Characterization of Medium-Scale Accidental Releases of LNG

Abstract

The need for sustainable energy sources has recently promoted the use of liquefied natural gas (LNG) as a low-carbon fuel. Although economic evaluations indicate the transportation of LNG as a convenient solution for long distances between markets and reservoirs, several concerns are still present regarding its safe use and transportation. The preliminary evaluations performed in this work indicate that credible releases deriving from real bunkering operations result in pools having a diameter smaller than 1 m, which has been poorly investigated so far. Hence, an experimental campaign devoted to the characterization of a medium-scale release of LNG was carried out either in the presence or absence of an ignition source. An evaporation rate of 0.005 kg s⁻¹ m⁻² was collected for the non-reactive scenario, whereas the measured burning rate was 0.100 kg s⁻¹ m⁻². The reduction factor of 20 demonstrates the inaccuracy in the commonly adopted assumption of equality between these values for the LNG pool. Flame morphology was characterized quantitatively and qualitatively, showing a maximum ratio between flame height and flame diameter equal to 2.5 and temperatures up to 1100 K in the proximity of the flame.

1. Introduction

The use of natural gas has recently been boosted by technical, environmental, economic, and political factors [1,2]. Indeed, natural gas is widely recognized as one of the most sustainable alternatives among fossil fuels for sustainable development [3,4,5]. In addition, natural gas can be conveniently considered a short-term solution for the energy transition, especially for hard-to-abate sectors such as trucks and ships. Regarding the storage and transportation alternatives, cryogenically liquefied natural gas (LNG) can be preferred to compressed natural gas (CNG) in case of long distances between reservoirs and markets or the absence of a suitable level of infrastructure. The increasing diffusion of LNG should be coupled with an accurate assessment of safety aspects based on an in-depth characterization of phenomenological aspects ruling the accidental release [6,7]. Particular attention should be posed to the interactions between cryogenic systems and the atmosphere [8,9] in the case of an accidental release on the ground or in water due to failure of the containment system (pipe, hose, or tank) or during the loading/unloading operations [10,11]. According to the literature and technical regulations, any accidental release of LNG can lead to several scenarios, including pool fire, flash fire, and jet fire, based on the presence of immediate or delayed ignition or just the atmospheric dispersion [12]. Due to its likelihood and its potential to trigger domino effects and have consequences for human beings in close surroundings [13], pool fires are of particular interest [14,15,16].

Regardless of the flammable liquid material and substrate where the release occurs, pool fires are typically characterized by three phases: initiation, pseudo-steady state, and termination. The first phase is related to the formation of fire above the pool, and it is characterized by an increased burning rate over time, resulting in a larger heat release rate. In the second phase, the presence of a stable flame can be observed. It can be fairly described as a cylinder tilted by the wind (Figure 1), thus with an almost constant mass burning rate. The third phase is associated with a reduction of flame size and mass burning rate due to the limited amount of fuel forming the liquid pool. A representation of the flame can be obtained by considering the flame dimensions (diameter D_f, height H_f, length L_f, and area A_f) and the tilt angle (δ). Quite obviously, atmospheric conditions can play a relevant role in the determination of flame geometry. Among the others, wind speed and direction represent the most influential parameters on pool fires [17]. In this sense, dedicated analysis based on laboratory scale tests is recommended to produce a stable and controlled atmosphere suitable for the realization of dedicated models. Examples of this approach can be found in the recently published literature [18]. Nevertheless, practical knowledge can be gathered utilizing on-field tests, as well as being a possible representation of a credible scenario of an accidental release in the atmosphere.

If the accidental release is unbounded, a circular pool having a diameter variating with time from the release can be formed until a steady state is reached [19]. Alternatively, a rectangular shape can be expected for non-movable industrial installation, where a catch basin is commonly used to collect the liquid spill. In this case, an equivalent diameter can be assumed for the sake of simplicity. Usually, the flame diameter D_f is supposed to be equal to D_p. In addition, the flame height H_f can be estimated using the mass burning rate per unit area (m″). Several correlations have been proposed for this aim, as reviewed by Raj [20], including the pioneering study performed by Thomas [21], leading to the homonymous correlation. The parameter m″ is typically employed to represent the consumption rate of the pool due to the combustion, whereas the evaporation rate per unit area ($m_{\mathit{e}\mathit{v}}^{\mathit{″}}$) represents the homologous parameter in the absence of an ignition source, i.e., the net phase change per unit area from a liquid pool to the atmosphere. It should be noted that in a complex processing facility, there is the likelihood of the occurrence of cascading scenarios, i.e., hydrocarbon release, fire, explosion, and dispersion of combustion products. It has empirically been proven that both properties can be significantly affected by the consequences generated by the release of liquid fuels [22]. This is particularly relevant for medium- and small-scale releases. Indeed, Babrauskas et al. [23] have proposed ~ 5 m as the infinite diameter of LNG ($D_{\mathit{\infty }}$), namely, the minimum diameter where the asymptotic value of burning rate per unit area ($m_{\mathit{\infty }}^{\mathit{″}}$) can be achieved. According to this model, the following equation is currently adopted for the definition of m″:

