Professional Documents
Culture Documents
RESEARCH REPORT
Kristina Larsson
PREFACE
This report is a literature review of the knowledge of fires in tunnels at present. It summarizes
part of the fire dynamics necessary to understand the development of fire in tunnels, state-ofthe-art of concrete research concerning fire behaviour, and tries to collect what is known
about how rock behaves at high temperatures. The report will form a basis for deciding the
direction of continued research in the field of rock behaviour at high temperatures. The report
is funded by Banverket.
ii
SUMMARY
During the past 15 years there have been several major fires in tunnels all over the world,
most of them caused by accidents. Many of these fires have occurred in road tunnels, and
many have involved heavy goods vehicles carrying goods not classified as dangerous. The
heat developed by these fires has been on par with that expected from a fire in a fuel tanker,
and the damage to the tunnel structure has in some cases been severe. Sweden has been spared
disasters so far, but since many tunnels are being built as parts of major rail and road routes
around and under cities, more transportation will take place underground so the risks are
increasing.
A brief investigation of Swedish railroad tunnels has shown that reinforcement is used in
almost all tunnels; only 6 % does not have any reinforcement at all. Selective reinforcement
(the tunnel carries all load and is only reinforced with spot bolting and/or shotcrete in a few
places) is used in 38 % of the tunnels. In 61 % of the tunnels interacting reinforcement is
used, which refers to systematic bolting, mesh and shotcrete that together with the rock mass
carries the load. Concrete constructions that carries load (supporting reinforcement) is used in
26 % of the tunnels, and then usually only on short sections. In Sweden most tunnels are
constructed in relatively strong rock, and at shallow depths, which makes concrete lining
excessive. It is therefore interesting to study the behaviour of strong, hard rocks subjected to
high temperatures from a tunnel fire.
The aim of this literature review was to find the state of the art of how rock behaves at high
temperatures caused by a fire, but very little information was found. Some large scale fire
tests have been performed in rock tunnels, but only a few of the reports from these tests
mention anything at all about what happened to the rock during the test. In a few cases it has
been noted that rock fell from the roof during and after the fire, but no investigation of the
cause of these rock falls has been made. During most of the large scale fire tests performed in
rock tunnels, the rock has been protected from the heat to prevent fall-outs. It seems no tests
have been performed to check the actual behaviour of rock during a fire.
Two areas of interest for future research can be noted. The behaviour of some typical Swedish
rock types found in tunnels should be investigated with focus on the behaviour at high
temperatures. For concrete there is a maximum temperature defined that the concrete can be
exposed to, to avoid spalling. A similar type of maximum temperature limit should if possible
be defined for different rock types. The other area of interest is the behaviour of typical rock
reinforcement used in Swedish tunnels. This reinforcement is usually a combination of rock
bolts and shotcrete, but the amount of reinforcement used varies. The effect of rock bolts must
be investigated, since they are mainly made of steel, which transfers heat very well. The bolts
iii
will conduct heat into the rock mass, forming local hot spots at some depth. The bolts will
also loose the majority of their strength at temperatures of about 300-400C, which means
that they will no longer carry load. The effect of this must be considered and the risk
estimated for the people escaping the tunnel, and the fire fighters entering the tunnel. It has
been shown by several authors that shotcrete can be used for fire protection of concrete. It
should be investigated if shotcrete can fill the same function for rock, and still have the
reinforcing capabilities required, and if so what the minimum thickness is to prevent spalling
of rock.
iv
SAMMANFATTNING
Under de senaste 15 ren har det intrffat flera svra brnder som en fljd av olyckor i tunnlar
ver hela vrlden. Mnga av dessa brnder har intrffat i vgtunnlar, och flera har omfattat
lngtradare med en last som inte klassificerats som farligt gods. Vrmeutvecklingen i dessa
brnder har varit av den storleksordning som man frvntar sig frn en brand i en
brnsletanker, och i vissa fall har det blivit omfattande skador p tunneln. Sverige har hittills
klarat sig undan strre brnder, men med den utbyggnad av infrastruktur i form av vg- och
jrnvgstunnlar runt och under strre stder och den kning av transporter under jord som
fljer, s kar riskerna fr att ngot ska intrffa.
En underskning av svenska jrnvgstunnlar i berg har visat att frstrkning av ngot slag
anvnds i nstan alla tunnlar, bara 6 % av tunnlarna r ofrstrkta. Selektiv frstrkning
anvnds i 38 % av tunnlarna och innebr att tunneln br all last och att bultning och/eller
sprutbetong anvnds p korta strckor. Samverkande frstrkning anvnds i 61 % av
tunnlarna, vilket innebr att bergmassan tillsammans med frstrkning i form av systematisk
bultning, nt och sprutbetong br lasten. Betongkonstruktioner som br last (brande
frstrkning) anvnds i 26 % av tunnlarna, och d vanligtvis bara p kortare strckor (t.ex. i
portaler). I Sverige r de flesta tunnlar byggda i relativt h llfast berg och ganska ytligt, vilket
innebr att betonglining r verfldig. Det r drfr intressant att studera hur hghllfast berg
beter sig vid de hga temperaturer som kan uppkomma vid en tunnelbrand.
Mlet med denna litteraturstudie var att inventera kunskapsnivn om bergs beteende vid hga
temperaturer (vid brand), men endast begrnsad information har hittats. Ngra storskaliga
brandfrsk har utfrts i bergtunnlar, men enbart ngra f av rapporterna frn dessa tester
nmner ngot alls om vad som hnde med berget under frsken. I ngra frsk har man
noterat att det fll ned berg frn taket under och efter branden, men man har inte underskt
orsakerna till nedfallen. Under de storskaliga brandfrsken har man skyddat berget mot
hettan fr att frhindra utfall. Det verkar inte som att ngra frsk alls har utfrts fr att
underska hur berget verkligen beter sig under en brand
Tv omrden har bedmts som intressanta fr vidare forskning. Man br bestmma
egenskaper och beteende fr ngra av de typiska svenska bergarter man kan frvnta sig i
tunnlar med avseende p hga temperaturer. Liksom fr betong s br man ta reda p den
maximala temperatur olika bergarter kan utsttas fr utan att spjlkning initieras, om en sdan
temperatur existerar. Det andra forskningsomrdet r hur typisk bergfrstrkning i en svensk
tunnel beter sig vid hga temperaturer. Frstrkningen r ofta en kombination av bergbultar
och sprutbetong, men mngden frstrkning varierar. Bergbultars pverkan mste underskas
eftersom de tillverkas av stl som r en vldigt god vrmeledare. Bultarna kommer att leda in
v
vrme i bergmassan s att man fr lokala vrmekllor p djupet. Vid temperaturer p ca 300400C har bultarna frlorat det mesta av sin hllfasthet, och kan allts inte lngre ta ngon
last. Vad detta innebr fr tunnelns stabilitet och vilken risk det innebr fr mnniskor som
ska utrymma tunneln i hndelse av brand och fr rddningspersonalen br underskas. Flera
forskare har visat att sprutbetong kan anvndas som brandskydd fr betong. Det br
underskas om sprutbetong kan fylla samma funktion fr berg och nd ha den frstrkande
effekt man efterstrvar, och i sfall vilken minsta tjocklek man mste ha fr att frhindra
spjlkning.
