Full-Scale Experimental Investigation on the Thermal Control of a Water Mist System in a Road Tunnel Under Single-Source and Double-Source Fire Scenarios

Abstract

This study investigates the thermal control effect of a water mist fire-extinguishing system in road tunnels under both single-source and double-source fire scenarios. A total of eight full-scale fire tests were executed in a physical tunnel, and the double-source fire scenarios were further subdivided into two spatial configurations, including fire sources close together and fire sources with a center-to-center distance of 2 m. During the fire tests, the evolution of fire, temporal and spatial temperature distributions of the tunnel ceiling, longitudinal and vertical temperature gradients, and smoke behavior within the tunnel were systematically recorded and interpreted. The results demonstrate that early activation of the water mist system optimizes its physicochemical mechanisms by preventing the transition from the growth phase of fire to a stable phase. In single-source fire scenarios, the water mist directly suppresses the flame and eliminates the high-temperature core, leading to a significant alteration in the vertical temperature gradient. For double-source fire scenarios, the high-temperature region on the ceiling is reduced upon the application of the water mist. However, when the fire sources are positioned in close proximity, they tend to merge into a larger fire source, with the water mist proving insufficient to prevent this fusion. Conversely, when the center-to-center distance between the fire sources is 2 m, the water mist effectively separates the sources, blocking thermal feedback between them and forcing the flames to develop vertically. This, in turn, accelerates the attenuation of the fire and the recovery of the ambient temperature. Additionally, within the effective coverage of the water mist, the longitudinal temperature distribution on the tunnel ceiling still follows an exponential attenuation pattern, with a significantly high rate of temperature decline.

1. Introduction

As a crucial component of modern transportation infrastructure, road tunnels play a pivotal role in overcoming topographic barriers, optimizing road network structures, and enhancing transportation efficiency [1,2]. However, with the continuous expansion of tunnel construction, fire safety issues have become an increasingly prominent concern, particularly regarding the potential threat that fires pose to human life, property, and transportation systems [3,4,5]. Tunnel fires exhibit distinct characteristics [6,7], and the types of fire sources in tunnels are diverse, including lithium battery fire [8], solid thermal insulation material fire [9], and liquid fire [8]. The gathering of high-temperature and poisonous smoke in the confined environment not only creates a thermal radiation barrier, directly endangering the lives of trapped individuals, but also significantly compromises the structural integrity of the tunnel [10,11]. Particularly, under the influence of longitudinal ventilation, an initial fire source triggered by a vehicle collision can cause the flames to tilt, thereby accelerating the spread and dispersion of high-temperature smoke, which can rapidly evolve into a complex and catastrophic scenario involving multiple fire sources [12,13]. The progression from a single-source fire to double-source or even multi-source fires can lead to an exponential rise in the heat release rate of fire, significantly complicating firefighting efforts and intensifying the overall damage [14,15]. A notable example occurred during the Futuyu 5 tunnel fire in China in 2017, when a hazardous chemical transport vehicle caught fire and ignited five surrounding coal trucks, resulting in 12 fatalities and damage to 9 vehicles [16]. This tragic incident starkly highlighted the destructive potential of multi-source fires in road tunnels.

Tunnel fires often give rise to disastrous outcomes due to the confined space of tunnels [17]. In the absence of effective fire suppression and smoke extraction systems, the rapid accumulation of poisonous combustion products at high temperatures presents a severe threat to human lives [18,19,20]. The concept of water-based fixed firefighting systems (WFFFSs) dates back over 60 years, and the design objectives of WFFFSs focused on three main functions, namely, fire scale control, spread retardation, and tunnel structure protection [21]. At present, many countries and regions are contemplating the implementation of WFFFSs in newly constructed tunnels, with some already having a substantial history of their application [22] (Häggkvist, 2009). In the Oceania region, New Zealand and Australia are actively working towards developing world-leading WFFFSs for tunnels [23] (Ernst et al., 2018). In Europe, tunnels with WFFFSs can be found in Austria, France, Italy, the Netherlands, Norway, Spain, and Sweden [24] (Arvidson, 2004). As an advanced form of this system, the water mist system (WMS) offers significant technical advantages by optimizing the droplet size distribution, enhancing heat exchange efficiency between water droplets and high-temperature gases or flame plumes, and achieving efficient water resource utilization [25,26]. In recent years, considerable research has been conducted on the use of water mist (WM) for fire suppression [27,28,29]. Ingason [30] reconstructed a truck fire scenario and confirmed that early activation of the sprinkler system was crucial in limiting fire spread and reducing the smoke temperature. Zhang et al. [31] conducted experiments on the interaction between fire and WM in a narrow and confined space, revealing a dual extinguishing mechanism involving radiant heat flux decline and fuel surface cooling. Liang et al. [32] further validated that flame cooling and fuel surface cooling are the core mechanisms driving the effectiveness of WM in extinguishing oil pool and PVC fires. Additionally, Li et al. [33] systematically assessed the effects of key parameters on the fire control efficiency of WMSs through a model test in a 1:4 scale tunnel. While existing studies have thoroughly demonstrated the excellent performance of WM in single-source fire scenarios, significant gaps remain in understanding its fire-extinguishing efficiency and fire evolution characteristics in double-source fire scenarios.