$$m^{″}=m_{\infty }^{″}·(1-e^{-k\beta D_p})$$

(1)

where k and β are empirical factors that depend on the investigated fuel. Specifically, k is the absorption-extinction coefficient, which is approximately 3.0 m⁻¹ for LNG [24], and β is the correction coefficient for the beam length. Most of the reported values for kβ range between 0.14 m⁻¹ and 0.46 m⁻¹. Although more recent studies have reported different trends for the mass burning rate per unit area, a consensus in the scientific community on the effects of pool diameter on this parameter has been achieved [25]. Starting from the ‘70s, several tests focused on increasing knowledge of accidental scenarios of LNG were conducted [20]. These tests considered either reactive (i.e., pool fire) or non-reactive (i.e., pool evaporation) scenarios deriving from different pool sizes up to 35 m. A collection of these data is reported in

Table 1. Overview of the main characteristics of selected experimental campaigns on the accidental release of LNG from the literature.

Scenario	Diameter	Results	Reference	Year
Pool fire on water	15	m″ = 0.180 *; 0.495 * kg s−1 m−2 Lf/Dp = 2.8; 4.4	[31,35]	1979
	30	Lf/Dp = 2.6	[36,37]	1980
Pool fire on insulated concrete	20	Lf/Dp = 2.15 m″ = 0.106 kg s−1 m−2	[38]	1982
	1.13 **	Lf/Dp = 3.54 m″ = 0.065 kg s−1 m−2	[39]	2011
Pool fire on soil	35	Lf/Dp = 2.2 m″ = 0.140 kg s−1 m−2	[40]	1989
	1.8, 6.1	Lf/Dp ϵ [1.84—3.05] m″ ϵ [0.052—0.104] kg s−1 m−2	[41]	1974
	7.4 **	m″ = 0.054 kg s−1 m−2	[42,43]	1984
Pool evaporationon water (no fire)	1.97, 3.63	m″ = 0.029 kg s−1 m−2	[44]	1973
	0.75, 6.06	m″ = 0.181 kg s−1 m−2	[45,46]	1970
	7, 14	m″ = 0.195 kg s−1 m−2	[47,48]	1972
	10	m″ = 0.085 kg s−1 m−2	[49,50]	1982
	6.82, 7.22	m″ = 0.120 kg s−1 m−2	[51]	1978
Trench fire	0.22—6.8	m″ ϵ [0.022—0.13] kg s−1 m−2	[23]	1983

* Calculated considering a density of 450 kg m⁻³. ** Equivalent diameter.

. A significant discrepancy can be observed among the different sources, although similar investigated conditions were reported at this stage [23]. Similar conclusions could be drawn for the L_f/D_p ratios reported so far. Moreover, comparable values for m″ and $m_{ev}^{″}$ are reported, despite the differences in the boundary conditions characterizing the system from a thermal point of view. As it is possible to note, the majority of data have been collected for D_p > 5 m (i.e., larger than $D_{\infty }$), whereas a dearth of studies can be observed for D_p < 1 m. Generally speaking, even if large-scale experimental scenarios have already been tested [26] and modelled [27], quite unexpectedly, small- and medium-scale experiments are rare, as recently observed also by Zhang et al. (2018) [28]. To this aim, data on the mass evolution over time can be conveniently integrated by qualitative (e.g., images) and quantitative (e.g., thermocouples and pyrometers) analyses on the thermal aspects in the proximity of a pool fire [29]. The construction of a comprehensive database including these scales can be valuable for the evaluation and comparison of different numerical approaches, such as computational fluid dynamics [30]. Furthermore, soot data are almost absent for the analysed compounds [31] and are lacking for medium scale [32], promoting the implementation of a dedicated experimental campaign [33]. In this sense, the analysis of the most relevant reactions ruling the overall reactivity and the formation of by-products at low-temperature is paramount [34].

