Zhenyi Liu， Yuanyuan Ma,， Yuan Ren， Mingzhi Li， Pengliang Li and Song Wan

（1. State Key Laboratory of Explosion Science and Technology，Beijing Institute of Technology，Beijing 100081，China；2. QHSE Department，Sino-Pipeline International Company Ltd.，Beijing 102206，China）

Abstract: Regulator station is an important part in the urban gas transmission and distribution system. Once gas explosion occurs, the real explosion process and consequences of methane gas explosion in the regulator station were not revealed systematically. In this study, a full-scale experiment was carried out to simulate the regulator station explosion process, and some numerical simulations with a commercial CFD software called FLACS were conducted to analyze the effect of ignition and vent conditions on the blast overpressure and flame propagation. The experimental results demonstrated that the peak overpressure increased as the distance from the vent increased within a certain distance. And the maximum overpressure appeared 3 m away from the door, which was about 36.6 kPa. It was found that the pressure-time rising curves obtained from the simulation are basically the same as the ones from the experiment, however, the time of reaching the peak pressure was much shorter. The numerical simulation results show that the peak overpressures show an increase trend as the ignition height decreased and the vent relief pressure increased. It indicates that the damage and peak overpressure of gas explosion could be well predicted by FLACS in different styles of regulator station. In addition, the results help us to understand the internal mechanism and development process of gas explosion better. It also offers technical support for the safety protection of the urban regulator station.

Key words: regulator station；gas explosion；full-scale experiment；numerical simulation；overpressure distribution

Natural gas is widely used in industrial production as a kind of clean energy source in the 21st century due to its low emission and high energy efficiency[1?3]. While regulator station is a facility used to regulate and stabilize the pressure of the urban gas pipe network[4], an important link connecting the gas source and thousands of households[5]. Actually in Chinese cities, some regulator stations often located in the center of the residential area. However, gas leakage and explosion occurred due to some factors such as pipeline corrosion, valve failure, external force,etc. Therefore, some efforts should be made to estimate the damage effect of gas explosion in the regulator station.

Actually, the impact of gas explosion is related to the concentration, activity, uniformity of the gas cloud, and also concerned with the chamber size, ignition location, initial pressure and temperature[6?7], which made it difficult to estimate the gas explosion hazards. Some analytical and empirical methods have also been developed to predict and evaluate the gas explosion[8?14].Meanwhile, standards EN-14994(2007) and NFPA-68(2013) also propose some prediction methods and prevention measures to reduce the hazards caused by overpressure rise. However,the application of mathematical models and empirical based guidance are not always appropriate to any kind of gas explosion because of its complexity.

Experimental study is the most effective way to estimate the explosion hazards and verify the numerical model. However, most scholars have focused on gas explosion tests in small pipes,tanks or enclosures[15?23]. The experiments’ sizes are so small that they can not provide representative data for regulator stations, while large or even full-size gas explosion experiments of methane have been hardly carried out. Bauwens et al.[24?25]used stoichiometric propane/air and methane/air mixture for the gas explosion test in a 63.7 m3vented confined space. Besides, Tomlin G et al.[26]carried out 9 m × 9 m × 4.5 m methane/air gas explosion test of different obstacles. Cao Yong et al. studied the effect of some important parameters on the internal pressure histories in a square vessel in 2018[27]. Experimental studies have found that reducing the vent size and increasing obstacles may accelerate the flame speed, which leads to a cumulative increase in overpressure. But few scholars have conducted corresponding research on gas leakage and explosion in regulator stations.

Above all, the experimental data obtained by large-scale vented confined gas explosion for the research of confined gas explosion hazards analysis is limited. Gas explosion simulation for the overpressure forecasting and calculation by the numerical commercial CFD software FLACS is accurate and mature[28]. In many studies, numerical simulations of gas explosions using FLACS have been validated to simulate the process of gas explosion and overpressure rise well[29?30]. Fan Xudong[31]simulated natural gas explosions in restaurant kitchens using software FLACS, and found that the peak overpressure was the greatest when the gas concentration was about 10.5%, almost the same at all locations in the room. A Floating Production Storage and Offloading(FPSO) flammable gas explosion model was established based on FLACS[32]. Finally, a method for determining explosion scene was proposed based on leakage scenario. Sun Qingwen[33]has studied the characteristics of gas explosion load in urban utility tunnel. Using FLACS, the numerical simulation of the explosion process of multiple combustible gases in confined space was carried out by Zhang Qun[34]. These studies all verify the feasibility of simulating combustible gas explosion by the software FLACS.

