Keywords

These keywords were added by machine and not by the authors. This process is experimental and the keywords may be updated as the learning algorithm improves.

Introduction

The scenario of a large vehicle fire remains one of the most troublesome issues for designers of tunnels and underground facilities. Two of the main questions which continue to be raised and debated at tunnel safety symposia are:

  • What should be used as ‘design fires’ for road, rail and subway tunnel design?

  • What fire protection and prevention measures are effective against fires in tunnels?

Most of the major experimental research projects investigating vehicle fires in tunnels carried out in the past two decades, including the EUREKA EU499 project [1], the Runehamar fire tests [2, 3] and the METRO project [4, 5], have shown that real vehicle fires in tunnels may be more severe than the scenarios which have historically been used as ‘design fires’ for vehicle tunnels, subway stations, etc. The reason is twofold;

  • The knowledge on fire development and sizes is limited due to limited number of tests leading to “expert views” on reasonable worst case fire sizes. These views tend to lower the values used in design as they usually are a product of a consensus between the experts.

  • The tests performed sometimes may exaggerate the real conditions. For example in the Runehamar tests, the fuel load was fully exposed to the wind conditions whereas in reality the load may be blocked by the truck or the front wall of the trailer resulting in a lower fire growth rate.

This chapter provides a review of the published data produced as part of these studies and identifies the characteristics common to most of these fire experiments. If design fires used for vehicle tunnels are to be in any way realistic, these characteristics must be taken into account. The chapter then looks at the question of fire protection for vehicle tunnels and discusses the current state-of-the-art.

Full Scale Fire Experiments in Tunnels

This section considers only very large vehicle (and solid fuel) fire tests carried out in full-scale tunnels, it is not intended to be a comprehensive review of all tunnel fire tests. In particular, the Memorial Tunnel Fire Ventilation Test Program (MTFVTP) results are not considered here as these tests involved fuel pools, not vehicles or realistic vehicle cargoes [6]. Similarly, smaller fire tests carried out as part of other test series will not be discussed. For a reasonably comprehensive summary of fire experiments at full and reduced scale, see Carvel and Marlair [7] or chap. 3 in Ingason et al [8]. As most design fires are specified in terms of the heat release rate (HRR) of the fire, and as this is a good measure of the severity of such fires, this review will focus on HRR data from the experiments considered. A summary with focus on HRR data from each vehicle type is given in Ingason and Lönnermark [9].

The EUREKA EU499 Fire Tests (1992)

The EUREKA EU499 ‘Fires in Transport Tunnels’ project involved researchers from nine European countries and the main data were gathered from a series of 21 fire tests which were carried out in a disused tunnel in Repparfjord in the north of Norway, near Hammerfest [1]. The tunnel has a rough-rock, irregular surface and varies between 4.8 and 6.0 m high and 5.3–7.0 m wide; the average cross-sectional area is about 34 m2; the tunnel is 2.3 km long and has a 1 % uphill gradient towards the ‘downstream’ end of the tunnel, which terminates in a shaft, open to the air. The large vehicles fire tested in this series (for which HRR data were estimated) were a subway carriage, two full railway carriages, one ‘joined’ railway carriage, one bus, and a heavy goods vehicle (HGV); a large ‘mixed load’ was also tested, which may be considered analogous to a vehicle cargo. In each case, the fire load was positioned approximately 295 m from the ‘upstream’ (downhill) end of the tunnel.

The bus fire test (B11) involved a single decker, Volvo school bus built in the early 1960s [10]. It was 12 m long and had 40 seats. During the fire test the luggage compartment door and the front door of the bus were both open, and the partition between the luggage compartment and the passenger compartment had been removed. 310 pine sticks, 800 × 40 × 40 mm, were distributed throughout the bus to represent luggage. The bus fire test was carried out without forced longitudinal ventilation. During the test, smoke from the fire travelled in the ‘upstream’ direction with a velocity of about 0.3 ms−1, while the smoke velocity in the ‘downstream’ direction was not recorded. Estimates of the HRR were made using oxygen and CO/CO2 concentration measurements taken on both sides of the fire location [11]. The fire reached a peak HRR of 29 MW within 8 min of ignition. By 18 min after ignition, the HRR had diminished to less than 13 MW, after which the fire gradually burned out over a period of more than an hour. Figure 88.1 shows a simplified representation of the measured HRR data [11].

Fig. 88.1
figure 1

Approximate HRR data from EUREKA EU499 bus fire test B11 (solid line), mixed load test LF1 (short dash) and subway carriage fire test F42 (long dash) (Adapted from [10])

Full size image

The mixed fire load test (LF1) involved a load of 2850 kg, three-fourth of which was wood, one-eight of which was tyres, and one-eight of which was other plastic materials. This was distributed in a stack approximately 2.4 m long, 2.2 m high and 2.4 m wide. No forced ventilation was used during the test and the smoke travelled in the ‘upstream’ direction until it reached about 115 m upstream of the fire, where the smoke front stopped. The smoke transport in the downstream direction was not recorded. The fire grew rapidly to a peak HRR of about 17 MW in about 15 min; the HRR data [11] are shown in Fig. 88.1.

The HGV fire test (HF1) is described in several places in the literature including [12, 13]. The fuel load in this instance consisted of a tractor unit (built by Leyland DAF in 1988) and a double axle trailer unit loaded with about 2 tonnes of wooden framed furniture. The trailer was 12.5 m long, 2.4 m wide and 2.5 m high. During the test, the HGV was positioned with the tractor unit facing upstream. A blockage of two shipping containers was positioned in the tunnel 29 m upstream of the HGV. The ventilation in the tunnel was forced longitudinally and was varied during the test. In the initial stages of the test the ventilation flow was about 6 ms−1, after the fire spread from the cab to the trailer the fan was switched off for 2 min, then restated at 2.9 ms−1 (the velocity corresponding to the intended emergency ventilation flow rate in the Channel Tunnel, which the test was designed to partially replicate). All smoke from the fire was blown downstream of the fire, except during the period that the fan was switched off. The ventilation system was switched off after about 15 min into the test while the fire was still growing, so it is not known how large the fire would have grown with a flow of 6 ms−1. Once the fans were established again at 2.9 ms−1 (after about 2 min), the fire reached a peak of 128 MW. The most reliable HRR data from this test were obtained by Grant and Drysdale and are shown in adapted form in Fig. 88.2 [12].