Under these premises, the identification of the sizes representative of credible accidental releases of LNG is an essential step from the safety and risk engineering perspective. In this sense, the accidental release of LNG during bunkering operation has been largely considered one of the most critical phases [6], taking into account their elevated likelihood to occur [52]. Furthermore, the transportation and bunkering processes are concerns of public authorities and the population [53,54], having the potential to generate vapour dispersion and pool fires [53]. To this aim, loading arms and flexible hoses are mentioned as typical transfer systems in the specific literature [6]. Hence, leak diameters varying from 5 to 20 mm, as well as a fully automatic emergency shutdown system, are considered; thus, a release of 30 s can be assumed and modelled following the current literature [55,56]. The full set of results is reported in the Supplementary Material (Table S1). However, it is worth noting that maximum D_p of 0.80 m and 1.40 m were obtained for release on the ground and water, respectively.

In this light, the experimental campaign reported in this work aimed to characterize reactive and non-reactive scenarios of small/medium scale (i.e., D_p < 1), which are most representative of common processes employed during bunkering or transfer of the fuel. Indeed, the collected measurements represent an essential database for the validation or modification of the existing models in the case of medium-scale accidental releases of LNG in poorly congested areas. The determination of mass burning rate and flame geometry in on-field tests can also be intended as a paramount step for the optimization of emergency response plans, since the gathered information also accounts for the possible spatial and temporal oscillations of the boundary conditions.

2. Methodology

The experimental tests reported in this work were conducted in collaboration with the Nuclear Biological Chemical and Radiological (NBCR) team of the Italian National Fire and Rescue Service (Minister of Interior). Either reactive or non-reactive scenarios resulting from an accidental release of LNG were tested in an on-field environment. The LNG was composed essentially of methane (89.48%v), ethane (8.76%v), propane (1.16%v), nitrogen (0.50%v), and a low amount of butane. The initial composition of the adopted LNG results in a density equal to 453.5 kg m⁻³ at a temperature of −160 °C. A preliminary investigation was devoted to the characterization of the most representative size for a credible accidental release of LNG. Based on the composition and boundary conditions of interests, the accidental release of LNG during bunkering operation (considered one of the most critical phases [6]) was characterized via state-of-art models [56]. To this aim, a vertical jet directed on the ground, wind velocity of 5 m s⁻¹, a neutral (D) stability class, a release of 30 s (i.e., a fully automatic emergency shutdown system was considered), and typical transfer systems were reported in the specific literature [6], and leak diameters of varying from 3 to 6 mm were assumed, following the values recommended by the literature [55]. The results of these calculations are reported in

Table 2. Estimation of pool dimensions in case of an accidental release of LNG during bunkering operations.

Transfer System	Size [in]	Pressure (P) [bar]	Substrate	Pool Diameter (Dp) [m]	Mass (m) [kg]
Loading arm	6	5.55	Concrete	0.80	8.4
			Water	1.40	5.4
Loading arm	12	7.55	Concrete	0.44	1.95
			Water	0.89	2.38
Flexible hose	2	12.0	Concrete	0.35	2.55
			Water	0.75	3.34

.

Under these assumptions, D_p is lower than 0.80 m in case of release on the ground and lower than 1.40 m for a release in water. Hence, the majority of experimental campaigns available in the current literature explore larger scenarios, possibly not representative of the bunkering operations. In this light, the experimental campaign reported in the following aimed to characterize reactive and non-reactive scenarios of small/medium scale (i.e., D_p < 1).

For all tests, the temperature field developed in the proximity of the cryogenic pool was measured by an arrangement of type-K thermocouples with an estimated accuracy of ±2.5 °C [57]. The acquisition rate variated between 6 and 30 samples per minute. In addition, visible and infrared (IR) digital images were recorded during the tests. A FLIR thermal imaging camera (model T1010) with an accuracy of ±2 °C and a frame rate of 30 Hz was employed at this scope. Using thermal and visible images, the flame characteristics (H_f, D_f, A_f, and δ) were analysed for the reactive scenarios. Furthermore, the burning rate per unit area ($m_{ev}^{″}$) was evaluated for the fire tests by assessing the liquid level evolution over time. For m″, data were filtered to remove noise and irregularities using a 100 points Savitzky–Golay (SG) method. The experimental procedure was designed to reduce the variability of atmospheric parameters within each test. However, small oscillations were observed (within 5% of the average value). For the sake of simplicity, only the average ambient conditions are reported in terms of wind velocity (u_w) and direction, ambient temperature (T_a), and relative humidity (

Table 3. Ambient conditions registered during the tests.