vi
TABLE OF CONTENTS
Preface ......................................................................................................................................... i
Summary.................................................................................................................................... iii
Sammanfattning.......................................................................................................................... v
Table of contents ...................................................................................................................... vii
1
Introduction ........................................................................................................................ 1
1.1
Background ................................................................................................................. 1
1.2
Swedish rock tunnels .................................................................................................. 1
1.3
Aim ............................................................................................................................. 2
1.4
Limitations .................................................................................................................. 3
2
3
4
5
6
7
8
Fires .................................................................................................................................... 5
2.1
General ....................................................................................................................... 5
2.2
Smoke stratification in tunnels ................................................................................... 6
2.3
Flame length............................................................................................................... 8
2.4
Large fires in tunnels with high longitudinal air flow.............................................. 10
2.5
Fire spread in tunnels ................................................................................................ 11
2.6
Vehicle fire load ....................................................................................................... 11
2.7
Heat release rate (HRR)............................................................................................ 12
2.8
Case studies of tunnel fires ....................................................................................... 13
2.8.1
Channel tunnel.................................................................................................. 14
2.8.2
Mont Blanc tunnel............................................................................................ 14
2.8.3
Tauern tunnel.................................................................................................... 15
2.8.4
Hong Kong ....................................................................................................... 16
2.8.5
St Gotthard Tunnel ........................................................................................... 17
2.8.6
Kaprun tunnel................................................................................................... 17
High temperature properties of rock................................................................................. 19
Concrete............................................................................................................................ 25
4.1
Effects of high temperatures..................................................................................... 25
4.2
Failure modes ........................................................................................................... 26
4.3
Prevention of spalling ............................................................................................... 28
4.4
Test methods............................................................................................................. 30
4.5
Shotcrete as fire protection....................................................................................... 30
4.5.1
Shotcrete on concrete ....................................................................................... 30
4.5.2
Shotcrete on rock .............................................................................................. 31
High temperature properties of steel ................................................................................ 33
Fire tests............................................................................................................................ 35
6.1
Fire resistance tests ................................................................................................... 35
6.2
Fire behaviour tests................................................................................................... 37
6.3
Full scale tests........................................................................................................... 38
6.3.1
Repparfjord tunnel............................................................................................ 39
6.3.2
Lappeenranta tunnel ......................................................................................... 39
6.3.3
Runehamar tunnel............................................................................................. 40
6.3.4
Blasted rock tunnel, Sweden ............................................................................ 42
Swedish tunnel specifications ........................................................................................... 45
Discussion and conclusions .............................................................................................. 47
vii
9
10
viii
1 INTRODUCTION
1.1 Background
Fire has been used by mankind for thousands of years, and its ability to break rocks is well
known and was used early for tunnelling and mining. The method of using fire to break rock
is called fire-setting and works as follows; a fire was started at the wall, and when the rock
was hot the embers were removed and water was thrown onto the wall, which cracked the
rock. Wedges were driven into the cracks and blocks of rock could be removed with an ironbar. The Egyptians built a tunnel in this way as early as 2000 BC (www.pbs.org, 2005). The
method was used in Europe up to the 18th century, when blasting with black powder took over
as the main mining method.
During the past 15 years there have been several major fires in tunnels all over the world.
Some have been caused by accidents, and some have been arsons. Many of these fires have
occurred in road tunnels, and many have involved heavy goods vehicles carrying goods not
classified as dangerous. These fires have turned into disasters and have killed several people.
The heat developed by these fires has been on par with that expected from a fire in a fueltanker, and the damage to the tunnel structure has in some cases been severe. Sweden has
been spared disasters so far, but since many tunnels are being built as parts of major rail and
road routes around and under cities, more transportation will take place underground so the
risks are increasing. The major part of European road and rail tunnels has concrete lining,
partly because of the weak rock under high pressure they are constructed in. In Sweden most
tunnels are constructed in relatively strong rock, and at shallow depths, which makes lining
excessive. Instead many of our tunnels are reinforced by a layer of shotcrete, which prevents
rocks from falling into the tunnels and also prevents water leakage into the tunnel.
After the tunnel fires in Europe several research projects have studied the behaviour of
concrete lining during and after a fire, and much have been learned on how to make the
concrete perform better. Addition of polypropylene fibres is one of the most efficient ways of
improving the spalling properties of the concrete, protecting the surface with fibre reinforced
(polypropylene) shotcrete is another.
and only 6 % have concrete portal constructions longer than 50 m. Reinforcement is used in
almost all tunnels, only 6 % does not have any reinforcement at all, and 38 % has selective
reinforcement. Selective reinforcement means that the rock mass around the tunnel carries all
the load and that the tunnel is only reinforced with spot bolting and/or shotcrete in a few
places. In 61 % of the tunnels interacting reinforcement is used, which refers to systematic
bolting, mesh and shotcrete that together with the rock mass carries the load. Reinforced
concrete constructions that carries load (supporting reinforcement) is used in 26 % of the
tunnels, and then usually only over short sections. Over 50 % of the tunnels are grouted, but
38 % are not, and for the remaining tunnels no information on grouting was found. The whole
list of tunnels can be viewed in Appendix 1.
Table 1-1. Reinforcement statistics for rock tunnels in the Swedish railnet.
Longest concrete constr [m]
0
1-9
10-49 > 50
no of
tunnels
No reinf. /
selective
Interacting
reinf
Supporting
reinf
Grouting
22
62
20
7 / 47
76
32
66 / 48
18 %
50 %
16 %
6%
6 % / 38 %
61 %
26 %
53 % / 38 %
The Swedish road administration (Vgverket) operates 28 rock tunnels on the national road
network (Freiholtz, 2005). Only two of these have concrete lining, and the rest are reinforced
with bolts and shotcrete.
1.3 Aim
Little research has been made on the behaviour of rock during and after a fire. There are many
tunnels in Sweden that are unreinforced or that have only bolts or plain shotcrete as
reinforcement. What would happen in these tunnels during a fire? How much work would it
require afterwards to make them safe for traffic again? This depends on how the rock and
shotcrete reacts to the high temperatures developed by a fire. The behaviour of shotcrete has
been studied in several European research projects, and it has been stated that polypropylene
reinforced shotcrete is a cheap and easy way of improving the fire resistance of concrete. The
behaviour of rock at these high temperatures does not seem to have been studied at all,
probably because it has not been considered necessary since the rock in European tunnels is
always protected by thick layers of concrete. The aim of this literature review is to summarize
the knowledge of the effects of fires in rock tunnels. Current status of research regarding
concrete fire behaviour is summarized. The main focus is road and rail tunnels. Basic fire
dynamics in enclosures and tunnels is also covered, but not in great detail. Cases of recent
tunnel fires are summarized to provide a background for the theory of fire dynamics and to
show which parameters are important to know about a tunnel fire. Some of the large scale
tunnel fire tests performed during the past 10 years are described to see what has been studied
and learned from them.
1.4 Limitations
The limitations of the literature review are that fluid dynamics of a fire, smoke flow, and
chemistry are not covered. Numerical simulations of fires are not covered either, since these
are mostly concerned with smoke propagation and the development of exact temperature
gradients of the fire gases.
The focus of the review is rock tunnels and the use of shotcrete for fire protection. This
literature review is not a complete state-of-the-art document of tunnel fire research and
knowledge, but it should provide a basis for understanding fire theory, summarize the most
important results of concrete research regarding fire behaviour, and should provide a starting
point for further research on the behaviour of rock during a tunnel fire.
Many Swedish rock tunnels are grouted to prevent water leakage. The effect of fire on the
grout is not covered in this literature review, and neither are the effects on frost insulation
used in tunnels today. Other materials for fire protection than shotcrete have not been
reviewed.
2 FIRES
2.1 General
A fire is a manifestation of a chemical reaction, but the mode of burning may depend more on
the physical state and distribution of the fuel, as well as its environment, than its chemical
nature (Drysdale, 1999). The understanding of fire behaviour in general requires knowledge
of chemistry, heat transfer, fluid dynamics, etc., and the behaviour of a fire in a tunnel is even
more complex. When comparing a tunnel fire to a fire in the open air, there are two important
differences (Ingason, 2005); the heat reflection to the burning vehicles is more efficient in a
tunnel fire because of the confinement, and the interaction of the ventilation with the growing
fire. The heat reflection causes an increase in intensity of the burning vehicle, and may
increase the heat release rate by up to a factor of three (Carvel et al., 2004), see section 2.7.
The interaction of ventilation and fire generates aerodynamic disturbances in the air flow
through the tunnel, which may cause changes in the ventilation pattern, such as throttling of
airflow, and reverse flow of hot gases and smoke from the fire into the ventilation stream
(backlayering) (Ingason, 2005). These effects complicate the fire-fighting, and also transport
toxic fumes and gases far away from the fire.
Ingason (2005) also compared tunnel fires to compartment fires (rooms in buildings) and
stated three major differences. First, the maximum heat release rate (HRR) of a compartment
fire depends on the natural ventilation, which is determined by the area and height of the
openings into the compartment. In tunnels the natural ventilation depends on the fire size,
slope of the tunnel, cross-sectional area, length of tunnel, type of tunnel (concrete lined, rock),
and meteorological conditions at the entrance to the tunnel. Tunnels often also have forced
longitudinal ventilation, which has an effect on the combustion efficiency as well (Ingason,
2005). Second ly, compartment fires can grow to flash-over within a few minutes, but this is
unlikely to happen in a tunnel fire because of great heat losses to the surrounding walls, and a
lack of containment of the hot fire gases. Inside a truck cabin or train compartment located
inside a tunnel, however, flash-over can occur. Thirdly, in the early stages of compartment
fires an upper layer of buoyant smoke is formed, with a cold smoke-free layer below. If there
is very low longitudinal ventilation in a tunnel the same type of smoke layer can be formed in
the early stages of a fire. Further away from the fire source, however, the smoke will descend
to the floor. The distance at which this occurs depends on the fire size, tunnel type,
circumference and height of the tunnel cross-section. If the ventilation is increased, the
stratification of the smoke will dissolve, and backlayering is formed on the upstream side of
the fire, and the stratification of the smoke downstream of the fire is determined by the heat
losses to the surrounding walls, and by the turbulent mixing of the buoyant smoke layer and
the opposite moving cold air below.
