See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/346529996

Damage assessment framework for tunnel structures subjected to fire

Conference Paper · December 2020

DOI: 10.14264/a0b3b36

CITATION READS
1 188

3 authors:

Nan Hua Anthony Tessari

University at Buffalo, The State University of New York University at Buffalo, The State University of New York
7 PUBLICATIONS   17 CITATIONS    33 PUBLICATIONS   114 CITATIONS   

SEE PROFILE SEE PROFILE

Negar Elhami Khorasani

University at Buffalo, The State University of New York
79 PUBLICATIONS   523 CITATIONS   

SEE PROFILE

Some of the authors of this publication are also working on these related projects:

Fire Fragility Functions for Risk Assessment of Buildings in Fire View project

Probabilistic Risk Assessment & Fire Safety Engineering View project

All content following this page was uploaded by Negar Elhami Khorasani on 01 December 2020.

The user has requested enhancement of the downloaded file.

SiF 2020 – The 11th International Conference on Structures in Fire
The University of Queensland, Brisbane, Australia, 30th November - 2nd December, 2020

DAMAGE ASSESSMENT FRAMEWORK FOR TUNNEL STRUCTURES

SUBJECTED TO FIRE

Nan Hua1, Anthony Frederick Tessari2, Negar Elhami-Khorasani3

ABSTRACT
Fire hazards are a major threat to tunnel structures. The rapid rise of gas temperatures exceeding 1000 °C
inside a confined tunnel space as well as long fire durations due to limited emergency responder access
necessitate special analyses and design considerations. Intense lining damage, mostly due to concrete
spalling, has been observed after major tunnel fires. Although collapse of the tunnel structure is rare, the
serviceability and resilience of the tunnel structure are challenged, as major fire events could lead to months
of downtime and millions of dollars of losses from repair and affected operations. This research will propose
an assessment methodology to quantify tunnel fire damage to tunnel linings. While the framework is applied
to railway tunnels, the approach is general and can be extended to roadway tunnels. The ultimate goal is to
minimize economic losses and improve resilience of the transportation network by ensuring that individual
structures, such as tunnels within the network, are resilient when subjected to extreme events.
Keywords: Tunnel fire; fire scenarios; concrete spalling; damage assessment

1 INTRODUCTION
Fire hazards are major threats and can cause severe damage to tunnel structures. There is a long list of
historical tunnel fires, including some extreme events with a large number of fatalities, injuries, and
significant socioeconomic impacts. On May 29, 1999, a truck crash resulted in a major fire in the Tauern
Tunnel connecting the industrial areas of southern Germany and northern Italy. A total of 24 cars and 16
trucks were completely burnt. Severe concrete spalling was observed on the tunnel ceiling and sidewalls.
In the area that underwent the highest temperatures, the concrete lining spalled to a depth of 400 mm over
a 100 m tunnel length. Around 600 m3 of chip-size spalled concrete had to be removed after the fire. The
downtime of the tunnel was three months and 6.5 million US dollars were spent on the repair work, while
another 19.5 million US dollars were lost via uncollected toll fees [1]. Similar lessons can be learned from
other major tunnel fire events, such as the Mont Blanc Tunnel fire in 1999 and the multiple Channel Tunnel
fires that occurred in 1996, 2006 and 2008. Among the long list of tunnel fire events, although collapse or
local failure of the tunnel structures due to fire are rarely observed, the serviceability and resilience of the
tunnel structures are impacted, as major fire events could lead to months of downtime and millions of
dollars of losses from repair and affected operations.
Tunnel fires have become a growing concern and received increasing attention in recent decades. On one
hand, an increasing density in the transport of goods, especially of flammable materials, increases the
amount of potential fuel, and the likelihood for a fast fire spread in the case of a fire within a tunnel space.

1 Ph.D.
Candidate, University at Buffalo
e-mail: nanhua@buffalo.edu, ORCID: https://orcid.org/0000-0002-3046-9331
2 Assistant Professor, University at Buffalo