In contrast, Liu et al. [34] investigated the flame interaction of two parallel rectangular fires and provided a physical interpretation of the mechanism of flame interaction. Meng et al. [35] studied the temperature of smoke flow beneath the tunnel ceiling induced by two fires in a naturally ventilated tunnel, and they established a formula to predict the longitudinal temperature distribution. The characteristics of double-source fires differ significantly from those of single-source fires due to their unique physical coupling mechanisms, primarily manifested in two key interaction processes [34,35,36,37]. The first is thermal feedback strengthening, where two fire sources exchange heat through radiation and convection, resulting in a substantial increase in combustion intensity. The second is the confined air entrainment effect, which results in phenomena such as flame tilting and plume merging because of the limited air entrainment caused by the spatial proximity of the fire sources. Together, these coupling mechanisms govern the distinct dynamic evolution of double-source fires. In research on tunnel fires, full-scale tests are considered a crucial basis for validating small-scale test results and guiding fire protection design, as they provide a more accurate reflection of actual fire behavior [38,39,40]. Despite challenges such as high costs and operational complexity, significant progress has been made through full-scale fire tests [27,41,42,43]. For instance, Ingason et al. [43] conducted tests on a firefighting system in the Runehamar Tunnel, demonstrating that WFFFSs effectively reduced the heat release rate from 75–100 MW under free combustion conditions to 20–45 MW. Correspondingly, Chang et al. [42] successfully verified the significant suppression effect on the ceiling temperature and radiant heat flux in a WMS test conducted in the Dong-Aw Tunnel. It is important to note that most existing full-scale experiments have focused on single-source fires, and no systematic experimental framework has been established for the more complex and realistic scenario of double-source fires. This research gap hinders the advancement of tunnel fire theory, underscoring the need for innovative full-scale experimental designs.

In this study, a WMS was installed in a physical tunnel, and two distinct fire scenarios were designed based on the fire characteristics of road tunnels. The double-source fire scenario was further subdivided into two spatial configurations, including fire sources that were close together and fire sources with a center-to-center distance of 2 m. By analyzing and comparing the fire evolution process, the temporal and spatial distribution of tunnel ceiling temperatures, the distribution pattern of the longitudinal temperature, the overall spatial temperature distribution within the tunnel, and vertical temperature gradients, the study investigates the thermal control effects of the WMS in different fire scenarios in road tunnels.

2. Description of the Fire Tests

2.1. Test Tunnel

Multiple full-scale fire tests were executed in the Longchi Tunnel (LC Tunnel) located in Sichuan, China. The LC Tunnel, with an overall length of approximately 1100 m, was abandoned upon completion and is currently repurposed as a test tunnel. As exhibited in Figure 1, the tunnel is a two-lane structure with an internal width of 9.45 m and a clear height of 7 m. The inlet section, spanning 300 m, features an asphalt pavement, while the remaining sections are constructed with concrete. The side walls of the LC Tunnel are also made of concrete, and LED luminaires are installed along the tunnel ceiling to provide illumination.

2.2. Water Mist System

The WMS utilized in this study consists of a high-pressure pump unit, intelligent control valves, a pipe network system, and atomizing nozzles. The spatial configuration of these components is depicted in Figure 2. The WMS in the test section extends 12 m longitudinally along the tunnel, with the pipe network installed using an embedded method along the tunnel’s vaulted ceiling, ensuring symmetrical distribution. Based on design principles of optimal fluid dynamics, the spacing between the main pipeline and the nozzles on each pipe is set at 2 m, forming a three-dimensional coverage network. A total of 35 atomizing nozzles are configured (see Figure 2 for the specific geometric parameters). The atomizing nozzles feature a porous swirl structure (as displayed in Figure 2), designed to produce a wide-angle cone spray (see Figure 3 for the fog field morphology). Pressure–flow calibration experiments confirmed that the flow rate of a single nozzle remains stable at 12.25 L/min under a working pressure of 0.8 MPa. Droplet size distribution characteristics were evaluated with a laser diffraction particle size analyzer (Malvern Mastersizer 3000), with a 99% droplet diameter (D_v99) ≤ 390 μm and a median volume diameter (D_v50) of 193 μm.

2.3. Fire Source

In this study, an oil pan fire was employed as the experimental fire source, using diesel as the fuel. Diesel, known for its high smoke and dust generation characteristics, was chosen to facilitate optical observation of flame morphology [16,28,44,45]. Although real tunnel fires may involve various fuels, including gasoline or battery materials, diesel is widely used in full-scale fire experiments due to its stable combustion characteristics and prevalence in both passenger and commercial vehicles. The combustion device was constructed from high-temperature-resistant stainless steel (3 mm plate thickness), with the diesel fuel contained within the combustion plate. The heat release rate (HRR) from an oil pan fire is primarily dependent on the burning area, and the fire intensity can be regulated by adjusting the size of the burning area. As exhibited in Figure 4a, the geometric parameters of the oil pan are 1.21 m (L) × 1.21 m (W) × 0.15 m (H).