There are two modes of combustion that are important to know and distinguish between
(Ingason, 2005), i.e., fuel-controlled and ventilation-controlled fire. Fuel-controlled fire
means that oxygen is in unlimited supply, and that the rate of combustion is independent of
the mass flow rate of air (oxygen supply rate), but is instead determined by the mass flow rate
of vaporised fuel (fuel supply rate). A ventilation-controlled fire has a limited oxygen supply,
and the combustion rate depends on both air and fuel supply rates. At the precise oxygen level
that enables complete combustion, the mixture is said to be stoichiometric. To determine
whether a fire is fuel- or ventilation-controlled, the air-to- fuel equivalence ratio ()
m&
= a ,
(2-1)
rm& f
can be used, where m& a is the mass flow rate of air (oxygen) supply, m& f is the fuel mass loss
rate (fuel supply), and r is the stoichiometric coefficient for complete combustion. If 1,
the fire is fuel-controlled, and if < 1, the fire is ventilation-controlled.
the third region (Fr > 10) there is little or no vertical temperature gradient, and consequently
little or no stratification.
h free = 0.2Q 2 / 5 ,
(2-3)
where hfree is the flame height, and Q is the heat release rate, and is valid for axisymmetric
fires. In tunnels there are two factors that must be considered to determine the flame length:
the presence of the ceiling, and the ventilation. When a non-combustible ceiling is present the
horizontal extension of the flames, hhor, can be related to the flame cut-off height, hcut, see
Figure 2-3.
Figure 2-3. Horizontal length of flame s under a ceiling, from Ingason (2005).
Babrauskas (1980) calculated the ratio hhor/ hcut for an unbounded horizontal ceiling and for a
corridor. The ratio was 1.5 for the unbounded ceiling. For the corridor the ratio was found to
be highly dependent on the width of the corridor, for a corridor of 3 m width the ratio was
1.81 and for a corridor width of 2 m, the ratio was 2.94. Assuming that a tunnel behaves
similarly to a corridor, Ingason (2005) states that the flames would extend along the ceiling to
a horizontal length of 1.5 3 times the cut-off height. The effect of the ventilation is not taken
into account in these calculations (Ingason, 2005).
Depending on the velocity of the air flow the effect on the flame length differs. High
longitudinal velocity creates a better mix of the oxygen supply with the fuel supply, and
thereby increases the efficiency of the combustion, which can lead to a shortening of the
flames. At moderate air velocities, an increase in the air velocity may cause a lengthening of
the flames, because at low velocities the volatiles must spread over a larger area before there
is sufficient oxygen to allow complete combustion (Ingason, 2005). At velocities slightly
below moderate, the flames will again grow shorter and interact to a larger degree with the
tunnel ceiling, which decreases the inflow of oxygen to the fire source. At air velocities lower
than moderate, the flames will become less horizontal and will interact with the ceiling to a
greater degree, which reduces the flow of oxygen into the fire source. At zero forced
ventilatio n (natural convection only) the flames will hit the ceiling and spread in both
directions.
Figure 2-4. Burning zones in a tunnel with high longitudinal ventilation, from Ingason (2005).
The burning process itself can be seen as stationary, but to explain the process, five different
zones are assumed in Figure 2-4;
1)
2)
3)
4)
5)
The zones move forward in a dynamic manner, provided that the vehicle density is high
enough in the area of the initial fire. In the burnt-out zone, the vehicles have been completely
consumed by the fire and the fire gases have already cooled down. In the glowing ember
zone, the vehicles have stopped burning and are in a late stage of the decay phase, and are
literally a pile of glowing embers. In the combustion zone (x = 0 in Figure 2-4) the vehicle
fire is fully developed, and flaming combustion is taking place in the whole zone. The flames
cause high heat transfer rates from the gas to the fuel, which leads to high fuel vaporisation
10
rates (Ingason, 2005). The temperature of the gas phase just beyond x = 0 increases rapidly,
and reaches a maximum at x = x1 , where the oxygen level reaches a minimum. The excess
fuel zone starts at x = x1 , at which the oxygen is depleted. In this zone, fuel vaporises from the
vehicles, but no combustion takes place due to a lack of oxygen. The fuel vaporisation takes
place up to the point where the temperature drops below the pyrolysis temperature, (Tvap
300C for most solid materials). Beyond this point (at x = x2 ) no vaporisation takes place, but
the gas flows on and loses its heat to the tunnel walls and preheats the vehicles within this
zone (preheating zone) (Ingason, 2005).
Remarks
Passenger car
3 3.9
2.5 5.0
Bus
Truck load
41
65
20
20 30
88
30
Railway car
Tanker with 50 m3
gasoline
41 77
12.5 22
1500
300
11
Figure 2-5. Variation of the enhancement coefficient with fire dimension, from Carvel et al.
(2004).
12
HRR is difficult to estimate for a fire test in a tunnel even with a full set of measuring
equipment, and for a real fire it is even more difficult to estimate when the fire has already
occurred. To address this problem Carvel et al. (2004) presented an empirical relationship
(Eq. 2-5)
W
= 24 F
WT
+ 1
(2-5)
where WF is the width of the fire object and WT is the width of the tunnel. The relationship is
valid for tunnels with flat roofs and all fuels (petrol, heptane, cars, etc.) except methanol. It is
possible that Eq. 2 underestimates the enhancement factor for a concave roof by up to 10 %,
since a concave roof is thought to give a lens-effect, thus concentrating the heat back toward
the source (Carvel et al., 2004).
Carvel and Beard (2005) states that the HRR is the most important factor contributing to the
severity of a fire. The higher the HRR the more severe the fire.
Country
Type of tunnel
Length
Fatalities
Year
Daegu metro
Gleinalm
Kaprun
St Gotthard
Tauern
Salerno
Mont Blanc
Gueizhou tunnel
Channel Tunnel
Isola Delle
Pfnder
Baku underground
Great Belt
Serra Ripoli
South Korea
Austria
Austria
Switzerland
Austria
Italy
France / Italy
China
UK / France
Italy
Austria
Azerbaijan
Denmark
Italy
metro
road
funicular
road
road
rail
road
rail
rail
road
road
metro
during construction
road
8.3 km
3.3 km
16.3 km
6.4 km
9 km
11.6 km
0.8 km
49.6 km
0.15 km
6.7 km
8 km
0.44 km
~ 200
5
155
11
12
4
40
> 80
5
3
289
4
2003
2001
2001
2001
1999
1999
1999
1998
1996
1996
1995
1995
1994
1993
13
Country
Moscow
New York
London
Gumefens
Caldecott
London
Kajiwara
Nihonzaka
Velsen
Russia
USA
United Kingdom
Switzerland
USA
United Kingdom
Japan
Japan
Netherlands
Type of tunnel
Length
Fatalities
Year
metro
metro
metro
road
road
metro
road
road
road
0.34 m
1.02 km
0.74 km
2.04 km
0.77 km
7
2
31
2
7
1
1
7
5
1991
1990
1987
1987
1982
1981
1980
1979
1978
eight HGVs and several passenger cars entered the tunnel before the entrance was closed, but
since they could see the smoke coming from the burning vehicle they could stop their vehicles
and escape. All passenger cars turned and drove out, while the eight HGVs were left behind
and later burned. During the same time 18 HGVs, 9 passenger cars, a motorcyclist and a pickup van entered the tunnel from the French side. Four of the HGVs ma naged to pass the
burning HGV and make it to safety, but all the others were trapped in the smoke, and later
burned. Totally 39 people died in the accident. The fire lasted for 53 hours. The tunnel vault
was severely damaged over a length of 900 m, and there was also damage to the roadway
pavement and slab. The secondary lining was also destroyed or badly damaged over a
considerable distance because of the high temperatures (LaCroix, 2001). The temperatures in
the tunnel during the fire have been estimated to between 800 and 1000C, with a maximum
temperature of 1200C (Abraham and Derobrt, 2003). The average concrete facing thickness
was 0.5 m, with no reinforcement and water tightness. Spalling does not seem to have
occurred, despite the high temperatures, which Abraham assumes can be explained by the
high permeability of the concrete (bad quality). The high permeability allowed water to
escape the concrete, which kept the stresses low enough for thermal spalling not to occur
(Abraham and Derobrt, 2003). The damage to the tunnel lining is instead caused by the
decreased strength of the concrete due to the high temperatures.
under the first lorry, and the other two smaller cars were pushed up against the tunnel wall.
The ramming lorry ended up with its cab right up against the back of the first lorry. In the
collision itself eight people were killed, and another four died as a result of deciding to stay in
their car during the ensuing fire (Eberl, 2001).
stop
stop
Figure 2-7. Schematic of accident causing the fire in the Tauern Tunnel.
