UNIVERSITY OF CATANIA

DEPARTMENT OF CIVIL ENGINEERING AND ARCHITECTURE

PHD COURSE IN
EVALUATION AND MITIGATION OF URBAN AND LAND RISKS

XXXV CYCLE

A FRAMEWORK FOR RISK-BASED DECISION MAKING FOR

LOCAL RAILWAYS MANAGEMENT

Final Thesis

PHD CANDIDATE: TUTOR:

Vincenzo MARCHETTA Prof. Ing. Alessandro DI GRAZIANO

A.A. 2022/2023
 ABSTRACT
"Local" or "isolated" railway networks are defined as railway lines that are not connected to the
main network and have peculiar and unique characteristics, serving short/medium distance travel
needs and touristic and historical routes. Local railways play a fundamental role in a country's
transportation serving difficult-to-reach contexts where physical and economic constraints have
prevented the development of the conventional railway network. Given their unique environment
and the isolation from the main railway network, local railway networks have developed their own
management strategies, design characteristics and safety standards over time.
The demand for higher and aligned safety standards in railway services has become
increasingly important in recent years, frequently requiring significant investment in
infrastructure, equipment, maintenance and personnel training. In the context of local railways,
where economic and management resources are not comparable to those of national-level
networks this process can result in problems and constraints that managers are faced with. Failing
to comply with the required safety standards, in fact, leads to imposing unfavourable but
necessary safety measures, such as slowdowns and closures, which inevitably reduce the
attractiveness of this mode of transport and gradually lead to its closure. Furthermore, knowledge
and procedures conceived for interconnected railways cannot be transferred to local ones because
of the differences due to their history, unique infrastructural characteristics and management
methods.
Research on decision making tools and studies taking into account local railways'
peculiarities and data collected on these networks are still lacking. For this reason, the objective of
the thesis is to propose a quantitative, case-specific, and risk-based approach for decision support
in local railway safety management and to analyse the effect of local railways design
characteristics on risk analysis and, consequently, on management decisions.
To do so, a risk management framework is proposed to quantitatively estimate the impact
on the risk of system modifications and management choices, prioritizing and optimizing the
intervention strategies. The methodology includes the identification and characterization of
hazardous events, the analysis of accidental causes and consequences using FTA and ETA
methodologies respectively, the calculation of the risk level and the optimization of the decision-

i
making process in limited budget environments in order to improve the system and keep it within
the safety boundaries.
In order to broaden the accidental database, a methodology for the calibration of data of
similar systems through the use of expert judgments and quantitative analysis is presented and
commented. The methodology allows to indirectly assess frequencies and accidental
consequences by comparing the effectiveness of safety barriers of a case study and a reference
system. Finally, With the aim of guiding the choices of local railways managers to optimize
investments in the presence of limited budgets, the RM process is integrated with a Benefit-Cost
Analysis (BCA).
The effects of different infrastructure and rolling stock are analysed to understand the
effects of local railway design characteristics on the risk management process. Data harvested by
two high-precision monitoring campaigns on a local railway line are the basis for the track
geometry degradation evaluation and its influence on derailment risk. The effects of curvature,
slope, type of track, and loads on the degradation of Gauge, Longitudinal Level, Alignment and
Twist are investigated. Additionally, the effect of the rolling stock characteristics on the stopping
distance is analysed. In order to calibrate the commonly used empirical formula for the case of
local railways, a total of 25 braking tests are carried out with the trains of a narrow-gauge local
railway.
Finally, in order to provide a practical example of how the findings of this work can be
transferred to real-world scenarios, the Risk Management framework is introduced for the
utilization and decision support tool for analysing monitoring and managing the safety in the
tunnels and level crossings of the Ferrovia Circumetnea (FCE), a narrow-gauge railway that
connects several small towns on the slopes of Mount Etna, Sicily. The application made it possible
to verify the effectiveness of the framework in assessing the level of risk and prioritizing possible
interventions to control and reduce it.

ii
 ACKNOWLEDGEMENTS
At the conclusion of the research activity that led to the production of this work, I would like to
extend my heartfelt thanks to those who have directly and indirectly contributed to the
achievement of this great objective.
I would like to express my deep gratitude to my supervisor, Professor Alessandro Di
Graziano, for his enthusiasm from the beginning of my PhD journey to the completion of this
research. I sincerely appreciate his support and invaluable contribution to my growth as a
researcher, professional, and individual.
The numerous years of study at DICAR introduced me to numerous people that has
inspired valuable ideas and discussions on multiple occasions. I would like to thank all the people
of the Department and, in particular, Professors Cafiso and Inturri for their valuable advice.
I would like to thank all my colleagues and friends for their support, advice, and the
friendly days we spent together during these years. I am grateful to Martina, Giovanni, Roberto,
Nadia, Michela, Giulia, Eliana and Pietro for the time we spent working and having fun together
During the three years, I had the opportunity to participate in the collaboration between
Ferrovia Circumetnea and DICAR, and I would like to thank all those involved for their valuable and
enjoyable discussions. In particular, I would like to thank Eng. Fiore, Eng. Gentile, Eng. Contino,
and the entire staff of FCE with whom I have engaged in numerous activities. Within the
collaboration with FCE, I would also like to thank Eng. Marchese from MERMEC SpA for the
precious advices.
I express my gratitude to ANFISA and, in particular, Ing. Tatarelli and Ing. Marzoni for the
valuable information and teachings I have received during our numerous meetings.
My doctoral journey led me to spend three months at Lulea University of Technology. I
extend a warm thank you to Prof. Ahmadi for his time and dedication to me. I also appreciate the
colleagues and friends that this experience has provided me, especially Iman and Mahdi, for the
extensive and valuable discussions we had together. Your company and your help made the few
months spent together invaluable.
A thought goes out to my new colleagues at Italferr SpA, who motivated me to complete
this journey, especially the entire Safety, Risk & Data Analysis team, who, despite the short time,

iii
have provided me with important and valuable insights to enhance this final work. May our future
work together be as fulfilling as the months we have spent so far.
The achievement of a challenging goal like this is not only the result of technical and
academic support but also personal support. Therefore, I would like to thank my father, Annalisa,
Valeria, Virginia, my entire family, and all the friends who consciously and unconsciously fuelled
my determination towards this objective.
Lastly, my warmest thanks go to Martina, always ready to listen to my complaints and my
rambling thoughts, the shoulder that supported me in moments of discouragement and the
strength that pushed me to always do my best. After all, you have contributed to writing this
thesis as well.

iv
 LIST OF PAPERS
[1] Di Graziano*, A.; Marchetta, V.; (2021) A risk-based decision support system in local
railways management, Journal of Rail Transport Planning & Management, DOI:
10.1016/j.jrtpm.2021.100284

[2] Marchetta, V.*, Di Graziano, A., Soleimanmeigouni, I., & Ahmadi, A. (2022). Railway
degradation behaviour analysis in narrow-gauge railways: A local-railway case study.
Proceedings of the Institution of Mechanical Engineers, Part F: Journal of Rail and Rapid
Transit. DOI: 0.1177/09544097221136912.

[3] Marchetta, V.*, Di Graziano, A., & Contino, F. (2022). A methodology for introducing the
impact of risk analysis in local railways improvements decisions. In T. R. Procedia (Ed.), TIS
Roma 2022 Conference Proceedings. DOI: 10.1016/j.trpro.2023.02.191.

[4] Di Graziano, A; Marchetta, V.* (2023) A local railway application of a quantitative risk-
based decision tool for level crossing management; (Under Revision)

OTHER PAPERS:

[5] Di Graziano, A.*; Mancini, G.; Marchetta, V.; Spinelli, M. (2020) The use of checklists for
verifying the design of a railway infrastructure. Ingegneria Ferroviaria,75, 17–37

[6] Di Graziano, A.*; Marchetta, V.; Cafiso D.S. (2020); Structural health monitoring of asphalt
pavements using smart sensor networks: a comprehensive review. Jour of Traffic and
Transportation Engineering (Eng. Ed). DOI: 10.1016/j.jtte.2020.08.001

[7] Di Graziano, A.*; Marchetta, V.; Grande, J.; Fiore, S. (2021). Application of a decision-
making tool for the risk management system of a metro railway. International Journal of
Rail Transportation. DOI:10.1080/23248378.2021.1906341

v
 TABLE OF CONTENTS

Abstract ...................................................................................................................................... i
Acknowledgements ................................................................................................................... iii
List of Papers ............................................................................................................................. v
Table of Contents ...................................................................................................................... vi
List of Figures ............................................................................................................................ ix
List of Tables ............................................................................................................................. xi
List of Abbreviations ............................................................................................................... xiii
1. INTRODUCTION ...................................................................................................................... 1
1.1 Background ............................................................................................................................ 2
1.2 Problem statement................................................................................................................ 5
1.3 Research purpose and objectives ......................................................................................... 6
1.3.1 Purpose and Objectives ................................................................................................. 6
1.3.2 Research questions ........................................................................................................ 6
1.4 Structure of the thesis ........................................................................................................... 7
2. BACKGROUND ...................................................................................................................... 10
2.1 Introduction ......................................................................................................................... 11
2.2 The Local Railways ............................................................................................................... 13
2.2.1 Definition and characteristics of Local Railways .......................................................... 13
2.2.2 Local Railways Problems and Challenges ..................................................................... 17
2.3 Regulatory Framework ........................................................................................................ 19
2.3.1 European Railway Safety Regulations.......................................................................... 19
2.3.2 Italian Railway Safety Regulation ................................................................................. 28
2.3.3 Local Railway Safety Regulation................................................................................... 32
2.4 The Railway Risk Management ........................................................................................... 35
2.4.1 The concept of Risk ...................................................................................................... 35
2.4.2 The Risk Management Process .................................................................................... 37
2.5 Risk based DSS for safety improvement in railway systems ............................................... 59
2.6 Effect of infrastructure behaviour on Risk Management ................................................... 61
3. DEVELOPMENT OF A FRAMEWORK FOR RISK BASED DECISION MAKING IN LOCAL RAILWAYS 66
3.1 Introduction ......................................................................................................................... 67
3.2 Accidental analysis and identification of hazardous events ............................................... 69
3.2.1 Hazardous Events Identification .................................................................................. 69
3.3 Frequency analysis .............................................................................................................. 72

vi
 3.3.1 Causes of Derailment ................................................................................................... 72
3.3.2 Causes of Collision........................................................................................................ 76
3.3.3 Causes of Fire ............................................................................................................... 78
3.3.4 Causes of Level crossing accidents .............................................................................. 80
3.4 Consequence analysis ......................................................................................................... 83
3.4.1 Identification of accident scenarios and calculation of frequencies: Event Trees ...... 83
3.4.2 Lethality models ........................................................................................................... 91
3.5 The role of isolated railway features in risk assessment .................................................. 101
3.5.1 Accident data calibration ........................................................................................... 101
3.5.2 Effect of infrastructure features ................................................................................ 104
3.5.3 Effect of local railway rolling stocks and operation features on risk ......................... 107
3.5.4 Planning and resources optimization under budget constraint ................................ 114
4. RISK MANAGEMENT FRAMEWORK APPLICATION IN AN ITALIAN LOCAL RAILWAY............... 119
4.1 Introduction ....................................................................................................................... 120
4.2 Circumetnea Railway ......................................................................................................... 122
4.2.1 Story and characteristics ............................................................................................ 122
4.2.2 The line ....................................................................................................................... 123
4.2.3 The tunnels................................................................................................................. 124
4.2.4 The level crossings ..................................................................................................... 125
4.2.5 Reference accidental data: Italian Railway Network ................................................. 134
4.3 Risk Management for FCE tunnels .................................................................................... 139
4.3.1 System definition ....................................................................................................... 139
4.3.2 Hazard identification .................................................................................................. 140
4.3.3 Frequency analysis ..................................................................................................... 142
4.3.4 Severity of accident scenarios ................................................................................... 144
4.3.5 Calibration of frequencies and consequences ........................................................... 147
4.3.6 Risk calculation ........................................................................................................... 152
4.3.7 Risk evaluation ........................................................................................................... 153
4.3.8 Risk treatment and evaluation of mitigating actions................................................. 154
4.3.9 Risk based resources optimization ............................................................................ 157
4.4 Risk Management for FCE level crossings ......................................................................... 160
4.4.1 System definition ....................................................................................................... 160
4.4.2 Hazard identification and analysis of the causes ....................................................... 162
4.4.3 Consequences analysis............................................................................................... 168
4.4.4 Risk calculation ........................................................................................................... 171
4.4.5 Decision making support tool for risk treatment ...................................................... 173
4.5 Analysis of the role of characteristics of the local railways .............................................. 178
4.5.1 Track geometric quality degradation on narrow gauge local railways...................... 178
4.5.2 Experimental measures for the stopping distance .................................................... 195
5. CONCLUSIONS .................................................................................................................... 204
5.1 Introduction ....................................................................................................................... 205

vii
 5.2 A risk management framework for local railway decisions support ................................ 206
5.3 Effect of local railways design characteristics on risk ....................................................... 209
5.4 Case study validation......................................................................................................... 212
5.5 Future works...................................................................................................................... 214
Reference Regulatory Framework .......................................................................................... 216
European regulation ..................................................................................................................... 217
Italian regulation .......................................................................................................................... 219
References ............................................................................................................................. 221

viii
 LIST OF FIGURES
Figure 1.1. Organization of the topics of the thesis ............................................................................................................ 7
Figure 1.2. Structure of the thesis ....................................................................................................................................... 8
Figure 2.1. Topics of Chapter 2 ......................................................................................................................................... 11
Figure 2.2. Active Italian local railway networks .............................................................................................................. 15
Figure 2.3. Risk management process flow-chart ............................................................................................................. 40
Figure 2.4. FN curves for accidents happened in Europe between 1980 and 2009 (Evans, 2011) .................................... 52
Figure 2.5. Individual risk acceptance criterion. ............................................................................................................... 56
Figure 2.6. Cumulated risk acceptance criterion. .............................................................................................................. 56
Figure 3.1. Topics of Chapter 3 ......................................................................................................................................... 67
Figure 3.2. Derailment Fault Tree structure ...................................................................................................................... 74
Figure 3.3. Appendix T2 to the Faut Tree of the derailment relating to the causes related to track geometry defects ... 76
Figure 3.4. Collision Fault Tree structure .......................................................................................................................... 77
Figure 3.5. Fire Fault Tree structure .................................................................................................................................. 79
Figure 3.6. Fault tree of the Hazardous Event ‘Hazardous crossing of the LC’ ................................................................. 82
Figure 3.7. Derailment event tree and final accident scenarios ........................................................................................ 86
Figure 3.8. Collision event tree and final accident scenarios ............................................................................................ 87
Figure 3.9. Fire event tree and final accident scenarios .................................................................................................... 89
Figure 3.10. Hazardous cross at LC event tree and final accident scenarios .................................................................... 90
Figure 3.11. Lethality for train (a) and vehicle (b) passengers in function of train speed in Level Crossin accidents ....... 99
Figure 3.12. Risk management calibration framework .................................................................................................. 102
Figure 3.13. Representation of the vehicle as a material point for the study of motion ................................................ 109
Figure 3.14.Cost-benefit optimization problem. Adapted from (Špačková & Straub, 2015) .......................................... 117
Figure 4.1. Topics of Chapter 4 ....................................................................................................................................... 120
Figure 4.2. Railway and underground line and related stops and stations managed by FCE ......................................... 123
Figure 4.3. Data and information for RM process implementation ................................................................................ 130
Figure 4.4. Number of LC faults (N) and mean time to repair (MTTR) per part of the system........................................ 132
Figure 4.5. Scheme of the road network involved in the analysis with relative assignment of vehicular flows ............. 133
Figure 4.6. Derailment rate interpolating function ......................................................................................................... 136
Figure 4.7. Collision rate interpolating function. ............................................................................................................ 136
Figure 4.8. Fire rate interpolating function ..................................................................................................................... 137
Figure 4.9. Accident at level crossings rate interpolating function ................................................................................. 138
Figure 4.10. Incidence of causes in accident frequency .................................................................................................. 143
Figure 4.11. Event Trees structures for FCE tunnels ........................................................................................................ 144
Figure 4.12. Example of tunnel modeling (a) and fire simulation (b) ............................................................................. 151
Figure 4.13. Cumulative risk for accident category and global curve for Tunnel 1 ......................................................... 153
Figure 4.14. Cumulated risk levels for Tunnel 1, 2 and 3 ................................................................................................ 154
Figure 4.15. Effect of the application of all safety measures in Tunnel 1 (a), Tunnel 2 (b) and Tunnel 3 (c) .................. 156
Figure 4.16. LC distribution and classes along the studied line ...................................................................................... 162
Figure 4.17. Fault tree of the Hazardous Event ‘Hazardous crossing of the LC’ ............................................................. 164
Figure 4.18. Hazardous cross at LC event tree and final accident scenarios .................................................................. 168
Figure 4.19. Expected accident lethality trend vs train speed for the passengers of the railway vehicle (a) and the road
vehicle (b) ........................................................................................................................................................................ 170
Figure 4.20. Cumulative risk curves for all 96 analysed LCs ............................................................................................ 172
Figure 4.21. Individual Risk level for all 96 analysed LCs ................................................................................................ 172
Figure 4.22. Spatial distribution of the level crossings with reference to the identified risk level .................................. 173
Figure 4.23. Actual CR curve of LC#34 ............................................................................................................................ 174
Figure 4.24. CR curve of LC#34 with speed reduction ..................................................................................................... 175
Figure 4.25. CR curve of LC#34 with improved Sight Distance ........................................................................................ 176

ix
Figure 4.26. CR curve of LC#34 with protection equipment class upgrade ..................................................................... 176
Figure 4.27. CR curve of LC#34 with all mitigation measures implemented ................................................................... 177
Figure 4.28. Data preparation methodology .................................................................................................................. 179
Figure 4.29. Positional error between (a) peaks and corresponding trend (b) ............................................................... 181
Figure 4.30. Longitudinal level dataset (a) before and (b) after alignment process ....................................................... 182
Figure 4.31. Box plots and Tukey test Grouping of geometry indexes of (a) Gauge, (b) Longitudinal Level, (c) Alignment
and (d) Twist for the three classes of Curvature, Slope, Track Type and Number of Trains in Year 1 dataset ............... 189
Figure 4.32. Box plots and Tukey test Grouping of degradation rates of geometry indexes of (a) Gauge, (b) Longitudinal
Level, (c) Alignment and (d) Twist for the three classes of Curvature, Slope, Track Type and Number of Trains ........... 193
Figure 4.33. The railway vehicle used for the tests (a) and the driver’s cab with all the acquisition system installed. .. 196
Figure 4.34. GPS antenna inside the train control cabin (a) and the positioning of the analogical sensor to detect the
position of the brake control lever (b). ........................................................................................................................... 197
Figure 4.35. Dewesoft DEWE-43A system (a) and the Sirius ACC + system (b). .............................................................. 197
Figure 4.36. Parameters during one of the emergency braking tests at 50 km/h with the opening of the sandboxes
during all the test (a) and at the beginning (b) ............................................................................................................... 200
Figure 4.37 Curves resulting from the application of the parameters in the three models for calculating the braking
distance with reference to the fitting operation with test 1. .......................................................................................... 202
Figure 4.38. Comparison between Literature formula and local railway empirical adaptation for Pedeluq (a) and
Minden (b) formulas ....................................................................................................................................................... 203

x
 LIST OF TABLES
Table 2.1. Comparison of the extension of the Italian isolated and interconnected railway network.............................. 15
Table 2.2. Isolated narrow-gauge local railways lines in Italy .......................................................................................... 16
Table 2.3. Operational and design characteristics of isolated local railways (ANSF, 2019b) ........................................... 17
Table 2.4. Transposition of the European legislation on liberalization and security in Italy ............................................. 29
Table 2.5. Italian isolated railway networks and linens according to the Italian Ministerial Decree no. 347 of 2 August
2019 .................................................................................................................................................................................. 33
Table 2.6. Hazard Identification techniques ..................................................................................................................... 43
Table 2.7. Qualitative classes of Hazard frequency and consequences definition. ........................................................... 45
Table 2.8. Quantitative classes of consequences [FWI] comparison between EN and Italian Infrastructure Manager
regulation ......................................................................................................................................................................... 46
Table 2.9. Quantitative classes of frequency comparison between EN and literature values .......................................... 46
Table 2.10. Methods for Causal and Frequency Analysis .................................................................................................. 47
Table 2.11. Methods for Consequence Analysis ................................................................................................................ 49
Table 2.12. Risk matrix from EN 50126 ............................................................................................................................. 50
Table 2.13. Risk matrix for risk evaluation as reported by European legislation .............................................................. 55
Table 2.14. Risk matrix for risk evaluation as reported by Italian IM ............................................................................... 55
Table 3.1. Description and incidence of Derailment root causes ...................................................................................... 75
Table 3.2. Causes of derailment related to track geometry defects ................................................................................. 76
Table 3.3. Description and incidence of Collision root causes ........................................................................................... 78
Table 3.4. Main Fire Causes in UK railway network (RSSB, 2018) ..................................................................................... 78
Table 3.5. Main Fire Causes in Italian railway network .................................................................................................... 79
Table 3.6. Description and incidence of Fire root causes .................................................................................................. 80
Table 3.7. Models that can be used for the calculation of the wheel-rail friction coefficient. ........................................ 111
Table 3.8. Values of λ for various types of rolling stock and brakes(Profillidis V.A., 2014). ............................................ 112
Table 4.1. Features of the LC along the line .................................................................................................................... 125
Table 4.2. Data on traffic speed and visibility distance in LCs ........................................................................................ 128
Table 4.3. Significant accidents by type recorded in Italy between 2010 and 2020, edited from (ANSFISA, 2021) ........ 134
Table 4.4. System Characterization ................................................................................................................................ 140
Table 4.5. Hazard identification and characterization .................................................................................................... 141
Table 4.6. Hazard occurrence rates adapted at the tunnels analysed ............................................................................ 142
Table 4.7. Severity of the accident scenarios obtained for the three tunnels through the application of lethality models.
........................................................................................................................................................................................ 147
Table 4.8. Effect and impacts of system safety measures for frequencies and consequences calibration ..................... 148
Table 4.9. Calibrated frequency and consequences ........................................................................................................ 149
Table 4.10. Comparison between lethality obtained through empirical formulas and FDS simulation .......................... 151
Table 4.11. Total expected risk and individual risk for accident category ...................................................................... 152
Table 4.12. Total Risk level reduction ............................................................................................................................. 157
Table 4.13. Contributions and costs of IGs ...................................................................................................................... 158
Table 4.14. IG priority, fist iteration ................................................................................................................................ 159
Table 4.15. IG priority. Second and third iteration .......................................................................................................... 159
Table 4.16. LC safety protection classes ......................................................................................................................... 161
Table 4.17. Basic Events of the Hazardous Event ‘Hazardous crossing of the LC’ .......................................................... 164
Table 4.18. Factor of TESEO model estimation ............................................................................................................... 165
Table 4.19. Frequency of occurrence of equipment failures per LC ................................................................................ 166
Table 4.20. Class of curvature, slope, track and number of trains .................................................................................. 183
Table 4.21. Track geometry parameter and track geometry quality indicator............................................................... 184
Table 4.22. Means and STD of indexes distributions into the classes of characteristics and results of ANOVA and Tukey
test for Year 1 dataset for each quality index and track characteristic combination ..................................................... 186

xi
Table 4.23. Means and STD of indexes distributions into the classes of characteristics and results of ANOVA and Tukey
test for Year 2 dataset for each quality index and track characteristic combination ..................................................... 187
Table 4.24. Means and STD of indexes distributions into the classes of characteristics and results of ANOVA and Tukey
test for Degradation Rates.............................................................................................................................................. 191
Table 4.25. Braking tests characteristics ........................................................................................................................ 198
Table 4.26. Summary of the values obtained from the emergency braking tests at 50 km / h with the opening of the
sandboxes. ...................................................................................................................................................................... 201

xii
 LIST OF ABBREVIATIONS
AADT Average Annual Daily Traffic
AAID Average Amplitude of isolated defects
ABCL Automatic Barrier Crossing Locally monitored by train crew
ADT Average Daily Traffic
AL Alert Limit
ALARP As Low As Reasonability Possible
ANN Artificial Neural Network
ANOVA Analysis of Variance
ANSF Italian National Railways Safety Agency
ANSFISA Italian National Agency for Railways, Road and Highway Infrastructures Safety
ATP Active Train Protection
BCR Benefit cost Ratio
CBA Cost-Benefit Analysis
CBM Condition Based Maintenance
CR Cumulated Risk
CSI Commons Safety Index
CSM Common Safety Method
CST Common Safety Target
CTC Centralized Traffic Control
ERA European Railway Agency
ERTMS European Rail Traffic Management System
ETA Event Tree Analysis
EUAR European Union Agency for Railways
FCE Ferrovia Circumetnea
FDS Fire Dynamic simulator
FMEA/FMECA Failure Mode and Effect Analysis/Failure Mode, Effects, and Criticality Analysis
FRA Federal Railroad Administration
FSE Fire Safety Engineering
FTA Fault Tree Analysis
FWI Fatalities and Weighted Injuries
GAMAB Globalment Au Moins Aussi Bon

xiii
HAZOP HAZard and OPerability
HBD Hot Box Detector
HE Hazardous Event
IAL Immediate Action Limit
IG Interventions Group
IL Intervention Limit
IR Individual Risk
ISO International Organization for Standardization
LC Level Crossing
LRT Light Rail Transit
MCDA Multicriteria Decision Analysis
MEM Minimum Endogenous Mortality
MGT Million Gross Tons
MTTR Mean Time To Repair
OC Open Crossing
PT Passenger train
QRA Quantitative Risk Assessment
RB-DSS Risk Based - Decision Support System
RE Radical Event
RFI Rete Ferroviaria Italiana
RM Risk Management
RPN Risk Priority Number
RSSB Rail Safety and Standards Board
SMS Safety Management System
SPAD Signal passed at danger
STD standard deviation
UIC International Union of Railways
USTIF Italian Special Office for Stationary Transport
VSL Value of Statistical Life

xiv
1. INTRODUCTION
CHAPTER 1 Background

1.1 BACKGROUND
"Local" or "isolated" railways play a fundamental role in a country's transportation systems by
integrating the standard railway network, interconnecting difficult-to-reach context due to
physical and economic limitations.
In general, local railway networks are defined as railway lines that are not connected to the
main network and have distinct and unique characteristics, serving short to medium distance
travel needs. Many examples of such local networks exist around the world, particularly in Europe.
In Italy, several local railway networks developed during the last decades, comprising
approximately 1600 kilometres of track many of those with narrow gauges of 1000 mm and 950
mm. The use of narrow gauges enables these railways to overcome the challenging terrain and to
contain construction costs.
Given their unique environment and the separation from the main railway network, local
railway networks have developed their own management strategies, design characteristics and
safety standards over time. These may differ from those of the main network, as they reflect the
specific needs and characteristics of each local network.
The demand for higher and aligned safety standards in railway services has become
increasingly important in recent years due to the need to ensure the safety of passengers and
goods. This demand has driven improvements in the quality of railway services, such as improved
infrastructure, modernized rolling stock, and more advanced signalling and control systems. These
improvements have led to greater efficiency, faster and more reliable travel times, and increased
capacity for transporting people and goods.
However, achieving higher safety standards requires significant investment in
infrastructure, equipment, maintenance and personnel training.
This issue is even more evident in the context of local railways, where economic and
management resources are not comparable to those of national-level networks. Failing to comply
with the required safety standards leads to impose detrimental but necessary safety measures,
such as slowdowns and closures, which inevitably reduce the attractiveness of this mode of
transport and gradually lead to its closure.
Therefore, decisions regarding safety measures and strategies aimed at controlling risks
must be evaluated both in terms of their effectiveness and economic feasibility. Over the years,

2
CHAPTER 1 Background

the Risk Management (RM) process has emerged as a critical tool for railway organizations to
evaluate safety performance and plan future actions (Aven, 2016). The relevance of risk
management is also widely recognized at the legislative level. Directive (EU) 2016/798 (European
Union, 2016) highlights the importance of risk-based safety management approach and Regulation
(EU) n. 402/2013 (European Union, 2013) defines the methodology for risk assessment and
management. These directives have been implemented in each European country through the
European Union Agency for Railways (ERA) and National Agencies for Railway Safety.
RM is in general defined as ‘the identification, analysis, and prioritization of risks followed
by coordinated and economical application of resources to reduce, monitor, and control the
probability and/or impact of unfortunate events’ (Hubbard, 2020) with the aim of maintaining a
particular process within boundaries of safe operation (Rasmussen, 1997). By using this process,
railway organizations can identify the most significant risks, prioritize actions to address them, and
allocate resources more effectively, coping with safety standards with strategies that are both
effective and economically feasible.
Risk Management process consists of the following phases:

• System definition phase: delimitation of the system, identification of its interfaces, and
description of all technological, operational, and organizational elements capable of
influencing the level of system safety (European Union, 2013).
• Risk Assessment phase: includes the phases of Hazard Identification, Risk Analysis and Risk
Evaluation. Hazards Identification involves the definition of what can go wrong in a system,
identifying and characterizing all possible sources of risk. The Risk Analysis phase aims to
describe the current risk level estimating probability and severity of all possible accident
scenarios (An et al., 2011; Andrews & Dunnett, 2000). In the Risk Evaluation phase, are risk
indicators are compared with criteria and thresholds to assess whether the estimated risk
level is unacceptable, tolerable or acceptable (Rausand & Stein, 2020).
• Risk Treatment phase: on the basis of the results of risk assessment phase, decisions are
taken to control, reduce or monitor the risk level.
• Monitoring and control: risk factors are monitored to identify changes in the overall risk
level.

3
CHAPTER 1 Background

The first applications of RM methodologies can be traced back to the eighteenth and early
twentieth centuries in the fields of insurance and banking (Hubbard, 2020). Thanks to the
flexibility and potential of the implemented methodologies, RM has found applications in
numerous engineering branches over the years, from nuclear to oil and gas.
The use of RM as a decision support tool has also found wide application in the railway
sector (Sasidharan et al., 2017). Already in the early 2000s, Muttram (2002a) proposed a railway
safety risk model to provide a structured representation of the causes and consequences of
potential accidents through the combined use of fault tree analysis (FTA) and event tree analysis
(ETA) models. The framework lays the foundations for a case-specific quantitative assessment of
risk levels and describes an application for assessing the risk of derailment.
An et al. (2011) propose a RM system that incorporates the fuzzy reasoning approach (FRA)
and the analytical hierarchy process (AHP) for estimating the level of risk for each hazard by
combining qualitative and quantitative basic information. The model presented is then applied for
the assessment of the risk level of a railway depot.
In Berrado et al. (2011) a step forward has been made. In particular, a framework for RM
based on historical databases and the combined use of FTA and ETA is presented and its
applicability is described within social and economic assessments through Cost-Benefit Analysis
(CBA). The framework allows, starting from historical data of the studied system, to evaluate the
risk according to frequency classes and consequences within a risk matrix, thus providing discrete
and qualitative information on the level of safety starting from quantitative data.
The analysis of the applications presented in the literature highlights the central role
played by the data relating to faults and accidents of the applications of the RM process. In fact,
each state can rely on rich incidental databases (Evans, 2011; Lin et al., 2020; RSSB, 2018). This
information, if not collected through monitoring and data collection, can be estimated, for
example, through methodologies that rely on the use of expert judgments (Cooke & Goossens,
2004; Smithson, 2014).
In this regard, a methodology that takes into account the peculiarities of isolated railways
in assessing accident frequencies and consequences are still lacking. Furthermore, no studies have
been done to understand how the different infrastructure (gauge, gradients, curves, etc.) and
rolling stock (weight, speed, dimensions, etc.) of local railways can impact the level of risk and the
reliability of the decision-making process.

4
CHAPTER 1 Problem statement

1.2 PROBLEM STATEMENT

The local railways play an important role in short and medium-range transportation as well as in
historical and touristic contexts. In recent years, the new regulatory framework requires these
networks to align both in terms of safety and management with the national network. Achieving
this alignment requires considerable investment and management efforts, which can be
impractical.
The new regulatory approach, in fact, requires that the management of a railway system is
based on the satisfaction of safety target and standards. These standards are well-established for
interconnected railways but not yet for local railways. One obstacle to the application of existing
standards in these networks is that it is not possible to transfer the knowledge and procedures
developed in the standard railway network to local ones because of the differences due to their
history, unique infrastructural characteristics and management methods.
The decision-making process underlying management and improvement must be able to
take these differences into account to optimize resources and safety level. This can be
accomplished by using specific tools based on local railways' peculiarities and specific data. Such
an approach will enable stakeholders to make informed decisions while minimizing the potential
negative impacts of changes.
Risk management (RM) is a powerful tool able to evaluate managerial choices, the safety of
the system and the feasibility of investments in relation to safety standards. While RM has been
applied in the context of interconnected railways, it is still lacking a framework to support the
decision-making in local railways.
Additionally, RM rely on detailed data on system accidental history and on the information
on the performance in terms of safety of its elements. The greater the knowledge of the analysed
system, the more reliable will be the analyses and forecasts. To date, there is a lack of in-depth
study on the effects of the unique infrastructural and rolling stock characteristics typical of local
railways on the system's risk level.
Therefore, the research work behind this thesis aim to propose a quantitative, case-
specific, and risk-based approach for decision support in local railway safety management.
Additionally, it aims to analyses the effect of local railways design characteristics on risk analysis
and, consequently, on management decisions.

5
CHAPTER 1 Research purpose and objectives

1.3 RESEARCH PURPOSE AND OBJECTIVES

In order to provide a guide for the correct reading of this work, the understanding of the purpose,
the objectives and the questions that guided the research activity is crucial.

1.3.1 Purpose and Objectives

The purpose of this work is to support the development of safety and competitiveness of local
railways through the definition of a risk-based decision support tool that takes into account the
peculiarities of this type of railway networks.
In order to pursue this purpose, the following objectives have been set and achieved:

I. To propose an approach quantitative, case-specific and risk-based for decision support in

local railway safety management;
II. To analyse the effect of local railways design characteristics on risk analysis and,
therefore, on management decisions.

1.3.2 Research questions

In order to fulfil the research objective, the following questions guided the research work:
1. How to optimize investments in local railways with the aim of aligning them with the
safety standards and management strategies of interconnected railways?
2. How to develop a quantitative risk management framework in the presence of a limited
accident history and constrained resources?
3. What is the effect of the design characteristics of local railways on the probability and
severity of railway accidents?

6
CHAPTER 1 Structure of the thesis

1.4 STRUCTURE OF THE THESIS

In order to comprehend the organization of this work, it is necessary to define the underlying logic
of the research. As shown in Figure 1.1, this work develops along two main dimensions that define
its structure in chapters and subchapters. The first "vertical" dimension identifies the three central
chapters of the thesis and is represented by the level of detail of the topics discussed. It begins
with an extensive exploration of the literature and regulations supporting the research, followed
by a focused examination of a quantitative methodology for risk management to aid decision-
making. Lastly, the framework proposed is applied to a case study.
On the other hand, the "horizontal" dimension repeats within each chapter and defines the
structure of the subchapters. It is characterized by the interplay between "Definition of a Risk
Management Methodology" and "Role of Local Railway Characteristics".

Figure 1.1. Organization of the topics of the thesis

With the aim of providing all the information necessary to understand the motivations,
steps and results that characterized the research work, the organization of the topics discussed
above guided the structure of the thesis as shown in Figure 1.2.

7
CHAPTER 1 Structure of the thesis

Figure 1.2. Structure of the thesis

In particular:

• Chapter 1: Introduction – This chapter presents a brief background on the topic and the
motivations that guided the research work. In particular, the isolated railways are
presented and their characteristics and the main management problems described. Risk
Management process as a support for the solution of these problems is than presented.
The objectives and questions of this research are then described and outlined.
• Chapter 2: Background - This chapter aims to describe in depth the characteristics of local
railways, provide the complete regulatory framework in which the topics covered are
placed and describe in depth the state of the art regarding Risk Management and its role as
a decision support tool for infrastructure managers.

8
CHAPTER 1 Structure of the thesis

• Chapter 3: Development of a framework for risk-based decision making in local railways -

This chapter focuses on describing the risk assessment methodology for local railways. The
methodology includes identifying and characterizing hazardous events, analysing incidental
causes and accident consequences using FTA and ETA methodologies respectively,
calibrating models based on the characteristics of local railways, defining the role of local
railways infrastructure and rolling stock in risk management, and optimizing decision-
making in limited budget environments based on risk analysis outcomes.

• Chapter 4: Risk Management framework application in an Italian local railway - This

chapter presents the application of the Risk Management framework to the Italian local
railway ‘Ferrovia Circumetnea’. The railway's history, physical characteristics, and
operational aspects are described, and the application of the presented framework is
described in the case of tunnel al level crossing safety management. The chapter then
investigates, starting from field-measured data, the effect of the different characteristics of
the infrastructure and rolling stock on risk.

• Chapter 5: Conclusions – This chapter summarizes the main results, contributions and
considerations obtained through this work as well as some suggestion for future works.

9
2. BACKGROUND
CHAPTER 2 Introduction

2.1 INTRODUCTION

CHAPTE 2
ocal railways de ni on problems and
iterature review Railway safety regula on . .
challenges .
regulatory RM process and methodologies .
ocal railway safety regula on . .
framewor RM ad decision support tool .

CHAPTE 3
Role of available accident data
Role of infrastructure
Methodology e ni on of the RM methodology
Role of Rolling toc
Role of limited resources

CHAPTE 4
ata calibra on
nalysis of infrastructure related ris s
Case study RM applica on Ferrovia Circumetnea
nalysis of vehicles related ris s
nterven on op mi a on

Figure 2.1. Topics of Chapter 2

National railway networks are made up of various types of railway lines, including local isolated
railways. Local isolated railways are defined as railway lines that are not connected to the main
network and with different and peculiar characteristics.
Due to the isolated nature, these railways developed their own management techniques
and safety standards over time. The differences in resources, environments and needs between
isolated and standard railways lead to unique track geometry, rolling stock, and operating
characteristics.
Transferring knowledge developed in the context of interconnected railway networks is not
always possible and can produce unexpected results, obtaining non-optimal management
strategies.
To overcome these challenges, specific knowledge of the level of risk of the line and the
actual effects of system changes is required. This knowledge can be obtained through the use of
tools and models related to railway risk, which will be further discussed later in this chapter.
In order to understand these models, a complete view of the regulatory framework on
railway safety and risk management in Europe and Italy is required. This will provide a foundation
for understanding the concept of risk in railway and the risk management process aimed at
evaluating, characterizing, and controlling it. Risk management can be used as a decision support

11
CHAPTER 2 Introduction

tool for introducing safety measures and optimizing maintenance in local isolated railway
networks, ensuring the safe and efficient operation of these railways and reduce the risk of
accidents.
In the following sections, are provided a definition of local isolated railways (Section 2.2); a
detailed discussion of the regulatory framework on railway safety (Section 2.3), the concept of risk
in railway and the risk management process (Section 2.4), and, finally, the use of risk management
as a decision support tool for safety measures introduction (Section 2.5) and maintenance
optimization (Section 2.6)

12
CHAPTER 2 The Local Railways

2.2 THE LOCAL RAILWAYS

A country's railway network comprises various types of networks that differ in hierarchy and
scope. Along with the main interconnected network, the railway transportation demand can be
met by integrating “isolated” or “local” railway networks.
These local networks emerged between the late 19th and early 20th century in response to
the need to connect locations in mountainous or inaccessible areas at a lower cost. Despite their
age, many of these networks are still in operation today and provide crucial services, including:

• Passenger services on short-medium-haul routes competing with road vehicles. Due to

their small size and track design flexibility, local railways integrate better than standard
rails into the urban context and provide a means of connection with suburban locations;
• Connection of mountainous and inaccessible areas overcoming otherwise inaccessible
routes;
• Touristic routes and museum trains. Local railways have a long history and have frequently
been supplanted by road transport. The lines that were born in suggestive environments or
with historical significance are still used for tourist purposes.

Of particular interest in this work are the lines that primarily provide passenger services. In
the following paragraphs is presented the history of local railways worldwide and in Italy and the
main issues regarding maintenance and safety management in these networks.

2.2.1 Definition and characteristics of Local Railways

The definition of a local railway varies from state to state. The term “local network” is often used
to identify the regional network and all lines dedicated to local-scale routes, including
interconnected networks owned by national infrastructure managers.
In this work, however, the term "local network" refers to a specific type of rail transport
network that is functionally separate from the national network and has different historical,
operational, and infrastructural characteristics from the interconnected networks.
Their function in the transportation network differs from that of regional or commuter
rails. Regional lines run between cities at medium-low frequencies, serving all (intercity) or part of

13
CHAPTER 2 The Local Railways

the stations on their route. Commuter rails, on the other hand, are mostly used in urban areas
with frequent trips.
Local networks are used connect large cities with sites in difficult-to-reach contexts, such
as mountain villages where the paths are tortuous and uneven, and where the demand for
transportation and available resources are insufficient to support significant engineering works to
overcome high gradients and natural obstacles. Furthermore, the small size of the infrastructure
and the reduced civil works allows the lines to fit into the urban context serving also as commuter
rail.
Local isolated narrow-gauge railway lines developed mainly between the end of the 1800s
and the first half of the 1900s, since the second half of the XX century due to increased
road vehicles competitiveness.
There are numerous examples in the world and Europe. Germany, for example, has several
forest and mountain railway networks that feature narrow gauges and winding tracks. In Austria,
lines with track widths between 760 mm and 1000 mm play mainly tourist and tram roles. Or
Switzerland, because of its predominantly mountainous terrain, has numerous narrow-gauge
networks often characterized by rack railway sections.
In Italy, the national railway safety agency (ANSFISA) surveyed 13 operating isolated local
railways, managed by 12 administrative regions (ANSF, 2019a) which are responsible for planning
and administration tasks.
As in the rest of Europe, the development of isolated local railways, especially narrow-
gauge networks, can be found in Italy at the end of the nineteenth century. Indeed, the Law of 29
July 1879 n. 5002, also known as the Beccarini Law, authorizes the construction of local and
secondary narrow-gauge lines in order to accelerate the completion of the national railway
network and containing construction costs. The majority of the networks built in accordance with
this law used ‘Italian narrow-gauge' equal to 950 mm . To keep costs down, these nets made use
of light rails, winding paths, and continuous variations in gradient.
Local networks which are still active today were born during the following decades.
Although many local networks lost their attractiveness and were decommissioned over time, now
these account for 5% of the total national railway network and are equally distributed all over the
Italian territory as shown in Figure 2.2.

14
CHAPTER 2 The Local Railways

Figure 2.2. Active Italian local railway networks

Table 2.1 compares the lengths in kilometres of the local networks the interconnected
network managed by the Italian Infrastructure Manager 'Rete Ferroviaria Italiana' (RFI).

Table 2.1. Comparison of the extension of the Italian isolated and interconnected railway network
Type km %
Isolated Local Railways* 1607 5%
RFI Network** 16832 48%
Fundamental lines** 6486 18%
Complementary lines** 9396 27%
Node lines** 950 3%
*(ANSF, 2019a)
**(RFI, 2022)

Table 2.2 details the Italian active isolated local lines, including infrastructure managers,
administrative regions of competence, lengths, and relative gauges.
Isolated local networks use mainly three gauges: 1435 mm, 1000 mm, and 950 mm. The
Italian narrow gauge (1000 mm), also called “metric gauge” accounts for 72% of the total 1600 km
of lines. The 950 mm is due to the practice of measure of the metric nominal gauge between the

15
CHAPTER 2 The Local Railways

rails' axes. The 1000 mm gauge networks are fewer in number and mostly concentrated in
northern Italy, in areas influenced by the German system of measuring the nominal gauge in the
inner edge of the rails heads (Federici, 1999).

Table 2.2. Isolated narrow-gauge local railways lines in Italy

Infrastructure Administrative Length

Line Gauge
Manager Region (km)

AMT Liguria Ferrovia Genova Casella 24.3 1000

Brescia-Iseo-Edolo and Bornato -
Ferrovienord Lombardia 108.4 1435
Rovato
GTT S.p.A. Piemonte Torino - Ceres 42.2 1435
SSIF S.p.A. Piemonte Domodossola - Confine Svizzero 32.0 1000
ASTRAL S.p.A. Lazio Roma - Lido 28.4 1435
ASTRAL S.p.A. Lazio Roma - Viterbo 101.9 1435
ATAC S.p.A. Lazio Roma - Giardinetti 5.4 950
Ferrovie della
Calabria Cosenza Vagliolise - Catanzaro Lido 115.1 950
Calabria S.r.l.
Ferrovia
Sicilia Catania Borgo - Riposto 110.0 950
Circumetnea
FAL S.r.l. Puglia-Basilicata Bari - Matera (Altamura - Gravina) 86.3 950
FAL S.r.l. Puglia-Basilicata Altamura - Avigliano Lucania 73.6 950
FAL S.r.l. Puglia-Basilicata Avigliano Città - Potenza Inf. Scalo 7.7 950
FAL S.r.l. Puglia-Basilicata Avigliano Città - Potenza Inf. Scalo 5.3 950
ARST S.p.A. Sardegna Monserrato - Isili 71.1 950
ARST S.p.A. Sardegna Mandas - Arbatax 159.0 950
ARST S.p.A. Sardegna Isili - Sorgono 83.0 950
ARST S.p.A. Sardegna Macomer - Nuoro 57.8 950
ARST S.p.A. Sardegna Macomer - Bosa 46.0 950
ARST S.p.A. Sardegna Sassari - Alghero 30.1 950
ARST S.p.A. Sardegna Sassari - Sorso 9.9 950
ARST S.p.A. Sardegna Sassari - Palau 150.7 950
EAV Campania Cumana 19.8 1435
EAV Campania Circumflegrea 27.0 1435
EAV Campania Circumvesuviana Napoli - Sorrento 42.4 950

16
CHAPTER 2 The Local Railways

Infrastructure Administrative Length

Line Gauge
Manager Region (km)

Circumvesuviana Napoli -
EAV Campania 14.3 950
Poggiomarino
Circumvesuviana Napoli - Ottaviano -
EAV Campania 34.0 950
Sarno
EAV Campania Circumvesuviana Napoli - Baiano 40.2 950
EAV Campania Circumvesuviana Napoli - Acerra 5.6 950
EAV Campania Circumvesuviana Napoli - San Giorgio 6.2 950
Trentino-Alto
Trentino Trasporti Trento-Malè-Mezzana 66.0 1000
Adige
Lombardia-
Rhaetian Railway Tirano-Campocologno 3.0 1000
Swizerlan

Over time, local railways have developed unique track design parameters due to their
distinct history and objectives, resulting in significant deviations from standard railways. Table 2.6
provides a comparison of key design parameters used in the construction of interconnected
railways and isolated narrow-gauge local railways.

Table 2.3. Operational and design characteristics of isolated local railways (ANSF, 2019b)
Parameters Italian interoperable railways Italian local railways
Max Speed 300 km/h 50-120km/h
Max gradient 35 mm/m 40 mm/m
Min horizontal curvature radius 150m 80 m
Gauge 1435mm 1000mm/950mm
Cant 160mm 110mm
Cant deficiency 153mm 86mm

2.2.2 Local Railways Problems and Challenges

Due to differences in resources and safety standards, isolated railways have developed their own
management techniques and infrastructure characteristics over time. However, in the last few
years a growing need for alignment with national railways standards is driving the management
approach in local railways.

17
CHAPTER 2 The Local Railways

Until today, safety systems and technologies available in local railways are usually less
complex than regional or high-speed networks. One example are the requirements for signalling
technologies systems that can be very different between 300 km/h high speed lines and local lines
where in some cases maximum allowed speed is near to 50 km/h.
Furthermore, the investments required to align design and management standards across
all railway networks may be unsustainable by local operators, intensifying preventive operational
limitations like speed reductions and line closures. In this way, the local railway service loses
competitiveness in favour of other modes of transportation (private car, road public transport,
etc.), potentially heralding the beginning of the end of these railways.
The closure of local networks not only represents the loss of historical and tourist routes
but causes an increase in passengers on the road with all the risks related to pollution and
exposure to the typical hazards of this mode of transport.
In addition, isolated railways often have track geometry, rolling stock, and operating
procedures that differ significantly from those of interconnected railways. Therefore, transferring
knowledge and solutions developed in the context of interconnected railways to isolated networks
may lead to unpredictable outcomes, and adopting the same solutions used in traditional railways
may result in suboptimal management strategies.
Additionally, the risks associated with local networks are heightened when taking into
account that less than half of the traffic on Italian isolated lines is equipped with active train
control systems. Furthermore, the level crossing density along the line is particularly high
compared to other lines, with an average of one level crossing per kilometre (ANSF, 2019a).
To address these issues, a solution can be found through a detailed understanding of the
risk level of the railway network and the potential effects of system changes. Therefore, this work
proposes a risk-based decision-making tool that can evaluate the actual risk level of the system,
estimate the impact of any modifications on risk, and determine the optimal combination of
interventions and their priority based on their costs and benefits to the system.

18
CHAPTER 2 Regulatory Framework

2.3 REGULATORY FRAMEWORK

Ensuring railway safety requires the proper implementation of regulations, guidelines, and
operating procedures before, during, and after every train movement. Before developing tools
and models for railway risk, it is crucial to have a comprehensive understanding of the regulatory
framework for railway safety. Specifically, discussing railway safety requires contextualizing the
European regulatory framework, which began with the initial efforts towards the liberalization of
the railway sector, and how this framework affects the management of isolated railways in the
national context today. Then its reflexes in Italian regulation and local railways regulatory context
are discussed.