Test #	Ambient Temperature (Ta) [°C]	Humidity [%]	Wind Velocity (uw) [m s−1]	Wind Direction [°]
Test A	6.6	32	0.7	96
Test B	3.7	44	0.5	322
Test C	7.0	32	1.0	286
Test D	7.2	31	1.7	22
Test E	6.9	31	1.4	307

).

2.1. Dispersion Test

For the dispersion tests, a steel rectangular-shaped drip tray with dimensions of 0.750 m · 0.485 m · 0.011 m was employed. Hence, an equivalent diameter (D_p) of 0.68 m was considered. This diameter has also been chosen for pool fire, as reported below, based on a preliminary evaluation obtained using integral models [56]. Additional details are reported in the Design of Experiment section of the Supplementary Material.

The mass evolution over time was measured using a load cell with a capacity of 60 kg and a resolution of 20 g. Values were acquired with a frequency of 0.1 Hz. Consequently, using the geometrical characteristic of the drip tray, the evaporation rate per unit area ($m_{ev}^{″}$) was calculated as the first derivative of the measured mass divided by the liquid–vapour interface at any given time.

2.2. Pool Fire

The pool fire tests were conducted employing the same drip tray (D_p ~ 0.68 m) for the sake of comparison with the non-reactive scenario (i.e., dispersion). The relative position of the adopted thermocouples is given in

Table 4. Position of thermocouples adopted for the characterization of temperature distribution from pool fire tests expressed (x, y, z) coordinates considering the centre of the drip tray as the origin.

	V1	V2	V3	V4	H5	H6	H7	H8
x [m]	0	0	0	0	0.375	1.375	2.375	3.375
y [m]	0	0	0	0	0	0	0	0
z [m]	4	3	2	1	2	2	2	2

, where the coordinates are expressed as x, y, and z, with the origin coinciding with the centre of the drip tray. More specifically, the positions referred to as V_i are located on the axis of symmetry of the adopted box, whereas the points reported as H_i are located at z equal to 2 m and horizontally distanced by 1 m each from the edge of the box. Further information on this aspect can be retrieved in the previously published literature [58]. In addition, IR images were collected during the experimental campaign to quantify the main geometrical characteristics of the undetached plume.

3. Results and Discussion

In the following sections, the results of the experiments are presented. First, the results obtained by the evaporation test (Test A) are reported and discussed. Then, pool fire scenarios are analysed and compared with the non-reactive scenario and data from the current literature.

3.1. Evaporation

In Test A, a quantity of 8.68 kg of LNG was spilt in the drip tray, and the non-reactive scenario was investigated using the load cell (Figure 2).

Figure 3 shows the mass profile (m) and the evaporation rate ($m_{ev}^{″}$) as a function of time.

From a phenomenological point of view, in the first part of the test, a peak in $m_{ev}^{″}$ is observed. Indeed, a maximum value equal to 0.023 kg s⁻¹ m⁻² is obtained at the very beginning of the experiment.

The initial peak can be attributed to two factors. The first is driven by the mass transfer and is related to the initial absence or limited presence of a vapour cloud in the proximity of the liquid–gas interface that brings a reduction in the effect of the condensation term in the net evaporation rate. Indeed, several theories, including the well-known Lee’s model [59], account for the contribution of two phenomena to the net evaporation rate: condensation of vapour particles in the proximity of the liquid–vapour interface and evaporation of liquid particles in the proximity of the liquid–vapour interface. The second is driven by the contribution of heat exchange by conduction, which becomes negligible in the stationary phase. In this regard, detailed discussions on the effects of the initial wetting and cool-down of the containing vessel on the evaporation rate of cryogenic liquids can be found in the literature [60]. Quite obviously, this aspect is potentially largely affected by the thermal properties of the substrate considered in the analysis. Indeed, a large share of the heat exchange between cryogenic liquids and the surrounding environment is typically attributed to the conductive term [61]. Afterwards, a linear decrease of liquid mass representative of a stationary phase can be observed. As a result, a $m_{ev}^{″}$ value of 0.005 kg s⁻¹ m⁻² is obtained. This trend can be associated with the achievement of an equilibrium phase between the evaporation and condensation terms; thus, it is reached as soon as a stable layer is formed. This phase lasts approximately 40:00 min. After this phase, a slowing in evaporation is registered due to the consumption of most of the initial mass, enrichment in heavier species in the liquid pool, and reduction in driving forces ruling the evaporation term.