The fire lasted for about 17 hours. During the extinction phase, engineers entered the tunnel
ahead of the fire fighters to ensure the stability of the roof. The tunnel (inner) ceiling sagged
in places and had to be propped. Soot cleaning was necessary in the entire tunnel, and about
350 m of inner ceiling had to be replaced (Eberl, 2001). The sidewalls over a length of 100 m
in the area of highest temperatures had spalled to a depth of 400 mm, and to a depth of 50 mm
over the whole sidewall surface or locally along another 450 m length. Totally some 600 m3
of spalled concrete was removed (Leitner, 2001). About 800 m of the roadway had to be
rehabilitated due to spalling. Here 5-8 cm of the surface was removed and a new pavement
added. The rescue niches in the tunnel worked as planned during the fire (radio and lighting).
The tunnel was closed to traffic for three months, during which also the ventilation system
was improved.
2.8.4 Hong Kong
The tunnel is a cross-harbour tunnel that was opened for traffic in 1972. It consists of two
double tubes, with two lanes each, having a roadway width of 6.6 m and a height of 5.1 m
(Chow and Li, 2001). The fire accident occurred in May 2000, and was caused by a passenger
car that caught fire. The driver tried to extinguish the fire himself, but failed. After about four
minutes firemen were on the scene, and started to evacuate people into the other tunnel tube.
After evacuation the fire was rapidly extinguished. The fire lasted for about 45 min, and 15
min after that the tunne l was reopened for traffic. No damage to the ceiling was reported, and
no remedial measures beyond soot cleaning were necessary.
16
17
18
Physical properties
Thermal properties
Youngs modulus
Poissons ratio
Tensile strength
Compressive strength
Viscosity
Thermal expansion
Density
Permeability
Melting temperature
Heat of fusion
Specific heat
Thermal conductivity
Thermal diffusivity
The most important mechanical properties to consider in regard to fires are Youngs modulus,
compressive strength and thermal expansion. Youngs modulus and compressive strength are
affected by both temperature and pressure. The variation of the normalized Youngs modulus
with temperature is shown in Figure 3-1. Increasing the surrounding pressure tends to delay
the decay of the normalized modulus when the temperature increases (Heuze, 1983).
Figure 3-1. Normalized Youngs modulus versus temperature, from Heuze (1983).
19
The compressive strength decreases with increasing temperature, as shown in Figure 3-2. At a
surrounding pressure of 500 MPa, the difference in strength is 700 MPa between samples at
300C and 900C, respectively.
Figure 3-2. Variation of compressive strength with temperature, from Heuze (1983).
Heating of rock leads to thermal expansion, which can be reversible or irreversible depending
on the heating rate and the maximum temperature. Richter and Simmons (1974) studied the
thermal expansion behaviour of igneous rocks, and found that for heating rates less than
2C/min, and a maximum temperature of 250C, the expansion curves are reproducible and
no permanent strains are induced. The thermal expansion of a rock type depends on its
constituent minerals. Quartz-rich rocks have a larger thermal expans ion than rocks with low
quartz content. The thermal expansions of some minerals are shown in Figure 3-3.
20
Figure 3-3. Thermal expansion of typical minerals found in granitic rocktypes, from Simmons
and Cooper (1978). The minerals are a) quartz, b) olivine, c) pyroxene, d) orthoclase, and e)
plagioclase.
A schematic of the variation of the linear thermal expansion with pressure and temperature is
shown in Figure 3-4. Increasing the pressure increase the --transition temperature, and
lowers the peak thermal expansion (Heuze, 1983). The --transition of quartz is an
important event, that occurs at a temperature of 573C. At this temperature quartz changes
from one form to another with an accompanying increase in volume. The physical properties
are not dependent on temperature in the short time interval considered here, i.e., the duration
of a fire.
Figure 3-4. Coefficient of thermal expansion versus temperature, from Heuze (1983).
21
Dry granite melts at about 1050C, but partially or fully saturated rock melts at lower
temperatures. The specific heat of granite is discontinuous around the --transition, but the
general trend is that the specific heat increases with increasing temperature (Heuze, 1983), see
Figure 3-5. The thermal conductivity decreases with increasing temperature, see Figure 3-6, in
the interval 20 to 300C. The thermal diffusivity also decreases with increasing temperature,
and shows a minimum at the --transition, see Figure 3-7 (Heuze, 1983).
22
23
24
4 CONCRETE
This chapter explains the behaviour of concrete when it is subjected to high temperature due
to fires. Some of the topics are the effects of raised temperatures, typical failure modes, and
how spalling can be prevented. Shotcrete is also covered as a means of protecting the
structural concrete from heat.
Hydrothermal reactions
Loss
of chemically
water starts.
Calcium
hydroxide bound
dissociates
Triple
point
of
water
Marked increase in creep
--inversion
quartz =>
expansion of aggregates
Dissociation ofofcalcium
carbonate
Ceramic binding
Total
loss
of water of hydration
Melting
starts
Concrete melted
The compressive strength of concrete decreases when the temperature is increased from room
temperature to about 80C, however, this decrease is largely reversible upon cooling. Above
300C most concretes show a strength reduction, how large depends on the aggregates and
cement paste used. Above temperatures of 550-600C the concrete has lost enough strength so
as not to be structurally useful (Khoury, 2000). In a fire, usually only the first few centimetres
closest to the surface are subjected to temperatures above 330C due to the low thermal
diffusivity of the concrete. This layer provides insulation to the inner concrete and the
reinforcing steel, although its structural strength is greatly diminished. This layer is usually
replaced after the fire.
25
Bostrm (2003) noted that the results from fire tests on concrete varied greatly with regard to
spalling. Several tests on three different kinds of concrete (normal, high strength, and self
compacting concrete) were performed to find the relation between spalling and water content,
water-powder ratio, and also to see the effect of addition of polypropylene fibres. The results
of testing of concrete columns showed a reduction of spalling when polypropylene fibres were
used, but a higher proportion of fibre did not give a better performance (Bostrm, 2003). The
optimum performance and proportion of fibres seemed to vary with concrete quality and
possibly with curing (in air or under water). The results from testing of slabs show a clear
reduction of spalling when polypropylene fibres were added. The conclusions from the tests
were that self compacting concrete spall much more than normal concrete with the same w/cratio, but that addition of polypropylene fibres or some kind of insulation reduces spalling
significantly.
Aggregate
Corner
Surface
Explosive
7-30
splitting
popping
superficial
30-90
non-violent
none
can be serious
7-30
violent
cracking
can be serious
7-30
violent
loud bang
Serious
26
Table 4-3. The types of spalling and their influencing factors, modified after Khoury (2000).
Aggregate
Corner
Surface
Explosive
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
Pore pressure spalling is influenced by the permeability of the concrete, the initial water
saturation level, and the heating rate (Khoury, 2000). Thermal stress spalling is explained by
the fact that heating of a material with low conductivity (like concrete or some ceramics)
creates temperature gradients that induce compressive stresses close to the heated surface and
tensile stresses in the cooler regions. The compressive stresses at the surface can be enhanced
by load. The theory behind the combined pore pressure thermal stress is that explosive
spalling is a result of pore pressure, compression in the region closest to the surface (caused
by thermal stresses and external loading), and internal cracking, see Figure 4-1. During
heating, cracks form in the material as the sum of the stresses exceeds the tensile strength of
the material. The formation of cracks is accompanied by a sudden release of energy and the
sudden failure of the heated surface material.
Tests performed by Jansson (2005) have shown that the pore pressures are not very high at the
onset of spalling. A fire test of a concrete panel showed that when spalling initiated 15 min
into the test, the temperature at 10 mm depth was about 200C and the pressure in the
concrete about 0.7 MPa. This pressure is lower than the tensile strength of concrete at the
measured temperature, and indicates that pore pressures probably are less important than
thermal stresses as a cause of spalling (Jansson, 2005).
27
Load, L, and
thermal stress, t
Pore pressure, p
Spall
Pores
Concrete
L + t
Figure 4-1. Forces in heated concrete, after Khoury (2000).
28
Table 4-4. Methods for preventing spalling of concrete, modified after Khoury (2000).
Method
Effectiveness
Polypropylene fibres
Air-entraining agent
Very effective
Thermal barrier
Very effective
Moisture content
control
Compressive stress
control
Comments
Choice of aggregate
Reinforcement
Supplementary
reinforcement
Choice of section
shape
After a fire incident the damage to the concrete can be assessed by concrete petrography
(Nijland and Larbi, 2001). High temperatures results in changes of some phases of both the
cement paste and the aggregates which may alter the colour of the concrete and the original
mineralogical composition of both the cement paste and the aggregates. These changes can be
used to find isograds (horizons of similar composition and/or appearance) in the concrete.
