Académique Documents
Professionnel Documents
Culture Documents
vt
V$r v
!
v
!
Vp V$s r g
!
F
!
(2)
s m
_
_
V v
!
V v
!
T
_
2
3
V$ v
!
I
_
(3)
where p is pressure, s is stress tensor, F
!
is the sum of body forces,
r g
!
means gravity works on dense gas cloud in vertical direction, m
is dynamic viscosity.
Energy equation
vrE
vt
V$ v
!
rE p V$
_
_
k
eff
VT
j
h
j
J
j
!
s
eff
$ v
!
_
_
S
h
(4)
E h
p
r
v
2
2
(5)
where E is total energy, k
eff
is effective conductivity, S
h
is energy
source (if chemical reaction exist).
The choice of turbulence model is a key in dispersion simulation
using CFD codes. In this study, realizable k model (Shih, Liou,
Shabbir, Yang, & Zhu, 1995) is applied, because this model
satises certain mathematical constraints on Reynolds stress, and is
consistent with the physics of turbulent ows. Besides, this model
exhibits an excellent performance in capturing the phenomenon of
gravity slumping associated with dense gas dispersion, and is
superior to the standard k viscosity model (Shih et al., 1995). It
has been reported that realizable k predicts the spatial and
temporal concentration prole of the vapour cloud in the presence
of obstacles more accurately than the hitherto oft-used standard
k model (Launder & Spalding, 1972; Tauseef, Rashtchian, &
Abbasi, 2011; Tauseef, Rashtchian, Abbasi, & Abbasi, 2011). Realiz-
able k model differs from the standard k model in two ways.
1) The realizable k model contains a new formulation for the
turbulent viscosity and a neweddy-viscosity formula involving
a variable C
m
, shown in Equation (6), whereas C
m
is a constant
and equals 0.09 in standard k and RNG k model.
2) A new transport equation for the dissipation rate has been
derived from an exact equation for the transport of the mean-
square vorticity uctuation, shown in Equation (8).
m
t
rC
m
k
2
(6)
vrk
vt
v
_
rku
j
_
vx
j
v
vx
j
_
_
m
m
t
s
k
_
vk
vx
j
_
G
k
G
b
rY
M
S
k
(7)
vr
vt
v
_
ru
j
_
vx
j
v
vx
j
_
_
m
m
t
s
_
v
vx
j
_
rC
1
S rC
2
2
k
v
p
C
1
k
C
3
G
b
S
(8)
where m
t
turbulence viscosity, k is kinetic energy, is turbulence
eddy dissipation, G
k
and G
b
are the turbulence production due to
viscous and buoyancy forces, S
k
and S
m
t
Sc
t
_
VY
i
D
T;i
VT
T
(10)
where Y
i
is Mass fraction of species i, J
j
!
is Mass diffusion, R
i
is Net
rate of production of species, S
i
is the source term, D
i,m
is mass
diffusion coefcient.
3. The Burro series test
The Burro tests were performed by the Lawrence Livermore
National Laboratory (LLNL) at the Naval Weapons Center, China
Lake, California in 1980, and sponsored by the US DOE and the Gas
Research Institute (Ermak, Chan, Morgan, & Morris, 1982; Koopman
et al., 1982a, 1982b; Koopman, Cederwall, et al., 1982). The main
purpose was to obtain adequate samples of data under different
meteorological conditions. These experiments involved eight spills
of LNG and one of liquid nitrogen. A total of eight LNG spills onto
water surface were performed with spill volumes ranging from
24 m
3
to 39 m
3
, spill rates from 11.3 m
3
/min to 18.4 m
3
/min, wind
speeds from1.8 m/s to 9.1 m/s, and atmospheric stability conditions
being fromunstable (Class B) to slightly stable (Class E). Water pond
and instrument array (Koopman et al., 1982a, 1982b; Koopman,
Cederwall, et al., 1982) are shown in Fig. 3. There were 25 gas
sensor stations were placed in arcs in the downwind distance 57 m,
140 m, 400 m and 800 m. Meanwhile, 6 turbulence stations and 20
wind eld anemometer stations were arranged in both upwind and
downwind directions. The centreline of instrument array was
oriented at 225
C)
37.6 33.8 35.4 40.5 39.2 33.7 33.1 35.4
Atmospheric stability B B C C C D E D
Fig. 4. Computational domain and boundaries (a) and hexahedral mesh rened near
the ground surface (b).
Fig. 3. Water pond (a) and instrumentation array (b) in Burro series dispersion tests
(overhead view).
B. Sun et al. / Journal of Loss Prevention in the Process Industries 26 (2013) 117e128 120
Non-dimensional wind shear
F
m
_
KZ
U
*
__
vU
vz
_
(13)
where F
m
is calculated differently upon different atmospheric
conditions, described as ows: For unstable conditions
F
m
_
1
15Z
L
_
0:25
(14)
For stable conditions
F
m
1
4:7Z
L
(15)
For neutral conditions
F
m
1 (16)
LNG evaporates very quickly fromthe water pond area. The heat
transfer between LNG and water is due to convection ow in the
water, with very minor quantities of ice formed (Opschoor, 1980).
The published LNG vaporization rate for leaking on water surface
varies between 0.029 and 0.195 kg/(m
2
s) (Luketa-Hanlin, 2006).