Fig. 88.2
figure 2

Approximate representation of HRR data from Hammerfest HGV fire test HF1 (Adapted from [11]). The HRR during the period when the fan was shut down could not be estimated, so is not presented

Full size image

A fire test was carried out using two joined half railway carriages as a fuel load. The half-carriage designated F2Al was an aluminium bodied carriage, while the half designated F7 was steel bodied. The carriages were joined by a corrugated bellows, similar to those common on railway carriages. The tunnel was initially longitudinally ventilated at about 6–8 ms−1. The first and second ignitions of the fuel load were rapidly extinguished following instrument failures. The third ignition of the fuel was carried out after adding about 360 wooden sticks to the carriages. The fire was ignited in the upstream aluminium carriage. After 32 min, the ventilation was stopped and restarted at a lower flow rate 4 min later. At approximately 48 min after ignition, the fire appears to have spread to the steel half carriage, and ‘the whole car seems to ignite immediately’ [1]. Shortly after this, the HRR peaked at about 43 MW, as shown in Fig. 88.3. The atypical fire behaviour during this test was due to the unusual test setup, ignition location and ventilation changes during the test.

Fig. 88.3
figure 3

Approximate HRR data from Hammerfest IC train test F11 (solid line) [10], ICE train carriage test FS2 (short dash) and joined half-carriage test FA3 (long dash) (Adapted from [13])

Full size image

Two fire tests were carried out with full railway carriages. The first test, designated FS1, involved an old steel bodied railway carriage which was refurbished with a ‘modern’ interior, similar to that used on a German ‘Inter City Express’ train. Tunnel ventilation was set on ‘low’ with the longitudinal airflow before the test being less than 0.4 ms−1. The first ignition of this fuel load burned out without growing significantly after 18 min. Following this, 170 wood sticks were added to the carriage, in the vicinity of the ignition location, and some of the windows were opened. The fire was reignited, designated Test FS2. The fire grew and consumed all combustibles in the vehicle, reaching a peak HRR of 19 MW after 80 min, as shown in Fig. 88.3.

The second rail carriage test also featured a steel bodied inter-city vehicle, but with an older style interior to that tested in FS2. This test was designated F11. Once again, the ventilation was set to ‘low’. The fire reached a peak of about 13 MW after about 25 min, which quickly diminished by about 50 % and then grew again to a second peak of about 12 MW after 90 min. A simplified representation of the HRR data from this test is given in Fig. 88.3.

The subway carriage tested was aluminium bodied and the seats in it were replaced with seats corresponding to the contemporary standards at the time of testing. Two ventilation panels and one of the carriage doors were left open for the test. Ventilation was set to ‘low’ as with other tests. The first ignition of the load (Test F41) used 0.7 kg of isopropanol in two fuel pans as an initial fuel, but this burned out before spreading to involve the rest of the carriage. The second ignition (Test F42) involved 6.2 kg of isopropanol in two larger fuel pans. This resulted in a very rapid fire growth and witnesses report the entire carriage was involved in the fire within 2 min. The peak HRR was about 35 MW after about 5 min, as shown in Fig. 88.1.

Large Fire Test in the 2nd Benelux Tunnel (2001)

A number of fire tests were carried out in the 2nd Benelux Tunnel, Rotterdam, the Netherlands, shortly before the tunnel was opened to the public [15]. The tunnel is rectangular in section and is approximately 5.1 m high by 9.8 m wide. While several of these fire tests involved vehicles and vehicle mock-ups, only one of the tests (Test 14) had a very large fire load using 72 wood pallets using a longitudinal velocity of 1–2 ms−1. In three tests, designated tests 8–10, half this fuel load was used (36 pallets) but the velocity was varied from test to test.

In Test 14, eight stacks, each containing nine wooden ‘Euro pallets’, were arranged in a 2 × 4 configuration to represent a HGV cargo. Six car tyres were added to the load and two small bowls of petrol were ignited at the centre of the load to start the fire. The ventilation used was longitudinal at about 1–2 ms−1. The primary aim of Test 14 was to investigate the capabilities of a sprinkler system in protecting a fuel tanker, which was positioned downstream of the fire. The sprinklers above the tanker were activated after 21 min into the test. The sprinklers cooled the tanker, but were not directed toward the fire source until after 31 min into the tests when they were manually activated, meaning it burned freely most of the test time. In Tests 8–10, the fuel load consisted of four stacks with nine pallets in each without any interaction of the sprinkler system or any cover. In Fig. 88.4 the adapted data from Lemair and Kenyon [16] is given.

Fig. 88.4
figure 4

Approximate representation of HRR data from [15]. Test 14 (solid line, 72 pallets, 1–2 m/s), Test 8 (short dash, 36 pallets, natural ventilation), Test 9 (double line, 36 pallets, 4–6 m/s) and Test 10 (long dash, 36 pallets, 6 m/s) of the Benelux tunnel fire tests

Full size image

When observing Fig. 88.4 two important observations can be made. The maximum heat release rate is affected by both the ventilation rate and the number of wood pallets (tilt the flames), (i.e. the amount of fuel). Higher ventilation rates tend to spread the flames faster inside the pallets, and thereby increase the fire growth rate considerably (the growth rate of the fires varies considerably depending on ventilation rate). Consequently the maximum heat release rate is increased as the tilted flames spread faster through the entire fuel surface.

The Runehamar Tunnel Fire Tests (2003)

A series of four fire tests was carried out in a disused two-lane road tunnel in Norway in September 2003 [2, 3]. The tunnel is 1.6 km long and has a rough rock cross-sectional area of about 47–50 m2. At the location of the fire experiments (approximately 1 km into the tunnel), a 75 m length of the tunnel was lined with fire-protective panels, and this reduced the cross-sectional area of the tunnel to 32 m2 in the vicinity of the fire.

Each of the tests comprised a fire load of equivalent size and shape to a standard HGV trailer (10.45 m long, 2.9 m wide, 4.5 m high). In test 1, the fire load consisted of 11.0 tonnes of wooden pallets and mattresses. There was no plastic material except the tarpaulin, but that was the case in all the tests and a ‘target’ object positioned 15 m downstream of the fire. In test 2, the fire load consisted of 6.9 tonnes of wooden pallets and mattresses. In test 3, the fire load consisted of 7.8 tonnes of furniture on wooden pallets, and ten tyres (800 kg) were positioned around the frame at the locations where they would be on a real HGV trailer. In test 4, the fire load consisted of 2.85 tonnes of plastic cups in cardboard boxes on wooden pallets. In each test the amount of plastic materials was estimated to be about 18–19 % (by weight). Similar tarpaulin coverings were used in all four tests. In each test, two mobile fans positioned near the tunnel portal were used to generate a longitudinal airflow, this was about 3 ms−1 at the start of each test, but reduced to about 2–2.5 ms−1 once the fires became fully active. The peak heat-release rates of the fires in tests one to four were 202, 157, 119 and 67 MW, respectively. The HRR data are shown in Fig. 88.5.