In order to study the explosion effect and overpressure distribution law of the regulator station, a 1:1 experimental building model of a real regulator station was constructed in this paper.Meanwhile, the experimental data were used to validate the results of FLACS simulations that would be applied towards the development of the vented methane-air explosion overpressure calculation. It provides a guarantee for the safety operation of urban regulator station, and forms a basis for the formulation of surrounding control scope and relevant standards’ revision. There is a great hope to reduce damage and hazards of gas explosion, and it has important and practical significance for countries and enterprises to develop disaster prevention and mitigation measures.

1 Experimental Gas Explosion Testing

1.1 Experimental details

We performed three full-scale explosion tests with stoichiometric methane/air in a 129.276 m3explosion chamber(Fig.1).The size of explosion chamber is 5.7 m×6.3 m×3.6 m and its thickness is about 15 mm, which is surrounded by some angle steel for support. The top of the chamber was constructed of five pressure relief panels of 1.2 m×0.6 m，with the failure overpressure of 50 kPa, which were fixed on the ceiling by bolts. The right face of the chamber was equipped with a 2 m×0.8 m door. The fore and rear face accommodated 4 windows symmetrically. Two kinds of pressure transducers of FPT and FPG were used. Five blast-pressure transducers were installed right to the door and windows to record the overpressure-time history(Fig.1), which were installed to stand (positioned perpendicular to the door and windows to measure overpressure) probably 1.8 m above the ground level. The explosions were also photographed by a high speed camera (2 000 frames per second) and Go Pro. Two electrodes were placed in the center of the regulator station and connected to the ignition trigger through the ignition line. The electric igniter was attached to the two electrodes and could release approximately 10 J of energy after igniting.

Fig. 1 Top view of the chamber, showing the locations of blast-wave pressure transducers (circles), and the ignition locations (triangles)

The initial mixture was supplied by injecting pure methane from the below of the chamber through gas pipeline connected to the methane gas cylinder. The methane cylinder is 10 MPa and was passed through a pressure reducing valve. The volume flow rate of methane was controlled by a flow meter and the gas/air concentration was measured by Altair5X gas concentration instruments in real time. In every experiment, six gas concentration instruments were used, one in the center of the chamber near the ignition while another five was put on the ground evenly. Once the gas/air concentration reached the stoichiometric limit, the pressure reducing valve of gas cylinder was closed and the pressure regulation station was allowed to become quiescent.

1.2 Experimental results and discussion

Fig.2 showed some of Go Pro snapshots of flame propagation from the outside. During the initial confined stage of gas explosion, the inner pressure was lower than the failure of the pressure relief panels, the flame front spreaded nearly spherically centering on the ignition location(Fig.2a). The flame surface was nearly smooth,which illustrated that the flame burning speed was laminar. Due to the output of burned gas volume, the pressure inside the chamber firstly reached the pressure failure of the ceiling relief panels, and vents began to depart from the enclosure, allowing the unburned methane-air mixture to escape from the vessel. As shown in Fig.2b, the flame front had not yet propagated to the ceiling when the ceiling pressure relief panels were vented. It was proved that the propagation velocity of the overpressure was higher than combustion flame front. Simultaneously, the original spherical flame smashed spark to ignite unburned gas around, forming a small new spherical flame closely and then both flame fronts crossexpanded. The effluence of the unburned gas distorts the flame front from its original spherical shape, extruding toward the venting direction(Fig.2c).

