Combustion Evolution of Aviation Kerosene Pools in Confined Spaces Under Mechanical Negative Pressure

Abstract

This study experimentally investigates the combustion behavior of RP-3 aviation kerosene pool fires (300~800 mm) within a confined space, specifically focusing on the complex interaction between buoyancy-driven plumes and mechanical negative pressure ventilation. By integrating high-precision mass loss measurements with multiple characteristic parameters, this research uniquely characterizes the transition of energy feedback mechanisms under confined suction flow. Results show that ventilation enhances combustion intensity and compresses the fire cycle. For an 800 mm pool, the peak mass loss rate rose by 57.1%, from 16.71 g/s to 26.25 g/s. This enhancement stems from boundary layer thinning, which transitions the combustion from diffusion-controlled to kinetics-controlled. Ventilation also induces severe flame tilt with a non-monotonic trend. The tilt angle peaks at 84° for 600 mm pools but drops to 64° at 800 mm as buoyancy momentum increases. Additionally, an energy contrast of vertical cooling and horizontal heating was observed. Axial peak temperatures decreased by 20%, while downwind thermal radiation flux increased by up to 125%. The ventilation system essentially acts as a directional energy projector, shifting heat loads toward the downwind region. These findings support the optimization of fire safety and detection designs for industrial ventilation systems. This study experimentally investigates the combustion behavior of RP-3 aviation kerosene pool fires (300–800 mm) within a confined space, specifically focusing on the complex interaction between buoyancy-driven plumes and mechanical negative pressure ventilation. By integrating high-precision mass loss measurements with multi-point thermal and imaging diagnostics, this research uniquely characterizes the transition of energy feedback mechanisms under confined suction flow.

1. Introduction

Aviation kerosene serves as a strategic energy source for modern aviation transportation and advanced manufacturing [1]. Its fire safety during production, transportation, and industrial application directly relates to the intrinsic safety of industrial facilities. As a typical hydrocarbon fuel, RP-3 aviation kerosene possesses significant thermophysical properties, including high calorific value, strong volatility, and intense combustion [2]. Once a leak ignites, the resulting pool fire releases massive energy within a very short period. This energy release occurs through intense thermal radiation feedback and convective heat transfer. Such events pose a severe threat to human life, precision machinery, and the structural integrity of buildings. In industrial plants or semi-enclosed spaces with mechanical exhaust systems, nonlinear coupling occurs between the fire source and the ventilation flow field. This interaction disrupts the traditional equilibrium of fire dynamics. It significantly alters the flame spread trajectory, heat release rate, and the distribution of the thermal radiation field. Although RP-3 aviation kerosene is irreplaceable in industrial scenarios such as cleaning, cutting, and surface treatment, the interference mechanism of the complex flow field induced by mechanical negative pressure exhaust on pool fire dynamics remains unclear [3]. Consequently, the accurate assessment and dynamic prevention of fire risks in confined spaces have become urgent scientific challenges in the field of engineering safety [4].

For a long time, pool fires have served as a fundamental research topic in fire science [5]. They represent the most established method for studying flame combustion characteristics. Existing studies have conducted systematic and in-depth investigations into the pool fire combustion behavior of typical liquid fuels. These fuels include heptane, diesel, methanol, and ethanol. The core of this research lies in revealing fundamental combustion mechanisms. For example, researchers examine the quantitative relationships between mass burning rate and factors such as pool size, fuel thickness, initial temperature, and ambient pressure [6]. They also analyze the variation patterns of key parameters such as flame height, plume temperature, and radiation fraction. Sun et al. [7] systematically studied biodiesel pool fires. They found that the combustion follows a typical three-stage pattern. Furthermore, they established predictive correlations for burning rate, average flame height, and axial plume temperature. Ge et al. [8] conducted multi-pool fire experiments. They discovered that reducing the flame spacing causes the tilt angle to increase initially before stabilizing. Consequently, they developed temperature distribution expressions considering tilt angles and a weighted multi-point source radiation calculation model. Tian et al. [9] researched methanol pool fires. They found that convective heat transfer from the burner significantly affects the burning rate under natural ventilation. They also determined the critical Richardson number to be approximately 1. Fang et al. [10] experimentally investigated the coupled combustion of storage tanks and thin pool fires. They observed that the regression rate of the storage tank undergoes unique stages of sudden increase, stabilization, and further increase. Additionally, the regression rate of the coupled pool fire is lower than that of an independent pool fire. Tao et al. [11] showed through experiments that burning rates for both annular and rectangular pool fires increase with the shape parameter. Moreover, these rates are significantly affected by ambient pressure. Zhao et al. [12] studied gasoline thin-layer pool fires. Their research indicated that penetrating thermal radiation is the primary cause of burning rate decline. This radiation increases exponentially as the liquid layer thins. Zhao et al. [13] also experimentally studied transformer oil pool fires. They found that parameters like mass burning rate and radiation fraction have quantitative relationships with the pool diameter. A constant boiling layer of approximately 2.6 mm exists within the fuel layer. Chen et al. [14] investigated the influence of initial fuel temperature on heptane pool fires. They found that the burning rate is independent of the initial temperature during the steady stage. However, the rate increases with temperature during the bulk boiling stage. Hu et al. [15] combined sub-atmospheric pressure and crossflow factors. They revealed that the flame base drag length is greater under lower pressures. They further explained and predicted the influence of pressure on flame base drag behavior.

To simulate the impact of ambient wind, numerous studies have conducted experiments in wind tunnels or open spaces using horizontal crossflows or sidewinds. These works systematically reveal the complex effects of uniform, stable crossflows on flame morphology and combustion behavior. Hu et al. [16] discovered that the burning rate of gasoline pool fires under crossflows is negatively correlated with pool size. They subsequently established a burning rate enhancement model based on heat transfer from the pool walls. Hu et al. [17] also revealed that the burning rate of medium-to-large heptane pool fires exhibits multi-stage, non-monotonic variations as crossflow velocity increases. This transition is closely related to pool size and buoyancy effects. Salvagni et al. [18] demonstrated that the burning rate of diesel pool fires oscillates within crossflows. They also noted that flame geometric parameters tend to stabilize beyond a specific wind speed. Tang et al. [19] quantified the behavior of near-wall flames under sidewinds. They found that these flames are longer than free flames. Because the two follow different variation patterns with sidewinds, separate prediction models were established. Li et al. [20] compared the spread characteristics of diesel flames under opposing and concurrent airflows. They found that spread rates exhibit opposite trends. Furthermore, the dominant heat transfer mechanism changes with wind speed. Woods et al. [21] indicated that the response of methanol pool fire burning rates to crossflows depends on pool size and shape. The streamwise length of the pool is the critical characteristic dimension for this response. Hu et al. [22] quantified the evolution of heat feedback in crossflows. They observed that while conduction and convection are enhanced, radiation is weakened. Consequently, they proposed the stagnant layer theory to describe mass burning flux. Ping Ping [23] investigated the non-monotonic response of crude oil boil-over fire burning rates to crossflows. This work included an improved flame length model for the boiling stage that considers initial oil layer thickness. Hu et al. [24] found that the elongation effect of crossflows on flame length can be predicted uniformly using a dimensionless model incorporating fuel thermochemical properties. Hu et al. [25] also clarified that crossflows significantly reduce radiation feedback by deflecting the flame. This causes the relative importance of conduction and convection feedback mechanisms to increase. Jiang et al. [26] categorized the variation of aviation fuel pool fire burning rates under sidewinds into three stages. They established separate prediction models for different wind speed intervals.