A high-precision weighing system was employed to monitor the mass change of fuel during the free combustion (natural ventilation and no water mist) of a single oil pan in real time, enabling the calculation of the HRR for the oil pan under actual combustion conditions. The system consists of four pressure sensors with a measurement precision of 0.1 kg, recording data at 1 s intervals. The weighing device is positioned at the bottom of the oil pan and is insulated from the pressure sensors by a fireproof plate, ensuring that the high-temperature environment does not affect the measurement accuracy. The HRR ($\dot{Q}$) for the free combustion of a single oil pan is calculated from the fuel mass loss rate, as illustrated in Equation (1).

$$\dot{Q}=\chi \dot{m}\Delta H_c$$

(1)

Here, χ represents the combustion efficiency, which is assumed to be 0.75 [46]; $\dot{m}$ is the real-time mass loss rate (kg/s); ΔH_c is the calorific value of combustion, with a value of 43 MJ/kg for diesel [45].

Based on the real-time fuel mass loss archived by the weighing system and Equation (1), the HRR curve for the fire source of a single oil pan is exhibited in Figure 5. The mass loss rate of fuel remains relatively stable during the steady combustion period of the fire source, which is referred to as the quasi-steady state [41]. The average HRR ($\overline{\dot{Q}}$) of the single oil pan is calculated using the average fuel mass loss rate ($\overline{\dot{m}}$) during the quasi-steady state and Equation (2), yielding an average HRR of 1.42 MW. This value can be used to characterize the scale of the fire test.

$$\overline{\dot{Q}}=\chi \overline{\dot{m}}\Delta H_c$$

(2)

In this fire test, two oil pans of identical size were utilized. The experimental design considered a parallel arrangement of the two fire sources, with fire scenarios created by adjusting the spacing and number of oil pans to form different combustion areas and fire configurations. Figure 4b illustrates the combustion scenario where the two oil pans are positioned close together, forming a larger fire source, while Figure 4c depicts the scenario where the oil pans are separated with a center-to-center distance of 2 m (S = 2 m), creating a double-source fire. This setup effectively simulates both single-source and double-source fire combustion phenomena, which could be commonly observed in tunnel fires.

2.4. Test Methods and Test Conditions

To comprehensively record the entire process of the fire test, high-definition cameras were utilized to capture footage from both longitudinal and horizontal perspectives. As displayed in Figure 6, one camera was positioned outside the coverage area of the WMS, primarily to monitor the overall development of the fire. Another camera was placed near the fire source to precisely document the flame morphology and its evolution throughout the combustion process. Additionally, smoke concentration detectors were installed on the tunnel ceiling, 50 m upstream and downstream of the fire source, to observe and compare changes in smoke concentration before and post the activation of the WMS.

Temperature measurements in the tunnel were conducted using K-type armored thermocouples (WRNT-35), which possess a measurement range of up to 600 °C, a precision of 0.1 °C, and a data acquisition frequency of 1 s. A total of 31 measuring points were evenly distributed along the longitudinal direction beneath the tunnel ceiling, with a 2 m spacing between adjacent points and a vertical distance of 0.05 m between the probes and the tunnel ceiling. To obtain a more comprehensive temperature distribution, an additional 27 measurement points were installed at heights of approximately 4 m, 5 m, and 6 m above the ground. This brought the total number of thermocouples in the tunnel to 58. The specific distribution of temperature measurement points is detailed in Figure 6.

All tests were conducted under natural ventilation conditions, with longitudinal wind speeds ranging from 0.3 to 0.8 m/s, and the ambient temperature was maintained between 6 and 10 °C. The same test methods and instruments were used across all conditions. A total of 8 sets of fire test conditions were designed, divided into two series (single fire source and double fire source). The single-fire-source test group (T1–T4) focused on evaluating the fire extinguishing efficiency of the WMS and the effect of its start-up time on fire development. T1 served as the control group with no WM, while T2–T4 were configured with varying start-up times for the WMS.

For single-fire-source scenarios, the WMS activation time was varied at 1 min, 2 min, and 3 min. The baseline of 1 min aligns with the requirement of the Chinese standard “Technical Requirements for Highway Tunnel Fire Alarm Equipment (JTT 610-2016)” [47], which mandates fire alarm activation within 60 s. The longer activation times (2 min and 3 min) were investigated to assess the system’s effectiveness under progressively more severe fire development conditions, where intervention is delayed. In contrast, for the double-fire-source scenarios, only the most severe case (a WMS activation delay of 3 min) was considered. This represents the worst-case scenario for fire control, allowing us to rigorously evaluate system performance under the most challenging operational constraints.

The double-fire-source experimental group included two spatial configurations. T5–T6 featured fire sources arranged in close proximity (with a center spacing (S) of 1.21 m), while T7–T8 had fire sources set 2 m apart (S = 2 m). To analyze the interaction between the double fire sources and the water mist system, each double-fire-source group included tests with and without WM. Specifically, T5 and T7 represented the control conditions without WM, while T6 and T8 involved the action of the WMS. The details of the 8 test conditions are summarized in

Table 1. Test conditions.