2.3.1 European Railway Safety Regulations

The liberalization process initiated by the European Community in 1991 has brought significant
changes in the conception and management of the railway sector. This process was triggered by
several factors, one of which is the decline of the railway market share compared to other modes
of transportation in the last few decades of the 1990s (Berrado et al., 2011). The fragmentation
due to non-interoperable networks, the loss of efficiency linked to the monopoly approach of the
Member States, has caused the loss of competitiveness of the railway mode in the short/medium-
haul trips in favour of the road vehicle and in the long-haul trips, i.e. over 200-300 km, in favour of
the air transport.
The improvement of security levels is another crucial reason behind the political project of
railway liberalization in Europe (Acquaro, 2019). In addition to the efforts made by the European
legislator to liberalize the railway market, there has also been significant work on the safety front,
promoting technical standardization and common methods for managing it in order to improve
the existing security levels in the various member states.
The initial modifications enforced by the European regulations aimed to eliminate railway
operators acting solely on a national level. This change was necessary because these operators
employed different methodologies, pursued varying objectives, and adhered to different technical
standards and regulations between state and state. For this reason, the European Union projected
a unified European space that featured an integrated infrastructure network, equipped with

19
CHAPTER 2 Regulatory Framework

interoperable equipment. This resulted in seamless transportation services across European

borders, ensuring uninterrupted connectivity.
The initial move towards standardization across the EU was initiated with the issuance of
the Directive on the liberalization of rail transport in Europe (Directive 91/440 / EEC), which aimed
to achieve several goals, including the separation of the railway system from state control, the
division between Railway Undertakings (RUs) responsible for providing rail transport services, and
Infrastructure Managers (IMs) responsible for establishing and maintaining railway infrastructure,
and the right of access to the infrastructure across all European countries for the provision of
freight and passenger transport services.
On the basis of what was imposed in '91, further directives (Directive 95/18/EC, Directive
95/19/EC, Directive 96/48/EC) have provided additions and improvements from the point of view
of free circulation, economic recovery and in the homogenization of safety levels.
The 'Infrastructure package', also known as the 'First railway package', marked a
significant milestone in the liberalization process. This package was the first of a series of four
successive packages, issued in 2001, and aimed to ensure equal and non-discriminatory access to
the railway network and its optimal use. The regulations included in this package were essential in
creating a competitive market for railway transport services by establishing fair competition
among railway undertakings and promoting the development of international rail transport.
The following are part of the first railway package:

• Directive 2001/12/EC of 26 February 2001 concerning the development of the Community

railways;
• Directive 2001/13/EC of 26 February 2001 concerning the licensing of Railway
Undertakings;
• Directive 2001/14/EC of February 26, 2001, concerning the allocation of railway
infrastructure capacity, the imposition of rights for the use of the railway infrastructure
and safety certification;
• Directive 2001/16/EC of 26 February 2001 concerning the interoperability of the
conventional trans-European system.

20
CHAPTER 2 Regulatory Framework

Recently, the directives that make up the package have been revised into Directive
2012/34/EC, improving access to rail services and strengthening the independence of national
regulatory bodies.
The 'Second railway package' marked an important milestone in the European railway
sector, introducing the first directives on railway safety. It established the European Railway
Agency (ERA) and granted the right of access to the entire Community railway network for all
types of international freight transport. The package also introduced critical measures concerning
the safety and interoperability of the entire trans-European rail system, including high-speed and
conventional railways. The framework laid out in the package allowed for the liberalization of
freight transport services, with each Railway Undertaking being recognized as an independent
operator:

• the access to the entire European railway network for the provision of all types of
international freight transport services, starting from January 1, 2006 (thus ahead of the
deadline of March 15, 2008 provided for by Directive 2001/12 /THERE IS);
• the right of access to the infrastructure in all Member States for the provision of all types
of freight transport services (not only international transport, therefore, but also national
and cabotage transport), starting from 1 January 2007.

The regulatory measures that are part of the second railway package are:

• Directive 2004/49/EC of 29 April 2004 relating to the safety of the Community railways
(Safety Directive);
• Directive 2004/50/EC of 20 April 2004 on the interoperability of the trans-European rail
system;
• Directive 2004/51/EC of 29 April 2004 relating to the development of the Community's
railways;
• the EC Regulation 881/2004 of 29 April 2004 establishing the European Railway Agency
(ERA).

Of these, the Safety Directive (Directive 2004/49/EC) played a crucial role in the European
railway safety regulatory framework and has paved the way for several subsequent legislative
measures. The directive mandates the maintenance or improvement of high safety standards in
the rail system, aiming to prevent scenarios where railway operators prioritize profit-related

21
CHAPTER 2 Regulatory Framework

objectives over safety. By doing so, the directive promotes a safe and secure railway transport
system for passengers and freight across the European Union.
The pursuit of these objectives is linked to:

• the harmonization of the regulatory structure;

• the definition of the responsibilities between the actors of the railway system;
• the development of Common Safety Targets (CST) and Common Safety Methods (CSM) to
harmonize the national rules;
• the establishment, in every Member State, of a safety authority (National Safety Agency -
NSA) and an accident and incident investigating body;
• the definition of common principles for the management, regulation and supervision of
railway safety.

The effective management of railway safety relies heavily on the implementation of

Common Safety Targets (CST) and Common Safety Methods (CSM). The CSTs represent the
minimum safety levels that must be achieved by all parts of the railway system and are established
by the European Railway Agency (ERA) based on statistical analysis of historical data on personal
injury. The CSTs consist of target risk values that are considered tolerable for the exposed
population. The CSMs, on the other hand, define the methods for assessing the safety levels
identified by the CSTs. The legislator has defined CSMs for various aspects, such as safety
certification, risk assessment, verification of compliance of Safety Management Systems (SMS),
monitoring, and supervision. Overall, the implementation of CSTs and CSMs plays a crucial role in
maintaining and improving the safety standards of the rail system.
Directive 2004/50/EC had the task of defining the essential requirements that the system
must meet in terms of safety, reliability, availability, health, environment and technical
compatibility and introduced the regulatory instrument of the 'Technical Specification of
Interoperability '(TSI), identifying the technical standards for the subsystems:

• Infrastructure,
• Energy;
• Maintenance;
• Control, command and signalling;
• Rolling Stock;

22
CHAPTER 2 Regulatory Framework

• Operation and traffic management;

• Telematics applications for passengers and freight transport.

The 'Third railway package', approved on 23 October 2007, intended to create an

integrated European railway area, with the aim of making rail transport more competitive and
attractive to users.
The package consists of two directives:

• Directive 2007/58/EC on the allocation of railway infrastructure capacity and the

imposition of rights for the use of the railway infrastructure.
• Directive 2007/59/EC on the certification of skills and responsibilities of train drivers;

Directive 2007/58/EC, in amending Directives 91/440/EEC and 2001/14/EEC, introduces

important innovations in terms of opening the market for international passenger transport rail
services within the Community. In fact, licensed and safety certified railway companies are
granted the right to access within all Member States for the operation of international passenger
transport services.
In the years between the third and fourth railway package, a series of directives and
regulations are then issued to carry out the liberalization process and define adequate safety
standards:

• Directive 2008/57/EC of 17 June 2008 and Directive 2009/131/EC of 16 October 2009, both
relating to the interoperability of the railway system;
• Directive 2008/110/EC of 16 December 2008, relating to the safety of Community railways:
it introduces the principle that keepers of freight wagons are no longer subject to the
obligation to register the wagons with an RU and are responsible for maintenance of the
wagons themselves;
• Directive 2012/34/EU of 21 November 2012 establishing a single European railway area
(recast) and repealing Directives 91/440 / EEC, 95/18 / EC and 2001/14 / EC;
• Regulation (EU) 1158/2010 of 9 December 2010 relating to a common security method for
assessing compliance with the requirements for obtaining safety certificates;
• Regulation (EU) 1169/2010 of November 16, 2012 on a common safety method for
supervision by the national safety authorities after the issue of a safety certificate or a
safety authorization;

23
CHAPTER 2 Regulatory Framework

• Regulation (EU) 445/2011 of 10 May 2011 relating to a certification system for persons
responsible for the maintenance of freight wagons;
• Regulation (EU) 1077/2012 of November 16, 2012 on a common safety method for
supervision by the national safety authorities after the issue of a safety certificate or a
safety authorization;
• Regulation (EU) 1078/2012 of 16 November 2012 on a common safety method for
monitoring that railway undertakings, infrastructure managers who have obtained a
certificate or a safety authorization and entities in charge of maintenance must apply;
• Regulation (EU) 402/2013 of 30 April 2013 relating to the common safety method for the
determination and assessment of risks and which repeals Regulation (EC) no. 352/2009;
• Regulation (EU) 1136/2015 of 13 July 2015 amending Regulation (EU) 402/2013 relating to
the common security method for the determination and assessment of risks;
• Regulation (EU) 995/2015 relating to the technical specification for interoperability
concerning the "Operation and traffic management" subsystem of the railway system in
the European Union (TSI OPE).

Of these, Regulation (EU) 402/2013 and the subsequent amendments introduced by

Regulation (EU) 1136/2015 are crucial for the risk assessment and management process. The
Regulation, in fact, on the basis of what was introduced by Directive 2004/49/EC of the Second
railway package, establishes the Common Safety Method (CSM) for the determination and
assessment of risks and harmonization of risk management procedures, the exchange of security
information and control of the application of the CSM.
The Regulation plays a central role in defining the risk assessment process in the event of
significant changes to the railway system. It establishes the criteria for identifying changes that can
affect the security of the system, describes the risk assessment process in all its phases (system
definition, hazard identification, risk analysis, risk evaluation) and defines the risk acceptance
criteria (application of code of practice, comparison with similar systems, explicit risk estimation).
In order to make the regulatory action on safety and interoperability more effective, the
European Commission issued the Fourth railway package on January 30, 2013.

24
CHAPTER 2 Regulatory Framework

The 'Fourth railway package' was necessary because, despite the high effort made up to
that moment, the railway network was still fragmented (Acquaro, 2019), with safety standards and
technical systems still different from state to state.
In order to find a solution to these problems, the fourth railway package proposes an
integrated approach aimed at revitalizing EU rail transport to foster the creation of a single
European railway area containing measures aimed at increasing the modal share of rail transport.
The main objectives of the fourth railway package are:

• The reduction of administrative costs for railway companies and facilitating the entry into
the market of new operators (the European Railway Agency - EUAR - becomes the only
place for issuing authorizations for vehicles and safety certificates for operators);
• The strengthening of the role of the Infrastructure Managers, guaranteeing their total
operational and financial independence from the railway operators;
• The opening of national passenger rail networks to new operators and services from
December 2019. Companies will be able to offer competitive services, such as new rail
services on a particular route, or to win public service contracts in the railway sector
through tenders. The proposed changes make competitive bidding procedures mandatory
for public service contracts in the rail sector in the EU;

In particular, the Fourth railway package consists of a Market Pillar and a Technical Pillar.
The Market Pillar includes:

• Directive (EU) 2016/2370 of 14 December 2016, amending Directive 2012/34 / EU as

regards the opening of the market for national rail passenger transport services and the
governance of the railway infrastructure;
• Regulation (EU) 2016/2338 of 14 December 2016 amending Regulation (EC) no. 1370/2007
relating to the opening of the market for national rail passenger transport services;
• Regulation (EU) 2016/2337 of 14 December 2016 which repeals Regulation (EEC) no.
1192/69 of the Council concerning the common rules for the normalization of the accounts
of railway companies.

The Technical Pillar, on the other hand, includes:

25
CHAPTER 2 Regulatory Framework

• Regulation (EU) 2016/796 of 11 May 2016 establishing the EUAR (European Union Agency
for Railways) and repealing Regulation (EC) no. 881/2004;
• Directive (EU) 2016/797 of 11 May 2016 on the interoperability of the railway system of
the European Union (repeals Dir. 2008/57 / EC);
• Directive (EU) 2016/798 of the European Parliament and of the Council of 11 May 2016 on
railway safety (repeals Dir. 2004/49 / EC).

In particular, Regulation (EU) 2016/796 suppresses the ERA and establishes the European
Union Agency for Railways (EUAR), which is entrusted with the task of issuing a single safety
certificate, provided for by the same railway package. In particular, the Regulation makes EUAR
the only body responsible for issuing vehicle authorizations and safety certificates for railway
operators and authorizations for control, command and signalling systems on the tracks.
The EUAR also takes on new tasks, including:

• monitoring of national railway rules and the activities of national authorities in the areas of
interoperability and railway safety;
• the provision of independent and objective technical support, in particular to the
Commission;
• a stronger role in ensuring the coherent development and rapid implementation of
telematics applications;
• a significant role in ensuring the coherent development of the European Rail Traffic
Management System (ERTMS).

The Regulation also supports the establishment of the single European railway area and
the objectives related to interoperability, safety of the railway system and certification of train
drivers.
Directive (EU) 2016/797 provides the conditions for the elaboration and revision of TSIs
with the aim of defining an optimal level of technical harmonization such as to facilitate, improve
and develop railway transport services at within the Union, complete the single European railway
area. The Railway System, the Subsystems and the Interoperability Components, including the
interfaces, must meet the Essential Requirements that concern them and, consequently, must
comply with the technical and functional requirements set out in the Technical Interoperability
Specifications (TSI). Specifically, the Essential Requirements are divided into:

26
CHAPTER 2 Regulatory Framework

• general requirements (Safety, Reliability and Availability, Health, Environmental Protection,

Technical Compatibility, Accessibility);
• specific requirements of each subsystem (Safety, Reliability and Availability, Technical
Compatibility, Accessibility);

The latter apply to the Infrastructure, Energy, Rolling Stock, Maintenance, Operation and
Traffic Management subsystems, Telematics applications for passengers and freight transport.
Finally, of fundamental importance from the point of view of the safety of the European
railway network is Directive (EU) 2016/798. This Directive amends Directive 2004/49/EC and
establishes provisions aimed at improving Union rail safety and improving access to the market for
the provision of railway services. The Directive redefines the role of National Safety Authorities
(NSAs) and reallocates the responsibilities between the latter and the EUAR.
The changes also take into account the evolution of the railway market and ensure the use
of safety monitoring and risk control measures.
The Directive focuses on the single safety certificate, which is the key to accessing
infrastructure exclusively by railway companies that possess it. The purpose of this certificate is to
provide concrete proof that the railway company has implemented SMS, meeting the
requirements of TSIs, CSMs and CSTs, and other legal provisions, and complying with the relevant
national standards. The single certificate will be issued by the Agency or the National Safety
Authority, as appropriate, after assuming their responsibilities and submitting the entire dossier to
the national authorities.
Another fundamental element is the safety authorization issued granted to Infrastructure
Managers, certifying their compliance with the SMS and acceptance of measures to meet the
specific safety requirements essential for the safe operation of railway services by the undertaking
on its own network. This authorization is issued by the National Safety Authority or the Agency,
based on a positive assessment of the infrastructure manager's SMS and its compliance with the
requirements of the CSTs and CSMs.
Common Safety Targets (CSTs) identify the minimum safety levels that must at least be
achieved by the system as a whole and, where possible, by the different parts of the railway
system. CSTs can be expressed in terms of risk acceptance criteria or levels of safety objectives
considering both the individual risks, to which any passenger is exposed, and the risks to society
and must be reviewed periodically, taking into account the global development railway safety.

27
CHAPTER 2 Regulatory Framework

To facilitate the assessment of CSTs, and to allow monitoring of safety developments,

Member States acquire information on so-called Common Safety Indicators (CSIs) relating to
accidents, dangerous goods, suicides, accident history and safety of the infrastructure and its
implementation.
Finally, the Common Safety Methods (CSMs) describe the assessment of safety levels, the
achievement of safety objectives (CSTs) and compliance with other safety requirements through
an Independent Assessment Body.
CSMs can be classified into (Acquaro, 2019):

• common methods for the accreditation of Railway Undertakings (Regulations (EU)

2018/762 and (EU) 2019/779);
• common method for risk assessment in case of changes (Regulation (EU) 402/2013);
• common method for monitoring the safety performance of the rail systems of the Member
States (Decision 2009/460/EC);
• common method for managing improvement (Regulation (EU) 1078/2012);
• common method for the supervision of NSAs (Regulation (EU) 1077/2012);
• common methods for technical and operational standardization (TSIs)

2.3.2 Italian Railway Safety Regulation

Over the years, the concept of safety in the Italian railway sector has undergone significant
changes. One of the most important developments is related to the shift from a deterministic
approach to a probabilistic one, which is in line with the European Community's current safety
standards. The idea of a safe system has also gained importance, emphasizing the importance of
considering the whole railway system as a complex and interconnected entity.
The Presidential Decree (DPR) 753/1980 'New rules on police, safety and regularity of the
operation of railways and other transport services', is guided by this deterministic view of risk as
it requires that all measures and precautions be taken in order to 'avoid incidents'. According to
this concept, therefore, a safe system does not admit the existence of accidents (Acquaro, 2019).
A step forward is already made by the Decree of the Ministry of Transport and Navigation
number 247/VIG3 of 2000 where the task of 'minimizing risks' is entrusted to the infrastructure

28
CHAPTER 2 Regulatory Framework

manager. By putting the problem in these terms, the legislator admits the possibility that a
residual risk of the system remains, approaching the probabilistic approach.
Nevertheless, the DPR plays an important role by defining roles and responsibilities for
defining the measures and precautions to be taken in the railways. In particular, it entrusts the
power to issue general technical standards on railway matters to the Ministry of Transport.
The subsequent Decree of the Minister of Infrastructure and Transport of 28 October
2005 'Safety of railway tunnels' was based on the provisions of the Presidential Decree, which
were enriched with a probabilistic view of risk. The decree plays a fundamental role not only for
tunnels but for the entire railway system. In the case of tunnels, it defines responsibilities and
technical and operational solutions for risk control. From the point of view of risk, however, it
describes in detail the railway risk analysis procedure, from the analysis of the causes to the
quantitative calculation of the level of risk. In addition, the document is an important reference for
the definition of criteria for the acceptance of risks. Its annexes, in fact, quantitatively define the
level of risk acceptability based on risk assessments 'freely assumed' in modern society.

Table 2.4. Transposition of the European legislation on liberalization and security in Italy
NAME EUROPEAN REGULATION ITALIAN REGULATION
Dir. 91/440/CEE* DPR n. 277/1998*
Dir. 2001/12/CE* D.Lgs. 188/2003
Dir. 2001/13/CE*
First railway package Dir. 2001/14/CE*
Dir. 2012/34/UE D.Lgs. 112/2015
Reg. (UE) 2017/2177
Dir. 2004/49/CE* D.Lgs. 162/2007*
Dir. 2004/50/CE* D.Lgs. 163/2007*
Second railway package Dir. 2004/51/CE*
Dir. 2008/57/CE* D.Lgs.191/2010*
Reg. (CE) n. 881/2004*
Dir. 2007/58/CE D.Lgs. 15/2010
Dir. 2007/59/CE D.Lgs. 247/2010
Third railway package
Reg. (CE) 1370/2007
Reg. (CE) 1371/2007
Reg. (UE) 796/2016
Fourth railway package Dir. (UE) 2016/797 D.Lgs. 57/2019
Dir. (UE) 2016/798 D.Lgs. 50/2019
*Repealed

29
CHAPTER 2 Regulatory Framework

As shown in Table 2.4, the liberalization process was welcomed at the national level
already by the first European legislative acts. In fact, in the national context, Directive 91/440/EEC
saw its implementation in the DPR 277/1998 which imposed the separation between the
Infrastructure Manager and the Railway Undertakings, generating four Divisions within the Italian
National Railway Network company:

• the Infrastructure Division, under the direct control of Holding Ferrovie dello Stato S.p.A.;
• the Passenger Division (medium and long distance), the Local and Regional Transport
Division and the Cargo Division, under the control of Trenitalia S.p.A..

The issue of railway packages had a parallel reference to Italian legislation, with the
transposition of directives and regulations in corresponding Legislative Decrees.
From the point of view of safety, the changes introduced by the first Safety Directive
(Directive 2004/49/EC) and by Directive 2004/51/EC were collected in Italy by Legislative Decree
162/2007.
In accordance with the provisions of European legislation, Legislative Decree 162/2007
identifies the Italian NSA by establishing the National Agency for the Safety of Railways (ANSF).
The decree identifies the tasks of the agency, attributable to the following three areas:

• Regulations: definition and modification of standards and directives on railway safety;

• Authorization: issue of certificates and authorizations to Infrastructure Managers, Railway
Undertakings
• Inspectorate and Control: verifying the state of the infrastructure, railway vehicles and the
work of managers and operators regarding the effectiveness of the actions taken to protect
safety.

In addition, the Decree defines the role in protecting the safety of the IMs, RUs system and
of all the subjects involved in the railway service.
As well as the Agency, an Investigating Body is established, entrusted to investigate and
provide subsequent recommendations regarding serious accidents occurring in the system and
defined in order to prevent their occurrence in the future.
As the Second Package was superseded by the Fourth, in parallel in Italy the Legislative
Decree 50/2019, which implements the Directive 2016/798, exceeds and replaces the Legislative
Decree 162/2007 in the issues of railway safety and interoperability.

30
CHAPTER 2 Regulatory Framework

With the Legislative Decree 14th May 2019, n. 50, a decisive step was taken for the
European railway area through the Cooperation Agreement between the National Agency for
Railway Safety (ANSF) and the European Union Agency for Railways (ERA).
The Decree applies to the entire railway system and concerns the Safety Requirements as a
whole, putting in place provisions to improve the safety of the railway system and to improve
access to the market for the provision of railway services. However, it does not apply to metros,
trams, light rail vehicles, and infrastructures used only by such vehicles, including those used
occasionally by rail vehicles in the operating conditions of the light rail system for connectivity
purposes only. Private railway infrastructures are also excluded.
A novelty with respect to the provisions of the Directive in general is that the provisions of
Legislative Decree n.50 also apply to the Isolated Railway Networks used for local railway services,
which have been identified with the Decree of the Minister of Infrastructure and Transport no.
347/2019.
The Decree also contains the indications regarding the safe management of the railway
infrastructure and traffic, as well as the interaction between the Railway Undertakings, the
Infrastructure Managers and the other subjects operating in the railway system.
These subjects are not only the maintenance managers, but also the manufacturers,
transporters, shippers, maintenance service providers, holders, service providers and contracting
entities, etc., who are entrusted with the responsibility of providing, to the other actors of the
system, information affecting security, in order to minimize the risk.
The decree specifies the key stakeholders responsible for the development and
improvement of safety within the railway system. These include the Ministry of Infrastructure and
Transport, as well as ANSFISA, which is tasked with developing a systematic approach to safety for
infrastructure managers.
Risk management is one of the novelties of the Directive, in fact it is required, as a general
principle, to all subjects operating in the sector, to take all the necessary measures to deal with
the risk and to report these risks to the interested parties, to ANSFISA and the National
Investigating Body. The Agency ensures that the managers implement the necessary risk control
measures, apply the European Union rules and the National Standards and develop the Safety
Management Systems.

31
CHAPTER 2 Regulatory Framework

2.3.3 Local Railway Safety Regulation

The regulatory history of isolated local railways has moved independently and in parallel with the
interconnected networks, with an alignment only in recent years.
The definition currently assumed by the Italian regulatory context is provided by Legislative
Decree No. 50 of 14 May 2019 where local isolated railways are identified as railway lines and
networks isolated from the functional point of view from the rest of the Community network and
the management and programming tasks are entrusted to the Italian regions
Historically, the authorization and control of isolated railways have been entrusted to the
USITF (Special Office for Fixed Systems Transport). Established by Italian Law no. 870 on 1
December 1986, the USTIF is a peripheral body of the Italian Ministry of Transport responsible for
issuing authorizations for the entry into service of transport means and granting technical and
economic approval for interventions. The USTIF was responsible for granting authorizations and
approvals for fixed transportation systems, such as railways (regional and isolated), subways,
tramways, ski lifts and chairlifts, funiculars, lifts, etc.
With the process of liberalization, and in particular with the implementation of the Second
railway package from Legislative Decree 162/2007, the safety management of the interconnected
railways was entrusted to the newly formed ANSF, leaving the isolated railways under the control
of USTIF.
Only in 2017, with the Law Decree n. 148 of 16/10/2017, converted into Law no. 172 of
04/12/2017, that isolated railways used exclusively for passenger services were included in the
scope of Legislative Decree 162/2016, and ANSF was entrusted with safety oversight for these
types of networks. The agency is now responsible for identifying technical regulations and safety
standards, and evaluating mitigating or compensatory measures proposed by service managers
based on a risk analysis that takes into account the characteristics of the railway section, rolling
stock, and transport service when issuing authorizations.
After these changes, the isolated railways were included in the scope of Legislative Decree
50/2019 with the implementation of the Fourth Railway Package. This legislation confirmed the
regulatory framework previously established and required the census of isolated networks to
which it applies.

32
CHAPTER 2 Regulatory Framework

Paragraph 4, article 2 of the decree specifies the identification of isolated railway

networks, which was carried out through Italian Ministerial Decree no. 347 of 2 August 2019. The
decree identifies the networks listed in Table 2.5, which had a total length of 955 km and a traffic
volume of 5.35 million train-km at the time of the survey in 2019 (ANSF, 2019a).

Table 2.5. Italian isolated railway networks and linens according to the Italian Ministerial Decree no. 347 of 2 August
2019
Operating company Railway line
AMT Genova – Casella
Ferrovienord Brescia – Iseo – Edolo
GTT S.p.A. Torino – Ceres
SSIF S.p.A. Domodossola – Swiss border
ASTRAL S.p.A. Roma – Lido
ASTRAL S.p.A. Roma – Civita Castellana – Viterbo
Ferrovie della Calabria S.r.l. Entire network
Ferrovia Circumetnea Catania Borgo – Riposto – Suburban line
FAL S.r.l. Entire network
ARST S.p.A. Entire network
EAV Circumvesuviana Railway
EAV Cumana and Circumflegrea Railways

To those in the Table 2.5, the Italian Law Decree 10 September 2021 n. 121 added a line of
3 kms of the Rhaetian Railway (Rhätische Bahn AG) connecting Tirano and Campocologno
(Lombardy-Switzerland).
In carrying out its regulatory mission, ANSF issues in 2019 the Decrees number 1 and 3
relating only to isolated networks deriving from the need to apply certain safety measures,
through a reorganization process during which railway service operators can use the support and
supervision of the Agency (ANSF).
The ANSF has defined a new regulatory framework on safety that includes both regulatory
and organizational adaptations, as well as significant adaptations of infrastructure and rolling
stock. The aim is to harmonize technical and safety standards across the entire national railway
network, including functionally isolated networks.
ANSF Decree no. 1/2019 sets out safety principles for rail traffic, essential requirements,
and technical standards applicable to railway subsystems of functionally isolated networks, as well

33
CHAPTER 2 Regulatory Framework

as to service managers operating on such networks. The Decree and its annexes provide a
comprehensive framework for ensuring safety in functionally isolated networks.
ANSF Decree 3/2019 titled 'Discipline of rules and procedures, pursuant to art. 16,
paragraph 2, letter bb) of the Legislative Decree 14 May 2019, n. 50, applicable to networks that
are functionally isolated from the rest of the railway system as well as to subjects operating on
such networks' sets out the rules and procedures to be applied to functionally isolated networks.
The Decree contains three annexes, which cover a range of topics related to the safe operation of
functionally isolated network.:

• Annex 1. "Rules on the requirements of the Safety Management System, for the
application of Common Safety Methods (CSMs), for the issue of qualifications to personnel,
for the issue of the certificate of suitability for operation and on supervision applicable to
networks functionally isolated from the rest of the railway system ";
• Annex 2. "Rules for the application and certification of maintenance management systems
for vehicles circulating on networks that are functionally isolated from the rest of the
railway system";
• nnex . “Rules for the registration of vehicles circulating on networ s that are functionally
isolated from the rest of the railway system”.

According to Annex 1, functionally isolated networks must promptly implement a Safety

Management System (SMS) in compliance with the principles established by Legislative Decree no.
50. This system ensures the safe operation of their part of the network and represents a shift from
a prescriptive to a performance-based approach. This proactive approach focuses on continuous
improvement to prevent unwanted events, rather than reactive management based on
discontinuous actions after undesirable events occur. To ensure proper safety management of the
isolated railway network and its parts, the ANSF has issued directives, recommendations,
guidelines, and notes in accordance with the requirements of Legislative Decree 50/2019.

34
CHAPTER 2 The Railway Risk Management

2.4 THE RAILWAY RISK MANAGEMENT

2.4.1 The concept of Risk

The investigation and definition of the concept of 'Risk', whether in general terms or specifically
applied to railways, is an essential step in the analysis of the risk management process. It is
imperative to thoroughly understand the concept of risk to effectively manage and mitigate
potential hazards in railway operations.
While the objective of defining the concept of risk may seem straightforward, it is
important to recognize that there are several shades that need to be discussed. The concept of
risk is interdisciplinary and can take on different meanings depending on the context in which it is
used. As a result, there are numerous distinct yet related definitions of the same concept within
the literature (Aven & Renn, 2009).
The concept of risk has deep historical roots. In the 'Pericle's Funeral Oration in
Thucydides' History of the Peloponnesian War' two millennia ago, the Greeks emphasised their
ability to take risks and assess them beforehand, in contrast to the <<Others [that] are brave out
of ignorance and when they stop to think, they begin to fear>>. (Aven, 2003).
The need to assess the risk of a breakdown is strongly linked to the modern world's
growing complexity and the increasingly serious consequences of a failure of the technologies that
are gradually developed (bridges, dams, means of transport, etc.). Risk in its modern meaning was
introduced during the Renaissance years, during the development of mathematics and probability
theory, and was associated with an evolution in people's perspectives, as they began to see future
events not as an immutable product of fate, but as a result of actions taken in the present. The
term 'risk' comes from the early Italian word 'risicare' which means 'to dare', emphasising the idea
that what awaits us is linked to the actions we dare to take. (Bernstein & Glenn, 1999).
This concept is still used today by a generic definition of the risk as the possibility
that human action or events can have consequences that can harm aspects of things that humans
value (Klinke & Renn, 2002). This definition links the concept of risk to the decision-making
process. Risk is seen as an evaluation of the future effects of present actions becoming a guide for
the analysis in the decision-making process (Bernstein & Glenn, 1999; Rausand & Stein, 2020).

35
CHAPTER 2 The Railway Risk Management

Moreover, the ISO 31000 (2018) underlines its role in the planification process defining risk as the
effect of the uncertainty on objectives.
Three concepts shared by modern definitions of risk inextricably linked are Hazard,
Probability and Consequences. Rausand and Stein (2020) describe risk with the combination of
three questions: Wat can go wrong? What is the likelihood? What are the consequences? The first
one refers to hazard identification, the second to frequency analysis and the last to consequence
analysis.
Also in Modarres (1993) the three basic concepts of risk are summarised into the definition
‘[Ris identifies – N/A] the potential loss or injury resulting from exposure to hazards’. Moreover
Sorrill et al. (1987) describe risk as the exposure to the possibility of an economic or financial loss
or gain, physical damage or injury, or delay, arising from the uncertainty associated with pursuing
a particular course of action. This definition summarizes risks deriving from various areas, such as
the financial one, that linked to people's health or the quality of a service.
In summary, the level of risk of a system is linked to the possible consequences that the
hazards could have if they gave rise to an accidental chain. Indeed, from a quantitative point of
view, the risk is identified by the combination of the likelihood at which accidents or harmful
incidents occur and the level of severity of the consequences (Hubbard, 2020).
In general, the likelihood can be expressed both in terms of frequency and probability. For
a random phenomenon E, the classical definition of frequency of occurrence of E is the ratio
between the number of times, nE, in which E occurs and the total number of observations n. In the
case of the risk study, it is preferable to define the probability using the Bayesian approach, i.e. a
measure between 0 and 1 of the degree of belief about whether or not an event will occur and the
frequency in terms of events that occurred in a given time period.
Severity, on the other hand, is a measure of the consequences of an accident. The
consequences vary in extent and may concern economic damage, damage to health, property, etc.
In the case of railway accidents, the consequences are mainly measured in terms of the number of
deaths or injuries and the economic damages to infrastructure or rolling stocks.
The Eq.(2.1) represents the mathematical function of risk, where p is the probability or the
frequency of the accident occurrence and N is a measure of the consequences. The total risk of the
system, R, is given by the sum of the risks of each accident scenario that is likely to occur (Bai &
Jin, 2016).

36
CHAPTER 2 The Railway Risk Management

𝑅 = ∑ 𝑓(𝑝, 𝑁) (2.1)
The commonly accepted function is the product of the probability or frequency of
occurrence and the expected consequences.
The p and N quantities are defined from risk metrics, which are estimators that offer
insights into the future level of risk based on data collected over time regarding the system being
analysed or similar systems (Rausand & Stein, 2020). Risk metrics are designed to identify a
quantity that provides information on the level of risk and a well-defined measurement
methodology that can be applied using the available data. In the context of railways, the definition
of risk metrics involves a comprehensive assessment of various factors, such as the frequency
exposure measures (time, distance, number of trains-km), the type of trains to be considered, the
unit of measurement for consequences (deaths, injuries, equivalent deaths, etc.), and the
definition of severity meters. The latter involves determining when a consequence is to be
considered related to the accident, for which the European and national regulations provide
important guidance. Additionally, it is necessary to consider accidents in extraordinary conditions,
such as those resulting from terrorism, environmental catastrophes, etc. The outcome of applying
risk metrics is referred to as a risk measure or risk indicator, which can take the form of a single
number, a vector, or a function.
Both the frequency and consequences of an accident are a function of the chain of events
that resulted in a specific accident scenario. In particular, once an hazardous event, known as the
Initiating Event, has been identified as capable of triggering an accident scenario, the probability
and consequences are linked to the frequency of the causes linked to the Initiating Event, as well
as the subsequent development of the accident.
The reduction of the level of risk, therefore, is possible through the analysis and
management of these three factors. The process of identifying hazards, analysing and controlling
risk through the assessment of new safety barriers is called Risk Management..

2.4.2 The Risk Management Process

Once the concept of risk is understood, it is necessary to define a process for evaluating,
characterizing, and controlling it. Furthermore, the increasing complexity of modern processes and
industries necessitates demonstrating the low probability of accidents in order to satisfy public

37
CHAPTER 2 The Railway Risk Management

opinion and avoid catastrophic consequences (Rasmussen, 1997). These requirements are met
through the use of Risk Management (RM) methodologies.
Between the eighteenth and early twentieth centuries, the first applications of RM
methodologies occurred in the fields of insurance and banking (Hubbard, 2020). Because of the
benefits associated with risk management, these methodologies expanded in the following years
in the fields of finance and public health.
The invention of computers and the ability to analyse a large number of complex scenarios
in a short period of time and with limited resources fuelled the spread of RM. This was one of the
reasons why, beginning in the 1960s, engineering and economics began to use quantitative tools
to apply RM (Hubbard, 2020). Nuclear power and oil and gas were among the first engineering
fields to adopt these approaches.
These methodologies evolved and spread over time, with changes to the process to be
adaptable from case to case and to meet the needs of the analysts as well as the end users to
whom the risk had to be communicated. These requirements also resulted in the development of
simplified methodologies such as risk matrices or ranking and scoring methodologies, which can
be used and understood immediately.
Many of the methodologies implemented within specific processes and then transferred to
multiple areas have merged into regulations and standards. ISO 31000: 2018 ‘Risk management –
Guidelines’ (International Organization for Standardization, 2018) is one of the most important
references of this type, providing organisations with the principles, framework, and process for
risk management. The ISO is not the only reference; numerous standards have been published
with regard to various application subjects such as occupational health, aviation, information
technology, and so on.
Risk Management for ISO 31000 (2018) is the combination of coordinated activities to
direct and control an organization with regard to risk. is the set of coordinated activities used to
direct and control an organization's risk management. The ISO, in particular, is intended for the
area to which it refers, namely risk in organisations, and addresses all of the activities that must be
planned and carried out in order to deal with risk.
RM è is in general defined as ‘the identification, analysis, and prioritization of risks followed
by coordinated and economical application of resources to reduce, monitor, and control the

38
CHAPTER 2 The Railway Risk Management

probability and/or impact of unfortunate events’ (Hubbard, 2020) with the aim of maintaining a
particular process within boundaries of safe operation (Rasmussen, 1997).
This final definition summarises the RM process's steps and purpose. Risk Assessment and
Risk Treatment are the two main interconnected process characters. The Risk Assessment
combines hazard identification and risk analysis (Di Graziano and Marchetta, 2021) and provides
information on the system's risk level. Their purpose is to provide the basis for the Risk Treatment
phase, i.e. the assessment of the need and the efficiency of risk control measures and strategies,
as well as to define a plan capable of allocating an organization's resources more efficiently (X. Liu,
2016; Sasidharan et al., 2017). Financial and human resources are limited if compared with the
complexity of the system and of the hazards involved and the correct understanding of the risk
allows to optimize resources (Muttram, 2002b).
Monitoring, or the control of how the risk assessment factors change over time, is also part
of the risk treatment. If on the one hand the Risk Treatment is fed with the information of the Risk
Assessment, on the other it provides information on how the system changes due to the decisions
and use of the system, starting a new evaluation phase.
Also in railway sector the RM has the ambition of predicting and quantifying system
failures with the aim of planning and prioritizing adequate actions (Sasidharan et al., 2017). The
main references in railway sector for RM are the European Regulation 402/2016 (The European
Commission, 2013) and the EN 50126 series (CEI CLC/TR 50126-2, 2017; CEI EN 50126-1, 2017).
Numerous quantitative or semi-quantitative applications are reported in the literature focused on
one or limited infrastructure elements, such as track (Lin & Saat, 2014), civil works olše 0 ,
electricity (Cosulich et al., 1996), level crossing (X. H. Liu et al., 2014) and so on.
What has been said thus far can be summarised in the steps represented Figure 2.3.

39
CHAPTER 2 The Railway Risk Management

Figure 2.3. Risk management process flow-chart

Before the actual RM process begins, preliminary planning phase of the RM process is
defined, with the goal of describing the purpose of the analysis, the external and internal context,
the data basis, and the risk criteria to be used.
The steps of the RM process following the planning phase are extensively discussed in the
following sections.

2.4.2.1 System definition

The first step in performing an accurate analysis is to create a preliminary exhaustive and
in-depth description of the system to be analysed. This entails investigating all systems, locations,
and activities that can contribute to risk levels and incidental chains from the standpoint of user
safety.
The system to be analysed must be properly defined and delimited. Rausand and Stein
(2020) summarise the key elements to consider during the system definition phase:

• Boundaries and interfaces;

• Interaction and constraints due to factor outside the identified boundaries;
• Technical, human and organizational relevant aspects;

40
CHAPTER 2 The Railway Risk Management

• External conditions (e.g. environment);

• Functions of the systems;
• Ordinary and extraordinary operating conditions.
• Safety and emergency procedures.

It is also possible to implement the system breakdown, or the division of systems into
subsystems, parts, or subparts based on the complexity of the subject of the analysis.
At this stage, it is also possible to evaluate the system breakdown, which is the subdivision
of the systems into subsystems, parts or sub-parts according to the complexity of the object of the
analysis. The role of the data available for analysis is crucial (Anderson & Barkan, 2004).
Quantitative assessments is based on two types of data: descriptive and probabilistic (Rausand
and Stein, 2020). The first relate to the characteristics of the system at the current state and
collect technical, organizational, operational, and environmental data and come into play in the
definition of the system and of the existing safety barriers. The probabilistic data, instead, concern
accidental events and in particular the probability that hazardous events, failures of system
elements or operational errors occur. This type of data is more difficult to find due to the high
safety of the railway systems and the rarity of the events studied and is collected by the
infrastructure managers in appropriate incidental databases. Historical data, in fact, are an
essential tool for managers as, through the statistical interpretation of past events, it is possible to
estimate and predict the future behaviour of the system (Aven, 2003; X. Zhang et al., 2016).

2.4.2.2 Hazard identification

The Hazard Identification, together with Risk Analysis and Risk Evaluation re the three
fundamental phases of the Risk Assessment (International Organization for Standardization, 2009).
The Hazard Identification phase aims to answer the questions:

• What can go wrong? (Rausand & Stein, 2020)

• What can happen and why? (International Organization for Standardization, 2009)

In other words, the goal of this phase is to identify, recognise, and record all potential
hazards, threats, initiating and hazardous events that can cause harm by triggering an accidental
sequence due to interference with the system's safe operation (Valdez Banda et al., 2014).

41
CHAPTER 2 The Railway Risk Management

The objectives of Hazard Identification can be summarised as follows:

• Create a comprehensive list of system-related hazards and hazardous events.

• Recognize critical and non-critical hazards;
• Describe, classify, and characterise each hazard identified;
• Describe how the hazard can cause an initiating event and how it contributes to the
incidental chain;
• Describe the potential interactions and combinations of hazards;
• Identify the safety barriers associated with each hazard.

The methodologies for hazard identification are numerous and are chosen based on the
information and knowledge available (Bai & Jin, 2016; Famurewa et al., 2015) and frequently are
not limited to the identification of hazards but also incorporates complete Risk Assessment
methodologies. ISO 31010 (International Organization for Standardization, 2009) categorises
commonly used methodologies into three groups:

1. Evidence based methods

2. Systematic team approaches
3. Inductive reasoning techniques

The first category includes all methods based on an examination of the incidental history of
the system under consideration or similar systems. Check-lists, historical accidental data, and
literature reviews are some examples of this type of method. The second group, on the other
hand, includes techniques in which the role of expert experience and knowledge is central, such as
brainstorming, interviews, audits, diagram techniques, and so on. Finally, the third category
includes inductive reasoning methodologies, such as the HAZOP methodology.
Table 2.6 lists and briefly describes the methodologies mainly belonging to the second and
third categories. This list is not meant to be exhaustive, but rather to provide an overview of the
most commonly used methodologies.

42
CHAPTER 2 The Railway Risk Management

Table 2.6. Hazard Identification techniques

METHOD DESCRIPTION
Check lists or Written list of hazards or hazardous events based on experience and past
Process reviews events. Can be structured as a list of questions and need to be built case-
specific.
• PRO: Easy to use and minimal information are required
• CONS: Trusting past experience and the absence of stimuli means
that events not yet recorded can be overlooked.
Brainstorming To identify potential hazards, uses free-flowing conversations between
knowledgeable groups of people.
• PRO: it is quick and simple to set up and allows for the identification
of risks that have not yet appeared in the system through the
stimulation of the imagination and collaboration between
stakeholders.
• CONS: it is unstructured, there is no guarantee that all risks have
been investigated, and it is heavily dependent on the personalities of
those involved in the process.
Delphi technique It is a specific type of involvement of expert groups (panels) in which the
experts express their opinions anonymously and everyone has access to a
summary of all opinions.
• PRO: it allows to obtain all opinions, including the most unpopular,
and allows all panel members to express their views equally.
• CONS: complex and time consuming.
Structured A set of predetermined questions is asked of different individual
Interviews interviewees with the goal of identifying the system's hazards.
• PRO: allow interviewees to conduct in-depth analysis and involve a
large number of people.
• CONS: it is time-consuming and only stimulates the imagination in a
limited way.
Preliminary Hazard Inductive method used in the early stages of design to identify hazards and
Analysis (PHA) hazardous events capable of causing damage. The method is able to rank the
risks based on their frequency and consequences.
• PRO: It is simple to use, provides a foundation for more in-depth
analysis, and, because it refers to the early stages of a project, allows
you to act in time.
• CONS: only preliminary information, and it loses efficiency as the
system and events analysed become more complex.
HAZOP (Hazard Qualitative technique that combines special adjectives and guide words with
and Operability) process conditions to evaluate deviation from intended outcomes

43
CHAPTER 2 The Railway Risk Management

METHOD DESCRIPTION
• PRO: structured, multidisciplinary, provides solution and risk
treatment actions and can take into account consequences of human
errors.
• CONS: time-consuming, high level of documentation required and
highly relies on designers expertise.
SWIFT (structured A simplified form of HAZOP in which a set of 'what-if' type phrases, along
what-if) with a set of words and expressions are submitted to a team to investigate
the impact of deviations from normal operations.
• PRO: easily and widely applicable.
• CONS: final results depend on the interviewer's and participants'
preparation..
FME(C)A (Failure The FMEA (or FMECA if the severity rank is expressed) is a methodology that
Modes and Effect was developed in the context of reliability analysis and consists in listing all
(and Criticality) possible failure modes, causes, and effects for each elementary subsystem.
Analysis) • PRO: simple to interpret and comprehend, systematic and structured,
automatable and adaptable to complex systems.
• CONS: it is based on analyst experience, is time-consuming, and has
difficulty identifying hazards due to the complex interaction of
multiple failures.

2.4.2.3 Risk Analysis

Risk Analysis covers the overall processes of Causal and Frequency analysis, Consequences
Analysis and Risk Calculation. It is frequently used as a synonym for Risk Assessment, but in this
work, it identifies the phases of characterization and calculation of the two quantities f and N, as
well as the calculation of the risk level, in accordance with the ISOs definitions (International
Organization for Standardization, 2009, 2018).
Risk analysis consists on the determination of:

• The causes and sources of a given risk event, as well as the frequency with which the
Initiating Event may occur;
• The relative probabilities and consequences of all possible developments of an Initiating
Event;
• All risk control measures that the system is equipped with to reduce the probability or
severity of possible accidents.

44
CHAPTER 2 The Railway Risk Management

ISO 31010 (International Organization for Standardization, 2009) divides all possible Risk
Analysis methods into:

• Qualitative;
• Semi-Quantitative;
• Quantitative.

Qualitative methods rely on frequency, consequences and risk levels categories identified
with predetermined and defined reference adjectives and words. EN 50126 (2017) provides an
example of a qualitative classification, which is summarised in Table 2.7.

Table 2.7. Qualitative classes of Hazard frequency and consequences definition.

Metric Class Definition
Likely to occur frequently. The hazard will be continually
Frequent
experienced
Will occur several times. The hazard can be expected to occur
Probable
often
Likely to occur several times. The hazard can be expected to
Occasional
occur several times
Frequency
Likely to occur sometime in the system life cycle. The hazard
Remote
can reasonably expected to occur
Unlikely to occur but possible. It can be assumed that the
Improbable
hazard may exceptionally occur.
Extremely unlikely to occur. It can be assumed that the hazard
Incredible
may not occur
Fatalities and/or multiple severe injuries and/or major damage
Catastrophic
to the environment
Single fatality and/or severe injury and/or significant damage
Critical
Consequences to the environment and/or Loss of a major system
Minor injury and/or significant threat to the environment
Marginal
and/or severe system damage
Insignificant Possible minor injury and/or minor system damage.

The Semi-Quantitative use numerical evaluation scales for frequencies and consequences
but the numerical values are brought back to classes similar to those in Table 2.7 and the final
level of risk is expressed qualitatively.
In fact, based on the definitions provided by EN (CEI EN 50126-1, 2017), intervals of
frequencies and consequences for each class can be estimated. In the case of consequences, the

45
CHAPTER 2 The Railway Risk Management

estimation is simpler. Translating definition into FW ‘fatalities’ means at least or more FW a

‘severe injury’ 0. FW and a ‘Minor injury’ 0.0 FW . Table 2.8 (Acquaro, 2019) lists other
precautionary restrictions that the Italian IM regulation (RFI) proposes in addition to current ones.

Table 2.8. Quantitative classes of consequences [FWI] comparison between EN and Italian Infrastructure Manager
regulation
CONSEQUENCES
EN 50126 RFI disposition 51/2007
CLASS
Catastrophic ≥ > 0.1
Critical [0.1; 2[ = 0.1
Marginal ]0.01; 0.1[ ]0; 0.1[
Insignificant ≤ 0.0 0

Applications examples can be used to gain some examples of frequency classes as well as
the implications. In actuality, how they are defined depends on the circumstances and the
features of the system under investigation. The CENELEC CEI CLC/TR 50126-2 (2017) or Rausand &
Stein (2020) criteria, which are highlighted in Table 2.9, give an indication of the limits of each
class.

Table 2.9. Quantitative classes of frequency comparison between EN and literature values
FREQUENCY
EN 50126-2 (Rausand & Stein, 2020)
CLASS
Frequent More than 4 events/year Between 1 and 10 events/year
Probable Between 0.8 and 4 events/year Between 0.1 and 1 events/year
Occasional Between 0.143 and 0.8 events/year Between 0.01 and 0.1 events/year
Remote Between 0.029 and 0.143 Between 0.001 and 0.01
events/year events/year
Improbable Between 0.006 and 0.029 Between 0.0001 and 0.001
events/year events/year
Incredible Less than 0.006 events/year Between 0 and 0.0001 events/year

Finally, the Quantitative methods calculate the actual frequency and consequences of all
the accident scenarios with appropriate units of measurement and provide a quantitative

46
CHAPTER 2 The Railway Risk Management

indicator of the level of risk. The frequencies are estimated as number of events/year, the
consequences in terms of FWI/event and the final risk in terms of FWI/year.
Regardless of their type, the steps of the Risk Analysis methodology include:

• Causal and frequency analysis;

• Consequence analysis;
• Risk calculation methods.

The objectives of the Causal and Frequency Analysis are to determine the causes of a
hazardous event and to establish the relationships, relative importance of the causes and any
safety barriers present to hinder their occurrence.
Generally, these methods fall into one of the following categories:

• Statistical analysis of relevant historical data;

• Predictive techniques based on the analysis of the parts and operations of the system;
• Expert elicitation.

The first class of methods are based on the evaluation of the possible causes and
frequencies on the basis of the information extracted from the analysis of the incidental databases
of the system analysed or from similar systems.
The second type consists of numerous approaches that have been employed over time in a
variety of industries, including the railway one. A collection of methods, indicative but not
exhaustive, is reported in Table 2.10 (International Organization for Standardization, 2009;
Rausand & Stein, 2020).