Fig. 88.5
figure 5

HRR data from the Runehamar fire tests [2, 3] with Test 1 (solid line), Test 2 (long dash), Test 3 (double line) and Test 4 (short dash)

Full size image

Fire Tests at Carleton University (2011)

Two full scale fire tests were carried out in a 10 m wide, 5.5 m high by 37.5 m long ‘tunnel’ facility at Carleton University in Canada [17]. The facility is open to the outside through a reduced height opening at the ‘upstream’ end and is blocked at the ‘downstream’ end, with an opening in the ceiling leading to smoke/air extraction ducts. The ventilation in the facility is generated by extraction fans, with a maximum capacity of 132 m3/s. One of the tests involved an intercity style railway carriage, 23 m long, 3 m wide and 3.7 m high, while the other test involved a subway car which was 19.7 m long, 3.15 m wide and 3.45 m high. The estimated fire load of the railway car was 50 GJ and the estimated fire load of the subway car was about half of that. In both tests the extraction system was initially set to 50 % capacity, but was increased to maximum capacity after 3 min into the subway car test and 6 min into the intercity car test. The longitudinal flow during the experiments is not reported in the paper, but is assumed to decrease as the fires grew, due to greater volumes of smoke generated, yet a constant extraction rate. The HRR data, estimated by oxygen consumption, are shown in Fig. 88.6.

Fig. 88.6
figure 6

Approximate representation of HRR data from the Carleton University fire tests. The rail car is indicated by the broken line and the subway car by the continuous line (Adapted from [16])

Full size image

Subway Car Tests During Project METRO (2011)

Two fire tests involving fires inside subway cars were carried out as part of the METRO project in Sweden in 2011 [4, 5]. The fire tests were carried out in the 276 m long abandoned ‘Old Brunsberg Tunnel’ in Western Sweden. The tunnel is an irregular horseshoe in shape and varies from 6.7 to 7.3 m high and from 5.9 to 6.8 m wide. One subway carriage was used as supplied, with older style seats and furnishings, the other was refurbished inside to resemble a modern ‘C20’ train carriage. This was obtained by cover the combustible wall and ceiling linings with 1.5 mm thick aluminium sheets. A realistic selection of luggage types were placed in the subway carriages as fire load. Longitudinal flow in the tunnel was generated by a mobile fan, positioned upstream of the fire location. This generated a flow of between 2 and 2.5 ms−1, before the fire was ignited. At peak burning, the fires both resisted the airflow, effectively reducing the flow velocity (see the discussion of the ‘throttling effect’ in section “Using Ventilation Systems for Fire Safety in Tunnels,” below). The fire tests were intended to simulate an arson fire on a subway carriage, so both fire tests were initiated with a small quantity of gasoline/petrol on one of the seats. In the test involving the old style carriage, the fire rapidly took hold and grew to flashover-like conditions within the carriage in a few minutes. In the test involving the refurbished carriage, the fire remained small and localised for over 100 min. However, this test also eventually reached flashover-like conditions and from there rapidly transitioned to a fully involved fire, much like the earlier fire test had done at about 12 min after ignition. This was due to the ‘melted’ aluminium sheets. Following this rapid fire growth, both tests exhibited very similar fire behaviour, as shown in Fig. 88.7, both reached a peak of about 77 MW, both burned at this level of intensity for about 12 min, before rapidly decreasing in intensity to below 20 MW, then slowly burning out over the next 30 min.

Fig. 88.7
figure 7

Approximate representation of HRR data from the METRO project fire tests [4, 5]. Test 2 is indicated by a solid line and Test 3 with a double line

Full size image

Observed Tunnel Fire Characteristics

In the design of fire safety systems in tunnels there is often a need for a “design fire”. This fire may be obtained from tabulated data or it may be based on engineering methods calculating the actual fire growth and maximum HRR. The tabulated data is often based on consensus of technical committees such as NFPA or PIARC. A round table discussion using the latest test results such as those presented above informs proposals about the highest value for different vehicle types. This gives guidance to engineers that need the information for design of ventilation systems, the tunnel structure or evacuation procedures. Traditionally it has been the design of the ventilation capacity that has been the primary focus of designers, although structural considerations have become more important in recent years. As advanced calculation models have been created and developed, the need for realistic design fires has increased.

Most fire protection engineers are familiar with the general picture of a compartment fire which grows in a ‘t2’ manner from ignition to flashover, burns at a more or less constant, ventilation controlled, peak heat release rate for a significant period of time, before beginning to burn out and entering a decay phase. Because of this familiarity, fire protection engineers have frequently tried to apply the same concepts and assumptions to tunnel fires. However, the above survey through the published tunnel fire heat release rate data should be sufficient to demonstrate that vehicle fires in tunnels do not conform to this common assumption.

It is clear from the above presented data that tunnel fires, particularly when influenced by longitudinal ventilation, do not grow in anything like a t2 manner. Instead the growth of a tunnel fire may more readily be characterised by a two step linear growth model [18]. The first step in this process, which is generally characterised by slow fire growth is variously described in the literature as the ‘incipient’, ‘incubation’ or ‘delay’ phase, may last for a very long time, as demonstrated by the METRO results (see Fig. 88.7), but may be as short as a few minutes. It is currently unclear exactly what factors govern how long the incipient phase lasts, although it has been suggested that variations in ventilation velocity may reduce or extend the duration of this phase [18]. The second step in this process is a very rapid growth phase which, in tunnels with significant ventilation flow, generally appears to reach peak fire size in just a few minutes. As is clear from Fig. 88.4, above, the rate of growth in this phase varies with longitudinal ventilation velocity [18]. Other models of fire growth are available, including the exponential model presented by Ingason [19] which has the benefit that the entire fire growth, peak and decay process can be described with a single mathematical expression.

Another way in which vehicle fires in tunnels differ from the common compartment fire model is in the duration of peak burning. By peak burning we mean the period where most of the fuel is fully involved in an intensive burning. In the majority of cases described above, the period of peak burning was relatively short, mostly less than 10 min and almost always shorter than 20 min, and was generally followed by a decay which was almost as fast as the rapid growth phase.

Another interesting aspect of vehicle fires is the observation that the HRR per square meter of exposed fuel is generally found to be in a very narrow range. This was first observed through analysis of the several tunnel fire tests [20] and it has been shown that the recent large scale tests in the METRO project confirm this trend. For most vehicles the maximum HRR per square meter of exposed burning fuel surface between 0.27 MW/m2 and 0.4 MW/m2, even for fully involved ‘post flashover’ railway carriages. The METRO tests were equivalent to 0.31 MW/m2 at maximum HRR. This information can be used to estimate the potential maximum HRR for different types of vehicles. In HGV cargo this value may vary from 0.1 to 0.5 MW/m2, where the cellulosic materials will tend to have values at the lower end of the range, whereas plastics will tend to have higher values.