Since these isograds depend on the temperature they will coincide with isotherms and can
therefore be used to trace the temperature variations with depth from the surface (Nijland and
Larbi, 2001). Concrete petrography can be done on three levels of increasing accuracy: visual
and stereomicroscopic inspection, fluorescent macroscopic analysis (FMA), and polarising
and fluorescent microscopy (PFM). The first is performed on drilled cores, FMA on flatpolished sections, and PFM on fluorescent thin sections. The visual inspection is aimed at
finding the colour variations with depth from the surface, patterns of cracking in and around
the aggregates, width and depth of cracks, dissolution and loss of bonding to the aggregate
particles, and the integrity of the cement paste (Nijland and Larbi, 2001). The isograds that
can be identified by the naked eye together with the corresponding temperatures are listed in
Table 4-5.
29
Table 4-5. Isograds in concrete with corresponding temperatures, after Nijland and Larbi
(2001).
Temperature (C)
< 300
300-350
573
> 800
Comment
Normal, no apparent macroscopic changes in concrete. The colour remains grey.
Oxidation of iron hydroxides like FeOOH in aggregate and cement paste to hematite,
-Fe2 O3 , causing a permanent change of colour of the concrete from grey to pinkish
brown.
Transition of -quartz to -quartz, accompanied by an instantaneous increase in
volume of about 5 %, resulting in a radial cracking pattern around the quartz grains in
the aggregate. This phase transition itself is reversible, but the radial cracking
provides a diagnostic feature that remains after cooling.
Complete disintegration of calcareous constituents of the aggregate and cement paste
due to both dissociation and extreme thermal stresses, causing a whitish grey
coloration of the concrete.
30
shotcrete. The samples with shotcrete thickness of 60 mm had a maximum spalling depth of
about 40 mm, i.e., spalling of the concrete had started (Bostrm, 2005).
4.5.2 Shotcrete on rock
Mangs and Keski-Rahkonen (1990) tested granite slabs protected by shotcrete according to
the ISO834 fire curve. The slabs had an area of about 1 m2 , and a thickness of 20 - 30 cm. The
shotcrete was applied in different thicknesses and with different reinforcement. In about half
of the tests the shotcrete was anchored to the rock slabs with bolts, which in some cases were
insulated, as indicated in Table 4-6. A summary of the composition of the samples is shown in
Table 4-6.
Table 4-6. Composition of samples, from Mangs and Keski- Rahkonen (1990).
Thickness
[mm]
dry mix
dry mix
3
4
wet mix
wet mix
Anchored
20 30
60 80
60 100
50 110
+, insulated
+, insulated
wet mix
80 100
6
7
wet mix
wet mix
75 115
20 60
+, not insulated
-, sample edges fixed to furnace
wet mix
40 - 120
+, not insulated
Changes of the shotcrete surface were observed visually (Mangs and Keski-Rahkonen, 1990).
The duration of the tests varied from 15 min to 4 hours. A summary of the results of the tests
and the observations made is shown in Table 4-7.
Table 4-7. Summary of shotcrete tests, from Mangs and Keski- Rahkonen (1990).
Sample Location Duration of test Test results
1
ceiling
16 min
wall
3h
ceiling
4h
wall
4h
wall
2 h 35 min
ceiling
53 min
ceiling
2 h 35 min
wall
55 min
31
Temperatures in the samples were measured at the shotcrete-rock interface and at halfthickness of the shotcrete layer. A summary of the temperatures for the different locations for
each sample are shown in Table 4-8. In some cases spalling of the shotcrete in the vicinity of
the thermocouples give anomalous temperature readings, these have been discarded in Table
4-8.
Table 4-8. Summary of temperature measurements by Mangs and Keski-Rahkonen (1990).
Sample
16 min
80
100
Temperature at 11 min.
3h
410-460
650-680
4h
360-440
420-470
4h
300-540
520-700
2 h 35 min
240-290
270-420
53 min
80-100
140-220
7
8
2 h 35 min
55 min
650-870
90-110
720-900
110-150
The temperature on the outside of the granite slabs (at the centre) was measured on samples 3
and 4. After 2 hours the temperature was about 30C, after 3 hours about 50C, and after 4
hours about 60C (Mangs and Keski- Rahkonen, 1990). No note of the exact thickness of these
slabs can be found in the article.
Mangs and Keski-Rahkonen (1990) summarized their observations from the tests:
- bond breakage occurred in all tests (both with and without mechanical anchoring),
- slight spalling of the shotcrete caused by water pressure was observed,
- when the shotcrete layer fell off, the hot fire gases caused a thermal shock to the granite
which started spalling, and
- large thermal stresses broke the shotcrete la yer and the granite slab in sample 7.
As a conclusion Mangs and Keski-Rahkonen (1990) say that mechanical anchoring of the
shotcrete to the rock is important, since bond breakage occurred in all samples, but that the
steel bolts have to be insulated from the heat.
32
33
34
6 FIRE TESTS
There are generally two types of fire tests, which will both be described in this chapter. The
first type is performed to determine the fire resistance of construction components such as
beams with regard to spalling, load carrying capacity and ability to function as a barrier. This
type of fire test is described in chapter 6.1. The second type of fire tests are performed to test
the behaviour of a fire with regard to different tunnel cross-sections, and can be performed on
different scales. These types of tests are described in chapter 6.2. Full scale tests performed
during the 90s and beginning of the 21st century are described in chapter 6.3. For more
details concerning tunnel fire experiments from the 1960s until present day, the reader is
referred to Carvel and Marlair (2005).
35
The importance of including the cooling-off phase was noted by Wetzig (2001). Heating of a
concrete sample up to 1600C for two hours produced no cracking or collapse, but 30 minutes
after the end of the test the sample exploded. Wetzig (2001) noted that as the sample cooled
glazing of the surface took place, thus preventing water vapour from escaping, which led to
pressure build- up and finally the explosive destruction of the sample.
1400
Furnace temperature [ C]
1200
1000
RWS
800
HC
ISO 834
EBA
600
RABT
400
200
0
0
20
40
60
80
100
120
140
160
180
Time [min]
Figure 6-1. European fire curves, after Khoury (2000) and Carvel (2005).
In 2004 concrete panels of four different qualities were fire tested at SP (Wickstrm, 2004). A
summary of the different qualities is shown in Table 6-1. A total of 16 samples were tested,
samples 1 to 10 were tested without load, and samples 11 to 16 were tested with a load of 6.25
MN, which gives a theoretical compressive stress of 9.6 MPa. The samples were tested
according to the RWS fire curve, see Figure 6-1
Table 6-1. Sample numbering and concrete qualities, after Wickstrm (2004).
Concrete
Sample
K65, no fibres
1, 2, 9 and 10
3, 4, 11 and 12
5, 6, 13 and 14
7, 8, 15 and 16
36
The unloaded samples (1 to 8) were tested first. It was noted that extensive spalling started
after only 2 minutes in samples 1 and 2 and after 3 minutes also sample 3 starts to spall
extensively. Spalling of sample 3 stopped after about 11 minutes. Spalling of samples 1 and 2
stopped after 27 minutes. After the test was concluded glazing of the surface as well as
formation of craters (diameter 5 30 mm) and stalactites (maximum diameter 10 mm and
maximum length 50 mm) was observed on all samples (Wickstrm, 2004). Craters and
stalactites were more common on the samples that did not spall.
The loaded samples (11 to 16) and samples 9 and 10 were tested together. It was noted that
extensive spalling started after only 1.5 minutes in samples 9 and 10 and after about 2 minutes
light spalling starts in sample 11. After about 2.5 minutes also samples 13 and 16 started to
spall lightly. After 3 minutes there is extensive spalling of sample 11. Spalling of sample 11
stopped after about 12 minutes. Spalling of samples 9 and 10 stopped after 23 minutes. After
the test was concluded glazing of the surface as well as formation of craters (diameter 5 30
mm) and stalactites (maximum diameter 10 mm and maximum length 50 mm) was observed
on all samples (Wickstrm, 2004). Craters and stalactites were more common on the samples
that did not spall.
The conclusion from these tests is that K65 concrete without fibres and with 1.0 kg/m3 of
polypropylene fibres spall more easily than K65 concrete with 1.5 kg/m3 of polypropylene
fibres. The influence of loading is uncertain since only one loaded sample (no. 11) spalled,
but the loaded sample showed spalling to a greater depth when compared with the unloaded
sample of the same quality. The samples with concrete K45 showed no spalling when
unloaded, and only light spalling (not measureable) with the load applied.
number, the Reynolds number, the Richardson number, and the Grashof number. In the ideal
case all of these numbers should be the same for the scale model and the real case. Typically
only conservation of the Froude number (see Eq. 6-1) is considered, since it is impossible to
conserve all the numbers (Carvel and Marlair, 2005). The Froude number is proportional to
HRR2 /L5 , where L is a characteristic dimension (usually the tunnel height) and is formally
defined as
Fr =
U2
,
gL
(6-1)
where U is the velocity of the gases, g is the acceleration due to gravity and L is a
characteristic dimension of the system (Ingason, 2005).
