1

LNG ACCIDENT DYNAMIC SIMULATION: APPLICATION FOR HAZARDOUS CONSEQUENCE REDUCTION

Biao Sun Kaihua Guo∗

SYSU-BP Center for LNG Training Education and Research

Sun Yat-Sen University, Guangzhou, China

KEYWORDS: LNG accident, dynamic simulation, consequence reduction

ABSTRACT

The definition of hazard area, centered inside the LNG stations, is essential for risk assessment in LNG industry. In this study, computational fluid dynamics (CFD) simulations have been conducted for the two main LNG hazards, LNG flammable vapor dispersion and LNG pool fire radiation, to determine the hazard exclusion area. The spatial and temporal distribution of hazard in complex spill scenario has been considered in CFD model. With the developed CFD code, the spray water curtains were studied as a shield to prevent LNG vapor dispersing. Two types of water spray curtain, flat curtain and cone, were analyzed to show their performance for reduction and minimization of the hazard influencing distance and area. The high expansion foam firefighting process was also studied with dynamic simulation of the foam action. The characteristics of the foam action on the reduction of LNG vaporization rate, vapor cloud and flame size as well as the thermal radiation hazard were analyzed and discussed.

1. INTRODUCTION

Large quantities of LNG liquid could give off cryogenic and flammable vapor once released. In LNG industry, safety concern for LNG station siting, construction and production has been raised highly because of potential risk of LNG release hazards [1, 2]. One of the most important factors in risk assessment is the prediction of the exclusive distance of inside and outside of LNG plant [3]. For public safety, the two typical hazards upon LNG release, LNG vapor dispersion and LNG pool fire radiation, have gained most attention [4].

Since 1960s, large-scale field tests of LNG release and its resultant hazards have been conducted to study the important physical phenomena [5, 6]. Specifically, experiments of LNG vapor dispersion, as well as LNG fires have been performed during later 1980s [7]. The experimental data have been used to validate mathematical models, including integral model and computational fluid dynamics (CFD) models. The CFD models, which include detailed description of flow physics, are more computationally intensive but are capable of modeling various phenomena adequately. The CFD models have gained increasing interest for analyses of LNG spill hazards. Sutton, Brandt, and White [8] were among the first to apply CFD method to simulate dense gas dispersion in a boundary-layer wind tunnel. Since 1990s, commercial CFD softwares, such as FEM 3 [9], FLACS [10], Star-CD, FLUENT [11] and CFX [12], have become popular and were widely applied to simulate LNG spill hazards.

This study deals with the spatial and temporal CFD simulations of the two typical LNG release resultant hazards, vapor dispersion and fire radiation. Besides, mitigation methods of LNG releases were studied, spray water curtain and high expansion foam working on LNG vapor dispersion and LNG pool fire, respectively. Commercial code FLUENT 13.0 was applied to establish the CFD model of LNG hazards.

∗ Corresponding author, email: guokaih@mail.sysu.edu.cn, tel: +86 20 39332893

2

Theory basis of these models can be referred to Ref [13].

2. VALIDATION OF CFD MODELING ON LNG HAZARDS

2.1 CFD model validation for LNG vapor dispersion Falcon series of LNG vapor dispersion field tests were performed in 1987 [14] by the Lawrence Livermore National Laboratory (LLNL). This study used the experimental data from Falcon tests to validate CFD model of LNG vapor dispersion. Another field test, Montoir series of LNG pool fire tests, performed in 1987 [15], was applied to validate CFD model of LNG pool fire. In the Falcon tests, LNG was released onto the surface of a rectangular water pond, 60 m long and 40 m wide. The water was recycled in order to fully vaporize LNG. A “billboard” stood in front of the water pond, which behaved as the LNG storage tank. The pond was surrounded by a fence, 8.7 m high, 88m long and 44 m wide, behaving as the impoundment wall.

Monin-Obukhov theory was applied to describe the atmospheric stability, in order to calculate the vertical profile of ambient wind velocity in the boundary layer [16]. A steady-state solution was first sought for the ambient velocity field, as shown in Fig. 1. Swirls were generated inside the impoundment area, which would influence the vapor dispersion.