The value acquired in Test A is significantly lower than the $m_{ev}^{″}$ reported in the literature, even though a more conductive substrate was utilized in this work. Indeed, according to Luketa-Hanlin et al. (2006) [44], the mass evaporation rate $m_{ev}^{″}$ can vary between 0.029 kg s⁻¹ m⁻² and 0.195 kg s⁻¹ m⁻². In addition, several numerical studies have adopted 0.181 kg s⁻¹ m⁻² as a reference value, following a conservative approach [62]. The difference between data collected in this work and retrieved in the literature can be mainly attributed to the adopted substrate and pool size. Indeed, it should be noted that most of the evaporation data from the literature were acquired employing water or soil as a substrate [45,47,48,49,50,52]. Hence, a correction factor between two and three is recommended to correlate water and land evaporation rates [63] to account for the differences in heat transfer from the substrate to the cryogenic pool. As a way of example, DNV PHAST uses a factor of 2.5, as reported in Equation (2) [64]:

$$m_{ev,water}^{″}\approx 2.5·m_{ev}^{″}$$

(2)

3.2. Pool Fire

Four tests were performed for the pool fire scenario (Test B–E). In Test B, 16.5 kg of LNG was spilled into the drip tray and consequently ignited. The test lasted 8 min. An initial growing phase was detected within the first 10 s after the ignition, as can be observed in Figure S6 included in the Supplementary Materials, followed by a pseudo-steady state where reduced oscillations are reported for each geometrical feature of the produced flame. Due to the wind effect, the flame was tilted for almost the entire experiment duration, so the thermocouples were rarely enveloped by the flame (Figure 4).

It should be noted that the formation of black, solid particles (i.e., soot) cannot be observed in the upper part of the flame. This observation is particularly remarkable if the sooting tendency of LNG pool fire reported in the literature is analysed. More specifically, field experiments reported by Suardin et al. (2011) [65] show light and clear flame, whereas a significant formation of soot particles has been observed by Luketa and Blanchat (2015) [66]. Although the liquid composition is unquestionably a potential factor for the soot formation, and these values are not available for all the cited cases, large differences in soot tendency cannot be associated only with the composition since the variations are constrained by dedicated standards [67]. The main difference between the abovementioned experimental campaigns relies on the size of the investigated pool, namely, in the first case, a medium scale has been investigated. In contrast, a large scale has been tested in the latter case. Typically, the formation of carbonaceous particles in pool fires has been attributed to low mixing levels in the core area, which generates a possible dearth of fuel vapour or oxygen content for the oxidation of pyrolysis products [31]. The comparison of the obtained results with data from the literature demonstrates that this condition can be related either to the pool diameter or the addition of diluents (e.g., water, nitrogen, and carbon dioxide). In the former case, the temperature and concentration profile within the flame may locally generate conditions promoting the formation of soot, whereas in the latter case, the formation of soot particles can also be attributed to modifications in the reaction pathways either due to a direct or indirect role of diluent agents. More specifically, the presence of a diluent can modify the reaction rates of activation steps at low temperatures (i.e., hydrogen abstraction) by affecting the availability of typical abstracting agents (e.g., OH, H, and O) [68,69], with significant impacts on the associated branching ratios. This phenomenon can be ascribed to a direct role. Conversely, the indirect role of diluents can be related to chemical dilution (i.e., decrease in the concentration of reactants), physical dilution (i.e., heat sink action), or effective collisions (i.e., promotion of termination reactions as third body agents).

Due to the wind direction and the relative position between thermocouples and the pool, the measured temperature has never exceeded the value of 180 °C, registered 1 m above the drip tray (Figure 5). Similar data collected for different tests are reported in Figures S3 and S5.

Along with the horizontal profile, the recorded temperature value was always lower than 50 °C (Figure 5b). This is particularly relevant if the corresponding radiative flux is calculated. Indeed, under the conservative assumptions posed by the Stefan–Boltzmann law, the corresponding maximum heat flux (q_max) is lower than 2.4 kW m⁻² (the thermal radiation required to generate a 2nd-degree burn [70]). On the other hand, if the IR images are considered (Figure 6), temperatures up to 650 °C can be observed in a position not covered by the thermocouples, corresponding to q_max equal to 41 kW m⁻², which can turn out in domino scenarios [71]. In addition, a structure characterized by three areas can be identified. More specifically, a cold pocket in the proximity of the pool, followed by the core combustion-driven area, where maximum temperatures are reached, and hot gas diffusion can be observed. These results demonstrate that the wind direction strongly affects the produced flame and the resulting area of concern.

