1. Introduction

Owing to recent environmental issues, the IMO has adopted regulations to address the emission of air pollutants from ships and has adopted mandatory energy efficiency measures to reduce the emissions of greenhouse gases from international shipping under Annex VI of the IMO’s pollution prevention treaty (MARPOL). In addition, IMO regulates air pollutants from international shipping, particularly nitrogen oxide (NOx) and sulphur oxide (SOx) emitted from ships.

To comply with recent IMO requirements, LNG fuel has entered limelight as an effective solution (Bengtsson et al., 2011;Burel et al., 2013;Hoenders, 2013). Therefore, many vessels equipped with LNG fuel are being built in the world. LNG can be considered the most environmentally friendly fossil fuel because it has the lowest CO₂ emissions per unit of energy, and because it is suitable for use in high-efficiency combined cycle power stations (IMO, MEPC.212(63), 2012a;2012b).

For an equivalent amount of heat, the burning of natural gas produces about 30 % less CO₂ than the burning of petroleum, and about 45 % less CO₂ than the burning of coal. However, LNG is easily vaporized to 600 times the volume, and changes into a flammable gas when mixed with air. The flammable range is about 5–15 vol%. (GECF, Gas Basic, 2013)

Most traditional classification rules are detailed prescriptive requirements for specific types of equipment or designs that must be adopted or functional requirements that must be attained, for all installations classed under the rules. This gives very clear instructions on how to design these aspects of the installation. It implies that the responsibility for safety in these areas rests mainly with the classification society because the designer is simply required to satisfy the application rules. In general, such rules have been developed by expert judgment, responding to previous accident experience. They are only rarely based on a risk assessment, and do not themselves satisfy the requirements of performing a risk assessment (Chang et al., 2008).

Many high-pressure gas-fueled LNG vessels which are combined with a high-pressure fuel gas compressor or high-pressure pump/vaporizer have been built because LNG fuel is regarded as environmentally friendly and satisfies the IMO requirements; however, there are many reasons to examine the safety requirements such as for a gas leakage, explosion, or fire. Because conventional LNG carriers applied by the IGC code have been under sail for a long period of time, there are many adequate safety regulations. However, in LNG fueled vessels including newly built LNG carriers, there are some incomplete safety regulations because LNG powered vessels applied by the new IGC code do not have sufficient references. Most safety regulations of the new IGC codes are applied in accordance with the requirements of the IGC codes, and thus there are many inadequate regulations for preventing the risk elements in LNG fueled vessels. In particular, the gas detection system applied in the machinery room based on the IGF and new IGC codes simply defines the number of gas detectors. There are no rules for their locations, and thus the gas detectors are installed based on an agreement among the ship owners, shipyard, and classification societies. The minimum number of detectors in a machinery room (cargo compressor room) by the new IGC code is three, but there are no rules for its detection points. Therefore, ship owners do not rely heavily on the detection system of the new IGC code.

Recently, 174K ME-GI LNG carriers used for carrying cryogenic liquefied natural gas at -163℃ have been designed, and nearly 30 vessels with typical cargo handling equipment and piping arrangement have been constructed according to the new IGC code. Because the vessels use LNG as fuel, there are many reasons to examine their safety requirements such as for a gas leakage, explosion, or fire (MAN B&W Diesel, 2006;Paltrinieri et al., 2015). Thus, it is positively necessary to identify the optimum number of gas detection sensors and their locations to prevent or mitigate a gas accident. In this paper, we present a reasonable method for identifying the risk of an explosion and to determine the optimal location of gas detecting devices. To examine the above-stated points, an LNG gas dispersion simulation for a high/low-pressure leakage in the cargo compressor rooms of 174K ME-GI LNG carriers was carried out according to the volume flow rate of the leak and the new IGC code. The ventilation capability and the locations of gas detection sensor were verified through comparison between actual gas detection sensor and virtual monitor points.

2. Methodology

2.1 System configuration

To conduct the LNG gas dispersion simulation in a cargo compressor room, a 174K LNG vessel built by DSME in Korea was selected to have the same 3D size, not only for the equipment but also for the compressor room geometry, and the initial conditions and leak scenario were defined according to the pinhole sizes.

Fig. 1 shows the general arrangement of an ME-GI and re-liquefaction system in the cargo compressor room of an LNG vessel. The main fuel gas consumers are two sets of an ME-GI main engine and four sets of a dual fuel generator engine (DFG) located in the engine room. Fuel gas of 305 bar is supplied by a high-pressure fuel gas compressor or high-pressure pump/vaporizer. Re-liquefaction equipment located in the cargo compressor room, and a gas combustion unit (GCU), which is located in the engine room, is used in the case of excessive BOG treatment.