In summary, although previous studies on liquid fuel pool fire combustion mechanisms and ambient wind effects have been highly effective, several critical issues remain unresolved. First, combustion behavior exhibits significant scale effects. For pool fires, the diameter D = 300 mm is widely recognized as a critical threshold where the flame transitions from a laminar or transitional regime to a fully turbulent regime [27]. Within the 300~800 mm range, thermal radiation becomes the dominant heat feedback mechanism, which more accurately reflects the physical characteristics of large-scale industrial fires compared to D < 200 mm small-scale wicks [5,28,29]. From an engineering perspective, this range covers the most frequent medium-scale spill fire scenarios encountered in aviation maintenance hangars, pump rooms, and fuel storage areas, where localized leakages often result in pool sizes of these dimensions. Consequently, systematic experimental data in this range is essential for bridging the gap between laboratory-scale fundamental research and full-scale fire safety engineering. Second, current ambient wind models mostly rely on uniform and stable crossflows in wind tunnels or open spaces. Physically, these represent idealized positive-pressure-driven flow fields. Real industrial fire scenarios, however, are often constrained by non-uniform and unsteady negative pressure fields from mechanical exhaust systems. These exhaust-induced confined flows differ fundamentally from uniform crossflows in entrainment dynamics and flame stability. Consequently, existing theoretical models face significant applicability challenges when addressing negative pressure ventilation in confined spaces.

In view of this, the present study established a confined space pool fire experimental platform. We selected six characteristic pool sizes: 300 mm, 400 mm, 500 mm, 600 mm, 700 mm, and 800 mm. Systematic combustion experiments were conducted using RP-3 aviation kerosene under mechanical negative pressure ventilation. This study focuses on deconstructing mass loss rate, spatial temperature field evolution, flame morphology dynamics, and directional thermal radiation flux. These analyses help reveal the intervention mechanism of negative pressure flow on medium-to-large pool fire characteristics. We also clarify the physical essence of the combustion mode transition from diffusion control to kinetics control. The research findings aim to fill the data gap for industrial fuels in complex confined flow fields. Finally, this work provides a theoretical basis and scientific support for accurate fire safety design, ventilation system optimization, and emergency response in industrial plants.

2. Materials and Methods

As shown in Figure 1, the pool fire experimental system was constructed inside a confined chamber. The chamber features a spatial radius of 4 m and a vertical wall height of 3 m. The ceiling consists of a conical funnel structure with a height of 1 m. A horizontal extraction fan is fixed at the top-right wall position. Its exhaust direction is collinear with the thermal radiation sensors. The fan provides an exhaust volume of 0.32 m³/s, with an adjustable maximum exit wind speed of 0.4 m/s. During the exhaust process, the chamber doors and windows allow for continuous fresh air intake to ensure a sufficient oxygen supply. To characterize the airflow distribution, a multi-point velocity calibration was performed using a hot-wire anemometer before the formal tests. Although the mechanical exhaust creates a non-uniform suction field toward the vent, the horizontal wind speed at the fuel pan center was maintained at a stable 0.4 m/s. The selection of 0.4 m/s as the target ventilation velocity was informed by three primary considerations. First, preliminary sensitivity tests indicated that at this velocity, the transition from diffusion-controlled to kinetics-controlled combustion was most pronounced across the 300~800 mm pool range. Second, this velocity represents the stable upper operating limit of the mechanical exhaust system (0.32 m³/s), ensuring a consistent and maximum sustainable negative pressure field within the 3 m high confined space. Third, 0.4 m/s at the exhaust vent mirrors the typical low-velocity suction flow fields encountered in industrial pump rooms or aviation hangars during standard mechanical air exchange, providing high engineering relevance.

The distance between the exhaust vent and the fuel pan was set at 2.8 m, which is sufficient to allow the airflow to develop into a relatively stable stream before interacting with the flame plume. Regarding the chamber constraints, the fixed position of the fan and the confined wall boundaries (3 m height) inevitably introduce localized entrainment effects. These constraints reflect real-world industrial ventilation scenarios, where negative pressure fields are often non-uniform. The study focuses on the ‘suction-induced’ tilt rather than an idealized wind tunnel crossflow, acknowledging that wall-induced turbulence may slightly influence the flame oscillation frequency.

In this study, the negative pressure environment was established and maintained by the constant volumetric exhaust rate of the mechanical fan (0.32 m³/s). Although direct pressure transducers were not employed, the negative pressure level was implicitly controlled by the fixed fan frequency and the specific geometry of the confined chamber (3 m height, 4 m radius). Before the formal tests, the exit wind speed was stabilized at 0.4 m/s, ensuring that the chamber reached a quasi-steady negative pressure state. This approach ensures that the suction-driven airflow, which is the primary factor affecting the mass loss rate (MLR) and flame dynamics, remains consistent across all experimental groups.

The fuel pool, the core of the experiment, is positioned at the center of the model. The empty pool has an edge height of 40 mm. RP-3 aviation kerosene is utilized as the primary fuel. An electronic balance (±0.1 g) is placed beneath the pool to transmit real-time weight data. To prevent high-temperature damage to the balance and data deviation, a square quartz plate is inserted between the pool and the balance. The kerosene is filled to an initial height of 30 mm. Eight K-type armored thermocouples (numbered 1# to 8#) are installed to monitor fuel and flame temperatures. These sensors, produced by Shanghai Songdao Heating Sensor Co., Ltd. (Model WRNK-191-1.5 mm), feature a measurement range of 0~1200 °C with an accuracy of ±1.0 °C. All thermocouple tips are collinear with the vertical centerline of the pool. Sensors 1# to 7# are spaced 0.5 m apart, while sensor 8# is positioned 1 m above sensor 7#. Real-time temperature data are transmitted via a data logger to a computer terminal for storage. Three thermal radiation sensors (Model TS-34C-34 mm, Beijing Oriental Top Technology Co., Ltd., Beijing, China) are placed in the downwind direction. These sensors have a range of 0~200 kW/m². They are positioned 1 m, 1.5 m, and 2 m from the pool extension, respectively. All radiation sensors are maintained at a uniform installation height of 1.2 m. Simultaneously, a digital camera (GoPro, 1080P, 60 fps) is fixed on a tripod for image acquisition. It records the entire combustion process and variations in flame morphology. The overall experimental layout strictly follows the spatial and equipment parameters described above. The control room was external, and a 45–60 min cooling period was mandated between runs to ensure initial condition consistency.