Code	Fire	Mass of Fuel	State of WMS	Description	Appearance
T1	Single fire source	9 kg	Off	-	Figure 4a
T2	Single fire source	9 kg	1 min on	-	Figure 4a
T3	Single fire source	9 kg	2 min on	-	Figure 4a
T4	Single fire source	9 kg	3 min on	-	Figure 4a
T5	Double fire sources	18 kg	Off	Close together (S = 1.21 m)	Figure 4b
T6	Double fire sources	18 kg	3 min on	Close together (S = 1.21 m)	Figure 4b
T7	Double fire sources	18 kg	Off	S = 2 m	Figure 4c
T8	Double fire sources	18 kg	3 min on	S = 2 m	Figure 4c

.

3. Results

3.1. Process of Fire Evolution

3.1.1. Different Activation Times of the WMS Under Single-Source Fire Scenarios

Figure 7 illustrates the effects of different activation times of the WMS on the fire evolution and extinguishing efficiency in the scenario of a single fire source. In T1, the fire source burned for more than 10 min without the intervention of the WMS. During the first 4 min, the flame height rose rapidly, accompanied by significant smoke spread. After 5 min, the fire began to weaken, and the flame tilted due to the influence of natural wind. In T2, the WMS was activated at 1 min, while the fire was still in its early growth stage. The WM quickly cooled the core area of the flame, inhibited its rise, significantly reduced smoke generation, and extinguished the fire within 4 min. In T3, the WMS was activated at 2 min, when the fire was nearing a stable combustion stage, with a layer of hot smoke forming at the top of the tunnel. The WM had to penetrate this layer to reach the fuel surface. Additionally, the high-temperature environment accelerated the evaporation of droplets, reducing the effectiveness of the WM coverage. Consequently, the fire extinguishment time increased to just under 6 min. In T4, the WMS was activated at 3 min, by which time the HRR of the fire source had reached its peak. Although the WMS suppressed the flame temporarily, the fire intensified after 6 min. Finally, the fire was extinguished after 8 min.

The overall trend indicates that earlier activation of the WMS leads to more effective extinguishing. Early intervention (within 1 min) maximizes the physical and chemical effects of WM, preventing the fire from transitioning into the stable combustion stage. Once the fire approaches or enters the stable combustion stage, factors such as thermal intensity, the resistance of the smoke layer, and the fuel state significantly complicate extinguishing efforts. This finding provides critical guidance for tunnel fire prevention and control strategies. Enhancing the sensitivity of fire detection systems and optimizing the response speed of the WMS are key to improving the success rate of tunnel firefighting.

3.1.2. Different Spatial Configurations in Double-Source Fire Scenarios

Figure 8 and Figure 9 illustrate the fire-extinguishing process of the WMS in the double-source fire scenario. Figure 8 exhibits a configuration where the two oil pans are placed close together, representing the test conditions of T5 and T6. In T5, without WM, the merging of the two flames results in a significant rise in the HRR, and the fire continues to burn for over 10 min. In T6, the WMS was activated at 3 min. Although the WM does not fully prevent the initial flame merger, it weakens the combustion intensity, disrupting the thermal feedback loop. This gradual decoupling of the flames ultimately leads to extinguishment within 6 min. This highlights that the extinguishing process for double fire sources involves two stages. First, the WM disrupts the thermal equilibrium, and then it isolates the fire source to extinguish the flames.

Figure 9 illustrates a scenario of the double fire sources with a center-to-center distance of 2 m, corresponding to the test conditions of T7 and T8. In T7, without WM, a low-pressure area was created by the hot plume from the burning double fire sources, causing the phenomenon of “flame attraction”. This results in the flames leaning toward the center, and the fire continues for over 10 min. Initially, the flames do not show significant attenuation due to the attraction phenomenon, but the fire weakens significantly after 8 min. In T8, the WMS was activated at 3 min. The wide coverage of the WM blocked the thermal radiation between the two fire sources, effectively cutting off the connection between them and eliminating the flame attraction effect. The two fire sources were then separated by the WM, and one of the fire sources was extinguished first. The variation in the fire behavior may be due to factors such as natural wind speed and direction, which could cause uneven distribution of the WM droplets.

In general, when the double fire sources are arranged close together, they merge into a larger fire source, and the WM is unable to prevent this fusion. However, when there is a certain distance (S = 2 m) between the two oil pans, the WM can effectively separate the double fire sources and extinguish the flames.

3.2. Spatiotemporal Distribution of Tunnel Ceiling Temperature

3.2.1. Single-Source Fire Scenarios

The 31 thermocouples arranged on the tunnel ceiling recorded real-time temperature changes along the longitudinal range above the fire source. Figure 10 reveals the impact of the WMS activation timing on the temporal and spatial distribution of the ceiling temperature in the single-source fire scenario.