This research analyzes the optimal gas detecting system through the gas leak and dispersion of both a high-pressure leak and a low-pressure leak according to the varying scenarios that can occur. The ventilation capability in the room and the gas detection sensor locations were verified through a comparison between a real gas detection sensor and the virtual monitor points.

2.2 Leak scenarios and boundary conditions

The new IGC code (2016) specifies that artificial ventilation inlet and outlets shall be arranged to ensure sufficient air movement through space to avoid the accumulation of flammable, toxic, or asphyxiant vapors, and to ensure a safe working environment. In addition, the ventilation system shall have a capacity of no less than 30 changes of air per hour, based upon the total volume of the space (Kang, 2012). According to the new IGC code 13.6.12, the rule requirement specifies that for every installation, the number and positions of the detection heads shall be determined with due regard to the size and layout of the compartment. Therefore, we should consider the reasonable means of achieving the safety regulations, as well as the number of gas detectors and their location, specified in the code.

To realize the gas dispersion simulation, the actual physical properties of LNG gas were applied, as defined in Table 1. The LNG consumed in the model ship was provided by the Korea Gas Corporation (KOGAS). LNG can be obtained from different sources and may have different compositions, implying that the heat value is a variable.

LNG gas dispersion simulations were carried out in a cargo compressor room in accordance with the pinhole sizes, and the boundary condition was set to two (2) "pressure in" natural vents, seventeen (17) "pressure out" mechanical ventilators, and a "mass-flow-inlet" at the leakage points. Moreover, to create a realistic gas dispersion simulation, the actual physical properties of LNG gas were used.

In this research, the LNG gas leak scenarios consisted of a high-pressure leak and a low-pressure leak. Fig. 2 shows the virtual space, which was designed to be 28.5 m (Width) × 17.5 m (Depth) × 7.5 m (Height) based on the arrangement of a 174K LNG vessel equipped with a high-pressure fuel gas compressor for the ME-GI main engines and re-liquefaction equipment. A high-pressure leak was assumed at the high-pressure pump/vaporizer discharge pipe located partially on the deck at the middle of the cargo compressor room (arrow no. 1), and a low-pressure leak was assumed at the vapor return compressor discharge pipe located on the floor deck at the left of the cargo compressor room (arrow no. 3). The leakage scenario for high-pressure gas was considered for three (3) cases, with pinhole diameters of 4.5, 5.0, and 5.6 mm, and the transient flow calculation was carried out until 503s. The low-pressure leakage scenario was composed of two (2) cases, with pinhole diameters of 100 and 140 mm, and a transient flow calculation was carried out until 4,200s.

In Table 2, a pinhole size of 5.6 mm, described in Case 1 ~ 3, was assumed as a rupture case at the maximum capacity of the high-pressure pump/vaporizer, and the pinhole sizes in Case 4-5 are the discharge pipe of the vapor return compressor. One(1) of the gas detection sensors which is the nearest point 30 % LFL (Lower Flammable Limit) of the total four(4) sets was alarmed after the gas leak and then leaked gas was continuously discharged during 10 s and then stopped. Mechanical ventilators were continuously operated before and after the leak, and the methane gas behavior and ventilation capabilities were monitored in this study.

The formulas below are used to estimate the required effective leakage when the flow is subcritical. Under this condition, the formulas (Crowl and Louvar, 1990) used for the calculation of the leaking gas can be assumed to obtain the mass flow rate, dm/dt.

Here, dm/dt, c_d, A_h, P₀, W_g, R, and T represent the mass flow rate, leakage coefficient (c_d = 0.97), area of the pinhole, internal pressure of the pipe, molecular weight, specific heat ratio (c_p/c_v), gas constant, and gas temperature, respectively. In addition, the constant K in Equation (2) is related to the gas leakage rate.

In CFD analysis, ‘Ansys Fluent Release v17.2’ was used as simulation and calculation tool.

3. Results and discussion

The position of a gas detector in a hazardous area is very critical since the activation of safety systems and functions requires fast detection of the gas (Luketa-Hanlin et al., 2007). In addition to the recommended detector locations presented in the IGC code provides a table with gas detection main principles covering several areas such as the engine room and the compressor room. This code will not be presented in detail because this paper emphasizes on the factors which must be considered in order to find the optimized detector positions. Bafjord (2011) has examined the most suitable location for gas detection in offshore installations for oil and gas production and evaluated the effects on the functionality and reliability of the gas detection system. Using FLACS, he studied the physical factors that affect the optimum behavior of the exhaust gas with wind speed, wind direction, source of leakage, leakage direction, rate of leakage, gas composition and geometry. Because rapid detection of escape gases is one of the key requirements associated with the gas detection system, the detection time is an important factor in the reliability of the system.