e-mail: atessari@buffalo.edu, ORCID: https://orcid.org/0000-0003-0231-345X

3 Assistant Professor, University at Buffalo

e-mail: negarkho@buffalo.edu, ORCID: https://orcid.org/0000-0003-3228-0097

https://doi.org/10.14264/a0b3b36
On the other hand, modern tunnel design uses higher strength concrete and thinner linings. Although the
use of high-strength concrete saves cost for materials, it may lead to severe spalling, and consequently loss
of a large fraction of lining sections in the case of a fire. While the historical events identified the disastrous
consequences of tunnel fires with respect to conventional materials, the newly raised concerns associated
with modern tunnels further emphasize the need to design tunnels for fire.
In the existing research areas of tunnel fires, there is a relatively large body of literature on fire propagation
and smoke control inside tunnels, as well as relevant provisions on evacuation routes in tunnel spaces to
ensure passenger safety. However, very little focus has been placed on the damage assessment of tunnel
structures due to fire. Meanwhile, fire is not widely considered in the current engineering practice of tunnel
design. There are limited official documents providing guidance on designing tunnels for fire, most of
which focus mainly on the design of emergency ventilation systems. No performance-based framework
exists in the guidelines to design tunnel liners for loss of functionality following a fire event.
The performance-based design (PBD) method provides improved engineered solutions based on realistic
characterization of fire hazards and allows higher design flexibility and a better understanding of the fire
performance of tunnel structures. A comprehensive performance-based approach requires a clear definition
of design objectives, assessment of performance, and demonstration of compliance with the required
performance objectives. Developing an understanding for tunnel fire damages, by investigating the
influences of the key parameters, provides references for PBD for fire in tunnels based on performance
requirements, such as the acceptable cost (for damage repair) given a fire demand.
The objective of this paper is to propose a framework to quantify fire damage to tunnel linings considering
the uncertainties in fire scenarios. While the following paper focuses on the implementation of the
framework on a rail tunnel, the methodology is general and can be extended to roadway tunnels. The
outcome can be used to assess downtime and decide on proper mitigation measures, such as the design of
passive fire protection of tunnel linings to minimize life cycle costs and improve resilience of the tunnel
structure subjected to fire.

2 METHODOLOGY
The framework for fire damage assessment of tunnels follows the three typical steps of solving structural
fire engineering problems: (1) determining the fire scenario (temperature-time curve), (2) performing
thermal (heat transfer) analysis, and (3) performing structural analysis that considers the thermal load. The
rest of this section will introduce the three steps for tunnel fire damage assessment using a case study of a
prototype rail tunnel. The geometry of the prototype cross-section is adapted from the Howard Street Tunnel
in Baltimore, Maryland, which serves as an integral part of CSX operations on the eastern coast of the US
[2]. As shown in Figure 1(a), the cross-section is horseshoe-shape, 6.7-meters tall and 8.2-meters wide. The
real Howard Street Tunnel was built in 1895 and is made of thick masonry. In this study, it is assumed that
the tunnel is instead made of concrete with a thickness of 0.3 m to reflect a modern design.
2.1 Computational fluid dynamics modelling of fire scenarios
A tunnel fire scenario describes the evolution of gas temperatures over time during a fire event, which is
primarily influenced by the tunnel geometry, fire size (i.e., energy released), ventilation conditions, and
characteristics of fire spread. Although the procedure of defining the fire scenario involves a high level of
uncertainty due to variation of the influencing factors, the design fire curves currently being used are
deterministic [3], which do not consider the fire duration, spread of fire within the tunnel, and the unique
environmental conditions during a tunnel fire. Limited studies are available in the literature on quantifying
uncertainties in tunnel fire demand (e.g. heat release rate) [4-6]. However, the uncertainty in temperature-
time evolution of a tunnel fire has not been investigated. In addition, the evolving spatial distribution of fire
temperatures within the tunnel space has not been widely studied, which impedes a realistic quantification
of fire damage to the whole structure.
While experimental studies are necessary to capture the real response of engineered systems, they are costly
in terms of both monetary value and time to successful execution. Numerical modelling can be used to run
a series of simulations with a range of input parameters, the results of which can be used for making design
recommendations and assisting with the decision-making process. Computational fluid dynamics (CFD)
modelling, though costly, provides a feasible approach to obtain both temporal and spatial fire temperatures
within the tunnel and to predict traveling fire scenarios when multiple railway cars are involved in a fire.
In this paper, a series of railway tunnel fire scenarios, considering the uncertainties in the heat release rate,
ventilation velocity, tunnel slope, ignition point, and ignition criteria for fire spread, are modelled using
Fire Dynamics Simulator (FDS), a CFD software package developed by the National Institute of Standards
and Technology [7]. The modelling approaches have been verified using the experimental data from the SP
small-scale test and the MTFVT program [8].