Table 2.10. Methods for Causal and Frequency Analysis

METHOD DESCRIPTION SCOPE
Cause and effectConsist of a graphical representation of the knowledge Causes
diagrams and ideas raised during a brainstorming. identification
• PRO: easy to apply;
• CONS: it responds only to the search for causes but
not to the calculation of their frequencies and does
not indicate the relationship between causes.
FTA (Fault Tree Top-down methodology which, starting from a hazardous Causes
Analysis) event, breaks it down into gradually simpler causes up to identification and
the Basic/Root Events evaluating the interrelationships. Frequency
• PRO: easy to employs also in complex systems, analysis

47
CHAPTER 2 The Railway Risk Management

METHOD DESCRIPTION SCOPE

qualitative and quantitative analysis in complex
systems.
• CONS: static due to a binary states and Boolean
logic
Bayesian Graphic methodology using a network consisting of nodes Causes
networks representing a state or a condition and arcs representing identification and
the reciprocal influences. Frequency
• PRO: flexibility. analysis
• CONS: complex and time-consuming
Markov methods Stochastic process of forecasting the evolution of system Frequency
states based only on the present state, without the analysis
influence of previous states.
• PRO: intuitive, quantitative and qualitative, deep
analysis of system properties and operation.
• CONS: time-consuming, complicated and not
suitable as initial method for causes identification

Finally, expert elicitation provides the tool to determine the causes and
incidental frequencies where data is lacking. The literature has validated the performance of the
use of expert judgment in quantitative risk assessment as a tool to overcome inaccuracies or
approximations of the available data (Jiang et al., 2018).
The purpose of the Consequence analysis, on the other hand, consists in understanding the
events that may happen following the occurrence of the Initiating Event and which contribute to
amplifying or preventing the evolution of the accident, acting on the possible consequences and
their probability.
In particular, the objectives of the Consequence Analysis concern:

• The determination of all the factors (internal and external) that contribute to influencing
the evolution of the accident;
• The determination, through the study of the combination of the identified factors, of the
possible accident sequences and the related final accident scenarios;
• The determination of the frequencies of each accident scenario;
• The determination of the consequence spectrum of each accident scenario;;
• The identification of the available safety barriers.

48
CHAPTER 2 The Railway Risk Management

Also in this case, the main methodologies used to identify the consequences relate to three
types:

• Statistical methods on incidental data;

• Predictive methodologies;
• Expert elicitation.

Table 2.11. Methods for Consequence Analysis

METHOD DESCRIPTION
ETA (Event Tree Graphical and probabilistic inductive method. ETA, starting from the
Analysis) Initiating Hazard, uses forward logic to identify and assess the role of
successive influencing events in term of probability and severity.
• PRO: widely used and documented methodology, clearance in
consequence development description.
• CONS: no standard for graphical layout; risk of not adequately
developing the end of the whole tree.

Event sequence
diagram Methods similar to ETA but with different layouts to simplify the structure
Cause-consequence according to the needs of the analyst and the analysis.
analysis

Table 2.11 provides an overview of some of the most well-known strategies employed in
the absence of data that have consistently produced outstanding outcomes because of their
robustness and reliability.
Once f and N have been evaluated, the third step of the Risk Analysis phase consists in
Calculating the Risk indicator. The frequency of each potential accident scenario and the
associated consequences are combined to determine the risk level.
In the qualitative methodologies the risk level is described by categories identified by the
combination of frequency and consequence classes.
The Risk Matrix, which consists of tables with distinct frequency classes on the rows and
classes of consequences for each column, is a tool for qualitatively characterising the risk of the
system. The risk level is obtained from the intersection of rows and columns.
The Risk Matrixes are not regulated, so in different studies and in different fields of
application, matrices with columns and rows of different numbers or called differently can be

49
CHAPTER 2 The Railway Risk Management

used. In the railway case, EN 50126 (CEI EN 50126-1, 2017) proposes the matrix shown in Table
2.12.

Table 2.12. Risk matrix from EN 50126

FREQUENCY RISK CLASS
Frequent
Probable
Occasional
Remote
Improbable
Incredible
SEVERITY: Insignificant Marginal Critical Catastrophic

The matrix also is used for Semi-Quantitative methodologies. By associating a numerical

indicator to the frequency and consequence classes, the combination makes it possible to
measure the level of risk and make it comparable.
To make risk reduction actions prioritised, a metric known as the Risk Priority Number
(RPN) must be defined, which is generally defined as the product of the category numbers of the
frequency and the consequence associated with a specific accident scenario. The analyst
determines the order of numbering of the classes, the values to be used, and whether to use the
product or the sum of the indices based on the needs and characteristics of the analysis.
Despite the speed with which the matrices are evaluated and communicated, they have a
number of limitations. One of them is related to the definition of remote events. Such events have
a 175-year return time (equal to a frequency of 0.006 events/year, see Table 2.9) or 104 years
(equivalent to 0.0001 events/year, see Table 2.9), therefore, there are a class of conditions that
have never been observed or recorded may be impossible to consider, rendering incidental
reference information insufficient or incomplete (Acquaro, 2019). This is a problem that affects all
approaches based on time series analysis.
The use of Quantitative Risk Analysis methodologies can provide an accurate measurement
of the level of risk. In this case, the risk is estimated as a mathematical function of the frequency
and consequence metrics, allowing quantitative indicators of risk to be obtained.
The risk associated to the i-th scenario (Ri) can be obtained through the Eq.(2.2) (Italian
Ministry of Infrastructures and Transport, 2005; Rausand & Stein, 2020), where 𝛼 is a safety level

50
CHAPTER 2 The Railway Risk Management

coefficient that controls the weight of accidental consequences which, in the first analysis, is set
equal to 1 and F and N are respectively the frequency and the severity of the scenario.
𝑅𝑖 = 𝐹𝑖 ∙ 𝑁𝑖 𝛼 (2.2)
The Total Expected Risk (R) is the sum of the risk of all scenarios as expressed in Eq.(2.3).
𝑛 𝑛

𝑅 = ∑ 𝑅𝑖 = ∑ 𝐹𝑖 ∙ 𝑁𝑖 (2.3)
𝑖=1 𝑖=1

Dividing R to the total number of passengers exposed, the Individual Risk (IR) is obtained.
Societal or Cumulative Risk (CR) level (Aloqaily, 2018; Cantino et al., 2016), instead,
assesses the probability of a damage to be greater than a certain tolerance threshold. As shown in
Eq.(2.4), CR is represented by a cumulative probability curve on the F-N plane which, at the
generic severity value h, associates the sum of the frequencies corresponding to severity values k
greater than h.

𝑓𝐶,ℎ = ∑ 𝑓𝑘 (2.4)
𝑘>ℎ

The use of cumulative risk is due to the fact that the possible combinations of frequencies
and consequences can vary widely. The Cumulative Risk, therefore, is not a single number but a
curve in a plane of consequences versus their frequencies. This type of representation, which was
first used in nuclear engineering in the 1960s (Farmer, 1967), is known as the Farmer Curve or FN
Curve, where F is the frequency and N is the number of fatalities.
A representation of this type provides an immediate indication of how the risk is
distributed across the various types of accident scenarios. Indeed, depending on the accident
chain that is triggered, a Hazardous event can result in accidents with minor but frequent
consequences or catastrophic accidents with very low frequencies. These types of accidents are
identified in the upper left and lower right areas of the FN plan, respectively. This characteristic is
evident in the decreasing trend of the curves obtained by Evans (2011) relating to fatal accidents
that occurred in European railway network between 1980 and 2009, as shown in Figure 2.4.

51
CHAPTER 2 The Railway Risk Management

Figure 2.4. FN curves for accidents happened in Europe between 1980 and 2009 (Evans, 2011)

2.4.2.4 Risk evaluation and acceptance

The Risk Analysis phase results provide the input data required for the decision-making process.
The decisions to be made, in particular, are divided into two levels:

• Is the risk level acceptable?

• How and to what extent can the risk be reduced?

The first question is answered by the Risk Evaluation and Acceptance phase, while the
second is related to the Risk Treatment phase, which is discussed in the following paragraph.
Exposure to all system risks cannot be avoided completely, for both financial and practical
reasons (Jones-Lee & Aven, 2011). As a result, hazard is configured as an intrinsic attribute of the
system, and it is neither desirable nor possible to define Risk Elimination methodologies, but
rather Risk Acceptance Criteria that identify the system's maximum risk level.
The Risk Acceptance Criteria define the risk levels considered tolerable based on, and the
need for further risk-reducing measures.

52
CHAPTER 2 The Railway Risk Management

Risk Acceptance Criteria are defined by the need to keep the frequency of the critical
accidents below appropriate thresholds, maintaining the level of risk at levels comparable to other
occupational hazards (S. Zhang et al., 2019)
The Risk Acceptance Criteria must be defined in accordance with commonly accepted
criteria (CEI EN 50126-1, 2017) and must take into account the safety objectives, system
characteristics, governmental legislation, standards, or experience gained over time.
The following are three of the most important risk acceptance principles that have
emerged in Europe:

• ALARP: The 'As Low As Reasonability Possible' principle born in the United Kingdom is
based on the distinction of three areas of increasing risk: Acceptable, Tolerable and
Unacceptable. In the zone of Acceptability, the risks are considered so low that any further
effort to reduce them would bring no improvement to the system, not satisfying the ALARP
criterion. In the zone of Unacceptability, the risks are so high as to be intolerable and any
effort is reasonable as long as it brings the level of risk back within the limits of tolerability,
otherwise, the service should not be implemented (CEI EN 50126-1, 2017).
• GAMAB: The 'Globalment Au Moins Aussi Bon' principle is a method used in France and is
based on comparison with existing systems. In particular, new systems are required to
have a level of security at least as good as an existing reference system.
• MEM: The 'Minimum Endogenous Mortality' principle is a method used mainly in Germany
and considers acceptable a system which is not able to significantly increase the
endogenous mortality of a society. In fact, the causes of death of an individual, in addition
to causes related to health, malformations or pathologies, are linked to 'technological
factors' or sports, machinery, do-it-yourself activities and transport. Those are the cause of
endogenous mortality, estimated at around 2 ∙ 10−4 𝑓𝑎𝑡𝑎𝑙𝑖𝑡𝑖𝑒𝑠/𝑝𝑒𝑟𝑠𝑜𝑛 ∙ 𝑦𝑒𝑎𝑟, and,
according to this principle, the introduction of the new transport system must not generate
a significant increase in this value.

The ALARP principle, of the three methodologies, allows for reflection on each individual's
or society's susceptibility to risk. In fact, a single person has a more or less marked risk tolerance
when the resulting benefit is high. The same can be said for a society in which a high level of risk
associated with a sufficient benefit is acceptable. The introduction of the concept of Tolerability

53
CHAPTER 2 The Railway Risk Management

allows for the identification of a company's willingness to live with a risk, ensuring that the sources
of this risk are appropriately analysed and controlled.
There are three possible positions on the level of risk:

• Acceptability: when the risk levels are low and do not constitute an obstacle to the safe
conduct of operations;
• Tolerability or acceptability on condition: risk levels are still low but not negligible. In this
case, the risk's acceptability is contingent on the fulfilment of predetermined conditions;
• Unacceptability: the risk levels are too high and must be reported within the limits of
acceptability.

The concepts underlying the risk acceptance principles are also reflected in the provisions
of the Common Safety Method for risk evaluation and assessment (European Union, 2013) which
identifies as risk acceptance criteria:

• The application of good practise codes: it is based on the assumption that certain risks can
be controlled by applying previously consolidated practises and rules.
• Comparison with reference systems: it is based on the belief that if a sufficient level of
safety is guaranteed in a system, it can be used as a reference system to demonstrate the
acceptability in other similar systems.
• Accurate risk assessment: if none of the previous methodologies have demonstrated the
risk's acceptability, an explicit risk assessment is used.

Accurate risk assessment methods can be qualitative or quantitative in regards to the

approach used to define frequencies and consequences.
The risk matrix presented above is used in qualitative applications to assess risk
acceptability. As previously stated, risk matrices are not regulated, so there are numerous
examples of applications. IEN 50126, for example, provides the matrix in Table 2.13 in which the
risk is considered in the railway sector:

• Negligible: Acceptable without further actions.

• Tolerable: Acceptable but monitoring and controls need to be carried out.
• Undesirable: Acceptable only if additional reductions are impractical;
• Intolerable: Risk need to be mitigated.

54
CHAPTER 2 The Railway Risk Management

Table 2.13. Risk matrix for risk evaluation as reported by European legislation
FREQUENCY RISK CLASS
Frequent Undesirable Intolerable Intolerable Intolerable
Probable Tolerable Undesirable Intolerable Intolerable
Occasional Tolerable Undesirable Undesirable Intolerable
Remote Negligible Tolerable Undesirable Undesirable
Improbable Negligible Negligible Tolerable Tolerable
Incredible Negligible Negligible Negligible Negligible
SEVERITY: Insignificant Marginal Critical Catastrophic

RFI, Italy's infrastructure manager, tightened up the risk categories in a internal regulation
in 2007 (RFI regulation 51/2007), setting the acceptability thresholds (Table 2.14) and the
definitions:

• Negligible: Acceptable, any other action is needed.

• Tolerable: Acceptable only if all measures and precautions suggested by technique and
practise have been implemented;
• Undesirable: The risk must be eliminated;
• Intolerable: The risk must be eliminated.

Table 2.14. Risk matrix for risk evaluation as reported by Italian IM

FREQUENCY RISK CLASS
Frequent Undesirable Intolerable Intolerable Intolerable
Probable Undesirable Undesirable Intolerable Intolerable
Occasional Undesirable Undesirable Undesirable Intolerable
Remote Tolerable Undesirable Undesirable Undesirable
Improbable Negligible Tolerable Tolerable Undesirable
Incredible Negligible Negligible Negligible Tolerable
SEVERITY: Insignificant Marginal Critical Catastrophic

In the case of quantitative Risk Evaluations, the acceptability or tolerability of the risk is
expressed through numerical thresholds. The definition of these thresholds depends on the
system in question as well as security objectives to be pursued.

55
CHAPTER 2 The Railway Risk Management

Different limits are defined according to the risk indicator analysed. Generally, the Italian
railway regulation provide limitations for both Cumulated Risk and Individual Risk indicators
(Italian Ministry of Infrastructures and Transport, 2005).
In the case of the IR, the limits, calculated on the basis of the endogenous fatality and the
average annual journey of each passenger, are set at 10−9 or the Acceptability and 10−11 for the
Tolerability of the risk. A representation of these limits is shown in Figure 2.5.

Figure 2.5. Individual risk acceptance criterion.

In the case of CR, on the other hand, acceptability is assessed by comparison with
thresholds in the FN plan. In particular, as shown in Figure 2.6, two thresholds identify three
regions on the plane. In the first region (Acceptability) the risk is under control and no further
measures are required. In the Unacceptability region, the risk is too high and all the necessary
actions must be implemented to bring the risk level back within the regions of acceptability,
introducing new risk reduction measures or increasing the performance of existing ones
(Melchers, 2001). Finally, between the two thresholds the Attention region identifies the risk
levels for which the application of the ALARP principle is required.

Figure 2.6. Cumulated risk acceptance criterion.

56
CHAPTER 2 The Railway Risk Management

2.4.2.5 Risk treatment and monitoring

For the RM management process to be effective, decisions must be made based on the results of
the Risk Assessment phase and their performance must be monitored over time.
The Risk Treatment's purpose is to define and evaluate one or more actions to be
implemented in order to limit the frequency or consequences of critical accident scenarios if a risk
exceeds an acceptable threshold.
The decisions made in this phase necessitate a reassessment of the risk level and, as a
result, the start of a new RM cycle that takes into account the system changes.
There are four types of strategies that can be implemented:

• Transfer risk: risk can be transferred to subjects with greater risk management skills where
possible and convenient;
• Mitigate risk: risk can be controlled by implementing actions aimed at reducing the
frequency and consequences of accident scenarios;
• Avoid risk: if a specific source of risk is considered unacceptable, alternative options with a
higher level of safety can be evaluated;
• Accept the risk: if the acceptability thresholds are satisfied, the representativeness of the
data used has been verified, the reliability of the assumptions underlying the analyses and
all the procedures and rules of good practice have been put in place, acceptance and
monitoring of risk can be considered.

The risk level mitigation is possible through the implementation of technological,

infrastructural or operational actions, called safety measures or safety barriers aimed at
preventing, controlling or mitigating the accidental events (Sklet, 2006), ensuring the appropriate
levels of safety.
The presence of efficient and effective safety measures ensures the safety level of the
system and allows for assessment, where necessary, further risk reduction measures.
Each safety barrier contributes to reducing the frequency or consequences of accidents by
acting on one or more of the elements that come into play in the evolution of the accident. This
contribution can be quantified through observations on real systems, expert opinion or from
referenced sources based on the performance characteristics, the function performed and the
integrity level of the barrier.

57
CHAPTER 2 The Railway Risk Management

Based on their function, barriers can be preventive or proactive barriers, that act before a
specific initiating event, characterized by a frequency-reducing effect, or protective or reactive
barriers, that act after the accident reducing and mitigating its, with a consequence-reducing
effect.
The International Union of Railways (UIC) (UIC, 2002) adds facilitation of escape measures
and facilitation of rescue measures Firsts allow passengers to leave the accident site as quickly as
possible, the latter to facilitate the intervention of rescue services.
Social and economic criteria influence the evaluation of which measures to implement and
the resulting benefit. The Cost-Benefit Analysis (CBA) is the most commonly used economic
criterion in the railway industry due to its versatility and immediacy. The CBA compares the total
expected costs of a set of system configurations with the total expected benefits associated with
them in order to determine the most profitable option (International Organization for
Standardization, 2009). In general, the costs are associated with the purchase, implementation,
and maintenance of specific measures, whereas the benefits are valued as a monetization of the
value of the avoided fatalities. The assessment of the economic value of a saved life has numerous
limitations, most of which are related to the difficulty of quantifying a complex concept like human
life in terms of currency.
Multicriteria Decision Analysis (MCDA) methods allow to assess attributes of different
nature and difficult to monetize. MCDAs are a category of decision-support methodologies that
use a range of criteria to prioritize a set of options. The criteria used can include different
attributes, quantitative and qualitative, and of different weights for different stakeholders
involved in the analysis (Odoki et al., 2015).

58
CHAPTER 2 Risk based DSS for safety improvement in railway systems

2.5 RISK BASED DSS FOR SAFETY IMPROVEMENT IN RAILWAY SYSTEMS

Historically, hazardous industries have operated in accordance with codes, standards, and
requirements. However, nowadays, the required approach is performance-based, i.e., aimed at
achieving and maintaining standards rather than applying rules (Aven, 2003).
The same happened in the railway system with the progressive introduction of the four
railway packages of the CST and CSM. In particular, the new regulatory framework underlines the
central role of risk identification and categorization to meet this type of approach and provide the
necessary support to the decision process. Risk management, in fact, as an assessment of the
future effects of present actions becomes the guide for the decision process (Bernstein & Glenn,
1999; Rausand & Stein, 2020)
The high complexity railway systems and the high safety standards required make all those
tools essential to support complex management processes. This translates into the definition and
application of Decision Supporting Systems (DSSs) in railway system management, that is, all
information systems that provide knowledge and processing capability as support for decision-
making activities (Burstein, W. Holsapple, & W. Holsapple, 2008).
DSSs for the management of structures and infrastructures cover numerous engineering
fields (Barriquello et al., 2017; Burstein, W. Holsapple, & Power, 2008; Cavalcante & Alencar, 2012;
Lins et al., 2012).
For over two decades tools of this kind have also been used in the management and
maintenance of railways. For instance, Zoeteman (2002) describes a DSS based on Life Cycle Costs
estimation for analysing the long-term impacts of railway design and maintenance decisions. In
(Meier-Hirmer et al., 2006) a computer application for track degradation surveillance and
maintenance intervention decision-making is presented. In (Guler, 2013), author presents a DSS
approach for track maintenance and renewal management system based on a set of rules
resulting from measured data, deterioration models and expert opinion.
The application of a DSS is divided into a series of successive phases (Covello, 1987):

1. Identification and definition of alternatives;

2. Decision problem structuring and goal setting;

59
CHAPTER 2 Risk based DSS for safety improvement in railway systems

3. Definition of performance measures or variables able to measure the degree of

satisfaction of the objectives;
4. Definition of the reliability of the variables;
5. Assessment of the probabilities;
6. Specification of value judgements, preferences and trade-offs;
7. Evaluation of alternatives;
8. Sensitivity analysis.

The Risk Management process, therefore, also on the basis of what is described in the
previous paragraphs, meets the definition of a Decision Supporting Tool (Deng et al., 2020; Jardine
et al., 2006). In a railway system, the management decision problem must bring together the
requirements related to (Sasidharan et al., 2022):

• Maintenance of asset performance;

• Minimization of risks;
• Allocation of economic and non-economic resources;
• Pursuit of business policies;
• Compliance with standards and codes.

In this context, the risk is configured not only as an objective but also as a means for
guiding decisions by evaluating the economic and social cost of the consequences, and indirectly,
allowing to influence the opinion of stakeholders (Rausand & Stein, 2020).
Risk management in a decision-making context has to take into account the mitigating
impact on the level of risk, the generalized cost and the perception of risk (Aven, 2003) of each of
the actions or interventions considered in the analyses. For those reasons, Risk-Based Decision
Support Systems (RB-DSSs) have found extensive development in the management processes of
the maintenance of the railway system (Meier-Hirmer et al., 2006) with the aim of identifying the
best time to intervene in the elements of the infrastructure (Power et al., 2016).

60
CHAPTER 2 Effect of infrastructure behaviour on Risk Management

2.6 EFFECT OF INFRASTRUCTURE BEHAVIOUR ON RISK MANAGEMENT

The exploitation of a railway system leads to the inevitable degradation of the reliability and
availability of the assets that compose it, compromising the safe performance of operations and
increasing the risk of accident.
The geometric quality of the track, i.e. the property of the rails to maintain their geometry
in space within the appropriate safety limits (EN 13848-1, 2019), is one of the characteristics that
is strongly affected by the number, load, and speed of railway vehicles.
Il Maintenance activities, i.e. all technical and administrative actions aimed at retaining or
restoring a system to a state in which it can perform its required function, are tasked
for preventing this phenomenon (CEI EN 50126-1, 2017; Gits, 1992).
On the one hand, continuous and complete maintenance necessitates a significant financial
commitment for a railway company (Lasisi & Attoh-Okine, 2018), but it also allows for the
prevention of degradation risks.
Track geometry irregularities (wide gauge, excessive twist, horizontal and vertical rail
defects) are a major contributor to railway accidents, causing derailments due to the influence on
the dynamic balance between wheel and rail, and an indirect impact, triggering processes capable
of degenerating into rail breakage (Chenariyan Nakhaee et al., 2019). To give an order of
magnitude to the phenomenon, a broken rail derailment occurs every 133 defects on average
(Zarembski & Palese, 2006).
Therefore, the link between maintenance, safety, availability and reliability of a system is
profound, starting from the design of the railway system. RAMS, acronym of Reliability,
Availability, Maintainability and Safety (CEI EN 50126-1, 2017), is a discipline deriving from System
Enginering and increasingly used within the European railway organizations, aimed at integrating
the concepts of safety, availability and cost effectiveness from the early stage of a railway project.
As a result, in a railway management focused on maximising safety while minimising costs,
it is impossible to ignore the understanding and control of track geometry degradation, as well as
the definition of counter-actions.
In fact, the understanding the degradation processes allows to optimise the efficiency of
maintenance actions, avoiding the costs of over-maintenance or the risk associated with the
arising of defects by defining the most appropriate time to implement the planned actions. The

61
CHAPTER 2 Effect of infrastructure behaviour on Risk Management

definition of the time when the maintenance activity is performed and the actual or presumed
occurrence of the defect is critical. One of the main classifications of maintenance strategies is
based on this distinction. In fact, there is a distinction between a reactive type of corrective
maintenance and a proactive type of preventive maintenance.
The first is based on the 'fix it when it breaks' concept, which entails acting after the
occurrence of defects, thereby saving resources for infrastructure monitoring but incurring
problems such as unscheduled downtime, potential serious safety violations, and potentially
significant damage (Gits, 1992). Railways have always taken a safety-first approach, ensuring
maximum reliability and availability of railway assets through scheduled maintenance actions with
large safety margins. However, with the development of private vehicles, particularly in the post-
war period, rail transport has had to compete with other modes of transportation. As a result,
managers of railway infrastructures have found it increasingly difficult to provide services with the
highest levels of safety while maintaining competitiveness among modal choices. This issue has
fuelled the search for methodologies capable of optimising all railway management processes,
including maintenance (Carretero et al., 2003). This phenomenon manifests itself in the definition
and investigation of preventive methodologies that provide for the implementation of actions to
anticipate the occurrence of defects and always keep the system within appropriate safety
margins.
Preventive maintenance strategies, in turn can be (Consilvio et al., 2021):

• Scheduled or Planned: based on time intervals or predetermined usage cycles;

• Condition-based: based on frequent monitoring and data collection to determine the state
of health of the various elements of a system and verify the possible exceeding of
threshold values and plan actions only when necessary;
• Predictive: involves the definition of forecasting models calibrated on datasets collected
less frequently than the condition-based approach capable of providing information to plan
interventions only when necessary.

Unlike other types of maintenance, the predictive approach ensures the lowest fault
probabilities while lowering costs through forecasting models and a limited number of monitoring
measures (Consilvio et al., 2019; Deng et al., 2020).

62
CHAPTER 2 Effect of infrastructure behaviour on Risk Management

The evaluation of the system's lowest fault probabilities cannot ignore assessments of
safety and risk. The maintenance intervals must be defined in such a way that they allow for as
little intervention as possible while maintaining the system's availability and reliability
(Soleimanmeigouni et al., 2016).
As a result, risk optimization and risk management enable management to efficiently guide
resource allocation decisions, as well as the management of infrastructure quality (Jardine et al.,
2006). Furthermore, the position of the set of options in terms of risk and cost allows for the
comparison of heterogeneous quantities, such as the greater tolerability of one type of defect
versus another based on the relative effect on the level of risk.
There have been numerous applications of Risk-Based Decision Support Systems (RB-DSS)
presented, particularly in the so-called hazardous industries. For example, in (Jardine, 2002) where
authors blend economic considerations and the risk estimate to establish optimal condition-based
maintenance (CBM) decisions and apply this methodology to decisions in food, coal mining and
mass transit industries, or Viana et al. (2022) propose a decision model for selecting a portfolio of
risk-based maintenance actions in natural gas pipelines.
DSS methodologies applied to the optimization of preventive maintenance have also been
proposed in the railway sector for almost three decades. Numerous examples are collected in (Soh
et al., 2012).
Consilvio et al. (2021), however, argue that many of the first models proposed in the
literature use cost minimization and time duration, to reduce the overall maintenance budget or
strategic maintenance crew scheduling as decision criteria. Instead, the optimization process must
also take into account the concepts of reliability, availability, safety and risk.
In (Bharadwaj et al., 2007) qualitative and quantitative risk analysis is used to identify and
assess ris levels for ‘high ris ’ components in railway structures and a Cost Ris Optimi ation
analysis is used to undertake run-repair-replace decisions. The methodology is then focused on
the degradation of railway system structural components due to corrosion and the optimum time
to replacement or repair is estimated under limited budget constraints.
Tezuka et al. (2015) propose a maintenance schedule optimization based on the Monte
Carlo approach for the evaluation of maintenance costs and the simulation of failures according to
the failure probability distribution.

63
CHAPTER 2 Effect of infrastructure behaviour on Risk Management

In (Consilvio et al., 2019) and (Consilvio et al., 2021) a mixed-integer linear programming
problem based on risk minimization is formulated for rail maintenance scheduling optimization
using quasi-real-time asset data. In particular, they introduce the concept of risk in railway
maintenance activities’ scheduling, determining the maintenance deadlines by means of a
predictive model of rail vertical deformation, characterized by spatial discretization and
uncertainties.
The general safety level of the system is firmly linked to the reliability of the railway system
assets. In fact, there are numerous studies that aim to maximize reliability at minimum of
maintenance costs, as in (Eker et al., 2011) where a simple state-based prognostic method that
aims to detect and forecast failure progression in railways turnout systems is presented, in
(Moghaddam & Usher, 2011) optimization models are developed to determine the optimal
preventive maintenance and replacement schedules in repairable and maintainable systems, in (T.
Zhang et al., 2013) an enhanced genetic algorithm approach is proposed to search for a solution
producing a minimum costs maintenance schedule in a finite planning horizon, or, finally, Sun et
al. (2017) who propose a mathematical model to minimize costs and ensure reliability for railway
turnout maintenance schedule.
The European Union itself has funded projects aimed at studying the application of
reliability-centred maintenance (RCM) techniques to railway infrastructure. One of them, named
'RAIL: Reliability centred maintenance approach for the infrastructure and logistics of railway
operation', is presented and tested in a large-scale railway network in (Carretero et al., 2003).
Many of the models presented take into consideration the predictive maintenance of the
railway network technological assets, not dealing with the quality of the track.
For example, Zarembski & Palese (2006) describe a management tool that deal with three
key track failure areas (broken rail, track buckling, track geometry irregularities), and a risk
approach is used effectively and economically focus resources on these high-risk areas to control
and reduce the number of derailments.
Rhayma (2013) presents a methodology to analyse the behaviour railway track based on
diagnostic data allowing reliability analysis of different maintenance operation.
(Khajehei, Ahmadi, Soleimanmeigouni, & Nissen, 2019; Letot et al., 2016; R. Li et al., 2017;
Soleimanmeigouni, Ahmadi, Khajehei, et al., 2020; Vale et al., 2011; Vale & M. Lurdes, 2013; Wen
et al., 2016) provide tamping optimization models taking into account only technical and economic

64
CHAPTER 2 Effect of infrastructure behaviour on Risk Management

aspects, such as degradation modelling, maintenance thresholds, time needed to carry out
maintenance actions, etc.
Ultimately, to date, the literature is full of studies and methodologies capable of optimizing
the maintenance processes of the elements of the railway infrastructure on the basis of risk-
related assessments. There are also numerous models for optimising predictive maintenance
actions for railway track management by minimising costs, times, and operational disruptions.
Despite this, the use of risk management for maintenance action optimization is still limited, and
none of these have ever been implemented and validated using data from systems such as local
railways.

65
3. DEVELOPMENT OF A FRAMEWORK FOR RISK BASED
DECISION MAKING IN LOCAL RAILWAYS
CHAPTER 3 Introduction

3.1 INTRODUCTION

CHAPTE 2
ocal railways de ni on problems and
iterature review Railway safety regula on
challenges
regulatory RM process and methodologies
ocal railway safety regula on
framewor RM ad decision support tool

CHAPTE 3
Role of available accident data . .
e ni on of the RM methodology Role of infrastructure . .
Methodology . Role of Rolling toc . .
Role of limited resources . .

CHAPTE 4
ata calibra on
nalysis of infrastructure related ris s
Case study RM applica on Ferrovia Circumetnea
nalysis of vehicles related ris s
nterven on op mi a on

Figure 3.1. Topics of Chapter 3

The creation of a risk management framework for local railways necessitates the identification of
methods and methodologies that best meet the requirements and needs of the analysis and
decision-making processes.
To make the process reliable, repeatable, and the results comparable, a single framework
must be defined that incorporates all Risk Management steps, from system definition to risk
assessment and decision analysis.
In general, the evaluation process begins with the collection of accidental data and its
statistical analysis. As stated in previous chapters, the local railways do not have enough accident
data to conduct an complete statistical analysis. To address this issue, similar reference systems
with a larger and more detailed database are required. In this regard, the extensive accidental
history of the national network in Italy provides a useful starting point.
Safety standards, operational characteristics, and infrastructure can vary significantly
between different railway systems, and these differences can have a considerable impact on
system safety. Therefore, to ensure accurate assessments of accident rates, it's crucial to calibrate
them based on the specific infrastructural and operational characteristics of the systems under
study. This calibration should involve a comparison of relevant safety characteristics between local
railways and reference railway systems.

67
CHAPTER 3 Introduction

Based on these considerations, this chapter aims to describe the risk assessment
methodology in local railways starting from reference systems, and in particular it highlights:

• The criteria for identifying and characterizing Hazardous Events (Section 3.2);
• The analysis of accidental causes using the FTA methodology (Section 3.3);
• Analysis of the accident consequences using the ETA methodology (Section 3.4);
• Calibration of models based on the characteristics of local railways (Section 3.5.1);
• Definition of the role of local railways infrastructure (Section 3.5.2) and rolling stock
(Section 3.5.3) in risk management;
• Definition of the decision optimization problem in limited budget environments as in local
railways (Section 3.5.4) based on risk analysis outcomes.

68
CHAPTER 3 Accidental analysis and identification of hazardous events

3.2 ACCIDENTAL ANALYSIS AND IDENTIFICATION OF HAZARDOUS EVENTS

The role of railway accident history is an important starting point for identifying the most common
railroad accidents in order to support the hazards identification phase. However, rail transport has
always been a highly safe mode of transport, with a significantly lower record of significant
accidents than other modes of transport (European Railway Agency, 2018).
The rarity of accidental events, combined with the lack of a systematic collection of the
precursors associated with such events, makes building a statistically significant data base difficult.
Furthermore, the local railways, due to the less line length all over the word, the peculiar
characteristics and the limited accidents precursors data, databases available for risk analyses are
inadequate.
As a result, in order to identify critical Hazardous Events, the reference database needs to
be expanded to wider railway contexts, with the infrastructure. The analysis of these databases
must take into account railway characteristics, and the information must be filtered based on
accidents that are actually compatible with local lines.

3.2.1 Hazardous Events Identification

The first step in accident data analysis is determining which accidents are critical to the system. In
other words, it is necessary to understand what accidents can occur in the system based on data
gathered over time in railway networks.
The goal of this step is to identify, through the analysis of national and international
accidental databases and of literature, the Hazardous Events that are responsible for the majority
of deaths and injuries on a railway system. The critical analysis of accidental databases then aims
to filter out accidents that are incompatible with local networks, such as freight train accidents
and hazardous materials train.
The data to be analysed refer to 'significant' accidents, i.e. those accidents whose
consequences (on people or things) has exceeded a certain threshold. In general, a significant
railway accident is considered if it involves at least one moving railway vehicle and has caused at
least one death, serious injury or economic damage (in terms of destruction of things or

69
CHAPTER 3 Accidental analysis and identification of hazardous events

interruption of the railway service) beyond a certain threshold (UIC, 2022). In particular, UIC and
the European Railway Agency (ERA), considers significant and accident that at list produced:

• A Death, considered only if happened immediately or within 30 days of the accident;

• A Serious Injury, that is a person hospitalized as a result of the accident for more than 24
hours;
• A Serious Damage defined as a damage to rolling stock or structures for more than 150,000
euros and/or prolonged traffic disruption for more than 6 hours.

These definitions exclude suicides and attempted suicides.

Railway bodies and statistical bodies classify significant accidents according to different
criteria. Based on the type, the UIC divides railway accidents into:

• Collisions: collision between rolling vehicles or vehicles and obstacles;

• Derailments: derailments of rolling stock;
• Accidents to people caused by moving rolling stock;
• Accidents at level crossings;
• Other accidents (including fires and explosions).

The UE adopts the same definition, groping main railway accident into collisions;
derailments; level crossing accidents; accidents to persons involving rolling stock in motion; fires
and others (European Parliament, 2004).
On the basis of these criteria, various databases are available for analyses suitable to
extend the statistical basis of railway accidents.
Main causes of death of passengers in railways registered in Europe between 1946 and
2009 are related to Collisions of trains, Derailments and LC accidents (Evans, 2011). Data provided
by ERA (European Railway Agency, 2018) highlights that in 2018, on the European network, among
all the significant accidents recorded, 1500 occurred at level crossings and 200 are attributable to
Derailments, Collisions and Fires.
Therefore, in order to build a useful Risk Management tool, in order to control and prevent
major accidents capable of affecting the safety of users of the railway system, it is useful to
analyse the Critical Hazardous Events:

• Derailment

70
CHAPTER 3 Accidental analysis and identification of hazardous events

• Collision
• Fire
• Accident at the level crossing

Furthermore, for the purposes of this work, since the analysis is aimed at local passenger
traffic railway lines, accidents involving freight trains are excluded.

71
CHAPTER 3 Frequency analysis

3.3 FREQUENCY ANALYSIS

In general, every Quantitative Risk Assessment (QRA) application, starts from the estimation of the
frequency of occurrence of the hazardous events previously identified.
The attribution of the causes to the various subsystems was carried out through a Fault
Tree Analysis as described by IEC 61025:2006 - Fault Tree Analysis (FTA)(IEC, 2006). This allows for
the evaluation of possible interrelationships between the causes and the clear determination of
the cause-effect chain that leads to the accident, as well as the identification of existing safety
measures (barriers), estimation of their impact, or identification of those to be introduced to
mitigate the risk.
Below are described the Fault Tree structures for each of Derailment, Collision, Fire, and
Accident at the level crossing.

3.3.1 Causes of Derailment

Among the accidents involving only rolling stock, derailment is the most common (X. Liu et al.,
2011). The literature has investigated the main causes of derailment with the aim of establishing
efficient strategies for reducing the probability of this HE.
In (Wang et al., 2014) an FTA is implemented for derailment in an Urban Rail Transit,
identifying signalling defects, natural disasters or engineering structure damages as macro causes.
In particular, the reporting category includes both technical faults and personal errors or malicious
behaviour. Damages to engineering structures, on the other hand, include all those tunnel or
online failures related to design or maintenance problems.
Also in (Dindar et al., 2017) e main causes triggering a derailment belong to signalling
defects associated with human errors or to infrastructure defects. In general, it is possible to trace
railway accidents to three main macro-categories of causes (P. Liu et al., 2015):

• Human error,
• Technical causes (failures of infrastructure or rolling stock),
• Causes external to the railway system.

72
CHAPTER 3 Frequency analysis

Britton et al. (2017) identify main derailment causes into Track related precursors, Train
equipment causes, Operation related causes related to human error of the driver or staff as well
as vandalism and external factors. In particular, technical causes are mainly responsible for
recorded derailments followed by accidents due to human errors (ANSFISA, 2021; X. Liu et al.,
2012).
The analysis of the sub-causes belonging to each macro-category allows for the
construction of a more ramified Fault Tree structure and for a more detailed description of the
derailment phenomenon. Liu et al.(2012) identify through the analysis of accident databases how
the main causes of derailment in line are associated, in order of frequency, to:

1. Broken rails or welds;

2. Track geometry;
3. Buckled track;
4. Obstructions;
5. Bearing failure (car);
6. Wide gauge;
7. Train handling;
8. Broken wheels (car)

This distribution of causes is also valid for the Italian national network, where the main
causes of derailment are classified in (ANSFISA, 2021; Marchetta et al., 2023):

1. Track geometry and roadbed irregularities

2. Mechanical or electrical defects/wear of rolling stock
3. Irregularities of the infrastructure (rail/switches/portals)
4. Abnormality due to external event/flooding
5. Failure/incorrect compliance with regulation
6. Failure/incorrect compliance with operational/technical prescriptions
7. Irregularities concerning work sites
8. Loss of rolling stock components
9. Landslides/boulders/trees on the railway site
10. SPAD
11. Wrong itinerary preparation

73
CHAPTER 3 Frequency analysis

12. Maximum permissible speed exceeding

13. Positive hot-box detectors (HBDs)
14. Breakage of the coupling devices of the rolling stock
15. Obstacles interfering with the loading gauge

Therefore, on the basis of these considerations, the tree shown in Figure 3.2 and described
in Table 3.1 was built for the framework proposed in this work.

Figure 3.2. Derailment Fault Tree structure

74
CHAPTER 3 Frequency analysis

Table 3.1. Description and incidence of Derailment root causes

Human Error
H1: Failure/incorrect compliance with regulation
H2: Failure/incorrect compliance with operational/technical prescriptions
H3: SPAD
H4: Wrong itinerary preparation
H5: Maximum permissible speed exceeding
Technical Causes
Infrastructure
T1: Irregularities of the infrastructure (rail/switches/portals)
T2: Track geometry and roadbed irregularities
Rolling Stock
T3: Mechanical or electrical defects/wear of rolling stock
T4: Positive hot-box detectors (HBDs)
T5: Loss of rolling stock components
T6: Breakage of the coupling devices of the rolling stock
Causes external to the railway system
E1: Landslides/boulders/trees on the railway site
E2: Abnormality due to external event/flooding
E3: Irregularities concerning work sites
E4: Obstacles interfering with the loading gauge
Other
U: Causes undetermined or unspecified

The T2 class 'Track geometry and roadbed irregularities', considering the importance of the
track geometry and the degradation of the infrastructure which will be better analysed later, can
be further explored.
In particular, this category includes bankruptcies related to:

1. Roadbed
2. Track geometry
3. Rail, Joint Bar and Rail Anchoring
4. Frogs, Switches and Track Appliances
5. Other

75
CHAPTER 3 Frequency analysis

Therefore, the appendix T2 of the FTA of Figure 3.2 can be decomposed as shown in Figure
3.3 and Table 3.2.

Figure 3.3. Appendix T2 to the Faut Tree of the derailment relating to the causes related to track geometry defects

Table 3.2. Causes of derailment related to track geometry defects

Track Roadbed and Structures Derailment Causes

T2.1 Roadbed
T2.2 Wide Gauge
T2.3 Twist
T2.4 Alignment
T2.5 Longitudinal level
T2.6 Other

3.3.2 Causes of Collision

Also for Collision, main triggering causes fall within three macro-categories (P. Liu et al., 2015):

• Human error,
• Technical Causes,
• External Causes.

Unlike the derailment, however, the weight of each of these macro-categories varies due
to the different dynamics of the accident.
Several studies have pursued the objective of identifying the main causes capable of
triggering a collision between trains. Turla et al. (2019) investigate more than 300 causes of

76
CHAPTER 3 Frequency analysis

collision between freight trains and Lin et al. (2020) among passenger trains in the US network in
recent years and identify as major causes, in order:

1. Failure to obey signals;

2. Violation of train speed rules;
3. Violation of mainline operating speeds.

Similar results are obtained by Hasheminezhad et al. (2021) through Fuzzy methodologies
for the prioritization of alternatives in complex problems, establishing a ranking of factors for the
collision through the use of questionnaires, expert judgment and literature studies. In particular,
according to the experts, they identify as major causes lack of attention at red lights (SPAD), lack
of issuing orders or compliance with regulations, and causes related to the unavailability of Active
Train Control systems (common condition in local railways - A/N).
In (Y.-F. Li et al., 2013) a Fault Tree is proposed for rear-end collision accidents and most of
the identified root causes are attributable to human errors or problems related to the signalling
system.
Based on these considerations and on the elicitation of expert groups it is possible to build
the Fault Tree for the Collision shown in Figure 3.4 and Table 3.3.

Figure 3.4. Collision Fault Tree structure

77
CHAPTER 3 Frequency analysis

Table 3.3. Description and incidence of Collision root causes

Human error
H1: Failure/incorrect compliance with regulation
H2: Failure/incorrect compliance with operational/technical prescriptions
H3: SPAD
H4: Wrong itinerary preparation
H5: Maximum permissible speed exceeding
Other (undetermined or unspecified)

3.3.3 Causes of Fire

Data on railway accidents show that in general fires constitute a less impactful event for the
system when compared with collisions and derailment (ANSFISA, 2021; RSSB, 2018). Modern
technologies and safety procedures reduced the frequency and consequences of this HE which,
however, still need to be accurately analysed, especially for parts of the line, such as tunnels,
which lead passengers to be more exposed to the effects of fire and its toxic products.
Also for Fire, the causes of an accident can be grouped within three macro-categories, namely:

• Human error;
• Technical causes;
• External causes.

In particular, Technical Causes represent the major cause. The data collected on the English
network (RSSB, 2018) underline how causes of fires in passenger trains (event identified by the
code HET17) can be summarized, in order of frequency, as in Table 3.4

Table 3.4. Main Fire Causes in UK railway network (RSSB, 2018)

RSSB Main Fire Causes
1. Main fire causes
2. Engine fires
3. Dragging brake fires
4. Exhaust systems fires
5. Traction motors fires
6. Electric equipment fires
7. Current collection fires
8. Resistance bank fires
9. Electric equipment fires
10. Cab fires

78
CHAPTER 3 Frequency analysis

11. Power car fires

12. Hot/failed axle box fires
13. Oil/diesel leak fires
14. Switch gear fires
15. H&V equipment fault
16. Mechanical equipment fires
17. Brake sparking fires
18. Control equipment fires
19. Transformer fires
20. Battery box fires
21. Locomotive fire

Table 3.5. Main Fire Causes in Italian railway network

RFI Main Fire Causes
1. Mechanical or electrical defects/wear of rolling stock
2. Defects of electrical system
3. Overheating of rolling stock components
4. Abnormality due to external event
5. Positive hot-box detectors (HBDs)

Similarly, as shown in Table 3.5 , the data collected on the Italian network show how
events involving rolling stock represent the first 5 causes of fire ignition. Based on these
considerations, the Fault Tree for HE Fire is shown in Figure 3.5 and Table 3.6.

Figure 3.5. Fire Fault Tree structure

79
CHAPTER 3 Frequency analysis

Table 3.6. Description and incidence of Fire root causes

Technical causes
Rolling Stock
T1: Mechanical or electrical defects/wear of rolling stock
T2: Defects of electrical system
T3: Positive hot-box detectors (HBDs)
T4: Overheating of rolling stock components
T5: Loss of rolling stock components
T6: Breakage of the coupling devices of the rolling stock
Causes external to the railway system
E2: Abnormality due to external event/flooding
E3: Irregularities concerning work sites
Other
U: Causes undetermined or unspecified

3.3.4 Causes of Level crossing accidents

evel Crossings C are one of the ‘blac spots’ of a railway line due to the high ris coming from
interaction of two very different transport systems: rail and road. Numerous studies that aim to
analyse the causes that lead a train and a car to occupy an LC at the same time.
The Hazardous Event at level crossing, i.e. the event to be studied that produces a
potentially hazardous situation, is the 'Hazardous crossing of the LC' by a vehicle or a train. This
event defines all the cases in which, due to the failure of one or more systems, the LC is unable to
prevent the passage of road vehicles when a train is approaching or fails to notify the train or
vehicle driver of the hazard (the LC remains unprotected).
In (Nowakowski et al., 2018) train-road vehicle collisions can occur through two means:
either by entering a protected level crossing due to driver error, or by entering an unprotected
crossing. In the latter case, the reason for the collision can be attributed to either driver error, or
the absence of warning systems caused by errors on the part of controllers or incorrect operation
of the level crossing barriers. This lack of warning may be compounded by the failure of detection
systems to identify an approaching train, malfunctions in control systems or actuating devices.

80
CHAPTER 3 Frequency analysis

In (Joung, 2005) an FTA is constructed on the basis of accident data collected in Korea. In
particular, the failure of LC to protect from approaching train is happens if at least one of the
following conditions occurs:

• Failure of track circuit or communication system;

• LC controller indicates route clear when occupied;
• Timing sequence failure.

Also in (Berrado et al., 2011) a list of 63 potential Basic Events for accidents at LC obtained
through several brainstorming sessions is provided. The Hazards identified are attributable to
problems related to the LC (malfunctions due to poor maintenance, lack of light or acoustic
warnings, barriers improperly closed or too slow to close), problems related to the train (train
brakes do not work, non-compliance with standards, etc.) or problems related to vehicles both
mechanical (e.g. breakdowns or problems to cross the LC) and behavioural (drivers ignore the
signal, low level of discipline, driver disregarding signals, etc.).
Therefore, from what emerged from the literature analysis, 'Hazardous crossing of the LC'
is possible if at least one of the:

• Railway failure: that is a breakdown or failure of the rail-side safety systems that allow the
train to cross the unguarded LC;
• Failure of the LC: i.e. a failure or malfunction of the road side LC protection systems that
allow vehicles to cross the LC with an approaching train;
• Hazardous behaviour of the vehicle: that is all those cases in which the vehicle invades the
track due to incorrect behaviour of the road vehicle driver.

In particular, the railway failure has been related to the occurrence of:

• Human error of the train driver (RE1);

• Train detection failure (RE2);
• Failure of the logic and control systems of the LC (RE3).

The failure of the LC is possible if the following REs occur:

• Barriers failure (RE5);

• Acoustic signals failure (RE6);
• Warning lights failure (RE7);

81
CHAPTER 3 Frequency analysis

• Vandalism (RE4).

Finally, incorrect behaviour of the road vehicle driver is attributable to:

• Vehicle breakdown (RE8);

• Rush or distraction of vehicle drivers (RE9).

The fault tree structure is summarized in Figure 3.6.

Figure 3.6. Fault tree of the Ha ardous Event ‘Ha ardous crossing of the C’

82
CHAPTER 3 Consequence analysis

3.4 CONSEQUENCE ANALYSIS

The analysis for determining the possible consequences of the scenarios deriving from the basic
events identified was defined through the following subsequent steps:

• Identification of subsequent events and definition of accident scenarios (Event Tree

Analysis);
• Calculation of probabilities and frequencies of final accident scenarios;
• Calculation of consequences through lethality models.

The identification of the accident scenarios can be carried out through the analysis of the
information reported in the accident databases integrated by experts judgement.
The probabilities if each event is assessed using statistical analyses as well as
considerations based on infrastructure, operation, and existing safety systems. Finally, the
consequences of the accident scenarios were quantitatively assessed using referenced calculation
models.

3.4.1 Identification of accident scenarios and calculation of frequencies: Event Trees

Once the HEs have been identified and the causes and their frequencies obtained, it is essential to
understand how the accident can develop and how the events following the HE can contribute to
influencing both the frequency and the consequences of the final accident scenarios.
In this sense, the Event Tree Analysis allows to determine the possible final accident
scenarios starting from each the HE. A final scenario is triggered by an accidental sequence made
by the combination of several successive events which independently and sequentially describe
the evolution of the emergency. The frequency associated with each evolutionary scenario
depends on the combination of the probabilities of each event of the sequence.
The following paragraphs comment on the considerations that led to the different
structures of the Event Trees starting from each HE identified.

83
CHAPTER 3 Consequence analysis

3.4.1.1 ETA Derailment

Considerations regarding the characteristics of the operation (speed and frequencies), the number
of trains involved, and the dynamics of the accident are used to identify and assess the
probabilities of the final scenarios of the basic derailment.
The necessary information for constructing the Event Tree can be obtained through
analysing literature and extracting relevant data from accident databases. In (Raza et al., 2020) an
Event Tree for train derailment is proposed which identifies as events subsequent to the
derailment:

1. Possibility that a train maintains the clearance (i.e. the train remains within safe limits);
2. Probability that the train derails towards the adjacent track or from the opposite side;
3. Probability that one or more wagons fall over;
4. Probability of a structure being hit;
5. Probability that a following train collides with the collapsed one.