Fig. 2 Flame development process in the explosion from Go Pro

As the flame surface area continued to increase, the net growth rate of volume by the gas burning exceeded the volume outflow rate through the vent and the internal pressure gradually rose. After that, the pressure exceeded the pressure failure of the door and door was pushed out in the Fig.2d. Meanwhile, some irregular small scale cellular structure were present on the flame surface depending on Taylor instability and Helmholtz oscillation. Taylor instability[35]was introduced when low density burned production entered into denser unburned mixture, accelerating pressure growth process through addition of the flame surface area. Generally, Helmholtz oscillation was amplified largely by Taylor instability and both acted in the same phase. The pressure transducers in the experiment were equidistantly arranged and different overpressure-time curves were recorded. In Fig.3,the pressure profile, recorded on pressure transducers ofP2,P3andP4, were shown for the experiment. Pressure sensorsP1andP2were located 2 m and 3 m from the door, in the meanwhile,P3andP4were located 1 m and 2 m away from the windows. The peak overpressure ofP3andP4were 13.9 kPa and 18.7 kPa, at 1.239 s and 1.277 s, respectively. It was obvious thatP3reached the peak pressure earlier thanP4, as the flame firstly propagated toP3and then passed throughP4. As the distance ofP3andP4was 1 m, the flame propagation speed was calculated 26.3 m/s. However, the peak overpressure ofP4was greater thanP3outside the window (verified in the study of Tomlin G in 2015), because the peak overpressure was mainly affected by the flame propagation speed rather than the distance from the vent in this circumstance. The gas explosion could be treated as a process of flame acceleration in confined space, inner pressure and temperature increasing intensely because of the expansion of combustion products. When the window was broken, internal unburned combustible gas rushed out from the vent at a higher speed and formed gas cloud further away, resulting in a higher flame propagation speed at the pressure transducer ofP4.

Fig. 3 Pressure-time profile of the monitor points

Similarly, a similar phenomenon occurred outside the door, in which the maximum pressure ofP2was also greater thanP1. In general,the peak pressure increased with distance within a certain distance outside the door and window.The peak overpressure ofP2was 36.6 kPa, appeared at 1.118 s owing to the door venting,which corresponded to the moment of door failure in Fig.4a. In addition, the time interval of peak pressure betweenP2andP3was about 0.121 s observed from the pressure-time curves.Compared with the time of door yielded and window broken in the high-speed camera, we could find that the door yielded and was pushed out firstly, and then the right side of window broke,thus the time interval between both was approximately 0.115 s, close to 0.121 s.

Fig. 4 Photographs from high speed photography

The overpressure-time history ofP4was shown in Fig.5 below, three distinct pressure peaks were evident,Pv1,Pv2andPmaxrespectively at 0.714 s, 1.145 s and 1.273 s. When the inner pressure reached the venting pressure of windows, the right side window ruptured firstly observed from a high-speed camera. A small size of break at the beginning, which caused a small amount of unburned gas rushed out of the vent,forming a certain volume of gas cloud in the external space. The flame was also ejected at the same time, igniting the external gas cloud and forming gas explosion, which caused the pressure outside to rise slowly to 1.19 kPa firstly, and(dp/dt)1was about 1.66 kPa/s.

Fig. 5 Pressure-time profile for P4

Then the right side of the window was further broken and the split was further enlarged,allowing more gas to escape the window vents to create larger range of gas cloud, which pressure rose to a greater valuePv2, about 8.91 kPa,(dp/dt)2was calculated 17.8 kPa/s, much larger than (dp/dt)1. As gas explosion progressed, the flame propagation speed and the flame area were further increased. Then the window on the left was also broken affected by the combination of the external and internal explosions, resulting in the volume of the external unburned gas further increased. Ultimately, the rate of pressure rise increased greatly, and in just 0.123 s, the pressure rose from 8.91 kPa to 18.70 kPa, with the rate of pressure rise about 79.60 kPa/s.

2 CFD Simulation and Validation

2.1 Basic methodology

The commercial software FLACS, specialized CFD solver for the prediction of overpressure of methane-air mixture explosion was used in this study. The FLACS is mainly targeted at simulating dispersion and gas explosion. Many scholars use FLACS to simulate gas explosion,for good validation of vented gas explosion,which has a reference for the prediction of accidental gas explosion. Thek-εturbulence model is used in FLACS[36]. This section describes the mathematical model for compressible fluid flow used in FLACS. Conservation principles have been applied to the following quantities in order to derive the conservation equations:

? Mass

? Momentum

? Enthalpy

? Mass fraction of fuel (or products)