The cross-section of the modelled tunnel in FDS is shown in Figure 1(b), where the curved ceiling is
constructed using rectangular obstructions. Thermal properties of the tunnel lining were defined in the FDS
model to capture the influence of thermal conduction of tunnel lining on the surface gas temperature. The
thermal conductivity of concrete is taken as 1 W/m-K, specific heat as 1000 J/Kg-K, and density as 2400
kg/m3. The model includes 260 m of the tunnel in length. The ambient temperature and background pressure
are set as 20 °C and 101325 Pa in the simulations.

Figure 1. (a) Cross-sectional geometry of the prototype tunnel, and (b) modeled tunnel lining in FDS
Tunnel slope affects the buoyancy force of the smoke flow and back-layering length, and consequently
alters the fire demand by influencing ventilation [9, 10]. Three different tunnel slopes are considered in this
study to investigate the influence of slope on fire gas temperature: upward 2.2%, downward 2.2%, and zero.
The selected values cover the range of practical and allowable slopes: 2.2% slope corresponds to the
steepest section of the Cascade Tunnel in Washington State, which serves both passenger and freight trains.

Figure 2. Randomly generated HRR curves based on available experimental data

Heat Release Rate (HRR) describes the rate of energy released during a fire and reflects the magnitude and
development of a fire. For train fires, the HRR largely depends on the combustion load of carriages while
it can also be influenced by the tunnel geometry, ventilation conditions, and fire propagation. A variety of
train body types, geometry of windows and openings, and the amount of combustible material in the interior
introduces a large uncertainty in determining the combustion load of a carriage. To capture the uncertainties
in HRR and establish the HRR demands for passenger train fires, 10 HRR curves are generated using Latin
Hypercube Sampling (LHS) [8], based on the distributions of available experimental data [6, 11]. The ten
generated HRR demands, as Figure 2 shows, will be used as inputs to the FDS parametric study to
investigate a series of potential fire scenarios.

Ventilation of a tunnel can be achieved by either natural or mechanical (forced) means. Mechanical
ventilation systems are generally classified as longitudinal or transverse based on the direction of airflow
in the traffic space. Longitudinal systems ensure a longitudinal flow along the axis of the tunnel with fans
at portals or distributed along the tunnel, whereas transverse systems use air ducts to circulate polluted and
fresh air perpendicular to the tunnel axis [12, 13]. Practically, longitudinal ventilation systems are easier to
adopt and cheaper compared to transverse systems [14]. This research only investigates tunnel fires under
longitudinal ventilation.
NFPA 502 [15] requires emergency ventilation in tunnels with lengths exceeding 1000 m. For life safety
concerns in tunnels operating with passenger trains, the ventilation system should be designed to create an
air velocity exceeding the critical velocity (Vc). The critical velocity is defined as “the minimum steady-
state velocity of the ventilation airflow moving towards the fire, within a tunnel or passageway, that is
required to prevent back-layering at the fire site” [15]. The critical ventilation velocity (Vc) has been widely
studied in terms of influencing factors (e.g. HRR, tunnel height, slope, etc.) and predictive models [16-21].
The critical velocity (Vc) of 2.65 m/s is calculated for the tunnel under study using Li’s equation [16],
assuming HRR curves of Figure 2 (limiting to HRR values larger than 20 MW) and tunnel slope of zero.
In practice, fans may be controlled by the tunnel operator, adding uncertainty in the process due to the
human-element. Given the lack of data to quantify the uncertainty in the ventilation velocity during a tunnel
fire, and given the variation in detection time, operator reaction, and time needed to change a fan’s operation
mode, a uniform distribution with a range of 0.8 Vc to 1.5 Vc is assumed for the ventilation velocity. The
realizations of ventilation velocity are generated using LHS, as 2.45 m/s, 3.08 m/s and 3.70 m/s, based on
a uniform distribution between 2.1 m/s (0.8 Vc) and 4.0 m/s (1.5 Vc). The air flow is assumed to blow at a
constant velocity from the left entrance of the tunnel, while the right portal is set to be open to ambient air.

The size of a railcar influences the distance between the fire source and the ceiling, resulting in different
blockage ratios. Considering that passenger railcar dimensions are relatively standard, the dimension of an
Amtrak “Superliner” railcar [22] was adopted in this study for the train model, which is 25.9 meters long,
4.9 meters tall, and 3.1 meters wide. Five identical railcars are aligned longitudinally inside the tunnel to
study the fire spread pattern (Figure 3). The distance between two adjacent railcars is 1 m. The distance
from the left portal of the tunnel to the first train carriage is 50 m. The surface material of the train cars is
assumed to be stainless steel, with a thermal conductivity of 15 W/m-K, specific heat of 500 J/Kg-K, and
density of 7800 kg/m3. The applied HRR values are normalized by the surface areas of each railcar (except
for the bottom surface), and the evolution of HRR over time is applied using piecewise linear functions.