Reduced scale tests have been used to, e.g., test sprinkler systems, provide data for
comparison with numerical models, test conditions in refuges during a fire, and to investigate
how under-ventilation affects the HRR of a fire (Carvel and Marlair, 2005).
Laboratory scale tests should ideally be used as a complement to full scale tests and numerical
simulations, since although they may give information on the behaviour of larger-scale fires,
the similarity to full scale fires is not necessarily good (Carvel and Marlair, 2005). The Froude
scaling criterion may not apply to fires in tunnels smaller than about 1 m in diameter, and
small scale pool fires behave very differently from full scale pools. The size of these small
scale tunnels range from 0.1 0.3 m high, 0.2 0.5 m wide, and the length varies between 1
m and about 30 m. These small tunnels have been used to test smoke flow, to find the critical
ventilation velocity, and to provide information for comparison with numerical simulation
results.
38
which ran between 1985 and 1995. The project was a joint research project between nine
European countries, among them Sweden, Norway and Finland. The project examined fires in
transport tunnels, and the research was based on full scale tests (Firetun, 1995). The
objectives of the project were to provide information on:
- fire phenomena,
- escape, rescue and fire extinguishing possibilities,
- the effect of the surrounding structural parts on the fire,
- reusing the structure,
- accumulation of theory on fire behavior, and
- formation, distribution and precipitation of the contaminants.
Not all of the objectives were realized in the project, but the results provide an important
foundation of data for future research.
The Runehamar Tunnel tests were performed by SP (Swedish National Testing and Research
Institute), and were aimed at obtaining knowledge of fire development in different HGV
cargos in a tunnel with forced longitudinal ventilation (Ingason and Lnnermark, 2005).
In 1997 the Swedish Defence Research Establishment (FOA) performed a fire test in a blasted
rock tunnel. The aim of the test was to determine the effect of ventilation on the heat release
rate and to find correlations between optical density and gas concentration at different
locations in the tunnel (Ingason et al., 1997).
6.3.1 Repparfjord tunnel
The test tunnel is located in a mine in Repparfjord, Norway, about 45 km south-east of
Hammerfest (Firetun, 1995). The tunnel is 2.3 km long, 5.3 - 7.0 m wide, and 4.8 - 6.0 m
high. The cross-section is horseshoe shaped, but with irregular sides and a slightly flattened
roof. At one end of the tunnel there is a shaft, about 3 m in diameter and 90 m high. The walls
and roof of the tunnel around the fire location was reinforced with 15 cm and 20 cm of steelfibre reinforced light-weight shotcrete, respectively. The fibre-reinforcement worked well
since no spalling was noted after the tests (Firetun, 1995), despite maximum temperatures of
1000C close to the roof in several of the tests.
6.3.2 Lappeenranta tunnel
The tunnel in which the tests were conducted is situated in Lappeenranta, Finland, in a
limestone quarry (Keski- Rahkonen et.al., 1986). The tunnel is 140 m long, 5.5 - 6.1 m wide
and 4.5 - 5.0 m high. The rock mass is to 90 % made up of calcite. Two tests were performed
in the tunnel, the first to simulate a fire in a subway train, and the other was a simulated small
39
car multiple fire. In the first test measurement of the temperature in the centre of the ceiling
failed because falling rocks destroyed the thermocouple (or cable), but halfway between midceiling and abutment the measured maximum temperature was 210C, and the maximum
temperature at the wall was 100C (Keski-Rahkonen et.al, 1986). The heat release rate was
about 1.8 MW. In this test the rock was unprotected and large amounts of rock fell down.
Already 25 minutes from ignition spalling from the roof started, spalling occurred over nearly
the whole length of the tunnel. Closest to the fire location the damage was worst, with damage
to a depth of about 20 cm in the roof and 10 cm in the walls. The largest single block weighed
about 400 kg.
The second test consisted of 8 wood piles where two piles were ignited at the start of the test
to see if the fire would spread to the other piles (Keski-Rahkonen et. al., 1986). Fire spread
did not occur, so piles 7 and 8 had to be ignited separately. During the first part of the test the
measured maximum air temperature in the ceiling was 800C, giving a rock surface
temperature of 450C in the ceiling and 210C in the wall. During the second part of the test
the maximum air temperature reached only about 400C. The heat release rate of the first half
of the experiment was 8 MW but only 4 MW for the second half. No explanation for this big
difference in two very similar set-ups is offered by Keski- Rahkonen et. al. (1986). In this test
some roof sections were protected by 10 cm mineral wool, which was sufficient to prevent
blocks from falling down on the weighing platform. Rock still spalled from the ceiling, but
the pieces were smaller and the total amount less. No information on support of the tunnel
was found in the report.
6.3.3 Runehamar tunnel
The Runehamar tunnel is located about 40 km south of Molde, Norway, and is a two-way
asphalt road tunnel that has been taken out of use (Lnnermark and Ingason, 2003). The
tunnel is 6 m high, 9 m wide, 1.6 km long and has a slope of 1-3 %. The rock around the
tunnel is hard gneiss (Opstad and Wighus, 2003). During the fall of 2003 a series of four fire
tests were performed in the tunnel. To protect the rock and provide a safe working
environment an inner shell of protection boards was erected over a length of 75 m. At both
ends an additional 10-25 m of the tunnel was protected by ceramic curtains to prevent hot
gases from penetrating behind the protective shell. The fire test consisted of fake HGVs with
different loads, see Table 6-2. The loads consisted of about 80 % cellulose and 20 % plastic,
which is a good representation of the load of an actual HGV (Ingason and Lnnermark, 2003).
40
Table 6-2. Description of fire loads and results from measurements (Ingason and Lnnermark,
2003).
Test Description of load
T1
T2
T3
T4
Weight
[kg]
Theoretical calorific
energy [GJ]
Measured peak
HRR [MW]
10911
240
203 (average)
6853
129
158 (average)
8500
152
125
3120
67
70
The time from ignition to peak HRR varied between 8 and 18.5 min in the four tests, and the
fire growth (between 5 and 100 MW) varied between 17 and 29 MW/min for tests T1, T2 and
T3, and for test T4 the fire growth was about 18 MW/min from HRR 5 MW to 70 MW
(Ingason and Lnnermark, 2003). The maximum air temperatures measured in the tests varied
between 1281C and 1305C. Lemaire (2003) calculated the radiation temperature on the
walls of the tunnel for each test. The radiation temperature is calculated to estimate the
thermal load on the wall, and uses the heat flux measured during the tests. The heat flux is
converted to the temperature of a black body radiating the same temperature that is received
by the wall (Lemaire, 2003). The error in the conversion is on the order of 20C to 50C. For
test T1 the thermal load is on average 900C for about 30 min, with peaks up to 1100C. For
tests T2 and T4 the average temperature is 800C for about 15 min, and in test T3 the average
temperature over 15 min is 700C.
During and after test T1 rocks fell from the roof both upstream and downstream of the fire
location (Lnnermark and Ingason, 2003). During test T2 rocks fell through the roof of
protective boards. Downstream of the fire location the rock was affected almost all the way to
the entrance of the tunnel, about 400 m away. Upstream of the fire, rocks fell over a short
length due to backlayering of hot gases from the fire. There is no information on tunnel
support reported in the articles describing the fire tests.
HRR values in technical specifications are often given as one value for a given vehicle, and
the time history of the HRR is not included. For a general case one European standard
(PIARC) gives a value of 20-30 MW, and the Swedish guideline for road tunnels (ATB
Tunnel, 2004) specifies that the ventilation should be able to control smoke from a 100 MW
fire. No specification of type of fire load is given. Ingason and Lnnermark (2005) made
comparisons to three other fire tests involving HGVs; Repparfjord, Mont Blanc, and 2nd
Benelux tunnel. A summary of the HRRs from those fire tests are shown in Table 6-3.
41
Table 6-3. Fire loads in tunne l fire tests Ingason and Lnnermark (2005).
HGV fire load
Ventilation
Repparfjord
Real HGV 87 GJ
Simulated HGV 64 GJ
Mont Blanc
Real HGV 76 GJ
2nd Benelux
Comment
Load consisted of furniture
Load consisted of wood, rubber tyres
and plastic
Fire load less than what was estimated
from the fire accident
Load consisted of wood pallets
The results from the Runehamar test series showed that the growth rate of the fire was
approximately linear for all tests between HRR 5 and 100 MW, see Figure 6-2. The maximum
HRR developed was between 66 and 202 MW, and the time from ignition to maximum HRR
was between 7 and 18.5 min (Ingason and Lnnermark, 2005).
Figure 6-2. HRR versus time from the Runehamar tunnel fire test series, from Ingason and
Lnnermark (2005).
