The use of water-based fire suppression systems or fixed firefighting systems (FFFS) in road tunnels has attracted much attention in recent years and regulations and standards are changing accordingly.
In Sweden, several projects have been planned with FFFS. The use of FFFS in road tunnels is related to zone-operated, water-based deluge systems. Activation is triggered either by an alarm system or by visual or detectable observations in CCTV.
If a thermal activation device is incorporated, the system is usually called an ‘automatic sprinkler system’, to use the same terminology as conventional sprinkler head manufacturers. Automatic sprinklers are common in commercial buildings, hotels and industrial premises but not in tunnels. They are generally activated by subsidiary thermal elements, e.g bulbs or links, and thus do not require other activation methods. In practice, such systems are simple and easy to use.
The main concern with automatic sprinklers is the potentially negative impact of tunnel ventilation on their activation and thus their performance. A previous model-scale study showed that automatic sprinkler systems may fail at high ventilation rates, and therefore their use is only recommended in tunnels with low ventilation (Li and Ingason, 2013).
The use of automatic sprinklers in low-ventilation tunnel conditions is attractive due to its simplicity. In 2016, the authors carried out a full-scale automatic sprinkler test in the Runehamar tunnel which showed the potential for their use in tunnels with relatively low-flow velocities, less than or equal to 2m/s (Ingason et al., 2016). The fuel load consisted of a heavy goods vehicle (HGV) mock-up made of wood pallets.
Feasibility pre-studies on installing automatic sprinklers in an existing underwater road tunnel in Sweden were carried out by Arvidson and Vylund 2017, and Li and Ingason, 2017. They found that the range of longitudinal ventilation varied between -1 m/s to +2 m/s, depending on direction. The study proposed that automatic sprinklers should be installed along the tunnel centre-line but questions such as the influence of initial operational water pressure, activation temperatures and longitudinal ventilation on a number of activated sprinklers remained to be answered.
Water distribution tests were carried out using commercial sprinkler nozzles (Arvidson and Vylund, 2017). This was to determine water distribution characteristics using different water pressures and sprinkler spacing with two selected ‘extended coverage’ (EC) upright sprinklers. The test set up simulated an HGV trailer positioned inside the tunnel. The test was carried out in the RISE fire laboratory with the HGV trailer assumed to be positioned close to the tunnel’s sidewalk. The rationale was that an HGV on fire would be parked as far as possible to the side of the road by an alert driver. A platform with a horizontal plane that measured 5m(L) by 2.6m(W) was constructed to simulate the top of a trailer. It was found that the Tyco EC-14 sprinkler can be installed at a spacing of 5m in the tunnel. For a design density of 10mm/min (l/m2/min), the operating pressure would be around 3.6bar.
As it was seen to be plausible to install automatic sprinklers in a tunnel in Sweden, it was decided to perform additional verification. Promising results were observed in 1:15 model scale tests (Li and Ingason, 2013), full-scale tests (Ingason et al., 2016) and the pre-study with numerical simulations (Arvidson and Vylund, 2017, Li and Ingason, 2017), therefore it was deemed prudent to do more tests using a 1:3 scale model.
The authors had previously carried out numerous fire tests using water-spray systems at different scales, 1:20, 1:15, 1:8 and 1:4 (Ingason et al., 2015). Carrying out full-scale tests was not possible due to the high costs so a 1:3 scale study was thought could provide a realistic simulation and was found to be appropriate. It also made it possible to achieve more accurate scaling of the sprinkler head, the thermal response device, the wood pallet fuel and the thermal response of the tunnel. A summary of the results follows.
Test Results Using a Scale-Model Tunnel
The test tunnel had a ceiling height of 1.7m (5.1m full-scale equivalent), a width of 2.35m (7.05m) and a length of 49.3m (148m). Details of the test tunnel including all test data can be found in the technical report (Ingason et al., 2019) with further analysis in Ingason et al., 2020.
To generate air-flows inside the tunnel, two fans were placed on a platform 2m away from the portal. In order to make the air-flow inside the tunnel more evenly distributed, a barrier to smoothen the flow was built using vertical gypsum boards with known porosity.