Figure 3. Schematic graph of the FDS model

When a fire occurs in a tunnel, the fire spread between railcars will exacerbate the fire size and the potential
damage. Extreme fire spread has been observed in almost all of the major tunnel fires, however, research
on tunnel fire spread is quite limited. Characterize the ignition condition for a real passenger carriage, due
to the varying ignition sources, paths and train materials, is a complex process. Andreini et al. [23] provided
ignition temperatures of the main components of a typical Italian passenger rail carriage, obtained from a
series of cone calorimeter tests. The test result shows that the ignition temperatures of the seat, floor curtain,
headrest, wall and ceiling material of the carriage ranges from 352-462 °C. In this study, a train carriage is
considered as a lumped mass and modelled as an object with bulk thermal properties and an ignition
temperature. When the surface of an object reaches the ignition temperature, it will ignite and follow one
of the HRR evolution curves as defined earlier. To cover the potential range and investigate the influence
of ignition conditions, three values of 300 °C, 400 °C, and 500 °C are selected as ignition temperatures.
The ignition temperature is also assigned to the surfaces of train carriages. The fire can start from either the
first or third car from the north portal to capture the fire initiation from one end and the middle of the train.
In summary, a total of 540 simulations, considering 10 HRR demands, three ventilation velocities, three
tunnel slopes, three ignition temperatures, and two fire initiation locations are conducted in FDS. For each
simulation, the fire temperature-time histories were recorded at 1) different cross-sections (10m interval)
to reflect temperature distribution along tunnel length, and 2) three different heights (6.4 m, 3.9 m and 1.3
m) at each cross-section to account for potential non-uniform fire damage within a cross-section.
2.2 Spalling incorporated heat-transfer analysis
The second step for damage assessment is to determine the temperatures within the tunnel lining sections.
The damage level of a certain tunnel location can be evaluated based on the maximum temperature reached
in a fire event. In case of a tunnel fire, different locations of the lining would be exposed to different fire
temperatures, and hence undergo varied levels of damages. To investigate the total damage of the 260m
length of the investigated tunnel, the fire temperature-time histories measured across the tunnel (from FDS
simulations) are transferred as input data to SAFIR heat transfer analysis [24].

Heat-induced spalling is a major and common issue for fire performance of concrete structures [25]. This
phenomenon is still not well-understood because of the complex mechanisms and large uncertainties
involved. The research focusing on the subject is mostly experimental and has shown that factors such as
the geometry of concrete elements, fire duration and temperature, boundary and restraint conditions, and
the externally applied loads could influence spalling (more details and references are provided in [26]).
Different spalling prediction models have been developed by studying macro- and micro- behaviours of
concrete [27-29], yet they are unable to accurately predict spalling for a given concrete mix under certain
mechanical and heating protocols, and cannot be applied to mainstream finite element software packages.
To incorporate spalling in the heat transfer analysis of SAFIR, a simplified spalling model is proposed. This
paper considers that, for a given location of tunnel lining, concrete spalling is a gradual and uniform process
with a constant spalling rate (i.e., describes spalling depth vs. time, with a unit of mm/min). Thus, a spalling
process can be characterized using three factors: spalling start time (tstart), spalling rate (Rs), and the end
time (tend) of the process. The determination of tstart, Rs and tend is based on the collected data from published
experiments. While the large-scale experiments provide the most reliable data on potential spalling of
tunnel segments, the number of investigated specimens is quite limited. This paper also investigates small-
and intermediate- scale tests to expand the data pool. Figure 4 shows the spalling process from the collected
data, where the tests using hydrocarbon fire curves (HC or RWS) are shown in red, and those using ISO
834 fire curve (with lower heating rate) are shown in black. The details of collected data and references are
provided in Hua et al. [26]. Figure 4 shows that specimens tested under Hydrocarbon (HC) or RWS fire
curves start spalling within the first five minutes of the fire because of the high heating rate. These
specimens also have a higher spalling rate (2-5 mm/min). Specimens under less-intense fire conditions
(e.g., ISO 834) mostly start to spall after 5-10 minutes and have a relatively mild spalling rate (1 to 3
mm/min).
 Figure 4. Experimental data for concrete spalling [26]

