guidelines and information from real evacuations in tunnels and tests suggest the

following values for the walking speeds [58]:

After a series of experiments in a smoke-filled corridor, it was concluded that in non-irritant smoke

walking speed can be reduced to 0.5 m/s whereas in irritant smoke the walking speed can be

reduced to 0.4 m/s. In addition, according to PIARC, the walking speed in a smoke-filled

environment can be considered to be between 0.5 and 1.5 m/s [4].

It can be concluded that faster fire detection and proper instructions to the occupants of the

tunnel could, significantly, reduce the first two phases. In that case, more time is available in the

third phase for the actual escape [14].

synthesis

Resulting from the literature review described above, many interesting aspects of road tunnel fire

safety were determined that are, still, not extensively researched upon. That consists a motivation

for this research study. That means that the present work will try to fill in certain literature gaps

that were mentioned in the literature review so that it can add to the body of knowledge.

First of all, there is a lack of information regarding the reasoning behind the current regulations.

What was considered interesting to investigate, thus, are the criteria for setting the regulations.

This way, the engineer and tunnel designer could have an insight on the scientific background of

the regulations and make a more elaborate estimation of each individual tunnel. Then, there will

be room for a more performance-based design that would take into account the individual

characteristics of each tunnel. In addition, as mentioned previously, the current regulations do not

extensively take into account the technological developments in the field of fire detection. As a

result, an investigation of the impact of those technologies within the context of this research

study would highly benefit a great number of stakeholders in the field of tunnel safety.

Moreover, as mentioned before, there are no specific requirements set regarding the

Table 2.1 Summary of recommended walking speed

34 performance of fire detection technologies. In addition, the fire detection technologies are tested

under specific scenarios while there is little experience and information regarding the response of

the fire detectors under different fire scenarios. For that reason, the current study will focus on

determining the performance requirements of the detectors and, at the same time, identifying

which technology best responds to certain characteristic fire scenarios.

As far as the evacuation of pedestrians is concerned, a lot of literature is available. Nevertheless,

there is little information on the human response in case of an emergency and specifically on the

aspects that affect the decision to start the evacuation. That is because of the difficulties involved

in conducting experiments and the reliability of the results of the experiments. On the one hand, it

is not allowed to expose people in toxic conditions as those in the case of fire because an ethical

issue arises. As a result, the experiments are conducted under not toxic conditions and the

results cannot be considered fully reliable. On the other hand, in most of the experiments the

participants are aware that the fire emergency is not real, a fact that affects the reliability of the

results. Under these circumstances, this research study will aim to combine the findings of the

latest existing research and provide qualitative recommendations related to the tunnel fire safety

design.

All in all, after the literature review was conducted, the basic points of interest were identified. It is

considered important to mention that throughout the research study many sources - scientific

articles, books, conference proceeding and more - were consulted and will be mentioned in the

report. Nevertheless, the main guidelines and basic knowledge regarding tunnel safety were

provided by the following sources:

The Handbook of tunnel fire safety by Alan Beard and Richard Carvel [6]: this book

covers a very wide range of knowledge related to tunnel fire safety.

NFPA 502 regulation [22]

NCHRP 415 Synthesis [4]

SFPE Handbook of Fire Protection Engineering [59]

At this point, it is considered necessary to mention that the topic of fire safety in tunnels is

complex and involves many fields of knowledge. As a result, before starting this thesis, an ample

research had to be realized including fields like the following:

fire dynamics in tunnels

fire detection technologies

human behavior in case of emergencies

evacuation strategies

ventilation in tunnels

Computational Fluid Dynamics (CFD)

software manuals

full-scale fire tests in tunnels

validation of simulation models and more

Therefore, mentioning all the relevant research in this chapter would result in a long literature

review with a load of information that would not be directly linked to the work of this thesis. For

that reason, in each of the following chapters, there is a section where the respective research is

cited and justifies the choices made throughout the research study.

35

CHAPTER 3: Tunnel operation

This chapter has a dual purpose: on the one hand, to offer an understanding of the normal operation of road tunnels and of all the systems involved; on the other hand, it will provide the information necessary to justify certain choices made throughout the research. These choices concern the type of tunnel geometry simulated, the chosen ventilation type, the fire scenarios and the emergency systems.

3.1 NORMAL OPERATION

The aim of this research study could not be accomplished without a thorough and comprehensive

understanding of the normal operation of a road tunnel. For that reason, the overall and initial aim

of this research study would be to investigate into the elements that comprise a complete and

safe tunnel concept in order to, subsequently, determine the operation of the tunnel systems in

case of an emergency.

3.1.1 CASE STUDY: HUBERTUSTUNNEL

In order to make the research study more applied and to use real data as an input to the

simulations, a road tunnel was selected as a case study. The selected tunnel is the

Hubertustunnel which is the last part of the Northern Ring Road in the Hague region (N14). This

ring road is the link between the A4 (at Leidschendam) and Scheveningen.

description of the Hubertus tunnel

The traffic in The Hague increases significantly. At the same time, because of the location of the

city close to the sea, it was not possible to guide the traffic around the city but, instead, all the

traffic passes through the city. This makes residential area in The Hague to be seriously

burdened by the traffic intensity causing traffic jams and delays. With the construction of the

Hubertustunnel, the traffic was redirected and many of these problems were solved. The tunnel

construction began in 2004 and the tunnel was open to public in 2008.

A characteristic of the Hubertus tunnel is that it was the first tunnel in the Netherlands to be built

underneath buildings. As a result, in order to cause minimal damage to the environment and the

buildings, it was constructed with a special drilling technique using a boring machine. The 63

meters long drill was drilling the tunnel and at the same time it was placing the tunnel walls.

The Hubertus tunnel consists of two unidirectional tubes and there are two lanes for each

direction. The total length of the tunnel is 1,600 meters.

Regarding the safety provisions, the Hubertus tunnel complies with the tunnel safety regulations

and standards. More specifically, it monitored in a 24-hour basis by a control center in

Scheveningen and the tunnel operator can intervene if necessary at any point. Special facilities

are integrated in the tunnel to ensure the safety of the users such as barriers in the entrances and

exits, a Speed Discrimination System to identify and locate any stationary vehicles and more.

Moreover, the fire safety systems installed in the Hubertus tunnel include a fire resistant coating

on the tunnel walls, fans mounted on the ceiling for smoke discharge, pumps of water in the

cellars and a fire extinguishing system [60].

36

More information for the Hubertus tunnel, together with reports of the construction method can be

found in the COB electronic platform [61].

traffic data for the Hubertus tunnel

At this point, it was considered necessary to determine the traffic intensity for the Hubertus tunnel

in order to provide this data as input for the evacuation simulations. The information necessary to

set up the evacuation simulations mostly concerns the number of vehicles crossing the tunnel on

a daily basis in order to estimate the number of people present in the tunnel the time of the fire

emergency in the scenarios investigated.

As a result, traffic data had to be retrieved for the Hubertus tunnel. As can be seen in the

following map, connected to the Hubertus tunnel is the road N440 that is part of the ring road of

the Hague and, also, bears the number S200. As is, also, illustrated in the map, the N440 starts

after the Hubertus tunnel. The traffic intensities, thus, for the N440 road correspond to the traffic

intensities for the Hubertus tunnel.

In the electronic platform of the Province of South Holland, the following traffic data were

recorded [62]:

So, in the year 2010, the total number of vehicles that crossed the N440 road was measured to

be 32,900 vehicles. Moreover, in the Hubertus tunnel the speed limit was increased in 2013 to 70

km/h [62].

Fig.3.1 Map of Hubertus tunnel, The Hague (source: www.maps.google.com)

Table 3.1 Traffic data for Hubertus tunnel vehicles/year

37 Another parameter that was considered

when choosing the tunnel to be used as

a case study was the wind conditions.

The wind is a crucial meteorological

factor in the Netherlands and can have

an impact on the climate in a tunnel. As a

result, an area in the country that is most

affected by the wind had to be chosen.

According to the Meteorological Institute

of the Ministry of Infrastructure and

Environment, the Netherlands is

separated in wind zones as shown on the

map below. The area with the most

severe wind conditions is the one on the

west coast of the Netherlands

represented by the dark pink color on the

map (zone 4). The average wind speed

at this area is around 6 m/s. The

Hubertus tunnel is located within this

wind area and is, thus, most affected by

the local wind conditions.

3.1.2 TUNNEL DESIGN CONSIDERATIONS

The specific characteristics of tunnels impose a set of design considerations that should be met.

Tunnels are characterized by the fact that they are enclosed structures and the road has many

restrictions in both the vertical and lateral direction. In addition, there are no crossroads or

pedestrians present. Furthermore, they require installations for artificial lighting and mechanical

ventilation.

These characteristics have an effect on the driver’s perception of the driving environment. The

tunnel user may be cautious and apprehensive on entering or exiting a tunnel and usually has the

tendency to drive far from the tunnel walls and has a wrong perception of the inclinations of the

tunnel. These result in decreasing the effective width of the lane and in underestimating the

braking distances. Despite all the above, accidents in tunnels occur less often than those in the

open road. That is because of the safety design and features of the tunnels, the absence of

obstacles at the sides of the road and, finally, the fact that the drivers are more cautious and alert

while entering the different environment of the tunnel.

The specific considerations involved in tunnel design are, mainly, related to the construction, the

operation, the safety systems and the communications. When designing a tunnel, an engineer

should take into consideration the following parameters [63]:

Design and construction: design speed, cross-section, grading, alignment, drainage,

structural requirements

Operations: emergency egress exits and plan, maintenance requirements, ventilation,

lighting, electrical requirements

Fig.3.2 Wind map of the Netherlands (source:www.knmi.nl)

38

Safety systems: fire safety design, emergency ventilation, emergency lighting

Communications: radio, voice alarm, notification of the fire brigade

tunnel structure

It is true that tunnels are important but very expensive parts of road infrastructure. Although they

can be up to ten times more costly to construct than a bridge at the same location, they are often

the only solution to overcome certain circumstances. Such circumstances would be to create new

networks and routes under dense urban areas with high costs of the land, to cross mountains

avoiding at the same time steep inclinations and longer routes, to cross rivers of large areas

covered with water, to protect areas of historical value or areas that are considered

environmentally sensitive.

It is very important to determine the type of tunnel that should be used as far as its construction

method and geometry is concerned. This depends on a range of operational and physical

circumstances. Some of the most common types of tunnel are:

cut-and-cover at shallow depth

cast-in-situ in a waterway

immersed tube for underwater crossings

bored tunnels and

tunnels excavated through rock

The two types of geometry that were chosen to be simulated for this research study were the cut-

and-cover and the bored tunnel. This was because these two types are two of the most

commonly used. The cut-and-cover tunnel is more appropriate for shallow depths and it is

constructed inside a trench that is excavated from the surface. This type of construction was

developed for dense and congested urban areas in order to avoid open excavation techniques

that could significantly disrupt traffic. In the Netherlands, this type of tunnel is common in urban

areas in cases where an under-water crossing has to be created [64].

The construction of a bored tunnel is realized by digging a tube-alike passage through the

ground. This method is most commonly applied to tunnels crossing mountains. Nevertheless, it is

also used for tunnels under large bodies of water. In order to avoid the conventional tunneling

techniques (which involve the use of explosives and manual labor), the bored tunnel construction

is realized with the use of Tunnel Boring Machines (TBM’s) [65].

choice of cross-section

Before making the models of the two alternative tunnel geometries, the cross-sections had to be

defined. In order to choose the cross-sections, three factors were considered:

the cross-sections that are more common for the two types of tunnel structure mentioned

above

the cross-section of the tunnel that was chosen as case study

the cross-sections of the tunnels where tests were performed and which are going to be

used for the validation of the results of the simulations

Taking into account these factors, the number of geometrical models that would need to be

created for the research study was reduced.

39 In general, the purpose of the tunnel defines the shape and size of the cross-section. In addition,

the geotechnical and structural characteristics of the ground influence the size of the cross-

section together with the alignment. Lastly, the design of the cross-section depends also on the

construction process.

In the context of this thesis, a road tunnel is studied and, thus, the cross-sections that correspond

to road tunnels were studied in order to choose the most representative ones. Generally, in road

tunnels, the traffic conditions should be in correspondence to the ones in the open road.

Nevertheless, more strict requirements should be met inside a road tunnel regarding its

construction, operation and, finally, maintenance in order to provide the necessary levels of

safety. A way to increase the safety levels in a road tunnel is to set a lower speed limit than that

in the open road. As a result, the maximum speed limit in a road tunnel is normally 80 km/h.

Road tunnels could be bi-directional (two-way traffic) or uni-directional (one-way traffic). Bi-

directional tunnels normally are constructed as a single tube and one lane for each direction. On

the other hand, uni-directional tunnels consist of two tubes and each tube can have one or more

traffic lanes of the same traffic direction. The cross-section of road tunnels depends on the traffic

intensity of the road, the operational equipment needed and the structure of the tunnel. More

detailed information on the design of road tunnel cross-sections can be found in the German

guidelines for the equipment and operation of road tunnels [66]. More specifically, the RABT

guidelines provide requirements for the following:

the standard cross-section

the structure or the vehicle gauge that needs to be maintained

the gradients (transverse and longitudinal)

the provision of breakdown bays and emergency exits

dimensions of cross-section

The cross-section of road tunnels can be selected according to Fig.3.3. Most commonly, for bi-

directional tunnels, the typical cross-section that is used is type 10.5T with a width of 7.50 meters.

In case of uni-directional tunnels with multiple lanes, a reduced standard cross-section without

hard shoulders is applied, type 26T (two lanes) or 33T (three lanes). The hard shoulder can be

justified by heavy traffic and steep gradients. The hard shoulder for the cross-section types 26T

and 33T should be 2.00 meters. The headroom that has to provided for road traffic should be

4.50 meters.

In either side of the carriageway, emergency pavements of 1 meter width have to be provided.

The clear headroom above the pavements has to be 2.25 meters. The emergency pavements are

separated from the carriageway with 7 cm high curbs.

The longitudinal gradient of a tunnel should not exceed 5% in order to ensure road safety and at

the same time to limit the chimney effect that can, severely, affect the smoke extraction process

in case of a fire.

The cross-slope of the tunnel carriageway should be a minimum of 2.5% in order to ensure the

drainage of the surface water [67].

40

Fig.3.3 Standard cross-sections for road tunnels

41

cross-section of case study tunnel

The road tunnel chosen as a case study is the Hubertus tunnel in The Hague. It is a two-tube

tunnel. Each tube consists of two lanes with one-directional traffic. The tunnel was constructed

with a boring machine and the cross-section is the typical one for a bored tunnel with an outside

diameter of around 10.30 meters.

The total length of the tunnel is 1,600 meters and the cross-section is as seen on the pictures on

the right.

cross-section of the test tunnel

Another tunnel that will concern this research study is the A2 Tunnel (Leidsche Rijntunnel) in

Utrecht. This is the tunnel in which tests were performed for a specific detection technology. As a

result, in order to validate the simulation models used for this thesis, the geometrical model of this

tunnel had to be created.

The A2 tunnel was constructed with the cut-and-cover method and has a rectangular cross-

section. It consists of four tubes in total: two main ones with three lanes each and two secondary

ones with two lanes each. Each tube has one-directional traffic. The total length of the Leidsche

Rijntunnel is 1,650 meters. The tunnel cross-section can be seen in the following pictures.

Fig.3.4-3.5 Cross section of Hubertus tunnel in The Hague (source1: www.siemens.com, source

2: www.denhaagfm.nl)

Fig.3.6 A2 tunnel in Utrecht (source1: www.schlijper.nl, source

2: www.artajasa.asia)

42

3.1.3 VENTILATION SYSTEM

The contribution of the ventilation system in tunnel safety is fundamental during normal operating

conditions as well as in case of a fire incident. Under normal operating conditions, the role of the

ventilation system is that of diluting the pollutants that are emitted by the vehicles, thus,

maintaining the required levels of air quality inside the harsh environment of the tunnel. At the

same time, the ventilation system improves the visibility conditions inside the tunnel. The need to

introduce a ventilation system in road tunnels emerged, initially, due to the negative impact of the

toxic emissions of internal combustion engines that resulted in a low air quality [68].

The exhaust emissions of the vehicles contain highly toxic gases, particulates and smoke. In a

tunnel, these can be harmful both for the tunnel structure and equipment and, more importantly,

for the health of the tunnel users. In short tunnels it is common to dilute those pollutants by using

longitudinal ventilation assisted by jet fans that are mounted on the ceiling and discharge the air

at the portals. The fans are controlled automatically depending on the traffic intensities, the levels

of pollution and the fire detection system. As a result, the tunnel should be equipped with systems

that will provide the necessary measurements to control the fans accordingly. Such systems are

traffic sensors, carbon monoxide and smoke monitors [63].

Short tunnels can be naturally ventilated under normal operation conditions. Nevertheless, there

is frequently the necessity to install mechanical ventilation systems in tunnels. The mechanical

ventilation systems for tunnel applications are classified as transverse or longitudinal. The

longitudinal systems create a longitudinal airflow within the tunnel while the transverse systems

are based on continuous uniform distribution and/or collection of air along the tunnel roadway

[22].

3.1.3.1 NATURAL VENTILATION

The basic principles of the natural ventilation systems in tunnels are based on the meteorological

conditions around the tunnel and the piston effect that is created from the moving vehicles. More

specifically, temperature differences and differences of the static pressure between the tunnel

portals can result in airflow along the tunnel that is, often, adequate to remove the pollutants and

provide the required level of air quality in the tunnel. Moreover, the prevailing wind can have a

significant impact on the discharge of the pollutants through the portals. It is possible to enhance

the natural ventilation by adding vertical shafts at certain points in the tunnel. This way the

chimney effect is further reinforced resulting in an even stronger airflow[68].

It is true, though, that natural ventilation systems are not always capable of achieving the

acceptable conditions within a tunnel. This is because these systems rely on a number of random

environmental and meteorological parameters that cannot be accurately predicted during the

design of a ventilation system. This is, also, the case with the piston effect by the moving

vehicles. Some of the factors that affect the piston effect are the geometry of the tunnel, the traffic

direction, the moving speed and the spacing of the vehicles. These factors are, in most cases,

unpredictable and cannot be controlled [68].

As a result, it is established that natural ventilation systems should, only, be applied in short

tunnels. According to national guidelines, short tunnels are the ones with a length that is less than

350 – 700 meters (German regulations) or 400 meters (UK regulations).

43

3.1.3.2 MECHANICAL VENTILATION

longitudinal ventilation systems

The longitudinal ventilation systems import air into the tunnel and, then, remove it usually through

the portals to create a longitudinal airflow along the tunnel length. Longitudinal ventilation

systems could be further categorized in the ones that use central fans and the ones that use local

fans or jet fans. Those two types of longitudinal ventilation system configuration are presented in

the following sketches.

In this case, centrally located fans are

used and the air is injected into the tunnel

through either a supply airshaft or a high-

velocity nozzle (for example, Saccardo

nozzle).

In this case, there is a series of jet fans

mounted at the ceiling that create a

longitudinal flow of air along the tunnel.

The exhaust air is discharged through the

exit portal or from exhaust airshafts.

transverse ventilation systems

The transverse ventilation systems uniformly collect and/or distribute air along the tunnel length

and can be distinguished in fully transverse and semi transverse (supply or exhaust type).

A fully transverse ventilation system consists

of supply and exhaust airshafts that run along

the tunnel roadway. In this case, most of the

pollutants are discharged through one or more

stacks while a small amount of pollutants is

discharged through the portals.

Semi-transverse ventilation systems can be

either supply or exhaust. In the case, of supply

semi-transverse systems, the fresh air is

supplied through an air-duct and the exhaust

is discharged through the tunnel portals.

In the case of the exhaust semi-transverse

ventilation systems, the fresh air enters the

tunnel through the portals and the exhaust air is

discharged through exhaust stacks as shown in

the sketch on the left.

Fig.3.7 Longitudinal ventilation with central fan

Fig.3.8 Longitudinal ventilation with jet fans

Fig.3.9 Fully transverse ventilation system

Fig.3.10 Supply semi-transverse ventilation system

Fig.3.11 Supply semi-transverse ventilation system

44

3.1.4 AIR QUALITY

In the confined environment of a road tunnel, maintaining the required levels of air quality is a

challenging task that is mainly performed by the ventilation system. In open air conditions, the air

pollutants that are emitted by the vehicles can be dispersed immediately with the help of the wind

and are rapidly mixed with fresh air due to turbulence. However, the environment of a tunnel is

sheltered by the wind and, despite the fact the turbulence still exists, the fresh air supply to be

mixed with the pollutants is limited. As a result, the concentrations of pollutants such as carbon

monoxide (CO) inside long urban tunnels (more than 1 km) can be up to 1-100 times higher than

those in the ambient atmosphere [69].

The major emissions in road tunnels are the exhausts from the tailpipe and are, mainly, carbon

monoxide (CO) and nitrogen oxide (NO2). Recent advances in technology have resulted in a

reduction of the CO emitted by the typical vehicle. However, this is not the case for the oxides of

nitrogen emissions. In addition, emissions consist of vehicle products such as dust from the tires

or the brakes. Moreover, the movement of the vehicles and the frictional contact between the tires

and the road surface result in the re-circulation of dusts that already rest in the tunnel surfaces.

Exposure to such an environment can have either long-term or even short-term effects on human

health. According to research studies of the effects of exposure to in-tunnel environments,

respiratory effects are observed in asthmatics after exposure of 30 minutes (Svartengren et al.,

2000) and 2 hours (Larsson et al., 2010). More specifically, exposure to the polluted tunnel

environment could cause asthma aggravation and cardiovascular disease exacerbation. What is

more, high CO uptake can cause headaches, dizziness, confusion, weakness and disorientation.

All these effects on health can significantly affect the tunnel users and delay the evacuation

process.

The importance of the air quality inside a tunnel, has led to the establishment of in-tunnel air

quality criteria ensuring the health and safety of the tunnel users. The criteria for in-tunnel air

quality permit the exposure to higher concentrations of pollutants compared to ambient air. That

is because the time of exposure to in-tunnel environment is considerably shorter (no more than a

few minutes). These criteria have, also, a significant impact on the performance requirements of

the ventilation system. As was, already, mentioned, the air quality in tunnels is constantly

monitored and the ventilation system is adjusted in order to maintain the required air quality

levels.

The World Health Association (WHO) has set standards and guidelines (WHO 2000, 2006) for

the acceptable concentrations of various pollutants that are referred to worldwide. These

guidelines resulted from a synthesis of research on the impact of these pollutants on the human

health. Amongst the pollutants mentioned in the guidelines, those that are related to vehicle

emissions are, also, included: CO, NO2 and particulate matter. The exposure limits that are most

commonly adopted for the assessment of the in-tunnel air quality are those for CO. That is

because the only traffic related pollutant for which WHO has established guidelines for short

exposure duration (like the passage through a road tunnel) is the CO. More specifically, the CO

concentration levels averaged in 15 minutes should be less than 87 ppm at 25oC [69].

The Permanent International Association of Road Congresses (PIARC) for road tunnels [21] has

adopted the WHO Guideline and the PIARC recommendations have been consulted worldwide

with certain adaptations. PIARC, also, introduced certain guidelines regarding the visibility limits

in road tunnels. It is true that visibility can affect the tunnel users by causing stress and obstruct

safe driving. Besides these effects, low visibility, also, implies the presence of particles such as

45 soot that have a direct effect on human health. The PIARC recommended limits for CO

concentrations and visibility levels are shown in the following tables [21]:

Table 3.2 CO concentration by traffic situation for the years 1995 and 2010

Table 3.3 Visibility conditions by traffic situation

Table 3.4 Threshold concentrations for CO and NO2 according to various regulations

46 Although in the past, it was considered that CO can be an accurate indicator for a wide range of

pollutants in the tunnel environment, nowadays this should be reconsidered. More specifically,

due to advances in vehicle technology, a significant reduction of the CO and other vehicle

emissions have been observed [70]. In the case of particulate matter and nitric oxide emissions,

though, the technological advances have not resulted in an important reduction of emissions.

Moreover, in Europe specifically, the wide use of diesel-powered vehicles has led to an increase

of the primary emissions that result in the formation of NO2 [71]. As a consequence, the NO2 and

particulate matter per amount of CO in tunnels is relatively more than it used to be. For that

reason, the guidelines based on the CO concentration alone should be revised and NO2

exposure limits should be introduced.

As can be derived from all the above, the air quality in road tunnels should be considered as a

major design factor. The tunnel environment should be constantly monitored and the ventilation

system should be designed in order to maintain the pollutant concentrations below the

aforementioned limits.

Currently, multi-gas sensors are available at the market that can be used to monitor the in-tunnel

air quality levels. Such sensors can be configured to monitor at the same time oxygen levels,

volatile organic compounds (VOCs), gases and vapors and, also, a variety of toxic gases. An

example of a multi-gas sensor is the PHD6 by Honeywell and its specifications regarding the

alarm thresholds set are shown in the following picture:

The brochures of the PHD6 multi-gas sensor including all the specifications of this technology can

be found in the Appendix B.1.