Same structure of Event Tree is proposed in (Bearfield & Marsh, 2005) in its study on an
urban railway.
The evidence obtained from Italian accident data shows how the events capable of
influencing the number of deaths resulting from a derailment are (Di Graziano & Marchetta,
2021):

1. Collision following derailment with a second train;

2. Fire triggered by derailment.

These considerations made it possible to build the Event Tree for Derailment as described
below.
As a first step in the accident's evolution, the derailment capable of producing serious
damage to the passengers were analysed, i.e. the accidents where the derailment affects first to
produce fatalities, then aggravated by subsequent events. This event is defined as "Serious
accident" and can be evaluated from the accident data.
As a further ramification, the possibility that the derailment is followed by a collision with a
second train was considered. This event is strongly influenced by the characteristics of the railway
operation (number of trains, speed, signalling, etc.), by the number of tracks (single or double) and
by the performance of the crew and rolling stock.

84
CHAPTER 3 Consequence analysis

The collision can occur either with a train traveling on the same track, in the opposite
direction or in the same direction, or with a train traveling on an adjacent track. The collision on
the same track occurs between a stationary non-signalled train and a second train which, arriving
on the same track, in the same direction or in the opposite direction in the case of single-track
lines, enters the block section occupied due to a failure of the signalling system, fails to brake in
time and hits the obstacle. The collision on the same track is a combination of two different
probabilities. The first is linked to the operation of the line and takes into account the time
distribution of the trains. For the collision to occur, the second train must be distant from the
derailed train for less than the alarm and stop time considering the least favourable conditions
combined with the probability of failure of the signalling system. The second probability is linked
to human error which, combined with the absence of protection systems, causes entry of the train
into the occupied section.
For the collision to occur with a train running on the adjacent track three conditions must
be met:

1. The train derails in the direction of the adjacent track (P1).

2. The train derails with a lateral displacement greater than the free clearance between the
tracks, interfering with the dynamic gauge of the other direction of travel (P 2);
3. There is an oncoming train in the other direction within a distance less than the braking
distance (P3).

Finally, a further evolution of the emergency is linked to the triggering of a fire caused by
the accident. The probability of triggering a fire following a derailment is linked to a large number
of factors, first of all the type of derailed train, and can be obtained from the accidental
information taken as a reference.
The latter event is critical in the tunnel, where the confined space and short escape
distances mean that exposure to high CO concentrations and temperatures can result in death. As
a result, it is critical to consider whether the event will occur in the tunnel or outside. Despite
derailment has a lower probability to occur in tunnels due to the protection from external events
and the homogeneity of the route, it can be assumed equally probable along the entire line,
assessing the derailment probability in tunnels as the ratio between the sum of the length of all
tunnels of the line and the total length of the track. Figure 3.7 shows the Event Tree obtained.

85
CHAPTER 3 Consequence analysis

Figure 3.7. Derailment event tree and final accident scenarios

Finally, the final scenarios identified can be described as follows:

• DER1: Serious derailment, subsequent collision with a second train, ignition of a fire in the
tunnel;
• DER2: Serious derailment, subsequent collision with a second train, ignition of a fire on an
open line;
• DER3: Serious derailment, subsequent collision with a second train;
• DER4: Serious derailment, ignition of a fire in the tunnel;
• DER5: Serious derailment, ignition of a fire on an open line;
• DER6: Serious derailment;
• DER7: Minor derailment.

3.4.1.2 ETA Collision

As for the HE Derailment, considerations regarding the characteristics of the operation (speed and
frequencies), the number of trains involved and the dynamics of the accident (invasion of adjacent
track, collisions, etc.) played an important role for the definition of the Event Tree of Collision.
Useful data were obtained from literature, accidental databases and expert opinions.

86
CHAPTER 3 Consequence analysis

The literature contains numerous studies examining potential collisions between several
types of trains. The purpose of this work, however, is to define a model for local passenger traffic
railways therefore information not compatible with this accident were not considered.
As with all accidents, even for collisions the recorded fatalities can be caused by the
primary event or by the chain of subsequent events. Thus, as first branch of the tree, among the
significant collisions are distinguished the severe accidents from those in which the recorded
fatalities are linked only to the evolution of the accident and not to the collision itself.
The subsequent development of the accident taken into consideration concerns the
triggering a fire due to the accident. In fact, Evans (2011) highlights how in Europe, between 1946
and 2009, at least two cases were recorded in which more than 30 fatalities were caused by
collision followed by fire. This probability is linked to a large number of factors, first of all the type
of collided trains, and can be obtained from the accident databases taken as a reference.
Also in this case, the fire event proves to be critical in the tunnel, where the confined space
and escape distances lead to higher concentrations of CO temperatures, causing the death of
passengers exposed. Therefore, also in this case, maintaining the assumptions described in the
previous paragraph, the probability that the accident occurred in the tunnel or outside is assessed.
Figure 3.8 shows the Event Tree thus obtained for the Collision event.

Figure 3.8. Collision event tree and final accident scenarios

The final scenarios thus identified can be summarized as follows:

• COL1: Serious collision, with ignition of fire in the tunnel;

• COL2: Serious collision, with fire triggering in an open line;
• COL3: Serious collision;

87
CHAPTER 3 Consequence analysis

• COL4: Minor collision.

3.4.1.3 ETA Fire

The ETAs for the fire event proposed in the literature are often constructed by analysing the
availability and reliability of emergency operations and systems (Leitner, 2017; Van Weyenberge
et al., 2016). Alarm or extinguishing systems, however, are protective measures that intervene on
the consequences and not on the frequencies of the scenarios (protective measures). In this phase
of the analysis, in fact, the objective is to build an Event Tree capable of explaining the probability
with which all the possible final scenarios of the fire may arise.
To this end, the identification of the branches of the tree took place through assessments
based on accidental data and expert judgement, identifying as possible worsening events.
To this end, the identification of the branches of the tree took place through assessments
based on accidental data and expert judgement, identifying as possible hazardous events:

• Starting a fire in a tunnel;

• The presence of other trains in the vicinity of the fire.

As regards the first point, in fact, fires result to be critical in tunnels where closed spaces
expose exodants to higher concentrations of CO and heat (Cheng et al., 2021).
Secondarily the presence of other trains within the range of the fire constitutes an element
capable of worsening the severity of the accident, as evidenced by some of the most serious
accidents related to the fire in long tunnels or subways (Won hua, 2004)
The range of fire (𝑅𝑎 ) is the distance at which the products fire (heat, CO, radiation, etc.)
can still have appreciable effects on exodants. This distance depends on the dynamics of the fire,
the geometry of the tunnel and the characteristics of the exodus. In conditions of poor visibility
and high power fires, this radius can also extend for several hundred meters.
The probability that at the time of the accident there is an oncoming train at a distance
equal to or less than 𝑅𝑎 + 𝑠𝑎 , i.e., respectively, the sum of the fire range and the braking distance
of the approaching train depends on the operation railway, by the alarm times, by the type of
signalling system and by the position of the trains in the tunnel. Figure 3.9 summarizes the
structure of the Tree of Events just described

88
CHAPTER 3 Consequence analysis

Figure 3.9. Fire event tree and final accident scenarios

3.4.1.4 ETA hazardous crossing at Level Crossing

All the possible accident sequences starting from the "Hazardous crossing at the LC" event need to
be evaluated on the basis of considerations deriving from the characteristics of the level crossing,
the rolling stock, the operation, the vehicular traffic and the safety systems present.
In this regard, Kim et al. (2009) identify the speed of the train on impact, the braking
distance, the possibility of road vehicle passengers to get to safety and the dimensions of the
impacted vehicle as hazardous events to be studied in order to build their Event Tree. Bally et al.
(2019), on the other hand, highlight the simultaneous presence of a train and a road vehicle on the
railway junction as the first event to be considered for the accident to occur.
Based on these considerations, the first event that must be considered and which is a
fundamental requirement for an accident to occur is that a road vehicle actually crosses the track
in hazardous conditions. This probability is mainly linked to the characteristics and distribution of
the traffic flows crossing the LC.
In the event that the vehicle is in a hazardous condition, a second subsequent event that
must be taken into consideration is whether or not the vehicle is able to move off the tracks to
avoid the hazard. This event is strongly influenced by the characteristics of the LC and by the
decisions taken by the driver of the road vehicle under stress conditions.
Following the events described above, if the vehicle remains stuck on the tracks, the
severity of the accident is influenced by whether or not the train initiates the braking manoeuvre,
reducing or the speed or totally avoiding the impact. This event is influenced not only by the

89
CHAPTER 3 Consequence analysis

factors affecting the stopping distance of the train and by the visibility distances but above all by
the driver's ability to perceive the hazard and start the stopping manoeuvres.
Finally, the severity of the accident is strongly influenced by the type of vehicle involved,
mainly influencing the number of people exposed.
All these considerations are summarized in the Event Tree shown in Figure 4.18.

Figure 3.10. Hazardous cross at LC event tree and final accident scenarios

Ultimately, depending on whether or not these events occur, 10 different accident

scenarios are identified:

• LC1: Car on the track does not move and the train does not brake;
• LC2: Motorcycle on the track does not move and the train does not brake;
• LC3: Heavy vehicle on the track does not move and the train does not brake;
• LC4: Bus does not move and the train does not brake.
• LC5: Car does not move and the train brakes;
• LC6: Motorcycle does not move and the train brakes;
• LC7: Heavy vehicle does not move and the train brakes;
• LC8: Bus does not move and the train brakes;
• LC9: Vehicle clears the track;
• LC10: No vehicles on the track.

90
CHAPTER 3 Consequence analysis

3.4.2 Lethality models

Due to the enormous complexity of the investigated scenarios, estimating their effects
necessitates a methodical approach based on lethality models. For each initiating event, the
models allow to calculate the expected number of fatalities taking into account:
• Type of trains involved;
• Infrastructural characteristics;
• Occupation of convoys;
• Characteristics of the signalling and traffic management system (speed, distances, etc.);
• Development of consecutive events;
• Characteristics of emergency plans (methods and times of intervention, etc.).

In particular, the consequences of Derailment, Collision and Accidents at Level Crossings

are mainly influenced by the operating parameters (speed of the trains), while the consequences
relating to Fire are strongly linked to the geometric characteristics of the tunnel (section, escape
distances, etc.) and the characteristics of emergency management.
Lethality of cold scenarios (i.e. excluding fires and explosions) can be expressed through
models depending on the speed at the moment of the accident. Lethality is evaluated as a
probability (between 0, no fatality, and 1 death of all passengers) which, if multiplied by the
expected number of passengers exposed, returns the level of the consequences (number of
fatalities or FWI).
Otherwise, hot scenarios lethality is estimated, in this phase, through simplified empirical
models which take into account, in their application, the value of the risk factors linked to the fire
and the characteristics of the exodus process.
The hypotheses and assumptions underlying the calculation of the consequences for each
accident scenario relating to each of the analysed HE are presented below, starting from the basic
scenarios and then calculating the consequences linked to the combination of all possible
consecutive event.

3.4.2.1 Lethality of Derailment

The lethality of each final derailment scenario is determined by the combination of events
synthetized by the event tree.

91
CHAPTER 3 Consequence analysis

In particular, since the events considered by the ETAs constructed in the previous
paragraphs are independent and successive, each final scenario is defined by the sum of all their
effects. In other words, the simple derailment is considered as the base event and all consecutive
scenarios start from it and add further risk factors.
As shown in Eq. (3.1) The number of fatalities for a passenger train derailment is given by
the number of passengers exposed to the accident times the probability for each individual of
being killed, called lethality:

𝑆𝑒𝑣𝑒𝑟𝑖𝑡𝑦𝑑𝑒𝑟 = 𝑁𝑝𝑎𝑥𝑃𝑇 ∙ 𝜆𝑑𝑒𝑟𝑃𝑇 (3.1)

where:
• 𝑁𝑝𝑎𝑥𝑇𝑃 is the number of passengers in a passenger train;

• 𝜆𝑑𝑒𝑟𝑇𝑃 is the lethality of the simple derailment of a passenger train

The survived passengers (total number of exposed reduced by the number of death due to
the first derailment) constitute the population exposed to the lethality of subsequent events.
Therefore, if the derailment is followed by a collision with another passenger train, two
contributions need to be evaluated:

• 𝑁𝑝𝑎𝑥𝑃𝑇 ∙ 𝜆𝑑𝑒𝑟𝑃𝑇 that is the number of fatalities expected following the derailment,
calculated by applying the 𝜆𝑑𝑒𝑟𝑇𝑃 to the population exposed to the derailment, i.e. the
passengers of the first train;
• 2 ∙ 𝑁𝑝𝑎𝑥𝑃𝑇 − 𝑁𝑝𝑎𝑥𝑃𝑇 ∙ 𝜆𝑑𝑒𝑟𝑇 that is the number of passengers exposed to the collision, i.e.
the "survivors" of the derailment and the total number of passengers in the second train.
The collision lethality function must be multiplied to this rate 𝜆𝑐𝑜𝑙𝑃𝑇 defined before.
Therefore, the scenario of derailment of a passenger train with subsequent collision with a
second passenger train is given by the sum of the two contributions, as shown in (3.2):

𝑆𝑒𝑣𝑒𝑟𝑖𝑡𝑦𝑑𝑒𝑟+𝑐𝑜𝑙 = 𝑆𝑒𝑣𝑒𝑟𝑖𝑡𝑦𝑑𝑒𝑟 + (2 ∙ 𝑁𝑝𝑎𝑥𝑃𝑇 − 𝑆𝑒𝑣𝑒𝑟𝑖𝑡𝑦𝑑𝑒𝑟 ) ∙ 𝜆𝑐𝑜𝑙𝑃𝑇 (3.2)

When the derailment is directly followed by a fire, the presence of seriously injured people
must be considered. Lethality from fire is different when considering moving people (exposed to
risk factors that vary in space and time and able to carry out self-rescue) and major injured (who
are assumed to remain stationary and unable to self-rescue). For the former, the probability of
survival depends on the dynamics of the exodus, for the latter on the intervention time of
intervention of rescue. For this reason, a lethality function for injured people of passenger trains

92
CHAPTER 3 Consequence analysis

(𝜆𝑓𝑖𝑟𝑃𝑇,𝑖𝑛𝑗 ) which considers the rescue intervention time is introduced. Therefore, the total of
fatalities due to derailment and subsequent fire is given by the sum of three contributions (3.3):
𝑆𝑒𝑣𝑒𝑟𝑖𝑡𝑦𝑑𝑒𝑟+𝑓𝑖𝑟 = 𝑆𝑒𝑣𝑒𝑟𝑖𝑡𝑦𝑑𝑒𝑟 + (𝑆𝑒𝑣𝑒𝑟𝑖𝑡𝑦𝑑𝑒𝑟 ∙ 𝑘1 ) ∙ 𝜆𝑓𝑖𝑟𝑒𝑃𝑇,𝑖𝑛𝑗 +
+[(𝑁𝑝𝑎𝑥𝑃𝑇 − 𝑆𝑒𝑣𝑒𝑟𝑖𝑡𝑦𝑑𝑒𝑟 − (𝑆𝑒𝑣𝑒𝑟𝑖𝑡𝑦𝑑𝑒𝑟 ∙ 𝑘1 ) ∙ 𝜆𝑓𝑖𝑟𝑃𝑇,𝑖𝑛𝑗 ] ∙ 𝜆𝑓𝑖𝑟𝑃𝑇 (3.3)

where 𝑘1 is the ratio, judged from statistical analysis, between serious injuries and fatalities
associated with accidents with the basic event of derailment.
Finally, the less frequent and most severe scenario is the occurrence of all the three
successive events. In this scenario, a collision and the triggering of a fire follow the serious
derailment. The fatalities associated with the entire sequence are represented as in the previous
case by the sum of all the contributions described, calculated by updating the number of exposed
and lethality functions event by event.
As in the previous case, since a heat event occurs, it is necessary to take into account the
different lethality between exodus and seriously injured passengers. Thus, the number of fatalities
for the scenario that sees the derailment, the collision and the fire in succession is obtained from
the Eq.(3.4):
𝑆𝑒𝑣𝑒𝑟𝑖𝑡𝑦𝑑𝑒𝑟+𝑐𝑜𝑙+𝑓𝑖𝑟 = 𝑆𝑒𝑣𝑒𝑟𝑖𝑡𝑦𝑑𝑒𝑟+𝑐𝑜𝑙 + (𝑆𝑒𝑣𝑒𝑟𝑖𝑡𝑦𝑑𝑒𝑟+𝑓𝑖𝑟 ∙ 𝑘1 ) 𝜆𝑓𝑖𝑟𝑃𝑇,𝑖𝑛𝑗 +
+[(2 ∙ 𝑁𝑝𝑎𝑥𝑃𝑇 − 𝑆𝑒𝑣𝑒𝑟𝑖𝑡𝑦𝑑𝑒𝑟+𝑐𝑜𝑙 − (𝑆𝑒𝑣𝑒𝑟𝑖𝑡𝑦𝑑𝑒𝑟 ∙ 𝑘1 ) ∙ 𝜆𝑓𝑖𝑟𝑒𝑃𝑇,𝑖𝑛𝑗 ]𝜆𝑓𝑖𝑟𝑃𝑇 (3.4)

3.4.2.2 Lethality of Collision

As in the previous case, the lethality models for the collision consider the type of train involved,
the number of occupants of the trains, the speed and the development of events after the
accident.
The basic event from which all possible scenarios arise is the collision between two
passenger trains and, starting from this, the magnitude of the possible consequences is evaluated.
The lethality function is defined as follows:

𝑆𝑒𝑣𝑒𝑟𝑖𝑡𝑦𝑐𝑜𝑙 = 2 ∙ 𝑁𝑝𝑎𝑥𝑃𝑇 ∙ 𝜆𝑐𝑜𝑙𝑃𝑇 (3.5)

where:
• 𝑁𝑝𝑒𝑟𝑠𝑇𝑃 is the number of passengers on a passenger train;

• 𝜆𝑐𝑜𝑙𝑇𝑃 is the lethality for collisions between passenger trains.

93
CHAPTER 3 Consequence analysis

If a fire due to the accident is also triggered following the collision, the lethality models
estimate the magnitude of the consequences of this combination of events as the sum of two
contributions. The first due to the collision alone and the second due to the lethality of the fire
considering on the survivors of the accident.
In particular, lethality between exodus and seriously injured people needs to be assessed
separately as shown in (17):
𝑆𝑒𝑣𝑒𝑟𝑖𝑡𝑦𝑐𝑜𝑙+𝑓𝑖𝑟 = 𝑆𝑒𝑣𝑒𝑟𝑖𝑡𝑦𝑐𝑜𝑙 + (𝑆𝑒𝑏𝑒𝑟𝑖𝑡𝑦𝑐𝑜𝑙 ∙ 𝑘1 ) 𝜆𝑓𝑖𝑟𝑒,𝑃𝑇,𝑖𝑛𝑗 +
(3.6)
+[2 ∙ 𝑁𝑝𝑎𝑥𝑃𝑇 − 𝑆𝑒𝑣𝑒𝑟𝑖𝑡𝑦𝑐𝑜𝑙 − (𝑆𝑒𝑣𝑒𝑟𝑖𝑡𝑦𝑐𝑜𝑙 ∙ 𝑘1 )𝜆𝑓𝑖𝑟𝑃𝑇,𝑖𝑛𝑗 ] ∙ 𝜆𝑓𝑖𝑟𝑒𝑃𝑇
where:

• 𝑘1 = 2,2 is the ratio, inferred from statistical analysis, between serious injuries and
fatalities associated with accidents with the basic event being a collision.

For the evaluation of the value of fire lethality in tunnels, refer to §3.4.2.3.

3.4.2.3 Lethality of Fire

The determination of the consequences of a fire event in a railway tunnel requires suitable models
to be formulated and solved to simulate the space-time evolution (hazard flow) of a probabilistic
set of critical events and the corresponding exodus processes of the exposed population, aimed at
determining the number of individuals saved, according to the safety measures, their availability
and reliability
Also, for Fire, the number of fatalities is estimated as the percentage of deaths out of the
total number of exposed passengers as expressed by the Eq.(3.7):

𝑆𝑒𝑣𝑒𝑟𝑖𝑡𝑦𝑓𝑖𝑟𝑒 = 𝑁𝑝𝑎𝑥𝑃𝑇 ∙ 𝜆𝑓𝑖𝑟𝑃𝑇 (3.7)

where:
• 𝑁𝑝𝑎𝑥𝑃𝑇 is the number of passengers exposed to the fire;
• 𝜆𝑓𝑖𝑟𝑒𝑃𝑇 is the lethality function.

The lethality function makes it possible to relate the magnitude of the impacts (for
example the quantity of CO produced or the thermal radiation) with the severity of the damage to
the exposed people (Vílchez et al., 2001). In particular λ is function of the exposition to an effect
with a given intensity x which generates a given damage. In particular, this probability is described
by the following relationship (Finney, 1971):

94
CHAPTER 3 Consequence analysis

𝑃𝑟−5 𝑥2
1 −
𝜆𝑓𝑖𝑟𝑒𝑃𝑇 = ∫ 𝑒 2 𝑑𝑥 (3.8)
√2𝜋 −∞
where x is intensity of the harmful exposition (e.g. CO concentration, temperature) and Pr is a
Probit variable from ‘Probability unit’ which follows a normal distribution with an average value
of 5 and a normal deviation of 1. The value of Pr is determined as a function of S according to the
expression:

𝑃𝑟 = 𝑎 + 𝑏 ∙ ln (𝑥) (3.9)

where a and b are variables obtained empirically through experiments or from the analysis of
information on accidents.
Therefore, in the event of a fire in a tunnel, the intensity S of the consequences of the
passengers exposed to the hazard flow is a function of:

1. Value of risk factors (i.e., CO concentrations, radiative heat flux, etc.);

2. Exposure time.

The dependence of the Probit function on the indicated variables varies according to the
combustion product assumed to be relevant for the salvageability of the exposed population. A
Probit function used for estimating the consequences determined by the concentration of carbon
monoxide (CO) is defined as follows:

𝑃𝑟 = 𝐴𝐶𝑂 + 𝐵𝐶𝑂 ∙ 𝑙𝑛[(𝐶𝐶𝑂 )𝑛𝑐𝑜 ∙ 𝑇𝑒 ] (3.10)

where:
• 𝐴𝐶𝑂 , 𝐵𝐶𝑂 and 𝑛𝐶𝑂 are empirical constants of the Probit function which, for a dose with 50%
lethality, assume values -37.98, 3.7 and 1.0 respectively (RFI, 2006);
• 𝐶𝐶𝑂 is the concentration of carbon monoxide expressed in ppm;
• 𝑇𝑒 it is the representative function of the exodus process.

The evolution of the fire, and therefore the assessment of the CO concentration instant by
instant, is influenced by a large number of factors related to the geometry of the tunnel and the
thermo-fluid dynamic characteristics of the fire event. In order for the model to be used for an
immediate assessment of the effects of the fire and of possible management decisions, a
simplified formula whose efficiency is, however, equivalent to the simulations for the purposes of
this work is used. Therefore the CO concentration (𝐶𝐶𝑂 ) is assessed through Eq.(3.11) (RFI, 2006):

95
CHAPTER 3 Consequence analysis

𝑤𝑓𝑖𝑟𝑒 ∙ 𝑥𝑠𝑚𝑜𝑘𝑒
𝐶𝐶𝑂 = (3.11)
𝑠𝑠𝑚𝑜𝑘𝑒 ∙ 𝐴𝑠𝑒𝑐𝑡𝑖𝑜𝑛

where:
• 𝑤𝑓𝑖𝑟𝑒 is the thermal power of the fire expressed in MW;

• 𝑥𝑠𝑚𝑜𝑘𝑒 is the smoke flow rate for each MW of thermal power expressed in [mg/MWs];
• 𝑠𝑠𝑚𝑜𝑘𝑒 is the propagation speed of the smoke in the tunnel;
• 𝐴𝑠𝑒𝑐𝑡𝑖𝑜𝑛 is the free area of the tunnel.
𝑇𝑒 , instead, it depends on the dynamics of exodus. In this regard, two categories of
passengers exposed to the flow of hazard are examined: passengers able to carry out self-rescue
(exodus), for whom exposure to hazard depends on the speed of the exodus; passengers unable to
leave (injured) for whom a variable exposure time is considered according to the intervention time
of the rescue teams.
For the first category of passengers, the representative function of the exodus process is
defined as follows:
𝑑𝑒
𝑇𝑒 = 𝑡𝑖 + (3.12)
𝑣𝑒
where:
• 𝑡𝑖 is the time interval between the stop of the convoy and the start of the exodus;
• 𝑑𝑒 is the maximum evacuation distance equal to half the distance between the usable exits
of the tunnel (portals, stations and stops);
• 𝑣𝑒 is the evacuation speed which generally assumes values between 0.3 m/s up to 1 m/s
depending on the size of the sidewalks and visibility conditions.
In addition to the simple case of fire in a passenger train in a tunnel, the possibility must be
assessed that, following the fire of the first passenger train, a second passenger train is present
within the range of action of the fire.
The radius of action is understood as that distance at which the risk factors relating to the
fire (temperature, radiation, smoke production) have a non-negligible effect on the exposed
population. This distance, even of the order of several hundred meters, is a function of the
characteristics of the fire, the fire and the tunnel.

96
CHAPTER 3 Consequence analysis

From the analysis of the accident sequence, the total number of fatalities can be obtained
from the sum of the contributions of the basic event plus the portion linked to the effects of the
fire on the second train. Therefore, the total of fatalities for this accident scenario is equal to:

𝑆𝑒𝑣𝑒𝑟𝑖𝑡𝑦𝑓𝑖𝑟+2𝑛𝑑 𝑡𝑟𝑎𝑖𝑛 = 𝑆𝑒𝑣𝑒𝑟𝑖𝑡𝑦𝑖𝑛𝑐 + 𝑁𝑝𝑎𝑥𝑃𝑇 ∙ 𝜆𝑓𝑖𝑟𝑒_2𝑛𝑑 𝑡𝑟𝑎𝑖𝑛

(3.13)

Where 𝜆𝑓𝑖𝑟𝑒_2𝑛𝑑 𝑡𝑟𝑎𝑖𝑛 is evaluated according to the possible position of the second train
and therefore the relative escape distances.

3.4.2.4 Lethality of Accident at Level Crossing

The lethality of complex events such as level crossing accidents is connected to a large number of
factors, many of which are difficult to identify and quantify. To overcome this obstacle and obtain
an estimate of the probability of death in a train-vehicle accident, the statistical results present in
the literature referred to systems with characteristics comparable to the one studied were used.
In order to obtain a better accuracy in the estimation of the consequences it is necessary to
consider separately the severity of the accident for the passengers of the train from the occupants
of the vehicles. For example, in Joung (2005) the consequences of the accident such as "train hits
vehicle" are analysed for the Korean railway network, evaluating an average of 0.11 FWI/accident
distributed as follows:

• 1 serious injury on the vehicle, i.e. 0.10 FWI/accident;

• 2 minor injuries on the train, or 0.01 FWI/accident.

Similar results are reported in the RSSB (Railway Safety and Standard Board) (RSSB, 2018)
databases relating to railway accidents in England updated to 2018. In fact, referring to the
accident category "HET10 - Passenger train collides with vehicle at level crossing", on average,
12.40 accidents occurred in one year with 0.19 FWI/accident for a total of 2.4 FWI/year. In
particular, the data on FWI are characterized between train passengers and subjects external to
the train (in the specific case, the occupants of the vehicle involved), and in particular 0.395
FWI/year on the train and 2.002 FWI/year on the vehicles are estimated.
Dividing by the number of events/year, we obtain a total of 0.19 FWI/accident of which:
• 0.16 FWI/accident among vehicle occupants;
• 0.03 FWI/accident among train passengers.

97
CHAPTER 3 Consequence analysis

It is possible to refine the analysis by characterizing the data according to the type of PL to
which they refer. In fact, among the types of English PL it is possible to identify some attributable
to those analysed in the present study (low rail traffic and average speeds below 100 km/h). In
particular, reference is made to level crossings of the OC (Open Crossing) type, characterized by
the absence of protections and speeds between 8 and 16 km/h, and ABCL (Automatic Barrier
Crossing Locally monitored by train crew), i.e. automatic level crossings but supervised by the train
crew, where the speed is limited to a maximum of 90 km/h.
Respectively recorded:
• OC (8-16 km/h): 0.008 FWI/vehicle crash and 0.001 FWI/train crash;
• ABCL (90 km/h): 0.0742 FWI/vehicle accident and 0.0113 FWI/train accident.

These values represent the number of average FWIs from an accident in England. To obtain
the lethality it is necessary to refer these indicators to the average number of occupants of the
vehicle and of the train in order to obtain the lethality for each individual person involved.
Data from Road Traffic Estimates (2018) show that 78% of English traffic is made up of
motor vehicles, 1% motorcycles, 21% heavy commercial vehicles and 1% buses. On average, the
occupants of the vehicles considered amount to 1.3 for motor vehicles, 1 for motorcycles, 1 for
commercial vehicles and 40 for buses, it is possible to obtain the following lethality factors for the
two speed bands:

• 8 – 16 km/h: 5.55E-3 FWI/vehicle accident and 7.26E-6 FWI/train accident;

• 90 km/h: 4.93E-2 FWI/vehicle accident and 8.51E-5 FWI/train accident.

For intermediate speeds a double linear trend was assumed between 0 and 12 km/h
(average of the speed range of the OC) and between 12 and 90 km/h.

98
CHAPTER 3 Consequence analysis

(a) (b)
Figure 3.11. Lethality for train (a) and vehicle (b) passengers in function of train speed in Level Crossin accidents

Therefore, each accident scenario differs from the others in the number of people exposed
and in the lethality factor value. Based on these hypotheses, the ten identified scenarios can be
grouped into three main groups:

3. Scenarios LC9 and LC10 which are defined as safe scenarios, i.e. not capable of producing
fatalities, therefore null consequences are assigned for all PLs and for all vehicle types.
4. Scenarios LC1, LC2, LC3 and LC4 which include all the cases in which the driver perceives
the hazard and starts to brake the train, therefore the expected consequences of the
accident are equal to or lower than the maximum possible ones according to the
dynamics of the braking manoeuvre.
5. Scenarios LC5, LC6, LC7 and LC8 which are characterized by an impact speed equal to the
line speed at the PL and therefore characterized by the maximum consequences.

For the second and third group of scenarios, the number of FWIs is calculated as the
number of people exposed to the accident times the lethality factor (𝜆) of the same:

𝑁 = 𝑃𝐴𝑋 ∙ 𝜆(𝑉) [𝐹𝑊𝐼] (3.14)

More specifically, it is useful to distinguish the consequences for the train occupants from
those for the passengers of the vehicle involved. Therefore, on the basis of the values obtained
from the literature analysis, the following relationship was used:

𝑁 = 𝑃𝐴𝑋𝑡𝑟𝑎𝑖𝑛 ∙ 𝜆𝑡𝑟𝑎𝑖𝑛 + 𝑃𝐴𝑋𝑣𝑒ℎ𝑖𝑐𝑙𝑒 ∙ 𝜆𝑣𝑒ℎ𝑖𝑐𝑙𝑒 [𝐹𝑊𝐼] (3.15)

99
CHAPTER 3 Consequence analysis

Lethality is strongly influenced not only by the vehicle, but also by the impact speed. This
speed is assumed equal to the maximum for the scenarios of the third group but is reduced by the
possible braking in the case of scenarios of the second group.
In addition, the distance necessary for the deceleration and stopping of the train must be
compared with the visibility distance actually present when passing through the PL.
Two situations can be distinguished:

1. The visibility distance is greater than the stopping distance: in these cases the train
manages to complete the braking manoeuvre without hitting the vehicle stationary on
the tracks, causing no consequences;
2. The visibility distance is less than the distance necessary for stopping: in this case, once
the space between the moment when the driver sees the vehicle and the vehicle itself is
known, the speed variation was calculated starting from the inverse formula of the
braking distance. This speed variation is the one considered to estimate the lethality of
the impact according to the previously introduced factors.

100
CHAPTER 3 The role of local railway features in risk assessment

3.5 THE ROLE OF LOCAL RAILWAY FEATURES IN RISK ASSESSMENT

The application of the risk management framework, as mentioned in the preceding paragraphs,
encounters some obstacles in the case of isolated railways. It is necessary to identify all the
distinctive elements of this type of line that differ from national lines, which can affect the results
of the risk management and decision-making process.
The first obstacle is related to the lack of a vast accident history and the need to adapt
accident data from standard railway systems. It is essential to define a methodology that takes
into account the peculiarities of isolated railways in translating accident frequencies and
consequences.
Secondly, it is crucial to understand how the different infrastructure (gauge, gradients,
curves, etc.) and rolling stock (weight, speed, dimensions, etc.) compared to those of reference for
national accident databases can impact the level of risk and the reliability of the decision-making
process.
Finally, one of the major constraints characterizing local railways is linked to the limited
budgets allocated to the management and improvement of the line. The need to optimize
resources through a structured and precise process is therefore more evident. In this regard, the
evaluations on the peculiarities of local railways presented above associated with a decision-
making system aimed at optimizing resources should be the ultimate goal of this framework.

3.5.1 Accident data calibration

The quantitative risk analysis is based on reliable and statistically significant accidental and fault
data observed in the analysed railway system.
Railway environment is commonly recognised as a safe system in which the set of safety
systems and operating procedures allows risk levels to be kept under control. This means that the
accident databases contain a limited number of accident records available for statistical analysis,
with many minor high-frequency and low-severity accidents, and a lack of rare and catastrophic
accidents (An et al., 2011).

101
CHAPTER 3 The role of local railway features in risk assessment

In addition, in railways such as local ones, alignment with national safety procedures and
standards has only been taking place in a few years, making safety databases inconsistent and not
adequate for the implementation of quantitative risk estimation.
To overcome this problem, the framework shown in Figure 3.12, was implemented and
comprises the following five main steps:

• Data collection;
• Construction of the bow-tie model for risk analysis;
• Data calibration;
• Calculation and risk assessment and Risk monitoring or treatment.

The data collection phase concerns not only accidental records but also the description of
the technical and operational characteristics of the system of the case studied and of the railway
networks taken as a reference. In this sense, it is essential to identify all safety barriers and
characterize them with their reliability and effectiveness in reducing the frequency (preventive
measures) or the consequences (protective measures) of accidents. This phase allows for the
collection of the knowledge necessary to properly build the Bow-Tie scheme of the HE considered.
The knowledge accumulated allows to define the structure of the Fault Tree, decomposing the
causes of the HE in gradually more in detailed faults up to the Root Events, and to develop the
scenarios in the Event Tree by identifying all possible successive Events.

Figure 3.12. Risk management calibration framework

102
CHAPTER 3 The role of local railway features in risk assessment

The quantitative estimation of frequencies and consequences can be challenging. For this
reason, for each cause or consequence of the HE the calibration phase requires the estimation of
the mitigating effect of different safety barriers.
The frequencies, the consequences and the safety barriers of the reference networks need
to be identified. The same need to be done for the studied system so to have for each cause and
consequence:

• the frequency or severity from the reference system;

• the type and efficiency of the related safety barrier in the reference network
• the type and efficiency of the related safety barrier in the studied system.

The risk reducing effect of each safety barrier is than quantified through an Impact Factor
(IF). The IF expresses, in percentage terms and in relation to a single cause or consequence, the
variation in frequency or severity between reference and studied system due to differences in
safety equipment or procedures. Therefore, if the studied network is equipped with more
performing safety barriers, the IF will assume values <1, reducing the consequences or frequencies
related to the event considered, otherwise IF will be >1.
The IF value can be estimated quantitatively or qualitatively, depending on the information
available. The first method is applicable to all safety barriers whose reliability is known and the
probability of failure can be calculated. In this case, the IF can be evaluated from the relationship
between the safety barriers failure probability in the examined system and in the reference ones.
When information on safety barriers are missing or incomplete, qualitative methods based
on expert judgement must be chosen. The literature has validated the performance of the use of
expert judgment in quantitative risk assessment as a tool to overcome inaccuracies or
approximations of the available data (Jiang et al., 2018).
The results obtained from the combination of both previous methods are collected into an
IF matrix for the calibration of accidental data. By doing so, through the development of the risk
assessment model it is possible to quantitatively estimate the current risk level and compare it
with the acceptability thresholds and assess the need to control or monitor the safety level (Di
Graziano et al., 2022).

103
CHAPTER 3 The role of local railway features in risk assessment

3.5.2 Effect of infrastructure features

The link between track degradation, i.e. the lowering of the geometric and structural quality, and
the risk of derailment is strong and is also demonstrated by the need to take it into account in the
FTAs shown in the previous chapters. The accidental statistics confirm this relationship,
highlighting how in the lines where the required track quality is higher the derailment rate is
reduced (X. Liu et al., 2011).
The geometric quality of the track, in particular, decreases with time and the use of the line
to the point of generating ‘defects’ capable of inducing, alone or in combination, instability of the
rolling stock capable of causing it to derail.
The speed and probability of occurrence of a geometric defect is influenced by the
characteristics of the materials, the geometry of the track, the characteristics of the rolling stock
and the operational characteristics of the line. Local railways differ significantly from national
railways in many of these elements, implying the need to understand how this affects degradation
and monitoring and maintenance strategies.
To date regulations and research have focused mainly on standard networks (Ahac &
a ušić 0 7 , providing an extensive knowledge and regulatory background for maintenance
design and approaches to prevent and manage track geometry degradation (Andrade & Teixeira,
2015; Falamarzi et al., 2019a; Ferreira & Murray, 1997; Soleimanmeigouni et al., 2016, 2018).
For this type of network, in fact, the new and increasingly developed monitoring systems
have made it possible to collect data with ever higher frequency and ever greater precision. In the
national networks, in fact, the so-called diagnostic trains are used intensively, equipped with
instruments capable of collecting hundreds of pieces of information even at over 200 km/h. Or,
again, technologies are beginning to be tested which in recent years are also spreading in the road
sector such as the integrated Wireless Sensors Networks, as widely described in (Di Graziano et al.,
2020). These sensors, some examples of which in the railway sector are shown in (Chapeleau et
al., 2015; Hodge et al., 2015; Velha et al., 2021), make it possible to create a continuous, real-time
and low-cost monitoring network through the integration of sensors within the elements of the
superstructure.
The same didn’t happened for local railway networks, where the differences in track key
factors influencing the degradation of track quality (dynamic forces, axle loads, speeds and track

104
CHAPTER 3 The role of local railway features in risk assessment

component characteristics) (Ferreira & Murray, 1997; Lyngby, 2009) makes the transfer of
maintenance strategies and limits implemented for standard railways non optimized (Andrade &
Teixeira, 2010; López-Pita et al., 2008).
Therefore, there is a need among local railways Infrastructure Managers to better
understand the mechanisms underlying the degradation behaviour of rail track quality and the
role of the peculiar characteristics of these types of networks hac a ušić 0 7 .
To do this and to lay the foundations for the analyses better described in the next Chapter,
it is essential to provide the main elements for understanding how track geometry is defined and
is measured, what factors influence it and how is linked to the risk of the system.

3.5.2.1 Railway track geometry degradation

The research has investigated standard railways degradation process as early as the 1980s
(Andrade & Teixeira, 2015; Corbin & Fazio, 1981; Hamid & Gross, 1981), driven by the need for
Infrastructure Managers to know the asset condition and to optimize maintenance activities and
costs.
Rail quality degradation is a failure process that, if not treated, can lead to defect and faults
(Elkhoury et al., 2018). Degradation can affect both structural quality, i.e. the performance of the
elements of the track, and the geometry quality (Soleimanmeigouni et al., 2016), i.e. the three-
dimensional position of the rails. In particular, the presence of geometric defects influences
comfort, safety and management costs (Vale & M. Lurdes, 2013) and, thus, need to be carefully
kept under control.
Track geometry is identified by a set of parameters related to the vertical and lateral
position of the rails. The EN 13848-1(EN 13848-1, 2019) distinguishes 5 parameters: Gauge,
Alignment, Longitudinal Level, Cross-Level and Twist. Gauge and Alignment describe the horizontal
position of the rails and, respectively, are the minimum distance between the inner surface of the
rails and the horizontal deviation of the track centreline(Soleimanmeigouni et al., 2018). The
vertical position of the rail is identified by the Longitudinal Level, i.e. the geometry of the track
centreline projected onto the longitudinal vertical plane, by the Cross-Level, i.e. the height
difference between adjacent running tables, and the Twist, i.e. the algebraic difference between
two Cross-Level measured on a specific base. Each of the aforementioned parameters is explained
in detail by the EN 13848(EN 13848-1, 2019)

105
CHAPTER 3 The role of local railway features in risk assessment

Every geometry parameter is expressed by indicators (Track Quality Indexes, TQIs) of

different degree of aggregation of information. For long-term planning a Macro-analysis are
preferred indicators with a higher degree of aggregation like Mean or Standard Deviation (STD)(EN
13848-6, 2021) or combinations of the STD of multiple parameters (Alemazkoor et al., 2018;
Andrade & Teixeira, 2010; Chang et al., 2010; El-Sibaie & Zhang, 2004; Falamarzi et al., 2019b;
Soleimanmeigouni et al., 2016). Micro-analysis, instead, requires discrete indicators as it is linked
to the analysis of single defect occurrence and development (Ahac & a ušić 0 7 .
The EN 13848-5(EN 13848-5, 2017) prescribes limitations both on the peak amplitude
value and on the aggregate indexes. Even though the EN standard does not apply directly to local
railways, it still constitutes a solid reference for the definition of specific limits for these networks
(ANSF, 2019b).
Three thresholds are defined according to the severity of the defect and the actions to be
taken: Alert Limit (AL), Intervention Limit (IL) and Immediate Action Limit (IAL).
AL or preventive maintenance limit is the limit that, if exceeded, requires the defect to be
analysed and the maintenance actions to be properly planned. IL or corrective maintenance limit is
that threshold that if exceeded requires maintenance actions to bring the defect within the safety
limits and prevent reaching the IAL. IAL or safety limits, is the threshold that if exceeded by a
defect involves high safety risks and restrictions on the operation or closure of the line
(Soleimanmeigouni, 2019).
Track quality can exceed safety limits due to the degradation linked to exploitation,
measured by the number of trains or Million Gross Tons (MGT) (Andrade & Teixeira, 2010) passed
on the line.
Track geometry degradation is influenced by numerous and complex factors and the
analysis approaches are various (Andrade & Teixeira, 2015; Soleimanmeigouni et al., 2016;
Yousefikia et al., 2014). In general, the models implemented range from mechanistic approaches,
which involves the study of the properties of track components, forces, stresses and deformations,
to empiric ones, where geometry measurements are studied and degradation is described by
statistical models hac a ušić 0 7; oleimanmeigouni et al. 0 ; Yousefi ia et al. 0 .
While mechanistic models provide a precise definition of the role of the elements of the track in
degradation (Soleimanmeigouni et al., 2016), empirical models are able to model uncertainty and
complex relations of the elements responsible for degradation through statistical and probabilistic

106
CHAPTER 3 The role of local railway features in risk assessment

analysis. The main limitation of empirical approaches is linked to data quality, affecting the results
quality by the datasets resolution and errors of different nature (Esveld, 2001).
To date the literature has not yet investigated the degradation behaviour in local railways
but similar problems have been addressed for narrow-gauge railways like Light Rail Transit (LRT)
and tram systems. hac and a ušić (Ahac & Lakusic, 2015) highlighted the problem of transferring
the knowledge accumulated in standard railways to narrow-gauge systems and laid the
foundations for predictive maintenance for tram track degradation by defining a mechanistic-
empirical model based on the modelling methodologies adopted for standard gauge
constructions. In (Yousefikia et al., 2014) a Markov model for track deterioration is applied to
Melbourne tram data for the determination of track conditions and optimal maintenance
operations. The importance of the factors influencing the degradation of light rail tracks and their
relationships is analysed in (Moridpour et al., 2016) and a high accuracy Artificial Neural Network
(ANN) model is proposed to predict gauge degradation. ANN and regression model for gauge
degradation of a tram system prediction are compared in (Falamarzi et al., 2017). In (Ahac &
a ušić 0 7 the effects of tram track design and construction elements and exploitation
characteristics on gauge degradation are evaluated and a deterministic mechanistic empirical
approach and statistical analysis in track degradation modelling are described. Data from
accelerometers are used to predict tram track degradation index in (Bocz et al., 2018) and
(Falamarzi et al., 2019b). A similar problem is addressed for LRTs in (Camacho et al., 2016) where
measured data from a track recording car are used to implement a regression model for
degradation rate and intervention effectiveness analysis.
According to the literature review, no specific study has been conducted on the
degradation behaviour of local railways. Considering that transferring the findings on other system
to local ones may be unreliable a specific study is required for proper maintenance planning.

3.5.3 Effect of local railway rolling stocks and operation features on risk

The weight and speed of rolling stock are two critical factors that influence the level of risk
associated with railway operations. These factors affect the safety of railway systems by impacting
the train's braking distance, track maintenance requirements, and the likelihood of accidents,
including derailments and collisions.

107
CHAPTER 3 The role of local railway features in risk assessment

Train stopping distance plays a critical role in ensuring the safety of railway operations
ubotić Vasiljević 0 . It refers to the distance required for a train to come to a complete
stop after the hazard is detected, sum of the space covered during the reaction and the braking
distance. The braking distance of a train is influenced by several factors, including the train's
speed, weight, braking system, and track conditions. Accurate braking distance calculations help to
prevent collisions, derailments, and other accidents. Therefore, precise train braking distance
calculations are essential for maintaining the safety and efficiency of railway operations.
In the context of local railways, it is particularly crucial to consider the unique challenges
posed by these networks. While various empirical computation models for train braking distance
have been developed over time, these models are limited in their applicability and typically
require the definition of multiple parameters. Additionally, the models in the literature have
largely been developed for national railways, which have very different characteristics in terms of
geometry and speed compared to local railways. As such, there is a need for local railways to
develop precise and reliable calculation tools that account for the specific factors of the trains,
braking systems, and unique characteristics of the routes on which they operate.

3.5.3.1 Definitions and models for train stopping distance

The train stopping distance, which is the distance that a train travels from when the driver applies
the brake until it comes to a complete stop, depends on several factors, including the train's speed
when the braking system is activated, the coefficient of friction between the wheel and rail, the
delay in the brake's activation, the state of wear of the brake pads, the air pressure in the brake
cylinders, the slope of the track, and the mass distribution of the train (Profillidis, 2017).
Due to these numerous variables, it can be challenging to calculate the precise stopping
distance, which can vary significantly based on the train and environmental conditions.
One method to calculate the stopping distance is to solve the general equation of motion,
which requires extensive knowledge of the braking system and a large number of parameters.
However, an alternative approach is to use empirical or semi-empirical parameters, which
simplifies the calculation of the stopping distance. It is important to note that these models are
typically developed based on a specific type of train.
In general, train stopping distance can be expressed as Eq.(3.16):
𝐿𝑠𝑑 = 𝐿𝑟𝑑 + 𝐿𝑏𝑑 (3.16)

108
CHAPTER 3 The role of local railway features in risk assessment

where:

• 𝐿𝑠𝑑 [m] is train stopping distance;

• 𝐿𝑟𝑑 [m] is reaction distance;
• 𝐿𝑏𝑑 [m] is braking distance.

The reaction distance is the length travelled by the train during the driver's reaction time
and is expressed as shown in Eq. (3.17):
𝐿𝑟𝑑 = 𝑡𝑟𝑡 ⋅ 𝑣 (3.17)
where:

• 𝑡𝑟𝑡 [s] is reaction time;

• 𝑣 [m/s] is the train movement speed.
• When studying motion, the vehicle can be examined either as a rigid body or, even more
accurately, as a material point. To analyse motion, Figure 3.13 illustrates the
representation of the vehicle as a material point.

Figure 3.13. Representation of the vehicle as a material point for the study of motion

The expression of the general equation of motion for a point can be expressed as equation
(3.18):

𝑑𝑣
𝑇 − 𝑅 = 𝑀𝑒 ⋅ (3.18)
𝑑𝑡
Where:

• 𝑇 [N] is the sum of all active forces;

• 𝑅 [N] is the sum of the resistances;
• 𝑀𝑒 [kg] is the equivalent mass which takes into account the rotating masses that need to
be decelerated, the sum of the mass of the vehicle or train and the contribution deriving
from the rotating inertia (wheels and/or other rotating parts), 𝑀𝑒=P/g(1+𝛽);
𝑑𝑣
• [𝑚/𝑠 2 ] is the acceleration.
𝑑𝑡

109
CHAPTER 3 The role of local railway features in risk assessment

Since the traction force T is zero during braking (T=0), a braking force (−𝐹𝑓 ) is added to the
resistances. The equation is the following (3.19):

𝑑𝑣
−𝑅 − 𝐹𝑓 = −𝑀𝑒 ⋅ (3.19)
𝑑𝑡
Considering all the involved resistances, assuming that the coefficient of friction varies
linearly and solving the equation the (3.20) is obtained:

1 1+𝛽 𝑣2
𝐿𝑏𝑑 = ⋅ ⋅ ⋅ (3.20)
2 𝑔 12,96(𝐾 ⋅ 𝑓 + 𝑟 + 𝑖)
where:

• 𝐿𝑏𝑑 [m] is train stopping distance;

• 𝛽 is the coefficient that takes into account the rotating masses;
• 𝑔 [𝑚/𝑠 2 ] is the acceleration of gravity;
• 𝑣 [𝑘𝑚 / ℎ] is the speed;
• 𝐾 is the percentage of real braked mass;
• 𝑟 is the coefficient which takes into account the rolling resistance;
• 𝑖 is the slope

The condition for the wheel to rotate without slipping in that the braking force satisfy the
(3.21):
𝐹𝑓 ≤ 𝑓 ⋅ 𝑃 (3.21)
where:

• 𝑓 is the coefficient of friction in the wheel-rail contact;

• 𝑃 [N] is the weight force;

If equation (3.21) is not verified the wheel translates without rotating. It is best to avoid
this situation as it not only increases the braking distance but also causes abnormal wear and tear
on the wheels. The friction coefficient, depend both from the conditions at the wheel-rail contact
and from the speed through non-linear relations. In this case, the resolution can be obtained
numerically using one of the many existing methods for solving ordinary differential equations
(ODE). Table 3.7 shows the models present in the literature that can be used for the calculation of
the wheel-rail friction coefficient (Yuan et al., 2021).