In T1, without WM, the ceiling temperature gradually increased after the ignition of the fire source. Under the influence of natural wind, the temperature field became asymmetrically distributed, with the peak temperature shifting from directly above the flame (0 m) to 2 m away. At 200 s after ignition, the peak temperature of the ceiling reached 65 °C, which remained stable for 130 s. However, the temperature continues to rise in the area extending 30 m on both sides. After 330 s, the temperature in the core high-temperature zone began to decrease, while the temperature on both sides remained relatively stable. In T2, where the WMS was activated at 1 min, the ceiling temperature peaked at 22 °C at 200 s, significantly lower than in T1. Notably, there is a clear temperature gradient inside and outside the effective WM zone, with a bimodal temperature distribution under the influence of natural wind. The monitoring data reveal that the ceiling temperature on the side of the fire source (from −30 m to 0 m) is generally lower than on the opposite side (0 m to 30 m), demonstrating the disturbance effect of natural wind on the temperature field. Conditions T3 and T4 (WMS activated at 3 min) show similar temperature trends. In T3, the peak ceiling temperature was 45 °C at 150 s, while in T4, it reached 48 °C at 200 s. After these peaks, the temperature decreased in the areas affected by the WM, although there was a brief temperature increase at 400 s. The temperature in the area outside the WM zone remained around 20 °C, highlighting the cooling effect of the WM on the high-temperature smoke. In T4, the ceiling temperature distribution on both sides of the fire source becomes asymmetrical, further indicating the influence of natural wind on the temperature field.

A comparative analysis reveals that when the WMS is activated after the ignition source has reached or entered the stable combustion phase (as in T3 and T4), the influence of the start-up timing on the temperature field becomes less significant. This suggests that the timing of the WM intervention in the early stages of the fire is crucial for achieving effective temperature control.

3.2.2. Double-Source Fire Scenarios

Figure 11 compares the spatiotemporal evolution of the tunnel ceiling temperature with or without WM under the scenario of double fire sources close together. In T5, where the WMS is not activated, the temperature distribution pattern of the ceiling is similar to that in the T1 condition. In T5, the peak ceiling temperature rose to 75 °C at 200 s after ignition and remained stable for the next 150 s. Notably, the proximity of double fire sources resulted in the phenomenon of flame merging, shifting the temperature peak from the initial 0 m position to 2 m. In T6, before the WMS was activated, the ceiling temperature evolution from 0 to 180 s closely mirrored that of T5. After activation of the WMS, the area of high temperature on the ceiling was significantly reduced compared to T5, and the duration of high temperatures was shortened. However, due to the blocking effect of WM on smoke spread, a high-temperature smoke accumulation occurs above the flame, causing the peak temperature to increase instead of decrease, reaching a maximum of 98 °C. Thanks to the combined cooling and washing effects of the WM, as well as its physical barrier to the high-temperature smoke, the temperature outside the effective range of the WMS remains below 32 °C. Despite this, the temperature remains much higher than that in the single fire source of T4, primarily due to the increased total HRR associated with double fire sources.

In accordance with the spatiotemporal distribution characteristics of tunnel ceiling temperature in the scenario of double fire sources (S = 2 m), as displayed in Figure 12, the thermal evolution with or without WM is compared and analyzed. In T7, where the WMS is not activated, the ceiling temperature distribution follows a pattern similar to that in T5. Notably, the double fire sources did not form a bimodal temperature field due to the flame fusion effect. Instead, the temperature peak shifted to a distance of 4 m from the fire source, with a maximum temperature of 70 °C. In T8, where the WMS was activated at 3 min, the system demonstrates a significant temperature control effect. Prior to activation of WMS, the ceiling temperature distribution exhibited a bimodal characteristic, which was notably different from T6. This difference is likely due to the interference of the natural wind field on the flame fusion process. After the WMS was activated, there was a brief rise in temperature directly above the fire source, followed by a rapid attenuation of the temperature gradient. Affected by the asymmetry of the natural wind field, the temperature distribution on the ceiling outside the coverage area of WM is uneven. The temperature on the windward side remained stable below 25 °C, while on the leeward side, it generally stayed below 35 °C.

A comparative study of T6 and T8 displays that when the distance between the double fire sources is reduced, the high-temperature area on the ceiling exhibits a clear spatial shrinkage effect. However, this phenomenon is accompanied by an increase in peak temperature and a prolonged duration of high temperatures.

3.3. Longitudinal Distribution Pattern of Tunnel Ceiling Temperature

Throughout the development of the tunnel fire tests, the fire plume engendered by the oil pan fire underwent a constant temperature decay process as it diffused longitudinally. This temperature decrease is attributed to the continuous thermal transfer between the smoke layer and the tunnel walls. When the WMS intervened, this temperature attenuation mechanism was enhanced by the heat absorption effect from the phase change evaporation of the WM, which added a composite cooling effect to the smoke layer. For scenarios with a single fire source, many scholars [48,49,50] have conducted fire tests at various scales, including both scaled-down model tests and full-scale tests, to examine the distribution of smoke temperature along the longitudinal direction of the tunnel ceiling. These studies revealed an exponential attenuation of the smoke temperature. In the case of double-source fires, several researchers [51,52] have verified that the longitudinal distribution of smoke temperature rise in the downstream zone of the tunnel ceiling also follows an exponential attenuation pattern.