Detailed distance layouts are illustrated in Figure 1.

Table 1. Measurement Equipment Parameter List.

Measured Parameter	Sensor	Version	Range	Accuracy/ Uncertainty	Manufacturer
Mass Loss	Real-time electronic balance	XY 20MB	0~21 kg	±0.1 g	Changzhou Lucky Electronic Equipment Co., Ltd. (Changzhou, China)
Temperature	K-type armored thermocouple	WRNK-191-1.5 mm High-temperature Probe Type	0~1200 °C	±1.0 °C	Shanghai Songdao Heating Sensor Co., Ltd. (Shanghai, China)
Heat Flux	Thermal Radiation Sensor	TS-34C-34 mm	0~200 kW/m2	±1 kW °C	Beijing Dongfang Dingpeng Technology Co., Ltd. (Beijing, China)
Flame Height/Angle	GoPro	HERO13Black	1080 pi	60 Hz	GoPro Inc. (San Mateo, CA, USA)

provides the measurement Equipment Parameter List.

The pool diameters used in this study are D = 300, 400, 500, 600, 700, and 800 mm. A long-arm igniter is employed for manual ignition. Six groups of experiments are conducted under both natural convection (fan off) and mechanical ventilation conditions (wind speed 0.4 m/s, flow rate 0.32 m³/s) to investigate flame combustion characteristics. Before the formal experiments, environmental inspections of the confined chamber are completed. The fans and extraction systems are confirmed to be in normal standby mode. The exhaust outlet is adjusted to stabilize the exit wind speed at 0.4 m/s. Subsequently, the electronic balance, thermocouple data acquisition system, thermal radiation data acquisition system, and digital cameras are calibrated to ensure normal data transmission and recording functions. During the preparation stage, RP-3 aviation kerosene is filled into the pool to a specified initial height of h₀ = 30 mm. The installation positions of thermocouples 1#–8#, located in the fuel layer and at vertical measuring points above the pool, are verified. The positions of thermal radiation sensors 1#–3# are also confirmed. After ensuring all detection equipment is correctly connected to the acquisition terminals, digital camera recording is initiated. At the start of the experiment, the igniter is used to light the RP-3 fuel. The electronic balance and thermocouple data loggers are activated simultaneously upon ignition to obtain real-time fuel weight changes, temperature data, and video footage. The operating status of the fan and extraction system is continuously monitored during the test to maintain a stable airflow environment within the chamber. Once the fuel is exhausted, data collection is terminated and saved, and the camera recording is stopped. After data archiving, the pool is allowed to cool naturally to room temperature. Finally, the stored temperature and weight data are organized in preparation for the next experimental condition. To ensure the statistical reliability and reproducibility of the experimental data, each test condition was repeated at least 3 times. The final results presented in this study are the average values of these repeated trials, with a maximum relative deviation of less than 5% observed between parallel tests. Taking into account the accuracy of various sensors and environmental fluctuations, the overall measurement uncertainty is estimated to be approximately 4.5%, which meets the accuracy requirements for large-scale combustion experiments.

Table 2. Summary of Experimental Conditions.

Exp. ID	Pool Diameters D (mm)	Wind Speed u (m/s)	Oil Temperature (°C)	Air Quantity (m3/s)	Air Pressure (kPa)
1	300	0	25 ± 3	0.32 ± 0.02	101 ± 5
2	400
3	500
4	600
5	700
6	800
7	300	0.4 ± 0.02
8	400
9	500
10	600
11	700
12	800

provides a summary of the experimental conditions.

3. Results

3.1. Mass Loss Rate of the Oil Pool Fire

The mass loss rate (MLR, defined as the negative time derivative of the residual fuel mass, dm/dt) of an aviation kerosene pool fire is a core physical quantity for characterizing heat release intensity and evolutionary dynamics. Its dynamic evolution directly reflects the interaction between fuel consumption processes and phase change dynamics within a confined space. The temporal evolution of the residual fuel mass was measured using K-type armored thermocouples. As shown in Figure 2 and Figure 3, in order to maintain the stability of the oil level (30 mm), the initial mass at time t = 0 for each operating condition will vary, which is related to the diameter of the oil pans to ensure the stability of the oil level in each oil pan. The mass loss rate is derived from the mass-time curve through numerical differentiation. Subsequently, the Heat Release Rate (HRR) is estimated using the oxygen consumption principle or the effective heat of combustion for RP-3 fuel (42.8 MJ/kg).

Figure 2 and Figure 3 display the variation curves of the mass loss rate (MLR) over time for RP-3 aviation kerosene pool fires with characteristic diameters of 600 mm, 700 mm, and 800 mm under different ventilation states. Combined with quantitative statistics, the entire combustion process of RP-3 aviation kerosene exhibits distinct staged characteristics across all conditions. These stages include the initial heating stage, the rapid development stage, and the stable decay stage. During the initial heating stage, the MLR remains at a low level, typically around 10 g/s. At this time, the heat flow feedback from the flame is primarily consumed to overcome the latent heat of vaporization and the sensible heat rise of the liquid layer. Since the liquid surface temperature has not yet reached a stable boiling state, the evaporation rate is limited by the saturated vapor pressure. Consequently, the evolution of MLR during this stage is relatively gradual. Upon entering the rapid development stage, the MLR shows a significant upward trend. The intervention of mechanical negative pressure ventilation significantly compresses the duration of this stage. For instance, after activating ventilation for the 600 mm pool, the duration of the rapid development stage dropped from 765 s to 620 s. The underlying physical mechanism is that the high-Reynolds-number turbulent field induced by the ventilation enhances the momentum exchange between the flame base and the air. This process accelerates the mass transfer of the fuel to the gas phase.

The enhancement effect of mechanical negative pressure ventilation on the MLR demonstrates clear scale dependency and dynamic response characteristics. As the pool diameter increases, the absolute value of the combustion intensity enhancement shows a non-linear gain. For the 800 mm pool, the average mass loss rate jumped from 5.925 g/s to 8.933 g/s after activating ventilation. This increase significantly exceeds that of the smaller-scale conditions. The core reason for this dynamic response lies in the fundamental switch of the combustion control mode. Under natural convection, combustion in the confined space is restricted by the limited oxygen supply flux, representing a typical “oxygen-limited” mode. The negative pressure flow field formed by the ventilation continuously extracts combustion products. Furthermore, it significantly thins the concentration boundary layer thickness above the liquid surface through forced convection. This improves the mass transfer capability and shifts the combustion mode toward a “power-controlled” mode dominated by the fuel evaporation rate. Additionally, Figure 2c shows that the MLR curves exhibit more intense instantaneous pulsations under ventilation conditions. These high-frequency oscillations are attributed to the instability of the shear flow field caused by negative pressure ventilation above the oil surface. This instability leads to periodic displacement of the flame base and triggers flame tilt. These factors cause the continuous spatial reorganization of the heat feedback field, forming complex turbulence-coupled combustion characteristics. In summary, mechanical negative pressure ventilation reshapes the boundary layer characteristics and oxygen supply modes in confined spaces. This induces a dynamic switch from low-efficiency diffusion combustion to high-intensity turbulence-enhanced combustion. Consequently, it forces RP-3 aviation kerosene to release chemical energy at rates far exceeding its natural state. In multiple repeated experiments, the maximum relative deviation of the mass loss rate (MLR) and the temperature peak was within 5%, which proved the excellent reproducibility of the experimental system.