Table 3.5 Alarm thresholds for various contaminants for multi-gas sensor PHD6 by Honeywell

47

3.1.5 TENABLE ENVIRONMENT

At this point, it is important to understand the conditions required to maintain a tenable

environment within a tunnel. This way a better understanding of the objectives of the fire

regulations can be achieved. What is defined as tenable environment is an environment where

humans can survive for a specific time period. Fire safety systems in tunnels aim to maintain a

tenable environment for so long as to achieve a safe evacuation of the tunnel users [4].

In case of a fire incident in a tunnel, human life is threatened by various ways: extremely high

temperatures, low oxygen concentration, low levels of visibility and various toxic gases that could

be lethal in high concentrations. Moreover, all the aforementioned phenomena can cause

damage to the construction, to the equipment installed and to the vehicles. As a result, certain

tenability limits have been set with which each tunnel should comply. Some of these tenability

limits by means of required environmental conditions are, shortly, discussed in the following

sections.

heat effects

A basic threat to human life in case of a fire is the extreme heat developed in the confined

environment of a tunnel. Heat threatens life in the following ways:

Hyperthermia

Body surface burns

Respiratory tract burns

The two criteria that have to be considered in relation to heat exposure are:

Threshold of the burning of the skin

Level of exposure at which hyperthermia can cause mental deterioration and threaten

survival

It is determined that respiratory tract burns appear after the body surface burns and for that

reason the tenability limit for skin burns is lower than that of respiratory tract burns.

In the event of a fire, radiation is created by temperature and is emitted both by the fire and by the

smoke layer. A tenability limit of 2.5 kW/m2 (for bare skin) is set for the exposure of skin to radiant

heat. Below this threshold, an exposure of 30 minutes can be tolerated without affecting the time

for escape. A radiant heat of 2.5 kW/m2 corresponds to a temperature of 200

oC which is,

normally, exceeded in the proximity of the fire. Firefighters that are equipped with special

protective clothing, can withstand a radiation value of 5 kW/m2 for a minimum of 7 minutes[4].

In addition, the unprotected human skin can tolerate up to approximately 120oC for convective

heat. Above this temperature and in only a few minutes, the body experiences intense pain and

burns appear. In addition, even for a lower temperature than 120oC, with longer time of exposure,

hyperthermia can be caused[4].

48 The following table illustrates the maximum exposure time without incapacitation for specific

temperatures [22]:

air CO content

The considerations for a tenable environment by means of air CO content, as mentioned in the

NFPA502 [22], are the following:

Maximum of 2000 ppm for a few seconds

Averaging 1150 ppm or less for the first 6 minutes of the exposure



Averaging 50 ppm or less for the remainder of the exposure

smoke obscuration levels

It is required that the levels of smoke obscuration should be below a certain level. Namely, a sign

with a luminance of 8.6cd/m2 should be visible at a distance of 30 meters and doors and walls

should be visible at a distance of 10 meters.

air velocities

The minimum air velocity within a tunnel should be 0.76 m/sec and the maximum air velocity

should be 11.0 m/sec. The maximum threshold was based on the ability of people to walk under

high air velocity.

geometric considerations

Certain geometric considerations should be taken into account so as to provide a safe evacuation

path. These considerations are the following:

An evacuation path of a clear height of 2.0 meters should be provided. It is considered

that the modeling methods that are currently in use can achieve a precision of 25%. For

that reason, when using a modeling method, the clear height of the evacuation path

should be 2.5 meters.

Table 3.6 Exposure time and Incapacitation limits

49

It is not considered practical to apply the tenability criteria close to the perimeter of the

fire. A zone of tenability should be determined which would be away of the perimeter of

the fire in a distance of 30 meters.

time considerations

As was already mentioned, in case of a fire emergency in a tunnel, the time aspect is of great

importance. Even a few extra seconds can have a significant impact in the total evacuation

process. That is why certain time considerations need to be fulfilled in order to establish the

tenability criterion. The following factors should, thus, be considered:

The time needed for a fire to ignite and become established

The time needed to notice and report the fire

The time needed for confirming the fire incident and to initiate a response

The time required for the self-evacuees to reach a place of safety

The time needed for the emergency personnel to arrive to the site

The time needed for the emergency personnel to trace and evacuate the ones who could

not self-evacuate

The time needed for fire brigade to start suppressing the fire

All the above should be considered when designing a tunnel. In case the time-of-tenability

criterion is not met in a tunnel, then the whole system should be designed that way so as to

maintain the tenability conditions for a minimum of 1 hour.

An interested reader can refer to the NFPA 502 [22], Annex B about the tenable environment

where more considerations are mentioned and analyzed.

3.2 EMERGENCY OPERATION

Fire safety management involves, basically, two categories of measures: the ones intending to

prevent a fire incident from taking place (prevention) and the ones intending to reduce the

damage after the ignition of a fire (protection). In order to incorporate the aspects of prevention

and protection in the fire safety design of a tunnel, many systems are employed and are working

together to assist in minimizing the loss of human lives. The tunnel operator depends on those

systems in order to determine and optimize the response in the most effective way possible. As a

result, it is understood that timely and coordinated activation of the systems for fire safety is a

determinant factor regarding incident response.

The active systems that are in use in case of an emergency in a tunnel are the following:

Emergency ventilation and smoke exhaust systems

Fire detection systems

Fire suppression systems

Automatic controls for the ventilation and smoke exhaust system

Radio telecommunication and telephone systems

Emergency lighting

Closed-Circuit Television

Systems for public address

Signage

Traffic management systems

50

Apart from the active systems mentioned above, passive measures of fire protection are also

applied and include the following:

Fire compartmentation

Evacuation passageways and exits

Use of fire resistant materials

As was, already, mentioned these systems should work together in a coordinated manner in

order to ensure a timely response to a fire incident. As a result, these systems are not

independent from each other and the operation of the one has an impact on the activation of the

other.

The emergency operation in a road tunnel, basically, consists of two phases: the evacuation

phase and the fire control phase. In the evacuation phase, self-evacuation and assisted

evacuation take place and the duration of this phase depends on the response of the systems

that detect the fire and notify the users. During the evacuation phase, the ventilation systems

should keep the smoke in a stratified layer and provide a smoke free egress path. The smoke

management should be implemented already in the stage of fire detection in order to ensure a

tenable environment within the tunnel for the various emergency phases.

In order to understand the interaction between the different systems involved in tunnel fire safety,

it is important to mention the emergency response process that is initiated during a fire:

Fire Detection: Many systems of fire detection can be combined in order to activate the

alarm and initiate the response plan. Manual fire alarm boxes, automatic fire detection

systems or closed-circuit television systems (CCTV) can be used to detect a fire.

Verification of the fire incident: Road tunnels are usually under 24-hour surveillance

and supervision through closed-circuit cameras. So, in case of a fire alarm, the person

supervising the tunnel visually confirms the fire emergency and gathers all the

information necessary to initiate the response plan.

Emergency response plan: This step includes notifying the fire brigade and initiating

the equipment necessary. The tunnel operator should apply the designated response

plan for the ventilation system in order to maintain a tenable environment within the

tunnel.

Since the focus of this research study is, mainly, on fire detection and evacuation, it is important

to gain an insight on the systems that most affect these fire safety design parameters. The

ventilation and smoke exhaust system have an impact on the fire detection system since they

significantly affect the fire phenomena that trigger fire detection (smoke, heat and flames).

Moreover, the egress provisions in a road tunnel determine the self-evacuation process and,

thus, the time needed for people to escape the tunnel in case of a fire incident. Finally, the

systems of fire detection, alarm and public address strongly affect the time needed to detect a fire

incident and, as a result, the time needed for the self-evacuation to start. Taking into account all

the aforementioned interrelations, the most important systems related to fire detection and

evacuation are described in the following sections.

51


The smoke produced by a fire in a tunnel can be a major threat for human lives. At this point, it

should be mentioned that, in case of a fire incident in a tunnel, most of the deaths are caused due

to smoke and not the heat. Smoke can threaten human lives in many ways since it reduces the

visibility and it is very toxic when inhaled. Especially in road tunnels, smoke propagates quickly

and threatens the users of the tunnel. In order to control the propagation of the smoke, the

ventilation system of the tunnel is used. As a result, the ventilation system is considered one of

the most significant safety systems in a tunnel.

More specifically, the aim of the ventilation system in case of an emergency is to extract the

smoke in the traffic direction (where there are not any people present). Apart from smoke

extraction, the ventilation system should prevent the backlayering effect (propagation of the

smoke towards the direction opposite to the traffic) and, at the same time, maintain the

stratification of the smoke layer. Then the smoke forms a stratified layer adjacent to the ceiling

and the people can, safely, self-evacuate. In addition, the fire fighters can access the tunnel and

approach the fire in order to extinguish it [72].

According to the NFPA 502 [22], the need for emergency ventilation depends on the length of the

tunnel. More specifically, emergency ventilation is not required in tunnels with a length less than

1000 meters provided that it can be proven by an engineering analysis that the required safety

levels are reached even without a ventilation system. For tunnels longer than 1000 meters, an

emergency ventilation system is required.

It is, also, prescribed within NFPA 502 that the emergency ventilation system should be able to

reach its full operational mode in a maximum time of 180 sec while reversible fans should be able

to achieve a full rotational reversal within 90 sec. In case of a fire in a road tunnel, the tunnel

operators should decide on a strategy for smoke control and management and select a sequence

of fan operation that needs to be modified as needed in order to make access to the fire site

easier. The effect of the location of the fans activated in case of a fire in a longitudinally ventilated

tunnel has been investigated through a sensitivity study [73]. More specifically, the impact of the

location of the fans on the distribution of temperature and structure of airflow within the tunnel

was investigated. This study concluded that the activation of the furthest group of fans can be

capable of preventing backlayering. However, in order to provide the required airflow speed, the

closest group of fans might need to be activated.

The most important factor when designing a ventilation system for a road tunnel is the critical

velocity. This is the minimum longitudinal air floe that is required in order to prevent the effect of

backlayering. Extensive investigation has been carried out regarding the critical velocities

together with the backlayering lengths in tunnels with longitudinal ventilation [74], [75], [76].

Based on former theoretical considerations and experimental results, a general rule could be that

the maximum value for the critical velocity is 2.5 – 3 m/s and can be applied to any fire scenario

in tunnels. These critical velocity values can be expected to prevent the backlayering effect and to

limit the smoke to the downstream region of the fire. Nevertheless, the stratification of the smoke

layer can be compromised.

In addition to this general rule, the following criteria for emergency ventilation and smoke control

can be applied [4]:

The longitudinal air velocity within the tunnel should not exceed 2 m/s in the proximity of

the fire. When this limit is exceeded, the smoke layer is mixed with fresh air and the

52

stratification is disturbed.

In the case when the longitudinal air velocity is almost zero, there is a backlayering effect

and the smoke can spread in a stratified way to both sides of the fire for approximately 10

minutes before starting to mix.

In case that the longitudinal air velocity is around 2 m/s and for a medium-size fire, the

smoke spreads only to the one side of the fire for about 400 to 600 meter before it starts

mixing. If the smoke extraction system activates early, then this mixing is prevented.

The presence of vehicles and other obstructions in the tunnel increases the vertical

turbulence and results in the mixing of the smoke.

In the case of transverse and semi-transverse ventilation systems, the fresh air that

enters the tunnel through the floor level or through ceiling openings respectively can

create airflow that tends to bring the smoke layer closer to the road surface. As a result, it

is recommended that the fresh air injection should be reduced or even stopped in a

smoke filled zone in order not to disturb the stratification.

According to Beard and Carvel, 2012 [6], longitudinal ventilation systems with jet fans were

proved to be highly effective for smoke control in case of fires up to 100 MW. It was, also,

concluded that these systems can only be applied in uni-directional tunnels. For this research

study, a uni-directional tunnel was chosen to be examined and a longitudinal ventilation system

with jet fans can be considered appropriate. As a result, in all the simulation scenarios performed,

longitudinal ventilation system is applied. Variations of this ventilation system were also examined

in terms of the number of jet fans installed and the longitudinal distances between them.

3.2.2 EGRESS EXITS

A complete and integrated fire safety design for a tunnel presupposes that the emergency

ventilation and fire suppression strategy is in full coordination with the emergency response plan

as well as the evacuation plan. Specially designated egress systems together with the response

plan should allow for a safe evacuation under a range of emergency scenarios and conditions.

In every tunnel escape routes should be provided and equipped with proper signage and lighting

in order to guide people during the self-evacuation. More specifically, these routes should guide

people to an emergency exit and, finally, to the open air either directly or through safe areas.

Depending on the tunnel construction, after reaching an emergency exit, the evacuees could be

directed to one of the following safety provisions [67]:

To escape rescue tunnels parallel to the road tunnel (egress corridors)

To the other tunnel bore through emergency cross-passages

To escape shelters

To a direct exit to the open air (shafts, portals)

In order to ensure a safe evacuation of the people in a road tunnel, emergency exits should be

provided within a specific distance throughout the tunnel. According to the NFPA 502, the

maximum spacing between the emergency exits should be 300 meters. In order to determine the

spacing between the exits, the following factors were taken under consideration:

The tunnel type

The chosen design fire size and the resulting smoke development

The egress analysis

The analysis of fire safety systems in order to maintain a tenable environment

53

The traffic management system

The emergency response plan

The uncertainties involved in human behavior during a fire emergency

3.2.3 FIRE SUPPRESSION SYSTEMS

Installing fire suppression systems in a road tunnel can contribute in preventing severe damages

to the tunnel structure and equipment provided that the system is activated at an early stage of

the fire development. Usually, in road tunnels fixed fire fighting systems are installed [4]. The

purpose of these systems is to limit the fire developments and prevent the fire from spreading to

other vehicles. With the activation of the fixed fire fighting systems (FFFS), the fire-fighters can

approach the fire location and, then, completely extinguish the fire. In addition, the tunnel lining is

protected and, thus, the structural damage can be limited which results in a shorter period of

disruption of tunnel operation for maintenance reasons. However, the use of a fire suppression

system could reduce the visibility in the tunnel which would significantly affect the evacuation

process and the intervention of the fire-fighters [4].

Usually, in tunnels deluge fire fighting systems are installed even though in some tunnels

automatic sprinkler systems have been installed. Deluge systems consist of a network of open

nozzles that are installed at the roof of the tunnel. The nozzles are divided into zones (usually

zones of 30 meters in order to cover the size of a HGV) and, in case of a fire, the zones above

and in each side of the fire are activated. The nozzles activate when a valve is opened and water

is sprayed in each zone.

Deluge systems are considered more appropriate for tunnels because of two reasons. First, the

ventilation system could transfer the heat away from the fire and, thus, the sprinklers that are not

above the fire could be activated. Second, a fire that is developing fast could result in a great

amount of heat over a large area and, thus, a large number of sprinklers could be activated

overcharging the water supply. Contrary to that, the deluge system uses a fixed amount of water

and the zones around the fire could only be activated through proper detection and monitoring.

Resulting from the above and as is, also, recommended by PIARC [21], a FFFS should be

installed in a tunnel only on the condition that an effective method for fire detection and

identification of the fire location is provided. Otherwise, its efficiency is not guaranteed and it

could, at the same time, compromise safety since it may result in reducing the visibility in the

tunnel. As a result, the FFFS should interact in an efficient way with the detection system installed

in the tunnel [77], [78].

An interested reader could refer to the report by PIARC, ROAD TUNNELS: AN ASSESSMENT

OF FIXED FIRE FIGHTING SYSTEMS, PIARC Technical Committee C3.3 Road tunnel

operations, where all the developments on FFFS are mentioned together with their recent

applications in tunnels. In addition, the types of FFFS are described in detail and the interactions

between the FFFS and the ventilation systems are analyzed.

54

CHAPTER 4: Computational Fluid Dynamics In this chapter, the basics of Computational Fluid Dynamics (CFD) will be shortly explained so that the reader can comprehend the nature of the simulations that were performed. In addition, the basic principles of the software used will be mentioned in order to better understand the input and output of the simulations. Finally, the crucial issue of validation of the numerical models will be dealt with and the validation cases will be presented.

4.1 INTRODUCTION

In case of a fire in a tunnel, the energy released creates buoyancy forces that give rise to three-

dimensional flows characterized by high complexity. The parameters that normally modify the

general behavior of a tunnel fire are the heat transfer and turbulence which are, in turn,

influenced by the geometrical characteristics of each tunnel and the operating ventilation system

[6].

In order to tackle the challenge of understanding the complex fire behavior, specialists have

developed and applied Computational Fluid Dynamics (CFD) techniques. Especially, during the

last two decades, the use of CFD as a tool for Fire Safety Engineering (FSE) has been widely

used in order to analyze fire and smoke behavior in tunnels. As a result, it is important to acquire

an insight into the basic principles of CFD and into how it can be used in tunnel applications [68].

4.2 THE BASICS OF CFD

CFD simulations, in general, require the solution of the complete set of partial differential

equations describing the conservation of mass, momentum and energy. This set of equations is

numerically solved and results in a detailed prediction of temperature and velocity distributions,

concentration of species, heat fluxes and other parameters. In CFD software, the computational

domain is separated in a high number of control volumes for which the conservation laws are

applied [68]. The governing equations are presented in the Appendix D.1.

limitations of cfd

Even though the CFD tools have rendered the modeling and understanding of fires and the

integration of complex physical phenomena easier, they still introduce certain limitations. The

most fundamental limitation for the use of CFD lies in the averaging procedure applied in the

model equations [79].

In general, two alternative approaches are used in CFD simulations. First, CFD models are based

on a framework developed through a time-averaged form of the Navier-Stokes equations (RANS)

and, more specifically, the k-e turbulence model which was initially introduced by Patankar and

Spalding [80]. Second, a turbulence modeling approach that includes spatial averaging and is

known as Large Eddy Simulation (LES) is used.

So, as described above, there are significant uncertainties involved in the modeling of turbulence

with CFD models. Added to that is the uncertainties that are related to the buoyancy, turbulent

55 combustion chemistry, radiative heat transfer and burning of condensed-phase fuels [68].

In the case of modeling fires in tunnels, under complex ventilation scenarios, uncertainties are

also introduced in determining the appropriate boundary conditions. More specifically, the

unknowns are usually related to the following:

Geometry and boundary conditions: meteorological conditions at the tunnel portals,

initial velocities of air in the tunnel, wall friction, wall heat transfer and presence of

obstructions and vehicles.

Modeling the fire: source of combustion, actual fire load, fire growth rate.

Grid: the choice of the right grid size

Resulting from all the above, CFD can be considered to be an appropriate tool for modeling the

complex flow behavior of fires in tunnels. Nevertheless, a lot of research has to be performed

because complete fundamental knowledge of the physics involved in fire does not, still, exist. In

order to compensate for this gap in knowledge, certain assumptions are introduced in the

mathematical process that lead to uncertainties and inaccuracies [6]. For that reason, when

performing a CFD analysis, the designer should take two additional actions, verification and

validation, in order to ensure the appropriate use of the models and the reliability of the results.

verification and validation

At this point, it is considered very important to understand the terms verification and validation in

order to form a strategy and apply them in the context of this research study. First, a definition for

each one of these terms needs to be given. The American Society for Testing and Materials

(ASTM) in 1997 was the first to publish an international standard regarding the evaluation of the

predictive capabilities of fire models. ASTM defines verification and validation as follows [81]:

Verification: The process of determining the correctness of the solution of a system of

governing equations in a model. With this definition, verification does not imply the

solution of the correct set of governing equations, only that the given set of equations is

solved correctly.

Validation: The process of determining the correctness of the assumptions and

governing equations implemented in a model when applied to the entire class of

problems addressed by the model.

A standard for the assessment on totally deterministic CFD modeling for fire was published by

ISO in the year 1999. In this standard, the definition for validation remained the same as defined

by the ASTM whereas the definition of verification changed into the following [81]:

Verification: The process of checking a mathematical fire model for correct physical

representation and mathematical accuracy for a specific application or range of

applications.

The process involves checking the theoretical basis, the appropriateness of the

assumptions used in the model, that the model contains no unacceptable mathematical

errors and that the model has been shown, by comparison with experimental data, to

provide predictions of the course of events in similar fire situations with a known

accuracy.

56 As mentioned in the definition itself, the process of verification is, basically, a check of the

mathematics included in the fire simulation. The developers of CFD software, usually, make a lot

of research and work in order to ensure that the equations are solved correctly within the

software. In addition, a lot of documentation is provided by the software developers together with

instructions on how to achieve proper verification for each model. The user of the model is

advised to perform sensitivity analysis for basic input parameters (grid sensitivity, sensitivity of

LES parameters, sensitivity of radiation parameters, sensitivity of thermophysical properties of

solid fuels).

Usually, in order to validate a computational model, a comparison is made between the model

itself and experimental data. The differences that are determined during this comparison and

cannot be justified by numerical issues are explained by mistakes in the hypothesis or

simplifications and assumptions in the process.

As can be derived by all the above, the verification and validation processes are necessary in

every fire simulation in order to render the results reliable and give value to the study results. That

is why a number of institutions are working hard to develop CFD tools appropriate for specific fire

applications [68].

4.3 FIRE DYNAMICS SIMULATOR (FDS) SOFTWARE

As was already mentioned, in this research study, numerical simulations were performed in order

to examine various scenarios of fires in road tunnels. The software that was chosen for the

simulations is the Fire Dynamics Simulator (FDS) developed by the National Institute of

Standards and Technology (NIST) of the United States Department of Commerce, in cooperation

with VTT Technical Research Centre of Finland.

general information about FDS

FDS is a Computational Fluid Dynamics (CFD) software that was developed to solve fire-driven

fluid flow. More specifically, it solves numerically a form (Large Eddy Simulation) of the Navier-

Stokes equations that are appropriate for low-speed, thermally-driven flow and that focus on

smoke and heat transport from fires. FDS software is accompanied by Smokeview, a software

that is used to provide a visualization of the output generated by FDS. The base model of the

tunnel that will be used for the research study will be created in PyroSim which is a graphical user

interface for the FDS [31], [82], [83].

The first version of FDS was released in February 2000 and since then a lot of progress have

been made to, finally, release the current version 6. At this point, FDS can be used for a number

of applications and for a number of purposes like [McGrattan et al., 2012c]:

Low speed transport of heat and combustion products from fire

Radiative and convective heat transfer between the gas and solid surfaces

Pyrolysis

Flame spread and fire growth

Sprinkler, heat detector, and smoke detector activation

Sprinkler sprays and suppression by water

The reason why this specific software was chosen is, mainly, because it was recommended by

fire safety specialists and, also, because there is a lot of documentation and research already

57 published [84], [82], [85], [86]. It is considered a flexible software since it can be applied to fires of

different scales. In addition, it can be used to simulate situations where a fire is not involved, such

as ventilation. Moreover, the software is distributed for free together with the EVAC component

which is a modeling tool for pedestrian flows. Finally, the FDS package contains detailed

Verification and Validation Guides developed by the software developers and the user community

that the user can consult in order to deal with the challenging issues of verification and validation.

Added to that, the FDS has constantly active forum discussions where the users can request for

advice and highlight any issues related to the software. The forum is supervised by the

developers that respond immediately to any questions and comments and update the issue

tracker section of their website.

basis of FDS

As mentioned in the previous section, CFD simulations are based on certain models regarding

the governing equations, the turbulence models and other parameters that concern the geometry

and boundary conditions. In FDS these models are the following [84]:

Hydrodynamic model

The FDS software solves a form of Navier-Stokes equations that are appropriate for low-

speed, thermally-driven flow and that focus on smoke and heat transport from fires. The

LES approach is used in order to treat turbulence. There is, also, the possibility of

performing a Direct Numerical Simulation (DNS) provided that the applied numerical

mesh is fine enough. The default mode is set to LES.

Combustion model

For the majority of the applications, FDS uses a single step, mixing-controlled chemical

reaction of three lumped species (air, fuel, products) which represent a mixture of

species. Namely, air is considered a lumped species that is a mixture of nitrogen,

oxygen, carbon dioxide and water vapor. The fuel and air species are, by default,

explicitly computed. There is, also, the option to include multiple reactions that are not

mixing-controlled.

Radiation Transport

In most of the cases, the radiative heat transfer is incorporated in the model through the

solution of a radiation transport equation for a gray gas. This equation is solved through a

technique that is similar to finite volume methods for convective transport which is named

Finite Volume Method (FVM).

Geometry

The FDS software, basically, performs an approximation of the governing equations on a

rectilinear mesh. Any rectangular obstructions should be aligned with the applied mesh.

Otherwise, they are forced to conform with the mesh by default.

Multiple meshes

FDS offers the option to apply more than one rectangular mesh in a calculation. This

option is chosen when the computational domain cannot be easily fitted within a single

mesh.

Boundary conditions

Every solid surface in a model is assigned with thermal boundary conditions together with

information about the burning behavior of its material. Empirical correlations are used in

58

order to deal with heat and mass transfer from and to solid surfaces. Nevertheless, in the

case of a Direct Numerical Simulation (DNS), it is not possible to compute directly the

heat and mass transfer [83].