Figure 1. Horizontal and vertical velocity field as initial conditions

t=10s t=150s t=400s （Falcon1: wind velocity 1.7 m/s, LNG spill rate 28.7 m3/min，temperature 28.7 ℃,

atmospheric stability C, spill duration 131 s）

Figure 2. Footprint of LNG vapor cloud dispersion in downwind distance (iso-concentration 2.5% (vol))

The time-dependent simulation was then performed, initialized by the steady-state simulation. Swirls dragged the dispersed cloud upwind at the beginning of LNG spill, shown in Fig. 2.

3

-60 -40 -20 0 20 40 60

0

2

4

6

8

10

12

14

16

Height (m)

Horizontal (m)

Downwind distance 150m, Time 180s

Falcon 1 (Experimental data)

-60 -40 -20 0 20 40 60

0

2

4

6

8

10

12

14

16

Downwind distance 150m, Time 180s

Height (m)

Horizontal (m)

Falcon 1 (Simulation results)

0.5

10

5.0

Figure 3. Experimental data (left) vs. predicted value (right) of gas concentration contours at 150 m downwind

Fig. 3 illustrated the comparison of vertical volumetric concentration between test and simulation, located 150 m downwind. The maximum value was almost consistent with experimental data.

0 100 200 300 400 500 600 7000

5

10

15

20 50 m downwind, 1 m above ground

Test Falcon 1

Wind velocity 1.7m/sTemperature 33.5℃

Spill rate 28.7m3/minSpill duration 131sAir stability G

Test value

Vol

umet

ric co

ncen

tratio

n (%

)

Time (s)

Predicted value

0 100 200 300 400 500 600 700 800

0

2

4

6

8

10

12

Wind velocity 1.7m/sTemperature 33.5℃

Spill rate 28.7m3/minSpill duration 131sAir stability G

Predicted value

150 m downwind, 1 m above ground

Volu

metr

ic Co

ncen

tratio

n (%

)

Time (s)

Test value

Test Falcon 1

Figure 4. Predicted vs. measured gas concentration at 1 m above ground, 50 m and 150 m downwind

Fig. 4 compared the evolution of the predicted value and measured value of gas concentration at 50m and 150m downwind of the impoundment, respectively. It showed that CFD simulation captured the general dispersion behavior of LNG vapor cloud.

2.2 CFD model validation for LNG pool fire radiation Among the different types of LNG fire, LNG pool fire happened in a higher frequency. The CFD model for LNG pool fire was verified by Montoir series of field test[15]. LNG spilled on a 35 m diameter insulated concrete dike. Tab. 1 listed the experimental conditions of three tests. Test 1 and 2 were applied to verify the CFD model of thermal radiation. Fig. 5 illustrated the temperature contours and radiant emissive power iso-profile in Montoir 1 test. A parameter study of flame length vs. iso-temperature surface was implemented and iso-temperature surface of 1100 K was defined as flame outer surface. Comparison of the predicted value of LNG pool fire characteristics and experimental data was listed in Tab. 2. The maximum relative error was lower than 10%.

Table 1. Initial conditions of Montoir series of LNG pool fire tests [15].

No. LNG Spill (m3) Component (% mole) Wind velocity (m/s) Temp. (℃)

1 238 CH4 90.33 C2H6 8.95

C3H6 0.342 N2 0.341 2.7-4.8 25

2 198 CH4 89.90 C2H6 8.70

C3H6 0.80 N2 0.42 7.0-10.1 21

3 196 CH4 90.10 C2H6 8.91 C3H6 0.64 N2 0.266 2.8-4.8 14

4

Figure 5. Temperature contour (left) and iso-radiant power distribution (right) of vertical central plane

Comparison of horizontal radiant power profile was demonstrated in Fig. 6. The downwind and crosswind fire radiant distance had a higher relative error. The total average relative error was 8.75%.