Under mechanical negative pressure ventilation at u = 0.4 m/s, the mass loss rate (MLR) of aviation kerosene pool fires exhibits significant pulsation enhancement and cycle compression characteristics. Figure 3 records the MLR evolution curves for six pool scales ranging from 300 to 800 mm under forced convection. Compared to natural convection, the intervention of ventilation causes the peak burning rate (r_max, the maximum MLR under each operating condition) of each pool diameter to undergo a magnitude leap. For the 300 mm pool, the peak rate is 4.12 g/s. As the diameter increases, the enhancement effect expands nonlinearly. The r_max for both 700 mm and 800 mm pools reaches 26.25 g/s, demonstrating extreme instantaneous heat release potential. Throughout the flame development period, the MLR curves under ventilation no longer maintain a stable plateau. Instead, they exhibit intense random pulsations and multi-peak features. This reflects the unsteady perturbation of the liquid surface evaporation process by turbulence coherent structures induced by the negative pressure flow field. Furthermore, forced ventilation significantly accelerates the heat feedback accumulation process of the system. This leads to a substantial increase in the slope of the rapid development stage and a highly compressed combustion cycle. These results reveal the explosive growth characteristics of confined space fires under forced ventilation.

Comparing the experimental data from Figure 2 (no ventilation) and Figure 3 (ventilation) reveals that ventilation has a dual effect on aviation kerosene combustion: increasing intensity and shortening the cycle. Under natural convection, oxygen flux in the confined space is limited by buoyancy plumes. This places the combustion process in a typical oxygen-limited mode. Consequently, the MLR curve evolution is relatively gentle and long-lasting. For example, the peak rate of a 700 mm pool without ventilation is only 14.77 g/s. When mechanical ventilation is activated, the peak burning rate for the same scale climbs to 26.25 g/s, representing a 77.7% increase. This drastic performance difference stems from the fundamental evolution of confined flow field dynamics. Negative pressure ventilation creates a directional pressure gradient within the chamber. The forced intake of fresh air greatly enhances the entrainment intensity at the flame base. This process shifts the combustion control mode from passive diffusion control to a power-enhanced mode dominated by the fuel evaporation flux.

Table 3. Summary Table of Peak Heat Release Rate (HRR).

Diameter (mm)	Wind Speed u (m/s)	MLR ${\dot{\mathit{m}}}_{\mathit{p}\mathit{e}\mathit{a}\mathit{k}}$ (g/s)	Peak Heat Release Rate HRR ${\dot{\mathit{Q}}}_{\mathit{p}\mathit{e}\mathit{a}\mathit{k}}$ (kW)
300	0	7.41	240.1
400		10.09	326.9
500		11.67	378.1
600		14.77	478.5
700		14.77	478.5
800		16.71	541.4
300	0.4 ± 0.02	4.12	133.5
400		25.69	832.4
500		11.55	374.2
600		18.59	602.3
700		26.25	850.5
800		26.25	850.5

is the summary table of Peak Heat Release Rate (HRR, the amount of heat released through chemical reactions within a unit of time) for 12 Operating Conditions of RP-3 Aviation Kerosene Pool Fire. For aviation kerosene (RP-3), its complete combustion heat $\Delta H_{c,eff}$ is typically within the range of 42.8~45 MJ/kg. Due to the incomplete combustion that occurs in pool fires, the combustion efficiency factor is set at 0.75 (typically ranging from 0.7 to 0.9). Among them, the effective combustion heat $\Delta H_{c,eff}$ of RP-3 aviation kerosene is set at 32.4 MJ/kg.

$$\dot{Q}=\dot{m}\cdot \Delta H_{c,eff}$$

(1)

where: $\dot{Q}$: heat release rate, kW; $\dot{m}$: g/s, measured rate of quality loss, g/s; $\Delta H_{c,eff}$: effective combustion heat, kJ/kg.

In summary, mechanical negative pressure ventilation injects kinetic energy into the system. This breaks the physical bottlenecks of product accumulation and oxygen deficiency in confined spaces. As a result, the thermal hazard of the fire source achieves an exponential leap in both scale and intensity simultaneously.

3.2. Geometric Characteristics of the Oil Pool Fire Flame

To accurately quantify the evolution of flame height and tilt angle for aviation kerosene pool fires under mechanical negative pressure ventilation, this study developed a digital image processing (DIP) program based on the MATLAB (MATLAB R2022a) platform. The program performs automated feature extraction from the recorded video sequences. Figure 4 illustrates the logic architecture and calculation flow of the algorithm in detail. It consists of four core modules: image calibration, region of interest (ROI) extraction, binary segmentation, and geometric feature quantification. The program extracts geometric parameters by calculating the pixel distribution within the white regions of the binarized images. To extract the flame tilt angle θ, the algorithm identifies the center point of the pool base $(X_b,Y_b)$ and the coordinates of the flame center of mass $(X_c,Y_c)$. Based on the vector relationship between these two points, the angle of the flame axis relative to the vertical direction is calculated using inverse trigonometric functions.

Figure 5 systematically illustrates the contrast in flame morphology evolution between natural convection u = 0 m/s and mechanical negative pressure ventilation u = 0.4 m/s for RP-3 aviation kerosene pool fires across a continuous scale of D = 300–800 mm. From the perspective of overall evolution, the flames under natural convection remain vertically upright due to strong buoyancy. The geometric structure displays typical axisymmetric diffusion flame characteristics, with the core height pulsating along the vertical axis according to variations in combustion intensity. In contrast, the introduction of mechanical ventilation completely disrupts the original buoyancy balance and imparts a significant horizontal momentum vector to the flame plume. This causes the flame envelope to deflect sharply toward the exhaust outlet throughout its life cycle. Experimental observations indicate that this deflection effect is particularly prominent during the growth stage. As the pool diameter increases, both the flame volume and the degree of turbulent wrinkling rise sharply. Consequently, the 800 mm large-scale flame demonstrates intense spatial aggressiveness in the horizontal direction.