Since LNG spill rates are 0.032 and 0.045 kg/(m
2
s) in B5 test and B8
test, therefore, this paper uses the spill rate as mass ow rate in
boundary conditions. The outlet boundary condition was set as
pressure outlet, since ow and energy eld were not known at this
boundary. A fully-developed ow was assumed at this boundary.
A wall boundary condition was applied to ground surface, since
no ow or energy exchanging occurs in this boundary. For side and
top surfaces, zero gradients of ow, energy and species variables
were assumed, because these surfaces are far enough from the
mass ow area.
4.2. Solution method
Prior to injection of LNG, a steady-state oweld was calculated.
Once a converged solution was obtained for steady state ow eld,
the time-dependent simulations were performed using the steady-
state as the initial condition. At time t 0 s, the injection of
natural gas was switched on and the volume concentration of
natural gas was obtained as a function of time. For the transient
vapour dispersion simulation, the residuals convergence criterion
was set as 1.0 10
4
, and the time step to 0.5 s. Approximately 20
iterations per time step were required to reach the limited residuals.
The total execution time of each transient simulation was about 5 h
on a 2.93 GHz Intel
Core
TM
i7 processor with 8.00 GB RAM.
5. Results and discussion
5.1. Crosswind dispersion
Fig. 5 shows vertical concentration proles at a downwind
distance of 57 m at different times after the LNG spill, 20 s, 70 s and
130 s, respectively. It compares the lateral dispersion distance
between experiment and simulation results. At the initial stage,
negative-buoyancy was dominated, because of heavier density of
LNG vapour than ambient air. The width of vapour cloud was wider
than the source diameter under the effect of gravity leading to lateral
spread. From 20 s to 130 s, lateral spreading distance increased
farther. It is obvious that the vapour cloud was driven by the gravity
and it moved laterally in both experiment and simulations as shown
in Fig. 5. This demonstrates that the CFD simulations were in a good
agreement with the theory of dense gas dispersion and eld
Fig. 5. FLUENT simulation of vertical volume fraction contours at 1 m height for Burro 5 test.
B. Sun et al. / Journal of Loss Prevention in the Process Industries 26 (2013) 117e128 121
experiment. In B5 test, iso-concentration contour 1% (volume
concentration) could cover the lateral distance as far as 76 min 130 s,
with the source radius of 29 m. However, the simulation predicted
a higher spreading distance of about 110 m. This was due to atmo-
spheric turbulence, and the boundary conditions applied (Koopman
et al., 1982a, 1982b; Koopman, Cederwall, et al., 1982). While, in the
eld tests, the wind speed and the wind direction were non-
uniform, in the simulation, the wind speed and its direction were
assumed to be uniform based upon the average wind speed and the
prevailing wind direction of experiments.
5.2. Horizontal dispersion
The most valuable data for model validation are considered to
be the gas concentration plume parameters (Luketa-Hanlin et al.,
2007). The four recommended values by Ermak, Chapman,
Goldwire, Gouveia, and Rodean (1988) are maximum gas concen-
tration, average ground-level plume centreline concentration,
plume half-width and plume height, all as a function of downwind
distance. Fig. 5 shows the plume height and spread in a time
dependent manner.
Fig. 6. CFD prediction of horizontal gas concentration prole at 1 m height for B5 test experiment. (a) Experiment (b) simulation (c) comparison.
B. Sun et al. / Journal of Loss Prevention in the Process Industries 26 (2013) 117e128 122
Fig. 7. CFD prediction of horizontal gas concentration prole at 1 m height for B8 test. (a) Experiment (b) simulation.
B. Sun et al. / Journal of Loss Prevention in the Process Industries 26 (2013) 117e128 123
In the B5 test, spill duration was 190 s, with a constant spill rate
11.3 m
3
/min. According to the experimental results, steady state
reached at about 90 s. The simulations achieved the steady state at
around 80 s and the dimension of the vapour cloud did not change
signicantly afterwards. Fig. 6 illustrates the horizontal iso-
concentration contour at 1 m height level in the downwind direc-
tion and compares with the experimental data with simulation
results at 20 s, 50 s, 90 s and 190 s. The experimental data are shown
in Fig. 6(a); the simulation results are in Fig. 6(b) and comparison is
shown in Fig. 6(c). As the time elapsed, the gas cloud grewbigger in
both downwind and crosswind direction. Lower volume concen-
tration in the downwind direction and larger covered area by the
dispersion cloud was also observed. The CFD simulations were able
to realistically capture these features. A very good quantitative
agreement was observed between the CFD simulations and the
experimental data for vapour cloud dimensions, especially as the
steady state was approached with an average relative error of
60.30% compared with experiment as shown in Fig. 6(c). The
gravity-induced effect can also be identied in Fig. 6 with the cloud
lateral distance wider than the source dimension. Because of the
Fig. 9. Computational domain of case study, and wind prole before LNG releasing. (a)
Geometry (b) mesh (c) wind prole.
a
b
Fig. 8. Comparison of steady-state results for experiment results, FLUENT simulation
and DEGADIS. (a) B5 test (b) B8 test.
B. Sun et al. / Journal of Loss Prevention in the Process Industries 26 (2013) 117e128 124
changes in the wind direction during the execution of experiments,
the plume shape deviates from the centreline by a small angle
(within 5