Table 2. Comparison of predicted value and experimental data in Montoir series tests

Montoir 1 test Montoir 2 test Montoir 3 test

Measured

data Predicted

value Relative error

（%） Measured

data Predicted

value Relative error

（%） Measured

data Predicted

value Relative error

（%） Flame length（m） 75.1±7.1 77.0 2.53 77.8±4.3 78.7 1.16 71.0±1.0 65.0 8.45 Average emissive power（kW/m2） 260.5±6.4 251.2 3.57 264.8±6.6 264.5 0.11 261.3±2.5 243.2 6.93

Flame tilted angle（°） 47.0±9.9 45.2 3.83 57.3±3.2 62.7 9.42 44.7±8.1 45.1 0.89

Experimental data Predicted value

Figure 6. Comparison of iso-radiant power in Montoir 2 test

3. MITIGATION METHOD ON LNG VAPOR DISPERSION

The mitigation of accidental LNG releases had gained a great concern in the LNG industry. Spray water curtain was one of the commonly used methods to block LNG vapor dispersion[17]. The mitigation effects of the two types of spray curtain, flat curtain and cone curtain were studied. The fan nozzle was capable to project a fan-shaped water curtain in 180º. According to the product specification from Angus [18], the flat curtain could cover approximate 24.4 m wide in crosswind and 7.6 m high. The other type of water curtain, 60º cone curtain, was constructed by a full cone spiral nozzle. The cone nozzles were usually designed in a line in the downwind of LNG vapor dispersion, which behaved as a porous block.

5

0 200 400 600 800 1000 1200 14000.0

0.5

1.0

1.5

2.0

2.5

3.0

D2/D1 =(P2/P1 )-0.3

q=k√P

Flow rate（

m3/min）

Pressure（kPa）

Flow rate (Flat) Flow rate (Cone)

Pipe diameter 0.0508m (2in)

0.00

0.25

0.50

0.75

1.00

1.25

1.50

1.75

2.00

2.25

Droplet (Flat) Droplet (Cone)

Wate

r dro

plet

diam

eter

0.1 1 10

101

102

103

104

105

106

107

108

109

1010

1011

1012

Resis

tanc

e coe

ffici

ent

Distance away from the spray center (m)

Viscous resistance (Flat) Inertial resistance (Flat)

Viscous resistance (Cone) Inertial resistance (Cone)

Pipe diameter 0.0508 m Outlet pressure 690 kPa

Figure 7. Flow rate and droplet size

vs. nozzle pressure Figure 8. Resistance coefficient

vs. water curtain size

The basic characteristics of the two types of nozzles were displayed in Fig. 7. The cone nozzles could create smaller water droplet. A semi-empirical correlation, Ergun equation [19], was referred to calculate the coefficients of viscous resistance, Cv, and inertial resistance, Ci, which were described in Eqn. 1.

2

2 3

1.5(1 )v

p

CD

εε−

= , 3

3.5(1 )i

p

CD

εε−

= （1）

Where Dp is the diameter of water droplet in curtain; ε is void fraction in curtain, which equals 1.0 at the spray center and equals 0.0 at the edge of curtain. The resistance coefficients were summarized in Fig. 8.

The steady-state velocity field was obtained first and the results were demonstrated in Fig. 9. Swirls were generated in both vertical and horizontal, which made the dispersed vapor to move towards the swirl’s center and mitigated the hazardous area to a large degree.

Cone water curtain Flat water curtain

Figure 9. Steady-state velocity field

Flat curtain Cone curtain No curtain

Figue10. Volumetric iso-concentration contours under different spill scenario

6

Fig. 10 compared the hazard influencing area of scenarios of flat curtain, cone curtain and no curtain, assuming the ambient conditions were the same. Iso-concentration surface of 2.5% was the most concerned, because exclusive distance would build at the downwind edge of this concentration, according to NFPA 59A (2009 Edition) [20]. The results showed that if there was no water curtain existed, the exclusive distance was 82.0 m downwind. The flat curtain helped to reduce the distance by 83.9 %, while the cone curtain reduced the distance by 61.0%.

4. MITIGATION METHOD ON LNG POOL FIRE RADIATION

Commonly used mitigation method in LNG industry was high expansion (HEX) foam, especially for reducing the hazard from LNG pool fire [21]. Previous experimental research had been showed that HEX foam with an expansion ratio of 500:1 had an optimum mitigation effects for LNG fire. Generally, HEX foam made the LNG burning rate reduced in a large amount. Considering of effect of the pool size, the burning rate could be written as [22]

( )Demm κβ−∞ −= 1"'' (2)

Where "m is the LNG burning rate (kg/(m2⋅s));

''m∞ is the maximum time-independent constant burning rate (0.08 kg/(m2⋅s)); D is the LNG pool diameter; κ is the absorption-extinction coefficient of flame and β is a mean-beam-length correction; κβ is equal to 1.1 [m-1] for LNG. For large pool, the wind may cause an increased burning rate. The formula suggested by Blinov and Khudiakov [23] may be applied, i.e.