For the sake of brevity, temperature, IR images, and temperature profiles measured for Test C, Test D and Test E were reported in Figures S1, S2, and S4 included in the Supplementary Materials. Starting from the visible features of the flames collected in this work, morphology data of the different tests were collected every 20 s. The minimum, average, and maximum values for the main characteristics of the flame are provided in

Table 5. Flame morphology characteristics (Figure 1) measured during different tests carried out in this work.

		Hf [m]	Df [m]	Af [m]	$\left|\mathit{\delta }\right|$ [°]	m″ [kg s−1 m−2]	T [°C] (# Thermocouple)
Test B	Max	1.68	1.13	1.09	55	0.042	173 (V4)
	Average	1.06	0.71	0.63	33		-
	Min	0.61	0.46	0.29	3		-
Test C	Max	2.39	1.31	1.88	42	0.093	780 (V4)
	Average	1.59	0.85	1.15	13		-
	Min	0.59	0.53	0.48	0		-
Test D	Max	2.43	1.00	1.20	50	0.119	780 (V4)
	Average	1.39	0.62	0.54	12		-
	Min	0.31	0.23	0.01	1		-
Test E	Max	2.43	0.96	1.29	19	0.099	790 (V4)
	Average	1.62	0.57	0.70	6		-
	Min	0.35	0.14	0.03	0		-

. Generally speaking, starting from these data, it is possible to conclude that an LNG pool of D_p ~ 1 burns at the average rate of 0.100 kg s⁻¹ m⁻² and produces a flame height of 1.6 m with a diameter roughly equal to the diameter of the pool at a maximum temperature of 800 °C at which corresponds a q_max equal to 75 kW m⁻², which can lead to domino effects for pressurized tanks [71]. Regarding the burning rates, it is possible to affirm that, except for Test B, values are in the range of 0.093 to 0.119 kg s⁻¹ m⁻². This is in good agreement with the literature data [23,36,41,42]. In addition, it is possible to note that m″ of Test B is more than half that of other tests. Indeed, due to the wind effect, the flame was tilted. This caused a worse heat exchange between the proximity of the free surface of the pool and the flame. The result is that m″ assumes an intermediate value between burning and evaporation rates. Regarding L_f/D_p, except for Test B, a range varying between 2.0 and 2.5 could be roughly identifiable, which is in line with values reported in the current literature [35,39,41].

The mass burning rate values were compared with the evaporation rate collected in this work and reported in the previous section. It is possible to conclude that a factor of 20 should be considered to switch between a non-reactive to reactive scenario (Equation (3)):

$$m^{″}\approx 20·m_{ev}^{″}$$

(3)

Indeed, during a pool fire, the heat from the fire normally radiates back into the pool, generating the vapour that burns in a quasi-steady state. Hence, increased efficiency of the heat exchange is expected. When pool fires are on soil or ground, a negligible amount of heat is transferred by conduction [64]. In addition, it should be noted that the difference between m″ and $m_{ev}^{″}$ evaluated in this work is considerably higher than the literature values.

The suggested modifications in the evaporation rate and mass burning rate associated with a medium-scale pool fire of LNG can significantly impact the consequence analysis of the corresponding scenarios. Regardless of the analysed scenario, the duration to be considered in the analysis shall be extended for any initial quantity of released LNG.

Concerning the consequence, the reduction of the mass burning rate results in smaller emissive power, under the reasonable assumption that the fraction of energy radiated by the fire and the heat of combustion are constant [72]. Hence, the stand-off distances for pool fire of LNG shall be reconsidered based on a more accurate mass burning rate. Concerning the non-reactive scenario, a decrease in evaporation rate limits the occurrence of a rapid phase transition [73] and affects the design of mitigation systems [74].

4. Conclusions

In this work, evaporation tests and pool fires of liquefied natural gas (LNG) were performed in a drip tray, representative of a credible release during typical bunkering operations. The pool evaporation, in the case of the absence of ignition sources, has shown a rate value that is about 0.005 kg s⁻¹ m⁻², which is considerably lower than the value typically assumed for numerical analysis and approximately 20 times lower than the measured mass burning rate, which indeed was estimated as 0.100 kg s⁻¹ m⁻². Quantitative and qualitative information gathered throughout the experimental campaign described in this work were utilized to characterize the flame geometry resulting from an LNG pool fire.