It is proposed that these features, common to all large fire experiments in longitudinally ventilated tunnels, be considered when defining ‘design fires’ for tunnels. This will be discussed in more detail below.

Multiple-Vehicle Fires and Fire Spread

All tunnel fire incidents which have been characterised under the heading of ‘catastrophic tunnel fires’ have involved fire spread from the initial vehicles involved in the incident, to other vehicles in the vicinity. Table 88.1 lists some of the most significant multiple vehicle fires of recent years. For a more comprehensive list, with details of these and many other incidents, see Carvel and Marlair [21].

Table 88.1 Significant recent multiple vehicle fires [20]
Full size table

In order to understand the fire spread in many large tunnel fires with multiple vehicles involved it is necessary to recognise how these fires progress and spread. Fires in road tunnels may start in one vehicle or in two vehicles due to a collision and then spread to the adjacent vehicles largely by radiation from the flames and the hot smoky gases. In rail vehicles at least one carriage needs to become fully involved in order to obtain fire spread between different carriages. It is not only the nature of the vehicles themselves that is required for a fire spread, but also the geometry of the tunnel, especially the tunnel height, and the ventilation conditions. The main mechanism of fire spread is through heat transfer by convection, radiation and conduction from the volume of flames and hot gases. The difference in the temperatures between the cooler solids and the hot gases drives this heat transfer.

Among the more spectacular fires that have occurred in tunnels are the large fires in the Channel Tunnel with HGVs on open train wagons in 1996 and 2008 [22]. Rew and Deaves [23] postulated five mechanisms for fire spread in tunnels in their analysis of the 1996 incident. These mechanisms are discussed in detail by Carvel [24] or chapter 11 in Ingason et al. [8]. The first mechanism of fire spread in large fires is flame impingement. In low ventilation conditions, the flames are deflected by the presence of the ceiling, mainly in the direction of the ventilation flow. The flames appear to ‘crawl’ along the ceiling above the vehicles, see Fig. 88.8. Rew and Deaves [23] presented a flame length model for tunnels, which included the heat release rate of the initial fire and the longitudinal flow velocity but not the tunnel width or height. Much of their work is based on the investigation of the Channel Tunnel fire in 1996, with test data from the HGV-EUREKA 499 fire test [1] and the Memorial Tests [6]. They defined the horizontal flame length, L f , as the distance of the 600 °C contour from the centre of the HGV or the pool, or from the rear of the HGV, see Fig. 88.8.

Fig. 88.8
figure 8

The flame impingement mechanism in tunnels with longitudinal flow. The target vehicle is about to ignite and Lf is the flame length [24]

Full size image

The other mechanisms of fire spread in tunnels are surface spread, that is to say flame spread across the surface of the fire load, remote ignition (or spontaneous ignition); the conditions at vehicles that are not very close to the initial fire are raised to the point of spontaneous ignition due to the high temperatures produced by the fire. Another is fuel transfer, since burning flammable liquids can spread from fuel tanks in the downhill direction, or flaming debris may blow downwind of the fire. The last mechanism postulated was explosion i.e. explosion of fuel tanks may spread burning fuel to adjacent vehicles. None of these other postulated mechanisms fully explain the fire spread between multiple vehicles which has occurred in large tunnel fires. The first mechanism postulated, i.e. flame impingement, in combination with spontaneous ignition when the flames virtually crawl over the target vehicle, as shown in Fig. 88.8, is generally the most plausible option. This has been confirmed through numerous full scale and model scale tests performed by SP, the Technical Research Institute of Sweden (SP) [3, 25–28]. Other works on fire spread in tunnels can be found in the work by Beard et al. [29–32].

The possibility of ignition of the target vehicle can be obtained by investigating whether or not the exposed surfaces of the vehicle would attain a critical ignition temperature, Tcr. There are two types of ignition, piloted ignition or spontaneous ignition. The possibility of ignition of the target vehicle can be obtained by evaluation of whether or not the exposed surface would attain a critical ignition temperature. The critical temperature, T cr , varies with the nature of the target material, but is generally estimated as follows: 600 °C for radiant exposure or 500 °C for pure convective exposure in the case of spontaneous ignition; or, 300–410 °C for radiant exposure or 450 °C for pure convective exposure in the case of piloted ignition. Note that these values are mostly deduced from experiments on small vertical specimens [33]. Newman and Tewarson [34] argued that, at ignition, \( {T}_{cr}\approx {T}_{avg} \) (the average gas temperature over the cross-section at a certain position) for a duct flow, i.e. when the T avg has obtained a critical ignition temperature, the material at that location will ignite. This would comply with the flame impingement mode discussed earlier and complies well with the experimental observations obtained by SP.

There is a strong correlation between the ignition and the ceiling gas temperature in a ventilated tunnel fire [3]. The critical ceiling gas temperature, i.e. the minimum ceiling gas temperature required to ignite the material, has been analyzed for full and model scale tests. In the vicinity of the flame volume, radiation dominates the heat transfer. Since the tunnel ceiling is enclosed by the flame and hot gases, the view factor can be regarded as being unity. Thus the incident heat flux at the ceiling can be simply expressed as [3]:

$$ {\dot{q}}_{inc}^{{\prime\prime} }={\varepsilon}_g\sigma \left({T}_g^4-{T}_s^4\right) $$
(88.1)

where ε g is the gas emissivity, σ Stefan-Boltzmann constant (kW/m2⋅K4), T g is the gas temperature in degrees Kelvin and T s is the solid surface temperature, also expressed in Kelvin. According to Ingason and Li [35], a gas temperature of 600 °C should be obtained for wood before its spontaneous ignition in model tunnels, see Fig. 88.9.

Fig. 88.9
figure 9

The critical gas temperature beneath ceiling when a second wood crib ignites [26]. The number on the horizontal axis refers to test number in the test series

Full size image

In a recent work by Ingason et al. [3] analyzing the Runehamar fire test data, it was concluded that the critical ceiling gas temperature is about 700 °C for fire spread to a wooden object at the floor level and about 600 °C for fire spread to a wooden object with surface close to the ceiling in a tunnel fire. The location of vehicle target surface relative to the smoke layer height plays therefore an important role in fire spread. Therefore, the authors argue that fire spread to subsequent vehicles is very difficult to occur if the vehicle height is small compared to the tunnel height, perhaps below about half the tunnel height [3].