4.4 VERIFICATION AND VALIDATION OF SIMULATION MODELS

In the context of this research study, verification and validation had to be performed in order to

make sure that the results were reliable and of high scientific value. As was already explained,

verification and validation are two different terms and, thus, processes and are explained very

briefly in the graph that follows.

In this section, the strategy followed in order to verify and validate the models used is going to be

described in detail.

4.4.1 VERIFICATION

The FDS software is internally verified and a lot of documentation is provided regarding the

mathematical models used together with verification cases and examples. The FDS Verification

Guide includes the work performed by NIST and VTT in order to verify the algorithms used. As a

result, the theoretical basis of the model is checked by the developers of the software and has

proved to correspond to the reality [87].

The verification of the theoretical basis and mathematics of FDS were checked by means of:

Analytical tests

Numerical tests

Sensitivity analysis and

RELIABILITY OF THE MODEL

VERIFICATION

Make sure that the equations

are solved correctly

FDS software is internally verified

VALIDATION

Comparison with experiment

Choice of full-scale fire test in

tunnel in Norway Fig.4.1 Validation and

verification strategy

59

Code checking

The last two steps are the responsibility of the user and, as a result, these were given special

consideration during this research study.

mesh sensitivity analysis

In every FDS simulation, the calculations should be performed in a domain that consists of

rectilinear volumes that are called meshes. Every mesh is separated into rectangular cells and

the number of the cells in each simulation depends on the required resolution of the flow

dynamics. It is suggested by the software developers that the mesh should be subdivided into

uniform cells that resemble cubes of the same width, length and height.

A challenge that FDS users have to face is to define the appropriate resolution for each

simulation which means defining the grid spacing. According to the FDS software developers, the

user should first create an input file with a relatively coarse mesh. Then, the mesh should be

gradually refined until there are no significant differences in the results. This process is known as

sensitivity study.

The size of the grid is considered the most significant numerical parameter in a simulation. This is

because it determines the accuracy of the discretized governing equations by means of spatial

and temporal averaging [79]. As a general rule, the finer the mesh, the better the solution of the

governing equations. The FDS software incorporates a second-order spatial and temporal

accuracy. This, actually, means that decreasing the dimensions of a grid cell by half will decrease

the discretization error in the equations by a factor of 4. At the same time, though, halving the

size of the grid cell, the computational time required for the simulation will increase by a factor of

24=16. As a result, during the sensitivity study of the mesh, the FDS user should consider a

balance between the resolution and the computational time needed.

A general rule is provided by the FDS User Guide according which, the dimensionless parameter

D*/dx is used to define the appropriate grid size. The value of this parameter should be between

4 and 16 where a low number stands for a coarse grid and a high number stands for a fine grid

[84]. The parameter D* is the characteristic diameter of the fire given by the following expression

[84]:

So, for each fire size (HRR) and each nominal size of a mesh cell, a check can be performed by

using the expression above in order to confirm that the cell size chosen is within the acceptable

limits.

A sensitivity analysis was, also, performed for the numerical grid in order to ensure that the

chosen size of the cells can generate reliable results. First, a coarse mesh with grid cells of 1*1*1

meters was applied on the geometrical model. All the simulations were performed for this mesh

and the relevant results were recorded. The expected outcome was that with a coarse mesh the

averaging that would take place within such a cell will result in values that would be far from the

reality. The comparison with a finer grid proved that expectation to be true.

Then, some characteristic simulations were performed with a finer mesh with grid cells of

0.5*0.5*0.5 meters. The computational time required for running one simulation where the whole

60 domain would be embedded within a mesh with such grid cells was significantly larger. In fact, it

was considered that running all the simulations with such a mesh would not be feasible within the

context of this research study. This is because it would require having access to advanced

hardware and a broader time scale.

The FDS software offers another alternative that could compromise for the use of a finer mesh in

the whole domain. Applying multiple meshes to the model is an option in which the computational

domain is embedded in more than one mesh. The multiple meshes could be connected even

though this is not necessary.

In the case of multiple meshes, the question of accuracy is set. This is because of the possible

loss of information in the boundaries between two consecutive meshes. According to the FDS

User Guide, an FDS user should place the boundaries away from the area of activity and assess

the importance of the information transferred from a finer mesh to a more coarse mesh.

In addition, in order to avoid mistakes and loss of information the consecutive meshes should be

properly aligned. The basic rule for a proper mesh alignment is that the cross-sectional area of

the abutting cells should be the same.

Following the instructions provided by the FDS User Guide regarding the use of multiple meshes

with different grid cell sizes, the mesh sensitivity study was performed. So, finally, three different

cases were examined:

A coarse mesh with a cell size of 1.0*1.0*1.0 meters which was applied to the whole

tunnel domain.

Three consecutive meshes of different grid cell size each. In the zone where the fire

phenomena were taking place a finer grid (0.5*0.5*0.5 meters) was applied and in the

rest of the tunnel domain coarser grid was applied. More specifically, two grids with a cell

size of 1.0*1.0*1.0 meters were constructed and applied in the beginning and end of the

tunnel.

A fine mesh with a cell size of 0.5*0.5*0.5 meters which was applied to the whole tunnel

domain.

As was already mentioned, a parameter that was taken under consideration for choosing the final

grid size for which all the simulations would be run was the time needed for completing one

simulation scenario. During the sensitivity study, it was observed that the time cost in the case of

the finer grid (0.5*0.5*0.5 meters) was approximately 115 hours or else more than 4 days of

continuous simulating. This means that running all the simulation scenarios with the finer grid

would not be feasible within the context of this research study. On the contrary, the duration of

the same simulation with a coarse grid (1.0*1.0*1.0 meters) was approximately 8 hours. Finally,

the duration of the same simulation with multiple meshes applied in the tunnel domain was

approximately 34 hours or else 1 day and a half of continuous simulating. This duration was

considered feasible for such a research study.

As a result, depending on the level of accuracy that each mesh size would provide and according

to the computational time needed for each case, the grid size to be used will be selected. The

simulation scenario that used as the case for the grid sensitivity study was a HGV fire in a cut-

and-cover tunnel. After running the simulation for the three different grid sizes, the following

graphs were generated. The graphs depict the temperature development in time in the tunnel in

three different measuring points in the tunnel. The points chosen to be presented in the graphs

61 were the ones for which the highest temperatures were recorded and, as a result, the ones in the

close proximity of the fire.

As can be observed in the graphs, the results from the simulation with the coarse grid have the

same trend with the other two cases but they mostly diverge from the others. In general, the

temperature in the case of the coarse grid was underestimated. On the other hand, the case with

the multiple meshes approaches the case with the finer grid both in values and in the general

trend. It can, also, be observed that in the earliest stages of the fire development (first 120

seconds) the temperature line of the multiple meshes case converges towards the one for the

finer grid case. This means that the results generated for a simulation with multiple meshes could

be considered reliable and similar to the ones for the finer grid.

Derived from all the above, the case with the multiple meshes was chosen to be used for

performing all the simulation scenarios in this research study. It was concluded that in this case

the simulation time is reduced and, at the same time, the desired levels of accuracy of the results

are ensured.

During the mesh sensitivity study, another alternative was also considered in order to balance the

accuracy of the results and the required computational time. For this alternative, the model could

0.00

50.00

100.00

150.00

200.00

250.00

0 10 20 30 40 50 60 70 80 90 100 110 120 130 140 150 160 170 180

Tem

pera

ture

GRID SENSITIVITY STUDY [FL.266]

Grid size 1.0*1.0*1.0 m

Multiple Meshes

Grid size 0.5*0.5*0.5 m

0.00

50.00

100.00

150.00

200.00

250.00

0 10 20 30 40 50 60 70 80 90 100 110 120 130 140 150 160 170 180

Tem

pera

ture


0.00

50.00

100.00

150.00

200.00

250.00

0 10 20 30 40 50 60 70 80 90 100 110 120 130 140 150 160 170 180

Tem

pera

ture


Fig.4.2 Temperature-time graphs for grid sensitivity study

62 be reduced by isolating the part of the tunnel where the fire is located and running the simulation

for a fine mesh. The information necessary for the exterior boundaries of the mesh would be

acquired by running a simulation for the whole tunnel with a coarser mesh.

code checking

The operation of the FDS software is based on a text input file that is either written by the user or

written through the graphical user interface for FDS which is the PyroSim software. The input file

is a text file which contains parameters that are organized into groups of namelists. All the

information necessary to describe a scenario is provided to the FDS software through this input

file [84].

It is considered important to provide a correct and carefully structured input file in order to have

an overview that allows the user to check the code and identify any mistakes. In this research

study, a certain strategy was followed in order to check the code in the input file. First, the FDS

User Guide was thoroughly studied and each namelist that defined a specific input was

determined and deeply understood. Then, a basic input file containing only the information

necessary to set up an FDS model was created with the help of a fire safety specialist in Deerns

Nederland B.V. with experience on the FDS software. Then, the necessary changes were made

depending on the requirements of each simulation. The code was written separately for each

component of the simulation and was added in steps so that the input file does not become very

long and full of information at once.

Moreover, the PyroSim software was used in order to have a visual overview of the model and

the parameters involved in each simulation. This way a better control over the model

characteristics and the changes performed was achieved. The basic input file that was written

was imported in PyroSim and the base model was, thus, generated. PyroSim has well-structured

menus, easy to be understood by even a new and inexperienced user and offers a 2D and 3D

visualization of the geometry of the model and of all the components added like, for example,

obstructions representing the vehicles or jet fans, the fire and the devices. After setting up the

model and inserting all the relevant parameters, PyroSim generates the text input file for FDS.

This file was checked for any mistakes before running each simulation.

4.4.2 VALIDATION

As was previously mentioned, the most common way to validate a simulation model would be to

compare the results with a known experiment. That was the strategy followed in this research

study in order to validate the simulations that were performed. More specifically, full-scale field

fire tests performed in tunnels that match the required simulations and that were well-documented

were chosen. These experiments were recreated with the use of the FDS software and the

results were compared.

Before choosing the experiment, a brief research regarding the parameters that need to be

recorded for a validation case was carried out. According to Smardz, 2006 [85], in order for an

experiment to be used effectively as a validation case, accurate information should be provided

regarding all the model inputs (initial and boundary conditions) as well as the model outputs (the

predicted quantities). Regarding the model inputs, the required information for simulating a fire in

a tunnel is the following:

A detailed description of the tunnel geometry

Information on the material of the tunnel walls and its properties (especially, its thermal

63

properties)

Detailed information on the size of the fire, the heat release rate (HRR), soot yield

Information on the conditions at the external boundaries (temperature, pressure, flow

velocities)

Information on the performance of active elements like, for example, ventilation system,

sprinklers or fire detectors

Regarding the results of the experiment, measurements should be recorded in specific locations

that are relevant to the case that is investigated. The experimental results should be some of the

following [85]:

Temperatures of gas and solids involved

Flow velocities

Measurements of heat flux

Concentrations of certain species (CO2, CO, O2)

Visibility measurements

Visual information of the behavior of fire and smoke

Taking all the above into consideration, the experiment that was chosen as reference for the

validation case was a full-scale field test performed in a tunnel in Norway.

4.4.2.1 RUNEHAMAR TUNNEL FIRE TESTS

In 2003, five large-scale fire tests were performed in the Runehamar tunnel situated in Norway.

The fire tests in the Runehamar tunnel were initiated and performed by the SP Technical

Research Institute of Sweden. In order to carry out the complex tests, a Consortium was

developed together with TNO in the Netherlands and SINTEF in Norway. It was acknowledged

that the tests in the Runehamar tunnel resulted in a significant contribution to the current

knowledge of fire dynamics in the case of large fires in tunnels. They have illustrated that a typical

truck loaded with heavy goods (Heavy Good Vehicle, HGV) produces a rapidly growing fire

reaching up to 200 MW. The maximum gas temperatures measured beneath the ceiling were

approximately 1350oC [27].

The reason that these tests were chosen as reference was that they were well-documented and

the results were presented in detail. There are various publications for the Runehamar fire tests

[88], [89-93]. In these publications the focus has, mainly, been on measuring HRR, temperatures

of gases in the ceiling level, fire spread and flame lengths, pulsations in the tunnel flow, heat

fluxes, humidity and toxicity. Although these publications are frequently referred to, the data

published were mostly preliminary and not deeply analyzed. For that reason, a report was

published by the SP Technical Research Institute of Sweden [Haukur Ingason, Anders

Lönnermark, Ying Zhen Li] were the preliminary data are further analyzed and corrections are

provided. The same report, also, enables the reader to make better use of the data by providing

more in depth information about the way that the tests were performed. Finally, within this report

there is a review of the articles that were published after the performance of the tests from 2003

and on. As a result, this report was thoroughly studied in order to perform the validation of the

models used in this thesis.

First, in order to reproduce the tunnel geometry and tests, a detailed study of the Runehamar

tunnel and the test had to be carried out. The Runehamar tunnel is a bi-directional tunnel

consisting of one tube and it stopped being used in the late 1980s. It has a length of

64 approximately 1600 meters, a width of 9 meters and a height of 6 meters. The total cross-section

is about 47 m2. In its longitudinal section, the tunnel has three different slopes:

An average uphill slope of 0.5% up to 500 meters from the east portal

A plateau of 200 meters

An average downhill slope of 1% for a length of 900 meters towards the west portal

fire location

The fire was located on the downhill section of the tunnel and, more specifically approximately

1037 meters from the east portal. It was mentioned in the SP report that this fire location was of

particular interest since the small downhill slope resulted in high pressure resistance for the

mobile fans used for ventilation. The fire was produced by a HGV trailer mock-up measuring

10.45 meters by 2.9 meters and with a total height of 4.5 meters. The platform floor was not

adjacent to the road surface but was 1.1 meters above it.

mobile fan units

In order to create a longitudinal flow in the tunnel, two mobile fan units were used. One was

located 12 meters away from the east tunnel portal, outside of the tunnel. The second one was

located inside the tunnel, around 50-60 meters from the east portal. The fans had a diameter of

1.25 meters and an engine of 100 HP providing 2600 N of axial thrust at 2000 RPM and the

primary air flow rate was 47.2 m3/s for each fan. Fig.4.4 [27] shows the location of the mobile fan

outside the tunnel.

1650 meters

Fig.4.3 Runehamar tunnel longitudinal section

Fig.4.4 Mobile fan units at the east entrance of the Runehamar tunnel

65

measurements

Before setting up the model in FDS, it was important to study the exact locations where

measurements were recorded. These were mentioned in detail in the SP report for the

Runehamar fire tests. Temperature measurements were recorder both upstream and

downstream of the fire:

Upstream of the fire: Thermocouples were positioned at -15 m, -25 m, -40 m, -70 m and

-100 m and 0.3 m beneath the ceiling.

Downstream of the fire: Thermocouples were positioned at 0 m, +10 m, +20 m, +40 m,

+70 m, +100 m, +150 m, +250 m, +350 m and +458 m and 0.3 m beneath the ceiling. At

the position +458 m, temperature measurements were recorded at five different height

positions: 0.7 m, 1.8 m, 2.9 m, 4.1 m, and 5.1 m, respectively.

Concentrations of certain gases (O2, CO2 and CO) were measured at the same locations

mentioned above. The measurement points setup is shown in Fig.4.5 [27]:

meteorological conditions

Next, in order to define the exterior boundary conditions in FDS, the meteorological conditions at

the time that the tests were performed had to be determined. It is recorded that the outside

temperature at the time of the tests varied between 9-14oC and the temperature in the tunnel at

the location of the fire was 10-11oC before the realization of the tests. Wind measurements were

performed before each test. Specifically, before test T1, the wind speed was measured to be

between 1.3-1.4 m/s blowing in the same direction as the fans.

chosen test

As was already mentioned, five tests were performed in the Runehamar tunnel each one resulting

in a different fire size. The test chosen as a validation case for this research study is T1. This

choice is justified by the fact that this test was previously simulated with the FDS software and the

existing report could be consulted in order to determine the input values to be used and to

provide a reference for checking the correctness of the output.

Fig.4.5 The measurement station 458 m downstream of the fire

66

The setup for the T1 included 360 wood pallets with dimensions 1,200*800*150 mm, 20 wood

pallets with dimensions 1,200*1,000*150 mm and, finally, 74 PE plastic pallets with dimensions

1,200*800*150 mm. All the pallets are covered with a 122 m2 polyester tarpaulin. The maximum

HRR at this test was measured to be 202 MW [27]. Also, an average longitudinal velocity of 2 –

2.5 m/s was measured and the ambient temperature was recorder to be 11oC. As was

determined by Ingason and Lönnermark, the average heat of combustion was estimated to be

18.5 MJ/kg, the fraction of soot production is 2.2% and the fraction of CO production is 0.88%.

4.4.2.2 FDS INPUT AND SIMULATIONS

At this point, taking under consideration the information regarding the geometry of the

Runehamar tunnel and the test setup and conditions, a model was created in FDS software in

order to run the simulation for the validation case. The model reproduced the tunnel geometry,

the fire size and location, the fire development and the meteorological conditions.

simulation time

In the tests performed in the Runehamar tunnel, most of the measurements lasted for about 60

minutes. Nevertheless, for the fire detection time, the general rule is that the first 2 minutes are

the most critical ones [6]. So, in order to reduce the required computational time and use a

relatively fine grid, the first 5 minutes (300 sec) were chosen to be simulated.

grid size

In order to determine the appropriate grid size, a research regarding previous attempts to

numerically simulate the Runehamar tunnel tests was performed. Ingason and Lönnermark for

the SP Technical Research Institute of Sweden, performed a numerical simulation of the

Runehamar tunnel test T1 with a grid of 20*20*20 cm which is a significantly fine grid. However,

the computational time needed to perform a simulation with this grid size was high and, thus, a

more coarse mesh had to be chosen.

A research study released on May 2015 by Jeroen Wiebes Kjos, determined that a grid of

50*50*50 cm provides enough accuracy for reproducing the Runehamar test for validation

purposes. This was proven through a grid sensitivity study where two simulations were run for a

grid size of 50*50*50 cm and a grid size of 40*40*40 cm. It was concluded that the results

converged towards a certain value and it was decided that the 50*50*50 cm grid size could be

used.

Taking into consideration all the above, a mesh of 50*50*50 cm was chosen in the area of the fire

and a mesh of 100*100*100 cm was chosen for the section of the tunnel where the fire

phenomena are not present. This way, the use of a multiple mesh can, also, be tested.

fire development

The fire development for the T1 is represented in the following graph as the HRR which was

obtained during the test using the oxygen consumption method. As can be derived by the graph,

the maximum HRR was 202 MW.

67

The HRR was provided as input to the FDS software as a fire ramp according to the above curve.

After running the simulation, the HRR that was given as output is represented by the following

graph.

fire area

In FDS, the fire is imported as a surface and, thus, an area has to be defined. More specifically,

the fire was imported as a rectangular obstruction and the top surface represents the fire. The

input parameters required are the surface area of the fire expressed in m2 and the HRR per Unit

Area (HRRPUA) expressed in kW/m2.

The fire area could be imported in the model as an obstruction with the same dimensions as the

test setup which has a width of 2.9 meters, a length of 10.45 meters and a height of 3.3 meters

with the fire load standing at a height of 1.1 meters above the road surface. However, during the

first 300 seconds of the fire development which are simulated, the fire does not develop fully and,

as a result, an estimation of the fire area should be made.

Fig.4.6 HRR-time (min) graph for Runehamar fire test T1

Fig.4.7 Comparison between simulation and Runehamar test results

69

The fire growth factor was translated into spread rate expressed in m/s by the fire specialist in

Deerns, Dave Hensen and the following model was created:

So, using a t2 curve approaching the given graph for the HRR measured during the Runehamar

tests, it was observed that the time to reach the 1 MW was 120 sec. As a result, in the model

presented above, the tc (which in the to described above) was chosen to be 120 sec and the fire

could be characterized as fast to ultra fast. The time for which the fire size has to be predicted is

300 sec and the fire spread rate was, already, imported in the model as approximately 0.0053

m/s. Finally, the surface area of the fire was calculated to be 12.5 m2.

The fire dimensions were rounded up in order to fit to the grid cell size. As a result, the fire

obstruction that was created in FDS has the following dimensions:

Width: 3 meters

Length: 4 meters

Height: 1 meter

The HRRPUA is defined by the maximum HRR (15 MW) and the fire area (12 m2) as follows:

HRRPUA = 15,000 kW / 12 m2 = 1250 kW/ m

2

This value was imported in the FDS when defining the fire surface.

heat of combustion (CO and soot yields)

As it was already mentioned in the description of T1, Ingason and Lönnermark determined that

the average heat of combustion was estimated to be 18.5 MJ/kg, the fraction of soot production is

2.2% and the fraction of CO production is 0.88%.

radiation

The radiation emitted from the fire is expressed as a function of the chemical composition and the

flame temperature. When simulating a large fire, the calculation of these parameters is not

reliable. For that reason, in order to tackle this problem, FDS introduces the radiative fraction of

the HRR. In open fires and for the most commonly used fuels, the radiative fraction ranges

between 20% and 40%. The default value used by FDS for LES simulations is 35% (or else,

0.35).

In the confined environment of a tunnel, however, the flame usually extends along the ceiling of

the tunnel. Compared to the flame height in an open fire, the flame length in case of a large fire in

Table 4.1 Results of the model for fast fire development based on quadratic fire curves

70 a tunnel is significantly larger. This, respectively, means that the radiative fraction is higher. As a

result, Ingason and Lönnermark suggested using a radiative fraction of 0.45 for reproducing the

Runehamar tunnel tests [94].

The focus of the validation study was, mainly, on the temperatures near the ceiling because this

is what affects the response of the fire detection system. Through the validation study, it was

observed that there is usually an overestimation of the developed temperatures at the first stages

of the fire development. However, the temperature development trend of the simulation follows

relatively accurately the one of the experimental measurements. Another observation is that the

FDS software predicts more accurately the temperatures in the area downstream of the fire

location than those in the area upstream the fire.

The validation study performed indicated the differences between the experimental data and the

simulations. The outcome of the validation study correlates with the outcome of previous studies

which implies that proper use of the FDS software was achieved. Since the objective of this thesis

is to investigate the performance of the detection systems, the focus of the validation process will,

mainly, rest on the validation of the simulation of the specific technology investigated. As a result,

this validation study was very important in order to reassure that the software was properly used

and to determine any points of special attention.

Fig.4.9 Temperature-time graphs for comparison of results of simulation with Runehamar test fire

oC o

C

oC

T (sec)

T (sec)

T (sec)

71

CHAPTER 5: Fire detection

This chapter offers an insight into the basics of fire detection: the role of fire detection in a road tunnel, the types of fire detection available in the market and their working principles and performance. In addition, the choice of specific types of fire detectors to be simulated will be justified.

5.1 INTRODUCTION

Fire detection systems are an essential part of the fire protection systems in road tunnels. Their

role is to identify an emergency, provide a timely warning of a fire incident, determine its exact

location and monitor the development of the fire in the tunnel [9]. As a result, the detection

system is capable of aiding in indicating the proper evacuation route and guiding the firefighting

operations [37]. Nevertheless, the most significant “duty” of a fire detection system in a tunnel is

to coordinate the activation of the smoke extraction system in an optimal way [10].This way, the

smoke stratification is maintained and the possibility to survive is significantly increased.

The selection of the appropriate fire detection system is made according to the fire safety goals

and objectives determined during the specific fire safety design. It is true, though, that fire is a

very complex phenomenon and, depending on the nature of each fire, different amount and

sequence of heat, smoke and flames are generated. As a result, by using detection systems with

various sensors, better automatic control is achieved [4].

5.2 DECISIONS RELATED TO FIRE DETECTION SYSTEMS

Firstly, in order to explore all the possibilities for fire detection in road tunnels, a literature review

has to be carried out. There are a lot of references in literature separating the detection

technologies depending on their detection principle.

During a fire, several material and energy conversions take place and result in end-products

referred to as fire phenomena. Automatic fire detection, in principle, converts certain fire

phenomena into electrical signals. As a result, the parameters of the fire that define the detection

principles are the following [6]:

Smoke

Smoke is usually detected by optical systems based on the principle of reflection. Light

emitters and light receivers are used and the loss of light intensity caused by smoke is

what triggers the alarm.

Heat

In this case, the ambient temperature is measured be the detector. The alarm threshold

is a maximum temperature value. When this is exceeded, the alarm is activated by the

system.

72

Flames (radiation)

Flame is a fire phenomenon that results in the emission of radiation. Flame detection

uses light-sensitive sensors that trigger an alarm when a certain radiation is received.

As a result, the fire detection technologies are developed based on these basic principles.

Specifically in road tunnels, the methods that are basically used for fire detection are:

Linear heat detection: a continuous heat detection cable able to detect fires across the

whole tunnel length

Closed-circuit television: using cameras to monitor the environment and detect fire

incidents

Video image smoke detection

Flame detection

Smoke and heat detectors

Spot-type heat and smoke detection

In order to decide on the most up-to-date detection systems that are used in road tunnels, a

market research had to be performed.