110
CHAPTER 3 The role of local railway features in risk assessment

Table 3.7. Models that can be used for the calculation of the wheel-rail friction coefficient.
N° Friction models References
∗
𝑓𝑠
1 𝑓= (1) (Bochet H, 1858)
1 + 0,03 ⋅ 𝑣
𝑓𝑠 ∗∗
2 𝑓= (2) (Chen HC, 1997)
1 + 0,23 ⋅ 𝑣
𝑓𝑠 (𝑣)∗∗∗
3 𝑓= (3) (W. Zhang et al., 2002)
1 + 𝛼(𝑣) ⋅ 𝑣
0,45
4 𝑓 = 0,15 + (Vollebregt & Schuttelaars, 2012)
3+𝑣
0,408
5 𝑓= (Chen K, 2017)
1 + 0,11 ⋅ 𝑣
0,03 15 (Croft et al., 2011; Vollebregt &
6 𝑓= + 2
0,2 + 𝑣 100 + 𝑣 Schuttelaars, 2012; Xie et al., 2006)
0,3 15
7 𝑓= + (Croft et al., 2012)
2 + 𝑣 100 + 𝑣 2
*very dry, 𝑓𝑠 = 0,31; dry, 𝑓𝑠 = 0,22; moist, 𝑓𝑠 = 0,14;
**𝑓𝑠 = 0,3;
***𝑓𝑠 (𝑣) and 𝛼(𝑣) are the functions of the speed and axle load.

3.5.3.2 Empirical stopping distance calculation models

While solving general equation of motion is often seen as more robust and accurate, in some cases
is preferred to apply empirical models for their advantages. Empirical models are often easier and
quicker to develop, require less specialized knowledge, and can capture complex, nonlinear
relationships that are difficult to model mechanistically.
Many empirical models for the determination of the train braking distance are presented in
literature, some more general, others specific for the type of train (freight and/or passengers) or
for a certain speeds.
The most commonly used include:

• Maison's formula;
• Pedeluq’s formula;
• Minden's formula;
• Belgian railways formula;
• Italian Infrastructure Manager formula.

The Maison's formula (Profillidis, 2017) was developed for freight trains with speed v<70
km/h and expressed by equation (3.22) as follows:

111
CHAPTER 3 The role of local railway features in risk assessment

4,24 ⋅ 𝑣 2
𝐿𝑏𝑑 =
1,00 ⋅ 𝜑 ⋅ 𝜆 + 0,0006 ⋅ 𝑣 2 + 3 − 𝑖 (3.22)

where:

• 𝐿𝑏𝑑 [m] is train stopping distance;

• 𝑣 [km/h] i strain speed;
• 𝑖 is trac gradient ‰ or in mm/m . Trac gradient is regarded positive downhill and
negative uphill;
• 𝜑 is the friction coefficient depending on gradient. Values of 𝜑 are: 𝜑=0 0 for i< ‰ 𝜑
=0 0÷0 00 for i> ‰;
• 𝜆 is braking percentage, defined as the ratio of the braking weight to total vehicle weight
and expressing the braking force required for braking one ton.

Braking percentage 𝜆 is a critical factor for the braking distance. Table 3.8 gives values of 𝜆
for various types of rolling stock and brakes. In any case equation (3.22) gives the possibility to
calculate the braking percentage 𝜆 in relation to the braking distance 𝐿𝑏𝑑 , the train speed 𝑣, the
gradient 𝑖 and the friction coefficient 𝜑.

Table 3.8. Values of λ for various types of rolling stoc and bra es(Profillidis V.A., 2014).
Type of braking Braking percentage 𝝀
Locomotives with axle load 𝑃 = 15 ÷ 20𝑡 80 ÷ 95%
Normal braking
Hauled vehicles with axle load 𝑃 = 15 ÷ 20𝑡 60 ÷ 90%
Locomotives vehicles 160 ÷ 220%
Emergency braking
Hauled vehicles 130 ÷ 220%

The Pedeluq’s empirical formula (Profillidis, 2017) changes for passenger trains (speed
between 70 and 140 km/h) and Diesel-electric passenger trains.
For the former braking distance is expressed as shown in Eq. (3.23):
𝜑 ⋅ 𝑣2
𝐿𝑏𝑑 =
1,09375 ⋅ 𝜆 + 0,127 − 0,235 ⋅ 𝑖 ⋅ 𝜑 (3.23)

where:

• 𝑖 is trac gradient ‰ or in mm/m . Trac gradient is regarded positive downhill and

negative uphill;

112
CHAPTER 3 The role of local railway features in risk assessment

• 𝜑 is the friction coefficient depending on gradient equal to 0 0 for i< ‰ and

0 0÷0 00 for i> ‰;
• 𝜆 is braking percentage, defined as the ratio of the braking weight to total vehicle weight
and expressing the braking force required for braking one ton.

For diesel-electric passenger trains Pedeluq’s model is expressed by Eq. (3.24):

0,0386 ⋅ 𝑣 2
𝐿𝑏𝑑 = (3.24)
𝑖
𝛾 − 100
where γ [ 𝑚/𝑠 2 ] is the deceleration.
Minden formula (Profillidis, 2017), commonly used in Germany, varies for passenger and
freight trains. For passengers trains is expressed by Eq. (3.25):
3,85 ⋅ 𝑣 2
𝐿𝑏𝑑 = (3.25)
𝜆
[6,1 ⋅ 𝜓 ⋅ (1 + 10)] + 𝑖
where:

• 𝜓 is a parameter in relation to the brake type characteristics. It values between 0,5÷1,25;

• 𝜆 is the braking percentage, defined as the ratio between the braking weight and the total
weight of the vehicle;
• 𝑖 is the gradient of the railway line.

For freight trains stopping distance is evaluated through Eq. (3.26):

3,85 ⋅ 𝑣 2
𝐿𝑏𝑑 =
5,1 ⋅ 𝜓 ⋅ √𝜆 − 5 + 𝑖 (3.26)

Eq. (3.27) is the reference for the Belgian national railways (Profillidis, 2017) for the
determination of the braking distance:
4,24 ⋅ 𝑣 2
𝐿𝑏𝑑 =
57,5 ⋅ 𝑣 (3.27)
[𝜆 ( 𝑣 − 20 )] + 0,05 ⋅ 𝑣 − 𝑖

where:

• 𝜆 is the braking percentage, defined as the ratio between the braking weight and the total
weight of the vehicle;
• 𝑖 is the gradient of the railway line.

113
CHAPTER 3 The role of local railway features in risk assessment

In Italy, Ministerial Circular CM 26 of 1971 (RFI, 1971) provides the following expression for
the calculation of the braking distance (3.28):
𝑣
𝐿𝑏𝑑 =
254 ⋅ 𝐾 ⋅ 𝑓 (3.28)

where:

• 𝐾 is the percentage of braked mass;

• 𝑓 is the coefficient of friction between the braking element and the friction surface;
• 𝑣 [m/s] is initial speed.

All the empirical methodologies presented thus far have been validated through the
extensive usage in standard railways. As shown above, many formulas are applicable for speeds
greater than 70 km/h, which usually corresponds to the maximum achievable velocity on local
railways. Moreover, resolving the general equation of motion can be impractical for decision-
making models and risk assessments on railway lines.
Hence, it is essential to expand the scope of application of empirical models to encompass
speeds and loads characteristic of isolated railways. This can be achieved through dedicated
investigations focused on this specific type of railway lines.

3.5.4 Planning and resources optimization under budget constraint

One of the main obstacles that local railways infrastructure managers face is the optimal allocation
of limited resources. Unlike national railways, local railways operate on a smaller scale and serve a
more limited population, which often translates into a smaller revenue stream. In fact, passenger
traffic on national networks can exceed the number of passengers transported each year in local
networks by two or three orders of magnitude (ANSF, 2019b; European Railway Agency, 2018). For
this reason, local railways may not have access to the same level of funding as larger, national
railways, making it challenging to keep up with necessary investments.
As a result, local railways often face budget constraints that limit their ability to invest in
essential upgrades or repairs. This can lead to a decline in the quality of service, increased
downtime due to maintenance issues, and safety concerns.

114
CHAPTER 3 The role of local railway features in risk assessment

Therefore, in order to be able to ensure high levels of service safety also in local railways
against an efficiency of investments, it is essential to adequately use the results obtained from the
risk assessment.
In this sense, Cost-Benefit Analysis (CBA) is a widely used tool for evaluating the
introduction of new risk reducing measures Špač ová traub 0 and in particular in the
railway industry (Ben Aoun et al., 2010; Mahboob et al., 2015; Rezvani et al., 2015). BCA is a
framework for comparing the benefits of a proposed project against its costs. The benefits are
estimated in terms of the value that they bring to the users of the system, such as improved safety
or faster travel times, while the costs are estimated in terms of the financial outlays required to
implement the measures. The resulting ratio of benefits to costs, known as the benefit-cost ratio
(BCR), is used to determine whether the proposed project is worth undertaking.
CBA is a critical tool to support decision-making and ensure that investments in rail
infrastructure and services are economically viable and provide maximum benefit to stakeholders.
CBA allows rail operators and policymakers to prioritize investment decisions, allocate resources
efficiently, and optimize the performance and sustainability of the rail network.
From the point of view of a short-medium term time horizon, the advantage of using CBA
for evaluating risk reducing measures in a railway system is that it provides an immediate,
systematic and objective method for decision-making. By weighing the potential benefits against
the costs, decision-makers can make informed choices about the best use of resources.
However, CBA is less effective for long term decisions, where intangible difficult to
monetize costs and benefits (environmental impact, social impacts, etc.) take on more
importance.
The first step in conducting a CBA for new safety measures in the railway industry is to
identify the potential costs and benefits. This may include the reduction of accidents, injuries, and
fatalities, as well as the reduction of potential liability costs. The direct and indirect costs of
implementing the new safety measures should also be considered, including the cost of new
equipment, training of personnel, and maintenance.
Once the costs and benefits have been identified, they need to be quantified and
monetized. This involves assigning a financial value to each cost and benefit, which allows for a
direct comparison of the two.

115
CHAPTER 3 The role of local railway features in risk assessment

The monetary value of benefits can be calculated by evaluating the likely cost savings or
other economic benefits that the new safety measures would provide. The monetary value of
costs may include the cost of equipment, engineering, and construction, and the impact on the
project's schedule.
One of the challenging aspects of determining the effectiveness of safety measures
involves quantifying the monetary value of preventing fatalities or the Value of Statistical Life
(VSL). Various approaches have been proposed in the literature to address this issue, such as the
ones based on the evaluation of the Loss of Production (Hanly & Sharp, 2014) or based on the
Willingness To Pay (Nimdet & Ngorsuraches, 2015; Ryen & Svensson, 2015), etc.
Once costs and benefits have been standardized for each system security improvement
strategy, the subsequent task is to prioritize and identify the most suitable interventions to
implement.
For each safety measure to be introduced, ordered according to increasing cost, the ratio
between the increase in benefits over the increase in costs (BCR) is evaluated, as shown in (3.29).

𝐵𝑒𝑛𝑒𝑓𝑖𝑡𝑠
𝐵𝐶𝑅 = (3.29)
𝐶𝑜𝑠𝑡𝑠
The same ratio can be written in terms of risk as follows.

𝑅𝑖𝑠𝑘 𝑟𝑒𝑑𝑢𝑐𝑡𝑖𝑜𝑛
𝐵𝐶𝑅 = (3.30)
𝐶𝑜𝑠𝑡𝑠
The strategy with the higher BCR is the better strategy.
By using the benefit-cost ratio, decision-makers are able to identify the interventions that
will provide the greatest benefits for the least amount of cost. This can help to ensure that
resources are allocated in a way that maximizes impact and overall effectiveness.
Once the first strategy to be implemented has been defined, in the evaluation of the
subsequent compatible alternatives, the new reference system is given by the analysed system
updated with the risk reduction strategy just found.
To evaluate the subsequent optimal action for risk reduction, the CBR analysis is repeated
for the remaining strategies. The procedure is iterative in which the system is updated with the
optimal strategy found in the previous iteration.
As shown in Figure 3.14, the progressive introduction of security measures defines an
increasing level of system protection. More numerous or more performing safety barriers lead to a
reduction in the risk level of the system with a corresponding increase in costs to be incurred.

116
CHAPTER 3 The role of local railway features in risk assessment

The optimal strategy identifies the minimum of the sum between intervention costs and
expected risk:

𝑂𝑝𝑡𝑖𝑚𝑚𝑎𝑙 𝑠𝑡𝑟𝑎𝑡𝑒𝑔𝑦 ≡ 𝑚𝑖𝑛[𝐶 + 𝑅] (3.31)

udget
costrained
op mum
Cost Ris

Op mum

Protec on level/ afety barriers

Figure 3.14.Cost-benefit optimization problem. Adapted from Špač ová traub 0

However, if the available budget is lower than that required to satisfy the (3.31), he
optimal solution is constrained by the maximum sustainable cost, bringing an higher risk level for
the system.
In conclusion, the limited resources faced by local railways infrastructure managers can
make it challenging to keep up with necessary investments. Benefit-Cost Analysis (BCA) is a critical
tool to support decision-making and ensure that investments in rail infrastructure and services are
economically viable and provide maximum benefit to stakeholders. The use of BCA provides an
immediate, systematic and objective method for decision-making, allowing decision-makers to
make informed choices about the best use of resources. However, the difficulty in monetizing
intangible costs and benefits can limit its effectiveness for long-term decisions, where multi-
criteria methodologies are more suitable (Odoki et al., 2015). Through the iterative process of
identifying and evaluating the optimal strategy for risk reduction, the infrastructure manager can
achieve the minimum of the sum between intervention costs and expected risk, taking into
account the budget available and the maximum sustainable expenditure. The use of BCA allows for

117
CHAPTER 3 The role of local railway features in risk assessment

efficient allocation of resources and optimized performance and sustainability of the rail network,
ensuring high levels of safety in local railways.

118
4. RISK MANAGEMENT FRAMEWORK APPLICATION IN
AN ITALIAN LOCAL RAILWAY
CHAPTER 4 Introduction

4.1 INTRODUCTION

Figure 4.1. Topics of Chapter 4

Exposing a case study allows to provide practical examples of how a theoretical framework can be
applied in real-world scenarios. The advantages of using a case study to understand a framework
are numerous. Firstly, case studies provide an in-depth understanding of the application of the
framework. Additionally, case studies provide a context for the framework, which enables the
researcher to evaluate its effectiveness in different scenarios.
The valuable and profitable agreement between Ferrovia Circumetnea (FCE) and the
Department of Architecture and Civil Engineering of the University of Catania allowed to apply the
Risk Management framework proposed in this work in the local railway line managed by FCE. The
case study is a narrow-gauge railway that connects several small towns on the slopes of Mount
Etna, Sicily. This railway is an essential transport system for local residents and tourists. The local
railway is a challenging system to manage due to its location and topography, and its management
requires a comprehensive understanding of the factors that impact its operations.
The main characteristics of the Ferrovia Circumetnea are described in this work. These
include the railway's history, its physical characteristics, and its operational aspects.
Understanding these characteristics is crucial to comprehending the challenges faced by the
infrastructure manager.
The framework presented in this work is applied to two specific cases in the Ferrovia
Circumetnea. The first case study focuses on the tunnels, where data from similar systems had to

120
CHAPTER 4 Introduction

be calibrated, and possible improvement actions were evaluated using cost-benefit analysis. The
second case study focuses on Level Crossing accidents, where the elements to be managed are
numerous and vastly different in terms of safety.
The last sections investigates, starting from field-measured data, the effect of the different
characteristics of the infrastructure and rolling stock on risk.

121
CHAPTER 4 Circumetnea Railway

4.2 CIRCUMETNEA RAILWAY

The Circumetnea Railway (FCE) is a narrow gauge (950 mm) Diesel traction railway line, about 110
km long, linking the city of Catania and Riposto, following a circular route connecting the main
inhabited centres around the volcano Etna .
The line, in addition to its important passenger transport service, has a strong historical
and tourist value. The following paragraphs aims to describe the historical evolution of the line
and the main characteristics central for risk management.

4.2.1 Story and characteristics

The line operates through the Etna area for over a hundred years. The concession of the works to
the Sicilian Society of Public Works dates back to 1889 and the inauguration of the first section,
the Catania Borgo-Adrano, to 18951. The remaining part of the line was finished in the following
few years.
Initially the service was carried out by steam locomotives spaced by a telegraph block
system between station and station with average speeds between 20 and 27 km/h.
Initially the traffic was mixed passengers and goods, thanks also to the connection with the
port of Catania, which is no longer used today.
Over the years, the line has undergone numerous changes linked to the need for
modernization and to the eruptive activity of Etna. One of the most important changes that
affected the historic route of the Circumetnea railway, started in the 80s of the last century,
involved some adjustments and construction underground of part of the line through the
construction of three new tunnels. The signalling system has been entrusted to the axle counter
block (BCA) between station and station, with the exception of a few parts of the low-traffic line
where the telephone block still exists.
The stretching of the line and the transition to diesel traction made it possible to improve
the railway service and reduce travel times, reaching maximum permitted speeds of 70 km/h and
commercial speeds of 40 km/h.

1
Source: FCE, The History of the Company, link: https://www.circumetnea.it/la-storia/# (last access 12/04/2023)

122
CHAPTER 4 Circumetnea Railway

The difficulties arising from the damage caused by the Second World War were one of the
main factors that placed the management of the company under the direction of the Italian
Ministry of Infrastructure and Transport.
Since the last years of the 1900s, a series of modernizations have been introduced,
generating a gradual transformation of the railway line into a double-track underground line,
standard gauge and 3kV direct current electric traction between Catania airport and Paternò city.
The operational line today is shown in Figure 4.2.

Figure 4.2. Railway and underground line and related stops and stations managed by FCE

4.2.2 The line

The FCE consists of a 110 km single-track line with a narrow gauge of 950 mm. The line operates in
heterogeneous conditions, going from zero to around 1000 m above sea level, through urban,

123
CHAPTER 4 Circumetnea Railway

suburban and rural environments. To overcome the external environment obstacles, the service is
carried out by smaller and lighter trains capable of dealing with high gradients up to 0‰ small
radius curves, up to 80 m and maximum permissible speed of 50 km/h. The line is a ballasted track
with 50 UNI and 36 UNI rails linked by concrete and wooden sleeper.
The line is operated in an unbalanced way, with traffic of approximately 19,000 trains/year
in the first 20 km of line, half of the trains in the subsequent 50 km and around one-sixth in the
remaining portion.
In particular, the operation characteristic are exposed below:
• Traffic Management: Operational Control Centre
• Control and regulation of traffic: Centralized Traffic Control (CTC) and Axle-counter Block
system between stations;
• Max allowed speed: Vmax=50 Km/h;
• Commercial speed: Vcom=40 Km/h;
• Daily traffic: 43 train/day;
• Annual passenger traffic: 649 – 718 pax per train-km

4.2.3 The tunnels

The narrow gauge typical of local railways allowed to design the route in mountainous areas
without tunnels or major civil works. For this reason, the historic line did not have long tunnels
along its extension.
With the recent modernization works, 3 tunnels longer than 2 km were built with the aim
of making the route of the line straighter and faster. These three tunnels develop under three
cities of the territory of the volcano Etna, specifically, under Adrano the hereinafter referred as
Tunnel 1 is 2.07 km long, under Biancavilla the hereinafter referred as Tunnel 2 of 3.17 km and
the one under Santa Maria di Licodia, called Tunnel 3, of about 2.50 km.
The three tunnels accepting the single track layout are made in a natural tunnel, with short
stretches in an artificial tunnel.
Their section of the three tunnels is represented by an arc of a circle having a radius of 3.49
m of 8.37 m in width and 7.3 m in heigh and an area of approximately 38 m2.

124
CHAPTER 4 Circumetnea Railway

The railway track is almost constantly uphill from the South (Santa Maria di Licodia) to the
North drano without slope inversions with gradients between ‰ and ‰.

4.2.4 The level crossings

The FCE line intercepts numerous inhabited centres along its route and various roads which
required the creation of several level crossings along the route. Today along the 110km line there
are 96 active level crossing whose characteristics and information are provided by FCE
documentation and by the main planning tools of the territory Catania.
A total 78 level crossings are protected with active safety measures, i.e. where there are
systems capable of protecting and/or warning users when the train is passing, and 18 level
crossings with passive safety measures.
Table 4.1 shows the complete list of level crossings analysed with the main characteristics
useful for characterization in terms of safety.

Table 4.1. Features of the LC along the line

Railway
Type of Optical and acoustic
n LC name Type of Barriers signalling
Manouver warnings
system

1 Catania Borgo Via Caronda Manual Full barriers Yes Yes

2 Catania Borgo Via Empedocle Manual Full barriers Yes Yes
3 Casello 5 bis (Via S. Sofia) Automatic Full barriers Yes Yes
4 Cibali I° (via S. Giovanni G.) Automatic Open crossing Yes Yes
5 Cibali II° Automatic Open crossing Yes Yes
6 PLA 6 Automatic Full barriers Yes Yes
7 PLA 6a Automatic Full barriers Yes Yes
8 Casello 6 bis Automatic Full barriers Yes Yes
9 PLA Nesima (via Amari) Automatic Half barriers Yes Yes
10 Casello 7 (Lineri) Automatic Full barriers Yes Yes
11 Garitta 8 Automatic Full barriers Yes Yes
12 Garitta 9 Automatic Full barriers Yes Yes
13 Casello 10 Automatic Full barriers Yes Yes
14 PL Misterbianco 1 Manual Full barriers Yes No
15 PL Misterbianco 2 Manual Full barriers Yes No
16 Casello 11 Automatic Full barriers Yes Yes
17 Casello 12 Automatic Full barriers Yes Yes
18 Casello 14 I° Automatic Full barriers Yes Yes
19 Casello 14 II° Automatic Full barriers Yes Yes
20 Casello 14 III° Automatic Full barriers Yes Yes

125
CHAPTER 4 Circumetnea Railway

Railway
Type of Optical and acoustic
n LC name Type of Barriers signalling
Manouver warnings
system

21 Casello 14 IV° (ex km 013+441) Private Gate Other No

22 Casello 14 bis Automatic Full barriers Yes Yes
23 Garitta 15 Automatic Full barriers Yes Yes
24 PLA 15 bis (via Ferrarotto) Automatic Full barriers Yes Yes
25 Casello 16 Automatic Full barriers Yes Yes
26 Garitta 17 Automatic Full barriers Yes Yes
27 Garitta 18 Automatic Full barriers Yes Yes
28 Garitta 19 Automatic Half barriers Yes Yes
29 Valcorrente Automatic Full barriers Yes Yes
30 Casello 20 Automatic Open crossing Yes Yes
31 Giaconia Automatic Full barriers Yes Yes
32 Casello 23 Automatic Half barriers Yes Yes
33 Casello 24 Manual Full barriers Yes No
34 Casello 25 Automatic Open crossing Yes Yes
35 Casello 40 Automatic Full barriers Yes Yes
36 Casello 43 I° Automatic Full barriers Yes Yes
37 Casello 43 II° Automatic Full barriers Yes Yes
38 Casello 44 Automatic Full barriers Yes Yes
39 Passo Zingaro Nessuna Open crossing Other No
40 Casello 48 Automatic Open crossing Yes Yes
41 Casello 49 Automatic Full barriers Yes Yes
42 PL Garitta 59 Nessuna Open crossing Other No
43 PL 60 Nessuna Open crossing Other No
44 PL Casello 61 Nessuna Open crossing Other No
45 Casello 63 Manual Chain Yes No
46 PLA 63 bis Maletto Automatic Half barriers Yes Yes
47 Casello 64 Automatic Full barriers Yes Yes
48 PL 64 II° Nessuna Open crossing Other No
49 Casello 65 Automatic Half barriers Yes Yes
50 PL Tartaraci Nessuna Open crossing Other No
51 PL Gurrida Nessuna Open crossing Other No
52 Casello 70 Automatic Half barriers Yes Yes
53 Casello 71 Automatic Half barriers Yes Yes
54 Casello 72 Automatic Open crossing Yes Yes
55 Casello 73 Automatic Full barriers Yes Yes
56 Casello 75 Automatic Half barriers Yes Yes
57 Casello 75 II° (ex km 076+538) Nessuna Open crossing Other No
58 Casello 76 (Castello Romeo) Automatic Open crossing Yes Yes
59 Casello 76 bis (Montelag.) Automatic Full barriers Yes Yes
60 PL 76 Ter (Borgo San Nicola) Nessuna Open crossing Other No
61 Casello 77 Automatic Full barriers Yes Yes

126
CHAPTER 4 Circumetnea Railway

Railway
Type of Optical and acoustic
n LC name Type of Barriers signalling
Manouver warnings
system

62 PL 77 II° (Croci Monaci) Nessuna Open crossing Other No

63 PL 77 III° (Cardile) Nessuna Open crossing Other No
64 PL 77 IV° Nessuna Open crossing Other No
65 PLA Moio Passopisciaro Automatic Full barriers Yes Yes
66 Casello 78 Automatic Full barriers Yes Yes
67 PL Torre Palino Nessuna Open crossing Other No
68 Rovittello 81 I° Automatic Open crossing Yes Yes
69 Rovittello 81 II° Automatic Open crossing Yes Yes
70 Casello 81 Automatic Open crossing Yes Yes
71 Casello 83 Automatic Open crossing Yes Yes
72 Casello 84 Nessuna Open crossing Other No
73 Casello 85 Automatic Open crossing Yes Yes
74 Mare Neve (ex km 099+343) Automatic Full barriers Yes Yes
75 Casello 87 (ex km 099+424) Automatic Full barriers Yes Yes
76 Casello 88 (ex km 100+098) Automatic Full and half barr. Yes Yes
77 Casello 89 (ex km 101+497) Automatic Full barriers Yes Yes
78 Terremorte (ex km 102+106) Automatic Full barriers Yes Yes
79 Terremorte II° (ex km 102+361) Nessuna Open crossing Other No
80 Casello 90 (ex km 103+113) Automatic Full barriers Yes Yes
81 Garitta 90 bis (ex km 104+163) Automatic Full barriers Yes Yes
82 PLA 91 (ex km 104+500) Automatic Full barriers Yes Yes
83 Garitta 92 (ex km 105+268) Automatic Full barriers Yes Yes
84 PLA 93 (ex km 106+112) Automatic Full barriers Yes Yes
85 Casello 94 (ex km 107+774) Automatic Full barriers Yes Yes
86 S. Venera (ex km 109+904) Automatic Open crossing Yes Yes
87 Casello 96 (ex km 110+429) Automatic Open crossing Yes Yes
88 PL 96 bis (ex km 110+300) Automatic Half barriers Yes Yes
89 Casello 97 (ex km 111+626) Automatic Half barriers Yes Yes
90 Via Badia (ex km 112+637) Automatic Full barriers Yes Yes
91 Casello 100 (ex km 112+775) Automatic Full barriers Yes Yes
92 PL curva 141 (ex 113+407) Private Gate Other No
93 PLA Cutula (ex km 114+241) Automatic Full barriers Yes Yes
94 Casello 102 (ex km 115+934) Automatic Full barriers Yes Yes
95 PLA Villa di Giarre (ex km 117+147) Automatic Half barriers Yes Yes
96 Casello 105 (ex km 118+389) Automatic Full barriers Yes Yes

Table 4.2, instead, shows the information on speed, sight distance and rail and road traffic
for each LC.

127
CHAPTER 4 Circumetnea Railway

Table 4.2. Data on traffic speed and visibility distance in LCs

Local speed Average speed Rail traffic Sight distance Road
n NAME
[km/h] [km/h] [train/gg] (2018) [m] ADT
1 Catania Borgo Via Caronda 30 30 41 42 2188
2 Catania Borgo Via Empedocle 30 30 41 18 2188
3 Casello 5 bis (Via S. Sofia) 30 45 41 0 1514
4 Cibali I° (via S. Giovanni G.) 5 45 41 12 5040
5 Cibali II° 5 45 41 21 5041
6 PLA 6 45 45 41 25 6365
7 PLA 6a 45 45 41 25 2334
8 Casello 6 bis 45 45 41 25 M
9 PLA Nesima (via Amari) 45 45 41 20 A
10 Casello 7 (Lineri) 50 50 41 200 1933
11 Garitta 8 50 50 41 25 B
12 Garitta 9 50 50 41 25 M
13 Casello 10 30 50 41 62,5 3960
14 PL Misterbianco 1 30 30 41 210 2515
15 PL Misterbianco 2 30 30 41 9 2515
16 Casello 11 30 50 41 37,5 2515
17 Casello 12 50 50 41 75 8665
18 Casello 14 I° 30 50 41 0 B
19 Casello 14 II° 30 50 41 100 B
20 Casello 14 III° 30 50 41 100 B
21 Casello 14 IV° (ex km 013+441) 0 50 41 250 B
22 Casello 14 bis 50 50 41 90 B
23 Garitta 15 50 50 41 62,5 M
24 PLA 15 bis (via Ferrarotto) 50 50 41 240 B
25 Casello 16 50 50 41 42 3212
26 Garitta 17 50 50 41 150 M
27 Garitta 18 50 50 41 87,5 1698
28 Garitta 19 50 50 41 62,5 A
29 Valcorrente 50 50 41 75 M
30 Casello 20 45 45 41 300 BB
31 Giaconia 45 45 41 87,5 M
32 Casello 23 45 45 41 39 6407
33 Casello 24 30 30 29 39 M
34 Casello 25 45 45 29 37,5 B
35 Casello 40 30 50 29 15 M
36 Casello 43 I° 50 50 29 8 B
37 Casello 43 II° 30 50 29 250 2320
38 Casello 44 45 50 23 125 B
39 Passo Zingaro 0 50 23 108 B
40 Casello 48 45 50 23 75 B
41 Casello 49 45 50 23 100 B
42 PL Garitta 59 0 50 21 237,5 B
43 PL 60 0 45 21 362,5 B
44 PL Casello 61 45 45 21 350 B
45 Casello 63 30 30 21 50 B
46 PLA 63 bis Maletto 45 45 21 48 726
47 Casello 64 50 50 21 37,5 M
48 PL 64 II° 0 50 21 300 M
49 Casello 65 50 50 21 225 B
50 PL Tartaraci 0 30 21 125 B

128
CHAPTER 4 Circumetnea Railway

Local speed Average speed Rail traffic Sight distance Road

n NAME
[km/h] [km/h] [train/gg] (2018) [m] ADT
51 PL Gurrida 0 50 21 175 B
52 Casello 70 50 50 21 23 925
53 Casello 71 50 50 21 25 B
54 Casello 72 50 50 21 150 B
55 Casello 73 30 30 11 24 M
56 Casello 75 50 50 11 25 2188
57 Casello 75 II° (ex km 076+538) 0 50 11 175 B
58 Casello 76 (Castello Romeo) 50 50 11 50 B
59 Casello 76 bis (Montelag.) 50 50 11 37,5 B
60 PL 76 Ter (Borgo San Nicola) 0 50 11 25 B
61 Casello 77 50 50 11 25 B
62 PL 77 II° (Croci Monaci) 0 50 11 125 832
63 PL 77 III° (Cardile) 0 50 11 125 B
64 PL 77 IV° 0 50 11 25 B
65 PLA Moio Passopisciaro 50 50 11 30 B
66 Casello 78 50 50 11 100 M
67 PL Torre Palino 0 50 11 50 B
68 Rovittello 81 I° 45 45 11 99 B
69 Rovittello 81 II° 45 45 11 100 B
70 Casello 81 45 45 11 50 B
71 Casello 83 45 45 11 50 B
72 Casello 84 0 45 11 81 B
73 Casello 85 45 45 11 62,5 B
74 Strada Mare Neve (ex km 099+343) 45 45 11 62,5 M
75 Casello 87 (ex km 099+424) 45 45 11 249 B
76 Casello 88 (ex km 100+098) 45 45 11 25 B
77 Casello 89 (ex km 101+497) 45 45 11 100 B
78 PLA Terremorte (ex km 102+106) 45 45 11 125 2457
79 PL Terremorte II° (ex km 102+361) 0 45 11 125 B
80 Casello 90 (ex km 103+113) 30 45 11 100 2457
81 Garitta 90 bis (ex km 104+163) 45 45 11 50 B
82 PLA 91 (ex km 104+500) 45 45 11 25 2457
83 Garitta 92 (ex km 105+268) 30 45 11 62,5 B
84 PLA 93 (ex km 106+112) 45 45 11 100 1330
85 Casello 94 (ex km 107+774) 45 45 11 37,5 B
86 S. Venera (Via Presa) (ex km 109+904) 45 45 11 87,5 B
87 Casello 96 (ex km 110+429) 45 45 11 62,5 B
88 PL 96 bis (ex km 110+300) 45 45 11 150 B
89 Casello 97 (ex km 111+626) 45 45 11 75 B
90 Via Badia (ex km 112+637) 45 45 11 50 B
91 Casello 100 (ex km 112+775) 45 45 11 50 220
92 PL curva 141 (ex 113+407) 0 45 11 24 B
93 PLA Cutula (ex km 114+241) 30 45 11 37,5 B
94 Casello 102 (ex km 115+934) 45 45 11 75 B
95 PLA Villa di Giarre (ex km 117+147) 45 45 11 37,5 A
96 Casello 105 (ex km 118+389) 45 45 11 250 6946

129
CHAPTER 4 Circumetnea Railway

4.2.4.1 LC data sources

The efficiency of the decision-making process is strongly influenced by the quality of the data
available for analysis. For this reason FCE provided all the necessary information for the analysis
collected over the years regarding the system. The process and type of data collected are shown in
Figure 4.3 and in particular:

• General information: name of the LC, name of the section, name of the road, position
along the line, urban or suburban area;
• Level crossing protection system: type of barriers, type of manoeuvre, presence of optical
or sound signals on the road side, signalling on the railway side, train-road visibility
distance.
• Operation data: Average Annual Daily railway Traffic (AADTrail), Average Annual Daily road
Traffic (AADTroad), average train speed.
• Faults data: log of all events related to the LC that led to operating conditions of the LC
elements other than those expected.

Figure 4.3. Data and information for RM process implementation

130
CHAPTER 4 Circumetnea Railway

The first type of data is intended to identify the LCs and the road section concerned. The
second type of information is crucial for the classification of LCs according. The road operational
data was obtained by merging information from key spatial planning tools and simulation models
of traffic on the network as shown in §4.2.4.2, which were used to evaluate the average daily
annual traffic (AADTroad). Conversely, data regarding the railway was provided by the infrastructure
manager, which included AADTrail and estimated train speeds at the LC.
The reliability and availability data of the LC protection systems were sourced from the
fault database, which documented all relevant events detected during monitoring and
maintenance activities carried out on the line. The database reports all the failures of the LC
systems from 2018 to 2021 for a total of 1227 reports, each characterized by:

• Number, description and date of input of the report;

• Status of the report (i.e. pending, solved, etc.);
• Date of the report and the time to start and complete the repair;
• Name and type of the LC involved;
• Type of failure, part of the system involved, possible cause and repair;

The faults described have been filtered in order to exclude non-safety-related events, that
is, not able to influence the risk level of the LC. The relevant faults refer to the following
categories:

• Train detection pedal failure;

• Faults in logic, control and power systems;
• Vandalism;
• Barriers failure;
• Acoustic signals failure;
• Warning lights failure;
• Accidents due to rush or distraction of car drivers.

131
CHAPTER 4 Circumetnea Railway

Figure 4.4. Number of LC faults (N) and mean time to repair (MTTR) per part of the system

According to Figure 4.4, the most common fault experienced by the system is associated
with the barriers, specifically the "Barrier does not lower" fault. However, this event is typically
resolved within a few hours. On the other hand, vandalism is a significant concern as it requires
the most time to repair, despite its average frequency.

4.2.4.2 Road Network

The set of roads intersecting the railway line have mainly local or suburban characteristics, with a
strong variability in traffic. The analysis of the road transport system was carried out using
calculation methodologies for estimating the performances and impacts that characterize it, with
reference to, for example, the traffic volumes affecting the road infrastructures intersecting the
railway line. The definition of a model of such specificity is possible through the use of software
already on the market that requires a significant and specific characterization of input data. For
this reason, in this phase, data and models already used for the drafting of the main planning tools
available in the metropolitan area of Catania (Mobility Plan of the Regional Province of Catania,
2012) were used, with the aim of estimating traffic (ADT) on roads affected by level crossings.

132
CHAPTER 4 Circumetnea Railway

Figure 4.5. Scheme of the road network involved in the analysis with relative assignment of vehicular flows

For the roads in which the model was not useful in obtaining the necessary information
due to a lack of information, a qualitative estimate of traffic was carried out through expert
judgements by indicating the average value of the following ranges: very low (BB): less than 2
vehicles/day; low (B): between 2 and 500 vehicles/day; medium (M): between 500 and 3000
vehicles/day; high (A): over 3000 vehicles/day.

133
CHAPTER 4 Circumetnea Railway

4.2.5 Reference accidental data: Italian Railway Network

In light of the fact that local railways do not have a long history of accidents, it is necessary to
reference similar systems in order to identify potential risks and develop appropriate safety
measures. In the case study presented here, data on accidents were gathered from the operator
of the Italian national rail network, Rete Ferroviaria Italiana (RFI), as a reference for understanding
the safety implications of local railway systems. Through an analysis of accident data collected by
RFI, insights can be gained into the types of accidents that may occur on local railways and the
potential risks associated with these accidents. By leveraging this information, it is possible to
develop effective strategies for mitigating the risk of accidents and ensuring the safe operation of
local railway systems.
In general, the data relating to the network managed by RFI, provided by the same
operator, by the ANSFISA and by the ERAIL database, make it possible to collect sufficient
information to evaluate the Hazardous Events to be analysed. In particular, as highlighted by the
annual report prepared by ANSFISA for 2020 (ANSFISA, 2021), the significant accidents recorded in
the decade between 2010 and 2020 and referred to all the networks under the control of the
agency are summarized in Table 4.3.

Table 4.3. Significant accidents by type recorded in Italy between 2010 and 2020, edited from (ANSFISA, 2021)

2010-
2010
2011
2012
2013
2014
2015
2016
2017
2018
2019
2020
2020
%

Collisions (trains, obstacles) 2 6 7 4 9 5 4 2 6 4 4 53 5%

Derailments 3 3 5 6 4 3 2 5 8 5 8 52 5%
Level Crossing accidents 15 18 13 14 16 19 15 12 3 5 8 138 13%
Accident to persons involving
moving rolling stock, except 80 77 79 71 74 67 72 75 83 51 56 785 71%
suicides and attempted suicides
Fire on board rolling stock 2 0 1 2 0 2 1 3 3 4 0 18 2%
Others 3 2 1 1 6 2 5 7 6 6 13 52 5%
TOT 105 106 106 98 109 98 99 104 109 75 89 1098

Most of the deaths and injuries (71%) that occurred in the decade 2010-2020 are linked to
the investment by moving vehicles of people unduly present on the tracks. The high incidence of

134
CHAPTER 4 Circumetnea Railway

significant accidents is linked to the strong difference in masses at the time of the accident and,
therefore, to the high probability of injury or death for the people involved.
Despite the number and significance of these accidents, they do not constitute an hazard
for other users of the railway system and, therefore, have not been evaluated as significant for the
definition of the railway risk management framework.
Accidents at level crossings rank first (13% of all accidents with consequences for railway
users). These accidents are characterized by a high probability and a severity that is often limited
to the injury or death of road vehicle occupants. The numerous points of contact between rail and
road (just over 4000 on the Italian network (ANSFISA, 2021), which is constantly decreasing) and
the unpredictable behaviour of road drivers lead to the numbers highlighted by the accident
statistics..
The other main causes of death in the railway sector for railway passengers in the last
decade are linked to Derailments (5%), Collisions between trains or with obstacles (5%) and Fires
(2%).

4.2.5.1 Derailment rate

In order to analyse the derailment rate and obtain a representative value for each year, the data
extracted from the databases relating to the Italian network were interpolated with an
exponential function (4.1):

𝑓(𝑡) = 𝑎 ∙ 𝑒 𝑏𝑡 (4.1)

where:
• 𝑓(𝑡): derailment rate expressed as events/train-km;
• 𝑡: reference year;
• 𝑎, 𝑏: parameters of the regression exponential function.
The curve that best describes the trend of the data is shown in Figure 4.6 and is described
by Eq.(4.2):

𝑅𝑎𝑡𝑒𝑑𝑒𝑟,𝑃𝑇 = 2 ∙ 10−8 ∙ 𝑒 −0,054∙(𝑡−1995) (4.2)

135
CHAPTER 4 Circumetnea Railway

Figure 4.6. Derailment rate interpolating function

4.2.5.2 Collision rate

Collisions with railway vehicles and collisions with obstacles interfering with the loading gauge of
the trains are among the collision data reported in the database relating to the collision event.
The collisions of interest for this study are the first ones and show the trend in annual rates shown
in Figure 4.7. The number of accidents per year shows a reduction in occurrences for this type of
event, made clear by the absence of significant collisions in the database since 2007, thanks to the
impact of the modernization of infrastructures and operating procedures.

Figure 4.7. Collision rate interpolating function.

This reduction is mainly due to the introduction of the SCMT system on the national
network between 2002 and 2006. As a result, for an analysis in for security and considering the
effective performance of the signalling system used in local networks, it is useful to consider the

136
CHAPTER 4 Circumetnea Railway

average of the annual rates up to 2006 as a representative value of the collision rate, excluding
data from subsequent years.

4.2.5.3 Fire rate

As for the derailment, fire rates have been decreasing over time. With the aim of forecast the rate
to 2018, the data were interpolated with the exponential law (4.3):
𝑓(𝑡) = 𝑎 ∙ 𝑒 𝑏𝑡 (4.3)
where:

• 𝑓(𝑡): collision rate expressed in events/train-km;

• 𝑡: reference year;
• 𝑎, 𝑏: parameters of the regression exponential function.
The curve that best fits the data is shown in Figure 4.8 and is described by Eq.(4.4):

𝑅𝑎𝑡𝑒𝑓𝑖𝑟,𝑃𝑇 = 9 ∙ 10−9 ∙ 𝑒 −0,033∙(𝑡−1995) (4.4)

Figure 4.8. Fire rate interpolating function

137
CHAPTER 4 Circumetnea Railway

4.2.5.4 Level crossing accidental rate

Also in the case of accidents at level crossings, the rates have been decreasing over time. This can
be explained by the increasing efficiency of safety systems and by the ongoing process of LC
suppression. The trend of the data can be described by an exponential law of the type (4.5):

𝑓(𝑡) = 𝑎 ∙ 𝑒 𝑏𝑡 (4.5)

where:

• 𝑓(𝑡): collision rate expressed in events/train-km;

• 𝑡: reference year;
• 𝑎, 𝑏: parameters of the regression function.
The curve that best fits the data is shown in Figure 4.9 and is described by Eq.(4.6):

𝑅𝑎𝑡𝑒𝑖𝑛𝑐,𝑃𝑇 = 10−7 ∙ 𝑒 −0,087∙(𝑡−2002) (4.6)

Figure 4.9. Accident at level crossings rate interpolating function

138
CHAPTER 4 Risk Management for FCE tunnels

4.3 RISK MANAGEMENT FOR FCE TUNNELS

The application of the RM framework in a context characterized by a limited data set on accident
history is carried out within the context of three tunnels, identified as Tunnel 1, Tunnel 2, and
Tunnel 3, located along the local narrow-gauge railway line managed by FCE. The aim of this
application is to validate the methodology for calibrating accident data derived from similar
systems with respect to the safety characteristics of the case under study and to describe its use
for the evaluation of the priority and cost-effectiveness of potential interventions applicable to the
three tunnels. By applying this methodology, it is possible to identify the most critical areas and
potential risks within the tunnels and develop appropriate interventions to mitigate these risks.
Through an analysis of the accident data, the effectiveness of these interventions can be
evaluated, and appropriate adjustments made to ensure the safe operation of the local railway
system.

4.3.1 System definition

The system analysed consists of three single-tube and single-track tunnels. Traffic control and
command is entrusted to the CTC system which manages 43 trains/day circulating at a commercial
speed of 40 km/h, transporting approximately 700,000 passengers every year. The distancing
between trains is ensured by an Axel-Counter Block system between station and station and the
line is not equipped with ATP or signal repetition systems. The service is operated by three
different types of diesel-powered locomotives.
Along tunnels, fire detection systems and extinguishing systems are installed but only
Tunnel 2 is equipped with smoke extraction systems.
For the purposes of the RM process, the definition of the system took place through the
subdivision of the system into gradually simpler parts, as shown in Table 4.4, in accordance with
the subdivision present in the Technical Specifications of Interoperability relating 'safety in railway
tunnels' (European Commission, 2014) suitably adapted to the case in question.

139
CHAPTER 4 Risk Management for FCE tunnels

Table 4.4. System Characterization

SYSTEM SUBSYSTEM PARTS
Superstructure
Civil works
Infrastructure
Track
Systems
Vehicles
Rolling Stock
Systems
Local railway tunnel
Control, command and signalling
Train management
Passenger
Traffic operation and
management
management
Procedures and
emergency plans

4.3.2 Hazard identification

In the risk assessment phase, accident historical data of the studied infrastructure were analysed.
Considering the rarity of the events studied, and therefore the lack of information, the
Hazard Log of the system was validated and integrated through comparisons with similar systems,
make it is possible to evaluate the hazards and the associated risk.
In the context of railway tunnels, and taking into account the characteristics of the three
tunnels analysed, the possible initiating events that must be considered include:
• Collision;
• Derailment;
• Fire.

The set of hazards considered significant for an isolated railway line was obtained through the
analysis of the data available and involving the experienced staff and other experts judgement.
Starting from the lists of causes shown in Chapter 3, as shown in Table 4.5, for each hazardous
events (Derailment D, Collision C, Fire F) or hazard was evaluated the applicability to the case
study, evaluating the relevance according to the characteristics of the portion of the system
analysed (i.e. Tunnels) and the accidental history of the infrastructure.
For this reason, it is necessary, for example, to mark as 'Not Applicable' (NA) the hazards
not compatible with the system characteristics and as ‘Not significant’ No the event where the
actual number of accidents recorded in the accidental databases or similar systems are equal to

140
CHAPTER 4 Risk Management for FCE tunnels

zero. The hazards marked by “No” or “N ” will not be ta en into consideration because not
significant for the case study.
This phase allows to associate each hazard to the respective safety barrier which
contribute to lowering the probability that they evolve into Initiating Events of an accidental
sequence.

Table 4.5. Hazard identification and characterization

Hazardous
Hazards Significant
Events (HE)
Human Error
H1: Failure/incorrect compliance with regulation D-C Only D
H2: Failure/incorrect compliance with operational/technical D-C Yes
prescriptions
H3: SPAD D-C Yes
H4: Wrong itinerary preparation D-C Only D
H5: Maximum permissible speed exceeding D-C Yes
Technical Causes
Infrastructure
T1: Irregularities of the infrastructure (rail/switches/portals) D Yes
T2: Track geometry and roadbed irregularities D Yes
Rolling Stock
T3: Mechanical or electrical defects/wear of rolling stock D-F Yes
T4: Defects of electrical system F Yes
T5: Positive hot-box detectors (HBDs) F Yes
T6: Overheating of rolling stock components F Yes
T7: Loss of rolling stock components D-F Yes
T8: Breakage of the coupling devices of the rolling stock D-F NS
Causes external to the railway system
E1: Landslides/boulders/trees on the railway site D NA
E2: Abnormality due to external event/flooding D-F Yes
E3: Irregularities concerning work sites D-F Yes
E4: Obstacles interfering with the loading gauge D NA
Other
U: Causes undetermined or unspecified D–C–F Yes

141
CHAPTER 4 Risk Management for FCE tunnels

4.3.3 Frequency analysis

4.3.3.1 Analysis of the causes

During the frequency analysis phase, occurrence frequencies are assigned to each basic event
based on the number of accidents recorded on the national network. The values shown in Table
4.6 were obtained by dividing the number of accidents caused by each Hazard by the total number
of trains, the number of years and the kilometres of line of system in which the accident occurred.

Table 4.6. Hazard occurrence rates adapted at the tunnels analysed

Rate
HE Hazards
[ev./tr-km-yr]
H1. Failure/incorrect compliance with regulation 5,36E-10
H2. Failure/incorrect compliance with oper./tech. prescriptions 2,30E-10
H3. SPAD 7,65E-11
H4. Wrong itinerary preparation 7,65E-11
H5. Maximum permissible speed exceeded 7,65E-11
T1. Irregularities of the infrastructure (rail/switches/portals) 6,12E-10
T2. Track geometry irregularities 1,22E-09
- T2.1 Roadbed 2.77E-10
Derailment

- T2.2 Wide Gauge 4.44E-10

- T2.3 Twist 2.22E-10
- T2.4 Alignment 1.66E-10
- T2.5 Longitudinal level 5.55E-11
- T2.6 Other 5.55E-11
T3. Mechanical or electrical defect/wear of rolling stock 6,89E-10
T4. Positive hot-box detectors (HBDs) 7,65E-11
T5. Loss of rolling stock components 1,53E-10
E2. Abnormality due to external event/flooding 6,12E-10
E3. Irregularities concerning work sites 2,30E-10
U (Other) Unknown causes 7,65E-11
H2. Failure/incorrect compliance with oper./tech. prescriptions 5,26E-10
Collis

H3. SPAD 4,20E-09

H5. Wrong itinerary preparation 2,63E-10
T3. Mechanical or electrical defect/wear of rolling stock 1,94E-09
T4. Defects of electrical system 1,45E-09
Fire

T6. Overheating of rolling stock components 3,63E-10

E2. Abnormality due to external event/flooding 2,42E-10
U (Other) Unknown causes 2,24E-10

142
CHAPTER 4 Risk Management for FCE tunnels

By combining the rates through the logical operations imposed by the FTA structure and
multiplying by the traffic and the length of each tunnel, the accidental frequency attributable to
the individual classes of causes and to each Top Event was estimated.
As better shown in Figure 4.10 for the analysed system, the Derailments is due for more
than a third of the causes to the failure of the infrastructure, the Collisions are almost entirely
attributable to the error human while the Fire cases are largely due to rolling stock failures.
Knowledge of the main sources of potential accidents is essential for choosing the correct
mitigating interventions.