+=

Du

mm

still

windy 15.01"

"

(3)

Where "windym and "

stillm are the wind-influencing and no-wind LNG burning rate; u is the wind velocity. A semi-empirical correlation was assumed to consider the burning rate when LNG pool was covered by HEX foam,

( )'' " 1 1 0.15vAt

D Vum m e eD

δκβ −−∞

= − +

(4)

δ is the HEX foam factor only related with foam expansion ratio; ν is the HEX application rate (m3/(m2⋅s)); A is the opening area of HEX foam generator (m2); t is the HEX foam working time; V is the LNG pool volume (m3). HEX foam factor δ was the key value for suppression of the burning rate of LNG pool fire. In a higher value δ, LNG pool fire was suppressed quickly and flame length was minimized in a short time. For foam expansion rate 500: 1, δ equaled to 30000 through comparing with Suardin’s experiments. The conditions of the experiments were summarized in Tab. 3. Test 3 and 4A were applied for simulation.

Table 3. Experiment conditions from Suardin’s study [21]

Test ID 1 2 3 4A 4B

Pool area (m2) 45 65 65 65 65

HEX foam application rate (L/(m2⋅min)) 10 3.5 7 10 10

Radiometer location from pool edge (m) 30 30 30 30 27

Average wind speed (m/s) 3.7 NA 1.2 2.2 3.7

Air temperature ( ℃) 15.8 NA 26.7 24.5 28.7

7

0 2 4 6 8 10

1000

2000

3000

4000

5000

Time (mins)

Predicted value

Experimental data

Heat flux data at 30 m from edge of LNG pool

HEX foam application rate of 7 L/(m2.min)

LNG pool area 65 m2

0

20

40

60

80

100

Heat

Flux

(W/m

2 )

Heat

Flux

(% o

f max

imum

)

Test 3

0 1 2 3 4 5

1000

2000

3000

4000

5000

6000

7000

Heat

Flux

(W/m

2 )

Test 4A

LNG pool area 65 m2

Heat flux data at 30 m from edge of LNG pool

HEX foam application rate of 10L/(m2.min)

Heat

flux

(% o

f max

imum

)

Time (mins)

Experimental data

Predicted value

0

10

20

30

40

50

60

70

80

90

100

Figure 11. Heat flux vs. time in the 65 m2 LNG pool fire with different HEX foam application rate

t=22 s t=40 s t=45 s t=1 min

t=2 min t=4 min t=7 min t=10 min

Figure 12. Evolution of LNG pool fire flame at different burning time

Fig. 11 illustrated the comparison of the predicted value and experimental data of Test 3 and Test 4A, respectively. Fire Control Time (FCT) was defined as the time required for 90% heat flux reduction, as specified in NFPA 11[24]. In the experiment, 90% heat flux reduction arrived at 100s and 60s, comparing of 130s and 55s in the simulation.

The sequence of plots in Fig. 12 demonstrated the pool fire flame changing with time in Test 3. The LNG pool fire started from t=0 s. At t=22 s, the fire was fully developed. Meanwhile, HEX foam was turned on. The fire flame length was declined quickly within 30 s. At t=4 min, the fire reached at a steady-state buring, with a consistant flame length. However, the fire was not extinguished by HEX foam although the time extended as long as 10 min.

5. CONCLUSION

In this study, LNG vapor dispersion and LNG pool fire radiation CFD models were validated through modeling historical field tests. The relative error of CFD model of pool fire radiation was lower than 10%. It concluded that CFD model could describe the complex terrain effectively. The two mitigation methods for LNG hazards, water curtain and HEX form, were studied. CFD simulation indicated that water curtains acted mainly as a porous barrier to block LNG vapor dispersion, which could reduce the hazardous area significantly. The effectiveness of HEX foam depended heavily on the foam expansion ratio and application rate. A new correlation was introduced to calculate the LNG pool fire burning rate. For foam expansion ratio 500:1, the HEX foam factor equaled 30000 through validating the experiments.