Through this market research, specific technologies were proven to be the most effective ones for

detecting fires in tunnels and are described in the following sections.

5.2.1 LINEAR HEAT DETECTION (LHD)

This type of fire detection technology is widely used in road tunnels and is recommended by

many professionals and companies. There are three main types of linear heat detectors: Analog,

Digital and Fiber Optic. The Fiber Optic Linear Heat detector is already used in many tunnels and

is proven to assure an accurate detection since it both detects the fire and indicates its location.

For that reason, this type of LHD was chosen to be studied in the context of this thesis.

The Fiber Optic LHD consists of a control unit and a fiberglass conductor inside a metal tube. A

laser light source is installed in the control unit that sends a light beam along the fiber optic cable.

The control unit, also, receives a specific response light signal which results from the reflection

inside the conductor [6]. During this transmission, a fraction of the light is lost through scattering

and the rest is received and processed by the control unit. This way, the LHD systems can sense

temperature changes based in the, as called, Raman Effect [4]. Namely, any fluctuation of

temperature results in changes of the properties (intensity, phase, polarization) of the light wave

that propagates within the fiber [95].

In general, in LHD technology, the alarms are activated based either on an absolute temperature

threshold or on a rate of temperature rise. Information on the location of the fire can, also, be

provided by this system [95].

73

The LHD systems have several advantages over other technologies and that is why they are

widely used in tunnel applications. The main advantages are the following:

The cable is resistant to harsh environmental conditions like high temperatures and

presence of toxic emissions and dust.

There is flexibility in the installation since it is a two conductor cable. This means that

space limitations can be overcome and the cable can be installed around any

obstructions.

Unlike other detection systems, the LHD system can define the location of the fire with

significantly high accuracy.

The LHD systems have higher life cycle which is approximately up to 30 years. This

results in a significant reduction of the maintenance cost.

Some of the disadvantages of the LHD systems are considered to be the following [4]:

In certain types of LHD systems, the cables are damaged after a fire and have to be

replaced. In long tunnels, this means replacing large lengths of cabling.

In large tunnel volumes that are ventilated, fire detection with LHD can become difficult

since less heat is developed at the ceiling height.

5.2.1.1 CHOSEN LHD SYSTEM

The LHD system that was chosen to be simulated for this research study is the FibroLaser III by

Siemens. Siemens has, already, installed more than 2,000 km of the FibroLaser cable and more

than 1,200 controllers. The FibroLaser provides effective fire detection under harsh environmental

conditions and for long or widespread systems. In has proven to offer fast detection of the fire and

accurate localization of the fire source. Added to this, the certification of the VdS according to Pr

EN 54-22, FM accreditation makes this system appropriate for a wide range of applications.

The working principle of the FibroLaser is that of a Fiber Optic LHD system as described above.

The system, basically, consists of a sensor cable that is installed along the tunnel and one or two

controllers that are centrally installed. The cable can reach a length of up to 10 km and is able to

provide consistent safety throughout the entire installation. The system can sense both

convective and radiative heat which means that it can detect a fire at its early stages and

Fig.5.1 Alarm thresholds for Linear Heat Detection (LHD) technology

74 eliminate false alarms. In addition, the system can be programmed to generate pre-alarms and

inform the control center before notifying the emergency services. Apart from the alarm, the

system provides important information that can help achieve a safer self-evacuation and

coordinate the interference of the fire department. The information that FibroLaser can provide

are:

The location of the fire

The extent of the fire

The direction of the propagation of the fire

With this information, the other systems of the tunnel such as the ventilation and smoke

extraction system, the guidance system and the fire extinguishing systems can be controlled and

coordinated depending on the fire development.

A sensor cable can be partitioned into up to 1,000 zones – either the one after the other or in

parallel - and each of these zones can be programmed with different alarm criteria. The alarm

thresholds are determined depending on the specific tunnel and the on-site conditions. There are

three options for the alarm:

Exceeding a defined absolute maximum temperature

Exceeding a defined deviation from the average temperature of a zone

Exceeding a defined maximum rise of temperature (rate of rise)

The output of the FibroLaser system is a profile (in figures and diagrams) containing information

regarding temperature peaks and the location and size of the fire. The time needed for the profile

to be generated is 8 to 15 sec.

FibroLaser system components

For the purposes of this research study, Siemens provided documentation regarding the

specifications and components of the FibroLaser III. Also, through an interview with Mr.Leo Knies

(fire specialist in Siemens), the basic types of sensor cables were explained together with their

main advantages and disadvantages.

One type of sensor cable is the one that can be

seen on the picture on the left. This type of cable

has an outer layer of infrared absorbing

insulation FRNC (Flame Retardant Non

Corrosive) and a subsequent layer of stainless

steel wires that attributes extended mechanical

properties to the cable. The optical fibers are

covered by a stainless steel tube in order to be

able to withstand a temperature of 400oC for 4

hours.

Another type of sensor cable is the one that can

be seen on the picture on the left. This type differs

from the previous one in terms of the material

used as a cover for the fibers. In that case, the

cover is a polyamide tube and between this tube

and the cable sheath is a layer of aramid fibers.

Another type of sensor cable is available that has Fig.5.3 FibroLaser sensor cable type (Type 2)

Fig.5.2 FibroLaser sensor cable type (Type 1)

75 neither a metal nor a plastic cover for the optic fibers. Comparing the aforementioned sensor

cable types, the following considerations can be derived:

The metal cover results in a slower response compared to the other two types

The plastic cover has an insulation effect on the fiber which has to be taken into account

in every application

The third type has the fastest response since the effect of electrical conduction does not

take place

Regarding cost considerations, the sensor cable costs between 5 and 10 euros per meter

depending on the type of cable and its durability is expected to be approximately 30 years with no

maintenance required. The components that is more costly is the controller.

The controller is the device equipped with the

laser transmitter and receiver. The controller is

centrally installed in the tunnel and gathers

information regarding the linear temperature

profile, the location of the fire and the direction

of the fire spread. It is suitable for wind speeds

up to 10 m/s. It is possible to achieve a system

configuration with high redundancy by using two

or three controllers. More specifically, the use of

an Optical Switch controller allows to attach a

sensor cable of double the previously allowable

length (2*10km = 20km).

configuration of FibroLaser system

Depending on the required redundancy of the FibroLaser system in each tunnel, one of the

following configurations can be selected.

One controller can be installed and connected to

one sensor cable that covers the whole tunnel

length. In this case, no redundancy is expected.

Two controllers can be connected to one sensor

cable which results in a fully redundant system.

A controller can be installed in the middle of the

tunnel length and two cables can be connected to

it. Each cable is installed along half the tunnel

length. In this case, the system is not redundant.

In case the previous system is equipped with two

extra controllers at the end of each of the cables,

the system acheives full redundancy.

Another configuration that can provide sensor cable

redundancy in installing the cable in a loop that runs

two times along the tunnel length. This setup has an

advantage in the case that maintenance of

Fig.5.4 FibroLaser controller unit

Fig.5.5 Alternative configurations for the FibroLaser controllers and sensor cable

76

replacement of the sensor cable is needed. In such a case, half of the tunnel can remain

closed and the other half can operate with safety.

5.2.1.2 MODELING OF THE LHD IN FDS

At this point, having discussed the basic considerations regarding the FibroLaser technology, it

was decided that the cable that will be simulated is the first type where the cover of the fibers is of

stainless steel. That was chosen in order to diminish the insulation effect caused by the plastic

cover and, thus, make the simulation input simpler. The third type could, also, be chosen but it

was rejected because it is more susceptible to damage from the harsh environment of a tunnel

and a more robust type was, thus, selected.

Within the FDS software, there is no default option for simulating the LHD technology as a

continuous cable. However, there is the option to use default devices like heat detectors by

defining certain properties. As a result, the FDS online forum was consulted in order to define the

best way to model the LHD. The software developers recommended that the LHD cable could be

modelled by defining many heat detectors in a row close to the tunnel ceiling and, thus, this way

was chosen for this research study.

In order to retrieve the information necessary to model the LHD, a supplier of this technology had

to be consulted. After contacting Siemens and specifically, Mr. Leo Knies, detailed information

regarding the specifications of the systems available in the market and full-scale tests conducted

were retrieved. As a result, decisions regarding the alarm thresholds and other specifications of

the LHD could be justified.

According to the FDS software manual, a heat detector is imported in the text file by the following

line:

&DEVC ID='LHD', PROP_ID='FibroLaser', XYZ=….. /

&PROP ID='FibroLaser', QUANTITY='LINK TEMPERATURE', RTI=30.0,

ACTIVATION_TEMPERATURE=58.0 /

The line QUANTITY='LINK TEMPERATURE' defines the heat detector. The RTI represents the

Response Time Index in units of √m*s. The ACTIVATION_TEMPERATURE=58.0 represents the

alarm threshold that is set for the heat detectors.

For the absolute temperature alarm threshold, a temperature of 58oC was chosen to be used in

order to match the alarm threshold used during the tests provided by Siemens. At this point, the

definition of the Response Time Index has to be defined. In the case of a heat detector, the

transfer of heat from the ceiling jet to the sensing element of the heat detectors does not happen

instantaneously but it lasts a certain period of time. The response Time Index (or thermal

response coefficient) is defined as the measure of the speed at which the heat transfer occurs

[96]. As stated in NFPA 72, the RTI of heat detectors should be listed together with its operating

temperature. As a result, for the purposes of this research study, the RTI of the FibroLaser III had

to be used and the value was 30 √m*s.

There is, already, available research regarding the Plunge tunnel method with which the RTI for

heat detectors is measured [97] as well as the most recent developments and observations on

predicting the RTI [98]. The interested reader could refer to the literature cited above for more

information regarding the RTI.

77

5.2.1.3 VALIDATION

In order to ensure that the LHD technology was properly modeled within the FDS software and, at

the same time, determine whether the results correspond to the reality, a validation study had to

be performed. The tests that were provided by Siemens were used as reference data for this

purpose. More specifically, the tests were reproduced by simulations and the results were

compared to the test measurements.

The tests performed by

Siemens were realized in the

Leidsche Rijn tunnel or else

known as A2 tunnel in Utrecht.

The tunnel has a length of 1650

meters which renders it the

second longest in the

Netherlands so far. It consists of

four tubes in total, two for each

traffic direction. The main tubes

consist of three lanes and the

secondary tubes consist of two

lanes each. The tunnel has a

traffic intensity of 188,300 motor

vehicles per day in the year

2013 [99].

description of the real tests

For the purposes of the fire test, 5 liters of ethanol were used in order to generate a fire that

would reach a HRR of 1.5 MW within 2 minutes. The ethanol was placed in a plate with

dimensions 1.5*1.5 meters. The LHD sensor cable was mounted near the ceiling and was

installed as a loop as shown in the following image.

The tests were performed for three different fire locations in which the fire was in the middle of the

tunnel cross-section and the location varied in terms of the position in the longitudinal dimension.

Along the length of the sensor cable, five zones were defined and the alarm thresholds for each

of these zones were differently adjusted as follows:

Zone 1: over the entire length of the sensor cable (51 - 3374 m) an alarm threshold of a

maximum of 58oC was set.

Fig.5.6 A2 tunnel in Utrecht

Fig.5.7 FibroLaser system configuration in the A2 tunnel tests by Siemens

78

Zone 2: over the first half of the sensor cable (51 - 1712 m) an alarm threshold of a rate

of rise of 5oC per 60 sec was set.

Zone 3: over the second half of the sensor cable (1712 - 3374 m) an alarm threshold of a

rate of rise of 10oC per 60 sec was set.

Zone 4: over the first half of the sensor cable (51 - 1712 m) an alarm threshold of 6oC

from the average zone temperature was set.

Zone 5: over the second half of the sensor cable (1712 - 3374 m) an alarm threshold of

10oC from the average zone temperature was set.

As can be derived from the report by Siemens, the fire that resulted in the highest temperature in

the tunnel is Fire 1 (the temperature reached approximately 36oC). This is the fire that was

simulated with the FDS software as a validation study. In this case, the fire is located in the point

1,163 meters of the sensor cable.

The airflow speed in the vertical and horizontal center of the tunnel was measured before each

set of measurements using an anemometer. During the test with Fire 1, the average airflow

speed in the tunnel was measured to be 0.57 m/s with the wind direction being from south to

north.

All the aforementioned data were imported in the FDS software in order to perform the validation

study. The first 2 minutes of the fire tests were simulated and a grid size of 0.5*0.5*0.5 meters

was used in order to achieve higher accuracy. The fuel used in FDS was ethanol and the

respective heat of combustion is 26.78 kJ/g, the soot yield is 0.008 and the CO yield is 0.001.

fire size

In the report of the fire tests in the A2 tunnel, it is mentioned that the fire reaches a HRR of 1.5

MW within the first 2 minutes of the fire development. After defining the necessary input in FDS

using ethanol as a fuel in a crib with dimensions 1.5*1.5 meters, the fire development graph that

was created is the following.

As it can be seen in the graph, the maximum HRR is achieved after 115 sec from the start of the

fire and is 0.88 MW which is lower than the 1.5 MW mentioned in the report. In order to determine

0.88 MW

0.00

0.10

0.20

0.30

0.40

0.50

0.60

0.70

0.80

0.90

1.00

HR

R (

MW

)

time (sec)

Fig.5.8 HRR-time graph for the simulation of the tests by Siemens

79 the reason behind this observation two sources were consulted and are described next.

First, an excel document was used that was published by the United States Nuclear Regulatory

Commission and, more specifically, by the Fire Protection Engineering and Special Projects

Section. This document is used to estimate the burning characteristics of liquid pool fires, the

HRR, burning duration and flame height. Using data from the SFPE Handbook of Fire Protection

Engineering regarding the properties of the fuels and the necessary formulas, it can estimate the

maximum HRR produced for different amounts of ethanol fuel for various plate sizes. So, it is

estimated that 5 liters of ethanol in a plate of 1.5*1.5 meters can produce a fire with a maximum

HRR of around 0.896 MW in about 2 minutes 25 seconds. This estimation is in accordance with

the simulation results. The calculation sheet for the ethanol fire size can be found in the Appendix

C.2.

Second, a report by Efectis Nederland BV was consulted that describes a number of fire tests

performed with ethanol in order to determine the HRR produced by an ethanol fire. Four tests

were performed in which the ethanol was contained in plates of different sizes (60*60 cm and

84*60 cm). The time interval that was examined starts at 200 seconds after ignition and lasts until

the fire begins to decay (this was between 500 seconds and 900 seconds). The HRR was

averaged over this time interval. A summary of the results of the Efectis ethanol tests can be

found in Appendix C.1.

In order to ensure that the results of the simulation were correct, Siemens was contacted and

informed about the difference in the HRR. It was, also, confirmed by Siemens that the size of the

fire was overestimated in the report and that they realized that the peak HRR was, indeed, lower.

It was, finally, agreed that the correct HRR corresponding to the large scale tests performed in A2

Tunnel was approximately 0.9 MW as given by the simulations and by the other sources

mentioned above.

The above issue served as a study which indicated that the FDS software is indeed a reliable tool

for simulating the LHD technology.

simulation of the validation case

In addition, in order to simulate the sensor cable, heat detectors had to be defined in FDS along

two times the length of the tunnel. This would result in a very large number of heat detectors that

would render the simulation very expensive in terms of computational time. As a result, the heat

detectors in FDS were chosen to be placed in the most important positions in the tunnel where

the temperature peaks were observed. The positions are all relative to the sensor cable length.

More specifically, heat detectors were placed in the following areas:

from 51 meters to 257 meters

from 989 meters to 1,405 meters

from 3,165 meters to 3,373 meters

The fire tests performed by Siemens generated a temperature profile along the sensor cable at

certain points in time. The same output was taken from the FDS software in order to compare the

data. The time periods for which the output is presented are the following:

at 1 min 10 sec from the ignition of the test fire



80


These specific points in time were chosen so that the observations could be made both at the

earlier stages of the fire and, also, at the stages where the temperature reaches its maximum

value. At the aforementioned points in time, the temperature peaks observed during the tests

were considerably high which makes these time periods more interesting to examine for the

purposes of the validation study.

The output of the FDS software has the form of numbers in an Excel file and, thus, graphs were

created in order to represent the temperature distribution across the sensor cable length. The

temperature distribution is almost uniform with peaks in certain positions (where the fire is

located). In the following graphs, the temperature profile of the tests by Siemens is compared to

the temperature profile generated by the simulations in FDS.

Fig.5.9 Temperature profile graph for comparison between simulation results and Siemens test result at t=1min25sec

81



82

A general observation that can be derived from the above graphs is that the temperature profile of

the simulation of the LHD correlates particularly well with the ones of the fire tests. A high level of

accuracy is achieved both in terms of the temperature profile in time and in terms of spatial

temperature distribution. The temperature peaks are achieved in the same points along the

sensor cable and, more specifically, in the points around the fire location (1,163 meters). The

maximum peak temperature difference observed is 2.8oC. This difference between the tests and

the simulation can be justified by external factors such as the specific material of the tunnel for

which there was no information or even the heat resistance of the sensor cable in practice. Along

the rest of the sensor cable length, were the distribution is more uniform and ranges around the

value of the ambient temperature, the prediction of the software is, also, accurate.

5.2.1.4 CONCLUSIONS

For the purposes of this research study, the accuracy with which the FDS software predicts the

performance of the LHD technology can be considered significantly high. The parameters that are

of interest for the study of fire detection in this case are the temperature peaks and the rate of rise

of the temperature. Both these factors are reproduced accurately since the trend of the

temperature distribution during the tests closely follows the one of the simulations.

Overall, the FDS software can be considered a proper tool for simulating the LHD technology.

Through this validation study it was proven that the way that the LHD was simulated represents

the reality and generates accurate output. In addition, it was determined that the input provided to

the FDS software was correct and representative of the reality. With these conclusions in mind,

the results of the simulations of the LHD technology in the context of this research study can be

considered accurate and trustworthy.


83

5.2.2 MULTIPLE GAS DETECTION (MGD)

Another state-of-the-art technology that is, currently, introduced into the market and is used in

tunnel applications is the Multiple Gas Detection (MGD). The MGD systems are generally

considered promising in terms of fire detection performance and as such were pointed out by fire

safety specialists. The interest to MGD technology has increased after attending the 3rd

Annual

Tunnels Fire Safety Forum 2015 where Firefly ab, a supplier of preventing fire protection systems

in Sweden presented the potentials of this technology.

In the context of this thesis, MGD technology was investigated in order to determine whether

there is enough evidence and material available that could be used to simulate and validate this

specific technology.

The MGD technology was initially developed for aerospace applications and its working principle

is based on detecting fire related gases. As was already mentioned, temperature detection is

effective to detect fires in tunnel applications. However, the earliest stage of the fire

developments is the release of gases. As a matter of fact, an open fire can result in the

immediate production of high levels of CO2 that are many times more than the CO2 produced by

the car engines. As was indicated by fire tests, the warm and highly concentrated CO2 cloud can

spread much faster than the smoke. In addition, according to a research by O. Linden and H.

Holemann, fire tests have proved that CO2 is the best fire indicator at this point.

Compared to smoke, gases are released in an earlier stage of fire development and are more

volatile. A multiple gas detection sensor is, alternatively, described as an “electronic nose”.

Chemical sensors are used for analyzing volatile compounds and work based on transforming

chemical interactions into electrical signals. It is possible to perform continuous measurements

and the use of chemical sensors is usually inexpensive. When chemical sensors are used in an

array for odor recognition and analysis, then the term “electronic nose” can be used [100]. An

interested reader could refer to the review by David et al., 2004 in order to be further informed on

the use and applications of the “electronic nose”. Moreover, the most recent and thorough review

on the developments regarding the “electronic nose” field is the “Handbook of Machine Olfaction”.

It was published in 2003 and elaborates on a wide range of aspects from sensor technologies to

signal processing and pattern recognition.

Fig.5.13 Principle of the electronic nose on which the MGD technology is based

84 Similarly, thus, to an electronic nose, with the use of sensors and through intelligent analysis, the

MGD can identify various types and sizes of fires (for example, even burning cables). As a result,

MGD can serve various purposes in order to guarantee a safe tunnel during both normal

operation and in case of a fire emergency:

Early detection of fire

Detection of leakages of fuels, solvents etc.

Monitoring of the in-tunnel air quality

Gas pattern recognition

Gas pattern suppression

The working principle of MGD technology is based on identifying the composition of gases

produced by a fire. In order to achieve that, six different gas sensors are used and measure,

compare and identify the gases and gas patterns. Then, through pattern classification, each gas

pattern is registered as a unique fingerprint. The recognition of different fingerprints can be used

for suppressing the gas patterns that are not fire related but still affect the in-tunnel air quality.

A MGD system available in the market is the Sentio Multi-Gas detector by Firefly ab. This system

was also presented in the 3rd

Tunnel Safety Forum in Amsterdam. The basic principles of gas

detection of this specific system are as described above. The brochure for this system can be

found in the Appendix B.4. However, the specifications of the Sentio system could not be found.

Unfortunately, within the context of this research study, specific information regarding the

performance and the pattern recognition process of MGD could not be retrieved. Relevant data

were researched for in literature and in the market. On the one hand, since MGD is a new and

developing technology, no relevant literature could be found. On the other hand, for the same

reason, the suppliers of MGD were reluctant to share any information regarding the types of

gases to be detected, the patterns used to detect fires and the detection times achieved until

now. Moreover, any attempt to retrieve validation cases with data and measurements form full-

scale fire tests performed with this technology, also, proved to be fruitless. As a result, one can

only but remain skeptical on the reason this information is by no means shared for research

purposes.

5.2.3 CCTV DETECTION Another fire detection technology that was chosen to be investigated is the CCTV video image

detection. This is a relatively new technology for fire detection and it uses real-time video images.

The working principle for fire detection is based on detecting and analyzing changes in contrast,

brightness, motion and loss of detail [4]. Originally, CCTV systems were designed to record video

signals and transform them into images that could be processed by the human eye. As a result,

these systems were relying on the response of the person supervising the images. In order to

eliminate the possibility of a mistake due to the human factor, it has been recently attempted to

develop automatic detection systems for flame detection [28].

flame detection

This is possible by a technique that combines CCTV with UV and IR radiation sensors for

detecting the fire and a camera that identifies regions with radiation and recognizes the fire.

Another technique that can be used for automatic flame detection is the Machine Vision Fire

Detection Systems (MVFDS). This system consists of video cameras, computers and systems

using artificial intelligence. The camera used in MVFDS is monitoring the environment and the

85 output is recorded in the computer as a function of space and time. Then, image processing and

pattern recognition methods are employed to analyze the images. There are three common

techniques for processing the image (histogram, temporal and rule based techniques) for fire

detection and an interested reader can refer to Wieser and Brupbacher for more information [28].

The CCTV flame detection systems perform well in detecting fires in tunnels since tunnel fires

involve flammable and combustible liquids as fuels. However, testing of these systems has

demonstrated that they are susceptible to nuisance alarms. In the continuously changing

environment of a tunnel, the natural or artificial lighting and the reflection from the metal surfaces

of the cars can be sources of false alarms. According to developers and suppliers of CCTV flame

detection systems (FLIR and Siemens) the most recent developments include systems that can

make a distinction between stable flames and unstable flames. For example, the flame of a lighter

could cause a false alarm. However, in contrast to a fire, the lighter produces a stable flame and,

by recognizing that, the system can prevent a false alarm. In addition, according to a fire

specialist in Siemens (Mr. Leo Knies), a time limit of usually 30 seconds is set. Then, if there is an

unstable flame for a time that exceeds this threshold, the alarm is activated.

smoke detection

As artificial intelligence progresses, digital imaging techniques can be used not only for flame

detection but also for smoke detection. An example of such a system is one developed by

Siemens that uses an algorithm that analyzes contrast in order to detect smoke. By using this

algorithm, problems related to moving vehicles and lights in tunnels can be overcome [28].

The main advantage of the CCTV video smoke detection technology is that the cameras are

already installed in tunnels for other purposes including traffic monitoring and security. Moreover,

in those systems, real time images are used to detect the fire and, thus, the tunnel operators and

emergency responders can have more direct information regarding the fire location and size as

well as the conditions in the tunnel. This way their response and intervention can be more

efficiently organized. In addition, the CCTV video smoke detection systems can identify fires in

moving vehicles [4].

In the environment of a tunnel, the exhaust fumes of the vehicles and the high pollution levels can

cause false alarms in the case of CCTV video smoke detection. In order to avoid nuisance

alarms, multiple fire detection systems need to be installed in the tunnel and confirmations have

to be taken before any system activation. The role of the tunnel operator is crucial to process the

image and identify a real emergency.

After consulting fire specialists and CCTV fire detection systems suppliers, it was concluded that

although these systems are becoming more popular, more research and efforts are needed to

avoid nuisance alarms. After an interview with a fire specialist in FireSense, Mr. Paul Wendt, it

was determined that in practice those systems cannot perform efficiently in a tunnel environment.