In figure 1, a photo of the fire source using scaled wood pallets is shown. The model tunnel was designed with two lanes, with the fuel placed on the centre line of one lane. Located 18.4m from the west portal, the fire source comprised 72 wooden pallets (figure 1) and is a fuel mock-up often used to simulate the payload of an HGV trailer. An uncovered target, comprising a pile of 12 wooden pallets was positioned 1.67m from the rear of the fuel mock-up in order to evaluate the risk of fire spread. The wood pallets were placed on lightweight concrete slabs.
In most of the tests, both the front side (upstream) and back side (downstream) of the fire source were covered with vertical steel plates, as was the area above the pallets using a horizontal steel plate. This arrangement made it difficult for water to directly penetrate the pallets, reducing the ability of the system to fight the fire from above and increasing the rigorousness of the test.
The upstream vertical steel plate at the fire source works as a wind break or wind barrier in relation to the longitudinal ventilation flow.
As a part of the test programme, a passenger car mock-up containing combustible plastic, wood, oil and electrical compounds was also used. The simulated total energy corresponded to an ordinary car.
In total, 28 upward model-scale sprinkler nozzles were installed at 1.67m centres (5m centres at full scale) along the centre line of the tunnel roof, 100mm below the gypsum-boarded tunnel ceiling. The pre-study (Arvidson and Vylund, 2017) used a full-size Tyco EC14 bulb sprinkler. However for the tests, a 1:3 scale 3D-printed titanium sprinkler was used.
The sprinkler supply system comprised a 37mm feeder pipe and 28 manifolds (12mm diameter) one connected to each sprinkler. Nozzles were at 1.67m centres. Assuming all sprinklers are activated, the maximum water flow varies depending on the water density in mm/min (l/min m2). An overview of the flow conditions is given in table 1 for each of the water densities tested. The values in brackets are corresponding full-scale values.
The automatic activation system of the sprinkler nozzles used in the test was technically very advanced and based on K-type sheathed 1.5mm thermocouples, regulator units and solenoid valves at each manifold connected to the sprinkler from the main feed pipe. The activation was therefore not based on real thermal devices such as bulbs or links. The thermocouple corresponded to an RTI value of 14 m1/2 s1/2 according to a prior wind-tunnel test, corresponding to 32 m1/2 s1/2 in full-scale. This RTI value is the same as for a 3mm glass bulb. Both the temperature rise and the activation time were logged during the experiments.
Each individual sprinkler would be activated when the temperature reached the activation temperature (68°C, 93°C or 141°C). All were connected to control and measuring boxes and activation times were logged and registered.
Undertaking 16 Tests
Table 2 shows the test sequence and physical parameters used in the tests. The numbers in brackets are corresponding full-scale values. The ventilation velocity was 1.2m/s in most of the tests, and in some tests, it was either 0.8m/s or 1.7m/s.
Corresponding velocities from the lowest to the highest were 1.4, 2.1 and 3m/s in full-scale. The sprinkler activation temperature was 93°C or 141°C for most tests, but 68°C for Test 5. Water density was varied within 2.9, 4.3, 5.8 and 8.7mm/min. In test 9, the upstream wind break (vertical steel plate on the upstream side) was removed. The total number of activated sprinklers in the experiment is shown, which clearly shows the effect of operational conditions on the sprinkler suppression performance.
Promising Results
Analyses of experimental results focus on the effect of ventilation velocity, sprinkler activation temperature, water-flow density and the presence of a wind break on the fire development of the HGV fuel load. The first activated sprinkler was at position x=-0.37m for Test 1, 2, 4, 5, 6, 7, 9, 10, 11, and at position x=1.3m for Test 3 and 8. Position x=0 is at the centre of the fire load and minus indicates being upstream of the fire. The first activated sprinkler does not show clear dependence on the ventilation velocity studied, but is always located close to the fire source. The passenger car (PB) did not show more than one activated sprinkler in the vicinity of the fire. In the following, a short summary of the tests with the HGV fire load (WP) is presented in tables 3 and 4, depending on the test conditions.