Figure 4.10. Incidence of causes in accident frequency

4.3.3.2 Analysis of the consequences

Severity analysis consisted in the identification of the pathways of subsequent hazardous events,
up to the identification of all possible final accident scenarios.
Since the analysis is aimed at a stretch of line entirely in tunnels, the subsequent events
taken into consideration are reduced to:

3. The probability that the accident is 'serious';

4. The probability that there are other trains involved;
5. The probability that a fire starts due to the accident.

Figure 4.11 shows the Event Trees obtained from these considerations and the probability
of each subsequent event are calculated as described in §3.4.1. In the case of a Collision, the
probability of another train being involved in the accident is considered negligible in the line's

143
CHAPTER 4 Risk Management for FCE tunnels

operating scheme and in the case of a Fire, the last of the three events is neglected for obvious
reasons.

Figure 4.11. Event Trees structures for FCE tunnels

The probability associated with each scenario was calculated as a combination of the
probabilities of occurrence (Y) or non-occurrence (N) of subsequent events that constitute the
evolution of the accident as imposed by the Event Tree structure.

4.3.4 Severity of accident scenarios

The lethality calculation models described in section 3.4.2 were used for each of the scenarios
identified for evaluating the severity of accidents. In the following, we describe the definition of
functions for lethality calculation and the process for estimating the expected number of fatalities.

144
CHAPTER 4 Risk Management for FCE tunnels

Once the lethality functions were defined, we applied them to each scenario to estimate
the expected number of fatalities. This involved quantifying the magnitude of each scenario in
terms of factors such as train speed, passenger occupancy, and the nature of the accident.

4.3.4.1 Lethality of Derailment as base event

The number of fatalities for the derailment of a passenger train is obtained from the following
relationship:
𝑆𝑒𝑣𝑒𝑟𝑖𝑡𝑦𝑑𝑒𝑟 = 𝑁𝑝𝑎𝑥𝑃𝑇 ∙ 𝜆𝑑𝑒𝑟𝑃𝑇 (4.7)
where:
• 𝑁𝑝𝑒𝑟𝑠𝑇𝑃 is the number of passengers in the passenger train;
• 𝜆𝑑𝑒𝑟𝑇𝑃 is the lethality of the derailment of a passenger train.

In particular, 𝑁𝑝𝑒𝑟𝑠𝑇𝑃 is estimated to be equal to 75% of the maximum capacity of the

trains.
The lethality can be obtained as a function of the speed of the train 𝑆𝑃𝑇 (expressed in
km/h) of the train at the time of the derailment as follows (Ernst-Basler, 2004):
𝜆𝑑𝑒𝑟𝑇𝑃 = 𝑎 + 𝑏 ∙ 𝑆𝑃𝑇 (4.8)
with a and b constants of the lethality function equal to -0,01 and 0.0002 respectively.
For speeds below 100 km/h, as in the case under examination, the lethality function is
assumed to be constant, equal to 𝜆𝑑𝑒𝑟𝑃𝑇,𝑆<100 = 0,01.

4.3.4.2 Lethality of Collision as base event

As for the previous case, the lethality function is defined as follows:

𝑆𝑒𝑣𝑒𝑟𝑖𝑡𝑦𝑐𝑜𝑙 = 2 ∙ 𝑁𝑝𝑎𝑥𝑃𝑇 ∙ 𝜆𝑐𝑜𝑙𝑃𝑇 (4.9)

where:
• 𝑁𝑝𝑒𝑟𝑠𝑇𝑃 is the number of passengers in the passenger train;

• 𝜆𝑐𝑜𝑙𝑇𝑃 is the lethality of the collision of a passenger trains.

Lethality is obtainable as a function of speed of the train, 𝑆𝑇𝑃 (in km/h), at the time of the
collision as follows (Ernst-Basler, 2004):

𝜆𝑐𝑜𝑙𝑇𝑃 = 𝑎 + 𝑏 ∙ 𝑆𝑇𝑃 (4.10)

145
CHAPTER 4 Risk Management for FCE tunnels

with a and b constants of the lethality function equal to -0,0075 and 0.000225 respectively.
For speeds below 100 km/h the lethality function is assumed to be constant, equal
to 𝜆𝑐𝑜𝑙𝑃𝑇,𝑆<100 = 0,015.

4.3.4.3 Lethality of fire as base event

The lethality of the basic Fire event was evaluated through the relationships set out in §3.4.2.3.
Specifically, for the assessment of the CO concentration and for the exodus the following were
considered:

• 𝑤𝑓𝑖𝑟𝑒 equal to 10 MW;

• 𝑥𝑠𝑚𝑜𝑘𝑒 equal to 3330 mg/MWs4 (RFI, 2006);
• 𝑣𝑠𝑚𝑜𝑘𝑒 in Tunnel 1 and Tunnel 3 equal to 1,5 m/s and 3 m/s in Tunnel 2 thanks to the
mechanical ventilation system;
• 𝑡𝑖 equal to 180 s;
• 𝑑𝑒 is the maximum evacuation distance equal to half the distance between the exits
(portals, stations and stops), equal to approximately 424 m for Tunnel 1, 370 m for Tunnel
2 and 606 m for Tunnel 3;
• 𝑣𝑒 is the evacuation speed which, considering the characteristics of the three tunnels, is
assumed to be equal to 0.6 m/s.

Injured passengers are assumed to be exposed to the effects of the fire for 1200 s,
approximately equal to the intervention times of the rescuers.

4.3.4.4 Severity results

Once the lethality has been defined, it is possible to apply the models for assessing the severity of
the compound scenarios exposed to §3.4.2.
The results obtained are presented in Table 4.7. However, it should be noted that these
results do not yet take into account the specificities of local railways and the presence of safety
measures. Therefore, it is crucial to calibrate both the frequencies and the consequences of each
accident scenario to adapt them to the case study.

146
CHAPTER 4 Risk Management for FCE tunnels

Table 4.7. Severity of the accident scenarios obtained for the three tunnels through the application of lethality
models.
HE Scenario Tunnel 1 Tunnel 2 Tunnel 3
D01 480 282 504
D02 11 11 11
Derailment D03 238 135 251
D04 3 3 3
D05 0 0 0
C01 477 272 503
Collision C02 8 8 8
C03 0 0 0
F01 178 98 188
Fire F02 118 65 125
F03 0 0 0

4.3.5 Calibration of frequencies and consequences

Since both the models and the consulted databases refer to frequencies and consequences
deriving from contexts with characteristics different from those of the Circumetnea Railway, it is
necessary to take into consideration the effects of existing safety measures in reducing the
consequences of accident scenarios.
In this sense, the UIC (UIC, 2002) provides the possible mitigation measures for railway
tunnels and each of these is characterized by the affected events in terms of mitigation impact
subdivided into three classes, namely low, medium, high, associated respectively with risk
reductions of less than 5%, between 5% and 25% and greater than 25%. n particular, these
measures refer to:

• Prevention of accidents;
• Mitigation of impacts;
• Facilitation of escape;
• Facilitation of rescue.

In particular, the former have an effect on the accidental frequencies, the seconds on the
consequences especially related to fire and the last two classes facilitate the escape of the
passengers and the rescue of the injured.
Each measure exerts its reducing effect on one or more causes and on one or more
accident scenarios, making it possible to estimate more likely levels of risk for the system in

147
CHAPTER 4 Risk Management for FCE tunnels

question. In particular, the process of calibration of frequencies and consequences followed the
following steps:

1. Assignment of the mitigating measures to the basic events and to the accident scenarios;
2. Evaluation of the mitigating impact of each measure from the indications provided by
the UIC Codex and based on the judgements of experts and professionals;
3. Application of reduction effects and risk calculation.

The safety measures of the three tunnels and which were taken into account in the
subsequent analyses, divided according to category, are shown in Table 4.8.

Table 4.8. Effect and impacts of system safety measures for frequencies and consequences calibration
UIC ID
(UIC, SAFETY MEASURES EFFECT IMPACTS
2002)
Prevention measures
Train control equipment (blocked brake, hot Frequency of Rolling Stock-
I-4 Medium
boxes) related basic events
Frequency of
Small/
I-6 Track inspection Infrastructure-related
Medium
basic events
Reduction of effects
I-22 Fire protection requirements for structures Variable
I-23 Fire, smoke and gas detection in tunnels Small
Fire extinguishing systems (sprinkler or similar Small/
I-24
installations) Medium
Severity of hot accident
I-25 Smoke extraction systems/ventilation system High
scenarios
Onboard fire extinguishing equipment (traction
R-12 Medium
units and/or coaches)
R-13 Central control of air conditioning Small
R-15 First aid equipment on board Small
Facilitation of escape
I-40 Escape routes (routes, handrails, marking) Medium
I-41 Emergency tunnel lighting Medium
I-42 Emergency telephones/communication means Small
I-44 Vertical exits/access Medium
I-45 Lateral exits/access Medium Severity related to the
Escape equipment and design of coaches (incl. escape of people in all
R-20 Small
access for rescue services) scenarios
Emergency information for passengers
O-20 Small
(preparation for emergencies)
Small/
O-21 Competence of train crew
Medium

148
CHAPTER 4 Risk Management for FCE tunnels

UIC ID
(UIC, SAFETY MEASURES EFFECT IMPACTS
2002)
Facilitation of rescue
Small/
I-61 Access to tunnel entrance and tunnel exits
Medium
I-64 Water supply (at access, in tunnel) Small
I-65 Electrical supply for rescue services Small
I-66 Radio installation for rescue services Medium
Reliability of electrical installations (fire resistance, Severity related to injured
I-67 Small
autonomy) people in all scenarios
I-68 Control system Small
O-30 Emergency and rescue plans Medium
Exercises with rescue services (railway/rescue
O-31 Medium
services communication and co-ordination)
O-33 Provision of rescue equipment Small

For each scenario a reduction in frequencies and consequences resulting from the
combination of the effects of the applicable safety measures was evaluated. The assessed impact
refers to the full efficiency of the single measure, whether material or immaterial, which therefore
must be maintained during all operational phases. The results obtained are shown in the Table 4.9.

Table 4.9. Calibrated frequency and consequences

Tunnel 1 Tunnel 2 Tunnel 3
HE Scenario
F N F N F N
D01 1,29E-12 114 2,26E-12 66 3,30E-12 128
D02 1,29E-08 7 2,26E-08 7 3,30E-08 7
Derailment D03 6,00E-10 53 5,99E-10 28 6,57E-10 60
D04 6,00E-06 2 5,99E-06 2 6,57E-06 2
D05 4,50E-05 0 4,50E-05 0 4,94E-05 0
C01 7,25E-10 106 7,25E-10 56 7,25E-10 121
Collision C02 7,25E-06 5 7,25E-06 5 7,25E-06 5
C03 5,07E-05 0 5,07E-05 0 5,07E-05 0
F01 8,45E-07 39 5,54E-07 20 7,01E-07 45
Fire F02 3,38E-06 26 3,67E-06 13 3,53E-06 30
F03 4,49E-05 0 4,49E-05 0 4,49E-05 0

Under the agreement between Ferrovia Circumetnea and the Department of Architecture
and Civil Engineering of the University of Catania, in order to validate the reliability of the results

149
CHAPTER 4 Risk Management for FCE tunnels

obtained, the consequences of the scenarios involving fire were compared with the results
obtained through the use of a numerical calculation code for the assessment of fire risk with an
engineering/performance approach (Fire Safety Engineering – FSE) for the quantitative assessment
of the development of the fire on the train, the exodus and the consequences. For the analysis,
field models of the FDS (Fire Dynamics Simulator) type were used for modelling the fire and three-
dimensional simulation models of the exodus which take into account the intervention times of
the phases (perception of danger and travel) of the process of evacuation from an environment,
time wasted to queue and individual and reciprocal behaviour simulated instant by instant. Figure
4.12 shows an example of modelling of the tunnel and the train (a) and the result of a generic
instant of the simulation of the exodus during the fire (b).

(a)

150
CHAPTER 4 Risk Management for FCE tunnels

(b)
Figure 4.12. Example of tunnel modelling (a) and fire simulation (b)

The comparison of the results obtained for the basic scenario characterized by the fire of a
passenger train (F02) allows to evaluate the reliability of what has been obtained with empirical
formulas and can be transferred to more complex scenarios. Table 4.10 summarizes for the three
tunnels analysed the lethality obtained respectively through the simplified empirical formula and
through fire and exodus simulation.

Table 4.10. Comparison between lethality obtained through empirical formulas and FDS simulation
Lethality TUNNEL 1 TUNNEL 2 TUNNEL 3
Empirical formula 0.09 0.05 0.11
FDS Simulation 0.002 > 0.001 0.014

What has been obtained allows us to make some considerations. First of all, both the
empirical formula and the simulations have returned the same order in terms of severity in the
three tunnels, identifying the conditions capable of determining greater lethality in the event of a
fire. The lethality obtained through the empirical formula is always greater than that obtained
through more in-depth simulations. This condition reduces the likelihood of underestimating the
risk level of the system and thus operating in favour of safety.

151
CHAPTER 4 Risk Management for FCE tunnels

4.3.6 Risk calculation

By combining the frequency and severity values of all the scenarios analysed, the main risk
indicators for each element of the line were estimated.
With reference to Tunnel 1, the values of total risk R and individual risk IR for the three
accident categories are calculated. The cumulative risk curve is obtained as the sum of the
contributions, in terms of cumulative frequency, of each value of FWI associated with each
scenario, obtaining a distribution of frequency of occurrence as a function of each value of the
consequence. Results are shown in Table 4.11 and the partial (a) and the global (b) cumulative risk
level are shown in Figure 4.13.
A similar methodology was used for the other two tunnels of the system and is not
reported as it is unable to add relevant information for the purposes of this application.

Table 4.11. Total expected risk and individual risk for accident category
Tunnel 1
Accident R IR
[FWI/km-yr] [FWI/km-pax-yr]
Derailment 2.52 ∙ 10−5 1.69 ∙ 10−11
Collision 7.54 ∙ 10−5 5.06 ∙ 10−11
Fire 2.51 ∙ 10−4 1.68 ∙ 10−10
Total 3.52 ∙ 10−4 2.36 ∙ 10−10

152
CHAPTER 4 Risk Management for FCE tunnels

(a)

(b)
Figure 4.13. Cumulative risk for accident category and global curve for Tunnel 1

4.3.7 Risk evaluation

For each tunnel, the Risk Evaluation consisted on the comparison of the risk indicators obtained
with acceptability criteria commonly used for railways.
Acceptability criteria of Italian legislation (Italian Ministry of Infrastructures and Transport,
2005) was ta en as a reference. n particular R is assumed ‘tolerable’ if less than 10−9 FWI per
passenger per ilometre per year and ‘acceptable’ if the indicator is below 10−11 FWI per

153
CHAPTER 4 Risk Management for FCE tunnels

passenger per kilometre per year. IR for Tunnel 1, 2 and 3 resulted respectively 2,36 ∙ 10−10 , 1,49 ∙
10−10 and 2,61 ∙ 10−10 FWI/pax-km-yr, falling within the tolerability thresholds.
Figure 4.14 highlights the thresholds for Cumulative Risk and the levels of the three
tunnels. The three CR curves fall in the right and central sections within the tolerability zone, also
defined as 'attention zone'.
The three tunnels present a similar risk level. Tunnel 2 is slightly different thanks to the
presence of smo e extraction systems that lower the severity levels of ‘hot’ accident scenarios.
From the results of both indicators all three tunnels require the evaluation of the
implementation of additional safety measures considering their cost-benefits performance.

Figure 4.14. Cumulated risk levels for Tunnel 1, 2 and 3

4.3.8 Risk treatment and evaluation of mitigating actions

The risk levels obtained imply the identification, evaluation and introduction of safety measures
such as to bring the indicators within the limits of tolerability.
Among all the interventions analysed, the following possible additional safety measures
were identified:

1. Train control equipment (blocked brakes, hot box detectors);

2. On-board derailment indicators;
3. On-board fire detection systems;
4. Training and updating of Emergency and Rescue Plans;
5. Smoke management (detection, ventilation, extraction) and fire extinguishing systems;

154
CHAPTER 4 Risk Management for FCE tunnels

6. Signalling and speed monitoring.

Each measure contributes to the prevention (measures 1, 4, 5) or the protection (measures

2, 3, 6) of accidents by reducing the frequency of one or more of the basic events of the FTAs or
the consequences and frequencies of one or more of the ETA scenarios. With reference to the
events affected by each measure, the UIC (UIC, 2002) evaluates a risk reduction medium for
measure 1, medium-high for measure 2, low for measure 3, medium-low for measure 4, high for
measure 5 and 6. Then, the effective reduction value was assessed according to the characteristics
of the tunnels and the type of intervention to be implemented.
Introducing the effect of these measures as reducing factors of frequency or severity
respectively, the results are shown in Figure 4.15, where the application of all measures shifts the
CR curve towards the area of acceptability. In particular, the reducing effect of each measure was
estimated by experts panel evaluations based on the list of measures and their impact provided by
the Union of Railways (UIC).

(a) Tunnel 1

155
CHAPTER 4 Risk Management for FCE tunnels

(b) Tunnel 2

(c) Tunnel 3
Figure 4.15. Effect of the application of all safety measures in Tunnel 1 (a), Tunnel 2 (b) and Tunnel 3 (c)

Single measure is not introduced individually but is linked to the implementation of specific
interventions by the Safety & Infrastructure Manager. In particular, the measures mentioned
above can be grouped into four main 'Intervention Groups' (IGs), characterized by a certain
investment:

• IG1: purchase of new trains, equipped with train control equipment (measure 1),
derailment indicators (measure 2) and fire detection systems (measure 3);
• IG2: staff updating activities (measure 4);
• IG3: interventions relating to smoke management systems (measure 5);
• IG4: interventions relating to the modernization of signalling systems (measure 6).

156
CHAPTER 4 Risk Management for FCE tunnels

4.3.9 Risk based resources optimization

The implementation of all IGs implies a significant economic investment that is not always
sustainable by the small operators of local rail networks. In this sense, the risk reduction
performance of each group of measures and the respective implementation cost must be taken
into consideration, so as to evaluate the priority of the interventions based on their cost-benefit
performance. For this reason, the methodology was applied again to the tunnel system
considering the implementation of each IGs at a time to evaluate their role in the overall
mitigating effect, to define an order of priority and the optimal investment for a system
characterized by economic constraints.
The ideal condition, characterized by the absence of economic constraints and the
implementation of all available measures, generates a maximum reduction in the total risk (Rmin) in
the system shown in Figure 4.15.

Table 4.12. Total Risk level reduction

Ractual Rmin Reduction
[FWI/km-yr] [FWI/km-yr] %
Tunnel 1 1,69E-04 3,13E-05 81%
Tunnel 2 9,83E-05 3,08E-05 69%
Tunnel 3 1,77E-04 3,77E-05 79%

Each of the four IGs contributes differently to the total risk reduction. In particular, through
the quantification of this contribution as a percentage weight of the effect of the measure on the
total reduction (4), the results shown in Table 4.13 were obtained. The implementation cost of
each IG has been estimated and provided by the Infrastructure Manager.

𝑅𝑎𝑐𝑡𝑢𝑎𝑙 − 𝑅′ (4.11)
𝐼𝐺 𝑐𝑜𝑛𝑡𝑟𝑖𝑏𝑢𝑡𝑖𝑜𝑛 [%] = ∙ 100
𝑅𝑎𝑐𝑡𝑢𝑎𝑙 − 𝑅𝑚𝑖𝑛

where:
• 𝑅𝑎𝑐𝑡𝑢𝑎𝑙 is the actual tunnel total risk level;
• 𝑅′ is the tunnel risk level resulting from the application of the IG considered;
• 𝑅𝑚𝑖𝑛 is the minimum tunnel risk level resulting from the application of all IGs.

157
CHAPTER 4 Risk Management for FCE tunnels

Table 4.13. Contributions and costs of IGs

Tunnel 1 Tunnel 2 Tunnel 3 os [ n €]
IG1 30% 28% 32% 5,44*
IG2 24% 20% 26% 0,10
IG3 58% 20% 56% 1,55
IG4 24% 49% 24% 3,50
* The cost is considered proportionate to the length of the tunnels compared to the
total length of the line.

The sum of the four interventions percentages always exceeds 100%, i.e. the combination
of different safety measures cannot be assessed by the sum of the individual contributions of each
one but must be estimated by updating the system information in the risk assessment
methodology. In other words, a safety measure applied in an unsafe system have different effects
than the same measure applied in a safer system.
Therefore, the evaluation of the order of introduction of the measures must consider the
iterative evolution of the system.
Since the analysed system operates in real conditions with cost constraints, following paragraphs
propose a methodology for assess the priority and the convenience of possible risk mitigation
interventions.

4.3.9.1 Intervention priority

The prioritization of the interventions was evaluated for each additional IG through a intervention
performance indicator as the ratio between the variation in the total specific risk of the three
tunnels (∆𝑅𝑡𝑜𝑡 ), expressed in FWI/year, and the related cost increment, namely:
∆𝑅𝑡𝑜𝑡
∆= (4.12)
∆𝐶𝑜𝑠𝑡
were 𝑅𝑡𝑜𝑡 is equal to the sum of the Total Risk R of the three tunnels multiplied by their length
expressed in kilometers.
As first iteration, actual system configuration is considered and the ∆ is calculated for each
IG implemented individually. Results are shown in Table 4.14. IG2 is the first intervention to be
implemented because of the highest BCR.

158
CHAPTER 4 Risk Management for FCE tunnels

Table 4.14. IG priority, fist iteration

∆𝑹𝒕𝒐𝒕 ∆𝑪𝒐𝒔𝒕 ∆
Priority
[FWI/yr] [M€] [10-9]
IG1 2,56E-04 5,44 0,05 4°
IG2 2,02E-04 0,10 2,02 1°
IG3 4,03E-04 1,55 0,26 2°
IG4 2,55E-04 3,50 0,07 3°

The second iteration of the risk assessment considers as reference the actual system plus
IG2 implemented. Similarly, at each subsequent iteration, the system is uploaded with the IG
identified in the previous step.
Since the IGs considered are four, a total of three iterations are needed to establish the
order of intervention. The results of Iterations 2 and 3 are summarized in Table 4.15.

Table 4.15. IG priority. Second and third iteration

∆ [10-9] Priority
Iteration 2
IG2+IG1 0,04 3°
IG2+IG3 0,19 1°
IG2+IG4 0,07 2°
Iteration 3
IG2+IG3+IG1 0,018 2°
IG2+IG3+IG4 0,073 1°

Therefore, the most suitable order of intervention in terms of benefits and costs is IG2,
IG3, IG4 and IG1.

159
CHAPTER 4 Risk Management for FCE level crossings

4.4 RISK MANAGEMENT FOR FCE LEVEL CROSSINGS

The Risk Management framework has been applied to the level crossings of the FCE line. The aim
is to demonstrate the use of the framework for managing several elements of the line with
different characteristics and varying levels of risk. Additionally, the application of the framework in
the presence of ad hoc databases and data collected on the line will be illustrated. Through this
application, the benefits of utilizing a risk management approach in railway infrastructure
management will be highlighted. The effective management of risk is essential for ensuring the
safety and reliability of railway operations, and the utilization of a comprehensive framework can
provide a structured approach to achieving this goal.

4.4.1 System definition

The description of the "Level Crossing" system cannot overlook the collection of information on
both the railway and road systems that are in communication with it. The following presents the
key features of the system under analysis and the sources of the data utilised.

4.4.1.1 The level crossings

The system under analysis is composed of 96 level crossings distributed along the 110 km of FCE.
These crossings exhibit a wide range of characteristics, which are largely dependent on the
protection systems that are installed.
The level crossings are equipped with various protection systems, ranging from fully
automatic LCs with barriers to LCs without any active protection. Notably, there are seven
different classes of level crossings, as outlined in Table 4.16.

160
CHAPTER 4 Risk Management for FCE level crossings

Table 4.16. LC safety protection classes

Class Description N
CLASS 1 Automatic LC 59
CLASS 2 LC with manual activated barriers, acoustic signals and warning lights 2
CLASS 3 LC with manual activated barriers without acoustic signals and warning lights 3
CLASS 4 Private LC protected by chains or gates 1
CLASS 5 LC with no barriers, with acoustic signals and warning lights 14
CLASS 6 LC with no barriers and no acoustic signals or warning lights 15
CLASS 7 Private LC with no barriers and no acoustic signals or warning lights 2
TOT 96

In Class 1 LCs, the passage of the train triggers a train detection pedal placed before the LC
which closes of the barriers switches on of the warning lights and acoustic signals. A second pedal
placed after the LC has the task of detecting the leaving train, opening the barriers and turning off
warning lights and acoustic signals.
In Class 2 LCs, barriers, acoustic signals and warning lights are activated and deactivated
manually by an operator. The same for Class 3 LCs but with no acoustic signals or warning lights.
Class 4 and Class 7 LCs are placed in private roads, used a few times a day by the road
owners. Finally, Class 5 and Class 6 LCs are open level crossing with and without acoustic or light
warnings respectively.
As shown in Figure 4.2, the average density of LC across the line is very high (0.87 LC/km
against the 0.35 LC/km Italian average (ANSFISA, 2021)), with high concentrations in urban areas
where traffic is higher and visibility is often impeded by constructions and obstacles along the line.

161
CHAPTER 4 Risk Management for FCE level crossings

Figure 4.16. LC distribution and classes along the studied line

4.4.2 Hazard identification and analysis of the causes

The analysis of causes, beginning with the study of the system being analysed, aims to identify
potential hazardous events that may lead to accidents and estimate their frequency of occurrence.
In this study, the analysis of causes is structured as follows:

1. Identification of the basic events that may lead to accidents based on the analysis of the
accident history of the studied systems and similar systems.
2. Examination of the causes that may lead to the initiator event using the Fault Tree
Analysis methodology.
3. Analysis of the frequency of the root events and the initiator event based on estimates
and information gathered from the fault register of the analysed system.

162
CHAPTER 4 Risk Management for FCE level crossings

4.4.2.1 Identification and Classification of Basic Events

The level of risk of a level crossing or any system in general is determined by the occurrence of
situations, known as initiator events, resulting from the combination of multiple causes, which can
escalate into hazardous scenarios.
The identification of the hazards intends to highlight all the events capable of generating
harmful consequences for the users of the system. The condition at level crossings that produces a
potentially hazardous situation has been identified in the Hazardous Event (HE) 'Hazardous
crossing of the LC'. This event defines all the cases in which a railway vehicle is approaching the LC
and the latter, due to the failure of one or more systems, is unable to prevent the passage of road
vehicles or fails to notify the train driver of the hazard.
The occurrence of this condition is the result of the failure of both road and railway safety
systems and could result in a vehicle being on the tracks at the same time as a train is passing.

4.4.2.2 Fault tree for Hazardous crossing at LC

The information collected and provided by FCE, briefly commented on the §4.2.4.1, made it
possible to classify the system failures into three main categories:

• Railway failure: that is a breakdown or failure of the rail-side safety systems that allow the
train to cross the unprotected LC;
• Failure of the LC: i.e. a failure or malfunction of the road side LC protection systems that
allow vehicles to cross the LC with an approaching train;
• Hazardous behaviour of the vehicle: that is all those cases in which the vehicle invades the
track due to incorrect behaviour of the road vehicle driver.

In particular, the Fault Tree structure in shown in Figure 4.17 and the description of each
Basic Event in Table 4.17.

163
CHAPTER 4 Risk Management for FCE level crossings

Figure 4.17. Fault tree of the Ha ardous Event ‘Ha ardous crossing of the C’

Table 4.17. asic Events of the Ha ardous Event ‘Ha ardous crossing of the C’
RE1 Human error of the train driver
RE2 Train detection failure
RE3 Failure of the logic and control systems of the LC
RE4 Barriers failure
RE5 Acoustic signals failure
RE6 Warning lights failure
RE7 Vandalism
RE8 Road vehicle breakdown
RE9 Rush or distraction of vehicle drivers

164
CHAPTER 4 Risk Management for FCE level crossings

4.4.2.3 Frequency analysis

Once the root events and their associated logical links have been described using the FTA
approach, it is possible to estimate the frequency of occurrence of the identified initiating event,
beginning with an assessment of the probability of occurrence of the causes..
In particular, starting from the information reported in the Breakdown Report and from the
study of the phenomena underlying the root events identified, the probability of occurrence of
each of these referred to the single passage of the train at the PL was evaluated. The annual
number of train passages multiplied by each of these probabilities allows one to calculate the
frequency in terms of events/year of the fault tree.
The absence of one or more protection systems (barriers, warnings, etc.) is introduced in
the frequency calculation by assigning a 100% failure probability where not present.

HUMAN ERROR OF TRAIN DRIVER (RE1)

The probability that a train driver crosses an unprotected level crossing has been estimated using
models for the quantitative assessment of human reliability available in the literature. In
particular, the TESEO model (Bello & Colombari, 1980) estimates this probability starting from the
product of five factors related to the specific work activity and conditions:
• K1 = factor linked to the type of activity to be carried out;
• K2 = factor linked to the time available to carry out the activity;
• K3 = factor linked to the human operator’s characteristics;
• K4 = factor linked to the operator’s emotional state;
• K5 = factor linked to the environmental ergonomics characteristics.

Considered the characteristics of the case study, values chosen for the analyses are shown
in Figure 4.19.

Table 4.18. Factor of TESEO model estimation

K1 Routine activity 0,001
K2 Minimum time to act 1
K3 Trained operator 0,5
K4 Situation of potential emergency 2
K5 Goof environmental ergonomics characteristics 0.7
𝑷𝑹𝑬𝟏 = 𝑲𝟏 ∙ 𝑲𝟐 ∙ 𝑲𝟑 ∙ 𝑲𝟒 ∙ 𝑲𝟓 0,0007

165
CHAPTER 4 Risk Management for FCE level crossings

EQUIPMENTS FAILURE (RE2, RE3, RE4, RE5, RE6, RE7)

The probability of a failure of LC safety equipment (basic events RE2, RE3, RE4, RE5, RE6 and RE7)
were evaluated based on the information reported in the failure logs provided by FCE. In
particular, the frequency of occurrence is evaluated with Eq. (4.13):
𝑁𝑒𝑣 ∙ 𝑛𝑑
𝑓= (4.13)
𝑛𝐿𝐶
where:
• 𝑁𝑒𝑣 is the number of failures events recorded;
• 𝑛𝑑 is the mean time to repair (MTTR) expressed in number of days;
• 𝑛𝐿𝐶 is the number of LCs that can be affected by that type of failure;

Table 4.19. Frequency of occurrence of equipment failures per LC

Failures Frequency
𝒆𝒗𝒆𝒏𝒕𝒔 Number of LC MTTR 𝒆𝒗𝒆𝒏𝒕𝒔
ID Basic Event
[ ] affected [𝒅] [ ]
𝒚𝒆𝒂𝒓 𝑳𝑪 ∙ 𝒚𝒆𝒂𝒓
RE2 Train detection failure 22 59 2.5 0.93
Failure of the logic and control
RE3 20 59 2 0.68
systems of the LC
RE4 Barriers failure 26 78 5 1.29
RE5 Acoustic signals failure 42 65 2 0.12
RE6 Warning lights failure 6 75 1.5 0.12
RE7 Vandalism 6 75 1.5 1.65

The frequency thus calculated takes into account the number of times in a year the failure
exposes the LC to an hazardous condition. On the other hand, the probability that the hazardous
condition occurs on a single passage of the train at the LC is necessary to divide this frequency by
the number of trains expected in a year as in Eq. (4.22).
𝑓
𝑝= (4.14)
𝑛𝑡,𝑦
where 𝑛𝑡,𝑦 is the number of train passages in a year specific of each of the 96 LCs.

ROAD VEHICLE BRAKEDOWN (RE8)

The probability that a vehicle brakes down on the tracks in correspondence with the passage of
the train was evaluated as the probability that a vehicle suffers a 'disabling' breakdown exactly in
the time taken to cross the LC.

166
CHAPTER 4 Risk Management for FCE level crossings

Based on this consideration, the probability (𝑝brakedown ) was estimated by the

combination of the probability P1 hat the vehicle suffers a disabling breakage, i.e. such as to
prevent it from moving, and the probability P2 that the vehicle is on the track.
The probability P1, was evaluated through the information collected in literature. In
particular, in (Chand et al., 2020) the probability of a vehicle breakdown is estimated in average
equal to 0,0011 breakdowns a year.
Probability P2, instead, is expressed by Eq. (4.15) assuming that, on average, a vehicle takes
about 5 seconds to cross the LC, assuming a homogeneous distribution of traffic during the 12
daily hours of line activity.
5
P2 = ∙ ADT12h (4.15)
86400
Finally, to homogenize this probability with those of the other basic events, was referred to
the single passage of the train. Therefore, the probability of finding a broken-down vehicle on the
tracks when the train passes is given by Eq.(4.16):
P1 ∙ P2
p𝑏𝑟𝑎𝑘𝑒𝑑𝑜𝑤𝑛 = (4.16)
nt,d
were nt,d is the average number of train crossing the LC in a day.

RE1. RUSH/DISTRACTION

The probability of level crossing violation by road vehicle drivers due to improper behaviour is
estimated based on data provided by FCE. However, these data refer only to PLs equipped with
protective systems (barriers), and the corresponding probability can be calculated using Eqs.(4.15)
and (4.16). Furthermore, this probability, for the purpose of this analysis, is representative of the
general tendency of drivers to violate the PL and therefore can be transferred to cases without
protective systems.
Breakdown databases reports an average of 34 events/year occurred due to rush or
distraction by drivers. Since these data refer to 67 level crossings, giving a frequency of 0.51
events per LC per year.
For classes 5, 6 and 7, i.e. barrier-free level crossings, the probability was increased by 30%
to take into account a greater possibility of incorrect behaviour.

167
CHAPTER 4 Risk Management for FCE level crossings

4.4.3 Consequences analysis

4.4.3.1 Event Tree Analysis

The possible accidental sequences initiated by the HE depends on the LC characteristics an class,
the rolling stock, the rail operation and the road traffic. The structure of the event tree developed
at §3.4.1.4 and shown in Figure 4.28 was applied to the case study.

Figure 4.18. Hazardous cross at LC event tree and final accident scenarios

The probabilities of each gate of the Event Tree are evaluated on the basis of the
characteristics of the level crossing, rail and vehicular traffic, general assumptions and human
behaviour in hazardous situations.

VEHICLE ON TRACK

The probability that a vehicle is on the tracks when the train crosses the level crossing is related to
the crossing time of the level crossing by the vehicle, the traffic intensity of the road involved and
how this is distributed during the day.
it is estimated that 70% of ADT of the latter transits during the 12 hours of activity of the
line and that, as a first approximation, the passage of vehicles is constant in this time interval.

168
CHAPTER 4 Risk Management for FCE level crossings

Furthermore, it is assumed that the PL is crossed at low speed, it is estimated that an average
vehicle takes about 5 seconds to cross the tracks.
Based on these hypotheses, the probability of finding a vehicle on the tracks was estimated
as the ratio between the total time in one day in which the track is occupied by a vehicle and the
total time of track activity expressed in seconds, i.e.:
ADT12H ∙ 5
𝑝vehicle on track =
12 ∙ 3600 (4.17)

VEHICLE DOES NOT MOVE

The "Vehicle does not move" event is linked to the possibility that due to a state of anxiety the
driver is unable to move from the track once the hazard has been perceived.
Therefore, the assessment of the probability of human error by the driver was evaluated
using the HEART methodology (Williams, 1986). This methodology estimates the probability of
human error starting from tabled probability values and influencing factors.
In particular, we consider the uncommon event, the lack of time to detect or correct the
error, the possible inexperience of the driver in handling similar situations and the discrepancy
between real and perceived hazard. According to these hypotheses, the methodology returns a
probability of error equal to 0,0004.

TRAIN DOES NOT BRAKE

The probability that the train driver does not perceive the hazard and does not start braking was
estimated as the possibility of human error occurring. In particular, this probability was estimated
with the HEART methodology (Williams, 1986) equal to 0,0001.

TYPE OF VEHICLE

The type of vehicle influences the severity of a possible accident at the LC due to the different
number of passengers exposed. In particular, the probability that the vehicle involved could be a
car, a motorcycle, a heavy vehicle or a bus is estimated in the analysed area following the
distribution (Comune di Catania, 2012):

• 70% cars;

169
CHAPTER 4 Risk Management for FCE level crossings

• 20,7% motorbikes;
• 8,6% heavy vehicles;
• 0,7% buses.

4.4.3.2 Severity analysis

The analysis of the consequences was based on the estimate of lethality, understood as the
probability of recording one or more fatalities, of each identified accident scenario. In particular,
the fatality estimation methodology is based on:

• Occupation of convoys;
• Type of vehicle involved;
• Average number of vehicle occupants;
• Speed at the time of the accident.

The severity of each accident scenario was estimated through the use of lethality factors as
described in §3.4.2.4. In particular, the number of expected FWIs is given by Eq.(4.23).

𝑁 = 𝑃𝐴𝑋𝑡𝑟𝑎𝑖𝑛 ∙ 𝜆𝑡𝑟𝑎𝑖𝑛 + 𝑃𝐴𝑋𝑣𝑒ℎ𝑖𝑐𝑙𝑒 ∙ 𝜆𝑣𝑒ℎ𝑖𝑐𝑙𝑒 [𝐹𝑊𝐼] (4.18)

The function of the lethality factors are shown in Figure 4.19 for the passengers of the
railway vehicle (a) and the road vehicle (b) respectively.

(a) (b)
Figure 4.19. Expected accident lethality trend vs train speed for the passengers of the railway vehicle (a) and the road
vehicle (b)

Since the lethality thus obtained is strongly influenced by the speed of the train, it is useful
to identify three main groups of final accident scenarios:

170
CHAPTER 4 Risk Management for FCE level crossings

• Scenarios S1 and S2 which are safe scenarios, i.e. not capable of producing fatalities,
therefore null consequences are expected for all LCs and for all types of vehicle.
• Scenarios S3, S4, S5 and S6 which include all the cases in which the train driver perceives
the hazard and begins to brake the train, therefore the expected consequences of the
accident are equal to or lower than the maximum possible according to the braking
dynamics.
• The S7, S8, S9 and S10 scenarios which are characterized by an impact speed equal to the
line speed at the LC and therefore characterized by the maximum possible severity.

The evolution of the S3, S4, S5 and S6 scenarios is a function of the Sight Distance (𝐷𝑠 ) and
the Braking Distance (𝑠𝑏 ) and in particular:

1. 𝐷𝑠 ≥ 𝑠𝑏 : train brakes in safe conditions, without hitting the road vehicle;

2. 𝐷𝑠 < 𝑠𝑏 : the train impacts the vehicle at a speed depending on the total possible train
deceleration allowed by the sight distance.

The Braking Distance, in particular, has been accurately estimated through the
experimental tests described in detail in §4.5.2, determining the empirical coefficients to expand
the range of application of the Pedeluq formula. In particular, the stopping distance according to
this model is given by the Eq.:
𝜑 ⋅ 𝑠2
𝐿𝑏𝑑 = (4.19)
1,09375 ⋅ 𝜆 + 0,127 − 0,235 ⋅ 𝑖 ⋅ 𝜑
where:
• s [km/h] is the train speed;
• i is the gradient.
• 𝜑 and 𝜆 are assumed equal to 0.10 and 2.38 respectively for local railways in case of
emergency breaking.

4.4.4 Risk calculation

The results from the previous steps provide the values of the accidental frequencies (F) and
consequences (N) for each of the identified scenarios. Based on the F-N pairs, the key risk
indicators have been calculated, including the total risk (R) of the single LC, the Individual Risk (IR)
and the Cumulated Risk curve.

171
CHAPTER 4 Risk Management for FCE level crossings

Figure 4.20 shows the 96 CR curves of the PLs on the line. It can be observed that there are
no catastrophic events, i.e., those characterized by more than one FWI (CEI EN 50126-1, 2017),
and the maximum severity for all PLs is equal to 1 FWI. Therefore, the PL system is sensitive to
more frequent accidents but with limited consequences. All level crossings are found to be safe,
i.e., below the tolerance limit, and only a few fall within the range of attention between the
acceptability and tolerability thresholds.
Similar results are obtained from the individual risk indicator shown in Figure 4.21.

Figure 4.20. Cumulative risk curves for all 96 analysed LCs

Figure 4.21. Individual Risk level for all 96 analysed LCs

The results of the RM methodology concern different stakeholders and must be easily read
by each of them. For this reason, the methodology has been integrated by a GIS support (Figure

172
CHAPTER 4 Risk Management for FCE level crossings

4.22), automatically assigning the yellow dot to LCs with at least one between IR and RC in the
attention zone and green dot to LCs with totally acceptable risk.

Figure 4.22. Spatial distribution of the level crossings with reference to the identified risk level

4.4.5 Decision making support tool for risk treatment

The analyses carried out up to this point made it possible to characterize each LC with its own level
of risk and to be able to assess its acceptability.
Level crossings falling between the acceptability and tolerability thresholds present a non-
negligible level of risk but considered acceptable only if the amount of effort to reduce it exceeds
the benefits that could be obtained. In this area, in fact, the ALARP (As Low As Reasonability
Possible) principle applies, that is the condition for which the risk is 'Tolerable only if risk reduction
is impracticable or cost is grossly disproportionate to the improvement gained' (CEI EN 50126-1,
2017).
Based on these assumptions, the framework described provides a powerful tool for the
qualitative assessment of risk and the comparison with the related investment of resources.

173
CHAPTER 4 Risk Management for FCE level crossings

In particular, the optimization of resources takes place on two levels: at line level, including
decisions of high extension that impact on different LCs, such as speed reductions or closures of
entire sections of line, or at a punctual level, i.e. intervention targeted on 'black spots', i.e. single
LCs which by their characteristics constitute an isolated element of risk. The implemented tool
allows to identify and evaluate corrections for both types of criticality.
Below we comment on a hypothesis of decision-making process for the evaluation of the
alternatives for the safety of one of the LCs for which further improvements must be evaluated
with the ALARP methodology.
The analysis highlighted a criticality to LC # 34 as shown in Figure 4.23.

Figure 4.23. Actual CR curve of LC#34

In particular, the LC is Class 5, characterised by a train speed of 45 km/h and a sight

distance of 37.5 m. The risk assessment showed a level of attention for both the CR and the IR.
The model allows to evaluate the benefits in terms of risk reduction deriving from the
following alternatives:

1. Speed limitation to the LC;

2. Improvement of visibility conditions.
3. LC safety barriers upgrade;

Fixing the other conditions, the risk analysis makes it possible to evaluate the maximum
speed at 20 km/h capable of ensuring the safety of the LC. The speed reduction as shown in Figure

174
CHAPTER 4 Risk Management for FCE level crossings

4.24, reduced the accidental frequency, lowering the curve and the consequences, shifting it to
the left.

Figure 4.24. CR curve of LC#34 with speed reduction

As shown in Figure 4.25, the sight distance for the safety of the LC is equal to 150 m. The
possibility of identifying the vehicle on the tracks in time has consistently reduced the frequency
of accident scenarios and in particular those characterized by high frequencies and low
consequences.

175
CHAPTER 4 Risk Management for FCE level crossings

Figure 4.25. CR curve of LC#34 with improved Sight Distance

Finally, the results shown in Figure 4.26 show how the provision of manual barriers is able
to reduce the risk of the LC within the thresholds of acceptability.

Figure 4.26. CR curve of LC#34 with protection equipment class upgrade

The evaluation of the best solution can also include a combination of the previous
interventions. In particular, as shown in Figure 4.27, the combined application of all three

176
CHAPTER 4 Risk Management for FCE level crossings

mitigating measures considered generates a lower risk level in the face, however, of greater
intervention costs and worsening of service quality.

Figure 4.27. CR curve of LC#34 with all mitigation measures implemented

The model therefore quantitatively provides the benefit in terms of risk reduction. In a
decision-making process, the comparison of this benefit with the economic costs related to the
upgrade of the LC class, or to the costs of expropriations and works for the increase of the visible
or, finally, of the economic and social costs (decrease of the competitiveness of the rail transport)
of speed reductions, allows IM to identify the best solution for optimizing the available resources.

177
CHAPTER 4 Analysis of the role of characteristics of the local railways

4.5 ANALYSIS OF THE ROLE OF CHARACTERISTICS OF THE LOCAL RAILWAYS

In the context of local railway systems, assessing and managing risk is of critical importance to
ensure the safety of passengers and employees. In order to properly evaluate the level of risk and
make informed decisions, it is necessary to understand the behaviour of two characteristics
different from standard railway systems:

• the infrastructure;
• the rolling stock.

The infrastructure can pose a significant risk to safety due to the risk of derailment
associated with the deterioration of the infrastructure. On the other hand, the rolling stock,
particularly the braking distance, is another important factor that must be evaluated to ensure the
safety of the system. In this way, a thorough understanding of these characteristics is essential in
order to properly assess and manage risk in railway systems.

4.5.1 Track geometric quality degradation on narrow gauge local railways

To define the role of the characteristics of narrow-gauge railways in the degradation behaviour the
monitoring data of a local railway in southern Italy were analysed.
The railway consists of a 110 km single-track line with a narrow gauge of 950 mm. The line
operates in heterogeneous conditions, going from zero to around 1000 m above sea level, through
urban, suburban and rural environments. To overcome the external environment obstacles, the
service is carried out by smaller and lighter trains, capable of dealing with high gradients, up to
0‰ small radius curves up to 80 m and maximum permissible speed of 0 m/h. The line is a
ballasted track with 50 UNI and 36 UNI rails linked by concrete and wooden sleeper.
The line is operated in an unbalanced way, with traffic of approximately 19,000 trains/year
in the first 20 km of line, half of the trains in the subsequent 50 km and around one-sixth in the
remaining portion.
The monitoring of the track is carried out using an automatic vehicle equipped with high-
efficiency laser systems. The vehicle collects high-quality data of the three-dimensional position of
the rails every 25 cm and the relative kilometre and the GPS position. Six geometry parameters are

178
CHAPTER 4 Analysis of the role of characteristics of the local railways

provided for each survey point: Gauge, Twist, Longitudinal Level of each rail and Alignment of each
rail. Two inspection runs have been carried out, producing around 5 ∙ 106 geometry data to be
analysed.
The limited number of inspections does not allow for detailed future predictions. Based on
this limitation, the quantity and quality of data available allows for an analysis of spatial rather
than temporal degradation, investigating the effects of different geometric and operational
characteristics of the track.

4.5.1.1 Data preparation

Measured data hides the effects of uncertainty and complex relations among the elements
responsible for degradation that can be interpreted through statistical and probabilistic
approaches. The obstacle of those approaches is linked to the quality of base data. Raw data are
affected by errors of different nature to be identified and eliminated and require filtering and
processing to define usable track geometry quality indexes (Esveld, 2001). The methodology for
the dataset preparation, are summarized in Figure 4.28.

Figure 4.28. Data preparation methodology

4.5.1.2 Data alignment methodology

Raw data from different monitoring runs are affected by positional errors due to shifted reference
points, different recording instruments set-ups or deformations due to slipping or sliding of the
wheels on the rail (Khosravi et al., 2021).
In the case study, the same monitoring vehicle was used in both inspection runs,
minimizing errors due to instrumentation set up and same error can be assumed for all geometry
parameter recorded within the same run. Therefore, by minimizing the positional error for one of
the parameters a good alignment is expected for the others. Gauge was selected as reference for

179
CHAPTER 4 Analysis of the role of characteristics of the local railways

alignment because is less affected by degradation occurring between inspections and is less
challenging for the alignment methods (Khosravi et al., 2021).
The positional error causes each data series to be misaligned with the actual position on
the line and between datasets. The first issue has an impact on the identification of the defect and
the scheduling of related maintenance activities. Misalignment between datasets, on the other
hand, precludes evaluation of the degradation path of defects. Absolute Position Based and
Relative Position Based methodologies, which are widely described in (Khosravi et al., 2021),
provide solutions to these problems. In order to trace the evolution of defects between datasets, a
RPB method based on the Year 1 dataset position as a reference is sufficient.
The alignment methodology analyse and minimize the positional error through the
alignment a set of corresponding peaks. Not all peaks can be considered for data alignment as
under a certain amplitude a strong variability makes it difficult to find the same peak into different
datasets. The minimum amplitude that produces a consistent number of peaks to be analysed and
a good ability to identify matches was found in 50% of AL.
Namely, as shown in Figure 4.29a, identifying two generic consecutive peaks A and B in
Year ' and ’ and Year ” and ” datasets that exceed the 0% of due to positional
error Eqs. (4.20) and (4.21) are verified.

𝑥′𝐴 ≠ 𝑥 ′′𝐴 𝑎𝑛𝑑 𝑥′𝐵 ≠ 𝑥′′𝐵

(4.20)

̅̅̅̅̅̅̅̅
𝑥′𝐴 𝑥′𝐵 ≠ ̅̅̅̅̅̅̅̅̅̅
𝑥′′𝐴 𝑥′′𝐵 (4.21)

In a plane with the Year 1 positions in the x-axis and the Year 2 positions in the y-axis, the
positional error causes corresponding peaks fall away from the bisector of the plane, on a line
defined by a slope β and intercept α Figure 4.29b . n particular α is constant for all points and
identifies the translation error; β measures the level of compression or stretching of the
segment.