This is because the detection principle is based on image processing and pattern recognition and

it has proven that in a tunnel the image is constantly changing because of moving vehicles,

exhaust fumes and pollution. As a result, those systems have not yet achieved the desired levels

of reliability for tunnel applications.

In addition, similarly to the MGD system, information necessary for modeling and validating the

CCTV detection technology could not be retrieved. Fire detection systems suppliers and fire

safety specialists were mostly reluctant to suggest the use of these systems for tunnel

applications based on the existing experience.

86

5.2.4 CONCLUSIONS

In general, heat was determined to be the most appropriate fire phenomenon for fire detection in

tunnels. Smoke can be detected by optical systems but they are proven to be unsuitable for the

harsh environment of a tunnel. Optical devices like optical smoke detectors or flame detectors

are, highly, affected by the ambient conditions. As a result, dirt is accumulated on the equipment

and electronic circuits are easily destroyed by corrosion. All the above can result in a frequent

activation of false alarms.

For all the aforementioned reasons, the fire detection system that was chosen to be investigated

was the LHD system. The other two systems described in the previous section were not further

investigated because of the lack of relevant information.

87

CHAPTER 6: Fire Simulations In this chapter, the fire simulation scenarios are going to be described one by one and the input of each simulation is going to be discussed. The basic choices on each simulation scenario will be justified and commented. Finally, the output of the simulations is, also, going to be presented and shortly analyzed. The most important results are going to be shown by means of visual output and tables.

6.1 INTRODUCTION

Within the context of this thesis, a number of different scenarios were examined and simulated in

order to determine the parameters of fire detection and evacuation in each one of them. These

scenarios were chosen according to two different criteria:

First, the scenarios which were mentioned as literature gaps in previous research studies

were examined. According to the literature review, thus, the scenarios that needed further

investigation were chosen to be simulated.

Second, the scenarios that were of interest both scientifically but, at the same time,

commercially were determined. More specifically, after consulting companies and

individuals involved in tunnel design, the aspects that are of interest and that are dealt

with in every tunnel design were identified.

As was mentioned in Chapter 3, all the variations were performed for two tunnel types: the cut-

and-cover tunnel with a rectangular cross-section and the bored tunnel with a horseshoe shaped

cross-section. In addition, all the sets of variations were performed for two fire sizes: a typical

passenger car fire and a HGV fire.

6.2 GENERAL INPUT

In general, the input varies for each simulation scenario and will be discussed separately for each

simulation. Nevertheless, some input parameters are in common for all the simulations and these

are going to be presented in this section. These parameters are the following:

Simulation time duration

Chemical reaction for fire (chemical formula of fuel, CO yield and soot yield)

Linear heat detection parameters (as described in Chapter 5)

Boundary conditions at the portals

Fire location

In the sections that follow, these input parameters are going to be described in detail in order to

understand and justify the relevant choices.

88

6.2.1 SIMULATION TIME DURATION

As was explained in Chapter 5, the computational time required in order to perform more complex

simulations is often very high. Especially in the case that a finer grid is chosen, one simulation

can last for many days or even weeks. For this research study, the computational domain where

the grid should be applied is a tunnel of a length of 1,600 meters and the resolution of the grid

should be fine in order to deliver accurate results. For that reason, a way had to be found in order

to ensure the required accuracy of the results while, at the same time, keeping the computational

time low enough in order to be feasible to carry out a significant number of scenarios within a

research study of 9 months.

In order to achieve the above, the objective of the research study had to be further considered.

Since the focus of the research study is on fire detection and the time needed to activate an

alarm, the most important time period would be the time between the ignition of the fire and the

first minutes of the fire development. More specifically, as was suggested by A.Beard and

R.Carvel, the general rule is that fire detection should be such that the smoke control system

should be operational within a maximum of 2 minutes from the start of flaming combustion. This

means that detection should take place within this time. As a result, for each scenario, the first 3

minutes from the ignition of the fire are going to be simulated.

This means that the input file for all the simulations will have the following command line:

&TIME T_END=180.0/

This line in the input file will instruct the FDS software to run the simulation with a timeline starting

at zero and ending at 180 seconds.

Fig.6.1 HRR-time graphs for a HGV fire and a passenger car fire respectively (based on test data)

89

6.2.2 CHEMICAL REACTION FOR FIRE

In order to import the fuel type in the fire simulations, the best defined fuel for which more

information is available had to be defined. In general, a combustion model has to consider

predictable and controllable fires which have been thoroughly investigated. These fires are mainly

cellulose (wood) fires, propane (gas) fires and hydrocarbon (cars, HGV) fires. Within the FDS

software, modeling a hydrocarbon fire requires more computational time and, at the same time, it

not very well-defined. For that reason, it was determined that a propane fire would be the best

option since it is well-defined and investigated and it is represented accurately within FDS.

The chemical formula for propane is C3H8. Within the FDS software, apart from defining the fuel,

the CO yield and the soot yield need to be specified. Consulting the NCHRP Synthesis, the

values for the CO and soot yield were imported as yco=0.043 g/g and ys=0.1 g/g. Nevertheless, in

the FDS software since propane is one of the default fuels and the properties can, also, be

automatically filled in. The default fuels can be found in a Table 11.1 within the FDS User Guide.

6.2.3 BOUNDARY CONDITIONS AT THE PORTALS

Defining the boundary conditions for an FDS simulation is a challenge. More specifically, in case

the location of the tunnel and, thus, the meteorological conditions are unknown, a sensitivity

analysis is needed in order to define the impact of each input on the results. For that reason, a

specific tunnel was chosen as a case study (Hubertus tunnel) as was already described in

Chapter 3. For this tunnel, real meteorological measurements had to be retrieved.

Most of the tunnels are equipped with sensors that measure the concentrations of the pollutants

at the portals. Measurements regarding the meteorological conditions are, also, recorded. In the

case of the Hubertus tunnel, measurements of the air speeds at the portals were carried out and

were, kindly, shared by Gemeente Den Haag. These measurements were carried out for two

periods:

Between 03/03/2010 and 18/03/2010

Between 24/03/2010 and 01/04/2010

The air speeds were measured at both tubes of the Hubertus tunnel at the south and north

portals. As a result, four locations of measurements were included in the data provided by

Gemeente Den Haag. The average values estimated for the total period of the measurements

were 2.4 m/s for the south portal and 2.8 m/s for the north portal.

In order to translate the wind speed into pressure, the following equation was used:

P=1/2*v2*ρ

where v, is the air speed in m/s

ρ, is the air density which is 1.2 kg/m3

As a result, at the south portal: Psouth=1/2*2.42*1.2=3.46 Pa

and at the north portal: Psouth=1/2*2.82*1.2=4.7 Pa

In order to describe a passive opening to the outside in FDS, an OPEN vent is applied to the

exterior boundary of the computational domain as suggested in the FDS User Guide. In the case

90 of the tunnel model, an OPEN vent is applied to each of the portals.

As it is further specified by the software developers, when modeling a tunnel in FDS, it is

recommended to determine a pressure rather than a velocity at a boundary. As a result, the

pressure values calculated above were imported in the FDS for each of the portals as:

&VENT SURF_ID='OPEN', XB=…. , DYNAMIC_PRESSURE=3.46/ SouthPortal

&VENT SURF_ID='OPEN', XB=….., DYNAMIC_PRESSURE=4.7/ NorthPortal

By defining the aforementioned pressure at the tunnel portals, the OPEN properties required as

input were determined.

alternatives considered

A research was carried out regarding how to implement the wind in the simulation model, before

deciding that the most appropriate solution would be to retrieve actual measurements. By

consulting a wind specialist within Deerns and by researching in relevant literature, it was

concluded that, in order to define the wind pressure in a specific location, the wind pressure

coefficient (Cp) had to be determined. This parameter is implemented in the wind pressure

equation:

P=1/2*v2*ρ*Cp

The Cp parameter expresses the pattern of wind flow around a structure and it is independent of

the wind speed. It varies according to the wind direction and the position on the structure surface

and it is, significantly, affected by the neighboring built environment.

Although there are many possible ways to predict the Cp parameter, their applicability is limited

because of the fact that the impact of the neighboring structures strongly affects the value. One

way of estimating the Cp is to perform full-scale measurements on the structure. However, this

way is complex and expensive and, thus, it is mainly used for validation purposes. Usually, the Cp

data are obtained through wind tunnel measurements using solid models of the structure and its

surroundings. Although this way is the most reliable one, it includes a very high cost for

constructing the models and for using the wind tunnel. Lately, CFD modeling was considered in

order to determine Cp. Nevertheless, the CFD models that are needed would require a large

computational domain and would be costly by means of time and resources. Moreover, the need

for validation makes CFD models harder to be applied for this purpose. Certain databases

already exist (AIVC, ASHRAE) where surface averaged Cp data can be found. These databases,

though, apply only to specific surfaces and do not take into consideration the adjacent structures.

Another alternative to define Cp is the use of analytical models including sets of equations. This

alternative is, also, considered unreliable due to the lack of data for certain parameters. Finally,

Cp generating software was developed to predict point values of the Cp on facades. However, as

was pointed out by a wind specialist in Deerns, this type of software was developed for buildings

and it is not certain if their application in tunnels (which are lower in height) would be appropriate.

After considering all the alternatives described above, it was concluded that the most accurate

way to implement the wind into the simulation model would be to research for data from real

measurements. The other alternatives would not be feasible within the context of a MSc thesis in

terms of time and resources.

91

6.2.4 FIRE LOCATION

Another parameter that had to be determined before setting up the simulation variations was the

fire location. A study was, thus, performed in order to specify the worst case scenario regarding

the fire location. During this study, the main criteria that were considered were:

The peak HRR of the fire

The maximum temperature developed

The backlayering effect

The stratification of the smoke layer

The number of potential vehicles trapped upstream of the fire

Three fire locations were studied, where the fire was placed in one of the three sections of the

tunnel respectively. In the first case, the fire was placed at the straight section of the tunnel that

has no longitudinal slope (location 1). In the second case, the fire was placed at the start of the

uphill section of the tunnel (location 2). In the third case, the fire was placed at the end of the

downhill section of the tunnel (location 3).

Each of these scenarios for the fire location was simulated for a passenger car fire and with a

relatively coarse grid (1*1*1 meters). The results can be observed in the following sketches:

location 1

In the case where the fire is located at the middle of the straight section of the tunnel, the peak

HRR was 1,000 kW and the maximum temperature was 73oC. The length of the tunnel

downstream the fire from where the vehicles can easily leave the tunnel is the same as the length

upstream the fire where the vehicles remain trapped. The backlayering effect is not intense and

the smoke layer remains stratified.

Fig.6.2 Alternative fire locations studied

92

location 2

In the case where the fire is located near the uphill section of the tunnel, the peak HRR was 800

kW and the maximum temperature was 75oC. The length of the tunnel downstream the fire from

where the vehicles can easily leave the tunnel is less than the length upstream the fire where the

vehicles remain trapped. The backlayering effect is not intense but the stratification of the smoke

layer is slightly disturbed.

location 3

In the case where the fire is located near the downhill section of the tunnel, the HRR was 1000

kW and the maximum temperature was 79oC. The length of the tunnel downstream the fire from

where the vehicles can easily leave the tunnel is more than the length upstream the fire where

the vehicles remain trapped. The backlayering effect is not intense but the stratification of the

smoke layer is significantly disturbed.

As a result, location 3 of the fire was determined as the worst case scenario. The impact of the

downhill slope on the smoke propagation is significant and the chimney effect disturbs the

stratification of the smoke layer. Moreover, the conditions inside the tunnel in this scenario result

in a more dangerous environment during self-evacuation.

6.2.5 FIRE SIZE

In FDS, the fire has to be defined in terms of peak HRR and fire area. More specifically, a solid

obstruction is defined and the top surface of the obstruction represents the fire. In the following,

the reasoning behind the choice of fire scenarios and the way these were imported in FDS is

going to be explained in detail.

fire size

The design fire size is one of the most significant parameters involved in the tunnel fire safety

design [4]. In the context of this research study, in order to determine the response time of the

detection system, a fire size had to be decided upon. In case of a road tunnel fire, the materials

that burn are, mainly, elements of the vehicles like, for example, tires, seats, materials of the

finishing (plastic), cargo or even fuel from vehicle tanks. Depending on the type of burning load,

fires can differ in terms of the total energy output and of the fire development.

The size of a fire is expressed as the peak fire Heat Release Rate (HRR) and is measured in

megawatts (MW) or MBtu/hr. A number of guidelines that resulted from the observations and

measurements of large-scale fire tests have suggested design values for the HRR according to

the type of vehicle [88] [101] [102] . Based on experimental data retrieved from fire tests, NFPA

502 suggests the design values for the HRR that are shown in the following table:

93

The experimental HRR represents the results of the measurements during the tests, whereas the

representative HRR is the design values suggested for each vehicle type.

Since time is a significant factor in case of tunnel fires, engineers combine the HRR with the

growth rate of the fire. This way it is possible to predict the fire development in time and

determine the way this will affect the evacuation process [4]. The Runehamar tunnel full-scale fire

tests, for example, indicated that in less than 10 minutes a fire can exceed the HRR of 100 MW.

In this tests it was, also, observed that the peak HRR ranges between 1.5 MW and 202 MW for

the most common types of road vehicles. Respectively, the gas temperatures at ceiling height

range between 110oC and 1365

oC.

For the purposes of this research study, two different fire scenarios were chosen to be simulated

representing two fires that differ significantly in size. In both scenarios, the maximum HRR

recorded was chosen. The following fire scenarios were simulated:

Fire scenario 1: a passenger car fire with a peak HRR of 10 MW

Fire scenario 2: a HGV fire with a peak HRR of 200 MW

A bus fire with a peak HRR was also considered to be simulated. For that reason, the HRR-time

curve for a bus fire had to be retrieved. After a literature research, it was concluded that the most

known fire tests that included a bus were a fire test within the EUREKA 499 project and a bus test

in the Shimizu tunnel in Japan.

Table 6.1 HRR for typical vehicles according to experimental data

94

Within the context of the EUREKA EU499 [101] project, 21 fire tests were performed in a tunnel in

Repparfjord in Norway. These tests included a bus fire with a single-decker Volvo school-bus that

was 12 meters long [103]. The HRR that was measured during the bus fire test can be seen in

the following graph and is represented by the continuous solid line.

Observing the trend of the HRR for the bus fire in the graph, it was concluded that the fire

development in the first 3 minutes from ignition (the time duration that was chosen for the

simulations) is similar to that of the HGV fire. As a result, for the purposes of this research study,

it was considered that the fire scenarios of a passenger car fire and of a HGV are representative

of a small and a large fire respectively.

A tanker fire was not chosen to be simulated because the fire development is extremely fast and

it is characterized as an explosion [104]. The chances of survival, thus, are almost zero and the

fire detection system could not contribute to a safe evacuation. Because of the severity of a fire

incident with tankers, regulatory measures are, usually, taken by forbidding the use of certain

tunnels to vehicles caring extremely flammable load [105].

For the Fire scenario 1 the HRR-time curve had to be used in order to import the fire

development in the FDS software. The curve that was used for the car fire is the one represented

by a dark solid line in the following graph. The maximum HRR at the end of the first 3 minutes of

the fire development is 1 MW.

Fig.6.3 Approximate HRR data from EUREKA EU499 bus fire test (solid line)

95

As can be seen on the graph, the line follows a linear trend and this was used in order to input the

fire ramp in the FDS software.

For the Fire scenario 2 the HRR-time curve that was used was the one of the Runehamar tunnel

test T1 that corresponds to a peak HRR of 200 MW. So, the curve that was used for the HGV fire

is the one that is represented in the following graph. The maximum HRR within the first 3 minutes

of the fire development is approximately 22.5 MW.

The fire development for the HGV fire was imported in the FDS software by observing the trend of

the line in the previous graph.

Fig.6.4 HRR-time curve for passenger car fire

Fig.6.5 HRR-time curve for HGV fire

96

fire area

As was already mentioned in Chapter 4, the fire is imported in FDS as a surface and, thus, an

area has to be defined. In the simulation model, the fire was imported as a rectangular

obstruction and the top surface represents the fire. The input parameters required are the surface

area of the fire expressed in m2 and the HRR per Unit Area (HRRPUA) expressed in kW/m

2.

The defined fire area is different for the two fire scenarios, the one of the passenger car and the

one for the HGV:

Fire scenario 1: In order to determine the fire area of a passenger can fire, the size of

the typical vehicle had to be defined. A typical passenger car, usually, has a width of

around 1.80 meters, a length of 4.75 meters and a height of 1.5 meters (Appendix A.2).

As a result, the total fire area is approximately 8.50 m2. An approximation of these

dimensions was chosen and, thus, an obstruction with a total area of 9 m2 (3*3 meters)

and a height of 1 meter was created. The top surface of this obstruction was defined as

the fire surface. The HRRPUA in this case is the maximum HRR (1 MW or 1,000 kW)

divided by the fire area (9 m2), which means it is approximately 115 kW/ m

2.

Fire scenario 2: The fire area in case of a HGV was determined in a different way. The

dimensions of a HGV are much larger which means that the fire does not cover the whole

area of the vehicle at the first stages of its development. As a result, similarly to the

validation case of Runehemar tunnel described in Chapter 4, an obstruction with a total

area of 12 m2 (3*4 meters) and a height of 1 meter was created. The HRRPUA in this

case is the maximum HRR (22.5 MW or 22,500 kW) divided by the fire area (12 m2),

which means it is approximately 1,885 kW/ m2.


The ventilation system to be used for the research study was determined after conducting a

literature research and consulting ventilation system suppliers for tunnel applications. The

literature research helped determine the ventilation systems that are most commonly used in road

tunnels as they were, already, described in Chapter 3. The longitudinal ventilation system was

determined as the most commonly used in tunnels. Moreover, the ventilation system supplier

NOVENCO suggested the use of jet fans in tunnels and provided brochures and specifications of

tunnel ventilation systems.

Resulting from the above, the ventilation system of the tunnel was modeled in the FDS software.

As was recommended by a fire safety specialist in Deerns, jet fans of high power should be

installed every 200 meters along the length of the tunnel. According to the brochures by

NOVENCO, two types of tunnel fans are manufactured: round type (AUR/ ARR) and rectangular

type for space saving (AUC/ARC). Both types can be designed as unidirectional or reversible.

The diameter of the rotors ranges between Ø630 to Ø1600 mm for round fans and from Ø500 to

Ø800 mm for the rectangular fans.

97

In the context of this research study, round jet fans of a diameter of 1,010 mm were chosen from

the NOVENCO catalogue. Moreover, the type of fans chosen is the AUR (unidirectional round

type) with a volume flow of 17.33 m3/s. The other characteristics and specifications of this jet fan

type are shown in the following picture from the NOVENCO brochure.

In the FDS software, the jet fans were modeled as rectangular blocks of 1,0*1,0*1,0 meters with

vents attached on them. In the following screenshot of the tunnel entrance, the way the jet fans

were modeled can be seen.

Fig.6.6 Jet fan sizes for unidirectional and reversible models from NOVENCO

Fig.6.7 Specifications for jet fan models by NOVENCO

98

6.2.7 GEOMETRICAL MODEL

A geometrical model of a typical and representative tunnel had to be created and imported in

FDS. The model was created according to the typical cross sections mentioned in Chapter 2. The

tunnel length was 1,600 meters for both tunnel typologies. The software offers the option of

importing CAD files and of creating the geometrical model in PyroSim, the visualization platform

for FDS. In the context of this research study, two geometrical models (the cut-and-cover and the

bored tunnel geometry) were created by a fire specialist in Deerns and were imported in FDS.

Additionally, the option of creating the models as CAD files and importing them in FDS was

explored. The Rhinoceros software was used and two models were created one for each tunnel

type.

It was observed that the rectangular geometry is more accurately represented after importing it in

the FDS software. The horseshoe tunnel geometry was altered when converted into blocks and

the model was uneven and at some points disconnected. This could be observed in SmokeView

before running the simulation setting as simulation time T_END=0.1 seconds. This way the

Fig.6.8 Configuration of jet fans in the simulation model

Fig.6.9 Geometrical models of cut-and-cover and bored tunnel typologies

99 geometrical model appeared in SmokeView and could be checked without running the simulation.

As a result, the geometry of the horseshoe tunnel was not imported from a Rhino model but was

created as a tube with a slab running along the tunnel length.

6.2.8 MESH

As was already mentioned in previous sections, the options for the mesh to be used were a

uniform coarse mesh, a uniform fine mesh or multiple meshes that would be finer in the area

affected by the fire phenomena and coarser in the area near the entrance and exit of the tunnel.

After the sensitivity study mentioned in Chapter 4, the option of multiple meshes was chosen.

More specifically, a cell size of 1*1*1 meters was used for the first 568.0 meters counting from the

entrance of the tunnel and the last 511.0 meters from the tunnel exit. For the middle part of the

tunnel (418.0 meters) where the fire is present and the fire phenomena are more intense, a finer

mesh with a grid cell size of 0.5*0.5*0.5 meters was used.

According to the FDS software User Guide, special attention should be given to the boundaries

between the different meshes. The alignment of the meshes was performed according to the

directions given in the FDS User Guide and it is illustrated in the following sketch.

Fig.6.10 Part of the tunnel where the finer grid is applied

Fig.6.11 Alignment of the meshes with different cell sizes in the model

100

6.3 FIRE SIMULATIONS

At this point, a number of variations for the fire simulations are going to be performed. The two

basic groups of simulations are the ones for a passenger car fire and the ones for a HGV fire. For

those basic groups, a series of simulations are performed for the two chosen tunnel typologies:

the cut-and-cover tunnel and the bored tunnel.

For each of these tunnel typologies, simulations for certain variations are performed regarding

different geometry of cross-section, different ventilation types and configurations and, finally, with

the presence of vehicles in the tunnel. All the variations performed are summarized in a table in

Appendix E.

6.3.1 PASSENGER CAR FIRE

The first basic group of simulations was the one performed for a passenger car fire. In this case,

the fire development is slower and the size of the fire is smaller. In the case of a passenger car

fire, the time needed to detect the fire is usually larger because the temperature on the ceiling

level needs more time to reach the absolute alarm threshold temperature.

6.3.1.1 CUT-AND-COVER TUNNEL

basic scenario

In the basic scenario, the input provided to FDS is the general input described in section 6.2. For

the basic scenario, the peak HRR was 1.07 MW and was reached in the first 175 seconds of the

fire development. The maximum recorded temperature in the whole tunnel was around 83oC.

Backlayering occurred for the first 110 seconds but the backlayering length was considerably

small. The ventilation system led the smoke upstream of the fire in a stratified layer.

T=90 seconds

T=120 seconds

Regarding the LHD system, the maximum detected temperature was approximately 43oC and it

was recorded in point 607.0 meters of the FibroLaser length at a distance of 2.0 meters

downstream the fire location. As a result, the alarm threshold for the absolute temperature of

50oC was not exceeded and the relevant alarm was not activated. On the other hand, the rate-of-

rise threshold was exceeded and an alarm was activated at 114 seconds.


101

The point of detection based on the rate-of-rise threshold for the basic scenario of the cut-and-

cover tunnel can be seen in the images above. The green dot represents the detection point of

the FibroLaser. As can be seen, the detection point is very close to the fire location and, thus, the

information that the FibroLaser system provides for the location of the fire is accurate.

The temperature distribution in the tunnel can be seen in the images below in three different

moments of the fire development. The highest temperatures can be observed upstream the fire

and the closer to the end of the tunnel, the highest temperatures are observed closer to the

ceiling.

T=90 seconds

T=110 seconds

T=180 seconds

RATE-OF-RISE ALARM: 114 sec

DETECTION POINT: 2.0 meters

downstream the fire

Alarm activation point in FibroLaser cable

FibroLaser cable

Fig.6.13 Detection point and detection time for basic scenario of cut-and-cover tunnel (PyroSim software)

Fig.6.14 Temperature distribution (oC) for the basic scenario of the cut-and-cover tunnel typology

(Smokeview software)

102 In the following sketches, the visibility levels can be observed around the fire. The visibility levels

downstream and in the close proximity of the fire are significantly low whereas upstream of the

fire there is no important reduction of the visibility. Since the backlayering effect upstream of the

fire is small and the smoke is limited on the ceiling level, the people trapped can evacuate without

visibility problems. The visibility levels that can be observed upstream of the fire are

approximately 28-30 meters which according to the tenability limits, it is considered appropriate in

order for the people to be able to recognize a safety sign or emergency exit.

On the other hand, downstream of the fire the smoke is not stratified for the first 85 meters after

the fire location and the visibility levels are as low as 7-10 meters at the eye level. Nevertheless,

the visibility conditions downstream the fire do not significantly affect the evacuation process,

since it is considered that the vehicles have already moved towards the tunnel exit portal and did

not stay close to the fire incident.