Figure 5. Probability density function of spalling start temperature [26]

This paper characterizes the spalling starting time as a function of gas temperature. That is, when the gas
temperature in proximity of the tunnel lining reaches the spalling starting temperature, the concrete lining
will begin to spall. Figure 5 shows the distribution of the spalling start temperatures calculated by on the
spalling start time and fire curves reported from the published experiments. Most test specimens start to
spall when fire temperatures reach 600-800 °C. For a practical implementation and a starting point, this
paper takes the spalling start temperature as 740 °C and the spalling rate as 2.6 mm/min, which are the
mean values of the collected data. However, data that could be used to determine the endpoint of the spalling
process is quite limited, as many experiments were terminated shortly after the first indication of spalling
(Figure 4). For modeling purposes, this paper assumes that spalling ceases when the first reinforcement
layer is exposed. The assumption can be adjusted by integrating additional future experimental results.

To incorporate the described fire spalling model within the thermo-mechanical finite element analysis, a
fine mesh has been created in SAFIR, for which the concrete cover of the reinforced concrete section is
composed of relatively fine layers of concrete fibres to enable proper removal of concrete when spalling
occurs. A program has been coded in MATLAB to control and update the input files of SAFIR over the
duration of fire as spalling occurs. The code automatically generates input files for the SAFIR thermal
analysis, decides whether spalling occurs, removes concrete layers based on the spalling rate, updates the
input file (and finite element mesh) with the removed nodes and elements, restarts SAFIR to continue the
thermal analysis, and finally reads output files from SAFIR for damage quantification.
In this paper, damage is defined as the volume of concrete that requires repair and replacement. While there
is no consensus on how to quantify fire damage to a tunnel lining, similar statements can be found in
existing standards where 300°C is recommended as a threshold for discolouring and reduced residual
strength of concrete [30-32]. Hence, the damaged volume of concrete is calculated as the volume of
concrete that experiences temperatures above 300 °C. This temperature threshold is also consistent with
existing experiments and observations, showing permanent material degradation above 300 °C [33, 34].
2.3 Coupled thermal-structural analysis
The third step in the framework is to investigate the structural performance of the tunnel under elevated
temperatures. This section focuses on the SAFIR thermo-mechanical analysis of a 2D model of the cross-
section of the prototype tunnel. The tunnel lining is constructed using 26 beam elements, and the earth
resistance is modelled using 25 compression-only spring elements, as Figure 6(a) shows. The tunnel base
is located 11 m below the grade and is modelled as fixed-fixed. One of the most intense temperature profiles
from the 540 FDS simulations is selected to study stability of the structure. If structural stability of the
tunnel is verified for an extreme fire scenario, it can be assumed that the tunnel structure will not collapse
in less severe cases. As Figure 6(b) shows, the tunnel section is divided into three regions and assigned with
different temperature boundaries to account for the variance in gas temperature across the height.

Figure 6. (a) Beam spring model of the tunnel for structural analysis, and (b) an extreme fire scenario
As for the material properties, at ambient temperature, the compressive strength of concrete is 45 MPa and
the tensile strength is assumed to be zero. The tensile yield strength of the reinforcement is 414 MPa. The
temperature-dependent material properties from EN 1992-1-2 [35] are applied to both concrete and steel.
The applied loads on the tunnel include self-weight, vertical earth pressure, horizontal earth pressure, and
surcharge. The stiffness of the compression-only springs is calculated based on Eq. 1, as a function of tunnel
radius R, modulus of elasticity E, and Poisson’s ratio 𝜈 of the surrounding soil [36]. Assuming a clay soil,
E and 𝜈 are taken as 30 MPa and 0.4, respectively.
𝑘 = ∙ (1)
The structural analysis is performed considering “with spalling” and “no spalling” cases, to obtain lower
and upper bound responses. The spalled layers cannot be removed from the finite element mesh during the
structural analysis in SAFIR. Thus, the spalling-induced section losses within the structural model is
represented by assigning a fibre temperature of 1200 °C to spalled layers – the temperature at which
concrete loses all material strength and stiffness [36] – to effectively remove the material from the analysis.