180
CHAPTER 4 Analysis of the role of characteristics of the local railways

(a) (b)
Figure 4.29. Positional error between (a) peaks and corresponding trend (b)

For a higher number of peaks, instead of the AB segment, the alignment error trend is
described by the interpolating line. Finally, the calculated aligned position for Year 2 was obtained
by subtracting the translation factor α from all points and dividing them by the deformation factor
β.
A total of 107 corresponding peaks were found in the two datasets. To increase the
effectiveness of the alignment, the line was divided into 4 same length sections. Error functions
and coefficients of determination, R2, of the four interpolating lines are shown in equations (4.22)
to (4.25).
𝐸𝑟𝑟𝑜𝑟 𝑆𝑒𝑐𝑡𝑖𝑜𝑛 1: 𝑥" = 1.0013𝑥′ + 0.0059 (𝑅 2 = 0.99) (4.22)
2
𝐸𝑟𝑟𝑜𝑟 𝑆𝑒𝑐𝑡𝑖𝑜𝑛 2: 𝑥" = 1.0013𝑥′ − 0.0183 (𝑅 = 0.99) (4.23)
2
𝐸𝑟𝑟𝑜𝑟 𝑆𝑒𝑐𝑡𝑖𝑜𝑛 3: 𝑥" = 1.0015𝑥′ − 0.0458 (𝑅 = 0.99) (4.24)
𝐸𝑟𝑟𝑜𝑟 𝑆𝑒𝑐𝑡𝑖𝑜𝑛 4: 𝑥" = 1.0011𝑥′ − 0.0673 (𝑅 2 = 0.99) (4.25)
The high values of R2 reflects the linear trend of the positional error in the four sections.
Year 2 dataset seems to be stretched (𝛽 > 1) in a similar way on the four sections. Finally, the
starting position of second year dataset is before the zero of the first year in Section 1 (α > 0) and
after in the other ones (α < 0).
The alignment process resulted in positional errors near to one meter in all geometry
parameters and a good alignment of the isolated defects. A sample of Longitudinal level raw data
extracted from the first of the four sections before (a) and after (b) data alignment are shown in
Figure 4.30.

181
CHAPTER 4 Analysis of the role of characteristics of the local railways

(a)

(b)
Figure 4.30. Longitudinal level dataset (a) before and (b) after alignment process

4.5.1.3 Segmentation

To cope with the huge amount of data to be analysed it is useful to divide the line into sections.
Segmentation was carried out according to the track characteristic. The information available on
the line made it possible to define homogeneous sections according to Curvature, Slope, Track
Type and Number of Trains. These characteristics were divided into three classes each as shown in
Table 4.20.

182
CHAPTER 4 Analysis of the role of characteristics of the local railways

Table 4.20. Class of curvature, slope, track and number of trains

CLASS: 1 2 3
≤ . 1.54 - 6.67 ≥ . 7
Curvature [1/km]
R≥ 0m (150 m < R < 650 m) R≤ 0m
Slope [‰] ≤ 0 10 - 30 ≥ 0
36UNI
50UNI 36UNI
Type of track mix concrete and
concrete sleepers concrete sleepers
wood sleepers
Number of trains
3000 9500 19000
[train/year]

For each characteristic considered, progressing from class 1 to 3, conditions are gradually
more aggressive for degradation. In addition, the class 3 sections of curvatures and slopes are
typical of local railways. Three classes for each of the four characteristics result in 3 4 combinations,
of which 48 actually present on the line. Identifying a section every time the combination of
characteristics remains constant, the 110 km of line can divided into 1477 sections of around 50 m
each.
Furthermore, the analysis highlighted several points subject to maintenance not reported.
Sections with abnormal behaviour, such as not reported maintenance actions points, have been
isolated. Also tunnels, bridges, level crossings, stations, or all special assets that would lead to
anomalous degradation behaviour were identified and analysed separately.
The soil is a fundamental characteristic in the degradation process (Falamarzi et al., 2017;
Lyngby, 2009; J. Xie et al., 2020). Nevertheless, soil data available are in low resolution and does
not allow to characterize reliably the single sections but highlights only a large-scale homogeneity
of volcanic soils. For these reasons, the soil was excluded from the analysis.

4.5.1.4 Track geometry quality indexes

Studies have usually focused on one or few aggregated indexes, such as STD of longitudinal level
(Khajehei, Ahmadi, Soleimanmeigouni, & ..., 2019; Sato, 1995; Soleimanmeigouni, Ahmadi, Nissen,
et al., 2020) or horizontal alignment Jovanović et al. 0 , gauge deviation hac a ušić
2017; Falamarzi et al., 2017) or a combination of few of them (Andrade & Teixeira, 2015;
Soleimanmeigouni, Ahmadi, Khajehei, et al., 2020; Wu et al., 2022). In order to provide a full view
of degradation behaviour, this work focuses both on aggregate and isolated defect indexes.

183
CHAPTER 4 Analysis of the role of characteristics of the local railways

European legislation (EN 13848-5, 2017) was taken in consideration for the definition of
the track geometry indexes. As highlighted in Table 4.21, for the aggregate description of the state
of health of the sections, the following indexes have been chosen:
• the mean of the deviations from the nominal value of the Gauge;
• standard deviation (mean of the two rails) on the section length of Longitudinal Level and
Alignment;
• EN 13848-5 provides limitations for Twist only for the amplitude of isolated defects,
confirming its the relevant role in the derailment risk (Bocz et al., 2018), especially when
combined with other defects (Wu et al., 2022).

European legislation (EN 13848-6, 2021) identifies as index for isolated defects their
number in a section or their amplitude. To obtain a homogeneous measurement between all the
indexes analysed, each section was characterized by the average Absolute Amplitude of major
Isolated Defects (AAID) defined as shown in (4.26).

∑𝑖=1,𝑛|𝑥𝑖 |
𝐴𝐴𝐼𝐷𝑘 = (4.26)
𝑛
where k is the section of the line, n is the total number of peaks of amplitude over 50% of AL, 𝑥𝑖 is
the actual peak amplitude.

Table 4.21. Track geometry parameter and track geometry quality indicator
Geometry parameter Track geometry quality index
• Mean
Gauge
• Average amplitude of isolated defects over 50% of AL (AAID)

• Standard deviation
Longitudinal Level
• Average amplitude of isolated defects over 50% of AL (AAID)

• Standard deviation
Alignment
• Average amplitude of isolated defects over 50% of AL (AAID)

Twist • Average amplitude of isolated defects over 50% of AL (AAID)

4.5.1.5 Effect of track characteristics on spatial distribution of track geometry quality degradation

The objective of this analysis was to identify, regardless of the history of the sections, the
combinations of classes of characteristic that are most sensitive to degradation and understand if
the peculiar classes of local railways exhibit significantly different behaviour.

184
CHAPTER 4 Analysis of the role of characteristics of the local railways

To do this, the means of the distributions of each class of characteristic were analysed in
Year 1 and Year 2. The ANOVA test was performed for each track geometry index (2 indexes for
Gauge, Longitudinal level, and Alignment and 1 for Twist) and track characteristic (3 classes for 4
characteristics) for both years, for a total of 56 tests. The null hypothesis of the ANOVA is that all
means of the distributions of the indexes in the three classes of each characteristic are equal. The
significance level was set at α= % so when the p-value obtained from the test is lower of α the
null hypothesis is rejected and at least one of the means of the classes is different from the others.
As presented in Table 4.22 and Table 4.22 aggregate indexes (i.e. means and STDs) showed
a significant difference within the distributions between the three classes of each track
characteristics. Presence of single defects showed no significant difference for Longitudinal Level
with reference to Curvature and Track Type and for Alignment with reference to Slope and Track
Type as well. This can be explained considering that the outer rail on a curve is affected by higher
horizontal forces due to centrifugal acceleration (Belalia et al., 2020; Lyngby, 2009), while slope
influence mostly the forces on the vertical plane. Nevertheless, the total number of isolated
defects of Longitudinal Level and Alignment is very low, making difficult to analyse the complexity
and variability of degradation. More data are needed in future analyses to improve the robustness
of those results.
For each test the mean values, the standard deviation and the total number of points (i.e.
sections) of the index distributions are presented. Finally, Tukey test was than carried out to
understand what classes are significantly different and how. In Table 4.22 and Table 4.23 is shown
the Grouping resulting from the Tukey test. Means of classes assigned to different groups
(identified by the letters A, B and C) are significantly different. Otherwise, when two classes share
a letter, the means are not significantly different. As an example, looking at the classes of
curvature with the distribution of the Mean Gauge index in Year 1 (Table 3), ANOVA p-value shows
that the variation in curvature has a significant role in Mean Gauge degradation and Tukey test
showed a statistically significant difference when comparing class 1 and 2, class 2 and 3 and class 1
and 3. Otherwise, in Gauge AAID vs Number of trains, ANOVA p-value is below the significance
level but Tukey test assigned class three both to and group highlighting that class doesn’t
significantly differ from class 1 (group A) and class 2 (group B).

185
CHAPTER 4 Analysis of the role of characteristics of the local railways

Table 4.22. Means and STD of indexes distributions into the classes of characteristics and results of ANOVA and Tukey
test for Year 1 dataset for each quality index and track characteristic combination
CURVATURE SLOPE TRACK TYPE NUMBER OF TRAINS
YEAR 1
1 2 3 1 2 3 1 2 3 1 2 3
Mean 0.35 11.90 14.48 6.45 7.21 10.98 8.77 4.70 10.56 9.63 6.31 8.95
STD 2.12 8.60 9.19 8.55 9.67 9.59 9.12 8.18 10.6 9.65 8.31 10.80
Mean

N. of Sections 585 440 451 464 565 447 749 388 339 569 615 292
ANOVA p-value 0.000 0.000 0.000 0.000
Gauge

Tukey test Grouping C B A B B A B C A A B A

Mean 2.38 13.68 16.58 10.67 12.42 14.77 13.82 9.75 13.16 14.15 11.31 12.45
STD 2.09 9.08 9.43 9.99 9.86 9.38 9.58 8.67 10.7 10.00 8.75 11
AAID

N. of Sections 184 362 399 280 341 324 451 213 281 383 360 202
ANOVA p-value 0.000 0.000 0.000 0.000
Tukey test Grouping C B A B B A A B A A B AB
Mean 2.42 2.60 2.87 2.74 2.48 2.63 2.37 2.68 3.05 2.90 2.30 2.70
STD 1.31 1.19 1.11 1.50 1.12 1.02 1.19 1.21 1.20 1.12 1.14 1.45
STD

N. of Sections 585 440 451 464 565 447 749 388 339 569 615 292
Longitudinal Level

ANOVA p-value 0.000 0.003 0.000 0.000

Tukey test Grouping C B A A B A C B A A B A
Mean 0.78 0.68 0.59 0.89 0.60 0.55 0.72 0.65 0.65 0.60 0.70 0.85
STD 0.69 0.55 0.59 0.79 0.51 0.45 0.66 0.53 0.64 0.57 0.58 0.73
AAID

N. of Sections 155 137 189 154 163 164 169 136 176 258 126 97
ANOVA p-value 0.023 0.000 0.476 0.003
Tukey test Grouping A AB B A B B A A A B AB A
Mean 1.73 3.14 4.39 3.09 2.50 3.42 2.80 2.46 3.91 3.30 2.48 3.32
STD 1.10 1.90 2.29 2.21 1.69 2.32 2.13 1.74 2.12 2.05 1.94 2.30
STD

N. of Sections 585 440 451 464 565 447 749 388 339 569 615 292
ANOVA p-value 0.000 0.000 0.000 0.000
Alignment

Tukey test Grouping C B A B C A B C A A B A

Mean 0.80 1.05 1.38 1.32 1.04 1.20 1.19 1.08 1.27 1.22 1.04 1.34
STD 0.75 0.92 1.11 1.14 0.85 1.04 1.06 0.97 1.01 1.03 1.02 1.01
AAID

N. of Sections 83 213 338 206 186 242 278 132 224 287 197 150
ANOVA p-value 0.000 0.026 0.245 0.020
Tukey test Grouping B B A A B AB A A A AB B A
Mean 0.29 0.59 0.78 0.43 0.46 0.73 0.34 0.59 0.88 0.34 0.59 0.88
STD 0.73 0.8 0.85 0.71 0.77 0.93 0.66 0.84 0.97 0.66 0.84 0.97
Twist

AAID

N. of Sections 586 440 451 465 565 447 749 388 340 749 388 340
ANOVA p-value 0.000 0.000 0.000 0.000
Tukey test Grouping C B A B B A C B A C B A

186
CHAPTER 4 Analysis of the role of characteristics of the local railways

Table 4.23. Means and STD of indexes distributions into the classes of characteristics and results of ANOVA and Tukey
test for Year 2 dataset for each quality index and track characteristic combination
CURVATURE SLOPE TRACK TYPE NUMBER OF TRAINS
YEAR 2
1 2 3 1 2 3 1 2 3 1 2 3
Mean 1.57 12.31 14.95 7.42 7.90 11.57 9.64 5.23 11.28 10.28 7.02 9.97
STD 1.98 8.32 8.99 8.11 9.15 9.44 9.03 7.34 9.83 9.43 7.81 10.20
Mean

N. of Sections 585 440 451 464 565 447 749 388 339 569 615 292
ANOVA p-value 0.000 0.000 0.000 0.000
Gauge

Tukey test Grouping C B A B B A B C A A B A

Mean 2.16 14.13 17.19 12.46 13.51 15.24 15.14 9.55 15.01 15.37 11.68 15.05
STD 1.73 9.04 9.54 10.11 9.87 9.72 9.75 8.72 10.26 10.17 8.88 10.69
AAID

N. of Sections 119 360 390 240 306 323 434 193 242 359 340 170
ANOVA p-value 0.000 0.003 0.000 0.000
Tukey test Grouping C B A B AB A A B A A B A
Mean 2.34 2.56 2.94 2.63 2.45 2.73 2.41 2.72 2.85 2.95 2.35 2.39
STD 1.17 1.12 1.11 1.35 1.04 1.08 1.15 1.22 1.05 1.06 1.15 1.21
STD

N. of Sections 585 440 451 464 565 447 749 388 339 569 615 292
Longitudinal Level

ANOVA p-value 0.000 0.000 0.000 0.000

Tukey test Grouping C B A A B A B A A A B B
Mean 0.74 0.67 0.62 0.8 0.61 0.62 0.71 0.7 0.6 0.59 0.78 0.74
STD 0.64 0.56 0.56 0.76 0.47 0.51 0.62 0.57 0.56 0.51 0.63 0.72
AAID

N. of Sections 155 139 217 151 167 193 191 154 166 288 145 78
ANOVA p-value 0.135 0.003 0.152 0.003
Tukey test Grouping A A A A B B A A A B A AB
Mean 1.63 2.69 4.08 2.63 2.38 3.15 2.61 2.37 3.26 3.26 2.21 2.62
STD 0.88 1.57 1.93 1.87 1.56 1.89 1.86 1.58 1.76 1.99 1.37 1.85
STD

N. of Sections 585 440 451 464 565 447 749 388 339 569 615 292
ANOVA p-value 0.000 0.000 0.000 0.000
Alignment

Tukey test Grouping C B A B B A B B A A C B

Mean 0.77 0.88 1.26 1.23 1.02 1.06 1.08 1.08 1.14 1.24 0.85 1.1
STD 0.64 0.82 1.02 1.15 0.83 0.88 0.96 0.98 0.95 1.03 0.76 0.93
AAID

N. of Sections 54 158 319 154 154 223 249 106 176 278 154 99
ANOVA p-value 0.000 0.131 0.805 0.000
Tukey test Grouping B B A A A A A A A A B AB
Mean 0.34 0.59 0.88 0.42 0.46 0.76 0.42 0.46 0.76 0.88 0.32 0.34
STD 0.66 0.84 0.97 0.71 0.86 1.01 0.71 0.86 1.01 1.07 0.65 0.63
Twist

AAID

N. of Sections 749 388 340 465 565 447 465 565 447 570 615 292
ANOVA p-value 0.000 0.000 0.000 0.000
Tukey test Grouping C B A B B A B B A A B B

187
CHAPTER 4 Analysis of the role of characteristics of the local railways

(a)

(b)

(c)

188
CHAPTER 4 Analysis of the role of characteristics of the local railways

(d)
Figure 4.31. Box plots and Tukey test Grouping of geometry indexes of (a) Gauge, (b) Longitudinal Level, (c) Alignment
and (d) Twist for the three classes of Curvature, Slope, Track Type and Number of Trains in Year 1 dataset

The distributions of the indexes analysed in each class of characteristic are shown in the
charts below where red letters call back the Tukey test grouping division.
Since the differences between the results obtained in the two years are negligible, for
convenience the distributions relating to Year 1 are commented.
Figure 4.31a highlights the relation between Gauge, both Mean and AAID, and Curvature.
Sections with curve radius below 150 m have the highest probability to show a worse general
quality and to conceal hazardous defects. This is partially linked to the presence of design gauge
widenings and no differences in maintenance limits due to track curvature, that makes more likely
that maintenance limits are reached, requiring a more careful monitoring. Also Class 3 Slope
affects gauge quality because of the increased stress and wear on railheads.
Longitudinal Level defects (Figure 4.31b) are mostly located in the sections with higher
curvature and lower class of the track type, where the magnitude and effect of the stresses due to
traffic are higher. The isolated defects, on the other hand, have a less marked behaviour,
underling the need of further data to define their behaviour.
Figure 4.31c shows the greater probability of getting worse Alignment quality conditions in
Class 3 curves due to the higher centrifugal forces. Also, the Class 3 Track Type generally has worse
quality due to the lower ability to bear the vehicles loads.
Twist (Figure 4.31d) tend to exceed maintenance limits in Class 3 of Curvature, Slope and
Track Type, i.e. where the stresses due to the traffic are higher and the ability of the
superstructure to bear them is less.
Finally, V-shaped behaviour of number of trains effect for all track geometry parameters
can be explained by the combination of two phenomena. The first leads to an increasing
degradation from class 1 to 3 and is linked to the logical increase in degradation where the use of

189
CHAPTER 4 Analysis of the role of characteristics of the local railways

the line is greater. The second goes against the trend and is linked to the greater frequency of
monitoring and maintenance of the most used parts of the line, bringing better conditions where
the number of trains is higher.

4.5.1.6 Effect of track characteristics on track geometry quality degradation rates

Once the role of the characteristics of local railways on the distribution of defects within the
sections was analysed, the role that these characteristics had on the degradation growth rate was
than evaluated.
To do this, for each section the variation of indexes values between the two datasets was
normalized by the amount of MGTs.
As shown in Table 4.24, ANOVA test p-values over the significance level (0,05) were found
for the Longitudinal Level indexes (both STD and AAID) versus Curvature and for the AAID index of
Longitudinal level versus Slope. AAID index showed non-significant difference for Alignment versus
Curvature and versus Slope classes and in AAID of Twist versus Curvature.
Tukey test was performed on the triplets of distribution of each combination of
characteristics of the line and geometry parameters and the results are shown in Table 4.24.

190
CHAPTER 4 Analysis of the role of characteristics of the local railways

Table 4.24. Means and STD of indexes distributions into the classes of characteristics and results of ANOVA and Tukey
test for Degradation Rates.
NUMBER OF
CURVATURE SLOPE TRACK TYPE
RATE TRAINS
1 2 3 1 2 3 1 2 3 1 2 3
Mean 3.97 3.84 4.54 3.89 3.95 4.50 3.49 4.5 5.34 6.40 2.94 1.78
STD 2.30 2.42 2.46 2.55 2.18 2.45 2.12 2.39 2.54 1.86 1.18 0.59
Mean

N 524 360 362 394 471 381 716 263 267 499 500 247
ANOVA p-value 0.000 0.000 0.000 0.000
Gauge

Tukey test Grouping B B A B B A C B A A B C

Mean 5.28 3.98 4.87 4.47 4.10 5.05 3.74 5.05 5.97 6.96 3.08 1.92
STD 3.92 2.58 2.65 3.21 2.51 2.83 2.39 3.31 2.91 2.30 1.71 0.83
AAID

N 99 294 322 202 243 270 408 107 200 316 257 142
ANOVA p-value 0.000 0.001 0.000 0.000
Tukey test Grouping A B A AB B A C B A A B C
Mean 1.04 0.84 0.83 0.71 0.91 1.14 0.74 1.29 1.13 1.29 0.65 0.32
STD 1.42 1.08 0.93 0.96 1.20 1.39 1.10 1.65 0.87 1.21 1.23 0.46
STD

N 260 139 153 144 233 175 328 117 107 271 223 58
Longitudinal Level

ANOVA p-value 0.151 0.008 0.000 0.000

Tukey test Grouping A A A B AB A B A A A B B
Mean 4.07 3.36 2.37 4.46 1.94 3.60 2.14 5.62 3.38 4.21 2.15 0.77
STD 5.24 3.87 2.15 5.51 1.98 3.34 2.98 4.32 4.58 4.21 3.66 0.68
AAID

N 21 12 22 19 22 14 28 12 15 28 12 15
ANOVA p-value 0.373 0.115 0.034 0.034
Tukey test Grouping A A A A A A B A AB A B B
Mean 0.69 1.18 1.59 0.79 0.98 1.39 0.90 0.79 1.87 1.74 0.42 0.36
STD 1.25 1.41 2.01 1.26 1.56 1.78 1.43 1.08 2.13 1.94 0.65 0.60
STD

N 260 139 153 144 233 175 328 117 107 271 223 58
ANOVA p-value 0.000 0.002 0.000 0.000
Alignment

Tukey test Grouping B A A B B A B B A A B B

Mean 6.08 3.84 6.17 5.33 5.81 5.55 3.92 6.49 7.33 6.56 3.62 2
STD 7.06 3.2 4.45 5.64 4.57 4.02 3.48 6.22 4.26 4.64 3.64 1.88
AAID

N 11 29 79 25 42 52 56 19 44 86 22 11
ANOVA p-value 0.057 0.914 0.000 0.000
Tukey test Grouping A A A AB B A B AB A A B B
Mean 1.38 1.2 1.06 1.15 1 1.45 0.99 1.24 1.59 1.58 0.49 0.57
STD 1.41 1.22 0.96 1.34 0.96 1.29 1.12 1.26 1.18 1.27 0.59 0.45
Twist

AAID

N 124 112 172 92 168 148 220 84 104 262 120 26

ANOVA p-value 0.078 0.003 0.000 0.000
Tukey test Grouping A A A AB B A B AB A A B B

191
CHAPTER 4 Analysis of the role of characteristics of the local railways

(a)

(b)

(c)

192
CHAPTER 4 Analysis of the role of characteristics of the local railways

(d)
Figure 4.32. Box plots and Tukey test Grouping of degradation rates of geometry indexes of (a) Gauge, (b) Longitudinal
Level, (c) Alignment and (d) Twist for the three classes of Curvature, Slope, Track Type and Number of Trains

Gauge (Figure 4.32a), both for the aggregate quality and the isolated defects, showed
significantly higher degradation rates in Class 3 of Curvature and Slope, typical of isolated railways.
This can be linked to the higher wear of the railheads in high slopes and in low radius curves due
to the difficult geometric inscription of the bogies. Also Track type presents a significant greater
degradation where the superstructure provides less resistance to traffic loads.
Longitudinal Level (Figure 4.32b) showed a significantly higher degradation rate of the
general quality of the section in Class 3 Slopes. The Type of track classes influenced both the STD
and AAID degradation rate and, in particular, the stiffer superstructure preserved the best quality
compared to the other two classes.
The quality of Alignment (Figure 4.32c) showed a significantly higher degradation rate in
the curved sections, due to the higher centrifugal forces. Class 3 Slope influenced both the STD
and AAID values. Finally, sections with wooden sleepers suffered a significantly greater
degradation.
The Twist degradation rate (Figure 4.32d) was higher on high Slopes (Class 3) and increases
going from heavier to lighter superstructure.
Some considerations common to all four track geometry quality parameters can be
discussed. The first concerns the number of trains. In all cases, the trends found in the spatial
analysis is further amplified, finding higher degradation rates where the train traffic is lower. This
can be attributed above all to the worse starting quality of the low traffic sections, demonstrating
the tendency of the degradation rate to increase when the initial quality is worse. Secondly, as
mentioned above, this result is a symptom of the higher attention of Infrastructure Manager for
high operated sections.

193
CHAPTER 4 Analysis of the role of characteristics of the local railways

The second concerns aggregate indexes. Despite the filter for the sections that were
maintained, slightly negative degradation values were detected. This is a demonstration of the
complex behaviour of small amplitude defects and the difficulty of aligning and analysing their
growth. This behaviour is absent in AAID indexes where such defects are excluded.
Finally, despite the low number of surveys available, the considerable amount of data
made it possible to identify the first effects due to the characteristics of the isolated railways.
Future studies will make it possible to validate the results and introduce solid forecasting models.

4.5.1.7 Discussion

Track geometric quality degradation is a crucial factor that influences the risk of derailment and
the level of safety of a railway line. Understanding the rate of degradation of geometric
parameters can allow for a more accurate assessment of the variation of risk level along the line,
based on the characteristics of the track, particularly in local railways where some of these
parameters reach extreme values.
The effect of curvature, slope, type of track, and loads on the line on the geometry quality
degradation behaviour was investigated. These characteristics contain some of the fundamental
differences between narrow-gauge and standard-gauges railway, like the presence of high
curvatures and slopes, heterogeneity in the track type and variability train traffic. These
characteristics were grouped into classes, isolating the peculiar characteristics of narrow-gauge
local railways.
The size and quality of available data allowed to obtain information on the spatial
distribution of degradation along the line and on the degradation rate occurred between different
surveys.
Section characterized by peculiar narrow-gauge characteristics resulted more likely to
show a worse quality condition both in terms of space distribution of defects and degradation
rates, requiring a greater attention to prevent maintenance thresholds exceeding. Therefore, in
these sections, the application of degradation models calibrated on standard railways could lead
to an underestimation of the degradation level, resulting in the occurrence of isolated defects and
higher need of corrective actions.
Future implementations of the analysis are related to upgrades of the input data. The
information provided by further surveys could consolidate the results obtained on the spatial

194
CHAPTER 4 Analysis of the role of characteristics of the local railways

distribution and provide a solid and reliable degradation rate analysis. More in-depth information
on the soil type, environmental conditions and maintenance history allows to improve the
interpretation of the variability of degradation.
Additionally, an increased number of monitoring runs and years of collected observations
can enable the establishment of a methodology in the future to correct the risk level of each
analysed line section based on its characteristics, and to more accurately evaluate decisions
related to the track maintenance strategy within the RM process.

4.5.2 Experimental measures for the stopping distance

Another factor that distinguishes local railways from interconnected railways is related to rolling
stock. In local railways the vehicles are generally smaller, less heavy and with different
characteristics compared to other railway networks. The rolling stock, particularly the braking
distance, is an important factor that must be evaluated to ensure the safety of the system.
The evaluation of stopping distances presents a level of complexity due to the multiple
parameters that come into play and influence the calculation. In order to identify a calculation tool
that is more suitable for the case of local railways, a series of experimental measurements carried
out with the trains of FCE.

4.5.2.1 Instrumentation

The tests were conducted on an ADE 19 diesel electric railcar equipped with two driving cabs
located at the heads with 31,9 tons of weight, an aerodynamic drag surface of 4 𝑚2 and 4 braked
wheels with a radius of 0,375 m and moment of inertia of 1 𝑘𝑔 ⋅ 𝑚2. Figure 4.33 shows the railway
vehicle used for the tests and the driver's cab with all the acquisition systems used.

195
CHAPTER 4 Analysis of the role of characteristics of the local railways

(a) (b)
Figure 4.33. The railway vehicle used for the tests a and the driver’s cab with all the acquisition system installed.

The instrumentation that was installed on the vehicle was composed by:

1. A high gain GPS (Sauchy Data System xPro) antenna with accelerometric and gyroscopic
platform in the vehicle cabin, digital communication via CAN bus at 25 Hz;
2. Analog pressure sensors in the braking system on:
(a) The main pipe;
(b) Front landing gear;
(c) Rear carriage;
3. An analogical sensor (wired potentiometer) to detect the position of the brake control
lever actuated by the driver.

The GPS antenna has fixed the longitudinal axis of the vehicle in the direction of travel as
the X axis, the traversal axis to the left as the Y axis and the vertical axis to the top as the Z axis.
Figure 4.34 shows the positioning of the GPS antenna inside the train control cabin and the
positioning of the analogical sensor to detect the position of the brake control lever.

196
CHAPTER 4 Analysis of the role of characteristics of the local railways

(a) (b)
Figure 4.34. GPS antenna inside the train control cabin (a) and the positioning of the analogical sensor to detect the
position of the brake control lever (b).

As regards the data acquisition phase, the following tools were used:

1. 1 DAQ Dewesoft DEWE-43A (8 channels, 24-bit sigma-delta with anti-aliasing filter, max
simultaneous sampling 200 kS / s);
2. 1 DAQ Sirius ACC + (8 channels, 24-bit sigma-delta with anti-aliasing filter, max
simultaneous sampling 200 kS / s);
3. Dewesoft-X software.

Figure 4.35 shows the Dewesoft DEWE-43A system and the Sirius ACC + system, both with CAN
input.

(a) (b)
Figure 4.35. Dewesoft DEWE-43A system (a) and the Sirius ACC + system (b).

197
CHAPTER 4 Analysis of the role of characteristics of the local railways

4.1.1.1 Tests and results

A total of 25 braking tests were carried out varying speed, wheel-rail contact (dry or sand)
characteristics and type of braking (emergency or service). The tests carried out are shown in
Table 4.25.

Table 4.25. Braking tests characteristics

Wheel-rail N° of
Test Initial speed
contact repetitions
Service braking Dry 5 50 km/h
Service braking Dry 3 60 km/h
Emergency braking with sand Dry with sand 4 50 km/h
Emergency braking without sand Dry 4 50 km/h
Service braking with electric brake Dry 3 50 km/h
Emergency braking with electric brake Dry with sand 3 50 km/h
Coasting Dry 3 50 km/h

In order to describe the methodology followed the results of emergency braking on sand
will be discussed below. The procedure has been the same for the other tests as well.
To perform the emergency breaking, the train was accelerated to the nominal speed of 50
km/h and maintained constant for 2 seconds, then the driver applied the brake lever. The
following parameters were monitored:

• Lever position: LEVER [mm];

• Pressure sensors: P_MOD, P_ANT, P_POST;
• Speed: GPS_Speed,
• Deceleration: Intern_AccelX_HighRes [𝑚/𝑠 2 ];
• GPS Position: GPS Longitude, GPS Latitude;
• Braking distance

The stopping distance and delay times were calculated from the instant 𝑡 = 0 in which the
start of the brake control lever is detected. Figure 4.36 shows the measured parameters during
one of the emergency braking with sand tests at 50 km/h.

198
CHAPTER 4 Analysis of the role of characteristics of the local railways

(a)

199
CHAPTER 4 Analysis of the role of characteristics of the local railways

(b)
Figure 4.36. Parameters during one of the emergency braking tests at 50 km/h with the opening of the sandboxes
during all the test (a) and at the beginning (b)

Figure 4.36a shows how the train, starting from the initial speed (blue dotted line) of 53,8
km/h, stops after a distance (SD, yellow solid line) of 110,7 m with a deceleration value (red solid

200
CHAPTER 4 Analysis of the role of characteristics of the local railways

line) at the maximum braking system pressure (green solid line) is at a maximum of -1,29 m/s^2.
Instead, Figure 4.36b shows a focus on the first part of the test. In particular, were obtained a
deceleration delay (DTAx0) equal to 0,802 s, a duration of brake lever application (DTB) equal to
0,448 s and a maximum delay of the front pressure (DTPAmax) of the brakes equal to 1,531 s.
Table 4.26 shows the summary of the values obtained from the emergency braking tests at
50 km/h with the opening of the sandboxes, of which the average and standard deviation were
calculated.

Table 4.26. Summary of the values obtained from the emergency braking tests at 50 km / h with the opening of the
sandboxes.
Measure Test 0 Test 1 Test 2 Test 3 Avg STD
Initial speed 𝑘𝑚
V0[ ℎ ] 53,8 53,3 52,4 53,9 53,4 0,7
Stopping
SD [m] 110,7 101,3 97,4 99,6 102,3 5,9
distance
Duration of
brake lever DTB [s] 0,448 0,29 0,344 0,405 0,372 0,069
operation
Deceleration
DTAx0 [s] 0,802 0,679 0,752 0,742 0,744 0,051
start delay
Max front
DTPAmax
pressure 1,531 1,419 1,37 1,471 1.448 0.069
[s]
delay
Max rear
DTPPmax
pressure 1,452 1,336 1,37 1,471 1,407 0,065
[s]
delay
Deceleration
DTAxpmax
delay at 1,802 1,808 1,642 1,742 1,749 0,077
[s]
Pmax
Deceleration Axpmax
value at 𝑚 -1,29 -1,37 -1,36 -1,33 -1,338 0,04
[𝑠2 ]
Pmax
Deceleration
𝑚
value at 10 Ax10 [𝑠2 ] -1,78 -1,68 -1,83 -1,78 -1,768 0,06
km / h
Initial Lat0 37.5370690 37.5370706 37.5371241 37.5370678
position GPS Long0 14.9771655 14.9771440 14.9768351 14.9771855
Final Lat1 37.5372810 37.5372643 37.5373117 37.5372571
position GPS Long1 14.9759495 14.9760313 14.9757665 14.9760903

The results obtained from the tests have allowed for the calibration of the constants of the
Pedeluq and Maiden formulas to the specific case under examination. In particular, the calibration
has provided the results shown in Figure 4.37 which demonstrate that the final stopping distance

201
CHAPTER 4 Analysis of the role of characteristics of the local railways

of test one coincides with the distance obtained from the two calibrated formulas (represented by
the circle and star at the bottom right of the figure).

Figure 4.37 Curves resulting from the application of the parameters in the three models for calculating the braking
distance with reference to the fitting operation with test 1.

From the analysis and fitting of the data it was possible to identify the parameters of the
different models. In particular, for the Pedeluq’s empirical model (Chapter 3, Eq. 3.23), the
following parameters were obtained:
• 𝜑=0,10;

• 𝜆=2,28 for service braking;

• 𝜆=2,38 for emergency braking.

The following for the Minden formula (Chapter 3, Eq. 3.25):

• 𝜓=0,69;

• 𝜆=228 for service braking;

• 𝜆=238 for emergency braking.

202
CHAPTER 4 Analysis of the role of characteristics of the local railways

4.5.2.2 Discussion

La Stopping distance is a fundamental parameter for risk management and for proper support of
the decision-making process. Underestimating the braking performance of rolling stock leads to a
lower risk assessment than what actually exists, which can result in tolerating higher risk levels
without mitigating interventions and exposing the system to vulnerabilities that are not correctly
identified. On the other hand, overestimating the braking distances can result in safety evaluations
with large safety margins, but at the decision-making level, it can lead to oversized or unnecessary
safety measures with negative impacts on the available resources, which are already limited in the
case of local railways.
The work presented in the previous paragraphs aimed to validate two of the most
commonly used formulations for stopping distance in railway design and safety evaluations
through experimental tests on a case study, and to calibrate them for proper application to local
railways. As shown in Figure 4.38, in the case of service braking on a level track, both for the
Pedeluq’s formula (a) and the Minden formula (b), the experimentally obtained formula yields
stopping distances below 70 km/h, which are almost half of those obtained from the extrapolation
of the formula results as proposed in the literature.
0 0
Pedeluc Minden
00 00
formula formula
0
ocal railway 0 ocal railway
parameters parameters
00 00
istance [m]

istance [m]

0 0

00 00

0 0

00 00

0 0

0 0
0 0 0 0 0 0 0 70 80 0 0 0 0 0 0 0 70 80
peed [ m/h] peed [ m/h]

(a) (b)
Figure 4.38. Comparison between Literature formula and local railway empirical adaptation for Pedeluq (a) and
Minden (b) formulas

203
5. CONCLUSIONS
CHAPTER 5 Introduction

5.1 INTRODUCTION
The local railways provide a fundamental role for the mobility, history and tourism of a
territory. Driven by the demand for increasingly higher safety standards, by the continuous
evolution of the regulatory framework and motivated by the lack of specific studies, the
purpose of this work was to support the development of safety and competitiveness of local
railways through the definition of a risk-based decision support tool that takes into account
the peculiarities of this type of railway networks.
This chapter aims to answer the research questions set out in Chapter 1 and to
summarize the significant findings in response to the research questions at the basis of this
work and reports the discussion and considerations regarding the results obtained. The main
contributions, limitations and possible future developments of which this research is
characterized are commented.

205
CHAPTER 5 A risk management framework for local railway decisions support

5.2 A RISK MANAGEMENT FRAMEWORK FOR LOCAL RAILWAY DECISIONS SUPPORT

1. How to optimize investments in local railways with the aim of aligning them with the
safety standards and management strategies of interconnected railways?

The first RQ asked for a tool capable of guiding the process of improving safety in local
railways, able to evaluate and optimize investments for the alignment with the standards of
the interconnected networks.
To do this, first of all, the reference regulatory framework was thoroughly analysed.
The whole path that the legislation on railway safety has followed starting from the beginning
of the liberalization process, which took place in the last decade of the 1900s, up to the
present day, was analysed. Thanks to this examination it was possible to understand the
safety standards adopted today, the process that led to their adoption and the main tools for
their satisfaction and verification. Secondly, the regulatory history of local railways was
examined in order to understand the reasons behind the infrastructural and organizational
structure that they have developed to date in parallel with the interconnected railways.
One of the tools available to infrastructure managers for assessing and controlling
safety levels is Risk Management. A critical literature review was conducted on the
frameworks and applications of RM, deepening the methodologies used to carry out each of
its steps and the results of the main applications. The study highlighted the lack of
applications in the management of local railways and the absence of methodologies that
allow the quantitative application of the RM process in environments with little information
on the accident history.
The first step of the framework presented is aimed at defining the system, identifying
the main characteristics of these networks. Than, in the hazard identification phase, the most
relevant Hazardous Events (HE) were identified. In particular Derailment, Collision, Fire and
Hazardous crossing at the LC were highlighted. For each of these, in the phase of analysis of
the causes, through investigation of literature and of accidental databases of similar systems,
the causes and their interrelationships were identified through Fault Tree Analysis (FTA). In
the phase of analysis of the consequences, all subsequent events were studied through Event
Tree Analysis and all the possible final accident scenarios were identified starting from each

206
CHAPTER 5 A risk management framework for local railway decisions support

HE. The identification of the accident scenarios was carried out through the analysis of the
information reported in the accident databases integrated by expert judgment. The
probabilities of each event was assessed using statistical analyses as well as considerations
based on infrastructure, operation, and existing safety systems. The consequences associated
with each accident scenario were evaluated using lethality models capable of taking into
account the speed of the train, the passengers occupancy, the development of the accident
chain and the characteristics of the fire using empirical formulas. Empirical formulas are able
to take into account a lower complexity than fire simulations but allow to obtain results in a
more immediate and simple way without compromising the reliability of the analysis.
Consequences and probabilities were then used to estimate the total, individual and
cumulative risk for each HE identified and to evaluate its acceptability.

2. How to develop a quantitative risk management framework in the presence of

limited accident history and limited resources?

The first research question allowed to build a RM framework capable of supporting

the decision-making process but, also, highlighted the problems that arise in the case of local
railways and synthetized by the second research question.
The reliability of accident frequencies and consequences calculation rely on the
analysis of accident data which are still scarce in the case of local railways. To overcome this
problem, the necessary information for the analyses were built starting from databases of
reference systems. A methodology for the calibration of accidental data collected in other
railway systems through the use of expert's judgments and quantitative analysis was
presented and commented. The methodology allowed to indirectly assess frequencies and
accidental consequences by comparing the effectiveness of safety barriers of a case study
and a reference system.
With the aim of guiding the choices of local railways managers to optimize
investments in the presence of limited budgets, the RM process has been integrated with a
Benefit-Cost Analysis (BCA). BCA allows rail operators and policymakers to prioritize
investment decisions, allocate resources efficiently, and optimize the performance and
sustainability of the rail network. From the point of view of a short-medium term time

207
CHAPTER 5 A risk management framework for local railway decisions support

horizon, the advantage of using BCA for evaluating risk reducing measures in a railway system
is that it provides an immediate, systematic and objective method for decision-making. By
weighing the potential benefits against the costs, decision-makers can make informed choices
about the best use of resources. However, BCA is less effective for long term decisions, where
intangible difficult to monetize costs and benefits (environmental impact, social impacts, etc.)
take on more importance.

208
CHAPTER 5 Effect of local railways design characteristics on risk

5.3 EFFECT OF LOCAL RAILWAYS DESIGN CHARACTERISTICS ON RISK

3. What are the effect of the design characteristics of local railways on the probability
and severity of railway accidents?

With the aim of responding to the third RQ, the main design characteristics of local
railways and their effect on risk were identified and analysed. In particular, there are two
elements that distinguish local railways from the point of view of design: the infrastructure
and the rolling stock.
The probability of derailment, and therefore the risk level of the system, is strongly
influenced by the quality of the track geometry and by the probability of occurrence of
defects. The literature analysis has highlighted how the behaviour of the track in standard
railways is a current and widely analysed topic. For local railways, however, no study has yet
been undertaken. In these railways the presence of unique values of the key factors of the
track geometry quality degradation (e.g. curvature, slopes) makes the onset of defects
unpredictable. With the aim of bridging this gap, data harvested by two high-precision
monitoring campaigns on a local railway line provided the basis for the track geometry
degradation evaluation. The effect of curvature, slope, type of track, and loads on the line on
the geometry quality degradation behaviour was investigated. These characteristics contain
some of the fundamental differences between narrow-gauge and standard-gauges railway,
like the presence of high curvatures and slopes, heterogeneity in the track type and variability
train traffic. These characteristics were grouped into classes, isolating the peculiar
characteristics of narrow-gauge local railways. First step was the data preparation for the
following statistical analysis. A data alignment methodology was implemented based on the
correspondence of isolated defects of gauge. The line was, then, divided into homogeneous
sections based on the characteristics of the line. Gauge, Longitudinal Level, Alignment and
Twist were considered to obtain an overall view of the track geometry quality conditions.
Both the aggregate description of the degradation and indicators related to isolated defects
were defined. The size and quality of available data allowed to obtain information on the
spatial distribution of degradation along the line and on the degradation rate occurred
between different surveys. ANOVA test was performed for each combination of track

209
CHAPTER 5 Effect of local railways design characteristics on risk

geometry index and track characteristic for both datasets and for the degradation rates.
Tukey test was carried out to understand what classes are significantly different in
degradation behaviour and how. Section characterized by peculiar narrow-gauge
characteristics resulted more likely to show a worse quality condition both in terms of space
distribution of defects and degradation rates, requiring a greater attention to prevent
maintenance thresholds exceeding. Therefore, in these sections, the application of
degradation models calibrated on standard railways could lead to an underestimation of the
degradation level, resulting in the occurrence of isolated defects and higher need of
corrective actions.
In the absence of ad hoc legislation and guidelines, this study allows to support good
practices for maintenance management of narrow-gauge local railways. Stricter maintenance
limits could be adopted in sections with more aggressive characteristics classes and specific
defect priority indexes based on the type of section and not related only to the severity of the
defect could be implemented. Furthermore, targeted monitoring should be reserved for
'sensitive' sections. Given the high costs of automatic monitoring vehicles and the limited
resources of local railways, the increase in the annual frequency of monitoring run may not
be achievable. In this sense, the use of manual systems for the sections subject to more
significant degradation could be effective, cheaper in economic and organizational resources
but equally accurate. The results obtained have highlighted the need to take into account the
different behaviour of degradation in local railways, but the limited data available does not
yet allow for the quantitative integration of these considerations within the risk management
framework.
Another factor that distinguishes local railways from interconnected railways is
related to rolling stock. In local railways the vehicles are generally smaller, less heavy and
with different characteristics compared to other railway networks. The rolling stock,
particularly the braking distance, is a crucial factor that must be evaluated to ensure the
safety of the system.
The evaluation of stopping distances presents a level of complexity due to the
multiple parameters that come into play and influence the calculation. In order to identify a
calculation tool that is more suitable for the case of local railways, a series of experimental
measurements were carried out with the trains of a narrow-gauge local railway. A total of 25

210
CHAPTER 5 Effect of local railways design characteristics on risk

braking tests were carried out on a diesel railcar varying speed, wheel-rail contact (dry or
sand) characteristics and type of braking (emergency or service). The stopping distance and
delay times were measured for each test and the results obtained allowed for the calibration
of the Pedeluq and Maiden formulas. New values for the constants of the two formulas were
proposed to adapt to local railways. The formulas obtained allowed to calculate the
probabilities and consequences of all those accident scenarios influenced by the stopping
distance.

211
CHAPTER 5 Case study validation

5.4 CASE STUDY VALIDATION

In order to provide a practical example of how the findings of this work can be transferred to
real-world scenarios, the Risk Management framework was applied to the Ferrovia
Circumetnea (FCE), a narrow-gauge railway that connects several small towns on the slopes
of Mount Etna, Sicily. FCE history, physical characteristics, and operational aspects were
presented to better understand the challenges faced by the infrastructure manager. The
accidental data of the Italian national interconnected railway managed by RFI taken as a
reference were also described.
The framework proposed in this work was applied to two specific cases in the Ferrovia
Circumetnea: the risk management for Tunnels and for Level Crossings.
The first case study focused on the decision support for the safety improvement of
three long tunnels. Data from similar railway systems were calibrated. Starting from the
frequency of the causes up to the assessment of the consequences for all possible accident
scenarios, the main risk indicators were evaluated. In order to improve the safety level, four
possible safety improvement interventions were evaluated. Through the iterative application
of the methodology within a cost-benefit analysis, the mitigating impact of all measures was
assessed and a prioritization of interventions was provided identifying the most convenient
interventions to be implemented. The analysis shown that the continuous updating of staff
produces the best results in terms of both prevention and protection action at highly
contained costs. Even the implementation of efficient smoke and fire management systems,
against a high economic commitment, produces a reduction in the risk levels that justifies the
cost. At the bottom of the priority scale are the improvement of signalling systems and the
improvement of the characteristics of the rolling stock through the purchase of new trains.
Nevertheless, this prioritization is intended to improve only the tunnels safety levels. Going
beyond the tunnels safety and extending the analyses to a line level the convenience of the
considered interventions could be inverted.
The second case study focused on the Level Crossings (LCs) risk management. The
framework was applied to the 96 LCs of FCE. All the necessary data were collected and the
LCs characterized according to their safety protection systems. The data available concerned
the number of failures per safety system, the description and cause of the failure and the

212
CHAPTER 5 Case study validation

time required to repair it, made it possible to quantitatively assess the causes and
consequences of all accidental scenarios for all LCs and to identify all of those in the risk
attention area for which further mitigating barriers must be assessed. Than, the use of the
framework as a tool to support decisions was described, evaluating the effect on the risk of
three different possible interventions to bring back the risk level within acceptable area and
their effect on all the LCs of the line. The presented methodology allowed to identify 'black
spots' of greatest risk, characterizing the risk factors to be mitigated. From the point of view
of the network, the tool allows to evaluate the effect of decisions such as increases in train
speed or in the number or type of trains, having an immediate overview of the effects on the
general risk level of the system. Through an iterative application of the methodology within a
cost-benefit analysis, the mitigating impact of all measures could be assessed and a
prioritization of interventions could be provided identifying the most convenient
interventions to be implemented.

213
CHAPTER 5 Future works

5.5 FUTURE WORKS

The topic analysed is extensive and the work presented only lays the foundations for the
improvement and optimization of the management of local railways. Despite the numerous
topics covered, not all the interesting research ideas have been addressed within this thesis
work. The following are some suggested research areas to improve and advance this work:

• A possible future development of the methodology is linked to the completeness of

the decision-making phase. The analysis presented considers the railway system as
isolated from the outside world, a hypothesis that could lead to inaccurate or inexact
assessments. A multimodal approach with a broader network perspective would
improve the final quality of the assessments. In fact, the Decision-Making Process
must not create situations in which the safety improvement is intended as the
transfer of the risk to other systems (e.g. the limitations to rail traffic generate a
modal shift on road transport, increasing accidents and pollution). Therefore, the
results provided by the RM process, if integrated with further non-homogeneous and
non-quantifiable decision criteria (such as the impact on user comfort or
environmental impact) within a Multi Criteria Decision Analysis methodology, can
provide a complete and reliable knowledge for the evaluation of the actions on the
system.
• A further development of this work can be given by the collection of data concerning
the local railways. From the point of view of accident databases, further information
would allow to validate the estimated frequencies and consequences and to evaluate
the risk with greater precision, improving the reliability of the decision-making
process.
• From the point of view of the knowledge of the effect of local railways design
characteristics on risk, future implementations of the analysis are related to upgrades
of the input data. The information provided by further surveys could consolidate the
results obtained on the spatial distribution and provide a solid and reliable
degradation rate analysis. More in-depth information on the soil type, environmental
conditions and maintenance history allows to improve the interpretation of the

214
CHAPTER 5 Future works

variability of degradation. Thanks to more precise data, it will be possible to define

appropriate corrective factors to take into account the characteristics of the track in
assessing the level of risk of the section analysed.
• Finally, the role of expert judgment was crucial for the calibration and adaptation of
the missing accidental information. A possible development in this sense may include
the use of more complex methodologies (such as methodologies based on Fuzzy
Reasoning) for a more accurate and systematic quantification of expert judgment.