To explore the longitudinal distribution pattern of the tunnel ceiling temperature in fire tests, this study conducts a comparative analysis between scenarios without WM (T1, T5, T7) and those with WM (T4, T6, T8). Considering both the thermal characteristics during stable fire combustion and the effective intervention time of WM, the average ceiling temperature from 200 s to 250 s after ignition was specifically selected for investigation. Based on fire plume theory [53], to better elucidate the temperature attenuation along the tunnel ceiling, the average temperature data in downstream ceiling regions during the 200–250 s interval for six experimental configurations (T1, T4, T5, T6, T7, T8) were normalized using dimensionless parameters. Through the introduction of a dimensionless longitudinal distance parameter, the exponential decay model depicted in Equation (3) was employed for fitting analysis.

$${\displaystyle \frac{\Delta T}{\Delta T_{max}}}=A{e}^{-B\frac{x-x_0}{H_c}}$$

(3)

where ΔT represents the temperature difference between the ceiling at the longitudinal position of x and the ambient environment; ΔT_max represents the maximum temperature difference between the ceiling and ambient environment; x₀ indicates the longitudinal location where the maximum ceiling temperature occurs; H_c corresponds to the clear height of the test tunnel; A and B are empirical constants.

Figure 13 demonstrates the dimensionless relationship between the average ceiling temperature and distance from the fire source under different experimental configurations. As depicted in Figure 13a, in scenarios without WM (T1, T5, T7), both single-source and double-source fires exhibit pronounced exponential decay patterns in average ceiling temperature. This phenomenon primarily stems from the thermal similarity between double-source fires (when treated as equivalent integrated sources) and single-source fires in terms of upstream or downstream heat transfer processes of smoke flow. Notably, the introduction of the WMS (T4, T6, T8) induces significant alterations in the temperature distribution characteristics, as displayed in Figure 13b. Within the effective coverage of WM, longitudinal temperatures still conform to an exponential decay pattern, but with markedly accelerated attenuation rates. This enhancement is principally attributed to the phase-change evaporation of water droplets that efficiently absorb thermal energy from the smoke layer. Beyond the influence zone of WM, two distinct temperature characteristics emerge. First, transient temperature rebound occurs immediately downstream of the WM coverage boundary, reflecting thermocouple measurements capturing the real-time temperature inside the tunnel space rather than actual smoke temperature. Second, oscillatory decay replaces exponential attenuation along the longitudinal axis, likely influenced by the disturbances of heterogeneous natural airflow within the tunnel and uneven distribution of residual water droplets in smoke. This atypical attenuation profile suggests that WM induces complex secondary effects on the thermodynamic behavior of smoke movement.

3.4. Temperature Distribution Inside the Tunnel Space

Figure 14 illustrates the longitudinal temperature field distribution characteristics in tunnel fire development at 250 s based on temperature data collected from thermocouple measuring points at varying heights. It should be specifically noted that the spatial density of measuring points (4–7 m intervals) limits this temperature distribution diagram to approximating thermal patterns within specific tunnel sections. Comparative analysis of fire scenarios without WM (T1, T5, T7) reveals that the high-temperature zone beneath the ceiling expands with increasing the quantity of fire sources. The thermal plume merging effect proves more pronounced in the T5 (double fire sources close together) than in the T7 (double fire sources with a center-to-center distance of 2 m). Notably, all scenarios exhibit asymmetric temperature distributions due to disturbances from natural wind fields.

In scenarios with WM, distinct temperature regulation performance was observed across configurations. For single-source fire scenarios, T2 restored ambient longitudinal temperatures after 3 min of spraying, while T3 retained localized heating above the fire source after the 2 min intervention. T4 achieved thermal field stabilization within 1 min, preserving residual heat solely above the ignition point. In double-source fire configurations, T6 (fire sources close together) maintained a confined high-temperature zone after 1 min of spraying, which is primarily attributed to the following: (1) the HRR of the combined flames is significantly higher than that of a single fire source, resulting in an increase in the temperature of the smoke under the ceiling; (2) momentum transfer from WM disrupts smoke stratification, causing the re-accumulation of high-temperature smoke under the ceiling. Similarly, T8 (fire sources with a center-to-center distance of 2 m) exhibited residual thermal zones above the fires. The temperature distribution in Figure 14h further reveals that the cooling and disturbance of the WM change the flame fusion characteristics of double fire sources (S = 2 m) and promote the combustion pattern to change from oblique fusion to a vertical development mode. Hence, a high-temperature area appeared in the 4 m height area above the fire source, which was produced by the high temperature of the flame.

A recognized limitation is the absence of direct smoke concentration visualization. While temperature distribution effectively indicates smoke dynamics and high-temperature zones, it cannot fully substitute for quantitative measurements of smoke layer height, optical density, or precise dispersion pathways. Future studies should incorporate laser-based techniques to complement thermal data with direct smoke metrics.

3.5. Vertical Distribution Pattern of Tunnel Temperature

The vertical temperature distribution pattern in tunnel fires was investigated with a focus on temperature variations above the fire source at an elevation of 0 m, primarily based on thermal data recorded by four thermocouple measurement points at this location. Figure 15 illustrates the vertical temperature evolution at an elevation of 0 m under single-source and double-source fire scenarios with and without WM. In T1 (single-source fire without WM), during 100–200 s, the fire intensity peaked with a thermal maximum at a height of 4 m, establishing a descending temperature gradient towards upper tunnel spaces. This thermal core at a height of 4 m originated from direct flame radiation. As combustion progressed (200–300 s), dual thermal cores emerged at heights of 4 m and 7 m, each generating temperature gradients upwards and downwards, respectively. The 4 m temperature remained flame-driven, while the 7 m thermal signature was derived from smoke accumulation. After 300 s, fire decay manifested as a singular downward gradient from 7 m with diminished thermal differentials. In T4 (single-source fire with WM activated at 3 min), flame suppression immediately eliminated the 4 m thermal origin. Residual thermal stratification persisted at 7 m from smoke accumulation, generating downward temperature gradients until 270 s, when thermal stratification collapsed, with all measurement points approaching ambient levels.