3 RESULTS

3.1 Fire scenarios

The results from the 540 FDS simulations of this study are organized and presented in two parts. The first
part focuses on capturing the control fire demand and providing design recommendations. The second part
investigates the spatial temperature distribution along the length and across sections of the prototype tunnel
for fire damage assessment.
This section investigates the temperature-time curve at the critical section of the prototype tunnel, where
the critical section is defined as the tunnel section that experiences the highest temperatures. Based on the
distributions of the maximum temperature (Tmax) data, the 540 fire scenarios have been divided into three
categories as a function of HRR to observe clear trends: low intensity fires (HRRmax ≤ 30MW), intermediate
intensity fires (30MW < HRRmax ≤ 40MW) and high intensity fires (HRRmax > 40MW). The mean of Tmax data
of low and high intensity fires are 245°C and 1007°C, respectively. The distribution of Tmax data and the
Weibull fits of low and high intensity fires are shown in Figure 7. While the distribution of Tmax data of low
and high intensity fires show distinct peaks, no clear trends are observed in the intermediate intensity fires
(not shown here), as the temperatures tend to be controlled by other parameters, such as ventilation velocity.

Figure 7. Tmax data and the Weibull fits of low intensity and high intensity fires
Table 1 provides the quantiles of Tmax for low and high intensity fires based on the Weibull distribution fits
(values are within ±10% of the generated data). More results of this study, such as the data on fire growing
rate and duration, can be found in Hua et al. [8].The study provides insight on the distribution of expected
fire temperature-time scenarios and demonstrates a clear trend among the proposed low and high intensity
fires. For design purposes, the intermediate-intensity fires with HRR values between 30 and 40 MW can
be conservatively assumed as high intensity fires.

Table 1. Quantiles of Tmax for low and high intensity fires

max (°C) for max (°C) for
Quantile
low intensity fires high intensity fires
80th 333 1125
85th 356 1145
90th 384 1170
95th 426 1204

The spatial and temporal distributions along the length and across sections of the prototype tunnel are also
studied. This part focuses on the unique features of the tunnel fire scenario and provides a basis to quantify
fire damage to the whole tunnel structure. The results are directly transferred as the input to the heat-transfer
analysis in SAFIR. Figure 8 shows the boxplots of peak temperatures across the tunnel length. The spatial
distribution pattern of the modelled tunnel fire scenarios can be summarized as follows: (1) fire
temperatures in low-intensity fires are mostly below 300 °C, while temperatures in high-intensity fires can
be as high as 1200 °C; (2) when studying temperatures at different heights, the tunnel ceiling (H = 6.4m) is
expected to have the most fire damage; and (3) upstream of the fire (x ≤ 0) remains at ambient temperature
in almost all the fire scenarios. Overall, the critical section with the highest temperatures, considering the
potential for fire spread, is 120 m downstream of the fire ignition point (x = 120 m).
 Figure 8. Boxplots of peak temperatures at different locations for (a) low- and (b) high- intensity fires

3.2 Damaged volume of concrete

This section studies the volume of concrete loss for the 540 fire scenarios based on the results of heat-
transfer analyses using SAFIR and incorporating spalling. The section losses can originate from two
criteria: spalling induced concrete losses and heat-induced losses where concrete temperature exceeds 300
°C, as discussed in Section 2.2. From the analysis, all the low intensity fires cause no damage to concrete
lining. Figure 9(a) shows the distribution of damaged volume of concrete for high intensity fires, and Figure
9(b) shows damage when not considering spalling.

Figure 9. Damaged volume of concrete (a) considering spalling and (b) not considering spalling
The results show that extensive repair work is needed after high intensity fires. The normalized damaged
volume of concrete for the worst-case scenario (226 m3), with respect to the total volume of concrete within
the affected 260 m of the tunnel section, is 16%. The volume of damaged concrete provides an
understanding of how much material should be replaced after a fire, as well as a reference for predicting
the potential labour cost, downtime, and economic losses from service interruptions. Comparing the results
with and without spalling shows that occurrence of spalling significantly increases damage. The observation
emphasizes the necessity of incorporating spalling into fire damage analyses for more realistic assessment.
3.3 Structural stability
The considered extreme fire scenario does not result in a structural failure of the tunnel, as the load is
redistributed within the cross section to the cooler side of the concrete lining. The arch shape of the tunnel
section transfers loads in compression and prevents a tensile failure of the structure. Displacements over
time for two locations have been recorded to understand the deformation of the tunnel structure under fire:
one is the vertical displacement of the crown (node 14 marked in Figure 6), and the other is the horizontal
displacement of the spring-line (node 6 marked in Figure 6). Figure 10 shows that the structure moves
upwards at the crown and bends inward at the spring-line during the heating phase. The delayed response
and propagation of heat within the section due to low thermal conductivity of concrete should be noted.
The FDS fire scenario in Figure 6(b) reaches peak temperatures at approximately 40 minutes, while the
maximum deflection occurs after 1 hour, during the cooling phase. Understanding the delayed response of
the concrete structure is significant when considering life safety of emergency responders.