? Mixture fraction

? Turbulent kinetic energy

? Dissipation rate of turbulent kinetic energy

2.2 Numerical modeling and comparison with experiments

The strength of methane-air explosions depends on a series of parameters, such as type and composition of fuel, vent area, ignition location,panel relief pressure, obstacles, internal layout and so on[37]. So it is critical to establish accurate physical models and set reasonable parameters for simulation[38]. Through the survey of a regulator station near residential buildings, a physical model was established on an equal scale firstly. The model of regulator station was shown in Fig.6, with a main chamber of 5.7 m × 6.3 m ×3.6 m (L×w×h) and three walls outside. Fig.7 showed the location of ignition. Concerning both the sequence of pressure relief panels’ release in actual experiments and Chinese building standards— “GBT 7106—2008 airtight, watertight,wind pressure resistance classification and testing method of building doors and windows” —failure values of pressure relief are present in Tab.1,besides some details of ceiling pressure relief panels, door and windows are shown in Fig.8. The vent was set as the POPOUT pressure relief panel, yielded and discharged the pressure when the internal pressure forced on the panel exceeds the certain limit.

Fig. 6 Physical model

Fig. 7 Ignition location

The boundary condition was Nozzle. The ambient temperature was 298 K, the initial pressure of chamber was 101.325 kPa and the total number of grid was 164 700. The unit grid was0.2 m, which not only ensured the calculation accuracy, but also shortened the calculation time,simultaneously meeting with the FLACS recommended guidelines for explosions[8]. Ceilings and walls were assumed to be rigid during the simulated explosion. The chamber was filled with methane, at equivalent ratio concentration of approximately 9.5%. The computational domain for scenarios of gas explosion was shown in Fig.9. An equal proportion scale model in this numerical simulation was adopted for keeping the impacts of domain boundaries such as pressure reflection smaller. The sensors were put in the domain to obtain the pressure-time and temperature during the numerical calculation, shown in Fig.10.

Tab. 1 Details of pressure relief panels

Fig. 8 Details of the pressure relief panels

2.3 Calculation results and analyses

The overpressure-time histories were extracted from monitor points outside the chamber at 1.8 m height for the gas explosion in FLACS.The overpressure-time curves of pressure transducers were shown in Fig.11. Pressure sensorsP1andP2underwent the peak of 25.4 kPa and 32.7 kPa, respectively andP3andP4’s peak overpressure were 16.8 kPa and 21.6 kPa. AndP1reached the peak earlier thanP2,P3earlier thanP4, besides peak overpressure of each pressure sensor was consistent with experimental results.In order to get access to compressive pressure distribution of gas explosion, another two pressure transducers were put in the numerical simulation (P5, 3 m away from the window;P6, 1 m away from the door). The peak overpressure ofP5was 17.1 kPa, while the pressure ofP6was 12.4 kPa. As a result, the peak overpressure first increased and then decreased with the distance outside the window, ultimately the maximum pressure was obtained at 2 m away from vents.

Fig. 9 Computational domain

Fig. 10 Monitoring point layout

Additionally, the specific values of peak overpressure both in the experiment and numerical simulation were present in Tab.2 and the deviation were calculated. In general, the peak value deviations of pressure monitoring point arecontrolled within approximately 20%, which indicates that the physical model established by FLACS is reasonable and can be used for the effective prediction of the peak pressure in the vented confined methane explosion, which has the same characteristics as the regulator station.

Fig. 11 Pressure-time profile for FLACS numerical simulation

Tab. 2 Comparison of experimental and numerical simulation data

The numerical results of overpressure-time history compared with experimental data were shown in Fig.12. The overall trend of pressure rise and peak pressure obtained were quite consistent. However, the time that pressure reaching the maximum in the actual experiments was more than one seconds, longer than the time for numerical simulation. Based on the ideal state of gas explosion in the numerical simulation, the heat and energy generated could propagate so quickly that the pressure at monitoring points would rise in a shorter time.

Fig. 12 Comparison of overpressure-time curves

The gas explosion generated a large amount of heat and energy, therefore the pressure rising sharply in the chamber with the burning area and speed of flame further increased. The rise of pressure further disturbing unburned gas flow field would improve the burning speed and flame area further. Finally, a positive feedback of the pressure rising and the speed and area of flame has formed.