When the fire starts to spread to the second vehicle the fire development is difficult to stop if there are more vehicles in the vicinity of the first two vehicles. Figure 88.10 gives a schematic description of a large tunnel fire with relatively high forced longitudinal ventilation rate and multiple vehicles. The burning process can be viewed as stationary and the fire-spread progress can be divided into different key zones [25, 36]:

Fig. 88.10
figure 10

Schematic presentation of a flame impingement fire spread process in a tunnel with multiple vehicles/coaches [24, 35]. Note: there may be multiple vehicles in each designated zone, and the zones are only shown as having the same length as each other for convenience

Full size image

Provided there are sufficient vehicles in the vicinity of the initial fire, these different zones move forward in a dynamic manner. The ‘burnt out zone’ involves vehicles that have been completely consumed in the fire and where the fire gases are now cool. The ‘glowing ember zone’ contains vehicles at a very late stage of the decay phase (predominantly a pile of glowing embers). The ‘combustion zone’, contains the violently burning vehicles (fully developed fire) where sufficient fuel is vaporising to support gas phase combustion. Flaming combustion is observed throughout this zone. The gas phase temperature in this zone increases rapidly. Simultaneously the oxygen is rapidly depleted and the temperature reaches a maximum just beyond the ‘combustion zone’. This gas temperature can vary between 900 and 1365 °C.

If all the oxygen is consumed within the combustion zone, an ‘excess fuel zone’ starts behind this. In the case where all the oxygen has been depleted we have a ventilation-controlled fire. In the case where useful oxygen is still available (due to flammability limits, oxygen concentrations below about 8–10 % may not be ‘useful’ unless the gas phase temperature is high enough to sustain combustion) we have a fuel-controlled fire and no excess fuel (no more vehicles in place to sustain the progress). In the case of ventilation-controlled fire, the fuel continues to vaporise from the vehicles throughout this zone up to a point along the tunnel where the gas stream has cooled to the fuel vaporisation (pyrolysis) temperature (Tvap ≥300 °C for the majority of solid materials). Beyond this point, no vaporisation of the vehicle materials occur but the gas flows into a ‘preheating zone’ and loses its heat to the tunnel walls and preheats the vehicle material within this zone.

de Ris [36] showed that the combustion and excess fuel zone lengths are proportional to the forced ventilation rate when the fire becomes ventilation-controlled and Comitis et al. that the fire will propagate at a constant speed [37] (when ventilation-controlled) provided that there is enough combustible material available. Delichatsios [38] concluded that active burning in a fibreglass-reinforced plastic duct will take place up to a maximum length corresponding to L c /D h  = 10 where D h is the hydraulic diameter of the duct (m) and L c is the combustion length (m). Assuming that this number is reasonable for tunnels the combustion zone, for a ventilation-controlled fire with forced ventilation, would not exceed 50–100 m for most common road and rail tunnels (assuming high vehicle density) where D h is ranging between 5 and 10 m.

Considerations for Design Fires for Vehicle Tunnels

Tunnel designers, ventilation engineers, suppression system manufacturers and others commonly need to define the characteristics of the fire which they are designing their facilities or systems to be able to manage. Thus, specification of realistic design fires is of great importance, if tunnel facilities and systems are to be made robustly.

A design fire need not be the most severe fire imaginable for a given situation, but should be at the upper bound of fire scenarios deemed credible, or likely to happen within a specified timescale. In other words, it need not be necessary to design a system to be able to manage a ‘once in 10,000 years’ incredible event, but consideration should be given to the more likely, though still improbable, ‘once in 1000 years’ type event.

Design fires should also be tailored to the system being designed. For example, a design fire for structural protection or structural resistance calculations should be radically different to the design fire used for the detection and alarm system.

There are a number of guidance documents which discuss design fires for tunnels, which go into the subject in far greater detail than can be devoted here, including [39–41]. These brief comments are intended to enable the designer to consider realistic fire features in design fires.

Design fires for tunnel structures, fire protection system design and smoke management should be based on the realistic maximum peak fire size. Such design fires need not consider the incipient stage of the fire, and assuming the growth phase to be of only a few minutes in duration, this can be neglected too. The design fire here should therefore be characterised by a maximum HRR plateau, from the outset, the duration of which should be prescribed on the basis of the maximum number of vehicles deemed likely to be involved in a fire. The duration of peak burning is limited by the availability of fuel and the availability of oxygen, and therefore this period of peak burning can be extended if additional fuel becomes available by means of the fire spreading to an adjacent vehicle or vehicles.

For property protection purposes it is also important to consider the cooling/decay phase of the fire, as it is well known that some structures which have endured throughout a fire or a furnace test at high temperature have failed as they cool, however this is rarely commented upon in the literature.

In situations like the experiments described above, where the fire load was large, relative to the tunnel, the peak HRR may be defined and limited by the available oxygen in the airflow. The maximum theoretical (ventilation-controlled) HRR may be estimated using the well known correlation between HRR and the mass of oxygen consumed in a longitudinal ventilation flow:

$$ \dot{Q}=\left(0.21-{\eta}_{O_2}\right)V\;\rho \varDelta {H}_{c, ox} $$
(88.2)

where ηo2 is the depleted oxygen level on the downstream side, V is the volumetric airflow, ρ is density of the inflowing airflow (approx. 1.2 kg/m3) and ΔHc,ox is the heat of combustion for oxygen, generally taken to be 13 kJg−1 [42].

Thus, for example, in a tunnel with a 40 m2 cross section and a 2 ms−1 longitudinal flow, we might expect the theoretical maximum ventilation controlled HRR to be about:

$$ \dot{Q}=\left(0.21-0.0\right)\times 2\times 40\times 1.2\times 13.02=262MW $$

It should be noted that in most circumstances with one or two collided vehicles involved in a fire, it is not likely that all the oxygen in the airflow will be usable by the fire as much of it may bypass the combustion zone. In the case when another vehicle is standing downstream a fire, the environment may not be combustible due to flammability limits (relation between the oxygen concentration and gas phase temperature).

It is important to note here that, provided sufficient fuel is available (for example, in multiple vehicle fire scenarios), it is the tunnel airflow which defines the theoretical peak fire size, and hence should define the design fire for structural purposes. A similar design fire should be used also for dimensioning the ventilation system. If fuel is limited, for example by considering only a single vehicle fire, then the fire may remain fuel controlled and the peak size may be estimated based on the fuel area, as described above [20].

However, this idea goes against the notion, so prevalent in the tunnel fire and design literature, that a ventilation system designer should first identify a ‘design fire’ size and then dimension the ventilation system in order to generate adequate smoke control. In the case of a multiple vehicle fire, the fire size does not define the required airflow, rather the airflow entirely defines the maximum possible fire size.

