7.1. Recommendations
2D calculation can initially be made to obtain far field vertical
profiles of wind velocity and turbulence parameters to be used as
inlet boundary conditions for the ,3D domain. Velocity power
profile or log profile can be used in the log profile including the
Monin-Obukhov stability functions. For the turbulent parameters,
relations from Han et al. (2000) can be used.
The velocity values should match the experimental velocity at
10 m above surface. Symmetry boundary conditions or inflow
boundary conditions are used as boundary conditions of velocity at
the lateral and top walls.
To avoid false outside transport due to finite size of the
domain and possible discrepancies originating in the 2D calcula-
tion, all inlet parameter horizontal gradients can be set to zero
and the vertical velocity at the top of the boundary can also be set
to zero. Mathematical self-consistency between boundary conditions
and internal 3D transport can be ensured by testing the corre-
sponding flat terrain case to check for undesirable sourcesJsinks.
Concerning the two-phase flow nature of the release, the
equilibrium thermodynamics simplification and the Raoult's law at
phase change give reasonable results in terms of temperature and
concentration predictions. The mixture mass, momentum and
energy transport equations can be employed in estimating the flow
and energy field. The heat transfer mechanism within the ground
can be limited to conduction.
To define the size of the domain (excluding any disturbances)
the following boundaries are recommended: upstream)8H;
downstream>15H and vertically>6H.
In order to limit the CPU time, a non-uniform grid is preferred,
with minimum grid size applied close to the ground, obstacles and
source of the release. The grid can be expanded with an expansion
ration no more than 1.2 as it draws far from these locations. The
underground domain could be assumed to have the size of 1 H, with
10 cells (vertically) of the same size. If a simple domain is modelled,
structured mesh is a better choice. For both methods, adoption of
the grid can be used to improve the results in areas with high
gradients. cane- or two-equation turbulence modelling can be used to
produce reasonable results. It is important that the selected model
considers the local stability effects in a multiphase medium as well
as the non-isotropic effects of eddy diffusivity.
Two approaches used for turbulent flow simulation in the
complex environment are Reynolds Averaged Navier-Stokes
(RANS) and large eddy simulations (LES). This work shows that
RANS is applicable to steady state or slowly evolving flows. It is
valid for simulations involving longer time scales, and is more
appropriate for continuous atmospheric release. On the other hand,
for highly accurate simulations the LES approach resolving the most
energetic eddies is needed.7. Conclusion and recommendations
This work shows that CFD models can play an important role in
prediction of toxic gases for emergency preparedness and
response. The models usually used are approximated from the
field experiments which done in flat plane fields. Case studies
based on the FLADIS experimental data were used to develop and
evaluate CFD simulations of ammonia plume dispersion. While
atmospheric turbulence is known to be non-isotropic, this paper
demonstrates that application of k-E turbulence closure model
appears to be sufficient for simulating plume dispersion. This
simulation is even more successful if the turbulent Schmidt
number is assumed not to be constant throughout the whole
atmospheric boundary layer.
2D calculation can initially be made to obtain far field vertical
profiles of wind velocity and turbulence parameters to be used as
inlet boundary conditions for the ,3D domain. Velocity power
profile or log profile can be used in the log profile including the
Monin-Obukhov stability functions. For the turbulent parameters,
relations from Han et al. (2000) can be used.
The velocity values should match the experimental velocity at
10 m above surface. Symmetry boundary conditions or inflow
boundary conditions are used as boundary conditions of velocity at
the lateral and top walls.
To avoid false outside transport due to finite size of the
domain and possible discrepancies originating in the 2D calcula-
tion, all inlet parameter horizontal gradients can be set to zero
and the vertical velocity at the top of the boundary can also be set
to zero. Mathematical self-consistency between boundary conditions
and internal 3D transport can be ensured by testing the corre-
sponding flat terrain case to check for undesirable sourcesJsinks.
Concerning the two-phase flow nature of the release, the
equilibrium thermodynamics simplification and the Raoult's law at
phase change give reasonable results in terms of temperature and
concentration predictions. The mixture mass, momentum and
energy transport equations can be employed in estimating the flow
and energy field. The heat transfer mechanism within the ground
can be limited to conduction.
To define the size of the domain (excluding any disturbances)
the following boundaries are recommended: upstream)8H;
downstream>15H and vertically>6H.
In order to limit the CPU time, a non-uniform grid is preferred,
with minimum grid size applied close to the ground, obstacles and
source of the release. The grid can be expanded with an expansion
ration no more than 1.2 as it draws far from these locations. The
underground domain could be assumed to have the size of 1 H, with
10 cells (vertically) of the same size. If a simple domain is modelled,
structured mesh is a better choice. For both methods, adoption of
the grid can be used to improve the results in areas with high
gradients. cane- or two-equation turbulence modelling can be used to
produce reasonable results. It is important that the selected model
considers the local stability effects in a multiphase medium as well
as the non-isotropic effects of eddy diffusivity.
Two approaches used for turbulent flow simulation in the
complex environment are Reynolds Averaged Navier-Stokes
(RANS) and large eddy simulations (LES). This work shows that
RANS is applicable to steady state or slowly evolving flows. It is
valid for simulations involving longer time scales, and is more
appropriate for continuous atmospheric release. On the other hand,
for highly accurate simulations the LES approach resolving the most
energetic eddies is needed.7. Conclusion and recommendations
This work shows that CFD models can play an important role in
prediction of toxic gases for emergency preparedness and
response. The models usually used are approximated from the
field experiments which done in flat plane fields. Case studies
based on the FLADIS experimental data were used to develop and
evaluate CFD simulations of ammonia plume dispersion. While
atmospheric turbulence is known to be non-isotropic, this paper
demonstrates that application of k-E turbulence closure model
appears to be sufficient for simulating plume dispersion. This
simulation is even more successful if the turbulent Schmidt
number is assumed not to be constant throughout the whole
atmospheric boundary layer.