For double-source fire scenarios, the flame merging characteristics and temperature field distributions exhibited significant variations under different experimental conditions. In T5 (fire sources close together without WM), flame merging induced a deviation from the vertical development trajectory of the oil pan flame, failing to establish a high-temperature core region at a height of 4 m. During 150–200 s, with the accumulation of the smoke layer under the ceiling, a stable temperature gradient structure formed descending from 7 m, demonstrating a vertical temperature differential of approximately 18 °C/m. In T5 (fire sources close together with WM activated at 3 min), although flame merging persisted between fire sources, momentum-induced disturbances from the WM created discontinuous temperature gradient distributions below 7 m, exhibiting a distinct contrast to the uniform gradient structure observed in T5. When the fire source separation increased to 2 m in T7 (without WM), flame merging phenomena remained detectable. The vertical temperature evolution at 0 m showed similarity to T5, though the increased spacing reduced the temperature elevation amplitude and weakened gradient stability. Conversely, in T8, WM (activated at 3 min) effectively disrupted thermal feedback mechanisms between fire sources, prompting a transition of the flame to vertical development patterns at 0 m. During 220–260 s, a temperature surge at 4 m established a thermal gradient upwards, a phenomenon strongly associated with secondary entrainment effects induced by WM momentum on flame plumes. Notably, sustained action of WM caused the separated fire source to decay rapidly after 270 s, with the temperature of each measuring point returning to ambient temperatures within 320 s.

3.6. Smoke Conditions Inside the Tunnel

This study investigates the diffusion characteristics of toxic gases (CO) and particulate pollutants (PM10) generated by diesel pool fires in the tunnel, comparing pollutant distribution patterns at different monitoring points with and without WM. Monitoring points were established at ceiling positions 50 m upstream and downstream of the fire source. Notably, due to natural wind direction reversal, the +50 m monitoring point in T1 was located downstream of the fire source, whereas in T4, the −50 m monitoring point transitioned to the downstream position relative to the fire source.

Figure 16 demonstrates the regulatory effect of the WMS on PM10 concentrations. Through physical adsorption mechanisms, WM effectively promotes particulate sedimentation, resulting in a general reduction of PM10 concentrations across all measurement points. Notably, in T4, the attenuation rate of PM10 concentrations at the downstream measurement point significantly exceeds that of the upstream point, indicating enhanced particle deposition efficiency along the primary diffusion direction. This phenomenon may be due to the coupling effect between the dominant smoke movement trajectory and the spray track of WM. Figure 17 reveals the chemical intervention effect of WM on the combustion process through CO concentration distribution. As a characteristic product of incomplete combustion, CO concentration variations indirectly reflect the inhibitory effect of WM on the combustion rate. The WMS achieves a reduction of approximately 15% in peak CO concentration downstream of the fire source. In T4, natural wind fields induce smoke accumulation in upstream areas, maintaining elevated CO levels, while improved dispersion conditions downstream facilitate rapid attenuation with over 68% reduction of CO concentration within 300 s.

4. Discussion

Observational analysis of fire evolution and temperature distribution in the tunnel demonstrates that the activation time (T1-T4) of the WMS exerts a significant impact on fire suppression effectiveness. Intervention implemented during stable combustion phases not only requires more water consumption but also extends fire suppression duration compared to early-stage deployment. Notably, when the burning area of the fire expands (T5, T6), the HRR exhibits a multiplicative growth trend, accompanied by abrupt temperature escalation that substantially challenges the cooling capacity of the WMS. Furthermore, fire propagation processes may induce multi-source fire scenarios. The conversion from single-source fires to double-source fires (T7, T8) generates interacting plumes that create bimodal temperature distributions, while the resultant nonlinear thermodynamic processes degrade the cooling efficiency of WM. The fire characteristics and thermal evolution of double-source fires under WM exhibit greater complexity, imposing elevated requirements for tunnel firefighting facilities.

Based on these findings, this study proposes that the fire detection response system in road tunnels should be fully optimized through the fusion of multiple sensor sources. Implementing integrated algorithms combining infrared thermal imaging, distributed optical fiber temperature measurement systems, and smoke concentration monitoring could achieve a substantial reduction in fire identification and response time, enabling the activation of WM within 1 min. This early-stage intervention strategy for fire incidents can significantly enhance the efficiency of fire suppression, effectively preventing single-source fires from evolving into double-source or even multi-source fires, thereby substantially mitigating structural damage risks and reducing personnel endangerment hazards. Guo et al. [16] hold the same view that longitudinal ventilation in a tunnel would alter the combustion state of a double-source fire. When the fire source spacing is smaller, there is a possibility of flame fusion between the two fire sources, resulting in a significant increase in the size of the tunnel fire accident.