Figure 10. Displacements at (a) spring-line and (b) crown under an extreme fire scenario

4 CONCLUSIONS
This paper provides a framework to quantify fire damage to concrete tunnel linings considering the
uncertainties of fire scenarios, with a spalling model incorporated. The fire scenarios were modelled using
FDS including five random variables: the heat release rate (HRR), ventilation, tunnel slope, starting ignition
point among the railcars, and fire spread criteria between the railcars. The spatial and temporal distributions
of fire temperature within the tunnel were recorded. Damage was defined as any concrete exceeding a
temperature of 300 °C. Spalling-incorporated heat transfer analyses were conducted to calculate the
damaged volume of concrete. Finally, the stability of the tunnel was evaluated based on structural analyses
of the tunnel section subjected to extreme fire scenarios.
Based on the results of the sample study conducted on a passenger rail tunnel, the maximum damaged
volume of concrete was calculated as 226 m3, meaning that 37 concrete trucks are needed to repair the
damaged sections of the tunnel. Structural analysis of the tunnel section confirmed stability of the tunnel
structure under an extreme fire scenario, as was also observed in historical events, where collapse has rarely
occurred. However, it was noted that the maximum demand within the tunnel section was not observed
until the fire entered into the cooling phase, due to the low thermal conductivity of concrete. The proposed
framework in this paper can be integrated within probabilistic risk-assessment methods to guide the design
of fire protection for concrete linings. The first step is to quantify the probability of a high intensity fire
based on the type of railcars passing through the tunnel. Given said probability (i.e., frequency of the fire
scenario and its intensity), concrete damage (as discussed in this paper) and consequently the corresponding
downtime of the tunnel can be evaluated. Combining the probability of hazard with the associated
consequences (i.e., losses) provides an assessment of risk. The decision to incorporate fire protection (if
any) and the corresponding design should be based on the comparison of the calculated risk versus the
acceptable risk that is agreed upon with the stakeholders.
ACKNOWLEDGMENT

The authors gratefully acknowledge the CAIT Region 2 UTC Consortium and the Institute of Bridge
Engineering at the University at Buffalo for their generous support. The views and conclusions contained
in this document are those of the authors and should not be interpreted as representing the opinions or
policies of the CAIT Region 2 UTC Consortium. Mention of trade names or commercial products does not
constitute their endorsement by the CAIT Region 2 UTC Consortium.