Comparing the characteristic time nodes reveals that the total duration from ignition to extinction is significantly reduced under the u = 0.4 m/s condition for all pool diameters. For instance, the extinction time for the 300 mm pool advances from 460 s under natural convection to 450 s under ventilation. The evolution of the 800 mm large-scale pool is even more explosive, with its extinction node compressed from 1300 s to 1000 s. This demonstrates the intense squeezing effect of forced convection on chemical reaction rates. Furthermore, the flame morphology during the decay stage under ventilation shows increased fragmentation. Because negative pressure suction maintains a steady oxygen supply at the flame base, the flame retains high brightness and turbulent pulsation frequencies until final extinction.

The physical mechanism for this morphological reshaping and accelerated evolution lies in the fundamental change in entrainment modes caused by the negative pressure field in confined spaces. Driven by the directional pressure gradient from mechanical exhaust, fresh air is forced toward the flame base instead of relying on passive buoyancy entrainment. This enhanced mass transfer mechanism thins the reaction-diffusion layer and shifts the combustion mode toward high-intensity reaction. As the pool diameter increases from 300 mm to 800 mm, the surface area exposed to the wind grows geometrically. This causes the drag moment generated by the horizontal airflow to far exceed the buoyant moment of the flame, driving a large deflection of the flame axis toward the exhaust side. This morphology reconstruction explains the surge in lateral thermal radiation measurements. The tilted flame physically brings the high-temperature radiation source closer to downwind sensors. This forcibly transfers the thermal hazard from the traditional vertical area to the downwind equipment zones, significantly increasing the dynamic risk of secondary ignition in confined plants.

A comparison of the temporal evolution of flame height H under natural convection (Figure 6a) and mechanical negative pressure ventilation (Figure 6b) clearly reveals the regulatory effects of confined space flow fields on fires of different scales. Under natural convection, the flame height increases nonlinearly as the pool diameter D grows. Specifically, the peak flame height for the 700 mm pool reaches 317.32 cm. Notably, as the diameter further increases to 800 mm, the height under natural convection drops back to 247.70 cm. This decrease reflects local oxygen depletion caused by intense combustion at the top of the fully enclosed space (the conical funnel structure). It also highlights the physical suppression of upward flame development by the accumulation of combustion products.

Upon activating mechanical negative pressure ventilation u = 0.4 m/s, the distribution characteristics of flame height undergo fundamental restructuring. For small-scale pools like 300 mm, the forced entrainment induced by ventilation significantly enhances the reaction intensity at the flame base. This causes the peak height to jump dramatically from 112.43 cm under natural convection to 195.98 cm. However, for large-scale pools (600 mm to 800 mm), the vertical flame height does not increase proportionally after ventilation is activated. In some cases, it even decreases; for instance, the peak height for the 700 mm condition drops to 207.63 cm. This phenomenon does not indicate a reduction in combustion intensity. Instead, it is a direct quantification of the previously mentioned flame tilt. The forced flow field stretches the flame plume horizontally, shifting flame energy from the vertical axis toward the exhaust side. Furthermore, ventilation significantly shortens the combustion cycle for all conditions. The effective combustion duration for the 800 mm case is compressed from approximately 1300 s to less than 1000 s, demonstrating extreme explosiveness.

The integrated analysis of flame morphology (Figure 5) and quantitative height (Figure 6) reveals a distinct “scale-flow field” coupling effect of mechanical negative pressure ventilation on confined space pool fires. At small scales, the primary roles of ventilation are “combustion assistance and stretching,” which significantly increase the flame volume by enhancing the oxygen supply. Conversely, at medium-to-large scales, its role shifts to “reshaping and transferring” by injecting horizontal kinetic energy to forcibly alter the heat feedback path. This transition in dynamic behavior reveals the core logic of fire safety design in confined spaces: while ventilation systems dilute harmful gases, they essentially transform the vertical thermal hazard into horizontal, high-intensity thermal aggression by thinning the heat and mass transfer boundary layers and altering radiation view factors. Consequently, for industrial plant designs involving large-diameter aviation kerosene storage areas, the flame tilt effect induced by negative pressure flow fields must be comprehensively considered. This is essential to prevent the risk of horizontal chain ignition of facilities caused by the restructuring of flame morphology.

The flame tilt angle θ is the most intuitive geometric parameter reflecting the force state of a fire plume within a confined space. Its magnitude directly determines the spatial directionality of the fire’s heat release and the resulting distribution of thermal hazards. Figure 7 records the real-time evolution of flame tilt angles for six pool scales, ranging from 300 to 800 mm, under a constant ventilation intensity of u = 0.4 m/s. Experimental data indicate that the horizontal flow field induced by mechanical negative pressure ventilation disrupts the original axisymmetric equilibrium of the flame. This causes significant downwind deflection across all experimental conditions. Furthermore, the degree of tilt exhibits complex unsteady characteristics that follow the fluctuations in combustion intensity.

According to the temporal distribution shown in Figure 7, the tilt angles for all pool diameters remain in an active state of oscillation throughout the entire combustion cycle. For the 300 mm and 400 mm small-scale pools, the oscillation center of the tilt angle is relatively low, primarily fluctuating within the 18° to 36° range. As the pool diameter D increases to the 500–700 mm range, the evolution of the tilt angle demonstrates a stronger correlation with the combustion phase. During the later stages of combustion, the depletion of the liquid fuel causes a sharp decrease in vertical buoyancy momentum. Consequently, the horizontal ventilation torque dominates the momentum balance, inducing an explosive increase in the tilt angle. The statistics for the maximum tilt angle in Figure 8 reveal a nonlinear trend that first increases and then decreases with pool diameter. The maximum tilt angle reaches a peak of 84° for the 600 mm pool, representing an extreme state where the flame is nearly horizontal. Conversely, when the diameter increases further to 700 mm and 800 mm, the maximum tilt angle drops back to 70° and 64°, respectively.

This complex scale-dependent response can be explained by the vector force model illustrated in Figure 9. Within the negative pressure flow field at the top of the confined space, the flame is simultaneously subjected to upward buoyancy V and the horizontal forced ventilation force F_w. The vector sum of these forces, the net momentum force F_t, determines the instantaneous direction of the flame tilt angle θ. At smaller scales, V is relatively low, allowing F_w to easily drive a large deflection. As the diameter increases to 600 mm, the growth rate of the flame’s windward area exceeds the increase in vertical lift, causing the tilt angle to reach its geometric limit. However, upon entering the 800 mm large-scale range, the mass loss rate improves significantly. This generates massive upward thermal buoyancy momentum, which greatly enhances the vertical stiffness of the fire plume. This increased stiffness provides stronger resistance to lateral wind disturbances, leading to a reduction in the maximum tilt angle compared to the 600 mm case.

Analysis of both the temporal variation (Figure 7) and the force mechanism (Figure 9) shows that the reshaping of flame morphology by mechanical negative pressure ventilation has distinct non-monotonic features. This dynamic change in tilting behavior directly drives the spatial reconstruction of heat feedback. When the flame tilts at a large angle, the high-temperature core physically approaches the downwind sensors. This leads to the previously noted doubling of thermal radiation intensity. This physical essence reveals the coupling mechanism between the ventilation system and fire source scale in confined plants. In the fire safety design of large-scale aviation kerosene storage areas, it is critical to evaluate the extreme morphological deflection induced by exhaust and its associated risks of directional thermal aggression.