6.3.4 Blasted rock tunnel, Sweden
In November 1997 the Swedish Defence Research Establishment (FOA) made a fire test in a
blasted rock tunnel (Ingason et al., 1997). The tunnel was 3 m wide, 3 m high, and 100 m
long. Closest to the fire source the roof was covered for 15 m by 50 mm thick light- weight
concrete slabs. The walls were protected by 100 mm of insulation for 3-4 m around the fire
source. At all other locations the walls were unprotected (Ingason et al., 1997). The testing
time was very short, only 20-25 min, and the average heat release rate was about 1.5 MW for
42
both tests. The maximum temperature 1 cm into the concrete (measured from the side closest
to the fire) was 220C in one of the tests. No remark about fallen rock is made in the report.
43
44
Figure 7-1. Fire curves specified in Swedish tunnelling specifications, from BV Tunnel
(2005).
Railway tunnels under water and under buildings should be designed for a cooling rate of
600C/h added to the fire curves in Figure 7-1, for other tunnels the load from the cooling
phase does not need to be considered (BV Tunnel, 2005). For road tunnels it must be checked
45
how the structure ais affected by a cooling phase corresponding to a cooling rate of 600C/h
(Tunnel 2004, 2004). It is stated in the technical specifications that components that are
necessary for ensuring safe escape and rescue operations must be proven to withstand fire
without local damage, such as fall outs, during the given escape and rescue time frame, which
is set to a minimum of 45 min. It is also stated that spalling of concrete can occur when the
200C isotherm has reached a depth of 2 cm (Tunnel 2004, 2004). It is noted that the moisture
content, porosity, heating rate, compressive stresses, and cross section has a large influence on
the occurrence of spalling. Measures that can prevent spalling of concrete are
- pre-investigation of the fire properties of the concrete,
- addition of polypropylene fibres, and
- using a non-combustible insulating material for insulation.
The ventilation velocity should be 3 m/s for fires with heat release rates up to 100 MW
(Tunnel 2004, 2004). If the standard fire has a larger heat release rate, special investigations
are necessary. For railway tunnels, no specification of the ventilation velocities is made (BV
Tunnel, 2005).
Doors that are a part of the escape route out of a tunnel should be dimensioned for gas
temperatures according to curve II in Figure 7-1, and with a fire duration of 60 min. Curve III
can be used for design of tunnels under water or tunnels that pass directly under a building
(BV Tunnel, 2005).
46
lining in tunnels is often protected by a layer of shotcrete. The shotcrete is used only as
insulation, and not as a load carrying reinforcement. The concrete lining can also be protected
by other means, but polypropylene-fibre reinforced shotcrete is a cheap and easily applied
protection. The polypropylene fibres melt at about 120-130C, and forms channels in the
shotcrete where the water vapour can escape, thereby reducing the pore pressure and
preventing spalling.
Spalling of rock probably occurs for slightly different reasons than spalling of concrete. First,
there will not be any great pore pressure build- up due to the presence of cracks. If the rock
surface is sealed, pore pressures can maybe build up behind the sealing. Differential
expansion of the minerals is one probable cause of rock spalling, another is that local heating
gives differential stresses due to constraints by the surrounding unheated rock. The thermal
stress caused by the expansion is added to the compressive stresses parallel to the rock
surface, which leads to a local increase of tensile stresses (small scale) in the direction
perpendicular to the rock surface. If the tensile stresses exceed the tensile strength, failure will
occur. This failure can both be violent and non-violent, but the exact reasons are not known.
Shotcrete is used for improving the fire resistance of concrete and the results from Mangs and
Keski-Rahkonen (1990) suggests that it can be used for protection of rock as well. The
shotcrete thickness used in these tests varied between 20 mm and 115 mm. The thickness at
which no spalling occurred depended on the reinforcement used, but regardless of
reinforcement 20-30 mm is too thin. The results of the tests by Mangs and Keski-Rahkonen
(1990) indicate that only the first few centimetres of rock closest to the surface will be
affected by any major temperature raise above ambient rock temperatures. The increase in
temperature, on the outer side of the samples, after four hours was only a 30C, but in these
measurements the shotcrete was 50-110 mm thick, and the thickness of the granite slab was
20-30 cm. The results of measurements with a thinner shotcrete layer would be interesting to
see.
Fire testing of concrete has been performed in laboratory scale, reduced scale and full scale
tests. The advantage of concrete is that it can be moulded into any shape and size that you
desire to test, but this cannot be made with rock. Laboratory tests on rock would have to be
performed with care, to take into account the importance of grain sizes and the fire scale.
The tunnel specifications used in Sweden today does not seem to consider rock tunnels
without concrete lining. The requirements posed for a structural component in a tunnel may
be hard to meet for a rock tunnel without reinforcement. The specifications state that the
structural component may not collapse, but there is no definition of collapse for a rock tunnel.
48
Is it when blocks fall down or when the whole roof collapses? For tunnels under water and
under buildings there is a cooling rate specified, but not for other types of tunnels. This means
that for the average railtunnel the load from the cooling phase does not need to be considered.
For road tunnels it must be checked how the construction components are affected by a
cooling phase corresponding to a cooling rate of 600C/h. It is stated in the specifications that
components that are necessary for ensuring safe escape and rescue operations must be proven
to withstand fire without local damage, such as fall outs, during the given escape and rescue
time frame, which is set to a minimum of 45 min. There is however nothing mentioned in the
specifications how this requirement should be checked for a rock tunnel with minimum
reinforcement.
49
50
51
to prevent spalling of rock? Are there any type of shotcrete that is better than others (with
regard to fibre-reinforcement etc.)?
- The effect of rock bolts must be investigated, since they are mainly made of steel, which
transfers heat very well. The bolts will also loose the majority of their strength at
temperatures of about 300-400C, which means that they will no longer carry load. The
effect of this must be considered and the risk estimated for the people escaping the tunnel,
as well as for the fire fighters entering the tunnel. Rock bolts are often cement- grouted,
how is this grout affected by the heat conducted by the bolt? Will the bolts function during
and after the fire?
Suggested methods
To answer the questions stated above, a series of tests should be performed on both laboratory
and on a larger scale. The tests should be used to find a correlation between the properties
determined in the laboratory and the damages noted in a real case. The laboratory tests should
include:
- Determination of Youngs modulus and Poissons ratio for dry and wet samples of some
typical rocks.
- Tests where part of a rock block is heated with e.g. a torch, to see differences in behavior
between:
o dry and wet samples
o loaded and unloaded samples
- Tests on jointed rock blocks.
- Tests on blocks containing natural joints, either dry or water saturated.
- Tests on rock blocks with a layer of shotcrete, where different types of shotcrete as well as
different thicknesses should be tested.
- Scale-tests of a fire in a tunnel (rock block with bore hole).
The results from the laboratory tests should be used in numerical analyses to model the rock
mass behavior when subjected to rapid temperature increase and high temperatures. Field
studies of rock tunnels subjected to fire should be made. The damages in the tunnel should be
documented and compared to the results of numerical analyses (coupled thermo- mechanical
models) to investigate the correctness of the models on a large scale
The Swedish tunnel specifications seem not to consider rock tunnels without heavy concrete
reinforcement when giving design requirements regarding fire. It should be investigated how
the requirements in the specifications compare to the real situation in the tunnels with regard
to standard reinforcement in rock tunnels.
52
10 REFERENCES
Abraham, O., Derobrt, X., 2003, Non-destructive testing of fired tunnel walls: The MontBlanc Tunnel case study. Non-Destructive Testing and Evaluation International vol 36, pp
411-418
Babrauskas, V., 1980, Flame Lengths Under Ceilings. Fire and Materials vol 4:3, pp 119126
Bostrm, L., 2003, Spalling of Tunnel Structure Linings New Swedish Results. Proc. Int.
Symp. Catatstrophic Tunnel Fires (CTF) / Bors / 20-21 November 2003, Ingason (ed.),
Bors: SP Swedish National Testing and Research Institute
Bostrm, L., 2005, Fire testing of concrete slabs isolated with shotcrete. SP Report P501398
BV Tunnel, 2005, Banverket (Swedish Railroad Administration) standard BVS 585.40 (in
Swedish)
Carvel, R., 2005, Chapter 6 in Handbook of Tunnel Fire Safety. Beard and Carvel (eds.),
London: Thomas Telford Publ. ISBN: 0-7277-3168-8
Carvel, R., Beard, A., 2005, Chapter 9 in Handbook of Tunnel Fire Safety. Beard and Carvel
(eds.), London: Thomas Telford Publ. ISBN: 0-7277-3168-8
Carvel, R. O., Beard, A. N., Jowitt, P. W., Drysdale, D. D., 2004, The Influence of Tunnel
Geometry and Ventilation on the Heat Release Rate of a Fire. Fire Technology vol 40, pp 526
Carvel, R., Marlair, G., 2005, Chapter 10 in Handbook of Tunnel Fire Safety. Beard and
Carvel (eds.), London: Thomas Telford Publ. ISBN: 0-7277-3168-8
Chow, W. K., Li, J. S. M., 2001, Case study: vehicle fire in a cross-harbour tunnel in Hong
Kong. Tunnelling and Underground Space Technology vol 16, pp 23-30
Drysdale, D., 1999, An Introduction to Fire Dynamics, 2nd edition. Chichester, UK: John
Wiley and Sons, ISBN: 0-471-97920-8
Eberl, G., 2001, The Tauern Tunnel incident what happened and what has to be learned.