Based on the Fig. 2, a total of 140 virtual points, as defined in Table 3, were selected, excluding real gas detection points 1, 2, 3, and 4. To analyze the CH₄ volume fraction, the number of virtual monitor points was five in the X-direction, four in the Y-direction, and seven in the Z-direction, and the number of virtual monitor points was from 5 to 144. Moreover, the naming in the Z-direction was from A to G, and thus 20 virtual monitor points were composed of seven X-Y planes.

3.1 Scenario 1: Gas detection positions for high-pressure leak

Fig. 3 shows a distribution plot of the CH₄ volume fraction for scenario 1 according to Table 2. The gas detection points made up a total four (4) sets, and a gas detection alarm is sounded at 30 % LFL (0.015). Thus, the gas cloud behavior was between 0.015 (15,000 ppm) to 0.05 (50,000 ppm). Moreover, all cases monitored showed similar gas behavior. A gas cloud was vertically positioned from the pinhole surface of the pipe and dispersed to the ceiling and walls. The gas cloud behavior was captured from the gas detection alarm up to 10s for each case. Their sizes are proportional to the leakage flow rate. All cases reached 30 % LFL (15,000 ppm) within 5s from the leak starting at gas detection point 1, and then gradually decreased, as shown in Fig. 4. The gas leakage stopped after 10s of the alarm. The CH₄ volume fractions at 7 and 1 m planes at the final measuring time all showed lower values of 10,230 ppm. The monitoring values for all gas detection points remained under 6,500 ppm after 508s. Fig. 5 shows the CH₄ volume fraction at after 10s (a) from the gas detection alarm, and 504s (b) for case 3. As shown in Fig. 4, for a high-pressure leak, the gas behavior for cases 1, 2, and 3 are similar, and thus case 3 with the highest leak rate was analyzed. The highest CH₄ concentration exceeding 80,000 ppm was found at virtual monitor points 84, 103, and 124, and the detected values were between 70,000 and 80,000 ppm. All of these points are located at the highest position, and it was verified that the gas leak at 305 bar under a temperature of 45℃ was distributed to the ceiling because the gas is lighter than air. Points 7, 8, 12, and 13 in Fig. 5(b) are located at the farthest and lowest positions from the ventilators, and they are on the right-hand side of the cargo compressor room because the leak point is located on the right-hand side. Points 32 and 53 are around the high-pressure fuel gas compressor, and it was verified that the remaining gas is not easily ventilated owing to a large obstacle with a size of 3,000 mm × 8,150 mm × 5,000 mm (W × D × H).

3.2 Scenario 2: Gas detection positions for low-pressure leak

The gas leak continued for 10s after the gas alarm sounded and stopped, after which only the mechanical ventilators were operated without a gas leak. The first gas detection alarm was first monitored at location no. 1 for cases 4 and 5, as shown in Fig. 6. Moreover, the highest CH₄ volume fraction at 10 s after the alarm was monitored at location no. 1 location for all cases. The final measurements at 2,312 and 4,697s for each case were conducted without the flammable concentration, and all measured values continued to be below 9,000 ppm. The leakage flow rate of case 5 is nearly two (2) times more than that of case 4, and it takes nearly two (2) times the time to ven comparing the final measurement time between cases 4 and 5. The transient calculation of case 5 is 1,910s, which is much more than in case 4.

Fig. 6 and 7 show distribution of the CH₄ volume fraction for scenario 2 of Table 2. The gas detection points and gas cloud were assumed to be the same as applied with the high-pressure leak, as already described. In Fig. 8, the gas cloud is vertically positioned from the pinhole surface of the pipe and dispersed to the ceiling and walls. The leak point is located partially under the deck, and the gas temperature inside the pipe is approximately -110℃. Thus, a comparatively heavier gas than a high-pressurized gas leak with a temperature of 43℃ at 305 bar cannot be easily dispersed to other spaces, and a dense high-flammable gas was found partially under the deck farthest from the ventilators. The gas cloud behavior was captured from the gas detection alarm up to 10s after the gas leak. Their sizes are proportional to the leak flow rate. The flammable gas cloud for two (2) cases between 50,000 and 150,000 ppm, at 500s in Fig. 7 (a) and 900s in Fig. 7 (b), shows separate plots, which remain at the farthest area from the mechanical ventilators. These areas are regarded as an accumulated flammable gas area, and thus we confirmed that additional gas detection devices are needed. All two (2) cases reached 30 % LFL (15,000 ppm) within 80 and 65s from the start of the leak, and there was no flammable gas cloud for the monitored gas detection points, after which the CH₄ volume fraction gradually decreased. In cases 4 and 5, the values are significantly decreased after 900 and 1,320s, respectively. The CH₄ volume fractions shown in Fig. 7(e) and (f) at 4 m are all lower values of 10,000 ppm.