T=90 seconds

T=110 seconds

T=180 seconds

Fig.6.15 Visibility conditions (soot visibility in m) for the basic scenario of the cut-and-cover tunnel typology (Smokeview Software)

103

geometry variations

The cross-sections of the tunnels are usually standard and depend on the tunnel typology and

construction method. However, at this point, the effect of a higher cross-section on the detection

time was examined. In this scenario, it was considered interesting to examine the height of the

smoke layer and whether a higher cross-section with a stratified smoke layer could allow for safer

evacuation. It should be noted, though, that constructing a tunnel with higher cross-section would

result in higher construction costs and the cost-benefit relation should be carefully examined.

In this scenario, the input provided to FDS is the general input described in section 6.2 but with a

different geometrical model. The smoke layer height in the basic scenario downstream of the fire

ranged between 1.50-2.0 meters. So, in order to provide to the evacuees a clear area of

adequate height where they could walk undisturbed by smoke, 2 meters of extra height were

added to the tunnel.

The peak HRR was 1.06 MW and was reached in the first 175 seconds of the fire development.

The maximum recorded temperature in the whole tunnel was around 76oC. Backlayering started

at the 90 seconds and reached the maximum length at 140 seconds before it started reducing

again. The maximum backlayering length reached approximately 17 meters. The smoke layer

upstream of the fire remains stratified while downstream of the fire the stratification is disturbed.

As a result, for the first 60 meters downstream of the fire there is a zone where the smoke

reaches the road surface. After this zone, the smoke layer remains stratified and allows for a

smoke free area of a height of 2.5 meters. The aforementioned observations can be seen in the

following sketches that were exported from the SmokeView software.

T=90 seconds

T=140 seconds


was recorded in point 604.0 meters of the FibroLaser length and right at the fire location. As a

result, the alarm threshold for the absolute temperature of 50oC was not exceeded and the

relevant alarm was not activated. On the other hand, the rate-of-rise threshold was exceeded and

an alarm was activated at 114 seconds.

Fig.6.16 Smoke propagation for the higher cross-section scenario of the cut-and-cover tunnel typology (Smokeview software)

104

The point of detection based on the rate-of-rise threshold for the higher tunnel cross-section can

be seen in the images above. The green dot represents the detection point of the FibroLaser. As

can be seen, the detection point is right at the fire location and, thus, the information that the

FibroLaser system provides for the location of the fire is accurate. In the case of the higher cross-

section the detection is more accurate than in the case of the basic scenario.




ceiling.

T=90 seconds

T=110 seconds

T=180 seconds


DETECTION POINT: at fire location


FibroLaser cable

Fig.6.17 Detection point and detection time for higher cross-section scenario of cut-and-cover tunnel (PyroSim software)

Fig.6.18 Temperature distribution (oC) for the higher cross-section scenario of the cut-and-cover

tunnel typology (Smokeview software)



fire there is no important reduction of the visibility. Since the backlayering effect upstream of the

fire is small and the smoke is limited on the ceiling level, the people trapped can evacuate without


approximately 28-30 meters which according to the tenability limits, it is considered appropriate in


On the other hand, downstream of the fire the smoke is not stratified for the first 60.0 meters after

the fire location and the visibility levels range between 10 and 15 meters at the eye level.

Nevertheless, the visibility conditions downstream the fire do not significantly affect the

evacuation process, since it is considered that the vehicles have already moved towards the

tunnel exit portal and did not stay close to the fire incident.

T=90 seconds

T=110 seconds

T=180 seconds

Fig.6.19 Visibility conditions (soot visibility in m) for the higher cross-section scenario of the cut-and-cover tunnel typology (Smokeview Software)

106

ventilation variations

In the basic scenario, a longitudinal ventilation system with jet fans was examined as described in

section 6.2.6. At this point, a longitudinal ventilation system with jet fans installed in closer

longitudinal distances was simulated. The jet fans were installed every 60 meters and were

chosen to be of lower power than in the case of the basic scenario. The jet fans chosen are form

the NOVENCO ventilation systems and the specific type is shown in the following figure:

In this scenario, the input provided to FDS is the general input described in section 6.2 but a

different configuration of the jet fans is used. The peak HRR was 1.18 MW and was reached in

the first 178 seconds of the fire development. The maximum recorded temperature in the whole

tunnel was around 82oC. Backlayering started at the 35 seconds and the smoke continued to

spread upstream of the fire until the end of the simulation at 180 seconds. The backlayering

length at this point reached approximately 65 meters. The stratification of the smoke layer is

disturbed both upstream and downstream of the fire. At approximately 25 meters downstream of

the fire the stratification tends to be restored but it is disturbed again right before the next set of

jet fans. As a result, the smoke layer covers the whole tunnel section both upstream and

downstream the fire. The aforementioned observations can be seen in the following sketches that

were exported from the SmokeView software.

T=35 seconds

T=180 seconds


was recorded in point 2.0 meters upstream the fire location. As a result, the alarm threshold for

the absolute temperature of 50oC was not reached and the relevant alarm was not activated. On

the other hand, the rate-of-rise threshold was exceeded and an alarm was activated at 68

seconds.

Fig.6.20 Smoke propagation for the alternative ventilation scenario of the cut-and-cover tunnel typology (Smokeview software)

107

The

point

of detection based on the rate-of-rise threshold can be seen in the images above. The green dot

represents the detection point of the FibroLaser. As can be seen, the detection point has moved

upstream of the fire compared to the previous scenarios. Nevertheless, the information that the

FibroLaser system provides for the location of the fire can still be considered accurate. The

different configuration and power of the ventilators has affected the distribution of the temperature

and the point of alarm activation has, thus, changed.


moments of the fire development. The highest temperatures are observed closer to the ceiling

both upstream and downstream the fire.

T=90 seconds

T=110 seconds

T=180 seconds



upstream the fire


FibroLaser cable

Fig.6.21 Detection point and detection time for alternative ventilation scenario of cut-and-cover tunnel (PyroSim software)

Fig.6.22 Temperature distribution (oC) for the alternative ventilation scenario of the cut-and-cover



both upstream and downstream of the fire are significantly low. More specifically, for the first 65

meters upstream the fire the visibility levels are around 15 meters. In addition, for the first 25

meters downstream the fire the visibility levels are around 20 meters at eye level.

Moreover, as it can be observed in

the sketch on the left, the jet fans

disturb the smoke layer reducing the

visibility under the ventilators. So,

even though the stratification of the

smoke layer tends to be restored, the

jet fans mix the smoke and reduce

the visibility levels at this point at

approximately 15 meters.

T=90 seconds

T=110 seconds

T=180 seconds

Fig.6.24 Visibility conditions (soot visibility in m) for the alternative ventilation scenario of the cut-and-cover tunnel typology (Smokeview Software)

Fig.6.23 Visibility conditions (soot visibility in m) around the ventilators (Smokeview Software)

109

traffic impact

As was already mentioned, previous research on fire detection systems [95] did not take into

account the presence of traffic obstructions in the tunnel and how these affect the response of the

fire detection systems. As a result, in this research study, a simulation scenario was examined

where traffic obstructions were added in the model. After consulting a fire specialist in Deerns and

supervisor of this research study, it was suggested that the obstructions should represent

passenger cars and trucks in rows in both lanes of the tunnel. This could be considered the worst

case scenario since more turbulence would be created between two consecutive vehicles of

different size and, more importantly, height.

In this scenario, the input provided to FDS is the general input described in section 6.2 and

rectangular obstructions in the dimensions of passenger cars and trucks were added with a

longitudinal distance of 4.0 meters the one from the next. The peak HRR was 1.16 MW and was

reached in the first 180 seconds of the fire development. The maximum recorded temperature in

the whole tunnel was around 81oC. Backlayering started at the 20 seconds and the smoke

continued to spread upstream of the fire until 175 seconds. Then, the smoke stopped spreading

upstream of the fire. The backlayering length at this point reached approximately 10 meters. The

stratification of the smoke layer is disturbed both upstream and downstream of the fire and smoke

is concentrated between the vehicles. As a result, the smoke layer covers the whole tunnel

section both upstream and downstream the fire. The aforementioned observations can be seen in

the following sketches that were exported from the SmokeView software.

T=20 seconds

T=170 seconds


was recorded in point 604.0 meters of the FibroLaser length and right at the fire location. As a

Fig.6.25 Modeling of traffic obstructions in the cut-and-cover tunnel typology (Smokeview software)

Fig.6.26 Smoke propagation in the presence of traffic obstructions in the cut-and-cover tunnel typology (Smokeview software)

110 result, the alarm threshold for the absolute temperature of 50

oC was reached and the relevant

alarm was activated at 175 seconds. Added to this, the rate-of-rise threshold was exceeded and

an alarm was activated at 65 seconds.

The point of detection based on the rate-of-rise and absolute temperature thresholds can be seen

in the images above. The green dot represents the detection point of the FibroLaser. As can be

seen, the detection point is right at the fire location and, thus, the information that the FibroLaser

system provides for the location of the fire is accurate. The presence of traffic obstructions does

not have an effect on detecting the location of the fire.

The temperature distribution in the tunnel can be seen in the following images in three different



ceiling.

T=90 seconds

T=110 seconds

T=180 seconds


ABSOLUTE TEMP. ALARM: 175 sec

DETECTION POINT: at fire location Alarm activation point in FibroLaser cable

FibroLaser cable

Fig.6.27 Detection point and detection time in the presence of traffic obstructions in the cut-and-cover tunnel (PyroSim software)

Fig.6.28 Temperature distribution (oC) in the presence of traffic obstructions in the cut-and-cover


111 In the following sketches, the visibility levels can be observed around the fire. Similarly to the

previous scenarios, the visibility levels downstream and in the close proximity of the fire are

significantly low whereas upstream of the fire there is no important reduction of the visibility.

Since the backlayering effect upstream of the fire is small (10 meters) and the smoke is limited on

the ceiling level, the people trapped can evacuate without visibility problems. The visibility levels

that can be observed upstream of the fire are approximately 28-30 meters which according to the

tenability limits, it is considered appropriate in order for the people to be able to recognize a

safety sign or emergency exit.

T=90 seconds

T=110 seconds

T=180 seconds

On the other hand, downstream of the fire the smoke is not stratified for approximately the first 85

meters after the fire location and the visibility levels are as low as 7-10 meters at the eye level. As

can be observed in the sketches below, the smoke gets trapped between the vehicles and loses

its stratification, thus, significantly reducing the visibility downstream of the fire. As a result, it can

be understood that the people in the proximity of the fire have to deal with extremely harsh

visibility conditions both because of the vehicles present in the tunnel and because of the trapped

smoke.

Fig.6.29 Visibility conditions (soot visibility in m) in the presence of traffic obstructions in the cut-and-cover tunnel typology (Smokeview Software)


112 Nevertheless, in a real fire scenario in a road tunnel, it is more likely that the cars downstream of

the fire will either move closer to the tunnel exit portal or completely leave the tunnel. As a result,

the visibility conditions downstream of the fire will be better than those presented by the

simulation scenario.

113

6.3.1.2 BORED TUNNEL

basic scenario




Backlayering started at the 80 seconds and the smoke continued to spread upstream of the fire

until the end of the simulation at 180 seconds. The backlayering length reached approximately 20

meters at 160 seconds and remained almost the same until the end of the simulation. The

stratification of the smoke layer is maintained upstream of the fire allowing for a smoke free area

of around 3.0 meters height. Downstream of the fire, the stratification of the smoke layer is

disturbed for the first 90 meters and the smoke covers the whole height of the tunnel section.

After this distance, the stratification is restored and there is a smoke free area of a height of

approximately 2.50 meters below the stratified smoke layer. The aforementioned observations

can be seen in the following sketches that were exported from the SmokeView software.

T=80 seconds

T=180 seconds


was recorded at a distance of 2.0 meters downstream the fire location. As a result, the alarm

threshold for the absolute temperature of 50oC was not exceeded and the relevant alarm was not

activated. On the other hand, the rate-of-rise threshold was exceeded and an alarm was activated

at 69 seconds.



downstream the fire Alarm activation point in FibroLaser cable

FibroLaser cable

Fig.6.31 Smoke propagation for the basic scenario of the bored tunnel typology (Smokeview software)

Fig.6.32 Detection point and detection time for the basic scenario of the bored tunnel typology (PyroSim software)

114 The point of detection based on the rate-of-rise threshold for the basic scenario in the bored

tunnel can be seen in the images above. The green dot represents the detection point of the

FibroLaser. As can be seen, the detection point is at the fire location and, thus, the information

that the FibroLaser system provides for the location of the fire is accurate.

The temperature distribution in the tunnel can be seen in the images above in three different



ceiling.

T=90 seconds

T=110 seconds

T=180 seconds

In the following sketches, the visibility levels can be observed around the fire. The visibility levels


fire there is no important reduction of the visibility. Since the smoke layer downstream of the fire

is highly stratified and limited on the ceiling level, the people trapped can evacuate without


approximately 30 meters which according to the tenability limits, it is considered appropriate in


On the other hand, downstream of the fire the smoke is not stratified for the first 90 meters after

the fire location and the visibility levels are as low as 7-10 meters at the eye level. After the first

90 meters downstream the fire, the visibility is restored to approximately 18-20 meters at eye

level. Nevertheless, the visibility conditions downstream the fire do not significantly affect the

evacuation process, since it is considered that the vehicles have already moved towards the

tunnel exit portal and did not stay close to the fire incident.

Fig.6.33 Temperature distribution (oC) for the basic scenario of the bored tunnel typology


115

T=90 seconds

T=110 seconds

T=180 seconds

Fig.6.34 Visibility conditions (soot visibility in m) for the basic scenario of the bored tunnel typology (Smokeview Software)

116

traffic impact

Similarly to the cut-and-cover tunnel, a simulation scenario that includes the presence of traffic

obstructions was examined for the bored tunnel. In this scenario, the input provided to FDS is the

general input described in section 6.2 and rectangular obstructions in the dimensions of

passenger cars and trucks were added with a longitudinal distance of 4.0 meters the one from the

next. The peak HRR was 1.11 MW and was reached in the first 166 seconds of the fire

development. The maximum recorded temperature in the whole tunnel was around 95oC.

Backlayering started at the 50 seconds and the smoke continued to spread quickly upstream of

the fire until the end of the simulation at 180 seconds. The backlayering length at this point

reached approximately 55 meters. The stratification of the smoke layer is disturbed both

upstream and downstream of the fire and smoke is concentrated between the vehicles. As a

result, the smoke layer covers the whole tunnel section both upstream and downstream the fire.

Nevertheless, downstream of the fire the stratification of the smoke layer was restored after

almost 25 meters. This was not the case in the cut-and-cover tunnel where the smoke remained

mixed downstream of the fire. The aforementioned observations can be seen in the following

sketches that were exported from the SmokeView software.

T=50 seconds

T=180 seconds


was recorded right at the fire location. As a result, the alarm threshold for the absolute

temperature of 50oC was exceeded and the relevant alarm was activated at 145 seconds. Added

to this, the rate-of-rise threshold was exceeded and an alarm was activated at 60 seconds.





FibroLaser cable

Fig.6.35 Smoke propagation in the presence of traffic obstructions in the bored tunnel typology (Smokeview software)

Fig.6.36 Detection point and detection time in the presence of traffic obstructions for the bored tunnel typology (PyroSim software)

117




system provides for the location of the fire is accurate. Similarly to the cut-and-cover tunnel, the

presence of traffic obstructions in the bored tunnel does not have an effect on detecting the

location of the fire.

The temperature distribution in the tunnel can be seen in the following images in three different



ceiling. The temperature is, also, high downstream of the fire and even closer to the road surface.

T=90 seconds

T=110 seconds

T=180 seconds

Fig.6.37 Temperature distribution (oC) in the presence of traffic obstructions for the bored tunnel

typology (Smokeview software)

118


both upstream and downstream of the fire are low. More specifically, for the first 55 meters

upstream the fire the visibility levels are around 18 meters at eye level. In addition, for the first 25

meters downstream the fire the visibility levels are around 20 meters at eye level.

T=90 seconds

T=110 seconds

T=180 seconds

Beyond this zone in the proximity of the fire, there is a smoke free area of a height of 2.5-3.0

meters where the visibility conditions are at the levels of 25-30 meters. These limits are

considered appropriate in order for the people to be able to recognize a safety sign or emergency

exit.

Fig.6.38 Visibility conditions (soot visibility in m) in the presence of traffic obstructions for the bored tunnel typology (Smokeview Software)

119

6.3.2 HGV FIRE SIMULATIONS

The second basic group of simulations was the one performed for a heavy goods vehicle (HGV)

fire. In this case, the fire was imported in the FDS software as described in Section 6.2.5. The

simulation scenarios for the HGV fire were performed for both tunnel geometries and are

presented and analyzed in the following sections.

6.3.2.1 CUT-AND-COVER TUNNEL

basic scenario





until the end of the simulation at 180 seconds. The backlayering length reached approximately 30

meters at 170 seconds and remained almost the same until the end of the simulation. The


of around 2.5 meters height. Downstream of the fire, the smoke fills the whole tunnel section and

covers the whole straight part of the tunnel moving to the exit portal. The aforementioned

observations can be seen in the following sketches that were exported from the SmokeView

software.

T=40 seconds

T=180 seconds


120 Regarding the LHD system, the maximum detected temperature was 230

oC. As a result, the

alarm threshold for the absolute temperature of 50oC was exceeded and the relevant alarm was

activated at 43 seconds and right at the fire location. In addition, the rate-of-rise threshold was

exceeded and an alarm was activated at 60 seconds.




system provides for the location of the fire is accurate also in the case of a much larger fire.


moments of the fire development. Downstream of the fire, a high temperature is maintained

throughout the whole height of the tunnel cross section. A tenability zone is only maintained

upstream of the fire where the temperatures are lower.

T=90 seconds

T=110 seconds

T=180 seconds


FibroLaser cable




Fig.6.40 Detection point and detection time for the basic scenario of the cut-and-cover tunnel typology (PyroSim software)

Fig.6.41 Temperature distribution (oC) of the basic scenario for the cut-and-cover tunnel typology




fire there is a reduction of the visibility for the first 30 meters. Since the backlayering effect

upstream of the fire is small and the smoke is limited on the ceiling level, the people trapped can

evacuate without visibility problems. The visibility levels that can be observed upstream of the fire

are approximately 28-30 meters which according to the tenability limits, are considered

appropriate in order for the people to be able to recognize a safety sign or emergency exit.

On the other hand, downstream of the fire the smoke covers the whole cross section of the tunnel

and, thus, the visibility is significantly restricted. More specifically, the visibility levels are less than

10 meters which means that the evacuation of the area downstream the fire is severely

obstructed.

T=90 seconds

T=110 seconds

T=180 seconds

Fig.6.42 Visibility conditions (soot visibility in m) of the basic scenario for the cut-and-cover tunnel typology (Smokeview Software)

122

geometry variations

In this scenario, similarly to the scenarios for a passenger car fire, the input provided to FDS is

the general input described in section 6.2 but with a different geometrical model. More

specifically, 2.0 more meters were added to the tunnel cross-section height.

The peak HRR was 18.44 MW and was reached in the first 180 seconds of the fire development.

The maximum recorded temperature in the whole tunnel was around 325oC. Backlayering started

at the 40 seconds and the smoke continued to spread upstream of the fire until the end of the

simulation at 180 seconds. The backlayering length reached approximately 80 meters at the end

of the simulation at 180 seconds. The smoke layer is stratified both upstream and downstream of

the fire but has an average height of 5.0-6.0 meters. As a result, the smoke free area below the

smoke layer has an average height of less than 2.0 meters. Downstream of the fire, the smoke

fills the whole tunnel section and covers the whole straight part of the tunnel moving to the exit

portal. However, since the smoke layer is stratified and is adjacent to the tunnel ceiling, a higher

cross-section is appropriate for allowing for a smoke free space under the smoke layer. The

aforementioned observations can be seen in the following sketches that were exported from the

SmokeView software.

T=40 seconds

T=180 seconds

Regarding the LHD system, the maximum detected temperature was approximately 241oC. As a

result, the alarm threshold for the absolute temperature of 50oC was exceeded and the relevant

alarm was activated at 52 seconds right at the fire location. Moreover, the rate-of-rise threshold

was exceeded and an alarm was activated at 60 seconds.


FibroLaser cable




Fig.6.43 Smoke propagation for the higher cross-section scenario of the cut-and-cover tunnel typology (Smokeview software)

Fig.6.44 Detection point and detection time for the higher cross-section scenario of the cut-and-cover tunnel typology (PyroSim software)

123 The point of detection based on the rate-of-rise and absolute temperature thresholds can be seen



system provides for the location of the fire is accurate also in the case of a much larger fire. The

higher cross-section of the tunnel does not significantly affect the detection point. The point is still

at the fire location with the difference that it was shifted towards the direction upstream of the fire.





T=90 seconds

T=110 seconds

T=180 seconds



upstream the fire the visibility levels are around 15 meters at eye level. In addition, downstream

the fire the tunnel is filled with smoke and the visibility levels are at approximately 10 meters at

eye level.

Fig.6.45 Temperature distribution (oC) of the higher cross-section scenario for the cut-and-cover


124

T=90 seconds

T=110 seconds

T=180 seconds

Beyond this zone in the proximity of the fire, there is a smoke free area where the visibility

conditions are at the levels of 25-30 meters. These limits are considered appropriate in order for

the people to be able to recognize a safety sign or emergency exit.

Fig.6.46 Visibility conditions (soot visibility in m) of the higher cross-section scenario for the cut-and-cover tunnel typology (Smokeview Software)

125

ventilation variations

In the basic scenario, a longitudinal ventilation system with jet fans was examined as described in

section 6.2.6. At this point, similarly to the simulation scenario for a passenger car fire, a

longitudinal ventilation system with jet fans installed in closer longitudinal distances was

examined.

In this scenario, the input provided to FDS is the general input described in section 6.2 but a

different configuration of the jet fans is used. The peak HRR was 19.87 MW and was reached in

the first 177 seconds of the fire development. The maximum recorded temperature in the whole

tunnel was around 313oC. Backlayering started at the 30 seconds and the smoke continued to

spread upstream of the fire until the end of the simulation at 180 seconds. The backlayering

length reached approximately 20 meters at the end of the simulation, at 180 seconds. The


of around 2.0 meters height. Downstream of the fire, the smoke fills the whole tunnel section and

covers the whole straight part of the tunnel moving to the exit portal. The aforementioned

observations can be seen in the following sketches that were exported from the SmokeView

software.

T=30 seconds

T=180 seconds

Regarding the LHD system, the maximum detected temperature was approximately 237oC. As a

result, the alarm threshold for the absolute temperature of 50oC was exceeded and the relevant

alarm was activated at around 56 seconds and 2.0 meters downstream of the fire location. Added

to that, the rate-of-rise threshold was exceeded and an alarm was activated at 60 seconds.


FibroLaser cable




downstream the fire

Fig.6.47 Smoke propagation for the alternative ventilation scenario of the cut-and-cover tunnel typology (Smokeview software)

Fig.6.48 Detection point and detection time for the alternative ventilation scenario of the cut-and-cover tunnel typology (PyroSim software)

126 The point of detection based on the rate-of-rise and absolute temperature thresholds can be seen


seen, the detection point is in a small distance from the fire location and, thus, the information

that the FibroLaser system provides for the location of the fire is accurate also in the case of a

different configuration of the ventilation system.

The temperature distribution in the tunnel can be seen in the images above in three different



ceiling.

T=90 seconds

T=110 seconds

T=180 seconds



fire there is a reduction of the visibility for the first 20 meters at approximately 18-20 meters.

Since the backlayering effect upstream of the fire is small and the smoke is limited on the ceiling

level, the people trapped can evacuate without visibility problems. The visibility levels that can be

observed upstream of the fire in the smoke free area are approximately 28-30 meters which

according to the tenability limits, are considered appropriate in order for the people to be able to

recognize a safety sign or emergency exit.

On the other hand, downstream of the fire the smoke covers the whole cross section of the tunnel

and, thus, the visibility is significantly restricted. More specifically, the visibility levels are less than

15 meters which means that the evacuation of the area downstream the fire is severely

obstructed.

Fig.6.49 Temperature distribution (oC) of the alternative ventilation scenario for the cut-and-cover


127

T=90 seconds

T=110 seconds

T=180 seconds

traffic impact

As was previously explained, in this scenario, the input provided to FDS is the general input

described in section 6.2 and rectangular obstructions in the dimensions of passenger cars and

trucks were added with a longitudinal distance of 4.0 meters the one from the next. The peak

HRR was 19.34 MW and was reached in the first 171 seconds of the fire development. The

maximum recorded temperature in the whole tunnel was around 307.5oC. Backlayering started at

the 25 seconds and the smoke continued to spread upstream of the fire until the end of the

simulation at 180 seconds. The backlayering length at this point reached approximately 45

meters. The stratification of the smoke layer is disturbed both upstream and downstream of the

fire and smoke is concentrated between the vehicles. As a result, the smoke layer covers the

whole tunnel section both upstream and downstream the fire. Nevertheless, downstream of the

fire the stratification of the smoke layer tended to be restored but was disturbed again by the jet

fans. The aforementioned observations can be seen in the following sketches that were exported

from the SmokeView software.