This relationship, where the theoretical maximum fire size is defined entirely on the basis of ventilation flow, will hold in general terms in situations like most of those described above, where the vehicle on fire is large, relative to the cross-section of the tunnel; that is to say, in single lane tunnels, which are generally subway and railway tunnels. The relationship is not likely to be as simple as this in two or more lane tunnels, such as are common for road transport. In larger tunnels, it is expected that a greater proportion of the longitudinal airflow will bypass, not directly feed the fire, so the resulting HRR could be significantly lower.

For single lane tunnels where multiple vehicle fires are possible, however, it is proposed that the peak HRR of a ‘design fire’ for structural and ventilation considerations, should be based entirely on the anticipated longitudinal airflow in the tunnel.

As noted above, the factors governing the duration of the incipient phase are not well understood at the present time, therefore, without further tests to demonstrate the performance of a particular system, it should be assumed—for design purposes of all tunnel systems—that the incipient stage will remain short, perhaps only of the order of a few minutes. Assuming an extended incipient phase as part of a design fire cannot be justified from the above data.

Design fires for detection systems and water spray systems need to consider the growth phase of the fire. Such systems are intended to detect the fire while it is as small as possible, and to suppress or control the fire, that is to say, prevent the fire from reaching its unsuppressed potential. The important characteristics of a fire for design of a water spray system are therefore those relating to fire growth [18]. Important parameters to consider here are the ventilation rate in m/s and the ceiling height. These parameters govern, together with the HRR at each time interval, the gas temperature at the tunnel ceiling. The location of the maximum gas temperature varies along the ceiling depending on these parameters. For further information see Li and Ingason [43].

Using Ventilation Systems for Fire Safety in Tunnels

The systems which we commonly consider to be safety systems for tunnels were not, in most instances, initially installed in tunnels as safety devices. The most common safety system in tunnels is the ventilation system, yet it must be recalled that the original and primary function of ventilation systems is not smoke management in the event of a rare event like a fire, but rather is pollution and environmental control in everyday usage.

The first ventilation system installed in a railway tunnel was in Liverpool’s Edge Hill tunnel in the UK, in 1870, although mechanical ventilation had been commonplace in mine tunnel networks for at least three centuries before this [44]. This was an exhaust fan for the removal of smoke from steam engines. In 1927, the Holland Tunnel in the USA became the first road tunnel equipped with a (fully) transverse ventilation system, that is, a system of ducts and openings provided fresh air into the tunnel at periodic locations along its length, while a second system of openings and ducts extracted polluted air from the tunnel at periodic locations. The first tunnel to have longitudinal ventilation with jet fans was the Bargagli-Ferriere tunnel in Italy in 1971. While all these systems were originally conceived as means of replacing polluted air with fresh air, these systems began to be understood as a means of controlling smoke in the event of a fire in a tunnel. The addition of jet fans to transversely ventilated tunnels in the 1980s (for example, the Dartford tunnel in the UK) reflects this change in mindset. In 1994, the Channel Tunnel opened with the world’s first ventilation system intended ‘from the drawing board’ as a smoke control system as well as a pollution control system [45].

Because almost all tunnels of significant length have ventilation systems, it is easy to understand why these systems became adapted to be emergency response devices. For the past two decades, perhaps longer, the tunnel safety industry appears to have been fixated on questions associated with using ventilation for smoke control in the event of a tunnel fire. The question of appropriate design for ventilation systems has therefore been generally reduced to the question of extract capacity, for transversely ventilated tunnels, or of identifying the ‘critical ventilation velocity’ (CVV) for tunnels with longitudinal ventilation.

The CVV is generally defined as the minimum longitudinal ventilation velocity required to blow all the smoke produced by a fire in a tunnel to one side of the fire location, that is, to prevent any smoke propagating upstream of the fire location, against the prevailing airflow. CVV varies with tunnel geometry and with fire size, up to a certain limit. A number of studies, including that of Oka and Atkinson [46], have identified that there is generally a ‘super critical ventilation velocity’ which is sufficient to control the smoke from all sizes of fire, for a given tunnel geometry. A review of tunnel CVV studies is presented by Ingason [47] and need not be repeated here.

The critical flow conditions are generally expressed in the literature in terms of a single ventilation velocity, which is commonly taken to be the average flow across the tunnel cross-sectional area, however, it may be better in some instances to speak in terms of a critical volumetric flow rather than a velocity. The focus on ventilation velocity also masks the throttling effect of tunnel fires, identified in the 1970s by Lee et al. [48]. In a given tunnel it may well be the case that, say, a 3 ms−1 airflow may be sufficient to control smoke in the event of either a 50 MW fire or a 100 MW fire, but expressing the situation in these terms hides the fact that more ventilation devices will be required in the case of a 100 MW fire to generate the 3 ms−1 airflow than would be required to generate the same magnitude flow in the case of a 50 MW fire. A computational example of this, derived from Vaitkevicius et al. [49], expanding on the work of Colella et al. [50], is shown in Fig. 88.11. Here, there is no increase in CVV beyond 31 MW, yet the number of fans required to generate the CVV continues to increase with fire size.

Fig. 88.11
figure 11

Critical Ventilation Velocity (CVV; dotted line) and the number of jet fans required to generate CVV (diamonds) vs. fire heat release rate, for a modelled rectangular tunnel (Adapted from Vaitkevicius et al. [49, 50])

Full size image

While our understanding of longitudinal ventilation for smoke control is reasonably advanced [51], our understanding of the influence of ventilation on fire behavior is considerably less mature. Various studies in the past decade have shown that peak fire size is influenced by longitudinal ventilation velocity [52] and by the geometry of the tunnel [53], other studies have shown that longitudinal ventilation also has a significant influence on fire growth rates [18, 27]. Each of these factors must be considered when designing ventilation systems for use in emergency situations.

As observed above, in tunnel fire experiments the fire tends to grow in two reasonably distinct stages. For the first few minutes after ignition there is an incipient phase, followed by a period of much more rapid fire growth. From observation of the experimental tunnel fire data, there appears to be a weak relationship between the length of the incubation phase and the longitudinal ventilation velocity [18]. Such limited data as are available suggest that the incipient phase is generally shorter at longitudinal velocities of about 2–3 ms−1 than it is at lower (0–1 ms−1) and higher (circa 6 m/s) velocities. This suggests that ventilation flows of about 2–3 ms−1 should be avoided in the initial stages of a fire. Unfortunately, 3 ms−1 is close to the CVV for design fires in many tunnels, so has commonly been used as an emergency ventilation flow. However, it should be noted that many different (and apparently contradictory) emergency ventilation strategies are used in tunnels worldwide, some opting to increase the flow greatly when a fire is detected, in order to dilute the smoke, while others opt to reduce the flow below about 1.5 ms−1, in order to maintain smoke stratification.