Fig. 8(a) shows the CH₄ volume fraction at a gas alarm point (80s), 10s after the gas detection alarm (90s), and at 900s for Case 4. The highest CH₄ concentration at 80 and 90s is between 250,000 and 450,000 ppm for virtual monitor points 8, 13, 29, and 34. All of these points are located around the leaked position and at the farthest and lowest positions from the ventilators partially under the deck. It was verified that the leaked gas of 1 bar at -110℃ was distributed to the lowest position because it is heavier than the air. Points 46, 51, 65 and 85 with the highest CH₄ concentration exceeding 25,000 ppm at 900s are located at the farthest and lowest positions from the ventilators. Points 101, 116, and 121 show the highest CH₄ concentration at 2,312s, the values of which are between 12,000 and 14,000 ppm, as shown in Fig. 8(b). They are around the high-pressure fuel gas compressor, and it was also verified that the remaining gas is not easily ventilated the same as a high-pressure leak.

Fig. 9(a) shows the CH₄ volume fraction at a gas alarm point (65s), at 10s after gas detection alarm (75s), and at 1320s for Case 2. The highest CH₄ concentration at 65 and 75s is between 350,000 and 450,000 ppm at virtual monitor points 9, 14, 29, and 34. All of these points are located around the leaked position and at the farthest and lowest positions from the ventilators partially under the deck, similar to case 4. Points 45, 46, 51, 65 and 85 with the highest CH₄ concentration at 1320s are between 35,000 ppm and located at the farthest and lowest positions from the ventilators. In Fig. 9(b), points 101, 116, and 124 show the highest CH₄ concentration at 4,697s, with values of between 8,000 and 10,000 ppm. They are around the high-pressure fuel gas compressor, and the remaining gas was mainly accumulated the same as the high-pressure leak, and as low-pressure leak of case 4.

4. Conclusion

The position of gas detectors affect the time from a gas leak starts until the escaped gas is detected. Potential leak locations and directions of air currents should be taken into consideration before detector positions are determined.

According to the Facility Regulations (PSA, 2011) the placement of detectors shall be based on relevant scenarios and simulations or tests. The use of CFD simulations is one way to find the best suited gas detector positions. Requirements such as fast and reliable detection are strongly influenced by the position of the gas detectors.

The LNG gas leak and dispersion were analyzed at high and low pressures according to the pinhole size for a cargo compressor room of a 174K ME-GI LNG vessel. Scenarios for a gas leak were examined for high pressure at 305 bar and for low pressure at 1 bar. High-pressure gas leak scenarios were examined for 4.5, 5.0, and 5.6 mm pinhole sizes (case 1~3), and low-pressure leak scenarios were examined for 100 and 140 mm pinhole sizes (case 4 and 5). Transient gas simulations were adopted to obtain the values of various time steps.

Through this study, we identified the ventilation capability with relatively cold and heavy LNG gas from low pressure partially under the deck and compared it with hot and light LNG gas at high pressure. The quantitative data obtained through the numerical simulation will help our understanding of the risk factors based on the flow characteristics of not only an ME-GI LNG ship but also similar ships. High-pressure gas leak scenarios show that the cargo compressor room of a 174K ME-GI LNG will not pose a serious risk problem regarding the flammable gas concentration because a ventilation assessment for a 5.6-mm pinhole size as the gas rupture condition was verified. Low-pressure gas leak scenarios show that the flow rate of a leak for case 2 was nearly two (2) times greater than that of case 4. Case 5 takes nearly twice the time of case 1 to ventilate the CH₄ gas for all gas detection points when comparing the final measurement time. The ventilation time of case 5 is much more 1,910s than in case 1. Thus, the more leakage gas that is present, the more time it takes to ventilate the gas.

Moreover, in the case of a high-pressure gas leak and dispersion, real gas detection sensors should be moved to the right-hand side of the cargo compressor room at around the ceiling height. In addition, in the case of a low-pressure leak and dispersion, CH₄ gas is not easily ventilated partially under the deck and in front of the high-pressure fuel gas compressor. Therefore, it was identified that additional gas detection sensors should be applied, or existing gas sensors should be moved to these points.