Apart from the effect of ideal energy release due to the flame acceleration mechanism, door and windows were set as pressure relief plates in the simulation, the internal pressure was immediately and completely released once the inner pressure reached the failure. But in actual experiments, windows were broken starting from a small crack, slowly expanding into a bigger vent,which also contributed to the slow rise of experimental pressure. Due to the combined effects of different manners of energy release and pressure relief, the experimental pressure rise rate was smaller.

Based on the consistency of numerical simulation and experimental results, some details of flame and temperature contours at height of 1.8 m can be provided in Fig.13. After ignition,flame and temperature expanded and spread out in a spherical shape (Fig.13a), besides the center temperature was the highest, forming a certain temperature gradient from the center to the surface. This forced the unburned gas ahead of the flame surface, therefore the flame constantly accelerated outward from the ignition point.Meanwhile, the burning gas moved rapidly throughout the chamber, causing a substantial increase in temperature (Fig.13b). Due to the pressure relief of south window, the combustible gas spread outward through the window and inner temperature around vents rose sharply(Fig.13c). The process of temperature diffusion agreed with the experimental high-speed photography, door was pushed out (Fig.13d) and left window ruptured, seen in Fig.13e. Ultimately,the entire chamber was maintained at a higher temperature (Fig.13f). In the numerical simulation, inner temperature reached 2391 K calculated as ideal state. However, in actual conditions, heat generated by gas combustion not only exchanged with the outside, but gas combustion reaction itself also dissipated part of the heat in the form of heat radiation. Therefore, the temperature would be slightly lower in actual gas explosion experiments.

3 Effect of Parameters on the Overpressure

The effects of ignition height and vent failure pressure on the overpressure were studied in the numerical simulation to study, which is significant for the study on internal influence mechanism of vented confined gas explosion.

3.1 Effect of ignition height

In the FLACS numerical simulation, gas was ignited at different heights at center of the room for the study of effect of ignition height on the overpressure and explosion hazards. In Fig.14,the overpressure-time curves of different ignition heights were shown for vented gas explosion. It could be seen that the overall trend of overpressure generated performed consistent at different ignition heights, however, the peak pressure that can be achieved decreased significantly with the ignition height. The higher the ignition height,closer to the ceiling, the faster the pressure was transmitted to the ceiling pressure relief plate,which resulted in a faster pressure relief in the explosion process and a drop in the peak of overpressure.

The pressure transducerP1was found to be sensitive to the ignition height, the degree of pressure drop was larger thanP2. Meanwhile, the time thatP1’s peak pressure seemed to push back as ignition height increased, which could not be found in pressure curve ofP2. As the ignition height increased, the peak overpressure outdoor decreased correspondingly shown in Fig. 15.

3.2 Effect of vent failure pressure

Fig. 13 Temperature contours at the typical times (Z=1.8 m)

Ventilation is an effective way to reduce damage to the building when gas explosion occurred. The pressure in the chamber could relieve once the vent (a fragile structure such as the door or window) was broken or popped.Therefore, studying the effect of failure pressure of vents on gas explosions is meaningful for reducing disasters and losses. When the door’s failure pressure changed, it would inevitably affect the combustion rate and pressure built-up process in the gas explosion. According to Chinese building standard—“GB 50016—2014 Code for design of building fire protection”, the area of pressure relief is specified, which can be calculated by the following formula

Fig. 14 Pressure-time curve of different ignition locations

Fig. 15 Effect of ignition height on overpressure

Moreover, the quality of lightweight roof panels and walls should not exceed 60 kg/m2.However, pressure release from the vent is not only related to the vent area, but also closely related to relief failure pressure decided by material and thickness, which is not mentioned in the standard. Thence, research on the impact of vent relief failure pressure on the gas explosion can provide a basis for the corresponding specifications of the explosion prevention.