In view of the evolution of high-temperature zones in tunnel ceilings and the variation in smoke concentration upstream and downstream of fire sources, and considering the potential influence of ambient conditions (such as the natural wind speed range of 0.3–0.8 m/s and temperature range of 6–10 °C in this study) on smoke behavior and WMS efficiency, it is imperative to investigate the integrated application of a WMS with longitudinal ventilation systems to achieve efficient fire suppression and temperature control in future research. The specific environmental conditions tested here may limit the direct generalizability of the results to tunnels with significantly stronger imposed ventilation or higher ambient temperatures. Therefore, under the premise of ensuring effective fire containment by the WMS, longitudinal ventilation should be leveraged to reduce the ambient temperature of the tunnel, inhibit upstream smoke propagation, and accelerate downstream smoke dissipation. Priority should be given to establishing multi-field coupling models that synergize WMSs and ventilation systems, combining full-scale experiments with numerical simulations to quantify the coordinated mechanisms among WM droplet size, spray intensity, ventilation velocity, and ambient temperature. Deng et al. [21] presented a review concerning the development and performance of water-based fixed firefighting systems in road tunnels, and they suggested that future research should focus on an optimal cooperation scheme between water-based fixed firefighting systems and longitudinal ventilation systems. Future research must explicitly explore a wider range of ventilation velocities (including stronger flows) and ambient temperatures to establish robust control strategies. Future research should ultimately develop dynamic regulation models based on real-time monitoring of temperature fields, smoke concentration distributions, ambient conditions (temperature, velocity), and airflow velocity profiles, thereby formulating intelligent coordinated solutions for fire prevention and control inside the tunnel.

5. Conclusions

This study conducted full-scale tunnel experiments involving single and double fire sources with WM. For single-source fire scenarios, the activation time of the WMS was varied. Double-source fire scenarios were further subdivided into two spatial configurations, including fire sources that were close together and fire sources with a center-to-center distance of 2 m. During the experimental process, comprehensive measurements were recorded and analyzed, and they included fire evolution characteristics, the spatiotemporal distribution of the ceiling temperature, longitudinal temperature patterns along the tunnel ceiling, the temperature distribution inside the tunnel space, vertical temperature profiles, and smoke conditions inside the tunnel. The principal findings of this investigation are summarized as follows:

(1)

The investigation of flame evolution with WM demonstrates that early intervention by the WMS (within 1 min) maximizes its physicochemical fire-extinguishing mechanisms by inhibiting the transition of fire development from the growth phase to the stable phase. In double-source fire scenarios, fire sources that are close together coalesce into a larger fire plume, a merged state that WM fails to dismantle. Conversely, when a separation distance (S = 2 m) is maintained between double fires, the WM successfully isolates the double fire sources, achieving effective compartmentalization.

(2)

The influence of WM activation timing on the ceiling temperature diminishes significantly when the activation of the WMS occurs after fire sources approach or enter stable combustion phases. In double-source fire scenarios, WM reduces the coverage of the ceiling high-temperature zone. However, due to the blocking effect of WM on the spread of smoke, thermal accumulation effects emerge above fire plumes, resulting in the peak temperature rising instead of decreasing. Without WM, both single-source and double-source fires exhibit distinct exponential decay patterns in the temperature distribution along the longitudinal ceiling. With the intervention of WM, the longitudinal temperature of the ceiling still conforms to the exponential attenuation law within the effective range of WM, and the attenuation rate is significantly increased.

(3)

The WMS fundamentally alters the temperature distribution of a tunnel fire by disrupting the heat feedback of the flame and perturbing the momentum of the smoke layer. In single-source fire scenarios, direct flame suppression eliminates high-temperature cores, restructuring vertical thermal stratification patterns. For separated double fire sources (S = 2 m), WM blocks the thermal feedback between fire sources, forcing vertical flame development that accelerates the attenuation of the fire source and the recovery of the ambient temperature. Notably, without WM, increased fire separation distances may destabilize vertical temperature gradients through competing plume interactions.

(4)

The WMS plays a pronounced role in regulating the dispersion of fire-related pollutants in the tunnel through a dual mechanism, including adsorption of particulate matter and inhibition of combustion. For PM10, the primary removal mechanism is physical sedimentation, and more effective particulate matter capture can be achieved by aligning the direction of smoke flow with WM. Regarding CO, the peak concentration is reduced by suppressing the combustion rate. However, due to disturbances in natural wind fields, higher CO concentrations may persist upstream of the fire source due to the accumulation of smoke. In contrast, downstream CO concentrations tend to decrease significantly due to the enhanced diffusion and dilution effects of the WM.

(5)

The activation timing of the WMS is closely associated with the dynamic evolution of the fire source. A delay in activation, especially after the fire has reached a stabilized combustion stage, leads to increased water consumption and prolonged fire control times. Additionally, the bimodal temperature distribution of the composite plume, resulting from multiple concurrent fire sources, significantly reduces the cooling efficiency of WM. In practical applications, it is recommended to enable early fire intervention through a multi-parameter fire alarm system integrated with the WMS. Future research should focus on exploring the synergistic control effects between WMSs and longitudinal ventilation, with the ultimate goal of developing an intelligent fire prevention and control system for tunnels.