There are also issues associated with reversing ventilation direction [54]. It appears, from an analysis of the Channel Tunnel fire incidents, that changes in the direction of the airflow, after a fire is fully developed, can lead to very rapid fire spread between vehicles and very fast fire growth rates, but further work is required to confirm these observations.

Passive Fire Protection: Tunnel Lining Systems

As with ventilation systems, the original intent of the tunnel lining system was not as a fire protection system, and many older tunnel lining systems which are still in use today serve only a cosmetic function and provide no fire protection. Tunnel lining systems were primarily installed to act as drip shields, to provide a visual aesthetic and, in some instances, for acoustic reasons.

Today, while the above reasons remain valid, the main design consideration for tunnel lining systems is how much thermal protection they can provide for the primary tunnel structure. Modern tunnels are generally made of concrete, although many older tunnels constructed out of cast iron segments or masonry are still in use. Concrete has excellent structural properties at ambient temperatures, but may fail when subjected to high temperatures and rapid heating, under fire conditions.

According to the guidance of NFPA 502 [55], the primary concrete structure in tunnels with cast in situ concrete provided with structural fire protection material temperatures shall not exceed 380 °C at the concrete surface and 250 °C at any of the rebar, when subjected to a time-temperature curve deemed equivalent to a tunnel fire. These requirements are mirrored in other standards [56]. Temperatures up to these limits will not significantly reduce the load carrying capacity of concrete.

One major concern with concrete structures is the still inadequately understood phenomenon of spalling [57]. Spalling is an explosive failure which commonly occurs when concrete is rapidly heated and steep temperature gradients are formed in the bulk material, leading to thermal stresses and high pore pressures in the material. Spalling has also been observed during rapid cooling, such as during fire-fighting activities [58]. Spalling has even been observed to occur at relatively low temperatures [59], so the function of a fire protection system is not merely to protect the structure from exposure to high temperatures, but also to protect the structure from rapid heating. As well as having excellent thermal barrier properties, it is vitally important that passive fire protection systems not be prone to spalling themselves.

Passive protection systems for tunnels are generally either a bolted on panel system or a sprayed on mortar system. Other options, such as intumescent coatings which are common in the building industry, do not seem to have been considered for tunnel applications. Research is on-going into the performance in fire of concrete containing various forms of fibres, some of which have been demonstrated to resist spalling and, thus, improve the fire resisting properties of concrete [60]. However, neither a spalling proof concrete mixture, nor a method for accurately predicting spalling behaviour has yet been developed. If suitable concrete mixes can be developed, this may have a considerable impact on the way that future concrete tunnel structures are made.

Finally, there is an increasing tendency in the tunnel protection industry to consider fire suppression systems as structural protection systems, and a number of recent papers have discussed ‘trade-offs’, that is, reducing or removing the requirement for a passive thermal barrier system if a suppression system is being installed [61, 62]. This will be discussed further in the next section.

Water Spray Systems

It is the opinion of one of the authors that the terminology used to describe water based fire protection systems these days is somewhat misleading [63]. The way the word ‘suppression’ is used by fire protection engineers strays considerably from the dictionary definitions of the word ‘suppression’ used by the rest of the world. Describing water spray systems as ‘suppression’ systems implies that these technologies have the ability to halt the growth of, or reduce the size of fires in tunnels. According to the definition given by NFPA 502, “Fire suppression systems are designed to arrest the rate of fire growth and significantly reduce the energy output of the fire shortly after operation”. In order to do that the system has to be able to deliver certain amount of water onto the burning fuel surface and actively cool it. Simultaneously it has to have the ability to hinder fire spread to adjacent fuel surfaces. As will be discussed below, neither of these abilities is guaranteed or even commonly observed with many ‘suppression’ systems promoted for use in tunnels. Such systems do not fulfill the basic requirements implied in the word ‘suppression’.

Similarly, the terminology which seems to be in vogue at present, ‘fixed fire fighting systems’ (FFFS), implies that the systems so described are actually able to ‘fight’ fires. Once again, this is not always guaranteed. There is a need for clear definitions that should be accepted by all involved partners. The term “deluge system” relates to systems with open nozzles operated on zone-to-zone basis creating water droplets in order of 1 mm, whereas “water mist system (WMS)” is a high-pressure deluge system creating very small droplets, in order of 50–250 μm. The blanket term FFFS is used to include all these system, together with foam based systems (high expansion, CAFS), which we do not consider in the short discussion here. The terms ‘suppression’ and ‘fire fighting’ both imply performance criteria which not all systems described as such can achieve. Thus, the term water spray system will be use here to include both deluge systems (even with foam additives) and water mist systems.

Although the benefits of fire sprinkler systems in buildings have been recognized widely for many years, the use of water spray systems in tunnels has generally been discouraged until comparatively recently. However, this generalisation is not universal. Deluge systems have been routinely used in Japanese tunnels since the 1960s, and in Australian tunnels since the early 1990s. There were hardly any European or North American tunnels with water spray systems until the last decade. In 1999, the World Road Association, PIARC reported that no road tunnels in Belgium, Denmark, France, Italy, the Netherlands, the UK or the USA were equipped with water spray systems, while Sweden had only one and Norway had two [64].

Deluge systems have been routinely installed in Japanese tunnels since 1963, when the first system was installed in the Tennozan Tunnel. Eighteen experimental test series were carried out in Japan between 1960 and 1985, the majority of these being carried out using fuel pans, although fire tests using cars, a small van, trucks and a bus were also carried out [65]. In general, these tests demonstrated the effectiveness of the deluge system in reducing temperatures in the tunnel and hindering fire spread. However it was observed that “the fire tends to expand less in vertical and more in horizontal direction” and that while fires of “wood or other open load” could be extinguished, fires inside or underneath vehicles “could not be extinguished at all”.

In practice, the Japanese experience is also good. Between 1963 and 1990 there were 90 fire incidents in tunnels operated by the Japan Highway Public Corporation. The vast majority of these fires (80 %) did not require the application of the (manually operated) water spray systems. In the 16 instances when the deluge system was used, the results were found to be “fairly satisfactory” [65]. No incident escalated to disastrous proportions. Australia has a shorter length of experience of deluge systems—the first system was installed in the Sydney Harbour Tunnel in 1992—but their experience is also positive [66]. The deluge systems have proven to be effective in vehicle fire exercises and during fire incidents, the highest profile of which was the fire in the Burnley Tunnel in March 2007 [67]. In general, testing and experience has demonstrated that water spray systems are reasonably able to contain vehicle fires and prevent fire spread in tunnels, allowing safe evacuation of people, effective fire fighting by the fire brigade and a rapid return to normal operation of the tunnel.