Numerical simulations with different relief failure pressures of ventilation were carried out in order to study the relationship between peak overpressure and relief failure pressure of the vent. In Fig. 16, the curves of peak overpressure were present under different relief failure pressure of the door. The peak pressure ofP2was the highest, which is higher thanP1andP0. This was consistent with the experimental results that the farther away from the vent, the larger the pressure was within a certain distance range. As the relief pressure of the door increased, the overpressure of the pressure ofP0(1 m away from the door),P1andP2outside the door also increased.Before vents such as doors and windows were removed, the gas explosion could be seen as a gas explosion in a confined space, resulting in pressure increasing over time until venting occurred.The increase of door’s relief pressure might delay venting, which caused higher burning speed and flame propagation velocity inside the chamber.Once the door yielded and popped up, the unburned gas flushed out at a higher speed causing the value of peak overpressure outside the door to increase.

Fig. 16 Effect of the vent failure pressure of door on overpressure

4 Conclusions

Results of vented confined explosion tests obtained in a vented enclosure with stoichiometric methane-air mixture have been presented,which includes the process of flame propagation and pressure relief, as well as the overpressuretime curves of external pressure sensors. The experiment results demonstrated that it is possible to generate the overpressures of 24.1 kPa and 36.6 kPa at 2 m and 3 m away from the door,13.9 kPa and 18.7 kPa at 1 m and 2 m away from the window. Within a certain distance, the farther away from the vent, the higher the overpressure value. That’s due to the reason that the flame propagation velocity ofP1was slower thanP2, which reveals the tendency of the flame propagation speed to increase with the distance.And the same phenomenon occurred, in which the peak overpressure ofP3was lower thanP4outside the window. The peak overpressure at a closer distance was smaller, however, the greater the impact of flame radiation and fragmentation,which meant that peak overpressure, flame radiation and fragmentation should be all taken into consideration on the hazards to surroundings.Moreover, results of the numerical simulations show a good agreement with the experimental results. The peak overpressure ofP1,P2,P3andP4were 25.4 kPa, 32.7 kPa, 17.0 kPa and 21.7 kPa,respectively. The deviations ofP1,P2andP3were controlled within 20%, which demonstrated that it was feasible to numerically simulate the peak overpressure and hazards of gas explosions in a ventilated confined space by FLACS, providing powerful data support and a prediction method for different forms of regulator station. In addition, data support was provided for further mechanistic studies of methane gas explosions in large, confined spaces.

Through several numerical simulations, it was found that the overpressure peak decreased as the ignition height increased at the center of enclosure. The higher the ignition location, the faster pressure relief in the chamber and a drop in the peak of overpressure. According to the damage threshold and effect to buildings and human body caused by the overpressure of gas explosion (Tab.3), the overpressure over 20 kPa would cause wall cracks of the buildings and human moderate injury. As the experimental simulation results showed the peak pressure ofP2exceeded 20 kPa when the ignition height ranged from 1.0 m to 2.2 m, which meaned that at least 3 m outside the door might be a slightly safe distance. In addition to the possible damage effect caused by flame radiation and fragmentation, appropriate margins should be left to ensure personal safety and reduce costs of construction and maintenance.

As the door’s pressure relief failure increased, the pressure outside the door increased.The increase of the venting pressure relief failure delayed venting of the door, thus the internal combustion speed, flame area and propagation speed reached so high that the overpressure in the chamber would get a higher value. As far as we know, the failure of pressure relief is relatedto both vent size and pressure relief failure in the real gas explosion process. Under the same size of the vent, the vent composed of different materials will correspond to different pressure relief failures, which would affect the process of gas expansion, flame propagation and pressure formation. However, Chinese building standard –“Code for building fire protection design” has been just specified the size of the pressure relief panel for the explosion-proof and venting. Therefore, it is recommended to increase the specification of the pressure relief material and classify the degree of the pressure relief failure within the standard.

Tab. 3 Explosion overpressure damage to buildings and damage to human body

Combined with the overpressure analysis of experiments and numerical simulations, it can provide a foundation for the distance between the wall, sidewalk or residential building from the regulator station. Considering the peak overpressure, the flame radiation and fragmentation damage caused by gas explosion at the regulator station, it can provide a basis for the development of the safety protection of the urban regulator station and the scope of the surrounding control and the revision of relevant standards. It is of great practical significance for enterprises to develop disaster prevention and mitigation control measures for the country.