In fire dynamics research, the introduction of dimensionless parameters for similarity analysis represents an effective method for revealing the fundamental combustion laws within complex flow fields. This study further explores the deep mechanisms by which mechanical negative pressure ventilation intervenes in aviation kerosene pool fires by constructing a momentum balance model. To address the flame tilt phenomenon, the Modified Froude Number (Fr) is defined to characterize the competition between the horizontal negative pressure flow momentum and the buoyancy momentum of the fire plume.

According to the law of conservation of momentum, the magnitude of the flame tilt angle θ depends on the vector synthesis of the horizontal forced force F_w and the vertical buoyancy V. To establish a quantitative relationship, the dimensionless characteristic velocity u* is introduced as a correlation term (Formula 2).

$$u^{\ast }={\displaystyle \frac{u}{{(\dot{Q}g/{\rho }_{\infty }{c}_p{T}_{\infty })}^{\frac{1}{3}}}}$$

(2)

where $\dot{Q}$ is the heat release rate of the fire source (kW); g is the gravitational acceleration (9.81\m/s²); ${\rho }_{\infty }$ is the ambient air density (kg/m³); C_p is the specific heat capacity of ambient air (kJ/(kg·K)); and $T_{\infty }$ is the ambient temperature (K).

Physically, the dimensionless velocity $u^{\ast }$ characterizes the momentum competition between the horizontal ventilation flow and the vertical buoyant fire plume [30]. In Equation (1), the numerator $u$ represents the inertial velocity of the mechanical cross-flow, which tends to deflect the flame horizontally. The denominator, ${(\dot{Q}g/{\rho }_{\infty }{c}_p{T}_{\infty })}^{{\displaystyle \frac{1}{3}}}$, represents the characteristic vertical ascent velocity induced by the thermal buoyancy of the fire. A larger $u^{\ast }$ indicates a flow regime dominated by the horizontal ventilation momentum, resulting in a more significant flame tilt. Conversely, a smaller $u^{\ast }$ implies that the vertical buoyancy driven by the intense heat release rate ($\dot{Q}$) is strong enough to maintain a relatively upright flame plume, enhancing its physical stiffness against cross-flow disturbances.

In this context, u represents the characteristic exhaust wind speed, and Q denotes the instantaneous heat release rate of the fire source. The experimental results indicate a significant power-function correlation between the tangent of the flame tilt angle tanθ and the dimensionless velocity u* [31]. Observations from Figure 8 and Figure 9 show that as the pool diameter D increases from 300 mm to 600 mm, the growth rate of the wind-exposed area exceeds the increment in initial momentum. This leads to an enhanced Froude number Fr effect, which increases the deflection sensitivity of the flame. However, when the diameter increases further to 800 mm, the vertical upward momentum generated by the sharp increase in burning rate significantly enhances the physical stiffness of the fire plume. This improvement in the flame’s resistance to disturbances from the lateral flow field causes the maximum tilt angle to decrease.

Regarding the enhancement mechanism of the mass loss rate, this study establishes a criterion equation for the dimensionless burning rate ${\dot{m}}^{*}$ rather than using absolute values. Based on boundary layer theory, the high-Reynolds-number turbulent field induced by mechanical negative pressure ventilation significantly thins the mass transfer boundary layer above the liquid surface. This process increases the Sherwood number Sh. The physical essence of this phenomenon is that forced convection exploits the evaporation potential of the fuel. By expressing the dimensionless burning rate as a function of flow field dynamics (Reynolds number, Re) and fire thermodynamics (Froude number, Fr) [32,33,34], it can be defined as:

$${\dot{m}}^{*}={\displaystyle \frac{{\dot{m}}^{″}}{{\rho }_{\infty }\sqrt{gD}}}=fr(\mathrm{Re},Fr)$$

(3)

According to Formulas (2) and (3) and Figure 9, the enhancement effect of ventilation on the burning rate does not follow a linear growth pattern. Rather, it is governed by the nonlinear coupling between flow field dynamics and fire thermodynamics. In the large-scale stage, the Modified Froude Number Fr is relatively low, indicating that buoyancy remains the dominant force. Under these conditions, the flame tends toward an axisymmetric distribution. This causes the disturbance effect of the lateral shear flow field on the boundary layer to converge compared to the medium-scale cases. This dimensionless analysis provides a theoretical explanation for the non-monotonic response characteristics observed in the experiments. It also establishes a physical foundation for the development of future cross-scale prediction models for aviation kerosene fires.

3.3. Characteristics of Temperature Variation of Flames

The flame temperature field is a critical physical parameter that reflects combustion intensity, heat release rate, and the spatial distribution of the gas-phase reaction zone. Its evolutionary characteristics provide a deep understanding of the dynamic response of fire plumes in confined spaces. Thermocouples 1# to 8# are arranged along the vertical centerline of the fuel pool. This configuration allows for the precise capture of spatiotemporal temperature gradients, ranging from the internal liquid domain to the far-field gas-phase plume. It should be noted that the temperatures recorded by the sheathed thermocouples represent local gas-phase values rather than the theoretical adiabatic flame temperature. This discrepancy is primarily attributed to the substantial radiative heat losses to the compartment boundaries and the cooling effect triggered by enhanced air entrainment under negative pressure ventilation. In industrial confined space fires, the effective heat field is significantly affected by radiant heat dissipation and non-uniform air dilution. For engineering safety design, the measured local gas temperature is the only reliable basis for evaluating structural integrity and fire protection requirements, because the theoretical maximum temperature value still has a considerable gap compared to the actual heat load. However, in the process of capturing the fire temperature in actual factories, the measurement data obtained by thermocouples tend to be closer to the actual values compared to the theoretical adiabatic flame temperature. Although the theoretical adiabatic flame temperature of RP-3 can be calculated using the thermodynamic equilibrium formula, in this analysis we have taken the actual measured temperature in the environment as the main angle of analysis.

Figure 10 illustrates the temporal evolution of temperatures along the vertical axis for pool diameters ranging from 300 mm to 800 mm under the baseline condition of natural convection u = 0 m/s. Experimental data show that all cases undergo a rapid temperature increase during the growth stage. Peak temperatures are primarily concentrated within the continuous flame zone, which extends from the fuel surface to a vertical height of 0.5 m. For the 600 mm, 700 mm, and 800 mm pools, the maximum axial temperatures T_max reach 823 °C, 903 °C, and 882 °C, respectively. Regarding spatial distribution, the axial temperature exhibits a clear monotonic decrease with increasing height. This trend is caused by the intense entrainment of surrounding cold air into the thermal plume. Such axisymmetric distribution reflects that the flame remains vertically upright under buoyancy control during natural convection. Consequently, the heat feedback path is concentrated directly above the fuel pool.