T=30 seconds

T=180 seconds

Fig.6.50 Visibility conditions (soot visibility in m) of the alternative ventilation scenario for the cut-and-cover tunnel typology (Smokeview Software)

Fig.6.51 Smoke propagation in the presence of traffic obstructions in the cut-and-cover tunnel typology (Smokeview software)




activated at around 50 seconds and right at the fire location. Added to this, the rate-of-rise

threshold was exceeded and an alarm was activated at 60 seconds.

The point of detection based on the rate-of-rise and absolute temperature thresholds in the case

of a large fire in a tunnel filled with traffic obstructions can be seen in the images above. The

green dot represents the detection point of the FibroLaser. As can be seen, the detection point is

right at the fire location and, thus, the information that the FibroLaser system provides for the

location of the fire is accurate. The presence of traffic obstructions does not affect the accuracy of

the detection location.





T=90 seconds

T=110 seconds

T=180 seconds


FibroLaser cable




Fig.6.52 Detection point and detection time in the presence of the traffic obstructions in the cut-and-cover tunnel typology (PyroSim software)

Fig.6.53 Temperature distribution (oC) in the presence of traffic obstructions for the cut-and-cover




upstream the fire the visibility levels are less than 5 meters at eye level. In addition, downstream

the fire the visibility levels are less than 10 meters at eye level.

T=90 seconds

T=110 seconds

T=180 seconds

As can be observed in the sketches below, the smoke gets trapped between the vehicles and

loses its stratification, thus, significantly reducing the visibility downstream of the fire. As a result,

it can be understood that, especially in a large HGV fire, the people in the proximity of the fire

have to deal with extremely harsh visibility conditions both because of the vehicles present in the

tunnel and because of the trapped smoke.

Nevertheless, in a real fire scenario in a road tunnel, it is more likely that the cars downstream of



simulation scenario. The problem, thus, lies in the area upstream of the fire where the evacuees

should be able to notice the exit signs and move to the appropriate exit.


Fig.6.55 Visibility conditions (soot visibility in m) in the presence of traffic obstructions and turbulence

130

6.3.2.2 BORED TUNNEL

basic scenario





until the end of the simulation at 180 seconds. The backlayering length kept increasing during the

simulation and reached approximately 105 meters at 180 seconds. The stratification of the smoke

layer is maintained upstream and downstream of the fire allowing for a smoke free area of around

2.5 meters height on average. The aforementioned observations can be seen in the following

sketches that were exported from the SmokeView software.

T=40 seconds

T=180 seconds

Regarding the LHD system, the maximum detected temperature was 200oC. As a result, the


activated at 51 seconds and right at the fire location. In addition, the rate-of-rise threshold was

exceeded and an alarm was activated at 60 seconds.


FibroLaser cable




Fig.6.56 Smoke propagation for the basic scenario of the bored tunnel typology (Smokeview software)

Fig.6.57 Detection point and detection time for the basic scenario of the bored tunnel typology (PyroSim software)

131 The point of detection based on the rate-of-rise and absolute temperature thresholds in the case

of a large fire in a bored tunnel can be seen in the images above. The green dot represents the

detection point of the FibroLaser. As can be seen, the detection point is right at the fire location

and, thus, the information that the FibroLaser system provides for the location of the fire is

accurate.


moments of the fire development. Both downstream and upstream of the fire, the highest

temperatures are concentrated very close to the ceiling. Nevertheless, downstream the fire, the

temperature close to the ceiling is high even in a large distance from the fire location. However,

upstream the fire, the highest ceiling temperatures are observed closer to the fire location.

T=90 seconds

T=110 seconds

T=180 seconds

In the following sketches, the visibility levels can be observed around the fire. Upstream of the

fire, for the first 105 meters there is a backlayering of the smoke and the visibility at eye level is

approximately 20 meters. The visibility levels that can be observed downstream of the fire are

similar and the visibility at eye level is approximately 20 meters. As a result, the people trapped

upstream of the fire will be able to evacuate moving in the smoke free area below the smoke

layer but the visibility conditions render the evacuation process more difficult. The exit signs and

the exit doors will be harder to recognize if installed higher in the tunnel cross-section.

Fig.6.58 Temperature distribution (oC) for the basic scenario of the bored tunnel typology


132

T=90 seconds

T=110 seconds

T=180 seconds

traffic impact

As was previously explained, in this scenario, the input provided to FDS is the general input

described in section 6.2 and rectangular obstructions in the dimensions of passenger cars and

trucks were added with a longitudinal distance of 4.0 meters the one from the next. The peak

HRR was 18.30 MW and was reached in the first 180 seconds of the fire development. The

maximum recorded temperature in the whole tunnel was around 349.5oC. Backlayering started at

the 30 seconds and the smoke continued to spread upstream of the fire until the end of the

simulation at 180 seconds. The backlayering length at this point reached approximately 120

meters. The stratification of the smoke layer is disturbed both upstream and downstream of the

fire and smoke is concentrated between the vehicles. As a result, the smoke layer covers the

whole tunnel section both upstream and downstream the fire. Nevertheless, downstream of the

fire the stratification of the smoke layer tended to be restored but was disturbed again by the jet

fans. The aforementioned observations can be seen in the following sketches that were exported

from the SmokeView software.

T=30 seconds

T=180 seconds

Fig.6.59 Visibility conditions (soot visibility in m) for the basic scenario of the bored tunnel typology (Smokeview Software)

Fig.6.60 Smoke propagation in the presence of traffic obstructions in the bored tunnel typology (Smokeview software)




activated at around 57 seconds right at the fire location. Added to this, the rate-of-rise threshold

was exceeded and an alarm was activated at 60 seconds.

The point of detection based on the rate-of-rise and absolute temperature thresholds in the case

of a large fire in a bored tunnel with the presence of traffic obstructions can be seen in the images

above. The green dot represents the detection point of the FibroLaser. As can be seen, the

detection point is right at the fire location and, thus, the information that the FibroLaser system

provides for the location of the fire is accurate. The presence of traffic obstructions does not affect

the detection of the fire location.


moments of the fire development. Both downstream and upstream of the fire, the highest

temperatures are concentrated very close to the ceiling. Nevertheless, downstream the fire, the

temperature close to the ceiling is high even in a large distance from the fire location. However,

upstream the fire, the highest ceiling temperatures are observed closer to the fire location.

T=90 seconds

T=110 seconds

T=180 seconds


FibroLaser cable




Fig.6.61 Detection point and detection time in the presence of traffic obstructions for the bored tunnel typology (PyroSim software)

Fig.6.62 Temperature distribution (oC) in the presence of traffic obstructions in the bored tunnel

typology (Smokeview software)

134 In the following sketches, the visibility levels can be observed around the fire. As can be observed

in the sketches below, the smoke gets trapped between the vehicles and loses its stratification,

thus, significantly reducing the visibility both upstream and downstream of the fire. Upstream of

the fire, for the first 120 meters there is a backlayering of the smoke and the visibility at eye level

is lower than 5 meters at eye level. The visibility levels that can be observed downstream of the

fire are, also, low and the visibility at eye level is approximately 10 meters. As a result, the people

trapped upstream of the fire will be able to evacuate moving in the smoke free area below the

smoke layer but the visibility conditions render the evacuation process more difficult. The exit

signs and the exit doors will be harder to recognize if installed higher in the tunnel cross-section.

T=90 seconds

T=110 seconds

T=180 seconds

Nevertheless, in a real fire scenario in a road tunnel, it is more likely that the cars downstream of



simulation scenario. The problem, thus, lies in the area upstream of the fire where the evacuees

should be able to notice the exit signs and move to the appropriate exit.

6.4 CONCLUSIONS

At this point, the conclusions derived from the simulation scenarios regarding the performance of

the fire detection system will be analyzed. The basic parameters that were evaluated in order to

formulate the conclusions were: the time of the alarm activation, the point of the fire detection and

the alarm threshold for which the alarm was activated.

Regarding the time of the alarm activation, it was observed that in all the simulation scenarios the

fire was detected within the first 2 minutes of the fire development. In the case of the passenger

car fire, the fastest detection time was 60 seconds (based on the rate-of-rise alarm threshold) and

was observed in the case of the bored tunnel that is filled with traffic obstructions. The highest

time of alarm activation was 114 seconds (based on the rate-of-rise alarm threshold) and was

Fig.6.63 Visibility conditions (soot visibility in m) in the presence of traffic obstructions in the bored tunnel typology (Smokeview Software)

135 observed in the case of the cut-and-cover tunnel in the basic scenario simulation and the higher

cross-section simulation scenario. In the case of a HGV fire, the fastest detection time was 43

seconds (based on the absolute temperature threshold) and was observed in the case of the cut-

and-cover tunnel during the basic scenario simulation. In addition, the alarm based on the rate-of-

rise threshold was activated at 60 seconds in all the simulation cases.

Regarding the point of fire detection, it was observed that in all the simulation scenarios the

detection point is right at the fire location and, thus, the information that the FibroLaser system

provides for the location of the fire is accurate. More specifically, the maximum distance that was

observed between the fire and the detection point where the fire was detected was 2.0 meters in

the direction downstream of the fire location. The most accurate detection of the fire location was

observed in the case of the cut-and-cover tunnel with a higher cross-section. It was concluded

that a different configuration of the ventilation system did not have a major effect on the detection

of the fire location. It was observed that when the jet fans are installed in closer longitudinal

distances and the power of the ventilators is lower, the detection point shifts to the side upstream

the fire location. Moreover, it was concluded that the presence of traffic obstructions does not

affect the accuracy of the detection of the fire location.

Regarding the alarm thresholds for which the alarm was activated, it was observed that the alarm

activated depended mainly on the size of the fire. More specifically, in the case of the passenger

car fire, the alarm was activated based on the rate-of-rise threshold. That was because within the

first 3 minutes of the fire development the temperature did not exceed the threshold of 50oC.

Nevertheless, it was observed that in the case where traffic obstructions are present in the tunnel,

the alarm based on the absolute temperature was also activated but later than the alarm based

on the rate-of-rise. Especially in the case of the cut-and-cover tunnel with traffic obstructions, the

alarm based on the absolute temperature threshold was activated after the first 2 minutes (at 175

seconds). On the other hand, in the case of a HGV fire, an alarm was activated based on the

rate-of-rise temperature usually within the first 60 seconds of the fire development followed by the

activation of the alarm based on the absolute temperature.

136

CHAPTER 7: Evacuation Simulations In this chapter, the evacuation simulations performed are going to be described and the input of each simulation is going to be discussed. The basic choices regarding the evacuation simulations will be justified and commented. Finally, the output of the simulations is, also, going to be presented and shortly analyzed. The most important results are going to be shown by means of visual output and tables.

7.1 INTRODUCTION

As was derived by the literature review, the human factor is the most significant parameter that

needs to be taken into account during a fire incident in a road tunnel. At the same time, it is the

one of the most unpredictable variables that a tunnel designer has to evaluate. As was already

mentioned, the lack of reliable experiments on human behavior during a fire incident makes it

difficult to predict the reaction of people. For example, it is not yet clear how toxic smoke affects

the physical condition of the evacuees. Added to that, there are factors that affect the human

reaction that are subjective and depend on the conditions during each incident. For instance, a

person reacts differently if alone or under the presence of other family members and especially

kids.

For all the above reasons, it is almost impossible to predict with accuracy the reaction time of

people present in a fire incident in a tunnel. Usually, it is mentioned in literature that the reaction

time can be between 5 to 15 minutes. With an early detection, this time range does not

necessarily change but it shifts earlier in the fire development making the actual evacuation

process easier.

In the following sections, thus, the human factor is going to be implemented in the simulation

scenarios performed. In order to achieve that, the pedestrian evacuation component of the FDS

software was used and a set of simulations were performed. As a result, conclusions could be

drawn regarding the impact of the detection time on the total evacuation process. During this

study, the basic parameters to be taken into account will be the tenability conditions in the tunnel

and the FED as generated by the FDS software. More specifically, the evacuation process will be

examined on the point of view of the evacuees and conclusions regarding the position of the

signage and the exit doors will be formulated.

All in all, this Chapter has a dual purpose. On the one hand, it aims to offer an insight on the

factors that affect the evacuation process so that certain design decisions could be better

justified. On the other hand, the Chapter intends to evaluate the evacuation module of the FDS

software and suggest an appropriate use and application for such evacuation simulation tools.

137

7.2 FDS+EVAC SOFTWARE

As was already mentioned, the simulations for the evacuation of the pedestrians were performed

through a module of the FDS software named Evac. By using FDS+Evac, fire and evacuation

processes can be simulated at the same time interacting with each other. Moreover, the human

evacuation process can be simulated without any fire phenomena and, thus, simulations of fire

drills can also be performed [106].

Each evacuee within FDS+Evac is treated as a separate agent with its own properties as well as

its own escape strategies. The agents are considered to move on a two-dimensional plane that

represents the road surface of the tunnel. According to the FDS+Evac User Guide, the algorithm

used to represent the egress movement works by solving an equation of motion for each of the

agents within a continuous two-dimensional space and time.

The Evac module is mainly developed as a tool for simulating the evacuation process in

buildings. It has been shown through comparison with experiments and other simulation tools that

it can produce reliable results. However, the developers warn that it is not yet completely

validated and that it should be used in combination with other egress simulation tools or methods.

For that reason, the simulations performed with the FDS+Evac within the context of this research

study are going to be used in order to evaluate the capabilities of the software and to understand

the conditions in the tunnel and their impact on the evacuees. As a result, the conclusions that

will be drawn from the simulation results will be mainly of a qualitative nature.

Before running the evacuation simulations with FDS+Evac, it is important to understand the

limitations of the software:

Geometry: Firstly, the simple rectilinear mesh used for the FDS simulations is efficient

for simpler geometries but can be limiting in case of more complex geometries. The

current Evac module uses two-dimensional planes and the mesh abides to these planes.

There is, also, the possibility to efficiently model inclined which is very important in the

case of the tunnel modeled in the research study.

Reduced visibility: The smoke has an effect on the visibility conditions and, thus, on the

walking speed of the agents. Within the FDS+Evac, the walking speeds of the agents are

reduced in the presence of smoke and depending on its concentration according to the

results of an experiment performed by Frantzich and Nilsson [107]. It is recognized,

however, that the experimental results are not adequate and new research is still needed.

Incapacitation: The model used for incapacitation is the Fractional Effective Dose (FED)

model by Purser [108]. However, there is a significant scatter between the different

agents that is not accounted for in the latest version of the FDS software. An agent is

considered to be incapacitated if the FED value is above 1.0. Then, the speed of the

agent is considered to be zero and it is assumed that the agent experiences no social

forces. It is suggested though in the FDS+Evac manual that this is a very rough approach

and it should be reconsidered in the next versions of the software.

Exit route selection: The algorithm used to describe the selection of the exit door is still

simplified. The control of the doors to be selected by the agents is basically on the user

which means that the responsibility lies on the user in order to produce reliable results.

This introduces an uncertainty in the whole process.

138

Detection and reaction times: The detection time can be interrelated to the FDS

software simulation results according to the calculated response of the smoke or heat

detectors. The pre-movement time is imported by the user for each category of

evacuating agents depending on their sex, age, physical condition and position in the

tunnel.

After running the evacuation simulations, the results can be visualized in the Smokeview software

similarly to the fire simulations performed with FDS.

7.3 EVACUATION SIMULATIONS

In this section the process followed in order to perform the evacuation simulations is going to be

described and the results of the simulation scenarios performed will be presented. First, the input

that was imported in the FDS+Evac software will be defined. Following, the scenarios that were

chosen to be simulated will be justified and the results are going to be presented.

7.3.1 FDS+EVAC INPUT

In the case of the FDS+Evac simulations, the user has to define most of the variables according

to the specific scenario. The parameters regarding the evacuation mesh, the number and type of

the evacuees, the walking speed of the agents, the positions of the agents of groups of agents in

the tunnel and the location of the emergency exits have to be imported before running the

simulations. The input parameters that were defined in the model of this research study are going

to be described in detail in the following sections.

mesh

A mesh had to be defined for the evacuation simulations that is different from the one for the fire

simulations. So, a two-dimensional mesh parallel to the road surface of the tunnel was defined in

the x-y directions. In the z direction, the mesh should have only one cell. It is important that the

mesh boundaries of the fire mesh and the evacuation mesh should not overlap.

In the example input file below, the fire mesh and the three evacuation meshes are indicated. The

fire mesh (grey colored text below) was chosen to be a coarse mesh (grid cell size: 1.0*1.0*1.0

meters) in order to reduce the required computational time. The three evacuation meshes (pink

colored text below) were also chosen to be coarse (grid cell size: 1.0*1.0*1.0 meters) for the

same reason.

&MESH ID='1x1x1', IJK=15,1497,37, XB=0.0,15.0,1.5,1498.5,-5.5,31.5/

&MESH ID='1x1x01', IJK=17,300,1, XB=-1.0,16.0,602.5,902.5,-3.5,-2.5, EVACUATION=.TRUE., EVAC_HUMANS=.TRUE., EVAC_Z_OFFSET=1.5/ &MESH ID='1x1x02', IJK=17,598,1, XB=-1.0,16.0,2.5,600.5,-3.5,-2.5, EVACUATION=.TRUE., EVAC_HUMANS=.TRUE., EVAC_Z_OFFSET=1.5/ &MESH ID='1x1x03', IJK=17,595,1, XB=-1.0,16.0,902.5,1497.5,-3.5,-2.5, EVACUATION=.TRUE., EVAC_HUMANS=.TRUE., EVAC_Z_OFFSET=1.5/

In order to fit the mesh to the geometry of the tunnel, the tunnel was separated into three sections

and a mesh was assigned in each of these parts. For the two inclined parts, an inclined mesh had

to be applied using the appropriate command in the input file. The command is the EVSS

139 namelist group which basically changes the z coordinates of the agents.

agent properties

The properties of the agents that have to be defined include body diameters, walking speeds and

pre-evacuation times [106]. The FDS+Evac has five agent types as default and according to the

agent type the body sizes and walking speeds are determined by default. The different default

agent types and the default values used by the software can be seen in the following table:

It is suggested by the FDS+Evac User Guide that the user should not change any of these

parameters. It is considered sufficient to define the agent types expected to be present and to set

a distribution for the detection and reaction time.

As a result, it was assumed that all the agent types were present in the tunnel during the fire

incident. In order to define the number of agents of each type, statistics regarding the population

in the Netherlands were consulted and the ratios indicated by the statistics were used [109]. Also,

the number of vehicles that would be present in the tunnel in case of a fire incident in any given

point in time was defined by consulting the measurement of the traffic in the case study tunnel.

Demographic data were consulted regarding the population by age and by gender in order to

define the percentage of women, men and children expected to be present in the tunnel at any

given moment. The percentage of men and women in the total population is almost 50%

respectively. It was assumed that every car has an average of 2 passengers and, as was already

mentioned, in 2010 the number of vehicles using the road that passes through the Hubertus

tunnel was 32,900 vehicles. So, the number of vehicles per minute was calculated and, taking

into account that the maximum speed allowed in the tunnel is 70 km/h it was calculated that a

vehicle needs almost 1.5 minutes to cross the tunnel. As a result, a rough estimation was

performed attributing 2 passengers in each vehicle and assuming that 50% of the total

passengers is men and the other 50% is women. Children and elderly were, also, added as

agents.

In addition, the tunnel was separated into three sections-two inclined parts and a horizontal one-

and evacuation meshes were assigned in each part. Then, the number of evacuating agents was

equally divided in these three sections of the tunnel and was defined in the input file.

Finally, the detection time was imported according to the results of the fire simulations. Before

running each evacuation simulation, the detection time that was calculated for each fire

simulation scenario was imported in the Evac module. Below is an example of an input file used

for one evacuation simulation scenario that indicates the agent types and the detection time. The

mean detection time for all the simulations was estimated to be 90 seconds.

&PERS ID='Women', DEFAULT_PROPERTIES='Female', DET_EVAC_DIST=0, DET_MEAN=90.0/

Table 7.1 Unimpeded walking speeds and body dimensions in FDS+Evac software

140 &PERS ID='Men', DEFAULT_PROPERTIES='Male', DET_EVAC_DIST=0, DET_MEAN=90.0/

&PERS ID='Children', DEFAULT_PROPERTIES='Child', DET_EVAC_DIST=0, DET_MEAN=90.0/

&PERS ID='Elderly', DEFAULT_PROPERTIES='Elderly', DET_EVAC_DIST=0, DET_MEAN=90.0/

In addition to the agent types, their exact number and their position in the tunnel had to be

defined. In the example below, the group “women” is defined:

&EVAC ID='Position 1', XB=1.0,14.0,603.5,901.5,-3.5,-2.5, NUMBER_INITIAL_PERSONS=15,

PERS_ID='Women', AVATAR_RGB=204,0,153, KNOWN_DOOR_NAMES='E02','E03','E04','E05','E06'/

exits and doors

In FDS+Evac, the exits are defined by the user that has to define the location and size of the

emergency exit doors. Through the exits, the agents are removed from the calculation. In the

context of this research study, the emergency exits were assumed to be placed every 60 meters

along the tunnel.

In the following input file example, two consecutive exits are defined by setting their coordinates,

the direction of the flow during evacuation (IOR=-1) and the mesh they are assigned to.

&EXIT ID='E02', XB=1.0,1.0,661.0,662.5,-3.5,-2.5, RGB=51,204,0, XYZ=1.0,661.75,-3.0, IOR=-1,

MESH_ID='1x1x01'/

&EXIT ID='E03', XB=1.0,1.0,721.0,722.5,-3.5,-2.5, RGB=51,204,0, XYZ=1.0,721.75,-3.0, IOR=-1,

MESH_ID='1x1x01'/

In the image below, a visualization of the evacuation simulation model as generated by the

Smokeview software can be seen at a specific point in time (t=85 seconds). The agents are still

not moving since the detection of the fire incident did not yet take place. The doors are

represented on the model with a green color.

Fig.7.1 Visualization of the evacuation simulation model (Smokeview software)

141

7.3.2 EVACUATION SIMULATION SCENARIOS

Similarly to the fire simulations, the evacuation simulations were performed for a passenger car

fire and a HGV fire for the cut-and-cover tunnel geometry. The simulation scenarios that were

examined were the basic scenarios and the ones that were expected to have a significant impact

on the evacuation process. By the fire simulations it was concluded that a higher cross-section

can provide better conditions for a safer evacuation. For that reason, this case was chosen to be

examined for the evacuation simulations. As a result, the scenarios simulated were the following:

Passenger car fire: basic scenario

Passenger car fire: higher cross-section

HGV fire: basic scenario

HGV: higher cross-section

In the following sections, thus, the evacuation simulation scenarios are going to be presented

separately for the passenger car fire and the HGV fire.

7.3.2.1 PASSENGER CAR FIRE

First, the evacuation process in the case of a passenger car fire was examined both for the basic

scenario and for the case of a higher cross-section. During the fire simulations, it was observed

that the conditions in the tunnel in this case were not very severe and the fire development and

growth were of a moderate rate. As a result, the evacuees are expected to be able to leave the

tunnel safely after being informed about the incident.

basic scenario

In this scenario, the evacuation process lasted approximately 65 seconds counting from the time

the first agents started moving until the last agent left the tunnel through the emergency exit

doors. As was observed in the fire simulations, the backlayering length was very small and, as a

result, the agents upstream the fire can easily evacuate the tunnel (Fig.7.2). On the other hand,

downstream the fire, the smoke is propagating faster but the agents have still time to locate the

exits and evacuate.

In the following image, the fire from the viewpoint of an agent standing upstream and in the close

proximity of the fire is presented at the time right before the start of the evacuation (at 85

seconds). It can be observed that there is a clear view in the tunnel and the signs and emergency

exits can easily be located. Also, at this point, the flames are already visible which means that the

evacuees can understand that there is a fire emergency.

142

In the next image, the fire from the viewpoint of an agent standing downstream the fire location is

presented at the time right before the start of the evacuation (at 85 seconds). It can be observed

that there is a clear view in the tunnel and the signs and emergency exits can easily be located.

At this distance, the agents cannot clearly identify the nature of the incident and, thus, the

information provided to them through the sound alarm message should be clear and provide the

evacuees with all the information necessary to lead them to a safe evacuation.

In Fig.7.4, a later stage of the evacuation process (at 110 seconds) is presented. The evacuees

have already started moving towards the closest exits and they can evacuate unobstructed by

Fig.7.2 Agent viewpoint upstream of the fire at t=85 sec in the basic scenario for passenger car fire

Fig.7.3 Agent viewpoint downstream of the fire at t=85 sec in the basic scenario for passenger car fire

143 smoke. Finally, in Fig.7.5 it can be seen that even when the last agent leaves the tunnel (at 130

seconds), the smoke is moving but has not occupied the whole tunnel.