However, elsewhere in the world, the negative attitude to water spray systems in tunnels extends back to the 1960s and 1970s, but was enshrined in international recommendations up until the late 1990s and early 2000s, as exemplified by this quote from the 1999 PIARC report “Fire and Smoke Control in Road Tunnels” [64]:

…the use of sprinklers raises a number of problems which are summarised in the following points:

  • water can cause explosion in petrol and other chemical substances if not combined with appropriate additives,

  • there is a risk that the fire is extinguished but flammable gases are still produced and may cause an explosion,

  • vaporised steam can hurt people,

  • the efficiency is low for fires inside vehicles,

  • the smoke layer is cooled and de-stratified, so that it will cover the whole tunnel,

  • maintenance can be costly,

  • sprinklers are difficult to handle manually,

  • visibility is reduced.

As a consequence, sprinklers must not be started before all people have evacuated.

Based on these facts, sprinklers cannot be considered as an equipment useful to save lives. They can only be used to protect the tunnel once evacuation is completed. Taking into account this exclusively economic aim (protection of property and not life safety) sprinklers are generally not considered as cost-effective and are not recommended in usual road tunnels.

The spate of catastrophic tunnel fires in road tunnels which occurred almost immediately following the publication of this report forced the transport tunnel industry to reconsider the importance of property protection and continuity of business, particularly on the trans-European road network, and the industry very quickly realised that water spray systems were a viable and economic option for improving life safety levels and ensuring business continuity.

A number of large, government funded, research projects were carried out in the 2000s to address the issues of tunnel safety, security and sustainability, including the UPTUN project [68], the SAFE-T project and the SOLIT project [69]. Integral to each of these projects was a fresh consideration of suppression technologies for tunnel applications, and as the focus of much of the research was on ‘novel’ technologies, deluge systems were generally overlooked in favour of water mist systems (WMS).

Detailed reviews of the findings of these projects are published and summarised elsewhere [e.g. 70] and do not need repeated here. To date, no WMS has been deployed during a real fire incident in a tunnel, so there is no information on their capabilities in practice.

Some testing of WMS for tunnels has been presented in the public literature. Almost all of these tests have involved either open pans of liquid fuel or stacks of wooden (occasionally plastic) pallets. In general, it is claimed that water mist systems have been demonstrated to: reduce the temperature in the vicinity of the fire, reduce visibility in the locality of the fire, improve visibility at locations distant from the fire, slow or halt the fire growth rate, reduce the ‘peak’ fire size, prevent fire spread and protect the tunnel structure.

However, in some instances (in the public literature, presumably also in some of the many unpublished tests), the WMS have not performed as expected [71]. For example, in one of the fire tests carried out in the Hagerbach tunnel complex in Switzerland for the Paris A86 project [72] the fire (initially involving three cars in a ‘collision’ configuration) spread to two adjacent cars while the WMS was operational. From the presented data, WMS seem unable to halt fire growth on solid cargo loads, but fire growth rate is, in general, observed to slow or, at least to be interrupted for a period of time upon WMS activation. Under certain circumstances (cargoes covered by tarpaulin, mixed load of wood and plastics) fires have been observed to grow considerably, over a period of minutes, during WMS operation. In one instance the WMS was switched on when the (covered, mixed load) fire was only about 10 MW in size, and the fire continued to grow to over 60 MW [73]. In another similar instance, the WMS was switched on when the (covered, wooden cribs) fire was about 10 MW in size and the fire grew to about 55 MW during system operation [74].

Thus, it appears that for all but the simplest, uncovered, fire loads, fire control cannot be guaranteed using a water mist system. However, such systems do appear to perform very well in terms of temperature control (hence structural protection) and, to some extent, for smoke control.

WMS (and other water spray systems) effectively cool the hot smoke layer in tunnel fire scenarios. This tends to destroy stratification, but also to reduce the momentum of the smoke. Thus, if water spray systems are operated in tunnels, considerably smaller ventilation velocities should be required to control smoke. This has yet to be studied in any detail. There may also be an optimum ventilation airflow for promoting the suppressing effects of water spray systems. This also has yet to be studied in any detail.

It should be stressed that the emphasis of this discussion is on the influence of the various systems on the fire itself. Even though the behavior of the fire itself may not be greatly impacted by the activation of certain types of FFFS, in some situations, such systems may have a large effect by mitigating the consequences of the fire on the structure and on persons and objects distant from the fire. Water sprays have excellent thermal radiation blocking properties, which serve to protect the structure, to allow the fire brigade to approach the fire to fight it and also to prevent fire spread to other vehicles, in conditions where there is no direct flame impingement [63].

At a workshop presenting the results of the SOLIT2 project in Gijon, Spain, in 2011 [75], one of the delegates, having heard the project results and having witnessed a very large fire (comprised of wooden pallets, approximately 20–25 MW at peak HRR) in a tunnel with an active water mist system, commented that the action of the WMS effectively made the magnitude of the fire (expressed as a heat release rate) an irrelevant concept for specifying the design fire for structural design of tunnels. While the WMS may not be able to suppress or reduce the absolute HRR of a vehicle fire in a tunnel, it has been demonstrated able to considerably mitigate the effects of such a fire, such that the thermal attack on the structure due to, say, a 25 MW fire in water mist may be less than that due to a 4 or 5 MW fire without water mist.

Observations such as these suggest that fixed water spray systems in tunnels should really be considered as ‘protection’ systems rather than ‘suppression’ systems. The ability to suppress a fire is not guaranteed, the ability to protect a structure, on the other hand, has been more than adequately demonstrated [76].

Of course, if we were to re-classify water spray systems as protection systems, then the question of ‘trade-offs’ mentioned above becomes less controversial. It is no longer a question of reducing passive structural protection if a suppression system is installed, the question resolves to which system is better in the instance considered—a passive thermal barrier, or an active water spray thermal management system? There is no ‘one size fits all’ solution for all tunnel environments, but in every instance it is a matter of engineering judgment and calculation.

Concluding Comments

Tunnel fire safety is a complex problem with no clear solution at this moment in time. Even in tunnels with cross passages to another tube or a safety gallery, egress distances can be several hundreds of metres. Vehicles carrying ‘non-hazardous’ cargoes can result in rapidly growing fires with peak heat release rates of over a hundred megawatts. Yet systems exist which can mitigate the effects of large fires in tunnels and provide adequate protection for escaping tunnel users and thermal protection for the tunnel structure. Few countries have prescriptive standards for tunnel fire safety, so defining how such systems are to be installed and used in the event of a fire inevitably involves a ‘performance based’ analysis of each tunnel on a case by case basis. The data and discussions presented here are provided in order that such analyses are based on realistic ‘design fires’ and realistic assessments of the performance of protection systems under consideration. In many rural and urban parts of the world, tunnels provide essential, efficient and sustainable transportation links. We can make them safer.