The introduction of mechanical negative pressure ventilation significantly alters the axial thermal evolution process of aviation kerosene pool fires within confined spaces. Figure 11 illustrates the real-time temperature distribution along the vertical axis for six pool scales ranging from 300 mm to 800 mm under an exhaust condition of u = 0.4 m/s. Compared to the natural convection condition in Figure 10, the axial temperature field under ventilation exhibits distinct peak suppression and cycle compression. Experimental data show that after activating the exhaust, the maximum axial temperatures for the 600 mm, 700 mm, and 800 mm pools dropped to approximately 715 °C, 700 °C, and 701 °C, respectively. Compared to the 823 °C to 903 °C range observed under no-wind conditions, the peak axial temperature decrease reaches up to 22.5%, demonstrating the potent thermal interference of the negative pressure flow field on the central axis region.

Regarding overall development trends, the temperature rise curves for the various measurement heights between 0.030 m and 3.000 m in Figure 11 exhibit higher instability and synchronicity under the influence of ventilation. Under natural convection, the fire plume is buoyancy-driven and vertically upright, transporting heat primarily upward along the central axis and forming a clear hierarchical temperature gradient. In the negative pressure flow field induced by mechanical exhaust, however, the temperature rise rates captured by thermocouples at all heights accelerate significantly. Because the entire flame deflects sharply toward the exhaust outlet—with tilt angles reaching 60° to 84°—the thermocouple array originally situated on the geometric center axis is effectively positioned at the edge or on the leeward side of the tilted flame. This spatial reconfiguration causes the high-temperature reaction core to physically deviate from the detection points, manifesting as a notable temperature drop in the monitoring data.

Another underlying dynamic cause for this axial cooling is the forced convection cooling effect. Driven by the directional pressure gradient of the exhaust system, a large volume of fresh ambient air is forced into the flame base, significantly enhancing turbulent entrainment and convective heat transfer at the flame front. This continuous cold air dilution not only lowers the gas temperature at the axial measurement points but also accelerates the discharge of gas-phase reaction products. Furthermore, looking at the time scales in Figure 11, the duration of the temperature rise for each condition is substantially compressed. For instance, the effective combustion duration for the 800 mm case under ventilation was shortened from 600 s in its natural state to approximately 400 s. This further corroborates the transition of the combustion mode toward a high-intensity, short-cycle power control mode under ventilation.

Moreover, the second temperature peak observed toward the end of the burning process (approx. 500~800 s) is attributed to the pan-bottom effect and thin-layer heating. As the fuel layer is depleted and becomes sufficiently thin, the thermal energy feedback from the flames is no longer absorbed by a bulk liquid volume but is instead conducted to the metallic pan bottom. The rapid temperature rise of the pan base provides secondary heat feedback to the remaining thin fuel layer, triggering a transient surge in the evaporation rate and burning intensity just before extinction. This terminal flare-up is further intensified in the confined space by the cumulative radiative feedback from the heated compartment walls.

In summary, mechanical negative pressure ventilation essentially achieves a directional reconstruction of the energy field within the confined space via kinetic energy injection. Although the apparent temperatures recorded at the axial monitoring points decrease due to flame deflection and cold air entrainment, the overall heat release intensity of the system is not weakened. Instead, the hazard load is forcibly transferred from the vertical overhead area to the horizontal downwind region. This coexistence of vertical cooling and horizontal thermal radiation surges reveals the spatial asymmetry of thermal threats in ventilated environments. Such findings provide critical engineering guidance for the layout of fire detection systems and the formulation of emergency response strategies in aviation kerosene storage facilities.

3.4. Thermal Radiation Characteristics of Flames

Thermal radiation flux serves as the core physical quantity for evaluating the thermal hazard and secondary ignition risk of confined space fires. Its magnitude is jointly governed by the heat release rate of the fire source, the geometric morphology of the flame, and the spatial positioning relationships. By deploying thermal radiation sensors 1#, 2#, and 3# in the downwind direction of the exhaust, the directional thermal aggression characteristics of the flame under various flow field interventions can be accurately captured.

Figure 12 illustrates the temporal evolution of thermal radiation for various pool scales under the baseline condition of natural convection u = 0 m/s. Experimental results indicate that the thermal radiation flux increases steadily with the pool diameter (D). In the small-scale range of 300 to 500 mm, the peak values at sensor 1# fluctuate between 4000 and 5500 W/m². As the diameter increases to 600 to 800 mm, the peak range rises to 7500 to 11,000 W/m². Regarding spatial distribution, since the flame remains vertically upright, heat is radiated outward axisymmetrically. This results in a clear attenuation characteristic as the distance increases, specifically 1# > 2# > 3#. Furthermore, the pulsation frequency is low, and the curves evolve relatively smoothly.

Figure 13 illustrates the temporal evolution of thermal radiation for various pool scales under the baseline condition of natural convection u = 0.4 m/s. A systematic comparative analysis of experimental data under natural convection (u = 0.4 m/s) and mechanical negative pressure ventilation (u = 0.4 m/s)) clearly reveals the regulatory laws of confined space flow fields on the dynamic behavior of aviation kerosene pool fires. Regarding the mass loss rate (MLR), mechanical exhaust demonstrates a significant combustion-assisting effect, with the peak burning rate (r_max) for all pool scales undergoing leapfrog growth. For instance, the 800 mm diameter pool jumps from 16.71 g/s under no-wind conditions to 26.25 g/s under ventilation, representing a 57% increase. Simultaneously, the combustion cycle is substantially compressed; the total combustion duration for the 800 mm case is reduced from 1300 s to 1000 s. This validates the physical essence of the negative pressure flow field pushing the combustion mode from diffusion control toward kinetic enhancement by increasing oxygen flux.

The comparison of flame geometric characteristics further reveals the logic of spatial energy redistribution. Under natural convection, the flame height H exhibits a strong linear correlation with the pool diameter D, with the 700 mm pool reaching a maximum vertical height of 317.32 cm. However, after activating the exhaust, the injection of horizontal momentum induces severe flame tilt, causing the peak vertical height of large-scale pools to decrease. This phenomenon is most extreme for the 600 mm pool, where the maximum tilt angle reaches 84°, exhibiting a nearly recumbent morphology. Analysis via the vector force model shows that this non-monotonic tilt response stems from the competition between vertical buoyancy momentum and horizontal exhaust torque. When the diameter reaches 800 mm, the massive upward lift generated by the surge in burning rate enhances the vertical stiffness of the fire plume, causing the maximum tilt angle to decrease to 64°.