As was observed in the fire simulations, there is a tenable zone in the proximity of the fire as far

as temperature is concerned. The maximum temperature is around 27oC which means that the

tunnel can be occupied during the first 180 seconds of the fire development. Until the end of the

simulation, thus, the temperature is not a hazard to the health of the evacuees. However, the

smoke threatens the health of the evacuees as can be derived from the evacuation simulation.



144 More specifically, at around 115 seconds, it was calculated that in the agents that are

downstream in the proximity of the fire the FED index is approximately 2.0*10-4

.

Moreover, the agents near the fire will have a difficulty locating the exits both because of the

reduced visibility and because of the impact of the smoke toxicity on their health and perception

of the environment. Nevertheless, the evacuation simulation showed that even these agents

managed to evacuate safely.

Fig.7.6 Temperature distribution (oC) in the proximity of the fire at t=115 sec for a

passenger car fire

Fig.7.7 Smoke propagation and toxicity (FED) in the proximity of the fire at t=115 sec for a passenger car fire

145

higher cross-section



doors. As was observed in the fire simulations, the backlayering length was very small and, as a

result, the agents upstream the fire can easily evacuate the tunnel (Fig.7.8). On the other hand,

downstream the fire, the smoke is propagating faster but the agents have still time to locate the

exits and evacuate. Compared to the previous fire scenario, however, the smoke layer is less

thick and the smoke free area under the smoke layer is higher.












Fig.7.8 Agent viewpoint upstream of the fire at t=85 sec in the higher cross-section scenario for passenger car fire

146

As was observed in the fire simulations, there is a tenable zone in the proximity of the fire as far

as temperature is concerned. The maximum temperature at 155 seconds is around 24oC which

means that the tunnel can be occupied during the first 180 seconds of the fire development. Until

the end of the simulation, thus, the temperature is not a hazard to the health of the evacuees.

Similarly to the basic scenario, the smoke threatens the life of the evacuees. However, in the

case of a higher tunnel, the FED calculated at the time of fire detection was around 1.5*10-4

.

Fig.7.9 Agent viewpoint downstream of the fire at t=85 sec in the higher cross-section scenario for passenger car fire

Fig.7.10 Temperature distribution (oC) in the proximity of the fire at t=155 sec for the

higher cross-section scenario for a passenger car fire

147

Moreover, the agents near the fire will have a difficulty locating the exits both because of the

reduced visibility and because of the impact of the smoke toxicity on their health and perception

of the environment. Nevertheless, the evacuation simulation showed that in the case of a higher

cross-section, the visibility conditions are better and the FED recorded was lower. It can, also, be

observed that the thickness of the smoke layer increases closer to ceiling level which means that

the signs would be more easily discernible if placed lower on the tunnel walls.

7.3.2.2 HGV FIRE

At this point, the evacuation process in the case of a HGV fire was examined both for the basic

scenario and for the case of a higher cross-section. During the fire simulations, it was observed

that the conditions in the tunnel in this case were extremely severe and the fire development and

growth were of a much faster rate. As a result, the evacuees are expected to face many

difficulties during the fire emergency and probably even not be able to leave the tunnel safely

after being informed about the incident. The difficulties mainly lie on the restriction of the visibility,

the underestimation of the speed of the fire development and the amount of smoke and its

propagation in the tunnel.

basic scenario

In this scenario, the evacuation process was not completed until the end of the simulation. So, in

the first 90 seconds after the evacuation process began, most of the agents have evacuated the

tunnel. The ones that were still left in the tunnel at that point, were those that were downstream

and in the close proximity of the fire. More specifically, the presence of smoke in the tunnel has

prevented the agents from noticing the exit closest to the fire. As a result, they run past it and

tried to reach the next visible exit. At the same time, their walking speed has decreased because

of the smoke and their physical condition and they could not reach the exit within the first 90

seconds of the evacuation process.

Fig.7.11 Smoke propagation and toxicity (FED) in the proximity of the fire at t=155 sec for the higher cross-section scenario for a passenger car fire

148 In the following image, the fire from the viewpoint of an agent standing upstream and in the close

proximity of the fire is presented at the time of the fire detection (at 43 seconds). This is when the

alarm will sound and the evacuees will get informed that there is a fire incident. It can be

observed that there is a clear view in the tunnel and the signs and emergency exits can easily be

located. Also, at this point, the flames are already visible which means that the evacuees can

understand that there is a fire emergency.

In the next image, the fire from the viewpoint of an agent standing downstream of the fire location

is presented at the time of fire detection (at 43 seconds). It can be observed that there is a clear

view in the tunnel and the signs and emergency exits can easily be located. At this distance, the

agents cannot clearly identify the nature of the incident and, thus, the information provided to

them through the sound alarm message should be clear and provide the evacuees with all the

information necessary to lead them to a safe evacuation. However, this is still a very early stage

of the fire development and the fast propagation of the smoke should not be underestimated.

Fig.7.12 Agent viewpoint upstream of the fire at t=43 sec (point of fire detection) in the basic scenario for HGV fire

Fig.7.13 Agent viewpoint downstream of the fire at t=43 sec (point of fire detection) in the basic scenario for HGV fire

149 In order to understand the speed of the fire development, the conditions in the tunnel at 115

seconds were, also, examined. The smoke layer is now very thick, especially, downstream of the

fire location and is propagating fast. By then, the evacuees appear to have taken their positions

by the emergency exits waiting to evacuate.

Fig.7.14 Agent viewpoint upstream of the fire at t=115 sec in the basic scenario for HGV fire

Fig.7.15 Agent viewpoint downstream of the fire at t=115 sec in the basic scenario for HGV fire

150 By the fire simulations it can be seen that the temperature near the fire at the point of fire

detection (43 seconds) has reached up to approximately 30oC. This means that until that point in

time, the environment in the tunnel is tenable as far as the temperature is concerned. However,

the smoke develops and propagates very fast and gradually fills the whole tunnel.

At 43 seconds when the fire detection takes place, the smoke layer in the proximity of the fire is

already covering the whole height of the tunnel and the people in this area are affected by the

toxicity of the smoke. It was calculated that the FED of the agents downstream the fire can be up

to approximately 0.5*10-4

.

Fig.7.16 Temperature distribution (oC) in the proximity of the fire at t=43 sec (fire

detection point) for the basic scenario for a passenger car fire

Fig.7.17 Smoke propagation and toxicity (FED) in the proximity of the fire at t=43 sec (fire detection point) for the basic scenario for a HGV fire

151 In addition, it can be observed that although the evacuation can start early with a fast fire

detection, the people evacuate slower due to the unfavorable visibility conditions. So, it can be

seen by the simulation results that until a person that was closer to the fire location reaches the

closest emergency exit, the FED has reached more than 2.0*10-4

.

higher cross-section



doors. Unlike the previous scenario, the higher cross-section allows for better visibility conditions

even in the proximity of the fire and the exits that are closer to the fire incident can more easily be

noticed. As a result, the evacuation process finished much faster and the agents downstream and

close to the fire evacuated through the closest exit.












Fig.7.18 Agent viewpoint upstream of the fire at t=55 sec (point of fire detection) in the higher cross-section scenario for HGV fire

152

By the fire simulations it can be seen that the temperature near the fire at the point of fire

detection (55 seconds) has reached up to approximately 45oC. According to the tenability limits, a

person can be exposed to such a temperature for a maximum of 26.9 minutes before the

incapacitation symptoms start to appear. Thus, this allowable exposure time is enough for the

evacuation process to be successfully completed. As a result, the temperature does not

constitute a threat for the human health even in the HGV fire scenario. Similarly to the basic

scenario, the smoke threatens the life of the evacuees. However, in the case of a higher tunnel,

the FED calculated at the time of fire detection was more than 0.6*10-4

.

Fig.7.19 Agent viewpoint downstream of the fire at t=55 sec (point of fire detection) in the higher cross-section scenario for HGV fire

Fig.7.20 Temperature distribution (oC) in the proximity of the fire at t=55 sec (fire

detection point) for the higher cross-section scenario for a HGV fire

153

In the following image it can be observed that the last person that leaves the tunnel at around 140

seconds is affected by the toxicity of the smoke. However, the FED that was calculated was

approximately 1.8*10-4

which is less than the FED observed in the basic scenario simulation.

As a result, it can be concluded that a higher tunnel cross-section has a positive impact on the

evacuation process. The smoke free area that is maintained is adequate for the evacuees to

reach the closest emergency exit. More specifically, the doors are visible and the passage way

that leads to the doors is clear enough. Once again, it can be also concluded that the sings would

be more visible if placed lower on the tunnel walls since the smoke layer is thicker closer to the

tunnel ceiling level.

Fig.7.22 Smoke propagation and toxicity (FED) in the proximity of the fire at t=140 sec for the higher cross-section scenario for a HGV fire

Fig.7.21 Smoke propagation and toxicity (FED) in the proximity of the fire at t=55 sec (fire detection point) for the higher cross-section scenario for a HGV fire

154

7.4 CONCLUSIONS

At this point, after examining the fire scenarios from the evacuation point of view, the following

conclusions can be drawn:

The evacuation models are not considered reliable and more research and development

has to be done in order to achieve adequate validation.

The Evac module of the FDS software has yet to be developed in order to provide more

reliable results. Also, the user interface should be more friendly and flexible.

Finally, the evacuation simulation module can be used in such a way that safe

conclusions can be drawn. It can be used in a more qualitative way in order to help the

designer understand the conditions in the tunnel and the point of view of the evacuees.

A way to make sure that the evacuation models correspond to reality would be to render

the real evacuation process more designated and, thus, more predictable. This way, the

unknown variables of the evacuation simulation scenarios are limited and the process

can be modeled without many uncertainties. In practice, this can be achieved by

providing the proper amount of information during a fire emergency in a tunnel and by

guiding the evacuees to achieve a safe egress. Then, the evacuation process could be

reproduced with the evacuation software in order to assess the efficiency of the

evacuation plan and the position of the signage and exits.

155

CHAPTER 8: Conclusions In this chapter, the conclusions regarding the outcome of the research are going to be presented. In addition, the connection between the research objectives and the conclusions is going to be made and the general impact of the conclusions is going to be discussed. Finally, recommendations for future research and recommendations for practice are going to be suggested.

8.1 INTRODUCTION

At this point, after completing the research study, the relevant conclusions are going to be

formulated. The conclusions were distinguished between those drawn from the fire simulations

and the ones from the evacuation simulations and mostly regard the FDS software reliability for

conducting fire simulations, the assessment of the Evac component, the detection time in the

various simulation scenarios, the impact of each change on the model on the detection time and

the performance of the Linear Heat Detection system used.

In addition, recommendations for further research and for practice, resulting from the research

study, are going to be suggested. The need for further research regarding the FDS software and

its Evac component, further validation cases, more detection technologies and more simulation

scenarios is going to be discussed. Moreover, the need for changes in the existing practice and

the regulations will be suggested.

8.2 CONCLUSIONS

8.2.1 FIRE SIMULATIONS

The conclusions drawn by the fire simulations are related to the FDS software, the Linear Heat

Detection performance and the detection time in the various simulation scenarios. The following

table provides an overview of the results regarding the fire detection time on which the

formulation of the relative conclusions was based.

Table 8.1 Overview of the results regarding the detection time in the various fire simulation scenarios

156 The conclusions formulated after conducting the various scenarios of the fire simulations using

the FDS software are summarized in the following:

Until recently, the regulations related to tunnel fire safety were determined according to

the experience and observations of past incidents with fire in tunnels. Using a tool like

FDS the regulations can be composed using a more performance-based approach.

The FDS software is a reliable tool for simulating the Linear Heat Detection system

provided that the proper validation process is followed using full-scale tests of the fire

detection technology.

In the FDS software there is no default option for modeling the LHD system. However,

modeling the LHD as multiple consecutive point heat detectors produces reliable results

as indicated by the validation case followed.

Modeling the LHD using multiple meshes can provide reliable results as indicated by the

grid sensitivity conducted provided that the rules for the application and alignment of the

meshes are strictly followed.

The LHD technology and specifically, the FibroLaser system was proven to be an

effective system for detecting both smaller and larger fires.

The absolute temperature alarm threshold was proven to be more appropriate for larger

fires and the rate-of-rise alarm threshold was more appropriate for smaller fires were the

absolute temperature was not exceeded.

The FibroLaser system detected fire in less than 2 minutes in every fire scenario that was

simulated.

The FibroLaser system detected the fire location with accuracy in every simulation

scenario. The maximum divergence from the fire location that was observed during the

simulations was 2.0 meters (downstream the fire).

A higher tunnel cross-section of a cut-and-cover tunnel results in more detection time but,

also, in a more accurate detection of the fire location.

The presence of traffic obstructions does not have a major impact on the fire detection

time. It was shown that with the presence of traffic, the detection of the fire location is still

very accurate.

Installing the jet fans in closer longitudinal distances slightly increases the detection time

in the case of a larger fire and, at the same time, results in shifting the detection point

upstream of the fire location.

8.2.2 EVACUATION SIMULATIONS

The conclusions formulated after conducting the various scenarios of the evacuation simulations

corresponding to the fire simulations using the FDS+Evac software are summarized in the

following:

The evacuation simulation tools are not yet reliable since a lot of research is still needed

in order to provide validation cases. Moreover, the validation cases cannot be

157

representative of the reality since there are many limitations involved in the conduction of

proper evacuation experiments (e.g. humans cannot be exposed to toxic smoke and,

thus, the effect of toxic smoke on the physical condition and perception of evacuating

agents cannot be accurately determined).

Moreover, the unpredictable human factor is a major impediment in developing reliable

evacuation simulation tools. The subjective judgment of each evacuee can affect the

whole evacuation process.

Smoke was proven to be a biggest threat than temperature because of the reduction of

visibility than can cause the evacuees in the proximity of the fire to lose the closest exit

and be exposed for more time to the toxic smoke.

A higher cross-section of the cut-and-cover tunnel provides better visibility conditions and

results in a more stratified smoke layer adjacent to the tunnel ceiling level. Thus, the

evacuees are less affected by the toxicity of the smoke and incapacitation is further

delayed.

8.3 RECOMMENDATIONS

8.3.1 RECOMMENDATIONS FOR FUTURE RESEARCH

At this point, the recommendations for future research as derived from this research study are

going to be presented.

The most state-of-the-art detection technologies available in the market should be

simulated in order to examine their performance under different scenarios.

More validation cases should be provided which means that full-scale tests of the specific

detection technologies should be performed and made accessible to the researchers.

The fire simulation scenarios and the evacuation simulation scenarios performed in the

context of this research study could be performed with a finer grid (0.5*0.5*0.5 meters).

Simulating a large tunnel domain, though, with such a fine grid will result in expensive

simulations in terms of computational time.

The impact of moving traffic obstructions on the fire phenomena as well as the detection

time would be interesting to be researched using the FDS software. This would help to

define the effect of the turbulence created by the moving vehicles on the detection time

and detection of fire location.

The fire simulation scenarios could be examined using the Multiscale Modeling Method

suggested by Francesco Colella [68, 110].

158

8.3.2 RECOMMENDATIONS FOR PRACTICE

Finally, the recommendations for practice as derived from this research study are going to be

presented.

The fire safety engineers designing the fire safety systems in a tunnel are recommended

to use the FDS as a tool to examine different scenarios of fire safety systems

configuration. Having such a tool offers flexibility in modeling more fire scenarios by

avoiding the costly, in terms of time and money, real scale tests. However, the FDS

should not be used unless a proper validation process is first followed in order to ensure

the reliability of the results.

Tests and specifications for the more state-of-the-art systems of fire detection for tunnel

applications should be made available by the suppliers of these technologies. The

working principles and performance specifications and details for Multiple Gas Detection

technology should be more transparent.

The FibroLaser detection system is a reliable system for fire detection in a tunnel, both

for smaller and larger fires. Before installing a LHD system in a tunnel, the FDS software

can be used in order to help define the appropriate alarm thresholds for the activation of

the detectors.

A way to make sure that the evacuation models correspond to reality would be to render

the real evacuation process more designated and, thus, more predictable. This way, the

unknown variables of the evacuation simulation scenarios are limited and the process

can be modeled without many uncertainties. In practice, this can be achieved by

providing the proper amount of information during a fire emergency in a tunnel and by

guiding the evacuees to achieve a safe egress. More specifically, if the evacuees are

specifically and clearly guided to a designated evacuation plan, then the factor of the

unpredictable behavior is significantly reduced. As a result, the proper emergency signs

should be installed and the sound alarm message should be clear and specific. Then, the

evacuation process could be reproduced with the evacuation software in order to assess

the efficiency of the evacuation plan and the position of the signage and exits.

The Evac module of the FDS software has yet to be developed in order to provide more

reliable results. Also, the user interface should be more friendly and flexible.

Finally, the evacuation simulation module can be used in such a way that safe

conclusions can be drawn. It can be used in a more qualitative way in order to help the

designer understand the conditions in the tunnel and the point of view of the evacuees.

http://www.denhaag.nl/home/bewoners/to/Hubertustunnel.htm

http://www.cob.nl/

http://www.zuid-holland.nl/

http://ponce.sdsu.edu/cive395fall04group07.html

162 [70] G. C. Ian Longley, Gustavo Olivares, "Guidance for the Management of Air Quality in

road tunnels in New Zealand: NIWA Research Report for NZTA," New Zealand Transport Agency, Auckland, New Zealand2010.

[71] D. C. Carslaw, "Evidence of an increasing NO2/NOx emissions ratio from road traffic emissions," Atmos.Environ, vol. 39, pp. 4793-4802, 2005.

[72] R. B. Petr Pospisil, "Smoke Control in Road Tunnels," ed, 2004. [73] E. W. M. L. Camby M.K. Se, Alvin C.K. Lai, "Impact of location of jet fan on airflow

structure in tunnel fire," 2012. [74] L. J. C. Chen F., "Smoke flow phenomena and turbulence characteristics of tunnel fires,"

Applied Mathematical Modelling, vol. 35, pp. 4554-4566, 2011. [75] C. L. F. Hua L.H., Wu L., Li Y.F., Zhang J.Y., Meng N., "An experimental investigation and

correlation on buoyant gas temperature below ceiling in a slopping tunnel fire," Applied Thermal Engineering, vol. 51, 2013.

[76] C. W. K. Li S.M, "Numerical studies on performance evaluation of tunnel ventilation safety systems," Tunnelling and Underground Space Technology, vol. 18, pp. 435-452, 2003.

[77] "4th International Symposium on Tunnel Safety and Security," Frankfurt am Main, GermanyMarch 17-19 2010.

[78] A. Brinson, "European Fire Sprinkler Network: Active Fire Protection in Tunnels ", ed. London, 2010.

[79] S. H. Kevin McGrattan, Randall McDermott, Jason Floyd, Craig Weinschenk, Kristopher Overholt, "Fire Dynamics Simulator Technical Reference Guide Volume 3: Validation," National Institute of Standards and Technology, Gaithersburg, Maryland, USA2014.

[80] P. S.V., "Numerical Heat Transfer and Fluid flow," ed: Hemisphere Publishing, New York, 1980.

[81] C. W. K. Mok W.K., ""Verification and Validation" in modeling fire by Computational Fluid Dynamics," 2004.

[82] J. G. Nielsen, "Validation Study of Fire Dynamics Simulator," Master of Science, Department of Energy Technology, Aalborg University, Aalborg, 2013.

[83] J. W. Karl Fridolf, "Predictive Capabilities of Computer Models for Simulation of Tunnel Fires," ed.

[84] S. H. Kevin McGrattan, Randall McDermott, Jason Floyd, Craig Weinschenk, Kristopher Overholt, Fire Dynamics Simulator User’s Guide. Gaithersburg, Maryland, USA: National Institute of Standards and Technology, 2013.

[85] P. Smardz, "Validation of Fire Dynamics Simulator (FDS) for forced and natural convection flows," ed. University of Ulster, 2006.

[86] J. W. Kjos, "Validation and application of Fire Dynamics Simulator (FDS) in tunnel fires," Master of Science, School of the Built Environment, Ulster University, 2015.

[87] S. H. Kevin McGrattan, Randall McDermott, Jason Floyd, Craig Weinschenk, Kristopher Overholt, "Fire Dynamics Simulator Verification Guide," National Institute of Standards and Technology, Gaithersburg, Maryland, USA2014.

[88] A. L. H. Ingason, "Heat Release Rates from Heavy Goods Vehicle trailer fires in tunnels," Fire Safety Journal, vol. 40, pp. 646-668, 1995.

[89] A. L. H. Ingason, "Large-scale Fire tests in the Runehamar tunnel - Heat Release Rate (HRR)," in International Symposium on Catastrophic Tunnel Fires (CTF), Boras, Sweden, 2003, pp. 81-92.

[90] H. I. A. Lonnermark, "Gas temperatures in heavy goods vehicle fires in tunnels," Fire Safety Journal, vol. 40, pp. 506-527, 2005.

http://www.cbs.nl/

i

APPENDICES

ii

APPENDIX A: GENERAL INFO

iii

A.1 DEMOGRAPHIC DATA FOR THE NETHERLANDS

In the following table the demographic data in the Netherlands for the year 2009 are presented.

The demographics regarding the population per gender and per age were used for the evacuation

simulations. The input regarding the number of agents present in the tunnel was defined

according to calculations based on the data from this table.

A.2 TYPICAL PASSENGER CAR DIMENSIONS

In order to define the fire area in the fire simulations, the dimensions of a typical passenger car

had to be known. So, in the case of the fire simulations for the passenger car fire, the following

dimensions were used as represented in the image below:

A.1 Demographic Data for the Netherlands (source: www.cbs.nl)

A.2 Typical dimensions for a passenger car (source:

iv

APPENDIX B: TECHNOLOGICAL SYSTEMS BROCHURES

v

B.1 PHD6 MULTI-GAS DETECTOR – HONEYWELL

In order to gain an insight on the alarm thresholds used for various contaminants, a market

research was performed regarding the available gas detecting technologies. The Six-Gas

detector by Honeywell was one representative technology providing a lot of information in their

brochures. In the following pages, the specifications of the PHD6 detector are presented.

B.1 PHD6 multi-gas sensor by Honeywell (source: www.honeywell.com)

vi

B.2 PHD6 multi-gas sensor by Honeywell (source: www.honeywell.com)

vii

B.2 NOVENCO JET FANS

The ventilation system in the fire simulations had to be defined according to the ones available in

the market. NOVENCO provides a variety of solutions for ventilation in tunnels and these are

presented in the following brochure.

B.3 Jet fans brochure by NOVENCO (source: www.novenco-building.com)

viii


ix


x


xi


xii


xiii


xiv


xv


xvi


xvii


xviii

B.3 FIBROLASER SENSOR CABLES

The chosen Linear Heat Detection system that was examined in this research study was the

FibroLaser system by Siemens. The technical specifications for this system were provided by Mr.

Leo Knies in the form of brochures. Two different types of sensor cables are presented in those

brochures.

B.13 FibroLaser sensor cable specifications brochure (source: www.siemens.com)

xix


xx


xxi


xxii


xxiii


xxiv

B.4 SENTIO MULTIPLE GAS DETECTION (MGD) – FIREFLY AB

Another technology that was examined was the MGD technology. A supplier of this technology is

the company Firefly AB in Sweden. The Sentio multiple gas detector is presented in the following

pages.

B.19 Sentio multi-gas detector by Firefly ab (source: www.firefly.se)

xxv


xxvi


xxvii


xxviii

APPENDIX C: ETHANOL HRR DATA

xxix

C.1 EFECTIS NEDERLAND REPORT

A report by Efectis Nederland written by Ir. V.J.A. Meeussen, Ir. S.D. Nieuwendijk in 2010 was

provided by a fire specialist in FireSense, Mr. Paul Wendt. Tests with ethanol were performed in

order to determine the HRR from burning ethanol in different plate sizes. The results of the tests

are summarized in the following table:

C.2 US NUCLEAR REGULATORY COMISSION CALCULATION SHEET

The US Nuclear Regulatory Comision published a calculation sheet for the HRR resulting from

different fuels. The calculation for burning ethanol as performed for this research study is shown

below:

C.1 Efectis ethanol tests results (source: Efectis)

C.2 US Nuclear Regulatory Comission ethanol fire size calculation: Input data

xxx

C.3 US Nuclear Regulatory Comission ethanol fire size calculation: Results

xxxi

APPENDIX D: FDS GOVERNING EQUATIONS

xxxii

D.1 THE FOUNDAMETAL CONSERVATION EQUATIONS

The conservation equations for mass, momentum and energy for a Newtonian fluid that are used

by the FDS software, according to the Fire Dynamics Simulator (FDS) Technical Reference

Guide, are shortly presented below.

Conservation of Mass equation:

Conservation of Momentum (Newton’s 2nd

Law):

Conservation of Energy (1st Law of Thermodynamics):

Equation of state for a Perfect Gas:

xxxiii

APPENDIX E: FIRE SIMULATION SCENARIOS

xxxiv

ROAD TUNNEL FIRE SAFETY - TU Delft Repositories

Documents