Proc. 4th Int. Conf. Safety in Road and Rail Tunnels / Madrid / 2-6 April 2001, Vardy (ed.),
Dundee: University of Dundee, pp 17-28
Egger, M., 2005, Chapter 16 in Handbook of Tunnel Fire Safety. Beard and Carvel (eds.),
London: Thomas Telford Publ. ISBN: 0-7277-3168-8
Firetun, 1995, Fires in transport tunnels: Report on full-scale tests. EUREKA-Project
EU499; Firetun, Studiengesellscha ft Stahlanwendung elV. D-40213 Dusseldorf
53
54
Lundman, P., 2005, personal communication: data from database on railway tunnels (BIS) of
the Swedish Railroad Administration
Lnnermark, A., Ingason, H., 2003, Large Scale Fire Tests in the Runehamar Tunnel Gas
temperature and Radiation. Proc. Int. Symp. Catatstrophic Tunnel Fires (CTF) / Bors / 2021 November 2003, Ingason (ed.), Bors: SP Swedish National Testing and Research
Institute
Mangs, J., Keski-Rahkonen, O., 1990, Fire testing of shotcrete samples in a small furnace.
Shotcrete for Underground Support V: Proc. Engineering Foundation Conference / Uppsala
/ 3-7 June 1990, Sharp and Franzn (eds.), New York: American Society of Civil Engineers
ISBN: 0-87262-944-9
Nijland, T. G., Larbi, J. A., 2001, Unraveling the temperature distribution in fire-damaged
concrete by means of PFM microscopy: Outline of the approach and review of potentially
useful reactions. Heron vol 46:4, pp 253-264
Opstad, K., Wighus, R., Fire Suppression Systems for Road Tunnels (UPTUN). Proc. Int.
Symp. Catatstrophic Tunnel Fires (CTF) / Bors / 20-21 November 2003, Ingason (ed.),
Bors: SP Swedish National Testing and Research Institute
Palmqvist, K., 1998, Utredning avseende brandskador p berg och bergfrstrkning.
Vgverket (Swedish Road Administration), BT Brand 00.0072 (in Swedish)
Richter, D., Simmons, G., 1974, Thermal Expansion Behavior of Igneous Rocks.
International Journal of Rock Mechanics and Mining Sciences and Geomechanics Abstracts,
vol 11, pp 403-411
Schupfer, H., 2001, Fire disaster in the tunnel of the Kitzsteinhorn funicular in Kaprun on 11
Nov 2000. Proc. 4th Int. Conf. Safety in Road and Rail Tunnels / Madrid / 2-6 April 2001,
Vardy (ed.), Dundee: University of Dundee, (paper not in proceedings)
Simmons, G., Cooper, H. W., 1978, Thermal Cycling Cracks in Three Igneous Rocks.
International Journal of Rock Mechanics and Mining Sciences and Geomechanics Abstracts,
vol 15, pp 145-148
Tunnel 2004, 2004, Vgverket (Swedish Road Administration) Publ 2004:124 (in Swedish)
Turner, S., 2001, St Gotthard Tunnel Fire. New Civil Engineer, 1 November, 2001, pp 5-7
Wetzig, V., 2001, Destruction mechanisms in concrete material in case of fire, and protective
systems. Proc. 4th Int. Conf. Safety in Road and Rail Tunnels / Madrid / 2-6 April 2001,
Vardy (ed.), Dundee: University of Dundee, pp 281-290
Wickstrm, U., 2004, Fire test of concrete. SP Report P401734
55
56
Appendix 1
Tunnels of the Swedish Railroad Administration (data from BIS-database)
Tunnel
length
265
1431
586
165
1000
732
1680
298
381
332
93
17
127
135
336
236
91
845
140
955
825
3850
450
468
71
49
167
159
152
122
218
590
535
440
421
227
210
320
210
45
200
37
118
32
261
100
65
133
139
93
385
381
243
Longest concrete-constr
0
1-9
10-49 > 50
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
No reinf. /
selective
x
x
x
x
x
x
x
x
0
0
x
Interacting
reinf
x
x
x
x
x
x
x
x
x
x
x
0
0
0
x
x
x
x
x
x
x
x
x
x
x
0
0
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
Support- Grouting
ing reinf
x
x
0
0
x
x
x
x
x
0
0
0
0
x
0
x
0
x
x
x
0
x
0
x
x
x
x
x
x
x
x
0
x
0
x
0
0
x
0
x
x
0
x
0
0
x
0
x
0
0
0
0
0
0
0
0
0
0
0
0
x
x
x
x
Comment
kbgjd
kbgjd
kvrig
kbgjd
kvrig
kvrig
bg + kbgjd + kvrig
kvrig
bg + kbgjd
bg + kbgjd
Appendix 1
Tunnel
length
239
577
146
1780
45
1295
114
895
52
377
185
657
691
307
201
520
411
321
574
180
429
132
163
537
96
127
679
182
285
2056
136
70
275
187
354
199
91
242
220
78
238
107
34
448
41
33
166
201
350
1840
530
220
365
76
106
76
Longest concrete-constr
0
1-9
10-49 > 50
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
-
No reinf. /
selective
Interacting
reinf
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
-
x
x
Support- Grouting
ing reinf
x
x
x
x
x
x
x
x
x
x
x
x
x
x
0
0
0
x
x
x
x
0
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
0
0
x
0
x
x
0
0
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
-
Comment
bg + kbgjd
kbgjd
bg + kbgjd
bg + kbgjd
bg + kbgjd
bg + kbgjd
bg + kbgjd
bg + kbgjd
bg + kbgjd
Appendix 1
Tunnel
length
53
68
62
181
380
156
466
242
280
79
431
196
1380
300
610
568
Longest concrete-constr
0
1-9
10-49 > 50
x
x
x
x
x
x
x
x
x
x
x
x
x:22
-: 15
62
20
No reinf. /
selective
x
x
x
x
x
x
x
x
x
Interacting
reinf
x
x
x
x
x
x
-
Support- Grouting
ing reinf
0
x
x
x
x
x
x
0
0
0
0
x
0
0
0
x
-
Comment
bg + kbgjd
bg + kbgjd
bg + kbgjd
bg + kbgjd
kbgjd + kvrig
kbgjd
kbgjd
x: 47
76
32
x:66
-: 11
-:11
0: 7
0:48
x; affirmative, -; information not available, 0; no reinforcement or no grouting, bg; rock carries load, kbgjd;
both sections where rock carries load and where load is carried by support, kvrig; other constructions
The total number of tunnels in the table is 125. There are more tunnels in the national railroad
network, but tunnels with no information on length have not been included.
Concrete constructions:
The longest concrete construction refers to constructed tunnel portals. Of 125 tunnels there
are 22 that have no constructed portals at all, 62 with portals between 1 and 9 m long, etc.
Reinforcement:
Selective reinforcement means that single bolts have been installed to prevent block fall
outs. "Interacting reinforcement means that the rock and the installed reinforcement together
carries the loads. This type can be systematic bolting, mesh and shotcrete. Supporting
reinforcement refers to constructions designed to carry the surrounding loads, and is often
some kind of concrete construction. 47 tunnels have selective reinforcement and 7 have no
reinforcement at all. Interacting reinforcement is used in 76 tunnels, and only 36 have
supporting reinforcement. The concrete constructions often occur only on short sections of the
tunnel.
Grouting:
Grouting with either cement based or chemical grout or both is used in 66 tunnels, and 48 of
the tunnels are not grouted at all.
Comment:
If there is no note, or the notation bg in the column comment it means that the rock mass
carries itself, possibly with selective bolting or very short sections with systematic bolting,
shotcrete and other load bearing constructions. The notation kbgjd means that most of the
tunnel has load bearing support, i.e., systematic bolting, shotcrete and other load bearing
constructions or concrete constructions. The notation kvrig means constructions of other
kinds than load bearing constructions. These notations can be combined, e.g., bg + kbgjd
Appendix 1
means that there are both sections where the rock mass carries itself, and sections where the
tunnel is supported with load bearing reinforcement or concrete constructions.