Effect Of Ventilation Velocity On Activation
A summary of the main effects of ventilation velocity on fire development is shown in table 3. The maximum temperature shown refers to the maximum value measured at position x=1.3m (spr11).
The tests show that the effect of ventilation velocity on fire development is limited. From table 3 it is clear that the water density is affecting a number of activated sprinklers. The ventilation velocity is not systematically affecting the number of activated sprinklers and the activation time of the first sprinkler appears not be clearly affected by the increased longitudinal ventilation velocity. The only conclusive result here is the effect of water density on the number of activated sprinklers. The influence of other parameters varies in the range of uncertainty found in the performance of this type of tests.
Effect Of Temperature On Activation
A summary of the main effects of the activation temperature on the fire development is shown in table 4.
Table 4 highlights that the number of sprinklers activated is not as sensitive to the activation temperature and shows the same tendency as in table 3, namely that water density is the governing parameter for the number of sprinklers activated. The results above show that the effects of water density are conclusive whereas the effects of other parameters are more unclear. Using 68°C yield, the fastest activation with the same RTI value is used. Between 93°C or 141°C it was hard to see any difference in the results, and thus either value could be selected as the best option. A lower temperature can help the system to activate earlier but the tests show no significant improvement on suppression performance, and it may even make the system too sensitive. In addition, if too many sprinklers are activated, the system may not work properly. On the other hand, a higher temperature makes the system harder to activate. A value of 93°C is therefore recommended for the investigated tunnel.
Effect Of Water Density
As the water flow increases, the number of activated sprinklers decreases and the heat release rate (HRR) also decreases. From tables 3 and 4, one can see that when water density increases from 4.3mm/min to 5.8mm/min, the maximum HRR falls from 1.63MW to 1.16 MW and the number of activated sprinklers from 12 to 5. Based on the test data that is presented here, the water density of the system is of crucial importance and is therefore recommended to be equal to, or greater than, 10mm/ min at full scale.
Presence Of Wind Break
The HRRs in test 4 had wind break, those in test 8 were without. Twelve sprinklers were activated in test 4, compared with only 8 in test 8. The maximum temperature was 155°C in test 4 and 210°C in test 8, but both dropped to a very low value after activation. Visibility quickly dropped to zero after three minutes in both tests. According to the limited test data, one may conclude that the removal of wind break in test 8 has a limited effect on the test results.
Conclusions
Early activation of automatic sprinklers in tunnels is of the utmost importance. Based on the model-scale tests performed, the activation time of the first sprinkler is found to be around 2.8-3.8 minutes (full scale 4.9-6.6 minutes), when the fire size is in a range of 0.1-0.2MW (full scale 1.6-3.1MW). In one typical case investigated where the fire was on the right side of a full-scale tunnel, the estimated tunnel velocity remains at 1m/s for the first five minutes and gradually increases to 2m/s after 10 minutes (Li and Ingason, 2017). Under such conditions, ventilation flow has limited effect on system activation. The passenger car activated only one sprinkler in the cases that were tested.
Activation temperature plays an important role on the system. Using 68°C yields the fastest activation with the RIT value used, but there is no consistency found in the results concerning 93°C or 141°C, and thus either value could be selected. A lower temperature can help the system to activate earlier but the tests show no significant improvement on the suppression performance, and it may even make the system too sensitive. In addition, if too many sprinklers are activated, the system may not work properly. On the other hand, a higher temperature makes the system harder to activate. A value of 93°C is therefore recommended for the investigated tunnel.
Water density is of crucial importance and is recommended to be equal to, or greater than 10 mm/min for the investigated tunnel. A lower water density can undermine the system suppression performance while only a small improvement is evident when adopting a greater water density.
The tests suggest that automatic sprinkler systems in full-scale tunnels with low ventilation rates (about 2m/s or less in full-scale) can be expected to perform relatively well. The model-scale technique used in the parametric study appears to have worked very well and shows the appropriateness of using this type of technique in future projects.