215
REFERENCE REGULATORY FRAMEWORK
 Reference Regulatory Framework

EUROPEAN REGULATION
Council Directive 91/440/EEC of 29 July 1991 on the development of the Community's railways
Council Directive 95/18/EC of 19 June 1995 on the licensing of railway undertakings
Council Directive 95/19/EC of 19 June 1995 on the allocation of railway infrastructure capacity and the
charging of infrastructure fees
Council Directive 96/48/EC of 23 July 1996 on the interoperability of the trans-European high-speed
rail system
Directive 2001/12/EC of the European Parliament and of the Council of 26 February 2001 amending
Council Directive 91/440/EEC on the development of the Community's railways
Directive 2001/13/EC of the European Parliament and of the Council of 26 February 2001 amending
Council Directive 95/18/EC on the licensing of railway undertakings
Directive 2001/16/EC of the European Parliament and of the Council of 19 March 2001 on the
interoperability of the trans-European conventional rail system
Directive 2001/16/EC of the European Parliament and of the Council of 19 March 2001 on the
interoperability of the trans-European conventional rail system
UIC-Codex 779-9 “ afety in Railway Tunnel” of /09/ 00 .
Directive 2004/49/EC Of The European Parliament And Of The COUNCIL of 29 April 2004 on safety on
the Community's railways and amending Council Directive 95/18/EC on the licensing of railway
undertakings and Directive 2001/14/EC on the allocation of railway infrastructure capacity and
the levying of charges for the use of railway infrastructure and safety certification (Railway
Safety Directive)
Directive 2004/50/EC Of The European Parliament And Of The Council of 29 April 2004 amending
Council Directive 96/48/EC on the interoperability of the trans-European high-speed rail system
and Directive 2001/16/EC of the European Parliament and of the Council on the interoperability
of the trans-European conventional rail system
Directive 2004/51/EC Of The European Parliament And Of The Council of 29 April 2004 amending
Council Directive 91/440/EEC on the development of the Community's railways
Regulation (EC) No 881/2004 Of The European Parliament And Of The Council of 29 April 2004
establishing a European Railway Agency (Agency Regulation)
Directive 2007/58/EC of the European Parliament and of the Council of 23 October 2007 amending
Council Directive 91/440/EEC on the development of the Community’s railways and irective
2001/14/EC on the allocation of railway infrastructure capacity and the levying of charges for the
use of railway infrastructure
Directive 2007/59/EC of the European Parliament and of the Council of 23 October 2007 on the
certification of train drivers operating locomotives and trains on the railway system in the
Community
Directive 2008/57/EC of the European Parliament and of the Council of 17 June 2008 on the
interoperability of the rail system within the Community (Recast) (Text with EEA relevance)
Directive 2008/110/EC of the European Parliament and of the Council of 16 December 2008 amending
irective 00 / 9/EC on safety on the Community’s railways Railway afety irective Text
with EEA relevance)
2009/460/EC: Commission Decision of 5 June 2009 on the adoption of a common safety method for
assessment of achievement of safety targets, as referred to in Article 6 of Directive 2004/49/EC

217
 Reference Regulatory Framework

of the European Parliament and of the Council (notified under document number C(2009) 4246)
(Text with EEA relevance)
Directive 2012/34/EU of the European Parliament and of the Council of 21 November 2012
establishing a single European railway area (recast) Text with EEA relevance
Commission Regulation (EU) No 1158/2010 of 9 December 2010 on a common safety method for
assessing conformity with the requirements for obtaining railway safety certificates Text with
EEA relevance
Commission Regulation (EU) No 1169/2010 of 10 December 2010 on a common safety method for
assessing conformity with the requirements for obtaining a railway safety authorisation Text
with EEA relevance
Commission Regulation (EU) No 1169/2010 of 10 December 2010 on a common safety method for
assessing conformity with the requirements for obtaining a railway safety authorisation Text
with EEA relevance
Commission Regulation (EU) No 1077/2012 of 16 November 2012 on a common safety method for
supervision by national safety authorities after issuing a safety certificate or safety authorisation
Text with EEA relevance
Commission Regulation (EU) No 1078/2012 of 16 November 2012 on a common safety method for
monitoring to be applied by railway undertakings, infrastructure managers after receiving a
safety certificate or safety authorisation and by entities in charge of maintenance Text with EEA
relevance
Commission Implementing Regulation (EU) No 402/2013 of 30 April 2013 on the common safety
method for risk evaluation and assessment and repealing Regulation (EC) No 352/2009 Text with
EEA relevance
Commission Implementing Regulation (EU) 2015/1136 of 13 July 2015 amending Implementing
Regulation (EU) No 402/2013 on the common safety method for risk evaluation and assessment
(Text with EEA relevance)
Commission Regulation (EU) 2015/995 of 8 June 2015 amending Decision 2012/757/EU concerning
the technical specification for interoperability relating to the ‘operation and traffic management’
subsystem of the rail system in the European Union (Text with EEA relevance)
Directive (EU) 2016/2370 of the European Parliament and of the Council of 14 December 2016
amending Directive 2012/34/EU as regards the opening of the market for domestic passenger
transport services by rail and the governance of the railway infrastructure (Text with EEA
relevance)
Regulation (EU) 2016/2338 of the European Parliament and of the Council of 14 December 2016
amending Regulation (EC) No 1370/2007 concerning the opening of the market for domestic
passenger transport services by rail (Text with EEA relevance)
Regulation (EU) 2016/2337 of the European Parliament and of the Council of 14 December 2016
repealing Regulation (EEC) No 1192/69 of the Council on common rules for the normalisation of
the accounts of railway undertakings (Text with EEA relevance)
Regulation (EU) 2016/796 of the European Parliament and of the Council of 11 May 2016 on the
European Union Agency for Railways and repealing Regulation (EC) No 881/2004 (Text with EEA
relevance)
Directive (EU) 2016/797 of the European Parliament and of the Council of 11 May 2016 on the
interoperability of the rail system within the European Union (recast) (Text with EEA relevance)
Directive (EU) 2016/798 of the European Parliament and of the Council of 11 May 2016 on railway
safety (recast) (Text with EEA relevance)

218
 Reference Regulatory Framework

Commission Delegated Regulation (EU) 2018/762 of 8 March 2018 establishing common safety
methods on safety management system requirements pursuant to Directive (EU) 2016/798 of
the European Parliament and of the Council and repealing Commission Regulations (EU) No
1158/2010 and (EU) No 1169/2010 (Text with EEA relevance)

ITALIAN REGULATION
Presidential Decree 11 July 1980, n. 753 New regulations on the subject of police, safety and regularity
in the operation of railways and other transport services.
Law 1 December 1986, n. 870 Extraordinary urgent measures for the services of the General
Directorate of civil motorization and transport under concession from the Ministry of Transport.
Presidential Decree 8 July 1998, no. 277. Regulation containing rules for the implementation of
directive 91/440/EEC relating to the development of the Community railways
Executive Decree 247/VIG3 of 22 March 2000 containing the definition of the standards and safety
regulations applicable to rail transport pursuant to article 5, paragraph 1, of the Presidential
Decree July 8, 1998, n, 277
Legislative Decree 8 July 2003, n. 188 Implementation of directives 2001/12/EC, 2001/13/EC and
2001/14/EC on railways
Legislative Decree 162/2007. Implementation of directives 2004/49/EC and 2004/51/EC relating to
the safety and development of the Community railways.
Legislative Decree 163/2007. Implementation of directive 2004/50/EC amending directives 96/48/EC
and 2001/16/EC relating to the interoperability of the trans-European rail system.
Ministerial Decree of 28 October 2005. Safety in railway tunnels
Legislative Decree 15/2010. Implementation of directive 2007/58/EC, which amends directives
91/440/EEC, relating to the development of the Community railways, and 2001/14/EC relating to
the allocation of railway infrastructure capacity and the imposition of charges for the use of the
railway infrastructure
Legislative Decree 247/2010. Implementation of Directive 2007/59/EC on the certification of train
drivers operating locomotives and trains on the Community railway system
Legislative Decree 191/2010. Implementation of directives 2008/57/EC and 2009/131/EC on the
interoperability of the Community rail system
Legislative Decree 14 May 2019, n. 50. Implementation of Directive 2016/798 of the European
Parliament and of the Council of 11 May 2016 on railway safety
Legislative Decree 14 May 2019, n. 57. Implementation of Directive 2016/797 of the European
Parliament and of the Council of 11 May 2016 on the interoperability of the rail system in the
European Union (recast)
Law Decree n. 148 of 10/16/2017 converted into Law no. 172 of 04/12/2017. Urgent provisions on
financial matters and for non-deferable needs.
Decree of the Minister of Infrastructure and Transport no. 347/2019. Identification of railway
networks functionally isolated from the rest of the railway system
Law Decree 10 September 2021 n. 121. Urgent provisions on investment and safety of infrastructure,
transport and road traffic, for the functionality of the Ministry of sustainable infrastructure and
mobility, the Superior Council of Public Works and the National Agency for Railway Safety and
road and motorway infrastructure.
RFI disposition 13/2001. Requirements for the adoption by the Railway Undertakings and the
Infrastructure Division of a safety management system

219
 Reference Regulatory Framework

RFI disposition 51/2007. Amendments to the Infrastructure Manager's Instruction no. 13 of 26 June
2001 and subsequent amendments
ANSF Decree no. 1/2019. Technical rules and safety standards applicable to networks functionally
isolated from the rest of the railway system as well as to service managers operating on such
networks
ANSF Decree no. 3/2019. Discipline of rules and procedures, pursuant to art. 16, paragraph 2, letter
bb), of the legislative decree 14 May 2019, n. 50, applicable to networks functionally isolated
from the rest of the railway system as well as to subjects operating on such networks

220
REFERENCES
 References

Acquaro, G. (2019). La sicurezza ferroviaria (CIFI, Ed.).

Ahac, M., & Lakusic, S. (2015). Tram track maintenance-planning by gauge degradation modelling.
Transport, 30(4), 430–436. https://doi.org/10.3846/16484142.2015.1116464
hac M. a ušić, S. (2017). Track Gauge Degradation Modelling on Small Urban Rail Networks:
Zagreb Tram System Case Study. Urban Transport Systems, January.
https://doi.org/10.5772/65086
Alemazkoor, N., Ruppert, C. J., & Meidani, H. (2018). Survival analysis at multiple scales for the
modeling of track geometry deterioration. Proceedings of the Institution of Mechanical
Engineers, Part F: Journal of Rail and Rapid Transit, 232(3), 842–850.
https://doi.org/10.1177/0954409717695650
Aloqaily, A. (2018). Pipeline Quantitative Risk Assessment. In Cross-Country Pipeline Risk Assessments
and Mitigation Strategies. https://doi.org/10.1016/b978-0-12-816007-7.00006-8
An, M., Chen, Y., & Baker, C. J. (2011). A fuzzy reasoning and fuzzy-analytical hierarchy process based
approach to the process of railway risk information: A railway risk management system.
Information Sciences, 181(18), 3946–3966. https://doi.org/10.1016/j.ins.2011.04.051
An, M., Lin, W., & Stirling, A. (2006). Fuzzy-reasoning-based approach to qualitative railway risk
assessment. Proceedings of the Institution of Mechanical Engineers, Part F: Journal of Rail and
Rapid Transit, 220(2), 153–167. https://doi.org/10.1243/09544097JRRT34
Anderson, R. T., & Barkan, C. P. L. (2004). Railroad accident rates for use in transportation risk
analysis. In Transportation Research Record (Issue 1863, pp. 88–98).
https://doi.org/10.3141/1863-12
Andrade, A. R., & Teixeira, P. F. (2010). Uncertainty in Rail-Track Geometry Degradation: Lisbon-
Oporto Line Case Study. Journal of Transportation Engineering, 137(3), 193–200.
https://doi.org/10.1061/(ASCE)TE.1943-5436.0000206
Andrade, A. R., & Teixeira, P. F. (2015). Statistical modelling of railway track geometry degradation
using Hierarchical Bayesian models. Reliability Engineering and System Safety, 142, 169–183.
https://doi.org/10.1016/j.ress.2015.05.009
Andrews, J. D., & Dunnett, S. J. (2000). Event-tree analysis using binary decision diagrams. IEEE
Transactions on Reliability, 49(2), 230–238. https://doi.org/10.1109/24.877343
ANSF. (2019a). Relazione annuale sulle reti ferroviarie isolate dal punto di vista funzionale di cui
all’art.2, comma4, del D.Lgs. 50/2019 (ANSF, Ed.).
ANSF. (2019b). Decree ANSF 1/2019: Technical regulations and safety standards applicable to
networks that are functionally isolated from the rest of the railway system as well as to the
service operators operating on these networks.
ANSFISA. (2021). Relazione annuale sulla sicurezza delle ferrovie anno 2020.
Aven, T. (2003). Foundations of Risk Analysis. In John Wiley & Sons (2nd ed.). Wiley.
Aven, T. (2016). Risk assessment and risk management: Review of recent advances on their
foundation. European Journal of Operational Research, 253(1), 1–13.
https://doi.org/10.1016/j.ejor.2015.12.023
Aven, T., & Renn, O. (2009). On risk defined as an event where the outcome is uncertain. Journal of
Risk Research, 12(1), 1–11. https://doi.org/10.1080/13669870802488883
Bai, Y., & Jin, W. L. (2016). Risk Assessment Methodology. In Marine Structural Design (Second edition)
(Issue 1999, pp. 709–723). https://doi.org/10.1016/b978-0-08-099997-5.00038-1
allay M. vente ová E. Urbancová Z. Monoši M. 0 9 . Application analyses of state of
evolution - ETA on selected extraordinary events. Transportation Research Procedia, 40, 1244–
1251. https://doi.org/10.1016/j.trpro.2019.07.173

222
 References

Barriquello, C. H., Garcia, V. J., Schmitz, M., Bernardon, D. P., & Fonini, J. S. (2017). A Decision Support
System for Planning and Operation of Maintenance and Customer Services in Electric Power
Distribution Systems. In System Reliability. https://doi.org/10.5772/intechopen.69721
Bearfield, G., & Marsh, W. (2005). Generalising Event Trees Using Bayesian Networks with a Case
Study of Train Derailment. In In: Winther, R., Gran, B.A., Dahll, G. (eds) Computer Safety,
Reliability, and Security. SAFECOMP 2005. Lecture Notes in Computer Science (Vol. 3688, pp. 52–
66).
Belalia, A., Kamla, Y., Amara, M., Meliani, M. H., el Azizi, A., & Azari, Z. (2020). Experimental and
numerical investigation of UIC 54 rail degradation. Engineering Failure Analysis, 111(August
2019), 104163. https://doi.org/10.1016/j.engfailanal.2019.104163
Bello, G. C., & Colombari, V. (1980). The human factors in risk analyses of process plants: The control
room operator model ‘TE EO.’ Reliability Engineering, 1(1), 3–14. https://doi.org/10.1016/0143-
8174(80)90010-4
Ben Aoun, R., El Koursi, E.-M., & Lemaire, E. (2010). The cost benefit analysis of level crossing safety
measures. 851–862. https://doi.org/10.2495/CR100771
Bernstein, P. L., & Glenn, B. J. (1999). Against The Gods: The Remarkable Story of Risk. The Journal of
Risk and Insurance, 66(3), 517. https://doi.org/10.2307/253563
Berrado, A., EL-Koursi, E.-M., Cherkaoui, A., & Khaddour, M. (2011). A Framework for Risk
Management in Railway Sector: Application to Road-Rail Level Crossings. The Open
Transportation Journal, 5(1), 34–44. https://doi.org/10.2174/1874447801105010034
Bharadwaj, U., Wintle, J., & Vadim, S. (2007). A risk based approach to maintenance optimisation of
business critical Railway Structure/equipment. Railway Network Younger Members Best Paper
Competition - Seminar on Application of New Technologies to Railways, 1–2. http://www.twi-
global.com/technical-knowledge/published-papers/a-risk-based-approach-to-maintenance-
optimisation-of-business-critical-railway-structures-equipment-november-2007/
Bochet H. (1858). Du frottement de glissement spe´cialement sur les rails des chemins de fer.
Bocz, P., Vinkó, Á., & Posgay, Z. (2018). Vibration-based condition monitoring of Tramway track from
in service vehicle using time-frequency processing techniques. Road and Rail Infrastructure V,
5(May), 631–638. https://doi.org/10.5592/co/cetra.2018.676
Britton, M. A., Asnaashari, S., & Read, G. J. M. (2017). Analysis of train derailment cause and outcome
in Victoria, Australia, between 2007 and 2013: Implications for regulation. Journal of
Transportation Safety and Security, 9(1), 45–63.
https://doi.org/10.1080/19439962.2015.1088906
Burstein, F., W. Holsapple, C., & Power, D. J. (2008). Decision Support Systems: A Historical Overview.
In Handbook on Decision Support Systems 1. https://doi.org/10.1007/978-3-540-48713-5_7
Burstein, F., W. Holsapple, C., & W. Holsapple, C. (2008). DSS Architecture and Types. In Handbook on
Decision Support Systems 1. https://doi.org/10.1007/978-3-540-48713-5_9
Camacho, D., Le, T. H., Rapp, S., & Martin, U. (2016). Light rail ballasted track geometry quality
evaluation using track recording car data. Computers in Railways XV: Railway Engineering Design
and Operation, 1(Cr), 303–315. https://doi.org/10.2495/cr160281
Cantino, V., Vincentiis, P. de, & Racca, G. (2016). Risk management. 51(4), 1–12.
Carretero, J., Pérez, J. M., García-Carballeira, F., Calderón, A., Fernández, J., García, J. D., Lozano, A.,
Cardona, L., Cotaina, N., & Prete, P. (2003). Applying RCM in large scale systems: A case study
with railway networks. Reliability Engineering and System Safety, 82(3), 257–273.
https://doi.org/10.1016/S0951-8320(03)00167-4
Cavalcante, C. A. V., & Alencar, M. H. (2012). Multicriteria decision model to support building
maintenance planning. 11th International Probabilistic Safety Assessment and Management

223
 References

Conference and the Annual European Safety and Reliability Conference 2012, PSAM11 ESREL
2012.
CEI CLC/TR 50126-2. (2017). Railway Applications. The Specification and Demonstration of Reliability,
Availability, Maintainability and Safety (RAMS) Systems Approach to Safety.
CEI EN 50126-1. (2017). Railway Applications. The Specification and Demonstration of Reliability,
Availability, Maintainability and Safety (RAMS) Generic RAMS Process.
Chand, S., Moylan, E., Waller, S. T., & Dixit, V. (2020). Analysis of vehicle breakdown frequency: A case
study of new south wales, Australia. Sustainability (Switzerland), 12(19), 1–14.
https://doi.org/10.3390/su12198244
Chang, H., Liu, R., & Wang, W. (2010). Multistage linear prediction model of track quality index.
Proceedings of the Conference on Traffic and Transportation Studies, ICTTS, 383, 1183–1192.
https://doi.org/10.1061/41123(383)112
Chapeleau, X., Cottineau, L.-M., Sedran, T., Cailliau, J., Gueguen, I., & Dumoulin, J. (2015).
Instrumentation by accelerometers and distributed optical fiber sensors of a real ballastless track
structure. In Geophysical Research Abstracts (Vol. 17).
Chen HC. (1997). On the rolling contact of high speed wheel/rail system [Doctoral dissertation.].
Academy of Railway Sciences of China.
Chen K. (2017). Experimental study on influencing factors of friction coefficient of wheel. [Master
thesis]. Lanzhou Jiaotong University.
Chenariyan Nakhaee, M., Hiemstra, D., Stoelinga, M., & van Noort, M. (2019). The Recent Applications
of Machine Learning in Rail Track Maintenance: A Survey. Lecture Notes in Computer Science
(Including Subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics),
11495 LNCS, 91–105. https://doi.org/10.1007/978-3-030-18744-6_6
Cheng, C. H., Chow, C. L., & Chow, W. K. (2021). A simulation study of tenability for passengers in a
railway tunnel with arson fire. Tunnelling and Underground Space Technology, 108.
https://doi.org/10.1016/j.tust.2020.103679
Comune di Catania. (2012). Relazione Generale del Piano Generale del Traffico Urbano.
Consilvio, A., di Febbraro, A., Meo, R., & Sacco, N. (2019). Risk-based optimal scheduling of
maintenance activities in a railway network. EURO Journal on Transportation and Logistics, 8(5),
435–465. https://doi.org/10.1007/s13676-018-0117-z
Consilvio, A., Febbraro, A. di, & Sacco, N. (2021). A Rolling-Horizon Approach for Predictive
Maintenance Planning to Reduce the Risk of Rail Service Disruptions. IEEE Transactions on
Reliability, 70(3), 875–886. https://doi.org/10.1109/TR.2020.3007504
Cooke, R. M., & Goossens, L. H. J. (2004). Expert judgement elicitation for risk assessments of critical
infrastructures. Journal of Risk Research, 7(6), 643–656.
https://doi.org/10.1080/1366987042000192237
Corbin, J. C., & Fazio, A. E. (1981). Performance-based track-quality measures and their application to
maintenance-of-way planning. Transportation Research Record, 802, 19–27.
Cosulich, G., Firpo, P., & Savio, S. (1996). Power electronics reliability impact on service dependability
for railway systems: a real case study. IEEE International Symposium on Industrial Electronics, 2,
996–1001. https://doi.org/10.1109/isie.1996.551080
Covello, V. T. (1987). Decision Analysis and Risk Management Decision Making: Issues and Methods.
Risk Analysis, 7(2), 131–139. https://doi.org/10.1111/j.1539-6924.1987.tb00978.x
Croft, B. E., Vollebregt, E. A. H., & Thompson, D. J. (2012). NNFM 118 - An Investigation of Velocity-
Dependent Friction in Wheel-Rail Rolling Contact. Rail Trans. Sys, 118, 33–41.

224
 References

Croft, B., Jones, C., & Thompson, D. (2011). Velocity-dependent friction in a model of wheel-rail rolling
contact and wear. Vehicle System Dynamics, 49(11), 1791–1802.
https://doi.org/10.1080/00423114.2010.543138
Deng, Y., Song, L., Zhou, J., Xu, N., Ni, G., & Wang, L. (2020). Analysis of Failures and Influence Factors
of Critical Infrastructures: A Case of Metro. Advances in Civil Engineering, 2020.
https://doi.org/10.1155/2020/2301276
Di Graziano, A., & Marchetta, V. (2021). A risk-based decision support system in local railways
management. Journal of Rail Transport Planning and Management, 20(September), 100284.
https://doi.org/10.1016/j.jrtpm.2021.100284
Di Graziano, A., Marchetta, V., & Cafiso, S. (2020). Structural health monitoring of asphalt pavements
using smart sensor networks: A comprehensive review. Journal of Traffic and Transportation
Engineering (English Edition). https://doi.org/10.1016/j.jtte.2020.08.001
Dindar, S., Kaewunruen, S., An, M., & Gigante-Barrera, A. (2017). Derailment-based Fault Tree Analysis
on Risk Management of Railway Turnout Systems. IOP Conference Series: Materials Science and
Engineering, 245(4). https://doi.org/10.1088/1757-899X/245/4/042020
olše M. 0 . implified method for seismic ris assessment of buildings with consideration of
aleatory and epistemic uncertainty. Structure and Infrastructure Engineering, 8(10), 939–953.
https://doi.org/10.1080/15732479.2011.574813
Eker, O. F., Camci, F., Guclu, A., Yilboga, H., Sevkli, M., & Baskan, S. (2011). A simple state-based
prognostic model for railway turnout systems. IEEE Transactions on Industrial Electronics, 58(5).
https://doi.org/10.1109/TIE.2010.2051399
Elkhoury, N., Hitihamillage, L., Moridpour, S., & Robert, D. (2018). Degradation Prediction of Rail
Tracks: A Review of the Existing Literature. The Open Transportation Journal, 12(1), 88–104.
https://doi.org/10.2174/1874447801812010088
El-Sibaie, M., & Zhang, Y. J. (2004). Objective track quality indices. Transportation Research Record,
1863, 81–87. https://doi.org/10.3141/1863-11
EN 13848-1. (2019). Railway applications - Track - Track geometry quality – Part 1: Characterisation of
track geometry.
EN 13848-5. (2017). Railway applications - Track - Track geometry quality - Part 5: Geometric quality
levels - Plain line.
EN 13848-6. (2021). Railway applications - Track - Track geometry quality - Part 6: Characterisation of
track geometry quality.
Ernst-Basler. (2004). Technical Documentation Identification of Safety Standards for Tunnels.
Esveld, C. (2001). Modern Railway Track 2nd Edition (Vol. 148).
European Commission. (2014). COMMISSION REGULATION (EU) No 1303/2014 of 18 November 2014
concerning the technical specification for interoperability relating to ‘safety in railway tunnels’ of
the rail system of the European Union. In Official Journal of the European Union (Vol. 57, Issue L
356, pp. 394–420).
European Parliament. (2004). DIRECTIVE 2004/49/CE of the European Parliament and of the Council of
29 april 2004. Official Journal of the European Union, 44–113.
European Railway Agency. (2018). Report on Railway Safety and Interoperability in the EU. European
Union agency for railways. https://doi.org/10.2821/30980
European Union. (2013). Commission implementing regulation (EU) No 402/2013 on the common
safety method for risk evaluation and assessment and repealing Regulation (EC) No 352/2009. In
Official Journal of the European Union.
European Union. (2016). DIRECTIVE (EU) 2016/798 of the European Parliament and of the Council of
11 May 2016 on railway safety. Official Journal of the European Union.

225
 References

Evans . W. 0 . Fatal train accidents on Europe’s railways 980-2009. Accident Analysis and
Prevention, 43(1), 391–401. https://doi.org/10.1016/j.aap.2010.09.009
Falamarzi, A., Moridpour, S., & Nazem, M. (2019a). A review of rail track degradation prediction
models. Australian Journal of Civil Engineering, 17(2), 152–166.
https://doi.org/10.1080/14488353.2019.1667710
Falamarzi, A., Moridpour, S., & Nazem, M. (2019b). Development of a tram track degradation
prediction model based on the acceleration data. Structure and Infrastructure Engineering,
15(10), 1308–1318. https://doi.org/10.1080/15732479.2019.1615963
Falamarzi, A., Moridpour, S., Nazem, M., & Hesami, R. (2017). Rail degradation predication:
Melbourne tram system case study. ATRF 2017 - Australasian Transport Research Forum 2017,
Proceedings, 2018.
Famurewa, S. M., Asplund, M., Rantatalo, M., Parida, A., & Kumar, U. (2015). Maintenance analysis for
continuous improvement of railway infrastructure performance. Structure and Infrastructure
Engineering, 11(7), 957–969. https://doi.org/10.1080/15732479.2014.921929
Farmer, F. R. (1967). Siting Criteria - A New Approach. Containment and Siting of Nuclear Power Plants
Proceedings of a Symposium, SM-89/34.
Federici, A. (1999). Lo scartamento ridotto in Italia. La storia. I treni. Le linee (Tuttotreno). Duegi
Editrice.
Ferreira, L., & Murray, M. H. (1997). Modelling rail track deterioration and maintenance: current
practices and future needs. Transport Reviews, 17(3), 207–221.
https://doi.org/10.1080/01441649708716982
Finney, D. J. (1971). Probit Analysis, Cambridge University Press. In Journal of Pharmaceutical Sciences
(Vol. 60, Issue 9).
Gits, C. W. (1992). Design of maintenance concepts. International Journal of Production Economics,
24(3), 217–226. https://doi.org/10.1016/0925-5273(92)90133-R
Guler, H. (2013). Decision Support System for Railway Track Maintenance and Renewal Management.
Journal of Computing in Civil Engineering, 27(3), 292–306.
https://doi.org/10.1061/(asce)cp.1943-5487.0000221
Hamid, A., & Gross, A. (1981). Track-quality indices and track-degradation models for maintenance-of-
way planning. Transportation Research Record, 802.
Hanly, P. A., & Sharp, L. (2014). The cost of lost productivity due to premature cancer-related
mortality: an economic measure of the cancer burden. BMC Cancer, 14(1), 224.
https://doi.org/10.1186/1471-2407-14-224
Hasheminezhad, A., Hadadi, F., & Shirmohammadi, H. (2021). Investigation and prioritization of risk
factors in the collision of two passenger trains based on fuzzy COPRAS and fuzzy DEMATEL
methods. Soft Computing, 25(6), 4677–4697. https://doi.org/10.1007/s00500-020-05478-3
Hodge V. J. O’Keefe . Wee s M. Moulds . 0 . Wireless sensor networ s for condition
monitoring in the railway industry: A survey. IEEE Transactions on Intelligent Transportation
Systems, 16(3), 1088–1106. https://doi.org/10.1109/TITS.2014.2366512
Hubbard, D. W. (2020). The Failure of isk Management: why it’s broken and how to fix it (2nd ed.).
John Wiley & Sons. https://doi.org/10.1002/9781119198536
IEC. (2006). IEC 61025: Fault tree analysis. In 61010-1 © Iec:2001.
International Organization for Standardization. (2009). ISO/IEC 31010:2009 Risk management - Risk
assessment techniques. Risk Management, 31010, 92.
International Organization for Standardization. (2018). ISO 31000:2018 - Risk management - Principles
and guidelines. In ISO 31000:2018.

226
 References

Italian Ministry of Infrastructures and Transport. (2005). Ministry of Infrastructures and Transport
decree of 28 october 2005 “Safety in railway tunnels” (Italian language).
Jardine, A. K. S. (2002). Optimizing condition based maintenance decisions. Proceedings of the Annual
Reliability and Maintainability Symposium, 90–97. https://doi.org/10.1109/rams.2002.981625
Jardine, A. K. S., Lin, D., & Banjevic, D. (2006). A review on machinery diagnostics and prognostics
implementing condition-based maintenance. Mechanical Systems and Signal Processing, 20(7),
1483–1510. https://doi.org/10.1016/j.ymssp.2005.09.012
Jiang, L., Wang, X., & Liu, Y. (2018). Reliability evaluation of the Chinese Train Control System Level 3
using a fuzzy approach. Proceedings of the Institution of Mechanical Engineers, Part F: Journal of
Rail and Rapid Transit, 232(9), 2244–2259. https://doi.org/10.1177/0954409718769193
Jones-Lee, M., & Aven, T. (2011). ALARP - What does it really mean? Reliability Engineering and
System Safety, 96(8), 877–882. https://doi.org/10.1016/j.ress.2011.02.006
Joung, E. J. (2005). Application of safety management process for the safety analysis of level crossing.
ICEMS 2005: Proceedings of the Eighth International Conference on Electrical Machines and
Systems, 3, 2493–2497. https://doi.org/10.1109/icems.2005.203024
Jovanović . Guler H. Čo o . 0 . Track degradation analysis in the scope of railway
infrastructure maintenance management systems. Gradevinar, March.
https://doi.org/10.14256/JCE.1194.2014
Khajehei, H., Ahmadi, A., Soleimanmeigouni, I., & ... (2019). Application of principal component
analysis and artificial neural network in prediction of track geometry degradation. International
Heavy Haul STS Conference 2019, June.
Khajehei, H., Ahmadi, A., Soleimanmeigouni, I., & Nissen, A. (2019). Allocation of effective
maintenance limit for railway track geometry. Structure and Infrastructure Engineering, 15(12),
1597–1612. https://doi.org/10.1080/15732479.2019.1629464
Khosravi, M., Soleimanmeigouni, I., Ahmadi, A., & Nissen, A. (2021). Reducing the positional errors of
railway track geometry measurements using alignment methods: A comparative case study.
Measurement: Journal of the International Measurement Confederation, 178(April), 109383.
Kim, M.-S., Wang, J.-B., Park, C.-W., & Cho, Y.-O. (2009). Development of the Risk Assessment Model
for Railway Level-Crossing Accidents by Using The ETA and FTA. Journal of the Korean Society for
Railway, 12(6), 936–943.
Klinke, A., & Renn, O. (2002). A new approach to risk evaluation and management: Risk-based,
precaution-based, and discourse-based strategies. Risk Analysis. https://doi.org/10.1111/1539-
6924.00274
Lasisi, A., & Attoh-Okine, N. (2018). Principal components analysis and track quality index: A machine
learning approach. Transportation Research Part C: Emerging Technologies, 91(April 2018), 230–
248. https://doi.org/10.1016/j.trc.2018.04.001
Leitner, B. (2017). A General Model for Railway Systems Risk Assessment with the Use of Railway
Accident Scenarios Analysis. Procedia Engineering, 187, 150–159.
https://doi.org/10.1016/j.proeng.2017.04.361
Letot, C., Soleimanmeigouni, I., Ahmadi, A., & Dehombreux, P. (2016). An adaptive opportunistic
maintenance model based on railway track condition prediction. IFAC-PapersOnLine, 49(28).
https://doi.org/10.1016/j.ifacol.2016.11.021
Li, R., Salling, K. B., Zoeteman, A., & Wolfert, A. R. M. (2017). Preventive condition-based tamping for
railway tracks: A decision support model. Life-Cycle of Engineering Systems: Emphasis on
Sustainable Civil Infrastructure - 5th International Symposium on Life-Cycle Engineering, IALCCE
2016. https://doi.org/10.1201/9781315375175-284

227
 References

Li, Y.-F., Mi, J., Huang, H.-Z., Zhu, S.-P., & Xiao, N. (2013). Fault tree analysis of train rear-end collision
accident cosidering common cause failure. Mantainance and Reliability, 15(4), 403–408.
Lin, C. Y., Rapik Saat, M., & Barkan, C. P. L. (2020). Quantitative causal analysis of mainline passenger
train accidents in the United States. Proceedings of the Institution of Mechanical Engineers, Part
F: Journal of Rail and Rapid Transit, 234(8), 869–884.
https://doi.org/10.1177/0954409719876128
Lin, C. Y., & Saat, M. R. (2014). Semi-quantitative risk assessment of adjacent track accidents on
shared-use rail corridors. 2014 Joint Rail Conference, JRC 2014, September, 2–6.
https://doi.org/10.1115/JRC2014-3773
Lins, P., Almeida-Filho, A., Alencar, M., & Almeida, A. (2012). A DSS for multiple dimensionrisk
evaluation of a gas pipeline. 11th International Probabilistic Safety Assessment and Management
Conference and the Annual European Safety and Reliability Conference 2012, PSAM11 ESREL
2012.
Liu, P., Yang, L., Gao, Z., Li, S., & Gao, Y. (2015). Fault tree analysis combined with quantitative analysis
for high-speed railway accidents. Safety Science, 79, 344–357.
https://doi.org/10.1016/j.ssci.2015.06.017
Liu, X. (2016). Analysis of collision risk for freight trains in the United States. Transportation Research
Record, 2546, 121–128. https://doi.org/10.3141/2546-15
Liu, X., Barkan, C. P. L., & Saat, M. R. (2011). Analysis of derailments by accident cause: Evaluating
railroad track upgrades to reduce transportation risk. Transportation Research Record, 2261,
178–185. https://doi.org/10.3141/2261-21
Liu, X. H., Han, M., Lin, X. H., & Yang, N. P. (2014). Quantitative safety assessment based on risk in
level-crossing between railway and highway. Applied Mechanics and Materials, 536–537, 854–
857. https://doi.org/10.4028/www.scientific.net/AMM.536-537.854
Liu, X., Saat, M. R., & Barkan, C. P. L. (2012). Analysis of causes of major train derailment and their
effect on accident rates. Transportation Research Record, 2289, 154–163.
https://doi.org/10.3141/2289-20
López-Pita, A., Teixeira, P. F., Casas, C., Bachiller, A., & Ferreira, P. A. (2008). Maintenance costs of
high-speed lines in Europe state of the art. Transportation Research Record, 2043, 13–19.
https://doi.org/10.3141/2043-02
Lyngby, N. (2009). Railway track degradation: Shape and influencing factors. International Journal of
Performability Engineering, 5(2), 177–186.
Mahboob, Q., Schöne, E., Maschek, U., & Trinckauf, J. (2015). Investment into Human Risks in Railways
and Decision Optimization. Human and Ecological Risk Assessment: An International Journal,
21(5), 1299–1313. https://doi.org/10.1080/10807039.2014.958375
Marchetta V. & Di Graziano A. & Contino F. (2022). A methodology for introducing the impact of risk
analysis in local railways improvements decisions. TIS Roma 2022 Conference Proceedings,
Transportation Research Procedia.
Marchetta, V.,Di Graziano, A., & Contino, F. (2023). A methodology for introducing the impact of risk
analysis in local railways improvements decisions. Transportation Research Procedia, 69, 424–
431. https://doi.org/10.1016/j.trpro.2023.02.191
Meier-Hirmer, C., Senée, A., Riboulet, G., Sourget, F., & Roussignol, M. (2006). A decision support
system for track maintenance. WIT Transactions on the Built Environment, 88, 217–226.
https://doi.org/10.2495/CR060221
Melchers, R. E. (2001). On the ALARP approach to risk management. Reliability Engineering and
System Safety, 71(2), 201–208. https://doi.org/10.1016/S0951-8320(00)00096-X

228
 References

Modarres, M. (1993). What Every Engineer Should Know about Reliability and Risk Analysis. In
Technometrics (Vol. 30). CRC Press. https://doi.org/10.2307/1270248
Moghaddam, K. S., & Usher, J. S. (2011). Sensitivity analysis and comparison of algorithms in
preventive maintenance and replacement scheduling optimization models. Computers and
Industrial Engineering, 61(1). https://doi.org/10.1016/j.cie.2011.02.012
Moridpour, S., Mazloumi, E., & Hesami, R. (2016). Application of artificial neural networks in
predicting the degradation of tram tracks using maintenance data. Applied Big Data Analytics in
Operations Management, 30–54. https://doi.org/10.4018/978-1-5225-0886-1.ch002
Muttram R. . 00 a . Railway safety’s safety ris model. Proceedings of the Institution of Mechanical
Engineers, Part F: Journal of Rail and Rapid Transit, 216(2), 71–79.
https://doi.org/10.1243/09544090260082317
Nimdet, K., & Ngorsuraches, S. (2015). Willingness to pay per quality-adjusted life year for life-saving
treatments in Thailand. BMJ Open, 5(10), e008123. https://doi.org/10.1136/bmjopen-2015-
008123
Nowa ows i W. Cis ews i T. Młyńc a J. Łu asi , Z. (2018). Failure Evaluation of the Level
Crossing Protection System Based on Fault Tree Analysis. Lecture Notes in Networks and
Systems, 21, 107–115. https://doi.org/10.1007/978-3-319-64084-6_10
Odoki, J. B., Di Graziano, A., & Akena, R. (2015). A multi-criteria methodology for optimising road
investments. Proceedings of the Institution of Civil Engineers: Transport, 168(TR1), 34–47.
https://doi.org/10.1680/tran.12.00053
Power, C., Mian, J., Spink, T., Abbott, S., & Edwards, M. (2016). Development of an Evidence-based
Geotechnical Asset Management Policy for Network Rail, Great Britain. Procedia Engineering,
143(Ictg), 726–733. https://doi.org/10.1016/j.proeng.2016.06.112
Profillidis, V. A. (2017). Railway Management and Engineering (4th ed., Vol. 15, Issue 2). Routledge.
https://doi.org/10.4324/9781351150842
Rasmussen, J. (1997). Risk Management in a dynamic society: a modelling problem. Safety Science,
27(2/3), 183–213.
Rausand, M., & Stein, H. (2020). Risk Assessment Theory, Methods, and Applications. In Statistics in
Practice (2nd ed., Vol. 42, Issue 4). Wiley. https://doi.org/10.1017/CBO9781107415324.004
Raza, S., Mahboob, Q., Khan, A., Khan, T., & Hussain, J. (2020). Computation of Importance Measures
Using Bayesian Networks for the Reliability and Safety of Complex Systems. International Journal
of Reliability, Risk and Safety: Theory and Application, 3(2), 99–111.
https://doi.org/10.30699/ijrrs.3.2.12
Rezvani, A. Z., Peach, M., Thomas, A., Cruz, R., & Kemmsies, W. (2015). Benefit-cost Methodology for
Highway-railway Grade Crossing Safety Protocols as Applied to Transportation Infrastructure
Project Prioritization Processes. Transportation Research Procedia, 8, 89–102.
https://doi.org/10.1016/J.TRPRO.2015.06.045
RFI. (1971). Circular 26: Railway Rolling Stock. Technical Procedures for Carrying out Tests on Braking
Systems (Italian Lenguage).
RFI. (2006). General methodology for the classification of infrastructures in terms of risk.
RFI. (2022). The network today.
Rhayma, N., Bressolette, P., Breul, P., Fogli, M., & Saussine, G. (2013). Reliability analysis of
maintenance operations for railway tracks. Reliability Engineering and System Safety, 114(1).
https://doi.org/10.1016/j.ress.2012.12.007
RSSB. (2018). Safety Risk Model: Risk Profile Bulletin.

229
 References

Ryen, L., & Svensson, M. (2015). The Willingness to Pay for a Quality Adjusted Life Year: A Review of
the Empirical Literature. Health Economics, 24(10), 1289–1301.
https://doi.org/10.1002/hec.3085
Sasidharan, M., Burrow, M. P. N., Ghataora, G. S., & Marathu, R. (2022). A risk-informed decision
support tool for the strategic asset management of railway track infrastructure. Proceedings of
the Institution of Mechanical Engineers, Part F: Journal of Rail and Rapid Transit, 236(2), 183–
197. https://doi.org/10.1177/09544097211038373
Sasidharan, M., Burrow, M. P. N., Ghataora, G. S., & Torbaghan, M. E. (2017). A Review of Risk
Management Applications For Railways. Railway Engineering, June.
https://doi.org/10.25084/raileng.2017.0065
Sato, Y. (1995). Japanese Studies on Deterioration of Ballasted Track. Vehicle System Dynamics,
24(sup1), 197–208. https://doi.org/10.1080/00423119508969625
Sklet, S. (2006). Safety barriers: Definition, classification, and performance. Journal of Loss Prevention
in the Process Industries, 19(5), 494–506. https://doi.org/10.1016/j.jlp.2005.12.004
Smithson, M. (2014). Elicitation. In Introduction to Imprecise Probabilities.
https://doi.org/10.1002/9781118763117.ch15
Soh, S. S., Radzi, N. H. M., & Haron, H. (2012). Review on scheduling techniques of preventive
maintenance activities of railway. Proceedings of International Conference on Computational
Intelligence, Modelling and Simulation, 310–315. https://doi.org/10.1109/CIMSim.2012.56
Soleimanmeigouni, I. (2019). Predictive Models for Railway Track Geometry Degradation.
Soleimanmeigouni, I., Ahmadi, A., Khajehei, H., & Nissen, A. (2020). Investigation of the effect of the
inspection intervals on the track geometry condition. Structure and Infrastructure Engineering,
16(8), 1138–1146. https://doi.org/10.1080/15732479.2019.1687528
Soleimanmeigouni, I., Ahmadi, A., Khouy, I. A., & Letot, C. (2018). Evaluation of the effect of tamping
on the track geometry condition: A case study. Proceedings of the Institution of Mechanical
Engineers, Part F: Journal of Rail and Rapid Transit, 232(2), 408–420.
https://doi.org/10.1177/0954409716671548
Soleimanmeigouni, I., Ahmadi, A., & Kumar, U. (2016). Track geometry degradation and maintenance
modelling: A review. Proceedings of the Institution of Mechanical Engineers, Part F: Journal of
Rail and Rapid Transit, 232(1), 73–102. https://doi.org/10.1177/0954409716657849
Soleimanmeigouni, I., Ahmadi, A., Nissen, A., & Xiao, X. (2020). Prediction of railway track geometry
defects: a case study. Structure and Infrastructure Engineering, 16(7), 987–1001.
https://doi.org/10.1080/15732479.2019.1679193
Sorrill, C. M., Cooper, D., & Chapman, C. (1987). Risk Analysis for Large Projects: Models, Methods and
Cases. The Journal of the Operational Research Society. https://doi.org/10.2307/2582754
Špač ová O. traub . 0 . Cost-Benefit analysis for optimization of risk protection under
budget constraints. Risk Analysis, 35(5), 941–959. https://doi.org/10.1111/risa.12310
ubotić T. Vasiljević M. 0 ecember . Influence of Train Stopping Distance and Overlap on the
Railway Traffic Safety. https://doi.org/10.20544/tts2021.1.1.21.p05
Sun, J., Zhang, R., & Qin, S. (2017). Turnout maintenance scheduling problem considering reliabilities:
Modeling and optimization. Chinese Control Conference, CCC, 10143–10148.
https://doi.org/10.23919/ChiCC.2017.8028973
Tezuka, M., Munakata, S., & Sawada, M. (2015). Maintenance schedule optimization based on failure
probability distribution. IEOM 2015 - 5th International Conference on Industrial Engineering and
Operations Management, Proceeding, 5–9. https://doi.org/10.1109/IEOM.2015.7093732

230
 References

The European Commission. (2013). Commission Implementing Regulation (EU) No 402/2013 of 30

April 2013 on the common safety method for risk evaluation and assessment and repealing
Regulation (EC) No 352/2009. Official Journal of the European Union, 56(L 121).
Turla, T., Liu, X., & Zhang, Z. (2019). Analysis of freight train collision risk in the United States.
Proceedings of the Institution of Mechanical Engineers, Part F: Journal of Rail and Rapid Transit,
233(8), 817–830. https://doi.org/10.1177/0954409718811742
UIC. (2002). Codex 779-9e: Safety in Railway Tunnels.
UIC. (2022). UIC safety report 2022.
Valdez Banda, O. A., Jalonen, R., Goerlandt, F., Montewka, J., & Kujala, P. (2014). Hazard Identification
in Winter Navigation. In Aalto University (Vol. 15, Issue November 2015).
Vale, C., & M. Lurdes, S. (2013). Stochastic model for the geometrical rail track degradation process in
the Portuguese railway Northern Line. Reliability Engineering and System Safety, 116, 91–98.
https://doi.org/10.1016/j.ress.2013.02.010
Vale, C., Ribeiro, I. M., & Calçada, R. (2011). Integer programming to optimize tamping in railway
tracks as preventive maintenance. Journal of Transportation Engineering, 138(1), 123–131.
https://doi.org/10.1061/(ASCE)TE.1943-5436.0000296
Van Weyenberge, B., Deckers, X., Caspeele, R., & Merci, B. (2016). Development of a Risk Assessment
Method for Life Safety in Case of Fire in Rail Tunnels. Fire Technology, 52(5), 1465–1479.
https://doi.org/10.1007/s10694-015-0469-y
Velha, P., Nannipieri, T., Signorini, A., Solazzi, M., Di Pasquale, F., Lupi, M., Lisi, S., Iacobini, F., & Firmi,
P. (2021). Monitoring of main railway infrastructures through a “R M N/F G” hybrid fiber optic
sensor system. Ingegneria Ferroviaria.
Viana, F. F. C. L., Casado, R. S. G. R., da Silva, L. B. L., Alencar, M. H., Ferreira, R. J. P., & de Almeida, A.
T. (2022). A hybrid multicriteria decision model for selecting a portfolio of risk-based
maintenance actions in natural gas pipelines. Journal of Natural Gas Science and Engineering,
103. https://doi.org/10.1016/j.jngse.2022.104655
Vílchez, J. A., Montiel, H., Casal, J., & Arnaldos, J. (2001). Analytical expressions for the calculation of
damage percentage using the probit methodology. In Journal of Loss Prevention in the Process
Industries (Vol. 14). www.elsevier.com/locate/jlp
Vollebregt, E. A. H., & Schuttelaars, H. M. (2012). Quasi-static analysis of two-dimensional rolling
contact with slip-velocity dependent friction. Journal of Sound and Vibration, 331(9), 2141–2155.
https://doi.org/10.1016/j.jsv.2012.01.011
Wang, M.-Y., Wang, H., & Liu, Z.-G. (2014). Reach on Fault Tree Analysis of Train Derailment in Urban
Rail Transit. International Journal of Business and Social Science, 5(8), 128–135.
http://ijbssnet.com/journals/Vol_5_No_8_July_2014/13.pdf
Wen, M., Li, R., & Salling, K. B. (2016). Optimization of preventive condition-based tamping for railway
tracks. European Journal of Operational Research, 252(2).
https://doi.org/10.1016/j.ejor.2016.01.024
Williams, J. C. (1986). HEART - A proposed method for assessing and reducing human error.
Proceedings - Advances in Reliability Technology Symposium.
Won hua, H. (2004, March). The progress and controlling Situation of Daegu Subway Fire Disaster. The
Proceedings of 6th Asia-Oceania Symposium on Fire Science and Technology.
Wu, Q., Azad, A. K., Cole, C., & Spiryagin, M. (2022). Identify severe track geometry defect
combinations for maintenance planning. International Journal of Rail Transportation, 10(1), 95–
113. https://doi.org/10.1080/23248378.2021.1871673
Xie, G., Allen, P. D., Iwnicki, S. D., Alonso, A., Thompson, D. J., Jones, C. J. C., & Huang, Z. Y. (2006).
Introduction of falling friction coefficients into curving calculations for studying curve squeal

231
 References

noise. Vehicle System Dynamics, 44(SUPPL. 1), 261–271.

https://doi.org/10.1080/00423110600870428
Xie, J., Huang, J., Zeng, C., Jiang, S. H., & Podlich, N. (2020). Systematic literature review on data-driven
models for predictive maintenance of railway track: Implications in geotechnical engineering.
Geosciences (Switzerland), 10(11), 1–24. https://doi.org/10.3390/geosciences10110425
Yousefikia, M., Moridpour, S., Setunge, S., & Mazloumi, E. (2014). Modeling Degradation of Tracks for
Maintenance Planning on a Tram Line. Journal of Traffic and Logistics Engineering, 2(2), 86–91.
https://doi.org/10.12720/jtle.2.2.86-91
Yuan, Z., Wu, M., Tian, C., Zhou, J., & Chen, C. (2021). A Review on the Application of Friction Models
in Wheel-Rail Adhesion Calculation. In Urban Rail Transit (Vol. 7, Issue 1). Springer Science and
Business Media Deutschland GmbH. https://doi.org/10.1007/s40864-021-00141-y
Zarembski, A. M., & Palese, J. W. (2006). Managing Risk on the Railway Infrastructure. World Congress
Railway Research, 1–18.
Zhang, S., Pedersen, P. T., & Villavicencio, R. (2019). Probability of ship collision and grounding. In
Probability and Mechanics of Ship Collision and Grounding (pp. 1–61).
https://doi.org/10.1016/b978-0-12-815022-1.00001-3
Zhang, T., Andrews, J., & Wang, R. (2013). Optimal scheduling of track maintenance on a railway
network. Quality and Reliability Engineering International, 29(2).
https://doi.org/10.1002/qre.1381
Zhang, W., Chen, J., Wu, X., & Jin, X. (2002). Wheel/rail adhesion and analysis by using full scale roller
rig. Wear, 253, 82–88.
Zhang, X., Deng, Y., Li, Q., Skitmore, M., & Zhou, Z. (2016). An incident database for improving metro
safety: The case of shanghai. Safety Science, 84, 88–96.
https://doi.org/10.1016/j.ssci.2015.11.023
Zoeteman, A. (2002). Computer assisted revision of infrastructure maintenance policy. Transport
Policy, figure 1.

232