REFERENCES
1. Leitner, A., The fire catastrophe in the Tauern Tunnel: experience and conclusions for the Austrian guidelines.
Tunnelling and Underground Space Technology, 2001. 16(2001): p. 217-223.
2. McGrattan, K. and Hamins, A., Numerical simulation of the Howard Street Tunnel fire. Fire Technology,
2006. 42(4): p. 273-281.
3. EFNARC, Specification and guidelines for testing of passive fire protection for concrete tunnels linings. 2006,
EFNARC.
4. Meng, Q. and Qu, X., A probabilistic quantitative risk assessment model for fire in road tunnels with parameter
uncertainty. Int. J. of Reliability and Safety, 2011. 5: p. 285-298.
5. Cheong, M.K., Spearpoint, M., and Fleischmann, C., Design fires for vehicles in road tunnels. 2008.
6. Li, Y. and Ingason, H., A new methodology of design fires for train carriages based on exponential curve
method. Fire Technology, 2015. 52.
7. McGrattan, K., Hostikka, S., McDermott, R., Floyd, J., Weinschenk, C., and Overholt, K., Fire dynamics
simulator user’s guide. NIST special publication, 2013. 1019(6).
8. Hua, N., Tessari, A., and Elhami-Khorasani, N., Quantifying uncertainties in the temperature-time evolution
of railway tunnel fires. Fire Technology, 2020.
9. Atkinson, G.T. and Wu, Y., Smoke control in sloping tunnels. Fire safety journal, 1996. 27(4): p. 335-341.
10. Chow, W., Wong, K., and Chung, W., Longitudinal ventilation for smoke control in a tilted tunnel by scale
modeling. Tunnelling and Underground Space Technology, 2010. 25(2): p. 122-128.
11. Lattimer, B.Y. and McKinnon, M., A review of fire growth and fully developed fires in railcars. Fire and
Materials, 2018. 42(6): p. 603-619.
12. Fire and smoke control in road tunnels. 1999, PIARC Committee on Road Tunnels: Paris, France.
13. Road tunnels: operational strategies for emergency ventilation. 2011, PIARC Committee on Road Tunnels:
Paris, France.
14. Design fire characteristics for road tunnels 2017, PIARC Committee on Road Tunnels: Paris, France.
15. NFPA, NFPA 502: Standard for road, tunnels, bridges, and other limited access highways 2017.
16. Li, Y.Z. and Ingason, H., The maximum ceiling gas temperature in a large tunnel fire. Fire Safety Journal,
2012. 48: p. 38-48.
17. Lee, S.R. and Ryou, H.S., An experimental study of the effect of the aspect ratio on the critical velocity in
longitudinal ventilation tunnel fires. Journal of Fire Sciences, 2005. 23(2): p. 119-138.
18. Lee, Y. and Tsai, K., Effect of vehicular blockage on critical ventilation velocity and tunnel fire behavior in
longitudinally ventilated tunnels. Fire Safety Journal, 2012. 53: p. 35-42.
19. Tanaka, F., Takezawa, K., Hashimoto, Y., and Moinuddin, K.A., Critical velocity and backlayering distance
in tunnel fires with longitudinal ventilation taking thermal properties of wall materials into consideration.
Tunnelling and Underground Space Technology, 2018. 75: p. 36-42.
20. Weng, M., Lu, X., Liu, F., Shi, X., and Yu, L., Prediction of backlayering length and critical velocity in metro
tunnel fires. Tunnelling and Underground Space Technology, 2015. 47: p. 64-72.
21. Wu, Y. and Bakar, M.A., Control of smoke flow in tunnel fires using longitudinal ventilation systems–a study
of the critical velocity. Fire Safety Journal, 2000. 35(4): p. 363-390.
22. Superliner(railcar).Access date: Dec. 2018; Available from: https://en.wikipedia.org/wiki/Superliner_(railcar).
23. Andreini, A., Da Soghe, R., Facchini, B., and Giusti, A. Fire scenarios modelling for the safe design of a
passenger rail carriage. in 9th World Congress of Railway Research. 2011. Lille, France.
 24. Franssen, J.-M. and Gernay, T., Modeling structures in fire with SAFIR®: Theoretical background and
capabilities. Journal of Structural Fire Engineering, 2017. 8.
25. Bisby, L., Mostafaei, H., and Pimienta, P., White paper on fire resistance of concrete structures. 2014, US
Department of Commerce, National Institute of Standards and Technology.
26. Hua, N., Tessari, A., and Elhami-Khorasani, N., Characterizing damage to a concrete liner during a tunnel fire.
Submitted to Tunnelling and Underground Space Technology, in revision.
27. Gawin, D., Pesavento, F., and Schrefler, B., Towards prediction of the thermal spalling risk through a multi-
phase porous media model of concrete. Computer methods in applied mechanics and engineering, 2006.
195(41-43): p. 5707-5729.
28. Ožbolt, J. and Bošnjak, J. Modelling explosive spalling and stress induced thermal strains of HPC exposed to
high temperature. in MATEC Web of Conferences. 2013. EDP Sciences.
29. Zeiml, M., Lackner, R., and Mang, H.A., Experimental insight into spalling behavior of concrete tunnel linings
under fire loading. Acta Geotechnica, 2008. 3(4): p. 295-308.
30. Concrete Society, Assessment, design and repair of fire-damaged concrete structures. 2008: Camberley, UK.
31. Federal Highway Administration, Tunnel Operations, Maintenance, Inspection, and Evaluation (TOMIE)
Manual. 2015: Washington DC, USA.
32. International Federation for Structural Concrete (fib), Fire design of concrete structures – structural behavior
and assessment. 2008: Lausanne, Switzerland.
33. Du, S., Zhang, Y., Sun, Q., Gong, W., Geng, J., and Zhang, K., Experimental study on color change and
compression strength of concrete tunnel lining in a fire. Tunnelling and Underground Space Technology, 2018.
71: p. 106-114.
34. Hager, I., Colour change in heated concrete. Fire Technology, 2013. 49.
35. European Committee, Eurocode 2: Design of concrete structures, in Part 1-2: General rules-structural fire
design. 2004.
36. Savov, K., Lackner, R., and Mang, H.A., Stability assessment of shallow tunnels subjected to fire load. Fire
safety journal, 2005. 40(8): p. 745-763.

View publication stats