The most significant comparative feature is reflected in the deconstruction and reconstruction of the spatial temperature and thermal radiation fields. A unique energy mutation phenomenon was observed: after activating the exhaust, temperatures at thermocouple points located on the vertical axis of the pool generally decreased; for the 800 mm case, the peak axial temperature dropped from 882 °C to 701 °C, a decrease of approximately 20%. Simultaneously, the values at the downwind radiation sensor 1# experienced exponential growth, with the peak surging from approximately 11,000 W/m² under natural convection to over 17,500 W/m² under ventilation. This contrast of vertical cooling and horizontal heating profoundly reveals the dual role of mechanical negative pressure ventilation in fire evolution: while it reduces local axial temperatures through forced convection cooling, it essentially projects extreme thermal loads directionally toward the downwind area by altering radiation view factors. In summary, mechanical negative pressure ventilation not only significantly exploits the combustion potential of aviation kerosene but also fundamentally changes the direction of thermal aggression, making the ignition risk for horizontal downwind facilities far exceed traditional prediction ranges.

4. Discussion

4.1. Mechanisms of Combustion Enhancement and Mode Transition Under Negative Pressure

In this study, when the pool diameter D reached 800 mm and the exhaust velocity was set at u = 0.4 m/s, the average mass loss rate (MLR) of RP-3 aviation kerosene underwent a dramatic leap from 16.71 g/s under natural convection to 26.25 g/s, representing a significant increase of 57.1%. This non-linear surge markedly deviates from the predictions of classical open-space crossflow models, such as the Babrauskas or Thomas correlations. In traditional open-field research [27], medium-to-large scale pool fires (D ≥ 300 mm) are typically categorized as diffusion-controlled combustion, where the burning intensity is governed by the rate of air entrainment into the flame reaction zone. However, the mechanical negative pressure environment within the confined space disrupts this inherent equilibrium.

This escalation in combustion intensity is attributed to the forced intervention of the negative pressure field on the heat and mass transfer boundary layer. Unlike uniform crossflows driven by positive pressure, mechanical exhaust creates a localized pressure gradient at the top or side of the confined space, generating a distinct suction effect. This effect first induces a high-Reynolds-number turbulent field, which drastically enhances the shear mixing efficiency between the fuel vapor and ambient air. Furthermore, due to the physical boundary constraints of the confined space, the negative pressure flow forces the flame to tilt and remain in close proximity to the liquid surface, significantly thinning the gas-phase boundary layer above the fuel. According to Stefan flow theory, the thinning of this boundary layer reduces the resistance to heat transfer, allowing the convective heat feedback from the flame to constitute a much larger proportion of the total heat feedback, thereby accelerating the fuel gasification rate.

Fundamentally, this phenomenon signifies a profound transition in the combustion mode: under the intense intervention of mechanical negative pressure, the process evolves from a traditional diffusion mode limited by oxygen supply to a mode driven by high heat feedback and dominated by fuel evaporation flux. This finding not only complements the discourse by Junjunan [28] and Deng [29] regarding the scale effects in compartment fires but also reveals the complex coupling between industrial ventilation systems and fire dynamics—namely, that while mechanical exhaust removes smoke, it may simultaneously trigger more aggressive fire behavior by intensifying thermal feedback.

4.2. The Mechanism of Combustion Enhancement and Energy Field Reconstruction Under Negative Pressure Intervention

The negative pressure field fundamentally reshapes the flame morphology and energy distribution. Contrary to the monotonic tilt angle increase observed in Huang [27], the decrease in tilt angle at D = 800 mm (from 84° to 64°) reveals that the surge in burning rate enhances plume stiffness, allowing the fire plume to resist lateral suction. Regarding the energy field, while forced convection causes axial temperatures to drop (from 823~903 °C to 700~715 °C) due to air entrainment, it triggers a 125% increase in downwind radiant heat flux. This extends the findings of Junjunan [28] and Deng [29] by demonstrating that mechanical exhaust acts as a “directional energy projector,” diverting the primary thermal load from the vertical axis to the horizontal downwind direction. Such energy reconstruction implies a heightened risk of failure for traditional overhead detectors while intensifying the ignition hazard for adjacent downwind facilities.

5. Conclusions

This study systematically investigates the combustion characteristics of RP-3 aviation kerosene pool fires in confined spaces under mechanical negative pressure ventilation. Through quantitative analysis of mass loss rate, flame geometry, spatial temperature fields, and directional thermal radiation flux for pool scales ranging from 300 to 800 mm, the following major conclusions are drawn:

• Mechanical negative pressure ventilation significantly enhances the combustion intensity of aviation kerosene and alters its dynamic evolutionary process. Experimental results indicate that under the u = 0.4 m/s exhaust condition, the average mass loss rate (MLR) for all pool scales undergoes a substantial leap. Specifically, the peak burning rate for the 800 mm pool increases from 16.71 g/s under natural convection to 26.25 g/s, representing a 57.1% increase. The high-Reynolds-number turbulent field induced by the exhaust effectively thins the heat and mass transfer boundary layer above the liquid surface. This promotes a transition in the combustion mode from diffusion control, limited by oxygen supply, to a power-enhanced mode dominated by fuel evaporation flux.
• The mechanical negative pressure flow field fundamentally reshapes the flame geometry, exhibiting significant scale dependence. As the pool diameter increases, the flame undergoes severe deflection under the influence of horizontal momentum, with the maximum tilt angle peaking at 84° for the 600 mm case. However, when the diameter further increases to 800 mm, the massive vertical buoyancy momentum generated by the surge in burning rate enhances the stiffness of the fire plume, causing the maximum tilt angle to decrease to 64°. Additionally, exhaust intervention at the top of the confined space leads to a decrease in the vertical flame height for large-scale pool fires compared to natural convection, forcibly diverting the energy feedback path from a vertical upward direction to a horizontal downwind direction.
• The energy field within the confined space undergoes drastic spatial reconstruction under the influence of exhaust, characterized by the coexistence of vertical axial cooling and horizontal radiation surges. Due to flame deflection and the forced convection cooling effect induced by forced ventilation, the peak temperatures along the vertical axis of the pool decrease from the 823–903 °C range under natural conditions to approximately 700–715 °C. Simultaneously, the directional thermal radiation flux in the downwind direction exhibits exponential growth, with peak increases reaching up to 125%. This indicates that mechanical exhaust essentially acts as a directional energy projector, forcibly transferring the fire’s thermal load from its original site to the horizontal downwind area and significantly increasing the risk of chain ignition for surrounding facilities.
• The research results reveal the coupling mechanism between industrial plant ventilation systems and fire dynamics, providing a new perspective for fire safety design. In the design of fire protection for aviation kerosene storage and application areas, warning strategies that rely solely on overhead fire detectors may produce false alarms or missed detections due to the temperature drop caused by flame deflection. The directional thermal radiation enhancement effect induced by the negative pressure flow field must be comprehensively considered. It is recommended to mitigate the extreme thermal aggression risks induced by mechanical exhaust in confined spaces through the scientific planning of vent layouts, the addition of downwind thermal radiation shielding devices, and the optimization of differentiated fire separation distances based on wind flow guidance.