8

REFERENCES

[1] J. Havens and T. Spicer. United States regulations for siting LNG terminals: Problems and potential. J. Hazardous Materials, 2007. 140(3): p. 439-443.

[2] R.P. Koopman, D.L.Ermak, Lessons learned from LNG safety research. J. Hazardous Materials, 2007. 140(3): p. 412-428.

[3] P.K. Raj and T. Lemoff. Risk analysis based LNG facility siting standard in NFPA 59A. J. Loss Prevention in the Process Industries, 2009. 22(6): p. 820-829.

[4] Mike Hightower. Guidance on Risk Analysis and Safety Implications of a Large Liquefied Natural Gas (LNG) Spill Over Water. SANDIA REPORT.

[5] Luketa-Hanlin. A review of large-scale LNG spills: Experiments and modeling. J. Hazardous Materials, 2006. 132(2-3): p. 119-140.

[6] P. Cleaver, M. Johnson, B. Ho. A summary of some experimental data on LNG safety. J. Hazardous Materials, 2007. 140(3): p. 429-438.

[7] D. L. Ermak, R.C., H. C. Goldwire, F. J. Gouveia, H. C. Rodean. Heavy Gas Dispersion Test Summary Report. Lawrence Livermore Laboratory, UCRL-21210, 1988.

[8] Sutton, S. B., H. Brandt, and B. R. White. Atmospheric dispersion of a heavier-than-air gas near a two-dimensional obstacle. Boundary-Layer Meteorology, 1986. 35(1): p. 125-153.

[9] Luketa-Hanlin, R. P. Koopman, and D. L. Ermak. On the application of computational fluid dynamics codes for liquefied natural gas dispersion. J. Hazardous Materials, 2007. 140(3): p. 504-517.

[10] S. Dharmavaram, S. R. Hanna, and O. R. Hansen. Consequence Analysis—Using a CFD Model for Industrial Sites. Published online 3 May 2005 in Wiley InterScience.

[11] Gavelli, F., E. Bullister, and H. Kytomaa. Application of CFD (Fluent) to LNG spills into geometrically complex environments. J. Hazardous Materials, 2008. 159(1): p. 158-168.

[12] Qi, R., et al., Numerical simulations of LNG vapor dispersion in Brayton Fire Training Field tests with ANSYS CFX. J. Hazardous Materials, 2010. 183(1-3): p. 51-61.

[13] ANSYS (2010). ANSYS FLUENT Theory Guide. USA: ANSYS Inc.

[14] T. C. Brown, R.T.C., S. T. Chan, D. L. Ermak, R. P. Koopman, K. C. Lamson, J. W. McClure, L. K. Morries, Falcon Series Data Report: 1987 LNG Vapor Barrier Verification Field Trials. America: (LLNL) Lawrence Livermore Laboratory, 1990.

[15] D. Nedelka, J. Moorhouse, et al. The Montoir 35m diameter LNG pool fire experiments, Midlands Research Station, 1989.

[16] Khaled S. M. Essa, M.E., Soad M. Etman, A notional variation of the wind profile power-law exponent as a function of surface roughness and staility. 4th Conference on Nuclear and Particle Physics. Fayoum, Egypt, 2003.

[17] Rana, M.A., et al. Experimental study of effective water spray curtain application in dispersing liquefied natural gas vapor clouds. Process Safety Progress, 2008. 27(4): p. 345-353.

[18] Angus. Product specification sheet. Oxford shire, UK: Angus Fire Armour Limited, 2005.

[19] Ergun. S. Fluid Flow through Packed Columns. Chem. Eng. Progr,, 1952. 48(2): p. 89-94.

[20] NFPA 59A. Standard for the Production, Storage and Handling of Liquefied Natural Gas. 2009.

[21] Suardin, J.A., et al. Field experiments on high expansion (HEX) foam application for controlling LNG pool fire. J. Hazardous Materials, 2009. 165(1–3): p. 612-622.

[22] VYTENIS BABRAUSKAS. Estimating Large Pool Fire Burning Rates. Fire techonology.

[23] V. I. Blinov, and G. N. Khudimakov. Diffusion Burning of Liquids. U.S. Army Translation, NTIS No. AD296762 (1961).

[24] NFPA 11. Standard for Low-, Medium-, and High-expansion Foam, 2005.